Strong variations in fault structure and fault geometry
Figures 3 and 4 show the map views and vertical profiles of the seismic parameters (Vp, Vs, and η) and porosity parameters (ε and ζ) in the crust. Here we define a rigid seismogenic zone as being characterized by high Vp and Vs values but low η, ε, and ζ values, indicating dry, dense, and strong rocks; in contrast, a ductile zone is defined by low Vp and Vs values but high η, ε, and ζ values, reflecting watery, porous, and weak rocks20,21. In accordance with the aforementioned definitions, the Southeast Anatolian Thrust (SAT) boundary demarcates significant characteristics of crustal structural variations. On the right-hand side, we observe high-Vs anomalies accompanied by low Vp/Vs, –ε, and –ζ anomalies. Conversely, opposite anomalies of these parameters are observed on the left-hand side. Although at certain depths or regions these differences may be less pronounced than anticipated (e.g., at 3 km depth as shown in Fig. 3), substantial variations do occur at greater depths. Specifically, within the depth range of 11 to 20 km, notable changes in Vs, Vp/Vs ratio, and fracture parameters are evident (as illustrated in Fig. 3). Generally, except for the Vp structure, other parameters exhibit distinct contrasts adjacent to the SAT boundary when depths exceed 3 km. These observations are commensurate with the anomalous structural patterns observed by e.g., shear wave velocity, magnetotelluric conductivity, Pn-wave speed, and Lg attenuation on either side of the SAT boundary22,23,24,25,26,27,28,29,30,31. Such mutually corroborated structural features from independent studies confirm that these changes are genuine, indicating that the East Anatolian Fault (EAF) likely plays a significant role in tectonic movements between the Anatolian and Arabian plates at shallow depths ( < 3 km) (Fig. 3a). Specifically, our seismic tomography results reveal distinct crustal structural variations along the EAF at a depth of 3 km (Fig. 3a), suggesting that this boundary is crucial for regulating stress distribution and crustal deformation in the shallow upper crust. Moreover, GPS data indicate that the slip rate of the EAF zone exceeds that of adjacent regions32, highlighting its pivotal role in plate tectonics. The spatial distribution of fault slip rates correlates strongly with areas of high seismic activity10,11,33, implying that this fault serves as both a primary locus for stress accumulation and a significant conduit for stress release. The coseismic slip distribution and crustal deformation patterns along the EAF suggest that it regulates relative plate movement in the upper crust and influences the tectonic evolution of surrounding regions. However, in the middle crustal depths, the EAF does not exhibit distinct characteristics of being a tectonic boundary between these two plates (Fig. 3b, c). The structure inconsistency observed at different depths can be attributed to the oblique collision and obduction of the Arabian plate by the Anatolian plateau4,5. This observation implies that rather than solely involving strike-slip faulting, the Mw 7.8 earthquake potentially occurred within a complex fault system comprising reverse-thrust and strike-slip mechanisms in the eastern Anatolian plateau (Fig. 1a).
Two red stars indicate the Mw 7.8 and Mw 7.6 hypocenters. Gray dots indicate the aftershocks with 2–5 km thickness bound to each cross-section. The dotted color lines represent the same meanings as those shown in Fig. 3. Surface topography is plotted atop each profile. The profile locations are shown in the lower-right insert map. a-d Indicate the L1-L4 vertical cross-section, respectively.
Another prominent feature is the identification of contrast seismic structures at the hypocenters of the doublet. The Mw 7.8 earthquake occurred at the Narlı-Sakçagöz segment in the northern tip of the DSF, where it coincides with a transition from negative to positive changes in Vp, Vs, and ε parameters (Figs. 3b and 4a, b). The Mw 7.6 event, in contrast, is located within a distinct belt (D in Fig. 3c), where it is characterized by low Vp and Vs, as well as high η, ε, and ζ; the belt is approximately bounded by the Çardak fault to the north and the SAT boundary to the south (Fig. 3c). This sandwiched belt extends into the lower crust beneath the Çardak fault (Fig. 4c). These observed seismic and porosity patterns are also consistent with previous results at the hypocentral depths from MT and crustal attenuation studies24,25,26. Meanwhile, the Checkerboard resolution tests for seismic tomography also show high resolutions along the two main rupture zones of the doublet with different lateral grid spacings (see Supplementary Note 2 for details), suggesting that the observed anomalous structures at these depths are genuine features rather than artificial (Supplementary Figs. 1–4). The map views also reveal another notable characteristic, namely the presence of similar patterns in the Vs and ε (Fig. 3). The similarity between Vs and ε can be attributed to two factors: 1) Vs is highly sensitive to the presence of fluids in the crust, similarly, ε serves as an indicator for rock crack density which is also influenced by fluid content. Therefore, it is not surprising that the imaged structures of these parameters exhibit similar features; 2) According to Eq. (7), the calculated value of ε primarily depends on Poisson’s ratio, which has a positive correlation with Vs. Hence, ε also exhibits a positive correlation with Vs. Therefore, we assert that the similarities observed in Vs and ε parameters are reliable features rather than artificial errors.
In order to understand the geometry of the rupture faults, we conducted a DD analysis (see Supplementary Note 3 for details). Our results show that the majority of the small aftershocks (Mw > 1.5) are predominantly located at a depth shallower than ~18 km (Fig. 5a). At these depths, we observe pronounced high Vs anomalies corresponding to low η, ε, and ζ, a parameter range referred to brittle-like rocks (Fig. 4). In contrast, only a limited number of aftershocks are observed within regions where the same parameters but referred to ductile-like rocks (Figs. 3, 4). In addition, we observe multiple clusters of aftershocks with deeply dipping fault planes at the hypocenter area of the doublet, suggesting that the initiation of the earthquake doublet may have involved multiple faults (Fig. 5d, f). For instance, the source area of Mw 7.8 exhibits a prominent cluster of aftershocks at depths between 10–16 km dipping to the northwest; while shallower than 10 km, the aftershocks show a nearly vertical lineation of clustering that likely ruptured in the mainshock10 (Fig. 5d). We also observe several adjacent clustering of aftershocks either trending northwest or southeast, indicating multiple faults may have reactivated during the mainshock (Fig. 5d). In contrast, at the hypocenter area of Mw 7.6 earthquake, the aftershock pattern is much simpler, which shows a nearly vertical clustering of aftershocks occupying at depths shallower than 18 km (Fig. 5f); this aftershock pattern suggests that a primary fault was likely reactivated during the aftershock. Another intriguing feature of the aftershock sequence is the different distribution characteristics in multiple segments along the dual rupture zones (Figs. 4, 5b, e). We observed sparse decentralized clusters of aftershocks within northern Seg. 3 (Amanos fault), compact clusters of aftershocks within Seg. 1 (Pazarcik fault), and dense aftershocks within Seg. 2 (Erkenek fault) from the relocated aftershock sequence (Fig. 5b), indicating increased geometric complexity from Seg. 3 to Seg.1 to Seg. 2 (Figs. 5b, 6a). Focal mechanisms of the aftershocks (Mw > 4.0) reveal left-lateral strike-slip motion with steeply dipping fault planes along these fault segments34, which coincides with the observed aftershock clusters with deeply dipping angles from the aftershock sequence (Fig. 5b, d). Meanwhile, dense aftershocks are observed within Eastern Seg. 4 (Releasing Bend) and Western Seg. 5 (Savrun fault), except for sparse aftershocks within the connection zone between Segs. 4 and 5 (Figs. 4c, 5e). These segmental aftershocks exhibit positive correlations with surface structure heterogeneities, fault geometry, and coseismic slips. For instance, high-Vs is associated with dense aftershocks while low-Vs is associated with sparse aftershocks along the dual rupture zones (Fig. 6a, d). The aforementioned observations unveil the interconnections among fault segmentations, structural heterogeneities, surface deformation, and coseismic slips along the two primary rupture zones, providing a new insight into a comprehensive understanding of intricate fault structures and aftershock behaviors associated with multiple fault segments.
a 3-D distribution of earthquakes after relocation using the DD method17,18. Gray circles indicate the earthquakes and the magnitude scale (gray circles) is shown in the lower-right corner of this map. b–f Vertical cross-sections of aftershocks along the lines shown in map a. The thickness of the aftershocks bond to each profile is 2-3 km. Red dashed lines in the d and f profiles show the speculated faults involved in the doublet ruptures. In this study, the focal depths of the Mw 7.8 and Mw 7.6 events (indicated by two red stars) relocated using the DD location method are 11.21 km and 15.75 km, respectively.
a and d The Vs structure at a depth of 3 km along the rupture zones of the Mw 7.8 and Mw 7.6 earthquakes. Colored solid lines indicate the ruptured fault segments along the EAF and CF main rupture zones. Red stars show Mw 7.8 and Mw 7.6 hypocenters, respectively. b and e Coseismic slip along the two main rupture zones, which are calculated using the rupture model developed by Ren et al.10 c, f Profiles of Vs structures along the double rupture zones. The dashed lines represent the same meanings as those shown in Fig. 3.
We also assess the correlation between the seismic structures, foreshock occurrences, and aftershock patterns of the 2023 Mw 7.8-7.6 earthquake doublet. The spatial distribution of foreshocks during the period from Jan. 1, 2019 to Feb. 6, 2023 indicates their predominant occurrence in three distinct regions: the Northeastern tip of the EAF zone, the western region near the dual rupture zones, and within close proximity to the hypocenter of the Mw 7.8 shock (highlighted by red dot lines in Supplementary Fig. 5). Over a span of four years, only two foreshocks with magnitudes around Mw 5 were recorded near the Northeastern region of this fault system (Supplementary Fig. 5). This implies that the plate stress accumulated by the EAF and Çardak fault has not been released through foreshocks. In such a scenario, the level of stress built up along these two fault zones could be sufficiently high to trigger a destructive earthquake doublet. On the other hand, our seismic models (Figs. 3, 4) indicate that the high-Vs and low-η body (C1) is located close to the Narlı and Pazarcık segments (Seg.1), with these anomalies extending southwestward along the EAF zone, albeit with lower amplitudes (Fig. 4a). In addition, robust high-Vs body (C2) is also observed at the terminus of the Amanos segment. Conversely, along the Erkenek fault (seg.2) in an NE direction, an inverse trend is evident as low-Vs and high-η values are prominently present. The rupture was observed to propagate bilaterally in the SW and NE directions along the EAF zone. Specifically, high-magnitude aftershocks initially propagated towards the NE for approximately 60 kilometers along the EAF within about 40 s, coming to a halt at the boundary of a high-Vs and low-η body (C1). Subsequently, they transitioned to the SW segment characterized by weaker seismic activity associated with warm structure anomalies. Finally, frequent aftershocks occurred south of Amano’s fault, which is distinguished by high-Vs and low-η values (C2). It is noteworthy that this spatial distribution aligns with both aftershock activities and regions exhibiting relatively high-Vs and low-η characteristics. These distinctive features play a significant role in influencing aftershock behaviors along the strike variation of the fault structure of the EAFZ, as indicated by back projection analysis and seismic tomography studies33,34.
Seismogenic structure of rigid-ductile variation in the M
w 7.8 earthquake
w 7.8 earthquake
The rupture area of the Mw 7.8 earthquake and its subsequent aftershocks are well located within a fault zone where it exhibits positive or positive-to-negative variational seismic attributes (Figs. 3b, 4a, b), indicating that except for a short portion of the fault exhibits a rigid-ductile transition property, majority of the EAF is rigid. The high Vs, low η, ε, and ζ anomalous belt with relatively low Vp (belt A in Fig. 3b) coincides with the presence of Late Eocene Metamorphic rocks, as reported by a previous geological study30. The morphological features along the East African Fault (EAF) exhibit a flat and low-lying Arabian plain in the southern segment, while the northern segment is characterized by east-west trending mountain ranges35 (Supplementary Fig. 6). Consequently, there exists a diverse range of sedimentary rocks on either side of the EAF (Supplementary Fig. 6). The presence of high-Vp, –Vs, and low-η, –ε, and –ζ anomalies in the belt (11 km depth in Fig. 3) can be attributed to the existence of the Arabian Platform which primarily comprises a marine sedimentary succession deposited from the early Cambrian to middle Miocene period35,36 (Fig. 3 and Supplementary Fig. 6). This geological history has resulted in elevated seismic velocity and reduced porosity and fracture characteristics. The region labeled as B, characterized by low-Vp and –Vs structures alongside relatively high-ε anomalies (B in Fig. 3), is considered to indicate the presence of scatted Neogene-Quaternary sedimentary rocks in the upper crust rather than melting associated with magmatism. The primary explanation for this phenomenon can be attributed to the collision between the Arabian and Anatolian plates, resulting in the formation of a Neogene-Quaternary sedimentary reverse nappe overlying the Paleozoic sequence of the Arabian platform35,36. The presence of sedimentary rocks may lead to a reduction in seismic velocities, but it is not necessarily indicative of a significant alteration in the Vp/Vs ratio3. Additionally, it arises from volcanic activity, typically resulting in low Vp and Vs but very high Vp/Vs (>1.8) within seismic structures. However, our seismic models demonstrate decreased velocities (Vp, Vs) and lower Vp/Vs (<1.7) (Fig. 3b), contradicting the presence of volcanism. Therefore, we posit that the observed low-V body B with mediate Vp/Vs ratio as the presence of sedimentary rocks at shallow crust with high rock porosity (high ε), rather than volcanic rocks3,35,36.
Belt A exhibits varying seismic velocity and Vp/Vs ratio, as well as coseismic slip at depths10, indicating changes in rock properties such as mechanical strength, fault coupling (locked asperity), and decoupling (unlocked) along the EAF (Fig. 3b). Based on these parameter observations, two primary asperities were identified within belt A: one is located at the Mw 7.8 hypocenter (C1 in Fig. 3b), and the other is situated at the southernmost part of the EAF (C2 in Fig. 3b). We consider that the rigid faulting zone A, which is characterized by discrete and localized asperities C1 and C2, is primarily caused by the different strain rates and crustal structures on the two sides of EAF zone due to the oblique collision between the East Anatolian plateau and the northwest-directed Arabian platform4,5,20,21. The right side of the EAF is typically characterized by prominent pre-Maastrichtian sedimentary rocks or Middle Eocene Madan rocks, while the opposite side exhibits notable Mesozoic sedimentary formations, as documented in a previous geological study35. The Mw 7.8 mainshock initiated at the transitional boundary between asperity C1 and plastic body B. Asperity C1 displays a distinctive tomographic structure characterized by significant high anomalies in seismic velocities (Vp, Vs) but low porosity (ε, ζ) values (Figs. 3b, 4a), representing the rigid undifferentiated sedimentary sequence associated with the Arabian platform according to Yilmaz’s classification system35. The coseismic slip revealed in the C1 segment displays larger amplitudes compared to that in C2, which is characterized by relatively low-Vs, but high-η, –ε, and –ζ values when compared to C1 (Fig. 1c). These observations suggest that in general, regions with higher energy release during rupture tend to have higher seismic velocities and lower crack density and saturation rates; whereas lower energy release regions may be associated with a structure of relatively lower seismic velocities33,37 (Figs. 4a, 6).
The M
w 7.8 mainshock unclamped the Çardak fault and facilitated fluid intrusion
w 7.8 mainshock unclamped the Çardak fault and facilitated fluid intrusion
From the evidence of our revealed 3D seismic structures, together with the comprehensive analysis of the static Coulomb stress changes and localized geological settings, we propose that the occurrence of the Mw 7.8 earthquake significantly influences the advanced failure of the Mw 7.6 shock through two primary key contributors. That is the Coulomb stress increase and a significant reduction in effective normal stress along the Çardak fault, imparted by the former event (Fig. 7).
a The Mw 7.8 earthquake caused changes in Coulomb stress. b–d The Mw 7.8 event resulted in changes in shear stress, normal stress, and Coulomb stress along the Mw 7.6 rupture zone. Note that positive normal stress changes in c indicate the unclamping of normal stress acting on the Çardak fault. Note that positive changes in normal stress indicate tensile stress acting on the Çardak fault.
The calculated Coulomb stress changes reveal a marginal increase ( < 0.1 MPaa) in shear stress (Fig. 7d); while the normal stress changes are substantially high ( > 0.8 MPa), unclamping the Çardak fault (Fig. 7c). We hypothesize that the primary cause of the normal stress change lies in localized tectonics caused by the northwestward motion of the Arabian plate colliding with the counterclockwise rotating Anatolian plate. The Eastern Anatolian Plateau, situated at the convergence of the Anatolian Plate and the Arabian Plate, is significantly influenced by tectonic activities. The region’s geological structure primarily consists of the East Anatolian Fault (EAF) and the Çardak Fault (CF) (Fig. 1a), with these two fault systems collectively governing the relative motion of the plates. The EAF predominantly exhibits left-lateral strike-slip movement (Fig. 1a), which can be attributed to the differential rotation between the Anatolian Plateau and the Arabian Plate. Similarly, the CF demonstrates comparable kinematic characteristics, resulting from varying displacement rates on either side of the fault. Following the occurrence of the first Mw 7.8 earthquake, characterized by a left-lateral strike-slip motion, the CF also experienced a similar kinematic response (Fig. 1c). Consequently, this interaction creates a torque couple system in a large-scale lateral deformation within a triangular region between the EAF and Çardak fault (Fig. 1d). Among this system, the observed clockwise rotation of the eastern Anatolian plateau is localized and short-term, resulting from the relative movement of materials on both sides of the fault caused by the earthquake doublet. The EAF is a left-lateral strike-slip fault, with the left plane moving south or southwest and the right plane of the Çardak fault also moving in a south or southwest direction. This leads to a significant reduction in normal stress on the right end of Seg. 4 within the Çardak fault at the junction between EAF and subsequent mainshock fault (Fig. 7c), consequently causing a noticeable decrease in effective normal stress on Çardak fault plane. Therefore, we conclude that shear stress increases and unclamping of the fault results in a significant Coulomb stress increase at the hypocenter area (Fig. 7). The Coulomb stress amplitude is larger than 0.4 MPa which is way much higher than a typical value (0.1 MPa) expected as a trigger of an earthquake38,39 (Fig. 7d). The unclamping effect could open the fault, which may significantly enhance the porosity of the Çardak fault, thereby facilitating fluid migration within permeable and extended fault rocks (Figs. 3c, 4d, 7b). Together with our tomographic observation, we will discuss how fluid migration and intrusion could occur and accelerate the failure of the Mw 7.6 aftershock in more detail in the next section.
Nucleation mechanism of the M
w 7.8 earthquake
w 7.8 earthquake
Based on the comprehensive analysis presented above, we propose a possible seismogenic mechanism for the Mw 7.8 earthquake (Fig. 8a). That is, the sharpened boundary between ductile body B (deformed or strained fast), and brittle body C1 (coupled, and deformed or strained slowly) introduces a shear stress concentration (high Vs); when such stress amplitude reaches the threshold, an earthquake starts to nucleate, and finally ruptures in the Mw 7.8 mainshock (Fig. 6c). The brittle body C1 may represent a high-strength material and thus an asperity that could accumulate large stress amplitude (Fig. 4A in Ren et al.10), promoting initiation of the supershear rupture process at the early stage during the Mw 7.8 earthquake10,33,40,41 (Fig. 7a). In addition, the northwestward collision front of the Arabian plate exhibits bending, which could further enhance complexity in fault geometry and thus, the stress concentration and a heterogenous stress field that are ideal for supershear rupture propagation along both the Narli F and Pazarcik-Erkenek segments42,43,44,45 (Seg 1 in Fig. 6a). The junction between the Narli and Pazarcik-Erkenek further provides another geometry complexity that could promote the initiation and sustained supershear rupture on the Pazarcik-Erkenek segment43,44,46 (Fig. 1), although some studies have not observed clear rupture in this region34. This speculation is supported by a previous study that revealed a higher supershear rupture speed (4.0–4.5 km s−1) in Seg 1 (Fig. 6) compared to other segments10. A high-stress accumulated area could be released over the Mw 7.8 event, resulting in the largest surface rupture (4.5–6.7 m) measured by both the field survey and satellite geodesy11 (Seg 1 in Fig. 6a). Following this logical interpretation, whereas it is difficult to explain the Erkenek F segment (Seg 2 in Fig. 6b). Because the Erkenek F segment is characterized by relatively low Vs, but high η, ε, and ζ corresponding to a ductile-like body. However, this segment receives the supershear rupture from the Pazarcik F segment and continues conveying the rupture propagation process, resulting in significant slip and surface displacements10,11 (Fig. 6a). We hypothesize that when the front of the supershear rupture propagates into the Erkenek F segment, the rapid motion further opens the dense cracks that channel the saturated fluid migrating on the fault; Consequently, a fluid-rich fault could further reduce the coefficient of friction47,48,49, allowing dynamic weakening mechanisms, e.g., thermal pressurization, to occur and, eventually, slip significantly ( ~ 6 m) similar to the shallow large slip observed in the 2011 Mw 9.1 Tohoku-Oki and 1999 Mw 7.7 Chi-Chi earthquakes50. For the Amanos F segment (Seg 3 in Fig. 6), in general, it is characterized by relatively high η, ε, and ζ (Fig. 4a), thus corresponding to a mechanically weak fault zone and, therefore, a partially coupled area that could slip less in comparison with segments 1 and 2 along the EAF (Fig. 6c). This weakened fault segment likely explains why it experienced a subshear rupture propagation and moderate slip as observed in seismo-geodetic constrained slip model10. Rather than the Erkenek F segment, the crack density and fluid saturation rate are significantly lower; thus, a dynamic weakening mechanism, e.g., thermal pressurization, is less likely. Our 3D tomographic model, from a different point of view, uncovers the heterogenic nature of the physical properties of the rocks in the fault zone, which allows rupture propagating rapidly at some segments while slowly at other segments as reconciled from the seismogeodetic rupture models10,50,51,52,53.
Red lines delineate the Anatolian and Çardak faults, while two red stars indicate the dual hypocenters of the 2023 Türkiye earthquakes. a Sharp contrast seismogenic structure triggered a brittle fault deformation of the Mw 7.8 earthquake. The enlarged section proposes the seismogenic structures of the source areas (see text in detail). b Fluid intrusion triggered the Mw 7.6 earthquake. The enlarged section illustrates the migration of fluids from deep depths to the source areas through either intra-faults, opened cracks, or both (see text in detail).
Fluid intrusion may have accelerated the failure of the M
w 7.6 earthquake
w 7.6 earthquake
The Mw 7.6 earthquake occurred within a low Vp and Vs, but high-η, –ε, and –ζ anomalous belt parallel to the Çardak fault, suggesting the presence of a ductile seismogenic zone along the SAT boundary (D belt in Fig. 3c). We interpret belt D as a ductile seismogenic zone based on variations of the seismic and porosity parameters (Figs. 3c, 4c, d). It should be noted that seismic velocity decreases can occur due to either mineral hydration or rock pore opening13,15,54. Mineral hydration beneath belt D is unlikely due to the dlnVs > dlnVp relationship13, along with a strong high-η value (Figs. 4c, 5c). Additionally, a geological study revealed a common Palaeozoic stratigraphy in the Eastern Anatolian plateau55, suggesting the absence of any significant geological discontinuities across the Çardak fault. Therefore, we believe that the above interpretations of these porosity parameters (ε, ζ) could be reasonable assuming there are no lateral heterogeneities in geological settings. Under these circumstances, when fluids intrude into the partially or fully porous rocks, a decrease in Vs and an increase in η can be observed15,54. This is supported by our tomographic models (Figs. 3b, 4c, d) as well as a previous geological study55. The ductile belt D is primarily attributed to a wedge-shaped accretion complex or thrust-fold nappe at the subduction front of the northward Cyprus slab56. This complex/nappe typically comprises weak and porous sediment with fluid retention capabilities, potentially resulting in reduced seismic velocity and an increased η ratio (Figs. 3c, 4c, d). Meanwhile, the dehydration process during the northward subduction of the Cyprus slab, followed by its detachment and sinking26,28,57,58,59, could release a substantial amount of fluids from the lower crust or uppermost mantle into the rocks beneath the hypocenter33. Alternatively, the presence of the ductile belt could also be attributed to the partial melting of crustal magma chambers and extensive occurrence of ophiolitic and young volcanic rocks in the Eastern Anatolian Plateau, as supported by previous high heat flow measurements60,61. The Cyprus slab’s northward subduction along the Anatolian plateau’s eastern margin leads to a volcanic arc formation along the SAT boundary due to subduction, accompanied by an accretion complex resulting from collision (Fig. 1a, b). In this instance, belt D (Fig. 3c) may indicate anomalous crustal heating and potential melting resulting from heightened heat flux originating in the lower crust. Indeed, recent petrological, geochemical, and seismological investigations suggest that during the Pliocene-Quaternary epoch, crustal rocks may have experienced partial melting at depths ranging from 20 km to 50 km within regions characterized by the low Vs and high η zones in the Tibetan Plateau3,62. The similar geological settings between our study region and the eastern Tibetan Plateau suggest that analogous temperature patterns may occur at similar crustal depths, indicating the likelihood of hydrous melting occurring in the middle-lower crust under the Çardak fault (Figs. 4c, d, 8b). At these depths, the magnetotelluric survey demonstrates that a high conductivity feature does exist in the eastern Anatolian plateau22,23, indicating that melting at the deeper portion of the Çardak fault is highly possible.
Therefore, we conclude that the rocks within the Çardak fault zone (belt D) are ductile-like, and are weaker than that in the EAF zone (Figs. 4, 6c). Underneath the Çardak fault, deeper than ~20 km, the ambient rocks are saturated with fluid and have high temperatures3,13,62. As the result of the unclamping effect from the Mw 7.8 mainshock, the Çardak fault is opened, and its porosity is significantly raised in the Mw 7.6 source area (D in Fig. 4c, d). Previous studies, such as those by Guglielmi et al.63, have demonstrated the effects of earthquakes on rock permeability and fluid migration, indicating that seismic activity can substantially modify these properties. Additionally, existing research highlights that changes in crustal normal stress and crack propagation induced by seismic events can significantly expedite fluid migration, especially within highly permeable fault zones where rapid fluid movement is observed64,65. In this framework, the saturated fluid from both the deeper depth and sides of the Çardak fault could easily migrate along the pathway through dense cracks on the fault towards the hypocenter and diffuse along the fault3,15,16,32,66. In addition, intra-crustal faults can act as efficient conduits for deep fluid migration within the crust57,58,59,61,66. Consequently, facilitated by the neighboring intra-crustal faults, the fluid migration and diffusion processes could occur more efficiently, enabling the Mw 7.6 earthquake to develop rapidly. Therefore, it is theoretically plausible that the fluid migration process, which enables such an event, could be completed within a 9-hour timeframe following the Mw 7.8 earthquake. Similarly, the foreshock of the continental 2019 Ridgecrest doublet induced a similar clamping effect on some portions of the mainshock fault that eventually ruptured during the mainshock (Fig. 8 in Qiu et al.19). However, the mainshock rupture was delayed by 34 h about 4 times longer than the Türkiye sequence. This disparity may suggest that saturation rate and crack density are not as fully developed as that in the Çardak fault area67; thus, a limited fluid concentration and lesser conduits tend to change the fluid diffusing slowly and ineffectively and eventually take a much longer time to trigger the mainshock.
The widely accepted notion is that a weak or ductile-like fault segment is expected to slip slowly or aseismically. The seismogenic zone of the Çardak fault is such a region that is characterized by low seismic velocities (Vp, Vs), and high η, ε, and ζ values; therefore, it should be too ductile-like to accumulate strain energy and a low coupling ratio during the interseismic period of seismic cycles. Surprisingly, this zone slipped a wide depth range with a nearly uniform slip amplitude during the Mw 7.6 earthquake (Fig. 6b). The maximum slip reaches 12 m, which is similar to that of the Mw 7.8 mainshock10. The observation contradicts our well-accepted expectation which makes it challenging to explain the mechanism. We hypothesize that the high saturation rate and high temperature60 offer sufficient hot fluid contents that intrude toward the seismogenic zone of the Çardak fault, as aided by the opening effect imparted by the Mw 7.8 mainshock. In this framework, when the fluid concentration is sufficiently high enough and reaches saturation in the seismogenic zone after a 9-hour migration and diffusion process. At this stage, the effective normal stress reaches the lowest level, particularly in the hypocenter area. Therefore, the rupture starts at the hypocenter area and propagates over the hot and saturated fluid area in the seismogenic zone. Under the shear-heating effect of the rupture propagation, the seismogenic zone is heated up, possibly resulting in a thermal pressurization weakening mechanism that pushes the fault slip largely similar to the large shallow slip observed in the 2011 Mw 9.1 Tohoku-Oki and 1999 Mw 7.7 Chi-Chi earthquakes50,51,68.
Implications for long-term seismic hazard
Our detailed 3D tomographic results facilitated a comprehensive understanding of the diverse and complex seismogenic mechanisms of the 2023 Türkiye earthquake doublet and shed light on why the Mw 7.6 quake occurred in a ductile-like seismogenic zone 9 h after the Mw 7.8 earthquake. Following the earthquake doublet, the regional stress field has been adjusted, with many places where the stress increases reaching 0.1 MPa (1 bar) or even higher critically to trigger future seismicity (Fig. 8a). Most of these stress-raised places are located on the EAF and DSF segments where it was ruptured by strong earthquakes historically, and now they experience a spatiotemporal variation of fault slip deficit4,6,8. The EAF has experienced a period of seismic quiescence ranging from 130 to 500 years since significant earthquakes occurred in different segments4,8,69: a Mw 7.1 earthquake in the northern segment in 1893 (Seg 2 in Fig. 1b), aMw 7.4 earthquake in the middle segment in 1513 (Seg 1 in Fig. 1b), and two Mw 7.2 and 7.5 earthquakes occurred in 1872, 1822, respectively in the southern segment (Seg 3 in Fig. 1b). If we calculate the slip deficit since the rupture of these earthquakes with a slip rate of 9.2 ± 0.5 mm yr−1, then slip release during the Mw 7.8 mainshock overshuts the slip deficit4,56, which suggests that the strain release on the EAF may obey the supercycle behaviors collectively releasing strain over multiple previous cycles70. In this framework, for some segments if it slipped less in the current cycle, then it would be dangerous for subsequent rupture; particularly, the segment between the northern tip of the Mw 7.8 rupture to the eastern tip of the 2020 Mw 6.8 rupture requires urgent attention for following comprehensive geophysical, geodetic and structural observations as it was stressed by the earthquake doublet (Figs.1a, 8a). Similarly, the DSF holds a well-identified seismic gap at the northern tip of the fault. In this gap, the latest earthquake occurred in 1170 with a Mw between 7.3–7.5; then, the gap kept quiet without rupturing as major earthquakes over the past ~850 years71. With an average slip rate of ~7 mm yr−1, this portion of the fault could have accumulated ~6 m of slip deficit71. After the failure of the 2023 Türkiye earthquake doublet, the gap was stressed and probably ready for the next great earthquake as occurred in history71. We call for following attention for future seismogenic and geodetic observations, and armed with our powerful tomographic technique, eventually gain a comprehensive understanding of the characteristics of tomographic structures, deformation patterns, and stress evolution processes for a better assessment of future seismic hazards and better preparedness for mitigation strategy in this region.