Onsager Symmetry in Linear Thermodynamics

• Onsager symmetry is the reciprocal structure governing linear relations between fluxes and forces near equilibrium, ensuring time-reversal invariance.
• It underpins the Onsager–Casimir relations in systems with time-reversal-odd fields, linking coupled transport phenomena across thermoelectric and charge–spin contexts.
• Advanced formulations extend its applicability beyond the linear Gaussian regime to generalized gradient flows, integrable lattice models, and nonequilibrium systems.

Searching arXiv for recent and foundational papers on Onsager symmetry and reciprocity. Onsager symmetry is the reciprocal structure of linear irreversible thermodynamics near equilibrium. In its classical form, if fluxes $J_i$ are related to generalized forces $X_j$ by $J_i=\sum_j L_{ij}X_j$, microscopic reversibility implies $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$, where $\varepsilon_i$ denotes time-reversal parity; for variables of the same parity this reduces to the familiar symmetric form $L_{ij}=L_{ji}$. In contemporary research, the term also covers Onsager–Casimir relations in the presence of time-reversal-odd fields, generalized gradient-flow formulations beyond the linear Gaussian regime, and, in integrable lattice models, the non-Abelian Onsager algebra generated by Dolan–Grady-type operators (Mielke et al., 2015, Brechet, 2022, Vernier et al., 24 Jun 2026).

1. Classical statement and thermodynamic content

In linear nonequilibrium thermodynamics, Onsager symmetry concerns the matrix of phenomenological coefficients connecting fluxes and forces. The standard setting is a neighborhood of equilibrium in which the entropy production rate is a positive quadratic form, $\sigma=\sum_i J_iX_i\ge 0$, and the constitutive law is linear. Under microscopic reversibility, the cross-couplings are reciprocal: the coefficient linking force $X_j$ to flux $J_i$ equals, up to time-reversal parity, the coefficient linking $X_i$ to $X_j$0 (Brechet, 2022, Mielke et al., 2015).

A purely phenomenological derivation expresses dissipated power densities as positive semi-definite quadratic forms of generalized forces. In that formulation, the antisymmetric part of the scalar or vectorial Onsager matrix does not contribute to dissipation, and the reciprocal relations arise by requiring invariance of those quadratic forms under time reversal. This produces the scalar and vectorial Onsager relations in the absence of time-reversal-odd fields and the Onsager–Casimir form when such fields are present (Brechet, 2022).

The same reciprocity appears in specific transport settings discussed across the literature summarized here: thermoelectric transport, coupled phase flows, heat engines, chemically active colloids, and charge–spin conversion. A recurring misconception is that reciprocity is a statement about any driven linear system. The literature instead treats it as a near-equilibrium statement tied either to microscopic reversibility or to an appropriate generalized antiunitary symmetry, depending on the system class (Huang et al., 18 Jun 2025, Corato et al., 2022).

2. Microscopic basis, Green–Kubo structure, and variational reformulations

The standard microscopic basis is the Green–Kubo representation. With equilibrium weighting, transport coefficients are time integrals of current–current correlation functions, and time-reversal invariance of the equilibrium measure yields the reciprocal symmetry of the integrated cross-correlations. This is the mechanism used both in textbook derivations and in explicit numerical studies of mesoscopic systems (Winkler et al., 2020, Mielke et al., 2015).

A mathematically sharper reformulation replaces the symmetry of $X_j$1 by self-adjointness of the kinetic operator $X_j$2 in the entropic inner product. For linear or linearized kinetics $X_j$3, $X_j$4 is symmetric with respect to the Hessian of the thermodynamic potential at equilibrium, and therefore the propagator $X_j$5 is symmetric in the same inner product. This yields reciprocal relations between entire kinetic curves: for a master equation started from pure state $X_j$6, the ratio $X_j$7 is constant in time and equals the ratio of equilibrium probabilities (Yablonsky et al., 2010).

Mielke, Peletier, and Renger generalized Onsager’s 1931 theorem beyond Gaussian invariant measures and linear macroscopic evolution by replacing a symmetric mobility matrix with symmetric convex dissipation potentials $X_j$8 and $X_j$9. In that framework, microscopic reversibility plus pathwise large deviations yields a generalized gradient-flow structure,

$J_i=\sum_j L_{ij}X_j$0

and the nonlinear analogue of Onsager symmetry becomes the evenness of the dissipation potential, $J_i=\sum_j L_{ij}X_j$1 (Mielke et al., 2015).

For adiabatically driven Gaussian heat engines, a related hierarchy connects local and global response. The local Onsager matrix $J_i=\sum_j L_{ij}X_j$2 obeys a symmetry/antisymmetry pattern in which the work–work block is symmetric while the cross blocks satisfy $J_i=\sum_j L_{ij}X_j$3. Cycle-averaging yields global coefficients $J_i=\sum_j L_{ij}X_j$4, and after a change of thermodynamic variables the reduced global matrix becomes symmetric with vanishing determinant, i.e. a tight-coupling form (Izumida, 2021).

3. Magnetic fields, broken time reversal, and generalized reciprocity

The conventional extension in magnetic fields is the Onsager–Casimir relation,

$J_i=\sum_j L_{ij}X_j$5

or, with parity factors, $J_i=\sum_j L_{ij}X_j$6. This has often been taken to mean that symmetry at fixed field is generically lost. Several of the cited works refine that conclusion (Luo et al., 2019, Carbone et al., 2020).

For classical interacting particles in two dimensions with a magnetic field $J_i=\sum_j L_{ij}X_j$7, an exact generalized time-reversal operation exists that combines reflection in $J_i=\sum_j L_{ij}X_j$8, partial momentum reversal, and $J_i=\sum_j L_{ij}X_j$9, without changing the sign of the magnetic field. In that case one obtains $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$0 at fixed field profile. Numerical evidence in a multi-particle-collision gas further indicates that the same fixed-field symmetry persists for generic $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$1 profiles in the two-terminal thermoelectric setup studied there (Luo et al., 2019).

A broader Hamiltonian analysis shows that the class of time-reversal operations in magnetic fields is much larger than the canonical one. The paper derives the most general form of generalized time reversal on phase space and gives necessary and sufficient conditions on $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$2 and on the potential $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$3 for the existence of a field-preserving involution $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$4 satisfying $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$5. Whenever such a generalized time-reversal invariance exists, the standard Onsager relations at fixed $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$6 follow; Casimir’s form is recovered only when one insists on the canonical operation or when no generalized symmetry exists (Carbone et al., 2020).

A different generalization replaces bare time reversal $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$7 by a combined symmetry $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$8. In spin–orbit-coupled charge–spin transport, the symmetry or antisymmetry of the cross-coupling coefficients remains in place provided that the unitary operator $L_{ij}=\varepsilon_i\varepsilon_j L_{ji}$9 satisfies the conditions identified in the analysis. This leads to generalized Onsager reciprocal relations in systems that break $\varepsilon_i$0 but preserve $\varepsilon_i$1 (Huang et al., 18 Jun 2025).

The same logic appears experimentally in a tilted Weyl semimetal. In Co$\varepsilon_i$2Sn$\varepsilon_i$3S$\varepsilon_i$4, the chirality-dependent Hall effect is antisymmetric in both in-plane magnetic field and magnetization, yet the measured resistances satisfy

$\varepsilon_i$5

and the local Hall responses satisfy the corresponding Onsager–Casimir relation. The interpretation given there is that anomalous-velocity and chiral-chemical-potential contributions are linked by reciprocity once all time-reversal-odd quantities are properly reversed (Jiang et al., 2022).

4. Nonequilibrium extensions, coarse graining, and effective asymmetry

Several papers distinguish sharply between fundamental reciprocity and effective asymmetry induced by coarse graining, driving, or elimination of variables. In periodically driven stochastic thermodynamics, a Fourier-resolved Onsager matrix can possess antisymmetric nondissipative parts, and reduced macroscopic Onsager matrices can be asymmetric under time-asymmetric driving. Nevertheless, the second law implies that all microscopic fluxes vanish in the zero-dissipation limit, so any reduced Onsager matrix must become singular; in the $\varepsilon_i$6 case this forces emergent symmetry $\varepsilon_i$7 (Proesmans et al., 2015).

Chemically active colloids provide a direct chemo-mechanical realization. For a spherical catalyst that mediates a reversible surface reaction $\varepsilon_i$8, the linear relation between particle velocity $\varepsilon_i$9, net reaction rate $L_{ij}=L_{ji}$0, mechanical affinity, and chemical affinity has a symmetric Onsager matrix around equilibrium, with the cross-coefficients satisfying $L_{ij}=L_{ji}$1. The analysis also shows that solute advection is structurally required to preserve this symmetry, even in regimes where propulsion speeds are weakly affected by advection. Around a nonequilibrium steady state, by contrast, the same paper finds $L_{ij}=L_{ji}$2, so the Onsager matrix for fluctuations becomes asymmetric (Corato et al., 2022).

Onsager’s variational principle can also be used to derive nonreciprocal effective equations in active matter. In the odd-elasticity construction, the extended system obeys Onsager symmetry at the level of the chosen variables and the dissipation function is symmetric, but elimination of an extra variable coupled to a nonequilibrium force yields an effective antisymmetric elastic response. The odd elastic constants are then proportional to the nonequilibrium force and the friction coefficients, providing a concrete route from reciprocal extended dynamics to nonreciprocal reduced equations (Lin et al., 2022).

These examples support a common distinction. Fundamental Onsager reciprocity is a statement about a particular microscopic or extended description; asymmetry in a reduced description need not contradict it. This suggests that many reported “violations” are more precisely reparameterizations, coarse-grained reductions, or expansions about nonequilibrium steady states rather than counterexamples to the classical theorem.

5. Entropy-weighted and geometric reinterpretations

A recent multiscale proposal recasts apparent Onsager violation as a consequence of entropy-weighted statistical ensembles and entropy-weighted reparameterization of thermodynamic variables, while leaving the microscopic Onsager theorem intact (Dabas, 21 Mar 2026). The microscopic starting point is a weighted ensemble with

$L_{ij}=L_{ji}$3

The corresponding weighted Green–Kubo coefficients satisfy the “Weighted Reciprocity Theorem”

$L_{ij}=L_{ji}$4

If $L_{ij}=L_{ji}$5, one recovers standard Onsager reciprocity. If $L_{ij}=L_{ji}$6, the weighted transport matrix acquires an antisymmetric part without altering the underlying time-reversible Hamiltonian dynamics (Dabas, 21 Mar 2026).

On the macroscopic side, the same work introduces entropy-weighted response variables

$L_{ij}=L_{ji}$7

with invariants

$L_{ij}=L_{ji}$8

In equilibrium they satisfy

$L_{ij}=L_{ji}$9

The pressure–volume sector is encoded by the accessibility $\sigma=\sum_i J_iX_i\ge 0$0-form

$\sigma=\sum_i J_iX_i\ge 0$1

with curvature

$\sigma=\sum_i J_iX_i\ge 0$2

Equilibrium corresponds to $\sigma=\sum_i J_iX_i\ge 0$3, while $\sigma=\sum_i J_iX_i\ge 0$4 is interpreted as thermodynamic curvature and as an effective asymmetry in the transformed coupling matrix (Dabas, 21 Mar 2026).

The paper then ties this geometric picture to two datasets. First, the Transforma model for the $\sigma=\sum_i J_iX_i\ge 0$5 transition series shows cross-derivative asymmetries peaking at Cr and Cu, the classic configuration anomalies. Second, temperature-dependent Raman spectroscopy of monolayer graphene exhibits statistically significant hysteresis loops, with loop areas up to $\sigma=\sum_i J_iX_i\ge 0$6, and the paper interprets the nonzero loop area through Stokes’ theorem,

$\sigma=\sum_i J_iX_i\ge 0$7

as direct evidence of thermodynamic curvature. The paper’s claims are explicitly framed as an effective or transformed-level violation, not as a refutation of the microscopic Onsager theorem (Dabas, 21 Mar 2026).

6. Representative realizations across scales and algebraic extensions

Onsager symmetry has been probed in systems far removed from its original molecular context. In incompressible, immiscible two-phase flow through porous media, a 2D network model shows Gaussian flow distributions at a sufficiently large representative elementary volume, convergence of Green–Kubo-type integrals, and symmetric integrated cross-correlations of wetting and non-wetting phase flows, $\sigma=\sum_i J_iX_i\ge 0$8, within statistical uncertainty. The result supports a nonequilibrium thermodynamic description of an athermal mesoscopic system (Winkler et al., 2020).

In synthetic-field optomechanics, a closed optomechanical loop with two mechanical resonators and a cavity exhibits directional signal and noise routing in the presence of a loop phase that implements synthetic magnetism. The system nevertheless obeys Onsager–Casimir symmetry once one compares responses with their time-reversed counterparts, namely under resonator interchange $\sigma=\sum_i J_iX_i\ge 0$9 together with loop-phase reversal. The paper introduces a non-reciprocity measure based on the difference between such time-reversed counterparts and shows that non-reciprocal noise flow is strongest for smaller intermechanical couplings, whereas higher coupling is more viable for sensing (Ege et al., 19 Aug 2025).

The term “Onsager symmetry” also has a distinct algebraic meaning in integrable lattice models. In the XXZ spin chain at roots of unity, non-commuting transfer matrices built from cyclic and nilpotent representations generate an explicit lattice realization of the Onsager algebra. The duality automorphism is represented by a matrix product operator related to the $X_j$0 model, and that operator obeys $X_j$1 Tambara–Yamagami fusion rules. In the scaling limit, this is identified with the topological defect lines of the free compactified boson conformal field theory. In that setting, Onsager symmetry refers not to reciprocity of transport coefficients but to an infinite-dimensional non-Abelian symmetry algebra and its duality structure (Vernier et al., 24 Jun 2026).

Taken together, these works show that “Onsager symmetry” now names a family of related but non-identical structures: reciprocal transport near equilibrium, generalized reciprocity under modified antiunitary symmetries, symmetric or even dissipation structures in generalized gradient flows, and Onsager-algebra symmetry in integrable models. A plausible implication is that the enduring utility of the concept comes from this common core: a symmetry that constrains cross-couplings, whether those couplings are transport coefficients, kinetic propagators, defect fusion rules, or transformed nonequilibrium response functions.