Effect of Statistical Noise on Simulation Results with a Plasma Fluid Code Coupled to a Monte Carlo Kinetic Neutral Code

Abstract

Power exhaust is one of the major challenges that future devices such as ITER and DEMO will face. Because of the lack of identified scaling parameters, predictions for divertor plasma conditions in these devices have to rely on detailed modelling. Most plasma edge simulations carried out so far rely on transport codes, which most of the times consist of a fluid code for the plasma coupled to a kinetic Monte Carlo (MC) code for neutral particles. One of the main difficulties in interpreting code results is the statistical noise from the MC procedure, which makes it difficult to define a convergence criterion for the simulations. In this work, we elaborate on similarities between noisy transport code simulations and turbulence simulations, and argue that the time averaged solution is a well defined stationary solution for the system. We illustrate these ideas with a simple slab test case with fluid neutrals, to which we add synthetic noise. In this case, the effects of noise are found to be significant only at high noise levels and for large enough correlations times.

Analytical study of statistical error in coupled finite‐volume/Monte Carlo simulations of the plasma edge

“…We expect that in more practical plasma edge simulations, the statistical error scales the same way. This work is complementary to the work of Marandet et al, where it is observed that the use of short cycles also leads to increased statistical noise.…”

Grid resolution study for B2-EIRENE simulation of partially detached ITER divertor plasma

“…Moreover, the presence of the sudden transition from the attached to the detached divertor regime can contribute to bad local estimates. When the estimate for p is inaccurate, p = 1 can be assumed in equation (7). Because of these inaccuracies, we take a safety factor of 1.25 into account.…”

“…Because of the long simulation times, relatively coarse discretization grids are typically used, with only a few 10 3 grid cells. However, recent developments show that an order of magnitude speed up can be obtained without losing accuracy [5][6][7]. By reducing the number of MC particles in each iteration and averaging the results over several iterations, it has been demonstrated that accurate solutions can be obtained in a significantly shorter computational time.…”

Self‐consistent coupling of the three‐dimensional fluid turbulence code TOKAM3X and the kinetic neutrals code EIRENE

“…The code is thus run in exactly the same way as a transport code as a first but key step of the verification procedure. Convergence towards a steady‐state solution (in fact, a statistically stationary state due to statistical noise from EIRENE) is assessed by checking for stationarity of all the fields and verifying the global particle and energy balances in the simulations. We thus check that the flux absorbed in the wall () plus the neutral flux at the core edge interface are equal to the influx from the core sustaining the simulation (i.e., the integral of the forcing).…”