Tracing and identifying recycled carbonates is a key issue to reconstruct the deep carbon cycle. To better understand carbonate subduction and recycling beneath the southeastern Tibetan Plateau, high‐K cal‐alkaline volcanic rocks including trachy‐basalts and trachy‐andesites from Tengchong were studied using Mg and Ca isotopes. The low δ²⁶Mg (−0.31 ± 0.03‰ to −0.38 ± 0.03‰) and δ^44/40Ca (0.67 ± 0.07‰ to 0.80 ± 0.04‰) values of these volcanic rocks compared to those of the mantle (−0.25 ± 0.07‰ and 0.94 ± 0.05‰, respectively) indicate the incorporation of isotopically light materials into the mantle source, which may be carbonate‐bearing sediments with low δ²⁶Mg and δ^44/40Ca values. In addition, no correlations of δ²⁶Mg and δ^44/40Ca with either SiO₂ contents or trace element abundance ratios (e.g., Sm/Yb and Ba/Y) were observed, suggesting that limited Mg and Ca isotopic fractionation occurred during cal‐alkaline magmatic differentiation. A binary mixing model using Mg–Ca isotopes shows that 5–8% carbonates dominated primarily by dolostone were recycled back into the mantle. Since Tengchong volcanism is still active and probably related to ongoing plate tectonic movement, we propose that the recycled carbonates are derived from oceanic crust related to the ongoing subduction of the Indian plate.

Mantle peridotites show that Ca is isotopically heterogeneous in Earth's mantle, but the mechanism for such heterogeneity remains obscure. To investigate the effect of partial melting on Ca isotopic fractionation and the mechanism for Ca isotopic heterogeneity in the mantle, we report high‐precision Ca isotopic compositions of the normal Mid‐Ocean Ridge Basalts (N‐MORB) from the southern Juan de Fuca Ridge. δ^44/40Ca of these N‐MORB samples display a small variation ranging from 0.75 ± 0.05 to 0.86 ± 0.03‰ (relative to NIST SRM 915a, a standard reference material produced by the National Institute of Standards and Technology), which are slightly lower than the estimated Upper Mantle value of 1.05 ± 0.04‰ and the Bulk Silicate Earth (BSE) value of 0.94 ± 0.05‰. This phenomenon cannot be explained by fractional crystallization, because olivine and orthopyroxene fractional crystallization has limited influence on δ^44/40Ca of N‐MORB due to their low CaO contents, while plagioclase fractional crystallization cannot lead to light Ca isotopic compositions of the residue magma. Instead, the lower δ^44/40Ca of N‐MORB samples compared to their mantle source is most likely caused by partial melting. The offset in δ^44/40Ca between N‐MORB and BSE indicates that at least 0.1–0.2‰ fractionation would occur during partial melting and light Ca isotopes are preferred to be enriched in magma melt, which is in accordance with the fact that δ^44/40Ca of melt‐depleted peridotites are higher than fertile peridotites in literature. Therefore, partial melting is an important process that can decrease δ^44/40Ca in basalts and induce Ca isotopic heterogeneity in Earth's mantle.

Full recovery from column separation and matrix effects are the two factors that need to be considered for the high-precision analysis of stable Ca isotopes, but generally they are difficult to balance. In many cases, to get a pure Ca fraction, the interference of the matrix elements is reduced at the cost of discarding a fraction of Ca overlapping with other elements (e.g. Sr and K). However, quantitative evaluation using this approach is challenging but greatly needed. Our study investigates the influence of low Ca recovery on δ^44/40Ca using thermal ionization mass spectrometry (TIMS) with the double spike technique. δ^44/40Ca of IAPSO seawater, ML3B-G and BHVO-2 in different Ca subcuts (e.g. 0–20, 20–40, 40–60, 60–80 and 80–100%), display limited variation after iterative correction by ⁴²Ca–⁴³Ca double spike with the exponential law. Notably, δ^44/40Ca of each Ca subcut with ∼20% recovery is consistent with that of the Ca cut with full recovery, within a margin of error. Our results indicate that ⁴²Ca–⁴³Ca double spike technique can simultaneously correct Ca isotopic fractionation, occurring during column separation, and TIMS determination, because both follow the exponential fractionation law well. Therefore, a “peak cut” procedure of column separation for Ca isotope measurement using the double spike technique on TIMS is proposed. Briefly, we can mix the double spike with the sample solution well before column separation, then collect the peak of the Ca cut and abandon both sides of the Ca eluate that may overlap with other elements. This procedure would eliminate matrix effects efficiently, especially for samples with low CaO contents which typically must be passed through the column twice (e.g. peridotite and dunite).

We report a modified procedure for separating Mg and Al from meteorites and terrestrial igneous rocks for high-precision analysis of Mg isotopes with multiple collector-inductively coupled plasma-mass spectrometry. The separating procedure was carried out in a single ion-exchange column filled with AG50W-X12 resin, and Mg was eluted with 1 M HNO₃ , followed by Al eluted with 4M HNO₃ . The modified procedure efficiently eliminates most matrix elements (except for Ni, Co, and Cu) with a recovery yield of Mg > 99%. Measurements of Ni-, Co-, and Cu-bearing simulation solutions revealed no detectable matrix effects. However, test runs demonstrated significant mass-dependent fractionation during the chromatographic process; consequently, a high recovery yield of Mg (>99%) is required to limit the deviation to less than 0.05‰. Furthermore, analysis of Mg standard solution with a wide range of concentrations demonstrated negligible deviation for samples with concentrations of 0.3–2.5 μg ml⁻¹Mg relative to Mg standard solution concentrations of 1 μg ml⁻¹Mg. The total procedure bank is less than 2 ng. The long-term reproducibility of instrumental measurements of Mg isotopes is ±0.05‰ for δ²⁵Mg, ±0.10‰ for δ²⁶Mg, and ±0.06‰ for δ²⁶Mg* (2 SD, n = 211), based on analyzing five pure Mg standard solutions. Analysis of three chondrites and seven igneous rock standards showed ranges of δ²⁶Mg_(DSM3) from 0.02‰ to 0.30‰ for the former and from 0.03‰ to 0.30‰ for the latter. This method is simple and actionable.