Back in July, we released a new version of the igneous-set x-eos, which I claimed would prevent the appearance of a high-Ca opx in most-stable peridotitic assemblages. Soon afterwards, Ben Klein pointed out that, in fact, high-Ca opx was alive and well in his Perple_X calculations (thanks for letting me know, Ben!).
I believe we’ve now genuinely solved this. You can now download yet another set of igneous input files.
So what was wrong? Initially, Tim and I looked for the problem in the pyroxene quadrilateral, which is the heart of the opx and cpx x-eos calibration. For our previous release of files back in July, Tim removed a gremlin from the enstatite-diopside binary system (introduced by me – yikes), in which the solvus in cpx could be metastable with respect to the equivalent solvus in opx. When this didn’t eliminate high-Ca opx from the 10-component peridotite system, I flushed the same gremlin out from the rest of the pyroxene quadrilateral. But the high-Ca opx continued to mock us.
The final(?) piece of the problem related to the introduction of Ti into opx via the end-member obuf, MgAl(MgTi)0.5SiO6. Ben Klein alerted me to this by mentioning the relatively high Ti content of his high-Ca opx. The end-member properties of obuf itself are essentially unknown, so it doesn’t appear in the dataset. Instead we make an end-member with the right composition by combining Mg-tschermak’s pyroxene, periclase, rutile and corundum, and adding a ΔG term to represent the difference between their combined G(P,T) curves and the (unknown) true G(P,T) of obuf:
Gobuf = Gmgts + 1/2 (Gper + Gru – Gcor) + ΔGobuf
Since we know so little about obuf, and given that it’s only ever present in small proportions, we would not try to fit for both ΔGobuf and the mixing properties of obuf, which are also unknown. We face this problem with many end-members in the more complex x-eos. A common solution is to assign values to the mixing properties, based on the mixing properties of similar end-members, then just fit for the ΔG value, in this case ΔGobuf. However, obuf appears in such tiny proportions that in Holland et al (2018) it was simply treated as if it mixed ideally. Apparently, this wasn’t good enough, so I have given it some more realistic mixing properties, and adjusted ΔGobuf to compensate, doing something similar with obuf‘s monoclinic counterpart cbuf in the cpx x-eos. This destabilises high-Ca opx, at least in Ben’s example.
Has stable high-Ca opx really gone? Let us know if you discover further problems!