Subduction erosion at convergent margins is a leading mechanism for the destruction (recycling and reworking) of continental crust. But because of the lack of direct evidence, it is not straightforward to identify erosive events and their intensities in fossil subduction zones. The Heilongjiang accretionary complex in northeastern China was formed during the early Mesozoic subduction of the Paleo-Pacific Ocean. We investigated amphibolites from this accretionary complex, whose protoliths (based on whole-rock trace elements and Sr-Nd-Hf isotopes) were mafic continental arc magmatic rocks (255–249 Ma; zircon core U-Pb ages) from the upper plate. Phase equilibria modeling constrained by mineral geochemistry indicates that the amphibolites and their wall rocks were first heated to low granulite facies (750–800 °C, ~7 kbar) at 251–244 Ma (zircon rim U-Pb ages) and then cooled to ~700 °C with increasing pressure (8–9 kbar) before 213–187 Ma (titanite and apatite U-Pb ages). To explain the occurrence of the lower arc crustal lithologies in the accretionary complex and their metamorphic history, we propose that the subducting plate strongly eroded the forearc crust, allowing the plate interface to advance landward and scrape the amphibolites and wall rocks formed under the old arc, which finally were exhumed along the subduction channel and became components of the complex. The case study exemplifies direct petrological evidence of strong subduction erosion occurring in an ancient orogen, thus implying that consumption of the entire forearc crust could occur within only ~50 m.y.

Subduction erosion is the process of scraping materials from the forearc wedge by the subducting plate (von Huene and Scholl, 1991; Stern, 2011). It is prevalent because friction at the plate interface exists ubiquitously, resulting in widespread recycling of crustal rocks into the mantle and involvement of eroded materials in arc magmatism (Straub et al., 2020). Considering that the generation of juvenile rocks by arc magmatism also takes place at subduction zones, identifying the occurrence and intensity of subduction erosion is important for understanding the balance between continental growth and consumption along convergent margins. In fossil orogens, the criteria used to unravel the history of subduction erosion include the absence of the forearc (Huang et al., 2022), landward migration of the continental arc (Goss et al., 2013), and increasingly crust-like characteristics of arc magmatic rocks (Holm et al., 2014; Straub et al., 2015). However, the practical application of these criteria is not straightforward and involves uncertainties due to complex orogenic deformation, the influence of changing subducting plate geometry, and the mantle heterogeneity of the arc source.

Petrological records of eroded materials may allow us to identify past subduction erosion events. Eroded materials are inaccessible if they sink into the mantle (Stern, 2011) or are underplated beneath the overriding plate (“relamination”; Hacker et al., 2011). However, some portions of the eroded materials might enter the subduction channel and then be exhumed at the accretionary complex along with ophiolitic fragments detached from the lower plate. To test this possibility, we carried out a detailed study of the Heilongjiang Complex, a blueschist-facies accretionary complex in northeastern China that developed in response to subduction within the northwestern Paleo-Pacific Ocean domain during the Mesozoic (Wu et al., 2007; Zhou and Li, 2017). We studied meta-mafic blocks (mainly amphibolites and wall rocks) from this complex. By tracking the magmatic and metamorphic history, we were able to reveal that they were parts of the lower continental arc crust, and their final exhumation in the accretionary complex was attributed to vast consumption of the forearc crust by strong subduction erosion.

The Heilongjiang accretionary complex is located along the western margin of the Jiamusi Block in northeastern China, where most outcrops occur in the Luobei, Yilan, and Mudanjiang areas (Fig. 1A). It was formed by the westward subduction of the Mudanjiang Ocean, which belonged to the Paleo-Pacific subduction system (Zhou et al., 2009; Wu et al., 2011; Liu et al., 2021). The Heilongjiang Complex shows typical block-in-matrix structure, with blocks of blueschist, greenschist, granulite, marble, and amphibolite surrounded by foliated meta-sedimentary schist and deformed gneiss. The blueschist and greenschist rocks have geochemical affinities of oceanic island basalt (OIB) and mid-ocean-ridge basalt (MORB), indicating an origin of ancient (Permo-Triassic) oceanic crust (Zhou et al., 2009). The continental arc associated with the subduction zone occurs west of the accretionary complex and is represented by a north-south–trending belt of Permian to Jurassic (ca. 260–170 Ma) calc-alkaline magmatic rocks (Fig. 1) (Wu et al., 2011). The timing of high-pressure and retrograde metamorphism (representing subduction and exhumation, respectively) has been constrained to 200–170 Ma, based on Ar-Ar dating of phengite and hornblende in meta-basaltic and meta-sedimentary rocks of the Heilongjiang Complex (Ge et al., 2017; Aouizerat et al., 2020).

Based on the metamorphic grades, we subdivide the meta-mafic rock samples from the Heilongjiang Complex in Luobei and Yilan into two groups. The first group consists of blueschist, representative of low-temperature and high-pressure metamorphism, i.e., low thermobaric ratio (temperature/pressure [T/P]), and greenschist. In Luobei, greenschist occurs as lenses and interlayers in mica-quartz schist. In Yilan, the occurrence of greenschist is associated with blocks within mica schist and rinds surrounding blueschist blocks (Figs. S1 and S2 in the Supplemental Material1). The second group of meta-mafic rocks are metamorphosed to high amphibolite to low granulite facies. These samples consist of garnet- ([Grt-])amphibolite, granulite, and Grt-plagiogneiss. Most (Grt-)amphibolite and granulite shows migmatization to various degrees, evidenced by leucosome patches or veins (Fig. S1) and melt pockets (pseudomorphs of quartz ± plagioclase) (Fig. S3). Outcrops of (Grt-)amphibolite and Grt-plagiogneiss in Luobei show weakly to highly deformed blocks in a matrix of gneiss and metapelite; the rocks are intersected by felsic veins (Figs. S1 and S4). In Yilan, amphibolites are mylonitic (showing S-C fabrics) and the granulite is weakly deformed (Figs. S1 and S3).

Geochemical and isotopic analyses reveal an oceanic crustal nature for the first group of samples and a continental arc affinity for the second. The blueschist and greenschist rocks have negative anomalies of large-ion lithophile elements (LILEs) and weakly positive anomalies of Nb, Ta, and Ti (Figs. 2A and 2B), similar to the patterns of OIB and/or MORB. Positive zircon εHf(t) (7.9–13.9) and whole-rock εNd(t)values (6.6–8.4) as well as low (87Sr/86Sr)i values (initial ratios of the protolith; 0.704–0.708) suggest a depleted mantle source (Figs. 2E and 2F). The protoliths of the second group of samples are mafic to intermediate in composition, belonging to the medium to high potassium calc-alkaline series (Fig. S5). They are enriched in LILEs and light rare earth elements (LREEs) and depleted in Nb, Ta, and Ti (Figs. 2C and 2D). Integrating the negative zircon εHf(t) (−5.9 to 0.3) and whole-rock εNd(t) values (−7.3 to −3.9) and the high (87Sr/86Sr)i values (0.707–0.712), the protoliths of the second group are interpreted as being derived from enriched continental arc magmas in the upper plate. The difference in the source setting between the two groups is further verified by the zircon Sc/Yb ratios (Fig. S14), with low values suggesting a plume-influenced source and MORB setting for the greenschist and blueschist zircons, and high values indicating a continental arc environment for the (Grt-)amphibolite and granulite (Grimes et al., 2015).

Oceanic crustal materials that have experienced low-T/P metamorphism can be detached from the lower plate and exhumed in accretionary complexes (Agard et al., 2009). We suggest that the first group of samples and other blueschist and greenschist rocks from the Heilongjiang Complex (Zhou et al., 2009) may represent such a scenario. Zircons from our samples show fluid-metasomatized textures in rims (Fig. S6), possibly indicating low-T hydrothermal alteration in the subduction channel (Rubatto, 2017). The cores show magmatic oscillatory zonation and yield concordant ages ranging from 316 ± 1 Ma to 256 ± 2 Ma in different samples (Figs. S6 and S7), consistent with the reported protolith ages (356–213 Ma) (Zhou et al., 2010; Ge et al., 2017; Aouizerat et al., 2020). These ages represent the age of the oceanic crust, which was formed at least from the early Carboniferous to the Late Triassic.

The samples from the second group show evidence that the rocks experienced a more complicated metamorphic history before being involved in the subduction channel. Most zircons from these samples have igneous cores with typical magmatic oscillatory zonation and metamorphic overgrowth rims with weak zoning (Fig. S6). The cores of zircons from (Grt-)amphibolite and granulite yielded concordant ages from 255 ± 1 to 249 ± 1 Ma, indicating that the protoliths (continental arc magmas) were produced during the late Permian and earliest Triassic. The rims have slightly younger ages from 252 ± 1 to 245 ± 2 Ma. In each grain, the core-rim ages differ by 1–10 m.y. (Fig. S8), indicating that metamorphic overgrowth occurred soon after magmatic zircon crystallization. This overgrowth was most likely caused by magmatic heating from further intrusion of arc magma into the lower crust. The petrological evidence of partial melting (Fig. S3) demonstrates that the peak metamorphic T exceeded the wet solidus (~700 °C; Fig. S9). The phase equilibria modeling constrained by mineral geochemical analyses of Grt-amphibolite samples suggests a peak T of 750–800 °C at 7–8 kbar (M2 in Fig. 3), a condition that more likely occurred in the arc setting rather than the cold subduction channel (Penniston-Dorland et al., 2015; Wang et al., 2023). The high-T metamorphic event is further evidenced by the Grt-plagiogneiss sample from Luobei. Some igneous zircons from this sample yielded concordant ages of 1.0–0.9 Ga (Fig. S8), consistent with the age of the continental basement of the upper plate (Zhou and Li, 2017), thus indicating that the protolith belonged to the wall rock of the mafic arc magmatic rocks. U-Pb ages of most other zircon cores are discordant in the concordia diagram (Fig. S8), indicating that the Neoproterozoic zircons lost part of the radiogenic Pb at ca. 250 Ma (the lower intercept age). This age is identical to the concordant age of the zircon rims (249 ± 1 Ma) from the same sample, suggesting that the wall rock was heated by the arc magmas at ca. 250 Ma. The phase equilibria modeling suggests that it was also heated to ~765 °C at ~7.2 kbar, consistent with the peak condition of the (Grt-)amphibolite. Because the coeval arc-like granitoids are now exposed in the eastern part of the Songnen Block (Fig. 1B), we infer that the meta-mafic rocks in the Heilongjiang Complex originated from the lower crust of the continental arc in the upper plate.

Seismic imaging of present-day subduction zones reveals that the horizontal distance between the arc axis and slab surface at 25–30 km depth (the formation depth of our samples) ranges from 95 to 180 km (Hayes et al., 2018) (Fig. S10). Assuming similar geometry for our study area, ~95–180 km of forearc crust must have been removed by the downgoing slab to have allowed for incorporation of lower arc rocks and their wall rocks from the upper plate into the subduction channel. This scenario can be explained by subduction erosion (Fig. 4A). The destruction of the forearc crust predicts an inland migration of the subduction interface and magmatic arc (Fig. 4B), which is recognized in the eastern Songnen Block by the westward younging of the Permian to Jurassic (ca. 260–170 Ma) magmatic rocks (Fig. 1B).

At 250 Ma, after peak T conditions (M2 in Fig. 3) of the meta-mafic rocks, P increased from ~7 kbar to 8–9 kbar (Fig. 3), as recorded by the garnet rims. Crustal thickening near the arc axis cannot explain the increase in P because such a process would also result in an increase in T due to the higher thermal flux in the arc. Our samples instead show evidence of cooling. We therefore suggest that the P increase likely reflects either enhanced compression exerted by the approaching subduction interface or a process whereby rocks were dragged down into the subduction channel (Fig. 4B).

The low U-Pb closure temperatures of apatite (Ap; ~450–550 °C; Cherniak et al., 1991) and titanite (Ttn; ~550–650 °C; Cherniak, 1993) allow us to discern the timing of exhumation. Ap and Ttn from the low-T rocks (our first group of samples) failed to yield concordant U-Pb ages due to their very low U concentrations. But Ap in high-T Grt-amphibolite, granulite, and Grt-plagiogneiss samples of our second group yield lower-intercept ages in the Tera-Wasserburg diagram ranging from 194 ± 3 Ma to 213 ± 17 Ma (Fig. S11), and the three samples yield Ttn ages of 191 ± 1 Ma, 206 ± 2 Ma, and 187 ± 3 Ma, respectively (Fig. S12). Within error, these results are similar to the published phengite and hornblende Ar-Ar (Zhou et al., 2009; Ge et al., 2017) and rutile U-Pb ages (Dong et al., 2019) from blueschist, greenschist, mica schist, and amphibolite of the Heilongjiang Complex (ca. 210–180 Ma). We therefore suggest that the Ap and Ttn U-Pb ages (ca. 213–187 Ma) probably record the time when the lower crustal arc rocks were incorporated in the subduction channel. Subsequently, these rocks were exhumed along the subduction channel and finally emplaced within the Heilongjiang accretionary complex (Fig. 4C).

Although the bulk of eroded continental crustal materials tends to get recycled into the mantle, remnants captured by the Heilongjiang Complex, as demonstrated in this study, provide crucial information on ancient subduction erosional events. The time interval between the high-T metamorphism in the lower crustal continental arc (ca. 250 Ma by zircon rim ages of the Grt-amphibolite and the Grt-plagiogneiss) and the cooling process in the subduction channel (ca. 200 Ma by averaging the Ap and Ttn ages) indicates that subduction erosion can remove a tract of 95–180 km of forearc crust (based on present-day forearc widths) over the course of only ~50 m.y. Assuming an original crustal thickness of 30 km in the upper plate, the erosion rate in our study area is 60–110 km3/km/m.y. along the trench. Such values are in agreement with erosion rate estimations based on forearc subsidence in active erosive subduction zones (Chile and Japan; 30–120 km3/km/m.y.) (Scholl and von Huene, 2007, 2009; Stern, 2011) and are also comparable to mantle-derived melt production rates for continental arcs (140–215 km3/km/m.y. during magmatic flare ups and ≤15 km3/km/m.y. during magmatic lulls) (Chapman et al., 2021). It appears, therefore, that while additional crustal material is produced via mantle input in arcs, the continental margins are consumed at a similar rate by subduction erosion. Accordingly, our results imply little to no net crustal growth in such arc settings. Considering that more than half of modern subduction zones are erosive, this may help explain why the rate of crustal growth decreased after the onset of modern plate tectonics (Cawood et al., 2013).

1Supplemental Material. Figures S1–S14, Tables S1–S10, and methods. Please visit to access the supplemental material; contact with any questions.

This work was funded by National Natural Science Foundation of China (NSFC) grants 41888101 and 42272252. We thank editor Rob Strachan, Anthony Ramírez-Salazar, and one anonymous reviewer for constructive comments that improved the science.

Gold Open Access: This paper is published under the terms of the CC-BY license.