Computers and Geotechnics 65 (2015) 126135

Contents lists available at ScienceDirect

Computers and Geotechnics

journal homepage: www.elsevier.com/locate/compgeo

Research Paper

Size effect, material ductility and shape of fracture process zone

in quasi-brittle materials
Mehdi Galouei 1, Ali Fakhimi
Department of Mineral Engineering, New Mexico Institute of Mining and Technology, Socorro, NM 87801, United States

a r t i c l e

i n f o

Article history:
Received 18 October 2014
Received in revised form 10 December 2014
Accepted 13 December 2014
Available online 31 December 2014
Keywords:
Bonded particle model
Tensile softening
Shape of fracture process zone
Size effect
Material ductility

a b s t r a c t
Numerical three-point bending tests were conducted on a softening material model to study the effect of
specimen size and material ductility on the shape of the fracture process zone. A bonded particle model
(BPM) was used for simulation of the rock. The particles at the contact points were allowed to follow a
softening behavior to be able to capture the initiation and development of the process zone. Five different
beam sizes of 20 (height)  60 (span), 40  120, 80  240, 160  480, and 320  960 mm2 were used. All
beams had a notch at their mid-span to study the mode I fracturing of rock. For each specimen size, six
different realizations were introduced to study the effect of particle arrangement on the induced damage
zone. The material ductility was controlled and modied by introducing different slopes for the post peak
behavior of the contact points between the particles. The shape of the process zone was obtained by
calculating both the width and the length of the process zone at the peak load. The results suggest that
as the specimen size increases, the process zone expands in its size. In addition, the results indicate that
for quasi-brittle materials, the length to width ratio of the process zone is greater compared to that of
ductile materials. Furthermore, it is shown that the sizes of the representative elemental volumes
corresponding to the width and the length of the process zone may not be identical in the quasi-brittle
materials.
2014 Elsevier Ltd. All rights reserved.

1. Introduction
It is well known that during crack propagation in quasi-brittle
materials such as rock and concrete, a localized damaged zone is
created which can affect the peak load and the deformational
behavior of the material. This damaged zone or process zone is
made of micro-cracks between the grains or through the grains.
With continuation of loading, the severely deformed and loaded
micro-cracks in the process zone are eventually extended and coalesced to form the macro-cracks which subsequently can result in
failure of the rock structure.
Due to the importance of the process zone in the fracture
mechanics of quasi-materials, this topic has been investigated by
many researchers. Otsuka and Date [1] used the X-ray and 3D acoustic emission (AE) techniques to investigate the effect of specimen
size and aggregate size on the fracture process zone in concrete.
Their results showed that as long as the maximum aggregate size
is identical, the width and length of the fracture process zone

Corresponding author. Tel.: +1 575 835 6577.

1

E-mail addresses: galavi@nmt.edu (M. Galouei), hamed@nmt.edu (A. Fakhimi).

Tel.: +1 575 418 1691.

http://dx.doi.org/10.1016/j.compgeo.2014.12.010
0266-352X/ 2014 Elsevier Ltd. All rights reserved.

increase as the specimen size increases. Their results also

demonstrated that the rate of change of the length of the fracture
process zone was much larger than that of the specimen size. On
the other hand, it was shown that the rate of change of the width
of the fracture process zone was smaller than that of the specimen
size. Consequently, the shape of the process zone is affected by
the specimen size.
The effect of aggregate size on the fracture process zone of
asphalt mixture at low temperature was monitored using an acoustic system in three point semi-circular bending tests by Li and Marasteanu [2]. Based on their results, the authors claimed that the size
of the fracture process zone depends on many parameters such as
aggregate particles, preexisting cracks and presence of voids in
the material. Their experimental data revealed the importance of
the temperature of the asphalt specimen on the size of the process
zone as well; the fracture process zone is longer at lower
temperatures.
Zhang and Wu [3] performed several physical three point bending tests using concrete beams with different notch depths. The
acoustic emission method was used to study the process zone.
The results revealed that the length of the fracture process zone
is not a material parameter as it is greatly inuenced by the
fracture ligament length.

127

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

A meso-scale approach was used by Grassl et al. [4] to investigate the effect of specimen size on the fracture process zone of concrete beams. Their study concluded that the width of the fracture
process zone for notched specimens does not depend on the specimen size. On the contrary, they showed that the fracture process
zone length along the ligament is dependent on the specimen size.
Finite element numerical simulations of fracture tests were carried out by Vesely et al. [5] to investigate the inuence of structural
size and geometry on the size and shape of the fracture process
zone. Two different types of stress constraint at the crack tip, single
edge notched beam under three point bending (SEN-TPB) and double edge notched panel under tension (DEN-T), were studied. Five
different specimen sizes were considered for each testing conguration. The results indicate that the size and shape of the fracture
process zone are affected by the specimen size and the structural
geometry.
In this paper, the effect of specimen size and material ductility
on the size and shape of the process zone is studied. Three-point
bending tests are conducted on notched beams to create crack
extension in mode I fracture. To capture the process zone, a softening contact bond model is used in the numerical bonded particle
model (BPM). Material ductility is controlled by changing the
micromechanical softening behavior of the particles at the contact
points. It is shown that both the specimen size and material ductility can affect the shape and size of the process zone. To the best
knowledge of the authors, the applications of the BPM in studying
the shape of the process zone in mode I fracture and the associated
representative elemental volume are being investigated for the
rst time in this study.

contact is forced to follow a Coulomb type behavior using the contact friction coefcient l. In addition to the above micromechanical
parameters, the particles radii (R) and genesis pressure (r0) must be
specied. The genesis pressure is the applied surrounding pressure
during sample preparation that allows some initial small overlap of
the particles which helps to more realistically simulate the macroproperties of the rock [8].
The simulated material was calibrated to mimic the mechanical
behavior of the Berea sandstone. The sandstone has an elastic modulus of 14 GPa, a Poissons ratio of 0.32, a uniaxial compressive
strength (qu) of 5560 MPa, and a bending tensile strength (for a
80 mm  240 mm beam) of 8.6 MPa. The calibration procedure
was conducted using the technique introduced by Fakhimi and
Villegas [8] and resulted in the following model properties:
kn = 22.0 GPa, ks = 5.5 GPa, nb = 2800 N/m, sb = 12300 N/m, l = 0.5,
r0 = 2.2 GPa, and knp = 1.83 GPa (kn/knp = 12). To obtain the appropriate kn/knp value, the calibration was conducted to get a realistic
width for the fracture process zone of the sandstone consistent
with that from the physical testing [9]. The circular particles or
disks in the bonded particle model were assumed to have radii
with uniform random distribution within the range of
0.270.33 mm (Rave = 0.3 mm). The simulated material with the
above micro-parameters was tested in uniaxial compression
(40 mm  80 mm
specimen)
and
three
point
bending
(80 mm  240 mm specimen) to verify the accuracy of the calibration procedure. An elastic modulus of 13.3 GPa, a Poissons ratio of
0.19, a uniaxial compressive strength of 60.5 MPa, and a tensile
strength of 8.7 MPa were obtained that are in reasonable agreement with the corresponding properties of the physical samples.

2. Numerical model

3. Numerical results

The computer program CA2 [6,7] which is a hybrid nite elementdiscrete element code for static and dynamic simulation of
geomaterials is used in this study. The rock is simulated by discrete
circular particles that are glued at the contact points (bonded particle model). The simple contact bond model for interaction of the
particles at the contact points in this numerical technique, has a
few micromechanical parameters including normal and shear stiffnesses (kn and ks), normal and shear bond strengths (nb and sb), and a
friction coefcient (l). A contact breaks in shear if the applied shear
contact force exceeds the bond strength. On the other hand, if the
applied tensile force exceeds the contact normal bond strength,
gradual loss of bond strength with further deformation is assumed
in our softening model (Fig. 1). With this technique, gradual weakening of the material in mode I fracture can be simulated. To change
the material ductility, the post-peak slope of the softening curve in
Fig. 1a (knp) is modied. Note that no softening in shear is studied in
this paper as all the observed micro-cracks in the conducted three
point bending tests are tensile; softening in shear was not included
because this effect is only pronounced for loading under signicant
(>10% qu) mean stress. Following complete failure of a contact, the

Three point bending tests (Fig. 2a) on ve different specimen

sizes (20  60, 40  120, 80  240, 160  480, and 320  960)
were conducted. The rst and second numbers for each sample size
are the height and the span of the beams in millimeters. Each beam
has a notch at its mid-span which is 3.6 mm in width. The notch
length to the beam height ratio (a/D) is constant for all beams
and is equal to 0.25. The numbers of discrete particles or disks
for the smallest and largest beams were 1850 and 465,280, respectively. For each beam size, 6 different random distributions of disks
locations were used; six different realizations were conducted. To
reduce the computational time (for the largest beam, a few weeks
were needed to nish the run using a desk top computer with an i7
processor. For each beam size, six different realizations and four
different ductility values were needed), each beam was divided
along its span to three equal pieces. The outer pieces were covered
by disks with the average radius four times the average radius of
the particles in middle part (Fig. 2b). The normal and shear stiffness of the disks in the three regions of the beam are the same
to assure the same elastic modulus for all parts of the beam [8].
Furthermore, the normal and shear bond values for the contacts

(a)
nb

1
E

kn knp
1

kn
A

Fn (Compression)

Fs
sb

D
kn

(b)

Fn (Tension)
C

F
Un

Fn

Us

Fig. 1. The relationship between (a) normal contact force and normal contact displacement and (b) shear contact force and shear contact displacement.

128

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

(a)

(b)

D
a
S

Fig. 2. (a) Three-point bending test set up and (b) the 40 mm  120 mm numerical specimen. Larger disk sizes were used for both ends of the beam to reduce the
computational time.

between the larger and smaller disks and for the contacts between
larger disks in the outer regions were four times that for the contacts between smaller disks in the middle of the beam to assure
consistent rock strength for all parts of the simulated rock [8]. Note
that in all the numerical simulations, the damaged contacts were
conned completely within the middle part of the beam.
In the numerical studies of this work, the micro-mechanical
parameters obtained for the Berea sandstone are used. The only
exception is the knp parameter; kn/knp is changed to control the
material ductility. Four different kn/knp = 20, 40, 60, 80 are used
in this study.
The load-notch mouth opening displacement curves for two
beam sizes 40  120 mm2 and 320  960 mm2 are shown in Fig. 3.
As expected, as kn/knp is increased, the pre-peak and post-peak parts
of the loaddisplacement curves show more ductile behavior suggesting less violent failure of the simulated specimen. The average
peak loads (average of six peak loads for the six different realizations) for each beam size and each kn/knp are reported in Table 1.
The nominal tensile strength (rN) of the beam was obtained using
the following equation:

rN

3Pmax S

in which Pmax is the peak load, S is the beam span, D is the beam
height, and b (assumed to be 1 in our analyses) is the beam thickness. The apparent fracture toughness (KICA) for a single-edge
cracked specimen under three-point bending for arbitrary height
to span ratio, could be calculated by:

K ICA

Specimen size (mm2)

kn =knp

P max (kN)

rN (MPa)

KICA (MPa m0.5)

20  60

20
40
60
80

31.47
45.85
55.28
63.58

7.08
10.32
12.44
14.30

0.87
1.28
1.54
1.78

40  80

20
40
60
80

46.45
70.60
89.13
104.10

5.23
7.94
10.03
11.71

0.92
1.39
1.76
2.05

80  240

20
40
60
80

66.73
109.10
143.30
177.70

3.75
6.14
8.06
10.00

0.93
1.52
2.00
2.48

160  480

20
40
60
80

97.43
180.83
247.83
298.67

2.74
5.08
6.97
8.40

0.96
1.78
2.45
2.95

320  960

20
40
60
80

146.33
274.50
387.50
468.67

2.06
3.90
5.45
6.59

1.02
1.92
2.70
3.27

2bD

p
rN D

Table 1
The average values of peak load, nominal tensile strength, and apparent fracture
toughness of the simulated materials in three-point bending tests.

1 2a1  a

1:5

PS=D a

in which rN is the nominal tensile strength, D is the specimen

height, a is the ratio of notch length to specimen height (a/D), and
PS/D(a) is a shape factor. The expression for PS/D(a) has been reported
by Pastor et al. [10]:

PS=D a P1 a 4

 
D
P4 a  P1 1
S

in which the expressions for P4(a) and P1(a) are:

h
i
P4 a 1:900  a 0:089 0:6031  a  0:4411  a2 1:2231  a3 4
h
i
P1 a 1:989  a1  a 0:448 0:4581  a  1:2261  a2
5

The values of apparent fracture toughness for different beam

sizes with different ductility (kn/knp) are reported in Table 1. To
obtain the fracture toughness values using the apparent fracture

Fig. 3. Load vs notch mouth opening displacement for materials with different ductility. (a) Specimen size 40  120 mm2 and (b) specimen size 320  960 mm2.

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

toughness KICA, (1/KICA)2 is plotted vs 1/D. According to Eq. (2), a linear relationship should be expected. From this plot, the KIC value
corresponding to very large specimens (1/D equal to zero) is
obtained [11,12]. The effect of specimen size and material ductility
on the apparent fracture toughness is shown in Fig. 4. For each
specimen size and kn/knp value, six points are shown in Fig. 4 corresponding to six different realizations. The mean value corresponding to each six apparent fracture toughness values is
shown with a dark circle and the associated standard deviation is
reported. Note that as the material becomes more ductile, the fracture toughness increases; the energy for unit extension of the crack
increases. The apparent fracture toughness increases with the sample size as well, but for more ductile materials, this dependency on
sample size is more severe. This suggests that for less brittle materials, greater specimen sizes need to be tested to obtain the true
fracture toughness of the material. Note that consistent with the
simulation results, size dependency of apparent fracture toughness
has been reported in physical fracture testing of rock specimen
[13]. Inspection of the standard deviation of the fracture toughness
values indicates that the data scatter is not a function of the material ductility and the specimen size.
4. Process zone around the notch tip
We assume the process zone around the notch tip is made of
micro-cracks (damaged contact points whose bond strengths have
been reduced to zero) and those contact points that have been partially damaged. A damaged contact is the one that is deforming in
its post-peak regime corresponding to a point such as point D in
Fig. 1a. To obtain a more accurate process zone, the numerical sample is loaded by moving the midpoint at its top part with a small
and constant quasi-static vertical velocity of 1.0  109 m/cycle.
As the peak load is approached, the applied velocity is even further

129

reduced to 0.05  109 m/cycle to allow for a better control of the

applied load and to detect the exact pattern of the process zone
around the peak load.
In general, the damaged contact points in the process zone are
scattered around the notch tip. This creates difculties in dening
the width and the length of the process zone at the peak load. In
the previous studies [14], an arbitrary rectangular region to contain a certain number of damaged points was selected around
the notch tip to dene the boundaries of the process zone. In this
study, a more objective approach in dening the boundaries of
the process zone is proposed. The histogram of the damaged contacts suggests that the damaged points are approximately following a normal distribution pattern across the process zone width.
We assume a normal distribution for the damaged contacts along
the length of the process zone as well, even though it is a less accurate model compared to that for the width of the process zone. The
histograms of damaged points along the width and length of the
process zone for beam size 80  240 mm2 with kn/knp = 40 are
shown in Fig. 5a and b, respectively. The pattern of the histograms
in Fig. 5 has led us to dene the width and length of the process
zone using a statistical approach. To this end, the mean values of
) of the damaged points were
x-coordinates (
x) and y-coordinates (y
obtained and the width (W) and length (L) of the process zone were
dened by the following equations:

W 3Sx
  ynotch 1:5Sy
Ly

6a
6b

in which ynotch is the y-coordinate of the notch tip, and Sx and Sy are
the standard deviations of the x and y values of the center of damaged contacts, respectively. By assuming a normal distribution for
the damaged points around the notch and using the above equations, about 87% of the damaged points will be contained within
the width or length of the process zone.

Fig. 4. Apparent fracture toughness vs specimen size (height of specimen) for different material ductility. (a) kn/knp = 20, (b) kn/knp = 40, (c) kn/knp = 60, and (d) kn/knp = 80. For
each specimen size, six different values corresponding to six different numerical realizations are shown. The mean value of each set of data is shown with the dark circle and
the corresponding standard deviation is reported in the gure.

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

200

(a)

Number of microcracks
and damaged contacts

Number of microcracks
and damaged contacts

130

150
100
50
0

9 10 11 12 13

Bin number

140
120

(b)

100
80
60
40
20
0

1 2 3 4 5 6 7 8 9 10 11 12 13

Bin number

Fig. 5. Histograms of damaged contacts for beam size 80  240 mm with kn/knp = 40, (a) across the width of the process zone and (b) across the length of the process zone.

In Fig. 6, the damaged contacts and the calculated process zones

for six different realizations of the 40  120 mm2 specimens with
kn/knp = 40 are shown. The width and length of the process zone
in this gure are (10.0  8.7), (7.8  6.5), (5.8  4.8), (8.6  8.0),
(8.0  8.8), and (10.0  6.3) mm2. The process zone dimensions
for the largest beam (320  960 mm2) which is made of the same
material
(kn/knp = 40)
are
(16.8  24.4),
(18.7  23.7),
(18.7  44.3), (17.6  20.4), (18.8  25.5), and (19.6  26.6) mm2
(Fig. 7). Notice that with the increase in the beam size, the size
of the process zone has increased substantially. In addition, the
shape of the process zone has been affected; the ratio of the length
to width of the process zone has increased with the increase in the
specimen size. This issue will be discussed in the next section in
more details. In Fig. 8, the process zones for the simulated beams
320  960 mm2 in dimension with kn/knp = 80 are shown. The calculated process zones dimensions are (63.7  65.7), (68.8  59.5),
(56.9  54.3), (60.2  52.6), (61.8  50.4), and (59.1  51.0) mm2
that are much larger in size compared to those for the same specimen size but with kn/knp = 40. This suggests that with increase in
the material ductility, the process zone grows in its size which is
expected; as the ductility increases, the material behavior shows
more deviation from the theory of linear elastic fracture mechanics
and the apparent fracture toughness becomes more size dependent
(Fig. 4).
5. Evolution of process zone with notch opening
In this section, the development of the process zone with the
notch opening is discussed. The length and width of the process
zone for the specimen sizes (20  60), (80  240), and
(320  960) mm2 were calculated and their changes with the
increase in the notch mouth opening were investigated. The results
for kn/knp = 20 and kn/knp = 80 are shown in Fig. 9. Note that irrespective of the specimen size, for the more brittle material (kn/knp =
20), the width of the process zone is developed almost instantly;
depending on the specimen size, there is a critical notch mouth
opening at which the process zone width suddenly develops. In
this case, further loading and notch opening has a small impact
on the width of the process zone. For the less brittle material (kn/
knp = 80), there is a sudden growth in the width of the process zone
as well, but in this case with further notch opening, the growth of

the process zone width continues. On the other hand, the length of
the process zone evolves smoothly with the notch opening for both
the brittle and ductile materials. Notice that for the more brittle
material, the length of the process zone grows faster close to the
applied peak load or critical notch mouth opening. This increase
in the growth rate of process zone length is particularly more
severe as the specimen size increases; controlling the testing stability in the physical tests is more difcult for larger specimens.
The situation is different for the more ductile material in Fig. 9b.
In this latter case, the increase in the process zone length follows
a relatively constant rate even though the overall growth rate of
the process zone length still shows some increase as the specimen
size increases. Note that consistent with the previous discussions,
Fig. 9 shows that as the material ductility increases, the size of
the process zone increases. In addition, Fig. 9 suggests that the sensitivity of the process zone dimensions to the specimen size
increases with increase in the material ductility.
6. Size effect on the shape of the process zone
As was pointed out previously, the specimen size not only
affects the size of the process zone, but also it has an impact on
its shape (its aspect ratio). To investigate this issue, the ratios of
the process zone length to its width at the peak load for the simulated beams with different ductility (different kn/knp values) are
shown in Figs. 10 and 11. The results for all six different realizations for each beam size are shown together with the average values (lled in symbols in the gures). In addition, to study the
scatter of the data, the standard deviation corresponding to each
column of the data is shown in the gures. Note that the data scatter is much greater in the more brittle materials indicating that it is
more difcult to obtain the accurate process zone dimensions in
the quasi-brittle materials such as rock and concrete; tests on
apparently similar materials can results in different process zone
dimensions. This can explain some of the discrepancies regarding
whether the process zone is a material property or not in the literature. An interesting observation in Figs. 10 and 11 is the change in
the process zone aspect ratio as the specimen size is increased. For
the more brittle material, this change in aspect ratio is more sever.
As the material ductility is increased, the ratio of the process zone
length to its width shows less variation with the specimen size and

Fig. 6. The process zones at the peak load for six different realizations of the 40  120 mm specimens with kn/knp = 40. The damaged points are shown in green while the
tensile micro-cracks are shown in red. From left to right, the process zone width and length are: (10.0  8.7), (7.8  6.5), (5.8  4.8), (8.6  8.0), (8.0  8.8), and (10.0  6.3).
The dimensions are given in mm. The parallel lines at the bottom of the FPZ region show the beam notch. (For interpretation of the references to color in this gure legend, the
reader is referred to the web version of this article.)

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

131

Fig. 7. The process zones at the peak load for six different realizations of the 320  960 mm specimens with kn/knp = 40. The damaged points are shown in green while the
tensile micro-cracks are shown in red. From left to right, the process zone width and length are: (16.8  24.4), (18.7  23.7), (18.7  44.3), (17.6  20.4), (18.8  25.5), and
(19.6  26.6). The dimensions are given in mm. The parallel lines at the bottom of the FPZ region show the beam notch. (For interpretation of the references to color in this
gure legend, the reader is referred to the web version of this article.)

Fig. 8. The process zones at the peak load for six different realizations of the 320  960 mm specimens with kn/knp = 80. The damaged points are shown in green while the
tensile micro-cracks are shown in red. From left to write, the process zone width and length are: (63.7  65.7), (68.8  59.5), (56.9  54.3), (60.2  52.6), (61.8  50.4), and
(59.1  51.0). The dimensions are given in mm. The parallel lines at the bottom of the FPZ region show the beam notch. (For interpretation of the references to color in this
gure legend, the reader is referred to the web version of this article.)

Fig. 9. Development of the process zone width (W) and length (L) as the notch opening increases for (a) kn/knp = 20, (b) kn/knp = 80. The results for three different beam sizes of
20  60, 80  240, 320  960 mm2 are shown. The solid and dotted lines show the evolution of the width and the length of the process zone, respectively. The end point of
each curve corresponds to the peak load situation for that particular beam size.

Fig. 10. The ratio of length to width of the process zone at the peak load vs the specimen height for the simulated beams with (a) kn/knp = 20, (b) kn/knp = 40. The average value
for each six realization values is shown with the lled in symbol. The standard deviation for each column of data is shown as well.

132

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

Fig. 11. The ratio of length to width of the process zone at the peak load vs the specimen height for the simulated beams with (a) kn/knp = 60, (b) kn/knp = 80. The average value
for each six realization values is shown with the lled in symbol. The standard deviation for each column of data is shown as well.

it reaches faster to a specimen size after which this ratio is not

affected by further increase in the specimen size. The important
observation is that for quasi-brittle materials, proportional
increase in specimen size does not guarantee proportional increase
in the size of the process zone; the shape of the process zone
changes as the specimen size is increased, i.e. the length and width
of the process zone do not change with the same rate. This nding
is consistent with the experimental tests conducted on concrete
beams in which it was shown that the rates of change of the process zone length and width are not identical as the specimen size
is modied [1].
Inspections of Figs. 10 and 11 indicate that the process zone for
more ductile materials and for large specimens has a square shape
while for more brittle materials, the length of process zone is
greater than its width; the process zone can be approximated with
a rectangle. Similar observation [2] has been reported in physical
testing of asphalt mixture in which the specimens with lower
temperature (more brittle specimens) have resulted in longer
process zones (greater process zone length).
7. Size of process zone as an intrinsic material property
In a previous study, Fakhimi and Tarokh [12] proposed the
following equation for the change of width (W) of the process zone
with the specimen size:

W 1b
1b

In Eq. (7), b = D/D0 is the brittleness number [11] and W1 is the

width of the process zone for very large specimens. A similar equation can be applied for the length of the process zone:

L1 b0
1 b0

In Eq. (8), b0 D=D00 is the brittleness number and L1 is the

length of the process zone for very large specimens. To verify the
appropriateness of Eqs. (7) and (8) in showing the evolution of
the process zone size with the specimen size, Eq. (7) is written in
the following form:

1
1
D0 1

W W1 W1 D

Note that in Eq. (9), 1/W is a linear function of 1/D. Similarly, for
the length of the process zone, we have:

1
1
D0 1

0
L L1 L1 D

10

Using the average lengths and widths of the process zones for
the six different realizations and for different beam sizes with different ductility (Table 2), linear regression analysis (using the data
in Table 2 and Eqs. (9) and (10)) was performed [12] and the best
values for parameters W1, L1, D0, and D00 were obtained. The values of these parameters are reported in Tables 3 and 4. In Tables
3 and 4, the approximate equations that dictate the change in process zone width and length with the specimen size are shown as
well. To realize the usefulness of Eqs. (7) and (8), the data points
together with the prediction of these equations for the simulated
materials with kn/knp = 20 and 80 are shown in Figs. 12 and 13.
Inspections of Figs. 12 and 13 show that the more ductile material
(kn/knp = 80) has less data scatter compared to that with the brittle
behavior (kn/knp = 20). Similar high data scatter have been reported
in experimental measurement of the process zone dimensions of
concrete [1]. It is evident from Fig. 12 that the data scatter is
greater for the length of the process zone compared to the width
values; sensitivity of the process zone length to the particle
arrangement and its sudden change close to the peak load can be
the cause of greater data scatter in the length of the process zone
(Fig. 9).
An interesting observation from Tables 3 and 4 is that while the
values of D0 and D00 are almost equal for more ductile materials,
they are different as the material becomes more brittle. Consequently, the L/W ratio is approximately independent of the specimen size for more ductile materials (kn/knp = 60, 80); the width
and length of the process zone grow with the same rate as the
specimen size is changed. On the other hand, the L/W ratio is a
function of the specimen size for more brittle materials (kn/knp =
20, 40). To realize the signicance of this nding, we dene DW
and DL as the specimen sizes at which W/W1 = 0.95, and L/L1 =
0.95, respectively. We call DW and DL the characteristic beam sizes
for width and length of the process zone. Alternatively, DW and DL
can be considered as representative elemental heights corresponding to the width and length of the process zone, respectively. Note
that for specimen sizes whose heights are greater than DW, the
change in the width of the process zone is very small; for D > DW,
the width of the process zone can approximately be assumed
to be an intrinsic material property. Similarly, for D > DL, the
length of the process zone is assumed to remain approximately
constant. Using Eqs. (7) and (8), it is easy to show that DW = 19
D0 and DL = 19 D0. Therefore, the ratio of the characteristic beam
sizes is equal to:

DL
D0
0
DW D0

11

133

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

Table 2
The average values of the length and the width of the process zones for different specimen sizes and kn/knp values.
Specimen size (mm)

kn =knp 20

20  60
40  120
80  240
160  480
320  960

kn =knp 40

kn =knp 80

Width (mm)

Length (mm)

Width (mm)

Length (mm)

Width (mm)

Length (mm)

Width (mm)

Length (mm)

4.6
5.2
5.2
5.4
6.2

3.3
4.3
4.2
5.8
9.6

7.0
8.4
10.0
15.2
18.4

4.9
7.2
9.9
17.4
27.5

8.6
12.0
17.4
28.9
40.0

5.7
9.6
17.0
28.4
38.6

10.1
14.8
24.4
40.4
61.7

6.4
11.6
21.6
34.9
55.6

Table 3
The values of W1, D0, and the approximate equations for the width of the process
zone for different kn/knp values.
kn =knp

kn =knp 60

W1 (mm)

D0 (mm)

Width

20

6.4

19.5

6:4 D
w D19:5

40

26.8

132.7

26:8 D
w D132:7

60

73.5

256.3

73:5 D
w D256:3

80

123.5

325.7

123:5 D
w D325:7

Table 4
The values of L1, D00 , and the approximate equations for the length of the process zone
for different kn/knp values.
kn =knp

L1 (mm)

D00 (mm)

20

11.3

143.9

11:3 D
L D143:9

40

69.4

481.3

69:4 D
L D481:3

60

70.9

251.3

70:9 D
L D251:3

80

108.7

324.7

108:7 D
L D324:7

Length

The variation of DL/DW vs kn/knp is shown in Fig. 14a. Alternatively, the same trend in variation of DL/DW can be observed if this
ratio is plotted as a function of the characteristic size or (KIC/rt)2
(KIC is the fracture toughness and rt is the tensile strength of the
simulated material). Notice that for kn/knp greater than 60, the representative elemental heights for the width and length of the process zone are approximately identical; the same representative
elemental volume exists for both the width and the length of the
fracture process zone. On the other hand, for more brittle materials, DL/DW can be substantially greater than 1; the specimen size
at which the length of process zone can be considered as a material
property is greater than that for having a size independent process
zone width. This observation can explain the discrepancies in the
literature regarding whether the size of the process zone is an
intrinsic material property or not. For small specimen sizes and a
high rock brittleness number, neither the width nor the length
may appear as intrinsic properties of the material as they change
with the change in the specimen size. For intermediate specimen
sizes, the width of the process zone may appear as an intrinsic
property, but the length may not show this feature. Only when
large specimens are used, both length and width of the process

Fig. 12. The width (a) and the length (b) of the process zone vs the specimen height for the simulated material with kn/knp = 20. The data points for the six different
realizations have been shown. The lled in symbols and the numbers are showing the average and standard deviation values for each column of the data.

Fig. 13. The width (a) and the length (b) of the process zone vs the specimen height for the simulated material with kn/knp = 80. The data points for the six different
realizations have been shown. The lled in symbols and the numbers are showing the average and standard deviation values for each column of the data.

134

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

Fig. 14. The ratio of DL to DW as a function of (a) material ductility (kn/knp) and (b) characteristic size.

zone can appear as size independent intrinsic properties of the

given material. Another consequence of having different values
for DL and DW for a material is the impact of specimen size on
the shape of process zone. As the material becomes more brittle,
it is expected that the process zone to gradually change from
approximately a square shape to a narrow rectangle; the length
to width ratio of the process zone grows as the specimen size
increases and this increase is greater as the material ductility is
reduced.
We now consider an extreme situation in which the material
ductility is very high (kn/knp = 1), i.e. the material behaves as an
elastic-perfectly plastic material. In this case, the role of fracture
mechanics diminishes or completely disappears and theory of plasticity (theory of strength of materials) governs the beam deformation. It is well known that in this case, the plastic region around the
notch tip must scale with the specimen size. Consequently, the
width and length of the plastic region will change linearly with
the specimen size; the curves in Fig. 13 must be replaced by
straight lines. In this situation, the brittleness numbers b and b0
are so small (D0 and D00 are very large) that they can be ignored
compared to 1 in the denominators of Eqs. (7) and (8). As a result,
in Eqs. (7) and (8), width and length of the plastic zone will be linear functions of the specimen size. In this extreme situation, a representative elemental volume corresponding to the size of the
plastic zone around the notch tip cannot be dened.
8. Comparison with physical tests results
There are not many studies in the literature that are concerned
with experimental measurement of dimensions of fracture process
zone with emphasis on the role of specimen size. A particular study
that shows physical evidences for the change in the size of FPZ
with the specimen size is the work conducted by Zietlow and
Labuz [14]. In this study, three point bending tests on rock specimens were conducted and AE technique was utilized to detect
the locations of the induced micro-cracks. Both the length and
the width of the FPZ were determined. Four different rocks namely
Sioux quartzite, Berea sandstone, Charcoal granite and Rockville

granite were tested. The FPZ dimensions were reported for two different specimen sizes. While for the Berea sandstone no change in
the size of the process zone was observed, for the other three rocks
either the length or width or both the length and width were changed as the specimen size was modied. The authors concluded that
the change in the width of the process zone with specimen size
was not signicant and therefore, the width of FPZ was considered
as an intrinsic material property. The work of Otsuka and Date [1]
reveals a better evidence for the size dependency of the FPZ. In this
work, tests were conducted on notched concrete specimens. X-ray
and AE techniques were employed to characterize the evolution of
the process zone. Several specimen sizes were examined and the
change in the width and length of the process zone with the specimen size was studied. In Fig. 15, the results of these physical tests
have been compared with the numerical data (for kn/knp = 40) from
this study. Note that no attempt was made to calibrate the numerical model for the concrete used by Otsuka and Date [1]. Nevertheless, it appears that the numerical model with a kn/knp of about 40
is able to duplicate with a reasonable accuracy the physical test
data. Notice that both the physical and numerical data show similar trends; the width and length of the FPZ increase with the
increase in specimen size.
The effect of material ductility on the shape of FPZ of quasibrittle materials has not been investigated extensively in the literature. A notable work was conducted by Li and Marasteanu [2]. In
their work, asphalt mixture at low temperature was studied. Bending tests on notch specimens were conducted and the FPZ was
detected using the AE technique. Specimens with different temperature were used and by plotting the load-line load displacement,
the authors demonstrated that the sample with the lowest temperature was more brittle. Their results showed that the ratio of the
length to width of the process zone for the more brittle specimen
was greater than that for the more ductile beam (with higher temperature). Note that our numerical results are consistent with this
physical observation. Nonetheless, further physical and numerical
tests are needed to more clearly understand the effect of material
ductility on the shape of FPZ. Furthermore, the size of the representative elemental volumes corresponding to the situations in which

Fig. 15. Comparison of physical tests results from Otsuka and Date [1] with the numerical data. (a) Width of FPZ, and (b) length of FPZ vs the specimen size. Numerical tests
results are for kn/knp = 40.

M. Galouei, A. Fakhimi / Computers and Geotechnics 65 (2015) 126135

the width or length of the process zone can be considered as intrinsic material properties need to be investigated in future physical
studies.
9. Conclusion
A bonded particle model with tension softening was used in
three point bending tests to study the effect of specimen size and
material ductility on the size and shape of the fracture process
zone. The results suggest that with the increase in the beam size,
the size of the process zone increases. The specimen size effect
on the size of the process zone is more pronounced as the material
ductility is increased. The specimen size and material ductility
affect the shape of the process zone as well. As the material
becomes more brittle and larger in size, the ratio of the length to
width of process zone increases; the shape of the process zone
can be approximated by a narrow rectangle around the beam
notch. An important observation from this study is concerned with
the representative elemental volume for the process zone. While it
appears that the representative elemental volumes corresponding
to the width and length of the process zone are identical for more
ductile materials, this does not seem to be the case for quasi-brittle
materials. The numerical results suggest that the representative
elemental volume corresponding to the length of the process zone
is larger than that for the width of the process zone; larger specimens must be used to obtain a size independent process zone
length for quasi-brittle materials. In the extreme situation that
the material is highly ductile, the size of the plastic zone around
the notch tip scales with the specimen size and an intrinsic fracture
process zone does not exist.

135

References
[1] Otsuka K, Date H. Fracture process zone in concrete tension specimen. Eng
Fract Mech 2000;65:11131.
[2] Li X, Marasteanu M. The fracture process zone in asphalt mixture at low
temperature. Eng Fract Mech 2010;77:117590.
[3] Zhang D, Wu K. Fracture process zone of notched three-point bending concrete
beams. Cem Concr Res 1999;29:188792.
[4] Grassl P, Gregorie D, Rojas-Solano L, Pijaudier-Cabot G. Meso-scale modelling
of the size effect on the fracture process zone of concrete. Int J Solid Struct
2012;49:181827.
[5] Vesely V, Routil L, Kersner Z. Structural geometry, fracture process zone and
fracture energy. In: Proceedings of the 6th international conference on fracture
mechanics of concrete and concrete structures, new trends in fracture of
concrete. Tylor and Francis/Balkema; 2007. p. 1118.
[6] Fakhimi A. Application of slightly overlapped circular particles assembly in
numerical simulation of rocks with high friction angle. Eng Geol
2004;74:12938.
[7] Fakhimi A. A hybrid discrete-nite element model for numerical simulation of
geomaterials. Comput Geotech 2009;36:38695.
[8] Fakhimi A, Villegas T. Application of dimensional analysis in calibration of a
discrete element model for rock deformation and fracture. Rock Mech Rock
Eng 2007;40(2):193211.
[9] Tarokh A, Fakhimi A, Labuz JF. Size of process zone in fracture testing of rock,
ARMA 12-258. In: 46th US rock mechanics/geomechanics symposium,
Chicago, IL, June 2427, 2012.
[10] Pastor JY, Guinea G, Planas J, Elices M. Nueva expression del factor de
intensidad de tensiones para la probeta de exion en tres puntos. Anales de
Mecanica de la Fractura 1995;12:8590 [a new expression for the stress
intensity factor of a three-point bending specimen in Spanish].
[11] Bazant ZP, Planas J. Fracture and size effect in concrete and other quasi-brittle
materials. CRC Press; 1998.
[12] Fakhimi A, Tarokh A. Process zone and size effect in fracture testing of rock. Int
J Rock Mech Min 2013;60:95102.
[13] Yi X, Sun Z, Ouchterlony F, Stephansson O. Fracture toughness of Kallax gabbro
and specimen size effect. Int J Rock Mech Min 1991;28(2/3):21923.
[14] Zietlow WK, Labuz JF. Measurement of the intrinsic process zone in rock using
acoustic emission. Int J Rock Mech Min Sci 1998;35(3):2919.