[mesa-users] Eq. 9 in Paxton et al. 2011

[Ehsan Moravveji]

Thanks Bill for your detailed explanations.

>         dP0 = s% surf_bc_offset_factor*s% cgrav(1)*s% mstar*s% dm(1)/(8*pi*r*r*r*r)
>         dT0 = dP0*s% gradT(1)*s% T(1)/s% P(1)
There still remains a puzzle for me to resolve:
I understand the necessity of the interlaced mesh for intensive and extensive quantities for the sake of numerical efficiencies. 
However, what I see is the blend of two variable types in e.g. dT0 equation above; here, gradT is defined at the face, whereas P and T are defined as cell averages. But, both are used on the RHS of dT0 equation.
Why is this an allowed operation?

Even further than that, such a blending occurs in a different style for the governing structure equations. Based on the first instrument paper, Eqs. (5) and (11) are defined/solved at cell centers, while Eqs. (7), (8) and (14) are solved at cell face. 
It is intriguing for me that it is possible to solve for the structure and evolution in this blend style. I was expecting that e.g. all equations be taken to the cell face, and be solved.
I'm just trying to gain deeper insight into the design of the code, so your comments will be enlightening.

I guess you have already addressed this topic somewhere in the mailing list, but I have no idea which keywords to use to retrieve it.
I will try test_suite/low_z with/without s% surf_bc_offset_factor and let you know.

Happy weekend,
Ehsan.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.mesastar.org/pipermail/mesa-users/attachments/20140912/65e320c8/attachment.html>