Enable XMH geometry for TGLF in TGYRO runs

[halpernf]

The following changes have been made

• Add an input parameter tgyro_tglf_geometry_flag in tgyro inputs
• Transfer the existing XMH coefficients from tgyro to tglf in tgyro_tglf_map if new input flag is set to 4

tgyro_tglf_geometry_flag has default 1 to preserve backwards compatibility.

[jcandy]

Thanks Federico!

Would it be better to call this TGYRO_TGLF_MXH_FLAG?

[halpernf]

Would it be better to call this TGYRO_TGLF_MXH_FLAG?

My original idea was to subsequently add a geometry_flag=4 option in tglf for XMH vs the current approach. Right now, XMH is on by default on standalone TGLF with geometry_flag=1, not backwards compatible.

Hence the name tgyro_tglf_geometry_flag

Maybe an even better idea is to match what neo and cgyro are doing at the moment. I'm happy to make modifications as people prefer.

[gmstaebler]

I am not sure a new flag is needed unless you want to have TGYRO reduce the XHM shape functions to the standard Miller (i.e. zero the extra ones) before calling tglf. I guess if the default is for all of the shape coefficients to be computed then it would be necessary to use this flag to recover the original Miller. Note that standard Miller has zmaj_loc=0, dzmajdx_loc=0 , zeta_loc=0 and zeta_s_loc=0. These were added by Jeff before XHM so there are really three version of Miller.

[jcandy]

@halpernf @gmstaebler

I would like to have you both agree about these changes, at least so you both know exactly what the new behaviour will be. I am happy with whatever is agreed upon.

[halpernf]

I am not sure a new flag is needed unless you want to have TGYRO reduce the XHM shape functions to the standard Miller (i.e. zero the extra ones) before calling tglf.

@gmstaebler that was my original idea -- the TGYRO regression tests are converging more slowly with the XHM shape functions. My concern is whether existing TGYRO+TGLF workflows would be disrupted by this upgrade. I'm inclined to at least have a parameter to preserve the existing behavior. Let me know what you prefer, also a shoutout @tomneiser who is another TGYRO/TGLF user. I'm happy to go with the flow

[jmcclena]

Looks fine to me, other than the discussion of tglf geometry variable flag names which I have no strong feeling one way or the other.

[gmstaebler]

It is good to be able to recover the previous default. The way I had miller_geo set up in TGLF before the XHM was added set zmaj_loc and dzmaj_dx_loc to zero. This should be done to the inputs from TGYRO as well as zeroing the shape and shape_s functions if the flag is set.

[gmstaebler]

to be clear the TGYRO flag TGYRO_TGLF_MXH_FLAG would be 1 if you want to use XHM and 0 if you want the previous Miller model. The coding in tgyro_tglf_map.f90 would need to be added to cover these options.

[halpernf]

to be clear the TGYRO flag TGYRO_TGLF_MXH_FLAG would be 1 if you want to use XHM and 0 if you want the previous Miller model. The coding in tgyro_tglf_map.f90 would need to be added to cover these options.

Changed the TGYRO inputs so that it follows the behavior preferred by Gary

• Use previous Miller model with TGYRO_TGLF_MXH_FLAG=0 (default)
• Use MXH model with TGYRO_TGLF_MXH_FLAG=1

Last comment: only harmonics up to sin3/cos3 are included in TGYRO, but can be added easily

[jcandy]

It would make the formulation more elegant if someday we could replace triangularity and squareness with s1 and s2 (now, delta = sin(s1) and zeta = -s2).