Localized Lagrange multipliers with weak enforcement of constraints and Mortar method equivalence
- 1Czech Academy of Sciences, Institute of Thermomechanics
- 2University of California San Diego
- 3Universidad de Sevilla
Abstract
Preprint of the article.
Abstract: Domain decomposition methods based on Lagrange multipliers are widely used to couple nonconforming finite element discretizations. Among them, the Mortar Method (MM) and the Localized Lagrange Multiplier (LLM) method are often regarded as fundamentally different approaches. While MM enforces interface constraints variationally, LLM is commonly associated with collocation constraints based on Dirac delta multipliers. This work revisits the original
concept of localized Lagrange multipliers and argues that the defining feature of LLM is the localization of multipliers to individual subdomains through a frame construction rather than the use of Dirac delta functions. Based on this interpretation, a weak localized Lagrange multiplier formulation is derived. It is shown that the derived method is mathematically equivalent to the classical mortar method, but only when appropriate interpolation spaces are employed. The solvability and stability of the resulting discrete system are investigated, and the frame-based formulation is shown to naturally accommodate crosspoints involving multiple subdomains. Numerical examples confirm the theoretical findings and demonstrate that weak LLM achieves the same accuracy as MM while preserving the implementation advantages associated with localized multipliers.
Related
- Degeneracies of Lagrange functions and geometric interpretations of Lagrange multipliers
- On the Choice of Continuous and Discontinuous Lagrange Multipliers in Fictitious-Domain Methods
- On error bounds and noncritical Lagrange multipliers
- Dual companions and improved regularity of Lagrange multipliers in optimal control
- Transport and thermalization under noncommuting localized constraints
BibTeX
@misc{dvorak2026localized,
title = {{Localized Lagrange multipliers with weak enforcement of constraints and Mortar method equivalence}},
author = {Radim Dvořák and Murtaza Bookwala and Petr Krysl and Radek Kolman and José A. González},
howpublished = {Zenodo},
year = {2026},
month = oct,
doi = {10.5281/zenodo.23692397},
url = {https://zenodo.org/records/23692397},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Localized Lagrange multipliers with weak enforcement of constraints and Mortar method equivalence
AU - Radim Dvořák
AU - Murtaza Bookwala
AU - Petr Krysl
AU - Radek Kolman
AU - José A. González
PY - 2026
DA - 2026/10/22
PB - Zenodo
DO - 10.5281/zenodo.23692397
UR - https://zenodo.org/records/23692397
AB - Preprint of the article. Abstract: Domain decomposition methods based on Lagrange multipliers are widely used to couple nonconforming finite element discretizations. Among them, the Mortar Method (MM) and the Localized Lagrange Multiplier (LLM) method are often regarded as fundamentally different approaches. While MM enforces interface constraints variationally, LLM is commonly associated with collocation constraints based on Dirac delta multipliers. This work revisits the original concept of localized Lagrange multipliers and argues that the defining feature of LLM is the localization of multipliers to individual subdomains through a frame construction rather than the use of Dirac delta functions. Based on this interpretation, a weak localized Lagrange multiplier formulation is derived. It is shown that the derived method is mathematically equivalent to the classical mortar method, but only when appropriate interpolation spaces are employed. The solvability and stability of the resulting discrete system are investigated, and the frame-based formulation is shown to naturally accommodate crosspoints involving multiple subdomains. Numerical examples confirm the theoretical findings and demonstrate that weak LLM achieves the same accuracy as MM while preserving the implementation advantages associated with localized multipliers.
ER -