[mesa-users] MLT++ ?= "black box"

Tue Apr 15 17:04:02 EDT 2014

Hi Tanja,
to visualize what Bill just nicely explained
these slides might be useful.

Cheers,
-M

On Apr 15, 2014, at 12:49 PM, Bill Paxton wrote:

> 
> On Apr 14, 2014, at 3:17 PM, Tanja Rindler-Daller wrote:
> 
>> I can see from your Paxton et al 2013 (MESA II) paper
>> 
>> http://adsabs.harvard.edu/abs/2013ApJS..208....4P
>> 
>> that including MLT++ ,or not, DOES make a difference, for sure....
>> 
>> The choice of the parameters is a 'little' black box for me. I can see the explanation in Sec.7.2 of your paper, but I do not understand the nature of
>> lambda1, lambda2 versus lambda_max and
>> beta1, beta2 versus beta_min.
> 
> Hi Tanja,
> 
> MLT++ with its many parameters isn't theory!  It is a placeholder waiting for a theory.  The parameters are just chosen based on what seems to be necessary for mesa/star to work for some cases that otherwise get bogged down in intolerably small timesteps.   It is possible that this is a numerical problem in 1D that we stumble on in mesa by taking small timesteps (for example, I have heard that some codes get through the evolution stages that need MLT++ in mesa simply by taking large timesteps that persumably smooth over the issue).    
> 
> Alternatively the need for MLT++ might be a clue that limiting energy transport to radiative diffusion and standard MLT is wrong in some extreme situations during stellar evolution -- something else is going in the star on that we've left out of our code.  MLT++ basically is a way to model the effects of such an alternative form of energy transport --- limiting superadiabaticity is equivalent to saying that some other mechanism carries energy in addition to radiative diffusion and convection so we don't need such a steep temperature gradient to get the luminosity out.
> 
> Of course you need to remember that these are just my thoughts on MLT++ and I'm a computer scientist, not an astrophysicist!   I wouldn't be surprised if some of the co-authors of the mesa paper disagree with some or all of what I've said above. ;D
> 
> 
> Okay --- with that preamble, let's dig out some more details about all those MLT++ parameters.   The routine mesa/private/mlt_info/set_gradT_excess_alpha is the place we'll look if you'd like to see for yourself.    Before doing anything else, it checks to see if MLT++ should be in effect according to these controls: okay_to_reduce_gradT_excess, gradT_excess_max_center_h1, and gradT_excess_min_center_he4.  If those tests pass, then it sets beta and lambda as follows where k is the cell index from 1 at surface to s% nz at the center.  (note that the k for beta need not be the same as the k for lambda -- i.e., beta_min and lambda_max can come from different cells).
> 
> 	beta_min = min over k of Pgas(k)/P(k)
> 	lambda_max = max over k of Lrad(k)/Ledd(k)
> 
> These are called beta_{min} and lambda_{max} in the 2nd instrument paper (eqn 38).
> 
> Now get out a piece of paper and draw axes with the y axis for lambda = 0 at bottom to 1 at top and the x axis for beta = 0 at left to 1 at right.  We'll set alpha according to where in this place our model's (beta,lambda) falls on this graph.   Recall that alpha = 0 means no boost in efficiency (i.e., no decrease in superadiabaticity) and alpha = 1 means max allowed boost in efficiency (i.e., max allowed decrease in superad).     So in your graph, the upper left where beta is near 0 and lambda is near 1 will give alpha = 1, and the lower right where beta is near 1 and lambda is near 0 will give alpha = 0.   The transition regions are determined by the points (beta1,lambda1) and (beta2,lambda2) where beta1 >= beta2 and lambda1 >= lambda2.    Draw a line connecting the points (0,lambda2), (beta2,lambda2), (beta1,lambda1), and (beta1,1).  The region to the upper left of the line gets alpha = 1 and the region to the lower right gets alpha = 0.  Instead of having an abrupt transition from 0 to 1 across the line, the parameters gradT_excess_dlambda and gradT_excess_dbeta are used to define the width in lambda and beta of a transition zone where the values gradually change.
> 
> I hope that gives you a picture of how lambda1, beta1, lambda2, beta2, dlambda, dbeta, beta_min, and lambda_max work to set gradT_excess_alpha.  If you want to keep going to see how alpha is used to adjust the gradT for each location in the model, take a look at the routine adjust_gradT_excess in the same file, star/private/mlt_info.   This is where the parameters gradT_excess_f1, gradT_excess_f2, and gradT_excess_age_fraction come in.  Let me know if you'd like more details on that.
> 
> Good luck!
> -Bill
> 
> 
>  
> ------------------------------------------------------------------------------
> Learn Graph Databases - Download FREE O'Reilly Book
> "Graph Databases" is the definitive new guide to graph databases and their
> applications. Written by three acclaimed leaders in the field,
> this first edition is now available. Download your free book today!
> http://p.sf.net/sfu/NeoTech_______________________________________________
> mesa-users mailing list
> mesa-users at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/mesa-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20140415/64e899a5/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: MLT++.pdf
Type: application/pdf
Size: 131947 bytes
Desc: not available
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20140415/64e899a5/attachment.pdf>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20140415/64e899a5/attachment-0001.html>