[mesa-users] On the MLT++

[Bill Paxton]

On Oct 28, 2014, at 4:08 PM, Mathieu wrote:

> Hi everyone,
> 
> I would like some more detailed information regarding the MLT++, that I 
> could not find neither in the second MESA paper, nor in the documentation.
> 
> First of all: Why call it MLT*++*? As far as I understand, it is just a 
> trick to help convergence and avoid pressure inversions which in 
> principle could be physical (or something like mass loss, or photon 
> bubbles could prevent them).

My naming was not meant to be profoundly significant.
But the idea is that MLT++ is MLT plus 1, (following C language notation),
where the addition is the option to limit the superadiabaticity --
or, equivalently, to limit the inefficiency.  

> 
> Second, and more important: What happens to the energy flux when MLT++ 
> sets in?

Recall that the MLT routine gets L as an argument and returns gradT,
the expected dlnT/dlnP, as a result. MLT++ can tweak the gradT, but
again, it doesn't directly set L.  L happens as part of the overall solution.
It emerges from the newton iterations along with the profiles for T and P.

> MLT++ artificially decreases the superadiabaticity to enforce a 
> near-adiabatic stratification of the structure, so what happens in 
> convective regions where the superadiabaticity is reduced? Will MESA 
> shut down convection there (it seems to me it doesn't)? And what about 
> the convective energy flux, which is proportional to the 
> superadiabaticity to the 3/2-power? Is it changed by the MLT++, and if 
> yes, does the reduced flux imply that a larger amount of energy is 
> trapped inside the star?

I hope that this question will be resolved by the previous comments.
L is set; MLT and MLT++ decide how it will be split between L_conv + L_rad.
The reduction in superadiabaticity means that L_conv gets larger and
L_rad gets smaller, but the sum remains = L.

Of course, that's on the local level -- for a particular point at a particular location.
The global impact is a different question.  The increase in convective efficiency
from MLT++ might well lead to structural changes at a global stellar level
that might includes changes in total luminosity --- since it can be expected
to lead to changes in T gradients, it wouldn't be surprise to me if it also changed L.
But keep in mind that I'm a computer scientist with just a tiny smattering
of knowledge about stars.  So I better stick just to answering questions about the code!  ;D
Perhaps someone else will step forward to explain how things might change
in response to a less steep temperature gradient.


> From this email exchange 
> http://sourceforge.net/p/mesa/mailman/message/32229940/ it seems that 
> the energy is nevertheless carried out by the unknown mechanism 
> represented by the MLT++, but I am still confused.

The idea is just that in MLT++ we artificially boost L_conv in order to reduce L_rad while keeping the sum L_conv + L_rad = L.
But one can equally well view this as leaving L_conv small as in standard MLT and introducing L_unknown so that
L_conv_MLT + L_rad + L_unknown = L.  This allows a smaller L_rad than the L_rad_MLT = L - L_conv_MLT.
Compared to standard MLT, we've reduced L_rad by the amount L_unknown.  In that view, convection remains inefficient,
L_conv remains small as set by MLT, but L_rad doesn't have to make up for all of the inefficiency of convection.
Some unknown mechanism comes into play to carry the difference L - (L_conv_MLT + L_rad).
This unknown mechanism seems to enter only when the convection region is radiation dominated and there is a high L/M ratio.
Strong mass loss may also be a part of all of this.  And maybe all of this is a numerical side effect of doing 1D.
Perhaps it would all go away in 3D, and something would prevent the situations that lead to problems we're solving with MLT++.
We really need something better to use in our 1D codes.  MLT has been great.  But we may be running into its limits.

But see my warning above -- don't believe anything I say about anything other than what is going on in the code. ;D

Does that help, or is it all just as confusing as before?

> 
> Thanks in advance,
> 
> Mathieu
> 

Cheers,
Bill





> 
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> mesa-users mailing list
> mesa-users at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/mesa-users