What a new partitioning dataset changes for modeling alkaline magma evolution

Why alkaline magmas needed their own partitioning data

Clinopyroxene and amphibole control the trace element budget of alkaline magmas, but the predictive models available for these minerals were built largely on experimental data from arc-related or anhydrous systems [1][3]. Beard et al. (2019) developed a clinopyroxene-melt partitioning model specifically for sodic alkaline magmas using clinopyroxene major element compositions, demonstrating that melt composition exerts a first-order control on partitioning [2]. Shimizu et al. (2017) parameterized lattice strain models for REE partitioning between amphibole and silicate melt, showing that amphibole composition and melt Si, Ti, and Ca contents govern D0, while E remains relatively constant [3]. These foundational and precursor studies established the theoretical framework, but experimental data for coexisting clinopyroxene and amphibole in hydrous alkaline melts remained scarce [1].

The new dataset addresses this gap directly by crystallizing both minerals together from natural alkali basalt and mugearite starting compositions at 500 MPa, 975–1100 °C, and 3.7–7 wt.% initial H2O [1]. The coexistence of clinopyroxene and amphibole in all eight experiments allowed the authors to derive partition coefficients specifically applicable to amphibole-clinopyroxene parageneses, which are common in alkaline magmas but difficult to interpret without internally consistent data [1].

What the experiments produced and how the data were validated

The experimental charges produced high-Al augite/diopside clinopyroxene, titanian pargasite to Ti-rich magnesio-hastingsite amphibole, titanomagnetite, and residual glass [1]. Clinopyroxene compositions show Ti + Al decreasing and Mg + Fe increasing with increasing Si, consistent with the Tschermak exchange reaction, while amphibole displays greater cation dispersion reflecting its more complex site occupancy [1]. Trace element systematics reveal anticorrelation between compatible Sc and incompatible Sr in clinopyroxene, and REE ratios remain largely constant despite increasing absolute concentrations from basalt to mugearite experiments [1].

To validate the partitioning results, the authors fitted measured partition coefficients to the lattice strain equation using an iterative Monte Carlo approach [1]. The fitted D0, r0, and E values for both minerals fall within ranges reported in previous studies, providing independent confirmation that the dataset is internally consistent [1][3]. Notably, r0 for trivalent cations is comparable between clinopyroxene M2 and amphibole M4 sites, but clinopyroxene shows slightly higher E3+ values, reflecting the greater rigidity of its M2 site relative to the amphibole M4 site [1]. This structural difference explains why clinopyroxene fractionates REE more efficiently than amphibole under equivalent conditions [1].

How the new data compare with earlier models and what they add

Shimizu et al. (2017) found that their melt composition model for amphibole-melt REE partitioning performed slightly better than a mineral composition model, with Pearson's Chi-square values of 41.3 and 93.7 respectively [3]. Their model reproduced 305 partitioning data from 100 experiments and generally fell between 1:2 and 2:1 correlation lines when plotted against measured values [3]. The new dataset extends this work by providing experimentally determined partition coefficients for both clinopyroxene and amphibole from the same experiments, eliminating the need to combine data from different studies with potentially inconsistent conditions [1].

The lattice strain parameters derived from the new dataset are consistent with those reported by Shimizu et al. (2017) for amphibole, but the new data also provide clinopyroxene parameters from the same experiments, enabling direct comparison of site elasticity between the two coexisting minerals [1][3]. The authors note that fitting monovalent cations in clinopyroxene was not successful due to extremely low compatibility of K, Rb, and Cs, a limitation that also affects earlier models [1]. Tetravalent cations yield high E4+ values for both minerals, reflecting the stiffness of octahedral sites associated with strong Coulombic forces [1].

Boundaries of the dataset and remaining uncertainties

The dataset is limited to two starting compositions from the Dunedin Volcano, a narrow pressure range (500 MPa), temperatures between 975 and 1100 °C, and initial water contents between 3.7 and 7 wt.% [1]. These conditions represent mid-crustal storage in an intraplate alkaline system, but extrapolation to other tectonic settings, deeper or shallower storage, or more evolved compositions requires caution [1]. The authors explicitly state that the dataset provides an empirical foundation for predictive models rather than a universal calibration [1].

Additional limitations include the unsuccessful fitting of monovalent cations in clinopyroxene and the small number of experiments for some cation groups, which required fixing r0 at constant values for tetravalent elements to reduce degrees of freedom [1]. LaTourrette and Burnett (1992) showed that U and Th partitioning between clinopyroxene and basaltic liquid is consistent with previous experimental results, but also noted that U in terrestrial magmas is not entirely tetravalent, introducing uncertainty when applying partitioning data to natural systems [5]. The new dataset does not resolve this valence-state complexity, and users should consider it when modeling U and Th in alkaline magmas [1][5].

Implications for volcanic hazard assessment and ore exploration

The new partition coefficients provide a more reliable basis for reconstructing liquid lines of descent in alkaline magmas, which directly supports volcanic hazard assessment and magmatic ore exploration [1]. The dataset includes a melt differentiation modeling worksheet that applies both batch equilibrium and Rayleigh fractional crystallization equations to the experimental data, with error estimates comparing predicted and observed residual glass compositions [1]. This allows users to test how well the new partition coefficients reproduce natural differentiation trends before applying them to specific volcanic systems [1].

The authors also provide a macro-enabled spreadsheet containing the Monte Carlo lattice strain fitting algorithm, enabling users to apply the lattice strain equation to their own clinopyroxene- and/or amphibole-melt partitioning data [1]. This tool, combined with the validated dataset, lowers the barrier for incorporating accurate partitioning parameters into models of magma differentiation and dynamics in active and quiescent alkaline plumbing systems [1]. However, the ultimate test of the dataset's utility will come from independent applications to natural samples and comparison with petrological and geochemical observations from other alkaline volcanoes [1].

About These Sources

This research page is built on 5 peer-reviewed studies — published from 1992 to 2026, 2 from 2024 or later, collectively cited 306 times — selected as the most relevant from 7 studies that passed quality screening, drawn from 75 papers retrieved from a database of over 500 million.

Sources used in this answer