betaFeedback

WorksPhysics and Astronomy

Thermodynamic Invariants of Coupled Channels: A Many-Channel Tolman-Ehrenfest Effect

Abstract

When multiple thermodynamic channels are coupled, single-channel equilibrium conditions fail. Extending the Tolman–Ehrenfest effect to the entropy manifold, we derive the unique nn-channel invariant ζiTi=Cζ_i T_i = C, where ζiζ_i is the holonomy of the Ruppeiner connection. For the granular volume–stress ensemble, Rowe's dilatancy ratio and energy restriction emerge as geometric consequences of the off-diagonal curvature gVσg_{Vσ}, and the 60-year puzzle of state-dependent KμK_μ is resolved: the correction ζVζ_V reaches O(1)O(1) near jamming. The prediction ζVχ=constζ_Vχ=\mathrm{const} across a shear band is experimentally testable.

Introduction

Classical thermodynamics achieves its economy by assuming the entropy surface is effectively diagonal: channels equilibrate independently. For dense granular media, reactive geophysical fluids, and active matter [1, 2], coupling between thermodynamic channels is the dominant physics. Existing frameworks include Blumenfeld–Edwards granular statistical mechanics [3, 4], gauge-theoretic thermodynamics [5], and the cross-diffusion reduction of Biktashev and Tsyganov [6]. These approaches identify the need for coupling corrections but do not derive the invariant that coupled channels share at equilibrium.

The Ruppeiner metric [7, 8], gij=−∂2S/∂qi∂qjg_{ij} = -\partial^2 S/\partial q^i \partial q^j, quantifies this failure: gijg_{ij} (i≠ji \ne j) measures how much injecting qjq^j shifts the conjugate intensity TiT_i at fixed qiq^i, so that ∂βi/∂qj=−gij\partial\beta_i/\partial q^j = -g_{ij}. It is the precision matrix of equilibrium fluctuations and diverges at critical points [8]. The Ruppeiner and Weinhold [9] metrics have been applied extensively to phase transitions, black hole thermodynamics, and information-geometric characterisation of thermodynamic systems [10, 11]; what has not been identified in this literature is the specific invariant that replaces the bare intensive variable when channels are coupled. In the relativistic Tolman–Ehrenfest (TE) effect [12] the conserved quantity in a gravitational field is the redshifted energy, forcing T(1+ϕ/c2)=const⁡T(1+\phi/c^2) = \operatorname{const}; here inter-channel coupling plays the role of the gravitational potential, and the geometric weight ζi\zeta_i is computed from the entropy-surface curvature.

MCTE relation

Let a body B\mathcal{B} possess nn thermodynamic channels with extensive variables q=(q1,…,qn)\mathbf{q}=(q^1,\ldots,q^n) and conjugate intensities βi=∂S/∂qi\beta_i=\partial S/\partial q^i, Ti=βi−1T_i=\beta_i^{-1}. The correct multi-channel equilibrium condition must be a first integral of the level-set flow on the entropy manifold: a function of the intensities {βi}\{\beta_i\} that is constant along every trajectory of the flow induced by the constraint S=const⁡S=\operatorname{const}.

On the level set dS=∑iβi dqi=0dS=\sum_i\beta_i\,dq^i=0, the Ruppeiner metric ∂βi/∂qj=−gij\partial\beta_i/\partial q^j=-g_{ij} drives the coupled flow

dβidqj∣dS=0=giiβj−gijβiβi,j≠i.(1)\left.\frac{d\beta_i}{dq^j}\right|_{dS=0} = \frac{g_{ii}\beta_j-g_{ij}\beta_i}{\beta_i}, \qquad j\ne i. \tag*{(1)}

A first integral F(β1,…,βn)F(\beta_1,\ldots,\beta_n) of Eq. (1) must satisfy ∇βF⋅β˙=0\nabla_\beta F\cdot\dot{\beta}=0 along every level-set trajectory for all gijg_{ij}. The unique solution of this integrability condition, derived in SM1 (Supplemental Material) by separation of variables on the level-set ODE is F=ζiβi−1F=\zeta_i\beta_i^{-1} for all ii, where ζi\zeta_i satisfies

dlog⁡ζi=∑j≠igiiβj−gijβiβi2 dqj=:ωi.(2)d\log\zeta_i=\sum_{j\ne i}\frac{g_{ii}\beta_j-g_{ij}\beta_i}{\beta_i^2}\,dq^j=:\omega_i. \tag*{(2)}

Setting F=CF=C then gives the MCTE relation

ζiTi=Con B,i=1,…,n,(3)\zeta_i T_i=C\quad\text{on }\mathcal{B},\quad i=1,\ldots,n, \tag*{(3)}

where CC is a single scalar invariant across all channels and all points: the multi-channel generalisation of the relativistic Tolman temperature T/g00T/\sqrt{g_{00}}. The multiplicative effective energy Eeff=∫ζaea dvE_{\rm eff}=\int\zeta_a e_a\,dv whose conservation enforces Eq. (3) is not assumed but follows from the first-integral requirement (see SM1 in Supplemental Material [13]). Equation (3) is also the Euler–Lagrange equation of L[ea]=∫s dv−β∫ζaea dv\mathcal{L}[e_a]=\int s\,dv-\beta\int\zeta_a e_a\,dv, confirming C=β−1C=\beta^{-1} (see SM2 in Supplemental Material [13]).

Coupling coefficient as Ruppeiner holonomy

The MCTE condition requires dlog⁡ζi=dlog⁡βid\log\zeta_i=d\log\beta_i along every entropy level set. The one-form ωi\omega_i defined in Eq. (2) integrates to give the explicit coupling coefficient

ζi(q)=λiexp⁡(∫q0qωi),(4)\zeta_i(\mathbf{q})=\lambda_i\exp\left(\int_{\mathbf{q}_0}^{\mathbf{q}}\omega_i\right), \tag*{(4)}

with λi\lambda_i fixed by a single boundary condition. When gij=0g_{ij}=0 for all i≠ji\ne j and giig_{ii} is position-independent, ζi=λi\zeta_i=\lambda_i (constant) and Eq. (3) reduces to the classical Ti=const⁡T_i=\operatorname{const}.

The one-form ωi\omega_i is the projection of the logarithmic Jacobian of the intensive-variable map onto the entropy level set. The integral ∮ωi\oint\omega_i around a closed loop is the holonomy of the Ruppeiner connection: it vanishes for a flat metric (decoupled channels) and is non-zero whenever inter-channel coupling is present, measuring the “thermodynamic twist” accumulated along a closed equilibrium path. In the Tolman–Ehrenfest analogy, ζi\zeta_i corresponds to g00\sqrt{g_{00}}, which is the norm of the time-translation Killing vector; here it is the norm of the vector field generating translations along effective-energy level sets on the entropy manifold.

Three-panel schematic of the MCTE framework

Figure 1. Three-panel schematic of the MCTE framework (column width). (a) Classical (gij=0g_{ij}=0): entropy level sets are straight lines; each intensity TiT_i is independently uniform. (b) MCTE (gij≠0g_{ij}\ne0): level sets curve; a closed loop accumulates a non-trivial Ruppeiner holonomy ∮ωi≠0\oint\omega_i\ne0 (red), and the coupling coefficient ζV\zeta_V grows in the direction of increasing off-diagonal curvature (orange). The unique first integral of this flow is ζiTi=C\zeta_iT_i=C. (c) ζV\zeta_V as a function of packing fraction ϕ\phi: unity in the dilute limit, O(1)O(1) departure at jamming ϕJ\phi_J, explaining the observed rise in Rowe’s KμK_\mu.

Granular realisation.—In Edwards’ statistical mechanics [4], the role of temperature is played by compactivity χ\chi (χ−1=∂S/∂V\chi^{-1}=\partial S/\partial V). The Blumenfeld–Edwards extension [3] introduces angoricity A\mathcal{A} (Aij=∂S/∂σij\mathcal{A}_{ij}=\partial S/\partial\sigma_{ij}) conjugate to the stress σ\boldsymbol{\sigma}, yielding the two-channel (N=2N=2) entropy surface S(V,σ)S(V,\boldsymbol{\sigma}) with block Ruppeiner metric

g=−(∂2S/∂V2∂2S/∂V∂σij∂2S/∂σij∂V∂2S/∂σij∂σkl).(5)g=-\begin{pmatrix} \partial^2 S/\partial V^2 & \partial^2 S/\partial V\partial\sigma_{ij}\\ \partial^2 S/\partial\sigma_{ij}\partial V & \partial^2 S/\partial\sigma_{ij}\partial\sigma_{kl} \end{pmatrix}. \tag*{(5)}

Remark 1 (System Parity). Although this paper addresses an even-parity system for illustration, where the skew-symmetric Onsager coupling block has no zero eigenvalue and inter-channel exchange is confined to closed reversible orbits, the general formalism is parity-agnostic. The N=2N=2 volume–stress granular ensemble is the canonical realisation; subtleties specific to odd-parity (N=3,5,…N=3,5,\ldots) systems, in which a topological zero mode opens an irreversibility mechanisms and drives non-associated plastic flow, will be treated in upcoming works. Crucially, the N=2N=2 MCTE equilibrium constitutes the thermodynamic background state upon which the N=3N=3 instability grows: shear-band nucleation requires a N=3N=3 realisation, but its energetics are set by the N=2N=2 Stable Layer from which it departs.

The off-diagonal block gVσij=−∂χ−1/∂σijg_{V\sigma_{ij}}=-\partial\chi^{-1}/\partial\sigma_{ij} measures how much the inverse compactivity shifts when stress is varied at fixed volume. It vanishes in the dilute limit and grows large near the jamming transition. A sign subtlety merits emphasis: because gij=−∂2S/∂qi∂qjg_{ij}=-\partial^2S/\partial q^i\partial q^j, the off-diagonal entry gVσg_{V\sigma} is negative when the cross-derivative ∂2S/∂V∂σ\partial^2S/\partial V\partial\sigma is positive (stress-softening coupling). Computing gVσg_{V\sigma} with the wrong sign breaks the MCTE invariance ζVχ=C\zeta_V\chi=C; in the toy model of the Supplemental Material [13], this error causes ζ1T1\zeta_1T_1 to drift by O(1)O(1) along the level set even though SS itself remains constant to machine precision, providing a stringent self-consistency diagnostic (see Supplemental Material, Section S10). The correct equilibrium condition is ζVχ=C\zeta_V\chi=C, not χ=const\chi=\mathrm{const}.

Applying Eq. (4) to the (V,σ)(V,\sigma) system and projecting onto the normal direction n^\hat{\boldsymbol{n}} gives the scalar coupling ODE

dlog⁡ζV=gVVA−1−gVσχ−1χ−2 dσ,(6)\mathrm{d}\log\zeta_V=\frac{g_{V V}\mathcal{A}^{-1}-g_{V\sigma}\chi^{-1}}{\chi^{-2}}\,\mathrm{d}\sigma, \tag*{(6)}

where gVσ≡n^in^jgVσijg_{V\sigma}\equiv\hat n_i\hat n_jg_{V\sigma_{ij}} and A≡n^in^jAij\mathcal{A}\equiv\hat n_i\hat n_j\mathcal{A}_{ij} are the normal projections of the off-diagonal curvature and angoricity respectively. The fractional correction to the single-channel condition is

∣∇ζVζV∣∼∣gVσgVV⋅χA∣,(7)\left|\frac{\nabla\zeta_V}{\zeta_V}\right|\sim\left|\frac{g_{V\sigma}}{g_{VV}}\cdot\frac{\chi}{\mathcal{A}}\right|, \tag*{(7)}

with gVσg_{V\sigma} and A\mathcal{A} the normal projections defined above. This correction is doubly suppressed in the dilute limit (gVσ→0g_{V\sigma}\to0, χ/A→0\chi/\mathcal{A}\to0 simultaneously) and becomes O(1)O(1) near isostaticity: precisely where the coupled Blumenfeld–Edwards description is most needed, imposing χ\chi and A\mathcal{A} independently introduces an O(1)O(1) systematic error. (7) is the first quantitative geometric estimate of this failure. A concrete evaluation on a toy entropy surface S(V,σ)=aln⁡(V−VJ)+bln⁡(σmax⁡−σ)−cσ/(V−VJ)S(V,\sigma)=a\ln(V-V_J)+b\ln(\sigma_{\max}-\sigma)-c\sigma/(V-V_J), detailed in the Supplemental Material [13] (Sections S8–S13), confirms that ζ1T1=const\zeta_1T_1=\mathrm{const} holds to machine precision (∼10−13\sim10^{-13}) and yields the following quantitative benchmarks: for moderate coupling (c=0.3c=0.3), the total MCTE correction ∣ζ1−1∣|\zeta_1-1| reaches 53% at V0=0.80V_0=0.80 and 63% near jamming (V0−VJ=0.04V_0-V_J=0.04); the purely cross-coupling contribution ∣ζ1(c)−ζ1(0)∣|\zeta_1(c)-\zeta_1(0)| (the irreducible multi-channel correction absent in single-channel theory) reaches 17% near jamming, rising to 28% at c=0.6c=0.6. Figure 2 illustrates the key result.

Figure 2. MCTE invariance on the toy entropy surface S=aln⁡(V−VJ)+bln⁡(σmax⁡−σ)−cσ/(V−VJ)S = a\ln(V-V_J) + b\ln(\sigma_{\max}-\sigma) - c\sigma/(V-V_J) with c=0.3c = 0.3, V0=0.78V_0 = 0.78. (a) Bare compactivity χ\chi varies by a factor of 3 along the entropy level set. (b) Coupling coefficient ζ1\zeta_1 compensates. (c) Product ζ1T1\zeta_1 T_1 is constant to 10−1310^{-13} (machine precision), verifying the MCTE relation. Full parameter study in the Supplemental Material [13].

Microscopic origin of the coupling.—The growth of gVσg_{V\sigma} near jamming has a concrete dynamical origin. Biktashev and Tsyganov [6] showed that adiabatic elimination of a fast diffusing intermediate species ww from a three-component self-diffusion system generates effective cross-diffusion Duv,Dvu≠0D_{uv},D_{vu}\ne0 in the reduced two-component system (their Eq. 12), with magnitude proportional to Dw′=Dw(1−k2Dw)D'_w = D_w(1-k^2D_w). Via the fluctuation-dissipation relation, off-diagonal diffusivity is the kinetic dual of off-diagonal Ruppeiner curvature: gVσ≠0g_{V\sigma}\ne0 arises precisely because the fast intermediate channel, i.e. the mechanical fabric degree of freedom, MM, has been adiabatically eliminated from the full N=3N=3 system, imprinting its coupling onto the N=2N=2 entropy surface. The mapping is structural: both systems share the same three-channel topology and the same elimination mechanism, though the physical variables differ.

Two testable consequences follow directly. First, the MCTE correction (7) becomes O(1)O(1) at the diffusion length ℓ∼Dw1/2\ell\sim D_w^{1/2} of the fast fabric channel, providing a microscopic length scale for the onset of the ζVKμ\zeta_VK_\mu departure in Prediction 3 below. Second, the B-T system supports quasi-soliton waves precisely at the wavenumber k∼Dw−1/2k\sim D_w^{-1/2} where Dw′→0D'_w\to0 [6]; in the granular context this predicts that spatially structured precursors to shear-band formation appear at the same length scale ℓ\ell as the MCTE correction, a coincidence testable in DEM simulations. The even-parity (N=2N=2) framework here describes the equilibrium structure of the reduced system; the N=3N=3 dynamics of the full system, which drives non-associated flow, will be addressed in detail in upcoming works.

Three predictions for granular plasticity.—All three results below follow from the N=2N=2 MCTE geometry without additional constitutive input; full proofs are in the Supplemental Material [13] (Sections S1–S6).

Prediction 1 (dilatancy ratio). Along a quasi-static path maintaining ζVχ=C\zeta_V\chi=C on S(V,σ)S(V,\sigma), the dilatancy ratio in the normal direction n^\hat n is

D=−dV/Vdσ/σ=gVσgVVχA+O((gVσgVV)2),(8)D=-\frac{\mathrm{d}V/V}{\mathrm{d}\sigma/\sigma}=\frac{g_{V\sigma}}{g_{VV}}\frac{\chi}{\mathcal{A}}+O\left(\left(\frac{g_{V\sigma}}{g_{VV}}\right)^2\right), \tag*{(8)}

where gVσ=n^in^jgVσijg_{V\sigma}=\hat n_i\hat n_j g_{V\sigma_{ij}}. Dilatancy is the off-diagonal Ruppeiner curvature, weighted by the ratio of intensive variables.

Prediction 2 (Rowe’s energy restriction). The non-negativity of frictional dissipation in Rowe’s theorem [14], 0≤DR≤1−Kμ−10\le D_R\le1-K_\mu^{-1}, is equivalent to positive semidefiniteness of the 2×22\times2 Ruppeiner metric:

det⁡(gVVgVσgσVgσσ)≥0.(9)\det\begin{pmatrix} g_{VV} & g_{V\sigma} \\ g_{\sigma V} & g_{\sigma\sigma} \end{pmatrix}\ge0. \tag*{(9)}

Equality is the critical state (det⁡g=0\det g=0, maximal channel coupling, maximum ∣ζV−1∣|\zeta_V-1|). Rowe’s restriction is therefore not an independent constitutive hypothesis but a consequence of thermodynamic stability (δ2S≤0\delta^2S\le0): any admissible entropy surface automatically satisfies the Rowe bound.

Prediction 3 (MCTE-corrected Rowe relation). Rowe’s stress-dilatancy law [14], σ1′/σ3′=KμR\sigma'_1/\sigma'_3=K_\mu R, has been observed for 60 years to require a state-dependent KμK_\mu near jamming, with no geometric explanation [3]. The MCTE framework resolves this directly: the correct relation is

σ1′σ3′=ζV(σ)KμR,(10)\frac{\sigma'_1}{\sigma'_3}=\zeta_V(\sigma)K_\mu R, \tag*{(10)}

where ζV(σ)\zeta_V(\sigma) is computable from S(V,σ)S(V,\sigma) alone with no free parameters. In the decoupled limit ζV→1\zeta_V\to1, recovering the classical result exactly. Near jamming ∣ζV−1∣=O(gVσ/gVV)=O(1)|\zeta_V-1|=O(g_{V\sigma}/g_{VV})=O(1), and ζVKμ\zeta_VK_\mu departs from the contact-scale constant by an amount set entirely by the off-diagonal entropy curvature. At the critical state, ζV\zeta_V achieves its maximum consistent with det⁡g=0\det g=0, and Eq. (10) predicts the critical-state stress ratio MM from geometry alone.

The three predictions are testable without free parameters: extract S(V,σ)S(V,\sigma) from discrete-element simulations of the Blumenfeld–Edwards ensemble, compute gijg_{ij} from Eq. (5), and compare DD, the Rowe bound, and ζVKμ\zeta_VK_\mu directly with simulation data. Crucially, ζV(σ)\zeta_V(\sigma) in Eq. (10) is extracted from the local curvature of the entropy surface at each stress state, making the prediction spatially resolved: different regions of a heterogeneous packing will have different ζV\zeta_V values, and the spatial map of ζVKμ\zeta_VK_\mu is a direct observable distinguishing the MCTE correction from any global (spatially uniform) empirical rescaling of KμK_\mu. The N=2N=2 framework has no zero mode, so these results hold independently of questions about non-associated flow, which require the odd-channel (N=3N=3) extensions.

Conclusions.—We have shown that when multiple thermodynamic channels are non-negligibly coupled, the correct equilibrium condition is ζiTi=C\zeta_iT_i=C: a single conserved scalar, with coupling coefficient ζi\zeta_i equal to the holonomy of the Ruppeiner connection on the entropy level set. This generalises both the classical uniform-temperature condition and the relativistic Tolman temperature, and reduces to each in the appropriate limit. For the two-channel granular ensemble, three classical results of granular plasticity: Rowe’s dilatancy ratio, his energy restriction, and the state-dependence of KμK_\mu emerge as parameter-free geometric consequences of the off-diagonal Ruppeiner curvature.

There are two most direct experimental tests: The CC-uniformity test: across a developing shear band the individual intensities χ\chi, A\mathcal{A} should be spatially non-uniform, while ζVχ\zeta_V\chi remains uniform, accessible to continuum simulations and discriminating the MCTE picture from any classical multi-temperature alternative. The length-scale test: the onset of the O(1)O(1) departure in ζVKμ\zeta_V K_\mu should occur at the diffusion length ℓ∼Dw1/2\ell\sim D_w^{1/2} of the fast fabric channel, matching the quasi-soliton wavenumber in the dual reaction-diffusion system [6] and providing a parameter-free prediction for DEM simulations.

The authors acknowledge support from the ARC Centre of Excellence for Carbon Science & Innovation (grant CE230100021) and Curtin University.

References

  1. [1]C. Scheibner, A. Souslov, D. Banerjee, P. Surówka, W. T. M. Irvine, and V. Vitelli, Nature Physics 16, 475–480 (2020).
  2. [2]M. Fruchart, R. Hanai, P. B. Littlewood, and V. Vitelli, Nature 592, 363–369 (2021).
  3. [3]R. Blumenfeld and S. F. Edwards, Physical Review Letters 90, 114303 (2003).
  4. [4]R. Blumenfeld, J. F. Jordan, and S. F. Edwards, Physical Review Letters 109, 238001 (2012).
  5. [5]M. Polettini, Europhysics Letters 97, 30003 (2012).
  6. [6]V. N. Biktashev and M. A. Tsyganov, Scientific Reports 6, 30879 (2016).
  7. [7]G. Ruppeiner, Physical Review A 20, 1608 (1979).
  8. [8]G. Ruppeiner, Reviews of Modern Physics 67, 605 (1995).
  9. [9]F. Weinhold, J. Chem. Phys. 63, 2479 (1975).
  10. [10]D. C. Brody and A. Riegler, J. Geom. Phys. 47, 207 (2003).
  11. [11]G. E. Crooks, Phys. Rev. Lett. 99, 100602 (2007).
  12. [12]R. C. Tolman, Physical Review 35, 904 (1930).
  13. [13]See Supplemental Material for complete proofs (Part I, Sections S1–S7) and the toy model demonstration (Part II, Sections S8–S13).
  14. [14]P. W. Rowe, Proceedings of the Royal Society of London, Series A 269, 500 (1962).

Paper details

vlnc:W_4mgbsf5e
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents