Speaker
Description
Kinetic neutral transport remains one of the least reducible components of fusion exhaust modelling. While coupled plasma-edge tools such as SOLPS-ITER/EIRENE remain the standard route for integrated divertor simulations, standalone kinetic-neutral calculations provide a complementary way to isolate rarefied-gas effects, test boundary-condition assumptions, and interpret neutral transport patterns without the full complexity of plasma feedback.
In this contribution, we discuss the use of SPARTA, a direct simulation Monte Carlo code for rarefied-gas dynamics, as a standalone tool for divertor-relevant neutral transport studies. The emphasis is not on replacing established plasma-edge workflows, but on identifying where a DSMC treatment can provide useful diagnostic and modelling insight. We first outline the practical advantages and limitations of this approach relative to EIRENE-like neutral models, including the use of binary neutral--neutral collisions rather than BGK-type closures, the treatment of neutral--wall interaction models such as CLL and TRIM-informed reflection, and the construction of plasma-mimicking boundary conditions from prescribed neutral sources and fluxes.
We then present a set of simplified private-flux-region and toroidal-plane test cases motivated by divertor exhaust applications. These cases show that neutrals need not behave as an isotropic diffusive background: ordered sources, wall interaction, geometry, and neutral--neutral collisions can generate structured flows, preferential capture, backflow, and non-trivial momentum transport. Such behaviour has direct implications for reduced exhaust models, including the common treatment of pump surfaces through fixed albedo or absorption coefficients. In particular, DSMC calculations allow pump capture, return flux, and effective exhaust conductance to be estimated from the local rarefied flow rather than prescribed as constants.
The talk will focus on lessons for kinetic-neutral and exhaust-code developers: which assumptions are exposed by standalone DSMC, which quantities can be passed back to integrated edge codes, and where systematic comparison against EIRENE-like models is still needed.