Introduction

As the main constituent of continental large igneous provinces (LIPs)1, continental flood basalts (CFB) are characterized by huge eruptive volumes within a relatively short time span. The estimated eruptive basalt volumes range from ~2 × 105  km3 for the Columbia River Basalts to > ~ 2 × 106  km3 for the Siberian Traps2,3. The time span is usually as short as <1–3 My4,5,6,7,8. These features imply a special geodynamic process in the mantle and may trigger prominent environmental effects (climate change, mass extinction, etc.) and contribute to the formation of giant metal ore deposits9.

In principle, the generation of CFB requires an abnormally high temperature, extended decompression, a certain amount of mafic rocks in the mantle source, or the addition of H2O and/or CO2 into the mantle source10,11. These four factors are not mutually exclusive and it is likely that several or all factors contribute together to generate CFB. The elevated H2O and/or CO2 content allows melting to start in the deeper mantle and enlarges the whole melting regime, consequently contributing to the enormous melt. The CO2 content is much less than the H2O content in the mantle12 and the magnitude of the lowering solidus of the upper mantle by CO2 is less than that of H2O13. Therefore, adding H2O is expected to be more important for the genesis of CFB. High temperature, decreased pressure and mafic source lithology have been extensively discussed, albeit debated, for three decades14,15,16,17,18,19, but the evidence of high H2O content is scarce.

Indeed, there were attempts to obtain H2O content of mineral-hosted melt inclusions in CFB, but the extent to which they can reflect the initial H2O content of primitive basaltic magmas (i.e. the magmas that after being extracted from their source regions have experienced little modification) was controversial. Stefano et al.20 and Cabato et al.21 measured melt inclusions hosted by olivine phenocrysts in the CFB of the Yellowstone hotspot track and the Columbia River, respectively. They found that the H2O content in melt inclusions is as high as 2.4 wt.% for the Yellowstone and 3.3 wt.% for the Columbia River. However, the melt inclusions with the highest H2O content are not hosted by the earliest-formed (i.e., with highest Fo value) olivine phenocrysts, so they may represent the H2O content of the evolved melts rather than that of the initial ones. Ten melt inclusions in olivine phenocrysts from the Siberian Traps basalts have H2O contents ranging from 0.01 wt.% to 1.6 wt.%22,23, almost falling in the range of mid-ocean ridge basalts (MORB, ~0.1–0.3 wt.%)24,25,26,27,28,29 and ocean island basalts (OIB, 0.3–1.0 wt.%)29,30,31,32,33. However, the possibility of loss of H2O due to late-stage degassing processes was not evaluated for the Siberian melt inclusions.

In addition, Michael et al.34 and Wallace et al.35 analysed basaltic glasses of the Ontong Java and Kerguelen oceanic plateaus (the oceanic counterpart of CFB), respectively. The H2O content in these glasses ranges from 0.13 to 0.49 wt.% for the Ontong Java and from 0.24 to 0.69 wt.% for the Kerguelen oceanic plateaus and is only slightly higher than that of MORB. However, the low Mg# (=100 Mg/(Mg + Fe) mol.) of 40–60 indicated that the basaltic glasses they analysed are evolved melts, again arguing against the representativeness of the initial melts. Overall, there is currently no unarguable evidence to show whether the generation of CFB is related to the high H2O content of the mantle source.

Here, we calculate the initial H2O content of the early Permian Tarim CFB in NW China (>200,000 km2 flood basalts)9,36 using the H2O content of clinopyroxene (cpx) macrocrysts crystallized from the primitive Tarim flood basalts and the H2O partition coefficient between cpx and basaltic melt. The inferred high H2O content in the initial basaltic melt provides the first firm evidence that H2O plays an important role in the generation of CFB.

Samples and previous study

Many cpx macrocrysts (1–15 mm of grain size) were hosted by one basaltic dyke that crosscuts to the Early Permian (~280 Ma) Xiaohaizi wehrlite intrusion in the Tarim large igneous province, NW China (Fig. 1a,b). They are fresh and usually prismatic and sub- to euhedral shapes (Fig. 1c) and they commonly have a high-Mg (Mg# = 80–89) core and a thin low-Mg rim (Mg# down to 70) that is resulted from the interaction with the host basalt37. Wei et al.37 carried out a detailed geochemical analysis on these macrocrysts. These cpx generally have low TiO2 (0.26–1.09 wt.%), Al2O3 (1.15–3.10 wt.%) and Na2O (0.16–0.37 wt.%) compared to the cpx in mantle peridotites (0.31–2.50 wt.% TiO2, 1.32–12.55 wt.% Al2O3 and 0.2–1.90 wt.% Na2O), so they are not likely to be xenocrysts from mantle peridotites. The macrocrysts have strong resorption textures and are not in chemical equilibrium with the host basaltic dyke, arguing against a phenocryst genesis. In addition, the cpx macrocrysts define a coherent compositional trend (e.g., negative correlations between Mg# and Ti, Al, Na, La, Nd, Yb)37 with the cpx from the wehrlites crosscut by the basaltic dyke hosting the cpx macrocrysts and these cpx have identical trace element distribution patterns, demonstrating a comagmatic origin. Accordingly, these macrocrysts have been ascribed to be antecrysts that crystallized from the earlier and more primitive melts and have been reincorporated into the host basalt dyke before intrusion. High-Mg values indicate that the cpx macrocrysts were formed from a nearly primary basaltic melt. Although an assimilation and fractional crystallization process may operate during the formation of the Xiaohaizi intrusion that was evidenced by higher 87Sr/86Sri (0.7038–0.7041) and lower εNdi (1.0–1.9), the preservation of the high-Mg feature and depleted Sr-Nd isotope compositions (87Sr/86Sri = 0.7035–0.7037 and εNdi = 4.5–4.8) suggests that the cores of these cpx macrocrysts may have recorded the composition of the primitive Tarim basaltic melts, with little crustal contamination37. The Cpx macrocrysts in this paper were from the same dyke studied by Wei et al.37.

Figure 1
figure 1

(a) Location of the Tarim LIP and Xiaohaizi intrusion. This figure is created by the software CorelDRAW(R) Graphics Suite 12 (http://www.corel.com/cn/). (b) Clinopyroxene macrocrysts in the Xiaohaizi dyke. (c) Backscattered electron (BSE) image of a typical clinopyroxene macrocryst (XU06-09) which has a gray core and a thin and bright rim. Red squares are five spots for FTIR and EPMA analyses. (d) IR spectra of five spots in c showing similarities both in OH band positions and heights.

Results

The chemical composition and H2O content in 10 cpx grains were obtained by an electron probe micro-analyzer (EPMA) and a Fourier transform infrared spectrometer (FTIR), respectively (see Methods). Wei et al.37 have shown that the rims of the Tarim cpx macrocrysts may have reacted with the host basalt, so only the clean core area of each cpx grain was measured here, in order to retrieve the information about the initial and primitive basaltic melts. 4–6 clean analysed spots in the core area were selected to run EPMA and FTIR for each grain (Fig. 1c) and in individual grains they show same chemical compositions and IR spectra (Fig. 1d). The average values of the analysed spots of each grain were, therefore, used to represent the element and H2O contents of that grain. Ten cpx grains have TiO2 (0.39–0.66 wt.%), Al2O3 (1.23–1.77 wt.%) and Na2O (0.15–0.26 wt.%) (Table 1), which is within the range reported by Wei37. The cpx Mg# values are 85.2 to 87.8 (Table 1), corresponding to a Mg# of ~70 for the equilibrated basaltic melts using the experimental Mg-Fe partition coefficient (0.34 ± 0.04)38. This suggests that the analysed cpx grains were crystallized from a nearly primary basaltic source39, in agreement with the trace element and Sr-Nd isotope characteristics of the Tarim cpx macrocrysts37.

Table 1 Chemical composition and H2O content of the Tarim clinopyroxenes and H2O content of the corresponding basaltic melts.

The IR absorption spectra of the Tarim cpx can be subdivided into four groups, namely: 3630–3620 cm−1, 3540–3520 cm−1, 3470–3450 cm−1 and 3360–3350 cm−1 (Fig. 1d). The band at 3630 ~ 3620 cm−1 and 3540–3520 cm−1 is always strongest and weakest, respectively and the band at 3360–3350 cm−1 occurs in few grains, consistent with the structural OH bands in the cpx phenocrysts in Mesozoic-Cenozoic basalts of eastern China40,41,42. The calculated H2O contents of 10 cpx grains are 300–550 wt. ppm (Average: 384 ± 83 wt. ppm) and the calculated H2O contents of the equilibrated basaltic melts are 3.69 wt.% to 6.61 wt.% (Average: 4.82 ± 1.00 wt.%) (Table 1). Within the 40% uncertainty (see Methods), the H2O contents of the equilibrated melts do not show significant variations when the Mg# of the cpx varies from 85.2 to 87.8, suggesting that the H2O content in the magma system remained almost constant at the early stage of magma evolution. Therefore, it is reasonable to use the calculated H2O content of the melts equilibrated with the analysed cpx to represent the H2O content of the initial and primitive Tarim basaltic melt. Although bearing an uncertainty of up to 40%, such an H2O content is apparently higher than those of MORB, OIB and back-arc basin basalts (BABB, 0.2–2.0 wt.%)43,44,45 and falls in the range of island arc basalts (IAB, 2.0–8.0 wt.%)46,47,48 (Fig. 2).

Figure 2
figure 2

Comparison of the initial H2O contents of the Tarim primitive basalts with those of MORB, OIB, BABB and IAB.

The range of MORB, OIB, BABB and IAB is from references 24, 25, 26, 27, 28, 29, 29, 30, 31, 32, 33, 43, 44, 45 and 46, 47, 48, respectively.

Discussion

The arc-like H2O contents in the early Permian Tarim primary basaltic melts indicate an addition of water from subduction-related processes. In the mid-Proterozoic, the Tarim was surrounded by subduction zones49. In addition, ophiolite mélanges and arc-like magmatic events along the northern margin of the Tarim were dated at 600–418 Ma and 422–363 Ma, respectively, suggesting an active convergent margin50. These subduction processes may have provided water to the source of the Tarim Early-Permian basalts. However, the Tarim basalts do not display arc-like geochemical signatures (i.e., LILEs-enrichment and HFSEs-depletion)36,51,52,53,54. This suggests that the extra water in the source of the Tarim basalts was not from the released fluids from the subducting plates, but was instead from the dehydrated plates stagnated in the deep earth. Experimental and natural investigations have demonstrated that minerals (cpx, garnet, olivine, etc.) in dehydrated plates can carry at least several thousands ppm (wt.) of H2O into the Earth’s mantle55,56. Garnets and omphacites from ultra-high pressure matamorphic eclogites have also been shown containing ~2000–3000 ppm wt. H2O57,58. If we consider that (1) the partition coefficient of H2O between the mantle rock (peridotite, eclogite, pyroxenite) and melt is ~0.0159 and (2) the degree of partial melting of the Tarim basalts is <10%51,53, then <5000 ppm wt. H2O in the source can produce 5% H2O in basaltic melts, regardless of the melting model (batch or fractional) involved.

The upper mantle can accommodate several hundred ppm (wt.) of H2O56,60 and the lower mantle contains much less61. Only the mantle transition zone (MTZ) can contain up to >1 wt.% H2O55,62. Several thousands ppm (wt.) of H2O in the source of the Tarim basalts is, therefore, likely from the MTZ where the subducted plates stagnated and provided water63. If so, the classic core-mantle boundary-derived plume model15,64 cannot be applied to the Tarim large igneous province.

In conclusion, the high water content in the primary early Permian Tarim basalts provides clear evidence that water, in addition to the temperature, pressure and source lithology, plays an important role in the generation of continental flood basalts. Furthermore, when high water content is considered, abnormally high temperature and extended decompression that are two critical factors in the widely accepted mantle plume model65 are not always to be prerequisites in the generation of CFB (and LIPs).

Methods

The H2O content of cpx was determined with a Nicolet iso50 FTIR coupled with a Continuμm microscope in School of Earth Sciences, Zhejiang University, following the unpolarized method described in Xia et al.40. For each cpx grain, several analysed spots (~50 μm × 50 μm) were set in the clean core area and they display almost same spectra, therefore the average spectrum was used to calculate the H2O content of that grain. The modified Beer-Lambert law [c = A/(I × t)] was used to calculate to H2O content, in which c is the content of water (H2O ppm wt.), A is the total integral absorption of OH bands (cm−2) that is 3 times of the integral area of unpolarized absorption66, I is the integral specific absorption coefficient (7.09 ppm−1 cm−2)67, t is the thickness (cm). The uncertainty of H2O content is less than 30%40.

The major element contents of cpx were analysed using a Shimadzu EPMA 1600 at University of Science and Technology of China. The 15 kV accelerating voltage, 20 nA beam current and 1 μm beam diameter were used. Standards are natural minerals and synthetic oxides. Data correction was obtained by a program based on the ZAF procedure. The reproducibility is <1% for elements with concentration >5% and <3% for elements with concentration > 1%. The analysed points were set within the FTIR analysed area. Several points in each cpx grain have homogeneous element contents and the average values were used (Table 1).

The H2O content of the basaltic melts equilibrated with cpx is estimated by the H2O content of cpx and the H2O partition coefficients (Dcpx/melt) between cpx and melt. Dcpx/melt can be calculated by the equation 10 in O’Leary et al.68: D = exp(−4.2 + 6.5*X(ivAl)-X(Ca)), where X(ivAl) and X(Ca) are the concentration of octahedrally coordinated Al3+ in tetrahedral site and Ca2+ in cpx calculated on the basis of 6 oxygen. This equation was derived by compiling experimental results run at temperatures between 1025 °C and 1440 °C, pressures between 0.5–5.0 GPa, melt H2O contents between 1.09 wt.% and 24.9 wt.% and cpx ivAl between 0.002 and 0.306. Considering the uncertainties from Dcpx/melt (~10%)68 and H2O content in cpx (<30%), the total uncertainty of H2O contents in melts is estimated to be less than 40%40,41,42.

Additional Information

How to cite this article: Xia, Q.-K. et al. High water content in primitive continental flood basalts. Sci. Rep. 6, 25416; doi: 10.1038/srep25416 (2016).