Re: [mesa-users] Convergence of models in the 11 - 14 Msun range

[Bill P. <pa...@ki...>]

Thanks Matteo.  I use "convergence" for 2 things, neither of which is what Manos is talking about.

The 1st most common use of "convergence" is for the successful outcome of the iterations of the newton solver for a single timestep.  

The 2nd use is for the convergence of various properties of a series of evolution runs with varying resolution in timesteps or cell size. (see for example, section 6.7 of 1st mesa instrument paper where it discusses resolution convergence of a 1Msun model.)

-B




On Jun 26, 2014, at 2:31 PM, Matteo Cantiello wrote:

> Hi Manos,
> 
>> Hi Bill,
>> 
>> It sure is a tough problem, my last run took > 10000 steps and was 
>> still not converging
> 
> Just wanted to stress terminology here:
> What Bill means is that MESA *is* converging, but this evolutionary phase is well known to require very short time steps.
> Therefore MESA is doing its job right, even if the evolution to core collapse is expected to be computationally expensive.
> 
>> that is why I thought to share the issue to see if
>> any one had better luck.
> 
> As Bill explained if you do not care too much about the details of the physics of the inward burning flame, you can probably 
> speed up things a bit including a large overshooting below the convective flame.
> 
> Let us know how it goes!
> 
> -M
> 
>> I ll just let it burn for more and see what 
>> happens. Thanks!
>> 
>> Manos
>>> Okay.   Finally, what do you mean by "convergence problems"?   It is going to need to take small timesteps if you have an inward burning front combined with limits on how much the abundances can change per step.  The problem again is that unlike the case with convection, this one has to burn down the abundance over a very thing layer as it moves toward the center.  That just takes a lot of steps.  It is indeed a problem for those of us who are impatient!  But it isn't something I can fix for you.  As I mentioned if you aren't interested in the exact details, you might be able to speed up the inward burn by turning up overshooting below the burning shell so you get a wider burn zone.
>>> 
>>> If you find something that looks like a bug in mesa, please let me know.   But otherwise, I'm afraid you just have a tough case.  Inward burning takes a lot of steps.   Same sort of thing happens for the helium flash at tip of RBG for solar like stars.   The burn starts off center and has to work its way inward.  Also see the test case cburn_inward.  It is doing the same thing.
>>> 
>>> -B
>>> 
>>> 
>>> 
>>> 
>>> 
>>> On Jun 26, 2014, at 2:11 PM, Manos Chatzopoulos wrote:
>>> 
>>>> Hi Bill,
>>>> 
>>>> Fair enough, attached is a saved model before major convergence issues start. At this point the time-step is still fair, but it starts
>>>> to slowly drop and I start to see this off-center burning you mentioned.
>>>> 
>>>> Manos
>>>>> Hi,
>>>>> 
>>>>> Inlist is good.  But please don't force me to start from the pre ms.   Send a saved model from a bit before it starts to get convergence problems.
>>>>> 
>>>>> -B
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> On Jun 26, 2014, at 1:51 PM, Manos Chatzopoulos wrote:
>>>>> 
>>>>>> Dear Bill et al.,
>>>>>> 
>>>>>> Attached is the inlist I am using, with the latest version of MESA (6596) compiled with the latest version of MESASDK
>>>>>> for MacOSX. Running on 4 threads. Thank you for looking into this, in the meantime I will try to play around with
>>>>>> the overshooting parameter as you mentioned.
>>>>>> 
>>>>>> Manos
>>>>>>> Hi Manos,
>>>>>>> 
>>>>>>> Please send inlists and all details necessary to reproduce a sample case.
>>>>>>> I'd like to know how far the evolution has gone when convergence fails.
>>>>>>> In the past, this mass range has ignited Ne20 off-center after C/O burning.
>>>>>>> It then takes a very large number of timesteps for the burning front to advance inward to the center.
>>>>>>> The advanced is limited by not being able to deplete the Ne20 over a significant mass range at once.
>>>>>>> With convection that works to help the burning go relatively quickly.
>>>>>>> For inward burning, there is no help from mixing --- unless you crank up overshooting below the burn.
>>>>>>> Without the help of overshooting to drive things inward faster, it has to creep inward bit by bit.
>>>>>>> 
>>>>>>> -Bill
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> 
>>>>>>> On Jun 26, 2014, at 1:38 PM, Manos Chatzopoulos wrote:
>>>>>>> 
>>>>>>>> Dear all,
>>>>>>>> 
>>>>>>>> Has anyone managed to have a converged MESA model all the way from
>>>>>>>> pre-MS to pre-supernova (core-collapse)
>>>>>>>> in this rather interesting mass range of 11 - 14 Msun? Stars around this
>>>>>>>> mass range are thought to make electron-capture
>>>>>>>> SN progenitors with O/Ne/Mg cores, but when I try running them they
>>>>>>>> always seem to fail to converge past
>>>>>>>> core C/O burning. Is this relevant to the core becoming significantly
>>>>>>>> degenerate? Any insight on this would be
>>>>>>>> appreciated.
>>>>>>>> 
>>>>>>>> Best,
>>>>>>>> 
>>>>>>>> Manos
>>>>>>>> 
>>>>>>>> ------------------------------------------------------------------------------
>>>>>>>> Open source business process management suite built on Java and Eclipse
>>>>>>>> Turn processes into business applications with Bonita BPM Community Edition
>>>>>>>> Quickly connect people, data, and systems into organized workflows
>>>>>>>> Winner of BOSSIE, CODIE, OW2 and Gartner awards
>>>>>>>> http://p.sf.net/sfu/Bonitasoft
>>>>>>>> _______________________________________________
>>>>>>>> mesa-users mailing list
>>>>>>>> mes...@li...
>>>>>>>> https://lists.sourceforge.net/lists/listinfo/mesa-users
>>>>>> <inlist_input.txt>
>>>> <13sm_1.mod>
>> 
>> 
>> ------------------------------------------------------------------------------
>> Open source business process management suite built on Java and Eclipse
>> Turn processes into business applications with Bonita BPM Community Edition
>> Quickly connect people, data, and systems into organized workflows
>> Winner of BOSSIE, CODIE, OW2 and Gartner awards
>> http://p.sf.net/sfu/Bonitasoft
>> _______________________________________________
>> mesa-users mailing list
>> mes...@li...
>> https://lists.sourceforge.net/lists/listinfo/mesa-users
>