J Appl Electrochem (2011) 41:1175–1182

DOI 10.1007/s10800-011-0335-7

ORIGINAL PAPER

Polarization curve analysis of all-vanadium redox flow batteries

Doug Aaron • Zhijiang Tang •
Alexander B. Papandrew • Thomas A. Zawodzinski

Received: 20 May 2011 / Accepted: 8 July 2011 / Published online: 18 August 2011
Ó Springer Science+Business Media B.V. 2011

Abstract We outline the analysis of performance of intermittent power sources, such as wind and photovoltaic
redox flow batteries (RFBs) using polarization curves. This cells, and for loading leveling for the electrical grid [4–6].
method allows the researcher immediate access to sources A commonly investigated RFB chemistry is the all-
of performance losses in flow batteries operating at steady vanadium system, for which sulfuric acid solutions of the
state. We provide guidance on ‘best practices’ for use of V2?/V3? and V4?/V5? (present as VO2?/VO2?) redox
this tool, illustrated using examples from single cells couples serve as the anolyte and catholyte, respectively.
operating as vanadium redox batteries. During battery discharge, VO2? is reduced to VO2? at the
cathode, accompanied by a concomitant oxidation of V2?
Keywords Flow battery  Polarization curve  to V3? on the anode; these reactions proceed in the
Vanadium redox battery  VRB  RFB opposite direction in the charging process.
Typical testing of modifications to RFBs involves
charge–discharge cycling to determine the voltage, charge,
1 Introduction and power efficiency. However, cycling experiments do not
provide direct information on which mechanisms result in
Redox flow batteries (RFBs) have drawn considerable RFB efficiency loss. Polarization curves are commonly
interest from energy storage researchers for a variety of used in fuel cell studies to analyze the losses in a cell.
reasons [1–3]. In contrast with batteries such as lead-acid, Polarization curves with associated cell resistance mea-
Ni–Cd and Li-ion that store charge in the solid state, charge surements provide a clear interpretive tool for identifying
in RFBs is typically stored in solution. Anolyte and catho- dominant limitations in performance. The primary losses
lyte solutions containing reversible redox couples are stored identified in fuel cells via analysis of polarization curves
separately, and passed through a flow cell for charge and include kinetic activation polarization, ohmic polarization
discharge. This effectively decouples the power density and due to DC resistance and catalyst layer mass transport
energy capacity of RFBs, resulting in flexibility in battery losses, and mass transport limitation arising from gas
system design. Currently, RFBs are widely considered a transport through gas diffusion media [10]. Kinetic acti-
promising energy storage candidate for coupling with vation polarization is a result of slow charge transfer
reactions at the interface between the electrode and elec-
trolyte and is evident at low cell operating current density.
Ohmic loss in the cell includes the resistance to ionic
transport through the electrolyte (electrolyte solution and
D. Aaron  Z. Tang  A. B. Papandrew  T. A. Zawodzinski (&)
Department of Chemical and Biomolecular Engineering, the separator-typically a polymer membrane), electrical
University of Tennessee, Knoxville, TN 37996, USA resistance in the electrodes, and contact resistance between
e-mail: [email protected] cell components. Finally, concentration polarization dom-
inates when mass transport of reactants limits the current
T. A. Zawodzinski
Physical Chemistry of Materials Group, Oak Ridge National density of a cell [11]. It should be noted that these polar-
Laboratory, Oak Ridge, TN 37831, USA ization losses in a cell are not discrete or exclusive, but that

123
1176 J Appl Electrochem (2011) 41:1175–1182

all of them contribute to loss in a cell whenever current is approximately 25%. Platinum wires (Alfa Aesar) were
flowing. However, each mechanism tends to dominate in inserted into the carbon felts through septa in the endplates
different regions of a polarization curve [12]. to serve as current collectors.
Flow batteries are similar to fuel cells in important The second cell used in this work was a modified direct
respects. Steady-state performance can be achieved in methanol fuel cell (DMFC) from Fuel Cell Technologies
RFBs given the constant supply of reagent during operation with an active area of 5 cm2, hereafter referred to as the ‘fuel
at a stoichiometric flow rate above that demanded by the cell battery’ (FCB). The FCB uses Poco graphite plates with
current density. Nonetheless, polarization curves are un- single-serpentine flow fields (0.787 mm wide and 1.02 mm
derutilized in RFB studies. In this contribution, we explore deep) for electrolyte delivery and gold-plated aluminum
the application of this simple yet powerful tool to unrav- current collectors. Toray carbon paper (0.200 mm thickness,
eling the performance limitations of vanadium redox cells. 0.08 X cm through-plane resistivity) or carbon felt (ini-
We illustrate the method and provide some guidance on its tially *2.5 mm thick, compressed to *0.500 mm thick,
use to extract maximum information from experimental 0.15 X cm uncompressed, through-plane resistivity) served
studies. as electrodes. We denote the FCB sub-configurations with
carbon paper or carbon felt electrodes FCB-P and FCB-F,
respectively. In both cases, NafionÒ117 was used as the
2 Method membrane material.
Figure 1 includes schematics of the FCB (Fig. 1a) and
2.1 Cell construction the SB (Fig. 1b) indicating the relative sizes of the elec-
trodes and membranes.
Two single-cell versions of VRB configurations were
utilized in this work. The first was a simple battery (SB) 2.2 Electrolyte system
with symmetric carbon felt electrodes housed in PVC
compartments. Square PVC pipe formed the electrode An all-vanadium electrolyte was used in this work. 99.9%
compartments, and PVC plates were used as endplates. VOSO4 (Alfa Aesar) was dissolved in 2.0 or 5.0 M H2SO4,
Viton gaskets were placed between all parts of the cell to at a concentration of 0.5 or 1.0 M, respectively. Both sides
ensure no leakage of electrolyte. Polypropylene compres- of the VRB were initially loaded with the V4? solution.
sion fittings (1/800 NPT 9 00 OD tube) were used to The first charging step converted V4? to V3? and V5? in
connect the electrolyte feed lines to the electrode com- the negative and positive electrode compartment respec-
partments. A 20 cm2 Nafion 117 membrane (Ion Power) tively. The V5? solution was removed and replaced with an
was used as the separator. Two pieces of graphitic carbon equal amount of V4? solution, and the VRB was charged
felt (CeraMaterials, 1.27 cm thick, 0.12 X cm uncom- again to reach the ‘‘fully charged’’ state consisting of V2?
pressed through-plane resistivity) were stacked in each on the anode side and V5? on the positive side. Diaphragm
electrode resulting in a total, uncompressed electrode metering pumps (STEPDOS 08 from KNF Lab) delivered
thickness of 2.54 cm. However, because the PVC electrode electrolyte to the batteries at a flow rate ranging up to
compartment was intentionally made slightly under- 30 mL/min. A nitrogen purge was fed to the negative
sized, the carbon felt electrodes were compressed by electrode reservoir (containing V2? and V3?) to avoid

Fig. 1 Schematics of the a FCB

and b SB with carbon felt
electrodes

123
J Appl Electrochem (2011) 41:1175–1182 1177

oxidation of the V2? when the battery was in a charged measurements, in the determination of sources of perfor-
state. mance loss in RFBs. This entails an explication of these
sources of loss. In general, voltage drops in the cell are
2.3 Electrochemical measurements associated with (1) electrode polarization or kinetic losses,
(2) iR losses, associated with contact and ionic resistances,
In this work, the cells were controlled via a Bio-Logic or ‘pseudo-iR’ losses, associated with mass transfer of
HCP-803 high current potentiostat/galvanostat/EIS. This redox-active species within active electrode layers and (3)
potentiostat had a maximum current of 400 mA on the mass transfer-limiting currents associated with bulk
main board and was capable of reaching 80 A via an reagent delivery to the electrode. These are listed in order
integrated booster. The working electrode lead from the of increasing overvoltage at which they become relevant
potentiostat was connected to the positive electrode (V4?/ and are illustrated in Fig. 2. We note that the two sources
V5?) of the battery while the counter electrode lead was of loss listed in item (2) could each result in a linear
connected to the VRB negative electrode. Polarization decrease in voltage with increasing current density in the
curves were generally measured using controlled current middle portion of the polarization curve. They are, how-
steps, measuring the cell potential at each steady-state ever, readily distinguished by correction of the polarization
current value. Cases employing potentiostatic control are curve voltage by iRdc where Rdc refers to the DC resistance
noted where appropriate. For a discharging polarization or HFR of the cell. The mass transfer losses within the
curve measurement, we generally started with a fully electrode do not contribute to this HFR resistance since the
charged battery. Charging polarization curves were per- electrode is electronically shorted in any DC or HFR
formed on either fully discharged or partially charged cells. measurement. We note in passing that the condition of
A steady current below 2 mA/cm2 at a cell potential of linear voltage drop from catalyst layer loss strictly occurs
1.8 V was taken to indicate ‘‘full charge.’’ The cell was only when one transport mechanism dominates the voltage
charged or discharged at the specified current for 30 s and drop. However, this linear behavior is often seen in
then allowed to rest for up to 2 min at its open circuit iR-corrected fuel cell polarization curves. Item (3) can be
voltage (OCV). Following this open circuit period, the next driven by mass transport within flow fields or within an
desired current in the polarization curve was set by the electrode structure and is generally associated with elec-
potentiostat. Cell potential measurements were averaged trode ‘starvation’, i.e. complete conversion of all available
over the 30 s of each current step to provide a point on the electroactive species at the demanded current density.
polarization curve. The electrolyte flow rate for these For analysis of RFBs, care must be taken to ensure that the
measurements ranged from 10 up to 30 mL/min. All flow rate of the electroactive electrolyte is substantially in
experiments were carried out at room temperature (25 °C), excess of any limiting currents (see below).
with no active control exerted on the temperature of the
cell or the electrolyte solutions. However, the temperature 3.1 Polarization curves for discharge
change between inlet and outlet at maximum power was
measured to be *1 °C. Thus, we can safely assume Our initial experiments focused on the SB. The electrolyte
approximately constant temperature conditions. was 0.5 M VOSO4 in 2.0 M H2SO4 fed at a flow rate of
Prior to and following each polarization curve, the high
frequency resistance (HFR) was measured for the VRB.
The potentiostat measured the HFR at 15–30 kHz using an
AC potential perturbation of 10 mV amplitude, measuring
the current response. This resistance was multiplied by the
active area of the membrane (5 or 20 cm2, depending on
the battery type) to determine the areal specific resistance
(ASR) of the cell and to iR correct the cell potential
measurements using current density. All reported current
densities are with respect to the geometric surface area of
the electrode.

3 Results and discussion

Our primary goal in this contribution is the illustration of Fig. 2 Generalized polarization curve for a VRB indicating the
the use of polarization curves, with cell resistance dominant source of overpotential in each region

123
1178 J Appl Electrochem (2011) 41:1175–1182

30 mL/min. Figure 3 shows the polarization curve results

for this experiment. This cell exhibited very little kinetic
polarization (*0.031 V drop at 10 mA/cm2), but a sub-
stantial ohmic ASR (4.57 X cm2) and a mass-transport
limited current density of 152 mA/cm2. For comparison,
well-hydrated fuel cell configurations using N117 exhibit
an ASR of roughly 0.2 X cm2. Clearly, there is a sub-
stantial excess resistance in the SB—roughly 200-fold
above that of a simple hydrated membrane. We note that a
Nafion membrane exposed to this electrolyte composition
exhibits comparatively slight changes in conductivity [13].
Nevertheless, the large iR correction applied to the curve
removes nearly all polarization, resulting in a flat corrected
curve. The predominantly ohmic nature of the voltage loss, Fig. 4 The effects of electrolyte concentration on the iR-free kinetic
given the limited ionic contribution from the membrane, region of polarization curves for the FCB-F. The electrolyte solvent in
both cases was 5.0 M H2SO4
suggests that contact resistance is the primary loss channel
in this device. We also largely discount a substantial con-
tribution to ohmic drop arising from the electronic resis-
tance of the carbon electrode material. Even for an
uncompressed electrode, the loss associated with this
source amounts to only \0.4 X cm2 given the resistivity of
material mentioned above. Upon compression, the thick-
ness is decreased as is the resistivity. It is also unsurprising
that using a simple wire as a current collector can con-
tribute such a large contact resistance. The impact of this
choice of current collector on current distribution in the
cell is unclear, but the in-plane resistivity is likely to be
substantially lower than that measured through plane.
Thus, the lateral iR drop across the electrode is likely Fig. 5 Discharging polarization curve for the FCB-P showing the
somewhat lower than that through the electrode. Further magnitude of the iR correction given an ohmic ASR of 0.66 X cm2.
study of this point is beyond the scope of this paper, The 1.0 M V/5.0 M H2SO4 electrolyte flow rate was 12 mL/min
especially since we regard this geometry to be an extreme
and not a viable approach to high-performance cells, in interpretation of the data when obtained as a polarization
spite of its prevalence in the literature. curve coupled with an HFR/ASR measurement. This
Again, the overall points of the above discussion are analysis reveals, virtually at a glance, the conclusion that
not only the cell design but also the relative ease of losses due to electrode kinetics and mass transport issues
are minimal compared to iR losses.
In experiments with both configurations of the FCB,
a higher concentration of electrolyte was used. For compari-
son purposes, the FCB-F was operated with 1.0 M VOSO4
in 5 M H2SO4 to examine whether electrolyte concentra-
tion had a significant effect on the kinetic region of the
polarization curve. Figure 4 illustrates that the kinetic
region of the polarization curve, up to *15 mA/cm2, was
largely unaffected by the increased concentration. Thus, we
consider differences in behavior between the SB and FCB
to be more due to cell architecture than to any concentra-
tion effect.
Figure 5 shows a polarization curve for the FCB-P with
a N117 membrane fed with the higher concentration
Fig. 3 Discharging polarization curve of the SB illustrating the
electrolyte (1.0 M VOSO4 in 5 M H2SO4) at 12 mL/min.
significant effect iR correction can have on the shape of the curve.
Here, the measured ohmic ASR was 4.57 X cm2. The 0.5 M V/2.0 M The electrodes in this study were the 0.200 mm thick Toray
H2SO4 electrolyte flow rate was 30 mL/min carbon papers. In this case, kinetic polarization is quite

123
J Appl Electrochem (2011) 41:1175–1182 1179

pronounced compared to the polarization curve for the

SB—a drop of 0.215 V at 10 mA/cm2 in the FCB-P
compared to 0.031 V at 10.0 mA/cm2 in the SB. In contrast
with the SB, the ohmic ASR of the FCB-P was 0.660
X cm2. Moreover, the effect of iR correction on the
polarization curve was quite different for the FCB-P
compared to the SB. In the case of the FCB-P, significant
voltage loss is evident after the iRdc correction, indicating
that the resistance to charge transport in the mixed con-
ductor is of greater magnitude than ionic or contact resis-
tances. Beyond 180 mA/cm2, the polarization of the FCB-P
becomes mass-transport controlled, reaching a diffusion-
limited current density of 250 mA/cm2. These results
suggest that performance in the FCB-P is controlled by a
combination of kinetic and active layer mass transport Fig. 6 iR-free discharging polarization curves illustrating the effect
of electrolyte flow rate on the FCB-P configuration. The electrolyte
losses. consisted of 1.0 M V/5.0 M H2SO4
The FCB-P exhibited much greater kinetic polarization
than the SB and iRdc correction did not remove all of the
ohmic polarization in the FCB-P. However, the mass significantly above 80% of full charge (see below). This
transfer limitation generally occurred at higher current points to the importance of operating at very low utilization
density in the FCB-P than it did in the SB. The greater of the overall inventory of the available redox active species
current density associated with mass transfer limitation over the course of a polarization experiment. In our small
could be due to improved electrolyte flow across the laboratory-scale set-up, we are operating with limited elec-
electrodes and membrane afforded by the graphite flow trolyte reservoirs of up to 250 mL volume. While utilization
fields in the FCB-P. of the available vanadium will lead to a drop in the OCV, the
We further investigated the effects of electrolyte trans- maximum drop in state of charge corresponds to a loss of
port on polarization in the FCB-P by systematically vary- only *30 mV. This magnitude of change does not signifi-
ing the electrolyte flow rate. These measurements were cantly alter any of the conclusions below.
performed with 1.0 M V/5 M H2SO4 electrolyte solutions, The pronounced effects of electrolyte flow rate on the
a N117 membrane, and carbon paper electrodes. The flow mass-transport limited current density was surprising and
rate was varied from 0.5 up to 25 mL/min. While all other motivated us to calculate the utilization factor for various
polarization curves in this work had step times of 30 s at a electrolyte flow rates and to compare these values with the
desired current or potential, this series of experiments observed limiting current density in the FCB-P.
required step times of 120–180 s to maintain adequate In Table 1, we present the theoretical limiting current
electrolyte circulation at low flow rates and to obtain stable density calculated by converting the delivery rate of elec-
current measurements. The results of these experiments are trolyte to the FCB-P to the number of electrons available if
shown in Fig. 6. all vanadium was converted in a single pass. The observed
As the flow rate was increased, the onset of mass
transport control occurred at progressively lower cell Table 1 Comparison of theoretical limiting current density and
potential and greater current density. Increasing the flow observed limiting current density in the FCB-P at various electrolyte
rate from 0.5 to 25 mL/min increased the limiting current flow rates
density from 40 to 325 mA/cm2, an eight-fold increase. For Flow rate Theoretical limiting Observed limiting Percent
comparison, the SB achieved a limiting current density of (mL/min) current density current density of max
approximately 165 mA/cm2 at an electrolyte circulation (mA/cm2) (mA/cm2) current
rate of 30 mL/min. 0.5 161 40 25.2
The high current region of the curves for flow rates of 20 2 643 105 16.3
and 25 mL/min displayed an unusual feature: the current 4 1287 159 12.4
density decreased as the cell potential was lowered. This is 8 2573 209 8.12
likely due to depletion of the electrolytes at very high current
12 3860 250 6.48
density. The experiments performed at 20 and 25 mL/min
16 5147 261 5.07
resulted in the state of charge ending below 80% of full
20 6433 306 4.76
charge (76 and 73%, respectively). All of the other polari-
25 8042 321 3.99
zation curves in this series ended with states of charge

123
1180 J Appl Electrochem (2011) 41:1175–1182

limiting current density values were taken from the results this system cannot account for the observed ASR. Because
shown in Fig. 6. At low flow rates, nearly 25% of all the the carbon felt electrodes were not compressed to a great
electrons available in the entering vanadium species were degree in the SB (only compressed to 75% of the uncom-
harvested in one pass through the FCB-P. However, the pressed thickness), contact resistance could be very sig-
limiting current density was quite small—40 mA/cm2. nificant in this system, as noted above. Significant contact
At the highest flow rate, single-pass utilization of the resistances may exist between the carbon felt electrodes
vanadium was much lower, with only 3.99% of the avail- and the N117 membrane and between the active electrodes
able electrons transferred to the electrode. This indicates and current collectors. The current collector was a thin
that a greater stoichiometric flow results in improved lim- platinum wire that did not share very much contact area
iting current density at the cost of utilizing very little of the with the carbon felt and lacked a robust mechanical con-
available vanadium in the electrolyte. Since the vanadium nection. If most of the ohmic polarization was primarily
is never really depleted at the electrode surface at high flow due to those two contact resistance mechanisms, such a
rate, mass transfer limitation happens at a much greater large effect from iRdc correction is reasonable. On the other
current. RFBs are generally recirculating systems, how- hand, the electrodes in the FCB (carbon paper or carbon
ever, making this only negative from the standpoint of felt) were under substantial compression against the
increased pump load. membrane and the planar flow field plate, and so contact
One striking difference evident between the SB and resistances were likely to be very small. In the case of the
FCB (regardless of FCB electrode material) is the lack of carbon paper electrodes, the gaskets around the 0.20 mm
kinetic polarization in the SB. While the carbon felt used in thick electrodes were 0.15 mm thick, enforcing compres-
both battery systems was not characterized by a technique sion of the carbon paper. Similarly, the 2.5 mm thick
such as nitrogen adsorption to determine surface area, the carbon felt electrodes had 0.500 mm of gasketing, with an
minimal kinetic polarization shown in Fig. 3 is likely to be attendant compression resulting. Since both electrodes had
due to the very large surface area afforded by *57 cm3 of substantial pressure on them (the tie-rods were tightened to
carbon felt. For comparison to the 19.1 mm electrode 11 Nm of torque), contact between the current collectors,
thickness in the SB, the electrodes used in the FCB had electrodes and membrane was expected to be quite good.
thicknesses of approximately 0.50 and 0.20 mm for the This is supported by the observation that iRdc correction
carbon felt and carbon paper, respectively. The mass of the did not remove all of the ohmic polarization in the FCB.
carbon felt electrodes in the SB was approximately 3.81 g The latter observation is itself an interesting point as it
on each side while the carbon felt electrodes in the FCB-F suggests that substantial ‘pseudo-iR’ loss is present due to
were approximately 0.10 g each. mass transport within the active layer of the FCB elec-
It is unlikely that the difference in kinetic region behavior trodes. These electrodes are relatively thin (compared to
between the SB and FCB originates from the use of different the SB electrode) but are quite thick when compared to a
electrolyte concentrations. Though the SB was operated with fuel cell electrode. Moreover, these Toray carbon papers
the less concentrated electrolyte (0.5 M VOSO4 in 2 M are wet-proofed with PTFE for the purpose of water
H2SO4) while the FCB was generally operated with the management in PEM fuel cells. In the case of RFBs, such a
higher concentration electrolyte (1.0 M VOSO4 in 5 M hydrophobic surface may prevent complete wetting of the
H2SO4). Figure 4 shows that the kinetic polarization in the electrode by the electrolyte. We thus assign the larger
FCB-F with carbon felt electrodes does not vary with elec- losses in the kinetic region to the smaller available surface
trolyte concentration. Carbon paper was used for the area in the FCB electrodes.
experiment shown in Fig. 5, and these curves show similar
kinetic polarization. Thus, regardless of electrode material 3.2 Polarization curves for charging cycles
and electrolyte concentration, substantial kinetic polariza-
tion occurred in the FCB but not in the SB. Carbon paper was In addition to discharging polarization curves, charging
not considered as electrode material in the SB since the point polarization curves have been considered. Figure 7
of the SB was to have a large, three-dimensional electrode. includes an example of charging and discharging polari-
We surmise that the observed difference is one related to zation curves for the FCB-F that have been corrected for
available surface area for the reaction. Detailed studies of iRdc. In this figure, we plotted overpotential versus current
electrode kinetics for various materials are underway and density for the convenience of the reader and to make the
will be reported elsewhere. comparison more transparent. The charging polarization
Another noticeable difference between the polarization curve was performed with electrolytes at an approximate
curve behavior between the SB and FCB was that iRdc SOC of 89% following 150 mA discharge for 2 h. The
correction removed nearly all of the ohmic polarization higher concentration electrolyte (1.0 M VOSO4 in 5 M
present in the SB. We note that the membrane resistance in H2SO4) was used at a flow rate of 10 mL/min. A N117

123
J Appl Electrochem (2011) 41:1175–1182 1181

Fig. 8 Cycling behavior of the FCB-F at 40 mA/cm2 and 30 mL/min

1.0 M V/5.0 M H2SO4 electrolyte flow rate. The voltage efficiency of
the first cycle was 79.9%

Table 2 Comparison of overpotential, ASR and limiting current for

various VRBs
Fig. 7 Absolute value of overpotential for iR-free charging and
discharging polarization curves for the FCB-F (the iR-free polariza- Battery gCell at ASR JLimiting
tion curves are included in the inset). The SOC of the battery was 10 mA/cm2 (mV) (X cm2) (mA/cm2)
100% during the discharge curve and 89% during the charge curve.
SB 31 4.57 160
The 1.0 M V/5.0 M H2SO4 electrolyte flow rate was 10 mL/min
FCB-F 184 0.65 178
membrane separated the two carbon felt electrodes. FCB-P (25 mL/min flow) 215 0.63 326
The kinetic polarizations observed were nearly identical, Ref. [15] *100 5.4 –
0.187 V at 10 mA/cm2 for the charging curve, compared to Ref. [16] 250 *3.5 –
0.184 V at 10 mA/cm2 for the discharging curve. Simi- Ref. [17] 61 – 326
larly, the ASR values measured are nearly the same; ASRs
of 0.670 and 0.650 X cm2 were measured during charge
and discharge, respectively. To add additional context to our investigations, we
Though the curves are very similar at low current sought polarization data on comparable all-vanadium
density, the charging overpotential deviates from the dis- RFBs. At this time, few such studies are publicly available.
charging overpotential at cell operating currents exceeding The three battery configurations we report had overpoten-
15 mA/cm2. The nearly identical activation overpotentials tials at 10 mA/cm2 that are similar to previously reported
and ohmic resistances observed during charge/discharge batteries [15–17], and exhibit comparable cycling charac-
suggest that ionic transport, and not charge transfer, is the teristics. The ohmic resistances reported by other investi-
source of the observed asymmetry in the data. The onset gators are very close to the values we measured for the SB,
of mass transfer limitation occurred at similar current but are nearly an order of magnitude greater than the FCB
densities for charging and discharging, though it did result values. Table 2 summarizes the key data extracted from
in a steeper curvature for the charging process than for our polarization studies of the SB, FCB-F, FCB-P and
discharging. Thus, in this case, the primary asymmetry includes applicable data from three other studies. Based
between the discharge and charge curves occurs in the on these data, we conclude that lowering activation over-
‘pseudo-iR’ region of the curve. potentials while maintaining a low ASR is a route to
In addition to our polarization studies, we also made improving the performance of RFBs. Though this may
standard battery cycling measurements. Figure 8 shows the seem a self-evident determination, we opine that it merely
cycling behavior of the FCB-F with a N117 membrane, appears so due to the clarity that polarization curve analysis
high concentration electrolyte, and carbon felt electrodes. brings to the table.
Cycling was performed at 40 mA/cm2. Over the course of We also point out that the present work was carried out
two cycles, the voltage efficiency of the FCB-F was 79.9 with two-electrode cells, as were virtually all contributions
and 78.8%, and the OCV was stable at approximately discussed in the literature. Under these conditions, it is
1.41 V, indicating little crossover through the membrane or difficult to directly infer much concerning the relative
via any cross-membrane leak. These performance metrics performance of the two electrodes. What data are available
are well in line with those reported by other investigators in the literature concerning electrode kinetics and diffu-
[2, 7–9] and indicate that the FCB is an appropriate model sivity for the various species suggest that the two elec-
system for the techniques described here. trodes will behave similarly. Yamamura et al. [18] show

123
1182 J Appl Electrochem (2011) 41:1175–1182

slightly better kinetics for the negative electrode but all design and a FCB based on modified DMFC hardware with
results discussed there appear to be highly dependent on thin electrodes and serpentine flow fields. Kinetic polari-
the exact carbon material used. Furthermore, reported dif- zation was nearly absent in the SB, but significant for all
fusion coefficients are generally similar. Different data conditions explored in the FCB, probably due to the much
taken from the literature indicates that one or the other D greater electrode surface area in the SB. In contrast, the iRdc
for different oxidation states is somewhat different from correction was dominant in the SB, but did not significantly
the others, but not systematically so. Thus, short of affect the polarization curves for the FCB. This was likely
developing and inserting a reference electrode and actually due to significant contact resistance losses in the SB while
measuring the difference or more thorough experimenta- the FCB exhibited more ‘pseudo-iR’ losses, i.e. limitations
tion to reveal physical parameters, it is difficult to speculate in ionic transport through the electrode layers, that were not
on the relative performance of the two electrodes. These addressed via iR correction. By identifying different origins
significant gaps in the literature will be addressed in of voltage loss in these ostensibly similar devices, we
subsequent contributions. demonstrate the central importance of polarization analysis
to the study of RFB systems.
3.3 Recommendations
Acknowledgments We gratefully acknowledge the Governor’s
Chair Fund from the State of Tennessee for support of this research.
The work presented here has resulted in the formulation of TZ also acknowledges the support of the Materials Science and
some ‘best practices’ that can be used as guidelines for Technology Division of the Physical Sciences Directorate at Oak
investigators beginning studies of RFBs. Ridge National Lab for support of this work through coverage of his
time. DA and ZT acknowledge SEERC and the Dept. of Chemical
(1) Galvanostatic control is preferable to potentiostatic and Biomolecular Engineering for their partial support of this work.
control for acquisition of polarization curves, result-
ing in a generally more stable response from the
device.
(2) HFR measurements at 15–30 kHz were found to most References
accurately approximate the DC resistance of our
RFBs. This value is somewhat higher than the value 1. Rychick M, Skyllas-Kazacos M (1988) J Power Sources
of *1–5 kHz used for the same measurement in 22:59–67
2. Skyllas-Kazacos M, Kasherman D, Hong DR, Kazacos M (1991)
PEM fuel cells [14].
J Power Sources 35:399–404
(3) Accurate recording of polarization curves requires 3. Dell R (2001) J Power Sources 100:2–17
electrolyte reservoirs containing sufficient volume 4. Joerissen L, Garche J, Fabjan C, Tomazic G (2004) J Power
that a polarization curve will not exhaust the stock of Sources 127:98–104
5. Ponce de Leon C, Frias-Ferrer A, Gonzalez-Garcia J, Szanto D,
charged vanadium species. Figure 6 shows potentio-
Walsh FC (2006) J Power Sources 160:716–732
static polarization curves that exhibit a decreasing 6. Fabjan C, Garche J, Harrer B, Jorissen L, Kolbeck C, Philippi F,
current density with a lowering of cell potential at the Tomazic G, Wagner F (2001) Electrochim Acta 47:825–831
higher limiting currents; this is attributed to exhaus- 7. Sun B, Skyllas-Kazacos M (1992) Electrochim Acta 37:
1253–1260
tion of the electrolyte.
8. Fang B, Wei Y, Arai T, Iwasa S, Kumagai M (2003) J Appl
(4) Polarization curve analysis does not allow the deter- Electrochem 33:197–203
mination of overpotentials at the individual electrodes 9. Sun B, Skyllas-Kazacos M (1992) Electrochim Acta 37:
without the inclusion of a reference electrode. Thus, 2459–2465
10. Arico V, Creti AS, Antonucci P, Antonucci PL (1998) Electro-
we cannot ascertain whether cell overpotential is
chem Solid State Lett 1:66–68
located primarily in the negative electrode, positive 11. Yoon KJ, Huang W, Ye G, Gopalan S, Pal UB, Seccombe DA
electrode, or is distributed between the two. For this (2007) J Electrochem Soc 154:B389
reason, reference electrodes are essential for more 12. Mench M (2008) Fuel cell engines. Wiley, Hoboken
13. Tang Z, Aaron D, Keith R, Papandrew A, Zawodzinski T Jr
detailed studies of polarization in RFBs.
(2011) Proceedings of the 220th Electrochemical Society meeting
(Submitted)
14. Cooper KR, Smith M (2006) J Power Sources 160:1088–1095
15. Kazacos M, Skyllas-Kazacos M (1989) J Electrochem Soc
4 Conclusion 136:2759–2760
16. Chen D, Wang S, Xiao M, Meng Y (2010) J Power Sources
In this work, polarization curves were utilized to analyze 195:2089–2095
17. Kjeang E, Michel R, Harrington D, Djilali N, Sinton D (2008)
the distribution of performance losses for two different all- J Am Chem Soc 130:4000–4006
vanadium RFB architectures. We considered a SB char- 18. Yamamura T, Watanabe N, Yano T, Shiokawa Y (2005)
acterized by thick carbon felt electrodes and a flow-through J Electrochem Soc 152:A830–A836

123