You are on page 1of 12

International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

Contents lists available at SciVerse ScienceDirect

International Journal of
Rock Mechanics & Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

The damage mechanism of rock fatigue and its relationship to the fracture
toughness of rocks
N. Erarslan n, D.J. Williams
Golder Geomechanics Centre, School of Civil Engineering, The University of Queensland, Australia

a r t i c l e i n f o abstract

Article history: This study presents the results of laboratory diametrical compression tests performed on Brisbane tuff
Received 26 August 2011 disc specimens to investigate their mode-I fracture toughness response to static and cyclic loading, as a
Received in revised form function of the applied load. Both the static and cyclic loading tests were carried out on Cracked
9 April 2012
Chevron Notched Brazilian Disc (CCNBD) rock specimens. Two different types of cyclic loading were
Accepted 24 July 2012
Available online 13 August 2012
applied: (a) cyclic loading with constant mean level and constant amplitude, termed sinusoidal cyclic
loading and (b) cyclic loading with increasing mean level and constant amplitude, termed increasing
Keywords: cyclic loading. The fracture toughness response to cyclic loading was found to be different from that
Rock fracture toughness under static loading in terms of the ultimate load and the damage mechanisms in front of the chevron
Rock fatigue
crack. A maximum reduction of the static fracture toughness (KIC) of 46% was obtained for the highest
Increasing cyclic loading
amplitude increasing cyclic loading test. Conversely, for sinusoidal cyclic loading, a maximum
CCNBD
SEM reduction of the static KIC of 29% was obtained. Detailed scanning electron microscope (SEM)
examinations revealed that both loading methods cause fatigue in the CCNBD specimens. When
compared with static rupture, the main difference with the cyclically loaded specimens was that
intergranular cracks were formed due to particle breakage under cyclic loading, SEM images showed
that fatigue damage in Brisbane tuff is strongly influenced by the failure of the matrix because of both
intergranular fracturing and transgranular fracturing.
& 2012 Elsevier Ltd. All rights reserved.

1. Introduction the significance of inherent flaws or microcracks in reducing a


material’s strength and in measuring its imperfection. Griffith [1]
Techniques for projects including the design and construction argued that brittle solids fail by incremental propagation of a
of underground openings, supports and rock pillars as well as multitude of randomly oriented, small pre-existing cracks. Griffith
drilling, blasting and haul roads are based on an understanding of cracks are common in rocks that contain both intragranular
the mechanical behaviour of rocks under various loading con- (outside; in grain boundary) and intergranular (inside; in grain)
ditions. Rock materials are discontinuous at all scales. At the microcracks and larger macroscopic or transgranular multiple
microscale, defects causing stress concentrations include micro- cracks.
cracks, grain boundaries, pores and bedding planes, while at the Linear Elastic Fracture Mechanics states that a crack will
macroscale geologic fractures are referred to as joints (opening) propagate when its stress intensity factor reaches a critical value
or faults (shearing), based on their genesis. Numerous experi- (KIC). The stress intensity factor depends on fracture displacement
mental and theoretical efforts have been devoted to the under- modes and crack geometry. A crack can deform in three basic
standing of crack initiation, propagation and coalescence in brittle modes: tensile (mode I), shearing (mode II) and tearing (mode III).
materials [1–3]. Under tension, these three processes take place This classification of fracturing is based on crack surface displace-
almost simultaneously in brittle rocks. However, the failure ment or crack tip loading [2]. Mixed-mode I–II fracture problems
process is more complex under compression. Under both kinds in compression are shown to be more complicated and quite
of loading, rupture (failure) of the material results primarily from different from those under tension. In this mode, tensile cracks
stable and unstable fracture propagation and crack coalescence, initially grow at an angle with respect to the direction of axial
rather than directly from fracture initiation. Griffith [1] realised compressive stress then rapidly grow along the axial compressive
stress [3–6].
Mechanical behaviour of rock under static loading has been
n
Correspondence to: The University of Queensland, School of Civil Engineering,
thoroughly investigated. However, rock reaction to cyclic, repeti-
St Lucia, Brisbane, 4072, Australia. Tel.: þ 61733653912. tive stresses resulting from dynamic loads has been generally
E-mail address: nazife.tiryaki@uqconnect.edu.au (N. Erarslan). neglected, with the exception of a few rather limited studies

1365-1609/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijrmms.2012.07.015
16 N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

[7–11]. Cyclic loading often causes brittle materials such as


ceramics and rocks to fail at a stress level lower than their
determined strength under monotonic conditions; a phenomenon
commonly termed ‘fatigue’. A principal objective of rock-fatigue
research, as found in the literature, has been to establish a
relationship between the number of cycles (N) and the reduc-
tion of applied stress amplitude (S); the S–N curve approach
[7–14]. However, most rock-fatigue research has focused on uni-
axial compressive strength degradation under cyclic loads
[7,10,14,23]. Therefore, information regarding the dynamic tensile
properties of rocks is of considerable importance in assessing the
stability of rock structures under dynamic loads. This is also
important for determining rock breakage and fragmentation
under explosive and percussive excavation [15–18]. There is very
limited research on the response of the tensile strength of rocks to
cyclic loading (as opposed to dynamic loading, such as explosive
loads and impact loading). Moreover, relatively little attention has
been given to investigating the damage mechanisms of rock
Fig. 1. The CCNBD specimen geometry with recommended test fixture.
fatigue. A novel outcome of this research is the observation of
the effect of indirect tensile cyclic loading on the fracture
toughness of rocks. One of the most fundamental parameters in required for testing. It was also unnecessary to perform pre-
fracture mechanics is critical fracture toughness (KIC), which cracking for a CCNBD specimen because it used a chevron notch
describes the resistance of a material to crack propagation. KIC is that self-pre-cracks during testing and leads to stable crack
an important material property, which corresponds to the critical propagation. Another advantage of the CCNBD method over other
state of the stress intensity factor required for crack initiation and ISRM methods involves increased precision; there is a higher load
subsequent propagation. Therefore, assessment of resistance to capacity and consistent results for each test. In our research, all
crack propagation is crucial for understanding the behaviour of the CCNBD samples showed high precision in maximum load
structures involving brittle materials. measurements. In addition, it was possible to measure mode I,
mode II and mixed-mode I–II fracture toughness by inclining the
notch at different angles with respect to the axis of diametral load.
2. Experimental procedure and tests The geometry of a CCNBD specimen is illustrated in Fig. 1. The
chevron notch causes crack propagation to start at the tip of the
2.1. Sample preparation V alignment and to proceed radially outwards in a stable fashion
until the point at which the fracture toughness is calculated.
Most of the tests in the current research were carried out on All dimensions of the geometry should be converted into
Brisbane tuff, because it is a host rock of Brisbane’s first motorway dimensionless parameters with respect to the specimen radius
tunnel, CLEM7, from which core samples were obtained. Brisbane and diameter. The suggested standard specimen dimensions are
tuff was chosen for several reasons. Firstly, being an ash deposit, it given in the ISRM suggested methods [19]. Other selections of
is a massive rock type with no bedding and it is easy to handle specimen geometrical dimensions are possible, but in order to have
and prepare for testing, minimising time constraints for the a valid test, the two most important selected dimensions: dimen-
research project. Moreover, Brisbane tuff’s massive character sionless final notched crack length (a1) and the dimensionless
gives less test result variability. Finally, it is a targeted rock type quantity (aB) must fall within the range outlined in the suggested
in the Brisbane area for its strength and stability parameters in ISRM methods [19]. The dimensionless initial crack length (a0 ¼ a0/R),
regard to tunnelling and excavation. Brisbane welded tuff is a dimensionless final notched crack length (a1 ¼a1/R) and dimen-
fine-grained, massive rock of rhyolitic composition with coarser sionless quantity (aB ¼B/R) are the three basic dimensions for the
grains imparting a porphyritic texture. Quartz and feldspar CCNBD parameters. All specimen geometries used in this research
phenocrysts 1–3 mm in size are embedded without interlocking were in the valid ranges indicated by ISRM [19]. The thickness of
in a matrix consisting of polycrystalline silica. Uniaxial compres- the notches, t, was 1.5 mm, the thickness of the samples, B, was
sive strength (UCS) and indirect Brazilian tensile strength (BTS) 26 mm and radius of samples, R, was 26 mm. The initial chevron
tests were conducted on the Brisbane tuff specimens to determine notch crack length, 2a0, was 16–18 mm and the final chevron
the mechanical characterisation of Brisbane tuff (see Table 1). notch crack length, 2a1, was 36–37 mm.
The Cracked Chevron Notched Brazilian Disc (CCNBD) speci- A circular 40 mm diamond saw was used to cut the required
mens were used in both the static and cyclic tests. The CCNBD notch. A specially designed jig recommended by the ISRM was
method had advantages over other International Society of Rock used to ensure that the chevron notches were exactly in the
Mechanics (ISRM) proposed fracture toughness tests in terms of centre of the disc. The crack displacement was measured as crack
the simplicity of sample preparation and the reduced material mouth opening displacement (CMOD) across the crack mouth.
A clip gauge for measuring the notch opening was attached to the
Table 1 knives. An Instron 2670 series crack opening displacement gauge
Mechanical characterisation results for Brisbane tuff. was used to measure CMOD. The gauge length was 10 mm and
maximum travel was 2 mm. The gauge met the requirements set
Rock sample Young’s modulus Poisson’s UCS BTS
(GPa) ratio (MPa) (MPa) out in American standard ASTM 399 70T.

Brisbane tuff NST-62 25 0.26 190 15 2.2. Static and cyclic tests
(average of five repeats)
Brisbane tuff NST-35 19 0.22 97 8.0
(average of five repeats)
Disc specimens were diametrically loaded parallel to the
diametral compressive loading directions with a crack inclination
N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26 17

Fig. 2. (a) Sinusoidal cyclic loading and (b) increasing cyclic loading.

angle of zero (b ¼01). A load-controlled testing manner was


adopted and loading was continued until failure. For compressive
loading tests, an Instron 6027 100 kN load cell hydraulic servo-
control testing system with Bluehill software was used and a
loading frame, configured separately according to the test type
(i.e., static loading or cyclic loading), was used. Load, diametral
displacement and crack mouth displacement were continually
recorded during the test using a computerised data logger. Static
tests were performed by following ISRM instructions [19].
Two different types of cyclic loading were used in this
research: cyclic loading with constant mean level and constant
amplitude, termed sinusoidal cyclic loading (see Fig. 2a) and
cyclic loading with increasing mean level and constant amplitude,
termed increasing cyclic loading (see Fig. 2b). Loading amplitudes
are constant in both types of cyclic loading. However, the mean
level of each cycle increases at a constant rate in increasing cyclic Fig. 3. Various amplitude increasing cyclic loading with same mean values.
loading tests, whereas the mean level of each cycle is constant in
sinusoidal cyclic loading tests. Table 2
Sinusoidal cyclic loading test series were conducted to obtain Mode I fracture toughness values of Brisbane tuff obtained by CCNBD tests.
an S–N curve, illustrative of the continuous weakening of rock
Rock type Pmax (kN) KIC (CCNBD)
with the increase in N required for failing a specimen loaded to a
certain upper peak stress (S). In the literature, the S–N curve CCNBD (MPaOm)
concept has been used for fatigue research under uniaxial
compressive loading with cylindrical rock samples [10,14,21]. Brisbane tuff-1 5.2 1.3
However, the current research is a first in obtaining an S–N curve Brisbane tuff-2 5.1 1.3
Brisbane tuff-3 4.6 1.2
for fracture toughness degradation because of fatigue under
Brisbane tuff-4 5.1 1.3
cyclic loads. Brisbane tuff-5 4.1 1.1
Ramp type cyclic compressive loading with increasing mean Brisbane tuff-6 4.0 0.9
level with 1 Hz frequency was used in the increasing cyclic Brisbane tuff-7 3.5 0.8
Average 4.5 1.1
loading tests. Four different amplitudes with the same mean load
were chosen for each increasing cyclic loading test series to
investigate the effect of fatigue on the fracture toughness of
rocks: 0.45 kN at 10% static ultimate load (SUL), 0.9 kN at 20% Two of the load–CMOD plots of CCNBD specimens are shown
SUL, 1.35 kN at 30% SUL and 1.8 kN at 40% SUL (see Fig. 3). in Fig. 4. The transition point from stable crack propagation to
unstable crack propagation can be determined by using load–
CMOD plots. Further, it is possible to show that there is a fracture
process zone (FPZ) in front of the tip of the chevron notch crack by
3. Results of tests obtaining the plastic deformation behaviour just before the
reaching of failure load. However, it was not possible to get
3.1. Results of static tests post-peak tensile softening behaviour after failure with all
specimens.
Five CCNBD samples were tested under load-control test
conditions. The specimens were placed under the platens with a 3.2. Results of cyclic tests
crack inclination angle of zero (b ¼01) to provide mode I loading
conditions. The loading rate was chosen as 9 kN/s to cause failure The main purpose of performing two types of cyclic loading
within 20 s as suggested by the ISRM [19]. The maximum was to find the most damaging cyclic loading type using the same
recorded load and the calculated KIC values obtained from both amplitudes. It is known that the act of applying a load provides
ISRM standard tests are shown in Table 2. energy for the crack initiation and propagation process in a
18 N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

Fig. 4. Load–CMOD plots of CCNBD specimens during monotonic testing.

Fig. 5. S–N curve for CCNBD specimens.


Table 3
Results of sinusoidal cyclic loading.
static fracture toughness values. The main purpose of this
Smax Fmax Number of cycles (N) up to comparison is to show the clear reduction of ultimate failure load
(as % of static failure load) (failure load, kN) failure that resulted in the reduction of the mode I stress intensity factor
because of rock fatigue. It is shown that crack propagation causing
93 4.2 835
failure is possible with lower stress intensity values (KI) at the
93 4.2 587
88 4 13 crack tip than the critical stress intensity value (KIC). This result
88 4 205 goes against the classical theory, which predicts that there will be
84 3.8 9 no crack growth as long as KI oKIC. According to the obtained
84 3.8 149 results, the maximum reduction, 46% of static KIC, was obtained
84 3.8 285
with the highest amplitude, 1.99 kN at 40% SUL, of cyclic loading.
80 3.6 13000
80 3.6 27000 The mode I fracture toughness value of Brisbane tuff (NST50) was
75 3.4 1761 reduced from 1.12 MPaOm to 0.61 MPaOm with the highest
75 3.4 4831 amplitude under increasing cyclic loading. This reduction has
75 3.4 1153
important implications for the investigation of the effect of cyclic
75 3.4 24175
71 3.2 6250 loading on the fracture resistance of cracks in rocks.
70 3.1 50000a To date, experimental observations of cyclic loading tests have
70 3.1 50550a indicated that dynamic cyclic loading seems to have a greater
a
effect on fracture toughness degradation of rocks than sinusoidal
No failure after that point.
compressive cyclic loading. The reduction of fracture toughness
was found to be 29% under sinusoidal loading tests as shown,
material. By removing the load, the energy supply driving crack whereas increasing cyclic loading caused a reduction of fracture
propagation is discontinued and the remaining excess energy in toughness at a maximum of 46%.
the system dissipates through the crack propagation processes. For a clear understanding of the effect of rock fatigue on
Hence, the points at which the loading and unloading of a cycle damage mechanisms, a comparison between static and dynamic
stop become important. The first important conclusion drawn cyclic loading tests is shown in Fig. 6 by plotting both results on
from the sinusoidal cyclic tests was that the failure load used in the same axes. The amplitude of the dynamic cyclic loading used
calculations of indirect tensile strength is definitely weakened in the test was 0.45 kN at 10% of SUL. As seen in Fig. 8, both failure
30% by repetitive loading. Results are given in Table 3. load values and damage mechanisms are quite different under
The S–N curve clearly shows that as the maximum applied static and dynamic cyclic loading. The failure load obtained from
diametral load decreases, the life expectancy of a CCNBD speci- the average of two static tests was reduced from an average of
men increases (see Fig. 5). Because the load range used in tests to 4.2 kN to an average of 2.1 kN because of rock fatigue. In both test
determine fracture toughness is very small, the data around high types, stable and unstable crack propagation stages were clear.
amplitudes is clustered. However, the S–N curve shows that the However, the resistance of crack propagation to cyclic loading
ultimate load causing failure is reduced by 30% (from 4.5 kN to with accumulation of plastic deformation (1 mm) was much
3.2 kN) because of rock fatigue. Therefore, it can be concluded greater than the relative value (0.025 mm) under static loading
that crack propagation resistance (fracture toughness) can be before failure (see Fig. 6). This behaviour shows that the devel-
reduced by 29–30% through repetitive loading. This means that opment of a large number of microcracks causing accumulation of
a crack can propagate under lower loads than the expected irreversible deformation is observed even prior to the appearance
ultimate loads under repetitive loading. of main cracks in a loaded specimen. This phenomenon is similar
The series of diametral compressive increasing cyclic loading to the subcritical failure of geomaterials commonly known as
tests was performed on 12 CCNBD Brisbane tuff samples. Fracture subcritical crack propagation. The mechanisms responsible are
toughness values of Brisbane tuff under static loading were discussed in detail below.
calculated as 1.12–1.5 MPaOm using the ISRM [19] suggested A load–CMOD plot further reveals that there is a tensile
methods as discussed above. The fracture toughness reduction softening behaviour with dynamic cyclic loading. Due to the
because of cyclic loading is given in Table 4 by comparing with different post-peak behaviour, the behaviour of the damage zone
N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26 19

Table 4
The fracture toughness reduction due to cyclic loading: a comparison with static fracture toughness values.

Sample Amplitude Ultimate Number of cycles Mode-I stress Fracture toughness Reduction
pffiffiffiffiffi
% SUL load (kN) (N) up to failure intensity factor (KI)a (KIC)b (MPa m) of KIC (%)
pffiffiffiffiffi
(MPa m) [22]

CCNBD-Rp1 10 2.39 2453 0.62 0.72 35


CCNBD-Rp2 10 1.95 1829 0.50 0.60 45
CCNBD-Rp3 10 2.69 2261 0.69 0.83 26
Average 2.34 2181 0.61 0.72 35

CCNBD-Rp1 20 2.16 1553 0.55 0.67 41


CCNBD-Rp2 20 2.10 1501 0.53 0.65 42
CCNBD-Rp3 20 2.24 1630 0.57 0.70 38
Average 2.17 1561 0.55 0.67 40

CCNBD-Rp1 30 2.40 1218 0.62 0.74 34


CCNBD-Rp2 30 2.5 1702 0.64 0.77 32
CCNBD-Rp3 30 2.42 1501 0.62 0.75 33
Average 2.44 1473 0.63 0.75 33

CCNBD-Rp1 40 1.86 319 0.48 0.57 50


CCNBD-Rp2 40 2.17 446 0.55 0.67 41
CCNBD-Rp3 40 1.95 136 0.50 0.60 47
Average 1.99 300 0.51 0.61 46

a
a is assumed at 0.5.
b
Y nmin is assumed at 1.21.

front of the chevron notch in CCNBD specimens is a useful means


of investigating fatigue mechanisms.
Fig. 8 shows the load–CMOD plots with different amplitudes of
increasing cyclic loading tests. To avoid bounding and moving of
the samples, a 0.5 kN minimum compressive load was applied at
the beginning of all tests. The loading–unloading frequency was
1 Hz. As the plastic deformation is very high compared with
elastic deformation in the plots, the CMOD axis was plotted as a
log scale. A clear tensile softening can be seen in all plots through
post-peak behaviour in load–CMOD plots after an accumulated
plastic deformation in the samples. This tension-softening beha-
viour is responsible for the development of the FPZ in front of and
around the crack tip. The cohesive crack model is able to describe
materials that exhibit strain-softening behaviour [25]. However, it
is hard to say whether there is a single crack plane, as defined
traditionally under static loading in fracture mechanics, which
takes place in front of the chevron tip under cyclic loading.
Therefore, ‘damage’ is the preferred term to use in rock-fatigue
research, rather than ‘failure plane’ (main crack). It is possible to
Fig. 6. Comparison of load–CMOD curves of CCNBD specimens tested under static
see quantified damage as the accumulation of permanent strain
and increasing cyclic loading.
within each cycle in all load–CMOD plots.
From these plots, it was found that clear tensile softening
at the tip of the chevron notch inside the sample is also different. behaviour took place for all post-peak behaviours after an
This may help to explain the fatigue mechanisms by using the FPZ accumulated plastic deformation in the sample. However, the
concept in front of the cracks. In contrast, no tensile softening microcracking damage process that causes the tension-softening
post-peak behaviour was observed with the static loading tests behaviour under cyclic loading needs to be determined. The
(see Fig. 6). These findings are significant and helpful in explain- cohesive crack model proposed in the literature may not be able
ing rock fatigue mechanisms. to explain the observed tension-softening behaviour under cyclic
Another important observation was made by examining the loading. This is because there are some limitations among
crack surfaces of failed specimens after the dynamic cyclic loading cohesive crack formulations, such as the crack tip face closing
tests. There was a clear crushed region including small particles smoothly (the stress intensity factor KI vanishes at the crack tip in
and dust in front of the chevron tip, as shown in Fig. 7b. However, mode I propagation) and the FPZ being of negligible thickness
no rock chips (small particles) were observed at the crack surfaces [25]. Conversely, it was not possible to get tensile softening or
of failed specimens under static loading, as seen in Fig. 7a. High post-peak behaviour after a plastic deformation with static load-
micro- and macrocrack density in front of the chevron tip shows a ing tests. A detailed discussion about possible fatigue damage
gradual separation of crack surfaces along the preferential path set mechanisms is given below.
up by microcracking. This damaged zone in front of the chevron
tip is called the FPZ. The mechanisms of rock fatigue under cyclic 3.3. The Mechanisms of rock fatigue damage
loading have been explained in the literature with many more
microcracks induced compared to the failure mechanisms under A typical feature of rock fatigue in experimental tests can be
static loading [23,24]. Therefore, observation of the FPZ zone in observed by producing a progressive accumulation of permanent
20 N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

Fig. 7. Failed specimens and damaged zone in front of the chevron tip (a) under static loading and (b) under cyclic loading.

Fig. 8. Load–CMOD plots of increasing cyclic loading with amplitudes (a) 10% SUL (0.45 kN), (b) 20% SUL (0.9 kN), (c) 30% SUL (1.35 kN) and (d) 40% SUL (1.8 kN).

strain in the specimen, rather than any significant decay in the as shown in Figs. 10 and 11 with different amplitudes. Plots of
material’s elastic modulus. Accumulation of plastic deformation is permanent damage show that both the CMOD and diametral axial
responsible for the fatigue damage; the magnitude and increasing displacement increase with increasing damage increments but at
trend of the irreversible deformation influences the cumulative different rates. Initially, irreversible CMOD deformation develops
fatigue damage. The displacements along both x and y directions quickly. This is followed by deformation increasing at a slow
represent CMOD and diametral axial displacement, respectively, constant rate before cumulative deformation begins to accelerate
N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26 21

rapidly until failure. Conversely, diametral axial displacement–N


plots show that the displacement rate increases with an increas-
ing rate up to failure (see Figs. 9 and 10). The opening of crack
faces parallel to the applied load and the closure of crack faces
perpendicular to the load causes certain changes in the relative
lateral and axial deformations, respectively. Behaviour of diame-
tral axial deformation results may show that irreversible damage
accumulation in front of the chevron tip is not caused by the
closing of pre-existing microcracks aligned parallel to the maximum
principal stress.
A closer examination of diametral axial displacement and CMOD
after the start of nonlinear behaviour in plots shows both diametral
axial displacement and CMOD increase with a concave upward rate
(see Fig. 11). Continuous irreversible damage occurs slowly up to a
certain point because new stress-induced microcracks are taking
place. Coalescence of the induced cracks starts after this point, with
the rate of cracking increasing in speed, up to failure.
The characteristics of the fracture surfaces were examined by
means of a JEOL JMS-6460 LA SEM. The JEOL JSM-6460 LA is a
tungsten low-vacuum analytical SEM. In this study, all rock
fracture images were obtained under low vacuum chamber
pressures, that is, 1–50 Pa (with adjustable pressure between
10 Pa and 270 Pa). This allowed certain samples to be observed
uncoated and reduced damage to the specimens caused by the

Fig. 10. (a) CMOD and (b) diametral axial displacement versus number of cycles
plots: same mean increasing cyclic loading (type II) tests with amplitude 20% SUL
(0.9 kN)

effects of high vacuum. In this study, the damaged zone in front of


the tip of the chevron notch cracks of the tested CCNBD speci-
mens were examined directly with a SEM, without preparation of
thin sections. Thin-section specimens were not used in this study
to avoid creating extra microcracks during preparation of the
specimens. Scanning electron micrographs of fracture surfaces of
Brisbane tuff CCNBD specimens tested under cyclic loading are
shown in Fig. 12.
Two fatigue mechanisms were observed in the cement of
Brisbane tuff: (1) grain decohesion in secondary microcrystalline
quartz cement intrusions (see Fig. 13) and (2) fatigue striations in
primary silica cryptocrystalline cement. Fig. 13 provides a closer
look at the cement and feldspar mineral. Some microcrystalline
quartz cement intrusions were seen in the primary silica crypto-
crystalline cement after petrographic analyses. Fig. 13b shows
those microcrystalline quartz cement intrusions. Closer examina-
tion of the damaged cement would appear to indicate that a
massive amount of loosened microcrystalline quartz minerals
resulted from grain decohesion. Further, there is a clear grain
boundary crack between the feldspar and cement (see Fig. 13a).
Similar to the fatigue damage in the cement, there were
two fatigue damage mechanisms seen with grain-related damage:
Fig. 9. (a) CMOD and (b) diametral axial displacement versus number of cycles
(1) intergranular cracks causing grain decohesion (a primary
plots: same mean increasing cyclic loading (type II) tests with amplitude 10% SUL mechanism) and (2) intragranular cracks (a secondary mechanism).
(0.45 kN). Brisbane tuff is composed mainly of quartz and K-feldspar minerals
22 N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

Fig. 11. Diametral axial displacement and CMOD after the start of nonlinear deformation in front of the chevron crack tip.

Fig. 12. Debris and small particles in front of the tip of the chevron notch as a result of cyclic loading.

with small amounts of siderite (Fe-carbonate) and zeolite minerals. presented at greater magnification in Fig. 14a. At this scale,
Fig. 14 shows the fatigue-induced cracking around the quartz and intergranular cracking around the grains is apparent. Under closer
feldspar minerals. Fatigue-related microcracking around tuff miner- examination, fine fragments around the grains were observed,
als causing grain decohesion as a result of the pulling out of grains which accumulate in the proximity of the grain corners. This is
by frictional sliding is visible in Fig. 14c and d. Fig. 14b and e are particularly evident in Fig. 14e, from which it can be inferred that
N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26 23

Fig. 13. Fatigue damage in the cement of Brisbane tuff resulting from decohesion of microscale quartz minerals.

Fig. 14. Grain decohesion and loosened grains (a)–(e) and fatigue striations in damaged grains under high magnification (  500) at the surface of the fatigue crack (f).
24 N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

such powder is the result of fatigue damage, probably due to local work on stress corrosion crack growth in rocks and ceramics
abrasion of the sharp corners of the quartz and feldspar grains. use methods originally developed for determining fracture tough-
SEM imaging revealed that the primary fatigue damage ness in metallic materials. The corrosive environment is water
mechanisms in front of a chevron notch crack are grain decohe- and there is a chemically active environment in those metallic
sion and intergranular cracks. Debris and dust are the results of material experiments. However, some mechanisms have been
fatigue damage in Brisbane tuff cement and around loosened accepted as additional mechanisms of subcritical crack growth,
grains. The fatigue cracks in the cement are restricted around the such as cyclic fatigue [23,20]. In this research, it is believed that
grains and cannot grow through the grains. Therefore, it can be fatigue cracks (e.g. intergranular cracks) are the dominant mech-
deduced that those fatigue cracks are stable (subcritical) cracks anisms of subcritical crack propagation. Thus, the main aim here
that coalescence to form macroscale fatigue cracks resulting in is to discover the possible mechanisms in microcracks causing the
failure. Further, each grain after decohesion may behave as an corrosive environment. Therefore, direct observation of the FPZ in
indenter to indent the surface of the weaker cement material front of the tip of the main crack in the tested CCNBD specimens
under far field cyclic diametral compressive loading. This would using the SEM was found a suitable observation technique for this
explain the mechanisms of the debris material around and in the research.
corners of the loosened grains. To investigate possible differences between static and fatigue
failure, the fracture surfaces of monotonically tested CCNBD
specimens were also observed by SEM. Scanning electron micro-
4. Discussion graphs of fracture surfaces of Brisbane tuff CCNBD specimens
tested under monotonic loading are shown in Fig. 15. When
The fracture toughness response to cyclic loading was found to compared with static rupture, the main differences are: (1) the
be different from that under static loading in terms of the ultimate number of fragments produced is much greater under cyclic
load and the damage mechanisms in front of the chevron crack. A loading than under static loading and (2) intergranular cracks
maximum reduction of the static KIC of 46% was obtained for the are formed due to particle breakage under cyclic loading, whereas
highest amplitude dynamic cyclic loading tested. For sinusoidal smooth and bright cracks along cleavage planes are formed under
cyclic loading, a maximum reduction of the static KIC of 29% was static loading. Further, the macroscale main crack causing failure
obtained. These reductions clearly illustrate the dramatic effect of is seen in the cement without any dust or debris material under
cyclic loading on the fracture resistance of cracks in rocks. This monotonic loading. Typical sparkling cleavage cracks resulted
means that crack propagation and damage can take place under when rupture of the crystals occurred along cleavage planes.
lower than expected ultimate loads compared with static loading.
The mode I stress intensity factor KI at the crack tip is known to
control crack growth. The stress corrosion lower limit is Ko and the 5. Conclusions
fracture toughness is KIC. When KI 4KIC, the crack grows rapidly at
approximately the speed of sound, when KI oKo, the crack does not The fracture toughness response to cyclic loading was found to
grow and when Ko oKI oKIC, the crack grows at a certain velocity be different from that under static loading in terms of the
with the stress intensity factor KI. However, this research shows ultimate load and induced plastic displacement. The maximum
that unstable crack propagation causing failure occurs with lower reduction of the static KIC of 46% was obtained for the highest
KIs at the crack tip than the KIC value. This result contradicts the amplitude dynamic cyclic loading tested. For sinusoidal cyclic
classical theory, which predicts that there will be no crack growth loading, a maximum reduction of the static KIC of 27% was
as long as KI oKIC. This phenomenon is known as ‘subcritical crack obtained. These reductions clearly illustrate the dramatic effect
growth’ [26,27]. Subcritical crack growth is one of the main of cyclic loading on the fracture resistance of cracks in rocks.
explanations for the creep damage mechanism in rocks. In some Damage was quantified as the accumulation of permanent strain
fatigue research, damage accumulation in brittle materials under in front of the chevron notch crack tip with each cycle of loading,
cyclic loading has been explained by the creep mechanism and because microfracturing introduces nonlinearity into the theoreti-
stress corrosion [28,23,29]. Despite using different test geometries cally elastic behaviour of the rock. A continuous irreversible accu-
and loading boundary conditions to those were used in past studies, mulation of damage was observed in dynamic cyclic tests conducted
this study found similar damage behaviour in front of the notched at different amplitudes. After the accumulation of irreversible
crack in our disc shaped samples as a result of dilatant creep damage and failure of the specimen, clear tensile softening was
damage behaviour. It was possible to measure the crack growth observed in cyclic loading tests carried out at different amplitudes
rate for subcritical crack growth in other studies under controlled on vertically aligned chevron notch cracks (mode I). However, no
stress corrosive environments with edge-cracked specimens. How- post-peak behaviour was observed in the CCNBD specimens tested
ever, it is not possible to measure the crack growth rate in our cyclic under static loading. Considering the shape of the load–CMOD plots,
loading experiments because of the unknown stress corrosive the accumulation of irreversible damage in front of the chevron
mechanism and the high-speed unstable crack propagation due to notch crack under cyclic loading was found to be similar to static
the embedded notch crack in the disc specimens. Further, there was fatigue (creep). Further, the subcritical crack mechanisms were
no pre-existing crack for proper monitoring in our three-dimen- verified from rock-fatigue research, with lower ultimate loads
sional sample geometry because chevron notched specimens do not causing smaller KI than the KIC due to rock fatigue.
need to be pre-cracked for rock fracture research [19]. The SEM results enable some of the qualitative features of the
Stress corrosion is the most common mechanism associated fatigue damage process in Brisbane tuff to be inferred. This
with subcritical crack growth in rock. However, it is certainly not research found that the failure of a CCNBD specimen under cyclic
the only mechanism by which subcritical crack growth occurs. loading is the result of the coalescence of many microcracks, not
Another mechanism that can be important, in certain circum- of the growth of a single macrocrack. SEM images showed that
stances, is fatigue crack growth. In this study, it has been shown fatigue damage in Brisbane tuff is strongly influenced by the
that the N to failure decreased when the amplitude of cycles failure of the matrix because of both intergranular fracturing and
increased. This effect may also be explained by stress corrosion. transgranular fracturing. The main characteristic is particle break-
Stress corrosion and fatigue represent the major mechanisms of age under cyclic loading, which probably starts at contacts
subcritical crack growth in rocks [26,23]. Most of the experimental between particles and is accompanied by the production of very
N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26 25

Fig. 15. Damage in the cement and cleavage cracks under monotonic loading.

small fragments, probably resulting from frictional sliding within [3] Hoek E. Brittle fracture of rocks. In: Stagg KG, Zienkiewicz OC, editors. Rock
the weak matrix. It is believed that point contacts at grain Mechanics in Engineering Practice. London: Wiley; 1968.
boundaries are regions of stress concentration (i.e. indenters). [4] Li YP, Chen LZ, Wang YH. Experimental research on pre-cracked marble under
Transgranular cracks may emanate from these regions and inter- compression. Int J Solids Struct 2005;42(9–10):2505–16.
[5] Al-Shayea NA. Crack propagation trajectories for rocks under mixed mode I–II
granular cracks sometimes pass through the contact points. This fracture. Eng Geol 2005;81(1):84–97.
stage can be correlated with a steady progression of damage and [6] Chang SH, Lee CI, Jeon S. Measurement of rock fracture toughness under
produces a general ‘loosening’ of the rock, which is a precursor to modes I and II and mixed-mode conditions by using disc-type specimens. Eng
the formation of intergranular cracks. When compared with static Geol 2002;66(1–2):79–97.
[7] Attawel PB, Farmer IW. Fatigue behavior of rock. Int J Rock Mech Min
rupture, the main differences are that a much greater number of
1973;10(1):1–9.
small particles and debris are created under cyclic loading than [8] Bagde MN, Pedros V. Fatigue properties of intact sandstone samples sub-
under static loading, and that intergranular cracks are formed due jected to dynamic uniaxial cyclical loading. Int J Rock Mech Min 2005;42(2):
to particle breakage under cyclic loading, whereas smooth and 237–50.
bright cracks are formed along cleavage planes under static [9] Gatelier N, Pellet F, Loret B. Mechanical damage of an anisotropic porous rock
in cyclic triaxial tests. Int J Rock Mech Min 2002;39(3):335–54.
loading. [10] Haimson BC, Kim CM. Mechanical behaviour of rock under cyclic fatigue.
Rock Mech 1971;3:845–63.
[11] Tao Z, Mo H. An experimental study and analysis of the behaviour of rock
Acknowledgements under cyclic loading. Int J Rock Mech Min Sci 1990;27:5.
[12] Burdine NT. Rock failure under dynamic failure conditions. Soc Pet Eng J
1963;3(1):1–8.
Acknowledgement is made to Leighton Contractors who pro- [13] Evans AG, Fuller ER. Crack propagation in ceramic materials under cyclic
vided core samples of Brisbane tuff from the CLEM7 Project and to loading conditions. Metall Trans 1974;5(1):27–9.
[14] Haimson BC. Effect of cyclic loading on rock. Dynamic Geotechnical Testing,
Ted Brown, Les McQueen, Mark Funkhauser and Rob Morphet of
654. ASTM STP; 1978 (p. 228–245).
Golder Associates Pty Ltd. for their assistance and advice. [15] Roxborough FF. Cutting rocks with picks. Min Eng 1963;132:445–52.
The work described forms part of the first author’s PhD research [16] Roxborough FF. Coal ploughing. Colliery Eng 1968;12(16):66–71.
carried out within the Golder Geomechanics Centre at The University [17] Hood M, Alehossein HA. Development in rock cutting technology. Int J Rock
Mech Min Sci 2000;37(1–2):297–305.
of Queensland. The first author was supported by an Australian
[18] Gottlieb L, Moore PJ. Vibratory cutting of brown coal. Int J Rock Mech Min Sci
Postgraduate Award/UQRS and the Golder Geomechanics Centre. 1981;18:335–9.
[19] ISRM. Suggested method for determining mode I fracture toughness using
cracked chevron notched Brazilian disk (CCNBD) specimens. Int J Rock Mech
References Min Sci 1995;32(1):57–64.
[20] Kim K, Mubeen A. Relationship between differential stress intensity factor and
[1] Griffith AA. The phenomena of rupture and flow in solids. Philos Trans R Soc crack growth rate in cyclic tension in Westerly granite. Fracture Mechanics
Lond 1920;221:163–98. Methods for Ceramics, Rocks, and Concrete, 745. ASTM STP; 1981 (p. 157–168).
[2] Lajtai EZ. A theoretical and experimental evaluation of the Griffith theory of [21] Singh SK. Fatigue and strain hardening behaviour of greywacke from the
brittle fracture. Tectonophysics 1971;11:129–56. flagstaff formation, NSW. Eng Geol 1989;26:171–9.
26 N. Erarslan, D.J. Williams / International Journal of Rock Mechanics & Mining Sciences 56 (2012) 15–26

[22] Shetty DK, Rosenfield AR, Duckworth WH. Fracture toughness of ceramics [26] Atkinson BK. Subcritical crack growth in geological materials. J Geophys Res
measured by a chevron notch diametral compression test. J Am Ceram Soc 1984;89:4077–114.
1985;68(12):C325–7. [27] Atkinson C, Smelser RE, Sanchez J. Combined mode fracture via the cracked
[23] Costin LS, Holcomb DJ. Time-dependent failure of rock under cyclic loading. Brazilian disk test. Int J Fract 1982;18(4):279–91.
Tectonophysics 1981;79(3–4):279–96. [28] Celestino TB, Bortolucci AA. Determination of rock fracture toughness under
[24] Hadley K. The effect of cyclic stress on dilatancy: another look. J Geophys Res creep and fatigue. In: Proceedings of the 35th US Rock Mechanics Sympo-
1976;81(14):2471–4. sium, Reno, 1995, p. 147–52.
[25] Dugdale DS. Yielding of steel sheets containing slits. J Mech Phys Solids [29] Evans AG. Fatigue in ceramics. Int J Fract 1980;16(6):485–98.
1960;8(2):100–4.