Liouvillian Spectral Topology

• Liouvillian spectral topology is the study of the full superoperator spectrum governing open-system dynamics, including quantum jumps and non-Markovian effects.
• It uncovers exceptional-point structures and hybrid winding invariants that distinguish Liouvillian behavior from effective non-Hermitian Hamiltonian models.
• The field employs advanced methods like algebro-geometric analysis and exact diagonalization to classify spectral phase transitions and dissipative topological phenomena.

Searching arXiv for papers on Liouvillian spectral topology and closely related Liouvillian spectral phenomena. Liouvillian spectral topology is the study of topological and topology-adjacent structure in the spectrum of the Liouvillian superoperator that governs open-system density-matrix evolution, rather than in an effective non-Hermitian no-jump Hamiltonian alone. In current usage, the subject includes at least three distinct but overlapping strands: Liouvillian exceptional-point topology, where branch-point singularities and defective decay modes organize the complex spectrum; Liouvillian point-gap and skin-effect topology, where spectral winding and boundary accumulation appear directly at the superoperator level; and topological fingerprints in relaxation spectra, where Hamiltonian or dissipative topology is encoded in Liouvillian gaps and mode structure without necessarily producing a standalone invariant. Recent work has emphasized that this Liouvillian perspective is not reducible to Hamiltonian topology dressed by decay, because quantum jumps, biorthogonal mode geometry, memory embeddings, and symmetry-sector structure can qualitatively alter the spectrum and its dynamical consequences (Zhang et al., 6 Dec 2025, Long et al., 25 Feb 2026, Kopciuch et al., 3 Jun 2025).

1. Liouvillian spectra as topological objects

The Liouvillian $\mathcal L$ is the generator of open-system dynamics in operator space, typically through a master equation of the form

$\frac{d\rho}{dt}=\mathcal L \rho .$

Unlike a Hamiltonian, $\mathcal L$ acts on density operators, so its eigenvectors are decay modes or eigenoperators, not state vectors. This distinction is central to the modern formulation of Liouvillian spectral topology: the relevant complex spectrum is that of the full generator of open-system dynamics, including quantum jumps, not merely of the effective non-Hermitian Hamiltonian governing no-jump trajectories (Zhang et al., 6 Dec 2025).

A recurring theme across the literature is that Liouvillian spectral structure is physically richer than the spectrum of an effective non-Hermitian Hamiltonian. In atomic-vapor models, the inclusion of quantum jumps can shift, split, lower the order of, or remove degeneracies predicted by the Hamiltonian description, and can even alter whether a given exceptional degeneracy exists at all (Kopciuch et al., 3 Jun 2025). In non-Markovian settings, the relevant object may be an extended Liouvillian superoperator acting on an enlarged Hilbert space that includes memory degrees of freedom, such as a pseudomode representation of a structured reservoir (Zhang et al., 6 Dec 2025).

This broader viewpoint has two immediate consequences. First, topological analysis often requires working in Liouville space rather than Hilbert space. Second, the physically meaningful spectral structures may depend on left-right biorthogonality, quantum-jump recycling terms, symmetry-sector decomposition, or memory embeddings in ways with no direct Hamiltonian analogue (Molina, 1 Feb 2026, Zhang et al., 6 Dec 2025).

2. Exceptional-point structures in Liouville space

One major branch of the subject concerns Liouvillian exceptional points (LEPs): parameter values where Liouvillian eigenvalues and Liouvillian eigenvectors coalesce. Because Liouvillian eigenvectors are superoperator eigenmodes of density-matrix evolution, LEPs describe defective decay-mode structure rather than defective wavefunction dynamics (Zhang et al., 6 Dec 2025).

A particularly sharp example is the non-Markovian qubit–reservoir system studied in "Exploring the topology induced by non-Markovian Liouvillian exceptional points" (Zhang et al., 6 Dec 2025). There, a qubit coupled to a structured Lorentzian reservoir is embedded into a pseudomode description, and the density operator of the combined qubit–pseudomode system evolves under

$\frac{d\rho}{dt}={\cal L}\rho ,$

with

$${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$

The effective non-Hermitian Hamiltonian

$$H_{NH}=\Omega \left(b^{\dagger }\left\vert g\right\rangle \left\langle e\right\vert +b\left\vert e\right\rangle \left\langle g\right\vert \right)+\Delta \left\vert e\right\rangle \left\langle e\right\vert -\frac{i\kappa }{2} b^{\dagger }b$$

captures only conditional no-jump evolution, whereas $\mathcal L$ includes the recycling term $\kappa b\rho b^\dagger$. The paper’s central claim is that this difference is topologically consequential: the Liouvillian hosts exceptional structures that are “significantly different” from those of the corresponding non-Hermitian Hamiltonian (Zhang et al., 6 Dec 2025).

In the experimentally relevant zero- and one-excitation sector, the extended Liouvillian becomes a $9\times 9$ non-Hermitian matrix. At

$\{\Omega=\kappa/4,\ \Delta=0\},$

the spectrum contains a twofold LEP2, meaning two distinct second-order Liouvillian exceptional points coincide in parameter space and in complex spectral position, at

$\frac{d\rho}{dt}=\mathcal L \rho .$0

The same model also contains an LEP3, but the topological analysis focuses on the twofold LEP2 (Zhang et al., 6 Dec 2025).

This example is important because it shows that Liouvillian exceptionality can be structurally more elaborate than Hamiltonian exceptionality. The coincident LEP2 is not a mere relabeling of a Hamiltonian EP, but a genuinely Liouvillian object produced by the combination of non-Markovian memory and jump dynamics.

3. Hybrid invariants and branch topology of Liouvillian EPs

The topological invariant used around a Liouvillian EP is a spectral winding number defined directly from the complex Liouvillian eigenvalues. For an eigenvalue branch $\frac{d\rho}{dt}=\mathcal L \rho .$1,

$\frac{d\rho}{dt}=\mathcal L \rho .$2

where $\frac{d\rho}{dt}=\mathcal L \rho .$3 is the smallest positive integer such that

$\frac{d\rho}{dt}=\mathcal L \rho .$4

This is the natural branch-point winding for non-Hermitian spectra: around an EP2, one generally needs $\frac{d\rho}{dt}=\mathcal L \rho .$5, since a single encircling exchanges the two branches (Zhang et al., 6 Dec 2025).

In the non-Markovian qubit–pseudomode system, a single loop in $\frac{d\rho}{dt}=\mathcal L \rho .$6 space enclosing $\frac{d\rho}{dt}=\mathcal L \rho .$7 produces two opposite half-integer windings simultaneously. The branch pair $\frac{d\rho}{dt}=\mathcal L \rho .$8 yields

$\frac{d\rho}{dt}=\mathcal L \rho .$9

while the branch pair $\mathcal L$0 yields

$\mathcal L$1

The paper interprets this as a hybrid topological invariant: the same physical loop around the same singular point carries two opposite half-integer charges depending on which Liouvillian branch sector is tracked (Zhang et al., 6 Dec 2025).

This differs sharply from the effective non-Hermitian Hamiltonian description, where the corresponding Hamiltonian EP2 has a definite winding number. The Liouvillian invariant is therefore not a scalar attached to the point independently of branch choice; it is sector-dependent. A plausible implication is that Liouvillian topology may require classification data richer than a single integer whenever coincident branch structures occur in Liouville space.

The mathematical origin of this hybrid behavior lies in the square-root branch structure of distinct Liouvillian eigenvalue pairs. The relevant pairs define overlapping two-sheeted Riemann surfaces with opposite orientations in the complex spectral plane. This is why the same loop can produce opposite windings without contradiction (Zhang et al., 6 Dec 2025).

4. Symmetry-protected winding and the Liouvillian skin effect

A second major strand of the subject concerns point-gap topology and the Liouvillian skin effect. Here the relevant object is often a translationally invariant Liouvillian or Floquet-Liouvillian whose complex spectrum winds around a reference point under periodic boundary conditions, with open boundaries producing macroscopic boundary accumulation of eigenmodes.

A particularly explicit invariant-based formulation appears in "Symmetry-protected control of Liouvillian topological phases via Hamiltonian band topology" (Long et al., 25 Feb 2026). The paper studies a 1D two-band lattice with coherent Hamiltonian

$\mathcal L$2

and Lindblad dynamics

$\mathcal L$3

When both the Hamiltonian and jump operators respect the same chiral symmetry, the Hamiltonian winding number

$\mathcal L$4

acts as a control parameter for Liouvillian point-gap topology (Long et al., 25 Feb 2026).

The Liouvillian point-gap invariant near the steady-state eigenvalue is defined as

$\mathcal L$5

In the symmetry-aligned class treated there, the Hamiltonian winding enters the Liouvillian through an exact gauge-like shift, yielding

$\mathcal L$6

with $\mathcal L$7 for a single-harmonic Hamiltonian (Long et al., 25 Feb 2026). This establishes a symmetry-protected correspondence between coherent band topology and Liouvillian spectral winding.

The physical consequence is a Liouvillian skin effect under open boundary conditions: the steady state and slow modes accumulate near one edge, with the localization direction controlled by the sign of $\mathcal L$8. The paper further shows that lattice parity influences the corresponding bulk-boundary relation and coherence properties of the steady state, so the dissipative bulk-boundary correspondence is more delicate than in closed Hermitian band theory (Long et al., 25 Feb 2026).

Closely related but less invariant-centered are the skin-effect studies "Liouvillian Skin Effect: Slowing Down of Relaxation Processes without Gap Closing" (Haga et al., 2020), "Optical pumping through the Liouvillian skin effect" (Cai et al., 2024), and "Quantum Pontus-Mpemba Effect Enabled by the Liouvillian Skin Effect" (Longhi, 20 Jan 2026). These works emphasize that asymmetric dissipation can produce boundary-localized left and right Liouvillian eigenmodes at opposite edges, leading to strong non-normality and boundary sensitivity. In the 2020 work, the longest observable relaxation time obeys

$\mathcal L$9

so it can diverge with system size even when the Liouvillian gap remains finite (Haga et al., 2020). In the 2026 Pontus–Mpemba work, the same left-right edge localization under asymmetric dissipation is exploited to suppress overlap with the slow skin mode and thereby accelerate relaxation without changing the asymptotic decay rate (Longhi, 20 Jan 2026).

These studies are topological in a broader, point-gap-and-skin sense rather than in a strict invariant-classification sense. They show that Liouvillian spectral topology is often inseparable from non-normality, left-right eigenmode separation, and boundary-conditioned overlaps.

5. Floquet, jumps, and many-body exceptional spectral phases

A third line of development concerns situations where Liouvillian topology is reshaped by quantum jumps or organized into phase-like many-body structures.

"Liouvillian topology and non-reciprocal dynamics in open Floquet chains" (Koch et al., 20 Nov 2025) starts from a driven dissipative lattice with known non-Hermitian Floquet topology in an effective no-jump approximation and then studies the full Floquet-Liouvillian. In the no-jump limit, the vectorized propagator factorizes as

$\frac{d\rho}{dt}={\cal L}\rho ,$0

embedding the non-Hermitian Floquet topology into Liouville space. But once quantum jumps are restored, this factorization is broken, spectral outliers appear, and a distinct jump-induced topological phase emerges. The paper defines windings of individual Liouvillian bands by

$\frac{d\rho}{dt}={\cal L}\rho ,$1

and argues that the sign of the resulting Liouvillian winding correlates with the direction of non-reciprocal transport and boundary localization (Koch et al., 20 Nov 2025). The central lesson is that full Liouvillian topology is not a trivial lift of no-jump topology; jumps can generate qualitatively new phases.

In a different direction, "Spectroscopic Signatures of a Liouvillian Exceptional Spectral Phase in a Collective Spin" (Molina, 1 Feb 2026) studies a many-body collective-spin Lindbladian with

$\frac{d\rho}{dt}={\cal L}\rho ,$2

and jump operators

$\frac{d\rho}{dt}={\cal L}\rho ,$3

The paper’s notion of exceptional spectral phase is thermodynamic: as $\frac{d\rho}{dt}={\cal L}\rho ,$4, a finite fraction of the Liouvillian spectrum becomes defective, consisting of an extensive set of second-order exceptional points rather than isolated degeneracies. The defective Jordan structure gives a resolvent

$\frac{d\rho}{dt}={\cal L}\rho ,$5

on the Jordan subspace, producing second-order poles and corresponding “super-Lorentzian” contributions to frequency-resolved spectra (Molina, 1 Feb 2026). Here the relevant “topology” is not a band invariant but the spectral organization of defectiveness across symmetry sectors and parameter regions.

Together, these works show that Liouvillian spectral topology now spans both few-mode branch-point topology and many-body spectral-phase organization. A plausible implication is that the subject is moving toward a broader conception in which defectiveness, point-gap winding, and dynamical accessibility all contribute to classification.

6. Methods, diagnostics, and scope of the field

Several methodological directions have become especially important.

One is algebro-geometric EP analysis. "Characterizing Liouvillian Exceptional Points Through Newton Polygons and Tropical Geometry" (P et al., 9 Oct 2025) shows that Newton polygons and tropical geometry can locate and classify Liouvillian EPs directly from the characteristic polynomial

$\frac{d\rho}{dt}={\cal L}\rho ,$6

The slope of a lower Newton edge determines the leading Puiseux scaling

$\frac{d\rho}{dt}={\cal L}\rho ,$7

with $\frac{d\rho}{dt}={\cal L}\rho ,$8 given by minus the edge slope, thereby identifying EP order and perturbation anisotropy (P et al., 9 Oct 2025). The method is explicitly applied to a dissipative spin-$\frac{d\rho}{dt}={\cal L}\rho ,$9 Liouvillian and a dissipative superconducting-qubit Liouvillian, where different perturbation directions reveal square-root, cube-root, linear, or invariant splitting sectors.

Another is exact Liouvillian diagonalization beyond quadratic models. "Exact diagonalization of a non-quadratic bosonic Liouvillian with two-body loss" (Tokieda, 29 Mar 2026) gives the full spectrum

$${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$0

for a single bosonic mode with one-body loss, two-body loss, and Kerr interaction, together with explicit left and right eigenoperators in terms of confluent hypergeometric functions (Tokieda, 29 Mar 2026). Although no topological invariant is defined, the paper provides a rare complete spectral resolution of a genuinely non-quadratic Liouvillian, which is exactly the sort of foundation needed for future invariant-based analyses.

A third important theme is the limitation of bare eigenvalue gaps as dynamical diagnostics. "Symmetrized Liouvillian Gap in Markovian Open Quantum Systems" (Mori et al., 2022) proves that steady-state autocorrelations obey the rigorous bound

$${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$1

in terms of a symmetrized Liouvillian gap $${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$2, while the standard eigenvalue gap need not provide a correct finite-time bound away from detailed balance (Mori et al., 2022). This is not a topology paper, but it is directly relevant to Liouvillian spectral topology because it clarifies that non-normal Liouvillian dynamics can invalidate naive eigenvalue-based intuition.

Finally, work such as "Photonic chiral state transfer near the Liouvillian exceptional point" (Gao et al., 17 Jan 2025) underscores that Liouvillian EP topology can be experimentally real yet dynamically transient. In that two-level open-system simulator, the chirality measure

$${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$3

is nonzero only for intermediate encircling times and collapses at long times because the system relaxes to the steady state whenever a finite Liouvillian gap is present (Gao et al., 17 Jan 2025). This sharply distinguishes Liouvillian EP topology from more familiar Hamiltonian EP scenarios.

6. Scope, limits, and open problems

The current literature does not yet furnish a single unified classification of Liouvillian spectral topology. The field contains at least three partially distinct notions:

Regime	Central object	Representative result
Exceptional-point topology	Branch points and defective Liouvillian modes	Hybrid $${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$4 winding at a twofold LEP2 (Zhang et al., 6 Dec 2025)
Point-gap/skin topology	Spectral winding and boundary accumulation	Symmetry-protected relation between $${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$5 and $${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$6 (Long et al., 25 Feb 2026)
Topological fingerprints in decay spectra	Gap structure and decay-mode organization	Topological phase dependence of the Liouvillian gap in an open Kitaev chain (Kavanagh et al., 2024)

"Topological fingerprints in Liouvillian gaps" (Kavanagh et al., 2024) is exemplary of the third category. In an open Kitaev chain with local Hermitian dephasing, the weak-dissipation Liouvillian gap is

$${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$7

which becomes insensitive to chemical potential throughout the topological phase $${\cal L}\rho =i\left[\rho H_{NH}^{\dagger }-H_{NH}\rho \right]+\kappa b\rho b^{\dagger }.$$8 but not outside it (Kavanagh et al., 2024). This is topology reflected in Liouvillian relaxation rates, not a topological invariant of the Liouvillian itself.

Several limitations recur across the literature. Many papers are model-specific and do not provide general classification principles. Skin-effect papers often study topology-adjacent non-normal geometry without computing explicit invariants (Haga et al., 2020, Longhi, 20 Jan 2026). Exceptional-phase papers analyze defectiveness and resolvent structure rather than band-topological indices (Molina, 1 Feb 2026). Non-Markovian work shows that enlarged Liouvillians can host new topology, but leaves open how universal such hybrid invariants are beyond pseudomode embeddings (Zhang et al., 6 Dec 2025).

Open directions repeatedly identified include broader classification of non-Markovian Liouvillian topology, many-body and higher-dimensional extensions, higher-order Liouvillian EPs, relations between point-gap topology and experimentally accessible observables, and dynamical protocols such as adiabatic encircling, mode switching, and geometric response (Zhang et al., 6 Dec 2025, Molina, 1 Feb 2026, Koch et al., 20 Nov 2025).

In that sense, Liouvillian spectral topology is best understood not as a finished classification scheme but as an emerging framework for organizing the complex spectra of open-system generators. Its defining insight is that the topology of dissipative quantum matter often resides not in an effective Hamiltonian but in the spectrum of the full Liouvillian, where jumps, memory, symmetry, and biorthogonal mode geometry become topological ingredients in their own right.