Mosco Convergence in Variational Analysis

• Mosco convergence is a variational concept for convex sets and lsc functionals that unifies weak and strong topologies.
• It ensures the convergence of resolvents, semigroups, and gradient flows across Hilbert and Banach spaces, providing rigorous analytical control.
• Its applications span nonlocal-to-local limits, moving domain problems, and quantum graph approximations, though it may not preserve global path properties.

Mosco convergence is a fundamental variational concept governing the limiting behavior of families of convex sets, lower semicontinuous functionals, and, most prevalently, closed (typically Dirichlet) forms on Hilbert and Banach spaces. This notion unifies weak and strong topologies to encode lower semicontinuity and recoverability of limiting energies, providing a precise analytic framework for establishing convergence of semigroups, resolvents, gradient flows, and Markov processes. It has wide-ranging applications in analysis, probability, geometric measure theory, PDEs, and the mathematical foundations of stochastic particle systems.

1. Core Definitions and Foundational Results

The classical setting involves a sequence of closed, convex, lower semicontinuous functionals or forms $(\mathcal{E}_n, \mathcal{F}_n)$ on a Hilbert space $H$, converging to $(\mathcal{E}, \mathcal{F})$. Mosco convergence is defined by the two conditions:

• (M1) Liminf inequality: For any sequence $u_n\rightharpoonup u$ (weakly) in $H$,

$$\liminf_{n\to\infty} \mathcal{E}_n(u_n) \geq \mathcal{E}(u).$$

• (M2) Recovery sequence: For any $u \in H$, there exists $u_n \to u$ (strongly) in $H$ such that

$$\limsup_{n\to\infty} \mathcal{E}_n(u_n) \leq \mathcal{E}(u).$$

For closed convex sets $H$0, Mosco convergence ($H$1) amounts to (i) every $H$2 is approximated strongly by some $H$3, and (ii) all weak limits of sequences $H$4 lie in $H$5 (Boccardo et al., 9 May 2025, Menaldi et al., 2021). The extension to varying spaces and non-Hilbert settings is codified via the Kuwae–Shioya framework, where convergence of underlying Hilbert spaces is prescribed through realization maps and intrinsic strong/weak topologies (Ayala, 2024, Grothaus et al., 2021).

Notably, Mosco convergence is equivalent to strong resolvent convergence and convergence of the associated linear/nonlinear semigroups:

• If $H$6, then the resolvent operators $H$7 converge strongly to $H$8 for the associated self-adjoint generators $H$9 (Löbus, 2012, Ayala, 2024).
• In Hadamard (CAT(0)) spaces, Mosco convergence of convex, lsc functions is equivalent to pointwise convergence of Moreau–Yosida envelopes and proximal mappings, with semigroup convergence for the associated nonlinear flows (Bačák et al., 2016, Bacak, 2012, Berdellima, 2020).

2. Methodological Framework and Criteria

The Mosco framework imposes two-sided control on variational sequences:

• Liminf conditions guarantee lower semicontinuity under weak convergence, ensuring tightness and energy bounds for all possible weak limit points.
• Recovery sequence construction demonstrates that any putative limit energy is actually achievable, implying the space of admissible functions is sufficiently rich even after taking limits.

For Dirichlet forms and energy functionals on $(\mathcal{E}, \mathcal{F})$0 spaces with varying measures, critical technical criteria include:

• Weak convergence of reference measures;
• Control of domain approximations via density or compactness;
• Fiberwise closability and uniform coercivity (Hamza's theorem, residual-perturbation bounds) (Grothaus et al., 2021);
• For jump and nonlocal operators: uniformity in kernel behavior, vanishing of long-range terms (localization), and translation invariance (Gounoue et al., 2019, Yang, 2019, Suzuki et al., 2014).

In problems involving variable spaces (thin domains, Gromov–Hausdorff limits), convergence is mediated by realization maps between reference spaces and uniform control on functionals or inner products (Kuroda, 2011, Nobili et al., 17 Nov 2025).

3. Mosco Convergence in Geometric and Nonlocal Settings

Mosco convergence describes both local-to-local and nonlocal-to-local limits, with robust consequences for the passage from discrete or nonlocal models to continuum variational structures.

• Nonlocal to Local: Families of stable-like, fractional, or truncated Dirichlet forms converge Mosco-wise to local forms as the jump parameter approaches the diffusive regime, under geometric hypotheses (volume doubling, Poincaré inequalities) (Yang, 2019, Gounoue et al., 2019). This transition aligns nonlocal Markov processes with their local diffusion limits, unifying jump and diffusion phenomena on metric measure spaces.
• Varying Domains: The Mosco framework controls the convergence of function spaces (e.g., Sobolev, H(curl)) over varying, possibly nonsmooth or perforated domains, ensuring stability of PDE solutions and well-posedness under complex boundary perturbations (Fornoni et al., 2022, Liu et al., 2016). In quantum graph/“fat graph” models, thin domain Dirichlet forms with delta potentials Mosco-converge to graph Laplacians with δ-coupling (Kuroda, 2011).
• Metric-Measure Limits: On spaces with Ricci curvature lower bounds in the sense of Lott–Sturm–Villani, Cheeger energies Mosco-converge under measured Gromov–Hausdorff convergence, establishing the analytic robustness of the RCD/CD(M,K) and BV theories (Nobili et al., 17 Nov 2025).

4. Impact on Evolution Equations, Variational Inequalities, and Particle Systems

Via equivalences between Mosco convergence and semigroup/resolvent convergence, this notion is pivotal for stability of elliptic and parabolic PDEs, quasilinear and variational inequalities, and stochastic processes:

• Domain Perturbation: Stability of parabolic variational inequalities under changing domains or constraints is characterized by Mosco convergence of the underlying convex sets in appropriate Bochner spaces (Ngiamsunthorn, 2011). The equivalence between Mosco convergence for elliptic and parabolic settings ensures solution convergence of both types of problems.
• Moving Sets and Obstacles: In obstacle and moving set problems for monotone operators, Mosco convergence ensures the stability of variational inequality solutions—even in the presence of natural-growth lower-order terms (Hamiltonians with critical growth) (Boccardo et al., 9 May 2025, Menaldi et al., 2021). The same theory extends to quasi-variational inequalities and impulse-control problems.
• Particle Systems and Self-duality: For $(\mathcal{E}, \mathcal{F})$1-particle interacting diffusions and Markov processes, Mosco convergence (often in the Kuwae–Shioya sense for varying spaces) provides the analytic basis for weak convergence to McKean–Vlasov or mean-field limits, for both symmetric and non-symmetric bilinear forms (Löbus, 2012, Ayala, 2024, Bacak, 2012). This underpins the hydrodynamic scaling of lattice particle systems and the analysis of correlation scaling limits for systems with Markov moment duality.

5. Stability, Instability, and Limitations

While Mosco convergence guarantees strong convergence of semigroups and thus convergence of finite-dimensional distributions for associated Markov processes, it does not preserve certain global path properties, such as transience/recurrence or conservativeness/explosion. Transition discontinuities—e.g., from recurrent to transient, or conservative to explosive processes—can occur under Mosco convergence of the energy forms, even with sequences obeying local $(\mathcal{E}, \mathcal{F})$2 convergence of coefficients, exponents, or jump kernels (Suzuki et al., 2014, Li et al., 2015). These results show that Mosco convergence is fundamentally a variational and local tool rather than a probabilistically global one.

Stability Property	Preserved under Mosco Conv.?	Reference
Semigroup/Resolvent convergence	Yes	(Löbus, 2012, Bacak, 2012)
Finite-dimensional distributions	Yes	(Löbus, 2012, Ayala, 2024)
Recurrence/Transience	No	(Suzuki et al., 2014, Li et al., 2015)
Conservativity/Explosion	No	(Suzuki et al., 2014, Li et al., 2015)

Uniform resolvent convergence or additional regularity (e.g., uniform ellipticity, strong coefficient convergence) may be used when global path properties are required to be preserved.

6. Advances and Extensions

Bridging variational analysis, convex geometry, and stochastic analysis, Mosco convergence has been generalized in several directions:

• Metric Spaces and Nonlinear Flows: In Hadamard (CAT(0)) spaces, Mosco convergence of functionals is characterized by convergence of Moreau envelopes, equivalence with Frolík–Wijsman convergence for sets, and is metrizable via a complete metric on the cone of convex lsc functions (Bačák et al., 2016, Berdellima, 2020).
• Varying Hilbert/Banach Spaces: Singular limits of gradient flows, including thin domain and discrete-to-continuum problems, require an abstract “Mosco along connecting operators” framework (Giga et al., 20 Nov 2025).
• Non-convex Potentials: Mosco criteria have been adapted to gradient Dirichlet forms with non-convex interaction potentials by controlling residual-perturbation operators and fiberwise approximations, relaxing the classical log-concavity assumptions (Grothaus et al., 2021).

7. Illustrative Examples and Applications

Nonlocal-to-local limits:

• Stable-like jump forms on metric measure spaces converge Mosco-wise to canonical local Dirichlet forms as the stability parameter approaches the critical diffusive value (Yang, 2019, Gounoue et al., 2019).

Moving obstacles:

• Stability of variational solutions to quasilinear obstacle problems is established under Mosco-convergence of the obstacle sets, applicable to obstacle homogenization and shape optimization (Boccardo et al., 9 May 2025).

Quantum graphs:

• Dirichlet forms on thin domains with delta potentials Mosco-converge to the corresponding Laplacians on graphs with delta couplings, rigorously justifying spectral approximations in quantum waveguides (Kuroda, 2011).

Particle systems:

• Mosco convergence in the sense of Kuwae–Shioya ensures that Dirichlet forms for independent or weakly interacting $(\mathcal{E}, \mathcal{F})$3-particle systems converge to the forms of the limiting stochastic process, with direct applications to hydrodynamic limits and self-duality in models such as the symmetric inclusion process (Ayala, 2024).

---

In summary, Mosco convergence is the central paradigm for variational convergence in modern analysis, providing robust mechanisms for passing to scaling limits, handling moving or rough constraints and domains, and optimizing over classes of energy functionals—while at the same time highlighting the need for finer analysis to control global dynamic properties of associated stochastic processes.