Speaker
Description
The plasma behaviour in the exhaust region of tokamaks is strongly influenced by atomic and molecular processes, as experimentally demonstrated on, e.g., MAST-U [1], JET, AUG [2], Alcator C-Mod [3]. Faithful representation of these processes is therefore necessary for robust exhaust simulation capability. However, fully-resolved atomic and molecular reaction systems typically include hundreds and even thousands of excitation and rovibration levels across ionisation stages [4], rendering direct inclusion into exhaust fluid or kinetic codes computationally unfeasible. At the same time, an asymptotic treatment is not applicable, as neither local-thermodynamic nor coronal equilibriums hold in the typical divertor conditions. This, combined with high stiffness due to extreme timescale ranges, warrants for reduced (effective) models accounting for both collisional and radiative processes.
CRFAX (Collisional Radiative modelling Framework in JAX), being developed at UKAEA, is a Python package for modelling, analysis, reduction and post-processing of atomic and molecular systems in plasmas. A non-exhaustive list of present capabilities includes reduction methods for linearised reaction systems (a method by Greenland [5] and quasi-steady-state assumption), automatic reduced-model identification method [5], calculation of validity timescales and metrics of reduced models, solvers for linear and non-linear models, as well as calculation of rate contributions (emission spectrum and energy balance). Multi-dimensional scanning, automatic differentiation [6] and gradient-based fitting applied to parametrised reaction systems allow for applications such as tabulation of effective rate coefficients, sensitivity analysis and spectrum fitting. The provenance of CRFAX objects is tracked using W3C Prov data model [7]. The future development of CRFAX is aimed at reduction techniques for non-linear systems [8, 9] and implementation of uncertainty quantification.
[1] K. Verhaegh et al., Nucl. Fusion 63, 016014 (2023).
[2] M. Bernert et al., Nucl. Mater. Energy 12, 111-118 (2017).
[3] A. Mathews et al., Rev. Sci. Instrum. 93, 063504 (2022).
[4] K. Sawada, M. Goto, Atoms 4(4), 29 (2016).
[5] P. T. Greenland, Proc. R. Soc. A: Math. Phys. Eng. Sci., 457, 1821-1839 (2001).
[6] R. Frostig, M. J. Johnson, C. Leary, SysML Conf. 2018, hal-05188750, (2019).
[7] L. Moreau et al., PROV-DM: The PROV data model, http://www.w3.org/TR/prov-dm/ (2013).
[8] U. Maas, S.B. Pope, Proc. Combust. Inst. 24, 103-112 (1992).
[9] S. H. Lam, D. A. Goussis, Int. J. Chem. Kinet. 26(4), 461-486 (1994).