pith. sign in

arxiv: 2607.00655 · v1 · pith:TTJIU4Q5new · submitted 2026-07-01 · 🧮 math.AP

The Schr\"odinger problem on metric graphs

Juliane Krautz , Jan-F. Pietschman This is my paper

Pith reviewed 2026-07-02 09:53 UTC · model grok-4.3

classification 🧮 math.AP
keywords Schrödinger problemmetric graphsentropic optimal transportWasserstein distanceGamma-convergenceBenamou-Brenier formulationdynamic formulation
0
0 comments X

The pith

On metric graphs the dynamic Schrödinger problem minima converge to the squared Wasserstein distance with minimizers approaching geodesics.

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

The paper starts from a static Schrödinger problem on metric graphs, recasts it as entropic optimal transport, and proves Gamma-convergence to ordinary optimal transport. It then derives an equivalent dynamic Benamou-Brenier formulation that extends earlier results known for RCD spaces. With the two versions shown to be equivalent, the authors conclude that the dynamic problem's minimum values converge to the squared Wasserstein distance and its minimizers converge to Wasserstein geodesics. They also prove existence for more general boundary data and illustrate the results numerically. A reader would care because the work supplies a regularized route to optimal transport quantities on graph structures that model networks and discrete geometries.

Core claim

Starting from a static version, we introduce an equivalent reformulation as entropic optimal transport and show Gamma-convergence towards static optimal transport. We then rigorously derive a Benamou-Brenier type dynamic version of the Schrödinger problem, thereby extending known results from RCD*(K,N)-spaces. With this equivalence at hand, we conclude that the minimum values of the dynamic Schrödinger problem converge towards the squared Wasserstein distance, and minimizers converge to Wasserstein geodesics. We also extend the dynamic formulation to a more general class of initial and final data and show existence of solutions in this setting using the direct method.

What carries the argument

The equivalence between the static Schrödinger problem (reformulated as entropic optimal transport) and its dynamic Benamou-Brenier version, which transfers convergence results from RCD spaces to metric graphs.

If this is right

  • The minimum values of the dynamic Schrödinger problem converge to the squared Wasserstein distance.
  • Minimizers of the dynamic problem converge to Wasserstein geodesics.
  • Existence of solutions holds for a broader class of initial and final data via the direct method.
  • The static-to-dynamic equivalence extends known results from RCD*(K,N)-spaces to the graph setting.

Where Pith is reading between the lines

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

  • The same static-to-dynamic route could be tested on other length spaces that are not RCD.
  • Numerical schemes developed for graphs might be adapted to approximate Wasserstein distances on real network data.
  • Existence proofs for general data open the possibility of studying time-dependent or evolving graphs within the same framework.

Load-bearing premise

Metric graphs possess enough length-space structure for the dynamic Benamou-Brenier formulation and the convergence results from RCD spaces to carry over directly.

What would settle it

A concrete metric graph on which the minima of the dynamic Schrödinger problem fail to approach the squared Wasserstein distance as the regularization parameter tends to zero.

Figures

Figures reproduced from arXiv: 2607.00655 by Jan-F. Pietschman, Juliane Krautz.

Figure 1.1
Figure 1.1. Figure 1.1: Relation between optimal transport and Schr¨odinger problems on an RCD [PITH_FULL_IMAGE:figures/full_fig_p003_1_1.png] view at source ↗
Figure 5.1
Figure 5.1. Figure 5.1: Star-shaped graph with three edges roots of the gradient. We distinguish two cases. First, assume that ρ u > 0. Then, the optimality conditions read   −λ |j v | 2+|g v | 2 2ρv + ρ v − ρ u λ j v ρv + j v − j u λ g v ρv + g v − g u   = 0. Similar to [PS22, Appendix A], direct calculations verify that this system is solved by ρ v = ρ ∗ , j v = ρ v j u ρv+λ and g v = ρ v g u ρv+λ if ρ ∗ is the largest … view at source ↗
Figure 5.2
Figure 5.2. Figure 5.2: Comparison between numerical results for [PITH_FULL_IMAGE:figures/full_fig_p029_5_2.png] view at source ↗
Figure 5.3
Figure 5.3. Figure 5.3: Comparison between numerical results in the setting of Example 5.2 for [PITH_FULL_IMAGE:figures/full_fig_p029_5_3.png] view at source ↗
read the original abstract

We study the Schr\"odinger problem on metric graphs and its different formulations. Starting from a static version, we introduce an equivalent reformulation as entropic optimal transport and show $\Gamma$-convergence towards static optimal transport. We then rigorously derive a Benamou-Brenier type dynamic version of the Schr\"odinger problem, thereby extending known results from ${\rm RCD}^*(K,N)$-spaces. With this equivalence at hand, we conclude that the minimum values of the dynamic Schr\"odinger problem converge towards the squared Wasserstein distance, and minimizers converge to Wasserstein geodesics. We also extend the dynamic formulation to a more general class of initial and final data and show existence of solutions in this setting using the direct method. Lastly, we illustrate our analytical findings by a numerical investigation.

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

1 major / 2 minor

Summary. The paper studies the Schrödinger problem on metric graphs. It begins with a static formulation shown equivalent to entropic optimal transport, establishes Γ-convergence to static optimal transport, derives a Benamou-Brenier dynamic formulation by extending results from RCD*(K,N)-spaces, concludes that dynamic Schrödinger minima converge to the squared Wasserstein distance with minimizers converging to Wasserstein geodesics, extends the dynamic formulation to general initial/final data with existence via the direct method, and includes numerical illustrations.

Significance. If the extension of the dynamic formulation holds, the work provides a rigorous bridge between the Schrödinger problem and Wasserstein geometry on metric graphs, which model networks and branched structures. Strengths include the outlined use of Γ-convergence, dynamic equivalence, and convergence to Wasserstein objects, plus the direct-method existence proof and numerical component. This could support further developments in optimal transport and PDE analysis on singular length spaces.

major comments (1)
  1. [Section deriving the dynamic formulation from the static one] The derivation of the Benamou-Brenier dynamic version (invoked to obtain the convergence of minima to W₂² and minimizers to geodesics) extends results known for RCD*(K,N)-spaces, but the manuscript provides no explicit verification that metric graphs satisfy the lower Ricci bound, doubling property, or heat-kernel estimates at vertices of degree ≠2. This is load-bearing for the dynamic-to-static equivalence and the limit passage.
minor comments (2)
  1. [Introduction and dynamic formulation section] The precise class of initial and final data for which the dynamic formulation is extended should be stated explicitly when first introduced, rather than only in the abstract.
  2. [Throughout] Notation for the entropic regularization parameter and the graph metric could be unified across static and dynamic sections to improve readability.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful reading of the manuscript and for highlighting the need for explicit verification in the derivation of the dynamic formulation. We address the major comment below.

read point-by-point responses

  1. Referee: [Section deriving the dynamic formulation from the static one] The derivation of the Benamou-Brenier dynamic version (invoked to obtain the convergence of minima to W₂² and minimizers to geodesics) extends results known for RCD*(K,N)-spaces, but the manuscript provides no explicit verification that metric graphs satisfy the lower Ricci bound, doubling property, or heat-kernel estimates at vertices of degree ≠2. This is load-bearing for the dynamic-to-static equivalence and the limit passage.

    Authors: We agree that the manuscript lacks an explicit verification of the conditions under which the Benamou-Brenier formulation extends from RCD*(K,N)-spaces to metric graphs. In the revised version we will insert a short subsection (or paragraph within the relevant section) that directly verifies the doubling property and heat-kernel estimates on metric graphs, including at vertices of degree ≠2, via explicit computation of the distance measure and the heat kernel on the graph. For the lower Ricci bound we will note that the derivation adapts the RCD* arguments by local analysis at vertices and does not require a uniform positive lower bound; we will make this adaptation explicit rather than relying solely on the RCD* literature. revision: yes

Circularity Check

0 steps flagged

No circularity; central claims rest on external RCD*(K,N) extensions

full rationale

The paper starts from a static Schrödinger problem, reformulates it as entropic OT, shows Gamma-convergence to static OT, then extends the dynamic Benamou-Brenier formulation from known RCD*(K,N) results to metric graphs. The convergence of dynamic minima to W_2^2 and minimizers to geodesics follows from that equivalence. All load-bearing steps invoke external theorems on RCD spaces and standard OT; no equation reduces a derived quantity to a fitted input by construction, and no self-citation chain supplies the uniqueness or the dynamic-static link. The derivation is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

Based solely on the abstract, the work relies on standard background results in optimal transport and analysis on metric spaces; no free parameters, invented entities, or ad-hoc axioms are identifiable from the provided information.

axioms (1)
  • domain assumption Metric graphs admit the structure needed to extend Benamou-Brenier formulations and Γ-convergence from RCD*(K,N)-spaces
    Invoked when deriving the dynamic version and stating the extension of known results.

pith-pipeline@v0.9.1-grok · 5661 in / 1304 out tokens · 26778 ms · 2026-07-02T09:53:38.919333+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

31 extracted references · 16 canonical work pages

  1. [1]

    Carrillo and Katy Craig and Li Wang and Chaozhen Wei , journal=

    Jose A. Carrillo and Katy Craig and Li Wang and Chaozhen Wei , journal=. Primal dual methods for. 2021 , doi=

    2021

  2. [2]

    Data-driven gradient flows , volume=

    Pietschmann, Jan-Frederik and Schlottbom, Matthias , year=. Data-driven gradient flows , volume=. doi:10.1553/etna_vol57s193 , journal=

    work page doi:10.1553/etna_vol57s193

  3. [3]

    Sitzungsberichte der Preussischen Akademie der Wissenschaften, Physikalisch-mathematische Klasse , year =

    Schrödinger, Erwin , title =. Sitzungsberichte der Preussischen Akademie der Wissenschaften, Physikalisch-mathematische Klasse , year =

  4. [4]

    Discrete and Continuous Dynamical Systems , volume =

    A survey of the. Discrete and Continuous Dynamical Systems , volume =. 2014 , issn =. doi:10.3934/dcds.2014.34.1533 , url =

    work page doi:10.3934/dcds.2014.34.1533 2014

  5. [5]

    A computational fluid mechanics solution to the

    Jean-David Benamou and Yann Brenier , journal=. A computational fluid mechanics solution to the. 2000 , volume=

    2000

  6. [6]

    Advances in Mathematics , volume =

    The dynamical. Advances in Mathematics , volume =. 2023 , issn =. doi:10.1016/j.aim.2023.109100 , author =

    work page doi:10.1016/j.aim.2023.109100 2023

  7. [7]

    Networks and Heterogeneous Media , volume =

    Gradient flow formulation of diffusion equations in the. Networks and Heterogeneous Media , volume =. 2022 , issn =. doi:10.3934/nhm.2022023 , author =

    work page doi:10.3934/nhm.2022023 2022

  8. [8]

    Journal of Scientific Computing , year = 2017, volume =

    Computations of optimal transport distance with. Journal of Scientific Computing , year = 2017, volume =

    2017

  9. [9]

    A new class of transport distances between measures , volume=

    Dolbeault, Jean and Nazaret, Bruno and Savaré, Giuseppe , year=. A new class of transport distances between measures , volume=. Calculus of Variations and Partial Differential Equations , publisher=. doi:10.1007/s00526-008-0182-5 , number=

    work page doi:10.1007/s00526-008-0182-5

  10. [10]

    Luigi Ambrosio and Nicola Gigli and Giuseppe Savaré , title =

  11. [11]

    doi:10.1007/s00526-021-02089-0 , year =

    Artur Stephan , journal =. doi:10.1007/s00526-021-02089-0 , year =

    work page doi:10.1007/s00526-021-02089-0

  12. [12]

    Le spectre du

    Roth, Jean-Pierre , year =. Le spectre du. Théorie du. doi:10.1007/BFb0100128 , note =

    work page doi:10.1007/bfb0100128

  13. [13]

    Networks & Heterogeneous Media , author =

    Gaussian estimates for a heat equation on a network , volume =. Networks & Heterogeneous Media , author =. 2007 , pages =. doi:10.3934/nhm.2007.2.55 , language =

    work page doi:10.3934/nhm.2007.2.55 2007

  14. [14]

    Heat kernels on metric graphs and a trace formula , language =

    Kostrykin, Vadim and Potthoff, Jürgen and Schrader, Robert , year =. Heat kernels on metric graphs and a trace formula , language =. Adventures in mathematical physics , series =

  15. [15]

    Spectral

    Kurasov, Pavel , year =. Spectral

  16. [16]

    Applied Mathematics and Optimization , author =

    Variational and. Applied Mathematics and Optimization , author =. 2007 , pages =. doi:10.1007/s00245-006-0887-9 , language =

    work page doi:10.1007/s00245-006-0887-9 2007

  17. [17]

    Nonlinear Analysis: Theory, Methods & Applications , author =

    New lower semicontinuity results for nonconvex functionals defined on measures , volume =. Nonlinear Analysis: Theory, Methods & Applications , author =. 1990 , pages =. doi:10.1016/0362-546X(90)90007-4 , number =

    work page doi:10.1016/0362-546x(90)90007-4 1990

  18. [18]

    Semigroup

    Mugnolo, Delio , year =. Semigroup

  19. [19]

    Benamou–

    Gigli, Nicola and Tamanini, Luca , year =. Benamou–. Probability Theory and Related Fields , publisher =. doi:10.1007/s00440-019-00909-1 , number =

    work page doi:10.1007/s00440-019-00909-1

  20. [20]

    Tamanini, Luca , title =

  21. [21]

    Convergence of

    Carlier, Guillaume and Duval, Vincent and Peyr\'. Convergence of. SIAM Journal on Mathematical Analysis , volume =. 2017 , doi =

    2017

  22. [22]

    2026 , issn =

    Weak curvature conditions on metric graphs , journal =. 2026 , issn =. doi:10.3934/nhm.2026031 , url =

    work page doi:10.3934/nhm.2026031 2026

  23. [23]

    Santambrogio, Filippo , year =. Optimal

  24. [24]

    Transactions of the American Mathematical Society , author =

    Differential one-forms on. Transactions of the American Mathematical Society , author =. 2018 , pages =. doi:10.1090/tran/7362 , language =

    work page doi:10.1090/tran/7362 2018

  25. [25]

    Dal Maso, Gianni , year =. An

  26. [26]

    and Pavon, Michele , title =

    Chen, Yongxin and Georgiou, Tryphon T. and Pavon, Michele , title =. Journal of Optimization Theory and Applications , year =

  27. [27]

    2026 , Eprint =

    Mattia Garatti and Luca Nenna and Simona Rota Nodari and Luca Tamanini , Title =. 2026 , Eprint =

    2026

  28. [28]

    The Annals of Probability , author =

    Bakry–. The Annals of Probability , author =. doi:10.1214/14-AOP907 , language =

    work page doi:10.1214/14-aop907

  29. [29]

    ESAIM: Mathematical Modelling and Numerical Analysis , volume=

    An optimal transport approach for solving dynamic inverse problems in spaces of measures , author=. ESAIM: Mathematical Modelling and Numerical Analysis , volume=. 2020 , publisher=

    2020

  30. [30]

    Dynamic Optimal Transport on Networks , volume =

    Burger, Martin and Humpert, Ina and Pietschmann, Jan-Frederik , year =. Dynamic Optimal Transport on Networks , volume =. doi:10.1051/cocv/2023027 , journal =

    work page doi:10.1051/cocv/2023027

  31. [31]

    European Journal of Applied Mathematics , volume =

    Ariane Fazeny and Martin Burger and Jan-Frederik Pietschmann , title =. European Journal of Applied Mathematics , volume =. doi:10.1017/s0956792525000051 , year =

    work page doi:10.1017/s0956792525000051