There is a spectrum of approaches to modelling the thermodynamic properties of silicate melts. At one extreme, components are chosen to mimic plausible mineral end-members, like Mg2SiO4, NaAlSi3O8 etcetera. This approach has worked well in reproducing cotectic phase relations, but has seen little utility in other applications, notably the effects of melt composition on trace-element partitioning, redox ratios (like Fe3+/Fe2+), and siderophile/chalcophile element and volatile-component solubilities, where simple oxide components are more convenient. Problematic has been the rich variety of unexamined assumptions, even to selecting the stoichiometry of the oxide components, with a choice between single cation (NaO1/2, MgO, AlO3/2, etc.), single-anion (Na1/2O, MgO, Al2/3O, etc.) and the DOME basis (Na2O, MgO, Al2O3 etc.), while in the metallurgical literature, components containing fictive vacancies are widely used. Parameterizations with melt structural descriptors like NBO/T (non-bridging-oxygens/tetrahedral-cations) persist, despite no experimental evidence for their effectiveness.

Here I examine published and new experimental measurements to see if there are empirical rules that may aid in predicting oxide component activity-composition relations. Valuable are measurements from metal solubility experiments under controlled fO2, which can cover a huge range of melt compositions. Components of cations with the same charge and similar ionic radius respond to melt composition similarly, e.g., the M2+O group of divalent cations (M = Mg, Fe, Ni, Co, Mn or Cr), despite the differences in crystal-field stereochemistry of the 3d2 cations. This consequence is little effect of melt composition on crystal/melt two-element exchange coefficients, most famously Mg-Fe2+ between olivine and melt. Exchange coefficients show that Zn belongs to this group. Similarities between geochemical twins (Cr3+/Ir3+, Zr4+/Hf4+, Mo6+/W6+) are unsurprising but that all these higher-charge oxide components can be parameterized in CaO-MgO-Al2O3-SiO2 melts by the “optical basicity” concept is. The Zr-Hf data exemplify that Henry's law persists to ~10 wt%. There are no counterexamples.