Inverse Currents in Coupled Transport

• ICC is defined as a regime where cross-coupling in particle and energy transport leads to a current flowing opposite to its conjugate thermodynamic force while preserving non-negative entropy production.
• Both classical Hamiltonian models and quantum dot systems illustrate ICC through negative off-diagonal Onsager coefficients and dynamic self-organization that produce counterintuitive transport behavior.
• ICC has practical implications for autonomous devices such as heat engines and refrigerators, enabling work extraction against conventional thermodynamic biases.

Searching arXiv for recent and foundational papers on inverse currents in coupled transport. Inverse currents in coupled transport (ICC) denote transport regimes in which coupling between distinct nonequilibrium channels causes one steady current to acquire a sign opposite to the direction naively associated with its conjugate thermodynamic force. In the one-dimensional interacting Hamiltonian system of "Inverse Currents in Hamiltonian Coupled Transport," ICC concerns coupled particle and energy transport and arises when perturbations of equilibrium by thermodynamic forces produce a negative off-diagonal Onsager coefficient (Wang et al., 2019). Subsequent quantum work established that inverse current also exists in Coulomb-coupled quantum dots (Zhang et al., 2021), and later formulations cast genuine ICC as a thermodynamic regime with mutually parallel energy and particle forces for which one current still flows opposite to both forces without violating the second law (Ghosh et al., 2024). The subject therefore sits at the intersection of nonequilibrium statistical mechanics, Onsager theory, and nanoscale transport thermodynamics, with direct relevance to autonomous heat-engine and refrigerator modes in coupled quantum-dot platforms (Ghosh et al., 6 Mar 2026).

1. Definition and scope

In the classical Hamiltonian formulation, the relevant densities are the particle density $\rho$ of the bullet species and the energy density $u$. Their conjugate thermodynamic forces are

$$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$

with $\beta_\alpha = 1/(k_B T_\alpha)$, while the corresponding steady-state fluxes are the bullet particle current $J_\rho$ and the total energy current $J_u$, taken as positive from left to right (Wang et al., 2019). In this setting, an inverse current means that the sign of an induced current is opposite to the applied force conjugate to that current.

The later quantum thermodynamic literature sharpens the definition. In ordinary coupled transport, two mutually parallel forces such as $X_E>0$ and $X_N>0$ drive both currents in the same sense, $J_E>0$ and $J_N>0$. ICC refers to the more counterintuitive regime in which one induced current flows opposite to both forces, for example $u$0 while $u$1 and $u$2 (Ghosh et al., 6 Mar 2026). For a two-terminal quantum device, the forces can be written as

$u$3

and the left-lead currents as

$u$4

This thermodynamic formulation makes ICC a precisely defined cross-effect rather than a generic sign reversal.

A useful conceptual distinction follows from these definitions. The classical Hamiltonian work emphasizes inverse response relative to the directly conjugate force, whereas the quantum thermodynamic framework emphasizes a stricter regime of mutually parallel forces and one current opposing both. The two uses are continuous rather than contradictory: both describe cross-coupled transport in which off-diagonal response overwhelms direct response.

2. Onsager structure and thermodynamic consistency

In the near-equilibrium regime, the classical Hamiltonian model is written in Onsager form as

$u$5

or equivalently $u$6, with $u$7 (Wang et al., 2019). Onsager reciprocity gives $u$8, the diagonal coefficients satisfy $u$9 and $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$0, and the crucial model-specific result is

$$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$1

This negative thermodiffusion cross-coefficient is the central linear-response signature of ICC in the Hamiltonian system.

The inverse-current criterion follows immediately. If $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$2 and $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$3, then

$$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$4

so the bullet current flows from cold to hot. If $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$5 and $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$6, then

$$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$7

so the energy current is opposite to the direction selected by the directly conjugate force. More generally, for $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$8, an inverse $$F_\rho \equiv \mu_L \beta_L - \mu_R \beta_R, \qquad F_u \equiv \beta_R - \beta_L,$$9 requires

$\beta_\alpha = 1/(k_B T_\alpha)$0

The mechanism is therefore not the negativity of the diagonal response, but the dominance of a negative cross-response over a positive direct response.

In the quantum formulation, the same logic appears in the linear-response equations

$\beta_\alpha = 1/(k_B T_\alpha)$1

with Onsager reciprocity $\beta_\alpha = 1/(k_B T_\alpha)$2 (Ghosh et al., 6 Mar 2026). Positivity of the entropy-production rate for arbitrary $\beta_\alpha = 1/(k_B T_\alpha)$3 requires

$\beta_\alpha = 1/(k_B T_\alpha)$4

For ICC in the particle current,

$\beta_\alpha = 1/(k_B T_\alpha)$5

Thus ICC is compatible with the second law because the total entropy production remains

$\beta_\alpha = 1/(k_B T_\alpha)$6

This thermodynamic point resolves the principal misconception surrounding ICC. The effect is not a breakdown of irreversibility or of Onsager theory; it is a regime in which cross-coupling is sufficiently strong, and sufficiently signed, that one channel runs inversely while the aggregate entropy production remains non-negative.

3. Hamiltonian realization: negative thermodiffusion and self-organization

The classical model is a one-dimensional box of length $\beta_\alpha = 1/(k_B T_\alpha)$7 containing two species: bullets of mass $\beta_\alpha = 1/(k_B T_\alpha)$8 and rods of mass $\beta_\alpha = 1/(k_B T_\alpha)$9. Bullets exchange both particles and energy with ideal-gas reservoirs, whereas rods exchange only energy and their total number is fixed. Interactions are defined by a finite barrier $J_\rho$0: whenever two particles meet, they pass through if their relative kinetic energy exceeds $J_\rho$1, otherwise they collide elastically (Wang et al., 2019). These ingredients are minimal but sufficient to generate coupled particle-energy transport with a nontrivial cross coefficient.

The microscopic picture of ICC is explicitly dynamical. Under a positive temperature bias $J_\rho$2, with the hot side on the left, light rods acquire, on average, higher velocities near the hot end and cross heavy bullets more readily in one orientation. This dynamically builds up a species-density asymmetry: rods concentrate near the cold end and bullets near the hot end. Because only bullets are injected and absorbed, a net bullet current from cold to hot emerges, $J_\rho$3, so $J_\rho$4 (Wang et al., 2019). The paper describes this mechanism as self-organization: the applied forces induce a redistribution of species that feeds back on transport coefficients.

The same logic applies when the particle-affinity channel is used as the perturbation. For $J_\rho$5, an entirely analogous picture holds, and the energy current can become inverse. In both cases, the key point is that the system does not merely transmit reservoir biases; it adaptively rearranges its internal composition, and that rearrangement changes the sign of the cross-response.

For stronger driving, the system undergoes phase separation. A pure-bullet domain forms at one end, further enhancing $J_\rho$6 and deepening ICC (Wang et al., 2019). This places self-organization at the center of the phenomenon: the inverse current is not an externally imposed anomaly, but the macroscopic transport consequence of a force-induced restructuring of the nonequilibrium steady state.

4. Quantum-dot realizations and symmetry breaking

Quantum ICC has been developed in several related double-dot settings, ranging from a Coulomb-coupled spinless model to strongly coupled spin-polarized dots and the three-terminal coupled-quantum-dot review framework of Ghosh and Ghosh (Zhang et al., 2021, Ghosh et al., 2024, Ghosh et al., 6 Mar 2026).

Platform	Microscopic condition	Inverse regime
Coulomb-coupled double quantum dot	Increasing $J_\rho$7; for $J_\rho$8, $J_\rho$9	$J_u$0 or $J_u$1
Strongly coupled spin-polarized dots	Attractive $J_u$2; sufficient condition $J_u$3	$J_u$4 or $J_u$5 with $J_u$6
Three-terminal CQD	Attractive $J_u$7 and $J_u$8	particle ICC $J_u$9 or energy ICC $X_E>0$0

In the Zhang-Xie Coulomb-coupled quantum-dot model, two single-level, spinless quantum dots, top $X_E>0$1 and bottom $X_E>0$2, are each tunnel-coupled to two leads and capacitively coupled to one another through an inter-dot Coulomb energy $X_E>0$3. The stationary transport problem is solved in the Coulomb-blockade regime by a master equation for the state probabilities $X_E>0$4. Numerically, for fixed $X_E>0$5 and $X_E>0$6, the particle current $X_E>0$7 first flows in the bias direction and then reverses sign once $X_E>0$8 exceeds a threshold $X_E>0$9. In the limit $X_N>0$0, one has $X_N>0$1 and $X_N>0$2, so the dominant transport paths flip sign. Inverse particle current and inverse energy current occupy complementary regions in the $X_N>0$3 plane (Zhang et al., 2021).

The exactly solvable model of "Inverse Current in Coupled Transport: A Quantum Thermodynamic Framework for Energy and Spin-polarized Particle Currents" uses two single-level quantum dots, one carrying spin-polarized $X_N>0$4 occupation and the other spin-polarized $X_N>0$5 occupation, with effective interaction $X_N>0$6. The four eigenstates $X_N>0$7 are connected by sequential tunneling to three reservoirs $X_N>0$8, and the steady-state entropy production takes the macroscopic form $X_N>0$9. The authors identify two conditions for genuine ICC: necessity, $J_E>0$0; and sufficiency,

$J_E>0$1

which is a level-swap condition. In that regime, one finds either $J_E>0$2 or $J_E>0$3 while $J_E>0$4 and $J_E>0$5, with disjoint inverse-current regions and $J_E>0$6 throughout (Ghosh et al., 2024).

The 2026 review extends the thermodynamic argument to coupled quantum-dot systems in general. In generic two-terminal single-QD models, one finds $J_E>0$7, so currents cannot be inverted independently. The key to ICC is therefore breaking the proportionality between energy and particle transport by introducing a second dot strongly capacitively coupled to the first in a three-terminal CQD geometry with

$J_E>0$8

For repulsive $J_E>0$9, the eigen-level ordering enforces particle and energy exchange to coincide and there is no ICC. For attractive $J_N>0$0 and sufficiently strong interaction, $J_N>0$1, the ordering of the two-electron $J_N>0$2 and one-electron $J_N>0$3 states inverts. Particle excitation can then occur together with energy de-excitation and vice versa, breaking the symmetry between energy- and particle-transfer channels and permitting ICC (Ghosh et al., 6 Mar 2026).

5. Device modes, experimental signatures, and operational meaning

In the quantum thermodynamic setting, ICC is not merely an anomalous sign reversal; it defines operating modes of autonomous devices. Beyond the threshold $J_N>0$4, the CQD model supports one regime in which the particle current satisfies $J_N>0$5, described as particle ICC or autonomous heat-engine mode, and another in which the energy current satisfies $J_N>0$6, described as energy ICC or autonomous refrigerator mode (Ghosh et al., 6 Mar 2026). The sign inversion is therefore operationally meaningful: it maps directly onto work extraction against a chemical-potential bias or heat extraction against a temperature bias.

In the spin-polarized model, the same interpretation is explicit. For ICC in the spin-polarized particle current,

$J_N>0$7

For ICC in the energy current,

$J_N>0$8

The paper emphasizes that both the engine and refrigerator operate autonomously, without external switching, and that they exploit the counterintuitive ICC effect (Ghosh et al., 2024). This places ICC within the broader theory of autonomous nanoscale thermoelectric engines and refrigerators.

The most direct experimental geometry proposed in the review is the three-terminal CQD, or Sánchez–Büttiker, configuration. One dot, $J_N>0$9, couples to left and right leads at $u$00 with $u$01; the other dot, $u$02, couples to a third heat lead at $u$03. By further tuning $u$04, one obtains mutually parallel $u$05 and $u$06, which is precisely the regime in which genuine ICC can be observed (Ghosh et al., 6 Mar 2026). A signature is the reversal of the sign of $u$07 or $u$08 while both $u$09 and $u$10 point in the same nominal direction.

The review also notes that spin-polarized reservoirs can realize effective $u$11 via spin-spin dressing, as demonstrated in carbon-nanotube double-dots by Hamo et al. 2016 (Ghosh et al., 6 Mar 2026). This is significant because attractive interaction is the necessary criterion in the exactly solvable framework and the sufficient condition for channel-symmetry breaking is a level inversion driven by that interaction.

6. Conceptual boundaries and related inverse-transport phenomena

A recurrent source of confusion is the conflation of ICC with any transport sign reversal. The ICC literature is defined by conjugate force-flux pairs, coupled response coefficients, and non-negative total entropy production. Several neighboring literatures exhibit inverse or opposite-sign transport, but not all are ICC in this thermodynamic sense.

In two coupled Josephson junctions driven by zero-mean ac signals, Machura, Spiechowicz, and Łuczka identified regimes in which the two junctions rectify dc voltages of opposite sign, including $u$12 and $u$13. In their split-drive scenario, the maximal $u$14 typically exceeds those of the single-junction biharmonic-drive scenario by factors of $u$15 (Machura et al., 2012). A related study of two coupled Josephson junctions under combined dc and ac forcing found anomalous negative response in which $u$16 despite a positive dc bias, with finite coupling, low frequency, and suitably chosen ac amplitudes and noise intensity playing essential roles (Spiechowicz et al., 2012). These are inverse-transport effects in nonlinear driven systems, but their organizing framework is directed transport and rectification rather than thermodynamic ICC of energy-particle coupling.

A different example is the longitudinal transport problem in type-II superconductors. Ruiz et al. showed that flux depinning and flux-line cutting thresholds can force a negative surface transport layer, $u$17, while maintaining a positive central current, together with collateral paramagnetic behavior (Ruiz et al., 2010). In a further example, Mukhopadhyay et al. showed that dimensional coupling in a two-dimensional ac-driven lattice permits trajectories to access phase-space regions that are inaccessible in the quasi-one-dimensional limit, producing a delayed current reversal with $u$18 (Mukhopadhyay et al., 2018). These cases again involve backflow or current reversal, but their microscopic structures differ from coupled thermodynamic force-flux transport.

Reciprocity subtleties in other coupled-transport fields sharpen this boundary. In orbital-charge-coupled transport, Go et al. argued that the conventional orbital current is ill-defined and that global reciprocity is restored only for the proper orbital current; even then, layer-resolved local responses can be strongly nonreciprocal because surface contributions differ between direct and inverse processes (Go et al., 2024). This is not ICC, but it shows that once a transport quantity is nonconserved, the precise definition of forces, fluxes, and conjugacy becomes decisive.

A plausible implication is that ICC belongs to a broader family of coupled sign-inversion phenomena, but with a stricter thermodynamic signature than most of its analogues. Its defining features are explicit force-flux pairing, a cross-response large enough to overcome the direct response of the conjugate force, and preservation of $u$19 despite the inverse sign of one current.