Pith. sign in

REVIEW 3 major objections 4 minor 35 references

The selection problem for some first-order stationary mean-field games

T0 review · 3 major / 4 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper proves that, as the discount rate vanishes, the limit of a first-order stationary mean-field game is the regular weak solution minimizing the weighted action functional $\int\langle u\rangle m\,dx$ among all regular weak…

desk verdict The weak-solution selection theorem is a genuine advance and looks right, but the classical existence results are stated more broadly than the proofs support: Proposition 3.4 needs β+1>0, which Assumption 2 does not guarantee. read the letter →

arxiv 1908.06485 v1 pith:H2FX2YUF submitted 2019-08-18 math.AP

Diogo A. Gomes , Hiroyoshi Mitake , Kengo Terai This is my paper · ORCID
classification math.AP MSC 35A0149L2591A13
keywords mean-fieldgamesHamilton-Jacobiequationsselectionproblemvanishingdiscountweaksolutionsregularityestimatesstationary
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper studies what happens to a stationary mean-field game on a torus when the discount rate tends to zero. It proves that, under polynomial growth and compactness assumptions, the discounted problem has a unique classical solution and that this solution converges to a unique classical solution of the limit problem. When uniqueness fails, the paper gives a selection rule: any weak limit of the discounted solutions must be a solution of the limit problem that minimizes the weighted action functional $\int\langle u\rangle m\,dx$ among all regular weak solutions. The rule is proven by constructing a phase-space measure from each discounted solution and passing to a limit measure that satisfies a holonomy constraint. An explicit one-dimensional example with non-unique solutions shows what the selection criterion predicts.

What carries the argument

The central object is a phase-space probability measure attached to each solution, the discounted holonomy measure $$\mu_\epsilon(dx,dv)=m_\epsilon(x)\,dx\otimes\delta_{-Du_\epsilon(x)}(dv).$$ It encodes the transport equation through the discounted holonomy condition $$\int(-\epsilon\phi+v\cdot D\phi)\,d\mu_\epsilon=-\epsilon\int\phi\,dx,$$ and its Lagrangian action equals $\epsilon\int u_\epsilon\,dx$. Passing $\epsilon\to0$ yields a holonomy measure $\mu$ satisfying $\int v\cdot D\phi\,d\mu=0$. Comparing the two measures through the Legendre transform of the quadratic Hamiltonian produces the monotone inequality whose $\epsilon\to0$ limit is (1.12), and hence the selection inequality (1.13). This is the mechanism that converts compactness of the discounted solutions into a variational selection principle.

What would settle it

Solve the explicit example $g(m)=m$, $V(x)=\pi\cos(2\pi x)$ numerically for a sequence $\epsilon\to0$ and test whether $u_\epsilon-\int u_\epsilon$ converges weakly in $H^1(\mathbb T)$ to the minimizer $\tilde u$ of $\int\langle u\rangle m\,dx$; convergence to the other regular weak solution $\hat u$ would disprove Theorem 1.9.

Watch

Extended reading notes

Core claim

The central discovery is a selection criterion for the vanishing-discount limit of first-order stationary mean-field games. If $(u_\epsilon,m_\epsilon)$ is a regular weak solution of the discounted problem and, after normalization, $u_\epsilon$ converges weakly in $H^1$ to $\bar u$ while $m_\epsilon$ converges weakly in $L^1$ to $\bar m$, then for any regular weak solution $(u,m)$ of the limit problem the difference term $\int(g(m_\epsilon)-g(m))(m_\epsilon-m)\,dx$ tends to zero. Since $g$ is strictly increasing, this forces $\bar m=m$. The same argument yields the variational inequality $\int\langle\bar u\rangle m\,dx\le\int\langle u\rangle m\,dx$ for every regular weak solution of the limit problem, so the vanishing-discount limit is the regular weak solution that minimizes the functional $u\mapsto\int\langle u\rangle m\,dx$. The paper also proves that when a uniform lower bound on the density and polynomial growth of $g$ hold, the discounted problem has a unique classical solution and the limit is classical and unique, with the refined rate $\|u_\epsilon-\bar H/\epsilon-u-\lambda\|_\infty+\|m_\epsilon-m\|_\infty\to 0$.

Load-bearing premise

The selection mechanism relies on the discounted densities $m_\epsilon$ converging strongly in some $L^p$ space (not merely weakly) as $\epsilon\to0$; if only weak convergence is available, the monotonicity argument cannot force the vanishing-discount density to coincide with the density of the limit solution.

Editorial extensions

If this is right

  • If the selection inequality holds, the vanishing-discount limit is singled out by a variational principle: among all regular weak solutions of the stationary problem, it minimizes $u\mapsto\int\langle u\rangle m\,dx$.
  • In the classical regime, the limit is unique up to constants and smooth, with the refined expansion $u_\epsilon=\bar H/\epsilon+u+\lambda+o(1)$ and $m_\epsilon=m+o(1)$.
  • The monotone term $\int(g(m_\epsilon)-g(m))(m_\epsilon-m)\,dx\to0$ forces the limiting density to match the density of any regular weak solution, so $\bar m$ is unique.
  • The explicit one-dimensional example shows that non-uniqueness arises where the limiting density vanishes, and the selection criterion identifies the minimizer among the possible solutions.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • This selection rule could be used as an a posteriori numerical check: compute any regular weak solution of the limit problem, evaluate $u\mapsto\int\langle u\rangle m\,dx$, and the vanishing-discount limit must be the minimizer.
  • The same holonomy-measure comparison may apply to other singular perturbations—small viscosity, finite horizon, or stochastic noise—provided a discounted holonomy condition with a similar limit exists; the paper does not prove those cases.
  • If the minimizing solution is not unique, the first-order correction $\lambda$ in the refined expansion would be the natural tie-breaker; the paper leaves that question open.
  • The explicit example suggests that non-uniqueness is tied to regions where the limiting density vanishes, so the selection criterion can be interpreted as choosing the solution whose active positive-density region realizes the lowest weighted action.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper studies the vanishing-discount limit of a first-order stationary mean-field game with quadratic Hamiltonian. Under Assumptions 1 and 2 it claims existence and uniqueness of classical solutions of the discounted problem and convergence to a unique classical solution of the ergodic problem (Theorem 1.1, Corollary 1.2), together with refined asymptotics via a linearized problem (Theorem 1.4). Under a different growth range, Assumption 3, it develops a weak-solution framework, constructs discounted and limit Mather measures, and proves a selection criterion (Theorem 1.9): any weak limit of regular weak discounted solutions is the regular weak solution of the limit problem that minimizes the Mather-type functional ∫⟨u⟩m. An explicit one-dimensional example with non-unique weak solutions is used to illustrate the selection rule.

Significance. If the selection theorem is correct, it is a genuinely interesting application of Aubry-Mather ideas to mean-field games and gives a sharp variational selection principle for the vanishing-discount limit. The paper also provides a welcome explicit example of non-uniqueness and a workable notion of regular weak solutions. The classical existence and refined-asymptotics part is also potentially valuable, and the Mather-measure construction in Section 7 is clearly the strongest and most original part of the paper. However, the classical part contains a real gap in the stated parameter range, and one load-bearing uniqueness assertion is not proved; these issues need to be addressed before the paper can be accepted.

major comments (3)
  1. [§3, Proposition 3.4, inequality (3.8)] The proof of the uniform L∞ bound for m uses the chain ∫m^{p+β}|Dm|² ≤ C∫m^{p+1} ≤ C∫m^{p+β+2}, with the last step justified by p+β+2>p+1. This ordering only holds when β+1>0, whereas Assumption 2 explicitly permits every β∈R. For β≤−1 the exponent ordering is reversed; the lower bound m≥m0 from Proposition 3.2 does not repair the estimate because m^{p+1}/m^{p+β+2}=m^{−β−1} is unbounded above when β<−1. In addition, the Moser iteration is initialized at r0=β+1, which is not a norm controlled by Lemma 3.3 when β+1≤0. Consequently Theorem 1.1 and Corollary 1.2 are not proved in the full stated range of Assumption 2. This does not affect the Section 7 selection argument, which is run under Assumption 3, but it is a genuine gap in the classical part of the paper.
  2. [§6, Proposition 6.5] The proof of Proposition 6.5 asserts 'Because the solution to (1.5) is unique' and uses that assertion to pass from subsequential convergence of (vϵ,θϵ) to full convergence. No proof or reference for uniqueness of Problem 3 is given. Uniqueness for the linear first-order system (1.5), including uniqueness of the constant λ under the normalization of u, is not evident and is load-bearing for Theorem 1.4; without it the refined asymptotics are only subsequential. The authors should either prove the uniqueness statement or replace it with a direct argument that does not require it.
  3. [§7, proof of Theorem 1.9, inequality (7.8)] The derivation of (7.8) is summarized as 'proceeding in a similar manner,' but the symmetric argument is not identical to the one leading to (7.7). It requires an estimate on εuε+1/2|D(ηδ*uε)|²+Wε integrated against the arbitrary regular weak solution m, whereas (1.6) is stated only in the sense of distributions. The mollification step introduces ηδ*m in place of m, so an additional argument is needed, for instance using the pointwise identity forced by (1.6)–(1.7) together with uniform integrability, or a different mollification procedure. Since the selection inequality (1.13) rests on (7.8), this step should be spelled out.
minor comments (4)
  1. [§7, equation (7.5)] The display on the left-hand side of (7.5) reads ∫ v·D(ηδ*u)−Lϵ+W−Wϵ dμϵ, but the left-hand side of (7.4) has −v·D(ηδ*u)−Lϵ+W−Wϵ; the minus sign is missing in the display, even though the subsequent computation appears to use it.
  2. [Assumption 2] The second condition in Assumption 2 is printed as g(z)≤C2+zg(z); the way this condition is used in Lemma 3.3 and later suggests that the intended inequality is g(z)≤C2+zg'(z) or a similar growth condition. Please correct or clarify the statement.
  3. [§6, Proposition 6.3] In the integration-by-parts step after equation (6.5) the displayed formula contains the fragment '−hzxi' with an undefined h; it should involve the function ψ_i defined in (6.5).
  4. [§7.4, Proposition 7.9] The proof states strict inequalities comparing ˜u and u on the two halves of the interval, but the preceding bounds (7.12)–(7.13) appear to justify only ≥/≤, with possible equality on parts of the support of m; the strict inequalities should be relaxed or justified.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the selection inequality is derived from Mather-measure comparison, not from its conclusion.

full rationale

The core selection claim (Theorem 1.9) is self-contained: it starts from the defining identities (1.6)-(1.8) for the discounted problem and (1.9)-(1.11) for the limit problem, constructs the phase-space measures μϵ and μ via μϵ = mϵ ⊗ δ_{−Duϵ} in Propositions 7.4 and 7.6, derives the holonomy and action identities (7.2)-(7.3), and then obtains inequalities (7.7) and (7.8) from the Fenchel inequality and Jensen's inequality. Adding them gives (7.9), which yields the monotone-convergence identity (1.12) and the selection inequality (1.13). No parameter is fitted, and no subsequential limit is chosen by fiat; the minimizing property (1.13) is a consequence of the comparison of the two Mather measures, not an assumed input. The cited works [16] and [19] supply, respectively, existence of regular weak solutions for the discounted problem and the Aubry-Mather viewpoint; neither contains the MFG selection statement, and neither is used to replace the derivation of (1.12)-(1.13). The flagged weakness in Proposition 3.4 concerning β ≤ −1 is a possible correctness gap in the classical-solution regularity range, not a circularity, and the weak-solution selection proof in Section 7 is independent of that Moser iteration. Therefore no significant circularity is present.

Assumptions & free parameters 0 free parameters · 8 assumptions · 0 invented entities

No fitted parameters appear; all hypotheses are stated assumptions. The central selection theorem uses prior results by overlapping authors [16] and [19] as tools, not as conclusions of this paper. Two unproved or implicit premises are flagged: uniqueness of the linearized problem and the condition β+1>0 in the Moser step.

assumptions (8)
  • domain assumption V ∈ C1,α(Td), g ∈ C1,α((0,∞)), g strictly increasing
    Standing hypotheses in Problem 1; used throughout for classical solutions.
  • domain assumption Assumption 1: g^{-1}(g(1) - osc V) > 0
    Ensures m has a positive lower bound in Proposition 3.2.
  • domain assumption Assumption 2: g'(z) ≥ C1 z^β and g(z) ≤ C2 + z g(z)
    Used for higher integrability and L∞ bounds of m, hence for classical existence in Theorem 1.1.
  • domain assumption Assumption 3: c1 m^{α-1} ≤ g'(m) ≤ c2 m^{α-1} with the stated range on α
    Used in Section 7 for weak solutions, compactness of mε, and the selection theorem.
  • domain assumption Existence and uniform estimates for weak solutions of Problems 1 and 2 from Ferreira-Gomes [16]
    Theorem 1.7 is quoted from [16]; the selection proof relies on those uniform bounds.
  • domain assumption Mather measure framework and discounted holonomy condition from Gomes [19]
    The selection proof constructs ε-Mather measures and uses the framework introduced in [19].
  • ad hoc to paper Uniqueness of solutions to Problem 3, Eq. (1.5)
    Invoked in Proposition 6.5 to make the limit independent of subsequence; no proof or reference is supplied.
  • ad hoc to paper Implicit condition β+1>0 in Proposition 3.4
    Inequality (3.8) relies on p+β+2>p+1; this is not stated in Assumption 2, which allows any β∈R.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The selection problem for some first-order stationary mean-field games." pith.science (2026). https://pith.science/paper/H2FX2YUF

@misc{pith2026190806485,
  author       = {Pith},
  title        = {Pith review of: The selection problem for some first-order stationary mean-field games},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/H2FX2YUF}},
  note         = {Machine review of arXiv:1908.06485}
}
read the original abstract

Here, we study the existence and the convergence of solutions for the vanishing discount MFG problem with a quadratic Hamiltonian. We give conditions under which the discounted problem has a unique classical solution and prove convergence of the vanishing-discount limit to a unique solution up to constants. Then, we establish refined asymptotics for the limit. When those conditions do not hold, the limit problem may not have a unique solution and its solutions may not be smooth, as we illustrate in an elementary example. Finally, we investigate the stability of regular weak solutions and address the selection problem. Using ideas from Aubry-Mather theory, we establish a selection criterion for the limit.

Figures

Figures reproduced from arXiv: 1908.06485 by the authors.

Figure 1
Figure 1. Density m for (2.2) which exhibits areas with no agents. (a) ux (b) u˜x [PITH_FULL_IMAGE:figures/full_fig_p007_1.png] view at source ↗
Figure 2
Figure 2. Two distinct solutions, u and ˜u, of the Hamilton-Jacobi equation in (2.2). Their gradients differ only when m vanishes. 3. Preliminary estimates In this section, we establish preliminary a priori estimates for solutions of Problem 1. To simplify the notation, we denote by (u, m) a solution of Problem 1, instead of (u  , m ). Here, we seek to establish bounds for (u, m) that are uniform in . Accordingly, the boun… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 34 canonical work pages

  1. [19]

    D. Gomes. Generalized Mather problem and selection principles for viscosity solutions and Mather measures. Adv. Calc. Var., 1(3):291–307, 2008

    work page 2008

  2. [16]

    Ferreira and D

    R. Ferreira and D. Gomes. Existence of weak solutions to stationary mean-field games through varia- tional inequalities. SIAM J. Math. Anal. , 50(6):5969–6006, 2018

    work page 2018

  3. [1]

    Al-Aidarous, E

    E. Al-Aidarous, E. Alzahrani, H. Ishii, and A. Younas. A convergence result for the ergodic problem for Hamilton-Jacobi equations with Neumann-type boundary conditions. Proc. Roy. Soc. Edinburgh Sect. A, 146(2):225–242, 2016

    work page 2016

  4. [2]

    Camilli, I

    F. Camilli, I. Capuzzo-Dolcetta, and D. Gomes. Error estimates for the approximation of the effective Hamiltonian. Appl. Math. Optim. , 57(1):30–57, 2008

    work page 2008

  5. [3]

    Cardaliaguet and P

    P. Cardaliaguet and P. J. Graber. Mean field games systems of first order. ESAIM Control Optim. Calc. Var., 21(3):690–722, 2015

    work page 2015

  6. [4]

    Cardaliaguet and A

    P. Cardaliaguet and A. Porretta. Long time behavior of the master equation in mean field game theory. Anal. PDE, 12(6):1397–1453, 2019

    work page 2019

  7. [5]

    Davini, A

    A. Davini, A. Fathi, R. Iturriaga, and M. Zavidovique. Convergence of the solutions of the discounted Hamilton-Jacobi equation: convergence of the discounted solutions. Invent. Math., 206(1):29–55, 2016

    work page 2016

  8. [6]

    Dieudonn´ e.Foundations of modern analysis

    J. Dieudonn´ e.Foundations of modern analysis. Academic Press, New York-London, 1969. Enlarged and corrected printing, Pure and Applied Mathematics, Vol. 10-I

    work page 1969

Show all 35 references
  1. [7]

    Evangelista, R

    D. Evangelista, R. Ferreira, D. Gomes, L. Nurbekyan, and V. Voskanyan. First-order, stationary mean- field games with congestion. Nonlinear Analysis, 173:37–74, 2018

    2018

  2. [8]

    Evangelista and D

    D. Evangelista and D. Gomes. On the existence of solutions for stationary mean-field games with congestion. Journal of Dynamics and Differential Equations , pages 1–24, 2016

    2016

  3. [9]

    L. C. Evans. Some new PDE methods for weak KAM theory. Calculus of Variations and Partial Differential Equations, 17(2):159–177, 2003

    2003

  4. [10]

    L. C. Evans and D. Gomes. Effective Hamiltonians and averaging for Hamiltonian dynamics. I. Arch. Ration. Mech. Anal., 157(1):1–33, 2001

    2001

  5. [11]

    L. C. Evans and D. Gomes. Effective Hamiltonians and averaging for Hamiltonian dynamics II. Archive for rational mechanics and analysis , 161(4):271–305, 2002. SELECTION FOR MFGS 25

    2002

  6. [12]

    A. Fathi. Solutions KAM faibles conjugu´ ees et barri` eres de Peierls.C. R. Acad. Sci. Paris S´ er. I Math., 325(6):649–652, 1997

    1997

  7. [13]

    A. Fathi. Th´ eor` eme KAM faible et th´ eorie de Mather sur les syst` emes lagrangiens.C. R. Acad. Sci. Paris S´ er. I Math., 324(9):1043–1046, 1997

    1997

  8. [14]

    A. Fathi. Orbite h´ et´ eroclines et ensemble de Peierls.C. R. Acad. Sci. Paris S´ er. I Math., 326:1213–1216, 1998

    1998

  9. [15]

    A. Fathi. Sur la convergence du semi-groupe de Lax-Oleinik. C. R. Acad. Sci. Paris S´ er. I Math. , 327:267–270, 1998

    1998

  10. [17]

    Ferreira, D

    R. Ferreira, D. Gomes, and T. Tada. Existence of weak solutions to first-order stationary mean-field games with Dirichlet conditions. To appear in Proc. Amer. Math. Society , April 2018

    2018

  11. [18]

    Gilbarg and N

    D. Gilbarg and N. S. Trudinger. Elliptic partial differential equations of second order. Classics in Math- ematics. Springer-Verlag, Berlin, 2001. Reprint of the 1998 edition

    2001

  12. [20]

    Gomes and H

    D. Gomes and H. Mitake. Existence for stationary mean-field games with congestion and quadratic Hamiltonians. NoDEA Nonlinear Differential Equations Appl. , 22(6):1897–1910, 2015

    1910

  13. [21]

    Gomes, H

    D. Gomes, H. Mitake, and H. Tran. The selection problem for discounted Hamilton-Jacobi equations: some non-convex cases. J. Math. Soc. Japan , 70(1):345–364, 2018

    2018

  14. [22]

    Gomes, L

    D. Gomes, L. Nurbekyan, and M. Prazeres. Explicit solutions of one-dimensional, first-order, stationary mean-field games with congestion. 2016 IEEE 55th Conference on Decision and Control, CDC 2016 , pages 4534–4539, 2016

    2016

  15. [23]

    Gomes, L

    D. Gomes, L. Nurbekyan, and M. Prazeres. One-dimensional stationary mean-field games with local coupling. Dyn. Games and Applications , 2017

    2017

  16. [24]

    Gomes, S

    D. Gomes, S. Patrizi, and V. Voskanyan. On the existence of classical solutions for stationary extended mean field games. Nonlinear Anal., 99:49–79, 2014

    2014

  17. [25]

    Gomes and H

    D. Gomes and H. S´ anchez Morgado. A stochastic Evans-Aronsson problem. Trans. Amer. Math. Soc., 366(2):903–929, 2014

    2014

  18. [26]

    Ishii, H

    H. Ishii, H. Mitake, and H. V. Tran. The vanishing discount problem and viscosity Mather measures. Part 1: The problem on a torus. J. Math. Pures Appl. (9) , 108(2):125–149, 2017

    2017

  19. [27]

    Ishii, H

    H. Ishii, H. Mitake, and H. V. Tran. The vanishing discount problem and viscosity Mather measures. Part 2: Boundary value problems. J. Math. Pures Appl. (9) , 108(3):261–305, 2017

    2017

  20. [28]

    Lasry and P.-L

    J.-M. Lasry and P.-L. Lions. Jeux ` a champ moyen. I. Le cas stationnaire.C. R. Math. Acad. Sci. Paris , 343(9):619–625, 2006

    2006

  21. [29]

    Lasry and P.-L

    J.-M. Lasry and P.-L. Lions. Mean field games. Jpn. J. Math. , 2(1):229–260, 2007

    2007

  22. [30]

    P. L. Lions. Coll` ege de France course on mean-field games. 2007-2011

    2007

  23. [31]

    P. L. Lions, G. Papanicolao, and S. R. S. Varadhan. Homogeneization of Hamilton-Jacobi equations. Preliminary Version, 1988

    1988

  24. [32]

    Ma˜ n´ e

    R. Ma˜ n´ e. On the minimizing measures of Lagrangian dynamical systems.Nonlinearity, 5(3):623–638, 1992

    1992

  25. [33]

    J. Mather. Action minimizing invariant measure for positive definite Lagrangian systems. Math. Z , 207:169–207, 1991

    1991

  26. [34]

    Mitake and H

    H. Mitake and H. Tran. Selection problems for a discount degenerate viscous Hamilton-Jacobi equation. Adv. Math., 306:684–703, 2017

    2017

  27. [35]

    Pimentel and V

    E. Pimentel and V. Voskanyan. Regularity for second-order stationary mean-field games. Indiana Univ. Math. J., 66(1):1–22, 2017. (D. A. Gomes) King Abdullah University of Science and Technology (KAUST), CEMSE Divi- sion, Thuwal 23955-6900. Saudi Arabia, and KAUST SRI, Center fo...

    2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.