The fluid flow properties of fault

damage zones

Thesis submitted in accordance with the requirements of the

University of Liverpool for the degree of Doctor of Philosophy by

Thomas Matthew Mitchell

July 2007

1
Dedicated to my Mum and Dad

2
Contents
CONTENTS................................................................................................................................... 3
ABSTRACT................................................................................................................................... 5
ACKNOWLEDGEMENTS.......................................................................................................... 7
1 INTRODUCTION ............................................................................................................... 9
1.1 NATURE OF THE PROBLEM ............................................................................................ 9
1.1.1 Permeability structure of faults and the importance of the damage zone ............. 10
1.1.2 Field studies of fault zone structure ...................................................................... 12
1.1.3 Laboratory studies of fault zone permeability ....................................................... 13
1.2 AIMS OF THE THESIS.................................................................................................... 14
1.3 THESIS STRUCTURE ..................................................................................................... 15
2 ON THE NATURE, DISTRIBUTION AND EVOLUTION OF DAMAGE
SURROUNDING STRIKE SLIP FAULT ZONES: A FIELD STUDY OF THE ATACAMA
FAULT ZONE, NORTHERN CHILE ...................................................................................... 17
2.1 ABSTRACT .................................................................................................................. 17
2.2 INTRODUCTION ........................................................................................................... 19
2.3 THE ATACAMA FAULT ZONE ...................................................................................... 23
2.3.1 General geology .................................................................................................... 23
2.3.2 Lithology and petrology ........................................................................................ 27
2.4 METHODS ................................................................................................................... 29
2.4.1 Macrofractures...................................................................................................... 29
2.4.2 Microfractures....................................................................................................... 30
2.5 RESULTS ..................................................................................................................... 33
2.5.1 General structure of faults studied ........................................................................ 33
2.5.1.1 Caleta Coloso Fault................................................................................................... 33
2.5.1.2 Cristales Fault ........................................................................................................... 39
2.5.1.3 Blanca Fault.............................................................................................................. 40
2.5.1.4 Small scale faults ...................................................................................................... 42
2.5.2 Damage zone macrofractures................................................................................ 43
2.5.3 Damage zone microfractures ................................................................................ 47
2.5.4 Damage zone microfracture and macrofracturing compared ............................... 53
2.5.5 Fault core zone...................................................................................................... 54
2.6 DISCUSSION ................................................................................................................ 56
2.6.1 Current models on the origin of off-fault damage ................................................. 56
2.6.2 Fault structure, scaling relationships and applicable models............................... 62
2.6.2.1 Fault damage zones................................................................................................... 62
2.6.2.2 Fault core zones ........................................................................................................ 71
2.7 CONCLUSIONS ............................................................................................................. 74
3 EXPERIMENTAL DESIGN AND THE MEASUREMENT OF PERMEABILITY .. 76
3.1 INTRODUCTION ........................................................................................................... 76
3.1.1 Basic design........................................................................................................... 76
3.2 EQUIPMENT SET-UP AND DESIGN ................................................................................. 79
3.2.1 Pressure vessel and sample assembly ................................................................... 79
3.2.2 Force gauge block (and force gauge).................................................................... 85
3.2.3 Axial loading system.............................................................................................. 90
3.2.4 Pore fluid controller system .................................................................................. 95
3.2.5 Confining pressure system..................................................................................... 99
3.2.6 Transducers, data logging system and computer control.................................... 100
3.2.7 Preparation of samples for testing ...................................................................... 102
3.2.8 Summary.............................................................................................................. 103
3.3 SUMMARY OF POTENTIAL SOURCES OF ERROR AND CALIBRATION ............................ 104
3.3.1 Calibrations......................................................................................................... 106
3.3.1.1 Confining pressure and pore fluid pressure............................................................. 106
3.3.1.2 Downstream storage capacity ................................................................................. 106

3
 3.3.1.3 Volumometer LVDT............................................................................................... 108
3.3.1.4 Force gauge calibration........................................................................................... 109
3.3.1.5 Elasticity of the loading column ............................................................................. 111
3.3.1.6 Axial displacement transducer calibration .............................................................. 113
3.3.1.7 Load on the force gauge from confining pressure................................................... 113
3.3.1.8 Thermal expansion effects on the force gauge........................................................ 114
3.4 THE MEASUREMENT OF PERMEABILITY ..................................................................... 116
3.4.1 The diffusion equation for the flow of compressible fluids through porous media
116
3.4.2 Methods for measuring permeability in low permeability rocks ......................... 120
3.4.2.1 The pore pressure oscillation technique.................................................................. 121
3.4.2.1.1 Fischer and Krantz solution............................................................................... 121
3.4.2.1.2 Bernabe’s solution ............................................................................................. 128
3.4.2.2 Processing techniques ............................................................................................. 130
3.4.2.3 Analysis of errors and minimization ....................................................................... 133
4 EXPERIMENTAL MEASUREMENTS OF PERMEABILITY EVOLUTION
DURING TRIAXIAL COMPRESSION OF INITIALLY INTACT CRYSTALLINE
ROCKS AND IMPLICATIONS FOR FLUID FLOW IN FAULT ZONES........................ 135
4.1 ABSTRACT ................................................................................................................ 135
4.2 INTRODUCTION ......................................................................................................... 136
4.3 EXPERIMENTAL TECHNIQUE ..................................................................................... 139
4.4 RESULTS ................................................................................................................... 144
4.4.1 Cerro Cristales Granodiorite .............................................................................. 145
4.4.2 Westerly Granite ................................................................................................. 150
4.4.3 Increasing amplitude cyclic loading test on Westerly Granite............................ 154
4.5 DISCUSSION .............................................................................................................. 157
4.5.1 Strength and permeability evolution of Cerro Cristales granodiorite and Westerly
granite. 157
4.5.2 Permeability evolution from induced cyclic damage........................................... 160
4.5.3 Crack network controlling permeability.............................................................. 164
4.5.4 Fault healing and permeability recovery ............................................................ 165
4.5.5 Implications for crustal fluid flow and fault zone hydraulics.............................. 168
4.6 CONCLUSIONS ........................................................................................................... 173
5 DISCUSSION AND FURTHER WORK....................................................................... 175
6 APPENDIX A – EXPERIMENTAL CALIBRATIONS............................................... 179
7 APPENDIX B - PAPERS PUBLISHED AND IN-PRESS............................................ 182
7.1 SLIP ON ‘WEAK’ FAULTS BY THE ROTATION OF REGIONAL STRESS IN THE FRACTURE
DAMAGE ZONE. NATURE, 2006............................................................................................... 183
7.2 ON THE STRUCTURE AND MECHANICAL PROPERTIES OF LARGE STRIKE-SLIP FAULTS.
GEOL. SOC. SPECIAL PUBLICATION, IN PRESS. ......................................................................... 187
8 REFERENCES ................................................................................................................ 205

4
Abstract

Quantification of the fluid flow properties of the Earth’s crust is an essential

precursor to the understanding of a wide range of geological processes, including
earthquake generation and crustal strength, and the recovery of natural resources.
Faults play a key role in the migration of fluids around the Earth’s crust, and
therefore the fluid flow properties of fractured rocks and how these properties
evolve with time are of major importance. This thesis aims to improve our
understanding of the hydraulic transport properties of large fault zones by
presenting a large dataset of detailed field and microstructural observations and
results from a suite of laboratory experiments to provide a basis for studying the
distribution, and fluid flow properties, of damage surrounding large natural fault
zones.

Damage surrounding the core of faults is represented by both microfracturing of

the rock matrix and by macroscopic fracture networks. Microfracture and
macrofracture densities and orientations have been analysed on strike slip faults
with displacements ranging over 3 orders of magnitude (~0.12 m – 5000 m).
These faults cut crystalline rock within the excellently exposed Atacama Fault
Zone, Northern Chile. All faults consist of a fault core and associated damage
zone. Damage zone width as defined by macrofractures and microfractures scale
with displacement and fault length. Both microfractures (specifically fluid
inclusion planes) and macrofractures within the damage zone show a log-linear
decrease in fracture density with perpendicular distance from the fault core. An
empirical equation for microfracture density distribution based on the evolution
of displacement has been derived for these faults. Preferred microfracture
orientations in the damage zone suggest that this damage may predominantly be
due to early processes related to enhanced stress at fault tips, in addition to
cumulative wear processes from the juxtaposition of geometrical irregularities on
the fault plane and damage from dynamic rupture. Fault core widths scale with
displacement, with the largest displacement fault showing a wide multiple core
zone.

Detailed experimental studies of the development of permeability of crustal rock

during deformation are essential in helping to understand fault mechanics and
constrain larger scale models that predict bulk fluid flow within the crust. The
strength, permeability and pore fluid volume evolution of initially intact
crystalline rock under increasing differential load leading to macroscopic failure
has been determined at water pore pressures of 50 MPa and varying effective
pressures from 10 to 50 MPa. Permeability is seen to increase by, up to, and
over two orders of magnitude prior to macroscopic failure, with the greatest
increase seen at lowest effective pressures. Post-failure permeability is shown to
be over three orders of magnitude higher than initial intact permeabilities and
approaches the lower the limit of measurements of in situ bulk crustal
permeabilities. Increasing amplitude cyclic loading tests show permeability-
stress hysteresis with high permeabilities maintained as differential stress is
reduced and the greatest permeability increases are seen between 90-99% of the
failure stress. Under hydrothermal conditions without further loading, it is

5
suggested that much of this permeability can be recovered by healing and
sealing, and pre-macroscopic failure fracture damage may heal relatively faster
than post-failure macroscopic fractures. Pre-failure permeabilities are nearly
seven to nine orders of magnitude lower than that predicted by some high
pressure diffusive models suggesting that microfracture matrix flow cannot
dominate, and agrees with inferences that bulk fluid flow and dilatancy must be
dominated by larger scale structures, such as macrofractures. It is suggested that
the permeability of a highly stressed fault tip process zone in low-permeability
crystalline rocks could increase by more than 2 orders of magnitude, while stress
drops related to fracture propagation close damage zone cracks, and some
permeability is maintained due to hysteresis from permanent microfracture
damage.

Future work should aim to quantify experimentally-induced microfractures and

associated permeability measurements, and by relating the fracture densities
surrounding natural fault zones with densities seen in experimental deformed
samples with known permeabilities, modelling techniques can then be applied to
gain estimates of bulk fluid flow of the fracture networks. This will provide a
basis for predicting the influence of pore fluid pressures on important geological
issues, such as crustal strength.

6
Acknowledgements

Firstly I’d like to say a big thanks to Dan Faulkner for being a great supervisor
and equally good friend. In many ways you’ve inspired me over the years to
work hard and get to where I am. We’ve travelled to some great places, done
some enjoyable work and without your enthusiasm in the lab and field and
encouragement to present at conferences, meet new people in the scientific
community (and your ability to get grant money to fund it all) would have
resulted in an entirely different few years! Also a special thanks goes to Dave
Prior who first inspired me to love all things geo, and it was because of him I
chose to do a PhD in the first place.

I’d like to thank everyone in the Rockdef lab at Liverpool over the years,
especially Armo who has been a great friend and help. Thanks to Sergio, Nicola,
Shanvas, Healy (for providing junky and jovial antics), Mike, Fil and Alex who
have helped to provide me with a brilliant environment to work in. Probably the
best and most memorable aspect of my PhD that will stay with me on my journey
into uncharted academic territory is the general day to day lab-life. All jokes
aside, there is nothing more motivating than working with a group of people who
are all enjoying what they are doing. Whether it’s arguing over who last had the
spanner, having a heated scientific debate, dressing up as transvestites or
ducktaping ourselves to the wall for charity, taking the rig to bits for the 20th time
because some Spaniard has broken it again, spending two solid days trying to
solder high pressure piping into the sample assembly to find it still leaks and you
want to cry, working for a solid month with no sleep to churn out experiments to
meet deadlines, fighting over whose turn it is to make the tea or rubbing out yet
more perverted graffiti in your lab book – there’s always a new experience.
Thanks to all the Chile crew, namely Jose, Erik, Nico and Felipe whose help,
friendship and cracking desert barbeques made fieldwork a lot of fun. Thanks to
Ian, Dave, Mark and Clive in engineering for providing crucial help

Thanks to others in the department for providing top nights out, good Boswell
banter and fun field trips over the years, including Vijay, Sneedham, Sasha,
McNamara, Craig, Chris Hunt, Roy, Worden, Wheeler, Dave Hodgson, Betty,
Neil, Ros, Scalv, Laura, Becky, Pottsy, Dan, Seaton, Rosie, Goz, Ceri, and many
more. Thanks to Mo for putting up with my banter and abuse, and helping to
make my stay in the department over the years very enjoyable. Thanks to Kav
for always being really helpful and proactive. Thanks to Helen, Jeanette and
Alan Mac, for always having time to help. Thanks to Richard Holme for offering
financial support and general help beyond the call of duty.

Big thanks to Matt, Ste, Rob, Will, Frank, Jonny and Rachel for being brilliant
mates and housemates. Thanks to all my other good pals for general
entertainment, hilarity and wild nights on the town that have kept me sane over
the years, such as Roger, Farrell, McNamara, Jen, Cat, Miles, Kim, Claire, Mary,
Ciara, Aoife, Meg and many more. Thanks to Jude for being a great friend and
girlfriend through the early years, and being so understanding on the many trips

7
away abroad. Thanks to Chris Oliphant for daily countdown-till-submission text
message reminders in the last few weeks up until submission.

And finally, thanks to Jess and Alice for being top sisters and providing me with
years of Mitchell mayhem and fun. Thanks to Grandma for supporting me for
cash when I’ve needed it, and a massive thanks to Mum and Dad for being
superb throughout all of my university career. I owe you a lot. And not just
money!

If I’ve forgotten anyone else, get over it, I’m tired.

8
1 Introduction

1.1 Nature of the problem

Quantification of the fluid flow properties of the Earth’s crust is an essential

precursor to the understanding of a wide range of geological processes, including
earthquake generation and crustal strength, and the recovery of natural resources.
Many economically important hydrothermal mineral deposits and hydrocarbon
reservoirs are hosted within fault fracture networks. Faults play a key role in the
migration of fluids around the Earth’s crust, and therefore the fluid flow
properties of fractured rocks and how these properties evolve with time are of
major importance.

An example of the importance of fault zone permeability comes from recent

studies that have shown that in intracontinental regions an interconnected
network of faults and their associated high permeability damage zones help to
maintain hydrostatic pore pressures (Townend & Zoback 2000). This leads to
‘normal’ crustal strength according to Coulomb failure criterion

τ = C + μ (σ n − Pf ) 1.1

where: τ = shear strength (Pa)

C = cohesion (Pa)
μ = coefficient of friction
σn = normal stress (Pa)
Pf = pore fluid pressure (Pa),

and Byerlee’s generalized rule for rock friction, in which 0.6 ≥ μ ≥ 0.85 .

In contrast to intraplate regions, anomalously low crustal strength has been

inferred for some plate boundaries, such as the San Andreas fault in California,

9
from geophysical observations (Lachenbruch & Sass 1980, Zoback et al. 1987).
It has been proposed that this weakness (with μ ~ 0.2, assuming hydrostatic pore
pressures) is in fact due to excess pore pressures produced by the retardation of
fluid flow by low permeability material within the fault zone (Blanpied et al.
1992, Faulkner & Rutter 2001). These seemingly contradictory observations
regarding the fluid flow properties of faults found in intraplate and interplate
settings illustrate the need for a better understanding of the permeability of
crustal materials. The principal aim of the thesis is to quantify the field structure
of faults and to provide experimental measurements that quantify how the fault’s
fluid flow properties might vary within the zone.

1.1.1 Permeability structure of faults and the importance of the

damage zone

Sense of fault movement In the simplest terms, faults

typically consist of two
parts; a fault core consisting
of a relatively narrow
localised slip zone
containing fine-grained
fault gouge, surrounded by
the damage zone where

Retarded
intact rocks are pervasively
fluid flow
fractured (Figure 1.1).
<50cm ~100m Enhanced
Fault Damage Country fluid flow These can act as fluid
core zone rock
barriers and conduits
Figure 1.1 - Conceptual model for a typical strike slip respectively (Figure 1.1);
fault zone illustrating permeability anisotropy. The fault
core will consist of high strain products such as fault fluid flow is enhanced
gouge, cataclasites and breccia. Damage zone will consist
of small faults, fractures and veins. Country rock will be parallel to the fault in the
relatively undamaged rock containing regional structures.
relatively porous fractured
“damage” zone surrounding faults whereas the core of a fault is often composed
of very fine-grained, low permeability material that impedes lateral fluid flow
(Figure 1.1). These zones are then surrounded by undeformed host rock. In
order to quantify the bulk permeability anisotropy of large fault zones,

10
knowledge of the permeability of the various components of the fault zone is
needed, and also the spatial variation in these components. In addition, temporal
changes in fluid flow properties can occur, particularly if the fault zone is subject
to earthquake events.

The damage zone has been shown to be of critical importance from recent work
conducted on the dynamic fluid flow properties of fault gouge (Faulkner and
Rutter, unpublished data). These experiments show that fault gouge permeability
increases ~10% per decade increase in slip velocity. Hence the creation of
highly permeable pathways through the low permeability gouge zone even
during fault slip is unlikely. Consequently, any significant co-seismic
permeability changes in the fault zone must occur in the damage zone. Recent
work by (Miller et al. 2004) have suggested that earthquakes can initiate large
fluxes of fluid in the same type of ‘fault valving’ process first suggested by
(Sibson 1990). Miller et al. noted that the aftershocks following the 1997
Colfiorito earthquake in the Italian Apennines showed a diffusive-like spreading
of seismicity along the fault plane over several days. These aftershocks were
modelled as being triggered by a high pressure fluid front that migrated up the
fault plane from depth, initiated by damage associated with the mainshock. This
hypothesis is given credence by the fact that all the aftershock events occurred in
the region where Coulomb failure stress was reduced following the mainshock,
suggesting that they might have been triggered by fluid pressure changes rather
than stress transfer. In this model, however, the fluid flow characteristics of the
fault zone, as well as the physical dimensions of the damage zone, were tailored
to fit the observations and still await experimental and field verification.

To understand the fluid flow properties in fault zones, it is clear that a

combination of field and laboratory techniques is required. Laboratory studies
can quantify the permeability characteristics of the fault zone as a function of
damage, and field investigations are required to place the experimental results
into a framework of constrained fault zone architecture.

11
1.1.2 Field studies of fault zone structure

Understanding fault zone structure gives an invaluable insight into the pathways
and barriers for fluid flow. Despite this, there have been remarkably few detailed
quantitative studies of the structure of faults (more specifically the damage zone),
particularly those making direct comparisons between faults with displacements
ranging from very small to very large. This is due mainly to the difficulties in
identifying faults that have recently been exhumed and hence now expose fault
rocks that previously formed at seismogenic depths. Also, faults that are exposed
in arid environments are a necessary precursor for the preservation of fragile
rocks that are extremely susceptible to the effects of weathering (Faulkner et al.
in press). It is therefore not surprising that there are few studies of detailed fault
structure. Some exceptions include (Chester et al. 1993, Faulkner et al. 2003,
Faulkner et al. in press, Hoffman-Rothe et al. 2004, Scholz et al. 1993, Schulz &
Evans 2000, Vermilye & Scholz 1998). Such examples studied faults that are
relatively well-exposed and have various exhumation histories, and can be used
to make predictions of the structure of faulting at depth. The hyperarid Atacama
desert in Northern Chile used in this study provides excellent exposure in the
crustal scale Mesozoic Atacama fault system in the Coastal Cordillera, and
provides a perfect opportunity to conduct such a field study due the to the well
preserved fault structures.

Probably the most poorly constrained aspect of fault zone structure at present is
the nature and origin of the damage zone. The physical dimensions and intensity
of damage within this zone as a function of distance from the fault core is fairly
well documented quantitatively for individual faults from both field studies
(Anders & Wiltschko 1994, Faulkner et al. 2006, Scholz et al. 1993, Wilson et al.
2003) and experimental studies (Janssen et al. 2001, Moore & Lockner 1995,
Zang et al. 2000). However, it is less clear as to when, in a fault’s history, the
bulk of this damage was created and how damage zones vary as a function of slip
displacement on faults of ranging sizes.

12
1.1.3 Laboratory studies of fault zone permeability

Previous experimental measurement of the fluid flow properties of fault zone

rocks has been heavily biased towards fault gouge materials. These studies have
usually been performed in ‘triaxial’ type deformation apparatuses, where
hydraulic fluid provides the confining pressure (σ2 = σ3) and an additional load is
applied axially (σ1). Synthetic quartz-rich and granite-derived fault gouge has
been shown to have permeability of the order of 10-16 m2, with a permeability
anisotropy of up to 1 order of magnitude (Zhang & Tullis 1998). Clay-rich fault
gouges have much lower permeability (~10-21 m2) and can exhibit permeability
anisotropy of 3 orders of magnitude (Faulkner & Rutter 1998).

In contrast to studies on gouge, fractured damage zone rocks have received little
attention. In the fault damage zone contained within crystalline rocks, fluid flow
will be controlled by both microfracturing (fractures that can only be seen under
a microscope) of the rock matrix and by macroscopic (fractures that can be seen
with the naked eye) fracture networks. In the laboratory, sample size restrictions
constrain study to the fluid flow properties of the microfracture network.
However, modelling techniques can be applied to gain estimates of bulk fluid
flow of the macroscopic fracture networks, provided there is prior knowledge of
the fracture density and orientation pattern. Studies on the San Andreas fault
around Tejon Pass in California have shown that pervasive microfracture damage
can dominate close to the fault (Wilson et al. 2005). Consequently, measurement
of permeability and porosity evolution associated with this type of damage is
important.

Zoback and Byerlee (1975) measured the effect of microfracture dilatancy on the
permeability of Westerly granite. In their experiments, however, the deformation
was only taken to 80% of the failure stress, and due to limitations in the
experimental techniques, deformation had to be stopped while permeability was
measured which may have time-dependent effects. Natural damage zone
permeability was measured by Wibberley & Shimamoto (2003) for various
components of the Median Tectonic Line in Japan, and these suggest
permeability variations ranging over several orders of magnitude, much greater

13
than those recorded in the experiments of Zoback and Byerlee (1975) who
suggested increases up to 300%.

1.2 Aims of the thesis

This study aims to improve our understanding of the hydraulic transport

properties of large fault zones by presenting a large dataset of detailed field and
microstructural observations and results from a suite of laboratory experiments to
provide a basis for studying the distribution, and fluid flow properties, of damage
surrounding large natural fault zones. This will provide a basis for predicting the
influence of pore fluid pressures on important geological issues, such as crustal
strength.

With the use of a detailed field study within the excellently exposed and
passively exhumed Atacama Fault Zone, Northern Chile, this study evaluates the
physical dimensions of fault damage zones surrounding strike-slip faults with
varying displacements in low porosity crystalline material. It aims to ascertain
the nature of the deformation within the damage zone and assess how is it
partitioned between macroscopic fractures and microscopic fractures, how the
intensity of these damage types vary with distance from the fault and with fault
maturity. Furthermore, the study tests some of the current models for fracture
damage accumulation surrounding fault zones in light of this dataset.
Experimental measurements were made on the permeability and porosity
evolution with progressive brittle deformation of initially intact
quartzofeldspathic crystalline rocks up to failure as a proxy for varying degrees
of fracturing found within fault damage zones. This is used as a framework to
assess the permeability structure surrounding fractures and faults and the
implications for fault zone hydraulics.

To summarise, the thesis will specifically address the following:

14
 • Evaluate the typical physical dimensions of fault damage zones
surrounding large strike-slip faults, and how they vary with increasing
displacement.

• Ascertain the nature of the deformation within the damage zone and
assess how is it partitioned between macroscopic fractures and
microscopic fractures, and how the intensity of these damage types vary
with distance from the fault.

• Measure experimentally the pre-yield permeability and porosity evolution

with progressive brittle deformation of initially intact quartzofeldspathic
crystalline rocks as a proxy for varying degrees of fracturing found within
fault damage zones.

• Characterize the development of permeability anisotropy with

progressive deformation.

• Develop a typical model for the structure and fluid flow properties of
large strike-slip fault zones, and apply this model provide a framework to
assess the permeability structure surrounding fractures and faults and the
implications for fault zone hydraulics.

1.3 Thesis structure

After this introduction, the core of this thesis (chapters 2 and 4) are presented as
journal papers and this necessarily involves some repetition of material as key
concepts are re-introduced in places. There have been changes compared to the
versions prepared for submission to journal, in that the figures, section headers
and formulae have been renumbered, and for brevity all references have been
combined into a single section at the end of the thesis. Both of these chapters
have their own abstract, introduction, results, discussion and conclusions. The
length of these chapters is too long for journal publication, and some reduction in
both chapters will be necessary for final submission. However, for the purposes

15
of the thesis, the level of information presented in each chapter was thought
better than to split the information further, leading to an overall longer thesis.
After this introduction, the thesis structure is as follows:

• Chapter 2 presents data and discussion on the nature, distribution and

evolution of damage surrounding strike slip fault zones, based on a
detailed field study of the Atacama Fault zone in Northern Chile.

• Chapter 3 introduces the purpose built experimental apparatus and used

to retrieve the data in chapter 4, and discusses the techniques used for
measuring permeability. It is not in the form of a journal manuscript.

• Chapter 4 presents data and discussion on the experimental

measurements of permeability evolution made during triaxial
compression of initially intact crystalline rocks as a proxy for varying
degrees of fracturing found within fault zones.

• Chapter 5 presents a brief discussion relating all previously discussed

data and conclusions, and makes suggestions for further work.

• Chapter 6 / Appendix A presents the various calibrations that are

discussed in chapter 3.

• Chapter 7 / Appendix B presents two published manuscripts to which the

author contributed.

16
2 On the nature, distribution and evolution of
damage surrounding strike slip fault zones: a
field study of the Atacama Fault zone, Northern
Chile

2.1 Abstract

Damage surrounding the core of faults is represented by both microfracturing of

the rock matrix and by macroscopic fracture networks. The spatial distribution
and geometric characterization of damage at various scales can help to predict
fault growth processes, subsequent mechanics and bulk hydraulic properties of a
fault zone. Within the excellently exposed Atacama Fault Zone, Northern Chile
micro- and macroscale fracture densities and orientation surrounding strike-slip
faults with well-constrained displacements ranging over 3 orders of magnitude
(~0.12 m – 5000 m) have been analysed. Faults have been studied that cut
crystalline rock (granodiorite) and have been passively exhumed from 6-10 km.
All faults studied consist of a fault core and associated damage zone.
Macrofractures are predominantly shear fractures orientated at high angles to the
fault. Microfractures are a combination of open, healed, partially healed and
fluid inclusion planes (FIPs). FIPs show a log-linear decrease in fracture density
with perpendicular distance from the fault core, and appear to be the youngest set
of fractures. FIPs may have remained open through most of the faulting, but
healed before late stage fault duplex development and exhumation in the area.
Later microfractures do not show a clear relationship of microfracture density
with perpendicular distance from the fault core. Damage zone widths defined by
FIPs density scale linearly with fault displacement. All faults appear to have a
critical microfracture density independent of displacement. An empirical
equation for microfracture density distribution based on the evolution of
displacement has been derived for these faults. Preferred microfracture
orientations in a mode I orientation suggest that this damage may predominantly
be due to early fault initiation processes related to enhanced stress at fault tips, in

17
addition to cumulative wear processes from the juxtaposition of geometrical
irregularities on the fault plane and damage from dynamic rupture. The fault
core zones generally consist of a single zone of finer grained fault rock;
including variably deformed protocataclasites to ultracataclasites with localized
bands of fault gouge reflecting high shear strains and localization of slip, and
evidence of enhanced fluid flow from hydrothermal precipitation. Fault core
widths scale with displacement, with the largest fault showing a wide multiple
core zone

18
2.2 Introduction

Typically, the structure of large fault zones is envisaged as a single fault core
(Figure 2.1) containing fault gouge, surrounded by a damage zone (Aydin 1978,
Bruhn et al. 1994, Bruhn et al. 1990, Byerlee 1990, Caine et al. 1996, Chester &
Chester 2000, Chester et al. 1993, Chester & Logan 1986, Cowie & Scholz 1992,
Dengo 1982, Evans 1990, Evans & Chester 1995, Flinn 1977, 1979, Little 1995,
Mitra 1984, Scholz 1987, Scholz et al. 1993, Sibson 1977, 1986). More complex
fault geometries have also been recognized (Faulkner et al. 2003).

Figure 2.1 - A typical strike-slip fault zone structure in a

quartzofeldspathic country rock (After (Faulkner et al. 2003)). Other
fault types such as normal and reverse have similar fault structure

A fault core is usually a relatively narrow, localized slip zone which is

surrounded by a damage zone which is a transitional region of deformed rock.
Kim et al. (2004) describe the damage zone as the volume of deformed wall
rocks around a fault surface that results from the initiation, propagation,
interaction and build-up of slip along faults. These zones are then surrounded by

19
undeformed host rock that contains little or no deformation features associated to
the faulting. Understanding each component of fault structure is important as it
gives an invaluable insight into the pathways and barriers for fluid flow (Caine et
al. 1996, Knipe 1992, Shipton et al. 2002), as well as the mechanical and
seismological properties. Faults are a major control on fluid flow in the crust,
which in turn can control crustal strength (Townend & Zoback 2000) and trigger
earthquakes (Miller et al. 2004). ‘Impermeable’ fault core material can act as
fluid barriers and can trap fluids within faults zones generating pressures greater
than hydrostatic (Byerlee 1990); high pressure fluid sealing and weakening
models (Blanpied et al. 1992, Faulkner & Rutter 2001, Rice 1992) rely on
interconnected fault networks within core and damage zones. This study
concentrates on the structure of strike-slip faults, primarily because these were
the types of faults available in the field area.

The most poorly constrained aspect of fault zone structure at present is the nature
and origin of the damage zone. The physical dimensions and intensity of both
microscale damage (fractures that can only be seen under a microscope) and
macroscale (can be seen with the naked eye) damage within this zone as a
function of distance from the fault core is fairly well documented qualitatively
(e.g. Kim et al. 2004) and quantitatively for individual faults from both field
studies (Anders & Wiltschko 1994, Brock & Engelder 1977, Engvik et al. 2005,
Faulkner et al. 2006, Scholz et al. 1993, Shipton & Cowie 2001, Vermilye &
Scholz 1998, Wilson et al. 2003) and experimental studies (Janssen et al. 2001,
Moore & Lockner 1995, Zang et al. 2000). However, general relationships are
difficult to achieve as many of these studies compare faults that crosscut
dissimilar lithologies, have varying grain sizes and (with a few exceptions)
concentrate on the structural development of relatively small faults with a small
range of displacements. These are important variables as fault zone structure is
dictated by (a) the depth of faulting (b) the protolith (c) the fault displacement
and (d) the interaction with other faults and/or pre-existing structures (Faulkner
et al. in press). Also, it is less clear as to when in a fault’s history the bulk of this
damage was created and how damage zones vary as a function of slip
displacement and time on faults of ranging sizes. As a result, it is not fully

20
understood what the dominant processes producing off-fault fracture damage are
and how this damage is partitioned between micro and macrofracturing.

In general, mechanical models of faulting suggest that the overall fault structure
of large displacement zones is established early in a fault’s history (Chester et al.
2004, Chester et al. 2005, Evans J.P. 2000). For example, once faulting has
initiated, some models suggest that the bulk of the fracture damage is imported
immediately prior to fault formation in the ‘process zone’ surrounding the fault
tip (Scholz et al. 1993, Vermilye & Scholz 1998). Fault wear models attribute
the bulk of fault damage to the cumulative fracture damage that results from
continued slip on pre-existing fault surfaces (Scholz 1987, Wilson et al. 2003).
More recently work has focused on fracture damage that is created by the elastic
strain energy released during dynamic rupture in the form of dynamic slip pulses
which can cause considerable off-fault damage in the form of general off-fault
secondary failure or pulverised rock at shallower levels (Dor et al. 2006b, Rice et
al. 2005). It is likely that some or all of the processes described by these models
contribute to bulk fault zone fracture damage, with the signatures of more recent
damage overprinting evidence of earlier processes, which makes it less clear as
to how fracture damage is partitioned between each fault process. As fault
damage zones are represented by both microfracturing of the rock matrix and by
macroscopic fracture networks, the empirical characterization of fracture patterns
at various scales, as well as understanding the geometry and timing of the fault
zone fracture population help to identify such fault growth processes. To
construct bulk fluid flow fault zone models, at a fundamental level this requires
knowing the properties of individual fractures at all scales, and knowing the
structural properties of the fracture population. Recent work has shown that
microfracture damage in the damage zone can play an important role in stress
rotation surrounding faults (Faulkner et al. 2006) due to the variation in elastic
properties resulting from the increasing density of microfracture damage towards
the fault core. This highlights the importance of such quantitative studies at
various scales and understanding such fracture patterns surrounding fault zones.

This work first introduces the geological setting of the faults studied, and then
presents detailed fault zone descriptions and mapping that elucidates the internal

21
structures of faults with varying displacements within the excellently exposed
Atacama fault system in northern Chile, as a context to then assess the
characteristics and origins of fault damage zones. Comparisons of strike-slip
faults that cut crystalline rock (granodiorite) at various stages of maturity are
made by characterizing bulk fault structure and completing micro- and
macroscale fracture density profiles within the damage zones of faults with well-
constrained displacements ranging over 3 orders of magnitude (~0.12 m – 5000
m). Following a brief review of existing fault-damage models, the implications
of observations for fault growth and damage models are discussed for faults at
such depths.

22
2.3 The Atacama Fault Zone

There have been remarkably few detailed quantitative studies of the structure of
faults (more specifically the damage zone), particularly making direct
comparisons between faults with displacements ranging over several orders of
magnitude. This is due mainly to the difficulties in identifying faults that have
recently been passively exhumed and hence now expose fault rocks that
previously formed at seismogenic depths. Also, faults that are exposed in arid
environments are a necessary precursor for the preservation of fragile rocks that
are extremely susceptible to the effects of weathering (Faulkner et al. in press).
Another key factor is locating faults of various sizes that crosscut similar
lithologies to aid comparison between data. However, recent studies of fault
zones include Cembrano et al. (2005), Chester et al. (1993), Faulkner et al.
(2003), Faulkner et al. (in press), Hoffman-Rothe et al. (2004), Jefferies et al.
(2006), Scholz et al. (1993), Schulz & Evans (2000), Shimamoto et al. (2001),
Shipton & Cowie (2001), Vermilye & Scholz (1998) and Wibberley &
Shimamoto (2003). All of these examples provide different levels of
information; they study faults that are relatively well-exposed and have various
exhumation histories and can be used to make predictions of the structure of
faulting at depth. The hyperarid Atacama desert in Northern Chile used in this
study provides excellent exposure in the Mesozoic Atacama fault system in the
Coastal Cordillera, and provides a perfect opportunity to conduct such a field
study due the to superbly well preserved fault structures that have been passively
exhumed from depth.

2.3.1 General geology

The Atacama Fault System (AFS) is situated in the continental margin of the
South American plate, beneath which oceanic lithosphere has been subducted
since early Palaeozoic time (Brown et al. 1993, Mpodozis & Ramos 1990), and is
a very important structure within the forearc of the Central Andes (Cembrano et
al. 2005, Scheuber & Gonzalez 1999). It is an arc-parallel strike-slip structure
that accommodates some of the oblique convergence between the Nazca and

23
South American plates (Figure 2.2), and it extends for ca. 1000 km between
Iquique (21°S) and La Serena (30°S) (Brown et al. 1993, Cembrano et al. 2005,
Scheuber & Gonzalez 1999) within the Cordillera de la Costa of the Central
Andes. The large scale geometry of the AFS was formed during the late Jurassic
and Early Cretaceous where large brittle structures in excess of 60km in length
were formed by sinistral strike-slip movement (Cembrano et al. 2005). Some
faults (e.g. the Caleta Coloso fault) have undergone more recent movements,
although these are limited in extent, and are in response to large subduction zone
earthquakes associated with the offshore trench (González et al. 2003). These
movements are easily recognized in the field and expressed as very localized,
narrow fault planes that from small fault scarps at the surface. Some of the NS-
striking master faults and subsidiary NW striking splay faults are organized into
strike-slip duplexes that occur at various scales from regional to local (Cembrano
et al. 2005). The faults studied in this work are all substructures of such a
structure; the Caleta Coloso Duplex (Figure 2.2). Figure 2.3a shows a 3D
topographic image of the field area, and Figure 2.3b is the same but has the map
shown in Figure 2.2 overlaid onto it. The major faults and units can be clearly
seen by comparing between Figure 2.3a & b. The fault zone runs through
quartzo-feldspathic rock, including diorites, granodiorites, pegmatites, gabbros
and metamorphic basement.

24
 Caleta Coloso Fault
5km displ.

Blanca Fault
35m displ. B

Cristales Fault
220m disp.

Figure 2.2 - Regional geological map (with inset) show broad scale features of the Atacama
fault system. a) Atacama fault system (AFS) in the Coastal Cordillera of Northern Chile.
Locations of 3 largests faults are labelled in red; b) Geology, geometry and kinematics of the
sinistral strike-slip Coloso duplex, showing location of faults studied. J = Jurassic,; EC =
Early Cretaceous; M = Miocene; P-P = Plio-Pliestocene. Maps simplified from (Brown et al.
1993, Cembrano et al. 2005, Scheuber & Gonzalez 1999). B marks the location where
background macro and microfracture counts were measured.

25
 a

Figure 2.3 – a) A Google earth 3D topographic image of the field area looking
downwards in a North West direction. Note that geological units are marked by
variations in colour and the faults have a distinctive topographic expression. b) as a) but
with geological map from Figure 2.2 overlaid to show relationship between topographic
expression of faults and geological features.

26
2.3.2 Lithology and petrology

The rocks in this study are all part of the Cerro Cristales pluton, which was first
described by Uribe & Niemeyer (1984). The Cerro Cristales Pluton includes
diorites, granodiorites, granodioritic diorites with a magmatic foliation and
isotropic tonalites and granodiorites. The pluton can be seen in the Figure 2.2 as
an elongate unit running N-S towards the east of the duplex zone. It varies in
width from north to south by 2.5 km to 4.3 km respectively. Towards the west, it
has been intruded by the Bolfin metadiorites of Lower Jurassic age, and towards
the east it is crosscut by the largest fault in this study, the Caleta Coloso fault.
Various NW running subsidiary faults including the Jorgillo (Figure 2.2), Blanca
(Figure 2.2 and Figure 2.10) and Cristales (Figure 2.2 and Figure 2.12) traverse
the centre of the pluton. The rocks within the pluton are cut by two generations
of dykes; Lower Jurassic microdioritic dykes and late Jurassic to early
Cretaceous granitic dykes. Unconformable Cenozoic sedimentary deposits
obscure the pluton in places. This pluton has been studied in detail by González
(1990, 1996, 1999) and consist of isotropic tonalites, granodiorites and quartzo-
feldpathic diorites that are classified from the variable amounts of plagioclase,
quartz, orthoclase, biotite and amphibole. Figure 2.4 shows examples of the
granodiorite through which all faults cross-cut, with an average grain size of
0.7mm and around 30% quartz.

27
 a b

Plg
Qtz

Aph

Bt

Figure 2.4 - a) Pristine granodiorite in hand specimen. b) Thin section of granodiorite in

transmitted cross-polarized light. Average quartz grain size is 0.7 mm. Major phases are
labelled. Qtz = quartz, Plg = plagioclase feldspar, Aph = amphibole and Bt = biotite.

Geobarometery suggest pressures of crystallization of up to 3.6 Kbar (360 MPa),

that would indicate a depth of crystallization of the order of 13 km (González
1996, 1999). In certain areas a foliation with a general NS to NE direction,
defined by the preferential orientation of elongate crystals of plagioclase is
interpreted as resulting from magmatic flow (González 1990, 1996, 1999) can be
seen. These studies suggest that the plutonic intrusion was syntectonic with the
sinistral activity of some pre-existing shear zones and that the magma came from
a deep source with a relatively fast injection rate into a confined space. This was
dictated by intrusion through N-S shears that promoted progressive N-S plutonic
growth that gave rise to the progressive expansion of the pluton, and this flow in
turn created the general N-S foliation. Within the studies, care was taken to
ensured sampled rocks were isotropic.

28
2.4 Methods

In order to characterize the spatial variation of damage within the fault damage
zone, detailed fault transects and maps were completed running normal to the
fault trace (beginning at the core/damage zone boundary) of each fault. Two
separate transects were completed for each fault to gain an average fracture
density. For each transect (starting at the fault core moving outwards into the
damage zone), macroscopic fracture counts and fracture orientation
measurements were completed at sampling stations located every ~5m. At each
sampling location, an orientated sample was collected for thin-sectioning, so that
microfracture densities and orientations could be measured under the
microscope. This section now details the methods used in this study to measure
macrofracture and microfracture orientations.

2.4.1 Macrofractures

All fracture orientations were measured, and various components of the fault
zone were identified, including any lithological variation; cataclasite/ultra
cataclasite zones were noted, filled/open fracture, type of fill (e.g. chlorite filled),
and any subsidiary faults. Fractures recorded (Figure 2.5) consisted of major
fault planes, and two types of fractures; open and closed (filled).

a b

Figure 2.5 - a) Characteristic meso-scale structures associated with the Caleta Coloso fault
- open and shear fractures, parallel sets. b) Open and shear fractures.

29
Orientated samples were taken along the transect for micro-scale analysis
between macrofracture counts (see section 2.5.3). Linear fracture density of
macroscopic fractures was determined by counting the number of fractures
intersecting along two orthogonal lines (parallel and perpendicular to the main
fault trace) contained in a horizontal plane. Each of these macrofractures
counted had their orientations measured using a compass clinometer. The study
was constrained to transects on only one side of the fault zones due to steep
topography or under alluvial cover, apart from the Blanca Fault where data from
transects taken on both sides of the fault were recorded. Background
macrofracture densities and orientations were measured in the area marked in
Figure 2.2, as this area was recommended to be an area that had no close
proximity to any faults and showed minimal fracture damage (pers. com.
Cembrano, 2005).

2.4.2 Microfractures

The microfracture density of each sample has been quantified following the
methods described by Anders & Wiltschko (1994) and Wilson et al. (2003) to
determine the intensity of microfracture occurrence relative to distance from the
fault core. Fifty evenly spaced quartz grains per sample were analysed, and each
of the transect samples were cut in a horizontal plane and section. Thin sections
were cut perpendicular to the fault plane and parallel to the slip direction. This
orientation provides the maximum visibility for fault-related microfractures
(Engelder 1974, Vermilye & Scholz 1998). Vermilye & Scholz (1998) showed
that measurements made only on slides of this orientation are consistent with
measurements made by combining data from three orthogonal sections for both
bedding and orientation, except some bedding-parallel microfractures. Hence
this method is taken to be representative of microfracture density and
microfracture orientation when comparing between all faults. All visible fracture
types were noted in the analysis and divided into (1) fluid inclusion planes
(FIPs), (2) healed, (3) partially healed and (4) open microfractures (Figure 2.6).
Quartz was selected for microfracture counts as it has little fracture anisotropy in
comparison to feldspar and amphibole (which can lead to difficulties

30
distinguishing between cleavage and fractures), and hence was considered a good
proxy for the total amount of microfracture damage the rock has sustained.

Fluid inclusion plane

Open fracture

0.1 mm

Figure 2.6 - a) Optical micrograph in crossed polar light showing types of

microfractures. Open-fractures, sealed fractures, partially open fractures and
fluid inclusion planes . b) Optical micrograph (high magnification) showing
open fractures and fluid inclusion planes in more detail.

31
The fracture density was determined by counting the number of microfractures
that intersected a line of length 1.5 times the average grain diameter (Wilson et
al. 2003). At each recording site, the microscope stage was rotated an arbitrary
amount in order to randomize the counting line orientation to minimize operator
sampling bias. The total number of microfracture intersections was divided by
the total counting line length to determine the average linear density of
microfractures (Wilson et al. 2003).

For the 3 smaller displacement faults, where damage zones were on the order of
1-10cm, a few thin sections covered the total damage zone width, so
microfracture density was measured in all quartz grains, and grains at the same
distance from the fault core had their fracture densities averaged. A key aspect
of this study is that grain size and mineralogy in the country rock is the same or
very similar for all faults, making it easy to compare results. It also alleviates
problems that arise from smaller grains being harder to fracture (Wong et al.
1997).

Background macrofracture densities and orientations were measured on

orientated samples that were collected from the area marked in Figure 2.2.

32
2.5 Results

Six faults which have displacements that range over 3 orders of magnitude
(~0.2m - 5000m) have been studied in detail, and section 2.5.1 describes the
general structure of each fault in turn. Although the fault cores in general are
fairly distinct in the field, understanding the size and dimensions of the fault
damage zones is relatively hard because of the microscopic nature of the damage.
On first inspection in the field, macro scale damage can be seen to decrease with
distance qualitatively from the fault, which was used initially to identify areas for
further study. Micro- and macroscale fracture densities in the damage zones
surrounding the fault cores of these faults are then characterized using several
detailed transects so as to define damage zone dimensions and fracture patterns.
The results of those analyses are presented in section 2.5.2 and 2.5.3.
Macroscale studies were only completed for the three largest faults.

2.5.1 General structure of faults studied

This section summarises the general fault structure of each of the six faults that
are studied in detail. The damage zones of the faults are discussed in later
sections. The locations of the first three large faults are indicated on Figure 2.2.

2.5.1.1 Caleta Coloso Fault

The Caleta Coloso fault is the largest displacement fault in this study, and has
been largely passively exhumed from ~6-9km depth and is reasonably well
exposed. It lies in a NNW orientation and it extends on land at least 80 km, has a
sub-vertical dip towards the west, and in plan view is slightly curved to the west
(Figure 2.2). Its structure is defined by a wide zone of ‘multiple’ fault cores of
cataclastic rocks (Figure 2.7a), with an average thickness of ~400m and a
surrounding damage zone. The rocks within the fault core (Figure 2.7a in plan
view and Figure 2.8 in hand specimen) are predominantly highly damaged
country rock bound by discrete units of cataclasites (Figure 2.8b),
protocataclasites (Figure 2.8c), and small units of ultracataclasites. The
cataclasites and ultracataclasites (Figure 2.8a, b, d & e) vary in width from a ~1-2

33
metres to 20, separated by highly damaged and veined ‘country rock’ (Figure
2.8d) in parallel units. Within the fault core, various discrete faults can be seen
to juxtapose the different units into each other, indicating various generations of
faulting. Substantial and pervasive chlorite and epidote veining affects the whole
fault core which gives rise to it overall distinct green colouration (Figure 2.7 and
Figure 2.8) that clearly distinguishes the contact between core and damage zone
(Figure 2.7b).

Between the boundaries of each of the cataclasite / damaged rock units it is

common to see localized zones of ultracataclasite sometimes associated with
fault gouge. Gouge zones generally vary from 5-20cm, but in certain areas
towards the north, gouge zones have been observed with maximum thicknesses
of up to 4 metres (Figure 2.9). Within such exceptional examples, the preserved
gouge matrix with strong vertical foliation and clasts/inclusions with
symmetrical tails can be observed (Figure 2.9). These gouges were probably
formed more recently at a shallower depth.

34
 a

Figure 2.7 – a) Detailed map of the Caleta Coloso fault around Quebrada Isabel within the Coastal Cordillera to the south of Antofagasta. The location of the field study area is
shown by the latitude/longitude GPS position on the left of the map. b) View looking south along the strike of the Caleta Coloso fault (but flipped horizontally to match the
geological map), showing the fault core zone to the left and fault damage zone from which samples were taken for analysis on the right.

35
 a b

c d e

Figure 2.8 - a) Core zone on smaller displacement fault with strands of cataclasite. (b) Cataclasite block, with distinct fine
grained green epidote/chlorite colouration. (c) Protocataclasite showing angular blocks of granodiorite contained within
an ultracataclasite matrix containing hydrothermally-precipitated chlorite and epidote. (d) Relatively undamaged unit
with veins of epidote. (e) Example of grain size reduction into cataclasite
36
Figure 2.9 - Large fault gouge zone on Caleta Coloso Fault, with excellently preserved
structure.

The nature of the strands of fault rocks within the Caleta Coloso fault zone show
that the distribution of strain is heterogeneous, with localized planes of
comminution typically confined to the edge of one of the strands of cataclasite.
This observation is consistent with the notion of the cataclasite layers
strengthening due to hydrothermal precipitation, with later movements
accommodated on new fault gouge strands confined to the boundary of the
strengthened old cataclasite layer. This is discussed in more detail in Faulkner et
al, (in press) (included in Appendix B). The striations contained on some of the
fault planes of main fault show two generations of movement, one subhorizontal
and another subvertical. The sub-horizontal striae are preserved on fault planes
with chlorite and epidote fill, and are cross cut by later sub-vertical
slickenlines/striations preserved in the haematite filled faults. The horizontal
fault striations suggest slip between 0-20ºN from horizontal and the sense of

37
movement, deduced from offset planar markers and R1 Riedel shears in fault
gouge (Rutter et al. 1986), is sinistral. Within the fault core and damage zone,
early sub-vertical microdioritic dykes in a NW-SE orientation with thicknesses
ranging from 1-2 metres are commonly seen, with sinistral horizontal
displacement up to 3 metres, and locally with sequential separations. The overall
displacement of the Caleta Coloso fault is difficult to quantify from the present
work, but previous work has suggesting a minimum strike-slip offset of >5 km
(Cembrano et al. 2005). The vertical fault movements postdate horizontal
displacement, and show a dip-slip fault sense and topographical fault scarps
suggest that they may cut the later alluvial deposits which implies normal
faulting occurred in the Miocene-Pliocene (pers. com. González, G. 2005).
Various large splay faults with unknown displacements branch off this fault and
cross the transects (see Figure 2.2). The structure of this fault can be taken to be
representative of faulting at depth, and can be assumed to be a crustal-scale
structure by virtue of its offset. The mineralogy of the fault rock indicates
faulting at ~4-6 km depth (Cembrano et al. in review).

38
2.5.1.2 Cristales Fault

The Cristales fault is an excellently exposed sinistral strike slip fault (Figure
2.10) that runs NNW and dips 75-80 º to the west (see Figure 2.2). In plan view
the fault appears to extend to a minimum of 7.5km in length. The single fault
core zone is on average 4 metres
a
thick, and cuts the Cerro Cristales
pluton and metadiorite Bolfin
complex (Figure 2.10b); the fault
transects in this study were within
the Cerro Cristales granodiorite.
The fault core is composed of
cataclasites, and in the centre of the
cataclasite is a discrete
development of ultracataclasites
that ranges from 10 to 14 cm
thickness, with a gouge core
(Figure 2.11) that ranges between
b 20cm and 80cm in width.
Slickenlines orientated at 11-20ºS
on the fault plane suggest strike-
slip movement, with sense of
sinistral movement which can be
determined from the P / R1 Riedel
foliation fabric within the
ultracataclasites, offset dykes and
the displaced geological contacts.
The magnitude of the horizontal
displacement was determined by
Figure 2.10 - a) View north along strike of the the offset contact between the
Cristales fault. b) Geological map of the area
surrounding the Cristales fault. Cerro Cristales Pluton and the
Bolfin metadiorites at 220m.

39
Figure 2.11 - Cristales fault core showing a localized layer of fault gouge (purple) within the
ultracataclasite

2.5.1.3 Blanca Fault

The Blanca Fault is well-exposed in the central part of the area in study, and is
orientated NNW and dips between 70-80ºW. Nearly 2.5km of the fault is
exposed in plan view, and it is defined by fault core composed of cataclasite that
varies in width varies between 4 and 5 metres and an associated damage zone. In
places in the centre of the core, fault gouge of up to 20cm in width can be seen. It
displaces Lower Jurassic Bolfin Metadiorites unit into the Jurassic – Lower
Cretaceous Cerro Cristales pluton granodiorites (see Figure 2.2). Sinistral strike-
slip movement can be deduced from horizontal slickenlines, offset geological
contacts and microdioritic dykes cross cutting the cataclasite core zone.
Slickenlines suggest slip vectors between 6-14º from the horizontal, and sinistral-
normal movement determined from offset geological contacts. The magnitude of

40
the horizontal displacement is well constrained at ~35m, shown by a displaced
felsic dyke very close to the studied transects (Figure 2.12).

Figure 2.12 – a) View south along strike of the Blanca fault. b)

Geological map of the area surrounding the Blanca fault. Dashed
lines indicate inferred features.

41
 2.5.1.4 Small scale faults

The three smaller scale faults were located within the granodiorite rocks in the
study region. Faults C1, FC2-8 and FC13 (Figure 2.13) are sinistral strike-slip
faults, and using offset veins and dykes, displacements of 1.2m, 2m and 0.13m
respectively can be deduced. Each fault has a fault core of 0.06, 0.03 to 0.002m
thickness respectively, each with associated damage zones. Figure 2.13 shows
an example of one of these small scale faults (FC13), where an offset quartzite
vein shows the sinistral displacement.

Figure 2.13 - FC-13 photo. Left lateral displacement can be seen from the displaced quartzite vein.

42
2.5.2 Damage zone macrofractures

As the general structure of the six faults studied has now been described, the
quantitative data collected from the fault damage zone transects will now be
presented. Macrofracture counts were only completed on the damage zones of
three largest faults, as the smallest faults do not have sufficient amounts of
displacement to accrue significant macrofracture damage. .

a b
Caleta Coloso
Fault.
5km offset

c d
Cristales Fault
220m offset

e f
Blanca Fault
35m offset

Figure 2.14 – Equal area projections showing orientations of all macrofractures measured
within the damage zone of the three largest faults. Blue great circles represent fault plane
orientation and filled circles show the orientation of fault b-axis, normal and slip direction.
Green great circles show orientation of major subsidiary faults. a) Caleta Coloso transect 1.
b) Caleta Coloso transect 2. c) Cristales fault transect 1. d) Cristales fault transect 2. e)
Blanca fault transect 1. f) Blanca fault transect 2.

43
These small faults (displacements ranging from 0.13 – 2m in displacement) can
in fact be viewed as macrofractures themselves relative to the macrofractures
there were quantified within the damage zone of the three largest faults.

Macrofractures measured are opening and shear mode fractures or a combination

of the two (Figure 2.5), and the damage zones of all faults appear to contain very
few filled fractures. Of the few filled fractures seen, the fill consisted of chlorite
and epidote and there was no preferred
orientation of filled fractures. A general
fracture orientation vs. distance
relationship was looked for, but no
systematic pattern was found. Hence for
each fault, all damage zone macrofractures
have been plotted in equal area projections
as shown in Figure 2.14. (Figure 2.14a-f)
shows the orientation of all macrofractures
measured within the damage zone of the
three largest faults (two transects for each).
Figure 2.14e & f differ in that each transect
was on opposite sides of the fault to look
Figure 2.15 – Schematic sketch of
for possible asymmetry of fracture damage subsidiary shear fractures surround
the faults R1 and R2 shears are Riedel.
surrounding the fault core.

The Caleta Coloso fault shows in both transects (Figure 2.14a & b) that a high
proportion of macrofractures are generally orientated between 35-45° to the main
fault plane, which may be a combination of synthetic P-shear and X shears
(Figure 2.15), with a considerably smaller number in an R1 Riedel and conjugate
R2 Riedel orientation. The Cristales fault Figure 2.14c & d a high proportion of
macrofractures in conjugate R2 Riedel and X shear orientations, with the rest of
the macrofractures lying in an R1 Riedel and P shear orientation. The Blanca
fault (Figure 2.14e & f), which shows data for each side of the fault appears to
show some asymmetry, in that transect 1 shows a high proportion of
macrofractures in a R1 Riedel shear orientation, and on the other side of the fault
macrofractures predominantly lie in a P shear orientation. Many of the fractures

44
may be due to exhumation and/or weathering, but it difficult to identify these
from fault related fracturing as background fracture counts show no preferred
orientation.

In contrast to the fault core, very few filled fractures were seen within the
damage zones. Figure 2.16 shows a graph of macrofracture density vs
perpendicular distance from the fault core for the three largest faults; Caleta
Coloso, Cristales and Blanca faults. Both transects for each fault showed very
similar results, and hence the transect data have been averaged to give one
relationship for each fault. It can be seen that there is a log linear decrease in
fracture density with perpendicular distance from the fault core for all three
faults. All three fault fits show a critical macrofracture density at 0m of around
100 fractures per metre, but show varying gradients.

Figure 2.16 - Graph showing macrofracture density vs perpendicular distance from the
fault core for the 3 largest faults.

45
Assuming that the point at where the macrofracture density decreases to
background levels defines the outer boundary of the damaged zone i.e. damage
related to the fault movement alone, then damage zone widths, as defined by
macrofractures, appear to scale with fault displacement. There are localized
spikes in the fracture density where subsidiary faults are approached (later
branches from the main fault trace that cut the damage zone; the location of such
faults are indicated on Figure 2.16), hence the data are somewhat semi-
quantitative. It should be noted that these points are ignored for the line of best
fit, as it is assumed that these spikes are damage related to the subsidiary faults.
Figure 2.16 shows that macrofracture density on the Caleta Coloso Fault rises
from 30 per metre to 90 per metre as subsidiary faults are approached. The
Cristales Fault shows a similar relationship in that macrofracture density drops to
around half the critical fracture density at 40-60m, but increases in density with
increasing proximity to the subsidiary fault at 100m distance. The Blanca fault
shows a similar decrease in fracture density, but no subsidiary faults are
intercepted and no density spikes are seen. Background fracture counts were
completed on samples from undeformed regions for both macro and
microfracture densities and orientation of background fractures. Figure 2.4a &
b). Although mapped in no significant detail, mapping within the area in other
studies have shown subsidiary faults have similar offsets to the Blanca fault
(~35m), in which case the fall off to background from the localized spikes is
comparable to the damage zone width of the Blanca fault.

46
2.5.3 Damage zone microfractures

A qualitative analysis of the cross-cutting relationships of the different

microfracture types (fluid inclusion planes (FIPs), healed, partially healed and
open microfractures - see Figure 2.6) indicates that each type is a different
generation of microfracture. Only the density of FIPs showed any distinct
discernable correlation with distance from the fault core, as can be seen in Figure
2.17 which shows the relationship between microfracture density of all
microfracture types and distance from the fault, and cross cutting microfracture
relationships suggest that FIPs predate all other fractures

Figure 2.17 - Densities vs distance for all four microfracture types from the Caleta Coloso
fault.

Although the presented data are only for the largest Caleta Coloso fault, it is
representative for all faults in that only the FIPs show a correlation with distance
from the fault. FIPs show an exponential decrease in log microfracture density
with perpendicular distance from the fault core from just above 20 per mm to
below 3 per mm. Partially open fractures show a general decrease in density
with distance, although this relationship is magnified by the log scale and only
shows a drop in density from 3/mm to below 1/mm. Open fractures and sealed
fractures show no relationship with distance from the fault. Locations of the

47
subsidiary faults have been noted on the graph, but it is hard to say that they have
any effect on the microfracture density. Some crosscutting relationships can be
seen between the fluid inclusion planes, although it is difficult to identify clearly
different generations. Preliminary cathode luminescence work (not shown here)
suggests that multiple generations of fill are not seen in the FIPs.

Microfracture orientations of FIPs were measured on selected slides for the two
largest faults (Caleta Coloso and Cristales fault) using an optical microscope and
universal stage (e.g. Friedman 1969). The technique used was similar to that of
Wilson et al. (2003), although no statistical weighting was applied as
microfracture orientations were only determined in one thin section plane.
Figure 2.18 shows FIP microfracture orientations within the damage zones of the
two largest faults at various distances from the fault core. The first two
projections show summary data for all the microfractures measured for all
distances for each fault. In general, microfractures in the damage zone of the
Caleta Coloso fault appear to be predominantly in an orientation of 20º to the
fault. However, the separate plots for each distance show that at 16m distance
from the fault plane there appears to be a set of microfractures at a steeper angle
(around 85º) to the fault plane that dominate over the microfractures orientated at
20º. At 59m the higher angle microfractures become less apparent while
microfractures at 20º to the fault appear more dominant, and at distances of 107m
and 139m the higher angle set of fractures disappear while 20º microfractures
dominate. It is clear that the higher angle microfractures are more common
closer to the fault core. Conversely, the Cristales fault microfractures
cumulatively appear to be predominantly in an orientation at an angle of 70º to
the fault plane. At 9m this orientation dominates the microfracture fabric, with
little change at 38m and 90m from the fault, although some microfractures with
orientations perpendicular to the fault occur. In addition to the trends for both
faults, there appears to be a background spread of microfracture orientations
ranging from fault parallel to perpendicular. Microfracture orientation
measurements made on the same intact samples that the background density
counts were made on showed a generally diffuse microfracture fabric with no
preferred orientation.

48
 a) Caleta Coloso Fault b) Cristales Fault
All All

16m 9m

59m 38m

107m 90m

139m
Figure 2.18 - Equal area projections showing some
preliminary measurements of the orientations of
FIP microfractures within the damage of the two
largest faults. Blue great circles represent fault
plane orientation and filled circles show the
orientation of fault b-axis, normal and slip
direction. Green great circles show orientation of
major subsidiary faults. The projection has the
distance from the fault core labelled. Red and
green lineations represent an overlay of expected
orientations of microfractures predicted by the
fracture damage models discussed later.

49
Taking into account that only FIPs show any clear relationship with
perpendicular distance from the fault core, Figure 2.19a & b show FIP
microfracture density variations with perpendicular distance from the fault core
for the three smallest displacement faults (Figure 2.19b - 2m, 1.2m and 0.13m
respectively) and the three largest displacement faults (Figure 2.19a - 5000m,
220m and 35m displacement respectively).

a b

Figure 2.19 - a) Graph showing FIP microfracture density vs perpendicular distance from the
fault core for the 3 smallest faults. (b) Graph showing FIP microfracture density vs
perpendicular distance from the fault core for the 3 largest faults. Error in fracture density is
typically ± 2 microfractures / mm.

The FIPs show a log-linear decrease in microfracture density with perpendicular

distance from the fault plane for all six faults, as seen with the macrofractures
(Figure 2.16). Figure 2.19a & b show that microfracture density with distance
for all 6 faults follow the relationship:

F = F0 e −αx 2.1

where F = fracture density (#/m),

F0 = critical fracture density (#/m)
-α = gradient
x = perpendicular distance from fault (m).

There appears to be a critical microfracture density shared by all six faults of

around 20-25/mm suggesting that the maximum fracture density for this rock is
independent of fault displacement. Background fracture density, Fb, counts were

50
completed on samples from undeformed regions (Figure 2.4) for both macro and
microfracture density. Assuming that the decrease in microfracture density to
background levels defines the outer boundary of the damaged zone, then damage
zone widths as defined by FIPs can be determined (Table 2.1):

Fault FIP Damage

Displacement gradients 1 zone width Core zone
Fault name D (m) F0 α α Xdz (m) width (m)
Caleta Coloso 5000 17.77 0.011 91.743 149.073 230
Cristales 200 18.84 0.015 68.493 115.275 4
Blanca 35 12.64 0.070 14.368 18.450 8
C1 2 20.22 31.589 0.032 0.056 0.06
FC2-8 1.2 22.35 20.774 0.048 0.089 0.03
FC-13 0.13 23.03 38.962 0.026 0.050 0.002

Table 2.1 - Summary of fault data

These damage zone widths that appear to scale with either displacement or size
of the fault. Therefore, to compare between the 6 faults, the gradients of each of
the empirical functions shown in figures Figure 2.19a & b have been plotted vs
displacement, as seen in Figure 2.20. The gradients are divided by 1 in order to
show a positive correlation.

51
 Figure 2.20 - gradients from functions in Figure 2.11a & b versus fault
displacement

Figure 2.20 shows a log-log (exponential) relationship between the slope of the 6
functions in Figure 2.19a & b and fault displacement, following the relationship
shown in equation 2.2:

1 1
= aD b or = 0.0913D 0.938 2.2
α α

where b = gradient (0.938)

α = gradient (of density distance functions from equation 2.1)
D = displacement
a = intercept (0.0913 at D=1).

By combining equations 2.1 and 2.2 an empirical function for fracture density F
with respect to displacement D, perpendicular distance to the fault x and critical
microfracture density F0 can be derived for these faults:

52
 x
−
F = F0e 0.0913 D 0.938 2.3

where F = fracture density

F0 = critical fracture density
x = perpendicular distance from fault (metres)
D = displacement (metres)

2.5.4 Damage zone microfracture and macrofracturing

compared

Figure 2.21a,b & c show microfracture and macrofracture density variations with
perpendicular distance from the fault core for each of the three largest faults
(5000m, 220m and 35m displacement respectively). This shows more clearly the

a b

Figure 2.21 –micro and macrofracture

density vs distance from the fault core for
the (a) Caleta Coloso fault, (b) Cristales
fault and the (c) Blanca fault. The
locations of large subsidiary faults have
c been marked along the transect for
reference.

spikes in the macrofracture densities caused by the subsidiary faults that were
discussed earlier. However, note that where subsidiary faults (marked on the
graph) intercept the transect, there are localized spikes in macrofracture densities,
which are not seen the FIPs relationships, suggesting that either the FIPs formed
before the subsidiary faults, or the fractures have a very localized zone of FIP

53
damage that is not recorded by our sampling. It might be argued that this is
evidence to suggest that the fluid inclusion planes formed before most of the
macrofracture damage.

Figure 2.22 shows the microfracture and macrofracture density for both sides of
the Blanca Fault. This was the only fault to be sampled on both sides due to time
and exposure issues. It can be seen that macro and microfracture densities are
approximately symmetrical on both sides of the fault.

Figure 2.22 - Micro and macrofracture density vs distance from the fault core
for both sides of the Blanca fault.

2.5.5 Fault core zone

It can be clearly seen from the photos and maps in Figure 2.7, Figure 2.10 and
Figure 2.12 and the data in table 1 that there is some variation in fault core width
and size of the fault. Figure 2.23 shows the relationship between core width and
displacement, and shows a positive log – log relationship between fault
displacement and fault core width. It must be noted that the upper limit on

54
displacement on the Caleta Coloso fault is unknown and what is used is a lower
limit (5km).

10-1

10-3

Figure 2.23 - Fault core widths vs. fault displacement. Lines of 10-1 and 10-3
refer to the scaling parameter k discussed later in the discussion.

55
2.6 Discussion

In this section an introduction to various existing models on the origin of off-

fault damage is provided as a context for discussing the results. Variations in
fault structure are first examined, followed by fracture damage surrounding the
faults and the various scaling relationships of the faults and finally the possible
origins of the damage surrounding these faults are commented on.

2.6.1 Current models on the origin of off-fault damage

It is likely that fracture damage surrounding fault zones is created by various

cumulative processes, ranging from early initiation and fault tip processes to
fracture damage induced by dynamic rupture events. Wilson et al. (2003)
showed that microfracture fabric and orientation can be used to investigate the
palaeostress conditions at the time when the microfractures were formed,
assuming that microfractures are mode I fractures and that microfractures will
form in an orientation perpendicular to the minimum compressive stress
(Engelder 1974, Friedman 1963). The microfracture orientations relating to
faulting events within the damage zone should therefore record variations in
stress states through the history of the fault. Various existing models for the
origins of off-fault damage are now each described in turn, as a context in which
the results can be discussed. Table 2.2 shows a summary of the fault damage
characteristics expected from each of the models.

(1) Fault initiation models. The Anderson model of faulting assumes simple,
homogenous stress states in the crust and Coulomb failure behaviour and
suggests faults will form at angles of ~25-30º with respect to the maximum
principal stress driving the deformation (Anderson 1942, Scholz 1990). It has
been demonstrated experimentally that such faults are formed through the
interaction and coalescence of many tensile microcracks (Lockner et al. 1991,
Lockner et al. 1992). These microcracks are orientated parallel to the principal
compressive stress, and hence microfractures related to Andersonian faulting
should lie at an angle of 25-30 º to the fault plane (Figure 2.24a). The model

56
predicts that microfractures should be diffusely distributed in the rock mass,
apart from very close to the fault plane where microfractures coalesce, and hence
no variation in fracture density with distance from the shear fracture should be
seen.

Figure 2.24 - Schematic diagrams illustrating mechanics-based models for microfracture

development associated with fault formation and wear. From (Wilson et al. 2003). (A) Fault
models assuming homogeneous stress. Anderson model for fault formation predicts
microfracture orientation at approximately 30º to fault. (B) Fault model for fault growth by
tip propagation. Microfractures form in the region of the fault-tip stress concentration. (C)
Fault model for wear along wavy, frictional fault surfaces.

(2) Once a fault has initiated, the migrating process zone model predicts
microfracture formation as the fault continues to propagate. It appeals to non-
linear or post-yield elastic fracture mechanics models (Barenblatt 1962, Dugdale
1960, Scholz et al. 1993, Vermilye & Scholz 1998) that predict a zone of
inelastic deformation in the form of intense microfracture damage surrounding
the propagating fault tip that is related to the taper of fault tip displacement

Figure 2.25 - Sketch showing relationship between fault displacement and damage
surround the fault tip. (a) a purely linear elastic fault tip, where the displacement gradient
at the fault tip is infinite and no zone of damage is produced (b) a non-linear elastic fault
with a finite displacement gradient that is related to the size of the damage zone, and (c)
illustration of how the damage zone might be produced as the fault propagates.
(Figure 2.25).

57
This zone is referred to as the process zone (Figure 2.25b), (Cox & Scholz 1988,
Reches & Lockner 1994, Scholz et al. 1993) or frictional breakdown zone
(Cowie & Scholz 1992a, 1992b). As this process zone migrates (Figure 2.25c), it
leaves a damage zone of microfractures in its wake. The higher the displacement
gradient at the fault tip is, the smaller the surrounding zone of deformation (and
hence consequent damage zone) should be (Cowie & Scholz 1992b, Scholz et al.
1993, Scholz & Lawler 2004). Hence a purely linear elastic fault would be
expected to have an infinite displacement gradient, a stress singularity and no
fracture damage surrounding the fault tip (Figure 2.25a), which is, of course,
unrealistic. This model that the width of the damage zone should scale with the
width of the process zone and microfracture density should decrease with
distance from the fault plane. The process zone size/damage zone width will be
determined by the stress at the fault tip which is proportional to active fault
length and the remote stress. Following the work of Lawn & Wilshaw (1975)
who calculated stress orientations and magnitude within the region of a
propagating tip of a mode II fracture, Scholz et al. (1993) calculated the stress
field and show the expected orientations of fractures. On the compressive side of
the fault tip σ1 is orientated at a low angle to the fault plane (~20º), whereas they
are at a high angle (70º) in the tensile side of the tip (Figure 2.24b). The
microfractures will therefore be orientated at these angles to the fault plane.
Various studies have looked at microfractures surrounding fault zones, both in
field studies (Anders & Wiltschko 1994) and experimental studies (Moore &
Lockner 1995), and showed data that fit with this model. Anders & Wiltschko
(1994) suggested that the bulk of fractures surrounding the faults they studied
may have been formed near the leading edge of a propagating mode II or mode
III fault tip [process zone] with microfracture orientations consistent with model
predictions. Experimental studies by Moore & Lockner (1995) where they
investigated microcracking related to the growth of a shear fracture in Westerly
Granite, showed results that were consistent with those found in the field by
Anders & Wiltschko (1994) . They showed that the predominantly tensile
microcracks, that were formed during an experiment, showed a rapid decrease in
microfracture density with increasing distance from the fault, and a significant
concentration of microfractures surrounding the dynamically propagating fault

58
tip. Similar results were seen by Janssen et al. (2001) in the processes zone
surrounding a controlled propagating shear fracture.

(3) Fault wear models generally suggest that fracture damage may occur due to
juxtaposition of fault irregularities and consequent stress cycling (Chester &
Chester 2000, Flinn 1977, Wilson et al. 2003). Chester et al. (2004) highlighted
that continued slip on irregular faults will produce stress cycling and damage in
the host rock that ultimately may overshadow the relict damage from tip
propagation. Power & Tullis (1991) and Scholz & Aviles (1986) have
demonstrated that fault surfaces at all scales will show roughness, and as a result,
outwardly expanding zones of damage might be expected as fault displacement
accumulates. This is because larger wavelength roughness elements will be
juxtaposed with greater slip (Scholz 1987); the result being that the thickness of a
damage zone should scale with total fault displacement. Observations on many
crustal-scale faults show that their structures evolve toward simplicity (e.g.
simple planar active fault strands) as a function of the fault displacement (Ben-
Zion & Sammis 2003, Chester & Chester 2000, Tchalenko 1970). Unlike the
process zone model that suggests fracture damage is likely to record early fault
fracture processes, fault wear processes will record continuous damage related to
the entire fault’s history. In many cases this may overshadow fault tip
deformation, which may explain why the fault tip model (1) has only been
successfully applied to the fabrics and fracture density distributions surrounding
small and immature faults. Microfractures are expected to form in the region of
stress concentrations formed by offset along rough surfaces, in orientations that
deviate from far field stress. Wilson et al. (2003) suggested that analysis of
stress states induced by slip along non-planar frictionless fault suggests that the
maximum compressive stress can vary locally from an orientation parallel or
perpendicular to the fault pane (e.g. Chester & Fletcher 1997, Saucier et al.
1992). Chester & Chester (2000) showed using a mechanical analysis of slip
along non-planar faults with friction, that fractures are most likely occur at
angles to the fault that are slightly greater than the far-field maximum stress.
Hence microfracture orientations will vary along the fault in a similar manner
(Figure 2.24c). Principal compressive stress orientations and magnitudes, and
hence microfracture orientations, depends therefore on the roughness of the fault

59
surface, the elastic modulus, fault friction, and far-field principal stress
orientation. For some loading conditions and weak faults (i.e. faults with the
principal compressive stress at a high angle to the fault plane resulting in a low
resolved shear stress), microfractures form approximately normal and parallel to
the fault surface.

(4) Earthquake rupture events will also create off-fault stress concentrations with
related fracture damage (Rudnicki 1980, Wilson et al. 2003). Recent work on the
San Andreas fault has shown that seismic rupture may have produced
‘pulverization’ of the surrounding rock, (Dor et al. 2006b, Wilson et al. 2003).
Generally the microfractures that are created by an earthquake rupture tip are
expected to be formed in a similar orientation to the fractures formed by the
migrating fault tip model, although Cowie & Scholz (1992) suggested that the
region of damage may be smaller that the fault tip model predicts. Brune et al.
(1993), suggested that fault breccia can be formed by dynamic stress changes
during an earthquake rather than just typical mechanical models. They suggested
that gouge could form explosively during the dynamic reduction of normal stress
across a fault accompanying the propagation of a “wrinkle pulse” in the fault as a
result of a large dynamic slip event. Transient decreases in normal stress during
rupture would predict microfractures with orientations at small angles to the fault
plane. Rice et al. (2005) showed that the dynamic stress field from a propagating
slip pulse can produce Coulomb slip on pre-existing fractures within the fault
damage zone. The magnitude and extent of this damage is controlled by the
stress drop of the slip event, static and dynamic friction coefficients, rupture
velocity, principal pre-stress orientation and poroelastic (Skempton) coefficients.
A change in the stress orientation is dependent on rupture velocity – if rupture
velocity is equal or greater to the Rayleigh wave speed then microfractures
would be expected at a high angle to fault microfractures. Using their model to
analyze slip pulses of previous recorded large slip events, their predictions are
consistent with those of with observations of Wilson et al. (2003) that a zone of
oriented microfracture damage extends to distances of about 30 m from the
exhumed Punchbowl Fault in southern California, and that other orientations of
microfractures are above the background level to distances of about 100 m. A
good field indicator of faults being subject to earthquake slip events is the

60
presence of slip induced frictional melt (pseudotachylytes). Pseudotachylytes are
locally seen at various scales in injection veins at generally high angles to the
fault plane, ranging from small fractures to up to a 1m from the fault (Sibson &
Toy 2006).

Table 2.2 - Summary of models of microfracture orientations within the damage zone of a
strike slip fault

Model of Microfracture Microfracture orientation wrt Microfracture

fault damage timing fault plane distribution and damage
zone
1. Andersonian Earliest ~25-30º to fault Diffuse
2. Migrating Early at fault ~70º on the tensile side and Damage zone scales with
process zone tips ~20º on the compressive side of displacement and fault
fault tip length. Decrease in
density with distance
from fault.
3. Fault wear Mid-late Varying orientations that can Decrease in density with
vary from fault normal to distance from fault,
parallel. Weak faults can show controlled by factors such
strong fault parallel as fault roughness.
orientations.
4. Earthquake Late Similar to (1) but also can be Decrease in density with
slip low angle due to sudden distance from the fault,
decrease in fault normal stress. although distance
Depends on rupture velocity predicted to be smaller
and stress drop. than for fault tip
Pseudotachylytes injection relationships.
veins often orientated at 90°

61
2.6.2 Fault structure, scaling relationships and applicable
models

All six faults studied share similar qualitative aspects in that they are sinistral
strike-slip faults composed of a 2 part structure; a fault core that has
accommodated the bulk of the total displacement with an associated damage
zone, both of which have variable sizes. In light of the fault damage models
previously discussed, fault structure and damage scaling relationships will now
be discussed.

2.6.2.1 Fault damage zones

Both micro and macrofracture densities were measured in detail throughout the
damage zone. Unlike the fault cores which were locally marked by distinct green
colouration, it is difficult to identify where the damage zone extends to.
However, using fracture density – distance profiles, damage zone widths can be
inferred from the distance at which fracture densities drop to background levels.
Chester et al. (1993) suggested that using microfracture density relationships to
infer effective fault zone widths in this way are likely to be consistent with
geophysical definitions of faults using seismic wave propagation or gravity
techniques of Feng & McEvilly (1983), Stierman (1984), Wang et al. (1986) and
Mooney & Ginzburg (1986). The results in this study show that both
macrofracture and microfracture densities decrease with increasing perpendicular
distance from the fault core. Such results have been well documented for
individual small faults both in experimentally deformed samples (Janssen et al.
2001, Moore & Lockner 1995), and in various field studies (Anders & Wiltschko
1994, Brock & Engelder 1977, Scholz et al. 1993, Vermilye & Scholz 1998,
Wilson et al. 2003). This will now be discussed in more detail.

The macrofracture densities for the three largest faults in Figure 2.16 show that
macrofractures decrease through the damage zone as a function of distance from
the fault core. There are three key observations; there is a significant difference
between the damage zone widths of the two largest faults and the smaller Blanca
fault, there are large localized spikes in fracture density on the transects where

62
subsidiary faults are approached / intercepted and there is a critical macrofracture
density. The damage zone width as defined by the macrofractures appears
therefore to scale with fault displacement, but macrofractures surrounding more
faults within this range of displacements need to be studied to improve
confidence in this deduction. Wilson et al. (2003) showed similar localized
spikes in macrofracture density due to the proximity of subsidiary faults. It is
expected that the subsidiary faults that intercept the fault transects will have their
own related fracture damage with similar density-distance profiles to the faults
studied in detail here; the displacements of the subsidiary faults, although not
mapped, are thought to be between 20-50m, so it is therefore no surprise that the
density spikes die off with comparable distances to the width of the damage zone
of the Blanca fault (~35m displacement). Figure 2.16 also shows that there
appears to be an critical macrofracture density of around 100/m, which might be
expected as there will be a certain critical level of fracturing before fracture
damage is so intense that brecciation and cataclasis begin. Observations of the
damage zone close to the core show very high degrees of fracturing and damage,
with decreasing grain sizes due to comminution into the fault core. Previous
studies have shown that other faults have initiated on pre-existing joint planes
(e.g. cooling joints), for example those described by Martel et al. (1988) for the
Mount Abbott quadrangle faults in the Sierra Nevada, or the Gole Larghe fault
within the Adamello batholith in the Italian Alps (Di Toro & Pennacchioni
2005). In this study background fracture counts and orientations show no
preferred orientation of joints in relatively undamaged areas of isotropic rock, it
is inferred that majority of macrofractures are related to faulting. It must be
noted that the background macrofracture count varied from 3/m to 30/m. Some
fractures will undoubtedly be due to exhumation and associated stress release.
Macrofracture orientations within the damage zone appear to be in a variety of
shear fracture orientations typical of shear fractures surrounding strike slip fault
zones, and do not seem to differ in orientation between the three faults.

Multiple generations of microfractures are represented by fluid inclusion planes

(FIPs), healed, partially healed and open microfractures. Only the FIPs show a
log-linear decrease in density with perpendicular distance from the fault plane
(Figure 2.19a & b), The occurrence of a critical microfracture density, F0, at 0 m

63
from the core shared by all six faults suggests that maximum microfracture
density of the FIPs is independent of fault displacement. Vermilye & Scholz
(1998) made similar observations where process zone microfractures show
logarithmic density increases with proximity to the fault and a constant
maximum density that is independent of fault length (which is proportional to
fault displacement). This critical density may well be a material-dependent
property, i.e. a maximum density of fracturing occurs before further processes
such as cataclastic deformation takes over, although our microfracture densities
are a lot lower than in previous studies, perhaps due to the smaller proportion of
quartz and preferential partitioning of fracturing within feldspars and
amphiboles. As Anders & Wiltschko (1994) speculated, because the
microfractures are preserved in, and the stresses are transmitted by, not just
quartz but other minerals present, their proportion of the total mineralogy should
affect microfracture density. Despite their study showing no relationship
between microfracture density and different modal mineralogy, it still remains a
useful constant in this study and suggest that the data are representative of the
relative amount of microfracture damage that the rock surrounding all faults
studied have sustained. The damage zone widths inferred for the 6 faults suggest
that they scale with fault displacement (Figure 2.19a & b and Figure 2.20),
following the empirical relationship shown in equation 2.3. Anders & Wiltschko
(1994) suggested that the effect of displacement on microfracture density is small
to non-existent for the rocks they studied, in that the damage zone width and
critical microfracture density did not appear to vary with increasing
displacement. This study agrees with Anders & Wiltschko (1994) in that there is
critical microfracture density, but differs in that fracture density-distance profiles
and hence damage zones widths clearly scale with fault displacement. This is
consistent with damage models of fault tip propagation, fault wear and
earthquake wear, but not with Andersonian faulting (see Table 2.2).

As discussed in the previous section, microfracture orientation can be used to

make inferences on palaeostress orientations at the time of microcrack
generation, which in turn enables inferences on what processes may have been
dominant in fracture damage generation. Wilson et al. (2003) investigated the
origin of microfractures fabrics surrounding the Punchbowl Fault by comparing

64
the orientation of the measured fabrics with the orientation of the fabrics
predicted on the basis of mechanical models of fault formation and wear. Using
this methodology, Figure 2.26a shows the microfracture orientations predicted on
the basis of the models discussed in section 2.6.1 and summarized in Table 2.2
with the orientation, sense of shear and slip direction appropriate for the Caleta
Coloso and Cristales Faults.

b c

Caleta Coloso Fault Cristales Fault

Figure 2.26 - a) Expected orientations of microfractures predicted by the models
discussed previously. Modified from Wilson et al. (2003). b) Comparison of fault model
predictions with orientations of preliminary study microfracture orientations for the
Caleta Coloso Fault and b) Cristales Fault.

For all models, the normal to the average microfracture orientation (the minimum
compressive stress direction) should lie in the same plane containing the slip
vector of the fault studied and the normal to the fault surface (Wilson et al.
2003)(Figure 2.26a). Figure 2.26a & b show a comparison between these model
predictions and the summary of all microfracture orientations measured in the
damage zones of the Caleta Coloso Fault and the Cristales Fault. The
predominantly low angle to the fault-plane microfracture-set seen in the Caleta

65
Coloso fault appear to be most consistent with predictions of damaged accrued
on the compressive side of a propagating fault tip, with a smaller set of
microfractures at a higher angle of around 70-85º to the fault plane being roughly
congruent with fault wear models. Figure 2.18a shows the latter set dominates
closer to the fault, whereas further out into the damage zone the higher angle
microfractures appear to dominate which would suggest that fault wear
microfractures dominate closer to the fault core in the damage zone, and further
out damage related to fault tip processes becomes more dominant. Within the
damage zone of the Cristales Fault, it can be seen that a single microfracture set
dominates at an angle of around 70° to the fault plane consistent with that
predicted for the tensile side of a process zone surrounding a propagating fault
tip. As shown in the results, Figure 2.18a illustrates that this microfracture set
occurs throughout the damage zone at all the intervals sampled. The background
spread of microfracture orientations for both faults ranging fault-parallel to fault-
perpendicular could be accrued from some or all of the damage models, and
therefore appears to be generally clustering around the predicted orientation for
Andersonian faulting. Wilson et al. (2003) showed that within the damage zone
of the Punchbowl fault (San Andreas system, California) the most distinct set of
microfractures were perpendicular to the slip direction of the fault and were
present throughout the damage zone, implying the principal compressive stress
was at right angles to the fault within the damage zone. They suggested that this
is most consistent with local damage accumulation from stress cycling associated
with slip on a geometrically irregular, relatively weak fault surface. Here, both
faults appear to show microfracture orientations more consistent with the fault tip
processes, although the Caleta Coloso Fault also shows predominant high-angle
microfracture orientations closer to the fault core that may be more consistent
with wear mechanisms as they are not seen throughout the damage zone. This
could be explained by localized stress concentrations near the fault core due to
juxtaposed geometric irregularities on the slip plane. It is difficult to identify
fracture related to earthquake rupture on either fault, although this would be
more likely to be seen on the Caleta Coloso fault as it is a much larger crustal
scale fault than the Cristales fault and as a result is more likely to have
undergone earthquake rupture. However, considerably more microfracture
orientations need to be measured at more sampling locations to get a more

66
statistically viable fabric dataset within the damage zone. Also, it must be noted
that it is possible that local stress variations within the damage zone could effect
microfracture orientations and may not be representative of true damage zone
palaeostress. However, greatest care was taken in sampling of rocks for thin
sectioning not to take samples in close proximity to highly damaged zones.

The FIPs appear to be the youngest set of fractures within all of the faults, as
they are cross cut by all other microfracture types, and are unaffected by large
subsidiary faults that crosscut the transect which appear to cause localized spikes
in macrofracture density. However, if these FIPs are were formed by early fault
tip processes as well as wear and earthquake rupture, then it suggests that the
fractures healed at a relatively late stage. Many of the subsidiary faults may be
related to the later fault duplex in the area, as described by Cembrano et al.
(2005). If the FIPs in this study do indeed record the microfractures created from
a migrating fault tip, then it would suggest that the size of the process zone (and
therefore damage zone) is controlled by the variation in fault length as inferred
by displacement. Figure 2.27 shows a schematic diagram of a possible model
based on the scaling results from this study and fault tip migration models.

67
The primary variables that control the size of the process zone are the half-length
of the active fracture, c, which is linearly proportional to the stress at the fault tip

Figure 2.27 - Schematic diagram of process zone / damage zone scaling with fault length.
Note that this is a simplified diagram as the process zone is shown to have no asymmetry as
for mode I, which may not be the case for a process zone of a mode II fracture.

as a function of the remotely applied stress (e.g. Lawn & Wilshaw 1975).
(Cowie & Scholz 1992) suggested that the zone of enhanced stress around a
rupture tip is approximately 10% of the length of the slip patch. Assuming that
the active fault length (i.e. the length of fault that is not locked and acts to induce
stress at the tip), 2c, is proportional to displacement e.g. studies showing that
max displacement scales to fault trace length (Cowie & Scholz 1992b, Dawers &
Anders 1995, Gillespie et al. 1992, Schlische et al. 1996, Vermilye & Scholz
1998, Walsh & Watterson 1988), then stress at the fault tip is controlled by active
fault length and the remote stress. This in turn will create a process zone of a
size that is scaled to these variables.

A model proposed by Cowie & Shipton (1998) conceived of fault growth as

occurring by repeated slip on many small patches on the fault surface. This
model successfully demonstrated that observed fault displacement profiles can be
modelled by the summation of many small slip events without creating
unrealistic stress concentrations at the fault tip (Cowie & Scholz 1992b). It must
be noted that the active fault length will be much smaller than the fault trace
length, but it can be assumed that the relationship is still proportional. As active

68
fault length increases with increased displacement, and the fault tip migrates
leaving microfracture damage in its wake, and stress at the fault tip will increase
but will be relaxed by an increase in process zone size, proportional to the new
active fault length. The faults shown schematically in Figure 2.27 will therefore
expect to have density – distance profiles similar to the ones shown within the
damage zone as shown in the results of this study. Vermilye & Scholz (1998)
showed similar relationships between fault length and process zone width. Of
course this is a very simplified case; process zones will probably be
asymmetrical in shape and there will be a maximum active fault length due to
critical stresses applied at the fault tip. Also, fracture mechanics assumes a
frictionless crack, hence why it is more applicable to mode I cracks. As well as
this, faults are not perfectly planar, and there will be sticking patches along the
fault where stress at the tip is retarded by friction along the fault plane. If
microfracture damage is accrued from the migrating process zone model, then it
would predict a dog-bone shape to microfracture damage surrounding a fault, as
shown in Figure 2.28. Further microfracture studies need to be completed at
several locations along strike of the faults to confirm this.

Figure 2.28 - Schematic diagram of process zone migration leading to a dog-bone shaped
damage structure

Despite microfractures and macrofractures being formed by different

mechanisms, during the same faulting episode, they are formed in response to the
same stress fields in the same rock type under the same temperature and fluid
pressures. Anders & Wiltschko (1994) pointed out that the form of
macrofracture and microfracture fall-offs with distance from fault cores are very
similar for various independent studies, and this suggests that a universal scaling
law between the two can be established. They identified a need for a more
thorough study between microfracturing and macrofracturing to be completed, as
previous comparisons were made more difficult as they were between faults in

69
various different lithologies and different faulting conditions. Rock type, pore
fluid and stress conditions at the time of failure for all faults in this study were
probably similar. Macrofracture-defined damage zone widths appear to be quite
similar to microfracture-defined damage zone widths, although if higher
background counts of fracture density are used, macrofracture damage zones are
somewhat smaller that microfracture damage zones. The Blanca fault shows
fairly symmetrical fault damage on both sides of the fault (Figure 2.22), and
damage zone widths defined by both microfractures and macrofractures are
similar in dimensions for both sides. If the bulk of microfracture damage is
predominantly related to fault tip processes as suggested by the results, the
majority of microfractures related to fault wear and rupture will never exceed
distances greater than those created by fault tip processes. The growth of an
individual earthquake rupture involves a breakdown process at the rupture tip
similar to the process of fault growth (Rudnicki 1980, Scholz et al. 1993,
Swanson 1992). As Chester et al. (2004) discussed, the structural signature from
passage of a seismic rupture probably is a narrow zone or zones of concentrated
shear demarcating the rupture surface within a broader zone of distributed
fracturing. They suggested that the cumulative effect of numerous ruptures of
large-displacement, seismic fault may be a damage zone characterized by a
thickness that scales to the breakdown dimension of earthquake ruptures. They
suggested that the breakdown dimension for an earthquake rupture on an existing
fault is probably less than that for formation of a new fault (Cowie & Scholz
1992b). Thus, the thickness of a damage zone produced during seismic slip may
be less than that produced during initial stages of fault formation and growth. If
larger estimates of background macrofracture densities are used, then it suggests
that macrofracture damage zones widths related to repeated rupture events are
indeed smaller than damage zone widths defined by FIP microfractures. As both
fault wear models and fault tip models predict damage zones of widths that scale
with displacement, despite implications from preliminary microfracture
orientation data, it is difficult to confidently identify which models are supported
by the data presented here.

As well as the models discussed in this study, the interaction of faults with other
faults at all scales must be considered. Many fault studies showed fault tips ,

70
fault linkages and terminations indicate that linkage zones can be on the scale of
up to 400m wide and 1km long (Evans et al. 2000), and consist of altered and
fractured rocks with numerous through-going slip surfaces. As faults grow, it is
important to understand that there will be a combination of in-plane fault
propagation as well as larger scale fault linkage that will further complicate off-
fault damage accumulation.

2.6.2.2 Fault core zones

Figure 2.23 shows a distinct relationship between fault core width and increasing
displacement; the cataclasite cores appear to widen with increased fault maturity
(see also Table 2.1). Such field relationships have been documented before,
where it has been shown that thickness, T, of the cataclasite and gouge zone
found in the fault core scales linearly with total fault displacement T = kd, where
the scaling parameter k ranges between 10-1 and 10-3 (Hull 1988, Marrett &
Allmendinger 1990, Robertson 1982, Scholz 1987) and are consistent with
steady state frictional wear mechanisms (Scholz 1987). These scaling
parameters are overlaid onto Figure 2.23, and show that the data fit well. Within
the fault cores of all six faults, strain has been localized to discrete gouge zones
ranging in size from mm up to 3 metres, although the wider gouge zones may be
representative of later faulting at shallower depths. The gouge layers probably
also scale with displacement, although gouge widths were not measured in any
detail here. This is consistent with Sibson (2003) who suggested that some fault
cores, although they are contained within damage zones of variably fractured
rock ranging up to hundreds of meters in thickness, will often comprise just a
few centimetres of gouge/ultracataclasite that have accommodated large finite
displacements (>1 km).

The most distinct difference between the various faults studied here is the
presence of a very wide multiple core zone on the largest Caleta Coloso fault
(Figure 2.7a & b). The core is made up of discrete bands of cataclasite and
ultracataclasites that bound units of highly fractured and veined country rock and
protocataclasite. Although not shown at the scale of mapping, these bands

71
anastamose and connect at greater distances along strike, and probably also with
depth. Figure 2.29 shows a schematic diagram of Caleta Coloso fault in direct
comparison to single core models, as seen in Figure 2.1. The large scale
precipitation of predominantly chlorite and epidote that gives rise to the dark
green colouration of the core (Figure 2.7b) indicates a high flux of hydrothermal
fluids. These fluids may perhaps be a result of the fault being a part of the
Andean magmatic arc, where geothermal gradients are elevated and
devolatilization of magmatic products may have led to enhanced fluid
movements (Faulkner et al. in press). The structure of the fault is interpreted to
be influenced strongly by precipitation hardening of the fault zone leading to
strengthening and subsequent failure along a new plane parallel to the fault zone
(Faulkner et al. in press), and crosscutting relationships seen by faults within the
core juxtaposing cataclasites against protocataclasites support this. There are
experimental data that support the idea of rapid mineral precipitation and
strengthening in quartz-rich fault gouges at hydrothermal conditions (Olsen et al.
1998, Tenthorey et al. 2003, Yasuhara et al. 2005), and over time periods
commensurate with the interseismic period. Other studies of large scale faults
have suggested that the thickness of the central gouge layer / core zone varies
significantly along strike but may reach a maximum width that is unrelated to the
total amount of slip (Hoffmann-Rothe 2002, Shipton & Cowie 2001). However,
this may be due to a variation in faulting mechanisms because of different
lithologies; grain rotation in sandstone as well as porosity collapse and formation
of deformation bands leading to fault strengthening behaviour. The models in
this study are more specific to crystalline rocks.

72
Figure 2.29 - Schematic sketch of the multiple core zone and damage zone seen on the
Caleta Coloso Fault. The shaded portion represents the core zone and zone of chlorite-
epidote mineralization.

73
2.7 Conclusions

The strike-slip faults seen in this study appear to share many similarities over a
range of sizes; all faults are sinistral strike-slip faults that crosscut rock of
granodiorite composition and consist of a fault core and damage zone of varying
widths. The fault core zone is defined by finer grain sizes, variably deformed
rock of proto to ultracataclasites with localized bands of fault gouge that reflect
high shear strains, evidence of enhanced fluid flow, fluid-rock interactions and
resultant mineral alterations. The mechanical behaviour of this low-porosity
crystalline rock is localized, as might be expected, showing brittle fracture with a
single core which is seen for the five smallest faults. The fault with the greatest
displacement has a wide multiple core, which is interpreted to be a product of
precipitation healing of the fault zone leading to strengthening that spreads the
deformation over a wider zone. Localization of slip is seen typically on the
boundaries of ultracataclasite layers. There is a distinct relationship between
both fault core and damage zone dimensions and fault maturity. Fault core width
scales with fault displacement. Fault damage zones are zones of both micro and
macrofracturing that show a log-linear decrease in fracture density, with
perpendicular distance from the fault core. Macrofractures are predominantly
shear fractures orientated at high angles to the fault. Microfractures are a
combination of open, healed, partially healed and fluid inclusion planes (FIPs).
Damage zone widths, defined by microfracture and macrofracture density scales
linearly with fault displacement and hence fault length, and an empirical equation
for fracture density based on the evolution of displacement has been derived for
these faults. Preliminary microfracture orientations suggest that this damage
may be due to processes related to enhanced stress at fault tips, in addition to
cumulative wear processes from the juxtaposition of geometrical irregularities on
the fault plane and damage from dynamic rupture, although with the current
dataset it is difficult to suggest which dominates. Only the FIPs show this
relationship with distance in the damage zone. Cross cutting relationships in the
FIPs and the lack of apparent effect on the FIP densities from the latest
subsidiary splay faults suggest that FIPs may have remained open through most
of the faulting, but healed before late stage fault duplex development in the area.

74
The bulk of fluid flow appears to have been largely restricted to fault core zones
as shown by localised hydrothermal precipitation seen within the fault core that
is not present in the damage zone, especially with increased fault maturity.
Although FIPs suggest the presence of fluid in fractures, relatively little fluid
flow was seen within the damage zone in comparison to fault cores. Whatever
the dominant processes of the generation of off-fault damage, the scaling
relationships seen in the study allow us to better understand and predict how fault
zone permeability evolves with time.

75
3 Experimental design and the measurement of
permeability

3.1 Introduction

All of the experiments reported in this thesis were conducted on the “high
pressure high temperature triaxial deformation apparatus with fluid flow” that
was constructed by the author of this PhD thesis as a critical part of this PhD in
the Rock Deformation Laboratory, Department of Earth and Ocean Sciences at
the University of Liverpool. Dan Faulkner was also heavily involved in the
construction of the apparatus, and was responsible for the design of the main
vessel and high pressure parts. This chapter outlines the technical specifications
and design of the apparatus, details the calibrations necessary to use the
equipment for deformation and permeability measurements and also presents
discussion on the aspects of the design and experimental procedure which
influence the quality of the mechanical data obtained. Finally, the methods used
in this study to determine permeability are discussed.

3.1.1 Basic design

The rig is a high pressure high temperature triaxial deformation apparatus with a
servo-controlled fluid flow system (Figure 3.1). This rig is capable of
performing triaxial deformation experiments at pressures up to 250 MPa
(approximately equivalent to 10km depth in the Earth's crust) and temperatures
up to 250 º C. It must be noted that the temperature elements of the rig have not
been included as temperature was not used in any of the experiments. The rig
has an independent pore fluid system capable of measuring ultra-low
permeability (10-22 m2) and sample volume change (~0.1 mm3) during
deformation experiments.

76
 Confining Pore pressure
pressure gauge gauge
Pore fluid and axial
Pressure vessel load controllers

Transducer
amplifiers
Control PC

Pore pressure
generator

Axial loading
system

Figure 3.1 – The high pressure high temperature triaxial deformation apparatus with fluid
flow

The rig can accommodate samples up to 25mm in diameter and has a servo-
controlled ball-screw driven actuator with a 30 tonne loading capacity. The
internal force gauge design utilizes high performance materials (maraging steel
C300 with a ~2 GPa yield strength) to maximize the load measuring sensitivity.
A balanced piston design allows the full 30 tonne capacity to be directly applied
to the sample. The balanced piston can be bypassed to allow deformation in
extension in addition to compression.

The servo-controlled pore fluid system controls pore fluid pressure at the
upstream end of the sample and also serves as a high precision volumometer.
The downstream end of the sample is connected to a low volume reservoir. This
system can be used to measure volume change and permeability via the pore
pressure oscillation technique (Fischer 1992) and the pulse transient technique
(Brace et al. 1968).

77
 a b c

d e f

Figure 3.2 - Construction of the rig. a) Rig base and castors. b) and c) Shielding with angle
iron support and bottom plate. d) Valve system in construction on front panel. e) Gearbox
in place, with pressure vessel to be attached. f) Construction.

78
3.2 Equipment set-up and design

The apparatus shares a similar design to that of the ‘Nimonic’ deformation

apparatus at the Rock Deformation Laboratory, University of Manchester (see
Covey-Crump (1992) for details). A primary difference is that this rig is capable
of applying up to 30 tonnes of axial load. The rig comprises three main parts; the
pressure vessel (containing the sample assembly), force gauge block (containing
the force gauge) and the loading system (comprised of the ball screw, gear trains
and drive motor) as shown in Figure 3.3. Unlike some deformation apparatus,
the axial load is applied from the bottom of the axial column, not the top. The
rig will now be described in detail from top to bottom.

3.2.1 Pressure vessel and sample assembly

The apparatus has a pressure vessel made from Jessop Saville H.50 (AISI H.13)
and uses silicon oil as its confining medium, driven by a compressed air-driven
hydraulic pump. Silicon oil is used in preference to water as a confining medium
to minimise rust in certain elements of the system, and enhance the life of the
vessel. The pressure vessel sits at the top of the apparatus (Figure 3.3 and Figure
3.4), into which the sample assembly (Figure 3.5) is inserted via the top opening
(Figure 3.4c). The sample assembly is secured by the threaded top nut (Figure
3.6c & d) and is sealed by an o-ring and brass mitre ring configuration that seals
the gap between the upper sample assembly and the inside of the pressure vessel
(sealing methods are described in more detail in section 3.2.2). Confining
pressure is measured during experiments by an analogue gauge and a pressure
transducer (see 3.2.6). The vessel is designed to allow confining pressures that
can be maintained for long periods of up to 250 MPa, but has been tested up to
375 MPa. The sample assembly (Figure 3.5) consists of three main parts; the
upper and lower sample assemblies and the sample that is sandwiched between
the two. It is designed to accommodate samples of 50mm in length and 20mm
diameter, which is an initial length / diameter ratio of 2.5 (Paterson & Wong
2005). The sample is kept separate from the confining medium by placing it in a
PVC jacket, (Figure 3.5)

79
Figure 3.3 - The deformation apparatus, drawn approximately to a scale of 1:3.3

80
 a Pressure vessel b Sample assembly
Top nut

Force gauge
block

Anti-rotation pin

Pressure
Force gauge vessel
Balanced piston
LVDT cable
valve

Plumbing: valves
on this front
panel

Figure 3.4 - External view of the upper rig. a) Pressure vessel and force gauge block. b)
Sample assembly being inserted into pressure vessel. c) Rig and plumbing arrangement

81
The PVC tube seals against the o-rings on the upper and lower sample
assemblies, as the confining pressure is higher than pore pressure. The pore fluid
system uses water or high purity argon gas as the pore fluid, although only water
has been used in this study. When measuring low permeabilities, it is essential to
have minimal leaks within the pore-fluid system. The sample assembly was
therefore designed to have a minimal amount of connections. Pore fluid pressure
is introduced into the sample assembly via thin bore (0.5mm) high pressure pipe
(Figure 3.5 and Figure 3.6) connected to the upstream connector in the top of the
upper sample assembly (the pore fluid system is discussed in section 3.2.4). Pore
fluid passes through the upper sample assembly and is introduced into the top of
the sample which it flows through and exits through the lower sample assembly.
Pore fluid exits on the downstream side via thin bore pipe that is silver-soldered
into the base of the lower sample assembly, passed up around the sample in the
PVC jacket and threads through the upper sample assembly, secured by a
connector. Porous alumina spacers (Figure 3.6) are inserted on to the upstream
and downstream faces of the sample which ensures that there is an even
distribution of pore pressure across the face of the sample.

82
Figure 3.5 - The sample assembly. Note that photo varies slightly from schematic diagram
due to the photo being of an older assembly arrangement

83
 a Copper
jacket

Sample
Upper sample (steel blank Lower sample
assembly shown here) assembly

Porous spacers

b PVC
jacket

Downstream d
c
Top nut
Pressure vessel

Upstream
Nut spacer

Figure 3.6 - Sample assembly arrangement. a) Sample assembly, porous disks and copper
jacket. b) Sample asselmbly, porous disks and PVC jacket. c) Sample assemby in pressure
vessel, showing nut spacer and upstream and downstream thin-bore piping. d) Top nut

84
 3.2.2 Force gauge block (and force gauge)

The force gauge block was designed to house the force gauge column. The axial
load is applied from below by means of the ball screw (driven by the drive motor
and gearbox). The force gauge column protrudes out of the top of the force
gauge block (Figure 3.7) directly into the pressure vessel, where it connects to
the lower sample assembly via means of the force gauge extension (Figure 3.3).
The pressure vessel is connected to the force gauge block via 8 high-tensile bolts
which pass through the flange ring (tensile collar) that is screwed onto the
pressure vessel, into 8 threaded holes located around the top of the force gauge
block (Figure 3.3). The seal is a delta ring that placed between the faces (in
angled seats on the base of the pressure vessel and the top of the force gauge
block), and deforms elastically against the walls of its seat by the force that is
exerted from the eight bolts.

Figure 3.7 - The force gauge block and force gauge column arrangement

85
The axial load that is exerted on the sample is supported by the entire loading
column. Load that is applied to the sample is measured by measuring the
distortion within the force gauge column by means of a Linear Variable
Differential Transformer (LVDT) displacement transducer, in the form of an
internal load cell as shown in Figure 3.8. As shown in Figure 3.8 the LVDT is
forced firmly against the base of a T-section tube, which in turn is pushed against
a shoulder mid-way up the interior of the force gauge column above the level of
the moving piston seal. During axial loading any elastic distortion of the piston
below the top of the T-section tube is accommodated by the spring, and the
displacement measured by the LVDT is due entirely to the shortening of the
column between the top of the T-section tube and the top of the un-stressed rod
(i.e. the active gauge length, Figure 3.8). Hence, the actively deforming length
of the force gauge lies entirely between the pressure seals and no correction for
seal friction for force measurement is required. The LVDT was calibrated using
a load cell, as is discussed in section 3.3.1.4

Although the force gauge column, at is smallest width, is 19mm in diameter

(with a 3mm internal hole for the LVDT armature and T-tube), unlike the rest of
the force gauge assembly the force gauge column is made from maraging steel
C300, heat treated with a solution treatment at 1050°C and a 1 hour age at
485°C. Maraging steel is an iron nickel alloy which is known for possessing
superior strength (ultra-high yield strength up to 3.5 GPa) and toughness without
losing malleability.

86
Figure 3.8 - the force gauge assembly

87
The force gauge assembly has three pressure seals, and between the one at the
top of the force gauge column and the one on the bottom piston (Figure 3.7)
marks an un-pressurised area. This un-pressurised area allows the LVDT cable
to exit the force gauge block via a slot in the side of the force gauge LVDT
housing (Figure 3.8) and then through the force gauge block via the anti-rotation
pin slots. The anti-rotation pins screw into the bottom piston and pass out
through the anti-rotation pin slots. These pins are necessary as when load is
applied from below via the ball screw, it rotates as it screws upwards applying
the load. The anti-rotation pins stop the force gauge column from rotating above
the bottom piston, which if it were allowed, would shear the LVDT cable as well
as passing the rotation up towards the experimental sample. As shown in Figure
3.7, the two confining pressure inlets supply two isolated sealed off volumes.
The top pressurised area consists of the pressure vessel (sealed at the top by the
upper sample assembly seals) and the top of the force gauge block, sealed on the
force gauge column against the force gauge block. Confining pressure is
supplied to this area via the upper confining pressure inlet. The lower
pressurised area consists of the lower part of the force gauge block. It is sealed
at the top of the bottom piston against the force gauge block and below between
the bottom piston and the sealing spacer. Confining pressure is supplied to this
area via the lower confining pressure inlet. As the sample assembly is held down
in the pressure vessel by the top nut, the sealing spacer (that contains the lower-
most seal) is held in place by the bottom nut. The base of the bottom piston has a
hemispherical seat, so it can connect to the ball screw (described in the following
section) which also has an identical hemispherical seat; the two parts contact via
a hardened steel ball that sits within the seats. The ball / seat arrangement
ensures that concentricity of loading is applied and the loading column stays
central.

The confining pressure sealing arrangement may be understood with reference to

Figure 3.3. The seals within the force gauge block (Figure 3.7 and Figure 3.8)
consist of an o-ring of the required diameter and a brass mitre ring (or anti-
extrusion ring). When confining pressure is applied, the o-ring is deformed
elastically outwards e.g. against the inner surface of the force gauge block and
inwards against the force gauge assembly / piston, sealing the vertical flow of

88
confining medium. The surfaces are finely polished to ensure a good seal.
Conventional hydraulic seals are not useable at higher pressure (i.e. over 100
MPa), and in this case the brass mitre ring prevents the o-ring from being
extruded through the fine gap. During a test, the deformation of the seal applies
a large frictional force onto the piston, and piston movement can only commence
once the axial load is large enough to overcome these frictional forces.

As mentioned earlier, the system has a balanced piston arrangement. As can be

seen in Figure 3.7, there are two confining pressure inlets to the force gauge
block. There is a valve between these two inlets so that confining pressure can
be supplied to both, or isolated to just the upper inlet. By allowing the confining
pressure into both inlets, this allows the full capacity of the ball screw (30
tonnes) to be applied axially to the sample, rather than it having to support the
load applied to the piston by the confining pressure. By closing the valve, tests
can be conducted in extension, although no extension experiments have been
made in this study.

With the internal configuration described above, the entire piston can move ±
10mm from the centre position, and hence the rig can accommodate sample
shortening of up to 20mm.

89
3.2.3 Axial loading system

Load is applied via a servo-controlled axial displacement system. This consists

of a ball screw, that is driven by a gear train and drive motor (Figure 3.3 and
Figure 3.9). The ball screw (Figure 3.9c) is, in essence, a ball bearing leadscrew
that is nearly frictionless in operation, provided by high precision preloaded ball
bearings running on a helical raceway. The load between the screw and nut is
carried by ball bearings which provide the only contact between the nut and
screw. The ball screw assembly can operate with either the nut rotating around
the screw or the screw rotating through the nut, the latter being employed here.
The gearbox and gear train were designed and custom built in-house. The gear
train consists of four primary gear couples (Figure 3.9c); the largest gear is
attached to the base of the ball screw (secured by a key and keyway), and
connects to a smaller gear which gives a gear ratio of 5:1. On the same spindle
sits a wheel connecting to a worm to give a gear ratio of 40:1. On the same
spindle as the worm, sits a gear that couples with a smaller gear (that connects to
the primary gearbox) that gives a gear ratio of 2:1.

90
 a b

Drive Primary
motor gearbox

Limit switches
c
Ball
screw

2:1
Primary
Gearbox
120:1
40:1
(worm wheel) Axial displacement LVDT Ball screw base
5:1

Figure 3.9 - Axial loading system. a) Drive motor, gear train and ball screw. b) side on view of drive motor, gear train and ball screw. c) Ball
screw, gear train and displacement LVDT.

91
The primary gearbox, that connects to the drive motor via a custom-built
coupling plate, gives a gear ratio reduction of 120:1. The total gear ratio from
the drive motor to the ball screw is 48000:1. The drive motor (Figure 3.9a and c)
is capable of giving a max torque of 131 Ncm, which is increased by the gearbox
to 6288000 Ncm (6288
Nm). This figure
discounts the friction /
torque loss in the gear
train. The calculated
torque required (from
the ballscrew
manufacturer) is 531
Nm at full 30 tonne
load, although this
calculation does not
take into account piston
friction, but still
remains well within
torque range of the gear

Figure 3.10 - Axial load limit switches system. The gearbox

frame (Figure 3.9)
consists of a 15mm thick lower plate, attached to the main loading frame plate by
4 separator bars / spacers. The vertical shafts sit between these plates, with their
ends sitting in either iolite fittings or deep-groove ball bearings. The horizontal
axels are held in place by specially made steel holding blocks (which are bolted
onto the main loading frame plate). Each gear is locked onto its spindle with a
keyway and feather key combination. Axial displacement is measured by a
displacement LVDT that measures the displacement of the base of the ball screw
(Figure 3.9b). The limit switches (Figure 3.9b and Figure 3.10) are wired into
the axial load controller as hard limits.

This setup allows the piston and hence sample displacement to be controlled by
the servo motor, which is controlled by the computer via the motion control card

92
and axial load controller. This and the following is shown schematically in
Figure 3.11.

Figure 3.11 - Schematic diagram of the axial loading system

A piston displacement rate is sent as a command from within the main LabVIEW
control program (see section 3.2.6). This signal is sent via the National
Instruments Motion Control card, which is a PCI card that connects to connector
panel located outside the PC (Figure 3.12). The motion control card also
receives a milivolt feedback from the axial displacement LVDT (Figure 3.9c,
Figure 3.11 and Figure 3.12). The motion control card compares the two signals
i.e. the required displacement rate and the actual displacement measured by the
LVDT to produce an error signal, and then drives the motor (and hence gear train
and ball screw) to reduce this error to zero. The axial motor is a brushed servo,

93
and the axial displacement is controlled with the National Instruments Motion
Control Card, via user interface commands and axial LVDT feedback, using
standard PID control. The PID (proportional band – time integral – time
derivative) control controls the motor output and overshoot, and this is tuned by
modifying the PID parameters within the LabVIEW control software. The black
control box (Figure 3.15a) has buttons on the front panel that allow the axial load
to be manually controlled in compression and extension and at slow or fast
speeds.

Figure 3.12 - Motion control box that connects to the PCI motion
control card within the PC. Inputs include excitation voltage, axial
displacement LVDT feedback and output to the axial load controller
box

94
3.2.4 Pore fluid controller system

Pore fluid pressure in the system is generated by a NovaSwiss hand pump (see
Figure 3.1 and Figure 3.4c). Once the desired pressure has been reached, the
pore fluid system is isolated by shutting off valve 2 (see Figure 3.13). The pore
fluid pressure can then be controlled using the upstream pore fluid controller
(Figure 3.15). The upstream and/or the downstream reservoirs can be isolated
from each other by the use of valves 1 and 3. The primary technique that is used
for the determination of permeability in samples in this study is the pore pressure
oscillation technique (see section 3.4.2.1). This technique involves the creation of
a oscillating sinusoidal pressure wave on one side of the sample. It is the pore
fluid controller that generates this pressure wave. The pore fluid controller
consists of a servo controlled NovaSwiss 5cc pump that is attached directly to the
upstream reservoir plumbing. This setup allows the pore fluid pressure to be

Figure 3.13 - Schematic diagram of the experimental apparatus used for the measurement
of permeability.
controlled via a servo-driven pump (Figure 3.15b,c & d) which allows precise
pressure changes with computer control. This is shown schematically in Figure

95
3.13. A computer generated sinusoidal signal is sent via the RS232 serial
interface to a Eurotherm 2604 controller. The upstream transducer (Figure
3.15b) provides a millivolt feedback, which is then amplified through the brown
box seen in Figure 3.17a) and the Eurotherm controller uses standard PID control
to control the output to the motor. The end result is that the Eurotherm controller
makes the upstream fluid pressure follow the sinusoidal signal output from the
PC causing a sinusoidal pressure wave in the upstream end of the sample. The
system must be tuned to control the motor output and overshoot, and this is done
by modifying the PID parameters (proportional band – time integral – time
derivative) on the Eurotherm controller. Pore fluid pressure can be manually
controlled at varying speeds by controls on the front panel.

96
The upstream and downstream pore pressures are measured by RDP pressure
transducers, as shown in Figure 3.14 and their outputs are amplified through
twin-channel DC RDP 611 amplifier modules (Figure 3.17a). An LVDT
measures the displacement of the NovaSwiss pump piston (Figure 3.15c, e).
This is then used to calculate the change in pore volume as the pump piston
moves. The pore fluid control system (motor, gearbox and pump units) is shown
in Figure 3.15c.

Upstream to sample To pore pressure To confining

gauge pressure gauge
Downstream from sample

Downstream
pressure
transducer
To pore fluid Upstream
controller pressure
transducer
Confining
pressure
To confining
pump
pressure
pump

Pore fluid
pressure
generator

Figure 3.14 - Pore fluid plumbing, including pressure transducers

97
 a b Drive motor

Servo amp

Eurotherm Pore fluid

controller

Axial load controller Pore fluid pressure transducer

c Drive motor LVDT d

Heat sink

1:1 gears
Gearbox Limit
120:1 switch

To upstream pore
pressure plumbing Pump clamp Pump
e Limit switches
Limit switch
contactor

LVDT LVDT Bolt connected to Pump

armature piston
Figure 3.15 - Pore fluid control system. a) Pore fluid control box. b) View of pore fluid control
box circuitry, connected to the pore fluid controller drive motor. c) Pore fluid pump, gearbox
and drive motor. d) Rear view of pore fluid control system. e) LVDT displacement transducer

98
3.2.5 Confining pressure system

Confining pressure is generated by use of an air-driven SC hydraulic pump

(Figure 3.16) using low viscosity silicon oil. Once the desired confining pressure
has been reached, valve 5 in Figure 3.13 can be closed to isolate the pump from
the confining pressure system. Confining pressure is measured by a RDP Z-type
DC transducer.

FRL
Airline in

Airline

Air driven Confining

hydraulic medium inlet
pump

High pressure confining

medium out

Figure 3.16 - Air-driven confining pressure pump. The air flow into the pump is
controlled by a manual valve on the front of the rig, and air passess through a
Filter Regulator Lubricator (FRL) that is designed to remove compressor
lubricants, dirt and water from the air stream

99
3.2.6 Transducers, data logging system and computer control

All data output from both pressure and displacement transducers is logged by a
computer during experiments. The data is logged using a National Instruments
Data Acquisition device (NI-DAQ PCI card in the PC), connected to a data
logging board with BNC connectors (Figure 3.17).

a b

Figure 3.17 - a) NI-DAQ data board, motion control box and amplifiers. b) NI-DAQ
board with BNC connections for amplified transducer outputs

The LabVIEW logging program communicates with this device to record

pressure and displacement data. The transducer outputs connect to this data
board. However, the maximum output ranges of the transducers is in the
millivolt range, so before they connect to the logging board they are first
amplified, as the maximum input range of the NI-DAQ board is -10V to 10V
range. This provides a much bigger range and greater sensitivity. For example,
the force gauge LVDT has a sensitivity of 158.03 mV / V / mm. If the maximum
displacement range of the LVDT were to be used (5mm) with an excitation
voltage of 5V, then the max output range would be ~3.95V, which is only 1/5th of
the range that the NI-DAQ board can measure. In practice, the force gauge
LVDT will be required to detect much smaller variations in displacement
(micron changes in force gauge length). Hence, by amplifying the signal, the
output voltage can amplified up to the -10V to 10V range.

100
A screenshot of the triaxial logging program written in LabVIEW 7.2 can be seen
in Figure 3.18, although the program will not be discussed in any detail here.
The program controls all parts of the system, logs data from all of the transducers
via the NI-DAQ card, applies the calibrations and outputs the results of upstream
and downstream pore pressures, confining pressure, force, displacement and
volume graphically and numerically on the front panel. It also controls the
upstream pressure generator, creating the oscillating sinusoidal pressure wave
required to measure permeability via the pore pressure oscillation technique (see
section 3.4.2.1). The wave amplitude and wave time period can be easily
modified at the touch of a button. The displacement rate of the loading piston
can also be controlled in the software, as the program controls the Motion
Control devices (see section 3.2.3).

Figure 3.18 - Screenshot of LabVIEW Triaxial Logging Program.

101
3.2.7 Preparation of samples for testing

The samples were cored from samples of rock to a nominal diameter of ~20 mm
using a diamond-tipped coring drill (Figure 3.19a). The machine is configured
for minimal vibration so as to reduce the breakage of cores, and using a water
swivel that uses water as coolant while cutting. The ends of the cores were cut to
approximate length using the diamond blade saw in Figure 3.19c. Samples were
squared using a grinder (Figure 3.19b) so that the ends were squared to the

a b

c d V-block

Magnetic clamp

Figure 3.19 - Sample preparation equipment. a) Coring drill. b) Grinder. c) Diamond

blade saw. d) V-block on magnetic clamp of grinder

102
sample long axis within a tolerance of 0.01mm or less. The sample length to
diameter ratio was approximately 2-2.5:1, as suggested in Paterson & Wong
(2005). Samples with this dimension ratio ensure the stress and strain in the
central portion of the plug are not affected by stress shadows. (Paterson & Wong
2005) discussed various arrangements for reducing the effect of the stress
shadow, such as placing a low friction spacer between the sample and loading
platens, matching the elastic properties of the platen to the sample, or varying the
ratio of the diameter of the sample to the loading platen. They conclude that the
most reliable method is to use loading platens of the same diameter as the
sample, without any type of spacer. This approach was adopted in these
experiments, as discussed in section 3.2.1.

3.2.8 Summary

This section has so far summarised the primary parts of the apparatus. It must be
made clear that many of the parts of the apparatus have gone unmentioned due to
space constraints. As many parts of the apparatus were not necessarily designed
to fit together, things such as connectors, adaptors, coupling plates, gearboxes,
spindle holding blocks, LVDT and pump clamps etc. had to be made in house.

The loading frame of the rig was constructed with a frame of 9mm thick angle
iron. Shielding walls made from mild sheet steel (5mm and 9mm in thickness)
were then bolted to the frame. The main loading plate that bears the weight of
the machine is 25mm thick.

103
3.3 Summary of potential sources of error and calibration

In addition to errors that are introduced into the permeability measurements by

factors such as transducer accuracy and length measurements that have been
mentioned so far, there are other possible factors that may cause values obtained
for permeability to be erroneous. There are various types of problem that can
have a negative effect on the quality of the data obtained with experiments. As
discussed earlier there are the technical problems associated with the need to
maintain confining pressure whilst keeping the confining medium out of the
sample. Therefore the true axial displacement of the sample being deformed
must be determined remotely by measuring external piston displacement and
subtracting the predetermined distortion of the loading column, hence
introducing calibration errors. These problems are unavoidable, but their
significance can be reduced by careful procedure, accurate and repeatable
calibrations (described in the next section). A degree of error is also introduced
into the determination of the differential load experienced by the specimen
because of the presence of the PVC jacket around the sample, which can in fact
support some of the load as elastomers can become quite strong and apply
appreciable constraint at higher confining pressures (commonly 200-500 MPa)
(Paterson 1964, Weaver & Paterson 1969a, 1969b).

As Faulkner (1997) discussed, the sample jacket (PVC in this case) has a finite
strength. When confining pressure is increased, there can be differential
compaction between the sample and the jacket, although this is generally more
applicable in gouge experiments where there is a much larger degree of sample
compaction. If this happens, ‘crinkling’ of the copper jacket could be seen, and
high permeability pathways can be created along the length of the sample, which
could lead to an overestimation of the sample permeability. Tests were
completed on an impermeable steel blank encased in the PVC jacket, and it was
show that within experimental error, fluid communication was not evident at
various effective confining pressures up to 200 MPa. The PVC jacket was thus
acting as a compliant material.

104
The second types of problems are associated with assumptions made about the
specimen deformation behaviour which can become important during the data
interpretation stage. For example, for convenience of data reduction and
parameter fitting, it might be assumed that the deformation through the sample is
homogenous. Also, that any variability in confining pressure and pore pressure
between tests under nominally the same condition with respect to those variables
is insignificant. These problems are avoidable, but if the approximations can be
made, they simplify the data analysis considerably.

So far some of the main potential sources of error that can arise from the
experimental measurements have been discussed. The calibrations that are
necessary in order to be able to process the pressure and displacement data and
obtain values for permeability and stress/strain will now be described.

105
3.3.1 Calibrations

3.3.1.1 Confining pressure and pore fluid pressure

The pressure transducers output raw millivolt values, that are in turn amplified to
a -10V – 10V range by the amplifier units, as mentioned in section 3.2.6. The
output voltage of the pressure transducers were calibrated against the Budenburg
pressure gauge at intervals of 10 MPa in the range of 0-200 MPa. It is important
to repeat these calibrations, and record values for pressure going up as well as
down, so that any hysteresis can be identified. As can be seen in the graph
shown in Appendix A, the dependences show excellent linearity, with a least
squares fits with R2 values of > 0.9998, and the gradients show a standard
deviation of 0.099.

3.3.1.2 Downstream storage capacity

To calculate the permeability using the pore pressure oscillation method (see
3.4.2.1), the downstream storage capacity (Bd) must be known. This storage
capacity is defined, at a given fluid density, as the increase in volume of pore
fluid stored per unit increase in pore fluid pressure (m3/Pa). This differs from the
sample storage capacity (an unknown parameter of the sample βc calculated
along with permeability k using the Fischer method) as it is measured relative to
the specific volume of the downstream reservoir, whereas the sample storage
capacity is expressed per unit volume, and hence as units Pa-1.

The downstream storage capacity therefore varies as a function of pore fluid

pressure due to the compressibility of pore fluid, and it was therefore calculated
at a variety of pore fluid pressures. This was done using the volumometer over a
small increment of pressure. This is done twice, once with the valve connecting
the downstream (valve 1, Figure 3.13) reservoir open (Bd + Bu, where Bu =

106
 upstream storage capacity) and once with the valve closed (Bu - isolating the
downstream reservoir). When completing these measurements, a steel blank was
used instead of a sample to ensure that there was no fluid communication
between the upstream and downstream reservoirs, and no sample volume was
measured. By subtracting the storage capacity of the upstream reservoir (Bu)
from the storage capacity of the downstream and upstream reservoirs combined
(Bd + Bu), the downstream storage capacity can be calculated. Using this
technique, the downstream storage capacity was calculated for pore pressures
from 20 MPa to 110 MPa, in 10 MPa steps, and the results are shown in Figure
3.20. Using the equation of best-fit line from this graph, the downstream storage
capacity can be calculated for any pore fluid pressure in the range.

Downstream Storage Capacity

1.4E-15

1.3E-15
Downstream Storage Capacity m^3 Pa^-1

1.2E-15

1.1E-15
-0.1929
y = 2E-15x

1E-15

9E-16

8E-16
0 20 40 60 80 100 120
Pore pressure (MPa)

Figure 3.20 - Downstream storage capacity values for water, calculated by the method
described in the text

107
3.3.1.3 Volumometer LVDT

To calculate the volume of fluid displaced by the volumometer (pore fluid

controller pump), the known surface area of the piston face must be combined
with the displacement of the piston. The piston displacement is measured by the
LDVT attached to the pore fluid controller pump, and it is calibrated using a
displacement dial gauge. The output of the LVDT was calibrated against the
displacement dial gauge at intervals of ~0.2 mm in the range of 0-3 mm. A
graph of LVDT voltage vs. volume of displaced fluid (the surface area of the
piston as a function of the displacement) can be plotted. The graph in Appendix
A shows that the dependence appears to show good linearity with an R2 value of
0.9997, and displaying no hysteresis between the piston in and piston out.

Volumometer data are very important, as not only do they give an indication into
the change in porosity and dilatancy within a sample, but it can also be used to
estimate volumetric strain (assuming all pore space is connected and not
accounting for the compressibility of the constituent minerals in the rock). When
the rock becomes more permeable, by measuring the amount of fluid moving in
or out of the saturated sample during the deformation gives a measure of the
volumetric strain associated with the accessible pores or cracks. After correction
for elastic changes in pore volume, the dilatancy can then be obtained.

108
3.3.1.4 Force gauge calibration

As discussed in section 3.2.2, the force gauge displacement LVDT reading was
used to measure the load applied in the loading column by measuring the elastic
change in length of the force gauge column. The force gauge needs to be
calibrated very precisely as it is required to measure small changes in force. It
was calibrated using an external load cell. The load cell was mounted externally
on the rig using the calibration setup shown in Figure 3.21. The entire sample
assembly was removed from the pressure vessel, and in place of the top nut, the
force gauge calibration plug was inserted. This is similar to the top nut, but
extends further out of the top of the pressure vessel so that the load cell flange
ring can also be screw onto it. The load cell sits sandwiched between the load
cell flange ring and the load cell holder, and then fastened together by 8 high
tensile bolts that pass through the flange ring and screw into the load cell holder.
Care was taken to tighten each of these bolts with a similar amount of torque so
that upon loading there is no uneven load. The load was transmitted from the
force gauge extension to the load cell via the force gauge calibration rod that
passess down through the pressure vessel.

The output voltage of the force gauge LVDT was calibrated against the output
voltage of the load cell (which is converted into a force by a calibrated function
supplied by the load cell manufacturer) by loading from 0-30 tons and back
down while continuously logging both outputs in the triaxial logging program on
the PC. The results of the calibration are shown in Appendix A. It must be
noted that this calibration was completed without confining pressure, as there is
no sealing arrangement built into the calibration headstock.

109
Figure 3.21 - Schematic diagram illustrating the force gauge calibration setup that is attached
to the top of the pressure vessel

110
3.3.1.5 Elasticity of the loading column

In order to gain correct results for the sample strain during experiments, the
measured axial displacement measurement must be corrected for the elastic
distortion of the loading column. The loading column is cumulatively long in
comparison to the sample and the axial load LVDT is measuring both the strain
in the sample as well as the strain in the entire loading column. This elastic
distortion is proportional to the force and the confining pressure. The loading
column stiffness was determined at twelve equally spaced confining pressures in
the range 0-142 MPa at room temperature, by loading from 0-10 tons with a steel
blank (of known length, diameter and elastic properties) in place of the sample.
At each pressure at least 20 readings were taken, and for each loading the output
of the force gauge was recorded. The results are shown in Figure 3.22, and
shows that force gauge stiffness (gradient of each fit) decreases with increasing
confining pressure.

1.0 rig distorsion

0.8

0.6
distorsion, mm

0.4

MPa 35
MPa 78
MPa 49
MPa 110
0.2 MPa 125
MPa 142
MPa 24
MPa 101
MPa 49
MPa 104.7
MPa 0.1
Calculated distortion of steel blank

0.0
0 50 100 150 200 250 300 350
differential stress, M Pa

Figure 3.22 - Elasticity of the loading column for various confining pressures

111
By applying best line linear fits to each of the traces, the gradients (stiffness) of
each of the best fit lines were then fitted as a function of confining pressure
(Figure 3.23).The theoretical distortion of the blank is calculated and subtracted
from this curve. The stiffness of the loading column with zero confining
pressure is 8.43 x 10-3 mm/kN.

9.0

8.5
slope distorsion, micron/kN

8.0 y = 3E-05x 2 - 0.0128x + 8.4785

R2 = 0.7304

7.5

7.0
0 20 40 60 80 100 120 140 160
confining P, MPa

Figure 3.23 – Gradients (stiffnesses) of the fits from Figure 3.22 plotted against
confining pressure.

112
3.3.1.6 Axial displacement transducer calibration

The axial displacement transducer output in Volts must be calibrated against

absolute displacement measured using a dial gauge held in a magnetic clamp.
The clamp can be attached to the bottom of the gearbox, and the dial gauge
aligned so that the moving armature of the dial gauge rests against the bottom of
the ballscrew. The output of the LVDT was calibrated against the displacement
dial gauge at intervals of ~0.2 mm in the range of 0-3 mm. A graph of LVDT
voltage versus axial displacement (mm) can be plotted. This calibration is shown
in Appendix A and shows that the dependence appears to show good linearity
with an R2 value of 0.9997, and displaying no hysteresis between the ballscrew
in extension and in compression.

3.3.1.7 Load on the force gauge from confining pressure

Once the force gauge LVDT has been calibrated and confining pressure (σ3) is
increased (without axial load - σ1), a force is exerted on the forge gauge. It is
important therefore to know how this force applied varies with confining
pressure, so that it can be subtracted to calculate differential load (σ1 -σ3). The
load applied to the force gauge was calibrated against the confining pressure at
intervals of 2 MPa in the range of 0-200 MPa, both going up pressure and down
pressure. The results can be seen in in Appendix A and can be seen to show
excellent linearity.

113
3.3.1.8 Thermal expansion effects on the force gauge

Following the calibration of the force gauge, it was apparent that between
experiments, the force gauge zero reading appeared to wander by +- 1-3kN,
above and below zero. After checking seals for sticking problems within the
loading column, it became apparent that there maybe other factors such as
fluctuations with room temperature may be the cause. The force gauge LVDT is
very sensitive with a post-amplification sensitivity of 0.014 mV/N; if 100 grams
in weight is added to the force gauge this results in a measurable change of 0.014
mV. As a result, tiny changes in the length of the force gauge due to thermal
expansion (the tendency of matter to increase in volume or pressure when
heated) could be registered. Thermal expansion typically causes solids to expand
in response to heating and contract on cooling; this response to temperature
change is expressed as its coefficient of thermal expansion. The linear thermal
expansion coefficient relates the change in temperature to the change in a
material's linear dimensions. It is the fractional change in length of a bar per
degree of temperature change, and is as follows:

1 ⎛ ∂L ⎞
α= ⎜ ⎟ or ∂L = Lα (∂T ) 3.1
L ⎝ ∂T ⎠

where α = linear thermal expansion coefficient,

L = unit change in length
T= temperature.

As the active length of the force gauge is 40mm, a 5ºC change in temperature
would result in a change in length of 0.00226 mm (assuming a typical thermal
expansion coefficient for maraging steel of 11.3 x 10-6 at 20ºC). By using this
change in length to calculate strain, the stress that would be required to create
such a change in length in the force gauge can be calculated using:

114
 ∂L
σ = Eε and ε = 3.2
L0

where σ = tensile stress (Pa)

Ε = Young’s Modulus of steel (Pa)
ε = strain
L = length (m)

Therefore the stress required to create a 0.00226 mm change in length is 11.018

MPa. By combining this stress with the surface area of the force gauge with
19mm outer diameter and a 3mm internal hole (0.00027646 m2) the force that
would act over this area is calculated using:

F = σA 3.3

where F = force (N)

σ = stress (Pa)
A = cross-sectional area (m)

which gives a force of 3045 N. Therefore, the force required to create a

0.00226mm change in length of the force gauge caused by a 5ºC change in
temperature due to thermal expansion is 3045N. This value compares well to the
observed force gauge zero offset wander of 1-3kN. It is therefore very important
to maintain a constant room temperature during experiments.

115
3.4 The measurement of permeability

This section aims to introduce the related equations that describe the flow of
compressible fluids through porous media, and are required to determine
permeability of a sample. The two main methods that can be used to measure
permeability are discussed, focusing primarily on the pore pressure oscillation
technique that this study employed as the primary method. The data processing
techniques, benefits and limitations with relevance to the experimental
equipment that has been previously described, and an analysis of errors are
discussed.

3.4.1 The diffusion equation for the flow of compressible fluids

through porous media

(Faulkner 1997) showed that Darcy’s law is a useful description for the flow of
fluids through porous media under steady flow conditions. However, a more
general description of the flow of fluids through porous media may be found in
the diffusion equation (Carslaw & Jaeger 1959), as summarized by (Faulkner
1997).

Compressible fluid flowing through a porous medium with a small matrix

compressibility is analogous to the conduction of heat through solid bodies. On
this basis, equations used to describe heat flow can similarly be used in the
analysis of fluid flow. The diffusion equation for fluid flow (derived from the
heat flow equation, see (Carslaw & Jaeger 1959)) is a second order partial
differential equation, and states that for all values of x, and for t < 0,

∂p ∂2 p
=κ 2 3.4
∂t ∂x

where κ = hydraulic diffusivity.

116
The hydraulic diffusivity, κ, of the rock is defined as the hydraulic conductivity,
K, divided by the storage capacity, βc:

K
κ= 3.5
βc

The storage capacity, βc, of a rock is defined as the additional volume of fluid
stored a unit volume of rock per unit rise in fluid pressure (Pa-1). (Faulkner
1997) stated that without the effect of storage, which arise from the elastic
compressibility of both fluid and rock, any increases in fluid pressure would be
transmitted with infinite velocity through the porous medium. (Brace et al. 1968)
expressed the storage capacity of rock as:

β c = β b + φβ f − (1 + φ )β m 3.6

Where βb = bulk compressibility of the porous medium, Pa-1

Βf = fluid compressibility, Pa-1
Βm = compressibility of the constituent minerals, Pa-1
= porosity

The hydraulic conductivity, K, in Darcy’s law is specific to the fluid flowing

through the porous medium, and the hydraulic conductivity may be divided thus
k
K= 3.7
η

Where k = permeability, m2
η = fluid viscosity, Pa.s.

From equations 3.5 and 3.7, hydraulic diffusivity can be seen to be related to the
permeability, k, the storage capacity βc, and the fluid viscosity, η, by:

k
κ= 3.8
β cη

117
If permeability is measured in m2, storage capacity in Pa-1 and fluid viscosity in
Pa.s, hydraulic diffusivity will have units of m2/s and dimensions of L2T-1, which
are the same dimensions of the analogous quantity, thermal diffusivity (Carslaw
& Jaeger 1959), pg 3.

To obtain values for the permeability of a porous medium, a complete solution to

the diffusion equation must be obtained, which requires knowledge of the storage
capacity of the rock, the compressibility and the viscosity of the permeating
fluid, the time, t, required for propagation of a given fluid pressure, p, over a
distance, x.

Simple solutions for equation 3.4 invariably involve the sum of a number of
exponential terms, and it is convenient to express these in terms of Gauss’s error
function:

2
erf ( x ) = x
π ∫ e ξ dξ
2
3.9
0

Where ξ = dimensionless variable, x/l, where –l < x < 1 (Carslaw &

Jaeger 1959) p.24.

The error function has the properties of erf(0) = 0, erf( ∞ ) = 1 and erf(-x) = -
erf(x). The expression erfc(x) may be defined as:

erfc( x ) = 1 − erf ( x ) 3.10

Thus, in the case of a semi-infinite porous medium, with homogenous physical

properties and x = 0, defined as our planar interface where there is a sudden
pressure increase at time, t, that is maintained at x = 0, then the solution for the
change in pressure at any distance, x, into the material, at time, t, is:

118
 ⎛ x ⎞
Δp( x ,t ) = erfc⎜ ⎟ 3.11
⎝ 2 κt ⎠

after (Carslaw & Jaeger 1959), p. 60. In using this analysis, the value for κ must
have been previously determined, involving the knowledge of the permeability
and the storage capacity of the material. Therefore, attention is now focused
employed for the experimental determination of these parameters.

119
3.4.2 Methods for measuring permeability in low permeability
rocks

The two most common methods of measuring the permeability of tight rocks are:

1. the pulse transient method / transient step method

2. the pore pressure oscillation technique

(Faulkner 1997) gave a detailed discussion assessing these two techniques for the
measurement of permeability, and here these techniques are summarised. The
pulse transient method was first introduced and developed by Brace et al. (1968)
as it is technically much easier to measure changes in pressure than flow rate in
low-permeability experiments. In the method, two reservoirs are connected
through a sample. A sudden increase of pressure is induced in one of the
reservoirs causing an instantaneous pore pressure gradient across the sample.
The pore pressure gradient decays as fluid flows through the sample and builds
up pressure on the low pressure side; the system is then allowed to equilibrate to
a new pressure level, which is determined by the magnitude of the step, the
storage capacity of the sample, and the storage capacities of the reservoirs. By
measuring the decay of the pore pressure gradient on both sides of the sample
with time, the permeability of the sample can be deduced graphically. Zoback &
Byerlee (1975) showed an example of how this technique can be used to measure
permeability under differential load. A major drawback to the pulse transient
method is that it requires the axial load to be stopped so that permeability can be
measured, which introduces various time dependent effects. In this study, the
pore pressure oscillation technique was used as the permeability and porosity
evolution can be measured during progressive deformation without stopping
axial loading. A brief overview of the technique will now be discussed in this
section.

120
3.4.2.1 The pore pressure oscillation technique.

The pore pressure oscillation technique was developed concurrently by (Kranz et

al. 1990) and (Fischer 1992). The primary aim was to be able to measure both
permeability, k, and storage capacity, S, simultaneously during the course of a
deformation experiment. The technique utilises the attenuation (amplitude shift)
and phase shift of a forced sinusoidal pressure wave created in the upstream
reservoir as it is transmitted through the sample to the downstream reservoir.
Two dimensionless parameters are used to calculate k and S (Fischer 1992). The
process by which this sinusoidal pressure wave is created has already been
discussed in section 3.2.4.

The pore pressure oscillation technique was presented and discussed in detail in
various publications (Bernabe et al. 2006, Faulkner 1997, Fischer 1992, Kranz et
al. 1990), and so a brief overview is provided here. A summary of the
mathematical formulation of the solution to the diffusion equation is given with
the imposed boundary conditions and description of the processing of the data.
Recent modifications by Bernabe et al. (2006) are also summarised, as the
program used to process all data contained in this work uses both the Fischer
(1992) and Bernabe (2006) method to output results for permeability. Hence,
both are now described.

3.4.2.1.1 Fischer and Krantz solution

Faulkner (1997) showed how the general solutions of Fischer (1992) can be used
as a starting point to be developed into useable equations so that permeability can
be determined by using the measured experimental parameters. Although both
Fischer (1992) and Kranz et al. (1990) use similar analytical approaches,
Faulkner (1997) started with the solution to the diffusion equation derived by
Fischer (1992). Both techniques yield the same values for permeability and
storage capacity using identical experimental parameters (Faulkner 1997). The
mathematical process by which Fischer (1992) and Kranz et al. (1990) arrived at
their general solution to the diffusion equation (along with the appropriate

121
boundary conditions) are not discussed in Faulkner (1997) or here, but for further
information one is referred to the papers and appendices of this literature. This
section summarises the main parts of the manipulation of the equations discussed
in Faulkner (1997).

The measurement system for this technique is shown in Figure 3.24, which is a
simplified version of Figure 3.13 for simplicity of describing the technique. As
discussed in previous sections, the jacketed samples sits within the pressure
vessel, connected to the upstream and downstream pressure reservoirs that are set
at the desired pressure conditions. During the test, the upstream and downstream
reservoirs are isolated from each other by the valve shown Figure 3.24
(equivalent of valve 3 in Figure 3.13).

volumometer
and control

P1 (u)
pressure
Valve
measurements
Confining
Pressure P2 (d)
Pc PVC jacket

Sample

Porous spacer

Axial compression

Figure 3.24 - Schematic diagram illustrating the experimental sample assembly, simplified from
Figure 3.13.

122
The upstream and downstream pressure transducers allow independent,
continuous measurement of the fluid pressure in both reservoirs. Firstly, a forced
sinusoidal oscillation of the upstream fluid pressure P1 was created and the
sinusoidal pressure oscillation is monitored in the downstream reservoir P2; the
two are related by:

P2 = χP1 sin(ωt − θ ) 3.12

Where P2 = pressure in the downstream reservoir, Pa

χ = amplitude ratio of the downstream and upstream pressure
oscillations
P1 = pressure in the upstream reservoir, Pa
2π
ω= , where T = period of oscillation
T
θ = phase difference between the upstream and downstream
pressure oscillations, in radians

This oscillation is created using the pore fluid controller (section 3.2.4), a servo
controlled pump. The pressure pulse passess through the sample and the
resulting pressure variations in the downstream reservoir P2 are logged on the
computer.

As Faulkner (1997) summarised, the only unknowns in Fischer’s equation 3.12

above are χ and θ, and following the solution of Fischer (1992), they may be
related to the permeability, k and the storage capacity, βc of a sample by the
modulus (χ) and the argument (θ) of the complex number D:

Bd iωη
v(1 + i ) cosh[vx(1 + i )] + sinh[vx(1 + i )]
D= kA
B iωη 3.13
v(1 + i ) cosh[vL(1 + i )] + d sinh[vL(1 + i )]
kA

where

123
 β cπη
v= 3.14
Tk
L = Length of the sample (metres)
Bd = Downstream storage capacity (m3 Pa-1)
A = cross sectional area of sample (m2)
η = fluid viscosity, Pa.s
T = period of oscillation, s

At this stage, these equations can be simplified by introducing two dimensionless

parameters, γ and ψ, where

Bd
γ = 3.15
ALβ c

and

1
⎛ πηβ c ⎞ 2
ψ = L⎜ ⎟ 3.16
⎝ Tk ⎠

Fischer (1992) merged the equations 3.15 and 3.16 with the complex number
equation 3.13, then expanded the denominator with the hyperbolic function with
the real and imaginary components separated. Faulkner (1997) showed that the
real and imaginary parts of the complex number (equation 3.13) can be re-
expressed and thought of in terms of an Argand diagram, finding the modulus
and argument using Pythagoras’ theorem and trigonometry so that the amplitude
ratio χ, and the phase shift θ, are expressed in terms of the two dimensionless
parameters, γ and ψ. The reader is referred to Faulkner (1997) for this method
fully explained.

In the next stage of the analysis, Faulkner (1997) manipulated the equations so
that the amplitude ratio, χ, and one of the dimensionless parameters, γ, are both
expressed in terms of the other dimensionless parameter ψ, and the phase shift, θ.

124
The values for the amplitude ratio, χ, and the phase shift, θ, can be determined
experimentally, and hence ψ is the only unknown in the equation. The
manipulation results in obtaining the amplitude ratio, χ, in terms of only ψ and θ,
as follows:

1
χ= 2
⎛ R + P tan θ ⎞ ⎛ R + P tan θ ⎞
( P + R ) − 2( PQ + RS ).⎜⎜
2 2

θ
(
⎟⎟ + Q 2 + S 2 )
.⎜⎜ ⎟⎟ 3.17
⎝ S + Q tan ⎠ ⎝ S + Q tan θ ⎠

where

P = coshψ cosψ
Q = ψ (sinh ψ cosψ − cosh ψ sin ψ )
R = sinh ψ sin ψ
S = ψ (cosh ψ sin ψ + sinh ψ cosψ )

The equation can now by solved by providing the value for the phase shift, θ
(determined experimentally) then iterating for the amplitude, χ (also determined
experimentally) by substituting the values for the dimensionless parameter ψ into
equation 3.17. It must be noted that more than one value of ψ can satisfy
equation 3.17. The reader is referred to Faulkner (1997) to how this is dealt with,
as the iterated value for 3.17 is used to compute γ from an equation not
mentioned here, and both these values need to be used to evaluate the phase shift
θ, which must be compared to the experimentally determined value for θ.

After converting the phase shift in radians, rather than cycles, the values for γ and
ψ are plotted in δ and χ space, the graph shown in Figure 3.25 is obtained, as
seen in (Fischer 1992).

125
 0.00

-0.50

-1.00 0.03
0.1
-1.50 0.3
1
Log (amplitude ratio)

-2.00
3

ψ=0.25
-2.50
10

0.5
-3.00
30

0.75
1.0
-3.50

1.25
100

1.5
1.75
-4.00
300

2.0

2.25

2.5
-4.50

2.75
γ=1000

3.0
-5.00
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Phase lag (cycles)

Figure 3.25 - Graph showing the two dimensionless parameters, γ and ψ used in the
analysis developed by (Fischer 1992) plotted as functions of the amplitude ratio, χ, and
the phase shift, δ. From(Faulkner 1997) .

Finally, permeability, k, and storage capacity, βc may be calculated from ψ and γ

using:

Bd
βc = 3.18
LAγ

πηBd L
k= 3.19
TAψ 2 γ

Where:

L = Length of the sample (metres)

Bd = Downstream storage capacity (m3 Pa-1)
A = cross sectional area of sample (m2)
η = fluid viscosity, Pa.s

126
T = period of oscillation, s

There are some limitations using the Fischer method for calculating permeability.
The dimensionless parameter ψ depends on the storage capacity Bc and
permeability k, therefore it does not represent independent material values and
hence is not a true dimensionless permeability. In addition to this, at high values
of amplitude ratio χ , values for permeability cannot be returned because the non-
linearity produces an extreme sensitivity to noise (e.g. small variations in
amplitude ratio when in the top left area of the plot in Figure 3.25 can lead to
non-physical values). These problems have never been explained or addressed
until Bernabe et al. (2006).

127
3.4.2.1.2 Bernabe’s solution

More recently Bernabe et al. (2006) have re-analysed the oscillating flow method
described in the last chapter, as well as determining the uncertainty on the
permeability and storativity, assessing the effect of transients, and deriving the
approximate solutions in the end-member cases (i.e. very small or very large
sample storage). The key elements of this technique will now be described, as
the data processing program also calculates permeability using this technique.

To determine permeability and storativity, Bernabe et al. (2006) used a different

approach to derive equations with two different independent dimensionless
parameters for storativity, ξ , and for permeability, η which have independent
material properties:

STk
η= 3.20
πLμβ D

and

SLβ
ξ= 3.21
βD

where: S = sample cross-sectional area, m2

L = sample length, m
β = sample storativity
βD = downstream storage capacity (m3 Pa-1)
k = sample permeability (m2)
T = sinusoidal wave period (seconds)
μ = fluid viscosity (Pa.s)

Their approach differs in that the end product after mathematical analysis (reader
is referred to Bernabe et al. (2006)) outputs an amplitude ratio A which is a ratio

128
between the downstream to upstream amplitudes. This can then be related to the
dimensionless permeability parameter, η, by

2A
η= 3.22
1 − A2

where A = amplitude ratio.

Permeability can then be calculated using the amplitude ratio and the period, T,
in conjunction with the sample and pore fluid constants also used in the Fischer
(1992) technique by combining equations 3.20 and 3.22 and rearranging for k.

129
 3.4.2.2 Processing techniques

The data that are collected during an experiment require considerable processing
so that permeability values may be obtained. This section briefly outlines the
main processing techniques that are incorporated into the processing program
written by Faulkner (1997). The program is designed specifically for the purpose
of processing permeability data. Faulkner (1997) details these techniques used
within the program, so only a summary is provided here.

Figure 3.26 shows results from a typical data file, with the forced sinusoidal
wave in the upstream reservoir pressure signal and the resultant pressure wave
that is measured on the downstream reservoir pressure signal. The data are
outputted as a milivolt signal that is then converted into pressure in MPa by
applying the appropriate calibrations.

51
Up.Pore.Press.(MPa)
Pore pressure MPa (PU and PD)

50.8 Dwn.Pore.Press.(MPa)

50.6

50.4

50.2

50

49.8
100 200 300 400 500

Figure 3.26 - Typical experimental results from a permeability measurement employing

the PPO technique. Millivolt data has been converted into pressure in MPa and plotted
against time. The larger amplitude pressure wave is the forced sinusoidal pressure wave
in the upstream reservoir of the sample, and the smaller amplitude pressure wave is that
measured in the downstream reservoir. Note the attenuation and phase shift between the
upstream and downstream plots.

Due to the small volumes pore fluid reservoirs (especially the downstream)
during experiments and the low compressibility of water, variations in room

130
temperature and dilatant changes in the sample can cause the pressure readings to
drift. Such variations can be removed from the data by the processing program
as the pore pressure oscillation technique relies on processing a oscillating
sinusoidal signal varying around a mean value. The reader is referred to
Faulkner (1997) for further details. Figure 3.27 shows a typical datafile from an
experiment on granodiorite during axial loading at a constant strain rate up to
sample failure. The mean downstream pore pressure varies throughout the
experiment, due to the change in volume created by microcrack dilatancy. The
drift routine in the program is able to remove this trend before the data are
processed.

53 300
Up.Pore.Press.(MPa)
Dwn.Pore.Press.(MPa)
52.5 Stress (MPa)
250

52

200

Differential Stress (MPa)

51.5
Pressure (MPa)

51 150

50.5
100

50

50
49.5

49 0
0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
Time (Seconds)

Figure 3.27 - Typical experimental results from a permeability measurement experiment

during axial loading

To enhance the signal-to-noise ratio of the raw data that are collected, the use of
a Fast Fourier Transform (FFT) is employed, as shown by Fischer & Paterson
(1992). Faulkner (1997) discusses how this is applied to the data by the
processing program. FFT data processing offers a powerful signal-to-noise ratio
enhancement due to it’s filtering effect, resulting in an increased resolution /
sensitivity. This aids the ability to measure very low permeabilities and storage
capacities. Owing to its steady-state character, systematic contaminations such
as leaks and modulation can easily be recognized and removed from the data
record.

131
Finally, the program output values of ψ and γ so that permeability, k, and storage
capacity, βc may be calculated using equations 3.18 and 3.19. The Bernabe
method is employed in the processing program in place of the Fischer method
when the dimensionless parameter γ (see Figure 3.25) is large (>10 or 100)

132
3.4.2.3 Analysis of errors and minimization

Fischer (1992) conducted a comprehensive parameter study for the pore pressure
oscillation method which identified many of the potential problems that can be
encountered. Following on from this, Faulkner (1997) provided an in depth
critical assessment of the pore pressure oscillation technique and detailed the
errors involved that had not been identified or fully explored by previous
investigations.

Other than sources of error such as length and diameter measurements of samples
and fluid viscosity which will involve small inherent uncertainties, Faulkner
(1997) suggested that the greatest uncertainties are related to the dimensionless
parameters ψ and γ (see equations 3.18 and 3.19). Faulkner (1997) made a
comparison of permeability measurements made using the pore pressure
oscillation method to measurements made on the same samples using the well-
established pulse-transient technique, and shows that they are in good agreement,
although it is thought that values for storage capacity are relatively less reliable
that permeability. Fischer & Paterson (1992) and Faulkner (1997) showed that
there is a dependence of apparent permeability on the frequency and amplitude of
the pressure wave used in the upstream reservoir. They show that this
dependence of permeability on the wave period diminishes as the amplitude of
the wave is reduced, and can be almost eliminated (Faulkner 1997).

With these main issues in mind, Faulkner (1997) summarised three main points
to optimize the pore pressure oscillation technique so as to minimize errors:

1. Some idea of what the permeability of the material to be tested might be

is necessary to allow choice of experimental parameters (e.g. period) that
ensure the amplitude ratio and the phase shift fall within a range that
minimizes subsequent error in the value for permeability.
2. In the case that the amplitude of the pressure wave in the downstream
reservoir is falling below the transducer resolution, rather than increasing
the amplitude of the wave in the upstream reservoir, increasing the period

133
 of the waveform is preferable. If this leads to prohibitive testing times,
then a reduction in the volume of the downstream reservoir is the best
method to remedy this.
3. The optimization of the experimental equipment to measure permeability
may mean that only semi-quantitative information is retrieved for the
storage capacity of the sample. Faulkner (1997) pointed out that this
hypothesis clearly requires further investigation, by both numerical
analysis and experimentation with samples of the same material but of
different volume.

Probably the most important difference in this study is that dynamic permeability
is being measured. Fischer & Paterson (1992) and Faulkner (1997) conducted all
of their experiments under constant hydrostatic load. In this study, the evolution
of permeability and porosity is continuously monitored as differential load is
increased right up till sample failure. The pore pressure oscillation technique is
capable of making measurements during axial loading.

It is important to note that as the pore pressure oscillation technique utilizes the
attenuation and phase shift of a forced sinusoidal wave, a minimum of two
waveforms are required to be able to calculate the permeability. The program
calculates the permeability from every data point in the datafile to the
corresponding end point 2 waveforms later. As the axial strain rate is kept
constant during experiments, the data for permeability will in fact be an average
over these two waveforms, which means that permeability calculations are
somewhat smoothed and small variations will not be detected. During
deformation the sample permeability is constantly changing and as such a steady
state is never reached, and hence the values for permeabilities will be subject to a
small error, although values should still be representative. A low strain rate was
chosen as they allow for slow permeability changes.

134
4 Experimental measurements of permeability
evolution during triaxial compression of initially
intact crystalline rocks and implications for fluid
flow in fault zones

4.1 Abstract
Detailed experimental studies of the development of permeability of crustal rock
during deformation are essential in helping to understand fault mechanics and
constrain larger scale models that predict bulk fluid flow within the crust. The
strength, permeability and pore fluid volume evolution of initially intact
crystalline rock under increasing differential load leading to macroscopic failure
has been determined at water pore pressures of 50 MPa and varying effective
pressures from 10 to 50 MPa. Permeability is seen to increase by up, to and
over, two orders of magnitude prior to macroscopic failure, with the greatest
increase seen at lowest effective pressures. Post-failure permeability is shown to
be over 3 orders of magnitude higher than initial intact permeabilities and
approaches lower the limit of measurements of in situ bulk crustal permeabilities.
Increasing amplitude cyclic loading tests show permeability-stress hysteresis,
with high permeabilities maintained as differential stress is reduced and the
greatest permeability increases are seen between 90-99% of the failure stress.
Under hydrothermal conditions without further loading, it is suggested that much
of this permeability can be recovered, and pre-macroscopic failure fracture
damage may heal relatively faster than post-failure macroscopic fractures. Pre-
failure permeabilities are nearly seven to nine orders of magnitude lower than
that predicted by some high pressure diffusive models suggesting that
microfracture matrix flow cannot dominate, and agrees with inferences that bulk
fluid flow and dilatancy must be dominated by larger scale structures, such as
macrofractures. It is suggested that the permeability of a highly-stressed fault tip
process zone in low-permeability crystalline rocks could increase by more than 2
orders of magnitude, while stress-drops related to fracture propagation close
damage zone cracks, and some permeability is maintained due to hysteresis from
permanent microfracture damage.

135
4.2 Introduction

Experimental constraints on the evolution of permeability of crustal rock with

deformation are essential in helping to understand bulk fluid flow within the
crust. Models predicting crustal fluid flow are important for a variety of
reasons;for example earthquake models invoking fluid triggering (e.g. Miller et
al. 2004, Sibson 1974), or modelling flow surrounding deep waste repositories.
Crustal fluid flow is controlled by both the bulk transport properties of rocks
(e.g. Stearns 1968), as well as heterogeneities such as faults. Fracture damage on
all scales affects the permeability of fault rock and understanding these variations
is important as it gives an invaluable insight into the pathways and barriers for
fluid flow (Caine et al. 1996, Knipe 1992, Shipton et al. 2002). Faults are a
major control on crustal fluid flow, which in turn can control crustal strength
(Townend & Zoback 2000) and trigger earthquakes (Miller et al. 2004).
Impermeable fault core material can act as fluid barriers and can trap fluids
within faults zones and allow overpressures greater than that of lithostatic
pressures to develop (Byerlee 1990); high pressure fluid sealing and weakening
models (Blanpied et al. 1992, Faulkner & Rutter 2001, Rice 1992) rely on
interconnected fault networks within core and damage zones.

It is well known that the pervasive damage of rock by dilatational microcracks

strongly influences rock strength and macroscopic elastic properties. It has been
demonstrated experimentally that faults in brittle rock are shear fractures formed
through the interaction and coalescence of many tensile microcracks (Healy et al.
2006, Lockner et al. 1991, Lockner et al. 1992, Reches & Lockner 1994). This
suggests that when increasing axial load is applied to intact rock samples,
microfracture damage increases as the rock approaches failure, and the resultant
progressive dilatancy will have a direct effect on both porosity and permeability.
Once faults have initiated and begin to grow, various mechanical and wear
models suggest ways in which fracture damage then accumulates. Various
mechanical models suggest intense zones of microfracturing in the ‘process
zone’ surrounding the highly stressed fault tip (e.g. Scholz et al., (1993) and see
also review by Dresen & Guéguen (2004)) which decrease in density with

136
distance from the fault and increase in size with increased fault length. Wear
models (e.g. Chester & Chester 2000) suggest continued slip on pre-existing fault
surfaces introduces cumulative damage into the surrounding rock and the elastic
strain energy released during the stick-slip nature of earthquakes introduces
cyclic damage (e.g. Rice et al. 2005). Whatever processes are dominant, a key
factor is that microfracture damage surrounding fault zones will vary
significantly which will have a direct effect on permeability surrounding faults at
all stages of fault evolution. This highlights the importance of understanding the
relationship between progressive fracture damage and permeability.

Permeability evolution of crystalline rocks during progressive deformation has

have has received remarkably little attention in the literature. Various studies
have examined the permeability evolution of porous clastic rocks (sandstones)
under increasing differential stress (Heiland 2003, Keaney et al. 1998, Zhu &
Wong 1997), both pre and post-failure with initial porosities ranging from 4.5 to
>14%. However few data exist for crystalline rocks such as granites which can
have porosities lower than 1%. Zoback & Byerlee (1975) produced one of the
first studies on the permeability evolution of granite under increasing differential
stress. They determined permeability under stepwise increasing axial stress at
effective confining pressures of 39 and 14 MPa. Rock specimens were loaded up
to a maximum differential stress of approximately 80% of their peak strength and
showed increases in permeability of around 300%.

In most previous work (e.g. Keaney et al. 1998, Zhu et al. 1997, Zoback &
Byerlee 1975) the pulse transient technique for permeability measurement was
used (Brace et al. 1968, Sanyal et al. 1972). This requires the axial load actuator
to be stopped while permeability is measured. This may introduce the problem
of creep, due to the fact that once dilatancy has initiated, rocks exhibit stable
crack growth (Atkinson & Meredith 1987) which also occurs while under
constant load which leads to ongoing deformation, especially in the lateral
direction (Heiland 2003). As a result of the actuator being stopped and hence
fixed, stress relaxation results in a decrease in the axial stress, which in turn leads
to increasing lateral strain that influences permeability. Keaney et al. (1998)
suggested stopping the actuator movement results in a non-steady-state

137
permeable microstructure in the rock, especially in the dilatant deformation
phase. In these studies, different techniques were used to reduce the effect of
creep on permeability measurements. For example, Keaney et al. (1998) reduced
the axial load during the permeability measurements during a standardized
procedure in order to achieve steady-state condition. Heiland (2003) tried to
suppress lateral deformation while the actuator was stopped by controlling the
lateral strain by varying the confining pressure.

Due to the nature of the experimental methodology i.e. stopping the loading to
measure a single static permeability reading, the amount of data that are recorded
is limited by the longer time taken to complete enough experiments to describe
the permeability variation. In this study, the pore pressure oscillation technique
(Fischer 1992, Kranz et al. 1990) is employed to measure permeability as it
allows the continuous measurement of permeability and porosity evolution
during progressive deformation without stopping axial load. In our experiments,
permeability was measured at various effective pressures during the progressive
deformation of samples of Westerly granite and Cerro Cristales granodiorite
(from the Coastal Cordillera, northern Chile) under constant confining pressure
and pore pressure. Firstly, the variations in permeability of initially intact
crystalline rock right up to macroscopic failure of samples are presented, then
cumulative damage and permeability hysteresis due to increasing amplitude
cyclic loading is presented, and these results are used as a framework to assess
the permeability structure surrounding fractures and faults and the implications
for fault zone hydraulics.

138
4.3 Experimental Technique

The experimental apparatus used in these experiments is shown schematically in

Figure 4.1. The equipment is a high pressure triaxial deformation apparatus with
a servo-controlled pore fluid pump. This rig is capable of performing triaxial
deformation experiments at confining pressures up to 250 MPa and has an
independent pore fluid system capable of measuring ultra-low permeability and
sample volume change during deformation experiments. The servo-controlled
pore fluid generator is capable of measuring permeabilities down to 10-22 m2, and
resolving porosity changes ~0.1 mm3. The system controls pore fluid pressure at
the upstream end of the sample (see fig 1). The downstream end of the sample is
connected to a low volume reservoir. This system is used here to measure
volume change and permeability via the pore pressure oscillation technique
(Fischer 1992), although it can be used for the pulse transient technique (Brace et
al. 1968).

Figure 4.1 - Schematic diagram of the experimental apparatus. The pore pressure
oscillation technique can be used on a sample loaded in the conventional triaxial
configuration and progressive permeability development can be measured during
constant axial deformation.

139
The rig can accommodate samples up to 25mm in diameter and has a servo-
controlled ball-screw driven actuator with a 30 tonne loading capacity. The
internal force gauge design utilizes high performance materials (maraging steel
with a ~2 GPa yield strength) in order to maximize the load-measuring
sensitivity. The balanced piston allows the full 30 tonne loading capacity to be
directly applied to the sample. The balanced piston can be bypassed to allow
deformation in extension in addition to compression.

The pore pressure oscillation technique was developed concurrently by Kranz et

al. (1990) and Fischer (1992). The technique is able to measure both
permeability, k, and storage capacity, S, simultaneously during the course of a
deformation experiment. The technique utilizes the attenuation (amplitude shift)
and phase shift of a forced sinusoidal pressure wave created in the upstream
reservoir as it is transmitted through the sample to the downstream reservoir to
obtain two dimensionless parameters which may be used to calculate k and S
(Fischer 1992). Recently Bernabe et al. (2006) re-analysed the oscillating flow
method, including determining the uncertainty on permeability and storativity.
They derived approximate solutions for these parameters in the end-member
cases where there is very small or very large sample storage with respect to the
downstream storage capacity. The improvements of Bernabe et al. (2006) have
been incorporated into the processing of the data in this study. It is important to
note that as the technique utilizes the attenuation and phase shift of a forced
sinusoidal wave, a minimum of two waveforms are required to be able to
calculate the permeability. During processing the permeability from every data
point in the datafile to the corresponding end point two waveforms later. As the
axial strain rate is kept constant at 9 x 10-8 s-1 during all experiments, the data for
permeability will in fact be an average over these two waveforms, which means
that permeability values are somewhat smoothed and small variations will not be
detected. During deformation the sample permeability is constantly changing
and as such a steady state is never reached, and hence our values for
permeabilities will be subject to a small error, although values should still be
representative. A low strain rate was chosen as they allow for slow permeability
changes. Figure 4.2 shows a typical datafile from an experiment on granodiorite

140
during axial loading at a constant strain rate up to sample failure. The mean
downstream pore pressure varies throughout the experiment, due to the change in
volume created by microcrack closure and dilatancy. These changes in mean
downstream pressure can be removed from the experimental data during
processing by assuming a constant rate of pressure change between the start and
end of each cycle and removing fractions of difference by linearly interpolating
the change for each pointing the record between the start and the end of the cycle
(Faulkner & Rutter 2000).

53 300
Upstream pore pressure
Downstream pore pressure
52.5 Stress (MPa)
250

52

Differential Stress (MPa)

200
51.5
Pressure (MPa)

51 150

50.5
100

50

50
49.5

49 0
0 5000 10000 15000 20000 25000 30000 35000 40000 45000 50000
Time (Seconds)

Figure 4.2 - Typical experimental results from a permeability measurement experiment

during axial loading.

Rock cores of 20mm in diameter and 50mm in length were cut from intact rock
blocks of Westerly granite and Cerro Cristales granodiorite, and placed in the
sample assembly as shown in Figure 4.1. Samples were saturated with de-
ionised distilled water as the pore fluid and placed in a PVC jacket. Porous
alumina spacers are inserted on to the upstream and downstream faces of the
sample which ensures that there is an even distribution of pore pressure across
the face of the sample. Confining and pore pressures were monitored using
strain gauge pressure transducers with accuracies of 50 Pa. Samples were
deformed at room temperature in the conventional triaxial configuration (axial
stress σ1 > σ2 = σ3) using silicon oil as a confining medium. The upstream pore

141
pressure (Pp) was kept controlled with oscillations with an amplitude of 1 MPa
around a mean of 50 MPa and the confining pressure Pc (=σ3) was varied
between 60 and 100 MPa to explore the effect of various effective pressures on
the permeability evolution. The oscillating wave is assumed to have a negligible
effect on the damage mechanics in the sample. A pore fluid controller
maintained oscillating pressure wave, and the pore volume change was recorded
by monitoring the piston displacement of the pressure generator using a
displacement LVDT. Axial load is measured by a calibrated internal force gauge
with an accuracy of just under 10 N. Once the specified confining and pore
pressures were attained, the oscillating upstream pressure wave was enabled and
once equilibrated (i.e. downstream pressure began to show a noticeable pressure
wave around a constant mean), differential load was applied via a servo-
controlled loading system as described above and shown in Figure 4.1.
Displacement was measured externally using a displacement LVDT in contact
with the moving piston. With a knowledge of the stiffness of the loading system
(8.43 x 10-3 mm/kN with zero confining pressure), the axial displacement of the
sample was obtained by subtracting the displacement of the loading system from
the apparent displacement measured by the LVDT, and axial strain was
calculated with reference to the initial length of the sample. In order to
investigate how the permeability hysteresis varied at different points in the strain
history, samples were subjected to increasing-amplitude cyclic loading
experiments (see Table 4.1). These consisted of five increasing and decreasing
stress cycles with increasing peak stresses of 60, 70, 80, 90 and 100% of the
failure stress for each cycle, with the lower limit of each cycle at 15% of the
failure stress.

142
 Plg
Qtz

Aph

Bt

Figure 4.3 - a) Example of the Cerro Cristales Granodiorite in hand specimen. b) Thin
section of granodiorite in transmitted cross-polarized light. Average quartz grain size is
0.7 mm. Major phases are labelled. Qtz = quartz, Plg = plagioclase feldspar, Aph =
amphibole and Bt = biotite.

Westerly granite (from south-east Rhode Island, USA) is commonly used in

experimental rock mechanics due to its small grain size, low initial crack density
and that it is generally considered to be isotropic. Hence it can yield a high level
of experiment reproducibility under carefully controlled test conditions (see
Haimson & Chang 2000, Lockner 1998). The Cerro Cristales granodiorite
(Figure 2.4) is taken from the Cerro Cristales pluton in the Coastal Cordillera
region of the Atacama desert in Northern Chile, which was first described by
Uribe & Niemeyer (1984). This pluton has been studied in detail by González
(1990, 1996, 1999) and consist of plagioclase, quartz, orthoclase, biotite and
amphibole with an average modal grain size of 0.7 mm.

The rationale behind using Cerro Cristales granodiorite in addition to Westerly

granite is that microfracture densities have previously been measured in natural
samples (e.g. Faulkner et al. (2006) and in section 2.5.3 of this work), and future
work will aim to quantify experimentally induced microfractures and associated
permeability measurements, and relate to these to a natural fault zone. Westerly
granite also provides an excellent comparison for the granodiorite results.

143
4.4 Results

The experimental details for all of the tests reported are summarized in Table 4.1.
In this section, comparisons are made between the variations in permeability
during progressive deformation until failure of samples of granodiorite and
granite at various effective pressures. All samples of both rock types showed
localized brittle failure in the form of a through-going shear fracture at an acute
angle to the principle stress (σ1). Results for Cerro Cristales granodiorite are first
presented, followed by results of Westerly granite and then results of the
increasing amplitude cyclic loading test.

Table 4.1 – Summary data for the 10 triaxial experiments

Test No. Pc (MPa) Pp (MPa) Peff (MPa) Sample Yield Failure Rock
failed? stress stress
1 100 50 50 N ~325 350 Grano
2 60 50 10 Y 190-200 220 Grano
3 75 50 25 Y - 275 Grano
4 100 50 50 N - 350 Grano
5 75 50 25 Y - 290 Grano
6 60 50 10 Y 180 216 Grano
7 60 50 10 Y 260 278 Westerly
8 (cycled) 60 50 10 Y - 276 Westerly
9 70 50 20 Y 341 356 Westerly
10 65 50 15 Y 298 330 Westerly
Grano = Cerro Cristales granodiorite. Westerly = Westerly granite.

144
4.4.1 Cerro Cristales Granodiorite

Figure 4.4 shows differential stress and permeability plotted against axial strain
for the six tests completed on Cerro Cristales granodiorite at effective pressures
of 10, 25 and 50 MPa. The stress-strain curves show typical behaviour of
crystalline rock with failure stress increasing with increased confining pressure
(or effective stress). For the first 50 MPa the stress-strain response is concave
upwards, becoming very nearly linear soon after. Starting values of permeability
vary over an order of magnitude from as low as 4 x 10-21 m2 to 7 x 10-20 m2
which appears to scale with effective pressure from 10 MPa to 50 MPa
respectively. As strain increases, permeability initially decreases for all effective
pressures up to a strain of 0.0015, which compares well to the concave section of
the stress-strain curve. Stress-strain behaviour for all effective pressures then
appears to be linear-elastic up to the yield point, and then concave down until
failure. Failure stresses range from around 220, 290 to 350 MPa for effective
pressures of 10, 25 and 50 MPa respectively with repeat tests at the same
pressures showing good reproducibility. For the lowest effective pressure of 10
MPa, after the initial decrease permeability steadily increases with an increasing
gradient right up till failure, with permeability varying from 3.5 x 10-21 to 9 x 10-
19
m2- over two orders of magnitude. For effective pressures of 25 MPa, the
permeability increases relatively slowly up until a strain of 0.0035, where there is
a relatively sharp turning point and permeability increases sharply from around
1.7 x 10-20 to 6.3 x 10-19 m2. For effective pressures of 50 MPa, again
permeability increases relatively slowly up to a turning point around a strain of
0.0038, where the permeability changes significantly with an increase from 10 x
10-21 to 1.8 x 10-19 m2. These results also seem to suggest that the final
permeability just pre-failure decreases as a function of the effective pressure.

145
Figure 4.4 - Experimental data for Cerro Cristales Granodiorite. The differential stress
and permeability are plotted versus axial strain. Effective pressures are indicated, and
pore fluid pressure was maintained at 50MPa.

Figure 4.5 shows permeability and pore volume change as functions of

differential stress for Cerro Cristales granodiorite. The volume is the amount of
fluid that moves in or out of the sample during deformation as measured by the
volumometer, and is proportional to crack volume and hence porosity. Positive
change is pore fluid being expelled (decrease in porosity) from the sample and
negative change is pore fluid being added to the sample (increase in porosity).

146
The stress – permeability relationships are similar to the relationships for strain
versus permeability shown in Figure 4.4, as might be expected, although the
curves for the repeated tests at the same effective pressures are slightly spread
out due to small variations in the peak failure stress.

147
 Porosity decrease
Porosity increase

Figure 4.5 - Permeability and volume as functions of differential stress for Cerro
Cristales granodiorite. The pore volume is the amount of fluid that moves in or out of
the sample during deformation as measured by the volumometer, and is proportional to
porosity. Positive change is pore fluid being expelled (decrease in porosity) from the
sample and negative change is pore fluid being added to the sample (increase in
porosity). Pore pressure is maintained at 50 MPa.

148
For all tests, the pore volume initially decreases up to the turning point at which
it then begins to increase at an accelerating rate. A key observation is that the
inflection point of the volume – stress curves i.e. where the porosity stops
decreasing and starts increase, does not appear to coincide with the change in
gradient of permeability increase as might be expected; permeability begins to
noticeable increase at a slightly lower strain.

Figure 4.6 - Experimental data for Cerro Cristales Granodiorite tests 2 and 3 at effective
pressures of 10 and 25 MPa. The differential stress and permeability are plotted versus
time and include measurements of permeability post-failure. Pore fluid pressure was
maintained at 50MPa.

Figure 4.6 shows a graph of differential stress and permeability plotted versus
time for tests 2 and 3. Post-failure permeability values have also been measured
and plotted for these two tests, where there has been a substantial drop in the
differential stress. Post-failure permeabilities were not measured in all samples,
although the interpretation of the post-failure permeability data is not
straightforward as the sample will have developed a highly heterogeneous
permeability structure during the formation of the shear fracture. As a sample

149
approaches failure stress and failure is imminent, there is a window where no
data can be gathered as the final stage of the loading occurs very rapidly making
it difficult to gain results for permeability in such a short space of time. The
period of the oscillation wave is reduced during the experiment (from 5000,
2000, 1000, 500, 250 and 100) as the sample becomes more permeable, and the
period is relatively small towards the end of the experiment, so only a small
segment of data will be missing. However, it can be seen that post-failure
permeability is higher for both tests than the previously measured data. For an
effective pressure of 10 MPa, permeability reaches as high as 4 x 10-18 m2 which
is three orders of magnitude change from the initial preloading permeability.

4.4.2 Westerly Granite

Figure 4.7 shows the differential stress versus axial strain for the 3 tests
completed on Westerly granite at effective pressures of 10, 15 and 20 MPa. Due
to the greater compressive strength of the Westerly granite in comparison to the
granodiorite, effective pressures were not taken higher than 20 MPa (70 MPa
confining pressure) as the failure stress could not be reached due to it being
beyond the maximum load capacity of the loading apparatus. As was found for
the granodiorite, the stress strain curves for Westerly granite show typical
behaviour for crystalline rock with failure stress increasing with increased
effective pressure. The stress-strain response is concave upwards for the first 40
MPa, becoming very nearly linear soon after. Starting values of permeability
range from 1 x 10-19 to 3.6 x 10-20 m2, which are towards the upper range and
slightly higher than the Cerro Cristales granodiorite starting permeabilities. This
range in permeability does not vary as much as the Cerro Cristales granodiorite,
but this is likely to be due to the smaller range of effective pressures used. As
strain increases, unlike with the Cerro Cristales granodiorite, there is a very small
decrease, if any, in permeability during the initial non-linear section of the stress
strain curve. Stress-strain behaviour for all effective pressures then becomes
linear elastic up to the yield point, and then slightly concave down until failure.
The failure stresses are 278, 330 and 356 MPa for effective pressures of 10, 15

150
Figure 4.7 - Experimental data for Westerly Granite. The differential stress and
permeability are plotted versus axial strain. Effective pressures are indicated, and pore
fluid pressure was maintained at 50MPa.
and 20 MPa respectively. As with the granodiorite, permeability steadily
increases and accelerates prior to sample failure. However, each of the three
permeability curves are similar in shape, whereas the granodiorite permeability
curves have varying slopes with distinct turning points that were proportional to
effective pressure.

151
For the lowest effective pressure of 10 MPa, permeability steadily increases at a
strain of around 0.002 from 5 x 10-20 to 1 x 10-18 m2, just over an order of
magnitude. For effective pressure of 15 MPa, the permeability begins to rise at a
strain of 0.0027 from 3.3 x 10-20 to 2.5 x 10-18 m2. For the highest effective
pressure of 20 MPa, permeability begins to rise at a strain of 0.0033 from 4 x 10-
20
to 2 x 10-18 m2. Unlike for the granodiorite, there is no relationship between
the final pre-failure permeability and effective pressure. In fact the lowest
effective pressure appears to be associated with the smallest change in
permeability.

Figure 4.8 shows permeability and pore volume change as functions of

differential stress for Westerly granite. The stress-permeability relationships are
similar to the relationships shown in Figure 4.7 for strain and permeability. The
volume-stress curves show concave down shapes that suggest volume initially
increases, and hence porosity decreases, to a peak level that corresponds well
with the onset of steepening gradient of the permeability curve. This varies from
the results of the Cerro Cristales granodiorite that showed the permeability began
to increase slightly before the porosity began to increase. The turning point /
yield stress increases as a function of effective pressure at 130, 155 and 180 MPa
for effective pressures of 10, 15 and 20 MPa respectively. The peak volume i.e.
maximum decrease in porosity also seems to vary as a function of effective
pressure, with the largest porosity increase of around 12mm3 for an effective
pressure of 20 MPa. This porosity decrease is comparable to that of the Cerro
Cristales Granodiorite.

152
 decrease
Porosity
Porosity
increase

Figure 4.8 - Permeability and volume as functions of differential stress for Westerly granite.
The volume is the amount of fluid that moves in or out of the sample during deformation as
measured by the volumometer, and is proportional to porosity. Positive change is pore
fluid being expelled (decrease in porosity) from the sample and negative change is pore
fluid being added to the sample (increase in porosity). Pore pressure is maintained at 50
MPa.

153
4.4.3 Increasing amplitude cyclic loading test on Westerly
Granite

In addition to the tests described above, an increasing amplitude cyclic loading

test was completed on Westerly granite in order to identify any permeability
hysteresis due to the loading and unloading of stress within a sample, in an
attempt to examine what effect such loading conditions may have on in-situ
materials that undergo such loading cycles such as in earthquake events. Figure
4.9 shows a graph of differential stress versus axial strain for an increasing stress
amplitude cyclic loading test on Westerly granite. The graph shows distinct
unloading hysteresis. A cumulative increase in strain with each cycle can be seen
by examining the base of each loading cycle, which suggests that permanent
strain has accumulated between each cycle due to crack damage.

Figure 4.9 - Experimental data for stress cycled Westerly granite. The differential stress is
plotted versus axial strain. Each cycle is indicated along with the approximate percentage of
the failure stress that each reached. Pore fluid pressure was maintained at 50 MPa with a
confining pressure of 60 MPa.

154
Figure 4.10 shows a graph of permeability versus differential stress for the same
cyclic test as shown in Figure 4.9. During the first loading cycle, the
permeability decreases slightly from an initial permeability of around 7 x 10-20
m2, again presumably due to the closure of pre-exisiting cracks. As differential
stress increases, the permeability increases correspondingly to 1.4 x 10-19 m2 at
the top of the stress cycle (60% of the failure stress). As differential stress is
removed, the permeability shows the same trend as shown in Figure 4.10, where
the permeability is initially maintained as effective pressure is reduced then
begins to decrease. The second cycle (70% failure stress) follows a very similar
permeability –stress path to the first. The third cycle (80% failure stress)
however shows an increase in permeability of 500%, with the permeability
gradient steepening somewhat at higher stresses. The concave-up shape of the
loading-permeability curve is mirrored in the unloading curve and the
permeability at the bottom of the cycle is 300% higher than the starting
permeability. The fourth cycle (90% failure stress) shows a permeability change
of nearly an order of magnitude, and the permeability at the end of the cycle is
400% higher than the permeability at the start of the cycle. The 5th and final
cycle shows a permeability increase of over an order of magnitude up to failure,
with the final pre-failure permeability being almost two orders of magnitude
higher than the initial permeability, at 4.4 x 10-18 m2.

155
 4

5
4

1 1
3 2

2
1

Figure 4.10 - Experimental data for stress cycled Westerly Granite. The permeability is
plotted versus differential stress. Each cycle is indicated along with the percentage of the
failure stress that each reached at the peak of each cycle. Pore fluid pressure was maintained
at 50 MPa with a effective pressure of 10 MPa.

156
4.5 Discussion

4.5.1 Strength and permeability evolution of Cerro Cristales

granodiorite and Westerly granite.

The stress-strain relationships show that Westerly granite Figure 4.7a appears to
be much stronger than Cerro Cristales Figure 4.4a granodiorite with failure
stresses being somewhat higher (summarized in Table 4.1). The plots of Mohr-
Coulomb failure criterion shown in Figure 4.11 suggest that Westerly granite has
a higher coefficient of internal friction. Although the mineralogy between the
two rock types is not dramatically different, Westerly granite has a much smaller
grain size, and it is well known that smaller grains are harder to fracture (e.g.
Wong et al. 1997).

Figure 4.11 - Mohr-Coulomb Failure Criterion for Cerro Cristales granodiorite and
Westerly granite.

157
The permeability results show many similarities between the two rock types.
With increased differential stress and strain, there appears to be an initial
transient of permeability decrease that is more apparent for the Cerro Cristales
granodiorite than the Westerly granite. The initial decrease is related to the non-
linear increasing stiffness part of the stress-strain curve is most likely due to the
closure of microfractures orientated obliquely and transversely to the applied
load. Crack closure causes a decrease in porosity and related decrease in
permeability. The Westerly granite shows no significant decrease in
permeability in the crack-closure phase, which suggests that the crack network is
not hydraulically connected or a concomitant decrease in permeability would be
seen. Also, the apparent lack of discernable decrease in permeability of the
Westerly granite may be because the Cerro Cristales granodiorite is more
weathered than the Westerly granite, as the Cerro Cristales samples were cored
from samples collected from surface rocks, whereas the Westerly granite was
cored from quarry cut samples. If there are cracks that have weathering products
in them within the Cerro Cristales, then these may hold open the cracks until a
small amount of shear helps to close them. The confining pressure will play a
part in how well these shearing cracks will close, hence the confining pressure
control. These effects are therefore not seen in the Westerly granite.

After the initial decrease, permeability increases by up to, and over, two orders of
magnitude until sample failure. The Cerro Cristales granodiorite shows the
greatest increases in permeability from 3.5 x 10-21 to 9 x 10-19 m2, well over two
orders of magnitude, and if post-failure permeability is taken into account, one of
the tests showed a total range in permeability of three orders of magnitude, as
high as 4 x 10-18 m2 once most of the stress was removed. The Westerly granite
shows increases of 4 x 10-20 to 2 x 10-18 m2, just under two orders of magnitude.
The schematic graph shown with the diagram in Figure 4.14 shows a summary
graph of the results now discussed. The stress-strain curves become
approximately linear once cracks have closed and this is related to the lowest
permeabilities measured. Once closed, linear elastic deformation takes place and
the permeability either remains constant or slightly increases up to the yield

158
point. The yield point of the stress-strain curve is defined as the progression
from elastic deformation alone to the onset of permanent crack damage. This is
reflected in the volume-stress relationship as the point of maximum curvature,
where pore volume change changes from contractional to dilatational (Figure
4.14) (volume of fluid expelled will be related to the porosity and volumetric
strain). From this point on as stress and strain increase, porosity increases
dramatically and accelerates up to sample failure. The Westerly granite shows
that the onset of permeability increase correlates well with the point at which the
volume change evolves from contractional to dilatational. This differs for the
Cerro Cristales granodiorite that shows permeability begins to increase at a
slightly lower strain before the pore volume change changes from contractional
to dilatational, which implies that the sample volume was deceasing while
permeability had started to increase. This suggests that sample anisotropy was
already developing, with closure of oblique and transverse cracks dominating the
pore volume but perhaps growth of the axial cracks, contributing less to the
overall pore volume change, are controlling the permeability. There appears to
be a relationship between final pre-failure permeabilities and effective pressure
and the Cerro Cristales granodiorite shows that the final permeabilities
immediately prior to failure appear to decrease systematically in tests conducted
at higher effective pressures. This may be due to the confining pressure having a
greater closing effect on axially orientated microcracks which is supported by
showing less pore volume increase in the tests conducted at higher effective
pressures. However, the Westerly granite data do not seem to show this
relationship between final pre-failure permeabilities and effective pressures,
which could be due to weathering products in cracks. Further tests need to be
completed to assess this. Effective pressure also has some effect on the shape the
permeability strain curves, with an increasingly sharper turning point with
increasing effective pressure. Also, there is a larger general increase in porosity
for the Westerly granite experiments. This could perhaps be due to a more
pervasive crack network forming which could be related to grain size, especially
if cracks are intragranular.

The results in this study show different permeability variations to studies that
have focused on the evolution of permeability of clastic rocks under increasing

159
differential stress (Heiland 2003, Keaney et al. 1998, Zhu & Wong 1997). Low
porosity rocks increase in permeability as a result of increased induced fracture
damage, whereas in higher porosity clastic rocks, reduction in permeability can
be seen due to a collapse in porosity and increased comminution in the form of
deformation bands (Aydin & Johnson 1978).

4.5.2 Permeability evolution from induced cyclic damage

Rocks do not show perfectly elastic stress-strain behaviour, as once the yield
point is exceeded there is hysteresis, which implies that some of the energy
stored during loading has been dissipated into the specimen (Jaeger et al. 1969,
Paterson & Wong 2005). In low temperature experiments this energy is
generally assumed to be used to create new crack surface area which has a direct
effect on both porosity and permeability. The results of the increasing amplitude
cyclic loading test on Westerly granite show both strain and permeability
hysteresis related to each stress cycle. The cyclic test data shows permeability
increases by 400-500% of the starting permeability of each of the first four
cycles, with starting permeabilities increasing by 200-300% each cycle. The
final cycle sees a permeability increase over an order of magnitude just before
sample failure, which is a total permeability change of two orders of magnitude
from the initial sample permeability at the beginning of the test. The total
change in permeability is similar to the total change of the non-cycled tests,
suggesting that pre-failure permeability is the same for cycled and non-cycled
tests and permeability controlled by the maximum stress that the sample has
experienced rather than number of cycles. The shape of the curve for each stress
cycle is very similar with the gradients of the unloading curve showing a very
similar pattern. This means that the permeability change with varying differential
stress is predictable – the sample can be taken to any percentage of the failure
stress and predictions can be made for what the permeability would do if it was
unloaded from that point. This has useful implications for modelling
permeability variation. Overall, the results shows that permanent strain has
accumulated between each stress cycle, and therefore crack damage mechanisms

160
must have been operative. The increasing amplitude stress cycling appears to
show an overall increase in permeability and reduction in sample stiffness.

Zoback & Byerlee (1975) determined the permeability of Westerly Granite under
stepwise increased axial stress at effective pressures of 39 and 14 MPa (pore
fluid pressure of 11 MPa) using the pulse transient technique of Brace et al.
(1968) and Sanyal et al. (1972). These effective pressures are slightly higher
than the 10 MPa used in this study. Figure 4.12 shows the results of Zoback &
Byerlee (1975) overlain onto the same results from Figure 4.10. Samples were
loaded up to a maximum differential stress of approximately 80% of their peak
strength.

Figure 4.12 - Experimental data for stress cycled Westerly Granite, as shown in
Figure 4.10 but with the results of Zoback and Byerlee (1975) overlain for
comparison. To aid comparison between the datasets, the permeability has been
plotted as a function of the percentage of failure stress.

161
Permeabilities published by Zoback & Byerlee (1975) compare well to the data
shown in this study, although starting permeabilities are slightly higher. The
starting permeability compares well to permeability at the end of our 3rd cycle
which, as in their study, reached a peak of 80% the failure stress. A key
difference is that where this study has taken stress to higher levels, above 80%
failure stress, permeability increases by over an order of magnitude before
failure.

Zoback & Byerlee (1975) pre-loaded their samples to un-specified high

differential stress over 20 times before starting their permeability measurements
in an attempt to simulate more closely in situ materials that may repeatedly
undergo large stresses, whereas in this study samples are initially intact. Even
so, permeabilities still seem to compare well to results in this study that have
been cycled only three times. (Heap & Faulkner In press) showed that cyclic
fatigue is not a major contributor to microfracture damage in Westerly granite, as
constant amplitude stress cycles produced no variations in elastic properties
which was interpreted as indicating that no additional microfracture damage is
imparted in the sample. It is only with increasing amplitude stress cycles that
further damage is created; until the previous cycle peak stress is exceeded, no
new microfracture damage is noticeably introduced. This is termed the ‘Kaiser
effect’, and their hypothesis is supported by acoustic emission analysis during
experiments by (Lavrov 2003) who demonstrated a paucity in acoustic emission
events during constant amplitude stress cycles. Therefore despite their samples
being cycled 20 times, our results compare well as cyclic fatigue is not very
effective without concomitant increases in differential stress for each cycle. A
key finding with our study over Zoback and Byerlee’s is that above 80% failure
stress, permeability increases by over an order of magnitude before failure,
suggesting that the final stages of the loading are very important for permeability
evolution.

The stress-permeability curves during increasing differential stress show a

concave-up shape, and during decreasing differential stress show much shallower
gradient resulting in a concave down shape. Zoback & Byerlee (1975) attribute

162
this increase to two competing effects that determine the permeability; flow
decrease through the axial cracks as they begin to close, but flow increases
through the opening non-axial cracks. In their study, they saw a slight increase
in permeability as differential stress was decreased which they interpret as the
nonaxial cracks temporarily dominating the behaviour of the sample.

The experiments in this study were completed at room temperature with water as
a pore fluid, unlike Zoback & Byerlee (1975) who used argon which is
chemically inert and hence only considered the mechanical effects of the pore
fluid. However, the data compare well, suggesting processes such as stress
corrosion do not have a large effect on the timescales of our experiments. Also,
water is arguably the most common geofluid in the upper crust, and this therefore
makes the results of this study more applicable to natural crustal conditions.

163
4.5.3 Crack network controlling permeability

It is well documented experimentally that faults in brittle rock are shear fractures
formed through the interaction and coalescence of many tensile microcracks
(Healy et al. 2006, Lockner et al. 1991, Lockner et al. 1992, Reches & Lockner
1994). This implies that with increasing stress on a sample, an increasingly
heterogeneous microfracture network must develop as a sample approaches
failure. In light of the increases in permeability with increasing differential stress
demonstrated in this study, it raises the question as to what extent such
heterogeneity is responsible for the variation in bulk permeability. (Lockner et
al. 1991) presented observations of quasi-static fault growth in Westerly granite
(Figure 4.13). By using an array of piezoelectric transducers, they recorded the
arrival times of acoustic emission (AE) events which indicate the impulsive
growth of microcracks. The arrival times were then inverted to obtain three-
dimensional locations of the events that occurred during fault nucleation and
growth.

a b

Figure 4.13 – Simplified key figures from Lockner et al. (1991). a) Axial stress plotted
against axial displacement for Westerly granite at Pc = 50 MPa. Letters along curve
correspond to plots in (b). b) Sequential plots of locations of acoustic emission. Stress
interval for each plot is shown in (a). Upper plots show events viewed along-strike of the
eventual fault plane (seen as diagonal band of events). Lower plots show same events when
the fault plane is viewed face-on. Fault nucleates in b and propagates across sample in c-f. A
distinct fracture front develops, in d-f, as fault grows.

By controlling the axial load using a fast acting control-system that adjusted the
load applied to the sample to maintain a constant emission rate, they were able to

164
study the post-peak-stress region, in which a macroscopic fault forms in the
sample. This allowed the imaging of the post-failure-stress curve, as the
technique extended to hours the fault growth processes which normally would
occur in a fraction of a second. Figure 4.13a shows the loading curve for the
sample, and it can be seen that region a represents the pre-failure region and the
region that permeability was measured in this study. Figure 4.13b shows that
region (a) corresponds to a relatively diffuse acoustic emission events that are
interpreted to be homogenous microfracturing occurring through the sample. It
is only at regions c-f that a distinct fracture front begins to develop and sample
failure occurs. The implications of this for the data in this study is that the
variation of permeability seen pre-failure is entirely due to a relatively
homogenous microfracture network developing throughout the sample, whereas
post-failure permeability is controlled by a strongly heterogeneous sample with a
thoroughgoing shear fracture.

4.5.4 Fault healing and permeability recovery

Within the seismogenic zone of the crust, elevated temperatures in hydrothermal

systems are known to affect the permeability of fault zones with time. Sibson
(1996) highlighted that a large proportion of the structural permeability
developed in the fault-fracture meshes is likely to be transient. This is because
microcracks, fractures and faults may all become infilled with low-permeability
hydrothermal precipitates, alteration products, and/or clay-rich gouge. Tenthorey
et al. (1998) and Aharonov et al. (1998) related decreases in permeability to
authigenic mineral formation during diagenesis experiments at elevated
temperatures. This has particular relevance to understanding earthquake
processes, as many recent models of the earthquake cycle invoke the formation
of low-permeability minerals seals within fault zones in the interseismic period
(e.g. (Blanpied et al. 1992, Byerlee 1990, Lockner & Byerlee 1995, Miller et al.
1996)). Such mineral seals may aid the development of high fluid pressure
conditions within the fault zone, which is one of the explanations for the apparent
weakness of the San Andreas fault that is implied by the lack of a measurable
heat flow anomaly (Brune et al. 1969, Lachenbruch & Sass 1980) and orientation

165
of principal stresses driving fault movement (Zoback et al. 1987). Such results
also support the hypothesis that fault seals can form in less time that the
recurrence interval for larger earthquakes at the base of the seismogenic zone.
Geophysical evidence exists for healing processes, for example the Landers fault
zone in California was shown to exhibit time-dependent increase in seismic
velocity following the 1992 earthquake, only interrupted by the nearby 1999
Hector Mine earthquake (Vidale & Li 2003). This healing and compaction of the
damaged rocks is thought to be a key process in the temporal weakening of large
faults by increasing the pore pressure over time (Sleep 1995).

Previous experimental studies investigating healing properties of fault zones

have concentrated on two problems; healing of shear fractures and healing of
microfracture networks. One is important for fault-zone healing, the other for
damage-zone healing. (Morrow et al. 2001) studied the reduction of
permeability in granite under hydrothermal conditions. They conducted
experiments on Westerly granite at effective pressures of 50 MPa and at
temperatures from 150º to 500ºC, simulating conditions in the earthquake-
generation portion of fault zones. They showed that fractured samples have
higher rates of permeability reduction than intact samples and that mineral
dissolution and precipitation rates are fast enough even at lower temperatures of
their experiments to produce significant changes in permeability over times of
years to decades. Their experiments showed that a fractured sample (i.e. sample
with a post-failure shear fracture) of Westerly granite exhibited a steady decline
in permeability due to hydrothermal dissolution and mineral growth, beginning at
a value almost 2 orders of magnitude higher than that of the intact sample they
tested and reaching a near constant rate of decline after ~30 days. In this period
the sample reduced in permeability by an order of magnitude and a half. The
initial permeability measurements of the samples with a pre-existed shear
fracture at the beginning of their experiments compare well to the values of
permeability shown at sample failure in this study. This would suggest that if the
samples deformed in this study were subjected to similar experimental conditions
and elevated temperatures described by (Morrow et al. 2001), if not subjected to
further loading the samples could regain up to 90% of its initial in a relatively
short amount of time.

166
The experiments of (Morrow et al. 2001) showed the healing potential of a shear
fracture in hydrothermal conditions. The healing potential of the rock mass pre-
failure also needs to be considered, i.e. before a macroscopic shear fracture
develops and fluid flow is controlled entirely by the microfracture matrix.
(Brantley et al. 1990) looked at the healing of microcracks in quartz at elevated
temperatures and discussed the implications for fluid flow. They suggest that the
rate of microcrack healing under hydrostatic stagnant conditions show
dependence on fluid pressure, chemistry and crack dimensions, and that when
new cracks are not being produced in rocks at elevated temperatures and
pressures, fractures will have a wide range of lifetimes. Their results
demonstrated that faster healing rates are observed in smaller cracks, and they
suggest macrofracture will transport most of the fluid volume and seal relatively
slowly. Microcracks will allow pervasive penetration of fluid into the rock mass,
but heal quickly. This suggests that the pre-failure microfractures that cause the
two orders of magnitude increase in permeability are likely to heal considerably
faster than post-failure shear-fracture permeability, suggesting pre-failure
permeability recovery occurs at a faster rate than post-failure permeability
recovery. This is also supported by supported by the results of (Blanpied et al.
1992), where they inferred that the material either side of a saw-cut fault healed,
leading to elevated pore pressures and fault weakening. Permeability
enhancement through faulting and fracturing in fault-related hydrothermal
systems competes with permeability reduction as a consequence of fluid flow
and precipitation (Henley 1984). Hence deformation within fault-fracture
meshes, either continual or intermittent, is probably necessary for them to remain
effective as high-permeability structural conduits (Sibson 1994).

167
4.5.5 Implications for crustal fluid flow and fault zone
hydraulics

As discussed in the introduction to this chapter, recent work by Miller et al.

(2004) has suggested that earthquakes can initiate large fluxes of fluid using the
‘fault valving’ process first suggested by Sibson (1990). Aftershocks were
modelled as being triggered by a high pressure fluid front that migrated up the
fault plane from depth, initiated by damage associated with the mainshock. In
this model, however, the fluid flow characteristics of the fault zone, as well as
the physical dimensions of the damage zone, were tailored to fit the observations
and still await experimental and field verification. In this model, values for
permeability are calculated to be 4 x 10-11 m2 in the damage zone of the fault.
The results in this study show that stressed crystalline rock pre-failure has
permeabilities nearly seven to nine orders of magnitude lower than that predicted
by Miller et al. (2004) suggesting that such fluid flow rates could not be
produced by the pervasive microfracture of the rock matrix. Manning &
Ingebritsen (1999) inferred permeabilities of 10-16 m2 for crustal permeability in
tectonically stable environments, and Townend & Zoback (2000) suggested in
situ bulk permeabilities of ~10-17 to 10-16 m2 which is still one to two orders of
magnitude higher than measured immediately before failure in this study. The
highest permeabilities measured were those immediately post failure at 4 x 10-18
m2, which is approaching the suggested crustal permeability. The data presented
here concur with inferences that bulk fluid flow and dilatancy must be dominated
by larger scale structures, such as macrofractures and jointing in the absence of
lithological variety. If diffusion models are to be used, then it is important to
base these models on macro-scale fracture geometries as well as matrix-
dominated flow. Studies have demonstrated high permeabilities on the scale of
that calculated by Miller et al. (2004) in rough fractures in granite and marble,
but at low effective normal stress (Lee & Cho 2002). Matthai & Belayneh
(2004) used field data-based finite-element simulations of flow partitioning
between fractures and a permeable rock matrix to reveal critical aperture values
that mark the transition from matrix- to fracture-dominated flow. They
-13 -16
suggested that matrix permeabilities of 0.001-1 D (9.87 x 10 to 9.87 x 10

168
m2) are required for the matrix to either dominate, or contribute significantly to,
the total flow. This also highlights the need for further quantitative studies of
macroscopic damage surrounding fault zones. As discussed in the previous
section, microfractures in hydrothermal seismogenic zones are short-lived in
comparison to macrofractures due to preferential healing rates in smaller
fractures. This serves to add to the dominance of macro-scale fluid flow, as even
after crack opening events such as earthquakes, microscale structures will heal
relatively quickly and recover permeability.

It has been shown that an exponential decrease in microcrack density with

distance from the fault plane is commonly recorded around faults in low porosity
rock, which is predicted by various models. Mechanical models of faulting
suggest that the overall fault structure of large displacement zones is established
early in a fault’s history (Chester et al. 2004, Chester et al. 2005, Evans et al.
2000). For example, once faulting has initiated, some models suggest that the
bulk of the fracture damage is imported immediately prior to fault formation in
the ‘process zone’ surrounding the fault tip (Evans et al. 2000, Scholz et al. 1993,
Vermilye & Scholz 1998). Fault wear models attribute the bulk of fault damage
to the cumulative fracture damage that results from continued slip on pre-existing
fault surfaces (Scholz 1987, Wilson et al. 2003). More recently work has
focused on fracture damage that is created by the elastic strain energy released
during dynamic rupture in the form of dynamic slip pulses which can cause
considerable off-fault damage in the form of general off-fault secondary failure
or pulverised rock at shallower levels (Dor et al. 2006b, Rice et al. 2005). It is
likely that some or all of the processes described by these models contribute to
bulk fault zone fracture damage, with the signatures of more recent damage
overprinting evidence of earlier processes, which makes it less clear as to how
fracture damage is partitioned between each fault process.

169
Figure 4.14 shows a schematic diagram summarizing the possible permeability
structure inferred from the results in this study surrounding a fault with fracture
damage variations as described by the previously described models. The upper
diagram shows a three dimensional block with a strike slip fault cutting through
it.

Figure 4.14 - Schematic diagram illustrating relative states of stress and permeability
surrounding a fault plane with a highly stressed fault tip and associated microfracture
damage. The diagram is related to the scale of the experiment.

170
Microfracture damage surrounds the fault zone, with the most intense
microfracture density closest to the fault plane, with microfracture density
decreasing with distance from the fault. At the fault tip, a process zone (marked
are region A) exists, as described above, that relaxes the local stress at the fault
tip. This is the pre-failure region, where stresses range from low at the edge of
the process zone to high at the fault tip. The pre-failure region is represented by
the section marked A on the stress strain curve. In this region, permeability
correspondingly increases with increased porosity due to creation of
microfracture damage, with the greatest increase in permeability being closest to
the crack tip and decreasing exponentially to the edge of the process zone. The
variation in permeability could be up to two orders of magnitude within the
process zone, with the greatest permeability variation at the tip of the fault and
the smallest at the outer limit of the process zone. When the stress at the tip
reaches critical level and the fracture propagates, there is a corresponding stress
drop and the strength of the same area where the fault now exists falls to around
the frictional strength of the fault (region B). Permeability is greatly enhanced in
this shear fracture, and fluid flow initially preferentially occurs here. The results
from this study suggest that within this shear fracture the permeability could be
up to three orders of magnitude higher than intact rock permeabilities.
Conversely, in region C (the damage zone) outside of the newly created
permeable shear fracture, the stress-drop acts to relieve stress on the
microfracture matrix to remote stress levels and cracks close, decreasing the
permeability. However, as shown in this study, permeability hysteresis due to
permanent damage induced in the rock will mean that permeability is not be
reduced to initial levels and can remain somewhat higher. Hydrothermal fluids
within the microfracture matrix will work to seal fractures through dissolution
and mineralization, while continued fault wear and stress cycling will serve to
induce damage in the wall rock surrounding the fault and raise permeability.
Microfracture damage surrounding the fault will be representative of peak stress
conditions, rather than repeated stress cycling causing cumulative damage. This
schematic model will be relatively scale- independent, as process zones (and
hence damage zone) can range from millimetres to hundreds of metres (Scholz et
al. 1993). However, this is rather a simplistic view, as fault zones evolve and
many more fault growth and evolution processes become involved.

171
It is important to note that the work discussed here applies primarily to faults that
crosscut low porosity crystalline rock. Although the enhancements of
permeability are significant, lithological variations will have a varying effect on
permeability development (Brace 1980), and more experimentation on other
crystalline rocks is also needed as the upper crust is not composed entirely of
granite. It is also important to consider fracture processes related to dynamic
rupture, as it has been shown that dynamic rupture lead to more pervasive
microfracture damage, such as shown by (Dor et al. 2006b) and (Wilson et al.
2005).

172
4.6 Conclusions

This study presents data on the continuous evolution of permeability of

crystalline rocks under increasing differential stress. Previous data have been
limited by experimental techniques that require loading to be stopped for
permeability to be measured, and have considerably less measurements. Use of
the pore pressure oscillation technique has allowed the measurement of
continuous permeability evolution without such limitations. Data suggest that
with increased differential stress and strain permeability increases by up to and
over two orders of magnitude up till sample failure. In tests where post-failure
permeability is was measured where shear fractures (macrofracture) had
developed in the samples, the permeability is over three orders of magnitude
higher than initial intact permeabilities, as high as 4 x 10-18 m2 which is
approaching measurements of in-situ bulk crustal permeabilities of ~10-17 to 10-16
m2 (Townend & Zoback 2000). Increasing amplitude cyclic loading test show
both strain and permeability hysteresis related to each cycle, with permeability
increasing by 400-500% of the starting permeability of each of the first four
cycles, with starting permeabilities increasing by 200-300% each cycle. Above a
peak cycle of 80% failure stress, permeability increases by over an order of
magnitude before failure, suggesting that the higher levels of differential stress
are very important for permeability evolution and have not been examined in
previous studies.

Permeability creation will be offset in nature by hydrothermal precipitation at

elevated temperatures in a relatively short space of time (e.g. relative to
earthquake cycles). Hence repeated fracturing would be required to keep the
rock mass acting as viable fluid conduits. Pre-failure microfractures that cause
the two orders of magnitude increase in permeability in the samples are likely to
heal considerably faster than post-failure shear-fracture permeability due to
studies suggesting faster healing rates are observed in smaller cracks.

The results show that stressed crystalline rock pre-failure has permeabilities
nearly seven to nine orders of magnitude lower than that predicted by high

173
pressure diffusive models such as Miller et al. (2004), suggesting that such fluid
flow rates cannot be maintained through the pervasive microfracture matrix.
Despite two orders of magnitude variation in permeability, pre-failure
permeabilities are still one to two orders of magnitude lower than in-situ crustal
permeabilities, although post failure at 4 x 10-18 m2 begin to approaching the
suggested crustal permeability. The data agree with inferences that bulk fluid
flow and dilatancy must be dominated by larger scale structures, such as
macrofractures and jointing. This highlights the need for further quantitative
studies of macroscopic damage surrounding fault zones so that diffusion models
can be based on macro-scale fracture geometries as well as matrix-dominated
flow. From the results shown in this study, it is suggested that the permeability
of the highly stressed fault tip process zone in low-permeability crystalline rocks
could increase by up to two orders of magnitude, and once the fracture
propagates the stress in the resulting damage zone is reduce causing decreased
permeability due to reduced stresses surrounding a fault plane. Permeability in
the damage zone will be higher than starting permeabilities due to hysteresis,
with fluid flow being preferable in the main through going fault with
permeability variations of up to three orders of magnitude.

174
5 Discussion and further work

The primary aims of this thesis and the specific problems to be addressed were
set out in chapter 1. Chapters 2 and 4 have presented results and discussion and
addressed these problems, and so will not be discussed in any great detail here.
Here a final summary will be presented, outlining suggestions for further work.
Chapter 2 evaluated the typical physical dimensions of fault damage zones
surrounding 6 strike-slip faults, and how they vary with increasing displacement.
Primary conclusions were that fault damage zone widths are shown to scale in
size with increasing fault displacement, and empirical relationships allow
damage zone widths and fracture density to be estimated within damage zones of
other faults with known displacements. Chapter 4 show data from experiments
that measured experimentally the pre-yield permeability and porosity evolution
with progressive brittle deformation of initially intact quartzofeldspathic
crystalline rocks. Primary conclusions were that permeability was shown to rise
by up to two orders of magnitudes before sample failure, and once sample failure
had occurred, permeability was seen to be up to three orders of magnitude higher
than intact sample permeability.

In this thesis, field studies have shown fault damage zones host both microscopic
and macroscopic fracture damage, and have shown empirical relationships that
allow fracture damage at all scales to be predicted. It is suggested that the
damage zone should exhibit a high permeability and hence contribute
significantly to the bulk transport properties of faults (Scholz 2002, Vermilye &
Scholz 1998). As fluid flow in fault damage zones will be controlled by a
combination of macroscopic fracture networks and microscopic damage, such
empirical relationships enables the quantification and prediction of bulk fluid
flow surrounding faults. The experimental measurements of permeability
evolution made during triaxial compression of initially intact crystalline rocks
can be used as a proxy for varying degrees of fracturing found within fault zones.
One of the final aims was to develop a typical model for the structure and fluid
flow properties of large strike-slip fault zones, and apply this model provide a
framework to assess the permeability structure surrounding fractures and faults

175
and the implications for fault zone hydraulics. The schematic diagram shown in
Figure 4.14 in chapter 4 shows a semi-qualitative model of how both field and
experimental data can be combined to make inferences on fluid flow properties
surrounding faults. However, with the data provided in this thesis and some
further thin section work, it is possible to link both lab and field data in a larger
scale model to predict bulk fluid flow surrounding faults of all sizes. This is
beyond the scope of this thesis, but future work should aim to quantify
experimentally-induced microfractures and associated permeability
measurements as were done for microfractures for natural samples in this study.
Microfracture counts on experimentally deformed samples that have been taken
to various percentages of peak stress while the permeability has been measured
could be completed, and microfracture counts within these samples could be
made. By relating the fracture densities surrounding natural fault zones with
densities seen in experimental deformed samples with known permeabilities,
modelling techniques can then be applied to gain estimates of bulk fluid flow of
the fracture networks. This will provide a basis for predicting the influence of
pore fluid pressures on important geological issues, such as crustal strength.

In addition to understanding fluid flow within fractured rocks surrounding faults,

a very important question is how rapidly faults seal after a large earthquake, and
whether fault zones can completely seal during interseismic period (i.e. return to
permeabilties similar to that of intact samples). This is important for both fluid
flow at depth in the Earth’s crust and for earthquake generation. Healing and
sealing in fault zones by secondary mineral precipitation as a direct consequence
of fluid flow can help maintain high pore fluid pressures which can weaken
faults and render them more susceptible to earthquakes. Geophysical evidence
exists for this process, where the Landers fault zone in California was shown to
exhibit time-dependent increase in seismic velocity following the 1992
earthquake, only interrupted by the nearby 1999 Hector Mine earthquake (Vidale
& Li 2003). This healing and compaction of the damaged rocks is thought to be
a key process in the temporal weakening of large faults by increasing the pore
pressure over time (Sleep 1995). Future work could test this experimentally by
the application of moderate temperature (200-250ºC - commensurate with those
found at seismogenic depths) once samples have been experimentally deformed,

176
while continuing to monitor the permeability and porosity with time, and will
help to determine rates of closure of earthquake induced microcrack dilatancy.
Continuous monitoring of the porosity and permeability of the deformed samples
at these moderate temperatures would help determine the rates of closure of
microcrack dilatancy due to low temperature processes over time periods ranging
from days to weeks. In addition to using water as the pore fluid, using the inert
gas argon can help identify physicochemical interactions between the pore fluid
and the rock. The role of subcritical crack growth is a potentially important
process, but how it affects the bulk permeability of the sample with time is not
currently understood.

Also, fault damage zones can run through porous sedimentary rocks in the near
surface or in subsiding sedimentary basins, or through crystalline basement rocks
at deeper crustal levels. In this study, focus is concentrated on compact,
crystalline basement rocks, as these are the rocks commonly found at
seismogenic depths. However, there are many fundamental differences in fault
zone structure between sedimentary rocks and basement rocks. Fault zones in
basement rocks will have permeable damage zones as shown in this study, but
those in porous sediments and sedimentary rocks often do not. The latter
properties are important for fault seal of hydrocarbons, for CO2 sequestration,
waste storage and for underground water flow. Fundamental questions are:
What are the differences in permeability structures between sedimentary rocks
and basement rocks and at what stages of sediment compaction and cementation
do the changes in permeability structure occur? What is the maximum seal
capacity of sedimentary rocks with faults and fractures? Fault zones in
sedimentary rocks are of vital significance for accretionary prism and subduction
zones (fluid flow and earthquakes). Hence, similar studies (both field and
experimental) on sedimentary rocks will yield important information.

More recently, many seismologists have worked on dynamic rupture propagation

on bimaterial interfaces, e.g., fault zones with different rock types (and hence
different seismic velocities) on both sides of fault zone, or the slip zone. There is
an incompatibility created along faults in such cases that causes ‘wrinkle-like’
pulses in the normal direction of fault zones (Andrews & Ben-Zion 1997). The

177
pulses may form asymmetric structures and "pulverized rocks" on the high
velocity side (e.g. Dor et al. 2006a). These wrinkle-like pulses can cause
instantaneous reduction in normal stress and fault can be dynamically weakened.
It has been suggested that this effect can be much more important than high
velocity weakening of faults. Thus studies on bimaterial rupture propagation,
both in the field and lab will be important.

178
6 Appendix A – Experimental calibrations
The following calibrations are referred to in section 3.3.1.

Calibration chart for the confining pressure transducer

300
y = 39.064x + 2.4085
2
R = 0.9998 Test 1 - Up
y = 39.266x + 2.2398 Test 1 - Down
250 R2 = 0.9999 Test 2 - Up
Test 2 - Down
y = 39.07x + 0.6913
2
Linear (Test 1 - Up)
R = 0.9999
Linear (Test 1 - Down)
y = 39.068x + 1.8565 Linear (Test 2 - Up)
200
R2 = 0.9999 Linear (Test 2 - Down)
Pressure (Mpa)

150

100

50

0
-1 0 1 2 3 4 5 6 7
Voltage (Volts)

Graph showing confining pressure transducer calibrations

Calibration chart for the upstream and downstream pore fluid pressure transducers

250 y = 24.103x + 7.7549

2
R = 0.9999

y = 23.561x + 7.5653 Increasing P - Upstream

2 Increasing P - Downstream
R = 0.9999
Decreasing P - Upstream
y = 24.235x + 7.9099 Decreasing P - Downstream
200 2
R =1 Linear (Increasing P - Upstream)
Linear (Increasing P - Downstream)
y = 23.687x + 7.6021
2 Linear (Decreasing P - Upstream)
R =1
Linear (Decreasing P - Downstream)

150
Pressure (Mpa)

100

50

0
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00
Voltage (Volts)

Graph showing upstream and downstream pore pressure transducer calibrations

179
 Displacement transducer calibration - Voltage vs Displacement volume

250

200
Displacement volume (mm^3)

Piston going in 150

Piston going out
y = -124.74x - 35.822 Linear (Piston going in)
R2 = 0.9996 Linear (Piston going out)

100

50

y = -125.94x - 39.71
R2 = 0.9997

0
-2.5 -2 -1.5 -1 -0.5 0
Voltage (Volts)

Graph showing LVDT calibration of pore fluid controller

Force gauge calibration (Voltage vs Load) using load cell

200

180

160
y = -71.347x + 83.978
R2 = 0.9998
Up
140
y = -70.839x + 85.204 Down
2
R = 0.9999 Linear (Up)
120 Linear (Down)
Load (kN)

100

80

60

40

20

0
-2 -1.5 -1 -0.5 0 0.5 1 1.5
Voltage (Volts)

Graph showing force gauge LVDT calibration

180
 Voltage vs axial displacement calibration

3.5

y = 0.8111x + 0.3816
2
R =1
3 Piston going in
y = 0.8029x + 0.4127 Piston going out
R2 = 0.9999 Linear (Piston going in)
Linear (Piston going out)
2.5
Axial displacement (mm)

1.5

0.5

0
0 0.5 1 1.5 2 2.5 3 3.5 4
Voltage (v)

Axial displacement LVDT calibration

Confining pressure vs load on force gauge

25

y = 0.1145x + 0.7363
R2 = 0.9995 Up
Down
20 y = 0.1144x + 0.8372 Linear (Up)
R2 = 0.9995 Linear (Down)
Load on force gauge (kN)

15

10

0
0 50 100 150 200 250
Confining pressure (MPa)

Load applied to the force gauge by confining pressure.

181
7 Appendix B - Papers published and in-press

This section enclosed two papers for which I am second author on. Both papers
present some data collected by me, which has been discussed already in this
thesis, and some of the figures are similar to ones that are presented earlier in this
thesis. . I was responsible for sections of the data, text and various figures.

In the paper entitled “Slip on ‘weak’ faults by the rotation of regional stress in
the fracture damage zone” (Nature, 2006) 30% of the data is from my own work,
with contributions from other authors.

In the paper entitled “On the Structure and Mechanical Properties of Large
Strike-slip Faults” (Geol. Soc. Special publication, in press), up to 50% of the
data is from my own work..

182
7.1 Slip on ‘weak’ faults by the rotation of regional stress
in the fracture damage zone. Nature, 2006.

183
184
185
186
7.2 On the Structure and Mechanical Properties of Large
Strike-slip Faults. Geol. Soc. Special publication, in
press.

D. R. Faulkner1, T. M. Mitchell1, E. H. Rutter2 and J. Cembrano3

1
Rock Deformation Laboratory, Dept. of Earth and Ocean Sciences, University
of Liverpool, Liverpool, UK
2
Rock Deformation Laboratory, School of Earth, Environment and Atmospheric
Sciences, University of Manchester, Manchester, UK
3
Departamento de Ciencias Geologicas, Universidad Catolica del Norte,
Antofagasta, Chile

Abstract

Elucidation of the internal structure of fault zones is paramount for

understanding their mechanical, seismological and hydraulic properties. In order
to observe representative brittle fault zone structures, it is preferable that the fault
be passively exhumed from seismogenic depths and the exposure must be in arid
or semi-arid environments where the fragile rocks are not subject to extensive
weathering. Field observations of two such faults are used to constrain their
likely mechanical properties. One fault is the Carboneras fault in southeastern
Spain, where the predominant country rocks are phyllosilicate-rich lithologies,
and the other is part of the Atacama fault system in northern Chile, where faults
pass through crystalline rocks of acidic to intermediate composition. The
Carboneras fault is a sinistral fault with several tens of kilometres offset
exhumed from approximately 4 km depth, and displays multiple strands of clay-
bearing fault gouge, each several metres wide, that contain variably fractured
lenses of protolithic mica schists. The strain is evenly distributed across the
gouge layers, in accordance with the measured laboratory mechanical behaviour
which shows predominantly strain hardening characteristics. The overall width of
the fault zone is several hundred metres. Additionally, there are blocks of
dolomitic material that are contained within the fault zones that show extremely
localized deformation in the form of faults several centimetres wide. These are
typically arranged at an angle of ~20 degrees to the overall fault plane. These
differing types of fault rock products allow for the possibility of ‘mixed mode’
seismicity, with fault creep occurring along the strands of velocity strengthening
clay-rich gouge, punctuated by small seismic events that nucleate on the velocity
weakening localized faults within the dolomite blocks. The Caleta Coloso fault in
northern Chile has a sinistral offset of at least 5 km and exhumed from 5 to 10
km depth. The fault core is represented by a 200 – 300 m wide zone of
hydrothermally-altered protocataclasite and ultracataclasite. This is surrounded
by a zone of micro- and macro-fractures on the order of 150 m thick. The fault
core shows a heterogeneous distribution of strain, with alternate layers of
ultracataclasite and lower strain material. The strain-weakening behaviour of
crystalline rocks might be expected to produce highly localized zones of

187
deformation, and thus the wide core zone must be a result of additional process
such as precipitation strengthening, or geometric irregularities along the fault
plane.

Introduction

Understanding the structure of fault zones is essential in order to help interpret

their mechanical, hydraulic and seismological properties. Faults are key
structures in the Earth's upper crust that control the strength of the lithosphere
(Kusznir and Bott, 1977; Kohlstedt et al., 1995; Townend and Zoback, 2000), the
flow of fluids (Caine et al., 1996; Knipe, 1992) and the nucleation and
propagation of earthquakes. Often we want to describe the structure of faults at
depth in the crust (for example at seismogenic depths), but these regions can only
be accessed by remote geophysical methods, unless deep drilling is an option
(Zoback et al., 2006; Ohtani et al., 2001) or observation from deep mines. Even
where drillholes are present, a one-dimensional view of a fault zone is obtained,
where a two- or three-dimensional view is necessary in order to understand fully
the fault properties.

Field studies of exhumed fault zones can aid in the understanding of fault zone
structure at depth, but must be approached with caution. Often, fault zones are
active during the exhumation process and as such have a full range of structures
that overprint those that formed at the depth of interest. The primary structures
(those formed at depth) and the overprinting structures may not be easily
distinguishable. Consequently structures formed under near-surface conditions
may be wrongly interpreted as key features in exhumed surface-exposed fault
zones. For the above reasons field studies on fault zones are often made on
passively-exhumed fault zones where it can be demonstrated that little or no
displacement and structural overprinting occurred during exhumation either from
the regional tectonics or from the nature and mineralogy of the fault rocks.

Faults that are not well-exposed can present problems as key parts of the fault
zone may be disregarded solely because they are not observable at the surface.
Another potential problem with using natural faults for study concerns collection
of material from surface exposures for laboratory testing or analysis. Near-
surface weathering can significantly alter the mineralogy, microstructure and
physical properties of rocks, as demonstrated by Morrow and Lockner (1994).
Again, incomplete exposure may lead to the over-emphasis of laboratory results
obtained from material collected from the exposed material, whereas the
mechanical and hydraulic properties may well be primarily controlled by
unexposed or poorly exposed parts of the fault zone.

Given the difficulties of studying large faults in the field, it is hardly surprising
that few detailed studies exist of their structure. Exceptions to this include work
by Chester et al. (1993), Schulz and Evans (2000), Wibberley and Shimamoto
(2003), Faulkner et al. (2003), Cembrano et al. (2005) and Jefferies et al. (2006).
All of these examples study faults with variable exposure levels, varying
constraints on their exhumation histories and as such give different degrees of
confidence when interpreting the structure of faulting at depth.

188
In this work, we aim to contribute to the database on the internal structure of
large strike-slip faults, and infer some aspects of the mechanical properties of
faults from observations of their internal structure. First we describe the structure
of two major strike-slip faults that cut through very different country rocks.
Aspects of the structure of the Carboneras fault in southeastern Spain are
described, then those from the Caleta Coloso fault in northern Chile. The likely
mechanical properties of the two faults are inferred from the macroscopic
structure of the two faults in the discussion.

Fault zone structure

Both the Carboneras and the Caleta Coloso faults have been largely passively
exhumed from depth, and they are excellently exposed in semi- and hyper-arid
environments. The mineralogy of the fault rock in the Carboneras fault and the
Caleta Coloso fault indicates fault activity at ~4 km and ~6 km depth
respectively. The depth range from which the Carboneras fault was exhumed is
estimated from the present height of uplifted basement rocks and Pliocene reef
complexes and the temperatures at which the neocrystallized clay phases
contained within the fault gouges formed (assuming an average geothermal
gradient). This yields values of exhumation between 4 and 1.5 km (Faulkner et
al., 2003), which is consistent with recent uplift rates estimated from marine
terraces on the order of 0.05–0.1 mm/year (Bell et al., 1997). The fault
accommodates several tens of km of movement across it (Rutter et al., 1986). On
the Caleta Coloso fault, fission track dating of fault-related fracturing has shown
that the presently exposed fault rocks formed at a temperature of at least 100ºC;
interpreted to indicate a minimum of 3 km depth (Cembrano et al., 2006). The
Caleta Coloso fault has undergone more recent movements, although these are
limited in extent, and are in response to large subduction zone earthquakes
associated with the offshore trench (Gonzalez and Carrizo, 2003). These
movements are easily recognized in the field and expressed as very localized,
narrow fault planes that form small fault scarps at the surface. The Caleta Coloso
fault has at least 5 km of strike-slip offset (Cembrano et al., 2005).

The results of the above analyses indicate that the structures of both faults can be
taken to be representative of faulting at depth. Both faults, by virtue of their
offset, are crustal-scale structures. The constituent structures of each fault are
now described in turn, and the interpretation of the mechanical significance of
these is analysed later in the discussion.

189
Carboneras fault, S.E. Spain

The macroscopic structure of the Carboneras fault has been described in detail
previously in Faulkner et al. (2003). Here some key aspects of the structure are
summarized. The fault cuts through Alpine age (Palaeogene) metamorphic rocks,
Miocene age marine sediments and volcanic rocks. In the area of study (see
figure 1) the fault cuts predominantly through the metamorphic basement rocks
that consist largely of graphitic mica schist. The southern edge of the fault is
obscured somewhat by later volcanic rocks (lava flows and lahars), but along-
strike observations suggest that very little of the total fault width is unexposed.
The igneous rocks contained within the fault zone are dolerites and show
intrusive relationships with very little deformation. They have been interpreted
previously as feeder systems for the later volcanics (Rutter et al., 1986). The fault
initiated in the early Miocene (Scotney et al., 2000) and, as stated previously,
ceased significant activity between 5 and 10 Ma ago. The fault has sinistral offset
on it, as determined from slickenlines and movements indicators constructed
from the measurement of the orientation of P foliations and R1 Riedel shears
present within the gouge layers.

Figure 1. Regional map showing the location of the Carboneras

fault within the tectonic framework of southeastern Spain. SLF =
Sierra de Los Filabres, SA = Sierra Alhamilla, SC = Sierra
Cabrera, NB = Níjar Basin, SB = Sorbas Basin, VB = Vera Basin.

The macroscopic fault zone structure is that of a fault consisting of multiple,

anastomosing strands of fault gouge that contain variably fractured lenses of the
country rock (figure 2). Overall the width of the zone of deformation is on the

190
order of several hundred metres. As the fault cuts through mica-rich schists, the
fault rocks within the gouge layers are phyllosilicate-rich, clay-bearing fault
gouges. The typical mineralogical assemblage in these gouges includes
muscovite/illite, chlorite, quartz with minor amounts of gypsum, albite and
graphite. The layers of fault gouge are typically on the order of several metres
wide and the distribution of strain within the gouge layers is very homogeneous
(figure 3 a and b).

Large blocks of dolomite are also included within the fault zone, and these
display a completely different style of faulting to that developed within the mica-
rich rocks. Instead of wide layers of deformed rock with an even distribution of
strain across them, the faults in the dolomite are very localized, narrow features.

Figure 2. Summary geological map of the Carboneras fault zone to the area immediately
northeast of the village of El Saltador, southeastern Spain. The anastomosing layers of fault
gouge are highlighted as are the very localized narrow fault planes developed within the blocks
of dolomite contained within the fault zone (after Faulkner et al., 2003).

Figure 3c shows one of these localized fault planes, that has a slickenlined
mechanical wear surface. The orientation of these fault planes is typically
oblique to the general strike of the Carboneras fault; they are arranged in an en
echelon pattern and make an angle of 10 to 25º measured anticlockwise from the
strike of the main fault zone (figure 2). Thus these faults are in a R1 Riedel
orientation.

191
Figure 3. Photographs of meso-scale deformation structures within the Carboneras fault zone. (a)
phyllosilicate-rich fault gouge layer developed within the graphitic mica schist (b) homogeneous
distribution of strain within each phyllosilicate-rich gouge layer. Hammer head is approximately
20cm in length. (c) localized, polished fault plane developed within one of the dolomite blocks
contained within the Carboneras fault zone (d) detail of one of the localized faults within the
dolomite. Hammer head is approximately 20cm in length.

Caleta Coloso fault, N. Chile

The large-scale structure of the Caleta Coloso fault in northern Chile has
previously been described by Cembrano et al. (2005). The fault zone is part of
the Atacama fault system, which is an arc-parallel strike-slip structure that
accommodated some of the oblique convergence between the Nazca and South
American plates (figure 4). Presently the Atacama fault system is largely passive,
with transcurrent plate motions occurring on fault systems much further to the
East (e.g. Falla Oeste). In this work, we describe the results of some new detailed
fault zone mapping conducted on part of the Caleta Coloso fault surrounding
Quebrada Isabel (figure 5). In the area of study, the fault zone cuts through
plutonic rocks of predominantly granodioritic composition. It has a sinistral
sense of shear determined from the orientation of slickenlines and offset planar
markers (e.g. dykes) within the fault zone.

192
 The fault core
structure is
approximately 200 to
300m wide and is
characterized by
zones of fractured
protolith, cataclasite
and ultracataclasite.
Hydrothermal
precipitation of
chlorite and epidote
within the fault core
has led to a
characteristic green
colouration that is
restricted to the fault
core (figure 6).
Geothermometry
conducted on chlorite
contained within
veins, some of which
have been affected by
later fault movements,
Figure 4. Regional tectonic framework of the Caleta Coloso fault in shows temperatures
northern Chile. It is part of the Atacama fault system. Nazca and around 250ºC.
South American plates labelled. Assuming that the
fluids that deposited these veins had not travelled a great distance, their
temperature gives an additional constraint on the maximum depth of faulting, if a
reasonable geothermal gradient is used. For a magmatic arc this gradient may be
up to 40ºC indicating a depth of around 6 km.

193
Figure 5. Detailed map of the Caleta Coloso fault around Quebrada Isabel within the Coastal
Cordillera to the south of Antofagasta. Drift deposits are indicated by a lighter shading of the
solid geology deposits. The specific location of the field study area is shown by the
latitude/longitude GPS position (S23º55.280 W070º24.809) on the map.

Figure 6. Photograph of the part of the Caleta Coloso fault to the south of Quebrada Isabel. The
fault core is marked and is characterized by a green colouration (indicated) related to the
precipitation of chlorite and epidote. The length of the arrow indicating the fault core is
approximately 150m long.

194
The distribution of strain within the zones of cataclasite (figure 7a) and
ultracataclasite (figure 7b) is not as clear as within the gouge layers of the
Carboneras fault zone. There are angular blocks of granodiorite contained within
an ultracataclasite matrix containing hydrothermally-precipitated chlorite and
epidote (figure 7a, c and e). At the edge of the cataclasite zones, strain is often
localized into a very narrow plane (figure 7c and d). The damage zone consists of
macro- and microscopic fracturing that decays with distance from the fault core
(Faulkner et al., 2006).

Figure 7. Photographs showing characteristic meso-scale structures associated with the Caleta
Coloso fault. (a) ‘network’ zone showing angular blocks of granodiorite contained within a
cataclasite and ultracataclasite matrix containing hydrothermally-precipitated chlorite and epidote
(b) ultracataclasite zone (c) and (d) localization of strain on the boundary of one of the strands of
cataclasite (e) gradation of strain into a cataclasite layer.

Discussion

The structure of the fault zones of the Carboneras and Caleta Coloso faults is, at
first glance, quite similar. Both have a wide fault core (on the order of several
hundred metres) that consists of multiple strands of fault-related rocks. However,
there are key differences between the faults and these are discussed below. First,
we concentrate on what might be the general controls on fault zone structure.
Second, the mechanical significance of the macroscopic structure of the two
faults is discussed.

195
Controls on fault zone architecture

The primary controls on fault zone architecture are (a) the depth of faulting (b)
the protolith (c) the fault displacement and (d) the interaction with other faults
and/or pre-existing structures. In the case of crustal, plate boundary-scale faults
their angle with the local plate motion vector can also have an effect on the
structure. One other possible control on fault zone structure is the mode of
faulting itself. Recent work on the San Andreas fault has shown that seismic
rupture may have produced ‘pulverization’ of the surrounding rock although this
has yet to be shown to occur at significant depth (Wilson et al., 2005; Dor et al.,
2006).

In this work we have described the fault rocks of the Carboneras and Caleta
Coloso faults that formed at ~4 and ~6 km depth respectively. The difference
between these depths is not large, and hence the control exerted by depth (control
(a)) on the relative structure of the two faults should not be a major factor. We
consider the protolith to be an important factor in the development of the
structure of the two faults (control (b)) and this is considered in more detail later
in the discussion. With regard to displacements across the two faults (control
(c)), they have ~40 (Carboneras) and >5km (Caleta Coloso) accumulated
displacement. They are both crustal-scale features. Although the displacements
of the two faults are different, we consider that displacements of greater than a
few km will produce a macroscopic structure that is fairly typical of the depth
and protolith, unless geometric factors or interactions with other faults occur.
The view of Ben Zion and Sammis (2003) that fault zones evolve towards
simplicity does not adequately explain the complexity of high strain features seen
across the Carboneras fault zone, for example (some of which must have
developed outside of the fault’s early history). The structure of the Carboneras
fault can be compared with that of the Punchbowl fault, a trace of the San
Andreas fault exhumed from similar depth and with a comparable displacement
(Chester et al., 1993), where almost all of the strain was accommodated within a
fault core of 20-30 cm width. Clearly, factors in addition to fault displacement
play a role in determining fault zone structure.

On the subject of interactions with other faults (control (d)), the observations in
this work were made in areas that were some distance from other faults with any
significant displacement on them. However, approximately 4 km to the northeast
of where observations and mapping of the Carboneras fault were made, the fault
is cut by the Palomares fault, another many-km-offset regional strike-slip fault
(see figure 1). However, the activity on this fault is thought to post-date
movements on the Carboneras fault (Faulkner et al., 2003) and consequently
should have little effect on the development of the structure of the fault.

We interpret the differences in the structure of the two faults arising from
variations in depth, displacement and interactions with other faults to be of
secondary importance. Hence we compare the structures of the two faults in
terms of the protolith (and its mechanical response to deformation) from which
they formed and post-formation processes. We first review briefly the factors
responsible for changes in the mode of failure of rocks before considering in

196
more detail the protolith and consequent mechanical response leading to
variations in structure of each fault.

As depth increases in the crust, so the strength of brittle rocks also increases. As
confining pressure increases, the mode of deformation of rocks can change from
localized to distributed (Rutter, 1986). At low temperatures, localized
deformation is often associated with a mechanical response whereby the rock

Figure 8. Shows the comparative appearance of experimentally compressed cylindrical rock

samples and how these commonly relate to stress/strain curves under moderate to low
temperature conditions. Thus ductile (distributed) deformation is often not related to a post-
failure drop in flow stress, whereas localized brittle deformation is usually associated with a
stress drop and continued focusing of deformation into the fault zone.
loses the ability to support as much differential stress once the failure stress has
been exceeded. This is seen as the development of a fault plane in experiments
(figure ). In contrast, distributed deformation is commonly associated with a
mechanical response in which the rock retains the ability to support differential
stress on the same order as the failure stress after this has been reached (figure 8).
However, it is worth noting that localized behaviour can develop in a strain
hardening regime (Hobbs et al., 1990), although it is more usually the case that
distributed deformation will occur. When we consider the mode of failure of
rocks, nothing is implied about the mechanism of failure.

For rocks deforming by brittle deformation mechanisms, a transition between

localized and distributed deformation can occur as a result of increasing effective
pressure (Bernabe and Brace, 1990) and would indicate that the strength of the
intact rock is approximately equal to that of the resulting fault rock. This change
can also be accommodated by the collapse of porosity within the intact rock
(Rutter and Hadizadeh, 1991), the presence of a ‘weak’ second phase (Byerlee,

197
1968) or a contribution from some crystal plastic mechanism (Fredrich et al.,
1989).

If we relate the mechanical data back to the fault zones studied, we may make the
following observations. First, all fault zones are likely indicative of strain
weakening, and the fact they exist as discrete structures within the surrounding
country rock indicates that they continue to be weak relative to the country rock.
However, the structure of the fault zone may give some idea as to the ‘ductility’
of the fault zone, i.e. the relative strength between the country rock and its
faulted equivalent within the fault core. The greater the contrast between the
strength of the intact rock and its faulted equivalent, the more localized the fault
zone. In the examples considered in this paper, the fault zones are relatively
wide, and contain multiple zones of fault gouge or cataclasite. We now seek to
explain the structure of these faults by consideration of the mechanical properties
of the faults and the protolith from which they are formed.

Carboneras Fault Structure; Mechanical Interpretation

In the case of the Carboneras fault, the mechanical strength contrast between the
mica schist country rock and the phyllosilicate-rich fault gouge produced from it
can inferred to be small, given the width of the fault zone. If the strength of the
country rock was much greater than the fault gouge, then a much more localized,
and hence narrow, zone of deformation might be expected.
The apparently even distribution of strain within the individual fault gouge layers
can be explained by the mechanical characteristics of phyllosilicate-rich fault
gouge. In experiments, strain hardening behaviour is typically seen (Morrow et
al., 1982; Rutter et al., 1986; Logan and Rauenzahn, 1987; Morrow et al., 1992),
although velocity weakening has been noted in some phyllosilicates, notably
serpentinites (Reinen, 2000) and under very low effective stress (Saffer at al.,
2001). The strength increase with strain within fault gouge layers may also be
partly responsible for the width of the fault zone (Faulkner et al., 2003).

Other explanations for the wide nature of the fault zones studied can be
considered. If a restraining or releasing bend occurs along the fault, then this also
may spread the deformation over a much wider zone (giving rise to positive or
negative flower structures). However, in the case of both the Carboneras fault
and the Caleta Coloso fault, regional mapping of the fault zones shows no such
restraining bend (as discussed previously), and qualitative observations along
strike for several kilometres along both faults indicate that the fault structure
mapped in detail is not unique and is characteristic of the fault zones along strike.

The predominant type of fault motion one might expect from a fault zone
containing the materials found within the Carboneras fault is fault creep. The
mechanical characteristics of phyllosilicate-rich fault gouge and the lack of any
localization within the gouge layers suggest that the fault was strain hardening,
which would preclude any periods of instability. The presence of the dolomitic
blocks within the fault zone, with their contrasting, localized style of faulting,
opens up the possibility of ‘mixed mode’ palaeo-seismicity on the Carboneras
fault. These types of fault rock indicate strain weakening behaviour, and might
well have been seismogenic. Hence one could envisage predominantly aseismic

198
creep on a fault such as the Carboneras fault, punctuated by seismic events. It is
interesting to note that this type of behaviour is the same as is seen on the San
Andreas fault around the area of Parkfield, and recent results from the SAFOD
borehole indicate that the fault zone at 3 km depth consists of multiple fault
strands, similar to that of the Carboneras fault (Zoback et al., 2006). A fuller
discussion of the similarities between the Carboneras fault zone structure and the
San Andreas fault structure at depth around the area of Parkfield (from
geophysical observations) is presented in Faulkner et al. (2003). The size of
earthquake that could arise from faulting on included brittle blocks contained
within a creeping fault depends on the size of the block. In the case of the
dolomites in the Carboneras fault zone, earthquake size would be small <Mw4
(Wells and Coppersmith, 1994).

Caleta Coloso Fault Structure; Mechanical Interpretation

Low temperature faulting in granites at any crustal depth is localized. It is

sometimes initiated on pre-existing joint planes, such has been interpreted by
Martel et al. (1988) for the Mount Abbott quadrangle faults in the Sierra Nevada,
or the Gole Larghe fault within the Adamello batholith (Di Toro and
Pennacchioni, 2005) in the Italian Alps. In laboratory experiments, low
temperature faulting produced in intact crystalline rocks at any confining
pressures such as would be encountered within the brittle crust display localized
behaviour (e.g. Lockner, 1998). Strain rate would also play a role, but the strain
rate in typical geological processes is such that the deformation mode is expected
to be localized brittle. Given the body of evidence from the field and laboratory,
faulting in crystalline rocks deformed within the brittle crust might be expected
to be highly localized in the absence of geometrical irregularities. However, the
nature of the faulting seen on the Caleta Coloso fault displays, at first glance, a
similar mesoscale structure to that on the Carboneras fault, that of a wide fault
zone consisting of multiple strands of fault rocks.

The character of the fault core seen on the Caleta Coloso fault, with large scale
precipitation of predominantly chlorite and epidote, indicates a high flux of
hydrothermal fluids (see figure 6). These fluids may perhaps be a result of the
fault being a part of the Mesozoic Andean magmatic arc, where geothermal
gradients were elevated and devolatilization of magmatic products may have led
to enhanced fluid movements. Other present day strike-slip faults in
geothermally active areas such as the Dead Sea Transform or the Salton Sea
cannot be compared to the Caleta Coloso fault as the seismogenic parts of the
fault have not been exposed at the surface. The near-surface structure of these
faults may not be representative of the structure of the deeper parts of the fault
zone.

The fluids in the Caleta Coloso fault do not appear to have reacted significantly
with the host rock as little phyllosilicate material or alteration is observed on the
mesoscale. We interpret the structure of the fault to be influenced strongly by
precipitation hardening of the fault zone leading to strengthening and subsequent
failure along a new plane parallel to the fault zone. There are experimental data
that support the idea of rapid mineral precipitation and strengthening in quartz-
rich fault gouges at hydrothermal conditions (Olsen et al., 1998; Tenthorey et al.,

199
2003; Yashuhara et al., 2005), and over time periods commensurate with the
interseismic period, processes such as these are very likely to happen.

The nature of the strands of fault rocks within the Caleta Coloso fault zone are
also very different to those seen on the Carboneras fault. The distribution of
strain is very heterogeneous, with very localized planes of comminution,
typically confined to the edges of the strands of cataclasite (figure 7c). This
observation is consistent with the notion of strengthening of cataclasite layers
due to precipitation, with later movements accommodated on new fault gouge
strands confined to the boundary of the strengthened old cataclasite layers.

Conclusions

We have studied in detail the internal structure of two large strike-slip faults; the
Carboneras fault in southeastern Spain and the Caleta Coloso fault in northern
Chile. The protoliths for each fault are very different. The Carboneras fault cuts
through predominantly phyllosilicate-rich metamorphic rocks, whereas the
Caleta Coloso fault affects igneous crystalline rocks. Both faults have been
largely passively exhumed from depth so that their present day surface structure
can be viewed as representative of the structure of the faults while active at
depth. Both faults are very well-exposed in arid environments ensuring the fault
rocks have minimal degradation from surface weathering effects.

The Carboneras fault displays multiple strands of phyllosilicate-rich fault gouge

in a fault zone with a total width approaching 750m. The anastomosing fault
gouge strands are typically less than 10m thick and they contain lenses of
variably fractured protolith between them. The strain is apparently evenly
distributed within each gouge strand. We interpret the structure of the fault to
represent a small strength contrast between the protolith and the resultant fault
gouge (semibrittle or ductile behaviour in the mechanical sense). The strain
hardening characteristics of phyllosilicate-rich fault gouge may also contribute
towards the bulk structure and the uniform distribution of strain across the gouge
layers.

The Caleta Coloso fault incorporates multiple strands of variably deformed rock
including protocataclasite and ultracataclasite in a zone at least 200m wide. The
fault core zone is characterized by a green colouration produced by the
precipitation of chlorite and epidote in veins and also in the fault rock.
Localization of slip is seen typically on the boundaries of ultracataclasite layers.
The mechanical behaviour of these low-porosity crystalline rocks would be
expected to be very localized and brittle, and hence we interpret the wide nature
of the fault zone to be a product of precipitation healing of the fault zone leading
to spatially and temporally localized strengthening that spreads the deformation
over a wider zone.

Acknowledgements

This work was supported by NERC grant NE/C001117/1 (to DRF). Erik Jensen
Siles and Nicolas Reyes are thanked for field assistance.

200
References

Bell, J.W., Amelung, F. and King, G., 1997. Preliminary late quaternary slip
history of the Carboneras fault, southeastern Spain. J. Geodynamics, 24: 51-66.

Bernabe, Y. and Brace W.F., 1990. Deformation and fracture of Berea sandstone.
In: The brittle-ductile transition in rocks – The Heard volume, eds. A.G. Duba,
W.B. Durham, J.W. Handin, H.F. Wang. American Geophysical Union
Monograph 56: 91-101

Ben-Zion, Y. and Sammis, C.G., 2003. Characterization of fault zones. Pure and
Applied Geophysics, 160: 677-715.

Byerlee, J.D., 1968. Brittle-ductile transition in rocks. J. Geophys. Res., 73:

4741-4750.

Caine, J.S., Evans, J.P. and Forster, C.B., 1996. Fault zone architecture and
permeability structure. Geology, 24: 1025-1028.

Cembrano, J., González, G., Arancibia, G., Ahumada, I., Olivares, V. and
Herrera, V., 2005. Fault zone development and strain partitioning in an
extensional strike-slip duplex: a case study from the Atacama Fault System,
northern Chile. Tectonophysics, 400: 105-125.

Cembrano, J., Faulkner, D.R., Herrera, V. and Mitchell, T.M., 2006. On the
formation of and fluid pressure within strike-slip dilational jogs: constraints on
crustal stresses at depth. Submitted to Geology, August 2006.

Chester, F.M., Evans, J.P. and Biegel, R.L., 1993. Internal structure and
weakening mechanisms of the San Andreas fault. J. Geophys. Res., 98: 771-786.

Di Toro, G., Pennacchioni, G., 2005. Fault plane processes and mesoscopic
structure of a strong-type seismogenic fault in tonalites (Adamello batholith,
Southern Alps). Tectonophysics, 402: 55-80.

Dor, O., Ben-Zion, Y., Rockwell, T.K. and Brune, J.N. 2006. Pulverized rocks in
the Mojave section of the San Andreas fault zone, Earth Planet. Sci. Lett., 245,
642-654,
doi:10.1016/j.epsl.2006.03.034.

Faulkner, D.R., Lewis, A.C. and Rutter, E.H., 2003. On the internal structure and
mechanics of large strike-slip faults: field observations from the Carboneras
fault, southeastern Spain. Tectonophysics, 367: 235-251.

Faulkner, D.R., Mitchell, T.M., Healy, D. and Heap, M.J., 2006. Slip on 'weak'
faults by the rotation of regional stress in the fracture damage zone. Nature, 444:
922-925. doi:10.1038/nature05353.

Fredrich, J. T., Evans, B., and Wong, T.-F., 1989. Micromechanics of the Brittle
to Plastic Transition in Carrara Marble. J. Geophys. Res., 94: 4129–4145.

201
Kohlstedt, D.L., Evans, B., Mackwell, S.J., 1995. Strength of thelithosphere:
constraints imposed by laboratory experiments. J. Geophys. Res., 100: 17587–
17602.

Knipe, R.J., 1992. Faulting processes and fault seal. Norwegian Petroleum
Society (NPF) Special Publication, 1: 325-342.

Gonzalez, G., and Carrizo, D., 2003. Segmentación, cinemática y cronología

relativa de la deformación tardía de la Falla Salar del Carmen, Sistema de Fallas
de Atacama, (23°40'S), norte de Chile. Revista Geológica de Chile, 30: 223-244.

Hobbs, B. E., Muhlhaus, H.B. and Ord, A., 1990. Instability, softening and
localization of deformation. In: Deformation Mechanisms, Rheology and
Tectonics, eds. R. J. Knipe and E. H. Rutter. Geol. Soc. Spec. Publ., 54: 143-165.

Jefferies, S. P., Holdsworth, R. E., Wibberley, C. A. J., Shimamoto, T., Spiers, C.

J., Niemeijer, A. R. and Lloyd, G. E., 2006. The nature and importance of
phyllonite development in crustal-scale fault cores: an example from the Median
Tectonic Line, Japan. J. Structural Geology, 28: 220-235.

Kusznir, N.J. and Bott, M.H.P., 1977. Stress-concentration in upper lithosphere

caused by underlying viscoelastic creep. Tectonophysics, 43 (3-4): 247-256.

Lockner, D. A., 1998. A generalised law for brittle deformation of Westerly

granite. Journal of Geophysical Research, 103: 5107-5123.

Logan, J.M. and Rauenzahn, K.A., 1987. Velocity-dependent behaviour of mixed

quartz-montmorillonite gouge. Tectonophysics, 144: 87-108.

Martel, S.J., D. D. Pollard, and P. Segall, 1988. Development of simple strike-

slip fault zones, Mount Abbotquadrangle, Sierra Nevada, California. Geol. Soc.
Am. Bull. 100: 1451-1465.

Morrow, C.A., Shi, L.Q. and Byerlee, J.D., 1982. Strain hardening and the
strength of clay-rich fault gouges. J. Geophys. Res., 87: 6771-6780.

Morrow, C.A., and Lockner, D.A., 1994. Permeability differences between

surface-derived and deep drillhole core samples. Geophys. Res. Lett., 21 (19):
2151-2154.

Morrow, C.A., Radney, B. and Byerlee, J.D., 1992. Frictional strength and the
effective pressure law of montmorillonite and illite clays. In: Fault Mechanics
and Transport Properties of Rocks, eds. Evans, B. and Wong, T.-F., pp. 69-88,
Academic Press Ltd.

Ohtani, T., Tanaka, H., Fujimoto, K., Higuchi, T., Tomida, N. and Ito, H., 2001.
Internal structure of the Nojima Fault zone from the Hirabayashi GSJ drill core.
The Island Arc, 10: 392. doi:10.1046/j.1440-1738.2001.00337.x

202
Olsen, M.P., Scholz, C.H. and Léger, A., 1998. Healing and sealing of a
simulated fault gouge under hydrothermal conditions: Implications for fault
healing. J. Geophysical Research, 103: 7421-7430. DOI: 10.1029/97JB03402.

Reinen, L., 2000. Slip styles in a spring-slider model with a laboratory-derived

constitutive law for serpentinite. Geophysical Research Letters, 27: 2037-2040.

Rutter, E.H. and Hadizadeh, J., 1991. On the influence of porosity on the low-
temperature brittle-ductile transition in siliciclastic rocks. J. Struct. Geol., 13:
609–614.

Rutter, E.H., 1986. On the nomenclature of mode of failure transitions in rocks.

Tectonophysics, 122: 381-387.

Rutter, E.H., Maddock, R.H., Hall, S.H., and White, S.H., 1986. Comparative
microstructures of natural and experimentally produced clay-bearing fault
gouges. Pure and Applied Geophysics, 124: 3-30.

Saffer, D.M., Frye, K.M., Marone, C., and Mair, K., 2001. Laboratory results
indicating complex and potentially unstable frictional behaviour of smectite clay.
Geophys. Res Lett., 28: 2297-2300.

Schulz, S.E. and Evans, J.P., 2000. Mesoscopic structure of the Punchbowl Fault,
Southern California and the geologic and geophysical structure of active strike-
slip faults. J. Structural Geology, 22: 913-930.

Scotney, P., Burgess, R. and Rutter, E.H., 2000. 40Ar/39Ar age of the Cabo de
Gata volcanic series and displacements on the Carboneras fault zone, SE Spain.
J. Geol. Soc. Lond., 157: 1003-1008.

Tenthorey, E., Cox, S. F. and Todd, H. F., 2003. Evolution of strength recovery
and permeability during fluid-rock reaction in experimental fault zones. Earth
Planet. Sci. Lett., 206: 161–172.

Townend, J., and Zoback, M.D., 2000. How faulting keeps the crust strong.
Geology, 28: 399-402.

Wells, D.L. and Coppersmith, K.J., 1994. New empirical relations among
magnitude, rupture length, rupture width, rupture area and surface displacement.
Bull. Seism. Soc., 84: 974-1002.

Wibberley C.A.J. and Shimamoto T., 2003. Internal structure and permeability of
major strike-slip fault zones: the Median Tectonic Line in Mie Prefecture,
Southwest Japan. J. Structural Geology, 25: 59-78.

Wilson, B., Dewers, T. Reches, Z. and Brune J. 2005. Particle size and energetics
of gouge from earthquake rupture zones, Nature, 434 (7), 749-752.

203
Yasuhara H., Marone, C. and Elsworth, D., 2005. Fault zone restrengthening and
frictional healing: The role of pressure solution. J. Geophysical Research, 110:
B06310, doi:10.1029/2004JB003327.

Zoback, M. D., Hickman, S. H. and Ellsworth, W., 2006. Preliminary results

from the SAFOD experiment. Geophys. Res. Abs., 8: 02474.

204
8 References

Aharonov, E., Tenthorey, E. & Scholz, C. H. 1998. Precipitation sealing and

diagenesis - 2. Theoretical analysis. Journal of Geophysical Research-
Solid Earth 103(B10), 23969-23981.
Anders, M. H. & Wiltschko, D. V. 1994. Microfracturing, Paleostress and the
Growth of Faults. Journal of Structural Geology 16(6), 795-815.
Anderson, E. M. 1942. The dynamics of faulting and dyke formation with
applications to Britain. Oliver and Boyd, Edinburgh, 191pp.
Andrews, D. J. & Ben-Zion, Y. 1997. Wrinkle-like slip pulse on a fault between
different materials. Journal of Geophysical Research-Solid Earth
102(B1), 553-571.
Atkinson, B. K. & Meredith, P. G. 1987. The theory of subcritical crack growth
with application to minerals and rocks. In: Fracture Mechanics of Rock
(edited by Atkinson, B. K.). Academic Press, London, 111–166.
Aydin, A. 1978. Small Faults Formed as Deformation Bands in Sandstone. Pure
and Applied Geophysics 116(4-5), 913-930.
Aydin, A. & Johnson, A. M. 1978. Development of Faults as Zones of
Deformation Bands and as Slip Surfaces in Sandstone. Pure and Applied
Geophysics 116(4-5), 931-942.
Barenblatt, G. I. 1962. The mathematical theory of equilibrium cracks in brittle
fracture. Adv. Appl. of Mech.(7), 25.
Ben-Zion, Y. & Sammis, C. G. 2003. Characterization of fault zones. Pure and
Applied Geophysics 160(3-4), 677-715.
Bernabe, Y., Mok, U. & Evans, B. 2006. A note on the oscillating flow method
for measuring rock permeability. International Journal of Rock
Mechanics and Mining Sciences 43(2), 311-316.
Blanpied, M. L., Lockner, D. A. & Byerlee, J. D. 1992. An Earthquake
Mechanism Based On Rapid Sealing Of Faults. Nature 358(6387), 574-
576.
Brace, W. F. 1980. Permeability of Crystalline and Argillaceous Rocks.
International Journal of Rock Mechanics and Mining Sciences 17(5),
241-251.
Brace, W. F., Walsh, J. B. & Frangos, W. T. 1968. Permeability of Granite under
High Pressure. Journal of Geophysical Research 73(6), 2225-&.
Brantley, S. L., Evans, B., Hickman, S. H. & Crerar, D. A. 1990. Healing of
Microcracks in Quartz - Implications for Fluid-Flow. Geology 18(2), 136-
139.
Brock, W. G. & Engelder, T. 1977. Deformation Associated with Movement of
Muddy Mountain Overthrust in Buffington Window, Southeastern
Nevada. Geological Society of America Bulletin 88(11), 1667-1677.
Brown, M., Diaz, F. & Grocott, J. 1993. Displacement History of the Atacama
Fault System 25-Degrees-00's-27-Degrees-00's, Northern Chile.
Geological Society of America Bulletin 105(9), 1165-1174.
Bruhn, R. L., Parry, W. T., Yonkee, W. A. & Thompson, T. 1994. Fracturing and
Hydrothermal Alteration in Normal-Fault Zones. Pure and Applied
Geophysics 142(3-4), 609-644.

205
Bruhn, R. L., Yonkee, W. A. & Parry, W. T. 1990. Structural and Fluid-
Chemical Properties of Seismogenic Normal Faults. Tectonophysics
175(1-3), 139-157.
Brune, J. N., Brown, S. & Johnson, P. A. 1993. Rupture Mechanism and
Interface Separation in Foam Rubber Models of Earthquakes - a Possible
Solution to the Heat-Flow Paradox and the Paradox of Large Overthrusts.
Tectonophysics 218(1-3), 59-67.
Brune, J. N., Henyey, T. L. & Roy, R. F. 1969. Heat flow, stress, and rate of slip
along the San Andreas Fault, California. Journal of Geophysical
Research 74, 3821-3827.
Byerlee, J. 1990. Friction, Overpressure and Fault Normal Compression.
Geophysical Research Letters 17(12), 2109-2112.
Caine, J. S., Evans, J. P. & Forster, C. B. 1996. Fault zone architecture and
permeability structure. Geology 24(11), 1025-1028.
Carslaw, H. S. & Jaeger, J. C. 1959. Conduction of heat in solids. 2nd Edition,
Oxford University Press.
Cembrano, J., Faulkner, D. R., Herrera, V. & Mitchell, T. M. in review. On the
formation of and fluid pressure within strike-slip dilatational jogs:
constraints on crustal stresses at depth.
Cembrano, J., Gonzalez, G., Arancibia, G., Ahumada, I., Olivares, V. & Herrera,
V. 2005. Fault zone development and strain partitioning in an extensional
strike-slip duplex: A case study from the Mesozoic Atacama fault system,
Northern Chile. Tectonophysics 400(1-4), 105-125.
Chester, F. M. & Chester, J. S. 2000. Stress and deformation along wavy
frictional faults. Journal of Geophysical Research-Solid Earth 105(B10),
23421-23430.
Chester, F. M., Chester, J. S., Kirschner, D. L., Schulz, S. E. & Evans, J. P. 2004.
Structure of large-displacement, strike-slip fault zones. in Rheology and
Deformation in the Lithosphere at Continental Margins (eds Karner, G.
D., Taylor, B., Driscoll, N. W. & Kohlstedt, D. L.) (Columbia Univ.
Press, New York).
Chester, F. M., Evans, J. P. & Biegel, R. L. 1993. Internal Structure and
Weakening Mechanisms of the San-Andreas Fault. Journal of
Geophysical Research-Solid Earth 98(B1), 771-786.
Chester, F. M. & Logan, J. M. 1986. Implications for Mechanical-Properties of
Brittle Faults from Observations of the Punchbowl Fault Zone,
California. Pure and Applied Geophysics 124(1-2), 79-106.
Chester, J. S., Chester, F. M. & Kronenberg, A. K. 2005. Fracture surface energy
of the Punchbowl fault, San Andreas system. Nature 437(7055), 133-136.
Chester, J. S. & Fletcher, R. C. 1997. Stress distribution and failure in anisotropic
rock near a bend on a weak fault. Journal of Geophysical Research-Solid
Earth 102(B1), 693-708.
Covey-Crump, S. J. 1992. Application of a state variable description of inelastic
deformation to geological materials. Unpublished unpublished Ph.D.
thesis thesis, University of London.
Cowie, P. A. & Scholz, C. H. 1992. Physical Explanation for the Displacement
Length Relationship of Faults Using a Post-Yield Fracture-Mechanics
Model. Journal of Structural Geology 14(10), 1133-1148.

206
Cowie, P. A. & Scholz, C. H. 1992a. Displacement Length Scaling Relationship
for Faults - Data Synthesis and Discussion. Journal of Structural Geology
14(10), 1149-1156.
Cowie, P. A. & Scholz, C. H. 1992b. Physical Explanation for the Displacement
Length Relationship of Faults Using a Post-Yield Fracture-Mechanics
Model. Journal of Structural Geology 14(10), 1133-1148.
Cowie, P. A. & Shipton, Z. K. 1998. Fault tip displacement gradients and process
zone dimensions. Journal of Structural Geology 20(8), 983-997.
Cox, S. J. D. & Scholz, C. H. 1988. On the Formation and Growth of Faults - an
Experimental-Study. Journal of Structural Geology 10(4), 413-430.
Dawers, N. H. & Anders, M. H. 1995. Displacement-Length Scaling and Fault
Linkage. Journal of Structural Geology 17(5), 607-&.
Dengo, C. A. 1982. A Structural-Analysis of Cores from the Leg-67 Transect
across the Middle America Trench - Offshore Guatemala. Initial Reports
of the Deep Sea Drilling Project 67(Nov), 651-666.
Di Toro, G. & Pennacchioni, G. 2005. Fault plane processes and mesoscopic
structure of a strong-type seismogenic fault in tonalites (Adamello
batholith, Southern Alps). Tectonophysics 402(1-4), 55-80.
Dor, O., Ben-Zion, Y., Rockwell, T. K. & Brune, J. 2006a. Pulverized rocks in
the Mojave section of the San Andreas Fault Zone. Earth and Planetary
Science Letters 245(3-4), 642-654.
Dor, O., Rockwell, T. K. & Ben-Zion, Y. 2006b. Geological observations of
damage asymmetry in the structure of the San Jacinto, San Andreas and
Punchbowl faults in Southern California: A possible indicator for
preferred rupture propagation direction. Pure and Applied Geophysics
163(2-3), 301-349.
Dresen, G. & Guéguen, Y. 2004. Damage and Rock Physical Properties. In:
Mechanics of Fluid Saturated Rocks (edited by Guéguen, Y. & Boutéca,
M.). International Geophysics 89. Elsevier Academic Press, 169-217.
Dugdale, D. S. 1960. Yielding of Steel Sheets Containing Slits. Journal of the
Mechanics and Physics of Solids 8(2), 100-104.
Engelder, J. T. 1974. Cataclasis and Generation of Fault Gouge. Geological
Society of America Bulletin 85(10), 1515-1522.
Engvik, A. K., Bertram, A., Kalthoff, J. F., Stockhert, B., Austrheim, H. &
Elvevold, S. 2005. Magma-driven hydraulic fracturing and infiltration of
fluids into the damaged host rock, an example from Dronning Maud
Land, Antarctica. Journal of Structural Geology 27(5), 839-854.
Evans J.P., S. Z. K., Pachell M.A., Lim S.J. and Robeson K. 2000. The structure
and composition of exhumed faults, and their implications for seismic
processes. In: Proceedings of the 3rd Conference on Tectonic Problems
of the San Andreas Fault System, . Bokelmann, G. and Kovach, R.L.
(eds.) Stanford University Publications, Geological Sciences 21, 14.
Evans, J. P. 1990. Thickness Displacement Relationships for Fault Zones.
Journal of Structural Geology 12(8), 1061-1065.
Evans, J. P. & Chester, F. M. 1995. Fluid-Rock Interaction in Faults of the San-
Andreas System - Inferences from San-Gabriel Fault Rock Geochemistry
and Microstructures. Journal of Geophysical Research-Solid Earth
100(B7), 13007-13020.
Evans, J. P., Shipton, Z. K., Pachell, M. A., Lim, S. J. & Robeson, K. 2000. The
structure and composition of exhumed faults, and their implications for

207
 seismic processes. In: Proceedings of the 3rd Conference on Tectonic
Problems of the San Andreas Fault System, . Bokelmann, G. and Kovach,
R.L. (eds.) Stanford University Publications, Geological Sciences 21, 14.
Faulkner, D. R. 1997. The role of clay-bearing fault gouges in controlling fluid
pressures in fault zones: implications for fault mechanics. Ph.D. Thesis,
University of Manchester, UK, 1997, 279pp.
Faulkner, D. R., Lewis, A. C. & Rutter, E. H. 2003. On the internal structure and
mechanics of large strike-slip fault zones: field observations of the
Carboneras fault in southeastem Spain. Tectonophysics 367(3-4), 235-
251.
Faulkner, D. R., Mitchell, T. M., Healy, D. & Heap, M. J. 2006. Slip on 'weak'
faults by the rotation of regional stress in the fracture damage zone.
Nature 444(7121), 922-925.
Faulkner, D. R., Mitchell, T. M., Rutter, E. H. & Cembrano, J. in press. On the
Structure and Mechanical Properties of Large Strike-slip Faults. GSL
Spec. Pub. paper.
Faulkner, D. R. & Rutter, E. H. 1998. The gas permeability of clay-bearing fault
gouge at 20ºC. in Faults, Fault Sealing and Fluid Flowin Hydrocarbon
Reservoirs, edited by G. Jones, Q. Fisher, and R.J. Knipe, Geol. Soc.
Spec. Pub. 147, 147-156.
Faulkner, D. R. & Rutter, E. H. 2000. Comparisons of water and argon
permeability in natural clay-bearing fault gouge under high pressure at 20
degrees C. Journal of Geophysical Research-Solid Earth 105(B7), 16415-
16426.
Faulkner, D. R. & Rutter, E. H. 2001. Can the maintenance of overpressured
fluids in large strike-slip fault zones explain their apparent weakness?
Geology 29(6), 503-506.
Feng, R. & McEvilly, T. V. 1983. Interpretation of Seismic-Reflection Profiling
Data for the Structure of the San-Andreas Fault Zone. Bulletin of the
Seismological Society of America 73(6), 1701-1720.
Fischer, G. J. 1992. The determination of permeability and storage capacity: Pore
pressure oscillation method. In: "Fault Mechanics and Transport
Properties of Rocks" Evans, B. and Wong, T.-F (eds). Academic Press
Ltd, 187-212.
Fischer, G. J. & Paterson, M. S. 1992. Measurement of permeability and storage
capacity in rocks during deformation at high temperature and pressure.
In: "Fault Mechanics and Transport Properties of Rocks" Evans, B. and
Wong, T.-F (eds). Academic Press Ltd, 214-252.
Flinn, D. 1977. Transcurrent fault and associated cataclasis in Shetland. J. Geol.
Soc. London 133, 17.
Flinn, D. 1979. The Walls Boundary fault, Shetland, British Isles. In: Proc. Conf.
VIIIm Analysis of Actual Fault Zones in Bedrockm U.S. Geol. Surv., Open
File Rep., 1160.
Friedman, M. 1963. Petrofabric Analysis of Experimentally Deformed Calcite-
Cemented Sandstones. Journal of Geology 71(1), 12-&.
Friedman, M. 1969. Structural analysis of fractures in cores from the Saticoy
Field, Ventura Co., California. American Association of Petroleum
Geologists Bulletin 53, 367-389.
Gillespie, P. A., Walsh, J. J. & Watterson, J. 1992. Limitations of Dimension and
Displacement Data from Single Faults and the Consequences for Data-

208
 Analysis and Interpretation. Journal of Structural Geology 14(10), 1157-
1172.
González, G. 1990. Patrones estructurales, modelo de ascenso emplazamiento y
deformación del Plutón Cerro Cristales, Cordillera de la Costa al Sur de
Antofagasta, Chile. Memoria para optar al título de Geólogo (inédita),
Universidad Católica del Norte, Antofagasta, 124.
González, G. 1996. Evolución tectónica de la Cordillera de la Costa de
Antofagasta (Chile). In: especial referencia las deformaciones
sinmagmáticas del Jurásico Cretácico Inferior. Berliner
Geowissenschaftliche Abhandlungen (A), Band 181, 111 p. .
González, G. 1999. Mechanismo y profundidad de emplazamiento del Plutón
Cerro Cristales, Cordillera de la Costa, Antofagasta, Chile. . Revista
Geológica de Chile 26, 23.
González, G., Cembrano, J., Carrizo, D., Macci, A. & Schneider, H. 2003. The
link between forearc tectonics and pliocene-quaternary deformation of the
Coastal Cordillera, northern Chile. Journal of South American Earth
Sciences 16(5), 321-342.
Haimson, B. & Chang, C. 2000. A new true triaxial cell for testing mechanical
properties of rock, and its use to determine rock strength and
deformability of Westerly granite. International Journal of Rock
Mechanics and Mining Sciences 37(1-2), 285-296.
Healy, D., Jones, R. R. & Holdsworth, R. E. 2006. Three-dimensional brittle
shear fracturing by tensile crack interaction. Nature 439(7072), 64-67.
Heap, M. J. & Faulkner, D. R. In press. Quantifying the evolution of static elastic
properties as crystalline rock approaches failure. International Journal of
Rock Mechanics and Mining Sciences.
Heiland, J. 2003. Permeability of triaxially compressed sandstone: Influence of
deformation and strain-rate on permeability. Pure and Applied
Geophysics 160(5-6), 889-908.
Henley, R. W. 1984. The geothermal framework of epithermal deposits. In:
Geology and Geochemistry of Epithermal Systems (edited byBerger, B.R.
& Bethke, P.M.). Soc. Econ. Geol. Rev. Econ. Geol 2, 1-24.
Hoffman-Rothe, A., Ritter, O. & Janssen, C. 2004. Correlation of electrical
conductivity and structural damage at a major strike-slip fault in northern
Chile. Journal Of Geophysical Research-Solid Earth 109(B10).
Hoffmann-Rothe, A. 2002. Combined structural and magnetotelluric
inversigation accross the West Fault Zone in northern Chile. Ph. D. diss.
University of Potsdam.
Hull, J. 1988. Thickness Displacement Relationships for Deformation Zones.
Journal of Structural Geology 10(4), 431-435.
Jaeger, J. C., Cook, N. G. & Zimmerman, R. W. 1969. Fundamentals of Rock
Mechanics. Blackwell Publishing.
Janssen, C., Wagner, F. C., Zang, A. & Dresen, G. 2001. Fracture process zone
in granite: a microstructural analysis. International Journal of Earth
Sciences 90(1), 46-59.
Jefferies, S. P., Holdsworth, R. E., Wibberley, C. A. J., Shimamoto, T., Spiers, C.
J., Niemeijer, A. R. & Lloyd, G. E. 2006. The nature and importance of
phyllonite development in crustal-scale fault cores: an example from the
Median Tectonic Line, Japan. Journal of Structural Geology 28(2), 220-
235.

209
Keaney, G. M. J., Meredith, P. G. & Murrell, S. A. F. 1998. Laboratory Study of
Permeability Evolution in ‘Tight’ Sandstone under Non-Hydrostatic
Stress Conditions. SPE/ISRM 47265, presented at the SPE/ISRM Eurock
’98 held in Trondheim, Norway, 8-10 July 1998.
Kim, Y. S., Peacock, D. C. P. & Sanderson, D. J. 2004. Fault damage zones.
Journal of Structural Geology 26(3), 503-517.
Knipe, R. J. 1992. Faulting processes and fault seal. In Norwegian Petroleum
Society (NPF) Special Publication 1, 325-342.
Kranz, R. L., Saltzman, J. S. & Blacic, J. D. 1990. Hydraulic Diffusivity
Measurements on Laboratory Rock Samples Using an Oscillating Pore
Pressure Method. International Journal of Rock Mechanics and Mining
Sciences 27(5), 345-352.
Lachenbruch, A. H. & Sass, J. H. 1980. Heat-Flow and Energetics of the San-
Andreas Fault Zone. Journal of Geophysical Research 85(Nb11), 6185-
6222.
Lavrov, A. 2003. The Kaiser effect in rocks: principles and stress estimation
techniques. International Journal of Rock Mechanics and Mining
Sciences 40(2), 151-171.
Lawn, B. R. & Wilshaw, T. R. 1975. Fracture of brittle solids. Cambridge
University Press, Cambridge.
Lee, H. S. & Cho, T. F. 2002. Hydraulic characteristics of rough fractures in
linear flow under normal and shear load. Rock Mechanics and Rock
Engineering 35(4), 299-318.
Little, T. A. 1995. Brittle Deformation Adjacent to the Awatere Strike-Slip-Fault
in New-Zealand - Faulting Patterns, Scaling Relationships, and
Displacement Partitioning. Geological Society of America Bulletin
107(11), 1255-1271.
Lockner, D. A. 1998. A generalized law for brittle deformation of Westerly
granite. Journal of Geophysical Research-Solid Earth 103(B3), 5107-
5123.
Lockner, D. A. & Byerlee, J. D. 1995. An earthquake instability model based on
faults containing high fluid-pressure compartments. Pure and Applied
Geophysics 145(3-4), 717-745.
Lockner, D. A., Byerlee, J. D., Kuksenko, V., Ponomarev, A. & Sidorin, A.
1991. Quasi-Static Fault Growth and Shear Fracture Energy in Granite.
Nature 350(6313), 39-42.
Lockner, D. A., Byerlee, J. D., Kuksenko, V., Ponomarev, A. & Sidrin, A. 1992.
Observations of quasi-static fault growth from acoustic emissions. in
Fault Mechanics and Transport Properties of Rocks, edited by B. Evans
and T.-f. Wong, 3-31, Academic, San Diego, Calif.,.
Manning, C. E. & Ingebritsen, S. E. 1999. Permeability of the continental crust:
Implications of geothermal data and metamorphic systems. Reviews of
Geophysics 37(1), 127-150.
Marrett, R. & Allmendinger, R. W. 1990. Kinematic Analysis of Fault-Slip Data.
Journal of Structural Geology 12(8), 973-986.
Martel, S. J., Pollard, D. D. & Segall, P. 1988. Development of Simple Strike-
Slip-Fault Zones, Mount Abbot Quadrangle, Sierra-Nevada, California.
Geological Society of America Bulletin 100(9), 1451-1465.
Matthai, S. K. & Belayneh, M. 2004. Fluid flow partitioning between fractures
and a permeable rock matrix. Geophysical Research Letters 31(7), -.

210
Miller, S. A., Collettini, C., Chiaraluce, L., Cocco, M., Barchi, M. & Kaus, B. J.
P. 2004. Aftershocks driven by a high-pressure CO2 source at depth.
Nature 427(6976), 724-727.
Miller, S. A., Nur, A. & Olgaard, D. L. 1996. Earthquakes as a coupled shear
stress high pore pressure dynamical system. Geophysical Research
Letters 23(2), 197-200.
Mitra, G. 1984. Brittle to Ductile Transition Due to Large Strains Along the
White Rock Thrust, Wind River Mountains, Wyoming. Journal of
Structural Geology 6(1-2), 51-61.
Mooney, W. D. & Ginzburg, A. 1986. Seismic Measurements of the Internal
Properties of Fault Zones. Pure and Applied Geophysics 124(1-2), 141-
157.
Moore, D. E. & Lockner, D. A. 1995. The Role of Microcracking in Shear-
Fracture Propagation in Granite. Journal of Structural Geology 17(1), 95-
114.
Morrow, C. A., Moore, D. E. & Lockner, D. A. 2001. Permeability reduction in
granite under hydrothermal conditions. Journal of Geophysical Research-
Solid Earth 106(B12), 30551-30560.
Mpodozis, C. & Ramos, V. A. 1990. The Andes of Chile and Argentina: Circum-
Pacific Council for Energy and Mineral resources. Earth Science Series
11, 31.
Olsen, M. P., Scholz, C. H. & Leger, A. 1998. Healing and sealing of a simulated
fault gouge under hydrothermal conditions: Implications for fault healing.
Journal of Geophysical Research-Solid Earth 103(B4), 7421-7430.
Paterson, M. S. 1964. Effect of Pressure on Youngs Modulus + Glass Transition
in Rubbers. Journal of Applied Physics 35(1), 176-&.
Paterson, M. S. & Wong, T.-F. 2005. Experimental rock deformation; the brittle
field. 2nd ed. . Berlin, Heidelberg, New York: Springer-Verlag. , 348pp.
Power, W. L. & Tullis, T. E. 1991. Euclidean and Fractal Models for the
Description of Rock Surface-Roughness. Journal of Geophysical
Research-Solid Earth and Planets 96(B1), 415-424.
Reches, Z. & Lockner, D. A. 1994. Nucleation and Growth of Faults in Brittle
Rocks. Journal of Geophysical Research-Solid Earth 99(B9), 18159-
18173.
Rice, J. R. 1992. Fault Stress States, Pore Pressure Distributions, and the
Weakness of the San Andreas Fault. in Fault Mechanics and Transport
Properties in Rocks (eds. B. Evans and T.-F. Wong), Academic Press, 28.
Rice, J. R., Sammis, C. G. & Parsons, R. 2005. Off-fault secondary failure
induced by a dynamic slip pulse. Bulletin of the Seismological Society of
America 95(1), 109-134.
Robertson, E. C. 1982. Continuous formation of gouge and breccia during fault
displacement. in: Issues in Rock Mechanics, edited by R.E. Goodman and
F. Hulse, American Institute of Mining and Engineering, New York, 397-
404.
Rudnicki, J. W. 1980. Fracture-Mechanics Applied to the Earths Crust. Annual
Review of Earth and Planetary Sciences 8, 489-525.
Rutter, E. H., Maddock, R. H., Hall, S. H. & White, S. H. 1986. Comparative
Microstructures of Natural and Experimentally Produced Clay-Bearing
Fault Gouges. Pure and Applied Geophysics 124(1-2), 3-30.

211
Sanyal, S. K., Pirnie, R. M., Chen, G. O. & Marsden, S. S. 1972. Novel Liquid
Permeameter for Measuring Very Low Permeability. Society of
Petroleum Engineers Journal 12(3), 206-&.
Saucier, F., Humphreys, E. & Weldon, R. 1992. Stress near Geometrically
Complex Strike-Slip Faults - Application to the San-Andreas Fault at
Cajon Pass, Southern California. Journal of Geophysical Research-Solid
Earth 97(B4), 5081-5094.
Scheuber, E. & Gonzalez, G. 1999. Tectonics of the Jurassic- Early Cretaceous
magmatic arc of the north Chilean Coastal Cordillera (22 degrees-26
degrees S): A story of crustal deformation along a convergent plate
boundary. Tectonics 18(5), 895-910.
Schlische, R. W., Young, S. S., Ackermann, R. V. & Gupta, A. 1996. Geometry
and scaling relations of a population of very small rift-related normal
faults. Geology 24(8), 683-686.
Scholz, C. H. 1987. Wear and Gouge Formation in Brittle Faulting. Geology
15(6), 493-495.
Scholz, C. H. 1990. The Mechanics of Earthquakes and Faulting. Cambridge
University Press, New York.
Scholz, C. H. 2002. The mechanics of Earthquakes and Faulting. 2nd edition,
Cambridge University Press, Cambridge.
Scholz, C. H. & Aviles, C. A. 1986. The fractal geometry of faults and faulting.
In Earthquake Source Mechanics, (eds. S. Das, J. Boatwright, and C.
Scholz). New York:Academic Press, 9.
Scholz, C. H., Dawers, N. H., Yu, J. Z. & Anders, M. H. 1993. Fault Growth and
Fault Scaling Laws - Preliminary-Results. Journal of Geophysical
Research-Solid Earth 98(B12), 21951-21961.
Scholz, C. H. & Lawler, T. M. 2004. Slip tapers at the tips of faults and
earthquake ruptures. Geophysical Research Letters 31(21), -.
Schulz, S. E. & Evans, J. P. 2000. Mesoscopic structure of the Punchbowl Fault,
Southern California and the geologic and geophysical structure of active
strike-slip faults. Journal of Structural Geology 22(7), 913-930.
Shimamoto, T., Takemura, K., Fujimoto, K., Tanaka, H. & Wibberley, C. A. J.
2001. Part II: Nojima Fault Zone probing by core analyses. Island Arc
10(3-4), 357-359.
Shipton, Z. K. & Cowie, P. A. 2001. Damage zone and slip-surface evolution
over mu m to km scales in high-porosity Navajo sandstone, Utah. Journal
of Structural Geology 23(12), 1825-1844.
Shipton, Z. K., Evans, J. P., Robeson, K. R., Forster, C. B. & Snelgrove, S. 2002.
Structural heterogeneity and permeability in faulted eolian sandstone:
Implications for subsurface modeling of faults. Aapg Bulletin 86(5), 863-
883.
Sibson, R. H. 1974. Frictional Constraints on Thrust, Wrench and Normal Faults.
Nature 249(5457), 542-544.
Sibson, R. H. 1977. Fault rocks and fault mechanisms. J. Geol. Soc. London 133,
22.
Sibson, R. H. 1986. Brecciation Processes in Fault Zones - Inferences from
Earthquake Rupturing. Pure and Applied Geophysics 124(1-2), 159-175.
Sibson, R. H. 1990. Conditions for fault-valve behaviour. Geol. Soc. Lond. Spec.
Pub 54(1), 15-28.

212
Sibson, R. H. 1994. Crustal stress, faulting, and fluid flow. in Geofluids: Origin,
Migration and Evolution of Fluids in Sedimentary Basins (edited by
Parnell, J.). Spec. Publs geol. Soc. Lond. 78, 69-84.
Sibson, R. H. 1996. Structural permeability of fluid-driven fault-fracture meshes.
Journal of Structural Geology 18(8), 1031-1042.
Sibson, R. H. 2003. Thickness of the seismic slip zone. Bulletin of the
Seismological Society of America 93(3), 1169-1178.
Sibson, R. H. & Toy, V. G. 2006. The Habitat od Fault-Generated
Pseudotachylyte: Presence vs. Absence of Friction-Melt. in Earthquakes:
Radiated Energy and the Physics of Faulting Geophysical Monograph
Series 170.
Sleep, N. H. 1995. Ductile Creep, Compaction, and Rate and State-Dependent
Friction within Major Fault Zones. Journal of Geophysical Research-
Solid Earth 100(B7), 13065-13080.
Stearns, D. W. 1968. Fracture as a mechanism of flow in naturally deformed
layered rocks. In: Proceedings af the Conference on Research in
Tectonics (edited by Baer, A. J. & Norris, D. K.). Geological Survey of
Canada, Ottawa, 79-89.
Stierman, D. J. 1984. Geophysical and Geological Evidence for Fracturing,
Water Circulation and Chemical Alteration in Granitic-Rocks Adjacent to
Major Strike-Slip Faults. Journal of Geophysical Research 89(Nb7),
5849-5857.
Swanson, M. T. 1992. Fault structure, wear mechanisms and rupture processes in
pseudotachylyte generation. Tectonophysics 204, 223-242.
Tchalenko, J. S. 1970. Similarities between Shear Zones of Different
Magnitudes. Geological Society of America Bulletin 81(6), 1625-&.
Tenthorey, E., Cox, S. F. & Todd, H. F. 2003. Evolution of strength recovery and
permeability during fluid-rock reaction in experimental fault zones. Earth
and Planetary Science Letters 206(1-2), 161-172.
Tenthorey, E., Scholz, C. H., Aharonov, E. & Leger, A. 1998. Precipitation
sealing and diagenesis - 1. Experimental results. Journal of Geophysical
Research-Solid Earth 103(B10), 23951-23967.
Townend, J. & Zoback, M. D. 2000. How faulting keeps the crust strong.
Geology 28(5), 399-402.
Uribe, F. & Niemeyer, H. 1984. ranjas miloníticas en la Cordillera de la Costa de
Antofagasta (Cuadrángulo de Cerro Cristales, 24º00'-24º15'S) y la
distribución del basamento precámbrico. Revista Geológica de Chile 23,
4.
Vermilye, J. M. & Scholz, C. H. 1998. The process zone: A microstructural view
of fault growth. Journal Of Geophysical Research-Solid Earth 103(B6),
12223-12237.
Vidale, J. E. & Li, Y. G. 2003. Damage to the shallow Landers fault from the
nearby Hector Mine earthquake. Nature 421(6922), 524-526.
Walsh, J. J. & Watterson, J. 1988. Analysis of the Relationship between
Displacements and Dimensions of Faults. Journal of Structural Geology
10(3), 239-247.
Wang, C. Y., Feng, R., Yao, Z. S. & Shi, X. J. 1986. Gravity-Anomaly and
Density Structure of the San-Andreas Fault Zone. Pure and Applied
Geophysics 124(1-2), 127-140.

213
Weaver, C. W. & Paterson, M. S. 1969a. Deformation of Cube-Oriented Mgo
Crystals under Pressure. Journal of the American Ceramic Society 52(6),
293-&.
Weaver, C. W. & Paterson, M. S. 1969b. Stress-Strain Properties of Rubber at
Pressures above Glass Transition Pressure. Journal of Polymer Science
Part a-2-Polymer Physics 7(3pa2), 587-&.
Wibberley, C. A. J. & Shimamoto, T. 2003. Internal structure and permeability
of major strike-slip fault zones: the Median Tectonic Line in Mie
Prefecture, Southwest Japan. Journal of Structural Geology 25(1), 59-78.
Wilson, B., Dewers, T., Reches, Z. & Brune, J. 2005. Particle size and energetics
of gouge from earthquake rupture zones. Nature 434(7034), 749-752.
Wilson, J. E., Chester, J. S. & Chester, F. M. 2003. Microfracture analysis of
fault growth and wear processes, Punchbowl Fault, San Andreas System,
California. Journal of Structural Geology 25(11), 1855-1873.
Wong, T. F., David, C. & Zhu, W. L. 1997. The transition from brittle faulting to
cataclastic flow in porous sandstones: Mechanical deformation. Journal
of Geophysical Research-Solid Earth 102(B2), 3009-3025.
Yasuhara, H., Marone, C. & Elsworth, D. 2005. Fault zone restrengthening and
frictional healing: The role of pressure solution. Journal of Geophysical
Research-Solid Earth 110(B6), -.
Zang, A., Wagner, F. C., Stanchits, S., Janssen, C. & Dresen, G. 2000. Fracture
process zone in granite. Journal of Geophysical Research-Solid Earth
105(B10), 23651-23661.
Zhang, S. Q. & Tullis, T. E. 1998. The effect of fault slip on permeability and
permeability anisotropy in quartz gouge. Tectonophysics 295(1-2), 41-52.
Zhu, W. L., Montesi, L. G. J. & Wong, T. F. 1997. Shear-enhanced compaction
and permeability reduction: Triaxial extension tests on porous sandstone.
Mechanics of Materials 25(3), 199-214.
Zhu, W. L. & Wong, T. F. 1997. The transition from brittle faulting to cataclastic
flow: Permeability evolution. Journal of Geophysical Research-Solid
Earth 102(B2), 3027-3041.
Zoback, M. D. & Byerlee, J. D. 1975. Effect of Microcrack Dilatancy on
Permeability of Westerly Granite. Journal of Geophysical Research
80(5), 752-755.
Zoback, M. D., Zoback, M. L., Mount, V. S., Suppe, J., Eaton, J. P., Healy, J. H.,
Oppenheimer, D., Reasenberg, P., Jones, L., Raleigh, C. B., Wong, I. G.,
Scotti, O. & Wentworth, C. 1987. New Evidence on the State of Stress of
the San-Andreas Fault System. Science 238(4830), 1105-1111.

214