Augmented Lagrangian preconditioners

Augmented Lagrangian preconditioners for systems arising from fictitious domain discretizations

Fictitious domain methods allow one to solve problems on complex or moving geometries without generating a body-fitted mesh: the constraint on the immersed boundary is enforced weakly through a Lagrange multiplier in the variational formulation, leading to large (and multiple) saddle point system that are notoriously hard to solve.

Together with Michele Benzi, Luca Heltai and Federica Mugnaioni, I develop augmented Lagrangian (AL) preconditioners for these systems that are optimal (iteration counts independent of the mesh size) and scalable in parallel (Benzi et al., 2026). The same ideas extend to immersed surfaces (Benzi et al., 2026), elliptic interface problems (Benzi et al., 2026), and a unified variational framework covers both fictitious domain and mixed-dimensional couplings (Benzi et al., 2026). Ongoing work explores the extension of these approaches to fluid-structure interaction problems.

The implementation is open and based on the deal.II library. It is available at fdrmrc/fictitious_domain_AL_preconditioners.

References

2026

  1. A unified variational framework for augmented Lagrangian preconditioning of saddle point systems: applications to mixed-dimensional and fictitious domain methods
    Michele Benzi, Marco Feder, Luca Heltai, and Federica Mugnaioni
    arXiv preprint, 2026
    Under review
  2. Scalable augmented Lagrangian preconditioners for fictitious domain problems
    Michele Benzi, Marco Feder, Luca Heltai, and Federica Mugnaioni
    Computer Methods in Applied Mechanics and Engineering, 2026
  3. Augmented Lagrangian preconditioners for fictitious domain formulations of elliptic interface problems
    Michele Benzi, Marco Feder, Luca Heltai, and Federica Mugnaioni
    Computers & Mathematics with Applications, 2026
  4. Operator preconditioning for fictitious domain discretizations
    Marco Feder
    2026
    In preparation