Origin of the primitive, strongly SiO₂-undersaturated alkalic rocks from the Deccan Traps by low-degree mantle melting and high-pressure fractional crystallization

Abstract

Strongly SiO₂-undersaturated alkalic rocks (Mg# > 50, SiO₂ ≤ 45 wt%, Na₂O + K₂O ≥ 3 wt%) occur in three early-stage (Sarnu-Dandali, Mundwara, Bhuj) and one late-stage (Murud-Janjira) rift-associated volcanic complexes in the Cretaceous-Paleogene Deccan Traps flood basalt province of India. Thermobarometry based on clinopyroxene-liquid equilibrium suggests that they mostly crystallized beneath the Moho at ~ 15 kbar/1270 °C to ~ 11–12 kbar/1115–1156 °C pressures and temperatures. Primary magma compositions in equilibrium with lherzolite were estimated through reverse fractionation calculations by incrementally adding equilibrium phases to the rocks in olivine:clinopyroxene:spinel:phlogopite = 12:68:20:15 proportions at low temperatures followed by olivine:clinopyroxene:spinel = 12:68:20 proportions at higher temperatures. A comparison of the primary magmas with experimentally generated melts shows that their compositions are consistent with an origin from garnet lherzolite sources with < 1 wt% H₂O and CO₂. Hornblendite, pyroxenite (except for some Bhuj rocks) and carbonated eclogite are unlikely sources for the Deccan alkalic rocks. The Sarnu-Dandali and Bhuj alkalic rocks and the Murud-Janjira lamprophyres probably originated by < 5% melting of ~ 1.3 times Ti-enriched lherzolitic sources compared to primitive mantle. The primary magmas of the Murud-Janjira basanites calculated through reverse assimilation-fractional crystallization by assimilating lower crustal and mantle xenoliths found in younger lamprophyre dikes of the same area indicate that contamination by the Indian lithosphere was unlikely during their ascent. The basanites evolved by mixing with phonotephritic melts, and they probably originated from a Ti-poor (0.7 times) lherzolite source. The temperature of the melts at the base of the lithosphere was ~ 1325 °C beneath Sarnu-Dandali and ~ 1285 °C beneath Bhuj and Murud-Janjira.

Introduction

Continental flood basalt provinces are largely composed of tholeiites that originate by voluminous melting of the mantle as plumes ascend through upper mantle and encounter temperatures above the solidus (Morgan 1972; McKenzie and Bickle 1988; Richards et al. 1989; Campbell and Griffith 1990). By contrast, small amounts of alkalic volcanic rocks (e.g. nephelinite, basanite, tephrite, phonolite, trachybasalt, trachyte, etc.) present in flood basalt provinces are the products of low-degree mantle melting (Gast 1968; Green 1973). Determining the mantle source characteristics, degree of melting, and the pressure–temperature (P–T) conditions for the origin of alkalic rocks is an active area of petrological research. According to some experimental studies, the source is deep (~ 2.5–3.0 GPa) garnet lherzolite ± H₂O ± CO₂ (Dasgupta et al. 2007, 2013; Baasner et al. 2016; Condamine et al. 2016), whereas other experiments suggest a relatively shallow (~ 1.5 GPa) metasomatized lithospheric mantle as the likely source (Pilet et al. 2008, 2010). In addition, experiments have been conducted on garnet pyroxenite (Kogiso et al. 2003; Hirschmann et al. 2003) and eclogite ± CO₂ believed to be recycled into the mantle (Pertermann et al. 2004; Dasgupta et al. 2006) to assess whether these lithologies can be the potential sources of alkalic rocks. This study attempts to determine the origin of primitive (molar 100.Mg/(Mg + Fe²⁺), or Mg# > 51), strongly SiO₂-undersaturated alkalic rocks (SiO₂ = 36–45 wt %, Na₂O + K₂O = 3–6 wt%, all oxides recalculated on a volatile-free basis) from the Deccan Traps of India by comparing the compositions of their bulk rock and calculated primary magmas (Mg# 72, in equilibrium with Fo₉₀ olivine) with melts produced in high-pressure experiments.

Although some rare Deccan picrites from Saurashtra (Fig. 1) containing Fo_88-92 olivine may be the progenitors of the locally associated tholeiitic basalts (Krishnamurthy and Cox 1977; Melluso et al. 1995), low-pressure reverse fractional crystallization modeling shows that the primary magmas of most Deccan tholeiites are primitive tholeiites that last equilibrated with lherzolitic mantle near the crust–mantle boundary (Chatterjee and Sheth 2015, using the methods of Till et al. 2012). A similar reverse fractionation approach is adopted here to calculate the primary magmas of the Deccan alkalic rocks. However, the Deccan tholeiites differentiated by fractionating gabbroic assemblages at low pressures (< 0.2 GPa, Bhattacharji et al. 1996; Sano et al. 2001), whereas thermobarometry in this study suggests that the Deccan alkalic rocks differentiated at much higher pressures (> 1.2 GPa). Hence, the appropriate assemblages and phase proportions that crystallize at high pressures from primitive alkalic melts (as in the experiments of Holbig and Grove 2008; Pilet et al. 2010) are recombined with the Deccan alkalic rocks to model their primary magmas. At the end of each calculation, the melt (Mg# 72) is in equilibrium with lherzolite containing Fo₉₀ olivine. The mantle melting conditions for the origin of the Deccan alkalic rocks constrained by lherzolite melting experiments and geophysical data are discussed.

Fig. 1

Google satellite image of the Indian Shield region overlain by a map showing the areal extent of Deccan Traps with rift zones, faults, location of alkalic complexes, and important places

Geological background

The Deccan Traps of central and western India constitute one of the largest (≥ 500,000 km²) continental flood basalt provinces on Earth with thick, near-horizontal tholeiitic lava sequences and rift-related minor alkalic rocks (Wadia 1975; Fig. 1). The bulk of the tholeiitic basalts at Western Ghats (Beane et al. 1986; Lightfoot et al. 1990) was erupted within a ~ 1 million year interval that straddles the Cretaceous-Paleogene boundary (66.4–65.4 Ma, Schoene et al., 2019; Sprain et al., 2019). By contrast, the alkalic rocks associated with the NNW–SSE oriented Barmer-Cambay Rift (68.6–68.5 Ma old Sarnu-Dandali and Mundwara complexes, Chandrasekaran et al. 1990; Basu et al. 1993) and the E–W-oriented Kachchh Rift (67.0–66.8 Ma old Bhuj alkalic plugs, Krishnamurthy et al. 1999; Courtillot et al. 2000) in the northwestern part were emplaced before, and the alkalic rocks at the rifted western continental margin (Murud-Janjira plugs and dikes, Dessai et al. 1990) were emplaced after the Western Ghats tholeiites. It is worth noting that the Sarnu-Dandali and Mundwara complexes have protracted magmatic histories and also contain older alkalic rocks unrelated to Deccan (c. 110–80 Ma, Pande et al. 2017; Sheth et al. 2017).

The Deccan flood basalt province probably formed when the northward drifting Indian plate was over a mantle hot spot (Deccan plume) (Devey and Lightfoot 1986; Mitchell and Widdowson 1991). Extreme upwarping of the Moho (~ 18 km depth) and lithospheric thinning (~ 40 km depth) are evident at the rifted western Indian continental margin including the Murud-Janjira area located ~ 300 km south of the postulated Deccan plume center (Negi et al. 1992; Bhattacharji et al. 1996 and references therein). Thinned crust (Cambay: 28–35 km, Kachchh: 29–43 km) and lithosphere (Cambay: 85–90 km, Kachchh: 63–69 km) within the rift zones of the northwestern Deccan province close to the northern periphery of the Deccan plume (at a distance of ~ 400–530 km from the center) have been also attributed to mantle upwelling (Rao et al. 2015; Radha Krishna et al. 2002; Mandal and Pandey 2011).

Strongly SiO₂-undersaturated Deccan alkalic rocks

The Sarnu-Dandali and Mundwara primitive alkalic rocks show Reunion plume-type high ³He/⁴He ratios (Basu et al. 1993). Their range of initial ¹⁴³Nd/¹⁴⁴Nd ratios is 0.5127–0.5128 (Simonetti et al. 1998). In an Sr–Nd isotope plot, the Sarnu-Dandali primitive rocks (initial ⁸⁷Sr/⁸⁶Sr: 0.7043–0.7045, Basu et al. 1993; Simonetti et al. 1998) plot near the depleted end (close to Reunion lavas) of a mixing array between Reunion-type mantle and continental lithosphere indicating that their source was not significantly influenced by the continental lithosphere. However, the higher initial ⁸⁷Sr/⁸⁶Sr ratios of the Mundwara samples considered in this study (0.7045–0.7048, Basu et al. 1993; Simonetti et al. 1998) indicate that their source formed by mixing of a Reunion-type mantle with a lithospheric component.

The Sarnu-Dandali alkalic rocks occur as lava flows, and plugs and dikes intruding older plutons (Chandrasekaran et al. 1990; Sheth et al. 2017). In a total alkali versus silica (TAS) diagram (Le Bas et al. 1986), most of the primitive samples (Chandrasekaran et al. 1990; Simonetti et al. 1998; Vijayan et al. 2016) plot in the foidite field (Fig. 2a), the evolved samples plot in the phonolite field, and a few other samples plot in the tephrite/basanite, trachyandesite and trachyte fields (not shown). They show high Ce/Y ratios compared to the common Deccan tholeiites (Wai Subgroup, Fig. 2f), a feature consistent with origin from a garnet-bearing mantle source. The Ni and Cr contents are positively correlated with MgO (Fig. 2g, h). The lowest-SiO₂ rocks with < 6.2 wt% Na₂O + K₂O (K₂O = 1.4–2.6 wt%, Na₂O/K₂O = 1.2–2.1) are also the most primitive (SiO₂ = 36.3–39.6 wt%, MgO = 8.2–11.5 wt%, Mg# 52–65, Fig. 2c, Table 1). These rare rocks are variously reported as melanephelinite (Vijayan et al. 2016), alkali pyroxenite, micro-melteigite (Chandrasekaran et al. 1990) and melilitite with fine-grained microporphyritic textures (Simonetti et al. 1998). Their TiO₂ (3.5–6.4 wt%) and CaO (13.8–18.1 wt%) contents, and CaO/Al₂O₃ ratios (1.3–1.6) are the highest among all Deccan rocks. They also contain 1.0–2.5 wt% H₂O and ~ 0.4 wt% CO₂, determined by gravimetric methods (Chandrasekaran et al. 1990). Carbonatites are commonly associated with the Sarnu-Dandali alkalic rocks (Chandrasekaran et al. 1990).

Fig. 2

Bulk compositions of the Deccan primitive alkalic rocks recalculated on a volatile-free basis and experimentally derived melts (refer to legends in b and d); data sources as in text for the alkalic rocks, Beane et al. (1986) and Lightfoot et al. (1990) for the average Ce/Y ratios of Wai Subgroup basalts, and the experiments of D06: Dasgupta et al. (2006), H03: Hirschmann et al. (2003) and P08: Pilet et al. (2008). In a, the boundaries between different rock-types are from Le Bas et al. (1986) and the subalkalic-alkalic boundary is from Macdonald and Katsura (1964). The dotted lines in (a, b and d) represent model trends for the evolution of strongly SiO₂-undersaturated Cape Verde lavas (Weidendorfer et al. 2016). The melting trends for hornblendite, Cpx-hornblendite, eclogite + CO₂, and garnet pyroxenite are shown by dashed lines with arrows. Hbdite hornblendite, Hbdite/DMM hornblendite in equilibrium with depleted MORB mantle, Eclog eclogite

Table 1 Summary of bulk compositions of the primitive, strongly SiO₂-undersaturated alkalic rocks from the Deccan Traps

The Mundwara alkalic rocks occur as dikes intruding three distinct plutons (Pande et al. 2017 and references therein). According to the TAS classification, they mostly belong to the tephrite/basanite-phonolite series (Fig. 2a), and a few are basalts and trachybasalts (Simonetti et al. 1998; Pande et al. 2017). The most primitive rocks contain 7.3–8.4 wt% MgO (Mg# 51–58) and 41.4–44.2 wt% SiO₂ (Table 1). They have higher SiO₂, but similar total alkali, K₂O, TiO₂, Ce/Y, Ni and Cr values compared to the Sarnu-Dandali rocks (Fig. 2). Their REE contents are consistent with 0.5–2.0% melting of a garnet peridotite source (Basu et al. 1993).

The Bhuj primitive alkalic rocks, some of which contain spinel lherzolite xenoliths, occur as plug-like bodies belonging to a monogenetic volcanic field in the central part of the Kachchh Rift (Kshirsagar et al. 2011 and references therein). In the TAS diagram, most of the primitive samples (Simonetti et al. 1998; Karmalkar et al. 2005; Kshirsagar et al. 2011) plot in the tephrite/basanite field, and a few plot in the foidite field (Fig. 2a). Other samples (Paul et al. 2008; Sen et al. 2009) plot in the basalt and trachybasalt fields (not shown). The samples from Simonetti et al. (1998) have similar initial ⁸⁷Sr/⁸⁶Sr (0.7036–0.7040) and ¹⁴³Nd/¹⁴⁴Nd (0.5128–0.5129) ratios to the Reunion and enriched Central Indian Ridge (CIR) basalts, indicating an absence of lithospheric contamination. The primitive samples are composed of fine- to medium-grained, feebly porphyritic lava with 10–20% phenocryst content (Simonetti et al. 1998; Krishnamurthy et al. 1999). They contain 9.3–15.9 wt% MgO (Mg# 57–72). Compared to the Sarnu-Dandali primitive rocks, the Bhuj primitive rocks contain higher SiO₂ (39.6–45.2 wt%), similar K₂O (0.7–2.5 wt%), and lower TiO₂ (2.6–3.8 wt%), CaO (10.0–12.3 wt%), CaO/Al₂O₃ and Ce/Y (Fig. 2a-f, Table 1). They cover a broader range of Na₂O/K₂O ratios (1.0–4.8) than the Sarnu-Dandali rocks, and their range of Ba/Nb ratios is 8–18. They contain 1.3–3.3 wt% H₂O and 0.07–0.24 wt% CO₂ determined by gravimetric methods (Karmalkar et al. 2005). They also contain high Ni (168–464 ppm) and Cr (334–995 pm) that are positively correlated with MgO (Fig. 2g,h) (Simonetti et al. 1998; Sen et al. 2009). Except for some diorites from Girnar (Bose, 1973), no other Deccan alkalic rocks have such high Ni contents. Incompatible element patterns of the Bhuj alkalic rocks generally resemble ocean island basalts (Sen et al. 2009), and their REE ratios indicate an origin by 2–6% melting of garnet peridotite (Karmalkar et al. 2005).

The Murud-Janjira lamprophyres, which also contain lower crustal and upper-mantle xenoliths derived from the Indian continental lithosphere, occur as N–S-oriented dikes cross-cutting older tholeiitic dikes and flows about 50–80 km south of Mumbai along the coast (Dessai et al. 1990, 2004; Melluso et al. 2002 and references therein). Four of the nine lamprophyre samples from Dessai and Viegas (2010) have primitive compositions (SiO₂ = 41.9–44.7 wt%, MgO = 9.2–12.1 wt%, Mg# 56–60, Table 1). They contain 3.5–4.8 wt % TiO₂ and 1.5–2.2 wt% K₂O (Na₂O/K₂O = 1.0–2.4), and show LILE enrichment (Ba/Nb = 4–11), high Zr (164–475 ppm) and Nb (39–209 ppm) contents with low Zr/Nb ratios (~ 2–4), and very high Ce/Y ratios indicating a garnet-bearing source (Fig. 2f). The range of Ni and Cr contents is 117–269 ppm and 111–584 ppm, respectively (Fig. 2g, h), though the Cr content of a basaltic lamprophyre is as high as 991 ppm (not shown). Based on their high Zr and Nb contents, Melluso et al. (2002) suggested low degrees (2–3%) of non-modal fractional melting of an enriched mantle source for the origin of the lamprophyres.

The Murud-Janjira nepheline-rich rocks are younger than the tholeiites and older than the lamprophyres. They occur in an E–W-oriented plug that cross-cuts the older tholeiitic dikes, and N–S-oriented dike or stock-like intrusions at the coast (Dessai and Viegas 2010; Melluso et al. 2002 and references therein). Two of the three samples from the plug (RG3 and RG1, Dessai et al. 1990; Dessai and Viegas 2010) are primitive basanites (Mg# 65–68) that contain 1.4–2.4 wt% TiO₂ and 3.0–5.2 wt% Na₂O (Na₂O/K₂O = 1.4–4.4) with highly enriched LILE patterns and variable Ce/Y ratios (Table 1, Fig. 2). Their Zr/Nb ratios are similar to the lamprophyres, but their Ba/Nb ratios (16–37) are much higher and they contain much lower Zr (59–195 ppm), Nb (13–68 ppm), La, Ce and Y. The third sample from the plug (RG2, Dessai et al. 1990) is a phonotephrite. Thin veins of carbonatite occur within the plug. The stock-like intrusion consists of melanephelinite, nephelinite and tephriphonolite (Dessai et al. 1990; Melluso et al. 2002). The melanephelinites and nephelinites show low initial ⁸⁷Sr/⁸⁶Sr (0.7045–0.7049) as well as low initial ¹⁴³Nd/¹⁴⁴Nd (0.5117–0.5119) ratios indicating mixing between primitive mantle and a low ¹⁴³Nd/¹⁴⁴Nd contaminant that may have been derived from delaminated continental mantle or “Lewisian-type” lower crust (Melluso et al. 2002).

Petrography and mineral chemistry

Mineral and groundmass compositions were analyzed on a JEOL JXA-8200 Superprobe electron probe microanalyzer (EPMA) at the Massachusetts Institute of Technology, Cambridge, MA, USA. Back-scattered electron (BSE) images and elemental X-ray maps were acquired with the same instrument. The EPMA was operated using an accelerating voltage of 15 kV and a beam current of 10 nA, and the raw data were corrected for matrix effects using the CITZAF package (Armstrong 1995). The compositions of the phases and groundmass are listed in Table 2 and the Online Resource Supplementary Tables S1–S3.

Table 2 Compositions of minerals and groundmass in the Sarnu melanephelinite sample SD11

The Sarnu melanephelinite sample SD11 has a porphyritic texture with large phenocrysts dominated by clinopyroxene (Cpx), and subordinate olivine (Ol), biotite (Bt) pseudomorphs, and spinel (Spl) (Fig. 3). Approximately, 50% by volume of the rock is composed of a fine-grained groundmass. Clinopyroxene is strongly zoned and shows distinct inner core (avg. En₄₇Fs₉Wo₄₅), outer core/mantle (avg. En₄₂Fs₁₀Wo₄₇) and rim (avg. En₃₉Fs₁₁Wo₅₀) domains with core-to-rim oscillatory trends of decreasing MgO and Na₂O and increasing Al₂O₃ and TiO₂ (Figs. 4a, 5). The jadeite content (< 5%) decreases from core to rim (Fig. 5d). Olivine is also zoned with a high-Mg inner core (avg. Fo₈₄) and a lower-Mg outer core (avg. Fo₈₁) and rim (Fig. 4b, 6a). The NiO content decreases, and the MnO and CaO contents increase from core to rim (Fig. 6b). The spinel phenocrysts consist of a Cr–Al-rich core and a Fe–Ti-rich rim (Figs. 4c, 6c). Spinel also occurs as inclusions in clinopyroxene and olivine. The biotite pseudomorphs are recrystallized to the groundmass phases (Fig. 4d) and show a Mg-rich average composition (Mg/(Mg + Fe^T) = 0.65, where Fe^T is total Fe). The groundmass contains nepheline, fluorphlogopite, biotite and Fe-Ti oxides. Nepheline (15.8 wt% Na₂O, 6.8 wt% K₂O) is partially altered to natrolite. Fluorphlogopite (Mg/(Mg + Fe^T) = 0.88 to 0.82) is rich in BaO (3.0–8.9 wt%), TiO₂ (4.3–5.9 wt%) and F (5.6–5.1 wt%) and shows Ba and Ti incorporation by coupled substitutions (Fig. 6d, e) (cf. Tischendorf et al. 2007). Neither the groundmass biotite (Mg/(Mg + Fe^T) = 0.64) nor the biotite pseudomorphs contain significant Ba, Ti and F.

Fig. 3

a BSE image and b Mg x-ray map of Sarnu melanephelinite SD11 showing a porphyritic texture with large phenocrysts surrounded by a fine-grained groundmass. In (a), the bright grains are spinel and Fe-Ti oxides. In (b), the yellow-green Mg-rich crystals are olivine, and the euhedral, light-blue Mg-poor crystals are clinopyroxene. The images have a resolution of 10 micron, and were acquired using a stage-raster mode with a stationary beam

Fig. 4

BSE images showing phenocrysts of a clinopyroxene, b olivine, c spinel, and d recrystallized biotite pseudomorph surrounded by a fine-grained groundmass. Core-to-rim radial lines in (a–c) indicate the locations of compositional transects in Figs. 5 and 6

Fig. 5

a, b Core-to-rim compositional variation in clinopyroxene along transect shown in Fig. 4a. c Clinopyroxene compositions plotted in the enstatite (En)-diopside (Di)-hedenbergite (Hd)-ferrosilite (Fs) pyroxene quadrilateral. d Clinopyroxene compositions expressed as quadrilateral and non-quadrilateral (Jd jadeite, and Ae aegirine) components and plotted according to Morimoto et al. (1988). In (c, d), the data from the experiments of HB08 (Holbig and Grove 2008) and EG03 (Elkins-Tanton and Grove 2003) are shown for comparison

Fig. 6

a–c Core-to-rim compositional variation in olivine and spinel along transects shown in Fig. 4b–e Compositions of mica showing incorporation of Ba and Ti by interlayer-tetrahedral and tschermak-type coupled substitutions

Thermobarometry

Thermobarometry was carried out using the clinopyroxene-liquid equilibrium formulations of Putirka (2008) with the bulk composition from Vijayan et al. (2016) (Table 3). Equilibrium was assessed on the basis of the mineral-bulk K_D(Fe²⁺–Mg) exchange coefficient with bulk Fe²⁺ estimated from the bulk Fe³⁺/$\sum {}$Fe ratio using the Fe–Mn–Mg exchange olivine-melt oxybarometer of Blundy et al. (2020). Since the oxybarometer formulation requires an equilibrium temperature, an initial temperature estimate was obtained by considering all Fe as Fe²⁺ and using the jadeite-diopside/hedenbergite (Jd-DiHd) formulations for Cpx-hydrous liquid equilibrium [Eqs. 30 and 33 of Putirka (2008), uncertainties: ± 3.6 kbar and ± 45 °C from regression analysis of experimental data] with loss-on-ignition as the H₂O content of the bulk. The Cpx-bulk K_D(Fe^T–Mg) for the average and the outer core of Cpx are 0.28 and 0.29, and the estimated P–T are 14.9 kbar/1266 °C (average Cpx) and 15.5 kbar/1272 °C (Cpx outer core) (Table 3, S1). At 1270 °C, the lattice strain model (Blundy et al. 2020) predicts an equilibrium K_D(Mn–Mg) of 0.25 for the outer core of olivine that is within the uncertainty of the observed value (0.32 ± 0.08, based on uncertainties in Mn measurements with XRF and EPMA). Hence, the olivine outer core (Fo_81, olivine-bulk K_D(Fe^T–Mg) = 0.27) was used to calculate the Fe³⁺/$\sum {}$Fe ratio of the bulk. The bulk Fe³⁺/$\sum {}$Fe ratio is 0.16, and the corresponding olivine-bulk K_D(Fe²⁺–Mg) of 0.32 is within the uncertainty of the equilibrium value (0.30 ± 0.03, Roeder and Emslie 1970). Considering a bulk Fe³⁺/$\sum {}$Fe ratio of 0.16, the K_D(Fe²⁺-Mg) values for average Cpx-bulk (0.33) and Cpx outer core bulk (0.34) are within the uncertainty of the equilibrium Cpx-liquid K_D(Fe²⁺–Mg) value (0.28 ± 0.08, or 0.295 calculated with Eq. 35, Putirka, 2008). Hence, the P–T results calculated above are valid for Cpx-liquid equilibrium.

Table 3 Thermobarometry of the Deccan alkalic rocks

Equation 30 of Putirka (2008) is based on a selected, high-quality experimental data set. When Eq. 31 (Putirka 2008, uncertainty: ± 2.9 kbar) based on a global experimental data set is used, the estimated pressures are lower (12.2 kbar/1250 °C–12.6 kbar/1255 °C) but within the ± 3.6 kbar uncertainty for Eq. 30 (Table 3). On the other hand, when the Jd-DiHd formulations for the Cpx-anhydrous liquid exchange (Putirka et al. 1996) are used, the pressures are higher (17.0–17.6 kbar) but also within the ± 3.6 kbar uncertainty of the pressures calculated with the Cpx-hydrous liquid exchange equilibrium. Furthermore, the Al-exchange formulation between Cpx and hydrous liquid [Eq. 32c of Putirka (2008), uncertainty: ± 5 kbar] yields higher pressures (21.3 kbar/1308 °C, though with overlapping uncertainties) than the pressures obtained using the Jd-DiHd formulations (Table 3). Thus, the P–T of Cpx crystallization is probably ~ 15 kbar/1270 °C within ± 3.6 kbar and ± 45 °C uncertainties.

Pressures and temperatures calculated simultaneously using only the Cpx composition [4.8 kbar/1123 °C-5.4 kbar/1131 °C, with Eqs. 32a and 32d of Putirka (2008), uncertainties: ± 3.1 kbar and ±  ~ 70 °C] are lower by ~ 10 kbar and ~ 150 °C than the P–T calculated with the Cpx-liquid formulations. Equation 32a is based on the model of Nimis (1995) that recovers anomalously low pressures, e.g., the mean pressure is ~ 3 kbar lower at the experimental temperature conditions for 15 kbar experiments on a variety of compositions (Putirka 2008). In particular, simultaneous application of the Cpx-only formulations yields anomalously low P–T values for clinopyroxene (similar to Cpx in SD11, Fig. 5c, d) in equilibrium with low-SiO₂ alkalic liquids in experiments at 14–16 kbar (Holbig and Grove 2008) (lower by 6–9 kbar/80–140 °C for) and 24–33 kbar (Elkin-Tanton and Grove 2003) (lower by 7–12 kbar/70–110 °C). It appears that simultaneous P–T calculation with the Cpx-only formulations underestimates pressure by ~ 10 kbar for the Deccan alkalic rocks.

Thermobarometry of samples from previous studies

Thermobarometry was also performed on other samples using previously published data. The Sarnu melilitite sample 95-BMR-37 (Simonetti et al. 1998) shows Cpx-bulk K_D(Fe^T–Mg) values of 0.28–0.30 for Cpx grains 1, 3 and 5, and the calculated P–T (13.6–15.2 kbar/1267–1284 °C) using Eqs. 30 and 33 of Putirka (2008) are similar to the calculated P–T for sample SD11 (Table 3). The lattice strain model (Blundy et al. 2020) predicts a K_D(Mn–Mg) of 0.25 at 1270 °C for olivine grain-3 (Simonetti et al. 1998) that is identical to the observed value (0.23 ± 0.08), and the calculated bulk Fe³⁺/$\sum {}$Fe ratio is 0.10. Using this Fe³⁺/$\sum {}$Fe ratio, the K_D(Fe²⁺–Mg) values for olivine-bulk (0.32) and Cpx-bulk (0.31–0.33) show equilibrium and indicate that the above P–T estimates are valid.

Two Mundwara samples (MM03 and MER06, Pande et al. 2017) show Cpx-bulk K_D (Fe^T–Mg) values of 0.23–0.30, and Eqs. 30 and 33 (Putirka 2008) yield P–T of 3.7–8.8 kbar/1123–1148 °C (Table 3). However, olivines in these samples with olivine-bulk K_D(Fe^T–Mg) of 0.36–0.87 are probably in disequilibrium, and the bulk Fe³⁺/$\sum {}$Fe ratios were not calculated. Assuming an average bulk Fe³⁺/$\sum {}$Fe ratio of 0.13, the Cpx-bulk K_D (Fe²⁺–Mg) are 0.26–0.35 indicating equilibrium, and the P–T estimates are probably valid. Pande et al. (2017) calculated similar pressures (3–7 kbar) at 935–1030 °C for amphibole crystallization in their Mundwara samples MM11 and MM20. However, pressure obtained with the Al-exchange barometer (Eq. 32c) along with the Jd-DiHd thermometer (Eq. 33) yields higher pressures of 6.3–12.3 kbar.

The Bhuj basanite 95-BHUJ-16 (Simonetti et al. 1998) and the Murud-Janjira basanite RG3 and phonotephrite RG2 (Dessai and Viegas 2010) show high Cpx-bulk K_D(Fe^T–Mg) values (0.47–0.55 and 0.44–0.48). If some Fe is considered as Fe³⁺, the Cpx-bulk K_D(Fe²⁺–Mg) values are even higher. Hence, Cpx does not have an equilibrium Fe–Mg distribution with the bulk and the Jd-DiHd Cpx-liquid formulations are not applicable to these samples. The Cpx-only formulations (Eqs. 32a and 32d, Putirka 2008) yield pressures of 1.2–2.2 kbar that are probably ~ 10 kbar lower than the actual values (see above). Thus, these samples probably crystallized Cpx at pressures of ~ 11–12 kbar and corresponding temperatures of 1115–1156 °C (Table 3). The high pressures for the Bhuj sample are supported by results obtained with the Al-exchange barometer for Cpx-hydrous liquid that is independent of the Fe–Mg distribution. This barometer (Eq. 32c) in conjunction with the Cpx-only thermometer (Eq. 32d) yields pressures of 13.8–15.3 kbar.

Primary magmas

The primary magma compositions for primitive rocks have been traditionally determined by adding equilibrium olivine to the rock to return it to its Fo₉₀-equilibrated composition appropriate for lherzolite. Although this method works for rocks that have undergone only olivine fractionation, Till et al. (2012) demonstrated that for more evolved rocks, it is necessary to add other phases like plagioclase (Pl) and Cpx that crystallized during fractionation (e.g., Ol + Pl and Ol + Pl + Cpx). Chatterjee and Sheth (2015) calculated the primary magma compositions of Deccan tholeiites that evolved at upper crustal pressures by adding Ol + Pl + Cpx followed by Ol + Pl, and Ol only in small increments while maintaining crystal–liquid equilibrium to backtrack along their low-pressure fractionation paths until the liquids were in equilibrium with phases in lherzolite (with Fo₉₀ olivine). A similar approach was used here to calculate the primary magmas of the Deccan alkalic rocks. Thermobarometry suggests that most Deccan alkalic rocks underwent fractional crystallization at pressures of 12–15 kbar, and according to experimental constraints, primitive alkalic magmas evolve by crystallizing olivine and clinopyroxene in proportions of 11:89 at 1.4–2.2 GPa and 15:85 at 1.2 GPa between ~ 1350 and ~ 1290 °C (Holbig and Grove 2008), and 14:86 to 23:77 at 1.5 GPa and > 1200 °C (Pilet et al. 2010). The average of these proportions is Ol:Cpx = 15:85. If spinel is part of the crystallizing assemblage, as observed in the Deccan alkalic rocks, the phases crystallize in Ol:Cpx:Spl = 12:68:20 proportions (Elkins-Tanton and Grove 2003). With decreasing temperature, phlogopite joins the crystallization assemblage at ~ 1270 °C (Holbig and Grove 2008), and it is present in all Deccan alkalic rocks except the Murud-Janjira basanites. Since the Sarnu-Dandali rocks equilibrated above the temperatures of phlogopite stability (~ 1270 °C, Table 3), only Ol:Cpx:Spl = 12:68:20 phase proportions were added to calculate their primary magmas. The other Deccan rocks show equilibration temperatures of ~ 1120–1150 °C (Table 3). Hence, the primary magmas of these rocks (except Murud-Janjira basanites) were calculated by adding phases in proportions of Ol:Cpx:Spl = 12:68:20 above 1270 °C, and Ol:Cpx:Spl:Phl = 12:68:20:15 below 1270 °C (phlogopite proportion from Holbig and Grove, 2008). Amphibole is present in some Mundwara rocks and Murud-Janjira lamprophyres, but it was not considered in the calculations because it is stable at lower temperatures (< 1130 °C at 15 kbar in experiments, Pilet et al. 2010; 935–1030 °C at ~ 3–7 kbar in Mundwara rocks, Pande et al. 2017) than the equilibration temperatures of the Deccan rocks (Table 3). However, the Mundwara primary magma estimates should be treated with caution because the crystallizing-phase proportions are not well known at their 4–9 kbar pressures of equilibration.

The primary magmas of the Murud-Janjira basanites RG3 and RG1 were calculated by adding phases in Ol:Cpx:Spl = 12:68:20 proportions. In addition, the primary magmas were calculated using a reverse assimilation–fractional crystallization approach (AFC, Thompson et al. 2002) by assimilating xenoliths occurring in the lamprophyres (Dessai et al. 2004). The mass ratio of assimilated xenoliths to fractionated crystals in each step (r = M_aⁱ/M_cⁱ) was fixed at 0.6, similar to olivine tholeiite fractionating at 1250 °C while assimilating tonalite at an initial temperature of 800 °C (cf. Thompson et al. 2002; the felsic granulite xenoliths show equilibration at ~ 800 °C, Dessai et al. 2004). The primary magmas of basanite were also calculated by incremental magma mixing between the primitive melts and an evolved melt represented by phonotephrite RG2 that occurs in the same intrusion as the basanites. The mixing proportions were adjusted (melt:RG2 = 55:45 for RG3, and melt:RG2 = 25:75 for RG1) until the SiO₂ contents of the primary magmas of basanite and lamprophyre were similar.

In the above calculations, average equilibrium K_D(Fe²⁺–Mg) values of 0.29 for Ol-melt, 0.28 for Cpx-melt, 0.95 for Spl-melt, and 0.21 for Phl-melt were used (Elkins-Tanton and Grove 2003; Holbig and Grove 2008), and the primary magmas (Mg# 72.3) were equilibrated with Fo₉₀ olivine. Where olivine composition in the sample is known, the Fe³⁺/$\sum {}$Fe ratio of the bulk was calculated with the olivine-melt oxybarometer of Blundy et al. (2020). The lattice strain model predicts K_D(Mn–Mg) values of 0.25 for sample SD11 (olivine outer core, Fo₈₁), and 0.25 and 0.20 for samples 95-BMR-37 (olivine grain-3, Fo₈₁) and 95-BHUJ-16 (olivine grain-3, Fo₈₈, Simonetti et al. 1998) that are within the ± 0.08 uncertainties of observed K_D(Mn–Mg) values (0.32, 0.23 and 0.29), and the Fe³⁺/$\sum {}$Fe ratios of these sample are 0.16, 0.10 and 0.14, respectively. For the other samples, an average Fe³⁺/$\sum {}$Fe ratio of 0.13 was assumed. The Fe₂O₃ contents of clinopyroxene and spinel were estimated using the data in Canil and O’Neill (1996), and the Fe³⁺/$\sum {}$Fe ratio of phlogopite was assumed to be the same as that of clinopyroxene.

The uncertainties in the calculations depend on the degree of reverse fractionation required to attain the primary magma composition for each sample. For high degree of fractional crystallization, such as in sample SD11 (49%), a 10% increase in the K_D(Fe²⁺–Mg) values results in a 1.6% decrease in MgO and 3.2% increase in FeO in the primary magma (Mg# 71.3) that equilibrates with Fo_88.6 olivine. A 10% decrease in the K_D(Fe²⁺–Mg) values results in a 0.2% increase in MgO and 0.3% decrease in FeO in the primary magma Mg# (72.4) that equilibrates with Fo_90.1 olivine. Increasing the olivine proportion in the crystallizing assemblage from Ol:Cpx = 15:85 to 23:77 results in a 4.1% increase, whereas decreasing it to 11:89 results in a 2.3% decrease in both MgO and FeO. The combined uncertainties are ± 6.4% in both MgO and FeO. For Bhuj sample B (Karmalkar et al. 2005) that underwent only 2% fractional crystallization, the combined uncertainties in MgO and FeO are much lower (± 0.2%).

The calculated compositions of the Deccan alkalic primary magmas (Mg# 72.3, MgO = 12.3–16.2 wt%, Murud-Janjira basanites modeled by magma mixing) are provided in Table 4 and S4, and shown in Fig. 7. The Sarnu-Dandali primary magmas contain 38.3–39.7 wt% SiO₂, 2.0–4.3 wt% total alkali, and 14.1.16.8 wt% CaO. They mostly plot in the foidite field of the TAS diagram (Fig. 7a). The Bhuj, Mundwara and Murud-Janjira primary magmas contain higher SiO₂ (40.0–44.9 wt%), and lower CaO (10.8–14.4 wt%) and CaO/Al₂O₃, and they mostly plot in the picrobasalt and basanite (rarely in foidite) fields. The primary magmas of Murud-Janjira basanite modeled by magma mixing contain lower (1.5–2.2 wt%) TiO₂ than the other primary magmas. The Ni contents of the Deccan alkalic primary magmas are similar to the Ni contents of model peridotite-source primary magmas with ~ 12–16 wt% MgO (Herzberg et al. 2016) (Fig. 7i). Furthermore, using D_Ni from Beattie et al. (1991), the NiO contents of olivine (0.17–0.58 wt%) in equilibrium with the Deccan primary magmas overlap with the estimated NiO contents of olivine from primitive basalts (Mg# 89) worldwide, and experimentally derived melts of peridotite at 1–4.5 GPa (0.2–0.5 wt%, Matzen et al. 2017). The Cr contents of the Deccan alkalic primary magmas (200–1310 ppm) also overlap with the Cr content of primary melts of lherzolite (~ 750–2500 ppm, or 0.11–0.37 wt% Cr₂O₃, Liu and O’Neill 2004 and references therein). Thus, the estimated primary magma compositions in this study are consistent with the compositions of primary melts of lherzolite. The Sarnu-Dandali and Mundwara primitive alkalic rocks and the Murud-Janjira lamprophyres evolved by higher degrees (30–59%) of fractional crystallization than the Bhuj rocks (2–46%, average 26%) and the Murud-Janjira basanites (16–28%, Table S4).

Table 4 Composition ranges of the Deccan alkalic primary magmas, experimentally derived melts and primitive rocks

Fig. 7

Bivariate plots showing the variation of different oxides and their ratios with SiO₂ and MgO in the calculated Deccan alkalic primary magmas and experimentally generated melts and primitive rocks with Mg# 70–75 [refer to legends in (a, b)]. In (a), the boundaries for different rock-types are as in Fig. 2a. In (i), the D_Ni values used are 2.6 (Cpx) and 1.3 (Phl) from GERM (2021), 10.1 (Spl) from Wijbrans et al. (2015), and from the expression in Beattie et al. (1991) for olivine; the dash-dot line represents the trend for primary melts saturated with fertile lherzolite (Herzberg et al. 2016). In (j), the D_Cr values used are 1.24 (Ol), 1.66 (Cpx), 3.43 (Spl) and 5.4 (Phl) from GERM (2021). Maximum uncertainties in the primary magma estimates from uncertainties in K_D(Fe²⁺–Mg) and crystallizing-phase proportions are shown by error bars; for Ni and Cr, the error bars include uncertainties arising from the range of D_Ni and D_Cr values reported in the literature. The dashed lines with arrow connect the bulk composition of SD11 (open square) to its primary magma. The data for primitive rocks and experimentally generated melts are from EG03—Elkins-Tanton and Grove (2003), E80—Edgar et al. (1980), D07—Dasgupta et al. (2007), and B16—Baasner et al. (2016) (H₂O and CO₂ contents: 0.4 wt% in 4H4C, 0.7 wt% in 7H7C)

Discussion

Non-lherzolitic mantle sources

Experiments on hornblendite at ~ 1.5 GPa (Pilet et al. 2008), garnet pyroxenite at 2–2.5 GPa (Hirschmann et al. 2003), and eclogite + CO₂ at 3 GPa (Dasgupta et al. 2006) have produced melts with a wide range of Mg# (36–77, assuming Fe²⁺ = Fe^T) that show increasing Mg# and MgO and decreasing SiO₂, TiO₂ and K₂O with the degree of melting (25–100%) (Tables 1 and 4, Fig. 2). These melts and their derivatives are comparable with the Deccan primitive alkalic rocks with similar Mg#s (51–72). The Sarnu-Dandali primitive rocks have similar MgO, total alkali, K₂O and Ce/Y, and lower SiO₂ and TiO₂ contents compared to the hornblendite melts (Fig. 2). They are lower in SiO₂ and MgO, and higher in total alkali, K₂O and TiO₂ than the pyroxenite and eclogite melts. Their CaO/Al₂O₃ ratios are higher than the hornblendite and pyroxenite melts, but lower than the eclogite melts. Primitive hornblendite, pyroxenite and eclogite melts will probably evolve by clinopyroxene-dominated fractional crystallization with spinel also playing an important role. This will result in a decrease in MgO and an increase K₂O and TiO₂, whereas the trend for CaO/Al₂O₃ will depend on the clinopyroxene:spinel ratio. The CaO/Al₂O₃ ratio will decrease if clinopyroxene is the dominant crystallizing phase, but it may slightly increase if significant spinel co-crystallizes (compare with SD11, Fig. 7). Such evolutionary trends are not likely to produce melts similar to the Sarnu-Dandali primitive rocks.

The Bhuj primitive rocks have similar SiO₂, MgO, total alkali and K₂O, and lower CaO/Al₂O₃ and TiO₂ contents compared to the Cpx-hornblendite melts (Fig. 2). Fractionation of a primitive hornblendite melt will not produce melts similar to the Bhuj primitive rocks. However, the Bhuj primitive rocks are higher in CaO/Al₂O₃, K₂O and TiO₂ than the pyroxenite melts, and fractionation of a primitive pyroxenite melt may produce melts similar to some of the Bhuj rocks. The high Ce/Y ratios of the Murud-Janjira lamprophyres and the low TiO₂ contents of the Murud-Janjira basanites (Fig. 2d, f) are not consistent with hornblendite or pyroxenite melts and their derivatives.

Whether high-Mg melts of pyroxenite can be parental to the Bhuj rocks needs to be confirmed through rigorous fractional crystallization modeling. This could not be done because Fe₂O₃ contents and proportions of phases (probably pressure-dependent) crystallizing from high-Mg melts of non-lherzolitic lithologies are not available. With the exception of some Bhuj rocks, it appears that non-lherzolitic sources for the origin of the Deccan primitive alkalic rocks are unlikely.

Lherzolitic mantle sources

High-pressure (~ 3 GPa) melting experiments have been performed on anhydrous and hydrous spinel- and garnet-lherzolite (Till et al. 2012; Grove et al. 2013 and references therein), garnet lherzolite with added CO₂ (Dasgupta et al. 2007, 2013) and H₂O + CO₂ (Baasner et al. 2016), phlogopite lherzolite (Condamine et al. 2016), and K-rich lherzolite + basalt mixtures (Davis and Hirschmann 2013). Low-degree melts produced in these experiments, and high-degree (near liquidus) melts produced in experiments on primitive basalt (Takahashi et al. 1998), hydrous basanite (Pilet et al. 2010) and olivine leucitite (Edgar et al. 1980; Elkins-Tanton and Grove 2003; Holbig and Grove 2008) are possible analogues of the calculated Deccan alkalic primary magmas in equilibrium with lherzolite.

The melts of basalt (Takahashi et al. 1998) are highly quartz normative (Grove et al. 2013) and the melts of hydrous basanite (Pilet et al. 2010) are highly enriched in Na₂O (3.5–9.4 wt%) and K₂O (1.6–3.0 wt%). These melts are very different from the Deccan primary magmas. A naturally occurring primitive olivine ugandite with Ol + Cpx + Spl + Phl on its liquidus (Edgar et al. 1980) contains similar total alkali, K₂O, FeO^T and CaO/Al₂O₃, but higher SiO₂ and MgO and lower TiO₂, CaO and Al₂O₃ than most Deccan primary magmas (Fig. 7a-h). A Sierra Nevada olivine leucitite and its near-liquidus melts in equilibrium with Ol + Cpx + Spl + Phl (Elkins-Tanton and Grove 2003) have Mg#s 70.0–74.4 (assuming Fe²⁺ = Fe^T) and similar CaO and Al₂O₃, but higher SiO₂, total alkali and K₂O and lower FeO^T and TiO₂ than most Deccan primary magmas (Fig. 7a-h). Thus, the high-degree melts of primitive basalt, hydrous basanite and olivine leucitite are not analogous to the Deccan primary magmas. The Deccan picrites from Dhandhuka, Gujarat, (Mg# 69–72, Krishnamurthy and Cox 1977) have many compositional similarities with the Deccan alkalic primary magmas (Table 4 ), but their high SiO₂ contents (47.6–48.7 wt%) imply that they are not analogous.

When the Deccan primary magmas are plotted in Ol-Pl-Cpx (from quartz, Qz), Ol-Cpx-Qz (from Pl), and Ol-Pl-Qz (from Cpx) pseudoternary projections (Fig. 8), they do not plot near their spinel- or garnet-lherzolite multiple saturation points (MSPs) predicted by the composition and pressure-dependent expressions of Till et al. (2012) and Grove et al. (2013). Compared to the Deccan primary magmas, the melts of phlogopite lherzolite (Condamine et al. 2016) and garnet lherzolite + basalt with added K₂O (Davis and Hirschmann 2013) are higher in total alkali (~ 5–11 wt%), SiO₂ (44.3–47.7 wt%) and Al₂O₃ (12.9–14.6 wt%), and the moderate-degree (16–18%) melts of lherzolite + 2.5 wt% CO₂ (Mg# 75.1–75.5, Dasgupta et al. 2007) are lower in SiO₂, total alkali, K₂O, TiO₂ and Al₂O₃ and higher in MgO and CaO (Fig. 7a-h). Thus, volatile-free lherzolite, phlogopite lherzolite, lherzolite + basalt + K₂O, and lherzolite + 2.5 wt% CO₂ are not possible sources of the Deccan alkalic primary magmas.

Fig. 8

Portions of the pseudoternary projections of a Ol-Pl-Cpx from Qz, b Ol-Cpx-Qz from Pl, and c Ol-Pl-Qz from Cpx showing the compositions of the Deccan alkalic primary magmas, experimentally derived liquids for different degrees of melting (labeled for Dasgupta et al. 2007) and primitive rocks plotted according to the scheme of Tormey et al. (1987) and Grove (1993) (symbols as in Figs. 2 and 7). Also shown are the predicted spinel lherzolite (9–21 kbar) and garnet lherzolite (25–40 kbar) multiple saturation points using the expressions of Till et al. (2012) and Grove et al. (2013) for the average Deccan alkalic primary magma on a volatile-free basis. The dot-dashed lines with arrow indicate the direction of displacement of the 25 kbar garnet lherzolite MSP with addition of CO₂ in the lherzolite source

The 2–7% melts of lherzolite + 0.4 wt% H₂O + 0.4 wt% CO₂ (Mg# 72.9–78.2, assuming Fe²⁺ = Fe^T) and the 2–18% melts of lherzolite + 0.7 wt% H₂O + 0.7 wt% CO₂ (Mg# 74.9–76.7) (Baasner et al. 2016) have overlapping SiO₂, FeO^T and CaO contents and CaO/Al₂O₃ ratios with the Deccan primary magmas, but they have higher MgO and lower Al₂O₃, K₂O and TiO₂ contents (Fig. 7). The 6–21% melts of lherzolite + 1 wt% CO₂ (Mg# 72.4–75.0, Dasgupta et al. 2007) show overlapping MgO, FeO^T, Al₂O₃ and total alkali contents and CaO/Al₂O₃ ratios with the Deccan primary magmas (Fig. 7). Their SiO₂ contents are higher than the Sarnu-Dandali primary magmas, but similar to the Bhuj primary magmas. Thus, although their K₂O and TiO₂ contents are lower, they are the most similar to the Deccan primary magmas among all of the experimentally derived melts of lherzolite considered above. In the pseudoternary plots (Fig. 8), the Deccan primary magmas partially overlap with the melts from Baasner et al.’s (2016) experiments as well as the melts from experiments on lherzolite + 1 wt% CO₂ (Dasgupta et al. 2007). The melts of lherzolite + H₂O + CO₂ with higher MgO contents plot closer to the olivine apex and away from the plagioclase apex compared to the melts of lherzolite + 1 wt% CO₂. The Sarnu-Dandali primary magmas appear similar to the melts of lherzolite + 1 wt% CO₂ in the Ol–Pl–Cpx and Ol–Pl–Qz projections (Fig. 8a,c), and to the melts of lherzolite + 0.4 wt% H₂O + 0.4 wt% CO₂ in the Ol–Cpx–Qz projection (Fig. 8b). The primary magmas of the Bhuj alkalic rocks and Murud-Janjira lamprophyres are similar to the melts of lherzolite + 1 wt% CO₂ in the Ol–Pl–Qz projection (Fig. 8c). The primary magmas of the Murud-Janjira basanites are similar to the melts of lherzolite + 1 wt% CO₂ in all three projections. Thus, the Deccan alkalic primary magmas probably originated by melting of a lherzolite source containing low H₂O and CO₂ (probably < 1 wt%). A comparison of the SiO₂ and CaO contents of the Deccan primary magmas with Fig. 14 of Dasgupta et al. (2007) suggests < 0.75 wt% CO₂ in their lherzolite sources. A deep lherzolitic source is consistent with the high ³He/⁴He ratios of the Sarnu-Dandali and Mundwara alkalic rocks, and the REE patterns of Mundwara and Bhuj rocks have been modeled by low-degree melting of garnet lherzolite (Basu et al. 1993; Karmalkar et al. 2005).

Degree of melting

Assuming a lherzolitic source, the TiO₂ and K₂O contents of the primary magmas of Sarnu-Dandali and Bhuj primitive alkalic rocks and Murud-Janjira lamprophyres can be modeled by ~ 1–5% batch melting of a ~ 1.3 times Ti-enriched garnet lherzolite compared to primitive mantle (0.2 wt % TiO₂ and 0.029 wt % K₂O, McDonough and Sun 1995) (Fig. 9). The primary magmas of Murud-Janjira basanite contain lower (1.5–2.2 wt%) TiO₂, and they require a Ti-poor (0.7 times) garnet lherzolite source for their origin. Alternatively, the primary magmas of the Murud-Janjira lamprophyres and basanites can be modeled by batch melting of a ~ 1.7 times Ti-enriched and primitive spinel lherzolite, respectively (Fig. 9b), though garnet lherzolite is a more likely source for their origin.

Fig. 9

The TiO₂ and K₂O contents of Deccan alkalic primary magmas (boxes show the ranges) compared with model non-modal batch melting curves for a primitive (solid line), Ti- poor (0.7 times, long dashed line) and Ti-enriched (1.2 times: dashed line, 1.3 times: dotted lines) garnet lherzolite with tick marks indicating melting percentages, and b primitive (solid line) and 1.7 times Ti-enriched (dot-dashed line) spinel lherzolite. For garnet lherzolite melting, the source contains 55% olivine, 20% Opx, 10% Cpx and 15% garnet, the melting reaction is 1 Cpx + 0.6 garnet + 0.3 olivine = 0.9 Opx + 1 liquid (Dasgupta et al. 2007), and D_Ti and D_K are from Davies and Hirschmann (2013). For spinel lherzolite melting, the source contains 52.5% olivine, 27% Opx, 17.5% Cpx and 3% spinel (Kinzler 1997), the melting reaction is 0.82 Cpx + 0.08 spinel + 0.4 Opx = 0.3 olivine + 1 liquid (Kinzler and Grove 1992), D_Ti values are from McDade et al. (2003) (olivine-melt and pyroxene-melt) and Davies et al. (2013) (spinel-melt), and D_K values are from Davies and Hirschmann (2013)

Origin of the Murud-Janjira basanites

The isotopic compositions of Murud-Janjira melanephelinites and nephelinites indicate mixing between primitive mantle and a low ¹⁴³Nd/¹⁴⁴Nd lower crustal component in their source (Melluso et al. 2002). Furthermore, magma mixing has been proposed as a mechanism for the origin of complex, reverse zoning patterns in clinopyroxene (Dessai and Viegas 2010). Hence, the compositional effects of assimilation of the lamprophyre-hosted crustal/mantle xenoliths, and mixing with evolved phonotephrite associated with the basanites (Dessai et al. 2004; Dessai and Viegas 2010) on the primary magmas of basanite are evaluated below.

The primary magmas of basanite RG1 and RG3 calculated by simple reverse fractionation show higher SiO₂, Na₂O and CaO, and lower MgO, FeO and TiO₂ than the primary magmas of lamprophyre (Fig. 10). When the primary magmas of basanite are modeled through AFC (Thompson et al. 2002) by assimilating an average felsic granulite xenolith (Dessai et al. 2004), their SiO₂ contents become lower, whereas their Na₂O and CaO contents remain high (Fig. 10). Assimilating mafic or ultramafic xenoliths results in even higher Na₂O and K₂O contents (not shown). When the primary magmas of basanite are modeled through incremental magma mixing with phonotephrite RG2 to match the SiO₂ contents of the primary magmas of lamprophyre, their Na₂O, FeO and MgO contents become similar, whereas TiO₂ remains low and CaO remains high compared to the primary magmas of lamprophyre (Fig. 10). Thus, the AFC-based model and the magma-mixing-based model generate distinctly different compositions of the primary magmas of basanite. Because RG1, RG2 and RG3 occur in the same plug, the magma mixing model for the origin of the basanites is preferred. The differences between the primary magmas of basanite modeled through magma mixing and lamprophyre may be attributed to source heterogeneities as shown above (Fig. 9).

Fig. 10

Bivariate plots showing model primary magmas of Murud-Janjira basanites RG1 and RG3 calculated through reverse fractionation (filled grey circles), reverse AFC by assimilating felsic granulite (open grey circles), and incremental magma mixing with phonotephrite RG2 (filled red circles). Basanites, phonotephrites and tephriphonolites (Melluso et al. 2002; Dessai and Viegas 2010), average xenoliths (Dessai et al. 2004), and the primary magmas of lamprophyre are also shown. The dotted lines represent model trends for the evolution of Cape Verde lavas from a primary melt (diamond) (Weidendorfer et al. 2016). The abbreviations are: pb picrobasalt, t/b tephrite/basanite, pt phonotephrite, tp tephriphonolite, tb trachybasalt, bta basaltic trachyandesite, ba basaltic andesite, c clinopyroxenite, w websterite, m mafic granulite, and f felsic granulite with 69.8 wt% SiO₂ (plots outside the figure areas in a–d)

The above modeling results suggest that the primary magmas of basanite were probably not contaminated by an AFC-type process during ascent through the Indian continental lithosphere (cf. Dessai et al. 2004). Although isotopic and some trace element evidence suggest mixing of the mantle source with a lower crustal component (Melluso et al. 1995), it is unlikely that this component was derived from the Indian lithosphere.

Melting conditions

The Deccan primary magmas ascended along adiabatic paths below the lithosphere–asthenosphere boundary (LAB), but followed the supra-solidus ~ 5% melt contour for the appropriate lherzolite composition above the LAB (Fig. 11). The approximate P–T path can be determined from the results of experiments on lherzolite + 1.0–2.5 wt% CO₂ that is 5% melted at ~ 4 GPa/1350 °C and ~ 2 GPa/1250 °C (Dasgupta et al. 2013), and lherzolite + 0.4 wt% H₂O + 0.4 wt% CO₂ that is ~ 4% melted at ~ 2.8 GPa/1345 °C (Baasner et al. 2016). The 5% melt contour appropriate for the lherzolite source of the Deccan primary magmas (thin red line, Fig. 11) is probably located approximately between the melt contours inferred from the above two studies. Geophysical studies constrain the depths of the LAB at 85–90 km (~ 2.7 GPa) below Sarnu-Dandali (Radha Krishna et al. 2002), 63–69 km (~ 1.9 GPa) below Bhuj (Mandal and Pandey 2011), and ~ 40 km (~ 1.3 GPa) below Murud-Janjira (Negi et al. 1992). At these depths, the temperatures at the top of the melting column were approximately 1325 °C, 1290 °C, and 1280 °C beneath Sarnu-Dandali, Bhuj, and Murud-Janjira, respectively.

Fig. 11

Approximate mantle melting conditions for the Deccan alkalic rocks. The volatile-free (S17—Sarafian et al. 2017) and CO₂-saturated (PG08—Presnall and Gudfinnsson 2008) peridotite solidi are shown for reference. The melt contours in black are for ~ 5% melted lherzolite + 1.0–2.5 wt% CO₂ (D13—Dasgupta et al. 2013) and ~ 4% melted lherzolite + 0.4 wt% CO₂ + 0.4 wt% H₂O (B16—Baasner et al. 2016), and in red for the approximately ~ 5% melted lherzolite source of the Deccan alkalic rocks. The thick dashed lines represent geophysically constrained lithosphere-asthenophere boundaries (LAB) beneath Sarnu-Dandali (black), Bhuj (red) and Murud-Janjira (blue). The colored boxes with error bars indicate the P–T of crystallization for the Deccan alkalic rocks determined through thermobarometry. The dotted lines with arrow indicate the approximate P–T paths followed by the Sarnu-Dandali (black), Bhuj (red) and Murud-Janjira (blue) magmas

Conclusion

The Deccan primitive alkalic rocks mostly crystallized at high pressures beneath the Moho with clinopyroxene equilibrating at ~ 15 kbar/1270 °C at Sarnu-Dandali and at ~ 11–12 kbar/1115–1156 °C at Bhuj and Murud-Janjira. Their primary magma compositions (Mg# 72.3, MgO = 12.3–16.2 wt%) in equilibrium with Fo₉₀ olivine were calculated by incrementally adding equilibrium phases in Ol:Cpx:Spl:Phl = 12:68:20:15 proportions at low temperatures and Ol:Cpx:Spl = 12:68:20 proportions at higher temperatures (only the latter proportions were used for the Sarnu-Dandali rocks that equilibrated at high temperatures and the Murud-Janjira basanites that lack phlogopite). The Sarnu-Dandali and Mundwara primitive alkalic rocks and the Murud-Janjira lamprophyres require higher degrees (30–59%) of reverse fractional crystallization than the Bhuj rocks (2–46%) and the Murud-Janjira basanites (16–28%) to achieve equilibration with lherzolite. In a total alkali-silica diagram, the modeled Sarnu-Dandali primary magmas plot in the foidite field, and the modeled Bhuj, Mundwara and Murud-Janjira primary magmas mostly plot in the picrobasalt and basanite fields.

With the exception of some Bhuj rocks, non-lherzolitic sources are unlikely for the Deccan alkalic rocks. The modeled Deccan alkalic primary magmas show Ni and Cr contents consistent with melts of lherzolite, and their major element compositions show similarities to melts derived in ~ 3 GPa experiments on lherzolite + 1 wt% CO₂ (Dasgupta et al. 2007) and lherzolite + 0.4–0.7 wt% H₂O + 0.4–0.7 wt% CO₂ (Baasner et al. 2016). Hence, they probably originated by melting of lherzolite containing low (< 1 wt%) H₂O and CO₂. The TiO₂ and K₂O contents of the primary magmas of the Sarnu-Dandali and Bhuj alkalic rocks and Murud-Janjira lamprophyres can be explained by < 5% melting of a ~ 1.3 times Ti-enriched garnet lherzolite source compared to primitive mantle. Reverse AFC modeling by assimilating lower crustal and mantle xenoliths suggests that the primary magmas of Muruj-Janjira basanites were not contaminated by the Indian continental lithosphere during ascent. Magma mixing with evolved melts was more important in their origin. Their primary magmas contain lower (1.5–2.2 wt%) TiO₂ than the other Deccan alkalic primary magmas, and they likely originated by melting of a Ti-poor (0.7 times) garnet lherzolite source. The experimental constraints on lherzolite melting (Dasgupta et al. 2007; Baasner et al. 2016) and geophysical constraints on the depth of the lithosphere–asthenosphere boundary indicate that the approximate temperatures at the top of the melting column were ~ 1325 °C beneath Sarnu-Dandali, and ~ 1285 °C beneath Bhuj, and Murud-Janjira.