v1.1.0

[jcitrin]

https://github.com/google-deepmind/torax/releases/tag/v1.1.0

[rrrttt123]

Hello, I would like to customize the initial input kinetic profiles (ne, Te, Ti, rotation) in TORAX, so that the pedestal of the kinetic profile can be a hyperbolic tangent function instead of a linear function. At the same time, fixing the pedestal profiles during the simulation can prevent the bootstrap current profile at the pedestal region from becoming strange. Does TORAX currently support such input of kinetic profiles?

[jcitrin]

Hello! Thanks for your interest in running TORAX. Apologies for the delay in replying.

1. It is certainly possible to initialize ne, Te, Ti to any profile. Either python primitives (dicts), numpy arrays, or xarray DataArray inputs can be provided to profile_conditions n_e, T_e, and/or T_i, all compatible as inputs for a TimeVaryingArray. If any of these equations are "evolving" according to settings in the numerics config, then these inputs will provide the profiles at t=t_intial, and the boundary conditions for all times existing in the input data. If profiles are not evolving (i.e. are prescribed), then the prescribed data will be interpolated from the input data for all times. See https://torax.readthedocs.io/en/v1.1.1/configuration.html#time-varying-arrays

2. We do not yet have rotation profiles in TORAX. This is coming later this year for prescribed toroidal rotation profiles, neoclassical poloidal rotation, and grad(Pi) contributions, as well as the impact of ExB shear on turbulence. Momentum transport is on the roadmap but not the short-term roadmap.

3a. Fixing pedestal profiles directly is not currently supported in TORAX. It is not possible to set an internal boundary condition and have profiles be fully fixed outside that radius, while evolving inside that radius. The entire PDE is always solved for all radii. A pedestal is set with an adaptive source term. It is possible to influence the profile in the pedestal by setting a prescribed pedestal transport coefficient model, which can have radial dependence. See https://torax.readthedocs.io/en/v1.1.1/configuration.html#combined on how to do that.

3b. A relatively easy feature extension is possible for supporting fixing pedestal profiles. We can make the Ti_ped, Te_ped ne_ped, etc input variables profiles and not scalars. The user could then set e.g. a tanh, or we even provide tanh parameters as an input option. We then extend the adaptive source routine to work for the profile for all points at rho_norm >= rho_ped, and not just the single point at rho_ped. If this is something you are interested in, please set up a feature request as an issue.

[rrrttt123]

Thank you very much for your detailed explanation. I now understand how to set up the base profile in TORAX currently. Regarding the setting of the simulation area, I indeed want to set it as non-global，for example, rho = 0.2-0.8；and I will set up a feature request as an issue about this later.