Available online at www.sciencedirect.

com

Acta Materialia 60 (2012) 2359–2368

www.elsevier.com/locate/actamat

Deformation and failure in nodular cast iron

A. Ghahremaninezhad, K. Ravi-Chandar ⇑
Center for Mechanics of Solids, Structures and Materials, The University of Texas at Austin, Austin, TX 78712-0235, USA

Received 5 October 2011; received in revised form 22 December 2011; accepted 22 December 2011
Available online 1 March 2012

Abstract

Ductile failure in nodular cast iron is explored through uniaxial tension and notched tension experiments. Specimens obtained
through tests interrupted at various stages of deformation and failure evolution were examined through quantitative microscopy to dis-
cern the mechanisms of failure and to quantitatively evaluate the local strain evolution. Fractographic observations were used to identify
the onset and evolution of damage processes during the deformation and failure of nodular cast iron. These tests and observations reveal
that void growth and coalescence occurred only within a narrow localized band, whose size is comparable with the size of the graphite
nodules; no statistically significant changes in the porosity were observed outside this zone.
Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.

Keywords: Damage nucleation; Failure mechanism; Porosity

1. Introduction models [4–6], and the general framework of these models is

now rather well established. Tvergaard [7] and Benzerga
The failure of many ductile materials used in structural and Leblond [8] provided comprehensive reviews of this
applications is typically considered to occur by the progres- approach to modeling of ductile failure. Calibration and val-
sive nucleation, growth and coalescence of damage. Micro- idation of these models have been attempted by numerous
scopic models of such damage are usually considered at investigators (see, for example, Brocks et al. [9], Dong
the continuum level, through the introduction of the idea et al. [10], Decamp et al. [11], Pardoen et al. [12], Steglich
of a representative volume element (RVE), where damage and Brocks [13], and Benseddiq and Imad [14]). However,
can be represented by internal variables. Homogenization this process is inherently non-unique, since many different
requires that the size of the RVE be selected so as to contain combinations of parameters appear to capture the softening
a statistically uniform representation of the damage process part of the material/structural response [15]. In this paper we
within the RVE [1,2]. The size of the RVE can vary from investigate the process of deformation and evolution of fail-
about 103 mm3 for metals and ceramics to about 106 mm3 ure in an initially porous material. This material was chosen
for concrete. The damage discontinuities in the RVE are with the goal of avoiding the difficulty of modeling the com-
considered to be “small” with respect to the size of the plex void nucleation process and examining the process of
RVE, but large compared with the atomic spacing and even void growth. For this purpose three different materials, sin-
grain size [3]. Numerous studies have been performed aimed tered iron, spheroidized sintered iron and nodular cast iron
at exploring the development of damage. The class of micro- (NCI), were considered as suitable candidates. However, the
mechanical models that incorporate void nucleation, growth sintered iron specimens could not sustain strains greater
and coalescence as the damage process are typically called than 2% and while the spheroidized sintered iron specimens
Rousselier or Gurson–Tvergaard–Needleman (GTN) could be strained up to 16%, the poorly controlled heat treat-
ment produced a highly non-homogeneous microstructure
in which no meaningful correlation could be made between
⇑ Corresponding author.
specimens interrupted at different stages of the deformation.
E-mail address: [email protected] (K. Ravi-Chandar).

1359-6454/$36.00 Ó 2011 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.actamat.2011.12.037
2360 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368

Therefore, sintered and spheroidized sintered iron specimens ultrasonic cleaner. The samples were finally etched to reveal
were not considered further in this study and attention is the grain structure. Thirty images were taken of each
restricted to the NCI specimens. sample using a Nikon Eclipse microscope at 200
This paper is organized as follows. The microstructure magnification and then stitched together to a size of
of NCI used in the present work is described in Section 2. 1.40  4.00 mm area. These images were then used to
The results of uniaxial tension tests and notched tension obtain quantitative measures of grain size, void distribution
tests are discussed in Sections 3 and 4, respectively. Partic- and other microstructural features. The unetched micro-
ular attention is paid to the development of microstructural structure and etched microstructure at a higher magnifica-
features corresponding to different macroscopic strain lev- tion of the NCI are shown in Fig. 1a and b, respectively.
els and triaxiality conditions. The conclusions from this As can be seen in the micrographs, in this hypoeutectic alloy
study regarding the growth of voids and their coalescence the graphite nodules are surrounded by grains of ferrite.
are summarized in Section 5. The dark phase identifiable outside the ferritic domains is
pearlite. The alternating layers of ferrite and cementite are
2. Material extremely fine and not resolved in this micrograph. Image
analysis software (NIH ImageJ) was used to obtain quanti-
Ductile cast iron or nodular cast iron (NCI) is a type of tative measurements of the microstructural features. It
cast iron in which the graphite nodules take on a nearly should be emphasized that even though extra caution was
spherical shape. Magnesium particles are usually added taken in polishing the specimens to protect the surface from
as an inoculant in the casting process to provide the nuclei polishing artifacts, due to the nature of the polishing pro-
from which the graphite nodules grow. The cooling time cess there exist a number of features that appear dark and
during the manufacturing process plays a very important resemble graphite nodules in the micrographs that should
role in determining the microstructure and the mechanical be excluded in order to obtain as accurate results as possi-
properties of the cast iron. Therefore, the microstructure of ble. Observations of the microstructure at higher magnifica-
different sections of the cast material varies in the size and tion showed that most of these types of impurities are
number of nodules, and hence in the mechanical behavior. smaller than 47 lm2, therefore, in all the microstructural
In order to ensure consistency in the material used in the analysis performed dark patches smaller than 47 lm2 have
experiments all the specimens were cut from the central been filtered out. A shape factor F ¼ 4pA=p2 where A and
portion of the blank. Metallographical observations and p are the area and perimeter of the nodules respectively, is
mechanical testing support consistency in the material defined to characterize the shape variation of the graphite
microstructure and material properties. nodules. A value of F ¼ 1 corresponds to a perfectly round
The nodular cast iron used in this study is an ASTM particle and the shape becomes more irregular with decreas-
A536 grade 65-45-12 (the grade code number indicates ing F . The average grain size of the ferrite and the diameter
the tensile strength (ksi), yield strength (ksi) and percent of the graphite nodules were found to be about 26  9 and
elongation at failure, respectively) ductile cast iron with a 27  25 lm, respectively. The mean spacing of the nodules
ferritic matrix that contains about 5–25% pearlite. The is about 95  60 lm.
chemical composition of the cast iron, as provided by the It has been documented in the literature [16,17] that a
supplier, is listed in Table 1. weak interface between the graphite nodules and the fer-
Metallographic samples were prepared by sectioning the ritic matrix triggers decohesion of the nodules at very low
tensile specimens in the longitudinal direction, mounting in plastic strain; however, there have also been investigations
epoxy resin, and polishing following the standard proce- that showed debonding occurring continuously over the
dures down to a final polish with 0.05 lm silica particles course of deformation [18]. We assume that the initial dis-
in colloidal suspension. Special care was exercised so as tribution of graphite nodules is equivalent to the initial dis-
not to cause flow of the material over the pores during the tribution of voids; in particular, we identify the nodule
polishing process, thereby enabling accurate measurements geometry with the void geometry.
of the porosity. In order to remove the polishing com- In order to obtain quantitative measures of the micro-
pounds from the surface the samples were cleaned in an structure, a MATLAB code was implemented to construct
a Voronoi tessellation diagram from the stitched optical
images from which the local porosities were measured
Table 1 (Fig. 2). Two measures of porosity are evaluated. The local
Chemical composition of the nodular cast iron. porosity is calculated for each pore by dividing the pore
Source: Dura-Bar, www.dura-bar.com. area Ap by the area of the assigned cell Av . The bulk poros-
Component Content (%) ity is calculated as the ratio of the total area of nodules to
Carbon 3.50–3.90 the area of the representative volume. Although the stere-
Silicon 2.25–3.00 ological characteristics of the pores are not included in
Manganese 0.15–0.35 the calculations, as pointed out by Argon and Im [19] the
Sulfur 0.025 max frequency distribution of a pore size measured on a sec-
Phosphorus 0.05 max
tioned plane is a good approximation of the density in a
 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368 2361

(a)

Graphite
nodule

Ferrite (b)

Pearlite

Fig. 1. Microstructure of the nodular cast iron in (a) the unetched and (b) the etched conditions. Different phases present in the microstructure are
identified in the lower image.

in the bulk porosity calculated as the ratio of the total area

of nodules in a sub-region to the area of the sub-region is
also included in the plot for comparison. Fig. 4b shows
Av the variation in the standard deviation of the local porosity
with RVE size. From these results it appears that above an
RVE size of about 1200 lm the mean local porosity
Ap
becomes less sensitive to the size of the RVE. The change
in standard deviation also seems to decrease at a length
scale of about 1200 lm. Based on the above observations,
a proper size of the RVE for the material at hand would
be 1200 lm. Note that this is an extremely “large” size in
comparison with the overall specimen dimensions typically
used in experiments on NCI. In the present work the spec-
Fig. 2. Voronoi tessellation diagram used in calculating the distribution of imen thickness comprises only two RVE. This estimate of
local porosity. the RVE is based entirely on the geometry of the porosity,
and does not indicate whether estimates of effective field
pffiffiffi quantities averaged with respect to this RVE could repre-
volume equal to 2 times the area of sectioned plane. sent actual material responses. This requires further exam-
However, the results reported here are not corrected for ination of the deformation and stress fields. However, all
this approximation. the microstructural analyses described in the following sec-
The length scale of a representative volume element tions were performed on an area of 1400  4000 lm2.
(RVE) plays a significant role in the determination of effec- In the following sections we examine the evolution of
tive properties of the material. In order to characterize the void size with deformation through interrupted uniaxial
effect of the length scale on the mean as well as the stan- and flat notched tension tests. While the macroscopic
dard deviation of the local porosity, sub-regions of sequen- deformation and stress in uniaxial tension tests are uniform
tially reduced dimensions (1/4, 1/9, 1/16, 1/25, 1/36, 1/49 throughout the gage length (before the formation of plastic
and 1/64 of the area of the largest region) were analyzed instabilities in the form of diffuse or localized necking) with
(see Fig. 3). The variation in the mean local porosity Ap/ a stress triaxiality of 1/3, deformation and stress in flat
Av calculated using the Voronoi tessellation procedure as notched specimens are non-homogeneous with a nearly
a function of RVE size is plotted in Fig. 4a. The variation constant stress triaxiality of about 0.577 for the particular
2362 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368

Fig. 3. Subregions of the initial microstructure with different length scales.

Local Porosity Standard Deviation

0.12 0.095
Bulk porosity
Local porosity
0.09
0.1
Mean Porosity

0.085
0.08
0.08

0.06
0.075

0.04
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5
Length Scale - mm Length Scale - mm
(a) (b)
Fig. 4. Variation of (a) the mean and (b) the standard deviation of porosity with RVE size.

notch geometry considered in this work. The flat notched 700

D
tests are particularly useful in assessing the effect of stress C
600
triaxiality on the evolution of void nucleation and growth B
in NCI. 500
True Stress - MPa

400
3. Uniaxial tension test
300
Uniaxial tension tests were performed on standard
dumb-bell shaped specimens in order to characterize the 200

elastic–plastic behavior of the materials. Dumb-bell speci- 100

mens of 2.54 mm thickness and 6.35 mm width were cut
in the longitudinal direction from a blank of 0
0 0.05 0.1 0.15 0.2
38.1  38.1  160.0 mm. Tensile tests were run on three True Strain
specimens in an Instron 4482 universal testing machine at
a cross-head speed of 0.254 mm min1, resulting in quasi- Fig. 5. True stress–true strain response of the nodular cast iron.
static loading at a strain rate of 104 s1. The strain and
displacement measurements were made over a gage length until abrupt failure. Thus, this specimen does not exhibit
of 20 mm using digital image correlation (DIC) (see Sutton plastic localization in the form of either diffuse or localized
et al. [20] for details of this technique). Specimen A is an necking. A power law relation was used to fit the true
unstrained sample to characterize the initial microstruc- stress–true strain curve as follows:
ture. Specimens B and C were unloaded before final failure p n
at different levels of homogeneous straining (eB ¼ 0:056, rNCI ¼ rNCI
0 ð1 þ be Þ ð1Þ
eC ¼ 0:094) in order to quantify the microstructural evolu- where rNCI
¼ 227:5 MPa is the initial yield strength,
0
tion. The final specimen (Specimen D) broke at an overall n ¼ 0:17 is the hardening exponent, ep is the plastic strain
true strain of eD ¼ 0:145. The true stress–true strain and b ¼ 2439. The Considère point for this material is at
responses determined from the tension tests are shown in en = n  1/b = 0.17, hence the absence of diffuse necking
Fig. 5. The Young’s modulus and Poisson’s ratio were suggests that fracture or failure by the linking up of micro-
found to be E ¼ 170 GPa and m ¼ 0:3, respectively. As scopic damage must have preceded the onset of diffuse
can be seen from Fig. 5, the load increases monotonically necking.
 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368 2363

Optical micrographs of a thick section of the specimen, standard deviation in the local porosity shown in Fig. 4b,
obtained by stitching together a number of micrographs, is the observed increase in bulk porosity cannot be statisti-
shown in Fig. 6 for specimens A–D. Higher magnification cally significant. This is not surprising in the light of the
images of portions of these regions are shown in the figures fact that the RVE is quite large. A slight decrease in the
in the bottom row. The evolution of the microstructure average shape factor from F ¼ 0:685 for specimen A to
with deformation is not clearly evident from the high mag- F ¼ 0:670 for specimen D can be seen as a subtle change
nification micrographs in Fig. 6. Careful examination indi- in the shape of the voids which visibly appear to be elon-
cates that the voids appear to grow by subtle elongation in gated in the loading direction (see Fig. 6). This indicates
a direction parallel to the applied tensile loading. Of a change in void shape in addition to volumetric enlarge-
course, the spatial distribution of the nodules does not ment during deformation of the microstructure, but all of
change during deformation; it is difficult to see growth of these changes are of very small magnitude in comparison
the pores or the evolution of the porosity of specimens with the change in pore structure corresponding to final
B–D corresponding to different strain levels in comparison failure.
with the initial microstructure. Evolution of the distribu-
tion of local porosity, estimated using the Voronoi tessella- 4. Flat notched tension test
tion procedure discussed above, is plotted in Fig. 7a, while
evolution of the distribution of void area is shown in Flat notched specimens cut longitudinally into strips of
Fig. 7b. While there does not appear to be a statistically 2.54 mm thickness, 24.13 width and 152.4 mm length from
significant change in the distribution of local porosity in the center portion of the cast iron blank were used in order
the specimens strained to different levels, the bulk porosity to extend the strain range over which the evolution of
increased from 0.10 to 0.13. However, in view of the porosity was examined (Fig. 8a). Two semicircular cut-outs

Fig. 6. Evolution of the microstructure at different stages of the tensile tests: (a) eA = 0, (b) eB = 0.056, (c) eC = 0.094 and (d) eD = 0.145. All images are to
the same scale; the scale bar is shown in (a). Note that these are stitched images and the dark horizontal bands are artifacts of image composition. Higher
magnification images are shown in the bottom row.
2364 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368

0.16 flat-notched specimens. Fig. 8c and d shows the contour

A
0.14
plots of the axial and transverse strains superimposed on
B
C
the speckled surface of the specimen for specimen A. The
0.12
D axial strain is seen to be concentrated near the notch
regions, reaching values as high as 0.165. The transverse
Frequency

0.1
strain is close to zero in the middle portion of the specimen
0.08 width, indicating a nearly plane strain state, and attains a
0.06
maximum value near the notch tip region, which is
expected to be in a state of uniaxial stress. It can be seen
0.04 from Fig. 8e that the axial strain on the surface of the spec-
0.02
imen across the minimum cross-section varies from 0.037
at the mid-point to 0.165 at the notch tip. The onset of final
0
0 0.2 0.4 0.6 0.8 1
failure occurs rather abruptly at the notch tip. A complete
Local Porosity numerical analysis of the problem is required to obtain the
variation in the stress field in the specimen, however, it can
(a) be noted that due to the geometrical constraint the region
0.14
close to the root of the notch will be in a stress state close
A to uniaxial tension, whereas the middle part is in a plane
0.12 B strain state. Therefore, a gradualp increase in stress triaxial-
C ity from 1/3 at the notch tip to 1/ 3 at the minimum cross-
0.1 D section is expected.
Frequency

0.08
Specimen A was sectioned along the lines a–a, b–b, and
c–c indicated in Fig. 8a to reveal the x–z planes. These sec-
0.06 tions were polished and observed using an optical micro-
scope to identify changes in the voids. These micrographs
0.04
are shown in Fig. 9, where the axial strain corresponding
0.02
to each image is also identified. The distributions of local
porosity and void area were obtained as discussed earlier
0 and are plotted in Fig. 10. As in the uniaxial tension tests,
0 1000 2000 3000 4000
these plots indicate that there is very little change in the
Void Area μm2
void size distribution, but a slight decrease in the shape fac-
(b) tor, indicating elongation of the voids. A small increase in
the bulk porosity varying from about 0.106 at section a–a
Fig. 7. Evolution with strain of the distribution of (a) local porosity and
(b) void area.
to about 0.14 at section c–c is observed as the strain varies
from 0.037 to 0.165, but once again this is not statistically
significant in view of the required size of the RVE and the
of radius 4 mm were introduced along the sides in order to standard deviation in the initial porosity. Interestingly, sec-
increase the stress triaxiality in the central portion of min- tions a–a and b–b do not indicate significant void growth
imum cross-section. This modification allows us to pro- even though the stress triaxiality is higher here than that
mote and evaluate the void growth process in a different at the notch. It should be noted, however, that the strain
stress state. Two tests were run with specimen B strained levels are quite low in these regions.
to failure and the straining of specimen A interrupted Fig. 11a shows scanning electron micrographs of the
slightly before final failure so that evolution of the void fracture surface of specimen B. The fracture surface con-
growth could be correlated with the macroscopic strain sists of three main features. First, equiaxed dimples are
fields measured by DIC. It should be noted that DIC mea- observed surrounding each graphite nodule (see Fig. 11a
sures the strains on the surface of the specimen, however, in and b). As discussed earlier, each nodule acts as a void that
general the strain in the specimen midplane could be subsequently grows with strain. Coalescence with neigh-
higher, particularly when sheet mode necking is triggered boring voids results in ridges that outline the large equi-
[21,22]. For our particular application, due to the absence axed dimples. Second, voids that are about one order of
of significant necking prior to failure, the thickness varia- magnitude smaller than the primary dimples are also seen
tion was indeed found to be small. The nominal stress on the fracture surface (see Fig. 11c). These secondary
(force/initial minimum cross-sectional area) vs. normalized voids are typically observed at the boundaries of the larger
gage displacement (over a gage length of 20 mm) responses dimples and are found at the ridges where the primary
for both specimens are shown in Fig. 8b. Just as in the uni- voids meet or link up. Such secondary fields of small voids
axial tension tests, these curves continue to exhibit a posi- have been experimentally observed in fracture surfaces of
tive slope until the onset of final failure. No instability in tensile specimens of steel and aluminum alloys (see, for
the form of necking or shear bands was observed in the example, Cox and Low [23], Hahn and Rosenfeld [24],
 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368 2365

600
x
B
500

Nominal Stress -MPa

400 A

a b c 300

y 200

a b c 100

0
0 0.01 0.02 0.03 0.04 0.05
Normalized Gage Displacement - Δ/L

(a) (b)

(c) (d)
0.18
0.16
0.14
Axial Strain

0.12
0.1
0.08
0.06
0.04
0.02
0
0 2 4 6 8 10
y - mm
(e)
Fig. 8. (a) Schematic diagram indicating sections a–a, b–b and c–c made across the minimum cross-section. (b) Nominal stress (force/initial minimum
cross-sectional area) vs. gage displacement (D) normalized to the gage length (L = 20 mm) response curves of the flat notched specimens. (c) Contour plot
of the axial strain at the onset of final failure. (d) Contour plot of the transverse strain at the onset of final failure. (e) Variation of the axial strain along
x ¼ 0.

and Marini et al. [25]), termed void sheets, occurring dur- examined. Optical micrographs of these sections, shown
ing coalescence of neighboring voids. Lastly, cleavage fac- in Fig. 12, reveal the morphology of the nodules (voids)
ets typical of ferrous alloys are also present on the fracture below the fracture surface. The highly localized nature of
surface, as indicated in Fig. 11d. The topology of the facets void growth and final failure is evident from a qualitative
hints at a transgranular mode of fracture. It is concluded comparison of the void dimensions in these sections with
that initially the nodules create voids that grow by plastic those on the fracture surface. This can be quantified in
flow, but coalescence can occur either by the formation two different ways. First, the void area distributions
of void sheets from a field of smaller voids or through acquired for section c–c of specimen A and from the frac-
cleavage in favorably oriented grains. However, the regions tographs of specimen B are shown in Fig. 13a. The mean
in which voids grow to such a large extent needs to be dimple in the fracture plane (the area inside the ridges
examined through additional micrography. where coalescence occurred) occupies an area of
In an effort to identify the spatial extent over which the 2500 lm2. In contrast, the mean void area in plane c–c of
damage processes that were identified from fractography specimen A at a distance of about 20 lm from the fracture
develop sections a–a, b–b and c–c of specimen B were surface is only about 1052 lm2, which corresponds to a
2366 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368

Fig. 9. Optical micrographs of the sections a–a, b–b, and c–c and the corresponding surface strain levels e = 0.037, 0.049 and 0.165, respectively.

0.25 0.14
a-a a-a
b-b b-b
0.12
0.2 c-c c-c
0.1
Frequency

Frequency

0.15
0.08

0.1 0.06

0.04
0.05
0.02

0 0
0 0.2 0.4 0.6 0.8 1 0 1000 2000 3000 4000
Local Porosity Void Area - μm2
(a) (b)
Fig. 10. Evolution of the distribution of (a) the local porosity and (b) the void area for specimen A.

mean diameter of 36 lm, comparable with the initial an example of a material with pre-existing porosity, thus
nodule size. Second, a tessellation diagram drawn on sec- avoiding the complexity of exploring the nucleation pro-
tion c–c, as shown in Fig. 13b, in order to evaluate the local cess. The NCI used in this study is an ASTM A536 grade
porosity Ap/Av at different distances from the fracture 65-45-12 ductile cast iron with a ferritic matrix that con-
plane indicates that the local porosity along lines 1–5, inter- tains about 5–25% pearlite. The average grain size of ferrite
preted as the average local porosity of every cell that is and the diameter of the graphite nodules were found to be
intersected by these lines, remains statistically unchanged about 26  9 and 27  25 lm, respectively. Interrupted ten-
from the initial porosity. The main upshot of this measure- sion tests and flat notched tension tests were performed in
ment is the implication that fracture through void growth order to relate the evolution of void growth with deforma-
and coalescence occurs in an extremely narrow zone near tion and stress state.
the “fracture plane” and that the porosity is not a uni- Optical image analysis software (NIH ImageJ) was used
formly varying quantity. Therefore, an appropriate to obtain the microstructural characteristics of the material
description of nucleation, growth and coalescence needs at different stages of deformation and different stress
to be generated in a localized region whose extent is driven states. A Voronoi tessellation diagram was employed to
by the microstructure of the material and, more impor- quantify the local porosity distribution. No statistically
tantly, by the statistical fluctuations in the microstructure. significant change in the evolution of the local porosity
or void size distribution was observed on macroscopic
5. Summary straining.
The fracture surface indicates the following features.
In this paper we investigated the process of deformation Each nodule acts as a void that subsequently grows with
and failure in initially porous materials. NCI was used as strain. Coalescence with neighboring voids results in ridges
 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368 2367

(a) (b)

(d) (c)

Fig. 11. (a) Scanning electron micrographs of the fracture surface of specimen B. (b) Equiaxed dimples surrounding each graphite nodule. (c) Secondary
smaller voids at the boundary of the primary dimples. (d) Cleavage facets seen on the fracture surface.

Fig. 12. x–z Micrographs of the sections a–a, b–b and c–c of specimen B showing the evolution of the microstructure below the fracture surface.

that outline large equiaxed dimples. Voids that are about on the fracture surface. It was concluded that initially the
one order of magnitude smaller than the primary dimples nodules create voids that grow by plastic flow, but this is
are typically observed at the boundaries of the larger dim- terminated by a transition to deformation in a very local-
ples and are found at the ridges of the primary voids. ized region over which coalescence occurs either by the for-
Cleavage facets typical of ferrous alloys are also present mation of void sheets or through cleavage.
2368 A. Ghahremaninezhad, K. Ravi-Chandar / Acta Materialia 60 (2012) 2359–2368

Line 1
0.25
Plane c-c of Specimen A,
Line 2
mean void area = 1052 μm2
Fracture surface, Line 3
mean dimple area = 2498 μm2
0.2 Line 4
Line 5
Frequency

0.15

0.1

0.05

0
0 1000 2000 3000 4000 5000 6000
Dimple or Void Size - μm2

(a) (b)
Fig. 13. (a) Distribution of void area acquired at section c-c of Specimen A and from the fractographs of Specimen B. (b) A Voronoi tessellation diagram
drawn on section c–c of Specimen B.

Acknowledgement [10] Dong MJ, Berdin C, Beranger AS, Prioul C. In: First European
mechanics of materials conference on local approach to fracture;
1996.
This work was performed during the course of an inves- [11] Decamp K, Bauvineau L, Besson J, Pineau A. Int J Fract 1997;88:1.
tigation into ductile failure under two related research pro- [12] Pardoen T, Doghri I, Delannay F. Acta Mater 1998;46:541.
grams funded by the Office of Naval Research: MURI [13] Steglich D, Brocks W. Fatigue Fract Eng Mater Struct 1998;21:1175.
project N00014-01-1-A00001 and FNC project: N00014- [14] Benseddiq N, Imad A. Int J Pres Ves Pip 2008;85:219.
08-1-0189. This support is gratefully acknowledged. [15] Bernauer G, Brocks W. Fatigue Fract Eng Mater Struct 2002;25:363.
[16] Dong MJ, Prioul C, Francois D. Metall Mater Trans A
1997;28A:2245.
References [17] Steglich D, Brocks W. Comput Mater Sci 1997;9:7.
[18] Guillemer-Neel C, Feaugas X, Clavel M. Metall Mater Trans A
[1] Beran MJ. Statistical continuum theories. New York: Wiley; 1968. 2000;31A:3063.
[2] Hashin Z. Appl Mech Rev 1968;17:1. [19] Argon AS, Im J. Metall Trans A 1975;6A:839.
[3] Lemaitre J, Desmorat R. Engineering damage mechanics: ductile, [20] Sutton MA, Wolters WJ, Peters WH, Ranson WF, McNeill SR.
creep, fatigue and brittle failures. Berlin: Springer-Verlag; 2006. Image Vis Comput 1983;1:133.
[4] Gurson A. J Eng Mater Technol 1977;99:2. [21] Dunand M, Mohr D. Int J Solids Struct 2010;47:1130.
[5] Needleman A, Tvergaard V. J Mech Phys Solids 1984;32:61. [22] Ghahremaninezhad A, Ravi-Chandar K. Int J Solids Struct
[6] Rousselier G. Nucl Eng Des 1987;105:97. 2011;48:3299.
[7] Tvergaard V. Adv Appl Mech 1990;27:83. [23] Cox TB, Low JR. Metall Trans A 1974;5A:1457.
[8] Benzerga AA, Leblond JB. Adv Appl Mech 2010;44:169. [24] Hahn GT, Rosenfeld AR. Metall Trans 1975;6:653.
[9] Brocks W, Sun DZ, Honig A. Comput Mater Sci 1996;7:235. [25] Marini B, Mudry F, Pineau A. Eng Fract Mech 1975;22:989.