ARTICLE
Fault rock heterogeneity can produce fault
weakness and reduce fault stability
John D. Bedford
1
✉
, Daniel R. Faulkner
1
& Nadia Lapusta
2,3
Geological heterogeneity is abundant in crustal fault zones; however, its role in controlling the
mechanical behaviour of faults is poorly constrained. Here, we present laboratory friction
experiments on laterally heterogeneous faults, with patches of strong, rate-weakening quartz
gouge and weak, rate-strengthening clay gouge. The experiments show that the hetero-
geneity leads to a signi
fi
cant reduction in strength and frictional stability in comparison to
compositionally identical faults with homogeneously mixed gouges. We identify a combi-
nation of weakening effects, including smearing of the weak clay; differential compaction of
the two gouges redistributing normal stress; and shear localization producing stress con-
centrations in the strong quartz patches. The results demonstrate that geological hetero-
geneity and its evolution can have pronounced effects on fault strength and stability and, by
extension, on the occurrence of slow-slip transients versus earthquake ruptures and the
characteristics of the resulting events, and should be further studied in lab experiments and
earthquake source modelling.
https://doi.org/10.1038/s41467-022-27998-2
OPEN
1
Rock Deformation Laboratory, Department of Earth, Ocean and Ecological Sciences, University of Liverpool, Liverpool, UK.
2
Department of Mechanical and
Civil Engineering, Division of Engineering and Applied Science, California Institute of Technology, Pasadena, CA, USA.
3
Seismological Laboratory, Division of
Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA, USA.
✉
email:
jbedford@liverpool.ac.uk
NATURE COMMUNICATIONS
| (2022) 13:326 | https://doi.org/10.1038/s41467-022-27998-2 | www.nature.com/naturecommunications
1
1234567890():,;
M
any large crustal faults have been shown to be fric-
tionally weak
1
–
6
when compared to laboratory mea-
surements of quasi-static fault friction. The coef
fi
cient
of friction
μ
¼
τ
=
σ
n
, where
τ
is the shear stress during slip and
σ
n
is the effective normal stress, of most geological materials is
typically measured in the laboratory to be between 0.6 and 0.85 at
slow slip speeds, independent of rock type
7
, with the exception of
a few weak minerals, predominantly phyllosilicates
7
,
8
. Possible
explanations for weak faults in nature, where the apparent
μ
at
which faults operate is often <0.5, include localization of weak
minerals along structural foliations
9
–
13
, dynamic weakening
during seismic slip
14
, and elevated pore
fl
uid pressure interpreted
as lower friction coef
fi
cients
15
,
16
. As well as being apparently
weak, many crustal faults also exhibit a spectrum of slip beha-
viour, with earthquake slip and aseismic creep often occurring on
the same fault
17
,
18
and slow slip phenomena being prevalent at all
crustal depths
19
. While the apparent weakness of faults and
spectrum of slip behaviour can be attributed to the effects of
spatially varying and temporally evolving con
fi
nement, tem-
perature, and pore
fl
uid pressure, it is clear that heterogeneity in
fault zone rocks (Fig.
1
a) can also play an important
20
,
21
, if not
dominant, role.
Geological investigations have shown that heterogeneity in
fault zone rocks occurs over many different scales, from
submillimetre-scale structural foliations
9
,
10
, centimetre- to meter-
scale blocks within a shear zone mélange
22
, hundreds-of-meters
scale where lenses of damaged protolith can be entrapped within
the core of wide (km-scale) fault zones
23
,
24
(e.g. Fig.
1
a), to tens-
of-kilometers scale variations in rock types
10
,
25
. The role of large-
scale fault rock heterogeneity has been highlighted in a number of
studies; for example, it has been suggested that heterogeneities
such as seamounts can act as earthquake nucleation sites and
control the seismogenic behaviour of subduction zone megathrust
faults
26
,
27
. However, the importance of small-scale fault rock
heterogeneity in controlling fault slip behaviour, average fault
strength, and fault stability is still uncertain.
Here, the effect of fault rock heterogeneity on fault strength and
slip behaviour is investigated by a series of laboratory friction
experiments on simulated laterally heterogeneous faults. The faults
consist of different sized patches of strong, rate-weakening quartz,
and weak, rate-strengthening clay fault gouges. Until now, the
majority of previous experimental investigations have been per-
formed using mixtures of different fault gouge materials with
varying frictional properties, where the materials are homo-
geneously mixed together
28
–
31
; intact wafers of natural gouge have
also been used
9
,
10
. In this work, experiments are performed on both
homogeneously mixed and spatially heterogeneous gouge layers
consisting of quartz, frictionally strong and rate-weakening, and
kaolinite clay powder, frictionally weak and rate-strengthening. The
fault gouge layers (50 mm long, 20 mm wide, with a thickness of
~1 mm at the onset of shear, after initial pressurization) are sheared
in a direct-shear arrangement (Fig.
1
b, see also Supplementary
Fig. 1) within a triaxial deformation apparatus (see
“
Methods
”
). The
heterogeneous gouge layers are constructed by placing different
sized patches of
fi
ne-grained quartz and clay powder (both <5 μm
grain size) adjacent to each other in a symmetrical pattern, with a
central quartz patch being bound by two clay patches (Fig.
1
b). This
symmetrical arrangement ensures that no misalignment between
the direct-shear forcing blocks would occur as a result of any dif-
ferential compaction between the different materials; furthermore,
the amount of gouge material used (measured by weight prior to
the experiment) was calculated so that the thickness of the quartz
and clay gouges were the same after initial pressurization and a
small amount of shear (Supplementary Fig. 2). The normal stress is
applied by the con
fi
ning pressure (
P
c
) in the triaxial apparatus, held
constant at 60 MPa for all tests in this study, and the pore
fl
uid
pressure (
P
f
) within the gouge is servo-controlled at a constant value
of 20 MPa, resulting in the effective normal stress
σ
n
=
40 MPa
(
σ
n
¼
P
c
P
f
). The gouge layers are sheared up to a maximum
displacement of 8.5 mm (shear strain
≈
10, given the
fi
nal layer
thickness of ~0.85 mm). Monitoring the evolution of shear stress
while applying velocity steps from 0.3 to 3 μm·s
−
1
and back allows
the experiments to quantify the rate-and-state friction parameters
that determine the stability of fault slip
32
. These sliding velocities
are suf
fi
ciently slow, given the gouge permeability, to ensure that
pore pressure transients do not build up within the gouge layer
during shearing
33
. The sizes of the strong yet unstable quartz and
weak but stable clay patches are varied to investigate the role of
different scales of heterogeneity on the magnitude and stability of
fault friction.
Results
Fault strength evolution
. The experimental results indicate
pronounced differences between the behaviour of laterally het-
erogeneous faults compared to the laterally homogeneous faults
with mixed gouge (Fig.
1
). All experiments are characterised by
an initially rapid increase in shear stress during the loading phase,
before the samples clearly yield
—
i.e., shear inelastically
—
after
~1 mm of displacement. After that, the friction coef
fi
cient
μ
of the
homogeneously mixed gouge layers remains relatively constant
(Fig.
1
d), with rate-and-state effects consistent with results from
previous experimental studies
29
–
31
. In contrast, the hetero-
geneous gouge layers all show ubiquitous weakening (Fig.
1
c),
with
μ
evolving towards the value of the weaker clay phase. To
ensure that the observed weakening was not caused by the
arrangement of the different gouge patches in the experiments,
tests were performed where the symmetry of the heterogeneous
layers was reversed (i.e. a central clay patch bound by two quartz
patches). These tests also exhibit similar weakening (Supple-
mentary Fig. 3) suggesting that it is the heterogeneity itself, not
the arrangement of the different materials, that causes the
weakening. Stable sliding is observed for all homogeneously
mixed faults and the majority of heterogeneous faults. However,
when the quartz patch in the heterogeneous layers comprises
≥
80% of the total sliding area, unstable stick-slip sliding emerges,
typically triggered by up-steps in the sliding velocity (Fig.
1
c).
The observed weakening of the heterogeneous faults is greater
than can be explained by the observed smearing of the clay
patches. Microstructural analysis of a heterogeneous layer
recovered at the end of an experiment (Fig.
2
a) shows smearing
of clay into localized boundary Y-shears that propagate into the
quartz patch. With progressive smearing and localization of the
clay phase (Fig.
2
b), the strength of the layer overall is expected to
decrease as a greater proportion of the slipping surface can be
located within the weak clay phase
34
. As the frictional strength of
the endmember gouge compositions is known (i.e. 100% quartz
and 100% clay in Fig.
1
c), the predicted weakening due to
smearing can be calculated (Fig.
2
c) by assuming that the overall
strength is determined by the strength of the two gouges acting in
series, based on their relative proportions (the arithmetic mean of
μ
, based on the proportions of clay and quartz within the layer).
The predicted weakening, associated with the relative increase in
length of the clay patches, is considerably less than the observed
weakening in the experiments (Fig.
2
c), suggesting that clay
smearing alone is not responsible for the progressive weakening
of heterogeneous faults.
An additional cause of the weakening could be differential
compaction between the different gouge materials resulting in a
redistribution of normal stress (see Supplementary Note 1 for full
discussion of this effect). The volumetric strain data from the
endmember quartz and clay gouge experiments show that the
ARTICLE
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-27998-2
2
NATURE COMMUNICATIONS
| (2022) 13:326 | https://doi.org/10.1038/s41467-022-27998-2 | www.nature.com/naturecommunications
quartz gouge experiences a greater layer thickness reduction of
about 20
μ
m than the clay gouge during slip (Supplementary
Fig. 2). In the heterogeneous layer experiments this would result
in an increase of normal stress on the weaker clay patches leading
to a progressive reduction in shear resistance, as observed in our
experiments. The magnitude of this effect is dependent on the
bulk (
K
) and shear (
G
) moduli
35
, which are poorly constrained
for the gouge materials in this study. Using plausible values for
the moduli (Supplementary Note 1) indicates that this differential
compaction effect could potentially explain a large component of
the weakening we observe in our experiments (Fig.
2
d).
Frictional stability
. The velocity steps from Fig.
1
c, d are used to
calculate the evolution in the rate-and-state friction
36
–
38
para-
meter (
a
—
b)
, which determines the frictional stability of the
fault
39
–
41
. When
(a
–
b)
> 0, the sliding behaviour is rate-
strengthening, suppressing instabilities and promoting stable
sliding, whereas when
(a
–
b)
< 0, the sliding behaviour is rate-
weakening which promotes unstable slip behaviour and the
occurrence of stick-slips in the laboratory. The values of (
a
–
b
) are
consistently lower (i.e. less rate-strengthening) in the hetero-
geneous faults throughout the experiments (Fig.
3
a
–
c). This
fi
nding indicates that the heterogeneous faults are closer to the
potentially unstable, rate-weakening regime than their homo-
geneous counterparts.
For the homogeneous faults with the pure quartz gouge and the
heterogeneous faults where the quartz patch comprises
≥
80% of
the total sliding area, only the
fi
rst velocity step can be used to
determine the rate dependence due to the occurrence of stick-slip
instabilities triggered by subsequent velocity steps. However, this
initial velocity step at 1.5 mm displacement does show negative
values of (
a
–
b
) associated with rate-weakening behaviour (Fig.
3
a),
which is consistent with the occurrence of stick-slip instabilities
later in the experiment. All of the calculated rate-and state
friction data are presented in Supplementary Table 1.
Discussion
Our experiments show that laterally heterogeneous fault gouge
layers weaken signi
fi
cantly in comparison to homogeneous layers,
pointing to heterogeneity-induced weakening effects. We
Fracture
damage
zone
Lenses of different materials
produce along fault
heterogeneity
Fault
core
Country
rock
Clay gouge
patches
Quartz
gouge patch
Gouges
homogeneously
mixed together
Soft silicone
spacer
Heterogeneous fault
experiments
Homogeneous fault
experiments
σ
n
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Friction coefficient,
μ
0123456789
Displacement (mm)
01 2 3 4 56 78 9
Displacement (mm)
Progressive weakening
Stick-slip instabilities
0.3
3
0.3
3
0.3
3
0.3
3
Clay fraction:
0%
20%
30%
40%
50%
60%
70%
80%
100%
Heterogeneous faults
Homogeneous faults
Increasing
clay patch
size
Increasing
clay content
Qtz
Qtz
Clay
Clay
-1
Sliding velocity (μm·s ):
0.3
3
0.3
3
0.3
3
0.3
3
-1
Sliding velocity (μm·s ):
Constant frictional strength
c
d
a
b
Fig. 1 Mechanical behaviour of laterally heterogeneous vs. homogeneously mixed clay-quartz fault gouge layers. a
Schematic diagram of a typical
natural fault zone showing how lenses of different materials trapped within the fault core produces a heterogeneous structure.
b
Simpli
fi
ed diagrams
of the experimental setup for the heterogeneous fault experiments, where quartz and clay gouges are separated into adjacent patches, and homogeneou
s
fault experiments where the two gouges are homogeneously mixed together.
c
Evolution of the friction coef
fi
cient (
μ
) with displacement for the
heterogeneous experimental faults and, (
d
) the homogeneous experimental faults. Heterogeneous faults show ubiquitous post-yield weakening with
increasing displacement, in contrast to homogeneous faults where
μ
remains relatively constant once the layer has yielded after approximately 1 mm of slip.
For pure quartz and heterogeneous faults where the quartz patch comprises
≥
80% of the total fault area, stick-slip instabilities occur, triggered by up-steps
in the sliding velocity.
NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-022-27998-2
ARTICLE
NATURE COMMUNICATIONS
| (2022) 13:326 | https://doi.org/10.1038/s41467-022-27998-2 | www.nature.com/naturecommunications
3