Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation

Abstract

Increasing fracture aperture by lowering effective normal stress and by inducing dilatant shearing and thermo-elastic effects is essential for transmissivity increase in enhanced geothermal systems. This study investigates transmissivity evolution for fluid flow through natural fractures in granodiorite at the laboratory scale. Processes that influence transmissivity are changing normal loads, surface deformation, the formation of gouge and fracture offset. Normal loads were varied in cycles between 1 and 68 MPa and cause transmissivity changes of up to three orders of magnitude. Similarly, small offsets of fracture surfaces of the order of millimeters induced changes in transmissivity of up to three orders of magnitude. During normal load cycling, the fractures experienced significant surface deformation, which did not lead to increased matedness for most experiments, especially for offset fractures. The resulting gouge material production may have caused clogging of the main fluid flow channels with progressing loading cycles, resulting in reductions of transmissivity by up to one order of magnitude. During one load cycle, from low to high normal loads, the majority of tests show hysteretic behavior of the transmissivity. This effect is stronger for early load cycles, most likely when surface deformation occurs, and becomes less pronounced in later cycles when asperities with low asperity strength failed. The influence of repeated load cycling on surface deformation is investigated by scanning the specimen surfaces before and after testing. This allows one to study asperity height distribution and surface deformation by evaluating the changes of the standard deviation of the height, distribution of asperities and matedness of the fractures. Surface roughness, as expressed by the standard deviation of the asperity height distribution, increased during testing. Specimen surfaces that were tested in a mated configuration were better mated after testing, than specimens tested in shear offset configuration. The fracture surface deformation of specimen surfaces that were tested in an offset configuration was dominated by the breaking of individual asperities and grains, which did not result in better mated surfaces.

Full text

ORIGINAL PAPER
Permeability Evolution in Natural Fractures Subject to Cyclic
Loading and Gouge Formation
Daniel Vogler
1
•Florian Amann
1
•Peter Bayer
1
•Derek Elsworth
2
Received: 6 September 2015 / Accepted: 3 June 2016
ÓSpringer-Verlag Wien 2016
Abstract Increasing fracture aperture by lowering effective
normal stress and by inducing dilatant shearing and thermo-
elastic effects is essential for transmissivity increase in
enhanced geothermal systems. This study investigates trans-
missivity evolution for fluid flow through natural fractures in
granodiorite at the laboratory scale. Processes that influence
transmissivity are changing normal loads, surface deforma-
tion, the formation of gouge and fracture offset. Normal loads
were varied in cycles between 1 and 68 MPa and cause
transmissivity changes of up to three orders of magnitude.
Similarly, small offsets of fracture surfaces of the order of
millimeters induced changes in transmissivity of up to three
orders of magnitude. During normal load cycling,the fractures
experienced significant surface deformation, which did not
lead to increased matedness for most experiments, especially
for offset fractures. The resulting gouge material production
may have causedclogging of the main fluid flow channels with
progressing loading cycles, resulting in reductions of trans-
missivity by up to one order of magnitude. During one load
cycle, from low to high normal loads, the majority of tests
show hysteretic behavior of the transmissivity. This effect is
stronger for early load cycles, most likely when surface
deformation occurs, and becomes less pronounced in later
cycles when asperities with low asperity strength failed. The
influence of repeated load cycling on surface deformation is
investigated by scanning the specimen surfaces before and
after testing. This allows one to study asperity height distri-
bution and surface deformation by evaluating the changes of
the standard deviation of the height, distribution of asperities
and matedness of the fractures. Surface roughness, as
expressed by the standard deviation of the asperity height
distribution, increased during testing. Specimen surfaces that
were tested in a mated configuration were better mated after
testing, than specimens tested in shear offset configuration.
The fracture surface deformation of specimen surfaces that
were tested in an offset configuration was dominated by the
breaking of individual asperities and grains, which did not
result in better mated surfaces.
Keywords Fracture mechanics Fracture transmissivity 
EGS Fracture surfaces Aperture Gouge
1 Introduction
Anthropogenic intervention and resulting perturbations
(e.g., hydraulic fracturing) in a rock mass at great depth
may result in complex thermal-hydro-mechanical response.
This is of particular relevance when dealing with enhanced
geothermal systems (EGS) and unconventional oil and gas
extraction utilizing hydraulic fracturing. Our focus here is
on EGS, which is considered to be a promising option for
generating electricity from hot but naturally low permeable
deep formations (Tester et al. 2006). Studying the key
THM coupled processes associated with specific reservoir
characteristics in an EGS is of foremost relevance to
establish a reliable heat exchanger capable of achieving the
target production rate while maintaining low background
seismicity. The triggering of elevated seismic activity can
cause the termination of EGS projects due to damage to
&Daniel Vogler
daniel.vogler@erdw.ethz.ch
1
Department of Earth Sciences, Swiss Federal Institute of
Technology Zurich, Sonneggstr. 5, 8092 Zurich, Switzerland
2
Department of Energy and Mineral Engineering, EMS
Energy Institute, Center for Geomechanics, Geofluids, and
Geohazards, Pennsylvania State University, University Park,
PA, USA
123
Rock Mech Rock Eng
DOI 10.1007/s00603-016-1022-0

facilities on the surface (e.g., the hot dry rock EGS project
in Basel in 2006). Rock transmissivity and flow rates in
fractures determine the productivity of geothermal reser-
voirs and are controlling factors of whether reservoirs are
economically feasible. Lowering effective normal stresses
by high-pressure injection, dilatant shearing of critically
stressed fractures or thermo-elastic effects (e.g., cooling)
can cause fracture conductivity to increase (Evans 2005).
To relate mechanical and hydraulic effects in fractures,
the mechanical and hydraulic aperture (amand ahyd ) are
generally considered separately (Witherspoon et al. 1980;
Barton et al. 1985; Esaki et al. 1991; Zimmerman et al.
1991; Park et al. 2013). While the mechanical aperture
describes the physical distance between two fracture sur-
faces, the hydraulic aperture describes the aperture
accommodating a particular flux assuming a parallel plate
model. With increasing mechanical aperture the hydraulic
aperture increases, but the relation between the mechanical
aperture and the hydraulic aperture is not one to one (Esaki
et al. 1991,1999; Rutqvist and Stephansson 2003; Xiong
et al. 2011 and McClure and Horne 2014).
The common constitutive equations are based on aper-
ture and linearly relate the flow rate in a fracture to the
pressure gradient. Fluid flow between parallel walls can be
derived from the simplified, incompressible Navier–Stokes
equations (Louis 1969), as,
Q¼a3
hydw
12lrpð1Þ
where Qrepresents the fluid flow rate, ahyd the hydraulic
aperture, wthe fracture width, lthe dynamic viscosity and
pthe fluid pressure.
HM-coupled laboratory investigations reveal that
assuming the same mechanical and hydraulic aperture is
not generally valid (Raven and Gale 1985; Brown 1987;
Cook 1992; Renshaw 1995; Hakami and Larsson 1996;
Oron and Berkowitz 1998; Esaki et al. 1999; Chen et al.
2000 and Lee and Cho 2002).
Laboratory tests on granite, marble and basalt specimens
with tension-induced artificial fractures by Witherspoon
et al. (1980) and on artificial resin fractures by Li et al.
(2008) found the approximation of fluid flow with a par-
allel plate model with an effective hydraulic aperture to be
suitable depending on the fracture roughness characteris-
tics. Findings by Raven and Gale (1985), Cook (1992),
Hakami and Larsson (1996), Oron and Berkowitz (1998),
Esaki et al. (1999), however, showed hydraulic apertures to
be consistently smaller than mechanical apertures. Studies
by Cook (1992) and by Oron and Berkowitz (1998) showed
decreasing mechanical aperture leading to a faster than
cubic decrease in hydraulic conductivity and a nonlinear
increase in contact area. Zimmerman et al. (1991)
demonstrated that the ratio of the hydraulic aperture to the
mean mechanical aperture is related to the ratio of the
mean mechanical aperture and its standard deviation. In
contrast, Renshaw (1995) found the hydraulic aperture to
be constant below a residual hydraulic aperture, while the
mechanical aperture could still decrease further when the
normal load was increased.
Analysis of experimental data by Cook (1992) and Park
et al. (2013) showed mechanical normal joint stiffness and
hydraulic aperture to be largely dependent on the contact
area between fracture surfaces. Joint stiffness, joint closure
and fluid flow were shown to behave highly nonlinearly.
Also, as normal loads on a specimen are cycled, hysteretic
behavior results. The exponent of the aperture is increas-
ingly higher than cubic with progressive fracture closing.
Experimental work by Durham and Bonner (1994), Esaki
et al. (1991), Esaki et al. (1999), Chen et al. (2000), Li et al.
(2008), Park et al. (2013) found the hydraulic conductivity of
natural fractures to be significantly larger with fracture shear
displacements. Esaki et al. (1991,(1999) and Lee and Cho
(2002) also found increases in transmissivity to become
smaller with ongoing shear displacement. This behavior was
controlled by the maximum asperity heights and the distance
between asperities. For repetitive forward and reverse shear-
ing, the development of wear products from fracture surface
deformation (gouge) was found to decrease transmissivity by
an order of magnitude. A commonly used relationship
between shear and normal stress to represent changes in
conductivity was developed by Barton et al. (1985). This
relationship was tested for rock specimens of various sizes and
utilizes surface roughness quantities [i.e., the joint roughness
coefficient (JRC)] to quantify surface roughness.
While the hydraulic aperture inferred from experiments
normally only represents an averaged hydraulic aperture,
fracture flow depends on fracture surface heterogeneity, which
is also strongly fracture size dependent (Renshaw 1995;Yeo
et al. 1998; Oron and Berkowitz 1998; Pyrak-Nolte and Morris
2000;Walsh2003;Lietal.2008,2014;Parketal.2013).
The findings of previous experiments demonstrate that
the cubic law and a linear relation between mechanical and
hydraulic aperture are only partially valid, which serves as
motivation for this study. In this study, several coupled
processes of importance for reservoir modeling are exam-
ined for a single fracture. The overall objective is to define
the evolution of fracture transmissivity in natural fractures
under applied normal and shear displacements as an analog
to the response of fracture networks to changes in applied
effective stresses driven by fluid pressures and thermal
loads. Changes in transmissivity due to changes in con-
fining stress on the specimen are studied for single loading
cycles. This includes an analysis of fracture surface dam-
age and the production of gouge material. The experi-
mental setup allows one to compare changes in mechanical
and hydraulic aperture during specimen loading.
D. Vogler et al.
123

For increases in confining stress, the effect of the mat-
edness of the two fracture surfaces is examined by per-
forming experiments with well-mated and offset fracture
surfaces. This provides qualitative as well as quantitative
information on the impact of dilation due to shear dis-
placement, production of gouge material, and quantitative
insight into the magnitudes of transmissivity and trans-
missivity changes during cyclic loading.
2 Methods
2.1 Specimen Preparation
Granodiorite specimens were provided by the Grimsel Test
Site, Switzerland. The specimens were obtained from cores
of the CRIEPI fractured rock study (Takana et al. 2014).
Twelve laboratory specimens with a diameter of 2.5 cm
and a length of 6 cm were produced by overcoring pre-
existing fractures in 10.5 cm cores (Table 1; Fig. 1). The
overcoring orientation was chosen so that fractures are
aligned parallel to the specimen’s main axis. The fractures
were classified as tensile (mode I) and shear (mode II)
fractures (Table 1). Classification of the fracture type was
performed by investigating fracture surfaces for slicken-
sides, plumose structures, mineralization and crack propa-
gation through individual grains.
2.2 Experimental Setup
The experimental setup consisted of a pressure-tapped core
holder of the DCH series (Fig. 2, Table 2), as produced by
Core Lab. The core holder setup was described in detail by
Wang et al. (2011).
One Isco pump (model 500D; pump PA) and two Isco
pumps (model 100DM; pumps PBand PC) were utilized to
control the fluid pressure at the specimen inlet pfp (pump
PC), the confining pressure pco (pump PB) and to measure
Table 1 Physical properties of
the specimen under
investigation with original
specimen length, shear
displacement offset, the
resulting effective specimen
length and the distinction
between tensile and shear
fracture mode
Test specimen Specimen length Shear offset Eff. length Frac. mode
(–) (mm) (mm) (mm) (–)
1 61 0 60 I
2 62 0 60 II
3 62 0 60 I
4 60 0 60 I
5 60 2 58 I
6 60 2 58 I
7 60 3 57 I
8 61 3 58 I
9 60 6 54 I
10 62 5 57 II
11 60 1 59 I
12 60 1 59 I
Fig. 1 Example of overcored specimen on top of one fracture surface
Fig. 2 Experimental setup of core holder and specimen. The
pressures were prescribed in pumps PA(axial pressure pax), PB
(confining pressure Pco) and PC(fluid inlet pressure pfp )
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

the axial pressure pax (pump PA). Similar setups for
investigation of rock specimen properties were employed
by Zhu et al. (2007), Wang et al. (2011), Shugang et al.
(2013) and Zhong et al. (2014), but focusing on different
research questions.
The specimen placement with the fracture oriented
parallel to the core holder axis produces fluid flow along
the fracture from the inlet to the outlet. Specimens were
loaded with an initial confining stress of 1 MPa, and
steady-state flow was established. At all times, the fluid
pressure at the outlet was kept at atmospheric pressure patm
and the inlet pressure was recorded. When initial equilib-
rium of fluid pressure for a given flow rate was reached, the
confining stress was increased incrementally by 1–2.5 MPa
up to a maximum of 68 MPa and subsequently decreased
back down to 1 MPa, thereby completing one loading
cycle. After each incremental increase in confining stress,
the confining stress was kept constant until the fluid inlet
pressure necessary to maintain a constant flow rate
Qthrough the specimen reached a steady-state condition.
During the experiment, the flow rate Qwas kept constant
for individual loading cycles. Accuracy of flow rate mea-
surements was 0.5 %of setpoint (i.e., the currently read
pressure in the pump). Between 5 and 10 loading cycles
were performed for each specimen.
2.3 Permeability Behavior of Natural Fractures
Under Various Loading Scenarios
The experimental setup was used to test transmissivity
changes of natural fractures under changing confining
stress conditions. Confining stress is applied radially on
the cylindrical specimen. The fracture surfaces were
investigated in mated and offset configurations with
offsets among fracture surfaces ranging between 1 and
6 mm. An example of open and closed fractures is
shown in Fig. 3. Offset of the fractures occurred in the
axial direction. Differentiating between mated and offset
fracture surfaces enabled us to investigate the effect of
shear displacement on transmissivity and surface
deformation.
The main measured properties are the inlet fluid pressure
pfp and the confining pressure pco. The effective confining
pressure pco;eff was calculated by
pco;eff ¼pco pfp þpatm
2ð2Þ
A representative transmissivity Twas derived from the
cubic law (Eq. 1) to allow one to investigate changes in
transmissivity, with the transmissivity being independent
of changes in flow rate in subsequent cycles. If the flow rate
Qis changed between load cycles, the representative
transmissivity Tdoes not change if the changes in pressure
gradient rpare proportional to the changes in flow rate.
Q¼a3
hd
12lrpð3Þ
T¼a3
h
12l¼Q
drp¼Q
d
DL
Dpð4Þ
where dis the specimen diameter, ahis the hydraulic
aperture, pthe flow pressure and lthe dynamic viscosity of
water. The transmissivity Tis used to make a comparison
between individual experiments because Trelates the flow
rate driven by a unit pressure gradient, rather than a head
gradient in the normal hydrogeological definition. Note that
the transmissivity Tis related to the normal hydrogeolog-
ical definition of transmissivity Thas Th¼Tcwhere cis
the unit weight of the flowing fluid. Reporting the trans-
missivity offers the advantage of reporting all properties
known in the experiment (e.g., Q,DL,Dpand d) in one
variable. This allows straightforward comparison of
experiments performed with different flow rates Qor on
fractures with varying width d. Similar forms of the
transmissivity were also used in Renshaw (1995), Gentier
et al. (2013) and Zimmerman and Bodvarsson (1996).
Besides the hydraulic analysis, surface deformation was
determined and mineralogical characterization of the
fracture surfaces was performed.
2.4 Surface Scans
Prior to testing, replicas of the fracture surfaces were
produced. These were utilized to obtain high-resolution
photogrammetric scans of the surface. Fracture surfaces
were evaluated with surface scans recorded with the ATOS
Table 2 Pumps used in the setup with model, minimum and maxi-
mum pressures pmin and pmax, standard pressure accuracy SPA
PAPBPC
pfp pco pax
Model (–) 500D 100DM 100DM
Min. pressure (MPa) 0.67 0.67 0.67
Max. pressure (MPa) 25.9 69 69
Volume (mL) 507 103 103
SPA (% FS) 0.1 0.1 0.1
Fig. 3 Example of overcored specimen aopen and bclosed
D. Vogler et al.
123

Core 3D scanner from GOM. The ATOS Core sensor
projects fringe patterns on the object surface, which are
recorded by two cameras. The patterns form a phase shift
that is based on a sinusoidal intensity distribution which
enables one to calculate the three-dimensional (3-D) sur-
faces. The photogrammetry scanner is calibrated with two
tests. The diameter and shape of a sphere and the distance
between two spheres that are mounted on a plate are
measured with the photogrammetry scanner to derive cal-
ibration errors and accuracy. All equipments used for cal-
ibration are specifically developed by the company GOM,
which manufactures the scanner. The photogrammetry
scanner was calibrated with length deviation errors
between 0.009 and 0.027 mm and optimized calibration
deviations of 0.014 ±0.001 pixels.
The same process was repeated for the damaged sur-
faces after testing to analyze the surface changes that
occurred during the experiments and compare these to the
changes in transmissivity during the experiment. Surface
damage during testing caused gouge material to accumu-
late in the fracture planes. Analyzing the gouge material
found in the fracture after testing can give insight into
possible relations between the grain size distribution of the
gouge material and transmissivity changes during
mechanical cycling.
A similar analysis is performed by computing the
asperity distributions on the fracture surfaces derived from
the fracture scans before and after the tests. The changes on
the fracture surfaces can also be used to determine probable
relations to transmissivity changes during testing.
2.5 Grain Size Analysis
The grain size distribution of the gouge material produced
during testing was measured by sieving samples larger than
1 mm and by using laser spectrometry (Malvern Instru-
ments Mastersizer 2000) for all gouge material below
1 mm. The gouge material is analyzed for volumetric dis-
tribution of gouge instead of weight. This is done because
laser spectrometry allows a very accurate measurement of
grain size volume down to around 10 micrometers.
2.6 Aperture Changes
The volume change of the confining fluid during cyclic
loading was recorded in each experiment and was used to
calculate the volumetric strain. After removing noise and
accounting for confining pressure changes, these data were
used to quantify relative mechanical aperture changes
during cycling, assuming that the volume change is entirely
associated with fracture deformations.
When rubber jacket and specimen volumes decrease
(Fig. 2), the pump controlling the confining pressure has to
adjust for this increase in volume of the confining pressure
fluid. The volume of the rubber jacket can decrease due to
deformation of the jacket and the specimen enclosed
within. The rubber jacket itself deforms elastically, which
allows one to separate the volumetric changes of the rubber
jacket and the specimen. These changes can be attributed to
the elastic deformation of the specimen itself as well as to
reversible and permanent changes of the fracture. To cal-
ibrate this, an additional test with an intact granodiorite
specimen was performed for loading between 1 and
68 MPa. Therefore, the changes in confining fluid volume
dðVp;conf Þare related to the changes in mechanical aperture
damech by
damech ¼dVp;conf
dlð5Þ
where dis the specimen diameter and lis the specimen
length. The change in hydraulic aperture derived from fluid
pressure increase can be compared to the mechanical
aperture decrease, and the hydraulic aperture can be cal-
culated with Eq. 3. Changes in hydraulic aperture are
measured by comparing the hydraulic aperture during the
experiment to the initial value at the start of the experiment
with 1 MPa confining pressure.
3 Results
In the performed experiments, we measured the fluid
pressure gradient response to changing confining pressure
on a natural fracture. The flow rate Qwas kept constant for
individual cycles, and the required fluid pressure gradient
rpto maintain Qwas measured. A representative fracture
conductivity can be calculated in form of the transmissivity
(Eq. 4). Hence, transmissivity is used to quantify fracture
conductivity changes during the experiments. This section
reports on the effect of cyclic loading on the fluid flow,
gouge material and the surface damage that can be evalu-
ated after completion of the experiments. To investigate the
effect of shear displacement on cyclic loading, two classes
of tests were performed on specimens with mated surfaces
(tests 1, 2, 3, 4) and on specimens with shear displacement
of 1–6 mm (tests 5, 6, 7, 8, 9, 10, 11, 12). A detailed
discussion follows in the subsequent Sect. 4. The feasible
number of loading cycles was determined for each speci-
men and the experimental device. Due to the large opening
of a natural fracture in comparison with a saw-cut fracture,
each specimen was wrapped in a smaller jacket within the
larger rubber jacket separating the specimen from the
confining fluid. Nonetheless, rupture of the rubber jacket
led to the termination of some tests and a maximum of 10
cycles was performed to limit wear on the material. Other
reasons for termination of experiments were failure of the
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

specimen under high normal loads and fluid inlet pressures
above 30 MPa. It should be noted here that the effective
confining stress in these extreme cases was very variable
along the specimen axis, due to the large pressure gradients
observed between inlet and outlet. The equilibration times
for fluid pressures strongly varied depending on the frac-
ture opening, existing gouge material and applied confining
pressures. These times ranged from minutes up to multiple
hours for individual changes in confining pressure that
were between 1 MPa for small pressures (i.e., 1–10 MPa)
and 2.5 MPa for larger pressures (i.e., 10–68 MPa).
Experimental data that are used for further analysis
include volume changes, time and pressures that are mea-
sured at each pump for confining and fluid flow pumps
(Fig. 2). Transmissivity (Eq. 4) and effective confining
pressure are chosen to represent the experimental data.
Changes in surface properties are characterized by changes
in asperity height and the standard deviation of asperity
height. These are obtained from inspection of the surface
scans. For a concise description of results, we select only
experiments 1, 2, 4, 7 and 10 (Fig. 4a–e). The selected
experiments are considered representative of hysteretic
behavior during normal load increase and transmissivity
changes with ongoing cycling. Tests 2 and 4 (Fig. 4b, c)
display strongly hysteretic fluid inlet pressures, which are
linearly related to the transmissivity. Specimens 2 and 4
were tested in a mated configuration. Specimens 7 and 10
were tested in an offset configuration (3 and 5 mm,
respectively) and show hysteretic fluid pressures responses
for late cycle numbers (e.g., cycles 4 and later for tests 7
and 10) as well.
3.1 Permeability Changes During Load Cycling
Figure 4a–e depicts the transmissivity changes as derived
from Eq. 4for changing effective confining pressure. The
color coding of the data curves ranges from the first cycle
(dark red) to the last cycle (dark blue), and the increasing
confining pressure path of a cycle is marked with a ., while
the decreasing confining pressure path is marked with a /.
The general trend shows fracture transmissivity decreasing
with increasing cyclic loading (Fig. 4a–c). During the first
initial cycles, transmissivity declines rapidly and converges
toward later cycling.
Most experiments (Fig. 4b–e) display hysteretic behav-
ior during cycling as fracture transmissivity is lower during
unloading of the specimen than during loading. The cyclic
loading also allows one to compare transmissivity decrease
due to changes in confining stress and increasing cycling
and accompanying surface damage. During experiment 1
(Fig. 4a), transmissivity decrease due to larger confining
stresses during individual test cycles is on the same order
of magnitude as decreases due to irreversible fracture
closure (e.g., due to surface damage) with repeated load
cycling. Load cycle 4 shows exceptional behavior with
significantly larger initial transmissivity decreases than all
other cycles. This is likely related to clogging of a main
flow channel in the fracture or clogging of the fluid pipes
upstream or downstream of the specimen. The effect of
changing confining stress during individual cycles is more
pronounced for experiments 2 and 4 (Fig. 4b, c) where
transmissivity decreases strongly during initial confining
pressure increase and converges against a constant trans-
missivity value between 10 and 20 MPa effective confining
pressure.
In experiment 10 (Fig. 4e), the fracture aperture is very
large and the pressure gradient is quite small until
47.5 MPa confining pressure is reached. The fluid pressure
gradient then rises quickly, which means that transmis-
sivity is decreasing quickly.
Depending on the surface of the fractures, different
response patterns of the transmissivity to increased con-
fining pressure can be observed. Specimen 1 (Fig. 4a)
exhibits a small slope that is almost linear on a semilog
plot, which stands in stark contrast to specimens 2 and 4
(Fig. 4b, c). The surface of specimen 1 shows significant
mineralization and slickensides (Sect. 1) with two well-
mated surfaces with a uniform distribution of contact area.
While this leads to almost no hysteretic effects between the
increasing and decreasing confining stress paths, it also
causes a significantly less steep transmissivity decline for
small confining stress (1–10 MPa). Tests with shear offset
show higher transmissivities (Fig. 4d, e). The large range
of transmissivity values for respective confining stresses
can also be observed for all specimens in Fig. 13. While
increased confining stress and surface damage cause
transmissivity values for individual specimen to vary up to
three orders of magnitude, the offset of the specimen only
seems to have an effect up to 1 mm. For offsets between 1
and 6 mm, the total transmissivity range is comparable for
all tests.
3.2 Aperture Changes
Measurements during the experiments only allow one to
deduct changes in mechanical aperture, as absolute values
are not locally known and measured changes are repre-
sented as averages across the whole specimen. Therefore,
changes in the mechanical and hydraulic aperture from the
starting value of each cycle at 1 MPa confining stress are
presented. As both aperture values generally decrease from
their starting value at 1 MPa, positive aperture changes
denote a reduction of the mechanical or hydraulic aperture.
After maximum effective confining stress is reached, the
mechanical aperture changes generally decrease, indicating
fracture opening. Changes in mechanical and hydraulic
D. Vogler et al.
123

aperture are compared in Fig. 5a–d. As for Fig. 4a–e,
stages in the experiment with equilibrated confining and
injection pressures are displayed with .and /during
increasing and decreasing confining pressure, respectively.
Closely spaced markers, therefore, indicate small changes
during one confining pressure interval change (2.5 MPa)
while markers spaced further apart indicate rapid changes
with confining pressure changes.
The first cycle is not always shown for reasons given
below. The specimen and the experimental equipment (i.e.,
the rubber jacket, see Fig. 2) deform during the first cycle,
which can lead to erratic aperture changes. This behavior
Fig. 4 Tests 1, 2, 4, 7 and 10 are shown in subfigures a–e,
respectively. The plots show transmissivity (Eq. 4) versus effective
confining pressure. The first number in each row in the legend denotes
the cycle number, with color coding of the data curves going from the
first cycle (dark red) to the last cycle (dark blue). The second number
in each row in the legend denotes the flow rate during the respective
cycle in (mL/min). The curve from 1 MPa to maximum effective
confining pressure is marked with .while the reverse curve is marked
with /(color figure online)
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

can be especially pronounced for the offset specimens
where the two fracture sides have a small contact area.
Small shear displacements at high confining stresses can
lead to rapid fracture normal closure. For offset fracture
surfaces (eg., tests 7 and 10), oscillating behavior and rapid
changes during initial loading cycles are especially pro-
nounced, due to large initial apertures and fluid flow
pressure oscillations.
During test 1, mechanical aperture changes are more
than one order of magnitude larger than hydraulic aperture
changes (Fig. 5a). For each cycle, the hydraulic aperture
changes increase initially drastically until they start to
converge. Hydraulic aperture changes for test 4 (Fig. 5b)
are also significantly smaller than mechanical aperture
changes. Maximum mechanical aperture changes decrease
with ongoing load cycling while maximum hydraulic
aperture shows no distinct trend. Mechanical aperture
changes (i.e., fracture closure) are largest for test 7 (i.e., a
specimen that contained a fracture that was tested with
3-mm offset, Fig. 5c). For test 10 (5-mm offset),
mechanical aperture initially changes drastically, marked
by wide marker spacing. While mechanical aperture
change increases further as the fracture closes, hydraulic
aperture changes start converging.
For all tests, maximum mechanical aperture changes
(i.e., maximum mechanical aperture closure) remain com-
parable between cycles (around 1 mm except for test 4
shown in Fig. 5b with 0.2 mm). The mechanical aperture
changes of tests 1, 4, 7 and 10 can also be interpreted when
comparing mechanical aperture changes to effective con-
fining pressure (Fig. 6a–d). All tests show larger mechan-
ical aperture changes during initial loading (up to 10 MPa),
which become smaller for higher effective confining
stresses. Other visible trends include small decreases of
mechanical aperture changes with ongoing load cycles and
hysteretic behavior of mechanical aperture changes. For all
tests, the mechanical aperture change is larger than the
hydraulic aperture change after the maximum effective
confining pressure has been reached and confining stress is
decreasing (Fig. 5a–d). Hysteretic effects are more pro-
nounced for mated specimen (Fig. 5a–b), but are still
observable for offset specimen (Fig. 5c–d).
(a)
ahyd change [mm]
0 0.01 0.02 0.03 0.04
amech change [mm]
0
0.2
0.4
0.6
0.8
1
2 - 20
3 - 20
4 - 20
5 - 20
6 - 20
7 - 20
8 - 20
9 - 20
10 - 20
Q [mL/min]
(1)
(b)
ahyd change [mm] ×10-3
02468
amech change [mm]
0
0.05
0.1
0.15
0.2
0.25
2 - 0.1
3 - 0.3
4 - 0.5
5 - 0.1
6 - 0.3
7 - 0.5
Q [mL/min]
(4)
(c)
ahyd change [mm]
0 0.02 0.04 0.06 0.08
amech change [mm]
0
0.5
1
1.5
2 - 10
3 - 10
4 - 10
5 - 10
Q [mL/min]
(7)
(d)
ahyd change [mm]
0 0.02 0.04 0.06 0.08 0.1
amech change [mm]
0
0.2
0.4
0.6
0.8
1
2 - 10
3 - 10
4 - 10
5 - 10
6 - 10
Q [mL/min]
(10)
Fig. 5 Mechanical versus hydraulic aperture changes. Changes are
calculated by comparison with the initial values at 1 MPa. Specimens
from tests 1, 4, 7 and 10 are assigned to a–d, respectively. The curve
from 1 MPa to maximum effective confining pressure is marked with
.while the reverse curve is marked with /
D. Vogler et al.
123

Here, it should be noted that for tests depicted in
Figs. 5a–d and 6a–d, only test 4 (Figs. 5b, 6b) experienced
fluid inlet pressures high enough to cause significant dif-
ferences between the confining stress and effective con-
fining stress.
3.3 Analysis of Gouge Material
Due to the experimental setup, fine and coarse grain
material was not collected at the outflow end of the
experiment. The gouge material collected after testing was
analyzed under a magnifying lens, which found the min-
erals quartz, feldspar, biotite and chlorite, all common in
granodiorite. Specimens 2 and 10 showed slickensides.
The collected volume of gouge material (Fig. 7) follows
a log-normal distribution. Since the tests in mated config-
uration did not lead to significant surface damage, the
amount of gouge material was not sufficient to derive a
distribution for specimens without shear offset. Specimen 7
had whole individual grains of rock breakouts with sizes up
to 6 mm that were not monominerals. Larger quartz
crystals were the dominant minerals on the fracture surface
of specimens 1 and 7. Overall, mineral distributions on the
fracture surface and of the gouge material revealed to be
comparable and did not differ largely for individual spec-
imens and between all specimens. No mineralogical
Fig. 6 Mechanical aperture changes versus effective confining
pressure. Changes are calculated by comparison with the initial
values at 1 MPa. Specimen from tests 1, 4, 7 and 10 are assigned to
a–d, respectively. The curve from 1 MPa to maximum effective
confining pressure is marked with .while the reverse curve is marked
with /
Size [mm]
10 -3 10 -2 10 -1 10 0
Volume Fraction [-]
0
2
4
6
8
10
12
14
05
06
07
08
09
10
11
12
Fig. 7 Grain size distribution of gouge material for tests 5, 6, 7, 8, 9,
10, 11 and 12
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

features larger than the average grain sizes exist in the
tested specimens. Changes in transmissivity during normal
load cycling are more likely connected to the surface
geometry and the specimen offset than the mineralogy of
the specimen. This is supported by comparable grain size
distributions across tests (Fig. 7). While mineralogy on
fracture surfaces was comparable across specimens, chan-
ges in transmissivity during cyclic changes in confining
stress showed strong correlation to shear offset during
testing (Fig. 4a–c for mated specimens versus Fig. 4d–e for
offset specimens and Fig. 13).
Larger rock pieces and individual grains that broke off
stayed in place during all tests. Smaller grains and fine
material were found evenly distributed across the fracture
after testing. Therefore, while larger breakouts of material
led to stronger surface alteration, fine material is more
likely to clog flow channels with gouge that leads to
increasing pressure gradients to maintain flow rates.
3.4 Comparison of Surface Scans Before and After
Testing
Due to the large maximum confining pressures, significant
surface deformation was expected in the fracture plane of
the specimens. For detailed analysis, photogrammetric
surface scans of the fracture specimens were generated
before and after the experiments.
The surface scans of the individual sides can be matched
to derive the fracture aperture by calculating the distances
between the two surface sides. This is fundamental to study
the effect of surface roughness, connectivity and the cor-
relation length of the fracture surfaces as well as the
respective aperture, which is especially crucial for fracture
transmissivity.
Surface roughness can be quantified with the standard
deviation hstd of the asperity height hin a distribution with
an average (mean) asperity height have .
hstd ¼X
N
i¼1
ðhave hiÞ2
Nð6Þ
Another measure to characterize surface roughness is the
correlation length of the asperity height, which gives an
indication of the directional dependence that is to be
expected for shear displacement, which affects fluid flow
through the fracture. The correlation length ncan be found
by studying the convergent behavior of the function H,
which is defined as
H¼Pn
i¼1ðzðcxyÞzðcxy þrÞÞ2
nð7Þ
where cxy is the spatial coordinate in the x- and y-direction
on the fracture surface, zis the asperity height at cxy and n
is the number of spatial locations that are at a distance of r
from cxy. The correlation function Hwas computed in the x
and y-direction, with the correlation length ndefined as the
point when Hdoes not increase further for r[n. For a
very small correlation length in the offset direction, the
effect of dilation due to the offset is expected to become
constant once the offset distance reaches the correlation
length. This has also been found by Yeo et al. (1998) and
Kim and Inoue (2003). Studies by Iwano and Einstein
(1993) and Hakami et al. (1993) determined the correlation
length in apertures of different rock types and linked
increasing correlation length to smaller areas of contact and
more pronounced flow channels.
The asperity distribution before and after testing shows
drastic changes when comparing the ranges observed for
the cumulative density functions of asperity height distri-
butions on the fracture surfaces (Fig. 11). Figure 11 shows
more comparable distributions of asperity heights across all
fracture surfaces before testing, with asperity height dis-
tributions becoming more heterogeneous after testing when
compared between all tested specimens. Photogrammetric
scans produced profiles of the surfaces, which were ori-
ented according to a best-fit plane in the x–ycoordinates.
Therefore, the mean asperity height is close to 0 (Figs. 8,
9). However, for visualization and comparison, an asperity
height of 0 mm is assigned to the lowest point of the sur-
face for Fig. 11. While asperity distribution for individual
specimens is similar to both fracture surfaces before test-
ing, the two sides show different distributions after testing.
This can be attributed to surface damage at individual
contact points occurring on the specimen side with the
lower asperity strength.
Surface damage also caused the measured correlation
lengths to change (Table 3). While the correlation function
tended to converge slightly slower for the mated tests (tests
1–4), convergence was changed more significantly for offset
tests, with the correlation length increasing (e.g., tests 7 and
8), or even decreasing for one of the surfaces (e.g., surface B
in test 9). An example of the correlation length function in x-
and y-direction for specimen 6 is shown in Fig. 12.
Transmissivity values for the maximum effective con-
fining pressure (varies between experiments) and for 1 MPa
after each load cycle are shown in Fig. 13a, b, respectively.
The marker size changes from early cycles (small circle) to
later load cycles (large circle), which enables one to observe
quantitative changes of the transmissivity with ongoing load
cycles and shear displacement. The impact of effective
confining pressure is more pronounced for mated fractures.
This becomes apparent when comparing Fig. 13a, b, where
specimen with shear offset show transmissivity changes
between maximum effective confining stress and at 1 MPa
confining stress of one order of magnitude or smaller.
Specimens in mated configuration, however, can show
D. Vogler et al.
123

transmissivity differences of up to two orders of magnitude
(e.g., specimens 3 and 4 in Fig. 13a, b). The transmissivity
behavior at maximum effective confining stresses does not
appear to be influenced by additional shear offset past 1 mm
in Fig. 13a, while there is a trend of increasing minimum
transmissivity values for 1 MPa effective confining pressure
after cycling (Fig. 13b). The general trend of transmissivity
decrease for ongoing load cycling is evident in all specimens
at maximum effective confining stress and at 1 MPa con-
fining stress after each cycle, with changes in magnitude
varying from half an order of magnitude (test 1, 0 mm shear
displacement) to three orders of magnitude (test 6, 2 mm
shear displacement).
4 Discussion
4.1 Fracture transmissivity
Two fundamental behaviors were observed during this
study. First, in the case of mated fractures the
transmissivity decreases rapidly during individual load
cycles for effective confining stresses below 10 MPa
(Fig. 4a–c). The minimum transmissivity observed during
each cycle only shows small changes significantly below
one order of magnitude between cycles. However, the
minimum transmissivity generally decreases with an
increasing number of load cycles (Fig. 4a–c). Secondly, in
the cases of offset fractures, the transmissivity does not
always decrease significantly during individual cycles, with
maximum differences between subsequent cycles of one
order of magnitude or lower (Fig. 4d, e). Exceptions are
cycles 2 and 6 in experiment 10 (Fig. 4e). Changes of
minimum transmissivity between cycles were more pro-
nounced than for individual cycles and mated specimen,
with observed changes of multiple orders of magnitude
(Fig. 4d, e).
The fact that fault gouge was produced (Fig. 7) during
cyclic loading suggests that the transmissivity decrease in
both cases is associated with the transport of fault gouge
material downstream after surface damage occurred.
Surface A Surface B
(a)
Test 1
(b)
Test 2
(c)
Test 4
(d)
Test 7
(e)
Test 10
(f)
Fig. 8 Surface scans before testing. Specimens from tests 1, 2, 4, 7
and 10 (top to bottom,a–e) with surfaces A (left) and B right before
testing. A reference figure for the surface scan dimensions and
asperity height colorbar is shown in f(color figure online)
Surface A Surface B
(a)
Test 1
(b)
Test 2
(c)
Tes t 4
(d)
Test 7
(e)
Tes t 1 0
(f)
Fig. 9 Surface scans after testing. Specimens from tests 1, 2, 4, 7, 10
(top to bottom) with surfaces A (left) and B (right) before testing. A
reference figure for the surface scan dimensions and asperity height
colorbar is shown in f(color figure online)
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

Gouge material transport likely occurs during decreasing
confining pressures when the fracture aperture increases
again and gouge material that was previously lodged in
place can be transported downstream. This observed gouge
material production could potentially cause the observed
hysteretic behavior by subsequently clogging flow paths,
thus lowering fracture transmissivity (Fig. 4b–e).
The results for the mated fractures in Fig. 4show that
most of the transmissivity changes between load cycles
occured during the initial cycling (Fig. 4a–c). Mated spec-
imen have more contact area, which can be seen in their
aperture fields (Fig. 10a–c). This likely leads to less surface
damage as indicated by the small gouge production mea-
sured for mated specimen (Fig. 7). The smaller changes
observed in transmissivity behavior between load cycles for
mated specimen could therefore be linked to less pro-
nounced surface damage. During individual cycles, high
effective confining stresses greater than 20 MPa (Fig. 4b,
c) do not lead to further transmissivity decrease for mated
specimens. As mechanical aperture changes were recorded
for effective confining stresses above 20 MPa, this indicates
that additional changes in mechanical aperture (Fig. 6a, b)
do not affect fluid flow. One possible explanation for this is
a crucial change in flow regime from distributed flow across
the whole specimen width to channelized flow. This
hypothesis is supported by the converging hydraulic aper-
ture changes while mechanical apertures are still decreasing
(Fig. 5a, c, d). The slow decrease in transmissivity observed
in specimen 1 (Fig. 4a) is in contrast to specimens 2 and 4
(Fig. 4b, c). The aperture fields (Fig. 10a–c) obtained from
specimens before and after testing give indications that this
could also be related to surface damage. When comparing
aperture fields before and after testing for specimen 1, there
is no significant aperture decrease across the fracture
(Fig. 10a). Specimen 2 is a shear fracture and has smaller
aperture values across the fracture than other specimens,
both before and after testing (Fig. 10b). Significant surface
damage in form of a reduced aperture across the fracture is
apparent for specimen 4, however (Fig. 10c). Observed
magnitudes in transmissivity changes during cyclic loading
could therefore potentially be strongly linked to the specific
fracture and corresponding aperture field configura-
tions (Fig. 13a, b).
Transmissivity changes during load cycles are not as
pronounced for shear offset specimens. However, shear
offset specimens experience large transmissivity changes
between cycles. For shear offset fractures, the aperture
fields with small regions of low aperture suggest larger
contact stresses between the fracture surfaces (Fig. 10d, e).
While small transmissivity changes during each cycle
could stem from relatively open aperture fields (Fig. 10d,
e), the more pronounced changes between cycles could
Table 3 Correlation length in direction of the specimen axis (X) and
normal to axial direction (Y) before and after testing for fracture
surfaces A/B of each specimen, respectively
Test Pre Pre Post Post
Specimen XYXY
(–) (mm) (mm) (mm) (mm)
124;19 þ;þ27;28 þ;þ
221;20 14;14 25;24 12;12
328;26 13;þ27;28 12;12
416;18 08;10 17;21 10;11
520;21 10;10 20;21 09;11
625;23 15;þ26;25 13;þ
707;07 13;08 15;20 þ;þ
808;08 09;þ09;24 09;þ
925;26 þ;þ27;19 þ;þ
10 þ;þ14;12 15;þþ;þ
11 22;20 þ;þ25;20 þ;þ
12 16;18 08;10 17;21 10;11
Correlation lengths that did not converge over half of the total length
in x- and y-direction are marked with a þ
Pre-Testing Post-Testing
Test 1
Test 2
Tes t 4
Tes t 7
Test 10
(a)
(b)
(c)
(d)
(e)
(f)
Fig. 10 Aperture fields as derived from surface scans (Figs. 8,9).
Specimens from tests 1, 2, 4, 7 and 10 (top to bottom) with apertures before
(left)andafter(right) testing. Reference figure for the aperture field
dimensions and aperture size colorbar is shown in f(color figure online)
D. Vogler et al.
123

result from larger contact point stresses that could result in
asperity failure and wear product removal.
The transmissivity for mated and offset fractures is
always larger after the first cycle than after subsequent
cycles (Fig. 13b), and the general trend is a decrease in
transmissivity from the first cycle (smallest circle) to the
last cycle (largest cicle). The finding of no significant
increase in transmissivity after initial shear displacements
(i.e., 1 mm, see Fig. 13) is congruent with previous find-
ings by Kim and Inoue (2003) and Esaki et al. (1999).
In most cases, the transmissivity decrease after the first
cycle is considerably larger then decreases for later cycles,
which agrees with the previous literature on fracture clo-
sure and hydraulic aperture changes during repeated load-
ing (Witherspoon et al. 1980; Gentier et al. 2013). This
illustrates the importance of reservoir history on expected
reservoir performance. Repeated fluid injection under high
pressures leads to fracture opening, which can potentially
cause a redistribution of gouge material along the main
flow paths. Lowering of injection pressure may cause
asperities and gouge material to be further comminuted,
which could close flow pathways.
4.2 Aperture Changes
Changes in mechanical and hydraulic aperture from the
initial values at 1 MPa confining stress are shown in
Fig. 5a–d, while changes in mechanical aperture versus
effective confining stress are shown in Fig. 6a–d. Note that
here one marker symbol (.and /for increasing and
decreasing confining pressure, respectively) represents an
increase in confining stress of 1 MPa (up to 10 MPa con-
fining stress) or 2.5 MPa (between 10 and 68 MPa con-
fining stress), respectively. Generally, changes in amech are
more drastic than those for ahyd for all tests. Similar to the
changes in transmissivity (Figs. 4a–e, 6a–d), aperture
changes display hysteretic behavior. Generally, changes in
mechanical and hydraulic aperture are larger during
decreasing effective confining stress (Figs. 5a–d, 6a–d).
Mechanical and hydraulic aperture changes in tests 1, 7
and 10 can be categorized similarly (Fig. 5a, c, d). For low
confining stresses, mechanical aperture changes increase
strongly (Figs. 6a–d, 5a–d) until aperture changes (equiv-
alent to aperture closure) increase monotonically (Fig. 6a,
c, d) or converge against a constant value (Fig. 6b).
During initial mechanical aperture changes, the
hydraulic aperture remains relatively unaffected, but
changes strongly once initial mechanical aperture change
has occurred (Fig. 5a–d). This change in behavior occurs
between mechanical aperture changes between 0.15 and
0.25 mm (test 1), 0.5 and 0.6 mm (test 7) and 0.1 and
0.2 mm (test 10) amech . The sudden decrease of ahyd could
be caused by the transition from uniform flow through most
of the fracture to channelized flow. Hydraulic aperture
changes become smaller after this increase in ahyd change
despite mechanical aperture changes increasing further
with effective confining stress (Fig. 6a, c, d). This behavior
can be explained by fluid flow confined to individual flow
channels, which are not closed despite further increase in
mechanical aperture changes. Once the maximum confin-
ing stress is reached and confining pressure is lowered
again, the mechanical aperture amech change reverts slower
than during initial loading (Fig. 6a–d). However, when
comparing changes in amech to ahyd (Fig. 5a, c, d), the
mechanical aperture change recovers much faster than the
hydraulic aperture change after peak loading. With fluid
flow potentially concentrated within single channels,
increasing amech (i.e., decreasing amech changes) would not
lead to a redistribution of fluid flow since the contact area
of the fracture surface is where the initial comminution of
asperities and the resulting formation of gouge material
would be located. Once amech opens sufficiently for the
failed asperties to be flushed from the system, ahyd
decreases significantly between a change in the mechanical
aperture of 0.2 and 0.4 mm. Mechanical aperture changes
do not decrease more rapidly during that regime (Fig. 6a–
d), since the gouge material is only flushed from the system
if compressive stresses on the gouge material are low and
therefore do not significantly contribute to the continued
propping of fractures. The rapid decrease in flow channels
for confining stress increases up to 10 MPa is consistent
with prior observations by Gentier et al. (2013).
4.3 Analysis of Gouge Material
Grain size distribution of the comminution products
(Fig. 7) organizes in two different families with a more
pronounced log-normal distribution for tests 5, 11 and 12
than for tests 6, 7, 8, 9 and 10. Tests 5, 11 and 12 have
small shear offsets (2, 1 and 1 mm, respectively), making
the breakout of monominerals much more likely than in the
other specimens, where large shear offsets lead to isolated
contact points that can cause more rupturing of asperities
during each loading cycle. With ongoing surface damage,
the number of contact points may be increased until the
local contact stress is then insufficient to overcome the
asperity strength. This hypothesis is consistent with the
analysis of mechanical and hydraulic aperture changes in
Sect. 4.2.
4.4 Surface Scans
Figures 8and 9compare the surfaces before and after
testing and show strong alterations in the surfaces, but not
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

decreasing asperity height during testing for all speci-
mens (Fig. 14). Due to the small length scale of the frac-
tures under investigation, the breaking of individual grains
and asperities can lead to fractures that fill with gouge
material, but do not develop better correlation between the
two sides than before testing. This phenomenon is influ-
enced by the displacement of the two fracture surfaces
against each other. Displacement offsets of a few mil-
limeters may lead to more point loads than in mated
specimens (Fig. 10a–e), which can cause higher stresses in
grains and asperities that can lead to failure, as mentioned
above. The relationship between the decrease in contact
area upon displacement can be related to the correlation
length in the x-direction (Table 3), which influences the
dilatancy of the fracture surfaces upon shear displacement.
The large correlation lengths indicated in Table 3could
suggest that mechanical apertures should continually
increase even for large shear displacements. However,
most specimens show a strong increase of the correlation
function H in the first few millimeters of specimen offset
(e.g., Fig. 12), with convergence only occurring for large
correlation lengths beyond the specimen scale. This indi-
cates that the correlation decreases strongly during the first
few millimeters of shear displacement and has no large
effect for displacements larger than 5 or 10 mm as the
correlation length has not converged yet. While the cor-
relation length can be used as an indicator after which shear
displacement of an offset specimen will not see an aperture
increase anymore, it cannot be used to estimate the amount
of contact area for a given displacement, which is what
ultimately effects contact stresses between the two speci-
men sides. The contact area plays a very significant role
since this determines the likelihood of surface damage in
(a)
Asperity Height [mm]
0246
CDF [-]
0
0.2
0.4
0.6
0.8
1
Asperity Height [mm]
0246
CDF [-]
0
0.2
0.4
0.6
0.8
1
(b)
Normalized Asperity Height [-]
0 0.5 1
CDF [-]
0
0.2
0.4
0.6
0.8
1
Normalized Asperity Height [-]
0 0.5 1
CDF [-]
0
0.2
0.4
0.6
0.8
1
Fig. 11 CDF of asperity height (a) and asperity height normalized to
the maximum asperity height (b) for specimens 1–12 for fracture
sides A and B before (left) and after (right) testing
r [mm]
0 5 10 15 20 25 30
H(z) [mm2]
0.2
0.4
0.6
0.8
1
1.2
r [mm]
0 5 10 15
H(z) [mm2]
0.2
0.4
0.6
0.8
1
1.2
Fig. 12 Example of the correlation length function in x- (left) and
y-direction (right) for specimen surface side A of test 6
Fig. 13 Transmissivity values for all normal loads versus specimen
offset for all specimens. Shown is one marker for each load cycle at:
amaximum effective confining stress and b1 MPa confining stress at
the end of each load cycle. Cyclic loading progresses from the first
cycle (smallest circle) to the last cycle (largest circle)
D. Vogler et al.
123

the fracture for a given normal load, which can counteract
changes in mechanical aperture caused by an increase in
offset of a millimeter or two more. Therefore, we deduce
that the correlation length is only a suitable metric to
determine transmissivity in fractures for the first few mil-
limeters of shear offset, with the amount of contact area
being more significant for defining the important role of
fracture damage under high normal loads.
Figure 11 shows a wider distribution of asperity heights
after testing when comparing the values for the CDF at 0.5
and 1.0. While all but two fracture surfaces in Fig. 11a
(left) have a maximum asperity height of 2 mm for the
midpoint of the CDF before testing, more than half of all
specimens surpasses this value for fracture surfaces after
testing. Asperity height distributions across specimens
show (Fig. 11) that maximum asperity height differences
become more pronounced after testing, showing larger
maximum values for asperity height and a weaker resem-
blance of the CDF between individual specimens. The
effect of increased local normal stresses observed in
Fig. 11a and b (e.g., Fig. 9c–e with strong surface changes)
shows that repetitive loading does not necessarily lead to
smoothed out asperity distributions and surfaces. While a
few specimen surfaces became more homogeneous, it was
observed on tests on offset specimens that cracks formed as
a consequence of normal loading and the associated sur-
faces were fairly rough. Such cracks are shown in Fig. 9c
(on surface B) where a crack parallel to the y-direction is
visible at an xvalue of around 20 mm and in Fig. 9e (on
surface A) where a crack parallel to the y-direction is
visible at an xvalue of around 10 mm. Mineral inclusions
of quartz and felspar in granodiorite have a high strength,
which makes crack propagation more likely to occur at the
boundaries with lower strength in between grains. Grain
boundaries and grain size will then dominate the effect of
surface deformation after increases in confining stress.
Iwano and Einstein found evidence that asperity distribu-
tion in tensile fractures strongly depends on the overall
grain size (Iwano and Einstein, 1993). Mineral grain sizes
of intact core material were estimated to be between 2 and
7 mm, which can explain the large asperity heights after
testing (Fig. 11b).
These findings indicate the influence of surface damage
on fracture transmissivity, especially for fractures that
experienced shear offset. Gouge production induced by
surface damage can strongly counteract the transmissivity
increase produced by shear offset (Figs.12,13,14).
5 Conclusions
Granodiorite specimens with natural tensile and shear
fractures were subjected to cyclic loading between 1 and
68 MPa confining pressure. Constant fluid flow rates
through the specimen were established, which made it
possible to measure the fluid pressure response to con-
fining pressure. A total of 12 tests were performed, with
four specimens tested in a mated configuration and eight
specimens tested with shear displacement between 1 and
6 mm. Fracture surfaces were scanned before and after
testing to give insight into surface deformation during
testing. The gouge material produced by asperity damage
was collected to study the impact of a gouge layer on
transmissivity.
While all specimens showed a decrease in transmissivity
with increased confining pressure, transmissivity decrease
in mated and offset specimens shows fundamentally dif-
ferent behavior. Mated specimens only show strong
decrease during individual load cycles for small confining
pressures, suggesting fluid flow confined to channels for
high confining pressures. Offset specimens show signifi-
cant transmissivity decrease between load cycles and sur-
face damage which leads to gouge production.
Permeability generally decreased with ongoing loading
cycles, indicating nonelastic deformation of the fracture
surfaces. Specimens 2, 4, 7 and 10 showed hysteretic
effects during individual loading cycles, with lower per-
meabilities during unloading of the specimen for respective
effective confining pressures. During testing, the normal
loads were large enough to keep most gouge material in
place, which led to increased fluid pressures to sustain
constant flow rates. It is hypothesized that breaking
asperities, which are only flushed out of the system during
decreasing normal loads, contribute strongly to the hys-
teretic transmissivity behavior observed during tests.
sample [-]
123456789101112
asperity height STD [mm]
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Pre (A)
Pre (B)
Post (A)
Post (B)
Fig. 14 Comparison of the standard deviation of asperity height on
the surfaces A and B before (red marker) and after testing (blue
marker) (color figure online)
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123

Mechanical and hydraulic aperture changes are compared,
showing more pronounced changes in the mechanical aper-
ture for all tests. Mechanical apertures change most drasti-
cally for low confining stresses. The most pronounced
changes in hydraulic apertures occur after initial mechanical
aperture closure can be observed. This could be explained by
mechanical aperture closure causing fluid flow displacement
to individual flow channels. This behavior changes again for
high confining pressures, when mechanical aperture changes
(i.e., closes) further while the hydraulic aperture does not
change significantly. These nonlinear changes in the rela-
tionship of mechanical and hydraulic aperture changes can be
attributed to increased surface damage and fracture closure
for high confining pressures, while fluid flow is already
confined to channel flow and is not strongly affected by
compression of the fracture.
This work illustrates the importance of fatigue behavior,
which is especially crucial for EGS with reservoir opera-
tion times of multiple decades. During operation of a
reservoir, interruption of injection wells can lead to cyclic
lowering and raising of the effective normal stress that
fractures are subjected to. On the laboratory scale, this
study shows differences of up to three orders of magnitude
of the transmissivity, which suggests that the history of the
fracture surfaces in a reservoir plays a significant role in the
evolution of fracture transmissivity after initial stimulation.
Acknowledgments The authors want to thank two anonymous
reviewers for their constructive suggestions which helped to improve
this work. The authors further want to thank the National Cooperative
for the Disposal of Radioactive Waste (Nagra), Switzerland, and the
CRIEPI fractured rock study Takana et al. (2014) for providing us
with the specimen material for our study. The authors further want to
thank the chair of geosensors and engineering geodesy at ETH Zurich
for their support with the photogrammetry scanner. This work was
partially supported by the GEOTHERM II project, which is funded by
the Competence Center Environment and Sustainability of the ETH
Domain. This project benefitted from partial funding from DOE DE-
FE0023354.
References
Barton N, Bandis S, Bakhtar K (1985) Strength, deformation and
conductivity coupling of rock joints. Int J Rock Mech Min Sci
22(3):121–140
Brown SR (1987) Fluid-flow through rock joints—the effect of
surface-roughness. J Geophys Res-Solid Earth Planets
92(B2):1337–1347
Chen Z, Narayan S, Yang Z, Rahman S (2000) An experimental
investigation of hydraulic behaviour of fractures and joints in
granitic rock. Int J Rock Mech Min Sci 37(7):1061–1071
Cook NGW (1992) Natural joints in rock*—mechanical, hydraulic
and seismic behavior and properties under normal stress. Int J
Rock Mech Min Sci Geomech Abstr 29(3):198–223
Durham WB, Bonner BP (1994) Self-propping and fluid-flow in
slightly offset joints at high effective pressures. J Geophys Res-
Solid Earth 99(B5):9391–9399
Esaki T, Hojo H, Kimura T, Kameda N, Deut Gesell E, Grundbau
(1991) Shear-flow coupling test on rock joints. In: Proceedings—
seventh international congress on rock mechanics, vol 1. Rock
mechanics and environmental protection
Esaki T, Du S, Mitani Y, Ikusada K, Jing L (1999) Development of a
shear-flow test apparatus and determination of coupled proper-
ties for a single rock joint. Int J Rock Mech Min Sci
36(5):641–650
Evans KF (2005) Permeability creation and damage due to massive
fluid injections into granite at 3.5 km at soultz: 2. critical stress
and fracture strength. J Geophys Res Solid Earth 110(B4):1–14
Gentier S, Hopkins D, Riss J (2013) Role of fracture geometry in the
evolution of flow paths under stress. In: Faybishenko B,
Witherspoon PA, Benson SM (eds) Dynamics of fluids in
fractured rock, American Geophysical Union, pp 169–184.
doi:10.1029/GM122p0169
Hakami E, Larsson E (1996) Aperture measurements and flow
experiments on a single natural fracture. Int J Rock Mech Min
Sci Geomech Abstr 33(4):395–404
Hakami, E and Einstein, H H and Gentier, S and Iwano, M (1993)
Characterisation of fracture apertures-Methods and parameters.
8th ISRM congress 1995
Iwano M, Einstein H (1993) Stochastic analysis of surface roughness,
aperture and flow in a single fracture. In: Proceedings of the
ISRM International Symposium Eurock ’93. Lisbon.
pp 135–1441
Kim HM, Inoue J (2003) Analytical approach for anisotropic
permeability through a single rough rock joint under shear
deformation. J Geophys Res Solid Earth 108(8):5-1–5-10
Lee HS, Cho TF (2002) Hydraulic characteristics of rough fractures in
linear flow under normal and shear load. Rock Mech Rock Eng
35(4):299–318. doi:10.1007/s00603-002-0028-y
Li B, Jiang Y, Koyama T, Jing L, Tanabashi Y (2008) Experimental
study of the hydro-mechanical behavior of rock joints using a
parallel-plate model containing contact areas and artificial
fractures. Int J Rock Mech Min Sci 45(3):362–375
Li Y, Chen Y, Zhou C (2014) Hydraulic properties of partially
saturated rock fractures subjected to mechanical loading. Eng
Geol 179:24–31
Louis C (1969) A study of groundwater flow in jointed rock and its
influence on the stability of rock masses. Rock mechanics
research, report 10
McClure M, Horne R (2014) Characterizing hydraulic fracturing with
a tendency-for-shear-stimulation test. Spe Reserv Eval Eng
17(2):233–243
Oron AP, Berkowitz B (1998) Flow in rock fractures: the local cubic
law assumption reexamined. Water Resour Res
34(11):2811–2825. doi:10.1029/98wr02285
Park H, Osada M, Matsushita T, Takahashi M, Ito K (2013)
Development of coupled shear-flow-visualization apparatus and
data analysis. Int J Rock Mech Min Sci 63:72–81. doi:10.1016/j.
ijrmms.2013.06.003
Pyrak-Nolte LJ, Morris JP (2000) Single fractures under normal
stress: the relation between fracture specific stiffness and fluid
flow. Int J Rock Mech Min Sci 37(1–2):245–262. doi:10.1016/
s1365-1609(99)00104-5
Raven KG, Gale JE (1985) Water-flow in a natural rock fracture as a
function of stress and sample-size. Int J Rock Mech Min Sci
22(4):251–261
Renshaw CE (1995) On the relationship between mechanical and
hydraulic apertures in rough-walled fractures. J Geophys Res-
Solid Earth 100(B12):24,629–24,636. doi:10.1029/95jb02159
Rutqvist J, Stephansson O (2003) The role of hydromechanical
coupling in fractured rock engineering. Hydrogeol J 11(1):7–40
Shugang W, Elsworth D, Jishan L (2013) Mechanical behavior of
methane infiltrated coal: the roles of gas desorption, stress level
D. Vogler et al.
123

and loading rate. Rock Mech Rock Eng 46(5):945–958. doi:10.
1007/s00603-012-0324-0
Takana Y, Miyakawa K, Fukahori D, Kiho K, Goto K (2014) Survey
of flow channels in rock mass fractures by resin injection. Asian
rock mechanics symposium 8
Tester JW, Anderson BJ, Batchelor AS, Blackwell DD, DiPippo R,
Drake EM, Garnish J, Livesay B, Moore MC, Nichols K, Petty S,
Tokso
¨z MN, Veatch RW, Baria R, Augustine C, Murphy E,
Negraru P, Richards M (2006) The Future of geothermal energy;
impact of enhanced geothermal systems (EGS) on the United
States in the 21st century. Technical report INL/EXT-06-11746,
Idaho National Laboratory
Walsh JB (2003) A theoretical analysis of sliding of rough surfaces.
J Geophys Res-Solid Earth 108(B8). doi:10.1029/2002jb002127
Wang S, Elsworth D, Liu J (2011) Permeability evolution in fractured
coal: the roles of fracture geometry and water-content. Int J Coal
Geol 87(1):13–25
Witherspoon PA, Wang JSY, Iwai K, Gale JE (1980) Validity of
cubic law for fluid-flow in a deformable rock fracture. Water
Resour Res 16(6):1016–1024
Xiong X, Li B, Jiang Y, Koyama T, Zhang C (2011) Experimental
and numerical study of the geometrical and hydraulic
characteristics of a single rock fracture during shear. Int J Rock
Mech Min Sci 48(8):1292–1302
Yeo IW, De Freitas MH, Zimmerman RW (1998) Effect of shear
displacement on the aperture and permeability of a rock fracture.
Int J Rock Mech Min Sci 35(8):1051–1070
Zhong Z, Elsworth D, Hu Y (2014) Evolution of strength and
permeability in stressed fractures with fluid-rock interactions.
Pure Appl Geophys 173(2):525–536. doi:10.1007/s00024-015-
1099-5
Zhu WC, Liu J, Elsworth D, Polak A, Grader A, Sheng JC, Liu JX
(2007) Tracer transport in a fractured chalk: X-ray ct character-
ization and digital-image-based (dib) simulation. Transport
Porous Media 70(1):25–42
Zimmerman RW, Kumar S, Bodvarsson GS (1991) Lubrication
theory analysis of the permeability of rough-walled fractures. Int
J Rock Mech Min Sci Geomech Abstr 28(4):325–331
Zimmerman RW, Bodvarsson GS (1996) Hydraulic conductivity of
rock fractures. Transport Porous Media 23(1):1–30
Permeability Evolution in Natural Fractures Subject to Cyclic Loading and Gouge Formation
123