Introduction

Along the subduction interface, the transition zone between the seismogenic zone (hosting regular earthquakes of up to M > 9) and the creeping zone (steady deformation accommodated by ductile shearing) embraces a depth range between about 30–40 km updip to 60–80 km downdip and is characterized by several instrumentally-constrained kinematic and physical conditions1,2. Kinematic observations include deep slow slip events3,4, which are sometimes seen to correlate with fluctuations of Vp/Vs, suggesting changes in pore fluid pressure or rock properties5,6. Another notable feature, particularly in fast convergence zones, is the occurrence of deep aftershocks (2010 Maule earthquakes7) and continuous deep interface seismicity (2014 Iquique earthquake8) in the transition zone, well below the locked seismogenic zone. This deep seismicity—localized within the plate interface zone—may be linked to the seismogenic zone or be separated by a “soft barrier” from the latter3,9. Stress drops in this region are around 1 MPa10, lower than those typically observed in shallower aftershocks along the seismogenic plate interface11,12. In contrast, the overall stress in the transition zone is higher by an order of magnitude (20–40 MPa), as recovered from lithosphere-scale rheological and force analysis3,9,13, bridging the downdip end of seismic coupling with the peak stresses to the much lower stresses in the continuously creeping zone below 60 km depth.

While these instrumental observations provide valuable insights into certain properties of the deep transition zone, the physical mechanisms governing the plate interface in this region are not as well understood or analyzed as their equivalent in the shallower depth range of the seismogenic zone14,15. A range of mechanisms has been invoked: the importance of pressure solution creep, as a form of viscous deformation at low temperatures (<350°C), has been recognized for decades16,17,18. On the other hand, recent studies have brought to light the significant potential of dislocation creep in accommodating the transient oscillations of higher-temperature SSEs occurring at the base of the seismogenic zone, without switching to seismic slip19,20,21. Furthermore, the widespread presence of free fluids and highlighted cyclic pore fluid pressure swings in the transition zones22,23 play an important role in the transition of aseismic-seismic behavior and the occurrence of tremor signals24,25. Based on the lithological heterogeneity of the subduction interface, numerical studies highlight stress accumulation formed by rigid block interaction hosted by the weak matrix, which could strongly alter the rheological properties of the interface26,27,28. Therefore, identifying the transient phenomena recorded in exhumed rocks from the transition zone of the subduction interface—where field and fabric observations are far more limited compared to the seismogenic zone—and determining their associated creep mechanisms are essential for advancing our understanding of the rheology of the transition zone.

Rheological constraints on exhumed fossil subduction channels from this depth domain can be derived from several observations, including the hydraulic state and its variations22,29, deformation fabrics diagnostic of the underlying physical deformation mechanisms17,18, the range of lithologies involved20, and the role of recrystallized grain size paleopiezometry30,31,32,33. However, the heterogeneity of subduction mélange rocks poses challenges for the accurate application of paleopiezometry developed for monomineralic rocks34. Metachert blocks locally preserved in paleosubduction channels, characterized by high quartz purity comparable to quartzite, could represent an ideal target for applying quartz paleopiezometry at outcrop scales in order to estimate shear stress and strain rate distribution across the block-in-matrix setting. Additionally, local stress heterogeneity at this temperature range26,35 limits the extrapolation of small-scale paleopiezometric data to the entire subduction channel.

Here, we investigate the North Qilian high-pressure/low-temperature (HP/LT) mélange (NE Tibet) that represents the depth range of the former transition zone along the plate interface during Early Paleozoic fast subduction of Proto-Tethyan Ocean. In this mélange, metachert blocks are embedded in a calcschist matrix and contain foliation-parallel ultramylonitic micro-shear zones, which are viewed as recording dislocation-based creep below the base of the seismogenic zone and associated with anomalously high stress and strain rates. This finding provides key insight into coupling heterogeneities in the deep SSEs region as well as transient block interactions along the subduction interface.

Results

Geological and structural setting

The WNW–ESE-striking North Qilian HP/LT mélange belt (NQMB) represents the exhumed Early Paleozoic paleosubduction interface and extends over 50 km along the northeastern margin of the Qinghai-Tibet Plateau (Fig. 1a). It resulted from progressive northward subduction of the Paleo-Qilian Ocean (a branch of the Proto-Tethyan Ocean) beneath the Alxa block, along with fast plate convergence velocities during its closure, demonstrated by the subduction erosion-controlled active continental margin36. The NQMB, exposed as a tectonic slice between shallower greenschist facies slivers, comprises two distinct geological units, representing segments of different depths along the subduction interface (Fig. 1a, b): the eclogite facies unit and the overlying blueschist facies unit (Fig. 1c). The blueschist facies unit represents a coherent tectonic slice composed mainly of foliated glaucophane-bearing micaschists, calcschists, metacherts, and minor epidote blueschists. Previous data, including Lu-Hf isochron ages, suggest that the blueschist facies unit underwent solely blueschist facies metamorphism (450–550 °C and 1.2–1.6 GPa) at 467–462 Ma37, with subsequent cooling ages spanning 460 to 440 Ma38.

Fig. 1: Geological background and outcrop of the studied metachert block in Qingshui Gorge mélange.
figure 1

a Geological map of the HT-LP mélanges in the Qilian orogenic belt; tectonic location in inset map (upper right). Mylonitic foliation and stretching lineation from the A–B profile shown in a stereographic projection (lower hemisphere) along the left edge. Note that only observed metachert blocks are shown on the map; other blocks exist, but their boundaries remain uncertain due to limited accessibility. b Observed profile (A-B) across the HT-LP mélanges along the Qingshui Gorge. c Satellite image of the studied block-in-matrix structure, illustrating the boundaries between different units and the locations of the analyzed samples. d Mylonitic foliations and stretching lineations in the micro-sheared metachert. e Foliated metachert forming a thrust-related fold, with the position of (f) highlighted within the rectangle. f The graphite-bearing metachert is interlayered within the coarse-grained metachert. g Large-scale exposure of micro-sheared metachert. h Penetrative foliation of the micro-sheared metachert, as shown in the box of part (g). i At the margin of the studied metachert block, the calcschist is in contact with the metachert. Foliation-parallel quartz veins within the calcschist are boudinaged along the foliation.

Full size image

Along the Qingshui Gorge profile, we focus on a ~ 500-meter-wide metachert block within the host calcschist matrix of the blueschist facies unit. It can be divided into two structurally distinct portions: a micro-sheared metachert (MSMC) and a coarse-grained metachert (CMC), based on integrated fine-scale field mapping, systematic sampling at 50 m intervals, and oriented thin-section observations (Fig. 1b, c; Supplementary Table S1). Pronounced ductile shearing, localized in the MSMC portion with an outcrop thickness of ~300 m, is indicated by outcrop-scale mylonitic foliation, stretching lineation (Fig. 1d, g, h), and pervasively developed foliation-parallel micro-shear zones (Supplementary Table S1). The portion of CMC ( ~ 200 m thick) is characterized by top-to-the-SSW thrust-related folds, enclosed along the foliation within the MSMC portion (Fig. 1e), with no evidence of micro-shearing observed (Supplementary Table S1). The overall fabric orientations of the block and matrix are consistent, with the foliation dipping moderately to the NNE (14/47 N) and the gently east-northeast–plunging stretching lineation defined by elongated quartz (Fig. 1a), which also corresponds well to the top-to-the-south shearing sense observed in the blueschist unit (Supplementary Fig. S1). At the flank of the thrust fold, the graphite-bearing metachert foliation is embedded and lies parallel to the coarse-grained metachert (Fig. 1f). Furthermore, quartz-filled sheared veins within the centimeter-layered blueschist facies calcschist are ductilely elongated, exhibiting boudinage and/or pinch-and-swell geometry along the foliation in the northern margin of the metachert block (Fig. 1i). The development of brittle fault/ fracturing zones is partially preserved within the calcschist matrix (Supplementary Fig. S1). Extensive field observations indicate that metachert blocks contain far fewer veins compared to the surrounding calcschist matrix (Fig. 1d–h; Supplementary Fig. S1). Pseudotachylyte has been observed in neither the matrix nor the metachert blocks.

Deformation of metachert block versus calcschist matrix

Electron backscatter diffraction (EBSD) and microstructural analysis reveal and quantify the deformation styles of quartz in the calcschist matrix and the metachert block. In the calcschist matrix, quartz sheared veins range in thickness from a few millimeters to a centimeter and host large quartz grains along with dispersed phengite platelets, which are loosely aligned with the main foliation (Fig. 2a, b). The surrounding matrix consists of near-equigranular calcite–quartz aggregates, where quartz grains are commonly in contact and form a load-bearing framework (Supplementary Fig. S2). Recrystallized quartz grains appear weakly lobate along the edges of relict quartz grains with undulose extinction (Fig. 2c). Viscous foliations in CMC are defined by coarse-grained quartz ribbons interspersed with sparsely distributed, recrystallized quartz grains (Fig. 2e–g), along with 1 vol.% elongated mica identified through rock surface scans (Fig. 2e; Supplementary Fig. S3). The microstructures of the oriented MSMC samples are marked by penetrative foliation-parallel micro-shear zones (2–6 vol.%), ranging from 50 to 400 μm in thickness (Fig. 3; Supplementary Figs. S4–5 and Table S1). These micro-shear zones commonly cross-cut coarse-grained ribbons composed of undulatory quartz, fragmented glaucophane, and elongated phengite (Fig. 3a–c). In addition, the boundaries between the quartz ribbons and micro-shear zones are commonly decorated by aligned mica aggregates (Fig. 3b, c), as further corroborated by TIMA analysis showing that the MSMC sample (E-52) contains an appreciable amount of mica (8 vol.%, Supplementary Fig. S6). Within the micro-shear zones, quartz grains appear as fine, equant, strain-free recrystallized grains mantling larger, undulatory relict grains with higher internal lattice distortions (Fig. 3d, e, Supplementary Figs. S4–5).

Fig. 2: Oriented photomicrograph and electron backscatter diffraction (EBSD) analysis of quartz shear veins in calcschist matrix and coarse-grained ribbons of metachert block.
figure 2

a Elongated phengites (Phe) in the quartz shear vein, which have undergone dislocation creep, are aligned with the foliation of the calcite (Cal), epidote (Ep), and pelitic minerals; crossed polarized light. b Grain orientation spread (GOS) map of quartz shear vein of the box in part (a). c The quartz shear vein, deformed by typical subgrain rotation recrystallization, exhibits a foliation consistent with that of the calcschist; crossed polarized light. d Quartz <C> axis pole figures (equal-area, lower-hemisphere), misorientation axis distributions (2–10°) for various grain types, crystallographic vorticity axis (CVA) analysis, and misorientation angle distributions (MAD) are presented for quartz shear veins within calcschist matrix (B61-A; location marked by the rectangle in part a). e Coarse-grained quartz ribbons of metachert block formed by elongated phengite and quartz deformed by subgrain rotation recrystallization; crossed polarized light. f GOS map of coarse-grained ribbons in the box of part (e). g Coarse-grained ribbons in micro-sheared metachert (rectangle), where the foliation is defined by elongated phengite aggregation; crossed polarized light. h EBSD analysis of the “coarse-grained ribbon” region of site C43-A (rectangle in part e). i, j Grain size distributions for relict (red column), recrystallized grains (blue column), and subgrains (green column) of quartz shear veins of calcschist matrix (i) and coarse-grained ribbons of metachert block (j). drex and dsubg represent the root mean square sizes of recrystallized grains and subgrains, respectively. Legend: oriented kinematic framework, representative colors for distinct grain boundaries and grain types, multiples of uniform distribution (M.U.D) of pole figures and inverse pole figures of misorientation axes, and the spatial relationship between the subhorizontal CVA (the red line terminated with black triangles) and the geometry of the North Qilian paleosubduction channel.

Full size image
Fig. 3: Oriented photomicrograph and electron backscatter diffraction (EBSD) analysis of the micro-shear zone within the micro-sheared metachert.
figure 3

a The alternation of coarse quartz ribbons and micro-shear zones, along with pronounced foliation consisting of elongated phengite (Phg) and glaucophane (Gln). White arrowheads indicate the microboudinage of Gln; crossed polarized light. b Typical subgrain rotation recrystallization (SGR) of quartz (Qz) occurs within the micro-shear zone, which is defined by elongated phengite aggregation (E52 site A); crossed polarized light. c Coarse quartz ribbons intersected by foliation-parallel micro-shear zones. Microboudinage of Gln is shown by white arrowheads; crossed polarized light. d Misorientation to mean grain orientation (Mis2mean) maps showing the intragranular misorientation patterns of quartz. e Misorientation traverses along the A–A’ and B–B’ profiles (locations marked in part d). The A-A’ misorientation profile angle across the subgrain boundaries in larger relict grains is associated with progressive jumps below 6°, while the abrupt alternation of misorientation angle from grain boundaries between small recrystallized grains ranges from 20–60° along the B-B’ profile. f, g Quartz c-axis pole figures, misorientation axis distributions (2–10°), crystallographic vorticity axis (CVA) analysis, and misorientation angle distributions (MAD) are shown for the regions of the micro-shear zone and coarse quartz ribbons in (e). h Grain size distributions of micro-shear zones from micro-sheared metacherts across the studied metachert block. See Fig. 2 for the legend.

Full size image

The analyzed sites, including quartz sheared veins, micro-shear zones, and coarse-grained ribbons, demonstrate consistent patterns in their crystallographic preferred orientations (CPO), misorientation angle distributions (MAD), and crystallographic vorticity axis (CVA) (Figs. 2–3; Supplementary Figs. S3–5). Additional details are available in the Supporting Information. According to the GOS-based separation method39, the root mean square (RMS) grain sizes of recrystallized quartz and subgrain populations were determined for both the matrix and the block: quartz sheared veins (100–112 μm and 98–105 μm; Fig. 2d), coarse-grained ribbons (16–34 μm and 17–29 μm; Fig. 2h), and micro-shear zones (7–10 μm and 6–12 μm; Fig. 3h).

Assessment of P-T Conditions for micro-shearing

Graphite occurs as flattened flakes and/or dispersed aggregates (~50–400 μm in size), aligned with the foliation, as identified by Raman spectroscopy in the coarse-grain ribbons of the CMC (G, D1, and D2 bands; Supplementary Table S2). The Raman Spectra of Carbonaceous Material (RSCM) geothermometry40 yielded peak temperatures at the analyzed sites (including sample C-61; Fig. 1e, f) ranging from 446 to 525 °C, with a mean temperature (RMS) of 480 °C and a standard deviation of 22 °C (Supplementary Table S2).

Electron probe micro-analyses of white mica from block and matrix samples reveal homogeneous chemical compositions, with no significant variations among micro-shear zones, coarse-grained ribbons, and quartz sheared veins (see detailed mineral chemistry in Supplementary information). White micas are phengites (solid solution between celadonite and muscovite) with silica per formula unit (p.f.u) values ranging from 3.42 to 3.46 in MSMC sample (E-52) (Fig. 4a, Supplementary Data 1). Phase equilibrium modeling using the Perple_X software41 for sample E-52 was conducted based on synkinematic petrological relationships and bulk-rock compositions calculated from TIMA estimates (Supplementary Fig. S6). By combining RSCM geothermometry, the RMS silica p.f.u. value in phengite, and the mineral assemblage of sample E52 observed through TIMA scanning, the average equilibrium P-T conditions of micro-shearing were constrained to 490 ± 10 °C and 15.0 ± 1 kbar (Fig. 4b). These P-T conditions are regarded as the synkinematic peak conditions during micro-shearing, supported by the absence of microstructures indicative of retrograde metamorphic reactions (e.g., consistent CPO of quartz, uniform silica content across phengite, and flat profiles of sodium content and XMg value across glaucophane; Supplementary Fig. S7 and Supplementary Data 1–2). Additionally, phengite 40Ar–39Ar dating on calcschist matrix yields two uniform plateau ages of 467 ± 3 Ma (Supplementary Fig. S8 and Table S5).

Fig. 4: White mica compositions from Qingshui Gorge mélange and P–T pseudosection of micro-sheared metachert.
figure 4

a Ternary diagram showing white mica compositions in micro-shear zones, coarse ribbons of metachert, and shear veins of calcschist. Cel. – Celadonite; Mus. – Muscovite; Prl. – Pyrophyllite. b P–T pseudosection calculated for the micro-sheared metachert (sample E-52), highlighting the best-fit area where foliation-defining phengite formed (white area outlined by black line), determined using the RSCM method (gray shaded rectangle), isopleths of Si-in-phengite (Si p.f.u), and analogous TIMA scanning-based silicate mineral phases and their normalized volume percentages shown in the pie chart. The black dotted arrow indicates the hypothetical clockwise P–T path after Cheng et al.37.

Full size image

Discussion

In the Qingshui Gorge mélange, the mechanisms and kinematic framework of viscous creep events are quantitatively extracted from the obtained quartz fabrics. No significant divergence in the active slip systems of various quartz grain populations was observed among the calcschist matrix, CMC, and MSMC (Figs. 2–3, Supplementary Figs. S3–5), all of which coincided with the shift from the rhomb <a> slip to the prism <a> slip42,43. This, combined with the strong CPO and intragranular distortion patterns of quartz aggregates, suggests that dislocation creep driven by subgrain rotation recrystallization (SGR) dominated deformation in both the calcschist matrix and the metachert block. The similar grain sizes and CPO patterns between recrystallized quartz grains and their subgrains imply negligible overprinting on these textures during late-stage exhumation. Furthermore, the near-horizontal vorticity axis and top-to-the-south shear sense, inferred from the CVA y-axis maximum and the counterclockwise rotation of asymmetric c-axis girdles, correspond well to the regional subduction kinematic framework (Fig. 2 Legend).

Outcrop-scale quartz piezometry conducted on the studied metachert block and quartz-rich sheared veins within the calcschist matrix provides insight for evaluating stress distribution during peak burial metamorphism. Based on the “sliding resolution” piezometer of Cross et al.39 and measured RMS recrystallized quartz grain sizes, shear stress estimates range from 23 to 25 MPa for quartz sheared veins, 57 to 75 MPa for coarse-grained ribbons, and 126 to 143 MPa for micro-shear zones (Supplementary Table S4). The shear stress values of quartz sheared veins are considered the upper bound for the strength of calcschist matrix, as the boudinage and/or pinch-and-swell geometry of the quartz veins suggests they are more competent than the host calcschist44. The quartz load-bearing framework observed throughout the calcschist thin-sections (Supplementary Figs. S2 and S7) suggests that the rock’s strength is mainly governed by quartz, thereby implying a relatively homogeneous strain distribution at the whole-rock scale45. Although these shear stress estimates, derived from micro-shear zones, exceed previously reported paleopiezometry-based values from exhumed subduction rocks18,30,32,33,46,47 ( ~ 10–100 MPa or less), they remain valid due to the observed dominance of dislocation creep. The presence of intragranular Dauphiné twins, which do not cross grain boundaries (yellow twin boundaries in Figs. 2–3, Supplementary Figs. S3–5), indicates differential stresses of ~50–145 MPa at ~500 °C, based on the study of Minor et al.48.

According to estimated deformation P-T conditions (490 °C and 15 kbar), the spread in flow stress translates to an approximately 600-fold difference in strain rate. Quartz sheared veins from the host calcschist exhibit rates on the order of 10−12 s−1, while estimates for block-hosted micro-shear zones locally exceed 10−9 s−1 (Fig. 5a, Table S4). The rheological results obtained through the application of various piezometers and flow laws are broadly consistent, as illustrated in Supplementary Fig. S9. On the other hand, given the closure temperatures of muscovite49 (425 °C ± 70 °C) being slightly lower than the constrained deformation temperatures, the 40Ar–39Ar plateau ages of 467 Ma are thought to represent the initial cooling age of the dislocation creep event. Our estimated P-T conditions and timing are both compatible with the reported blueschist facies metamorphism in the NQMB37. Consequently, the studied calcschist matrix and the adjacent metachert block are regarded as part of the North Qilian paleosubduction channel and are here interpreted to have coevally recorded a single deformation event in the Lower Ordovician. Under the lithostatic gradient of the overriding continental plate, the viscous deformation is inferred to have occurred at a depth of 50-60 km in this paleo-subduction channel, which falls at the base of the modern deep SSEs zone, typically situated between 40 and 60 km1.

Fig. 5: Deformation mechanisms and rheological properties of metachert block and calcschist matrix.
figure 5

a Deformation mechanism map under constrained P-T conditions (~1.5 GPa, 490 °C) of the metachert block. The red line represents the boundary between fields of pressure solution and dislocation creep, while the black dotted line represents the “sliding resolution” piezometer of Cross et al.39. Discussions and derivation details are provided in the main text and supplementary material, respectively. b Rheological diagram for flow stress versus depth in the quartz veins (matrix) and metachert (block). Byerlee’s Law is given for pore fluid factor values (λ) of 0.86 and 0.97. Strain rate (\(\dot{\varepsilon }\)) contours of 10-8–10-12 s-1 are calculated using quartz flow laws from Hirth et al.81. The frictional Byerlee’s Law83 is calculated using a friction coefficient of 0.7, with the geostatic gradient is 27.5 MPa/km. This diagram suggests that the block within the viscous deformation regime (λ < 0.86) is less capable of withstanding pore fluid pressure than the matrix with vein-openings (λ > 0.97).

Full size image

Among viscous creep mechanisms, pressure solution has been reported as active along the shallow subduction interface17,18, particularly favored in areas with small grain sizes50. The application of the dislocation creep-based flow law in this study may be influenced by the unaccounted role of pressure solution creep, which is likely leading to grain size re-equilibration and further misestimation. Even so, the deformation mechanism map (Fig. 5a) shows that our obtained values for quartz veins (matrix) and metachert (block) fall within the dislocation creep field (observed dominance of dislocation creep; Figs. 2–3, Supplementary Figs. S3–5), suggesting that pressure solution creep has little to no impact on our rheological estimates. Furthermore, recent numerical and microphysical models predict the prevalence of dislocation creep within the high-temperature SSEs zone, highlighting its potential to accommodate transient perturbations through accelerated creep without switching to seismic slip19,20,21. Here, the deformation style identified in the metachert block and calcschist matrix provides compelling geological evidence of dislocation-based creep within the SSEs zone, bridging the gap between model predictions and field observations.

Within the metachert block, a positive correlation between phyllosilicate content and creep rate exists, i.e., micro-shear zone formation and strain rate increase in the mica-rich MSMC (8 vol.%) while micro-shearing is absent in the mica-poor CMC (1 vol.%). The role of mica in facilitating strain localization is commonly demonstrated by deformation experiments51 and natural observations52. The addition of a mere 10 vol.% of mica content can significantly alter the rheological behavior of quartz, and increase the strain rate by an order of magnitude relative to pure quartzite. Furthermore, there is no evidence of pre-existing foliation-parallel brittle fractures within the metachert that could have been exploited by micro-shearing53. This textural variation is therefore interpreted as the product of strain localization driven by variations in phyllosilicate content.

Quartz micro-shearing, as a snapshot of transient deformation conditions, has been evidenced by both experimental results54 and geological observations31,55,56, including for other HP/LT paleo-accretionary systems worldwide57. The overprinting of coarse-grained ribbons by micro-shear zones, along with their consistent CPO, can therefore be interpreted as two stages of a deformation event: coarse-grained ribbons represent steady-state creep, which subsequently transitioned to transient higher stress and strain rate events that are recorded in micro-shear zones. We emphasize that the micro-sheared samples (collected across the ~300-meter-wide MSMC) contain penetrative, foliation-parallel micro-shear zones (~2–6 vol.%, based on optical scanning of oriented X-Z thin sections; Fig. 3; Supplementary Fig. S4–5 and Table S1). Rheological envelope calculations (Fig. 5b) show that the pore fluid pressure factor (λ) required for stable plastic behavior is lower within the micro-sheared metachert block (λ < 0.86) compared to the vein-bearing calcschist matrix (λ > 0.97), implying that micro-shearing is activated within the block under physical conditions (i.e., at lower fluid content and higher differential stresses) distinct from those in the matrix. This variability in pore fluid pressure (between the block and the matrix) is independently confirmed by (i) the greater outcrop-scale abundance of quartz veins in the matrix (Fig. 1d–i; Supplementary Fig. S1) and (ii) a higher amount of hydrated minerals (e.g., epidote and mica; Figs. 2–3; Supplementary Table S1) of the matrix. Elevated pore fluid pressure (λ > 0.97) within the calcschist matrix is regarded as a key factor promoting unstable slip and subsequent cataclasis. However, direct evidence of cataclasis is rarely preserved in nature, as it is prone to being overprinted by deformation-assisted recrystallization during multiple seismic cycles, ultimately erasing the original brittle markers14,58,59 (e.g., Fig. S1e, f). In this context, although the occurrence of former SSEs within the calcschist matrix cannot be directly demonstrated, they remain the most likely mechanism of stress release in this segment of the subduction interface—an area known to host such events2,3,4,5.

Deep SSEs occur beneath the seismogenic zone, manifesting as transiently elevated slip velocities that exceed the long-term plate motion rate with durations ranging from days to years60. Modern subduction interfaces are characterized by relatively low bulk strength level3,9,10,13 (1–40 MPa) in the transition zone compared to estimates from paleopiezometry18,30,32,33,46,47, as well as heterogeneous plate coupling along strike, influenced by variations in the effective strength of locked patches61,62. From our geological perspective, these discrepancies likely arise from sheared, heterogeneous lithologies at small-scale26,28, such as the presence of “strong” metachert/basaltic blocks within a “weak” metasedimentary matrix20,63,64. Here, the power-law equation for dislocation creep indicates that shear stress and strain rates in the stronger, metachert block are significantly higher than those in the calcschist matrix, unexpectedly underscoring their contrasting rheological behavior. The presence of various metachert blocks exposed within the studied blueschist facies mélange (Fig. 1a–c) suggests that these relatively strong blocks could have mechanically interacted together during deep-seated shearing along the subduction interface, leading to the amplification of localized stress19,28. This transient configuration is closely linked to their local interconnected geometric arrangement, which creates a force-chain-like phenomenon, as previously suggested by numerical studies19,27. Meanwhile, the integration of stress over the block surface28,35 and/or the indentation of relatively strong blocks within a weaker matrix could lead to higher internal stresses65 within blocks than in the matrix, possibly also playing a role in strengthening this phenomenon. Alternatively, the strength-inversion model66 suggests that comparably high stress concentration and associated brittle failure may also arise from a potentially local configuration, in which mechanically weak blocks are embedded within a stronger, viscous matrix—without requiring direct mechanical coupling between blocks.

Numerical modeling studies19,28 have demonstrated that a critical threshold of approximately 50–55% block content is required to maintain Voigt-type iso-strain-rate behavior, which in turn leads to the mechanical protection of the matrix through a well-connected network of strong blocks. This highlights the broader role of lithological heterogeneity, as also evidenced by geophysical observations67 showing that subduction zone mechanics are strongly influenced by the structural complexity of the downgoing plate (e.g., asperities, seamounts, fracture zones, mélanges, and olistostromes). At the North Qilian paleo-subduction interface, our field observations (Fig. 1a) indicate that the block content in the mélange is heterogeneous and does not systematically reach the critical threshold (~50%), implying that such localized force-chain phenomena65 could not have persisted over prolonged durations. Instead, transient and localized stress transfer along force chains may have occurred only episodically, during temporary enrichment of the downgoing crust in strong blocks, as illustrated in the inset sketch of Fig. 6. While during the rest of the time, the block content was too low to generate interaction or exert a first-order control on matrix rheology, as supported by previous modeling studies19,28. Alternatively, it cannot be excluded that other events, such as asperity subduction62, could also have occasionally contributed to the stress transfer recorded in the metachert blocks. In any case, our observations suggest that stress increase associated with the activation of micro-shear zones was not sustained over geological timescales of millions of years but rather occurred over a shorter duration (i.e., within a timeframe of ~5 to 40 years; see calculations in Method), likely on the long-term SSEs timescales68,69 (5–15 years). Shear strain rates (~10−8 s−1–10−4 s−1; see Method)14 of SSEs below the seismogenic zone can be estimated from geodetic slip rate70 (300–1500 mm/yr) and defined slip thickness of SSEs (10-4–50 m)71, with the lower limit almost reaching the constrained strain rate of micro-shearing in the blocks (10−9 s−1). Linking the exhumed geological evidence to present-day geophysical observations of SSEs is intrinsically hampered by (i) a pronounced mismatch in spatial and temporal scales; (ii) the scarcity of unequivocal rock fabrics and/or structures attributable to single events; and (iii) the prolonged history of subduction, accretion, and exhumation, which complicates the interpretation of the rock record14,72. Nevertheless, constrained analogies with modern subduction channels remain an indispensable—though indirect—framework for deciphering how transient SSEs may be archived in the deep geological record20. Here, extrapolating from observations in modern subduction analogs, such a slip transient would be spatially embraced by the large-scale (several to hundreds of kilometers), multi-fault network59,71,73,74 of the weaker matrix (Supplementary Fig. S1), which is expected to host the SSEs. The aforementioned associations in space and strain rate imply that the stress perturbation observed within the block may be related to, or even triggered by, large-scale SSEs occurring in the surrounding faulted matrix network. Notably, large megathrust ruptures are also known to be followed by viscoelastic relaxation along the subduction interface that can persist for several decades (such as for the 1960 Great Chile Earthquake75), in line with the duration estimated from observed microstructures. Such post-seismic stress relaxation processes could also have contributed to further block interaction and the subsequent generation of the micro-shear zones observed within the metachert blocks.

Fig. 6: Sketch illustrating the strong coupling model of the studied metachert block along the North Qilian paleo-subduction channel at a depth of approximately 55 km.
figure 6

The high differential stress recorded in the metachert block, with a lower pore fluid factor (λ) compared to the calcschist matrix, likely results from transient block interactions driven by the stress amplification effect along the subduction channel. Such a phenomenon led to the widespread formation of foliation-parallel micro-shear zones at strain rates of 10−9 s−1 (as shown in the inset microstructure). These micro-shear zones preserved in the block provide direct geological evidence of the transient dislocation-based creep, interpreted here as related or triggered by the surrounding large-scale multi-fault network associated with the SSEs, and represented by the plateau-like peak of the red strain rate curve. Note that the inset sketch depicts the geometric configuration where block-block interaction occurred. Such interaction, which requires a high abundance of blocks within the mélange, is proposed to have occurred only transiently during the lifespan of subduction channel activity.

Full size image

In summary (Fig. 6), the studied micro-shear zones formed during transient block interactions in a paleo-subduction channel exhumed from below the downdip end of the seismogenic zone. Block interaction processes aseismically generated dislocation-based creep while recording anomalously high stress and strain rates within the block compared to the adjacent, more fluid-rich, weaker calcschist matrix. These findings offer unprecedented high-resolution insights into the mechanical feedback between hydraulic state, stress heterogeneities, and mixed lithologies along the deep SSEs region in the transition zone of the subduction interface.

Methods

Electron backscatter diffraction (EBSD)

Thin sections (X–Z plane) were prepared by polishing with the Buehler MasterPrep 0.05-μm alumina polishing suspension and the Buehler EcoMet300-AutoMet300 grinder–polisher, followed by coating with 7.5 nm layers of carbon to reduce charging. Quartz crystallographic orientation mapping was also performed using the FEI Quanta 450 FE-SEM equipped with an Oxford Symmetry S3 EBSD detector in high vacuum mode at the Institute of Geology, Chinese Academy of Geological Sciences (CAGS), Beijing. The orientation of the thin sections in the analysis chamber was set with the stretching lineation parallel to the SEM X-axis and a tilt angle of 70°. Operating conditions for electron backscatter analysis were: accelerating voltage of 20 kV, beam current of 110–120 μA, step size of 1.5 μm for quartz sheared veins, 0.75–1.0 μm for coarse ribbons, and 0.18–0.9 μm for micro-shear zones. The image collection was completed using the AztecCrystal software, and data were processed with the open-source toolbox76 for MATLAB (https://mtex-toolbox.github.io/; v. 5.8.1). Grains were defined as areas of a single phase with a misorientation of less than 10°, and the crystallographic vorticity axis (CVA) patterns were determined after77.

TESCAN integrated mineral analyzer (TIMA)

Mineral identification and mica content measurement of coarse-grained metachert (CMC) and micro-sheared metachert (MSMC) thin sections were conducted using the TIMA mineralogy system at the Institute of Geology, CAGS, which comprises a Tescan Mira Schottky field emission scanning electron microscope equipped with four silicon drift energy dispersive (EDS) detectors arranged at 90° intervals around the chamber. The analyses were performed with an accelerating voltage of 25 kV, a beam current of 8 nA, a working distance of 15 mm, and a spot size of 90 nm. The classification of mineral phases and the calculation of their volume ratios were performed by matching the obtained spectroscopic data with built-in mineral definition files.

Electron probe micro-analyzer (EPMA)

Mineral chemical compositions of the calcschist matrix and the metachert block were acquired on the JEOL JXA-iHP200F field emission electron probe micro-analyzer at the Institute of Geology, CAGS. An accelerating voltage of 15 kV, a probe current of 20 nA, and a beam diameter of 5 μm were used for acquisitions. Ten oxide compositions were measured with the following standards: SiO2, TiO2, Al2O3, Cr2O3, FeO, MnO, MgO, CaO, Na2O, and K2O. The lines and spectrometer crystals used for the major elements are: Si (Kα, TAP), Ti (Kα, PETJ), Al (Kα, TAP), Cr (Kα, LIFL), Fe (Kα, LIFH), Mn (Kα, LIFH), Mg (Kα, TAPL), Ca (Kα, PETJ), Na (Kα, TAPL), and K (Kα, PETJ). Count times were set to 10 s for the peak and 5 s for the background for each element. Utilizing natural minerals for standardization and performing ZAF corrections yielded an estimated precision of ±2% for major elements.

Raman spectroscopy

Raman spectra of graphite on CMC thin sections were obtained using a Horiba LabRAM HR Evolution spectrometer, coupled with an Olympus BX41 microscope at the Institute of Geology, CAGS. A 532 nm air-cooled frequency-doubled Nd:YAG laser was used, delivering 20 mW on the sample with a ×50 objective lens. The scattered light was received by an 1800 gr/mm optical grating through a confocal 100 μm pinhole. After calibration with a silicon standard, we use an acquisition time of 10 s and 2 scans were accumulated within the Raman shift range of 1000 cm-1 to 3500 cm-1. The raw spectral data were fitted using the Voigt function in Fityk 1.3.1 software78. Three peaks, corresponding to the G (1580 cm⁻¹), D1 (1350 cm⁻¹), and D2 (1620 cm⁻¹) bands were identified. Based on the determined peak parameters (Table S2), the quantitative parameter of the graphitization degree (R2 ratio) and related temperature are estimated following40.

Phase equilibria modeling

The major element composition of the MSMC sample E52 was calculated based on the volume fractions of classified mineral phases and their corresponding chemical formulas, which in weight percent is: SiO2 (90.48), Al2O3 (4.13), FeO (1.56), MnO (0.07), MgO (1.31), CaO (0.05), and Na2O (0.69). The phase equilibria modeling of sample E52 was conducted using Perple_X version 6.8.141 and the internal thermodynamic database79 in the MnO-Na2O-CaO-K2O-FeO-MgO-Al2O3-SiO2-H2O (MnNCKFMASH) system. Ferric iron has been neglected in sake of simplicity because analyzed minerals exhibit low ferric iron contents. The a–x relations are provided in the Supporting Information. Mineral abbreviations are given following80. The isopleths of silica content per formula unit in phengite are also constrained in the pseudosection. The comparison between EPMA analyses from the MSMC sample (E52-B site) and modeled mineral compositions (for the best-fit area shown in Fig. 4b), presented in Table S3, reveals minor yet acceptable deviations.

Flow law and slip rate estimation

The EBSD-derived RMS recrystallized grain sizes of quartz (\({d}_{{{\rm{RMS}}}}\)) were converted into differential stress (\(\sigma\)) using the “sliding resolution” piezometer39, expressed as: \({d}_{{{\rm{RMS}}}}={10}^{4.22}\times {\sigma }^{-1.59}\). The estimated differential stress values were multiplied by \(2/\surd 3\) to obtain the equivalent shear stress under plane stress conditions46, which was confirmed by the finite strain estimation (Supplementary Fig. S11). The “dislocation creep” flow law from Hirth et al. 81 and the derivation details of the deformation mechanism map are included in the Supporting Information.

Assuming the North Qilian paleo-subduction channel underwent down-dip simple shear deformation, the shear strain rate (\(\dot{\gamma }\)) can be expressed in terms of the shear strain (\(\gamma\)) accumulated across an idealized shear width (\(w\)), resulting from a finite slip displacement (\(d\)) parallel to the subduction interface over a given shearing duration (\(t\)). Thus, shear strain (\(\gamma\)) is defined as \(\gamma={{\rm{d}}}/{{\rm{w}}}\) and the shear strain rate (\(\dot{\gamma }\)) is given by \(\dot{\gamma }=\gamma /{{\rm{t}}}\) 82. Strain estimates for the micro-sheared metachert range from 0.71 to 2.69, as constrained by syn-kinematic microstructures shown in Supplementary Fig. S10 and the calculation method detailed in the Supporting Information. With constrained strain rates of 10−9 s−1, the process of micro-shearing is estimated to last 5–40 years. Using the relationships described above, the shear strain rate (\(\dot{\gamma }\)) along the subduction interface can be expressed in terms of slip rate (\({V}_{S}={{\rm{d}}}/{{\rm{t}}}\)) and idealized shear width (\(w\)) as: \(\dot{\gamma }=\,{V}_{S}/{{\rm{w}}}\).

40Ar–39Ar dating

Syn-kinematic phengite grains were manually selected under a binocular microscope from two crushed blueschist facies calcschist samples (B61 and B62). The phengite grains were ultrasonically cleaned with deionized water and methanol, weighed, placed into aluminum packets, and loaded into quartz vials for irradiation. Irradiation was conducted for 24 hours in the B4 position of the nuclear reactor at the Chinese Institute of Atomic Energy, Beijing, with a neutron flux density of ~2.65 × 1013 n cm−2 s−1. After irradiation, 40Ar–39Ar standard step analyses were performed using a Helix MC mass spectrometer at the Institute of Geology, CAGS. Correction factors for interfering isotopes were calculated during irradiation using irradiated K2SO4 and CaF2. Their values were: (36Ar/37Ar)Ca = 0.0002398, (40Ar/39Ar)K = 0.004782, and (39Ar/37Ar)Ca = 0.000806. All isotopic data are provided in Table S5.