Multicomponent diffusion in basaltic to andesitic melts: Toward a temperature- and composition-independent eigenvector matrix

Multicomponent diffusion in natural silicate melts controls many igneous processes such as magma mixing and crystal growth. In N -component silicate melts, multicomponent diffusion is characterized by an N - 1 × N - 1 diffusion matrix [ D ] , whose eigenvectors define N - 1 eigen-components that diffuse independently. Each eigenvalue is the diffusion coefficient for the corresponding eigen-component. Previous studies demonstrated that diffusion eigenvectors in 8-component SiO 2 -TiO 2 -Al 2 O 3 -FeO-MgO-CaO-Na 2 O-K 2 O basaltic melts are roughly invariant with temperature. In this work, we extend this investigation to andesitic compositions to test the temperature independence of diffusion eigenvectors in more evolved melts. Thirteen diffusion couple experiments were carried out at 1250, 1350, and 1500 °C: twelve andesite-andesite couples and one basalt-andesite couple. Diffusion profiles from the twelve andesite-andesite diffusion couple experiments can be fit simultaneously with a single eigenvector matrix, supporting the hypothesis that the eigenvectors are temperature invariant in andesitic melts. The eigenvector matrix obtained from the andesitic melt is similar to that in basalt, indicating possible compositional invariance across the basaltic–andesitic range. To test compositional independence, we simultaneously fit andesitic diffusion couple profiles in this study and basaltic diffusion couple profiles from previous investigations using a single eigenvector matrix. All diffusion profiles can be well reproduced using a single eigenvector matrix, supporting the hypothesis that a common eigenvector/eigensubspace structure can characterize multicomponent diffusion in basaltic to andesitic melts over the investigated temperature and compositional range. The fitted eigenvector matrix was further tested using diffusion profiles from olivine dissolution experiments. The model reproduces the major features of the fitted profiles and can predict diffusion profiles at higher pressures that were not used in fitting. The results provide additional support for the broader applicability of the proposed multicomponent diffusion model. The extracted eigenvector matrix and eigenvalues are incorporated in an updated open-access diffusion calculator to compute multicomponent diffusion profiles in basaltic to andesitic melts.