Lithium cannot be determined by electron microprobe, but it may be an essential component in tourmalinesupergroup minerals. Therefore, its estimation is important for structural formula calculation and nomenclature. In this paper, we present a method to estimate Li content in tourmaline from microprobe data based on a multiple linear-regression model, which is not reliant on a particular normalization scheme. The results derived from this model are reasonably accurate, particularly for low-Mg tourmalines (<2 wt.% MgO) with Li2O contents higher than ∼0.3 wt.%. Furthermore, it provides a better fitness compared with estimations of Li assuming that Li fills any cation deficiency at the Y site.
