Non-alkali metals titrate silica out of carbonatite melts into insoluble silicate minerals

Silica solubility in crustal carbonatite melts is a controversial topic with suggested solubility limits varying wildly from nearly 0 % up to 15 % SiO 2 . Here, we present experiments utilising alumina crucibles and CO 2 atmospheres. We tested low pressure phase relations and compositions in simplified Na–K–Ca–Fe 3+ –Al–Si–F systems. We found that substantial mutual solubility of CaO and SiO 2 was not possible. Instead, they reacted to form combeite and other Na–Ca silicates. Kalsilite formed on crucible walls, and occasionally within the liquid pools. In Fe 2 O 3 -rich systems, SiO 2 was likewise poorly soluble in the carbonatite melt with both components partitioning to separate silicate phases. Silica is substantially soluble in carbonatite melts only if they contain Na 2 O or K 2 O and no other non-alkali cations that can react with SiO 2 to form refractory silicates. In these alkali-dominated systems, immiscible silicate liquids form in equilibrium with Si-bearing, Ca-poor carbonatite melts. Our findings agree with experimentally-derived carbonatite compositions obtained over the past several decades, particularly in decreasing temperatures and increasing alkali regimes. The Ca-poor character of both experimental immiscible liquids may explain the unusual alkali-rich composition of carbonatites and nephelinites in Oldoinyo Lengai, formed at near-atmospheric pressures. Since virtually all natural carbonatite melts contain much more Ca (and other non-alkali metals such as Mg or Fe) than Si, actual natural melt compositions will be SiO 2 -free for all practical purposes at temperatures below approximately 1000 °C, resulting from SiO 2 being titrated out to refractory silicates. We propose that low-temperature silicocarbonatites are likely to be antiskarns, igneous silicate mineral assemblages within carbonatites where SiO 2 was externally supplied by adjacent silicate rocks. Likewise, we suggest that many carbonatite-associated silicate rocks in ring complexes did not entirely form as cumulates from SiO 2 -bearing carbonatite melts. Instead, some formation by antiskarn metasomatism is a plausible mechanism. We suggest that currently-observed quartz in carbonatites forms post-magmatically. Therefore, previously hypothesised silica-dependent ore-forming processes may not occur in natural carbonatites, because SiO 2 cannot attain sufficient concentrations in primary carbonatitic melts and mineralising fluids.