A fast direct solver for network matrices

Per-Gunnar Martinsson

Home/Archive/arXiv

arXiv · 0706.4348

A fast direct solver for network matrices

Per-Gunnar Martinsson

View original source ↗Read source document ↗Export metadata ↓

Abstract

A fast direct inversion scheme for the large sparse systems of linear equations resulting from the discretization of elliptic partial differential equations in two dimensions is given. The scheme is described for the particular case of a discretization on a uniform square grid, but can be generalized to more general geometries. For a grid containing $N$ points, the scheme requires $O(N \log^{2}N)$ arithmetic operations and $O(N \log N)$ storage to compute an approximate inverse. If only a single solve is required, then the scheme requires only $O(\sqrt{N} \log N)$ storage; the same storage is sufficient for computing the Dirichlet-to-Neumann operator as well as other boundary-to-boundary operators. The scheme is illustrated with several numerical examples. For instance, a matrix of size $10^6 \times 10^6$ is inverted to seven digits accuracy in four minutes on a 2.8GHz P4 desktop PC.

Explore related subjects

math.NA cs.NA

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Per-Gunnar Martinsson. 2007-06-29. A fast direct solver for network matrices. https://arxiv.org/abs/0706.4348

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related papers

arXiv2026

DDFEM: A Python Package for Diffuse Domain Methods

Solving partial differential equations (PDEs) on complex domains can present significant computational challenges. The Diffuse Domain Method (DDM) is an alternative that reformulates the partial differential equations on a larger, simpler domain. The original geometry is embedded into the problem by representing it with a phase-field function. This paper introduces DDFEM, an extensible Python package to provide a framework for transforming PDEs into a Diffuse Domain formulation. We aim to make the application of a variety of different Diffuse Domain approaches more accessible and straightforward to use. The DDFEM package includes features to intuitively define complex domains by combining signed distance functions and provides a number of Diffuse Domain transformers for general second-order evolution equations. In addition, we present a new approach for combining multiple boundary conditions of different types on distinct boundary segments. This is achieved by applying a normalised weighting, derived from the component signed distance functions, to combine the additional boundary terms in the Diffuse Domain formulations without requiring additional phase fields. The domain definition and Diffuse Domain transformation provided by our package are designed to be compatible with a wide range of existing finite element solvers without requiring code alterations. Both the original (non-linear) PDEs provided by the user and the resulting transformed PDEs on the extended domain are defined using the Unified Form Language (UFL). Here we focus on the software implementation and demonstrate its effectiveness across a range of problems.

math.NA↗

arXiv2026

Diffuse Domain Methods with Dirichlet Boundary Conditions

The solution of partial differential equations (PDEs) on complex domains often presents a significant computational challenge by requiring the generation of fitted meshes. The Diffuse Domain Method (DDM) is an alternative which reformulates the problem on a larger, simple domain where the complex geometry is represented by a smooth phase-field function.This paper introduces and analyses several new DDM methods for solving problems with Dirichlet boundary conditions. We derive two new methods from the mixed formulation of the governing equations. This approach transforms the essential Dirichlet conditions into natural boundary conditions. Additionally, we develop coercive formulations based on Nitsche's method. We provide proofs of coercivity for all new methods and also for some approximations found in the literature for a wide range of linear advection-diffusion problems.Numerical experiments demonstrate the improved accuracy of the new methods, and reveal the balance between $L^2$ and $H^1$ errors. The practical effectiveness of this approach is demonstrated through the simulation of the incompressible Navier-Stokes equations on two fluid dynamics benchmark problems.

math.NA↗

arXiv2026

A second-order energy-stable sequential regularization scheme for the Navier-Stokes equations

In this paper, we construct a second-order decoupled scheme for the Navier-Stokes equations based on the idea of the auxiliary variable and the sequential regularization method (SRM). The proposed scheme enjoys unconditional energy stability, which is generally unavailable for the existing SRM, while preserving its key advantage of avoiding artificial pressure boundary conditions and maintaining the original velocity boundary conditions. We establish a viscosity-independent low-order stability estimate and a higher-order stability estimate for the proposed scheme. Furthermore, we derive global-in-time second-order error estimates in two dimensions and corresponding local-in-time estimates in three dimensions, without imposing any time-step restrictions. A series of numerical experiments confirm the convergence and stability, and demonstrate improved accuracy and incompressibility control compared with the projection method and the conventional SRM. In particular, both high-Reynolds-number and three-dimensional numerical simulations further demonstrate the robustness and reliability of the proposed method for complex incompressible flows.

math.NA↗