mesa
mesa copied to clipboard
MESAHub
Can’t generate starting model for gyre_in_mesa_wd
The comments in the inlist say this was generated by running make_co_wd with an initial mass of 4.1. I just tried this on 9e988c4 and hit a max_model_number termination in inlist_remove_env_header. I haven’t checked to see if more steps resolves this.
At the moment this is okay: the initial model is not in particular need of revamping. But we should make sure we can regenerate it at some point.
I've started digging into this, and it looks like this issue has been around for quite a while. It also appears to affect a pretty wide range of initial masses. I tried initial mass 6.1 that should produce around a 0.95 solar mass C/O WD, and I ran into a similar problem.
I was not able to find an easy fix. I'll look into it further at some point.
Bisecting along main points to 1550ee1 as the change that broke things for making massive C/O WDs. It still works for making standard 0.6 solar mass WDs, but anything around 0.8 or above stalls out right after the envelope is stripped before the model can contract to become a WD. So it's entirely possible that massive C/O WDs haven't been possible with make_co_wd since then. Any idea how to go about debugging that?
Uhhhhh a lot has changed in the code handling MLT since then. Not sure how useful bisecting is here.
Is it using TDC? If so, maybe try turn it off during the contraction and see if that helps? I could believe that the stripping leaves the model in a bad state that breaks TDC....
Otherwise, do you see what kind of failure it hits?
Yeah, I hoped bisecting might point to something more useful like an EOS change, but no luck. At least we've now established that this has been a problem for several months.
I did try flipping MLT back to Cox, and that didn't seem to help at all.
It's not obvious what's causing the failure. The model is going to timestep hell with lots of retries, so I guess I'll have to find time to try hydro debugging to point me toward the problematic zones/equations.
🙁 Yeah. If I remember correctly the stripping routine is followed immediately by a relax_entropy phase. The relax stuff is all a bit of a black box to me, but I wonder if it's better to just directly evolve it and let it cool rather than trying to artificially speed it up...
