Isobaric Crystallization - Negative Mole Fraction

[Kimberlite]

Hello everyone.  I am attempting to use pMELTS to crystallize a "kimberlite-like" melt from 1400C to 900C under isobaric conditions at 3GPa to determine what the product phases would be.  Whenever I execute the calculated melt file at these P/T conditions, I get the following error, which prevents pMELTS from executing:

CAUTION: Liquid component Si4O8 has negative mole fraction.
CAUTION: Liquid component Al4O6 has negative mole fraction.

The .melts file I am using is as follows:

Title: Kimberlitic Fluid
Initial Composition: SiO2 24.0000
Initial Composition: TiO2 1.7000
Initial Composition: Al2O3 1.6000
Initial Composition: FeO 8.2000
Initial Composition: MgO 24.0000
Initial Composition: CaO 13.5000
Initial Composition: Na2O 2.4000
Initial Composition: K2O 1.5000
Initial Composition: H2O 7.1000
Initial Composition: CO2 16.0000
Initial Temperature: 1400.00
Final Temperature: 900.00
Initial Pressure: 30000.00
Final Pressure: 30000.00
Increment Temperature: 10.00
Increment Pressure: 0.00
dp/dt: 0.00
log fo2 Path: None

Would anyone happen to know what would be causing this issue and how I can go about resolving it?  Thanks!

[Paula]

Hi,

pMELTS isn't calibrated for CO₂-bearing systems yet (see Ghiorso et al. 2002 and Antoshechkina et al. 2018). Also, as described in the alphaMELTS documentation, it is generally it is easier to treat K as a trace element for pMELTS because there are few solid phases that K can go in to at pressures where feldspar is not stable. That advice doesn't really work for kimberlite-type compositions where phlogopite crystallization may be important though.

The problem is that the composition doesn't lie within the MELTS liquid composition space. The liquid end members for K and Na are KAlSiO4 and Na2SiO3 respectively. So that's why you are ending up with negative mole fractions of some end members (and, unlike solid phases, it is not OK to have negative mole fractions of liquid end members; compare this post for the solids case).

You might not be able to model this exact composition but if you increase the Al₂O₃ and/or SiO₂ content a little you might be able to model something close - hopefully close enough to get the approximate assemblage.

Best wishes,
Paula

[Paula]

One other thing: you should put some of the Fe as Fe2O3, or turn on the fO2 buffer (or are you doing that after loading the file?)

Cheers,
Paula