[mesa-users] Hook in HR when hitting max_age

[Bill Paxton]

Hi Radek,

Like you, I find it disturbing that we get such a big effect from change of timestep.  Perhaps there is a bug lurking in there somewhere, but it isn't obvious.  

For now, interpolating to the max age is probably the best work around -- with checks to make sure the results are stable when you go to a smaller max timestep (convergence, convergence, convergence).  ;p

-Bill




On Jan 31, 2014, at 12:08 AM, Radek Smolec wrote:

> Hi Bill,
> 
> Thanks for your explanantion. I will stop at some larger max_age and interpolate. There is however one think I don't understand. By default, MESA instantly adjust the time-step. Usually the changes are much smaller than in the case of interpolation to hit the max_age. I infer that the problem remains but is much smaller. I guess that for a longer time, the effect cancels out rather than accumulates. Am I right?
> Cheers,
> Radek 
> 
> 
> 2014-01-30 Bill Paxton <paxton at kitp.ucsb.edu>:
> 
> On Jan 30, 2014, at 1:09 AM, Radek Smolec wrote:
> 
>> the last time step is adjusted to match the desired age. But in HR diagram it produces an artificial hook - see an example  Figure (magnified at the end of calculation) and inlist I used. The hook remains there for all the parameters I tried, sometimes it is smaller, sometimes it is very large. How can I get rid of it?
>> 
> 
> Hi Radek,
> 
> Short answer: to get rid of the hook you need to reduce the size of changes in timestep at the end of the run:
> either use much smaller timesteps or (perhaps better) keep the same timestep size and let the last step go beyond the target age; then interpolate the values for things lke L, R, Teff to the target intra-step time.
> 
> Longer answer: changes in timestep leads to changes in eps_grav because of changes in 1st order estimates for Lagrangian time derivatives for lnT, lnRho, or lnPgas.  Changes in eps_grav lead to changes in temperature, density, luminosity, and radius since everything depends on everything else.
> 
> [Someday it would be interesting to see if the eps_grav calulations could be improved by going to a 2nd order backward difference estimate for the time derivatives -- but I haven't gotten to that yet.]
> 
> Here are plots from your case.   The biggest changes happen at the base of the convection zone and actually result in a small movement of the base (too small to see in these plots).    (model 1730 is the final model; it has a slightly smaller timestep than 1729 -- log(dt/yr) of 6.433119 vs 7.000 for previous steps)
> 
> First, here's the HR hook for the final model (1730):
> 
> 
> 
> 
> 
> 
> Here are profiles for the final 2 models:
> 
> 
> 
> 
> Those changes in eps_grav are reflected in the luminosity (more negative eps_grav in 1730 gives lower L)
> 
> 
> There are similar changes in the time derivatives for lnT, lnPgas, and lnR
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> 
> If the change in dt is small relative to the previous value, then I'd expect smaller effects -- conversely, if you have to make a big reduction in the final timestep to hit the specified max age, then I'd expect that to give a large "hook".
> 
> 
> -Bill
> 
> 
> 
> 
> 
> 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20140131/b9a75669/attachment.html>