Hamiltonian Lift of Bures--Wasserstein Covariance Dynamics with a Spectral Floor

Christian Kerskens

Hamiltonian Lift of Bures--Wasserstein Covariance Dynamics with a Spectral Floor

T0 review · 0 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read The paper establishes a Hamiltonian parent theory for Bures–Wasserstein covariance relaxation, with explicit H=2tr(ΠΣΠ)+V and a spectral-floor stiffness that diverges as (s−ν)^−2.

desk verdict A clean, correct little derivation of the Hamiltonian structure behind Bures–Wasserstein covariance flows; worth reading and citing, nothing earth-shaking. read the letter →

arxiv 2607.16380 v1 pith:WH7VQF2V submitted 2026-07-17 quant-ph math-phmath.MP

Christian Kerskens This is my paper · ORCID

classification quant-phmath-phmath.MP

keywords Bures-WassersteinmetricHamiltonianliftLyapunovoperatorcovariancedynamicsgradientflowspectralfloorRayleighdissipationGaussianfluctuations

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

The paper constructs a finite-dimensional Hamiltonian description of covariance matrices evolving under the Bures–Wasserstein metric, which is usually studied only through gradient flows. It derives canonical momentum Π=½LΣ[Σ̇] and Hamiltonian H(Σ,Π)=2tr(ΠΣΠ)+V(Σ), giving explicit equations of motion. Adding Rayleigh dissipation and taking the overdamped limit recovers the standard Bures–Wasserstein gradient flow, so the Hamiltonian system acts as a conservative parent for dissipative covariance relaxation. For a spectral-floor potential, the Hessian at the isotropic equilibrium separates trace and traceless fluctuations, with a stiffness C=κ/[2(s−ν)²] that diverges as the equilibrium approaches the floor.

What carries the argument

The Lyapunov operator LΣ[A], defined as the unique symmetric solution of ΣLΣ[A]+LΣ[A]Σ=A, encodes the Bures–Wasserstein metric and its inverse. Its trace-pairing self-adjointness converts the kinetic Lagrangian into the closed-form Hamiltonian 2tr(ΠΣΠ), and the Lyapunov form of the kinematic equation Σ̇=2(ΣΠ+ΠΣ) guarantees positivity preservation by congruence. Rayleigh dissipation with coefficient γ provides the mechanism that collapses the conservative flow onto the gradient flow in the overdamped limit.

What would settle it

Numerically integrate the linearized fluctuation equation Q¨ = -4sK[Q] for the spectral-floor potential at several values of s−ν and check that the traceless-mode frequency scales as (s−ν)^−1; if it does not, the stiffness divergence claim fails. Alternatively, simulate the full damped Hamiltonian equations in the large-γ limit and check convergence to the gradient flow of V/γ.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the Bures–Wasserstein metric on the cone of positive-definite covariance matrices, usually used only to define gradient flows, admits a finite-dimensional Hamiltonian lift. The natural Lagrangian gives canonical momentum Π=½LΣ[Σ̇], where LΣ is the Lyapunov operator, and the kinetic Hamiltonian closes to H=2tr(ΠΣΠ)+V(Σ) with canonical equations Σ̇=2(ΣΠ+ΠΣ) and Π̇=−2Π²−∂V/∂Σ. Adding Rayleigh dissipation and taking the overdamped limit recovers the familiar Bures–Wasserstein gradient flow, so the Hamiltonian system is a conservative parent rather than a replacement. For a spectral-floor trace potential, the Hessian at the isotropic equilibrium separates tra

Load-bearing premise

The entire canonical structure depends on the Lyapunov operator being self-adjoint under the trace pairing, a property that is true for positive-definite Σ but is invoked without proof in the Legendre-transform step.

Editorial extensions

If this is right

The Bures–Wasserstein gradient flow is not the Legendre transform of the lift but its overdamped Rayleigh reduction, so the Hamiltonian system serves as a conservative parent theory for dissipative covariance relaxation.
The Hamiltonian flow preserves the canonical symplectic form and preserves positivity of Σ along finite regular solutions, because Σ̇ has Lyapunov form.
For the spectral-floor trace potential, the Hessian at the isotropic equilibrium separates trace and traceless fluctuations, with baseline stiffness C=κ/[2(s−ν)²].
All local fluctuation modes acquire unbounded restoring stiffness as the equilibrium approaches the floor, with frequencies diverging as (s−ν)^−1.
The spectral barrier prevents finite-energy trajectories from crossing the floor because V→+∞ as the smallest eigenvalue of Σ−νI tends to zero.

Reading between the lines

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

A natural extension is to quantize this explicit Hamiltonian on the cone; the spectral floor would then act like a hard-wall boundary condition, and the (s−ν)^−2 stiffness suggests the ground-state wavefunction is suppressed near the floor.
The Lyapunov-operator representation may generalize to other affine-invariant metrics on symmetric positive-definite matrices, giving each such metric a closed-form kinetic Hamiltonian, not just the Bures–Wasserstein one.
A direct numerical check of the overdamped reduction for n>1 would simulate the damped Hamiltonian equations with a simple polynomial potential and compare long-time trajectories with the gradient flow; the paper gives an analytic argument but no numerical demonstration.
The paper notes that the ordinary spectral floor does not enforce the Robertson–Schrödinger condition for quantum covariance matrices; a testable extension is to replace the barrier with one acting on symplectic eigenvalues and see whether the same (s−ν)^−2 divergence appears.

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

Quick take: I agree with the reader. This is a solid, self-contained piece of classical mechanics on the covariance cone. The central new object — the explicit Hamiltonian H=2tr(ΠΣΠ) with the Lyapunov momentum — is real and not in the cited prior work. The equations check out, the scalar n=1 sanity check is right, and the overdamped reduction correctly recovers the BW gradient flow via Rayleigh dissipation, not by Legendre transform. The spectral-floor Hessian calculation is straightforward and the (s−ν)⁻² stiffness divergence follows cleanly. The stress-test note is also right: the self-adjointness assumption in Proposition 1 is unstated but true by cyclicity for positive-definite Σ, so no gap there.

Where are the soft spots? Modest significance. This is a useful construction rather than a deep discovery. The singular limit in Section 3 is heuristic — "γ→∞ with F fixed" is a plausible scaling but not justified rigorously. That is a real but minor caveat, since the center of the paper is the Hamiltonian lift, not the limit. The potential (13) is explicitly a toy model; the author says so, and the Discussion honestly flags that it doesn't enforce the Robertson–Schrödinger condition. Good. The citation pattern is clean: standard references, plus the López-Saldívar/Man'ko line, positioned as complementary rather than claiming priority. No circularity; the parameters are inputs.

For a referee: I'd send it out. It is short, correct, and gives people working on Gaussian flows or BW geometry a working Hamiltonian they didn't have. It won't reshape anything, but it doesn't need to. The only things I'd ask in revision: make the self-adjointness explicit, and either prove the overdamped limit properly or label it as formal.

Who reads it: people in quantum Gaussian dynamics, optimal transport on Gaussians, and covariance estimation who want inertial corrections or fluctuation spectra. I'd take it if offered.

Referee Report

0 major / 4 minor

Summary. The manuscript constructs a finite-dimensional Hamiltonian lift of covariance dynamics on the positive-definite cone equipped with the Bures–Wasserstein metric. Starting from the mechanical Lagrangian L = (1/2) g_BW(Σdot, Σdot) − V(Σ), it defines canonical momentum Π = (1/2) LΣ[Σdot], where LΣ is the Lyapunov operator, and obtains the explicit Hamiltonian H(Σ, Π) = 2 tr(ΠΣΠ) + V(Σ) with Hamilton equations (5). It then shows that adding Rayleigh dissipation and taking the overdamped limit recovers the Bures–Wasserstein gradient flow (Section 3). For the spectral-floor/trace potential (13), it computes the isotropic equilibrium, the Hessian at equilibrium, and the quadratic fluctuation Hamiltonian, obtaining separated trace and traceless stiffnesses and the boundary scaling C = κ/[2(s−ν)^2]. I checked the central computations: the Legendre transform, the Hamilton equations, the n=1 scalar reduction, the variation of the potential, the Hessian, and the linearized fluctuation equations. They are internally consistent.

Significance. If the result holds — and the calculations in the manuscript bear it out — the paper provides a compact conservative parent theory for Bures–Wasserstein covariance relaxation, with an explicit closed-form kinetic Hamiltonian rather than an abstract cotangent-bundle construction. The spectral-floor potential yields a concrete, falsifiable local prediction: the stiffness of all covariance fluctuation modes diverges as (s−ν)^{-2}, with frequencies diverging as (s−ν)^{-1}. The trace/traceless decomposition of the quadratic Hamiltonian is simple and potentially useful. The derivation is self-contained, includes a correct scalar check, and does not fit parameters to the claimed scaling; ν, κ, α, λ enter as inputs. The manuscript also states its limitations honestly, including the fact that the spectral floor does not by itself enforce the Robertson–Schrödinger condition or Gaussian separability.

minor comments (4)

[§2, Proposition 1] The proof of Proposition 1 invokes the trace pairing to identify symmetric matrices with their duals, which requires the Lyapunov operator LΣ to be self-adjoint with respect to the Hilbert–Schmidt inner product. This is true for Σ ≻ 0 by a short cyclicity argument, but it is not stated or proved. Adding a one-line proof or an explicit citation before Eq. (4) would make the derivation fully self-contained.
[§3, Eq. (11)] The overdamped reduction is formally correct but is presented as a terse singular limit. It would help to state explicitly that the term d/dt(∂L/∂Σdot) is being neglected on the slow manifold, with γ → ∞ and F = V/γ held fixed. As written, Eq. (11) is a heuristic derivation; the conclusion (12) is correct, but the limiting statement deserves one or two sentences of justification.
[§2.1, Eq. (10)] The isotropic-sector equations (10) are presented as a reduction of the matrix equations. If a reduced canonical Hamiltonian structure is intended, the normalization relative to the symplectic form tr(dΠ ∧ dΣ) should be clarified, since with Π = pI and Σ = σI the reduced symplectic form carries a factor of n. As written, Eq. (10) is best read as a direct projection of the matrix equations; a short remark would prevent confusion.
[§5, Eq. (16)] The notation Q:K:Q for the fourth-order Hessian tensor is used without an explicit component definition. A brief parenthetical definition, e.g. Σ_{ijkl} Q_{ij} K_{ijkl} Q_{kl}, would improve readability, especially since the paper is otherwise very explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the derivation is self-contained from the stated Bures–Wasserstein metric and potential.

full rationale

The chain of derivation is fully internal and does not reduce to its inputs. The Hamiltonian H(Σ,Π)=2tr(ΠΣΠ)+V(Σ) is obtained by a direct Legendre transform of the natural Lagrangian L=1/2 g_BW(Σ̇,Σ̇)−V(Σ), with the canonical momentum Π=1/2 L_Σ[Σ̇] and the kinematic relation Σ̇=2(ΣΠ+ΠΣ) following from the definition of the Lyapunov operator. No fitted parameter is renamed as a prediction: the potential parameters ν, κ, α, λ are inputs, and the claimed stiffness divergence C=κ/[2(s−ν)^2] is the explicit second derivative of that chosen potential, not a quantity inferred from data. Similarly, the overdamped reduction to the Bures–Wasserstein gradient flow is obtained under the clearly stated rescaling V=γF and limit γ→∞ with F fixed; it is a consequence of the Rayleigh dissipation model, not an import of the target result. The only technical assumption flagged by reviewers, the self-adjointness of L_Σ with respect to the trace pairing, is true for symmetric matrices on positive-definite Σ and is used only in the standard identification of symmetric matrices with their duals. No load-bearing self-citations or imported uniqueness claims appear; references are standard external sources. The paper is self-contained against analytical benchmarks, so the circularity score is 0.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central Hamiltonian construction rests only on standard matrix analysis and the Bures–Wasserstein metric. The four potential parameters are ad hoc inputs for the example potential; they are not fitted or inferred. No new physical entities are introduced.

free parameters (4)

ν (spectral floor)
Lower bound of the logarithmic barrier; the stiffness constant C=κ/[2(s−ν)^2] depends on it.
κ (barrier strength)
Sets the prefactor of the barrier stiffness; chosen by hand in the potential.
α (trace curvature)
Strength of the fourth-order trace potential; contributes 3αs² to the trace-mode stiffness.
λ (trace target)
Preferred value of Φ³; equilibrium condition α(s³−λ)=κ/[2(s−ν)] uses it.

assumptions (5)

standard math Lyapunov equation ΣX+XΣ=A has a unique symmetric solution for Σ≻0, and LΣ is invertible and self-adjoint under the trace pairing.
Used to define the metric, canonical momentum, and Hamiltonian; invoked in Section 2 and the proof of Proposition 1.
standard math Tangent and cotangent spaces of Sym+_n are identified with symmetric matrices via the Hilbert–Schmidt trace pairing.
Basis for the Legendre transform; explicitly used in the Proposition 1 proof.
domain assumption The Bures–Wasserstein metric is g_BW,Σ(A,B)=1/2 tr(LΣ[A]B) and is a valid Riemannian metric.
Taken from earlier work as the starting point; Section 2.
domain assumption The overdamped limit γ→∞ with F=V/γ fixed is valid and inertial terms become negligible.
Needed in Section 3 to recover the BW gradient flow; standard but unproved singular-perturbation step.
domain assumption Finite-energy trajectories cannot cross the logarithmic barrier because V→∞ at the boundary.
Used in Section 5/Corollary 1 to argue the floor is not crossed; relies on conservation of H.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Hamiltonian Lift of Bures--Wasserstein Covariance Dynamics with a Spectral Floor." pith.science (2026). https://pith.science/paper/WH7VQF2V

@misc{pith2026260716380,
  author       = {Pith},
  title        = {Pith review of: Hamiltonian Lift of Bures--Wasserstein Covariance Dynamics with a Spectral Floor},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WH7VQF2V}},
  note         = {Machine review of arXiv:2607.16380}
}

abstract

Covariance dynamics on the positive-definite cone are commonly described by gradient flows, which encode dissipative relaxation but obscure the underlying phase-space structure. We construct a finite-dimensional Hamiltonian lift of covariance dynamics on Sym$^+_n$ equipped with the Bures--Wasserstein metric. The natural mechanical Lagrangian yields canonical momentum $\Pi=\tfrac12 L_\Sigma[\dot\Sigma]$, where $L_\Sigma$ is the Lyapunov operator, and explicit Hamiltonian $\mathcal{H}(\Sigma,\Pi) = 2{\rm tr}(\Pi\Sigma\Pi)+V(\Sigma)$. Adding Rayleigh dissipation recovers the Bures--Wasserstein gradient flow in the overdamped limit. For a spectral-floor and trace-control potential, the quadratic fluctuation Hamiltonian around the isotropic equilibrium separates trace and traceless modes; the baseline stiffness diverges as $(s-\nu)^{-2}$ as the equilibrium covariance approaches the floor. The construction identifies a conservative parent system for constrained Bures--Wasserstein covariance relaxation and fixes the local stiffness scale induced by the spectral floor.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

26 extracted references · 3 canonical work pages

[1]

C. R. Givens, R. M. Shortt, A class of wasserstein metrics for probability distributions, The Michigan Mathematical Journal 31 (2) (1984) 231–240

1984
[2]

Takatsu, Wasserstein geometry of gaussian measures, Osaka J

A. Takatsu, Wasserstein geometry of gaussian measures, Osaka J. Math. 48 (2011) 1005–1026

2011
[3]

Villani, Optimal Transport: Old and New, Springer, 2009

C. Villani, Optimal Transport: Old and New, Springer, 2009

2009
[4]

Malagò, L

L. Malagò, L. Montrucchio, G. Pistone, Wasserstein rie- mannian geometry of gaussian densities, Information Ge- ometry 1 (2018) 137–179

2018
[5]

Bhatia, T

R. Bhatia, T. Jain, Y . Lim, On the bures–wasserstein dis- tance between positive definite matrices, Expositiones Mathematicae 37 (2) (2019) 165–191

2019
[6]

Bengtsson, K

I. Bengtsson, K. ˙Zyczkowski, Geometry of Quantum States: An Introduction to Quantum Entanglement, 2nd Edition, Cambridge University Press, 2017

2017
[7]

Jordan, D

R. Jordan, D. Kinderlehrer, F. Otto, The variational formu- lation of the fokker–planck equation, SIAM J. Math. Anal. 29 (1998) 1–17

1998
[8]

Otto, The geometry of dissipative evolution equations: the porous medium equation, Comm

F. Otto, The geometry of dissipative evolution equations: the porous medium equation, Comm. Partial Differential Equations 26 (2001) 101–174

2001

Show all 26 references

[9]

Seifert, Stochastic thermodynamics, fluctuation the- orems and molecular machines, Reports on Progress in Physics 75 (12) (2012) 126001

U. Seifert, Stochastic thermodynamics, fluctuation the- orems and molecular machines, Reports on Progress in Physics 75 (12) (2012) 126001

2012
[10]

Ambrosio, N

L. Ambrosio, N. Gigli, G. Savaré, Gradient Flows in Metric Spaces and in the Space of Probability Measures, 2nd Edition, Birkhäuser, 2008

2008
[11]

H. L. Van Trees, Optimum Array Processing: Part IV of Detection, Estimation, and Modulation Theory, Wiley, 2002

2002
[12]

Ledoit, M

O. Ledoit, M. Wolf, A well-conditioned estimator for large- dimensional covariance matrices, Journal of Multivari- ate Analysis 88 (2) (2004) 365–411. doi:10.1016/S0047- 259X(03)00096-4

2004 doi
[13]

Y . Chen, A. Wiesel, Y . C. Eldar, A. O. Hero, Shrinkage algorithms for MMSE covariance estimation, IEEE Trans- actions on Signal Processing 58 (10) (2010) 5016–5029. doi:10.1109/TSP.2010.2053029

2010
[14]

Bhatia, Positive Definite Matrices, Princeton University Press, 2009

R. Bhatia, Positive Definite Matrices, Princeton University Press, 2009

2009
[15]

R. A. Horn, C. R. Johnson, Topics in Matrix Analysis, Cambridge University Press, 1991

1991
[16]

V . I. Arnold, Mathematical Methods of Classical Mechan- ics, 2nd Edition, Springer, 1989

1989
[17]

Goldstein, C

H. Goldstein, C. P. Poole, J. L. Safko, Classical Mechanics, 3rd Edition, Addison-Wesley, 2002

2002
[18]

S. Boyd, L. Vandenberghe, Convex Optimization, Cam- bridge University Press, 2004

2004
[19]

Dutta, N

Arvind, B. Dutta, N. Mukunda, R. Simon, The real sym- plectic groups in quantum mechanics and optics, Pramana 45 (6) (1995) 471–497

1995
[20]

Adesso, S

G. Adesso, S. Ragy, A. R. Lee, Continuous variable quan- tum information: Gaussian states and beyond, Open Sys- tems & Information Dynamics 21 (2014) 1440001

2014
[21]

Yurke, S

B. Yurke, S. L. McCall, J. R. Klauder, Su(2) and su(1,1) interferometers, Physical Review A 33 (6) (1986) 4033– 4054

1986
[22]

J. A. López-Saldívar, M. A. Man’ko, V . I. Man’ko, Differ- ential parametric formalism for the evolution of gaussian states: Nonunitary evolution and invariant states, Entropy 22 (5) (2020). doi:10.3390/e22050586

2020 doi
[23]

J. A. López-Saldívar, Differential formalism and the thermodynamic description of multimode gaus- sian equilibrium states, Physica A: Statistical Me- chanics and its Applications 617 (2023) 128676. doi:https://doi.org/10.1016/j.physa.2023.128676

2023
[24]

Weedbrook, S

C. Weedbrook, S. Pirandola, R. García-Patrón, N. J. Cerf, T. C. Ralph, J. H. Shapiro, S. Lloyd, Gaussian quantum information, Reviews of Modern Physics 84 (2) (2012) 621–669. doi:10.1103/RevModPhys.84.621

2012 doi
[25]

Simon, Peres–horodecki separability criterion for con- tinuous variable systems, Physical Review Letters 84 (12) (2000) 2726–2729

R. Simon, Peres–horodecki separability criterion for con- tinuous variable systems, Physical Review Letters 84 (12) (2000) 2726–2729. doi:10.1103/PhysRevLett.84.2726

2000 doi
[26]

R. F. Werner, M. M. Wolf, Bound entangled gaussian states, Physical Review Letters 86 (16) (2001) 3658–3661. doi:10.1103/PhysRevLett.86.3658. 4

2001 doi

Pith tools

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

[1] [1]

C. R. Givens, R. M. Shortt, A class of wasserstein metrics for probability distributions, The Michigan Mathematical Journal 31 (2) (1984) 231–240

1984

[2] [2]

Takatsu, Wasserstein geometry of gaussian measures, Osaka J

A. Takatsu, Wasserstein geometry of gaussian measures, Osaka J. Math. 48 (2011) 1005–1026

2011

[3] [3]

Villani, Optimal Transport: Old and New, Springer, 2009

C. Villani, Optimal Transport: Old and New, Springer, 2009

2009

[4] [4]

Malagò, L

L. Malagò, L. Montrucchio, G. Pistone, Wasserstein rie- mannian geometry of gaussian densities, Information Ge- ometry 1 (2018) 137–179

2018

[5] [5]

Bhatia, T

R. Bhatia, T. Jain, Y . Lim, On the bures–wasserstein dis- tance between positive definite matrices, Expositiones Mathematicae 37 (2) (2019) 165–191

2019

[6] [6]

Bengtsson, K

I. Bengtsson, K. ˙Zyczkowski, Geometry of Quantum States: An Introduction to Quantum Entanglement, 2nd Edition, Cambridge University Press, 2017

2017

[7] [7]

Jordan, D

R. Jordan, D. Kinderlehrer, F. Otto, The variational formu- lation of the fokker–planck equation, SIAM J. Math. Anal. 29 (1998) 1–17

1998

[8] [8]

Otto, The geometry of dissipative evolution equations: the porous medium equation, Comm

F. Otto, The geometry of dissipative evolution equations: the porous medium equation, Comm. Partial Differential Equations 26 (2001) 101–174

2001