“…This update might make the relation between f and g consistent at the end of the second step but it leads to undesirable numerical divisions by the small parameter ε; but, for well-prepared initial data, this consistency can be imposed at the beginning. Similar arguments were already given in [30,31] to avoid this update of the fluctuations g. Before proceeding further with the analysis of this kinetic method, we show in Figure 1 a comparison between the results of the three discussed kinetic methods: a semi-lagrangian PWENO6,4-interpolation scheme [10] (SL-WENO), the asymptotic preserving method without update of g proposed above and the asymptotic preserving method with update of g. In all cases, we show the L 2 t,x,v -error between the kinetic results and the solutions of the heat equation, its ε → 0 asymptotic limit, in a log-plot depending on ε. The results show that the kinetic scheme proposed in this paper works perfectly in the ε → 0 regime while both the updated scheme and the SL-WENO scheme do not describe well the asymptotic limit.…”