[Paper Review] A new practical framework for the stability analysis of perturbed saddle-point problems and applications

Qingguo Hong, Johannes Kraus+2 more|arXiv (Cornell University)|Mar 16, 2021

Advanced Numerical Methods in Computational Mathematics44 references4 citations

TL;DR

This paper introduces a novel norm-fitting framework for analyzing the stability of perturbed saddle-point problems in Hilbert spaces, enabling shorter, more transparent proofs of inf-sup conditions. The method constructs problem-specific norms from seminorms, leading to parameter-robust preconditioners and successfully applied to generalized Poisson, Stokes, vector Laplace, and Biot's equations with uniform stability results.

ABSTRACT

In this paper we prove a new abstract stability result for perturbed saddle-point problems based on a norm fitting technique. We derive the stability condition according to Babuska's theory from a small inf-sup condition, similar to the famous Ladyzhenskaya-Babuska-Brezzi (LBB) condition, and the other standard assumptions in Brezzi's theory, in a combined abstract norm. The construction suggests to form the latter from individual fitted norms that are composed from proper seminorms. This abstract framework not only allows for simpler (shorter) proofs of many stability results but also guides the design of parameter-robust norm-equivalent preconditioners. These benefits are demonstrated on mixed variational formulations of generalized Poisson, Stokes, vector Laplace and Biot's equations.

Motivation & Objective

• To develop a constructive, abstract framework for stability analysis of perturbed saddle-point problems with symmetric positive semidefinite perturbations.
• To generalize classical Brezzi and Babuška stability conditions to include non-zero C operators in symmetric two-by-two block systems.
• To guide the design of norm-equivalent, parameter-robust preconditioners through a systematic norm-fitting technique.
• To simplify and unify proofs of inf-sup conditions across mixed finite element formulations of complex PDEs.
• To demonstrate the framework's effectiveness on generalized Poisson, Stokes, vector Laplace, and Biot’s consolidation models.

Proposed method

• The framework constructs a combined abstract norm from individual, problem-specific seminorms fitted to the operator structure.
• It establishes a generalized Brezzi-type condition that implies Babuška's big inf-sup condition under a small inf-sup condition and norm-fitting assumptions.
• The method uses a variational formulation in Hilbert spaces and derives stability from the interplay between the A, B, and C operators in the saddle-point system.
• It introduces a norm-fitting technique that ensures the induced norms are equivalent to the natural energy norms, enabling robust preconditioning.
• The approach is applied to continuous variational formulations, with direct implications for discrete mixed finite element methods.
• Theoretical results are validated through detailed analysis of four model problems, including Biot’s equations with three-field formulations.

Experimental results

Research questions

• RQ1 Can a unified, constructive framework be developed to analyze the stability of perturbed saddle-point problems with non-zero C operators?
• RQ2 How can norm-fitting techniques simplify and shorten proofs of inf-sup conditions in mixed finite element methods?
• RQ3 What conditions ensure parameter-robustness in the stability analysis of saddle-point systems arising from mixed formulations?
• RQ4 To what extent can the proposed framework guide the construction of norm-equivalent preconditioners for complex PDE systems?
• RQ5 Does the framework preserve uniform stability across different parameter regimes in models like Biot’s consolidation equation?

Key findings

• The proposed framework generalizes Brezzi's theorem to include symmetric positive semidefinite perturbations, ensuring the Babuška condition holds under a small inf-sup condition.
• The norm-fitting technique leads to shorter, more transparent proofs of stability for mixed formulations of generalized Poisson, Stokes, vector Laplace, and Biot’s equations.
• For the three-field Biot model, the framework yields a uniform stability result with a parameter-robust preconditioner involving inverse operators of div-ε and div-div terms.
• A norm-equivalent preconditioner is explicitly constructed for the Biot model, based on the fitted norms, ensuring robust convergence independent of physical parameters.
• The framework enables the derivation of a lower bound on the inf-sup constant that is uniform in key parameters such as the Biot coefficient and fluid-to-solid stiffness ratio.
• The method provides a constructive path to preconditioners that are robust with respect to physical parameters like λμ and Rp, as demonstrated in the three-field Biot formulation.