problem for Morb.melts at 10 kbar

[roman@ipgp.fr]

Hi have installed the 64-bit RedHat Linux 6.6 rhyolite-MELTS-1.0.2/pMELTS on my computer.
While the exemple morb.melts works fine at 500 bar pressure, at 10 kbar I am facing some issues.
The liquidus is calculated at 1348.05 degrees. When I perform the calculation between the liquidus and 800 degrees, at 800 degree I still have 1.79 g liquid.
If I set the final temperature at 600 degrees, melts starts getting slow very below 800 degrees, and at about 630 degrees, with still 0.56g liquid, it fails to converge.
What I would be interested in to know is
1) if  it's normal that at high pressure the solidus gets so depressed.
2) If it's possible to reach a stable solid assemblage at the solidus temperature at high pressures (typically 10kbar, say the base of the crust).

By the way, when descending below 800 I get this message on the terminal in which melts is running

Warning:     
Name: Scale
    Class: XmScale
    The specified scale value is greater than the maximum scale value.



Thank you very much!
Alberto

[asimow]

Dear Alberto,

You are not doing anything wrong, as such. MELTS is doing as well as it can with this calculation.

The Scale warnings are irrelevant; they do not indicate any problem.

Let's look at what you have: the Morb.melts composition includes a lot of minor components such as P2O5, K2O, and H2O that are very difficult to put into high-pressure minerals. This is the reason for the extreme solidus depression - if there is no other place to put these minerals then they will stabilize the liquid down to extremely low temperatures. In reality there must be trace or hydrous phases that help to take up these components, but MELTS doesn't know about them. So liquid has to persist until biotite, leucite, and apatite all appear. Hence, also, your last liquid is an insane composition (<19 wt.% SiO2, >14 wt.% P2O5, >28 wt.% H2O...). Clearly, there are no experiments at high pressure that actually constrain the saturation surfaces of these phases in such a liquid, and so you are taking MELTS into the land of extrapolation, and we shouldn't expect the temperatures to be very meaningful. If it had better amphibole, mica, and apatite models then presumably they would form at higher temperature. Also, by the time this calculation stops, you have liquid+garnet+(normal)cpx+feldspar+(strange Ti,Al-rich)cpx+quartz+corundum+biotite+leucite, and the reason it is getting slow is that such a complicated assemblage is hard to equilibrate, numerically. Also, you are at conditions where the model was never meant to run, and some of the numerical operations are hard under these conditions.

If you do want a stable solid assemblage at 10 kbar, you have two choices:
(1) remove the troublesome components. If I just delete the P2O5 then I can get this composition to freeze, because you can reach biotite-in before it crashes. But the solidus is still down around 650°C and the last liquid has 35% H2O. That is what it takes to stabilize phlogopite and give a solid host for the H2O. If you also delete the water, of course, you'll have no trouble getting it to freeze above 1000°C - if you don't care about the water, that's your best bet. If you do, you're going to have trouble stabilizing a nominally hydrous solid host for the water. Using alphaMELTS, you can treat trace water in nominally anhydrous phases, but that also is probably not the right approach for a basaltic composition. alphaMELTS also has a chlorite model that might solve this problem, though.
OR, (2) construct your subsoil's assemblage by taking the major solids from a calculation like yours and putting them together to reconstruct a well-behaved solid. That is, delete the liquid which is hosting all the excess H2O, K2O, P2O5, etc. What is left will crystallize nicely.