Sharp Continuity Moduli for Dirichlet Heat Flow in Boundary-Reservoir Transport Metrics

Maja Gwozdz

REVIEW 2 major objections 4 minor 22 references

Sharp Continuity Moduli for Dirichlet Heat Flow in Boundary-Reservoir Transport Metrics

T0 review · 2 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The killed Dirichlet heat semigroup is exactly 1/p-Hölder in boundary-reservoir transport distances for every p>1, with the exponent optimal among power moduli.

desk verdict A genuinely new sharp Hölder-modulus theorem for the killed heat flow in boundary-reservoir metrics, with a solid proof and one load-bearing definitional error that must be corrected before acceptance. read the letter →

arxiv 2607.16874 v1 pith:AHF6ZG6N submitted 2026-07-18 math.AP

Maja Gwozdz This is my paper · ORCID

classification math.AP MSC 35K0535K2047D0749Q2235B65

keywords Dirichletheatsemigroupboundary-reservoirtransportdistanceHöldermodulusboundaryamplificationHopflemmaKantorovich-RubinsteindualityEVImasssublevel

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

At a fixed positive time, the paper asks how the heat semigroup with zero Dirichlet boundary condition distorts the boundary-reservoir transport distances W_{b,p} on finite measures. It proves that at p=1 the semigroup is globally Lipschitz; for every p>1, on each total-mass sublevel it is exactly 1/p-Hölder, and no better power modulus exists. On the full finite-measure space, the map is discontinuous at the zero measure. The quadratic case p=2 then inherits an obstruction: no standard finite-lambda EVI_lambda gradient-flow semigroup on a W_{b,2}-metric domain can contain the affine constant-boundary heat flow. The mechanism is boundary-layer amplification: a packet initially at distance ε from the boundary has W_{b,p}-distance O(ε) from zero, but after any positive time its p-th boundary moment is at least cε.

What carries the argument

The argument rests on two tools. A dual representation of W_{b,1} over boundary-vanishing Lipschitz test functions, combined with a fixed-time gradient estimate for the Dirichlet heat semigroup, yields the global Lipschitz bound at p=1. For p>1, a Hopf-type linear lower bound (Lemma 3.5) shows that the positive-time solution of the homogeneous Dirichlet problem with nonnegative nonzero initial data is bounded below by a positive constant times the distance to the boundary in a collar. This converts an initial boundary p-moment of order ε^p into a positive-time boundary p-moment of order ε, forcing the sharp 1/p-Hölder exponent, the discontinuity at zero, and the EVI obstruction.

What would settle it

On a domain with a corner, such as the square (0,1)^2 (which is not C^2), place a unit-mass packet at distance ε from a corner, run the killed heat semigroup for a fixed time t>0, and measure the p-th boundary moment of the solution. If that moment decays faster than linearly in ε—so the lower bound is not ε but ε^q with q>1—then the ratio W_{b,p}(P_t μ_ε,0)/W_{b,p}(μ_ε,0)^α would remain bounded for some α>1/p, contradicting the claimed sharp exponent. Equivalently, on a C^2 domain one can directly compute this ratio for a sequence of packets approaching the boundary; the paper asserts it dive

Watch

Extended reading notes

Core claim

The central claim is that for every fixed t>0, the killed Dirichlet heat semigroup P_t is globally Lipschitz with respect to W_{b,1}, and for every p>1 and m>0 there is a constant C_{t,p,m,Ω} such that W_{b,p}(P_t μ, P_t ν)^p ≤ C W_{b,p}(μ,ν) for all finite measures μ,ν with total mass at most m; equivalently, P_t is 1/p-Hölder on each mass sublevel. The exponent 1/p is sharp in the power-modulus scale: for any α>1/p the ratio W_{b,p}(P_t μ, P_t ν)/W_{b,p}(μ,ν)^α is unbounded. On the full finite-measure space, P_t is discontinuous at zero for every p>1. As a corollary in the quadratic case p=2, no standard finite-λ EVI_λ semigroup on a W_{b,2} metric domain can both contain the affine consta

Load-bearing premise

The load-bearing premise is the Hopf-type linear lower bound (Lemma 3.5): at a fixed positive time, a nonnegative nonzero solution of the homogeneous Dirichlet problem is bounded below by a positive constant times the distance to the boundary in a boundary collar; this requires a C^2 boundary (interior ball condition) and parabolic smoothing up to the boundary, and if it fails the sharpness, discontinuity, and EVI-obstruction conclusions collapse.

Editorial extensions

If this is right

For p=1, the semigroup has a finite global Lipschitz constant on all finite measures, with no convexity or curvature assumptions on the domain.
For every p>1, on each total-mass sublevel the map is 1/p-Hölder, and this is the best possible power modulus: any exponent larger than 1/p fails.
On the full finite-measure space, P_t is discontinuous at the zero measure for every p>1, so any fixed-time regularity statement must restrict the total mass.
In the quadratic case p=2, the affine constant-boundary Dirichlet heat flow cannot be realised as the restriction of a standard finite-λ EVI_λ gradient-flow semigroup on a W_{b,2}-metric domain.
The same boundary-layer amplification applies to smooth uniformly elliptic operators, yielding infinite W_{b,p}-Lipschitz constant and discontinuity at zero for the density-induced maps.
A forward nondecreasing orientation of the Ambrosio–Gigli open problem in the quadratic metric fails already for smooth data.

Reading between the lines

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

If the Hopf-type linear lower bound is the true driver, the sharp 1/p exponent should appear for any parabolic flow that amplifies boundary layers linearly, suggesting the result is generic beyond the Laplacian and divergence-form operators.
The discontinuity at zero indicates a fundamental incompatibility between the boundary-reservoir topology and unbounded mass concentrated in thin boundary layers; a modified metric that weights boundary-layer mass differently might restore continuity at zero.
The EVI obstruction suggests that gradient-flow formulations of Dirichlet problems with a boundary reservoir must either restrict to mass-bounded sets, alter the metric, or abandon the standard finite-λ EVI framework.
The C^2 boundary assumption is likely essential: on domains with corners (e.g., a square, which is C^{1,1} but not C^2), the linear lower bound may fail and the exponent could change; computing the p-th boundary moment of a packet near a corner would reveal whether 1/p remains sharp.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

Two things up front. The main theorem is genuinely new and essentially correct: for the killed Dirichlet heat semigroup on a bounded C^2 domain, the fixed-time map is Lipschitz in W_{b,1}, 1/p-Hölder on mass sublevels for p>1, and discontinuous at zero on the full finite-measure space; the 1/p exponent is optimal. The mechanism is the boundary-layer amplification: a unit packet at distance ε has W_{b,p} distance O(ε) from zero, but after positive time the Hopf-type lower bound gives boundary p-moment of order ε, hence W_{b,p} distance of order ε^{1/p}. That is a real insight, and the proof is careful.

The second thing is a real flaw in the written definition. Definition 2.2 puts admissible plans on Ω×Ω, with both marginals entirely on Ω. That is not the boundary-reservoir metric. Under that definition, for any nonzero μ the set Adm(μ,0) is empty, so Lemma 2.6's formula W_{b,p}(μ,0)^p = ∫ δ^p dμ is false; Lemma 2.3's construction even places mass at a boundary point b_0, which is outside Ω×Ω. The stress-test note describes this exactly. The intended correction is unambiguous: plans should live on Ω̄×Ω̄, with marginals equal to μ and ν on Borel subsets of Ω. Boundary mass is free. With that change, Lemma 2.6, the comparison lemmas, the duality statement, and the boundary-layer witnesses are all consistent. So this is a fixable definitional slip, not a hole in the mathematics.

Credit where due: the p=1 endpoint uses the collapsed-boundary Kantorovich–Rubinstein duality plus a fixed-time parabolic gradient estimate; the comparison to p>1 is a short chain; the lower-bound witnesses in Proposition 3.6 are explicit; and Appendix A really proves the Hopf-type lower bound with the regularity details. The uniformly elliptic extension in Section 5 and the EVI obstruction in Section 6 are honest consequences, not afterthoughts. I do not see fitted parameters or circular dependencies.

Beyond the definition, the soft spots are minor. The C^2 assumption is explicit and is needed for the Hopf lower bound; on rougher domains the sharpness argument may fail. The exposition is dense but not misleading.

This paper is for researchers working in optimal transport with boundary reservoirs, Dirichlet gradient flows, or metric-pair transport. It deserves a serious referee. Send it to review, with a request to fix Definition 2.2 (and re-check the statements that depend on it) before acceptance.

Referee Report

2 major / 4 minor

Summary. The paper studies the fixed-time regularity of the killed Dirichlet heat semigroup P_t on the space of finite nonnegative measures over a bounded C^2 domain Ω, equipped with the Figalli–Gigli boundary-reservoir transport distances W_{b,p}. Theorem 1.1 claims: (i) P_t is globally Lipschitz for W_{b,1}; (ii) for p>1, on each mass sublevel M_{≤m}, W_{b,p}(P_tμ,P_tν)^p ≤ C W_{b,p}(μ,ν), i.e. 1/p-Hölder continuity; (iii) the exponent 1/p is optimal in the power scale; (iv) on the full finite-measure space P_t is discontinuous at the zero measure. The proof combines metric comparison between W_{b,1} and W_{b,p}, a Kantorovich–Rubinstein duality for W_{b,1}, a positive-time global gradient estimate for the Dirichlet semigroup, and a parabolic Hopf-type linear lower bound near the boundary. Lower-bound witnesses are single-mass packets placed at distance ε from ∂Ω: their input distance to zero is O(ε), while after positive time the p-th boundary moment is ≥ cε. Section 5 extends the lower-bound obstruction to uniformly elliptic operators, and Section 6 applies the p=2 case to exclude any standard finite-λ EVI_λ semigroup whose domain contains the affine constant-boundary data class X_c and which restricts to the affine Dirichlet heat flow.

Significance. If the theorem is accepted after the necessary correction to Definition 2.2, the paper settles the fixed-time modulus question for Dirichlet heat flow in the boundary-reservoir metric family. The p=1 endpoint is clean and uses only the shortcut/Kantorovich–Rubinstein representation; the p>1 upper bound is a short and transparent consequence of the p=1 estimate plus mass-sublevel comparison. The lower-bound strategy—Hopf amplification of boundary layers—is explicit, with no fitted parameters, and yields quantitative witnesses for sharpness, infinite Lipschitz constants, and discontinuity at zero. The EVI_λ obstruction in the quadratic case is a credible answer to a question in the Ambrosio–Gigli user's guide and clarifies that the original finite-measure W_{b,2} metric cannot support a standard EVI realization containing the affine constant-boundary flow. The main caveat is that the present statement of W_{b,p} in Definition 2.2 is not the metric actually used in the proofs; once that is repaired, the chain is coherent.

major comments (2)

[Definition 2.2; Lemmas 2.3, 2.6; throughout] Definition 2.2 takes admissible plans to be finite Borel measures on Ω×Ω whose marginals equal μ and ν on Borel sets E⊂Ω. Under this definition no plan can charge ∂Ω: if ν=0, Adm(μ,0)=∅ for μ≠0, and the formula W_{b,p}(μ,0)^p=∫δ^p dμ in Lemma 2.6 is false. The construction in Lemma 2.3, γ=(Id,b0)#μ+(b0,Id)#ν with b0∈∂Ω, is not a measure on Ω×Ω. Lemmas 2.6, 2.3, Remark 2.4, Lemma 2.8, Appendix B, and the sharpness witnesses all require plans with boundary mass, so Theorem 1.1 as stated is not a theorem about the metric of Definition 2.2. The intended correction is unambiguous: take plans on Ω̄×Ω̄ with (π1)#γ|_Ω=μ and (π2)#γ|_Ω=ν (boundary mass free). Please correct Definition 2.2 and adjust the statements that quote it.
[Lemma 3.5 and Section 5] Lemma 3.5 assumes A,b,q∈C∞(Ω), but the proof and Lemma 4.4 rely on global W^{2,1}_r and Hopf estimates that require the coefficients to be bounded (and sufficiently regular) up to ∂Ω; C∞ on the open set does not imply boundedness near the boundary, and the proof uses ∥b∥∞ and ∥q∥∞. This does not affect the Laplacian endpoint, but it underpins the uniformly elliptic extension in Proposition 5.3 and Theorems 5.4–5.5. Please add explicit smoothness-up-to-boundary/boundedness hypotheses for A,b,q in Lemma 3.5 and Section 5.

minor comments (4)

[Lemma 2.6 proof] After the definitional correction, the sentence 'γ(Ω×Ω)=0' should read 'γ(Ω̄×Ω)=0'; the current wording is false even in the intended boundary-reservoir metric.
[Theorem 5.4] In the displayed ratio, '(P_t^L (muε0))dx' should be '(P_t^L (m u_ε^0))dx'.
[Corollary 6.5 proof] 'forcequal to the entropy-selected boundary constant' should read 'for c equal to the entropy-selected boundary constant'.
[Appendix A] The parabolic domain is denoted 'P F' in the barrier argument, which is easily confused with the semigroup P_t; consider renaming it (e.g. Q or D).

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is self-contained.

full rationale

The paper's derivation chain does not reduce to its inputs. The p=1 Lipschitz endpoint is proved from the Kantorovich–Rubinstein duality for the boundary-reservoir metric (Theorem 4.1, proved in Appendix B from the shortcut metric) and from the positive-time gradient estimate (Theorem 4.5), both of which are independent of the target modulus. The p>1 upper bound follows by comparing W_{b,p} with W_{b,1} on mass sublevels (Lemma 2.8) and then applying the p=1 endpoint; this is a derivation, not a restatement. Sharpness uses the boundary-layer packets of Proposition 3.6, whose only inputs are the distance-to-zero formula W_{b,p}(μ,0)^p=∫δ^p dμ (Lemma 2.6) and the Hopf-type linear lower bound (Lemma 3.5, proved in Appendix A by standard parabolic Hopf arguments). Neither assumes Theorem 1.1. The discontinuity at zero and the EVI obstruction are corollaries of the same lower-bound construction. There are no fitted parameters, no author-self-citations, and no imported uniqueness theorem. A separate, non-circular defect is that Definition 2.2 restricts admissible plans to Ω×Ω, so Adm(μ,0)=∅ for μ≠0 under the literal wording; Lemmas 2.3 and 2.6 require plans supported on Ω̄×Ω̄. This is a well-posedness/correctness issue, not a circularity, and it does not raise the circularity score.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The proof is built on standard tools: parabolic L^p boundary regularity, maximum principles, Kantorovich–Rubinstein duality, and Riesz representation. No free parameters are fitted to data; the constants (α, r, L_{t,Ω}, C_{t,p,m,Ω}) are existential outcomes of the estimates. No entities are invented. The boundary-reservoir metric is prior art (Figalli–Gigli), not introduced here.

assumptions (5)

standard math Global W^{2,1}_r estimates for zero-initial Cauchy–Dirichlet problems for uniformly parabolic operators on C^{1,1} cylinders (Lieberman [17]).
Invoked in Lemma 4.4 and used for the fixed-time gradient estimate (Theorem 4.5) and for the boundary regularity in Lemma 3.5; if these estimates fail, the p=1 Lipschitz bound and the Hopf lower-bound proof collapse.
standard math Parabolic maximum principle, strong maximum principle, and parabolic Hopf boundary lemma for uniformly parabolic operators with bounded coefficients.
Used in Lemma 3.5/Appendix A to obtain w(t0,x) ≥ α δ(x) near ∂Ω; this is the load-bearing lower-bound input for sharpness and the EVI obstruction.
standard math Kantorovich–Rubinstein duality for the 1-Wasserstein distance on a compact metric space.
Used in the proof of Theorem 4.1 (Appendix B) to derive the dual formula for W_{b,1}, which is essential for the p=1 Lipschitz endpoint.
standard math Riesz representation theorem identifying positive linear functionals on C_0(Ω̄) with finite Radon measures.
Used in Definition 3.3 to define the dual action of the killed semigroup on M(Ω).
standard math Existence, C_0-contractivity, and strong continuity of the Dirichlet heat semigroup on C_0(Ω̄), plus L^1–C_0 duality (Lemma 3.4).
Background for Definition 3.3 and the duality identity used in the lower-bound estimates.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Sharp Continuity Moduli for Dirichlet Heat Flow in Boundary-Reservoir Transport Metrics." pith.science (2026). https://pith.science/paper/AHF6ZG6N

@misc{pith2026260716874,
  author       = {Pith},
  title        = {Pith review of: Sharp Continuity Moduli for Dirichlet Heat Flow in Boundary-Reservoir Transport Metrics},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/AHF6ZG6N}},
  note         = {Machine review of arXiv:2607.16874}
}

abstract

Let $\Omega\subset\mathbb R^n$ be a bounded $C^2$ open set and let $P_t$ be the killed Dirichlet heat semigroup. We prove the sharp fixed-time power-scale modulus of $P_t$ for the Figalli--Gigli boundary-reservoir transport distances $W_{b,p}$. For every $t>0$, $P_t$ is globally Lipschitz with respect to $W_{b,1}$. For every $p>1$, and on every total-mass sublevel $\{\mu:\mu(\Omega)\le m\}$, it is $1/p$-H\"older: \[ W_{b,p}(P_t\mu,P_t\nu)^p \le C_{t,p,m,\Omega} W_{b,p}(\mu,\nu). \] For $p>1$, we show that the exponent $1/p$ is optimal in the scale of power moduli. On the full finite-measure space, $P_t$ is discontinuous at the zero measure. To establish the lower bound, we rely on the amplification of the boundary layer. More precisely, a unit mass initially placed at distance $\varepsilon$ from $\partial\Omega$ has input $W_{b,p}$-distance $O(\varepsilon)$ from zero, whereas after any fixed positive time, its $p$-th boundary moment is bounded below by $c\varepsilon$. As a result, in the quadratic case and in the original finite-measure $W_{b,2}$ metric, there does not exist a standard finite-$\lambda$ $\mathrm{EVI}_\lambda$ semigroup on a $W_{b,2}$-metric domain which would contain the affine constant-boundary data class and could restrict to the affine constant-boundary Dirichlet heat flow. Finally, we also describe the corresponding lower-bound obstruction for smooth uniformly elliptic perturbations in divergence form.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

22 extracted references · 4 linked inside Pith

[1]

Figalli and N

A. Figalli and N. Gigli. A new transportation distance between non-negative measures, with applications to gradients flows with Dirichlet boundary conditions.J. Math. Pures Appl., 94(2):107–130, 2010

2010
[2]

Profeta and K.-T

A. Profeta and K.-T. Sturm. Heat flow with Dirichlet boundary conditions via optimal transport and gluing of metric measure spaces.Calc. Var. Partial Differential Equations, 59:Article 117, 2020

2020
[3]

J. Morales. A new family of transportation costs with applications to reaction–diffusion and parabolic equations with boundary conditions.J. Math. Pures Appl., 112:41–88, 2018

2018
[4]

Erbar and G

M. Erbar and G. Meglioli. Gradient flow for a class of diffusion equations with Dirichlet boundary data,
[5]

Quattrocchi

F. Quattrocchi. Variational structures for the Fokker–Planck equation with general Dirichlet boundary conditions.Calc. Var. Partial Differential Equations, 65:Article 23, 2026

2026
[6]

D. Kim, D. Koo, and G. Seo. A gradient flow for the porous medium equations with Dirichlet boundary conditions, 2025. arXiv:2212.06092v3 [math.AP]

arXiv 2025
[7]

Ambrosio and N

L. Ambrosio and N. Gigli. A user’s guide to optimal transport. InModelling and Optimisation of Flows on Networks, volume 2062 ofLecture Notes in Mathematics, pages 1–155. Springer, Berlin, Heidelberg, 2013
[8]

Casteras, L

J.-B. Casteras, L. Monsaingeon, and F. Santambrogio. Sticky-reflecting diffusion as a Wasserstein gradient flow.J. Math. Pures Appl., 199:Article 103721, 2025

2025

Show all 22 references

[9]

Bate and A

D. Bate and A. L. Garc ´ ıa Pulido. Bi-Lipschitz embeddings of the space of unorderedm-tuples with a partial transportation metric.Math. Ann., 390:3109–3131, 2024

2024
[10]

Optimal partial transport for metric pairs, 2025

Mauricio Che. Optimal partial transport for metric pairs, 2025. arXiv:2406.17674v2 [math.MG]

2025 arXiv
[11]

Relative optimal transport, 2026

Peter Bubenik and Alex Elchesen. Relative optimal transport, 2026. arXiv:2411.05678v3 [math.MG]

2026 arXiv
[12]

Savar´ e and G

G. Savar´ e and G. E. Sodini. A relaxation viewpoint to unbalanced optimal transport: duality, optimality and Monge formulation.J. Math. Pures Appl., 188:114–178, 2024

2024
[13]

Bormann, L

M. Bormann, L. Monsaingeon, D. R. M. Renger, and M. von Renesse. A gradient flow that is none: heat flow with Wentzell boundary conditions.Indag. Math. (N.S.), 37(3):1042–1063, 2026

2026
[14]

R. L. Foote. Regularity of the distance function.Proc. Amer. Math. Soc., 92(1):153–155, 1984

1984
[15]

S. G. Krantz and H. R. Parks. Distance toC k hypersurfaces.J. Differential Equations, 40(1):116–120, 1981

1981
[16]

E. M. Ouhabaz.Analysis of Heat Equations on Domains, volume 31 ofLondon Mathematical Society Monographs Series. Princeton University Press, Princeton, NJ, 2005

2005
[17]

G. M. Lieberman.Second Order Parabolic Differential Equations. World Scientific, 1996

1996
[18]

F.-Y. Wang. Gradient estimates of Dirichlet heat semigroups and application to isoperimetric inequali- ties.Ann. Probab., 32(1A):424–440, 2004. 38

2004
[19]

Ambrosio, N

L. Ambrosio, N. Gigli, and G. Savar´ e.Gradient Flows in Metric Spaces and in the Space of Probability Measures. Lectures in Mathematics ETH Z¨ urich. Birkh¨ auser, Basel, 2nd edition, 2008

2008
[20]

Santambrogio

F. Santambrogio. Euclidean, metric, and Wasserstein gradient flows: an overview.Bull. Math. Sci., 7(1):87–154, 2017

2017
[21]

Divol and T

V. Divol and T. Lacombe. Understanding the topology and the geometry of the space of persistence diagrams via optimal partial transport.J. Appl. Comput. Topol., 5(1):1–53, 2021. 39

2021
[2025]

arXiv:2408.05987v2 [math.AP]

arXiv

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [1]

Figalli and N

A. Figalli and N. Gigli. A new transportation distance between non-negative measures, with applications to gradients flows with Dirichlet boundary conditions.J. Math. Pures Appl., 94(2):107–130, 2010

2010

[2] [2]

Profeta and K.-T

A. Profeta and K.-T. Sturm. Heat flow with Dirichlet boundary conditions via optimal transport and gluing of metric measure spaces.Calc. Var. Partial Differential Equations, 59:Article 117, 2020

2020

[3] [3]

J. Morales. A new family of transportation costs with applications to reaction–diffusion and parabolic equations with boundary conditions.J. Math. Pures Appl., 112:41–88, 2018

2018

[4] [4]

Erbar and G

M. Erbar and G. Meglioli. Gradient flow for a class of diffusion equations with Dirichlet boundary data,

[5] [5]

Quattrocchi

F. Quattrocchi. Variational structures for the Fokker–Planck equation with general Dirichlet boundary conditions.Calc. Var. Partial Differential Equations, 65:Article 23, 2026

2026

[6] [6]

D. Kim, D. Koo, and G. Seo. A gradient flow for the porous medium equations with Dirichlet boundary conditions, 2025. arXiv:2212.06092v3 [math.AP]

arXiv 2025

[7] [7]

Ambrosio and N

L. Ambrosio and N. Gigli. A user’s guide to optimal transport. InModelling and Optimisation of Flows on Networks, volume 2062 ofLecture Notes in Mathematics, pages 1–155. Springer, Berlin, Heidelberg, 2013

[8] [8]

Casteras, L

J.-B. Casteras, L. Monsaingeon, and F. Santambrogio. Sticky-reflecting diffusion as a Wasserstein gradient flow.J. Math. Pures Appl., 199:Article 103721, 2025

2025