The modal amounts of olivine (Ol) in the studied harzburgites vary within 73%–86%, that of orthopyroxene (Opx)—26%–12.5%, clinopyroxene (Cpx)—0.6%–2%, Cr-spinel (Cr-Sp)—0.4%–0.5%. The rocks exhibit porphyroclastic textures, with Ol and Opx porphyroclasts up to 1–3 mm in size, and small grains of Ol and Opx 0.1–0.3 mm in size (Figures 2A,B). A linear orientation of elongated Ol porphyroclasts is observed (Figure 2A). The porphyroclasts of Ol and Opx exhibit kink-bands and undulose extinction (Figure 2B). The observations using SEM/EDX revealed narrow (less than 1 µm wide) oriented ingrowths of Cr-Sp, which are interpreted as exsolution lamellae. The porphyroclasts of Opx often have rounded shapes and rarer concave margins, form grain clusters. The Opx porphyroclasts also contain clinopyroxene exsolution lamellae, which lack in small Opx grains; in some cases bear Ol inclusions in their central parts (Figure 2B). More often, small Ol grains are developed around Opx porphyroclasts and along cracks, which cross the porphyroclasts. Newly-formed, small grains of Cpx, Cr-Sp, rarely edenite, and Ol are observed close to the Opx porphyroclasts and among the small grains of olivine and orthopyroxene (Figures 2D,E). The newly-formed small Ol grains bear small inclusions of Cr-Sp.

Clinopyroxene composes irregular grains (10–100 µm in size) without exsolution lamellae, which are located near to small Opx and locally indicate replacement of Opx by Cpx (Figure 2E). In addition, Cpx forms intergrowths with small irregular Cr-Sp. Homogeneous porphyroclasts of Cr-Sp generally have irregular shape and size up to 1.5 mm, and are located in the interstitial space (Figure 2F). Accessory sulfides are represented by pentlandite and found as inclusions in silicates and in the interstitial space. The secondary minerals are represented by antigorite, lizardite, talc, fibrous chlorite, formed due to low-temperature serpentinization. The amount of secondary minerals in harzburgites does not exceed 1%.

Most of dunites are composed of Ol (>90 vol%) and Cr-Sp (1-3 vol%) and lack pyroxenes, except the S21-89 sample, which has 8.5 vol% of Opx and 0.5 vol% of Cpx. Dunites have porphyroclastic textures (Figures 3A–C), with Ol porphyroclasts up to 1–5 mm and small grains of Ol 0.1–0.5 mm in size. Kink-bands in olivine are also observed. Ol porphyroclasts contain narrow oriented plates of Cr-spinel. In the S21-89 sample, Cr-Sp has irregular shape, whereas in other dunites, Cr-spinel is anhedral to subhedral. Cr-Sp grains form chains, which are located both in porphyroclasts and small grains of olivine in some samples (S18-11, S18–12, S18-12/6). The veinlets of antigorite are present in dunites as well.

Cr-Sp in duinte bear silicate inclusions, which are interpreted as trapped and crystallized batches of the former silicate melt. Such melt inclusions were found in three dunites; one sample (S18-11) only has small inclusions cut by cracks and completely replaced by chlorite, whereas in the S18–12 and S18-12/6 dunites, the inclusions are irregular-shaped and orbicular (Figure 4). The inclusions are arranged irregularly within the host Cr-Sp, and their size ranges from 20 to 40 µm (generally single-phase) to 150–200 µm (polyphase inclusions). The multiphase inclusions are composed of Ol, pyroxenes (diopside and enstatite), Amp (Mg-pargasite, Mg-hastingsite, edenite, tschermakite), and phlogopite (Figure 4; Supplementary Table S2). A single melt inclusion contain up to 3 exposed phases (e.g., Cpx+Opx+Phl or Cpx+Ol+Phl), while some of the inclusions are composed of two phases (most often Cpx+Opx, Cpx+Amp, or Opx+Amp), or a single phase (Ol, Cpx, Opx, or Amp). Some of the inclusions are cracked and underwent secondary alteration, expressed as development of serpentine and chlorite (Figures 4E,H). Secondary magnetite develops on the contact between serpentinized inclusions and host Cr-Sp (Figures 4C,H) and forms rims around the Cr-Sp.

The Khara-Nur harzburgites have very low losses on ignition (LOI) within 0–0.41 wt% (Supplementary Table S1). Relative to depleted mantle (DM) composition, the rocks are richer in MgO (45.9–48.4 wt%) and strongly depleted in Al₂O₃ (0.39–1.05 wt%), CaO (0.27–0.49 wt%), Na₂O (b.d.l.), and K₂O (b.d.l.). In dunites, the LOI values are slightly higher (0.84-1.89 wt%). As compared to harzburgites, the dunites exhibit lower SiO₂ (40–42 wt%), CaO (<0.1–0.4 wt%), Al₂O₃ (<0.1–0.5 wt%), but higher MgO (48.3–51.7 wt%) and Cr₂O₃ (0.43-1.94 wt%). NiO abundances in the samples of two types are similar (0.3–0.33 wt% in harburgites versus 0.29–0.36 wt% in dunites).

Cr-Spinel from harzburgites has high values of Cr# (Cr/(Cr + Al)) within 0.51–0.66 and Mg# (Mg/(Mg + Fe²⁺)) within 0.58–0.66 (Supplementary Table S2; Figure 5). In some dunite samples (S18–12, S18-133), Cr-Sp has the composition similar to that from harzburgites, while higher Cr# (0.87–0.89) and lower Mg# (0.52–0.65) are typical for the rest dunites (samples S18–11, S21-116). The TiO₂ contents in Cr-Sp from harzburgites is below 0.1 wt%, while slightly higher values (from <0.1 wt% to 0.15 wt%) were revealed occasionally in dunites. In harzburgites, there is a systematic decrease in Cr# and an increase in Mg# towards the rims of Cr-Sp porphyroclasts; newly formed, small Cr-Sp grains resemble the composition of porphyroclast rims or exhibit even lower Cr# and higher Mg#. Cr-Sp from both harzburgites and dunites are characterized by low Fe³⁺/∑Fe ratios of 0.02–0.12 obtained with correction methods using a set of secondary standards with Fe³⁺/∑Fe ratios measured by Mössbauer spectroscopy (Supplementary Table S2). In small Cr-Sp grains from harzburgites, Fe³⁺/∑Fe ratios are slightly higher (up to 5-30 rel. %) (Supplementary Table S2).

Olivine porphyroclasts in harzburgites are homogeneous, display high Mg# within 0.912–0.927 and NiO within 0.37–0.47 wt% (Supplementary Table S2; Figure 6). Ol porphyroclasts from some dunites (samples S18–12, S18-133) have Mg# of 0.916–0.925, similar to that of Ol from harzburgites. In the other dunite samples (S18–11, S21-116), Ol porphyroclasts have higher Mg# near 0.95. In addition, Ol porphyroclasts in dunites and harzburgites are characterized by similar NiO contents. In harzburgites, the relation of Cr# in Cr-Sp and Mg# in Ol is similar to that of forearc peridotites (Figure 7).

Orthopyroxene from harzburgites (Supplementary Table S2; Figures 8A,B) is high-Mg# (0.918–0.935) and characterized by low contents of Al₂O₃ (0.5–2.6 wt%), Cr₂O₃ (0.4–1.0 wt%) and very low TiO₂ and Na₂O (b.d.l.). The rims of Opx porphyroclasts have lower Al₂O₃ and Cr₂O₃ contents at similar Mg# relative to the cores. The small grains of Opx have the composition of porphyroclasts or even lower Al₂O₃ and Cr₂O₃. Cpx from harzburgites (Supplementary Table S2; Figures 8C,D) are high-Mg# (0.942–0.957), and have low contents of Al₂O₃ (0.6–2.5 wt%), Cr₂O₃ (0.4–2.0 wt%) and Na₂O (b.d.l.—0.3 wt%).

Orthopyroxene and Cpx from melt inclusions in Cr-Sp from dunites (Supplementary Table S2; Figure 8) have high Mg# (0.925–0.935 and 0.94–0.95, respectively), varying contents of Al₂O₃ (0.6–1.48 wt% and 0.81–3.06 wt%, respectively) and Cr₂O₃ (0.57–1.20 wt% and 0.8-2.0 wt%, respectively). In terms of Mg#, pyroxenes in inclusions closely resemble the composition of matrix pyroxenes from the Khara-Nur harzburgites and differ from that of the Eastern Sayan ophiolite cumulates (Figure 8).

Amphibole from harzburgites is represented by pargasite and edenite. Amphibole from inclusions in Cr-spinel of dunites is represented mainly by pargasite and edenite, and rarely by magnesian hornblende (Supplementary Table S2). Pargasite and edenite have high Mg# (0.936–0.944 and 0.939–0.953, respectively) and low TiO₂ content (∼0.4–0.52 and 0.37–0.52). Phl in Cr-Sp-hosted inclusions is rare and possesses high Mg# (0.97–0.98) and low TiO₂ content (∼0.55 wt%).

Orthopyroxene from harzburgites (Supplementary Table S2) has low contents of trace elements (Figures 9A,B), which are, in some samples, below the detection limits for some elements. The chondrite-normalized rare-element element (REE) patterns have positive slopes of HREE and flat or slightly negative slopes of LREE, with LREE_N<HREE _N (Figure 9A), which indicate more depleted REE composition of the Khara-Nur Opx relative to that from abyssal peridotites (Figure 9A). Trace-element patterns are U-shaped with Cs and Ba enrichment, prominent positive Pb anomaly, and weak negative Nb and Zr anomalies (Figure 9B). The observed positive anomalies of Ti support highly depleted nature of the rocks (McDade et al., 2003). More specifically, trace element patterns are similar to that of Opx from supra-subduction peridotites (Figure 9B), for instance, from peridotite xenoliths of the Bismark arc (Tollan et al., 2017) and forearc peridotites of Egypt (Khedr et al., 2022). Opx from Cr-Sp-hosted inclusions have similar HREE contents and higher L-MREE contents relative to Opx from harzburgites (Figure 9C), while their trace element patterns demonstrate positive anomalies of Pb (Figure 9D).

Clinopyroxene from harzburgites is characterized by low contents of trace elements (Supplementary Table S2). In most samples, Cpx has similar REE patterns with a decrease from HREE to MREE and LREE variations, generally with LREE_N<HREE_N. Almost flat REE distribution patterns are observed in Cpx from three harzburgite samples. All Cpx demonstrate enrichment in Cs and positive anomalies of Nb and Pb (Figure 10B). HREE contents in Cpx of the Khara-Nur harzburgites are lower than in Cpx from abyssal peridotites and approach that of peridotites from forearcs and supra-subduction settings (Figure 10A).

Clinopyroxene from inclusions in Cr-Sp exhibit two types of trace element distribution. Except one analysis, it displays flat distribution of HREE, enrichment in MREE and LREE over HREE, and generally demonstrates a positive slope from La to Nd (type 1; Figure 10C). All but one Cpx exhibit weak negative anomalies of Eu. Most Cpx are enriched in Cs and Ba, show negative anomalies of Nb, Zr, Ti and Sr One inclusion-hosted Cpx demonstrates negative anomaly of Ba, and two others lack Sr anomaly. Moreover, Pb shows the most variable behavior with positive, negative, or lacking anomalies. Only one inclusion-hosted Cpx revealed almost flat distribution of REE (Figure 10C) at enrichment in Cs and Ba (type 2; Figure 10D). HREEs contents in Cpx of both types are similar and vary within 2–5 chondrite levels, while more significant variations (1–20) of LREE were found. Importantly, Cpx with both types of trace element patterns may be found in a single Cr-Sp (Figures 10C,D).

The equilibrium temperatures were calculated using several geothermometers based on rock-forming and accessory phases (Supplementary Table S1). The highest temperature values were obtained by Opx-Cpx thermometer of Liang et al. (2013), based on REE+Y distribution between two pyroxenes (862°C–1,093 °C). The most well reproduced temperature values were calculated from HREE and Y distribution, whereas LREEs distribution coefficients for pyroxene pairs deviate from the regression lines due to their slower diffusion rates. Similar temperatures were obtained by Ca-in-Opx and Al-in-Opx contents thermometers (1,000°C–1,114 °C and 935°C–1,056 °C, respectively) (Brey and Köhler, 1990; Köhler and Brey, 1990). The lowest temperatures were obtained from thermometers using Fe²⁺-Mg diffusion exchange. Two-pyroxene Fe-Mg thermometry (Brey and Köhler, 1990) yielded a range of 767°C–888 °C, whereas the values from Ol–Sp pairs (O’Neill and Wall, 1987; Ballhaus et al., 1991) are even lower (693°C–829 °C). In dunites, the temperature values from Ol-Sp thermometry are within 770°C–859 °C, which is similar to those estimated for harzburgites by the same thermometer. In Cr-spinel-hosted inclusions in S18-12 dunite, the clinopyroxene crystallization temperature estimated from the geothermometer (Wang et al., 2021) is 1,258 °C, and the amphibole crystallization temperature calculated from the geothermometer (Ridolfi et al., 2010) is 940 °C (Supplementary Table S1). The equilibrium temperatures calculated from two-pyroxene Fe-Mg (767°C–888 °C) and REE+Y geothermometry (891°C–1,089 °C) for Khara-Nur harzburgites are within the range for supra-subduction peridotites and lower than estimates for the abyssal peridotites (Dygert and Liang, 2015; Tollan et al., 2017).

The Khara-Nur harzburgites have weakly correlated SiO₂ (42.5–45.0 wt%) and Al₂O₃ (0.48–1.05 wt%) contents, with the latter being a proxy for melting degree (Figure 11). The observed SiO₂—Al₂O₃ systematics allows splitting the mantle rock compositions into two groups (Figure 11B). The Group 1 samples have elevated SiO₂ contents; their compositions variably shift from the compositions of residues from anhydrous melting of fertile lherzolites. Elevated SiO₂ contents are characteristic of arc-related harzburgites (e.g., Parkinson and Pearce, 1998; Song et al., 2009; Birner et al., 2017), which differ from the oceanic harzburgites (Herzberg, 2004). Such peridotites could be residues from hydrous melting of a specific, SiO₂-rich mantle source called “hybrid mantle wedge” (HMW) (Bénard et al., 2017), which forms through enrichment of typical depleted-mantle substrate by slab-derived fluids/melts before or during partial melting. In contrast, the Group 2 harzburgites resemble major-element compositions of residues from anhydrous melting of a source compositionally similar to DM, which was not affected by subduction-zone crust-mantle interactions.

The two groups of Khara-Nur mantle harzburgites and dunites have chemistries consistent with two distinct melting trends (Figure 11B). The modal compositions of harzburgites agree well with calculated amounts of Opx in residues from hydrous and anhydrous melting (Figure 11B; Bénard et al., 2017). For instance, the S12-27 harzburgite contains 26% Opx, which is consistent with the recalculated 23% Opx in a hydrous melting residue. Similarly, for samples S12-27/1 and S12-28 with 15.7% and 15.5% Opx, respectively, predicted Opx in anhydrous melting residue is around 20%. The composition of two dunite samples (S21–89 and S21-78) also tend to follow the mentioned melting trends (Figure 11), and thus have a residual origin. On the contrary, for the rest two dunites, major-element compositions deviate from melting trends and have lower NiO than that of residual dunites (Figures 11B,C,E). The formation of these low-NiO dunites could relate to melt-harzburgite interaction, while fine-grained Cr-Sp and Opx enclosed in Ol of such dunites (Figure 3A) point out Ol crystallization at the expense of Opx both during partial melting and melt-rock interaction.

Based on the whole-rock abundance of Al₂O₃ in residual peridotites, the Khara-Nur harzburgites were produced by ∼30–40% melting (Figure 11) in accordance with the lack of residual Cpx in the rocks. Notably, harzburgites of both Groups 1 and 2 have high and similar melting degrees. Within each of the two groups, the samples demonstrate a decrease in modal Opx accompanied by growth in Cr# of Cr-Sp (Figure 12), which is consistent with partial melting trends. The high melting degrees of both groups of Khara-Nur harzburgites are confirmed by low reconstructed HREE in bulk-rock samples estimated from mineral trace-element abundances and their modal compositions (Figure 13). Similar melting degrees were estimated for typical supra-subduction zone peridotites, including, for instance, the Izu-Bonin-Mariana forearc and Kamchatka arc (within 25%–35%), as well as Bismark arc (to 30%–40%) (Figure 11), whereas the Khara-Nur harzburgites specifically possess calculated HREE values similar to or lower than that of the most depleted forearc peridotites from Izu-Bonin-Mariana (Figure 13).

Alternatively, melting degrees can be estimated from Cr# of Cr-spinel. In case of multi-stage melting started in the garnet-facies and continued in the spinel-facies, Cr-spinel retains Cr# corresponding to a lower melting degree than the bulk melting degree. In both groups of the Khara-Nur harzburgites, the less depleted samples (S12-27 in Group 1 and S12-27/1 in Group 2) preserve Cr-Sp with Cr# of ∼0.55 and ∼0.56, respectively, which correspond to melting degrees of 18% versus 25%–28% estimated from whole-rock Al₂O₃ abundances. Therefore, two-stage melting scenario should be assumed. The studied harzburgitic Cpx exhibits low MREE/HREE ratios, which are potential indicators of an early-stage, garnet-facies melting episode (Hellebrand, 2002). The estimated equilibrium temperatures for harzburgites obtained from two-pyroxene REE thermometry (862°C–1,093 °C) and Ca-in-Opx and Al-in-Opx thermometry (1,000°C–1,114 °C and 935°C–1,056 °C, respectively) are generally lower than temperatures of melt-harzburgite interaction and magmatic Cpx crystallization (Tollan et al., 2017; Le Roux and Liang, 2019), and should reflect the subsolidus transformation of the rocks. As shown by Liang et al. (2013)and Dygert and Liang (2015), during cooling and subsolidus re-equilibration, REE are redistributed between Opx and Cpx, especially at low Cpx modes and lower temperatures, may also lead to both low MREE/HREE ratios in Cpx. Thus, the trace-element composition of the Khara-Nur Cpx cannot be used unambiguously for modeling of the peridotite partial melting. Nevertheless, the difference in melting degrees estimated from Cr-spinel and whole-rock compositions of Khara-Nur peridotites testifies to the initiation of melting in the garnet-facies.

Previous studies of forearc peridotites documented that their oxygen fugacity values vary from those of mid-ocean ridge mantle to arc xenolith peridotites (Parkinson and Pearce, 1998; Pearce et al., 2000; Dare et al., 2009; Birner et al., 2017), being controlled mainly by a contribution from subducted slabs. Thus, the processes of flux melting and melt metasomatism, which involve slab-derived volatile-rich (most importantly, H₂O and S) fluid of melts should generally yield a more oxidized conditions for supra-subduction zone peridotites (Parkinson and Arculus, 1999). As shown above, the Khara-Nur harzburgites of Group 1 have chemical compositions of supra-subduction peridotites enriched in SiO₂, and therefore should be more oxidized with respect to their chemically depleted counterparts. This, however, contradicts the observations of Cr-Sp with Fe³⁺/∑Fe ratios lower relative to typical forearc and supra-subduction samples (Figure 14).

Three different approaches are used for estimating Fe³⁺/∑Fe ratios in Cr-Sp, including 1) stoichiometric calculation from EPMA data, 2) direct analysis of Fe³⁺/∑Fe ratio by Mössbauer/XANES spectroscopy, and 3) a correction method of combined EPMA analysis of unknowns and a set of standards with the reliably estimated Fe³⁺/∑Fe ratios (Davis et al., 2017). On Figure 14 we used the Fe³⁺/∑Fe ratios in Cr-Sp, obtained by method 3). The data obtained by methods 1) and 3) are consistent (Supplementary Table S2). This is caused by low uncertainties of stoichiometric calculations for Cr-spinels with low ∑FeO abundances and moderate values of Cr#, as well as determination of minor elements (e.g., V, Zn, Ni) (Quintiliani, 2006; Davis et al., 2017; Jia et al., 2022). The low Fe³⁺/∑Fe ratios in Cr-spinel are in agreement with V-Yb systematics (Pearce et al., 2000) of the Khara-Nur peridotites exhibiting lower oxygen fugacity conditions compared to that of forearc peridotites of Torishima seamount (Supplementary Figure S2).

Partial melting of the forearc mantle proceeds in two stages, with the early episode of decompression melting in the asthenosphere, which yields only moderate, MOR-like melting degrees and does not enrich the residue in SiO₂ (e.g., Herzberg, 2004). For the Tonga peridotites (e.g., Birner et al., 2017), this early melting stage is recorded by Fe³⁺/∑Fe ratios in Cr-Sp, which are transitional between the Khara-Nur harzburgites and Izu-Bonin-Mariana forearc (Torishima) (Figure 14). The second, major stage of melting assisted by slab-derived fluids and hydrous partial melts produce forearc peridotites with higher melting degrees, SiO₂ enrichment and more oxidized conditions (Parkinson and Pearce, 1998).

The Khara-Nur harzburgites of Group 2 are not enriched in SiO₂ but demonstrate high melting degrees, therefore they did not undergo flux melting. The SiO₂-enriched Khara-Nur harzburgites of Group 1 possess high melting degrees, but their Fe³⁺/∑Fe ratios in Cr-Sp are too low to support the flux melting. If the Group 1 harzburgites indeed formed due to flux melting, then some subsequent process must necessarily decrease their Fe³⁺/∑Fe ratios in Cr-spinel. One possible explanation for decrease of Fe³⁺/∑Fe ratios in Cr-spinel during partial melting was suggested for Bismark arc (Bénard et al., 2018), where some harzburgite xenoliths are SiO₂-enriched (Figure 11B), have high melting degrees (30%–40%; Figure 11A), and are relatively reduced as indicated by Cr-Sp composition (Figure 14). These xenoliths were interpreted as products of two-stage melting: flux melting of spinel harzburgites followed by second-stage adiabatic decompression melting (F∼5–8%) of the residues experienced a convective ascent in the mantle wedge (Bénard et al., 2018). Such second-stage melting is not fluid-buffered and preferentially introduces Fe³⁺ into the melt, thus decreasing the estimated Fe³⁺/∑Fe ratios (Bénard et al., 2018). However, the driving force of such second-stage melting of the already depleted substrate without a slab-derived flux remains uncertain. Such a model cannot be applied to the studied rocks, since the less depleted harzburgites of Khara-Nur Group 1 (samples S12–27, S21-88) have lower Fe³⁺/∑Fe ratio in Cr-spinel as compared to Bismark xenoliths with similar modal orthopyroxene (Figure 14), for which the fisrt stage of flux melting is suggested (Bénard et al., 2018). Alternatively, more reduced subduction-related fluids could be considered. For example, such fluids were assumingly involved into redox melting of supra-subduction peridotites of the Yushigou ophiolite (Tibet) with fO₂ < FMQ, as evidenced by primary methane-bearing fluid inclusions in Ol (Song et al., 2009). It is unlikely that the methane input and redox-melting is accompanied by rock enrichment in SiO₂. If the abovementioned SiO₂ enrichment of the Khara-Nur Group 1 rocks occurred before their introduction to a subduction zone, then redox melting assisted by moderately reduced fluids would be a plausible mechanism for generating harzburgites of Groups 1 and 2 with high melting degrees and low Fe³⁺/∑Fe ratios in Cr-Sp. However, the model of redox melting is not favorable for suprasubduction settings (Foley, 2011).

To resume, the compositions of Group 1 and 2 of Khara-Nur harzburgites cannot be explained by the formation via the melting models applied for subduction zones. The Khara-Nur peridotites did not experienced melting in the supra-subduction zone where the Eastern Sayan ophiolites formed.

Most of the provided data favor the origin of the Eastern Sayan ophiolites and—in particular—the Khara-Nur mantle peridotites in a forearc area (Khain et al., 2002; Belyaev et al., 2017, this study). However, multiple petrological and geochemical evidences imply a protracted history of the parental mantle before its entrainment in a subduction-zone setting: low Fe³⁺/∑Fe ratios in Cr-spinel, whole-rock chemical pre-enrichment (e.g., in SiO₂), and also ancient Paleoproterozoic to Mesoproterozoic Re-Os model ages of Khara-Nur peridotites (Wang et al., 2017). The protracted and multi-stage mantle evolution is commonly implied for subduction-zone peridotites based on their Re-Os model age characteristics, which variably pre-date their formation. For instance, the ancient Re depletion ages of ∼1.2 Ga for the forearc peridotites from the Torishima seamount (Eocene Izu-Bonin-Mariana forearc) were interpreted to reflect mantle source heterogeneity and an early stage of melting and subduction-driven enrichment before second-stage melting during subduction initiation (Parkinson et al., 1998). Furthermore, similar inferences were made for highly refractory peridotites from modern oceanic lithosphere and oceanic islands (e.g., Simon et al., 2008; Neumann and Simon, 2009; Dijkstra et al., 2010), which have major-element composition similar to that of forearc peridotites (Parkinson and Pearce, 1998) and suprasubduction xenoliths (Ionov, 2010), and exhibit Fe³⁺/∑Fe ratios in Cr-Sp similar to that in forearc and supra-subduction peridotites (Figure 14). Simon et al. (2008) and Neumann and Simon (2009) supposed that the highly refractory peridotites of oceanic islands originate from two-stage melting (MOR-type + flux melting), similar to that of forearc peridotites, and may be accreted to the basement of oceanic lithospheric mantle through the ongoing subduction. This scenario could be viable for the Khara-Nur peridotites if they would not have such low Fe³⁺/∑Fe ratios in Cr-Sp.

Alternatively, highly refractory peridotites of oceanic islands could represent the remnants of subcontinental lithospheric mantle. O’Reilly et al. (2009) provided a few lines of evidence that refractory and buoyant ancient cratonic lithospheric mantle blocks could be incorporated into the oceanic mantle during continental rifting and opening of oceanic basins. This would generally explain the ancient Re-Os ages of peridotites and enriched Nd-Sr isotopic signatures of oceanic intraplate basalts (O’Reilly et al., 2009). Such interpretations were favored, for instance, in case of peridotites from Lac Michèle locality of Kerguelen archipelago (Wasilewski et al., 2017) or the Yunzhug ophiolite (Tibet) (Huang et al., 2020), assumed to represent blocks of Proterozoic SCLM. It is known that the Precambrian SCLM (as well as supra-subduction lithosphere) has a specific, SiO₂-enriched composition (Herzberg, 2004). A comparison of Khara-Nur harzburgites with Lac Michèle (Kerguelen) xenoliths (Wasilewski et al., 2017) and Yunzhug peridotites (Huang et al., 2020) demonstrates their similarity in terms of whole-rock major elements (Figure 15), olivine (Figure 6), ortho- and clinopyroxene compositions (Figure 8). Moreover, in Lac Michèle and Yunzhug peridotites the Cr-spinel has very low Fe³⁺/∑Fe ratios (Figure 14). Foley (2011) reviewed the available data on peridotites from different tectonic settings and demonstrated that the oceanic and continental (cratonic and off-cratonic) mantle has variably reduced composition (fO₂<FMQ-1). In opposite, the subduction-related peridotite xenoliths are relatively oxidized, with average fO₂ value of FMQ+0.51 and considerable number of samples with fO₂>FMQ+1. In the Khara-Nur harzburgites and dunites, the Fe³⁺/∑Fe ratios of Cr-spinel are transitional between those in Cr-spinel from Proterozoic SCLM mantle block in the Yunzhug ophiolite (Huang et al., 2020) and typical forearc (Parkinson and Pearce, 1998) and supra-subduction peridotites (Ionov, 2010). This potentially indicates a transitional composition and redox state of the Khara-Nur peridotites due to melt interaction of the former refractory subcontinental mantle in the mantle wedge (Figure 14).

To resume, the compositional features of the studied Khara-Nur peridotites correspond best to their origin from more ancient Proterozoic mantle with sub-continental lithospheric affinity interacted with supra-subduction melts in a mantle wedge. Boninites and IAT in Eastern Sayan ophiolites formed due to melting of asthenospheric mantle, which experienced upwelling as a result of intra-oceanic subduction initiation. At the same time, the upward flows of the asthenospheric mantle could entrap refractory and buoyant blocks of SCLM incorporated into the oceanic mantle via mechanism suggested by O’Reilly et al. (2009) and further transport them in the mantle wedge. Later, the refractory Khara-Nur peridotites were tectonically juxtaposed with the crustal rocks of the Eastern Sayan ophiolite.