The FreeBoundaryStability/eigenmode_energies datasets in gpec.h5 (power-normalized
total / plasma / vacuum energies et[1], ep[1], ev[1]) read ~1e-15, i.e. effectively
zero, while the ideal-stability physics is correct: the in-memory eigenvalue et[1]
is order unity (about 0.8 for the DIII-D-like example), matching DCON (total dW about
+0.76, stable) and the parallel-integration test (et_par about 0.80).
So this is an output / serialization issue in the eigenmode-energy HDF5 write (or its
normalization), not a physics error. It is pre-existing: the regression has tracked
these quantities as ~1e-15 since at least the develop-era baseline.
Fix: write the correct power-normalized eigenmode energies to gpec.h5, and re-pin the
corresponding regression quantities once corrected.
The FreeBoundaryStability/eigenmode_energies datasets in gpec.h5 (power-normalized
total / plasma / vacuum energies et[1], ep[1], ev[1]) read ~1e-15, i.e. effectively
zero, while the ideal-stability physics is correct: the in-memory eigenvalue et[1]
is order unity (about 0.8 for the DIII-D-like example), matching DCON (total dW about
+0.76, stable) and the parallel-integration test (et_par about 0.80).
So this is an output / serialization issue in the eigenmode-energy HDF5 write (or its
normalization), not a physics error. It is pre-existing: the regression has tracked
these quantities as ~1e-15 since at least the develop-era baseline.
Fix: write the correct power-normalized eigenmode energies to gpec.h5, and re-pin the
corresponding regression quantities once corrected.