Approach to equilibrium for a particle interacting with a harmonic thermal bath

Alberto Mario Maiocchi; Federico Bonetto

arxiv: 2510.20003 · v2 · submitted 2025-10-22 · ❄️ cond-mat.stat-mech · math-ph· math.MP

Approach to equilibrium for a particle interacting with a harmonic thermal bath

Federico Bonetto , Alberto Mario Maiocchi This is my paper

Pith reviewed 2026-05-18 04:13 UTC · model grok-4.3

classification ❄️ cond-mat.stat-mech math-phmath.MP

keywords harmonic oscillatorheat baththermalizationcorrelation functionstochastic thermostatfrequency spectrumapproach to equilibriumweak coupling

0 comments

The pith

A probe harmonic oscillator coupled to a large oscillator chain reaches the bath temperature in its position variance when its natural frequency lies inside the bath spectrum.

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

The paper follows the long-time position-position correlation of a single oscillator initially at temperature T_P that is weakly coupled by strength α to a chain of N oscillators at temperature T_B. For times scaled with N the correlation approaches a well-defined infinite-bath limit. When the probe frequency Ω sits inside the bath spectrum this limit reproduces, at order zero in α, the exact statistics of an oscillator driven by white noise and friction, so that the variance settles to T_B/Ω² and the time-difference correlation decays exponentially. At higher orders in α the same limit still contains oscillatory or power-law tails in the time separation, showing that the bath never becomes a memoryless thermostat.

Core claim

In the N-to-infinity limit the correlation C_α(s,t) equals the stochastic-thermostat correlation at leading order in α whenever Ω belongs to the bath spectrum, yielding lim t→∞ C_α(t,t) = T_B/Ω² together with exponential decay in the time lag; higher-order terms in α nevertheless retain oscillations or algebraic decay in |t-s|. When Ω lies outside the spectrum the variance does not approach the bath value and no exponential relaxation appears.

What carries the argument

The N→∞ limit of the position-position correlation C_α(s,t) evaluated at times t and s proportional to N, which encodes both the leading white-noise behavior and the persistent non-Markovian corrections.

If this is right

The probe variance reaches the bath value T_B/Ω² in the long-time limit when frequencies overlap.
At leading order the time-lagged correlation decays exponentially, reproducing Markovian Langevin dynamics.
Persistent oscillations and slow algebraic tails survive at higher orders in coupling even for an infinite bath.
No relaxation to bath temperature occurs when the probe frequency lies outside the bath spectrum.

Where Pith is reading between the lines

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

Frequency resonance appears necessary for effective dissipation in purely harmonic systems.
Simulations that aim to mimic a thermostat must ensure spectral overlap in addition to large bath size.
The same mechanism may govern how weakly coupled harmonic subsystems exchange energy in larger networks.

Load-bearing premise

The finite-N chain can be replaced by its infinite-size limit for times proportional to N, producing an effective white-noise drive already at the lowest order in the coupling strength.

What would settle it

Numerically integrate the finite but large-N chain equations for a probe frequency inside the bath band, extract the long-time variance of the probe position, and verify whether it approaches T_B/Ω² while the time-lag correlation still exhibits slow power-law or oscillatory remnants at moderate α.

read the original abstract

We study the long time evolution of the position-position correlation function $C_{\alpha,N}(s,t)$ for a harmonic oscillator (the {\it probe}) interacting via a coupling $\alpha$ with a large chain of $N$ coupled oscillators (the {\it heat bath}). At $t=0$ the probe and the bath are in equilibrium at temperature $T_P$ and $T_B$, respectively. We show that for times $t$ and $s$ of the order of $N$, $C_{\alpha,N}(s,t)$ is very well approximated by its limit $C_{\alpha}(s,t)$ as $N\to\infty$. We find that, if the frequency $\Omega$ of the probe is in the spectrum of the bath, the system appears to thermalize, at least at higher order in $\alpha$. This means that, at order 0 in $\alpha$, $C_\alpha(s,t)$ equals the correlation of a probe in contact with an ideal stochastic {\it thermostat}, that is forced by a white noise and subject to dissipation. In particular we find that $\lim_{t\to\infty} C_\alpha(t,t)=T_B/\Omega^2$ while that $\lim_{\tau\to\infty} C_\alpha(\tau,\tau+t)$ exists and decays exponentially in $t$. Notwithstanding this, at higher order in $\alpha$, $C_{\alpha}(s,t)$ contains terms that oscillate or vanish as a power law in $|t-s|$. That is, even when the bath is very large, it cannot be thought of as a stochastic thermostat. When the frequency of the bath is far from the spectrum of the bath, no thermalization is observed.

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.

Desk Editor's Note private letter to a colleague

The paper shows that a harmonic probe reaches bath temperature in the long-time limit only at leading order in coupling when its frequency sits in the bath spectrum, while higher orders produce persistent oscillations or power-law tails that survive the N to infinity limit.

read the letter

The main takeaway is that for times scaled with N, the position-position correlation of the probe approaches the ideal stochastic thermostat result at order zero in the coupling α, including the equilibrium variance T_B/Ω² and exponential decay of the stationary part. Higher orders in α, however, add terms that oscillate or decay as a power law in the time difference, so the bath never becomes a perfect white-noise thermostat even after N goes to infinity. When the probe frequency lies outside the bath spectrum, thermalization fails altogether. This is worked out explicitly for the linear chain model by replacing the finite-N correlation with its infinite-N version at scaled times and then analyzing the resulting function order by order in α. The calculation is straightforward from the normal-mode solution of the coupled oscillators, and it gives a clean, solvable example of how the stochastic approximation works and where it breaks. That is the useful part: it pins down the regime of validity for a common modeling assumption in classical statistical mechanics. The soft spot is the uniformity of the N to infinity approximation. The argument takes the limit at times of order N and then lets the observation time go to infinity inside that limit. If the error between the finite-N and infinite-N correlations grows with the total time, even mildly, the long-time claims could be affected by residual mode discreteness. The stress-test note flags exactly this ordering issue, and without explicit remainder bounds that hold uniformly for all s, t up to order N, it is not obvious that the exponential decay survives at finite but large N. Since the system is fully linear, such bounds should be derivable from the mode sums, but they need to be stated clearly. Readers working on rigorous derivations of effective stochastic dynamics or on the foundations of thermalization in harmonic systems will find this worth reading. It is a concrete, checkable calculation rather than a general claim, so the results are reproducible in principle. I would send it to peer review; the core limits are stated sharply enough that referees can verify the error control and the higher-order corrections without much trouble.

Referee Report

2 major / 2 minor

Summary. The paper studies the long-time evolution of the position-position correlation function C_{α,N}(s,t) for a harmonic probe oscillator coupled with strength α to a finite chain of N oscillators as a heat bath, with the probe initially at temperature T_P and the bath at T_B. It claims that for times t and s of order N, the finite-N correlation is well approximated by the N→∞ limit C_α(s,t). When the probe frequency Ω lies in the bath spectrum, the system appears to thermalize at higher order in α: at leading order in α, C_α(s,t) matches that of a probe coupled to an ideal stochastic thermostat (white noise plus dissipation), yielding lim_{t→∞} C_α(t,t) = T_B/Ω² and exponential decay of the stationary correlation lim_{τ→∞} C_α(τ,τ+t). Higher-order terms in α introduce oscillations or power-law decay in |t-s|. No thermalization occurs when Ω is outside the bath spectrum.

Significance. If the central limits hold with the stated uniformity, the work provides a concrete, solvable example of how a large but discrete harmonic bath induces effective thermalization at leading order in coupling while retaining non-Markovian corrections at higher orders. The explicit scaling with N and the distinction between order-0 thermostat behavior and persistent bath-induced oscillations or power laws are useful for clarifying the validity of stochastic-thermostat approximations in harmonic systems. The focus on verifiable correlation limits and the condition that Ω must lie in the bath spectrum adds falsifiable structure to the analysis.

major comments (2)

[Abstract and the derivation of the N→∞ limit] The central thermalization claims (lim_{t→∞} C_α(t,t) = T_B/Ω² and exponential decay of the stationary correlation) rest on first replacing C_{α,N}(s,t) by its N→∞ limit for t,s = O(N) and then taking t→∞ inside that limit. The manuscript does not supply explicit remainder bounds on |C_{α,N}(s,t) - C_α(s,t)| that are uniform for all s,t = O(N) (including large |t-s| or min(s,t)). Without such bounds independent of observation time, residual discrete-mode oscillations arising from the ~1/N spacing of bath normal modes could survive at finite N and only vanish after an additional N→∞ limit taken after t→∞, reversing the order of limits used in the argument.
[Abstract (thermalization at order 0 in α)] The statement that 'at order 0 in α, C_α(s,t) equals the correlation of a probe in contact with an ideal stochastic thermostat' is load-bearing for the thermalization interpretation, yet the manuscript provides no explicit derivation or error estimate showing how the leading-order term is obtained from the underlying linear oscillator equations or how higher-order corrections in α are controlled uniformly in time.

minor comments (2)

[Abstract] Notation for the two-time correlation C_α(s,t) and the stationary limit lim_{τ→∞} C_α(τ,τ+t) should be defined more explicitly in the main text to avoid ambiguity when discussing the order of limits.
[Introduction or methods] The manuscript would benefit from a brief statement of the precise sense in which the N→∞ approximation holds (e.g., pointwise, in L², or uniformly on compact sets) before the long-time limits are taken.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thorough review and insightful comments on our manuscript. The points raised regarding the uniformity of the N to infinity limit and the explicit control of the leading-order term in alpha are important for clarifying the scope of our results. We address each major comment below and indicate where revisions will be made to strengthen the presentation.

read point-by-point responses

Referee: [Abstract and the derivation of the N→∞ limit] The central thermalization claims (lim_{t→∞} C_α(t,t) = T_B/Ω² and exponential decay of the stationary correlation) rest on first replacing C_{α,N}(s,t) by its N→∞ limit for t,s = O(N) and then taking t→∞ inside that limit. The manuscript does not supply explicit remainder bounds on |C_{α,N}(s,t) - C_α(s,t)| that are uniform for all s,t = O(N) (including large |t-s| or min(s,t)). Without such bounds independent of observation time, residual discrete-mode oscillations arising from the ~1/N spacing of bath normal modes could survive at finite N and only vanish after an additional N→∞ limit taken after t→∞, reversing the order of limits used in the argument.

Authors: We agree that a more explicit discussion of remainder estimates would improve clarity. In the manuscript the N→∞ limit is obtained by writing the exact solution for the finite-N chain in terms of its normal modes, passing to the continuum limit for the bath spectral density, and verifying that the difference vanishes pointwise for each fixed s,t with t,s = O(N). The frequency spacing of O(1/N) produces oscillations whose amplitude is controlled by the smoothness of the initial data and the spectral density; these contributions are shown to average to zero in the integrated sense relevant for the long-time limits. While we do not supply fully uniform-in-time remainder bounds with explicit constants, the explicit mode expansion already demonstrates that any surviving discrete-mode effects are suppressed by the same 1/N factor that defines the limit. We will add a dedicated paragraph in the revised manuscript discussing the order of limits and the justification for interchanging N→∞ and t→∞ within the stated regime. revision: partial
Referee: [Abstract (thermalization at order 0 in α)] The statement that 'at order 0 in α, C_α(s,t) equals the correlation of a probe in contact with an ideal stochastic thermostat' is load-bearing for the thermalization interpretation, yet the manuscript provides no explicit derivation or error estimate showing how the leading-order term is obtained from the underlying linear oscillator equations or how higher-order corrections in α are controlled uniformly in time.

Authors: The leading-order (order-zero in α) term is derived in Section 3 by solving the coupled linear system and isolating the contribution in which the probe feels the bath only through the effective friction and noise kernels obtained from the bath Green's function in the N→∞ limit. At this order the back-action of the probe on the bath is neglected, yielding precisely the stationary Ornstein–Uhlenbeck process whose position autocorrelation is T_B/Ω² and whose time correlations decay exponentially with the damping rate fixed by the bath spectral density at Ω. Higher-order terms in α are generated by the iterative solution of the integral equations and introduce the memory kernel whose Fourier transform produces the persistent oscillations or power-law tails. We will expand the derivation of the order-zero term with an explicit error estimate (O(α)) that holds uniformly on compact time intervals and indicate how the uniformity extends to the long-time regime under the spectral condition that Ω lies inside the bath continuum. revision: yes

Circularity Check

0 steps flagged

Derivation from linear oscillator equations is self-contained

full rationale

The paper computes the correlation C_α(s,t) explicitly from the linear system of coupled harmonic oscillators in the N→∞ limit, then extracts the stated limits and exponential decay directly from the resulting integro-differential equation and the bath spectral density. No parameters are fitted to data and then relabeled as predictions, no self-citation supplies a load-bearing uniqueness theorem, and the central claims do not reduce to a redefinition of the inputs. The derivation remains independent of the target thermalization statements.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The analysis rests on the exact solvability of linear harmonic systems, the thermodynamic limit taken with times scaled to system size N, and the assumption that the bath chain possesses a continuous spectrum overlapping the probe frequency.

free parameters (2)

coupling strength α
Treated as a small parameter for perturbative expansion in interaction strength; its value is not fitted to data but controls the order of corrections.
probe frequency Ω
Chosen relative to the bath spectrum to decide whether thermalization occurs.

axioms (2)

domain assumption All oscillators are harmonic and the bath is a finite chain of N coupled oscillators with a well-defined frequency spectrum.
This defines the microscopic model whose correlation functions are computed.
domain assumption Initial state: probe in equilibrium at temperature T_P, bath at T_B, with times t,s scaled as order N.
Sets the nonequilibrium starting point and the long-time regime under study.

pith-pipeline@v0.9.0 · 5849 in / 1512 out tokens · 43955 ms · 2026-05-18T04:13:44.695789+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

16 extracted references · 16 canonical work pages

[1]

Rieder, J

Z. Rieder, J. L. Lebowitz, and E. Lieb. Properties of a harmonic crystal in a stationary nonequi- librium state.Journal of Mathematical Physics, 8(5):1073–1078, 05 1967

work page 1967
[2]

The Kac model coupled to a thermo- stat.J

Federico Bonetto, Michael Loss, and Ranjini Vaidyanathan. The Kac model coupled to a thermo- stat.J. Stat. Phys., 156(4):647–667, 2014

work page 2014
[3]

D. J. Evans and B. L. Holian. The nose–hoover thermostat.The Journal of Chemical Physics, 83(8):4069–4074, 10 1985

work page 1985
[4]

Chernov, G.L

N. Chernov, G.L. Eyink, J.L. Lebowitz, and Ya.G. Sinai. Steady state electric conductivity in the periodic lorentz gas.Comm. Math. Phys., 154:569–601, 1993

work page 1993
[5]

Bonetto, D

F. Bonetto, D. Daems, J.L. Lebowitz, and V. Ricci. Properties of stationary nonequilibrium states in the thermostatted periodic Lorentz gas: The multiparticle system.Phys. Rev E, 65:05124, 2002

work page 2002
[6]

Eckman, C.A

J.P. Eckman, C.A. Pillet, and L. Rey-Bellet. Non-equilibrium statistical mechanics of anharmonic chains coupled to two heat baths at different temperatures.Comm. Math. Phys., 201:657–697, 1999

work page 1999
[7]

Uniform approx- imation of a Maxwellian thermostat by finite reservoirs.Comm

Federico Bonetto, Michael Loss, Hagop Tossounian, and Ranjini Vaidyanathan. Uniform approx- imation of a Maxwellian thermostat by finite reservoirs.Comm. Math. Phys., 351(1):311–339, 2017

work page 2017
[8]

Toda and Y

M. Toda and Y. Kogure. Statistical dynamics of systems of interacting oscillators.Supplement of the Progress ot Theoretical Physics, 23:157–171, 1962

work page 1962
[9]

Takeno and J

S. Takeno and J. Hori. Continuum approximation for the motion of a heavy particle in one- and three-dimensional lattices.Supplement of the Progress ot Theoretical Physics, 23:177–184, 1962

work page 1962
[10]

G. W. Ford, M. Kac, and P. Mazur. Statistical mechanics of assemblies of coupled oscillators.J. Math. Phys., 6:504–515, 1965. 33

work page 1965
[11]

Davis and Philip Rabinowitz.Methods of Numerical Integration

Philip J. Davis and Philip Rabinowitz.Methods of Numerical Integration. Elsevier, 1984

work page 1984
[12]

Benettin, A

G. Benettin, A. Carati, and G. Gallavotti. A rigorous implementation of the Jeans-Landau-Teller approximation for adiabatic invariants.Nonlinearity, 10:479–505, 1995

work page 1995
[13]

Carati, P

A. Carati, P. Cipriani, and L. Galgani. On the definition of temperature in FPU systems.Journ, Stat. Phys., 115:1119–1130, 2004

work page 2004
[14]

Maiocchi

A.M. Maiocchi. Freezing of the optical–branch energy in a diatomic FPU chain.Comm. Math. Phys., 372:91–117, 2019. A Technical Lemmas In this appendix we collect few results that are used many times in the course of the paper. As discussed in Remark 5.2, we want to compute (5.4) by shifting the integral from a line with real partξ >0 to a line with real pa...

work page 2019
[15]

the product ofC 1 with its complex conjugate where the denominator is (ω 2 −Ω 2 +(α))(ω2 − Ω∗ +(α) 2): this gives rise to an oscillating term in (t−s), exponentially decreasing as Kα −2e−ξ(α)(t+s). 38

work page
[16]

Concerning the product ofC 3 withC 2 or its complex conjugate, we can always apply Lemma A.1 to bound the contribution asK/ p min (t, s)

the product ofC 1 timesC ∗ 3 (orC ∗ 1 timesC 3), where the denominators (ω 2 −Ω 2 +(α))Dα(−iω+ 0+) or (ω2 − Ω∗ +(α) 2)Dα(iω+ 0 +) appear: here we get a term bounded byKα −2e−ξ(α) min (t,s). Concerning the product ofC 3 withC 2 or its complex conjugate, we can always apply Lemma A.1 to bound the contribution asK/ p min (t, s). The same applies to the produ...

work page

[1] [1]

Rieder, J

Z. Rieder, J. L. Lebowitz, and E. Lieb. Properties of a harmonic crystal in a stationary nonequi- librium state.Journal of Mathematical Physics, 8(5):1073–1078, 05 1967

work page 1967

[2] [2]

The Kac model coupled to a thermo- stat.J

Federico Bonetto, Michael Loss, and Ranjini Vaidyanathan. The Kac model coupled to a thermo- stat.J. Stat. Phys., 156(4):647–667, 2014

work page 2014

[3] [3]

D. J. Evans and B. L. Holian. The nose–hoover thermostat.The Journal of Chemical Physics, 83(8):4069–4074, 10 1985

work page 1985

[4] [4]

Chernov, G.L

N. Chernov, G.L. Eyink, J.L. Lebowitz, and Ya.G. Sinai. Steady state electric conductivity in the periodic lorentz gas.Comm. Math. Phys., 154:569–601, 1993

work page 1993

[5] [5]

Bonetto, D

F. Bonetto, D. Daems, J.L. Lebowitz, and V. Ricci. Properties of stationary nonequilibrium states in the thermostatted periodic Lorentz gas: The multiparticle system.Phys. Rev E, 65:05124, 2002

work page 2002

[6] [6]

Eckman, C.A

J.P. Eckman, C.A. Pillet, and L. Rey-Bellet. Non-equilibrium statistical mechanics of anharmonic chains coupled to two heat baths at different temperatures.Comm. Math. Phys., 201:657–697, 1999

work page 1999

[7] [7]

Uniform approx- imation of a Maxwellian thermostat by finite reservoirs.Comm

Federico Bonetto, Michael Loss, Hagop Tossounian, and Ranjini Vaidyanathan. Uniform approx- imation of a Maxwellian thermostat by finite reservoirs.Comm. Math. Phys., 351(1):311–339, 2017

work page 2017

[8] [8]

Toda and Y

M. Toda and Y. Kogure. Statistical dynamics of systems of interacting oscillators.Supplement of the Progress ot Theoretical Physics, 23:157–171, 1962

work page 1962

[9] [9]

Takeno and J

S. Takeno and J. Hori. Continuum approximation for the motion of a heavy particle in one- and three-dimensional lattices.Supplement of the Progress ot Theoretical Physics, 23:177–184, 1962

work page 1962

[10] [10]

G. W. Ford, M. Kac, and P. Mazur. Statistical mechanics of assemblies of coupled oscillators.J. Math. Phys., 6:504–515, 1965. 33

work page 1965

[11] [11]

Davis and Philip Rabinowitz.Methods of Numerical Integration

Philip J. Davis and Philip Rabinowitz.Methods of Numerical Integration. Elsevier, 1984

work page 1984

[12] [12]

Benettin, A

G. Benettin, A. Carati, and G. Gallavotti. A rigorous implementation of the Jeans-Landau-Teller approximation for adiabatic invariants.Nonlinearity, 10:479–505, 1995

work page 1995

[13] [13]

Carati, P

A. Carati, P. Cipriani, and L. Galgani. On the definition of temperature in FPU systems.Journ, Stat. Phys., 115:1119–1130, 2004

work page 2004

[14] [14]

Maiocchi

A.M. Maiocchi. Freezing of the optical–branch energy in a diatomic FPU chain.Comm. Math. Phys., 372:91–117, 2019. A Technical Lemmas In this appendix we collect few results that are used many times in the course of the paper. As discussed in Remark 5.2, we want to compute (5.4) by shifting the integral from a line with real partξ >0 to a line with real pa...

work page 2019

[15] [15]

the product ofC 1 with its complex conjugate where the denominator is (ω 2 −Ω 2 +(α))(ω2 − Ω∗ +(α) 2): this gives rise to an oscillating term in (t−s), exponentially decreasing as Kα −2e−ξ(α)(t+s). 38

work page

[16] [16]

Concerning the product ofC 3 withC 2 or its complex conjugate, we can always apply Lemma A.1 to bound the contribution asK/ p min (t, s)

the product ofC 1 timesC ∗ 3 (orC ∗ 1 timesC 3), where the denominators (ω 2 −Ω 2 +(α))Dα(−iω+ 0+) or (ω2 − Ω∗ +(α) 2)Dα(iω+ 0 +) appear: here we get a term bounded byKα −2e−ξ(α) min (t,s). Concerning the product ofC 3 withC 2 or its complex conjugate, we can always apply Lemma A.1 to bound the contribution asK/ p min (t, s). The same applies to the produ...

work page