mesa-users] Hydro convergence failure at end of main sequence.
paxton at kitp.ucsb.edu
Tue Nov 2 18:08:06 EDT 2010
On Nov 2, 2010, at 6:25 AM, Sam Jones wrote:
> I am using MESA version 2664 (compiled with ifortran 11.1) on Ubuntu 10.04 to run a 20Mo model up to collapse, but at the end of the main sequence I get an error "hydro_failed_to_converge". I have tried this with two different timestep parameters and attach the inlists for both runs - maybe these values for varcontrol are inappropriate?
From the inlists you sent, I see that you've tried changing the following controls:
varcontrol_h1 = 100.
varcontrol_he4 = 100.
varcontrol_c12 = 100.
varcontrol_o16 = 100.
Let me clarify the situation with those parameters a bit. First, in recent versions, I've moved varcontrol_h1 and it's friends from public/star_defaults.dek to private/private_defaults.dek just to avoid the problem you have had! Those parameters (and several others) are leftovers from various experiments I've done, some of which worked and some of which didn't. So I'd suggest updating to the current mesa and limiting yourself to controls listed in star/public. My apologies for not cleaning things up sooner.
Second, as the comments in the file explain, the role of those parameters is to include certain abundances in the decision for the default timestep. However in my experience, it has worked better to keep the default varcontrol simple and add the abundance limits as a "special" test that applies to reduce the default timestep (that's why I've moved those controls to private; next step will be to delete them). There are a ton of special limits, and I think the ones you want are there in star_defaults: seach for the string "limits based on max decrease in mass fraction at any location in star" and check things you find. In particular, you may want to set dX_limit and dX_limit_min_X (and also dH_limit, dHe_limit, etc.). Take a look at star/test_suite/15M_si_burn/inlist_15_z2m2 for an example.
As to why those settings caused a failure to converge, I cannot guess. But of course there is NO way that mesa will be able to survive all variations on parameter settings, and failure to converge is its way of dying. Rather than spend time on this particular case, please give your 20M case another try using the controls I've mentioned above. I think you might want to adopt the strategy of starting from the inlist from the test_suite and making small changes from that to see what happens.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Mesa-users