Top: comparison of multiple methods for calculating the view factor F₂₁ for two parallel facets (the case A) in Fig. 2) as a function of N_s (Eq. (9)). Bottom: relative deviation from M₄ as a function of N_s for different methods. The standard approach, M₁, is accurate to within 1% when N_s is larger than about 100, while the subdivision methods, M₃-Linear and M₃-R2, are in accurate within 1% for the entire range of N_s. Therefore, these methods are clearly superior to the M₁ and M₂ methods.