Ni partitioning in cpx (oxide vs trace element)

[BDreyer]

Hi All,
I'm fractionating mafic basalts and noticed that NiO increases as MgO decreases during cpx and plag fractional crystallization. Only when olivine appears does NiO decrease. It seems true then, that Ni isn't partitioned into cpx, and perhaps more realistic results would be obtained treating Ni as a trace element in alphaMELTS? I am aware that there are known issues with the partitioning of Cr2O3- what about CoO? Thanks for any insight.
-Brian

[asimow]

Hi Brian,

You are correct. As a major oxide component, NiO can ONLY go into olivine and liquid (though if you are reduced enough, Ni can go into Fe-Ni metal alloy, too). MELTS does not know how to put it in cpx. The same for CoO.

Cr2O3 is a similar but not identical situation -- it only goes into spinel and liquid, no other phases.

So, yes, you can treat Ni as a trace element. But the results will not be that accurate, because Ni partition coefficients are not very constant. Currently, D_Ni^(ol/liquid), D_Ni^(cpx/liquid) and D_Ni^(opx/liquid) ALL default to 9.4. Ni default to perfectly incompatible in every other phase. Currently, you can change these to new constant values (and make them different from one another), but you cannot implement anything better ... Ni is not one of the elements that can be treated by lattice strain theory. In the near future, we will be implementing extensions to alphaMELTS that will let you define general functions for such partition coefficients, but not yet. So, your best bet is to do your calculations of fractional crystallization without considering Ni, and then post-process a trace element treatment in outside software such as Excel or MATLAB, where you can implement something like the Beattie-Jones parameterization for D_Ni as a function of D_Mg.