It was first formulated by H. A. Schwarz1 and served as a theoretical tool: its convergence for general second order elliptic partial differential equations was first proved much later, in 1951, by Solomon Mikhlin.2
The original problem considered by Schwarz was a Dirichlet problem (with the Laplace's equation) on a domain consisting of a circle and a partially overlapping square. To solve the Dirichlet problem on one of the two subdomains (the square or the circle), the value of the solution must be known on the border: since a part of the border is contained in the other subdomain, the Dirichlet problem must be solved jointly on the two subdomains. An iterative algorithm is introduced:
At convergence, the solution on the overlap is the same when computed on the square or on the circle.
The convergence speed depends on the size of the overlap between the subdomains, and on the transmission conditions (boundary conditions used in the interface between the subdomains). It is possible to increase the convergence speed of the Schwarz methods by choosing adapted transmission conditions: theses methods are then called Optimized Schwarz methods.3
Original papers
Conformal mapping and harmonic functions
PDEs and numerical analysis
See his paper (Schwarz 1870b) - Schwarz, H. A. (1870b), "Über einen Grenzübergang durch alternierendes Verfahren", Vierteljahrsschrift der Naturforschenden Gesellschaft in Zürich, 15: 272–286, JFM 02.0214.02 https://www.biodiversitylibrary.org/item/34472#page/280/mode/1up ↩
See the paper (Mikhlin 1951): a comprehensive exposition was given by the same author in later books - Mikhlin, S.G. (1951), "On the Schwarz algorithm", Doklady Akademii Nauk SSSR, n. Ser. (in Russian), 77: 569–571, MR 0041329, Zbl 0054.04204 https://mathscinet.ams.org/mathscinet-getitem?mr=0041329 ↩
Gander, Martin J.; Halpern, Laurence; Nataf, Frédéric (2001), "Optimized Schwarz Methods", 12th International Conference on Domain Decomposition Methods (PDF) http://www.ddm.org/DD12/Gander.pdf ↩