Shallow creep and confined ruptures in frictionally unstable mica schist Authors: Sezim E. Güvercin, Sylvain Barbot, Lei Zhang, Mehmet Köküm, Zekang Yang, Cengiz Zabcı Published: 2026 ABSTRACT Crustal faults across geological settings are often enriched in phyllosilicates because of the hydrothermal alteration of wall rocks. The frictional behavior of these altered rocks at mid-crustal conditions remains poorly understood. Here, we show that mica-schist gouge under crustal hydrothermal conditions can host mixed behavior promoting shallow aseismic creep and confined seismic ruptures. We conduct triaxial deformation experiments on natural samples from the East Anatolian Fault, Turkey, and simulate the resulting dynamics on the Pütürge segment during seismic cycles using a physical model of rock friction and a realistic thermal structure. Mica schist exhibits potentially unstable weakening between 300° and 400°C, promoting depth-confined earthquakes. The frictional properties of mica schist are compatible with the 2020 moment magnitude (Mw) 6.8 Elazığ earthquake and the northern rupture arrest of the 2023 Mw 7.8 Kahramanmaraş event, illustrating how the thermal and lithological structure affects rupture style and segmentation. INTRODUCTION The metamorphic and metasomatic alteration of granitic fault zones involves the formation of phyllosilicates, typically micas and chlorites (1–3). Phyllosilicates in the shallow crust are mechanically stable, featuring dominantly aseismic deformation (4–11). However, single minerals (12, 13), synthetic mixtures (14–16), and natural samples (17–21) under hydrothermal conditions can feature unstable sliding. Understanding the frictional behavior of phyllosilicate-rich, altered rocks under mid- to lower-crustal pressures and temperatures is essential for assessing the seismic potential of metamorphic lithologies found in major faults at typical seismogenic depths. The East Anatolian Fault in eastern Turkey constitutes a unique natural laboratory to investigate the seismogenic potential of phyllosilicate-rich fault zones in a continental setting. The Pütürge metamorphics are exhumed along the >2400-km-long Bitlis-Zagros suture between the Arabian and Eurasian plates within the Pütürge segment (22–26) and are primarily composed of Upper Cretaceous metamorphic rocks in greenschist facies (27–29), including a Precambrian basement of gneiss, amphibolite, and mica schist, overlain by a Paleozoic-Mesozoic sedimentary sequence dominated by schist, phyllite, and marble (30) (Fig. 1). Geology and tectonics of the East Anatolian Fault. (A) Simplified geologic map of the Pütürge segment, highlighting lithological units, major fault traces, and coseismic surface ruptures (thick red lines) associated with the 2023 moment magnitude (Mw) 7.8 Kahramanmaraş and the 2020 Mw 6.8 Elazığ (epicenter in red star) earthquakes (32, 75). Geological mapping modified from (30). Active faults are shown by thin black lines (26). Inset shows the emplacement of Precambrian metamorphics along the Bitlis-Pütürge Suture Zone (pink shaded area). (B) Geological cross section along N-S modified from (76). The lithology of the Pütürge segment contrasts with adjacent segments dominated by carbonate, ophiolitic, and volcanic rocks. The resulting along-strike variations in fault-zone frictional properties may play an important role in the northern termination of the 2023 moment magnitude (Mw) 7.8 Kahramanmaraş earthquake and the confined rupture of the 2020 Mw 6.8 Elazığ earthquake (31, 32). The thermal structure of the East Anatolian Fault also showcases marked along-strike changes evidenced by surface heat flow measurements and Curie point depth estimates (33, 34). The latter reaches ∼23-km depth in the northeastern Pütürge segment, near the 2020 Mw 6.8 Elazığ earthquake, and decreases to ∼18 km toward the southwest, where the main fault trace intersects the Bitlis-Zagros suture and where the 2023 Mw 7.8 Kahramanmaraş rupture terminated. Here, we study the frictional behavior and seismic potential of mica schist, a laterally persistent component of the Pütürge metamorphics, as a phyllosilicate-rich end-member lithology. We conduct triaxial velocity-step experiments under hydrothermal conditions on a powder derived from Pütürge mica schist from room temperature to 500°C at effective normal stress of 100 and 200 MPa under 30 MPa of pore-fluid pressure. Using a calibrated physical model of rock friction and the three-dimensional thermal structure of the East Anatolian Fault, we simulate the dynamics of the Pütürge segment during seismic cycles. The simulations reproduce key characteristics of the 2020 Mw 6.8 Elazığ earthquake, including depth-confined coseismic rupture and subsequent shallow afterslip. Our laboratory experiments and numerical simulations indicate that mica-rich fault zones host preferentially confined seismic ruptures with a depth extent strongly modulated by crustal temperatures. RESULTS AND DISCUSSION Brittle rheology of Pütürge mica schist We initially consider two samples of Pütürge mica schist. The PTR1 outcrop at 38.3850°N, 39.1836°E is adjacent to the main branch of the East Anatolian Fault zone. We collected PTR2 farther south along the Elazığ rupture, at 38.34945°N, 38.97076°E (Fig. 1). Thin-section analysis of the samples reveals a foliated texture dominated by quartz and muscovite, visible under plane-polarized and cross-polarized light (fig. S1). X-ray diffraction analysis reveals similar compositions of the phyllosilicate-rich samples with 39–wt % quartz, 21–wt % albite, 2–wt % microcline, 22–wt % muscovite, 11–wt % clinoclore (magnesium-rich chlorite), and 4–wt % ettringite (a hydrous calcium aluminum sulfate) for PTR1 and 33–wt % quartz, 18–wt % albite, 2–wt % microcline, 28–wt % muscovite, and 19–wt % clinoclore for PTR2, with an estimated uncertainty of ±5%. The similarity in mineralogy suggests that both samples are representative of the Pütürge metamorphic belt. The presence of quartz, alkali feldspar, and plagioclase mixed with mica, chlorite, and ettringite indicates a metagranite under zeolite or prehnite-pumpellyite facies. Except for ettringite, the mineral composition is stable throughout the greenschist metamorphic grade, covering the depth range of the seismogenic zone. The formation of muscovite within granitic wall rocks is frequently observed in fault zones, associated with the alteration of potassium-feldspar or the circulation of potassium-rich fluids (1–3, 9, 18). Hence, the rheological behavior of Pütürge mica schist may be representative of altered rocks in granitic fault zones. We characterize the frictional properties of Pütürge mica schist using PTR1 as a representative specimen. We conduct triaxial deformation experiments under various pressure, temperature, and kinematic conditions. We apply two repeated sequences cycling through axial loading rates of 1.0, 0.2, and 0.04 μm/s at temperatures ranging from 100° to 500°C in 100°C increments under a constant pore fluid pressure of 30 MPa with an additional run at room temperature. We follow the experimental protocol twice under effective normal stress of ∼100 and 200 MPa, which increases slightly with cumulative slip (Fig. 2, A and B, and Materials and Methods). In total, we perform 12 experiments, each lasting 8 to 10 hours, depending on temperature, including time for sample preparation, loading, and velocity step execution. We analyze the frictional response using the RSFit3000 methodology (35) (Materials and Methods). The velocity-step experiments reveal steady-state velocity-strengthening friction at low temperatures and high slip rates, compatible with previous findings on phyllosilicate-rich gouge (11, 36, 37). However, PTR1 features steady-state velocity-weakening friction between 300° and 400°C, potentially enabling seismogenic behavior at mid-crustal depth. Friction properties of Pütürge mica schist. (A) Friction coefficient measured in triaxial deformation experiments [inset in (B)] at different temperatures (color) under 100-MPa effective normal stress and 30-MPa pore-fluid pressure. (B) Friction coefficient at different temperatures under 200-MPa normal stress. Loading rates include 0.05, 0.24, and 1.22 μm/s for all experiments and additional 0.0065, 4, and 20 μm/s steps at room temperature experiments. Each experiment consists of two consecutive sequences of velocity steps. (C) Experimental data (circles colored by temperature) and constitutive model (solid lines colored by temperature) (38, 39) for 0.24 to 0.05 μm/s velocity down-steps at different temperatures under 100-MPa effective normal stress. (D) Experimental data and constitutive model for velocity down-steps at 200-MPa effective normal stress. We reproduce the experimental results under various temperature and pressure conditions with a physical model of rock friction (38, 39). The model consistently describes the direct and evolutionary effects caused by background and perturbation of temperature, effective normal stress, and slip rate (Materials and Methods). The evolutionary effects originate from the flattening of contact junctions accommodated by crystal plasticity at the contact scale that increases the real area of contact. With a single set of constitutive parameters, the model reproduces the mechanical data under all experimental conditions (Fig. 2, C and D). The low-temperature transition from velocity strengthening to velocity weakening at steady state is associated with a switch of microasperity creep mechanism that only affects the intensity of the evolutionary effects. The transition to velocity-strengthening friction at higher temperatures and low slip rates corresponds to a switch to semibrittle deformation where the direct effect increases. With increasing effective normal stress, the magnitude of the direct and evolutionary effects decreases. Temperature structure of the East Anatolian Fault The laboratory experiments and constitutive law capture the temperature and pressure dependence of mica schist friction. We estimate the thermal structure along the East Anatolian Fault combining regional heat flow and Curie point depth, an independent proxy for the thermal base of the magnetized crust (Fig. 3A and Materials and Methods). After spatial interpolation of heat flow measurements (33), we optimize depth-dependent thermal profiles using a one-dimensional steady-state conductive geotherm with shallow radiogenic heat (Materials and Methods). We assume a surface temperature of 20°C, a 10-km-thick radiogenic layer producing 2.25 mW/m2, and spatially variable basal heat flow ranging from 50 to 110 mW/m2. We fit the surface heat flow measurements within ±3 mW/m2 along the fault zone (fig. S2). We estimate the uncertainty in the resulting thermal profile using a Monte Carlo approach (Materials and Methods and fig. S3). Temperature structure along the East Anatolian Fault. (A) Background heat flow distribution (33). Colored circles represent modeled surface heat flow, while black thin lines denote active faults. Thick black lines indicate the surface ruptures of the 2020 Mw 6.8 Elazığ and 2023 Mw 7.8 Kahramanmaraş earthquakes. The Pütürge metamorphics are outlined in the shaded area. (B) Thermal profile of the Pütürge segment with 50°C isotherms, 1-m and 50-cm coseismic slip contours of the 2023 Mw 7.8 Kahramanmaraş and 2020 Mw 6.8 Elazığ earthquakes (hypocenter in yellow star), respectively (32, 75), 2020 to 2023 seismicity (41, 42) (black circles), and along-fault Curie point depth (33, 34) (thick black line). Segment boundaries (Palu, Pütürge, Erkenek, and Pazarcık) are labeled at the top. The resulting thermal structure reveals a pronounced temperature contrast between the northeastern and southwestern segments of the East Anatolian Fault (Fig. 3B) of ∼100°C at 10-km depth, particularly prominent at the transition between the Erkenek and Pütürge segments, where the fault trace intersects the Bitlis-Zagros thrust zone. The southern segments exhibit higher surface heat flow, thinner crust, shallower Curie point depths, and elevated geothermal gradients. The Pütürge metamorphics are characterized by lower surface heat flow, thicker crust, and lower geothermal gradients (40) (fig. S4). The transition to a high background seismicity between 10- and 15-km depth including repeating earthquakes (32, 41–43) (Fig. 4) is plausibly driven by elevated pore-fluid pressures in phyllosilicate networks, distributed shear in a mylonitic fault zone, and creep transients at the boundary between velocity-weakening and velocity-strengthening domains (44, 45). Thermal and pressure controls on seismic cycles in mica schist. (A) Schematic illustration of lithological and frictional controls on rupture along the Pütürge segment. The blue shaded area represents the mica schist zone, with the black box marking the frictionally unstable zone constrained by laboratory experiments and the calibrated physical model of rock friction. Geological units are adapted from (30). Interseismic creep (red curve) is from (77). Black dashed lines indicate the coseismic slip of the 2020 and 2023 earthquakes. The Pütürge segment hosts a cold, thick seismogenic zone, in contrast to the warmer, thinner zone of western segment. (B) Temperature and effective normal stress distribution across the cold northeastern vertical profile. (C) Seismic-cycle simulation along the cold profile with coseismic (red) and interseismic (blue) slip. Hypocenters marked by yellow stars. (D) Comparison of coseismic slip distribution (red curve) from the 2020 Mw 6.8 Elazığ earthquake (32) with the experimentally derived steady-state friction velocity dependence parameter (a-b) of Pütürge mica schist (solid and dashed curves for slip rates of 1 and 0.1 μm/s; blue and orange curves for 100- and 200-MPa effective normal stress. The background histogram shows the normalized number of earthquakes with M > 2 between 2020 and 2023 as a function of depth between 38°E and 38.5°E (43). (E to G) Same as (B) to (D) for the warm-temperature profile. The background histogram in g shows the normalized number of earthquakes between 39° and 39.5°E. The simulations reproduce the first-order depth confinement of the Elazığ rupture, despite minor differences in absolute depth extent. Seismic cycle simulations with mica schist We incorporate the experimentally derived constitutive parameters of PTR1 to simulate the seismic cycle on the Pütürge segment. Although other rocks can be found in the fault zone, we consider mica schist as an end-member lithology illustrative of near-complete hydrothermal alteration of potassium feldspar into muscovite. Amphibolite, another component of the Pütürge metamorphics that exhibits velocity-strengthening friction at seismogenic depths (15), inhibits earthquake nucleation and rupture propagation. We simulate the evolution of fault slip in a two-dimensional elastic half space using the boundary integral method with the radiation damping approximation implemented in “Unicycle,” capturing the dynamic feedback between slip rate and temperature associated with elastic interactions and shear heating during the seismic cycle (Materials and Methods). The background temperature and effective normal stress govern the depth of the seismogenic zone and the amplitude of coseismic slip (fig. S5). We consider two vertical profiles of the East Anatolian Fault at the northeastern and southwestern sections of the Pütürge segment characterized by warmer and colder conditions, respectively (Fig. 4A). In both cases, the effective normal stress increases gradually from 10 MPa at the surface to a maximum of 250 MPa at 12-km depth, representative of lithostatic pressure under hydrostatic pore-fluid pressure (Fig. 4, B to E). The simulations reveal a complex sequence of creep episodes and earthquakes with overlapping footprints (Fig. 4, C to F). The ruptures are confined to the seismogenic zone and followed by shallow afterslip for a few years. Shear heating generates a transient temperature rise of ∼100°C during coseismic ruptures (fig. S6), with negligible impact on fault dynamics. However, the background thermal profile exerts an important control on the depth of earthquakes. Along the northeastern cold profile, simulated earthquakes are confined between 6- and 14-km depth, produce of the order of 1 m of coseismic displacement, and are followed by afterslip in the top 7 km, similar to the coseismic and postseismic sequence of 2020 Mw 6.8 Elazığ earthquake. Each seismic rupture is preceded by a slow-slip event at seismogenic depth. Along the southwestern warm profile, the ruptures produce a similar coseismic slip but are confined to a narrower 4- to 10-km depth range. Although creep propagates into the locked zone during the interseismic period, no slow-slip events occur before seismic ruptures. Seismic cycles with depth-confined ruptures and pervasive shallow creep are a robust feature of mica-schist friction. Numerical simulations with the empirical slip rate and state-dependent friction law (46) with depth-dependent parameters reproduce these characteristics (fig. S7). However, assuming a physical model of rock friction leads to a more complex, nonrepetitive sequence of ruptures confined to variable depths and preceded by accelerated creep waves, highlighting the velocity dependence of frictional stability and the sensitivity of rock friction to the underlying temperature and pressure of the continental crust. Other rheological effects can further influence dynamic ruptures due to shear heating, including dehydroxylation of phyllosilicates (47, 48), decomposition of carbonates (49, 50), pressurization of pore fluids (51, 52), and other forms of thermal weakening (53), possibly affecting the rupture style and size of earthquakes. Lithological and frictional controls of the Pütürge segment Phyllosilicate-rich rocks are found in many strike-slip systems, including the Alpine Fault, Red River Shear Zone, and San Andreas Fault (54–56), and in deep, fault-cutting boreholes (57, 58), highlighting their potential importance on seismogenesis. The exhumed metamorphic core under greenschist facies along the Bitlis-Zagros suture produces persistent rheological contrasts along the East Anatolian Fault that fundamentally alter the seismic cycle. Phyllosilicate-rich metamorphic rocks may actively contribute to seismogenesis depending on the ambient hydrothermal and pressure conditions. Mica schist, a key lithology of the Pütürge metamorphics, showcases velocity-weakening steady-state friction within a narrow range of temperatures, similar to the Pelona and Portal greenschists of the San Andreas Fault (20) and other phyllosilicate-rich natural samples (17–19, 21). This rheological framework may explain the peculiar rupture style of the Pütürge segment combining shallow aseismic creep and confined ruptures. The narrow temperature window for unstable friction in mica schist promotes depth-confined ruptures with a vertical extent controlled by the local geothermal gradient and pore-fluid pressure. Conversely, the wide temperature range of velocity-strengthening friction favors interseismic creep and widespread afterslip at shallow depths. The velocity dependence of frictional stability induces complex noncharacteristic sequences of slow-slip events, earthquakes, and afterslip with overlapping footprints, a behavior that challenges conventional models of the seismic cycle based on persistent unstable asperities surrounded by stably sliding regions. The northeastern termination of the 2023 Kahramanmaraş earthquake was likely controlled by a combination of structural, rheological, thermal, and kinematic contrasts as the nearby Bitlis-Zagros suture exhumes the Pütürge metamorphics and affects the geometry of the East Anatolian Fault (26, 41, 42, 59, 60). In addition to structural changes, the propagation of the Kahramanmaraş rupture into the Pütürge segment was further hindered by the presence of mica schist in a warm crust associated with a thin seismogenic layer between 4- and 10-km depth. The 2020 Mw 6.8 Elazığ rupture also took place in mica schist but in a colder crust that offers a wider seismogenic zone, demonstrating how along-strike thermal gradients can modulate rupture dimensions within the same lithology. Given the velocity-weakening behavior of mica schist at seismogenic depths, the remaining Pütürge seismic gap bears the potential for a future large earthquake (Fig. 4A). Our findings confirm the seismogenic potential of phyllosilicate-bearing faults. The Pütürge mica schist illustrates the thermal and lithological controls on the seismic cycle, governing earthquake size and recurrence in metamorphic terranes and altered fault zones. Extending this framework to other settings may help reconcile diverse fault dynamics with their physical environment. MATERIALS AND METHODS Sample preparation for triaxial deformation experiments We conduct velocity-stepping experiments in a triaxial deformation apparatus. We first crush and sieve the PTR1 sample of Pütürge mica schist to obtain a gouge material with particles smaller than 200 μm. The gouge is confined between two cylindrical gabbro blocks with a precut fault plane inclined at 35°. To enhance mechanical interlocking, the edges of both gabbro blocks are roughened using a 200# abrasive paper and further textured with emery powder. Two fluid-access holes are drilled into the upper gabbro block, each packed with a brass filter to prevent gouge intrusion while allowing pore fluid entry. The PTR1 powder is mixed with deionized water to form a consistent paste, which is applied as a 1-mm-thick layer onto the fault surface of the lower gabbro block. The assembly is encased in a 24-mm copper jacket, with alumina and carbide spacers positioned symmetrically around the saw-cut sample to ensure even loading. To isolate the sample chamber, double O-ring seals are installed at both ends of the copper jacket. The jacket was then connected to the furnace and packed with boron nitride powder. Boron-nitride-saturated cotton is used to seal the top of the assembly, minimizing heat loss during the experiment. Additional details of the experimental configuration are described in (61). We conduct the velocity steps experiments under 30-MPa pore fluid pressure maintained by servo control. We maintain the pore-fluid pressure, confining pressure, and temperature constant and observe the evolution of shear stress versus displacement during velocity steps. Each experiment at a given confining pressure and temperature takes about 1 day. Because of the coupling between shear and normal stress in a triaxial setting, the effective normal stress increases slightly with increasing slip. We measure the velocity dependence of the friction coefficient, defined as the ratio of shear stress to effective normal stress, accounting for the effect of pore-fluid pressure, during stepwise variations of the axial loading rate. During postprocessing, we correct the stress measurements for the reduction of contact area along the saw-cut as a function of load-point displacement. Modeling with an empirical friction law We describe the temperature and pressure dependence of the frictional behavior; we analyze the friction measurements with an empirical friction law. We model the frictional response of PTR1 using the RSFit3000 methodology (35) using a slip-rate and state-dependent friction law under isothermal, isobaric conditions (46), providing the evolution of the direct effect and steady-state velocity dependence parametersa=∂μ∂lnVa‐b=∂μss∂lnV(1)where μ and μss are the friction coefficient and its steady-state value, respectively, as a function of background slip rate, temperature, and effective normal stress (fig. S8). Fit with the empirical law also provides the characteristic weakening distance L for the evolutionary effect. The best-fitting parameters a, a-b, and L are shown in fig. S9 for the aging law and the slip law. We do not use the empirical friction law in the seismic cycle simulations (Fig. 4). Instead, we use the physics-based constitutive law described in the next section. Pressure-dependent and thermally activated mica schist friction We interpret the mechanical behavior of mica schist using a constitutive law incorporating the thermal activation and pressure dependence of rock friction (39, 62–67). The mechanical response is described as a flow lawVV0=∑k=1M(τμ0σ¯)nk(dd0)−αnk(σ¯σ0)βnk−ζkexp[−QkR(1T−1Tk0)](2)where V≥0 and V0=1 μm/s are the norm of the slip velocity vector and a reference value, respectively; M = 2 is the number of fault-parallel shear deformation mechanisms; μ0 is reference friction coefficient; τ≥0 and σ¯ are the norm of the shear traction vector and the effective normal stress, respectively, with the power-law exponent nk; d≥0 and d0=1 m are the size of microasperities forming the real area of contact and a reference value, respectively, with the power-law exponent α≥0; σ0 is a reference effective normal stress with power-law exponents β and ζk; T is the absolute temperature expressed in Kelvin; and Qk and Tk0 are the activation energy and activation temperature, respectively, with the universal gas constant R. The subscript k refers to a given deformation mechanism. The model captures the brittle to semiductile transition by integrating M = 2 stress, state, temperature, and pressure-dependent deformation mechanisms for fault-parallel shear. The size of microasperities evolves because of plastic deformation around contact junctions and the random rejuvenation of contacts during fault slip (62). Normal stress drives the thermally activated creep at contact junctions that leads to gouge compaction (65). The evolution law that governs the size of microasperities can be cast as the aging-law end-member, given by (66)d˙d=∑k=1Nf0(dd0)−pk(σ¯σ0)qkexp[−HkR(1T−1Tk)]−λV2h(3)where d is the average microasperity size of contact junctions; f0 is a reference strain rate for plastic compaction of microasperities; Hk and Tk are the activation enthalpy and activation temperature for the thermal activation of healing, the power-law exponents pk and qk for the size and effective normal stress, respectively; h is the gouge layer thickness, and 1/λ is a characteristic strain for contact rejuvenation, involving N = 2 healing mechanisms. Alternatively, the evolution law can be expressed as the slip-law end-member, following (66)d˙d=λV2hln2hλV∑k=1Nf0(dd0)−pk(σ¯σ0)qkexp[−HkR(1T−1Tk)](4)involving the same constitutive parameters. Equations 3 and 4 produce the same microasperity size at steady state. Together, Eqs. 2 and 3 or Eqs. 2 and 4 explain the frictional behavior of rocks for a wide range of physical conditions, including the direct and transient response to temperature steps (62, 66), normal stress steps (65), and velocity steps across the brittle-to-flow transition at various background slip rates and temperatures (63, 64, 66, 68). We model the evolution of the frictional resistance of PTR1 upon velocity steps using a spring-slider system with a stiffness of 0.2 MPa/μm, where the slider obeys the constitutive framework described above. The frictional response of PTR1 to velocity steps engages M = 2 shear deformation mechanisms and N = 2 healing mechanisms that dominate the mechanical behavior in different ranges of temperature, pressure, and slip rate, including intermediate situations where multiple mechanisms contribute to deformation in coeval proportions. The switch of dominant healing mechanism leads to a transition from stable steady-state friction at 25° to 200°C to unstable friction at 300°C by increasing the evolutionary effects while keeping a similar direct effect (Fig. 2). The switch of fault-parallel shear deformation mechanism triggers a transition from velocity weakening at 300°C to velocity strengthening at 400°C and above by increasing the direct rate-strengthening effect and reducing the evolutionary effects. We model healing of mica schist gouge using two mechanisms, each governed by pressure- and temperature-sensitive kinetics. We compare the model predictions of the frictional parameters with the laboratory observations. The direct and steady-state velocity dependence parameters can be obtained using Eq. 1, allowing direct comparison with laboratory data (figs. S10 and S11). We reproduce the observed transitions: from velocity strengthening at low temperatures to velocity weakening around 300° to 400°C, and ultimately to semibrittle deformation at higher temperatures. Some residuals remain because of the unaccounted dependence of the frictional properties with total strain and trade-offs arising during parameter estimation. The laboratory measurements can be explained consistently using constant parameters by invoking M = 2 brittle deformation mechanisms and N = 2 healing mechanisms with the constitutive parameters shown in Table 1. Table 1.Constitutive model parameters for mica schist PTR1, considering the reference parameters d0 = 1 m, V0 = 1 m/s, and σ0 = 100 MPa. Area of contact ParameterValueReference friction coefficientμ00.5Microasperity size exponentα0.1Normal stress power exponentβ0.4 Flow law Brittle deformation Stress power exponent n 1 80Activation energy (kJ/mol) Q1 10Reference temperature (°C) T10 20Compressibility factor ζ1 6 Semibrittle deformation Stress power exponent n2 34Activation energy (kJ/mol) Q2 100Reference temperature (°C) T20 425Compressibility factor ζ2 4 Evolution law Low-temperature mechanism Size power exponent p1 18Normal stress power exponent q1 4Activation energy (kJ/mol) H1 10Activation temperature (°C) T1 0 High-temperature mechanism Size power exponent p2 5.6Normal stress power exponent q2 0Activation energy (kJ/mol) H2 90Activation temperature (°C) T2 290 Contact rejuvenation Characteristic strain 1/λ 3Shear zone thickness (mm) h 1 Constitutive model parameters for mica schist PTR1, considering the reference parameters d0 = 1 m, V0 = 1 m/s, and σ0 = 100 MPa. Thermal model of the East Anatolian Fault zone We use a steady-state solution of the heat conduction equation incorporating radiogenic heat production to estimate the thermal structure along the East Anatolian Fault. In a one-dimensional, depth-dependent model, the conservation of heat in a stable continental lithosphere is governed by (69)k∂2T∂z2+Q=0(5)where k is the thermal conductivity, T is the temperature, z is the depth, and Q is the volumetric radiogenic heat production (70, 71). Radiogenic heat production is assumed to decay exponentially with depth following (72)Q(z)=Q0exp(−zhr)(6)where Q0 is the surface radiogenic heat production and hr is the characteristic decay depth. The downward heat flow is defined asq=−k∂T∂z(7) We assume that the upward heat flow at great depth is given byql=−limz→∞q(z)(8) Combining Eqs. 5 and 6 and integrating once, we obtain the heat flowq(z)=−ql−Q0hrexp(−zhr)(9) Hence, the upward surface heat flow q0=−q(z=0), a measured quantity, is given byq0=ql+hrQ0(10) A second integration of Eq. 5 yields the temperature profileT(z)=T0+qlkz+Q0hr2k[1−exp(−zhr)](11)where T0 = 20°C is the surface temperature. Thermal models compatible with measured heat flow and surface temperature can be constructed for any combination of basal heat flow ql and radiogenic-layer thickness hr. We discuss various end-member thermal models on the basis of the Curie point temperature Tc = 570°C and Curie point depth zc. For any choice of Curie point depth and radiogenic-layer thickness, the corresponding heat flow at the base of the radiogenic crust is given byql=kTc−T0zc−Q0hr2zc[1−exp(−zchr)](12) This approach enables the calculation of a laterally variable temperature distribution along the fault based on regional heat flow data, which is then used to estimate depth-dependent frictional properties in our earthquake cycle simulations. The residual heat flow varies within ±3 mW/m2, and its spatial distribution is shown in fig. S2. Uncertainty in the thermal model Uncertainty in the temperature profiles is estimated using a Monte Carlo approach that propagates uncertainties in both observational constraints and model parameters. For each location, we generate an ensemble of thermal models by perturbing surface heat flow, Curie point depth, thermal conductivity, radiogenic-layer thickness, surface temperature, and Curie temperature within prescribed ranges. For each realization, the basal heat flow is recalculated by enforcing the Curie temperature constraint, and the corresponding temperature-depth profile is computed using the same steady-state formulation. Uncertainty is quantified using the median and percentile ranges (16 to 84% and 2.5 to 97.5%), providing a nonparametric description of the temperature distribution at each depth. This approach captures the combined effect of uncertainties in both the thermal boundary conditions and the parameters controlling heat production and conduction. Seismic cycle simulations with a physical model of rock friction We use two-dimensional numerical simulations of the seismic cycle along the Pütürge segment of the East Anatolian Fault. The model consistently captures all phases of the seismic cycle, encompassing the nucleation, propagation, and arrest of earthquakes, relocking, and possible fault creep. To study dynamic rupture behavior, we model faults in an elastic half-space using the boundary integral method (73, 74). We discretize vertical strike-slip fault with 1000 patches of 20-m width, sufficiently fine to resolve the cohesion length at the rupture front during frictional instabilities without modifying laboratory-derived rock properties. The simulations incorporate the constitutive framework described in the previous section, where temperature is treated as a dynamic variable. Temperature evolution is approximated using a membrane diffusion model, following the governing equationT˙=−DW2(T−Tb)+τVwρc(13)where D is the average thermal diffusivity in the surrounding damage zone, the product ρc is the specific heat per unit volume in the fault zone, and Tb is the bath temperature outside the fault zone, which is considered in drained isothermal conditions. W is the characteristic width of heat diffusion, and w is the thickness of the active shear layer.