Porosity reduction in rocks from a fault core can cause elevated pore fluid pressures and consequently influence the recurrence time of earthquakes. We investigated the porosity distribution in the New Zealand's Alpine Fault core in samples recovered during the first phase of the Deep Fault Drilling Project (DFDP-1B) by using two-dimensional nanoscale and three-dimensional microscale imaging. Synchrotron X-ray microtomography-derived analyses of open pore spaces show total microscale porosities in the range of 0.1 %–0.24 %. These pores have mainly non-spherical, elongated, flat shapes and show subtle bipolar orientation. Scanning and transmission electron microscopy reveal the samples' microstructural organization, where nanoscale pores ornament grain boundaries of the gouge material, especially clay minerals. Our data imply that (i) the porosity of the fault core is very small and not connected; (ii) the distribution of clay minerals controls the shape and orientation of the associated pores; (iii) porosity was reduced due to pressure solution processes; and (iv) mineral precipitation in fluid-filled pores can affect the mechanical behavior of the Alpine Fault by decreasing the already critically low total porosity of the fault core, causing elevated pore fluid pressures and/or introducing weak mineral phases, and thus lowering the overall fault frictional strength. We conclude that the current state of very low porosity in the Alpine Fault core is likely to play a key role in the initiation of the next fault rupture.
Fault mechanics, fault structure, and fluid flow properties of damaged fault
rocks are intimately related (e.g., Gratier and Gueydan, 2007; Faulkner
et al., 2010). Fault rupture is associated with intense brittle fracturing
that enhances porosity, and thus permeability, and therefore also possible
rates and directions of fluid propagation within fault zones (e.g., Girault
et al., 2018). Conversely, post-seismic recovery mechanisms (gouge compaction
and pressure solution processes) result in reductions in porosity,
permeability, and fluid flow (Renard et al. 2000; Faulkner et al., 2010;
Sutherland et al., 2012). These processes may cause elevated pore fluid
pressures within fault cores and trigger frictional failure (e.g., Sibson,
1990; Gratier et al., 2003; Zhu et al., 2020). Therefore, the state of
porosity within rocks from fault cores can play a key role in fault slip.
The Alpine Fault of New Zealand is late in its seismic cycle (Cochran et al.,
2017), so studying it allows us to investigate pre-earthquake conditions that
may influence earthquake nucleation and rupture processes. Recently, drilling
operations were undertaken in this fault zone to investigate the in situ
conditions (Sutherland et al. 2012, 2017). Slug tests in the DFDP-1B borehole
(Sutherland et al., 2012) and laboratory permeability measurements of core
samples (Carpenter et al., 2014) indicate permeability decreases by 6 orders
of magnitude with increasing proximity to the fault. Furthermore, Sutherland
et al. (2012) documented a 0.53
In this study, we investigate the porosity distribution in rocks from the Alpine Fault core and consider the potential effects of this porosity on fault strength. We have measured open pore spaces in these rocks from X-ray computed tomography (XCT) datasets and examined pore morphology by implementing quantitative shape analyses. Lithological and microstructural characteristics of these samples were performed by using scanning electron microscopy (SEM) and transmission electron microscopy (TEM).
New Zealand's Alpine Fault (Fig. 1a) is a major active
crustal-scale structure that ruptures in a large earthquake every
Most of the brittle shear displacement along the fault has been accommodated
within the fault core, which includes PSZ gouges and
cataclasite-series rocks (Toy et al., 2015). Both in surface outcrops and
drill core samples, the Alpine Fault manifests itself as a thin (5 to 20
Porosity analyses were performed on four samples representing PSZ gouges and
cataclasites of the Alpine Fault core, which were recovered from the DFDP-1B
borehole (Fig. 1b, c; Sutherland et al., 2012). These are DFDP-1B 58_1.9,
DFDP-1B 69_2.48, DFDP-1B 69_2.54, and DFDP-1B 69_2.57. Sample
nomenclature includes drill core run number, section number, and centimeters
measured from the top of each section. These samples were recovered from
drilled depth of 126.94, 143.82, 143.88, and 143.91
Detailed lithological and microstructural descriptions of the DFDP-1B drill core were carried out simultaneously with and after the drilling operations by the DFDP-1 Science Team, and these data were later summarized by Toy et al. (2015). Samples DFDP-1B 58_1.9 and DFDP-1B 69_2.48 belong to foliated cataclasite units (Fig. 1b, c; Toy et al., 2015), described as ultracataclasites with gouge-filled shear zones located above PSZ-1 and PSZ-2, respectively. Sample DFDP-1B 69_2.54 represents the gouge layer that defines PSZ-2, whereas sample DFDP-1B 69_2.57 is composed of brown ultracataclasites that belong to the lower-cataclasite unit (Fig. 1b, c; Toy et al., 2015).
We imaged the samples using X-ray absorption tomography, where the signal
intensity depends on how electron density and bulk density attenuate a
monochromatic X-ray along its path through the material (e.g., Fusseis et al.,
2014). We acquired the X-ray microtomography data for this study at the 2-BM beamline of the Advanced Photon Source, Argonne National Laboratories, USA, in
December 2012. The non-cylindrical samples of
Data analyses and image processing were performed using the commercial software Avizo 9.1™ (Fig. 2). Initially, the datasets were rescaled to 8 bit grayscale volumes for enhanced computer performance. In addition, small volumes of interest were cropped from the whole volume before a non-local means filter was applied to reduce noise (Buades et al., 2005). For each voxel, this filter compares the value of this voxel with all neighboring voxels in a given search window. A similarity between the neighbors determines a correction applied to each voxel (e.g., Thomson et al., 2018).
X-ray tomography data processing workflow.
On the filtered grayscale images, pores were identified as disconnected
materials of the darkest grayscale range (Figs. 2a and S1 in the
Supplement). The corresponding grayscale values were thresholded, and the
datasets were converted into binary form. This step is called segmentation.
Several segmentation techniques exist, from thresholding at a given grayscale
value (e.g., Ianossov et al., 2009; Andrew et al., 2013) to deep-learning
algorithms (Ma et al., 2020). It is up to the user to choose the segmentation
technique that is most appropriate to analyze a given dataset. To our
knowledge, no single segmentation technique can be generalized and universally
used independently of the nature of the samples. In the present study, we have
chosen a simple segmentation technique by applying a threshold to the grayscale images to separate the void space from the solid. This technique has
been used in many studies in the last 2 decades to characterize porosity in
rocks, including some very recent studies in rock physics (Macente et al.,
2019; Renard et al., 2019). The segmented porosity volume depends strongly on
the choice of the threshold and some studies have demonstrated that the final
porosity estimated by different segmentation methods can vary by 20 %
(Andrä et al., 2013). However, when the level of noise in the data is low,
the differences in porosities estimated by different segmentation techniques
are negligible (Andrew, 2018). Our data were acquired at a synchrotron where
the parallel beam and high photon flux ensured a low level of the noise in the
images. In addition, application of a non-local-means filter applied to our
data reduced the noise level. For these reasons, we consider that it was
robust to apply a simple thresholding technique to this dataset but
acknowledge that the porosity values we estimate could differ by
However, our segmentation procedure also captured cracks within a sample,
which are likely to result from depressurization during core recovery
(Figs. 2b and S1 in the Supplement). To omit the cracks, we utilized the
morphological operation “connected components” available in the software
Avizo 9.1, which allows volumes larger than the selected number of connected
voxels to be excluded from the binary label images. To each sample we applied
upper limits of 20 (43.94
Plots of pore volume vs. number of pores for each sample. Estimates of total porosity and size of the maximum expected pore are also shown, as well as the curve fitting function for each dataset.
Instead, the volumes and shape characteristics of segmented materials
(including cracks, i.e., without any data limitation) were exported from the Avizo software in numerical format, and volume distributions within a sample were
plotted on a logarithmic scale (Fig. 3). Data up to a specific volume size
were fit to a polynomial curve, and then the curve was extrapolated to the
Bivariate histograms showing elongation vs. pore
volume (
Pore shapes were analyzed on bivariate histograms plotted by using the
numerical pore characteristics, previously extracted from the Avizo software.
Only pore volumes between 21.97
Bivariate histograms showing sphericity vs. pore
volume (
The angles
Bivariate histograms showing flatness vs. pore volume
(
Distribution of pore unit orientations plotted on a lower-hemisphere equal-area stereographic projection with a probability density contour.
SEM images were collected on Zeiss Sigma-FF-SEM at the University of Otago's
Centre for Electron Microscopy. The SEM was operated at a working distance of
8.5
TEM images were collected on a FEI Tecnai G2 F20 X-Twin transmission electron
microscope, located at the German Research Centre for Geosciences (GFZ),
Potsdam, Germany (Fig. 9). The instrument is equipped with a field-emission gun
(FEG) electron source and a high-angle annular dark-field (HAADF)
detector. Images were collected from samples placed on a Gatan double-tilt
holder at an accelerating voltage of 200
Scanning electron images collected from sample DFDP 1B
69_2.48 showing the existing mineral associations.
All samples contain low total porosities, ranging from 0.1 % to 0.24 %
(Fig. 3). If different segmentation techniques were applied, a variability in
the range that Andrew (2018) demonstrated would be reasonable (from nearly 0 %
to 20 %) and would correspond to porosities between 0.08 % and 0.29 %
in our samples. It can be noted that the lower-cataclasite sample (DFDP-1B
69_2.57) has twice as much pore space (Fig. 3d) as any of the other
samples. The characterized pore volume distributions range over almost 3 orders of magnitude for all samples (Fig. 3). Furthermore, the expected
maximum pore volume was estimated to be largest in the PSZ-2 sample (DFDP-1B
69_2.54), reaching 862
Transmission electron microscopy images collected from
the gouge sample DFDP-1B 69_2.54 (PSZ-2). Panels
In all samples, shape analyses of pores with volumes between
21.97
The orientations of the individual pore units show two distinctive peaks with opposite vergence, defining bipolar distributions of pore orientations (Fig. 7). The observed bipolarity is subtle in samples DFDP-1B 58_1.9 (Fig. 7a) and DFDP-1B 69_2.48 (Fig. 7b) and more obvious in samples DFDP-1B 69_2.54 (Fig. 7c) and DFDP-1B 69_2.57 (Fig. 7d).
To demonstrate the microstructural arrangement of the cataclasites, we show
representative SEM images from sample DFDP-1B 69_248 (Fig. 8), previously
described as a “lower foliated cataclasite” by Toy et al., 2015. SEM images
presented here reveal rounded to subrounded crystalline clasts up to
100
TEM characterization of the gouge material from PSZ-2 (sample DFDP-1B
69_2.54) reveals that the Alpine Fault gouges are composed of angular
quartz and/or feldspar fragments (
The gouge material also demonstrates phyllosilicate-rich areas, defined by an
increase in the clay
Porosity analyses of samples from or in close proximity to the two PSZs
encountered in the DFDP-1B drill core reveal total pore volumes between
0.1 % and 0.24 % (Fig. 3). These values are significantly lower than
the porosity estimates from other active faults in the world, such as 0.2 to
5.7 % total porosity in the core of the Nojima Fault, Japan (Surma et al.,
2003) and 0 % to 18 % in the San Andreas Fault core (Blackburn et al.,
2009). The Alpine Fault core contains total pore space volumes comparable only
with the lower porosities in these previous studies. It should be noted that
the smallest pore spaces captured in the XCT datasets are 1.3
Foliation in the upper cataclasites is defined by clay-sized phyllosilicates that become more abundant with proximity to the PSZ (Toy et al., 2015), where a weak clay fabric is developed (Schleicher et al., 2015). This gradual enrichment in clay minerals coincides with the subtle development of bipolar distributions of pore orientations with increasing sample depth (Fig. 7). This observation and the fact that pores are mainly distributed along grain boundaries of clays (Fig. 9) suggest that the distribution of clay minerals also controls pore orientations within the Alpine Fault core. Previously, the phyllosilicate foliation in the Alpine Fault cataclasites has been used to define shear direction (Toy et al., 2015). Thus, we speculate that pore orientations in these rocks are also systematically related to the kinematic framework of the shear zone. If these pores represent remnants of fluid channels, their spatial orientation is likely to reflect the fluid flow directions during deformation. To address this possibility more data for systematic analyses of pore orientations are needed.
The comparatively lower porosity estimates of the Alpine Fault core compared to other active faults (e.g., the Nojima Fault, Surma et al., 2003, and the San Andreas Fault, Blackburn et al., 2009) could be attributed to the fact that the Alpine Fault is late in its c. 300-year seismic cycle and the last seismic event occurred in 1717 (Cochran et al., 2017). Thus, we propose that the fault has almost completely sealed. Porosity of fault cores is believed to evolve during the seismic cycle, since fault rupture can cause porosities to increase up to 10 % (Marone et al., 1990), and subsequent healing mechanisms (such as mechanical compaction of the fault gouge and/or elimination of pore spaces within the fault core due to pressure solution processes) cause porosity to decrease over time (Sibson, 1990; Renard et al., 2000; Faulkner et al., 2010). SEM data presented here show that fine-grained chlorite and muscovite grains formed as a cement in the cataclastic matrix (Fig. 8c). Our TEM data reveal the abundance of newly precipitated authigenic clays, wrapped around coarser clay minerals (Fig. 9b). Furthermore, delicate clay minerals form fringe structures (Fig. 9a) and strain shadows (Fig. 9c) around larger quartz–feldspar grains. These microstructural observations demonstrate that pressure solution processes operated within these rocks (Toy et al., 2015).
Evidence for pressure solution processes has been previously documented in all units comprising the Alpine Fault core (Toy et al., 2015). Abundant precipitation of alteration minerals (Sutherland et al., 2012), calcite-filled intragranular and cross-cutting veins (Williams et al., 2017), and the occurrence of newly formed smectite clays (Schleicher et al., 2015) indicate extensive fluid–rock reactions. In addition, anastomosing networks of opaque minerals (such as graphite; Kirilova et al., 2017), which define foliation in the upper cataclasites (Toy et al., 2015), have been interpreted as being concentrated by pressure solution processes during aseismic creep (Toy et al., 2015; Gratier et al., 2011). The petrological characteristics of the Alpine Fault core lithologies indicate that solution transfer was likely the dominant mechanism for pore closure within these rocks.
Porosity estimates presented here are so low that presumably negligible variations in between samples can represent significant gradients in porosity. For example, the increase in total porosity in sample DFDP-1B 69–2.57 with only 0.14 % manifests itself as twice as many open pore spaces in comparison to the rest of the analyzed samples (Fig. 3). In addition, this is the only footwall sample analyzed here and, as already mentioned in Sect. 3.1, does not contain any gouge material. Post-rupture porosity reduction is known to operate 3 to 4 times faster within fine-grained fault gouges than in coarser-grained cataclasites (Walder and Nur, 1984; Sleep and Blanpied, 1992; Renard et al., 2000), which may explain the porosity differences demonstrated above. Furthermore, previous studies documented less carbonate and phyllosilicate filling of cracks in the Alpine Fault footwall cataclasites than in the hanging wall cataclasites (Sutherland et al., 2012; Toy et al., 2015), suggesting more reactive fluids are present and isolated within the hanging wall of the Alpine Fault. Thus, more intense dissolution–precipitation processes took place in the fault's hanging wall, which very likely resulted in more efficient porosity reduction, as demonstrated by our porosity estimates (Fig. 3).
Very low-porosity estimates are presented here (Fig. 3). Very low
permeabilities of 10
Previous studies and the observations presented here show that fluids were present in the Alpine Fault rocks. Fluid-filled pores represent a favorable environment for mineral precipitation, which can affect the fault strength in two ways: (i) a very small decrease in these critically low total porosities due to mineral precipitation would cause fluid pressurization, which is a well-known fault-weakening mechanism described by Byerlee (1990) and Sibson (1990); however, this pressure increase could be slightly offset by the inclusion of fluids into new hydrous minerals; (ii) deposition of frictionally weak phases (such as clay minerals and graphite), especially if they decorate grain contacts and/or form interlinked weak layers, would lower the overall frictional strength (Rutter et al., 1976; Niemeijer et al., 2010).
Precipitated authigenic clay minerals were identified in our TEM data (Fig. 9) and also documented by previous studies (Schleicher et al., 2015). As well as having low frictional strengths (Moore and Lockner, 2004), clay minerals may also contribute to the formation of an impermeable seal if they form an aligned fabric, which can enhance the likelihood of fluid pressurization in the fault rocks (Rice, 1992; Faulkner et al., 2010). In addition, graphite, which was previously documented in these rocks (Kirilova et. al., 2017), may effectively weaken the fault due to mechanical smearing (Rutter et al., 2013) and/or localized precipitation within strained areas (Upton and Craw, 2008). Such graphite precipitation within shear surfaces was previously documented by Kirilova et al. (2017).
In summary, the presence of trapped fluids in the low-porosity rocks of the Alpine Fault core possibly controls the mechanical behavior of the fault and could be responsible for future rupture initiation due to fluid pressurization and/or precipitation of weak mineral phases. This hypothesis is further supported by an experimental study showing that the DFDP-1 gouges are frictionally strong in the absence of elevated fluid pressure (Boulton et al., 2014).
Analyses of XCT datasets and TEM images of borehole samples from the core of the Alpine Fault reveal micro- and nanoscale pores, distributed along grain boundaries of the constituent mineral phases, especially clay minerals. The tendency of these pores to ornament clays defines their predominantly non-spherical, elongated, flat shapes and the bipolar distribution of pore orientations. The documented extremely low total porosities (in the range 0.1 %–0.24 %) in these rocks suggest effective porosity reduction and fault healing. Microstructural observations presented here and documented in previous studies indicate that pressure solution processes were the dominant healing mechanism and that fluids were present in these rocks. Therefore, fluid-filled pores may be places where elevated pore fluid pressures develop, due to further mineral precipitation that decreases the already critically low total porosities. Alternatively, these pores may also facilitate the deposition of weak mineral phases (such as clay minerals and graphite) that may very effectively weaken the fault. We conclude that the current state of the fault core porosity is possibly a controlling factor in the mechanical behavior of the Alpine Fault and will likely play a key role in the initiation of the next fault rupture.
Avizo screenshots, total porosity estimates, Matlab script, and numerical data of pore volumes can be found in the Supplement.
The supplement related to this article is available online at:
MK reconstructed, processed, and analyzed the XCT datasets presented here, interpreted the TEM data, and prepared the paper. Most of this work was performed during MK's PhD under the academic guidance of VT. VT and KG collected the XCT data with technical support by XX. FR and KS contributed with valuable discussion about XCT data analyses and edited the paper. RW enabled TEM data acquisition and provided his expertise on TEM data interpretation. RM collected and analyzed the presented SEM data. The final version of this paper benefits from collective intellectual input.
The authors declare that they have no conflict of interest.
This research used resources of the Advanced Photon Source, a US Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Argonne National Laboratory under Contract No. DE-AC02-06CH11357. We thank Sherry Mayo for helping with the reconstruction process of XCT data and Andrew Squelch for providing use of the Avizo workstation, located at CSIRO, Perth, Australia during the initial data analyses. Special thanks to Reed Debaets for assistance with the development of Matlab code. Klaus Gessner publishes with permission of the Executive Director, Geological Survey of Western Australia.
This research has been supported by the Advanced Photon Source, Chicago (grant no. GUP 31177), Nvidia Corporation, Royal Society of New Zealand’s Rutherford Fellowship (grant no. 16-UOO-001), the Ministry of Business and Innovation's Endeavor Fund, New Zealand (grant no. C05X1605/GNS-MBIE00056), the Tectonics and Structure of Zealandia Program at GNS Science (grant no. GNS-DCF00020), and publishing bursary funding provided by the University of Otago. This open-access publication was funded by Johannes Gutenberg University Mainz.
This paper was edited by Florian Fusseis and reviewed by James Gilgannon and Michel Bestmann.