---
title: 'Strange Correlator: Topological Phase Diagnostic'
url: https://www.emergentmind.com/topics/strange-correlator
type: topic
---

# Strange Correlator: Topological Phase Diagnostic

The strange correlator is a normalized mixed matrix element between a target many-body state and a reference state, most commonly written as
\[
C(r,r')=\frac{\langle \Omega|\phi(r)\phi(r')|\Psi\rangle}{\langle \Omega|\Psi\rangle}.
\]
Its defining feature is that it is not an expectation value within a single state. Instead, it compares a nontrivial state \(|\Psi\rangle\) with a trivial reference \(|\Omega\rangle\), and thereby converts bulk wave-function data into a probe of the effective interface between the two phases. In one and two spatial dimensions, the foundational result is that if \(|\Psi\rangle\) is a nontrivial short-range entangled state and \(|\Omega\rangle\) is a trivial disordered state on the same Hilbert space, then the strange correlator either saturates to a constant or decays as a power law at large separation, even though both states are themselves short-range correlated [1312.0626]. Subsequent work extended the concept from symmetry-protected topological phases to string-net topological phases, mixed-state average symmetry-protected topological phases, MERA bulk-boundary constructions, non-invertible symmetry-protected phases, and even overlap-based analogues in ASEP/DSSYK duality [1801.05959].

## 1. Conceptual origin and boundary interpretation

The original motivation for the strange correlator was the absence of a universally reliable bulk diagnostic for interacting symmetry-protected topological phases. Entanglement-spectrum diagnostics can fail in some cases, and one-dimensional string order parameters are symmetry-specific and do not naturally generalize to higher dimensions. The strange correlator addresses this by comparing the state of interest with a trivial symmetric product state rather than by evaluating an ordinary bulk expectation value [1408.1512].

Its physical interpretation is a Euclidean-interface construction. One imagines \(|\Psi\rangle\) prepared from imaginary time \(\tau=-\infty\) to \(0\), and \(\langle\Omega|\) prepared from \(\tau=+\infty\) back to \(0\). Then \(\langle\Omega|\Psi\rangle\) is a path integral with a domain wall at \(\tau=0\), and the strange correlator is the correlation function of operators inserted on that domain wall. Under spacetime rotation, the temporal boundary becomes a spatial interface between a trivial phase and a nontrivial short-range entangled phase. Because the nontrivial phase has anomalous or protected boundary physics, the interface correlator is forced to be long-ranged or algebraic rather than exponentially decaying [1312.0626].

This boundary logic underlies later formulations as well. In projective quantum Monte Carlo, the strange correlator is measured at the temporal ends of the operator string, which makes the “temporal boundary” interpretation explicit. In tensor-network constructions, the overlap between a topological PEPS and a product state becomes a lower-dimensional partition function whose transfer matrix inherits nonlocal symmetries from the topological bulk. In both settings, the strange correlator is not merely an unusual overlap: it is a computable boundary observable encoded in bulk data [1408.1512] [1801.05959].

## 2. Formal definitions and operator constructions

The standard pure-state definition is
\[
C(r,r')=\frac{\langle \Omega|\phi(r)\phi(r')|\Psi\rangle}{\langle \Omega|\Psi\rangle},
\]
with \(|\Psi\rangle\) the state to be diagnosed, \(|\Omega\rangle\) a trivial reference state, and \(\phi\) a local operator chosen to overlap with the relevant boundary degrees of freedom. The denominator is essential: in the thermodynamic limit both numerator and denominator may vanish, while the ratio remains finite, so the ratio must be evaluated at finite size before taking the limit [1312.0626].

Several concrete implementations instantiate this definition in different many-body formalisms. In projective stochastic series expansion QMC for spin-1 Heisenberg antiferromagnets, the strange correlator is
\[
C(r-r')=\frac{\langle\Omega|\phi(r)\phi(r')\left({\cal H}-{\cal C}\right)^{2m}|\psi\rangle}{\langle\Omega|\left({\cal H}-{\cal C}\right)^{2m}|\psi\rangle},
\]
with
\[
\phi(r)\phi(r')=S_{r}^{+}S_{r'}^{-}+S_{r}^{-}S_{r'}^{+},
\qquad
|\psi\rangle=|\Omega\rangle=\prod_i|0\rangle_i.
\]
Here \(({\cal H}-{\cal C})^{2m}|\psi\rangle\) projects onto the ground state for large \(m\), so the strange correlator is a mixed overlap between a trivial symmetric product state and the projected ground state [1408.1512].

For one-dimensional fermionic symmetry-protected topological phases, a single strange correlator is not generally sufficient. The classification data split into fermionic decoration \(n_1\) and bosonic cocycle data \(\nu_2\), so the diagnosis requires both a fermionic strange correlator and a bosonic strange correlator. The bosonic form remains
\[
C(i,j)=\frac{\langle\Omega|\hat O_i\hat O_j|\Psi\rangle}{\langle\Omega|\Psi\rangle},
\]
while the fermionic version is built from bond fermion annihilation operators,
\[
C(i,j)=\frac{\langle\Omega|c_{i,i+1}\, c_{j,j+1}|\Psi\rangle}{\langle\Omega|\Psi\rangle}.
\]
In fixed-point wave functions these reduce to transfer-matrix correlators of one-dimensional statistical models, and nontrivial phases give constant or oscillatory long-range behavior rather than exponential decay [2312.01310].

For mixed states, the fidelity strange correlator generalizes the construction to density matrices:
\[
C(r,r')=\frac{F\left(\rho,\phi(r)\phi(r')\rho_0\phi(r)\phi(r')\right)}{F(\rho,\rho_0)},
\]
where \(F(\rho,\sigma)=\mathrm{Tr}\sqrt{\sqrt{\rho}\sigma\sqrt{\rho}}\). This is basis independent and was introduced for average symmetry-protected topological phases, where the diagnostic must operate on a bulk density matrix rather than on a pure wave function [2210.17485].

A distinct but now standard extension uses “strange correlator” for pure overlaps that generate classical partition functions. For string-net states this takes the form
\[
Z_{\mathrm{SC}}=\langle \Omega \mid \Psi_{\mathrm{SN}}\rangle,
\]
with \(|\Psi_{\mathrm{SN}}\rangle\) a string-net ground state and \(|\Omega\rangle=|\omega\rangle^{\otimes N}\) a product state. Contracting each physical PEPS leg with \(\langle\omega|\) yields a local Boltzmann tensor and therefore a two-dimensional partition function [1801.05959].

## 3. Phase diagnostics and critical behavior

The most basic diagnostic pattern is threefold: long-ranged or quasi-long-ranged in nontrivial symmetry-protected topological phases, exponentially decaying in trivial gapped phases, and algebraic at topological phase transitions. This structure is established explicitly for spin-1 Heisenberg antiferromagnets with uniaxial single-ion anisotropy. In the Haldane phase of the spin-1 chain at \(D=0\), the strange correlator approaches a nonzero constant,
\[
C(\infty)=0.64.
\]
In the trivial large-\(D\) phase at \(D=2\), it is short-ranged,
\[
C(r)\sim e^{-r/\xi}+e^{-(L-r)/\xi},
\qquad
\xi=3.76.
\]
Near the Haldane–large-\(D\) transition, at \(D=1\) in the reported data, it becomes algebraic,
\[
C(r)\sim r^{-\eta}+(L-r)^{-\eta},
\qquad
\eta=1.00,
\]
and the actual transition is a Gaussian quantum critical point with \(D_c=0.96845(8)\) [1408.1512].

The same work verified the even/odd ladder effect. For the two-leg spin-1 ladder with \(J=K=1\), \(D=0\), the strange correlator decays to zero in the thermodynamic limit and identifies the phase as topologically trivial. For the three-leg ladder with the same couplings, the strange correlator tends to a nonzero constant for correlations within a leg, between neighboring legs, and between outer legs, identifying a non-trivial SPT phase [1408.1512].

The original wave-function analysis already exhibited the same asymptotic dichotomy in explicit models. For the \(1d\) Haldane phase represented by the AKLT state, with \(|\Omega\rangle=|0\,0\,0\cdots\rangle\) and operators \(S_r^+\), \(S_{r'}^-\), the strange correlator is exactly
\[
C(r,r')=\frac{\langle \Omega| S^+_r \ S^-_{r^\prime} |\Psi \rangle } { \langle \Omega |\Psi \rangle }  = 2,
\]
independent of separation. For the \(2d\) quantum spin Hall insulator with strong Rashba coupling, the momentum-space strange correlator is singular as
\[
|C_k| \sim \frac{1}{|k|},
\]
which Fourier transforms to
\[
|C(r,r^\prime)|\sim \frac{1}{|\vec r-\vec r^\prime|}.
\]
For the \(2d\) spin-2 AKLT state on the square lattice, the envelope follows a power law
\[
|C(r,r')|\sim |r-r'|^{-\eta},
\qquad
\eta \simeq 0.32.
\]
For the Levin–Gu \(Z_2\) SPT, the strange correlator is long-ranged in one regime and critical near \(\beta_c \simeq 0.521\), with
\[
C(r,r')\sim |r-r'|^{0.199}.
\]
These examples established early that the strange correlator can be constant or algebraic even when both states individually have short-range bulk correlations [1312.0626].

In interacting fermionic systems, the strange correlator was used to diagnose the quantum spin Hall to antiferromagnetic Mott insulator transition in the Kane-Mele-Hubbard model. The single-particle strange correlator
\[
C^{\sigma}_{\mathbf{k}AB}= \frac{\langle\Omega|c^{\dagger}_{\mathbf{k}A\sigma} c_{\mathbf{k}B\sigma}|\Psi\rangle}{\langle\Omega|\Psi\rangle}
\]
is singular near one \(M\) point in the quantum spin Hall phase, while the singularity disappears in the antiferromagnetic Mott insulator. In the interacting topological phase, the strange correlators in the single-particle, spin, and pairing sectors reproduce the helical Luttinger-liquid scaling forms
\[
C^{\sigma}_{\mathbf{r}AB} \sim \mathbf{r}^{-g/2-1/2g},
\qquad
S^{\pm}_{\mathbf{r}AA} \sim \mathbf{r}^{-2g},
\qquad
D_{\mathbf{r}AA} \sim \mathbf{r}^{-2/g},
\]
so the strange correlator captures not only the presence of a topological interface but also the interaction-renormalized boundary exponents [1506.00549].

The same diagnostic logic has been extended to subsystem symmetry-protected topological phases. In the two-dimensional cluster model with subsystem \(\mathbb Z_2^{sub}\) symmetries, the dimer strange correlator
\[
C_D(\Delta r)=\frac{\langle\Omega|D_{i+\Delta r} D_{i}(-H)^{2n}|\Psi(0)\rangle}{\langle\Omega|(-H)^{2n}|\Psi(0)\rangle}
\]
is strongly anisotropic at the exactly solvable point:
\[
C_D(\Delta r)=1 \ \text{if}\ \Delta r_y=0,
\qquad
C_D(\Delta r)=0 \ \text{if}\ \Delta r_y\neq 0.
\]
The stabilizer-channel strange correlator \(C_B\) is isotropic and long-ranged in the subsystem SPT phase, while both strange order parameters vanish in the trivial paramagnet above the first-order transition at \(h_{x,c}=1\) [2209.12917].

For AKLT states on chain, square, honeycomb, cube, diamond, and hyperhoneycomb lattices, direct evaluation in the valence-bond loop gas framework confirmed the expected even/odd pattern. In one dimension, even \(S\) gives short-ranged strange correlations and odd \(S\) gives long-range order. On the square lattice, odd multiplicity \(M=2S/z\) gives algebraic strange correlations, while even multiplicity is short-ranged. In three dimensions, the diamond \(S=2\), \(M=1\) AKLT state was identified as a nontrivial SPT paramagnet by a nonzero strange winding fraction, whereas even-multiplicity disordered states were trivial [1603.03444].

## 4. Tensor-network, topological, and conformal generalizations

A major expansion of the subject came from tensor-network formulations of topological order. For string-net PEPS, the overlap
\[
Z_{\mathrm{SC}}=\langle \Omega \mid \Psi_{\mathrm{SN}}\rangle
\]
is a two-dimensional tensor-network partition function whose transfer matrix inherits all matrix product operator symmetries of the underlying string-net state. Because the PEPS is non-injective and carries virtual MPO symmetries satisfying the pulling-through property,
\[
\mathrm{MPO}_a\,\mathrm{MPO}_b=\sum_c N_{ab}^{\,c}\mathrm{MPO}_c,
\]
the transfer matrix cannot generically flow to a unique symmetric injective fixed point. The resulting partition functions are therefore either critical or symmetry broken. At criticality, the inherited MPO symmetries are interpreted as lattice topological conformal defects, and central idempotents of Ocneanu’s tube algebra organize the transfer-matrix spectrum into topological sectors that become conformal sectors or conformal boundary conditions in the emergent conformal field theory [1801.05959].

Two explicit examples establish this correspondence concretely. For the Fibonacci string-net, projecting every physical degree of freedom onto the \(\tau\)-label produces a lattice model identified with the critical hard-hexagon model, with critical fugacity
\[
z_c=\phi^5.
\]
For the symmetry-enriched Ising string-net, the product-state bra
\[
\langle \omega(\beta)| = \sqrt{2}\bigl(\cosh(\beta)\langle 1|+\sinh(\beta)\langle \psi|\bigr)
\]
yields the self-dual critical Ising point
\[
\beta_c = \frac{1}{2}\log(1+\sqrt{2}),
\]
and the transfer matrix carries the full Ising MPO symmetry algebra, including the \(\sigma\)-twist corresponding to Kramers–Wannier duality [1801.05959].

The same topological-holographic viewpoint underlies later “CFT factory” constructions. In a square-octagon discretization of a \((2+1)d\) topological state, the strange correlator
\[
\bra{\Omega(\{r_i\})}\Psi\rangle
\]
generates a family of \((1+1)d\) transfer matrices on a four-site circle, where the boundary state \(\bra{\Omega(\{r_i\})}\) is built from competing condensates associated with Frobenius algebra objects \(\mathcal A_i\). Scanning the entropy function
\[
S_{\Delta}(\ket{\psi}) := (S_{AB} + S_{BC}) - \eta(S_A + S_C) - (1-\eta)(S_B+S_{ABC})
\]
over the boundary couplings \(\{r_i\}\) gives a cost-effective way of identifying critical boundary conditions, estimating central charges, and plotting phase diagrams in multi-parameter spaces [2509.04947].

A related tensor-network development views MERA itself as a holographic strange correlator. If \(|\Psi\rangle\) is a one-dimensional critical state represented by MERA, lifting the network by inserting a basis-independent intertwiner tensor on every bond produces a two-dimensional bulk state \(|\Psi^{(\mathrm{lift})}\rangle\) with exponentially decaying bulk correlations. The original critical state is then recovered by projection with a bulk product state:
\[
|\Psi\rangle = \left(\bigotimes_{\text{all bulk sites} \eta\,\chi^{-1/2}\,\langle +|\right) |\Psi^{(\mathrm{lift})}\rangle,
\qquad
|+\rangle = \sum_{j=1}^{\chi}|j\rangle|j\rangle.
\]
At the holographic point
\[
\eta_{\mathrm{Holo}}=\frac{\sqrt{2\chi}}{\sqrt{\chi+1}},
\]
the lifted state exhibits holographic screens and an RT-like relation between boundary entanglement and bulk-screen entropy [1812.11644].

Strange correlators have also become a mechanism for constructing non-invertible defects. In generalized Ising models formulated from chain complexes, the overlap between a topologically ordered CSS-code-like state and a product state yields the partition function of a generalized Ising model. Partial gauging of a subregion then produces an interface partition function with a Kramers–Wannier duality defect. In the ordinary two-dimensional Ising case this reproduces the non-invertible fusion rule
\[
\mathcal{D}\times\mathcal{D}=\mathrm{I}+D_{\eta},
\]
with quantum dimension
\[
d_{\mathcal D}=\sqrt{2}.
\]
Here the strange correlator serves as the map from topological bulk data to classical partition functions with defects [2511.13833].

## 5. Numerical methodologies and practical diagnostics

The strange correlator has been implemented in several numerical frameworks, with operator choice emerging as a decisive issue. A general prescription based on bulk-boundary correspondence is to choose operators that create the edge modes of the topological phase and then continue them into the bulk. In free-fermion systems this means operators with overlap onto both filled and empty bands, so that the strange correlator probes bulk particle-hole excitations corresponding to edge modes. If
\[
|s[\hat{o},\hat{o}']_{{\bf r},{\bf r}'}|\sim |{\bf r}-{\bf r}'|^{-2\alpha},
\]
then the sum of moduli scales as
\[
\bar{S}[ \hat{o} , \hat{o}^{\prime}]_L \sim \left\{ \begin{array}{cc}
L^{d-2  \alpha}    &  \mathrm{for} \, \alpha<\frac{d}{2} \\
\ln L    & \mathrm{for}\, \alpha=\frac{d}{2}  \\
\mathrm{const} + O(L^{d-2\alpha})\, & \mathrm{for}\, \alpha>\frac{d}{2} \, .
\end{array} \right.
\]
Integrating moduli substantially reduces cancellations and finite-size effects, and the resulting \(\bar S\) provides a robust finite-size diagnostic [2209.04283].

Projective quantum Monte Carlo is particularly natural because ordinary observables are measured at the middle of the open operator string, whereas strange correlators are measured at the ends. In the spin-1 Heisenberg study, ground-state convergence was obtained with projection length \(m\propto L^2\), using practical choices such as \(2m=L^2\) for chains and two-leg ladders and \(2m=3L^2\) for three-leg ladders. A structure-factor-based quantity,
\[
\Psi_L=\frac{S(0)}{N},
\]
serves as an order-parameter-like measure of the long-distance strange correlator, with exponential finite-size form
\[
\Psi_L=\Psi_{\infty}+\left(a-be^{-L/\xi}\right)/L
\]
and critical form
\[
\Psi_L=a/L+b/L^{\eta}.
\]
This makes the strange correlator useful not only as a phase classifier but also as a finite-size scaling observable at topological transitions [1408.1512].

In fermionic projector determinantal QMC, the method is technically easier to implement than entanglement-spectrum diagnostics because only static correlations are needed. In the Kane-Mele-Hubbard model, the same projector-QMC machinery used for ordinary equal-time observables directly measures strange correlators in single-particle, spin, and pairing channels, without analytic continuation and without open boundaries [1506.00549].

For tensor-network strange correlators arising from topological PEPS, the transfer matrix is often treated as an iMPO, with its dominant eigenvector approximated by an iMPS using VUMPS. In the \(SU(2)_4\) strange correlator, the product state
\[
x\ket{2}+y\ket{4}+z\ket{0},
\qquad
x^2+y^2+z^2=1,
\]
defines a spherical two-dimensional parameter space. The phase diagram was mapped by the half-infinite-chain entanglement entropy of the transfer-matrix dominant eigenvector, using finite-entanglement scaling
\[
S \sim \frac{c}{6}\log \xi + \mathrm{const}.
\]
The resulting diagram contains six gapped phases, Ising-type critical points with \(c\approx \frac12\), an extended \(c=1\) critical phase, and critical lines with apparent \(c=2\) [2502.14556].

A complementary small-system workflow combines strange-correlator transfer matrices with the entropy-function criterion of Lin–McGreevy. On a four-site circle with \(\eta=\frac12\), one expects at criticality
\[
S_\Delta \approx \frac c3,
\]
so peaks in \(S_\Delta\) over boundary-coupling space provide a fast first-pass scan for critical points, critical lines, and multicritical structures. In the reported examples, every data point in the \(A\)-series plots and in the \(A_5\) two-dimensional phase diagram could be generated “within one second on a normal laptop” or “within one second on a basic CPU laptop” [2509.04947].

## 6. Limitations, ambiguities, and modern extensions

A central limitation is that the strange correlator is not a property of the target state alone. It probes the relation between the target and the reference state. Recent MPS analysis showed that ill-chosen reference states can induce spurious long-range strange correlators in trivial SPT phases. The true transfer-matrix criterion is magnitude-degeneracy of the leading eigenvalues of the mixed transfer matrix \(M_\omega\), and three distinct mechanisms can generate such degeneracy without nontrivial SPT order: high-dimensional irreducible representations in the entanglement space, phase mismatch in symmetry representations between target and reference states, and long-range order arising from symmetry breaking [2512.06691].

This result sharpens an older caveat: long-range strange correlators are a reliable positive signal only when the reference state is chosen properly. A valid reference state must be trivial or short-range entangled, preserve the protecting symmetry, match the target state’s one-dimensional symmetry representation labels, and avoid activating accidental high-dimensional irreducible representations in the entanglement space. The connected strange correlator is correspondingly more reliable than the ordinary one [2512.06691].

There are also dimensional limitations. In one and two dimensions, a nontrivial SRE/SPT state compared to a trivial disordered reference gives a strange correlator that either saturates to a constant or decays as a power law. In three dimensions, however, short-ranged strange correlations no longer guarantee triviality, because a two-dimensional boundary may be a gapped topological phase rather than gapless or symmetry-breaking. For that reason, the original wave-function work proposed additional overlap-based diagnostics for \(3d\) short-range entangled states [1312.0626].

Modern extensions of the strange-correlator idea have broadened both its scope and its meaning. For non-invertible symmetry-protected topological phases in \(1+1d\), strange correlators are defined between two arbitrary NISPT states rather than between a target state and a trivial product state, because a trivial product state is not always a valid symmetric reference. The inserted “strange charged operators” are constructed from the interface algebra, and the strange correlator has long-range order when evaluated between two distinct NISPTs and decays exponentially otherwise [2505.00673].

For open or disordered quantum systems, the fidelity strange correlator extends the construction to density matrices and average symmetry-protected topological phases. In one and two dimensions, nontrivial ASPTs exhibit long-range or power-law FSCs, and in several two-dimensional examples the FSC maps to nonlocal observables of \(O(n)\) loop models with “quantum corrections” from decorated defects [2210.17485].

The term has also acquired a broader overlap-based usage outside conventional topological diagnostics. In the ASEP/DSSYK duality, the normalization of the ASEP stationary state is written as
\[
Z=\langle \Omega|\Psi\rangle,
\]
where \(\langle\Omega|\) is a product bra and \(|\Psi\rangle\) the stationary ASEP MPS. This overlap is presented not as the conventional operator-valued strange correlator of topological phases, but as an analogue with the same “nontrivial bulk state + trivial product state \(\to\) boundary partition function” architecture [2606.18739].

Taken together, these developments define the strange correlator less as a single formula than as a family of overlap-based diagnostics governed by one persistent principle: compare a nontrivial state to a trivial or otherwise controlled reference, and read off the topology from the long-distance behavior of the induced interface observable. In its most reliable uses, long-range or algebraic strange correlations reflect protected boundary physics encoded in bulk wave functions or bulk density matrices; in its least careful uses, the same behavior can be a false positive generated by reference-state choice or by symmetry-breaking structure.

Source: https://www.emergentmind.com/topics/strange-correlator