Introduction

Gaining insight into the seismic and volcanic eruptions that occur in arc formations necessitates a comprehensive knowledge of the mechanisms responsible for the movement of material between the mantle wedge and subducted slab in subduction zones. In particular, the transfer of volatile from Earth’s surface to the deep mantle are largely controlled by those processes with vital implications for plate tectonics1, long-term global climate2, and the evolution of Earth’s heat budget3. Nonetheless, the transport mechanisms in which slab material is moved into the mantle wedge is widely disputed4,5,6,7,8,9, lead to an unclear awareness of global volatile cycling. Besides, the mechanism of fluid flow propagation (i.e., pervasive versus channelized10,11,12) and the fact of Phanerozoic secular cooling of Earth mantle13 also hinder the efficiency of volatile mobilization beneath the subarc, further making the feeding of subducted volatile into the mantle wedge remains ambiguous14. Flawed conclusions regarding both the long-term evolution of the crust and mantle, as well as the short-term recycling of volatiles shall raise, if geodynamic models used to assess geochemical and geophysical data related to subduction zones are of questionable accuracy.

Direct geophysical observations of subarc processes, however, presently lack the resolution to obtain fine-scale information about the material transfer processes at the slab-mantle interface. The chemical and isotopic compositions of arc lavas offer chance to extract critical information from the key regime of subduction zones (e.g., the subarc).

Research conducted over the past decade has demonstrated that a combination of partial melting of subducted sediments and metamorphic dehydration of altered oceanic crust (AOC) is capable of producing the required trace element fractionations (e.g., Nb-Ta negative anomaly, LILEs enrichment, and HFSEs depletion)7,15,16. Consequently, a metasomatized mantle model (e.g., refs. 17,18) is applied by most studies of subduction zone processes as the primary vector to explain material transfer between the descent slab and overlaying mantle wedge, with a common trait that the diagnostic trace element fractionation of arc lavas is produced in the slab before mixing with the mantle wedge takes place5,19. However challenges have raised, from experimental petrology, in simultaneously reproducing both major and trace element features of the most common types of arc magmas (i.e., tholeiites and calc-alkaline melts)20. Furthermore, argument has been recently proposed, for the debate of the processes by which typical trace element signatures are bred to arc magmatism, that the trace element and isotope variability of the magmas of global arc could not be simply settled with the classic model which introduces a hybridized mantle wedge by discrete feedings of fluids and/or aqueous melts from slab components9,19. Instead, reconciliation can only be achieved if physical mixing of depleted mantle (DM), AOC, and sediments arises early within the plate interface during subduction prior to any melting takes place5. The order of events is thus redefined, as a prerequisite, to back up the fundamental role of mélange for the genesis of arc magmatism. Laboratory21 and numerical modelings22,23,24,25 of subduction process have revealed that gravitational instabilities could be produced by hydration and partial melting at the slab-mantle interface, with further development into diapiric structures composed of partially molten materials20. Yet these diapirs have not been unambiguously imaged and accepted in global subduction zone architecture, probably due to the following reasons: (1) Along arc geophysical studies are rare, and the current resolution of seismic techniques may not be appropriate to image mixed mélange-peridotite lithologies; (2) Magnetotelluric approach, sensitive to interconnected free fluids, would not easily detect the presence of mélanges, where most of the water may be crystallographically bounded. It is also not able to detect the difference between fluids generated by a dehydrating mélange diapir and those released at the slab-top, which percolate the mantle wedge; (3) Several km-thick low-seismic velocity regions observed at the slab-top in subduction zones worldwide indicate the persistence of mélange zones (mainly hydrated rocks and sediments) in the slab-mantle interface26,27,28. Field evidence reveals multiple mélange outcrops worldwide (see refs. 5,29 for comprehensive review) but voluminous deeply burial and exhumation of sediments is of scarcity (10–15 km, Apennines, Italy30; 50–60 km, Raspas complex, Ecuador31), especially for those exhumed from subarc depths (i.e., ~85–140 km, >2.6 GPa12). This, in turns, obstructs direct evaluation of the degree of chemical and/or mechanical disruption of subducted material along the plate interface.

Here we present a comprehensive study on eclogites with distinct pressure-temperature-protolith histories from a deeply buried mélange “package” in the Atbashi low-temperature (LT)-high-pressure (HP) metamorphic complex, Kyrgyzstan section of the South Tianshan Metamorphic Complex (STMC), southern Altaids (Fig. 1). Recent studies32,33 in the Chinese section of the STMC disclose massive sediment accretion at ~80 km depth along the subduction interface, suggesting continuous refrigeration, by incoming cold material from the slab, and juxtaposition to the “cold nose” of mantle wedge. In addition, transient thermal excursion was revealed, in region, from strikingly concordant chemical zonation of garnet in coesite-bearing oceanic eclogites34, disclosing the potential translation of ultra-high-pressure rocks (UHP) refrigerated slices near to a relatively hotter mantle wedge. In this study, field mapping, bulk-rock geochemistry, metamorphic petrology, Zr-in-rutile & Ti-in-quartz thermobarometers, thermodynamic modeling, rutile & zircon trace elements, and U-Pb chronology analyses have been conducted to provide the first tangible eclogitic rock evidence recording mélange diapir melting signal (MDP) and experiencing substantial thermal excursion in a well-preserved refrigerated subduction plate interface, as confirmed by the pervasive presence of low-temperature eclogitic rocks (refs. 33,35,36,37,38,39,40 and also this study). Additional multi-disciplinary data, especially those Late Carboniferous ones, are also compiled from regional various lithologies to fingerprint the temporal-spatial variations of mantle signal and crustal feedback during which the eclogitic mélange rocks contemporaneously formed and their fate during substantial thermal excursion. Available data provide insights into a model of mélange diapir melting in refrigerated subduction plate interface as substantiated in the STMC. Implications for such process with a momentous contribution in transferring crustal volatile from slab surface to arc lava, regulating terrestrial geochemical cycle, are thus discussed.

Fig. 1: The Kembel mélange in the Atbashi range of the STOB.
figure 1

a Location of the research area and geological map with close-up view showing the regional geological setting and sampling location (modified after ref. 52). b, c Two cross-sections constructed, along the Kembel valley, across the Atbashi HP/UHP mélange with key structural elements and detailed sample sites.

Full size image

Results

The Kembel mélange in the Atbashi metamorphic complex

The construction of the Altaids involves the welding of the North China, Alxa, Karakum-Tarim, Siberia, and East European cratons (Fig. 1)41,42. The 2500 km-long South Tianshan Orogenic Belt (STOB, see Supplementary Note 1 for detailed geological background in Supplementary Information), is located in the southern Altaids, extending from the western deserts of Uzbekistan, Tajikistan, Kyrgyzstan, and Kazakhstan to northern Xinjiang in China, and formed in response to the final amalgamation between the Tarim and Siberian cratons43,44,45. The STOB is comprised of the Kazakhstan-Yili-Central Tianshan Continent (KYCTC) in the north and South Tianshan accretionary complex (AC, including the South Tianshan Belt, STB) in the south, and is bounded to the south by the North Tarim Fault (NTF) with the North Tarim Craton46,47. The Atbashi metamorphic complex (ATMC) represents the central part of the South Tianshan LT-(U)HP metamorphic complexes (STMC) within the southern Altaids and extends for about 120 km along the Atbashi-Inylchek-South Central Tianshan Fault and correlates with the Akeyazi33,37, Chatkal48,49 and Fan-Karategin50 (U)HP metamorphic complexes, which crop out in China, Kyrgyzstan and Tajikistan segments of the STOB, respectively. Numerous previous studies32,33,35,36,48,51 focused on the P-T-isotopic ages history of high-grade metamorphized lithologies had confirmed the existence of a refrigerated subduction plate interface during the recovery of the STMC (500–600 °C, 2.0–3.0 GPa, Fig. 1). Our studied area, named as the Kembel mélange51,52,53,54, lies close to the north boundary of the ATMC and is clamped in between the North and South regional greenschist-facies units. It was crosscut by the Kembel River as a ~1 km wide N-S across, deeply buried mélange “package” (Fig. 1b, c) predominantly composed of strongly schistose, high-grade metamorphosed, meta-volcanoclastics hosting eclogitie-/blueschist-facies mafic metavolcanics, as well as serpentinite, as pods, lenses, boudins, thin layers or massive blocks and marble horizons (Fig. 2).

Fig. 2: Outcrop features.
figure 2

a, b Outcrop observation of representative HT and LT eclogites wrapped within meta-sedimentary mélange. c Figure shows LT eclogitic pillows. d Representative outcrop of HT eclogite with contrasting bottle green color.

Full size image

Sampling strategy

To determine the petrogenesis, as well as the protolith and metamorphic ages, of eclogitic blocks (their contrasting features are described in the following context) wrapped within the Kembel mélange, twenty samples were collected, including four HT and eight LT eclogites (highlighted with red and green boxes/colors in Fig. 2 and following figures, respectively, Supplementary Data 1), several serpentinite blocks and meta-volcanoclastic host rocks were sampled (Fig. 2), on outcrops along the Kembel West and East cross-sections, for petrographic, mineralogical, and geochemical investigation. Of these, representative samples of HT and LT eclogites were selected for bulk-rock geochemistry (major and trace element and Sr-Nd isotopic composition, Supplementary Data 3, 4), rutile and quartz trace element composition (Supplementary Data 5, 6), and rutile, as well as zircon, U-Pb isotopic chronology analysis (Supplementary Data 7–9).

Petrography, petrology, and P-T estimates

Two types of eclogites (Supplementary Data 1) were recognized by color on a hand-sample scale: one has a bottle-green and the other a pale-green appearance (Figs. 2, 3, Supplementary Fig. 1). The bottle green eclogite (i.e., HT one, Fig. 3a, Supplementary Fig. 1a–c) is comprised of moderate size pink garnet porphyroblast (1–2 mm in diameter) and coarse-grained grass green clinopyroxene matrix (50–100 µm) painted with massive irregular patches of quartz (Fig. 3a–c, Supplementary Figs. 1a–c, 2) and opaque ilmenite (with few relic rutile) as well as minor dark-blue amphibole and phengite. Some pseudomorphs55, as “nano-diorite” inclusions with 100–200 µm size in garnet, after films of melt (Supplementary Fig. 2g, h) were identified with a fine-grained aggregate of quartz + clinopyroxene ± ilmenite ± epidote, yet no glassy was present. Orthopyroxene, despite predicted to be stable during decompression (Supplementary Fig. 8c), was not detected too, likely due to the potential limited inconsistency between the chosen bulk composition and artifacts of the solution models used to estimate mineral assemblages. In contrast, the pale-green eclogite (i.e., LT one, Fig. 3d, Supplementary Fig. 1d–f) is filled with fine-grained pink-white garnet porphyroblast (0.2–0.6 mm) and grayish green clinopyroxene matrix (20–50 µm) associated with minor blue amphibole, phengite, paragonite and rutile armored by titanite. Epidote is mainly present, as retrograde phase after lawsonite breakdown, in the pale-green eclogite (i.e., LT one, Supplementary Fig. 1d, e).

Fig. 3: Distinct petrological features.
figure 3

a Representative petrography of HT eclogite on microscope scale; more detail is presented in Supplementary Fig. 1 in Supplementary Information. b, c Si and Fe microprobe mapping show the localities of key minerals (i.e., quartz and ilmenite) and massive irregular patches of quartz inclusions in the garnet porphyroblast of HT eclogites; measurements of Ti-in-quartz and Zr-in-rutile are highlighted; see Supplementary Fig. 2 for corresponding BSE image and microprobe Mn, Ca, Mg elemental mapping. d Petrographical feature of LT eclogite on microscope scale.

Full size image

Mineral chemistry analysis (Supplementary Fig. 3b, e, Supplementary Data 2) discloses that garnet in HT eclogite is generally enriched in Mg (i.e., Xpyrope ~0.14–0.18) but depleted in Fe and Ca (i.e., Xalmandine ~0.65–0.72 and Xgrossular ~0.12–0.19), compared to that in LT eclogite (Xpyrope ~0.04–0.08, Xalmandine ~0.68–0.78 and Xgrossular ~0.18–0.21); In addition, clinopyroxene in HT eclogite is principally of diopside-aegirine-augite (Xjadeite ~0.21–0.26, Xdiopside ~0.52–0.56 and Xaegirine-augite ~0.18–0.26), countering to that from LT one (Xjadeite ~0.42–0.52, Xdiopside ~0.42–0.47 and Xaegirine-augite ~0.04–0.15); Besides, amphibole in HT eclogite displays solid-solution composition of winchite-barroisite while that from LT one falls within the range of glaucophane-barroisite; White mica in both HT and LT eclogites are mainly phengite (Si4+ ~3.32–3.50 a.p.f.u.); Contrasting Zr content (~600–750 versus ~20–35 µg g−1, Fig. 4a, Supplementary Fig. 4, Supplementary Data 5), which may be temperature dependent56, characterizes rutile from HT and LT eclogites. Besides, Ti-in-quartz solubility was evaluated, ranging from 2 to 5 µg g−1 (Fig. 4a, Supplementary Fig. 5, Supplementary Data 6), as a thermobarometer in combination with Zr-in-rutile thermometer and independent P-T estimates via thermodynamic modeling.

Fig. 4: Contrasting P-T histories.
figure 4

a Summary of P-T-time histories of studied eclogites; age constraints are from SIMS rutile U-Pb chronology in Supplementary Fig. 7; red and green trajectories are mainly derived from thermodynamic modeling with mineral compositional constraints (Supplementary Figs. 8 and 9); isopleth of Ti-in-quartz and Zr-in-rutile are plotted via calibrations of refs. 57,132 with pressure set to 2.7 GPa; solidus of various compositions are cited from thermodynamic modeling145,149 and experimental petrology150,151; comprehensive reviews32,33,35,36,74 of the P-T estimates, via thermodynamic modeling, quartz elastic barometer, and RSCM & Zr-in-rutile thermometers, of the STMC, which suggests a deep refrigerated condition (i.e., ~2.5 GPa, 540 °C), are also plotted for comparison. b P-T characteristics of the STMC compared to the global compilation30,58; worldwide representative fossil subduction suture zone is marked by bold text & rectangle; KO Kokchetav, SQD South Qinling-Dabie, China; NQDS North Qinling-Dabie-Sulu, China; DM Dora Maira, Italy; ZS Zermatt, Switzerland; VO Voltri, Italy; FC Franciscan C., California; SI Sistan, Iran; SLC Schistes Lustrés, Corsica; SLA Schistes Lustrés, Western Alps; ZA Zagros, Iran. Coesite-Quartz transformation follows the one that ref. 152 calibrated experimentally.

Full size image

Rutile in HT eclogite records temperature of 808 ± 15 °C (n = 72, with calibration of ref. 57), while in LT eclogite, the temperature is averaged at 555 ± 18 °C (n = 49). None of correlation, between recorded Zr-in-rutile temperatures and the presence or absence of garnet armoring (820 ± 9.3 °C versus 814 ± 9.1 °C, Supplementary Data 5), is observed (Supplementary Fig. 4a). Such a HT condition is also echoed by temperature estimate ~791 ± 38 °C by Ti-in-quartz thermobarometer with pressure set to 2.7 GPa (n = 69, Fig. 4a, Supplementary Data 6). P-T evolution of HT eclogites, as constrained by pesudosection modeling and chemical zonation of key minerals (Supplementary Figs. 8, 9), is characterized by near isobaric burial with substantial heating from 720, 2.6 to 820 °C, 2.7 GPa, and subsequently isothermally decompressing, crossing the wet solidus (Fig. 4a, Supplementary Fig. 8c), to 805 °C, 2.0 GPa. By comparison, LT eclogite experiences slightly decompression with heating from 450 °C, 2.6 GPa to its peak metamorphism at 530 °C, 2.2 GPa (Fig. 4a, Supplementary Fig. 9). P-T characteristics of the STMC (Fig. 4b) is also compared to the global compilation30,58.

Bulk-rock geochemistry

Bulk-rock major and trace elements, as well as Sr-Nd isotopes, analysis (Supplementary Data 3, 4) were conducted to investigate the protolith nature of studied eclogites. Both HT and LT eclogites are substantially of basaltic composition, but the latter is more mafic (Supplementary Fig. 6a) with different SiO2 (~52 versus ~44 wt %), CaO and FeOtot but comparable MgO and Al2O3 concentrations. HT eclogites show broadly positive slope in LREEs to MREEs and flat in HREEs, with (La/Sm)N ~1.59–3.46 and (Gd/Yb)N ~0.94–1.576, as well as (86Sr/87Sr)i ~0.703923–0.707086 and (143Nd/144Nd)i ~0.512352–0.512488 isotopic compositions. By contrast, patterns of depleted LREEs-MREEs and horizontal HREEs ((La/Sm)N ~0.28–1.43, (Gd/Yb)N ~0.7–1.05) are tied to LT eclogites with ~0.702494–0.705392 and ~0.512370–0.512678 in (86Sr/87Sr)i and (143Nd/144Nd)i ratios, respectively.

Rutile and zircon U-Pb isotopes

Forty-seven and twenty-three rutile grains, with diameters about 75.3 ± 20.0 and 136 ± 57.0 µm (Supplementary Fig. 4a, b), from two HT and LT eclogites, respectively, were analyzed (Supplementary Data 9). The measured U contents of all rutiles range from 0.05 to 4.1, mostly <1 µg g−1 with f206 values of common lead between 1 and 88%. On the Tera-Wasserburg linear regression plot, HT eclogites (AT54 and 50, Supplementary Fig. 7a, b) yield lower intercept ages of 312 ± 2.7 Myr (with upper intercept at 207Pb/206Pb = 0.86 ± 0.03, MSWD = 1.0, n = 26) and 313 ± 7.4 Myr (with upper intercept at 207Pb/206Pb = 0.96 ± 0.04, MSWD = 2.1, n = 21), broadly consistent with their weighted mean 206Pb/238U ages of 311 ± 5.8 Myr (MSWD = 0.21) and 307 ± 14.6 Myr (MSWD = 0.22), within errors, using the 207Pb-based common-lead correction59. Similarly, a lower intercept age of 315 ± 8.2 Myr (with upper intercept at 207Pb/206Pb = 0.97 ± 0.06, MSWD = 1.4, n = 23) and a weighted mean 206Pb/238U age of 303 ± 9.0 Myr (MSWD = 0.83) were obtained for LT eclogite (AT47, Supplementary Fig. 7c). Eleven and twelve spots of zircon grains, respectively, from HT and LT eclogites (AT54 and 47) were measured for trace element compositions and U-Pb dating (Supplementary Data 7–8). The Th/U ratio of all zircons vary from 0.64 to 2.43, mostly >0.85. Blurred core-rim structure was observed for a few grains but most of rims were too narrow to be analyzed (<10 µm). Core domains of zircons from HT and LT eclogites (Supplementary Fig. 7d, e) give concordia ages of 350 ± 5.1 Myr (MSWD = 6.2, n = 11) and 398 ± 4.1 Myr (MSWD = 3.6, n = 12), respectively, in accordance with their weighted mean ages of 349 ± 6.9 Myr (MSWD = 7.9) and 394 ± 4.4 Myr (MSWD = 2.2). To be stressed, REEs patterns of analyzed zircons (Supplementary Fig. 7f) uniformly show positive slope from LREEs to HREEs with obvious Eu negative abnormal, typical of zircon with igneous origin60.

Compilation of age and geochemical data of regional rocks

To elucidate the nearly simultaneous crustal and mantle responses in upper plate (i.e., KYCTC) during which the thermal excursion occurred and witnessed by the formation of HT eclogites within the subduction plate interface, we fingerprint the spatial and temporal variation and evolution of regional magmatism, deformation and metamorphism, based on systematic geochemical and chronological data collection of various lithologies (Relevant references are listed in Supplementary Data 10–13) among the STOB. This includes: (1) 403 arc intermediate rocks (including adakite, Figs. 5, 6b, e, Supplementary Fig. 11, Supplementary Data 10) and 753 granites (Fig. 6d, Supplementary Fig. 12, Supplementary Data 11) with complete major and trace element compositions, radiometric ages, and bulk-rock Sr-Nd isotopic data; (2) 74 metamorphic rocks with radiometric ages and P-T estimate from the Atbashi and Akeyazi LT-(U)HP metamorphic complexes in the STOB (Fig. 6a, Supplementary Data 12); (3) A compilation (Fig. 6c, d, Supplementary Data 13) of 142 age data of regional metamorphism, fossils in depositing basin, and main strike-slip faulting & thrusting events was also conducted for chronological comparison. Bulk-rock Hf/Nd, Nd/Sr, Sr/Y, 87Sr/86Sr, and 143Nd/144Nd elemental ratios were then utilized as proxies (Figs. 5, 6) to differentiate metamorphic and magmatic products with distinct source signatures and their spatial-temporal evolution, along with crust-mantle responses, during the continuous subduction of the STO. The compilation of zircon Lu-Hf isotopic data, together with ages, from igneous and sedimentary rocks37 and the regional ophiolite ages46,61,62 are presented for comparison. Compositions of relevant geochemical endmember mixing components are present in Supplementary Data 14.

Fig. 5: Plot of Sr-Nd isotopes versus key trace elements ratios.
figure 5

a, b, c 143Nd/144Nd, 87Sr/86Sr against Hf/Nd, Nd/Sr ratio plots; gray arrows paralleled to axes of Hf/Nd and Nd/Sr ratios indicate substantial fractionation caused by mélange melting with limited variations in 143Nd/144Nd and 87Sr/86Sr; literature data are collected from the GEOROC database (http://georoc.mpch-mainz.gwdg.de/georoc/); the compilation of STOB arc intermediate rocks is present in Supplementary Data 10; mixing lines between the mantle and bulk sediment and between the mantle and sediment melts are shown as examples; the black-gray bar illustrates the range of sediment melts possible for sediment melting down to a degree of as little as 1%; sediment melts were computed by using relative partition coefficients of DSr = 7.3, DNd = 0.35 and DNd/DHf = 0.9–4.3 in accordance with the data observed in experiments by ref. 16 over the temperature range of 750 to 900 °C; sources and compositions of endmember mixing components are list in Supplementary Data 14.

Full size image
Fig. 6: Plots of compiled multi-disciplinary evidence.
figure 6

a The temporal evolution of the peak temperature of metamorphic rocks in the STMC; geothermal gradient is shown as color bar; ages of seafloor thermal alteration in rodingite, as marked by black arrows, are from ref. 153. b, c Ages against the Hf/Nd and Nd/Sr ratios of arc intermediate rocks in the STMC; 143Nd/144Nd and 87Sr/86Sr are highlighted as color bar; ages, marked by black arrows, of fossil in submarine/hemipelagic basins are from refs. 54,154. The temporal evolution of d the Sr/Y ratio of arc intermediate rocks and e detrital & inherited zircon εHf(t) of (meta-)sedimentary & igneous rocks in the STMC; data distribution of scatter plot is evaluated by Voronoi density155. Histograms are also plotted from compiled regional data, listed in Supplementary Data 10–13, for comparison.

Full size image

Discussion

Eclogitic mélange rocks and thermal excursion

Direct geophysical investigation, although facilitating observations of subarc processes, does not provide the resolution required to obtain detailed micro-scale information of material transfer at the slab-mantle interface19. In contrast, the chemical and isotopic composition of metamorphic products and arc lavas serve as the best candidate to extract information on this critical realm of the subduction zone. Completely distinct geochemical features characterize studied eclogites in the Kembel mélange of the ATMC. Of these, HT eclogites are of intermediate composition (SiO2 ~52 wt %) with moderately varying values of TiO2, FeOtot, MgO, and CaO (Supplementary Data 3), and exhibit moderate enrichments in LILEs (e.g., Rb, Ba, Sr), HFSEs (e.g., Th, Nb, Ta) and LREEs, with relatively higher 143Nd/144Nd but lower 87Sr/86Sr isotopic ratios, compared to those of relatively mafic LT ones (Fig. 5, Supplementary Fig. 6b, d, e). In addition, HT eclogites present slightly Nb-Ta negative abnormal with ambiguous similarity to those of subduction-related lavas63,64, average mélange rocks5 and corresponding compositions of their modeled65 and experimental9 melts in trace element pattern (Supplementary Fig. 6d). Whereas close resemblance to those of depleted mantle66 (DM) and altered oceanic crust67 (AOC), without any Nb-Ta negative abnormal, is tied to LT eclogites. Further evidence comes from their sharp discrimination in Th/Yb-Nb/Yb diagram (Supplementary Fig. 6b) and εNd(t) values (~2.2–4.9 versus ~2.6–8.6; Supplementary Data 4), calling for in-depth inquiry to the petrogenesis and protolith as echoed by the fractionation of key trace elements and isotopes during which normal melting processes (e.g., at mid-ocean ridges68) and/or material transport from descent slab to overlaying metasomatized mantle63,64,69 took place. In Sr-Nd isotopes, Nd/Sr and Hf/Nd diagrams (Fig. 5), LT eclogites predominantly gather around the domain of DM and AOC with broadly unperturbed trace element and radiogenic isotope ratios, implying an origin of slab mafic components. By comparison, noteworthy deviation, aloof from DM and AOC endmembers, is bound to HT eclogites, suggesting a protolith of hybrid chemical composition with the addition of substantial sedimentary component. Since AOC-derived fluids generally have too low Hf/Nd and Nd/Sr ratios and distinctly higher 87Sr/86Sr19,67, this eliminates fluid from slab as the major endmember responsible for generating the Sr-Nd elemental & isotopic ternary mixing space overlapped with studied HT eclogites. Experimental works7,15,16 demonstrate that a combination of partial melting of subducted sediments and AOC metamorphic dehydration could be of capability to produce required trace element fractionations as observed in global arc lavas19,69,70 and HT eclogites in this study (Fig. 5, Supplementary Fig. 6). Yet, fairly high 87Sr isotopic composition and sharp Sr-Nd isotopic curvature, as well as extremely low Nd/Sr ratio, of simple sediment melts71 would yield ternary mixing domain far deviated from data of studied HT eclogites. In short, it follows that the Sr-Nd isotope systematics of HT eclogites evaluated here (Fig. 5b, c) denote a potential physical mixing26,72 of DM and AOC with ~2–4 weight percent of bulk sediments, which leads to the formation of hybrid rock types, rather than the direct addition of melts and/or fluids from slab components (e.g., sediment, AOC). Noteworthy, the relatively minor sediment component in some natural mélange, which is also mirrored by the amounts of sediment contribution in global arc lavas (Fig. 5), does not preclude the formation of accessory minerals and corresponding trace elements fractionation if mélange contains substantial AOC component9,19.

The finding of key index minerals (i.e., the existence of Mg-rich garnet, Fe-rich clinopyroxene, ilmenite, and high Zr content rutile, Fig. 3a–c, Supplementary Figs. 1, 4) and contrasting petrological features (i.e., the preservation of massive irregular patches of quartz and pseudomorphs after melt films, Fig. 3a, b, Supplementary Figs. 1a–c, 2g, h) as well as P-T constraints retrieved from multi-methodology approaches (Fig. 4a, Supplementary Figs. 4c, 8, 9), collectively suggest a contrasting thermal history of studied eclogitic mélange rocks with substantial heating, after peak burial, to HT condition (~810 °C, Fig. 4a, Supplementary Fig. 8). Alternatively, oversteeping breakdown of epidote group mineral via a reaction like cpx + zo = grt + qz + H2O may generate observed patchy quartz textures in garnets of HT eclogites. However, their relatively poor concentration of bulk-rock CaO (~7.8 wt%, compared with ~12.3 wt% of that of LT one; Supplementary Data 3) could essentially rule out this possibility, especially to condition >700 °C (Supplementary Figs. 8, 9; ref. 73). To be stressed, numerous previous estimates (in average of 2.8 GPa and 540 °C, as reviewed by refs. 33,74; Fig. 4) and plenty of lawsonite39,75,76,77 and coesite32,35,38,78 reports, as well as the presence of closely associated LT eclogites in the region (Figs. 1, 2), had confirmed the STMC is of close resemblance to a deep preserved subduction plate interface32,79, plausibly standing in contradiction with such an extraordinary eclogite-facies thermal excursion (∆T ~250 °C, Fig. 4a). Several processes can be envisioned for “unusual” heating: (1) Fluid, released from descant slab, could drive exothermic carbonation and hydration reactions in the subduction plate interface or overlying mantle wedge80 with temperature increasing up to 75–200 °C at high convergent rate ~10 cm yr-1 and P ~1.0 GPa81. However, pervasive hydration seems impossible at such depths in UHP eclogite-facies condition33; (2) Heat advection from syn-metamorphic mafic magmas together with potential fluid flow and/or radioactive decay could contribute to the thermal “budget”, but is likely to be <30 °C along the subduction plate interface80,82; (3) Shear heating, which is proportional to velocity times rock strength, could in principle modify the thermal budget along the subduction thrust by up to ∼200 °C80,83, but would be limited in fluid-rich environments or buffered by endothermic dehydration reactions (i.e., lawsonite and chlorite) predicted across the refrigerated temperature range of the STMC. Besides, none of eclogitic mylonites and/or ultra-mylonites, advocating for high strain, are observed in region33,51,74; (4) Convective motion of subducted materials could also strongly affect heat transport and temperatures80,84,85. It is effectively limited to depths greater than 80 km, as the “cold nose” with low-seismic attenuation is frequently observed86,87 and expected88,89 to be a comparatively rigid part of the upper mantle wedge, indicating relative isolation from heat flow88. Although deeply tectonic slicing and stacking of UHP metamorphic volcanoclastic sequence were locally envisioned in the Akeyazi area32,33,74, the LT-(U)HP metamorphic complexes (i.e., the Atbashi51, Chatkal48,49 and Fan-Karategin50 sections) across the entire STMC are mainly dominated by the exposure of blueschist-/eclogite-facies mafic blocks wrapped by meta-sedimentary host rocks (500–600 °C, 2.0–3.0 GPa). They constitute the architecture of deeply subducted and cooled sedimentary mélange with relatively low bulk viscosity32,84,85 (compared to rigid overlaying and descending plates), which could facilitate the efficient fluid circulation90 and ensure convective motion91 within subduction plate interface and/or short-lived translation34 toward the hot corner of mantle wedge. Thus, substantial thermal excursion after peak burial, which was exactly across the wet solidus of basic to intermediate composition (Supplementary Fig. 8), as witnessed by studied HT eclogitic mélange rocks could, in fact, disclose the potential translation of mechanical mixed subducted materials nearer to a relatively hotter mantle wedge and meanwhile farther away from the refrigerated subduction plate interface.

Mélange diapir melting in refrigerated subduction plate interface

Plentiful studies69,92 tie the chemistry of the arc magma to a variable contribution from the DM, AOC, and slab sediments. However, experimental petrology studies have faced challenges in simultaneously reproducing both major and trace element features of the most common types of arc magmas20. Nevertheless, recent debates challenge the traditional model, which attributes characteristic trace element signal of arc magmas solely to hybridized mantle wedges formed through discrete feedings of fluids and/or aqueous melts from slab components9,19. Alternatively, trace element and isotope variability in global arc magmas can be adequately explained only if physical mixing of DM, AOC, and sediments occurs early within the plate interface during subduction, before melting begins5. This revised prerequisite essentially substantiates the momentous role of mélange in arc magmatism. Although kilometers-thick low-seismic velocity regions atop subducting slabs worldwide could imply the existence of mélange zones at the slab-mantle interface26,27,28, direct field evidence confirming extensive exhumation of buried sediments along the plate interface remains scarce, further hampering a comprehensive assessment of the extent of chemical and/or mechanical disruption of subducted materials within the plate interface. To be stressed, massive sediment accretion at ~80 km depth along the subduction interface has been recently disclosed in the STMC32,33, denoting continuous refrigeration, by incoming cold material from the slab, and juxtaposition to the “cold nose” of mantle wedge. In addition, short-lived thermal pulse (~80 °C, to 600 °C within ~300 Kyr) was revealed from unusual garnet zonation in coesite-bearing oceanic eclogites34, advocating potential translation of UHP refrigerated slices to a relatively hotter mantle wedge. In particular, the trace element and Sr-Nd isotope systematics of HT eclogites evaluated here (Fig. 5b, c) denote a probable physical mixing26,72 of DM and AOC with bulk sediment, leading to the formation of mélange rocks, instead of the addition of melts and/or fluids from slab components. The discovery of massive irregular patches of quartz, which resembles granitoid products from melt experiments for pelitic to felsic lithologies6, and “nano-diorite” pseudomorphs after melt films in studied HT eclogites (Fig. 3a, b, Supplementary Figs. 1a–c, 2g, h) as well as P-T constraints retrieved from multi-methodology approaches (Fig. 4a, Supplementary Figs. 4c, 8, 9), jointly suggest a striking thermal history of studied eclogitic mélange rocks with substantial heating, after peak burial, to the condition rightly crossing the wet solidus on pseudosection of basic to intermediate composition (~810 °C, Fig. 4a, Supplementary Fig. 8). Such translation toward the hot corner of mantle wedge could be short-lived around several hundred thousand to several million years if considering the limited gap between the ~312 ± 2.7 Myr cooling timing of HT eclogitic mélange rocks and the ~315 ± 8.2 Myr crystallization age of LT oceanic eclogites (Fig. 4a, Supplementary Figs. 4c, 7a–c). It also broadly meets the timeframe of 0.01–1.0 Myr21,25, with a range of modeling parameters, for the mantle wedge diapirs rising from the slab to the magma chamber. This short-lived ~250 °C thermal excursion lasting ~1.65 Myr in average would correspond to displacements on the order of ~0.2 to 2 cm yr-1 assuming a thickness32,79 between 0.5 and 5 km for the subduction plate interface (in the broad range of geophysical imaging93, i.e., 2–5 km), compatible, in any case, within the range of typical exhumation rates worldwide, either for (U) HP oceanic or continental rocks (~1-5 mm yr-1, or ≥ cm yr-1, respectively)30. Yet, extremely high exhumation velocity, to ~40–45 cm yr-1, of diapiric stokes flow, between depths of 50 to 80 km within the subduction plate interface, was locally reported from the Akeyazi UHP metamorphic complex in the China segment of the STOB94.

It is noteworthy that almost all (>95%) Late Carboniferous arc intermediate rocks, which constitutes a flare-up period of Late Paleozoic magma emplacement in the STOB (i.e., the Central Tianshan Arc95), uniformly show signal of MDP with horizontal Hf/Nd arrays but little variation in 143Nd/144Nd (Figs. 5, 6b, c), serving as potential magmatic products in lower crust echoing with the simultaneous direct HT eclogitic records (Figs. 3a–c, 4a) near subarc depth supporting mélange diapir took place and subsequent trace element fractionation (main Hf/Nd, Nd/Sr) caused by mélange partially melting19. Such a Late Carboniferous flare-up of arc magmatism with MDP signal is coupled with noteworthy crustal thickening associated with production of high magma volumes96 (~40 km3 (km Myr-1)-1) along with relatively long-term compressional environment33,97 during the subduction of the STO (Fig. 6d). Thick crust, which will suppress mantle wedge melting and enhance intra-crustal differentiation65,98, could also offer a favorable condition for the generation of intermediate melts (e.g., sedimentary mélange melting during diapirism6). In addition, growing feed of juvenile slab components (Fig. 6e) almost contemporaneously follows with this period, suggesting a potential tectonic scenario with melted equivalents of mélange rocks buoyantly rise toward hot mantle wedge, generate magmas with MDP signal and occasionally leave partially melted HT eclogitic relics exhumed physically to shallow depths along the refrigerated subduction plate interface after thermal excursion (Fig. 4a, 7), with subsequent latent oceanic slab roll-back and break-off.

Fig. 7: 3D schematic drawing of the first tangible rock records substantiating mélange diapir melting model.
figure 7

Deeply accreted mélange and its buoyantly raised diapir are marked in gray with melt zones highlighted in red. Red arrow indicate trajecty of studied HT eclogites entrained in a mélange diapir within refrigerated subduction plate interface, and green path of LT eclogite is also show for comparison with insets depict their contrasting petrological and petrographic features.

Full size image

Implications for volatile recycling

The devolatilization of subducting slabs serves as a critical role in the melting of mantle wedge, generation of arc magmas, and regulation of global geochemical cycles. Petrological Modeling12,99,100, based on geophysical observation of thermal structure of global subduction zones, discloses that volatile (up to ~33–66% mineralogically bound H2O and ~65–95% CO2 in slab lithospheric components) could efficiently pass through old and fast, namely cold, subduction zones (e.g., Honshu in the western Pacific). Whereas much weighty proportion of volatile derived from the slab could be lost beneath the forearc in hot subduction zones (e.g., Cascadia). On average, as representative of the range of thermal conditions in modern arcs89, about four fifths of CO2 and one third of H2O are retained within slab beyond the subarc12,99, regulating the Earth with secular cooling13,101 as the efficiency of volatile recycling, to deep mantle (e.g., ≥240 km99), increased over the Proterozoic and Phanerozoic. The principal volatile components in subducting lithologies are respectively bound in carbonated and hydroxylated minerals within marine sediments, hydrothermally AOC, and serpentinized upper oceanic mantle12,102. Though controversy exists regarding the melting of subducted lithologies as a potential source for arc volcanism and the mobilization of volatile, nearly all melting in subduction settings requires the presence of fluid70,103. Yet the mechanism of fluid flow propagation (i.e., pervasive versus channelized10,11,12) and the fact of Phanerozoic secular cooling of Earth mantle13, indeed, hinder the efficiency of volatile mobilization beneath the subarc, further making the feeding of subducted volatile into the mantle wedge remains ambiguous14.

Alternatively, considerable amounts of volatile (H2O ~1.0–11.8, CO2 ~1.2–9.7 wt%; see ref. 5 for systematic review) are enriched in mélange rocks, comparable to those in oceanic slab components14,102,104. It could potentially act as a temporary carrier of H2O, CO2, and sulfur with relatively low bulk viscosity and density25 facilitating the hybridization of mantle wedge and the transformation of volatile from descent slab to arc magmas. Such process, as widely refers as buoyant diapirism5,20 of high-pressure mélange, invokes physical mixing of slab components and mantle wedge and generates arc magmatic products, with MDP signal19 (e.g., Figs. 5, 6) which were identified globally from representative arcs (e.g., Tonga19, Aleutians70 and Marianas69). In addition, mélange melting at the top of the slab could potentially record residual garnet fingerprint19, such as high Sr/Y and HREEs depletions as observed in many arc lavas, although it have been classically interpreted to reflect melting of the subducted basaltic crust at high pressure105. Experimental20,23,106 and field-based studies107,108,109,110 point to the presence of orthopyroxenites may also potentially record the hybridization of mantle wedge by mélange materials. Besides, the abundance of 10Be and the common existence of positive Sr anomalies in arc lavas indicate non-negligible contribution from subducted ocean-floor materials (sedimentary and oceanic crustal sources) to arc magmas111,112. Locally, channelized HP veins113,114, carbonated eclogites115 and graphite-rich eclogitic meta-sediments and meta-basalts116,117 were well preserved in the STMC, making a valuable nature “laboratory” of volatile-rich deeply buried mélange “package”. The fact that carbonates in eclogitic meta-basalts show mantle-like C-O isotopic features while marbles have sedimentary carbonate isotopes115,116 essentially disclose the necessity of a momentous contribution of carbonate-rich oceanic slab components in generating alkaline arc lavas65. CO2-bearing reactive fluid flows were emphasized118 only at the slab-wedge interface without being emitted to the volcanic arc or transported deeper into the mantle, further implying a great capacity of mélange diapir scenario in recycling volatile. In particular, the trace element and Sr-Nd isotope systematics of HT eclogites evaluated here (Fig. 5b, c) denote a probable physical mixing26,72 of DM and AOC with bulk sediment, leading to the formation of mélange rocks, instead of the addition of melts and/or fluids from slab components. The discovery of key mineralogical evidence (Fig. 3a, b, Supplementary Figs. 1a–c, 2g, h), together with P-T constraints (Fig. 4a, Supplementary Figs. S4c, 8, 9), jointly suggest a striking thermal history of the mélange “package” (as witnessed by studied eclogitic mélange rocks) with substantial heating, after peak burial, to the condition rightly crossing the wet solidus of basic to intermediate composition (~810 °C, Fig. 4a, Supplementary Fig. 8). Such translation, which could be short-lived around several hundred thousand to several million years, toward the hot corner of mantle wedge could serve as the first tangible rock records (Fig. 7) supporting mélange diapirs may, indeed, propagate, and dynamically mix with the overlying mantle as conceptual vitalization5,19,20 and modeling23,25 predicted. Coincidently, the contemporaneous Late Carboniferous flare-up of regional arc magmatism with MDP signal, which is coupled with production of huge magma volumes96, collectively suggests a viable and non-negligible process of the hybridization of mantle wedge by buoyant mélange materials, to transfer volatile, generate arc lavas and regulate terrestrial geochemical cycles, stands.

Methods

Microprobe analysis

The major element compositions of pivotal minerals in the eclogites investigated were determined through electron microprobe analyses of polished thin sections. These analyses were conducted at Wuhan Sample Solution Analytical Technology Co., Ltd., China, utilizing a Jeol JXA-8100 instrument. Quantitative measurements were performed with wavelength dispersive spectrometers, featuring an acceleration voltage of 15 kV, a beam current of 15 nA, a 3 μm beam size, and a counting time of 30 s. Reference standards, both natural minerals and synthetic oxides, were employed, and data correction was carried out using a program based on the ZAF procedure. Detailed information regarding the samples, their mineral assemblages, and the results of representative microprobe analyses of key minerals are presented in Supplementary Data 1 and 3, respectively.

Bulk-rock major and trace elements analysis

The analysis of major element compositions was carried out at the Institute of Geology and Geophysics, Chinese Academy of Sciences. Major oxide concentrations were determined using a Phillips PW1480 X-ray fluorescence spectrometer (XRF) on fused glass disks. Loss on ignition (LOI) was measured following heating to 1000 °C. Most major oxides had uncertainties of approximately 2%, while MnO and P2O5 displayed deviations of around 5%. The total measurements consistently fell within the range of 100 ± 1 wt%. Whole rock Fe2O3 content was determined via potassium permanganate titration119. Trace element concentrations were assessed through sector field inductively coupled plasma mass spectrometry (ICP-MS) using a Finnigan MAT ELEMENT spectrometer at the Institute of Geology and Geophysics, Chinese Academy of Sciences. Relative standard deviations (RSD) were generally within ± 10% for most trace elements, though they extended to ± 20% for V, Cr, Co, Ni, Th, and U, based on rock standard analyses. Detailed major and trace element analyses, as well as the comprehensive bulk-rock major element composition for thermodynamic modeling, are presented in Supplementary Data 2.

Bulk-rock Sr-Nd isotope analysis

High-precision isotopic measurements for the eclogites under investigation, specifically Sr and Nd isotopes, were performed at Nanjing FocuMS Technology Company. The methodology involved the mixing of geological rock powders with 0.5 ml of 60 wt % HNO3 and 1.0 ml of 40 wt % HF in high-pressure PTFE bombs. These steel-jacketed bombs, designed for safety, were subjected to 195 °C in an oven for a duration of 3 days. Following this, the digested samples were desiccated on an HTplate and reconstituted in 1.5 ml of 1.5 N HCl before undergoing ion exchange purification. The elutions containing Sr and Nd were gently evaporated to dryness and subsequently re-dissolved in 1.0 ml of 2 wt % HNO3. To further enhance precision, a diluted solution (0.05 µg g-1 Sr, 0.05 µg g-1 Nd) was introduced into the Nu Instruments Nu Plasma II Multi-Collector ICP-MS (MC-ICP-MS) in Wrexham, Wales, UK, through the Teledyne Cetac Technologies Aridus II desolvating nebulizer system in Omaha, Nebraska, USA. Raw data for isotopic ratios were internally corrected for mass fractionation by normalizing to specific ratios, namely 86Sr/88Sr = 0.1194 for Sr and 146Nd/144Nd = 0.7219 for Nd, using the exponential law. International isotopic standards, such as NIST SRM 987 for Sr and JNdi-1 for Nd, were periodically analyzed to rectify instrumental drift. Quality control measures were upheld through the analysis of geochemical reference materials, namely USGS BCR-2, BHVO-2, AVG-2, and RGM-2. The isotopic results obtained during this study were in agreement with those reported in prior publications, and the analytical uncertainties were well within acceptable limits120. Detailed Sr-Nd isotopic data of studied samples and modeling parameters are given in Supplementary Data 4.

Laser ICP-MS trace element analysis of rutile and quartz

Quantitative trace element analysis of rutile and quartz was performed in situ using separate laser ablation systems. For rutile analysis, the ESI New Wave NWR 193UC (TwoVol2) laser ablation system was employed, connected to an Agilent 8900 ICP-QQQ at Beijing Quick-Thermo Science & Technology Co., Ltd. Simultaneously, the analysis of quartz was carried out using a double focusing sector field ICP-MS instrument, specifically the Finnigan MAT model-ELEMENT, at Wuhan Sample Solution Company. These analyses were conducted on polished 50 µm-thick thin sections and selected mineral grains embedded in epoxy resin from the studied eclogites. Helium served as the carrier gas to enhance the transport efficiency of the ablated material, which was mixed with makeup gas (Ar) prior to entering the ICP, ensuring the maintenance of stable and optimal excitation conditions.

In the case of rutile analysis, a 193 nm (ArF excimer) argon fluoride New Wave Research Excimer laser ablation was conducted, lasting for 30 s. This process featured a 40 μm diameter beam size, delivering an energy of approximately ∼4 J cm-2 and operating at a repetition rate of 5 Hz. Rutile analysis was carried out on the selected grains and reference materials after an initial 10-s baseline signal collection and 3 pulses of pre-ablation. Calibration of trace elements was achieved using NIST SRM 610 glass, with the internal standard major element Ti. These reference materials were analyzed twice, both before and after each analytical session, covering 6–8 spots on mineral samples. The data underwent background subtraction and correction for laser-induced elemental fractionation downhole, a process facilitated by the Iolite data reduction package integrated within the Wavemetrics Igor Pro data analysis software121.

For quartz, laser ablation was conducted using a Finnigan MAT UV laser probe operating at 266 nm. The laser was operated with a pulse width of 3 ns (Q-switched), a sHT frequency of 20 sHTs s-1, a pulse energy of 0.5 mJ, and a spot size of 30 µm on rasters not exceeding 300 × 300 µm. To optimize the operating parameters of the ICP-MS, ablation of the reference material NIST 612 was carried out using 139La, with a focus on achieving maximum sensitivity and stability. Si contents, determined from the chemical formula, served as the internal standard for evaluating element abundance.

The acquisition time was structured with 20 s allotted for background measurements and 120 s for mineral analyses. External calibration standards were established using glass reference materials NIST SRM 610 and NIST SRM 612. The reproducibility and accuracy, as determined with NIST SRM 610 and NIST SRM 612, typically yielded results of less than 8% and 6%, respectively. Trace element concentrations were subsequently calculated using GLITTER Version 3122. Summary of apparent concentration of Zr and Ti of rutile and quartz, respectively, from different domains (i.e., in matrix and as inclusion in garnet) in studied eclogites were giving in Supplementary Data 5 and 6.

Laser ICP-MS U-Pb isotope analysis of zircon

Zircons were separated from samples processed by crushing, heavy-liquid, and magnetic methods and then were mounted in epoxy resin and polished to expose the interior. Subsequent to this preparation, cathodoluminescence (CL) imaging of the zircons was carried out using a scanning electron microscope at the Institute of Geology and Geophysics, Chinese Academy of Sciences. This step was integral in the selection of suitable zircon grains and the identification of optimal target sites for U-Pb dating. The U-Pb dating and the concurrent analysis of trace element composition in both rutiles and zircons were conducted using LA-ICP-MS at Wuhan Sample Solution Analytical Technology Co., Ltd., China. The operating conditions for the laser ablation system, the ICP-MS instrument, and the data reduction procedures were consistent with those described in ref. 123. Laser sampling was executed with a GeolasPro laser ablation system, comprising a COMPexPro 102 ArF excimer laser (with a 193 nm wavelength and a maximum energy of 200 mJ) and a MicroLas optical system. Ion-signal intensities were measured using an Agilent 7700e quadrupole ICP-MS instrument, with helium employed as the carrier gas. To maintain stability, argon was introduced as the makeup gas and mixed with the carrier gas via a T-connector before entering the ICP. Significantly, the laser ablation system was equipped with a “wave” signal smoothing device124. the laser’s spot size was set at 60 µm, and the frequency was maintained at 6 Hz. External standards, specifically zircon 91500 and glass NIST610, were used for U-Pb dating and trace element calibration, respectively. Each analysis began with a background acquisition period of approximately 20–30 s, followed by 50 s of data acquisition from the sample. To ensure accuracy, an Excel-based software, ICPMSDataCal, was employed for off-line selection and integration of background and analyzed signals, time-drift correction, and quantitative calibration for trace element analysis and U-Pb dating125. Concordia diagrams and weighted mean calculations were made using Isoplot/Ex_ver3126. Relevant data is listed in Supplementary Data 7 and 8.

SIMS U-Pb isotope analysis of rutile

U-Pb isotope measurements were conducted using the Cameca IMS 1280 ion microprobe at the Institute of Geology and Geophysics, Chinese Academy of Sciences. utile crystals were meticulously prepared by embedding them in transparent epoxy alongside reference standards. These reference standards included the R10 rutile standard (~30 µg g-1 U, Concordia age = 1090 ± 5 Myr)127, 99JHQ-1 rutile (highly variable U content with average of 2 µg g-1, 206Pb/238U age = 218 ± 1.2 Myr)128, and an in-house rutile megacrystal standard (JDX) (~6 µg g-1 U, 207 Pb/206 Pb age = 521 Myr, 206Pb/238U age = 500–520 Myr, unpublished TIMS data). The samples were carefully polished to expose the pristine interior of the crystals. Subsequently, the mount was coated with high-purity gold prior to ion probe analysis. The primary ion beam employed was O2, accelerated to 13 kV with an intensity of approximately 15 nA. The aperture illumination mode (Kohler illumination) was utilized, featuring a 200 μm aperture to ensure uniform sputtering across the entire area of analysis. The spot size was approximately 20 × 30 μm, displaying an ellipsoidal shape. Positive secondary ions were extracted using a 10 kV potential. The mass resolution of ~6000 was used and the magnet was cyclically peak-stepped through a sequence including the 206Pb+, 207Pb+, 208Pb+, U+, UO+, ThO+, UO2+ and 49TiO4+ to produce one set of data. A single ion-counting electron multiplier (EM) was used as the detection device. The 49TiO4+ signal was selected as a reference peak for aligning the secondary ion beams due to its robust intensity and lack of interference from ZrO. Each measurement cycle consisted of 10 cycles, and the total analytical time was approximately 15 min, inclusive of 2 min of rastering before the actual analysis to reduce the impact of surface-contaminant Pb. It’s worth noting that the consideration of mass fractionations of Pb isotopes and Pb hydrides, which would necessitate a mass resolution exceeding 30000, was omitted. Several studies have corroborated the negligible impact of these effects and their mutual cancellation129. Relevant data is listed in Supplementary Data 9.

Zr-in-rutile thermometer and Ti-in-quartz thermobarometer

Three commonly applied calibrations56,57,130,131 were used to estimate temperatures in dependent of Zr concentrations in rutile. We considered a pressure value of 2.7 GPa (for the calibration of ref. 57) in accordance with P-T estimates via thermodynamic modeling (Supplementary Figs. 8, 9) for both HT and LT eclogites, and with the finds of coesite in adjacent area (i.e., the Akeyazi metamorphic complex in the STOB, as reviewed by ref. 33). The activity of ZrSiO4 was set to unity because minor amount of zircon occurs in most samples. Only minor changes, around 15 °C, exists if pressure variations (i.e., from 2.4 to 2.8 GPa) and silica polymorph transition (α-quartz to coesite) are considering34. More scattered values were provided, by the calibrations of refs. 56,130,131, within the ranges of 50–100 °C in temperature difference (Supplementary Data 5).

Temperature was also evaluated for quartz crystallization specifically in HT eclogites, using Ti-in-quartz thermometers of refs. 132,133. The activity of TiO2 is set to unity (because of the widely present of Ti-rich phases) for the calibration of ref. 132 with a peak pressure value of 2.7 GPa. Considering the apparent changes in Ti-in-quartz solubility behavior that occurs at 2.0 GPa, extrapolating the calibration to higher pressures could yield inaccurate estimates of P-T conditions of crystallization132. Yet, the P-T estimate, if combined with Zr-in-rutile thermometer, is broadly consistent with that derived from thermodynamic modeling (within 50 °C, 0.1 GPa, Fig. 2e). In contrast, a relatively lower value (~70 °C) is given by the calibration of ref. 133, without taking into account of the effects of pressure and TiO2 activity (Supplementary Data 6).

Thermodynamic modeling

Pseudosection modeling for both HT and LT eclogites (AT54 and AT47, Supplementary Figs. 8, 9) were performed in the system NMnCKFMASHTO. TiO2 was considered due to the presence of rutile/ilmenite/titanite in matrix or as inclusions in garnet porphyroblasts. The fluid phase is assumed to be pure H2O and was set according to its exact amount in XRF data and effective bulk-rock compositions. CO2 was neglected as only small amounts of carbonate occur as thin secondary veins. Fe2O3 was set according to the bulk-rock data (Supplementary Data 3). To take into account the sequestration of elements induced by the growth zoning of garnet porphyroblast, effective bulk compositions (EBCs) were adjusted from XRF compositions by removing part of the garnet modal abundance, following the quantitative method of ref. 134 derived by the Mn zoning of garnet and calculated modal garnet trendline for each major element (Supplementary Fig. 10).

The P-T pseudosections were calculated using the software Perple_X 6.87135,136 and an internally consistent thermodynamic dataset (hp62ver.dat)137 based on the effective recalculated bulk-rock compositions. Mineral solid-solution models are Gt(HP)138 for garnet, Omph(GHP)139 for omphacite, Amph (DP)140 for amphibole, Mica(CHA)141 for white mica, Chl(HP)142, Ep(HP)138 for epidote/clinozoisite, Opx (W)143 for orthopyroxene, Bi (W)143 for biotite, melt(G)144 for silicate melt, Ilm (WPH)145 for ilmenite, Pl(I1, HP)146 for ternary-feldspar and H2O-CO2 fluid solution model is from ref. 147. The pseudosections are dominated by tri- and quadri-variant fields with a few di- and quini-variant fields. P-T conditions were further constrained by comparing predicted isopleths (Phengite Si4+, Xjadeite, Xpyrope, and Xgrossular) with measured mineral compositions incorporating typical uncertainties on EMPA analyses (ca. 3% to 5%)148.