Stability analysis of Crank-Nicolson and Euler schemes for time-dependent diffusion equations (Q2350724)

For the named schemes, the authors take, for the staggered grid approximation, two versions of discrete boundary conditions (an explicit and an implicit version) to solve the 1D heat equation along with a diffusion coefficient depending only on time. Using explicit formulae for certain tridiagonal matrices they show that the backward Euler scheme is stable in both cases whereas the Crank-Nicolson scheme needs an additional condition for the explicit boundary conditions. These results and much more general ones are contained in a nearly 50 years old paper by \textit{A. A. Samarskij} [Sov. Math., Dokl. 9, 946--950 (1968); translation from Dokl. Akad. Nauk SSSR 181, 808--811 (1968; Zbl 0179.20201)], which appeared in Russian and in English, without proofs -- as usual in Doklady's papers. Proofs are contained e.g. in Russian books by the same author from 1971 and 1973 (see, e.g. [\textit{A. A. Samarskii}, Einführung in die Theorie der Differenzenschemata. (Russian), Moskau: Verlag ``Nauka'', Hauptredaktion für physikalisch-mathematische Literatur (1971; Zbl 0239.65065)]) (and show stability in the discrete energetic norm whereas the discrete space in which the present authors show stability remains unclear). Further, the Samarskij result gives, e.g., the condition \(0\leq \sigma^j(1-\theta)\leq 1\) (in the notation of the present authors) for the \(\theta\)-scheme on the staggered grid and for the explicit version of the boundary conditions.