GetBcyl Issue in FIELDLINES

[WillT22]

When running previously successful simulations on FIELDLINES, I am now running into an initialization loop error. Attached is the error file when run on NoMachine. The error looks to be with GetBcyl specifically.
NX_error.txt

[zhucaoxiang]

@WillT22 From the source, the error probably indicates that a point isoutside the VMEC domain.

               ! The GetBcyl Routine returns -3 if cyl2flx thinks s>1
               ! however, if cyl2flx fails to converge then s may be
               ! greater than 1 but cyl2flux won't throw the -3 code.
               ! In this case GetBcyl returns br,bphi,bz = 0.  So
               ! bphi == 0 or ier ==-3 indicate that a point is
               ! outside the VMEC domain.

 ERROR in GetBcyl Detected
 R,PHI,Z   1.7075000000000000        1.6755161285400391       0.60000000000000009
 br,bphi,bz,myworkid   7.9050503334599447E-323   6.9257608337865693E-310   6.9257608337865693E-310           6

[zhucaoxiang]

This might be also related to #142 and can be merged if so.

[lazersos]

When running previously successful simulations on FIELDLINES, I am now running into an initialization loop error. Attached is the error file when run on NoMachine. The error looks to be with GetBcyl specifically. NX_error.txt

@WillT22 Can you upload the VMEC wout file? I'd like to check this against the #142 issue as @zhucaoxiang suggests.

[WillT22]

@lazersos Here you go. Let me know if you need anything else.

wout_fb_original.txt

[lazersos]

Running this equilibrium through my routine, the issue at the axis appears a bit differently.

Most likely your error is due to your choice of grid size. Adjust NR or NZ slightly and see if the error goes away. GetBCyl relies on an inverse look-up which can fail sometimes for highly cusped equilibria such as the one you supplied.

[WillT22]

For the NR and the NZ parameters, I was using 201 divisions which is the number provided in the example on the VMECwiki. I have adjusted these values down to 180 and up to 220 individually and together in the FIELDLINES input file on a simulation that ran before this issue arose. All tests yielded the same error as shown above.

[lazersos]

@WillT22 Can you upload the input file (assume it's fixed boundary). I'd like to run this case. Also please post your command for running FIELDLINES.

I suspect what you're running into is an edge related issue which arrises for strongly cusped equilibria. The fix for the axis problem will most likely also provide improvements for the edge.

Unfortunately, I'm busy with hardware from now till Christmas so I don't think I'll have much time to work on this.

[WillT22]

@lazerso Here is the input file and the .batch file I use to run in FIELDLINES modified for the fixed boundary condition whenever you can get to it.

fixed_boundary_input.zip

[lazersos]

Sorry, took a while for me to get back to this. Here are the issues I see

1. You're running a field line diffusion calculation but provide no vacuum field source (coils or mgrid). This is not possible.
2. You PHIMAX is 2pi but for an NFP stellarator should be 2pi/NFP unless you're attempting to calculate error field response. But then you need many more NPHI planes.
3. I ran what you provided on the Lazerson branch and found no issue.