Rapid endogenic rock recycling in magmatic arcs

In subduction zones, materials on Earth’s surface can be transported to the deep crust or mantle, but the exact mechanisms and the nature of the recycled materials are not fully understood. Here, we report a set of migmatites from western Yangtze Block, China. These migmatites have similar bulk compositions as forearc sediments. Zircon age distributions and Hf–O isotopes indicate that the precursors of the sediments were predominantly derived from juvenile arc crust itself. Using phase equilibria modeling, we show that the sediments experienced high temperature-to-pressure ratio metamorphism and were most likely transported to deep arc crust by intracrustal thrust faults. By dating the magmatic zircon cores and overgrowth rims, we find that the entire rock cycle, from arc magmatism, to weathering at the surface, then to burial and remelting in the deep crust, took place within ~10 Myr. Our findings highlight thrust faults as an efficient recycling channel in compressional arcs and endogenic recycling as an important mechanism driving internal redistribution and differentiation of arc crust.

M agmatic arcs witness the interplay between endogenous and exogenous processes, including magmatism, crustal thickening, uplift, erosion, sedimentation and burial of detritus [1][2][3][4] . Magmatism produces new crust, which later interacts with the hydrosphere and atmosphere through erosion and weathering. On the other hand, crustal materials from the surface are recycled to Earth's interior. This chain of processes in magmatic arcs play important roles in driving much of the mass exchange between Earth's interior and surface. The inward transport of surface materials, including volatiles, has profound influence on the cycling of carbon, oxygen, sulfur, etc. on Earth's surface and may alter the chemical and physical properties of the deep crust and even mantle.
Nearly every Phanerozoic arc in the world exhibit crustal signatures in geochemistry, suggesting pervasive crustal recycling in the formation of arc crust. Conventional views link crustal recycling processes to slab subduction, including sediment subduction and subduction erosion (± relamination) have been widely invoked to explain the crustal signatures seen in most arc magmas 5,6 . Yet the recent work on continental arcs hints at thrust faults as potential recycling channels [7][8][9][10] .
Here, we examined a suite of migmatites from a Neoproterozoic magmatic arc in western China. We used combined petrologic, geochronologic and geochemical studies of these samples to understand the nature of the recycled materials and evaluate how thrust faults may contribute to rock recycling in compressional arc settings.

Results
Geological setting and samples. The Yangtze Block in Eastern Asia consists of Archean-Paleoproterozoic crystalline basement surrounded by Neoproterozoic fold belts. It is bounded by the Tibetan Plateau to the west, the North China Block to the north and the Cathaysia Block to the southeast. It was placed in a marginal position in Rodinia supercontinent and has underwent a long-term evolution and complex tectonic-magmatic processes in a continental margin setting during Neoproterozoic [11][12][13] . The western margin of the Yangtze Block became tectonically active since the early Neoproterozoic; it started with intra-oceanic arc magmatism before 971 ± 16 Ma (ref. 14 ) and then transitioned to Andean-type magmatism at ca. 870 Ma (ref. 15 ). This ancient subduction relic was lately imaged by deep seismic reflection profile 16 . The prolonged magmatic history gave rise to linearly distributed Neoproterozoic arc magmatic rocks spanning over 800 km (Supplementary Fig. 1a). The Longmenshan Thrust Belt to the northwestern margin of the Yangtze Block exposed abundant Neoproterozoic plutonic complexes due to the major Miocene extrusion and thrust process 17 , of which the largest one is known as the Pengguan Complex, comprising voluminous 860-750 Ma plutonic rocks ( Supplementary Fig. 1b, c). The Huangshuihe Group in the core region of the Pengguan Complex exists as a huge roof pendant of the plutonic rocks and consists of metamorphic rocks of schist, metapelite, quartzite and meta-pyroclastic rock. Ductile deformation, faults, mylonite with S-C fabric, and migmatitic lineation are extensive in the sequences.
Two main types of migmatites were identified in the field (Supplementary Fig. 2): the inhomogeneous migmatites (or diatexite) contain abundant blocks of melanosome and associated aplite vein; the stromatic migmatites preserve a regular layered structure and are characterized by centimeter-thick, foliationparallel leucosome, melanosome and mesosome. Patch-shaped neosomes are abundant in stromatic migmatites and formed during incipient partial melting. Large leucosomes (~50 cm in width) occur occasionally and are usually fed by a few small leucosome veins. The stromatic migmatites have a NNW-dipping foliation (S 1 ) (~355°/48°) defined by oriented biotite or feldspar augen. The S 1 foliation is parallel to bedding planes defined by the metapelite (Supplementary Fig. 2c) and is folded locally by synanatectic deformation on varying scale ( Supplementary Fig. 2e, f). The fold axial planes (S 2 ) generally display E-W striking, Sdipping orientation. Besides, the study area was superimposed by massive high-angle, S-verging thrust faults ( Supplementary  Fig. 2g), which should be linked with post-Mesozoic structural tectonics 17 .
Zircon U-Pb-Hf-O isotopes. Most zircon grains in the Pengguan migmatites have core-rim structures. The zircon core domains, presumably derived from arc magmatic detritus, show limited variation in their ages, concentrating at~830-870 Ma, with few at~930 Ma (Fig. 1a), and have mantle-like to slightly elevated δ 18 O values (5.3 to 7.4‰) (Fig. 1b). Their ε Hf (t) values vary from -3 to +13, with most being positive, indicative of heterogeneous but generally juvenile sources. Zircon overgrowth rims are slightly younger than the maximum depositional age for each sample, with U − Pb dates generally ranging from~815 Ma to~860 Ma (Supplementary Data 3). The overgrowth rims have significantly higher δ 18 O values (9.3 to 13.3‰) compared with those of core domains, despite their similar ε Hf (t) range (-3 to +8 except one analysis of -9) as core domains (Supplementary Data 2). Zircon grains from the leucosome sample show homogeneous δ 18 O values (11.1 to 13.4‰) with a large range of ε Hf (t) values (-6.9 to +8.4) (Supplementary Data 2). All ε Hf (t) values were calculated to t = 850 Ma in order to facilitate comparison.
Anatexis P-T conditions. We reconstructed the metamorphic P-T conditions for the Pengguan migmatites using Perple_X 6.9.0 (http://www.perplex.ethz.ch). The bulk rock composition of sample 16YX-1-1 was chosen for calculation because this sample clearly documents: (1) mineral-melt interaction; (2) coexistence of minerals (Mn-rich garnet + biotite + muscovite + quartz + plagioclase + K-feldspar + Fe-oxides) and; (3) minor partial melting with no evident melt migration. In the calculated P-T pseudosection, the mineral assemblage of the Pengguan migmatite falls in a narrow domain (domain 1 in Fig. 2a) near the solidus. Using Si pfu in muscovite from 16YX-1-1 (3.08 to 3.14, in moles per formula unit; Supplementary Data 4), which is sensitive to pressure in the K-feldspar + phlogopite + quartz system 19 , we further constrained the anatexis P-T conditions to~670°C and 5.9−8.1 kbar (Fig. 2). The low anatexis temperature is also consistent with the extremely low Th/U ratios of the zircon overgrowth rims (Fig. 1a). At near-solidus temperatures, Th concentration in the melt is largely buffered by Th-rich accessory minerals (such as monazite and allanite) 20 .

Discussion
The Pengguan migmatites are peraluminous with aluminum saturation indices (ASI) of 1.10-1.44 (Supplementary Data 1). Muscovite and peritectic garnets are observed in all samples, indicating peraluminous composition of the protoliths. These migmatites also show geochemical signatures similar to those arc magmatic rocks and forearc sediments from Peninsular Ranges batholith, but distinct from those of MORB and intraplate volcanics ( Fig. 3), suggesting that the protoliths of these migmatites are dominated by arc-related magmatic detritus. This view is also consistent with the observation that the magmatic cores of zircon in the migmatites have very similar age distributions to that of the arc-related magmatic rocks in the study area (Fig. 1a). The absence of pre-Neoproterozoic zircon xenocrysts hints that forearc magmatic detritus may have served as the protoliths of the Pengguan migmatite. The consistent and juvenile Hf isotopes of the zircon cores and overgrowth rims lends further support for the arc origin of the migmatite precursor materials. We thus suggest the Pengguan migmatites documented a process that recycled the arc crust itself, and we refer to this process as endogenic recycling to distinguish it from recycling of oceanic sediments or oceanic crust into subduction zones.
We note that the zircon overgrowth rims have systematically higher δ 18 O values than the magmatic cores ( Fig. 1b), which is indicative of equilibrium with high-δ 18 O anatectic melts during crystallization. High δ 18 O is a diagnostic signature of lowtemperature water-rock interaction at Earth's surface. Thereby the protoliths of the Pengguan migmatites must have undergone some extent of chemical weathering and O isotope exchange at low temperatures before being buried and remelted. Downward infiltration of meteoric water may be another important mechanism to introduce oxygen isotopic heterogeneity to the deep crust. But this mechanism would likely cause water-rock interaction at high temperatures and impart low δ 18 O signature to the rocks, as has been clearly seen in the lower oceanic crust 21 . In addition, the maximum penetration depth of meteoric water ranges from 5 to 18 km (ref. 22 ), which is less than the depth of anatexis (~18 to 24 km) calculated for our migmatites. We thus exclude interaction with downward infiltrated meteoric water as a likely mechanism to explain the high δ 18 O recorded by the zircon rims of this study.
An important question pertains to how the magmatic detritus that had been initially deposited at the surface was transported to the hot deep crust. In magmatic arc settings, recycling of surface rocks has generally been associated with slab subduction. Subducting slabs can directly bring trench sediments to the deep crust or even mantle 5,6 . Subduction erosion has also been recognized as an important mechanism for downward transport of shallow crustal materials 5 . Slab tops are cold (dT/dP = <34°C/ kbar) ( Fig. 2b; estimated from Peacock 23 ) and melting of the sediments deposited at the slab surface is generally considered difficult at crustal depths 24 . Phase equilibrium modeling shows that the Pengguan migmatites formed at~670°C and 5.9−8.1 kbar. These P-T conditions translate into a hot geothermal gradient of 83−114°C /kbar or 25 to 34°C /km, considerably hotter than slab top geothermal gradients but consistent with those seen in arc crust with continuous magmatic inflation 25 (Fig. 2b). This would imply that the Pengguan forearc detrital sediments, shortly after their deposition, were rapidly transported to the deep crust beneath the active arc volcanic front. We suggest the most likely recycling mechanism is via deep thrust faults in the upper continental plate rather than by slab subduction (Fig. 4). Downward flow of wall rocks during magma ascent 26 could be another mechanism in transporting surface materials to the deep crust, but we think it less likely occurred because: (1) wall rock xenoliths were not seen in the plutonic rocks, and (2) vertical flow foliation or lineation is absent in the wall rocks. In compressional magmatic arcs, including mature island arcs and continental arcs, fold The nearly identical age distributions of the zircon cores and overgrowth rims (Fig. 1a) hint at a fast rock cycle, from arc magmatism to water-rock interaction, then to burial and remelting. To estimate the timescale and rate of burial for the Pengguan forearc magmatic detritus, we took the weighted average value of the 50% of youngest U − Pb dates with concordant U − Th−Pb isotopes from zircon core domains as the maximum depositional age, and the weighted average age of zircon rims as the remelting (anatexis) age (Supplementary Data 2 and Fig. 1a). In doing so, we see that the maximum depositional ages are less than 1 to 14 Myr older than the remelting ages for each sample. With the errors of zircon dating taken into account, the magmatic protoliths of the Pengguan forearc sediments must have been exhumed, deposited in a sedimentary basin, and then buried to the depth of crustal anatexis on a~10 Myr timescale. This would imply an efficient burial process with minimum burial rate of 2-3 mm/y.
Our findings point to endogenic recycling as an important mechanism driving internal redistribution of arc crustal materials. The role of such endogenic recycling in the formation of arc crust has been largely overlooked in the past. Radiogenic isotopes are widely employed to constrain crustal recycling processes, but given the short residence time (e.g., 10 Myr), radiogenic isotopes can be completely blind to endogenic recycling. We speculate that extensive endogenic recycling may also generate significant decoupling between radiogenic and stable isotope compositions in recycled materials and their derivative melts, which may further complicate the use of isotope-based proxies in tracing crustal recycling in arc settings.
Rapid endogenic recycling may be facilitated by thrust fault networks. Deep thrust faults may serve as critical transport channels connecting the surface and deep arc crust with ongoing magmatism. And by transporting hydrated surface crustal materials to the deep crust, endogenic recycling enhances the overall differentiation of arc crust. Because large-scale thrust faults necessarily form in compressional settings, efficient endogenic recycling may partly explain why thick arc crust formed in compressional settings (e.g., continental arcs) tend to be more differentiated than thin arc crust formed in extensional settings (e.g., immature island arcs) 29 .

Methods
In situ Zircon isotopes. Zircon grains were separated using conventional density and magnetic techniques, mounted in epoxy resin disk, and polished to expose their internal texture. In situ U-Th-Pb-Hf-O isotope analyses were carried out guided by cathodoluminescence (CL) images and transmitted and reflected photographs. The CL images were taken with a Carl Zeiss Supra 55 field-emission scanning electron microscope (FE-SEM) coupled to a GATAN MonoCL4 detector at the State Key Laboratory for Mineral Deposits Research in Nanjing University (MiDeR-NJU) under following conditions: an accelerating voltage of 3 kV, working distance of 11.5 mm, and aperture size of 30 μm.
Zircon U-Pb isotopes were analyzed using the Cameca IMS-1280HR second ion mass spectrometry (SIMS) at the Guangzhou Institute of Geochemistry, Chinese Academy of Sciences (SKLabBIG GIG CAS) under the following operating conditions: 7 scan cycle,~8 nA primary O 2beam, 20 × 30 µm spot size, and~5400 mass resolving power. Calibration of Pb/U ratios is relative to the primary standard zircon Plešovice 30 and is based on an observed linear relationship between ln ( 206 Pb/ 238 U) and ln ( 238 U 16 O 2 / 238 U) 31 . A long-term uncertainty of 1.5% (1 RSD) for 206 Pb/ 238 U measurements of the standard zircon was propagated to the unknowns, despite that the measured 206 Pb/ 238 U error in a specific session is generally around 1% (1 RSD) or less. U and Th concentrations of unknowns were also calibrated relative to the standard zircon Plešovice, with Th and U concentrations of 78 and 755 ppm, respectively 30 . Measured compositions were corrected for common Pb using non-radiogenic 204 Pb. A secondary standard zircon Qinghu 32 was analyzed as unknown to monitor the reliability of the whole procedure. Data reduction was carried out using the Isoplot/Ex 3 software 33 . Eleven measurements of the Qinghu zircon standard during the course of the study yielded a weighted mean 238 U/ 206 Pb age of 159 ± 2 Ma (MSWD = 0.72), consistent with its recommended value of 159 ± 0.2 Ma (ref. 32 ).
After U-Pb dating, the sample mount was re-ground for~5 µm to ensure any oxygen implanted in zircon surface from the O 2beam used for U-Pb analysis is completely removed. Zircon oxygen isotope analyses were also conducted using SIMS at SKLabBIG GIG CAS. The 133 Cs + primary ion beam was accelerated at 10 kV, with an intensity of~2 nA and focused to an area of ɸ 10 μm on the sample surface and the size of analytical spots is about 20 μm in diameter (10 μm beam diameter +10 μm raster). Oxygen isotopes were measured in multi-collector mode using two off-axis Faraday cups. The measured oxygen isotopic data were corrected for instrumental mass fractionation (IMF) using the Penglai zircon standard 34 (δ 18 O VSMOW = 5.31 ± 0.1‰), which was analyzed once every four unknowns, using sample-standard bracketing method. The internal precision of a single analysis generally was better than 0.1‰ (1σ) for the 18 O/ 16 O ratio. As discussed by Kita et al. 35 and Valley and Kita 36 , internal precision for a single spot (commonly < 0.1‰, 1σ) is not a good index of analytical quality for stable isotope ratios measured by SIMS. Therefore, the external precision, measured by the spot-to-spot reproducibility of repeated analyses of the Penglai standard, 0.30‰ (2σ, n = 24) is adopted for data evaluation. The Qinghu zircon was used as secondary zircon standard, and seventeen measurements of the standard yielded a weighted mean value of δ 18 O = 5.39 ± 0.08‰ (2σ; MSWD = 1.3), consistent with the reported value of 5.4 ± 0.2‰ (2σ) 32 .
Whole-rock geochemistry. Major elements were analyzed using a Thermo ARL9900XP X-ray fluorescence spectrometer (XRF) at the MiDeR-NJU. The analytical precision is generally better than 2% for all elements. Whole-rock rare earth and other trace elements were analyzed using an ICP-MS (Finnigan MAT-Element II) instrument at MiDeR-NJU. Each sample was precisely weighted 30 mg and then was put into a 15 ml Savillex digestion vessel. After being dissolved by HNO3 and the injection of 1 ml 500 ng/ml internal standard Rh solutions, the samples are ready for analyzing. Analytical precision for most elements by ICP-MS is better than 5%. Major and trace element composition data of the migmatite and leucosome samples are provided in Supplementary Data 1.
Mineral composition. The mineral major element compositions were determined using a JEOL 53 JXA-8100 electron probe microanalysis (EPMA) at the MiDeR-NJU. The instrument was operated in wavelength-dispersion mode with a beam diameter of 1-2 µm, a 15 kV accelerating voltage, and a 20 nA beam current. Element peaks and backgrounds were measured for all elements with counting times of 10 and 5. Natural and synthetic standards were used. Detection limits were better than 0.02 wt % for the oxides of most elements. All EPMA data were automatically reduced using the ZAF correction program. Mineral major content results are provided in Supplementary Data 4.