Levitov–Lesovik Formula: Charge Transfer Stats

• The Levitov–Lesovik formula is a canonical expression that calculates the full counting statistics of charge transfer in non-equilibrium steady state free-fermion systems.
• It computes a scaled cumulant generating function using an energy integral of logarithms built on transmission probabilities and reservoir occupation factors, reducing to binomial or Poissonian limits.
• Its determinant structure underpins extensions to interacting systems, monitored transport, and energy transfer analogues, offering versatile applications in mesoscopic physics.

The Levitov–Lesovik formula is the canonical full-counting-statistics expression for the long-time distribution of charge transfer in non-equilibrium steady transport of free fermions. In its standard form, it writes the scaled cumulant generating function as an energy integral of a logarithm built from transmission probabilities and reservoir occupation factors; in the long-time, zero-temperature single-mode limit it reduces to a binomial law, while in the weak-tunneling, elastic-scattering limit it becomes Poissonian in the transferred charge (Bernard et al., 2011, Beenakker et al., 10 Apr 2025, Skorobagatko, 2018).

1. Standard formulation

A convenient definition of the scaled cumulant generating function is

$$\chi(\lambda)=\lim_{t \rightarrow \infty} \frac{1}{t}\ln \left[\sum_{N=-\infty}^\infty P(N,t)e^{\lambda N}\right],$$

where $P(N,t)$ is the probability that the net transferred charge in time $t$ equals $N$. For noninteracting transport, one exact form used in a two-reservoir geometry is (Ptaszynski, 2018)

$$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$

with $f_\alpha(\omega)$ the Fermi occupations and $g_\alpha(\omega)=1-f_\alpha(\omega)$. The scaled cumulants follow from derivatives at $\lambda=0$,

$$c_i=\left.\frac{\partial^i\chi(\lambda)}{\partial\lambda^i}\right|_{\lambda=0},$$

so the formula organizes the entire hierarchy of current cumulants within a single generating object (Ptaszynski, 2018).

A frequently quoted single-channel steady-state form is (Bernard et al., 2011)

$$F(\lambda)=\int \frac{d\omega}{2\pi}\, \log\!\Big( 1+T(\omega)\big[ n_1(\omega)(n_2(\omega)-1)(1-e^{i\lambda}) +n_2(\omega)(n_1(\omega)-1)(1-e^{-i\lambda}) \big] \Big),$$

with Fermi occupations

$P(N,t)$0

In this representation the formula is a large-deviation object: it gives the long-time rate function for the charge-transfer statistics in terms of the transmission coefficient $P(N,t)$1 and the left/right thermal data (Bernard et al., 2011).

The same structure contains important limiting distributions. For a single detected outgoing mode with $P(N,t)$2 incident electrons and transmission probability $P(N,t)$3, the long-time zero-temperature law is binomial,

$P(N,t)$4

whereas in the weak-tunneling, long-time, elastic-scattering limit the cumulant generating function becomes Poissonian in the transferred charge (Beenakker et al., 10 Apr 2025, Skorobagatko, 2018).

2. Measurement prescriptions and determinant structure

The historical formulation distinguished two charge-measurement protocols. One is a von Neumann or two-time projective measurement of the reservoir charge; the other is the Levitov–Lesovik spin-measurement protocol, in which an interacting spin detector is coupled to the current. In the resonant-level model, these prescriptions differ at finite time because the projective protocol contains two insertions of the initial charge projector whereas the spin protocol contains only one, but both give the same large deviation function in the large-time steady-state limit (Bernard et al., 2011).

In free-fermion settings, the many-body generating function reduces to a determinant over one-particle Hilbert space. One form is

$P(N,t)$5

where $P(N,t)$6 is the counting operator reorganized into a form effectively built from local operators, and $P(N,t)$7 is the one-particle steady-state correlator (Bernard et al., 2011). In the standard unitary scattering setting, the moment generating function is likewise given by a Levitov–Lesovik determinant formula in terms of the single-particle scattering matrix $P(N,t)$8, the projector $P(N,t)$9 selecting the counted outgoing modes, and the equilibrium occupation matrix $t$0 of the incoming modes (Beenakker et al., 10 Apr 2025).

This determinant structure is the technical core of the formula. It explains why the statistics factorize into mode-by-mode contributions and why the same basic logarithmic form recurs across charge transport, heat transport, monitored transport, and higher-dimensional free theories. The determinant representation also makes precise that the formula is a statement about the full probability law, not only about mean current or zero-frequency noise (Bernard et al., 2011, Beenakker et al., 10 Apr 2025).

3. Exact proofs and free-fermion realizations

A rigorous derivation in a concrete interacting-impurity geometry was given for the resonant-level model, where two decoupled reservoirs prepared at different chemical potentials and temperatures are coupled through a single fermionic level. The steady state is described by a Hershfield-type density matrix,

$t$1

with conserved nonlocal charges $t$2 and $t$3, and the Levitov–Lesovik formula is obtained by recasting the charge-counting problem into averages of finitely supported operators (Bernard et al., 2011). In that model the transmission coefficient is

$t$4

in the conventions of the paper, with $t$5 the dot energy and $t$6 the tunneling amplitude (Bernard et al., 2011).

The same logic extends beyond one dimension. In the $t$7-dimensional free massive Dirac theory, the scaled cumulant generating function for transferred $t$8 charge in the non-equilibrium steady state takes a higher-dimensional Levitov–Lesovik form,

$t$9

with an overall factor $N$0 for spin, a sum over particles and antiparticles, and effectively perfect transmission $N$1 (Yoshimura, 2018). In the massless case, only the first four scaled cumulants are nonzero (Yoshimura, 2018).

The formula is also used as an exact transport tool in noninteracting mesoscopic thermoelectrics. For a two-level bridge of tunnel-coupled orbitals, each coupled to one reservoir, the transmission function entering the Levitov–Lesovik formula is (Ptaszynski, 2018)

$N$2

From the same generating function one obtains the mean particle current, particle-current variance, and heat currents. In the interpretation given there, coherent inter-orbital tunneling is associated with unitary evolution rather than a stochastic Poisson transition, and this suppresses current and power fluctuations relative to purely stochastic sequential transport (Ptaszynski, 2018).

4. Interacting extension and self-equilibration

The standard Levitov–Lesovik formula is a scattering-theory result for non-interacting electrons. A major extension is the exact Keldysh-time formulation for weak tunnel contacts between interacting Luttinger liquids, where the generating function is written as

$N$3

with bosonized tunnel operator

$N$4

The cumulant generating function is defined through

$N$5

and the cumulants are

$N$6

(Skorobagatko, 2018).

The central result is the self-equilibration theorem, which proves exact re-exponentiation of the full Keldysh series,

$N$7

with

$N$8

The associated differential form,

$N$9

is interpreted as a dynamical self-equilibration law for the Keldysh partition function (Skorobagatko, 2018).

In the long-time limit, the generating function becomes

$$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$0

with

$$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$1

This is presented as a long-time generalization of the Levitov–Lesovik formula to interacting Luttinger liquids. The standard result is recovered for $$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$2, corresponding to Fermi-liquid leads, most directly in the long-time asymptotic and lowest-order weak-tunneling expansion; the paper also states that the $$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$3 finite-time expression reproduces the familiar finite-time logarithmic correction structure for the lowest cumulants (Skorobagatko, 2018). In this construction, the Levitov–Lesovik result is not discarded but embedded as a limiting case of a broader exact full-counting-statistics theory.

5. Heat-transfer analogues

The Levitov–Lesovik structure also appears when the counted quantity is energy rather than charge. In extreme-near-field radiative heat transfer between two metallic bodies, the full counting statistics of transferred energy $$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$4 are defined by

$$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$5

and, after a Keldysh nonequilibrium Green’s function treatment with random phase approximation, the long-time scaled cumulant generating function becomes (Tang et al., 2018)

$$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$6

Here the Fermi factors of charge transport are replaced by Bose functions, and each transfer event carries energy $$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$7. The same SCGF yields the Caroli-type heat current and the second cumulant of heat-current fluctuations; in linear response the paper derives

$$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$8

with $$\chi(\lambda)=\int_{-\infty}^{\infty} \frac{d \omega}{2\pi} \ln \left \{ 1+\mathcal{T}(\omega) \left[ \left (e^{\lambda}-1 \right) f_L (\omega) g_R (\omega) +\left (e^{-\lambda}-1 \right) f_R (\omega) g_L (\omega) \right] \right \},$$9 the thermal conductance (Tang et al., 2018).

An analogous energy-counting formulation holds for heat transport in Josephson junctions. There the heat is carried by Bogoliubov quasiparticles, while Andreev reflections do not transfer energy, so heat-current fluctuations follow the Levitov–Lesovik relation (Virtanen et al., 2014). The generating function factorizes into independent quasiparticle transmission events and takes a multinomial form,

$f_\alpha(\omega)$0

where $f_\alpha(\omega)$1 labels left-to-right transfer, right-to-left transfer, or no net transfer (Virtanen et al., 2014). The essential transport objects are the energy- and phase-dependent BdG transmission eigenvalues $f_\alpha(\omega)$2. The framework gives explicit formulas for the mean heat current, noise, and fluctuation relation, and it yields a notable multichannel result: for short diffusive contacts without Dynes broadening, the heat-current statistics are independent of the superconducting phase difference, whereas phase dependence reappears when broadening is introduced (Virtanen et al., 2014).

6. Monitored transport and entanglement generalizations

A recent generalization incorporates projective or weak monitoring into the Levitov–Lesovik framework for chiral quantum Hall transport. The monitored evolution is written as a sequence of unitary propagators and occupation-number projectors,

$f_\alpha(\omega)$3

so the outgoing state is a quantum channel, not a single unitary map (Beenakker et al., 10 Apr 2025). The resulting moment generating function becomes a sum over measurement histories rather than a single determinant. Nevertheless, for a single detected outgoing mode the cumulant generating function remains binomial,

$f_\alpha(\omega)$4

and therefore

$f_\alpha(\omega)$5

Monitoring alters the effective transmission probability $f_\alpha(\omega)$6 and suppresses Aharonov–Bohm visibility in a Mach–Zehnder interferometer, but it does not destroy the binomial counting structure (Beenakker et al., 10 Apr 2025).

A different extension arises in nonequilibrium entanglement. For a one-dimensional fermionic chain with a scattering region under DC bias at zero temperature, the generating function

$f_\alpha(\omega)$7

has asymptotics

$f_\alpha(\omega)$8

The linear term is explicitly identified as a generalization of the known full counting statistics formula found by Levitov and Lesovik in the case of a $f_\alpha(\omega)$9-independent transmission factor (Fraenkel et al., 2021). The corresponding von Neumann entropy scales as

$g_\alpha(\omega)=1-f_\alpha(\omega)$0

with

$g_\alpha(\omega)=1-f_\alpha(\omega)$1

In the interpretation given there, the linear term is an entanglement analogue of Levitov–Lesovik: each momentum mode in the bias window contributes the binary entropy of the scattering-induced occupation probabilities, while the logarithmic term is a zero-temperature effect arising from sharp Fermi discontinuities (Fraenkel et al., 2021).

7. Scope, limiting cases, and recurring misunderstandings

The formula is sometimes treated as synonymous with noninteracting scattering theory, and that characterization is accurate for its standard use: the original result is a steady-state, noninteracting, elastic-scattering formula, and in weak tunneling it has Poissonian asymptotics (Skorobagatko, 2018). At the same time, several later works do not replace that formula so much as embed or generalize it. In the interacting weak-link Luttinger-liquid problem, the $g_\alpha(\omega)=1-f_\alpha(\omega)$2 limit reproduces the familiar Levitov–Lesovik structure; in free Dirac theory, the same logarithmic mode-by-mode construction survives in higher dimensions; in monitored quantum Hall transport, the determinant is replaced by a sum over measurement histories but the single-mode binomial law survives (Skorobagatko, 2018, Yoshimura, 2018, Beenakker et al., 10 Apr 2025).

A second recurrent misunderstanding is to identify the formula only with charge current. The cited literature shows that its mathematical structure extends to energy transfer by Coulomb-mediated near-field radiation and to quasiparticle heat transfer in Josephson junctions, with Bose or BdG statistical data replacing the original fermionic charge occupations (Tang et al., 2018, Virtanen et al., 2014). These are not merely analogies at the level of mean current: they are full scaled cumulant generating functions with fluctuation symmetries and higher cumulants.

A third misunderstanding concerns decoherence and phase averaging. In the monitored Mach–Zehnder problem, reduced interference visibility does not force the transferred-charge distribution away from a binomial form; by contrast, the dephasing-probe model can reproduce the average current while failing for the full counting statistics, giving the wrong noise in the symmetric interferometer with maximal dephasing (Beenakker et al., 10 Apr 2025). Likewise, in superconducting heat transport, phase dependence is not universal: it is present in general through the quasiparticle transmission eigenvalues, absent in the diffusive case without inelastic broadening, and restored once Dynes broadening is introduced (Virtanen et al., 2014).

Within the range covered here, the enduring content of the Levitov–Lesovik formula is therefore structural rather than tied to a single model: a counting field, a generating function, a logarithmic transmission-dependent form, and a direct route from microscopic scattering or Keldysh data to the full hierarchy of transport cumulants (Bernard et al., 2011, Skorobagatko, 2018).