|
MPI-AMRVAC 3.2
The MPI - Adaptive Mesh Refinement - Versatile Advection Code (development version)
|
The legacy MHD magnetofriction relaxation, weighted optimization extrapolator, and Grad–Rubin extrapolator all write
with the same schema:
The diagnostics are evaluated on all active cells using the magnetic field in AMRVAC code units:
CW_sin_theta, epsilon_force, and epsilon_div should approach zero for a well-resolved force-free, solenoidal field. Magnetic energy is not a residual; it records which energy branch the iteration approaches.
The meaning of iteration follows the method's natural outer update:
The common CSV is always written. Method-specific detailed histories are disabled by default and can be enabled independently:
The corresponding public optimization and Grad–Rubin configuration types use the common member name write_detailed_history. When enabled, the detailed files remain the authoritative source for method-specific convergence and control quantities: MFR artificial time step and flux metric, optimization functional and accept/reject history, and Grad–Rubin field/energy change, P/N discrepancy, alpha change, and field-line classifications.
The common file uses active cells only. The optimization-specific weighted functional additionally includes its fixed lower node and taper weights, while the legacy MFR log uses an inner-domain mask. Their method-specific values are therefore not identical to the common active-volume diagnostics. Magnetic energy is in AMRVAC code units and can only be compared directly when runs use the same units and domain.