Sintering of chromite-silica sand

[MCOVIAN]

Hello,

I'd like to study the sintering behaviour of a mixture of chromite and silica sand (mostly quartz) in rhyolite-MELTS v1.0.2 (since I am not considering water or any volatile phases). These sands are used as filler sands for steel ladle nozzles and their degree of sintering is key for a good performance (free open ratio).

I am able to find the Liquidus temperature at atmospheric pressure and run the simulations (from 800 C to past Liquidus to assess 100 wt% liquid phase) if I input the silica sands alone (which contain +98 % SiO2, some Al2O3 and trace amounts of Na and K feldspars). My problem arises when I input any amount of chromite in the composition (40-50% Cr2O3, 25-30% Fe2O3, 6-10% MgO and 10-15% Al2O3).

I am using no constraints (both fO2 and in the Options tab).

If I run the system with a say 75:25 ratio of chromite:silica sand MELTS crashes with the error

CAUTION: Liquid component Mg2SiO4 has negative molar fraction.
Aborted (core dumped)

The problem persists with different ratios of chromite:silica. Is there something than can be done? I understand this is not a natural rock composition but according to the literature the sintering of these sand mixtures proceed via liquid phase generation (induced by the alkali feldspars) with SiO2 content increase with temperature. The spinel phase should not see any significant mass transfer to the liquid phase (which I'm assuming is the problem here).

Thank you very much!

[Paula]

Hi,

Sorry for the delay in replying. I'm assuming that you are using the graphical user interface (GUI) for rhyolite-MELTS(?) The Cr-bearing liquid end member in rhyolite-MELTS is MgCr2O4, so if your liquid has more Cr2O3 than MgO by moles, then you will run into problems when it tries to calculate the configurational entropy.

The issue is that the graphical user interface uses a superliquidus initial guess for the equilibration routine. So even though, as you say, the chromite component should mostly be in the solid spinel, rhyolite-MELTS tries to 'melt' it initially in order to have a omnicomponent phase against which to calculate affinities. Phases that are saturated are then added back into the assemblage, and compositions and masses adjusted to achieve equilibrium.

alphaMELTS has the option to choose a subsolidus (norm-calculated) initial guess. It is geared towards peridotite or basalt compositions and probably won't work for this case, though you could try it. Between the GUI and alphaMELTS there are various options for assimilation, but none that does quite what you need. As currently implemented, the system always goes through a completely molten or completely solid (metastable) state, en route to equilibrium.

The subsolidus initial guess in alphaMELTS 2 is a bit more robust than it is in alphaMELTS 1.X, and could be made more so to deal with a pyroxene-free system. I was also going to work on the assimilation routines in alphaMELTS 2 sometime soon too, and one of the options I was thinking of implementing might work here. The two things are kind of related. If you send me .melts files for the sands and chromite I'll use them as a test.

One other issue is that adjustments to the solid phases, made in rhyolite-MELTS so that the phase relations of hydrous silicic systems are more closely modeled, also affect the melting temperatures of pure sanidine and quartz. Was the liquidus temperature you got for the silica sands alone roughly correct? alphaMELTS 2 contains the rhyolite-MELTS models but it might be that we need backport it to MELTS for this particular application. (I can just build you a separate executable, if needed.)

Best,
Paula