Magmatic activity in the syn-collision stage is key for net crustal growth. To understand the mechanism of accretion–differentiation and compositional change of the continental crust, it is important to focus on the magmatic activity during the syn-collision stage. Early Eocene mafic–ultramafic rock assemblages found in the western part of the Tengchong Block resulted from a continuous series of arc magmatic evolution, thoroughly recording the continental arc magmatic system during the subduction of the Neo-Tethys Ocean and syn-collision of the Indian-Asian continents. Early Eocene hornblende gabbro–diorite in the Tengchong Block formed at 53 Ma, and the primitive magma was derived from an enriched mantle source due to the enriched Nd–Hf isotopes. The amphibole and biotite thermobarometer measurements indicate that the mafic magma reservoirs in the Tengchong Block occurred at a mid-upper crust. Petrography, amphibole Fe/Mg exchange coefficient (KD), Rayleigh fractionation, and equilibrium melt calculation indicate that the Early Eocene hornblende gabbro–diorite in the Tengchong Block was created due to plagioclase-dominated accumulation at the mid-upper crust level. Based on the calculation, the corresponding amphibole equilibrium melt is more silicic (dacitic–rhyolitic in composition) than the bulk rocks, indicating a more evolved composition in the mid-upper crust. Three types of plagioclases reveal the multi-recharging and dissolution–reprecipitation promoting the further evolution of these mafic rocks. Therefore, this study concludes that magma recharge and plagioclase-dominated accumulation processes may be important mechanisms for the formation and evolution of mafic magma and the further crustal differentiation at the mid-upper crust level in a continental margin arc.

Arc is an important region for the growth and differentiation of the continental crust, and arc igneous rock is ideal for understanding the evolution of magma reservoirs and the transcrustal magmatic system in depth. Increasing level of attention is being given to the evolution process and the corresponding mechanism of magmatism in the continental arc [1-6]. However, the andesitic component of the continental crust cannot be completely derived from the single-stage melting of mantle-derived magmas; it requires additional magmatic evolutionary processes. These processes include gradual differentiation of mantle-derived basaltic magmas by fractional crystallization within the upper mantle and lower crust, and/or partial melting of the early mafic crust [1, 2, 7-10]. Recently, studies on the construction and differentiation of the continental crust have focused on the genesis of granite [11-16], whereas the vertical multistage differentiated evolution of mantle magma within the continental crust can be considered as the fundamental process of magmatic evolution from the mafic lower crust to the felsic upper crust [7, 8, 17-19]. The pyroxene-/amphibole-rich mafic intrusive and accumulated rocks, such as the hornblendite, pyroxenite, hornblende gabbro, and diorite, record the evolution and characteristics of arc magma, such as formation environment, cumulative properties, and recharge of magma [17, 20, 21]. The Gangdese arc in southern Tibet, the Kohistan arc in Pakistan, and the Cordilleran arc in the western border of South and North America have been studied in detail [2, 22, 23]. The results show that the mafic–ultramafic rocks are generally derived from the lithospheric mantle association with fluid and/or melt metasomatism in the subduction zone and primitive mantle-derived basaltic magma differentiation induced by slab-pull forces. These processes form mafic–ultramafic cumulate rock, which is, in turn, recycled back into the mantle through eclogitization and/or delamination, resulting in the transformation of the bulk continental crust to andesitic composition [9, 24-30]. Although subduction–accretion is the dominant mode of the growth of continental crust, subduction can lead to crustal accretion as well as crustal extinction, resulting in no net growth of the continental crust. The syn-collision stage, characterized by magmatic activity, is the key period for crustal net growth [2, 31]. Therefore, to understand the mechanism of accretion–differentiation and compositional change of the continental crust, focusing on the magmatic activity during the syn-collision stage is important. In addition, most studies on the accumulation of mafic rock consider this accumulation at the arc root; however, a few studies have focused on this accumulation at the mid-upper crust level. Furthermore, while investigating the compositional transformation mechanism of the continental crust, the existence, differentiation, and evolution of arc magmatic mafic rocks in the mid-upper crust may shed important light on this mechanism.

The genesis of mafic rocks intruding into the crust is considerably controversial. Some scientists believe that mantle-derived hydrous basaltic magma formed mafic cumulates by fractional crystallization [32-35], whereas others believe that the mafic rocks in the crust are the residue of partial melting, crustal contamination, and/or magmatic mixing within the arc crust [1, 36]. The transcrustal magmatic system of arc magmatic evolution indicates that mantle-derived magmas are stratified through crystal accumulation and melt extraction from magma reservoirs connected at different depths, leading to evolution from mafic to silicic composition, conceivably involving assimilate-contamination and other processes during the evolution [5, 37, 38]. Experimental petrology and natural studies also indicate that hydrous mafic–ultramafic magma is formed owing to polybaric differentiation by mineral assemblages at different depths in the crust [33, 39]. Therefore, systematic investigation of the genesis of the mafic intrusion in the continental margin arc at the syn-collision stage is important to reveal the characteristics of the magmatic reservoir and to further identify magmatic evolution and vertical differentiation in the continental arc crust. Typical arc mafic rock at the syn-collision stage can be considered an ideal object for studying magma evolution and differentiation in the continental arc crust.

The Tengchong Block is a tilted continental margin arc associated with the eastward subduction of the Neo-Tethyan Ocean [27, 28, 40-43]. The Meso-Cenozoic magmatic rocks are considered to be the southeastern extension of the Gangdese magmatic arc, and a large number of complete Early Eocene mafic–ultramafic rock assemblages found in the western part of the Tengchong Block are typical continental margin arc intrusive rocks at the syn-collision stage (50, 55 Ma) related to the Indo-Asian continental collision [13, 42, 43]. The exposed rock types are primarily hornblendite, hornblende gabbro, and diorite, and different lithologies represent the products of the evolution of the mafic magmatic system at different depths. Previous studies on the origin of different mafic rocks have attributed the differences in the source region properties caused by different metasomatic media in subduction zones to differences among these mafic rocks [42, 43]. However, the evolution process, especially differentiation, within the crust also plays an important role in the petrogenesis and geochemical properties of mafic rocks. Therefore, the Longpen hornblende gabbro–diorite in the western part of the Tengchong Block was selected as the research object of this study. Through petrography, chronology, whole-rock geochemistry, Sr–Nd isotopic composition, Lu–Hf isotope of zircon, Sr isotope of plagioclase, and major and trace element analyses of major rock-forming minerals, the genesis and evolution of these mafic rocks and their role in the arc magmatic system are discussed. The results indicated that the Early Eocene hornblende gabbro–diorite in the Tengchong Block is a product of evolved mantle-derived magma in the continental arc magmatic system; the results showed that it developed in the middle and upper crust, undergoing plagioclase-dominated accumulation, which could be an effective process involved in the chemical differentiation within the middle and upper continental arc crust during the syn-collision stage.

The Tengchong Block is located on the southwestern margin of the Sanjiang Tethyan orogenic belt, extending to the southeastern margin of the Lhasa Block on the Tibetan Plateau [44-47]. It is bounded by the Mogok metamorphic belt and the Sagaing fault from the west Burma Block to the west and by the Gaoligong belt from the Baoshan block to the east (Figure 1(a)). The Tengchong Block has a Mesoproterozoic–Neoproterozoic metamorphic basement—the Gaoligong Group, the upper part of which is metamorphic, siliceous, and schistose, and the lower part contains a small amount of mafic–ultramafic granulite [46, 47]. The Tengchong Block was a member of the Sibumasu Block that formed the northern margin of the Gondwana Continent in the early Palaeozoic, it collage with the Eurasia Continent during the late Mesozoic after the break-up of the Gondwana Continent [42, 47]. The Late Cretaceous–Early Eocene Neo-Tethys Ocean underwent eastward subduction and closure, resulting in the development of a substantial amount of mafic–felsic arc magmatism in the Yingjiang–Longchuan area of the Tengchong Block [27, 28, 48-50].

For this study, samples were collected from the Longpen area located in the western part of the Tengchong Block. The rock is exposed as small stocks adjacent to the Early Eocene Xima–Tongbiguan granitic pluton (Figure 1(b)). The lithology primarily comprises hornblende gabbro and diorite (Figure 2(a)–(c)), with medium- to fine-grained texture. The petrography showed a cumulate structure dominated by plagioclase and amphibole. The mineral assemblages are comprised of plagioclases (45–55%), biotites (15–25%), amphiboles (10–15%), minor amounts of interstitial quartz (–5%), and alkaline feldspar (–5%) (Figures 2 and 3). The accessory minerals (<5%) included magnetites, apatites, zircons, orthites, etc. (Figure 2(d) and (e)). The postcumulus crystals, formed by the interstitial melt at a later stage, accounted for ~45% of the rock, which is characteristic of orthocumulates. The plagioclases in the sample are classified into three types—Type I and Type II as euhedral–subhedral crystals and Type III as an interstitial phase (hereinafter referred to as “interstitial phase plagioclase”)—based on their different particle sizes and structural characteristics. The larger plagioclase crystals are stumpy, with lengths of 0.5–2 mm, forming a mineral framework texture within the whole rock, accompanied by certain directional arrangement characteristics (Figures 2(c) and 3). Some of the larger plagioclase crystals are predominantly in point contact with each other, whereas others are in contact with smaller interstitial plagioclase along common boundaries because of late growth. The Type I and II plagioclases generally show compositional zoning, patchy texture, sieve texture or dissolved texture in the core, and/or sieve texture in the mantle (Figure 2(d)–(f)). The small and interstitial plagioclases are subhedral–anhedral crystals, with weak component zoning and lengths of 0.01 to 0.1 mm (Figure 2(d)). Amphiboles have euhedral–subhedral granular structures that appear olive-green or brown under unipolar light. Some amphiboles occur as typical simple twins with a grain size in the range of 0.1–0.5 mm (Figure 2(e)). No obvious ductile deformation was observed in amphiboles, consistent with relatively uniform composition, indicating a stable crystallization environment. Amphibole polycrystals often occur around plagioclase crystals (Figure 2(g)), with single-particle sizes of <0.01 mm. Biotites present a subhedral–anhedral foliated in shape, appearing brown-yellow under unipolar light and yellow-green to light pink under orthogonal light, with particle sizes ranging from 0.1 to 2 mm (Figure 2(c) and (d)). Some amphiboles and biotites having a small crystalline form, with sizes of <0.01 mm, are disseminated and enclosed in plagioclases (Figure 2(e) and (f)).

3.1. Major and Trace Element Analyses

Major, trace, and rare earth element (REE) analysis tests were conducted at Northwestern University’s Continental Dynamics Laboratory, where any weathered shells were removed to 200 mesh before elemental and isotope analyses were performed using an agate mill. The main elements were determined using a wavelength-dispersive X-ray fluorescence spectrometer (Rigaku RIX2100), spectrometer, with an analytical accuracy and accuracy of >2%. The trace element and REE tests were performed using inductively coupled plasma mass spectrometry (ICP-MS) with an analytical accuracy of 10%. The analytical standards were derived from the United States Geological Survey and China National Rock Standards (BCR-2, GSR-1, and GSR-3).

3.2. Zircon U–Pb Chronology Analyses

Zircon U-Pb isotope dating and trace element content were simultaneously analyzed using LA-ICP-MS (Agilent7900) at the State Key Laboratory of Continental Dynamics, Department of Geology, Northwest University. In this study, the spot size and frequency of the laser were set to 30 µm and 6 Hz, respectively. Zircon 91500 and glass NIST610 were used as external standards for U–Pb dating and trace element calibration, respectively. Each analysis incorporated a background acquisition of ~20–30 seconds followed by 50 seconds of data acquisition from the sample. The internal morphology and structure of the representative zircons were examined by cathodoluminescence (CL). The Excel-based software ICPMSDataCal was employed to perform offline selection and integration of background and analyzed signals, time-drift correction, and quantitative calibration for trace element analysis and U–Pb dating. Concordia diagrams and weighted mean calculations were performed using Isoplot/Ex_ver3. Details of experimental principles and test procedures can be found in a paper by Bao et al. [51].

3.3. Whole-Rock Sr–Nd Isotope Analyses

The whole-rock Sr–Nd isotope data were analyzed by Nu Plasma HR Multi-collector ICP-MS at Wuhan Sample Solution Analytical Technology Co., Ltd., Wuhan, China. Sr and Nd isotope fractionation were corrected to 87Sr/86Sr = 0.1194 and 146Nd/144Nd = 0.7219, respectively. During the analysis period, the mean values of the NIST SRM 987 standard were 87Sr/86Sr = 0.710250 ± 12 (2σ, n = 15), and the mean values of the La Jolla standard were 146Nd/144Nd = 0.511859 ± 6 (2σ, n = 20). Specific test procedures and analysis methods can be found in a previous study [52].

3.4. Zircon In-Situ Lu–Hf Isotope Analyses

In situ, Lu–Hf isotope ratio analysis was conducted using a Neptune Plus MC-ICP-MS (Thermo Fisher Scientific, Germany) in combination with a Geolas HD excimer ArF laser ablation system (Coherent, Göttingen, Germany) that was hosted at the State Key Laboratory of Continental Dynamics, Department of Geology, Northwest University. The laser ablation rate was 6 Hz and the beam spot size was 30 µm. The 176Hf/177Hf ratio of standard zircon (91,500) was 0.282294 ± 15 (2σ) and the 176Lu/177Hf ratio was 0.00031. For the above test analysis, refer to the aforementioned experimental principle; the test process can be found in a previous study [53]. Later, data processing was completed using the Isoplot3.0 software [54].

3.5. Mineral Chemistry Analyses

The JXA-8230 electron probe microanalyzer (EPMA) from Japan’s JOEL Company and LA-ICP-MS were employed at the Ministry of Education, Key Laboratory of Western China’s Mineral Resources and Geological Engineering, School of Earth Science & Resources, Chang’an University for mineral major and trace element analyses of minerals. The operating conditions included an acceleration voltage of 15 kV, a beam current of 10 nA, and a beam diameter of 1 µm. Natural and synthetic microprobe standards of major elements were supplied by SPI, including jadeite for Si, Al, and Na; diopside for Ca; olivine for Mg; sanidine for K; hematite for Fe; rhodonite for Mn; and rutile for Ti. The United States Geological Survey and China National Rock Standards (BHVO-2G, BIR-2G, and BCR-2G) were adopted herein as the standard for trace element analyses. Subsequent data processing and plotting were completed by using the ICPMS-DataCal software [55, 56] and Isoplot3.0 software [53].

3.6. TESCAN-Integrated Mineral Analyzer

At Xi’an Mineral Spectrum Geological Exploration Technology Co., Ltd., TESCAN automatic mineral analysis system (TIMA) was used to conduct integrated mineral analysis for thin section scanning. After establishing a mineralogical database and comparing the results of different analysis models, a high-resolution cartographic model was chosen for the analysis. The acceleration voltage was 25 keV, the probe current was 10.56 mA, the beam intensity was 19.07 nA, and the beam spot size was 142.51 nm. For better mineral classification, the scan was set at a dot spacing of 9 µm, a working distance of 15 mm, a pixel size of 3 µm, an X-ray count of 1200, and a brightness between 15% and 100%. The total scanning time for a sample was ~5 hours, and ~7000 counts were collected for each analyzed sample. The TIMA2.3 software was employed for data processing [53].

3.7. In Situ Sr Isotope Analyses

The Sr isotope ratios of plagioclases were measured using the Neptune Plus MC-ICP-MS (Thermo Fisher Scientific, Bremen, Germany) in combination with the Geolas HD excimer ArF laser ablation system (Coherent, Göttingen, Germany) at the Wuhan SampleSolution Analytical Technology Co., Ltd, Hubei, China. In the laser ablation system, helium was used as the carrier gas for the ablation cell. For a single laser spot ablation, the spot diameter ranged from 60 to 160 µm, depending on the Sr signal intensity. The beam spot size was 90 µm, the frequency was 8 Hz, and the energy intensity was 10 J/cm2. All data reduction for the MC-ICP-MS analysis of Sr isotope ratios was conducted using the Iso-Compass software [57], and the standard sample used was YG4301 [58, 59].

4.1. Zircon U-Pb Dating

Here, two mafic rock samples of mafic rocks were selected to analyze zircon morphology, U–Pb chronology, and trace elements. The results are presented in online supplementary Table S1 and Figure 4. Most zircons in the hornblende gabbro sample (XM18-86) are subhedral, short columnar, and light brown or colorless, with lengths of 100–200 µm and aspect ratios of 1:1–2:1. The CL images show distinctly broad and gentle magmatic zoning or sectoral zoning (Figure 4(a)), suggesting a magmatic origin. Twenty-seven spots were selected for U-Pb isotope and trace element analyses (Figure 4), yielding the 206Pb/238U ages ranging from 50.1 to 56.7 Ma, with a weighted mean age of 52.8 ± 0.8 Ma (n = 27, MSWD = 4.50). The Th contents were 29–2060 ppm, and the U contents were 186–4974 ppm, with Th/U ratios ranging from 0.10 to 1.39. The zircon REE pattern exhibited depletion in light REE (LREE) and strongly strong enrichment in heavy REE (HREE) (Figure 4(c) and (f)), with negative Eu anomalies (Eu/Eu* = 0.10–0.40).

The zircon in the diorite sample (XM18-101) is also euhedral–subhedral crystal, short columnar, colorless, and transparent, with lengths of 150–400 µm, and aspect ratios of 1:1 to 3:1. The CL images of zircon revealed wide compositional zoning (Figure 4(a) and (d)), indicating a magmatic origin for the sample. Thirty-three points were selected from the diorite for U-Pb isotope and trace element analyses (Figure 4(d) and (e)). The 206Pb/238U ages obtained were 55.4–50.9 Ma, with a weighted mean age of 53.8 ± 0.4 Ma (n = 33, MSWD = 1.19). The analyzed zircon points yielded Th contents ranging from 72 to 1578 ppm, U contents ranging from 154 to 1028 ppm, and Th/U ratios of 0.29–1.53. The REE distribution pattern in zircon showed depletion in LREE and enrichment in HREE (Figure 4(c) and (f)), with negative Eu anomalies (Eu/Eu* = 0.45–0.56).

4.2. Major and Trace Elements

Eight samples were analyzed to determine the contents of major and trace elements; the results are presented in online supplementary Table S1. The samples showed low SiO2 (48.87, 54.52 wt%), and MgO (2.83, 3.20 wt%) contents but high Al2O3 (17.66, 20.52 wt%), Fe2O3t (6.84, 10.45 wt%), and K2O (1.90, 2.77 wt%) contents. The total alkali (Na2O + K2O) contents were 1.25–6.19 wt%, and the Na2O/K2O ratios were 1.23–1.63. The samples exhibited low Mg# values (41–49). On the TAS diagram, the samples fell into fields of monzodiorite, monzogabbro, and gabbroicdiorite (Figure 5). In the primitive mantle normalized spider diagram, the samples showed enrichment in large ion lithophile elements (LILE) (such as Pb, Rb, Th, and Nd) and depletion in high field strength elements (HFSE) (such as Nb, Ta, P, and Ti) (Figure 6(a)), similar to modern arc magmas along the Pacific subduction zone [60-62]. In the chondrite-normalized REE diagram, all samples exhibited a right-leaning REE pattern, enriched in light REEs (LREEs) (Figure 6(b)), with high (La/Yb)N values (7.61, 19.01) and negative Eu anomalies (Eu/Eu* = 0.62–0.90).

4.3. Whole-Rock Sr–Nd and Zircon In Situ Lu–Hf Isotope

Herein, whole-rock Sr–Nd isotope analyses were performed on four samples, including three hornblende gabbro samples and one diorite sample. During the calculation of the initial isotope ratios, all samples were based on the zircon U–Pb chronology analysis results at 53 Ma. The isotopic results are presented in online supplementary Table S2 and Figure 7, showing that 87Rb/86Sr = 0.403–0.422, and the initial 87Sr/86Sr(t) = 0.708787–0.709441, 147Sm/144Nd = 0.1065–0.1296, 143Nd/144Nd = 0.51287–0.512359, and εNd(t) = −6.3 to −4.8, corresponding to two-stage model ages of 1.07–1.18 Ga.

In situ, Lu–Hf isotope analyses were performed on 39 zircon samples. The results are presented in online supplementary Table S3 and Figure 8. The corresponding zircon U-Pb ages were used to calculate the initial Hf isotope ratios. The analysis results indicated that 176Hf/177Hf = 0.282383–0.282642 and εHf(t) = −12.59 to −3.54, corresponding to two-stage model ages of 1.35–1.92 Ga.

4.4. Mineral Chemistry and In Situ Sr Isotope Analyses

Major elements were determined from amphiboles, plagioclases, and biotites in samples. The results are presented in online supplementary Table S5-S8. Three plagioclase particles were selected for in situ Sr isotope analyses, and the results are presented in online supplementary Table S1 and Figure 9.

Type I plagioclase are euhedral–subhedral crystals, with lengths range of 0.8–1 mm and aspect ratios of 1:1–1.5:1. They are characterized by a brighter core and irregular dark rims in the backscatter images (BSE; Figure 10(a1) and (b1)). The core has a sieve texture and the rim has an overgrowth edge (Rim2) with a discontinuous interface, showing an abrupt composition between the core and the rim. Type I plagioclase has high An values ranging from 69.4 to 83.0 in the core belonging to bytownite, and relatively lower An values in the rim, decreasing to 42.4–55.2, belonging to andesine-labradorite; the lowest An values are observed in the overgrowth edge, decreasing to 29.9–38.4, belonging to andesine–oligoclase (Figure 11(a)).

The shape of Type II plagioclase particles is also euhedral–subhedral, with lengths ranging from 0.8 to 2 mm and aspect ratios of 1.5:1–3:1. It is characterized by a typical core–mantle–rim texture, showing a bright core color in the BSE image with abundant amphibole and biotite inclusions. A sieve texture is observed within the mantle and a dark thin layer is observed at the rim (Figure 10(c1) and (d1)). The An values at the core of Type II plagioclase in the hornblende gabbro are found to be 25.6–30.8, belonging to oligoclase-andesine; the An values at the mantle increase to 49.8–62.2, belonging to labradorite, while the An values at the rim gradually decrease to 30.3–42.6, belonging to andesine; the lowest An values are observed to be in the overgrowth edge, ranging from 20.1 to 22.3, belonging to oligoclase (Figure 11(a)). The An values at the core of Type II plagioclase in diorite range from 44.2 to 49.1, belonging to andesine, and the An values at the mantle increase from 72.4 to 82.0, belonging to bytownite. The An values at the rim sharply decrease from 45.1 to 57.6, belonging to andesine, and the lowest An values are observed in the overgrowth edge, ranging from 30.2 to 33.5, belonging to oligoclase (Figure 11(a)). The changes in An values show the characteristics of the “M” type (Figure 10(c2) and (d2)).

Type III interstitial phase plagioclase with a subhedral to anhedral shape is characterized by irregular banded zones at the rim (Figure 10(e1)). The BSE image shows a bright core and a dark rim. The An values at the core of Type III plagioclase range from 45.0 to 61.9, belonging to the andesine–labradorite; the An values at the rim decrease in a range of 33.4–35.2, belonging to andesine (Figure 10(a)).

The results of in situ Sr isotope analyses of plagioclase show that the Sr isotopic composition from core to mantle, and subsequently to the rim of the large crystal plagioclase is uniform (Figure 9), with the overall trend tending to be flat, with 87Sr/86Sr = 0.708431–0.709377, and Rb/Sr = 0.000266–0.074223, similar to the initial 87Sr/86Sr of the whole rock (0.708787, 0.709441).

Regarding amphiboles, all samples show low Mg (MgO = 7.01–11.07 wt%) and high Fe (FeOt = 16.23–23.34 wt%) contents. According to the classification and nomenclature of amphiboles in a previous study [63], the main amphibole in the hornblende gabbro samples is hastingsite (VIAl < Fe3+), with a small portion being ferrohornblende and ferrotschermakite (Figure 11(c)), showing Mg# = 0.35–0.49, Altot = 1.39–2.47 a.p.f.u., total alkali content (Na+K)A = 0.32–0.70 a.p.f.u., higher Ti (0.13, 0.27 a.p.f.u.), lower Cr (–14.89ppm), and Ni (–3.96ppm). The composition of amphibole in diorite samples is relatively complex (Figure 11c), including magnesiohornblende, tschermakite, ferrohornblende, and ferrotschermakite. The remaining components range from magnesio hastingsite to hastingsite (VIAl < Fe3+). The results of individual minerals of amphibole show moderate Mg# = 0.40–0.61, with average Altot of 1.77 a.p.f.u., Altot = 0.89–3.55 a.p.f.u., and total alkali content (Na + K)A = 0.13–0.62 a.p.f.u..

The spider diagrams of trace elements in amphiboles show Sr, Nd, Zr, and Ti depletions. The REE patterns of chondrite-normalized show depletion in LREEs (LaN/SmN = 0.19–1.37) and heavy REEs (HREEs) (GdN/YbN = 1.25–5.92) relative to the medium REEs (MREEs;Figure 12). Most samples exhibit considerable negative Eu anomalies (Eu/Eu* = 0.26–0.80), indicating that plagioclase had already crystallized in large quantities before the crystallization of amphiboles. Only two measurement points located in the core of the amphibole show no Eu or positive Eu anomalies (Figure 12(a)), with lower ∑REE, Nb, Ta, and Ti values compared with those of other amphiboles, indicating that the amphibole core may have formed before the crystallization of plagioclase.

Biotites, primarily comprise ferro-biotite (Figure 11(b)), with an SiO2 content ranging from 33.13 to 38.05 wt%, a relatively high FeOt content of 18.14–24.41 wt%, TiO2 content of 1.27–4.22 wt%, MnO content of 0.18–0.39 wt%, and relatively low MgO content (8.45, 10.55 wt%). The CaO content is below the limit of detection. The FeO, MgO, and K2O contents of all biotites from the core to the rim are relatively uniform, and the overall trend is flat, indicating a primary magmatic origin with a low degree of chloritization and sericitization caused by later magmatic-hydrothermal alteration.

5.1. Characteristics of the Magma Source and Storage Conditions of Magma Reservoirs

The Tengchong Block is located in the Sanjiang Tethyan Structure Domain, a southeastern extension of the Gangdese magmatic belt in southern Tibet, where a substantial amount of Mesozoic and Cenozoic igneous rocks are exposed (Figure 1(a)). A considerable amount of Late Cretaceous–Early Eocene continental arc magmatic rocks developed owing to the subduction of the Neo-Tethys Ocean and the Indo-Asian continental collision [42, 43, 45, 48, 49]. The ages of mafic plutons in the Nabang, Tongbiguan, Nanjingli, and Jinzhuzhai areas range from 50.1 to 55.3 Ma [42, 43], and those of granitic plutons in the Bangwan and Xima-Tongbiguan range from 53 to 54 Ma [27, 64, 65], it is thought to be a magmatic product of the syn-collision between the Indian and Asian continents. Two zircon U–Pb age data (XM18-86 and XM18-101) were obtained for the Longpen mafic rocks investigated in this study, with weighted average ages of 52.8 ± 0.8 and 53.8 ± 0.4 Ma, respectively. Thus, the mafic rocks developed as a response to magmatic activity during the closure of the Neo-Tethys Ocean in the Early Eocene; these are contemporaneous with the syn-collision magmatic activity (~50 Ma) of the Gangdese magmatic arc (Figure 8(a)) [2, 13, 45, 66].

The mafic rocks of the continental arc inevitably undergo strong magmatic processes, and the geochemistry is no longer in the original magmatic state. Therefore, based on geochemistry, the inversion of source area attributes is not sufficiently intuitive. Previous studies have proposed that the Early Eocene mafic rocks in the Tengchong area underwent some arc magmatic processes [42, 43]. However, to understand whether they underwent fractional crystallization, crustal contamination, partial melting, or magma mixing, further analysis and consideration are needed to determine the specific evolutionary processes. Trace elements of the Early Eocene hornblende gabbro and diorite samples are enriched in LREEs and LILEs, including Rb, Sr, and K; however, they are depleted in HREEs and HFSE, including Nb, Ta, Ti, and P. Since (Nd/Yb)N = 4.47–6.90, and (La/Sm)N = 2.08–3.97, the samples have the characteristics of enriched arc volcanic rocks in the source area (Figure 6). The Sr–Nd and zircon Lu–Hf isotopic compositions show clear enrichment features (Figure 7 and Figure 8), with initial 87Sr/86Sr (t) = 0.708787–0.709441, εNd(t) ranging from −6.3 to −5. 5, and εHf(t) ranging from −12.59 to −3.54, thus indicating that the primitive magma was derived from the enriched mantle related to the long-term metasomatism by the Neo-Tethys Ocean subducting slab and/or sediments [27, 28, 40-43, 45, 66].

Accurate analysis of pressure and temperature is crucial to understanding the storage status of magma reservoirs in the magma system [67]. Amphiboles and biotites are ubiquitous and have diverse compositions in hydrous arc magmas [63]. Therefore, they have substantial potential for temperature and pressure measurements [68-71]. The mineralogical study of mafic rocks in the Tengchong Block can be used to constrain the physical environment for the emplacement and differentiation of arc magmatic mafic rocks. This study employs machine learning thermobarometry to compute temperature and pressure during amphibole crystallization [72]. The crystallization temperature of amphiboles is 762–875°C, with an average value of 843℃; the crystallization pressure of amphiboles is 2–6.7 kbar, with an average value of 3.92 kbar (Figure 13(a)). The chemical composition of biotites can also be used to estimate their environment of origin. Based on biotite Ti saturation thermometer [68], the crystallization temperature of biotites is found to be between 556 and 720℃, with an average value of 664. According to the full aluminum barometry of biotites, the crystallization pressure of biotites is 1.81–2.53 kbar, with an average value of 2.12 kbar (Figure 13(b)) [68]. Therefore, the emplacement level of the magma reservoir associated with hornblende gabbro–diorite in the Tengchong Block is in the middle to upper crust.

5.2 Petrogenesis and Magmatic Evolution Processes

The Early Eocene hornblende gabbro–diorite of the Tengchong Block showed low MgO (2.83, 3.20 wt%) and Mg# (41.28, 49.15) values; Cr = 3.00–14.55 ppm, Co = 25.86–45.57 ppm, Ni = 3.26–5.39 ppm, and Nb/U = 2. 31–12.36, indicating disequilibrium with the mantle-derived primitive melt, suggesting that it may have undergone intracrustal evolutionary processes such as fractional crystallization or crustal contamination. However, the variable major and trace elements with consistent Sr-Nd isotopes in the sample (Figure 7(b)) indicate that the influence of continental crustal contamination on its geochemical properties can be considered negligible. The in situ Sr isotopes of plagioclase show 87Sr/86Sr = 0.708431–0.709377, similar to the initial 87Sr/86Sr of the whole rock (0.708787, 0.709441). The overall homogeneity of the amphibole and biotite compositions also indicates that crustal contamination exhibits an inconsiderable effect on its geochemistry. The imbalance between the sample and the primitive mantle melt indicates that the chemical composition of the Early Eocene mafic rocks in the Tengchong Block is due to the crystallization differentiation in the crust, the primitive magma may have undergone the crystallization differentiation before emplacing in the mid-upper crust. Therefore, the hornblende gabbro–diorite of the Tengchong Block may be the product of an evolved mafic (intermediate) magma. Additionally, the accumulation and differentiation of evolved mafic magma may also occur at the emplacement level.

The petrography of hornblende gabbro–diorite shows that some small-particle biotites and amphiboles are occasionally distributed in the core of the plagioclases, and other amphiboles and biotites in the euhedral–subhedral granular form surround the plagioclases, thus indicating that the crystallization time of amphiboles is not earlier than that of plagioclases, whereas the crystallization time of zircons is earlier than that of plagioclases (Figure 2(d)–(g)). Magnetite is distributed as an interstitial phase, and all samples showed high Fe2O3t (6.84, 10.45 wt%) and TiO2 (0.81, 1.63 wt%) contents. Based on the mineral characteristics within the thin section scale, the crystallization sequence of the major minerals in the hornblende gabbro–diorite could be concluded as plagioclase → amphibole/magnetite → biotite → quartz. In addition, the Early Eocene Longpen mafic rocks in the Tengchong Block contain a large number of touching euhedral crystals of plagioclases along with amphiboles and biotites as the crystal framework and orientation arrangement, this also led to imbrication of plagioclases. The mineral framework texture and imbrication of undeformed plagioclases are interpreted as a typical cumulative texture formed through crystal accumulation and interstitial melt removal, that may have resulted from compaction [73]. The subhedral–anhedral Type III plagioclase, polycrystalline amphibole, and quartz grains as the interstitial phase are densely packed within the framework, indicating a low melt extraction rate or low interstitial melt undercooling rate. Evolved melt remains caged between the accumulation of mineral phase particles [73, 74]. Furthermore, interstitial melt may interact and alter the morphology of the early-formed bulk crystals because of the low undercooling rate of intergranular melts. Some Type I and Type III plagioclases form fainter boundaries with each other (Figure 2(h)), suggesting that contact melting may have occurred during physical compaction [74, 75]. Some plagioclase crystals exhibit compressive deformation and mosaic textures (Figure 2(d)) indicating that physical fractures or compositional mutual infiltration may have occurred during the physical compaction process associated with plagioclase [76]. Additionally, the accumulation of feldspar is the only petrogenetic mechanism for producing positive Eu anomalies in igneous rocks during the process of arc magma fractionated crystallization or partial melting processes. However, in a feldspar-dominated closed system, the positive Eu anomaly of the residual melt is not an inevitable result of feldspar accumulation, and its absence does not rule out the possibility of feldspar crystal accumulation [77, 78]. Feldspar cumulate rocks with negative Eu anomalies may have been formed through high oxidation and magmatic dissociation of the original magma, a long history of feldspar fractional crystallization, mixing of recharged magma, and feldspar reabsorption processes [77, 78]. The trace element characteristics of the Early Eocene hornblende gabbro–diorite in the Tengchong Block show obvious negative Eu anomalies (Eu/Eu* = 0.62–0.90) and low MgO content (2.83, 3.20 wt%; Figure 6(a)); this may be attributed the depletion of Eu in the magma during early plagioclase fractionated crystallization. Therefore, the evolutionary mafic melts do not result in a positive Eu anomaly during the plagioclase-dominated accumulation.

In the study of mafic magmatic systems, magnesium–iron exchange between minerals and melts is usually used to assess the compositional balance between crystals and melts, thereby reflecting the crystal accumulation of minerals and rocks [69, 70, 79]. Amphibole is a typical mafic silicate mineral, and the Fe/Mg equilibrium of amphibole–melts can be used to assess the equilibrium state of amphibole–whole rocks [69, 70]. The most typical characteristics of Early Eocene mafic rocks in the Tengchong Block are high Fe (FeOt = 6.15–9.48 wt%) and low Mg (MgO = 2.83–3.20 wt%). According to Putrika [70]:

KDAmp=(FeOtAmp/MgOAmp)/(FeOtLiquid/MgOLiquid)

where the FeOt represents total FeO. The Fe/Mg exchange coefficient during the amphibole–equilibrium melt is KDAmp = 0.28 ± 0.11 [69, 70]. The calculation results in this article show that the KD>0.39 (Figure 14(a)) indicating that the amphibole in Early Eocene amphibole-rich mafic rocks is in disequilibrium with the whole rock, indicating that the rock underwent crystal accumulation that the amphibole could be in equilibrium with more evolved melts. For further evaluation of the evolved melts, the major elements of the amphibole equilibrium melt based on the calculations of Higgins et al. [72] are shown in online supplementary Table S9. The calculation of the trace element content in the amphibole equilibrium melt using the trace element partition coefficient is provided in online supplementary Table S12; this is formulated as follows:

CiMelt=CiMineral/DiMineral/Melt

where, CiMelt and CiMineral represent the concentrations of i element in the melt and mineral, respectively, and DiMineral/Melt represents the partition coefficient of i element in the mineral and melt [80], listed in online supplementary Table S11. To assess the reliability of the balanced melt composition, the SiO2 content in the equilibrium melt was calculated using another method by Zhang et al. [81], which showed excellent agreement with the calculated SiO2 content (Figure 14(b), online supplementary Table S10). In addition, the recalculated Fe/Mg exchange coefficient between the amphibole and the calculated melt composition is equilibrium (Figure 14(a)).

The amphibole in the Early Eocene mafic rocks of the Tengchong Block has the characteristics of low SiO2 and high TiO2 levels. The remaining melt generated via the separation and crystallization of the amphibole evolves toward higher SiO2 and lower TiO2 levels. Melts in equilibrium with amphibole generally exhibit the characteristics of enriching incompatible elements and depleting compatible elements as the temperature decreases [79]. The SiO2 content (70.5, 77.3 wt%) of the amphibole equilibrium melt is relatively high, indicating a more silicic melt (rhyolite–dacite) than the bulk rocks in composition (Figure 14c). The chondrite-normalized REE diagram of the equilibrium melt shows enrichment of LREEs, (La/Yb)N = 4.1–27, and small negative Eu anomalies (Eu* = 0.27–0.76). The content of compatible elements in the equilibrium melt is close to the whole-rock composition of the sample (Figure 13), whereas the concentration of incompatible elements in the equilibrium melt is generally higher than that in the whole rock. This may have resulted from the migration of interstitial melts containing incompatible elements toward the upper magma reservoir after the crystal accumulation.

The decreasing trend of the Dy/Yb and Dy/Dy* ratios with magma evolution is considered as important evidence for the fractionated crystallization behavior of amphibole during the evolution of arc magma [7, 8, 82]. To further evaluate the crystal–melt evolution in mafic magma reservoirs, this study uses amphibole and its equilibrium melt composition to simulate its evolution process. According to Ersoy [83], we used the REEs (La, Dy, and Yb) of the amphibole to evaluate the crystallization behavior of amphibole during its evolution. The trace elements were calculated using the basis of the Rayleigh fractionation equation (equations 1 and 2):

CL/C0=F(D1)
(1)
CR/C0=(1FD)/(1F)
(2)

where CL is the concentration of trace elements in the fluid, C0 is the initial concentration of trace elements in the fluid, CR represents the concentration of trace elements in the total residual solid, dominat F is the differentiation index (F = 0–1), and D is the integral of the product of all mineral ratios and partition coefficients. Assuming that amphibole is the major separated crystalline phase, the most basic component in the sample is used as the initial value for simulating the evolution. The simulated Rayleigh fractionation curve of the evolved melt and the cumulus evolution curve related to amphibole are shown in the figure (Figure 14(d)). The results show that the DyN/YbN and Dy/Dy* values of the equilibrium melt and accumulated matter gradually decrease with the increasing of the evolution degree, indicating that amphibole may have undergone fractional crystallization. However, the abnormal distribution between the DyN/YbN ratios and the mineral accumulation evolution curve as well as the compositional difference in the evolution between amphibole and whole rock indicate that amphibole could not be the major accumulation mineral. Combining the proportion of the primary rock-forming minerals in the sample and the petrographic perspective, it is suggested that amphibole be considered as the accumulation mineral, and that the plagioclase is the dominat accumulated phase in Early Eocene mafic rocks.

Combining petrography and equilibrium melt calculations of amphibole, we conclude that the hornblende gabbro–diorite in the Tengchong Block formed via crystal accumulation, in which plagioclase is the major accumulation mineral, along with a small amount of amphibole within the mid-upper crustal mafic magma reservoir.

5.3 Zoning Texture and Crystallization Path of Plagioclase

Plagioclase is not only one of the most abundant rock-forming minerals in the crust but also the most important indicator mineral when investigating the characteristics and evolution process of magma. Because of changes in temperature, pressure, volatiles, and melt composition during crystallization, plagioclase is often classified by multiple structural features and zonal patterns. For example, magma recharge and interparticle melt infiltration can lead to partial or complete remelting or dissolution and reprecipitation of plagioclase, resulting in the formation of unique core–rim or core–mantle–rim texture of plagioclase [84, 85]. The plagioclase in the mafic rocks of the Tengchong Block exhibits a typical core–mantle–rim texture. Hence, according to the texture, it can be divided into Type I plagioclase with a core–rim texture, Type II plagioclase with a core–mantle–rim texture, and Type III plagioclase with interstitial phase plagioclase. The BSE images of the three types of plagioclases are shown in Figure 12.

Type I plagioclase exhibits high An values at the core (69.4, 83.0) with a sieve texture, lower An values at the rim (42.4, 55.2), and the lowest An values at the overgrowth edge (29.9, 38.4), with a content difference of up to 40.6 between the core and rim. Additionally, the cores of Type II plagioclase have low An values (44.2, 49.1), subsequently abruptly increasing in the mantle (An=72.4–82.0) with a sieve texture, similar to the cores of Type I plagioclase, surrounded by a rim with lower-An values (30.3, 42.6) and with lowest-An values at the overgrowth edge (20.1, 33.5). The sieve texture developed in the cores of Type I plagioclase and in the mantles of Type II plagioclase may be generated by the reaction of precrystallized plagioclase at the dissolution interface with a higher temperature and water-rich mafic melt [86]. Natural samples and experimental petrology suggest that the dissolution event may have been caused by the decompression of water-unsaturated magma during its rapid ascent [67, 87]. However, the calcium content of plagioclase increases with decreasing pressure at a rate of ~3 mol% per kbar at a given temperature in a water-undersaturated melt [88]. Therefore, the difference in An content between the core and rim cannot be solely explained in terms of decompression–crystallization and convective self-mixing or stirring of the system [74, 88]. Batch replenishment of hydrous magma can also increase H2O pressure, reducing the stability of plagioclase and leading to its dissolution [87]. Thus, the sieve textures in the plagioclase may have resulted from crystal–melt interactions or from the reaction of plagioclase with low An values in crystalline mush with more mafic, high-temperature, and water-rich melts along the crystal–melt interface after batch magma recharge. This indicates that the Type I plagioclase cores and Type II plagioclase mantles may have reprecipitated from a mixed magma (i.e., the evolutionary parent magma reaction with a higher temperature water-rich mafic melt) [86], and that the Type II plagioclase with low An values at the core could be obtained from this reaction, representing the products directly crystallizing from the evolutionary parent magma (intermediate). The irregular interface between the core and mantle of large crystalline plagioclase (Figure 10, white arrow) may indicate the presence of one or more dissolution–reprecipitation reactions. The An values for the core of Type III plagioclase (45.0–61.9%) are similar to those for the rims of Type I and Type II plagioclases; there are similar to the low-An values at the rim of Type III plagioclase (33.4, 35.2), indicating that the plagioclase core could crystallize from a mixed magma related to the evolutionary parent magma and recharged magma, and that the rim of all types of plagioclase may have crystallized from the interstitial melt. The similar Sr isotope compositions among the core, mantle, and rim (Figure 9) also indicate that the evolutionary parent magma and recharged magma are homogeneous [89].

5.4 Genesis and Importance of Plagioclase-Rich Cumulate Rocks in the Mid-Upper Crust of the Early Eocene Arc of the Tengchong Block

Early Eocene magmatism in the Tengchong Block records the closure of the Neo-Tethys Ocean and syn-collision of the Indo-Asian continents [2, 27, 28, 42, 43, 46, 47, 49, 64, 65]. Although the fractionation crystallization of precollisional arc magmas has resulted in the formation of numerous amphibole-rich mafic–ultramafic cumulate rocks at the arc root along the continental arc margin related to the Neo-Tethys (e.g., hornblende rock and hornblende gabbro), the growth of the continental margin arc primarily occurred during the syn-collisional stage [2]. The syn-collisional period of the continental margin arc of the Tengchong Block is ~51 Ma, The slab located at the bottom of the Tengchong Block lithosphere in the Early Eocene break-off; as a result of the thermal event, the mantle wedge and accumulation rock located at the bottom of the continental lithosphere, as a result of the thermal event, partially melted and transferred the melt to the magma reservoir in the mid-upper crust [2]. A genesis study of Early Eocene mafic rocks in the Tengchong Block indicates that the creation of mafic rocks of the arc magmatic system cannot be attributed to simple fractional crystallization processes alone [12, 17, 90]. The results of petrography, Rayleigh fractionation, calculation of the amphibole Fe/Mg exchange coefficient (, and equilibrium melt composition indicate that the Early Eocene hornblende gabbro–diorite in the Tengchong Block is an enriched mantle-derived evolved mafic magma product related to plagioclase-dominated accumulation in the mid-upper crust magma reservoir (Figure 15(b)). Based on the different structural and compositional characteristics of the three plagioclase types (Figure 15(c)), we propose a model for the formation and magmatic evolution of mafic rocks in the Tengchong Block. The crystallization and magmatic evolution process of the hornblende gabbro–diorite in the Tengchong Block can be divided into fourtages (Figure 15(d)). In the first stage, the mafic magma originates from the deep enriched mantle and fractionation in the lower crust, subsequently, the evolved magma emplaces at the mid-upper crust and is thus intermediate in composition. It could be crystallized as the earliest plagioclase core (core of Type II, andesine–oligoclase in composition). In the second stage, accompanied by the recharging of more mafic, high-temperature, water-rich magma, the plagioclase formed in the first stage undergoes partial or complete dissolution. The partially dissolved crystals (residual core) subsequently recrystallize to form the Type II plagioclase sieved mantle and the completely dissolved crystals form the sieved core of Type I plagioclase. In the third stage, with the occurrence of accumulation, the crystals physically settle owing to gravity, causing the replenished magma to mix with the evolved basic magma; further, the crystallized plagioclase recrystallizes to form the rim of Types I and II plagioclases (Rim1). The core of Type III plagioclase also nucleates at this stage. In the last stage, some boundaries of accumulated plagioclase crystallize in the intercumulus melt to form the irregular overgrowth edge (Rim2), and Type III plagioclase forms the relatively complete overgrowth edge (Rim2). The multistage crystallization process of plagioclase in mafic rocks proves that the recharged magma process controls the crystallization of plagioclase; it also shows that the multistage crystallization–accumulation process is an effective mechanism for the formation of mafic rocks in the mid-upper crust of the continental margin arc. It further shows that the complexity of the crystallization process in the mush is crucial for understanding the formation and evolution of minerals in the multistage crystallization–accumulation process [14, 15, 33] (Figure 15(b)). Additionally, the plagioclase-dominated crystallization–accumulation during the vertical differentiation of arc magmas has been confirmed to be probably located at the mid-upper crust based on amphibole and biotite thermobarometer measurements (Figure 15(a)), which is of great importance for understanding the genesis of some arc mafic rocks [6, 28].

Previous studies of the arc magmatic systems, such as the Gangdese, Kohistan, and Cordilleran arcs, indicate substantial amounts of pyroxenite or hornblendite in the lower crust. Because of their higher density, these rocks tend to recycle back into the mantle, causing the composition of the residual arc to approach that of the continental crust [2, 12, 82, 90, 91]. Experimental petrology indicates that arc magma undergoes polybaric differentiation at different depths, and further differentiation in the shallow crust may be an important mechanism for the formation of granitic upper crust [32, 33]. Here, the investigation of Early Eocene hornblende gabbro–diorite in the Tengchong Block shows that the evolved mafic magma is stored as mush in the mid-upper crust, and undergoes plagioclase-dominated accumulation, whereas the evolved melt is more silicic (felsic), which is an important pathway for further differentiation of the continental crust in the mid-upper crust.

  1. The age of the Early Eocene hornblende gabbrodiorite in the Tengchong Block is ca. 53 Ma, indicating that the mafic rocks in the Tengchong Block are associated with magmatic activities related to the subduction and closure of the Neo-Tethys Ocean, and are contemporaneous with the syn-collision magmatic activity of the Gangdese magmatic arc (~50 Ma). The primitive magma originated from the mantle-enriched source area and was formed in the mid-upper crustal magma reservoirs.

  2. The Tengchong Block hornblende gabbrodiorite is a cumulate rock developed in a mafic magma reservoir (crystal mush) in the mid-upper crust; it is dominated by plagioclase and contains less amphibole, which makes up the main cumulate minerals. Plagioclases are classified into three types: Type I, Type II, and Type III, undergoing four stages of crystallization–accumulation process.

  3. The magma recharge of the arc magma at the mid-upper crust and multistage crystallization–accumulation process in the crystal mush are important mechanisms for the formation of arc magma mafic rocks; they are also an important pathway for further differentiation of the continental crust in the mid-upper crust.

All the initial data and calculation results are provided in the Supplementary Materials, including the whole-rock major and trace elements data, isotopic data, and mineral major and trace elements data.

The authors declare no conflicts of interest.

The authors are grateful for the important comments provided by the chief editor and the anonymous reviewers, aiding in substantially enhancing the manuscript quality.

Supplementary Table S1 Zircon U–Pb isotopic data.

Supplementary Table S2 Whole-rock major and trace element data and whole-rock Sr–Nd isotopic compositions.

Supplementary Table S3 Zircon Lu–Hf isotopic compositions.

Supplementary Table S4 In situ Sr isotopic compositions of plagioclase.

Supplementary Table S5 Amphibole major elements compositions, thermobarometer, and Fe/Mg exchange coefficient data.

Supplementary Table S6 Trace element composition of amphiboles.

Supplementary Table S7 Major element composition and An value data of plagioclase.

Supplementary Table S8 Major element composition of biotites.

Supplementary Table S9 Major element composition of the amphibole equilibrium melt (according to Higgins et al. [72]).

Supplementary Table S10 Major element composition of the amphibole equilibrium melt (according to Zhang et al. [81]).

Supplementary Table S11 Partition coefficient data of the amphibole equilibrium melt.

Supplementary Table S12 Trace element composition of the amphibole equilibrium melt.

Exclusive Licensee GeoScienceWorld. Distributed under a Creative Commons Attribution License (CC BY 4.0).

Supplementary data