2025-11-22T17:46:15.249041

Uniqueness of solutions to MFG systems with large discount

Cirant, Continelli
We prove that solutions to a class of Mean Field Game systems with discount are unique provided that the discount factor is large enough, and the Lagrangian term is (proportionally) small enough. This identifies an asymptotic uniqueness regime that falls outside the usual ones involving monotonicity.
academic

Uniqueness of solutions to MFG systems with large discount

Basic Information

  • Paper ID: 2510.09280
  • Title: Uniqueness of solutions to MFG systems with large discount
  • Authors: Marco Cirant, Elisa Continelli
  • Classification: math.AP (Mathematics - Analysis of Partial Differential Equations)
  • Publication Date: October 13, 2025
  • Paper Link: https://arxiv.org/abs/2510.09280

Abstract

This paper establishes the uniqueness of solutions to a class of discounted Mean Field Game (MFG) systems under the condition that the discount factor is sufficiently large and the Lagrangian term is (relatively) sufficiently small. This identifies an asymptotic uniqueness region that differs from the conventional uniqueness regions involving monotonicity conditions.

Research Background and Motivation

Problem Background

Mean Field Game theory was independently proposed by Lasry-Lions and Huang-Caines-Malhamé to describe Nash equilibria among large populations of indistinguishable rational agents. However, the uniqueness of solutions to MFG systems has remained a central challenge in the theory.

Research Motivation

  1. Theoretical Gap: While MFG systems are known to possess unique solutions under certain monotonicity conditions, the uniqueness problem remains unresolved for general cases that do not satisfy these conditions.
  2. Limiting Behavior: Work by Bardi and Cardaliaguet demonstrates that as the discount parameter λ→∞, solutions to MFG systems converge to solutions of nonlinear parabolic PDEs, which possess uniqueness. This inspires the idea of "borrowing uniqueness" from the limiting problem.
  3. Dual Role: The parameter λ simultaneously acts as a time discount factor and as the reciprocal of a control cost coefficient. As λ increases, agents place less weight on future events while "reaction costs" become cheaper.

Limitations of Existing Approaches

Traditional MFG uniqueness results primarily rely on:

  • Lasry-Lions monotonicity conditions
  • Displacement monotonicity
  • Small parameter assumptions (such as short time horizons or weak interactions)

These conditions are often difficult to satisfy in practical applications, limiting the applicability of the theory.

Core Contributions

  1. Establishment of a New Uniqueness Mechanism: Proves that MFG systems possess unique solutions when the discount factor λ is sufficiently large, representing a novel uniqueness region that does not depend on traditional monotonicity conditions.
  2. Quantitative Convergence Results: Provides quantitative estimates for the convergence of λDu_λ to DF(x,m_λ(t)): ∣λDuλ(x,t)−DF(x,mλ(t))∣≲1λ|λDu_λ(x,t) - DF(x,m_λ(t))| \lesssim \frac{1}{\sqrt{λ}}
  3. Unified Analytical Framework: Develops technical tools for handling MFG systems over infinite time horizons, including non-local Grönwall-type lemmas and refined a priori estimates.
  4. Viscosity Independence: Proves that the critical constant Λ is independent of the diffusion parameter, which is significant in the vanishing viscosity limit.

Detailed Methodology

Problem Formulation

Consider the following MFG system (MFG_λ):

-\partial_t u_λ - \Delta u_λ + \frac{1}{λ}H(λDu_λ, x) + λu_λ = F(x, m_λ(t)) \\ \partial_t m_λ - \Delta m_λ - \text{div}(m_λ D_p H(λDu_λ, x)) = 0 \\ m_λ(x,0) = m_0(x) \end{cases}$$ The objective is to prove that when λ ≥ Λ, solutions satisfying specific growth conditions are unique. ### Core Technical Framework #### 1. A Priori Estimates Solutions are constructed by taking limits from finite time horizon problems, establishing key uniform estimates: - $|u_λ(x,t)| \leq \frac{C}{λ}(1 + |x|)$ - $\|Du_λ\|_∞ \leq \frac{C}{λ}$ - $\|D^2u_λ\|_∞ \leq \frac{C}{λ}$ #### 2. Representation Formula Using Duhamel's principle and duality methods, an integral representation of λDu is obtained: $$λ\frac{\partial u}{\partial x_i}(x,t) = \int_t^∞ \int_{\mathbb{R}^n} \left(λe^{-λ(s-t)}\frac{\partial F}{\partial x_i}(y,m(s)) - e^{-λ(s-t)}\frac{\partial H}{\partial x_i}(λDu(y,s),y)\right) ρ^{x,t}(y,s) dy ds$$ #### 3. Convergence Analysis Proves that λDu_λ converges uniformly to DF(·,m_λ(·)) with convergence rate O(1/√λ). This indicates that for large λ, Du_λ(t) depends primarily on m_λ at time t, rather than on future behavior. #### 4. Uniqueness Proof Strategy The key step establishes difference estimates between two solutions: $$\|λDu_1(·,t) - λDu_2(·,t)\|_∞ \lesssim d_1(m_1(t),m_2(t)) + \int_t^∞ d_1(m_1(r),m_2(r))e^{-(λ-η)(r-t)}dr$$ Combined with the non-local Grönwall lemma, when λ is sufficiently large, one can prove m_1 ≡ m_2, and consequently u_1 ≡ u_2. ### Technical Innovations 1. **Handling Infinite Time Horizons**: Develops analytical techniques applicable to MFG systems over infinite time intervals, overcoming limitations of traditional finite-time methods. 2. **Non-local Grönwall Lemma**: Proposes and applies a Grönwall-type inequality incorporating "future integral" terms, which is the key tool for handling non-local temporal dependence in MFG systems. 3. **Refined Second-Order Derivative Estimates**: Obtains uniform boundedness of D²u through maximum principles, which is rarely seen in previous literature. ## Main Assumptions ### Assumptions on F (F1) F(·,m) is of class C^{2,α} for all m ∈ P₁ and has linear growth (F2) F satisfies Lipschitz conditions with respect to both spatial and measure variables ### Assumptions on H (H1) H is of class C^{2,α} and convex in the first variable (H2) H satisfies quadratic growth and uniform ellipticity conditions ## Main Results ### Theorem 3.4 (Main Uniqueness Result) There exists a constant η' such that when λ ≥ Λ = max{λ₀, η'}, the MFG_λ system possesses a unique solution. ### Proposition 3.1 (Uniform Convergence) $$\|λDu - DF(·,m(·))\|_∞ \leq \frac{C}{\sqrt{λ}}(L+1)$$ ### Proposition 3.3 (Difference Estimates) For two solutions, there exist constants K, η such that: $$\|λDu_1(·,t) - λDu_2(·,t)\|_∞ \leq Kd_1(m_1(t),m_2(t)) + K\int_t^∞ d_1(m_1(r),m_2(r))e^{-(λ-η)(r-t)}dr$$ ## Technical Details ### Key Lemma 2.5 (Non-local Grönwall Lemma) Let f be a non-negative continuous function, and let a, b, δ > 0. If $$f(t) \leq a\int_0^t f(s)ds + b\int_t^∞ f(s)e^{-δ(s-t)}ds$$ and δ ≥ 4a + 4b, then f ≡ 0. This lemma is the key tool for handling inequalities containing "future integral" terms. ### Proof Strategy 1. Analyze density evolution using SDE methods 2. Establish integral expressions for Du through representation formulas 3. Estimate differences between distinct solutions 4. Apply the non-local Grönwall lemma to derive a contradiction ## Related Work ### MFG Uniqueness Theory - **Lasry-Lions Monotonicity**: Classical uniqueness condition requiring monotonicity of coupling terms - **Displacement Monotonicity**: Geometric methods developed in recent years - **Small Parameter Methods**: Obtaining uniqueness by restricting time horizons or interaction strength ### Large Discount Limits - **Bardi-Cardaliaguet (2021)**: Establishes convergence of MFG systems to agent-based models - **Weighted Energy Dissipation Mechanisms**: Similar phenomena in variational methods ## Conclusions and Discussion ### Main Conclusions This paper identifies a new uniqueness region for MFG systems: solutions are unique when the discount factor λ is sufficiently large. This result: 1. Does not depend on traditional monotonicity conditions 2. Has a critical value independent of the viscosity coefficient 3. Provides new perspectives for understanding asymptotic behavior of MFG systems ### Limitations 1. **Special Solution Classes**: Uniqueness holds only within solution classes satisfying specific growth estimates 2. **Non-local F**: Assumes F depends on m in a non-local manner (e.g., convolution-type) 3. **Large λ Restriction**: Results apply only to sufficiently large discount factors ### Future Directions 1. **Local Coupling**: Study cases where F depends locally on density m(x,t) 2. **Long-Time Behavior**: Analyze asymptotic behavior and steady states of solutions 3. **Multiple Solutions Selection**: Selection mechanisms when the limiting problem admits multiple solutions 4. **Numerical Verification**: Develop corresponding numerical methods to verify theoretical results ## In-Depth Evaluation ### Strengths 1. **Theoretical Innovation**: Opens new directions in MFG uniqueness research, independent of traditional monotonicity 2. **Technical Rigor**: Develops a complete technical framework for handling infinite-horizon MFG systems 3. **Profound Results**: Reveals the dual role of the discount parameter and its impact on solution structure 4. **General Methods**: Tools such as the non-local Grönwall lemma have broad applicability ### Weaknesses 1. **Limited Scope**: Applies only to large discount cases, potentially limiting practical applications 2. **Solution Class Restrictions**: Cannot exclude the existence of other solutions 3. **Non-constructive**: Proofs are existential and do not provide algorithms for constructing solutions ### Impact This paper makes important theoretical contributions to MFG theory, particularly: - Extends the applicability of uniqueness theory - Provides new methods for studying limiting behavior of MFG systems - May inspire research on related nonlinear PDE systems ### Applicable Scenarios This theory is particularly suitable for: - Economic models where agents heavily weight immediate returns - Systems with relatively low control costs - Theoretical analyses requiring study of limiting behavior ## References The paper cites core literature in MFG theory, including: - Lasry-Lions (2007): Foundational work on MFG theory - Bardi-Cardaliaguet (2021): Convergence in large discount limits - Important advances in related uniqueness research