SCA2015- 001

1/12

APPLICATIONS OF THE INTERCEPT METHOD TO

CORRECT STEADY-STATE RELATIVE PERMEABILITY
FOR CAPILLARY END-EFFECTS
Robin Gupta and Daniel Maloney
ExxonMobil Upstream Research Company
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
During coreflood tests in the laboratory to determine relative permeability, capillary
discontinuities at sample ends influence fluid flow and retention. When this influence or
end-effect artifact is appreciable, the laboratory data incorrectly models the reservoircondition scenario, which may result in serious errors in reservoir performance
predictions. The end effect artifact is a well-known problem with unsteady-state test data.
It is also an issue with steady-state data; it is typically handled by increasing sample
length or experiment flow rates. Increasing sample length by arranging a series of core
plugs to create a long composite is not a perfect fix to the end-effect issue, because endeffects can still exist between core plugs as well as at the end of the composite.
Increasing flow rate may not be possible for gas-liquid or gas-condensate tests in which
the experimentalist limits pressure drop because of mass transfer considerations.
The Intercept Method is a modified steady-state approach that corrects data for endeffect artifacts while conducting the test. Corrections are determined from simple
calculations based on multiple rates versus pressure drop measurements at each test
fractional flow condition. Application of the method does not depend upon a-priori
characterization of capillary pressure versus saturation. This work focuses on the
application of the Intercept Method and demonstrates lab examples of gas-liquid, gascondensate and liquid-liquid systems in which end-effect errors induce the artifact of
flow rate dependence in relative permeability measurements. However, after applying the
Intercept Method on the same data, the rate-dependent family of curves collapse into a set
of unique, flow rate independent, end-effect corrected curves. Along with end-effect
correction, the method also simultaneously corrects errors from capillary discontinuities
between core plug faces in a composite and pressure transducer zero-errors. The
workflow of the method is explained and demonstrated through several lab test examples.

INTRODUCTION
The steady-state relative permeability method consists of co-injecting fluids in steps of
increasing or decreasing fractional flow, allowing sufficient time at each step to establish
equilibrium before recording data. Pressure and saturation data at steady-state conditions
are used to generate relative permeability versus saturation curves for each fluid phase.

SCA2015- 001

2/12

Capillary end-effects (CEE) yield pressure drop and saturation artifacts that cause errors
in laboratory-measured steady-state relative permeability functions [1, 2]. Deviations of
relative permeability with rate have been observed only where boundary effects are
known to exist, and disappear as the boundary effect vanishes [2, 3]
The end-effect decreases as the length of the flooded system, rate of injection, or fluid
viscosities are increased [4, 5]. However, the end effect can occur in the individual
segments of a long composite core [1, 6] due to capillary discontinuities. Performing
high injection rate or pressure drop tests might not be possible in many cases due to phase
behavior (e.g., gas-condensate) or rock reactivity (e.g. clay-rich cores) considerations [1].
High pressure drop can yield problematic gas compressibility effects, phase behavior
changes in gas-condensate tests, and fines migration issues in a clay-rich rock. For
laboratory tests on tight rocks, pressure drops high enough to mask end-effects may be
impractical to attain without causing other experimental artifacts.
Numerical approaches to correct relative permeability data for capillary end-effects are
often complex and require additional information [7] (in-situ saturations, independent
measures of capillary pressure versus saturation functions), and might not provide unique
solutions. Another end-effect correction approach is to use internal pressure taps in
combination with in-situ saturation monitoring [3], Using internal pressure taps is
challenging for reservoir condition tests in which core is jacketed with metal foil to
prevent gas permeation through the core sleeve. Also, additional pressure taps do not
guarantee to eliminate end-effect errors between core plugs in stacked composites.
A plot of pressure drop (P) versus flow rate (Q) is frequently used as a diagnostic in
routine and special core analyses. If the data is linear but the intercept is not zero
(Figure1a), the experimentalist may interpret that the offset is the result of a transducer
zero shift or gravity head. When the data is subsequently corrected for the offset
(Figure 1b), each measurement yields the same permeability. If an offset correction is
not applied, a different permeability is calculated from each measurement (Figure 1c) and
it will seem that permeability depends upon flow rate.
We have found in multiphase steady-state lab tests that when the length over which P is
measured is the entire length of the sample, capillary end-effects cause a positive or
negative P intercept shift similar to the illustration of Figure 1a. The effect is easy to
identify when multiple rates are tested at each steady-state flow ratio or fractional flow
condition. A simple saturation correction is also easily determined and applied. This
work is a companion to a previous publication1 that focused on descriptions of
background, theory, and simulation followed by brief examples from two gas-condensate
systems. The focus of this work is from an experimentalist perspective, with brief
description of the method followed by examples of its use in gas-condensate, gas-water,
and oil-water systems using published and in-house data.

SCA2015- 001

3/12

Figure 1: Conceptual relationships between pressure drop and rate for different flow scenarios.

CONCEPT
The capillary end-effect (CEE) results from a capillary discontinuity at the core outlet
that causes accumulation of one phase relative to the other. Figure 2 is a schematic of
water saturation in a water-oil steady-state coreflood test. In the figure, saturations are
SwCEE, average water saturation in the CEE region; SwMeasured, average measured water
saturation; and SwTrue, water saturation in the non-CEE region. The CEE region, which
develops because of the capillary discontinuity at the outlet end of the sample, influences
both pressure drop and saturation measurements in a steady-state coreflood test. When a
CEE artifact is appreciable, the laboratory data incorrectly models the reservoir scenario,
which may result in erroneous reservoir performance predictions. Hence, it is important
to estimate and correct CEE-related errors in lab tests.

Figure 2: Schematic of water saturation in an oil-water steady-state coreflood1, where SwMeasured = core average
saturation, Sw CEE = average saturation inside CEE region, and Sw True= saturation outside CEE region.

The Intercept Method corrects CEE errors from both pressure and saturation
measurements for each fractional flow condition during a steady-state coreflood test.
Gupta and Maloney have described the theory in detail [1]. In a conventional steady-state
relative permeability test, phases are co-injected with increasing or decreasing fractional
flow, with a steady-state condition achieved at each fractional flow. For the Intercept
Method, in addition to the conventional approach, the steady-state condition is achieved
at multiple flow rates at each fractional flow. Trends in pressure drop and saturation
versus rate are used to correct the data set for the capillary end effect at the current
fractional flow. The corrected pressure and saturation data are subsequently used to
calculate CEE-corrected relative permeability curves using Darcys law.

SCA2015- 001

4/12

Pressure Correction
The measured pressure drop across the core (Pmeasured) is the sum of the theoretical
pressure drop (PCorrected) which would occur if there were no CEE in the core plus the
pressure drop resulting from the CEE (PCEE). Gupta and Maloney1 demonstrated
through analytical calculations that the PCEE is independent of the total flow rate (for
Stokes flow or Darcy flow) for a given fractional flow (assuming CEE region is smaller
than the core length) in a steady-state coreflood test. For the Intercept Method
application, steady-state pressure drop versus total flow rate is plotted for a given
fractional flow. The plot normally has a linear trend with a non-zero intercept (PI). This
linear trend between Pmeasured and total flow rate has been observed in lab tests discussed
later in this paper and previous work [1]. Based on the above discussion, PI is equal to
PCEE, assuming Pmeasured are corrected for gauge zero offset. If a pressure gauge offset
exists, then PI will equal PCEE plus the offset error. The offset error can be determined
at the beginning or end of the test and can be discounted from PI to get the true PCEE
value for each fractional flow. The CEE-corrected pressure drop is the difference of the
measured pressure drop across the core (Pmeasured) and the intercept (Figure 3a).
PCorrected = Pmeasured - PI

(1)

Figure 4 is a schematic of phase pressures inside a composite at a steady state condition

for a typical oil-water coreflood. CEE results in additional positive pressure drop
resistance to one phase and a reduced resistance to the other. However, the CEEcorrected pressure drop of each phase is the same. The CEE-corrected phase pressures
differ by a constant value equivalent to capillary pressure at CEE-corrected saturation
(SwTrue). Hence, the Intercept Method can be applied using an apparatus in which
pressure drop is measured for one phase instead of both phases.
Saturation Correction
Gupta and Maloney [1] demonstrated that the average saturation in the CEE region is
independent of the total flow rate for a given fraction flow. They derived that the CEEcorrected saturation (SwTrue) for a given fractional flow is the intercept of the plot of
Swavg/(1-) [y-axis] and /(1-) [x-axis], where is PI/PCorrected and Swavg is the
average saturation in the core (Figure 3b). The derivation assumes that drop
measurements across the core are corrected for gauge zero offset. The Intercept Method
corrects not only the capillary discontinuity at the core outlet end, but also the capillary
discontinuities between the plug junctions in a composite when multiple core plugs are
stacked in series (Figure 5). The additional pressure resistance from capillary
discontinuities (inside and at the end) is reflected in the intercept (PI) of the pressure
correction plot (Figure 3a). Subtracting the intercept (PI) from the lab measured
pressure drop across the composite gives the CEE-corrected pressure drop.

SCA2015- 001

5/12

(a)

(b)

Figure 3: Schematic plots used during the Intercept Method application, (a) pressure drop versus total flow rate
plot, where CEE-corrected pressure drop is the difference between the lab-measured pressure drop across the
core and the intercept, and (b) saturation plot, where intercept is the CEE-corrected saturation.

CEE
Region

Figure 4: Schematic of phase pressures with and without CEE at steady-state condition inside a composite core.

For saturation correction, the capillary discontinuities between plugs only change the
slope of the saturation correction plot (Figure 3b), but the intercept remains the same and
equal to the CEE-corrected saturation. Similarly, the Intercept Method also corrects for
the apparatus related errors, e.g., gauge zero errors, and from the pressure drop
measurements. However, the apparatus-related pressure drop is subtracted from the
intercept of pressure drop versus flowrate plot before performing saturation correction.

Figure 5: Schematic of water saturation in a composite core during an oil-water steady-state test with
capillary discontinuities between the plugs.

METHOD APPLICATION AND DISCUSSION

The Intercept Method is applicable for liquid-liquid, gas-liquid, and gas-condensate
steady-state relative permeability tests. The method provides CEE-correction for tests
performed at low rates and pressure drops. Data are corrected as measured, without
requiring in-situ saturation monitoring, internal pressure taps, or simulation. A common
manifestation of CEE is the flowrate dependence artifact in Stokes flow8 relative
permeability measurements. Chen and Wood [3] demonstrated that steady-state relative
permeability results were independent of test flow rates. The Intercept Method addresses
the flow rate dependent artifact in measured relative permeability curves.

SCA2015- 001

6/12

In a steady-state relative permeability test, application of the Intercept Method requires

attaining steady-state at multiple rates for each fraction flow. A minimum of two total
flow rates per fractional flow is required to apply the method; however, 3 to 4 total flow
rates per fractional flow is recommended to generate high-confidence pressure and
saturation correction plots (Figure 3). The time to attain additional steady-state points is
very small compared to the first total flow rate for each fractional flow point. In theory,
the additional steady-state total flow rates are instantaneous. Hence, the Intercept Method
application adds relatively small incremental test time over that from a conventional
steady-state test. Total flow rates are increased in steps to avoid hysteresis effects for the
same fractional flow. The total flow rate can be reduced concurrently with stepping to the
next fractional flow. Since the saturation change is significant between two consecutive
fractional flows, reducing total flow rate between consequent fractional flows imposes
minimal hysteresis. The Intercept method application requires CEE region to be shorter
than the core length. Below a critical total flow rate, which is typically a low value, CEE
region can encompass the entire core length. Below this critical rate, the pressure and
saturation correction plots (Figure 3) are not linear [1]. While applying the method, it is
recommended to discard the low total flow rate data that is off the linear trend [1].
Laboratory Application on Gas-Condensate System
Results of Henderson et al. [9] are evaluated by the methods of this paper. The rate,
pressure drop, and saturation data are interpreted from figures in their paper. Fluid
viscosities are assumed the same as those from Jamiolahmady et al. [10] The data sets are
from steady-state gas-condensate relative permeability measurements with fluids of 0.14
mN/m and 0.9 mN/m interfacial tension (IFT).
Figure 6 shows pressure drop versus total flow rate for the two systems for three
condensate-to-gas flow ratios (CGR). Trend lines through the data sets for each CGR are
linear, but in each case, P intercepts are non-zero. These non-zero intercepts (Figure 6)
and shifts in saturation with increasing rate at constant CGR (Figure 7) are indicative of
capillary end-effects. When intercept corrections are applied to the data sets, the net
result is a set of relative permeability curves that are rate-insensitive rather than a family
of curves that appear to be rate-sensitive. This is shown in Figure 7, which compares
curves from the original work (white and grey data points) with those after correction for
capillary end-effects (final data points). The corrected curve is close to that from the
highest total flow rate test, which corroborates with the theory [4]. For a given IFT, the
gas-condensate curve is unique (rate independent); however, change in IFT influences the
curve. As expected, the relative permeability is higher at lower IFT (Figure 7).

SCA2015- 001

7/12

Figure 6: Pressure drop versus total flow rate for 0.14 and 0.9 dynes/cm interfacial tension systems.

Figure 7: Gas and condensate relative permeability versus normalized condensate saturation.

Another condensate example that shows rate-independent relative permeability curves

after the Intercept Method correction is provided by Gupta and Maloney [1]. Figure 6 of
Ayyalasomayajula et al. [11], which is approximated as Figure 8a below, presents what is
described as typical rate versus pressure drop data measured during gas condensate
relative permeability measurements for samples from a deep marine sandstone reservoir.
The data were used to show that condensate relative permeability curves are sensitive to
rate and capillary number. In Figure 8b, stabilized pressure drops from Figure 8a are
plotted against rate. The three data points are collinear (blue dashed line) with a P
intercept of almost 19 psi. Correcting for this non-zero intercept removes the ratesensitivity that otherwise would be interpreted when the data points are considered
independently (grey dashed lines on Figure 8b).

SCA2015- 001

a) Pressure drop versus PV injected

8/12

b) Pressure drop versus rate

Figure 8: (a) Pressure drop versus pore volumes injected11, and (b) interpretation of results.

Laboratory Application on Gas-Water System

This example illustrates a laboratory application of the Intercept Method on a gas-brine
drainage steady-state relative permeability test. The composite permeability was 8.7 mD.
At the beginning of the test, the composite was completely saturated with synthetic
formation brine. During the test, gas and synthetic formation brine were co-injected
through the composite with increasing gas-to-brine flow ratios ranging from 0:1 to 1:0.
At each steady-state condition, 3 to 4 sets of total flow rates were tested while
maintaining the same gas-to-brine flow ratio. The Intercept Method was applied to
correct capillary end-effect related errors in pressure drop and saturation data, which
were later used to calculate corrected relative permeability curves.
Figure 9 shows steady-state relative permeability results with and without capillary endeffect corrections. Application of the Intercept Method resulted in up to 51 % pressure
drop correction and brine saturation correction of up to 0.06 saturation fraction units.
Because of the significant end-effect correction in this example, there is a substantial
difference between uncorrected and corrected relative permeability curves (Figure 9). In
this experiment, the coefficient of determination (R2) was greater than 0.99 for all
pressure and saturation correction plots except the last set (lowest Sw). Figure 10 shows
the pressure and saturation correction plots for gas-to-brine flow ratio of 99:1. The
experimental data follow linear trends for both the plots with about 50 % correction in
pressure drop. The last point (lowest Sw) is off-trend potentially due to wrong phase
pressure measurement or other experimental errors. Overall, this case study clearly
demonstrates that capillary end-effects can be significant for a gas-water system, and that
the Intercept Method can be applied to obtain capillary end-effect corrected relative
permeability curves.

SCA2015- 001

9/12

Figure 9: Gas-water drainage steady-state relative permeability test. Test data and Intercept Method corrected
results are shown along with % pressure drop correction.

(a)

(b)

Figure 10: Pressure drop (a) and saturation correction (b) plots using the Intercept Method at gas:brine flow
ratio of 99:1.

Another example of P intercept correction is shown by Grattoni et al. [12]. From

transient pressure decay tests in rock partially saturated with water, they found that gas
stopped flowing before pressure drop decayed to zero. They termed this non-zero P the
water-blocking pressure. From multi-rate steady-state gas injection tests with constant
water saturation, they also found that plots of pressure drop versus rate were linear but P
intercepts were positive. They suggested a saturation-dependent correction to the Darcy
equation to correct for non-zero P intercepts. Although there are several possible
reasons for their non-zero intercepts, the data closely resembles that described in this
paper.
Laboratory Application on Oil-Water System
This example illustrates a laboratory application of the Intercept Method on a data set
from an oil-brine primary drainage steady-state relative permeability test. The data are
from Virnovsky et al. [13] from a study in which steady-state drainage relative

SCA2015- 001

10/12

permeabilities were measured during repeat tests with three distinct total flow rates (low,
medium and high rates of 0.2, 0.5, and 5.0 mL/min). This data set was selected because
it is descriptive and readily available from the literature. Relative permeability test results
from low, medium, and high rate tests are shown in Figure 11 (grey data points).
Without CEE correction, there appears to be flow rate dependence in the relative
permeability curves of this test. The Intercept Method was applied. Corrected results are
shown in Figure 11 (green and blue data points). The Intercept Method successfully
collapsed the family of lab-generated curves into a unique, rate-independent set of
relative permeability curves.

Figure 11: Lab generated, and the Intercept Method corrected primary drainage relative permeability curves.

Figure 12 shows the pressure correction plots for different water fractional flows for this
test. The trends are linear with non-zero P intercept. Similarly, saturation correction
plots also were of good linearity (not shown). Pressure drops from measurements with the
highest total flow rate (5 mL/min) were significantly higher than the non-zero P
intercepts in this test. For this reason, relative permeabilities from the highest total flow
rate test (5 mL/min) are similar to the CCE-corrected curves.
The Intercept Method of this paper was developed after collectively reviewing data from
a variety of multiphase steady-state laboratory flow tests in which several sets of rates
and pressure drops were measured for each steady-state fractional flow. In each case,
non-zero P intercepts and saturation shifts were revealed, which led to the development
of methods described in this paper.
We speculate that the need for an end-effect correction to steady-state relative
permeability data has received little attention previously because most multiphase steadystate tests are not conducted with multiple rates at each fractional flow. In general, most
studies that use multiple rates for each fractional flow look for rate effects, such as a
velocity enhancement effect in gas-condensate systems or visco-inertial flow effects in
high rate gas-liquid systems. Our impression is that when multi-rate data is available, the
end-effect artifacts described herein will be evident.

SCA2015- 001

11/12

Figure 12: Pressure drop versus total flow rate at different water fractional flows for oil-water system.

Examples of this paper were drawn from the literature and in-house tests to demonstrate
that end-effects influence multiphase flow laboratory data irrespective of fluids used and
techniques applied in gaining the measurements. The methods of this paper are simple to
apply in gaining steady-state data corrected for capillary end-effects.

CONCLUSIONS
This work demonstrates the Intercept Method that can correct capillary end-effect errors
in steady-state relative permeability test data.
The method enables attainment of CEE corrected steady-state relative
permeability in systems where performing high pressure drop or high total flow
rate tests would be challenging, such as in gas-liquid systems or when testing tight
or clay-rich rocks.
The Intercept Method can be applied during a test to correct for CEE errors as
data is measured without necessitating the use of additional simulation, internal
pressure taps, or in-situ saturation monitoring effort.
The Intercept Method can be applied to liquid-liquid, gas-liquid and gascondensate steady-state relative permeability data to correct for CEE artifacts.
The CEE corrected plots are unique and independent of flow rates.
The velocity enhancement effect that has been described in the literature from
laboratory gas-condensate relative permeability measurements is likely the result
of CEE artifacts. A gas-condensate curve is unique and flowrate independent;
however, it is sensitive to IFT changes.

ACKNOWLEDGEMENTS
We thank Lisa Lun, Jim Kralik, Ted Braun (consultant), Bo Gao, David Laverick, Robert
Longoria, Daniel Berry, Larry Poore, Brad Milligan, Loan Vo, Sergio Leonardi and
Prateek Patel of ExxonMobil URC for their valuable contributions to this work.
Appreciation is also expressed to ExxonMobil Upstream Research Company for the
opportunity to present this work.

SCA2015- 001

12/12

REFERENCES
1. Gupta, R., and Maloney, D. R.: Intercept Method - A Novel Technique to Correct
Steady-State Relative Permeability Data for Capillary End-Effects, Society of
Petroleum Engineers, (November 10, 2014), doi:10.2118/171797-MS.
2. Osoba, J. S., Richardson, J. G., Kerver, J. K., Hafford, J. A., and Blair, P. M.:
Laboratory Measurements of Relative Permeability, Society of Petroleum
Engineers, (February 1, 1951), doi:10.2118/951047-G.
3. Chen, A. L, and Wood, A. C.: Rate Effects on Water-Oil Relative Permeability,
Paper SCA2001-19 presented at the 2001 Symposium of the Society of Core
Analysts, Edinburgh, Scotland, (September 19, 2001).
4. Rapoport, L. A., and Leas, W. J.: Properties of Linear Waterfloods, Society of
Petroleum Engineers, (May 1, 1953), doi:10.2118/213-G.
5. Richardson, J. G.: The Calculation of Waterflood Recovery From Steady-State
Relative Permeability Data, Society of Petroleum Engineers, (May 1, 1957),
doi:10.2118/759-G.
6. Hinkley, R. E., and Davis, L. A.: Capillary Pressure Discontinuities and End Effects
in Homogeneous Composite Cores: Effect of Flow Rate and Wettability, Society of
Petroleum Engineers, (January 1, 1986), doi:10.2118/15596-MS.
7. Qadeer, S., Dehghani, K., Ogbe, D. O., and Ostermann, R. D.: Correcting Oil/Water
Relative Permeability Data for Capillary End Effect in Displacement Experiments,
Society of Petroleum Engineers, (January 1, 1988), doi:10.2118/17423-MS.
8. American Petroleum Institute (1998, February). Recommended Practices for Core
Analysis, Recommended Practice 40, 2nd Edition, p. 107.
9. Henderson, G. D., Danesh, A., Tehrani, D. H., Al-Shaldi, S., and Peden, J. M.:
Measurement and Correlation of Gas Condensate Relative Permeability by the
Steady-State Method, Society of Petroleum Engineers, (1996, June 1),
doi:10.2118/31065-PA .
10. Jamiolahmady, M., Sohrabi, M., and Ireland, S.: Gas Condensate Relative
Permeability of Low permeability Rocks: Coupling Versus Inertia, Society of
Petroleum Engineers, (January 1, 2009), doi:10.2118/120088-MS.
11. Ayyalasomayajula, P., Silpngarmlers, N., Berroteran, J., Sheffield, J., and Kamath, J.:
Measurement of Relevant Gas Condensate Relative Permeability Data for Well
Deliverability Predictions for a Deep Marine Sandstone Reservoir, Paper SCA200333 presented at the International Symposium of the Society of Core Analysts, Pau,
France, (September 21-24, 2003).
12. Grattoni, C., Al-Hinai, S., Guise, P., and Fisher, Q.: The Role of Interstitial Water in
Hydrocarbon Flow for Tight Rocks, Paper SCA2007-14 presented at the
International Symposium of the Society of Core Analysts, Calgary, Canada,
(September 10-12, 2007).
13. Virnovsky, G. A., Vatne, K. O., Skjaeveland, S. M., and Lohne, A.: Implementation
of Multirate Technique to Measure Relative Permeabilities Accounting for Capillary
Effects, Society of Petroleum Engineers, (January 1, 1998), doi:10.2118/49321-MS.

SCA2015- 002

NUMERICAL SIMULATION OF COMBINED

CO-CURRENT/COUNTER-CURRENT SPONTANEOUS
IMBIBITION
Douglas W. Ruth1, Geoffrey Mason2, Martin A. Fern 3, smund Haugen4
Norman R. Morrow2, Rasoul Arabjamaloei1
1
Centre for Engineering Professional Practice and Engineering Education, University of
Manitoba, Canada
2
Department of Chemical and Petroleum Engineering, University of Wyoming, USA
3
Department of Physics and Technology, University of Bergen, Norway
4
Statoil , Sandslihaugen, Bergen, Norway
This paper was prepared for presentation at the International Symposium of the Society
of Core Analysts held in St.Johns, Newfoundland and Labrador, Canada, 16-21 August,
2015

ABSTRACT
This paper describes a numerical study of combined co-/counter-current spontaneous
imbibition of water into a core sample that is initially fully oil saturated with the twoends-open-free (TEO-free) boundary geometry, as given by a sample with one end in
contact with brine and the other in contact with oil. The present study uses an explicit
simulator with an upstream differencing scheme which allows flows of the oil in both
directions without blockage because of zero-saturation cells. The boundary conditions
are modelled using zero-width boundary cells with fixed saturations. The saturation at
the end face in contact with brine fixes the capillary bubble pressure (the pressure
required to produce non-wetting fluid) at this face. All the simulations used a single set
of relative permeability curves. The experimental results were matched using an
automatic search technique with the capillary pressure curve as the fitting parameter. In
all cases, tabular values of capillary pressure were used, with six points in the curves and
linear interpolation between points. Fits were achieved using the oil productions at the
two faces as the test data. It was found that all the data could be fitted using comparable
capillary pressure curves. A comparison between the results obtained using the
numerical simulator and a piston-like flow model published with the original
experimental data provides support for the assumption of piston-like flow. The paper
demonstrates both the power of numerical simulation and the urgent need for a method to
either predict or accurately measure the imbibition face saturation during spontaneous
imbibition.

INTRODUCTION
The problem of co-current and counter-current imbibition has attracted much recent
interest in the petroleum community because these are the mechanisms whereby matrix

1/12

SCA2015- 002

2/12

oil can be produced into fractures in a reservoir [1,2]. There is now a body of
experimental data for short samples undergoing these processes as well as an
approximate theory to explain the observed behaviour [3,4,5]. The present paper reports
on a numerical-simulation study of a selection of these data. The numerical simulator is a
further development of work reported previously [6,7,8,9].
The problem under consideration is shown schematically in Figure 1. A sample,
originally saturated with oil, is confined radially by a sealed surface. It is then set up
axially between two reservoirs, a reservoir that is filled with water and contacts the left
face of the sample and a reservoir that is filled with oil and contacts the right face of the
sample. The oil in the reservoir is identical to the oil that saturates the sample. It is
assumed that the pressures of the fluids in the two reservoirs remain constant and equal
throughout an experiment. It is assumed that the sample is strongly water-wet, and that
the porosity and permeability are constant and uniformly distributed.
In this
configuration, it is possible for water to spontaneously imbibe into the sample, displacing
oil in both a counter-current fashion (causing oil production at the left face of the sample)
and a co-current fashion (causing oil production at the right face of the sample). A
pressure tap is located toward the right face of the sample.
Pressure Tap

Water

Oil

Figure 1. A schematic depiction of the problem under consideration during spontaneous

imbibition with the TEO-free boundary condition.
Figure 1 shows the grid system that is used in the numerical simulation. At both ends of
the sample, zero-width grid blocks are located to facilitate the definition of boundary
conditions. The sample is then divided into a number of equal-width grid blocks. The
simulation accounts for changes in saturation in a grid block by taking the difference
between the flow rates across the boundaries. The flow rates at the boundaries are
calculated using the modified-Darcy law equations. In the absence of gravity, these
equations are:

SCA2015- 002

3/12

(1)

= +

(2)
(3)

(4)
=

In these equations, is volumetric flow rate, is viscosity, is permeability, is crosssectional area, is relative permeability, is pressure, and is axial location. With
regard to the subscripts, denotes water, denotes oil, denotes total, and denotes
capillary. Equations 1 through 4 may be combined to obtain the following equation for
the flow of oil:


(5)
=

+ +
Because the simulation depends on calculating saturation changes, it follows that
saturations will be known for each grid block at any time. However, this does not allow
direct evaluation of Equation 5 because the total volumetric flow rate is not known a
priori for spontaneous imbibition problems. This shortcoming can be overcome by
combining Equations 2 and 5 to obtain

(6)

+ +
The pressures in the water reservoir at the left face of the sample and the oil reservoir at
the right face of the sample are known (they have been assumed to have the same value
and this value can arbitrarily be set to zero). Further, for the simulation model, if the
saturation at the left face can be assumed or calculated, then the pressure in the oil at the
two faces will be known. The right-hand-side of Equation 6 is a function only of
saturation with being a constant value throughout the sample at any given time.
Knowing the saturation profile at any time, integration of Equation 6 (in the simulator
this integration is performed numerically) yields the following equation for the total
volumetric flow rate:

( )
+
=
(7)
1

( + )
Here the subscripts and refer to the right and left faces of the sample.
A critical feature of the simulation is the assumption for the saturation used in calculating
relative permeabilities for the flows across the grid-cell boundaries. In all cases, the
saturation for a fluid is assumed to be the value for the upstream cell, the so-called

SCA2015- 002

4/12

upstream-differencing assumption. That is, if the flow rate of fluid crossing a cell
boundary is left-to-right, the saturation of the left-hand cell is used. It follows that, for
the case of counter-current imbibition at a grid-block boundary, the relative permeability
of the water is calculated using the saturation of water in the left-hand cell while the
relative permeability of the oil is calculated using the saturation of oil in the right-hand
cell. The fundamental importance of this assumption can be seen by considering the lefthand sample face condition. Upstream-differencing means that the flow into the sample
is controlled by the saturation of the fluid saturating the zero-volume surface cell. As this
cell has zero-volume, it can instantly take on any saturation. For the present study, the
saturation of the left-face is assumed to be the final average saturation of the sample.
This means that there is a finite relative permeability for water to enter the sample despite
the fact that the sample is fully saturated with oil. (It is noted that a theoretical
determination of the value of saturation and bubble pressure (the pressure required to
produce non-wetting fluid) at the left-face in spontaneous imbibition problems is the most
important unresolved issue for the problem of spontaneous imbibition.)

THE PROGRAM OF STUDY

For the present study, three previously published data sets [3,4] were considered. As can
be seen in Table 1, these tests used samples with similar petrophysical properties, the
same water, but oils with a wide range of viscosities.
Table 1 The petrophysical characteristics of the three samples under study.

Sample

CHP07
CHP11
CHP25

0.467
0.466
0.458

L
(cm)
14.25
6.10
5.75

D
(cm)
3.79
3.81
3.70

Tap

(cm) (kg/m3)
11.25
1005
3.23
1005
3.43
1005

(cp)
1.09
1.09
1.09

(kg/m3)
740
870
840

(cp)
1.47
137.0
83.3

K
(md)
5.50
4.40
4.08

The published results [3,4] report the range for the permeability to be 3 to 6 mD. For the
purpose of simulation, more accurate values are needed. These values were determined
by examining the early-time data (before a front could pass the pressure tap) for the
pressure and right-face production rates. This calculation is illustrated in Figure 2 for
Sample CHP07; the permeabilities calculated in this manner are shown in Table 1.
The history matches were achieved by using a single set of relative permeability curves
and modifying the capillary pressure curve. Pressure behavior is more complex than
production; the errors between the experimental and simulated results were based on the
productions at the each face of the samples. Comparison of the predicted and the
experimental pressures provides an independent test of the quality of the analysis.

SCA2015- 002

5/12

14

Permeability, mD

12
10
8
6

4
2
0
0

200

400

600

800

1000

Time, min

Figure 2. The calculated permeability for Sample CHP07 showing the range of times (the red line) used to
determine the mean permeability.

The simulations used the relative permeabilities reported by [3]. These curves are shown
in Figure 3. The history-matching process was straightforward. The capillary pressures
were input as a 6-entry table. For the search, various stages of keeping the saturations
constant and varying the capillary pressures, and vice versa were employed. For each
assumed capillary pressure curve, a simulation was run and the errors between the
simulated productions and the values calculated from experimental data were computed.
By using an exhaustive search technique, the capillary pressure curve that led to the
minimum error was determined.
1

Relative Permeability

0.9
0.8
0.7
0.6
0.5
0.4
0.3

0.2
0.1
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Water Saturation

Figure 3. The relative permeability curves used in the simulations.

All simulations were run using a total of 50 grid blocks. Stability was controlled by
limiting saturation changes in any one grid block to less than 0.0001. No stability
problems were encountered and each test simulation ran in under 20 seconds.

RESULTS
Figure 4 shows the capillary pressure curves that lead to the best fits for the three
samples. Also shown is the initial guess for the capillary curve which was based on the

SCA2015- 002

6/12

data from [3]. All three simulation curves show similar shapes and magnitudes and do
not greatly differ from the initial curve. It is noted that the small negative region at high
water saturation are not physically reasonable for a strongly water-wet sample. They
were included here because they were present in the original initial curve. In all cases,
the simulations predicted capillary pressures in the positive region and water saturations
never exceeded the zero-capillary pressure saturation. The shapes of these curves may
appear to be unusual; however, it must be recalled that the region from 0 < < 0.3
does not actually occur in the tests and these values are really an artifact of the fitting
technique.
The quality of the curve fits for the three samples can be seen in Figures 5 through 7. In
all cases, agreement between the predicted and measured productions is very good. Of
special interest is the remarkable agreement between the predicted and measured
pressures. Recall that these data were not used in the history matches. In Figure 5, the
spike in pressure after 600 minutes is clearly matched. For both Figures 6 and 7 the bimodal nature of the pressure behavior is captured although agreement between the two
data sets is not nearly as good as that demonstrated in Figure 5.

Capillary Pressure (kPa)

200
Initial
CHP07
CHP11
CHP25

150
100

50
0
-50
0

0.1

0.2

0.3

0.4
0.5
Water Saturation

0.6

0.7

0.8

Figure 4. The capillary pressure curves that lead to the best fits of the experimental data. Also shown is
the capillary curve from [3] (labelled Initial).

In the case of Sample CHP07 (Figure 5), equal to 1.35, production from the left
hand face is minimal. The simulation was very successful in predicting both the small
counter-current left-face production and the more substantial co-current right-face
production. Production at the right-face became very slow near the end of the
experiment; this feature was also captured by the simulator. This slowing of production
corresponds to the arrival of water at the right-face. The pressure history of the
experiment is quite complex with a sharp drop at early times, then a relatively gradual
linear decline followed by a sharp peak just before the end of the imbibition process.. All
of these features are predicted by the simulator. It must be emphasized that pressure was
not used during the history-match. The agreement between experiment and simulation

SCA2015- 002

7/12

provides confidence in the simulation. Note that for Sample CH07, production dropped
off quickly after the water first arrived at the right face.
45

50
Prod Left Exp

45

Prod Left Sim

40

35

Prod Right Sim

Press Sim
30

Press Exp
30

25
25
20
20

Pressure (kPa)

35

Production (cc)

40

Prod Right Exp

15

15

10

10

0
0

100

200

300

400

500

600

700

800

Time (min)

Figure 5. A comparison of experimental and simulated results for sample CHP07. Predominantly cocurrent oil production with a viscosity ratio ( / ) equal to 1.35.

In the case of Sample CHP11 (Figure 6), equal to 125, production at the left-face
is more substantial. Again the simulation captures the relative productions very well.
Unlike the low viscosity ratio case (Sample CHP07) the production at the right end of the
sample continues after about 600 minutes, although at a lower rate. This point roughly
corresponds to a local minimum in the pressure it also corresponds to the time the water
first reaches the right-face of the sample. Although not a perfect match, the simulated
pressure history shows all the features of the experimental curve: an initial rise, a drop to
a local minimum, a second rise and an eventual tailing off. A significant difference
between these results and the results for the low viscosity ratio case is that oil production
at the right face continued, albeit at a reduced rate, after the arrival of water at this face.
In the case of Sample CHP25 (Figure 7), equal to 76, all the main features of
Sample CHP11 are demonstrated. The simulation does not match the experimental
results as well as in the first two cases but the agreement is still very good.

SCA2015- 002

8/12

14

160

12

140

Production (cc)

100

Left Prod Exp

Left Prod Sim
Right Prod Exp
Right Prod Sim
Press Exp
Press Sim

6
4

80
60

Pressure (kPa)

120

10

40

20

0
0

200

400

600

800

1000

1200

1400

0
1600

Time (min)
Figure 6. Results for sample CHP11. Sustained counter-current imbibition, but still predominantly cocurrent oil production with a viscosity ratio ( / ) equal to 125.
14

160

12

140

Production (cc)

100
8

Left Prod Exp

Left Prod Sim

80

Right Prod Exp

Right Prod Sim

60

Press Exp

Pressure (kPa)

120

10

Press Sim

40

20

200

400

600

800

1000

1200

1400

0
1600

Time (min)
Figure 7. Results for sample CHP25. Sustained counter-current imbibition, but still predominantly cocurrent oil production with a viscosity ratio ( / ) equal to 76.

SCA2015- 002

9/12

Comparison with Piston-Like Flow Theory

Figure 8 shows the calculated saturation profiles for Sample CHP25. A model which
assumes piston-like flow has been published for both counter- and co-current imbibition
[4]. This figure clearly shows that displacement takes place in an essentially frontal
manner (tied to the low-to-zero permeability to brine for < 0.3). However, with a
wedge-shaped region extending from the imbibition face to the front, this is a clear
departure from the piston model. However, the maintained similarity of the profiles is
consistent with the close match to experimental results given by a piston-like flow model.
0.6

0.5

Water Saturation

0.4

0.3

0.2

0.1

0.0
0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Scaled x-Position
Figure 8. Water saturation profiles for CHP25.

Further comparisons between the results of the present study and the piston-like flow
model are given in Table 2. These results show various levels of agreement. The
permeability results differ because the present paper uses a different averaging procedure.
However, these results are very similar. For the piston-like flow, the front is perfectly
defined. However, in the simulations, fronts are generally smeared and a single
saturation is hard to quantify. The capillary pressures at the front and at the inlet show
reasonable agreement. The results for the relative permeabilities at the front do not show
particularly good agreement. However, it must be noted that the simulation used a set of
input curves based on independent experiments while the piston-like flow uses a single
point calculation. In light of this, the agreement between the results for , in Samples
CHP11 and CHP25 is satisfactory.

SCA2015- 002

10/12

Figure 9 gives a visual depiction of the agreement between the capillary pressure curves
determined through simulation and the frontal saturation/capillary pressure points
predicted by the piston-like flow theory. The piston-like flow theory is based on a single
saturation point but the points for the two samples with published data agree very well
with the points on the capillary pressure curves found by history matching with the
simulator.
Table 2 A comparison between the results obtained by simulation and results obtained using piston-type
flow theory.

Property
CHP07: Simulation
CHP07: Piston-like flow
CHP11: Simulation
CHP11: Piston-like flow
CHP25: Simulation
CHP25: Piston-like flow

() ,
()
5.6
85
5.0
34
4.49
55
4.58
80
4.08
117
4.83
80

,
0.48
NA
0.38
0.45
0.31
0.45

,
()
96
78
127
128
166
118

,
0.07
0.2
0.2
na
0.023
na

,
0.28
na
0.56
0.77
0.64
0.51

Capillary Pressure (kPa)

200
150
Simulation (CHP11)

100

Simulation (CHP25)
50

Piston-Like Flow (CHP11)

Piston-Like Flow (CHP25)

0
-50
0

0.1

0.2

0.3

0.4
0.5
WaterSaturation
Figure 9. Comparison of Piston-like flow and Simulation Results.

0.6

0.7

0.8

Overall, the simulations support the conclusion that the piston-like flow gives a very
good first estimate of the saturation functions.

DISCUSSION
Despite the very good history matches obtained in the present study, there are a number
of issues that need further attention. The most important issue is that the present study
uses relative permeability curves obtained in an independent experiment. The results
therefore are founded on the assumption that the relative permeability curves are unique
functions of the sample, not of the process. A related concern is that the relative

SCA2015- 002

permeabilities obtained in the present study vary considerably from those obtained by the
piston-like flow theory. Another important issue is that the saturation at the inlet face of
the sample was simply set to the final average saturation of the sample. A method to
theoretically predict this saturation, and its detailed physical significance, in a unique
manner does not yet exist. A third issue is that experimental saturation profiles are not
available for the comparisons in the present work. New, related imbibition results, with
imaging of the advancing imbibing front, are now available [10] Close observation (not
detailed above) of the various test values used in obtaining the best history matches
shows that significantly different capillary pressure curves can lead to quite comparable
levels of error. Until simulations that include production/pressure/saturation profile
information are available, based on results such as those presented in [10], results from
modeling this problem must be used with caution.

CONCLUSIONS
The following conclusions can be made based on the present study:
1. The results of three water-oil, combined counter- and co-current experiments, using a
range of oil-viscosities, were successfully matched using a single set of relative
permeability curves. The resulting, fitted capillary pressure curves were very
comparable in both shape and magnitude.
2. The results of the simulations were largely consistent with the previously published
results produced using a piston-like flow theory (see [4]).
3. Because of its complex behavior, the TEO-free boundary condition offers
opportunities for evaluation several core/fluid properties from a single experiment. If
a range of fluid viscosity is used then it is possible that more variables can be inferred
[11].

ACKNOWLEGEMENTS
This work was supported in part by a grant from the Natural Sciences and Engineering
Research Council of Canada.

11/12

SCA2015- 002

REFERENCES
1. Mason, G. and N.R. Morrow, Recovery of Oil by Spontaneous Imbibition, Current
Opinions in Colloid Interface Science, (2001), Vol.6, pp.321-337.
2. Mason, G. and N.R. Morrow, Developments in Spontaneous Imbibition and Possible
Future Work, Journal of Petroleum Science and Engineering, (2013).
3. Haugen, ., M.A. Fern, G. Mason, N.R. Morrow, The Effect of Viscosity on
Relative Permeabilities Derived from Spontaneous Imbibition Tests, Transport in
Porous Media,(2015), Vol.106, pp.383-404.
4. Haugen, ., M.A. Fern, G. Mason, N.R. Morrow, Capillary Pressure and Relative
Permeability Estimated from a Single Spontaneous Imbibition Test, Journal of
Petroleum Science and Engineering, (2014), Vol.115, pp.66-77.
5. Fern, M.A., , Haugen, G. Mason, N.R. Morrow, Measurement of Core Properties
Using a New Technique Two Ends Open Spontaneous Imbibition, SCA2014-006,
International Symposium of the Society of Core Analysts, (2014), Avignon, France, 811 September.
6. Ruth, D.W., G. Mason, N.R. Morrow, A Numerical Study of the Influence of
Sample Shape on Spontaneous Imbibition, SCA2003-016, International Symposium
of the Society of Core Analysts, (2006), Pau, France, 21-24 September.
7. Ruth, D.W., N.R. Morrow, Y. Li, J.S. Buckley, A Simulation Study of Spontaneous
Imbibition, SCA2000-024, International Symposium of the Society of Core Analysts,
(2000), Abu Dhabi, UAE, 18-22 October.
8. Ruth, D.W., G. Mason, N.R. Morrow, Y. Li, The Effect of Fluid Viscosities on
Counter-Current Spontaneous Imbibition, SCA2004-011, International Symposium
of the Society of Core Analysts, (2004), Abu Dhabi, UAE, 5-9 October.
9. Li, Y., D.W. Ruth, G. Mason, N.R. Morrow, Pressures Acting in Counter-Current
Spontaneous Imbibition, Journal of Petroleum Science and Engineering, (2006),
Vol.52, No.1-4, pp.87-99.
10. Fern, M.A., Haugen ., Bratteks, B., Mason, G. and Morrow, N.R. Quick and
Affordable SCAL: Spontaneous Core Analysis. SCA2015-A055, International
Symposium of the Society of Core Analysts, (2015), St. Johns, Canada, 16-21 August.
11. Meng, Q., H. Liu, and J.Wang, Entrapment of the Non-wetting Phase during Cocurrent Spontaneous Imbibition Energy & Fuels, 2015. 29(2): p. 686-694.

12/12

SCA2015-003

1/12

QUICK AND AFFORDABLE SCAL:

SPONTANEOUS CORE ANALYSIS
M. A. Ferna, .Haugenb, B. Bratteksa,c, G. Masond and N. Morrowd
a
Dept. of Physics and Technology, University of Bergen, Norway
b
Statoil, Bergen, Norway; c The National IOR centre, Dept. of Petroleum Technology,
University of Stavanger; d Dept. of Chemical & Petroleum Engineering, University of
Wyoming, USA
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
We present a quick and affordable special core analysis method that uses spontaneous
imbibition to estimate absolute permeability, capillary pressure and relative permeability.
The method uses the two-ends-open free (TEO-free) boundary condition where one end
of the core is in brine and the other in oil. Counter-current and co-current flow are easily
measured and quantified and from them we estimate end-point relative permeability and
capillary pressure using, by most oil recovery standards, very simple analysis. The
analysis is based on an assumption of a piston-like displacement and includes the
capillary back pressure at the open face. Piston-like displacement was verified using insitu imaging by positron-emission-tomography (PET) to visualize the imbibing water
front during TEO-free imbibition.

INTRODUCTION
A widely used assumption is that the imbibition from a fracture to the rock matrix is
counter-current. The most studied experimental approach is under counter-current flow
[1] - a term we here use when oil and water flows through a cross-section in opposite
directions. The core plugs used for analysis are small and gravity forces are negligible
compared to capillary forces, in contrast to matrix blocks in oil producing fractured
reservoirs, for which the height differences in the formation will promote significant cocurrent flow by gravity forces. The term co-current flow implies that oil and water both
flow in the same direction through the rock cross-section. Water drive will generally
involve both counter-current and co-current flow in various portions of the reservoir,
depending on the magnitude of capillary to gravity forces [2]. Co-current imbibition
prevails if the matrix blocks are partially exposed to water, for example in gravity
segregated fractures, where oil will flow preferentially towards the boundary in contact
with oil [3]. Co- and counter-current imbibition may coexist during imbibition in
fractured reservoirs [4], but the flows are different: counter-current flow has lower fluid
mobility, lower mobile saturations (lower relative permeabilities), and higher viscous
interaction between the phases [5]. Consequently, counter-current laboratory

SCA2015-003

2/12

measurements on small cores may underestimate both production rate and ultimate
recovery when scaled to field conditions.
Studied Boundary Geometries
In standard experiments, core plug sample boundaries are either sealed or fully
submerged in brine at all times. Several boundary conditions have been studied
extensively in the literature; the most frequent are All-Faces-Open (AFO) and One-EndOpen (OEO). During AFO, an oil saturated core plug, normally with an initial water
phase established by forced displacement with oil, is submerged in water and the water
imbibes freely. The flows are a mix of counter-current and co-current. For OEO, the
curved surface of the cylindrical core and one end face are sealed and only one end face
is open for flow. A partially or fully oil saturated core plug is submerged in water and
spontaneous imbibition starts. For OEO only counter-current flow will occur because
water enters and oil escapes through the same open end. Other, less frequently studied,
boundary conditions include Two-Ends-Open (TEO) and Two-Ends-Closed (TEC), see
e.g. Mason et al. 2009 [6]. For the TEO boundary condition, oil production can be
measured separately for each end and often features asymmetry [7]. Using the TEO
boundary condition in horizontally positioned core plugs, the amount of oil produced
from each open end face was often unpredictably asymmetrical, even though the amount
of water imbibed from each end face was equal and symmetric with respect to the center
of the core (sometimes referred to as the no-flow boundary). Hence, oil was produced by
both co- and counter-current imbibition, even in a short system not influenced by gravity.
The experimentally observed asymmetric oil production can be explained by the need to
overcome the capillary back pressure at the outlet face of the matrix.
The Capillary Back Pressure
During imbibition, the capillary back pressure is defined as the pressure difference
between the non-wetting phase (Pnw,o) and the wetting phase (Pw,o) at the brine face of the
porous media. The wetting phase pressure at the open face is zero. For spontaneous
imbibition to progress, the imbibition pressure has to exceed the capillary back pressure
required for the production of the non-wetting phase from out of the largest pores at the
open face. The capillary back pressure may also be referred to as the bubble pressure [8].
It exists because the production mechanism of the non-wetting phase at the open end face
is similar to a drainage process. The capillary back pressure is determined by the largest
pores at the surfaces where non-wetting phase is produced as droplets [9, 10].
The asymmetric oil flows from each open end face reported by Mason et al. 2010 [7] was
shown to occur simultaneously with symmetric influx of brine from each end as
quantified by nuclear tracer imaging [11]. Hence, pore-scale heterogeneities at the rock
surface, related to production paths for oil, may have a disproportionately large effect on
the recovery and flow of oil, whereas water influx was much less affected by small
changes in the capillary back pressure This illustrates that the properties of the
fracture/matrix boundary have significant impact on the exchange of fluids between

SCA2015-003

3/12

fracture and matrix. Measurement of the simultaneous production of oil and influx of
water provides an account of water and oil flow across each open end face.

Relative Permeability and Capillary Pressure from Spontaneous Imbibition

The experimentally observed asymmetrical oil production is related to the values of the
capillary back pressures acting at the outlet faces of the core. A small difference in
capillary back pressure between the two open end faces may consume a substantial
fraction of the capillary pressure driving the flow of oil. To study the capillary back
pressure in a more controlled manner, the TEO-free boundary condition was
implemented, whereby one of the core faces is in contact with oil rather than water
Although this boundary condition has been studied previously [5, 12], an analysis of the
governing parameters, including the capillary back pressure, has only been reported
recently with the M2HF model [13, 14].
Imbibition always starts out counter-currently for the TEO-free boundary condition. How
long this mode continues depends on the mobilities of the two fluids. If the mobility of
the oil is high (i.e. low viscosity oil) then counter-current imbibition soon ceases and
imbibition becomes co-current with production varying approximately as the square root
of time. However, if the resistance is mainly in the oil phase (high oil viscosity compared
with water), the resistance to co-current flow decreases over time because the distance oil
has to flow to reach the oil-filled end face decrease, i.e. imbibition rate increases with
time.
Two-phase flow functions can be estimated during TEO spontaneous imbibition using
three simple measurements that are individually standard, but are rarely, if ever,
combined. First, the ratio between counter-current flow and co-current flow for the twoends-open free boundary condition is quantified by measuring oil production separately
from each end face. Second, the pressure development during the dynamic spontaneous
displacement process is measured along the length of the core using pressure transducers.
Thirdly, imaging of the advancing fluid fronts are obtained during the spontaneous
imbibition tests to validate the assumption that displacement is piston-like.

PROCEDURES
Positron Emission Tomography
Techniques for in situ visualization are vital tools for studying details of flow within
opaque systems such as porous rock samples. Non-invasive, non-perturbing methods
include X-ray, X-ray computed tomography (CT), nuclear tracer imaging (NTI), nuclear
magnetic resonance (NMR), nuclear tracer imaging (NTI) and magnetic resonance
imaging (MRI). Positron emission tomography (PET) is a frequently used explicit
method for clinical oncology and clinical diagnosis, where detectors register gamma ray
pairs emitted indirectly by a positron-emitting radionuclide in a tracer fluid injected into
the patient. Although primarily used as a clinical diagnostic tool, PET has also more

SCA2015-003

4/12

recently been used to visualize, and partly quantify, fluid transport in geomaterials [15,
16].
In positron decay, a positron is emitted from the nucleus accompanied by an electron to
balance atomic charge. After emission, the positron loses kinetic energy by interactions
with the surroundings, and at near-zero momentum, the positron combines with an
electron and annihilates at a finite distance outside the radioactive nucleus. The
electromagnetic radiation, in the form of two 511 keV photons emitted in opposite
directions to conserve momentum, is registered only if the photon pair is within the
coincidence window and the line-of-response (LOR) acceptance angle [17]. PET scanners
are well suited for flow studies because 511 keV photons penetrate the aluminum
confinement vessel holding the rock sample at elevated pressures. For practical purposes,
the beta decay is insensitive to temperature and pressure [18], making PET suitable for
visualization of flow in porous rocks held under pressurized in confinement vessels.
Preparing core plugs with the TEO-FREE boundary geometry
With the TEO-free condition one end is in contact with brine and the other end in contact
with oil at the same pressure. Under these circumstances brine can only enter one end of
the core, but the oil can leave from either one or both ends. Ratios of oil/brine flow can
be readily measured for each end. The experimental procedure using the TEO-free
boundary geometry was as follows:
1. Cylindrical chalk core plugs of different lengths were cut on a lathe. Sandstone
cores were drilled from quarried slabs. The core plugs surfaces were gently
washed using water to remove loose particles
2. Cores were dried at room temperature for at least 24 hours before drying them in a
heated oven until stable weight
3. Diameters and lengths were measured to obtain bulk volume
4. The circumference of a core was sealed by coating with a thin layer (less than 5
mm) of epoxy resin
5. An end-piece with a void was attached at one the end face (later filled with oil and
serving as the oil-filled end face). The end piece was machined from
polyoxymethylene (POM)
5*. With PET (or other) imaging: Attach an end piece at the inlet with void for
accumulation and transport of liquids
6. Attach a short stainless steel tube at the end piece for oil production
6*. With pressure measurements: Drill a 10 mm hole into the curved surface of the
epoxy-coated surface of the cylindrical core to allow insertion of a steel tube
(1/8in OD). Cement the outside of the steel to the rock with epoxy.
7. Record the dry weight of the core and end piece. Saturate the core and end pieces
with oil under vacuum. Check that the rock is fully saturated with oil by
calculating the average porosity of the rock from the increase in weight. (The
volume of the end piece and the pressure and production tubes are excluded).
Core permeabilities were not measured: however, values of 3-8mD (chalk) and
1200mD (sandstone) have previously been measured.

SCA2015-003

5/12

8. To reduce induction time effects, start the imbibition experiments as soon as

possible after filling with oil, preferably on the same day. Submerge the oilsaturated assembly in brine and record the oil produced in inverted graded funnels
set at each end of the core.
Spontaneous imbibition in PET/CT
The epoxy-coated oil-filled (decane o=0.74) core was placed within a rectangular
Plexiglas box on the standard patient table on the combined PET/CT scanner. Preinstalled software and lasers were utilized for accurate placement of the rock system with
respect to the CT gantry and the PET detector array. The rock sample was placed in the
center of the bore (diameter 700 mm) to optimize CT and PET images. The rock system
was stationary within PET detector array, with an axial field of view of 169 mm.
Temporal resolution was determined in post processing for optimal visualization of the
displacement process. Changes in pressure and volume were monitored and recorded.
Radioactive 1wt% NaCl brine with water-soluble 18F-FDG (w=1.09 cP) was circulated
through the inlet end piece using a constant rate injection pump to start the imbibition
process, see Figure 1. External pressure transducers measured pressure changes during
imbibition, and volumetric oil measurements were registered manually over time. Fluid
saturations were found from the linear relationship between the number of disintegrations
and the saturation of the labelled fluid.

Figure 1. Experimental setup for spontaneous imbibition with PET imaging using the TEO-free boundary
condition. The CT image shows a horizontal chalk plug with three pressure tappings and POM end pieces
for water inflow and counter-current and co-current oil production. Volumetric measurements were
performed outside the PET ring, and a cylinder pump ensured continuous cycling of water into the end
piece for the duration of the test [19].

Calculating Relative Permeability and Capillary Pressure

The M2HF model based on piston-like displacement was previously developed for pure
1D counter-current or co-current spontaneous imbibition [13, 14]. During co-current
imbibition, the expected position of the imbibition front at time t can be found from

SCA2015-003

6/12

= + 2 + > 0

(1)

where is the distance of the imbibition front (cm), L is the sample length (cm) and t is
the imbibition time (min).
The constant D is defined as
=

( )( )1

(2)

where is the wetting phase viscosity (mPa s), is the non-wetting phase viscosity
(mPa s), is wetting phase relative permeability behind the imbibition front and
is the non-wetting relative permeability ahead of the imbibition front.
The constant E is defined as
2

= 1002,

(3)

where is the porosity, K is the absolute permeability, Swf is the average wetting phase
saturation behind the imbibition front, and Pc,f is the capillary pressure at the imbibition
front (kPa). Equations 13 can be used to calculate the distance of advance of the
imbibition front if imaging is not available. In this work, local saturation measurements
are measured directly with PET. The samples do not contain an initial water saturation as
is usual in practice. The (krnw) ahead of the imbibition front is unity (1). The average
water saturation (Swf) behind the imbibition front can be obtained from the amount of oil
production divided by the pore volume behind the imbibition front. Hence, there are only
two unknown parameters in Eq. 2 and 3; the relative permeability to brine behind the
imbibition front (krw) and the capillary pressure at the imbibition front (Pc,f).
The capillary back pressure at the open face was obtained from the oil production curve
by recording the pressure at which counter-current production stops. This occurs when
the non-wetting phase pressure at the front equals the capillary backpressure at the open
face and there is no differential pressure to drive the non-wetting phase towards the inlet.
The pressure measured with a transducer can be scaled to the pressure at the front and
thus to the capillary back pressure by using simple geometry, assuming linear pressure
drops,

, = , = ,
(4)

where Pnw,f is the non-wetting pressure at the imbibition front, Pc,o is the capillary back
pressure at the open face, Pnw,f is measured (oil) pressure in the transducer, and Xt the
position of the transducer.

RESULTS
Figure 2 shows results using the TEO-free boundary conditions for two chalk plugs
CHP26 and CHP27. The low viscosity mineral oil used (n-decane) promotes co-current

SCA2015-003

7/12

imbibition with this boundary geometry because the dominant resistance is in the aqueous
phase.

Figure 2. Spontaneous imbibition using the TEO-free boundary condition in core plugs CHP26 (top) and
CHP27 (bottom). The counter-current inlet (end face in contact with brine) oil production (solid red line
with crosses) was less compared to the co-current outlet (end face in contact with oil) oil production (solid
blue line with open circles). Counter-current oil production was 10% for CHP26 and 11.2% for CHP27.
Imbibition time increases with core length. Total production (CHP26: RF=61 %OOIP; CHP27: RF=58
%OOIP), with a uniform distribution. The locations of pressure ports are included for both tests. The
recorded pressures are used to calculate the capillary back pressure.

Access to local water saturation during the imbibition process using PET/CT imaging
allowed us to test the assumption of piston-like displacement that forms the basis for
calculating relative permeability and capillary pressure from these tests. The front
position can be found (without imaging) from the amount of brine imbibed (equal to the
total oil produced) divided by the brine imbibed when the front reaches the end of the
core. Figure 3 (left) shows these calculations for CHP26 and CHP27 versus square root of
time compared to the tracked imbibition front using PET. These two independent
measurements fell into alignment by the time counter-current imbibition ceases at the

SCA2015-003

8/12

inlet end. These results verify that the imbibition front is piston-like and co-current.
Figure 3 (right) shows the calculated front position Xf given by Eq. 1 plotted against the
experimentally measured Xf using PET. A value of D can be found that gives a straight
line. From D, a value of E can be determined that gives a gradient of unity. Once the
values D and E are determined, values of krw and , are found using Eq. 2 and Eq. 3.
The , is calculated using Eq. 4. Table 1 lists relative permeabilities and capillary
pressures.

Figure 3. Left: Front position based on piston-like displacement (red open circles: CHP26; black crosses:
CHP27) and visualized front position (solid red line: CHP26; solid black line: CHP27) vs. time. Right:
Calculated front position (Xf) using Eq. 1 vs. measured front position using PET for core plugs CHP26
(top) and CHP27 (bottom).
Table 1. Core properties, calculated relative permeability and capillary pressures

CHP26
CHP27

L [cm]
8.88
11.8

D [cm]
5.08
5.08

Por [frac]
0.455
0.468

krw
Pc,f [kPa]
0.33
64.2
0.31
78.2

Pc,o [kPa]
21.8, 22.2
33.7, 32.7, 28.1

Pc,oPc,f
0.34
0.40

Figure 4 shows the development of imbibing water (the labelled phase) during
spontaneous imbibition with the TEO-free boundary geometry using PET/CT. The core is
positioned horizontally during the experiment. The PET signal only measures the water
and is overlaying a CT image to align with system geometries. The locations of the three
pressure ports can be seen on the CT images, and appear brighter than surrounding rock
due to higher density. Shadows on the CT image below the pressure ports are artefacts
and does not represent areas of low CT numbers. The porosity was uniformly distributed
along the length of the core. Water was injected into the void of the inlet end piece and

SCA2015-003

9/12

appears bright yellow (high intensity) on the images. Each image is an accumulation of
PET signal during 10 min of imbibition and the advancing imbibition front may look
slightly dispersed on the images. Figure 4 shows the position of the front at 4 selected
times, and the advancing front progress as a piston.

Figure 4. Qualitative PET visualization (color) overlaying CT (gray scale) of spontaneous imbibition with
the TEO-free boundary condition for core plug CHP27.The location of the front is shown at four times
(TOP: 40 min (left), 110min (right); BOTTOM: 200min (left), 300min (right)). Each image shows four
orientations: cross-section (XY) at imbibition from (upper right); 2D slice (YZ) with pressure ports (upper
left); bird eye 2D slice (XZ); and 3D image (XYZ). Colors on 2D images: high intensity (yellow) area is
brine-filled void in inlet end piece; red is advancing water in core. Colors on 3D image: darker green is
brine in void spacer and light green is the advancing water in the core. The water advances as a near-perfect
piston from the inlet towards the outlet.

Figure 5 shows quantitative one dimensional saturation profiles for selected times during
the imbibition process for core plug CHP27. The calculations are based on the PET signal
and have been reduced from 3D to 1D by averaging. No additional smoothing has been
applied. An assumption of uniform water distribution at the end of the test (Sw=0.58) is
used to scale the normalized PET signal. We observe that this assumption was valid as
the last signal intensity at the end was (within experimental uncertainty) a straight line.
The front was not a perfect piston-like displacement, but the imbibition front was selfsimilar and maintained its shape throughout the test. The profiles slightly dispersed

SCA2015-003

10/12

because they are based on 10 min time-averaged PET images rather than snapshots of the
water-distribution.

Figure 5. One dimensional water saturation profiles during TEO-free spontaneous imbibition in core plug
CHP27. A sharp front advances through the core during imbibition; its shape is maintained during the
displacement process.

The results in Figure 5 demonstrate the use of PET as an imaging tool for oil-water
displacement in porous media with quantified local fluid saturations. Because PET relies
on direct measurement of the labelled phase (water in this work) PET is superior
compared to CT for determining front progression during flow injection tests and for
determining spatial fluid saturation. We also learned that PET imaging has a great
potential for study of low porosity samples because of the explicit fluid measurements
provide high image quality and high signal to noise ratio (SNR) independent of low pore
volume.

DISCUSSION
Imbibition is usually modelled by numerical solution of a set of differential equations.
The solution requires functions of capillary pressure, and the relative permeabilities to
both flowing phases, which are not usually known. If they are known, the values are
commonly determined from separate experiments which may not be representative of
flow during spontaneous imbibition [5].
A relatively straightforward mathematical analysis of co-current imbibition predicts that
the volume of oil produced will depend on the square root of time. This is useful for
correlation purposes. However, this behavior only allows determination of a single
parameter which involves a group of core and liquid properties. The experimentally
observed oil flows with the TEO-free boundary geometry, in this work and elsewhere,
can only be explained by the influence of the capillary back pressure at the brine end of
the core. The observation that counter-current oil production stops entirely after just 10%
of total production implies that the driving pressure (the capillary pressure at the front) is

SCA2015-003

11/12

lower than some threshold. We describe this threshold as the capillary back pressure. The
back pressure resists oil production because oil is produced as bubbles on the open end
face during imbibition. This pressure was estimated using pressure measurements in
combination with the position of the imbibing front when the counter-current production
stopped. Although small in absolute value, the capillary back pressure was approximately
1/3 of the driving pressure.
Recently, the M2HF-model [13, 14], initially developed in high-capillary chalk, was
demonstrated in unconsolidated sand packs by Meng et al. [20]. By varying the liquid
viscosity ratios, the water saturation behind the front was varied to establish a wider
range of saturation on the relative permeability curve. They installed a disc with smaller
pore sizes at the inlet end face in contact with water to force pure co-current imbibition.
This demonstrates that the method is applicable in media for which acting capillary
pressures are low. We expect the methodology to work equally well, as indicated by
initial tests, in low-permeability/porosity rocks such as tight gas sands where standard
core analysis is tedious and often inaccurate.

CONCLUSIONS
Using the M2HF-model, the oil production measured separately from each end face
during spontaneous imbibition using the TEO free boundary geometry can be used to
estimate the relative permeability to brine behind the front and the effective capillary
driving pressure at the front.
The capillary back pressure at the open face is not zero, as is commonly assumed. The
capillary pressure is an essential factor in analysis of imbibition data.
The assumption of piston-like displacement during spontaneous imbibition (without
initial water) was confirmed with PET imaging. We demonstrate the use of PET as a
novel methodology to study flow in geomaterials, for explicit flow visualization.

REFERENCES
1. Morrow, N.R. and G. Mason, Recovery of oil by spontaneous imbibition. Current
Opinion in Colloid & Interface Science, 2001. 6(4): p. 321-337.
2. Bourbiaux, B.J., Understanding the Oil Recovery Challenge of Water Drive Fractured
Reservoirs, in International Petroleum Technology Conference, 2009. International
Petroleum Technology Conference: Doha, Qatar.
3. Pooladi-Darvish, M. and A. Firoozabadi, Cocurrent and Countercurrent Imbibition in
a Water-Wet Matrix Block. SPE Journal, 2000. 5(1): p. 3-11.
4. Karpyn, Z.T., P.M. Halleck, and A.S. Grader, An experimental study of spontaneous
imbibition in fractured sandstone with contrasting sedimentary layers. Journal of
Petroleum Science and Engineering, 2009. 67: p. 48-65.
5. Bourbiaux, B.J. and F.J. Kalaydjian, Experimental study of cocurrent and
countercurrent flows in natural porous media. SPE Reservoir Eval. & Eng., 1990. 5: p.
361 368.

SCA2015-003

12/12

6. Mason, G., H. Fischer, N.R. Morrow, D.W. Ruth, and S. Wo, Effect of sample shape
on counter-current spontaneous imbibition production vs time curves. Journal of
Petroleum Science and Engineering, 2009. 66(3-4): p. 83-97.
7. Mason, G., H. Fischer, N.R. Morrow, E. Johannesen, A. Haugen, A. Graue, and M.A.
Fern, Oil Production by Spontaneous Imbibition from Sandstone and Chalk
Cylindrical Cores with Two Ends Open. Energy & Fuels, 2010. 24: p. 1164-1169.
8. Mason, G. and N.R. Morrow, Developments in spontaneous imbibition and
possibilities for future work. Journal of Petroleum Science and Engineering, 2013.
110(0): p. 268-293.
9. Li, Y., D. Ruth, G. Mason, and N.R. Morrow, Pressures acting in counter-current
spontaneous imbibition. Journal of Petroleum Science and Engineering, 2006. 52(1-4):
p. 87-99.
10. Unsal, E., G. Mason, N.R. Morrow, and D.W. Ruth, Bubble Snap-off and CapillaryBack Pressure during Counter-Current Spontaneous Imbibition into Model Pores.
Langmuir, 2009. 25(6): p. 3387-3395.
11. Graue, A. and M.A. Fern, Water mixing during spontaneous imbibition at different
boundary and wettability conditions. Journal of Petroleum Science and Engineering,
2011. 78(3-4): p. 586-595.
12. Dong, M., F.a.L. Dullien, and J. Zhou, Characterization of Waterflood Saturation
Profile Histories by the Complete Capillary Number. Transport in Porous Media,
1998. 31(2): p. 213-237.
13. Haugen, A., M.A. Fern, G. Mason, and N.R. Morrow, The Effect of Viscosity on
Relative Permeabilities Derived from Spontaneous Imbibition Tests. Transport in
Porous Media, 2015. 106(2): p. 383-404.
14. Haugen, ., M.A. Fern, G. Mason, and N.R. Morrow, Capillary pressure and
relative permeability estimated from a single spontaneous imbibition test. Journal of
Petroleum Science and Engineering, 2014. 115(0): p. 66-77.
15. Dechsiri, C., A. Ghione, F. van de Wiel, H.G. Dehling, A.M.J. Paans, and A.C.
Hoffmann, Positron emission tomography applied to fluidization engineering. Canadian
Journal of Chemical Engineering, 2005. 83(1): p. 88-96.
16. Kulenkampff', J., M. Gruundig, M. Richter, and F. Enzmann, Evaluation of positronemission-tomography for visualisation of migration processes in geomaterials. Physics
and Chemistry of the Earth, 2008. 33(14-16): p. 937-942.
17. Bailey, D.L., D.W. Townsend, P.E. Valk, and M.N. Maisey, eds. Positron Emission
Tomography. Basic Sciences. 2005, Springer.
18. Emery, G.T., Perturbation of Nuclear Decay Rates. Annual Review of Nuclear
Science, 1972. 22(1): p. 165-202.
19. Fern, M.A., . Haugen, G. Mason, and N.R. Morrow. Spontaneous Imbibition
Revisited: A New Method to Determine Kr and Pc by Inclusion of the Capillary
Backpressure at IOR 2015 18thEuropean Symposium on Improved Oil Recovery,
Dresden, Germany, 14-16 April, 2015.
20. Meng, Q., H. Liu, and J. Wang, Entrapment of the Non-wetting Phase during Cocurrent Spontaneous Imbibition. Energy & Fuels, 2015. 29(2): p. 686-694.

SCA2015-004

1/12

ION TUNING WATERFLOODING IN LOW

PERMEABILITY SANDSTONE:
COREFLOODING EXPERIMENTS AND
INTERPRETATION BY THERMODYNAMICS AND
SIMULATION
Quan Xie; Desheng Ma; Qingjie Liu; Weifeng Lv
Research Institute of Petroleum Exploration and Development of CNPC
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
In recent years, ions tuning water flooding (ITWF) has been a promising technique to
recover oil in sandstone reservoirs. In view of research results for the last decade, it is
acknowledged that substantial oil recovery beyond conventional waterflooding from
sandstone is wettability alteration. However, the major contributor to wettability
alteration is still uncertain. Therefore, this paper investigates this major mechanism and
shows how it is due to involved in the process of IOR. Rock and oil surface chemistry
were tested to explain the influence of zeta potential on the disjoining pressure. Coreflood
experiments with permeability less than 1mD were performed with two different brines.
Moreover, the impact of different wettabilities ranging from strong water- to slight water
-wet on ITWF recovery was investigated with combination of thermodynamic theory and
corefloods experiments. Relative permeability curves were obtained by history matching
the corefloods experiments for both oil-wet and water-wet cores with consideration of
salinity effect. Thermodynamics of wettability by ion tuning waterflooding was analyzed
to characterize the surface forces between the surfaces of oil/water and water/rock.
Zeta potential results showed that decreasing divalent cations and salinity makes the
electrical charges at both oil/brine and brine/rock interfaces become strongly negative,
which results in elevation of the repulsive forces between oil and rock, and as a result the
rock turns more water-wet, which was confirmed by thermodynamics characterization.
Coreflooding experiments showed that a high potential in oil-wet reservoirs can be
achieved by ion tuning waterflooding due to the double layer expansion. The relative
permeability curves obtained by history matching showed that ITW improves oil
recovery by accelerating oil production (relative perm changes) and reducing residual oil
saturation in oil-wet rock but not in water-wet rock. Thermodynamics of wettability
analysis indicated that the mechanism of ion tuning waterflooding can be interpreted by
disjoining pressure calculation. These findings can help in composition design of ion
tuning water to maintain higher potential to recover oil in oil field.

SCA2015-004

2/12

INTRODUCTION
Waterflooding technology has been the most successful approach to improve oil recovery.
The key point to reach this success of waterflooding is that the differential pressure can
be formed by the water injection which is necessary to displace oil out of formation. And
also, waterflooding involves much lower cost investment and convenient operation.
However, it was found that water chemistry and salinity level have a significant influence
on oil recovery from the experiment in the laboratory and field trials [1, 2]. In recent
years, several mechanisms were proposed to account how the ions tuning waterflood to
recover additional oil. (1) Fines migration and clays swelling caused by ions tuning
waterflood are the main mechanism of improved oil recovery [3, 4]. (2)Multi-component
ionic exchange between the rock minerals and the injected brine was proposed to be as
the major mechanism to enhance oil recovery [5, 6] . (3) Expansion of the double layer to
be as the dominant mechanism of oil recovery improvement[7]. The general agreement
among researchers is that low salinity waterflooding cause reservoirs become more
water-wet [1]. The general agreement among researchers is that low salinity
waterflooding cause reservoirs become more water-wet. The main objectives of this work
are (1) to examine the performance of LSW in sandstone rocks with different wettability,
and investigate the role of salinity and brine composition on the performance of LSW on
the secondary and tertiary recovery modes. (2) To derive the relative permeability curves
from corefloods experiments to better understand the LSW performance in different
wettability reservoirs. (3)Moreover, in order to have deeper understanding the mechanism
of the low salinity EOR-effect, the thermodynamics of wettability was analyzed with
investigation of surface chemistry of interfaces of oil/water and water/rock.

EXPERIMENTAL
Fluids
In order to investigate the impact of the rock wettability on the low salinity EOR-effect,
the crude oil from the Changqing Oil Field was used in zeta potential test and corefloods
experiments. The ingredients of the experimental oil from Changqing Oilfield were
shown in Table 1. The experimental oil was rich in aromatic hydrocarbon and asphaltic
bitumen, the main source of the carboxy and amino groups. The density of oil sample was
0.81 g/cm3 with viscosity at 9.0 cp at temperature of 65oC.
Two different brines were used in the corefloods and zeta potential tests; the one was
synthetic brine according to the formation brine formula. Another is the low salinity brine,
which was diluted 10 times by adding deionized water. The composition of the formation
brine and LSW were given in Table 2. The total salinity of the brine water was
57114mg/L with the concentration of Ca2+ and Mg2+ at 2460mg/L and 317mg/L,
respectively.
Mineralogy of Reservoir Cores

SCA2015-004

3/12

The core plugs for coreflood experiments were extracted from the Chang Qing oilfield
and outcrop in Chang Qing. The content of clays was analyzed by X-ray test to unveil the
importance of the clays on the low salinity EOR-effect in low permeability reservoirs, as
shown in Table 3. The reservoir cores were rich in clays, which was more than 23% in
total. 92.8% quartz and 4.2% clay content were observed from the outcrop core plugs.
The composition of the outcrop indicated that it could be water-wet because of high
content of quartz in the rock.
Table 1 Ingredients of the oil sample from Changqing Oilfield
Ingredients

Saturated
hydrocarbon

Aromatic
hydrocarbon

Non-hydrocarbon

Asphaltic
bitumen

wt%

65.05

23.30

6.68

4.97

Table 2 Composition of the formation brine and low salinity water (LSW)
Ingredientsmg/l)

Brines

K +Na
19249
1924.9

Formation brine/HSW
Ion tuning water/ITW

2+

2+

Ca
2460
246

Mg
317
31.7

HCO
308
30.8

3-

Cl
34780
3478

TDS
(mg/l)
57114
5711.4

Table 3 Mineralogy of the experimental core plugs

Relative content of
clay minerals%

Mineral types and content%

Sample

Outcrop

Total clay
minerals
%

Mix

quartz

potassium
feldspar

plagioclase

calcite

25

25

45

92.8

1.4

1.6

4.2

20.9

0.4

23.6

Reservoir
32
12 28 28
43.2
11.9
core
Note: I-illite, K-Kaolinite, C-Chlorite, Mix-Illite/Smectite

Zeta Potential
Zeta potential technique was applied to understand the relation between electric double
layer and wettability regarding the charges at interfaces of oil/brine and solids/brine.
Rock wettability is closely related to the thickness of water film between rock surface and
crude oil, which depends on the electrical double layer repulsion and Van der Waals force.
Wettability of the rock will be determined by the stability of the water film which is
bounded by the interfaces of oil/brine and solid/brine [8, 9]. An unstable water film, or
thin water film, may cause the wettability preferential to be oil-wet. Therefore, injection
water with different ions types and concentration will trigger the alternation of the surface
charges at both interfaces oil/brine and solids/brine. Then, oil film may detach from the
pore wall by increase of electrical double layer repulsion and the oil recovery will be
improved.

SCA2015-004

4/12

In this study, the protocol of zeta potential tests was shown in a published paper [10].
Results of zeta potential of oil/brine and solid/brine interfaces were given to investigate
the influence of the double layer expansion on low salinity EOR-effect in oil-wet and
water-wet core plugs, as shown in Figure 1. The pHs of the two different brines with
adding powders of rocks and crude oil were measured after equilibrium, as shown in
Figure 2, due to the zeta potential magnitude varying with pH value.
Figure 1 shows that LSW resulted in stronger negative charges at the surface of reservoir
rock/brine, outcrop rock/brine and crude oil/brine. This is consistent with most of the zeta
potential test for low salinity waterflooding. However, LSW increased the negative
surface charges at interface of outcrop rock/brine, compared to HSW, This result shows
that the mineral and clay content of the rock is closely related to the zeta potential for
certain brine. Basically, the larger the clay content, the more sensitive the zeta potential to
the brine salinity, regarding to the variation of the brine salinity. Zeta potential of
reservoir rock with LSW and HSW also proves that salinity strongly impact on the
electric surface charge at the oil/brine and rock/brine interfaces [7].
Figure 2 shows that pH of outcrop rock/HSW and outcrop rock/LSW was about 7.0 and
7.5, respectively. However, the pH of reservoir rock/HSW and reservoir rock/LSW was
about 6.2 and 8.0, respectively. The pH increase in the LSW compared with HSW was
interpreted by the chemical equilibrium given as below moving to the right because of the
low salinity water invasion [11].
NaAlSi3O8 +H2O <> HAlSi3O8 + Na+ + OHClay-Ca2+ +H2O <> Clay-H+ + Ca2+ +OHR-COOH <> R-COO- + H+

(1)
(2)
(3)

Chemical Equation (1) and (2) also indicate that the higher pH increase could be
observed if the rock is rich in clays and NaAlSi3O8, which is consistent with the rock
mineralogy, as shown in Table 3.

Figure 1 zeta potential results of oil/brine and

solid/brine interfaces

Figure 2 pHs of the two different brines with

adding powers of rocks and crude oil

SCA2015-004

5/12

Wettability Measurement
The Amott test is one of the most widely used empirical wettability measurements for
reservoir cores in petroleum engineering. Wettabilities of preserved core plugs were
measured with the improved Amott water index method which is slightly different from
the conventional procedure. This Amott water index approach combines spontaneous
imbibition and forced imbibition processes. The experimental procedures adopted in this
paper were introduced in a literature [12]. Crude oil from Changqing oil field was used to
desaturate the core plugs, then they were put into an oven at 65oC for aging four weeks to
restore the wettability.
The magnitude of Iw is from 0 to 1. The higher Iw , the more water-wet. The results of
wettability test were given in Table 4, which shows that the outcrop core plug OC-1# was
strongly water-wet and reservoir rock RC-1# was slightly water-wet. The two types of the
core plugs with different wettability were used to conduct corefloods experiments for
investigation of LSW performance, which was discussed later in this study.

Core
sample

OC-1#
RC-1#

Table 4 Physical properties of cores and the Amott test results

Core properties
Amott test
Diameter
(cm)

Length
(cm)

2.5
2.5

7.15
7.2

Pore
volume
(mL)
4.60
4.45

Porosity
(%)

Swi
(%)

Sws
(%)

Sor
(%)

Iw

13.1
12.6

26.8
31.1

60.2
41.2

64.9
54.8

0.88
0.43

Swi, initial water saturation; Sws, water saturation after spontaneous imbibition of water; Sor,
residual oil saturation after forced displacement.

COREFLOODING EXPERIMENTS AND SIMULATIONS

Coreflooding Experiments Protocols
In order to investigate the influence of wettability on low salinity water EOR effect,
corefloods experiments were executed by using two reservoir cores and two cores from
the outcrop. For each type of wettability, low salinity water was used under secondary
and tertiary mode with the core plugs, which were initially saturated by formation brine.
Due to the small changes in saturation followed by low salinity waterflooding under
tertiary mode, unsteady state (USS) under tertiary and secondary by using low salinity
water flooding were history matched to derive the low salinity water relative permeability
curves [13]. All of the coreflood experiments were conducted at 65oC. The brine
permeability of the experimental cores was lower than 1 mD, as shown in Table 5 and
Table 6. The injection rate was 0.025ml/min. The procedures of corefloods experiments
and set-up were selected from literature [10].
Methodology of History Matching in Core Scale
A one-dimensional homogeneous permeability core model was established to simulate
the characteristics of oil recovery by forced imbibition with the finite difference simulator
ECLIPSE 100. There were 22 equal-sized grid blocks in the core model. The first grid

SCA2015-004

6/12

which was located at the upstream of the core was saturated with 100% formation brine
with 100% porosity, but 1000 D permeability was assigned to simulate the experimental
injection. The last grid at the outlet of the core was saturated 100% oil with 0.001%
porosity and 1000 D permeability since at beginning of the forced imbibition the outlet of
the core was filled with 100% experimental oil. Porosity and permeability of each of the
cells except of inlet and outlet cells was assumed to be equal. Porosity was obtained by
Helium Porsimeter (PHI-220) under room temperature and permeability was measured
through coreflooding experiment by injection of formation brine with saturated core
plugs. The summary of composite core model for experiments and simulation model were
given in Table 5.
According to the waterflooding simulation through ECLIPSE 100, the significant
parameters dominating oil recovery and differential pressure by forced imbibition are the
capillary pressure and relative permeability curves. The functional forms in Eqs. (4)-(6)
are
S S wi
k rw k w

1 Sor S wi

nrw

rw

Sw Swi
kro k ro 1

1 Sor S wi

(4)
nr 0

P
c

S S wi
J 1 w

k
S w0 S wi

(5)
n pc

(6)

Where krw is the water relative permeability and kro is the oil relative permeability. Pc is
the capillary pressure, and S is the phase saturation. The subscripts w, wi and wo represent
water, initial water, residual oil at water saturation where the capillary pressure is equal to
zero, respectively. The superscript denotes the end-point.
Table 5 Summary of composite core model for four experiments
Properties
Outcrop
Samples
OC-2#
OC-3#
Flooding sequences
HS + LS
LS
Porosity
0.136
0.140
Kw (mD)
0.331
0.340
Pore volume (cm^3)
4.770
4.810
Length (cm)
7.115
6.990
Diameter (cm)
2.504
2.504
Bulk volume (cm^3)
35.03
34.37
Cross sectional area (cm^2)
4.924
4.917
Grid block dimensions
20 1 1
20 1 1
Width (cm)
0.35575
0.34955
Height (cm)
2.219009
2.217431

Reservoir cores
RC-2#
RC-3#
HS + LS
LS
0.129
0.123
0.92
0.82
4.3665
4.2029
6.867
6.927
2.5059
2.5068
33.849
34.170
4.9292
4.9328
20 1 1
20 1 1
0.34335
0.34635
2.2202
2.2210

SCA2015-004

Table 6 Summary of composite core model for four experiments

Properties
Outcrop
Samples
OC-2#
OC-3#
Flooding sequences
HS + LS
LS
no
2.2
2.1
nw
6.2
6.3
Krw(Sorw)
0.09
0.09
Kro(Swc)
1.0
1.0
Sorw
0.363(HS)/0.331 (LS)
0.339
Swc
0.268
0.291

7/12

Reservoir cores
RC-2#
RC-3#
HS + LS
LS
5.2
3.2
3.1
4.8
0.35
0.2
1.0
1.0
0.412
0.382
0.322
0.315

Effect of Low Salinity Waterflooding on Reservoir Cores

Low salinity EOR-effect has been closely related to the change of wettability of
reservoirs according to studies during the past more than a decade. However, the
influence of the wettability in low permeability reservoirs on the low salinity
waterflooding is rarely reported with combination of corefloods experiments, core-scale
history matching. Therefore, this paper unveils the deeper understanding of low salinity
EOR-effect through corefloods experiments on various wettability cores, core-scale
history matching and thermodynamics analysis, which will be discussed at the last part of
this study. Two reservoir cores, either flooded by low salinity water under secondary or
tertiary mode, were implemented to conduct coreflooding. Both oil recovery and
differential pressure by high and low salinity waterflooding were successfully history
matched to derive the relative permeability curves, as shown in Figure 3, Figure 4 and
Figure 5.
Figure 3 and Figure 4 show that low salinity waterflooding can improve oil recovery
about 10% of OOIP under both secondary and tertiary mode, compared with high salinity
waterflooding. Both of the Figures also indicate that the low salinity water can accelerate
the oil production rate and reduce the residual oil saturation, with comparison of the high
salinity waterflooding. This coreflood phenomenon proves that the great potential of low
salinity waterflood effect may be observed as the reservoir is slightly water wet. This
difference between LSWF and HSWF might be interpreted as a change in reservoirs
wettability with invasion of low salinity brine. This can be proved by relative
permeability curves derived by core-scale history matching, which was discussed below.
In conclusion, low salinity water can accelerate the oil production rate and decelerate the
water production rate by reduction of the residual oil saturation in reservoirs.
Low salinity EOR-effect on reservoir cores were also observed by the relative
permeability curves (Figure 5), which were derived from the core-scale history matching
by ECLIPSE 100 with low salinity model. Figure 5 shows that relative permeability
curves for low salinity waterfood moves toward to the right, which might be interpreted
as the transition of the wettability towards more water-wet. The capillary pressure curves
were not shown here, due to the insensitivity to the core-scale history matching in the low
permeability corefloods. As the corefloods experiments shown, the differential pressure
was 1 to 2 MPa, which was much higher than the capillary pressure.

SCA2015-004

8/12

The incremental oil recovery by LSW compared to HSW could be attributed to the
expansion of the electric double layer, which is caused by LSW as a result of increasing
the magnitude of the negative electric charge at the interfaces of oil/brine and rock/brine.
HSW caused weak surface charges at both oil/brine and rock/brine interfaces. However,
more negative charges at oil/brine and brine/rock were generated by LSW. According to
the theoretical calculation of thermodynamics, which will be discussed at the last part of
this study, more negative charges at the interfaces of oil/LSW and LSW/rock and
corresponding increase of the repulsive forces might detach the oil film from the rock
surface.

Figure 3 Low salinity waterflooding

under secondary mode with the core
from the reservoir (slightly water-wet)

Figure 4 Low salinity waterflooding

under tertiary mode with the core from
the reservoir (slightly water-wet)

Figure 5 Relative permeability curves of low salinity and high salinity waterflooding (reservoir cores)

Effect of Low salinity Waterflooding on Water-wet Reservoir

According the the literature, the low salinity EOR-effect was rarely observed in the
water-wet reservoirs. In order to further confirm and reveal the mechanism behind his
phenomenon, two core plugs extracted from the outcrop were used to perform
coreflooding with low salinity water, either in secondary mode or tertiary mode (Figure 6
and Figure 7). The experimental oil was from the Changqing Oil Field and the core plugs
were put into the oven with 65oC for four weeks after the cores desaturated with
experimental oil to restore the wettability. Additionally, the corefloods experiments were

SCA2015-004

9/12

successfully history matched to acquire the relative permeability curves, as shown in

Figure 8.
Figure 6 and Figure 7 show that slight low salinity EOR-Effect was observed from low
salinity waterflooding under secondary and tertiary mode. Figure 6 presents that 50.5%
oil recovery of OOIP was accumulated by low salinity waterflooding. Figure 7 also
shows that 50.0% oil recovery of OOIP was obtained by the high salinity waterflooding
under secondary mode. Additional 3.4% oil recovery of OOIP was observed by low
salinity waterflooding under tertiary mode. Relative permeability curves illustrates that
the wettability of the outcrop was preferential to water-wet and minor wettability change
during the low salinity waterflood was observed (Figure 8). Coreflooding experiments, by
using water-wet outcrop core plugs, verified that low salinity effect may not work in the
water-wet reservoirs. Moreover, with comparison of Figure 4 and Figure 7, piston-like
displacement was found for high salinity water flooding as the core was water-wet, while
the oil was produced over much longer periods for high salinity water flooding as the
core was slightly water wet. This could be due to the more favorable mobility ratio
formed by the low salinity water flooding in the reservoir [14].

Figure 6 Low salinity waterflooding under

secondary mode with the core from outcrop
(water-wet)

Figure 7 Low salinity waterflooding under

tertiary mode with the core from outcrop
(water-wet)

Figure 8 Relative permeability curves of low salinity and high salinity waterflooding (outcrop, water-wet)

THERMODYNAMICS ANALYSIS

SCA2015-004

10/12

Thermodynamics Background
Hirasaki has investigated the thermodynamics of the thin films to determine the
interdependence of spreading, contact angle and capillary pressure using the DLVO
theory and Laplace-Young Equation [8]. The intermolecular forces comprise of the van
der Waals, electrical and structural forces.

Total Van der Waals electrical structural

(7)

Where, Total is the disjoining pressure of the specific intermolecular interactions which
reflects the interactive forces between the interface of water/oil and water/rock. In this
study, dielectric constants for each of medium were tested through N5224A microwave
network analyzer, 2.8410-21J was used as the Hamaker constant for outcrop/brine/oil
system, 3.7010-21J was used for reservoir rock/brine/oil system, A brief introduction of
the forces and calculation procedures are presented in literatures [15].
Thermodynamic Calculation on Reservoir Cores and Outcrop
According to the theoretical model, structural forces were assumed to be as same with
respect to the variation of molarity and zeta potential. Therefore, the variations of
disjoining pressure could be explained by only considering the electrostatic interactions
and Van der Waals forces, calculated with consideration of mineralogy difference. With
combination of zeta potential measurement and composition of the brines, the disjoining
pressure versus water film thickness was calculated (Figure 9) to unveil the deeper
mechanism behind the low salinity waterflooding.
Figure 9 shows that the formation brine resulted in attractive force between the interface
of oil/brine and interface of brine/rock in the slightly water wet rock. However, the
repulsive force between the interfaces of oil/brine/rock was formed by low salinity water
due to the low salinity and highly negative zeta potential for both interfaces of oil/water
and water/rock. Figure 9 also shows that the water film was thicker in high salinity brine,
compared with the low salinity brine due to the double layer expansion. Thicker water
film indicates that the rock surface is preferential to water-wet with lower surface energy
with presence of low salinity brine. Moreover, the thermodynamic isothermal calculation
is consistent with the results of wettability test and corefloods experiments for the
reservoir rocks.
Figure 10 illustrates that both formation brine and low salinity water causes the repulsive
force between the interfaces of oil/water and water/rock in the water-wet rock. This also
fits the results of wettability test and corefloods experiments for the water-wet outcrop
rocks. Even though the thicker water film and repulsive force were formed as result of
double layer expansion in low salinity brine, repulsive force still remained in the
formation brine with thinner water film. The thermodynamic analysis for water-wet rock
discloses that low salinity EOR-effect would not be observed in the water-wet reservoirs.
Therefore, the conclusions can be drawn are: firstly, the thermodynamic isothermal can
be applied to unveil the mechanism behind the low salinity EOR-effect. Secondly, it can

SCA2015-004

11/12

be used to screen the candidate reservoirs for low salinity waterflooding. Thirdly,
composition of injection brine can be manipulated by the calculation of thermodynamics.

Figure 9 Disjoining pressure versus film

thickness with formation brine and LSW in oilwet rock

Figure 10 Disjoining pressure versus film

thickness with formation brine and LSW in
water-wet rock

CONCLUSION
In this study, the influence of rock wettability on the low salinity EOR-Effect was
investigated by coreflood and history matching through ECLIPSE 100. Several
observations were made in this study.

Low salinity EOR-effect was observed during the low salinity waterflooding at both
secondary and tertiary mode in the slightly water wet reservoir rock.
Low salinity water can accelerate the oil production rate and decelerate the water
production rate by reduction of the residual oil saturation in slightly water wet
reservoirs.
Slight wettability change and additional oil recovery were observed during the low
salinity waterflood in the water-wet outcrop rock.
Low salinity EOR-effect might be interpreted by thermodynamics of wettability.
Thermodynamic calculation could help design the composition of the injection brine
to enhance oil recovery.

ACKNOWLEDGEMENTS
The authors would like to thank Research Institute of Petroleum Exploration and
Development of PetroChina for permission to publish this paper.

REFERENCES
1.

2.

Myint, P.C. and A. Firoozabadi, Thin liquid films in improved oil recovery from
low-salinity brine. Current Opinion in Colloid & Interface Science, 2015. 20(2): p.
105-114.
Sheng, J.J., Critical review of low-salinity waterflooding. Journal of Petroleum
Science and Engineering, 2014. 120(0): p. 216-224.

SCA2015-004

3.

4.
5.
6.
7.

8.
9.

10.

11.

12.

13.
14.
15.

12/12

Sohrabi, M. and A. Emadi, Visual Investigation of Oil Recovery by Low Salinity

Water Injection: Formation of Water Micro-Dispersions and Wettability Alteration,
in SPE Annual Technical Conference and Exhibtion. 2013, 2013, Society of
Petroleum Engineers: New Orleans, Louisiana, USA.
Morrow, N. and J. Buckley, Improved Oil Recovery by Low-Salinity
Waterflooding. Journal of Petroleum Technology, 2011. 63(5): p. 106-112.
Lager, A., et al., Low Salinity Oil Recovery-An Experimental Investigation1.
Petrophysics, 2008. 49(1).
A.Lager, K.J.W., C.J.J.Black, M.Singleton, K.S.Sorbie, Low salinity oil recoveryan experimental investigation. SCA2006-36, 2006.
Nasralla, R.A. and H.A. Nasr-El-Din, Double-Layer Expansion: Is It A Primary
Mechanism of Improved Oil Recovery by Low-Salinity Waterflooding?, in SPE
Improved Oil Recovery Symposium. 2012, Society of Petroleum Engineers: Tulsa,
Oklahoma, USA.
Hirasaki, G.J., Wettability:fundamentals and surface forces. SPE Formation
Evaluation, 1991. 6(2): p. 217-226.
Tang, G.-Q. and N.R. Morrow, Influence of brine composition and fines migration
on crude oil/brine/rock interactions and oil recovery. Journal of Petroleum Science
and Engineering, 1999. 24(2-4): p. 99-111.
Xie, Q., et al., Ions tuning water flooding experiments and interpretation by
thermodynamics of wettability. Journal of Petroleum Science and Engineering,
2014. 124(0): p. 350-358.
Austad, T., A. Rezaeidoust, and T. Puntervold, Chemical Mechanism of Low
Salinity Water Flooding in Sandstone Reservoirs, in SPE Improved Oil Recovery
Symposium. 2010: Tulsa, Oklahoma, USA.
Xie Quan, H.S., Pu Wanfen, The Effects of Temperature and Acide Number of
Crude Oil on the Wettability of Acid Volcanic Reservoir Rock From the Hailar
Oilfield. Petroleum Science, 2010. 7: p. 93-99.
Masalmeh, S.K., et al., Low Salinity Flooding: Experimental Evaluation and
Numerical Interpretation. 2014, International Petroleum Technology Conference.
Jerauld, G., et al., Modeling low-salinity waterflooding. SPE Reservoir Evaluation
& Engineering, 2008. 11(6): p. 1000-1012.
Hirasaki, G.J., Wettability:fundamentals and surface forces. SPE Formation
Evaluation, 1991c. 6(2): p. 217-226.

SCA2015-005

1/12

CONTRADICTORY TRENDS FOR SMART WATER

INJECTION METHOD: ROLE OF PH AND SALINITY
FROM SAND/OIL/BRINE ADHESION MAPS
Mathilde Didier, Annabelle Chaumont, Thibaut Joubert, Igor Bondino, and Grald Hamon
TOTAL SA, CSTJF, Avenue Larribau, Pau, France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Smart Water Injection Methods (SWIM) requires a good understanding of the
evolution of wettability not only as a function of salinity but also of pH. In order to
have a better understanding of this parameter, adhesion tests were performed on crude
oil/sand/brine systems. These tests were developed according to an in-house
experimental protocol which guarantees full control of pH during its various stages and
when crude oil/brine/rock are in equilibrium. Fontainebleau and Ottawa sands were
used to understand the effect of pH and salinity on wettability. The brine was
composed of a mixture of NaCl and CaCl2 (ratio 9/1). The salinity varies from 7.9 to
785 mM and pH from 2 to 12.
Firstly, our adhesion tests are in good agreement with previous adhesion maps from the
literature, although the experimental techniques are very different. Moreover our tests
incorporate the effect of divalent ions and long term interactions between oil/water and
the solid, which were not accounted for in previous studies. Our results confirm that at
low salinity and acidic pH (pH < 7), oil adhesion increases for both sands, although
quite more importantly for Fontainebleau. In that respect, our adhesion maps highlight
contradictory results compared to the recent literature regarding the effect of salinity on
wettability.
Secondly we find three distinct oil adhesion areas (for salinity from 0.5 to 50 g/L)
which vary with equilibrium pH: from pH = 2.5 to 6 and pH = 8.5 to 12 showing oil
wettability, pH = 6 to 8.5 showing water wettability. Interestingly it is found that
critical pH thresholds exist in the range pH = 6 to 8.5 at which large variations in the
adhesion between oil and sand are seen. This critical pH might be rock dependent: these
three areas are specifically pronounced for Fontainebleau sand compared to Ottawa
sand where the water wet area is more important (pH = 5.5 to 10). Considering the fact
that both sands are mainly composed of quartz, it would imply that the predominant
mineral species is not an impacting factor on wettability and therefore on SWIM.

INTRODUCTION
SWIM (Smart Water Injection Methods) have been widely investigated in the last 15
years. This is an emerging EOR (Enhanced Oil Recovery) technique highlighted by
Tang and Morrow [1] in Berea sandstone, which is more and more studied due to its

SCA2015-005

2/12

economical aspect and feasibility. There are in the literature, numerous papers on this
topic in order to understand the mechanism, but at this time, the real mechanism or
combination of mechanisms which lead to the increase of the oil recovery have not
been clearly defined. Several authors have various approaches to explain this effect as
fine migration [2] controlled by clay particles, alkaline-flooding behavior which relies
on interfacial tension reduction caused by pH increase [3], Multiple-Ion-Exchange
(MIE) process [4], pH elevation [5] and formation of water micro-dispersions in the oil
phase [6]. None of these mechanisms on its own can be retained to explain the increase
in oil recovery: rather it is nowadays accepted that a combination of several factors and
a mix of all these mechanisms is behind the SWIM effect [7]. A general consensus is
that low salinity water may induce a change in the wettability towards a more water wet
condition, which would somehow help in releasing additional oil from the pore space.
But this mechanism has not yet been visualized or demonstrated in a convincing
manner.
Since wettability is an important parameter which determines the fluids distribution and
has an effect on oil recovery, several authors performed adhesion experiments in order
to understand the interaction between crude oil, brine and rock [8-12]. Usually the
reservoir sandstone rock is represented by crystalline quartz or various glass surfaces
and classically two kinds of experiments are performed: contact angle [9, 10, 12]
measurement and adhesion of a drop of crude oil on the surface [8, 10, 12]. In both
cases, crude oil is surrounded with a brine composed of sodium chloride at various
salinities and with a pH which varies from 2 to 10 and several parameters are normally
tested as pH, Temperature, I (ionic strength), nature of oil, amongst others. From these
results, adhesion maps are built which give a rapid and semi-quantitative means for
characterizing crude oil interactions with a brine/solid system. Most researches have
highlighted three areas in the adhesion map as a function of pH, brine salinity [9, 10]
whose extension depends on crude oils nature, brines ionic composition, temperature
and nature of solid:
-

Adhesion at low pH (pH < 6) and low salinity (TDS < 5.8 g/L)
Non-adhesion for pH comprised between 6 and 9 at moderate salinity (TDS
50 g/L)
Adhesion at high pH (pH > 9) and high salinity (TDS > 50 g/L)

Lebedeva and Fogden [13] have shown similar results for kaolinite. It is seen that pH
has a non negligible effect on adhesion. These results also highlight that there is more
adhesion at lower salinity, which is difficult to reconcile with the claim that at low
salinity, the system become more water wet. Nevertheless all these previous
experiments were performed in a very short time for the contact between oil and glass
(less than 10 min) which means that the system equilibrium may not have been reached.
Moreover, no divalent ions were present in surrounding brine and for some authors [14,
16, 24-26], their presence in the brine is essential to observe a low salinity effect.
In this paper, we developed adhesion tests with a home-made protocol which allow the
equilibrium of the system for various pH and salinities. Simple solids were used as pure

SCA2015-005

3/12

sands. This study allows us to evaluate the combined effect of salinity and pH on oil
adhesion to the rock grains.

EXPERIMENTAL PROCEDURES
Materials
Minerals
For this study, two pure sands were used, Fontainebleau (from VWR) and Ottawa
sands (from Fischer Chemical) in order to mimic sandstone reservoirs. These samples
have been chosen for their purity in terms of mineralogy. Both sands are comparable
in terms of mineralogy and are mainly composed of quartz (SiO2) (Table 1).
Table 1: Mineralogy of sands from XRF measurements

Sands
Fontainebleau
Ottawa

SiO2
99.914
99.896

CaO
0.070
0.071

K2O
0.006
0.013

P2O5
0.014
0.019

Fe2O3
0.017
0.021

TiO2
0.017
0.007

Ba
0.045
0.000

These samples were also characterized with a binocular magnifier to evaluate their
external structure (Figure 1). We can remark that the sands have completely different
structure. These sands were also characterized by laser granulometry. Ottawa sand has
bigger particles with a size of 880 m compared to Fontainebleau which has a particle
size of 260 m.

Figure 1: Sand observations with the binocular magnifier, Ottawa sand (left) and Fontainebleau
sand (right)

Crude Oil
The same crude oil was used for all the adhesion tests. The oil has been filtered at 40
m before use. Table 2 highlights its properties.
Table 2: Analysis of crude oil

Property
TAN
TBN
Density @ 40 C
Density @ 60 C
Viscosity @ 40 C
Viscosity @ 60 C
% Asphaltenes
TCC*

Units
mg/g KOH
mg/g KOH
kg/m3
kg/m3
cSt
cSt
weight %
C

*TCC: Critical Temperature of Crystallization

Crude oil
0.2
1.77
834
820
6.26
3.94
2.9
35

SCA2015-005

4/12

Brines
Synthetic brine solutions were prepared at various salinities with 90 wt% of NaCl and
10 wt% of CaCl2. NaCl(s) and CaCl2(s) were manufactured by Sigma Aldrich with a
purity higher than 99.9%. Demineralized water was used. Brine solution pH was
adjusted with HCl(aq) or NaOH(aq) solutions at 0.1 or 1 M. After stirring, all the
solutions were filtered at 0.45 m, pH varies from 2 to 12 and it was controlled during
manipulation. Tested TDS (Total Dissolved Salts) were 0.5, 1, 3, 7, 11, 15, 25 and 50
g/L. The properties of the brine used in this study are summarized in the Table 3.
Table 3: Salinity, molar concentration and ionic strength of used brines

TDS(g/L)
0.5
C (mM)
7.85
I*
0.019
*: Ionic Strength

1
15.7
0.036

3
47.1
0.11

7
11
0.25

11
17.3
0.40

15
23.6
0.55

25
39.3
0.91

50
78.5
1.82

Experimental Protocol
Sand Cleaning
In order to eliminate all contaminations, mainly by iron, the sand was washed
beforehand with an acid solution (HCl(aq), 1M), then neutralized with sodium
bicarbonate (NaHCO3(aq), 1M), rinsed thoroughly with demineralized water and
finally oven dried at 60C for 24 h.
Protocol of Adhesion Tests
The test tube experiment is a fast screening of sand/oil/brine interactions permitting to
focus on the effect of one parameter at the time. These batch experiments use a static
procedure which is described below.
1. Prepare brine at the chosen salinity.
2. Place a known mass of cleaned sand in contact with a known volume of brine.
The brine has to be in excess compared to the sand.
3. Add HCl(aq) at C = 0.1 M or 1 M or NaOH(aq) at C = 0.1 M to adjust the pH
to the chosen value.
4. Leave the solution at 60C during 24h for equilibration; regularly shake
manually to homogenize the sand/brine system.
5. Measure the pH.
If the target pH is not reached, repeat the steps 3 and 4 to eliminate the
buffering effect of sand.
6. When the target pH is reached and is seen to be stable, mark this pH as initial
pH (pHi). Now add a known volume of oil to the system.
7. Shake manually the glass vial to free the oil that is trapped under the weight of
the sand and not actually adhering to it.
8. Leave the system at 60C during 48 h under continuous stirring.
9. Observe wetting state at equilibrium at T = 60C. The wetting state is the
evaluation of the oil quantity which is in adhesion on and in the sand compared
to the oil which stays above the brine phase (Figure 4 for example). Three
states were defined; adhesion, transition and non-adhesion. When the oil
quantity in the sand is higher than in the brine, then there is adhesion. If there is

SCA2015-005

5/12

quite the same quantity of oil in the sand and in the brine phase, then it is
transition state. Finally when no oil adheres in the sand, then it is non-adhesion
state.
10. Shake manually, wait for equilibrium and observe the wetting state. As
previously, manual shake makes the system more homogeneous and helps to
estimate visually the actual sand and oil adhesion by freeing the oil that is
trapped under the weight of the sand.
11. Measure the final pH (pHf) after filtration at 0.45 m of the brine in the system.
12. Report the observed wetting state at fixed pH in an adhesion map
13. Perform this protocol for another initial pH until the adhesion map is complete
with all the tested salinity and initial pH
Remark: Equilibrium was estimated with tests at various experiment durations. It was
estimated that after 48h of stirring, the interaction between oil, sand and brine stays the
same even after more hours of stirring at temperature.
This protocol allows removal of the buffering effect of the sand (which can be quite
important) and to control the initial and final pH of the system. The key parameter in
this study is the control of pH. The final pH could be estimated with the pHi-pHf
diagram of the used crude oil (Figure 2). We can remark in this graph that the
evolution of pH is the same at low salinity (1 g/L) as at high salinity (50 g/L). For this
studied oil, when initial pH is comprised between 4 and 10, the final pH is between 6
and 8. This oil is mainly a basic one.

Figure 2 : pHi-pHf diagram of the tested crude oil at 45C in NaCl/CaCl2 brine at low and high
salinity (TDS = 1 and 50 g/L)

SCA2015-005

6/12

RESULTS AND DISCUSSIONS

The adhesion map of Fontainebleau sand is displayed in Figures 3a and 3b.

Figure 3a: Adhesion map of Fontainebleau sand

Figure 3b: Adhesion map of Fontainebleau sand (zoom on low salinity (TDS<10 g/L))

SCA2015-005

7/12

The information in this map can be approximated by three main adhesion areas
roughly situated at:
-

2 < pH < 6 : oil adheres to grains

6 < pH < 8 : oil does not adhere to grains
8 < pH < 12 : oil adheres to grains

An example of test tube experiments is displayed in Figure 4 to illustrate this adhesion

map at 7 g.L-1. In this example, adhesion occurs at final pH equal to 5.42 and at higher
pHf (10.17), whilst non-adhesion is seen for pHf = 7.23.

5.42

7.23

10.17

Figure 4: Glass vial adhesion map

results at fixed salinity (7 g.L-1)
and for three pH (pHf) values.
From left to the right: adhesion,
non-adhesion, adhesion.

6.38

7.11

7.74

Figure 5: Glass vial adhesion map

results at fixed salinity (3 g.L-1) and
for three pH (pHf) values. From
left to the right: non-adhesion,
transition, non-adhesion.

By decreasing the pH, whatever the salinity, a greater adhesion of oil on the
Fontainebleau sand is visible, a result in agreement with the work of Yang et al. [10]
and Drummond et al. [9]. Oil adhesion is systematic for pH below 5 as well as for pH
higher than 10. In addition, at low salinity (TDS < 3 g/L), the zone of oil adhesion
extends towards higher pH. In this area (6 < pH < 8) small variations of salinity for a
given pHf can create very different adhesion results (Figure 5).
This observation is very important since pH values expected at reservoir conditions
would cover the interval 5 < pH < 7 [22]: wettability being so sensitive in these pHsalinity ranges, could create an important variability of outcomes and would render
SWIM performance not trivial to assess.
In conclusion, the result obtained for Fontainebleau shows that the rock becomes more
oil wet when salinity decreases, in contradiction to comments in literature on smart
water techniques ([14-17]) where when decreasing the salinity, the water wetness of
the solid increases, leading to the increase of oil recovery.
In Figure 6 the adhesion map for Ottawa sand is displayed. There are three main areas
of adhesion, mainly controlled by pH, as for Fontainebleau, but here there is much less
influence of salinity. The oil adheres when pH decreases (pH < 5) and for low salinity
(TDS < 10 g/L) as well as for basic pH (> 11) for all salinities. The last area in this
adhesion map shows water wet condition.

SCA2015-005

8/12

Figure 6: Adhesion map of Ottawa sand

Since both sands share a very similar mineralogy, then we would conclude that for this
case sample mineralogy does not have an important effect on wettability.
The three areas in the adhesion maps of Fontainebleau and Ottawa sands could be
partially explained by adsorption of oil components on sand and electrostatic
repulsion, at least at low pH (Figure 7).

Figure 7: Interpretation of adhesion maps. Continuous and dotted green curves show the transition from
water-wet to oil-wet for Fontainebleau and Ottawa sand respectively

The results of this study could be compared with those from Bondino et al. [18], where
the authors used similar experimental setup to evaluate oil adhesion on sand. They
highlighted also oil adhesion at high and low salinity conditions (TDS = 1 and 50 g/L)

SCA2015-005

9/12

and for acidic pH (pHi < 5). Water wetness is observed at basic pH (pH > 7) and for
low salinity. Nevertheless the main difference between the two studies is that in [18],
the pH was not controlled and measured at the end of the experiment (therefore the pH
discussed in the paper happens to be the initial one). Regarding our results, we
demonstrate pH is a key parameter in wettability variation and has to be controlled and
measured all along the experiment.
These results are comparable with those from the literature [8-12] where adhesion
maps highlight also three areas which depend on pH and salinity range. Oil adhesion
occurs also for low pH (< 6) and high pH (> 9) and non-adhesion for intermediate pH
(6< pH < 10) and salinity ( 50 g/L). Transition zone for low salinity (TDS < 5.8 g/L)
is located for pH close to 6 and 8, as observed in this paper. This pH area is therefore
critical and pH seems to be a key parameter in oil adhesion. This study confirms these
previous contradictory results where oil adhesion increases for low salinity range.
Nevertheless the test tube experiments presented in this paper seem to be more
representative of SWIM compared to the previous adhesion tests. Indeed, compared to
the results of [8-12], duration of experiments is longer and led to equilibrium within
the system oil/brine/sand. Moreover brines were composed of mono and divalent ions
as Ca2+ which is favourable for low salinity effect [4, 16] compared to brines with only
monovalent ions in the cited literature.
Adhesion maps (Figures 3 and 6) could be explained by electrostatic interaction
between oil and sand. The isoelectric point of quartz is estimated to be close to 2.5
([19, 20]) which means that quartz (Qz) is negatively charged for pH higher than 2.5
(Figure 8). Usually, acid and base components of oil are simplified by RNH2, RCOOH
and Quinoline. pKa of these redox couples are illustrated in Figure 8. At low pH
(< 4.9), oil components are essentially positive and inversely at basic pH (> 8.1).

Figure 8: pKa of oil components (RNH2, RCOOH, Quinoline) and isoelectric point of Quartz (Qz)

Figure 7 summarizes adhesion maps with adhesion areas for Fontainebleau and
Ottawa sands. In Area 1, adhesion occurs at low pH (2.5 < pH <5) due to reduction of
electrostatic interaction between sands negatively charged and oil components
positively charged (R+NH3, Quinoline positively charged). In Area 2, non adhesion
occurs at higher pH (pH > 4.9) due to repulsion between sands still negatively charged
and neutral (RNH2, Quinoline) and negative (CH3COO-) oil components. However,
when salinity increases in Area 1, oil adhesion occurs at lower pH for Fontainebleau
sand but not for Ottawa sand. One hypothesis is the influence of roughness and
granulometry which could increase specific surface area of sand grains and therefore
increase the probability of adhesion. The roughness of Fontainebleau sand seems
higher than Ottawas, therefore its specific surface area could be higher and this will
increase the probability of oil adhesion. Similar observations were already highlighted

SCA2015-005

10/12

in the literature [21]. In the Area 3, oil adhesion at very basic pH (pH > 10) is more
difficult and complex to explain: further investigations are ongoing.
This variation of adhesion with pH suggests also that not only the final pH is
necessary to know, but also the initial pH of the system. If there is a pH evolution
during the waterflooding experiment in the system rock/brine/oil, we could have a
change of wettability which could have an impact on the recovery.
This study could be useful for understanding spontaneous imbibitions tests, as
adhesion maps were elaborated with test tube experiments without dynamic aspect. In
the literature, some authors highlight high response variability for spontaneous
imbibitions tests with various carbonates rocks [22] or various type of oil on Berea
sample [23] for example. Romanuka et al. [22] show various wettability modification
in function of rock samples or ionic composition of injected water. Suijkerbuijk et al.
[23] found very different oil recovery during spontaneous imbibitions with ten
different oils on Berea sandstones. Adhesion tests as described in this study could be
useful screening tests for understanding the variability in response of SWIM
spontaneous imbibitions experiments.

CONCLUSIONS
In this paper, test tubes experiments were performed in order to implement adhesion
map for crude oil/sand/brine system. A home-made protocol allows the system to
reach equilibrium at temperature with full control of the pH. Indeed, pH appears to be
an essential parameter in wettability, more than the salinity. Adhesion map of
Fontainebleau sand highlights oil adhesion at low salinity (TDS < 3 g/L) in reservoir
pH area (6 < pH < 8) and non adhesion for a salinity higher than 3 g/L in this pH area.
Moreover, this domain is very sensitive to pH variation. A little variation in pH will
change drastically the wettability. This highlight that pH has to be well monitored
during waterflooding experiments.
The fact that low salinity increases the oil wettability is contradictory to what is
expected in the literature for SWIM, where low salinity waterflooding is seen to
increase water wetness and then the oil recovery. In light of these results, we think that
it would be more appropriate to state that oil recovery depends from the transition
from some initial to some final state, and not simply from a system becoming more
water wet.
It was rather interesting to note that Ottawa and Fontainebleau highlight different
adhesion maps despite the fact that their mineralogy is quite similar. Therefore the
mineralogy may not always be a determinant parameter in low salinity effect.
Moreover both sands have various roughness, which could be an important parameter
for adhesion behavior. The effect of surface roughness and granulometry is currently
under investigation. Finally electrostatic interactions between acid and base
components of the oil seem to control the adhesion on the solid.

SCA2015-005

11/12

ACKNOWLEDGEMENTS
The authors would like to gratefully acknowledge TOTAL SA for funding this project
and for the technical support.

NOMENCLATURE
SWIM
EOR
TCC
TDS

Smart Water Injection Methods

Enhanced Oil Recovery
Critical Temperature of Crystallization
Total Dissolved Salts

I
wt%
C
M

Ionic Strength
Weight %
Molar Concentration
mol/L

REFERENCES
[1] G.Q. Tang, N.R. Morrow, Salinity, Temperature, Oil Composition, and Oil
Recovery by Waterflooding, in: SPE Reservoir Engineering, Society of Petroleum
Engineers, Denver, Colorado, 6-9 October, 1997, SPE-36680-PA.
[2] G.-Q. Tang, N.R. Morrow, Influence of brine composition and fines migration on
crude oil/brine/rock interactions and oil recovery, Journal of Petroleum Science and
Engineering, 24 (1999) 99-111.
[3] P.L. McGuire, J.R. Chatham, F.K. Paskvan, D.M. Sommer, F.H. Carini, Low
Salinity Oil Recovery: An Exciting New EOR Opportunity for Alaska's North Slope,
in: Society of Petroleum Engineers, Society of Petroleum Engineers, Irvine,
California, 30 March - 1 April, 2005, SPE 93903.
[4] A. Lager, K.J. Webb, C.J.J. Black, M. Singleton, K.S. Sorbie, Low Salinity Oil
Recovery - An Experimental Investigation1, in: SCA2006-36, Society of
Petrophysicists and Well-Log Analysts, Trondheim, Norway, 12-16 September, 2006.
[5] T. Austad, A. Rezaeidoust, T. Puntervold, Chemical Mechanism of Low Salinity
Water Flooding in Sandstone Reservoirs, in: S. 129767 (Ed.), Society of Petroleum
Engineers, 2010.
[6] A. Emadi, M. Sohrabi, Visual Investigation of Oil Recovery by LowSalinity Water
Injection: Formation of Water Micro-Dispersions and WettabilityAlteration, in: S.
166435 (Ed.) Society of Petroleum Engineers, Society of Petroleum Engineers, New
Orleans, Louisiana, USA, 2013.
[7] B. Soraya, C. Malick, C. Philippe, H.J. Bertin, G. Hamon, Oil Recovery by LowSalinity Brine Injection: Laboratory Results on Outcrop and Reservoir Cores, in,
Society of Petroleum Engineers.
[8] J.S. Buckley, N.R. Morrow, Characterization of Crude Oil Wetting Behavior by
Adhesion Tests, in: S. 20263 (Ed.) Society of Petroleum Engineers, Society of
Petroleum Engineers, Tulsa, Oklahoma, 22-25 April, 1990.
[9] C. Drummond, J. Israelachvili, Fundamental studies of crude oilsurface water
interactions and its relationship to reservoir wettability, Journal of Petroleum Science
and Engineering, 45 (2004) 61-81.
[10] S.Y. Yang, G.J. Hirasaki, S. Basu, R. Vaidya, Mechanisms for contact angle
hysteresis and advancing contact angles, Journal of Petroleum Science and
Engineering, 24 (1999) 63-73.
[11] S.Y. Yang, G.J. Hirasaki, S. Basu, R. Vaidya, Statistical analysis on parameters
that affect wetting for the crude oil/brine/mica system, Journal of Petroleum Science
and Engineering, 33 (2002) 203-215.

SCA2015-005

12/12

[12] L. Liu, J.S. Buckley, Alteration of wetting of mica surfaces, Journal of Petroleum
Science and Engineering, 24 (1999) 75-83.
[13] E.V. Lebedeva, A. Fogden, Wettability alteration of kaolinite exposed to crude oil
in salt solutions, Colloids and Surfaces A: Physicochemical and Engineering Aspects,
377 (2011) 115-122.
[14] M. Rotondi, C. Callegaro, F. Masserano, M. Bartosek, Low Salinity Water
Injection: enis Experience, in, Society of Petroleum Engineers (2014).
[15] R. Julija, H. Jan, L. Dick Jacob, S. Bart, M. Fons, O. Sjaam, B. Niels, L. Hilbert
van der, A. Hakan, A. Tor, Low Salinity EOR in Carbonates, in: SPE Improved Oil
Recovery Symposium, Society of Petroleum Engineers, Tulsa, Oklahoma, USA
(2012).
[16] S. Strand, E.J. Hognesen, T. Austad, Wettability alteration of carbonates - Effects
of potential determining ions (Ca2+ and SO42-) and temperature, Colloids and Surfaces
A: Physicochemical and Engineering Aspects, 275 (2006) 1-10.
[17] A.A. Yousef, S.H. Al-Saleh, A. Al-Kaabi, M.S. Al-Jawfi, Laboratory
Investigation of the Impact of Injection-Water Salinity and Ionic Content on Oil
Recovery From Carbonate Reservoirs, in.
[18] I. Bondino, S. Doorwar, R. Ellouz, G. Hamon, Visual microscopic investigations
about the role of pH, salinity and clay on oil adhesion and recovery, in: SCA, NAPA
Valley, California, 2013.
[19] F.C. Schoemaker, D.M.J. Smeulders, E.C. Slob, Electrokinetic Effect: Theory
And Measurement, in, Society of Exploration Geophysicists.
[20] P.L. Churcher, P.R. French, J.C. Shaw, L.L. Schramm, Rock Properties of Berea
Sandstone, Baker Dolomite, and Indiana Limestone, in, Society of Petroleum
Engineers.
[21] M.C. Michalski, S. Desobry, J. Hardy, Adhesion of Edible Oils and Food
Emulsions to Rough Surfaces, LWT - Food Science and Technology, 31 (1998) 495502.
[22] J. Romanuka, J. Hofman, D.J. Ligthelm, B. Suijkerbuijk, F. Marcelis, S. Oedai, N.
Brussee, H. van der Linde, H. Aksulu, T. Austad, Low Salinity EOR in Carbonates, in,
Society of Petroleum Engineers, 2012.
[23] B. Suijkerbuijk, J. Hofman, D.J. Ligthelm, J. Romanuka, N. Brussee, H. van der
Linde, F. Marcelis, Fundamental Investigations into Wettability and Low Salinity
Flooding by Parameter Isolation, in, Society of Petroleum Engineers, 2012.
[24] P. Zhang, M.T. Tweheyo and T. Austad, Wettability Alteration and Improved Oil
Recovery in Chalk: the Effect of Calcium in the Presence of Sulfate, Energy & Fuels,
20 (2006) 2056-2062
[25] R. Gupta et al., Enhanced Waterflood for Middle East Carbonate Cores Impact
of Injection Water Composition, Manama, SPE 142668 (2011).
[26] S.J. Fathi, T. Austad and S. Strand, Smart Water as wettability Modifier in Chalk:
The effect of salinity and ionic composition, Energy Fuels, 24 (2010) 2514-2419.

SCA2015-006

1/12

LOW SALINITY WATER FLOODING: FACTS,

INCONSISTENCIES AND WAY FORWARD
G. Hamon
TOTAL
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
15 years after the first experimental evidence of increased oil recovery by low salinity
water injection (LSWI), clear understanding of the mechanisms has not emerged yet out
of more than 500 published laboratory experiments.
Firstly, it is shown that there is increasing experimental evidence that published tertiary
LSWI core floods do not often succeed in increasing significantly the recovery within the
2-3 first PVs of tertiary injection, despite strong claims of positive results.
Then, this paper focuses on sandstones and mostly on studies where secondary LSWI
performs better than secondary high salinity water injection (HSWI). Even in such cases,
some examples show that the efficiency of tertiary LSWI may range from poor to nil.
These cases satisfy all required conditions, such as presence of clay, of connate water,
and mixed wettability.
Conditions of existence of a double saturation shock, effects of dispersion in water phase at
SorwHS, and the direction of wettability modification are the hypotheses discussed in this
paper to understand poor performance of tertiary LSWI. Some key experimental
observations are then compared to these possible explanations. They may explain the vast
majority of published studies, but counter examples can also be produced for any single
proposed mechanism.
This paper also puts in evidence that some types of experimental measurements have been
neglected and would deserve more attention, comments on the effect of interfacial tension
and suggests a new approach for investigating the efficiency of tertiary low salinity water
flooding.

INTRODUCTION
There is a clear evidence that:
Spontaneous imbibition by LSWI is able to increase oil recovery compared to high
salinity [21, 44, 46, 59]. In such cases, LSWI can be used to increase oil recovery of
very heterogeneous matrix reservoirs or highly fractured reservoirs.
Secondary LSWI is able to increase oil recovery compared to HSWI in a large number
of studies [11, 17, 18, 22, 23, 28, 29, 40, 45, 48, 51, 52, 55]. Significant additional oil
is often observed at water breakthrough or immediately after. In such cases, LSWI can

SCA2015-006

2/12

be used to increase recovery of undeveloped fields. The comparison of secondary

recoveries between LSWI and HSWI also showed cases with either very weak benefit
of LSWI [55] or no difference [34] or even negative result [19, 37, 52]. Some authors
[20, 51] suggested that LSWI was always successful but they also showed a wide
variation in positive results. Other authors [5, 55] highlighted that the main challenge
would be to explain the very large scatter in incremental recoveries by LSWI: from nil
to Sorw = -13 saturation units. Morrow [27] concluded that Identification of the
sufficient conditions for LSE and understanding the circumstances under which there is
little or no LSE remain as outstanding challenges.
A large number of laboratory studies consist of tertiary LSWI after secondary HSWI. The
motivations behind tertiary floods are:
1. Reproduce the saturation history of mature, waterflooded reservoirs and check whether
LSWI can be an efficient EOR process,
2. Cope with scarce availability of reservoir cores. Comparison of secondary waterfloods
requires at least twin samples or more if different recipes of injected water are tested
Almost all laboratory tertiary LSWI tests start at Sorw HS and consist of continuous
injection of low salinity afterwards. A lot of authors conclude that tertiary LSWI has
positive effects, based upon final incremental recovery after very long injection periods. In
fact this conclusion is much more debatable than for secondary LSWI. If we decide that a
realistic, positive effect means 1) The arrival of the oil bank at approximately 0,5-0,6 PV
injected or 2) Additional recovery of at least +5%ooip after 1 PV injected, the scene looks
quite different:
Studies showing production of significant incremental oil after a short tertiary injection
period of low salinity water are very scarce [20, 22, 38, 45, 51].
When significant incremental oil is produced during tertiary LSWI, oil is very often
delayed or oil is produced at a low pace over several PVs of injection [3, 10, 11, 21, 22,
23, 31, 32, 33, 45, 49, 55, 56]
Nil or very poor additional recovery by tertiary LSWI has now been largely reported. It
has been suggested that such failures were related to use of outcrop cores, rather than
reservoir cores [52]. In fact, very poor additional recoveries are shown for both:
o Outcrop: [16, 19, 28, 34, 35, 40, 41, 43, 52, 53, 55]
o And reservoir cores [5, 15, 17, 33, 37, 42, 52].
o Taking into account the overrepresentation of outcrop cores in published
laboratory work about LWSI, the conclusion about outcrop versus reservoir
cores remains questionable.
It is worth noting that the three necessary conditions, as defined in [27] for positive effect
of LSWI, were almost always satisfied in these studies.
In the following, the possible reasons of the poor performance of tertiary LSWI are
investigated. Unfortunately, the vast majority of these experiments are not accompanied
by any other type of tests: neither secondary LSWI nor wettability, etc. It prevents detailed
analysis. Therefore, this paper focuses mainly on the scarce laboratory studies which: 1)

SCA2015-006

3/12

Report both secondary and tertiary LSWI 2) Show larger oil recovery by LSWI than by
HSWI on secondary floods. These studies represent crude oil/brine/rock systems where
LSWI effect does exist in secondary mode and we can eliminate debates about causes
which trigger positive low salinity effect.

FACTS, HYPOTHESES AND INCONSISTENCIES

Existence of a Double Saturation Shock
Firstly, we might hypothesize that the evolution of oil production during tertiary LSWI,
starting at SorwHS, strongly depends on conditions which control the existence of a double
saturation shock [30], particularly on fractional flows for secondary high and low salinity
water floods. When reviewing literature LSWI results, viscosity ratio and secondary
water/oil relative permeabilities (Krs) for both LSWI and HSWI would be necessary to
evaluate these conditions for each study. In fact, they are almost never reported. However,
coarse estimation can be done when the viscosity ratio and the evolution of differential
pressure during secondary floods are reported.
Simulations in figure 1a have been performed using a set of Krs for secondary water floods
where Sorw = -0,1 between high and low salinity waters:
With favorable fractional flows, here o/w =2, an oil bank is created in tertiary mode
by LSWI and breaks through rather early, at about 0,5-0,6 PV injected, as suggested
also by other simulations [26] or analytical calculations.
At unfavorable conditions, here for o/w=7, the oil bank vanishes. Incremental oil is
delayed and is produced very slowly. It represents the slow arrival at the outlet of the
spreading wave related to the LSWI. For o/w=18, tertiary oil production is
postponed to very late times.
Experimental observations of a clear oil bank during tertiary LSWI are very scarce in the
literature. When this behavior is reported [20, 22, 45, 51], lack of information often
prevents the comparison of experimental conditions with theoretical estimations about the
existence and stability of the second saturation shock. Oil banking during tertiary LSWI is
consistent with low viscosity oils and with the type of Kr curves for secondary floods
shown by Webb in [51] on three reservoirs: very small difference in Krw(Sorw) between
HS and LS despite significant reduction in residual oil saturation: Sorw~ -0,05 to -0,07.
In fact, most of published tertiary LSWI have been carried out at large [16, 31, 32, 33, 35]
to very large (o/w~18 [40], ~23 [49], ~32 [55], ~40 [52, 53], ~54 [56]) viscosity ratios.
They might explain the absence of oil bank, the delayed oil bump if any, as well as very
slow oil production. Shehata [40] performed tertiary LSW on Bandera sandstone.
Significant benefits were clear in secondary mode by LSWI: Rec~+9% oip. The delayed
arrival of the oil bump, after 1,5 PV injected in figure 1b, is consistent with the
unfavorable viscosity ratio. In the same study, he obtained even better results in secondary
mode on Buff Berea: Rec~+16% OIP, but no tertiary incremental oil by LSWI. Again,
this is consistent with large viscosity ratio.

SCA2015-006

4/12

More surprising are some rare core floods where a fast and strong oil bank is reported in
conditions where it should disappear. Loahardjos [22] cores 2065/1 R1/C1 and 2060/1
R1/C1 did produce a strong oil bank at 0,42 PV tertiary LSWI despite very unfavorable
viscosity ratio (o/w~30). However a 5-10 fold increase in differential pressure is
reported immediately after starting LSWI, which may be related to dispersion of kaolinite
with salinity shock. Although endpoints are not explicitly reported, it shows that
Kw(SorwLS) <<Kw(SorwHS), despite Sorw~ -0,05. This reduction in total mobility might
contribute to the stabilization of tertiary saturation front.

Figure 1: a) simulations of tertiary LSWI as a function of viscosity ratio b) delayed oil recovery
during tertiary LSWI from Shehata, 2014

Can we assume that the evolution of oil production during tertiary LSWI is only controlled
by the fractional flows at high and low salinities?
Cissokho [11] presented positive results of secondary LSWI compared to HSWI, and a
wide range of results in tertiary LSWI on companion sandstone samples with the same
reservoir oil. Unreleased results and analysis are offered in the following. Two tertiary
LSWI at different temperatures, (DU3-D-9, o/w=5,4; DU3-0-5, o/w=7,5) showed
large differences in the amount and timing of incremental oil by LSWI, as illustrated in
figure 2a. It can hardly be explained by conditions related to the existence of a double
saturation shock, as the most significant bump in tertiary oil production is observed for:
The coreflood with the largest viscosity ratio,
For DU3-0-5, where
evolutions of differential pressure show that
Kw(SorwHS)=Kw(SorwLS), whereas the smallest oil bump is observed for DU3-D9, where Kw(SorwHS)>Kw(SorwLS)
It shows that other parameters than the fractional flows at high and low salinities also
control the tertiary behavior during LSWI.

SCA2015-006

5/12

Effect of Dispersion in Water Phase at SorwHS

We might also hypothesize that the evolution of oil production during tertiary LSWI
depends on dispersion in the water phase.
During tertiary LSWI (1g/l) by Cissokho on samples DU3-0-5 (fig. 2b), salinity shock
arrives at the outlet almost simultaneously as the oil bump, after approximately 1 PV
injected. The evolution of chloride concentration shows that 2 PV of LSWI are required
before the produced brine has decreased below 2g/l. The threshold salinity which triggers
tertiary incremental oil is 2,5 g/l in this set of experiments [11]. This observation shows
that dispersion at SorwHS is significant and might delay the effect of LSWI.

Figure 2: a) Evolution of tertiary recovery during LSWI at slightly different viscosity ratios b)
Evolution of recovery, differential pressure and salinity during secondary HSWI and tertiary LSWI

Figure 3a shows the comparison of results of tertiary LSWI carried out by Cissokho on
samples DU3-0-5 and DU3-A-4. Although there is indirect evidence of the salinity shock
through the increase in differential pressure as soon as the low salinity water contacts the
sample in both experiments, there is no incremental oil in tertiary LSWI on DU3-A-4.
Figure 3b shows the comparison between the normalized concentrations during brine tracer
tests which have been performed on sample DU3-A-4 at Sw=1 and at Sorw [11]. At Sw=1,
the effluent curve is symmetrical and injected concentration was produced after 1,7 PV of
injection. It confirms that the core is rather homogeneous. At Sorw, the effluent curve is
largely skewed, and injected concentration was not produced before more than 3,5 PV of
injection. It shows again that dispersion at Sorw might be quite significant, even in
homogeneous rock samples. This observation is in agreement with previous studies, such
as tests PF8A and PF8C in [24]. It shows that residual oil distribution is responsible for
increased dispersion. Significant incremental oil on DU3-0-5 by tertiary LSWI may be
correlated to limited dispersion in water phase at SorwHS whereas much larger dispersion
on DU3-A-4 correlates with no tertiary recovery.

SCA2015-006

6/12

DU3_A_4

Normalized concentration C/C0

1
0,9

0,8
0,7
0,6
0,5
0,4

0,3
0,2

Tracer test Sw=1

Tracer test at Sorw

0,1
0

0 0,5 1 1,5 2 2,5 3 3,5 4 4,5 5

Cumulative injection PV

Figure 3: a) recovery during secondary and tertiary LSWI; b), tracer tests at Sw= 1 and at Sorw
(from Cissokho)

Nasralla [28] shows water floods on four companion Berea sandstone samples, with
different brine salinities: 174, 55, 5,5 Kppm as well as distilled water. Distilled water
injection recovered +19% oip in secondary mode than the injection of 174Kppm brine, as
shown in figure 4a. At the end of each secondary flood, tertiary injection of distilled water
was performed over several PVs, but never exhibited any incremental oil. Although the
viscosity of crude A is measured at ambient temperature only: 7,2 cP, the usual
extrapolation at core test temperature: 100C suggest that the viscosity ratio may partially
explain the very poor performance of tertiary LSWI. Moreover, the evolution of ionic
composition during tertiary LSWI shows (fig 4b) that the composition of injected water
(deionized water) is not reached at the outlet after more than 5 PV injected, on these 15 cm
long cores. Reported information does not allow to separate core heterogeneity from added
dispersion by residual oil, but clearly shows strong dispersion at SorwHS. Strong dispersion
in brine phase at SorwHS delays the decrease towards very low salinity all over the core. It
suggests that the combined effect of unfavorable viscosity ratio and strong dispersion at
Sorw might totally suppress positive effect of tertiary LSWI, even when secondary floods
put in evidence very significant benefits.
Which factors may have large effects on brine/brine displacement at Sorw?
Single phase dispersion might already be large. Sample heterogeneity along flow axis
would be a major cause. Rock characteristics may also have strong effects: 15 fold
differences in single phase dispersion coefficient between various outcrop sandstones at
the same velocity have been reported [13]. Berea sandstone (300 mD) exhibited the
sharpest displacement front and the least amount of tailing. Wide pore-size distributions in
sandstones lead to higher dispersion coefficients [6].

SCA2015-006

7/12

Figure 4: a) Comparison between HS, LSWI in secondary and tertiary modes; b) evolution of ionic
concentration during tertiary LWSI ; from Nasralla, 2011

However, single phase tracer tests are almost never reported in experimental studies
devoted to tertiary LSWI. This is surprising, particularly when authors specify that samples
are taken parallel to bedding.
Core lengths might also have a significant impact, as dispersive flows depend on the ratio
between the dispersion length and the length of the system. In that respect, short cores (5,
7,5 cm) are most at risk. Assuming a dispersion length of 6-7mm at SorwHS, consistently
with observations made in [50] on mildly wet sandstone, this ratio would approach or
exceed 0,1 for 5 and 7,5 cm long cores. It might be even worse when HSWI is carried out
at unfavorable viscosity ratios which trigger water fingers. This is a significant concern for
a large number [10, 16, 19, 31, 32, 35, 41, 49, 52, 53, 55, 56] of tertiary LSWI studies. On
the other hand, too short core lengths cannot be systematically called to explain the
absence of any tertiary response during LSWI. Shehata [40] performed tertiary LSWI,
observed significant benefits in secondary mode, but no incremental oil during tertiary
LSWI on 50 cm long cores.
Others parameters may also have an effect on the evolution of salinity during laboratory
tertiary LSWI. Difference in viscosity between secondary HSW and tertiary LSW may
reach unfavorable values of viscosity ratios up to 1,5-1,6. This range of viscosity ratio
may trigger fingering in the water phase during tertiary LSWI and the salinity profile might
be smeared. Flow velocity might also be important, as dispersion coefficient may increase
as interstitial velocity increases, both at Sw=1 [4] and at Sorw [36].
Effect of Wettability Modifications Between HSWI and LSWI
Published tertiary LSWI results are very rarely accompanied by secondary HSWI or
wettability tests. None is reporting a full Amott Harvey or USBM test. In the absence of
wettability tests, it is often assumed that LSWI shifts systematically wettability towards
more water wet, based on observations of increased recovery during LSW spontaneous
imbibition compared to HSW. Evolution of Krs between LSW and HSW is then guessed
according to this assumption. In fact, this spontaneous recovery might be sometimes
misinterpreted as demonstrated below using Zhous results [57]. Zhou changed the

SCA2015-006

8/12

wettability of Berea sandstone by changing the length of aging time at 88C with Prudhoe
Bay crude oil. Three data sets are reported, showing the same behavior as shown in figures
5a, b. Water imbibition rate decreases, but final oil recovery by imbibition increases as the
ageing time increases from 0 to 24 hours. In the same way, waterflood oil recovery
increases as the ageing time increases from 0 to 24 hours. It is obvious that outcrop
samples cannot become more water wet by increasing ageing time with reservoir oil... It
clearly means that increase in recovery by spontaneous water imbibition does not always
mean more water-wet. Zhous experimental results were successfully simulated using
distribution of contact angles [0-89] in [14]. Consequently the very similar results of
LSWI by Tang [46, 48, 49], within the water-wet domain, might have been misinterpreted,
and might not represent a systematic shift towards water wet. Moreover, other sources of
information, such as adhesion maps [7] or flotation tests clearly show increased oil
adhesion, when salinity decreases. This topic is addressed in a companion paper [12].

Figure 5: evolution of spontaneous (left) and forced (right) oil recoveries as a function of ageing time
(from Zhou, 2000)

Then, in the absence of full wettability test or of any comparison between secondary LSWI
and HSWI, there is a doubt about the direction of change of wettability when the rock is
contacted by LSWI. Figure 6 shows simulations results of incremental recovery during
tertiary LSWI when the viscosity ratio equals 2. Two cases are compared: LSWI might
change wettability towards either more water wet (SorwHS=0,3; SorwLS=0,2; Krw(SorwHS)
= Krw(SorwLS) =0,3), as often assumed in the absence of both secondary LSWI Krs and
wettability tests, or less water-wet, (SorwHS=0,3; SorwLS=0,2; Krw(SorwHS)=0.3
Krw(SorwLS) =0,6) as suggested by the general trend of adhesion maps. In such a case
Krw(Sorw)LS can be much larger than Krw(Sorw)HS. Capillary pressure is not taken into
account.
The reduction in Sorw during secondary floods is 10 saturation units in both cases. Even
with a very favorable viscosity ratio, tertiary incremental is delayed and poor if LSWI
shifts wettability towards less water wet. Note that situation is totally consistent with
experimental results on secondary waterfloods [59]: when the initial wettability is within
the water-wet domain, a shift towards less water wet increases secondary recovery, but

SCA2015-006

9/12

may lead to poor recovery in tertiary LSWI. Moreover, shifting towards less water wet
may increase dispersion within the water phase, as observed in [36, 50].

Figure 6: simulations of tertiary LSWI as a function of direction of wettability change

CONCLUSIONS AND WAY FORWARD

Conditions of existence of a double saturation shock, effects of dispersion in water phase at
SorwHS, and uncertainty in the direction of wettability change may explain a lot of
experimental studies, where tertiary LSWI performs poorly. Counter examples can also be
produced for any single proposed mechanism, but their combination deserves further
attention. The absence of relevant experimental data in published studies is the main
difficulty when trying to decipher the causes of poor tertiary LSWI results. An ideal set of
experiments should incorporate:
A brine/brine tracer test at Sw=1 on each rock sample used for corefloods
Secondary HS and LS water floods, including evolution of recovery, differential
pressure, in situ saturations, as well as salinities after breakthrough
Full Amott Harvey wettability tests for both HS and LS brines. The negative part of the
forced imbibition capillary pressure curves can be derived from the forced
displacements if carried out by centrifugation.
Tertiary LS water flood including evolution of recovery, differential pressure, as well
as salinities and ionic compositions
Fluid characteristics at test temperature and pressure, including the evolution of wateroil interfacial tension (IFT) as a function of salinity. This last item has received very
little attention. However, several studies [9, 47, 54, 25, 2] reported decrease in IFT as
salinity decreases. Based on conventional understanding of oil mobilization, water-oil
IFT must be reduced by several orders of magnitude to mobilize oil and the reductions

SCA2015-006

10/12

of IFT with salinity are often deemed too small to have any significant effect. On the
other hand, when successful, tertiary LSWI illustrates that oil can be mobilized at very
low capillary numbers. In that respect, small variations of IFT should deserve attention.

ACKNOWLEDGEMENTS
The author would like to thank TOTAL for permission to publish this paper.

REFERENCES
1. Alotaibi M, Azmy R, Nasr El Din H: Wettability studies using low salinity water in
sandstone reservoirs OTC20718, 2010
2. AlShaikh M, Mahadevan J: Impact of brine composition on carbonate wettability: a
sensitivity study SPE 172187-MS, April 2014
3. Austad T, Rezaeidoust A, Puntervold T: Chemical mechanism of low salinity water
flooding in sandstone reservoirs SPE129767 2010
4. Bear J: Dynamics of Fluids in Porous Media1972 American Elsevier. NY.
5. Boussour S, Cissokho M, Cordier P, Bertin,H, Hamon G: Oil recovery by low salinity
brine injection: Laboratory results on outcrop and reservoir cores SPE 124277, 2009
6. Bretz R, Specter R, Orr F: Effect of pore structure on miscible displacement in laboratory
cores SPERE, August 1988 p 857-866
7. Buckley J, Morrow N: Characterization of crude oil wetting behavior by adhesion tests
SPE 20263, 1990
8. Buckley J, Fan T: Crude oil/brine interfacial tensions SCA 2005-01, International
symposium of the Society of Core Analysts 2005
9. Cai B, Yang J, Guo T: interfacial tension of hydrocarbon +water/brine systems under high
pressure Journal of chemical and engineering data 41 (3) p 493-496
10. Callegaro, C, Bartosek M, Masserano F, Nobili M, et al: Opportunity of enhanced oil
recovery by low salinity water injection: from experimental work to simulation study up to
field proposal SPE 164827, 2013
11. Cissokho, M, Boussour S, Cordier P, Bertin H, Hamon G: Low salinity oil recovery on
Clayey sandstone: Experimental study PETROPHYSICS, VOL. 51, N. 5 (Oct 2010); p
305-313
12. Didier M: Contradictory trends for Smart Water Injection Method: role of pH and salinity
from sand/oil/brine adhesion maps SCA 2015 A14
13. Donaldson E, Kendall R, Manning F: Dispersion and tortuosity in sandstones SPE 6190,
1976
14. Dixit A, McDougall S, Sorbie K: Pore scale modeling of wettability effects and their
influence on oil recovery SPE/DOE 35451, 1996
15. Fjelde I, Asen S, Omekeh A: Low salinity water flooding experiments and interpretation
by simulations SPE 154142 2012
16. Gamage P, Thyne G: Comparison of oil recovery by low salinity waterflooding in
secondary and tertiary modes SPE147375, 2011
17. Hadia N, Lehne H, Kumar K, Selboe K, Stensen J, Torsaeter O: Laboratory investigation
of low salinity waterflooding on reservoir rock samples from the Froy field SPE 141114,
2011
18. Hadia N, Ashraf A, : Laboratory investigations on effects of initial wettabilities on
performance of low salinity waterflooding JPSE 105 (2013) 18-25

SCA2015-006

11/12

19. Hamouda, A, Valderhaug O, Munaev, R, Stangeland H: Possible mechanisms for oil recovery from
chalk and sandstone rock by low salinity water SPE 169885, 2014
20. Lager, A., Webb, K.J., Black C.J: Impact of brine chemistry on oil recovery 14 th
symposium on Improved oil recovery Cairo, Egypt, 22-24 April 2007
21. Ligthelm D, Gronsveld J, Hofman, J, Brussee N, et al: Novel waterflooding strategy by
manipulation of injection brine composition SPE 119835, 2009
22. Loahardjo N, Xie X, Yin P, Morrow N: Low salinity waterflooding of a reservoir rock
SCA 2007-29, 2007
23. McGuire, PL, Chatham, JR,: Low Salinity Oil recovery: An Exciting New Opportunity
for Alaskas North Slope SPE 93903, 2005 Western Regional Meeting, Irvine, April 2005
24. Meyers, K.; Salter, S: Concepts pertaining to reservoir Pretreatment for chemical
flooding SPE/DOE 12696, April 1984
25. Moeini F, Hemmati-Sarapardeh A., Ghazanfari M, et al : Toward mechanistic
understanding of heavy crude oil/brine interfacial tension: The roles of salinity,
temperature and pressure Fluid Phase Equilibria 375 (2014) 191200
26. Mohammadi H, Jerauld G: Mechanistic modeling of the benefit of combining polymer
with low salinity water for enhanced oil recovery SPE 153161, 2012
27. Morrow N, Buckley J: Improved oil recovery by low salinity water flooding JPT, May
2011 p 106-112
28. Nasralla R, Alotaibi M, Nasr el din H: Efficiency of oil recovery by low salinity water
flooding in sandstone reservoirs SPE 144602, 2011
29. Nasralla R, Nasr el Din H: Double layer expansin: is it a primary mechanism of
improved oil recovery by low salinity wter flooding? SPE 154334, 2012
30. Pope G: The application of fractional flow theory to enhanced oil recovery SPEJ June
1980, p191-205
31. Pu H, Xie X, Yin P, Morrow N: Application of coalbed methane water to oil recovery by
low salinity waterflooding SPE 113410, 2008
32. Pu H, Xie X, Yin P, Morrow N: Low salinity waterflooding and mineral dissolution SPE
134042, 2010
33. RezaeiDoust A, Puntervold T, Austad T: A discussion of the low salinity EOR potential
for a North Sea sandstone field SPE 134459, 2010
34. Rivet S, Lake L, Pope G: A coreflood investigation of low salinity enhanced oil
recovery SPE 134297, 2010
35. Romero M, Gamage P, Jiang H, Chopping C, Thyne G: Study of low salinity
waterflooding for single and two phase experiments in Berea sandstone cores JPSE 110
(2013) p 149-154
36. Salter, S.; Mohanty, K: Multiphase flow in porous media: 1: Macrocopic observations and
modelling SPE 11017, 1982
37. Sandengen K, Tweheyo M, Raphaug M, et al: Experimental evidence of low salinity
waterflooding yielding a more oil-wet behavior SCA2011-16
38. Sohrabi M, Mahzari P, Farzaneh S, Mills J, Tsolis P, et al: Novel insights into
mechanisms of oil recovery by low salinity water injection SPE 172778, 2015
39. Sandengen K, Tweheyo M, Raphaug M, et al: Experimental evidence of low salinity
water flooding yielding a more oil wet behavior SCA 2011-16
40. Shehata A, Nasr el Din H: Role of sandstone mineral compositions and rock quality on
the performance of low salinity waterflooding IPTC 18176, 2014
41. Shiran B., Skauge A.: Wettability and oil recovery by low salinity injection SPE
155651, 2012

SCA2015-006

12/12

42. Skrettingland K, Holt T, Tweheyo M: Snorre low salinity water injection core flooding
experiments and single well field pilot SPE 129877, 2010
43. Spildo K, Johannessen A, Skauge A: Low salinity waterflood at reduced capillarity
SPE154236, 2012
44. Suijkerbuijk B, Hofman J, Ligthelm D, Romanuka J et al: Fundamental investigations into
wettability and low salinity flooding by parameter isolation SPE 154204, 2012
45. Suijkerbuijk B, Sorop T, Parker A et al.: Low salinity waterflooding at West Salym :
Laboratory experiments and field forecasts SPE 169102, 2014
46. Tang G, Morrow N: Salinity, temperature, oil composition, and oil recovery by
waterflooding SPERE, November 1997, p269-276
47. Tang, G : Brine Composition and waterflood recovery for selected crude Oil/Brine/Rock
systems Ph.D Dpt of Chemical and Petroleum Engineering UW, May 1998
48. Tang G, Morrow N: Influence of brine composition and fines migration on crude
oil/brine/rock interactions and oil recovery 1999, JPSE 24 (2-4) p99-111
49. Tang, G, Morrow N: Oil recovery by waterflooding and imbibitions invading brine cation
valency and salinity SCA 9911, 1999
50. Wang F: Effect of wettability alteration on water/oil relative permeability, dispersion, and
flowable saturation in Porous Media SPERE, May 1988,
51. Webb K, Lager A , Black C: Comparison of high/low salinity water/oil relative
permeability SCA 2008-39, 2008
52. Winoto W, Loahardjo N, Xie X, Morrow N: Secondary and tertiary recovery of crude oil
from outcrop and reservoir rocks by low salinity waterflooding SPE 154209, 2012
53. Winoto, Loahardjo N, Morrow N: Assessment of oil recovery by low salinity
waterflooding from laboratory tests SPE 169886, 2014
54. Yousef A, Al Kaabi A, Al Jawfi M: Laboratory investigation of the impact of injection
water salinity and ionic content on oil recovery from carbonate reservoirs SPE Reservoir
Evaluation & Engineering October 2011 578-593
55. Zhang Y Morrow N: Comparison of secondary and tertiary recovery with change in
injection brine composition for crude oil/sandstone combinations SPE 99757, 2006
56. Zhang Y Xie X Morrow N: waterflood performance by injection of brine with different
salinity for reservoir cores SPE 109849, 2007
57. Zhou W, Morrow N, Ma S: Interrelationship of wettability, initial water saturation, aging
time, and oil recovery by spontaneous imbibition and waterflooding SPEJ 5 (2) June
2000, p199-207
58. Wickramathilaka, S, Morrow N, Howard J : Effect of salinity on oil recovery by
spontaneous imbibition SCA 2010-12, 2010
59. Jadhunandan P, Morrow N : Effect of wettability on waterflood recovery for crude oil
brine rock systems SPERE, February 1995, pp40-46

SCA2015-007

1/12

THE FATE OF OIL CLUSTERS DURING FRACTIONAL

FLOW: TRAJECTORIES IN THE SATURATIONCAPILLARY NUMBER SPACE
M. Rcker2,1, S.Berg1, R. Armstrong3,1, A. Georgiadis1,5, H. Ott1,6, L. Simon2, F.
Enzmann2, M. Kersten2, S. de With4,1
1
Shell Global Solutions International, Kesslerpark 1, 2288 GS Rijswijk, Netherlands
2
Geosciences Institute, Johannes-Gutenberg University, 55099 Mainz, Germany
3
School of Petroleum Engineering, University of New South Wales, New South Wales,
Sydney, Australia
4
Technical University Delft, Stevinweg 1, 2628 CN Delft, Netherlands
5
also at: Department of Chemical Engineering, Imperial College London, SW7 2AZ UK
6
also at: Department of Earth Science and Engineering, Imperial College London, SW7
2AZ UK
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Fractional flow has been studied at the pore scale under dynamic flow conditions by
using fast synchrotron-based X-ray computed micro-tomography. The pore-scale flow
regimes have been mapped in a phase diagram where the regimes of connected
pathway flow and ganglion dynamics are characterized by fractional flow and capillary
number. The regimes are identified from the respective pore scale dynamics that can be
conveniently characterized by using a saturation-(cluster-based) capillary number
diagram. Therein connected pathway flow is represented by a fixed point (because all
parameters are constant over time) and in ganglion dynamic regime the oil clusters follow
trajectories because saturation and cluster length are changing over time. Ganglion
dynamics is composed of breakup and coalescence processes. During coalescence
processes, both cluster volume and length increases, i.e. clusters move up the
trajectory. During break-up processes, on the other hand, both properties decrease and
clusters move down the trajectory.
Ganglion dynamics occurred even though the (cluster-based) capillary number of the
average flow field was at least two orders of magnitude smaller than unity, i.e. the
average flow field indicates capillary-dominated regime. However viscous mobilization
can also be triggered by more complex break-up and coalescence processes that have
much higher local flow velocities than the average flow field suggests. Most situations
encountered are a combination of connected pathway flow and ganglion dynamics, where
a combination of viscous and capillary-driven processes accounts for the net transport of
oil. Static simulation approaches are not capable of capturing such regimes, as they
require connected pathway flow.

SCA2015-007

2/12

INTRODUCTION
During the recovery of oil from an underground reservoir by water flooding, the nonwetting oil phase is displaced by the wetting water phase. Corresponding to the Buckley
Leverett equation [1], the two fluids flow simultaneously through narrow pore space once
the saturation front has passed. In order to determine the relative permeability of this
simultaneous flow, commonly steady-state core flooding experiments (e.g., experiments
on 5cm-long rock samples) are conducted [2]. However, core flood experiments often
show fluctuations of parameters such as saturation, pressure and resistivity depending on
the fractional flow (Figure 1). The reason for these fluctuations is not fully clear yet, but
may be related to different flow regimes [3, 4, 5].
Fractional flow

Fluctuations

1.00
0.90

Ganglion
dynamics ?

Normalized dimension

0.80
0.70
0.60

0.50

Saturation

0.40

Pressure

0.30

0.20

Resistivity

0.10
0.00
0

1000

2000

3000

4000

5000

6000

7000

0.48 < Sw < 0.70

Pore volume injected

Figure 1 Results of a core flood experiment measured on a 5 cm rock sample (right) show a variation of
fluctuations in different parameters such as saturation, pressure, and resistivity depending on the fractional
flow (left).

On the pore scale, Avraam and Payatakes [3] observed three types of flow regime, which
are (i) connected pathway flow, (ii) drop traffic flow, and (iii) ganglion dynamics. During
connected pathway flow, the non-wetting and the wetting phases move mainly through
their respective connected networks [3]. The drop traffic regime describes the flow of oil
through disconnected droplets smaller than the diameter of the pore throats [3]. These are
formed by vigorous break-up of the oil ganglia and are stabilized by the flow of the
surrounding water, which prevents the growing of ganglia via coalescence [3]. The most
complex flow regime is the ganglion dynamics, whereby oil is disconnected into ganglia
with sizes larger than a pore. These ganglia move or get immobilized in narrow passages,
they break into smaller ganglia and get trapped even more easily, or they collide with
other ganglia, grow and get remobilized [3]. The thus prevailing flow regime depends on
different parameters such as viscosity, wettability, surface tension, flow rate ratio,
capillary number, etc. [4]. For specific fluids in a specific porous medium most
parameters are constant and the relative permeability appears to be a function of capillary
number and flow rate ratio [8].

SCA2015-007

3/12

While most previous studies performed on dynamic two-phase flow are either based on
2D models [3,9] or on 3D but indirectly measured [4, 5] or on simulations [6, 7], in this
study we image directly the 3D development of clusters in real time by fast synchrotron
based X-ray CT. Recent studies using this technique allowed the first insights into porescale processes during drainage and imbibition [10, 11, 12, 13] such as Haines jumps [14,
15, 16], coalescence [17,18] and snap-off [18, 19]. In this study we analysed ganglion
dynamics by co-injection of two fluids at different ratios and velocities in porous media
in order to determine the prevailing flow regime.

MATERIALS AND METHODS

For this experiment, n-decane was used as nonwetting phase and brine as wetting phase.
A strongly water-wet sintered glass sample (4 mm diameter, 20 mm length, embedded
into a polycarbonate tube by heat-shrinking) with a porosity of 35% and permeability of
222 D was mounted at the top of a flow cell. The flow cell was specially designed for
fractional flow experiments and contained two remotely controlled micro piston pumps,
which enables continuous rotation of the sample and constant data acquisition in the field
of view during the experiment. A full 3D image with a voxel size of 2.2 m was obtained
in 1 min. The measurements were performed at a fast synchrotron-based X-ray computed
microtomography facility (TOMCAT beamline, Swiss Light Source, Paul Scherrer
Institute). Details on the flow cell and the experimental settings were described
previously [20]. The reconstructed CT images were filtered, segmented and processed
with the software package AVIZO 8 (Visualization Science group).
Subsequently, the resulting binarized images containing the porous medium and the oil
and water phases were used for further analyses, e.g., for estimation of the relative
permeability. The single-phase flow simulation by the software package GeoDICT
(Math2Market) was used in the Navier-Stokes mode to calculate the permeability through
the connected portion of the oil- and water-phase for each time step.
Furthermore, all flow events occurring during the experiment were determined and
separated into oil-filling and water-filling events. While the former event type describes a
situation in which pores drained from water are filled by oil, the latter type corresponds to
oil being replaced by water. The events are determined by generating a differential image
of two consecutive 3D binary images containing only the oil phase. The differential
image gives the positon and size of both water- or oil-filling events. A single event is
defined as a spatially separated fluid change at a specific position. The event size is
defined as the volume of the spatially separated change. To reduce effects due to noise,
small events (< (21 m)3 corresponding to 10 voxels in each direction) are ignored.
In addition, the development of clusters was observed. To follow a chosen cluster
through the different time steps, the space occupied by this cluster in the first time step is
compared to the same space in the next time step. All clusters inside this space or
connected to this space are then used as a mask for the next time step. This procedure is

SCA2015-007

4/12

iterated over all measured time steps, enabling clusters to be tracked over time and
statistics to be obtained.

RESULTS AND DISCUSSION

The flow regime depends on the fractional flow and the capillary number. The fractional
flow is given by the injection rate ratio of each fluid to the total fluid, while the capillary
number in this study is determined by the macroscopic cluster-based definition [9, 24]:
=

,
,

(1)

Flow rate in m/s

Injection rate in l/min

For a sample diameter= 4mm

With the cluster length ( ), radius of a pore throat ( ), porosity (), viscosity of the
wetting phase ( ), interstitial velocity of the wetting phase ( ), and interfacial tension
(, ). The radius of the peak pore size of the pore size distribution calculated by the
software code GeoDICT was used as pore throat radius. The velocity is derived from the
injection rate and represents the average flow field, which is often used to characterize a
was calculated by averaging lcl of all clusters as
flow regime. An overall
described in [24]. The fractional flow-capillary number diagram (Figure 2) shows how
the experiments were conducted, starting at a high water fraction and low total injection
. The total injection rate increased
rate of 4x10-6 m/s corresponding to a low
-4
stepwise up to 4x10 m/s before changing to a lower oil fraction. For all experiments
< 1 indicating capillary dominated flow regime. However, since local velocities

may deviate from the overall velocity, the transition from capillary dominated to viscous
dominated regimes is rather smooth.

Camacro

Cluster length in m
area covered by
experiments
Camacro=1
crucial for viscous
Oil mobilization

Figure 2 The experiments started with a high water fraction (80% water) and a low flow rate (4x10 -6 m/s).
The total injection rate was increased stepwise up to 4x10 -4 m/s before repeating the experiments at a lower
oil fraction (left). In all experiments, a
< 1 corresponding to a capillary-dominated system (right).

In our experiments, we captured the transient regime when changing flow rates followed
by the quasi steady-state regime. This transition could be recognized by the initial change
in saturation. At low flow rates, the transition took longer and, therefore, was ideal for

SCA2015-007

5/12

observing growing and decreasing clusters. An increase in oil-saturation with a high

frequency of oil-filling events appeared as Haines jumps and coalescence, while a
decrease in saturation was dominated by snap-offs and led to disconnected clusters. For
this reason, initial drainage led to a high connectivity, while in general a low saturation
corresponded to low connectivity.
Increase
in
flow rate

pressure [mbar]

100

v=
4x10-5 m/s

90

Decrease
in
flow rate

quasi steady state

Ganglion
dynamic

v=
4x10-4 m/s

v=
0 m/s

80
70
60
0

10

Saturation in %

100

12

total oil
connected oil

80
60
40

20
0
0

10

water filling events

oil filling events

800

N(events)

12

600

400
200
0
0

10

12

time [min]

Figure 3 Pressure (1), saturation (2), and event statistics (3) for an experiment at fw = 0.5 and a total flow
rate of 4x10-4 m/s. Transitions between different fractional flows / flow rates where the target conditions
had not been fully reached are marked by the grey bars. In these transition regimes, we observed natural
changes in pressure, saturation, and pore-filling events, because flow conditions have been changed
externally. However, in the quasi steady-state regime there are also pressure data and saturation
fluctuations, which coincide with pore-scale water- and oil-filling events as signature of ganglion dynamics
behavior.

The term quasi steady-state is chosen to indicate that fluctuations and events may still
occur even though the average saturation has adjusted to the new injection rate. These
fluctuations may be indicative of ganglion dynamics. At quasi steady-state conditions the
saturation remained mostly stable and the number of events was smaller. Nevertheless,
some fluctuations could be observed at water fraction fw = 0.5 and high total flow rates
(4x10-4 m/s, Figure 3). A sudden pressure drop followed by an increase occurred between
the sixth and eighth minute (Figure 3.1). In the same time period, the oil volume dropped
(Figure 3.2) and the number of water- and oil-filling events increased (Figure 3.3).

SCA2015-007

6/12

This observation is very similar in terms of saturation and pressure fluctuation to the
results of core flood experiments at fw = 0.5 and 5 cm long rock samples.
Growing and Shrinking Clusters
We tracked individual clusters to gain further insights into ganglion dynamics. In Figure
4, the development of a cluster that grows over time is shown.
1.

time: 5min

time: 6min

1000m

time: 4min

2.

time: 8min

time: 9min

coalescence
100

V(cluster) / V(pores)

3.

selected ROI

coalescence (2)
10-1
time [min]
1
2
3
4
5
6
7
8
9

growing
Cluster (1)
10-2
1x10-4

2x10-4

Figure 4 Tracking a growing cluster at fw = 0.8 and v=4x106 m/s: In the first phase (1), the cluster
continuously increases in volume. For this reason, the corresponding cluster-trajectory in the cluster
volume-capillary number (V-Ca) space moves slowly towards high volume and capillary number, before it
suddenly jumps (3) due to a coalescence with another cluster (2).

The cluster grows first through pathways outside the field of view by invading adjacent
pores (Figure 4.1) before it coalesces with another cluster (Figure 4.2). In the volumeCamacro space, this development is visible in a trajectory moving from low V-Camacro to
high V-Camacro regime.
A shrinking cluster behaves differently, because in addition to a decrease in volume
due to water film swelling [11] (and oil moving out the field of view), we also observe
fragmentation. As shown in Figure 5, the cluster first decreases until snap-off events lead
to fragmentation. The evolution of the saturation distribution of a shrinking cluster is

SCA2015-007

7/12

shown in Figure 5.1. The initial cluster A snaps off apart into two smaller clusters A and
B. After redistribution of oil the volume apparently further decreases which is an imaging
artifact from moving oil due to snap-off (few milliseconds [22]) during the acquisition
time of a tomogram (60 s). Then cluster B snaps off into B and C. The corresponding
trajectory in the V-Camacro space is in this case moving, vice versa, towards low volume
and Camacro.
1.

1000m

A snap off

A
B

time: 5min

time: 6min

snap off

B
C

time: 7min

time: 8min

10-2

2.

V(cluster) / V(pores)

snap off
10-3

A A
B
10-4

10-5
5x10-6

snap off

10-5

B
shrinking
B
cluster
time [min]
3
4
5
6
7
8
5x10-5

Figure 5 Tracking a decreasing cluster at fw = 0.5 and high Ca. First, the cluster slowly decreases due to
film swelling, before snap-off events fragment it into smaller ones, which further decrease in size and break
apart. The clusters shown in (1.) correspond to the clusters shown in the trajectory (2.), where the single
new appearing fragments are named A-C to indicate which fragment in later time steps is represented by
which point in the diagram.

The cluster trajectories in the V-Camacro space move upwards when they are growing by
coalescence or Haines jumps, and downwards while they snap off during shrinkage.
However, all clusters have to be taken into account in order to determine the flow regime.

SCA2015-007

8/12

10-0

10-2

limited by sample size

V(cluster) / V(pores)

( )

2.5

10-4

max.
Camacro

10-6

10-6

10-5

10-4

time [min]
1
2
3
4
5
6
7
8
9
10
11
12
10-3

Figure 6 All clusters for each time step at fw = 0.5 and v = 4x10-4 m/s follow a power-law distribution with
an exponent 2.5, which gives information about the shape of the clusters. The capillary number of the
largest clusters is limited by the sample size leading to a bias.

Figure 6 shows the distribution of all clusters for all time steps at fw = 0.5 and v = 4x10-4
m/s. The distribution of the clusters appears to follow a power-law like behavior,
() ( ) , with an average exponent 2.5. Since , the
exponent gives information regarding the shape of the cluster. The exponent of = 2.5
indicates a cluster shape between an area-filling shape, where volume scales as (length)2,
and a volume-filling shape (such as spheres) where volume scales as (length)3
Furthermore, the exponent corresponds to the fractal dimension of invasion percolation
and indicates that the volume of clusters during fractional flow scales within percolation
universality class. The length of the largest cluster is restricted by the length of the
sample and/or field of view and, therefore, the largest clusters do not follow the general
distribution. Once a cluster percolates from bottom to the top of our field of view, we can
observe volume increase but cannot observe increase in length. This finite size effect is
displayed as the maximum in all following figures.
Pathway flow and Ganglion dynamics in the S-Ca-diagram
So far only individual clusters have been analyzed, while in the following we aim to
characterize the complete flow regime involving all clusters. Therefore we consider
, taking
average properties like saturation but also the averaged capillary number
into account that each cluster has a different length and volume.
are shown in Figure 7. When ganglion
Trajectories based on the average
dynamics occurs where the connected oil (grey) gets disconnected (yellow) and

reconnected (Figure 7.1), the trajectory moves up and down in saturation and
(Figure 7.2). The reason for this is that when more clusters are disconnected, the
saturation decreases, since the narrow pore throats connecting the oil phase are filled by
water instead of oil. Moreover, the volume of the disconnected clusters and their

SCA2015-007

9/12

. When
corresponding length becomes also smaller, which leads to a decrease in

the oil clusters become reconnected, the saturation and

increases again. Since

is limited due to the sample size, the trajectory is deflected at high saturations.
On the other hand, the trajectory in connected pathway flow does not show any changes
(Figure 7.4). Since the oil and water are moving through independent pathways, the
cluster shape is not changing as shown in Figure 7.3.
100

2.

1.

90

time: 8min

time: 6min

Disconnection

Oil-Saturation [%]

Ganglion dynamic

80

Reconnection

70

time [min]

60
50

movement
up and down

40
30
20

max.
Camacro

10

fw=0.5
Camicro=1.57x10-5

0
5x10-4

1
2
3
4
5
6
7
8
9
10
11
12

10-3

5x10-3

time: 7min

3.

4.

100
90

time: 7min

time: 3min

Oil-Saturation [%]

Connected pathway flow

80

no changes

70
60
50
40
30
20

max.
Camacro

10

fw=0.2
Camicro=6.31x10-6

0
5x10-4
time: 5min

10-3

time [min]
1
2
3
4
5
6
7
8
9
10
11
12
5x10-3

Figure 7 It depends on fw whether ganglion dynamics (fw = 0.5) or pathway flow (fw = 0.2) occurs. While at
fw = 0.5 the oil gets disconnected (yellow) and reconnected (grey) during quasi steady-state conditions (1),
the oil distribution remains constant at fw = 0.2. The S- diagram shows that the clusters at fw = 0.5
move up and down the trajectory (2), while the clusters at fw = 0.2 remain at one point (4). However, in both
cases
is close to the maximal possible Camacro, which is limited by the sample size.

Contribution to the relative permeability

diagram correlates with the
Since the ganglion dynamics observed on the S-
pressure and saturation fluctuations, all observations in core flood experiments appearing
at the same water fraction (fw = 0.5) may be also associated with ganglion dynamics.
However, at least at the pore scale, the impact on relative permeability is not as clear so
far. In Figure 8, the development of the relative permeability through the continuous oil
and water phase with time is shown, excluding the contribution of disconnected clusters
via ganglion dynamics. Since the injection rate and water fraction fw was constant over
time, the deflection in relative permeability between the sixth and eighth minutes
represents the contribution of ganglion dynamics.

relative permeability : kr

SCA2015-007

10/12

0.5
water
oil

0.4
0.3
0.2
0.1
0
0

10

12

time [min]
Figure 8 Relative permeability at fw = 0.5 and high Ca during time. At this high flow rate, the quasi steadystate conditions are reached already after one minute. As the saturation (Figure 2) and the relative
permeability of the oil and the water remains stable, a disconnection and reconnection (Figure 7)
corresponding to a moving ganglion can be observed, whereby a disconnection causes an increase of the
relative permeability of the water and a decrease for oil.

CONCLUSIONS
Flow regimes during fractional flow have been characterized by imaging the pore-scale
fluid distribution using synchrotron-based fast X-ray computed microtomography. We
observed regimes of connected pathway flow but also ganglion dynamics regime. These
flow regimes can be categorized in the fractional flow vs. capillary number (fw-Ca) space.
In order to construct a phase diagram indicating which flow regime prevails at which
conditions, we sampled this space by conducting steady-state CT flow experiments,
which in addition to pressure and saturation data allows us to identify the respective flow
regimes from the pore-scale displacements and oil cluster (ganglion) movement.
Oil clusters follow distinct trajectories in the saturation-Ca space, which is caused by
pore-scale displacement events. Pore-scale processes such as cluster coalescence and
growing cause an increase in cluster volume and length so that the clusters move up a
trajectory in a saturation and capillary number diagram. In break-up processes, both
volume and length decreases and clusters move down the trajectory. Averaging over all
clusters of a measurement for each time step allows us to characterize the flow regime
involving all clusters in our field of view. Since almost no events occur at connected
pathway flow, no changes can be observed at the saturation-Ca diagram. On the other
hand, ganglion dynamics cause a movement both up and down the trajectory.
These pore-scale processes clearly affect two-phase flow at larger scales. Fluctuations in
saturation and pressure, as observed in Darcy-scale steady-state relative permeability
experiments, have very similar characteristics to those observed in our CT flow
experiments. That clearly indicates that the pressure and saturation fluctuations observed
on a 5 cm-long rock sample at intermediate fractional flow and saturation ranges are
likely to be related to ganglion dynamics. The different flow regimes contribute
differently to the total flow. Due to break-up and coalescence in a ganglion dynamics

SCA2015-007

11/12

flow regime, the connectivity is changing what affects the flow and causes change in
relative permeability. This has to be considered for describing two-phase flow by
numerical models.

ACKNOWLEDGEMENTS
We would like to acknowledge Kevin Mader and Marco Stampanoni (Tomcat beamline
at the Swiss Light Source of Paul Scherrer Institute, Villigen, Switzerland). We thank
Alex Schwing and Rob Neiteler for the design of the flow set-up and instrumentation, Ab
Coorn, Fons Marcelis and Niels Brussee for sample preparation, Hilbert van der Linde,
Niels Brussee and Sebastiaan Pieterse for the steady-state relative permeability
measurements, and Axel Makurat for helpful discussions and leadership support. We
gratefully acknowledge Shell Global Solutions International B.V. for permission to
publish this work.

REFERENCES
1. Buckley, S. and Leverett, M., Mechanism of fluid displacement in sands,
Transactions of the AIME (1942) 146.01, 107116.
2. Berg, S., Cense, A. W., Hofman, J. P., Smits, R. M. M., Two-Phase Flow in Porous
Media with Slip Boundary Condition, Transp Porous Med, (2007) 74:275292.
3. Avraam, D. G. and Payatakes, A. C., Flow regimes and relative permeabilities
during steady-state two-phase flow in porous media, Journal of Fluid Mechanics
(1995) 293, 207236.
4. Tsakiroglou, C. D., Aggelopoulos, C. A., Terzi, K., Avraam, D. G., Valavanides, M.,
Explicit correlation of the steady-state-two phase relative permeability functions of
porous media with the local flow rate, SCA2014-041.
5. Datta, S. S., Dupin, J. B. and Weitz, D. A., Fluid breakup during simultaneous twophase flow through a three-dimensional porous medium. Physics of Fluids (2014)
26(6), 062004.
6. Raeini, A. Q., Blunt, M. J., Bijeljic, B, Direct simulations of two-phase flow on
micro-CT images of porous media and upscaling of pore-scale forces, Advances in
Water Resources (2014) 74, 116-126.
7. Lopez, O., Mock, A., Skretting, J., Petersen, E.B.Jr, ren, P.E., Rustad, A.B.,
Investigation into the reliability of predictive pore-scale modeling for siliciclastic
reservoir rocks, SCA2010-41.
8. Valavanides, M. S. (2014), Operational efficiency map and flow characterization for
steady state two-phase flow in porous media, SCA2014-047.
9. Armstrong, R. T. and Berg, S., Interfacial velocities and capillary pressure gradients
during Haines jumps, Physical Review E (2013) 88(4).
10. Youssef, S., Bauer, D., Bekri, S., Rosenberg, E., Vizika, O. 3D In-Situ Fluid
Distribution Imaging at the Pore Scale as a New Tool for Multiphase Flow Studies,
SPE (2010) 135194.

SCA2015-007

12/12

11. Georgiadis, A., Berg, S., Makurat, A., Maitland, G., Ott, H., Pore-scale microcomputed-tomography imaging: Nonwetting-phase cluster-size distribution during
drainage and imbibition, Physical Review E (2013), 88, 033002.
12. Youssef, S., Bauer, D., Peysson, Y., Vizika, O., Investigation of Pore Structure
Impact on the Mobilization of Trapped Oil by Surfactant Injection, SCA2014-064.
13. X. Pak, T., Butler, I.B., Geiger, S., Van Dijke, M.I.J. and Sorbie, K.S., Droplet
fragmentation: 3D imaging of a previously unidentified pore-scale process during
multiphase flow in porous media. (2015), Proceedings of the National Academy of
Sciences, 112(7), 1947-1952.
14. Haines, W. B., Studies in the physical properties of soil. V. The hysteresis effect in
capillary properties, and the modes of moisture distribution associated therewith, The
Journal of Agricultural Science, (1930) 20(97), 97116.
15. Berg, S., Ott, H., Klapp, S. A., Schwing, A., Neiteler, R., Brussee, N., Makurat, A.,
Leu, L, Enzmann, F., Schwarz, J.-O., Wolf, M., Kersten, M., Irvine, S., Stampanoni,
M., Real-time 3D imaging of Haines jumps in porous media flow, Proceedings of
the National Academy of Sciences, (2013) 110(10), 37553759.
16. Berg, S., Armstrong, R. T., Georgiadis, A., Klapp, S. A., Schwing, A. Neiteler, R.,
Brussee, N., Makurat, A., Leu, L, Enzmann, F., Schwarz, J.-O., Wolf, M., Khan, F.,
Kersten, M., Irvine, S., Stampanoni, M., Multiphase flow in porous rock imaged
under dynamic flow conditions with fast x-ray computed microtomography,
SCA2013-011.
17. Youssef, S., Rosenberg, E., Deschamps, H., Oughanem, R., Maire, E., Mokso, R.,
Oil ganglia dynamics in natural porous media during surfactant flooding captured by
ultra-fast x-ray microtomography, SCA2014-023.
18. Berg, S., Armstrong, R. T., Georgiadis, A., Ott, H., Schwing, A. Neiteler, R., Brussee,
N., Makurat, A., Rcker, M., Leu, Wolf, M., Khan, F., Enzmann, F., Kersten, M.,
Onset of oil mobilization and non-wetting phase cluster size distribution,
SCA2014-026.
19. Roof, J. G., Snap-off of oil droplets in water-wet pores, SPEJ (1970) 10(1), 8590.
20. Armstrong, R. T., Ott, H., Georgiadis, A., Rcker, M., Schwing, A., Berg, S.,
Subsecond pore-scale displacement processes and relaxation dynamics in multiphase
flow, Water Resources Research (2014) 50, 91629176.
21. Tuller, M. and Or, D., Hydraulic conductivity of variably saturated porous media:
Film and corner flow in angular pore space, Water Resources Research, (2001)
37(5), 12571276.
22. Morrow, N., Physics and thermodynamics of capillary action in porous media, Ind.
Eng. Chem. (1970) 62(6), 3256.
23. Wildenschild, D.; Culligan, K. A.; Christensen, B. S. B., Application of x-ray
microtomography to environmental fluid flow problems, In U. Bonse (Ed.),
Developments in X-Ray Tomography IV Proceedings of SPIE (2004) 5535, pp. 432
441.
24. Armstrong R. T., Georgiadis, A., Ott, H., Klemin, D., Berg, S., Desaturation studied
with fast X-ray computed microtomography, Geophysical Research Letters (2014),
41, 5560.

SCA2015-008

1/12

CAPILLARY DESATURATION CURVE PREDICTION

USING 3D MICROTOMOGRAPHY IMAGES
Souhail Youssef, Yannick Peysson, Daniela Bauer , Olga Vizika
IFP Energie nouvelles, 1, avenue de Bois-Prau, 92852 Rueil-Malmaison, France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
In this work, we investigated experimentally the relationship between the Capillary
Desaturation Curve (CDC) and microscopic properties at the pore scale: the oil cluster
size distribution and the porous structure. Experiments were performed on a set of waterwet sandstones with different petrophysical properties. Synchrotron based fast X-ray
microtomography was used to capture the dynamics of oil cluster displacements and to
get insight into the mechanisms that govern trapped oil mobilization by surfactant
injection. Oil cluster size distribution as well as pore geometrical properties were also
quantified using lab based microtomography (-CT) images at the pore scale. Results
showed how the CDC depends on the pore structure and on the oil cluster size
distribution at residual oil saturation. We show that rescaling CDC at the macroscopic
scale using the relative permeability allows the curves to collapse into one curve. Finally,
we propose a new method to predict CDC based on structural properties determined
experimentally by -CT. We demonstrate, that based on these microscopic properties,
the CDC measured on macro plugs can be directly predicted from 3D images.

INTRODUCTION
Enhanced Oil Recovery (EOR) with surfactant can improve significantly the total volume
of oil produced from a reservoir. In water-wet reservoirs, at the end of the water flooding,
the capillary trapping of the oil phase within the rock micro-structure can lead to a high
residual oil saturation. At the pore scale, the oil is trapped at the center of the pores in the
form of disconnected droplets or clusters. This remaining oil can be produced if the
capillary barrier is suppressed by surfactant injection. In this case, the viscous stresses
developed by the water flooding are high enough to mobilize the disconnected oil phase.
The evolution of the residual oil saturation as a function of the trapping number Nt
(capillary number plus Bond number), is known as the Capillary Desaturation Curve
(CDC) and it constitutes an important input parameter in chemical EOR flooding. In this
work, we investigate experimentally the relationship between CDC and microscopic
properties at the pore scale: the oil cluster size distribution and the porous structure.
In the Oil & Gas area, first attempts to predict CDC from structural parameters have been
proposed by Stegemeier [1] or Chatzis et al. [2] in the late seventies. They proposed
correlations based on the pore throat radii distribution to evaluate the capillary
desaturation curve. They obtained, in the domain of application of their correlation, a

SCA2015-008

2/12

quite good estimation of measured CDC. However, these approaches using correlations
are difficult to extrapolate to other rock types. Yet, these pioneer works show the
dependency of CDCs on two main structural parameters: the throat size distribution and
the trapped cluster size, which impacts the accessibility path of the flooding fluid.
Nowadays, with the renewal of the interest of Chemical EOR methods, new experimental
techniques have been developed aiming at the visualization of the oil clusters and the
determination of CDC in the case of model porous structures (glass beads or microfluidic
network) [3-4]. In the present work we used X-ray tomography [5-7] at two scales
(micron and centimeter scale) to provide information on the dependency of capillary
desaturation curve with the local microstructure.

MATERIAL AND METHODS

Fluid and Rock Properties
Displacement experiments were conducted using potassium iodide brine KI (4 wt. %) and
n-Decane. These two fluids have a significant X-ray absorption contrast, allowing a
satisfactory estimation of the two-phase saturations using CT-scan or -CT. The
interfacial tension between the brine and n-Decane was 40 mN/m. Lower interfacial
tension was obtained by adding 0.025 wt. % of Sodium Dodecyl Benzene Sulfonate
(SDBS). The interfacial tension between n-Decane and surfactant-brine solution,
measured by Spinning Drop method, was 0.3 mN/m. Highly consolidated, water-wet,
clean Berea, Bentheimer, Fontainebleau and Clashach sandstones were investigated.
Their main petrophysical properties namely permeability (K0) and porosity () as well as
the residual oil saturation (Sor_macro) are reported in Table 1. Sor_macro, are obtained after
spontaneous imbibition for at least 72 h. The core is then scanned and then water flooded
at very low Nc (7.6 10-8) et re-scanned again to check for the stability.
Table 1. Petrophysical properties of the cores used in coreflood experiments
Core type
Berea
Bentheimer
Fontainebleau
Clashach

K0 (mD)
208
2676
304
426

Porosity (%)
19.4
22.1
11.9
14.1

Sor_macro (%)
48.2
37.2
25.0
38.4

Core Flood Experiments

Coreflood experiments on macro-plug (33 mm in diameter) were combined with CT-scan
imaging to accurately measure the mean residual oil saturation. Samples were first
saturated with brine and drained by n-Decane injection using a centrifuge at a Bond number
around 10-6 . Then, we displaced the oil by injecting brine (or a surfactant added brine) at
different capillary numbers and the pressure between the inlet and outlet was
continuously measured. More details on experimental procedure and sample properties
can be found in [6]. We have also conducted dynamic experiments (drainage, imbibition
and surfactant injection) at the TOMCAT synchrotron beam line of the Swiss Light
Source (SLS) on a Bentheimer sandstone mini-plug (5.8 mm in diameter and 8 mm in

SCA2015-008

3/12

length). 3D images of oil clusters mobilization during the surfactant injection sequence
have been captured at different capillary numbers. 3D images were taken with a voxel
size of 5 m and a time interval of 3s. This experiment is described with more details in
[8]. Finally mini-plugs (5.6 mm in diameter) of each rock type, were also submitted to
drainage under n-Decane by centrifugation followed by a free spontaneous imbibition.
Samples were then imaged by a lab -CT at residual oil saturation conditions with 3 m
resolution. Porous structure statistics data were determined using the pore extraction
methodology developed by Youssef et al. [5]. Different structural parameters were
computed: pore radii, pore throat radii, pore to throat aspect ratio. 3D volume
reconstruction at residual oil saturation conditions permitted then the determination of the
oil cluster size distribution for each sample.

EXPERIMENTAL RESULTS
Capillary Desaturation Curves at The Core Scale
Figure 1.a shows the normalized residual oil saturation measured on the four sandstones
as a function of the trapping number. Normalized residual oil (S*or) is expressed by:
S
S or* or
(1)
S ori
where Sori is the residual oil saturation after the first waterflood (at very low trapping
number 10-7). The trapping number (Nt) is introduced to combine capillary and Bond
effect; it is expressed by the following equation :
(2)
Nt Nc Nb
where Nc and Nb are respectively capillary number and Bond number given by:
g K 0 K rw
V w
; Nb
Nc
cos
cos

(3)

where V is the Darcy velocity, w the viscosity of the brine, the oil/brine interfacial
tension, the contact angle, K0 the absolute permeability, Krw the brine relative
permeability and the oil/brine density difference.
Residual oil saturation mobilization occurs at trapping number values higher than 10-6.
CDC shapes are quite close for all cores. The Fontainebleau sample exhibits a more
pronounced slope due to its low Sori plateau. We also observe a significant shift (one
decade in Nt) between the decreasing part of all CDC. This shift is attributed to pore
structure differences between the samples, since experimental conditions and fluids
composition were identical.
Brine Relative Permeabilities at Residual Oil Saturation
Brine relative permeabilities at Sor were computed from each imposed flow rate Q and the
corresponding measured stabilized pressure difference P by:

SCA2015-008

4/12

K rw

w Q L

(4)

K 0 A P

where A is the section of the plug, L its length, Q the flow rate.
Figure 1.b shows Krw as a function of the trapping number Nt for the different samples. In
order to better visualize the curve shape, experimental data were fitted by a polynomial
fit. Relative permeability curves are distinct for each sample, their characteristic shape
depends on the specific microstructure of the samples. As can be seen, for the low range
of trapping numbers, Krw curves remain constant at a low value and independent of Nt.
These low values are due to large clusters situated in the center of the pores, that have not
yet been mobilized. Then, for higher values of Nt, a strong increase in Krw is observed,
that corresponds to the increase of the number of paths accessible to the wetting phase as
oil clusters have been evacuated. Once a major part of the oil clusters has been mobilized
and evacuated (high Nt values), increase of relative permeability becomes slower and Krw
tends to 1. The dependence of Krw on Nt in the intermediate trapping number range stands
for the existence of a regime in which the pressure difference does not scale linearly with
the flow rate as it is the case for the classical Darcy law. These observations are
confirmed by Sinha et al. [9], Tallakstad et al. [10] and Yiotis et al. [11] that investigate
steady state immiscible two-phase flow. They observed the existence of three regimes,
when expressing the normalized pressure gradient as a function of the capillary number.
1.0

100
0.8

K rw

Sor* (%)

80
60
40
20
0
1.E-08

Berea (BE1)
Bentheimer (BH1)
Fontainebleau (FB2)
Clashach (CL2)

0.6

0.4
Berea (BE1)
Bentheimer (BH1)
Fontainebleau (FB2)
Clashach (CL2)
1.E-07

1.E-06

0.2

(a)
1.E-05

Nt

1.E-04

1.E-03

1.E-02

(b)
0.0
1.E-08

1.E-07

1.E-06

1.E-05

1.E-04

1.E-03

1.E-02

Nt

Figure 1: Capillary desaturation curves (a) normalized Sor vs. Nt (Sor is normalized by Sori which is the
residual oil saturation at the end of the waterflood) (b) Brine relative permeabilities as a function of the
trapping number Nt for the four sandstones investigated.

Pore Scale Dynamic Observations

To study the behavior of oil clusters during surfactant injection, we have analyzed the
size of disconnected oil clusters. These data are extracted from 3D images of the
surfactant injection time-series captured during the dynamic experiment described in [8].
In Figure 2.a we report the mean cluster size as well as the size of the largest cluster as a
function of the time. The surfactant injection rate was increased each 30 s allowing an
increase in the trapping number. This graphic shows that at the beginning of the second
and the third stage of surfactant injection at respectively 30 s and 60 s, as soon as the

SCA2015-008

5/12

injection rate is increased the size of the largest cluster increases immediately then it
decreases to reach almost a constant plateau. This behavior can be explained as follows:
once the capillary number has reached a sufficiently high value (critical capillary
number) oil cluster menisci begin to move. As the interfacial tension is relatively low,
menisci can merge as soon as they get in contact, leading to cluster coalescence and
increasing in this way the size of the largest cluster. Afterwards this largest cluster is
partially evacuated from the sample and then breaks up in smaller clusters. This is
evidenced by Figure 2.b where we can see the evolution of the largest cluster in the time
interval from 33 s to 42 s.
In parallel to that, we can see that the mean cluster size decreases rapidly once the
capillary number is increased and then converges to a constant value. One important
observation is that the level of this plateau decreases for the mean cluster size as well as
for the largest cluster size when the capillary number is increased. These observations
suggest that statistically the capillary number imposes an upper threshold to the cluster
size distribution. This confirms a first order dependency of the cluster size to the capillary
number. This point will be developed theoretically in the last part of this paper.

(a)

(b)

Figure 2 (a) Evolution of the largest cluster size (Rmax) and the mean cluster size (R) as a function of the
time. (b) Formation and progression of the largest cluster in the time interval from 33s to 42s.

Pore scale statistics

Figure 3 shows 3D images of the four rock types at Sor. Minerals and brine are
respectively represented in light gray and dark gray. Oil clusters are represented in color.
The color code corresponds to a size classification beginning with the largest clusters in
yellow followed by red, green, orange, dark blue and finally the smallest clusters are
represented in light blue. The different mean properties computed from these images are
reported in Table 2. Image-based porosities are comparable with the macroscopic
porosity measurements except for the Clashach sample, certainly due to a local
heterogeneity of the sample. The residual oil saturations estimated from the mini-plugs of

SCA2015-008

6/12

the samples are systematically higher than those obtained from the macro-plugs (with the
exception of the Berea sample where we obtained a similar value). This can be explained
by the following fact: spontaneous imbibition of mini-plugs were realized with both faces
open, which can create a side effect and increase Sor above the macroscopic value.
Finally we also observe that the mean cluster sizes are very close to the mean pore sizes
for the four samples.

Figure 3: 3D -CT images of the four sandstones at residual oil saturation state. Minerals are represented in
light gray, brine in dark gray and oil clusters in color. Color code corresponds to a size classification from
the largest to the smallest: yellow, red, green, orange, dark blue and light blue.

Table 2. Statistic data of samples used in -CT experiments.

Core type
Berea
Bentheimer
Fontainebleau
Clashach

img (%)
17.6
23.1
12.0
10.5

Sorimg (%)
45.2
45.2
37.2
58.2

R (m)
28.6
36.9
41.4
38.7

r (m)
10.5
14.8
13.0
12.7

Rb (m)
31.5
40.1
38.6
37.2

Figure 4 shows the radius distributions of the oil cluster Rb, pore bodies R, pore bodies
containing water and oil and pore bodies containing only water. It can be seen for the four
rock types, that mainly small pores contain only water (pink curves). On one hand this is

SCA2015-008

7/12

due to the fact that during drainage oil does not invade the smallest pores. On the other
hand, water invades pores during spontaneous imbibition in the case of a water wet
system in increasing order. The equivalent radius Rb of each oil cluster was computed
assuming that the cluster shape is spherical. Mean values of Rb are comparable to mean
pore body radii (cf. Table 2) but the largest clusters can extend over hundreds of pores.

Figure 4: Radius distributions of the oil clusters (Rb), pore bodies (R), pore bodies containing water and oil
and pore bodies containing only water (Results are obtained at the mini-plug scale).

Trapping of the oil phase due to spontaneous imbibition might be described by a

percolation process [12], as the wetting phase invades first the small pores, leaving
disconnected oil clusters in larger pores. Percolation theory suggests that the number of
oil clusters of volume V follows a power law given by N(v)~v- with = 2.189. This has
been confirmed by Datta et al. [13] and Tallakstad et al. [10] for glass beads and Iglauer
et al. [12] for sandstones. They all observed a power law behavior for the range of lower
clusters volume followed by an exponential cutoff for larger clusters. We have fitted our
data by a power law using a Levenberg-Marquardt algorithm, that is generally used for
non-linear fitting problems (cf. Figure 5). We made a cutoff at 100 m3 as it is the
minimum volume we can measure and found the following exponents: Bentheimer

SCA2015-008

8/12

=2.37, Berea =2.31, Clashach =2.05, Fontainebleau =2.07. Also, our data seem to
show an exponential cutoff, however the number of data points for large clusters is not
sufficient to validate this behavior.

Number of ganglia

1000

1000

Berea
fit
=2,31

100

=2,37

100

10

10

100
1000

Number of ganglia

Bentheimer
fit

1000
Clashach
fit

1000
Fontainebleau
fit

=2,05

100

100
1000

=2,07

100

10

10

1
100

1000
Volume (m3)

100

Volume (m3)

1000

Figure 5: Number of oil ganglia as a function of the ganglia volume and corresponding power law fit.
(Results are obtained at the mini-plug scale).

2Rb

Capillary Desaturation Curve and The Clusters Size Distribution at Sor

We have investigated the relation between the clusters size distribution measured at Sor
and the measured capillary desaturation curve. Figure 6 shows a schematic depiction of
an oil ganglion of length 2Rb trapped in a single pore of radius R. The pore is oriented in
vertical direction. In the pore, pressure and gravity forces act in the same direction in
favor of the mobilization of the oil ganglion through the pore throat (blocking pore
throat) of radius rp, while capillary forces hinder the mobilization.

Oil
Water

360 m

Figure 6: 3D view of trapped oil blobs

from Clashach sample and the
equivalent pore model scheme.

SCA2015-008

9/12

At the pore scale, we can write a force balance to evaluate the threshold of clusters
mobilization [14]. The minimum pressure drop (Pmin) required to mobilize the oil blob
is equal to the Laplace pressure given by:
Pmin Pw1 Pw2 o g(2R b ) 2 cos (

1 1
)
rp R

(8)

where o represents the oil density, g the gravity acceleration, R the pore radius and rp the
blocking pore throat radius. Darcy's law is expressed by:
V =

K 0 K rw
(Pw2 Pw1 + w g (2R b ))
L

(9)

where V is the Darcy velocity.

Substituting (Pw1 - Pw2) deduced from equation (9) in equation (8) leads to:
( o )gK 0 K rw K 0 K rw 1 1
V
+ w
=
( )
cos
cos
R b rp R

(10)

Using the definition of the trapping number equation (10) becomes:

Nt

K 0 K rw 1 1
( )
Rb
rp R

(11)

As a consequence, and considering 1/rp >> 1/R, an oil blob can be mobilized if the
following inequality is respected:
K 0 K rw
(12)
rp R b
This equation shows that the trapping number at the threshold of clusters mobilization is
linked to three main parameters: the effective brine permeability (K0Krw), the throat
radius of the trapping pores and the clusters size. However, these three parameters are not
independent. Indeed, the effective permeability depends on the available path that is
controlled at the pore scale by the pore throat radius and the size of the blocking oil
cluster. It is also obvious, that during a capillary desaturation process the mean size of the
trapped oil cluster changes. Indeed, experimental observations in [7,8,10,14] show that
the mean size of the trapped clusters decreases when the capillary number increases
during the capillary desaturation process. Equation (5) can be written as:
Nt
1

(13)
K 0 K rw rp R b
Parameters of the left hand side of the latter equation are defined at the macro plug scale
whereas parameters of the right hand side are defined at the pore scale (mini plug). We
now assume that the aspect ratio () between the oil cluster radius and the corresponding
trapping radius is constant (rp = Rb/) and multiply the two sides of equation (6) by the
square of the mean pore throat radius < r >. We can then define two equivalent scaling
groups :
Nt

SCA2015-008

10/12

N *t

N t r 2 r 2

K 0 K rw
R 2b

(14)

Figure 7 shows the CDC obtained on macro plugs and plotted as a function of the
reduced trapping number. Using this new scaling group the four CDC quasi
superimposed with a critical capillary number around Nt*=10-2 and a total desaturation at
Nt*=1.
100

Figure 7 Comparison of capillary

desaturation curves plotted using Sor*
vs Nt*. Data are collected on the macro
plug experiments for the four
sandstones (Sor* is the residual oil
saturation normalized by the residual
oil saturation after the first waterflood,
see [6] for the experimental details).

Sor* (%)

80
60
40
20
0
1.E-05

Berea (BE1)
Bentheimer (BH1)
Fontainebleau (FB2)
Clashach (CL2)

1.E-04

1.E-03

1.E-02

1.E-01

1.E+00

Nt*

According to equation (13) we can consider that after flooding at a given reduced
trapping number Nt* all the clusters with a size greater than Rb are removed. At this
trapping number the normalized residual oil saturation S*or(Nt*) in the mini plug can be
expressed as:
4
S*or ( N t* )
R 3bi f ( Rb R bi )
(15)

3Vori R bi Rb
where Rbi is the ith class size of ganglion radius distribution function f(Rb) and Vori is the
total volume of oil in the mini plug after spontaneous imbibition. One can notice that the
function S*or (Nt*) is the volume weighted cumulative function of the ganglion size
N r 2
distribution. In Figure 8 we plot S*or obtained on macro plugs as a function of t
K 0 K rw
and S*or obtained by equation (15) as a function of

r 2
which correspond
R 2b

respectively to the CDC at the macro and mini plug scale.

For this study we consider that is constant. The aspect ratio is computed from the
ratio between <Rb> and < r >, supposing that < r >~< rp > as < rp > cannot be obtained
easily from the -CT images since the flow direction is not known.From these curves we
can see that the CDCs defined at the two different scales match very well for all samples.
These results show that CDCs can be estimated using structural parameters and that
ganglion size distribution is a first order parameter.

SCA2015-008

11/12

Macro plug
Mini plug

100

80

Sor* (%)

Sor* (%)

80

60

60

40

40
20

Macro plug
Mini plug

100

20

Berea

Bentheimer
0
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00

0
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00

Nt*

Nt*
Macro plug
Mini plug

100
80

Macro plug
Mini plug

100

Sor* (%)

Sor* (%)

80

60

60

40

40

20

20

Clashach

Fontainebleau
0
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00

Nt*

0
1.E-04 1.E-03 1.E-02 1.E-01 1.E+00

Nt*

Figure 8: Capillary desaturation curves as a function of the modified trapping number (from macro plug
experiments (points) and from mini plug image properties computed at Sor (triangles).

CONCLUSION
In the present work, we performed brine and surfactant flooding experiments in Berea,
Bentheimer, Fontainebleau and Clashach sandstone in order to investigate the influence
of the pore structure on the mobilization of oil clusters. X-ray Computed Tomography at
two scales was used to provide information on the rock local microstructure and to
measure the mean residual oil saturation at different trapping number values. A modified
trapping number was introduced, that takes into account effective brine permeability, or,
in other words, the size of the effective pathways through which fluid flow occurs. Use of
the modified trapping number permits rescaling CDC on quasi unique curves.
Experimental results showed that by using pore scale geometrical properties extracted
from 3D X-ray images we were able to predict satisfactorily the capillary desaturation
curve for the different samples from the size distribution of the trapped oil at initial
residual saturation. Our observation highlights the importance of the initial distribution of
disconnected oil clusters to understand the properties of capillary desaturation curves.
From the observation of the relative permeability variation with the trapping number for
all rock types, we deduced also a specific scaling for the trapping number. All these
results give new insights in recovery mechanisms and interpretation of capillary
desaturation of a discontinuous non-wetting trapped phase in a porous media.
Introducing this new experimental approach in a chemical EOR process design allows to
save time compared with conventional core flood methods and gives valuable
information on the rock properties at the early beginning of the workflow.

SCA2015-008

12/12

REFERENCES
1. Stegemeier, G.L., Mechanisms of Entrapment and Mobilization of Oil in Porous
Media, D.O. Shah and R.S. Schechter, Editors, Improved Oil Recovery by Surfactant and
Polymer Flooding, Academic Press, Inc., New York, 55-91, (1977)
2. Chatzis I., Morrow N.R., Lim H.T., Magnitude and detailed structure of residual oil
saturation SPEJ, 23 (2), 311-326, SPE-10681-PA, (1983)
3. Krummel A.T., Datta S.S., Munster S., Weitz D.A. Visualizing Multiphase Flow and
Trapped Fluid Configurations in a Model Three-Dimensional Porous Medium. AIChE
Journal, 59: 1022-1029, (2013)
4. Romano M., Chabert M., Cuenca A., and Bodiguel H., Strong influence of geometrical
heterogeneity on drainage in porous media, Phys. Rev. E, 84: 065302, (2011)
5. Youssef S., Rosenberg E., Glan N., Bekri S., Vizika O., Quantitative 3D
characterization of the pore space of real rocks : improved -CT resolution and pore
extraction methodology, SCA2007-17, Int. Sym. of the Society of Core Analysts, Calgary,
Canada (2007)
6. Oughanem, R., Youssef S., Peysson Y., Bazin B., Maire E., and Vizika O., Pore-scale
to core-scale study of capillary desaturation curves using multi-scale 3D imaging,
SCA2013-27, Int. Sym. of the Society of Core Analysts, Napa Valley, California (2013)
7. Berg S., Armstrong R., Ott H., Georgiadis A., S. Klapp A., Schwing A., Neiteler R.,
Brussee N., Makurat A., Leu L, Enzmann F., Schwarz J.-O., Wolf M., Khan F., Kersten
M., Irvine S., Stampanoni M., Multiphase flow in porous rock imaged under dynamic
flow conditions with fast x-ray computed microtomography, SCA2013-11, Int. Sym. of
the Society of Core Analysts, Napa Valley, California (2013)
8. Youssef S., Rosenberg E., Deschamps E., Oughanem R., Maire E., Mokso R., Oil
clusters dynamics in natural porous media during surfactant flooding captured by ultrafast x-ray microtomography, SCA2014-23 Int. Sym. of the Society of Core Analysts,
Avignon, France (2014)
9. Sinha, S. and A. Hansen: "Effective rheology of immiscible two-phase flow in porous
media," Europhysics Letters (2012), 1-5
10. Tallakstad K.T., Knudsen H.A., Ramstad T., Lvoll G., Maly K.J., Toussaint R., and
Flekky G.E., Steady-state, simultaneous two-phase flow in porous media: An
experimental study, Phys. Rev. E, 80: 036308, (2009)
11. Yiotis, A.G., Talon, L., Salin, D.: Blob population dynamics during immiscible twophase flows in reconstructed porous media. Phys. Rev. E, 87, (2013)
12. Iglauer S., Favretto S., Spinelli G., Schena G., and Blunt M.J., X-ray tomography
measurements of power-law cluster size distributions for the non-wetting phase in
sandstones, Phys. Rev. E, 82: 056315 (2010)
13. Datta S.S, Ramakrishnan T.S., David A. and Weitz D.A. Mobilization of a trapped
non-wetting fluid from a three-dimensional porous medium. Phys. Fluids 26: 022002,
(2014)
14. Lenormand, R. and Zacone C., Physics of Blob Displacement in a Two-Dimensional
Porous Medium, SPE Formation Evaluation, 3 (1): 271-275, (1988)

SCA2015-009

1/12

DYNAMIC DRAINAGE AND IMBIBITION IMAGED

USING FAST X-RAY MICROTOMOGRAPHY
M. Andrew, H. Menke, M. J. Blunt & B. Bijeljic
Imperial College London, Department of Earth Sciences & Engineering, London SW7
2AZ, UK
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland, 16-21 August, 2015

ABSTRACT
Recent developments in X-ray microtomography have allowed for multiple fluids to be
imaged directly in the pore space of natural rock with temporal resolutions of tens of
seconds, allowing for the direct observation of pore-scale displacements . This technique
was used to image, with a spatial resolution of 3.64 m and a temporal resolution of 45
seconds, the injection of supercritical CO2 (drainage) and brine (imbibition) into a
carbonate sample at conditions of pressure, temperature and salinity representative of
subsurface flow (10 MPa, 50oC, 1.5 M) using a novel technique for flow control at
extremely low flow rates. Capillary pressure was measured from the images by finding
the curvature of terminal menisci of both connected and disconnected CO2 clusters,
identified using local curvature anisotropy.
Snap-off in this system was examined by analysing a single snap-off event during
drainage and imbibition. Capillary equilibrium concepts do not explain the low capillary
pressures seen in the snapped off regions of the pore-space during drainage. The
disconnected region created during drainage instead preserves the extremely low dynamic
capillary pressures generated during a drainage event (Haines jump). Imbibition appeared
isotropic with no clear macroscopic gradients in saturation developed. Snap-off in this
system appears to be an equilibrium process, with no observable difference in interface
curvature between the connected and disconnected non-wetting phase regions generated
after snap-off.

INTRODUCTION
Recent developments in rapid synchrotron-based microtomography have, for the first
time, allowed researchers to examine systems with a temporal resolution sufficient for the
imaging of dynamic displacement processes, with pioneering papers by Berg et al.[1] and
Armstrong et al. [2] examining Haines jumps and capillary desaturation respectively. At
the same time recent developments have shown that, as microtomography is noninvasive, it is possible to conduct experiments at representative subsurface conditions,
investigating capillary trapping [3, 4, 5, 6] and measuring contact angle [7, 8].
Synchrotron imaging at representative subsurface conditions has been used to examine
multiple points of changing capillary pressure during drainage [9], however temporally
resolved studies have still remained a challenge.

SCA2015-009

2/12

Multiphase flow is controlled by a complex set of variables, such as pore structure,

interfacial tension and wettability. These are in turn a function of system conditions such
as pressure, temperature, and salinity, e.g. Espinoza & Santamarina [10]. In such a
complex system the experiments of flow processes at representative subsurface
conditions becomes highly attractive. We present results from experiments examining
drainage and imbibition conducted on the CO2-brine-carbonate system at 10 MPa and
50oC, conditions representative of subsurface flow in reservoirs and aquifers.
Drainage, or the injection of non-wetting phase (CO2) into a porous medium saturated
with wetting phase (brine), is extremely complicated in reservoir rocks. A qualitative
understanding of the process at the pore-scale can be found using a network-based
approximation of the pore-space, where the arbitrarily complicated pore-space is
simplified into geometric and well-defined pores, connected by equally well-defined
throats. Using this description drainage can be described using invasion percolation
theory where the non-wetting phase (CO2) displaces the wetting phase (brine) from each
pore in a sequence of pore-scale events called Haines jumps [11]. These jumps represent
the change between two states at capillary equilibrium, however as the jumps tend to
proceed on the millisecond time-scale [12, 13], the dynamics of the transition between
these two equilibrium states may display disequilibrium features [14]. Although these
features (such as low dynamic capillary pressure) are transient, disappearing in connected
regions of non-wetting phase as the system re-equilibrates after the transition (and so
inaccessible to dynamic tomography with a temporal resolution of 45s), they may be
preserved if regions of the pore-space filled with CO2 are disconnected from the rest of
the connected non-wetting phase prior to interface re-equilibration [15]. We present an
example of such a phenomenon, showing that snap-off during drainage is due to a
disequilibrium process.
In contrast, during imbibition (brine injection) regions of the pore-space full of wetting
phase appear to swell uniformly over the length scale of the sample, with no clear fluid
front developing. In this case snap-off occurs when the capillary pressure in a throat falls
below the threshold snap-off capillary pressure of that throat, causing the throat to rapidly
fill with brine, disconnecting a region of non-wetting phase from the rest of the
(connected) non-wetting phase. We show that in this case there is no dis-equilibrium
between the connected and disconnected non-wetting phase regions, at the point of snapoff, shown by the fact that there is no difference in their interface curvatures after snapoff is complete.

EXPERIMENTAL METHODS
The core assembly is shown in figure 1 and experimental apparatus shown in figure 2.
The principal challenges when imaging the dynamic displacement of fluids at
representative subsurface flow conditions are associated with system dead volume. The
equipment used to control flow at high pressure includes large syringe pumps, connected
to the core using long lengths of flexible tubing. Imaging the CO2-brine system increases

SCA2015-009

3/12

these problems, as system dead volume is further increased by the incorporation of large
batch reactors used to control fluid-fluid and fluid-solid reactions. This means that the
dead volume of the system (defined as all the volume of fluid outside the rock core,
estimated at around 1L) is many orders of magnitude greater than the internal porevolume of the core (<0.5ml). Small changes in ambient temperature (or errors in the
temperature control of heated elements of the system) cause changes in volume of the
fluids in the system, potentially causing flow through the core. At high flow rates, such as
those used in traditional core-flooding experiments, and in experiments imaging fluid
distributions at the end of displacement processes, this is typically not an issue, as any
induced flow by these changes in temperature is typically small compared to the flow rate
applied across the core [4].
At the very low flow rates required for the examination of dynamic flow, however, flow
induced by these changes in system dead-volume becomes important. In order to better
constrain the flow boundary conditions for the core we propose a new micro-flow
arrangement. A constant pressure drop is applied across a low permeability porous plate
on the outlet face of the core (Figure 1). As this plate has a much lower permeability than
the rock (with a rock to disk permeability ratio of around 1:200,000) essentially the entire
of the applied pressure drop occurs across the plate. As the plate has a high capillary
entry pressure, it remains saturated, resulting in a steady wetting phase (brine) flow
through it. In this method a constant and extremely low flow rate boundary condition is
applied directly at the outlet face of the core. In this way, changes in system volume are
accommodated by changes in pump volume (pumps 1 and 3), rather than across the core.
For this a hydrophilic modified semi-permeable disk (aluminium silicate, Weatherford
Laboratories, Stavanger, Norway) of dimensions 4mm in diameter and 4mm in length
and a permeability of 1410-17 m2 (1410-6 D) was used. In this experiment flow rates of
1.75 10-15 m3/s, equating to a capillary number of 1.10 10-11.
Experiments were conducted on a single carbonate sample, Ketton Limestone, an Oolitic
Grainstone from the Upper Lincolnshire Limestone Member (deposited 169-176 million
years ago). Small cores around 4mm in diameter and 10-20mm in length were drilled
from a large core plug around 38mm in diameter. The low permeability porous plate was
mounted at the outlet face of the core and both were wrapped in aluminium to prevent
diffusive CO2 exchange across the confining sleeve. An X-ray transparent 4mm outer
diameter polymeric tube was placed at the inlet face of the core so that the CO2-brine
interface could be easily monitored during pressurization prior to drainage. The core was
then placed in a flouro-polymer elastomer (Viton) sleeve, which was attached to metal
fittings connecting the core to the pore-fluid flow lines. A thermocouple was mounted
near the base of the core outside the viton sleeve, which was then wrapped twice more
with aluminium foil and placed within a high-temperature high-pressure Hassler type
flow cell. The thermocouple was connected to a custom PID temperature controller,
which heated the cell using an external Kapton insulated flexible heater. The position of
the thermocouple is crucial for accurate temperature control, and by siting it in the

SCA2015-009

4/12

confining annulus by the base of the core it is possible to have a constant temperature xin
the cell throughout the experiment.
The brine used was Potassium Iodide (KI) with a salinity of 1.5M. In order to prevent
reaction between the carbonate rock and carbonic acid formed when scCO 2 is mixed with
brine the three phases were mixed together prior to injection in a heated reactor. Highpressure syringe pumps were used to maintain system pressure and control the flow of
fluids with an incremental displacement resolution of 25.4nL.
The experimental protocol is as follows:
Raise the pressure and temperature in the reactor to that desired for the pore fluid
during the experiment (50oC and 10MPa).
Load the core into the coreholder without the flowlines connected to the pumps, and
establish a confining pressure of 1 MPa in the cell. This is to establish the same
conditions of differential pressure (with the confining pressure 1 MPa above the pore
pressure) as during the experiment.
Centre the core in the field of view.
Image the core along its entire length using a large number of projections (3600 per
180o rotation). This creates a high quality unsaturated image from which the porespace could be analysed in detail and to which saturated images could be compared.
Bring the temperature of the coreholder up to the desired reservoir temperature.
Bring the core up to reservoir pressures.
At reservoir conditions there may be some constant offset between the readings in
pumps 1 and 3, even if they were correctly calibrated at ambient conditions. In order
to find this difference stop pump 3 while running pump 1 in constant pressure mode.
Close valves 6, 7 and 8 then open valve 9. The difference between the pressure
readings in pumps 1 and 3 will then be the pressure offset between the two pumps.
Close valve 9 and open valves 6, 7 and 8. Reduce the pressure in pump 3, considering
the transducer offset between pump 1 and 3, until there is a pressure drop of 5 kPa
across the porous plate. This begins drainage.
Begin taking scans. Exposure time and the number of projections per tomographic
scan will depend specifically on the detector and synchrotron light source used. The
results presented in this study were acquired using 800 projections with an exposure
of 0.04s. from each image.
The qualitative progress of the drainage process can be monitored without stopping
the tomographic sequence by monitoring changes in the first radiograph.
After drainage is complete, increase the pressure in pump 3 until it reaches 5 kPa
pressure drop across the porous plate, in the opposite direction to that applied during
drainage.
Begin taking scans. Exposure time and the number of projections per tomographic
scan will depend specifically on the detector and synchrotron light source use.

IMAGE ACQUISITION AND PROCESSING

SCA2015-009

5/12

Imaging was performed at Diamond Imaging Beamline I13 (Diamond Light Source,
UK). Imaging was conducted under pink-beam conditions, with each scan taking around
45 seconds to acquire, with 32 seconds projections and around 13 seconds returning to
the initial state and preparing for the next scan. To reduce the impact of beam heating on
fluid distributions the pink-beam was then filtered by 2 mm of in-line graphite, 2 mm of
in-line aluminium and 10 m of in-line gold.
All image processing was conducted using Avizo Fire 8.0 and imageJ programs. Each
image in the time-series was reconstructed using a filtered back projection algorithm[16],
binned then cropped such that each image consisted of around 110011001100 voxels
with a voxel size of 3.64 m. The images were filtered using a non-local means denoising
filter and registered to the first image in the dynamic sequence. The difference between
each image and the first in the sequence (prior to non-wetting phase invading the porespace during drainage) was computed for both drainage and imbibition, giving a high
contrast map of non-wetting phase distributions. The time sequence of these maps of the
non-wetting phase was then rendered in 3D to identify individual events of interest. Once
these events had been identified subvolumes were extracted from the 4D data temporally
located around each event. These subvolumes were then segmented using a seeded
watershed algorithm [17]. Surfaces were generated across the non-wetting phase using a
smoothed marching cubes algorithm [18], and interface curvature were calculated by
approximating the surface locally as a quadratic form [15]. Terminal menisci from the
wetting-non-wetting phase interface were identified using the anisotropy of the curvature
vector (only taking measurements where the smallest principal radius of curvature of the
interface was very close in magnitude to the largest principal radius of curvature). The
radii of curvature of these regions of the wetting-non-wetting phase interface were
maximised relative to the voxel size, making the estimation of mean curvature more
reliable and were furthest away from the three phase contact line, minimising the impact
of smoothing on resulting curvature distributions. Details about this technique will be
included in an upcoming paper.

DRAINAGE
During drainage non-wetting phase invasion does not occur uniformly, but instead as a
discrete set of events where regions of the pore-space which were previously occupied by
wetting phase (brine) become occupied by non-wetting phase (CO2) (Haines jumps [19]).
The progress of these events is not resolved with the temporal resolution provided by
synchrotron based dynamic tomography. Micromodels have been used to show that pores
drain on the millisecond timescale, and that the time taken for the jump to occur is
independent of the macroscopic capillary number [12, 13]. This behavior can be
understood by describing these events and changes between two states at capillary
equilibrium; before and after each of the non-wetting phase jumps. Dynamic tomography
therefore allows us to see the sequence of these jumps during drainage, and by looking in
detail at the fluid distributions before and after each of the jumps, we can examine and
test some of the assumptions in descriptions of multi-phase flow. Specifically it allows us
to examine the relative processes by which regions of non-wetting phase can become

SCA2015-009

6/12

isolated from the rest of the pore-space occupied by connected non-wetting phase in the
process of snap-off. This is particularly interesting as the disconnected non-wetting phase
region preserves the conditions present at the point of snap-off, even as flow continues
around it.
Snap-off occurs during drainage by the following mechanism. As the wetting-nonwetting interface emerges from a pore throat during a drainage event, the curvature across
that interface will decrease. This causes curvature disequilibrium between the leading
region and other regions of the wetting-non-wetting phase interface. As the interface
attempts to re-equilibrate, the curvature in the pore-throat connecting the leading portion
on the non-wetting phase from the rest of the connected non-wetting phase starts to
decrease. If the curvature in this throat decreases below the threshold snap-off capillary
pressure for that throat, the leading portion of non-wetting phase will snap-off,
disconnecting it from the rest of the connected CO2.
This process can be seen in the experiments examined in this study in two successive
tomographic scans in the 4D dataset, separated by 45 seconds (figure 3, A-B). Surface
curvature is measured, using curvature anisotropy (the relative magnitude of the larger
and smaller principal curvature vectors), for both the connected and disconnected CO2
phase before and after the jump (figure 3C).
The interface curvatures measured on the disconnected ganglion are much lower than
those of the connected CO2 region either before or after the invasion snap-off sequence
has occurred. We propose that this is due to significant dynamic capillary pressure
gradients generated in the leading region of the non-wetting phase during the drainage
event, out of capillary equilibrium. These cause the low capillary pressures required for
throat snap-off, which are then preserved in the disconnected ganglion. The capillary
pressure of the ganglion is defined by the threshold snap-off capillary pressure of the
disconnecting throat, defined by the local pore geometry, consistent with the findings of
Andrew et al. [6], where the capillary pressures of a population of disconnected ganglia
were observed to be controlled by local pore geometry (and so threshold snap-off
capillary pressures). After the invasion/snap-off event has completed the rest of the
connected non-wetting phase interface will re-equilibrate to some new stable interface
curvature, however the disconnected region will not, thereby retaining and preserving the
low dynamic curvatures generated during the invasion event on the time scale resolvable
with synchrotron based microtomography.

IMBIBITION
During imbibition the wetting phase invades the pore-space in a uniform fashion, with no
clear front developing over the length scale observable in the experiment. No
macroscopic gradients in saturation (front) develop on the length-scale of the image as
the connected wetting-non-wetting interface remains at capillary equilibrium (displaying
no gradients in the interface curvature of the connected non-wetting-wetting phase
interface). As capillary pressure decreases, corresponding to an increase in wetting phase

SCA2015-009

7/12

saturation, wetting phase regions appear to swell uniformly throughout the imaged
volume.
If we look at snap-off in this system (during imbibition) the connected non-wetting phase
curvature is not distinctly different before or after the snap-off event, and the
disconnected non-wetting phase region curvature is not distinctly different to the
connected non-wetting phase region either before or after the event (figure 4). These
results can be understood by thinking of imbibition (and resulting snap-off) as an
equilibrium process. As wetting layers swell during wetting phase invasion, capillary
pressure slowly decreases across the entire wetting-non-wetting phase interface. When it
gets below the threshold snap-off capillary pressure for a particular throat, that throat
snaps off rapidly, but because prior to snap-off the entire wetting-non-wetting interface
was at equilibrium, there is no gradient in capillary pressure between the connected and
disconnected non-wetting phase regions. The connected non-wetting phase will then
continue to be displaced out of the system by the invading wetting phase, reducing
connected surface capillary pressure. The disconnected ganglion, however, will retain the
capillary pressure present at snap-off.
This is in contrast to snap-off occurring during drainage, where dynamic forces generated
by a Haines jump decrease curvature in a throat to below its threshold snap-off curvature.
Because these changes in capillary pressure are due to dynamic forces (occurring on the
millisecond time-scale), they are not represented in the curvature of the connected phase
in the next image in the time sequence (with an image time resolution of around 45
seconds). As the capillary pressure of the disconnected phase is primarily controlled by
the threshold snap-off capillary pressure, as defined by the local pore geometry, dynamic
curvatures are preserved in the disconnected ganglion, creating a difference in interface
curvature between connected and disconnected non-wetting phase regions.

CONCLUSIONS
We have used synchrotron based dynamic tomography to image the drainage (nonwetting phase invasion) and imbibition (wetting phase invasion) process in a carbonate
sample at representative subsurface conditions with a time resolution of around 45
seconds. Snap-off has been observed in both processes, disconnecting non-wetting phase
regions. Curvature distributions show a difference in interface curvature between the
disconnected and connected non-wetting phase region after snap-off, whereas during
imbibition there is no clear difference between the interface curvature distributions. We
propose that this difference in interface curvature is due to a difference in the physics
causing the snap-off in each system.
Snap-off in the drainage process is due to dynamic forces, arising due to low interface
curvatures generated as a non-wetting phase (CO2) region invades a region of the porespace occupied by wetting phase (brine) in a Haines jump. This reduces interface
curvatures in the invaded pore throat to below its threshold snap-off capillary pressure.
Snap-off during imbibition, however, is due to the equilibrium swelling of wetting phase
regions during brine invasion, until the threshold snap-off capillary pressure of a throat is

SCA2015-009

8/12

reached. At this point the throat rapidly fills with wetting phase, however prior to snapoff the interface was at equilibrium, meaning that the resulting connected and
disconnected non-wetting phase regions have no difference in interface curvature.
Future work could focus on examining the relative statistical importance of snap-off
processes during drainage, or the correlation of modelled snap-off capillary pressures to
experimentally observed interface curvatures. Another interesting area for future research
is how the results of these experiments at extremely low flow rates correspond to more
rapid displacements seen in realistic subsurface conditions, and how the interaction
between viscous pressure gradients, causing flow, and capillary forces can cause changes
in pore-scale displacement processes on the individual pore and ganglion scales.

ACKNOWLEDGEMENTS
We thank Diamond Light Source for access to beamline I13 and especially the assistance
of Dr. Christoph Rau and Dr. Joan Vila-Comamala who contributed to the results
presented here. We acknowledge funding from the Imperial College Consortium on PoreScale modelling. We also gratefully acknowledge funding from the Qatar Carbonates and
Carbon Storage Research Centre (QCCSRC), provided jointly by Qatar Petroleum, Shell,
and Qatar Science & Technology Park.

REFERENCES
1. Berg, S. et al. Real-time 3D imaging of Haines jumps in porous media flow. Proc.
Natl. Acad. Sci. U. S. A. 110, 37559 (2013).
2. Armstrong, R. T., Georgiadis, A., Ott, H., Klemin, D. & Berg, S. Critical capillary
number: Desaturation studied with fast X-ray computed microtomography. Geophys. Res.
Lett. 41, 5560 (2014).
3. Andrew, M., Bijeljic, B. & Blunt, M. J. Pore-scale imaging of geological carbon
dioxide storage under in situ conditions. Geophys. Res. Lett. 40, 39153918 (2013).
4. Andrew, M., Bijeljic, B. & Blunt, M. J. Pore-scale imaging of trapped supercritical
carbon dioxide in sandstones and carbonates. Int. J. Greenh. Gas Control 22, 114
(2014).
5. Chaudhary, K. et al. Pore-scale trapping of supercritical CO2 and the role of grain
wettability and shape. Geophys. Res. Lett. 40, 38783882 (2013).
6. Andrew, M., Bijeljic, B. & Blunt, M. J. Pore-by-pore capillary pressure measurements
using X-ray microtomography at reservoir conditions: Curvature, snap-off, and
remobilization of residual CO2. Water Resour. Res. 50, 87608774 (2014).
7. Aghaei, A. & Piri, M. Direct pore-to-core up-scaling of displacement processes:
Dynamic pore network modeling and experimentation. J. Hydrol. 522, 488509 (2015).
8. Andrew, M., Bijeljic, B. & Blunt, M. J. Pore-scale contact angle measurements at
reservoir conditions using X-ray microtomography. Adv. Water Resour. 68, 2431
(2014).

SCA2015-009

9/12

9. Herring, A. L., Andersson, L., Newell, D. L., Carey, J. W. & Wildenschild, D. Porescale observations of supercritical CO2 drainage in Bentheimer sandstone by synchrotron
x-ray imaging. Int. J. Greenh. Gas Control 25, 93101 (2014).
10. Espinoza, D. N. & Santamarina, J. C. Water-CO2-mineral systems: Interfacial
tension, contact angle, and diffusionImplications to CO2 geological storage. Water
Resour. Res. 46, 110 (2010).
11. Wilkinson, D. & Willemsen, J. F. Invasion percolation: a new form of percolation
theory. J. Phys. A. Math. Gen. 16, 33653376 (1999).
12. Mohanty, K. K., Davis, H. T. & Scriven, L. E. Physics of Oil Entrapment in WaterWet Rock. SPE Reservoir Engineering 2, (1987).
13. Armstrong, R. T. & Berg, S. Interfacial velocities and capillary pressure gradients
during Haines jumps. Phys. Rev. E - Stat. Nonlinear, Soft Matter Phys. 88, 19 (2013).
14. Moebius, F. & Or, D. Pore scale dynamics underlying the motion of drainage fronts
in porous media. Water Resour. Res. 50, 84418457 (2014).
15. Armstrong, R. T., Porter, M. L. & Wildenschild, D. Linking pore-scale interfacial
curvature to column-scale capillary pressure. Adv. Water Resour. 46, 5562 (2012).
16. Titarenko, V., Titarenko, S., Withers, P. J., De Carlo, F. & Xiao, X. Improved
tomographic reconstructions using adaptive time-dependent intensity normalization. J.
Synchrotron Radiat. 17, 689699 (2010).
17. Jones, A. C. et al. Assessment of bone ingrowth into porous biomaterials using
MICRO-CT. Biomaterials 28, 24912504 (2007).
18. Hege, H., Seebass, M., Stalling, D. & Zockler, M. A Generalized Marching Cubes
Algorithm Based On Non-Binary Classifications. ZIB Prepr. sc-97-05, (1997).
19. Haines, W. B. Studies in the physical properties of soil. V. The hysteresis effect in
capillary properties, and the modes of moisture distribution associated therewith. The
Journal of Agricultural Science 20, 97 (1930).

SCA2015-009

10/12

Figure 1: Core Assembly, showing the relative positioning of the low permeability
porous plate, the core and the plastic space (allowing for the wetting-non-wetting
interface to be seen prior to invasion in the core.

Figure 2: Experimental flow apparatus, showing how high-pressure syringe pumps were
connected to the flow cell.

SCA2015-009

11/12

Figure 3: A-B: Connected and disconnected non-wetting phase (CO2) after invasion &
snap-off during drainage, taken from two subsequent tomographies. In this rendering the
wetting phase (brine) and the rock grains are see-through. The surfaces are coloured
according to the magnitude of the mean interface curvature. C: The distribution of
curvatures, as measured on terminal menisci identified using curvature anisotropy. The
disconnected non-wetting phase region formed by snap-off has a lower interface
curvature than the connected non-wetting phase region either before or after the snap-off
event.

SCA2015-009

12/12

Figure 4: A-B: Connected and disconnected non-wetting phase (CO2) after invasion &
snap-off during imbibition, taken from two subsequent tomographies. In this rendering
the wetting phase (brine) and the rock grains are transparent. The surfaces are coloured
according to the magnitude of the mean interface curvature. C: The distribution of
curvatures, as measured on terminal menisci from the wetting-non-wetting fluid-fluid
interface, identified using curvature anisotropy. The disconnected non-wetting phase
region formed by snap-off has an indistinguishable interface curvature to the connected
non-wetting phase region either before or after the snap-off event.

SCA2015-010

1/12

WETTABILITY ANALYSIS USING MICRO-CT, FESEM

AND QEMSCAN, AND ITS APPLICATIONS TO DIGITAL
ROCK PHYSICS
Nasiru Idowu1, Haili Long1, Pl-Eric ren1, Anna M. Carnerup2, Andrew Fogden3, Igor
Bondino4 and Lars Sundal5
1
FEI Trondheim AS, Trondheim, Norway; 2 FEI Australia Pty Ltd, Canberra, Australia;
3
Australian National University, Canberra, Australia; 4 Total E&P, Pau, France;
5
DEA Norge AS, Oslo, Norway
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada 16-21 August, 2015

ABSTRACT
This study presents an integrated methodology for determining the pore-scale distribution
of wettability of rock samples to guide pore network modeling. Wettability was
characterized by spatial registration of rock images from X-ray micro-computed
tomography (MCT), Field Emission Scanning Electron Microscopy (FESEM), and
Quantitative Evaluation of Minerals by SEM (QEMSCAN). The approach was applied to
miniplugs of an outcrop and a reservoir sandstone, which were drained and aged in oil
and underwent spontaneous and forced imbibition of brine. Tomogram acquisition after
each preparation step showed that, from similarly low initial water saturation, oil
recovery by spontaneous imbibition was high in the outcrop and virtually zero in the
reservoir sample, while the additional recovery by forced imbibition was the opposite.
These results and the pore-scale distributions of remaining oil suggest different
wettabilities for each sample though their petrophysical properties (porosity and
permeability) are similar.
The cleaned miniplugs were then subjected to FESEM mapping of raw cut faces to
visualize local wettability alteration, and to QEMSCAN of polished faces to relate this to
surface mineralogy. The established wettability information was then used to assign
plausible wettability parameters to pores and throats of topologically equivalent
networks. The simulated oil/water displacement results for the reservoir sandstone
showed good agreement with available SCAL data.

INTRODUCTION
Optimal recovery from hydrocarbon reservoirs during waterflooding and enhanced oil
recovery techniques depends largely on a good understanding of crude oil/brine/rock
interactions wettability [1]. To classify reservoir wettability either as water- or oil-wet
is a gross oversimplification that can lead to unexpectedly low oil recovery and
inefficient reservoir management [2, 3]. A mixed-wet state is often inferred, but without
sufficient information to determine its detailed distribution. The three standard

SCA2015-010

2/12

measurement methods for quantitatively characterizing wettability of rock samples are

contact angle goniometry [4], and the Amott [5] and USBM [6] tests based on
spontaneous and forced imbibition/drainage. Goniometry is restricted to a prepared rock
section, which often limits measurement precision. Amott and USBM tests only yield an
averaged wettability measure of a core sample, and the former is relatively insensitive
near neutral wettability. Complementary wettability information can be obtained from,
e.g. spontaneous imbibition rates, relative permeability, calorimetry, flotation of crushed
rock or microscopy of model substrates, although the results are often less quantitative or
less directly relevant. Nuclear magnetic resonance (NMR) methods have also been used
to provide quantitative wettability indices [7, 8].
Advanced, high-resolution imaging techniques such as MCT and FESEM have recently
been used to determine the pore-scale distribution of oil in rock plugs before, during or
after waterflooding and to visualize the asphaltene films remaining on cleaned rock
surfaces due to local wettability alteration [9-13]. To date, establishment of quantitative
links between these imaging studies and connections to prediction of residuals are
limited. Andrew et al. [14] recently presented a method for contact angle measurement
from MCT of a supercritical CO2-brine system in a Ketton limestone. However, this rock
has simple mineralogy, comprising 99.1% calcite and 0.9% quartz [14].
In this study, we establish the wettability of reservoir and outcrop sandstone samples by
integrating imaging results from MCT, FESEM and QEMSCAN. Secondly, we use this
information to assign plausible wettability parameters to pores and throats of
topologically equivalent pore networks extracted from the segmented pore space. Finally,
we simulate oil/water displacements with a quasi-static pore network model and compare
the results with available experimental data.

MATERIALS AND METHODS

Two sandstone samples were used in this study. MCT (by HeliScan) of the received core
plugs (25 mm) showed that the outcrop Sample 1 was fairly homogenous while the
reservoir Sample 2 was highly heterogeneous with several laminations. Locations within
these low resolution tomograms were chosen for subsampling of two miniplugs of 6 mm
15 mm and 9 mm 10 mm from each, using a manually-fed drill press. The miniplugs
were cleaned by hot Soxhlet extraction in toluene, methanol and chloroform/methanol
azeotrope, and then dried. Mercury injection capillary pressure (MICP) measurements
were performed on the 9 mm diameter miniplugs while the 6 mm diameter miniplugs
were used in the remainder of the study. The measured petrophysical properties are
summarized in Table 1.
Tomograms of the pair of 6 mm miniplugs were acquired (at 3.7 or 3.9 m/voxel
resolution for Samples 1 and 2, respectively) in a series of five prepared states, starting
with their cleaned, dry state. Each miniplug was then infiltrated, under vacuum followed
by high isostatic pressure, with brine comprising 5 g/l CaCl2, doped with 1.5 M NaI, after
which this brine-saturated state was scanned. Primary drainage was then carried out by

SCA2015-010

3/12

centrifugation in the supplied crude oils to a maximum capillary pressure of 500 kPa for a
minimum of 12 h, followed by 14 days of ageing at 60oC or 83oC for Samples 1 and 2,
respectively. The third tomogram was acquired in this state of established initial water
saturation and wettability alteration. The miniplug was then immersed in the same brine
and temperature as mentioned above for 7 days of spontaneous imbibition, followed by
re-scanning. Forced imbibition was then performed by centrifugation in this brine to a
maximum pressure of 17.5 or 14.0 kPa for Samples 1 and 2, respectively, for at least 18
h, after which this fifth state was imaged.
Table 1. Measured petrophysical properties of the two sandstone samples.

Sample No.
1 (outcrop)
2 (reservoir)

Porosity (%)
22.5
22.7

Kabs (mD)
524
853

Each miniplug was then cleaned of its remaining oil using decalin and heptane, followed
by centrifugation in the undoped brine, prior to being cut in half crosswise using a Struers
saw with water-based coolant. After removal of salt by soaking in methanol, the bottom
half was re-scanned and then imaged by FESEM (Helios 600 NanoLab, FEI) using a
secondary electron detector. Its raw cut face was first mapped at 0.59 m/pixel, followed
by acquisition of higher magnification micrographs at selected locations to resolve the
asphaltene footprint (or lack of) on pore walls. The top half of the miniplug was resin
embedded for preparation of a polished face for analysis by SEM (Quanta 650 FEG, FEI)
using a backscattered electron detector for imaging and Energy-dispersive X-ray
spectroscopy for mineral identification, automated by MAPS software and QEMSCAN.

RESULTS
MCT Imaging and Analysis
For each of the two 6 mm miniplugs, their series of five tomograms were post-processed
by Mango software for 1) spatial registration to within one voxel, 2) correction of beam
hardening artifacts, 3) cylindrical masking of the damaged periphery, and 4) rescaling of
attenuation. The processed tomograms and their differences were then segmented using
converging active contours algorithms to identify solid grains, pores and sub-resolution
microporosity. The oil and brine occupancy of each voxel of these latter two categories at
the end of primary drainage and ageing (PD), and spontaneous (SI) and forced imbibition
(FI) was identified. For the segmented porosity and water saturation of Samples 1 and 2,
Table 2 lists their tomogram averages, Figure 1 plots their (radially-averaged)
longitudinal profiles, and Figure 2 shows a representative cross-sectional slice, in which
rock, oil and water correspond to grayscales, red and blue, respectively. The top ( 5 mm)
and more porous part of the Sample 2 miniplug broke after PD and the remaining piece
was used for aging, SI and FI. The lower segmented porosity in Table 2 compared to
Table 1 and the shorter profile length in Figure 1 are due to this breakage.

SCA2015-010

4/12

Table 2. Average segmented porosity and water saturation from MCT of 6 mm diameter miniplugs.

Sample No.
1
2

Porosity (%)
20.7
17.4

Swi_PD (%PV)
3.0
3.2

Sw_SI (%PV)
59.3
4.5

Sw_FI (%PV)
62.9
34.1

The two miniplugs have similar measured petrophysical properties (Table 1) and
comparable Swi after primary drainage and aging (Table 2, which may be underestimated
due to limitations in resolving brine thin films). However, the corresponding profiles in
Figure 1 show that the outcrop Sample 1 is homogeneous while the reservoir Sample 2 is
heterogeneous with tighter and more open laminations. The two miniplugs also exhibit
very different imbibition response. Sample 1 shows very strong spontaneous imbibition,
while Sample 2 has almost none, aside from the rise near the ends which is common to
both. Additional recovery by forced imbibition is limited for Sample 1, but substantial for
Sample 2, and with heterogeneities partly mirroring those of its porosity profile. The
computed Amott water indices (Iw) from segmentation of the oil and water phases in
Samples 1 and 2 are in the ranges 0.60-0.94 and 0.01-0.04, respectively.
Figure 2a shows that the irreducible water of the 6 mm Sample 1 miniplug resides as
expected in the tightest locations and in the relatively rare clay-aggregate particles. Given
the predominance of grain-lining clays, much of the water may remain connected, in spite
of its overall low saturation. By contrast, the irreducible water of the 6 mm Sample 2
miniplug in Figure 2d appears less connected, due to its smoother grains with less clay
lining and the occlusion of tighter pores by a slightly lower-attenuating (darker) mineral
cement. However, it should be borne in mind that Swi of this slice is only 1.5 %, as it
comes from the lowest-saturation dip of the blue curve (Swi_PD profile) in Figure 1b.

(a)

(b)

Figure 1. Longitudinal profiles of porosity (% bulk volume) and Sw (% PV) after primary drainage,
spontaneous and forced imbibition, for the miniplugs of (a) Sample 1 and (b) Sample 2.

The substantial spontaneous imbibition into Sample 1 in Figure 2b recovers virtually all
oil from smaller pores and near grain surfaces to leave large blobs in pore bodies (e.g.
highlighted with green boxes), presumably due to snap-off in throats. Virtually no further
change is seen on forced imbibition in Figure 2c. Conversely, the change in Sample 2
from Figure 2d to 2e due to spontaneous imbibition is almost imperceptible, while
subsequent forced imbibition in Figure 2f results in complete recovery of oil from many

SCA2015-010

5/12

of the larger pores (e.g. highlighted with yellow circles). The remaining oil generally
occupies narrower pores closest to grain surfaces, and retains higher connectivity than for
Sample 1. Larger uninvaded clusters, e.g. at upper left in Figure 2f, may be due to the
miniplug heterogeneity. The computed Iw and the pore-scale remaining oil distributions
suggest that Sample 1 is water-wet while Sample 2 is oil-wet.

(a)

(b)

(c)

(d)

(e)

(f)

Figure 2. Registered tomogram slices of (a-c) Sample 1 and (d-f) Sample 2, showing oil (red) and
brine (blue) distribution after (a, d) primary drainage, (b, e) spontaneous and (c, f) forced imbibition.

QEMSCAN
Figure 3 shows registered automated maps from backscattered electron SEM imaging and
QEMSCAN mineral identification on a polished embedded section within the upper half
of the miniplugs of Samples 1 and 2, which do not coincide with the tomogram slices in
Figure 2. From Figure 3b, almost 40% of all grains in Sample 1 comprise Na plagioclase,
although these are indistinguishable from the quartz grains in Figures 2a-c and 3a owing
to their similar X-ray attenuation and electron density, respectively. Feldspar corresponds
to the slightly brighter grains in MCT, which are relatively rare. The prevalent flaky
particles are mainly muscovite with occasional biotite, and particles and seams of detrital
chlorite are fairly common. These clays are the brighter mineral features in Figures 2a-c
and 3a, while the brightest grains correspond to the densest trace minerals. Illite and
biotite in the form of grain-lining plates are under-represented due to resolution
limitations of both QEMSCAN (2.4 m/pixel) and MCT. Figure 3d shows that Sample 2
is also dominated by quartz and Na plagioclase. The latter is the cement-like phase in

SCA2015-010

6/12

Figures 2c-f and 3c of slightly lower X-ray attenuation and electron density, respectively,
than the surrounding quartz grains. Feldspar is more prevalent than in Sample 1, while
clays are less common, although their presence as grain linings may again be
underestimated. Figure 3c shows quite substantial damage to Sample 2 during resin stub
preparation, while Sample 1 in Figure 3a remained intact.

Figure 3. Registered maps of a polished cross-section of the 6 mm miniplugs of (a-b) Sample 1 and (cd) Sample 2, by (a, c) backscattered electron SEM mapping and (b, d) QEMSCAN mineral mapping.

FESEM Imaging
The FESEM map of a rectangular subarea of the raw cut face of the cleaned, dry
miniplug of Samples 1 and 2 is shown in Figures 4a and 4c, respectively. This 2D map
was registered into the 3D tomogram series via the intermediate step of registering the
tomogram of the cut (half-) miniplug to its uncut counterpart [12]. The uppermost voxel
at each position on the cut face was identified by segmentation to yield its height map.
The FESEM map was then 2D-2D registered to the vertical projection of these uppermost
voxels onto a flat plane. The voxels corresponding to this same height map in the original
miniplug tomogram, and all others in the series, were then similarly projected. Smoothing
of the grain-hugging height map was also performed so that it spans the pore bodies.
Figures 4b and 4d show the resulting registered projection of the tomograms after
spontaneous and forced imbibition of Samples 1 and 2, respectively. This procedure
served to identify subregions in the FESEM map which were free from debris and

SCA2015-010

7/12

plucking damage and where the oils pore occupancy and grain contact displayed
interesting imbibition history, for inspection by higher-resolution FESEM imaging.

Figure 4. Low resolution FESEM mosaic of cut face of (a) Sample 1 (3.3 mm x 4.5 mm) and (c)
Sample 2 (3.8 mm x 4.5 mm), and registered smoothed tomogram projection of (b) Sample 1 after
spontaneous imbibition and (d) Sample 2 after forced imbibition, with brine in white and oil in black.

The micrographs in Figure 5 are from the center of the circles marked in Figure 4, all of
which were oil-filled after primary drainage. Within the yellow circle in Figure 4b of

SCA2015-010

8/12

Sample 1, the oil blob trapped in this pore body after spontaneous imbibition is locally
anchored to a grain. The corresponding close-up of a quartz overgrowth of this grain in
Figure 5a shows a light scattering of asphaltene nanoparticles, suggestive of a weakly
water-wet surface and possibly accounting for the blob attachment there. Moreover, none
of the multitude of high resolution FESEM images acquired of Sample 1 showed
substantially greater asphaltene coverage than this, implying that the majority of its
surfaces are weakly to strongly water-wet. Figure 5a also exhibits another common
feature of Sample 1, namely debris in the form of very fine clay platelets on oil-exposed
regions. Within the red circle in Figure 4b, spontaneous imbibition detached the oil from
an illite-coated grain. The corresponding close-up in Figure 5b, and others in this
neighborhood, displays no asphaltene deposit, in line with the expectation that grainlining clays aid retention of water and water-wetness to cause oil snap-off.

Figure 5. High resolution FESEM images, all with 0.5 m scale bars, of raw cut face of Sample 1 at
the location of the (a) yellow and (b) red circle in Figure 4a-b, or of Sample 2 at the location of the (c)
yellow and (d) red circle in Figure 4c-d.

Within the yellow circle in Figure 4d of Sample 2, the pore remains oil-filled after forced
imbibition. The close-up in Figure 5c of a grain with quartz overgrowths bounding this
pore is representative of the much heavier deposits of asphaltene than on Sample 1. Film
coverage is almost complete, aside from the so-called dalmation pattern of bright
perforations from brine nano-droplets trapped under the film during primary drainage and
aging (arrowed). Other occasional bare patches (also arrowed) may be caused by fines
detachment by moving contact lines during cleaning.

SCA2015-010

9/12

Within the red circle in Figure 4d, forced imbibition removes oil from a grain that is
partly smooth and partly lined by clay, a close-up from the latter of which is given in
Figure 5d. In contrast to Sample 1, the clay linings of Sample 2, in this and all other
inspected regions, appear completely covered by asphaltene. This is again consistent with
the oil-wetness inferred from MCT, and suggests that local retention of oil, such as in the
vicinity of Figure 5c, is due to reasons other than mineralogy. QEMSCAN in Figures 3b
and 3d showed that Samples 1 and 2 are both dominated by quartz and plagioclase, lined
by illite to varying extents. Further, the same brines were used for both samples and were
drained to similar Swi values. Thus the strong wettability differences observed in MCT
and FESEM are most likely due to the different crude oils, with density of 835 and 900
kg/m3, used for Samples 1 and 2, respectively.

COMPARISON OF SIMULATED RESULTS WITH

EXPERIMENTAL DATA FOR SAMPLE 2
Topologically equivalent pore networks [15] were extracted from the segmented pore
space of the miniplug dry tomogram. To establish whether or not the pore networks are
representative, oil/water primary drainage displacement was simulated using a quasistatic pore network model [16]. The resulting oil/water Pc Sw curves were scaled into
mercury/air Pc Sw curves using interfacial tensions and contact angles [17]. Figure 6a
compares the measured MICP data on the 9 mm diameter miniplug with the simulated
data on the 6 mm miniplug, and Figure 6b shows the oil/water primary drainage relative
permeability curves. The measured and simulated MICP are in good agreement within the
limits of image resolution, suggesting that the pore networks are representative.

(a)

(b)

Figure 6. (a) Comparisons of simulated (Sim) and experimental (Expt) MICP data, and (b) primary
drainage oil/water relative permeability curves for Sample 2. The blue circles and red squares are the
simulated data points fitted to the corresponding lines.

Based on the wettability results presented above, we then assigned plausible wettability
input parameters to pore network elements of Sample 2 for simulation of imbibition
processes. In particular, the fraction of pores contacted by oil that changed wettability

SCA2015-010

10/12

from water- to oil-wet was 1.0, and the minimum and maximum advancing contact angle
for oil-wet pores was 100o and 150o, respectively. Simulations of oil/water displacements
were then performed for comparison with available data from special core analysis.
The experiments used a composite core of four plugs, from the same formation as Sample
2, with a porosity of 24% and effective oil permeability of 854 mD at Swi of 0.251. The
simulation was on pore networks extracted from the Sample 2 miniplug with a porosity of
17.4%, computed grid permeability of 699 mD and simulated Swi of 0.20. Figure 7a
compares the porous plate primary drainage (PD) and spontaneous imbibition (SI)
capillary pressure data, using dead crude oil on the four plugs, with simulated data.
Forced imbibition (FI) measurement carried out on one of the four plugs gave one
additional data point at -60 kPa as shown in Figure 7b. The volume of water
spontaneously imbibed and the calculated Iw from the porous plate experiments, MCT
imaging and simulations are summarized in Table 3, with good quantitative agreement
between the data.

(a)

(b)

Figure 7. Comparisons of simulated (dashed line) with porous plate experimental capillary pressure
data (solid line) for Sample 2. (a) Primary drainage (PD) and spontaneous imbibition (SI), and (b) SI
and forced imbibition (FI).
Table 3. Volume of brine spontaneously imbibed and calculated Amott water indices.

Porous plate experiment

MCT imaging
Simulation

Volume of brine spontaneously

imbibed (% PV)
1.0-5.0
1.3
1.2

Calculated Iw
0.014
0.010-0.040
0.018

To eliminate the influence of the different Swi values, the imbibition relative permeability
data were normalized. The composite core measured data were history-matched with a
Sendra core flooding simulator to produce the experimental LET [18] curve fit. Figure 8
compares the normalized simulated relative permeability results to those from
experiment. Considering the heterogeneity of Sample 2 and the difference in scale

SCA2015-010

11/12

between the core experiment and the miniplug simulation, the measured relative
permeability compares qualitatively well with the simulation.

Figure 8. Comparisons of simulated (Sim) and experimental (Expt) oil/water relative permeability
curves for Sample 2. The blue circles and the red squares are the simulated data points fitted to the
corresponding lines.

CONCLUSIONS
The pore-scale distribution of water in the initial state, and spontaneously and forcibly
imbibed, from MCT imaging and registration can provide useful information on the porescale distribution of wettability. Integration of FESEM images at raw surfaces provides
complementary insight at higher resolution, and combination with QEMSCAN of
embedded sections can shed light on mineral-specific wettability. Although the two
samples studied proved to be fairly uniformly water-wet and oil-wet, this imaging
framework can greatly benefit the interpretation of more complex mixed-wet samples.
Network modelling of multiphase transport properties incorporating this wettability
identification showed good quantitative/qualitative agreement with experimental data,
despite the difference in scale.

ACKNOWLEDGMENTS
The authors acknowledge FEI Trondheim AS for granting permission to publish this
paper, which is partly funded by DEA Norge AS and Norwegian Research Council
through PETROMAKS Project Number: 208772/E30, titled: Physically based threephase relative permeability relations. We also thank STATOIL, TOTAL and DEA for
sharing their samples and data with us. Jill Middleton (ANU) is thanked for assistance
with FESEM-MCT registration.

REFERENCES
1. Anderson, W.G., Wettability Literature Survey Part 1: Rock/Oil/Brine Interactions
and the Effects of Core Handling on Wettability, J. Pet. Tech. (1986), 38: 11251144.

SCA2015-010

12/12

2. Morrow, N.R., Wettability and its Effects on Oil Recovery, J. Pet. Tech. (1990),
42: 1476-1484.
3. Abdallah, W., Buckley, J.S., Carnegie, A., et al., Fundamentals of Wettability,
Oilfield Review. Summer 2007: 44-61.
4. Anderson, W.G., Wettability Literature Survey Part 2: Wettability Measurement,
J. Pet. Tech. (1986), 38: 1246-1262.
5. Amott, E., Observations Relating to the Wettability of Porous Rock, Trans. AIME,
(1959), 216: 156-162.
6. Donaldson, E.C., Thomas, R.D. and Lorenz, P.B., Wettability Determination and its
Effect on Recovery Efficiency, SPE Journal, (1969) 9: 13-20.
7. Freedman, R., Heaton, N., Flaum, M., et al., Wettability, Saturation, and Viscosity
From NMR Measurements SPE Journal, (2003) 8: 317-327.
8. Looyestijn, W. J., and Hofman, J., Wettability Index Determination from NMR
Logs, SPE-93624-MS (2005). doi:10.2118/93624-MS.
9. Kumar, M., Fogden, A. and Senden, T., Investigation of Pore-Scale Mixed
Wettability, SPE Journal, (2012) 17: 20-30.
10. Fogden, A., Kumar, M., Morrow, N.R., et al., Mobilization of fine particles during
flooding of sandstones and possible relations to enhanced oil recovery, Energy
Fuels, (2011) 25: 1605-1616.
11. Marathe, R., Turner, M.L., and Fogden, A., Pore scale distribution of crude oil
wettability in carbonate rocks, Energy Fuels (2012), 26: 6268-6281.
12. Dodd, N., Marathe, R., Middleton, J., et al., Pore-Scale Imaging of Oil and
Wettability in Native-State, Mixed-Wet Reservoir Carbonates, Paper IPTC 17696,
(2014).
13. Lebedeva, E.V. and Fogden, A., Micro-CT and wettability analysis of oil recovery
from sandpacks and the effect of waterflood salinity and kaolinite, Energy Fuels,
(2011), 25: 5683-5694.
14. Andrew, M., Bijeljic, B., and Blunt, M.J., Pore-scale Contact Angle Measurements
at Reservoir Conditions Using X-ray Microtomography, Advances in Water
Resources. (2014), 68: 24-31.
15. Bakke, S., and P.E. ren, 3-D pore-scale modelling of sandstones and flow
simulations in the pore networks, SPE Journal, (1997), 2, 136-149.
16. ren, P.E., S. Bakke, and O.J. Arntzen, Extending Predictive Capabilities to
Network Models, SPE Journal, (1998), 3, 324-336.
17. Idowu, N., Nardi, C., Long, H., et al., Improving Digital Rock Physics Predictive
Potential for Relative Permeabilities From Equivalent Pore Networks, Paper
SCA2013-17, Proceedings of the 2013 SCA Symposium, Napa Valley, California,
USA, 16 -19 September , 2013.
18. Lomeland, F., Ebeltoft, E., and Thomas, W. H., A new versatile relative
permeability correlation, Paper SCA2005-32. Proceedings of the 2005 SCA
Symposium, Toronto, Canada, 21-25 August, 2005.

SCA2015-011

1/10

IMPACT OF DISPLACEMENT RATE ON WATERFLOOD

OIL RECOVERY UNDER MIXED-WET CONDITIONS
Yukie Tanino1, Blessing Akamairo2, Magali Christensen1, Stephen A. Bowden2
1
School of Engineering, University of Aberdeen, UK, 2School of Geosciences, University
of Aberdeen, UK
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Under uniformly water-wetting conditions, it is well established that waterflood oil
recovery remains independent of flood rate while the flow is capillary-dominated and,
furthermore, that the rate dependence emerges when the microscopic capillary number
exceeds O(10-5). In contrast, there is no equivalent framework for interpreting the flood
rate-dependence of oil recovery under mixed-wet conditions representative of oil
reservoirs. Indeed, not even the appropriate definition of capillary number under mixedwet conditions is established. In this paper, we focus specifically on oil recovery from
mixed-wet porous media at high initial oil saturation and its dependence on water
injection rate. We present laboratory measurements of oil distribution and its evolution
during secondary waterflood at 2.5, 5.0, and 500 ft day-1. Oil distribution was directly
imaged at the pore scale using a high-speed camera coupled to an optical microscope in a
bed of calcite grains packed into a microfluidic chip. These measurements are
benchmarked against bulk measurements of remaining oil saturation in limestone cores.
Data to date suggest that remaining oil saturation, So, after the same volume of water
injection, displays a dependence on injection rate qualitatively similar to the classic
capillary desaturation curve for uniformly water-wet media, with the rate dependence
emerging at a higher injection rate under mixed-wet conditions than under water-wet
conditions. However, unlike in uniformly water-wet media, the long-time (residual) So
does not display a dependence on the rate. These findings may have important
implications not only for oil recovery, but for a wide range of engineering applications
involving mixed-wet porous media, e.g., geological CO2 storage, remediation of nonaqueous phase liquid (NAPL)-contaminated soils and aquifers, and irrigation.

INTRODUCTION
Even in the absence of heterogeneities that may give rise to poor sweep, only 40 to 70%
of oil can be recovered by waterflooding. This incomplete recovery is attributed to
capillary trapping, the naturally-occurring phenomenon whereby the non-wetting phase in
the larger pores of a porous medium is immobilized against advection in the form of
pore-scale ganglia by capillary forces.

SCA2015-011

2/10

Capillary trapping is controlled by the pore structure, wettability, properties of the fluids,
the initial oil saturation Soi, and the flow rate. This paper focuses on the impact of flow
rate on capillary trapping during secondary waterflood of rock with high Soi. The
dependence of oil recovery on flow rate is traditionally characterized by the capillary
number, most commonly defined as
U
Ca w w ,
(1)

where w is the dynamic viscosity of the flood water, Uw is its Darcy velocity, and is its
interfacial tension with the oil phase. While more relevant definitions of capillary
number have been derived (see Ref. [1] for a detailed discussion), Eq. (1) has the
advantage of being a function only of fluid properties. For convenience, the microscopic
capillary number as defined in Eq. (1) will be used in the present paper.
While many studies report measurements of remaining oil saturation at different Ca,
literature that focus on the Ca dependence controlled by variations in flow rate are limited
(e.g., Ref. [2] and references therein). If we further restrict our consideration to
laboratory experiments in which (a) the wetting state of the model reservoir is
unambiguous, (b) Soi at the onset of waterflood was established by displacement of the oil
into a porous medium the scenario relevant to oil recovery, geological carbon storage,
and NAPL contamination of groundwater aquifers and (c) is representative of field
conditions, available data are limited (Table 1). This paper presents new laboratory
measurements of oil recovery to complement the existing data set. We consider recovery
from two limestones and a packed bed of calcite grains at injection rates corresponding to
Uw = 0.0015 to 12 mm s-1. In particular, we focus on mixed-wet conditions established
by wettability-altering constituents in the oil phase. The results are compared to the
behaviour of uniformly water-wet media.

METHODS
Two techniques were used to measure the remaining oil saturation, So, as a function of
time during waterfloods: lab-on-a-chip methods and conventional corefloods. For both,
each experiment comprised three stages: complete saturation of the porous medium with
the model connate water, primary drainage, and secondary waterflood at constant
injection rate. To remove ambiguity in the wettability of the system, a new porous
medium was used in each experiment.
Fluids
The oil phase was a degassed sample of a stock oil held at University of Aberdeen in the
lab-on-a-chip experiments and a 6.62 10-2 M solution of either cyclohexanepentanoic
acid (Sigma-Aldrich, 98%) or cyclohexanepropionic acid (99%) in n-decane ( 99%) in
the corefloods. The crude sample was topped (light components removed) prior to use by
heating at 40C under a nitrogen stream to yield an oil with an API of 28 to 30. The
aqueous phase was seawater with a salinity equivalent of 34 ppt in the lab-on-a-chip
experiments and an aqueous solution of 5wt.% NaCl and 1wt.% KCl in the corefloods.

SCA2015-011

3/10

The latter was further saturated with the rock to prevent dissolution of the core during the
coreflood. Basic properties of the fluids are summarized in Table 2.
The interfacial tension of the two oil/water pairs considered in the corefloods are
approximately the same at 30 mN m-1 [7, 8]; the measurement of of the crude oil
with the seawater is pending. The organic acids naturally found in the crude oil sample
and added artificially to the synthetic oils have been shown to alter the wettability of
carbonates (e.g., [9]). Preliminary measurements of contact angle on a calcite substrate
using the sessile drop method confirm the wettability alteration capacity of the oil phases
used presently, with the crude oil, 6.62 10-2 M cyclohexanepentanoic acid in n-decane,
and 6.62 10-2 M cyclohexanepropionic acid in n-decane yielding a static contact angle
of 60 , 140 and 110, respectively.
Table 1 Studies that report oil recovery by waterflood at constant injection rate for multiple injection rates.
Only experiments in which (a) the wetting state of the test samples are well defined, (b) initial oil
saturations were larger than Soi > 0.6 and (c) were established by displacement of oil into a water-saturated
core, (d) remaining oil saturation at the limit of low Ca ( 10-5) was measured, and (f) > 30 mN m-1 are
listed.

reference

aqueous
phase

non-aqueous
phase

[mN
m-1]

porous
media

52.3
0.4 [7]

Ketton
limestone

Tanino &
Blunt [3]

5wt.% NaCl,
1wt.% KCl

n-decane

Tie &
Morrow [4]

5% CaCl2;
seawater

Cottonwood
crude

29.7
32.8;
50.0
50.0

Abrams [5]

Chatzis &
Morrow [6]
Fig. 6

various
synthetic

2% CaCl2

various
synthetic

Soltrol 130
or 170 oil

50.0;
33.7;
35.5
32.8;
50.0
50.0;
33.7;
32.0;
36.3
34.8
35.9
35.9
34.8
34.8

Whitestone
limestone
Edwards
limestone
Berea
sandstone
Bandera
sandstone

k [1015
m2]

marker
in Fig. 5

0.226

2880

blue

0.26
0.21

7.4;
13.5
11.3;
13.5

red
red

0.209

306

blue +

0.226

31.6

blue -

Paluxy
sandstone

0.268

1840

blue dot

Dalton
sandstone

0.276

484

blue

Gallup
sandstone

0.267

1130

blue *

Berea
sandstone

0.214
0.204
0.211
0.187
0.177

501
419
479
115
75.0

blue
blue
blue
blue
blue

SCA2015-011

4/10

Table 2 Properties of the fluids used in the waterfloods. All values were measured at ambient temperature
(22C).

fluid
topped crude oil
6.62 10-2 M cyclohexanepentanoic acid in n-decane
6.62 10-2 M cyclohexanepropionic acid in n-decane
seawater
synthetic brine used in the corefloods

density
viscosity
-3
[kg m ] [mPa s]
874.3
80
731.6
0.950
731.3
0.910
1023.4
0.950
1041.5 1.0315

Lab-on-a-chip Experiments
The porous medium was a quasi-two-dimensional bed of granulated calcite, with an
average diameter of 35 3 m, packed in a 1200 m-wide and 50 m-deep channel
etched in soda lime. The fabrication of the microfluidic chip (Dolomite Centre Ltd.) is
described by Bowden et al. [10]. The porosity of the packed bed was = 0.45 and the
pore volume was 0.05 L in all experiments.
Waterflood procedure
The packed bed was first saturated with the seawater to establish connate water saturation
(Fig. 1a). Next, oil was injected (Fig. 1b) until a uniform oil saturation of Soi 1 was
established (Fig. 1c). To establish mixed wettability, the packed bed was aged at
elevated temperature for 30 minutes. Once the system cooled to ambient temperature,
seawater was injected at constant flow rate using a high precision syringe pump (Fig. 1d).
At the end of a waterflood, the bead pack was dissolved with 10% hydrochloric acid (Fig.
1e). The channel was further cleaned by flushing dichloromethane and methanol through
it before a new bed was packed with fresh calcite grains.

Figure 1 The packed bed at different stages of a lab-on-a-chip experiment: connate water injection (a); oil
injection (b) to establish initial oil saturation (c); waterflood (d), and acid injection in preparation for the
next experiment (e). Flow direction is bottom to top.

SCA2015-011

5/10

The evolution of the depth-integrated oil distribution inside the packed bed during the
waterflood stage was captured using a high-speed camera coupled to a microscope at a
resolution of 3.3 to 3.4 m pix-1 (Fig. 2); the chip was lit from below. The acquired still
images were first converted to grey scale. Next, the still image corresponding to the
instance immediately prior to the onset of waterflood was subtracted from all subsequent
images. The remaining oil saturation at a given instance is then given by So(t) = 1 - nw /
(N ), where N is the total number of pixels in the region of interest (the packed bed) and
nw is the number of pixels in the same region with a greyscale level larger than a
threshold value.
Camera

Syringe
pumps

Computer

bead
pack

Fluidic Chip

Figure 2 Setup for the lab-on-a-chip experiments.

The primary source of error in the lab-on-a-chip experiments is the non-uniform

propagation of the flood water in the microfluidic channel as it approaches the packed
bed. Specifically, the flood water tends to enter the packed bed from one part of its
cross-section at early times in the waterflood. The errors arising from this deviation were
quantified by evaluating So(n) for the left and the right half of the packed bed separately
for each waterflood.
Corefloods
Experiments were performed on cores from two naturally water-wet rocks: Ketton
limestone ( = 0.23; permeability to brine k = 3 10-12 m2) and Indiana limestone [ =
0.15; k = (1.9 to 14) 10-15 m2]. The pore structure of the two rocks differ considerably:
Ketton limestone has a distinct bimodal pore (entry) size distribution as measured by
mercury injection porosimetry, while Indiana limestone displays a more complex
structure (Fig. 3). All cores were 37.6 37.7 mm in diameter and L = 76 mm (Ketton) or
L = 89 mm (Indiana) in length.

SCA2015-011

6/10

rp f (rp)

0.4

0.2

0
0.001

1000

rp [m]
Figure 3 Pore entry size distribution for Ketton limestone (solid line) and Indiana limestone (dashed) as
determined from mercury injection capillary pressure measurements. rp is the equivalent pore throat radii
to the applied capillary pressure, Pc, Hg, and is given by rp = 2 Hg |cos Hg | / Pc,Hg, where the interfacial
tension and contact angle of mercury with its vapour are taken to be Hg = 485 mN m-1 and Hg = 135,
respectively [11]; f (= dSw /drp) is the probability distribution function for pores with a capillary entry
pressure corresponding to rp.

Waterflood procedure
Corefloods were performed under radial confining pressure in custom-made horizontal
Hassler-type cells. Initial oil saturations of Soi = 0.83 to 0.93 were established using the
porous plate method, whereby oil is injected into a brine-saturated core at constant
pressure against a hydrophilic permeable disk that retains the oil in the core while
allowing the brine through. This technique ensures that a uniform oil saturation is
established across the core. Subsequently, brine was injected into the core at a constant
volumetric flow rate of Qw = 0.750, 100, and 800 ml min.-1 for Ketton limestone and Qw
= 0.100, 0.200, and 2.000 ml min-1 for Indiana limestone. These flow rates correspond to
Ca ranging from Ca = 6 10-8 to 8 10-4. After a selected volume of water, n [pv], was
injected the core was removed from the cell and weighed to determine So(n). The relative
permeability of a core to the flood water during waterflood was estimated as
U L
k rw n w w ,
(2)
k pn

SCA2015-011

7/10

where pn is the pressure drop across the length of the core temporally averaged over
a window of 1 pv. Additional details of the coreflood procedure are provided in Ref. [3].

RESULTS
Remaining oil saturations normalized by their initial saturation are presented in Figs. 4
and 5 for the unconsolidated bed of calcite chips and the two limestones, respectively.
Note that, in Fig. 5, the ordinate has been further normalized by So / Soi at the lowest Ca
considered for that rock to facilitate comparison between rocks of different pore size
distributions.

1
0.8
0.6
0.4
0.2
0
0.000001

0.00001

0.0001

0.001

0.01

0.1

wUw [mN m-1]

Figure 4 Normalized remaining oil saturation, So / Soi, at different wUw [mN m-1] in a mixed-wet bed of
calcite chips after n = 1 (triangle), 5 (), and 10 (circle) of water injection. Soi 1.

During any waterflood, So decreases with time until it asymptotes to its residual state.
After n = 1 pv of water injection, So in the packed bed was constant at So/Soi 1 between
wUw = 810-6 to 210-5 mN m-1, then decreased to So/Soi = 0.76 at wUw = 210-3 mN m1
(Fig. 4, triangle). This dependence is qualitatively similar to the capillary desaturation
curve for uniformly water-wet media. However, unlike in water-wet media, So/Soi
approached the same value (So/Soi 0.5) at all Uw, with this limit being achieved earlier
(i.e., at lower n) at higher Uw, such that the dependence on Uw can no longer be discerned
after n = 10 pv (circle).
We now consider recovery from consolidated rock. In the two limestones considered
presently So/Soi remained constant within experimental uncertainty from Ca = 6 10-8 to
Ca = 10-4, then decreased between Ca = 10-4 and 8 10-4 (Fig. 5). Thus the transition to
Ca dependence occurs at a Ca two orders of magnitude larger than in uniformly waterwet rock (Fig. 5, blue markers), for which the threshold generally falls between Ca = 10-6
and 10-5.

SCA2015-011

8/10

1
0.8
0.6
0.4
0.2
0
1E-08

0.0000001 0.000001 0.00001

0.0001

Ca

0.001

0.01

0.1

Figure 5 Normalized remaining oil saturation, So/Soi, at different Ca under mixed-wet conditions (red) and
uniformly water-wet conditions (blue) measured in Ketton (solid circle, n = 105 - 122) and Indiana (solid
square, n = 17 - 19) limestones from the present study, in Whitestone and Edwards limestones by Tie and
Morrow [4], in Berea sandstone by Chatzis & Morrow [6], and in five sandstones by Abrams [5]. The
ordinate is normalized by So/Soi at the lowest Ca considered for that rock in a particular study. For marker
definition of data from literature, see Table 1.

Interestingly, the two mixed-wet limestones considered by Tie & Morrow [4] do not
display a Ca-independent regime even at Ca as low as Ca = 10-7 (red open circle, open
square). The discrepancy with the present experiments may be attributed to the degree of
wettability alteration by the constituents in the oil phase: oil-contacted grain surfaces are
rendered oil-wet (i.e., > 90) in the present corefloods, while wettability alteration is
inferred to have been relatively minor ( < 90) in Tie & Morrow [4]s corefloods.
Table 3 Summary of experimental conditions for the corefloods and the measured remaining oil saturation
(So) and krw.

rock

Ketton
limestone
Indiana
limestone

0.233
0.230
0.235
0.218
0.143
0.154
0.153

oil
(concentration of
acid in n-decane)
6.62 10-2 M
cyclohexanepentanoic acid
6.62 10-2 M
cyclohexanepropionic acid

Ca

So(n)

7.1 10-7
9.4 10-5
7.6 10-4
7.3 10-4
6.0 10-8
1.1 10-7
1.1 10-6

0.38 (106)
0.39 (114)
0.31 (105)
0.35 (122)
0.28 (18)
0.28 (19)
0.29 (17)

krw(n)

marker
in Fig. 5

0.55 (77-107)

red

0.56 (17)
0.61 (18)

red

SCA2015-011

9/10

The relative permeability to the flood water, at n corresponding to the So measurements,

was approximately krw 0.6 in all three corefloods for which it could be measured (Table
3). This value corresponds to saturation-normalized relative permeabilities of krw/(1-So)
0.8 (Indiana limestone) and 0.9 (Ketton), which are significantly larger than krw/(1-So)
0.2 measured previously in the same rocks under uniformly water-wet conditions [3, 7].
The elevated krw indicate that the flood water was not restricted to the smallest pores,
presumably because a significant fraction of the surface of the grains that constitute the
cores has been rendered oil wet [7].

CONCLUSIONS
Laboratory measurements of remaining oil saturation established under mixed-wet
conditions in two limestones and in a bed of unconsolidated calcite chips were presented.
As Ca increased from Ca 10-8 to 10-3, remaining oil saturation in the limestones
remained constant up to Ca = 10-4, and then decreased from Ca = 10-4 to 10-3. Analogous
measurements in packed beds suggest that in contrast to uniformly water-wet
conditions this Ca dependence is a transient phenomenon, and that the residual (longtime) saturation depends only weakly on Ca. Future laboratory work will focus on larger
Ca ( > 10-3) and a wider range of .

ACKNOWLEDGEMENTS
This material is based on work supported by the University of Aberdeen Chevron Fund, a
Society of Petrophysics and Well Log Analysts grant, and an Aberdeen Formation
Evaluation Society scholarship. M.C. was supported by the University of Aberdeen
College of Physical Sciences scholarship. The authors thank undergraduate student
Matthew P. Wilkie for the measurements of density and viscosity of the synthetic brine
and Colin Taylor for the mercury injection porosimetry measurements on Indiana
limestone.

REFERENCES
1. Hilfer, R. & Oren, P.-E. (1996) Dimensional analysis of pore scale and field scale
immiscible displacement. Transp. Porous Media 22(1), 5372.
2. Lake, L.W. (1989) Enhanced Oil Recovery. Old Tappan, NJ: Prentice Hall.
3. Tanino, Y. & Blunt, M.J. (2012) Capillary trapping in sandstones and carbonates:
Dependence on pore structure. Water Resour. Res. 48(8), W08525. doi:
10.1029/2011WR011712.
4. Tie, H. & Morrow, N.R. (2005) Low-flood-rate residual saturations in carbonate
rocks. Proc., International Petroleum Technology Conference, 21-23 Nov., Doha,
Qatar.
5. Abrams, A. (1975) The influence of fluid viscosity, interfacial tension, and flow
velocity on residual oil saturation left by waterflood. SPE J. 15(5), 437-447.
doi:10.2118/5050-PA.
6. Chatzis, I. & Morrow, N.R. (1984) Correlation of capillary number relationships
for sandstone. SPE J. 24(5), 555-562. doi:10.2118/10114-PA.

SCA2015-011

10/10

7. Tanino, Y. & Blunt, M.J. (2013) Laboratory investigation of capillary trapping

under mixed-wet conditions. Water Resour. Res. 49(7), 4311-4319.
8. Christensen, M., Dufour, R. & Tanino, Y. Non-aqueous phase recovery from
mixed-wet media: dependence on pore-scale wettability contrast. Water Resour.
Res., in prep.
9. Wu, Y., Shuler, P.J., Blanco, M., Tang, Y. & Goddard, W.A. (2008) An
experimental study of wetting behavior and surfactant EOR in carbonates with
model compounds. SPE J. 13(1), 26-34. doi:10.2118/99612-PA.
10. Bowden S.A., Cooper, J.M., Greub, F., Tambo, D. & Hurst, A. (2010)
Benchmarking methods of enhanced heavy oil recovery using a microscaled beadpack. Lab Chip 10(7), 819-823.
11. Good, R.J. & Mikhail, R.S. (1981) The contact angle in mercury intrusion
porosimetry. Powder Technol. 29(1), 53 62. doi = 10.1016/00325910(81)85004-8.
12. Moore, T.F. & Slobod, R.L. (1955) Displacement of oil by water: effect of
wettability, rate, and viscosity on recovery. Proc., Fall Meeting of the Petroleum
Branch of AIME, 2 - 5 Oct., New Orleans, Louisiana.

SCA2015-012

1/12

COMBINED CT AND MAGNETIC SCANNING

TECHNIQUES FOR MULTIMODAL IMAGING OF FLUID
FLOW IN POROUS MEDIA: APPLICATION TO HEAVY
OIL WATERFLOODING
Petar Petrov1, David K. Potter1, Shauna Cameron2, Mike London2,
James Donald2 and *Wade Waterman2
1
Department of Physics, University of Alberta, Edmonton, Alberta, Canada
2
Alberta Innovates - Technology Futures, Edmonton, Alberta, Canada
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Waterflooding is one oil recovery process used in heavy oil reservoirs. However, little
has been published in terms of imaging heavy oil waterfloods, mainly due to the similar
densities of water and heavy oil, which makes it difficult to track the advancement of the
water/heavy oil front. The present study used computer tomography (CT) and magnetic
susceptibility techniques to attempt to quantitatively image waterflooding of a heavy oil
saturated sandpack, and monitor the progress of the water/heavy oil front in real time. A
low concentration of superparamagnetic nanoparticles (20 nm diameter maghemite) was
added to water during the flooding. These particles act as dual response contrast agents,
having an extremely high magnetic susceptibility that can be monitored magnetically via
a surrounding sensor, and an increased X-ray attenuation over water alone for CT
scanning. Jar tests were first undertaken to establish the optimum conditions for both CT
and magnetic susceptibility scanning.
Waterflooding, with the dilute nanoparticle suspension, of a heavy oil saturated sandpack
revealed the formation, growth and movement of a significant positive magnetic
susceptibility peak. This formed at the injection end of the flow cell and migrated
towards the production end as the waterflooding progressed. The peak was likely due to a
higher concentration of nanoparticles collecting at the main water/heavy oil front. This
appears to provide a means of quantitatively tracking the position of the front in real time.
The increased accumulation of nanoparticles at the main front was further supported by
material collected in the production jars. Whilst the CT attenuation profiles in part of the
sandpack showed some correspondence with the magnetic results, the CT profiles did not
show a recognizable front. This may be due to the low contrast between the water +
nanoparticles and the heavy oil, to the presence of trapped gas, and to the shallow and
diffuse nature of the front in a waterflood with heavy oil. Once the main water/heavy oil
front had passed through the sandpack the magnetic profiles had a constant shape similar
to the porosity profile (confirmed by the CT derived porosity variation). The magnetic
technique has potential for monitoring larger scale commercial waterflooding operations.

SCA2015-012

2/12

INTRODUCTION
There is little in the literature regarding quantitatively imaging waterflooding of a heavy
oil saturated sandpack (simulating one of the oil recovery processes employed in the
oilsands of Northern Alberta and elsewhere), although there are related studies on viscous
fingering [1] and using CT scanning of waterflooding in low permeability chalk [2]. The
main aim of this project was to provide a means of quantitatively imaging the
waterflooding of a heavy oil saturated sandpack, and tracking the water / heavy oil front
in real time. Monitoring such waterflooding using CT scanning, without the addition of a
contrast agent, is very difficult due to the similar densities of water (1.00 g/cc) and heavy
oil (0.99 g/cc). Therefore it was proposed that the addition of superparamagnetic
nanoparticles to the water (or brine) phase might improve the CT contrast between the
water (or brine) phase and the heavy oil. Moreover, these nanoparticles would act as dual
contrast agents, and their progress can be also be independently monitored by a magnetic
susceptibility sensor [3]. The advantage of the nanoparticles is that they have an
enormously higher magnetic susceptibility compared to the sandpack, the water or the
heavy oil. The sandpack (quartz) and fluids are dominantly diamagnetic, which means
they have very low negative values (quartz theoretically has a mass magnetic
susceptibility of -0.62 x 10-8 m3 kg-1 and the magnetic susceptibility of typical reservoir
fluids is given in [4]). The superparamagnetic maghemite nanoparticles, on the other
hand, have a value around 55,000 x 10-8 m3 kg-1 (which varies a bit depending upon how
they are dispersed). A series of jar tests (without porous media) were initially conducted
in order to identify the most appropriate nanoparticles for both CT and magnetic use, and
to determine the optimum conditions for their dispersion and stability over prolonged
time periods in water and various brines. A waterflooding experiment was then
undertaken on a heavy oil saturated sandpack to evaluate the effectiveness of the
nanoparticle injection for the CT and magnetic scanning techniques.

METHODS
Jar Tests Prior to Waterflood Experiment
A series of superparamagnetic nanoparticle dispersions (testing maghemite, magnetite
and nickel ferrite nanoparticles) were first prepared for magnetic susceptibility scanning
and CT scanning jar tests without porous media. We prepared brine samples where the
nanoparticles were dispersed in sodium chloride or in sodium iodide. In each case a small
amount of dispersant, sodium dodecylbenzeno sulfonate (DDBS), was added to each
sample. This was followed by sonication for several minutes. Tests had shown that this
anionic dispersant (rather than a cationic dispersant such as cetyltrimethyl ammonium
bromide), followed by sonication, was the most effective way of dispersing the
nanoparticles. The maghemite nanoparticles remained in suspension longer than the other
types, and were thus chosen for the subsequent waterflood experiment. Nevertheless,
these maghemite nanoparticles still tended to settle out in a timescale of around 1 hour.
CT contrast tests indicated that there could potentially be enough X-ray contrast to
observe a waterflood with any of the nanoparticles, but that the variability of measured
values, combined with the expected porosity and saturation contrast, made such a

SCA2015-012

3/12

conclusion uncertain. Substituting iodine for chlorine in the brine enhanced the oil-brine
contrast.
A second set of jar tests were carried out to determine the optimum conditions required
for creating long term stable maghemite nanoparticle suspensions. The stability with time
of different solution compositions (nanoparticle, DDBS dispersant, sodium iodide and
sodium chloride concentrations), as well as different mixing processes (mechanical and
sonication), was determined using a Bartington MS2C sensor. The jar tests revealed that
the addition of sodium iodide or sodium chloride, either before or after the formation of
the suspension, caused the nanoparticles to agglomerate and settle out relatively fast. The
jar tests composed of only nanoparticles and DDBS dispersant mixed in deionised water
displayed remarkable stability over long periods of time, consistent with our previous
stability and flow experiments [3]. Therefore we decided to use deionised water rather
than brine in the waterflood experiment, and the CT contrast would be provided by the
nanoparticles themselves.
Experimental Set-up for Waterflood Experiment
A PEEK flow cell was assembled and leak tested in preparation for the waterflood
experiment. We used a Hassler-type vessel with a confining pressure (radial) of 600 kPa
during the waterflooding. A Viton sleeve was packed with around 78 g of sand and then
inserted into the flow cell. The sandpack was 114.3 mm long and approximately 22.2 mm
in diameter. The core material is water wet Ottawa sand, F110 from the U.S. Silica
Company. F110 is a pure quartz sand (99.8% SiO2) with minor amounts of Fe2O3, Al2O3
(<0.1% each), and other oxides. Grain diameter is 50150 m. Particles are subangular,
and the size distribution is as follows: 8% 5375 m, 25% 75106 m, 44% 106150
m, 18% 150212 m, and 4% 150212 m, with <1% beyond the upper and lower
limits. The median particle diameter is 85 m, the median pore diameter 46 m, and the
median throat diameter 18 m. The absolute permeability was estimated to be 5 Darcy.
Dry CT scans were performed as a baseline measure at 135 kVp and 100 mA with an
Aquilion One CT scanner. All CT scans for the remainder of the experiment, were
collected at these settings. The experimental setup of the flow cell, magnetic
susceptibility sensor and CT scanner for the waterflood experiment is shown in Figures 1
and 2. A Bartington MS2C coil magnetic susceptibility sensor, connected to an MS2
meter, surrounded the flow cell. The coil sensor could be moved so as to make
measurements at any desired point along the flow cell. The majority of the magnetic
susceptibility signal is contained within a thin disc-shaped slice approximately 16 mm
wide, 8 mm either side of the measure point (the centre of the plane of the sensor coil).
New non-metallic (PEEK) end fittings were manufactured for the flow cell, which
reduced the background noise signal during the magnetic susceptibility measurements.
The flow cell assembly was modified so that the magnetic sensor could be removed after
each CT scan, allowing the re-zeroing of the magnetic sensor prior to each magnetic
scanning sequence.

SCA2015-012

4/12

Table 1 summarises the main conditions of the waterflood experiment. Two pore
volumes (PV) of deionised water were first injected into the sandpack at approximately
60 mL/hour. Each pore volume was approximately 15 cc. CT scans were again performed
at this stage for baseline values. The MS2C magnetic sensor was also employed to gather
baseline readings at this stage. Measurements were taken every 0.5 cm along the 15 cm
length of the sandpack within the vessel. An oil flood, using Lloydminster heavy oil of
20,000 cP viscosity, was then undertaken at an average flow rate of approximately 0.5
mL/hour for 1.2 pore volumes, followed by a set of baseline oil CT scans and a set of
MS2C magnetic sensor measurements. The oil flood is a kind of Swi setting.
Table 1. Summary of the main conditions of the waterflood experiment.

Conditions for Waterflood Experiment

Heavy oil saturation:
Injection rate
Total injected

0.5 mL/hr
1.2 PV

Waterfloods:
Maghemite nanoparticle concentration
DDBS concentration
Sonication
Sodium iodide
FLOPAAM
Shaker for accumulator
Injection rate
Breakthrough (produced) volume
Total deionised water injected
Total oil produced

0.6 wt %
0.81 wt %
20 min
none
none
no
1 mL/hr
0.04 PV
1.75 PV
0.2 PV

For the waterflood, maghemite nanoparticles were added to deionised water at 0.6 wt%
(Table 1) and mechanically agitated. The dispersing surfactant sodium
dodecylbenzenesulfonate (DDBS) was then added at a concentration of 0.81 wt%. The
solution was mechanically agitated and sonicated for 20 minutes in an ice bath prior to
injection into the sandpack. The target injection rate was 1 mL/hour for the duration of
the experiment. A total of 1.75 pore volumes of solution were injected into the sandpack
and 13 sets of CT scans and magnetic sensor measurements were collected at regular
intervals during injection. The nanoparticle tracer is not expected to go into the oil. One
Dean Stark test was performed post experiment to determine oil, water, and solids.

RESULTS AND DISCUSSION

Figure 3 shows the produced oil versus the injected water during the waterflood. Figure
4 shows a few smoothed CT attenuation profiles of differences from the oil saturated scan
for the first 60mm of the sandpack, and Figure 5 shows the final change from the oil

SCA2015-012

5/12

saturated state for the entire length of the sandpack. The vertical scale on Figure 4 has
been expanded to show the differences. Based on the calibrated nanoparticle-oil contrast
of 24 Hounsfield units (HU) and a porosity of about 1/3, the peak value of 4 HU
corresponds to an increase in saturation of 50%. The CT profiles clearly showed an
increase in the attenuation between profiles 1 and 7 consistent with an increased magnetic
susceptibility (due to injection of nanoparticles) as shown in Figure 6. Moreover, the CT
attenuation dropped once the main front (identified from the magnetic susceptibility
results and the material collected in the production jars) passed through the sandpack,
consistent with the magnetic susceptibility profiles of Figure 7. However, the CT profiles
did not exhibit a clearly recognisable front to the waterflood (neither did an earlier
waterflood experiment using a sodium iodide solution). The CT profiles also showed a
region of high attenuation (a hump at 70-80 mm from the inlet) followed by low
attenuation in the downstream part of the sandpack (Error! Reference source not
found.). The latter prevented any meaningful calculation of saturation profiles
downstream of the hump. The low attenuation may have been caused by gas that was
trapped in the pack during the initial water saturation and not fully swept out by the oil
saturation. The CT attenuation responds to the nanoparticles and the fluids, including any
trapped gas. The origin of the hump feature is unclear at present. It remained in the same
position in each CT profile and did not appear to correlate with the magnetic
susceptibility peak (Figure 6) which evolved and moved during the waterflooding as
detailed below.
Eight volume magnetic susceptibility profiles were measured during the first day of nanofluid injection (Error! Reference source not found.6 shows profiles 3-8). The points on
the profiles directly represent the content of maghemite nanoparticles at each point in the
sandpack along the flow cell. The profiles revealed the progressive formation, growth and
movement of a significant positive magnetic susceptibility peak, which initially formed
on the left injection side of the flow cell (Error! Reference source not found.6, profiles 3
and 5) and migrated towards the right production (outlet) side as the injected volume of
nanoparticle suspension increased (Error! Reference source not found.6, profiles 6-8).
The peak is likely the result of a higher concentration of nanoparticles collecting at the
site of the main water-heavy oil front during the flooding process. The peak may not
necessarily reflect a higher water saturation at the peak (compared to portions of the
sandpack closer to the inlet), since it appears that the nanoparticles agglomerate at the
front (from material collected in the production jars as discussed later). First
breakthrough (nano-fluid recovered at the production end) was observed after an
injection of approximately 8.46 ml (0.2 PV injected or 0.04 PV produced) halfway
between profiles 5 and 6. This is consistent with the observed magnetic susceptibility
profiles. Profile 5 shows negative magnetic susceptibility at the production end (right side
of the graph) indicative of the absence of nanoparticles, while profile 6 has a clear
positive magnetic susceptibility signal on the production end, which can only occur if
nanoparticles are flowing through that section. Therefore it appears that breakthrough of
the nanoparticle suspension occurred before the main front, whose progress was tracked
by the magnetic susceptibility peak, reached the production end. This may happen if, for

SCA2015-012

6/12

instance, viscous fingering is occurring. The results of Error! Reference source not
found.6 show how magnetic sensing of a nanoparticle suspension can track the progress
of the main front during water flooding of a heavy oil saturated sandpack. This would
appear to be a significant result. The shape of the front is likely to be complex, but the
overall position of the main front appears to be tracked by the peak in magnetic
susceptibility in each of the curves of Figure 6. The physical mechanism as to why the
nanoparticles agglomerate at the front is not completely understood at present. A review
of processes affecting nanoparticles at fluid interfaces is given by Bresme and Oettel [5].
In our case we think it may be due to a weakening of the repulsive electrical double layer
around the nanoparticles, making them more likely to agglomerate. One possible cause of
the weakening of the electrical double layer could be due to adsorption of hydroxyl ions
at oil-water interfaces as described by Marinova et al. [6].
After the main front had passed through the production end of the flow cell the magnetic
profiles of Error! Reference source not found.7, taken on the second day, show that the
maximum magnetic susceptibility was lower than the peak observed in Error! Reference
source not found.6. This would be expected if nanoparticles are no longer collecting at a
major front. The magnetic profiles shown in Error! Reference source not found.7 also had
a relatively constant shape, which was expected to reflect the porosity profile. Higher
porosity areas should give larger magnetic signals due to the higher volume of
nanoparticles. CT scanning (Error! Reference source not found.8) confirmed that the
porosity profile was very similar to the magnetic profiles. In particular, the decrease in
magnetic susceptibility (Error! Reference source not found.7) just before 60 mm from the
inlet end seems to correspond with a similar decrease in the CT porosity profile (Error!
Reference source not found.8). There is also a correspondence because the Sor value is
quite homogeneous within the sandpack because the brine volume measured is locally a
function of phi*(1-Sorw) and not phi only. Note that the magnetic and CT values are the
result of different thickness slices at each point (16 mm for the magnetics versus 0.35
mm for the CT), and this might explain differences between the two types of profile.
Note also that the magnetic susceptibility values decrease slightly at both the inlet and
outlet ends of the flow cell because the sensor is sensing outside the region of the
sandpack at these points. Figure 7 also indicates that there were slight increases in
magnetic susceptibility with time. This is likely due to an increase in the volume of
magnetic nanoparticles within the pore spaces, as nano-fluid was continuously displacing
the oil.
The growth and migration of the magnetic susceptibility peak shown in Error! Reference
source not found.6 strongly suggests that the nanoparticles accumulated at the main front.
This was further supported by observations of the material in the production jars (Error!
Reference source not found.9). The production jar relating to the main front (JAR #5)
contained nanoparticle agglomerates that had settled out of suspension. Such
agglomerates were not subsequently seen in the later production jars after the main front
had passed through the production end. These later jars (JARS #6 and #7) had
nanoparticles still in suspension.

SCA2015-012

7/12

CONCLUSIONS
1. The formation, growth and movement of a magnetic susceptibility peak during the
waterflood is consistent with a higher concentration of nanoparticles collecting at the
site of the main front during the flooding process. This appears to provide a means of
quantitatively tracking the progression of the front in real time. The aggregation of
the nanoparticles at the front, however, may mean that the magnetic susceptibility
values might not quantitatively relate to the water saturation as the front is passing
through the sandpack.
2. Further independent evidence for a higher concentration of nanoparticles
accumulating at the main front was provided by the material collected in the
production jars.
3. Once the main water/heavy oil front had passed through the production end of the
flow cell the magnetic profiles had a relatively constant shape, which reflected the
porosity profile of the sandpack as confirmed by the CT results. After the main front
has passed through the sandpack the magnetic results should potentially provide a
quantitative measure of water saturation, since the nanoparticles were dispersed and
in suspension at this stage (as seen from the material collected in the production jars)
similar to the originally injected nano-fluid.
4. The CT attenuation profiles do not presently show a recognizable front. Part of the
reason for this could be the relatively low contrast between the water + nanoparticles
and the heavy oil, whereas for the magnetic results there is a substantially larger
contrast between the magnetic susceptibility of the nanoparticles and that of the
sandpack or fluids. It may also be due to the smallness of the change in saturation that
the front represents. In addition, the CT attenuation results appear to have been
corrupted by trapped gas in the downstream part of the sandpack.

ACKNOWLEDGEMENTS
We thank Alberta Innovates - Technology Futures (AITF) for funding for this project, in
particular Marlene Huerta of AITF for all her support.

REFERENCES
1. Da Costa e Silva, A. (1995). Analysis of viscous fingering reproducibility in
consolidated natural porous media. Society of Core Analysts conference, Paper
Number 9504, pp. 110.
2. Mogensen, K., Stenby, E. H., & Zhou, D. (2001). Studies of waterflooding in lowpermeable chalk by use of X-ray CT scanning. Journal of Petroleum Science and
Engineering, 32, 110.

SCA2015-012

8/12

3. Khan, S., Potter, D. K. & Kuru, E. (2015). Quantifying the transport of

superparamagnetic nanoparticles in porous media using an acrylic flow cell and
integrated magnetic susceptibility sensor technique. Transport in Porous Media, 106
(issue 3), 691-705.
4. Ivakhnenko, O. P. and Potter, D. K. (2004). Magnetic susceptibility of petroleum
reservoir fluids. Physics and Chemistry of the Earth, 29, 899-907.
5. Bresme, F. and Oettel, M. (2007). Nanoparticles at fluid interfaces. Journal of
Physics: Condensed Matter, 19 (issue 41), Article Number 413101 (33pp). DOI:
10.1088/0953-8984/19/41/413101.
6. Marinova, K. G., Alargova, R. G., Dencov, N. D., Velev, O. D., Petsev, D. N.,
Ivanov, I. B. & Borwankar, R. P. (1996). Charging of oil-water interfaces due to
spontaneous adsorption of hydroxyl ions. Langmiur, 12, 2045-2051.

MS2C coil magnetic

susceptibility sensor
CT scanner

Flow cell

MS2 meter

Figure 1. The experimental setup on the CT couch for the waterflood experiment, with the flow cell and
magnetic sensor in place.

SCA2015-012

9/12

MS2C coil magnetic

susceptibility sensor
Flow cell

MS2 meter

Figure 2. Close up of the flow cell and magnetic sensor on the CT couch for the waterflood experiment.
Water injection is from left to right.

SCA2015-012

10/12

Figure 3. Produced oil versus injected water during the waterflood. Breakthrough occurred at about 0.04
PV produced or 0.2 PV injected. This was observed before the first production jar was removed, therefore
before the first point on the production curve in Figure 3. Before and after breakthrough, production
behaved as if there were compressible fluid present. Our analysis of the pressure drop across the pack
suggests between 0 and 1 ml of gas (depending on where it may have been lodged) trapped in the pack. The
rest (accounting for the discrepancy between injected and produced volumes) could have been in either the
production or the injection plumbing (or both).

Figure 4. CT attenuation profiles for the upstream portion of the sandpack. Attenuation values are given in
Hounsfield units (HU). Profile 13 was taken about a day after profile 1. Injection is from left to right.

Figure 5. Final change in attenuation from the oil saturated scans observed in the medical CT scanner after
1.75 pore volumes of water + dispersed nanoparticles were injected into the heavy oil saturated sandpack.
The figure shows a small increase in attenuation in the first 60 mm, due mainly to the nanoparticles. The

SCA2015-012

11/12

features beyond 60 mm appear to be artefacts. The hump feature at 70-80 mm is presently unexplained, but
does not correlate with the magnetics. The low attenuation beyond 80 mm appears to be due to trapped gas.

Figure 6. Volume magnetic susceptibility profiles taken during day one of the nano-fluid flooding of a
heavy oil saturated sandpack.

Figure 7. Volume magnetic susceptibility profiles taken during day two of the nano-fluid flooding.

SCA2015-012

12/12

Figure 8. CT scanning porosity profile of the sandpacked flow cell. Note the overall similar shape to the
magnetic susceptibility profiles in Figure 7.

Agglomerated
nanoparticles

Nanoparticles in
suspension

Figure 9. Production jars from the waterflood experiment: initial waterflood JAR #1, oil flood JARS #2
and #3, nanoparticle suspension flood JARS #4 to #7. Production JAR #5 clearly shows that nanoparticles
have agglomerated and come out of suspension (lower thin brown layer). Above this is a yellow layer
which comprises water plus some nanoparticles, and this is overlain by a heavy oil layer (black layer). The
later production JARS #6 and #7 clearly show nanoparticles in suspension and not separated out.

SCA2015-013

1/12

CHARACTERIZING FLUID PRESENCE AND

TRANSPORT IN ROCK CORES AT RESERVOIR-LIKE
CONDITIONS VIA SPATIALLY RESOLVED NMR
RELAXATION/DIFFUSION MAPS
Huabing Liu1, Mark Hunter1,2, Sergei Obruchkov1, Evan McCarney2, Mitch Robison3,
Robin Dykstra1,2, Petrik Galvosas1
1
Victoria University of Wellington, SCPS, MacDiarmid Institute for Advanced
Materials and Nanotechnology, Wellington, New Zealand,
2
Magritek Ltd., Wellington, New Zealand
3
University of Auckland, Auckland, New Zealand
This paper was prepared for presentation at the International Symposium of the Society of
Core Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
The properties of fluids saturating rocks depend strongly on temperature and pressure.
Therefore, ambient laboratory conditions may not be desirable for the investigation of
fluids in reservoir rocks. To mimic the reservoir, a pressurized and temperature
controlled (overburden) cell, compatible with Nuclear Magnetic Resonance (NMR),
was assembled within a 2 MHz NMR Rock Core Analyzer. 1D NMR relaxation and
2D diffusion-relaxation correlation distributions of fluid-saturated rock cores were
measured in conjunction with 1D NMR imaging. By performing these spatially
resolved NMR relaxometry and diffusometry experiments within the environment of
the overburden cell, it is possible to obtain porosity, fluid saturation and residual fluid
content profiles under reservoir-like conditions. Here we show for the first time that
these spatially resolved NMR relaxometry and diffusometry experiments could be
performed under elevated pressures and temperatures for fluids flooding rock cores in
the laboratory. Experiments performed with this NMR setup may allow one to study
the oil properties under reservoir conditions, which may inform oil recovery
enhancement strategies.

INTRODUCTION
NMR is one of the non-invasive techniques to deliver substantial petrophysical
parameters, such as pore volume, permeability and fluid saturation in rock plugs [1].
In particular, low-field 1H NMR has become an important tool to study rock core
plugs from oil reservoirs, and is established as an industry standard to calibrate NMR
well-logging data [2]. In this context, the relaxation time/diffusion coefficient
distributions, as well as their multi-dimensional correlation maps, are routinely
utilized to characterize pore size, wetting state, fluids types, and quantify individual
fluid phase saturations [3-5]. Meanwhile, magnetic resonance imaging (MRI) was
introduced in rock core analysis to monitor real-time fluid invasion profiles during
core flooding experiments [6]. To reveal the spatial fluid properties with
saturating/flooding multi-phase fluids, it is desirable to combine NMR imaging
techniques with 1D relaxation time/diffusion coefficients or 2D diffusion-relaxation
correlation measurements. Subsequently, it allows one to extract the local profiles of

SCA2015-013

2/12

pore structure, saturation of individual fluid phases as well as wettability information

[7-9].
Most reservoir rocks are buried in the formation approximately thousands meter deep,
where the in-situ temperatures and pressures are much higher as compared to
conditions on the surface [10, 11]. In this case, fluid properties (e.g. viscosity) in
porous rocks behave differently according to local pressure and temperature, leading
to a complex mechanism of oil recovery in oilfields [12]. Therefore, in order to
characterize and model fluid properties and transport, it is necessary to confine and
flush rock cores at reservoir-like temperatures and pressures whilst NMR experiments
are performed in laboratory measurements.
In this work, an overburden system was assembled to pressurize rock cores at
required temperatures and pressures. Furthermore, spatially resolved NMR techniques
were combined with relaxation time (T1 and T2) and relaxation-diffusion correlation
(D-T2) experiments and implemented on a low-field NMR rock core analyzer. Since
NMR signal intensity is inversely proportional to temperature, imaging was restricted
to one dimension (1D) only, to ensure sufficient Signal-to-Noise Ratio (SNR). The
property and distribution of oil saturating rock core were studied at different
temperatures before flooding experiment. Subsequently, the conditions of secondary
oil recovery were investigated by flooding water through an oil-bearing rock plug.

HARDWARE CONFIGURATION
A commercial 2 MHz NMR analyzer equipped with a shielded 1D gradient coil
(Magritek, Ltd.) is used to acquire NMR data in this work [13]. The RF probe has a
dimension of 10 cm length and 5.4 cm diameter. This setup is complemented with a
Daedalus Innovations overburden rock holder (Figure 1 a), which maintains required
temperatures and pressures for the rock plug under study [14]. It is worth noting that
the materials of the rock holder, especially the section close to the detection volume
of the NMR system, are chosen to be non-magnetic to avoid disturbing NMR signals.
Before the experiments, the rock core is placed between two PEAK mounts first,
wrapped by heat shrink sleeve, and then loaded in the chamber of overburden holder
(Figure 1 a). Two parts are screwed in each end of the holder housing to seal the
overburden cell. Subsequently, the overburden cell within rock plug is inserted in the
bore of NMR system, and is connected to two external circuits. These two circuits
will provide pressurized fluid flooding through rock samples (left plumbing in Figure
1 b), and the confining pressure and temperature for rock plug (right loop in Figure 1
b), respectively. To achieve the desired confining temperature and pressure for the
rock plug, two ISCO syringe pumps are used to drive confining fluid flow (right part
in Figure 1 b). Perfluro polyether (PFPE, non-hydrogen signal) is chosen as the
confining fluid to circulate in the plumbing and rock holder chamber. A heating bath
after the syringe pump ensures the required temperature of PFPE before entering the
pressure cell. Another bath after the cell will cool down the confining fluids before
entering a back pressure regulator (BPR 1) which controls the pressure surrounding
the rock core. On-site pressures and temperatures are logged to monitor the real-time
confining conditions of the loaded rock core. Furthermore, a high-pressure crank

SCA2015-013

3/12

pump is connected to the inlet end of the flooding channel of the overburden rock
holder to provide flooding conditions. Water can be imbibed and then pressurized
through the rock core in this work.

(a)
(b)
Figure 1. Schematic of overburden rock holder (a) and its combination with low-field NMR system for
high pressure and temperature experiments (b). Red and blue arrows indicate the flow direction of
flooding fluid and confining fluid, respectively. The bold line in the confining part represents the
tubing insulation. P in the circuit stands for pressure gauge and T for thermocouple.

Thermal conditions arising from the core holder within the NMR system have been
modeled and simulated for different temperatures with SolidWorks. Results obtained
by these simulations confirm the compatibility between operational NMR system and
the overburden system.

NMR METHODOLOGIES
MRI methods are generally based on either frequency or phase encoded imaging [15].
A magnetic field gradient is applied during signal acquisition for frequency encoded
imaging, ensuring spatially distributed resonance frequencies which return the image
of the sample upon Fourier transform of the acquired MRI signal. The field of view
(FOV) for this imaging method is determined to be: 2/gt, where g is the gradient
strength and t is the time interval between neighboring acquired points (aka dwell
time). In order to obtain an image using the phase encoded MRI method, a set of
gradients within variable amplitudes is applied during a fixed time period before
acquisition. This returns a set of experiments with varying phases in the acquired MRI
signal, which again provides an image after Fourier transform. In this case, the FOV
is determined as m/gmax with gmax being the maximum intensity of the phase
gradient. The imaging resolution Z is FOV/m, where m is the number of acquired
points in frequency-encoded imaging and the number of gradient steps in phase
encoded imaging, respectively.
Since the spatial information can be encoded in a one-shot signal acquisition under
fixed amplitude of the imaging gradient, frequency-encoded methods are generally
faster as compared to phase-encoded. However, T2 effects will irreversibly impact the

SCA2015-013

4/12

obtained imaging profile during the acquisition time, leading to the absence of signal
with short T2 relaxation time. The phase-encoded method on the other hand uses short
signal acquisition time on the cost of experimental time, and can therefore offer more
accurate imaging profiles. Both methods were employed in this work, depending on
required time efficiency and accuracy.
In the context of petrophysical analysis using low-field NMR measurements, the
longitudinal relaxation T1 and transverse relaxation T2 are mostly studied for fluidsaturated rock cores. This may provide information on the pore volume and pore size
distributions. The 1D phase encoded imaging technique combined with T2 relaxation
measurements, as shown in Figure 2, can yield spatially resolved T2 profiles [16]. The
first period covered by TE1 is the phase encoded imaging part of the experiment,
providing the 1D image (profile). The following section, consisting of the 180 pulse
train acquires the NMR echo decay, which is determined by the transverse relaxation
T2. Similarly, spatially resolved T1 profiles can be obtained by measuring T1 using
inversion/saturation recovery or rapid acquisition schemes as suggested in [17, 18],
again in conjunction with the 1D imaging method.

Figure 2. NMR pulse sequence for spatially resolved T2 profiles by using phase encoded methods [16].
g is the imaging gradient which has variable intensities with a duration of . TE is the echo spacing
during the acquisition of the T2 measurement.

The acquired data for spatially resolved T2 experiments can be expressed as

M ( k, nTE ) =

F ( z,T2 ) ei2p kz e

nTE
T2

dzdT2 ,

(1)

while the data obtained by spatially resolved T1 techniques (using rapid acquisition) is
M ( k, nTACQ ) =

F ( z,T ) e
1

i2 p kz

- nTACQ

T1
n-1

e
sin a cos a dzdT1

(2)

with k=g/2. z is the gradient direction which coincides with the cylindrical axis of
the rock cores in this work. represents the small tip angle and TACQ is the signal
evolution time in rapid encoding T1 method.
Both spatially resolved T1 and T2 imaging methods yield two-dimensional datasets
and a two-step data process is needed to extract the spatial relaxation distribution

SCA2015-013

5/12

function F(z, T1) or F(z, T2). Firstly, Fourier transform is employed to transform the
encoded phase information into positions, returning the fluid content in each resolved
slice (pixel). The Inversion Laplace transform algorithm is then consecutively applied
slice by slice, yielding the spatial resolved relaxation distribution function. It is worth
noting that F(z, T1) and F(z, T2) are akin to the curves from NMR well logging
operated in the oilfield, where the sensor is pulled from the down-hole up to the
surface and provide relaxation distributions from each vertical layer in the formation.
Besides relaxation measurements, diffusion coefficients of fluids can be uniquely
measured by NMR techniques [15, 19]. This allows one to identify different fluid
types and quantify fluid saturation in rock samples. In this case, 2D D-T2 NMR
correlation technique provides substantial information in rock core analysis, such as
fluid typing and wettability identification in multi-phase experiments. Furthermore,
spatially resolved D-T2 distribution provides the aforementioned information along a
certain core axis and potentially indicates progress of oil recovery or monitors the
saturation level of rock cores [8]. While incorporating 2D relaxation-diffusion
correlation experiment with 1D imaging technique, the entire experimental time
should be taken into account for practical reasons. Therefore, the frequency encoding
method is preferred because of the equivalent experimental time as compared to 2D
correlation experiment without spatial resolution.
The pulse sequence for spatially resolved D-T2 distribution is shown in Figure 3 and
the signal decay is expressed as
M ( k,G, nTE ) =

F ( z, D,T ) e
2

i2 p kz

d
-g 2d 2G 2 D D-
3

nTE
T2

dz dD dT2

(3)

where k=gtp/2. is diffusion observation time and is gradient duration. The result
is a 3D data matrix acquired during the spatially resolved D-T2 experiment. To obtain
the final map, Fourier transform is performed on each acquired echo to obtain the
spatial imaging profile. Then 2D ILT algorithm is applied subsequently for the 2D
exponential decay data in each slice, in order to extract the local D-T2 distribution
function F(z,D,T2) [20].

Figure 3. NMR pulse sequence for spatially resolved D-T2 map, details are shown in [8]. is the
diffusion observation time and is the gradient duration. G and g are the gradients for encoding
diffusion coefficient and position, respectively. TE is the echo spacing and tp is the echo acquisition
time.

MATERIALS AND EXPERIMENTAL STRATEGIES

Sample. A Edwards Brown limestone core was used in this work to study fluid
presence and multi-phase fluid flooding at different temperatures. The core porosity is

SCA2015-013

6/12

36.5% and the permeability is 90 mD. The rock core has a dimension of 6.3 cm length
and 3.84 cm diameter. Before the experiment, the core plug was dried in an oven at
60C for 48 h, until no hydrogen signal was detected in NMR measurements.
Afterwards it was saturated with diesel (Hydrogen index HI=1.023).

(a)
(b)
(c)
Figure 4. Edwards Brown plug used in this work (a), its porosity profile (b) determined by 1D
imaging, and its pore length scales (c) determined by DDIF technique [21]. The heterogeneity of the
rock sample can be recognized from the comparison of the acquired porosity profile and weighing
porosity indicated by a dashed line. Large grains existing close to the plug middle might lead to the
section with reduced porosity. This rock plug has a wide range of pore length from approximately 1
m to 100 m, indicating a highly heterogeneous pore system.

Basic parameters in NMR experiments. The duration of 90 and 180 pulses in the
NMR experiments were kept to be 25 s (to ensure the same RF-pulse frequency
bandwidth) while the amplitudes were adjusted accordingly. T1, T2 and D-T2
experiments were performed as routine measurements for determining the global
properties of the fluid in the rock core. Relaxation time T2 distributions were
determined using the CPMG [22] pulse sequence with an echo spacing of 150 s and
12000 echoes. A dataset with SNR=450 was acquired within a 4-scan measurement
lasting 45 s. In the fast encoding T1 measurement, the small tip angle was set to be
5.3 and the number of small tip angle pulse is 250. The acquisition time TACQ was
optimized to be 25.6 ms. The entire experiment time was 6 min within 32 scan
acquisition. D-T2 correlation distributions were obtained with a gradient duration
= 5 ms and diffusion observation time = 40 ms. The gradient G for encoding
diffusion varied up to 0.4 T/m in 40 steps linearly. The echo spacing was 150 s and
the number of echoes was 12000.
Before flooding: static measurement. Because of the smaller impact of T2 relaxation
during the imaging encoding period, the phase-encoded T2 method was used to extract
spatially resolved T2 profiles in the case of static fluid measurement. The spatial
resolution was 1.5 mm and the number of imaging gradients m was 64. The echo
spacing in the first stage TE1 was set to be 500 s and the duration of phase gradient
was 100 s. The echo spacing in the second stage was set to be 150 s and the
number of echoes was 12000. With an 8-scan measurement for each imaging gradient
step, phase-encoded T2 profile datasets have been acquired within 1.5 h. For phase
encoded T1 measurement, the spatial resolution was 3 mm and the number of imaging

SCA2015-013

7/12

gradients m was 32. The data can be acquired within 3.5 h with a 32-scan
measurement for each imaging gradient step.
Before flooding, the confining pressure was set to be 1000 psi by adjusting BPR 1 in
the confining circuit. To study the temperature-dependence of oil properties and
distribution in rock plug, the temperatures of rock sample were set to be 25C, 55C
and 80C with a constant flow rate of 150 mL/min for the confining fluid PFPE
heated in the heating bath. During this measurement, the inlet and outlet of flooding
channel were both switched off to avoid the saturated oil being driven out during
providing confining pressure.
Flooding measurements. In the flooding experiment, to obtain spatially resolved DT2 distribution efficiently, frequency encoded imaging method was used. Parameters
have been optimized to ensure comparable results obtained from frequency encoded
T2 profiles as compared to phase encoded T2 method. The intensity of the frequency
gradient g was chosen to be 12 mT/m. The number of acquired data points for each
echo was 256 and the dwell time was 2 s. This ensured the acquisition of the full
echo (as determined by T2*, which is close to T2 at low-field NMR measurement). The
echo spacing with TE = 800 s was chosen accordingly to suit the acquisition of the
full echo. The number of echoes was 3000. For the acquisition in the diffusion
dimension, the gradient duration was 5 ms and the diffusion observation time was
40 ms. The gradient G for diffusion encoding varied up to 0.4 T/m in 12 steps linearly.
The experimental time was 20 min with a 16-scan acquisition.
To investigate the oil recovery procedure, water was injected into the oil-saturated
rock core confined in the overburden cell. During the flooding experiment, to avoid
water bypassing the core and travel between the core sample and confining fluid, the
overburden pressure in this work was set to be at least 300 psi higher than the inlet
pressure of crank pump. Therefore, the pressure before BPR 1 was kept to be 1000 psi
and the initial pressure for crank pump was 700 psi. The pressure before BPR 2 was
set to be 500 psi in this work. The entire flooding procedure is carried out in small
individual steps as visualized on the horizontal axis in Figure 8. The crank pump
needs to be suspended after each flooding step. Spatially resolved D-T2 NMR
measurements are carried out after there are no reading changes from these two
pressure gauges between the confined rock core. The above procedure is repeated till
there is no noticeable decrease of the oil signal detected as estimated from spatially
resolved D-T2 maps. Subsequently, the confining temperature of the prepared rock
plug is increased and a further set of flooding and NMR measurement is performed, in
order to investigate the temperature dependence of water-flooding-oil process of rock
core. Three experimental temperatures, 25C, 55C and 80C, will be adopted in this
work.

POROSITY PROFILE AND FLUID DISTRIBUTION

Before spatially resolved measurements were performed, T2, T1 and D-T2 distributions
at three temperatures under 1000 psi were acquired (Figure 5). T2 distributions at the
three temperatures indicate two components of oil in the pore space (light component
corresponds to larger T2 value while heavy components has a smaller T2 value). With

SCA2015-013

8/12

increasing temperature, T2 distribution gradually shifts towards longer relaxation time,

which implies decrease of oil viscosity [23, 24]. When the sample confining
temperature was increased from 55C to 80C, the heavy component was less shifted
as compared to the light counter-part, which might be caused by weaker temperature
dependence as compared to light one. Compared to T2 distributions, T1 results
obtained from rapid encoding method have only the distribution from light component.
This is caused by long signal evolution time TACQ used in the signal acquisition.
Likewise, T1 values progressively increase with heating up the rock core, which
suggest the decrease of oil viscosity.

(a)
(b)
(c)
Figure 5. (a) Transverse relaxation T2 and (b) longitudinal relaxation T1 distributions of oil
saturating rock under 1000 psi. The dash, solid and dot dash lines represent the distributions at
25C, 55C and 80C respectively. The regularization factor during the inversion is 1 in T2 and 33
in T1 data process. (c) Diffusion-relaxation time correlation D-T2 distribution under 1000 psi. The
slope dash line represents hydrocarbon correlation line of D=510-10 T2. The T2 distribution of bulk
diesel at 25C is given in the inset of (a).

The diffusion-relaxation time correlation distribution were obtained and presented in

Figure 5 c. Because of relaxation effect during the diffusion observation time, the
heavy component with short T2 relaxation time in the correlation experiments were
missing compared to 1D relaxation results. Nevertheless, a reasonable prediction of
oil properties is still accessible since the light component predominates this oil as
shown in 1D T2 spectra. Meanwhile, these 2D distributions exhibit a strong
correlation between diffusion coefficient D and relaxation time T2, a feature that is
similar to that in bulk oil. It indicates non oil-wet status for this rock plug. This is
probably ascribed to the thin water film absorbed at the pore walls that have not
entirely dried out during the sample preparation. This amount of water, although was
hardly to be detected during the experiment, results in the wettability of this rock plug.
The distributions shift along the hydrocarbon line even at temperatures of up to 80C,
which proves a decrease of oil viscosity. While the shape of D and T2 distributions
remain almost the same, indicating a nearly stable composition of oil in pore space
within our temperature range [24].
Spatially resolved T1 and T2 profiles at 25C are presented in Figure 6 a and b,
respectively. Similar features, such as the heterogeneity of porosity profile, can be
observed along the cylindrical axis in both maps. Because of the long TACQ time
adopted here, heavy component of oil is absent in all resolved T1 profiles, while is still
visible in T2 profiles. Both projected T1 and T2 distributions are comparable with the
bulk results shown in Figure 5 a and b. Furthermore, spatially resolved T2 experiments

SCA2015-013

9/12

were performed at confining temperatures of 55C and 80C in order to investigate

the oil temperature-dependent behaviour. The varying T2 and spatial features as seen
at 55C and 80C are not understood yet and remain subject to further research.
However, projections onto the spatial and T2 domains remain reasonable and confirm
a shift of T2 towards higher temperatures as expected.

(a)
(b)
(c)
(d)
Figure 6. (a) Spatially resolved T1 profile at 25C and T2 profiles at (b) 25C, (c) 55C and (d) 80C
under 1000 psi. The top panel presents the projected T2 relaxation distribution and right panel gives the
porosity profile. The blue dashed line in the right panel corresponds to the weighing porosity value.

In summary, the oil situated in the rock core becomes less viscous with increasing the
confining temperature, which is revealed from T1, and T2 distributions as well as from
the D-T2 correlation distributions. The spatial T2 profile resolves the T2 distributions
along cylindrical core dimension, also exhibiting distinct oil properties and
distributions at different confining temperatures.

IN-SITU WATER-FLOODING MEASUREMENTS

In this section, multi-phase fluid transport processes at different temperatures are
observed by using the spatially resolved D-T2 technique. The oil-saturated rock plug
was firstly flooded by water at 25C and 1000 psi. Subsequently, two further flooding
steps at 55C and 80C were performed to investigate the temperature dependence
during the flooding process. The spatially resolved D-T2 profiles before, during and
after water flooding at 25C are compared and presented in Figure 7. In order to
visualize the existence of two different fluid phases, two reference planes indicating
oil (left) and water (right) were added in to the figures as grey planes.
The distribution function F(z,D,T2) before water flooding (shown in Figure 7 a)
mainly extends along the hydrocarbon plane indicating oil saturating from top to
bottom in the rock plug. After flooding an amount of water corresponding to 0.1 pore
volume through the rock (0.1 PV water flooding), there appears a distinct distribution
neighbouring the oil signal and laying along water reference plane (Figure 7 b). This
signal extends along the water plane after 1 PV water flooding (shown in Figure 7 c).
From the reference planes, it can be distinguished that the left column represents
residual oil while the right one is the injected water. Therefore, it is feasible to extract
the spatially resolved residual oil and water saturation from this distribution, and
estimate oil recovery efficiency. The volume of flooding water was incremented up to
2.2 PV and two subsequent flooding experiments at 55C and 80C were performed
with equal amount of water flooding.

SCA2015-013

10/12

(a)
(b)
(c)
Figure 7. Spatially resolved D-T2 profile before water flooding (a), and after 0.1 PV (b) and 1 PV (c)
water flooding at 25C. Two reference planes indicating hydrocarbon and water were plotted in
order to identify the spatial distributions of oil and water signals.

The intensities of oil and water signals in the spatially resolved D-T2 maps of each
water flooding step can be readily summed up and plotted as relative water and oil
saturation (shown in Figure 8).
Figure 8. Relative water (red diamond) and oil
(green square) saturation estimated from the
spatially resolved D-T2 maps. The entire
flooding procedure is performed under three
confining temperatures (25C, 55C and 80C).
The amount of flooding water is normalized
with the pore volume (PV) of the rock plug
here. During the flooding in each temperature,
the water volume is incremented to 2.2 PV.

For the water-flooding procedure at 25C, the relative oil saturation drops rapidly
with increasing the flooding water volume, and reaches to a plateau level after 1 PV
water injection. The further flooding shows no obvious oil yield in this rock plug. By
increasing the rock temperature, the relative oil saturation continues to decrease
because of the higher oil mobility, and quickly converges to a constant level
afterwards. The residual oil saturation is still above 50% even at rock temperatures up
to 80C, which is a higher value as compared to common water-flooding-oil
experiments (lower than 40% of residual oil saturation). This may due to the
relatively high flow rate of the flooding water (approximately 0.25 mL/s) provided by
the crank pump in this case. In order to investigate the multi-phase fluid interaction,
all distributions under three temperatures were projected into the D-T2 dimension and
shown in Figure 9.
Regardless of longitudinal relaxation during the diffusion observation time, the ratio
of water and oil varies gradually with increasing the rock temperature, which can be
observed from the projected diffusion coefficients in the right panel of each
distribution. These results exhibit the same features as shown in the relative saturation
data in Figure 8, and imply a process of enhanced oil recovery by flooding the rock

SCA2015-013

11/12

plug at higher temperatures. Further oil recovery strategy, such as gas, thermal or
chemical polymer injection may be justified.

(a)

(b)

(c)

Figure 9. Projected diffusion-relaxation correlation distribution at (a) 25C, (b) 55C and (c) 80C. The
horizontal lines represent water diffusion coefficient at 25C (=2.310-9 m2/s), at 55C (=4.410-9 m2/s),
and at 80C (=6.610-9 m2/s) [25]. The slope dash line represents hydrocarbon correlation line of
D=510-10 T2. The top panel presents the projected relaxation distribution and right panel shows the
projected diffusion coefficient distribution.

CONCLUSION
Based on the constructed high pressure/temperature NMR measuring system, spatially
resolved T1 and T2 relaxation techniques provides direct insight into local properties
of fluid situated in the pores of rock core at reservoir-like conditions. Furthermore, a
spatially resolved D-T2 method yields substantial information regarding flooding
experiments, such as wetting status, residual oil/water saturation profiles, and their
dependence on temperature. The current hardware set-up and the implemented NMR
methods provide the base for further investigations of the residual oil saturation and
displacements in the context of enhancement oil recovery.

ACKNOWLEDGEMENTS
The project was supported by the New Zealand Ministry of Business, Innovation, and
Employment via the Grant "New NMR Technologies". Magritek Ltd. is
acknowledged for providing the overburden cell. R. Peterson (Daedalus Innovations
LLC), S. Buchanan (Magritek Ltd.) and T. Brox (VUW) are acknowledged for the
technical suggestions. H-B. Liu thanks the financial supports from Chinese
Scholarship Council (CSC).

REFERENCES
1.
2.
3.
4.

Dunn, K.-J., D.J. Bergman, G.A. Latorraca, Nuclear magnetic resonance:

Petrophysical and logging applications, Pergamon, (2002).
Mitchell J., L.F. Gladden, T.C. Chandraseker, E.J. Fordham, Low-field
permanent magnets for industrial process and quality control, Progress in
Nuclear Magnetic Resonance Spectroscopy, (2014) 76, 160
Hrlimann M.D, The diffusionspin relaxation time distribution function as an
experimental probe to characterize fluid mixtures in porous media, The Journal
of chemical physics, (2002) 117, 22, 1022310232.
Sun, B.Q., M. Olson, J. Baranowski, S. Chen, W. Li, D. Georgi, Direct fluid
typing and quantification of Orinoco Belt heavy oil reservoirs using 2D NMR
logs, SPWLA 47th Annual Logging Symposium, Veracruz, Mexico, June 4-7,
(2006).

SCA2015-013

5.

6.

7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.

12/12

Hrlimann, M.D., M. Flaum, L. Venkataramanan, C. Flaum, R. Freedman, G.J.

Hirasaki, Diffusion-relaxation distribution functions of sedimentary rocks in
different saturation states, Magnetic Resonance Imaging, (2003) 21,3-4, 305
310.
Kim, K., S.-H. Chen, F.-F. Qin, A.T. Watson, "Use of NMR Imaging for
determining fluid saturation distributions during multiphase displacement in
porous media." International Symposium of the Society of Core Analysis, (1992)
9219.
Sun B.-Q., K.-J. Dunn, G.A. LaTorraca, D.M. Wilson, NMR imaging with
diffusion and relaxation, International Symposium of the Society of Core
Analysts, Pau, France, September 21-24, (2003), SCA2003-24.
Rauschhuber, M., G. Hirasaki, Determination of saturation profiles via lowfield NMR imaging, International Symposium of the Society of Core Analysts,
Noordwijk, Netherlands, September 2730, (2009), SCA2009-09.
Mitchell, J., Magnetic resonance core analysis at 0.3 T, International
Symposium of the Society of Core Analysts, Avignon, France, September 811,
(2014), SCA2014-10.
Utracki, L.A., Temperature and pressure dependence of liquid viscosity, The
Canadian Journal of Chemical Engineering, (1983) 61, 5, 753-758.
Dandekar A.Y., Petroleum reservoir rock and fluid properties, CRC press,
(2013).
Tiab D., E.C. Donaldson, Petrophysics: theory and practice of measuring
reservoir rock and fluid transport properties, Gulf professional publishing,
(2011).
http://www.magritek.com.
http://www.daedalusinnovations.com/apparatus/rockcore.html.
Callaghan, P.T., Principles of nuclear magnetic resonance microscopy, Oxford
University Press, Inc., New York, (1991).
Perlo J., F. Casanova, and B. Blmich, 3D imaging with a single-sided sensor:
an open tomograph, Journal of Magnetic Resonance, (2004) 166, 228235.
Hsu, J.-J., I. J. Lowe, Spin-lattice relaxation and a fast T1-map acquisition
method in MRI with transient-state magnetization, Journal of Magnetic
Resonance, (2004) 169, 270-278.
Chandrasekera T.C., J. Mitchell, "Rapid encoding of T1 with spectral resolution
in n-dimensional relaxation correlations", Journal of Magnetic Resonance, (2008)
194, 1, 156-161.
Tanner, J.E., "Use of the stimulated echo in NMR diffusion studies," The
Journal of Chemical Physics, (1970) 52, 5, 2523-2526.
Venkataramanan, L., M.D. Hrlimann, Y-Q. Song, Solving Fredholm integrals
of the first kind with tensor product structure in 2 and 2.5 dimensions, IEEE
Transactions on Signal Processing, (2002) 50, 5, 10171026.
Song, Y.-Q., S. Ryu, P. N. Sen, Determining multiple length scales in rocks,
Nature, (2000) 407, 178-181.
Meiboom, S., D. Gill, "Modified spin-echo method for measuring nuclear
relaxation times," Review of Scientific Instruments, (1958) 29, 8, 688-691.
Freed, D.E., M.D. Hrlimann, One-and two-dimensional spin correlation of
complex fluids and the relation to fluid composition, Comptes Rendus Physique,
(2010) 11, 2, 181191.
Freed, D.E., Temperature and pressure dependence of the diffusion coefficients
and NMR relaxation times of mixtures of alkanes, The Journal of Physical
Chemistry B, (2009) 113, 13, 4293-4302.
Holz, M., S. R. Heil, A. Sacco, Temperature-dependent self-diffusion
coefficients of water and six selected molecular liquids for calibration in
accurate 1H NMR PFG measurements, Physical Chemistry Chemical Physics,
(2000) 2, 4749-4742.

SCA2015-014

1/12

TRACER DISPLACEMENTS FOR EOR EXPERIMENTS:

THEORY, MODELING, AND EXPERIMENTS.
Roland Lenormand, Guillaume Lenormand and Fabrice Bauget
Cydarex
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August 2015

ABSTRACT
During Enhanced Oil Recovery (EOR) experiments, front spreading is due to various
mechanisms: pore-scale mixing, channeling due to permeability heterogeneity, viscous
fingering, capillary effects, etc. This paper gives a review of existing models for the
simplest case of tracer spreading and a discussion of their validity based on comparison
with laboratory experiments.
We first recall the definitions and the physical mechanisms describing pore-scale mixing
and spreading, and their modeling at the scale of a Representative Elementary Volume
(REV, local or Darcy-scale approach) called "microdispersion". When used as local input
in 3-Dimensional (3-D) simulations, the Darcy-scale approach can describe the plug-scale
spreading related to permeability heterogeneity. In a 3-D heterogeneous medium, even a
displacement started as a piston-type at the Darcy-scale leads to a dispersion behavior.
However, laboratory experiments are rarely simulated with 3-D models, but only with 1D models. A local microdispersion displacement in a 1-D simulation cannot model the
large spreading observed at the plug-scale. Consequently, additional parameters must be
introduced in 1-D local laws to compensate for the missing information on
heterogeneities, and we recall here the existing up-scaled (or homogenized) models
(microdispersion, macrodispersion and convective channeling).
These models are compared to our tracer experiments. Displaced and injected fluids have
different salinities, and local saturation profiles were measured using ten electrodes along
the sample. Our results confirm previously published results that channeling convective
models are more suitable than standard dispersion models. The best result is obtained by
calculating a permeability distribution function in a stratified model. This model accounts
for the long tail in the effluent production and the in-situ concentration profiles.

INTRODUCTION
During Enhanced Oil Recovery (EOR) experiments, front spreading is due to various
mechanisms: pore-scale mixing, channeling due to permeability heterogeneity, viscous
fingering, capillary effects, etc.
The main objective of this paper is to clarify the mechanisms linked to the spreading of a
tracer in a porous medium and to present the equations used to model the concentration as
a function of time and space (transport equation) at the scale of a plug of real rock. We

SCA2015-014

2/12

will show that a transport equation written at the scale of a REV needs to be modified
when applied to a plug, due to the presence of heterogeneities. This effect, called
homogenization (up-scaling), will be discussed, and existing models will be compared to
our experiments. We first give some definitions, then present existing models used at the
REV scale, and finally we describe homogenized transport equations.
Definitions
A tracer is a soluble substance, like dye or salt, which at low concentration does not
change the fluid properties. Dispersion is used as a general term equivalent of
spreading, independently of fluids characteristics: miscible or not, with or without
viscosity contrasts. Originally diffusive spreading was referring to the physical
mechanism of molecular diffusion leading to a transport equation with a second order
derivative with respect to space. By extension, this term is now generally used when
spreading is described by a term of transport equation with a second order spatial
derivative, leading to the width of the front being proportional to the square root of the
travelled distance, even if the physical mechanism is not molecular diffusion. We will see
that an uncorrelated velocity field also leads to a front spreading with square root of
travelled distance. When molecular diffusion is negligible, spreading is due to spatial
differences in velocities, either at pore level or at larger scale (permeability
heterogeneities) and dispersion is called hydrodynamic (or advective).
In general, a transport equation in concentration as function of time and distance contains
three terms: the transient term with a derivative with respect to time, a term with a first
order derivative in space, called the "convective" term, and a term with a second order in
space, called the dispersive term.
Flux and concentration
All the equations are written at the Darcy scale, meaning that a REV of porous medium
contains a large number of pores (such as the micromodel in Figure 1).
The tracer concentration is the mass of tracer per unit volume. The flux is the
mass of tracer crossing a given surface per unit of time. The flux density is the flux
per unit surface.
In an elementary volume (length dx in the direction of flow, cross-section area dA,
porosity , and volume = ), the mass balance equation implies that the
accumulation of solute during , , is equal to the quantity of tracer
entering in , (), minus the quantity of tracer leaving the volume, ( +
). This leads to the 1-D mass balance equation, valid with no tracer adsorption:

1
+ = 0.
(1)

In order to derive the concentration as function of time and distance, the dispersion
equation, relating the flux density to the concentration is required. In addition to the
mass balance and the dispersion equation, Darcy's equation is used to calculate the local
velocity as function of fluid properties and pressure gradient.

SCA2015-014

3/12

In the following sections, we will study the local dispersion equation for tracer flow. Then
we will describe the integration of this local equation for a 2 or 3-D flow, leading finally
to the 1-D homogenization, average in a 1-D flow as commonly used to represent
laboratory experiments.
Tracer local equation (Microdispersion)
Spreading in the direction of the mean
flow is due to the combined effect of the
variation of velocities inside the pores
(the tracer follows the streamlines) and
molecular diffusion between streamlines
that smears the color near the entrance
(Figure 1). The molecular diffusion
allows the invasion of the tracer in deadend pores and zones with very low
velocities.
Figure 1- Tracer flow experiment in a micromodel. We can derive the dispersion equations
The tracer in red is injected at a constant flux on the (relationships
between
flux
and
left
concentration) in some limiting cases:
Piston-type flow: A theoretical case of transport without any dispersion (not possible in
porous media). All the particles are transported at the same Darcy velocity U. The flux
density is therefore: = .
Pure molecular diffusion: With no velocity, spreading is due to molecular diffusion and
Ficks law gives the flux density proportional to the concentration gradient. In a liquid,
the coefficient of proportionality is the molecular diffusion . In a porous media, this
coefficient is an apparent diffusion coefficient . Its ratio with reflects the presence
of solid and pore space tortuosity: = 1 where is formation factor: =

.
Hydrodynamic dispersion: With no molecular diffusion, the spreading due to pore scale
variations in the velocity field can be described using a stochastic approach. If the
distribution of pore size is random with no spatial correlation, the invasion of the tracer
follows the statistic law of large numbers of uncorrelated events, and the spreading is
proportional to the square root of distance (similar to the macrodispersion described
below). This mechanism can be represented by a transport equation, sum of a convective
term (plug flow without spreading) and a diffusive term leading to the spreading in square
root of travelled distance with a dispersion term :

= .
(2)
When both molecular diffusion and pore scale velocity field act together, the mechanism
is called microdispersion and the flux density is assumed to have the same form as Eq.

SCA2015-014

4/12

2 but with a dispersion coefficient D that includes molecular diffusion and hydrodynamic
dispersion.
Effect of molecular diffusion on microdispersion
Microdispersion involves both molecular diffusion and hydrodynamic dispersion
(Pfannkuch in 1963 [1]; Fried and Combarnous [2]). The balance between molecular
diffusion and hydrodynamic dispersion is quantified using the molecular Pclet number,
defined as the ratio of the molecular diffusion characteristic time 2 / to the advective
characteristic time /: = / , where is a characteristic velocity (i.e. Darcy
velocity U or front velocity = /). is a characteristic length, i.e. mean grain
diameter for unconsolidated porous media or of the order of for a consolidated media.
Pfannkuch presented a graph with dispersion coefficient normalized by molecular
diffusion as function of Pclet number (Figure 2).

Figure 2 Dispersion coefficient as function of Pclet number (after Pfannkuch [1]).

At very low flow rate (P < 1) The dimensionless dispersion coefficient is constant and
equal to the apparent molecular diffusion coefficient, a regime of pure diffusion
controlled by the apparent diffusion coefficient described above. For P > 1000: the
dispersion coefficient is roughly proportional to the Pclet number. In this regime the
dispersion coefficient is independent of molecular diffusion and proportional to the front
velocity:
=
(3)
is called the dispersivity and is around 1.8 times grain diameter for granular material.
is the front velocity = /, with the Darcy velocity. In laboratory, the determination
of the dispersion coefficient is performed in this regime.

DISPERSION THROUGH A PLUG

In laboratory, the dispersion coefficient is derived from tracer injection through a
cylindrical plug with uniform injection and production on the inlet and outlet faces. The
concentration is measured in the effluent and sometimes locally using in-situ

SCA2015-014

5/12

measurements (X-rays, resistivity, etc.). A constant concentration C0 starting at time t=0

is injected continuously. Several models have been proposed for the boundary conditions:
Inlet boundary condition
In laboratory, even if the tracer is injected at constant concentration, it is the flux that is
imposed at the entrance and not the concentration. This is well illustrated in a micromodel
(Figure 1): A tracer is injected at a constant concentration on the left-hand side of the
picture. Considering this picture as a REV, the concentration is constant at the pore scale,
in the first rows of pores, but not at the REV scale. Consequently, the boundary condition
at the REV scale should be a constant flux density = 0 . However, experimental
data are often interpreted with a constant concentration boundary condition. We will
compare these two assumptions when presenting the experimental results, but the
difference is insignificant.
Outlet boundary conditions
In pure molecular diffusion (without flow), the outlet boundary condition should be the
concentration imposed at the outlet face. In hydrodynamic regime, the tracer follows the
streamlines from inlet to outlet with no backflow. Velocity and dispersion inside the plug
are not influenced by the value of concentration at the outlet or the length of the plug. The
equation is the same as for an infinite medium, with the condition of zero flux at infinite
distance (non infinite mass).
Dispersion through a homogeneous plug
In a homogeneous plug, permeability is uniform and flow lines are parallel to the axe of
the plug. The velocity is uniform and a 1-D mass balance equation can be written
(equation 1). Using the microdispersion relationship between flux and concentration leads
to the standard 1-D dispersion equation.
C/t+U/ C/x=D ( 2 C)/x 2
(4)
Analytical solutions are obtained using Laplace's transforms. With the classical
assumption of constant concentration at entrance, the solution is the following (Fried &
Combarnous [2]):
C
xVt
C
Vx
x+Vt
C = 20 erfc (2(Dt)1/2 ) + 20 exp ( D ) erfc (2(Dt)1/2 )
(5)
With the condition of constant flux at the entrance, the solution is:
f
xVt
f
Vx
x+Vt
f = 20 erfc (2(Dt)1/2 ) + 20 exp ( D ) erfc (2(Dt)1/2 )

(6)

Then the concentration is calculated using the mass balance equation with the boundary
condition:
1

0
2

Vx

V2 t 2

(erfc(z1 ) (1 + ( + )) exp ( D ) erfc(z2 ) + ( D ) exp(z1 2 )),

(7)

SCA2015-014

xVt

6/12

x+Vt

with 1 = 2(Dt)1/2 and 2 = 2(Dt)1/2

At the outlet face, the flux density is derived from the measurement of the effluent
concentration . The calculation is similar to the inlet flux: (, ) = /. The
concentration of the effluent is not equal to the concentration at the outlet inside the plug
(difference between flux and concentration), such as immiscible flow where fractional
flow (flux) is not equal to saturation.
Dispersion through a heterogeneous plug
In heterogeneous media, the permeability is non-uniform and spreading is mainly
controlled by the variation of velocity due to permeability heterogeneity. This type of
dispersion is called macrodispersion because mechanisms are at a higher scale than the
pore scale. Except in artificial media such as filters or chromatographic columns,
spreading due to microdispersion is generally negligible compared to macrodispersion.
The 2 or 3 -D approach
Figure 3(a) shows two examples of heterogeneous permeability fields with the calculation
of flow lines (b) solving Darcy's equation for an uncompressible fluid and the simulation
of injection of a tracer in black (c) [3]. The tracer is displaced along the flow lines and is
transported according to the local velocities. In these simulations, microdispersion is
assumed to be negligible and the tracer is a piston-type displacement along the flow line
(either black and white, no grey).

Figure 3- Two examples of heterogeneous permeability fields with

the corresponding flow lines and tracer injection assuming pistontype displacement along the flow lines.

Figure 4 Principle of stochastic

calculation for spreading in a
heterogeneous plug

Homogenization for 1-D flow

Macrodispersion can be described by hydrodynamic dispersion model (Eq. 2) that
accounts for 2 or 3 -D local heterogeneities. However, in laboratory studies, plugs are
rarely described using 2 or 3 -D permeability fields. Instead 1-D models with variables
averaged over a cross-section are used, and the 1-D permeability becomes uniform, this is
called homogenization. (Figure 3 d). With the local plug-flow rule = in a 1-D

SCA2015-014

7/12

model, there will be no spreading. The information on heterogeneity is lost, and the local
dispersion equation must be changed. We study below three cases of homogenization:
random, stratified, and intermediate medium.
Random permeability field in the flow direction
Gelhar and Axness [4] assumed a diffusive process with proportionality between the flux
and the concentration gradient (as in Fick's law) with a lognormal permeability
distribution. They derived an expression for the dispersion coefficient = , where the
dispersivity is related to the variance of the logarithm of permeability and the
2
correlation length :
. However, mechanism at the origin of dispersion is not
the concentration gradient but the spatial variations of the velocity field.
A model only based on hydrodynamic spreading can leave to the same result The
principle is represented in Figure 4 [5]: the complex stream tubes geometry (a) is modeled
by a bundle of parallel tubes (b) with variable cross-section area of elementary volumes
(c). The tracer front is irregular due to randomness in the elements volumes. The
spreading of the front is calculated as a function volumes linked to the permeability
distribution. The result is similar to Gelhar if randomness is assumed (correlation length
small compared to travelled distance) and transport equation is similar to the
microdispersion case equation (4), but with a dispersion coefficient linked to the size of
heterogeneity (correlation length) instead of pore size. The spreading is diffusive and
the width of the front grows with the square root of travelled distance. This model will be
noted "D" for the simulations, with the two possibilities for the inlet boundary condition:
constant flux or constant concentration. The corresponding flux and concentrations are
given by Eq. 5 to 7.
Homogenization: stratified medium
1

F*

C*
0

Figure 5 - Perfectly layered medium: principle of the calculation of

tracer transport

Figure 6 - relationship between

flux and concentration for a
stratified medium

Stratified media have been extensively used as model of heterogeneous media either for
immiscible or miscible flows (for instance Fourar [6], and with a simplified model of 2
layers by Dauba et al. for carbonate studies [7]).
The heterogeneous porous medium is represented by a perfectly layered medium with a
permeability distribution function (). Microscopic dispersion is assumed to be
negligible and the tracer follows the streamlines. To simplify the calculation, the layers

SCA2015-014

8/12

are re-arranged with increasing permeability (Figure 5). At a given time, the front of
tracer is calculated using the flow rate in each layer and the concentration, by averaging
over the cross-section:

(, ) = 0 (),
(8)

where = is the lower limit of the invaded layers and is the effective
permeability. The flux at the same position is calculated by summing the flow rates in all
invaded layers:

1
(, ) = 0 = 0 () .
(9)

In the general case, there is no analytical relationship between flux and concentration, but
the curve can be determined numerically by elimination of the parameter . The shape is
similar to the fractional flux curve for immiscible flow, but without the front discontinuity
(Figure 6). The transport equation is convective with no second order derivative and the
width of the front is proportional to travelled distance.
Fourar [6] used a lognormal permeability distribution and defined a heterogeneity factor
as the ratio of the standard deviation to the effective permeability. Here we will keep the
general form of the permeability distribution.
Homogenization: general case
For real rock samples, the spreading is a combination of convective flow in the
channels of various permeabilities (preferential paths) and diffusive spreading inside
these channels due to random heterogeneities.
One approach is to use the stochastic tube model with long-range correlations to represent
the transition between diffusive and layered. Results are close to the Continuous Time
Random Walk model introduced by Berkowitz [8]. Models with non-integer derivative
order (between 1 for convective and 2 for diffusive) have also been proposed [9].
In this study, we tested a simpler empirical model called H-D [10], built as the sum of a
convective term and a diffusive term:

= ( ) , with = +(1)/ .
(10)
This approach is justified by assuming that convective effects take place at a larger scale
than dispersive effects. The calculation is similar to the microdispersion case, but instead
of the front velocity in the flux equation, an apparent front velocity is used since the
injected fluid is moving faster.

EXPERIMENTAL RESULTS
Tracer are injected at constant concentration, using 35 and 20 g/l NaCl solutions. Several
plugs have been leading to similar results. Experiment 8 described here is performed on a
5 cm long Claschasch sandstone plug, 1 inch in diameter, with porosity 0.14 and
permeability 300 mD. In-situ concentration profiles are measured along the plug with 10
electrodes. The effluent concentration is measured with a conductimeter placed 5 cm from

SCA2015-014

9/12

the outlet face. Injection is performed through a spiral groove, flushed before the
beginning of the experiment. Injection rate is 10 cc/h with a Pharmacia pump.
Temperature is controlled at 0.2 C accuracy (ion conductivity is strongly dependent on
temperature).
Effluents
The measured effluent flux normalized between 0 and 1 is plotted in Figure 7 as function
of dimensionless time (unity corresponds to the time needed to displace the pore volume).

Figure 7 - Effluent flux profiles: experiment and

models.

Figure 8 Permeability distribution G(K)

calculated from effluent flux and in-situ
concentrations.

For the stratified model, the permeability distribution from effluent is derived by
plotting the experimental flux (figure 7) as a function of / and calculating () after
fitting by spline functions, following Eq. 9. The G(K) from concentration is calculated
from the measured concentration using Eq. 8. The results are plotted in Figure 8 in
dimensionless form. Curves noted D, HD, G(K) are best fits with the corresponding
models (models defined below).
Concentration profiles
Only 5 profiles over the 10 measured are represented in Figure 9 for better readability of
the figures. Symbols represent experiments and the solid lines are the results of the 4
models using CYDAR:
- D model with the two boundary conditions at entrance: constant concentration
(conc. BC, Eq. 5) or constant flux (flux BC, Eq. 7)
- HD model with convective and dispersive terms (Eq. 10), manually adjusted for
the best fit.
- Stratified model with G(K) calculated using Eq. 10. In this equation, the variable
is proportional to /. Therefore, all concentration profiles are plotted as
function of the scaling variable / and data are fitted using splines functions
(Figure 10). The permeability distribution () derived from concentration
profiles is plotted in Figure 8 for comparison to the value derived from the effluent
flux.

SCA2015-014

10/12

In Figure 10, we have also plotted the concentration profiles for the model D with the
scaling variable ( )/. If the experiment follows the model D, all points should
collapse on the same curve.

Figure 9 In situ concentration profiles measured along the plug at different times. Symbols represent the
experiment and the solid lines are the results of the various models at the same times. For the flux and
concentration D model, D= 0.05 (dimensionless). For the HD model, D=0.025 and H=1.1.

Figure 10 a) Scaling laws for the measured in-situ concentration profiles for the D model (scaling in
/) - b) the stratified G(K) model (scaling in x/t).

DISCUSSION
Effluent profiles
Models D and HD cannot reproduce the dissymmetry and the long tail of the effluent flux
production accurately (Figure 7). 2-D numerical simulations performed on heterogeneous

SCA2015-014

11/12

permeability field with intermediate correlation lengths permeability field show that the
injected fluid starts to flow in permeable zones of the medium (channeling), leading to a
fast displacement. At the end of the displacement, the injected fluid flows in much lower
permeability zones and the production is slower, explaining the dissymetry of the curve.
Recently Zik et al. [11] used a model involving exchange by molecular diffusion between
flowing and stagnant zones (model generally called Coats and Smith). This model allows
the representation of long tails by diffusion, but the representation of the concentration
profiles have not been experimentally verified. From the estimate of the diffusion time
scale, this effect might be valid in chromatographic columns, but comes into play at a
different scale for displacement in heterogeneous plugs. The agreement between
experiments and results of G(K) model is good since this distribution is calculated from
the effluent. The small differences are introduced by the fit with splines functions.
Concentration profiles
1) From Figure 9, we see that the best fit is obtained with the stratified model, using the
G(K) permeability distribution. However, the first profiles are not well represented; the
model gives the maximum concentration in contradiction with the experiment. This
discrepancy is also visible in Figure 10, where the brown squares represent the first
profile. This discrepancy could be an experimental artifact during flushing of the inlet
end-piece spiral, with displacing fluid entering partially into the plug.
2) D model with the flux boundary condition is the only model that represents the increase
of concentration at the entrance. However, after a small distance, both boundary
conditions give similar results.
3) Scaling laws for the D and G(K) models (Figure 10) lead to similar results. Therefore
this test cannot efficiently discriminate between models.
4) The G(K) distributions determined from the flux and the one determined from the
concentration (Figure 8) are different, although G(K) should theoretically be unique. The
observed difference may arise from a delay in effluent production in the outlet spiral. The
volume of the outlet spiral is not negligible (0.3 cc compared to 5 cc of pore volume).
This dead-volume introduces a delay that would lead to a lower observed permeability.

CONCLUSION
When used as local input in 3-D simulations, the Darcy-scale approach can accurately
describe the plug-scale spreading related to permeability heterogeneity. In a 3-D
heterogeneous medium, even a displacement started as piston-type at the Darcy-scale
leads to a dispersion behavior due to the spatial differences in velocities.
We have presented several up-scaled (or homogenized) models:microdispersion,
macrodispersion, convective channeling).
Our results confirm previously published results that channeling convective
models are more suitable than the standard dispersion models.
The best result is obtained by calculating a permeability distribution function
() in a stratified model. This model captures both the long tail in the effluent
production and the in-situ concentration profiles.

SCA2015-014

12/12

REFERENCES
[1]

[2]
[3]

[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
[13]

[14]
[15]
[16]

[17]
[18]
[19]

[20]

[21]

H.-O. Pfannkuch, Contribution l'tude des dplacements de fluides miscibles dans un milieu
poreux, Revue de l'Institut Franais du Ptrole et annales des combustibles liquides, vol. 28, pp. 154, 1963.
J. J. a. C. M. A. Fried, Dispersion in Porous Media, Advances in Hydroscience, vol. 7, pp. 169282, 1971.
R. Lenormand and M. Thiele, Homogeneization of Flow equations with viscous Instabilities:
Numerical Validation of the MHD model, European Conference on the Mathematics of Oil
recovery (ECMOR V), 1996.
L. W. Gelhar and C. L. Axness, Three-dimensional stochastic analysis of macrodispersion in
aquifers, Wat. Resour. Res., vol. 19, pp. 161-180, 1983.
R. Lenormand, A stream tube model for miscible flow. Part 1: Macrodispersion in homogeneous
porous media, IEA Collaborative Project on Enhanced oil Recovery, Salzburg, Austria, 1993.
M. Fourar, Characterization of heterogeneities at the core-scale using the equivalent stratified
porous medium approach, SCA2006-49, Trondheim, Norway, 2006.
C. Dauba, G. Hamon, M. Quintard and F. and Cherblanc, Identification of parallel heterogeneities
with miscible displacements SCA paper 9933,, Golden, USA, 1999.
B. a. S. H. Berkowitz, The role of probabilistic approaches to transport theory in heterogeneous
media, Transport in Porous Media, vol. 42, pp. 241-263, 2001.
R. Lenormand, Use of Fractional Derivatives for Fluid Flow in Heterogeneous Media, 3rd
European Conference on the Mathematics of Oil Recovery, 1992.
R. Lenormand, Determining Flow Equations from Stochastic Properties of a Permeability Field,
SPE Journal, vol. 1, 1996.
B. Vik, K. Djurhuus, D. Doublet, K. Spildo and A. Skauge, Characterization of vuggy carbonates
by miscible process, SCA2008-28, Abu Dhabi UAE, 2008.
F. A. L. Dullien, Porous Media: Fluid Transport and Pore Structure, Academic Press ed., San
Diego, California, 1992, p. 574.
K. Sorbie, H. Zhang and N. Tsibuklis, Linear viscous fingering: new experimental results, direct
simulation and the evaluation of averaged models, Chemical engineering science, vol. 50, n %14,
pp. 601-616, 1995.
A. Koval, A Method for Predicting the Performance of Unstable Miscible Displacement in
Heterogeneous Media, SPE Journal, pp. 145-154, June 1963.
M. Sahimi, M. R. Rasaei and M. Haghighi, Gas injection and fingering in porous media, Gas
Transport in Porous Media, C. Ho and S. Webb, Eds., Springer, 2006, pp. 133-168.
M. Fourar, G. Konan, C. Fichen, E. Rosenberg, P. Egermann and R. Lenormand, Tracer tests for
various carbonate cores using X-Ray CT, SCA 2005-56, Society of Core Analysts, Toronto,
Canada, 2005.
A. Fick, Ueber Diffusion, Annalen der Physik, vol. 170, pp. 59-86, 1855.
K. H. Coats and B. D. Smith, Dead-end pore volume and dispersion in porous media, Society of
Petroleum Journal, pp. 73-84, 1964.
G. Taylor, Dispersion of soluble matter in solvent flowing slowly through a tube, Proceedings of
the Royal Society of London, Series A, Mathematical and Physical Sciences, vol. 219, pp. 186-203,
August 1953.
G. Taylor, The Dispersion of Matter in Turbulent Flow through a Pipe, Proceedings of the Royal
Society of London A: Mathematical, Physical and Engineering Sciences, The Royal Society, vol.
223, pp. 446-468, 1954.
R. Aris, On the Dispersion of a Solute in a Fluid Flowing through a Tube, Proceedings of the
Royal Society of London A: Mathematical, Physical and Engineering Sciences, The Royal Society,
vol. 235, pp. 67-77, 1956.

SCA2015-015

1/12

EFFECT OF THE PRESENCE OF INTERSTITIAL BRINE

ON GAS-OIL CAPILLARY PRESSURE
F. Pairoys, D. Simons, K. Bohn, M. Alexander, V. Odu, R. DeLeon, J. Ramos
Schlumberger Reservoir Laboratories, Houston, Texas, 77041, United States
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in Newfoundland and Labrador, Canada, 16-20 August, 2015

ABSTRACT
Gas-oil centrifuge capillary pressure experiments were performed on two outcrop
samples in the presence and absence of a third immobile phase in order to highlight
effects on final recovery.
Two core plugs with different porosity and permeability properties were respectively
drilled from an Indiana limestone block and Berea sandstone block. They were solvent
cleaned and measured for routine core analysis before starting the SCAL program. Two
centrifuge capillary pressure experiments were run on the samples at different saturation
conditions. Both samples were first saturated with brine and displaced by oil up to
irreducible brine saturation using centrifuge (primary drainage). Gas-oil gravity drainage
was then performed using multistep centrifuge method for capillary pressure
measurement. The samples were then cleaned, measured again for basic properties, and
saturated with oil. The second experiment was a new gas-oil multistep centrifuge
drainage using same capillary pressure steps than in the first test, but directly performed
on the samples initially saturated with oil.
The objective was to highlight the effect of the interstitial brine on the capillary pressure
curves and on the oil recovery. It is finally shown that the use of only two phase capillary
pressure may result in an invalid capillary pressure relationship between the two mobile
fluids in the presence of a third. The Hassler-Brunner method for the determination of the
local capillary pressure curves in presence and absence of a third phase is discussed. The
study also confirms the necessity of using the Forbes method to avoid inaccurate data
results.

INTRODUCTION
Almost all reservoirs contain three fluid phases within their pore network, respectively
water, oil and gas. Water and oil are immiscible together, whereas gas can be miscible
with oil or/and can be dissolved in the aqueous phase according to the reservoir
conditions and fluid properties. Assuming that miscibility and dissolution are not
involved, the three phases may be mobile or static according to their saturations; in a
three phase system, there is at least one mobile phase. The capillary pressure Pc is an
important parameter for the reservoir engineers who are using this petrophysical input in
their numerical simulations, in addition to the relative permeability Kr, to anticipate the

SCA2015-015

2/12

reservoir behavior. Capillary pressure Pc is a two-phase phenomenon resulting from a

discontinuity in pressure across the interface between two immiscible fluids. Capillary
pressure is affected by a multitude of parameters; fluid properties, saturation, saturation
history [1], pore size, pore shape, pore distribution, tortuosity, wettability [4]. It is evident
that oil recovery in a reservoir is affected by the capillary pressure Pc since trapping
mechanisms depends on the listed parameters.
Gravity drainage is the drainage process of gas displacing oil in a reservoir or sample. In
certain cases, it is an efficient recovery mechanism which can result in very low residual
oil saturation compared to waterflooding displacements [2]. Three phase capillary
pressure test as described by Kantzas et al. [3] with three mobile phases is not considered
in this study; two phase -gas/oil- capillary pressure experiments in presence or absence of
a third immobile phase -brine- are investigated. Hassler et al. [4] and other others such as
Kyte et al. [5] or Hagoort [6] concluded that the gravity drainage could be considered as
two phase flow displacement at low water irreducible saturation. There is a lack in
experimental investigation of the effect of connate water on the final oil recovery during
gravity drainage. Dumore [7] showed that the presence of connate water during gravity
drainage test in permeable medium leads to very low residual saturations due to film flow
[8]. Hagoort [6] observed a decrease in residual oil saturation with an increase of water
saturation due to more streamlined flow channels. One study from Carlson [9] showed
that the presence of irreducible water had no effect on the residual oil saturation. Earlier,
a same observation was made by Tiffin and Yellig [10].
The purpose of this paper is to compare gas-oil capillary pressure in presence or absence
of connate water during a multistep centrifuge experiments. The centrifuge tests were
performed at ambient conditions and immiscible conditions. Because capillary pressure
is, inter alia, affected by pore size, two different rock types were tested to compare the
effect of irreducible water on the final oil recovery during drainage cycle. The HasslerBrunner method [4] to calculate the local capillary pressure at the inlet of the core
samples is also discussed in this paper. The core analysis software CYDAR was used for
data processing.

BACKGROUND
In centrifuge gravity drainage experiments, the fluid flow is controlled by the density
difference between fluids. The local velocity u for each fluid is given by the generalized
Darcys law:
KK r P

(1)
u
2 r

With K the absolute permeability in m2, Kr the relative permeability, the fluid viscosity
in Pa.s, P the fluid pressure in Pa, the fluid density in Kg/m3, the rotational speed in
rad/s and r the distance between the position on the rock sample and the centrifuge axis in
m.

SCA2015-015

3/12

The capillary pressure Pce at the inlet of a sample during a centrifuge experiment can be
expressed as follows:
1
Pce 2 r22 r12
(2)
2
With Pce the inlet capillary pressure in Pa, the density contrast between the two fluids
in Kg/m3, r2 the distance between the outlet of the sample and the centrifuge axis in m,
and r1 the distance between the inlet of the sample and the centrifuge axis in m.
A sketch of rock sample loaded in centrifuge core holder is shown below:

Figure 1: Sketch of core plug loaded in centrifuge core holder

During a multi-speed experiment, when the production of the displaced phase ceases at
the end of each rotational speed, the average saturation <S> can be calculated from the
effluent production. <S> is linked to the local inlet capillary pressure Pce at equilibrium
by the following Hassler and Brunner equation:
r
1 2
S Pce
S ( Pce )dr
(3)
r2 r1 r1
The Hassler-Brunner method [4] is commonly used in the industry to approximate the
local saturation at the capillary pressure Pce using the following equation:
S Pce

d
Pce S Pce
dPce

(4)

The differentiation of this equation was performed by differentiating a modified

hyperbolic function which best fits the product Pce<S(Pce)>. The Forbes method [11] after
smoothing the data [13] was used when the Hassler-Brunner method did not give
satisfying local capillary pressure curve.
The Hassler-Brunner method is valid for short cores spinning far enough from the
centrifuge axis to assume that the variation of the centrifugal field is negligible; in this
study, the ratio r1/r2 in Figure 1 was found to be equal to 0.6. The maximum Bond
number Nb obtained at the highest speed was less than 10-4.

SCA2015-015

4/12

ROCKS, FLUIDS AND EXPERIMENTAL SETUP

Two rock types were tested respectively Berea sandstone and Indiana limestone. Berea
sandstone is unimodal in pore and pore throat sizes whereas Indiana limestone is
bimodal. It is confirmed by mercury injection capillary pressure (MICP) and NMR T2
distributions obtained respectively from two rock chips and two 100% brine saturated
core samples cut from each block (Figure 2):

Figure 2: Capillary pressure (left), pore throat size distribution (middle) and NMR T2 distribution (right)

Figure 2 shows that the two rock types have different signatures as expected and also
shows that measurements on twin plugs obtained from the same block are almost
identical. The two samples tested in this study came from the same blocks as the samples
tested for MICP and T2 NMR from Figure 2. There is a clear correlation between the
NMR pore size distribution and the MICP pore throat distribution for each rock type.
Both MICP and NMR highlight respectively the unimodal pore structure of the Berea
sandstone and bimodal pore structure of the Indiana limestone. Additional rock
information obtained from the same Berea sandstone and Indiana limestone blocks can be

SCA2015-015

5/12

found in Pairoys et al. [14].

Several core plugs were obtained from the Berea sandstone Indiana limestone blocks, but
only data results of sample B2 (Berea sandstone) and sample IL20 (Indiana limestone)
are reported. After having Soxhlet cleaned the samples, routine core analysis was
performed on the samples. Data results are listed in Table 1:
Table 1: Dry core plug properties
Sample Id

Type

g (g/cc)

(%)

PV (cc)

Kg (mD)

Kkl (mD)

B2

Berea sandstone

2.65

25.23

10.473

735.7

727.8

IL20

Indiana limestone

2.69

15.43

6.772

31.8

30.4

Fluid properties (brine, oil and air) at ambient temperature and atmospheric pressure are
reported in Table 2:
Table 2: Fluid properties
Fluid

Salinity
(Kppm NaCl)

Density
(g/cc)

Viscosity
(cp)

Brine

200

1.130

1.345

Isopar-L

0.762

2.049

Air

0.001

0.018

The capillary pressure experiments were run with an Ultra Rock Centrifuge from
Beckman allowing monitoring fluid production using a camera system. The two samples
were first saturated with Isopar-L before being loaded in drainage centrifuge cells and
tested for multistep capillary pressure Pc. Ten Pc steps were applied, from 0 to 84 psi.
At the end of the drainage cycle, the samples were Soxhlet cleaned using Toluene and
Methanol, dried, and re-measured for gas permeability and porosity. Because the rock
properties did not change after the first Pc test and cleaning, they were used again for the
second Pc test in presence of connate water. The two cleaned samples were then saturated
with brine, loaded in drainage cells for brine-oil primary drainage up to irreducible water
saturation Swi using one single maximum step (8000 rpm around 100 psi for all
samples). Then, they were tested in the same drainage core holders for gas-oil
displacement. They were spun at the exact same capillary pressure steps than those
applied during the first test with samples initially saturated with oil.

EXPERIMENTAL RESULTS
Centrifuge gas-oil capillary pressure experiments were run on the two samples, IL20 and
B2, initially 100% saturated with oil and later saturated with oil at irreducible water
saturation Swi. The main objective is to see if the capillary pressures curves and
productions are affected by the presence of irreducible water saturation.

SCA2015-015

6/12

Gas-Oil Drainage at Initial Oil Saturation So=100%:

The samples B2 and IL20 were directly tested for gas-oil multistep capillary pressure.
Ten capillary pressure steps were applied; they were chosen according to a preliminary
designing test on sample IL20 100% oil saturated only (using CYDAR software). The
average saturations were fitted with a bi-exponential curve to obtain the average
saturation at equilibrium. Figure 3 represents the oil production curves during the
centrifuge Pc experiment.

IL20

B2

Figure 3: Drainage oil productions (raw data) for samples IL20 (left) and B2 (right)

Sample IL20: A total of 3.82 cc of oil was produced, leading to residual oil saturation Sor
close to 43.6%. Sor was found to be high due to the presence of the oil-filled nanopore
population highlighted in Figure 2; at the applied pressure steps, it is not possible to
displace the oil in these tiny pores. The production curve also shows that equilibrium was
not completely reached at all Pc steps. The oil production was fitted with multistep biexponential function to obtain the saturation at equilibrium.
Sample B2: A total of 8.40 cc of oil was produced from sample B2. The residual oil
saturation was found to be equal to 19.8%, significantly lower than for the Indiana
limestone IL20; these two rocks have different pore size and throat signatures, partially
responsible for this difference.
In Figure 4, the capillary pressures Pc versus average oil saturations So at equilibrium are
plotted for both samples (in blue). Modified hyperbolic functions were used to fit the
capillary pressure curves before calculating the local saturations (in red). Local capillary
pressure curve is obtained from Hassler and Brunner method in green). As QC technique,
the average saturations <So> were recalculated (in purple) and compared to the analytical
Pc curves (in red) in order to assess the Hassler-Brunner method.

SCA2015-015

7/12

IL20

B2

Figure 4: Measured, analytical fit, local Hassler-Brunner and recalculated capillary pressure curves

Figure 4 shows that the Hassler and Brunner method provides more acceptable result for
sample B2 than for sample IL20; the recalculated average saturation matches well the
analytical fitted curve. The exact same experimental protocol was performed on the same
cores but at irreducible saturation Swi as initial condition.
Gas-Oil Drainage at Initial Irreducible Water Saturation Swi:
For the sake of more clarity, an asterisk * was added to the plug number containing
irreducible water at initial conditions. At the end of the previous test, samples were
Soxhlet cleaned, dried and measured for routine core analysis; porosity and permeability
were found were close to the ones reported in Table 1. After brine saturation of sample
B2* and IL20*, a primary drainage, oil displacing brine, was performed. They were spun
in centrifuge at a single speed of 8,000 rpm, so approximatively 100 psi of oil-brine
capillary pressure Pc up to irreducible water saturation Swi. Cores were also flipped to get
a more uniform fluid distribution along the samples [15]. The irreducible water saturation
Swi for sample IL20* was found to be equal to 36.1% whereas Swi for sample B2* was
found to be equal to 16.6%. The presence of nanopores (Figure 2) explains the high Swi
value for sample IL20*. The samples were then loaded in the drainage centrifuge cells for
air-oil multistep capillary pressure experiment. Tests were run with the same capillary
pressure steps than those applied on the samples initially saturated with oil. No brine
production was observed during the gas-oil centrifuge Pc experiments.

IL20*
The same increasing capillary pressure steps were applied during the gas-oil drainage
displacement of sample IL20*.

IL20*
Figure 5: Drainage oil productions (raw data) for sample IL20* (left) and B2* (right)

B2*

SCA2015-015

8/12

Figure 5 represents the oil production versus time during the multistep centrifuge
experiment with samples initially at Swi. The production curves show that equilibrium
was not completely reached at all Pc steps, leading to an overestimation of the residual oil
saturation Sor. The same fitting process than the one described on the experiment with
samples initially saturated with oil was used to obtain the saturation at equilibrium.
Sample IL20*: A total of 3.92 cc of oil was produced, leading to residual oil saturation
Sor close to 6.1%. The produced volume was found to be almost the same than the one
obtained on the same sample initially saturated with oil. It is explained by the bimodal
porosity system of this limestone rock. Only oil occupying the micro-pore population
highlighted in Figure 2 can be mobilized and produced. The main difference is the final
recovery of the original oil in place (OOIP), which is obviously much higher when the
water occupied the small pores (nano-pores population for this Indiana limestone rock).
Sample B2*: A total of 7.68 cc of oil was produced from sample B2, leading to 10.1% of
residual oil saturation. The oil production was found to be less than the one obtained for
the same sample at 100% oil initial condition. The capillary pressure Pc versus average
oil saturation So at equilibrium is then plotted for both samples (blue curve in Figure 4):

IL20*

B2*

Figure 6: Measured, analytical fit, local H-B and recalculated capillary pressure curves

As observed in Figure 4 for samples initially saturated with oil, the back calculation of
average oil saturation <So> for sample IL20* does not perfectly fit the analytical curve
used for fitting the Pc data points. The non-monotonic local Pc of sample B2* obtained
with the Hassler-Brunner method may be explained by the choice of the modified
hyperbolic fitting function and the imperfect equilibrium of the Pc data points. Despite of
this observation, an acceptable match between analytical data fit Pc (red) and recalculated
Pc at average saturation (purple) is obtained for sample B2*.
For the local Pc calculation, whatever the initial conditions (with or without presence of
irreducible water saturation), the Hassler and Brunner method seems more appropriate for
the sandstone sample B2/B2*; the match between the analytical function used to fit the Pc
data points and the recalculated <So> function for the Indiana limestone is less good than
the match of the same curves for the Berea sandstone. The reason could be the bimodal
character of the limestone rock. Because of the non-monotonic Hassler-Brunner local Pc
curve of sample B2* shown in Figure 6, the Forbes method with initial splines fit was

SCA2015-015

9/12

tested (Figure 7):

Figure 7: Measured, analytical fit, Forbes-splines and recalculated <.So> Pc curves for sample B2*

In addition to the observed monotonic local Pc curve, the match between the analytical
and recalculated Pc was improved using Forbes method with splines. All capillary
pressure curves Pc were then normalized for saturation (reduced oil saturation So* on X
axis), plotted on a same graph, and compared (Figure 8).

Figure 8: Comparison of local Pc curves for sample IL 20 and B2, So=100% versus initial Swi

Figure 8 shows that, whatever presence or absence of irreducible water, the local
capillary pressure curves of IL20/IL20* are overlapping. Using two phase capillary
pressure centrifuge test without connate water may be representative of the two phase
capillary pressure with connate water for this sample. However, for sample B2/B2*, the
capillary pressure curves are different. In that case, the use of two phase capillary
pressure without connate water may not be representative of the two phase capillary
pressure with connate water. One of the reasons may be the change of gas-oil interface
curvature in presence of irreducible water.
Additional tests need to be performed to confirm that the Hassler-Brunner outflow
condition was not violated due to high Bond number Nb, invalid constant centrifugal field
condition (geometry), or presence of irreducible water at the outlet of the samples. NMR
or X-ray measurements before and after the gas-oil drainage displacement could help in
ensuring that the outlet saturation does not change. The failure of Hassler-Brunner
method for obtaining reliable local Pc curve for B2* may also be due to one or
combination of the enounced factors, showing the complexity of such experiments.

SCA2015-015

10/12

Recovery Analysis:
Table 3 compares the productions at different initial conditions, with and without
irreducible water saturation. For the sake of more clarity, an asterisk * was added to the
plug number containing irreducible water at initial conditions.
Table 3: Summary of production according to the initial conditions
Sample
Id
IL20

PV
(cc)
6.77

Swi
(frac.)
0.000

So
(frac.)
1.000

Vo prod.
(cc)
3.82

ORF
(%)
56.4

Sor
(frac.)
0.436

Sg
(frac.)
0.564

Kg(Sor)
(mD)
32.9

IL20*

6.77

0.361

0.639

3.92

90.5

0.061

0.579

31.8

B2

10.47

0.000

1.000

8.40

80.2

0.198

0.802

747.8

B2*

10.47

0.166

0.834

7.68

87.9

0.101

0.734

727.0

Sample IL20/IL20*:
The presence of irreducible water Swi in IL20* results in very low residual oil saturation
Sor close to 6.1%. It can be explained by the rock water wettability. First, the small pores
are initially saturated with brine; no oil could displace the water in the nanopore zone.
Then, water spreads on the pore wall, making easy the oil displacement by gas in the
large pores. It also explains the large oil recovery factor ORF of IL20* with almost
90.5% of the original oil in place produced by gas. The gravity drainage in IL20 (no Swi)
is much less efficient. The main reason is that oil initially trapped in nanopores (see
Figure 2) cannot be mobilized by gas at the applied capillary pressure steps, in addition to
the oil wettability of the rock in presence of oil and gas only (no lubrication effect like
IL20*). Around 56.4% of oil in place was recovered; this value can be considered low
compared to the generally high recovery (>60% from case histories) observed during
gravity drainage at secondary conditions (no waterflooding before gas injection in oil
reservoirs).
Sample B2/B2*:
The residual oil saturation and recovery factor of sample B2/B2* are less affected by the
presence or absence of irreducible water saturation Swi compared to IL20/IL20*. One of
the reasons is the unimodal signature of the pore space. The oil recovery factor is higher
in presence of connate water. It is also observed that the values at Sor of B2 and IL20 are
in the same range but slightly higher than the values of Swi of B2* and IL20*; same range
because the capillary trapping mechanism is similar for both tests (B2 being oil-wet in
presence of gas only and B2* being water-wet in presence of gas and oil) and slightly
higher gas/oil Sor than brine/oil Swi because of the difference in maximum applied Pc (100
psi for oil displacing brine up to Swi and 85 psi for gas displacing oil up to Sor).
Steady-state gas permeability measurements were performed on the samples at the end of
the gas-oil drainage to obtain the relative permeability kg at Sor. Results are reported in
Table 3 as well. It shows that for sample IL20/IL20*, the presence of a third immobile
phase does not change the effective permeability to gas; both values at Sor and at Sor and
Swi are similar. One of the reasons is that, in presence or absence of irreducible water, the

SCA2015-015

11/12

couple IL20/IL20* and the couple B2/B2* have close final gas saturation Sg; whatever
liquids (oil or water + oil) are inside the rock, they do not affect the flow of the strongly
non-wetting phase such as gas. The effective gas permeabilities are also very close to the
absolute gas permeabilities (Table 1). The residual oil and irreducible water do not
hamper gas flow because gas is strongly non-wetting the rock compared to liquids.
Effective permeability values are even slightly higher than the absolute one; this
phenomenon is known as lubrication effect [16].
At the end of the experiment, the samples initially saturated with brine were Dean Stark
extracted to validate the irreducible water saturation Swi obtained by material balance
during the primary drainage. Irreducible water saturation Swi obtained by material balance
during the centrifuge tests and irreducible water saturation obtained by Dean-Stark were
found to be very close, ensuring that no connate water was dried or produced during the
gas injection. Moreover, to ensure that the samples were not damaged during the
centrifuge tests, porosity and absolute gas permeability were measured; no significant
rock property changes were observed (Table 4).
Table 4: Dean Stark results and dry rock properties before and after the tests
Sample Id

Material balance
Swi (frac.)

Dean Stark
Swi (frac.)

Initial
(%)

(%)

Final

Initial
Kg (mD)

Final
Kg(mD)

B2*

0.166

0.159

25.23

25.40

735.7

728.6

IL20*

0.361

0.355

15.43

15.24

31.8

31.0

CONCLUSION
Two-phase, gas-oil, centrifuge capillary pressure measurements were performed on
different rock types in presence or absence of interstitial water. Two different rock types
were tested to assess the effect of pore size distribution on final oil recovery. Results of
the investigation show that a third brine phase saturation can significantly change the
residual oil saturation Sor and the oil recovery factor ORF, the shape of the capillary
pressure curve and saturation distribution in the cores. It is also shown that the Hassler
and Brunner method, used for determining the inlet saturation in centrifuge experiments,
can be improved using Forbes method with splines. It is finally concluded that the
presence of a third phase can have a significant effect in determining the capillary
pressure-saturation relationships between two fluids in a three phase system. Additional
work is required for validating the Hassler-Brunner outlet boundary condition using
NMR or X-ray spectroscopy.
For oil recovery, the rock saturation history is an important parameter: more oil volume is
obtained from samples initially saturated with oil, but higher oil recovery factor ORF is
observed in presence of irreducible water. It is also very important to know the pore
size/pore throat distributions to get a better understanding of the oil production scenario.
The gas floods leads to low residual saturation in swept zones but poor volumetric rock
sweep in the bimodal Indiana limestone sample initially saturated with oil.

SCA2015-015

12/12

REFERENCES
1. Brown, R.J.S., Fatt, I., Capillary Pressure Investigations, Trans. AIME, 192, pp. 6774, 1951
2. Leverett, M.C., Capillary Behavior in Porous Solids, Trans. AIME, 142, pp. 152169, 1941
3. Kantzas, A., Chatzis, I., Dullien, F.A.L., Enhanced Oil Recovery by Inert Gas
Injection, SPE/DOE Paper 17379, 1988
4. Hassler, G.L., Brunner, E., Measurement of Capillary Pressure in Small Core
Samples, Trans. AIME 160:114-123, 1945
5. Kyte, J.R., Stanclift, R.J. Jr., Stephan, S.C., Rapoport, L.A., Mechanism of Water
Flooding in the Presence of Free Gas, Trans. AIME, 207, pp. 215-221, 1956
6. Hagoort, J., Oil Recovery by Gravity Drainage, SPE Journal, pp. 139-150, June
1980
7. Dumore JM, Schools RS, Drainage Capillary Pressure Functions and the Influence
of Connate Water, SPE 4096, SPE Journal, pp. 437-444, October 1974
8. Nenniger, E. Jr., Storrow, J.A., Drainage of Packed beds in Gravitational and
Centrifugal-Force Fields, AIChE, Vol. 4, Issue 3, pp. 305-316, September 1958
9. Carlson, L.O., Performance of Hawkins Field Unit Under Gas Drive Pressure
Maintenance Operations and Development of EOR Project, SPE/DOE paper 17324,
1988 SPE/DOE Symposium on EOR, Tulsa, Oklahoma, US, 17-20 April 1988
10. Tiffin, D.L., Yellig, W.F., Effects of Mobile Water on Multiple-Contact Miscible
Gas Displacement, SPE Journal, pp. 447-455, June 1983
11. Forbes, P., Simple and Accurate Methods for Converting Centrifuge Data into
Drainage and Imbibition Capillary Pressure Curves, Society of Core Analysts,
SCA1991-07, 1991
12. Nordtved, J.E., Kolitvelt, K., Capillary Presure Curves from Centrifuge Data by Use
of Spline Functions, SPERE, pp. 497-501, November 1991
13. Bauget, F., Gautier, S., Lenormand, R., Samouillet, A., Gas-Liquid Relative
Permeability from One-Step and Multi-Step Centrifuge Experiments, Society of
Core Analysts, SCA2012-13, Aberdeen, Scotland, 27-30 August, 2012
14. Pairoys, F., Nadeev, A., Kirkman, K., Poole, G., Bohn, K., Alexander, M., Radwan,
N., Abdallah, W., Akbar, M., Assessment of Sidewall Cores for Routine and Special
Core Analyses, Society of Core Analysts, SCA2015-A003, Newfoundland and
Labrador, Canada, 16-20 August 2015, Scotland, 27-30 August, 2012
15. Pairoys, F., Al-Zoukani, A., Nicot, B., Valori, A., Ali, F., Zhang, T., Ligneul, P.,
Multi-Physics Approach for Aging Assessment of Carbonate Rocks, SPE 149080,
Society of Petroleum Engineer, ATSE, Al-Khobar, Saudi Arabia, 15-18 May 2011
16. Berg, S., Cense, A.W., Hofman, J.P., Smits, R.M.N., Flow in Porous Media with
Slip Boundary Condition, Society of Core Analysts, SCA2007-13, Calgary, Canada,
10-12 September 2007

SCA2015-016

1/12

REMOVAL OF MUD COMPONENTS FROM RESERVOIR

SANDSTONE ROCKS
Ingebret Fjelde, Aruoture Voke Omekeh and Mona Wetrhus Minde
International Research Institute of Stavanger (IRIS)
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Laboratory experiments using reservoir rock/oil are used extensively to estimate reservoir
oil recovery processes. Mud contamination of the reservoir rock can change the flow
conditions by changing the rock surface chemistry and permeability. Representative data
can then only be generated if mud contamination is removed. The objective for study was
to investigate the removal of these mud components during cleaning of sandstone
reservoir core plugs. Effluent samples were analysed, and wettability of cleaned core
plugs was characterized. Samples of core plugs after cleaning and after water flooding
were analysed by scanning electron microscopy (SEM) and the oil residue in cleaned
reservoir rock was analysed by gas chromatography.
When formation water (FW) was injected to uncleaned or solvent cleaned reservoir core
plugs with potential contamination by water-based mud, ions from the mud were
produced for relatively long periods. SEM analyses showed that polymers, clay and barite
from the mud were present in the pores of cleaned high permeability reservoir core plugs.
Even after water flooding, mud components were found in this rock. Polymers and
particles can reduce the permeability of the rock. Invasion of clay from the mud will
increase the surface area and ion-exchange capacity of the rock. Clay invasion and/or
wrong FW composition can affect the established wettability and thereby the estimated
potential for water flooding and EOR methods.
Contamination of sandstone rocks with oil-based mud was shown to alter the wettability.
Emulsifiers from the mud were slowly removed during cleaning by crude oil injection.
The cleaning procedure for reservoir rock with potential oil-based mud contamination,
was shown to affect the measured capillary desaturation curve.
Mud components should be removed during cleaning of the reservoir core plugs,
otherwise the core plugs should not be used. Analyses of effluent samples for these
components during core plug preparation should be included in the standard preparation
procedures. For some mud components, chemical analysis can be challenging. The
cleaning procedure should also be evaluated for the actual reservoir rock by analysing
cleaned core plugs for wettability, oil residue and mud components. Important minerals
should not be removed from the original reservoir rock during cleaning.

SCA2015-016

2/12

INTRODUCTION
Both pressure and temperature are reduced during retrieval of reservoir rock samples [1].
This can change the solubility of minerals and organic and inorganic components in the
residual reservoir fluids. A gas phase will also be formed when the pressure is reduced
below bubble point. Invasion of mud components can alter the rock properties [2].
Wettability can be changed by adsorption of mud components. Invasion of particles,
polymer and resins from the mud can also reduce the permeability of the rock. The ionic
composition of the brine in the mud system is different from the formation water (FW)
composition. The surface properties of the minerals can therefore be changed due to mud
filtrate invasion. In the preparation of reservoir rock in the laboratory it is important to
establish wettability conditions representative of the oil reservoir [2]. Mud contamination
and precipitates from the reservoir fluids should be removed, prior to testing. It is
important that the selected composition of the FW is as similar as possible to the real
reservoir FW. Preparation of the reservoir rock with incorrect compositions of the FW
can give unrepresentative surface properties of the minerals and thereby wettability
conditions which are not representative of the reservoir.
The first aim of the presented study was to evaluate the cleaning procedure of high
permeability reservoir sandstone core plugs with potential contamination by water-based
mud. The second aim was to explore cleaning of sandstone rock (medium permeability)
with oil-based mud contamination. Another reservoir sandstone rock with the potential
contamination with oil-based mud was cleaned by different procedures before
determination of the capillary desaturation curve.

METHODS
Water-Based Mud: Evaluation Cleaning Procedure
The procedure for cleaning of high permeability reservoir sandstone rock with potential
contamination by water-based KCl mud was evaluated using core plugs drilled from the
same seal peel parallel to bedding.
Cleaning
The core plugs were first flooded with synthetic formation water (SFW, Table 1) in triaxial core-holders at 60oC using injection rate of 0.5ml/min. The core plugs were then
cleaned by methanol/toluene cycles at 60oC with an injection rate of 0.5ml/min until the
effluent was colourless, 4 cycles with 5 pore volumes (PV) of each solvent in each
cycles. The total period with cleaning was approximately 6 weeks, including injection
and storage. The cleaning was finished with a 17 PV injection of methanol. The methanol
was then replaced with SFW at room temperature at an injection rate of 0.5 ml/min,
before absolute permeability was determined. Effluent brine samples were analysed for
elements by Inductively Coupled Plasma (ICP) in both SFW injection steps.
Analyses Of Cleaned Reservoir Cores
Spontaneous imbibition and forced imbibition experiments were performed to
characterize the wettability of cleaned core plugs. After drainage to Swi by Isopar H

SCA2015-016

3/12

(synthetic oil) using the confined porous plate method the spontaneous imbibition of
SFW was studied in Amott cells at room temperature. Water flooding with SFW was then
carried out in tri-axial core-holder at room temperature. Samples from inlet, mid and
outlet of one cleaned core plug were extracted with dichloromethane, and the extracts
were analysed by gas chromatography to characterize the oil residue. Distribution of mud
components from inlet to outlet of cleaned reservoir core plug was studied by a ZEISS
Supra 35 VP field emission SEM. An EDAX Energy-dispersive X-ray spectroscopy
(EDS)-detector was used for analyses of the elemental composition.
Oil-Based Mud: Evaluation Cleaning Procedures
Cleaning By Stock Tank Oil Injection
Sandstone core plugs (100-200mD) drained to Swi were aged with STO at 90C for 14
days. Oil-based mud filtrates were prepared by stepwise filtration from 120m down to
5m. The mud filtrate was injected to core plugs through an inlet filter of 0.45m using
an injection rate of 0.5ml/min. The rock was then aged again for 14 days at 90C before
cleaning by STO injection. Emulsifier production was studied by determination of the
interfacial tension (IFT) between the effluent oil phase and SFW using the spinning drop
technique [3]. Wettability after this cleaning was then characterized by spontaneous
imbibition of SFW.
Capillary Desaturation Curve With Different Solvent Cleaning
To evaluate the potential for surfactant flooding in a sandstone reservoir, capillary
desaturation curve (CDC) was determined. Two different cleaning procedures were used
for the reservoir rock with potential contamination by oil-based mud: 1. Injection of
methanol until colourless effluent, and 2. Injection of methanol/toluene (33pore volume
(PV)), acetic acid (33PV) and ethanol (33PV). After cleaning, the composite reservoir
core was saturated with isopropanol. Fluid systems of different IFT were prepared using
solvent systems of 2w%CaCl2/iso-octane/isopropanol [4, 5]. It was assumed that the
solvent system has similar flooding behaviour at reduced IFT as surfactant systems.

RESULTS
Water-Based Mud: Evaluation Cleaning Procedure
Analysis of effluent samples and cleaned reservoir rock were carried out in the evaluation
of the cleaning procedure for high permeable reservoir rock with potential contamination
with water-based mud.
Element Composition Of Effluent During SFW Injections
SFW was first injected before cleaning with solvents to remove water-soluble mud
components. In the beginning the effluent potassium (K) concentration was much higher
than in injected SFW (Figure 1a) The K concentration gradually reduced, but was still
higher than in SFW at the end of this injection step. In the beginning of the second SFW
injection after solvent cleaning, the K concentration was also higher than injected before
it became similar in the levels in SFW. These results show that the core plugs were
contaminated by KCl-mud, and that all the contamination was not removed during the

SCA2015-016

4/12

solvent cleaning. Before solvent cleaning the effluent sulphur (S) concentration was also
much higher than in SFW at the start of injection, and was still higher than in SFW at the
end of this step (Figure 1b). When SFW was injected the second time, the Sconcentration was still higher than in SFW at the start of injection but became similar at
the end of the injection. The development in effluent concentrations of K and S were
rather similar in the first SFW injection (Figure 1), and the sulphur source is therefore
probably the mud.
During the first SFW injection the effluent calcium (Ca) concentration was lower than in
SFW, but at the end of this injection step was similar to the concentration in SFW (Figure
1c). This also confirmed the mud invasion, because the concentration of Ca in the mud
filtrate is lower than in SFW. In the second SFW injection after solvent cleaning, the Ca
concentration was largely stable. In the first SFW injection the effluent magnesium (Mg)
concentration was also lower than injected and was still approaching the SFW level at the
end of this step (Figure 1d). This again confirmed the mud contamination. At the end of
the second SFW injection, the Mg concentration was similar as in SFW. The mud
contamination was also confirmed by low effluent sodium (Na) concentration in the
beginning of the first SFW injection and was still approaching SFW level at the end of
this step (Figure 1e). At the end of the second SFW injection, the effluent Na
concentration was as in SFW.
It is the water soluble mud components that should be easiest to remove during injection
of SFW, especially the ions. The ICP-analysis of effluent samples during the two steps
with SFW injection, clearly show that it is necessary to inject several pore volumes to
remove the mud contaminations. It would have been more difficult to remove these mud
contaminations from more heterogeneous core plugs. The results in the first SFW
injection also indicated that cleaning of core plugs with only injection of SFW and crude
oil, will require many pore volumes of SFW to remove the mud contaminations.
SEM-Analysis
After cleaning with solvents, inlet, mid and outlet samples of reservoir core were
analysed by SEM. Examples of SEM-images are shown in Figure 2. Clusters of barite,
clay and polymers were found in some samples. Bentonite clay in the mud has similar
elemental composition as other clay minerals, and is difficult to identify by the used
SEM-method. Since clay was found together with barite particles, it is likely that the
samples also contained bentonite. In the images, polymers similar as in the reference
samples were found. The SEM-images showed that the reservoir core plugs contained
mud components even after cleaning with solvents. The elemental compositions based on
EDS-spectra of the cleaned reservoir rock samples are given in Table 2. Since all samples
contained barite, Ba and S were present in the samples. No systematic variation of
elemental concentrations from the inlet to outlet was found. The reservoir core plugs
were also drilled from the seal peel in the longitudinal direction. In another study
reservoir cores were also cleaned by toluene/methanol cycles before they were drained to
Swi, aged with STO and used in water flooding experiments. After new toluene/methanol

SCA2015-016

5/12

cleaning after water flooding, the core plugs were also found to contain mud
contaminations. See examples of SEM-images in Figure 3.
Oil Extract
Chromatogram of crude oil showed typical crude oil pattern of n-alkanes at least up to
C41 (Figure 4a). The GC-analysis of extracts showed that crude oil components were still
present in the core plug even after cleaning with toluene/methanol cycles (Figure 4b, c
and d). Since mainly high molecular weight components were present in the cleaned
reservoir rock, it indicated precipitates of these components which were difficult to
remove during cleaning with toluene/methanol cycles. Since the amount of extractable
material was increasing from inlet to outlet in the core plug (Table 3), it appears that the
solubility with these oil components has been low.
Spontaneous Imbibition And Forced Imbibition
To confirm that toluene/methanol cleaning resulted in water-wet rock, spontaneous and
forced imbibition were studied by preparing two cleaned reservoir core plugs using
Isopar H (synthetic oil) as oil. Established Swi was similar for the two core plugs, 0.18
and 0.19 respectively. The spontaneous imbibition was rather high and fast for both core
plugs (Figure 5). When viscous flooding was carried out with SFW after the spontaneous
imbibition experiments, no additional oil was produced. Increase in the injection rate
from 3 to 7ml/min gave no oil production and the differential pressure across the core
plugs was rather stable at the different rates. No capillary end effects were therefore
observed. These results showed that the core plugs were water-wet.
Even though the cleaned core plugs were found to be water-wet, the wettability after
aging with crude oil may be affected by invasion of mud particles. Barite has rather small
area/weight ratio, and will be water-wet. Clay has high area/weight ratio and invasion of
bentonite clay will increase the surface area. Aging with crude oil may therefore give less
water-wet conditions than without invasion. In addition, invasion of mud particles may
have altered the permeability.
Oil-Based Mud: Evaluation Cleaning Procedures
Cleaning With STO Injection
Sandstone core plugs were first aged with STO at Swi. When STO was displaced by the
mud filtrates, the oil permeability was only slightly changed (Table 4). After new aging
for 14 days, the mud filtrate was replaced by STO. The oil permeability then became only
slightly higher than before the exposure to the mud filtrates. Minor changes in oil
permeability was observed when STO was injected instead of mud filtrate in the
reference core plug. For the core plug exposed to the mud filtrate with mineral base oil
(MFM), the STO injection was stopped after approx. 4PV, because the differential
pressure across the core plug was stable. IFT between the oil-phase effluent and SFW
was at this point in the same range as between MFM and SFW (Figure 6a). This means
that at least one emulsifier was still released from the rock, and equilibrium conditions
were not established inside the core plug. For the core plug exposed to the mud filtrate

SCA2015-016

6/12

with synthetic base oil (MFS), IFT was found to increase from approx. 3PV to approx.
7.3PV (Figure 6a). IFT was then still lower than between STO and SFW. This also
showed that at least one emulsifier was produced during a rather long period with backproduction. The release of emulsifier(s) was therefore slow in both plugs exposed to mud
filtrate. After cleaning with STO injection, the wettability of the core plugs was
characterized by spontaneous imbibition of SFW. The spontaneous imbibition of SFW
was in the beginning slower for the core plugs exposed to mud filtrates than for the
reference core plug not exposed to mud filtrate (Figure 6b). This means that the mud
exposed core plugs were less water-wet than the reference core plug.
Capillary Desaturation After Mild And Strong Cleaning
An atypical CDC was determined after methanol cleaning of the composite reservoir core
(Figure 7). The same reservoir core was then prepared by cleaning with
methanol/toluene, acetic acid and ethanol. The CDC profile was then as expected for
water-wet rock where the residual oil at low capillary number is distributed as isolated
droplets (Figure 7). This showed the importance of the type of cleaning procedure. When
these measurements were carried out, the understanding was that sandstone rock should
be water-wet. Since atypical CDC was measured after methanol cleaning, it was based on
results presented by Garnes et al. [6] assumed that the reservoir core was contaminated
by oil-based mud. During injection of acetic acid the mineral composition of the reservoir
rock was probably altered, e.g. dissolution of carbonate minerals. Today many of the
sandstone oil reservoirs are characterized as mixed-wet. In recent years it has been shown
that atypical CDC can be determined for the wetting phase [5] and also for mixed-wet
rocks [7].

DISCUSSION
All mud components should ideally be removed during preparation of reservoir core
plugs. The main challenge is that the muds and mud filtrates contain many components. It
has been shown in the present study that analysis of effluent samples during cleaning of
core plugs can confirm that the mud components that are easiest to remove are removed,
e.g. ions in water-based mud and emulsifiers in oil-based mud. If these components are
still present in the rock, the rock is probably not representative. It has also earlier been
reported that emulsifiers can adsorb onto the rock and thereby alter the wettability [8].
Emulsifiers will not give any colour to the effluent, but the effluent can be analysed for
these components or IFT can be measured. If the reservoir rock has been invaded by
water-based mud, it is important that the composition of brine in the rock is changed to
FW composition. It has been shown in other studies that potassium can reduce the
concentration of divalent cations onto clay surfaces and thereby alter the wettability to
more water-wet [9, 10]. Differential pressure across the core plugs cant be used to
confirm that mud contaminations have been removed, because low concentrations of such
contamination will have minor effects on viscosity.
After cleaning, the rock can be crushed and analysed for mud components (e.g. by SEM)
and oil residue (e.g. by analysing oil extracts on GC). These methods are destructive.

SCA2015-016

7/12

Often extra core plugs are cleaned and prepared in experimental programs, and the
additional plugs can be analysed to confirm the quality of the cleaning. An alternative is
to analyse cores/plugs after the SCAL- or EOR-experiments in the same way. It was not
possible to identify bentonite clay by the used SEM method, but invasion of clay from
water-based muds will increase the surface area. This can affect the established
wettability conditions during aging with stock tank oil. The organoclay in oil-based muds
are oil-wet, and invasion of this clay will also alter the wettability of the reservoir rock.
It is challenging to analyse for all the mud components, because the muds their filtrates
contain many components of technical grade (i.e. standard set for quality is not
established). If mud components are present, the question is to what extent do they affect
the results. If several SCAL- or EOR-experiments have been carried out, it can be
investigated whether the results depend on the amount of mud invasion in the different
cores/plugs. The best chance will be to get as low mud invasion as possible during
coring. Invasion will depend on the sampling conditions, mud design, and over balance
used in the coring process, but will in general decrease with decreasing permeability.
The cleaning procedure for reservoir core plug can be improved by including chemical
analyses. Details in the preparation procedure will depend on the composition of the rock,
reservoir fluids and mud system. In the evaluation of the cleaning procedure for waterbased muds, oil components with high molecular weight were found in the oil residue.
The solvent cleaning was carried out at 60C. It will probably be possible to remove
these components more efficiently by cleaning the core plugs at higher temperatures.
Cleaning at higher temperature will increase the diffusion rate and the solubility of the
high molecular oil components. This is an example of how chemical analysis can be used
to improve the core preparation procedure.

CONCLUSIONS
In evaluation of the cleaning procedure for high permeability sandstone reservoir core
plugs with potential contamination by water-based mud, injection of several PV with
SFW were required to change the brine composition in the core plugs back to SFW
composition. Mud components were found by SEM-analysis to be present in cleaned
reservoir rock, also after water flooding. The cleaned reservoir rock was water-wet, but
mud components present in the rock affect the wettability conditions prepared by aging
with crude oil. Cleaned reservoir rock was also found to contain oil components of high
molecular weight.Slow release of emulsifiers from mud filtrate of oil-based muds was
observed during cleaning of sandstone with STO injection. After this cleaning the core
plugs exposed to mud filtrate were found to be less water-wet than reference core plug
without exposure.
The measured CDC for sandstone reservoir rock with potential contamination with oilbased mud, was found to depend on the cleaning procedure. Atypical CDC was measured
after methanol cleaning, and typical CDC was measured for water-wet rock after strong
cleaning (methanol/toluene, acetic acid and ethanol). The strong cleaning probably
altered the rock composition.

SCA2015-016

8/12

Mud components with the potential to affect the flow conditions should be removed
during cleaning of reservoir core plugs, otherwise, ideally, these core plugs should not be
used. In preparation of core plugs it is recommended to analyse effluent samples for mud
components to confirm that they are removed, at least the easiest removable components.
The cleaning procedure should be evaluated for the actual reservoir rock by analysing
cleaned plugs for wettability and mud components. For some mud components, chemical
analysis can be challenging because they are of technical grade, have similar structures as
other mud components and are in complex mixtures. Important minerals should not be
removed from the original reservoir rock during cleaning. It is recommended to analyse
core plugs in experimental programs to investigate whether the presence of mud
components may have affected the results.

ACKNOWLEDGEMENTS
The authors acknowledge Lundin Norway for financing and supporting the work on
water-based mud presented in this paper. They also acknowledge the Research Council of
Norway and the industry partners; ConocoPhillips Skandinavia AS, BP Norge AS, Det
Norske Oljeselskap AS, Eni Norge AS, Maersk Oil Norway AS, DONG Energy A/S,
Denmark, Statoil Petroleum AS, GDF SUEZ E&P NORGE AS, Lundin Norway AS,
Halliburton AS, Schlumberger Norge AS, Wintershall Norge AS of The National IOR
Centre of Norway for support.

REFERENCES
1 Skopec, R. A. 1994. Proper Coring and Wellsite Core Handling Procedures: The First
Step Toward Reliable Core Analysis. SPE-28153. JPT. 46 (04).
2 Anderson, W. G. 1986. Wettability Literature Survey- Part 1: Rock/Oil/Brine Interactions and the Effects of Core Handling on Wettability. SPE-13932. JPT: 38 (10).
3 Adamson, A.W., Physical Chemistry of Surfaces. 5th ed. New York: John Wiley and
Sons, Inc., ISBN 0-471-61019-4, 1990, p.777.
4 Morrow, N.R., Chatzis, I., & Taber, J.J. 1988. Entrapment and Mobilization of
Residual Oil in Bead Packs. SPE-14423 SPE Res Eng 3 (3): 927-934.
5 Chukwudeme, E.A., Fjelde, I., Abeysinghe, K.P., & Lohne, A. 2014. Effect of
Interfacial Tension on Water/Oil Relative Permeability on the Basis of History
Matching to Coreflood Data. SPE-143028. SPE Res Eval & Eng 17 (01): 37-48.
6 Garnes, J.M., Mathisen, A.M., Scheie, A., & Skauge, A. 1990. Capillary Number
Relations for Some North Sea Reservoir Sandstones. SPE-20264, IOR Symp., Tulsa,
22-25 April.
7 Fjelde, I., Lohne, A. & Abeysinghe, K. P. 2015. Critical Aspects in Surfactant
Flooding Procedure at Mixed-wet Conditions. SPE-174393. Europec 2015, Madrid, 14 June.
8 Yan, J. N., Monezes, J. L. & Sharma, M. M. 1993. Wettability Alteration Caused by
Oil-Based Muds and Mud Components. SPE-18162. SPE Drill&Compl. 8 (01): 35
44.

SCA2015-016

9/12

9 Fjelde, I., Polanska, A., Asen, S. M. & Taghiyev, F. 2013. Low Salinity Water
Flooding: Retention of Polar Oil Components in Sandstone Reservoirs. A24. 17th
European Symp. IOR. St Petersburg, Russia. 16-18 April.
10 Omekeh, A., Friis, H. A., Fjelde, I. & Evje, S. 2012. Modeling of Ion-Exchange and
Solubility in Low Salinity Water Flooding. SPE-154144. IOR Symp., Tulsa, 14-18
April.
Table 1. Composition of SFW.
Concentration
Ion
(mg/l)
Na+
16105.6
K+
278.0
Mg2+
581.0
Ca2+
2920.0
Sr2+
103.5
Ba2+
5.6
Cl31744.4
NO3146.5
SO42281.1

Table 2. Elemental compositions (w%) of inlet, mid and outlet

samples of cleaned reservoir core plug based on EDS-spectra.
Element
O
Na
Mg
Al
Si
S
K
Ba
Fe

Inlet
36.3
1.2
0.1
6.9
45.5
1.0
3.8
3.7
1.5

Mid
37.7
0.6
0.1
6.6
43.6
0.9
5.3
3.5
1.8

Outlet
38.5
0.4
0.2
5.4
45.2
1.0
4.0
3.5
1.8

Table 3. Amount of oil extracted Table 4. Less water-wet core plugs (by aging with STO) and exposed to
from cleaned reservoir core plug. mud filtrates with mineral and synthetic base oils (MFM and MFS).
Sample
Extracted material
Ref. core plug MFM core plug MFS core plug
id
(mg/kg Rock)
Step
keo(Swi)
keo(Swi)
keo(Swi)
[mD]
[mD]
[mD]
Inlet
62
After first aging with STO
96
112
130
Mid
132
After aging with mud filtrate ----------------96
136
Outlet
604
(reference aged with STO)
After second aging STO
99
119
144
Element
a. K

b. S

First FW injection

Second FW injection

SCA2015-016

10/12

c. Ca

d. Mg

e. Na

Figure 1. Effluent element concentrations during SFW injection before and after solvent cleaning of four core plugs.

Quartz grain with clusters of clay,

polymers and barite

Barite and clay

Quartz grain with clay and polymer

Barite and polymer

Inlet

Mid

SCA2015-016

Outlet Quartz with coating of clay and barite

11/12

Feldspar with coating of clay, polymers & barite

Figure 2. Examples of SEM-images of samples from cleaned reservoir core.

Barite and clay

Barite and clay

Figure 3. Examples of SEM-images of samples from water flooded reservoir core plugs.
a.

Crude oil

b.

Inlet sample

c.

Mid sample

d.

Outlet sample

SCA2015-016

12/12

Figure 4. Gas chromatogram of crude oil and oil extracts from inlet, mid and outlet of cleaned reservoir core plug.

Figure 5. Oil saturation (So) as function of time during spontaneous imbibition of SFW into reservoir core plugs.
a.

IFT between effluent samples and FW

b.

Spontaneous imbibition of formation water

Figure 6. IFT between effluent samples and FW during soft cleaning by STO and the following spontaneous
imbibition experiments with SFW. (MFM and MFS: Mud filtrates with mineral and synthetic base oils.

Figure 7. CDC, normalized residual saturation vs capillary number (S or/Sorw vs Nc) after methanol cleaning (soft) and
after cleaning with methanol/toluene, acetic acid and ethanol cleaning (strong) of reservoir core.

SCA2015-017

1/12

PRESERVING INITIAL CORE WETTABILITY DURING

CORE RESTORATION OF CARBONATE CORES
Paul Hopkins, Tina Puntervold and Skule Strand
University of Stavanger, Norway
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
A correct restoration of carbonate cores is very important because the wettability will
influence the results of relative permeability, capillary pressure and oil recovery tests.
The acidic polar components in the crude oil are the main wetting parameter for
carbonates, and will dictate the wettability of the rock surface. This work aims to
determine the suitability of using mild core cleaning with kerosene and heptane during
core restoration to preserve the initial reservoir wettability in the reservoir cores during
core restoration. Initially a water-wet outcrop chalk core with Swi was flooded with crude
oil. Effluent samples of produced oil were collected and the acid number (AN) was
measured as a function of pore volume injected. The results showed that the Acidic polar
components in the crude oil were adsorbed onto the rock surface, and gradually an
equilibrium between the AN in the injected oil and AN in the effluent oil was achieved. A
wettability test confirmed that the water-wetness of the core was drastically reduced. The
chalk core was then put for restoration. First, the core was mildly cleaned by flooding
with kerosene to displace the crude oil, followed by heptane for removal of kerosene.
During the second crude oil saturation and flooding, the adsorption of acidic organic
polar components onto the rock surface was drastically reduced, confirming that the rock
surface had maintained approximately 3/4ths of the initially adsorbed acidic polar
components during the mild core cleaning process.

INTRODUCTION
In carbonate reservoirs, the success of an oil recovery is primarily linked to the initial
wetting conditions of the rock. In an oil reservoir, the initial wettability is established by a
chemical equilibrium between oil, brine and rock over millions of years of interaction.
The properties of the initial wetting state will decide physical parameters such as the
capillary pressure, the relative permeabilities and fluid distribution. The AN, acid
number, is a measure of the content of surface-active components, and is regarded as one
of the most important parameters dictating the initial wetting of a carbonate rock. The AN
is related to the amount of carboxylic groups that are present in the oil. It may be
measured and quantified by titration and has the unit mg KOH/g oil. When analysing
reservoir core material it is essential that the initial wetting is preserved, so that the
wettability of the core may be investigated in the most representable state. Therefore, in a
laboratory, it is crucial that steps are taken during preparation to preserve the initial core

SCA2015-017

2/12

wettability. Core material is often cleaned to a completely water-wet state and the core is
aged with the formation water and reservoir oil to build up the initial conditions [1]. In
practice, it is common to clean core material with Toluene followed by methanol.
Toluene should remove hydrocarbons including asphaltenes and adsorbed polar
components. Methanol displaces both water and toluene. The questions asked in this
paper are: Will a more representative wettability be achieved by using mild cleaning, and
can we perform wettability tests to validate that polar components have been preserved
on the surface?
The wettability will be determined by a chromatographic wettability test at Sor (residual
oil saturation) [2]. In this chromatographic wettability test, a separation of ions will only
take place on the water-wet areas of the pore surface. The area between a non-adsorbing
tracer and sulphate is proportional to the water-wet area. The ratio of this area against the
area of a completely water-wet core will determine a wettability index in chalk, ranging
from 0-1.

EXPERIMENTAL
Materials
Outcrop chalk from Stvens Klint, Copenhagen, Denmark, was used as model rock. The
chalk has high porosity (45-50%) and low permeability (2-3 mD) and is relatively
homogeneous. The outcrop chalk core consists of 98% pure biogenic CaCO3 and is
considered similar to North Sea chalk reservoirs. The properties of the core are given in
Table 1.
Brine Composition
Synthetic formation water, VB0S, modelled on the Valhall field, was the brine used as
the formation water in the experiments. SW0T and SWT are brines used for the
chromatographic wettability test. Brine compositions are reported in Table 2.
Oil Properties
A crude oil was diluted with heptane in the ratio of 60/40 respectively. The oil was
centrifuged and filtered through a 5m Millipore filter. An acid number (AN) of 1.7
mgKOH/g oil was measured. A part of this oil was treated with silica gel to remove
surface active polar components and thereafter filtered, giving an AN of ~0.0 mgKOH/g
oil.
A mixture of treated and untreated oil was prepared at a specific ratio and analysed. The
designed oil, RES40-0.35, had an AN of 0.35 mg KOH/g and a base number (BN) of
0.25mg KOH/g. The oil properties are given in Table 3.
Analyses Of Acid And Base Numbers In Crude Oil Samples
The acid and base numbers, AN and BN, of the crude oil samples were analysed by
potentiometric titration. The methods used are modified versions of ASTM 665 and

SCA2015-017

3/12

ASTM 2898 [3]. The reproducibility of the acid number titration tests is presented in
Table 4.
Preparation Of Chalk Core
The cores were prepared following a procedure set by Puntervold et al.[4]. All cores used
were initially flooded with 5 pore volumes (PV) of de-ionized water (DI) for a removal of
easily dissolvable salts and sulphate. The cores were dried to a constant weight at 90 C.
Establishing Initial Water Saturation
The initial water saturation, Swi, was established using the desiccator technique [5]. After
the core had achieved a Swi of 10%, the core was equilibrated in a sealed container for 4
days to allow the even distribution of salts within the core.
Oil Saturation And Establishing Core Wetting
The chalk core with Swi =0.1 was mounted in a Hassler core holder with an overburden
pressure of 20 bar and with a back pressure of 6 bar. The core was oil flooded at 50 C at
a rate of 0.1 ml/min. Effluent samples of produced oil were collected in sealed vials using
a fraction-collector.
The core was flooded with oil until the AN in the effluent samples reached the same
concentration as initially present in the oil, AN = 0.35 mg KOH/g. A total of 15 PV of oil
was flooded through the core. By plotting the AN against PV of oil injected, the
adsorption of acidic oil components onto the rock surface can be expressed.
Core Aging
Once the core had been flooded with 15 PV of crude oil, it was wrapped in Teflon tape.
The wrapped core was placed in an aging cell surrounded by oil, and aged at 50 C for
two weeks to achieve a homogeneous core wetting.
Core Wettability By Chromatographic Wettability Test
The chromatographic wettability test was designed by Strand et al.[2] and is based on the
chromatographic separation between sulphate and a non-adsorbing tracer. The test is
performed at Sor at ambient temperature. The chromatographic separation of the sulphate
and the tracer is proportional to the water-wet area of the chalk core surface.
The core is flooded with the brine SW0T. SW0T is depleted in sulphate and is used to
displace the remaining fluid in the core to bring the sample to Sor, as well as to establish a
stable flow. Then the SWT brine, containing SCN- and SO42-, is injected and effluent
samples are collected into sealed vials. SO42- ions have an affinity towards the water-wet
chalk surface, and the chromatographic separation area, Awet, between SCN- and SO42eluent curves gives a relative measure of the water-wet rock surface. Compared with the
area determined from a completely water-wet chalk core, AWW, the wettability index ICW
describing the water-wet fraction of the chalk surface can be calculated according to
equation 1 below.

SCA2015-017

4/12

(1)

Ionic Concentration Analyses

The ionic concentrations of the effluent samples from the chromatographic wettability
test were analysed using an ion chromatograph, ICS 3000, delivered by Dionex
Corporation, USA. The samples were diluted 1:200 times prior to the analysis. The
anionic concentrations in the effluent were calculated using an external standard method.
=

Core Restoration
After the first polar component adsorption test, the core was restored. A mild core
cleaning procedure with low aromatic kerosene and heptane was performed to remove the
residual oil. The first step of the procedure is a miscible fluid displacement using
kerosene. When the core is clean and the effluent is clear or consistent in colour, the
second step of the procedure is to displace the kerosene with n-heptane. As the final step,
the core was flooded with DI water to remove heptane and easily dissolvable salts. The
water and heptane was afterwards removed by evaporation in a standard heating chamber,
at 90C, until constant weight. The procedure was then repeated, establishment of initial
water saturation and core saturation with oil, ageing, water flooding to Sor, and wettability
determination.

EXPERIMENTAL RESULTS
A correct initial wetting condition is crucial for achieving reliable experimental results
from core analyses. The initial wetting conditions in the core dictates capillary pressure
measurements and are very important for relative permeability calculations which is an
important parameter in most reservoir modelling and simulations. Oil recovery tests and
Smart Water effects are also dictated by the initial wetting properties of the restored
cores. An optimized core restoration laboratory procedure that preserves the initial
reservoir wetting during core cleaning is of great interest for the industry.
In this work, a water-wet outcrop chalk core was used. The effect of core cleaning and
core restorations on rock wettability is discussed.
Initial Wetting of the Outcrop Chalk Core
The outcrop chalk cores used in these experiments are initially completely water-wet.
Cores saturated with oils without polar components, like mineral oils, will not change the
initial wettability. Restored cores imbibe water very rapidly giving oil recoveries above
70 % OOIP. The chromatographic wettability test was performed at ambient temperature
on an outcrop core at 100% initial water saturation, Figure 1. The core details are
presented under reference core on Table 1. The results show a large separation between
the non-adsorbing tracer thiocyanate and sulphate which adsorbs to water-wet chalk
surfaces. The calculated area between the tracer and sulphate eluent curves is Aww =
0.254, represents a very water-wet rock surface area.

SCA2015-017

5/12

Initial Adsorption of Carboxylic Material Onto Water-Wet Chalk Surface

The chalk core used was initially very water-wet. After establishing Swi = 0.1, the core
was mounted in a core holder with a confining pressure of 20 bar and with a back
pressure of 6 bar. The core was flooded at 50 C with the oil containing polar
components with an AN=0.35 mg KOH/g oil. Effluent samples of the produced oil were
collected and analysed for the content of acidic polar components, AN. Figure 2 shows
the AN in the effluent with increasing PV injected.
The results confirm adsorption of polar acidic components towards the pore surface in the
chalk core. The initial measurements, the first 2 PV, show that the effluent oil has a very
low AN, Figure 2. This shows that the acidic polar material in the oil adsorbs onto the
surface of the chalk core, which results in an effluent oil with minimal acidic
components. As more oil is flooded into the chalk core the surface becomes more
saturated with acidic polar material. There is less ability for the acidic polar groups to
adsorb on to the surface, due to the lack of available surface sites. As a consequence, the
AN in the oil effluent increases. After 13PV of injected oil, the system establishes an
equilibrium, where the AN in the effluent samples is equal to the AN of the injected oil,
0.35 KOH/g oil.
Core Wetting Of The Outcrop Chalk Core After Oil Flooding
A new chromatographic wettability test was performed on the oil flooded core. The core
was successively flooded with SW0T to Sor, 0.25, at ambient temperature, and the
chromatographic test was performed by switching to SWT at ambient temperature. The
results from the chromatographic wettability test is presented in Figure 3.
The chromatographic separation between SCN- and SO42- is given in Figure 3. The
calculated area between the tracer curve and sulphate curve is now dramatically lower
compared to the initially water-wet core, in Figure 1. The calculated area is now Awet =
0.055, confirming a dramatic reduction in water-wet sites on the pore surface after
flooding 15 PV with an acidic oil.
Adsorption Of Acidic Organic Compounds In A Second Oil Flood
The Core#1 was then prepared for a second oil flooding test. The core, at Sor, was first
mildly cleaned by kerosene, heptane and then flooded with DI water and dried. Initial
Swi=0.1 was established and the core#1 was mounted in a core holder. The core was once
again flooded at 50 C with a RES40-0.35 oil containing polar components. Effluent
samples of produced oil were collected and then analysed for the content of acidic polar
components. The results of the adsorption of acidic components on a chalk surface that
has previously seen oil are shown in Figure 4. Unlike the first adsorption test, the AN in
the effluent samples starts relatively high, 0.13 mg KOH/g. This indicates that there is
something preventing a full adsorption of the carboxylic material onto the chalk surface
as seen in the first adsorption test on a water-wet sample. This further implies that the
pore surface already has acidic organic components adsorbed to the pore surface, and

SCA2015-017

6/12

reduces the adsorption of more acidic organic material. The slope of the AN curve is
relatively sharp, and is brought to a new equilibrium within 6 PV of oil flooded.
Core Wetting of the Outcrop Chalk Core after Oil Flooding
The core wetting after the secondary Oil Flooding was tested by using the
chromatographic wettability test. The core was flooded with SW0T to Sor, 0.2 at ambient
temperature, and the chromatographic test was performed flooding SWT at ambient
temperature. The results from the chromatographic wettability test are presented in Figure
5. This result shows a very low separation between the tracer and sulphate curve. The
calculated area, Awet = 0,008, is dramatically reduced compared with the previous
wettability tests for water wet core, the core that has been injected 15 PV of crude oil.
The test shows that the wettability of the core has become even less water-wet.

DISCUSSION
In this experimental work, we started with an initially water-wet core sample. This could
simulate the initial situation in a reservoir before crude oil accumulates into the water
filled reservoir trap. The first oil flooding we performed on a water-wet core sample
could represent the process when the crude oil migrates from source rock into a water
saturated reservoir trap. This oil-flooded core could represent the initial situation inside
the reservoir regarding adsorption of acidic polar components and the initial reservoir
wetting properties that could be established. Initial content of acidic polar components in
the crude oil that migrates into the reservoir, and number of crude oil reservoir refilling
during geological time will definitely contribute with large uncertainties to a correct
wettability description.
Results show that the initial adsorption of acidic organic compounds increases with
increasing PV of acidic oil injected, Figure 2, and that the adsorption at the end reaches
an equilibrium. The adsorption of polar components onto the rock surface, reduced the
water-wet surface area of the core dramatically, which is clearly seen when we compare
Figure 1 and 3. The area between these curves in the water-wet core is Aww = 0.254 and
for the core flooded with crude oil, the area is Awet=0.055 which corresponds to a fraction
of water-wet surface area:
=

0.055
=
= 0.22
0.254

(2)

The mild core cleaning that was performed on the oil flooded core, preserved large
amount of polar components that were adsorbed to the pore surface. This is clearly shown
when we compare Figure 2 and 4, Figure 7.
During the second oil flooding far less acidic polar components was adsorbed compared
to the first oil flooding. The area above the adsorption curves represent the total amount
of acidic organic material adsorbed onto the surface, AAT. The Acid Adsorption (AA)
was determined for both adsorption tests, before and after mild cleaning.

SCA2015-017

7/12

(3)

AN plateau is the average highest acid number when the oil reaches equilibrium with the
surface. ANx is the acid number value of a given sample number. Giving the total acid
adsorption, AAT:
=

(4)

The calculated area above the curve shown in Figure 2 was AAT = 1.47. The
corresponding area in Figure 4 was AAT = 0.37. This partly represents the amount of
acidic organic components removed from the pore surface during the mild core cleaning,
and partly the amount of acidic components that could be re-adsorbed during the second
core restoration. Approximately 75% of the initial adsorbed material was still present on
the pore surface. An increased number of PV injected with oil, will reduce the water-wet
surface area inside the core. This could have dramatic effects on initial wetting and on the
capillary pressure.
There is not a direct link between the AAT and the water-wet surface area measured by the
chromatographic wettability test when we compare wettability tests from 1st restoration,
Figure 3, and the 2nd restoration, Figure 5, shown in Figure 8.
The separation area between these curves are now Awet= 0.055, 1st restoration, and
Awet= 0.008, 2nd restoration, which corresponds to the water-wet fraction of Icw= 0.22
and Icw=0.03 respectively.
Even though the total acidic adsorption capacity, AAT, was reached in both experiments,
the wettability test showed a reduction in available water-wet surface area, ICW , from
22% to 3 % after the second oil flood. This could be explained by the presence of acidic
organic components in the crude oil with different surface reactivity. The more PV of oil
flooded through the core, the more higher reactive acidic components will be
concentrated at the rock surface, compared to what is initially present in the crude oil [6,
7]. Injection of 15 PV of crude oil in the second restoration was too large to reproduce
the initial wetting achieved in the first restoration.

CONCLUSIONS
In this work, a water-wet outcrop chalk core was used. The effect of core cleaning and
core restorations on rock wettability is discussed.
The results showed that the adsorption of acidic organic compounds onto the pore
surface increases with increasing PV of acidic oil injected.
The adsorption of acidic organic compounds reached an equilibrium after 13 PVs of
crude oil flooded in the first restoration.

SCA2015-017

8/12

The adsorption of polar components to the pore surface reduced the water-wet core
surface.
Mild core cleaning preserved approximately 3/4 of the polar components that were
adsorbed to the pore surface.
In the second restoration acidic the adsorption equilibrium was reached after only 8
PVs of crude oil.
After 15 PV crude oil injected in the second restoration, the water surface area had
been reduced from 22% to 3%, giving a less water wet core compared to the first
restoration.

The mild core cleaning procedure preserved most of the adsorbed acidic components.
Lower crude oil volume was needed to restore the initial core wetting.

NOMENCLATURE
AN
BN
ASTM
D
L

PV
Swi
Kro
Krw
Awet
AWW
VB0S
TDS
SW0T
SWT
C/Co
ICW

Acid number
Base number
American Society for Testing and Materials
Core diameter
Core length
Core porosity
Pore volume
Initial water saturation
Relative permeability of oil
Relative permeability of water
The area between the thiocyanate and sulphate curve of a sample
The area between the thiocyanate and sulphate curve of a water-wet
sample
Valhall brine with no sulphate
Total dissolved salt
Seawater without thiocyanate tracer
Seawater that contains thiocyanate tracer
Relative concentration of ion in effluent fractions
Chromatographic wettability index

AKNOWLEDGEMENTS
BP is acknowledged by the authors of this work for financial support and for the
permission to publish this work.

REFERENCES
1. Shariatpanahi, S.F., et al., Wettability restoration of limestone cores using core
material from the aqueous zone. Petroleum Science and Technology, 2012. 30(1-9).

SCA2015-017

9/12

2. Strand, S., D.C. Standnes, and T. Austad, New wettability test for chalk based on
chromatographic separation of SCN- and SO42-. Journal of Petroleum Science and
Engineering, 2006. 52(1-4): p. 187-197.
3. Fan, T. and J.S. Buckley, Acid Number Measurements Revisitied Society of Petroleum
Engineers, 2007.
4. Puntervold, T., S. Strand, and T. Austad, New method to prepare outcrop chalk cores
for Wettability and oil recovery studies at low initial water saturation. Energy &
Fuels, 2007. 21(6): p. 3425-3430.
5. Springer, N., U. Korsbeck, and K. Aage. H, Resistivity index measurement without the
porous plate: A desaturation technique based on evaporation produces uniform water
saturation profiles and more reliable results for tight North Sea chalk. Prepared for
presentation at the International Symposium of the Society of Core Analyst, 2003.
6. Fathi, S.J., T. Austad, and S. Strand, Wettability alteration in Carbonates: The Effect
of Water-Extractable Acids in Crude Oil. Energy & Fuels, 2010. 24(5): p. 2974-2979.
7. Fathi, S.J., T. Austad, and S. Strand, Effect of Water-Extractable Carboxylic Acids on
Wettability in Carbonates. Energy & Fuels, 2011. 25(6): p. 2587-2592.

SCA2015-017

10/12

Table 1. Core data for chalk cores used.

Core material
A1
A1b
96,78
92,28

Stvens Klint Chalk

Dry weight, g
Length, cm

6,21

Diameter, cm
3,74
Bulk Volume, cm3
68,22
Sat. Weight, g
128,56
3
Density of FW 10xD g/cm 1,00
Pore Volume, ml
31,69
Porosity %
46 %

Reference core
84,27

6,21

5,43

3,74
68,22
119,79
1,00
27,43
40 %

3,70
58,33
112,01
1,02
27,14
47 %

Table 2. Compositions of the brines used.

Brine Reference VB0S
Salt
mole/l

SW0T
mole/l

SW1/2T
mole/l

0,009

0,002

0,002

Cl

1,066

0,583

0,583

2SO4

SCN

0
0

0
0

0,012
0,012

Mg2+

0,008

0,045

0,045

Ca
Na+

0,029

0,013

0,013

0,997

0,460

0,427

0,005

0,01

0,022

0,012

1,022

1,022

HCO3
-

2+

K
+
Li
3

Density g/cm

Table 3. Properties of the oil used.

Crude oil
AN mgKOH/g
0,33-0,35
BN mgKOH/g
0,24
Viscosity cP 25C
3,25
Density g/cm3 25C
0,808

SCA2015-017

11/12

Table 4. The reproducibility of Acid Number titration tests.

AN range
Trial 1 Trial 2 Trial 3
Sample of low AN
0,06
0,06
0,06
Sample of medium AN
0,27
0,27
0,26
Sample of high AN
0,34
0,34
0,34

Stdev.
0,00
0,01
0,00

Figure 1. Chromatographic wettability test on a 100%

water saturated core, Core#1. The relative effluent
concentrations of SCN- and SO42- are plotted vs. PV
injected.

Figure 2. The adsorption of carboxylic material

during flooding of an oil with AN=0.35 mgKOH/g
into chalk Core#1 at 50C, with a flow rate of 0.1
ml/min (4 PV per day). The AN in the effluent
samples is presented vs. PV injected.

Figure 3. Chromatographic wettability test of Core#1

after the first oil flooding of 15 PV with an oil having
AN = 0.35 mgKOH/g oil. The test on Core#1 was
performed at Sor. The relative effluent concentrations of
SCN- and SO42- are plotted vs. PV injected.

Figure 4. The adsorption of Acidic organic material

during the second Oil flooding of Chalk Core#1 at
50C with RES40-0.35 oil, with a flow rate of 0.1
ml/min (4 PV per day). The AN in effluent samples
is presented vs. PV injected.

SCA2015-017

12/12

Figure 5. Chromatographic wettability test of Core#1

after the second oil flooding of 15 PV with RES400.35 oil. The test on Core#1 was performed at Sor. The
relative effluent concentrations of SCN- and SO42- are
plotted vs. PV injected.

Figure 6. Chromatographic wettability test of a

reference core and Core#1 after the first oil flooding of
15 PV with RES40-0.35 oil. The reference core was
100% saturated. The test on Core#1 was performed at
Sor. The relative effluent concentrations of SCN- and
SO42- are plotted vs. PV injected.

Figure 7. A comparison for the adsorption of Acidic

organic material for both first and second Oil flooding
of Chalk Core#1. The tests were performed at 50C
with RES40-0.35 oil, with a flow rate of 0.1 ml/min (4
PV per day). The AN in effluent samples is presented
vs. PV injected.

Figure 8. Chromatographic wettability test of

Core#1 for both the first oil flood and second oil
flood. 15 PV was flooded with RES40-0.35 oil. Both
tests on Core#1 was performed at Sor. The relative
effluent concentrations of SCN- and SO42- are plotted
vs. PV injected.

SCA2015-018

1/12

THE IMPACT OF RESERVOIR CONDITIONS ON

WETTING AND MULTIPHASE FLOW PROPERTIES
MEASUREMENTS FOR CO2-BRINE-ROCK SYSTEM
DURING PRIMARY DRAINAGE
Ali Al-Menhali1,3, Ben Niu2,3, Samuel Krevor1,3
Department of Earth Science & Engineering, Imperial College London
2
Department of Chemical Engineering, Imperial College London
3
Qatar Carbonates and Carbon Storage Research Center, Imperial College London
1

This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in Newfoundland and Labrador, Canada, 16-20 August 2015

ABSTRACT
The wettability of CO2-brine-rock systems will have a major impact on the management
of safe carbon sequestration in subsurface geological formations. The wetting properties
of a system controls the mobility and trapping efficiency of CO2 during injection
processes for carbon storage as well as for enhanced oil recovery operations. Recent
contact angle measurement studies have reported significantly different wetting
behaviour with regard to pressure, temperature and water salinity. We report the results of
an experimental investigation into the wetting properties of CO2 and brine solutions in a
single Berea sandstone utilising measurements of the multiphase flow properties under a
wide range of reservoir conditions with pressures (5 to 20 MPa), temperatures (25 to 50
C) and ionic strengths (0 to 5 M kg-1 NaCl). Primary drainage capillary pressure curves
were measured in a horizontal core flooding apparatus using the semi-dynamic method to
investigate the wetting properties during CO2 injection process. The observations were
made using a reservoir condition core-flooding laboratory that included high precision
pumps, temperature control, the ability to recirculate fluids for weeks at a time and in situ
saturation monitoring with x-ray CT scanner. The wetted parts of the flow-loop are made
of anti-corrosive material that can handle co-circulation of CO2 and brine at reservoir
conditions. Measurements in the Berea sample were made using CO2-brine and N2-water.
The capillarity of the system, scaled by the interfacial tension, were equivalent to the N2water system. Thus reservoir conditions did not have a significant impact on the capillary
strength of the CO2-brine system through a variation in wetting. The capillarity is
consistent with general characteristics of drainage in strongly water-wet rocks. The fluid
distributions were observed using x-ray computed tomography and the spatial saturations
were investigated and were found to be invariant with different reservoir conditions in
homogeneous samples.

INTRODUCTION
While recent contact angle measurements have shown CO2 to generally act as a nonwetting phase in siliciclastic rocks, some observations report significantly different

SCA2015-018

2/12

wetting behaviour, if not entirely contradictory results, with regard to pressure,

temperature and water salinity of CO2-brine-rock systems. Additionally, there is a wide
range of reported contact angles for this system, from strongly to weakly water-wet. In
the case of some minerals, intermediate wet contact angles have been observed. Using
contact angle measurements on smooth crystal surfaces, [1] found that the CO2-brine
system was strongly water-wet on quartz and calcite with no impact on pressure and a
small shift with salinity up to 3.5M NaCl. [2] have also observed the system to be waterwet on quartz with little impact of both pressure and salinity from atmospheric to 10 MPa
and 0.01-1M NaCl solutions. In the case of mica, however, the system shifted from water
wet to intermediate wet with increasing CO2 pressure (> 10 MPa). Water wettability was
reduced by 25o with salinity change from 0.1 to 1 M NaCl. However, the results at 0.01
M NaCl didnt obey the trend. [3] have found that contact angles on silica increased up to
17.6 2.0 as a result of reactions with supercritical CO2. The observed increase
occurred primarily within the pressure range 710 MPa, but remain nearly constant at
pressure greater than 10 MPa. They also observed that the contact angle increased with
ionic strength nearly linearly with a net increase of 19.6 2.1 at 5.0 M NaCl. Pressure
had only a minor influence on the ionic strength effect. [4] have observed the system to
be water-wet on quartz with contact angle <30o for both low and high temperature and
pressure (T=30o P=7 MPa vs. T=50o P=20 MPa). However, they observed that the
contact angle decrease with ionic strength for both conditions. In microfluidic
experiments with silica micromodels, [5] observed contact angles ranging from water-wet
to mixed wet for the supercritical CO2-brine system as the salinity is increased from 0.015M NaCl.
In this study, we make several primary drainage (CO2 displacing brine) capillary pressure
measurements of at different reservoir conditions to investigate the effective wettability
during CO2 injection process that can last for decades in CCS projects. This process is a
very important if not the most critical stage in CCS projects for safe CO2 injection. A
solid and clear understanding of the wettability of the system is essential for proper
reservoir management and simulation of multiphase flow (CO2 plume) during CO2
injection. Figure 1 show that capillarity and multiphase flow effects may or may not be
important depending on the rock properties of the storage target and thus it can be
important to understand these properties in some detail. All of those graphs, however,
were scaled from the mercury-air system assuming that the CO2-brine system is strongly
water wet. This assumption presumes maximum capillarity in the system. If the system is
in fact less strongly water wet this manifests as a weakening of capillarity. As the system
becomes less wet with respect to a given phase, it matters less and less what the particular
pore structure looks like, the multiphase flow effects will be largely nullified.

SCA2015-018

3/12

Figure 1. Reproduced from [6] showing the impact of capillarity on the height of 2-phase region. On the
left, capillary pressure functions reproduced from plotted from observations with mercury-air displacement
(MICP) scaled to interfacial tension values relevant to the CO2-brine system. The Pc was converted to
height to demonstrate the height of the 2-phase region in a CO2 plume for different rock types. On the right,
thin sections under cross-polarized light to compare grain size and distribution of the four sandstone rock
types showing in the Pc curves graph. Blue die is used to visualize the pores. Clockwise from upper left:
Berea, Paaratte, Tuscaloosa, Mt. Simon.

Figure 2 shows two end member examples reproduced from the model developed in [7,
8] of a buoyant CO2 plume migrating upwards and along an impermeable boundary.
Weak capillarity results in a very narrow range in the plume over which the saturation
varies, corresponding to a relatively compact and fast moving plume. On the other hand,
strong capillarity results in a much larger range over which the saturation varies, and an
overall larger plume volume for the same mass of fluid and thus a more slowly migrating
system. Once the injection phase is completed, imbibition plays a major role in CO2
residual trapping. The impact of different reservoir conditions on residual trapping and
hysteresis is investigated on the same Berea sample and conditions and discussed in
details in earlier work [9].

Figure 2. Two end member examples reproduced from the model developed in [7, 8] of a buoyant CO2
plume migrating upwards and along an impermeable boundary.

The measurements of this work uses X-ray CT imaging in a state of the art core flooding
laboratory designed to operate at high temperature, pressure, and concentrated brines.

SCA2015-018

4/12

This is the first study looking systematically at the impacts of reservoir conditions on the
effective wettability in the CO2-brine-sandstone system in a single rock sample. We also
make measurements using N2 and water to assess the ability to use analogue fluids to
represent CO2 and brine systems and developed a method to quantify shifts in effective
wetting properties with changing reservoir conditions. We find no significant impact on
wetting during primary drainage within the range of reservoir and flow conditions
relevant to CO2 storage, consistent with traditional multiphase flow theory but despite
observations by others suggesting that wetting properties and multiphase flow in this
system are sensitive to pressure, temperature and brine salinity [10]. In this paper, we
report some of the results showing the impact of pressure on effective wettability and on
spatial saturations distributions.

MATERIALS AND EXPERIMENTAL CONDITIONS AND SETUP

The Rock Sample and Fluids
A single homogeneous 20.32 cm long, 3.81 cm diameter Berea sandstone core sample
was used for the entire tests performed in this study. Berea sandstone is an outcrop rock
from the United States which is widely used in studies as a benchmark and for its utility
in making comparisons with other studies. The sample was fired at 700 oC for 4 hours to
prevent fines migration by stabilizing swelling clays that otherwise would cause
permeability changes during core flooding tests. The faces of the core were machined flat
to ensure good contact with the end-caps. The sample was vacuum dried in air at 70 oC
overnight before each test. The absolute permeability of the sample to water was 212 mD
with an average porosity of 21% as measured by x-ray CT scanning by following a
protocol described in [11]. The skeletal density of the sample is 2.601 g/cm3 measured
with Helium pycnometer (AccuPyc 1330, Micromeritics) at 24 oC and 20 psi on a subsample (~ 0.8 cm3) that was cut from a section adjacent to the inlet face of the core
sample. Carbon dioxide and Nitrogen were used as the non-wetting phases in the coreflooding experiments both with 99.9% purity (BOC Industrial Gases, UK). The wetting
phase fluids used were deionized water or brine. The brine solutions were made of
deionized water and NaCl with total salt molality ranging from 0 to 5 mol kg-1. In this
paper, we report the capillary pressure experimental results investigating the impact of
pore pressure on wetting properties and CT x-ray images. The fluids interfacial tension
values as well as density and viscosity ratios for the experimental conditions are
summarized in Table 1. Experiment N. 4, as described in the table was the only
experiment where the wetting phase solution was deionized water doped with 5 wt%
Sodium Iodide. This was necessary to compare and quantify the saturation distribution
for the high density CO2 experiments as will be discussed in a later section.

SCA2015-018

5/12

Table 1. Experimental conditions and fluids properties of experimental results investigating the
impact of pore pressure on wetting properties and CT x-ray images

CO2 IFT from correlation developed in [12], b N2 IFT from [13], c CO2 densities and viscosities from
NIST chemistry web book [14], d Water and brine densities and viscosities from [15]

Core-flooding Experimental Setup

The experimental work was conducted using a state of the art multi-scale imaging
laboratory (core and pore scale) recently developed at Imperial College London designed
to characterise reactive transport and multiphase flow, with and without chemical
reaction for CO2-brine systems in both sandstone and carbonate rocks at reservoir
conditions [16]. The experimental setup is designed to replicate in-situ conditions of up to
120 oC and 30 MPa. Details of the experimental apparatus are reported in [11].

METHODOLOGY
Core-flooding was performed for capillary pressure measurements. Total injection flow
rates varied from 0.5 up to 50 mL min-1. The dimensionless capillary number, =
/, was used to ensure that local capillary equilibrium conditions applied during all of
the experimental flow rates. Where [m/s] and [Pa s] are the CO2 superficial velocity
and dynamic viscosity, respectively, [N/m] is the interfacial tension between CO2 and
brine. Capillary numbers around 10-4-10-5 indicate that viscous and capillary forces are
equivalent for the reservoir rock [17]. In this study, the capillary numbers for the highest
flow rates applied ranged between 10-7 to 10-6 and the pore-scale fluid distribution was
thus controlled by capillary forces.
We used the semi-dynamic capillary pressure method to measure the capillary pressure in
the whole core. The technique was developed by [18] based on a model proposed by [19]
and has been recently applied to measurements in CO2-brine systems by [20, 21]. The
technique applied here is described in [11] but we do not fit MICP to measurements for
effective contact angle estimation. We use a more practical reference know to be strongly
water-wet to evaluate wetting strength. We do this by measuring the semi-dynamic Pc of
N2-water on the same core sample and use it as a benchmark. The technique was
developed further in this work and the main modifications were (1) the use of a spacer at
the downstream end of the core to control the outlet boundary condition and (2) the use of
a numerical simulator to design the parameters for the test, e.g., the injection flow rate.
Brooks-Corey curves were fit to the core-flooding capillary pressure measurements [22].

SCA2015-018

6/12

S S

1/

w
w,irr
Pc = Pe ( 1S
)
w,irr

(1)

Where Pe [Pa] is the entry pressure (i.e. the minimum pressure required for the entry of
CO2 into the pores of the rock), Sw represent the water saturation, while Sw,irr represent
the irreducible water saturation, is a fitting parameter known as the pore size
distribution index.
Experimental Procedures
The first step was to expel air from the flow loop by flowing CO2 through all parts of the
experimental setup at an elevated pressure and exhausting to the atmosphere. This was
done at different stages by isolating different sections of the flow loop to insure that air is
expelled entirely. Then the system was pressurized with CO2 and heated to the
experimental pressure and temperature. This step was required to obtain the CO2
saturated core background scan. Next, water or brine, depending on the experimental
fluid, was injected into the loop to displace and dissolve all of the CO2 out of the core at
experimental pressure and temperature. A scan of the brine saturated core was obtained
for porosity measurements. Porosity and saturations calculations with full experimental
procedures for absolute permeability, fluids equilibration and semi-dynamic capillary
pressure are described in details in earlier work [11].

RESULTS
A stable pressure drop was achieved for each CO2 injection flow rate applied for semidynamic Pc before taking x-ray scans to measure saturations. The horizontal and vertical
error bars of the Pc measurements represent uncertainty caused by pressure fluctuations
during X-ray imaging time and saturation uncertainty based on the range of saturations
seen from 10 repeated scans, respectively. The measured Pc curves were fitted with
Brooks-Corey model, Eq. (1), objectively using a MatLab code. A single value was
assigned for as a single rock sample was used for the entire study.
The Impact of Reservoir Pressure on Wettability
To evaluate the effect of pressure on wettability three experiments were carried out at
pressures of 5, 10 and 20 MPa pore pressure at 50oC and 0 Mol kg-1. The pressures range
such that CO2 phase is transitioning from gas to low density scCO2 and finally to high
density scCO2. With increasing pressure, the interfacial tension between the fluids
decreased from 47 to 30 mN/m, Table 1. This was reflected in the capillary pressure and
observable in the measurements with a lower entry pressure as the interfacial tension
decreases, Figure 3a. Each Pc curve was scaled by its respective interfacial tension to
allow for seeing if there is a weakening of the capillarity of the system through a change
in wettability, Figure 3b. The scaled data set collapse nicely suggesting no change in the
wettability with regard to pressure.

SCA2015-018

7/12

Analogue Fluids
A core-flood was performed using N2 and deionized water at lab temperature, 25oC, and
10 MPa pore pressure for the purpose of comparison with a different fluids pair. The N2water system experiment is quicker and much easier than CO2-brine experiments. Carbon
dioxide-brine system is corrosive fluids pair and requires the entire wetted parts in the
experimental apparatus to be anti-corrosion such as Hasteloy and Titanium. The
solubility of CO2 and N2 at 25oC and 10MPa are 31.75cm3/g (cm3 of gas at standard
conditions per gram of water) [23] and 1.264cm3/g [24] respectively, i.e., the solubility of
N2 is more than 25 times less than CO2. Therefore, N2 equilibrates with water much faster
compared to CO2. The thermophysical properties of N2 at 25oC and 10MPa were similar
to CO2 at 5MPa and 50oC, as shown in Table 1. Nitrogen and CO2 capillary pressure
curves scaled by their respective IFT are similar with slight shifts with regard to
wettability as discussed in previous sections. Therefore, it is possible that the capillary
pressure for CO2-brine in sandstone to be estimated using N2-water fluids pair
particularly where the experimental apparatus for handling CO2-brine systems is not
readily available.

Figure 3. [10] a. Core-flood capillary pressure measurements of CO2-water system at 50 oC, 0 mol/kg and
pressures ranging from the gas phase (5 MPa) to low density scCO 2 (10 MPa) to high density scCO2 (20
MPa). Nitrogen-water capillary pressure was included for comparison and to show the effect of IFT on Pc
as measured by semi-dynamic technique. The repeats demonstrate the reproducibility of the semi-dynamic
Pc measurements performed in this study. b. The Pc curves were scaled by IFT values representative of the
experimental conditions, Table 1, and show no significant change in wettability with regard to pressure.

SCA2015-018

8/12

CT Images Accuracy as a Function of Fluids Density Ratio

Fluids saturations were measured in-situ by CT scanner using a combination of
experimental and background scans as described in a previous section. This method
depends on the contrast in density between the imaged phases (rock, wetting and nonwetting fluids). The density ratio between the wetting and non-wetting fluids varies
depending on the experimental conditions as shown in Table 1. Thus, CT image quality is
dependent on the density ratio of the experimental fluids and the noise associated with
CT images can affect the corresponding reconstruction of 3-D saturation map. Depending
on the significance of noise, a false fluids saturation distribution can be measured without
proper coarsening. Figure 4 shows the saturation distribution of experiments 1 to 4 where
CO2 phase transfers from the gas phase to scCO2 and finally to high density scCO2, the

range from 9.5 to 1.3. Experiment 4 was made at the same condition as 3 (50oC,
2
20 MPa) with the exception of doping the water with 5 wt% Nal to enhance the contrast
between the fluids for the CT image.

Figure 4. a. 2D saturation map of the inlet slice for 4 different experiements examining the impact of
pressure on multiphase flow properties. The 2D slices have similar average saturations. b.
Saturation
distributions of the 2D inlet slices are directly proportional to experimental pressure and as CO 2 becomes
denser. This is not representing insitue saturations but rather noise in CT measurements due to low
water/CO2 density ratio. This was confirmed by observing the saturation distribution at the high density
CO2 experiment after dopping water with Nal.

This test proves that the actual saturation distribution in the core sample remains
unchanged and that the difference in the saturation distribution between experiments 1 to
3 is mainly due to the noise associated with lower density contrast. This noise has random
nature and is irrelevant to the experimental conditions, as shown in Figure 5a. By
subtracting two CT images at same conditions and location, the noise at voxel by voxel
level was obtained. Usually, the noise has normal distribution and no relationship with
fluid density. The uncertainty on calculated saturation induced by random noise for CO 2brine system can be calculated by [21].

SCA2015-018

9/12

1 + (

+2
2

(2)

where is the uncertainty on saturation, and is the standard deviation of random

noise. The average uncertainty at various experiment ranges from 0.115 to 0.379, which
is large enough to influence the comparison of saturation maps at different conditions.
Therefore, different coarsening schemes have been applied to images from 1x1 to 15x15,
as shown in Figure 5b. With increasing degree of coarsening, the level noise can be
reduced. Therefore, the images at different experimental conditions have been processed
with proper coarsening scheme before the reconstruction of saturation maps. By using the
above approach, the range of the average uncertainty on saturation at various experiments
has been reduced to 0.115~0.129, which is acceptable for the comparison between
saturation maps.

Figure 5. a. A histogram of CT numbers obtained by taking the difference of two scans of the same
location. Various shades of grey represent difference values obtained at various conditions of pressure and
temperature but they are indistinguishable indicating that the nature of the noise was intrinsic to the x-ray
scanner. From light to dark the histogram bars are taken from: CT images of CO 2 saturated core at 5MPa50oC, 10MPa-25oC, and 20MPa-50oC respectively, Grey line: fitted normal distribution curve with mean
value = 0.04 and standard deviation = 19.9.
b. Random noise vs. coarsening schemes.

Quantify the impact of density ratio on error from image noise is crucial for studies
targeting voxel resolution saturations from CT x-ray scanning. In this study, coarsening
the images is not necessary for obtaining accurate saturation values as the slice average
saturation remains unaffected. This analysis was made to inspect dissolution or gravity
segregation and assure that the experiments were performed as designed by visualizing
comparable images with less noise. This was also useful to confirm that the saturation
distribution remains unaffected with different experimental conditions.

CONCLUSION

SCA2015-018

10/12

Wettability determines the efficiency of enhanced oil recovery operations as well as our
ability to inject and store CO2 in geological formations. This study investigated the
effective core-scale wetting properties at a wide range of reservoir conditions to observe
the impact of pressure, temperature and brine salinity on the wetting properties of the
CO2-brine system in siliciclastic rocks during primary drainage. Core-flood semidynamic capillary pressure were measured to observe the impact of different reservoir
conditions on effective wettability from observations of multiphase flow properties. This
allowed for measuring the core-scale effective and dynamic wetting properties
representative of CO2 injection processes such as EOR and carbon storage. No significant
change in capillarity observed with regard to pressure even with different CO2 phases
(gaseous and supercritical). The saturation distribution of the fluids remained similar
within the core sample with varying water/CO2 density ratios. This was confirmed by a
repeat of high density supercritical CO2 and doped brine measurement. The error
associated with noise from density ratio has to be considered in studies looking into voxel
size saturations from CT imaging or spatial saturations distribution and up scaling
resolution is advised. Analogue fluids can be used for characterising multiphase flow
properties when reservoir conditions cannot be replicated. In general, reservoir conditions
are not having a major impact on the wettability of CO2-brine-sandstone system during
CO2 injection process and the system remains strongly water-wet during primary
drainage for the experimental conditions investigated in this study.

ACKNOWLEDGEMENT
We gratefully acknowledge funding from the Qatar Carbonates and Carbon Storage
Research Centre (QCCSRC), provided jointly by Qatar Petroleum, Shell, and Qatar
Science & Technology Park.

REFERENCES
1. Espinoza, D., and J. Santamarina. "Water-CO2-Mineral Systems: Interfacial Tension,
Contact Angle, and DiffusionImplications to CO2 geological Storage." Water
Resources Research 46, no. 7 (2010).
2. Chiquet, P., D. Broseta, and S. Thibeau. "Wettability Alteration of Caprock Minerals
by Carbon Dioxide." Geofluids 7, no. 2 (2007): 112-22.
3. Jung, J., and J. Wan. "Supercritical CO2 and Ionic Strength Effects on Wettability of
Silica Surfaces: Equilibrium Contact Angle Measurements." Energy & Fuels 26, no.
9 (2012): 6053-59.
4. Wang, S., I. Edwards, and A. Clarens. "Wettability Phenomena at the CO2Brine
Mineral Interface: Implications for Geologic Carbon Sequestration." Environmental
Science & Technology (2012).
5. Kim, Y., J. Wan, T. Kneafsey, and T. Tokunaga. "Dewetting of Silica Surfaces Upon
Reactions with Supercritical CO2 and Brine: Pore-Scale Studies in Micromodels."
Environmental science & technology 46, no. 7 (2012): 4228-35.

SCA2015-018

11/12

6. Krevor, S., R. Pini, L. Zuo, and S. Benson. "Relative Permeability and Trapping of
CO2 and Water in Sandstone Rocks at Reservoir Conditions." Water Resources
Research 48, no. 2 W02532 (2012).
7. Golding, M., J. Neufeld, M. Hesse, and H. Huppert. "Two-phase gravity currents in
porous media." Journal of Fluid Mechanics 678:248-270 (2011).
8. Golding, M., H. Huppert, and J. Neufeld. "The effects of capillary forces on the
axisymmetric propagation of two-phase, constant-flux gravity currents in porous
media." Physics of Fluids 1994-present, 25 (3):036602 (2013).
9. Niu, B., A. Al-Menhali, and S. Krevor. "The Impact of Reservoir Conditions on the
Residual Trapping of Carbon Dioxide in Berea Sandstone." Water Resources
Research (2015).
10. Al-Menhali, A., B. Niu, S. Krevor. Capillarity and Wetting of Carbon Dioxide and
Brine During Drainage in Berea Sandstone at Reservoir Conditions [in preperation]
Water Resources Research (2015).
11. Al-Menhali, A., and S. Krevor. "Pressure, Temperature and Ionic Strength Effects on
the Wettability of CO2-Brine-Sandstone System: Core-Scale Contact Angle
Measurements." Society of Core Analysts Symposium SCA2013-003, no. Paper
presented at the 2013 SCA in Napa Valley, California (2013).
12. Li, X., E. Boek, G. Maitland, and J. Trusler. "Interfacial Tension of (Brines + CO2):
(0.864 Nacl + 0.136 KCl) at Temperatures between (298 and 448) K, Pressures
between (2 and 50) Mpa, and Total Molalities of (1 to 5) MolKg1." Journal of
Chemical & Engineering Data 57, no. 4 (2012).
13. Yan, W., G. Zhao, G. Chen, and T. Guo. "Interfacial Tension of (Methane +
Nitrogen) + Water and (Carbon Dioxide + Nitrogen) + Water Systems." Journal of
Chemical & Engineering Data 46, no. 6 (2001): 1544-48.
14. Lemmon, E., M. McLinden, and D. Friend. "Thermophysical Properties of Fluid
Systems in Nist Chemistry Webbook; Nist Standard Reference Database No. 69,
Eds. P. Linstrom and W. Mallard, National Institute of Standards and Technology,
Gaithersburg Md, 20899, Http://Webbook.Nist.Gov." (retrieved May 6, 2013).
15. Kestin, J., H. Khalifa, and R. Correia. Tables of the Dynamic and Kinematic
Viscosity of Aqueous Nacl Solutions in the Temperature Range 20-150 C and the
Pressure Range 0.1-35 Mpa. American Chemical Society and the American Institute
of Physics for the National Bureau of Standards, 1981.
16. Al-Menhali, A., C. Reynolds, P. Lai, B. Niu, N. Nicholls, J. Crawshaw, and S.
Krevor. "Advanced Reservoir Characterization for CO2 Storage." International
Petroleum Technology Conference, (2014).
17. Hassanizadeh, S., and W. Gray. "Thermodynamic Basis of Capillary Pressure in
Porous Media." Water Resources Research 29, no. 10 (1993): 3389-405.
18. Lenormand, R., A. Eisenzimmer, and C. Zarcone. "A Novel Method for the
Determination of Water/Oil Capillary Pressures of Mixed-Wettability Samples."
Paper presented at the SCA, (1993).
19. Ramakrishnan, T., and A. Cappiello. "A New Technique to Measure Static and
Dynamic Properties of a Partially Saturated Porous Medium." Chemical Engineering
Science 46, no. 4 (1991): 1157-63.

SCA2015-018

12/12

20. Pini, R., and S. Benson. "Simultaneous Determination of Capillary Pressure and
Relative Permeability Curves from Core-Flooding Experiments with Various Fluid
Pairs." Water Resources Research (2013): 49, doi:10.1002/wrcr.20274.
21. Pini, R., S. Krevor, and S. Benson. "Capillary Pressure and Heterogeneity for the
CO2/Water System in Sandstone Rocks at Reservoir Conditions." Advances in Water
Resources 38 (2012): 48-59.
22. Brooks, R., and A. Corey. "Hydraulic Properties of Porous Media, Hydrology
Papers, No. 3, Colorado State University, Ft." Collins, Colo (1964).
23. Wiebe, R. "The Binary System Carbon Dioxide-Water under Pressure." Chemical
Reviews 29, no. 3 (1941): 475-81.
24. Wiebe, R., V. Gaddy, and C. Heins. "The Solubility of Nitrogen in Water at 50, 75
and 100 from 25 to 1000 Atmospheres." Journal of the American Chemical Society
55, no. 3 (1933): 947-53.

SCA2015-019

1/12

THE IMPACT OF RESERVOIR CONDITIONS ON

MULTIPHASE FLOW IN THE CO2-BRINE SYSTEM IN
PERMEABLE SANDSTONE
Samuel Krevor, Catriona Reynolds, Ali Al-Menhali, Ben Niu
Department of Earth Science & Engineering
Qatar Carbonates and Carbon Storage Research Centre
Imperial College London
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Capillary strength and multiphase flow are key parameter inputs for modeling CO2
injection processes for enhanced oil recovery and CO2 storage. Experimental
observations of the multiphase flow properties in this system reported over the past 10
years have raised important questions about the impact of reservoir conditions on flow
through effects on wettability, interfacial tension and fluid-fluid mass transfer. In this
work we report the results of a three year investigation aimed at resolving many of these
outstanding questions for flow in sandstone rocks. The capillary pressure, relative
permeability and residual trapping characteristic curves have been characterized in
Bentheimer and Berea sandstone rocks across a pressure range 5 20 MPa, temperatures
25 90 C and brine salinities 0-5M NaCl. In total over 30 reservoir condition core flood
tests were performed evaluating these properties with techniques including the steady
state relative permeability test, the semi-dynamic capillary pressure test and a novel test
for the rapid construction of the residual trapping initial-residual (IR) curve. Test
conditions were designed to isolate effects of interfacial tension, viscosity ratio, density
ratio and salinity. The results of the tests show unequivocally that reservoir conditions
have minimal impact on relative permeability and residual trapping, consistent with
continuum scale multiphase flow theory for water wet systems. The invariance of the
characteristic curves is observed across the range of conditions, including transitions of
the CO2 from a gas to a liquid to a supercritical fluid, and in comparison with N2-brine
systems. Variations in capillary pressure curves are generally well explained by
corresponding changes in IFT although some further variation may reflect small changes
in wetting properties that have been observed in sessile drop experiments. As with gasbrine systems, the low viscosity of CO2 at certain conditions results in particular
sensitivity to rock heterogeneity. We show that (1) heterogeneity is the likely source of
much of the uncertainty around relative permeability observations for this system and (2)
that appropriate scaling of the driving force for flow by a quantification of capillary
heterogeneity allows for the selection of core flood parameters that eliminate this effect.
Similarly this scaling can be used to approximate the effect of small scale heterogeneity
on flow for real reservoir systems.

SCA2015-019

2/12

INTRODUCTION
Fluid flow during CO2-injection processes, for enhanced oil recovery and CO2 storage, is
governed by the multiphase flow properties - capillary pressure, relative permeability,
hysteresis and residual trapping. A sketch of fluid flow during buoyant CO2 migration in
a reservoir is shown in Figure 1. The leading edge and upper portions of a plume
maintain high capillary pressures and CO2 saturation is high in these regions leading to
high values of relative permeability. In the lower and distal sections of the plume a
capillary fringe will appear as saturation tapers to the residual. The disconnection of CO2
ganglia results in hysteresis in the multiphase flow functions.
CO2
permeability

- Drainage
- - Imbibition

102 m
C

Water saturation
Imp

oc
le r
eab
er m

CO2 saturation

Maximum

Fringe
C

Injection
well

Residual

Rock
CO2
Zero

Brine
10-5 m

Figure 1. A sketch of key flow processes during CO 2 storage. During drainage, at the leading edge of the
plume, CO2 displaces resident brine to near the connate water saturation and relative permeability is high.
Where the CO2 plume is receding, brine imbibes and results in a disconnected fluid saturation. Ultimately,
the saturation is lowered to residual immobile CO2 which remains trapped. Figure from [25]

These relationships between the capillarity of a geologic system, and the saturation and
permeability of the fluids have a direct bearing on the migration speed and reservoir
sweep of a CO2 plume (Figure 2, [1]), immobilisation through capillary trapping [2] and
are used directly in assessments of the capacity of a storage site to contain CO2 over
geologic timescales [3].

SCA2015-019

3/12

Figure 2. The impact of capillarity on multiphase flow in systems with strong capillarity, i.e. strongly
water wet systems, there is a larger plume sweep and lower forward propagation of CO2 compared to
systems with weaker capillarity, i.e. less strongly water wet systems. Figure adapted from [1].

Previous and ongoing experience with industrial scale carbon dioxide injection has come
mostly from operations of enhanced oil recovery in the United States although there are
pilot CO2 storage projects around the world [4].
Despite the long history of the use of CO2 for enhanced oil production there are
longstanding uncertainties about the multiphase flow properties underlying descriptions
and predictions of CO2 migration and trapping in the subsurface. Major outstanding
uncertainties include the impact of reservoir conditions of pressure, temperature and brine
salinity on the wetting state and the relative permeability characteristic curve.
A large number of studies using sessile drop observations to characterise wetting have
observed effects of pressure, temperature and brine salinity on wetting in this system with
silicate and carbonate minerals, but the direction of the effects (increasing or decreasing
wetting strength) and the magnitude have at times been contradictory [5-10]. A review of
this issue has recently been published by [11].
Similarly, there are a large number of relative permeability curves reported in the
literature (Figure 3, [12]). Based on these observations studies have variously described
the CO2-brine system as both less [13] and equivalently [14, 15] water wetting than
hydrocarbon systems, to have relative permeabilities sensitive to variation in reservoir
conditions [16], and for the system to have unusually low endpoint CO2 relative
permeabilities [17].
The work of [14] stands alone as a comprehensive evaluation of the sensitivity of the
multiphase flow properties of the CO2-brine system to reservoir conditions. In this
approach, capillary pressure and relative permeability curves of rocks were characterised

SCA2015-019

4/12

with mercury-air and nitrogen-water observations respectively. These properties were

shown to lead to accurate modelled predictions of observations of unsteady state
drainage, CO2 displacing brine, at various temperatures and pressures. This lead to the
conclusion that the CO2-brine-rock system was water wet and the multiphase flow
characteristics insensitive to reservoir conditions.
P

11

30

Pressure [MPa]

20
15
10
5
0

0.6

0.4

0.2

280

300

320

340

Temperature [K]

360

380

400

All published CO2/brine

drainage relative
permeability curves

0.8

Relative permeability [-]

25

P,T conditions of
experiments reported
in the literature

0.2

0.4
0.6
Brine saturation [-]

0.8

Figure 3. Compilation of drainage relative permeability curves published for CO 2-brine systems. From
[12].

Characterisation of the capillary pressure characteristic curve can provide useful insight
into the wetting state of a system and its impact on multiphase flow. Figure 4 shows
sketches of capillary pressure primary drainage and secondary imbibition curves for
water and mixed-wet CO2-brine systems. In the mixed wet system during drainage there is
a lower capillary entry pressure, a lower range of pressures required for full brine
desaturation compared with the water wet system. During imbibition, there is more
trapping in the water wet system and the residual is approached at capillary pressures
close to zero. In the mixed-wet system, significant desaturation occurs at negative
capillary pressures.

Figure 4. Primary drainage and secondary imbibition capillary pressure characteristic curves for water wet
(left) and mixed-wet systems (right).

SCA2015-019

5/12

The residual trapping characteristic of a rock can thus be used partly as a diagnostic for
the wetting state of the system. As described above, it is also a key reservoir
characteristic, with primary controls on plume migration and storage capacity. The
characteristic curve described by the initial residual relationship shown in Figure 5 is
also often used to parameterise hysteresis models for capillary pressure and relative
permeability, i.e. the Land model shown in Figure 5 [18].

Figure 5. Initial-residual characteristic curves are shown on the left, constructed from the Land model with
various values of the constant C [18]. On the right are the corresponding relative permeability curves with
hysteresis tracking curves constructed with the Land model.

In this study we have sought to address outstanding questions around the response of
multiphase flow properties in the CO2-brine-sandstone rock system to reservoir
conditions. We have evaluated the drainage capillary pressure, steady state relative
permeability, and initial-residual curves characterising residual trapping at a range of
reservoir conditions. These curves were compared with observations of the characteristic
curves using the N2-water system, known to be water-wet.
MATERIALS AND METHODS
A homogenous Berea sandstone rock core was used for the studies on capillary pressure
and residual trapping. The homogeneity of the rock was most evident in the highly
uniform saturation distribution of CO2 and brine during multiphase flow displacement.
The core was 3.8cm in diameter and 20cm in length and had a porosity and permeability
of 0.19 and 210 mD respectively. The rock core was heated in an oven at 700 C to
stabilise mobile clays. A Bentheimer sandstone sample was used for the observations of
relative permeability. The rock core was 3.8 cm diameter, 23 cm in length and had a
porosity and permeability of 0.22 and 1.8 D respectively.

SCA2015-019

6/12

The ranges of pressures, temperatures and brine salinities for all of the tests are
summarised in Table 1.
Table 1. Conditions and rock samples used in the multiphase flow experiments.

Observation

Pressures
[MPa]

Rock core

Technique

5-20

Temperature Salinity
[oC]
[M
NaCl]
25, 50
0,1,3,5

Capillary
pressure
(drainage)
Relative
permeability
(drainage)
Residual
trapping

Berea
sandstone

Semidynamic Pc

10 21

30 - 90

0,1,3,5

Bentheimer
sandstone

Steady state

5- 20

25, 50

0,1,3,5

Berea
sandstone

[19]

We used a variation of the semi-dynamic capillary pressure method [20-22] to measure

the capillary pressure characteristic curves. A detailed description of the procedure is
included in another paper in this symposium [23].
A steady state technique was used to characterise relative permeability. The saturation in
the core sample was controlled by the ratio of injected phases, i.e. the fractional flow. Insitu saturation was measured along the entire sample with x-ray CT for each fractional
flow once a constant pressure drop (i.e. steady state) was achieved. Evaluating the impact
of reservoir conditions on multiphase flow in the CO2-brine system is not a trivial task
given that multiple fluid properties (density, viscosity) are changing over the same
conditions of pressure, temperature and brine salinity. Experiments were designed to
make observations of the relative permeability by varying interfacial tension while
holding one or more of these parameters constant.
To construct the initial-residual capillary trapping characteristic curve, a core flood test
was used to exploit the presence of capillary end effects to rapidly observe the
relationship across a wide range of saturations at reservoir conditions [19]. In a given test,
a range of initial CO2 saturations along the length of the core were created by performing
drainage at a flow rate that maximised the capillary end effect. Subsequently, imbibition
was performed at a flow rate representative of reservoir conditions until the CO2
saturation reduced to the residual. A range of residual saturations observed along the
length of the core corresponded to the range in initial saturation existing prior to
imbibition. As a result, a single core flood could be used to parameterise the initialresidual curve along a range of saturations.
RESULTS
One group of capillary pressure characteristic curves for tests evaluating the impact of
pressure on capillarity are shown in Figure 6. In the right graph of the figure, the capillary

SCA2015-019

7/12

pressures are scaled by the respective interfacial tensions for each condition. Variation in
the capillary pressure characteristic curves are accounted for by the variation in
interfacial tension for each condition suggesting that any variation in wettability is not
making an impact on capillarity for this system. This was generally the case for all of the
conditions tested, although there was some variation in the scaled curves. The
consistency with the N2-water system indicates that the CO2-brine system is water wet
during drainage. Further results are detailed in another paper in this symposium, [23].

Figure 6. Capillary pressure characteristic curves evaluating the impact of pressure variation on capillarity
in the system. Curves measured at pressures ranging from 5 MPa to 20 MPa are shown on the left and
scaled by interfacial tension on the right. [23].

In the relative permeability tests the relative permeability of the rock core obtained when
fluid saturations were unaffected by end effects, gravity segregation, and rock
heterogeneity, was found to be invariant with reservoir conditions and equivalent to the
N2-water relative permeability. This will be referred to as the intrinsic relative
permeability.
One group of results is shown in Figure 7. Two of the tests were performed with CO2 and
brine at reservoir conditions of pressure, temperature and brine salinity, while a third test
was performed using a nitrogen-brine system at ambient conditions. The range of
interfacial tension spanned from 37 mN m-1 to 62 mN m-1. The results show little
variation in the relative permeability between the tests.

SCA2015-019

8/12

0.4

CO 37 mN/m
2

CO2 41 mN/m

0.3

N2 65 mN/m

0.2

0.1

0
0.2

0.4

0.6

0.8

Figure 7. The intrinsic relative permeability of the rock core obtained from CO 2 and N2 systems.

At conditions that resulted in CO2 viscosities less than 30 Pa s, the fluid distributed
heterogeneously throughout the core due to natural heterogeneity in the rock. Under these
conditions, the derived relative permeability will be referred to as the effective relative
permeability. We use this as the observational counterpart of a pseudo relative
permeability. See [24] for a study in which pseudos have been applied to investigate core
scale rock heterogeneity.
The effective relative permeability was observed to be sensitive to the test conditions,
pressure, temperature, brine salinity and fluid flow rate. This is consistent with past
observations and modelling of the impact of rock heterogeneity on observed flow
properties [24]. Figure 8 shows one example of the intrinsic curve measured compared
with an effective curve in which the interfacial tension was constant, but the viscosity of
the CO2 was less than 30 Pa s in the measurement affected by rock heterogeneity.

Figure 8. An intrinsic relative permeability curve (grey) and an effective relative permeability curve (white)
measured at the same interfacial tension but distinct values of pressure and temperature. The difference in
CO2 viscosity led to homogenous and heterogeneous fluid distributions in the rock core.

The relative permeability core floods were all performed at a constant total flow rate of
20 ml min-1, this being sufficient to minimise capillary end effects. Thus the total viscous
pressure differential across the core, Pinlet Poutlet, varied primarily depending on

SCA2015-019

9/12

the viscosity of CO2 for a given test. When the viscous force was high enough to support
the capillary pressure gradients prevalent at constant saturation (Figure 9), a homogenous
saturation was obtained and the intrinsic relative permeability was achieved. When the
viscous force was low compared with the capillary heterogeneity in the core, capillary
driven flow redistributed fluid towards a system with constant Pc, but heterogeneous
saturation distribution and an effective relative permeability was observed.

Figure 9. Conceptualising heterogeneity in the rock core with distinct capillary pressure characteristic
curves. Low viscous force leads to a system of constant capillary pressure, but heterogeneous saturation
distribution. A high viscous force leads to a system of homogenous saturation but capillary pressure
gradients.

This study confirms the importance of evaluating the capillary heterogeneity in reservoir
rock cores. We recommend in particular a characterisation of spatial heterogeneity in
capillary pressure characteristic curves, e.g. [22]. A comparison of this with the estimated
driving force for flow e.g., pressure gradients, buoyancy - in a reservoir allows for one
to design laboratory experiments either so that (1) an intrinsic relative permeability curve
may be obtained or (2) effective flow functions representative of reservoir conditions can
be observed. If capillary heterogeneity is significant, this provides quantitative guidance
as to the conditions at which core tests should be operated to obtain representative
effective relative permeabilities.

SCA2015-019

10/12

Figure 10. In-situ saturation profiles of CO2 saturation after drainage (Initial saturation), after 0.1 pore
volume of brine injection (PVI) and greater than 0.3 PVI showing the residually trapped saturation. From
[19].

Observations of the saturation profile in a residual trapping experiment after drainage and
during imbibition is shown in Figure 10. A compilation of the initial-residual data from
all of the floods reported in [19] is shown in in Figure 11, along with the data obtained
with an N2-water system. There was little, if any, impact of reservoir conditions on the
residual trapping characteristics of this system. The N2-water curve was also
indistinguishable from the CO2-brine curves suggesting that this system was water wet
during imbibition.

Figure 11. Initial-residual data from CO2-brine core floods at all of the conditions are shown in black. Light
colored symbols show trapping data obtained from a N2-water system. From [19].

SCA2015-019

11/12

CONCLUSION
The body of observations demonstrate definitively that (1) the CO2-brine system is water
wet, and (2) the intrinsic multiphase flow properties of this system are invariant across
the range of pressure, temperature and brine salinity in subsurface reservoirs worldwide.
These two conclusions are consistent with the findings of [14] where a comprehensive
evaluation of CO2-brine multiphase flow was evaluated in carbonate rocks. It was also
observed that (3) relative permeability in the CO2-brine system may be sensitive to the
presence of rock heterogeneity. In this case effective relative permeabilities should be
obtained with appropriate scaling to reservoir conditions by comparing capillary
heterogeneity to the viscous flow force representative of the reservoir system.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge funding support for this work from the Qatar Carbonates and
Carbon Storage Research Centre provided jointly by Shell, Qatar Petroleum and the Qatar
Science and Technology Park.

REFERENCES
1. Golding, M.J., H.E. Huppert, J.A. Neufeld (2013) The effects of capillary forces on the
axisymmetric propagation of two-phase, constant-flux gravity currents in porous media.
Physics of Fluids. 25, 3, 036602
2. Hesse, M.A., F.M. Orr Jr., H.A. Tchelepi (2008) Gravity currents with residual trapping.
Journal of Fluid Mechanics 611, 35-60
3. Szulczewski, M.L., C.W. MacMinn, H.J. Herzog, R. Juanes (2012) Lifetime of carbon
capture and storage as a climate-change mitigation technology. Proceedings of the
National Academy of Sciences of the United States of America. 109, 14, 5185-5189
4. Global CCS Institute (2014) The Global Status of CCS: 2014
5. Chiquet, P., D. Broseta, S. Thibeau (2007) Wettability alteration of caprock minerals by
carbon dioxide. Geofluids. 7, 112-122
6. Espinoza, D.N. and J.C. Santamarina (2010) Water-CO2-mineral systems: Interfacial
tension, contact angle, and diffusion Implications to CO2 geological storage. Water
Resources Research 46, 7, W07537
7. Jung, J-W. and J. Wan (2012) Supercritical CO2 and Ionic Strenght Effects on Wettability
of Silica Surfaces: Equilibrium Contact Angle Measurements. Energy & Fuels. 26, 9,
6053-6059
8. Wang, S., I.M. Edwards, A.F. Clarens (2013) Wettability Phenomena at the CO2-brinemineral interface: Implications for Geologic Carbon Sequestration. Environmental
Science & Technology. 47, 1, 234-241
9. Kim, Y., J. Wan, T. J. Kneafsey, T. K. Tokunaga (2012) Dewetting of Silica Surfaces
upon Reactions with Supercritical CO2 and Brine: Pore-Scale Studies in Micromodels.
Environmental Science & Technology. 46, 7, 4228-4235
10. Broseta, D., N. Tonnet, V. Shah (2012) Are rocks still water-wet in the presence of dense
CO2 or H2S? Geofluids. 12, 4, 280-294
11. Iglauer, S., C. Pentland, A. Bush (2015) CO2 wettability of seal and reservoir rocks and
the implications for carbon geo-sequestration. Water Resources Research. 51, 1, 729-774
12. Benson, S., Pini, R., Reynolds, C., and Krevor, S. (2013) Relative permeability analyses
to describe multiphase flow in CO2 storage reservoirs. Global CCS Institute

SCA2015-019

12/12

13. Berg, S., Oedai, H., Ott, H. (2011) Displacement and mass transfer between saturated and
unsaturated CO2-brine systems in sandstone. International Journal of Greenhouse Gas
Control. 12, 478-492.
14. Egermann, P., Chalbaud, C., Duquerroix, J.-P., Gallo, Y.L. (2006) An Integrated
Approach to Parameterize Reservoir Models for CO2 Injection in Aquifers. SPE Annual
Technical Conference and Exhibition, San Antonio, Texas, 24-27 September 2006. SPE
102308
15. Krevor, S.C.M., R. Pini, L. Zuo, S.M. Benson (2012) Relative Permeability and Trapping
of CO2 and Water in Sandstone Rocks at Reservoir Conditions. Water Resources
Research. 48, 2, W02532
16. Bachu, S. and B. Bennion (2008) Effects of in-situ conditions on relative permeability
characteristics of CO2-brine systems. Environmental Geology. 54, 1707-1722
17. Levine, J., Goldberg, D.S., Lackner, K.S., Matter, J.M., Supp, M.G., Ramakrishnan, T.S.
(2014) Relative permeability Experiments of Carbon Dioxide Displacing Brine and Their
Implications for Carbon Sequestration. Environmental Science & Technology. 48, 1, 811818.
18. Land, C. (1968) Calculation of Imbibition Relative Permeability for Two- and ThreePhase Flow From Rock Properties. SPE Journal. 8, 02, 149-156
19. Niu, B., Al-Menhali, A., Krevor, S.C. (2015) The impact of reservoir conditions on the
residual trapping of carbon dioxide in Berea sandstone. Water Resources Research. In
Press. doi: 10.1002/2014WR016441
20. Ramakrishnan, T.S., Cappiello, A. (1991) A New Technique to Measure Static and
Dynamic Properties of a Partially Saturated Porous Medium. Chemical Engineering
Science. 46, 4, 1157-1163
21. Lenormand, R., A. Eisenzimmer, C. Zarcone (1993) A Novel Method for the
Determination of Water/Oil Capillary Pressures of Mixed-Wettability. Society of Core
Analysts Conference Paper. SCA 1993-9322
22. Pini, R., S.C.M. Krevor, S.M. Benson (2012) Capillary pressure and heterogeneity for the
CO2/water system in sandstone rocks at reservoir conditions. Advances in Water
Resources. 38, 48-59
23. Al-Menhali, A., Niu, B., Krevor, S. (2015) The Impact of Reservoir Conditions on
Wetting and Multiphase Flow Properties Measurements for CO2-Brine-Rock System
During Primary Drainage. International Symposium of the Society of Core Analysts,
Newfoundland and Labrador, Canada, 16-20 August, 2015. SCA-A058
24. Yokoyama, Y., Lake, L.W. (1981) The Effects of Capillary Pressure on Immiscible
Displacements in Stratified Porous Media. 56th Annual Fall Technical Conference and
Exhibition of the Society of Petroleum Engineers of AIME, San Antonio, Texas, Oct. 5-7,
1981. SPE 10109
25. Krevor, S, Blunt, M.J., Benson, S.M., Pentland, C.H., Reynolds, C, Al-Menhali, A., Niu,
B (2015) Capillary trapping for geologic carbon dioxide storage From pore scale
physics to field scale implications. International Journal of Greenhouse Gas Control. In
Press

SCA2015-020

IMPROVED CORE ANALYSIS MEASUREMENTS IN LOW

PERMEABILITY TIGHT GAS FORMATIONS
S. Kryuchkov1,2, J. Bryan1,2, L.Yu1, D. Burns3 and A. Kantzas1,2
1
PERM Inc., Calgary, Canada; 2University of Calgary, Canada
3
TAQA North Ltd, Calgary, Canada
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Recovery of hydrocarbon gas in micro-Darcy rock is now becoming more common-place
for gas producers. As part of the reservoir characterization associated with these systems,
it is required to understand parameters such as the total porosity and permeability of the
system, but also the initial water saturation and the effective gas permeability at this
connate water saturation. This is still under the realm of what is termed routine core
analysis, but in these tight formations it is imperative to understand that these tests are
actually very non-routine by nature. Current work presents a case study of core analysis
that was run on plugs from a tight gas producing formation, and illustrates some of the
pitfalls and corrections associated with measurements of this low permeability rock.
The case study combines low field NMR measurements with core flooding measurements
of porosity and permeability. Cores are initially tested at their connate water saturations:
NMR measures the initial water present in each core, and effective porosity and
permeability to gas are measured through different core analysis techniques. By
considering the time scales required for pressure to propagate within the cores, it is
shown that routine measurement of permeability through multi-rate Darcy flow of gas
is simply not adequate to properly characterize the core. Instead, core properties are
measured through analysis of transient pressure data. Cores are then run through a DeanStark cleaning process, and NMR measurements are used to verify the ability of DeanStark to remove all water out of the cores. In this manner, post-cleaning measurements
of porosity and permeability can be understood as effective values rather than their
assumed absolute properties. This case study illustrates the challenges in making
accurate measurements in micro-Darcy rock systems, and helps end-users to understand
the dangers in using routine data without knowing how the data was collected.

INTRODUCTION
The objective of this test was to measure effective porosity and permeability in tight
cores containing some residual water content. These values were then to be compared to
measured porosity and permeability after attempting to clean water out of the cores. In
order to assess the objective of this project, six core plugs from the Ferrier field operated
by TAQA North Ltd. were cut from as received (pseudo-native-state) core and were sent

1/12

SCA2015-020

to PERM Inc. The field is a dry gas field with water present. The mineralogy is primarily
carbonate.
The following work plan was defined at the start of the project:
Measure dimensions of cores.
Measure NMR spectra of initial cores with parameters of 0.16ms TE, 5000
Echoes, 12,000ms waiting time, and 128 trains.
Measure pore volume and porosity with gas expansion and effective permeability
to gas under 14.9MPag (2160psig) overburden pressure and near ambient pore
pressure for the cores.
Dean-Stark the cores, making sure to spend sufficient time to try to remove all
water from the samples.
Measure NMR spectra of cores after Dean Stark extraction with parameters of
0.16ms TE, 5000 Echoes, 2,000ms waiting time, and 128 trains.
Measure pore volume and porosity with gas expansion and effective permeability
to gas under 14.9MPag (2160psig) overburden pressure and near ambient pore
pressure for the cores.
Dual-energy CT scanning of the cores.

PROCEDURE
1) In order to execute the work plan, the following steps/modifications were carried out:
Cut both ends of as received cores and obtain cylindrical plugs. The cylindrical
plugs were thereafter named as initial cores.
2) Re-measure dimensions and mass of initial cores.
3) Make up brine that is 125,000ppm (12.5%) of NaCl, and measure its density at 25C.
This is the expected salinity of the connate water in the cores.
4) NMR measurement of initial cores, as well a known amount of water to establish the
amplitude index of the brine.
5) Place the cores under 14.9MPag (2160psig) overburden pressure and measure
effective permeability to gas and porosity with gas expansion.
6) Again weigh the cores and measure NMR spectra of the cores after permeability and
porosity measurement.
7) Dean-Stark the cores with Toluene for two weeks and then Acetone / Methanol for
one week. Dry the cores in a conventional oven for 2 days at 105C.
8) Take the mass of cores after extraction, and measure the NMR spectra of the cores.
9) Record the saturation values (Sw) before and after Dean-Stark extraction for all plugs.
10) Put the cores back to core holder under 14.9MPag (2160psig) overburden pressure
and measure permeability to gas and porosity with gas expansion. Based on the postcleaning NMR spectra, this step is either an absolute permeability or an effective
permeability at lower water saturations.
Please note that the Dean Stark extraction had to be extended much longer than the
standard method of 48 hours per solvent to ensure maximum contact of the solvents with

2/12

SCA2015-020

resident fluids. When comparing the results with routine core analysis measurements,
care must be taken to reconcile the methodology used by the routine core analysis
laboratory.
The porosity and permeability measurements were done following the methodologies of
gas expansion and steady state permeability measurement. However, the results were
analyzed with two different methods. One was the traditional approach. The second
method relied on conducting a history matching of the pressure decay data as observed in
the gas expansion method and by fitting different values of permeability until a match
was found. The results were compared for all the plugs and all conditions.
Equations for the non-stationary gas flow are nonlinear, which calls for numerical
solution for calculation of permeability. The total pressure change from initial state to
final stationary state is used to characterize the porosity of the core.
The slope of the pressure curve as a function of square root of time at the origin depends
on gas parameters, the geometry of the system, and the core porosity and permeability.
This simplifies the formulation of the algorithm for finding permeability from the history
match of the pressure decay data.

EXPERIMENTAL RESULTS AND DISCUSSION

Tested Core Properties
Tabulated values for all measured properties are included as well as sample graphs for
individual plug measurements along with summary graphs that show trends.
Figure 1 shows the setup for porosity and permeability measurements. Tests were run at
ambient temperature. Due to the relatively high permeability range measured in the core
plugs (i.e. micro-Darcies compared to nano-Darcies expected for shale), pressure decay
tests can be run relatively quickly and a more thorough temperature control was not
considered necessary. In these small pore volume systems, the key to being able to
measure pressure changes is to keep all dead volumes small, on the same order of
magnitude as the pore volumes of the core. Traditional gas expansion core analysis
equipment will not register an accurate change in pressure as gas enters the core, since
the dead volumes are so much larger than that of the pores.
Table 1 summarizes the dimensions of each plug. The grain density numbers are
estimated from bulk density and porosity calculations with liquid saturation included.
Table 2 summarizes the porosity and apparent permeability results for the as received
cores. The term effective porosity refers to the porosity measured with gas expansion, i.e.
the pore space that is saturated with gas only. The term total porosity is the sum of the
gas saturated porosity (measured by gas expansion) and the liquid-filled porosity
measured by NMR. Total porosity is then taken to be the sum of the gas expansion
(effective) porosity and the bound water porosity, and fluid saturations are also calculated

3/12

SCA2015-020

on this basis. Without any additional information about the drilling mud used and the
pressure conditions while drilling it is uncertain whether the determined liquid saturation
values represent the true irreducible wetting saturation numbers; in the absence of any
other data, the liquid in the core is assumed to be the saline brine present for this system
(125,000 ppm salt).
Table 3 summarizes the porosity and apparent permeability results for the core after Dean
Stark extraction. Despite the extended cleaning and drying there are still some notable
amounts of liquid left in the core samples. This is due to the presence of small pores that
hold water or solvent through capillary forces. In small pores the vapor pressure starts
dropping because of the high curvature of the concave interfacial surfaces. This slows
down the evaporation rate. In addition, surface area of pores with small radii is
increasingly important, because of its relatively high contribution to the total surface area
and because of the water films, which are hard to evaporate due to the surface forces.
Drying this remaining trapped water would require substantially higher temperatures for
the cores to dry. This cannot be allowed because drying at extreme temperatures would
damage the clays and the pore structure of the core would change. Thus the residual
liquid volume has to be considered when the porosity is calculated. Once again, these
liquid volumes are measured through NMR analysis of the cores; the effective porosity
measured through gas expansion is still lower than the true porosity of the core.
Estimates of total porosity are similar in Tables 2 and 3, indicating that this methodology
is accurate.
The other significance of the remaining water saturation after cleaning is that the
measured permeability values are not absolute permeability, but rather are still effective
gas permeability measurements at some (albeit reduced) liquid saturation. Table 4
presents the apparent vs. Klinkenberg corrected permeability values for all plugs, all
conditions for reference purposes. There is very little difference between apparent and
Klinkenberg corrected permeability for the samples and pressure ranges tested here.
Discussion
Figure 2 shows a typical NMR spectrum for plug SP16A at the different stages of the
experimental process. Figure 3 shows the effective porosity change before and after the
Dean Stark extraction process. Figure 4 shows the total porosity before and after the
extraction process. The fact that the total porosity measurements match is very
encouraging and justifies the use of the NMR technology in compiling supplementary
measurements even for routine purposes in the calculation of total porosity.
Figure 5 presents the permeability data before and after cleaning. An increase in
permeability is seen for the cleaned core. Figure 6 presents rudimentary relative
permeability curves for gas for different plugs lumped together based on their
extrapolated absolute permeability values. It is fascinating that the data are lumped into
three sets that can be fitted by straight lines. Extrapolation of these line to zero

4/12

SCA2015-020

permeability provides predictions for critical gas saturation that appears to be inversely
proportional to permeability. In Figure 7 the permeability data are compared against the
history matching predictions of the pressure transient data. The agreement is fairly good,
although it becomes worse as the 1D value is reached.
Figure 8 and Figure 9 show samples of the permeability measurements for plug SP16.
They are presented as examples in order to illustrate the linearity of the steady state
results. Note that routine core analysis will not be conducted under the steady state
mode. For cores with permeability larger than 1D, the steady state method remains the
most reliable method for permeability measurement. However, the transient approach for
cores with permeability below 1D appears to be quite reliable.
Figure 10 and Figure 11 are examples of the pore volume measurement and the transient
data used in the complex transient permeability predictions. The line in red is the
pressure transient match if the steady state permeability is used and the black line is the
permeability obtained by better history matching with permeability being the adjustable
parameter.
As discussed above, the slopes of the pressure transient curves at short times are directly
related to the permeability of the core and help to accelerate the history match procedure
for permeability. Straight lines on the graph are trend lines based on few initial
experimental and calculated points.
Some mismatch of the fitted curve and the observed pressures is due to the variation of
the sample temperature during this experiment.
The combination of NMR and Dean Stark was very useful in calculating the saturations
in the core. Table 5 summarizes the residual saturation after Dean Stark as calculated
from mass balance and NMR. The agreement is good. It should be noted that there is no
evidence of hydrocarbons in the reservoir (it is considered a dry gas reservoir).
However, the NMR could pick up the presence of residual hydrocarbons. Discriminating
hydrocarbons from water in small pores and pore wedges or physically trapped in clay
structures would require NMR with gradients. This can be done but it was not addressed
in this paper.
Furthermore it should be noted that the reservoir cores were quite heterogeneous. Image
examples of the different plugs are shown in Figure 12. The whole plugs are used for the
testing under net overburden pressure. Thus core heterogeneity is maintained and it is
utilized in the calculations.
PERM has developed numerous algorithms for the determination of special core analysis
information from limited data through pore network modelling and pore level multiphysics programs [1-3]. This Digital Core Analysis approach would be ideal to provide

5/12

SCA2015-020

estimates of capillary pressures, resistivity indices, and relative permeability curves. The
analysis can be extended to the utilization of NMR spectra as input [4]. However, this
was deemed outside the scope of this project and will be pursued as future research.

CONCLUSIONS

Deviations from routine methods (steady state testing) increase when permeability
falls below the 1D range.
Transient methodology becomes more accurate and it can be used to measure
permeability values below 1D.
The combined porosity measurements from gas expansion and NMR converge to the
same total porosity value, irrespective of the saturation conditions of the core.

ACKNOWLEDGEMENTS
Funding of this project from TAQA North, as well as permission to publish this paper are
gratefully appreciated.

REFERENCES
1. Ghomeshi, S., Bashtani, F., Bryan, J., Kryuchkov, S. and Kantzas, A.
Determination of Physical Properties of Tight Porous Media Using Digital Core
Physics/Analysis, accepted to Society of Core Analysis Symposium 2015, St.
John's, NL, Canada, August 16-21, 2015.
2. Ghomeshi, S., Bashtani, F., Taheri, S., Skripkin, E., Kryuchkov, S., Bryan, J. and
Kantzas, Predicting Physical Properties of Porous Media at Sub-Pore Scales
from Natural to Complex Heterogeneous Systems: Application in Tight
Reservoirs, GeoConvention 2015, Calgary, AB, Canada, May 4-8, 2015.
3. Rubin M., Wickens S., Leung S., Akinbobola O., Kent T. and Kantzas A.
Network Modeling and Two-Phase Flow in Tight Reservoirs GeoConvention
2014, Calgary, Canada, 1214 May 2014.
4. Talabi, O. and Blunt, M.J. Pore-Scale Network Simulation of NMR Response in
Two-Phase Flow, J. Pet. Sci. Eng., (2010), 72, 1-9.

6/12

SCA2015-020

7/12

Table 1: Core ID and Dimensions

Sample ID

Depth (m)

Diameter
(cm)

Length
(cm)

Grain Density
(kg/m3)

SP16A

2565.70

3.81

6.05

2788.1

SP18A

2566.74

3.81

7.21

2779.4

SP19A

2567.70

3.81

7.53

2811.5

SP30A

2578.68

3.81

7.17

2750.8

SP33A

2580.30

3.81

7.27

2748.1

SP34A

2581.30

3.81

7.03

2992.1

Table 2: Initial Core Properties

Mass
(g)

SP16A

175.68

5.57

10.62

0.81

SP18A

220.77

1.29

4.68

SP19A

226.34

3.75

SP30A

204.47

SP33A
SP34A

Effective Porosity
(%)

Total Porosity
(%)

kgas (D)
Steady
state
Transient

Sample
ID

Sliquid

Sgas

2.07

0.48

0.52

0.12

0.41

0.73

0.27

7.70

0.09

0.89

0.51

0.49

6.17

10.82

1.60

1.60

0.43

0.57

210.82

4.84

9.18

2.36

2.36

0.47

0.53

229.53

1.63

5.82

0.32

0.42

0.72

0.28

Table 3: Core Properties after Cleaning with Dean-Stark Extraction

Sample
ID

Mass
(g)

SP16A

172.82

10.10

SP18A

218.15

SP19A

Effective Porosity
(%)

Total Porosity
(%)

kgas (D)
Steady state

Transient

Sliquid

Sgas

10.59

6.71

7.56

0.05

0.95

4.17

4.72

1.10

1.70

0.12

0.88

222.99

7.42

7.92

1.03

1.84

0.06

0.94

SP30A

200.75

10.29

10.52

7.71

11.82

0.02

0.98

SP33A

207.41

8.86

9.26

12.51

13.60

0.04

0.96

SP34A

227.02

4.86

5.73

2.88

3.45

0.15

0.85

SCA2015-020

8/12

Table 4: Core Apparent and Klinkenberg Corrected Permeability Data

Before Cleaning (D)
Sample ID

After Cleaning (D)

kg (D)

k (D)

kg (D)

k (D)

SP16A

0.81

0.73

6.71

5.38

SP18A

0.12

0.10

1.10

0.91

SP19A

0.09

0.05

1.03

0.79

SP30A

1.60

1.59

7.71

6.54

SP33A

2.36

2.35

12.51

10.71

SP34A
0.32
0.29
k is Klinkenberg corrected permeability.

2.88

2.52

Table 5: Post Dean-Stark Water Mass Balance

Water Extracted
Sample ID
(cm3)
SP16A
2.27

Residual water in
core (cm3)

NMR Residual water in

core (cm3)

0.32

0.34

SP18A

1.68

0.53

0.45

SP19A

2.70

0.43

0.44

SP30A

3.17

0.17

0.19

SP33A

2.57

0.31

0.33

SP34A

1.62

0.71

0.70

Figure 1: Core Testing Apparatus Rig Setup

SCA2015-020

Figure 2: NMR Spectra for Plug SP16A

Figure 3: Effective Porosity of Initial Core and Core after Cleaning with Dean-Stark

9/12

SCA2015-020

Figure 4: Total Porosity of Initial Core and Core after Cleaning with Dean-Stark

Figure 5: Gas Permeability of Initial Core and Core after Cleaning

10/12

SCA2015-020

11/12

SP18A/SP19A
SP30A/SP33A

0.9

SP16A/SP34A
Linear (SP18A/SP19A)

0.8

Linear (SP30A/SP33A)

Linear (SP16A/SP34A)
Relative permeability to gas

Relative permeability to gas

0.7

0.1
y = -1.8827x + 0.9978
R = 0.9898
y = -1.0833x + 0.9997
R = 0.9999

y = -1.8827x + 0.9978
R = 0.9898

0.5

y = -1.3035x + 0.9767
R = 0.9695

0.4

y = -1.0833x + 0.9997
R = 0.9999
0.3

SP18A/SP19A
y = -1.3035x + 0.9767
R = 0.9695

0.6

SP30A/SP33A

0.2

SP16A/SP34A
Linear (SP18A/SP19A)
Linear (SP30A/SP33A)

0.1

Linear (SP16A/SP34A)

0.01
0.00

0.20

0.40

0.60

0.80

1.00

0.00

0.20

0.40

0.60

0.80

1.00

Liquid Saturation (Fraction)

Liquid Saturation (Fraction)

Figure 6: Gas Relative Permeability with Liquid Saturation for Each Core

100.0

Initial Core
Core AfterDean Stark

Kg, Trannsit (D)

10.0

1.0

1:1 Line

0.1
0.0

0.1

1.0
Kg, Steady State (D)

10.0

100.0

Figure 7: Comparison of Gas Permeability of Steady State with Transient Pressure Method

Figure 8: Gas Permeability Measurement

Sample

Figure 9: Klinkenberg Correction for Gas

Permeability Measurement

SCA2015-020

12/12

Figure 11: Permeability Calculation from

Transient Pressure Data

Figure 10: Porosity Measurement with Gas

Expansion with Transient Data

SP18A

SP19A

SP30A

SP6A

SP34A

SP33A

Figure 12: CT image reconstructions of the tested plugs. Plug 6A includes more than one pieces. Plug
34A is the most heterogeneous (tightest) of all samples

SCA2015-021

1/12

LOW PERMEABILITY MEASUREMENTS: INSIGHTS

Sandra Profice, Grald Hamon and Benjamin Nicot
TOTAL CSTJF Avenue Larribau 64018 Pau Cedex France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
An accurate determination of the intrinsic permeability kl is essential for characterizing
hydrocarbon production from shale. However, recent studies have reported unacceptable
discrepancies (up to several decades) between kl measurements obtained by different
laboratories. These discrepancies could be explained by:
i.
The diversity of experimental protocols, since no standards of measurement have
been clearly defined for poorly permeable samples.
ii. The validity of interpretative models, which are rarely published.
Firstly, this paper presents the results of an experimental study comparing:
i.
Values of the intrinsic permeability kl from Step Decay (gas), Pulse Decay (gas)
and steady-state (both gas and liquid) tests
ii. Values of the Klinkenberg coefficient b from Step Decay and steady-state tests
iii. Values of the porosity from Step Decay and pycnometry tests.
On a homogeneous material principally composed of clay (pyrophyllite, 10 nD < kl < 60
nD), no matter which property is being measured (kl, b or ), different testing techniques
achieve similar results. Thus, our Step Decay method provides simultaneous and accurate
estimations of low porosities ( down to 2 %) and permeabilities (kl down to 10 nD).
Besides, the accurate prediction of gas flow behavior by the Step Decay and steady-state
interpretative models proves the relevance of the Darcy-Klinkenberg equation. In other
words, gas flows in tight rock can be described as viscous flows with slippage at pore
walls. To check these conclusions for reservoir rocks, the study was repeated on shale.
Secondly, we provide the results of a Round-robin test where Total and three commercial
laboratories performed unsteady-state and, when possible, steady-state measurements on
identical plugs of pyrophyllite. All laboratories chose their techniques and experimental
conditions but had to work on the whole plugs at a given effective pressure, to provide
estimations of kl. Furthermore, no prior treatment was carried out on the plugs in order to
prevent any bias in the results due to preparation. The comparison of the collected series
of kl values reveals that the results agree in a satisfactory manner.
Thus, after showing the validity of the classical interpretative models through preliminary
studies on pyrophyllite and shales, the results of the round robin test emphasize the need
to properly define experimental procedures in order to provide a framework in which to
compare low permeability results produced by different laboratories.

SCA2015-021

2/12

INTRODUCTION
Oil and gas shales have become over the past ten years a topic of real interest due to the
large amounts of hydrocarbons they could potentially produce. Their characterization is a
challenging task since these unconventional reservoirs have tight pore throats (a few tens
of nanometers) and low permeabilities (from microdarcies to nanodarcies). Consequently,
shales require careful identification of tractable and reliable methods to identify their onephase flow properties.
Steady-state techniques are the oldest and simplest ones. They allow the determination of
the intrinsic permeability kl and, when applied with gas at different mean pore pressures,
provide the Klinkenberg coefficient b in addition [1, 2]. Until recently, the principal
drawback of such methods lay in the time needed to achieve steady-state at each new
measuring point. The sample characterization used to be a long process, requiring several
hours or even days when the material was extremely tight. Alternative methods dedicated
to a faster analysis of ultra-low permeable porous media have been developed since the
early 50s. Bruce et al. [3] were the first authors to propose in 1953 an unsteady-state
technique commonly referred to as Pulse Decay. This pioneering work gave birth to
numerous other studies on the technique, mainly aiming at deriving interpretative models
[4, 5, 6]. To summarize, in most of the studies, k, apparent permeability, or kl is estimated
separately from the porosity and, in rare cases, simultaneously with b. The Pulse Decay
technique is still widely used in the petroleum industry, either on core plugs or on drill
cuttings. This cheaper and faster option consisting in working on cuttings, known as Gas
Research Institute (GRI) technique, was first described by Luffel and Guidry in 1992
[7]. Even if unsteady-state methods relying on the application of a pulse of pressure are
popular, they are not the only methods enabling a rapid and accurate characterization of
low permeable rocks. Indeed, some oil and gas companies have made the choice to be
equipped with different in-house techniques such as the improved steady-state technique
proposed by IFPEN [2], the Step Decay technique [8] developed by Total or the Pore
Pressure Oscillation technique [9] adopted by Shell. The first technique allows a fast
estimation of kl with a liquid. The second one delivers kl, b and simultaneously using a
series of pulses of pressure to excite the plug while the last one provides k and (b too if
several tests are done) using a sinusoidal pressure wave.
All techniques presented in the previous paragraph involve interpretative models based
on the first assumption that Darcys law is still valid when modeling fluid flows in poorly
permeable porous media. The second assumption that Klinkenbergs law [10] is valid too
is made when b is determined. However, many publications question these assumptions
since the Knudsen numbers typical of shales are out of the range of validity of the DarcyKlinkenberg law. Karniadakis and Beskok [11] and also Javadpour [12], who worked on
networks composed of micropores and nanopores respectively, rejected Darcys law and
suggested new formulations of the gas flow rate. Javadpour even derived an expression of
k depending not only on the material specificities but also on the fluid properties at given
values of temperature and pressure. In his approach, the notion of intrinsic permeability is
completely lost. Using Karniadakis and Beskoks theory, Civan [13] found a relationship

SCA2015-021

3/12

between k and kl applicable to the whole Knudsen number range. More recently, Fathi et
al. [14] have determined a similar relationship by theorizing the phenomenon of double
molecular slippage at the pore scale.
From these last studies, a natural question arises. Are interpretative models relying on the
Darcy-Klinkenberg law well-suited for shales? This is the first question this paper will try
to answer. The second question regards the great number of methods used in the industry
for routine measurements. Are the discrepancies between the results found on an identical
sample by different laboratories explained by the diversity of their interpretative models?
Indeed, several authors [15, 16] recently mentioned discrepancies between permeability
estimations up to several decades. The question remains whether this wide dispersion of
results is due to interpretation or to sample preparation.

STUDY
Three experimental studies were devised to answer the questions above. Study 1 aimed at
comparing the values of kl, b and estimated for homogeneous plugs of pyrophyllite with
common methods of the industry to those estimated with our Step Decay method. To be
more precise, the comparison was made between:
i.
The values of kl given by a Step Decay test, a Pulse Decay test, a steady-state test
with gas and a steady-state test with oil
ii. The values of b given by a Step Decay test and a steady-state test with gas
iii. The values of given by a Step Decay test and a pycnometry test.
The same approach was then applied in Study 2 to two shale plugs sourced from an actual
development target area, in order to check whether the results from the pyrophyllite study
could be corroborated by a similar study with reservoir rocks. Study 3 was a Round-robin
test involving three commercial laboratories and Total. Its goal was to identify the main
cause of the discrepancies often observed between the kl estimations found by different
laboratories for an identical sample.
Plugs
Study 1 was carried out on five plugs of pyrophyllite named Pyro 1, Pyro 2, Pyro 3, Pyro
4 and Pyro 5 respectively. Pyrophyllite is a homogeneous quarry rock sourced from the
United States. It is mostly composed of clay and consequently has a low permeability. All
plugs were successively subjected to Step Decay tests, Pulse Decay tests and steady-state
tests using gas and. Pyro 5, the least permeable of the five plugs, was also analysed with a
liquid at steady-state, at the end of the experiments with gas. No special treatment, such
as cleaning or drying, was performed on the pyrophyllite prior to the measurements with
gas, which were conducted therefore on plugs containing some water. Before starting the
steady-state experiment with liquid, Pyro 5 was saturated at 400 bar during two weeks,
after having created a vacuum over a period of four days. Pyro 1, Pyro 2, Pyro 3 and Pyro
4 were the four plugs sent to each of the three commercial laboratories selected for the
Round-robin test of the Study 3.

SCA2015-021

4/12

The two shales involved in Study 2, Shale 1 and Shale 2 respectively, are of different
origins. Shale 1 comes from a wet gas well and Shale 2 from an oil well. Both samples
were subjected to all the gas tests in native state. Step Decay, Pulse Decay and steadystate. No steady-state analysis with liquid was undertaken since the oil phase existing in
the plug would have probably moved with the flow. This would have led to an estimation
of kl that was not comparable with the results derived with gas.
A Mercury Injection Capillary Pressure (MICP) analysis was carried out on a fragment of
rock from the same block that provided Pyro 1. Figure 1, which shows the pore throat
size distribution from the MICP analysis, reveals that the most represented pore throat
radius in the material is around 20 nm. This result is in line with the range of pore radius
of 10 to 40 nm found by using the Density Functional Theory (DFT) method on a crushed
sample taken from the Pyro 1 block. Pores in shales are likewise characterized by radii of
several tens of nanometers.
Methods and Interpretative Models
The Pulse Decay technique [4] consists in applying a pulse of pressure on one face of the
plug and recording the differential pressure P calculated from measurements taken at
both extremes of the plug. As shown on Figure 2, the plug is confined in a Hassler sleeve
core holder connected to two tanks. The pulse of pressure is prepared in the upstream
tank V0 and released at the plug entrance by opening the valve v0. The recording of P is
triggered simultaneously with the pulse emission. The downstream tank V1 collects the
gas flowing out of the plug. All Pulse Decay experiments were conducted with the Pulse
Decay Permeameter 200, an automated device manufactured by Corelab, which works
with nitrogen and at high mean pore pressure Pp to eliminate the Klinkenberg effect. Pp
was initially stabilized at:
Shale 1: 31 bar (450 psi)
Plugs of pyrophyllite and Shale 2: 69 bar (1000 psi).
The gas flow from V0 to V1, which are both around 20 cm3, is started by depressurizing
V1 until P reaches 3 bar (40 psi). The effective pressure Peff, defined as the difference
between the confining pressure Pc and Pp, was:
Plugs of pyrophyllite: 60 bar
Shales: 100 bar.
The temperature was fixed at 20 C for all our tests (unsteady-state and steady-state). The
interpretation of P is based on Jones approach described in reference [5]. Jones devised
a method of calculating kl by using an approximate analytical solution of the Pulse Decay
problem that combines the traditional equations of fluid mechanics in porous materials
(including Darcys law) and excludes the Klinkenberg effect. The solution is an infinite
sum of exponentials decreasing over time. At long times, the first term predominates over
the other terms. As a consequence, the curve of the logarithm of P evolves linearly over
time, allowing the derivation of kl from the slope.

SCA2015-021

5/12

In the Step Decay technique [8], the plug is subjected to a series of pulses of pressure. In
terms of the device, the Step Decay differs from the Pulse Decay only by the existence of
an additional buffer tank Vb located at the plug entrance, as represented in Figure 3. The
first pulse of pressure is prepared by filling V0 and Vb with nitrogen. Once the selected
pressure setpoint is reached, valve vb is closed to isolate Vb from V0. The pulse is then
emitted by opening v0. This operation triggers the simultaneous recording of upstream
and downstream pressures, P0 and P1 respectively. All pulses of pressure from the second
pulse are produced by pressurizing Vb and are liberated by briefly opening vb. The pulse
pressures and waiting time intervals used for the Step Decay tests were the following:
Plugs of pyrophyllite: 6 bar 16 min / 10 bar 30 min / 31 bar 5 to 15 hours
Shale 1: 3 bar 1 hour / 5 bar 1.5 hours / 9 bar 2 hours
Shale 2: 6 bar 15 min / 13 bar 2 hours / 22 bar 20 hours
For each plug, the same Peff was applied for both Pulse Decay and Step Decay tests. The
values of V1 were of:
Pyro 1, Pyro 2, Pyro 3 and Pyro 4: 8.43 cm3
Pyro 5, Shale 1 and Shale 2: 12.46 cm3.
In terms of the interpretation of the raw data, the Step Decay model assumes the validity
of the Darcy-Klinkenberg law, as do most Pulse Decay models found in existing studies
of experiments at low pore pressure. The Step Decay provides kl, b and by matching a
numerically simulated P1 profile with the recorded P1 data. Note that a particular strength
of the method resides in the fact that the P0 profile acts as an input for the interpretation.
One consequence of that is the removal of V0 from the series of parameters required for
the inversion. As a result, the interpretation is only concerned with the measurement of
V1. Regarding , as evoked in reference [8], an increase in the sensitivity of P1 to was
necessary to ensure its accurate estimation by history matching. This was achieved thanks
to the techniques major strength resulting from the fact that P0 is an input of the history
matching: the modulation of the excitation. By emitting successive pulses of pressure, the
phenomenon of gas accumulation occurring right after the pulse emission and creating
sensitivity to is generated several times, compared to only once in a Pulse Decay test.
The results from unsteady-state experiments were compared to the results from a steadystate analysis carried out with nitrogen. In the case of Pyro 5, characterization at steadystate was also repeated with Isopar L. Figure 4 shows a diagram of a typical steady-state
device. Regardless of the fluid used, one point of measurement was made by regulating
P0 and recording at steady-state P0, P1 and the volume flow rate Qv. For all tests involving
nitrogen, several points were taken at increasing mean pore pressures to derive kl and b
independently. For the test with Isopar L, only one point was needed to determine kl. The
parameters chosen for the characterization with nitrogen were:
Plugs of pyrophyllite: 40 bar P0 70 bar, P1 = Patm (atmospheric pressure), Peff
= 60 bar
Shale 1: 40 bar P0 70 bar, P1 = Patm, Peff = 100 bar
Shale 2: 6 bar P0 11 bar, P1 = Patm, Peff = 100 bar.

SCA2015-021

6/12

In the steady-state test with Isopar L on Pyro 5, P0 was set at 80 bar, P1 at 2 bar and Peff at
60 bar. The interpretation of the steady-state raw data recorded with nitrogen relies on:
b
(1)
k = k l (1 + )
Pm
In this relationship presented first by Klinkenberg [10], k is calculated from Equation 2,
which was derived for an isothermal steady-state gas flow by integrating the differential
form of Darcys law including the Klinkenberg effect.
L P1 QV
(2)
k=
S Pm P
Note that is the viscosity, L the length, S the cross-section area and Pm the mean pore
pressure. According to Klinkenbergs law, it appears that the separate estimation of kl and
b requires the plot of k versus 1/Pm. Indeed, this plot displays a linear behaviour with a
slope equal to kl*b and an intercept with the Y-axis equal to kl. For the characterization
with Isopar L, the calculation of k is immediate with Darcys law:
L QV
(3)
k=
S P
Across all tests carried out with gas on a given plug, Pp varied approximately between 1
bar and 70 bar. For these extreme values of Pp, the mean free path defined by Equation
4 is in the range [1 nm - 94 nm].
=

P M

(4)

is the ideal gas constant, T the absolute temperature and M the molecular mass. For
such a range of and for pyrophyllite mean pore radius Rp of 25 nm (10 nm Rp 40
nm), the Knudsen number Kn derived from Equation 5 is between 0.04 and 3.8.

Kn =
(5)
Rp
Consequently, in pores having radii of a few tens of nanometers, either a slip flow regime
or a transition flow regime arises, depending on the level of pressure [17]. For the latter,
both Darcys law and Klinkenbergs law are seriously put into question knowing that gas
molecules collide principally with pore walls and no longer with other molecules.
Round-Robin Test
After characterization, Pyro 1, Pyro2, Pyro 3 and Pyro 4 were sent successively to three
different commercial laboratories for a round Robin-test comprising unsteady-state and,
when possible, steady-state experiments. The specifications emphasized conducting them
with nitrogen, on the whole plugs (no crushing), by regulating Peff at 60 bar and without
carrying out any treatment on the plugs (cleaning or drying) in order to prevent any bias

SCA2015-021

7/12

in the results due to preparation. The laboratories were free to choose their own methods
and experimental conditions (pressures, flow rates, temperature...).
Laboratory 1 performed unsteady-state tests with the Pulse Decay technique and steadystate tests following a procedure similar to ours. The interpretation of the Pulse Decay
raw data relied on the methodology implemented in the Pulse Decay Permeameter 200,
except that the analytical solution was not Jones but Brace et al.s [4]. The unsteadystate measurements were started by increasing P0 by a few bars (between 2.5 and 3.5 bar),
after the stabilization of Pp between 15.5 and 18 bar. V0 and V1 were of 58 cm3 and 47
cm3 respectively. Estimations of kl were deduced from the tests. Each analysis at steadystate involved several points of measurement to enable the separate identification of kl
and b. The mean Pp chosen for a given plug varied between a few bars (from 1.5 to 4 bar)
and 20 bar. Both types of test were carried out at a temperature of 24 C.
Laboratory 2 delivered values of kl and b from unsteady-state experiments done with the
Automated Permeameter - 68 manufactured by Coretest Systems. Pp was initially set at
14 bar. Once equilibrium was achieved, the gas flow was started by opening v1. Hence,
P1 was kept at Patm and P0, recorded in V0 of 6 cm3, was the only signal introduced in the
history matching procedure to obtain kl and b. The numerical model assumed as usual the
validity of Darcys law and Klinkenbergs law. The temperature was of 20 C.
Laboratory 3 provided estimations of kl from steady-state and unsteady-state analyses.
Steady-state tests relied on a unique point of measurement obtained by applying a similar
procedure to our own, P1 and P being of 6 bar and 5 bar respectively. The value of k
was then corrected with an unknown empirical correlation to derive kl. Unsteady-state
tests used the Pulse Decay method. The plug was first pressurized at 40 bar and the pulse
of pressure was then prepared by increasing P0 to 60 bar. V0 and V1 were of 15 cm3 and
13 cm3 respectively. The model, assumptions and procedure on which the interpretation
was based were not given. All tests were carried out at a temperature of 22 C. At the
moment of writing the present paper, the results of Laboratory 3 were not available but
will be presented during the conference.

RESULTS AND DISCUSSION

Results of the Comparative Studies
Table 1 lists the results of all measurements conducted on the five plugs of pyrophyllite.
The estimations of given by the Step Decay method were compared to those provided
by pycnometers employing helium at low pressure. The subscripts PD, SD, SSG,
SSO and Pyc refer to: Pulse Decay, Step Decay, Steady-State Gas, SteadyState Oil and Pycnometry respectively, in Table 1 as below. Moreover, the deviation
indicator D quantifying the discrepancy between 1 and 2, both estimations of (= kl, b
or ), and used throughout the present development has been defined as:
|1 2 |
(6)
D = 100
(1 + 2 )/2

SCA2015-021

8/12

For Pyro 1 to Pyro 5, Dkl derived for kl,SD and kl,PD is in the range [5 % - 17 %], against
[19 % - 40 %] when calculated for kl,SSG and kl,PD. Consequently, the values of kl from
experiments performed with gas, i.e. kl,PD, kl,SD and kl,SSG, agree in a satisfactory manner.
As highlighted in Table 1, this agreement can be extended to estimations of kl from tests
carried out with gas and oil, at least for pyrophyllite. Indeed, in the case of Pyro 5, Dkl
determined for kl,SSO and kl,SSG is about 26 %. Regarding b and , bSD compares well with
bSSG as well as SD with Pyc, Db ranging from 10 % to 43 % and D from 0 % to 19 %.
Figure 5.a presents, for Pyro 1, the normalised difference (P1,rec - P1,sim) / P1,rec where P1,rec
is the recorded profile P1 and P1,sim the profile P1 simulated with the Step Decay model.
Similar graphs were obtained for Pyro 2 to Pyro 5. The signal is flat and centered on 0,
excepted during a period of less than half an hour right after the pulse emission. Figure 6
gives an example of k plotted against 1/Pm. This graph results from the processing of the
steady-state raw data collected for Pyro 1. A similar linearity was observed for Pyro 2 to
Pyro 5.
Table 2 provides the results from the experiments conducted on Shale 1 and Shale 2. As
emphasized for pyrophyllite plugs, the same gas-based techniques used on shale plugs to
identify a given property, kl or b, lead to consistent estimates. Dkl is about 9 % for Shale
1 and about 19 % for Shale 2, when comparing kl,SSG and the mean of kl,SD and kl,PD, while
Db is about 21 % for Shale 1 and about 26 % for Shale 2. The estimations of SD are not
available since the characterization of with the Step Decay technique is uncertain when
dealing with shales. The bias in is mainly due to the fact that one major assumption of
the Step Decay model, which is the homogeneity of the studied rock, is rarely respected
in the presence of shales. In the case where the rock is highly heterogeneous, it has been
proven with numerical simulations and practical tests that the method delivers an accurate
estimation of kl, an acceptable estimation of b and an unreliable estimation of [18].
Results of the Round Robin Test
Table 3 lists the values of kl obtained for Pyro 1, Pyro 2, Pyro 3 and Pyro 4 by Total and
by the two laboratories (Laboratory 1 and Laboratory 2) which participated in the Roundrobin test and could deliver their results in time for the writing of the paper. Estimations
of b were provided by some of the laboratories but are not reported here since focus was
primarily on the characterization of kl. Regarding the values of kl measured by Laboratory
1 and by Laboratory 2 for a given plug, those obtained at unsteady-state were compared
to kl,SD while those obtained at steady-state were compared to kl,SSG. Dkl ranges from 31 %
to 118 % at unsteady-state and from 9 % to 27 % at steady-state, leading to a maximum
discrepancy factor of 3.8. Therefore, the discrepancies noted here are much more modest
than the two decades reported in the literature [15].
Discussion
As demonstrated by the first series of comparative measurements performed on the plugs
of pyrophyllite, our in-house Step Decay method enables the simultaneous and accurate
estimation of kl, b and , in the case of homogeneous and low permeable media. From the
tests on both pyrophyllite and shale plugs, it results that a satisfactory agreement between

SCA2015-021

9/12

the estimations of kl found with different methods used in the industry can be achieved.
This conclusion is also applicable to estimations of kl determined by different laboratories
working with their own protocols and interpretative models, as highlighted by the Roundrobin test. Therefore, the discrepancies of several decades reported in references [15, 16]
and noted between values of kl derived for a same plug by different laboratories cannot be
explained by the diversity of the experimental conditions selected for the design and use
of the devices (pressures, flow rates, volumes...). Similarly, these discrepancies cannot be
justified by the diversity of the interpretative models and of their assumptions: negligible
, no Klinkenberg effect, constant gas compressibility Consequently, the main cause of
divergence of the results delivered for an identical sample by several laboratories is likely
the sample preparation. This last conclusion is supported by reference [19] where Tinni et
al. evidence the great variability of kl according to the sample specificities when dealing
with cuttings. In addition, Darcys law and Klinkenbergs law remain usable in the case
of nanoporous rocks, like pyrophyllite or shale. This is clearly proven by the linear
behavior observed in Figure 6. In Figure 5.b are shown the discrepancies at short times
between the recorded and simulated P1 signals. They are due to the poor quality of the
recorded response P1 and probably to the inability of the Darcy-Klinkenberg model to
predict the gas flow since the values of Pp of a few bars at short times induce Knudsen
numbers higher than 0.1, i.e. out of the range of validity of the Darcy-Klinkenberg law.
However, despite the fact that the model is not well-adapted to predict the flow during a
short period, kl,SD remains reliable and accurate. Thus, the need to revise classical fluid
mechanics when testing rocks of low permeability is not justified in the unsteady- and
steady-state ranges of Pp considered in this paper.

CONCLUSION
Redundant estimations of kl can be determined for an identical poorly permeable plug, by
using different techniques of the industry. More precisely, steady-state and unsteady-state
methods applied with gas lead to results in excellent agreement on pyrophyllite and shale.
The convergence of the results from all measurements with gas and from a measurement
at steady-state with a liquid is true for pyrophyllite. The verification of this last point in
the case of shale is ongoing. Moreover, comparable estimations of kl can be determined
by different laboratories working with their own techniques, experimental conditions and
interpretative procedures, provided that sample preparation is carefully defined. Knowing
the numerous difficulties tied to the characterization of cuttings, we recommend carrying
out permeability measurements on whole plugs as much as possible. Finally, Darcys law
and Klinkenbergs law remain valid when modeling gas flows in nanoporous rocks such
as pyrophyllite and shale. In other words, gas flows in tight formations can be described
as viscous flows with slippage at pore walls.

ACKNOWLEDGEMENTS
The financial support from Total is gratefully acknowledged. Besides, we thank Ghislain
Pujol, Jean-Michel Kluska, Frdric Plantier, Sonia Vincent-Gill and Antoine Delafargue
for their contribution to the work presented in this paper.

SCA2015-021

10/12

REFERENCES
1. Rushing, J. A., Newsham, K. E., Lasswell, P. M., Cox, J. C., Blasingame, T. A.,
Klinkenberg-Corrected Permeability Measurements in Tight Gas Sands: Steady-State
Versus Unsteady-State Techniques, SPE Annual Technical Conference and
Exhibition, Houston, Texas, USA, 26-29 September 2004.
2. Boulin, P. F., Bretonnier, P., Gland, N., Lombard, J. M., Contribution of the SteadyState Method to Water Permeability Measurement in Very Low Permeability Porous
Media, Oil and Gas Science and Technology - Rev. IFP Energies nouvelles, (2012)
67, 3, 387-401.
3. Bruce, G. H., Peaceman, D. W., Rachford Jr., H. H., Rice, J. D., Calculations of
Unsteady-State Gas Flow Through Porous Media, Journal of Petroleum Technology,
(1953) 5, 3, 79-92.
4. Brace, W. F., Walsh, J. B., Frangos, W. T., Permeability of Granite Under High
Pressure, Journal of Geophysical Research, (1968) 73, 6, 2225-2236.
5. Jones, S. C., A Rapid Accurate Unsteady-State Klinkenberg Permeameter, SPE
Journal, (1972) 12, 5, 383-397.
6. Hsieh, P. A., Tracy, J. V., Neuzil, C. E., Bredehoeft, J. D., Silliman, S. E., A
Transient Laboratory Method for Determining the Hydraulic Properties of Tight
Rocks - 1. Theory, International Journal of Rock Mechanics and Mining Sciences &
Geomechanics Abstracts, (1980) 18, 3, 245-252.
7. Luffel, D. L., Guidry, F. K., New Core Analysis Methods for Measuring Reservoir
Rock Properties of Devonian Shale, Journal of Petroleum Technology, (1992) 44, 11,
1184-1190.
8. Lasseux, D., Jannot, Y., Profice, S., Mallet, M., Hamon, G., The Step Decay: A
New Transient Method for the Simultaneous Determination of Intrinsic Permeability,
Klinkenberg Coefficient and Porosity on Very Tight Rocks, International Symposium
of the Society of Core Analysts, Aberdeen, Scotland, 27-30 August 2012.
9. Wang, Y., Knabe, R. J., Permeability Characterization on Tight Gas Samples Using
Pore Pressure Oscillation Method, International Symposium of the Society of Core
Analysts, Halifax, Nova Scotia, Canada, 4-7 October 2010.
10. Klinkenberg, L. J., The Permeability of Porous Media to Liquids and Gases, API
Drilling and Production Practice, (1941), 200-213.
11. Karniadakis, G., Beskok, A., Aluru, N., Microflows and Nanoflows, Fundamentals
and Simulation, Springer-Verlag, New-York, 2005, 818 p.
12. Javadpour, F., Nanopores and Apparent Permeability of Gas Flow in Mudrocks
(Shales and Siltstones), Journal of Canadian Petroleum Technology Distinguished
Author Series, (2009) 48, 8, 16-21.
13. Civan, F., Effective Correlation of Apparent Gas Permeability in Tight Porous
Media, Transport in Porous Media, (2009) 82, 2, 375-384.
14. Fathi, E., Tinni, A., Akkutlu, I. Y., Shale Gas Correction to Klinkenberg Slip
Theory, Americas Unconventional Resources Conference, Pennsylvania, USA, 5-7
June 2012.
15. Passey, Q. R., Bohacs, K. M., Esch, W. L., Klimentidis, R., Sinah, S., From OilProne Source Rock to Gas-Producing Shale Reservoir - Geologic and Petrophysical

SCA2015-021

11/12

Characterization of Unconventional Shale Gas-Reservoirs, CPS/SPE International

Oil & Gas Conference and Exhibition, Beijing, China, 8-10 June 2010.
16. Sondergeld, C. H., Newsham, K. E., Comisky, J. T., Rice, M. C., Rai, C. S.,
Petrophysical Considerations in Evaluating and Producing Shale Gas Resources,
SPE Unconventional Gas Conference, Pittsburgh, Pennsylvania, USA, 23-25
February 2010.
17. Ziarani, A. S., Aguilera, R., Knudsens Permeability Correction for Tight Porous
Media, Transport in Porous Media, (2012) 91, 1, 239-260.
18. Profice, S., Mesure de Proprits Monophasiques de Milieux Poreux Peu Permables
par Voie Instationnaire, PhD Thesis, Universit de Bordeaux, France, 2014.
19. Tinni, A. O., Permeability Measurements of Nanoporous Rocks, Master Thesis,
University of Oklahoma, USA, 2012.
1.5
1.25

dSHg/dLog(Pcap)

0.75
0.5

0.25

0
0.001

0.1

10

Pore throat radius (m)

Figure 1 : Pore throat size distribution

Figure 3 : Step Decay device

Figure 2 : Pulse Decay device

Figure 4 : Steady-state device

Figure 5.a : Normalised difference between P1,rec and P1,sim versus time

SCA2015-021

12/12

Figure 5.b : P1,rec and P1,sim signals at short times

3.3E-13

k*Pm (m2.Pa)

3.1E-13
2.9E-13
2.7E-13
2.5E-13
2.3E-13
2.0E+06

Pm (Pa)
2.4E+06

2.8E+06

3.2E+06

3.6E+06

Figure 6 : k*Pm versus Pm

Table 1 : Results of the comparative measurements for the plugs of pyrophyllite
Plug
Pyro 1
Pyro 2
Pyro 3
Pyro 4
Pyro 5

Plug
Shale 1
Shale 2

kl (nD)
kl,PD
70
35
67
51
11

kl,SD
59
33
56
46
10

kl,SSO
8

b,SD
19.6
24.0
18.5
20.2
33.3

b,SSG
22.4
26.6
23.0
24.6
21.5

,SD
4.2
3.8
4.0
3.7
2.5

,Pyc
4.1
3.5
4.0
3.5
2.1

Table 2 : Results of the comparative measurements for the shales

kl (nD)
b (bar)
(%)
kl,PD
kl,SD
kl,SSG
kl,SSO
b,SD
b,SSG
,SD
,Pyc
2031
2222
2303
4.3
3.5
10.1
46
58
43
7.4
9.6
0.7

Plug
Pyro 1
Pyro 2
Pyro 3
Pyro 4

(%)

b (bar)
kl,SSG
56
24
53
42
14

kl,PD
70
35
67
51

kl,SD
59
33
56
46

Table 3 : Results of the round robin test

Total
Laboratory 1
kl (nD)
kl,SSG
kl,SSO
kl,PD
kl,SSG
56
115
48
24
45
22
53
121
59
42
78
55

Laboratory 2
kl,PD
203
127
185
104

kl,SSG
-

SCA2015-022

1/12

UNCONVENTIONAL TIGHT OIL RESERVOIRS:

A CALL FOR NEW STANDARDIZED CORE ANALYSIS
WORKFLOWS AND RESEARCH
Gary A. Simpson and Neil S. Fishman
Hess Corporation
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August 2015

ABSTRACT
Recommended Practices for Core Analysis is firmly established in the American
Petroleum Institutes RP 40 Second Edition, February 1998 document. Standardized
workflows for the core analysis of unconventional shale reservoirs are deeply rooted in
analytical techniques that were originally developed for coal bed methane, shale gas and
tight gas reservoirs. These techniques were researched and developed by the Gas
Research Institute (GRI) and are outlined in the GRI final report: GRI-95/0496,
Development of Laboratory and Petrophysical Techniques for Evaluating Shale
Reservoirs. These methodologies were extremely successful in evaluating cores from
gas shale reservoirs. As operators shifted their development focus from gas shale
reservoirs to unconventional oil producing reservoirs, these same gas core analytical
techniques were used to analyze unconventional tight oil cores.
This paper will discuss the results of tests and experiments that were made on core from
the Bakken petroleum systems Middle Bakken and lower Three Forks formations. These
series of tests were made following a very poor core-to-log data water saturation
comparison. The core analysis measured 40% to 60% water saturation, and the logs
showed the reservoir to be 100% water saturated.
This investigation started by looking closely at the solvent extraction methods with Dean
and Stark [3] using toluene followed by chloroform/methanol azeotrope extraction.
Preliminary tests focused on the possibility of anhydrite dissolution, which would have
created excess porosity, thus making the core analysis incorrect. Additional investigation
of the problem included measurements with nuclear magnetic resonance (NMR), retort
extraction methods and possibilities of errors in the computations due to extremely high
water salinities. While this investigation has resulted in applying some new protocols in
the analysis of tight oil cores, it has also demonstrated that more research is needed and
new standard core analysis workflows need to be developed for unconventional tight oil
reservoirs.

SCA2015-022

2/12

INTRODUCTION
One of the most important aspects in the development of unconventional reservoirs is a
comparison of core measurements with wireline logging measurements. This comparison
primarily accomplished by cutting whole core and completing extensive wireline logging
in pilot hole wells, is particularly important in the development of unconventional tight
oil reservoirs. Unconventional tight oil reservoirs typically have a porosity of less than
10% and a permeability of less than 0.01 mD.
The Bakken petroleum system, one of the largest tight oil deposits in the world, is
situated in the Williston Basin [Fig. 1]. The Bakken petroleum system is generally
considered a hybrid type of unconventional tight oil reservoirs: it consists of two layers of
organic-rich shale (the Upper and Lower shales) sandwiching a dolomitic-siltstone
interval called the Middle Bakken [Fig. 2]. Included in the Bakken petroleum system
below the Lower shale is the Three Forks formation, composed of laminations of
dolostone and dolomitic-mudstone. Developing large acreage positions in the Bakken
petroleum system to maximize hydrocarbon production requires a thorough
understanding of the reservoir parameters across the play and an understanding of the
individual reservoirs of the system.
To evaluate the production potential more accurately, a series of five wells [Fig. 3] were
cored and logged in the Bakken petroleum system starting at the end of 2013 and through
2014. The wells were positioned over a broad section of leases that represented the
acreage being developed. During this process, routine core analysis (RCA) saturation
data from parts of the middle Bakken and the lower Three Forks formations were
compared with wireline logging data. The workflows recommended by API RP 40 [1]
were implemented to acquire the RCA saturation data. The comparison revealed obvious
discrepancies between the core and log data sets.
Advanced logging measurements using dielectric and NMR logs had been run and
analysed over these formations. The NMR logging tools measure total porosity, while the
dielectric logging tools measure the total water-filled porosity. Therefore, by plotting the
NMR total porosity with the dielectric water-filled porosity, the difference between the
two measurements is a hydrocarbon-filled porosity. A plot of the porosities from the log
measurements is shown in Fig. 4 in the second log track from the right, labeled
Porosity. The dark green shading between the two curves represents oil-filled porosity.
The teal blue color represents water-filled porosity. The maroon dots are core porosity
measurements.
Due to these log measurements, there was a high level of confidence that log data in the
lower portion of the Three Forks formation indicated 100% water saturation, as
represented in Fig. 4 by the blue curve in the middle log track labeled Water
Saturation. In the same intervals, water saturations from the RCA data, shown as
maroon dots, largely ranged from 40% to 60% and as such were highly questionable.

SCA2015-022

3/12

This discrepancy triggered an inquiry into its sources, with a focus on the core data. The
authors started with the core service company working on the core in which a large
discrepancy in the data sets was observed; a second inquiry began into other cores from
the same formations in process at a second core service company. Additionally, the
authors examined historic core data. It was concluded that this was probably not the first
time that a discrepancy between the data sets existed.
Research into the reliability of the core data led, over time, to a more comprehensive
understanding of analytical procedures and lab workflows for data acquisition from the
Bakken petroleum system. A series of lab visits, detailed discussions with lab analysts
and a review of industry-accepted protocols were all-necessary to define the scope of the
problems and develop tests to address analytical procedures and lab workflows. The
authors focus has been on the following:

Solvent extraction procedures during RCA on both core plugs [1]; as well as
crushed rock using the GRI method [2], or a modified version of GRI (following a
proprietary lab workflow that diverges from the original GRI workflow), which
itself was developed for characterizing reservoir properties of gas shales.
NMR as an independent check of plug porosity and fluid saturations.
The possibility of anhydrite dissolution during solvent extraction. In the absence
of resolution to this investigation, processing of previously unanalyzed core
materials has been suspended. Cores 1 and 2 had already been analyzed, Core 3
had had plugs cut and Dean Stark had been started on some, and Cores 4 and 5
were being evaluated for sampling intervals. [Fig. 3]

DISCUSSION
Solvent Extraction
Cleaning of core materials was undertaken on both 1.5 x 2-inch core plugs and on
material that was crushed. Cleaning of the core plugs followed the procedures originally
developed by Dean and Stark [3] and a workflow outlined in API RP 40 [1]. For both
Dean Stark and GRI, the samples are subjected to extraction using toluene at 110oC.
Questions were raised regarding the efficacy of the Dean Stark workflow because it was
developed for conventional reservoir rocks, but applied here to the Bakken and Three
Forks formations, considered unconventional tight oil reservoirs [4].
A quick test of the Dean Stark method was made on six plug samples of Bakken and
Three Forks rocks from Core 4 [Fig. 3]. Core plugs were first subjected to RCA and Dean
Stark cleaning. Lab protocols for cleaning required no measurable increase in water level
in the extraction apparatus after at least 24 hours. Upon completion of the cleaning
process, the plugs, still saturated with toluene, were removed from the extractor and
immediately crushed. The crushed material was placed back into the extractor, and
additional hydrocarbons and water were removed during this second episode of
extraction [Figs. 5A and 5B]. Preliminary results from this quick test indicate that as
much as 15% more water was removed from the crushed material, which would suggest

SCA2015-022

4/12

that there was incomplete cleaning of the core plugs through the normal RCA Dean Stark
process. Additional testing is currently underway to confirm the degree to which the
Dean Stark procedures on the Bakken and Three Forks rocks are inadequate to fully clean
core plugs, and to verify the volume of water left in the core plugs after the normal RCA
workflow.
While attempting to understand the efficacy of RCA on plugs from the Bakken and Three
Forks formations, an unexpected error was uncovered in calculating water saturations
from the volume of extracted distilled water using Equation 1:
Sw = (VDW VBrine /VPure Water )/VPore

(1)

Where the brine volume is calculated as follows:

VBrine = [(VPureWater Pure Water )/Brine ][1,000,000/(1,000,000 Csalinity )]

(2)

(Sw, water saturation; VDW, volume of distilled water; VBrine, volume of brine; VPure Water,
volume of pure water; VPore, pore volume; , density)
In the absence of client guidance, two labs used different default salinity values in
Equation 2 to calculate water saturations in Equation 1: one lab applied a default salinity
value of 50,000 ppm (density of ~1.03 g/cc); and the second lab applied a default salinity
value of 30,000 ppm (density of ~1.02 g/cc). Pore waters in the Bakken petroleum system
are extremely saline, with salinities measured on produced waters at the surface ranging
up to 360,000 ppm (density of ~1.29 g/cc) or greater. As a result, using the correct values
for water salinity resulted in as much as a 15.34% difference in the calculated water
saturations, especially in rock intervals that are dominantly water-wet. Therefore, the
simple matter of using correct values for pore water salinity allowed for a significant
change in calculated saturation values, and thereby corrected core data to match log
data more closely. Use of the appropriate salinity affected saturation calculations for
samples subjected to either RCA or GRI. Fig. 6 is a plot of water saturation for 350 core
plugs. The blue line is computed water saturation assuming a formation water salinity of
50,000 ppm; the red line is computed water saturation assuming a formation water
salinity of 350,000 ppm.
Given the extreme salinity of formation waters in the Bakken petroleum system, there
was a suspicion that salt (halite) precipitation was a possibility in core material. Analysis
confirmed the presence of halite occluding pores in core material [Fig. 7] as well as in at
least some X-ray diffraction analyses. The authors now believe that the halite precipitated
after the core was brought to the surface and subjected to ambient conditions. (Formation
temperatures are approximately 250oF, and pressures are as high as 7,000 psi.) The halite
not only occludes pores but also occurs as a partial coating on diagenetic minerals in the
rocks. This conclusion corroborates the assumption that the halite is a contaminant
formed after the cores were retrieved. The presence of halite in the rocks calls into
question whether salt can be completely removed during the chloroform/methanol
azeotrope and subsequent methanol cleaning steps of RCA, especially given that

SCA2015-022

5/12

incomplete cleaning of water and hydrocarbons during RCA toluene extraction had been
observed [Figs. 5A and 5B]. Further, because the GRI protocol as written [2] does not
include either a methanol/chloroform or methanol cleaning after toluene extraction,
incomplete removal of halite can be expected in crushed rock analyses. Indeed, initial lab
tests showed that both post-toluene cleaning steps are required on crushed rock samples
to remove the contaminant halite completely from the rocks; the methanol/chloroform
azeotrope cleaning proved insufficient to remove all halite, and a methanol cleaning was
required. Because incomplete removal of halite will affect subsequent measurements of
materials, it is likely that the core measurements are in error from the lack of thorough
salt removal. The significance of this source of error is being evaluated.
Comparison of Nuclear Magnetic Resonance and Extraction Saturation Analysis
Core NMR measurements were run to attempt an independent analysis of the core
saturations without using an extraction method. The NMR measurements required a
1-inch diameter core plug. This was accomplished by under coring preserved 1.5-inch
diameter plugs that were twins to original core analysis. These 1-inch plugs were
measured in native saturation using a high field 20 MHz NMR spectrometer by acquiring
both T2 and T1 datasets. From these measurements, porosity, oil and water saturations
were calculated. Following NMR measurements, the 1-inch plugs were placed in Dean
Stark apparatus and normal Dean Stark measurements were made on the plugs. The
remnants from this core were used to make retort extraction analysis measurements.
Table 1 shows a comparison of the retort, NMR and Dean Stark measurements for
porosity, oil saturation, and water saturation. The water and oil saturations calculated
from Dean Stark uses water salinity of 360,000 ppm and oil gravity of 45 API.
Fig. 8 contains plot of all of the data in Table 1. Fig. 8A is a plot of porosity and shows
that Dean Stark and NMR porosities are similar whereas Retort porosities are generally
lower in value. Fig. 8B is a plot of oil saturations for all of the methods and shows that
Retort gives lower values, Dean Stark the highest values and NMR is generally in
between the other two. Fig. 8C, is a plot of the water saturations and shows that Dean
Stark has the lowest saturation values, Retort the highest and the NMR saturations values
varying both higher and lower when compared to the other two methods. Therefore, the
variation in the results of this data would seem to indicate that there is uncertainty as to
which method is more accurate in determining saturations in unconventional tight oil.
It should be mentioned that the current version of API RP 40 protocol for retort
procedures does not include a brine-density volume correction factor as required in the
Dean Stark procedures using Equations 1 and 2. Additionally, this protocol does not have
any procedures for the removal of salts from the sample during the test.
Anhydrite Dissolution
Early in the deliberations about possible sources of error in core measurements, the
likelihood of gypsum dissolution by methanol during RCA was discussed. This question
arose from the observation that the interval where the difference in log and core
saturations was occurring started where anhydrite was appearing in the log mineral
analysis. Although little or no indication of gypsum was observed in the samples,

SCA2015-022

6/12

anhydrite is present, particularly in the Three Forks. That gypsum is methanol soluble is
expected based on relevant literature [5, 6]. Although lab experiments have shown that
anhydrite solubility is higher than that of gypsum [6], it was unclear whether some of the
anhydrite in the Three Forks rocks was dissolving during the cleaning process when
methanol was used. Thus, the service labs were instructed to perform bench top tests;
anhydrite from the Three Forks was carefully weighed and then bathed in methanol. After
approximately two weeks, the samples were carefully removed from the beakers, dried
and reweighed, with little to negligible weight loss from the beginning of the experiment.
From these test results, the concerns regarding anhydrite dissolution during the Dean
Stark cleaning process seem unfounded.

CONCLUSION
An ongoing evaluation to understand possible reasons why core data are mismatched to
log data from the Bakken and Three Forks formations has uncovered several issues that
are being considered as sources of errors in the original core data. Most notable is the
problem with core analysis that may have roots in the salinities of the pore waters in these
rocks. The very high salinities require special attention in the calculations of water
saturation. In addition, salt precipitation in the cores on retrieval to the surface can lead to
porosity occlusion and therefore to incomplete cleaning during routine core and crushed
rock analysis. The salt precipitation may be considered a principal hurdle to overcome
when performing lab tests on either core plugs or crushed samples. Diligence is required
to inform core service labs of formation water salinity and produced oil density values
before commencing lab work, and to ensure that errors are minimized in calculating
water and oil saturations.
The characteristically low porosity and permeability of Bakken and Three Forks rocks,
coupled with the concerns regarding salt precipitation, calls into question the use of
existing core analytical approaches and workflows to obtain accurate core data from
unconventional tight oil rocks. Existing methods were developed for conventional
reservoir rocks (RCA) and gas shales (GRI). Consequently, the authors propose research
and development of analytical protocols and workflows specific to unconventional tight
oil reservoirs (low porosity/permeability) and associated rocks. Finally, our studies
suggest that the API RP 40 is due for a review and revision, and should be updated to
include detailed analytical techniques suitable for unconventional reservoirs.

ACKNOWLEDGEMENTS
The authors thank Hess Corporation and the Bakken Asset Team for permission to
publish this paper, and the service company representatives for putting up with relentless
questioning and insistence to try different tests to evaluate the nature and degree of the
problems. Specifically, the authors sincerely thank Pat Lasswell and Ron Martino at
Weatherford, and John Dacy, Craig Merritt, Michael Long, Stacy McWhorter, Greg
Salter, Tom Swisher and Wayne Sealey at Core Laboratories.

SCA2015-022

7/12

REFERENCES
1. API RP 40, Recommended Practices for Core Analysis. 1998. Washington, DC: API.
2. Luffels, D. L. and F. K. Guidry. 1992, New core analysis methods for measuring
reservoir rock properties of Devonian shale. Journal of Petroleum Technology 44:
1184-1190.
3. Dean, E. W. and D. D. Stark, 1920. A convenient method for the determination of
water in petroleum and other organic emulsions. Industrial and Engineering
Chemistry 12: 486-490.
4. Gaswirth, S. B. and K. R. Marra, 2015. U.S. Geological Survey 2013 assessment of
undiscovered resources in the Bakken and Three Forks Formations of the U.S.
Williston Basin Province. American Association of Petroleum Geologists Bulletin 99:
639-660.
5. Marshall, W. L. and R. Slusher, 1966. Thermodynamics of calcium sulfate dehydrate
in aqueous sodium chloride solutions, 0-110. Journal of Physical Chemistry 70:
4015-4027.
6. Kan, A. T., R. Gongmin, and M. Tomson, 2003. Effect of methanol and thylene
glycol on sulfates and halite scale formation. Industrial and Engineering Chemistry
Research 42: 2399-2408.

Fig. 1 - Map showing Williston Basin (outlined in red) in North America

SCA2015-022

8/12

Fig. 2 - Generalized stratigraphic column of the Williston Basin

Fig. 3 - Map showing the location of five cored wells in the Bakken petroleum system taken in 2013 - 2014

SCA2015-022

9/12

100 Sw 0

Fig. 4 - This example shows a poor log to core water saturation comparison in the middle water-saturation
log track, primarily over the Three Forks Zones 3 and 4

SCA2015-022

Fig. 5A - Dean Stark extraction apparatus showing

orange discoloration of toluene in boiling flask. The
discoloration resulted from hydrocarbons released
after crushing a core plug that had already been
subjected to RCA Dean Stark cleaning

10/12

Fig. 5B - Water level in the calibrated receiver of

the Dean Stark extraction apparatus. This additional
water was extracted from the sample after crushing
a core plug that had already been subjected to
complete RCA Dean Stark cleaning

Fig. 6 - Comparison of 350 RCA cores, changing formation water salinity from 50,000 ppm to 350,000 ppm

SCA2015-022

11/12

Fig. 7 - Scanning electron micrograph showing halite crystals (H) that partly occlude pores and coat
diagenetic minerals

Retort
Sample

Porosity

Number
%
30Rb

2.14

48Rb

3.22

53RB

4.60

58RB

2.58

62Rb

6.45

84Rb

4.69

108Rb

3.46

139Rb

2.82

161RB

5.75

183Rb

5.86

195RB

3.09

207Rb

3.91

216Rb

6.59

241RB

5.72

247Rb

4.95

Porosity
NMR DeanStark
Total
Porosity
%
2.2
4.9
6.0
2.5
8.4
6.1
4.6
6.4
5.4
4.6
3.8
1.2
4.7
7.0
4.9

Retort

Oil Saturation
NMR
DeanStark

Water Saturation
Retort
NMR DeanStark

360K & 45API

Porosity

Oil

% PV

1.94

0.00

4.91

4.48

5.58

13.57

1.98

12.67

8.74

19.62

7.20

0.00

5.54

3.64

7.89

3.67

4.91

2.71

10.69

0.00

4.26

0.00

2.25

0.00

6.29

2.71

6.84

2.95

3.88

0.00

Oil
% PV
9.4
13.9
18.3
18.4
18.0
10.6
14.2
19.5
3.5
2.6
5.6
10.2
6.7
2.7
6.0

360K

Oil

Water

% PV

% PV

28.7

52.45

20.1

77.87

27.3

68.91

9.7

44.33

11.3

71.39

10.1

79.88

11.0

76.97

20.2

58.47

41.5

85.08

9.2

80.46

29.6

72.58

20.2

72.58

17.7

66.17

17.9

90.37

12.8

76.06

Water
% PV
42.7
32.5
52.8
69.0
25.7
40.3
22.6
18.8
94.9
48.6
90.3
36.9
44.2
74.3
45.5

Water
% PV
21.3
31.2
43.5
44.5
39.7
56.0
33.4
8.2
69.0
54.9
63.4
50.8
42.6
58.2
28.9

Table 1 Comparison of porosity, oil saturation and water saturation measured on preserved core plugs using three
different analysis techniques. Values in the green columns were from retort extraction methods; values in blue were
from a 20 MHz lab NMR spectrometer; and values in yellow were computed from Dean Stark solvent
extraction methods

SCA2015-022

Fig. 8 Porosity, Oil and Water Saturation plots from Retort, NMR and Dean Stark

12/12

SCA2015-023

1/12

COMPUTERIZED X-RAY MICROTOMOGRAPHY

OBSERVATIONS AND FLUID FLOW MEASUREMENTS
OF THE EFFECT OF EFFECTIVE STRESS ON
FRACTURED LOW PERMEABILITY POROUS MEDIA
Nathan J. Welch1, John P. Crawshaw1, and Edo S. Boek2
1
Department of Chemical Engineering, Qatar Carbonates and Carbon Storage Research
Centre, Imperial College London, South Kensington Campus, London SW7 2AZ, U.K.
2
Department of Chemistry, Cambridge University
e-mail: [email protected]
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
This work shows the observed changes in sample permeability and fracture geometry that
occur within a fractured piece of low permeability shale with different effective stress
gradients applied to the plug sample. These observations were made using an in-situ
micro-CT fluid flow imaging apparatus with a carbon fibre Hassler-type core holder.
Changes within the preferential fluid flow path under different stress regimes as well as
physical changes to the fracture geometry are reported. Lattice-Boltzmann single phase
permeability flow simulations were performed on the extracted flow paths and compared
to experiment permeability measurements.
Experimental flow measurements showed significant changes in the permeability of the
fractured system with varying stress states. Imaged fracture apertures were also used in
simple parallel plate flow calculations, and compared to results from fluid flow
simulations on segmented sample void space and experimental fluid flow measurements.

INTRODUCTION
Ensuring the long-term safe geologic storage of carbon dioxide has become an important
scientific and political issue for wide spread implementation of carbon dioxide
sequestration operations. An intact and substantial seal formation above a storage
reservoir is required for a significant portion of the initial sealing mechanisms believed to
occur during carbon dioxide storage operations [1]. Shales and evaporites are common
seal formation rock types found above numerous hydrocarbon reservoirs, as well as
potential saline aquifer storage locations [2]. These seals are known for having very low
permeabilities, however, some also have the tendency to be quite fissile, and the
formation of fractures within these seals can have a significant detrimental effect on the
sealing potential of a reservoir and amount to large areas of high permeability and low
capillary pressures compared to the surrounding intact rock [3].

SCA2015-023

2/12

Work has continued to progress the rapid determination of these sealing parameters in the
selection of suitable sequestration sites. A recent review of all available literature for
capillary entry pressure measurements in an attempt to predict sealing characteristics of
these seal formations shows the substantial lack of data available for interpretation [4].
Numerous experimental techniques have been developed for quickly determining these
properties [5-9] with an overall comparison of each method presented by Boulin et al.
[10]. Intact seal samples typically have nanometre sized pores as well as very low
permeabilities, and fluid flow through fractures would largely dominate any observed
fluid behaviour in the system. Fractures in these systems therefore allow for application
of more traditional core analysis techniques [11] due to the rise in permeability and
decrease in capillary pressure.
In this work, we were able to image the individual fractures found to be present in a
fractured sample using our recently developed in-situ microtomography fluid flow
apparatus. The purpose of this work is to attempt to relate observed changes in the sample
fracture network geometry with varying system stress to changes in recorded fluid flow
behaviour, and also to compare predicted fluid flow behaviour using simple equations for
fluid flow between parallel plates and numerical fluid flow simulations, to experiment
results.

EXPERIMENT DESIGN
The fractured sample used in all of these observations was a cored plug sample taken
from a large carboniferous shale quarry sample received from Todhills Long Lane Quarry
in the UK. The sample was extracted via bulldozer from one of the working benches of
the quarry and shrink-wrapped to try and prevent loss of natural pore fluid during
shipping. The approximately 20 kg sample arrived after 2 days of shipping still in the
original packaging material. Upon removal, the sample remained in one large roughly
cube-shaped mass. The sample was then immediately wrapped in several layers of heavy
black plastic and stored for later use.
High pressure mercury intrusion (Micromeritics AutoPore IV 9500, Cup and stem
penetrometer) measurements were recorded on oven dried fragments collected from the
larger mass shown in Figure 1. After drying, these fragments were further broken down
into pieces no longer than 6 mm of any side length to fit into the mercury cup of the
penetrometer. The majority of pores of the intact sample pore structure were <100nm in
size, with a total porosity of 9.77%.

SCA2015-023

3/12

Figure 1: HPMI results for Long Lane Quarry shale

Large chunks of the original sample were broken away by hand with little force along
natural cleavage planes of the sample for plug drilling. The samples reactivity to water
was immediately apparent on contact with tap water used for drilling. The sample began
to fail along multiple parallel cleavage planes due to clay swelling and resulted in
laminate sheets of about 1 cm in thickness. These smaller cleaved sheets still had a fairly
high strength to them, and could not be broken by hand. In order to avoid the clay
swelling, later plugs were instead drilled using compressed air to remove debris and cool
the coring bit. The sample used in this study was drilled and then came into contact with
untreated water resulting in the observed fracture network. This caused the sample to
immediate begin to break apart due to swelling and it was then wrapped in PTFE tape to
maintain the original configuration of fractured pieces.
Fluid flow measurements and x-ray microtomography observations were collected in our
[12] micro-CT fluid flow apparatus designed for high resolution in-situ imaging of fluid
flow behaviour in porous samples. Tomograms were collected with an Xradia XRM-500
computerized microtomography apparatus. A specially designed Hassler-type flow cell
was used for this system designed by Airborne International. The flow cell has a
maximum internal diameter of 19 mm, and can accommodate samples up to 212 mm in
length. The flow cell is constructed of high strength carbon fibre for low x-ray
absorbance, and sealed with stainless steel end caps. ISCO 260D syringe pumps were
used to maintain all fluid pressures and record flow rates.
The fractured sample was then carefully removed from the PTFE tape and placed inside
Viton tubing, while maintaining the original alignment of the fractured pieces. Current
capabilities of the core holder system limits the system to only applying uniform
confining pressure around the whole of the sample with no other axial stress provided.

SCA2015-023

4/12

SINGLE PHASE PERMEABILITY OBSERVATIONS

After loading the sample into the fluid flow cell, water was used the annulus of the flow
cell as the systems confining fluid. This was done to try to ensure the x-ray transmission
of the dry scan was similar to that during fluid flow. A dry tomogram was recorded to
show the zero-stress geometry of the system.
Avizo 9.0 produced by FEI [13] was used for initial tomogram visualization and
manipulation. Orthogonal slices of the dry fractured sample can be seen in Figure 2. To
avoid interference with image segmentation and later analyses, a region similar to the one
represented with the white cornered rectangles in each plane was extracted from each 3D
reconstruction recorded. This region of interest was selected to exclude end piece
artefacts and the core sleeve, being the largest inscribed rectangular shape that could be
fit within the sample. Registration and exact extraction of a similar region would have
been more feasible if not for the rather high anisotropy of compressibility exhibited by
the sample.
A)

B)

Figure 2: Orthogonal slices of the dry scan tomography highlighting the region of interest extracted
avoiding x-ray artefacts in full 3D space. A) x-y and B) x-z orthogonal slice. All scale bars shown are
a length of 5000 m.

Following the extraction of the region of interest, a noise-reducing, edge-preserving nonlocal means filter [14] was applied to the image data set to assist in image segmentation
and analysis for the dry scan. Following dry scan imaging, 5wt% NaCl, 1wt% KCl brine
was introduced to the sample and pressurized to an initial system pressure of 6 bar
confining pressure, 5 bar upstream pressure, and 4 bar downstream pressure. This brine
was selected to try and reduce further reactivity with changes in ion concentration within
the shale matrix and pore fluid. No further fracturing was observed from the introduction
of this wetting phase brine.
System stress was then varied, initially holding a constant pressure drop of 1 bar across
the fractured sample, and incrementally increasing confining pressure of the system from
6 bar to 120 bar. After each change in system stress, a tomogram was recorded for the

SCA2015-023

5/12

system and permeability measurements recorded based on volume displacement in

system pumps over the course of the scan. System scan times took on average 90-120
minutes for a 13.2-13.7 m cubic voxel side length tomogram.
A slice orthogonal to the flow direction from approximately the same location in the
fractured system can be seen in Figure 3 under no-stress, a low confining stress of 5 bar,
and a high confining stress of 120 bar. The plane parallel to the major fractures has a
much lower resistance to stress than those opposing the major fractures. This change in
overall sample geometry also caused issues with direct visual comparison as the overall
geometry of the system had changed between each change in system stress.
A)
B)

C)

Figure 3: Orthogonal x-y slices from midsection of fracture system tomograms with
middle segmentation value overlaid in red.
Fluid flow occurred in the direction of into the
page. A) Dry Scan at 13.7x13.7x13.7 m3, B)
Low confining pressure at 13.2x13.2x13.2 m3,
and C) High confining pressure at
13.2x13.2x13.2 m3. Scale bar in red is 5000
m in length.

Sample permeabilities were also recorded with changes in system stress with a varying
pressure drop across the core. Finally, system permeabilities and tomograms were
recorded for decreasing effective stress, or rising pore pressure. The permeability results
for each different system stress can be seen in Figure 4. Initially confining fluid pressure
was increased in logarithmic intervals from 6 to 120 bar. Then, only the upstream
pressure was increased from 5 to 119 bar, with the maximum pressure selected to remain

SCA2015-023

6/12

just below the confining pressure of the system. Finally the downstream pressure was
increased in similar intervals until a final pressure of 118 bar was reached.
The effective pressure of the system is defined in this work as the confining pressure
minus the average of the inlet and outlet pore pressures. Error bars are reported as
discrepancies found between upstream and downstream measurements, with minimal
contribution to error with the in terms of the volume resolution of the pumps. One set of
error bars on the highest stress data point for the rising confining pressure data series was
omitted due to non-realistic flows found in the downstream pump. This non-realistic flow
behaviour may have been due to either compression of gases trapped in the pore space or
thermal variations within experimental environment.

Figure 4: Changes in permeability of the fractured system due to changes in system stress.

The initial changes in permeability with increased confining pressure are expected as the
overall aperture of the observed fractures was seen to decrease as the effective pressure of
the system increased. During increasing upstream pressure, and overall differential
pressure, permeability was seen to initially decrease with increasing pressure drop across
the sample. This decrease in permeability may be caused by further narrowing of
downstream fractures under the new stress of the high upstream pressure. There also exist
several cross bedding fractures within the sample that may have been closed, decreasing
permeability, due to the new stress applied from the increased inlet pressure.
Finally, while raising the downstream pressure of the sample, a recovery of the samples
initial permeability is seen. This shows that the decrease in permeability recorded for the
increasing pressure drop across the sample was not caused by fines migration. The final
recovered permeability in a stress state similar to that of the first recorded permeability is

SCA2015-023

7/12

expectedly lower, as some of the larger features holding the fracture open for fluid flow
may have been compacted over the course of the tests.
The fracture void space from each tomogram was then segmented from the rest of the
solid material. This segmentation was accomplished using a simple thresholding method
instead of a more advanced watershed based algorithm, since the selection of appropriate
phase label seeds within narrow near- or sub-resolution features proved to yield
erroneous results due to an insufficient gradient between phases. Label seeds would also
propagate inappropriately far into solid phase regions due to the lack of a significant
greyscale gradient between suspected void space and solid. To assess the significance of
the selected threshold value, three segmented images were created and used for flow
simulations using a subjective optimal threshold value and deviations 200.
This segmentation method was first applied to the dry scan images followed by the low
confining pressure, and then high confining pressure system tomographies. As the stress
applied to the system is increased, the aperture of the imaged fractures decreases and the
accuracy of the segmentation of the fracture void became increasingly challenging as
seen in Figure 3. Threshold segmentation was still found to yield the most representative
of fracture segmentations, but the inclusion of numerous regions of noise within the solid
regions of the sample significantly decreased the accuracy of the void space in each
tomogram.
These erroneous regions were largely eliminated from the final fracture segmentation by
computing the connected flow path along the flow direction of the sample using a
neighbourhood selection allowing segmented voxels to be considered connected across
edges, but not through vertexes. This analysis module then removes all disconnected
components that are smaller than 10 voxels in a given volume. This greatly reduces the
inclusion of non-connected, improperly segmented noise regions, but incorrect regions
near the connected fracture space remain problematic. Previous work has attempted to
tackle this issue of selecting a proper threshold value for features below image resolution
[15] with the inclusion of looking at the connectivity factor or Euler-Poincar number for
2D and 3D systems [16]. However, the distinction between erroneous regions attributed
to the void space segmentation and simply void space regions below image resolution
may have a significant effect and the analysis was not implemented here.
The high stress system proved to be beyond segmentation capabilities, as once a
threshold segmentation value was found that had connectivity through the sample in the
direction of flow, large errors were evident in the segmentation and excluded. A different
greyscale value was selected for each tomography as the reconstruction bit scaling was
different for each system based on how much overall resolvable fracture void space could
be observed in the final reconstruction, as well as known x-ray source issues.
The segmented fracture void space data sets from the zero and low effective pressure
conditions were processed using a Lattice-Boltzmann based flow simulation for the

SCA2015-023

8/12

determination of local velocity fields and single phase absolute permeability as shown in
Error! Reference source not found.. The calculations were performed by mirroring the
geometry about the flow face and using a constant uniform fluid body-force [17].
Experiment fluid flow and flow simulations were performed along the axial direction of
the sample, the direction into the page for the shown images in Figure 3.
The preferential fluid flow path for the most conservative threshold image of the dry scan
system can be seen in Figure 5 following the results of the fluid flow simulation. The
fluid flow path appears to be rather uniform throughout the fractured system flow
network, with slight cross-bedding fluid exchange via smaller interlayer fractures. The
two largest of fractures running through the sample appear to contribute the most to all
fluid flow through the sample.

Figure 5: 3D representation of single phase fluid flow through segmented fracture system. Fluid flow
occurred from the left front face to the rear right face of the extracted void space. Lighter colours
represent areas of higher flow velocity with the darker blue being areas of lower velocity.

The two major lateral fractures running through the entire length of the observed region
highlighted in Figure 5 were extracted for analysis compared to the known solution for a
system of two infinite parallel plates.
The flow rate between parallel plates is given by [18]:

3
=

12

(1)

Where Q is the volumetric flow rate, l is the length normal to flow, P is the pressure
difference across the length of the system, is the distance between the two plates, is
the fluid viscosity, and L is the length of the sample in the direction of flow. Then by

SCA2015-023

9/12

equating with Darcys Law and solving for permeability, the following relation can be
found:
3
(2)
=
12
Where k is the sample permeability, and A being the sample area normal to flow.
Predicted permeabilities for both the upper and lower major fractures as seen in the fluid
flow representation are presented below in Table 1 based on this calculation using the
measured mean aperture from the original greyscale images. Crack apertures are a
calculated mean value measured from greyscale images across 6 different orthogonal
slices along the length of the sample.
Table 1: Predicted permeability of fractured system using parallel plate model and LatticeBoltzmann flow simulations compared to measured values. Note: The values presented for the flow
simulation permeability calculations are based on the range of threshold values used for fracture
segmentation.

System
Top
Crack
Dry
Scan Bottom
Crack
Top
Crack
Low
Stress Bottom
Crack

Combined
Flow
Crack
Crack
Predicted
Simulation
Measured
Aperature Permeability Permeability Permeability Permeability
(m)
(Darcy)
(Darcy)
(Darcy)
(Darcy)
101.5

10.16
14.7

77.2

4.46

23.8

0.14
0.71

38.4

15.3-44.6

0.63-2.52

N/A

0.001

0.57

These results show that the resulting permeabilities from the fluid flow simulations and
those predicted by the parallel plate model correspond quite well. However, the predicted
permeability at the low stress condition was much higher than that measured in the
experiment. This was due to the fracture width being below the voxel size of the
tomograms. To obtain a permeability of 1 mDarcy, the estimated aperture width of a
planar fracture is 3.7 m, well below the 13.2 m voxel size of the scan. The fracture
segmented from the scan has a width of at least one voxel, much too large to correctly
represent the actual fracture system.
The effect of the matrix between fractures was assumed to be negligible, although fluid
flow experiments on intact samples could not be completed as no unfractured plugs could
be obtained. This assumption is supported by considering the orders of magnitude
difference that exists between the imaged fracture apertures and the first major pore size
peak appearing near 100 nm as seen in Figure 1. The lower values of measured
permeability approaching the 1 Darcy magnitude may be approaching that of the intact

SCA2015-023

10/12

matrix, and then would need to consider the effect of flow within the porous matrix
structure.
A technique was then developed to measure qualitatively changes in overall fracture
aperture along the length of the sample, making use of changes within the greyscale
values for features below the resolvable limit. The mean greyscale value for the entire
volume of interest (660x635x600 m3) at each effective pressure was calculated and used
to normalize the mean grey value for each slice in the plane orthogonal to flow along the
length of the sample. Fluid flow occurs in direction of increasing slice number, or left to
right in Figure 6. This analysis gives a qualitative analysis of sample void space for both
above and below resolvable fracture features.
1.006
Initial System

Normalized Mean Value

1.005

Max Confining Pressure

1.004

Maximum Pressure Drop

1.003

Max Pore Pressure

1.002
1.001
1
0.999
0.998
0.997
0

100

200

300

400

500

600

Slice #
Figure 6: Normalized mean grey scale value of x-y plane orthogonal slices along length of sample for
different stress systems.

The initial profile shows a high level of fluctuations with each advancing slice. These are
caused by the large open and fully resolved fractures observed, which contribute a large
portion of the potential fluid flow path. The profile at maximum confining pressure
shows the homogenisation along with a smaller total spread of values caused by the
closure of the fracture system. The maximum pressure drop data set shows the widening
of upstream fractures with the slight dip seen in the left hand side of the figure compared
to that of the maximum confining stress system, but with still a highly closed downstream
portion. The final maximum pore pressure profile was similar to the maximum stress
system, but has noticeably higher amounts of void space in the initial upstream and final
downstream portions of the sample leading to the higher permeabilities. This final state
also shows the unrecoverable change in system geometry following compaction by not
returning to the level of fluctuations seen initially in the system.

SCA2015-023

11/12

The overall U shape of the final curves may be a result of beam hardening or lack of
uniform x-ray delivery from the source cone shape leading to slight deviations from an
overall mean value. This shape may also have been an effect of the compression in the
axial direction of the sample caused by the unrestrained fluid inlets of the core-holder
assembly, dissipating with depth into the fractured sample.

CONCLUSIONS AND FUTURE WORK

The measurement of fractured system permeability under varying systems stresses was
successfully recorded in a hassler type flow cell. Fractured sample permeability was
found to decrease with increasing confining pressure under a constant, small pressure
gradient. The near complete recovery of initial permeability was found to be possible by
raising fluid pore pressure to the same initial effective pressure state. Full 3D x-ray
micro-tomograms were collected after each change of system stress, and used to show
qualitative changes in fracture void space. The fractured system was observed to exhibit
no major change in permeability under a rising pressure gradient, despite evidence
showing the widening of initial upstream fracture apertures. Upon raising pore pressure to
reduce the effective pressure to close to the initial value, the sample recovered most of its
initial permeability, with a slight decrease thought to be due to sample compaction. The
limitations of system resolution for imaging the entire diameter of the sample limited the
effectiveness of segmenting the fracture void space. The actual fracture aperture for even
the lowest confining pressure was estimated to be well below the voxel size of the CT
scan, and the image segmentation technique therefore overestimated the fracture width
subsequently used in the Lattice-Boltzmann simulations.

ACKNOWLEDGEMENTS
The authors would like to acknowledge the support of the Qatar Carbonates and
Carbon Storage Research Centre (QCCSRC) for the funding of this project, as well as
Shell, Qatar Petroleum, and Qatar Science and Technology Park for the funding of
this program. The authors would also like to acknowledge to fellow PhD candidate
Farrel Gray for his contribution of the Lattice-Boltzmann fluid flow simulation
results.

REFERENCES
1. S. M. Benson and D. R. Cole, "CO2 sequestration in deep sedimentary formations,"
Elements, vol. 4, pp. 325-331, 2008.
2. F. K. North, Petroleum Geology, 2nd ed. Winchester, Mass. 01890, USA: Allen &
Unwin, 1985.
3 K. Edlmann, S. Haszeldine, and C. McDermott, "Experimental investigation into the
sealing capability of naturally fractured shale caprocks to supercritical carbon dioxide
flow," Environmental earth sciences, vol. 70, pp. 3393-3409, 2013.
4 A. Busch and A. Amann-Hildenbrand, "Predicting capillarity of mudrocks," Marine
and Petroleum Geology, vol. 45, pp. 208-223, 2013.

SCA2015-023

12/12

5 A. Amann, M. Waschbsch, P. Bertier, A. Busch, B. M. Krooss, and R. Littke,

"Sealing rock characteristics under the influence of CO2," Energy Procedia, vol. 4,
pp. 5170-5177, 2011.
6 A. Hildenbrand, S. Schlmer, and B. Krooss, "Gas breakthrough experiments on
finegrained sedimentary rocks," Geofluids, vol. 2, pp. 3-23, 2002.
7 P. Egermann, J. Lombard, and P. Bretonnier, "A fast and accurate method to measure
threshold capillary pressure of caprocks under representative conditions," SCA2006
A, vol. 46, 2006.
8 N. J. Welch, J. P. Crawshaw, and E. S. Boek, "COMPARISON OF THE CHANGE
OF CAP ROCK PERMEABILITY AND CAPILLARY ENTRY PRESSURE WITH
VARYING EFFECTIVE STRESS," SCA2014, 2014.
9 S. Li, M. Dong, Z. Li, S. Huang, H. Qing, and E. Nickel, "Gas breakthrough pressure
for hydrocarbon reservoir seal rocks: implications for the security of longterm CO2
storage in the Weyburn field," Geofluids, vol. 5, pp. 326-334, 2005.
10 P. Boulin, P. Bretonnier, V. Vassil, A. Samouillet, M. Fleury, and J. Lombard, "Entry
Pressure Measurements Using Three Unconventional Experimental Methods,"
presented at the International Symposium of the Society of Core Analysts, Austin,
TX, USA, 2011.
11 A. RP40, "Recommended practices for core analysis," Feb, 1998.
12 M. Andrew, B. Bijeljic, and M. J. Blunt, "Pore-scale contact angle measurements at
reservoir conditions using X-ray microtomography," Advances in Water Resources,
vol. 68, pp. 24-31, 2014.
13 V. S. Group, "Avizo," 9.0 ed: FEI 2015.
14 A. Buades, B. Coll, and J.-M. Morel, "A non-local algorithm for image denoising," in
Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer
Society Conference on, 2005, pp. 60-65.
15 A. Samoulian, H.-J. Vogel, and O. Ippisch, "Upscaling hydraulic conductivity based
on the topology of the sub-scale structure," Advances in water resources, vol. 30, pp.
1179-1189, 2007.
16 H.-J. Vogel, "Topological characterization of porous media," in Morphology of
condensed matter, ed: Springer, 2002, pp. 75-92.
17 S. M. K. Shah, J. P. Crawshaw, O. Gharbi, E. S. Boek, and J. Yang, "Predicting
Porosity and Permeability of Carbonate Rocks From Core-Scale to Pore-Scale Using
Medical CT Confocal Laser Scanning Microscopy and Micro CT," in SPE Annual
Technical Conference and Exhibition, 2013.
18 R. W. Fox, A. T. McDonald, and P. J. Pritchard, Introduction to Fluid Mechanics 8th
ed.: John Wiley & Sons, Inc., 2010.

SCA2015-024

1/12

ESTIMATING SATURATIONS IN ORGANIC SHALES

USING 2D NMR
Benjamin NICOT1, Nopparat VORAPALAWUT2, Berengere ROUSSEAU1, Luis F.
MADARIAGA1, Gerald HAMON1, Jean-Pierre KORB 3
1
Total E&P, Pau, France, 2 WellStaff, Pau, France
3
PMC, Ecole Polytechnique-CNRS, Palaiseau, France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Tight organic shales are a challenge to the oil industry in several ways. An accurate
petrophysical evaluation of these rocks includes porosity, permeability, hydrocarbon
saturation and fracability. The evaluation of these petrophysical parameters has appeared
more challenging in tight organic shales than in conventional reservoirs. In this paper, we
describe a non destructive method allowing the measurement of hydrocarbon saturation
based on two dimensional T1-T2 NMR.
First, we give the experimental evidence of an NMR contrast between oil and water in
organic shales. Contrary to the conventional reservoirs, the contrast between oil and
water in shales is not based on diffusion, but on T1/T2. Various imbibition tests with
water/light oil/D2O were performed. These tests prove unambiguously that the oil and
water NMR signals can be assigned unambiguously in 2D T1-T2 NMR maps. They also
prove that the high T1/T2 in organic pores is not due to bitumen (high viscosity), and that
it can be achieved by light oil (isopar L). The high T1/T2 observed is only due to
confinement in the organic pores. In order to understand how confinement only can lead
to such a high T1/T2, multi-frequency NMR dispersion (NMRD) experiments were
performed. These experiments allow us to propose an interpretation that explains the
unexpected dynamical behavior of the light oil in organic pore leading to high T1/T2.
Hydrocarbon saturation in organic shales is commonly measured by Dean Stark, or retort
method. Saturation is the ratio between a volume of fluid and a total pore volume. If
thermal method may be good at estimating the volume of fluid produced, its major
drawback is that it is destructive, therefore preventing the direct measurement of total
pore volume. The total pore volume is then estimated by adding water, oil and gas
volumes: this leads to an accumulation of errors in the estimation of hydrocarbon
saturation in shales. The NMR method being non destructive, it allows performing a total
porosity measurement afterwards; therefore providing a much more accurate saturation
estimation. The porosity estimation being challenged for the thermal method, we
compare water volumes measured by NMR and a thermal method for validating the
NMR approach.

SCA2015-024

2/12

INTRODUCTION
Petrophysical properties including porosity, permeability, fracability and hydrocarbon
saturation are crucial for hydrocarbon producibility and economic assessment in tight
organic shales. Therefore, a technique to evaluate these petrophysical properties in tight
organic shales has become more challenging. It has been demonstrated extensively that
porosity measurement could have significant bias in these rocks [1-2]. Therefore, a
proper experimental protocol is a required to obtain usable data. The saturation is the
ratio between fluids (water or hydrocarbon) volume and a total pore volume. Standard
methods for evaluating water saturation are Dean Stark and retort fluid extraction. Dean
Stark can be time consuming while the Retort method can be relatively fast. However, the
hydrocarbon saturation estimated by retort method may be erroneous because the total
pore volume is a summation of gas, water and oil volumes [3]. Nuclear magnetic
resonance (NMR) is a non destructive method that is commonly used in the industry to
provide the porosity. Advanced NMR techniques such as 2D D-T2 maps are also
commonly used to estimate saturations in conventional reservoirs. There is an effort to
apply this technique in unconventional rocks [6] and they have shown the difficulty to
measure diffusion coefficients in nanoporous materials. Recently, 2D T1-T2 NMR has
been used by different authors to characterize the fluids contained in shale samples [5-9].
To summarize the finding of these different authors: (1) oil exhibits a high T1/T2, (2)
water exhibits a low T1/T2, (3) bitumen is identified by Singer P.M. et al. [7] and
Washburn K.E. et al [8] on their samples (Eagle Ford and Green River formation) and,
(4) bitumen is not found in samples used by Korb J-P et al. [10].
Here, we use 2D T1-T2 NMR to estimate water and hydrocarbon saturation in tight
organic shales. A contrast between water and oil signals in 2D T1-T2 maps is proved by
spontaneous imbibitions test in cleaned and as received shales by various liquids, i.e.
heavy water (D2O), light oil (isopar L) and water (70 g/L NaCl). Moreover, dynamic
behavior of light hydrocarbon confined in organic pore leading to high T1/T2 is proposed.
This model is confirmed by multi-frequency NMR dispersion (NMRD) and 2D T1-T2
maps at 2.5 MHz and 23 MHz. We therefore propose a comprehensive approach,
combining experiment and theory, allowing us to estimate accurately water and oil
saturations by NMR. The saturation from 2D T1-T2 NMR is then compared to gravimetric
method for the imbibitions test and by Thermo Gravimetric Analysis coupled to a mass
spectrometer (TGA) for 15 shale samples (as received) from the Vaca Muerta formation
in Argentina.

EXPERIMENTAL
Our samples were source rocks from the Vaca Muerta formation in Argentina. We
worked with two different wells, shale A which comes from the oil window, and shale B
which comes from the wet gas window.
On these samples we performed different types of measurements: NMR, mass balance
and Thermo Gravimetric Analysis (TGA):

SCA2015-024

3/12

2D T1-T2 maps were acquired at 2.5 MHz and 23 MHz on Oxford Instrument
spectrometers with TE = 200 S and inversion recovery varying from 70 s to 1 s in
200 values. The results were processed using an in-house 2D inverse Laplace
transform. The temperature of the samples was 21 1 C. 2.5 MHz measurements
were performed on cylindrical samples of 30 mm diameter by 50 mm height, while
the samples for 23 MHz measurements were 10 mm in diameter by 15 mm height.
Multi-frequency NMR dispersion (NMRD) was performed on a fast field cycling
spectrometer from Sterlar s.r.l., Mede, Italy. The measurements were performed on
samples of 9mm in diameter and 15mm height. At each frequency (varying from
10kHz to 35MHz), a full T1 measurement is performed, and processed using an inhouse 1D inverse Laplace transform.
A TGA instrument (NETZSCH-STA 449) coupled with mass spectrometer (AeolosOMS 403 D) was used. Thermo gravimetric analysis (TGA) is an analytical technique
used for accurately measuring mass changes as a function of increasing temperature,
while the MS is used to analyze the produced gases. The 10 x 15 mm samples were
placed in the TG furnace and heated under helium gas. The samples were heated from
25 C to 315 C at 5 C/min heating rate and held at this temperature for 30 min.
Final temperature of 315 C was chosen after previous tests in order to evaporate all
the water and the mobile oil, while minimizing the cracking of kerogen and the
production of structural water. This temperature is consistent with retort method for
total oil extraction [3-4]. The water molecule at mass 18 was monitored by mass
spectrometer. An integrated area of water mass peak allows us quantifying the water
mass. The oil mass was calculated by mass subtraction between total mass loss and
water mass. Samples used for TGA were 10 mm in diameter by 15 mm.

SPONTANEOUS IMBIBITION TESTS: EVIDENCE OF AN NMR

CONTRAST BETWEEN WATER AND OIL
Samples of 30 x 50 mm were used for imbibition tests. Firstly, shale A followed the
following experimental procedure (detailed in Figure 1):
As Received NMR measurement (Fig. 1a)
Cleaning by chloroform for 4 days and iso-propanol for 4 days.
Drying at 60 C for 1 day.
Spontaneous imbibition by water (70 g/L NaCl) (Fig. 1b and Fig. 2 )
Cleaning 4 days isopropanol and drying (60C)
Spontaneous imbibition by mineral oil (isopar L) (Fig. 1c and Fig. 2 )
Figure 1 shows the 2D T1-T2 maps for shale A: as received (a), cleaned and imbibed by
water (b), and cleaned and imbibed by isopar L (c). In as received shale A (Fig. 1a), we
noticed two separated signals. We have already shown [10] that the signal at low T1 and
T2 was water and the other elongated signal with T1/T2 varying from 5 to 10 was oil.
During imbibition by water, the signal at low T1 and T2 increased. On the contrary,
during imbibition by isopar L, it was the signal with high T1/T2 that increased.

SCA2015-024

4/12

Fig. 1: 2D T1-T2 2.5 MHz NMR for shale A: as received (a), cleaned/dried and imbibed by water (b) and
cleaned/dried and imbibed by isopar L (c). The diagonal lines correspond to T 1/T2=1 and T1/T2=2.

Fig.2: 2.5 MHz T1 and T2 projections of 2D T1-T2 maps for cleaned shale A imbibed by water (Fig 1b) and
isopar L (Fig 1c).

Figure 2 shows T1 and T2 projections of the 2D T1-T2 maps obtained for shale A during
spontaneous imbibitions by water (left) and isopar L (right). During imbibition by water,
the T2 distribution is shifted to longer T2s, while the T1 distribution seems stable.
Basically such a net difference can be explained by the basic relaxation features of T2 and
T1. Intrinsically, T2 is dominated by the slow surface dynamical contributions driven by
the dipolar fluctuations at zero Larmor frequency, while T1 is only due to the much faster

SCA2015-024

5/12

volume dynamical contributions driven by the dipolar fluctuations at the Larmor

frequency. T2 is thus more sensitive than T1 to the slow surface relaxation processes.
Owing to the biphasic fast exchange conditions, the evolution of the T2 distribution to
longer T2s thus reveals the progressive enhancement of the volume contributions in the
relaxation. On the contrary, the T1 distribution which is less sensitive to the surface
contributions mainly reveals the enhancement of the volume contributions with the
imbibition processes.
One can wonder if these results from spontaneous imbibitions are affected by the partial
saturation of the rock i.e. by the presence of gas in the sample. To answer that question
we performed NMR T1-T2 maps on another sample of shale A, and compared the results
for as received and 100% saturated with water and isopar L (Fig. 3). Fluid saturations
were performed by vacuuming the sample for 4 days, then saturating it with the fluid
under 100 bars of pressure.
First, the as received NMR T1-T2 map (Fig. 3a) of sample 2 is very similar to the one
obtained for sample 1 (Fig. 1a). Second, when saturating the shale sample with 100%
water (Fig 3b), we notice that water penetrates pores that were not penetrated under
spontaneous imbibition, and exhibit an elongated NMR signal along the T1/T2=1-2.
Third, when saturating the shale with 100% isopar L (Fig 3c), we notice that the NMR
signature is very similar to the one obtained after spontaneous imbibition of isopar L (Fig
1c).
One can interpret that water spontaneously imbibes only in the mineral porosity,
corresponding to the signal observed on Figure 1b (rounded signal at T1/T2=1-2). When
water is forced into the organic porosity, it appears as an elongated signal at T 1/T2=1-2
(Fig. 3b). For isopar L, it always exhibits an elongated signal at high T1/T2.

Fig.3 2D T1-T2 2.5 MHz NMR for shale A: as received (a), 100 % water saturation (b) and 100 % isopar L
saturation (c).

SCA2015-024

6/12

Additionally, we confirm the contrast between water and oil signal in 2D T1-T2 maps by
other imbibition tests: shale A (sample 3) was submitted to imbibition tests directly from
the as received state. The 2D T1-T2 maps for as received shale A imbibed by D2O (for 9
hours) and by isopar L (for 8 hours) are presented in Figure 4. The 2D T1-T2 map of the
as received sample is very similar to the one presented in Figure 1a. In fact, most of our
samples exhibit such 2D T1-T2 contrast between oil and water in the as received state.
The imbibition test by D2O (D2O has no NMR signal) shows that the signal of water at
low T1 and T2 has disappeared. However, the resolution of the oil signal (elongated signal
at high T1/T2) seems to be lost. After 9 hours of D2O imbibitions, isopar L was then
imbibed. During oil imbibition, it is clear that only the oil signal (elongated signal at high
T1/T2) increased.

Fig. 4: 2D T1-T2 2.5 MHz NMR for as received shale A (a) and imbibed by D2O (b) and isopar L (c).

Figure 5 shows T1 and T2 projections of the 2D T1-T2 maps for as received shale A
imbibed by D2O and isopar L. It obviously shows a continuous decrease of the signal at
low T1 and T2 when imbibed by D2O. During D2O imbibition, the peak of the T2
distribution is slightly shifted to longer T2 whereas the T1 distribution remains centered
en the same value. This was already observed in Figure 2 during H2O imbibition. These
different behaviors in the 1H T2 and T1 distributions have already been explained above
on the basis of the origin of the 1H T2 and T1 relaxation processes. The progressive
imbibition by the heavy water (with quadrupolar 2H nuclei) also explains the net decrease
of the 1H T1 and T2 intensities.
Moreover, T1 and T2 projections for isopar L imbibitions present only an increase of oil
signal at high T1 and T2. The water signal at low T1 and T2 was stable.

SCA2015-024

7/12

Fig. 5: 2.5 MHz T1 and T2 projections for as received shale A imbibed by D2O (Fig 4b) and isopar L (Fig
4c).

Isopar L could imbibe very fast into both cleaned and as received shales. Particularly in
the case of as received shale, an equivalent of 8.1 p.u. of isopar L imbibed the rock, and
only air was expelled from the sample, no water was produced. This demonstrates that
isopar L imbibed pores that were not accessible to water. This could be explained by a
well connected network of strongly oil-wet organic pores.
Moreover, from isopar L (light oil) imbibitions test in cleaned shale A, we still observed
a high T1/T2 for oil signal. Therefore, we proved that such a high T1/T2 is not due to
bitumen (as reported in Green River formation [7]) but due to light oil confined in
organic pore.

UNDERSTANDING THE NMR CONTRAST BETWEEN OIL AND

WATER IN 2D T1-T2 MAPS
Using NMR 2D T1-T2 maps at 2.5 MHz, we have identified a contrast between water and
oil. The water appears at short T1 and T2 with T1/T2 ~1, while oil exhibit a surprisingly
high T1/T2 ratio. In order to understand why such a high T1/T2 ratio can be achieved with
light oil like isopar L we performed multi-frequency NMR dispersion (NMRD).
At each frequency, a T1 measurement is performed and the T1 distribution is calculated.
The T1 distribution appears bimodal at all the frequencies. The peak corresponding to the
water signal is identified by performing the same experiment on a 100% water saturated
sample. A typical multi-frequency NMR dispersion plot is made by plotting the
relaxation rates R1 (R1_oil=1/ T1peak_oil and R1_water=1/ T1peak_water).

SCA2015-024

8/12

The multi-frequency NMR dispersion obtained on an as received sample of shale A

(Figure 6) clearly shows a very different frequency behavior for oil and water: we
observe a very strong frequency dependence of the relaxation of the oil (red squares)
compared to the one for the water (blue circles). We have proposed theoretical models
[10] for oil (red dash line) and water (blue solid line) allowing us to fit the experimental
data. The details of the theory are not the purpose of this article and are explained in
details in ref [10].

Fig. 6: Multi-frequency NMR dispersion plot showing the relaxation rates R1=1/T1 as a function of the
Larmor frequency for as received shale A. (Figure taken from [10]).

We have applied our relaxation theory for T1 and T2 relaxation for the two different
cases. We found a constant value of T1/T2 ~1.36 for water. While we found T1/T2 for oil
varies with frequency. According to this theory, we display in Figure 7 the evolution of
the ratio T1oil/T2oil with frequency, showing that the T1/T2 ratio for oil is expected to
increase linearly with the square root of the frequency.

T1/T2

100

10

1
0.001

0.01

0.1

Frequency (MHz)

10

100

1000

Fig. 7: T1oil/T2oil versus frequency, showing that the T 1/T2 for oil is expected to increase with frequency.

Since the oil-water contrast we observed in the previous section is based on the T1/T2
ratio, this suggests that this contrast should increase with frequency. This led us to
investigate higher frequencies and to perform NMR measurements at 23MHz.

SCA2015-024

9/12

Figure 8 shows 2D T1-T2 maps obtained at 2.5 MHz and 23 MHz for cleaned shale A
imbibed by isopar L and as received shale B. We observe that the T1/T2 ratio for oil
(isopar L) obtained at 2.5 MHz is around 4 to 5. On the other hand the T1/T2 ratio for oil
obtained at 23 MHz is the order of 10. The theoretical predictions T1/T2 for oil (Figure 7)
are in excellent agreement with the experimental values found: T1/T2([email protected])~4-5
while T1/T2(oil@23MHz)~10.

Fig.8: 2D T1-T2 2.5 MHz and 23 MHz NMR for cleaned shale A imbibed by isopar L (a) and as received
shale B (b).

High T1/T2 is usually attributed to the presence of bitumen. In this case we prove that a
signal with very high T1/T2 ratio can be obtained only by putting light oil (isopar L)
inside a nanoporous system. On the other hand, the high T1/T2 ratio observed on as
received samples (Fig. 1a) completely disappeared after cleaning with chloroform (Fig.
1b). While the definition of what is called bitumen is still highly debated [11], this
confirms that the high T1/T2 ratio comes from a fluid that is soluble in chloroform. While
previous authors observed bitumen at high T1/T2 ratio[7] we demonstrate that in our case,
high T1/T2 ratio is only due to light oil in a high confinement. We also show that after
cleaning and imbibing with isopar L (Fig. 1c) the signal at low T1 and T2 has
disappeared; this proves that our samples do not contain bitumen as described in [7].
In other words:
The T1/T2 for water is almost constant with frequency.
The T1/T2 for oil varies strongly with frequency.
Therefore, the higher the frequency the stronger the contrast between oil and
water in 2D T1-T2 maps.
We thus recommend using 23 MHz NMR 2D T1-T2 data in order to measure the quantity
of water and oil contained in a given sample. The next section aims at comparing water
and oil volumes obtained by NMR and other techniques (mass balance and TGA).

SCA2015-024

10/12

COMPARISON OF WATER AND OIL VOLUMES BETWEEN NMR

AND OTHER TECHNIQUES
Several imbibition tests have been performed on shale A samples and were reported in
the previous section. For two imbibition test (water imbibition from Fig. 2 and isopar L
imbibition from Fig. 5), the amount of fluid imbibed was monitored using NMR and also
by weighting the sample (mass balance). In Figure 9, we compared the water and isopar
L imbibed volume calculated from 2D T1-T2 2.5 MHz NMR with mass balance. Water
and oil volumes obtained from 2D T1-T2 maps have been calculated by integrating the
NMR signal attributed to each fluid in the 2D T1-T2 map. The excellent agreement
between NMR and mass balance measured volumes for both water and oil shows that 2D
T1-T2 NMR could be an accurate technique to estimate volumes of fluids confined in
porous media.
NMR cumulative volume, cm 3

2.0
1.6
1.2
0.8

Isopar L
0.4

Water

0.0
0.0

0.4

0.8

1.2

1.6

2.0

Gravimetric cumulative volume, cm 3

Fig. 9: Cross plot of water and isopar L volumes obtained from gravimetric and 2D T 1-T2 NMR methods.

The performance of the NMR approach was tested versus TGA on a set of as received
samples from a different well (shale B). The water and oil volume of 15 as received
samples from shale B obtained from 2D T1-T2 23 MHz NMR were compared with TGA.
The 2D T1-T2 maps at 23 MHz acquired before and after TGA measurement are shown in
Figure 10. Before TGA (Fig. 10a) the 2D T1-T2 map is characteristic of a shale sample
containing water (low T1/T2) and oil (high T1/T2). After TGA (Fig. 10b) the measured
NMR porosity is less than 1 p.u., meaning that water and oil were almost totally lost after
TGA analysis. This could confirm that high T1/T2 in our samples is not due to bitumen
because boiling point of bitumen is higher than 315 C [12].
Since TGA measures the mass loss, the density of water and oil is needed in order to
convert the mass to volume unit. We used the following values for the density: 1.05
g/cm3 for water (70 kppm NaCl brine) and 0.85 g/cm3 for oil (light crude oil). The cross
plot between volumes calculated from TGA and 2D T1-T2 NMR is presented in Figure
11a. The water, oil and total volumes obtained from 2D T1-T2 NMR and TGA are in very
good agreement. In order to estimate saturations, one needs to measure porosity. TGA
being a destructive measurement, we had to measure porosity on a set of twin samples.
15 twin samples were saturated by 100 % water to measure total porosity. Then, water

SCA2015-024

11/12

and oil saturation were estimated by dividing fluid volume with total pore volume. We
present the water and oil saturation for 15 as received shales in Figure 11b.

Fig. 10: 2D T1-T2 maps at 23 MHz for as received shale B before (a) and after (b) TGA analysis.

Fig. 11: Comparison of the results obtained by TGA and 23MHz NMR: a) volumes and b) saturations.

CONCLUSION
We proposed the 2D T1-T2 NMR technique as a non destructive technique to estimate the
water and oil saturation in tight organic shales. We interpreted the contrast between water
and oil in 2D T1-T2 maps on the basis of excellent agreement between experimental and
theoretical arguments (multi-frequency NMR dispersion). According to our theory of
NMRD in shales, oil and water are undergoing completely different dynamics, leading to
a high T1/T2 ratio for oil, and a low T1/T2 ratio for water. We also showed in our samples
that such a high T1/T2 ratio is not due to bitumen but comes from light oil (isopar L)
confined in organic pore. The T1/T2 ratio of oil depends on the frequency whereas T1/T2
of water is invariant with frequency. Therefore, using 23 MHz NMR water and oil signals
could be easily separated in 2D T1-T2 maps. The water and oil volumes obtained from 2D
T1-T2 maps were compared with Thermo Gravimetric Analysis. The fluid volumes
obtained from both techniques are in excellent agreement. Furthermore, the volume of

SCA2015-024

12/12

imbibed isopar L from 2D T1-T2 maps is also correlated well with gravimetric method
(mass balance). The water and oil saturation is calculated by dividing water or oil volume
with total pore volume. Because NMR is a non destructive method, the total pore volume
could be directly determined after 2D T1-T2 maps acquisition. In our case, because the
thermal method is destructive, the total pore volume was determined by T 2 measurement
of 100 % water saturated twin samples. If NMR only was performed, one could even
measure saturation and porosity (NMR acquisitions on as received and 100% water
saturated sample) on the same sample. This would be a strong advantage compared to
commonly used thermal methods (retort). Therefore, our proposed technique could give
more accurate determination of water and oil saturation in tight organic shales.

ACKNOWLEDGEMENTS
The authors would like to thank J.F. Perrette and Gregory Dubes for their help.

REFERENCES
1. Sondergeld C.H., Newsham K.E., Comisky J.T., Rice M.C., Rai C.S.,
Petrophysical Considerations in Evaluating and Producing Shale Gas
Resources, SPE, (2010), 131768.
2. Le Bihan A., Nicot B., Marie K., Thbault V., Hamon G., Quality Control of
Porosity and Saturation Measurements on Source Rocks, SPWLA, (2014).
3. Handwerger, D. A., Keller, J., & Vaughn, K. Improved Petrophysical Core
Measurements on Tight Shale Reservoirs Using Retort and Crushed Samples,
SPE, (2011), 147456
4. Handwerger D.A., Willberg D., Pagels M., Rowland B., Keller J.F., Reconciling
Retort versus Dean Stark Measurements on Tight Shales, SPE, (2012), 159976.
5. Kausik R., Cao Minh C., Zielinski L., Vissapragada B., Akkurt R., Song Y., Liu
C., Jones S., Blair E., Characterization of Gas Dynamics in Kerogen Nanopores
by NMR, SPE, (2011), 147198.
6. Ozen A.E., Sigal R.F., T1/T2 NMR Surface Relaxation Ratio for Hydrocarbons
and Brines in Contact with Mature Organic-Shale Reservoir Rocks,
Petrophysics, (2013), 54 no. 1, pp. 11-19.
7. Singer P.M., Rylander E., Jiang T., McLin R., Lewis R.E., Sinclair S.M., 1D and
2D NMR Core-Log Integration in Organic Shale, SCA, (2013), 018.
8. Washburn K.E., Birdwell J.E., A New Laboratory Approach to Shale Analysis
Using NMR Relaxometry, SPE, (2013), 168798.
9. Fleury, M. Characterization of Shales with Low Field NMR, SCA, (2014), 014
10. Korb J-P., Nicot B., Louis-Joseph A., Bubici S., Ferrante G., Dynamics and
Wettability of Oil and Water in Oil Shales, The Journal of Physical Chemistry C
(2014), 118, pp. 23212-23218.
11. Landis C.R., Castano J.R., Maturation and Bulk Chemical Properties of a suite of
solid hydrocarbons, Org. Geochem., (1995), 22 no. 1, pp. 137-149.
12. Etherington J.R., McDonald I.R., Is Bitumen a Petroleum Reserve?, SPE,
(2004), 90242.

SCA2015-025

1/12

COMPARISON OF PORE SIZE DISTRIBUTION BY NMR

RELAXATION AND NMR CRYOPOROMETRY IN
SHALES
M. Fleury1, R. Fabre1, J.B.W. Webber2
1

IFP Energies nouvelles, 1 avenue de Bois-Prau, 92852 Rueil-Malmaison, France.

Lab-Tools Ltd, Lakesview International Business Park, Hersden, Canterbury, Kent CT3 4JZ,
UK

This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
The characterization of shales is challenging due to their very small pore sizes. In a
previous paper, we have shown that Nuclear Magnetic Resonance relaxation techniques
are well suited for detecting and quantifying water in nanopores. However, the
distribution of relaxation times does not necessarily represent the distribution of pore
sizes in the sample due to pore diffusive coupling. Indeed, all pore sizes below the
diffusion length will be averaged around a single relaxation time representing an average
volume to an average surface ratio. We illustrate the above-mentioned effects by
comparing NMR relaxation and cryoporometry data on the same sample. The NMR
cryoporometry experiment relies on the shift of the melting temperature of the saturating
liquid, which is itself a function of pore size according to Gibbs-Thomson theory. The
experiments were performed first on a model system made of a mixture of two calibrated
porous glass (15 and 139 nm) and secondly on a shale. On the model system, the
cryoporometry results are in quantitative agreement with the expectations but the T2
distribution is strongly modified by pore coupling and only the 15nm pore size population
can be identified. Pore coupling is also clearly identified by T2-store-T2 experiments. On
the shale sample, the pore size distribution obtained from cryoporometry experiments is
quasi uniform from 2 up to 100 nm, while the T2 distribution is log-normal and relatively
narrow. The disagreement between these two results is also explained by pore coupling.

INTRODUCTION
The characterization of shales requires revisiting petrophysical techniques. For example
when considering porosity, strong disagreement between standard techniques can result if
inappropriate protocols are used [1]. For the pore size distribution, the mercury injection
technique may also fail to capture the very small range of pore sizes below 6nm. Also,
the sample drying can interfere with the PICP measurement in the presence of clays. For
this point of view, NMR techniques requiring saturated samples are non-disturbing and
well suited for pore sizes down to nanometers. For example we have shown that NMR
relaxation techniques are adequate for detecting and quantifying water in the interlayer of
smectites [2].

SCA2015-025

2/12

For the range of pore size considered, 1nm up to 1m, other techniques routinely used in
the field of catalysis are available. Gas adsorption and thermo-porometry are some
examples. NMR cryoporometry is not new [3] but has not been widely used although it
has several advantages. In low field systems, it only requires a device to control the
temperature of the sample very accurately. Very few results have been published so far
about this technique for characterizing shales and we believe it has a great potential.
In this work, we emphasize that standard T2 distribution cannot always capture the true
pore size distribution due to strong pore coupling effects. These effects are ideally
demonstrated by comparing NMR cryoporometry and relaxation results on the same
sample. For this purpose, we used calibrated porous glasses as a model system, and a
shale sample from a tight clayey formation.

BACKGROUND: NMR CRYOPOROMETRY

NMR cryoporometry is based on the shift of the freezing or melting point of the fluid
saturating the porous media according to:

() =
(1)

where kGT is the so-called Gibbs-Thomson constant and x is the pore size. The above
equation has been established rigorously in the case of a cylinder, hence x may
correspond to the diameter of this cylinder. However, when the shape is undefined, the
pore size x will rather correspond to a curvature V/S in the melting cycle [4]. In this
work, we only consider the melting cycle to avoid hysteresis problems and to allow a
large pore size to be explored given the capabilities of our instrument. If one detects the
amount of liquid volume V(T) present in the sample at a function of temperature T, one
can obtain the pore size distribution P(x) from the derivative of the V(T) curve using the
following equation:

(2)
() =
= 2 .

Since pore size distributions are usually plotted using logarithmic scales, it is more useful
to plot the following distribution:

(3)
() =
=
.
()

in a similar way as in mercury intrusion experiments for example.
The constant kGT actually contains complex and multiple physical phenomena but it has
been shown that coherent values are obtained on a series of model materials characterized
by other experimental techniques, such as gas adsorption [5]. In this work we will take
the value kGT=58.2 K.nm for water. With this value and the range of temperatures
available in our instrument, the range of pore sizes that can be explored is about 2 to 600

SCA2015-025

3/12

nm (Figure 1). For the upper value, we assumed that a temperature resolution of 0.1C is
achieved. Due to a 1/x sensitivity, the method has a non-uniform resolution (excellent
around 10 nm and potentially very poor around 500nm) if the temperature is increased
linearly, as explained later.
Water

10

X: -0.1
Y: 582
2

size (nm)

10

10

X: -30
Y: 1.94
0

10
-30

-25

-20
-15
-10
Tm variation (C)

-5

Figure 1: range of pore sizes that can be explored with water considering a variation of the melting
temperature Tm from 0 down to -30C.

A very common interrogation is whether the pore size is modified during the freezing
cycle due to the formation of ice with a smaller density. NMR cryoporometry is actually
only one branch out of many cryoporometry techniques and freezing-melting cycles are
used routinely for example in Differential Scanning Calorimetry (DSC) techniques
without any damage to samples. A key aspect is certainly to avoid freezing by nucleation
and this can be achieved by surrounding the sample with bulk water and performing a
freezing-melting-freezing cycle prior to the main melting cryoporometry experiment,
without melting the bulk phase at the second step [4]. Also, the ice formed in pores has a
different structure referred as plastic ice [6].

MATERIALS AND NMR METHODS

NMR Relaxation Methods
The experiments have been carried out on a Maran Ultra proton spectrometer from
Oxford Instruments with a proton Larmor frequency of 23.7 MHz. Beside standard
CPMG measurements, 2D T2-store-T2 exchange maps were measured using the
methodology presented elsewhere [7].
For the cryoporometry experiments, the liquid volume quantity is measured using a single
Hahn echo chosen at an appropriate echo spacing to avoid detecting ice. Indeed,
although normal ice has a T2 relaxation time of 8 s (and T1=70 s), some of the ice

SCA2015-025

4/12

formed in pores may have different characteristics. The value is also selected as a
function of the sample in order to maintain an adequate signal to noise ratio. For the
calibrated porous glass, =500s, and for the shale, =100s (i.e. the NMR amplitude is
taken at 1ms and 0.2ms respectively). Unfortunately, a CPMG sequence in which one
would choose any appropriate time cannot be used because the radio-frequency pulses
produce a significant and measurable heating of the sample, interfering with the imposed
temperature ramps.
NMR Cryoporometry Device
The cryoporometry device (Figure 2) designed by Lab-Tools Ltd is composed of a
special probe that can be inserted inside the magnets, replacing standard probes from the
manufacturer of the NMR instrument. In this probe, a glass tube of diameter 5mm o.d.
can be inserted, and the sample can have a maximum height of about 20 mm. This tube is
in close contact with a copper tube connected to a Peltier system for regulating very
precisely the temperature. Heat resulting from the cooling of the sample is evacuated
from the Peltier system using water circulating at a temperature of 5C. Temperature is
measured using thermocouples constantly calibrated with a reference bath at 0C. They
are placed at two locations : as closed as possible to the sample above the NMR antenna,
and close to the Peltier system. With a maximum current of 3A at 12V, the temperature
can decrease down to -31C. A software controlling the power supply allows
programming temperature ramps as small as 0.002 C/min. Such small ramps are only set
close to the bulk melting point and non-linear temperature ramps are generally used to
speed up experiments (Figure 3). Since a NMR measurement may take 2 minutes and
temperatures are also averaged over 2 minutes, the corresponding NMR signal may be
averaged over a temperature gradient of 0.004C. From this principle, a theoretical
resolution curve can be calculated as a function of pore size (Figure 3). For example, at
600 nm, the resolution is limited to about 11nm.

SCA2015-025

5/12

Volt meter
Temperature
controller

Reference
bath
0 C

Cooling
bath

Water

Power
supply

Control Temperature
Sample Thermocouple

Peltier current

NMR
Magnet

NMR
Magnet

Peltier
System

NMRC probe

NMR device

Figure 2: Schematic of the NMR cryoporometry probe and associated control components.

10
1

10

5
0

r (nm)

T (C)

-5
-10
-15
-20

10

-1

10

-25
-30
-35

10
15
20
Elapsed time (hr)

25

30

10

10
r (nm)

Figure 3: Typical temperature profiles for shale samples (left). Temperature increase is reduced to
0.002C/min close to the bulk melting point. Theoretical pore size resolution assuming a measuring time of
120 s (right).

Samples
We used a model system composed of calibrated porous glass (CPG from BAM, Federal
Institute for Materials Research and Testing, Germany) of mean pore sizes
d50=15.10.2nm (ERM-FD121, pore volume 624.6mm3/g) and d50=1393.7nm (ERMFD122, pore volume 924.4mm3/g) as measured by high pressure mercury injection
experiments. Taking a glass density of 2.5, the porosity of these medium are 25% and
37% respectively. The shape of the pores in these medium are tubular like. These
reference materials are delivered as grains of mean diameter 55m (ERM-FD121) and

SCA2015-025

6/12

90m (ERM-FD122). We made a grain pack from these materials by mixing 80.5 mg of
ERM-FD121 and 55.0mg of ERM-FD122 in a 5mm tube (Figure 4).
The shale sample originates from the Callovo-Oxfordian (COx) formation at a depth of
604m (EST433 borehole of Montiers-sur Saulx, France). At this depth of the COx
formation, a high clayey unit is present with carbonate nodules dispersed in the medium
(20%wt) playing a minor role in terms of porosity or permeability [8]. On a plug size, the
porosity measured by NMR and water permeability measured by steady-state were
respectively 14.8% and 7.3nD. We extracted from the plug a small cylinder of diameter
4mm and length 20mm for the cryoporometry experiments.

RESULTS
We show the results obtained on the above-mentioned samples. For the CPG medium, the
objectives are to test the cryoporometry device as well as comparing the measured pore
size distributions with NMR relaxation.
Calibrated Porous Glass
The cyoporometry experiments on the CPG mixture gave the expected results (Figure 5);
we observe 3 plateaus that correspond to the successive melting of the 15nm CPG, the
139nm CPG and finally the inter-grain phase surrounding the CPG grains. Each plateau
value allows calculating the fraction of each phase as detailed later. Then, the curve was
fitted with spline functions in order to extract the derivative and obtain the pore size
distribution, as detailed in equation 3. The pore sizes obtained fall very close to the values
given by the manufacturer using MICP measurements: 14 nm and 147 nm (Figure 6).
First, although we compare two measurements with different physical mechanisms, the
good agreement comes from the fact that both techniques use a cylinder shape for the
pores. Second, the pore size values obtained from cryoporometry depend (i) on the
constant kGT that can vary from one author to another, (ii) on the fine adjustment of the
bulk melting temperature which is not known very precisely (although thermocouples are
calibrated very accurately). The increase of temperature vs. time is a minor source of
uncertainty, as shown above. Qualitatively, the narrow width of the 15nm CPG medium
is in agreement with the MICP distribution given by the manufacturer (not shown), and
the width of the 139nm CPG is larger. Concerning the fraction of each population,
cryoporometry results are in good agreement with simple weight measurement of water
and CPG materials inserted in the tube (Table 1)
The T2 distribution measured for the CPG mixture at 30C is essentially bimodal (Figure
7). At first glance, the peak at T230ms corresponds to the 15nm population, whereas the
wider peak centered around T2220ms correspond to the 139nm population. However,
the inter-grain water is not identified and the tail at T2>1s correspond in fact to a small
amount of supernatant water at the top of the CPG grain pack in the tube (it was later
removed for performing the T2-store-T2 exchange experiments). Given the weak surface
relaxivity of this clean material (0.25 m/s), the inter-granular water is expected close

SCA2015-025

7/12

the bulk value. In fact, the T2 distribution is strongly influenced by pore coupling, i.e.
molecules are diffusing in and out of the porous glass and exchange with inter-granular
water. Hence, none of the peaks are representative of the different population fractions. A
strong evidence is given by the T2-store-T2 map (Figure 8) measured separately at a
temperature of 30C. On such map, pore coupling is unambiguously indicated by the
presence of off-diagonal peaks located symmetrically on both sides. The exchange time
was chosen here at 300ms but other smaller values give similar (but weaker) off-diagonal
peaks. However, the details of the exchanges cannot be known precisely. Taking a
diffusion coefficient of water of 0.5x 10-9 m2/s, the diffusion length at 300ms is about
30m. Therefore, inter-granular water can partially exchange with the porous grains for
each size. Most probably, exchanges between grains are also possible. Further evidence is
given by the T2 distribution measured at -0.15C (Figure 5) when the inter-granular water
is frozen but the 15nm and 139nm CPG are melted (Figure 9). We obtain in this case two
narrow peaks that are truly representative of the relaxation time and fraction of each
population. Focusing at the fraction of each population (Table 1), we retrieve the
appropriate values measured by cryoporometry and weight.
800

NMR echo amplitude (a.u.)

700
600
500
400
300
200
100
0
-7

Figure 4: Micro-CT image of the CPG mixture.

The diameter is 4 mm. Small grains: 15 nm,
large grains: 139nm

-6

-5

-4

-3
T (C)

-2

-1

Figure 5: Hahn echo amplitude at 1ms as a function of

temperature. The 3 plateaus correspond to the successive
melting of the 15nm CPG, the 139nm CPG and finally
the bulk phase surrounding the CPG grains around 0C.

SCA2015-025

8/12

35

4
3.5

30

14nm
3
25
A(T ) (a.u.)

2
147nm

1.5

dV/d(ln(x))

2.5

20
15
10

0.5
5

0
-0.5
0
10

10

10

0
0
10

10

10

x(nm)

Figure 6: Pore size distribution on the CPG mixture

obtained from the cryoporometry experiment (raw
data shown in Figure 5).

10
T2 (ms)

10

10

Figure 7: T2 distribution of the CPG mixture at

30C.

35
te=300ms =0.46416

30

10

A(T ) (a.u.)

T2 (ms)

25

10

20
15
10
5

10

10

10
T2 (ms)

10

Figure 8: T2-store-T2 exchange experiment t 30C

on the CPG mixture performed at an exchange time
te=300 ms.

0
0
10

10

10

10

T2 (ms)

Figure 9: T2 distribution obtained when the intergrain water is frozen at -0.15C, inhibiting diffusive
exchange. The two narrow peaks are located at
T2=20ms and 138ms.

Shale Sample
The cryoporometry experiment was performed on the shale sample using the largest
possible range of temperature (-31C). The measured pore size distribution (Figure 10) is
relatively uniform from 2 up to 100nm, in qualitative agreement with results obtained on
a Barnett shale [9]. Due to the low permeability of this sample (7 nD), pore sizes larger
than 100nm are not expected. As opposed to this large distribution, the T2 distribution

SCA2015-025

9/12

(Figure 11) has a more classical shape close to log-normal (note here that no shape is
imposed in the data processing). Compared to the model system, the T2 values are much
shorter due to a larger surface relaxivity (typically of the order of 5 m/s) but also due to
a much larger surface to volume ratio generated by pore size as small as 2 nm. The
strong disagreement with cryoporometry results can be explained by pore coupling as
discussed further later. We also measured the T2 distribution at several temperature (but
in a separate experiment) and found that the distribution is simply shifting as the
temperature is increasing and the largest pore sizes gradually melting. The small pores
are not described by the tail at small values inside the T2 distribution at full saturation.
0.3
18

0.25

16
-16C
14

0.2

-11C

A(T ) (a.u.)

dV/d(ln(x))

12

0.15
0.1

10

-6C

8
6

0.05

-1C

4
2

0
-0.05
0
10

10

10

0
-2
10

-1

10

10

10

x(nm)

Figure 10: Pore size distribution obtained on the

shale sample, cryoporometry technique.

Figure 11: T2 distribution obtained from relaxation

measurements (black curve at 30C) and at different
temperature during melting.

0.8
0.7

P * dS/dP

0.6
0.5
0.4
0.3
0.2
0.1
0
-3
10

-2

10

T2 (ms)

-1

10
10
Pore entry diametre ( m)

10

Figure 12: Pore size distribution obtained on the shale sample, high pressure mercury injection technique.

SCA2015-025

10/12

DISCUSSION
Pore size distribution deduced from NMR relaxation measurements relies on a diffusion
process of molecules interacting with the pore surface yielding the well-known
relationship with the surface to volume ratio. However, when molecules diffuse over
distances that are larger, the surface to volume ratio is averaged over several pores. For
standard reservoir samples in the mD or D permeability range, pore coupling is inherently
present in many NMR relaxation measurement, especially at high temperature, but to a
relatively small degree, except in certain cases. Hence, MICP measurements are often in
agreement with NMR T2 distributions and their direct comparison allows the calculation
of an apparent surface relaxivity. Indeed, very often, the different pore size populations
are grouped in well-defined regions of the porous media and therefore, the pore coupling
is very limited. For example in carbonates, microporosity may be due to intra-grain
porosity and these grains are large enough to avoid diffusive coupling. In shales, the pore
sizes are roughly 2 or 3 orders of magnitude smaller. The permeability is then reduced to
the nD range but this not the case for the diffusion coefficient. Indeed the latter primarily
depends on porosity at first order. Indeed, from several measurements of diffusion in
caprocks [10] and shales [11], we found that a reasonable estimation of the effective
diffusivity can be made using an Archie relationship Deff/Dm =m-1 where Dm is the
molecular diffusion coefficient of bulk water, with m varying between 1.8 and 2.5, as
observed for electrical measurements. Hence, the diffusion coefficient decreases roughly
linearly with porosity and the diffusion length (6Defft)1/2 calculated at the peak relaxation
time is still large. However, if the sample is heterogeneous at the micron scale (e.g. thin
laminations or organic matter channels), one may find large T2 distributions representing
indeed the different pore populations. The correct interpretation of T2 distributions in
shales must be complemented by high resolution visualization of the pore network
structure.
For the homogeneous shale sample studied here, the diffusion length at 0.7ms is about
1.4m when taking a diffusion coefficient of 0.5x 10-9 m2/s,. Since we have a pore size
distribution extending up to 100 nm, a strong coupling effect exist. Furthermore, in such
case, a T2-store-T2 experiment cannot further demonstrate the coupling because an
intermediate level of coupling is needed in such experiment (i.e. a bimodal or wide
distribution is needed). Also, as suggested during the review of this manuscript,
heterogeneous surface relaxivities within the porous media could also be a source of
discrepancy when comparing different pore size distribution measurements. This effect
would create larger T2 relaxation distribution and to some degree compensate for the
coupling effect. In the present situation the width of the distribution obtained by
cryoporometry is about 2 decades, whereas it is about 1 decade or less for T2 relaxation
time distribution.
Interestingly, the pore entry size distribution obtained by high pressure mercury injection
(HPMI) on a sister sample is relatively narrow with a peak at 40 nm, unlike the
cryoporometry result (Figure 12). Although we often observe a correlation between pore
size and pore throats, this correspondence is not necessarily true and is highly dependent

SCA2015-025

11/12

on the pore structure and geological history. At least, in terms of maximum size, the two
distributions agree: the largest throat size is about 100 nm, while the largest pore size is
about 300 nm (a ratio of 3 is the theoretical result for a sphere pack).

CONCLUSION
We show that pore size distribution obtained from NMR cryoporometry and relaxation
experiments on the same sample can strongly disagree. While the cryoporometry
technique is able to give accurate results from 2nm up to about 1m, T2 relaxation is
strongly influenced by pore coupling. As a result, the distribution obtained from
relaxation is representative on an average volume to surface ratio; in shales, this
distribution also depends on the length scales of the heterogeneities present in the porous
medium, compared to the diffusion length.

AKNOWLEDGEMENTS
We thank F. Norrant for providing the calibrated porous glasses, Elizabeth Rosenberg for
performing the microCT images, and J. Labaume for the HPMI experiments.

REFERENCES
1 B. Lalanne, A. Le-Bihan, E. Elias, R. Poyol, L. Martinez, How to Cope with some of
the Challenges Associated with Laboratory Measurements on Gas Shale Core
Samples, in: Proceeding Soc. Pet. Eng. SPE 167709, 2014: pp. 117.
2 M. Fleury, E. Kohler, F. Norrant, S. Gautier, J. MHamdi, L. Barr, Characterization
and Quantification of Water in Smectites with Low-Field NMR, J. Phys. Chem. C.
117 (2013) 45514560.
3 J. Strange, M. Rahman, E. Smith, Characterization of porous solids by NMR, Phys.
Rev. Lett. 71 (1993) 35893591. doi:10.1103/PhysRevLett.71.3589.
4 O. V. Petrov, I. Fur, NMR cryoporometry: Principles, applications and potential,
Prog.
Nucl.
Magn.
Reson.
Spectrosc.
54
(2009)
97122.
doi:10.1016/j.pnmrs.2008.06.001.
5. J. Mitchell, J. Webber, J. Strange, Nuclear magnetic resonance cryoporometry, Phys.
Rep. 461 (2008) 136. doi:10.1016/j.physrep.2008.02.001.
6 J.B.W. Webber, Studies of nano-structured liquids in confined geometries and at
surfaces, Prog. Nucl. Magn. Reson. Spectrosc. 56 (2010) 7893.
doi:10.1016/j.pnmrs.2009.09.001.
7 M. Fleury, J. Soualem, Quantitative analysis of diffusional pore coupling from T2store-T2 NMR experiments., J. Colloid Interface Sci. 336 (2009) 2509.
doi:10.1016/j.jcis.2009.03.051.
8 P. Cosenza, J.C. Robinet, D. Prt, E. Huret, M. Fleury, Y. Graud, et al., Indirect
estimation of the clay content of clay-rocks using acoustic measurements: New
insights from the Montiers-sur-Saulx deep borehole (Meuse, France), Mar. Pet. Geol.
(2013). doi:10.1016/j.marpetgeo.2013.07.004.
9 J.B.W. Webber, P. Corbett, K.T. Semple, U. Ogbonnaya, W.S. Teel, C. a. Masiello, et
al., An NMR study of porous rock and biochar containing organic material,

SCA2015-025

12/12

Microporous
Mesoporous
Mater.
178
(2013)
9498.
doi:10.1016/j.micromeso.2013.04.004.
10 M. Fleury, S. Gautier, N. Gland, P. Boulin, B. Norden, C. Schmidt-Hattenberger,
Advanced and integrated petrophysical characterization for CO2 storage: application
to the Ketzin site., Oil Gas Sci. Technol. (2013).
11 M. Fleury, Characterization of shales withlow field NMR, in: Proceeding SCA Annu.
Conf., 2014.

Table 1: Fraction in % of the different water populations from different methods (CPG sample).

Method
Weight
Cryoporometry
Relaxation
Relaxation (-0.15C)

15 nm
27.0
26.0
21.5
28.4

139nm
27.3
27.0
24.6

Inter-granular+bulk
45.6
47.0
88.5
-

SCA2015-026

1/11

MULTI-SCALE AND UPSCALING OF DIGITAL ROCK

PHYSICS WITH A MACHINE THAT CAN LEARN ABOUT
ROCKS
R. Sungkorn, A. Morcote, G. Carpio, G. Davalos, Y. Mu, A. Grader, N. Derzhi, J. Toelke
Ingrain Inc., 3733 Westheimer Rd., Houston, Texas, 77027, U.S.A.
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Digital rock physics (DRP) is becoming a standard tool for rock characterization. DRP
utilizes 2D and 3D digital images of rock samples to analyze petrophysical and
geological properties. The ability to apply DRP to a large rock sample opens a way for
economic exploration and recovery of hydrocarbon. Nevertheless, due to the well-known
multi-scale nature of rocks and limitations in imaging technology, less than 1% by
volume of a rock sample will be digitally acquired and analyzed. Undoubtedly, relevancy
and representativeness of DRP remain hotly debated topics in oil and gas industry.
Machine learning (ML) has recently accelerated advances in many industries. ML brings
together multiple disciplines such as computer science, statistics, and natural science to
create algorithms that can learn from data. DRP can harness the power of ML to learn
from its data, the digital image of rocks, to generate breakthroughs in the oil and gas
industry.
In this paper, we present a framework that combines advances in DRP and ML to
characterize rock samples at a large scale, not only a tiny part of it. The framework is
based on an understanding that a rock consists of multi-scale rock fabrics intermixed
spatially. These rock fabrics are captured as groups of patterns within a digital image
when they are smaller than the image resolution being used. We developed ML
algorithms that can automatically learn about rock fabrics and their patterns. This
learning process can be iteratively repeated down to an image resolution that resolves the
smallest or the most significant rock fabrics. Thus, the framework integrates DRP
paradigm to achieve a truly multi-scale analysis. Also, DRP and ML analysis determine
the optimum number and optimum locations for further acquisition and analysis of rock
fabrics at a higher resolution.

INTRODUCTION
Rocks are well-known to inherit complex heterogeneous structures with a broad spectrum
of scales. For example, pores within a rock sample can range in size from nanometers to
millimeters. Ehrenberg [1] carried out laboratory measurements of porosity and
permeability of the same rock at different sample scales; the smaller plug samples were

SCA2015-026

2/11

drilled from the larger cores. The study found that the larger core samples had generally
lower porosity but higher measured permeability values than the plugs. It can be implied
that small-scale rock samples do not adequately represent all features found in large-scale
samples. In addition, it is rarely feasible to perform laboratory measurements of largescale rock samples and, measurements of small-scale rock samples are typically limited
to a small amount of samples due to extensive time and cost.
Digital rock physics (DRP) aims at providing qualitative and quantitative understanding
of flow transport units as well as geometrical properties of rocks. Some of the rock
properties are extremely difficult, if not impossible, to measure in the laboratory. Thus,
DRP in conjunction with laboratory measurements, will compliment and complete well
log analysis with not only detailed information but also with new kinds of insights. Such
well logs enhance analysis of reservoirs and will open a way for economic exploration
and better recovery of hydrocarbons.
The use of DRP involves three steps: (a) digital imaging to create a digital representation
of a rock in 2D and 3D at a scale and resolution that will resolve rock features such as
pores, organics, and grains; (b) digital image processing to categorize pixels/voxels in 2D
and 3D respectively, with similar properties, and (c) digital rock analysis to digitally
model desired rock properties using the digital image of the rock [2, 3, 4].
The following discussions are applicable for both 2D and 3D images. For the sake of
simplicity, the term image refers to 2D and 3D images and pixel refers to both image
pixel and voxel, unless otherwise stated.
Scientists and researchers have been analysing rock properties using multi-scale DRP [5]
[6, 7, 8, 9]. Figure 1 shows on the left a typical multi-scale DRP paradigm and on the
right the DRP paradigm introduced in this paper. The main differences are the use of rock
fabrics instead of rock features and the recursive process to obtain information from the
small-scale rock fabrics. A rock fabric is defined as a combination of rock features.
Similar rock fabrics have similar properties or follow similar property trends.
The process begins with an image of a large rock sample acquired at a relatively coarse
resolution to cover a large field of view. At this stage, rock fabrics larger than the image
resolution are resolved while smaller ones are unresolved. A rock fabric is considered
resolved when it is represented, in every direction, by at least two pixels. Then, the
unresolved rock fabrics are segmented into groups. Information concerning the
unresolved rock fabrics is analyzed from additional images acquired at a finer resolution
and smaller field of view. The information from resolved and unresolved rock fabrics are
fused and populated back into the large-scale image. DRP analysis, of desired properties,
is carried out using the large-scale image.

SCA2015-026

3/11

Multi-scale DRP provides a promising method to characterize rocks. Nevertheless, due to

limitations in imaging technology, currently available multi-scale DRP methods still
suffer from shortcomings. They can be roughly summarized as follow:
Image scale: despite advances in image processing and imaging technology, the size
of a large rock sample is still limited, at best, to several centimeters and often only
to a few millimeters range. Properties derived from a large sample provide higher
reliability and relevancy.
Image resolution: in rocks, pore size may span several orders of magnitudes. For
complex rocks, such as shales small scale pores play an important role in transport
properties as well as in the total porosity of the rock. Therefore, an image resolution
adequately small to resolve these small pores should be used. However, such image
will have a drastic reduction in the field of view.
Lemmens & Richards [10] created an impressive high-resolution, 12 mm in length 2DSEM image. 12800 images where stitched and tiled together. It resolved pores from
millimeter to nanometer-scale. Their approach accurately provides detailed rock
properties with extensive time and resources requirements. Set aside the practical
aspect in 2D imaging, the approach is undoubtedly infeasible for 3D imaging.
Representative elementary area/volume (REV): REV is defined as the size of subsamples in which a measured property is approximately independent from location.
The definition is arguably invalid when a rock has a mild level of heterogeneity.
Unresolved rock fabrics sampling locations: unresolved rock fabrics are divided into
groups of similar properties, e.g. pixel intensity and CT number. Images of these
groups are acquired at a higher-image resolution to determine their properties. The
image locations are typically chosen manually and qualitatively. However, it is
extremely difficult to make a reasonable and consistent selection in 3D due to complex
process in human perception of volume from texture [11].
Recent advances in computing hardware and machine learning (ML) have accelerated
innovations and breakthroughs in many industries. ML brings together multiple
disciplines such as computer science, statistics, and natural science to create algorithms
that can learn from data. These algorithms have the ability to build a model from data
and/or training data without strict instructions. Detailed discussion regarding ML can be
found in [12] and [13]. Examples of ML-based computer vision applications include
autonomous vehicle technology [14], automatic tumor detection [15], and object
recognition [16]. Digital images produced in DRP can be also considered as data. Based
on this perspective, DRP can harness the power of ML to discover and learn from its
data.
In the computer vision community, a texture is loosely defined by complex visual
patterns formed by distinct features. Such features can be extracted using various
mathematical models such as intensity histogram, co-occurrence matrix, and Gabor filter.
Detailed discussion regarding texture analysis can be found in [17].

SCA2015-026

4/11

A paradigm for texture analysis usually involves four steps: (a) keypoint detection to limit
the analysis only to meaningful areas, (b) feature extraction to quantitatively represent
texture using appropriate models, (c) feature classification defines clusters of keypoints
corresponding to a perceptually homogeneous texture, and (d) texture segmentation to
construct area/volume based on the feature clusters. Texture in computer vision is similar
to the rock fabrics in DRP. Therefore, the texture analysis paradigm can be used with
some modifications to discover rock fabrics in DRP.

Figure 1: Left: generalized paradigm for multi-scale digital rock physics. Right: generalized paradigm for
multi-scale digital rock physics using machine learning. Steps are highlighted according to the tasks shown
in the bottom. Dotted block indicates a nested digital rock physics paradigm for unresolved rock features.

ROCK IMAGING
Rocks inherit complex multi-scale heterogeneous structures. A variety of imaging and
detection techniques have been used to gain insights into rocks. Ideally, the image
resolution being used should resolve all significant rock features and provide a
reasonably large field of view (i.e. image scale). Due to limitations in imaging technology
both image resolution and image scale are overly compromised.
Figure 2a shows a schematic image with multi-scaled objects. Overlaying the objects, we
have grid cells. Large objects encompass significant amount of grid cells and will be
resolved. In contrast, small objects are significantly smaller than the grid cells and will be
unresolved. Figure 2b shows a digitized representation of Figure 2a image. The gray

SCA2015-026

5/11

scale (i.e. intensity) of the grid cells directly relates to the object area covered by the grid
cell. A grid cell has low intensity (dark gray) when completely covering a feature while
high intensity (light gray) is presented in a grid cell that covers only a tiny part of a
feature. It can be seen that larger objects are fairly well represented by digitized grid
cells. On the other hand, small objects are smeared out in digitized grid cells and are
unrecognizable.
Nevertheless, digitized grid cells are generated based on the interaction between an object
and the physics of the imaging technique being used. A group of unresolved grid cells
contains information regarding patterns of unresolved objects. A sophisticated
mathematical model which quantifies patterns of grid cells intensity, e.g. co-occurrence
matrix, local binary patterns and Gabor filter, must be used [15] [17]. Since averaged
intensity of these patterns is approximately similar, the widely used averaging approach
cannot distinguish them. Figure 3 shows an example of multi-scale rock image. A largescale rock image, in the center, contains unresolved rock fabrics which have different
image intensity patterns. On the sides, we have high-resolution images

Figure 2: Illustration of a comparison between image resolution (grid size) with resolved rock features on
the left of (a) and unresolved rock features on the right of (a). Illustration similar rock features acquired and
digitized (b).

Figure 3: An example of unresolved rock fabrics in a large-scale image (a). Image of the unresolved rock
fabrics are acquire at higher resolution (b) and (c). Image courtesy of ADCO Ltd.

SCA2015-026

6/11

Figure 4: Eagle Ford sample. Segmented rock fabric overview (left) with optimum size and number of
areas to capture all rock fabrics (right).

MULTI-SCALE DRP WITH ML

DRP has been improved significantly in recent years. Variations of multi-scale DRP
(Figure 1a) are widely adopted [5] [6] [7] [8] [9]. However, applications of DRP for
reservoir characterization are still limited due to shortcomings discussed earlier. In this
section, we discuss a scalable multi-scale DRP paradigm with ML techniques for rocks.
The differences from previous multi-scale DRP are the use of ML to learn about resolved
and unresolved rock fabrics presented in a rock sample. We also use ML to identify
optimum size and location for further analysis of the unresolved rock fabrics.
The present multi-scale DRP paradigm (Figure 1b) begins with a digital imaging of a
rock sample at a large scale. This image will be called overview throughout this section.
The overview (Figure 4, right) might contain resolved and/or unresolved rock fabrics.
In the second step, rock fabrics (Figure 4, left) in the overview are detected and
segmented using the texture analysis method discussed earlier. However, rock fabrics are
different from image texture commonly encountered in computer vision. The main
difference is that rock fabrics tend to have pattern at individual pixel level not at edge or
blob level [13] [14] [15] [16] [17].
We developed a novel rock fabric analysis method, based on the texture analysis
paradigm (Figure 1b). Rock fabrics key-points are detected using the method discussed in
Appendix A. They are mostly located within an area with a rock fabric.
Then, rock fabric features of the area around the keypoints are computed using the
method discussed in Appendix B. In this method, four rock fabric features or attributes
are used: contrast, homogeneity, entropy, and variance.

SCA2015-026

7/11

Consequently, keypoints are clustered using the four rock fabric features. We developed
an unsupervised ML method for clustering high-dimensional data. It automatically learns
data and finds an appropriate number of clusters. Based on the understanding of rock
images, it is reasonable to postulate that keypoints within similar clusters have similar
rock fabrics. For example, in Figure 5, green represents areas with medium to high
intensities (unresolved organics and pores) while magenta represents areas only with
medium intensities (might be unresolved organics or very small pores). These clusters
can be used as a model for segmentation of the whole overview. During segmentation,
rock fabric attributes for every pixel/voxel are computed. Distance from clusters is
computed using, for example, Euclidean and Mahalanobis distance. A general discussion
regarding data clustering can be found in [12].

Figure 5: Close-up view of overview image (left) and its corresponding segmented fabrics (right).

The goal of rock fabric segmentation is to gain knowledge about unresolved rock fabrics.
Detailed information of the unresolved fabrics is needed in order to characterize the
large-scale image. This can be done by sub-sampling unresolved rock fabrics. An
optimum amount of sub-sampling is desired to minimize expenses while keeping high
level of accuracy and reliability. We developed an optimization algorithm for spatial data
analysis which it determines the most suitable locations for further analysis (Figure 4).
The algorithm finds a combination of areas that contains all fabrics with broadest variety
within the fabric. Information from sub-sampling areas can be used in fusion of multiscale information later.
An illustration of an extraction of multi-scale correlation extraction is shown in Figure 7.
We give an example of 2D porosity upscaling in this paper. The method can be used
directly to upscale properties in 2D and 3D. The suggested area within an overview
image is acquired at a resolution that adequately resolved rock features and segmented
into phases (e.g. pore and organic matters). A multi-scale correlation is obtained by
correlating, for each fabric, the intensity of the overview image pixels to the porosity
obtained from the area cover by the pixel in the high-resolution image. Examples of the
extraction are shown as plots in Figure 7. A multi-scale correlation for each fabric is

SCA2015-026

8/11

derived by fitting a function to the data (shown as solid lines in the plots). Then, the
porosity is upscaled by populating information from the multi-scale correlations for each
fabric to all pixels in the overview image (Figure 7).
A truly multi-scale DRP is achieved by repeating the procedure above recursively on the
unresolved rock fabrics until all rock fabrics are resolved. A method designed for
upscaling and fusion of multi-scale rock properties is discussed in [9]. It is noted that, the
multi-scale DRP with ML technology discussed here can be directly applied to multidimensional rock images. We use 2D images in this paper only for the sake of simplicity.
It is also worth to note that, the methods present in this paper are implemented using
graphics processing unit (GPU), which results in a computational time of approximately
100 seconds for a 2D image with a dimension of 2000 x 3000 pixels.

Figure 7: Illustration of multi-scale correlation extraction based on fabrics (bottom right) and population of
the multi-scale correlation back on the overview image (bottom left).

CONCLUSION
We present a multi-scale digital rock physics (DRP) method using machine learning
(ML) for rock fabrics characterization and scaling of rock properties. The method
integrates knowledge in geology, physics, and computer science. We developed new
algorithms based on the ones used in computer vision and pattern recognition
communities. The rock fabric analysis discovers rock fabrics both resolved and
unresolved by the image resolution being used. Also, it has the capability to characterize
large-scale rock images by iteratively learning about fabrics.

SCA2015-026

9/11

Additionally, it can be applied for selecting optimum number of meaningful

areas/volumes for laboratory measurements using whole core and cuttings.
It is also important to note that, despite the use of unsupervised ML methods, we
designed the method based on the concept of intelligence augmentation (IA) [21]. Thus,
experts can integrate their knowledge into the analysis to maximize benefits.

REFERENCES
1. Ehrenberg, S.N., Whole core versus plugs: Scale dependence of porosity and
permeability measurements in platform carbonates, AAPG Bulletin, (2007) 91,
pp. 835-846.
2. Dvorkin, J., Derzhi, N., Diaz, E., Fang, Q., Relevance of computational rock
physics, Geophysics, (2011) 76, pp. E141-E153.
3. Andr et al, Digital rock physics benchmarks Part I: Imaging and
segmentation, Computers & Geosciences, (2013) 50, pp. 25-32.
4. Andr et al., Digital rock physics benchmarks Part II: Computing effective
properties, Computers & Geosciences, (2013) 50, pp. 33-43.
5. ren P.E., Bakke, S., Process based reconstruction of sandstones and prediction
of transport properties, Transport in Porous Media, (2002) 46, pp. 311-343.
6. Grader, A.S., Clark, A.B.S., Al-Dayyani, T., Nur, A., Computations of porosity
and permeability of sparic carbonate using multi-scale CT images, Society of
Core Analysis International Symposium 2009, SCA2009-10.
7. Sok, R.M., Vaslot, T., Ghous, A., Latham, S., Sheppard, A.P., Knackstedt, M.A.,
Pore scale characterization of carbonates at multiple scales: Integration of microCT, BSEM and FIBSEM, Society of Core Analysis International Symposium
2009, SCA2009-18.
8. Khalili, A.D., Arns, C.H., Arns, J.-Y., Hussain, F., Cinar, Y., Pinczewski, W.V.,
Permeability upscaling for carbonates from the pore-scale using multi-scale
Xray-CT images, SPE/EAGE European Unconventional Resources Conference
and Exhibition 2012, pp. 606-622.
9. Sungkorn, R., Gundepalli, V.A., Mu, Y., Grader, A., Sisk, C., Bhakta, S.,
Dvorkin, J., Toelke, J., Multiscale permeability trends using digital rock
physics, Society of Core Analysis International Symposium 2014, SCA2014-43.
10. Lemmens, H., Richards, D., Multiscale imaging of shale samples in the scanning
electron microscope, in Camp, W., Diaz, E., Wawak, B., Electron microscopy of
shale hydrocarbon reservoirs: AAPG Memoir, (2013) 102, pp. 27-35.
11. Tory, M., Mller, T., Human factors in visualization research, IEEE
Transactions on Visualization and Computer Graphics, (2004) 10, pp. 1-13.
12. Bishop, C.M., Pattern Recognition and Machine Learning, Springer, Cambridge
U.K., (2006).
13. Bengio, Y., Learning deep architectures for AI, Foundations and Trends in
Machine Learning, (2009) 1, pp. 1-127.

SCA2015-026

10/11

14. Li, Q., Chen, L., Li, M., Shaw, S.-L., Nchter, A., A sensor-fusion drivableregion and lane-detection system for autonomous vehicle navigation in
challenging road scenarios, IEEE Transactions on Vehicular Technology, (2014)
63, pp. 540-555.
15. Menze et al., The multimodal brain tumor image segmentation benchmark
(BRATS), IEEE Transactions on Medical Imaging, (2014), pp. 33.
16. He, K., Zhang, X., Ren, S., Sun, J., Delving deep into Rectifiers: Surpassing
human-level performance on ImageNet Classification, arXiv:1406.4729, (2015).
17. Materka, A., Strzelecki, M., Texture Analysis Methods A Review, Technical
University of Lodz, Institute of Electronics, COST B11 report, Brussels, (1998).
18. Lowe, D.G., Distinctive image features from scale-invariant keypoints,
International Journal of Computer Vision, (2004) 60, pp. 91-110.
19. Yang, J., Yu, K., Gong, Y., Huang, T., Linear spatial pyramid matching using
sparse coding for image classification, Proceedings of CVPR, (2009).
20. Haralick, R.M., Statistical and structural approaches to texture, Proc. IEEE,
(1979) 67, pp. 1786-804.
21. Licklider, J.C.R., Man-computer symbiosis, IRE Transactions on Human
Factors in Electronics, (1960) 1, pp. 4-11.

APPENDIX
A. Rock keypoint detection
Natural rock images inherit highly heterogeneous arrangements of rock features.
Dividing images into tiles leads to an overestimation and classification of rock fabrics. In
order to (1) limit further analysis only to meaningful areas, (2) keep computational
expense tractable and (3) obtain reasonable number and classification of rock fabrics,
keypoints are detected. Concepts of keypoint detection are widely used in computer
vision community [18] [19]. Since rock fabrics have different characteristics from
commonly encountered image texture, previous methods for keypoint detection are not
applicable. Our rock keypoint detection algorithm (step b in Figure A1) begins with (a)
discrete wavelet decomposition of the image up to desired level. This step ensures that
rock fabrics at multiple scales will be captured. Then, (b) pixel/voxel gradient of images
obtained from previous step is computed. (c) Laplacian of Gaussian (LoG) is computed
on the gradient images to locate points of variation of pixel/voxel intensity. (d) Keypoints
detection within rock fabrics are detected by locating maxima in LoG images. Keypoints
at the edges of rock features are eliminated by limiting keypoints within certain value of
maxima (e.g. 80% of maxima).
B. Rock fabric features
There are models for image texture available in literature [17]. It is known that features of
rock fabrics, especially the unresolved rock fabrics, are in pixel/voxel level. Additionally,
similar rock fabric may have different orientations in an image. Therefore, a model that
quantitatively describes rock fabric features and is rotation invariance must be selected.
Note that, for the sake of naming consistency, rock fabric features (in DRP) are used
interchangeably with texture features (in computer vision). We use Haralick texture

SCA2015-026

11/11

features [20] which based on gray-level co-occurrence matrix (GLCM). They can be used
to quantify spatial distribution and auto-correlation of pixel/voxel pairs. The GLCM, Pi,j,
is constructed from probability of intensity j next to intensity i in defined directions and
distance. In this paper, we select an appropriate set of Haralick texture features to obtain
maximum separation between rock fabrics (for following features classification) namely,
(1) contrast, (2) homogeneity, (3) entropy, and (4) variance. Their mathematical
description can be found in [20].

Figure A1: Flow diagram for rock fabric recognition and segmentation.

SCA2015-027

1/12

A MULTI-SCALE, IMAGE-BASED PORE NETWORK

MODELING APPROACH TO SIMULATE TWO-PHASE
FLOW IN HETEROGENEOUS ROCKS
1

Tom Bultreys1, Wesley De Boever1, Luc Van Hoorebeke2 and Veerle Cnudde1
UGCT/PProGRess, Dept. of Geology and Soil Science, Ghent University, Krijgslaan
281 (S8), 9000, Ghent, Belgium
2
UGCT/Radiation Physics, Dept. of physics and astronomy, Ghent University,
Proeftuinstraat 86, 9000 Ghent, Belgium

This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Despite the large interest in the multi-phase flow properties of rocks with broad pore size
distributions, most digital rock physics approaches struggle with the presence of multiple
pore scales. In this work, we present a method to estimate relative permeability (Kr) and
resistivity index (RI) curves of such heterogeneous rocks during drainage. In our dual
pore network model (DPNM), macropores are represented as pores and throats , while
unresolved microporosity is treated as a continuous porous medium. The scales are
coupled by including microporosity as symbolic network elements in the DPNM, based
on 3D image analysis. The validity of the method is investigated by treating two
carbonate rocks (Estaillades and Savonnires limestone). We present a sensitivity
analysis of the drainage behaviour of these networks on the microporositys petrophysical
properties, which are provided as input. While a number of challenges persist, the
presented examples show how DPNM can help increase the understanding of two-phase
flow in complex carbonate rocks.

INTRODUCTION
Pore scale modeling is becoming an important tool to study how multi-phase transport
properties of core samples depend on the rocks microstructure and on the rock-fluid
system (e.g. wettability). Thanks to the recent developments in imaging techniques like
micro-computed tomography (micro-CT) [1], the pore space of many rock types can be
imaged in 3D. In pore network modeling, a simplified network of pores and throats is
extracted from these images. For low capillary numbers, two-phase flow can then be
simulated on the pore network model (PNM) by applying quasi-static invasion
percolation. The loss in detail caused by the simplification of pore space geometry and
fluid mechanics is compensated by the methods high computational efficiency. Despite
the large interest in image-based pore scale modeling, modeling multi-phase flow in
heterogeneous materials (e.g. many carbonates and clay-bearing sandstones) with broad
pore sizes remains problematic. The difficulties are mostly related to the sample
size/resolution trade-off in both imaging and modeling. PNM are the most likely
candidate for this type of modeling, due to their high computational efficiency and their

SCA2015-027

2/12

infinite resolution. In some dual pore network modeling (DPNM) approaches, individual
pores from the different scales are coupled into one multi-scale network [2,3]. This
approach offers a detailed description of the microporosity present, but quickly becomes
computationally infeasible when pore sizes span several orders of magnitude. In contrast,
the sub-resolution porosity can be treated as a continuous porous medium, characterized
by upscaled properties (e.g. [4]). In the method presented here, we use the continuum
approach for the microporosity to generate image-based multi-scale networks which
allow simulations even when the macroporosity does not percolate.

MATERIALS AND METHODS

We present simulations performed on two monomineralic, calcitic French limestones.
Estaillades limestone has a porosity of about 25 %, which contains intergranular
macropores and intra-granular micropores in fossil grains [5]. Savonnires limestone is a
layered oolithic limestone belonging to the Oolithe Vacuolaire, a stratigraphical unit
which also includes Brauvilliers stone (among other varieties) [6]. Its pore structure
contains 4 types of porosity: inter-granular and intra-oolithic microporosity, and intergranular and micro-connected macroporosity (hollow ooliths) [7]. The hollow ooliths are
only connected to the other macro-pores by microporosity in the oolith-shells. Depending
on local variations, the permeability ranges from 115 mD to more than 2000 mD (based
on TinyPerm mini-permeameter measurements), and porosity ranges from 22% to 41%.
Imaging
The pore space of an Estaillades sample of 7 mm and a Savonnires sample of 6 mm
diameter were imaged with UGCTs HECTOR micro-CT scanner (developed in
collaboration with XRE, Belgium) [8]. The former sample was scanned at an accelerating
high voltage of 100 kV and a target power of 10W. 2800 projections (2s illumination
each) were reconstructed with Octopus Reconstruction software (Inside Matters,
Belgium). The scan had a voxel size of 3.1 m. A 1000x1000x1001 section of the scan
was treated with a binormal noise filter and segmented into 3 phases (pore, solid and
microporous voxels) with a hysteresis thresholding algorithm in Octopus Analysis. The
Savonnires sample was scanned at an accelerating voltage of 120 kV and a target power
of 10W. 2400 projections (illumination time 1s) were acquired, resulting in a scan with
3.8 m voxel size. A 1000x1000x1000 subsection was segmented with a gradient
watershed thresholding algorithm in Avizo (FEI, France), after performing an anisotropic
diffusion filter.
DPNM Extraction
To extract a DPNM, a single-scale PNM was first extracted from the macroporous voxels
with a maximal ball algorithm [9]. Then, microporous voxels in the micro-CT image
were clustered into connected regions. Any two nodes in the macropore network which
touch the same microporous cluster were connected by adding a special network element,
henceforth called micro-link. The multi-scale PNM thus consist out of three types of
network elements: macroporous nodes, macroporous links and micro-links.

SCA2015-027

3/12

For each micro-link, the contact surface area of the microporous region with the
connected macropores is calculated by applying a marching cubes (MC) algorithm on the
3D three-phase image. Micro-link lengths are defined as the corresponding pore-to-pore
distances. Figure 1 illustrates the imaging and network extraction workflow. Note that
due to the applied network extraction method, microporous connections both parallel and
in series to macroporous links are automatically taken into account. This is important, as
serial microporous conductivity (dominant in rocks with pore-filling microporosity due to
e.g. clay) was found to have different effects on the transport properties than parallel
microporous conductivity (dominant in rocks with grain-filling microporosity due to e.g.
partial grain dissolution) [10].
DPNM Simulations
To calculate capillary pressure (Pc) curves, Kr-curves and RI-curves on multi-scale
networks, we modified the quasi-static network simulator developed at Imperial College
[11]. This model performs an invasion-percolation algorithm to simulate the pore filling
sequence during drainage, while neglecting the influence of viscous forces. At several
points during this invasion simulation, the flowrates of both fluid phases are calculated
separately by imposing a pressure difference over the network and assuming mass
conservation of the fluids in each pore. This leads to a set of linear equations which can
be solved for the pressure:
Pi Pj
qij g ij
Lij
(1)
qij 0, i
j

With qij the flow from pore i to pore j, gij the electrical or hydraulic conductance between
pores i and j, Lij the distance from pore i to pore j and Pj the pressure in pore j. Network
elements have square, triangular or circular cross-sections and retain wetting films in the
corners, which is important to correctly model the wetting phase connectivity.
The multi-scale PNM simulation extends this framework by taking into account microlinks. Flow through the microporosity is taken into account by assigning upscaled
transport properties to these network elements. Therefore, the multi-scale simulations
take a number of variable parameters of the microporosity as input rather than calculating
them: porosity, Pc-curve, (relative) permeability, breakthrough capillary pressure (BCP)
at a certain micro-link length, formation factor and saturation exponent. The advantage is
that the model does not have to take each separate micropore into account, allowing to
simulate larger volumes and thus incorporate more rock heterogeneity in the simulations.
The transport properties of the coupled network are then calculated as described in the
rest of this section (more details provided in [12]).
The saturation is calculated by summing the macroporous and the microporous
saturation, the latter is extracted from the microporositys Pc-curve. The porosity of the
microporous phase is supplied as input, while the total volume of microporosity is

SCA2015-027

4/12

calculated on the micro-CT image. Macropores can be invaded with non-wetting fluid
through a micro-link if a BCP is exceeded. This type of invasion is usually only
important for macropores which are only connected to the inlet face of the network by
taking into account microporous connectivity, as other macropores can be invaded
through macro-throats (which have larger radii than the throats in the microporosity). If
one would assume one single BCP for all micro-links, each macropore would be invaded
instantaneously and simultaneously at a certain capillary pressure. To come to a more
realistic drainage behavior, we try to mimic the behavior that a network with individual
micropores would display. Each microlink is therefore regarded as a cuboid domain filled
with a regular network of micropores. The hight of the cuboid is determined as the poreto-pore distance, while the square base is set to match the average pore-microporous
contact surface area of the micro-link. Scaling relations can then be deduced from
percolation theory to find the breakthrough saturation at which the non-wetting phase
percolates through this geometry, if a breakthrough saturation for a certain domain size is
provided as input. This input value can be calculated from a representative single-scale
network for the microporosity, informed by e.g. SEM imaging. The percolation threshold
saturation is translated to a BCP for each micro-link by using the input capillary pressure
curve for the microporosity. It should be noted that the approach followed here does not
take correlated heterogeneity in the microporous network into account.
Permeabilities and electrical resistivities of the multi-scale network are calculated in the
classical way (equation 1, [11]), but micro-links are also assigned a conductivity. Since
the microporosity is treated as a continuous porous medium with known upscaled
properties (which are provided as input), these conductivities can be calculated if a
geometric shape is assigned to each micro-link. Micro-links are treated as truncated
cones: the surface areas of the bases are set equal to the pore-micropore contact surface
areas, and the length of the cone is set equal to the pore-to-pore distance. The electrical
or hydraulic conductivity of the micro-link can then be calculated by taking into account
this geometry, in combination with the microporositys effective electrical conductance
or effective permeability at its current wetting phase saturation:
gij

Ai Aj

micro ( S w ) Lij

(2)

With Ai the contact surface area of the microporosity cluster to pore i, Lij the distance
from pore i to pore j, and micro(Sw) the hydraulic or electric resistance of the
microporosity at the current wetting saturation:

hydr ,p ( S w )

k kr , p (Sw )

elec (Sw ) R w FFmicro Sw n

(3)
(4)

SCA2015-027

5/12

In these equations, p and kr,p(Sw) are the viscosity and the relative permeability of fluid
phase p (wetting or non-wetting), k is the absolute permeability, Rw is the resistivity of
the wetting fluid, FFmicro the formation factor, and n the saturation exponent.For
computational reasons, a maximum length for micro-links has to be set. All micro-links
larger than this length are removed. This cut-off length also influences the results, as the
simplifications made in the network model cause an overestimation of the microporous
conductivity in long micro-links. Therefore, the cut-off length is chosen as the shortest
length for which all micro-connected macropores can be drained appropriately.
Eliminating this user-defined parameter is the subject of further research.

RESULTS AND DISCUSSION

For Estaillades, the microporositys input Pc-curve was deduced from a mercury intrusion
porosimetry (MIP) experiment. For Savonnires, this Pc-curve was based on the pore size
distribution in [7]. For both cases, the input Kr-curve and BCP were assessed by
simulations on an artificial network with a pore throat size distribution representative for
the microporosity in each rock. The other microporosity input properties (Table 2) were
based on [4] and [7], respectively.
Estaillades Limestone
In the Estaillades model, all macro-pores and throats were scaled with a factor 1.25 to
correct for an overestimation of intrusion capillary pressure (likely due to segmentation
uncertainty). Single-phase transport results calculated on the DPNM can be found in
Table 1. Kr and RI results are compared to experimental measurements from [13,14] in
Figure 2. DPNM simulations match the experimental trends better than single-scale
PNM, showing the importance of including microporosity in the simulations. Differences
between the experimental measurement and our results can be explained by sample scale,
heterogeneity, and by the difficulty of assessing the input parameters. In this sample, the
microporosity conducts mainly in parallel to the macroporosity. The model presented
here and the model in [4] mainly differ in their treatment of serial microporous
conductivity. As can be expected then, the RI results are in good agreement with the
results presented in [4] and show double-bending behavior [14]. The Kr-results agree
with the observations in [16] that added microporous pathways increase Kr,nw and
decrease Kr,w, and that this effect is stronger for higher microporous conductivities (see
further).
Savonnires Limestone
In the investigated sample, the macropores in the hollow ooliths make up 30% of the
macropore space. As these pores are only connected through microporosity, they are not
taken into account when calculating transport properties with conventional, single-scale
PNM. The Pc-curve simulated with the DPNM therefore matches the experimental Pccurve (scaled from an MIP experiment on a different Savonnires sample) better than the
single-scale PNM (Figure 3). The hollow ooliths start to drain at Sw = 0.61, when nearly
all inter-granular macropores have been filled with non-wetting phase. The Kr-curve
(Figure 3) is in qualitative agreement with measurements on Brauvilliers limestone [15],

SCA2015-027

6/12

which has a similar lithology than Savonnires but is tighter. The saturations at the crossover Kr and at the onset of non-wetting phase flow are matched better with the DPNM
than with the PNM, mostly due to the inclusion of the wetting saturation in
microporosity. In the multi-scale network, the inter-granular macropores dominate the
Kr,nw behavior (Kr,nw = 0.83 at Sw = 0.61). At Sw < 0,61, Kr,nw bends down due to the
lower importance of the microporosity/hollow ooliths pore system for the conductivity. In
the RI-simulation (Figure 3), the bending-up behavior at Sw=0.3 in the single-scale
simulation contradicts the classification of Brauvilliers as a texture III rock characterized
by a double bending RI [14]. In the DPNM simulation however, the curve weakly bends
up until reaching the macropores percolation threshold (Sw = 0.61), and bends down
when the pressure is further increased because the wetting films in the macropores
become the dominant electrical pathways when the microporosity drains. The weakness
of the bending-up may be an artifact of the selection of the length cut-off parameter. The
simulations on Savonnires illustrate how our DPNM can be used when microporosity
provides both parallel and serial connectivity to macroporosity.
Sensitivity study of Pc and Kr on microporosity properties
In this section, the influence of the microporositys permeability, breakthrough saturation
and capillary pressure curve on the relative permeability behavior of the multi-scale
networks is investigated. Varying these properties helps to better understand the influence
of microporosity on petrophysical properties of complex rocks. Furthermore, it allows to
probe the sensitivity of the model to the microporositys input properties which are often
difficult to assess.
Influence of the microporositys permeability
The microporositys permeability was varied over 4 orders of magnitude for the
Estaillades and the Savonnires network (Figure 4), while keeping all other parameters
constant. With increasing microporous permeability, the non-wetting phase relative
permeability decreases and the wetting phase relative permeability increases. This is due
to the increasing importance of fluid flow through microporosity, which is only invaded
by the non-wetting phase at Sw < 0.5 for Estaillades and Sw < 0.6 for Savonnieres. The
behavior fits well with the observation that non-wetting phase mobility is enhanced and
wetting phase mobility decreased by the existence of high-permeability pathways [16], in
this case caused by the macroporous network.
Influence of the microporositys breakthrough saturation
To test the influence of the input breakthrough saturation (and equivalently BCP) on a
network where a significant amount of micro-links conduct in series to macro-throats,
this parameter (defined for a 100x100x100 m geometry) was varied for simulations with
the Savonnires network. Drainage results for the different breakthrough saturations
(Figure 5) show identical Pc-curves and Kr-curves for Sw down to 0.6, as macro-pores
cannot be invaded through micro-links as long as the capillary pressure is below the
microporositys intrusion capillary pressure. The well-connected macropores are thus
drained first. Below Sw = 0.6, Pc and Kr curves show minor differences, related to the fact

SCA2015-027

7/12

that disconnected macro-porosity (mostly hollow ooliths) start to fill at different capillary
pressures for different microporous breakthrough saturations. The results show that this
parameter is of particularly minor importance to the Kr-calculations in this network, if the
pore sizes of the microporosity and the macroporosity do no significant overlap.
Influence of the microporositys input capillary pressure curve
To assert how the microporositys pore size distribution affects the drainage behavior in a
multi-scale carbonate, the input capillary pressure curve in the Savonnires simulation
was varied while keeping the other input parameters constant. Results with input
capillary pressure curves corresponding to a scaling of the microporous throat sizes with
factors 0.1, 2, 3, 4 and 5 were compared. Size distributions for micro-and macroporosity
are shown in Figure 6. When microporous throat sizes are scaled with a factor 3 or more,
there is significant overlap with the throat sizes of the macroporous network. This is an
interesting test case, as it indicates how the model responds to situations where there is no
clear distinction between the peaks in a rocks multimodal pore size distribution, or when
the resolution of the imaging method (in this case micro-CT) cannot be tuned to a value
in between the peaks. Scaling factors larger than 5 were not tested, as microporosity with
throat sizes larger than the macroporosity would be unphysical.
When microporosity is scaled with a factor 0.1, all well-connected macropores are filled
before the microporosity is invaded. This is clearly reflected in the Pc-curve by a strong
increase in the capillary pressure at Sw = 0.6. The new flow paths which are opened to
non-wetting phase flow below this saturation are less efficient, as they are associated with
drainage of macropores which only contribute to the flow through connections with lowpermeability micro-links. In the Kr,nw-curve (Figure 7), this point in the drainage
sequence is characterized by a sharp bend. The influence on Kr,w is much smaller, as the
wetting phase flow is already limited by passage through wetting films in the (fully
drained) well-connected macro-pore network.
For scaling factors larger than 1, the rise in the Kr,nw-curve shifts to lower wetting
saturations (Figure 7). This behavior can be analyzed by looking at the relative
permeability behavior of the macro-pore network separately (classical network in
Figure 3). Due to its heterogeneous nature, the relative permeability of the macronetwork has a sharp increase at a (macroporous) wetting saturation of 0.5. At this point,
70% of the well-connected macro-pore space has been drained. The slight shift for the
multi-scale simulation with scaling factor 2 (Figure 7) can then be explained by taking
into account that the microporositys invasion capillary pressure is reached at S w = 0.72
for this simulation. Therefore, at a given total wetting saturation below 0.72, the wetting
saturation of the well-connected macro-network is higher than in the multi-scale
simulation with the original microporositys throat sizes (due to drainage of
microporosity and, at lower saturation, of isolated macropores). For larger scaling factors,
the microporosity also starts to significantly affect the invasion sequence of the wellconnected macropores, due to the increased overlap in microporous and macroporous
throat sizes. For scaling factor 3, the strong increase in permeability (at total Sw = 0.65)

SCA2015-027

8/12

only happens when 82% of the well-connected macropore space has been drained
(compare to 70% in the original multi-scale network). This means that the well-connected
macroporosity is drained through macro-throats and micro-links concurrently, and
suggests that pores are not strictly drained in order of decreasing size of their connected
macro-throats. More validation work is needed to investigate the physical relevance of
the model in such situations.

CONCLUSION
In this work, we show a method to perform DPNM simulations with upscaled
microporosity properties based on micro-CT scans of heterogeneous rocks. The
employed method takes microporous conductivity into account by the concept of
microlinks, and allows pores to be drained through microporous connectivity in the
drainage simulation. While the method cannot be considered truly predictive yet, the
presented simulations on real carbonates illustrate how the model can help to understand
the influence of microporosity on transport properties, by investigating the influence of
microporosity properties. Major remaining challenges are the current need to include a
user defined micro-link cut-off parameter and the difficulty of assessing the upscaled
petrophysical parameters for the microporosity.

ACKNOWLEDGEMENTS
Prof. Dr. M. Blunt and Dr. A.Q. Raeini are thanked for helping with the maximal ball
code and the microPoreFlow code. Dr. H. Derluyn and Dr. M.A. Boone are
acknowledged for their help with the Savonnires sample and its micro-CT scan, and Dr.
J. Dils and Prof. Dr. G. Deschutter for the MIP measurements on Estaillades. The agency
for Promotion of Innovation by Science and Technology in Flanders (IWT) is
acknowledged for T. Bultreys PhD grant.

REFERENCES
1. Cnudde V., M.N. Boone, "High-resolution X-ray computed tomography in
geosciences: A review of the current technology and applications", Earth-Science
Rev. (2013) 123, 117.
2. Jiang Z., M.I.J. van Dijke, K.S. Sorbie, G.D. Couples, "Representation of multiscale
heterogeneity via multiscale pore networks", Water Resources Res. (2013) 49, 5437
5449.
3. Prodanovi M., A. Mehmani, A.P. Sheppard, "Imaged-based multiscale network
modelling of microporosity in carbonates", Geol. Soc. London, Spec. Publ. (2014)
406, SP4069.
4. Bauer D., S. Youssef, M. Fleury, S. Bekri, E. Rosenberg, O. Vizika, "Improving the
Estimations of Petrophysical Transport Behavior of Carbonate Rocks Using a Dual
Pore Network Approach Combined with Computed Microtomography", Transp.
Porous Media. (2012) 94, 505524.

SCA2015-027

9/12

5. Youssef S., M. Han, D. Bauer, E. Rosenberg, S. Bekri, M. Fleury, O. Vizika, "High

resolution -CT combined to numerical models to assess electrical properties of
bimodal carbonates", (2008) 37, in: Int. Symp. Soc. Core Anal., Society of Core
Analysts, Abu Dhabi UAE.
6. Derluyn H., M. Griffa, D. Mannes, I. Jerjen, J. Dewanckele, P. Vontobel, A.
Sheppard, D. Derome, V. Cnudde, E. Lehmann, J. Carmeliet, "Characterizing saline
uptake and salt distributions in porous limestone with neutron radiography and X-ray
micro-tomography", J. Build. Phys. (2013) 36, 353374.
7. Roels S., J. Elsene, J. Carmeliet, H. Hens, "Characterisation of pore structure by
combining mercury porosimetry and micrography", Mater. Struct. (2001) 34, 7682.
8. Masschaele B., M. Dierick, D. Van Loo, M.N. Boone, L. Brabant, E. Pauwels, V.
Cnudde, L. Van Hoorebeke, "HECTOR: A 240kV micro-CT setup optimized for
research", J. Phys. Conf. Ser. (2013) 463, 012012.
9. Dong H., M. Blunt, "Pore-network extraction from micro-computerized-tomography
images", Phys. Rev. E. (2009) 80, 036307.
10. A. Mehmani, M. Prodanovi, "The effect of microporosity on transport properties in
porous media", Adv. Water Resources. (2014) 63, 104119.
11. Valvatne P.H., M.J. Blunt, "Predictive pore-scale modeling of two-phase flow in
mixed wet media", Water Resources Res. (2004) 40, W07406.
12. Bultreys T., L. Van Hoorebeke, V. Cnudde, "Multi-scale, micro-computed
tomography-based pore network models to simulate drainage in heterogeneous
rocks", Adv. Water Resources. (2015) 78, 3649.
13. Ott H., C.H. Pentland, S. Oedai, "CO2brine displacement in heterogeneous
carbonates", Int. J. Greenh. Gas Control. (2015) 33, 135144.
14. Han M., M. Fleury, P. Levitz, "Effect of the pore structure on resistivity index
curves", (2007) 34, in: Int. Symp. Soc. Core Anal., Society of Core Analysts, Calgary
CA.
15. Moctezuma-Berthier P., O. Vizika, P. Adler, "Water-oil relative permeability in
vugular porous media: experiments and simulations", (2002) 06, in: Int. Symp. Soc.
Core Anal., Society of Core Analysts, Monterey USA.
16. Bekri S., C. Laroche, O. Vizika, "Pore-network models to calculate transport
properties in homogeneous and heterogeneous porous media", Computational
Methods in Water Resources, Proceedings of the XIVth International Conference,
Elsevier, Delft, Holland, (2002) 2, 1115, Hassanizadeh, Schotting, Gray and Pinder,
Elsevier

SCA2015-027

10/12

TABLES AND FIGURES

Table 1. Single phase flow properties of the two carbonates calculated with single-scale PNM, with DPNM
and measured experimentally (experimental values for Estaillades from [4]).
K PNM (mD)
K DPNM (mD)
K Exp. (mD)
FF PNM
FF DPNM
FF Exp.

Estaillades
281
289
273
36.8
19.3
24

Savonnires
244
268
903 (but
heterogeneous)
85.7
13.5
-

Table 2. Input parameters for the DPNM.

micro
Kmicro(mD)
FFmicro
BCP

Cut-off length (m)

Estaillades
0.36
8
7.72
0.3 for
(300m)
network
67.5

Savonnires
0.35
4.1
8.2
0.3 for
(100m)
network
160

Figure 1. Illustration of DPNM extracted from avonnires limestone. Microporous zones are in dark grey,
micro-links in blue.

Figure 2. RI and Kr-curves for drainage of Estaillades calculated with a single scale maximal ball based
PNM and with a DPNM, compared to experimental results from resp. [14] and [13] (the former estimated
from the graph).

SCA2015-027

11/12

Figure 3. Pc-curve (left), RI-curve (middle) and Kr-curves (right) for drainage of Savonnires, calculated
with a single scale maximal ball based PNM and with the proposed DPNM. The experimental Pc-curve is
obtained from an MIP measurement.

Figure 4. Relative permeability results from drainage simulations on the multi-scale Estaillades (left) and
the Savonnires (right) networks, with different input properties for the microporositys permeability.

Figure 5. Drainage capillary pressure and relative permeability results obtained on the multi-scale
Savonnires network, for different microporosity breakthrough saturations (defined at a 100x100x100 m
geometry). This parameter has a minor effect on relative permability results.

SCA2015-027

12/12

Figure 6. Cumulative pore size distributions and multi-scale drainage capillary pressure curves for
different input microporosity Pc-curves in Savonnires.

Figure 7. Drainage relative permeability curves calculated on the multi-scale Savonnires network with
different microporosity Pc-curves as input. These input Pc-curves correspond to scaling of the throat sizes in
the microporosity with different factors. The permeability of the microporosity was kept constant, to allow
better interpretation of the results.

SCA2015-028

1/12

NANOFLUID ENHANCED OIL RECOVERY MOBILITY

RATIO, SURFACE CHEMISTRY, OR BOTH?
A. Khezrnejad, L.A. James, T.E. Johansen
Memorial University of Newfoundland
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
The goal of enhanced oil recovery (EOR) is to manipulate the fluid-fluid properties and
fluid-rock properties between the injected fluid and the residual oil phase to improve
recovery efficiency. Water enhanced with nanoparticles (nanofluids) has recently gained
research interest for enhanced oil recovery because of the possible physical and chemical
properties imparted by the nanoparticles. The application of nanofluids in enhanced oil
recovery is strongly dependent on the resulting nanofluid properties. The research
question asked is whether oil recovery using nanoparticle enhanced water is due to a
more favorable mobility ratio (increased water phase viscosity) or is it due to the effect of
the enhanced surface chemistry? In this study, we examine the role of increased viscosity
of the water phase on oil recovery using nanoparticle enhanced water and polymer
enhanced water with similar viscosity. First, the nanoparticle enhanced water is
characterized. A statistical design of experiments technique, Response Surface
Methodology, is used to investigate the effect of the type of nanoparticles (silicon oxide
and aluminum oxide nanoparticles), concentration of the nanoparticles, pressure, and
temperature on viscosity. The effect of interactions between the factors on viscosity is
also studied. Second, the viscosity measurement results are used to plan micromodel and
coreflooding laboratory scale enhanced oil recovery experiments at low pressure and
temperature conditions. The results can be used to help elucidate the role of increasing
viscosity versus surface chemistry on oil recovery.

INTRODUCTION
The two main forces controlling fluid flow in porous media are viscous and capillary
forces. Literature suggests that recovery efficiency can be improved for waterflooding by
increasing the viscosity of injected fluid (improving the macroscopic sweep efficiency) or
through improving the microscopic efficiency via wettability alteration or interfacial
tension reduction [1, 2, 3].
Polymers are commonly used to increase viscosity of the injected phase [4, 5, 6, 7].
However, challenges such as the stability of polymers in harsh reservoir conditions, cost,
and required facilities hinder the wider use of polymer flooding [8, 9]. Adding
nanoparticles to water increases the viscosity; hence, improves the mobility ratio [10, 11].
Nanofluids have been shown to be very effective in terms of wettability alteration and
interfacial tension reduction [12, 13, 14]. Moreover, Zhang et al. [15] along with other

SCA2015-028

2/12

researchers demonstrated that specially designed nanoparticles are significantly more

stable than polymers or surfactants in harsh reservoir conditions [16, 17]. All these
features make nanofluids a very promising EOR technique for improving both
microscopic and macroscopic sweep efficiency.
In this study, nanoparticle enhanced water flooding is compared to polymer enhanced
water flooding to examine the role of mobility ratio, surface chemistry or both on oil
recovery efficiency. Two different types of nanoparticles were added to deionized (DI).
The results of the nanoparticle enhanced water flooding experiments are compared to
polymer flooding using a polymer solution with the same viscosity as the nanofluid.
Moreover, response surface methodology, a statistical design of experiment technique,
was used to investigate the effect of nanoparticle concentration, pressure, and temperature
on the viscosity of DI water. The effect of interactions between the factors on the
viscosity was also studied. Interfacial tension between the nanofluid and oil, and the
polymer solution and oil were also measured to better understand the possible
mechanisms of oil recovery.

MICROMODEL EXPERIMENTS
Etched micromodels were used to examine the effectiveness of injecting nanoparticle
enhanced water and polymer water solutions on oil recovery. The properties of the
dispersed nanoparticles (manufactured by US Research Nanomaterials, Inc.) are tabulated
in Table 1. It should be noted that the silica (SiO2) nanoparticles used in the experiments
were amorphous, and the alumina (Al2O3) nanoparticles were gamma type. In order to
completely disperse the nanoparticles in the DI water, an ultrasonic device was used. The
sonication process was performed on the nanofluid for 30 minutes. Homogeneity and
stability of the prepared solution were confirmed by placing the nanofluid solution in a
closed transparent bottle away from degrading factors such as light and heat for two
weeks. Visual inspection showed neither precipitation nor other visible alterations
indicating a stable nanoparticle suspension. The polymer used in these set of experiments
was Flopaam 3430S (manufactured by SNF Floerger). The hydrocarbon fluid used in the
experiments was stock tank crude oil from offshore Newfoundland with approximately
32-35 API.
Type
Al2O3
SiO2

Description
Gamma
Amorphous

Table 1: Properties of nanoparticles

Average Particle Size (nm)
Purity (%)
10
99.9
5-35
99.9

pH value
2-5
8-11

A micromodel fabricated from polymethyl methacrylate (PMMA) was used as the porous
medium. Figure 1 shows a photo of the micromodel saturated with oil. Table 2 shows the
properties of the micromodel. Different scenarios of waterflooding were performed on
the glass micromodel using: DI water, polymer solution, silica nanofluid (5 wt%), and
alumina nanofluid (5 wt%). Figure 2 shows a schematic of the micromodel visualization
setup. The different injection scenarios are defined in Table 3.

SCA2015-028

3/12

Figure 1: PMMA Micromodel

Figure 2: Schematic diagram of micromodel visualization setup

The following procedure was used to conduct the nanofluid/polymer solution/DI

waterflooding experiments in the low pressure (ambient conditions) micromodel system:
1. The micromodel was cleaned by flushing with hexane (2 pore volumes).
2. The micromodel was completely dried using compressed air.
3. The injection fluids were loaded in the syringe pumps.
4. DI water was injected up to the inlet through the temporary line.
5. The micromodel inlet valve was closed.
6. The micromodel and downstream tubing were vacuumed to remove any air and
reduce the probability of trapping air during the primary imbibition.
7. The outlet valve was closed.
8. By opening the inlet, the system was drained by DI water.
9. The outlet valve was opened.
10. Oil was loaded into a temporary line.
11. Oil was injected into the micromodel to the outlet.
12. Slugs of desired fluid were injected in the micromodel.

SCA2015-028

Table 2: Micromodel dimensions and

characteristics
Description
Dimension
Length (cm)
25.6
Width (cm)
6.4
Average depth (m)
160
Porosity
0.43
Pore Volume (cm3)
1.15
Permeability
2.9
(Darcy)

4/12

Test
#
1
2
3
4
5
6
7
8

Table 3: Micromodel Experimental Conditions

Viscosity Injection flow
Injected Fluid
(cP)
rate (ml/min)
DI water
1.00
0.010
Polymer (10 ppm)
1.75
0.010
Polymer (10 ppm)
1.75
0.005
Polymer (10 ppm)
1.75
0.010
Polymer (20 ppm)
2.30
0.010
Polymer (20 ppm)
2.30
0.005
SiO2 nanofluid (5 wt%)
1.06
0.010
Al2O3 nanofluid (5 wt%)
1.75
0.010

Oil saturation in the micromodel experiments can be measured at any time by image
analysis. Standard image analysis using Matlab software was used to determine the oil
recovery. The difference between the initial state of the black pixels and the final state
was interpreted as oil recovery.
Figure 3 shows the oil recovery at different pore volumes injected. The recovery values
used for the polymer 10 ppm case are the average value of test # 2 and 4. The standard
deviation was measured to be 1.89% recovery according to recovery values measured for
these two tests. The standard deviation is also shown in Figure 3. As shown in the figure,
the recovery values for alumina, silica and DI water are outside of this standard deviation
giving us confidence in the experimental results. Comparing the recovery of alumina
nanofluid and polymer solution (10 ppm) both with = 1.75 cP at experimental
conditions, we can see that the oil recovery for the alumina nanofluid is higher. Moreover,
we can see that silica nanofluid with =1.06 cP but the lowest interfacial tension (will be
discussed more) has the highest oil recovery. This higher oil recovery might result from
improvement of microscopic sweep efficiency. Nanoparticles have the ability to decrease
the interfacial tension between oil and water and improve microscopic efficiency, which
will be discussed in detail in IFT measurements section later.
Figure 4 shows the ultimate oil recovery, oil recovery at breakthrough, and breakthrough
time for different injection scenarios. As shown in the figure, oil recoveries obtained for
the alumina and silica nanofluids are 8 and 11% higher than the oil recovery by DI water
injection respectively, showing a significant improvement of oil recovery. Moreover, the
recoveries from injecting alumina and silica nanofluids were 5 and 8% higher
respectively than the recovery from injecting 10 ppm polymer. The oil recovery from
injecting 10 ppm polymer compared to DI water injection was approximately 3% higher,
which illustrates the effect of the increased viscosity of the injected fluid on oil recovery.
As shown in Figure 4, injected fluids with the same viscosities have approximately the
same breakthrough time.

SCA2015-028

5/12

60

50

Recovery (%)

40
Silica nanofluid, m = 1.06 cP, 0.01 ml/min

+ Polymer 20 ppm, m = 2.30 cP, 0.01 ml/min

30

Alumina nanofluid, m = 1.75 cP, 0.01 ml/min

Polymer 20 ppm, m = 2.30 cP, 0.005 ml/min
Polymer 10 ppm (avg), m = 1.75 cP, 0.01 ml/min
Polymer 10 ppm, m = 1.75 cP, 0.005 ml/min
DI Water, m = 1.0 cP, 0.01 ml/min

20

10

0
0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Injected pore volume

Figure 3: Oil Recovery vs. injected pore volume

Ultimate recovery (%)

Breakthrough time (min)

120

50

100

40

80

30

60

20

40

10

20

Test2#3
Test1#6
Polymer
Polymer
10 ppm
20 ppm
m = 2.30 cP m = 1.75 cP
0.005 ml/min 0.005 ml/min
0.005 ml/min

Test 3#2, 4
Test4 #5
Polymer
Polymer
10 ppm
20 ppm
m = 1.75 cP m = 2.30 cP
0.010 ml/min 0.010 ml/min
0.010 ml/min

Test #8
Test6#7
5
Alumina
Silica
nanofluid
nanofluid
m = 1.75 cP m = 1.06 cP
0.010 ml/min 0.010 ml/min

Test #1
7
DI
water
m = 1.00 cP
0.010 ml/min

Time (min)

Recovery (%)

Recovery at breakthrough (%)

60

Figure 4: Ultimate recovery, recovery at breakthrough, and breakthrough time for the experiments

One replicate test was performed for the polymer (10 ppm) solution to better estimate the
error in the experiments. The difference in recovery of tests two and four with a polymer
concentration of 10 ppm at 0.010 ml/min was approximately 2%. This difference in
recoveries might be due to experimental errors or image analysis errors. An additional
experiment at half flow rate was also performed to see the effect of flow rate on

SCA2015-028

6/12

breakthrough time and ultimate recovery. Decreasing the injection flow rate causes a
more stable front movement, which increases the breakthrough time significantly.
However, ultimate recovery did not change by decreasing the flow rate.
Overall, nanofluid injection has higher recovery compared to polymer waterflooding. The
fact that nanoparticles are surface active at the oil-water interface might be the reason for
this improvement. Therefore, we measured the interfacial tension between the nanofluid
and oil as well as the interfacial tension between the polymer solution and oil to better
understand the role of nanoparticles in possibly reducing the interfacial tension between
the immiscible phases. This is discussed in more detail subsequently. Mobility control
might be another reason for enhanced oil recovery of nanofluid or polymer injection over
simple water flooding. In the next section the viscosity measurements for nanofluid,
polymer solution and DI water are discussed.

NANOFLUID VISCOSITY MEASUREMENTS

As discussed in the introduction, nanoparticles are capable of increasing the viscosity of
water. However, most of these viscosity measurements were conducted under fixed
pressure and temperature. In order to understand the behavior of viscosity with respect to
different factors and their interactions, we need to change all the factors at the same time.
Response surface methodology (optimal design) was employed to investigate the effect of
each factor: concentration of nanoparticles, pressure, temperature, nanoparticle type, and
their interaction on the response (viscosity).
In this study, Design Expert Software was used for the design of experiments. Table 4
shows the 24 viscosity measurements based on optimal design. Concentration, pressure,
and temperature are quantitative factors varying from 0 to 5 wt%, 20 to 8000 psia, and 20
to 80C respectively.
The VISCOlab PVT viscometer (manufactured by Cambridge Viscosity) was used for
measuring the viscosity. A billet for the range of 0.25 to 5 cP was used for the
experiments. After loading the pump, the system was set to the desired temperature and
pressure. Before measuring any data, the system was bled through the relief valves to rid
the system of any air. Then, viscosity was measured under stable pressure and
temperature conditions. After running each test, the system was flushed with an
appropriate solvent to clean all the lines and fittings, and then vacuumed. Bias was
avoided by performing the experiments in random order. As shown in the table, *
indicates replicate runs. A standard deviation of 0.01cP was calculated based on the
replicate values. The viscometer was calibrated using DI water. As shown in Table 4, the
viscosity measured for the DI water at ambient condition was 1.07 cP (run #15), which is
slightly different than available data for DI water viscosity in literature. This discrepancy
might be due to experimental errors.

SCA2015-028

Run
*1
*2
*3
*4
5
6
*7
8
9
10
11
*12
13
14
15
16
17
18
*19
20
21
22
23
24

7/12

Table 4: Optimal design of viscosity measurements (* denotes replicate runs)

Concentration
Pressure
Temperature
Nanoparticle
Viscosity
(wt%)
(psia)
(C)
type
(cP) 0.01
5.00
8000
55.4
Al2O3
0.78
5.00
8000
55.4
Al2O3
0.80
2.50
4010
50.0
SiO2
0.66
2.50
4010
50.0
SiO2
0.66
0.00
4110
79.8
DI water
0.36
5.00
20
80.0
Al2O3
0.53
2.50
4010
50.0
SiO2
0.66
5.00
20
20.0
SiO2
1.06
0.00
4848
22.4
DI water
0.86
5.00
4010
24.4
Al2O3
1.82
5.00
5207
79.8
SiO2
0.40
2.90
8000
80.0
Al2O3
0.40
1.85
20
33.8
SiO2
0.81
0.00
8000
22.6
DI water
0.84
0.00
20
20.0
DI water
1.07
3.80
3611
61.6
Al2O3
0.74
5.00
8000
21.3
SiO2
0.92
1.00
4010
38.0
Al2O3
0.75
2.90
8000
80.0
Al2O3
0.43
1.43
20
61.7
Al2O3
0.51
0.00
8000
79.5
DI water
0.34
0.00
20
80.4
DI water
0.37
0.55
7880
42.5
SiO2
0.57
5.00
4010
27.8
Al2O3
1.69

Table 5 shows the analysis of variance (ANOVA) results for the viscosity measurements.
The prediction interval provides the upper and lower levels for 95% confidence level.
The p-value represents the probability of the occurrence of a given event. When the pvalue is less than 0.05 (1-95% confidence) the factor is considered significant. The
analysis of variance is model dependent, so it is up to the user to suggest models that
describe the data. We systematically tried and compared several different models (linear,
quadratic, etc.) with the inclusion and elimination of higher order and interaction terms.
The goal was to find the simplest model to best fit the results. The results of the ANOVA
table demonstrate that all the individual factors have significant effect on viscosity (pvalue < 0.05). Moreover, the interaction between concentration and nanoparticle type and
second order terms of pressure and temperature were shown to be important and should
be considered in the model.
Figure 5 shows how the viscosity predicted by the model matches the actual experimental
data. As shown in the figure, the data points fall very close to the 45 slope line, which
confirms that lack of fit is not significant as it was shown in the ANOVA table.
Figure 6 shows the effect of nanoparticles concentration on the viscosity. The black
points are indicating the experimental measurements. Figure 6a demonstrates that by

SCA2015-028

8/12

increasing the concentration of alumina (Al2O3) nanoparticles in the DI water, the

viscosity increases. However, Figure 6b shows that adding silica (SiO2) nanoparticles to
DI water does not change the viscosity significantly. As shown in the figure, viscosity
decreases with increasing temperature. The effect of pressure on the viscosity of
nanofluid is slightly more significant at lower temperatures. In fact, the viscosity values
at high temperature (80C) are almost the same for different states of pressure, which is
showing the insignificance of pressure effect on viscosity at high temperature. The
dashed lines indicate the 95% confidence interval bands for the predictive model.
Table 5: Analysis of variance for viscosity experiments
Sum of
Source
Mean Square
Squares
Model
2.320
0.330
Concentration (C)
0.160
0.160
Pressure (P)
0.023
0.023
Temperature (T)
1.840
1.840
Nanoparticle type (N)
0.079
0.079
Concentration x Nanoparticle Type (CN)
0.069
0.069
Pressure2 (P2)
0.011
0.011
Temperature2 (T2)
4.323x10-3
4.323x10-3
Lack of fit
0.019
1.096x10-3

F Value

p-value

359.11
176.93
25.09
1990.49
85.15
74.98
12.14
4.68
3.26

<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
<0.0001
0.0021
0.0416
0.0976

Figure 5: Predicted viscosity vs. experimental viscosity measurements

SCA2015-028

Figure 6a: Effect of nanoparticles concentration on

Alumina (Al2O3) nanofluid viscosity

9/12

Figure 6b: Effect of nanoparticle concentration on

Silica (SiO2) nanofluid viscosity

Figure 7 shows the effect of interaction between pressure and temperature on the
measured viscosity of nanofluid. As shown in the ANOVA table, the effect of second
order terms of pressure and temperature are significant. We can see this non-linearity
effect in Figure 7. Again, pressure is shown to have little effect. Figure 7b indicates that
viscosity behaves more non-linear while using alumina nanoparticles at higher
concentration. Moreover, by looking at Figure 7b, we can see that the maximum viscosity
was obtained while using alumina nanoparticles at medium pressure, low temperature,
and high concentration. For silica nanofluid increasing the concentration of nanoparticles
in DI water does not affect the viscosity significantly. However, viscosity increases
significantly by increasing the concentration of alumina nanoparticles in DI water.
Increasing the concentration of alumina nanoparticles increases the effect of non-linearity
behavior of viscosity (Figure 7a and 7b). Overall, the model predicts that the viscosity of
DI water can be increased to a maximum value of approximately 2 cP using alumina
nanoparticle.

Figure 7a: 3D map of viscosity vs. pressure and

temperature (Al2O3, 1 wt%)

Figure 7b: 3D map of viscosity vs. pressure and

temperature (Al2O3, 5 wt%)

SCA2015-028

10/12

Figure 7c: 3D map of viscosity vs. pressure and

temperature (SiO2, 1 wt%)

Figure 7d: 3D map of viscosity vs. pressure and

temperature (SiO2, 5 wt%)

INTERFACIAL TENSION (IFT) MEASUREMENTS

An Interfacial Tension Meter (IFT 700, manufactured by Vinci Technologies) was used
to determine interfacial tension between the oil and DI water, polymer solution, and the
nanofluid (liquid-liquid interface) at ambient (experimental) conditions. The pendant
drop method was used for IFT measurements. An oil drop was created and put in contact
with the nanofluid in a cell. A camera connected to a computer records the shape of the
oil droplet to derive the interfacial tension. The results of IFT measurements are tabulated
in Table 6 where the value reported is the average value for approximately 70 runs. The
standard deviation was calculated for each test separately. As shown in the table,
interfacial tension decreases significantly by adding nanoparticles to DI water. Moreover,
the minimum IFT was obtained while using silica nanoparticles.
Table 6: Interfacial tension measurements
Fluid
IFT (mN/m)
DI Water and Oil
29.00
Silica nanofluid (5 wt%) and Oil
6.56 1.06
Alumina nanofluid (5 wt%) and Oil
12.71 0.35
Polymer Solution (10 ppm) and Oil
21.47 1.30

CONCLUSIONS
In conclusion, the results of this experimental work shows that nanoparticle have the
ability to increase oil recovery by improving both microscopic and macroscopic sweep
efficiencies. The results of viscosity measurements demonstrated that alumina
nanoparticles can increase the viscosity of deionized water. The viscosity of the silica and
alumina nanofluids was measured at different conditions of pressure, temperature and
nanoparticle concentration. Interfacial tension (IFT) experiments show that surface
chemistry plays an important role when using nanoparticle enhanced water compared to
polymer water solutions of the same viscosity in micromodel water flooding experiments.
The results of IFT measurements indicate that the IFT between oil and DI water can
decrease from 29 to 6.56 and 12.71 for silica (5 wt%) and alumina (5 wt%), respectively.

SCA2015-028

11/12

The results of micromodel experiments show that this IFT reduction causes higher oil
recovery using nanofluid injection compared to polymer flooding with the polymer
solution having the same viscosity as the nanofluids. Oil recoveries using polymer
injection with concentrations of 10 and 20 ppm were 3 and 9% higher respectively than
oil recovery using DI water injection, which shows the effect of viscosity improvement
on oil recovery. The recoveries obtained from silica and alumina nanofluid injection were
also higher than the recovery of DI water injection by 11 and 8% respectively. More
investigation is required but our results indicate that surface chemistry does seem to play
a role in oil recovery using nanofluids. This experimental work shows that nanoparticles
are very promising for EOR purposes due to their specific chemical and physical
properties, and the fact they have the ability to improve oil recovery through viscosity
improvement and surface chemistry.

ACKNOWLEDGEMENTS
The authors thank the Hibernia Management and Development Company (HMDC),
Chevron Canada, the Natural Sciences and Engineering Research Council of Canada
(NSERC), and the Research and Development Corporation (RDC) for their support,
without which this work could not have been performed.

REFERENCES
1. Chatzis, I., and Morrow, N. R., Correlation of Capillary Number Relationships for
Sandstone, SPE Journal, (1984) 24, 05.
2. Lake, L. W., Enhanced Oil Recovery, Prentice Hall, New Jersey, USA, (1989).
3. Nguyen, P. T., Do, B. P. H., Pham, D. K., Nguyen, Q. T., Dao, D. Q .P., and Nguyen,
H. A., Evaluation on the EOR Potential Capacity of the Synthesized Composite
Silica-Core/ Polymer-Shell Nanoparticles Blended with Surfactant Systems for the
HPHT Offshore Reservoir Conditions, Paper presented at the SPE International
Oilfield Nanotechnology Conference and Exhibition, 12-14 June, Noordwijk, The
Netherlands, (2012).
4. Wang, J., Wang, D., Sui, X., Bai, W., Combining Small Well Spacing with Polymer
Flooding To Improve Oil Recovery of Marginal Reservoirs, SPE-96946-MS Paper
presented at SPE/DOE Symposium on Improved Oil Recovery, 22-26 April, Tulsa,
Oklahoma, USA, (2006).
5. Wang, Y., Zhao, F., Bai, B., Optimized Surfactant IFT and Polymer Viscosity for
Surfactant-Polymer Flooding in Heterogeneous Formations, SPE-127391-MS Paper
presented at SPE Improved Oil Recovery Symposium, 24-28 April, Tulsa, Oklahoma,
USA, (2010).
6. Morelato, P., Rodrigues, L., Romero, O. L., Effect of Polymer Injection on the
Mobility Ratio and Oil Recovery, SPE-148875-MS Paper presented at SPE Heavy
Oil Conference and Exhibition, 12-14 December, Kuwait City, Kuwait, (2011).
7. Fletcher, A. J. P., Weston, S., Haynes, A. K., Clough, M. D., The Successful
Implementation of a Novel Polymer EOR Pilot in the Low Permeability Windalia
Field, SPE-165253-MS Paper presented at SPE Enhanced Oil Recovery Conference,
2-4 July, Kuala Lumpur, Malaysia, (2013).

SCA2015-028

12/12

8. Kurenkov, V. F., Hartan, H. G., Lobanov, F. I., Degradation of Polyacrylamide and

Its Derivatives in Aqueous Solutions, Russian Journal of Applied Chemistry, (2002)
75, 07, 1039-1050.
9. Ayatollahi, S. and Zerafat, M. M., Nanotechnology-assisted EOR techniques: New
solutions to old challenges, Paper SPE 157094 presented at the SPE International
Oilfield Nanotechnology Conference and Exhibition, Noordwijk, 12-14 June, (2012).
10. Fei, D., Dingtian, K., and Alexandru, C., Viscosity affected by nanoparticle
aggregation in Al2O3-water nanofluids, Nanoscale Research Letters, (2011) 6, 1,
X1-5.
11. Zhi-Yong, L., Dong-Jian, S., Zhong-Qiang, Z., Jian-Ning, D., Guang-Gui, C., Long,
Q., and Rui, Z., Effect of temperature and nanoparticle concentration on the
viscosity of nanofluids, Gongneng Cailiao/Journal of Functional Materials, (2013)
44, 1, 92-95.
12. Maghzi, A., Mohebbi, A., Kharrat, R., and Ghazanfari, M. H., Pore-scale monitoring
of wettability alteration by silica nanoparticles during polymer flooding to heavy oil
in a five-spot glass micromodel, Transport in Porous Media, (2011) 87, 3, 653-664.
13. Roustaei, A., Moghadasi, J., Bagherzadeh, H., and Shahrabadi, A., An Experimental
Investigation of Polysilicon Nanoparticles Recovery Efficiencies through Changes in
Interfacial Tension and Wettability Alteration, Paper SPE 156976 presented at the
SPE International Oilfield Nanotechnology Conference and Exhibition, Noordwijk,
12-14 June, (2012).
14. Khezrnejad, A., James, L. A., and Johansen, T. E., Water Enhancement Using
Nanoparticles in Water Alternating Gas (WAG) Micromodel Experiments, Paper
SPE 173484-STU presented at the SPE Annual Technical Conference and Exhibition,
27-29 October, Amsterdam, The Netherlands, (2014).
15. Zhang, T., Espinosa, D., Yoon, K. Y., Rahmani, A. R., Yu, H., Caldelas, F. M., Ryoo,
S., Roberts, M., Prodanovic, M., Johnston, K. P., Milner, T. E., Bryant, S. L., and
Huh, C., Engineered Nanoparticles as Harsh-Condition Emulsion and Foam
Stabilizers and as Novel Sensors, OTC-21212-MS Paper presented at Offshore
Technology Conference, 2-5 May, Houston, Texas, USA, (2011).
16. Jafari, S., Khezrnejad, A., Shahrokhi, O., Ghazanfari, M. H., and Vossoughi, M.,
Experimental Investigation of Heavy Oil Recovery by Continuous/WAG Injection of
CO2 Saturated with Silica Nanoparticles. International Journal of Oil, Gas and Coal
Technology, (2015) 9, 2, 169-179.
17. Yu, J., Wang, S., Liu, N., and Lee, R., Study of Particle Structure and
Hydrophobicity Effects on the Flow Behavior of Nanoparticle-Stabilized CO2 Foam
in Porous Media, Paper SPE-169047-MS presented at the SPE Improved Oil
Recovery Symposium, 12-16 April, Tulsa, Oklahoma, USA, (2014).

SCA2015-029

1/12

AN EXPERIMENTAL INVESTIGATION OF
NANOPARTICLES ADSORPTION BEHAVIOR DURING
TRANSPORT IN BEREA SANDSTONE
Shidong Li, Miaolun Jiang and Ole Torster,
Norwegian University of Science and Technology (NTNU)
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Nanoparticles fluid (nanofluid) has shown its potential for increasing oil recovery during
last few years. Due to very small size (D~ 1 to 100 nm), nanoparticles can pass through
reservoir, while the huge specific surface area of porous media and nanoparticles result in
significant adsorption of nanoparticles inside reservoir. The behavior of this adsorption
plays very important role for enhanced oil recovery (EOR) mechanisms of nanofluid, it
can alter reservoir wettability and change permeability. Investigation of nanoparticles
adsorption behavior leads to better understanding of nanofluid EOR process.
The objective of this experimental study is to investigate nanoparticles adsorption
behavior during transport in Berea sandstone, analyses of pressure drop and nanoparticles
concentration of effluent fluid were used to evaluate nanoparticles adsorption. Three
different wettability core plugs (water, oil, and neutral wet) with 8cm length and 3.8 cm
diameter were employed. Hydrophilic nano-structure particle and colloidal nanoparticle
were used in this experiment and they were dispersed in 3 wt. % brine. Nanofluid was
injected into each core plug saturated with brine for several pore volumes, and brine was
injected afterwards for post-flush. Pressure drop across core was recorded during whole
injection process. Nanoparticles concentration of effluent fluid was measured to plot
adsorption curve.
The results showed that nano-structure particle and colloidal nanoparticles undergo
adsorption during transport inside core, but nano-structure particle has larger adsorption
amount than colloidal nanoparticle. For nano-structure particle more nanoparticles can be
adsorbed if the core is neutral wet, nanoparticles desorption was not observed inside
water wet cores. Injection of high concentration of nano-structure particle fluid can block
core channels and result in permeability impairment, while for colloidal nanoparticles
fluid injection does not reduce permeability dramatically, on the contrary it makes core
more permeable for some cases.

INTRODUCTION
During last decade nanotechnology was proposed can be utilized in oil and gas industry
for different disciplines [1]. Particles can show some special properties when size reduce
down to nanoscale, like surface activity and huge specific surface area, so nanoparticles

SCA2015-029

2/12

have promising future to be a new EOR agent. The nanoparticles suspension fluid, so
called nanofluid, is a fluid containing nanometer-sized particles, and the dispersing
liquids can be water for hydrophilic nanoparticles. Based on many publications [2, 3, 4, 5]
addressed on this topic, nanofluid has already been proven to have good potential for
EOR.
Miranda et al. [6] have mentioned that silica nanoparticle has many advantages as EOR
agent, for instance, 1) 99.8% of silica nanoparticle is silicon dioxide (SiO2), which is
main component of sandstone, so silica nanoparticle is an environmentally friendly
material compared to chemical substance; 2) nanoparticles dispersion has good stability
because surface forces easily counterbalance the force of gravity; 3) the properties of
thermal, stressstrain and rheology strongly depend on size and shape of the
nanoparticles, and can be tailored during their production; 4) the chemical behavior of the
nanoparticle is correlated to the chemical substance of surface coating, the chemical
properties of nanoparticle can be easily controlled by changing surface coating chemical;
5) the price of silica nanoparticle is cheaper than chemical, which makes silica
nanoparticle can be widely applied for EOR at oil field.
The EOR mechanisms for nanofluid have already been discussed in previous authors
papers [7, 8], which include disjoining pressure, interfacial tension reduction, wettability
alteration, pore channels plugging and emulsification. The effect of nanoparticle
adsorption inside core on wettability alteration is highlight of these mechanisms, since
hydrophilic nanoparticles can alter oil wet and neutral wet core to water wet [9, 10]. The
adsorption of nanoparticles might also affect permeability of core, so this paper focuses
on adsorption behavior of hydrophilic silica nanoparticles inside different wettability core
and its effect on permeability.
Adsorption and Transport of Nanoparticles inside Porous Medium
After hydrophilic nanofluid is injected into porous medium, five phenomena will occur:
adsorption, desorption, blocking, transportation and aggregation of nanoparticles. Since
the particle size of nanoparticle is less than 1 micron, so they are Brownian particles, and
five forces dominate the interactions between nanoparticles and pore walls: the attractive
potential force of van der Waals, repulsion force of electric double layers, Born repulsion,
acid-base interaction, and hydrodynamics. When the total force of five forces is negative,
the attraction is larger than repulsion between nanoparticle and pore walls, which leads to
adsorption of nanoparticle on the pore walls. Otherwise desorption of nanoparticle from
the pore walls will occur at the same time. Adsorption and desorption is a dynamic
balance process controlled by the total force between nanoparticle and pore walls. Zhang
et al., [11] discussed that both reversible and irreversible adsorption of nanoparticles
occurs during transport through porous medium. Blocking will take place if the diameter
of the particle is larger than the size of pore throat, or when some nanoparticles aggregate
to form bigger particle at the pore throat. The aggregation of nanoparticles happens if the
previous equilibrium of the nanoparticle dispersion system breaks up and nanoparticles

SCA2015-029

3/12

form clusters to block some pore channels. Some images of adsorption and aggregate of
nanoparticles in porous medium are shown in Figure 1.

Figure 1 Nanoparticles adsorption inside porous medium.

Left: ESEM image of nanoparticle adsorption inside core [12]; right: Microscope image of
nanoparticles adsorption in glass micromodel

Transportation of nanoparticles in porous medium is governed by diffusion, convection

and hydrodynamics. After adsorption and desorption reach the equilibrium state,
nanofluid can flow through the porous medium without too much adsorption and
retention. The equilibrium adsorption is estimated to be 1.27 mg/g for 5000 ppm
nanofluid [13].

EXPERIMENTAL MATERIALS
Nanoparticle
Both hydrophilic silica Nano-Structure Particles (NSP) and hydrophilic silica Colloidal
NanoParticles (CNP) were employed in this experimental study. They are produced by
Evonik Industries. NSP and CNP are supplied as powder and highly concentrated
dispersion fluid respectively. They have been characterized by Transmission Electron
Microscope (TEM) and TEM images are shown in Figure 2. NSP have average primary
particle size of 7 nm and specific surface area of 300 m2/g, but they can aggregate to form
bigger particles, where particle size might be higher than 100nm. CNP have average
single particle size of 18 nm and specific surface areas of 350 m2/g, and this type of
nanoparticles dont aggregate in dispersion due to adding of special stabilizer. The reason
for NSP and CNP looking similar in Figure 2 is that CNP dispersion fluid was dried
before TEM imaging, so nanoparticles reaggregated and formed soft agglomerate, while
NSP always form hard agglomerate both in dispersion and powder status.

SCA2015-029

4/12

Figure 2 TEM images for NSP (left) and CNP (right) (Provided by Evonik)

Nanofluid
Three nanoparticles concentrations (0.05 wt. %, 0.2 wt. % and 0.5 wt. %) were utilized
for nanofluid, and 3 wt. % NaCl brine was used as dispersion fluid. NSP was weighed
and dispersed in brine by sonicator, while CNP nanofluid was diluted from concentrated
dispersion. Two types of nanofluid with different concentration are shown in Figure 3, as
we can see there is no big difference between different concentrations of CNP nanofluid
due to good nanoparticle dispersion, but for NSP the higher concentration the milkier
nanofluid will be. Fluid properties of each fluid are shown in Table 1.
Table 1 Fluid properties
Fluid

Density, g/cm3

Viscosity, cP

Brine Nacl 3 wt. %

1.022

1.0026

NSP Nanofluid 0.05 wt. %

1.021

1.0858

NSP Nanofluid 0.2 wt. %

1.022

1.1550

NSP Nanofluid 0.5 wt. %

1.022

1.5627

CNP Nanofluid 0.05 wt. %

1.022

1.0331

CNP Nanofluid 0.2 wt. %

1.022

1.0342

CNP Nanofluid 0.5 wt. %

1.022

1.0372

0.05
wt. %

0.2
wt. %

0.5

0.05

0.2

0.5

Figure 3 Nanofluid for NSP (left) and CNP (right)

SCA2015-029

5/12

Porous Medium
In total, 18 long core plugs drilled from one block of Berea Sandstone were used for this
nanoparticles transport experiment. The average porosity and permeability are 19.5% and
352 mD respectively. The diameter is 3.83cm and length is 8 cm. The pore volume (PV)
is about 16ml. Two groups (6 cores for each) were aged to be oil wet and neutral wet
under the same condition and the wettability index for oil wet core and neutral wet core is
-0.39 and -0.09 respectively. The wettability index for original water wet core is 0.8.
Flooding Setup
Figure 5 shows schematic of flooding setup. The pump injected Exxol D-60 as pumping
fluid to push the piston located inside the reservoir. There are 3 reservoirs filled with
brine, oil and nanofluid respectively. The pressure drop across the core plug during
nanoparticle transport experiments was recorded by precision pressure gauge.
15
10
4

4
8

17

11

18

6
5

6
5

13
12

14

16

19

1) Pump fluid (Exxol D60); 2) injection line; 3) Quzix Pump; 4) Valve; 5) Pump Fluid in reservoir; 6) Piston plate; 7) Brine in
reservoir-A; 8) Oil in reservoir-B; 9) Nanofluid in reservoir-C; 10) Oil line; 11) Brine/Nanofluid line; 12) Bypass Valve; 13)Hassler
Core Cell; 14) Core plug ; 15) Pressure gauge; 16) Sleeve pressure; 17) connection cable; 18) Computer; 19) Accumulator

Figure 4 Schematic of flooding setup

EXPERIMENTAL METHODS
The core plugs were saturated by 3 wt. % brine using a vacuum pump to ensure there was
no trapped air inside. Firstly, about 1 PV of brine was injected to measure core absolute
permeability, and then about 4 or 5 PVs NSP or CNP nanofluid injection with different
concentrations was followed to evaluate effect of nanoparticles adsorption and retention
on permeability. Finally, brine injection was conducted as post-flush to observe
desorption of nanoparticles, flow rate of 2 ml/min was utilized. Pressure drop across the
core was recorded during whole injection process. Effluent fluid was collected every 4ml
(1/4 PV) for NSP nanofluid flooding experiments, nanoparticles concentration was
measured by using UV Spectrophotometer afterwards, 4-5 measurements were conducted
for one sample to get average result.

SCA2015-029

6/12

RESULTS AND DISCUSSION

Oil Wet Core
Figure 5 shows the pressure drop of oil wet core flooding experiments for NSP and CNP
nanofluid with different concentrations. There is big difference of pressure drop curve
between NSP and CNP nanofluid injection cases. For NSP injection case pressure drop
increased rapidly after nanoflud injection and the higher concentration the higher
pressure drop, at end of nanofluid injection pressure was still continuing to climb and far
away from equilibrium. During post-flush brine injection of NSP case pressure drop
decreased gradually for all of three cores. Compared to adsorption process desorption
process is slow but quite significant, pressure drop decreased about 30% during postflush injection. CNP nanofluid injection showed different pressure drop curves, as Figure
5 right shows. There is no rapid pressure increase after CNP nanofluid injection and
pressure still kept the same during post-flush period. The different nanoparticles
adsorption behavior indicates the adsorption of NSP nanoparticles is multilayer while
CNP might be monolayer.

Figure 5 Pressure drop curves for oil wet cores (left: NSP, right: CNP)

Figure 6 shows dimensionless nanoparticles concentration curve of NSP nanoparticles

transport through oil wet core plugs, the dimensionless nanoparticles concentration is
defined as the ratio of effluent nanoparticles concentration to the injection nanoparticles
concentration. As shown in Figure 6, 0.05 wt. % NSP nanofluid have earliest
breakthrough while 0.5 wt. % NSP nanofluid have latest breakthrough. Effluent
nanoparticles concentration can stay on plateau for about 3 to 4 PV and the lower
concentration the longer plateau will be. For 0.2 and 0.5 wt. % cases concentration
vibration is observed during plateau, the reason might be due to discontinuous adsorption
and desorption, which means for injection of higher concentration nanofluid large amount
of nanoparticles adsorbed and trapped inside core and resulted in blocking of pore
channels as well as increase of pressure drop. While when pressure drop increased high
enough adsorption or retention of nanoparticles can be detached and flushed out of core
and then more adsorption and retention will happen again to recover previous balance
until next breakthrough, this lead to decline of the effluent concentration. Similar
vibration was also observed in pressure drop curve. The effluent concentration of three

SCA2015-029

7/12

cases decreased at about 5 PV during post-flushing. A tail of concentration curve was

shown up after brine breakthrough, which means desorption of nanoparticles during postflush injection.
Dimensionless Nanoparticles Concentration

O.W. NSP 0.05 wt.%

O.W. NSP 0.2 wt.%

0.9

O.W. NSP 0.5 wt.%

0.8
0.7
0.6
0.5
0.4
0.3
0.2

Post-flush Injection

Nanofluid Injection
0.1
0
0

Pore Volume

Figure 6 Effluent NSP nanoparticles concentration curves for oil wet cores

Neutral Wet Core

Pressure drop curves were plotted in Figure 7 for neutral wet core NSP nanofluid
injection, three curves were plotted separately due to big scale difference. Similar with
previous case, pressure drop increase very fast after nanofluid injection. The higher
concentration the faster pressure drop increase will be. For 0.5 wt. % case both nanofluid
injection and post-flush injection were terminated earlier because pressure almost reached
to maximum limit of pressure gauge. Pressure drop decline result from nanoparticles
desorption was observed. Permeability change before nanofluid and after post-flush was
list in Table 2.

Figure 7 Pressure drop for neutral wet cores NSP nanofluid (left: 0.05 wt.%, middle: 0.2 wt.%, right: 0.5 wt.%)

Figure 8 shows pressure drop for neutral wet CNP injection, the pressure drop didnt
change too much before and after nanofluid injection. Which means for neutral wet core
adsorption of CNP is still monolayer while NSP adsorption is multilayers. For 0.2 and 0.5

SCA2015-029

8/12

wt. % cases after nanofluid injection pressure drop decrease somehow, which might mean
that permeability increased. Table 2 shows that percentage of permeability change.

Figure 8 Pressure drop for neutral wet cores CNP nanofluid

Injection scenario
O.W. NSP 0.05 wt.%

Table 2 Permeability of core plugs

K1, mD (before
K2, mD (after
NP injection)
NP injection)
326.2
85.5

K2/K1, %
26.20

O.W. NSP 0.2 wt.%

361.8

41.1

11.37

O.W. NSP 0.5 wt.%

526.3

33.0

6.27

O.W. CNP 0.05 wt.%

428.9

509.0

118.70

O.W. CNP 0.2 wt.%

321.6

303.2

94.26

O.W. CNP 0.5 wt.%

210.5

228.8

108.66

N.W. NSP 0.05 wt.%

511.4

47.1

9.20

N.W. NSP 0.2 wt.%

326.7

1.8

0.55

N.W. NSP 0.5 wt.%

232.9

0.7

0.29

N.W. CNP 0.05 wt.%

247.6

237.6

95.96

N.W. CNP 0.2 wt.%

427.7

871.3

203.70

N.W. CNP 0.5 wt.%

522.8

691.9

132.35

W.W. NSP 0.05 wt.%

269.3

18.1

6.72

W.W. NSP 0.2 wt.%

463.2

5.9

1.28

W.W. NSP 0.5 wt.%

326.2

1.6

0.48

W.W. CNP 0.05 wt.%

367.6

361.8

98.44

W.W. CNP 0.2 wt.%

165.4

171.5

103.70

W.W. CNP 0.5 wt.%

308.8

361.8

117.19

Figure 9 shows effluent concentration of NSP nanoparticle for neutral wet core, 0.5 wt. %
was not presented because not enough effluent samples were collected. 0.05 wt. % case

SCA2015-029

9/12

have earlier breakthrough than another one and for both of curves effluent concentration
started to decrease slowly when it reach to the peak. The dimensionless concentration
reached only to about 40% maximally, which means most of nanoparticles were trapped
inside core. During post-flush large amount of nanoparticles can be measured even at late
of post-flush for 0.2 wt. % case.
1

Dimensionless Nanoparticles Concentration

N.W. NSP 0.05 wt.%

N.W. NSP 0.2 wt.%

0.9
0.8
0.7
0.6

0.5
0.4
0.3
0.2

Post-flush Injection

Nanofluid Injection

0.1
0
0

Pore Volume

Figure 9 Effluent NSP nanoparticles concentration curves for neutral wet cores

Water Wet Core

Figure 10 left shows NSP nanofluid injection for water wet core, after nanofluid injection
pressure drop behavior is similar with previous flooding cases. At end of nanofluid
injection pressure curves of 0.5 wt. % case is much higher than another two cases. During
post-flush pressure drop increased to plateau and kept constant until end of flooding for
all of three concentrations nanofluid cases, which might mean that for water wet core
there is no significant desorption of nanoparticles. Figure 10 right indicates that there is
no significant pressure drop difference between CNP nanofluid injection and post-flush
injection, meaning CNP adsorption is monolayer and will not impair permeability of core.
Detail data of effect of nanoparticles adsorption on permeability change can be found in
Table 2.

Figure 10 Pressure drop curves for water wet cores (left: NSP, right: CNP)

SCA2015-029

10/12

Effluent NSP nanoparticles concentration curves were plotted versus time and shown in
Figure 11. Higher concentration nanofluid case had earlier breakthrough than lower
concentration cases, but lower concentration nanofluid curve reached to peak faster than
higher cases. The effluent concentration of all the curves decreases immediately after
reaching the peak and concentration almost reduced to 0 at 5 PV, which shows a big
amount of adsorption and retention of nanoparticles inside the core. The possible reason
might be self-adsorption of nanoparticles, which means that the previous adsorbed
nanoparticles can adsorb nanoparticles injected afterwards, so when adsorbed
nanoparticles accumulate more enough the following injected nanoparticles cannot pass
through core easily as before. During post-flush injection almost no desorption of
nanoparticles after 5 PV, this is consistent with the conclusion from pressure drop.
1

Dimensiionless Nanoparticles Concentration

W.W. NSP 0.05 wt.%

Nanofluid Injection

0.9

W.W. NSP 0.2 wt.%

W.W. NSP 0.5 wt.%

0.8
0.7
0.6
0.5

0.4
0.3
0.2

Post-flush Injection
0.1
0
0

Pore Volume

Figure 11 Effluent NSP nanoparticles concentration curves for water wet cores

Comparison of Nanoparticles Adsorption Behavior Against Different Wettabilities

As shown in above figures nanoparticles adsorption behavior is different for different
wettability core. First of all, for higher concentration NSP nanofluid injection, neutral wet
cores have higher pressure drop than other wettability cores at the same injection volume,
while oil wet cores have lowest pressure drop during nanofluid injection. As shown in
Figure 12 oil wet cores have highest recovery for nanoparticles, 80 to 90% nanoparticles
can pass through core plug. Water wet cores have different recovery for various
concentrations of nanofluid, the higher concentration of nanofluid the more nanoparticles
can be recovered. While the neutral wet cores have lower dimensionless concentration,
the maximum value is only 40%. Based on all of pressure drop data and effluent
concentration curves, we can conclude that oil wet cores have lowest adsorption ability
for nanoparticles and for higher concentration nanofluid neutral wet cores have highest
adsorption ability of nanoparticles but for lower concentration of nanofluid water wet
core can adsorb more than others. Desorption of nanoparticles was observed in oil wet
and neutral wet cases, but was not presented in water wet cores as discussed above. The
core permeability for each core at end of post-flush injection was calculated and listed in
Table 2. The permeability for all of core plugs injected by NSP nanofluid has been

SCA2015-029

11/12

impaired dramatically, some of them reduced to less than 1% compared with original
value.
1

Nanofluid Injection

O.W. NSP 0.05 wt.%

Dimensiionless Nanoparticles Concentration

0.9

O.W. NSP 0.2 wt.%

O.W. NSP 0.5 wt.%

0.8

N.W. NSP 0.05 wt.%

N.W. NSP 0.2 wt.%

0.7

W.W. NSP 0.05 wt.%

W.W. NSP 0.2 wt.%

0.6

W.W. NSP 0.5 wt.%

0.5
0.4

Post-flush Injection
0.3
0.2
0.1
0
0

Pore Volume

Figure 12 Comparison of effluent nanoparticles concentration curves for different wettability

For CNP nanofluid injection, nanoparticles adsorption is monolayer and will not impair
core plugs permeability significantly, which means that after nanoparticles breakthrough
most of particles can recovered. During post-flush pressure drop didnt change too much
meaning desorption of CNP might be minor. There is no significant difference of CNP
adsorption for all three different wettability core plugs. Permeability of cores was
calculated and shown in Table 2, as we can see for some core plug permeability increased
significantly, especially for neutral wet 0.2 wt. % case permeability double increase. The
reason is still unclear, more experiments need to be done to find out the mechanism.

CONCLUSION
1. NSP and CNP have different adsorption behavior inside core plug. Adsorption of NSP
is multilayer while CNP adsorption is monolayer. Amount of NSP adsorption is much
larger than CNP and result in permeability impairment.
2. Adsorption behavior of NSP is various with different wettability of core. For oil wet
core, most of nanoparticles can pass through core and plateau was present in
concentration curves. However, neutral wet core adsorbs a large amount of nanoparticles
resulting in high pressure drop. For water wet the higher nanoparticles concentration the
more nanoparticles can be recovered, and nanoparticles self-adsorption was observed.
3. NSP has different desorption behavior against different wettability core plug,
desorption happened during post-plush for both oil wet and neutral wet core, but there is
no significant desorption of nanoparticles was observed for water wet core.
4. CNP adsorption is independent of wettability and nanoparticles concentration.
Injection of CNP nanofluid will not impair core permeability, on the contrary it will
increase permeability for some cores.

SCA2015-029

12/12

REFERENCES
1. Xiangling, K. and Ohadi, M.M. 2010. Applications of Micro and Nano Technologies
in the Oil and Gas Industry - Overview of the Recent Progress. Paper presented at Abu
Dhabi International Petroleum Exhibition and Conference, 1-4 November, Abu Dhabi.
2. Binshan, J., Tailiang, F. and Mingxue, M. 2006. Enhanced Oil Recovery by Flooding
with Hydrophilic Nanoparticles. China Particuology 4, 41-46.
3. Suleimanov, B. A., Ismailov, F.S., and Veliyev, E.F. 2011. Nanofluid for enhanced oil
recovery. Journal of Petroleum Science and Engineering 78, 431-437.
4. Hendraningrat, L., Li, S., and Torster, O. 2013. Enhancing Oil Recovery of LowPermeability Berea Sandstone through Optimized Nanofluid Concentration. Paper
presented at SPE Enhanced Oil Recovery Conference, Kuala Lumpur, Malaysia, 2-4 July.
5. Skauge, T., Spildo, K., and Skauge, A. 2010. Nano-sized Particles For EOR. Paper
SPE 129933-MS presented at SPE Improved Oil Recovery Symposium, 24-28 April,
Tulsa, Oklahoma, USA
6. Miranda, C.R., De Lara, L, S., and Tonetto, B, X. 2012. Stability and Mobility of
Functionalized Silica Nanoparticles for Enhanced Oil Recovery Application. Paper
presented at SPE International Oilfield Technology Conference, 12-14 June, Noordwijk,
7. Li, S., Hendraningrat, L., and Torster, O. 2013. Improved Oil Recovery by
Hydrophilic Silica Nanoparticles Suspension 2-Phase Flow Experimental Studies. Paper
IPTC-16707 presented at the International Petroleum Technology Conference held in
Beijing, China, 2628 March 2013.
8. Li, S., and Torster, O. 2014. An Experimental Investigation of EOR Mechanisms for
Nanoparticles Fluid in Glass Micromodel. Paper SCA2014-022 was prepared for
presentation at the International Symposium of the Society of Core Analysts held in
Avignon, France, 8-11 September, 2014
9. Li, S., Torster, O. The Impact of Nanoparticles Adsorption and Transport on
Wettability Alteration of Intermediate Wet Berea Sandstone. Published on SPE Middle
East Unconventional Resources Conference held 26 28 January, 2015 in Muscat, Oman.
10. Li, S., Torster, O. Experimental Investigation of the Influence of Nanoparticles
Adsorption and Transport on Wettability Alteration for Oil Wet Berea Sandstone.
Published on 19th MEOS held 8 11 March, 2015 in Manama, Kingdom of Bahrain
11. Zhang, T., Murphy, M., Yu, H., Bagaria, H. G., Huh, C., and Bryant, S.L.
Investigation of Nanoparticle Adsorption during Transport in Porous Media. Paper was
prepared for presentation at the SPE ATCE held in New Orleans, 30 Sept.2 Oct. 2013.
12. Kanj, M.Y., Funk, J. J., and Yousif, Z. A. 2009. Nanofluid Coreflood Experiment in
the ARAB-D. Paper SPE-126161 was prepared for presentation at the SPE Saudia Arabia
Section Technical Symposium, 9-11 May, AlKhobar, Saudi
13. Yu, J., An, C., Mo, D., Liu, N., Lee, R. Study of Adsorption and transportation
behavior of Nanoparticles in Three Different Porous Media. Paper was prepared for
presentation at the SPE Improved Oil Recovery Symposium, 14-18 April, Tulsa, USA
(2012).

SCA2015-030

1/12

EFFECT OF HYDROSTATIC LOADING ON THE PORE

STRUCTURE AND TRANSFER PROPERTIES OF A
TIGHT GAS SANDSTONE
Yi WANG(1), Catherine DAVY(1), Franck AGOSTINI(1), Frdric SKOCZYLAS(1),
Laurent JEANNIN(2)
(1): LML UMR CNRS 8107/Ecole Centrale de Lille, CS20048,
59651 Villeneuve dAscq Cedex, France
(2): GDFSUEZ E&P International SA, 1 place Samuel de Champlain,
92930 Paris La Dfense cedex - France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Our study focuses on tight gas sandstones from a field located in North Africa and
explored by GDFSUEZ E&P. This rock is characterized by its low porosity (below 10%)
and its low absolute permeability (below 0.1mD i.e. 10-16m2 under ambient conditions). A
dedicated experimental setup has been designed to allow simultaneous measurements of
gas permeability, connected porosity and poro-elastic properties, under given hydrostatic
loading (up to 60 MPa). In the dry state, the decrease in accessible porosity is
demonstrated experimentally: for one sample, accessible porosity is reduced by more
than 10% (relative value) under 40 MPa hydrostatic loading. Simultaneously, gas
permeability reduction is observed of a factor 8. In the partial water saturation state, the
decrease in accessible porosity and effective gas permeability is enhanced. This shows an
important stress sensitivity of the petrophysical properties of the sandstone: it is
interpreted as pore entrapment under hydrostatic loading.

INTRODUCTION
Tight gas reservoirs are unconventional gas reservoirs constituted of low permeability
sandstones, which are characterized by low connected porosity (lower than 10%), low
absolute permeability (below 0.1mD i.e. 10-16m2 under ambient conditions), and strong
sensitivity to in situ stress as compared to conventional reservoirs [1-3]. Although these
unfavorable properties make them more complicated to explore than conventional ones,
tight reservoirs have a great potential in terms of gas production [4,5].
For their economical exploitation, characterization of the petrophysical properties of tight
reservoirs is necessary. This paper presents the assessment of the effect of hydrostatic
stress on the petrophysical properties of a particular sandstone, in both dry state and
partially water-saturated state. We also relate the effect of hydrostatic stress to the
changes in pore structure, by assessing the changes in connected pore volume.

SCA2015-030

2/12

Firstly, gas permeability and connected pore volume are investigated in the dry and in the
partially water saturated state, under different confining pressures. Secondly, we compare
variations in pore volume during loading and unloading to sample deformation, in order
to confirm our interpretation of pore entrapment. Thirdly, we measure sandstone poroelastic properties, in order to obtain a relationship between the externally applied stress
and the drained bulk modulus K b (or the solid matrix bulk modulus K s ). This allows
getting further insight into the pore structure changes.

EXPERIMENTAL METHODS
Sample Preparation And Water Porosity Measurement
The samples of tight gas sandstone used in this study originate from a field located in
North Africa [6], from depths between -2000 m and -2550 m. They are received by our
laboratory as cylinders with a diameter of 37 mm, and a length cut to 60 mm.
Water porosity w is defined as:

w msat mdry / wVsample

(1)

where msat and mdry are respectively the water-saturated mass and the dry mass, w is
water density and Vsample is the sample bulk volume. mdry is obtained by oven-drying the
sample at 105C, and regular weighing until the mass does not change by 0.01g for one
whole week. Similarly, for msat , samples are water-saturated under vacuum until mass
stabilization.
Partial Water Saturation State And Sample Conditioning
Partial water saturation S w is defined as:

S w m mdry / msat mdry

(2)

In order to achieve a given Sw, a specific method is used, which corresponds to that
recommended by RILEM [7], as follows. Firstly, sample water-saturated and dry masses
are assessed; the dry sample is then water saturated until reaching the desired mass m. It
is then sealed in three aluminium layers, and one layer of paraffin. Following this, it is
kept in a climatic chamber at 40 C for at least 14 days, to allow water to have a
homogeneous distribution in the sample.
Gas Permeability Measurement
Gas permeability of each sandstone sample is measured with the quasi-stationary fluid
flow method [8,9]. The test is performed in a hydrostatic cell where the sample is sealed
in a Viton Jacket (Figure 1). A buffer reservoir with a manometer is set up outside the
hydrostatic cell between the source of gas and the sample. At first, when valve V3 is
closed and V1, V2 and V4 are open, gas flows from the source of gas, via the buffer
reservoir, until reaching a steady state. At steady state, the upstream side of the sample is

SCA2015-030

3/12

at pressure P1 , while the pressure on the downstream sample side P0 is equal to the
atmospheric pressure. Then, valve V1 is closed at time t 0 , and we measure the time
necessary for upstream gas pressure to decrease by P1 (P1 is less than one 1% of P1).
During t , gas is assumed to flow steadily at an average pressure Pmean P1 P1 / 2 on
the upstream side, and the average volume flowrate is given by:
V P
Qmean 1 1
(3)
Pmean t
where V1 is the buffer reservoir volume. Finally, gas permeability is given by simplified
Darcys law [10]:
Qmean 2 LPmean
Kx
(4)
2
A Pmean
P02
where L is sample length, A is its sectional area and is the dynamic viscosity of gas
(here Argon is used with 2.2 10 5 Pa s ).

Figure 1 Principle for the quasi-stationary fluid flow method for gas permeability measurement

SCA2015-030

4/12

Measurement of Gas-accessible Pore Volume

This test quantifies the accessible pore volume under a given confining pressure [11,12].
The interpretation of this method is based on the ideal gas law:
(5)
PV nRT
As shown in Figure 1, the sample is placed in the hydrostatic cell under given confining
pressure. The buffer reservoir, of known volume Vr , is isolated from the rest of the
circuit (by closing valves V2 and V3), and it is filled with gas at an initial pressure Pinitial
(this is achieved by closing valve V1 when reaching the desired pressure) while the
sample is at atmospheric pressure. While valve V4 is closed, the opening of valves V2 and
V3 connects the buffer reservoir to the sample by both ends, while all other valves (V1
and V4) remain closed. From this moment, gas enters the sample connected pores and
pressure P measured in the buffer reservoir decreases until reaching stabilization at a
value Pfinal . Owing to the conservation of the gas mass initially located in the buffer
reservoir, the relationship between Pinitial and Pfinal writes:

PinitialVr = nRT = Pfinal (Vr +Vpore )

So that the accessible pore volume is determined by:

P
Vpore = ( initial -1)Vr
Pfinal
If the sample is in the dry state, the porosity accessible to gas is deduced as:
V
g pore
Vsample

(6)

(7)

(8)

where V pore is calculated by Equation (7) and sample volume Vsample is determined by
hydrostatic weighing (using Archimedes law): first, the dry sample is weighted, then it is
submerged in water, and its weight under water (equal to its dry weight minus the weight
of water occupying the volume of the sample) is measured. Sample volume is the
difference between the two weights divided by the density of water.
Poro-Elastic Property Measurement
The measurement of poro-elastic properties is based on the theory of Biot [14]. The
saturated porous material is considered as the superposition of two continua: the solid
matrix and the fluid present in the pores. In the following, in a first approach, the porous
medium (solid matrix+pores) is considered isotropic.
The drained bulk modulus K b represents the deformability of the whole material i.e. of
the so-called solid skeleton. The solid skeleton is constituted of connected and nonconnected pores, and of the solid matrix [13]. K b is measured under constant pore
pressure (often equal to atmospheric pressure, as used here), by varying the hydrostatic
stress by a small Pc , so that K b is given by:

SCA2015-030

5/12

Pc
(9)
v1
is the volumetric strain variation due to the variation in confining pressure
Kb

where v1
Pc .

The solid matrix bulk modulus K s represents the rigidity of the solid matrix, which is
composed of the solid and of the non-connected porosity alone [14]. In order to
determine directly K s , the variation in sample volumetric strain De v is measured when
the same variation in interstitial pore pressure Dp and in confining pressure DPc = Dp is
achieved:
Dp
Ks =
(10)
De v
However, in order to avoid dealing with equal gas and confining oil pressures (which
would mean extensive test leakage) , p Pc may be difficult to obtain. K s is rather
calculated from the measurement of modulus H, which is by varying pore pressure by an
amount p when hydrostatic stress is kept constant [11-13], as:
p
H
(11)
v 2
where v 2 is the volumetric strain variation due to the variation in pore pressure p .
Finally K s is given by [11-13]:
1
1
1

(12)
K s Kb H
Generally, strain gauges are used to measure sample longitudinal and radial deformations.
In this study, four LVDT sensors are used to measure longitudinal strains (Figure 2). As
our sandstone samples are considered isotropic, the volumetric deformation v is
calculated by:
3
v 1 2 3 4
(13)
4
When compared to strain gauges, using LVDT sensors has several advantages:
1) Sample length changes are measured directly without sample surface preparation
and the assembly of the hydrostatic cell is simplified.
2) The problem of strain gauge bonding, especially on partially saturated samples is
avoided.
3) The bonding of strain gauges and electric wire soldering can suffer from
saturation/drying cycles, and from confining pressure changes.

SCA2015-030

6/12

4) By avoiding sample surface preparation and the use of glue, the measurement of
sample mass variation before and after the experiment is more accurate.

Figure 2: LVDT measurement system

RESULTS AND DISCUSSION

Water porosity
Water porosity results (Table 1) indicate that the tested samples have small water
porosity values, with variations from 1.52% to 4.96%. No clear correlation is found
between water porosity and drilling depth. This absence of correlation is expected [15,
17, 18]. Drilling depth has some influence on porosity, but it is not the determining
factor: porosity also depends on diagenetic conditions (e.g. in situ chemical
transformations of minerals) [6].
Table 1 Water porosity of tight sandstone samples
Number
Water porosity (%)
Depth (m)

2335
4.96
2359

3248
3.37
2181

3249
3.83
2080

3250
3.88
2197

3372
2.05
2501

3375
2.93
2496

3377
1.81
2511

3379
1.52
2516

4456
2.54
2312

4458
2.78
2311

SCA2015-030

7/12

Gas Permeability In The Dry State

Normalized dry gas permeability is defined as K drynorm K Pc / K Pc 3MPa , where
K Pc is dry gas permeability under confining pressure Pc , K Pc 3MPa is that under

confining pressure Pc 3MPa . Figure 3 illustrates the evolution of normalized dry gas
permeability over a range of confining pressures from 3 MPa to 40 MPa. A decrease in
Kdrynorm by 90% is observed for the three samples presented in Figure 3, when hydrostatic
stress reaches 40 MPa. This underlines the important stress sensitivity of the material.
Samples 3249 and 3248 show almost the same type of Kdrynorm variation while 3379
seems to be more stress-sensitive. This is compared to the accessible porosity of those
samples: samples 3249 and 3248 have comparable porosities (3.83% and 3.37%
respectively), while sample 3379, which is more sensitive to Pc changes, has a lower
porosity (of about 1.52%). This has already been observed in a previous study on tight
Rotliegend sandstones from Germany [15].
1
3249

Normalized gas permeability

K(Pc)/K(Pc=3 MPa)

0.9

3248

0.8

3379

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

20
30
Confining pressure (MPa)

40

50

Figure 3: Normalized gas permeability for dry sandstone samples

Porosity And Pore Volume Variation In The Dry State

Similar stress-sensitivity is observed with normalized porosity, which is defined as
norm Pc / Pc 3MPa (Figure 4). Samples 3249 and 3248 lose nearly 10% of their
porosity when confining pressure varies from 3 MPa to 40 MPa, while sample 3379 loses
25% of porosity. The loss of porosity is so large, that it is difficult to explain by the poroelastic behaviour of the sandstone in terms of volume changes. It may well be indicative
of the entrapment of pores in the solid skeleton.
To confirm this assumption, let us take sample 3248 as an example. The variation of pore
volume as a function of sample volume variations (given by LVDT sensors) is shown in
Figure 5. According to the poro-mechanics theory for isotropic media, a value of Biot
coefficient equal to 1 implies that the variation in pore volume is identical to the variation

SCA2015-030

8/12

in sample volume: this is generally observed for granular media. For cohesive rocks such
as tight sandstones, Biots coefficient is smaller than 1, so that the variation in sample
volume is greater than the variation in pore volume [14]. However, the reduction in pore
volume is greater than the reduction in sample volume (Figure 5), even for moderate Pc
(between 3 MPa and 10 MPa). This is explained by the closing of micro-cracks in the
pore structure, which induces an occlusion of pores in the solid skeleton [12,15]. Above
Pc = 10 MPa, accessible pore volume variation is parallel to the theoretical line
corresponding to Biots coefficient=1. Our interpretation is that micro-cracks are closed
below Pc = 10 MPa: above this value, the material behavior is governed by its poroelastic properties.
1

Normalized porosity
(Pc)/(Pc=3 MPa)

0.95
0.9
0.85
0.8
0.75
0.7
3249

0.65

3248
3379

0.6
0

10

20
30
Confining pressure (MPa)

40

50

Figure 4 Normalized porosity for dry samples

Sample volume variation (cm3)

-0.25

-0.2

-0.15

-0.1

-0.05

0
3 MPa

Load

0
-0.05

Unload

3 MPa

Biot coefficient=1

-0.1
-0.15

10 MPa
20 MPa

40 MPa

-0.2
10 MPa

20 MPa

-0.25
-0.3

Pore volume variation (cm3)

-0.3

-0.35
-0.4
Figure 5 Comparison between pore volume variation and sample volume variation

SCA2015-030

9/12

Poro-elastic Properties In The Dry State

In order to confirm the changes in pore structure (entrapment under increasing Pc), poroelastic properties K b and K s are assessed under different confining pressures. As shown
in Figure 6, for the three samples, the effect of Pc is more important for low confinement.
Between Pc = 3 and 10 MPa the increase in K b reaches about 200%, and increases an
additional 50% between Pc = 10 and 40 MPa. Finally at Pc = 40 MPa, K b increases to
about 20 GPa. Similar stiffness values are also observed for other such tight sandstones
[14].
25

Kb (GPa)

20

15

10

3249
3248
3379

0
0

10

20
30
Confining pressure (MPa)

40

50

Figure 6 Drained bulk modulus K b as a function of confining pressure

However, difficulties have been encountered when measuring the solid matrix bulk
modulus K s , which are quite high (above 20 GPa). When an interstitial pore pressure is
applied, only small deformations of the solid matrix occur, which induces an important
uncertainty to the measurement. Further effort shall be dedicated to accurately assess the
value of K s , as in [14].
Gas Permeability In A Partially Water Saturated State
In the partially water saturated state, normalized gas permeability is the ratio between gas
permeability in the partially water saturated state at given Pc and gas permeability in the
dry state at Pc =3MPa: K norm K Pc / K Pc 3MPa . It is plotted under increasing
confining pressure for sample 3249 in Figure 7. It is observed that Knorm is more stresssensitive than in the dry state. For Sw = 25.6%, between Pc = 3 and 10MPa, normalized
permeability has decreased significantly, by 70%. The sharp decrease continues until
confining pressure reaches Pc = 20 MPa. Above this confining pressure value, normalized
permeability shows little variation. This behavior is highly dependent on the structure of
the pore network and on pore size distribution. The decrease in Knorm when Pc increases
indicates that the pores, which have closed under increasing Pc, were significantly

SCA2015-030

10/12

participating in gas flow, so that they were not filled with water, although the material is
in a partial water saturated state. Further interpretation will be conducted on this aspect
with the help of Mercury Intrusion Porosimetry (MIP), Scanning Electron Microscopy
(SEM) investigations and micro mechanical modelling [16].

Normalized gas permeability

K(Pc)/K(Pc=3 MPa)

1
0.9

dry state

0.8

Sw=0.256

0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0

10

20

30

40

50

Confining pressure (MPa)

Figure 7 Normalized gas permeability of sample 3249 in the dry and partially water saturated states

Normalized gas-accessible pore volume

Vpore(Pc)/Vpore(Pc=3 MPa)

Gas Accessible Pore Volume In A Partially Water Saturated State

The water saturation has a greater influence on gas-accessible pore volume, than on
normalized permeability (
Figure 8). Large reductions are observed for Sw = 25.6%, especially when confining
pressure Pc increases from 10 MPa to 20 MPa: a large portion of nearly 30% of the gasaccessible pore volume is lost.
1
0.95
0.9
0.85
0.8
0.75
0.7
0.65
0.6

dry state

0.55
Sw=0.256

0.5
0

10

20

30

40

50

Confining pressure (MPa)

Figure 8 Normalized gas-accessible pore volume of sample 3249 in the dry and partially water saturated
states

SCA2015-030

11/12

From Figure 7 and

Figure 8, we conclude that in the partially water saturated state, the evolution of
permeability and gas-accessible pore volume do not evolve so smoothly as in the dry
state: in the partially water saturated state, the major decrease in permeability occurs at
the beginning of loading, while relative pore volume decreases mainly under intermediate
confining pressures (10MPa to 20MPa). Further investigations and a dedicated micromechanical model are required for a proper interpretation.

CONCLUSION
This experimental study deals with samples from a tight gas reservoir. First, gas
permeability is investigated under different confining pressures. For samples in the dry
state, we show a great stress-sensitivity of normalized gas permeability. Additionally,
sample 3379, with a smaller initial porosity, is more sensitive to external stress than the
other two samples. The partial water saturation of the samples increases the sensitivity of
permeability to changes in Pc. Secondly, gas accessible pore volume is measured under
confining pressure. Large losses in pore volume are observed in dry state, which is
attributed to pore entrapment, which occurs when micro-cracks close under increasing
confining pressure. This phenomenon is more visible in a partially water saturated state.
The variation in pore volume is compared with that of sample volume: under increasing
Pc, a large loss of pore volume is observed, which is also an evidence of pore entrapment.
Thirdly, poro-elastic experiments are conducted to provide further information on sample
structure: it is observed that the drained bulk modulus K b increases steadily under
increasing Pc loading.
The interpretation of those tests requires further investigation. Rock micro-structure shall
be assessed by Scanning Electron Microscopy (SEM) and Mercury Intrusion Porosimetry
(MIP). The use of a micromechanical model will provide a prediction of petrophysical
properties for the reservoir.

ACKNOWLEDGEMENTS
We gratefully acknowledge financial support by GDFSUEZ E&P International SA.

REFERENCES
1. Cluff, R. M., and A. P. Byrnes. "Relative Permeability In Tight Gas Sandstone
Reservoirs-The Permeability Jail Model." SPWLA 51st Annual Logging Symposium.
Society of Petrophysicists and Well-Log Analysts, (2010).
2. Blasingame, T. A. "The characteristic flow behavior of low-permeability reservoir
systems." SPE Unconventional Reservoirs Conference. Keystone, Colorado, USA.
(2008).
3. Ward, J. S., and N. R. Morrow. "Capillary pressures and gas relative permeabilities of
low-permeability sandstone." SPE Formation Evaluation, (1987) vol.2, no.03, 345356.

SCA2015-030

12/12

4. Jiang, L. Z., J. Y. Gu, and B. C. Guo. "Characteristics and mechanism of low

permeability clastic reservoir in Chinese petroliferous basin." Acta Sedimentologica
Sinica, (2004) vol.22, no.1, 13-18.
5. Kang, Y. l., and P. Y. Luo. "Current status and prospect of key techniques for
exploration and production of tight sandstone gas reservoirs in China." Petroleum
Exploration and Development, (2007) vol.34, no.2, 239.
6. Tournier, F., et al. "Relationship between deep diagenetic quartz cementation and
sedimentary facies in a Late Ordovician glacial environment (Sbaa Basin, Algeria)."
Journal of Sedimentary Research, (2010) vol.80, no.12, 1068-1084.
7. Hilsdorf, H. K., and J. Kropp. "Permeability of Concrete as a criterion of its
durability." Report of RILEM Technical Committee TC, (1992) vol.116.
8. Skoczylas, F., Ecoulements et couplages fluide-squelette dans les milieux poreux:
tudes exprimentales et numriques., Dissertation, (1996).
9. Meziani, H., and F. Skoczylas, "An experimental study of the mechanical behaviour
of a mortar and of its permeability under deviatoric loading." Materials and
structures, (1999) vol.32, no.6, 403-409.
10. Davy, C. A., F. Skoczylas, P. Lebon, Th. Dubois, "Permeability of macro-cracked
argillite under confinement: gas and water testing." Physics and Chemistry of the
Earth, (2007) vol.32, no.8, 667-680.
11. Chen, X. T., Effet du chauffage sur le comportement mcanique et poro-mcanique
de matriaux cimentaires: proprits hydrauliques et changements morphologiques.,
PhD Dissertation, (2009).
12. Chen, X. T., G. Caratini, C. A. Davy, D. Troadec, F. Skoczylas, "Coupled transport
and poro-mechanical properties of a heat-treated mortar under confinement." Cement
and Concrete Research, (2013) vol.49, 10-20.
13. Coussy, O., Poromechanics. John Wiley & Sons, Chichester, (2004).
14. Duan, Z., C. A. Davy, F. Agostini, L. Jeannin, D. Troadec, F. Skoczylas "Gas
recovery potential of sandstones from tight gas reservoirs." International Journal of
Rock Mechanics and Mining Sciences, (2014) vol.65, 75-85.
15. Fu, X., et al. "Effect of mechanical loading and water saturation on the gas recovery
of tight gas: experimental study." 2012 International Symposium of the society of
Core Analysts. Aberdeen, SCA, (2012), 27-30.
16. Dormieux, L., L. Jeannin, and N. Gland. "Homogenized models of stress-sensitive
reservoir rocks." International Journal of Engineering Science, (2011) vol.49, no.5,
386-396.
17. Ramm, M., and K. Bjrlykke, 1994, Porosity/depth trends in reservoir sandstones:
Assessing the quantitative effects of varying pore-pressure, temperature history and
mineralogy, Norwegian shelf data: Clay Minerals, v. 29, p. 475 490.
18. Ehrenberg S. N., Nadeau P. H., Sandstone vs. carbonate petroleum reservoirs: A
global perspective on porosity-depth and porosity-permeability relationships, AAPG
Bulletin, v. 89, no. 4 (April 2005), pp. 435 445.

SCA2015-031

1/12

MICRO-SEISMIC EVALUATION OF FRACTURING IN

CORES DURING TRIAXIAL COMPRESSION TESTS
Y. Xiao, A. Abugharara and S.D. Butt
Advanced Drilling Laboratory, Memorial University of Newfoundland, St. Johns, NL,
Canada
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
This study is an evaluation of rock cracking and failure by means of laboratory standard
strength tests and real time micro-seismic or acoustic emission (AE) monitoring. Three
groups of rock-like materials were cast using fine aggregate and Portland cement, out of
which standard test specimens were cored. Confined compression strength (CCS) tests
were conducted on those cores while two non-destructive testing (NDT) sensors were
placed in end platens used to compress the core. Conventional rock mechanics results
were obtained such as stress-strain response. Hundreds of micro-seismic events were
recorded in the process of rock deformation and especially when the core failed. Seismic
data processing indicated the synchronization of event occurrence rate with correlated
material deformation. Also, micro-seismic properties were analyzed such as dominant
frequency (DF), event energy and cumulative AE counts. Event energy was found closely
related to the peak amplitude of seismic waves. Under the same confining pressure, DF
was prone to decrease with increase of deformation until the core failed. This correlated
with the higher AE event rate when deformation increased. High strength material tended
to generate higher DF than that of low strength material. For the same strength material,
increasing confining pressure played different roles on the dominant frequency. Finally,
AE event occurrence locations were determined along the core length which was
compared with the observation of core surface cracks.

INTRODUCTION
Micro-seismic events or AE are the elastic waves produced when rock undergoes internal
change, such as micro-crack initiation and propagation. Piezoelectric transducers are
commonly employed in detecting and monitoring micro-crack propagation [1]. As an
alternative way to see micro-crack initiation and propagation, AE detection was applied
in triaxial compression tests to monitor the whole deformation process [2]. Some AE
parameters such as DF, event energy and cumulative AE counts are related to the
different deformation stages [3, 4]. Source location was conducted to identify the
approximate location of AE events applied in hydraulic fracturing researches [5, 6, 7].
Numerical simulation was used to simulate the deformation process and predict the
failure [8, 9]. Crack type was also classified based on detected acoustic emissions for
failure prediction [10, 11, 12, 13].

SCA2015-032

2/12

EXPERIMENTAL SETUP
Monitoring AE during the CCS test was scheduled. The CCS tests were conducted using
a servo-controlled axial loading frame and a Hoek triaxial pressure cell. AE were
recorded using a two-sensor AE system.
Axial Loading Frame
During the CCS test, axial load was applied by the Instron load frame (Figure 1). The
maximum loading could be 250 kN. By setting the loading rate of 1 mm/min, extension
and load were recorded until the core specimen failed. The core was put into the Hoek
triaxial cell and specific confining pressure was loaded by a manually operated pump.

Figure 1. Strength test apparatus with the Hoek triaxial cell (left) and working schematic (right).

Monitoring Acoustic Emission

Two Panametrics p-wave sensors were placed into steel platens on both ends of core
specimens (Figure 1). Sponge material was put between the sensors and the loading
frame for complete contact assurance and p-wave couplant was put between the steel
platens and core ends and between sensors and steel platens to make better signal
transmittal. The central frequency of the p-wave sensors was 1.14 MHz with working
bandwidth of from 0.65 to 1.63 MHz at -6dB attenuation. Two PAC 2/4/6 preamplifiers
were utilized and gain selection of 40dB. Two customized power supply adapters were
connected to the preamplifiers with output voltage of 20 volt. The DAQ system was
comprised of GaGe CompuScope 8280 eight-channel board and its included DAQ
software. The trigger sensor was always located on the top of cores. Inputs for AE
detection are listed in Table 1.
Table 1. Inputs for monitoring acoustic emission.
Sampling frequency P-wave sensors #
Gain
10 MHz
2
40 dB

Peak-peak input
10 V

Trigger level
0.15 V

SCA2015-031

3/12

Test Materials
Three groups of rock-like materials were used with UCS (0 confining pressure) at 20,
55.5 and 87.5 MPa, designated as low, medium and high strengths (L, M and H) in
Figure 2. These materials were made of fine aggregate, Portland cement and water. This
type of rock-like materials has been used in all previous lab tests in this project based on
the ability of the reproducibility. Such tests were performed to study the drill-ability
including AE. In this paper, AE tests were conducted to investigate the deformation and
cracking properties which were valuable for future bit-rock interaction investigation, but
petroleum cores were not involved. Standard NQ cores were drilled with core diameter of
47.6 mm and the minimum ratio of height to diameter was 2:1. All the coring process and
requirements were done in accordance to ASTM D4543 [14]. Averaged core dimensions
and the loading plan are given in Table 2. To be consistent with the investigation of rocklike materials properties and mechanical response, confining pressures were chosen in
accordance to previous rock characterization tests.

Figure 2. Tested concrete cores (low, medium and high strength).

Table 2. Loading plan for triaxial compression test and CCS results.
Core
Length (mm)
Diameter (mm)
P-wave velocity (m/s)
#
L1
111.64
47.23
4304.0
L2
105.04
47.22
4304.0
M1
100.79
47.38
4785.4
M2
103.47
47.18
4785.4
H1
108.39
47.43
4710.4
H2
104.88
47.51
4710.4

Conf. Pres. (MPa)

2
4
2
4
2
4

CCS
(MPa)
33.90
41.48
59.84
75.36
105.65
116.11

Overall Workflow
The overall flow chart for this test is shown in Figure 3. Acoustic emission signals from
cracking were automatically detected and saved to the AE computer disk. At the same

SCA2015-032

4/12

time, CCS test was being conducted and the stress and strain data were recorded into the
other computer. Both the computers were synchronized before each test began.

Figure 3. Flow chart of CCS test with monitoring acoustic emissions.

RESULTS
The CCS tests [15] were conducted with the results in Table 2 and acoustic responses
were also obtained. Analysis of events includes DF, event energy, peak amplitude,
cumulative AE number and event location interpretation from the relative travel time was
numerically processed.
Single Acoustic Emission
A single event from a CCS test was obtained in Figure 4. The top sensor was always set
as the trigger channel and two bursts of signals were captured at both sensors. The
different first arrival time demonstrated that the AE source located closer to the top
sensor. AE parameters were calculated based on methodology previously developed [3].

Figure 4. One burst event from top sensor (upper) and bottom sensor (lower) detected from the high
strength material with 4 MPa confining pressure.

SCA2015-031

5/12

Single Test Inspection

AE event locations were determined and only those from inside the cores were kept. The
mechanical response and acoustic properties of low, medium and high strength cores
were plotted together for comparison in Figure 5-7. Stress and cumulative AE counts
were plotted on the same time base. Before the linear loading section, few AE events
were detected. AE events initiated at the end of the linear loading sections and the
number increased within non-linear ductile deformation section. This was explained by
the initiation of micro-cracks and micro-crack connection. AE rate dramatically increased
before and after the core failed. This was due to the micro-crack propagation and crack
nucleation that was continuously generated.
Event energy and peak amplitude were also investigated from both sensors. Event energy
was found closely correlated with peak amplitude. And energy from both sensors
correlated with each other. This indicated that a single channel of signals could be used
for AE analysis.
For medium strength cores, limited AE events were detected due to less capability of
signal transmittal between core surfaces and sensors.

SCA2015-032

6/12

Figure 5. Mechanical and acoustic responses of low strength cores with acoustic properties comparison
from both sensors.

Figure 6. Mechanical and acoustic responses of medium strength cores with acoustic properties
comparison from both sensors.

SCA2015-031

7/12

Figure 7. Mechanical and acoustic responses of high strength cores with acoustic properties comparison
from both sensors.

Dominant Frequency and Source Energy

DF and AE event energy of signals from the top trigger sensor were compared for all
tests in Figure 8 and Figure 9, separately. Linear fit was plotted for each group of
scattered points to the test time. Under the same confining pressure, DF tends to decrease
with the increase of core deformation. For the low strength material, DF stays constant
during the rock deformation process. There is no evidence to show the relation of
material strength and the DF level.
AE event energy generally decreases with increased DF under the same confining
pressure. An exception exists that event energy increases with increased DF for the
medium strength material under the confining pressure of 2 MPa.

SCA2015-032

8/12

Figure 8. DF versus time for low, medium and high strength cores under confining pressure of 2 MPa
(upper) and 4 MPa (lower). Linear fit was provided for each group of scattered points.

SCA2015-031

9/12

Figure 9. AE energy versus dominant frequency for low, medium and high strength cores under confining
pressure of 2 MPa (upper) and 4 MPa (lower). Linear fit was provided for each group of scattered points.

Source Location and Failure Observation

The AE source location was obtained and plotted versus the test time in Figure 10. The
vertical axes are scaled to the actual core lengths. For low strength material, the majority
of AE events were located in the upper half of the core under both confining pressures.
This coincides with the observed results. For medium strength material, there was limited
number of AE sources due to lack of good sensor contact. For high strength material, AE
sources were distributed more uniformly along the length of the core.
The above AE distribution characteristics or cracks were observed from the post failure
demonstration in Figure 11. The cores failed due to shear cracking and macroscopic
cracks propagate along all the length of the cores. Cracks mainly distributed along one
portion of core length for low strength material, which was possibly due to unevenly
distributed axial stress.

SCA2015-032

10/12

Figure 10. One dimensional acoustic source distribution versus test time for low, medium and high
strength cores. Vertical axes were scaled to actual core heights.

CONCLUSIONS
AE provides one feasible technique of characterizing rock deformation and failure in the
laboratory. AE event rates correlate with the rock failure.
AE event DF tends to decrease with increased deformation. Also, event energy tends to
decrease with increased dominant frequency. There is no evident relationship between
DF and CCS.
AE source location was plotted versus time along the length of the cores. For low
strength cores, AE sources were mainly distributed on one end which was observed in the
failed specimens. For medium strength and high strength cores, AE sources were more
uniformly distributed along the core length.

SCA2015-031

11/12

Figure 11. Post failure illustration for low, medium and high strength cores under confining pressure of 2
MPa.

ACKNOWLEDGEMENTS
The project is funded by Atlantic Canada Opportunity Agency, Husky Energy, Suncor
Energy and Research and Development Corporation (RDC) of Newfoundland and
Labrador under AIF contract number: 781-2636-1920044.

REFERENCES
1. Hardy, H.R. Jr., "Applications of Acoustic Emission Techniques to Rock and
Rock Structures: A State-of-the-Art Review", Acoustic Emissions in Geotechnical
Engineering Practice, American Society for Testing and Materials, (1981), 4-92.
2. Lockner, D., The role of acoustic emission in the study of rock fracture, Int. J.
Rock Mech. Min. Sci. & Geomech. Abstr., (1993) 30, 7, 883-899.
3. Butt, S.D. and P.N. Calder, Experimental Procedure to Measure Volumetric
Changes and Microseismic Activity During Triaxial Compression Tests, Int. J.
Rock Mech. Min. Sci., (1998) 35, 2, 249-254.
4. Keshavarz, M., F.L. Pellet and K.A. Hosseini, Comparing the effectiveness of
energy and hit rate parameters of acoustic emission for prediction of rock failure,

SCA2015-032

12/12

in ISRM International Symposium on Rock Mechanics, 2009, The University of

Hong Kong, China.
5. Damani, A., A. Sharma, C.H. Sondergeld and C.S. Rai, Mapping of Hydraulic
Fractures under Triaxial Stress Conditions in Laboratory Experiments using
Acoustic Emissions, In SPE Annual Technical Conference and Exhibition, 2012,
San Antonio, Texas.
6. Chitrala, Y., C.H. Sondergeld and C.S. Rai, Microseismic Studies of Hydraulic
Fracture Evolution at Different Pumping Rates, in SPE Americas
Unconventional Resources Conference, 2012, Pittsburgh, Pennsylvania.
7. Fortin, J., S. Stanchits, G. Dresen and Y. Gueguen. Acoustic Emissions
Monitoring during Inelastic Deformation of Porous Sandstone: Comparison of
Three Modes of Deformation, Pure and Applied Geophysics, (2009).
8. Cai, M., P.K. Kaiser, Y. Tasaka, H. Kurose, M. Minami and T. Maejima,
Numerical Simulation of Acoustic Emission in Large-scale Underground
Excavations, in The 42nd US Rock Mechanics Symposium, 2008, San Francisco,
CA.
9. Tang, C., P.K. Kaiser and G. Yang. Numerical Simulation of Seismicity In Rock
Failure, 2nd North American Rock Mechanics Symposium, 1996, 19-21 June,
Montreal, Quebec, Canada.
10. Zang, A., C. Wagner, F., S. Stanchits, G. Dresen, R. Andresen, and M.A.
Haidekker. Source analysis of acoustic emissions in Aue granite cores under
symmetric and asymmetric compressive loads, Geophysical Journal, (1998) 135,
3, 1113-1130.
11. Aggelis, D.G. Classification of cracking mode in concrete by acoustic emission
parameters, Mechanics Research Communications, (2011) 38, 3, 153157.
12. Ohno, K. and M. Ohtsu. Crack classification in concrete based on acoustic
emission, Construction and Building Materials, (2010) 24, 12, 23392346.
13. Chow, T.M. Concurrent ultrasonic tomography and acoustic emission in solid
materials, Doctoral dissertation, Queens University at Kingston, 1992.
14. ASTM D4543-08, Standard Practices for Preparing Rock Core as Cylindrical
Test Specimens and Verifying Conformance to Dimensional and Shape
Tolerances, ASTM International, West Conshohocken, PA, 2008, www.astm.org.
15. ASTM D7012-14, Standard Test Methods for Compressive Strength and Elastic
Moduli of Intact Rock Core Specimens under Varying States of Stress and
Temperatures, ASTM International, West Conshohocken, PA, 2014,
www.astm.org.

SCA2015-032

1/11

SIMULATING IN-SITU CONDITIONS

FOR DIGITAL CORE ANALYSIS
S. Linden1,2, T. Cvjetkovic2, E. Glatt2, J.-O. Schwarz2 and A. Wiegmann2
1
Fraunhofer ITWM, Fraunhofer-Platz 1, 67665 Kaiserslautern, Germany
2
Math2Market GmbH, Stiftsplatz 5, 67655 Kaiserslautern, Germany
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Digital rock physics (DRP) for routine and special core analysis is usually based on 3D
digital images acquired under ambient conditions. Hence it does not account for the
pressure release the rocks suffer during the uplift from reservoir to laboratory conditions.
The resulting change in the pore space affects all physical properties significantly. In this
paper we present an efficient workflow to perform the full digital core analysis under insitu conditions. The in-situ conditions are modeled by numerically compressing a scan
taken under ambient conditions rather than performing the scan under in-situ conditions.
The workflow is illustrated using a segmented 3D image of a Berea sandstone derived
from a CT-scan under ambient conditions. Our results confirm that high geostatic
pressures lead to significant changes of the physical rock properties.

INTRODUCTION
In the oil and gas industry, digital rock physics (DRP) on 3D images is a fast growing
research topic. It is a standard tool to supplement and replace routine core analysis (RCA)
and special core analysis (SCAL). However, most simulations are performed on 3D
images taken under ambient conditions by CT or FIB-SEM. The uplift of rocks from
high geostatic pressures to ambient conditions enlarges the overall pore space. This
change in pore space affects all physical properties significantly and has to be taken into
account. Another challenge for DRP is the need to work on representative elementary
volumes (REV) and hence very large images due to the inhomogeneity of the samples. In
this paper we study the influence of increasing the geostatic pressure on
1.
2.
3.
4.

Porosity & pore size distributions,

Absolute & relative permeability,
Electrical conductivity, resistivity index & formation factor and
Capillary pressure.

There are two approaches to simulate in-situ conditions: (i) performing a scan under insitu conditions or (ii) numerically compressing a scan taken under ambient conditions.
We present an efficient workflow that allows computing these properties under in-situ
conditions by numerically compressing a scan. The simulations are performed using the
digital material laboratory GeoDict. Its built-in highly optimized numerical solvers are

SCA2015-032

2/11

capable of performing simulations on very large REVs (e.g. 1000 voxels). The workflow
is illustrated using a segmented 3D image of Berea sandstone [1, 2] derived from a CTscan under ambient conditions.

METHODS
The physical properties mentioned above are computed on 3D digital images that may
stem from CT-scans, FIB-SEM, or virtual rock models. The outline of the proposed
workflow is
1.
2.
3.
4.

Acquisition of CT-scans or FIB-SEM images under ambient conditions

Segmentation of the image into solid phases and pore space
Numerical compression of the 3D image (NEW STEP)
Computation of single-phase properties
a. Porosity & pore size distributions
b. Absolute permeability
c. Electrical conductivity and formation factor
5. Computation of the distribution of the two liquid phases in the pore space
6. Computation of two-phase properties on varying saturation states
a. Capillary pressure
b. Resistivity index
c. Relative permeability
The calculations are performed by highly optimized numerical solvers that work directly
on 3D digital images, without additional meshing. It is also not required to assemble
matrices that describe discretized physical equations. That approach reduces the memory
consumption dramatically. All but one solver use Fourier transforms to reduce the
number of iterations, and thus achieve very low runtimes.
Numerical compression
The prediction of mechanical properties and the geostatic correction, i.e. uniaxial
compression of the sample, is computed with the FeelMath [3,4] solver. The input
parameter for the compression step is the compression rate in percent or a given geostatic
pressure. In this section we summarize the basic ideas of the implemented algorithms.
The equations of linear elasticity for a constant macroscopic strain are defined by
=0
= :
2 = 2 + + ( )

Equilibrium
Hookes law

(1)

for the stress-field , strain-field , and the displacement-field . By introducing a

reference material of homogeneous stiffness 0 these equations can be reformulated into
the Lippmann Schwinger equation

SCA2015-032

3/11

+ 0 (( 0 ): ) =

(2)

where the convolution of the Green operator 0 can be solved by using Fast Fourier
transforms. The solver is able to handle linear and non-linear (i.e. replacing Hookes law
with a non-linear formulation) material laws assigned to mineral phases. The number of
iterations required for computation depends on the largest phase contrast in the structure.
Then the displacement field is used to predict a compressed structure [5]. The voxels
of the original image are moved along the displacement field and cut with a reduced
boxel image. The size of the boxel image can be derived from the input parameters. The
result of that procedure is a gray-value image where a global threshold is used to perform
a segmentation of the different phases. The threshold is chosen such that either mass or
volume is preserved. Then the boxel image is resampled to a voxel image again.
For the mechanical properties it is important to capture the complete mineralogy in the
segmentation process to predict representative results. Another important aspect to
improve the validity is the correct modelling of contacts between grains. The stiffness at
grain-to-grain interfaces is much lower compared to the stiffness within the grains. We
approximate that aspect with an increased uniaxial macroscopic stress. The compressed
3D image is then used in further computations of the workflow.
Distribution of the Two Liquid Phases
The pore morphology method [6], also known as method of maximal inscribed spheres
[7] predicts the distribution of a wetting phase (WP) and a non-wetting phase (NWP)
inside a porous medium. The method distributes the phases by using morphological
operations rather than solving partial differential equations. For drainage, spheres that
represent the NWP are placed inside the pore space where the pore size is greater than a
certain radius. The radius is decreased in an iterative process, i.e. the capillary pressure is
increased. Additional connectivity checks with respect to NPW and WP reservoirs can be
used to increase the validity of the distributions. These connectivity checks allow the
algorithm to introduce residual phases where parts of the NWP are trapped and cannot
leave the simulated domain.
The output of the algorithm is a finite sequence of quasi-stationary states. Each state is a
3D image again that encodes the solid phase, WP and NWP. In a post-processing step the
Young-Laplace equation based on the radii of the inscribed spheres and the interfacial
tension is used to predict the capillary pressure curve. The sequence of 3D images is then
the input data to predict relative properties.
Absolute and Relative Permeability
Absolute and relative permeability on the WP and NWP are computed with the SIMPLEFFT and LIR-Stokes [8] methods. The SIMPLE-FFT is an enhancement of the original
SIMPLE algorithm that solves the pressure correction within a single step using Fourier
transforms. The LIR-Stokes solver uses an adaptive tree structure and solves the

SCA2015-032

4/11

discretized Stokes equations at once for each cell. The methods are very fast and in
particular the latter has very low memory requirements.
Both methods assume no-slip boundary conditions at pore-solid interfaces for the
absolute permeability and also no-slip boundary conditions at WP-NWP interfaces for the
relative permeability. At the domain boundaries we assume periodic boundary condition
with inlet (e.g. 16 voxel) in flow direction. This flow boundary condition connects all
pores from inlet and outlet and predicts permeabilities between those of purely periodic
boundary conditions and symmetry boundary conditions.
Electrical Conductivity and Resistivity Index
The electrical conductivity, resistivity index and formation factor are computed with the
Explicit Jump Immersed Interface Method [9, 10]. Additional jump variables at material
interfaces are introduced to represent discontinuities inside the flux. The physical
problem can be reformulated from a three-dimensional to a two-dimensional problem
where the method solves for the jump variables. The number of iterations depends on the
number of interfaces. The chosen boundary conditions for the electrical properties are
similar to the flow properties. For the resistivity index we also assume that the NWP is
non-conductive.

RESULTS
In our experiments we use a binary segmented 3D image of a computed tomography scan
from a Berea sandstone under ambient conditions. The image consists of 720 x 720 x
1024 voxels, has a resolution of 0.74 m, and is shown in Figure 1 together with the
original gray value image. The origin and segmentation of the dataset is described in the
first part of the benchmark study [1]. The second part of the benchmark study [2]
includes the numerical computation of physical properties but considers permeability,
electric conductivity and elastic moduli only.
Table 1 shows runtimes and memory consumptions for single computations of the
numerical solvers. Depending on the available computational resources, all simulations
can be performed in a couple of days. Figure 5 shows the graphs of porosity, pore sizes,
formation resistivity factor, absolute permeability, capillary pressure, and resistivity
index with respect to the pressure states.
Mechanical Properties and Numerical Compression
First, we compute a sequence of images which have been corrected for different geostatic
pressures, i.e. numerical compression with the states: 0.0, 0.12, 0.24, 0.48, 0.71, 0.95, and
1.43 GPa. Pores are assumed to be filled with air ( = 0 GPa and = 0) and the solid
phase is assumed to be quartz ( = 94.5 GPa and = 0.074). Then the core analysis is
performed on each compression step. The pressures correspond to the compression ratios:
0.0, 0.25, 0.5, 1.0, 1.5, and 3.0%. The computed isotropic elastic modulus is E = 45.9
GPa and the Poisson ratio is = 0.108 for the uncompressed Berea sandstone. The visible

SCA2015-032

5/11

difference of the uncompressed and compressed image can be seen in Figure 2 and
indicates a significant change of effective physical properties.
Porosity and Pore Size Distribution
The porosity of the uncompressed sample is 18.4 % and decreases linearly to 15.7 %. The
(relative) pore size distribution (PSD) obtained by maximum inscribed spheres does not
change significantly under increasing geostatic pressure. In contrast to the Porosimetry
simulation where the most frequent pore diameter 8.8 m is decreased to 7.4 m. A
reduction between 2% and 3% is a reasonable value for sandstone and agrees with
measured values in literature [12].
Absolute Permeability
The absolute permeability is 108 mD under ambient conditions and decreases with a high
slope to 66 mD under in-situ conditions. This decline of permeability agrees with the
relation of velocity and pore diameter. Figure 3 shows the velocity field of a Stokes flow
under ambient conditions. The trend agrees to the measured stress dependence of
permeability reported in [12].
Absolute Electrical Conductivity and Formation Factor
The electrical conductivity of the fully Brine saturated Berea sandstone is simulated to
0.17 S/m, assuming that Brine has a conductivity of 5 S/m and Quartz is non-conductive.
Figure 3 shows the potential field for the fully saturated medium under ambient
conditions. The corresponding formation resistivity factor increases almost linear from 27
to 39 when air drains brine as illustrated in Figure 4. That behavior coincides with the
linear decrease of the pore diameters.
In contrast to the permeability, the number of iterations for computing electrical
properties is almost the same from ambient to in-situ conditions. This can be explained
by the fact that the number of iterations depends on the brine surface area which does not
change significantly during the numerical compression.
Two-Phase Distribution
The pore morphology method is used to perform a drainage simulation. Figure 4 shows
different stationary saturation states of the drainage process where air drains brine. In that
experiment we assumed a non-wetting phase (Air) reservoir at the top and a wetting
phase (Brine) reservoir at the bottom of the domain.
Since that method does not solve a partial differential equation the computation is much
faster compared to the computations of single-phase flow, mechanical, and electrical
properties. A subset of these quasi stationary states is used to predict relative properties.
Capillary Pressure
Similar to the pore size distribution, the capillary pressure curve computed from the twophase distribution does not change significantly within the investigated geostatic pressure

SCA2015-032

6/11

range. The irreducible wetting phase saturation is 17.3% but the displacement pressure
increases from 24 kPa to 29 kPa.
Relative Permeability
The relative permeability of the wetting phase is almost the same for ambient and in-situ
conditions other than the relative permeability of the non-wetting phase. Figure 6 shows
the relative permeability of both phases with respect to the normalized wetting phase. The
decline occurs more early and can be explained by delayed breakthrough of the nonwetting phase.
The relative permeability is the most expensive property to compute depending on the
number of saturation states. The number of iterations increases significantly in lower
saturation states even though the number of computational cells decreases. The SIMPLEFFT method cannot profit from the decreasing number of cells while the LIR method can
and hence the runtime per iteration decreases.
The number of iterations needed until the same stopping criterion is reached increases by
25% for computing the absolute permeability from ambient to in-situ conditions. This can
be explained by the increased complexity of the pore space. That behavior can be
observed for the SIMPLE-FFT, LIR, and also other flow solvers.
Resistivity Index
The electrical conduction simulation is performed for the wetting phase on 8 saturation
states from 0.3 to 1.0. Figure 5 shows the resistivity index for the Berea sandstone under
ambient and in-situ conditions together with an exponential fitting curve. The saturation
exponents of Archies law are 2.38 and 2.04, respectively. Similar to the relative
permeability, the absolute resistivity of the wetting phase under ambient condition is
greater compared to the in-situ conditions.

CONCLUSION
The integrated approach of highly optimized solver technologies allows performing the
full digital core analysis under in-situ conditions in a short amount of time and memory.
Our results confirm that high geostatic pressure leads to different physical rock
properties. The impact is significant on the single-phase properties: porosity, the most
frequent pore diameter, permeability, and formation factor. The capillary pressure,
resistivity index, and relative permeability are also influenced by increasing pressure but
to a lesser extent. The next step is a comparative study of digital and laboratory in-situ
conditions, also for other types of rocks. In addition, numerical simulations and
modelling will be improved in at least three ways: 1. segmentation of all mineral phases,
2. modelling and consideration of grain contacts, and 3. improving the computational
runtimes for relative properties.

SCA2015-032

7/11

REFERENCES
1. H. Andr, N. Combaret, J. Dvorkin, E. Glatt, H. Junehee, M. Kabel, Y. Keehm, F.
Krzikalla, M. Lee, C. Madonna, M. Marsh, T. Mukerji, E. Saenger, R. Sain, N.
Saxena, S. Ricker, A. Wiegmann and X. Zhan, Digital rock physics benchmarks Part
I: Imaging and segmentation, Computers & Geosciences, 2013 (43), pp. 25-32.
2. H. Andr, N. Combaret, J. Dvorkin, E. Glatt, H. Junehee, M. Kabel, Y. Keehm, F.
Krzikalla, M. Lee, C. Madonna, M. Marsh, T. Mukerji, E. Saenger, R. Sain, N.
Saxena, S. Ricker, A. Wiegmann and X. Zhan, Digital rock physics benchmarks Part
II: Computing effective properties, Computers & Geosciences, 2013 (43), pp. 33-43.
3. M. Kabel and H. Andr, Fast numerical computation of precise bounds of effective,
Report of Fraunhofer ITWM, 2013 (224).
4. M. Kabel, T. Bhlke and M. Schneider, Efficient fixed point and Newton-Krylov
solvers for FFT-based homogenization of elasticity at large deformations,
Computational Mechanics, 2014 (54), pp. 1497-1514.
5. M. Kabel, H. Andr, F. Hahn and M. Lehmann, Simulating the Compression of
Filter Materials, Proceedings of FILTECH, 2013
6. M. Hilpert and C. T. Miller, Pore-morphology-based simulation of drainage in
totally wetting porous media, Advances in Water Resources, 2001 (24), pp. 243-255.
7. D. Silin, L. Tomutsa, S.M. Benson and T.W. Patzek, Microtomography and PoreScale Modeling of Two-Phase Distribution, Transport in Porous Media, 2011 (86),
pp. 495-515.
8. S. Linden, A. Wiegmann and H. Hagen, The LIR space partitioning system applied
to the Stokes equations, Dagstuhl Geometric Modelling 2014 Special Issure of
Graphical Models, to be published.
9. A. Wiegmann and K. P. Bube, The Explicit-Jump immersed interface method: Finite
difference methods for PDE with piecewise smooth solutions, SIAM Journal on
Numerical Analysis, 2000 (37 No 3), pp. 827-862.
10. A. Wiegmann and A. Zemitis, EJ-Heat: A fast Explicit-Jump harmonic averaging
solver for the effective heat conductivity of composite materials, Fraunhofer ITWM
Technical Report No 94, 2006.
11. J.J. Dong, J.Y. Hsu, W.J. Wu, T. Shimamoto, J.H. Hung, E.C. Ye, Y.H. Wu and H.
Sone, Stress-dependence of the Permeability and Porosity of Sandstone and Shale
from TCDP Hole-A, International Journal of Rock Mechanics and Mining Sciences,
2010 (47), pp. 1141-1157
12. S. Linden, T. Cvjetkovic, E. Glatt and A. Wiegmann, An integrated approach to
compute physical properties of core samples, Proceedings of SCA Avignon, 2014,
Paper 057.

SCA2015-032

8/11

FIGURES AND TABLES

Table 1 Runtime and Memory requirements of the solvers on a computer with 16 cores and 128 GB
RAM. The runtimes account for a single simulation on the uncompressed Berea sandstone.
Electrical
Two-Phase
Flow
Flow
Mechanics
Method
Conduction
Distribution
(SIMPLE-FFT) (LIR)
(FeelMath)
(Explicit Jump) (Pore Morphology)
Runtime [h]
3.6
3.1
0.6
0.8
8.3
Memory [GB]
42.3
5.4
9.4
5.0
97.1

Figure 1 Gray value image from the Berea sandstone (left and center) with a binary segmentation (right).
The gray value image consists of 1024x1024x1024 voxel and the cropped segmentation consists of
720x720x1024 voxel.

Figure 2 Cut-out from the segmented image of Berea sandstone before (left) and after (right) numerical
compression (1.43 GPa). The pore diameter in the blue region is visibly decreased while even the
connectivity of the pores is reduced in the two red regions.

SCA2015-032

9/11

Figure 3 Visualization of the von-Mises stress (left), the velocity field of a Stokes flow (center) and
electrical potential (right) on a smaller cut-out. A compression, drop in pressure, and potential difference
are applied in y-direction from top to bottom.

Figure 4 Air drains brine and we visualize air saturations of 25%, 50% and 75% respectively.

SCA2015-032

10/11

60

Volume Fraction [%]

19

Porosity [%]

18
17
16
15

50
PSD 0.0 GPa
PSD 1.5 GPa
Poro. 0.0 GPa
Poro. 1.5 GPa

40
30
20
10
0

0.5

1.5

10

46

120

41

100

36
31
26
21

40

50

80
60
40
20

16

0
0

0.5

1.5

Geostatic Pressure [Gpa]

0.5

1.5

Geostatic Pressure [Gpa]

160

100

140

0.0 GPa

120
100

0.0 GPa

80

1.5 GPa

60
40

Resistivity Index

Capillary Pressure [kPa]

30

Pore Size [m]

Permeability [mD]

Formation Resistivity Factor

Geostatic Pressure [Gpa]

20

1.5 GPa
10

20
0

1
0.00

50.00
Saturation WP [%]

100.00

0.10

1.00
Saturation WP [%]

Figure 5 Porosity, pore size distribution, porosimetry, formation resistivity factor, absolute permeability,
capillary pressure, and resistivity index for different applied geostatic pressures.

SCA2015-032

11/11

Relative Permeability

1.0

WP 0.0GPa
NWP 0.0GPa
WP 1.5GPa
NWP 1.5 GPa

0.8
0.6
0.4
0.2
0.0
0.0

0.2

0.4

0.6

0.8

1.0

Saturation WP
Figure 6 Relative permeability of the WP and NPW for ambient and in-situ conditions.

SCA2015-033

1/7

CARBONATE NMR MEASUREMENTS IN A COMBINED

AMOTT-USBM WETTABILITY METHOD
Edmilson Helton Rios, National Observatory; Vinicius de Frana Machado, Bernardo
Camilo Coutinho dos Santos, Willian Andriguetto Trevizan, Andr Luiz Martins
Compan, Dario Abilio Cruz, Rodrigo Skinner, Petrobras; Felipe Moreira Eler, Gorceix.
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
NMR measurements were performed along with the stages of a combined Amott-USBM
method for some Brazilian deep-water carbonates. Guided by the T2 signature of nonconfined brine and dead crude oil, monophasic and biphasic T2 distributions were
interpreted revealing pore occupancy and wettability of micro and meso-to-macro pores.
Consistent T2 shifts were observed after core aging, imbibition and second drainage. The
results indicate that simple 1D-NMR shall support standard methods that cannot resolve
saturation and wettability per porosity type.

INTRODUCTION
Shortly after NMR phenomenon discovery in the late forties, the enhancement of
relaxation time rates for the wetting phase was noticed [1]. Since then, many studies have
been performed on NMR wettability based either on modeling or experimental
approaches [2-4]. Although most of these works introduce NMR indices and compare
them to the industry standard indices, they do not necessarily must have the same
response because these techniques rely on different physical grounds. Instead, for
improving conventional methods response, this work evaluates transverse relaxation
times (T2) in the different saturation and wettability conditions availed from the stages of
a combined Amott-USBM method [5,6]. As indicated in Figure 1a, T2 measurements
were acquired after 1) total water saturation (Sw=100 %), 2) first drainage with dead
crude oil up to irreducible water saturation (Swi), 3) three-months aging in the crude oil,
4) spontaneous plus forced water imbibition up to residual oil saturation (Sor) and, 5)
spontaneous plus forced drainage with crude oil up to final water saturation (Swf).
As indicated in Figure 1b, the water Amott index (Iw) measures the displacement-bywater ratio, which is the oil volume displaced by spontaneous water imbibition (Sws),
relative to the total oil volume displaced by spontaneous and forced water imbibition (1Swi-Sor). The oil index (Io) is analogously defined and the difference between water an oil
indices is known as the Amott-Harvey index (IAH) which is bound by -1< IAH<+1, Figure
1b. In contrast, the US Bureau of Mines (USBM) index is calculated with the areas under
the capillary pressure curves that describes displacement of oil by water Ai (forced
imbibition) and the displacement of water by oil Ad (second forced drainage), Figure 1c.

SCA2015-033

2/7

For example, a carbonate with index -1 would require 10 times more energy for
imbibition than for drainage (and vice versa for index +1). Although the USBM index is
more sensible to neutral wettability, it cannot differ between neutral (Iw and Io are equal
and low) and mix (Iw and Io are equal and high) wettability.

Figure 1: a) Complete cycle of the Amott-USBM method showing the five stages where T2 measurements
were aquired. The classical wettability indeces are also defined: a) Amott-Harvey and c) USBM.

EXPERIMENTAL
X-ray microtomography (CT) and routine core analyses (RCAL) were used for selection
of seven representative core plugs from a Brazilian Pre-Salt carbonate reservoir. Special
core analyses (SCAL) were performed at the reservoir temperature of 60 C. Brine
salinity in NaCl equivalent was 182 g/l; brine density and viscosity was 1.09 g/cm3 and
0.65 cP, respectively, and dead crude oil density and viscosity was 0.86 g/cm3 and 11.23
cP, respectively. Spontaneous imbibition and drainage volumes were acquired until
production stabilization, between 25 and 75 days. Without confinement, forced
imbibition and drainage were performed with a centrifuge in 15 increasing capillary
pressure steps up to 100 psi (at the face of the core plug). The Carr-Purcell-Meiboom-Gill
(CPMG) pulse sequence were used for measuring T2 in a 2.2 Mhz (for 1H) benchtop
NMR analyzer (Maran DRX-HF, Oxford instruments, UK). A full polarization time and a
minimum signal to noise ratio of 100:1 were used. T2 distributions were inverted in the
WinDXP program (the Maran accompanying software) using 256 bins which were then
normalized to a unitary sum of amplitudes. Further details on NMR phenomenon, data
acquisition and inversion can be found in the reference [7].

SCA2015-033

3/7

RESULTS AND DISCUSSION

Bulk transverse relaxation time T2b of the brine and the crude oil are shown in Figure 2a
(non-confined in the rock). In brine, 1H is present only in the molecules of water and thus
T2bw distribution is quite narrow. In contrast, dead crude oil presents a very broad T2bo
distribution due to the diversity of hydrocarbon molecules (between tens and hundreds of
thousands) that contains 1H. Long T2bo is related to light components (e.g. paraffins and
aromatics) holding small molecules with high mobility. Short T2bo is associated to heavy
components (e.g. resins and asphaltenes) holding big molecules with low mobility.
However, if water and oil are confined in a porous media, relaxation time rates are
enhanced due to the interaction with pore surface. Considering a fast diffusion regime
and assuming no diffusional relaxation effects due to magnetic field gradients nor
diffusion pore coupling [7], T2 of water and oil confined in a pore can be described as
follows [3]:
1
T2

=T

2bw

+ w S

Aw
w Vp

+T

2bo

+ o S

Ao
o Vp

(1)

Where w and o are the T2 surface relaxivity for water and oil, respectively; Vp is the
pore volume; Sw and So are the fractional water and oil pore saturation (Sw+So=1),
respectively and; Aw and Ao are the fractional water and oil pore surface area
(Aw+Ao=A), respectively. From equation 1, the characteristic length (R) of a spherical
pore site (/3 = /) occupied by a fluid which has a single T2b is R(T2 ) =
3(T2 T2b T2b T2 ). Figure 2b plots R(T2 ) for brine and each of the three crude oil
components with w = 0.006 m/ms (average values based on mercury-injection data
comparison) and assuming o = w /3 as in reference [8]. Characteristic lengths of 1 and
10 m were considered as thresholds among micro, meso and macro pore types.

Figure 2: a) Bulk T2 for brine (T2bw) and crude oil (T2bo). b) Characteristic length (R) of spherical pore sites
as a function of monophasic T2 for brine and crude (although the oil components cannot be separated, their
relaxation behavier is illustrated separately, as if they were three different oils).

The studied carbonates are high quality reservoir facies with good porosities (15-20 %)
and permeabilities (8-1552 mD) responding with long monophasic (brine) T2
distributions, Figure 3a. Higher T2 are associated with higher permeabilities which is
explained by the bigger characteristic lengths (brine curve of Figure 2b). When these

SCA2015-033

4/7

carbonates were desaturated by crude oil (Swi ranging from 13 to 25 %), previous T2based differences in pore structure became less evident and all biphasic T2 distribution
reduce to a similar shape of the crude oil bulk response, black dotted curve in Figure 3b.
This indicates that crude is in meso-to-macro pores relaxing almost like T2bo distribution
(Figure 2b shows that surface relaxation is low in such pore types even for light oil
components). After brine imbibition (Sor ranging from 3 to 29 %), some previous features
of macropores (300 to 3000 ms in Figure 3a) are clearly recovered in Figure 3c between
600 and 4000 ms. The observed shift towards T2bw indicates a decrease of the brine
surface relaxation component (less contact with the pore walls) compared to the bulk
relaxation component (Equation 1). Although residual oil ocuppies smaller sites after
brine imbibition, characterist lengths of theses sites would be in the same order of
magnetude (lets say, for instance, a quarter of the meso-to-macro pore sizes). Imbibed
brine may also reduce contact with pore walls further explaining why it still relaxes as
T2bo distribution (which is sumed with the irreducible brine signal from micro pores,
T2<500 ms). Imbibed brine is not totaly drived out after second drainage (Swf-Swi>0) and
its portion more isolated from the pore walls, responding in the T2 range of 600 to 2000
ms (Figure 3d), prevents hysteresis back to the distribution shapes of Figure 3b.

Figure 3: T2 distributions for the carbonate plugs (continuous curves) after: a) fully water saturation, b)
drainage up to Swi and aged, c) imbibition up to Sor and d) drainage up to Swf. T2b distribution of nonconfined brine and crude oil are also shown as reference (black and blue dotted curves, respectively).

The curves for the three first stages are plotted together for samples C4 and C5 in Figure
4a and 4b, respectively. Strictly, the standard wettability methods started only after native
wettability restoration, which was performed by aging the samples in crude oil for three
months. The bigger shift to short T2 agrees with the more oil-wet conditions indicated by

SCA2015-033

5/7

the Amott and USBM methods of sample C4. After aging, the surface wetted by crude oil
in meso-to-macro pores (Ao ) increases and so does T2 rates (Equation 1). The four stages
presented in Figure 3 are plotted all together in Figure 5. Partial exchange between oil
and water is highlighted in black shading for sample C4 after imbibition (Figure 5a)
whereas the opposite exchange is highlighted in blue shading for sample C5 after second
drainage (Figure 5b). The biphasic distributions of Figure 5 are shown in Figure 6
subtracted by the oil signal which was considered to have the same shape of T2bo
distribution (sum of amplitudes normalized to 1-Swi, Sor and 1-Swf for stages 3, 4 and 5,
respectively). According to the classical cutoff interpretation, calculated with the Swi
value on the monophasic T2 distribution, irreducible brine is in micro pores and free brine
is in meso-to-macro pores. While subtracted irreducible brine in stage 3 is within the
monophasic T2 distribution agreeing especially for shorter T2, its apparent increase in the
stage 4 and 5 is probably brine coating the walls of meso-to-macro pores or occupying
reduced sites in such pore types. In contrast, free brine in pore sites with less contact with
the pore walls can have longer T2 than the monophasic response which is due to a
decrease of surface relaxation and consequent bigger influence of bulk brine components.

Figure 4: Monophasic (S1) and biphasic (S2 and 3) T 2 distribution for carbonates plugs a) C4 and b) C5.

Figure 5: Monophasic (S1) and biphasic (S3,4 and 5) T2 distribution for a) C4 and b) C5.

SCA2015-033

6/7

Figure 6: Monophasic (S1) and brine signal of biphasic (S3,4 and 5) T2 distribution for a) C4 and b) C5.

CONCLUSION AND PERSPECTIVES

NMR transverse relaxation times (T2) were interpreted for some Brazilian deep-water
carbonates under the different saturation and wettability conditions of a combined AmottUSBM method. While monophasic T2 distribution was more sensible to the samples
variation in characteristic pore lengths, the biphasic T2 distributions had signals
overlapped from the fully-brine-saturated micro pores (thus water-wet) and the brine and
crude oil presence in meso-to-macro pores. Several arguments were presented showing
that the crude oil relax quite close to its bulk T2 signature which allowed a simple signal
separation in T2-domain. Samples with the Amott-USBM indices indicating a preferably
oil-wet condition presented T2 distributions shifted to shorter values after restoring native
wettability. With the aging, crude oil increased its contact with the surface of the mesoto-macro pores. After imbibition and second drainage, the presence of residual crude oil
makes brine relax in different sites with more or less contact with the pore walls (i.e.
shorter or longer T2, respectively). This work is been extended for different carbonate
reservoirs with a bigger number of samples. The evaluation shall be improved with others
NMR techniques such as spatial T2 and difusion-T2 measurements.

ACKNOWLEDGEMENTS
The authors thank Petrleo Brasileiro SA (PETROBRAS) for releasing publication.
Edmilson Helton Rios is grateful to CNPq and Petrobras for his research fellowship.

REFERENCES
1. Brown, R.J.S and Fatt, I.: Measurement of Fractional Wettability of Oilfield Rocks
by the Nuclear Magnetic Relaxation Method, Trans. AIME, (1956) Vol. 207, p.262
264.
2. Al-Mahrooqi, S.H., Grattoni, C.A., Muggeridge, A.H. and Jing, X.D.: Wettability
Alteration During Aging: The Application of NMR to Monitor Fluid Redistribution,
Society of Core Analyst, (2005) Toronto, August 21-25.
3. Looyestijn, W. and J. Hofman, Wettability Index Determination by Nuclear
Magnetic Resonance, Proceedings of SPE Middle East Oil and Gas Show and
Conference, (2005).
4. Al-Muthana, A.S., Hursan, G.G., Ma, S.M., Valori, A., Nicot, B. and Singer, P.M.,
Wettability as a function of pore size by NMR, (2012) SCA2012-31.

SCA2015-033

7/7

5. Donaldson, E., R. D. Thomas, and P. B. Lorenz, Wettability Determination and Its

Effect on Recovery Efficiency, SPE Journal, (1969), 9, pp. 13-20.
6. Anderson, W.G. Wettability Literature Survey - Part 1 to 6, Journal of Petroleum
Technology, (1986-1987), vol. 38-39.
7. Dunn, K.J., Bergman, D.J. & Latorraca, G.A., 2002. Nuclear Magnetic ResonancePetrophysical and Logging Applications, Pergamon Press, Oxford.
8. Machado, V.F., Ramos, P.F.O., Netto, P., Azeredo, R.B.V., Boyd, A., Souza, A.A.,
Zielinski, L. & Junk, E., 2012. Carbonate Petrophysics in Wells Drilled with OilBased Mud, Petrophysics, 53(4), 285-292.

SCA2015-034

1/6

MULTI-SCALE ROCK ANALYSIS FOR IMPROVED

ROCK TYPING IN COMPLEX CARBONATES
Moustafa Dernaika, Yasir Naseer Uddin, Safouh Koronfol, Osama Al Jallad and
Gary Sinclair, Ingrain Inc-Abu Dhabi
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
The evaluation of carbonate reservoirs is a complex task because of the inherent
heterogeneities that occur at all length scales. Rock types may be defined differently at
different scales and this introduces a challenge in capturing heterogeneity in a single rock
volume. Heterogeneities at smaller length scales must be upscaled into larger scale
volumes to better predict reservoir performance. The objective in this study is to define
carbonate rock types at multiple scales and then upscale those rock types to the whole
core level.
Representative core plugs were selected in a heterogeneous reservoir interval based on
statistical distribution of litho-types in the core. The litho-types were described by
porosity and mineralogy variations along the core length using advanced dual energy
XCT imaging. Plug-scale rock types were defined on the basis of petrophysical data and
geological facies. High-resolution micro to nano XCT images were integrated in the rock
typing scheme. Those rock types were upscaled to the whole core level by linking the
core litho-types to the plug data.
The core litho-types (porosity and mineralogy) gave good representation of the whole
core heterogeneity and were reliable for selecting representative samples. This was
qualified by X-ray diffraction mineralogy and plug porosity measurements. This allowed
establishing the link between plugs and whole cores and hence upscaling rock type
information to the whole core scale. The high-resolution digital images emphasized the
different pore geometries in the samples and improved the definition of the rock types.
Accurate porosity and permeability logs were derived along the core length and gave
very good match with the plug data.
The paper presents an advanced and quick tool for representative sample selection and
statistical core characterization in heterogeneous reservoirs. The identified rock types at
multiple scales provided new insights into carbonate heterogeneity and gave upscaling
options for rock types and petrophysical data. The upscaled rock types at the whole core
level enhance the prediction of dynamic imbibition data along the reservoir column for
improved reservoir performance.

SCA2015-034

2/6

INTRODUCTION
Proper core characterization is often overlooked in rock typing and sample selection
studies. Random core sampling is usually performed and the selected plugs are not
associated to the heterogeneity at the whole core level. This leads to unrepresentative
selection of the core samples and their rock types, and hence raises questions about the
effectiveness of the core data in reservoir models and their calibrations.
In this study, rock typing was first established at the whole core level by advanced dual
energy CT scanning. Three main lithologies were identified that appeared to be the main
control of the petrophysical variations along the core length. Representative plugs were
selected from each lithology and analysed to confirm the rock types at the micro level.
This was performed by Poro-Perm measurements, thin-section photomicrograph
description, mercury injection analysis and high resolution XCT imaging.

DETAILED WHOLE CORE CHARACTERIZATION

X-ray CT imaging is a powerful non-destructive method used in the oil industry to
evaluate the internal structures of reservoir cores. When the core is imaged by the Dual
Energy technique it provides two distinct 3D images by using a high- and a low- energy
setting. The high-energy images are more sensitive to bulk density (Compton Scattering
effect) and the low-energy images are more sensitive to mineralogy (Photoelectric
Absorption effect). The bulk density (RHOB) and effective atomic number (Zeff) values
are computed independently for each CT slice along the core length [1].
Lithology Log and Statistical Analysis
In this study, 300 feet of reservoir core was imaged by the dual energy technique, and the
bulk density (BD) and Zeff were calculated at 0.5 mm spacing. Figure 1(a) shows the
bulk density and Zeff variations with three-color mineralogy log. The detected minerals
were calcite (for Zeff larger than 14), dolomite (at Zeff around 13) and partially
dolomitized core (at Zeff around 14) [2]. This mineralogical variation was also confirmed
by XRD analysis performed on rock samples taken from each Zeff response along the
core length. The identification of these minerals was obtained at very early stage of the
study while the core was still in the barrel, hence providing a quick tool for detailed core
characterization and representative sample selection. The volume percent of each
mineralogy with respect to the total core volume was quantified from the whole core CT
data and is presented in the pie chart in figure 2(a). The core is mainly composed of
calcite with 66 vol%, 15 vol% dolomite and 19 vol% partially dolomitized core.
Core-Scale Porosity and Permeability
Representative samples were taken from each mineralogy along the core and were
measured for helium porosity and gas permeability. The poro-perm data for the acquired
plugs are plotted in figure 2(b), and shows two distinct correlations that are controlled by
the different characteristics of the mineralogy in the reservoir. The dolomite and partially
dolomitised plugs gave single trend with the dolomite samples at the higher poroperm
range. Porosity ranges from 12 to 30 porosity unit and the permeability varies from 5 mD

SCA2015-034

3/6

to 1000 mD. On the other hand, the calcite samples gave a lower permeability trend
varying from 0.02 mD to 4 mD, while the porosity ranging from 8 to 25 porosity unit.
Figure 1(b) compares the plug porosity data with the core porosity log derived from the
XCT bulk density data along the core length. The core porosity was calculated from the
bulk density log and by using different grain density (GD) values for each mineralogy
(GD: 2.71 g/cc for calcite, 2.85 g/cc for dolomite and 2.79 g/cc for partially dolomitized
(dol/cal) intervals). There is an overall reasonable match between the plug helium
porosity and the XCT-derived whole core porosity. Many plugs were taken in the core
and gave different porosity values at almost the same depth intervals because of the core
heterogeneity that cannot normally be characterized properly by plug-scale
measurements. The comparison shows better matches for lower porosity range (i.e.
<15%) at which the core shows less heterogeneity. In direct conventional comparison
studies between whole core and plug helium porosity measurements, the whole core
porosity data would normally be lower than the plug porosity [3,4], confirming our
findings in this study.
Figure 1(c) compares the plug gas permeability with a permeability log derived along the
core length from the porosity log and by using the plug poro-perm relations from figure
2(b). A very good comparison is observed between the core permeability log and the plug
data. Very heterogeneous plugs overestimate core permeability by short-circuiting flow
either because of fractures or vuggy porosity channels that do not represent the full
diameter core [3,4]. Around two to three order of magnitude increase in permeability is
seen in the dolomite intervals, which suggests large improvement in the rock transport
properties with dolomitization for this reservoir under study.

ROCK TYPING
In order to understand the structure and porosity type of the rock types, thin-section and
MICP wafers were cut from selected representative locations along the plug lengths.
Figure 3 represents typical rock types identified from the different mineralogical
variations at the whole core level. The figure presents (from left to right, in each RRT)
2D whole core XCT image, high-resolution micro CT image, mercury-derived porethroat size distribution (PTSD) from mercury injection experiments and thin-section
photomicrograph. The whole core image represents the location from which the plug was
taken. The high-resolution micro images for the three rock types are taken at different
resolutions to resolve the pore system. The dolomite rock type needed lower resolution at
4 micron/voxel where the corresponding mercury peak is at 10 micron pore throat radius
and thin-section photomicrograph shows a complete dolomitised grainy system with
inter-crystalline porosity. On the other hand, the pore system of the calcite rock type was
only resolved by nano imaging at a voxel resolution of 0.064 micron. This muddy calcitic
rock type is almost completely made of micrite with inter-particle porosity and a mercury
peak at 0.7 micron. The partially dolomitized rock type gives the mercury peak at 3
micron pore throat radius and was imaged at 2 micron/voxel. Both the thin-section
photomicrograph and the micro CT image show the dolomitized rock with precursor

SCA2015-034

4/6

limestone as micritic particles. This integration of data confirms the rock types and
enhances our understanding of these rock types at the micro level that can be upscaled to
the whole core by the link of mineralogy.
The rock typing scheme presented in figure 3 can be populated in the whole core by the
identified mineralogy as depicted in figure 1(a). In this perspective, the different minerals
in this core are the main control of the petrophysical and fluid flow in the reservoir. The
dolomitization that occurred has been a great improvement in the rock properties in the
core, which is mainly composed of low permeability calcite. The dolomitization and
partial dolomitization in the core have enhanced the permeability by 2 to 3 orders of
magnitude and this was crucial for reservoir modeling and field development plan.

SUMMARY AND CONCLUSION

Dual Energy CT scanning was used to identify three different lithological variations
along the core length by bulk density and effective atomic number. The core was mainly
composed of low permeability calcite and higher permeability dolomitized intervals,
which were crucial for enhanced petrophysical and fluid flow properties. Representative
plugs were cut from the different lithologies and were grouped in three different rock
types based on micro CT imaging, MICP analysis and thin-section description. The
following can be concluded from this study:
1. Dual Energy CT scanning is a quick and powerful tool to accurately identify porosity
and mineralogy variations along reservoir core lengths at early stages of the analysis.
2. The identified mineralogy in this core was found to be the main control of transport
properties and fluid flow.
3. The CT-derived porosity and permeability logs at the core level showed good match
with the plug data and gave averaged properties in more heterogeneous intervals.
4. Two distinct poro-perm correlations were established for the calcite and dolomitized
rock types.
5. Rock typing was attempted at the following different scales, and was shown to give an
excellent tool for improved understanding and upscaling:
a. Whole core RRT by porosity and mineralogy logs
b. Plug RRT by poro-perm
c. Micro-level RRT by high-resolution CT, MICP and thin-section

ACKNOWLEDGEMENTS
The authors wish to acknowledge Ingrain Inc for the permission to publish the results of
this study.

REFERENCES
1. Wellington, S.L. and Vinegar, H.J.(1987): X-Ray Computerized Tomography, J. Pet
Tech 39 (8): 885-898. SPE-16983-PA.

SCA2015-034

5/6

2. Dernaika, M.R., Basioni, M., Dawoud, A., Kalam, M.Z., and Skjveland, S.M.
(2013):Variations in Bounding and Scanning Relative Permeability Curves with
Different Carbonate Rock Types J. Res. Eval. & Eng. (August 2013) paper SPE162265-PP presented at the Abu Dhabi International Petroleum Exhibition &
Conference held in Abu Dhabi, UAE, 1114 November (2012).
3. Dernaika, M., Serag, S. and Kalam, M.Z., (2011): The Impact of Heterogeneity and
Multi-Scale Measurements on Reservoir Characterisation and STOOIP Estimations
paper SCA2011-46 presented at the International Symposium of the Society of Core
Analysts held in Austin, Texas, USA 18-21 September.
4. Ehrenberg, S.N., 2007, Whole core versus plugs: Scale dependence of porosity and
permeability measurements in platform carbonates: AAPG Bulletin, V. 91, NO. 6
(June 2007), pp. 835 846.
(a) BD and Zeff logs

(b) Porosity log

(c) Permeability log

Calcite

Dol/cal
Dolomite

Figure 1 (a) XCT Dual Energy-derived bulk density and Zeff logs along the core length with colored
mineralogy log based on Zeff variation. (b) Comparison of DE-derived porosity with plug porosity. (c)
Comparison of DE-derived permeability with plug permeability.

SCA2015-034

(a)

6/6

(b)

Figure 2 (a) Dual Energy-derived statistical analysis of the litho-types in the core (data from figure 1a).
(b) Poro-perm relations for the calcite and the dolomitized plugs.

Figure 3 From left to right (in each RRT): 2D whole core XCT image, high resolution micro CT image,
mercury-derived pore-throat size distribution (MICP) and thin-section photomicrograph.

SCA2015-035

1/7

ASSESSMENT OF SIDEWALL CORES FOR ROUTINE

AND SPECIAL CORE ANALYSES
F. Pairoys1, A. Nadeev1, K. Kirkman1, G. Poole1, K. Bohn1, M. Alexander1, N. Radwan1,
W. Abdallah2, M. Akbar3
1

Schlumberger Reservoir Laboratories, Houston, Texas, 77041, United States

Schlumberger Dhahran Carbonate Research, Dhahran Techno Valley, Saudi Arabia
3
Ex Schlumberger - Independent O&G Consultant, Toronto, Canada

This paper was prepared for presentation at the International Symposium of the Society
of Core Analysts held in Newfoundland and Labrador, Canada, 16-20 August, 2015

ABSTRACT
This paper assesses the authenticity of the petrophysical measurements made on core
plugs taken with a new downhole sidewall coring tool that can acquire core-plugs of 1.5
diameter and lengths of 2.5, 3.0 and 3.5. For this purpose, one block of Indiana
limestone and one block of Berea sandstone were used. These blocks were characterized
with a minipermeameter through acquisition of minipermeability grid points.
Two types of twin plugs were cut from each block at locations having similar
permeability range. The first set of plugs was cut from each block using the new
downhole rotary sidewall coring tool with a dimension of 1.5 diameter and 2.5 length.
The second, twin set of plugs was cut with a conventional rotary core-plugging device
used by commercial core laboratories. Both types of plugs were full-size scanned using
micro-CT and measured for routine core properties. Resistivity and NMR measurements
were also made after brine saturation.
The results of the measurements made on the new sidewall-coring plugs were found to be
comparable with the measurements made on the plugs acquired with the conventional
core-plugging device. It can be concluded that the new sidewall cores are suitable for
RCA and further SCAL analyses.

INTRODUCTION
Routine (RCA) and Special Core Analyses (SCAL) are important for field development
projects. One of the main sources of data for this purpose comes from rock samples
which are acquired in the form of drill cuttings, continuous whole cores and sidewall
cores. The most established and well accepted of these rock sample sources is the
continuous whole core - sidewall cores have not gained such acceptance yet. The most
probable reasons may include non-standard dimensions of such cores and their potential
alteration in the process of acquisition.
In the past, the samples from rotary sidewall coring were limited in size, resulting in
limited laboratory evaluation. Recently, a new coring tool has been developed to obtain

SCA2015-035

2/7

1.5 diameter sidewall core samples equivalent to standard laboratory core plugs. With
this new coring capability, operators can fully characterize an extended reservoir interval
in a single sidewall coring descent instead of multiple sidewall coring descents or
multiple stands of whole core retrieval. There are no published documents proving that
these core samples can provide good routine core analysis data. The purpose of this study
is to compare core analysis data from samples obtained from a conventional lab plugging
machine and from the rotary sidewall coring tool (Figure 1):
A

Figure 1: Conventional laboratory plugging machine (A), rotary sidewall coring tool (B), 1.5 diameter
core within the drill bit (C) and 3.5 length sidewall core (D)

ROCK SAMPLES
Two rectangular cross section outcrop blocks were selected for the study: one block of
Indiana limestone and one block of Berea sandstone. These blocks were indexed into a
grid system (rows R and columns C (Figure 1). For each grid section local gas
permeability data were obtained with a mini-permeameter in a first step with the
objective to identify several zones with similar permeability prior to coring. This ensured
the reliability of the comparisons between the core analysis data obtained from twin
samples plugged with the two different techniques.
The permeability range for the Berea sandstone block was found to be between 500 and
800 mD whereas the permeability for the Indiana limestone block ranged from 15 to 150
mD. Figure 2 shows a picture of the Indiana limestone block which was plugged with
both laboratory plugging machine and rotary sidewall coring tool:

Sidewall
Lab
Figure 2: Indiana limestone block and mini-permeability gridding (left), block after coring with
conventional laboratory plugging machine and rotary sidewall coring tool (right)

Four core plugs were cut from each block (Berea and Indiana): on each block, two core

SCA2015-035

3/7

plugs were cut using the conventional laboratory plugging machine, two using the rotary
sidewall coring tool. They were all drilled in zones of close permeability. Tap water was
used as drilling fluid.
The integrity of the plugs did not seem to be affected by the way the plugs were drilled
(Figure 1, D), at least qualitatively. Results of additional measurements are described in
detail below to prove that the properties of plugs cut with the rotary sidewall coring tool
and with conventional plugging machine are not affected by the coring method.

EXPERIMENTAL RESULTS
All outcrop samples were first Soxhlet cleaned and dried before being measured for
routine core properties.
Grain density , ambient helium porosity (based on Boyles law), gas and Klinkenberg
permeabilities Kg and Kkl (using the pressure falloff method), were performed on all
samples. Data results are listed in Table 1:
Block

Mineralogy

Sample Id

Coring type

g
g/cc

Kg
mD

Kkl
mD

Berea

sandstone

BS1

lab

2.64

25.09

704

696.6

4.77

Berea

sandstone

BS2

lab

2.65

25.95

706

698.0

---

Berea

sandstone

BS3

sidewall

2.65

24.84

592

585.2

---

Berea

sandstone

BS4

sidewall

2.65

25.32

699

690.9

4.68

Indiana

limestone

IL4

lab

2.69

15.75

22.4

21.2

8.13

Indiana

limestone

IL10

sidewall

2.69

15.65

58.2

56.21

---

Indiana

limestone

IL14

sidewall

2.70

16.48

29.10

27.74

8.18

Indiana

limestone

IL20

lab
2.69
15.74
Table 1: Dry core plug properties

33.46

31.97

---

Grain density and dry matrix permittivity cannot highlight the presence or absence of
micro-fractures. The measured grain density ranges from 2.64 g/cc to 2.65 g/cc for the
sandstone samples and from 2.69 g/cc to 2.70 g/cc for the limestone samples. The dry
matrix permittivity of four samples was also measured and was found to be in the
expected range (around 4.7 for pure quartz and 8.5 for pure calcite from literature
values in dielectric unit).
As to porosity and permeability, the presence of induced fractures can increase both
values. The porosity ranges from 24.8% to 25.3% for the sandstone samples and from
15.4% to 17.2% for the limestone samples, gas permeability from 585 mD to 728 mD for
the sandstone samples and from 30.4 mD to 56.2 mD.
While three out of the four samples within a group (group BS for Berea Sandstone, group
IL for Indiana Limestone) show very good agreement in permeability and porosity, the

SCA2015-035

4/7

overall range of porosity and permeability data obtained within the group can be
attributed to natural anisotropy. No evidence is apparent that the sidewall coring tool has
damaged the rock during the coring process.
The core plugs were then scanned using X-ray micro-Computed Tomography (micro-CT)
techniques. Figure 3 shows typical 2D cross-sections of two Indiana limestone samples,
one obtained from the conventional laboratory plugging machine (IL4), one from the
rotary sidewall coring tool (IL14):
A

Figure 3: Typical 2D micro-CT cross-sections obtained from conventional sample IL4 (A) and from
sidewall sample IL14 (B)

Micro-CT was used to image core plugs with resolution of 42m/voxel. A high voltage
X-ray tube and different zoom-in objectives allow samples with higher resolution to be
scanned (5m/voxel without physical core mini-plug cutting). Micro-CT image
processing and analysis were carried out by using commercial software and in-house
developments. 3D pore structure representations of samples IL4 and IL14 are shown in
Figure 4:
A

Figure 4: 3D volume rendering from conventional sample IL4 (A) and from sidewall sample IL14 (B)

Both images and 3D volumes in Figure 3 and Figure 4 are very similar. There is no visual
evidence of rock structural alterations such as mini-fractures induced by the coring
method.
MICP data can also highlight the presence of tiny fractures due to the coring process:
tests were run on four end trims from rocks drilled with the two different methods. Figure
5 represents both capillary pressure and pore throat size distribution of two Indiana

SCA2015-035

5/7

limestone and two Berea sandstone rock chips obtained from the end-trims of the samples
cut from the conventional laboratory plugging machine (IL4 and BS1) and from the
rotary sidewall coring tool (IL14 and BS4):
Berea sandstone samples (BS1 in blue & BS4 in red)

Indiana limestone samples (IL4 in blue & IL14 in red)

Figure 5: Mercury injection capillary pressure (MICP) and pore throat size distribution

Figure 5 shows that capillary pressure curves Pc and pore throat size distribution curves
of sidewall and laboratory samples are very similar - there is no evidence of potential
micro-fracture or rock damage due to the coring process.
The MICP parameters obtained from both types of curves are presented in Table 2:
Sample
Id
IL4
IL14
BS1
BS4

Hg

(%)
15.8
16.7
26.3
25.4

Nanopores
1nm<Dia.<1m
(%PV)
42.7
41.0
19.3
19.0

Micropores
1nm<Dia.<1m
(%PV)
57.3
59.0
80.0
80.1

Mesopores
62nm<Dia.<4m
(%PV)
0
0
0.74
0.85

Hg/air
Pe
(psi)
6.24
5.55
1.68
1.34

Swanson
Kg
(mD)
35.7
38.3
518
616

SCA2015-035

6/7

Table 2: MICP data results

Note that MICP porosity and permeability are close to the ones obtained on core plugs.
Four core samples were then saturated with brine (200 kppm equivalent NaCl). Brine
permeability was run on the samples in addition to resistivity and NMR measurements.
Block

Sample Id

Berea
Berea
Indiana
Indiana

BS1
BS4
IL4
IL14

Coring type

Brine perm

FF

mD
@ 1 KHz
@ 1 KHz
lab
536
13.5
1.96
sidewall
532
13.4
1.96
lab
24.0
53.0
2.16
sidewall
33.2
48.5
2.10
Table 3: Brine saturated core plug properties

NMR
%
24.2
24.0
14.8
14.3

Table 3 shows that all values (sidewall versus laboratory) are essentially the same for
each couple of cores, with cementation factors equal to 2.0 for sandstone and 2.1 for
limestone. If fractures would have been induced, the permeability and cementation factor
on sidewall cores would have been found respectively higher and lower.
The NMR porosity was obtained from T2 distribution measurements (Figure 6).

Figure 6: NMR T2 distributions of BS1 and BS4 (left), and IL4 and IL14 (right)

The NMR porosity is slightly lower than the helium porosity, maybe due to some
evaporation during the ambient NMR tests. However, the results are still in agreement
with the laboratory porosity. T2 distributions do not show a difference between sidewall
and conventional samples, showing again that the rocks were not altered by the way they
were cut.

CONCLUSIONS
Core plugs of different rock types were cut using both a standard laboratory plugging
machine and a new rotary sidewall coring tool. All samples were full-size scanned using
micro-CT and measured for routine core properties, resistivity, NMR and dielectrics. The
results of the measurements made on the new sidewall-coring plugs were found to be
comparable with the measurements made on the plugs acquired with the conventional

SCA2015-035

7/7

core-plugging device. It is concluded that the sidewall cores are suitable for RCA and
SCAL analyses. The SCAL work is in progress at the Houston Schlumberger Reservoir
Laboratory; the preliminary results confirm the good quality of the sidewall cores. To
summarize, the new rotary sidewall coring tool is now an alternative for acquiring high
quality downhole core samples with proper core size for laboratory analyses.

SCA2015-036

1/6

MAGNETIC SUSCEPTIBILITY OF DRILL CUTTINGS IN

A NORTH SEA OIL WELL: A RAPID, NONDESTRUCTIVE MEANS OF CHARACTERIZING
LITHOLOGY
1

Arfan Ali, 2David K. Potter and 3Andrew Tugwell

1
Shell UK Limited, Aberdeen, UK
2
Department of Physics, University of Alberta, Edmonton, Canada
3
Advanced Downhole Petrophysics Limited, Aberdeen, UK
This paper was prepared for presentation at the International Symposium of the Society
of Core Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Magnetic susceptibility measurements provide a non-destructive method to rapidly
characterize drill cuttings at the wellsite or laboratory. Our previous studies on core
plugs, slabbed core and whole core have demonstrated that magnetic measurements can
identify lithological variations, estimate mineral content, and correlate with key
petrophysical properties (such as permeability) and with downhole gamma ray data. In
the present study volume magnetic susceptibility measurements were undertaken on drill
cuttings from a North Sea oil well using a portable low field magnetic susceptibility
sensor. The values were then converted to mass magnetic susceptibility by dividing by
the bulk density of each sample. The results clearly indicated the main lithological
zonations in the well, and provided estimates of basic mineral type (diamagnetic versus
paramagnetic or ferrimagnetic) significantly quicker and cheaper than undertaking XRD
measurements. The magnetic results also showed a correlation with the downhole LWD
(logging while drilling) gamma ray profile. Interestingly, the correlation was the opposite
way round to that observed in most other reservoirs we have studied. However, this
provided additional mineralogical information for the well in the present study. Normally
a low gamma ray signal (e.g., in a clean sandstone interval) would give a low or negative
magnetic susceptibility due to diamagnetic quartz, whereas a high gamma ray signal (e.g.,
in shale) would give a higher magnetic susceptibility signal due to paramagnetic clays
etc. In the present study many of the low gamma ray sandstone intervals exhibited a
higher magnetic susceptibility signal, which indicated that there are additional
paramagnetic and/or ferrimagnetic minerals present in those intervals in addition to the
main diamagnetic matrix mineral (quartz). These additional higher magnetic
susceptibility minerals can, for example, be due to a strongly paramagnetic mineral such
as siderite, small amounts of a ferrimagnetic mineral such as magnetite or the canted
antiferromagnetic mineral hematite. These minerals can affect the permeability, and may
explain why productivity has been lower in this well.

SCA2015-036

2/6

INTRODUCTION
Drill cuttings have generally been a highly under utilized resource in the petroleum
industry. Apart from some exceptions, such as the DarcylogTM method to determine
permeability [1] and porosity from drill cuttings, there have been very few published
studies that have derived mineralogical or petrophysical properties from drill cuttings.
Our previous magnetic susceptibility work on conventional core plugs [2], slabbed core
[3,4] and whole core [5] proved fruitful in demonstrating correlations between magnetic
susceptibility, mineralogy (especially clay content) and key petrophysical properties such
as permeability. For instance, our work on North Sea oilfields [2,3] showed strong
correlations between magnetically derived illite content and permeability. The present
paper describes how magnetic susceptibility measurements on drill cuttings can rapidly
and non-destructively identify mineralogical / lithological variations. In the present study
we made 421 measurements from 157 bags of drill cuttings from a North Sea oil well.
We were told by the operating company that there was an issue in this well, in that the oil
production was lower than expected compared to other wells in the same field. We were
therefore asked to see if the drill cuttings measurements might provide a possible reason
for the lower production.

METHODS
The drill cuttings measurements were performed at Iron Mountain in Dyce, Aberdeen,
UK. The sample boxes containing bags of drill cuttings were arranged in order of depth
(Figure 1 (a)). Each sample bag contained drill cuttings from a particular depth interval.
10cc plastic vials were used as sample pots and were filled with randomly selected
spoonfuls of cuttings from each sample bag (Figure 1 (b)). Volume magnetic
susceptibility measurements were undertaken on the vials containing the drill cuttings
using a small portable low field Bartington MS2B magnetic susceptibility sensor
connected to laptop via an MS3 meter (Figure 2). Each sample vial containing drill
cuttings was also accurately weighted. Volume magnetic susceptibility measurements
were then simply converted into mass magnetic susceptibility as follows:

=/

(1)

where is the mass magnetic susceptibility, is the measured volume magnetic

susceptibility, and is the bulk density of the drill cuttings in the vial. The advantage of
using mass magnetic susceptibility is that it removes any small effects due to porosity,
which can affect the volume magnetic susceptibility measurements. This includes the
intrinsic porosity of the individual drill cuttings and also the porosity between
individual drill cuttings. This ensures that drill cuttings with an identical mineralogy will
give exactly the same mass magnetic susceptibility value (whereas they would give
different volume magnetic susceptibility values if the amount of those cuttings is
different in each sample vial). During the measurement procedure a calibration sample
was also measured every hour to check whether there was any drift in the sensor.

SCA2015-036

3/6

RESULTS AND DISCUSSION

The results showed that the mass magnetic susceptibility was positive throughout the
9,000 ft section of the well. Moreover, an analysis of the magnetic results with the LWD
(logging while drilling) gamma ray data (Figure 3) shows that the sandstone intervals
identified by the low gamma ray sections (shaded yellow on the gamma ray log) have
anomalously high mass magnetic susceptibility values (shaded yellow on the mass
magnetic susceptibility data). This was very unexpected, since pure quartz sandstone
should have a very low, negative mass magnetic susceptibility (since quartz is
diamagnetic). The large positive mass magnetic susceptibility values immediately tell us
that there are additional paramagnetic and /or ferrimagnetic minerals contained within the
sandstones. Possible candidates for these minerals are siderite (a strongly paramagnetic
iron carbonate), the ferrimagnetic mineral magnetite, or the canted antiferromagnetic
mineral hematite. The additional minerals in the sandstone intervals are not likely to be
paramagnetic clays such as illite or chlorite, otherwise they would have given a much
higher gamma ray signal. Whilst there was no X-ray diffraction (XRD) data available in
this well, there was XRD data for just 10 samples from another well in the same oilfield.
The XRD data indicated average values of around 85% quartz, around 4% K-feldspar, 12% plagioclase, 1-2% pyrite, around 6% kaolinite (which is diamagnetic) and only trace
amounts of illite. No evidence for siderite was seen in this data. Of course the XRD data
from the other well may not necessarily be representative of the mineralogy in the present
studied well. However, it seems more likely that the observed high mass magnetic
susceptibility values are due to a mineral like magnetite or hematite. Small amounts of
these minerals would not necessarily be readily identified by XRD, yet would give the
observed high magnetic susceptibility values.
The additional minerals identified by the magnetics in the sandstone intervals may have
important implications for the petrophysical properties. We have previously demonstrated
[6] that small amounts of fine-grained hematite, for instance, can have a very dramatic
effect on reducing the permeability. The additional minerals in the sandstones may
therefore help to explain why the well in our present study was not as good a producer as
other wells in the same oilfield. It is also important to be able to readily identify
additional paramagnetic and ferrimagnetic minerals, since nuclear magnetic resonance
(NMR) log data can be significantly affected by these minerals.

CONCLUSIONS
1. The drill cutting measurements gave generally high positive mass magnetic
susceptibility values in the sandstone intervals (low gamma ray). This is very unusual
since pure quartz sandstone is diamagnetic with a low, negative magnetic susceptibility.
The magnetic measurements demonstrated that there must be significant additional
paramagnetic and/or ferrimagnetic minerals in the sandstone intervals. Potential
candidates could be the paramagnetic mineral siderite, the ferrimagnetic mineral
magnetite, or the canted antiferromagnetic mineral hematite.

SCA2015-036

4/6

2. The presence of these additional minerals in the sandstones is likely to affect the
permeability of the sandstones, and this may in turn be responsible for the lower
productivity that has been observed by the operating company in this well.
3. The study demonstrated that magnetic measurements on drill cuttings provide a rapid
means of identifying of mineralogical variations over large intervals and are a potentially
important supplement to XRD data.

FUTURE WORK
Either high field magnetic susceptibility or magnetic remanence measurements would
enable us to determine whether the additional minerals in the sandstones were due to, for
instance, small amounts of magnetite or hematite (note that XRD would not necessarily
be helpful in this respect). Also it would be useful to undertake magnetic susceptibility
measurements on drill cuttings in one of the good producing wells in the same field. If we
obtained a normal diamagnetic signal in the sandstones of the good producer then it
would be further evidence to suggest that the additional minerals identified in the present
well were responsible for its lower productivity. We are waiting for permission from the
operating company to undertake these extra studies.

REFERENCES
1. Lenormand, R., Bauget, F., Ringot, G., 2010. Permeability measurement on small
rock samples. 2010 International Symposium of the Society of Core Analysts, Halifax,
Nova Scotia, Canada. Paper SCA2010-32 (12 pages).
2. Potter, D. K., 2007. Magnetic susceptibility as a rapid, non-destructive technique for
improved petrophysical parameter prediction. Petrophysics, 48 (issue 3), 191-201.
3. Ali, A., Potter, D. K. and Tugwell, A., 2014. Correlation between magnetic properties
and permeability: results from a new case study in the North Sea. 2014 International
Symposium of the Society of Core Analysts, Avignon, France. Paper SCA2014-077 (6
pages).
4. Agbo, B. C. and Potter, D. K., 2014. Novel high resolution probe magnetic
susceptibility and comparison with wireline gamma ray and grain size in an Albertan
oil sand well. SEG Technical Program Expanded Abstracts: pp. 2590-2594. doi:
10.1190/segam2014-1140.1.
5. Potter, D. K., Ali, A., Imhmed, S. and Schleifer, N., 2011. Quantifying the effects of
core cleaning, core flooding and fines migration using sensitive magnetic techniques:
implications for permeability determination and formation damage. Petrophysics, 52
(issue 6), 444-451.
6. Potter, D. K., Ali, A. and Ivakhnenko, O. P., 2009. Quantifying the relative roles of
illite and hematite on permeability in red and white sandstones using low and high
field magnetic susceptibility. 2009 International Symposium of the Society of Core
Analysts, Noordwijk aan Zee, The Netherlands. Paper SCA2009-11 (12 pages).

SCA2015-036

5/6

(a)

(b)

Bag of drill
cuttings
Vials of drill cuttings
ready for measurement
in the sensor shown in
Fig. 2

Figure 1. (a) Bags of drill cuttings arranged in order of their depth in the well. (b) Samples of drill cuttings
being put into the vials for measurement.

Portable
MS2B sensor

Figure 2. View of the portable MS2B magnetic susceptibility sensor, connected to a laptop.

SCA2015-036

6/6

Mass Magnetic Susceptibility

1
10-8 m3 kg-1
10000

Figure 3. Mass magnetic susceptibility results for 421 measurements over 9,000 ft are shown in the left
hand column (scale runs from 110,000 x 10-8 m3 kg-1). An arbitrary cut-off (green/yellow) is given at
around 17 x 10-8 m3 kg-1 to compare easily with the low gamma ray (yellow) values in the adjacent column.

SCA2015-037

1/7

CARBONATE NMR PERMEABILITY ESTIMATES BASED

ON THE WINLAND-PITTMAN MICP APPROACH
Edmilson Helton Rios1, Adam Keith Moss2, Timothy Neil Pritchard2, Ana Beatriz
Guedes Domingues2 and Rodrigo Bagueira de Vasconcellos Azeredo3
1
National Observatory
2
BG Group plc
3
Fluminense Federal University
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Nuclear magnetic resonance (NMR) logging is widely used for continuous downhole
permeability estimates. But long before NMR, laboratory measurements of capillary
pressure curves were used for the same purpose. Remarkably, the mercury injection
capillary pressure (MICP) technique was developed for predicting the permeability of
small and irregular-shaped samples, such as drilling cuts. In the Winland-Pittman MICP
approach, pore-throat radii corresponding to different Hg saturations were individually
correlated with porosity and permeability. The saturation where the best correlation is
found defines the critical aperture. By adapting this methodology for NMR context, this
work introduces a novel concept of cumulative saturation cutoff applied on the transverse
relaxation time (2 ) distributions. The performance and the fitting coefficient as a
function of saturation cutoff were graphically analyzed for a group of North Sea and
Middle East Carbonate core plugs. Critical saturation cutoffs for both pure and sizescaled NMR distributions delivered better permeability estimates when compared with
the standard logarithmic mean estimator.

INTRODUCTION
Kozeny developed a theory, later improved by Carman, in which porous rocks
are regarded as bundles of capillary tubes [1]. The so-called Kozeny-Carman
equation takes many forms, including the following:

= ( ) ,

(1)

where , and are general coefficients for adjustment, is porosity, is

volume-to-surface-area ratio, also known as the hydraulic radius. For simple pore
geometries, can be written as , where is the pore radius and is a pore
shape factor (e.g. 1 represents flat or flake-like pores, 2 represents open-ended
cylindrical pores and 3 represents spherical pores). From Equation 1, Winland [2]

SCA2015-037

2/7

and then Pittman [3] related MICP-derived pore-throat radius with

permeability and porosity, as following:
= 1 + 2 + 3 ,

(2)

where 1 = 1 , 2 = and 3 = (log) + ( ), with being the

pore body-to-throat ratio ( ). = 2 , with and being the airmercury interfacial tension and the surface contact angle, respectively, and is
the capillary pressure, selected for a specific saturation of the MICP curve ( ).
The best-performing correlation defines the aperture that best describes
permeability.
Seevers [4] and then Kenyon [5] showed that nuclear magnetic resonance (NMR)
porosity and an average of the longitudinal 1 or transverse 2 NMR relaxation
times could be indirectly used as the independent variables of Equation 1 if
considered the fast diffusion regime and assuming no diffusional relaxation
effects due to magnetic field gradients nor diffusion pore coupling [6]. The
Seevers-Kenyon NMR estimator can be written for 2 (the widely used in
modern NMR logging) as follows:

(2 ) ,
=

(3)

where = 2 is a pre-multiplier factor that includes the 2 surface relaxivity

(2 ), = (2 ) is the NMR-derived total porosity and 2

(12 )( ) is the logarithmic mean of the (2 ) distribution such that 2

=
(2 )
2
. Literature coefficients are commonly used in the logging suites under an
2
4
estimator usually called as Schlumberger-Doll-Research model, =
2
.
In this work, the Winland-Pittman approach was adapted for the NMR context as
follows:

(2 ) ,
=
(4)
where = 2 = (103 2 ) , = 2 and = 11 are fitting
coefficients, the 2 is a relaxation time selected for a fixed point in the cumulative
saturation curve. The saturation cutoff is applied to the cumulative curves computed
from the normalized (or saturation-based) NMR distributions (2 ), such that
(2 ) = 1. Several cutoffs are tested using Equation 6 and the -based
estimates are compared to the standard estimator.
NMR carbonates are generally less sensitive to variations in pore body size, thus 2
distributions were size-scaled based on a NMR-MICP data integration. Because true
surface relaxivity is very complex to be determined and depends on the method
used, a simple scaling factor defined as the ratio of the logarithmic means
between MICP and NMR distributions were employed. This size-scaling factor

SCA2015-037

3/7

encompasses both the surface relaxivity and the geometry of the pore system
(both assumed to be constant within all pore-size families) as follows:

= 2 ,

(5)

where is the logarithmic mean of the pore radius distribution, is a pore

shape factor for simple geometries and is the pore body-to-throat ratio. By
solving for 2 in Equation 5 and inserting it into Equations 3 and 4, size-scaled
NMR permeability estimators can be obtained where the pre-multiplier
incorporates the geometry of the pore system such that = ( ) .

EXPERIMENTAL
Carbonates rocks from Cretaceous chalk fields were studied: 14 from Norwegian
reservoirs in North Sea (the Valhall, Hod and Tommeliten fields) and 15
diagenetic chalks from Middle East (the Thamama C Formation, Bu Hasa Field,
Abu Dhabi). Core plugs (1.5 x 1.8) and small core-end trims (visually
representative of the plugs) were cleaned of native fluids via Soxhlet extraction
and then dried in an oven [7]. The MICP measurements were performed in an
AutoPore II 9220, with a filling pressure of 0.1 bar followed by one hundred
pressure steps with maximum pressure of 4,130 bar. Plugs were routine core
analysed (RCAL) using helium gas at an overburden pressure of 20 bars and then
they were fully saturated with a 50,000 ppm brine solution for NMR
measurements. The 1H NMR T2 measurements of core samples were performed
on a bench-top NMR analyser at 2 MHz and 30oC using the CarrPurcell
MeiboomGill (CPMG) technique [8], with 4,050 spin echoes spaced by 700 s
(). Under sufficient time for fully spin polarization 10s (), signal averages
were performed until a minimal signal-to-noise ratio (SNR) of 100. A distribution
window with one hundred relaxation-time bins logarithmically spaced from 0.1 to
10,000 ms was set for 2 .

RESULTS
The normalized MICP, NMR, and size-scaled NMR distributions are shown on
Figure 1a, b and c, respectively. Left side is the incremental curves and right side
is the cumulative curves (two saturation cutoffs are showed as examples),
respectively. For MICP and size-scaled NMR, micro-, meso- and macro-porosity
ranges are outlined with the 0.3 and 3 m threshold, in accordance with [9]. All the
curves are labelled based on the increasing values of the plugs. Small
apertures and pores corresponds to plugs with low (predominantly North
Sea Chalks), whereas large apertures and pores corresponds to plugs with high
(predominantly Middle East Chalks). The similarity between routine and
special core analysed porosity in Figure 2a, indicates that gas, Hg and brine
probed representative pore spaces and that isolated fluid-filled pores may not
have been present. Figure 2b well illustrate that a simple power regression cannot

SCA2015-037

4/7

simultaneously handle the different pore-perm trends exhibited by the chalk

groups. Additional information such as apertures and pore sizes may help in
obtaining a global permeability estimator.
The r-squared and the fitting coefficients for the estimators and (with 50 and
85%) are presented in Table 1 for pure and size-scaled 2 . Figure 3 plots the R-square
and multiple linear regression coefficients obtained with saturation cutoffs from 5
to 85% under increments of 5%. The y-axis lower limit (0.37) corresponds to the
particular case where coefficient is null (Figure 2b). The performance of for
each case is also indicated in the plot. Size scaling improved permeability estimation
considerably for all case, producing lower (for porosity) and higher (for
2) coefficients values. It indicates that the size-scaled 2 distributions are
more effective to explain permeability and maybe can mitigate some of the
difficulties encountered in carbonate NMR, such as variations in surface
relaxivity and diffusive pore coupling. Figure 5 shows the response curves for the
and 85% before and after size scaling. In their original works with sandstones,
Winland and Pittman described an equivalent cumulative saturation (from micro
to macro) of 65% and 75%, respectively.

CONCLUSION
Inspired by the work of Winland and Pittman on MICP data, a novel NMR
permeability estimator was introduced based on new concept of cumulative
saturation cutoffs. 2 associated with macroporosity reveals much better results
than the standard 2 logarithmic mean estimator. To overcome the lower
sensitivity of carbonate NMR to pore and throat variation, this new approach was
also evaluated after 2 distribution size scaling. The results exceeded the
performance compared to the estimators with pure 2 . The study confirms the
potentiality of saturation-based NMR permeability estimator and further
evaluation is in course for logging application.

ACKNOWLEDGEMENTS
ANP (Compromisso com Investimentos em Pesquisa & Desenvolvimento) and BG Group
plc. for funding and providing the database. Derrick Green for reviewing and suggesting
improvements in the manuscript. Edmilson Helton Rios is grateful to CNPq and
Petrobras for his research fellowship.

REFERENCES
1. Carman, P.C., 1956. Flow of gases through porous media, Butterworths Scientific
Publications, London.
2. Kolodzie Jr., S., 1980, Analysis of pore throat size and use of the WaxmanSmits
equation to determine OOIP in spindle field, Colorado, SPE 55th Annual Fall
Technical Conference.

SCA2015-037

5/7

3. Pittman, E.D., 1992. Relationship of porosity and permeability to various parameters

derived from mercury injection-capillary pressure curves for sandstone, American
Association of Petroleum Geologists (1992), 76 (2), 191198.
4. Seevers, D.O., 1966. A nuclear magnetic method for determining the permeability of
sandstones, Trans. 7th Ann. SPWLA Logging Symp.,Tulsa.
5. Kenyon, W.E., 1992. Nuclear magnetic resonance as a petrophysical measurement,
Nucl. Geophys., 6(2), 153-171.
6. Dunn, K.J., Bergman, D.J. & Latorraca, G.A., 2002. Nuclear Magnetic Resonance
Petrophysical and Logging Applications, Pergamon Press, Oxford.
7. American Petroleum Institute, 1998. Recommended Practice for Core Analysis
Procedure, 2nd edn, API RP 40, (Washington, DC:API)
8. Meiboom, S. & Gill D., 1958. Modified spin-echo method for measuring nuclear
relaxation times, Rev. Scien. Inst., 29, 688-691, doi:10.1063/1.1716296.
9. Marzouk, I., Takezaki, H., & Suzuki, M., 1998. New classification of carbonate rocks
for reservoir characterization, 8th SPE Inter. Petrol. Exhib. Confer.,10p
Table 1: Performance and regression coefficients for pure and size-scaled NMR permeability estimators.

50%
85%

Pure 2

0.70 4E-05 2.82 1.33

0.71 3E-05 2.76 1.35
0.81 1E-05 2.87 1.37
2

Size-scaled 2

0.89 7.65 0.01 2.54

0.88 7.65 -0.04 2.53
0.95 7.14 1.01 2.10
2

SCA2015-037

6/7

Figure 1: a) ( ) and b) (2 ) with 3D plots showing each chalk group as inserts. c) Size-scaled NMR
distribution ( 2 ) with the thin section of the lower and higher permeability chalks as inserts.

Figure 2: a) Routine core analysed porosity (standard reference) versus special core analysed porosities
(MICP & NMR). b) RCAL permeability (standard reference) versus NMR porosity.

Figure 3: a) Performance and b) regression coefficients of the saturation-based estimator for both pure and
size-scaled 2 distribution.

SCA2015-037

7/7

Figure 4: a) Standard log mean estimator and b) the best saturationbased estimator response curves.

SCA2015-038

1/7

ON THE MEASUREMENT OF PORE GEOMETRY: A

COMPREHENSIVE PETROPHYSICAL STUDY OF
CONVENTIONAL ROCKS
Babak Salimifard1; Douglas W. Ruth1; Derrick Green2; Brent Nassichuk3
1
University of Manitoba, 2Green Imaging Technologies, 3Trican Geological Solutions
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Various techniques have been developed over the years for characterizing pore structure beyond
a simple visual description. These tests provide qualitative data for both reservoir evaluations in
the short run and reservoir simulation in the long run. In this study, mercury porosimetry (MP),
low field (2MHz) nuclear magnetic resonance (NMR) relaxometry, centrifuge drainage tests and
flow tests were run on 11 plugs of a mix of sandstones, limestones, dolomites and chalk.
Initially, a representative elemental volume (REV) which uses pore size distribution (PSD) data
and porosity to simulate the pore network is discussed. The model is later used to predict
permeability and predictions were compared with gas flow measurements. NMR and centrifuge
data are coupled to derive capillary pressure curves and the results are compared with MP
derived capillary curves. The results indicate that there is significant difference between the two
capillary curves based on the degree of heterogeneity of the samples. NMR and centrifuge data
are also used to come up with a method to study the degree of heterogeneity of the plugs based
on which it was decided which PSD (MP, NMR or centrifuge) would be the better representative
of the pore network. This PSD is then used to make new permeability predictions using the REV
model and the results show significant improvement over previous predictions. This study proves
that MP, NMR and centrifuge tests provide complimentary information that is crucial to pore
network simulation and analysis. This study also highlights the fact that rocks are complex and
one test cannot represent the properties of the pore network, making it is necessary and beneficial
to run complimentary tests to better understand the pore network. It also verifies the applicability
of methods that combine results of these tests to assess the rock pore network.

INTRODUCTION
Pore network simulation and reservoir rock characterization have been of interest to researchers
for almost a century. Initial approaches were mainly focused on permeability prediction. There is
a vast literature that deals with permeability prediction; however most pioneering methods were
based on experimental correlations [1]. Porosity-permeability cross plots are a good example of
these approaches. Later researchers tried to predict permeability by simulating the pore network
and applying fluid flow principles to the simulated network. Carman and Kozeny suggested that
the porous network can be simulated using a bundle of non-interconnected tubes of varying radii
[2]. This work has been the bedrock for future modeling by various researchers. Berg and Van
Baaren used the same mentality and tried to correlate tube diameter with grain size distributions
[3, 4]. Works of Leveret, Purcell, Thomeer and Swanson on the other hand dealt directly with
pore entry pressures derived through Mercury Porosimetry (MP) [5, 6, 7, 8]. Following the work
of Purcell, Ruth et al. suggested that the pore network can be simulated using a single REV,

SCA2015-038

2/7

correlating permeability with an average diameter and an average length of the conduit as
follows [9]:
2
(1)
k=
32 2
Here k is permeability, is porosity, is the representative diameter and is tortuosity. In their
model they used Purcells mean tube diameter as follows:
1
2

= (4 cos)
0

dSv
(Pc )2

(2)

Here, Sv is vacuum saturation, Pc is capillary pressure and and are surface tension and
contact angle in the MP experiment respectively. They also incorporated electrical properties to
account for the extended length of the mean flow path due to tortuosity. Archies formation
factor was used to define tortuosity as a function of porosity, cementation factor (m) and
lithology factor (a) as follows:
2 =

a
m1

(3)

Eventually, they rewrite their model as:

k=

m ( cos)2 1 dSv

2
2a
0 (Pc )

(4)

PRELIMINARY WORK
This REV model was initially tested against a mix of 25 sandstone and carbonate samples from
an offshore Ghana formation of Turonian age. Once provided with the mercury intrusion and
formation factor data, the method was able to predict permeability with a mean error of less than
35%. Results of this study are presented in Figure 1. The two dashed lines represent the 50% and
100% error bars and the solid line is the one-to-one line where prediction matches measurement.

Figure 1 - Comparison between the calculated permeability and the measured permeability

In two other attempts, the REV models predictions were compared with predictions of CarmanKozeny, Swanson, Timur-Coates and Schlumberger-Doll-Research (SDR) methods [10, 11]. In

SCA2015-038

3/7

all cases except Swansons, predictions made using the REV model outperforms predictions
made using other methods. Swansons however, makes better predictions than the REV method
because it is optimized for the data set, thus making it less reliable when not optimized for the
specific data set or formation.
After all, out of all the compared models, the REV model is the only one that solely relies on
petrophysical properties of the rock and does not incorporate any fitting parameter. However, a
claim cannot be made about the general applicability of this method particularly because the
mean diameter of the conduit is calculated based on MP measurements done on small samples
which are not always representative of the pore network. To start with, MP data is representative
of the pore volumes accessible through specific pore throat sizes. In clean, homogenous
sandstones where pore throats are supposed to be the main obstacles against fluid flow, and when
pore throat to pore body ratios are not large, using throat size distribution (TSD) instead of pore
size distribution (PSD) would not result in misinterpreting the pore network. However, in
carbonates, where there is a wider variety of pore throat to pore body ratios using TSD would
result in unreliable predictions.
Also, when it comes to field applications, MP is not the most applicable method because its not
easy to generate a continuous TSD profile for the entire height of the reservoir. However,
logging techniques such as Nuclear Magnetic Resonance (NMR) can produce a continuous PSD
profile for the entire height of the reservoir adjacent to the well bore.

RESULTS AND DISCUSSION

Techniques have been developed to derive capillary pressure from NMR data [12, 13]. Green
[14] developed a model to produce capillary pressure data by coupling centrifuge tests with low
field NMR measurements. In their method they calculated the capillary pressure distribution at
each centrifuge spin by using Hassler-Brunner boundary conditions and applying it to Darcys
law to end up with the following formula:
1
Pc (r) = 2 (r02 r 2 )
2

(5)

Here Pc(r) is the capillary pressure at each point along the sample, is the different in the
densities of saturating and displacing fluids, is the spin velocity and r0 is the distance from
center of rotation. After each spin the sample is removed from the centrifuge and brine saturation
along the sample is measured using low field NMR. Finally saturation at each r is coupled with
capillary pressure at that point and a global capillary pressure curve is plotted for the sample.
This test was run on the 11 samples used by Salimifard et al. [11] and results show that for the
more homogeneous samples, MP capillary pressure curves match with the NMR capillary
pressure curves. However, for the less homogeneous samples, i.e. samples E and J, the two
capillary pressure curves differ. Figure 2 shows the results of the comparison. It should be
mentioned that choosing the right contact angle for analyzing centrifuge capillary pressures is of
utmost importance. In this study a fixed value of zero degrees was used for brine-air system
which might not be the case for some rocks.

SCA2015-038

4/7

Figure 2 a comparison between capillary pressures derived from MP and Centrifuge tests. Figure a (on the left)
shows how MP and centrifuge capillary pressures match for a homogeneous sample while Figure b (on the right)
highlights the fact that for less homogeneous samples, proper sampling can play an important role in capillary
pressure determination.

Results of the same test are used to quantify the degree of inhomogeneity of the 11 samples.
Each sample is divided in 4 segments and the total change in brine volume in each segment is
monitored after each centrifuge spin. An average capillary pressure was calculated for each
segment using Greens approach and the change in brine volume was coupled with that capillary
pressure to construct the capillary pressure curve for each segment. The inhomogeneity of the
sample then can be analyzed based on how well the capillary pressure curves for the four
segments match. Obviously a good match means the sample is fairly homogenous and indicates
that the end piece used for mercury porosimetry is a good representative of the core plug. As
expected, capillary pressures for the four segments of samples E and J do not match which
highlights the degree of inhomogeneity of the two samples.

Figure 3 a comparison between capillary pressures calculated for the four sections of a homogenous and a less
homogenous sample. Figure a (far left) shows how the four segments are arranged on the core plug, segment 1 being
the closest to the center of rotation and segment 4 being the farthest. Figure b (middle) shows capillary pressures of
the four segments of a homogenous plug compared with centrifuge capillary pressure of the plug. Figure c (far right)
shows the same results for a less homogenous sample.

In presence of heterogeneity, it is believed that the NMR derived PSD is a better representative
of the pore network as it sees the entire plug, not just an end piece. Marschalls and Kleinbergs
methods are used to convert NMR T2 distributions to capillary pressure data by matching the
NMR T2 distribution curves with MP capillary pressure curves. As Marschall explained, NMR

SCA2015-038

5/7

data can be correlated with pore size when the plug is 100% brine saturated using the following
equation:
1
S
=
T2
V
S

(6)
2

Here is relaxivity and V is the surface to volume ratio, which can be replaced by r when a
cylindrical pore network (a bundle of capillary tube) is assumed. Here is calculated using
=

cos
PT2

(7)

where is defined as the effective relaxivity to account for the fact that NMR responds to pore
bodies whereas MP responds to pore throats. An effective relaxivity is calculated for all 11
samples by finding the best match between the NMR T2 and MP capillary pressure distributions.
These relaxivities match with the values reported by Marschall and Kleinberg and are tabulated
in Table 1. Relaxivities are used to produce capillary pressure curves from NMR T2 distributions
and permeability is recalculated using the REV method based on the new capillary pressure
curves.
Table 1 a compilation of sample properties and prediction results using different techniques
Sample Name
A
B
C
D
E
F
G
H
I
J
Porosity (%)
17.3
22.1
19.9
19.9
28.8
15.5
23.3
22.3
25.9
9.2
Measured
62.8 1060 7050
107
58.9
106
52.3 3810 1150 0.45
Permeability (mD)
MP-Predicted
210
1090 2500
908
123
336
52
1220
892
6.13
Permeability (mD)
Calculated
12.8
17.1
21.3
2.7
13.1
2.7
24.0
12.8
17.6
2.4
Relaxivity (m/s)
NMR-Corrected
180
1780 3960
184
590
250
86.9 3030 1030 4.38
Permeability (mD)

K
18.3
283
302
15.5
295

CONCLUSION
Results shown in the table above clearly show that excluding sample E, using NMR derived
capillary pressure curves improves the predictions for less homogeneous samples, namely for
samples D and J. Even though the NMR T2 distributions are calibrated with MP capillary
pressure curves, this approach improves the predictions by 113% (from 245% using MP data to
132% using NMR data, excluding sample E for both cases). Also, the T2 distribution curve was
recalibrated with centrifuge capillary pressures for sample J and a new permeability prediction of
1.07 mD was achieved.
Sample E is a chalk sample which is highly compressible and sensitive to confining pressure.
This is confirmed both by pressure-adjusted flow tests and by the NMR inhomogeneity test. As a
result, because permeability tests are performed under confining pressure, NMR PSD might not
be the best representative of the pore network when compared to confined flow tests, because
NMR tests are done under no overburden pressure. Using NMR PSD would result in over
estimating the pore sizes and over estimating the resultant permeability.

SCA2015-038

6/7

In conclusion NMR derived capillary pressures have proven to be properly representative of the
pore network and are capable of making reliable permeability predictions. However, theres still
a need for calibrating the T2 relaxation times with either centrifuge or MP capillary pressures.
N2 adsorption techniques on the other hand can provide a direct measurement of surface to
volume ratios, through which relaxivity values for each sample can be calculated [15]. In another
approach, Fleury discussed a novel method to directly measure relaxivity values from NMR
diffusion curves and concluded that the results compare with values derived through N2
adsorption measurements [16]. Unfortunately there wasnt sufficient data to apply these methods
to the current set of samples. Once relaxivity is obtained for a formation, T2 relaxation times can
directly be converted to PSD data and a continuous PSD profile can be generated for the entire
height of the reservoir from NMR logs.

REFERENCES
1. Nabovati A. and A. C. M. Sousa, "Fluid Flow Simulation in Random Porous Media At Pore
Level Using Lattice Boltzman Method," Journal of Engineering Science and Technology,
vol. 2, no. 3, pp. 226-237, 2007.
2. Carman P. C., Flow of Gasses Through Porous Media, NewYork: Academic Press Inc.,
1956.
3 Berg R. R., "Method for Determining Permeability From Reservoir Rock Properties,"
Transactions - Gulf Coast Association of Geological Societies, vol. 20, pp. 303-335, 1970.
4 Van Baaren J. P., "Quick-Look Permeability Estimates Using Sidewall Samples and Porosity
Logs," in 6th Annual European Logging Symposium, Society of Professional Well Log
Analysts, 1979.
5 Leverett M. C., "Capillary Behavior in Porous Solids," Transactions of AIME, vol. 142, pp.
152-169, 1941.
6 Purcell W. R., "Capillary Pressure - Their Measurement Using Mercury and the Calculation
of Mercury Therefrom," Journal of Petrleum Technology, vol. 1, no. 2, 1949.
7 Thomeer J. H. M., "Introduction of a Pore Geometrical Factor Defined by the Capillary
Pressure Curve," Journal Of Petroleum Technology, vol. 12, no. 3, 1960.
8 Swanson B. F., "A Simple Correlation Between Permeability and Mercury Capillary
Pressure," Journal of Petroleum Technology, vol. 33, no. 12, pp. 2488-2504, 1981.
9 Ruth D. W. , C. Lindsay and M. Allen, "Combining Electrical Measurements and Mercury
Porosimetry to Predict Permeability," in International Symposium of the Society of Core
Analysts, Aberdeen, Scotland, 2012.
10 Salimifard B., D. W. Ruth, M. Allen and C. Lindsay, "Applicability of Carman-Kozeny
Equation for Well-Consolidated Samples," in International Symposium of the Society of
Core Analysts, Napa, California, 2013.
11 Salimifard B., D. W. Ruth, D. Green and D. Veselinovic, "Developing a Model to Estimate
Permeability from Other Petrophysical Data," in International Symposium of the Society of
Core Analysts, Avignon, France, 2014.
12 Marschall D., J. S. Gardner, D. Mardon and G. R. Coates, "Method for Correlating NMR
Relaxometry and Mercury Injection Data," in Annual Simposium of the Society of Core
Analysts, 1995.
13 Kleinberg R. L., "Utility of NMR T2 Distributions, Connection with Capillary Pressure,

SCA2015-038

7/7

Clay Effect, and Determination of the Surface Relaxivity Parameter," Magnetic Resonance
Imaging, vol. 14, no. 7/8, pp. 761-767, 1996.
14 Green D. P., "Capillary Pressure Curves Determined by Direct Measurement of the
Saturation using Magnetic Resonance Imaging," Canaadian Well Logging Society, vol. 28,
no. 1, pp. 20-25, May 2009.
15 Saidian M., U. Kuila, S. Rivera, L. J. Godinez and M. Prasad, "Porosity and Pore Size
Distribution in Mudrocks: A Comparative Study for Haynesville, Niobrara, Monterey and
Eastern European Silurian Formations," in Unconventional Resources Tecnology
Conference, Denver, Colorado, 2014.
16 Fleury, M. "NMR Surface Relaxivity Determination Using NMR Apparent Diffusion Curves
and BET Measurements" in International Symposium of the Society of Core Analysts,
Calgary, Canada, 2007

SCA2015-039

1/6

ANALYTICAL INTERPRETATION METHODS FOR

DYNAMIC IMMISCIBLE CORE FLOODING AT
CONSTANT DIFFERENTIAL PRESSURE
Jie Cao, Xiaolong Liu, Lesley James, Thormod Johansen
Memorial University of Newfoundland
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Relative permeabilities are usually determined from unsteady state immiscible core
flooding experiments under constant differential pressure or constant flow rate conditions.
In the latter case, the classical fractional flow theory can be applied to interpret the results.
Recently, a new analytical interpretation method was presented for core flooding
experiments under constant differential pressure. This new interpretation method is based
on an analytical solution published recently. The analytical nature improves the accuracy
of the interpretation. It eliminates the need of using numerical differentiation, and
therefore reduces the overall numerical error. Under constant differential pressure, the
new theory determines the flow rate as a function of time, which is used to interpret core
flooding experiments under constant differential pressure conditions. In this paper, we
will first briefly review the interpretation method. Next, unsteady state core flooding
experiments are conducted under constant differential pressure conditions; the results
from different interpretations are compared and analyzed in terms of accuracy and
efficiency.

INTRODUCTION
Core flooding experiments are essential to Special Core Analysis (SCAL). These
experiments provide valuable information such as relative permeabilities and recovery
factors. Interpretation of the experimental results is generally based on fractional flow
theory of an immiscible fluid flow system. Combining Darcys Law and a material
balance equation for each phase, the general fractional flow model can be written as

( )

( )

],

(1)

where is the phase saturation; means the water phase; Pc is capillary pressure; uT is
the total volumetric flux; the water fractional flow function is given by =
( + ); where the water mobility is = and the oil mobility =
, = + . In analytical solutions to fractional flow theory, capillary
pressure is ignored, i.e. the right hand side of Eq. (1) equals zero. The classical fractional
flow theory by Buckley and Leverett [1] assumes that the total volumetric flux is

SCA2015-039

2/6

constant. A similar theory was recently presented by Johansen and James [2] assuming
that the inlet and outlet pressures are constant rather than the flow rate. In the latter
solution, the total flow rate is a function of time, which is determined analytically.
Johnson et al. [3] first presented an analytical method (JBN method) to determine relative
permeabilities based on the fractional flow theory developed by Buckley and Leverett [1]
and Welge [4]. This interpretation method is based on the assumption that the total flow
rate is constant during the core flooding experiments. However, core flooding
experiments are sometimes conducted under constant differential pressure conditions.
One of the main inaccuracies using JBN method is from the numerical calculation of the
differentiation of the cumulative water injection terms. Jones and Roszelle [5] provided a
similar method to JBN method. In the Jones-Roszelle method, the reciprocal relative
mobility at the outlet is determined by the intercepts of the tangent of the effective
viscosity curves. Since the tangent is determined from the discrete effective viscosity data
points, this introduces numerical instability and inaccuracy. Recently, Cao et al. [6]
presented a new analytical interpretation method for core flooding experiments under
constant differential pressure conditions. This method applied the novel solution to
fraction flow theory under constant pressure boundaries by Johansen and James [2]).
Contrary to the assumption of constant total flow rate, the total flow rate is a function of
time under constant differential pressure conditions and it is determined analytically as
part of the generalized solution. In the formulation by Cao et al. [6] pressure and
saturation profile at each point in time are solved analytically, hence this new method
eliminates numerical differentiation and provides stable and accurate results.
In this paper, the new interpretation method for unsteady state core flooding experiments
under constant pressure boundaries is first briefly reviewed. Then, it is applied to
determine relative permeabilities from the core flooding experiments. The main purpose
of this paper is to verify the accuracy of the new interpretation method and compare it to
the JBN and Jones-Roszelle interpretation methods. Two cases are presented in this paper:
Case 1: high viscosity oil in a low permeability core; and Case 2: low viscosity oil in a
high permeability core. In both cases, the core flooding experiments are conducted under
constant pressure condition and interpreted by the new method and JBN and JonesRoszelle methods. The results show that the new method is more accurate and stable as a
consequence of eliminating numerical differentiation.

METHODOLOGY
The core flooding experiments used two sandstone cores, synthetic oil, filtered crude oil
(from Hibernia oil field offshore Newfoundland, Canada), and brine. The experiments
were performed at room temperature. The schematic of the core flooding experiment is
shown in Fig. 1.
The pressure differential on the core was held constant by running the Quizix 20K pump
at constant pressure delivery mode for the inlet and using a back pressure regulator (BPR)
at the outlet. A calibrated density meter was connected to the outlet of the core holder to

SCA2015-039

3/6

monitor the density change in real time in order to determine the fraction of each
producing phase.

Core holder

BPR

Oil
Brine

Separator

Fig. 1 Schematic of the core flooding experiment

The interpretation methodology for core flooding experiments under constant differential
pressure conditions was presented by Cao et al. [6]. During the core flooding
experiments, differential pressure, total fluid density at the outlet and cumulative
production were recorded in time. Basic information of the core dimension and fluid
properties (viscosity and density) are shown in Table 1. The methodology involves
determining the pressure and saturation profiles analytically along the core length at the
time of water breakthrough analytically using the method in Johansen and James [2]. The
saturation profile is calculated when each saturation point arrives at outlet. In this process,
the total flow rate as a function of time can also be determined analytically and injection
accumulation is easily integrated. The new method for relative permeability calculation in
this paper can be summarized in the steps described below with the detailed procedures
and equations described by Cao et al. [6].
1. Calculate the saturation at the outlet at selected times post water breakthrough:
calculate the water fractional flow function for each saturation point.
2. Determine the saturation and pressure profiles along the length of the core at the
time of water breakthrough analytically using Johansen and James [2].
3. Calculate the spatial pressure derivative (pressure gradient) throughout the core.
4. Calculate the relative permeability.
5. Calculate the absolute permeability with brine when the single phase flow is
steady state. End points of the relative permeability curves are determined by
continuing the steady state two phase flow until connate water and residual oil
saturation is reached.

SCA2015-039

4/6

RESULTS
In both cases (Case 1: high viscosity oil and low permeability core; and Case 2: low
viscosity oil and high permeability core), the core flooding experimental results are
interpreted using the JBN method, the Jones-Roszelle method and the new method. The
experimental data is shown in Table 1. The lack of data regularity in both cases stems
from the fact that the core floods are performed under unsteady state conditions.
Case 1. High viscosity oil and low permeability core
In Case 1, a viscous crude oil (32 cP) and a tight sandstone (10 mD) core were used. As
shown in Fig. 2, the relative permeability results calculated from different methods are
compared. It is shown in Fig. 2. that the water relative permeability results from the new
method have a good agreement with most of the data points with the results from JonesRoszelle method, where oil relative permeability results are perfectly matched in the
water saturation range from 0.68 to 0.80. Water relative permeability results from JBN
methods shows minor deviation from the new method in the water saturation ranged from
0.20 to 0.58. The results from both the JBN and Jones-Roszelle methods show a nonmonotonic derivative variation deviating from the new method curve ranged from water
saturation of 0.55 to 0.75, while the results from the new method gives a smooth relative
permeability curve.
Table 1 Experimental data
Parameter
Oil viscosity (cP)
Brine viscosity (cP)
Core sample length (cm)
Core cross-section area (cm2)
Core sample porosity
Absolute oil permeability (mD)
Connate water saturation
Differential Pressure (psi)

Symbols

Swc

Case 1
Value
32.0
1.0
30.48
10.01
0.24
10
0.35
243.0

Case 2
Value
3.0
1.0
30.48
10.01
0.24
876
0.29
1.0

Case 2 Low viscosity oil in a high permeability core

The comparison of the results of Case 2 using a low viscosity oil (3 cP) in a high
permeability core (876 mD) from the three methods is shown in Fig. 3. It is shown that
the relative permeability curve obtained from the new method still shows a smooth
continuous curve. Similar to Case 1, the results from both the JBN and Jones-Roszelle
methods show a non-monotonic derivative variation deviating from the new method
curve in the water saturation range of 0.45 to 0.55, where the non-monotonic derivative
variation is more significant compared to Case 1. During the calculation using the JonesRoszelle method, the effective viscosity calculated for Case 2 is scattered as shown in Fig.
4 which makes it difficult to calculate the tangents on the curve numerically and leads the
instability of the relative permeability curve. It is relatively easier to calculate the
tangents for the smooth effective viscosity curve in Case 1 compared with Case 2 as

SCA2015-039

5/6

shown in Fig. 4, but the calculation of the tangents using discrete points numerically still
introduces errors.
1.00
New method_Kro

Relative permeability

0.90

New method_Krw

0.80

JBN_Kro

0.70

JBN_Krw
Jones and Roszelle_Kro

0.60

Jones and Roszelle_Krw

0.50

Fitted new method_Kro

Fitted new method_Krw

0.40
0.30
0.20
0.10
0.00
0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

Water saturation

Fig. 2 Relative permeability for Case 1: 32 cP oil and 10 mD rock

1.00
New method_kro

Relative permeability

0.90

New method_krw

0.80

JBN_kro

0.70

JBN_krw
Jones & Roszelle_kro

0.60

Jones & Roszelle_krw

0.50

Fitted new method_Kro

0.40

Fitted new method_Krw

0.30
0.20
0.10
0.00
0.20

0.30

0.40

0.50

0.60

0.70

0.80

Water saturation

Fig. 3 Relative permeability for Case 2: 3.0 cP oil and 876 mD rock

0.90

6/6

Effective viscosity for Case 1

350

Case 1

300

3.5

Case 2

3.0

250

2.5

200

2.0

150

1.5

100

1.0

50

0.5

0.0
0.0

0.5

1.0

1.5

2.0

2.5

3.0

Effective visccosity for Case 2

SCA2015-039

3.5

Pore volume injected

Fig. 4 Effective viscosity for the Jones-Roszelle method

CONCLUSIONS
As discussed in the comparison of the results using different methods, both JBN and
Jones-Roszelle methods give instable relative permeability results shown as nonmonotonic derivative variation on the curves due to numerical differentiation. By
eliminating the numerical differentiation calculation, relative permeability results from
the new method yield smooth curves for both cases. We show that the new method
provides a robust and stable methodology to interpret unsteady state core flooding data
for determining relative permeability. The superiority of the new method is that the
numerical differentiation is eliminated and the overall numerical errors are reduced.

ACKOWLEDGEMENTS
The authors thank the Hibernia Management and Development Company (HMDC),
Chevron Canada, the Natural Sciences and Engineering Research Council of Canada
(NSERC), and the Research and Development Corporation (RDC) for their support
without which this work could not have been performed.

REFERENCES
1. Buckley, S. E. and Leverett, M. C.: Mechanism of Fluid Displacement in Sands,
Trans. AIME 146 (1942), 107-116.
2. Johansen, T. E. and James, L. A.: Fractional Flow Analysis for Multi-Component
Problems with Constant Pressure Boundaries, To appear in Journal of Engineering
Mathematics, Springer Verlag. 2015.
3. Johnson, E. F., Bossler, D. P. and Naumann, V. O.: Calculation of Relative
Permeability from Displacement Experiments, Trans. AIME 216 (1959), 370-372.
4. Welge, H.G.: A Simplified Method for Computing Oil Recovery by Gas or Water
Drive, Petrol. Trans AIME 195(1952), 91.
5. Jones, S.C. and Roszelle, W.O.: Graphical Techniques for Determining Relative
Permeability From Displacement Experiments. Society of Petroleum Engineers of
AIME. 1978.
6. Cao, J., James, L.A. and Johansen, T.E.: Determination of Two Phase Relative
Permeability from Core Floods with Constant Pressure Boundaries. Society of Core
Analysis Symposium, Avignon, France. 2014.

SCA2015-040

ANALYZING THE OPEN FACE CONDITION AND NONEQUILIBRIUM EFFECTS IN CO- AND COUNTERCURRENT SPONTANEOUS IMBIBITION WITH
EXPERIMENTS AND NUMERICAL SIMULATION
Rasoul Arabjamaloei1, Douglas W. Ruth1, Derrick Green2, Dragan Veselinovic2
1
Centre for Engineering Professional Practice and Engineering Education, University of
Manitoba, Canada
2
Green Imaging Technologies INC., Fredericton, NB, Canada
This paper was prepared for presentation at the International Symposium of the Society
of Core Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August,
2015

ABSTRACT
In this work a set of co- and counter-current spontaneous imbibition experiments were
conducted on a sandstone core sample using Decane and D2O as displacing/displaced
fluids. Using an Oxford Maran DRX-HF instrument, the saturation profiles of the fluids
inside the core were measured and recorded during the experiments. The capillary
pressure-saturation curves for the sample rock were obtained by MICP and centrifuge
tests. The accumulative production of non-wetting phase, determined by integrating the
obtained saturation profiles, shows a square root of time behavior while the whole
process was seen to be affected by non-equilibrium effects at the inlet. It was assumed
that all the production was from the upstream face in the counter-current case, and from
the downstream face in co-current test. The experiments also showed an inlet water
saturation of around 50%, which is in contradiction with the assumption of capillary
continuity at the core face. The results were analyzed using an explicit simulator to obtain
Corey-type relative permeability functions. It was found that the counter-current
experimental results showed a close match between the simulated and the experimental
results. Further, the relative permeability curves obtained by history matching of
counter-current imbibition test results were used to simulate and predict the results for cocurrent imbibition test.

INTRODUCTION
Both co-current and counter-current spontaneous imbibition are complicated two phase
processes which are not yet well understood. One of the main processes involved in oil
production from fractured reservoirs is spontaneous imbibition (SI) which is driven by
capillary forces [1]. Experimental work has been done to study SI processes and obtain
relative permeability and capillary pressure curves; however exact measurements of
saturation distribution profiles are not very common [1, 2, 3]. The SI process could be
modeled analytically if the proper sets of relative permeability curves are known for the
porous medium. However, it is not yet known with certainty if a unique set of relative

1/7

SCA2015-040

2/7

permeability curves can be implemented for the mathematical modeling of all the
processes happening in a porous medium. In the present work, experiments were run to
measure the saturation distribution profiles in the sample core during co-current and
counter-current spontaneous imbibition processes. The experimental data was then input
to an explicit numerical simulator to achieve relative permeability curves by history
matching. The core properties and the capillary pressure curves are taken from a previous
study [4].

THE EXPERIMENTAL STUDY

In this study two co- and counter-current SI tests were performed on a core sample of
tight rock. Table 1 shows the rock type and properties.
Table 1: Rock properties of the sample used for co-current and counter-current experiments

Rock type
Sandstone

Porosity
(%)
23

Permeability
(md)
2

Length
(cm)
2.5

Cross Section
(cm2)
9.61

Teflon tape was wrapped tightly around the cylindrical sample plug and then the plug
was sealed with heat shrink plastic tubes. Figure 1 shows how the core was prepared
before the experimental work and figure 2 shows a diagram of experiment set-up. A
constant volume D2O reservoir is always maintained at the top of the sample by adding
droplets. The length to diameter ratio of the sample might be small, that was due to the
limitations of accessibility to big range of samples.
The two ends of the plug were open in co-current experiments while one end of the
sample was sealed with liquid glue for counter-current test. The sample was fully
saturated with decane by using centrifuge before the start of the experiment. That leads to
zero initial water saturation which might affect the results of the experiment. To study the
effect of initial water saturation another broad study is needed. Deuterium oxide (D2O)
was used as the injected fluid because this substance is not seen by the NMR apparatus.
The NMR measurements were done using an Oxford Maran DRX-HF instrument at 30 C
and 2MHz frequency. Green Imaging Technologies (GIT) software was used to measure
the decane saturation distribution. By proper design of the core holder it was possible to
continue the experiment in the machine while the measurements were done.
The simulator uses 50 central grid blocks system with equal distances except for the
boundary node which is a zero width block. The saturation change at each grid is
calculated based on the flow rate difference across the block boundaries. The flow rates
are calculated by modified Darcys law and gravity is ignored. The saturation profiles, the
capillary pressure curves and rock and fluid properties were provided to the simulator as
inputs. The program uses those data to find the best Corey-type relative permeability
curves for the sample. The simulation was fully explicit and the upstream saturation of
each grid block was used to calculate the Corey-type relative permeability of the
corresponding block. Further details of the model may be found in [6].

SCA2015-040

History matches were achieved by minimizing the error between the experimental and
simulated saturation profiles. Neither production data nor pressure data were used for the
process of history matching.

DATA ANALYSIS
Figure 3 shows the counter-current SI experimental data and the best fit curves resulting
from the simulations. As observed in figures 3 and 4, there is a very good match between
the experimental and simulation results, especially for the production curve. However,
the measured saturation profiles might not show the well-known patterns expected for
this displacement process. It is seen that water imbibes through the core very fast and
compensate for the zero initial water saturation. It is also observed that the NMR readings
dont show a strong water front during the co-current process at the early times. But it is
seen that a front forms at latter times. The non-wetting phase production was calculated
from NMR outputs. The relative permeability curves found for counter-current imbibition
were next used to predict the results for the co-current imbibition test. Figure 5 shows the
calculated relative permeability curves.
The objective was to determine if the history matched relative permeability curves are
suitable for the co-current imbibition process. The simulator was used to produce and
compare the saturation distribution and production curves. Figures 6 and 7 show the
comparison of results.

DISCUSSIONS AND CONCLUSIONS

As observed in figures 6 and 7, a Corey-type relative permeability curve can be used to
numerically simulate the counter-current spontaneous imbibition. The predicted
saturation distribution curves do not perfectly match the experimental results; however
the predicted production has an excellent match with the experimental results. Because
the NMR measurements take 1-2 minutes each time, this is a potential source of error in
the saturation profiles and might cause the difference between experimental and
simulation results. It is also observed in the experimental results that the inlet face
saturation is changing with time and is not a unique value. This shows the role of nonequilibrium effects in a spontaneous imbibition process. However, at later times the
saturation values stabilize and agree approximately with results predicted by the method
of Arabjamaloei et al [7].
The co-current experimental production results did not show as close a match as for the
counter-current case. This problem is likely related to the fact that in these experiments
the two faces were open and production was happening at both faces as it was observed
in the experiments; in the present experiments only the total production could be
determined. Parts of the saturation curves show reasonable agreement; however, the
upstream face saturation is not well predicted. This is further evidence that the face
condition during spontaneous imbibition is poorly understood. It should be noticed that

3/7

SCA2015-040

the produced decane floating on the top of the water reservoir in counter-current
experiment or collected at the bottom in co-current experiment might affect the readings,
but the effect was reduced by taking the decane out of the set-up every once in a while.
The results are encouraging since they show that, at least for one case, a single set of
relative permeability curves based on counter-current test can be used to approximately
predict the co-current test results.

ACKNOWLEGEMENTS
This work was supported in part by a grant from the Natural Sciences and Engineering
Research Council of Canada. The simulator used in this study was PORLAB and is used
by permission of D&B Ruth Enterprises Inc.

REFERENCES
1. Mason, G. and N.R. Morrow, Developments in Spontaneous Imbibition and Possible
Future Work, Journal of Petroleum Science and Engineering, (2013).
2. Baldwin, B.A., Spinler, B.A., 2002. In situ saturation development during
spontaneous imbibition. Journal of Petroleum Science and. Engineering, 35 (1-2), 23
32.
3. Li, Y., Ruth, D., Mason, G., and Morrow, N.R., 2004, Pressures acting in countercurrent spontaneous imbibition, Proceedings of the 8th International Symposium on
Reservoir Wettability, Houston, TX, May 17-18.
4. Mason, G., Ferno, A., Morrow, N.R., Ruth, D.W., 2012, Spontaneous CounterCurrent Imbibition Outward from a Hemi-Spherical Depression, Journal of
Petroleum Science and. Engineering, 90-91 (July), pp.131-138.
5. Babak Salimifard, Douglas Ruth, Derrick Green, Dragan Veselinovic, Developing a
Model To Estimate Permeability from Other Petrophysical Data, SCA2014-063,
International Symposium of the Society of Core Analysts, (2014), Avignon, France, 811 September.
6. Ruth, D.W., Mason, G., Fern, M.A., Haugen, A., Morrow, N.R., Arabjamaloei, R.,
Numerical Simulation of Combined Co-Current/Counter-Current Spontaneous
Imbibition, International Symposium of the Society of Core Analysts, (2015),
St.Johns, Newfoundland and Labrador, Canada, 16-21 August.
7. Arabjamaloei, R. and Shadizadeh, S. R., A New Approach for Specifying Imbibition
Face Boundary Condition in Countercurrent Spontaneous Imbibition, Petroleum
Science and Technology, (2010), 28(18): 1855-1862.

4/7

SCA2015-040

5/7

Figure 1: A: first step, the plugs are wrapped with Teflon tape, B: at the second step the cores are fitted
inside a plastic tube

Figure 2: Diagram of the experimental set-up

1
0.8
0.6

Sw
0.4
0.2
0
0

0.5

1.5

2.5

7.5min
21min
64min
85min
113min
175min
395min
815min
7.5
21
64
85
113
175
395
815

Distance (Cm)
Figure 3: Comparison of the simulated (dash lines) and experimental (solid lines) saturation profiles for the
counter-current spontaneous imbibition test

SCA2015-040

6/7

2
1.5

Production
(ml)

Experiment
Simulation

0.5
0
0

200

400

600

800

1000

Time (min)
Figure 4: Comparison of non-wetting phase production data of experiments and simulation for countercurrent test
1
0.8

Injected phase

0.6

Relative
permeability 0.4

Displaced phase

0.2
0
0

0.2

0.4

0.6

0.8

Water saturation
Figure 5: Relative permeability curves obtained by history matching the Counter-current imbibition test
results. These relative permeability curves were also used to simulate non-wetting phase production of the
co-current imbibition test.

1
0.8
0.6

Sw
0.4
0.2
0
0

0.5

1.5

2.5

1
3.5Min
9.5Min
23.5Min
44Min
71Min
93Min
123Min
160Min
1
3.5
9.5
23.5
44
71
93
123
160

Distance (Cm)
Figure 6: Comparison of the simulated (dash lines) and experimental (solid lines) saturation profiles for cocurrent spontaneous imbibition test

SCA2015-040

7/7

4
3

Production
2
(ml)

Prediction

Experiment

0
0

50

100

150

200

Time (min)
Figure 7: Comparison of non-wetting phase production data of experiments and simulation for co-current
test

SCA2015-041

1/6

APPLICATION OF COMPRESSED SENSING MRI TO

LABORATORY CORE-FLOODS
N.P. Ramskill, I. Bush, A.J. Sederman, M.D. Mantle, L.F. Gladden
Department of Chemical Engineering and Biotechnology, University of Cambridge, New
Museums Site, Pembroke Street, Cambridge, Cambridgeshire, CB2 3RA.
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 September 2015

ABSTRACT
Imaging of fluid distributions is essential to enable the unambiguous interpretation of
core flooding data. In the present study, a rapid and robust magnetic resonance imaging
(MRI) approach to provide 3D images of the fluid saturation in rock core samples during
laboratory core floods has been demonstrated. MRI has been widely used to image fluid
saturation in rock cores; however, the conventional acquisition strategies are typically too
slow to capture the dynamic nature of the displacement processes that are of interest.
Using Compressed Sensing (CS), it is possible to reconstruct a near-perfect image from
significantly fewer measurements than was previously thought necessary, and this can
result in significant reductions in the image acquisition times. Using a CS-MRI approach,
3D images of the fluid saturation in the rock core have been acquired in minutes as
opposed to hours, as is the case with the conventional methods.
As a proof-of-principle, the CS-MRI technique has been applied to image the residual
water saturation in the rock during a water-water displacement core flood. The
enhancement in the temporal resolution that has been achieved using the CS-MRI
approach will enable dynamic transport processes pertinent to laboratory core floods to
be investigated on a time-scale that, until now, has not been possible.

INTRODUCTION
Laboratory-scale displacements in rock core-plugs (core floods) are widely used to
develop the understanding of oil recovery mechanisms [1,2]. Magnetic resonance
imaging (MRI) and X-ray tomography (CT) are the most widely used techniques for
imaging in situ core flood fluid distributions of which both can non-destructively image
multiphase fluid systems in porous media [3-6]. CT images bulk densities and effective
atomic numbers, distinguishing multicomponent systems based on atomic density
differences. While this enables detection of both the rock matrix and imbibed fluids, this
presents a challenge in providing contrast between fluids within the pore space, i.e. oil
and brine, which, without the addition of dopants, the difference in atomic densities is
small. In contrast, MRI has the advantage that there are numerous contrast mechanisms
that can be implemented to provide contrast between different chemical species, namely:
chemical selectivity of NMR-active nuclei detected (e.g. 1H, 23Na), spectroscopic

SCA2015-041

2/6

chemical shift, relaxation time (T1 and T2) weighting, diffusivity contrast and use of
lower gyromagnetic species, such as D2O water [7].
However, conventional 3D MRI suffers low temporal resolution and hence, the dynamic
nature of core flood displacements cannot be effectively monitored. In conventional MRI,
data are uniformly sampled at the Nyquist rate (at a rate of at least twice the frequency of
the highest frequency components present in the signal of interest). This is determined by
the desired field of view (FOV) and image resolution. Therefore, when multi-dimensional
and high spatial resolution images are sought, this can result in long acquisition times. It
has been shown that using Compressed Sensing (CS), a near-perfect reconstruction can
be obtained from a number of measurements sampled below the Nyquist rate. Therefore,
by applying CS to MRI, reducing the number of data points sampled would lead to a
reduction in the image acquisition time [8].
CS has previously been applied to pure phase-encoding techniques for the study of
porous materials [9,10]. Due to their robustness in the presence of paramagnetic
impurities and magnetic susceptibility gradients, the pure phase encode methods have
proven to be suitable for providing quantitative measurements of the fluid content in
particularly challenging systems, such as rock cores. However, even with under
sampling, these techniques are too slow for studying dynamic displacement processes,
particularly when 3D images are required. Depending on the system under investigation
and the information sought, the choice of MRI pulse sequence is a trade-off between how
quantitative it is and its achievable temporal resolution. In this work, Rapid Acquisition
with Relaxation Enhancement (RARE) [11] with CS has been used to image the residual
fluid saturation during a laboratory core flood.

PROCEDURE
In the present study, a Bentheimer sandstone plug, 38 mm in diameter and 68 mm in
length has been used. The pore volume of the rock was determined to be 18 ml
corresponding to a porosity of ~24 %. The rock was initially saturated with deionised
water, which was displaced by a gadolinium chloride (GdCl3) doped-water solution
during the core flood. The concentration of the GdCl3 (aq) solution (~8 mM) was chosen
to ensure that the transverse relaxation time (T2) was sufficiently short that it could not be
detected in the images. The sample was held in-place by an Aflas sleeve within a PEEK
rock core holder (ErgoTech, Conwy, UK). A constant confining pressure was applied to
the outside of the Aflas sleeve by per-fluorinated oil (Fluorinert FC-43) using a Gilson
307 (Gilson Inc., USA) HPLC pump maintained at 25025 psig by a back pressure
regulator (Idex Health and Science, USA). The injectant was pumped through the rock at
a flow rate of 0.025 ml min-1 using a Quizix QX1500 (Chandler Engineering, USA) dualsyringe pump. The corresponding interstitial pore velocity was ~0.4 ft day-1.
The MRI experiments were carried out on a 2 T (85 MHz for 1H) horizontal bore magnet
controlled by a Bruker AV spectrometer. Prior to the start of the core flood, a fullysampled 3D image was acquired as a reference and has been used for CS simulations.

SCA2015-041

3/6

During the core flood, under-sampled 3D images were acquired. For the acquisition of
both fully and under-sampled 3D images, the RARE pulse sequence has been used. The
field of view (FOV) was 80 mm 50 mm 50 mm in the z, x and y directions
respectively. Correspondingly, for a data matrix size of 256 128 128 pixels in the
read (z) direction and first (x) and second (y) phase encoding directions, the image
resolution is 0.31 0.39 0.39 mm pixel-1. For the fully-sampled 3D RARE images, the
acquisition time was 2 hours and 9 minutes whereas for the 3D CS-RARE images, with
25 % sampling the acquisition time was approximately 16 minutes.

RESULTS AND DISCUSSION

Three-dimensional Compressed Sensing Simulations
In order to demonstrate the performance of CS, a qualitative comparison has been made
between an image reconstructed from the fully-sampled data set and that from a CS
reconstruction. For details on the CS reconstruction method used herein, the reader is
directed to [12]. A more detailed description of the CS acquisitions and reconstructions
employed in the present study will be reported in a future publication. Figure 1 a) shows
a 2D xz slice taken from the reconstruction of the fully-sampled 3D reference image. For
the CS simulation, a 30 % sampled data set was created by replacing 70 % of the data
points, from the fully-sampled data, with zeros. Figure 1 b) and c) show 2D xz slices
taken from the 3D images obtained from the zero-filled Fourier transform and CS
reconstructions of the simulated under-sampled data, respectively.
80

b)

1
20
0.5

Lz / mm

40
1
20
0.5
0
0 10 20 30 40 50
Lx / mm

saturation / a.u.

Lz / mm

40

0
0 10 20 30 40 50
Lx / mm

60

60

saturation / a.u.

Lz / mm

60

80

c)

40
1
20
0.5
0
0 10 20 30 40 50
Lx / mm

saturation / a.u.

80

a)

Figure 1. 2D xz slices taken from the centre of 3D images of the water-saturated Bentheimer rock prior to
the core flood obtained by a) Fourier transform of the fully-sampled data, b) Fourier transform of the 30 %
sampled data and c) CS reconstruction of the 30 % sampled data.

It can be seen that, due to the aliasing artefacts that arise from under-sampling, the
contrast in the image the reconstruction from the zero-filled Fourier transform (Figure 1
b)) is somewhat reduced with respect to the fully-sampled case (Figure 1 a)). However,
due to the bias towards a sparse solution in the CS reconstruction (Figure 1 c)), an
image with greater contrast than the zero-filled Fourier transform is obtained and is,
visually, much closer to the fully-sampled image. A quantitative assessment of the CS

SCA2015-041

4/6

methodology that has been implemented in the present study will be discussed in detail in
a future publication. However, the qualitative comparison shown in Figure 1 has
demonstrated that a near-perfect image has been recovered from data sampled
significantly below the Nyquist rate, and would therefore allow for significant
improvements in the temporal resolution for the acquisition of 3D images.
Application of 3D CS-MRI Imaging to a Water-Water displacement core flood
The fully and under-sampled 3D images of the initially water-saturated rock cores, prior
to the start of the core flood, are shown in Figure 2 a) and b), respectively.
a)

b)

Figure 2. 3D images of the initially water-saturated Bentheimer rock from a) fully-sampled and b) 25 %
sampled data set for which the acquisition times were 2 hours 9 minutes and 16 minutes, respectively.

By using CS combined with RARE, an enhancement in the temporal resolution by a

factor of 8 has been achieved. It should be noted that the reduction in acquisition times
relative to a standard spin-echo sequence or purely phase-encoded would be significantly
greater. For instance, a 3D image of the same resolution as those presented herein
acquired with a standard spin-echo sequence would take around 60 hours to acquire.
Figure 3 shows the 3D CS images of the residual saturation of the water that was initially
present in the rock at various times following injection of the injectant solution. The flow
of the injectant is from right-to-left, i.e. in the z direction.
a)

b)

SCA2015-041

5/6

c)

d)

e)

f)

Figure 3. 3D CS images of the residual water in the Bentheimer rock core after a) 82, b) 180, c) 279, d)
377, e) 476 and f) 574 minutes of flowing a ~8 mM GdCl3 (aq) solution at 0.025 ml min-1. The flow of the
injectant is from right-to-left, in the z direction. The concentration of the GdCl3 (aq) solution was chosen
as such to ensure that the transverse relaxation time (T2) was sufficiently short thus it was invisible in the
images. From the series of images, the invasion front throughout the core flood is evident.

From the series of images presented in Figure 3, the progress of the displacement front
through the rock during the core flood can be clearly observed.

CONCLUSIONS
In this paper, a CS MRI method for 3D imaging of the fluid saturation in rocks during a
laboratory core flood has been demonstrated. Firstly, it has been shown that, visually, a
near-perfect image can be reconstructed from data sampled in violation of the Nyquist
criteria, thus allowing for significant reductions in the acquisition times. Using CS with
RARE, an eight-fold improvement in the temporal resolution has been achieved, relative
to the fully-sampled RARE acquisition. However greater time savings are possible over
more standard pulse sequences. Secondly, the CS-RARE technique has been applied to
image the residual water saturation during a water-water displacement core flood.
The enhancement in the temporal resolution obtained with CS will unlock the potential to
observe phenomena during a core flood on a time scale that would not be possible using
conventional MRI protocols. It is therefore the aim of future work that this method will
be applied to investigate pertinent issues such as ganglion dynamics [6] and the influence
of capillary end effects [13] during an oil recovery core flood.

SCA2015-041

6/6

REFERENCES
1. Mitchell, J., Staniland, J., Chassagne, R. and Fordham, E.J., 2012a. Quantitative In
Situ Enhanced Oil Recovery Monitoring Using Nuclear Magnetic Resonance.
Transport in Porous Media, 94, pp. 683-706.
2. Mitchell, J., Wilson, A., Howe, A., Clarke, A., Fordham, E.J., Edwards, J., Faber, R.
and Bouwmeester, R., 2012c. Magnetic Resonance Imaging of Chemical EOR in
Core to Complement Field Pilot Study. SCA 2012-30. In Proc: International
Symposium of the Society of Core Analysts. Aberdeen, Scotland, UK, 27-30 August
2012. Society of Core Analysts.
3. Vinegar, H.J., 1986. X-ray CT and NMR Imaging of Rocks. Journal of Petroleum
Technology, (March), pp. 257-259.
4. Yuechao, Z., Yongchen, S., Yu, L., Lanlan, J. and Ningjun, Z., 2011. Visualization of
CO2 and oil immiscible and miscible flow processes in porous media using NMR
micro-imaging. Petroleum Science. 8(2), pp. 183-193.
5. Fern, M. A. Ersland. G., Haugen, , Johannesen, E., Graue, A., Stevens, J. Howard,
J., 2007. Impacts from fractures on oil-recovery mechanisms in carbonate rocks at
oil-wet and water-wet conditions visualizing fluid flow across fractures with MRI.
In Proc: International Oil Conference and Exhibition. Veracruz, Mexico, 27-30 June
2007. Society of Petroleum Engineers.
6. Youssef, S., Rosenberg, E., Deschamps, H., Oughanem, R., Maire, E. and Mokso, R.,
2014. Oil ganglia dynamics in natural porous media during surfactant flooding
captured by ultra-fast X-ray microtomography. SCA paper 2014-023. In Proc:
International Symposium of the Society of Core Analysts. Avignon, France, 11-18
September 2014. Society of Core Analysts.
7. Mitchell, J., Chandrasekera, T.C., Holland, D.J., Gladden, L.F. and Fordham, E.J.,
2013a. Magnetic resonance imaging in petrophysical core analysis. Physics Reports,
525, pp. 165-225.
8. Holland, D.J. and Gladden, L.F., 2014. Less is More: How Compressed Sensing is
Transforming Metrology in Chemistry. Agewandte Chemie Int. Ed. 53, 2-13.
9. Xiao, D., Balcom., B.J., 2012. Two-dimensional T2 distribution mapping in rock core
plugs with optimal k-space sampling. Journal of Magnetic Resonance, 220, pp 70-78.
10. Xiao, D., Balcom., B.J., 2014. k-t Acceleration in pure phase encode MRI to monitor
dynamic flooding processes in rock core plugs. Journal of Magnetic Resonance, 243,
pp 114-121.
11. Hennig, J., Nauerth, A., Friedburg, H., 1986. RARE imaging: A Fast Imaging Method
for Clinical MR. Magnetic Resonance in Medicine, 3, pp 823-833.
12. Benning. M., Gladden, L.F., Holland, D.J., Schnlieb, C.B., Valkonen, T., 2014.
Phase reconstruction from velocity-encoded MRI measurements A survey of
sparsity-promoting variational approaches. Journal of Magnetic Resonance, 238, pp
26-43
13. Huang, D.D., Honarpour, M.M., 1996. Capillary end effects in coreflood calculations.
SCA 1996-34. In Proc: International Symposium of the Society of Core Analysts.
Montpellier, France, 8-10 September 1996. Society of Core Analysts.

SCA2015-042

1/7

STUDY OF ROCK PORE SPACE BY COMBINATION OF

DIRECT AND INDIRECT TECHNIQUES
Aleksandr Denisenko, Ivan Yakimchuk, and Boris Sharchilev
Schlumberger
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
During the last few decades, Nuclear Magnetic Resonance (NMR) has become known in
the petrophysics community as a convenient and efficient technique for studying pore
space structure. NMR is an indirect method of pore size evaluation based on a
proportionality between relaxation times and surface-to-volume ratio for each pore. The
coefficient of proportionality (surface relaxivity) reflects the ability of mineral pore
surfaces to increase the relaxation rates. It is common to use a single value of relaxivity
for each core, which assumes that the pore surface properties within a small sample are
sufficiently homogeneous [1]. However, in some cases the sample might contain pores
with different surface properties that result in a range of relaxivity values characterizing
the rock. Induced magnetic susceptibility arising from different sources, for instance
metal ions, paramagnetic (Cu2+, Mn2+, Fe2+, Fe3+) or, in a worst-case scenario,
ferromagnetic metallic iron (Fe), leads to interpretation uncertainty of T2 spectra due to
the influence of the internal gradient effects on the transverse relaxation rate. Fortunately,
constructed T1T2 2D dependences help to monitor the discrepancies in T1 and T2 spectra,
which are affected by surface property changes and magnetic susceptibility contrast
within the sample structure.
In this study, we implemented several proven techniques designed to evaluate pore
structure directly. This approach allows for checking consistency of pore size distribution
obtained by NMR T1 at each pore scale. The results of X-ray microtomography (XmCT)
imaging validated the large pore scale or slow T1-decays and Scanning Electron
Microscopy (SEM) results confirmed the small pores described by the fast NMR
responses. The repeatability of all calculated distributions indicates similar abilities of all
pore surfaces with the sample to relax the hydrogen nuclei or close relaxivity values
responsible for T1-relaxation in large and small pores for the observed sandstone rock.

INTRODUCTION
One of the fundamental pieces of information derived from laboratory core analysis is the
pore size distribution. The classical and widespread methods used for pore size
distribution are thin-section petrography and mercury intrusion porosimetry. Recently, a
number of alternative technically complicated approaches have been proposed by the
scientific community, including nuclear magnetic resonance (NMR), X-ray
microtomography (XmCT), and scanning electron microscopy (SEM). Each of these

SCA2015-042

2/7

methods, as with any experimental technique, has its advantages and drawbacks. For
example, the modern nano-CT devices provide resolution down to 50 nm, but the small
field-of-view for such scanners raises concerns whether the imaged volume is
representative of the overall sample. In contrast to XmCT, laboratory NMR machines
detect hydrogen nuclei of fluid molecules constrained in pores of any geometry.
Therefore, NMR is an indirect method for pore size evaluation based on a nearly linear
relation between relaxation times and surface-to-volume ratio of a pore NMR distribution
covers the broadest range of pore sizes, up to 5 orders of magnitude, far greater than any
one direct technique.
Every pore-scale measuring method has its own specific limitation to the size range each
can detect. A possible solution that eliminates the limitations of individual method is to
perform the study by using a combination of methods instead. The key questions in such
a combination of techniques are the analysis accuracy and defining the roles for each
method. Research scientists have to understand the interpretation workflow and justify
the applicability of each method for the particular case. In other words, the combination
of different methods should lead to collaborative and consistent results instead of loosely
combined data fragments.

METHODS AND APPARATUS

NMR All of the NMR measurements were carried out using an Oxford Instruments lowfield spectrometer, which includes a MARANi-Pharmasence magnet block equipped with
a 2222 mm probe, and DRX-HF electronic control system. The resonance frequency of
20.6 MHz corresponds to hydrogen nuclei spin precession in a ~ 0.5 T magnetic field.
The T2 and T1 relaxation decays were recorded by ordinary Carr-Purcell-Meiboom-Gill
(CPMG) and inversion-recovery with echo detection. We used an echo time of 100 s
and numbers of echoes sufficient to complete a decay, 48 logarithmically spaced delay
times over the range 10-4 to 101 s, and a relaxation delay of 10 s. The measurement
temperature was 34C. Methodically, all NMR procedures and consequent collation with
XmCT data were performed in accordance with our previous work [2].
XmCT We used two laboratory table-top micro-CT scanners in this study [3], [4]. The
Using a 100-keV radiation energy for both systems a pixel size of 2.2 m was obtained
for the Bruker SkyScan 1172 and 0.5 m for the ZEISS XRadia Versa XRM 500 on
studied 8 mm diameter plug without contrast agents. The reconstructed 3D image of a
core sample was segmented into two classes of objects: pores and minerals. The most
intuitively obvious approach for individual pore analysis consists of separating a whole
pore space into a set of individual pores and analyze each pore separately. General
algorithms for pore separation are based on watershed techniques [2]. The equivalent
spherical diameter is one of the most commonly used quantities for estimating the
effective size of the object body. Finally, the histogram of the size values can be
constructed.

SCA2015-042

3/7

An alternative method for calculation of the object size distribution is based on sphere
fitting inside of the 3D structure [2]. The approach is also applicable for 2D images by
using circles instead of spheres. In our opinion, this algorithm more accurately describes
the pore space and more correctly fits with the physical phenomena in support of the
NMR method, as the magnetic spin relaxation dominated by the nuclei diffusion and
subsequent collisions with the grain surfaces. Therefore, the limits of molecules spreads
correspond to spherical areas at any time and the sizes of spheres are constrained with the
pore body ones.
SEM In comparison with XmCT the SEM method provides much higher image
resolution (up to 1 nm per pixel), but only the surface of the sample is investigated. For
accurate studying of pore space structure, a special sample preparation procedure should
be performed. This special procedure includes epoxy filling, mechanical cutting,
multistage polishing and final coating of the surface by a conductive material. Using an
image-stitching option makes it possible to cover rather large surfaces limited only by the
SEM chamber and the sample holder. For example, scanning an 88-mm area with 1-nm
resolution is essentially feasible, but would produce an enormous image of 8 million 8
million pixels (>50 TB). In the work being reported in this paper, a 100-nm resolution
was used (image size = ~80,000 80,000; file size = ~8 GB).
Focused Ion Beam coupled with SEM The focused ion beam inside SEM (FIB-SEM)
allows for obtaining a stack of 2D images of the samples internal structure of some
subsurface volume by ion etching of the surface slice-by-slice [5]. Thus, this method
produces 3D images with resolution near that of the SEM. Unfortunately, the field of
view is rather small (10 20 m).
Sample preparation The sample for study was cleaned and dried sandstone rock of
West Siberia province, 8 mm in diameter and 10 mm long. The plug was saturated with
20 g/l NaCl solution, and placed in hermetically sealed cell for NMR measurements.

EXPERIMENTS AND DISCUSSION

Before consolidation the different pore structure evaluation methods, we analyzed the
sample size representativeness and the influence of core preparation techniques on rock
integrity and invariance. The comparison of NMR evaluation result on 22-mm and 8-mm
samples cut from a single core gave rise to this investigation. The dissimilarity in short
relaxation times of T1T2 maps for different cylinder diameters indicated alteration of
microporosity structure compared with the consistency and stability of large pores. Plugs
drilling and especially a cooling fluid could cause the effect as a tap water was used
during the coring. The swelling of clay minerals and consequent dispersion of clay
particles along with migration of disrupted small solids at the cylinder near-surface area
alter rock properties and consequently, pore size distribution. Using the SEM, we
analyzed the structure and elemental alteration near the edges and in the central part of
the 22-mm diameter sandstone core (Figure 1). As is clearly shown, the internal core
structure is complicated, with rough and sharp points protruding from it. Many clay and

SCA2015-042

4/7

mud particles are observed near the edges of core, although rounded and clean grains
prevail in the center.
Therefore, many factors determine the representativeness of cores and we had to be
confident with our investigation, even with very homogeneous rock. For that reason, we
performed all of the experiments on single-core cylinders or their fragments.
As a result, the study of void space structure was performed on 8-mm diameter by 10 mm
in length cylindrical samples. This core size allows for performing experiments on the
same sample using the NMR, XmCT and 2D SEM techniques. The FIB-SEM as well as
Xradia XmCT scans were conducted on a small part of the rock cylinder. The collation of
different methods of observing pore structure was studied on well-sorted sandstone rock
samples. Initially, we adjusted the T1 distribution with two XmCT spectra in terms of
relaxation activity value, which was a tuning parameter of NMR and XmCT spectra
coincidence.

Figure1: SEM images of edge (a) and central parts (b) of granular sandstone core. Note the visible
alteration and contamination of pores near the core edges.

Regardless of the applied model, the relaxation activity value mathematically represents
the proportionality coefficient between pore sizes and relaxation times. We adopted the
spherical pore model assumption: r = 3T1, where r is a pore radius, a relaxivity; as
the inscribed sphere method used to calculate the pore sizes using XmCT data. It is
essential that the coefficient of proportionality between relaxation times and pore
diameters, a complicated value, be influenced by pore surface magnetism and wettability,
as well as pore shape. Therefore, it is more correct to name the coefficient as being
pseudorelaxation activity or cumulative parameter responsible for surface relaxation
properties defined by mineralogical composition of rock and a value of specific surface.
The fitting of 1 with XmCT data (Figure 2) provides a value of tuning parameter,
relaxivity 1 = 7.4m/s and a volume of invisible microporosity equal to 29.7% of the
total capacity; i.e., 8.2 p.u. The microporosity is the amount of the volume, which is
below the XmCT scanner resolution (~5m), i.e., the difference between the NMR and
the XmCT porosities.

SCA2015-042

5/7

Combining 2 spectra with XmCT data leads to a relaxivity value 2 = 24 m/s. Higher
values of the relaxation activity of the pore surface in the second case were caused by the
various physical mechanisms of longitudinal and transverse relaxation. Overestimation of
2 values is caused by the diffusion relaxation mechanism that originated from
significant internal magnetic gradients, as T2 is strongly influenced by echo time. These
gradients exert tremendous influence on NMR studies with high-value of magnetic field
(20.6 MHz). Differences in 1 and 2 spectra are clearly observed on a 2D 12 map
(Figure 2). Notice the area with an inflated 1/2 ratio in the upper-center region on the
map. This separate region can be associated with the presence of the hydrocarbons
remaining after extraction and/or contrasting pore surface areas with high-content
paramagnetic centers that enhance local diffusion processes, and hence, drastically reduce
2.
The differences of the T1 and T2 spectra described previously forced us to use
longitudinal relaxation as internal-free gradients; therefore, a more reliable mechanism
for observing pore size characteristics of multimineral rock composition.
The Xradia machine provides enhanced image resolution at the cost of sample size. We
use this scanner to verify the consistency of the XmCT results. Both XmCT machines
provide similar results; i.e., precisely duplicating large pores (right T1 mode) in the
calculated NMR pore size distribution (Figure 2).

Figure2: (a) Pore size distributions measured by different analysis methods (NMR T1, XmCT, SEM, and
FIB-SEM); (b) T1T2 map for the sandstone rock, West Siberia
(Quartz=62%, Feldspar=20%, Clays=18%, =27.8%, Permeability=407mD).

After conducting XmCT and NMR measurements on the cylinder, we obtained a small
portion of the sample and performed an FIB-SEM study to characterize microporosity,
which is invisible to common XmCT machines. Figure 2 also shows the resulting pore
size distribution. The reconstructed image of the sample 3D pore geometry structure
helped to visualize and detail small grains, their surfaces, internal structure of feldspars,

SCA2015-042

6/7

and clay mineral skeletons. The high-scale resolution of this method allowed for us to
construct the distribution of tiny pores with sizes in the range of 50 nm to 1000 nm. The
disadvantage of the FIB-SEM technique is its very small field of view (1515 m in our
case), which could reduce measurement representativeness of a specimen, making
volumetric evaluation difficult. Fortunately, the sample we studied was very
homogeneous sandstone rock, and similar to the FIB-SEM field of investigation areas,
which could be estimated on the 2D SEM image with a larger scale. The FIB-SEM
method was used for an approximate estimation of the volume of rock, which is
attributed to microporosity and construct the pore size distribution in porosity units
(Figure 3). The sphere fitting technique was also applied during the evaluation of pore
size distribution based on SEM data. Suitible repeatability of small pore sizes
distributions constructed by NMR and FIB-SEM identified a similarity in the surface
abilities to relax the hydrogen nuclei saturating the core. In other words, the
pseudorelaxivity value obtained by XmCT and NMR spectra matching is near the small
pore region. This result indicates a consistency by combining the input of mineral
composition and a specific surface value of the rock matrix.
The SEM image itself contains information about pore scaling. The advantage of the
SEM method is the larger field of view compared with the FIB-SEM image, but the data
are 2D only.
a

Figure3: (a) SEM slice image examples and resulting pore geometry evaluation technique based on
inscribed spheres (b). Each color defines a sphere diameter

This explains why we had to estimate the pore areas instead of volumes and used a circlefitting technique, which is analogous to previous FIB-SEM technique. Constructed
distribution reflects approximate characteristics of pore structure over a wide range of
sizes (0.4 40 m).

CONCLUSIONS
The results obtained in this study describe the advantages and drawbacks of different
methods of pore structure evaluation. These results point out the necessity for using an
integrated study of the pore space structure of rocks by the different methods to investe
pore sizes over a wide scale range and resolution. This detailed investigation using

SCA2015-042

7/7

several methods allows for avoiding uncertainties in the interpretation of NMR spectra
and consequent evaluation of relaxation activity factors for different pore scales. Sample
size as well as core preparation techniques have a strong influence on pore structure and
their consistency in different acquisition methods.

ACKNOWLEDGEMENTS
We would like to thank Schlumberger for permission to publish this paper. We also
acknowledge Dmitry Korobkov and Igor Varfolomeev for their support in performing the
experiments.

REFERENCES
1. Kleinberg, R., 1994, Pore Size Distribution, Pore Coupling, and Transverse
Relaxation Spectra of Porous Rocks, Mag.Res.Imag., 12, 2, 271 - 274.
2. Denisenko, A. and Yakimchuk, I., 2014, X-Ray Microtomography and NMR:
complimentary tools for evaluation of pore structure within a core, Society of Core
Analysts Symposium, 811 September, Avignon, France.
3. High-resolution micro-CT. Bruker Micro-CT. SkyScan 1172,
http://www.skyscan.be/products/1172.html.
4. High resolution 3D X-ray Microscopes. ZEISS Xradia,
http://microscopy.zeiss.com/microscopy/en_de/products/x-ray-microscopy.html
5. Lee, E., et al, 2004, 3D Materials Characterization using Dual-Beam FIB-SEM
Techniques. Microscopy and Microanalysis, 10 (Suppl. 02), 1128-1129.

SCA2015-043

1/8

MEASUREMENTS OF MOBILE WATER SATURATION IN

OIL SANDS
J. Butron1, J. Bryan1, 2, Y. Duan1 and A. Kantzas1, 2
1. PERM Inc.
2. University of Calgary
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
The presence of mobile water saturation in heavy oil and bitumen systems is of
considerable importance when understanding how different reservoir recovery methods
will work. In non-thermal heavy oil production, the presence of mobile water saturation
has extreme significance to the production of cold heavy oil (CHOPS) wells. Some wells
are able to produce oil under relatively low and constant water cuts, while other wells
produce significant water and quickly need to be shut in due to high production water
cuts. In this study, an NMR and core analysis approach is used to shed some light into
the properties of oil sand with or without mobile water present. Specifically, tests were
run in order to study if mobile water is a localized pore scale phenomenon or if it can be
understood through a more macroscopic view of the reservoir.
Samples of core are taken from two heavy oil producing wells: one that experienced very
high production water cuts, and the other that produced mainly heavy oil with minimal
water. The samples (containing heavy oil and connate water) are flooded with water in
order to measure effective permeability to water. NMR spectra are also obtained after
flooding and after cleaning of the cores, in order to understand the pore-scale location of
water in different effective permeability samples. The outcome of this study is insight
into what is the difference in core properties between wells with vs. without mobile
water. The key conclusion reached is that wells behave similarly at the pore/core scale,
but differences are observed macroscopically. The well with high historical water
contains thin zones of high water saturation (high water effective permeability) which
were missed at the resolution of the logs. Water production may be due to the presence
of these heterogeneous zones.

INTRODUCTION
Cold heavy oil production with sand (CHOPS) is a unique non-thermal reservoir recovery
strategy that is applied in many unconsolidated oil reservoirs in Alberta and
Saskatchewan. This recovery technique involves providing maximum drawdown to pull
oil to production wells, and is based on the concepts of process enhancement through
foamy oil flow [1,2] and sand production [3]. During CHOPS production, the high
pressure drawdowns lead to the generation of discontinuous gas droplets within the oil
phase, which keep the differential pressure high in the system and drive foamy oil to the

SCA2015-043

2/8

production wells. Furthermore, sand production creates regions of high permeability

(wormholes) throughout the reservoir, which act as conduits for further production of oil.
CHOPS production is run in unconsolidated oil sand reservoirs, where sand permeability
is on the order of Darcies and oil viscosity is on the order of 1,000 50,000 mPas at
reservoir temperatures. The high oil viscosity is the biggest challenge to production; if
there is mobile water present in these reservoirs, water will tend to flow preferentially to
oil. Accordingly, oil production will decline and wells will water out. Identifying if a
given heavy oil pool has mobile water is very important for assessing the chance of
economic success for CHOPS wells drilled into that pool.
This document is a case study of two CHOPS wells from northern Alberta. The oil in
these wells is on the order of 30,000 50,000 cP at reservoir conditions, and has an insitu solution GOR on the order of 15 m/m. In the initial identification of oil pay, the
determination for production was made on the basis that both wells have porosity > 30%
and measured resistivity (Rt) values greater than 10 ohmm. In terms of production,
Figure 1 shows that Well 1 produced oil under relatively low and constant water cuts,
while Well 2 exhibited much higher water production and, despite attempts to control the
production, the water cut spiked again and eventually Well 2 was shut in. This historical
production was observed despite the fact that the resistivity (indicative of oil saturation)
is higher in Well 2 vs. Well 1 (Figure 2 and Figure 3). The goal of this study was to
investigate if the differences between these wells can be understood on the basis of porescale water location within the sand of the two formations, or if the high water production
in Well 2 is due to a more macroscopic phenomenon.

EXPERIMENTAL PROCEDURE
The study of pore-scale water location was made through NMR and permeability
measurements in oil sand cores. 1 core plug samples are taken from stored frozen full
diameter core intervals from the two heavy oil wells (Well 1 and Well 2), and samples
are shown in Table 1. NMR spectra on the initial cores showed that the samples were
partially dried out. The samples were vacuum saturated with brine. The effective steady
state permeability to water was measured by running brine through the core at low DP/L
values, in order to measure water permeability without producing any oil. The NMR of
the cores were then re-measured and Dean-Stark was done to measure the connate water
present in each vacuum saturated core. The cleaned sand was saturated with water to
measure the true pore size distribution of the sand in each core with NMR.
Furthermore, core analysis (Dean-Stark) water saturations were present over the oilbearing interval in both wells. Based on correlations of water effective permeability vs.
water saturation on the tested samples from Table 1, a macroscopic study was also run to
output profiles of predicted effective water permeability as a function of depth (and water
saturation) in each well.

SCA2015-043

3/8

RESULTS
Table 1 shows the core samples that were taken from Wells 1 and 2. The Dean-Stark
water saturations are measured on the cores after they have been vacuum saturated and
water effective permeability has been measured. These permeability values are also
listed in Table 1. Effective permeability values are low in the core samples with higher
oil saturation, which is expected.
Two types of flow behaviour were observed in the tested cores. One flow type is shown
in Figure 4, which is the spectra from a low oil saturation lean zone in the reservoir.
Water vacuum saturated and flooded through the core is in the same T2 ranges as the
actual NMR pore sizes from this sand. In other words, water exists in the core in direct
contact with the sand grains, so water flows as a wetting fluid [3,4]. In contrast, Figure 5
is a sample containing higher oil saturation. In this case, water from saturation and
flooding exists mainly as a slow-relaxing peak at longer T2 values than the true pore sizes
for this sample. In oil-rich samples, water injected even at low injection pressures tends
to finger through the oil, i.e. viscous forces dominate flow patterns in high oil saturations.
These figures demonstrate that there are two different flow types present in oil sand, and
the physics of displacement will be different for both flow types. Unfortunately, both
wells exhibited this same behaviour for high Sw vs. low Sw samples. While the NMR
shows different flow behaviour in oil rich vs. lean zones, this cannot be used to infer
microscopic differences between Wells 1 and 2. Furthermore,
Figure 6 shows that both wells have a similar behaviour in terms of effective water
permeability vs. water saturation. In other words, the high water production in Well 2 is
not a pore-scale difference in wettability or water location within this formation.
A macroscopic view of the wells provides better insight into the production response of
the two wells. Error! Reference source not found. is the resistivity log (calculated) and
core (measured) water saturation profile over the producing zone of this well. The black
box at the left of the figure indicates the perforation zone within the oil formation. Water
saturations from log and core were used to estimate effective permeability to water by
applying the trend line from the core measurements in
Figure 6. Error! Reference source not found. shows that this well has low effective
permeability to water: values are generally 10 mD or less, as predicted from the water
saturation profile. In contrast, Error! Reference source not found. plots the resistivity
log (calculated) and core (measured) water saturation profile for Well 2. The log initially
provides a tightened water saturation profile with depth, but from the finer resolution core
measurements, it is observed that there are thin intervals present in this formation that
have much higher water. When water saturations are used to predict effective water
permeability, Figure 7 shows core predictions of water permeability that can be in the
range of 100 mD or higher in these thin lean streaks. The actual resistivity data (Figure
3) was noisy, compared to the more gradual resistivity changes in Well 1 (Figure 2).
These changes were initially smoothed out in the log predictions of saturation, but the
inclusion of core Dean-Stark data shows that these Rt variations are physically present,
and should be included in the calculated water saturation profile for this well. A better

SCA2015-043

4/8

log Rt model is shown that has higher fidelity to the Rt variations measured in this well,
and greater variations in the predicted permeability as a direct result of these R t
heterogeneities.
The results of this study can provide insight into the ability of these unconsolidated
reservoirs to flow non-thermal oil. With absolute permeability on the order of 100
1000 mD and water permeability only around 10 mD, heavy oil can be produced even
though its viscosity is so much higher than that of water. As water permeability increases
by an order of magnitude the mobility of water is higher than that of oil, and the well will
preferentially start to produce water.

CONCLUSION
Water effective permeability at connate water saturations is on the order of 1 10 mD in
oil rich zones, and can approach 100 mD or higher in lean oil zones. The response of
high vs. low water producing CHOPS wells is not due to microscopic (pore level)
differences between the wells. Instead, the high water production can be related to local
variations in water saturation within the oil zone. Thin streaks of high water saturation
can be quite permeable to water, and these zones may be missed from logging tool
interpretations. Proper reservoir description requires an understanding of heterogeneities
in fluid saturations, either through the collection of core or by ensuring that the well log
models have fidelity to the variations in measured Rt in the producing intervals.

ACKNOWLEDGEMENTS
The authors gratefully acknowledge Devon Canada for providing core samples and
logging tool information for these case study wells. Financial support for this work has
come partially from the NSERC Chair in Fundamentals of Unconventional Resources
(FUR), the University of Calgary, and its industrial sponsors: Laricina Energy, Husky
Energy, Athabasca Oil Corp, Suncor, Brion Energy, CNRL, Devon, Foundation CMG
and Alberta Innovates.

REFERENCES
1. Maini,B, Foamy Oil in Heavy Oil Production, J. Can. Pet. Tech., 35 (6), 21
24, Jun 1996.
2. Firoozabadi, A., Mechanisms of Solution Gas Drive in Heavy Oil Reservoirs, J.
Can. Pet. Tech., 40 (3), 15 20, Mar 2001.
3. Tremblay, B., Cold Flow: A Multi-Well Cold Production (CHOPS) Model, J.
Can. Pet. Tech., 48 (2), 22 28, Feb 2009.
4. Bryan, J., Mai, A. and Kantzas, A., Investigation into the Processes Responsible
for Heavy Oil Recovery by Alkali-Surfactant Flooding, SPE 113993, 2008 SPE
Improved Oil Recovery Conference, Tulsa, OK USA, Apr 19 23, 2008.
5. Al-Mahrooqi, S.H., Grattoni, C.A., Muggeridge, A.H. and Jing, X.D.,
Wettability Alteration during Aging: the Application of NMR to Monitor Fluid

SCA2015-043

5/8

Redistribution, SCA 2005-10, International Symposium of the Society of Core

Analysts, Toronto, ON Canada, Aug 21 25, 2005.

SCA2015-043

Well
No

Sample

6/8

kw
(mD)

Dean-Stark post
waterflood
So
Sw
(fraction) (fraction)

46.4

0.284

0.716

150.9

0.524

0.476

162.4

0.509

0.491

0.644

0.356

2.8

0.516

0.484

0.682

0.318

1.4

0.712

0.288

42.7

0.012

0.988

5 1300
0.058
0.942
Table 1: Core Analysis Samples Tested for NMR
and Water Permeability

Figure 2: Well 1 Resistivity Profile

Figure 4: NMR spectra for low oil saturation core

Figure 1: Historical water cuts of Well 1 and Well 2

Figure 3: Well 2 Resistivity Profile

Figure 5: NMR spectra for high oil saturation core

SCA2015-043

7/8

Figure 6: Water effective permeability vs. water

saturation

Figure 7: Log and core Sw profile for Well 1

Figure 8: Log and core predictions of effective

kw for Well 1

SCA2015-043

Figure 9: Log and core Sw profile for Well 2

8/8

Figure 7: Log and core predictions

of effective kw for Well 2

SCA2015-044

1/6

STUDY OF PETROPHYSICAL PROPERTIES

ALTERATIONS OF CARBONATE ROCKS UNDER
CARBONATE WATER INJECTION
Eric Yuji Yasuda1, Erika Tomie Koroishi2, Osvair Vidal Trevisan3 and Euclides Jos
Bonet3.
1
Mechanical Engineering, State University of Campinas, UNICAMP; 2Center for
Petroleum Studies CEPETRO; 3 Department of Energy, State University of Campinas
UNICAMP.
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Carbon dioxide (CO2) injection in reservoirs promotes rock-fluid interactions, which
depends on rock nature, brine composition, partial pressure of CO2, temperature as well
as operational conditions. The rock-fluid interactions cause changes in the petrophysical
properties, modifying both porosity and permeability of the rock. The present study aims
to study the effects of water injection with dissolved CO2 on the petrophysical properties
of carbonate rocks. The effects are evaluated experimentally by displacement runs on a
long core of an outcrop coquina. The work emphasizes the evaluation of permeability
changes along the length of the core, using a coreholder equipped with multiple pressure
taps. The experiments were conducted in dynamic regime, at T=22C and P=2,000 psi
and flow rates of 0.5, 1 and 2 cc/min. X-ray Computerized Tomography (CT) was used to
determine porosity alterations during the experiments and the differential pressure drop
was used to calculate permeability and its changes along the rock sample. Results show
that porosity increased steadily with the accumulated volume of water injected.
Permeability increased sharply at the higher flow rates and the changes were unevenly
distributed along the core.

INTRODUCTION
The occurrence of reactions between the CO2 enriched brines and carbonate rocks, has
been reported in several applications. The interaction between fluids containing CO2 and
the rock is often found in the literature associated with matters of capture and
underground storage of gas. In accordance to Ott et al. 2013 [1] reactions may influence
the fluid-flow field, i.e. reactive transport, and the mechanical rock properties, which
might degrade, leading to uncertainties with respect to the rock integrity in the affected
region. Luquot and Gouze [2] analyzed the dissolution of carbonate rocks submitted to
flow of aqueous solutions. In the experiments, measurements of salinity and pH were
carried out at different positions along the flow and then it was possible to classify the
mass transfer processes both near wellbore and further, where the dissolution becomes
increasingly uniform. Zekri et al. [3] observed that the CO2 injection at supercritical state

SCA2015-044

2/6

alters rock permeability and that the alteration is related to the rock composition. Taking
it into account, the changes must be evaluated along time for the particular rock in
consideration, due to the heterogeneities present in the reservoirs.
An important characteristic associated to the CO2 injection in reservoirs is the presence of
chemical interaction between the fluid and the rocks. With the CO2 injection in the
reservoir, a fraction of the gas dissolves into the formation brine or into the injection
water, producing carbonic acid (H2CO3), which dissociates as charged particles with
potential to interact chemically with the rock minerals (calcite, dolomite, anhydrite).
These reactions are known to geochemists to provoke dissolution or precipitation,
depending on the direction they manifest. They are a function of the rock nature, the
brine composition, the partial pressure of CO2 and the thermodynamic conditions. The
movement of the fraction of CO2 dissolved in the formation water is dependent on the
transport mechanisms (diffusion or convection) and the reaction kinetics [4]. The mass
transfer between the fluid and the rock relies on numerous parameters, such as: partial
pressure of CO2, concentration of cations in the formation water, the injection flow rate,
the reactive surface area, porosity, permeability and tortuosity of the medium, among
others.
The presence of the rock-fluid chemical interaction may change the pore structure and
modify the permeability of the rock, which is a key parameter to the flow in reservoir
engineering. The present paper reports an experimental investigation on the permeability
and porosity changes of a carbonate rock with the injection of carbonate water.

PROCEDURES
Materials
The carbonate rock used in the study was extracted from a coquina outcrop, from Morro
do Chaves formation, in the Sergipe-Alagoas basin, Brazil. Three fluids were employed
in the experiment: a sodium chloride (NaCl) brine at 35 kppm concentration as the initial
saturation fluid, a sodium iodide (NaI) brine at 35 kppm as the dopant and a sodium
chloride brine at 35 kppm saturated with CO2 at 2,000 psi as injection fluid.
Methods
The experiments were performed on the experimental setup made up basically by the
following devices: a positive displacement pump to guarantee a measurable and
continuous flow, high pressure vessels for the fluids conditioning, a special coreholder
with multiple pressure taps connected to differential transducers referenced at the inlet
point and a backpressure system at the outlet port. The coreholder was composed of an
aluminum cylinder, with an epoxy and carbon fiber jacket around the sample. The
experimental apparatus is schematized in Figure 1.

SCA2015-044

3/6

Figure 1. Scheme of the experimental apparatus.

The sample was saturated with the 35 kppm NaCl brine and then flushed with the
contrast fluid, the 35 kppm NaI brine. After that, the tests with carbonated water were
performed at the flow rates of 0.5, 1 and 2 cc/min. The operating conditions were 2,000
psi and 22C. Differential pressures were measured at points along the core as shown in
Figure 1 by appropriate sensors (nVision Crystal Engineering Corporation). The
permeability was evaluated by the pressure drop along the sample. The pressure was
measured in six different points along the sample and also on the inlet and outlet valves,
resulting in eight pressure taps, as schematized in Figure 2.

Figure 2. Scheme of the permeabilities evaluated during the test.

Porosity was evaluated by X-ray computerized tomography (Siemens CT scanner model

SWFVD30C), using the software Syngo. The images had a 0.5 mm resolution and each
scan acquired 69 images along the sample.

RESULTS
Porosity
A MatLab routine was used to calculate the porosity of each section. These data were
acquired by the continuous CT scan along each test cycle. Porosity was calculated in
accordance to Equation 1, where CTn, CTrock and CTfluid represent the mean CT
number for the system in a given time, the rock CT number and the fluid CT number
respectively.

(1)

SCA2015-044

4/6

Figure 3 shows the behavior of the mean porosity, which showed a steady increase of its
values with the increase of the accumulated volume of water injected. A slightly greater
increase was observed at the beginning of the test, when a flow of 1 cc/min was used.
The overall increase across the entire test was from 13% to 17%, a substantial change if
one considers the volumes of a whole reservoir.

Figure 3. Mean porosity by volume of carbonated water injected.

Using data from each transversal section (69 total slices) of the CT-scan, porosity was
calculated at each particular time and position. It was possible to verify porosity
evolution through time along the entire core. The evolution and distribution of porosity is
depicted in the 3D plot shown in Figure 4.

Figure 4. Evolution of porosity along whole sample through time.

During the injection test, small solid white particles were observed to accumulate at
bottom of the container of the produced fluids, evidencing the occurrence of dissolution
of the rock. In accordance with Grigg et al. 2013 [5], carbonate dissolution caused
changes in core permeability and porosity.

SCA2015-044

5/6

Permeability
Figure 5 depicts the variation of the average permeability of the rock sample along the
duration of the experiment. Permeability remained basically unchanged up to 120 PVI
(pore volumes of water injected). It is important to notice that from the beginning of the
test up to 100 PVI the test was carried out at a flow rate of 0.5 cc/min. A steep increase in
permeability, from 0.2 D to 12 D, was observed between 120 to 170 PVI. Such
remarkable increase is associated to the formation of a preferential path to the flow which
was visible at the CT images.

Figure 5. Mean permeability by pore volumes of carbonated water injected.

The significant change in permeability happened soon after the flow rate was raised to 1
cc/min. At 12 D the permeability curve leveled off and remained flat with the continued
injection of carbonate water even when the flow rate was increased to 2 cc/min.
Through the experiments, the permeability of all regions from K0 to K6 could be
evaluated along the test, and the result is shown in Figure 6. Evolution of permeability
was not even through the rock sample. Sections K2 and K6 showed the greatest
permeability changes at about 120 pore volumes injected, and it probably occurred due to
fluid breakthrough in a wormhole which increased the permeability of the entire core. For
these sections, the permeability dropped afterwards probably because of the movement of
fine grains of rock which clogged some permeable channels.

Figure 6. Permeability by sections of the sample against pore volumes injected.

SCA2015-044

6/6

CONCLUSION
It was possible to verify that both injection rate and also the interaction between rock and
fluid promoted alterations on the rock petrophysical properties. The porosity showed a
progressive increase through the tests. It can be verified the porosity evolution through
time along the entire core. The mean porosity performance showed a steady increase of
its values with cumulative volume of water injected. Porosity indicated a considerable
change along all test. Initially, the outcrop porosity was 13% and in the final test it
showed 17% that it represents a considerable change for a whole reservoir.
Permeability remained basically unchanged up to 120 PVI (pore volumes of water
injected). It was verified that from the beginning of the test up to 100 PVI the test was
carried at a flow rate of 0.5 cc/min. The experiments showed the creation of a preferential
path, which increased greatly the rock permeability, showing evident when the flow rate
was increased to 1 cc/min.

ACKNOWLEDGEMENTS
The authors acknowledge PETROBRAS and CNPq for financial support of the present
study.

REFERENCES
1. Ott, H et al, CO2 Reactive transport in limestone flow regimes, fluid flow and
mechanical rock properties. International Symposium of the Society of Core
Analysts, 2013, SCA-2013-029.
2. Luquot, L. and Gouze, P., Experimental determination of porosity and permeability
changes induced by injection of CO2 into carbonate rocks. Chemical Geology 265,
148-159, 2009.
3. Zekri, A. Y., Shedid, S. A., Almehaided, R. A. Investigation of supercritical carbon
dioxide, aspheltenic crude oil and formation brine interactions in carbonate
formations. Journal of Petroleum Science and Engineering, v. 69, 63-70, 2009.
4. Grigg, R. and Svec, R., Co-injected CO2-brine interactions with Indiana limestone,
International Symposium of the Society of Core Analysts, 2003, SCA-2003-19.
5. Mangane, P. O. Caractrisation des changements dans les proprits de rservoir
carbonat induits par une modification dans la structure des pores lors dune injection
de CO2: application au stockage gologique du CO2. Tese, Sciences de Techniques
du Languedoc-Roussillon, Universite Montpellier II, 2013.

SCA2015-045

1/6

CORNER OIL FILM ELEVATION ABOVE THE GAS-OIL

INTERFACE IN WATER-WET CAPILLARIES
Hossein Khorshidian, Lesley James and Stephen Butt
Memorial University of Newfoundland
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
In gas assisted gravity drainage, very high oil recovery factors can be obtained when a
thin oil film is formed over the water surface in water wet porous media. The formation
of an oil film, as an intermediate wetting phase, over the water (wetting phase) surface in
the presence of gas (non-wetting phase) is typically linked to the spreading coefficient
which is the gas-water interfacial tension subtracted by the sum of the oil-water and oilgas interfacial tensions. However, the oil film characteristics not only depends on the
interfacial tension, but also depend on the geometry of the phase contacts on solid surface
which is affected by the geometry and wettability of the solid surface. The oil film, which
may exist between the water and gas in the corner of capillaries, can play an important
role in oil recovery by gravity drainage during immiscible gas injection. The thin oil film
can be the path to transfer the trapped oil in smaller pore-throats, which is left after gas
injection, toward the oil bank beneath the gas front. Additional oil recovery can be
obtained by gravity drainage if the gas-oil capillary pressure is high enough for the entry
of non-wetting phase into the smaller pore-throats. In this article, the oil film
characteristics in the corner of a simple square capillary tube are modelled, based on the
size of the capillary tube, phase interfacial tension, phase differential density and the
distance between the gas front and oil front, through balancing the gravity and capillary
pressure under equilibrium condition. The results show that having a higher gas-oil
interfacial tension, lower gas-oil differential density and larger oil bank size provides an
intermediate phase (oil) with better hydraulic communication over a greater elevation.
The proper hydraulic communication above the gas front assists the gravity drainage
mechanism to recover more oil from smaller pore-throats in particular scenarios. These
scenarios are identified and discussed in this article.

INTRODUCTION
Oil displacement by gas-assisted gravity drainage in a water wet porous media often
results in a high recovery factor. In gas-assisted gravity drainage, both gravity and
capillary phenomena affect the oil recovery. The conditions that provide an efficient oil
recovery factor can be identified by characterizing mechanisms of fluid displacement in
gas-assisted gravity drainage. Oren and Pinczewski [1] investigated three phase fluid
flow in porous media and drew a relationship between the phase interfacial tensions and
the state of the fluids contacts configuration. In many investigations, it has been stated

SCA2015-045

2/6

that high oil recovery by vertical gas injection in water-wet porous media can be obtained
if the oil phase spreads over the water surface in the presence of gas, and the spreading
condition of oil over the water surface is only controlled by the interfacial tension
between each pair of fluids [2, 3, 4]. Further investigations showed that the oil
displacement mechanism is also affected by capillary pressure rather than by interfacial
tension alone [5]. The capillary pressure in addition to the fluids interfacial tensions is a
function of the curvature of the interfaces between fluids, which is affected by the
capillary geometry and phase contact angle. A stable oil layer over the water surface has
also been observed in a system with a negative oil spreading coefficient [5, 6]. Dullien et
al. [7] showed that the rate of oil recovery in porous media, when an oil film forms on the
smooth glass bead surfaces having positive spreading coefficient, is very low and
unmeasurable. Conversely, an oil film that is formed in corners of wedges and crevices
on the surface of scratched glass beads is thicker. The thick oil films yielded higher rate
of oil recovery that could be measured. Blunt et al. [8] calculated the height of the oil
film in porous media for the three phase system. They compared the oil-water and oil-gas
contact radii in capillary corners evaluating capillary pressures based on the fluids
interfacial tensions and densities. Depending on the fluid interfacial tensions and
differential densities, two different types of the oil film are formed in the corner of
capillaries. In the first type, the oil film can be present over the water surface at any
elevation, and in the second type, the presence of the oil film is limited to a critical
elevation. Blunt et al. [8] predicted that for conditions in which the oil height is limited,
the oil saturation after vertical gas injection is very low. However, a reverse result was
observed for a system in which oil has a negative spreading coefficient.
In this article, the equilibrium height of the oil film in the corner of a water-wet and
square shaped capillary tube is calculated varying the fluid characteristics. The
relationship between the residual oil saturation after gas breakthrough and the state of the
thin oil film is discussed. The results of this study provide insight for the phenomena
affecting the gas-assisted gravity drainage oil recovery method.

THEORY
The oil film in a water-wet and square shaped capillary tube with sharp corners can be
characterized if the capillary pressure between each pair of phases is known. Fig. 1
schematically shows a square capillary tube containing gas (non-wetting phase), oil
(intermediate wetting phase) and water (wetting phase). The equilibrium capillary
pressure between oil and water at point 1 in Fig. 1a, (P*cow), is shown in Eq. 1 [8]
P*
cow

ow

cos
D

ow

(1)

where, ow is the oil-water interfacial tension, ow is the oil-water contact angle, and
D is the size of the square tube side. The capillary pressure (Pc) at any point inside the
tube is the local differential pressure between the non-wet phase (Pnw) and wetting phase
(Pw) as indicated by Eq.2 [8].

SCA2015-045

3/6

P P P
c
nw
w

(2)

Oil and water hydrostatic pressure along the corners of the vertical tube vary differently
depending on their densities. The water and oil pressure above point 1 (oil-water contact
front) in Fig. 1a are calculated by Eq. 3 and 4;
P P* g ( H L)
w
w
w
*
P P g ( H L)
o
o
o

(3)
(4)

where, Pw and Po are water and oil pressures along the tube, P*w and P*o are the
water and oil hydrostatic pressure at point 1 in Fig. 1a, w and o are oil and water
densities, H is the height of the oil film above the gas-oil contact front (point 2 in Fig.
1a), L is the oil bank length (distance between point 2 and point 1 in Fig. 1a), and g is
gravitational constant, which is assumed to be 9.8 m/s2 in the calculations.
The oil-water capillary pressure at any location above point 1 can be calculated by
substituting Eq. 3 and 4 in Eq. 2. Knowing the gas-oil interfacial tension (go) and gas-oil
contact angle (go), the gas-oil capillary pressure above their contact at point 2 in Fig. 1a
can be calculated in the same manner. The oil-water and gas-oil capillary pressures,
Pcow and Pcgo, are given in Eqs. 5 and 6, respectively;
P
P* g ( H L)
cow
cow
ow
P
P* gH
cgo
cgo
go

(5)
(6)

where, ow and go, are the oil-water and gas-oil differential densities respectively,
P*cow, is the local oil-water capillary pressure at point 1, and P*cgo is the local gas-oil
capillary pressure at point 2. Neglecting the fluid contact curvature along the capillary
corner, the oil-water and gas-oil contact radii can be calculated using Eqs. 7 and 8.
cos
ow
ow
r
ow
*
P
g ( H L)
cow
ow

(7)

cos
go
go
r
og
*
P
gH
cgo
go

(8)

It can be seen in Fig. 1a and Fig. 1b, the fluids contact radii become smaller by moving
up along the tube corners. In a water-wet capillary tube with sharp corners, depending on
fluids interfacial tensions and differential densities, two different configurations of oil
film can be formed. In the first configuration, the oil-water contact radius shrinks toward
corner of the tube with a smaller rate rather than the gas-oil contact radius. Since at gasoil contact front (H=0) the gas-oil contact radius is larger than oil-water contact radius,
there is an elevation above the gas-oil contact front at which the contact radii of both gas-

SCA2015-045

4/6

oil and oil-water pairs are equal (point 3 in Fig. 1a). The second configuration of the oil
film is formed when the gas-oil contact radius along the tube corner remains larger than
the oil-water contact radius, as depicted schematically in Fig. 1b. Such a condition is
likely to occur when the differential density of the gas-oil pair is lower than the oil-water
pair. The elevation of the oil film in this configuration is no longer limited by the oilwater contact radius, and the oil film exists at all heights above the water surface in the
tube corners.
The maximum equilibrium oil film height above the gas-oil contact front (Hmax) for the
first configuration is shown by Eq. 9, which is derived by equating the oil-water and gasoil contact radii.
( cos .L)
og
go ow
(9)
H

max ( cos ) ( cos )

ow

ow

go

go

go

ow

DISCUSSION
We can see from Eq. 9 that the size of tube has no effect on the maximum oil film
elevation. Fig. 2 shows the variation of Hmax vs. gas-oil interfacial tension and the gas
density when the oil-water interfacial tension (ow) is 0.03 N.m-1, oil density (o), is 700
kg.m-3, water density (w) is 1000 kg.m-3, length of oil bank (L) of 0.05 m, and all contact
angles are zero degrees. It can be seen that the
Fig. 3 shows the maximum oil film elevation (Hmax) vs. the oil bank length when the gasoil interfacial tension is fixed at 30 mN.m-1, gas density at 300 kg.m-3 and other
parameters at their previous fixed levels. A greater oil film elevation can be expected for
the first configuration when the length of the oil bank is longer. Fig. 4 shows the gas-oil
capillary pressure at the top of the oil film in a square tube with the size of 100100 m,
versus the gas density (g) and gas-oil interfacial tension (go), and keeping all other
parameters fixed at their levels, previous indicated in Fig. 3. It is shown that the gas-oil
capillary pressure is higher when the gas-oil interfacial tension and gas density are
increased. Fig. 3 and 4 also illustrate that the effect of gas density on the oil film
elevation and gas-oil capillary pressure is more significant when the gas-oil interfacial
tension is higher.
In porous media, a thin oil film can be formed in the crevices and corners between the
rock grain and on their surface roughness. During a vertical gas injection, thick oil films
are the major paths for both hydraulic communication and recovery of trapped immobile
oil. Increasing the gas-oil capillary pressure provides a potential for the gas to enter the
small size pore-throats and deplete any residual oil through these paths. Consequently,
the overall oil recovery can be higher. Since the fluids contacts are not under equilibrium
conditions during vertical gas injection, the oil recovery from small pores should be
studied in a dynamic state of capillary pressure. The dynamic state of the oil film during
vertical gas injection is different from the equilibrium state. The non-equilibrium state of

SCA2015-045

5/6

oil film in gravity drainage is currently under experimental investigation at Memorial

University of Newfoundland.

CONCLUSIONS
The oil film elevation above the gas-oil contact front in a water-wet, non-circular
capillary tube is a function of the oil-water and gas-oil interfacial tensions and their
differential densities. Additionally, the oil film elevation can be limited due to the
presence of wetting phase in the sharp corners of the capillary tube. It is expected that a
longer oil film elevation results in higher oil recovery, since, it yields a higher gas-oil
capillary pressure, which allows the gas to enter the small pore-throats that are not swept
by primary gas injection. The analysis of the static oil film characteristics implies that the
injection of a gas with a higher density and having a greater gas-oil capillary pressure can
potentially improve the oil recovery factor in gas-assisted gravity drainage EOR methods.

ACKNOWLEDGEMENTS
The authors would like to thank Petroleum Research Newfoundland and Labrador
(PRNL), the Research Development Corporation of Newfoundland (RDC), Hibernia
Management and Development Company (HMDC), and the Natural Sciences and
Engineering Research Council of Canada (NSERC) for their technical and financial
support in this research.

REFERENCES
1. ren, P.E., & Pinczewski, W.V. 1995. Fluid distribution and pore-scale
displacement mechanisms in drainage dominated three-phase flow. Transport in
Porous Media, 20(1-2), 105-133.
2. Chatzis, I., Kantzas, A., & Dullien, F.A.L. 1988. On the investigation of gravityassisted inert gas injection using micro models, long Berea sandstone cores, and
computer-assisted tomography. In SPE 18284 Presented at the 1988 SPE Annual
Technical Conference and Exhibition, Houston, TX.
3. Vizika, O., & Lombard, J.M. 1996. Wettability and spreading: two key parameters in
oil recovery with three-phase gravity drainage. SPE Reservoir Engineering, 11(1).
4. Oren, P. E., & Pinczewski, W.V. 1994. Effect of wettability and spreading on
recovery of waterflood residual oil by immiscible gasflooding. SPE Formation
Evaluation, 9, 149-149.
5. Dong, M., Dullien, F.A.L., & Chatzis, I. 1995. Imbibition of oil in film form over
water present in edges of capillaries with an angular cross section. Journal of Colloid
and Interface science, 172(1), 21-36.
6. Keller, A.A., Blunt, M.J., & Roberts, A.P.V. (1997). Micromodel observation of the
role of oil layers in three-phase flow. Transport in Porous Media, 26(3), 277-297.
7. Dullien, F.A.L., Zarcone, C., Macdonald, I.F., Collins, A., & Bouchard, R.D. 1989.
The effects of surface roughness on the capillary pressure curves and the heights of
capillary rise in glass bead packs. Journal of Colloid and Interface Science, 127(2),
362-372.

SCA2015-045

6/6

8. Blunt, M., Zhou, D., & Fenwick, D. 1995. Three-phase flow and gravity drainage in
porous media. Transport in Porous Media, 20(1-2), 77-103.
9. Dullien, F.A.L. 1991. Porous media: fluid transport and pore structure. Academic
press, 2nd edition.

Fig. 1: Water, oil and gas in a non-circular tube

a) oil film elevation limited by water film, and
b) oil film elevated along the tube corner

Fig. 2: Oil film elevation above the gas-oil contact

front vs. gas density and oil-gas interfacial tension

Fig. 3: Oil film elevation vs. oil bank length

Fig. 4: Gas-oil capillary pressure at the top of the

oil film vs. gas density and oil-gas interfacial
tension

SCA2015-046

1/8

GAS MIGRATION THROUGH SELF-SEALED OR INTACT

CLAYSTONE
Yang Song, C. A. Davy, F. Skoczylas
Ecole Centrale de Lille, LML UMR CNRS 8107, CS20048,
59651 Villeneuve dAscq Cedex, France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
This contribution investigates gas migration through fully water-saturated CallovoOxfordian (COx) claystone.
Our originality is to evidence the nature of progressive passage through initially macrocracked and water-saturated (i.e. self-sealed) claystone, or through intact (undisturbed)
matter. The sample is subjected to hydrostatic pressure, identical to the mean in situ
principal stress. We use a mix of gases detected on the sample downstream side by a
mass spectrometer accurate to 2-5 ppm (i.e. 2-5 10-6 particles). We show that gas
breakthrough occurs first, by discontinuous capillary passage, i.e. snap off, followed by
continuous breakthrough (i.e. permeation), whether the upstream pressure value is
increased or kept constant. Snap off is characterized by small amounts of gas detected at
the outlet at random frequency and amplitude, until all gases in the mix are detected
continuously and simultaneously.
At continuous breakthrough, gas permeability is on the order of 10-21 m2 (1nD). For selfsealed claystone, gas breakthrough pressure (GBP) varies hugely in the low range 1.45-3
MPa. For undisturbed claystone, GBP ranges between 4.38 MPa (20 mm thickness) to
5.43 MPa (30 mm thickness), in good accordance with former research. It is concluded
that self-sealed COx claystone has significantly weaker gas breakthrough properties than
undisturbed matter.

INTRODUCTION
This contribution was originally devised for engineering applications related to deep
underground nuclear waste storage. In France, the latter is planned at 420-550 m depth,
and represents a network of horizontally-drilled galleries of several kms long, starting
from a main vertical shaft and an auxiliary vertical shaft [1]. Storage tunnels are made
within a claystone of Callovo-Oxfordian age (the so-called COx claystone), which has an
undisturbed water permeability of the order of 10-21 to 10-23 m2 (0.01 to 1 nanoDarcy). In
the Excavated Damaged Zone (EDZ) around the tunnels, macro-fracturing is observed.
However, underground water seepage is shown to allow self-sealing [2], whereby water
permeability recovers values on the order of 10-21 to 10-23 m2, similar to undisturbed
claystone.

SCA2015-046

2/8

Further, after drilling, filling and closure of storage tunnels, hydrogen gas may develop
and progressively pressurize inside the repository, due to varied physico-chemical
phenomena (anaerobic corrosion of carbon steel canisters, water radiolysis, etc.).
The industrial issue is to ensure the conditions for proper gas immobilization within the
underground site, both in the self-sealed zone and in the undisturbed claystone, away
from the repository: this is investigated here by identifying the Gas Breakthrough
Pressure (GBP) across fully water-saturated claystone. This study can be extrapolated to
the migration of a non wetting fluid (e.g. gas here, or oil in the petroleum engineering
context) through a wetting fluid filling the pores (water here, and water too in the
petroleum engineering context).
Extensive research has been performed on gas migration (or breakthrough) through COx
claystone with the step-by-step method [3, 4, 5] or with quicker, transient techniques
[6,7].
With the step-by-step method, progressive gas passage through water-saturated matter is
identified, so that both discontinuous and continuous migrations can be observed [8].
Through the test series presented here, our aim is to assess:
1. the difference between discontinuous and continuous gas passage: does the former
occur consistently before the latter?
2. is discontinuous passage mainly a feature of our experimental set-up, so that argon
gas detected in the downstream chamber is due to argon accumulation by diffusion
phenomena, or is discontinuous passage a capillary phenomenon, the so-called snap
off (or discontinuous capillary digitation)? For this part of our experimental
campaign, gas detection on the downstream side uses a mass spectrometer accurate to
2-5 ppm, and gas is either 100% argon or a mix of 50 mol% argon/50mol% helium.
3. is discontinuous or continuous GBP repeatable, i.e. after a first breakthrough at a
given gas pressure, whenever claystone is re-saturated with water until being selfsealed, does it have the same GBP again?

MATERIALS AND METHODS

Table 1: Main features (sample size, water permeability at saturation) and GBP test results for COx
claystone sample series S3.

SCA2015-046

3/8

Sample Origin and Preparation

Each of the five samples tested is cored and macro-cracked by Brasilian test (or splitting
test of a circular cylindrical sample), or machined by turning to 37 mm diameter and
varying height, from 9.5 to 30 mm, from horizontal core EST34386 or vertical cores
EST44331 and EST44140, see Table 1. Turning allows to minimize sample damage and
provides us with so-called undisturbed claystone, whereas macro-cracking provides
samples similar to damaged claystone from the EDZ.
Water Saturation. Prior to GBP, full water saturation is achieved as follows. The sample
is placed in a hydrostatic cell, and subjected to a confining pressure of 12 MPa (on the
order of in situ principal stress levels). Water is injected on the upstream side at 4 MPa
pressure, which is the lithostatic water pressure, until it is fully saturated. Full saturation
state is achieved when water permeability Kwater values fall below 10-20 -10-21 m2, see
Table 1. Owing to the water flow duration and to the stability in water permeability
values, it is thought that either all potentially trapped gas within the pore network has
evacuated, or has been dissolved in flowing water.
Single-gas GBP Experiment
Following water saturation, upstream pipes are emptied from water, while the sample is
kept at a constant confinement Pc=6 or 12 MPa, in order to avoid its premature failure.
The downstream chamber (of a volume of 2 cl) is closed by a dedicated valve, and its
pressure Pdownstream is recorded with a pressure transducer accurate to +/-100Pa. On the
upstream side, the argon gas pressure Pupstream is increased very slowly, at a rate of 0.5-1
MPa per day (unless otherwise stated in the following), and up to 14 days for a few tests.
Upstream gas pressure is given by a pressure transducer accurate to 1 kPa. Gas detection
on the downstream side is performed using both the downstream pressure transducer and

SCA2015-046

4/8

a dedicated argon gas detector accurate to +/-10-7 l/sec. At constant Pupstream value, gas
detection is performed every 24 to 72 h as follows: the downstream chamber valve is
opened and gas is detected (or not) by placing the detector nozzle at ca. 5 mm from the
valve opening. Gas passage is considered discontinuous whenever gas presence is
detected for less than 10-60 sec; gas passage is considered continuous when gas outflow
is recorded for more than 10 sec, three times every five minutes. This method means that
downstream gas pressure goes back to zero after each downstream valve opening. Also, it
does not ensure whether gas actually passes through the porous medium by dissolution
and diffusion in pore water, or by capillarity. Therefore, both mechanisms are accounted
for in the following.
Two-gas GBP Experiment
This test aims at determining what phenomenon is at the origin of gas breakthrough, i.e.
either percolation or dissolution and diffusion. For this purpose, a mix of gases is injected
through water-saturated claystone: it is chosen to ensure a significant difference between
both their solubility in water and their diffusivity, i.e. a significant difference in the
product of their solubility and diffusivity. If the gas with the greatest solubility and
diffusion migrates first, these phenomena will be privileged to explain breakthrough.
At 15 C, the solubility is 3.00x10-10 mol/Pa for argon and 7.20x10-11 mol/Pa for helium;
the diffusion coefficient is 2.00x10-9 m2/sec for argon and 6.30x10-9 m2/sec for helium
[9]. The product of solubility and diffusion coefficient at 15 C is of 5.92x10-19
mol/N.sec for argon and 4.48x10-19 mol/N.sec for helium: helium has lower solubility
and diffusivity by a coefficient of 1.32, when compared to argon. Although this
difference may not appear large enough, these gases have been chosen in a first approach
given their excellent availability and harmlessness.
After imposing the gas mix on the sample upstream side at given pressure, the presence
of each gas is detected individually on the downstream sample side by a mass
spectrometer accurate to 2-5 ppm.

RESULTS AND DISCUSSION

In the following, all pressure values are expressed in MPa abs.
Evidence of Discontinuous and Continuous Gas Passage
When the GBP test starts, whatever the initial Pupstream value chosen (as low as 0.2 MPa),
Pdownstream increases. This is attributed to water expelled on the downstream side, pushed
by gas on the upstream side. Therefore, potentially, gas entry may have begun, yet no
device is available in this experiment to check it with adequate accuracy. The argon
detector does not record any gas passage after 24 h at Pupstream = 2.02 MPa, whereas it
detects a discontinuous argon passage after 24 h at Pupstream =2.25 MPa. Simultaneously, at
each Pupstream value, downstream pressure data display a linear increase with time (Fig. 1a),
until going back to zero as soon as the downstream valve is opened. Therefore, the rate of
increase of downstream pressure is plotted vs. imposed upstream gas pressure (Fig. 1b).
An inflexion point is observed in these data, at Pupstream=2.22 MPa. This is attributed to the
start of an additional phenomenon, which induces a greater rate of increase in

SCA2015-046

5/8

downstream pressure. As it occurs at a very close Pupstream value as that when the argon
detector evidences gas passage, it is attributed to the same phenomenon. This means that
we have two independent means of detecting gas passage: the argon gas detector, and the
rate of increase in downstream pressure, both occurring due to gas passage from the
upstream side (where gas pressure is imposed) and downstream side (where all
measurements are made).

Figure 1 : Sample S3-n.2: rate of increase of downstream pressure (in kPa/day) vs. imposed upstream gas
pressure (MPa).

After discontinuous passage has been observed, Pupstream is increased regularly until
continuous breakthrough (Fig. 2). Continuous breakthrough is measured by (1) the argon
detector, (2) the rate of increase of Pdownstream and also (3) when the rate of decrease in
upstream pressure is greater than that due to thermal variations. This second criterion
allows for measuring gas permeability Kgas, as described in [Davy et al. 2007] (with an
assumption of quasi-static gas flow). As for discontinuous passage, an inflexion point is
observed in the rate of increase in Pdownstream vs. Pupstream, which occurs at a slightly greater
Pupstream,=3.59 MPa than continuous passage measured by the argon detector (at
Pupstream=3.35 MPa). The upstream pressure transducer allows for gas permeability
measurement at an even greater Pupstream,=4.24 MPa than with the two other devices: Kgas
is of 1.6x10-21 m2 at Pupstream,=4.24 MPa. Whenever Pupstream, is increased further, Kgas
increases monotonously [4].

SCA2015-046

6/8

Figure 2. Sample S3-n.2: Evolution of downstream pressure increase rate (red markers) and gas
permeability (blue markers) measured on upstream side vs. applied upstream gas pressure

Repeatability of GBP
Samples S3-n.1, S3-n.3 and S3-n.5 have been subjected to several water-saturation
phases followed by a GBP test (Table 1). It is observed that GBP values are not
repeatable, with greater or lower values after a first GBP test. For instance, for sample
S3-n.1, GBP is of 1.6 MPa after a first water saturation phase, whereas it is at a lower
1.45 MPa value after a second water saturation phase. For sample S3-n.5, GBP is of 2.2
MPa after the first water saturation phase, whereas it is at a greater 2.9-3MPa value after
the second and third water saturation phases. This is attributed to different gas pathways
from one GBP test to the other, in relation with statistically-varying capillary digitation.

Figure 3 : Sample S3-n.5: (left): Helium and argon concentrations (in %) vs. after 24 h at Pupstream=3.7 MPa
and downstream valve opening (at 10 sec); (right): Argon and helium concentrations (in %) vs. time, when
the downstream valve has been kept closed for 24 h at a value of 3.555 MPa down to 3.504 MPa. The
downstream valve is open at time t=500 sec.

SCA2015-046

7/8

Effect of Sample Thickness and Damage State Upon GBP

For test series S3 (Table 1), GBP is below 3 MPa for 9.5-10.9 mm thick samples, below
5.8 MPa for 19.7 mm thick sample, and below 5.4 MPa for 30 mm thickness. Let us
compare these results with existing data. GBP results on so-called undisturbed COx
claystone have been obtained by BGS (UK) and Laego (France) on 55-85 mm long
samples machined by turning, at values ranging between 5.5 and 6 MPa [4]. For 18 mm
long COx samples, studies at RWTH (Germany) measure GBP values at 4.8-5.3 MPa,
which are also attributed to undisturbed matter [4]. Results by [10] on macro-cracked
claystone range between 0.8-2 MPa for 75-103 mm thickness. This means that all the
values of our test series S3, and those of test series S1 and S2 (given in [5]) correspond to
initially macro-cracked claystone. Higher GBP values observed for thicker samples are
attributed to too quick an increase in Pupstream, which does not allow for sufficient time to
record gas breakthrough, rather than to actual unability for gas to breakthrough.
Evidence of the Nature of Discontinuous Gas Passage
After a first GBP test, sample S3-n.5 has been re-saturated with water until 16x10-21 m2
water permeability, and further injected with the mix of gases (argon+helium). At
Pupstream=2.96 MPa, the presence of helium is recorded as a peak concentration value upon
downstream valve opening, which decreases down to zero after several minutes: it is a
first evidence of discontinuous gas passage. Argon concentration is too noisy to show any
significant concentration difference upon valve opening. In order to allow for shorter
experimental duration, Pupstream is then increased slightly, up to 3.7 MPa and kept at that
value for 24 h. Fig. 3a shows argon and helium concentrations vs. time after valve
opening (at 10 sec): a peak in concentration is observed for both gases upon valve
opening, with a regular decrease down to zero for helium after 5500 sec (i.e. 1 h 31 min),
and with a highly fluctuating decrease for argon, down to ca. 2.8%. After 8 h at Pupstream
=3.7 MPa with the downstream valve kept open, the mass spectrometer records a peak in
helium concentration, which decreases down to zero. This peak in helium concentration
is followed by other evenly distributed peaks of varying amplitude, which frequency
increases with time. These peaks are attributed to progressive gas passage by a
discontinuous phenomenon, which cannot be attributed to progressive dissolution and
diffusion through the claystone pore network: no physical reason justifies discontinuity
due to dissolution and diffusion (continuous phenomena). Rather, it is the demonstration
of capillary snap off, i.e. of gas passage by a discontinuous progression through pores of
sufficiently varying size (gas is regularly blocked in its progression through the pore
network by smaller pore throats, which requires an increase in gas pressure to be
overcome [11]. Further to this, Pupstream is kept at its value of 3.5 MPa +/- 0.05 (no
increase back to 3.7 MPa is performed here), and the downstream valve is closed for 24
h. Upon re-opening (Fig. 3b), helium passes again in a discontinuous manner, with
several peaks in concentration, which decrease progressively with time, yet not back to
zero due to a higher frequency of occurrence. After a total of 64 h 19 min at Pupstream =3.7
-> 3.5 MPa, helium finally passes continuously through to the downstream sample side,
together with argon (Fig. 3a) again. It is a continuous capillary gas breakthrough.

SCA2015-046

8/8

CONCLUSION
This contribution has shown that gas migration through initially water-saturated and
macro-cracked COx claystone occurs consistently by discontinuous capillary snap off,
followed by continuous passage (i.e. percolation) (answer to the first question in the
introduction). Discontinuous passage is not related to dissolution and diffusion of gas
through to the downstream chamber, which is opened every 24 h (answer to question 2)
in the introduction). GBP values are significantly below those of undisturbed claystone,
which is evidence of its inability to self-heal, despite excellent self-sealing ability
(answer to 3rd question in the introduction).

ACKNOWLEDGEMENTS
The authors are grateful to Andra for financial support of this research.

REFERENCES
1. R. De La Vaissire, G. Armand, J. Talandier, Journal of Hydrology, 521 (2015), 141156.
2. C. A. Davy, F. Skoczylas, J.-D. Barnichon, P. Lebon, Physics and Chemistry of the
Earth, pp. 667-680 Vol. 32, Issue 8-14, 2007.
3. S. T. Horseman, J. F. Harrington, P. Sellin, Engineering Geology, 54 (1999), 139149.
4. G. Duveau, S. M'Jahad, C. A. Davy, F. Skoczylas, J.-F. Shao, J. Talandier, S. Granet,
Proceedings of the international conference 45th US Rock Mechanics Geomechanics
Symposium ARMA San Francisco, USA, 26-29th, June 2011.
5. C. A. Davy, S. M'Jahad, F. Skoczylas, J. Talandier, M. Ghayaza, 5th International
Meeting on Clays in Natural and Engineered Barriers for Radioactive Waste
Management, October 22-25, 2012.
6. P. Egermann, J.-M. Lombard and P. Bretonnier. paper SCA A46, 2006 SCA
International Symposium, Trondheim, Sept. 18-22, 2006.
7. B. M. Krooss, A. Amann-Hildenbrand and A. Ghanizadeh. Marine and Petroleum
Geology, 31/1 (2012): 90-99.
8. Marschall, P., Horseman, S., Gimmi, T., 2005. Oil Gas Sci. Technol. 60 (1), 121-139.
9. P. Marschall and I. Lunati editors, Technical Report 03-11, Grimsel Test Site
Investigation Phase V, GAM, December 2006.
10. C.-L. Zhang, T. Rothfuchs, Physics and Chemistry of the Earth, 33 (2008) S363S373.
11. W. R. Rossen. Colloids and Surfaces A, 166 (2000), 101-107.

SCA2015-047

1/5

ROCK PROPERTY CHANGES FOR CARBONATE

RESERVOIR DURING CO2 INJECTION
M. Shimokawara, S. Takahashi, F. Kono, A. Kato, K. Tsushima
JOGMEC
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in Newfoundland and Labrador, Canada, 16-20 August, 2015

ABSTRACT
Carbonate minerals can be dissolved when carbon dioxide CO2, in presence of aqueous
solution, is injected into carbonate rocks. It is therefore known that porosity increases by
dissolution as well as porosity decreases by compaction. Porosity changes by dissolution
and compaction are one of the most important factors in order to understand fluid flow
characteristics and monitor CO2 invasions. Especially, we need to consider that the
change of rock properties is affected during CO2 injection into the carbonate reservoirs,
when we evaluate seismic data. However it has not been clarified quantitatively.
In this study, we conducted core flooding tests with carbonate rock samples under the
cyclic load test to clarify the effects of porosity changes. Carbonate rock samples were
collected from a Middle East reservoir. In the core flooding test, CO2 saturated water was
injected into the rock samples for 200 Pore Volume Injected (PVI). Porosity,
permeability, and elastic wave velocities were measured every 50 PVI. Porosity was
measured using conventional Helium porosimetry and NMR methods. The amount of the
dissolved carbonate minerals that is related to porosity increase was evaluated based on
the amount of the dissolved cations in the CO2 saturated water. In addition, the change of
carbonate surfaces was also observed with SEM during CO2 saturated water flooding.
Results show that carbonate dissolution occurred at the early stage of CO2 saturated water
flooding. Porosity decreased under compaction when rock frame was weakened by
dissolution. The elastic wave velocities reduce with or without compaction effects when
carbonate minerals are dissolved by CO2 saturated water flooding.

INTRODUCTION
In the field operation of CO2-EOR as well as carbon capture and storage (CCS), it is
important to accurately understand the interaction between rock and fluids in addition to
the behavior of the injected CO2. In crosshole seismic data acquired in the field, cases of
significantly decrease of seismic velocity in rocks has been reported after CO2 was
injected [4]. These velocity changes were substantially higher than those due to fluid
substitution, as predicted by the Gassmann theory; consequently, it has been argued that
these changes were caused by dissolution of carbonate rock minerals.

SCA2015-047

2/5

It is also known that carbonate minerals are dissolved due to CO2. The effects of mineral
dissolution on the rock properties have been investigated by laboratory measurements,
and the results have been reported. Vanorio et al. [3] and Kono et al. [2] conducted
experiments on carbonate rock samples that were injected with CO2. Changes in physical
properties such as acoustic velocity and porosity were assessed and changes in rock
microstructures were observed with SEM images. The results showed that micritic
carbonate minerals were dissolved with aqueous solutions of CO2, leading to increased
porosity and decreased acoustic velocity. Grombacher et al. [1] also conducted integrated
laboratory measurements including NMR, SEM and CT-scan for carbonate rock samples
injected with CO2. They reported changes in acoustic velocity, permeability, and
microstructures.
This paper presents the results of core flooding tests of carbonate rock samples under the
cyclic load test to understand the changes of the rock properties and wave velocities
during CO2 injection.

PROCEDURE
1.5-inch-diameter plug samples were used for this study. Samples were carbonate rocks
collected from a reservoir in a Middle Eastern oil field. Properties of each plug sample
are shown in Table 1.
The laboratory measurement workflow is shown in Figure 1. CO2 saturated water was
prepared by bubbling CO2 in water. The pH (potential hydrogen) of the distilled water
decreased from 6.5 to 3.9. The rock samples were placed inside a pressure vessel to be
arranged in series, and the CO2 saturated water was allowed to be flooded through them.
CO2 saturated water was injected at the rate of 0.1 ml/min. For each 50 PVI flooded, the
samples were removed from the pressure vessel. NMR analysis was conducted for the
samples fully-saturated with CO2 saturated water. After drying and basic measurements,
elastic wave velocity measurement was conducted. Then, the samples were returned to
the pressure vessel, and the experiment procedure was repeated up to 100 PVI. In
addition to NMR and elastic wave velocity measurements, observations of carbonate
mineral dissolution were performed with a scanning electron microscope (SEM:
QUANTA600), and analyses of the amount of dissolved cations in the effluent were also
conducted. The data results were used for evaluating effects of dissolution due to CO2
saturated water on rock properties.

RESULTS
Figure 2 shows the change of the ratio of the pore volume to the initial pore volume
(PV/PV0) during the CO2 saturated water flooding. The porosity of Core_B without the
compaction effects increased by 3.6% at the first 50 PVI, and the porosity increased by
6.0% from 50 to 100 PVIs. On the other hand, the porosity of Core_A with the
compaction effects decreased by 4.3% at the first 50 PVI, and then the porosity was kept.
At the end of 100 PVI, the difference of porosity between Core_A and Core_B was about
8%.

SCA2015-047

3/5

Figures 3 and 4 show the change of normalized elastic wave velocities when CO2
saturated water was injected. The normalized P-wave velocities decreased by 10%
compared to the initial values in the both samples at the first 50 PVI and were constant
from 50 to 100 PVIs. The normalized S-wave velocities of Core_B also decreased by
10% as the same as the normalized P-wave velocities whereas the normalized S-wave
velocities of Core_A decreased by 5% at 100 PVI of the CO2 saturated water. Dissolution
of carbonate minerals was measured by produced water analysis. The amount of
dissolution was confirmed to be substantially equal to the change of porosity.

DISCUSSION
In the case of Core_B without the compaction effects, the change of porosity at the first
50 PVI was greater than after 50 PVI. So the dissolution of the carbonate minerals by
CO2 saturated water flooding occurred at the early stage of flooding. On the other hand,
Core_A with the compaction effects showed decrease in porosity. This suggested that the
decrease was caused by compaction load because rock frame was weakened by
dissolution.
The elastic velocities of Core_B without the compaction effects decreased when the bulk
density decreased with porosity increase. Core_A with the compaction effects also
showed elastic velocities decrease. This indicates that the elastic wave velocities reduce
with or without compaction effects when carbonate minerals were dissolved by CO2
saturated water flooding.

CONCLUSION
We conducted CO2 saturated water flooding test for carbonate and evaluated the change
of rock properties. Results showed that carbonate dissolution occurred at the early stage
of CO2 saturated water flooding. It is suggested that the porosity decreased under
compaction because rock frame was weakened by dissolution. In addition, results showed
that the elastic wave velocities reduce with or without compaction effects when carbonate
minerals are dissolved by CO2 saturated water flooding.

ACKNOWLEDGEMENTS
The authors were favored to have the assistance of Yasuyuki Akita, Emiko Shinbo and
Akira Hanyu who contributed their experimental skill, sustained effort, and grasp of
objectives to the accomplishment of the experimental program.

SCA2015-047

4/5

REFERENCES
1. Grombacher, D., T. Vanorio, and Y. Ebert, 2012, Time-lapse acoustic, transport, and
NMR measurements to characterize microstructural changes of carbonate rocks
during injection of CO2-rich water, Geophysics, 77, no. 3, WA169-WA179.
2. F. Kono et al., 2014, Laboratory Measurements on Changes in Carbonate Rock
Properties due to CO2-saturated Water Injection, SPE172013.
3. Vanorio, T., A. Nur, and Y. Ebert, 2011, Rock physics analysis and time-lapse rock
imaging of geochemical effects due to the injection of CO2 into reservoir rocks,
Geophysics, 76, no. 5, O23O33.
4. Wang, Z., M. Cates, and R. Langan, 1998, Seismic monitoring of a CO2 flood in a
carbonate reservoir: A rock physics study, Geophysics, 63, 16041617.
Table1 The properties of rock core samples.
Sample ID
Core_A
Core_B

Diameter

Length

Dry weight

cm
3.804
3.790

cm
4.076
6.154

g
89.92
147.20

Pore
volume
ml
12.56
15.34

Grain
density
g/cm3
2.66
2.72

Porosity

Kair

fraction
0.271
0.221

mD
20.9
28.4

With
compaction
yes
none

1.06

1.04

PV/PV0, frac.

1.02
The difference of
compaction effects
1

0.98

0.96
Core_A
Core_B
0.94
0

50

100

150

PVI, frac

Figure.1 Laboratory measurement flow.

Figure.2 The comparison of PV/PV0 between with

compaction effects.

SCA2015-047

5/5

1.02

1.02
1

Core_A

Core_B

0.98

Core_A
Core_B

0.98
0.96

Vs/Vs0, frac.

Vp/Vp0, frac.

0.96

0.94
0.92

0.94
0.92
0.9
0.88

0.9
0.86
0.88
0

50

100
PVI, frac.

Figure.3 Normalized P-wave velocities.

150

0.84
0

50

100
PVI, frac

Figure.4 Normalized V-wave velocities.

150

SCA2015-048

1/7

TWO SCALE PORE ANALYSIS APPLIED TO LOW

PERMEABILITY SANDSTONE
C. A. Davy1, P.M. Adler2, Yang Song1, L. Jeannin 3, D. Troadec 4, G. Hauss 1, Thang
Nguyen Kim2
1
LML UMR CNRS 8107/Ecole Centrale de Lille, CS20048,
59651 Villeneuve dAscq Cedex, France
2
Metis/UPMC, Paris, France
3
GDFSUEZ E&P International SA, 1 place Samuel de Champlain,
92930 Paris La Dfense cedex - France
4
IEMN, UMR CNRS 8520, BP60069, 59652 Villeneuve dAscq Cedex, France
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
The purpose of this research is 1) to predict fluid transport (permeability) from pore
structure measurements of a tight gas sandstone and 2) to compare predictions to actual
permeability measurements on centimetric plugs. This contributes to explain what is (or
are) the relevant pore scale(s) for fluid transport, and should help improve gas production.
Tight sandstones are low porosity media, with small fluid transport ability. The sample T2390-82 used in this contribution has a water porosity of 2.8-3.2%, and a dry gas
permeability of the order of 10-17 m2 (10 microDarcy).
The pore size distribution of T-2390-82 is measured by Mercury Intrusion Porosimetry
(M.I.P.) with a peak at 350 nm. This justifies the use of Focused Ion Beam/Scanning
Electron Microscope (FIB/SEM) to describe the 3D pore structure. FIB/SEM provides
the 3D geometry of a single joint (between quartz grains). However, 2D SEM shows the
presence of numerous joints of both sub-micrometric and micrometric sizes; 2D SEM is
used to describe the joint network topology (density, joint length, etc.). X-ray microtomography (micro-CT) is used in an attempt to describe a representative pore network in
3D. If a denotes the voxel size, no connected pore network is measured with a = 13
micron, and it is hardly connected with a = 4.4 micron. A voxel size of 600 nm is
necessary to image micrometric joints, yet only individual ones.
In order to predict fluid transport, 2D SEM is used to describe the joint network topology
and macroscopic permeability, combined to either (1): micro-CT or (2): FIB/SEM for the
individual joint transmissivity. At the joint network scale and at the 3D joint scale, twoscale numerical modelling shows that micrometric joints (observed with micro-CT)
control macroscopic permeability, although they have a partial volume contribution to
porosity, according to MIP. However, sub-micrometric joints (imaged by FIB/SEM)
which contribute significantly to porosity (up to 1.27%) are slow paths for fluid transport

SCA2015-048

2/7

(with permeabilities smaller by two to three orders of magnitude, when compared to that
given by micrometric joints).

INTRODUCTION
Low permeability materials are often composed of several pore structures of various
scales, which are superposed one to another. It is often impossible to measure and to
determine the pore geometry properties relevant for fluid transport in one step, or with a
unique experimental device.
In low permeability sandstones, the pore space is essentially made of micro-cracks
between grains. These macro-cracks (or fissures) are two dimensional structures, which
aperture is roughly on the order of one micron. When considering the grain scale, i.e., on
the millimetric scale, fissures form a network.
These two structures (individual crack and fissure network) are measured by using
different tools [1]. The density of the fissure network is estimated by trace measurements
on the two dimensional images provided by classical 2D Scanning Electron Microscopy
(SEM) with a pixel size of about 2 micron. The three dimensional geometry of fissures is
measured by both (1) Focused Ion Beam/Scanning Electron Microscopy (FIB/SEM) with
voxel size a = 15 nm and (2) X-Ray micro-tomography (micro-CT) in the laboratory
(www. http://isis4d.univ-lille1.fr), with a = 0.6 micron, or 4.37 micron or 13 micron.
A two-step methodology is proposed to predict the transport properties of tight
sandstones. It is applied on the example of tight sandstone T-2390-82. Transport is
characterised on two different scales; the pore network, representative of the macroscopic
scale, is imaged by 2D SEM in terms of crack organisation, average length, density, etc.;
the aperture b of the individual cracks is characterized on the scale given by FIB/SEM for
sub-micrometric joints, and by micro-CT for micrometric joints. Numerical calculations
are performed first, for the whole pore network (given by 2D SEM), as a function of
network characteristics; they yield a permeability proportional to the fracture
transmissivity. This individual crack transmissivity is determined by using the geometry
of the individual cracks (given by FIB/SEM and micro-CT).

EXPERIMENTAL
Experiments were performed on a single sample (37.7 mm diameter and 34.6 mm height)
of tight sandstone reference T-2390-82, after oven-drying at 105C until mass
stabilization (dry state).
Macroscopic Dry Gas Permeability. Details of the experimental method are given in
[3]. The 37.7 mm diameter sample is subjected to a quasi-static argon gas flow, at an
upstream gas injection pressure of 2 to 4 MPa. The apparent gas permeability Kgas is
measured during a small decrease in upstream gas pressure by 0.05 MPa, during two
successive cycles of external hydrostatic stress loading up to 45 MPa. For such important
gas injection pressures, the Klinkenberg effect is generally negligible [4]. After these

SCA2015-048

3/7

cycles, the sample is over-cored transversely to a sample of 20mm diameter and 30.8 mm
height, and tested for gas permeability again.
Fig. 1 shows that gas permeability Kgas decreases significantly with increasing hydrostatic
stress Pc; this is attributed to the progressive closure of fissures under loading. For further
permeability predictions, an average Kgas of 2.1+/-0.4 10-17 m2 (21 +/-4 micro Darcy) is
considered at low Pc=2.5 to 5MPa.

Figure 1: Gas permeability for 37.7mm diameter and

perpendicular 20mm diameter T-2390-82 sandstone
sample

Figure 3: Original 2D SEM image of sample T2390-82.

Figure 2: MIP results for T-2390-82 sandstone, plotted

after blank cell and conformance corrections [5,6]

Figure 4: Corresponding trace map of the SEM image

(left) used for permeability prediction

Indirect Pore Structure Assessment. Porosity is measured by the water imbibition

method on the 37.7 mm sample and on its 20 mm over-cored part; values of 2.8-3.2% are
obtained.
Mercury Intrusion Porosimetry is performed using a MICROMERITICS AutoPore IV
9500 up to 200 MPa, which corresponds to intruded pores down to 4 nm. Results (Fig. 2)
show a characteristic pore size of about 350 nm, with a limited contribution of pores

SCA2015-048

4/7

between 1-10 microns to cumulative pore size distribution. This justifies the use of
FIB/SEM imaging to describe sub-micrometric pores in 3D.
Direct Characterization of the Pore Structure: 2D SEM. The pore network cannot be
characterized by FIB/SEM alone. 2D SEM measures directly the topology (average
length, density, connectivity) of the fissure network of T-2390-82 sandstone. SEM
images are taken at low magnification (x40-45), with a pixel size of 2.15-2.33 micron, on
the top surface of the 37.7 mm diameter sample, where several zones of different fissure
densities are observed. A high fissure density zone and a medium density zone are
computed in the following. Therefore, a trace map of each individual SEM image is made
by manually locating each individual fissure (Figs. 3 and 4).

Figure 5: Pore network given by FIB/SEM (3D

reconstruction with Amira software, FEI)

Figure 6: Pore network given by micro-CT on a 0.8mm

rod (3D reconstruction with Amira software, FEI)

Direct Characterization of the Pore Structure: FIB/SEM. The focused ion beam (FIB)
cuts a U-shaped hole in order to isolate a plane-parallel sandstone volume [7]. Following
this, the FIB cuts regularly-spaced 50nm thick slices from the plane-parallel volume,
perpendicularly to the sample polished surface. Between each FIB cutting, the sandstone
matter perpendicular to the polished surface is observed with an electron detector of the
in lens type. A stack of 300 images is computed. Sample T-2390-82 has a voxel size of
14.65x35.71x50.0 nm3, and a total investigated volume of (9874x20069x15000) nm3. The
3D pore network is obtained from the raw FIB/SEM image stack, by filtering and
segmentation of each individual image using the ImageJ software. The 3D object is
generated by using the Amira (FEI) software, by using the method described in [8] (Fig.
5). The 2D and 3D Continuous Pore Size Distributions (CPSD) are deduced form the
binary images, as described by [9].
Direct Characterization of the Pore Structure: micro-CT. For comparison purposes,
three image series of the same T-2390-82 sample are acquired with three voxel sizes of
13 microns (measured on the 20 mm diameter sample used for gas permeability), 4.4

SCA2015-048

5/7

microns (on a parallelepipedic sample of 5 mm width) and 0.6 microns (on a thin rod of
0.8mm width). Image filtering and segmentation by Image J and 3D visualization by
Amira (Fig. 6) are performed as for FIB/SEM.
Comparison between the Various Pore Structure Assessment Methods. Indirect and
direct methods for pore structure assessment are compared in Table 1. For a = 13 micron
and 4.4 micron, pore volumes of micro-CT samples are not connected from one surface
to another, despite porosities of 1.016% +/-0.36 and 1.88% +/-0.17. Compared to water
porosity (2.8-3.2%), M.I.P. provides a greater value, possibly due to sample damage. On
the opposite, micro-CT of the 0.8mm rod provides a connected pore volume (Fig. 6),
corresponding to an individual fissure between two quartz grains, of porosity 1.34% +/0.15 and average opening of 4 micron. Although significantly smaller, the FIB/SEM
sample provides a porosity of 1.25%, which is similar to micro-CT, and a typical opening
on the order of 60-100nm. At this stage, it is difficult to determine whether or not both
3D fissures (given by micro-CT and FIB/SEM) contribute to fluid transport.

Table 1: Main characteristics of the pore structure of the same T-2390-82 sandstone sample.

PERMEABILITY PREDICTION
Macroscopic permeability K is calculated in two steps. On the small scale, the fracture
transmissivity is calculated by solving the Stokes equation on several portions of the
measured connected fissures by micro-CT and FIB/SEM. A single phase Lattice
Boltzmann code is used; according to the standard terminology, it is a D3Q19 code with
two relaxation times; the classical bounce-back condition is used at the solid interface.
Each sample is completed by its mirror image in order to avoid overall boundary effects.
More details can be found in [10].
On the large scale, the density of fissures is estimated by three different means based on
the number of intersections with scanlines, on the surface density of fissures and on the
intersections between fissures per unit surface. These three means show that the network
is relatively isotropic. They provide very close estimations of the density. Then, a general
formula derived from systematic numerical computations [2] is used to derive the
macroscopic dimensionless permeability which is proportional to the fracture
transmissivity K= 2/[R(1+ )] where =0.037, =- R is a

SCA2015-048

6/7

measure of the lateral fracture extension. is the average number of intersections per
fracture. A closely related application is given in [11].
The combination of the two previous results yields the dimensional macroscopic
permeability. For the combination of 2D SEM and FIB/SEM, permeability predictions K
range between 3 10-21- 6.2 10-20 m2 (3 to 62 nanoDarcy). This is three to four orders of
magnitude smaller than actual gas permeability measurements. For the combination of
2D SEM and micro-CT, permeability is predicted at values ranging between 6 10-18 - 7.4
10-16 m2 (6 to 740 microDarcy). This is in better agreement with experimental gas
permeability on centimetric samples.

CONCLUSION
For tight sandstone T-2390-82, the relevant pore structure for fluid transport is well
described on two separate scales: (1) the fissure network is imaged and characterized by
2D SEM, (2) the individual fissure between quartz grains is characterized at a 350 nm
size by M.I.P., which is the same order of magnitude as the 100nm opening given by
FIB/SEM. This fissure scale contributes by three-four orders of magnitude less to
transport than the micrometric fissure aperture obtained by micro-CT (of 4 microns
wide). The latter provides macroscopic permeability prediction consistent with gas
permeability identification on centimetric samples. This original approach needs to be
validated on other tight sandstones to prove its generality.

ACKNOWLEDGEMENTS
Financial support by GDFSUEZ E&P International SA is gratefully acknowledged. The
SEM national facility available at the Centre Commun de Microscopie de Lille (France)
is supported by the Conseil Regional du Nord-Pas de Calais, and by the European
Regional Development Fund (ERDF).

REFERENCES
1. Z. Duan, C. A. Davy, F. Agostini, L. Jeannin, D. Troadec, F. Skoczylas, International
Journal of Rock Mechanics and Mining Science, Vol.65, pp.75-85, 2014.
2. P.M. Adler, J.-F. Thovert, V.V. Mourzenko: Fractured porous media, Oxford
University Press, 2012.
3. X. T. Chen, Th. Rougelot, C.A. Davy, W. Chen, F. Agostini, F. Skoczylas and X.
Bourbon, Cement and Concrete Research, Vol.39 (12), pp. 1139-1148, 2009.
4. W. Chen, J. Liu, F. Brue, F. Skoczylas, C. A. Davy, X. Bourbon, J. Talandier, Cement
and Concrete Research, Vol.42, pp. 1001-1013, 2012.
5. Shafer J. and Neasham J. Mercury Porosimetry International Symposium of the Society
of Core Analysts, 2000, SCA 2021.
6. Dewhurst D. N., Jones R. M. and Raven M. D. Petroleum Geoscience, 2002;8:371
383.
7. Holzer, L., Indutnyi, F., Gasser, P., Munch, B., and Wegmann, M. Journal of
Microscopy, 216(1):84-95 (2004).

SCA2015-048

7/7

8. Y. Song, C. A. Davy, D. Troadec, A.-M. Blanchenet, F. Skoczylas, J. Talandier, J.-C.

Robinet, Marine and Petroleum Engineering, 65 (2015) 63-82.
9. Munch, B. and Holzer, L. (2008). Journal of the American Ceramic Society,
91(12):40594067.
10. A. Pazdniakou, P.M. Adler, Dynamic permeability of porous media by the lattice
Boltzmann method, Advances in Water Resources, 62, 292, 2013.
11. I. Malinouskaya, J.-F. Thovert, V.V. Mourzenko, P.M. Adler, R. Shekhar, S. Agar, E.
Rosero, M. Tsenn, Fracture analysis in the Amellago outcrop and permeability
predictions, Petroleum Geoscience, 20, 93, 2014.

SCA2015-049

1/7

NUMERICAL MODELING OF NMR SURFACE

RELAXATION IN POROUS MEDIA WITH IMPROVED
PORE SURFACE AREA EVALUATION AND DERIVATION
OF INTERFACIAL ABSORPTION PROBABILITY
Danyong Li1, Xuefeng Liu2, Feng Huang3, Tianpeng Zhao1, Weifeng Lv4, Sven Roth1
1
iRock Technologies, Beijing, China; 2 China University of Petroleum, Tsingtao,
Shandong, China; 3 Wuhan University of Technology, Wuhan, Hubei, China; 4 Research
Institute of Petroleum Exploration and Development, Beijing, China
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
In digital rock physics, surface relaxation of NMR decay on a 3D porous media is often
simulated with Random-Walker methods, where Brownian walkers representing the
magnetized nuclei are absorbed by the solid surface under an interfacial absorption
probability. Previous simulations usually neglected the blocky pixel appearance of 3D
pore space images, which results in over estimation of the surface area, particularly for
unconventional reservoir rocks with large fractions of micropores, such as tight
sandstones or shale. This leads to arbitrary adjustments of the interfacial absorption
probability by tuning the value of the surface relaxivity strength (SRS) in order to match
simulation results with experimental measurements. However, this arbitrary adjustment
violates the fact that SRS is a rock dependent physical property. In this study a new
interfacial absorption probability for NMR simulations is presented that honours the
physical principles. This new absorption probability computation depends on accurate
evaluation of the surface area, which is assured by a new approach of 3D pore space
surface area evaluation. The new algorithm is verified by performing NMR
Random-Walker simulation tests on both standard geometries and realistic 3D
heterogeneous porous media.

INTRODUCTION
In Random-Walk NMR simulation the decrease of random walkers, which simulates the

SCA2015-049

1/7

transverse decay of magnetization, is controlled by an absorption probability of the solid

grain surface, usually called the killing probability, which is proportional to the surface
relaxivity, the random walker step size, and is reversely proportional to the bulk
diffusivity [1]. Traditionally, the expression of killing probability derived by Bergman
and Dunn [1] was used as an input to the simulation program [2-4]. However, this
expression of killing probability, based on the assumption of random walkers hitting a
planar surface was neglecting the discretized voxel effect and thus, required unphysical
adjustment of input parameters. This effect is increasingly evident in unconventional rock
types, with high surface-to-volume ratio (S/V).
In this study two aspects are discussed to reduce the uncertainties from the previously
used killing probability approaches. Firstly, an updated killing probability is derived
based on the physical process of transverse magnetization decay. The new killing
probability effectively includes the surface-to-volume ratio and the global hitting
probability of the entire pore space geometry and is adaptable to any rock type. Secondly,
an algorithm is proposed, which more accurately evaluates the surface area and reduces
the voxel discretization effect. The new NMR simulation is verified by testing on both
standard geometrical pore spaces and digital images of realistic rock samples of various
rock types and shows that the previously shifted T2 distribution is successfully corrected.

THEORY AND METHODS

Nuclear magnetic resonance occurs when an ensemble of spinning hydrogen protons
performing Larmor precession in a uniform magnetic field are tilted by an
oscillating magnetic field pulse () with oscillation frequency the same as the
Larmor precession frequency and perpendicular to . The net macroscopic
magnetization previously along the direction of is tilted away from its
longitudinal direction to the transverse plane. As the magnetic pulse is turned off, the
tilted magnetization relaxes back to its longitudinal direction and the transverse
component of macroscopic magnetization follows an exponential decay [5],
() = (0) /2

(1)

where 2 is determined by three effects, bulk relaxation, surface relaxation and bulk
relaxation, which is defined as the combination of [6]
1
2

1
2

1
2

1
2

(2)

It is well understood that in pore space, where the scale satisfies the fast diffusion limit,

SCA2015-049

1/7

the surface relaxation time 2 is related to the mineralogy on the pore solid
surface and the surface-to-volume ratio of the pore space [7],
1
2

(3)

where is the surface relaxivity, the pore surface area and the pore space volume.
In this study, the surface relaxation described by Eq. (1) is simulated by Random-Walk
methods, in which an ensemble of random walkers representing the spinning hydrogen
protons is seeded into porous space to imitate diffusive Brownian motion [1]. The
random walkers travelling to the image pixels near to the solid-void boundary are
absorbed by the solid interface, leading to the decrease of random walkers, analogous to
the surface relaxation process in which the spinning nuclei is relaxed by interaction with
the solid grain surface.

Figure 1. Scheme of Random-Walk NMR simulation on MCT image: (a) a typical MCT greyscale image
with lighter area indicating the high density grains and darker area the pores. (b) segmented pore space
extracted from the greyscale image. (c) zoom-in of a part of the pore space shows the pore space is
compiled by cubic voxels. (d) random walkers (blue) seeded into the pore space. (e) the track of one
random walker in one pore space.

Initially, a number of random walkers are uniformly seeded into a 100% wetting
phase saturated pore space. After an infinitesimal time , the number of random
walkers decreases by , where is the global probability of random walkers hitting
the pore surface and the probability of random walkers being absorbed by the surface.
Relating the decrease of random walkers with the decay of magnetization in Eq. (1), we
obtain
=

(4)

In our simulation the time step is determined by the law of Brownian motion of

SCA2015-049

1/7

spatial diffusive particles in a 3D space, 2 = 60 , where 0 is the bulk fluid

diffusivity and the diffusion distance of random walkers within a time step. Therefore
Eq. (4) can be rewritten as the expression for killing probability,
2

= 6

(5)

Effective implementation of Eq. (5) into the whole process of NMR simulation requires
both accurate evaluation on the surface area and the global hitting probability of
the pore space. In this study, an optimized marching cubes algorithm [8] is employed to
evaluate the pore space surface area by interpolating a virtual smoothed surface along the
discretized solid-pore interfaces and random walkers hitting the surface are determined
by judging whether the random walkers cross this virtual surface. The hitting probability
depends on the structural geometry of the pore space and is calculated as

6
8
12
=
=1(1 )( =1 + =1 + =1 )

(6)

where is the pixel registration of the ith pixel of a 3D image with = 0 for void
and = 1 for solid grain. ( or ) are the 6 (8 or 12) sided (corner or edge)
neighboring pixels of , and , and are the constant probabilities of random
walkers in pixel jumping into sided, corner and edge pixels, respectively.

RESULTS
Standard Geometrical Benchmarks
The validity of the NMR simulation program with updated killing probability and
smoothed surface area evaluation is first verified by performing tests on standard
geometries and comparing the simulated results with the theoretical ones. The
consistency between simulation and theoretical results ensures the reliability of
performing the simulation on realistic pore spaces, where the porous structures are highly
irregular and the pore sizes are multiple-scale.
The simulation results (Table 1) show significant improvement of pore space surface area
evaluation and better consistency between the predicted and theoretical T2s than the
previous version of NMR simulations. It is obvious that while the previous program
works well on cubic pore space consisting of planar surfaces, it fails to correctly evaluate
the surface area of spherical and irregular pores which are dominant in most rock
samples.
Application on Digital Rock Images

SCA2015-049

1/7

The new simulation is further tested on realistic digital images of various rock types
including sandstone, carbonate and tight sandstone (Table 2). The validity of NMR
simulation on realistic rock samples can only be verified by matching with experimental
measurements. However, the improvement of the simulation algorithm can be verified by
comparing the simulation results with the previous ones. Figure 2 shows one testing
result, including decaying magnetization and its corresponding T2 distribution. With
more accurate killing probability realized by corrected evaluation of the pore space
surface area and more physical based random walker diffusion scheme, the T2
distribution of all rock types shift rightwards, which indicates that the previous over
estimated surface area is effectively corrected. The level of correction increases as the
pore space becomes more tight and tortuous.
Table 1. Simulation results on standard geometrical pore space shows that the updated simulation algorithm
significantly improves evaluation of surface area and T2 for spherical pores. The testing are perform on a
2003 image with pixel size 1.0m and surface relaxivity 3.0m /s.

Cubic Pore
Length(m)

Sth

(m2)

T2th(ms)

Sold(m2)

T2old(ms)

Snew(m2)

T2new(ms)

10.0

600.0

555.6

600.0

543.3

564.1

605.5

20.0

2400.0

1111.1

2400.0

1117.7

2329.0

1174.6

50.0

15000.0

2777.8

15000.0

2802.9

14820.0

2923.1

100.0

60000.0

5555.6

60000.0

5672.9

59650.0

5906.9

Spherical Pore
Radius(m)

Sth(m2)

T2th(ms)

Sold(m2)

T2old(ms)

Snew(m2)

T2new(ms)

5.0

314.2

555.6

486.0

424.0

341.0

516.5

10.0

1256.6

1111.1

1902.0

937.3

1372.0

1047.9

25.0

7854.0

2777.8

11770.0

2380.8

8539.0

2654.9

50.0

31415.9

5555.6

47070.0

4827.9

34140.0

5332.8

Figure 2. Comparisons of previous and updated works of NMR simulation on digital images of a 2mm
diameter tight gas sandstone subsample.

SCA2015-049

1/7

Table 2. Simulation results of various rock types and scales. The surface relaxivity is taken as 4.0m /s.

More accurately simulated T2 distributions provide several important potential practical

applications. The SRS is largely unknown but is needed to convert the experimental T2
distribution from time scale to length scale. Another example is to provide accurate
porous surface area which is important for some surfactant experiments.

CONCLUSIONS
We propose two important updates to previous NMR simulations. One is a newly derived
killing probability, based on the physical process of an ensemble of random walkers
diffusing, hitting and being absorbed by a porous space solid surface, to replace the
previous killing probability which fails to recognize the global irregularity of the pore
shapes. The other is an optimized algorithm to increase the accuracy of porous space
surface area evaluation, which was over estimated, previously. Both successfully reduce
the uncertainties resulting from previous works. With the new killing probability and
smoothed surface area evaluation, the NMR simulation can be reliably matched with
experimental measurements and thus, provides a more accurate surface relaxivity
strength and surface-to-volume ratio. Based on these improvements, NMR simulations
can be extended to images with different saturation stages. Families of T2 distributions
under various saturations can be simulated to calibrate NMR well logging data and to
obtain more accurate estimates of irreducible water saturation.

REFERENCES
[1] Bergman, D.J., Dunn, K.J., Schwartz, L.M. and Mitra, P.P., Self-diffusion in a
periodic porous medium: A comparison of different approaches, Physical Review E 51,
3393 (1995).
[2] Talabi, O., Alsayari, S., Iglauer S. and Blunt, M.J., Pore-scale simulation of NMR
response, Journal of Petroleum Science and Engineering 67, 168-178 (2009).
[3] Toumelin, E., Torres-Verdin, C. and Chen, S., Modeling of Multiple Echo-Time
NMR Measurements for Complex Pore Geometries and Multiphase Saturations, SPE
Reservoir Evaluation & Engineering, 234 (2003).

SCA2015-049

1/7

[4] Toumelin, E., Torres-Verdin, C. Sun, B. and Dunn K.J., Random-walk technique for
simulating NMR measurements and 2D NMR maps of porous media with relaxing and
permeable boundaries, Journal of Magnetic Resonance, 188, 83-96 (2007).
[5] Bloch, F., Nuclear Induction, Physical Review 70, 460 (1946).
[6] Bloembergen, N., Purcell, E.M. and Pound, R.V., Relaxation Effects in Nuclear
Magnetic Resonance Absorption, Physical Review 73, 679 (1948).
[7] Brownstein, K.R. and Tarr, C.E., Importance of classical diffusion in NMR studies of
water in biological cells, Physical Review A 19, 2446 (1979).
[8] Lindblad, J. and Nystrom I., Surface Area Estimation of 3D Digitized Objects Using
Local Computations.

SCA2015-050

1/6

SYSTEMATIC PARAMETER STUDY FOR FORMATION

FACTOR MODELING AT THE PORE SCALE
Matthias Halisch (1), Sven Linden (2), Jens-Oliver Schwarz (2), Sarah Hupfer (1), Andreas Wiegmann (2)
(1)
Leibniz Institute for Applied Geophysics (LIAG), Stilleweg 2, D-30655 Hannover, Germany
(2)
Math2Market GmbH, Stiftsplatz 5, D-67655 Kaiserslautern, Germany

This paper was prepared for poster presentation at the International Symposium of the Society of
Core Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Digital Rock Physics (DRP), in combination with state of the art 3D imaging techniques, such as
high resolution micro CT (-CT) or Synchrotron CT (Sr-CT), has progressed to a well
established methodical instrument for petrophysical process modeling and simulations at the pore
scale, respectively. Besides hydraulic and acoustic (elastic) rock properties, which probably form
the largest field of interest for DRP, electric conductivity / resistivity is also of great interest since
it is essentially used to determine (e.g.) water saturation, formation producibility, and at least the
volume of oil- and/or gas in place of the formation. For this, formation resistivity factor (F) is
measured by default during classic core analysis in the laboratory, in order to evaluate the
parameters as mentioned above and to characterize the pore system of the rock in more detail.
Nevertheless, to provide reliable statistics, experimental time increases significantly, same as the
amount of cores that need to be investigated. This is where DRP and non-destructive 3D imaging
become relevant. Though the experiment as well as the analytical calculation of F is straight
forward, numerical modeling of this parameter by using CT obtained geometries is significantly
affected by different boundary conditions, such as:
a) The 3D scan resolution (voxel resolution);
b) The image processing / phase segmentation;
c) The ratio of mineral to water resistivity;
d) The overall computational domain size;
e) The micro-scale sample (formation) heterogeneity.
The amount of pores that can be detected is directly linked to the resolution of the imaging, since
F is mostly dominated by the porosity. False or inaccurate segmentation of phases can increase
this effect in addition. The parametric setup of the model greatly increases computational power
and time. Finally, the domain needs to be representative, which is especially challenging for
heterogeneous (e.g. layered) samples. In this systematic parameter study, the authors showcase a
selection of these effects by using data from three different rock samples (i. young sandstone; ii.
reservoir sandstone; iii. porous carbonate). The results are compared to experimental, i.e.
laboratory obtained data, and discussed critically in order to evaluate and develop better and
optimized numerical tools and workflows for this specific modeling purpose.

INTRODUCTION
The classical and empirical relationship between rocks and the saturation fluids is essential for
the interpretation of electrical borehole measurements since the early 1940s [1, 2]. The

SCA2015-050

2/6

fundamental proportionality between the resistivity of a water saturated rock (0) and the
resistivity of the pore fluid (w) led to the infamous and so called Archies law:
0 = F w

(E1)

The introduced parameter F, the formation resistivity factor, is a measure of the increase in
resistivity of a pore fluid by presence of a non-conductive rock matrix. Formation factor is
dependent on the geometrical (spatial) distribution of the electrolyte within the rock, related to
the internal structure of the rock, of the pore space and of the wettability conditions therein [3].
Hence, F mostly but not exclusively is assigned to the (effective) porosity of the investigated
rock. This finding led to the first Archie equation [2]:
F

1
,
m

(E2)

where m denotes the so called cementation exponent. Since porosity can be easily assessed by
modern 3D digital imaging techniques and evaluation software tools, reliable modeling results
are still challenging to derive, and strongly depend upon the boundary conditions as mentioned in
the introducing abstract. The toolbox GeoDict (developed and distributed by Math2Market
GmbH) is a generic approach for computer aided material engineering, and has been used for the
formation factor modeling. It contains geometric models for virtual material synthesis and
provides numerical tools for the prediction of material properties on the micro-scale. The basic
concept is the simulation of physical processes on digital objects that are derived from either CT images, or from synthesized images and models. This toolbox is able to predict a broad range
of mechanic, electric and hydraulic rock properties, including parameters of routine and special
core analysis [5, 6].
Explicit Jump Immersed Interface Method for Electrical Conduction
The Explicit-Jump immersed interface method is a very fast finite difference method to solve
partial differential equations [7, 8]. Here it is applied to solve the stationary electrical conduction,
heat conduction and diffusion equation in reservoir rocks:
( U) = f ,

(E3)

where U denotes the electric potential and > 0 the piecewise constant electric conductivity. The
equation is composed of Ohms and Kirchhoffs law. The stationary electrical conduction
equation is directly solved on segmented -CT images, where each individual phase is assigned
an individual electrical conductivity value. Harmonic averaging is used to derive conductivities
between material interfaces. The potential is discretized in a regular grid even in solid phases.
Additional jump variables across material interfaces are introduced, where the continuity of the
electrical flux yields the need for extra equations. A so called Schur complement formulation for
these new variables is derived such, that fast Fourier transforms (FFT) and BiCGStab methods
can be used to solve the linear system. If the values of the jump variables are known, the FFT
solves the electrical conduction equation directly. This means that instead of solving a 3D
problem in order to find the electrical potential, a 2D problem is solved derive the jump variables.
Hence, computational time is reduced significantly. In addition, this method has a very good
convergence rate even in geometries with high contrast in conductivity, such as it is mandatory

SCA2015-050

3/6

for natural rocks. The convergence rate solely depends on the number of jumps, i.e. the surface
area of the material interfaces.

SAMPLE MATERIAL, -CT IMAGING & MODEL SETUP

For this study, three different rocks have been used (Fig.1, left to right): a highly porous
carbonate, a young tertiary sandstone, and a cretaceous reservoir sandstone. The carbonate
(denoted as EBK) is characterized by high porosity (up to 48 %) composed of inter- and intragranular void space, and by a high degree of mineralogical pureness (> 99 % CaCO3). The
tertiary sandstone (denoted as IG) features large and sharp edged quartz grains that are covered
by cristobalite needles, an overall low clay mineral content (< 2 %) and few but large muscovite
crystals. The porosity can reach 37 % and is dominated by intergranular voids only. The reservoir
sandstone (denoted as BE) features a porosity of up to 25 %, which is composed of inter-granular
voids of the quartz grain matrix as well as of micro-pores, which are related to the clay
agglomerations (up to 9 vol.-%) within.

Figure 1: -CT cross sections of the three investigated samples at a scanning resolution of 2 microns each (from left
to right hand side): carbonate, young (tertiary) sandstone, reservoir (cretaceous) sandstone.

From all materials, a small plug (4 mm diameter, 8 mm length) has been prepared for the -CT
imaging. The samples have been entirely scanned with a GE phoenix nanotom S tomograph, with
three different voxel resolutions: 2 m, 5 m and 10 m. Within the 2 m scan, a region of
interest of 1000 has been cropped and registered to the 5 m scan, which then has been cropped
at the same relative domain size and registered to the 10 m scan, which has been cropped in the
same way. Hence, for all samples, the exact same region of interest, i.e. 2 mm edge length (=
1000 @ 2 m, 400 @ 5 m and 200 @ 10 m) has been extracted for three different scan
resolutions. The 3D data sets have been processed individually, which means that they have been
filtered (non local mean) and registered with scanning electron data for a best fit segmentation
solution by using Otsu [9] and watershed algorithms [10]. For the formation factor modeling, the
following conductivities have been used [3, 11]: w = 5 S/m, matrix_EBK = 1e-10 S/m; matrix_IG =
1e-12 S/m; matrix_BE = 1e-12 S/m. Boundary conditions have been set to periodic, stopping
criteria was a model (0) accuracy equal or better than 0.001 S/m and a maximum of 10,000
iterations or 240 h maximum run time, respectively.

RESULTS

SCA2015-050

4/6

First of all, results of segmentation errors are showcased (Fig.2), since the segmentation threshold
directly influences the amount of porosity of the 3D data set, and hence of the modeled F-value.
The best fit threshold has been used as starting (center) point to vary the threshold stepwise for
selected percentages to this value. It is clearly visible, that the segmentation threshold, as a
function of scan resolution, has a tremendous influence upon the measurable porosity of the 3D
data sets, even for very small segmentation misfits. The resolution effect is especially for the
carbonate detectable, due to the clearly visible spreading of the error curves, caused by the loss of
intra-granular voids with increasing voxel resolution. 2 % threshold error may cause already
2.5 % to 6 % of porosity change. For the IG sample, this effect is below 2.5 % change in
porosity, due to the absence of micro porous structures.

Figure 2: Results of the segmentation error (x-axis = rel. threshold deviation related to the best fit threshold) on the
segmented data set porosity (y-axis: rel. porosity deviation related to the best fit threshold) for the carbonate (top),
tertiary (middle) and cretaceous sandstone (bottom). The right column is zoomed in ( 7.5 % on x-axis) for a better
view on deviations at small changes for each sample. Hence, the left hand column showcases the full investigated
deviation range ( 25 % on x-axis).

SCA2015-050

5/6

Figure 3: Relationship between scan resolution and segmented porosity (left hand side) and modeled F (right hand side),
respectively, for each of the three samples.

Figure 4: Scan dependent formation factor values versus porosity, including cementation exponent ranges related to
(E2), for the two different investigated sandstones. Diamonds = BE; triangles = IG.

CONCLUSIONS & OUTLOOK

This study has shown that formation factor modeling is very case sensitive towards the 3D scan
resolution, the segmentation threshold, and hence the segmented porosity. As a result, additional
information of the sample material (e.g. scanning electron microscopy images for a highly
reliable segmentation procedure; mineralogical information to estimate the amount and expected
types of pore structures) in close combination with high 3D image resolution is highly
recommended. Though modeling time may be drastically decreased (factor 50 for the EBK, 130
for the IG and for the BE) by using coarse image resolution, i.e. smaller model domains, results
may lead to significantly different interpretation of the reservoir characteristics.

SCA2015-050

6/6

This data base is being constantly increased for a broader range of rocks in order to transfer this
knowledge to get reliable interpretation results. For this, ongoing studies on representativeness of
the 3D data sets, on impact of the pore fluid (conductivity) and on heterogeneous (anisotropic)
rock types are performed and hopefully published in the near future.

REFERENCES
1. K. Sundberg: Effect of impregnating waters on electrical resistivity of soils and rocks.
Trans. AIME. Geoph. Prosp. (1932), p. 367.
2. G.E. Archie: The electrical resistivity log as an aid in determining some reservoir
characteristics. Trans. Americ. Inst. Mineral. Met. 146 (1942), p. 54 62.
3. J. Schn: Petrophysik Physikalische Eigenschaften von Gesteinen und Mineralen.
Ferdinand Enke Scientific Publishers, Stuttgart, 1983.
4. V.N. Dachnov: Promyslovaja geofizika. Izdat. Gostoptechizdat, Moscow, 1959.
5. H. Andr, N. Combaret, J. Dvorkin, E. Glatt, J. Han, M. Kabel, Y. Keehm, F. Krzikalla,
M. Lee, C. Madonna, M. Marsh, T. Mukerji, E. Saenger, R. Sain, N. Saxena, S. Ricker, A.
Wiegmann, X. Zhan: Digital Rock Physics benchmarks Part 1: Imaging and
Segmentation. Computers & Geosciences 50 (2013), p. 25-32.
6. H. Andr, N. Combaret, J. Dvorkin, E. Glatt, J. Han, M. Kabel, Y. Keehm, F. Krzikalla,
M. Lee, C. Madonna, M. Marsh, T. Mukerji, E. Saenger, R. Sain, N. Saxena, S. Ricker, A.
Wiegmann, X. Zhan: Digital Rock Physics Benchmarks Part 2: Computing effective
properties. Computers & Geosciences 50 (2013), p. 33-43.
7. A. Wiegmann and K. P. Bube: The Explicit-Jump immersed interface method: Finite
difference methods for PDE with piecewise smooth solutions. SIAM Journal on
Numerical Analysis, 1997 (37), p. 827-862.
8. A. Wiegmann and A. Zemitis: EJ-Heat: A fast Explicit-Jump harmonic averaging solver
for the effective heat conductivity of composite materials. ITWM report Fraunhofer, 2006.
9. N. Otsu: A Threshold Selection Method from Gray-Level Histograms. IEEE Transactions
on Systems, Man, and Cybernetics, Vol. SMC-9 (1979), 1: 62-66.
10. J. Ohser & K. Schladitz: 3D Images of Material Structures Processing and Analysis.
Wiley-VCH Publishers, Weinheim, 2009.
11. D. Tiab, E. Donaldson: Petrophysics Theory and Practice of Measuring Reservoir Rock
and Fluid Transport Properties. Second edition, Elsevier, Amsterdam, 2003.

SCA2015-051

1/6

MICRO-CT IMAGING AND MICROFLUIDICS FOR

UNDERSTANDING FLOW IN COAL SEAM RESERVOIRS
Peyman Mostaghimi1, Ryan T. Armstrong1, Alireza Gerami1, Majid Ebrahimi Warkaini2,
Hamed Lamei Ramandi1, and Val Pinczewski1
1
School of Petroleum Engineering, University of New South Wales, NSW 2052, Sydney,
Australia.
2
School of Mechanical and Manufacturing Engineering, Australian Centre for
Nanomedicine, University of New South Wales, NSW 2052, Sydney, Australia.
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in Newfoundland, Canada, 16-21 August 2015

ABSTRACT
We investigate transport through fractured coal samples using microfluidic experiment.
Firstly, micro-CT imaging at dry and wet conditions are performed to identify subresolution natural fractures in coal (cleats). Scanning Electron Microscopy (SEM) is also
performed to image cleats at resolution of 100 nm. A novel image segmentation method
is developed to segment the micro-CT image and measure the fracture apertures based on
SEM data. We fabricate micro-models based on observed geometrical features on the
segmented coal images. Microfluidic facility and high speed video microscopy are used
to capture displacement of gas by brine on the fabricated model. This is a first step for
understanding and analysis of high-viscosity ratio displacement in complex coal cleat
systems and to optimize gas recovery from coal beds.

INTRODUCTION
Coal seam methane is a form of natural gas stored in coal beds and is one of the most
important unconventional resources of energy. The flow and transport in coal beds occurs
in a well-developed system of natural fractures that are also known as cleats. The cleat
systems in coals have been the subject of intensive studies in recent years. The techniques
used range from two-dimensional visual observations of coal outcrops and mine sites,
high magnification and scanning electron microscopy (SEM) studies of rock fragments
[3, 7, 8] to three-dimensional studies using X-ray micro-computed tomography (microCT) imaging with resolutions down to approximately a micron [2, 5, 9]. These studies
show that the natural cleat system in coals spans all length scales from meters to
hundreds of meters for exogenous fractures resulting from tectonic activity down to
micro and nano-meters for endogenous micro and sub-micro fissures formed during
coalification.
Although cleat orientation, density, aperture, height, length and connectivity are
suggested to be important to the development of permeability none of the existing studies
attempt to relate these features to flow properties of coal [1, 4]. This is a major

SCA2015-051

2/6

shortcoming in the current understanding of the production characteristics of coal seam

gas and in the ability of the industry to assess the value of laboratory measured core data.
In this paper, we discuss a workflow where coal samples are imaged twice, i.e. without
and with an X-ray attenuating fluid present in pore spaces. This will allow for
visualisation of micro-cleats under the resolution of the micro-CT scanner. Following
this, a novel image segmentation method is developed that employs data from SEM to
find the cleat aperture on the micro-CT image. We fabricate micro-models based on
observed geometrical features in the coal images and study transport phenomena at
micrometer scales using microfluidic technology. This will assist in understating
transport phenomena in coals and optimising recovery from coal seam gas reservoirs.

MATERIALS AND METHODS

Imaging
The coal sample is a volatile bituminous coal from Queensland, Australia. Micro-CT
imaging was performed using a high-resolution, large-field, helical scanning instrument
at Australian National University. The sample was saturated with 1.5 molar Sodium
Iodide (NaI) brine to increase the X-ray attenuation of the coal fractures. The sample was
a full size core plug, i.e. 25 mm diameter and 35 mm length, imaged at a resolution of
16.5 micrometers. The collected images were segmented using a watershed-based
segmentation technique. After micro-CT imaging the sample was cut, parallel to butt
cleats, and polished for the SEM imaging. SEM images were registered to the micro-CT
data using the method developed by Latham et al. [6].
Flow Simulations
We calculate permeability and find the pressure and velocity profiles of coals on images
obtained from micro-CT imaging. After image segmentation, the pore space is used as a
computational domain to solve for Stokes flow. The governing equations for flow at the
pore scale are:
(1)
v 0
2
(2)
v P
where v is the velocity vector, P is pressure and is viscosity of the fluid flowing in the
porous medium. The equations are discretized using the finite difference method. The
main numerical challenge in solving the Stokes equation Eq. (2) is the weak coupling
of the pressure and velocity fields. For decoupling pressure and velocities, the Semi
Implicit Method for Pressure Linked Equations (SIMPLE) is used.
Experimental Setup
To perform the microfluidic experiment, a CAD (Computer-Aided Design) model based
on the observed coal features is designed. A micro-model is fabricated using
Polydimethylsiloxane (PDMS) based on the CAD design using softlithography technique.
Then 1/16 PTFE tubes are connected to the flow inlet and outlets and a syringe pump

SCA2015-051

3/6

with the accuracy of 0.35% and step resolution of 0.046 microns is used for flow
injection. The injection rate is 1 l/min. To visualise the flow pattern inside the
microfluidic channels, an inverted microscope with equipped high-speed and CCD
camera (Oympus IX81 and DP80) with 13601024 pixels resolution is used.

RESULTS
Figure 1 shows the workflow used in this study for creating the micro-models and
running the experiments. Figure 1(A) shows the grey scale image obtained from microCT imaging of the coal sample. The image has the resolution of 16 microns and imaging
has been performed at wet and dry conditions. The grey scale image is obetained by
subtracting the dry image from the wet image. This will enable us to visualise features
smaller than the scanner resolution [5].
B

Figure 1: Workflow for the fabrication of micromodels based real cleat geometries. Micro-CT images of
coal at wet and dry conditions are used to obtain a grey scale image (A). The image will be segmented (B)
and the cleat system features will be visualised to be fabricated (C). The fabricated micro-model is used for
microfluidic flow experiments using advanced microscopy tools (D).

SCA2015-051

4/6

The segmentation of coal images is of high importance due to existence of very narrow
connectivities in coal samples. Figure 2 shows results of flow simulation over a region of
interest (ROI) on the coal image. As can be seen from the velocity profile, there is a
single voxel connecting the cleat system. Therefore, losing a single voxel in segmentation
may lead to losing the connectivity.

Figure 2: Single-phase flow simulation on extracted fracture geometry (Left: normalized pressure profile,
Right: normalized velocity vectors in the black rectangle).

To segment the grey scale image, we use data from SEM imaging performed at a
resolution of 100 nm and relate the grey scale micro-CT values to the measured aperture
for the cleat system on the SEM image. Figure 3 shows a single cleat as observed in
micro-CT and SEM imaging. The aperture of the cleat has been measured on SEM image
and is related to the grey scale value. Repeating this process for different apertures will
result in a curve that shows the corresponding aperture size for each grey scale value on
the micro-CT image (Figure 3). The curve is used for image segmenation and the grey
scale image is segmented into four different phases: (1) resolved fractures, (2) subresolution porous regions, (3) minerals, and (4) solid regions.

Figure 3: Registered micro-CT and SEM data for measurement of coal fracture apertures.

SCA2015-051

5/6

After segmentation, the resolved cleat system can be visualised and we generate CAD
models based on the observed features in the cleat system. In this paper, we focus on a
small region of interest where a junction of four fractures is observed (Figure 4). A
micromodel is fabricated with similar topology and fracture aperture. Brine is injected
from the widest fracture to the junction at the rate of 1 l/min and the displacement of air
is captured using the high speed camera. Figure 4 shows the movement of interface
between fluids accurately with a contact angle of almost 90. In our future works, we will
use plasma treatment to change the wettability of surfaces and study the effect of
wettability on gas recovery in coal seam reservoirs at micrometer scale. In addition, we
will extract several unique features observed on segmeneted coal images and fabricate
micro-models based on them. Then, displacement at each of these micromodels will be
captured and investigated. This provides a comprehensive framework for investgating
flow characteristics of coals and validation data for numerical computations.

Figure 4: A 2D slice of the micro-CT image of coal where a region of interest (ROI) is chosen to be
fabricated. The displacement of air by water at three stages is shown on the fabricated model representing a
junction of coal cleat system. The dimension and the shape of the fabricated micromodel is chosen based
on measurement of the fracture aperture size of the micro-CT image.

SCA2015-051

6/6

CONCLUSION AND FUTURE WORK

A novel workflow for investigating the effect of cleat topology and geometry on fluid
flow in coals is discussed. Coal is imaged at dry and wet conditions. This allows for
visualisation of sub-resolution cleats in the image. An original segmentation method is
used to extract the cleat network from the micro-CT data. Numerical simulation of flow
shows that the cleat systems have very narrow connection that needs to be honoured in
the segmentation. A unique feature of the cleat system is fabricated for microfluidic flow
experiments. Preliminary flow experiments are performed to capture the interface
between brine and air and to understand transport at cleat junctions. In future, we will
conduct several experiments on a range of observed topological features in coals. This
data presented herein is the first stage of an ongoing research effort to better understand
flow in coal seam gas reservoirs.

REFERENCES
1. Aminian, K., Rodvelt, G., Chapter 4 Evaluation of Coalbed Methane Reservoirs, in
Thakur P., Schatzel S. and Aminian K. (Eds) .Coalbed Methane, Elsevier,
Oxford, (2014), 63-91.
2. Cai, Y., Liu, D., Mathews, J., Pan, Z., Elsworth, D., Yao, Y., Li, J. Guo, X.,
Permeability evolution in fractured coal, International Journal of Coal Geology,
(2014) 122,91-104.
3. Dawson, G., Esterle, J., Controls on coal cleat spacing, International Journal of
Coal Geology, (2010) 82, 213-218.
4. Flores, R., Chapter 5 Coal composition and reservoir characterization, in Thakur P.,
Schatzel S. and Aminian K. (Eds). Coalbed Methane, Elsevier, Oxford, (2014), 93100.
5. Golab, A., Ward, C., Permana, A., Lennox, P., and Botha, P., High-resolution threedimensional imaging of coal using microfocus X-ray computed tomography with
special reference to modes of mineral occurrence, International Journal of Coal
Geology, (2013) 113, 97-108.
6. Latham, S., Varslot, T. and Sheppard, A., Image registration: enhancing and
calibrating X-ray micro-CT imaging, International Symposium of the Society of
Core Analysts, (2008) Abu Dhabi, UAE.
7. Laubach, S., Marrett R., Olson, J., and Scott A., Characteristics and origin of coal
cleat: A review, International Journal of Coal Geology, (1998) 35, 175-207.
8. Liu, S., Sang S., Liu H., and Zhu, Q., Growth characteristics and genetic types of
pores and fractures in a high rank coal reservoir of southern Qinshui basin, Ore
Geology Reviews, (2015) 64, 140-151.
9. Yao, Y., Liu, D., Che, Y., Tang, D., Tang, S., and Huang, W., Non-destructive
characterization of coal samples from China using microfocus X-ray computed
tomography, International Journal of Coal Geology, (2009) 80, 113-123.

SCA2015-052

1/7

DETERMINATION OF PHYSICAL PROPERTIES OF

TIGHT POROUS MEDIA USING DIGITAL CORE
PHYSICS/ANALYSIS
Shahin Ghomeshi, Saeed Taheri, Evgeny Skripkin, Sergey Kryuchkov,
Apostolos Kantzas
University of Calgary
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
In this study we use pore-scale SEM images of tight porous media, and generate
threholded binary images which are then used to reconstruct three-dimensional (3-D)
pore structures. Computational physics is then employed in order to calculate the
physical properties of the porous media, such as the porosity, permeability, electrical
resistivity and formation factor, and the NMR spectra. This involves generating an
unstructured 3-D grid in the pore space in which we solve the corresponding governing
equations. The original thresholded images have high resolution which allows a more
precise masking of the pore spaces. However, there is a high computational cost
associated with the high resolution images. Reducing the resolution will reduce the mask
precision in the pore areas and this will lead to a different numerical solution. Therefore,
we much optimize between the computational cost and accuracy when performing the
numerical simulation. The threshold values for the samples at hand were selected by
trying to match porosity and permeability of a neighboring plug. The choice of threshold
value has a profound effect on the porous medium properties. For the fluid flow
calculations, we solve the velocity and pressure in the pore space by solving the NavierStokes equations and using the results to obtain the absolute permeability. The electrical
resistivity is obtained by solving for the current density through Ohms law, and for the
NMR study we solve for the equations of molecular diffusion. This approach can be
described as a version of digital core analysis (DCA) or digital core physics (DCP).

INTRODUCTION
The main goal of this work is to characterize and determine the physical properties in
tight porous media from actual pore samples of a tight formation. The method we employ
is based on a computational physics approach where we discretize the 3-D structure into a
computational mesh where the governing equations are then solved.
This study uses three scanning electron microscope (SEM) images of the tight sample as
in Figure 1, and poses a method for numerically determining the physical properties of
the reconstructed image, which includes the porosity, absolute permeability, electrical

SCA2015-052

2/7

resistivity and formation factor, and the NMR spectra. We then compare and validate the
results to the experiment.
There are several challenges associated with this task. The main challenges include: the
3-D reconstruction of the sample with only three 2-D images, the effect of different
threshold of the 2-D images on the calculation of the physical properties, and the
computational cost involved in the solution to the 3-D heterogeneous (therefore noted as
upscaled) geometry.
The work here relies on 3-D reconstruction based on only three SEM images. Previous
approaches of reconstruction include the work of Okabe and Blunt [1, 2] which
reconstructed porous media based on only a single training image. A major assumption
for them is that they assumed isotropy where the measured statistics on the XY plane is
transformed to the XZ and YZ planes. We do not make such assumption in this work as
several images were extracted directly from the given SEM images and used to recreate a
3-D digital porous media pattern.
The SEM images are 5120 3828 pixels (~ 637 476 microns) and are highly
heterogeneous with large range of grain sizes. We provide a given tolerance for the
different grain sizes into small, medium, and large. The smallest grains have bounding
boxes ranging in size from 22 microns to 1010 microns, the medium sized grains are
from 10 10 microns to 50 50 microns, and the large grains have bounding boxes
larger than 50 50 microns. The images are then decomposed into permeable domains
and non-permeable domains, where the non-permeable domains include the large and
medium sized grains. The permeable domains are at sub-pore scales where we extract
from the original image to include the small grain sizes with tiny pores in between
(Figure 2). We crop 33 of these images which are 256 256 pixels (or ~32 32
microns) from the three available SEM images. More images are required in order to
reproduce a 3-D structure where we will perform pore-scale simulations. We do a spline
interpolation between the greyscale of each 2-D image cropped images, in order to
produce more 2-D images which we use to reconstruct the 3-D structure as show in
Figure 3.a.
We perform pore-scale numerical simulations to calculate the permeability, resistivity
and formation factor for the reconstructed geometry. The reconstruction, modeling and
simulating process is an approach that is used in the areas of digital rock physics (see for
example, [3, 4]) for understanding the macroscopic rock properties. Here we ignore the
effects of adsorption, and assume that we are in the regime where the Navier-Stokes
equations apply. For the permeability calculations we impose a simple pressure inlet
condition and a 0 outlet condition, and solve the Stokes equation to calculate average
velocity of the flow within the pores as shown in Figure 3.b. Using the velocity and
pressure profiles, we then back calculate the Darcy equation to obtain the permeability.
We change the inlet and outlet, and calculate the permeability in different directions.
Resistivity and formation factor calculations are obtained by solving for the current
density using Ohms law, see Figure 3.c. The calculated properties of the porous media

SCA2015-052

3/7

are tabulated in Table 1. We also solve the equations of molecular diffusion and evaluate
the NMR spectrum which is shown in Figure 3.d.
Following the pore-scale computations, we use the obtained values as properties of the
permeable domain, while the grain properties are set to zero. We then solve for the fluid
properties of the complex upscaled geometry. The upscaled geometry is constructed by
using imageJ and tracing out the large and medium sized grains from the three SEM
images, as seen in Figure 4.a. In order to represent the 3-D structure of the grains, we
apply a dilation / erosion algorithm in several slices for both the medium and large sized
grains as seen in Figure 4.b. The final set of images is then assembled by merging the
medium and large grains together as shown in Figure 4.c. We then used these composite
images in order to re-construct the upscaled representative geometry of the original image
(see Figure 5.a) based on the three SEM images. In between the medium and large grains
we assume a permeability of 3 D in the z-direction as seen from Table 1. The porosity
will be given by
={

0
0

On medium and large grains

Everywhere else

(1)

where 0 = 0.09 from Table 1. The porosity of the upscaled geometry will then be the
averaged porosity over the entire volume V given by
=

2.015 1012 3

= 2.8663 1011 3 0.07

(2)

We assume Darcy flow where the permeability is given by

={

0
0

On medium and large grains

Everywhere else

(3)

where 0 is the permeability of the cropped region (Figure 3.a). This resulting pressure
and velocity profiles is then used to calculate the average permeability of the medium
(Figure 5). Using Darcys law, assuming incompressible fluid, we have
. = 0 ,

u=-

p,

(4)

where is the viscosity of the fluid, and is the density. We inject a velocity of 1.0
107 / at the inlet in the x-direction and a pressure of 0 Pa at the outlet. This will
result in a p = 6818 Pa (see Figure 5. b). The average permeability will then be given
by

= ,
(5)
where = is the Darcy velocity at inlet, which will lead to the result of 6.9 .

SCA2015-052

4/7

CONCLUSION
A new approach of imaging and computing is presented in this paper for determining the
petro-physical properties in tight formations using SEM images. SEM-CT images give
higher resolution than micro-CT images and can handle smaller sized pores specifically
in tight pores. This approach allows us to reconstruct the 3-D image based on only three
SEM images and to numerically simulate various physical properties in this digital object
such as permeability, porosity, electrical conductivity, and NMR.
The new component of the work relies firstly on the fact that images were not statistically
created and we do not make any isotropy assumptions. Images are extracted directly from
three SEM images provided. The only assumption is that the correlation length in the zdirection is the same as in the x and y directions.
Secondly, the reconstructed cropped images which are 256 256 256 pixels lead to
relatively high number of pores, approximately 10,000 15,000. At this number of pores
the properties of the sample is same as what will be observed at much bigger sample.
Therefore we performed the pore-scale simulations at the lower scales and then attempted
early upscaling where we started describing the flow as Darcy flow.
We note that these results were based on a given threshold, and for very tight media we
must have a very high resolution in order to be able to capture very low porosities
without losing connected pores. Moreover, based on our grey scale, we assume pore
regions which instead could have been organic material with nano sized pores. This
would complicate gas flow simulations that would be necessary to take into account the
effects of adsorption. This would in turn lead to different set of governing equations.

ACKNOWLEDGEMENTS
We would like to acknowledge the financial support from NSERC AITP/i-CORE
Industrial Research Chair in Modeling of Fundamentals of Unconventional Resources,
and the sponsoring partners: Alberta Innovates, Athabasca Oil Corporation, Brion
Energy, Canadia Natural, Devon Foundation CMP, Husky Energy, Laricina Energy
LTD., NSERC, Schulich School of Engineering-University of Calgary.

REFERENCES
1. Okabe, H., Blunt, M. J., Prediction of permeability for porous media reconstructed
using multiple-point statistics, Phys. Rev., (2004) E70, 6, 066135.
2. Okabe, H., Blunt, M. J., Pore space reconstruction using multiple-point statistics, J.
Petroleum Science and Engineering, (2005) 46, 121-137.
3. Andra , H. et al., Digital rock physics benchmarks-Part I: Imaging and
segmentation, Computers and Geosciences, (2013) 50, 25-32.
4. Andra , H. et al., Digital rock physics benchmarks-Part II: Imaging and
segmentation, Computers and Geosciences, (2013) 50, 33-43.

SCA2015-052

5/7

Table 1: Tabulated results from the pore-scale simulations

Porosity
0.09

Resistivity,
.m
145

Kx
3.0

Permeability, D
Ky
2.3

Figure 1: Three SEM images of a tight formation

Figure 2: Several Cropped images from SEM images

Kz
9.2

SCA2015-052

6/7

(a)

(b)
40

Amp

30
20
10
0
0.1

(c)

T2 (ms)

10

(d)

Figure 3: (a) computational domain (b) permeable regions, (c) current density (d) NMR spectra

(a)

(b)

100

SCA2015-052

7/7

(c)

Figure 4: (a) large and medium sized grains, (b) dilation sequence (c) merging of the large and medium grains

(a)

(b)

Figure 5: a. Upscaled geometry with dimensions 637 475 96 , b. Surface pressure profile for the
upscaled geometry

SCA2015-053

1/6

PORE-SCALE OBSERVATION OF RESIDUAL OIL IN

LOW SALINITY SURFACTANT FLOODING
Hamid Hosseinzade Khanamiri, Ole Torster, Norwegian University of Science and
Technology; Jan ge Stensen, Sintef Petroleum Research
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Low salinity water injection provides favorable conditions for chemical enhanced oil
recovery (EOR). Lower surfactant adsorption and lower chemical consumption is an
advantage of combined low salinity water and surfactant injection. However at lower
ionic strength the value of interfacial tension (IFT) is usually higher than the IFT at
higher or optimal salinities. Recently it has been found that existence of small amount of
calcium in low salinity surfactant can reduce the oil/water IFT. In this work, a
coreflooding experiment is performed at the micro-CT scale. In tertiary injection the
surfactant solution contained only NaCl, while the after-tertiary surfactant injection
contained both NaCl and CaCl2. After every step, the sample was scanned to observe the
saturation and distribution of residual oil. It was observed that calcium-enhanced low
salinity surfactant mobilized and recovered mainly the residual oil clusters which were
affected but not recovered by tertiary surfactant. The oil which was not affected by
tertiary surfactant mainly remained unaffected. It was also observed that oil recovery by
low salinity surfactant in pore scale was higher than in normal core scale flooding
experiment. In other words, microscopic sweep efficiency by low salinity surfactant was
stronger than the macroscopic sweep efficiency.

INTRODUCTION
Both surfactant flooding and low salinity waterflooding may improve oil recovery. The
goal of this work was to search for possible improvements in the performance of the
combined low salinity water and surfactant by only modifying the ionic composition
instead of adding extra chemicals. Observation of residual oil in pore scale by micro-CT
was used to examine outcomes of the flooding experiment.

EXPERIMENTAL MATERIAL
Rock. The rock sample had a diameter of 4mm and a length of about 10mm. The sample
was extracted from a Berea outcrop sandstone plug with permeability in the range of 200300mD. Porosity of the scanned portion was 15.9%.
Oil. The oil phase is mixture of a crude oil (Table 1) and 1-Iododecane. The volume
fraction of 1-Iododecane and crude oil is 6/4 in the mixture. 1-Iododecane increases the
x-ray attenuation coefficient of oleic phase.

SCA2015-053

2/6

Table 1. Composition in weight % and physical properties of the crude oil, [1]
Composition (%)
Physical properties
Saturates
61.19
TAN [mg KOH/g]
1.08
Aromatics
32.42
TBN [mg KOH/g]
1.160.35
Resins
4.93
Density, 15C [g/cm3]
0.8582
Asphaltenes 1.46
Density, 60C [g/cm3]
0.8252
API gravity [API]
33.5
Viscosity, 15C [mPas]
19.90
Viscosity, 60C [mPas]
4.07

Brines and Surfactant. Details of the experimental brines are given in Table 2. 500 mg/l
of sodium dodecylbenzenesulfonate (SDBS) is used in surfactant injection steps.
Table 2. Ionic composition of the brines and IFTs; IFT of doped crude was impossible to be measured by
spinning drop as denser phase (oil) is not transparent and water/surfactant drop is invisible.
Brine
NaCl
CaCl2.2H2O
SDBS TDS
Ionic
IFT, only
IFT, only 1mg/l
mg/l
mg/l
mg/l
strength
crude
Iododecane
mmol/l
mN/m
mN/m
32500
32500 556.1
84
in-situ
3250
3250
55.61
88.7
LSW-Na
3250
500
3250
57.05
1.6
15.6
LSS-Na
3087.5
136.2
500
3690
57.05
0.33
LSS-Ca
32500
32500 556.1
84
HSW-Na

EXPERIMENTAL PROCEDURE
The rock sample was mounted in the micro-CT core holder, vacuumed and then saturated
with in-situ brine (Table 2). Oil was then injected to drain mobile water. EOR injection
was started with low salinity sodium chloride brine (LSW-Na in Table 2). Low salinity
surfactant injection with sodium chloride (LSS-Na) was then performed. The next step
was also low salinity surfactant injection. The difference was that small amount of
sodium was replaced by calcium (LSS-Ca). Final injection was high salinity sodium
chloride brine (HSW-Na). The experiment was performed at 27C and atmospheric
pressure. At the end of LSW-Na, LSS-Na, LSS-Ca and HSW-Na the injection was stopped
and core holder was isolated for about 1hour to scan the sample. The sample was also
scanned before saturation (dry scan) and afterward at initial state with connate water and
initial oil. The details of injection experiment are given in Table 3. The scans were
performed by micro-CT Skyscan 1172 with a resolution of 5.44m/pixel.
Table 3. Details of EOR injection, the PVI (pore volume injected) is calculated roughly based on the pore volume
of the whole sample (0.02cm3) with length of 10mm, while the scanned length is 6mm.
Injection
Flow rate
Equivalent pore
Duration PVI
Scan name Initial Final
Rec. %
(ml/min)
velocity (ft/day)
(min)
So
So
OOIP
Oil injection
0.004-0.04
10-95
88
36.6
1-soi
0
0.864
0.002
4.7
60
6
LSW-Na
0.004
9.4
30
6
0.008
18.8
30
12
2-sorlsw
0.864
0.636
26.4
LSS-Na
0.008
18.8
93
37.2
3-sorlssna
0.636
0.529
12.4
LSS-Ca
0.008
18.8
93
37.2
4-sorlssca
0.529
0.426
11.9
HSW-Na
0.008
18.8
30
37.2
5-sorhsw
0.426
0.317
12.7
Total
0.864
0.317
63.3

SCA2015-053

3/6

RESULTS AND DISCUSSION

Image processing was performed using CTAn. The difference of gray scale index (GSI)
between oleic and aqueous phases is about 100 units in the interval of 0-255. This large
difference makes the fluids segmentation less subjective.
Although measurement of produced oil in micro-CT flooding experiments are almost
impossible, oil recovery during all four steps of injection was observed. It was also
observed that the recovered oil by HSW-Na in the last injection was in form of emulsion
as opposed to larger oil clusters produced during low salinity water and surfactant
flooding steps. The emulsions were formed at improved phase behavior of surfactant at
higher salinity. The salinity of HSW-Na is close to the optimal salinity of surfactant
solution with un-doped crude at elevated temperature of 60C. Interfacial tension
between crude and surfactant was about 0.005mN/m in that condition.
The saturation profiles are given in Figure 1. The average saturations and recoveries
based on original oil in place are also given in Table 3. The recovery of low salinity
waterflood was 26.5% which seemed to be lower than waterflooding recovery in normal
cores. The reason might be the high density of oil phase which is 1.1g/cm3 due to
addition of 1-Iododecane. High density difference may not cause strong segregation in
micro scale but it may bring fingering of the injected water. Another reason would be the
high IFT of dopant agent with brine which is about 80mN/m (Table 2). On the other hand
the recovery of LSS-Na is 12.4% which is 9% higher than the recovery in normal core
scale performed using the same rock type, though the oleic phase and experimental
conditions were different. Specifications of normal core flooding experiment are
summarized in Table 4. There is also 11.9% oil recovered in the LSS-Ca step. This high
difference of recoveries by surfactant in pore scale and normal core implies that
microscopic sweep efficiency of oil by surfactant is stronger than its macroscopic sweep
efficiency.
Table 4. Specifications of the normal coreflooding experiment both CT sample and normal core were
extracted from an identical Berea block
9.9cm
Crude
crude in Table 1
length
In-situ brine
in-situ in Table 2
diameter 3.8cm
309mD
1st flood (LSW-Na) recovery
51.5% OOIP
k
15.5%
2nd flood (LSS-Na) recovery
3% OOIP

76.7%
Flooding at 60C
Aging after drainage: 3 weeks at 80C
Soi

Further, the saturation profiles in Figure 1 show that although in many parts of the sample
oil mobilization occurred; in the two intervals of 3200-3500m and 5800-6100m low
salinity surfactant with sodium (LSS-Na) could not mobilize oil. However in the next step
(LSS-Ca) oil mobilization occurred. This extra oil mobilization at 3200-3500m was
most probably produced. If it was trapped again, the oil saturation at some part of the
sample ahead of this point would have increased resulting in higher residual oil saturation
than the previous surfactant injection (LSS-Na).

Oil saturation (fraction)

SCA2015-053

4/6

1
0.8

1-Soi

0.6

2-Sor-LSW-Na
3-Sor-LSS-Na

0.4

4-Sor-LSS-Ca
5-Sor-HSW-Na

0.2
0
0

1000

2000

3000
4000
5000
Vertical position (m)

6000

7000

Figure 1. Oil saturation profile at initial condition (1-Soi) and after every injection (Sors); the two dashed ovals
represent two parts of the sample where LSS-Na was ineffective while LSS-Ca could mobilize and recover oil.
3000
1-soi
Frequency

Frequency

2000

1000

0
1.E+03

1.E+05

1.E+07

size of oil cluster (m3)

1.E+09

2-Sor-LSW-Na

2000

3-Sor-LSS-Na
4-Sor-LSS-Ca
1000

5-Sor-HSW-Na

0
0

0.2

0.4
0.6
sphericity

0.8

Figure 2. Oil cluster size (left) and sphericity (right) distribution at initial conditions and after every flooding

As shown in Figure 2 (left), number of oil clusters is minimum at initial conditions (1Soi). The number has a sharp rise after low salinity waterflooding (2-Sor-LSW-Na).
Afterwards, number of clusters increases slightly during low salinity surfactant with
sodium (3-Sor-LSS-Na), while it decreases after the second surfactant injection with
calcium (4-Sor-LSS-Ca). The sphericity of the oil clusters, Figure 2 (right), has similar
trend in different steps of the experiment. Sphericity of a normal sphere is 1 and as the
cluster becomes more irregular in shape, sphericity tends to 0. Normally sphericity of
smaller clusters is more than that of larger ones.
1-Soi and 2-Sor-LSW-Na. Comparison of the low salinity water images with initial
conditions shows that whenever oil is mobilized in a small pore the whole oil is often
swept out the pore (Figure 3-A) while in medium pores some oil is left behind in form of
trapped oil in the corner of pore (Figure 3-B). In larger pores the residual oil has more
rounded profile and is mainly in the middle of the pore (Figure 3-C). This means that
although the viscous force was high enough to invade the large pore it just peeled and
displaced layers of oil instead of pushing the whole drop out of the pore. It was also
observed that there was residual oil at some points where it was initially occupied by
connate water (Figure 3-D). This type of residual oil was certainly mobilized initially but
re-trapped later.

SCA2015-053

5/6

E
B

1-Soi

C, F

2-Sor-LSW-Na

F, H

3-Sor-LSS-Na

H
J

4-Sor-LSS-Ca

5-Sor-HSW-Na

Figure 3. A vertical cross section of the cylindrical rock sample (D4000m & L6000m) after
different steps; the injection was from top to bottom. Oil is the white phase and water is dark one.

2-Sor-LSW-Na and 3-Sor-LSS-Na. There are many small and medium pores where the
whole residual oil after waterflooding or major part of it was displaced by first surfactant
flooding, Sor-LSS-Na (Figure 3-E), while in larger pores oil was disintegrated into
smaller (increase in number of oil clusters, Figure 2) and more rounded clusters (increase
in sphericity, Figure 2) after layers of oil was peeled off (Figure 3-F). At some points, the
mobilized oil was re-trapped again (Figure 3-G). The amount of this oil was 12.9%OOIP.
3-Sor-LSS-Na and 4-Sor-LSS-Ca. It seems that major part of the swept oil by calciumassisted surfactant flooding (LSS-Ca) is the more spherical residual oil drops which
changed their shape during previous surfactant injection (LSS-Na) (Figure 3-H).
Reduction in sphericity and number of residual oil clusters after LSS-Ca was also

SCA2015-053

6/6

represented in Figure 2. Oil was rarely swept out of the pores which were not
considerably affected by the first surfactant injection (LSS-Na). Moreover less oil was retrapped after mobilization (only 4.8%OOIP compared to 12.9% in previous step). This is
another advantage of LSS-Ca to LSS-Na. Although LSS-Ca could not effectively invade
the relatively unaffected oil by LSS-Na, its oil recovery was 11.9% of OOIP which was a
considerable amount of difficult residual oil after one step of surfactant injection (LSSNa). This extra oil recovery was gained by only a minor change in ionic composition.
4-Sor-LSS-Ca and 5-Sor-HSW-Na. The mobilized oil by HSW-Na at some points is
similar to the last surfactant injection where the more spherical oil clusters which were
affected previously but not produced, were mobilized (Figure 3-I). However, oil
mobilization was also observed in some pores where the two surfactant injections were
ineffective (Figure 3-J). Coalescence of small and more rounded clusters occurred also
(Figure 2). HSW-Na mixes with the surfactant from previous injection making a lower
IFT surfactant solution. It seemed the mixing happened quickly as oil production was
observed in form of emulsion just after some minutes of HSW-Na injection.

CONCLUSION
In the studied rock-fluid system with the mentioned experimental conditions it was
observed that: calcium-enhanced low salinity surfactant recovered mainly part of the
residual oil which was affected but not recovered by tertiary surfactant with only sodium.
Oil re-trapping after mobilization by LSS-Ca is less than by LSS-Na. Comparison of
recoveries in micro scale with those of normal core flooding implied that microscopic
sweep efficiency of low salinity surfactant was higher than its macroscopic sweep
efficiency. Therefore investigating the effect of mobility control during or after low
salinity surfactant seems to be necessary as it may help producing the bypassed oil.

ACKNOWLEDGEMENT
We would like to thank Lundin Norway AS, Statoil ASA, Det Norske Oljeselskap, GDF
Suez, Unger Fabrikker AS and the Research Council of Norway for their financial
support.

REFERENCES
1. Tichelkamp, T., Vu, Y., Nourani, M., ye, G. 2014. Interfacial tension between low
salinity solutions of sulfonate surfactants and crude and model oils. Energy Fuels 28,
24082414.

SCA2015-054

1/7

A MULTI-SCALE DYNAMIC PREDICTIVE MODEL FOR

DRAINAGE AND IMBIBITION CAPILLARY PRESSURE
IN HETEROGENEOUS ROCKS
X. Nie, C. Baldwin, Y. Mu, A. Grader and J. Toelke
Ingrain Inc
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
A multi-scale model for multiphase flow in heterogeneous rocks has been developed to
predict drainage and imbibition porous plate capillary pressures. The need for such a
simulation model arises from the inability to image and simulate at high-resolution large
enough volumes that are representative of multi-scale rocks. The overall distribution of
the different resolutions is defined by multi-scale x-ray imaging. A representative volume
from the micro-CT image is segmented into pores, solids and Darcy regions with
unresolved pores. The properties of Darcy region are further determined by separate
high-resolution imaging and multiphase flow simulations for a smaller volume. Twophase flow in the pore space is directly simulated using a lattice-Boltzmann model. Pore
size in the Darcy region is smaller than the resolution of the lattice-Boltzmann
simulation. The flow entering or exiting a Darcy region is determined by the fluid phase
and its pressure around the Darcy region and the intrinsic properties of the Darcy region
like the porosity and the Pc-Sw relationship. The boundary condition of the latticeBoltzmann simulation on the surface of a Darcy region depends on the solid wettability
and the fluids present in the Darcy region. The fluid distributions in pores and Darcy
regions, and therefore the boundary conditions for different regions dynamically change
with the applied capillary pressure. The multi-scale model has been used to compute the
displacement parallel and perpendicular to a layered system packed with solid spheres of
different size and compared to a fully resolved simulation to validate the model.

INTRODUCTION
The capillary pressure (Pc) and relative permeability in drainage and imbibition processes
are important inputs for reservoir simulations. The capillary pressures have been
experimentally measured in a core or a plug [1-3] for many decades. Recently digital core
analysis has become an important alternative method. In the digital analysis, the rock
geometry is digitized into a 3D matrix data to represent solids and pores by computerized
tomography (CT). To calculate the transport of multiphase flow, either pore network
model or direct simulation can be used. The pore network model replaces real pore
bodies and throat shapes with simple geometric shapes [4, 5]. The fluid transport and
distribution are determined by coupling semi-analytical solutions for the individual

SCA2015-054

2/7

elements. The lattice-Boltzmann method (LBM) is a direct simulation method [6, 7]. The
LBM solves a simplified kinetic equation and directly simulates the fundamental
equations of multiphase flow in pores in a digitized rock sample.
This study develops a dynamic multi-scale model to calculate the capillary pressures in
drainage and imbibition processes in heterogeneous rocks based on the LBM method.
Different digital resolutions are assembled into a representative flow unit volume. The
model is validated by digital porous plate experiments of a system packed with different
sizes of solid spheres.

MATERIAL AND SIMULATION METHODS

In our multi-scale simulation, the distribution of different resolutions of the structure of a
rock sample is determined by multi-scale x-ray imaging. A representative volume from
the coarse micro-CT image is segmented into pores, solids and Darcy regions. Pore size
in the Darcy region is smaller than the coarse micro-CT. Small volumes of Darcy regions
are further imaged by a separate higher-resolution micro-CT or nano-CT acquisition and
segmented into pores and solid. The Pc curves and the breakthrough pressures of Darcy
regions are determined by the LBM simulation in the high-resolution volumes. These
properties are assigned to the Darcy regions.
Now the two phase flow in the whole representative volume is simulated using the
dynamic multi-scale LBM simulator. Two-phase flow in the resolved pore space is
directly simulated using the LBM. The flow entering or exiting a Darcy region is
determined by the fluid phases around the Darcy region, the Pc curves and the
breakthrough pressures of the Darcy region. In the drainage process the Darcy region is
first occupied by water. If a Darcy region contacts with oil phase, the oil enters the Darcy
region gradually as the capillary pressure increases. The amount of oil entering the Darcy
region at each time step of the LBM simulation is determined by the Pc curve of the
Darcy region and the capillary pressure at that time step. When the capillary pressure
becomes larger than the breakthrough pressure the oil is allowed to pass through the
Darcy region. In the imbibition process, a Darcy region may be first completely or
partially occupied by oil. If the Darcy region is connected to water through bulk water or
a water film, the water enters the region gradually as the capillary pressure decreases. The
amount of water entering the Darcy region is determined by the imbibition part of the Pc
curve of the Darcy region.
We used the model in [8] to assign a contact angle at the interface between resolved pore
space and Darcy region. The contact angle depends on the porosity, the intrinsic
wettability and the fluid phases present in the Darcy region and is imposed through the
interaction of LB nodes at the interface between Darcy region volume and resolved pore
volume. The fluid distributions in pores and Darcy regions, and therefore the boundary
conditions and the contact angle dynamically change with the applied capillary pressure.

SCA2015-054

3/7

In order to validate the multi-scale model, we simulated a digital porous-plate experiment

[9] of a rock sample in a fine resolution that completely resolves all pores of the rock
sample using a recently developed LBM solver [10]. Then we used the dynamic multiscale model to simulate the porous-plate experiment of the same rock sample in a 10
times coarser resolution and compared the two results obtained from the two simulations.
In the digital porous plate experiment [9], the one surface of the rock is attached to an oil
reservoir and the opposite surface is attached to a water reservoir. The water pressure is
kept constant throughout the simulation. In the drainage process the oil pressure increases
step by step till the desired residual water saturation is reached. In the imbibition process
the oil pressure decreases till it reaches the water pressure to simulate controlled
spontaneous imbibition. In the recently developed LBM solver [10], we developed
special boundary conditions at the solid wall to model thin water-film flows that allow
transport of fluids through those thin films. These boundary conditions prevent e.g.:
unrealistic high values of trapped water in the drainage and allow snap-off effects
through the films in imbibition.

Fig. 1. The rock sample synthetized by packed

solid spheres. The pixel size is 2 microns. The
sample is a cube of 6003 pixels. Three
different pore sizes are present in the sample.

Fig. 2. The rock sample in figure 1 shown in a 10

times coarser resolution. The cube consists of 603
pixels. The pixel size is 20 microns. The pores in the
Darcy regions colored by blue and red are
unresolved.

As shown in Fig. 1, the rock sample was synthetized by packed solid spheres to simulate
a multi-scale rock. The sample is a cube with side length of 600 pixels and is composed
of 5 regions. The pixel size is 2 microns. In the resolution of 2 microns, all pores are
resolved. Starting from the bottom, the thickness of the second region is 200 pixels and
100 pixels for each of the other four regions. The first, third and fifth regions have the
largest pores. The second region has smaller pores and the fourth has the smallest. The
average pore throat radii in the three different fabrics are 46, 9.2 and 4.6 microns
respectively. The average pore radii in the three fabrics are 67, 13.4 and 6.7 microns
respectively. The porosity of all regions is 0.4. Fig.2 shows the same rock sample in a 10
times coarser resolution. The sample is a cube with the side length of 60 pixels. The pixel
size is 20 microns. The largest pores in the first, the third and the fifth region are still

SCA2015-054

4/7

resolved with the 20 micron resolution. The pores in the second and the fourth regions are
unresolved. The capillary pressures as functions of water saturations and the
breakthrough pressures of the two unresolved regions have been obtained by separate
LBM simulations for the tighter rock fabrics at a resolution of 2 microns and were
assigned to the corresponding Darcy regions in the multi-scale predictive LB model to
simulate the whole sample.

RESULTS AND DISCUSSIONS

Two digital porous plate experiments of the rock sample shown in Fig. 1 and 2 were
carried out (Case A and Case B) to validate the multi-scale model. In Case A, the bottom
surface of the rock sample is attached to an oil reservoir and the top surface is attached to
a water reservoir. In Case B, the left-front surface is attached to an oil reservoir and the
right-back surface is attached to a water reservoir. The other side are walls with the
properties of the rock. The multi-scale predictor was using the system with 603 grids
points depicted in Fig.4 and the single-scale LBM solver was applied to the rock cube
with 6003 grids depicted in Fig.3. The interfacial tension was set to 0.03 Nm-1. Strongly
water wet conditions were assumed and a contact angle of 30 degree was used in the
simulations. The sample is initially filled with water, then a drainage with oil as invading
fluid follows and finally an imbibition with water as invading fluid is performed.

Fig. 3. Oil distribution after imbibition Fig.4 Oil distribution after imbibition from the multi-scale
from the fully resolved LBM simulation simulation for case A. Left Pc curves are from the separated
for case A.
high-resolution simulations in two Darcy regions.

Figs. 3 and 4 show the residual oil distributions after imbibition for Case A. The trapped
oil distributions in regions 3 and 5 for the two simulations are very similar. The oil in
pores of these two regions are almost completely trapped since the pore body to throat
ratio is very large for this synthetic rock sample. As is shown in [10, 11] the residual oil
saturation highly correlated with the pore body to throat ratio. The average pore body
radius in the largest-scale pores is 67 microns and the pore throat radii in the two regions
with small-scale pores are 4.6 and 9.2 microns respectively. The effective pore body to

SCA2015-054

5/7

throat ratio is 14.6 for the imbibition process in regions 4 and 5, and 7.3 for the
imbibition process in regions 2 and 3. The trapped oil in the two small-scale pore regions
is shown in Fig.3 for the fully resolved LBM simulation. In the multi-scale simulation the
amount of residual oil in the two unresolved regions, colored in Darcy in Fig.4, are given
by the residual oil saturation in the Pc functions shown in Fig.4 which were computed by
separate fine scale simulations. Fig.5 shows the capillary pressures as functions of water
saturations of the two simulations in Case A. Each region has a similar porosity of about
0.4, so that each region 1,3,4,5 contributes 17% and region 2 contributions 33% to the
pore volume. After an incremental fill-up of region 1 (Sw=0.83..1), the breakthrough of
oil in the second region is clearly visible. The third region has large throats and no
incremental pressure is needed to enter that region. The second and third region
(Sw=0.33..0.83) are filled after the breakthrough in the second region. The region 4 and 5
(Sw=0..0.33) are filled as the pressure further increases. The amounts of residual oil and
the whole Pc curves of the two simulations agree very well. Fig.6 shows the capillary
pressures as functions of water saturations of the two simulations in Case B. The two
simulations produce fairly consistent capillary pressures. The first plateau of the pressure
curve in the drainage part (Sw=0.5..1) indicates the breakthrough of oil in the largest
pores in the first, the third and the fifth regions. Then a breakthrough in the second region
(Sw=0.17..0.5) occurs and finally in region 4 with the smallest pore throats. The amounts
of residual oil obtained from the two simulations agree very well. The amount of residual
oil in Case B is smaller than in Case A since the flow direction in Case B is parallel to the
layers and in Case A perpendicular. The imbibition process first happens in small-scale
pore regions and then the large-scale pore regions. The effective pore body to throat ratio
inside these regions is about 1.5. For Case A the effective ratio is much larger as
discussed earlier.

Fig. 5 Capillary pressures as functions of water Fig. 6 Capillary pressures as functions of water
saturations in Case A of the digital porous-plate saturations in Case B of the digital porous-plate
experiments.
experiments.

SCA2015-054

6/7

CONCLUSION
We have developed a dynamic multi-scale predictive model to determine two-phase flow
distributions at certain capillary pressure for complex multi-scale rocks. The model was
validated by two digital porous plate experiments of a synthetic multi-scale rock sample.
The results of the multi-scale mode in a coarse grid agree well with these of a latticeBoltzmann simulation in a 10 times finer grid that resolves all pores in the rock sample.
The new multi-scale model differs from a common assembling approach that computes
regions with different resolutions independently and then upscales in a post-processing
operation to obtain the properties of the representative volume. The dynamic exchange of
information about pressures and saturations between different regions in the multi-scale
model improves the accuracy of the prediction.

REFERENCES
1. Jerauld G.R., Gas-oil relative permeability of Prudhoe bay, SPE 35718 presented at
the western regional meeting held in Anchorage, Alaska, (1996), pp 653-670. Text for
References in Times New Roman 12 Point (margins justified),
2. Suzanne, K., Hamon, G., Billiote, J. and Trocme, V., Distribution of Trapped Gas
Saturation in Heterogeneous Sandstone Reservoir, Proceedings of the Annual
Symposium of the Society of Core Analysts, SCA2001-14, Edinburgh, Scotland, UK.
3. Pentland, C., Tanino, Y., Iglauer, S., Blunt, M., Capillary Trapping in Water-Wet
Sandstones: Coreflooding Experiments and Pore-Network Modeling. Paper SPE
133798 Proceedings of SPE Annual Technical Conference and Exhibition, Florence,
Italy. 2010.
4. Spiteri, E.J., Hamon, Juanes, R., Blunt, M.J., and Orr, F.M. Jr. A New Model of
Trapping and Relative Permeability Hysteresis for All Wettability Characteristics.
SPE J. 13 (3) 277-288. SPE-96448-PA. doi: 10.2118/96448-PA.
5. Al-Futaisi1 A. and Patzek T. W., Impact of wettability alteration on two-phase flow
characteristics of sandstones: A quasi-static description, Water Resources Research,
V. 39, pp. 1042-1051, 2003.
6. Chen S, Doolen GD. Lattice Boltzmann method for fluid flows. Ann Rev Fluid Mech
1998; 30:32964.
7. Nie, X., Martys, N.S., Breakdown of Chapman-Enskog expansion and the anisotropic
effect for lattice-Boltzmann models of porous flow, Physics of Fluids, 19 (1),
011702-4, 2007.
8. A. B. D. Casseie and S. Baxter, Trans. Faraday Soc., 40, 546-551 (1944).
9. Mu, Y., Fang, Q., Baldwin, C., Toelke, J., Grader, A., Dernaika, M., and Kalam. Z.,
Drainage and imbibition capillary pressure curves of carbonate reservoir rocks by
digital rock physics. Proceedings of the Annual Symposium of the Society of Core
Analysts, SCA 2012-056, Aberdeen, Scotland, UK.
10. Nie X., Mu Y., Sungkorn R., Gundepalli V. and Toelke J., Numerical Investigation of
the Dependence of Residual Oil Saturation on Geometry, Wettability, Initial Oil
Saturation and Porosity, Proceedings of the International Symposium of the Society
of Core Analysts, SCA2014-39, Avignon, France.

SCA2015-054

7/7

11. Aissaoui A., Etude thorique et exprimentale de lhystrsis des pressions capillaires
et des permabilits relatives en vue du stockage souterrain de gaz, Thesis Ecole des
Mines de Paris (1983), 223 p.

SCA2015-055

1/6

SAND CONTROL FOR UNCONSOLIDATED HEAVY OIL

RESERVOIRS: A LABORATORY TEST PROTOCOL AND
RECENT FIELD OBSERVATIONS
U.G. Romanova, M. Piwowar and T. Ma
Weatherford Laboratories Canada, Calgary, Canada
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Sand production is one of the major challenges in oil and gas industry, in particular for
unconsolidated heavy oil reservoirs. Achieving maximum well productivity without
excessive sand production or plugging in such deposits requires new technical solutions
and better understanding of reservoirs and sand control phenomena.
A laboratory protocol test has been developed for the sand control design for the heavy
oil deposits located in unconsolidated reservoirs, both clastic and carbonate. The protocol
allows evaluation of the different type of the sand control media (slotted liners, gravel
packs, wire wrapped screens and premium screens). The protocol is used in sand control
studies for thermal production operations and water disposal wells. Recent field data
indicate that results of the laboratory tests are in a good agreement with field
observations. The paper provides details of the laboratory test protocol and recent field
data for Western Canada.

INTRODUCTION
Production of formation sand is one of the oldest problems in oil and gas industry [1].
Heavy, extra heavy oil and bitumen are often found in the unconsolidated, shallow,
geologically young formations. Thermal technology based on steam injection such Cyclic
Steam Stimulation (CSS) and Steam Assisted Gravity Drainage (SAGD) is effective for
in situ production of heavy oil crude. As large volumes of water are utilized in such
operations, disposal wells are often required. Sand control media which will allow
maximum flow rates and minimal plugging or pressure build up is needed for both
injection and production wells. Efficiency of sand control depends on a variety of factors
such as geology, fluid type, fluid viscosity, production rates, thermal effects, etc.
Slotted liners (SL), gravel packs, wire wrapped (WWS) and premium screens are
commonly used for sand control in unconsolidated heavy oil reservoirs. Straight cut (SC)
and rolled top (RT) SL, standalone or with gravel packs, are commonly used in heavy oil
and bitumen production operations. Pressure build up and lost production due to liner

SCA2015-055

2/6

plugging, corrosion and scale [2] are the main challenges for SL. WWS and premium
screens, in particular as standalone screens, are becoming more popular due to a larger
open flow area and resistance to corrosion.
Laboratory work to select sand control devices prior to well completion is required. Lost
production and additional cost associated with acidizing, perforation, re-entry and side
wells, etc. can be avoided with proper planning and laboratory testing.

LABORATORY TEST PROTOCOL

A laboratory protocol test has been developed for the sand control design for
unconsolidated heavy oil reservoirs [3, 4]. The protocol was first used to evaluate SL for
SAGD operations in unconsolidated sandstone. Later, the protocol proved to be
successful in the evaluation of WWS and premium screens for both injection and
production wells in thermal production operations other than SAGD, for both sandstone
and carbonates, and for water disposal wells.
The protocol includes the following:
1. understanding geology of the reservoir and lithological facies, with the focus on
particle size distribution;
2. obtaining typical reservoir core for each lithological facies;
3. core preparation for the study (cleaning, homogenization, etc.);
4. sand control coreflood tests; tests are conducted with a synthetic brine of the same
composition as formation water, mineral oil of the same viscosity as viscosity of
heavy oil/bitumen in situ under steam injection conditions in case of thermal
production, laboratory test rates representing field rates;
5. analysis of produced fine solids and/or post test analysis of sand control coupons
by epoxy impregnation and by means of scanning electron microscopy and X-ray
energy dispersive spectrometry.
In a coreflood test, pressure differential across the sand pack, on top of the sand control
media and the amount of produced fine solids versus production rates are measured. A
schematic of the test apparatus is provided in Figure 1. Performance is evaluated by
measuring weight of produced fines and increase in delta P. If more than 0.1 g of fines is
produced or more than 5 psi delta P is developed, the sand control device is judged to be
inadequate. These parameters are somewhat arbitrary and are used for the purpose to
compare relative performance of different sand control devices.
Tests results provide data which allow the best possible sand control media for a
particular reservoir and specific operational conditions to be determined. Such sand
control media would minimize pressure build up and would provide the lowest amount of
produced solids at the maximum production rates. The protocol also allows a better
understanding of the sand control phenomena. The protocol has been used extensively for
a variety of projects in Western Canada, California, Latin America and the Middle East.

SCA2015-055

3/6

Over 1000 tests have been conducted. In Western Canada, the protocol became an
industry standard [5, 6].
Axial Overburden
Pressure Source
Application Point

Water Injection Pump

Oil Injection Pump

Pressure Chamber for
Application of Axial
Overburden Pressure
Gas Injection Pump

Pressure Transducer Total Pressure

Core Holder

Axial Overburden
Piston

Sand Pack

Sand control
media coupon
(liner or screen)

Slotted Coupon
Assembly

Pressure Transducer for

Pressure Tap Inside Slot

Pressure Transducer for

Pressure Tap Above Slot

Fluid Collection and Separation

(Four Phase Oil-Water-Gas-Solids
Separation System)

Figure 1. Schematic of sand control apparatus

TYPICAL TEST RESULTS

Formation of sand bridges on top of the sand control media and their stability under
different flow conditions can be determined in the sand control experiments. Test results
show that the rules of thumb commonly used for completion design for convention oil
and gas do not apply to heavy oil. For example, instead of the rule of 2 x D10, where D10
is the grain-size diameter from the distribution scale where 10% by weight of the sand is
of a larger size and 90% is of a smaller size, to determine the aperture of SC slot, one
typically needs to use 1.8 x D10. This applies only to sand which has less than 5 wt% of
fines, particles with less than 44 micron diameter. If sand has more than 5 wt% of fines,
SL will likely fail and screens are a better option.
Typical results of sand control tests to simulate SAGD operations for unconsolidated
sandstone (oil sands) from Western Canada are provided below. Delta P on top of sand
control media versus different flow regime and production rates is shown in Figure 2.
Amount of produced solids versus different flow regime and production rates is shown in

SCA2015-055

4/6

Figure 3. The test results presented in Figures 2 and 3 are reservoir specific and can be
used as a reference only. However, the test results show a few typical trends.

Figure 2. Delta P on top of sand control media versus different flow regime and
production rates, SAGD operations, oil sands from Western Canada

Figure 3. Produced solids versus different flow regime and production rates,
SAGD operations, oil sands from Western Canada

SCA2015-055

5/6

Test results show that:

1. Type and aperture of sand control media is specific to the reservoir and
operational conditions.
2. RT outperforms SC in terms of both delta P and amount of produced fine solids.
3. WWS outperforms SC and RT SL from a plugging perspective.
4. Solids production for WWS is marginally higher than for SL at the same aperture
width but the aperture is chosen correctly, only colloidal size particles are
produced.
5. WWS designs, even with smaller aperture width, have a lower pressure drop than
SL and the smaller aperture width of the WWS results in an effective sand control
design.
6. Premium screens have in general similar performance as WWS.

RECENT FIELD OBSERVATIONS

Recent field data from heavy oil projects in Canada confirm trends observed in the
laboratory experiments. A good example is the Leismer SAGD Demonstration project of
Statoil [5, 6]. Twenty three SAGD well pairs were completed with both liners (SC for
injectors and RT for producers) and WWS. First production started in January 2011. The
majority of SL producers show a gradual increasing drawdown with time which could be
indicative of gradual plugging. Another example is the JACOS Hangingstone SAGD
project [7]. Field data collected over the period of 15 years of production show low
pressure drop for WWS and a premium screen. Nine years of production for SL show
initially good sand control and good conformance but multiple liner failures and high
pressure drops.
The laboratory test protocol allows determination of the type and aperture of the sand
control media for a specific reservoir which would provide economical production rates
without pressure build up or excessive sand production. Sand control is one of the factors
which need to be taken into account for well design. Other factors such as operational
practices and mechanical stresses also need to be taken into consideration. Work on the
correlation of field data and lab data is ongoing.

CONCLUSIONS
1. A laboratory test protocol has been developed for sand control design for
unconsolidated heavy oil deposits. The protocol allows evaluation of the different
type of the sand control media with reservoir core material at test conditions
simulating field conditions.
2. Test results provide recommendations for the sand control media which would be
the most suitable for a particular reservoir and would provide maximum flow
rates and minimal plugging or pressure build up.
3. Recent field data indicate that results of the laboratory tests are in a good
agreement with field observations.

SCA2015-055

6/6

ACKNOWLEDGEMENTS
The authors would like to express their gratitude to the late Dr. Brant Bennion for the
contribution to the study.

REFERENCES
1. Coberly, C.J., Selection of Screen Openings for Unconsolidated Sands, American
Petroleum Institute, 1937, API-37-189.
2. Romanova, U.G. and Ma, T. An Investigation of the Plugging Mechanisms in a
Slotted Liner from the Steam Assisted Gravity Operations. SPE paper 165111MS, SPE European Formation Damage Conference and Exhibition, Noordwijk,
The Netherlands, June 5-7, 2013.
3. Bennion, D.B., Ma, T., Thomas, F.B., and Romanova, U.G. Laboratory
Procedures for Optimizing the Recovery from High Temperature Thermal Heavy
Oil and Bitumen Recovery Operations. Paper 2007-206, Canadian International
Petroleum Conference, Calgary, Canada, June 12 14, 2007.
4. Bennion, D.B., Gupta, S., Gittins, S., and Hollies, D. Protocols for Slotted Liner
Design for Optimum SAGD Operations. Paper 2008-186, Canadian International
Petroleum Conference, Calgary, Canada, June 17 19, 2008.
5. Romanova, U.G., Gillespie, G., Sladic, J., Ma, T., Solvoll, T.A., and Andrews,
J.S. A Comparative Study of Wire Wrapped Screens vs. Slotted Liners for Steam
Assisted Gravity Drainage Operations. Paper WHOC14-113, World Heavy Oil
Congress 2014, New Orleans, USA, March 5-7, 2014.
6. Romanova, U.G., Gillespie, G., Sladic, J., Solvoll, T.A., Andrews, J, and
Thomson, S., A Comparative Study of Wire Wrapped Screens vs. Slotted Liners,
Leismer Demonstration Project, SPE Thermal Well Design and Integrity
Workshop, Banff, Canada, November 18 20, 2014.
7. Park, B., Back to the Future: JACOS Hangingstone SAGD Well Completion
Design, SPE Thermal Well Design and Integrity Workshop, Banff, Canada,
November 18 20, 2014.

SCA2015-056

HETEROGENEITY ANALYSIS OF OIL SANDS CORES

A. Kantzas1, S. Kryuchkov1, J. Vanegas Prada2, M. Parliament2
1
PERM Inc., 2Athabasca Oil Corporation
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Thermal recovery of bitumen in oil sands requires the ability to propagate steam within
the formation, and to have drainage of heated fluids to production wells. As such,
reservoir heterogeneities can play a very significant role in these processes. Many of
these heterogeneities are not easy to define at the scale of logging tools. The objective of
this study is to develop a work flow and set of output parameters that can provide a much
higher resolution characterization of heterogeneities present in an oil sand, based on nondestructive testing of core. Oil sands cores with variable quality of sand and shale were
acquired and tested using an ensemble of techniques including CT scanning at 1mm
intervals, continuous NMR and EM sweeps. The continuous NMR is done by moving
the core in 1 cm intervals and then de-convoluting the signal from subsequent echo trains
to extract information in high resolution. The EM sweeps are done using two different
designs of probes (one induction and one plate design). Density and porosity are
extracted from CT. Oil and water contents are extracted from NMR and CT porosity.
Resistance is extracted from the induction probe and frequency dependent capacitance is
extracted from the plate probe. All the information is then reprocessed together to extract
a number of data in a high resolution log of the core. Although the initial interest is in the
quantitative distribution of the V-shale parameter, the results can be used to characterize
heterogeneity in both porosity and fluid distributions within the core.

INTRODUCTION
When conducting experiments in oil sand cores we often experience very poor injectivity
and very low permeability to water at initial core conditions. When core selection is done
based on logs, the core intervals appear homogeneous but at the core scale, local
heterogeneities are observed. As oil and gas development moves towards less
homogeneous sands, the log resolution for heterogeneity detection is inadequate. An
alternative process is demonstrated in this paper. Several core pieces from an oil sand
well were collected to demonstrate the work flow intended in high resolution
heterogeneity detection. The core pieces represent various oil sand layers containing
inclined heterolithic stratifications (IHS) some with more sand and some with more mud,
as well as breccia and cap rock. The core pieces were slabbed and kept frozen, but the
core was found to be desiccated at the start of the tests. Thus the core had to be hydrated.
All pieces were soaked with predefined amounts of fresh water, which was dripped onto
the surface of the core. Following that, the pieces were weighted and wrapped to
maintain moisture. All core pieces underwent CT scanning, continuous NMR scanning

1/10

SCA2015-056

2/10

and two novel EM sweep procedures. Each of these modalities was analyzed
independently and the results were combined to provide an integrated characterization.
The difference in resolution of the different modalities does not allow for simultaneous
treatment of all the results. Thus, for the purposes of this work the modalities are
interpreted sequentially.

SCANNING METHODS
X-ray Tomography
The core was scanned in 1mm intervals at two energies (120 and 100kV) [1, 2].
However, dual energy contrast was not sufficient for atomic number characterization.
Thus only the high energy scans (proportional to density) were used in the interpretations
of this study.
CT image analysis creates density maps for each slice. The density numbers for each
pixel of a given slice are plotted and a density distribution is generated for each slice. In
a uniform sand (40% porosity) this distribution is a Gaussian with a mean at around
2,000kg/m3 and a fairly narrow standard deviation (approximately 20kg/m3). Lower
densities denote more porous sands and at some point lower densities denote sand parting
(or coring induced dilation) and presence of fractures. When higher densities are present
in the distributions, depending on the shape and size of the density areas, one can get
boundaries for shale (clay laminations), pebbles, and inclusions such as siderite or pyrite.
The precise cut-off points depend on the machine used and the size of the core sample.
Thus it is very difficult to provide unilateral numbers. The most precise method for
identification of each density domain (sand, clay, pebble and siderite) would be through
statistical deconvolution, coupled with pairing against other data (e.g. particle size
analysis). This method is currently under development. For the current report the
simplification of cut-off points (used commonly in NMR for phase determination) is
used. After the cut-offs are determined, the density images are reprocessed and the four
domains are created. Currently, the sand is split in two domains sand and the rest as
shale. V-shale is determined as:
=

(1)

Vshale is then plotted as a function of depth.

Low Field Magnetic Resonance
An NMR scanning system for full diameter core was created for testing native cores and
also for core used in full diameter core floods. The system does not have gradients for
NMR imaging. Instead spectra and T2 relaxation time distributions are measured. The
system used in this study scans core at 30cm intervals (NMR sensitivity area sweet
spot). In order to generate higher resolution measurements, the core was scanned, moved
in 1cm intervals and scanned again. Thus in two successive NMR scans, 29cm of the
core within the sweet spot would be the same and 1cm would be different.

SCA2015-056

3/10

A proprietary algorithm was created that takes these successive scans and de-convolutes
them to provide spectra of approximately 3cm intervals. These spectra are then further
decomposed to provide amplitudes of oil, clay bound and capillary bound water. When
this is done, the NMR estimate of Vshale is calculated by normalizing the clay bound water
amplitude to the total amplitude as follows:

+ +( /)

(2)

Given the appropriate cut-off points the only unknown is the value of the relative
hydrogen index (RHI) which denotes the relative signal strength of oil as compared to the
same volume of water. If the reservoir oil is homogeneous with depth then an oil sample
can be tested and the RHI number can be extracted. However, if the oil is not
homogeneous then Equation (2) is not sufficient to provide an accurate Vshale. As an
alternative, a clay indicator (CI) can easily be created through Equation (3).
=

(3)

Small values of CI (CI0) indicate sandy environment, while large values (CI1)
indicate shale.
Electromagnetic Sweep Measurements
With the increasing popularity (at least in the research world) of electromagnetic heating
technologies for oil sands heating and bitumen recovery [3], there has been an increased
effort in understanding the dielectric properties of reservoir rocks and in particular oil
sands. When a rock sample is exposed to electrodes and then a variable frequency sweep
is applied on to it, measurements of capacitance and resistance can be made. Given
appropriate calibrations, these measurements can be translated into resistivity and
dielectric constant measurements as functions of frequency. If the electrodes cannot
touch the core, then the resistivity measurement becomes more problematic, but there are
still ways to calculate dielectric constant [4].
When a partially water saturated sand is exposed to an EM sweep, the resulting
conductivity and dielectric constant dispersions with frequency are recorded. The
conductivity (real component of complex impedance) dispersion is relatively flat and is
greatly affected by the amount and salinity of the water present in the core. The
dielectric constant (extractable from the imaginary component of complex impedance) is
also affected by the same parameters but the dispersion is much stronger. This is the
principle of the new dielectric logs [5] but the frequencies used for the EM sweeps in this
work are a very small fraction of the log sweeps. When it comes to clay, conductivity
dispersions should be different than sands and the dielectric constant dispersions should
be distinctly different than those of sand.

SCA2015-056

4/10

In order to extract conductivity and dielectric constant from resistance and capacitance,
details of what is the equivalent circuit of the core and the environment must be
constructed [6, 7]. This aspect for a complete core with vertical heterogeneity is still
under investigation. Thus in this paper data of capacitance and resistance are presented.
In order to capture resistance data, an induction coil probe was designed and
implemented. In order to capture capacitance data, a parallel plate probe was designed
and implemented. All the core pieces were scanned at the maximum spatial resolution
possible of approximately 1cm per measurement. The frequency range used was between
0.08MHz and 20MHz. The useful data were between 0.1MHz and 10MHz.
The Core Log Suite
When combining all the recovered data, and allowing for adaptations that will give
comparable resolutions, the following information can be obtained.
CT Scanning: Density profile, separation in different domains, with knowledge of grain
density per domain Porosity profile. From density profile and domain identification VShale profile.
NMR Scanning: Water amplitude profile and subsequently (through combination with
porosity) Water Saturation profile. Oil amplitude profile. If RHI is known then Oil
Saturation and V-Shale can be calculated but at this point RHI is considered unknown.
The Geometric mean relaxation times from water and oil phases are calculated.
Combination of CT and NMR: From the water saturation profile and the porosity profile
the Oil Saturation profile is extracted. From oil amplitude profile and using the oil
saturation profile and V-shale profile the RHI profile is extracted. From the Oil
Amplitude geometric mean relaxation time and RHI profile the Oil Viscosity profile [8]
is extracted.
EM Sweep: Resistance and Capacitance profiles.
capacitance profile and clay content.

There is correlation between

RESULTS
Figure 1 shows the photos and CT reconstructions of several pieces of core. It is evident
that the photography, no matter how high quality, does not capture all the heterogeneities
of the samples. This is especially evident in oil-bearing regions of the core, which show
up as simple black uniform-looking zones in the photographs. Figure 2 shows the
detailed NMR spectrum of a sample of bitumen and water representative of the area. The
fastest relaxing peak represents oil, the slowest relaxing peak represents water in bulk and
the smaller peaks in between represent various water in oil emulsion droplets.
Figures 5 and 6 show the calculated shale information from CT (Equation 1) and NMR
(Equation 3) for the four core pieces. The bulk heterogeneity levels are captured. The
porosity variability was captured as expected and in not shown here for brevity. The fact

SCA2015-056

that the core is not perfectly preserved does not allow for reservoir predictions but
demonstrates clearly the capability of the technique. Equation 3 rather than equation 2 is
used here for many reasons. In the clay zones in this core, the NMR clay water content
increases and the capillary water drops. When comparing CT and NMR we get
qualitative agreement but quantitative agreement was not always achieved. Trend wise
CT and NMR calculations are aligned. In the case of cap rock (piece 9) and IHS, both
CT and NMR are successful. CT with its finer resolution provides better description of
the IHS. NMR does not have the spatial resolution. However, NMR shows high Clay
Index in areas (particularly in piece 16A), where CT shows nothing. Utilization of
equation 2 would be much better but the inter echo time values used in full diameter coils
is not sufficient to extract enough points for a clear description of the oil peak of the
equivalents of Figures 2-4. Alternative approaches are sought. The NMR signal of high
clay content shown in piece 16A is believed to be correct. But since the scanner does not
pick up distinct laminations we believe that the clay is more dispersed rather than
layered.
Figures 5 and 6 show two samples of capacitance measurements from the EM sweeps. It
can be seen that sand and shale have distinctively different capacitance signatures. The
observed scatter of the experimental data below 80 kHz is due to experimental limitations
and the data at those frequencies are not used. By looking at the capacitance profiles we
can detect two distinct regions which can be fit in simple power functions. From the
presented examples we can see that the slopes of these functions are different for IHS and
oil sand. However, since the salinity is not independently measured it is difficult to
quantify the results into a predictive tool. Figure 7 shows a few such distinct profiles
from different parts of the core samples. The signatures are distinct and can be collected
in 1cm intervals. However, at this point there is not much more that can be said.
Moreover, it is noted that the core areas that are expected to have dispersed clay have a
capacitance signature that is a mixture of the two distinct shale vs. oil sand signatures
shown in Figure 7. Work in this area continues.
The major drawback of this work was that the core was dehydrated to begin with and that
did not allow for more quantitative water/salinity based measurements. It will be very
productive to repeat such measurements in fresh core. Currently direct contact with the
core provides much more accurate measurements but this is difficult to achieve and
guarantee that oil sand core integrity will be maintained. The smaller the core diameter
the easier the smaller the coil that can be used. Subsequently the inter echo time can be
faster and the method will be able to capture more of the bitumen signature (relaxing
faster than 1ms and even below 0.1ms) for a better description of the fluid distribution.
The capacitance measurements show great promise even though they are at much lower
frequencies than those of the dielectric logs.

5/10

SCA2015-056

CONCLUSIONS

Single energy CT scanning with density profile deconvolution provides detailed Vsand vs. V-Shale discrimination.
Discrimination of clay bound water vs. capillary bound water provides a second Vsand vs. V-Shale discrimination.
The EM sweeps indicate that shale and sand have distinctly different capacitance
responses with varying frequency.

ACKNOWLEDGEMENTS
The contributions of Jim Wang, Jon Bryan, Abdi Yussuf, Jennifer Amaechi and Jessica
Butron of PERM Inc. are gratefully appreciated. Funding for this project from Athabasca
Oil Corporation is also gratefully appreciated.

REFERENCES
1. Kantzas, A. Investigation of Physical Properties of Porous Rocks and Fluid Flow
Phenomena in Porous Media Using Computer Assisted Tomography, (1990), In Situ,
14(1), 77-132.
2. Kantzas, A., Marentette, D.F. and Jha, K.N., Computer Assisted Tomography: From
Qualitative Visualization to Quantitative Core Analysis, (1992), J. Can. Pet. Tech.,
31(9), 48-56.
3. McPherson, R.G., Chute, F.S. and Vermeulen, F.E., Recovery of Athabasca Bitumen
with the Electromagnetic Flood (EMF) Process, (1985), J. Can. Pet. Tech., 24,
4. Taherian, M.R., Kenyon, W.E. and Safinya, K.A., (1990), Measurement of
Dielectric Response of Water-Saturated Rocks, Geophysics, 55(12), 1530-1541
5. Hizen, M., Budan, H., Deville, B., Faivre, O., Mosse, L., Simon, M., Dielectric
dispersion; a new wireline petrophysical measurement, (2008), SPE116130 SPE
ATCE Colorado, USA
6. Mazzagatti, R.P., Dowling, D.J., Sims, J.C., Bussian, A.E. and Simpson, R.S.,
Laboratory Measurement of Dielectric Constant near 20 MHz, (1983), SPE 12097,
presented at the 58th Annual Technical Conference and Exhibition held in San
Francisco, CA
7. Josh, M., Esteban, L., Delle Diane, C., Sarout, J., Dewhurst, D.N., Clenell, M.B.,
Laboratory characterization of shale properties, (2012) J. Pet. Sci. Eng., 88-89,
p.107-124.
8. Bryan, J., Moon, D. and Kantzas, A., In-situ Viscosity of Oil Sands Using Low Field
NMR, CIM 2003-107, (2005), J. Can. Pet. Tech., 44(9), 23-30

6/10

SCA2015-056

7/10

Core 9

Core 14

Core 16-1

Core 16-2

Figure 1: Core photos and corresponding CT image sequences

400000
Produced Oil Emulsion

350000

Oil
T2gm=0.5867ms
RHI=0.19

Amplitude

300000
250000
200000
150000
100000
50000
0

0.1

10
100
T2 Relaxation, ms

1000

10000

Figure 2: T2 relaxation distribution of produced bitumen emulsion

Average Amplitude distribution Core 16-1

250000

Core 16-1

OIL
200000
WATER IN CLAY
150000

Amplitude, AU

WATER IN SAND

100000

50000

0.1

10
100
Relaxation time, ms

1000

10000

Figure 3: T2 relaxation distribution of oil rich core

SCA2015-056

8/10

Average Amplitude distribution Core 9

140000
120000

Core 9

OIL

100000

WATER IN CLAY
WATER IN SAND

Amplitude, AU

80000
60000
40000
20000
0

0.1

10
100
Relaxation time, ms

1000

10000

Figure 4: T2 relaxation distribution of clay rich core

Shale Indicator, Core14-1

Shale indicators, Core 9-1
0

0.2

0.4

0.6

0
0.8

0.2

0.4

0.6

0.8

20
20

40
40

Core length, cm

NMR Clay Indicator

V-Shale CT

80

Core Lengh, cm

60

60

Clay Indicator
V-Shale CT
80

100

100

120

120

140

140

Figure 5: CT V-Shale and NMR Clay Indicator Comparison Cores 9 and 14

SCA2015-056

9/10

Shale Indicator, Core 16-1

0.2

0.4

0.6

0.8

Shale Indicator, Core16-2

0
0

0.2

0.4

0.6

0.8

20
20

40
40

60

80

Water, g/cm
V-Shale CT

Core Length, cm

Core Length, cm

60

Clay Indicator
80

V-Shale CT

100
100

120
120

140
140

Figure 6: CT V-Shale and NMR Clay Indicator Comparison Cores 16A and 16B

SCA2015-056

40

10/10

30

Core 16-1 Bottom

C ||, pF

C ||, pF

25

35
30

20

25

15

20

y = 231.03x-0.175

R = 0.9978
High Frequency Power

10

15

Low Frequency Power

Power (High Frequency Power)

10
5

10000

100000

Freq, Hz

1000000

Power (Low Frequency Power)

10000

10000000

100000

Freq, Hz

1000000

Figure 7: EM sweep and power fit from an oil sand sample

C ||, pF

40

50

35
30

y = 60.752x-0.047

25

R = 0.9937

20

40
30
y = 600.39x-0.198

15

R = 0.9928

High Frequency Power

Low Frequency Power

10

Power (Low Frequency Power)

10000

100000

Freq, Hz

20
10

Power (High Frequency Power)

Core 16-1 top

C ||, pF

1000000

10000000

10000

100000

Freq, Hz

1000000

10000000

Figure 8: EM sweep and power fit from an inter-bedded shale sample

C ||, pF

Core 16-1 Shale

Core 16-1 Sand

Core 9 Caprock-Shale

Core 14 Sand

40
35
30
25
20
15
10
5
0

10000

100000

1000000
Freq, Hz

Figure 9: Capacitance comparison for shale vs. sand

10000000

y = 67.668x-0.095
R = 0.9683
10000000

SCA2015-057

1/6

CORRELATION OF COMPOSITIONAL PHASES AND

WETTABILITY-BASED FLUID-TRACERS IN FINEGRAINED RESERVOIRS
Qinhong Hu1; Jamil J. Clarke2; Rod Baird2; Chad Ostrander2; Zhiye Gao3; Yuxiang
Zhang1; Kibria Golam1
1
The University of Texas at Arlington, Arlington, TX 76019, USA.
2
Hitachi High Technologies America, Inc., Clarksburg, MD 20871, USA.
3
State Key Laboratory of Petroleum Resources and Prospecting, China University of
Petroleum, Beijing 102249, China.
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns, Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Fine-grained reservoirs can contain distinct compositional phases, with oil-wetting
organic matter/carbonate and other water-wetting mineral phases. Despite their
importance in hydrocarbon storage and production, the intertwined wettability, nanopore
distribution, and pore connectivity of these phases are not well understood. The approach
presented here involves tracers in two fluids (API brine and n-decane) that are used to
specifically interrogate the spatial wettability of organic-matter and mineral phases,
through an integrated approach of phase imaging by Scanning Electron Microscopy
(SEM) and tracer mapping with laser ablation-inductively coupled plasma-mass
spectrometry (LA-ICP-MS). A core sample from Eagle Ford was used for a testing area
of 2.1mm1.8mm that was argon-ion milled and imaged with a Hitachi SU3500 SEM
with image montage (Zig-Zag) capabilities for compositional distribution. The same
sample area was then treated with tracers in brine and n-decane fluids, followed with LAICP-MS mapping for tracer distribution at different spatial resolutions (8m, 12m, and
75m). A negative correlation between brine-tracers (ReO4- and Eu3+) and n-decanetracer of 1-iododecane was observed.

INTRODUCTION
Since 2000, the technological advances of horizontal drilling and hydraulic fracturing in
North America have led to a dramatic increase in hydrocarbon (gas and oil) production
from shale formations, changing the energy landscape in the U.S.A. and worldwide.
However, the sustainable shale resource development is implicated by the steep
hydrocarbon production decline and overall low recovery. For example, the top five US
resource plays typically produced 8095% less gas after 3 years, and the productivity of
new wells in two leading tight-oil plays (Bakken and Eagle Ford) dropped by about 60%
within the first year [1]. Total gas recovery from the Barnett, the most developed shale
play, was reported to be only 815% of gas-in-place in 2002, and 1230% in 2012 [2].
The recovery rate for tight oil is even lower at 510% [3].

SCA2015-057

2/6

Hydrocarbon storage and production in hydraulically-stimulated shale are affected by the

intertwined correlation of compositional phases, wettability, nanopore distribution, and
pore connectivity of shales. As shales contain distinct oil-wetting organic-matter (e.g.,
kerogen) and water-wetting mineral phases, the novel method of unique tracers in two
fluids to interrogate the wettability and connectivity of organic-matter and mineral pore
spaces is realized in this work. The organic fluid (n-decane) is expected to be
preferentially attracted to the hydrophobic component (e.g. organic particle) of the shale
matrix, with reported sizes ranging from less than 1m to tens of m [4]. Organic grains
are found to be dispersed through the matrix of shales such as Barnett [2], but their
connection through surrounding mineral phases is unknown, despite its implication in
steep initial decline and low overall recovery.
This paper presents our collaborative studies of correlating compositional phases from
SEM imaging (performed at Hitachi High Technologies America of Clarksburg, MD)
with wettability-tracers and laser ablation-ICP-MS (LA-ICP-MS) elemental mapping at
the University of Texas at Arlington.

PROCEDURES
In this work, we used a proprietary core sample from Eagle Ford to develop the technical
approaches and showcase the findings. The sample was cut into a rectangular bar with a
width and length of ~2.2 mm and height of ~10 mm. The sample surface was first milled
with a Hitachi Ar-ion milling system (Model IM4000, Hitachi High Technologies
America, Inc.). At a maximum milling rate of 300 m/h for silicon, this broad ion milling
system was used to prepare a high-quality cross-section. An area of 4.867mm2 was
processed by two 4-hr mill sessions at a voltage acceleration energy of 6 kV with a stage
swing of 40 (Fig. 1A). The area of 2,147m1,799m (the shaded region of Fig. 1B)
was then used for producing an image montage by Zig-Zag function of a Hitachi SU3500
SEM. The SU3500 SEM has a variable-pressure design that eliminates the need for
coating non-conductive materials for image acquisition. An acceleration energy of 5kV
was applied to the SEM for acquiring images at a magnification of 1,000 times. Each
image was captured at a resolution of 2,5601,920, with 8-bit grayscale, 512512 dpi,
and 121.3 m horizontal field width (Fig. 1C). The image montage was composed of 460
images that were stitched together, and the total time to acquire the entire image tile array
was ~7 hrs with the integrated Zig-Zag function which automates the entire process. The
image tiles were acquired along with a 20% margin of overlap, and this margin is needed
for the stitching software to align each image tile on all four sides. In addition, this
margin also includes some mechanical stage movement backlash which is why the
margin percent is needed to compensate for this deficit. Therefore, each image tile is not
a true side by side acquisition but a series of semi-overlapping images.

SCA2015-057

3/6

Figure 1. Work flow of Ar ion milling, SEM imaging, and elemental mapping of applied wettabilitytracers (scale bars are shown for each graph).

We used 0.1 mM each of NaReO4 and EuBr36H2O (Sigma-Aldrich Co., St. Louis, MO)
dissolved in API brine (8% NaCl and 2% CaCl2 by weight, with perrhenate anion (ReO4-)
and europium cation (Eu3+) as non-sorbing and sorbing tracers, respectively. For organicphase n-decane tracer, organic-I (1-iododecane) [CH3(CH2)9I, molecular weight of
268.18 g/mol] (>99% pure, Sigma-Aldrich Co.] was directly added to n-decane fluid at a
volume ratio of 1%. The elements rhenium (Re), europium (Eu), and iodine (I) of these
tracer chemicals are readily detected by LAICPMS.
To investigate the association of wettability to compositional phases, we applied one drop
(0.16 mL) of brine tracer solution in the middle of an imaged area, and waited for an hour
for the beaded-up fluid to slowly evaporate and spread on and within the sample. Next,
an additional drop of brine tracer (Fig. 2A-B) was applied. After another hour, some
localized zones of salt precipitates were observed and removed with a razor blade (Fig.
2C-D). Two drops of n-decane tracers were then applied, in a similar fashion as brine at
two settings; n-decane was found to quickly spread on the sample surface (Figure not
shown), which was much different from API brine. After seven hours that the 1st
application of brine tracer was dropped, LAICPMS mapping analyses were performed
on the sample (Fig. 1D).

SCA2015-057

4/6

Figure 2. Spreading behavior of traced-API brine on an Eagle Ford sample (a scale bar is shown on each
picture).

The laser ablation system (UP-213, New Wave; Freemont, CA) used a 213 nm laser to
vaporize a hole in the rock sample at sub-micron depth for each laser pulse; elements
entrained in the vapor were analyzed with ICP-MS (PerkinElmer/SCIEX ELAN DRC II;
Sheldon, CT). For different research purposes, this LA-ICP-MS approach can generate 2D and 3-D maps of chemical distributions in rock at a spatial resolution of microns, and a
concentration limit of low-mg/kg [5]. For this work, two mapping-area approaches were
used. The first, and a wider, mapping area (2,175m1,800m; called Grid A) nearly
matched the SEM-imaged area (2,147m1,799m), and we used 75m as the laser spot
size and spacing between spots; the laser was fired for 60 pulses at each spot (Fig. 1D).
This produced a total of 750 data spots, a duration of 10 hours for completion. The 2nd,
and a finer, grid (Grid B) was located near the middle of the mapped SEM region, and
was selected at 500m below, and to the right of, the starting point of Grid A (Fig. 1D).
This finer grid covered an area of 612m240m, and used smaller spot sizes and larger
laser pulses (8m and 400 pulses; 12m and 200 pulses) in order to collect more sample
mass for ICP-MS detection; a total of 26 hrs was taken to complete the mapping of Grid
B. During the data processing, the differences in spot size and laser pulse were corrected.

RESULTS
LA-ICP-MS elemental mapping results for Grids A and B are shown as Figs. 3-4. For
these figures, the scale bars indicate the measured elemental background level of the
analyte in the sample (as the lowest value, and cold color, in scale bar) and detected high
concentrations of tracer element (as the highest value, and warm color, in scale bar).
Therefore, any color difference in these figures indicates the presence of tracers.

SCA2015-057

5/6

Figure 3. Tracer distribution in Grid A.

Figure 4. Tracer distribution in Grid B with higher mapping resolutions; the 1st 240 m (from the left) and
2nd 360 m zones were mapped at 8 m and 12 m spatial resolutions, respectively.

DISCUSSION
The Eagle Ford sample has a high carbonate content [6], which was confirmed from the
SEM images showing an abundant presence of coccospheres (Fig. 1C). Carbonate rock
tends to exhibit an oil-wetting characteristics[7], consistent with our observed spreading
behavior of API brine and n-decane. The application of wettability-based tracers and
resultant LA-ICP-MS mapping further illustrate the spatial variability of wettability
regions, on the scale of microns. Fig. 3 was mapped at the scale of 75 m and shows
relatively less spotty behavior of high-concentration tracer regions than Fig. 4 at finer (8
and 12 m mapping resolutions). Furthermore, for both Figs. 3-4, brine tracers of ReO4and Eu3+ follow the similar pattern, while n-decane tracer of organic 1-iododecane shows
higher concentrations at different regions from brine tracers. The overall results illustrate
the presence of spatially variable water-wetting mineral phases and oil-wetting phase
(carbonate for this Eagle Ford sample), which is consistent with other tests (imbibition,

SCA2015-057

6/6

diffusion, and vacuum-saturation for edge-accessible pore spaces) of the sample (results
not shown).

CONCLUSIONS
Using an integrated approach of SEM phase imaging and wettability-based tracer
mapping of LA-ICP-MS, we illustrate the seemingly negative correlation of API brinebased tracers with n-decane tracer to indicate the spatial variability, at microns scale, of
water- and oil-wetting compositional phases of an Eagle Ford sample. Using these
developed approaches, additional tests will be conducted for shale samples with a range
of total organic content to interrogate the wettability patterns, and associated pore sizes,
of compositional phases of shale. The integrated studies on intertwined correlation of
compositional phases, wettability, nanopore distribution, and connectivity of fine-grained
reservoirs will help understand the hydrocarbon storage and production in stimulated
reservoirs.

ACKNOWLEDGEMENTS
Funding for this project was partially provided by the Foundation of State Key
Laboratory of Petroleum Resources and Prospecting, China University of Petroleum,
Beijing.

REFERENCES
1. Hughes, J.D. 2013. Drill, Baby, Drill: Can Unconventional Fuels Usher in a New Era
of Energy Abundance? Post Carbon Institute, 2013, 178 pp.
2. Hu, Q.H., and R.P. Ewing. Integrated Experimental and Modeling Approaches to
Studying the Fracture-Matrix Interaction in Gas Recovery from Barnett Shale. Final
Report, Research Partnership to Secure Energy for America (RPSEA), National
Energy Technology Laboratory, Department of Energy, 2014, 80 pp.
3. Hoffman, T. Comparison of various gases for enhanced oil recovery from shale oil
reservoirs. This paper was presented for presentation at the Eighteenth SPE Improved
Oil Recovery Symposium held in Tulsa, OK, USA, 14-18 April 2012, SPE 154329.
4. Loucks, R.G., R.M. Reed, S.C. Ruppel and D.M. Jarvie. 2009. Morphology, genesis,
and distribution of nanometer-scale pores in siliceous mudstones of the Mississippian
Barnett Shale. J. Sed. Res., 2009, 79(11-12), 848-861.
5. Hu, Q.H., T.J. Kneafsey, J.S.Y. Wang, L. Tomutsa, and J.J. Roberts. 2004.
Characterizing unsaturated diffusion in porous tuff gravels. Vadose Zone J., 2004,
3(4), 1425-1438.
6. Slatt, R.M., N.R. O'Brien, A. Miceli Romero, H.H. Rodriguez. Eagle Ford condensed
section and its oil and gas storage and flow potential. AAPG Annual Convention and
Exhibition, Long Beach, California, April 22-25, 2012. Search and Discovery Article
#80245, 2012.
7. Roychaudhuri, B., T.T. Tsotsis, and K. Jessen. An experimental investigation of
spontaneous imbibition in gas shales. J. Petro. Sci. and Eng., 2013, 111, 87-97.

SCA2015-058

1/7

THRESHOLD PRESSURE IN TIGHT GAS RESERVOIRS

OF CENTRAL EUROPEAN FORMATIONS
Budak P., Cicha-Szot R., Leniak G., Majkrzak M., Such P.
Instytut Nafty i Gazu - Pastwowy Instytut Badawczy, Krakow, Poland
This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Increasing development activities for tight gas reservoirs during recent years have
stimulated basic research on the flow characteristics of this kind of reservoirs.
Investigated tight gas reservoirs are characterized by low porosity, low permeability and
occasionally high water saturation which impact flow characteristics of gas reservoirs. In
this kind of reservoirs so called nonlinear flow or non-Darcy flow occurs. Despite recent
progress, the flow characteristics of tight gas reservoirs have not been thoroughly
investigated. One of the important parameter to characterize reservoir in the context of
production and basin modeling is threshold capillary pressure [1].
Threshold pressure is defined as the ability of a porous medium saturated with a wetting
phase to block the flow of a non-wetting phase. Its value corresponds to the size of the
largest pore throat in the porous medium. Therefore, the pressure difference between the
non-wetting and the wetting phase must exceed the threshold pressure before the nonwetting phase can start draining the porous medium and flow.
There are several methods for evaluating threshold pressure which has their own
advantages and disadvantages [2]. In this study, we present the comparison of estimated
threshold pressure from corrected high pressure MICP (Mercury Injection Capillary
Pressure) and measured directly by displacement methods. The application of the detailed
integrated petrophysical and petrographic data obtained by CAMI (Computerized
Analysis of Microscopic Images) show the main factors which affect the values of
obtained threshold pressures.

INTRODUCTION
The main target for tight gas exploration in Central Europe is the Rotliegend eolian and
fluvial sandstones. The possibility of this type of gas deposits are mainly associated with
eolian sandstone complexes with originally fair porosity values. Simulations of reservoir
parameters of the Rotliegend sandstone indicate porosity up to 12 % on the burial depth
up to 4500-5500 m [3]. It can be associated with the development of secondary porosity
formed as a result of the dissolution and/or transformation of cement and detrital grains
(feldspars), often leading to inversion of reservoir properties [4]. Although, this formation
has been studied for a long time determination of correlation between time of gas

SCA2015-058

2/7

saturation and time of partial loss of porosity and permeability mainly due to burial is still
key issue [5]. Accumulations of tight gas might exist also in the Cambrian sandstones of
the East European Craton. Although, numbers of diagenesis alterations, mainly quartz
cementation, affect vertical and lateral heterogeneity of reservoir properties, prospects are
promising for this structure mainly due to: 1) the spread over a wide area, 2) a simple
tectonic structure (Baltic Basin), 3) an increase in thickness in the area of gas window.
The Rotliegend and Cambrian sandstones are characterized by low porosity, low or ultralow permeability and high water saturation which lead to flow characteristics
significantly different from that the ones in conventional gas reservoirs.
Threshold Pressure is defined as the overpressure needed for the non wetting phase to
start flowing against the capillary forces. If the fluid flow is a linear, the pressure is
converted into a pressure per unit length and is called Threshold Pressure Gradient
(TPG). As one of the most important parameter to characterize flow, TPG has been
studied extensively for a long time and several approaches for TPG estimation (like
mercury intrusion, continuous injection, step-by-step, residual capillary pressure,
dynamic threshold) were introduced that give results with good medium or poor
accuracy.
In this study we compare two tight reservoir rocks in order to show which parameters
should be taken into consideration before choosing proper approach for TPG estimation.

EXPERIMENTAL
There are several laboratory methods used for evaluation threshold pressure gradient or
threshold pressure each having its advantages and disadvantages. In this study, mercury
intrusion, continuous flow and step-by-step approaches were applied to estimate
threshold pressure which give results with good or medium accuracy. As Egermann
(2006) reported mercury intrusion approach ignores the influence of the overburden
pressure and dry sample is used which may affect pore space properties. Appropriate
synthetic brines for each kind of rock samples, i.e. for each reservoir type, were used. For
MICP based method raw data were corrected according to the gas/brine interface and
temperature by using in calculations values of IFT measured for the specific system
brine/rock taking into account reservoir conditions [6].
In order to see the impact of sample conditions on threshold pressure continuous flow and
step-by-step approach was used. All above mentioned methods were extensively
described elsewhere [2,7,8]. Accuracy of the measurement was 0.07 kPa for flow
methods and 0.01% for MICP.
Experiments were carried out using plugs and cuttings from Cambrian and Rotliegend
sandstone reservoirs which are typical tight gas reservoirs located in Central Europe.
Selected properties of cores used in this study are shown in Table 1.

SCA2015-058

3/7

RESULTS AND DISCUSSION

The most significant difference between analysed reservoir rocks is porosity which affect
other petrophysical properties.
Generally, Cambrian sandstones (quartz arenites) are composed of quartz grains in 95 %
and are strongly cemented by quartz (quartz overgrowths and quartz basic cement),
resulting in almost total destruction of porosity (Fig.1). Some of Cambrian sandstones
(samples 10-15 Table 1) show fracture permeability. The fractures are either empty or
filled with clays/mudstone, quartz or bitumens. The content of pores >1 m is between 8
- 94% and practically there is no microporosity in that rocks.
Pore space of the Rotliegend sandstone (lithic, sublithic, subarcose arenites) is much
more complicated and consists mainly of intergranular pores and variable number of
micropores (Fig. 1). These sandstones are composed mainly of quartz, feldspars and
fragments of rocks. Cement is represented by ferruginous-clay overgrowths, quartz,
calcite and anhydrite (basic cement). There is a great diversity in the mineral
compositions which affects the rocks and cement as well as diagenesis processes
(dissolution, crystallization) and the creation of different amounts of micropores in the
analyzed rocks (the content of pores <1 m is between 44 - 89%).
Table 1 Specification of cores

Cambrian

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24

Formation

Rotliegend Sandstone

Core #

Sample depth Plug length

[m]
[cm]
2324.60
2327.20
2327.80
2328.70
2330.40
2331.20
2333.20
2339.70
2340.90
2282.40
2289.50
2293.00
2290.10
2292.20
2299.00
4245.15
4351.25
4359.50
4460.60
4551.00
4558.25
4642.20
4650.25
4656.40

5.43
5.50
5.32
5.28
5.53
5.53
5.55
5.20
5.17
5.23
5.37
5.16
5.23
5.04
4.95
4.59
5.04
5.19
4.92
5.03
4.44
5.52
4.67
5.05

Porosity
[%]
3.86
4.39
6.25
3.74
5.31
3.11
3.54
2.29
2.17
4.03
5.50
3.45
3.77
4.33
6.02
3.89
5.52
6.18
13.75
7.09
3.23
12.02
14.04
9.83

Avarage
capillary
[m]
2.53
0.47
0.40
0.68
0.89
0.37
0.90
0.15
0.17
0.13
1.13
0.42
0.17
0.11
0.64
0.09
0.09
0.12
0.23
0.11
0.05
0.16
0.19
0.12

Specific area
[m2/g]
0.02
0.15
0.25
0.08
0.09
0.11
0.06
0.24
0.20
0.45
0.08
0.12
0.14
0.47
0.10
0.69
0.99
0.84
1.04
1.03
0.98
1.28
1.30
1.37

Treshold
diameter
[m]
12.0
1.2
1.0
4.5
4.5
0.9
4.5
1.2
0.9
0.9
3.8
2.0
0.9
0.4
6.0
1.0
2.0
3.0
5.0
4.0
0.4
4.0
3.0
3.0

Permeability
[mD]
0.63
0.08
0.03
0.11
0.32
0.02
0.24
0.06
0.02
0.01
0.24
0.16
0.07
0.40
0.01
0.49
0.38
0.29
0.85
0.13
0.13
4.33
1.18
0.76

In Cambrian sandstones, the results of TPG obtained by mercury intrusion and step-bystep approach were quite similar for samples with average values of porosity for this
formation. The highest discrepancy was in the case of samples with relatively high

SCA2015-058

4/7

microporosity (surface area ca. 0.5 m2/g) where mercury intrusion overestimates values
of TPG. Underestimation of TPG applying MCIP approach occurs when samples are
mesoporous (average capillary over 0.6 m). These general observations were confirmed
by direct correlation of TPG and surface area and inverse correlation with average
capillary radius (Table 2).
In the case of samples 10 15, the highest discrepancy between analysed methods was
observed. It may be caused by distinct potential pathways for hydrocarbon migration in
this samples which are presumably fractures together with microstylolites.
In the case of Rotliegend sandstone, an inverse correlation of TPG with depth was found
what suggest that, except compaction, diagenesis processes and consequently secondary
porosity affect flow through such reservoir (Table 4). Effect of reservoir conditions
during TPG estimation increases with decreasing porosity and specific area. In such cases
flow methods are more relevant. Underestimation of threshold pressure using mercury
injection approach may occur also in the case of very heterogeneous parts of the reservoir
especially when very thin beds of various petrophysical properties exist (sample 19).

Figure 1. Selected thin sections (sample impregnated with blue resin) and pore size distribution for
samples of Cambrian sandstone (sample 14, 15) and Rotliegend sandstone (sample 21 and 17). In the
case of Cambrian rocks intergranular pores are observed. The Rotliegend sandstone shows diverse
porosity with quite complex microporosity which steers the values of threshold pressure.

SCA2015-058

5/7

Table 2. Correlation coefficients of petrophysical parameters with threshold pressure for Cambrian
sandstones
Depth
Pth MICP
Pth flow
Pth step-by-step
TPG flow
TPG step-by-step

-0.23
0.72
0.53
0.71
0.58

% of
Treshold Absolute
Pth
pores
diameter permeability MICP
>1 m
-0.11
-0.68
-0.70
1.00
-0.31
-0.30
-0.41
0.14
-0.30
0.12
-0.84
0.64
-0.32
-0.31
-0.41
0.14
-0.34
0.09
-0.91
0.62

Effective Average Specific

porosity capillary surface
-0.27
0.07
-0.48
0.06
-0.99

-0.62
-0.25
-0.81
-0.26
-0.94

0.82
0.06
0.70
0.07
0.69

Pth
flow

Pth
step-bystep

1.00
0.18
1.00
0.30

1.00
0.35
1.00

TPG
flow

1.00
0.68

Table 3. Correlation coefficients of petrophysical parameters with threshold pressure for Cambrian
sandstones (with no fractures Samples 1-9)
Depth
Pth MICP
Pth flow
TPG flow

0.44
0.35
0.38

Effective Average
porosity capillary
-0.18
-0.72
0.01
-0.65
-0.01
-0.65

Specific
surface
0.76
0.92
0.93

% of pores
>1 m
-0.77
-0.91
-0.91

Treshold
diameter
-0.83
-0.88
-0.88

Absolute
permeability
-0.80
-0.72
-0.72

Pth
MICP
1.00
0.84
0.83

Table 4. Correlation coefficients of petrophysical parameters with threshold pressure for Rotliegend
sandstones
Depth
Pth MICP
Pth flow
Pth step-by-step
TPG flow
TPG step-by-step

-0.69
-0.17
-0.92
-0.64
-0.91

Effective Average Specific

porosity capillary surface
-0.67
-0.34
-0.85
-0.76
-0.84

-0.57
-0.40
-0.57
-0.66
-0.56

-0.64
-0.06
-0.93
-0.56
-0.92

% of
pores
>1 m
-0.63
-0.28
-0.61
-0.52
-0.60

Treshold
Absolute
Pth
diameter permeability MICP
-0.89
-0.19
-0.80
-0.43
-0.79

-0.26
0.15
-0.74
-0.20
-0.71

1.00
0.33
0.95
0.52
0.96

Pth
flow
1.00
0.46
0.72
0.44

Pth
step-bystep

1.00
0.80
1.00

TPG
flow

1.00
0.80

*In the Tables 1 3 above: direct correlation blue, indirect correlation red, no correlation black. If the
correlation coefficient is close to 1, it would indicate that the variables are positively linearly related (both variables
are increasing or decreasing) and the scatter plot falls almost along a straight line with positive slope. For -1, it
indicates that the variables are negatively linearly related (one variable is increasing, second decreasing and vice
versa) and the scatter plot almost falls along a straight line with negative slope. Correlation coefficient equal 0
indicates no linear relationship between the variables.

On Figures 2 and 3 the results of TPG calculations obtained on the basis of using MICP
and flow methods are presented. Figure 2 shows the cross plot for Cambrian sandstones
(with no fractures) and Figure 3 the cross plot for Rotliegend sandstones. Squares of
correlation coefficients for both data sets show good correlation of results.

Figure 2. TPG flow vs. TPG MICP

for Cambrian sandstones

Figure 3. TPG flow vs. TPG MICP

for Rotliegend sandstones

SCA2015-058

6/7

CONCLUSION
The analyses conducted in tight sandstones show completely different structure of the
pore space due to mineral composition and diagenesis processes which has a strong
impact on threshold pressure and consequently on its gradient.
Observed weaker correlations with continuous flow method may be caused by limitation
of laboratory equipment (too high minimum flow rate), which generally implies higher
values of threshold pressure.
In Cambrian sandstones direct correlation between depth, specific surface and threshold
pressure obtained by flow and step-by step method was observed. Estimated correlation
between specific surface and threshold pressure for Cambrian rocks without fractures
(samples 1 9, Table 3) is not typical. Probably, this direct correlation is caused by very
low value of specific surface (below 0.2 m2/g) and suggests slight shift of pore size
distribution to lower pore diameters of mesopores.
Moreover, indirect correlations of average capillary, percent of pores > 1 m and Pth
(MICP) and Pth as well as TPG (step-by step) were found.
In the Rotliegend sandstones, indirect correlations between depth, pore space parameters
and threshold pressures and their gradients (measured using MICP and step by step
method) were observed. Correlation between depth and other analysed parameters shows
that evolution of pore space is dependent not only on compaction but also on diagenesis
processes (development of secondary porosity).
Obtained results and correlation showed that in the case of Cambrian sandstones without
microfactures, mercury intrusion gives very reliable results and might be used as a
standard method. For this type of sample, low value of specific surface is a good
indicator for choosing MICP approach. In the case of Rotliegend sandstone with more
complex pore space distribution, the mercury intrusion approach underestimates the
threshold pressure values and another method with higher accuracy should be used to
achieve more reliable results.

SCA2015-058

7/7

ACKNOWLEDGEMENTS
The research leading to these results was partially founded from the Polish-Norwegian Research
Programme operated by the National Centre for Research and Development under the Norwegian
Financial Mechanism 2009 - 2014 in the frame of Project Contract No Pol-Nor//196923/49/2013.

REFERENCES
1. Djaouti K., Egermann P., Hennebelle P. An intergrated approach to estimate threshold
capillary pressure from core and log data, paper prepared for presentation at the
International Symposium of the Society of Core Analysts held in Napa Valley, California,
USA, 16 - 19 September 2013
2. Egermann P., Lombard J.M., Bretonnier P. A fast and accurate method to measure
threshold capillary pressure of caprocks under representative conditions, paper
prepared for presentation at the International Symposium of the Society of Core
Analysts held in Trondheim, Norway 12-16 September, 2006
3. Leniak G., Such P., Sota-Valim M. Quantitative porosity and permeability
characterization of potential Rotliegend tight gas reservoirs, Polish Geological Review,
(2010) 58, 4, 345351.
4. Poprawa P., Kiersnowski H., Potential for shale gas and tight gas exploration in
Poland, Biuletyn PIG-NRI, (2008), 429, 145-152.
5. Schwarzer D., Littke R., Petroleum generation and migration in the Tight gas area
of the German Rotliegend natural gas play: a basin modelling study,
Petroleum Geoscience, (2007) 13, 1, 37-62.
6. Cicha-Szot R., Steinbach A., Marsh M., Gelb J., Linden S., Wiegmann A., Glatt E.,
Leniak G., Survey of recent advances in digital rock physics benefits of DRP and
their application to reservoir characterization, paper prepared for presentation at the
Geopetrol Conference held in Zakopane-Kocielisko, Poland 15 -18 September 2014
7. Hildenbrand A., Ghanizadeh A., Krooss B.M. Transport properties ofunconventional
gas systems, Marine and Petroleum Geology, (2012) 31, 1, 90 - 99.
8. Hildenbrand A., Schlmer S., Kroos M. Gas breakthrough experiments on fine grained
sedimentary rocks, Geofluids, (2002) 2, 1, 3-23.

SCA2015-059

1/7

ROCK SURFACE CHARACTERIZATION BY

IMMERSION CALORIMETRY: WETTABILITY AND
SPECIFIC SURFACE AREA
E. Dyshlyuk1, D. Korobkov2, V. Pletneva2
Schlumberger, Dahran Carbonate Research and 2-Moscow Research Center
1-

This paper was prepared for presentation at the International Symposium of the Society of Core
Analysts held in St. Johns Newfoundland and Labrador, Canada, 16-21 August, 2015

ABSTRACT
Petrophysical properties such as wettability and surface area control hydrocarbon storage
and transport properties. In the work being reported in this paper, we present the
theoretical aspects as well as experimental technique and results of testing and validating
a proposed calorimetry technique for specific surface area and wettability
characterization. Identical immersion calorimetry experiments with slightly different
initial conditions of the sample are analyzed to obtain these two properties of a core
sample. Validation of the experimental technique is an important step, which is
performed in this work by comparing the surface areas of the samples obtained by the
proposed technique with the surface area obtained by the well-known Brunauer-EmmettTaylor (BET) gas-adsorption technique. Surface area measurements are performed on a
set of core plugs and reference samples, including hydrophilic and hydrophobic powders
of pure substances, minerals, and clays.

INTRODUCTION
This paper is a continuation and enhancement of our previous studies on rock
characterization by calorimetry [1, 2]. The differential scanning calorimetry (DSC)
technique, which is used in the current study, allows one to measure transient differential
heat flows between the two cells of a differential scanning calorimeter. Implementation
of the DSC technique to laboratory core analysis can provide a significant amount of
valuable information with unprecedented accuracy. In particular, the DSC technique can
help to determine wettability heterogeneity of a core sample from adsorption calorimetry
[2] pore size distribution from the thermoporometry method, as well as the surface area
and average wettability state of a rock from immersion calorimetry [1].
In the immersion calorimetry experiments, a core sample is immersed in liquid and the
associated small heat effect (immersion enthalpy) is measured with a calorimeter. The
heat effect is related to the alteration of the surface energy of the rock surface during the
immersion process. Immersion experiments are commonly conducted with the use of a
sealed glass ampule containing the sample, which is broken inside the calorimeter cell
filled with liquid. After the ampule is broken, liquid fills the surface of the sample; thus,
changing the surface energy of the sample. Depending on the choice of initial conditions

SCA2015-059

2/7

of the rock surface, the associated heat effect is related either to the wettability state of
the surface or to the surface area of the sample. As a result, two important petrophysical
parameters can be measured by an accurate thermodynamic technique with calorimetry.
The total heat that evolves, H, during an immersion experiment or the enthalpy of
immersion is related to the variation of the Gibbs free energy (G) of the system by the
following expression: H S G T G / T , where S is the sample surface area and
T is temperature. The variation of the Gibbs energy of the system is in fact equal to the
variation of the surface energy of the system. If a solid is immersed from vacuum
conditions, then G S SL , where S is the solid-vacuum interface surface tension
and SL is the solid-liquid interface surface tension (Figure 1a). At the same time, if a
liquid film is on the surface of the sample prior to immersion, the variation of the Gibbs
energy is given by G SL LV SL LV (Figure 1b), where LV is the liquidvapor interfacial surface energy/surface tension.

G S SL

G SL LV SL LV

(a) Wettability characterization: cos

(b) Surface area characterization: S

Figure 1. Illustration of the two immersion experiments(a) immersion from vacuum for wettability
characterization;(b) immersion after precoverage of the sample with a liquid film for surface area
characterization.

The first experiment (solid is immersed from vacuum conditions) can be used to
characterize the wettability of a sample [1, 3, 4]. Since the surface tension of a liquid,
LV , (e.g., water surface tension ~ 72 mN/m at 25C) and its variation with temperature,

LV T , can be measured or are known from the literature, the second experiment (also
known as the modified Harkins-Jura procedure [5], which was developed after the
original work [6]) can be used for sample surface area measurements as follows :
S H

LV

T LV T .

(1)

Although the assessment of wettability of a core sample is the primary target of these
experiments, it is quite difficult to perform any other quantitative experiments on
wettability measurements to compare the results. Thus, standard core analysis methods
such as USBM, Amott, or their modifications do not provide information on the contact
angle, while the contact angle method is difficult to implement on porous media. Because
the surface area measurement is much easier to quantify by other methods, we decided to
validate this technique first by quantitative comparison of surface areas of different

SCA2015-059

3/7

samples measured by the proposed method and the well-known BET gas adsorption
technique [7]. The next step of our procedure would be to compare wettability obtained
by this method with other known methods.

EXPERIMENT DESCRIPTION AND RESULTS

Immersion experiments are commonly conducted with the use of a sealed glass ampoule
containing the sample, which is broken inside the cell that contains the immersion liquid.
Another variant of the cell is a membrane cell [8], which consists of two compartments
divided by thin metal foil; the upper compartment contains liquid and the lower
compartment contains the sample. During the experiment, the membrane is broken by a
rod and the liquid wets the sample. For each of these configurations, it is necessary to
take into account additional thermal effects appearing in the experiment due to breaking
of the ampoule or rupture of the membrane. The disadvantage of using these cells for our
future applications in the determination of wettability is that they are not capable of
working in high-pressure environments, which is required for wettability experiments at
reservoir conditions. To overcome this problem, a special calorimetry cell was developed
[9].

Figure 2. Experimental cell design: 1 external tube; 2 internal moveable tube; 3 wetting liquid
compartment; 4 friction-less valve; 5 sample compartment.

The special calorimetry cell provides: 1) a means for connecting the sample to the
vacuum/ vapor system for pretreatment the sample; i.e., vacuum the sample and/or adsorb
the liquid film on the surface; 2) both the immersion liquid and the sample had to be in
the same calorimetric cell during the experiment in order to exclude mass-heat transfer
during immersion of the sample; 3) for the heat effect due to combining both volumes of

SCA2015-059

4/7

the sample and the liquid so as to not influence the results and be reproducible and for
being systematically accounted.
The designed cell (Figure 2) is equipped with two coaxially arranged tubes (1 and 2 in
Figure 2) (1/16-in. tube into a 1/8-in. tube, respectively), independently connected to
external volumes. The internal tube is used for evacuation of the sample and vapor
adsorption, while the external tube is used for filing of the cell with the wetting liquid.
The inner tube can move in the vertical direction; i.e., when it is in the bottom position,
the sample is in a contact with a line for evacuation and vapor adsorption, but when it is
lifted up vertically (with a special pneumatic device), the cell compartments are
connected and the liquid fills the sample volume through the valve. The sample
compartment is a 9-mm diameter, 20-mm tall cylinder.
Prior to the experiment, a sample in the lower compartment should be dried in a vacuum
under a specific temperature ( typically 85C) to remove any loosely adsorbed vapors
from the surface. An experiment is initiated under a zero pressure condition. The
preliminary evacuated cell containing the sample to be studied is filled with the wetting
liquid vapor to create continuous film on the sample surface. The heat effect due to vapor
adsorption can be used for surface wettability heterogeneity evaluation [2] (not described
here). The immersion step is now completed, resulting in the second heat effect peak
(Figure 3), which is used for surface area assessment in accordance with Eq. (1).

Figure 3. Adsorption-immersion heat flow curve

It is important with the proposed method that the absorbed liquid layer be sufficiently
thick so that the liquid layer screens the surface from the other molecules. In this case, the
variation of the Gibbs energy during the immersion step is equal to LV and the
interpretation of the results of experiments with Eq. (1) is valid. This condition is fulfilled

SCA2015-059

5/7

when the thickness of the adsorbed layer is equal to approximately 1.5 monolayers of
adsorbed species [5]. When using water as a liquid, this condition is most often fulfilled
for hydrophilic samples if the liquid layer is adsorbed with humidity of ~ 45 % [5]. At the
same time, this condition might not hold with the hydrophobic samples or when the
adsorption isotherm of a hydrophilic sample is starting very slowly during the first half of
the adsorption isotherm, as for example probably the case with the carbonate samples [2].
All of the experiments performed in this work were with distilled water, and the liquid
film is adsorbed from the 45% humidity atmosphere, which is created by an external
water-filled and temperature-controlled tank.
Surface area measurements with calorimetry were performed on artificial and natural
powders and carbonate rocks (Table 1). The measurement time at each step of ordinary
adsorption-immersion experiment was approximately 2 hours. In the case of clays, we
had to increase the experimental time (three to six times until the baseline stabilized) for
both the adsorption and immersion step to measure complete adsorption and immersion
heat effects accurately. The specific surface area of the porous samples was also
measured by the BET method (some by two independent laboratories to evaluate the
measurement precision).
As shown in Table 1, the correspondence between the measured specific surface areas is
very good for a majority of the samples (including some mesoporous controlled pore
glass CPG powders) except for those samples marked in gray. Poor correspondence in the
surface areas is indicated with the bentonite powder, most likely due to the swelling of
bentonite in water. Thus, the BET method is able to measure only the external surface of
the bentonite, while the calorimetry method can also measure internal surface of the clay.
From the opposite point of view, we see a good correspondence between the
montmorillonite sample measurements. Montmorillonite is also a swelling clay (actually,
the main constituent of bentonite) and one would expect (at least we did) it to behave
similar to the bentonite behavior. Our present conclusion is that the montmorillonite
powder sample could have gone through an irreversible change in its structure due to
eventual water vapor adsorption/condensation. Now, the full surface, including internal
one, is exposed to both of the methods.
As was explained previously, proposed method cannot be successfully implemented with
hydrophobic samples (carbon) or with samples having a low adsorption of vapor during
the first part of the adsorption isotherm (carbonate samples). It is difficult to reach some
exact conclusions from the data of these carbonate core sample experiments because the
specific surface areas of these samples are very small and consequently, the accuracy of
both the BET and calorimetry experiments is not overly high. For now, we can observe
that both techniques give the same order of magnitude values and show the same
systematic variation; i.e., an increase in the surface area after extraction of the cores. The
difference between the measurements with cores can be also explained by the presence of
clays (about 5%) and halite (about 5%).

SCA2015-059

6/7

CONCLUSION
Immersion calorimetry experiments allow measuring the surface area of a core sample
and its wettability state by two identical experiments with slightly different initial
conditions. A special calorimetry cell was developed to allow for determining core
sample wettability at reservoir conditions in this experimental work and for future tests.
Although understanding wettability is a primary goal of the ongoing study, in the present
phase of this work, validation of the measurement technique was required. It is difficult
to validate wettability measurements by independent tests; therefore, the current work
focused on the validation of the technique by comparing the surface areas of different
artificial and natural samples as well as actual core samples. The values for comparison
were obtained by a well-known BET gas adsorption technique. Respectable
correspondence between the measured surface areas was obtained for most hydrophilic
samples, which proves the accuracy of the procedure. Hydrophobic samples and
carbonate samples show a more complex behavior, which is anticipated from the theory
of the method. Special care should be taken in the future tests during the measurements of
samples that contain a considerable amount of swelling clays. Future tests will include
analogous experiments with nonpolar hydrocarbons, which will hopefully allow us to
overcome the mentioned hindrances when implementing the method for surface area
measurements, followed by experiments on core sample wettability measurements.

REFERENCES
1. E. Dyshlyuk, "Application of Differential Scanning Calorimetry to Core Analysis," in
Society of Core Analysts, SCA2013-055, Napa Valley, California, USA, 2013.
2. E. Dyshlyuk, N. Bogdanovich, D. Korobkov and V. Pletneva, "The measurements of
core sample surface wettability heterogeneity by adsorption calorimetry," in Society of
Core Analysts, SCA2014-088, Avignon, France, 2014.
3. E. Dyshlyuk, "Method for Determining Wettability of a Surface". Russia Patent Claim
# 2014109083, February 2015.
4. E. Dyshlyuk, "Method for Reservoir Wettability Properties Determination". Russia
Patent 2497098, 2013.
5. S. Partyka, F. Rouquerol and J. Rouquerol, "Calorimetric determination of surface
areas Possibilities of modified Harkins and Jura procedure," Journal of Colloid and
Interface Science, vol. 68, pp. 21-32, January 1979.
6. W. Harkins and G. Jura, "An Absolute Method for the Determination of the Area of a
Finaly Divided Crystalline Solid," Am. Chem. Soc., vol. 66, pp. 1362-1366, 1944.
7. S. Brunauer, P. Emmett and E. Teller, "Adsorption of Gases in Multimolecular
Layers," J. Am. Chem. Soc., vol. 60, p. 309, 1938.
8. P. Aukett, "A New Membrane Cell for the Determination of Heats of Immersion Using
the Setaram C-80 Microcalorimeter," J. Thermal Analysis, vol.33, p.323, 1988.
9. E. Dyshlyuk and V. Baldygin, "Device for Measurement of Heats of Adsorption and
Wetting". Russia Patent 2524414, 2014.

SCA2015-059

7/7

Table 1. Specific surface area of powders and core plugs obtained by different methods.

Surface area,
Surface area
from
by immersion
manufacture,
calorimetry,
m2/g
m2/g
Reference samples (artificial and natural powders)
CPG 3000C (pore : 300 nm), borosilicate glass
8.4
9.3
CPG 1000C (pore : 100 nm), borosilicate glass
26.6
23
CPG 500C (pore : 50 nm), borosilicate glass
49.5
43
Corundum, Al203
155
197
Corundum, Al203
11.3
Powder and core samples name, material

Corundum, Al203

11.5

Quartz, SiO2
6.0
Kaolinite
10.3
Montmorillonite
265
Bentonite
250
Calcium carbonate, CaCO3
1.2
Carbon, C
1.0
Core plugs, carbonate rock, before/after extraction (described in [2])
Sample 1
0.20 / 0.54
Sample 2
0.07 / 0.40
Sample 3
0.57 / 0.62
Sample 4
0.63 / NA
Sample 5
NA / 0,63
* BET measurements by two independent laboratories

Surface
area by
BET,
m2/g
9.5
27
54
153
12.4 14.5*
12.2 14.0*
6.9 - 8.3*
10
232
15.6
3.8
1.75
0.04 / 0.27
0.00 / 0.30
0.18 / 0.33
0.20 / 0.28
0.12 / 0.24