Description
In plasma experiments temporally and spatially resolved heat loads on material surfaces are of interest. These can be derived from surface temperature measurements from infrared (IR) cameras by solving the heat diffusion equation. This involves non-trivial boundary conditions and non-linearities of material properties. A widely used tool for this task is the THEODOR (THermal Energy Onto DivertOR) code. Many upgrades and improvements have been introduced in recent years, such as transitioning from an explicit to an implicit solver, variable-depth discretisation, including inhomogeneities in geometry and material properties, and an extension from 2D to 3D.
These updates were necessary for heat flux evaluation in the new upper divertor in AUG capable of accessing alternative divertor configurations (ADCs). This is due to the initial formation of a single X-point followed by a transition to magnetic equilibria with potentially multiple strike lines in different locations [1], and often a strong flux expansion variation. A 2D solver leads to biased results due to the more complex tile geometry, even when toroidally symmetric heat loads are assumed. This is because the tile design itself is not toroidally symmetric with cut-outs for mounting points significantly influencing heat flow during experiments. In areas with low incidence angles the use of a 3D solver is also required to quantify toroidal variations and leading edges.
Evaluating THEODORs in-tile temperatures enables direct comparison to thermocouple data. This allows verification of the total energy input in the presence of potentially spurious IR radiation from the plasma or reflections, and calibration of surface emissivity. It also adds to the comparison with heat flux profiles from Langmuir probes with systematic differences to the IR analysis.
The updated THEODOR is about to be open-sourced under an MIT licence with version control to support the fusion community. This is a Python implementation with uniform interfaces for the 1D, 2D and 3D solvers. It provides comparable speed to the C++ version and parallelisation. Evaluation speeds of around 100 Hz for 3D geometries of 200x100x50 cells make it compatible with online analysis.
[1] T Lunt et al 2025 FEC https://conferences.iaea.org/event/392/contributions/37792/attachments/21076/36009/Lunt_synopsis_IAEA_submitted.pdf