Title: Robust isogeometric preconditioners for the Stokes system based on the Fast Diagonalization method

Abstract: In this paper we propose a new class of preconditioners for the isogeometric discretization of Stokes system. %Their application involves the solution of a Sylvester-like equation or a linear system with a Kronecker matrix. Their application involves the solution of a Sylvester-like equation, which can be done efficiently thanks to the Fast Diagonalization method. Complicated computational domains may have an influence on their performance, since the preconditioners can incorporate only partially the information on the geometry parametrization, in order to maintain the structure above which is the key ingredient for their performance. On the other hand, the preconditioners are robust both with respect to spline degree and mesh size. In our numerical tests and comparison to an alternative approach based on the incomplete factorization of the system matrix, the preconditioner setup and application is always orders of magnitude faster. Moreover, for high-degree high-regularity splines (the so-called $k$-method) the computational cost of the Krylov solver is mainly related to the matrix-vector multiplication, indicating that further significant improvements are possible shifting to a matrix-free implementation: the proposed preconditioners are very well suited for this purpose.