[mesa-users] Radiative Turbulent Diffusion

[Bill Paxton]

Okay -- one more time.  7627 has the following in place of the cycle and exit lines:

            if (D > s% D_mix(k)) s% mixing_type(k) = anonymous_mixing
            D = D + s% D_mix(k)

Cheers,
Bill




On Jun 4, 2015, at 12:40 PM, Ehsan Moravveji wrote:

> Dear Umberto and Bill,
> Thanks for your clarifications and kind advices. I shall immediately read the paper …
> 
> Indeed, the list of proposed mixing mechanisms in radiative layers is non-exhaustive, and it is gradually growing over time. Even more serious than that is the interaction between different mixing mechanisms; e.g. what D_mix to adopt in a layer where semi-convection efficiency is roughly the same order of magnitude as radiative diffusion? This is indeed non-trivial, and the user can implement his subjective idea using extra hooks based on his physical insights.
> 
> Having said that, let’s take another look at lines 72 and 73 of the new turbulent_diffusion.f90
>             if (D <= s% D_mix(k)) cycle
>             if (D < s% D_mix_ov_limit) exit
> The two “IF” clauses here are imposing two critical decisions. 
> 
> - Firstly, the associated radiative turbulent diffusion coefficient, D here, can be better added to s% D_mix(k), rather than just being compared and perhaps ignored (again my example above). If so, then D will serve as a background mixing coefficient and it can be easily overcome by more stronger mixing coefficients.
> 
> - Secondly, the s% D_mix_ov_limit has a default value of 100 cm^/sec which is too strong for some stars. Living in the era of high-precision frequencies from Kepler, we see evidences that D_mix_ov_limit should be fixed to much lower values to give room to smaller D_mix values to play. Thus, this default value sweeps out a whole deal of valuable information in the model. Additionally, I think the check is not needed at all. A user already has enough freedom to fix local changes of D_mix (also in the overshoot layer) through extra hooks.
> 
> Please consider my above words as few suggestions from a humble user; the Captain knows better how to steer the wheel ;-)
> 
> Best regards
> Ehsan.
> 
> 
>> On 04 Jun 2015, at 20:17, Bill Paxton <paxton at kitp.ucsb.edu> wrote:
>> 
>> Hi Ehsan,
>> 
>> On Jun 3, 2015, at 9:39 AM, Ehsan Moravveji wrote:
>> 
>>> Dear mesa-users,
>>> Today, I’ve been fiddling around the radiative turbulent diffusion (Morel & Thevenin, 2002, A&A), and its influence on the net mixing coefficient. I came up with few questions that I would like ask. 
>>> 
>>> One only needs to set radiation_turbulence_coeff to something greater than 0 to activate this. The relevant routine sits in star/private/turbulent_diffusion.f90. Reading through lines 64 to 84, you see (briefed in here):
>>> 
>>>        do k = 2, nz
>>>           if (s% mixing_type(k) /= no_mixing) cycle
>>>           …
>>> 
>>>           D = coeff*4*crad*T_face*T_face*T_face*T_face/(15*clight*opacity_face*rho_face*rho_face)
>>>           if (D < s% D_mix_ov_limit) exit
>>>           …
>>> 
>>>           s% cdc(k) = cdc
>>>           s% D_mix(k) = D
>>> 
>>>           s% conv_vel(k) = vc
>>>           s% mixing_type(k) = anonymous_mixing
>>>        end do
>>> 
>>> - Based on the original paper, the radiative turbulent mixing coefficient (D_rad) should be used across the whole star, although its effect will be swamped by the convective mixing in convective layers; thus, I would like to know why it is ignored if a zone is not flagged as no_mixing? I also cannot convince myself why D_rad cannot be “added” to other mixing coefficients, say semi-convective, thermohaline, etc.? Indeed, we do not know well how different mixing mechanisms interplay/interact. Some words from black belts would be very appreciated here.
>> 
>> good point.  the code is assuming that any other form of mixing will dominate, so it can skip calculating radiative turbulence at that cell boundary.
>> i'll change it to calculate the radiative turbulence mixing in all cases and use it if it is stronger than the previous mixing. 
>> i.e., delete the initial test of mixing_type, and after calculation of D, add "if (D <= s% D_mix(k)) cycle".
>> 
>>> 
>>> - Furthermore, why the associated convective diffusion coefficient and velocity, cdc and vc here, are stored in s% cdc and s% conv_vel, whereas these variable names are reserved for convection-related quantities? Are there specific design reasons behind this?
>> 
>> those names go back to the dark ages of mesa when all mixing was convective.  i haven't renamed them.
>> 
>>> 
>>> - And finally, is the add_radiation_turbulence() subroutine identical to the original CESAM version? 
>> 
>> I'll leave that one for you to check.
>> 
>> The above changes are in version 7626 if you'd like to test them.
>> 
>> Bill
>> 
>> 
>> 
>> 
>>> 
>>> Best regards
>>> Ehsan.
>>> 
>>> 
>>> 
>>> ------------------------------------------------------------------------------
>>> _______________________________________________
>>> 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/20150604/10e19cde/attachment.html>