New features in the release v25.1
The production version of TRANSP has been updated. This release includes the February 2026 update of the TRANSP build, CI/CD, and production systems to run on PPPL’s new Flux cluster following the decommissioning of the old mccune cluster.
This release includes several updates to predictive TRANSP capabilities,
especially in PT_SOLVER. Users running predictive cases should review
the updated PT_SOLVER
options and the profile
relaxation and filtering options before rerunning older input files that set
solver residuals, Peclet-number-related factors, or profile-filtering controls
explicitly.
- New Features
-
PT_SOLVER refactoring. The interpretation of enhancement
factors and residuals for
PT_SOLVERdiffusivity calculations has been updated to better match the published description of the algorithm and to reduce purely numerical contributions to diffusivity. Default values have been modified accordingly. - Enabled saturation rules 2 and 3 for the TGLF predictive transport model.
-
Added filtering options to
PT_SOLVERto smooth experimental profiles before they are supplied to the predictive transport calculation. These options are intended to improve numerical robustness when profiles contain localized spikes, frame-to-frame jitter, or weakly conditioned gradients. -
Added option
NMODVPH=4for a poly-exponential representation of the Prandtl number. - Re-enabled the NEO bootstrap current calculation.
- Generalized the ST pedestal model.
- Added the option to use FUSE GKNN models as an alternative surrogate to TGLFNN for fast predictive runs.
-
Enabled use of the
gnuplotpackage for therplotutility. -
Made the minimum
Zefftreatment zone-local. -
General
PT_SOLVERsetup, including predictive-mode selection, equation selection, transport-region setup, model selection, residuals, and numerical diffusivity parameters, is described on the PT_SOLVER options page. -
The relaxation/filtering controls, including
relx_profile,lam_t_base,lam_r_base,shape_mode,lam_m_base,delta_base, andbdy_weight, are described on the profile relaxation and filtering options page. -
Additional background on the
PT_SOLVERalgorithm and the updated interpretation of residual and diffusivity terms is available in arXiv:2605.09720. - Bug Fixes
-
Fixed normalization of the
zmajTGLF input variable inPT_SOLVER. - Fixed an array allocation issue for TGLF with adiabatic electrons.
- Fixed an initialization bug in the EC heating calculation.
-
Moved
i2mexto an external library. -
Removed some
PT_SOLVERdebugging output to reduce slowdown in I/O-limited jobs.
Important note for PT_SOLVER users
Because of the PT_SOLVER refactoring, the default settings for
residuals and Peclet-number-related factors have changed. Most users rely on
the default values and do not need to modify these settings. Users who
explicitly set these variables should review the updated documentation and use
the recommended values.
Users should pay particular attention to older predictive input files that
explicitly set pt_residual%RES_TE, pt_residual%RES_TI,
pt_residual%RES_NE, pt_residual%RES_NMAIN,
pt_residual%RES_NIMP, pt_residual%RES_PPHI, or
pt_num_diffusivity parameters. Values copied from older runs may not
have the same practical meaning after the refactoring.
PT_SOLVER profile filtering and relaxation
The new filtering options are intended for cases in which fitted experimental profiles or prescribed profiles introduce numerical artifacts into predictive transport calculations. The filter modifies the profile passed to the anomalous transport model during the solver step. For profiles that are not evolved, the filtered result is used only inside the transport step and the original experimental profile is restored afterward. Repeated filtering therefore does not accumulate as a persistent change in the stored profile.
A practical starting point for reduced-transport cases with experimental
profiles that are restored after each step is to use soft shape control with
modest radial smoothing. In typical cases this corresponds to
shape_mode = 2, modest lam_r_base, moderate
lam_m_base, nonzero delta_base, bdy_weight
near unity, and little or no lam_t_base. Stronger filtering should be
used only after checking that important physical profile structure and gradient
scale lengths are not being removed.
Filtering may be combined with temporary relaxation of nonlinear convergence controls during the transition from interpretive to predictive evolution. This can reduce time spent resolving short-lived numerical transients at the first predictive steps. After the transition, users should return to the standard solver settings unless a convergence study supports a different choice.
Miscellaneous changes
Recommended user action
Most users can continue to use the release defaults. Users rerunning older
predictive input files should review the updated documentation before using
custom PT_SOLVER residuals, Peclet-number-related parameters, or
profile relaxation/filtering settings. After modifying these parameters, compare
against a baseline run by checking evolved temperature, density, and rotation
profiles, effective diffusivities, convective velocities, nonlinear iterations,
time-step reductions, and total runtime.