Pith. sign in

REVIEW 2 major objections 4 minor 12 references

Large systems of symmetrized trapped Brownian Bridges and Schrodinger processes

T0 review · 2 major / 4 minor · reviewed 2026-08-12 · 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 $N$ grows, the empirical path measure of symmetrized trapped Brownian bridges converges to a unique minimizer, a Schrödinger process: uniform diffusion for hard walls, eigenfunction-driven for soft walls.

desk verdict Hard-wall Theorem 3.3 fails because qΛ is not a probability measure; the soft-wall and LDP parts look sound but the advertised uniform minimizer does not hold. read the letter →

arxiv 2411.18359 v2 pith:Q6BAXWCK submitted 2024-11-27 math.PR math-phmath.MP

Stefan Adams , Spyros Garouniatis This is my paper · ORCID
classification math.PRmath-phmath.MP MSC 60F1060J6560J6082B10
keywords symmetrizedBrownianbridgesSchrödingerprocesslargedeviationsDonsker-VaradhanratefunctionoccupationmeasuretrappotentialBose-Einsteincondensationoptimaltransport
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

This paper studies a system of $N$ non-interacting bosons in a trap, represented as $N$ Brownian bridges with symmetrized initial and terminal points, weighted by the trap potential $W$ over the time interval $[0,\beta]$. The central claim is that as $N\to\infty$ the empirical path measure $L_N$ of the bridges converges in probability to a single probability measure $\mu^*_{g,m,W}$, the unique minimizer of a large-deviation rate function, and that this minimizer is a Schrödinger process: the most likely collective path law under the symmetrized constraint. For hard-wall traps the minimizer is an ergodic Markov diffusion with uniform stationary density on the trap box, solving the Schrödinger bridge problem with uniform marginals; for soft-wall traps it is the diffusion with drift $\nabla\varphi/\varphi$ and invariant density $\varphi^2$, where $\varphi$ is the principal eigenfunction of $-\Delta+W$. The paper also proves that the rate function governing occupation measures is $\beta$ times the Donsker–Varadhan functional, yielding a formula for the large-$N$ logarithm of the symmetrized trace $\mathrm{Tr}_+(e^{-\beta H_N})$.

What carries the argument

The engine is the variational representation of the rate function as an infimum over shift-invariant pair measures $q$ on $\mathbb{R}^d\times\mathbb{R}^d$, $$\inf_{q\in $M_1^{{(s)}}$(\mathbb{R}^d\times\mathbb{R}^d)}\Big\{H(q|q\otimes m)-\int\int \log E^\beta_{x,y}[$e^{{-\tilde W}}$]\,q(dx,dy)-\langle q,\log g\rangle\Big\},$$ together with the observation that the unique minimizer $\mu^*$ is a mixture of trap-tilted Brownian bridges $Q^\beta_{x,y}$. The crucial step is that for $m=\mathrm{Leb}$ and $g\equiv p_\beta$, the minimizing pair measure $q^*$ factorizes through the principal eigenfunction $\varphi$ of $-\Delta+W$, converting the path-dependent tilt $e^{-\tilde W}$ into a boundary factor and making the lifted process a Schrödinger process. For hard walls the relevant object is the killed bridge kernel $p_{\beta,\Lambda}$ on the box $\Lambda$, with $q_\Lambda(dx,dy)=|\Lambda|^{-1}p_{\beta,\Lambda}(x,y)\,dx\,dy$; the paper uses this kernel to represent the limit as reflected Brownian motion on $\Lambda$ with uniform invariant measure.

What would settle it

On a finite interval $\Lambda=[0,L]$, compute the first marginal of $q_\Lambda(dx,dy)=|\Lambda|^{-1}p_{\beta,\Lambda}(x,y)dxdy$: for killed Brownian bridges the total bridge mass is $\int_\Lambda p_{\beta,\Lambda}(x,y)dy=P_x(\tau_\Lambda>\beta)<1$, so the marginal is not $|\Lambda|^{-1}dx$ and the equal-marginal condition fails. Showing this calculation explicitly would settle whether the hard-wall minimizer is the claimed uniform Schrödinger process or, instead, the Dirichlet ground state $\varphi_1^2$.

Watch

Extended reading notes

Core claim

The paper's core discovery is that the variational problem behind the large-$N$ limit collapses to a tractable entropy minimization over pair measures. Specifically, the rate function for $L_N$ takes the form $$$I^{{(\mathrm{sym}}$)}_{m,W,g}(\mu)=\inf_{q\in $M_1^{{(s)}}$(\mathbb{R}^d\times\mathbb{R}^d)}\{H(q|q\otimes m)+$I^{{(q)}}$_W(\mu)-\langle q,\log g\rangle\},$$ subtracted by its infimum, and the paper shows this has a unique minimizer $$\mu^*_{g,m,W}=\int_{\mathbb{R}^d}\int_{\mathbb{R}^d} Q^\beta_{x,y}\, q^*_{g,m,W}(dx,dy),$$ where $Q^\beta_{x,y}$ is the Brownian bridge tilted by $e^{-\tilde W}$. When $m=\mathrm{Leb}$ and $g\equiv p_\beta$, the minimizing pair measure factorizes through the principal eigenfunction $\varphi$ of $-\Delta+W$, so the path-level tilt becomes a boundary term; the resulting $\mu^*$ is a Schrödinger process. In the hard-wall case $\Lambda$ the paper claims the minimizer is reflected Brownian motion on $\Lambda$ with uniform invariant measure, and in the soft-wall case it is the diffusion $dX_t=(\nabla\varphi/\varphi)(X_t)\,dt+dB_t$ with invariant density $\varphi^2$. A direct corollary is the identification $J(p)=\beta I_W(p)$ for the occupation-measure rate function, with $I_W(p)=\|\nabla\sqrt{dp/dx}\|_2^2/2+\langle W,p\rangle$ when $dp=\varphi^2\,dx$.

Load-bearing premise

The hard-wall theorem stands on the unstated normalization $\int_\Lambda p_{\beta,\Lambda}(x,y)dy=1$ for the killed bridge kernel, which is what makes $q_\Lambda(dx,dy)=|\Lambda|^{-1}p_{\beta,\Lambda}(x,y)dxdy$ a probability measure with equal uniform marginals; if that equality fails, the proposed minimizer is not admissible.

Editorial extensions

If this is right

  • For large $N$, sampling the symmetrized bridge system is equivalent to sampling a single ergodic diffusion, so Monte Carlo simulations of boson paths can be replaced by one SDE.
  • The occupation-measure rate function is $\beta I_W(p)$, so the exponential cost of observing a density profile $p$ is $\beta(\|\nabla\sqrt{dp/dx}\|_2^2/2 + \langle W,p\rangle)$.
  • For hard-wall traps, the limiting occupation profile inside the trap is flat, matching the uniform-density prediction for the boson cloud in a box.
  • The symmetrized trace satisfies $(1/N)\log \mathrm{Tr}_+(e^{-\beta H_N})\to -\beta\lambda_\Lambda(W)$, a closed formula for the partition function in terms of the principal eigenvalue.
  • For soft-wall traps, the limit solves the optimal transport problem with marginals $\varphi^2\,dx$, connecting trap geometry to Schrödinger's entropy minimization.

Reading between the lines

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

  • If the hard-wall normalization is instead taken to be the killed kernel's total mass, the same variational calculation points to the Dirichlet ground state $\varphi_1^2$ as the invariant density—an alternative limit not pursued in the paper.
  • Tracking how the unique minimizer changes as the temperature $1/\beta$ varies could give a thermodynamic criterion for Bose–Einstein condensation: condensation would appear when the rate function's minimizer ceases to be unique or the ground-state density $\varphi^2$ localizes.
  • The same variational representation should extend to interacting systems along the lines of the paper's conjecture, with the pair-interaction term in $H_N$ producing a mean-field drift; testing this against known stochastic descriptions of condensates would require adding interactions to (2.1) and re-running the LDP.
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.

Referee Report

2 major / 4 minor

Summary. The paper studies the large-N behavior of N symmetrized Brownian bridges on [0,β] in a trap potential W. It derives large deviations principles for the empirical path measure L_N and for the mean occupation measures Y_N, extending results of Adams and König. The main new claims identify the unique minimizer of the rate function with a Schrödinger process: for hard-wall traps the minimizer is asserted to be an ergodic Markov process with uniform invariant measure on the box Λ and to solve the optimal transport problem with uniform marginals; for soft-wall traps the minimizer is asserted to be an h-process with invariant density φ², where φ is the principal eigenfunction of −Δ+W. The paper also connects the resulting rate function to the Donsker-Varadhan functional and to the large-N asymptotics of the symmetrized trace of e^{−βH_N}.

Significance. If correct, the paper would give a path-level, probabilistic description of the bosonic symmetrized trace and a concrete identification of the large-N minimizer with Schrödinger bridges, extending [AK08] to the full empirical path measure. The LDP framework is set up carefully, with explicit rate functions and variational arguments that do not rely on fitted parameters; the soft-wall identification via an h-transform is a natural and potentially useful construction. However, the hard-wall theorem rests on a false normalization of the killed heat kernel. Because that error invalidates the paper's main advertised example, the central claims of the paper are not currently supported.

major comments (2)
  1. [Section 5.4, Eqs. (5.19)–(5.20)] The proof of Theorem 3.3 asserts that qΛ(dx,dy)=|Λ|^{-1}p_{β,Λ}(x,y)dxdy has marginal qΛ(dx)=|Λ|^{-1}dx. This is false for the killed heat kernel defined in the same section: p_{β,Λ}(x,y)=μ^β_{x,y}(Cβ;B[0,β]⊂Λ), so ∫_Λ p_{β,Λ}(x,y)dy = P_x(τ_Λ>β) < 1 for x∈Λ. Hence qΛ has total mass |Λ|^{-1}∫_Λ P_x(τ_Λ>β)dx < 1, is not an element of M^{(s)}_1(R^d×R^d), and the measure μ*_{qΛ} in (5.19) is a sub-probability. It therefore cannot be the unique minimizer of the rate function on M_1(Cβ), and the uniform-marginal, uniform-invariant-measure, and optimal-transport claims of Theorem 3.3 are unsupported.
  2. [Theorem 3.3 and Corollary 3.2] The hard-wall conclusion also contradicts the paper's own general variational result. Applying Corollary 3.2 with m=Leb_Λ, g≡p_β, and W the hard-wall potential, the operator T in (3.4) has kernel p_{β,Λ}(x,y)m(dy), and its principal eigenfunction φ_T is the ground state of the Dirichlet operator on Λ. Corollary 3.2 then gives a minimizer with endpoint density φ_T(x)^2m(dx), not the uniform density, and the corresponding h-transform of killed Brownian motion is not the reflected Brownian motion with uniform stationary density advertised in Theorem 3.3. The uniform endpoint marginals would require the Neumann heat kernel on Λ, not the killed kernel used in Section 5.4.
minor comments (4)
  1. [Section 3.1] The sentence "The proof of Theorem 3.11 can be found in Section 5.6" should refer to Theorem 3.5, not Theorem 3.11.
  2. [Introduction and Section 6] Internal cross-references are inconsistent: the introduction refers to "Section 6.12" for Schrödinger processes, while the appendix section on Schrödinger processes is numbered 6.3; similarly, the LDP background is said to be in "Section 6.2" and appears there, but earlier references point to different subsections.
  3. [Abstract and Section 4] The abstract promises a proof of the large-N asymptotics of Tr_+(e^{−βH_N}), but the body only discusses this in prose in Section 4 via Varadhan's lemma; no numbered theorem or complete proof is provided for that trace asymptotics.
  4. [Equations (3.8) and Section 5.5] The notation for the soft-wall minimizer q* is garbled in places (e.g., "dy dxdy" in (3.8)), and the signs in the exponent defining the martingale D^{(W)}_β in (5.27) are difficult to reconcile with the Feynman–Kac identity used in (5.28)–(5.29). The soft-wall proof needs a careful rewriting before its claims can be checked.

Circularity Check

1 steps flagged · score 6.0 of 10

Hard-wall minimizer in Theorem 3.3 is built on an assumed uniform normalization, so the uniform Schrödinger claim reduces to the ansatz.

  1. self definitional [Section 5.4, proof of Theorem 3.3 (Eq. (5.19)-(5.21))]
    "It is easy to see that qΛ(dx, dy) := 1/|Λ|p_{β,Λ}(x,y)dxdy, x,y ∈ Λ minimizes (5.20) as qΛ is symmetric and qΛ(dx) = 1/|Λ| dx, x ∈ Λ."

    The uniform marginal qΛ(dx)=1/|Λ|dx is the conclusion of Theorem 3.3 (uniform invariant measure and UΛ endpoints), not a consequence of p_{β,Λ}. By the paper's own definition, p_{β,Λ}(x,y)=μ^β_{x,y}(Cβ;B[0,β]⊂Λ) is the killed heat kernel, so ∫_Λ p_{β,Λ}(x,y)dy=P_x(τ_Λ>β)<1 for x∈Λ; hence the first marginal of 1/|Λ|p_{β,Λ}(x,y)dxdy is 1/|Λ|P_x(τ_Λ>β)dx, never the uniform measure. The proof therefore assumes exactly what it needs to prove: the normalized product measure has equal uniform marginals. All subsequent claims—that qΛ minimizes (5.20), that μ*_{qΛ} is a Schrödinger process, that X0 is uniform on Λ, and that the invariant measure is uniform—rest on this assumed normalization.

full rationale

The LDP machinery in Propositions 2.1 and 2.3 and Corollary 2.2 is derived from a mixture-of-LDP-systems argument in Appendix 6.1 and the contraction principle, with no fitted parameters; Proposition 3.1 and Corollary 3.2 are obtained by convex duality and relative-entropy lower bounds. Theorem 3.4 (soft wall) constructs q* from the eigenfunction φ and verifies by martingale/Girsanov arguments that the resulting path measure is the φ²-stationary diffusion; that is a legitimate construction followed by verification. The circularity is localized to Theorem 3.3 and its proof in Section 5.4: the hard-wall minimizer qΛ is posited with a uniform normalization 1/|Λ|, and the uniform marginal qΛ(dx)=1/|Λ|dx is asserted rather than derived. Since p_{β,Λ} is the killed heat kernel, the asserted equality is false, so the uniform-invariant-measure and UΛ-endpoint conclusions are not established by the variational argument; they are built into the ansatz. The [AK08] citations support the occupation-measure rate-function identification in Theorem 3.5 and are independently published, so they do not by themselves make the paper circular. The score reflects one central result reducing by construction, while the remainder of the paper is largely self-contained.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

No empirical fitting appears; the free parameter list is empty. The main axioms are confining-potential assumptions, standard Girsanov and Feynman-Kac facts, the Schrödinger bridge characterization, and one unstated normalization in the hard-wall proof that is not satisfied by the paper's own kernel.

assumptions (7)
  • domain assumption Large deviations estimates from [AK08], including the base LDP for symmetrized bridges with compactly supported initial measure m.
    Proposition 2.1 is proved by adapting [AK08] Theorem 1.1; the paper does not re-derive the base LDP from scratch.
  • standard math Feynman-Kac representation (1.2) for the symmetrized bosonic trace.
    Used in the introduction to connect the path model to Tr_+(e^{-βH_N}).
  • domain assumption Assumption (2.6) on the confining trap W: coercive as |x|→∞, bounded below, continuous on {W<∞}.
    Ensures compactness and exponential tightness in the large deviations arguments.
  • domain assumption For soft-wall potentials: existence of a positive L^2-normalized ground state φ∈H^2(R^d) with Wφ∈L^2.
    Invoked via Reed-Simon Theorem XIII.72 in Section 3 to justify φ∈H^2 for smooth positive W.
  • standard math Girsanov theorem and the martingale property of D^W_t = e^{∫ W - βλ}φ(B_β)/φ(B_0).
    Used in Section 5.5 to identify the soft-wall minimizer with the h-transform diffusion.
  • standard math Schrödinger bridge characterization from [FG97]: dQ/dP_m = S1(ω(0))S2(ω(β)) implies Q is a Schrödinger process.
    Used to certify that the constructed path measures solve the optimal transport problem.
  • ad hoc to paper Unstated normalization ∫_Λ p_{β,Λ}(x,y)dy = 1 used in Section 5.4 to conclude qΛ(dx)=1/|Λ|dx.
    False for the killed Brownian bridge kernel defined in the same section; this is the load-bearing error in Theorem 3.3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Large systems of symmetrized trapped Brownian Bridges and Schrodinger processes." pith.science (2026). https://pith.science/paper/Q6BAXWCK

@misc{pith2026241118359,
  author       = {Pith},
  title        = {Pith review of: Large systems of symmetrized trapped Brownian Bridges and Schrodinger processes},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/Q6BAXWCK}},
  note         = {Machine review of arXiv:2411.18359}
}
abstract

Consider a large system of $N$ Brownian motions in $\R ^d$ fixed on a time interval $[0,\beta]$ with symmetrized initial and terminal conditions, under the influence of a trap potential. Such systems describe systems of bosons at positive temperatures confined in a spatial domain. We describe the large $N$ behavior of the averaged path (that is, their empirical path measure) and its connection with a well known optimal transport problem formulated by Erwin Schr\"odinger. We also explore the asymptotic behavior of the Brownian motions in terms of Large Deviations. In particular, the rate function that governs the mean of occupation measures turns out to be the well-known Donsker-Varadhan rate function. We therefore prove a simple formula for the large $N$ asymptotic of the symmetrized trace of $e^{-\beta \Hcal_N}$, where $\Hcal_N$ is an $N$ particle Hamilton operator in a trap

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

12 extracted references · 12 canonical work pages

  1. [1]

    S. Adams,W. K\"onig , Large deviations for symmetrised Brownian bridges Prob. Theory Relat. Fields. in press, published online DOI 10.1007/s00440-007-0099-5

    work page doi:10.1007/s00440-007-0099-5

  2. [2]

    Billingsley , Convergence of probability measures 2nd edition Wiley Series in Probability and Statistics (1968)

    P. Billingsley , Convergence of probability measures 2nd edition Wiley Series in Probability and Statistics (1968)

    work page 1968

  3. [3]

    Dembo and O

    A. Dembo and O. Zeitouni , Large Deviations Techniques and Applications, 2nd ed., Springer, New York (1998)

    work page 1998

  4. [4]

    ollmer and N. Gantert , Entropy minimization and Schr\

    H. F\"ollmer and N. Gantert , Entropy minimization and Schr\"odinger processes in infinite dimensions, Ann. Probab. 25:2 , 901--926 (1997)

    work page 1997

  5. [5]

    Chung and Z

    K. Chung and Z. Zhao , From Brownian motion to Schr\"odinger’s equation, Springer Berlin (1995)

    work page 1995

  6. [6]

    Ugolini , Stochastic Description of a Bose–Einstein Condensate, Ann

    L.M Morato and S. Ugolini , Stochastic Description of a Bose–Einstein Condensate, Ann. Henri Poincar´e 12 (2011), 1601–1612

    work page 2011

  7. [7]

    Dorlas and S

    T. Dorlas and S. Adams , Asymptotic Feynman–Kac formulae for large symmetrised systems of random walks, www.imstat.org/aihp Annales de l’Institut Henri Poincaré - Probabilités et Statistiques 2008, Vol. 44, No. 5, 837–875

    work page 2008

  8. [8]

    K\"onig , J.B

    W. K\"onig , J.B. Bru and S. Adams , Large systems of path-repellent Brownian motions in a trap at positive temperature, Electronic journal of Probability Vol. 11 (2006), Paper no. 19, pages 460–485

    work page 2006

Show all 12 references
  1. [9]

    Revuz and M

    D. Revuz and M. Yor , Continuous Martingales and Brownian Motion, Springer Berlin (1999)

    1999

  2. [10]

    Kohlmann , Schr\"odinger Operators and their Spectra, Lecture notes (2017)

    M. Kohlmann , Schr\"odinger Operators and their Spectra, Lecture notes (2017)

    2017

  3. [11]

    Alain-Sol Sznitman , Brownian Motion, Obstacles and Random Media, Springer Berlin, Heidelberg (1998)

    A. Alain-Sol Sznitman , Brownian Motion, Obstacles and Random Media, Springer Berlin, Heidelberg (1998)

    1998

  4. [12]

    M. Reed. and B. Simon. , Methods of Modern Mathematical Physics, Elsevier Science (1978)

    1978

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.