Calculation variations in different versions of alphaMELTS?

[scotthull]

I have a set of data that seems to only work with with Linux VM version of alphaMELTS v 1.4.1.  An attempt to run a simulation with the same data in a virtual machine on a separate computer with alphaMELTS v 1.5 AND yet another computer with alphaMELTS v 1.3.2 results in an initial calculation failure.  Here is my environment file followed by the input file I'm using in both versions:

Environment file:

ALPHAMELTS_VERSION         MELTS
ALPHAMELTS_MODE          isobaric
ALPHAMELTS_MAXT 3000
ALPHAMELTS_DELTAT        -2
ALPHAMELTS_MINT       1000
ALPHAMELTS_FRACTIONATE_SOLIDS true
ALPHAMELTS_CELSIUS_OUTPUT true
ALPHAMELTS_SAVE_ALL true
ALPHAMELTS_SKIP_FAILURE true
Supress: alloy-liquid




MELTS Input File:

Title: Input
Initial Composition: FeO 36.0612834498681
Initial Composition: Na2O 0.227249926193714
Initial Composition: MgO 24.3219743151707
Initial Composition: Al2O3 2.12838816487186
Initial Composition: SiO2 32.2591902818902
Initial Composition: CaO 2.03908555040175
Initial Composition: NiO 2.25905648294724
Initial Composition: TiO2 0.112981457074925
Initial Composition: Cr2O3 0.590790371581494
Initial Temperature: 2500
Final Temperature: 1000
Increment Temperature: -5
Initial Pressure: 4200
Final Pressure: 4200
dp/dt: 0
Mode: Fractionate Solids
Mode: Isobaric



Any idea what could be causing this problem?  I really need these results to match across versions.

[Paula]

Hi Scott,

Are you switching on the fO2 buffer at the command line? What to? There is a serious bug in ALPHAMELTS_ALTERNATIVE_FO2 that will be fixed in v1.6 (coming soon but I can send you a beta version if that's what you need).

Either way, try putting a tiny bit of FeO as Fe2O3 in your starting composition - the effects of having no Fe2O3 (or no FeO) initially can be unpredictable and I'm not that surprised that it varies between versions.

Paula

[scotthull]

Hi Paula,

Yes, I am switching fO2 at the command line, which is facilitated by a script that I wrote.  The calculations are running as expected in v1.4, so I think I'll stick with that version unless it would be a serious issue.

[Paula]

Hi Scott,

OK, good. Glad you have something that works for now. Version 1.6 should be ready late May / early June. If you let me know which version of the VM you are using (32-bit or 64-bit), I'll send you a beta nearer the time. That way we can verify whether the calculation works in 1.6 and debug as necessary before release.

Cheers,
Paula