Phase Transitions in Turnpike Theory For Mean-Field Games

Siddharth Karuturi

arxiv: 2605.20213 · v1 · pith:N4P6WKIRnew · submitted 2026-04-24 · 🧮 math.GM

Phase Transitions in Turnpike Theory For Mean-Field Games

Siddharth Karuturi This is my paper

Pith reviewed 2026-05-21 09:16 UTC · model grok-4.3

classification 🧮 math.GM

keywords mean-field gamesturnpike theoryphase transitionspitchfork bifurcationstability analysispropagation of chaostorusinteraction kernel

0 comments

The pith

In mean-field games on the torus a critical interaction strength triggers a pitchfork bifurcation from uniform to nonuniform stationary solutions.

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

The paper examines a translation-invariant mean-field game on the flat torus where the interaction strength γ multiplies a convolution with a smooth even mean-zero kernel. Linear stability analysis around the uniform equilibrium produces a dispersion relation whose sign changes at a finite critical value γ_c, after which the uniform state loses stability. For subcritical γ the uniform state remains exponentially stable in the turnpike sense for finite-horizon problems, while at criticality the decay slows to algebraic order T^{-1/2} and above criticality a family of nonuniform stationary solutions branches off via a pitchfork amplitude equation. The work also shows that standard asymptotic-consistency assumptions on N-player equilibria yield qualitative propagation of chaos below γ_c. These results link linear spectral properties directly to the existence of patterned equilibria and to the duration of turnpike behavior.

Core claim

For the mean-field game with interaction F(x,m) = γ(K ∗ m)(x) on the torus, the linearized operator around the uniform state admits a family of 2 × 2 mode-wise systems whose dispersion is σ_ξ(γ) = ν²(2π|ξ|)⁴ + γ(2π|ξ|)² K̂(ξ). The first negative value of K̂ determines the threshold γ_c; when γ exceeds γ_c a pitchfork-type amplitude equation produces a branch of nonuniform stationary solutions whose translates fill the orbit, while the spectral gap scales as C_* √(γ_c − γ) near criticality. Below γ_c the uniform equilibrium is exponentially stable for finite-horizon problems with rate ρ(γ), at γ_c the midpoint decay is algebraic of order T^{-1/2}, and subcritical N-player equilibria satisfy a

What carries the argument

The dispersion relation σ_ξ(γ) obtained by Fourier-mode linearization of the stationary mean-field game system, which identifies the stability threshold γ_c and supplies the center-manifold reduction that yields the pitchfork amplitude equation.

If this is right

For γ < γ_c the uniform equilibrium is exponentially stable in the turnpike sense, so finite-horizon solutions remain close to uniformity for most of the time interval.
At γ = γ_c the spectral gap vanishes and, after phase fixing and center-manifold reduction, the midpoint value of solutions decays only algebraically as T^{-1/2}.
For γ > γ_c a continuous branch of nonuniform stationary solutions bifurcates from the uniform state; translations of any solution on the branch also solve the system.
Under the stated asymptotic-consistency assumptions, symmetric N-player equilibria converge qualitatively to the mean-field limit in the subcritical regime.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same critical value γ_c may mark a change in long-time behavior for infinite-horizon problems, where nonuniform equilibria could become the relevant attractors.
Numerical schemes that discretize the torus could test the predicted square-root scaling of relaxation rates by tracking the decay of perturbations as γ approaches γ_c from below.
The bifurcation mechanism suggests that similar phase transitions could appear in mean-field games on other compact manifolds once the kernel admits negative Fourier modes.

Load-bearing premise

The kernel K is smooth, even, and mean-zero, and the N-player equilibria obey standard asymptotic-consistency assumptions that allow only qualitative propagation of chaos.

What would settle it

A direct numerical solution of the stationary system for a value of γ slightly larger than the computed γ_c that finds no nonuniform equilibria, or a computed spectral gap that fails to scale as the square root of the distance to γ_c.

read the original abstract

We study a translation-invariant mean-field game on the flat torus with interaction $F(x,m)=\gamma (K*m)(x)$, where $K$ is smooth, even, and mean-zero. The interaction is of potential type, arising as the first variation of a quadratic energy, though the stationary system is not treated variationally. Linearizing around the uniform equilibrium yields mode-wise $2\times 2$ systems with dispersion $\sigma_\xi(\gamma)=\nu^2(2\pi|\xi|)^4+\gamma(2\pi|\xi|)^2\hat K(\xi)$. If $\hat K$ is negative for some mode, a finite threshold \[ \gamma_c=\min_{\hat K(\xi)<0}\frac{\nu^2(2\pi|\xi|)^2}{|\hat K(\xi)|} \] marks loss of stability; otherwise $\gamma_c=+\infty$. Near criticality, the spectral gap scales as $\rho(\gamma)\sim C_*\sqrt{\gamma_c-\gamma}$. For $\gamma<\gamma_c$, the uniform state is exponentially stable in the turnpike sense for finite-horizon problems, with rate $\rho(\gamma)$. At $\gamma=\gamma_c$, the gap closes and, after phase fixing and center-manifold reduction, one obtains algebraic midpoint decay of order $T^{-1/2}$. For $\gamma>\gamma_c$, a branch of nonuniform stationary solutions bifurcates via a pitchfork-type amplitude equation, with translations generating the full family. Finally, under standard asymptotic-consistency assumptions on symmetric $N$-player equilibria in the subcritical regime, we obtain qualitative propagation of chaos, without quantitative rates.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives an explicit gamma_c threshold and claims a pitchfork bifurcation to nonuniform turnpike states in a translation-invariant MFG, but the center-manifold step may need a genericity condition on the critical mode.

read the letter

The main point is that this paper locates a concrete threshold gamma_c where the uniform state loses turnpike stability in this mean-field game, then uses center-manifold reduction to produce a pitchfork branch of nonuniform stationary solutions for stronger interactions, together with the square-root scaling of the spectral gap near criticality. The dispersion relation and the min formula for gamma_c are written out clearly from the linearized 2x2 blocks per Fourier mode. That calculation is straightforward and useful to have for this setting on the torus. The algebraic midpoint decay of order T to the minus one half at criticality also follows directly once the gap closes. These pieces are the actual new content relative to standard MFG turnpike work. The linear stability analysis holds up without obvious gaps. The mode-wise dispersion sigma_xi(gamma) is standard for translation-invariant kernels, and the square-root behavior of rho(gamma) is the usual local picture when a simple eigenvalue crosses zero. The paper states the assumptions on K plainly and keeps the propagation-of-chaos claim qualitative, which matches what the abstract promises. The soft spot sits in the bifurcation reduction. The abstract invokes phase fixing and center-manifold reduction to reach a pitchfork-type amplitude equation, yet supplies no genericity hypothesis that the critical eigenspace is exactly two-dimensional. If several distinct modes or non-collinear xi pairs hit the minimum at the same gamma_c, the kernel dimension exceeds two and the reduced dynamics on the center manifold need not stay a simple pitchfork. The stress-test note flags exactly this point. If the full proofs contain an implicit radial-K assumption or a separate argument ruling out simultaneous criticality, that would close the issue; otherwise it is a minor but real caveat worth adding explicitly. This work is for readers already comfortable with nonlocal PDEs, turnpike problems, and bifurcation methods in mean-field games. A specialist who wants explicit thresholds or wants to extend the analysis to other kernels will find the linear part and the scaling immediately usable. It is worth sending to a serious referee. The claims are specific enough to check, the linear analysis is clean, and the potential gap on genericity is narrow enough that referees can address it directly rather than reject outright. I would recommend peer review with a note to verify or add the condition on the critical mode.

Referee Report

1 major / 1 minor

Summary. The paper analyzes phase transitions in the turnpike property for a translation-invariant mean-field game on the flat torus with interaction F(x,m) = γ(K ∗ m)(x), where K is smooth, even, and mean-zero. Linearization around the uniform state produces mode-wise dispersion relations σ_ξ(γ) = ν²(2π|ξ|)⁴ + γ(2π|ξ|)² K̂(ξ), yielding a critical threshold γ_c as the minimum over modes with K̂(ξ) < 0. For γ < γ_c the uniform equilibrium is exponentially stable in the turnpike sense; at γ = γ_c the spectral gap closes with algebraic midpoint decay of order T^{-1/2} after center-manifold reduction; for γ > γ_c a branch of nonuniform stationary solutions bifurcates via a pitchfork-type amplitude equation, with translations generating the family. Qualitative propagation of chaos is obtained for N-player equilibria under standard asymptotic-consistency assumptions.

Significance. If the central claims hold, the work supplies a concrete link between mean-field game turnpike theory and classical bifurcation analysis, with explicit scaling ρ(γ) ∼ C_* √(γ_c − γ) near criticality and a reduced amplitude equation governing the post-critical branch. The combination of linear spectral analysis, center-manifold reduction, and propagation-of-chaos statements under minimal assumptions on the N-player equilibria constitutes a technically coherent contribution to the study of long-time behavior and pattern formation in potential-type MFGs.

major comments (1)

[Abstract] Abstract (bifurcation paragraph): the claim that a pitchfork-type amplitude equation governs the branch of nonuniform solutions for γ > γ_c presupposes that the critical eigenspace at γ = γ_c is exactly two-dimensional (real and imaginary parts of a single ±ξ pair). No genericity hypothesis is stated ensuring that the minimizing mode ξ is unique up to sign and that no other |ξ'| or non-collinear vectors achieve the same minimal value of ν²(2π|ξ|)² / |K̂(ξ)| simultaneously. When the kernel dimension exceeds two, the center-manifold reduction yields a higher-dimensional system whose normal form need not be a simple pitchfork, undermining the asserted branch structure and the translation-generated family.

minor comments (1)

[Abstract] The dispersion formula is written with (2π|ξ|)^4 and (2π|ξ|)^2 factors; a brief remark on the origin of the fourth-order term (likely from the Laplacian or higher-regularity cost) would aid readers unfamiliar with the precise MFG Hamiltonian.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading and constructive criticism. The observation regarding the bifurcation analysis is well taken, and we will revise the manuscript to incorporate an explicit genericity hypothesis.

read point-by-point responses

Referee: [Abstract] Abstract (bifurcation paragraph): the claim that a pitchfork-type amplitude equation governs the branch of nonuniform solutions for γ > γ_c presupposes that the critical eigenspace at γ = γ_c is exactly two-dimensional (real and imaginary parts of a single ±ξ pair). No genericity hypothesis is stated ensuring that the minimizing mode ξ is unique up to sign and that no other |ξ'| or non-collinear vectors achieve the same minimal value of ν²(2π|ξ|)² / |K̂(ξ)| simultaneously. When the kernel dimension exceeds two, the center-manifold reduction yields a higher-dimensional system whose normal form need not be a simple pitchfork, undermining the asserted branch structure and the translation-generated family.

Authors: We agree that an explicit genericity assumption is required to guarantee that the critical eigenspace is precisely two-dimensional. In the revised version we will add the standing hypothesis that the minimum defining γ_c is attained at a unique pair ±ξ (modulo the natural action of the torus) and that this value is strictly smaller than the corresponding quantity for every other mode. Under this condition the linearized operator at criticality has a two-dimensional kernel, the center-manifold reduction reduces to a scalar pitchfork amplitude equation, and the resulting branch is parameterized by translations, exactly as claimed in the abstract. The same hypothesis will be stated clearly in the introduction and in the bifurcation section, together with a brief remark that it holds for generic smooth, even, mean-zero kernels K. revision: yes

Circularity Check

0 steps flagged

No significant circularity; standard linearization and bifurcation analysis

full rationale

The paper derives its results from explicit linearization of the stationary mean-field game system around the uniform state, producing the mode-wise dispersion relation σ_ξ(γ) and the explicit threshold γ_c as the minimum of that relation over modes with negative Fourier coefficients. The subsequent claims of pitchfork bifurcation for γ > γ_c, spectral gap scaling, and algebraic decay at criticality follow from classical center-manifold reduction and phase fixing applied to this linear operator; these steps invoke standard external theorems rather than any self-definition, fitted-parameter renaming, or load-bearing self-citation. The propagation-of-chaos statement rests on stated asymptotic-consistency assumptions that are independent of the bifurcation analysis. No step reduces the claimed output to an input by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claims rest on standard PDE assumptions for mean-field games on the torus together with smoothness and symmetry of the kernel; no new free parameters are fitted and no new entities are postulated.

axioms (2)

domain assumption The domain is the flat torus with translation invariance
Used to justify Fourier-mode analysis and the form of the dispersion relation.
domain assumption K is smooth, even, and mean-zero
Required for the interaction to be of potential type and for the Fourier multiplier to be well-defined.

pith-pipeline@v0.9.0 · 5833 in / 1319 out tokens · 32516 ms · 2026-05-21T09:16:03.849713+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

18 extracted references · 18 canonical work pages

[1]

Carr,Applications of Centre Manifold Theory, Springer, New York, 1981

J. Carr,Applications of Centre Manifold Theory, Springer, New York, 1981

work page 1981
[2]

Cardaliaguet and A

P. Cardaliaguet and A. Porretta, Long time behavior of the master equation in mean field game theory,Anal. PDE12(2019), 1397–1453

work page 2019
[3]

Cardaliaguet, F

P. Cardaliaguet, F. Delarue, J.-M. Lasry, and P.-L. Lions,The Master Equation and the Convergence Problem in Mean Field Games, Annals of Mathematics Studies 201, Princeton University Press, 2019

work page 2019
[4]

Cecchin and F

A. Cecchin and F. Delarue, Weak solutions to the master equation of potential mean field games,Ann. Appl. Probab.32(2022), 1441–1512

work page 2022
[5]

J. A. Carrillo, K. Craig, F. S. Patacchini, and G. A. Shpak, Aggregation-diffusion equations: dynamics, asymptotics, and singular limits, inActive Particles, Vol. 2, Birkh¨ auser, 2020

work page 2020
[6]

M. G. Crandall and P. H. Rabinowitz, Bifurcation from simple eigenvalues,J. Funct. Anal.8(1971), 321–340

work page 1971
[7]

Dorfman, P

R. Dorfman, P. A. Samuelson, and R. M. Solow,Linear Programming and Economic Analysis, McGraw-Hill, New York, 1958

work page 1958
[8]

D. A. Dawson and J. G¨ artner, Large deviations from the McKean–Vlasov limit for weakly interacting diffusions,Stochastics20(1987), 247–308

work page 1987
[9]

Fournier and A

N. Fournier and A. Guillin, On the rate of convergence in Wasserstein distance of the empirical measure,Probab. Theory Relat. Fields162(2015), 707–738

work page 2015
[10]

Huang, R

M. Huang, R. P. Malham´ e, and P. E. Caines, Large population stochastic dynamic games: closed-loop McKean–Vlasov systems and the Nash certainty equivalence principle,Commun. Inf. Syst.6(2006), 221–252

work page 2006
[11]

Kielh¨ ofer,Bifurcation Theory: An Introduction with Applications to PDEs, Springer, New York, 2004

H. Kielh¨ ofer,Bifurcation Theory: An Introduction with Applications to PDEs, Springer, New York, 2004. PHASE TRANSITIONS IN TURNPIKE THEORY FOR MFG 19

work page 2004
[12]

Lacker, Quantitative approximate independence for continuous mean field games, Ann

D. Lacker, Quantitative approximate independence for continuous mean field games, Ann. Appl. Probab.33(2023), 3173–3245

work page 2023
[13]

Lasry and P.-L

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

work page 2007
[14]

Porretta and E

A. Porretta and E. Zuazua, Long time versus steady state optimal control,SIAM J. Control Optim.51(2013), 4242–4273

work page 2013
[15]

Sznitman, Topics in propagation of chaos, in ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XIX—1989, Lecture Notes in Mathematics1464, Springer, 1991, pp

A.-S. Sznitman, Topics in propagation of chaos, in ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XIX—1989, Lecture Notes in Mathematics1464, Springer, 1991, pp. 165– 251

work page 1989
[16]

A. M. Turing, The chemical basis of morphogenesis,Philos. Trans. R. Soc. London Ser. B237(1952), 37–72

work page 1952
[17]

Tr´ elat, Linear turnpike theorem,Math

E. Tr´ elat, Linear turnpike theorem,Math. Control Signals Systems35(2023), 685–728

work page 2023
[18]

Tr´ elat and E

E. Tr´ elat and E. Zuazua, The turnpike property in finite-dimensional nonlinear optimal control,J. Differential Equations258(2015), 81–114. [Illinois Mathematics and Science Academy] Email address:[skaruturi@imsa.edu]

work page 2015

[1] [1]

Carr,Applications of Centre Manifold Theory, Springer, New York, 1981

J. Carr,Applications of Centre Manifold Theory, Springer, New York, 1981

work page 1981

[2] [2]

Cardaliaguet and A

P. Cardaliaguet and A. Porretta, Long time behavior of the master equation in mean field game theory,Anal. PDE12(2019), 1397–1453

work page 2019

[3] [3]

Cardaliaguet, F

P. Cardaliaguet, F. Delarue, J.-M. Lasry, and P.-L. Lions,The Master Equation and the Convergence Problem in Mean Field Games, Annals of Mathematics Studies 201, Princeton University Press, 2019

work page 2019

[4] [4]

Cecchin and F

A. Cecchin and F. Delarue, Weak solutions to the master equation of potential mean field games,Ann. Appl. Probab.32(2022), 1441–1512

work page 2022

[5] [5]

J. A. Carrillo, K. Craig, F. S. Patacchini, and G. A. Shpak, Aggregation-diffusion equations: dynamics, asymptotics, and singular limits, inActive Particles, Vol. 2, Birkh¨ auser, 2020

work page 2020

[6] [6]

M. G. Crandall and P. H. Rabinowitz, Bifurcation from simple eigenvalues,J. Funct. Anal.8(1971), 321–340

work page 1971

[7] [7]

Dorfman, P

R. Dorfman, P. A. Samuelson, and R. M. Solow,Linear Programming and Economic Analysis, McGraw-Hill, New York, 1958

work page 1958

[8] [8]

D. A. Dawson and J. G¨ artner, Large deviations from the McKean–Vlasov limit for weakly interacting diffusions,Stochastics20(1987), 247–308

work page 1987

[9] [9]

Fournier and A

N. Fournier and A. Guillin, On the rate of convergence in Wasserstein distance of the empirical measure,Probab. Theory Relat. Fields162(2015), 707–738

work page 2015

[10] [10]

Huang, R

M. Huang, R. P. Malham´ e, and P. E. Caines, Large population stochastic dynamic games: closed-loop McKean–Vlasov systems and the Nash certainty equivalence principle,Commun. Inf. Syst.6(2006), 221–252

work page 2006

[11] [11]

Kielh¨ ofer,Bifurcation Theory: An Introduction with Applications to PDEs, Springer, New York, 2004

H. Kielh¨ ofer,Bifurcation Theory: An Introduction with Applications to PDEs, Springer, New York, 2004. PHASE TRANSITIONS IN TURNPIKE THEORY FOR MFG 19

work page 2004

[12] [12]

Lacker, Quantitative approximate independence for continuous mean field games, Ann

D. Lacker, Quantitative approximate independence for continuous mean field games, Ann. Appl. Probab.33(2023), 3173–3245

work page 2023

[13] [13]

Lasry and P.-L

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

work page 2007

[14] [14]

Porretta and E

A. Porretta and E. Zuazua, Long time versus steady state optimal control,SIAM J. Control Optim.51(2013), 4242–4273

work page 2013

[15] [15]

Sznitman, Topics in propagation of chaos, in ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XIX—1989, Lecture Notes in Mathematics1464, Springer, 1991, pp

A.-S. Sznitman, Topics in propagation of chaos, in ´Ecole d’ ´Et´ e de Probabilit´ es de Saint-Flour XIX—1989, Lecture Notes in Mathematics1464, Springer, 1991, pp. 165– 251

work page 1989

[16] [16]

A. M. Turing, The chemical basis of morphogenesis,Philos. Trans. R. Soc. London Ser. B237(1952), 37–72

work page 1952

[17] [17]

Tr´ elat, Linear turnpike theorem,Math

E. Tr´ elat, Linear turnpike theorem,Math. Control Signals Systems35(2023), 685–728

work page 2023

[18] [18]

Tr´ elat and E

E. Tr´ elat and E. Zuazua, The turnpike property in finite-dimensional nonlinear optimal control,J. Differential Equations258(2015), 81–114. [Illinois Mathematics and Science Academy] Email address:[skaruturi@imsa.edu]

work page 2015