Mixed-State Topological Phases

• Intrinsically mixed-state topological phases are defined by the topological features inherent to density matrices rather than pure-state ground states, leveraging fast local Lindbladian evolution.
• They utilize distinct strong and weak symmetry concepts and doubled Hilbert space formulations to classify phase behavior and uncover unique spectral and dynamical properties.
• In 2+1 dimensions, these phases reveal novel phenomena such as strong-to-weak 1-form symmetry breaking and premodular anyon theories, suggesting robust topological orders even under decoherence.

Intrinsically mixed-state topological phases are phases of density matrices whose defining topological structure is formulated at the level of mixed states rather than by reference to a parent pure-state ground state alone. Across the recent literature, the term encompasses several partially overlapping ideas: phase equivalence under fast local open-system dynamics, mixed-state symmetry patterns built from strong and weak symmetry, topological order encoded by premodular rather than modular anyon data, and axiomatic notions of locally recoverable, spatially uniform mixed-state fixed points. The central question is whether mixedness merely degrades known pure-state topological phases or can instead generate phases with no pure-state counterpart. Current work gives concrete realizations of the latter in two dimensions, especially through decoherence- or disorder-induced $1$-form symmetry patterns, while also showing that some purported mixed-state phases are better understood as extensions of ordinary pure-state SPT logic (Coser et al., 2018, Zhang et al., 2024, Ellison et al., 2024, Yang et al., 4 Jun 2025).

1. Foundational definitions and the move from pure states to density matrices

An important early many-body extension of topology to mixed states treated a density matrix $\rho$ through the spectral manifolds of $-\log \rho$, with the physically motivated regime being quasi-thermal states for which $\hat h\equiv -\log\rho$ is local or quasi-local. In that framework, topological equivalence required the same spectral structure and local-unitary equivalence of corresponding spectral manifolds, so the topology was intrinsic to $\rho$ but still tied to effective Hamiltonian structure rather than to arbitrary density matrices (Grusdt, 2016).

A more explicitly dynamical formulation defines phases directly on density operators by reachability under local, time-independent, fast Lindbladian evolution. Two states are in the same phase if

$\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$

with $\fastto$ meaning convergence in trace norm in sublinear time, typically $t\gtrsim \mathrm{poly}\log N$, under a local Lindbladian semigroup, optionally with locality-preserving ancillas. This relation is an equivalence relation, supports a partial order interpreted as “topological complexity,” and yields smooth intra-phase paths because local observables vary analytically along finite-time Lindbladian evolution (Coser et al., 2018).

That definition has two lasting consequences. First, it makes mixed-state topology operational: a density matrix is topologically nontrivial if local dissipative dynamics cannot generate its long-range structure in less than the Lieb-Robinson time scale. Second, it is generally coarser than pure-state Hamiltonian classification. In particular, while same Hamiltonian phase implies same Lindbladian phase for pure states, the converse fails sharply in one dimension: symmetric 1D SPT distinctions collapse, and there is only one SPT phase in 1D independently of the symmetry group. In two dimensions, by contrast, the paper proves for $\mathbb Z_n$ quantum doubles that richer order cannot be created rapidly from poorer order, which already suggests that mixed-state topology may be genuinely nontrivial in $2+1$ dimensions even when it ceases to coincide with pure-state classification (Coser et al., 2018).

2. Strong and weak symmetry, average symmetry, and doubled-space formulations

A recurring mixed-state-specific ingredient is the distinction between strong and weak symmetry. For a density matrix $\rho$0, strong symmetry means

$\rho$1

whereas weak symmetry means

$\rho$2

These notions coincide for pure states but differ for mixed states, and that difference underlies both mixed-state SPT theory and intrinsically mixed topological order. In particular, average symmetry is a symmetry of the ensemble or density matrix rather than of each component state, and it has no pure-state analogue in the same form (Ma et al., 2024, Zhang et al., 2024).

One systematic route uses the doubled Hilbert space. A density matrix is mapped to its Choi state

$\rho$3

and a mixed state is called short-range entangled when the normalized Choi state is an SRE pure state in $\rho$4. Exact symmetry $\rho$5 doubles to $\rho$6, average symmetry $\rho$7 acts diagonally, and Hermiticity produces an antiunitary modular conjugation $\rho$8. Within this framework, mixed-state SPT phases are classified not by all doubled-space SPT data, but by those compatible with Hermiticity and positivity. For unitary exact and average symmetry, the allowed classification is

$\rho$9

A central conclusion is negative: positivity rules out many naively allowed doubled-space invariants, including phases protected solely by average symmetry, joint $-\log \rho$0-$-\log \rho$1 invariants, and $-\log \rho$2-only SPT sectors (Ma et al., 2024).

The same strong/weak distinction appears in the theory of average symmetry-protected topological phases. In one dimension, decohered ASPT phases with exact symmetry $-\log \rho$3 and average symmetry $-\log \rho$4 are classified by

$-\log \rho$5

where the mixed-state-specific term decorates average-symmetry defects with exact-symmetry charges. More generally, bosonic decohered ASPTs with $-\log \rho$6 are organized by a defect-decoration spectral sequence, while disordered ASPTs differ because $-\log \rho$7-dimensional charge decorations become trivial as phase labels. The concept of an intrinsically mixed ASPT is then precise: it exists only when part of the symmetry is average, and it cannot be realized as an ordinary pure-state SPT with all relevant symmetries exact (Ma et al., 2023).

3. Strong-to-weak $-\log \rho$8-form symmetry breaking and intrinsically mixed topological order in $-\log \rho$9 dimensions

The sharpest proposals for intrinsically mixed-state topological order in $\hat h\equiv -\log\rho$0 dimensions identify it with strong-to-weak spontaneous symmetry breaking of $\hat h\equiv -\log\rho$1-form symmetries. In this picture, pure topological order is a phase with spontaneously broken anomalous $\hat h\equiv -\log\rho$2-form symmetries, but mixed states admit an additional pattern: a symmetry can be strong at the level of the density matrix while only weak symmetry remains exact after decoherence. The resulting SW-SSB pattern has no pure-state analogue and is taken as the hallmark of intrinsically mixed topological order (Zhang et al., 2024).

The corresponding phase notion is finer than two-way channel connectivity. Two locally correlated density matrices are equivalent only if they are connected by finite-time local Lindbladian evolution while maintaining finite Rényi-2 Markov length. This refinement separates the toric-code “classical memory” state from the infinite-temperature state even though the two are two-way channel connected. In the toric-code example, the pure topological phase has Choi state equal to two copies of toric code, the intrinsically mixed SW-SSB phase has Choi state equal to one copy of toric code, and the weakly symmetric trivial phase has trivial Choi order. Correspondingly, the mixed-state Levin-Wen Rényi-2 conditional mutual information takes the fixed-point values $\hat h\equiv -\log\rho$3, $\hat h\equiv -\log\rho$4, and $\hat h\equiv -\log\rho$5, so the intrinsically mixed phase is characterized by a “half-reduction” of the Choi-state topological entanglement structure (Zhang et al., 2024).

A concrete stabilizer realization appears in the toric code under stochastic unmonitored $\hat h\equiv -\log\rho$6-diagonal projective measurement without monitoring. The fully decohered state is

$\hat h\equiv -\log\rho$7

so decoherence reorganizes the stabilizer structure into composite vertex-plaquette operators rather than simply destroying order. The transition occurs around $\hat h\equiv -\log\rho$8, the long-range entanglement diagnosed by negativity persists while the toric-code logical qubit disappears, and the phase is characterized by restoration of a weak $\hat h\equiv -\log\rho$9-form symmetry generated by

$\rho$0

together with proliferation of weak fermionic anyons $\rho$1. Finite-size scaling of the disorder susceptibility gives

$\rho$2

with $\rho$3 very close to the 2D percolation value $\rho$4, which motivates the interpretation of the transition as “percolation of decoherence” (Kuno et al., 2024).

A complementary route starts from pure topological order and interprets local correlated decoherence as “gauging out” anyons rather than condensing them in the ordinary sense. For a parent anyon theory $\rho$5 and a decohered subset $\rho$6, the surviving coherent topological order is the centralizer

$\rho$7

Because gauging out confines nontrivially braiding anyons without identifying the gauged anyon with vacuum, $\rho$8 is generically non-modular and can contain transparent anyons. This leads to intrinsically mixed-state topological order in the sense that the mixed state’s strong $\rho$9-form symmetry category is premodular rather than modular. In Abelian stabilizer models this construction produces topological subsystem codes; examples include $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$0, $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$1, $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$2, non-Abelian mixed descendants of doubled Ising, and chiral mixed-state phases obtained by decohering one sector of a parent theory (Sohal et al., 2024).

4. Premodular anyon theories, axiomatic fixed points, and mixed-state diagnostics

A broader classification program argues that mixed-state topological orders in two dimensions should be organized by emergent strong generalized $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$3-form symmetry and associated premodular anyon theories. The phase relation is two-way connectivity by quasi-local quantum channels, but restricted to mixed states that are Rényi-1 and Rényi-2 locally correlated and purifiable to a gapped ground state. In the Abelian case, if two states are two-way connected and have anyon theories $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$4, then their minimal theories obtained after quotienting transparent bosons agree,

$\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$5

The paper identifies two mechanisms by which quasi-local channels act on topological data: incoherent proliferation of anyons, which reduces to a commutant of the proliferated anyons, and “classical gauging,” which symmetrizes anyons and extends the theory by transparent bosons. On that basis it conjectures that mixed-state topological orders are classified by premodular anyon theories (Ellison et al., 2024).

An alternative, explicitly axiomatic framework starts from three fixed-point conditions: local recoverability $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$6, absence of long-range correlations $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$7, and spatial uniformity $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$8 in the appropriate geometries. The resulting “mixed-state bootstrap program” defines topological mixed-state fixed points and then promotes them to phases by coarse-graining. The fundamental object is the information convex set $\rho_0\sim \rho_1 \quad\Longleftrightarrow\quad \rho_0\fastto \rho_1\ \text{and}\ \rho_1\fastto \rho_0,$9, the convex set of globally distinct but locally indistinguishable states compatible with the same local topological fabric. This yields a notion of classical and quantum logical memories directly for mixed states and quantifies them by the memory capacity

$\fastto$0

In this framework, dephased toric-code fixed points are topologically nontrivial even though they are classically preparable by finite-depth local channels, and non-Abelian decohered quantum doubles retain a hierarchy of secret-sharing constraints because simultaneous conjugacy-class data remain robust under decoherence (Yang et al., 4 Jun 2025).

Mixed-state chirality introduces an additional difficulty. In pure states, chirality is tied to gapless edges, modular commutators, entanglement-spectrum structure, and thermal Hall response. For mixed states, these diagnostics can fail simultaneously: the paper constructs chiral mixed-state phases with short-range-correlated boundaries, commuting-projector reduced density matrices, and vanishing modular commutator. The proposed replacement is parent-relative and nonlinear. If $\fastto$1 is a decohered state and $\fastto$2, then the Umegaki relative entropy

$\fastto$3

distinguishes surviving anyons from proliferated ones by whether it grows linearly with $\fastto$4. The mixed-state chiral central charge is then reconstructed from the surviving anyon set via

$\fastto$5

This establishes that mixed-state chirality demands intrinsically new diagnostics beyond direct analogues of pure-state probes (Sun et al., 24 Jun 2026).

5. Mixed-state SPT phases and the boundary of intrinsicity

Not every mixed-state topological phase is intrinsically mixed in the strongest sense. Some results extend ordinary pure-state SPT structures into the density-matrix setting without producing new topological sectors. A representative example uses finite-depth local channels as the mixed-state analogue of finite-depth local unitary circuits and studies one-dimensional spin chains with strong $\fastto$6 and weak $\fastto$7 symmetry. The topological invariant is the twist expectation

$\fastto$8

and for weak-$\fastto$9-mSRE states one has

$t\gtrsim \mathrm{poly}\log N$0

This yields a mixed-state Lieb-Schultz-Mattis theorem that does not rely on Hamiltonians or spectral gaps, but the realized phases are explicitly interpreted as ASPT phases and remain closely tied to the same $t\gtrsim \mathrm{poly}\log N$1 dichotomy familiar from pure-state $t\gtrsim \mathrm{poly}\log N$2 SPT physics (Li et al., 25 Mar 2026).

A more intrinsically mixed SPT construction uses modulated symmetries such as dipole or subsystem $t\gtrsim \mathrm{poly}\log N$3. Here the key mechanism is that a modulated symmetry can be weak on the density matrix while a global charge remains strong. In a 2D topological dipole insulator and a 3D subsystem-symmetric higher-order phase, disorder averaging or infinite-depth channels produce short-range-correlated mixed states whose boundaries carry anomalous symmetry patterns forbidden in any equilibrium short-range-entangled pure state or thermal state of a gapped local Hamiltonian. The edge diagnostics are nonlinear: operators charged under the weak symmetry have algebraic ordinary correlators, while operators charged under the strong symmetry exhibit algebraic order only in a Rényi-2 correlator (You et al., 2024).

Related holographic constructions begin with higher-order subsystem SPT phases protected by non-invertible Kramers–Wannier symmetry and trace out the bulk to obtain one-dimensional mixed states. The resulting phases display coexistence of mixed-state SPT order and strong-to-weak spontaneous symmetry breaking, termed doubled average SPT (DASPT). These are mixed-state descendants of higher-dimensional pure states rather than intrinsically mixed phases in the strongest sense, but they show that non-invertible symmetry can refine mixed-state phase distinctions beyond ordinary invertible-symmetry SPT logic (Mana et al., 3 Mar 2026).

6. Tensor-network constructions, robustness, and open problems

A recent tensor-network program provides exact fixed-point tensor-network representations for a broad family of intrinsically mixed-state topological orders. The construction starts from a pure topological phase, passes to its Choi state, implements suitable anyon-condensation or projective-decoherence constraints there, and maps back to a density-matrix tensor network. In the $t\gtrsim \mathrm{poly}\log N$4 toric-code setting the mixed-state tensor network takes the form

$t\gtrsim \mathrm{poly}\log N$5

where the connector tensors $t\gtrsim \mathrm{poly}\log N$6 encode the local decoherence constraints. The method applies to $t\gtrsim \mathrm{poly}\log N$7 decoherence of $t\gtrsim \mathrm{poly}\log N$8 toric code, decohered non-Abelian $t\gtrsim \mathrm{poly}\log N$9 quantum double, pure $\mathbb Z_n$0 decoherence of arbitrary CSS codes, and a chiral semion example that the authors present as a fixed-point mixed-state realization unavailable in ordinary pure-state local commuting-projector form (Aldossari et al., 30 Jul 2025).

Across these frameworks, several misconceptions are repeatedly corrected. Intrinsically mixed-state topological order is not synonymous with “a topological pure state plus weak noise,” because weak correctable noise often leaves the state in the same phase as the parent pure topological order. It is also not synonymous with arbitrary two-way channel connectivity, because that notion can collapse the distinction between trivial states and mixed states with nontrivial logical or $\mathbb Z_n$1-form symmetry data. Conversely, not every mixed-state phase constructed with channels or doubled-space methods is intrinsically mixed: some are better regarded as mixed-state continuations of pure-state or average-SPT structures (Coser et al., 2018, Ma et al., 2024, Li et al., 25 Mar 2026).

The present status is therefore mixed but increasingly structured. The strongest evidence for genuinely intrinsic mixed-state topological phases comes from $\mathbb Z_n$2-dimensional constructions with strong-to-weak $\mathbb Z_n$3-form symmetry breaking, non-modular or premodular anyon theories, and axiomatic fixed points supporting classical and non-Abelian logical data directly at the level of density matrices. The main unresolved issues are a universally accepted equivalence relation, a complete classification beyond partial premodular or symmetry-based schemes, a fully intrinsic diagnosis of chirality not relative to a known parent state, and the extension of these ideas beyond fixed-point or strongly constrained models. Intrinsically mixed-state topological phases are thus no longer merely speculative, but neither are they yet governed by a single settled formalism.