TJLF: The quasi-linear model of gyrokinetic transport TGLF translated to Julia*

In order to study present fusion experiments, and plan future fusion reactors, it is important to accurately model the heat and particle losses due to turbulence in magnetically confined plasmas. The trapped gyro-Landau fluid code (TGLF) is a state-of-the-art model of turbulent transport that is widely used in the fusion community. We have enhanced the capabilities of TGLF by translating the code from Fortran 90 to the Julia Programming language (www.julialang.org). Julia's auto-differentiation capability allows testing the sensitivity of the heat flux calculated with TGLF in Julia (TJLF) to the free parameters of the model, which can be used to direct model development and facilitate recalibration of the model against databases of gyrokinetic simulations. Benchmarks against the original TGLF model demonstrate matching accuracy at accelerated convergence in a flux-matcher using the FUSE integrated modeling framework. Ongoing work leveraging the native GPU compatibility of Julia together with energetic particle physics of TGLF-EP will also be discussed. TJLF promises more efficient generation of databases to train machine learning representations of TGLF, which can be used for real-time control of experiments and reactor optimization. We plan to open source TJLF to facilitate community contributions to this critical tool for fusion energy research.