Frictional properties and microstructural evolution of dry and wet calcite-dolomite gouges

Calcite and dolomite are the two most common minerals in carbonate-bearing faults and shear zones. Motivated by field examples from exhumed seismogenic faults in the Italian Central Apennines, we investigated the frictional and microstructural evolution of gouge mixtures consisting of 50 wt.% calcite and 50 wt.% dolomite using a rotary-shear apparatus. The gouges were sheared at a range of slip rates (30 μms-1 – 1 ms-1), displacements (0.05–0.4 m), and normal loads (17.5–26 MPa), under both room humidity and water-dampened conditions. The frictional behaviour and microstructural evolution of 5 the gouges were strongly influenced by the presence of water. At room humidity, slip strengthening behaviour was observed up to slip rates of 0.01 ms-1, which was associated with gouge dilation and the development of a 500-900 μm wide slip zone cut by Y-, R-, and R1-shear bands. Above a slip rate of 0.1 ms-1, dynamic weakening accompanied the development of a localised <100 μm thick principal slip zone preserving microstructural evidence for calcite recrystallization and dolomite decarbonation, while the bulk gouges developed a well-defined foliation consisting of organized domains of heavily fractured 10 calcite and dolomite. In water-dampened conditions, evidence of gouge fluidization within a fine-grained principal slip zone was observed at a wide range of slip-rates from 30 μms-1 to 0.1 ms-1, suggesting that caution is needed when relating fluidization textures to seismic slip in natural fault zones. Dynamic weakening in water-dampened conditions was observed at 1 ms-1, where the principal slip zone was characterised by patches of recrystallized calcite. However, local fragmentation and reworking of recrystallized calcite suggests a cyclic process involving formation and destruction of a heterogeneous slip zone. 15 Our microstructural data show that development of a well-defined gouge foliation at these experimental conditions is limited to high-velocity (>0.1 ms-1) and room humidity, supporting the notion that some foliated gouges and cataclasites may form during seismic slip in natural carbonate-bearing faults.


Friction evolution with slip and slip rate
The evolution of the effective friction coefficient (µ) with slip and slip rate was influenced by the availability of water during deformation (Figs. 2,3). In room humidity conditions and slip rates ≤0.01 ms -1 , the calcite-dolomite mixtures showed a progressive increase of µ (slip strengthening behaviour) up to 0.75-0.80 (measured between 0.15 m and 0.35 m of slip) following 130 an initial peak fiction (µ peak ) of 0.64-0.71 (Figs. 2,3). At a slip rate of 0.1 ms -1 , a substantial decrease of µ was observed (slip weakening behaviour to steady state µ ss of 0.55±0.01) following a prolonged initial strengthening phase (c. 0.062 m) that reached µ peak of 0.68 (Fig. 2b). Significant dynamic weakening was observed at a slip rate of 1 ms -1 at 17.5 MPa and 26 MPa normal stress (experiments s1221 and s1324, respectively), following a short initial strengthening phase (lasting c. 0.005-0.008 m) that was followed by a steady state of µ ss of 0. 3b). In these two experiments, a re-strengthening phase 135 (final µ up to c. 0.56-0.59) was observed during deceleration of the rotary column.
In water-dampened conditions, the gouge mixtures showed a similar evolution of friction at slip rates ≤0.1 ms -1 , characterized by slight slip strengthening to slip neutral behaviour (Fig. 2c). Notably, µ peak and µ ss were lower than in room humidity experiments, with µ peak = 0.61-0.64 and µ ss = 0.62-0.70 (Fig. 3). At a slip rate of 1 ms -1 , the initial strengthening phase was much shorter than in room-humidity conditions (c. 0.003 m), and dynamic weakening resulted in µ ss of 0.31±0.02. 140 Re-strengthening was also observed during deceleration, with an increase in µ up to 0.57.

Gouge thickness evolution with slip rate
No significant gouge loss was observed during the experiments, with the exception of those performed at V = 0.1 ms -1 discussed below. Therefore, the evolution of axial displacement is interpreted to result from changes in gouge layer thickness due to dilation and compaction. In room humidity conditions, the evolution of gouge layer thickness depends on slip rate (Fig. 4a). At 145 V ≤ 0.001 ms -1 , the gouge layers show a three-stage evolution: (i) initial compaction of c. 90-120 µm at the onset of sliding, (ii) dilation of c. 50-70 µm during the slip strengthening phase, and (iii) approximately constant thickness once the steady state friction coefficient is reached. Overall compaction of c. 30-60 µm is recorded. At V = 0.01 ms -1 , initial compaction of 100 µm is followed by approximately constant thickness (Fig. 4a). At higher slip rates (V ≥ 0.1 ms -1 ), continuous compaction was observed throughout the experiments (up to c. 300 µm of axial shortening at V = 1 ms -1 ), and compaction rate increased with 150 slip rate (Fig. 4a).
Under water-dampened conditions, the gouge mixtures exhibit a similar evolution of thickness irrespective of slip rate (Fig.   4b). Compaction was initially rapid in the first few cm of sliding, and then reached an approximately constant compaction rate that was similar in all experiments. Total compaction of c. 200-250 µm was recorded (Fig. 4b).
3.3 Temperature evolution and CO 2 emissions 155 Figure 5a shows maximum temperatures measured by the thermocouple located closest to the gouge layers ( Fig. 1b; Demurtas et al., 2019a, described temperature evolution with slip). The maximum temperature (621 • C) was achieved in experiment s1221 performed under room humidity conditions at V = 1 ms -1 (Fig. 5a). For the same slip rate and normal stress, but in water-dampened conditions, the maximum temperature was 210 • C (Fig. 5a). Temperature increases were detected in all experiments at slip rates ≥ 0.01 ms -1 , and the maximum temperature increased with increasing slip rate (Fig. 5a). 160 CO 2 emissions above ambient levels were only detected in experiments at slip rates ≥ 0.1 ms -1 (Fig. 5b). Because the mass spectrometer was not calibrated and the sample holder was open to the laboratory, the data can only be used in a qualitative way. In room humidity conditions, the intensity of the CO 2 peak was significantly higher at 1 ms -1 than at 0.1 ms -1 . In waterdampened conditions, the CO 2 peaks were substantially smaller than at equivalent room humidity conditions.

165
Compared to the starting materials, no mineralogical changes were detected in any of the deformed bulk gouges (see Supplementary Material). In room humidity experiment s1210 (30 µms -1 ), a slight broadening of the main peak for calcite was observed (Fig. 6a), and to a lesser degree also for dolomite. XRPD analysis of cohesive chips recovered from the slip surface of water-dampened experiment s1214 (V = 30 µms -1 ) indicates the presence of aragonite ( Fig. 6b). At V = 1 ms -1 and room humidity conditions, the recovered slip surface was composed of dolomite, Mg-calcite, and periclase (MgO) (Fig. 6c). Mg-calcite and 170 periclase are two of the main products of dolomite decarbonation that starts at c. 550 • C (MgCa(CO 3 ) 2 → MgO + (Ca, Mg)CO 3 + CO 2 , Samtani et al., 2002;De Paola et al., 2011a, b). In the present set of experiments, the microstructural domains and the principal slip zone varied in thickness at different slip rates (Fig. 7). In Figure 8, the thickness of the principal slip zone has been tracked at different deformation conditions (i.e. slip rate and presence of water).
Each individual shear band is associated with a very fine-grained matrix (grain size <1 µm) composed of calcite and dolomite.
The presence of multiple interlinked shear bands contributes to a weak foliation within the slip zone that lies sub-parallel to 185 gouge layer boundaries (Fig. 9a). Y-, R-, and most notably R 1 -shears, gradually decrease in abundance with increasing slip rate. The transition from fine-grained slip zone to highly fractured bulk gouge is typically well-defined (see upper part in Fig.   9d). The bulk gouge shows widespread cataclasis and intragranular fracturing, which is focussed preferentially into calcite grains (Smith et al., 2017;Demurtas et al., 2019a). Fractures that cut relatively large grains of calcite in the bulk gouge often exploit cleavage planes (e.g. Fig. 9d; Smith et al., 2017;Demurtas et al., 2019a).

190
At a slip rate of 0.1 ms -1 , the bulk gouge develops a weak foliation defined by compositional banding of heavily fractured calcite-and dolomite-rich domains, which lie adjacent to a localized principal slip zone c. 110 µm thick (Fig. 9e). The foliation is inclined 25-30 • to the principal slip surface and appears to form by disaggregation and shearing of originally intact calcite and dolomite grains (Fig. 9e). Locally, the principal slip surface is associated with discontinuous lens-shaped patches (up to 15-20 µm thick) of calcite with irregular boundaries and negligible porosity (Fig. 9f). 195 In experiments conducted at 1 ms -1 , the bulk gouges developed a well-defined foliation across most of the thickness of the layers (Smith et al., 2017;Demurtas et al., 2019a, b). The foliation is defined by alternating calcite-and dolomite-rich domains inclined at c. 40 • to the principal slip surface ( Fig. 10a-b), which become progressively rotated as they approach the slip surface (Fig. 10c). Large remnant grains (up to 200 µm) in the bulk gouge are often rimmed by fractured tails of finergrained aggregates (grain size <10 µm), and resemble mantled porphyroclasts in mylonites (e.g. Snoke et al., 1998;Trouw and 200 Passchier, 2009) (arrow in Fig. 10a). At distances of <400 µm from the principal slip surface, the mean grain size decreases substantially, there are very few large surviving grains (up to c. 100 µm in size), and there is a greater degree of mixing between calcite and dolomite (see more uniform colouring in the upper part of EDS map in Fig. 10b). The principal slip zone consists of a 15-20 µm thick, extremely fine-grained layer ( 1 µm in size) composed of calcite, Mg-calcite, dolomite, and periclase (EDS and XRPD analysis; Figs. 6 and 10c-d). Calcite forms elongate aggregates with negligible porosity that display 205 an aggregate preferred orientation with the long axes sub-parallel to foliation (Fig. 10c-e). Dolomite-rich domains show higher porosity and preserve distinct grain structures (Fig. 10c-d). EBSD analysis of elongate calcite aggregates within the principal slip zone (Fig. 10e) shows a distinct crystallographic preferred orientation with c-axes inclined sub-perpendicular to gouge layer boundaries ( Fig. 10f; see also Demurtas et al., 2019b). Adjacent to the principal slip zone, a c. 30-40 µm thick layer includes dolomite grains with diffuse internal cracking, clusters of small holes, and vesicular rims previously interpreted as 210 resulting from degassing during decarbonation of comminuted dolomite grains ( Fig. 10c; Mitchell et al., 2015;Demurtas et al., 2019b). At 1 ms -1 and 26 MPa, the foliation was found only within 400 µm of the principal slip surface (Supplementary In the bulk gouges, the region furthest from the slip zone is composed of grains that show very limited fracturing and resemble the starting materials (Fig. 11a,d; compare with Fig. 1e). Towards the slip zone, grains are increasingly fractured and become rounder. As in the room humidity experiments, most of the larger "surviving" grains are composed of dolomite (Fig. 11e), consistent with data showing that calcite undergoes more efficient grain size reduction compared to dolomite (Smith et al., 2017;Demurtas et al., 2019a). Domain boundaries (e.g. between intact bulk gouge and comminuted gouge) are often gradational ( Fig.   220 11d), and the total thickness of the comminuted zone is observed to decrease at higher slip rates (from c. 1500 µm thick at 30 µms -1 to c. 150 µm thick at 1 ms -1 ). The principal slip zone consists of an ultrafine-grained matrix (grain size <1 µm) composed of a mixture of calcite and dolomite, with a few well-rounded surviving dolomite grains up to 20-30 µm in size ( Fig.   11b-c). At the lowest slip rate (i.e. 30 µms -1 ), the principal slip zone has a sharp boundary with a characteristic wavelength with the underlying gouge (Fig. 11d), and contains irregular flame-like structures defined by subtle variations in the content of of the experiments (c. 0.5 s) (Sarnes and Schrüfer, 2007). Recent studies in which the temperature during experimental seismic slip was measured with optical fibres located inside the slip zone (in-situ measurements at acquisition rates of 1 kHz) detected 280 temperatures 300-400 • C higher than those measured with thermocouples (Aretusini et al., 2019). Temperatures in the slipping zone substantially higher than 621 • C would make grain size-and temperature-dependent deformation mechanisms more efficient. Instead, in the case of experiment s1218 performed at V = 0.1 ms -1 , the moderate dynamic weakening (µ ss = 0.55) can be related to more limited frictional heating within the principal slip zone both in time (max temperature measured of 190 • C, Fig. 5a) and space (patchy recrystallized areas in Fig. 9e-f). However, at least locally, the temperature increase was 285 sufficiently large to decompose dolomite, as testified by the clear CO 2 peak measured during shearing at this velocity (Fig. 5b).
In water-dampened conditions, the mechanical behaviour of the calcite-dolomite mixtures is similar (slight slip strengthening to slip neutral) at all slip rates up to 0.1 ms -1 (Fig. 2c). The thickness of the principal slip zone decreases log-linearly with increasing slip rate, indicating a progressively higher degree of localization (Fig. 8). However, this has no obvious effect on the steady state friction coefficient (Fig. 3b), possibly suggesting that the steady state is controlled by strain and that strain is 290 kept constant by microstructural reorganization distributed within the slip zone. The principal slip zone is composed of a very fine-grained ( 1 µm) matrix of calcite and dolomite that includes a few well-rounded dolomite clasts up to 20-30 µm in size (Fig. 11). The similarity in microstructure at all investigated slip rates suggests that water has a major role in promoting faster grain size reduction at the onset of slip, possibly by decreasing the surface energy and yield stress of calcite and dolomite (Risnes et al., 2005;Røyne et al., 2011). XRPD analysis of the slip surface of experiment s1214 (30 µms -1 ) showed the 295 formation of aragonite (Fig. 6b). Given that the starting materials were composed of calcite and dolomite only, the aragonite must have formed during deformation. Li et al. (2014), documented polymorphic transformation of calcite into aragonite due to mechanical grinding in a dry (i.e. room humidity) environment. Our observations therefore suggest that relatively dry patches could develop in the gouge layer during slip (or were present at the onset of slip), or that such transformation is also possible under water saturated conditions.

300
The slip zone of the water-dampened experiment performed at 30 µms -1 is characterized by flame-like structures, and domain boundaries that display a characteristic wavelength (Fig. 11b,d). Similar structures are typical of soft sediment deformation (Allen, 1985) and have also been described within fault cores (Brodsky et al., 2009), where they are interpreted to result from fluid mobilization and a difference in viscosity between two adjacent layers during deformation. Additionally, the occurrence of grain size grading within the water-dampened principal slip zone formed at a slip rate of 0.1 ms -1 is indicative of grain 305 rearrangement due to frictional sliding (see Masoch et al., 2019, and reference therein), referred to as the "Brazil nut" effect, a phenomenon observed when large grains move to the top of a fluidized layer due to differences in dispersal pressure between large and small particles (Williams, 1976). Grain size grading was reported by Boullier et al. suggests that water-dampened gouges experienced fluidization at slip rates between 30 µms -1 and 0.1 ms -1 . This is significant because textures related to fluidization in natural gouges and cataclasites are often interpreted to form during coseismic slip at 315 high velocities (e.g. Monzawa and Otsuki, 2003;Rowe et al., 2005;Boullier et al., 2009;Brodsky et al., 2009;Demurtas et al., 2016;Boulton et al., 2017;Smeraglia et al., 2017). Various mechanisms have been proposed to account for fluidization of granular materials in fault zones, including (i) frictional heating and thermal pressurization (Boullier et al., 2009), (ii) dilation that limits grain-grain contacts (Borradaile, 1981;Monzawa and Otsuki, 2003), and (iii) focussed fluid flow along slip zones during and after coseismic sliding (e.g. fault-valve mechanism of Sibson, 1990). In our experiments, temperature measurements 320 made during slip at 30 µms -1 suggest that significant frictional heating is unlikely, and therefore thermal pressurization is an unlikely mechanism within the slip zone. Water-dampened experiments are characterized by continuous compaction, which also excludes the dilation-related hypothesis of Borradaile (1981). We propose that fluidization in the principal slip zone might be caused by local fluid pressure increase within water-saturated patches as a result of continuous compaction combined with minimal fluid loss during deformation. A sudden release of water from the pressurized patch could result in gouge mobilization 325 and injection of material in to the adjacent slip zone.
In water-dampened experiments at 1 ms -1 , abrupt dynamic weakening preceded by a very short-lived strengthening phase has previously been documented in experiments on calcite gouges (Rempe et al., 2017) and calcite marbles (Violay et al., 2014). In gouges, Rempe et al. (2017) suggested that the rapid onset of dynamic weakening could be related to faster grain size reduction in the presence of water, leading to an early switch from brittle deformation to grain size sensitive creep in the principal slip 330 zone, analogous to the process suggested to occur in dry gouges (De Paola et al., 2015;Demurtas et al., 2019b;Pozzi et al., 2019). However, there is an apparent discrepancy between the relatively low maximum temperature measured close to the principal slip zone in water-dampened experiments (200 • C at V = 1 ms -1 ; Fig. 5a), and the observed CO 2 production ( Fig.   5b) combined with microstructural evidence for recrystallization during deformation ( Fig. 11f-g). As previously discussed, this could be due to an underestimate of peak temperature (Aretusini et al., 2019). Alternatively, Ohl et al. (2020) proposed that 335 mechanical liming (see Martinelli and Plescia, 2004) along natural faults could be a possible slip weakening mechanism that does not necessarily involve a macroscopic temperature increase of >500-600 • C.