INTRODUCTION

The Yoko-Dovyren layered massif is located 60 km north of the northern tip of Baikal Lake and belongs to the Dovyren intrusive complex, which is dated at ~728 Ma (Ariskin et al., 2013; Ernst et al., 2016). The massif is clearly discernible in the local topography: it composes a range up to 2114 m high, which extends northeastward for approximately 26 km and is cut by the valley of the Ondoko River, thus being divided into two mounts: Yoko and Dovyren (Gurulev, 1965; Konnikov, 1968). The nearly vertical dips of the rocks, which is conformable with the dips of the host carbonate–terrigenous rocks of the Middle–Late Riphean Olokit complex (Rytsk et al., 2002) allowed for a detailed mapping of the modal and compositional layering (Kislov, 1998; Ariskin et al., 2018a). In its thickest (up to 3.4 km) central part, the rock sequence of this massif consists of a basal horizon of plagioperidotites, which gives way upward to plagioclase and adcumulus dunites, with a sharp transition to troctolites (varying from melano- to leucocratic lithologies), and the vertical section is crowned by gabbroids, which vary from leucotroctolites and olivine gabbro to olivine and olivine-free gabbronorites and (at the roof) to pigeonite-bearing gabbro. Dunites are practically absent from the northeastern and southwestern closures of the intrusion, and these portions of the massif are dominated by melanotroctolites and leucogabbroids (NE, Shkolnyi section). Mount Yoko is made up of interbedding troctolites and olivine gabbro (Fig. 1). The basal horizons in the margins are similar in structure: in both of them, the rocks composing the very bottom of the sequence are plagioperidotites, which are a little bit less primitive that in the core of the intrusion. Correspondingly, pigeonite-bearing gabbro are more widely spread near the top portion of the massif.

Fig. 1.
figure 1

Schematic geological map of the Yoko-Dovyren massif [based on geological maps by A.G Krapivin (Bushuev and Tarasova, 1985) and A.G. Stepina (Stepin and Vlasenko, 1994)]. Insets show summarizing vertical sections and a schematic location map of the sampling sites). (I) Southwestern part of the massif (Yoko vertical section), (II) central part (near Bolshoi and Tsentralnyi streams), (III) northeastern termination of the massif (vertical section in the Shkolnyi area). Legend: (1) Quaternary rocks (QII–IV); (2) Kholodninskaya Formation (V – C1); (3–11) Synnyr–Dovyren volcanic–plutonic complex (R2–3): (3) Inyaptuk Formation, (4–11) Dovyren intrusive complex: (4) Marginal Group: picrodolerites, diabases, and plagioclase lherzolites near the bottom of the massif and ophitic diabases in the near-roof part, (5) dunites, (6) contaminated dunites, (7) troctolites, (8) olivine gabbro, (9) olivine gabbronorites and norites, (10) quartz–pigeonite gabbronorites; (11) not subdivided; (12) carbonate–terrigenous rocks of the Synnyr rift (R2–3); (13) terrigenous and carbonate–terrigenous rocks of the Olokit trough (R2–3); (14) geological boundaries; (15) faults; (16) elevations (Mt. Yoko 1762 m, Mt. New Dovyren 2030 m, and Mt. Dovyren 2114 m). (17) transects: Ia ans Ib are Yoko, IIa ans IIb are Tsentralnyi, and III is Shkolnyi.

Full size image

The reasons for long-lasting interest in this massif are as follows. First of all, this is because the Dovyren intrusive complex is accompanied by Cu–Ni sulfide ores, which were found in underlying sills and tongues of the main body (Gurulev, 1965; Konnikov, 1986). Also, domains with low-sulfide PGE mineralization were found in the leucotroctolites and anorthosites and are referred to as Reef I (Distler and Stepin, 1992; Orsoev et al., 1995; Konnikov et al., 2000) or the Main Reef (Ariskin et al., 2016, 2020). These facts and age correlations of the rocks provided grounds to attribute the Dovyren massif to the eastern branch of the Late Riphean Cu–Ni–PGE East Siberian metallogenic province (Polyakov et al., 2013).

Another reason for the aforementioned interest in this massif is that it shows a set of typomorphic features that indicate that the Dovyren magma chamber was open during the early and intermediate evolution of the massif (Ariskin et al., 2018a). This follows from the dominance of ultradepleted rocks and from the evaluated (for individual geological sections) weighted mean concentrations of incompatible elements, which are three to five times higher than in the inner-contact rocks. Neither volcanics nor sills with such geochemical characteristics have ever been found in the northern West Baikal area, and this leads to the conclusion that the Dovyren rock sequence is a sort of a “residue sequence”, which is made up of the crystalline residues of olivine and olivine–plagioclase cumulates that have lost up to 60–70% gabbronorite melts. In view of this, information on the composition of the parental, most primitive magmas is recorded not in the average composition of the intrusion in its various vertical sections but in the picrodolerites from its lower contact zones.

The third reason is the anomalous isotope characteristics of rocks of the Dovyren intrusive complex. Data on the Rb–Sr and Sm–Nd isotope systems of ultramafic sills underlying the Dovyren massif and the chilled gabbronorites, picrodolerites, dunites, troctolites, and gabbroids of the layered complex show that three-fourths of the rock samples define a stable cluster of compositions with extremely enriched 87Sr/86Sr(t) ratios (0.7094–0.7149 at t = 728.4 Ma) and low εNd(t) = –16.1 to –13.2 (Ariskin et al., 2015). The most strongly altered, mostly serpentinized rocks lie outside this range. These facts lend support for the hypothesis that the Dovyren magmas were derived from an isotopically anomalous source (Amelin et al., 1996) and led to the conclusion that its protolith in the Neoproterozoic sublithospheric mantle may have been products of partial melting and/or metasomatic transformations of the Archean mantle at ~2.7–2.8 Ga (Ariskin et al., 2018a). It is still uncertain whether the protolith was a heterogeneous mixture of in-situ crystallized partial melts and mantle residues in supersubduction (?) mantle or it was produced by the contamination of primary mantle melts with Archean crustal material. An extrapolation of \({{\varepsilon }}_{{{\text{Nd}}}}^{{}}\left( t \right)\) values for rocks of the Dovyren complex to the intersection with the mantle evolution trend in the approximation of the primary ratio Sm/Nd = 0.350 suggests that the Sm/Nd ratio of the source was close to 0.221 (Ariskin et al., 2015), which is close to the average composition of the continental crust (Taylor and McLennan, 1985). This is a serious argument in favor of the involvement of continental crustal material in forming the mantle source of the Dovyren magmas.

The Dovyren intrusive complex thus offers a unique example of distinct open-type magma chamber, whose evolution resulted in a large layered intrusion and differentiated rocks with Cu–Ni–PGE mineralization. The source of this magmatic material in the Neoproterozoic was the sublithospheric mantle, whose isotope composition (Sm–Nd system) carries information on a much older (perhaps, Neoarchean) protolith. The contrasting superposition of the problems of evolution of the magma chambers and the processes of chamber differentiation and the origin of the sublithospheric source are in the mainstream of fundamental problems of modern petrology and the theory of processes forming Cu–Ni–PGE sulfide mineralization. A solution of these problem requires further studies of the composition of the Dovyren rocks, including their stable-isotope composition, first of all, the sulfur and oxygen isotope composition. The need to update the data on the isotope composition of these elements on the basis of a larger set of rock compositions was partly dictated by progress in modern techniques of isotope analysis and the improvement of their sensitivity. The petrological–geochemical aspect is related to difficulties in distinguishing between the effects of the early crustal contamination during the origin of the mantle source and the Late Riphean assimilation of the wall rocks, a fairly probable process during magma emplacement and the development of the chamber. This problem is made even more pressing by the fact that the Dovyren intrusion hosts xenoliths of skarns that replaced carbonate rocks and contains relicts of the host rocks (Pertsev and Shabynin, 1979; Pertsev et al., 2003). The solution of the problems formulated above is even further complicated by hydrothermal–metasomatic processes, which modified large volumes of the ultramafics and gabbroids approximately 70 Ma after the solidification of the intrusion (Ariskin et al., 2013), perhaps, simultaneously with multistage epigenetic low-grade burial metamorphism (Spiridonov et al., 2019). This paper presents high-precision data on the oxygen isotope composition of olivine grains from cumulates, which were sampled along three transects across the central part of the Yoko-Dovyren massifs and its southwestern and northeastern closures.

EARLIER STUDIES

The first estimates of the oxygen isotopes in olivine were obtained in (Ustinov et al., 1980) for monomineralic fractions from dunites, troctolites, olivine gabbro, and gabbronorites from the central portion of the intrusion. Oxygen was extracted at interaction of the samples with BrF5 or ClF5 (Taylor and Epstein, 1962; Clayton and Mayeda, 1963), including a number of experiments on faster fluorination in autoclave reactors. The isotope ratios were measured on a MI-1201 mass spectrometer, the reproducibility of δ18O values was approximately ±0.2‰. The olivine had δ18O within the range of 8.1–9.2‰. With regard to additional data on the plagioclase and pyroxene, these authors have calculated the oxygen isotope composition of the original Dovyren magma (δ18O = 8.5–8.7‰), i.e., notably higher values than estimates for primitive mantle magmas (Taylor, 1980), including MORB (δ18O = 5.7 ± 0.2‰; Ito et al., 1987). The high δ18O values have not been confirmed later, and later measurements (at the same laboratory) of olivine from picrodolerite from the lower contact zone of the massif yielded much weaker enrichment in the heavy oxygen isotope: δ18O = 6.4‰ (Krivoplyasov et al., 1984).

No less contradictory data were presented in (Goncharenko et al., 1992). These data were acquired by means of volumetric fluorination and subsequent analysis on a MI-1309 spectrometer (the declared reproducibility was 0.3–0.5‰). The oxygen isotope composition was measured in olivine monomineralic fractions from fifteen samples taken in the central part of the intrusion: one plagiolherzolite sample, four dunite samples, five troctolite samples, one olivine gabbronorite sample, and four olivine gabbro samples. Although the authors claimed that “the least altered rocks were used”, whose LOI were lower than 2.5 wt %, the results showed widely scattered δ18O values of 0.6 to 6.4‰, which were lower than the earlier measurements. Therewith two types of olivine compositions were identified: one with a mode at δ18O = 5.5 ± 0.7‰ (two olivine gabbro samples, two melanotroctolites, and one olivine gabbronorite) and δ18O = 3.0 ± 1.1‰ (the other ten samples, including all dunites and one plagiolherzolite from the basal zone). Based on these data, it was hypothesized that the Dovyren chamber was formed when magma with mantle signatures of oxygen isotope composition was emplaced, whereas the shift toward low δ18O values correlated with traces of plastic deformations in the olivine grains. This fact was regarded as an argument that oxygen isotopes were redistributed at the syntectonic recrystallization of the rocks, perhaps, with the involvement of fluids with δ18O values (Goncharenko et al., 1992).

It should be mentioned that data on the oxygen isotope composition of olivine and other refractory minerals obtained by volumetric fluorination (particularly with the application of faster decomposition of the minerals, e.g., Ustinov et al., 1980) may have introduced uncertainties related to that the minerals were not completely decomposed at external heating. As a result, the modern literature conventionally operates with data acquired either with laser heating (mostly in publications after 1995) or with SIMS.

Data on the isotope composition of 17 monomineralic fractions (olivine, plagioclase, clinopyroxene, and orthopyroxene) and eight samples from the so-called Critical Zone of the Dovyren massif were presented in (Orsoev et al., 2010). The Critical Zone is a unit of varying mineral composition and thickness (about 200 m on average) in the transition zone from the troctolites to olivine gabbro. The top part of the Critical Zone is made up of complicatedly interbedding meso- and leucotroctolites, leucogabbro, and related gabbro-pegmatites, which are complicated by anorthosite schlieren with disseminated low-sulfide but locally abundant PGE mineralization (Orsoev, 2019). For the first time for the Dovyren massif, the isotope composition of minerals was measured using heating with a Nd–YAG laser. The data were calibrated on the NBS-28 and NBS-30 internationally certified standards. The five \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) values obtained for the olivine varied from 4.80 to 5.8‰ (at an average of 5.1 ± 0.4‰). With regard to data on the other rock-forming minerals, the values for the rocks lied within the range of δ18O = 5.4–6.1‰ (5.8 ± 0.3‰ on average, except the gabbro-pegmatite). The relatively low \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) values and the absence of oxygen isotope equilibrium with other phases were interpreted by the authors as resulting from interaction between the rocks and meteoric waters.

MATERIALS AND ANALYTICAL TECHNIQUES

Samples for the isotope study were selected with the aim of solving two problems: determine the variations in the \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) values of the olivine along the strike of the massif and across it, following the complete stratigraphic cross-section. Considering the distribution of the least altered rocks, five samples were collected in the northeastern termination of the massif, in the Shkolnyi section, which spans a range of stratigraphic levels of approximately 800 m, i.e., more than a half of the vertical section of this part of the intrusion (Table 1). In the central part of the Dovyren massif (its generalized vertical section along the Bolshoi and Tsentralnyi transects; Ariskin et al., 2018a), twelve samples were taken, which represent more than 1000 m of the vertical section of the massif (one-third of its total thickness) up from its lower contact. In the southwestern part (Yoko section), sixteen samples were collected, which represent the practically whole stratigraphic sequence. One half of the studied rocks are ultramafics (plagiolherzolites, plagiodunites, and dunites), which are contaminated in the central part because of reactions of carbonate in host-rock xenoliths with the ultramafic magma (Wenzel et al., 2002). The other half is troctolites (varying from melano- to leucocratic) and olivine gabbro (Table 1). The losses on ignition for the whole set were 0.6 ± 1.2 wt % or, with a correction for the excess mass at the complete Fe oxidation to Fe2O3, 1.8 ± 1.3 wt % (Table 2)Footnote 1. For the sixteen samples from the Yoko section, LOI values were 0.0 ± 0.5 and 1.1 ± 0.5 wt %. Note that the 2.5 wt % constraint placed on LOI in (Goncharenko et al., 1992) to select the “freshest” samples is equivalent to ~4 wt % with a correction for complete Fe oxidation (Table 2), which corresponds to serpentinized rocks.

Table 1.   Samples in which the oxygen isotope composition of olivine was studied
Full size table
Table 2.   Petrochemical and mineralogical features and δ18O of olivine of rocks from the Yoko-Dovyren massif
Full size table

Methods. Olivine for studying its oxygen isotope composition was separated from samples in which the composition of olivine has been studied and whose whole-rock composition has been analyzed at the Central Science Laboratory at the University of Tasmania in Hobart, Australia (Ariskin et al., 2018a). Monomineralic olivine fractions (+0.25 to –0.5 mm) were hand-picked under a binocular (Zeiss Stemi 2000C) to get rid of aggregates with other minerals and admixtures, washed in acetone to remove dust, and dried. The isotope analysis was carried out with 1 to 1.5 mg of the material.

The oxygen isotope composition was analyzed at the Laboratory for Isotope Geology at the Institute of Geology of Ore Deposits, Petrography, Mineralogy and Geochemistry, Russian Academy of Sciences, by means of fluorination with laser heating (Sharp, 1990). In contrast to other methods previously applied to analyze mineral and rock samples from the Dovyren massif, this technique is free of errors related to the incomplete decomposition of the samples (this problem pertains very much to olivine) and those related to measurement standardization. The samples were decomposed with an IR CO2 laser 30 W in power (NewWave Research, MIR10-30), with subsequent oxygen extraction with bromine pentafluoride. The mass-spectrometric measurements of the extracted oxygen were made on a DELTAplus (Thermo, Germany) using a dual inlet system. The errors of the measured \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) values were no higher than ±0.1‰ (1σ). The measurements were calibrated (on the V-SMOW scale) using two internationally certified standards: UWG-2 garnet and NBS-28 quartz (Spicuzza et al., 1998).

MEASUREMENT AND CALCULATION RESULTS

The measured data on the oxygen isotope composition of the olivine are presented in Table 2, which additionally displays information on the rocks: their normative olivine concentrations, the average composition of olivine in the samples and the Cr# of the coexisting aluminochromite. Olivine from the rocks that do not exhibit any indications of contamination demonstrates \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) within a narrow range of 5.8 ± 0.2‰, whereas olivine in the contaminated dunites is slightly enriched in the heavy oxygen isotope (6.2 ± 0.3‰). These are not the highest values ever obtained for olivine from the Dovyren rocks. When the immediate contacts of the magnesian skarns were studied, it was determined that the \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) of the olivine may reaches 12–14‰, and that of forsterite in the spinel–forsterite skarns is up to 19‰ (Krivoplyasov et al., 1982). In our situation, the increase in the \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) value by ~0.4‰ led us to distinguish a tendency toward olivine enrichment in the 18О isotope where xenoliths of carbonate-replacement skarns occur in the massif, toward their contact zones with the host olivine cumulates. This tendency correlates with mineralogical indications of rock contamination with carbonates (Ariskin et al., 2018a), including elevated CaO concentrations in the olivine and a systematic decrease in the Cr# of the spinel (Fig. 2). In the central part of the Dovyren massif, such contamination effects are discernible within a broad range of depths: from the level of ~500 m to 970 m (Table 1).

Fig. 2.
figure 2

Correlation between the δ18О of olivine, CaO concentration of the olivine, and Cr# of spinel in rocks of various type from the Dovyren massif (data from Table 2). Legend: (1) uncontaminated rocks from the Yoko and Bolshoi-Tsentralnyi vertical sections; (2) contaminated dunites; (3) rocks from the Shkolnyi vertical section.

Full size image

It seems to be important that the most homogeneous oxygen isotope composition of the olivine, δ18O = 5.8 ± 0.1‰, is typical of sixteen freshest rock samples from Mount Yoko. This composition can be assumed as the oxygen isotope composition of olivine that was in equilibrium with the original Dovyren magmas. The \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) values of six rocks from the Shkolnyi section is 6.0 ± 0.4‰. The reason for the significant standard deviation in this situation is the contrasting differences between the three samples with δ18O = 5.6–5.9‰ (which is close to compositions from the Yoko section) and two samples with 6.4‰ (Table 2). The Cr# of spinel in the latter two samples indicates that the rocks enriched in the 18O isotope are candidates to rocks weakly contaminated with carbonate material, but this is inconsistent with the primary magmatic CaO concentrations of the olivine (Fig. 2). In this instance, it seems to be reasonable to suggest contamination with siliceous terrigenous material: schists or quartzites from the host rocks of the Olokit complex (Rytsk et al., 2002).

Data on the original \({{{{\delta }}}^{{{\text{18}}}}}{\text{O}}\) of the olivine can be used to evaluate the local extent of contamination of the primitive magmas and their cumulates with carbonate material when the Dovyren chamber developed and/or in the course of differentiation in this chamber. To do this, it is necessary to proceed from the oxygen isotope composition of the olivine to that of the melt. This can be done for each of the rocks if data are available on the closure temperatures of the cumulus systems in terms of oxygen isotope exchange between the olivine and cumulus melt near the solidus. This problem was solved using modeling results on the equilibrium crystallization of virtual melts whose composition corresponds to the bulk composition of the samples.

Simulation of the equilibrium crystallization of the rocks. The petrological meaning of these calculations is analogous to that of mineralogical thermometry and is underlain by the following three postulates (Ariskin and Barmina, 2004): (1) the bulk composition of a rock sample reflects the composition of a mixture of the cumulus phases and intercumulus liquid; (2) the temperature of the original cumulus mixture corresponds to that of the melt; and (3) the solidification process is associated with systematic changes in the compositions of all phases according to the thermodynamic laws of mineral–melt equilibria. Thus, algorithms for simulating magma crystallization enable one to calculate the trajectory of the compositional evolutions of the melt and equilibrium phases up to complete solidification. The problem is complicated by that the temperature of the original mixture (magma or cumulate) is usually not known, and hence, the complete crystallization trajectory shall be calculated, starting at 100% rock melting (including segments of virtual heterogeneous equilibria) to the solidus. Our situation is simpler: it is important to reach the lowest temperatures, when the last melt portions occur in equilibrium with the whole volume of solids.

The equilibrium crystallization of the rock melts (Table 2) was simulated with the COMAGMAT-5 program package (Ariskin et al., 2018b) in its version 5.2.2.1 (https://comagmat.web.ru/apps-comagmat.html) for normal anhydrous conditions, at a pressure of 1 atm and oxygen fugacity corresponding to QFM – 1 (Ariskin et al., 2017). These parameters are close to the crystallization parameters in the chamber, which were estimated by an independent technique (P ≤ 0.5–1 kbar, water content in the melt <0.5 wt %, oxygen fugacity ≤ QFM) and have already been used in a set of simulations of the parameters of the most primitive Dovyren magmas (Ariskin et al., 2016). This was done using bulk rock compositions from Supplement A in (Ariskin et al., 2018а); the maximum crystallinity was 98%. Figure 3 shows some of the calculated crystallization trajectories. The parameters of the model melts, and the calculated temperatures and olivine and plagioclase compositions near the solidus are reported in Table 3.

Fig. 3.
figure 3

Simulated crystallization sequences and the lowest temperature mineral associations of the primitive picrodolerite from the lower contact zone of the Dovyren massif (DV30-2) and representative samples in Tables 1 and 2. Simulations with the COMAGMAT-5 program package (Ariskin et al., 2018b) in an approximation of dry systems under atmospheric pressure and QFM – 1 oxygen fugacity. Only results at temperatures below that of the parental magma (1290oC) have a petrological meaning.

Full size image
Table 3.   Parameters of model melts, calculated temperatures, the olivine and plagioclase compositions, and the oxygen isotope composition of the residual melts near the solidus of the cumulates
Full size table

The crystallization sequence of sample DV30-2 is shown in Fig. 3 for comparison, because the sample is picrodolerite from the vicinity of the lower contact and is orthocumulate with the most magnesian intercumulus melt, which contained ~11 wt % MgO and ~52 wt % SiO2 at the temperature of the original magma of 1290oC (Ariskin et al., 2016, 2018a). The evolution of the intercumulus melt during the solidification of the cumulus mixes generating crystallized rocks are calculated at the lowest temperatures, and this pertains to all of the samples.

Except only one sample from the top portion of the Yoko section (07DV229-1), our simulations have demonstrated a broad virtual crystallization field of the olivine (as a result of its accumulation). The samples can be classified into two groups, including “uncontaminated” and “contaminated” rocks. In the former instance, the second phase to crystallize was plagioclase, which was followed by clinopyroxene ± orthopyroxene, i.e., in exact compliance with what was predicted by the crystallization of primitive orthocumulates from the lower contact (Fig. 3). This succession is different from that of the contaminated rocks: the second mineral to crystallize was clinopyroxene (for example, sample 07DV309-1 in Fig. 3). Correspondingly, the near-solidus residual melts evolved in the region of melts strongly undersaturated in silica (<40–30 wt % SiO2) at temperatures above 1300–1400°C. We regard these estimates as an artifact of the model, although the expansion of the stability field of Ca-pyroxene in systems contaminated with carbonate is far from surprising (Gaeta et al., 2009; Mollo et al., 2010; Di Rosso et al., 2012). This is confirmed by the fact that the rocks contain magnesian diopside (up to diopsidites) and fassaite (Wenzel et al., 2002; Ariskin et al., 2018а). With regard for the uncertainties of the simulations of melt–mineral equilibria in similar systems with the COMAGMAT-5 model, these results were rejected from further considerations. The only exception was sample 07DV309-1, data on which are also placed in Table 3.

For the uncontaminated rocks, the COMAGMAT-5 program package predicts the solidus association Ol + Pl + Cpx ± Opx, occasionally with minor sulfide and ilmenite. The temperatures at which crystallization terminated were from 1131 to 1266oC (1176 ± 34°C on average) (Table 3). These variations reflect two factors: (1) the amount of melt buried in the cumulates and (2) the extent of its fractionation in the course of solidification of the massif. Correspondingly, the minimum temperature estimates are typical of the porous basal cumulates and rock units near the roof of the massif, which contain the least magnesian olivine and pyroxenes (samples 14–16 in Table 2). The maximum estimates, above 1200oC, are typical of the olivine and troctolite cumulates with the minimum content of intercumulus material.

The plausibility of the modeling results can be estimated based on data in Fig. 4, in which near-liquidus (according to the calculations) and observed (in the rocks provisionally recognized as “uncontaminated”) olivine compositions are compared. Although the calculated compositions are obviously shifted to forsterite-poorer compositions, the average deviations are small: 0.96 ± 0.59 mol % Fo, and the discrepancies are greater (>2 mol % Fo) for a single sample of the uppermost leucotroctolite (sample 07DV229-1, Δ = –2.2%). A systematic character of this shift may be explained by a slight overestimation of Fe2+ in the melt, Fe2+/ΣFe (the simulations were conducted for fairly reducing conditions of QFM – 1) or by the absence of complete equilibrium between the very last melt portions and the crystalline cumulate masses. In any event, these discrepancies do not affect significantly the composition of the model melts, whose temperature variations depending on the redox parameters are commonly do not exceed a few grades (Ariskin and Barmina, 2004). This led us to believe that the parameters of the near-liquidus equilibria of the rocks are predicted reasonably accurately (Table 3).

The oxygen isotope composition of the parental melt was calculated based on the approximation (1)

$${{\delta }^{{18}}}{\text{O}}\left( {{\text{Ol}}} \right) - {{\delta }^{{18}}}{\text{O}}\left( {{\text{melt}}} \right) \approx 1000\ln \alpha ,$$
(1)

where α is the isotope fractionation factor between the melt and olivine, which is related to the closure temperature of isotope exchange as

$$1000\ln \alpha = {{10}^{6}}A/{{T}^{2}},$$
(2)

where A is a semiempirical constant, and T is the temperature in K. The calculations were carried out with А = –1.70, which was suggested in (Zhao and Zheng, 2003) for volcanic rocks representing systems olivine phenocryst–basaltic glass/lava. Thereby the temperature at which effective isotope exchange ceases to proceed in uncontaminated rocks was assumed as the model temperature in a closed system corresponding to equilibrium between 2% residual melt and 98% crystalline mass (Table 3). For the six contaminated dunites, the temperature of isotope exchange termination was postulated to be the near-liquidus temperature in the closed system, i.e., 1176°C. These assumptions imply that the exchange by oxygen isotopes mostly terminates when the “fluid-melt” is exhausted, and there are no reasons to believe that contaminated melts evolved in another temperature field than that of the host dunites. Note that two-thirds of the samples contained more than 70% olivine (Table 2), which implies that postsolidus reactions (such as olivine–plagioclase) did not any significantly affect the isotope composition of the olivine that crystallized during the late magmatic process.

Table 3 reports the d18О values calculated for all of the samples. According to these data, the average d18О of the intercumulus melts is 6.6 ± 0.2‰ of the uncontaminated rocks and 7.1 ± 0.3‰ for the rocks with traces of contamination. According to the specified range of the near-solidus temperatures (1131–1266oC), the average isotope shift between the olivine and model melts is 0.81 ± 0.03‰.

Fig. 4.
figure 4

Correlation between the calculated and observed olivine compositions in uncontaminated rocks of the Dovyren massif (based on data in Tables 2 and 3).

Full size image

DISCUSSION

Comparison with data on other layered massifs. Figure 5 compares data on the isotope composition of olivine from the Dovyren massif and other layered intrusions and/or mineralized complexes. According to these data, δ18О of olivine tends to increase from intrusive bodies related to tholeiitic basalts (mantle sources of MORB type) to mineralized intrusions related to reservoirs with evidence that crustal material was involved in the magma-generating processes and/or evolution of the parental magmas (Wilson, 2012; Maier et al., 2016; Ariskin et al., 2018a; Tang et al., 2018). Our data are the closest to the isotope characteristics of the Paleoproterozoic Bushveld complex, and the average δ18О values of olivine from the uncontaminated Dovyren rocks practically exactly coincide with the maximum δ18О values of olivine from the mineralized Jinchuan complex (Ripley et al., 2005). The latter has an age of 831.8 ± 0.6 and was formed, similar to the Dovyren massif, at the breakup of the Rodinia supercontinent (Zhang et al., 2010).

Fig. 5.
figure 5

Oxygen isotope composition of olivine from widely known layered intrusions and mineralized complexes: Bushveld complex (Нarris et al., 2005; Gunther et al., 2018); Junchuan (Ripley et al., 2005); Freetown intrusion, Sierra Leone (Chalokwu et al., 1999); South Kawishiwi Intrusion, Duluth complex (Lee and Ripley, 1996); Skaergaard intrusion (Taylor, 1959; Bindeman et al., 2008); and the Kiglapait intrusion (Kalamaradis, 1984). Probable sources: (1) likely subcontinental lithospheric mantle (SCLM), (2) MORB-type depleted mantle, (3–6) Yoko-Dovyren massif: (3) uncontaminated rocks, (4) average composition of the uncontaminated rocks, (5) two “likely contaminated” rocks of the Shkolnyi section, (6) contaminated dunites.

Full size image

More and more arguments are lately published that intrusive magmatism that produced the mineralized Bushveld and Jinchuan complexes had certain analogous features (Eales and Costin, 2012; Tang et al., 2018). These analogies include two contamination stages of the parental magmas (in the lower and middle crust), which may take place either before the chamber stage or when the magma chamber was formed. The mantle source of the most primitive magmas is uncertain. This source was asthenospheric or metasomatized subcontinental lithospheric mantle (SCLM). The former scenario is discussed with reference to komatiite sills in the bottom part of the Bushveld complex (Wilson, 2012; Maier et al., 2016), and the latter one was suggested for the Jinchuan ultramafic rocks. Isotopic–geochemical data on the Dovyren massif may also suggest such a double-stage scenario for the involvement of crustal material in processes that produced the mantle source and controlled differentiation in the chamber.

The very low \({{\varepsilon }}_{{{\text{Nd}}}}^{{}}\left( t \right)\) values of the Dovyren rocks (–14.8 on average, t = 728.4 Ma) cannot be explained by the assimilation of carbonates from the host rocks, which have εNd close to –7 (Amelin et al., 1996). This leaves the only possibility of admitting the occurrence of an older mantle protolith with a crustal Sm/Nd ~ 0.22 (Ariskin et al., 2015). Thereby the high Mg# of the original olivine (which was calculated for rocks in the basal zone as ~Fo88) indicates that the Dovyren magmas with crustal signatures possessed a picritoid composition at temperatures of about 1300oC (Ariskin et al., 2018a). These signals of “mantle–crustal” processes correlate with the elevated (relative to the MORB system) δ18О values of olivine from the uncontaminated rocks (Fig. 5).

At the same time, in addition to the fairly broad variations in the δ18О of olivine from the uncontaminated rocks, the Dovyren intrusive complex shows evidence of heterogeneities of the Rb–Sr and Sm–Nd systems (Ariskin et al., 2015), which also cannot be explained by the prehistory of the source or by crystallization differentiation alone. This is particularly well seen in the dunite units with xenoliths of carbonate rocks (Wenzel et al., 2002). Despite the long-lasting discussion, the extent of contamination with marble/dolomite and the effects of these processes on differentiation in the Dovyren chamber, i.e., eventually on the development of the phase layering of the massif, are still largely uncertain.

Estimation of the extent of contamination with carbonate material. Interaction between basaltic magmas and carbonates is commonly described by desilication reactions, including the thermal decomposition of carbonates with the formation of Ca-pyroxene solid solutions (e.g., Mollo et al., 2010):

$${{L}_{1}} + {\text{CaC}}{{{\text{O}}}_{3}} \to {{L}_{2}} + Di + CaTs + {\text{C}}{{{\text{O}}}_{2}} \uparrow ,$$
(3)

where L1 is the uncontaminated melt, Di is diopside (CaMgSi2O6), CaTs is pyroxene rich in the Tschermakite component, and L2 is the newly formed melt depleted in SiO2 and somewhat enriched in CaO. The parental picritoid magmas of the Dovyren massif contained about 20–30% olivine (Ariskin et al., 2018a), and hence, such reactions should also take into account the presence of cumulus olivine and changes in its composition: an increase in its Mg# and CaO concentration (Wenzel et al., 2002; Di Stefano et al., 2018). This heterogeneity of the reactants and interaction products hampers the successive solution of the oxygen isotope balance with the aim of estimating how much the cumulus associations are contaminated. In a simplified form, this material balance can be described by the equation

$${{\delta }^{{18}}}{\text{O}}\left( {{{L}_{2}}} \right) = f{{\delta }^{{18}}}{\text{O}}\left( {{\text{cont}}{\text{.}}} \right) + \left( {1 - f} \right){{\delta }^{{18}}}{\text{O}}\left( {{{L}_{1}}} \right),$$
(4)

where δ18O(L1) and δ18O(L2) are the oxygen isotope compositions of the uncontaminated and contaminated melts, respectively, δ18O(cont.) characterizes the oxygen isotope composition of the carbonate material, and f is the fraction of the contaminant (i.e., the total amount of CaO + MgO + FeO provided by carbonate decomposition for the newly formed silicate phases). Assuming the average δ18О values for the contaminated (7.1‰) and uncontaminated (6.6‰) melts and δ18О = 23‰ for the carbonates (Krivoplyasov et al., 1982), we estimated the contribution of the contaminant at ~3.2%. For the heterogeneous cumulus system, the estimated fraction of the contaminant is 0.3–0.4% lower at ~40% olivine crystals and 0.7–1% lower at ~90% crystals.

The aforementioned balance assumes that the oxygen isotope composition of the carbonate decomposition products is either constant or varies insignificantly [see reaction (3)]. In fact, theoretical estimates indicate that oxygen isotope fractionation at temperatures of about 1200°C in the CO2–CaCO3 system may reach ~2‰ (Chacko and Deines, 2008). This increases the maximum estimated fraction of the contaminant to ~5 and to ~4% at a content of olivine crystals of 90%. Hence, the new melt or cumulus system whose isotope characteristics correspond to equilibrium with olivine from the contaminated dunite zone (Tables 2 and 3) can be obtained by at least 3–5% contamination with dolomite.

Assuming the isotopically heaviest oxygen isotope composition of the contaminated melt (say, δ18O = 12‰) in equilibrium with olivine from contact zones with the carbonate-replacement skarns (δ18O ≈ 11; Krivoplyasov et al., 1982) shall require approximately 30% contaminant. However, such significant contamination is local and is seen as the development of contaminated dunite aureoles around skarnized marble and dolomite relics. This is also evident from the intercumulus composition in these aureoles, in which diopside and fassaite (pyroxene with high alumina concentrations; Wenzel et al., 2002) are the most magnesian, the olivine has higher CaO concentrations and Mg#, and the Al-spinel composition changes (Kislov, 1998; Ariskin et al., 2018a). Reactions between the original magma and carbonates are reflected only in the expansion of the clinopyroxene stability field, as was determined by simulation the equilibrium crystallization of contaminated rocks with the COMAGMAT-5 program package.

CONCLUSIONS

(1) Fluorination with laser heating was applied to study the oxygen isotope composition of olivine in plagiolherzolites, dunites, troctolites, and olivine gabbronorites of the Yoko-Dovyren layered massif. Most of the δ18O values of olivine from the freshest and uncontaminated rocks (n = 27) lie within the range of 5.8 ± 0.2‰. The weakly and strongly contaminated dunites show a notably heavier oxygen isotope composition: up to 6.2 ± 0.3‰. Such compositions are demonstrated to be the closest to those in other mineralized complexes such as Bushveld and Jinchuan.

(2) The COMAGMAT-5 software (Ariskin et al., 2018b) was applied to simulate the crystallization trajectories of the cumulus rocks, including the equilibrium phase proportions at temperatures corresponding to 2% residual melt. These near-solidus temperatures span the range of 1131–1266°C (1176 ± 34°C on average) and are viewed as the temperatures at which the cumulus systems became closed with respect to oxygen isotope exchange between the olivine and melt after the intercumulus melt as mobile “fluid” has been exhausted.

(3) Data on the oxygen isotope composition of the olivine and model temperatures of the closure of this mineral, as well as δ18O are calculated for each of the rocks for melts in equilibrium with the olivine. The average δ18O value was 6.6 ± 0.2‰ for the intercumulus melts of the uncontaminated rocks, and 7.1 ± 0.2‰ for the melts with traces of contamination with crustal material.

(4) According to estimates with the balance model for melt contamination with dolomite, the main dunite volume in the core of the massif was no more than 3–5 wt % contaminated. The isotope characteristics of melts in aureoles near contacts with xenoliths of carbonate-replacement skarns indicate that these melts may have dissolved up to 30% carbonates.