Disorder-Averaged Dynamical Map

• Disorder-averaged dynamical map is a linear quantum channel that emerges by averaging unitary evolutions over an ensemble of disordered Hamiltonians, leading to effective decoherence.
• The channel formulation employs a time-local Lindblad generator and level-spacing functions that capture dephasing rates and coherent shifts under different disorder models.
• Mapping to a semi-infinite lattice offers a computationally efficient, geometric interpretation where unitary propagation into auxiliary sites explains the observed loss of phase coherence.

Searching arXiv for the cited work and closely related papers on disorder-averaged dynamical maps. Search query: (Gestsson et al., 2024) disorder averaged dynamical map quantum channel semi-infinite lattice Search query: (Kropf et al., 2015) effective dynamics of disordered quantum systems disorder average Lindblad A disorder-averaged dynamical map is the linear evolution law obtained by averaging unitary dynamics over an ensemble of disorder realizations. For a family of Hamiltonians $\{H(\omega)\}$ with probability density $P(\omega)$, the averaged state is

$$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$

In quantum settings this object is a completely-positive trace-preserving quantum channel; in related many-body and mean-field settings, disorder averaging can also produce effective single-site stochastic equations or self-consistent dynamical maps. Recent work has emphasized that the same averaged dynamics may be formulated as a quantum channel, as a time-local generator in Lindblad form, or as an exactly equivalent single-particle propagation problem on a semi-infinite lattice (Gestsson et al., 2024, Kropf et al., 2015).

1. Definition and basic structure

The standard construction begins from an ensemble of time-independent Hamiltonians $\{H_\lambda,p_\lambda\}$, or in the continuum $\{H(\omega),P(\omega)\}$. For each realization, the evolution is unitary: $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$ A common initial state $\rho(0)$ is mapped to the ensemble average

$$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$

or equivalently to $\mathcal E(t)[\rho(0)]$ in the notation above. Because the map is an average of unitary conjugations, it is CPTP; in the formulation of Kropf, Gneiting, and Buchleitner it is also unital (“bistochastic”) (Kropf et al., 2015).

This definition separates microscopic reversibility from macroscopic coherence loss. Each realization evolves unitarily, but the average need not preserve phase relations visible in a single sample. In the simplest spectral-disorder setting, this produces pure dephasing along a common eigenbasis. In more symmetric ensembles, such as unitarily invariant disorder, the same averaging procedure yields an isotropic depolarization channel rather than basis-selective dephasing (Kropf et al., 2015).

A common misconception is that disorder averaging is merely a numerical post-processing step. The modern literature instead treats it as an autonomous dynamical object: a map $\Phi_t$ or $P(\omega)$0 with its own exact structural representations, generators, symmetries, and computational schemes (Gestsson et al., 2024, Erpelding et al., 13 Jul 2025).

2. Channel formulation and effective generators

Once the averaged evolution is identified as a channel $P(\omega)$1, it can be differentiated and inverted to define a time-local generator

$P(\omega)$2

Except at isolated singular points where $P(\omega)$3 is not invertible, this produces an exact time-local equation. In vectorized form, if $P(\omega)$4, then $P(\omega)$5 is the matrix representation of the generator. Expansion in an orthonormal operator basis $P(\omega)$6 yields the standard Lindblad decomposition

$P(\omega)$7

with time-dependent $P(\omega)$8, $P(\omega)$9, and $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$0 (Kropf et al., 2015).

For spectral disorder, where all $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$1 commute and share the same eigenvectors, the map is diagonal in the projector basis. The central objects are the level-spacing characteristic functions

$$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$2

These determine both coherent energy shifts and dephasing rates. In the qubit case one obtains a pure-dephasing master equation in the $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$3 basis with

$$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$4

The cumulants of the level spacing then separate into even cumulants feeding $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$5 and odd cumulants feeding $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$6 (Kropf et al., 2015).

For isotropic disorder of the form $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$7 with $$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$8 Haar-random, averaging enforces a depolarizing channel

$$\bar\rho(t)=\int d\omega\,P(\omega)\,e^{-iH(\omega)t}\,\rho(0)\,e^{+iH(\omega)t}\equiv \mathcal E(t)[\rho(0)].$$9

with a single depolarization rate $\{H_\lambda,p_\lambda\}$0. Here the eigenvalue statistics enter only through the scalar function $\{H_\lambda,p_\lambda\}$1, whereas the Lindblad operators are the fixed generators of $\{H_\lambda,p_\lambda\}$2 (Kropf et al., 2015).

3. Exact mapping to a semi-infinite lattice

A central development is the exact equivalence between disorder-averaged dynamics and single-particle motion on a semi-infinite lattice. In the formulation of “Equivalence of dynamics of disordered quantum ensembles and semi-infinite lattices,” one embeds the full ensemble into an enlarged Hilbert space,

$\{H_\lambda,p_\lambda\}$3

where $\{H_\lambda,p_\lambda\}$4 is $\{H_\lambda,p_\lambda\}$5-independent and $\{H_\lambda,p_\lambda\}$6 carries all disorder dependence. One then introduces orthonormal polynomials $\{H_\lambda,p_\lambda\}$7 with respect to the measure $\{H_\lambda,p_\lambda\}$8,

$\{H_\lambda,p_\lambda\}$9

The unitary basis change

$\{H(\omega),P(\omega)\}$0

transforms $\{H(\omega),P(\omega)\}$1 into a tight-binding Hamiltonian on the semi-infinite index $\{H(\omega),P(\omega)\}$2 (Gestsson et al., 2024).

For linear disorder $\{H(\omega),P(\omega)\}$3, the resulting lattice has onsite terms $\{H(\omega),P(\omega)\}$4 and nearest-neighbor hoppings $\{H(\omega),P(\omega)\}$5, where $\{H(\omega),P(\omega)\}$6 and $\{H(\omega),P(\omega)\}$7 are the recurrence coefficients of the orthogonal polynomials associated with $\{H(\omega),P(\omega)\}$8. In monic normalization,

$\{H(\omega),P(\omega)\}$9

with

$$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$0

Equivalently, $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$1 and $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$2 may be expressed in terms of disorder moments or cumulants via determinantal formulas (Gestsson et al., 2024).

The operational statement is exact: propagate a single particle along the semi-infinite lattice and then trace out the lattice coordinate $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$3; the result reproduces the original disorder-averaged channel $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$4. The reverse map also exists: lattice dynamics can be computed by averaging over an appropriate disorder ensemble (Gestsson et al., 2024).

4. Geometric interpretation of coherence loss

The lattice representation turns disorder averaging into a one-dimensional geometry. If the initial state is independent of $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$5, only the polynomial mode $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$6 is populated, so the state is localized at the surface site $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$7. As time evolves, amplitude propagates into sites $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$8. Tracing out $$U_\lambda(t)=\exp[-(i/\hbar)H_\lambda t],\qquad U(\omega,t)=\exp[-iH(\omega)t].$$9 then destroys coherence exactly as dephasing would. In this picture, the random phases carried by $\rho(0)$0 are converted into spatial information stored in the bulk coordinate $\rho(0)$1, and motion away from $\rho(0)$2 is the loss of phase information (Gestsson et al., 2024).

This geometric interpretation clarifies a recurrent conceptual point. The loss of coherence in the disorder-averaged state does not imply non-unitarity at the level of each realization. Rather, the averaged description forgets which disorder realization was present. The semi-infinite lattice makes that forgetting explicit: what appears as dephasing in the reduced channel is unitary transport into auxiliary degrees of freedom (Gestsson et al., 2024).

A plausible implication is that disorder-induced decoherence can often be analyzed with tools native to one-dimensional tight-binding problems rather than only with direct sampling over random Hamiltonians. This is precisely the computational logic of the map.

5. Canonical examples

The qubit pure-dephasing ensemble is the simplest explicit case. For

$\rho(0)$3

the averaged channel acts on

$\rho(0)$4

as

$\rho(0)$5

The corresponding lattice description consists of two decoupled chains for the $\rho(0)$6 and $\rho(0)$7 sectors. The $\rho(0)$8 chain is trivial, while the $\rho(0)$9 chain carries the onsite coefficients $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$0 and hoppings $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$1. The amplitude returning to $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$2 on the $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$3 branch is exactly $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$4. The spectrum of that chain is the support of $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$5, for example a finite band for uniform or semicircle disorder (Gestsson et al., 2024).

The reverse-map example starts from a semi-infinite uniform chain with constant hopping $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$6 and zero onsite terms. Its local propagator at $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$7 can be reinterpreted as the disorder-averaged return amplitude of an ensemble whose $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$8 is the Wigner semicircle of radius $$\bar\rho(t)=\langle U_\lambda(t)\rho(0)U_\lambda^\dagger(t)\rangle$$9. Local density and related lattice observables may then be computed by sampling $\mathcal E(t)[\rho(0)]$0 according to the semicircle law and averaging $\mathcal E(t)[\rho(0)]$1; closed-form integral representations, including Bessel functions, arise naturally (Gestsson et al., 2024).

A different exact family is provided by $\mathcal E(t)[\rho(0)]$2-potent Hamiltonians satisfying

$\mathcal E(t)[\rho(0)]$3

In this case the infinite series for the averaged map collapses to a finite expansion over the $\mathcal E(t)[\rho(0)]$4 linearly independent operators $\mathcal E(t)[\rho(0)]$5. The resulting map is exact for arbitrary evolution times, and one can write Kraus forms with at most $\mathcal E(t)[\rho(0)]$6 terms and corresponding time-local generators when the superoperator is invertible. The qubit involutory case $\mathcal E(t)[\rho(0)]$7 reproduces an exact dephasing channel with decoherence function $\mathcal E(t)[\rho(0)]$8; Gaussian and uniform disorder then separate CP-divisible from non-Markovian behavior in the sense of the time-dependent rates and standard witnesses (Santra et al., 2024).

6. Stochastic, influence-functional, and DMFT generalizations

The disorder-averaged dynamical map also admits exact stochastic unravelings. For Gaussian operator disorder

$\mathcal E(t)[\rho(0)]$9

the averaged channel

$\Phi_t$0

can be written as

$\Phi_t$1

where $\Phi_t$2 obeys an Itô SDE and the average is over Brownian-bridge processes. The corresponding exact time-local master equation has generator

$\Phi_t$3

which is manifestly Lindblad. This representation supports a convergent stochastic Dyson series whose first term already captures diffusive disorder broadening, and it has been applied to density of states, spectral form factor, and out-of-time-order correlators in the Anderson model (Kurecic et al., 2018).

In disordered Floquet many-body systems, exact averaging can instead be encoded in an influence matrix $\Phi_t$4, a functional of forward and backward trajectories for a selected degree of freedom. For the kicked Ising chain, averaging over on-site disorder yields a non-local-in-time constraint and a linear self-consistency equation

$\Phi_t$5

whose solution determines temporal correlators. In the many-body localized phase, the temporal entanglement of $\Phi_t$6 grows very slowly, enabling efficient matrix-product-state computations beyond exact-diagonalization limits (Sonner et al., 2020).

In non-equilibrium DMFT for the Anderson-Hubbard model, averaging static Gaussian disorder on the Keldysh contour produces an extra nonlocal-in-time density-density interaction

$\Phi_t$7

Under the standard non-equilibrium DMFT approximation, this yields a closed impurity problem with self-energy

$\Phi_t$8

together with the Bethe-lattice self-consistency $\Phi_t$9. The resulting closed set of equations constitutes the disorder-averaged dynamical map for the quench problem and captures suppression of the weak-coupling prethermal plateau and damping of strong-coupling collapse-and-revival oscillations (Rangi et al., 2024).

A related but classical-statistical construction appears in compact DMFT for oscillator networks with quenched randomness. After averaging over random couplings in the thermodynamic limit, the $P(\omega)$00-oscillator problem reduces to a single-oscillator SDE on $P(\omega)$01 driven by a deterministic mean field and a self-consistent colored Gaussian process $P(\omega)$02, with covariance fixed by a circular two-time correlator $P(\omega)$03. In the limit $P(\omega)$04, the formalism reproduces Ott-Antonsen and Kuramoto reductions (Reddy, 10 Mar 2026).

7. Computational significance and emergent structure

The semi-infinite-lattice equivalence has immediate numerical consequences. Instead of sampling many disorder realizations or using high-order quadrature, one simulates a single sparse lattice Hamiltonian with only three-term couplings, truncated at some $P(\omega)$05. The construction is exact up to truncation, and a Lieb-Robinson-type bound can be used to estimate how far amplitude can propagate in time $P(\omega)$06, thereby fixing $P(\omega)$07 (Gestsson et al., 2024).

Disorder averaging can also restore symmetry at the level of superoperators. In the all-to-all random Ising model with transverse field, permutation-invariant disorder implies that the averaged map commutes with the symmetric-group action, so the dynamics preserves irreducible $P(\omega)$08-sectors. The fully symmetric sector can be represented in the orthonormal basis of totally symmetric Pauli strings $P(\omega)$09, and its dimension is

$P(\omega)$10

in place of the full operator-space dimension $P(\omega)$11. Short-time generalized-cumulant expansions and weak-disorder expansions then produce polynomial-cost approximations to the disorder-averaged dynamics for expectation values (Erpelding et al., 13 Jul 2025).

These developments collectively shift the interpretation of disorder averaging. It is not only a way to smooth sample-to-sample fluctuations; it is a structured dynamical object with exact channel representations, exact or time-local generators, auxiliary-lattice realizations, stochastic unravelings, self-consistency equations, and symmetry sectors. Across these formulations, the central physical content remains the same: static heterogeneity in microscopic Hamiltonians is re-expressed as effective decoherence, temporal memory, or auxiliary-coordinate propagation in the averaged dynamics (Gestsson et al., 2024, Kropf et al., 2015).