pith. sign in

arxiv: 2604.12061 · v1 · submitted 2026-04-13 · 🧮 math.OC

Integral equations for the optimal boundary surface of a mean-field game of capacity expansion

Tiziano De Angelis , Giulia Livieri , Maddalena Ghio This is my paper

Pith reviewed 2026-05-10 15:02 UTC · model grok-4.3

classification 🧮 math.OC
keywords mean-field gamescapacity expansionoptimal boundary surfaceVolterra integral equationfree boundary problemsstochastic controlItô formula extension
0
0 comments X

The pith

The optimal boundary surface in a mean-field game of capacity expansion is the unique continuous solution of a nonlinear Volterra integral equation.

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

The paper proves that the surface dividing action and inaction regions for many agents expanding capacity under mean-field interaction is uniquely characterized as the continuous solution to a nonlinear integral equation of Volterra type. This characterization is obtained by first showing the surface is continuous and then deriving it via an extended Itô formula that relaxes standard smoothness requirements on the value function's time and space derivatives. A reader would care because the result converts a high-dimensional stochastic game into a deterministic equation that supports numerical solution and explicit control computation, directly building on an earlier model of capacity expansion.

Core claim

We prove that the optimal boundary surface that splits the action and inaction regions in a mean-field game of capacity expansion studied in Campi et al. (2022) is the unique continuous solution of a nonlinear integral equation of Volterra type. In order to do that, we first establish continuity of the optimal surface. Then we develop an extension of Itô's formula which weakens assumptions required in the existing literature on the first-order time-derivative and/or second-order space derivative of the value function. The paper also provides an algorithm for the numerical solution of the integral equation and we compute optimal controls numerically for the mean-field game.

What carries the argument

The nonlinear Volterra integral equation for the optimal boundary surface, obtained after proving continuity and applying an extended Itô formula with relaxed regularity conditions on the value function.

If this is right

  • The optimal boundary surface can be computed by numerically solving the integral equation.
  • Optimal controls for the agents follow directly once the surface is known.
  • The extended Itô formula applies under weaker derivative assumptions than standard versions.
  • Numerical examples confirm that the equation yields explicit strategies for the capacity expansion game.

Where Pith is reading between the lines

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

  • The Volterra-equation reduction may extend to other mean-field games that feature free-boundary structures in stochastic control.
  • Algorithms developed for this integral equation could be reused for similar optimal-boundary problems arising in competitive resource allocation.
  • The continuity proof indicates that the boundary remains regular despite the mean-field coupling.

Load-bearing premise

The value function satisfies the regularity conditions needed for the extended Itô formula, and the model parameters and dynamics from the cited prior work hold without contradiction.

What would settle it

A numerical simulation of the mean-field game that produces an optimal boundary surface differing from the unique continuous solution of the proposed Volterra equation would falsify the claim.

Figures

Figures reproduced from arXiv: 2604.12061 by Giulia Livieri, Maddalena Ghio, Tiziano De Angelis.

Figure 1
Figure 1. Figure 1: Left panel: The figure displays the boundary b (0) 0 (t, y) (in red) and the boundary b (5) 0 (t, y) (in blue). Right panel: Optimal surface at game iteration n = 0. Similarly, [PITH_FULL_IMAGE:figures/full_fig_p031_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Left panel: The figure displays the boundary b (5) 5 (t, y). Right panel: The figure displays the generalized inverse c (5) 5 (t, x) [PITH_FULL_IMAGE:figures/full_fig_p032_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The figure displays m[n] (t) for game iteration n ∈ {0, . . . , 5} [PITH_FULL_IMAGE:figures/full_fig_p032_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: The figure displays the numerical convergence of the boundaries b (k) 0 (t, y) for k ∈ {1, . . . , 5}, where the numerical error at step k is defined as the 2-norm ∥b (k) 0 −b (k−1) 0 ∥2 [PITH_FULL_IMAGE:figures/full_fig_p033_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: The figure displays the numerical convergence of the boundaries b (k) 5 (t, y) for k ∈ {1, . . . , 5}, where the numerical error at step k is defined as the 2-norm ∥b (k) 5 −b (k−1) 5 ∥2 [PITH_FULL_IMAGE:figures/full_fig_p033_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: The figure displays, for each n ∈ {1, . . . 5}, the convergence of ∥b (5) n − b (k) n ∥2 for k ∈ {1, . . . 5} [PITH_FULL_IMAGE:figures/full_fig_p034_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: The figure displays the distance in 2-norm between the optimal boundaries of subsequent game iterations, i.e. ∥b (5) n − b (5) n−1 ∥2, n ∈ {1, . . . , 5} [PITH_FULL_IMAGE:figures/full_fig_p034_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Left panel: The sample path X⋆ t . Center panel: The reflected state Y ⋆ t together with the moving target c ⋆ (t, X⋆ t ). Right panel: The cumulative control ξ ⋆ t and the distance from the base capacity Y ⋆ t − c ⋆ (t, X⋆ t ) [PITH_FULL_IMAGE:figures/full_fig_p035_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Left panel: The figure shows R(ti , yj ) = |Φ [PITH_FULL_IMAGE:figures/full_fig_p035_9.png] view at source ↗
read the original abstract

We prove that the optimal boundary surface that splits the action and inaction regions in a mean-field game of capacity expansion studied in (Campi et al.,\ Ann.\ Appl.\ Probab.,\ {\bf 32}(5),\, pp.\,3674-3717, 2022) is the unique continuous solution of a nonlinear integral equation of Volterra type. In order to do that, we first establish continuity of the optimal surface. Then we develop an extension of It\^o's formula which weakens assumptions required in the existing literature on the first-order time-derivative and/or second-order space derivative of the value function. The paper also provides an algorithm for the numerical solution of the integral equation and we compute optimal controls numerically for the mean-field game.

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.

Referee Report

2 major / 2 minor

Summary. The manuscript proves that the optimal boundary surface separating the action and inaction regions in the mean-field game of capacity expansion from Campi et al. (2022) is the unique continuous solution of a nonlinear Volterra-type integral equation. The argument first establishes continuity of the boundary surface, then applies a newly developed extension of Itô's formula (with relaxed assumptions on the first-order time derivative and/or second-order space derivatives of the value function) to derive the integral equation from the HJB equation, proves uniqueness of the continuous solution, and concludes with a numerical algorithm for solving the equation together with computed examples of optimal controls.

Significance. If the central claims hold, particularly the validity and applicability of the extended Itô formula to the MFG value function, the work supplies a useful integral characterization of the free boundary in a singular-control mean-field game. This could support further theoretical study and numerical methods for capacity-expansion problems with mean-field interactions. The inclusion of a numerical algorithm and explicit computations is a positive feature that enhances practical relevance.

major comments (2)
  1. [Section 3, Theorem 3.1] Section 3, Theorem 3.1 (extended Itô formula): The theorem weakens the classical C^{1,2} requirements, but the subsequent application in Section 4 to pass from the HJB equation to the Volterra integral equation does not contain an explicit a-priori verification that the MFG value function satisfies the precise integrability/continuity conditions stated for the formula. The argument instead invokes the continuity result of Section 2 together with regularity inherited from Campi et al. (2022); if those weakened conditions fail to hold, the derivation of the integral equation contains a gap.
  2. [Section 4] Section 4 (derivation of the integral equation): The passage to the Volterra representation relies on the extended Itô formula being applicable once continuity of the boundary is known. Because the paper provides no separate regularity lemma confirming that the value function meets the exact (weakened) hypotheses rather than stronger classical ones, the load-bearing step from the HJB equation to the integral equation is not fully closed.
minor comments (2)
  1. [Introduction] The model setup in the introduction would benefit from a brief self-contained recap of the exact dynamics and mean-field interaction term taken from Campi et al. (2022), rather than relying solely on the citation.
  2. [Section 6] Section 6 (numerical algorithm): The description of the discretization and solution procedure for the Volterra equation lacks pseudocode or explicit convergence/error bounds, which would improve reproducibility of the reported optimal-control computations.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the detailed and constructive report. The two major comments both concern the need for an explicit verification that the mean-field game value function satisfies the hypotheses of the extended Itô formula before it is applied in Section 4. We address each point below and will revise the manuscript accordingly.

read point-by-point responses

  1. Referee: [Section 3, Theorem 3.1] Section 3, Theorem 3.1 (extended Itô formula): The theorem weakens the classical C^{1,2} requirements, but the subsequent application in Section 4 to pass from the HJB equation to the Volterra integral equation does not contain an explicit a-priori verification that the MFG value function satisfies the precise integrability/continuity conditions stated for the formula. The argument instead invokes the continuity result of Section 2 together with regularity inherited from Campi et al. (2022); if those weakened conditions fail to hold, the derivation of the integral equation contains a gap.

    Authors: We agree that an explicit verification step would remove any ambiguity. The continuity of the free boundary proved in Section 2, together with the C^{1,2} regularity away from the boundary and the integrability properties established in Campi et al. (2022), imply that the value function satisfies the precise (weakened) hypotheses of Theorem 3.1. Nevertheless, to make this transparent, the revised manuscript will include a short lemma immediately preceding the derivation in Section 4 that confirms the required integrability of the time derivative and the continuity of the spatial derivatives on the relevant domains. revision: yes

  2. Referee: [Section 4] Section 4 (derivation of the integral equation): The passage to the Volterra representation relies on the extended Itô formula being applicable once continuity of the boundary is known. Because the paper provides no separate regularity lemma confirming that the value function meets the exact (weakened) hypotheses rather than stronger classical ones, the load-bearing step from the HJB equation to the integral equation is not fully closed.

    Authors: We accept the referee's observation. The derivation in Section 4 will be reorganized to begin with the new regularity lemma (as noted in the response to the first comment) that directly checks the hypotheses of Theorem 3.1 against the properties inherited from Campi et al. (2022) and our continuity result. This will close the logical step from the HJB equation to the Volterra integral equation without relying on implicit appeal to stronger classical regularity. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation proceeds from continuity proof to new extended Itô formula to integral equation without reduction to inputs or self-citations.

full rationale

The paper first proves continuity of the optimal boundary surface independently, then introduces an original extension of Itô's formula with weakened derivative assumptions (developed in this work, not imported via self-citation), applies it to obtain the Volterra integral equation representation from the HJB equation, and separately establishes uniqueness. The underlying MFG dynamics are cited to Campi et al. (2022), an external reference with no author overlap indicated, and no fitted parameters or self-referential definitions are used. The regularity conditions for the extended formula are asserted as part of the proof setup rather than derived from the target result itself. This chain is self-contained and does not reduce any claimed prediction or uniqueness statement to its own inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on model assumptions inherited from the 2022 Campi et al. paper and on the validity of the extended Itô formula under weakened regularity; no new free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption The value function admits an extension of Itô's formula with weakened assumptions on time and space derivatives
    Invoked to derive the integral equation from the dynamic programming principle.
  • domain assumption The mean-field game model and dynamics from Campi et al. (2022) are well-posed
    The optimal boundary is defined within that model.

pith-pipeline@v0.9.0 · 5431 in / 1396 out tokens · 47017 ms · 2026-05-10T15:02:19.406280+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

  1. [1]

    , (2017)

    Baldi, P. , (2017). Stochastic Calculus: An introduction through theory and exercises. Universitext. Springer, Cham

    work page 2017

  2. [2]

    , (2023)

    Cai, C., De Angelis, T. , (2023). A change of variable formula with applications to multi-dimensional optimal stopping problems. Stoch.\ Process.\ Appl., 164, 33--61

    work page 2023

  3. [3]

    Cai, C., De Angelis, T., Palczewski, J. (2023). On the continuity of optimal stopping surfaces for jump diffusions. SIAM J.\ Control Optim. 61(3), 1513-1531

    work page 2023

  4. [4]

    , (2022)

    Campi, L., De Angelis, T., Ghio, M., Livieri, G. , (2022). Mean-field games of finite-fuel capacity expansion with singular controls. Ann. Appl. Probab., 32(5), 3674--3717

    work page 2022

  5. [5]

    , (2020)

    Campi, L., De Angelis, T., Ghio, M., Livieri, G. , (2020). Mean-field games of finite-fuel capacity expansion with singular controls. arXiv preprint arXiv:2006.02074

    work page arXiv 2020

  6. [6]

    Cao, H., Guo, X., Lee, J. S. , (2023). Approximation of N -player stochastic games with singular controls by mean field games. Numer. Algebra Control Optim., 3(4), 604--629

    work page 2023

  7. [7]

    , (2022)

    Cao, H., Guo, X. , (2022). MFGs for partially reversible investment. Stochastic Process. Appl., 150, 995--1014

    work page 2022

  8. [8]

    , (2017)

    Cardaliaguet, P., Hadikhanloo, S. , (2017). Learning in mean field games: the fictitious play. ESAIM Control Optim. Calc. Var., 23(2), 569--591

    work page 2017

  9. [9]

    , (2015)

    De Angelis, T. , (2015). A note on the continuity of free-boundaries in finite-horizon optimal stopping problems for one-dimensional diffusions. SIAM J. Control Optim., 53(1), 167-184

    work page 2015

  10. [10]

    , (2017)

    De Angelis, T., Federico, S., Ferrari, G. , (2017). Optimal boundary surface for irreversible investment with stochastic costs. Math.\ Oper.\ Res. , 42(4), 1135-1161

    work page 2017

  11. [11]

    , (2025)

    De Angelis, T., Gensbittel, F., Villeneuve, S. , (2025). Nash equilibria for dividend distribution with competition. Math.\ Oper.\ Res. DOI:10.1287/moor.2023.0374

    work page doi:10.1287/moor.2023.0374 2025

  12. [12]

    , (2025)

    Denkert, R., Horst, U. , (2025). Extended mean-field games with multidimensional singular controls and nonlinear jump impact. SIAM J.\ Control Optim. 63(2), 1374--1406

    work page 2025

  13. [13]

    , (2023)

    Dianetti, J., Ferrari, G., Fischer, M., Nendel, M. , (2023). A unifying framework for submodular mean field games. Math.\ Oper.\ Res. , 48(3), 1679--1710

    work page 2023

  14. [14]

    Preprint, available at arXiv:2509.18821 , year=

    Dianetti, J., Dumitrescu, R., Ferrari, G., Xu, R. , (2025). Entropy regularization in mean-field games of optimal stopping. arXiv:2509.18821

    work page arXiv 2025

  15. [15]

    , (2007)

    Elworthy, K.D., Truman, A., Zhao, H. , (2007). Generalized It\^o formulae and space-time Lebesgue–Stieltjes integrals of local times. Sém.\ Probab. \ XL, 117-136. Berlin, Heidelberg: Springer Berlin Heidelberg

    work page 2007

  16. [16]

    , (2023)

    Fu, G. , (2023). Extended mean field games with singular controls. SIAM J. Control Optim., 61(1), 285--314

    work page 2023

  17. [17]

    , (2017)

    Fu, G., Horst, U. , (2017). Mean field games with singular controls. SIAM J. Control Optim., 55, 3833–3868

    work page 2017

  18. [18]

    , (2012)

    Gu \'e ant, O. , (2012). Mean field games equations with quadratic Hamiltonian: a specific approach. Math. Models Methods Appl. Sci., 22(09), 1250022

    work page 2012

  19. [19]

    , (2021)

    Lauriere, M. , (2021). Numerical methods for mean field games and mean field type control. Proc. Sympos. Appl. Math., 78, 221-282

    work page 2021

  20. [20]

    Karatzas, I., Shreve, S. E. , (1988). Brownian motion and stochastic calculus. Graduate Texts in Mathematics, 113. Springer, New York

    work page 1988

  21. [21]

    , (1969)

    Kruse, R., Deely, J. , (1969). Joint continuity of monotonic functions. Amer. Math. Monthly, 76(1), 74--76

    work page 1969

  22. [22]

    , (1980)

    Krylov, N.V. , (1980). Controlled diffusion processes. New York, NY: Springer New York

    work page 1980

  23. [23]

    , (2025)

    Kwon, H.D., Palczewski, J. , (2025). Exit game with private information. Math.\ Oper.\ Res., 50(4), 2433--2469

    work page 2025

  24. [24]

    , (2005)

    Peskir, G. , (2005). On the American option problem. Math. Financ., 15(1), 169–181

    work page 2005

  25. [25]

    , (2005)

    Peskir, G. , (2005). A change-of-variable formula with local time on curves. J. Theor. Probab., 18(3), 499-535

    work page 2005

  26. [26]

    , (2007)

    Peskir, G. , (2007). A change-of-variable formula with local time on surfaces. Sém.\ Probab. \ XL, pp. 70-96, Berlin, Heidelberg: Springer Berlin Heidelberg

    work page 2007

  27. [27]

    Shiryaev , (2006)

    Peskir, G., A. Shiryaev , (2006). Optimal stopping and free-boundary problems. Birkh\"auser: Basel

    work page 2006

  28. [28]

    , (1990)

    Protter, P. , (1990). Stochastic integration and differential equations. Applications of Mathematics (New York) . 21. Springer, Berlin

    work page 1990