Topological Order Preview

Intrinsic topological order is a robust pattern of long-range quantum entanglement in a gapped many-body phase that cannot be classified by spontaneous breaking of an ordinary global symmetry.

In the standard two-dimensional setting, its mutually reinforcing signatures include:

• a ground-state sector that local bulk operators cannot distinguish;
• dependence of that sector on the topology of the spatial manifold;
• noncontractible string or loop operators acting within the ground space;
• quasiparticles with anyonic fusion and braiding;
• universal long-distance entanglement data.

No item in this list should be used as a universal one-line definition without hypotheses. Degeneracy can come from symmetry breaking. Edge states can produce low-energy levels without bulk topological order. A system can lack a local order parameter yet still be trivial, symmetry protected, invertible, gapless, or simply poorly understood. A reliable identification combines bulk spectra, excitation properties, the algebra of extended operators, and entanglement.

The $\mathbb Z_2$ toric code makes these ideas concrete. The discussion concerns the usual gapped, non-invertible two-dimensional phases; fractional quantum Hall fluids provide a physical realization with additional charge and response structure.

Helpful background. Gapped phases explains locality and allowed deformations, while entanglement entropy supplies the reduced-state measure used below.

Gapped ground sectors on closed surfaces

Unless a qualification is stated, this page concerns:

• local or sufficiently short-range Hamiltonians;
• zero-temperature ground states;
• a nonzero bulk spectral gap in the thermodynamic limit;
• ordinary intrinsic topological order in two spatial dimensions;
• closed orientable surfaces when discussing topology-dependent degeneracy;
• finite correlation length away from boundaries and quasiparticles.

The default Hamiltonian family has the form

$$H_L = \sum_{X\subset\Lambda_L} h_X,$$

where $\Lambda_L$ has linear size $L$, the interaction range is uniformly bounded, and the sum of interaction strengths touching any one site is bounded independently of $L$. Its low-energy projector is denoted by

$$P_0(L) = \sum_{a=1}^{N_0} \lvert\psi_a(L)\rangle \langle\psi_a(L)\rvert.$$

At an exactly solvable fixed point, the states in this subspace may be exactly degenerate. In a generic Hamiltonian in the same phase, it is safer to speak of a quasi-degenerate ground-state band whose width vanishes rapidly with $L$ while a nonzero bulk gap separates it from excitations.

Why Landau Order Is Not Enough

The successful local paradigm

For a conventional ordered phase, one often finds a local observable $\Phi(\mathbf x)$ whose expectation value transforms nontrivially under a global symmetry. Distinct thermodynamic branches satisfy

$$\langle\Phi\rangle_\alpha \ne \langle\Phi\rangle_\beta.$$

Equivalently, a local measurement can identify which branch was selected. A ferromagnet illustrates the pattern:

$$\langle M_z\rangle_+ = -\langle M_z\rangle_- \ne 0.$$

Long-distance correlations preserve this local distinction,

$$\lim_{\lvert\mathbf r\rvert\to\infty} \langle \Phi(\mathbf r)\Phi(\mathbf 0) \rangle \ne 0,$$

after the appropriate thermodynamic and source-selection limits are taken.

Landau theory organizes such phases through an order-parameter field and its symmetry-allowed free-energy functional. This framework is powerful, but its input already presumes that local symmetry data capture the essential organization.

The obstruction

There are gapped phases for which every strictly local bulk measurement looks the same throughout the ground-state sector. For a local operator $O_X$ supported in a contractible region $X$,

$$\langle\psi_a\lvert O_X \rvert\psi_b\rangle \approx c_X\,\delta_{ab}.$$

The label $a$ is physically real, but it is stored nonlocally. No local order parameter reads it. Operators that distinguish or transform the sectors must wrap a noncontractible cycle, connect boundaries, or otherwise have support growing with system size.

The nonlocal operator algebra and excitation structure supply the positive characterization; absence of a local order parameter alone is insufficient.

Absence is not a diagnosis

A missing local order parameter does not imply topological order. It can also occur in:

• a trivial symmetric product phase;
• a symmetry-protected topological phase;
• an invertible phase with a quantized response;
• a stable gapless phase;
• a crossover or finite-size regime;
• a system for which the correct local observable has not been identified.

The correct lesson is narrower: local order parameters are not universal classifiers.

Complementary properties of a topological phase

For a conventional gapped intrinsic topological phase in two dimensions, the following diagnostics are often related manifestations of one infrared structure.

Diagnostic	What it probes	Essential qualification
Local indistinguishability	local reduced states of the ground sector	region must remain small compared with system size
Topology-dependent ground space	global sectors on a closed surface	examine a thermodynamic sequence, not one exact crossing
Loop-operator algebra	action of noncontractible strings	operators are defined modulo local deformations
Anyon fusion and braiding	superselection sectors of excitations	requires isolated, mobile, gapped quasiparticles
Long-range entanglement	obstruction to shallow local disentangling	specify allowed ancillas, symmetry, and locality
Topological entropy	universal subleading entanglement data	requires controlled region combinations and scale separation
Quantized response or boundary data	coupling to probes and anomaly structure	not every topological order has the same response

Different models expose different subsets most cleanly. A commuting-projector model gives transparent ground sectors and string operators. A fractional quantum Hall fluid gives transport, quasiparticle charge, interferometric phases, and topology-dependent degeneracy. A numerical tensor-network study may access entanglement, modular transformations, and minimally entangled states.

A strong claim identifies several compatible diagnostics and checks that ordinary alternatives have been excluded.

Local Indistinguishability

Projector formulation

Let $P_0$ project onto the low-energy ground-state band. A compact local-indistinguishability condition is

$$P_0 O_X P_0 = c_X(O_X)P_0 + \varepsilon_X(L),$$

where $X$ is contractible and

$$\lVert\varepsilon_X(L)\rVert \longrightarrow 0$$

as $L\to\infty$ with $X$ fixed. In favorable gapped models,

$$\lVert\varepsilon_X(L)\rVert \lesssim C_X e^{-L/\xi_{\mathrm{top}}},$$

although the precise bound depends on geometry, interaction range, and the definition of the low-energy band.

In a ground-state basis,

$$\langle\psi_a\lvert O_X \rvert\psi_b\rangle = c_X\delta_{ab} + O\!\left( e^{-L/\xi_{\mathrm{top}}} \right).$$

The diagonal part says that local expectation values do not reveal the sector. The off-diagonal part says that a local operator does not coherently transform one topological sector into another.

Reduced-state formulation

For a region $X$, define

$$\rho_X^{(a)} = \operatorname{Tr}_{\bar X} \lvert\psi_a\rangle \langle\psi_a\rvert.$$

Local indistinguishability implies

$$\lVert \rho_X^{(a)} - \rho_X^{(b)} \rVert_1 \longrightarrow 0$$

for every fixed contractible $X$. Indeed, the optimal bias for distinguishing two reduced states by a measurement in $X$ is controlled by their trace distance.

For equal prior probabilities, the optimal local discrimination success is $p_{\mathrm{succ}}=1/2+\|\rho_X^{(a)}-\rho_X^{(b)}\|_1/4$. Identical reduced states in one chosen basis establish only the diagonal part of the projector condition; the off-diagonal matrix elements must also vanish.

For example, the two GHZ states $(|0\cdots0\rangle\pm|1\cdots1\rangle)/\sqrt2$ have identical reduced states on every proper subsystem, but $\langle\mathrm{GHZ}_+|Z_i|\mathrm{GHZ}_-\rangle=1$. A local $Z_i$ acts nontrivially in their span, so this subspace fails the full local-indistinguishability condition. Its locally distinguishable basis is simply $|0\cdots0\rangle,|1\cdots1\rangle$.

Region size matters

The condition is not intended for arbitrary $X$. A strip that winds around a torus is not contractible. A region whose diameter grows to the system size can support a logical operator. A useful scale hierarchy is

$$\xi \ll \operatorname{diam}(X) \ll L$$

when one wants a region large compared with microscopic correlations but too small to wrap the system.

At a fixed-point stabilizer code, indistinguishability is exact for any operator whose support contains fewer sites than the code distance; geometric diameter and number of supported qubits should not be confused. Away from the fixed point, exponentially small corrections are generic.

Contrast with symmetry breaking

Suppose $\lvert+\rangle$ and $\lvert-\rangle$ are two broken-symmetry branches. A local order parameter obeys

$$\langle+\lvert\Phi_X\rvert+\rangle - \langle-\lvert\Phi_X\rvert-\rangle \ne 0$$

in the thermodynamic limit. Their reduced states are locally distinguishable.

For topological sectors,

$$\rho_X^{(a)} \approx \rho_X^{(b)}$$

for every contractible bulk $X$. The difference is structural:

• Symmetry-breaking sectors: a local order parameter distinguishes the branches, the symmetry can permute them, and global spatial topology is not essential to their existence.
• Topological sectors: local bulk probes do not distinguish them, no ordinary global symmetry is required, and spatial topology controls the sector structure.

Finite-volume symmetric cat states can temporarily hide the local distinction between broken-symmetry branches. The thermodynamic phase structure, source response, and matrix elements of local order parameters reveal it. That is why one exact finite-system eigenbasis is never enough for the comparison.

Ground-State Degeneracy and Topology

Topology-dependent sectors

Place the same local phase on different closed spatial manifolds. Intrinsic topological order can produce a ground-space dimension

$$N_0(\Sigma) = \dim\mathcal H_0(\Sigma)$$

that depends on the topology of $\Sigma$ rather than on its local geometry.

For the $\mathbb Z_2$ toric-code phase,

$$\dim\mathcal H_0(T^2) = 4$$

on a torus, while on an orientable surface of genus $g$,

$$\dim\mathcal H_0(\Sigma_g) = 4^g.$$

A sphere has $g=0$ and therefore a unique ground state in this idealized closed-system setting.

The number is not a local geometric response. Smoothly stretching the torus does not change it while the phase assumptions remain valid.

Why the torus is useful

A torus has two independent noncontractible cycles, conventionally denoted $\alpha$ and $\beta$. For $\mathbb Z_2$ order, choose the commuting electric loop operators along the two cycles; each has a binary eigenvalue. The four combinations span the ground space:

$$\lvert w_\alpha,w_\beta \rangle, \qquad w_\alpha,w_\beta\in\{+1,-1\}.$$

These labels cannot be read from a small disk. They are measured by loop operators winding around the system.

Exact degeneracy versus a ground-state band

At a commuting-projector fixed point,

$$E_1 = \cdots = E_{N_0}.$$

A generic weak local perturbation can give a small splitting

$$\delta E_{\mathrm{top}}(L) := E_{N_0}-E_1,$$

while the excitation gap remains

$$\Delta(L) := E_{N_0+1}-E_{N_0}.$$

The phase signature is the scale separation

$$\frac{ \delta E_{\mathrm{top}}(L) }{ \Delta(L) } \longrightarrow 0, \qquad \Delta(L)\longrightarrow\Delta_\infty>0.$$

For many local gapped realizations,

$$\delta E_{\mathrm{top}}(L) \sim e^{-L/\xi_{\mathrm{tun}}}.$$

The splitting arises because a virtual quasiparticle pair can be created, one member can wind around a noncontractible cycle, and the pair can annihilate. The process has order proportional to the loop length and is exponentially suppressed.

Degeneracy alone is insufficient

A low-energy multiplet can instead come from:

• spontaneous symmetry breaking;
• Kramers degeneracy;
• an exact microscopic symmetry representation;
• boundary or defect zero modes;
• disconnected components;
• accidental level crossings;
• gaplessness with finite-size level crowding.

The questions to ask are:

1. Does the multiplicity depend on spatial topology?
2. Are the states locally indistinguishable in the bulk?
3. Is the ground band separated by a stable bulk gap?
4. Do noncontractible operators act within the band?
5. Does the structure persist under generic weak local perturbations?

Noncontractible String and Loop Operators

Local deformation invariance

A string operator $W_a(C)$ transports a quasiparticle of type $a$ along a path $C$. If $C$ is deformed locally without crossing another excitation or changing its homotopy class, its action on the low-energy subspace is unchanged up to local details and phase conventions:

$$P_0 W_a(C)P_0 \simeq P_0 W_a(C')P_0.$$

What survives is the global winding or linking information.

For a contractible closed loop in an empty region, the action can reduce to a scalar within the ground space. For a noncontractible loop on a torus, it can act nontrivially:

$$P_0 W_a(\alpha)P_0 \not\propto P_0.$$

Intersecting loops

Two loop operators whose cycles intersect can obey a noncommutative algebra. In the $\mathbb Z_2$ toric-code phase,

$$W_e(\alpha) W_m(\beta) = - W_m(\beta) W_e(\alpha)$$

when $\alpha$ and $\beta$ cross once.

This algebra prevents a one-dimensional representation of all loop observables and forces a multidimensional ground space. It also encodes the mutual braiding phase of the $e$ and $m$ excitations.

Opposite edges of the upper rectangle are identified with matching orientation, so both displayed strings are closed noncontractible loops. An electric and magnetic loop with one intersection anticommute. A complete electric-anyon winding around a magnetic anyon gives the corresponding mutual phase $-1$. A contractible disk $X$ cannot distinguish the ideal toric-code ground sectors.

Logical operators

From the quantum-information viewpoint, noncontractible loops are logical operators. If the ground space encodes a logical qubit, one can identify

$$\bar Z \sim W_e(\alpha), \qquad \bar X \sim W_m(\beta),$$

with

$$\bar Z\bar X = - \bar X\bar Z.$$

A local operator has support too small to realize either logical action. The minimum support of a nontrivial logical operator defines a code-distance scale.

This analogy is structural, not merely rhetorical: local indistinguishability is closely related to the Knill–Laflamme condition for correcting local errors.

Anyons in Two Dimensions

Why two dimensions are special

In three spatial dimensions, exchanging two point particles twice can be continuously undone, and ordinary point particles fall into bosonic or fermionic exchange classes. In two dimensions, worldlines can braid around one another. The relevant topology is richer because paths with different winding cannot generally be deformed into one another without particle collisions.

For $n$ identical particles in the plane, exchanges are organized by the braid group $B_n$, generated by $\sigma_i$ with relations

$$\begin{aligned} \sigma_i\sigma_{i+1}\sigma_i &= \sigma_{i+1}\sigma_i\sigma_{i+1}, \\ \sigma_i\sigma_j &= \sigma_j\sigma_i, \qquad \lvert i-j\rvert\ge2. \end{aligned}$$

Unlike the permutation group, the braid group does not impose

$$\sigma_i^2 = 1.$$

That missing relation permits exchange statistics beyond bosons and fermions.

Superselection sectors

An anyon type $a$ labels a topological superselection sector. Local operators cannot change the total topological charge inside a region without also creating compensating charge or moving charge across the boundary.

The vacuum sector is denoted by $1$. Every type $a$ has an antiparticle $\bar a$ such that

$$a\times\bar a \supset 1.$$

The Superselection Sectors Preview develops the general operator-algebraic idea. Here the sectors are emergent quasiparticle types of a two-dimensional topological phase.

Fusion rules

Bringing quasiparticles together can produce more than one possible total charge:

$$a\times b = \sum_c N_{ab}^{\ \ c}\,c.$$

The nonnegative integers $N_{ab}^{\ \ c}$ count fusion channels. If every pair has at most one fusion outcome and all sectors have quantum dimension one, the anyon theory is Abelian. Multiple fusion channels can produce a protected multidimensional state space and non-Abelian statistics.

Associativity requires

$$(a\times b)\times c \cong a\times(b\times c),$$

but the two bases need not be identical. Their change of basis is encoded by $F$-moves in a full anyon theory. Anyons and Braiding develops the operational $F$- and $R$-move language, explicit non-Abelian examples, and braid protocols; this preview keeps only the phase-level diagnostic structure.

Quantum dimensions

The quantum dimensions $d_a$ are positive numbers satisfying

$$d_a d_b = \sum_c N_{ab}^{\ \ c} d_c.$$

The total quantum dimension is

$$\mathcal D = \sqrt{ \sum_a d_a^2 }.$$

For an Abelian theory, every $d_a=1$, so $\mathcal D$ is the square root of the number of sectors. A non-Abelian anyon has

$$d_a>1,$$

signaling asymptotic growth of the fusion-space dimension as more such anyons are added.

Abelian braiding

For Abelian anyons, exchanging or winding quasiparticles changes the state by a phase. A full counterclockwise winding of $a$ around $b$ can give

$$\lvert\Psi\rangle \longmapsto e^{i\theta_{ab}} \lvert\Psi\rangle.$$

For ordinary bosons or fermions, the exchange phase is restricted to $+1$ or $-1$. For anyons it can be a more general phase, subject to consistency with fusion and locality.

Non-Abelian braiding

If a collection of anyons has a degenerate fusion space $\mathcal V$, a braid acts by a unitary matrix:

$$\lvert\Psi\rangle \longmapsto U(\mathcal B) \lvert\Psi\rangle, \qquad U(\mathcal B)\in U(\mathcal V).$$

Two braid operations can fail to commute:

$$U(\mathcal B_1) U(\mathcal B_2) \ne U(\mathcal B_2) U(\mathcal B_1).$$

This is the origin of the term non-Abelian anyon. It refers to the braid representation, not to a non-Abelian microscopic symmetry group.

Exchange, winding, and topological spin

Three related notions should not be conflated:

• exchanging two identical anyons;
• winding one anyon completely around another;
• rotating one anyon by $2\pi$, encoded by its topological spin.

Their phases are related by the consistency structure of the anyon theory, but they are not interchangeable in every convention. When quoting a statistical angle, state which process it describes.

The Toric Code as a Fixed-Point Example

Degrees of freedom

Put one qubit on each edge $e$ of a square lattice embedded on a closed surface. Define star and plaquette operators

$$A_s = \prod_{e\ni s} X_e, \qquad B_p = \prod_{e\in\partial p} Z_e.$$

Every star and plaquette commute:

$$[A_s,B_p] = 0.$$

A star and a plaquette share either zero or two edges. On each shared edge, $X_eZ_e=-Z_eX_e$; two minus signs cancel.

The Hamiltonian is

$$\begin{aligned} H_{\mathrm{TC}} &= -J_e\sum_s A_s -J_m\sum_p B_p, \\ J_e,J_m &> 0. \end{aligned}$$

Its ground states satisfy

$$A_s\lvert\psi\rangle = \lvert\psi\rangle, \qquad B_p\lvert\psi\rangle = \lvert\psi\rangle.$$

Counting the torus ground space

For an $L_x\times L_y$ square lattice on a torus,

$$N_e = 2L_xL_y$$

qubits live on edges. There are

$$N_s = L_xL_y, \qquad N_p = L_xL_y$$

star and plaquette constraints.

Two global products are redundant:

$$\prod_s A_s = I, \qquad \prod_p B_p = I.$$

For the connected closed torus lattice, these are the only independent relations, so the number of independent stabilizers is

$$N_s+N_p-2.$$

The encoded-qubit count is

$$\begin{aligned} k &= N_e - \left( N_s+N_p-2 \right) \\ &= 2. \end{aligned}$$

Hence

$$\dim\mathcal H_0 = 2^k = 4.$$

This count uses periodic topology. A planar patch has boundary-dependent stabilizer relations and a different encoded dimension.

Why a local Pauli operator cannot read the sector

Consider a Pauli operator $P$ supported inside a contractible disk. If it anticommutes with a stabilizer $S$, then $P_0PP_0=P_0SPSP_0=-P_0PP_0$, hence its ground-space projection is zero. If it commutes with every stabilizer, its closed strings inside the disk are products of local star and plaquette operators, so it acts as a scalar in the ground space. Every operator on the disk is a linear combination of Pauli operators. This proves $P_0O_XP_0=c_XP_0$ for such regions at the toric-code fixed point.

Electric and magnetic excitations

A violated star constraint,

$$A_s = -1,$$

is an electric excitation $e$. A violated plaquette constraint,

$$B_p = -1,$$

is a magnetic excitation $m$.

Open strings create excitations at their endpoints. For a direct-lattice path $\Gamma$,

$$W_e(\Gamma) = \prod_{e\in\Gamma} Z_e$$

anticommutes with the endpoint star operators and creates an $e$ pair. For a dual-lattice path $\widetilde\Gamma$,

$$W_m(\widetilde\Gamma) = \prod_{e\perp\widetilde\Gamma} X_e$$

creates an $m$ pair on endpoint plaquettes.

The energy cost depends on the endpoints, not on the path length:

$$\Delta E_e = 4J_e, \qquad \Delta E_m = 4J_m$$

for a well-separated pair in the bulk. This is deconfinement at the fixed point.

Fusion

Applying the same $\mathbb Z_2$ string twice gives the identity. The fusion rules are

$$\begin{gathered} e\times e = 1, \qquad m\times m = 1, \\ e\times m = \epsilon, \qquad \epsilon\times\epsilon = 1. \end{gathered}$$

All four sectors

$$\{1,e,m,\epsilon\}$$

have quantum dimension one. Therefore

$$\mathcal D = \sqrt{ 1^2+1^2+1^2+1^2 } = 2.$$

The toric-code anyons are Abelian.

Mutual braiding

Suppose an $e$ string and an $m$ string cross once. At the crossing edge,

$$Z_eX_e = -X_eZ_e.$$

All other factors commute, so

$$W_e(\Gamma) W_m(\widetilde\Gamma) = - W_m(\widetilde\Gamma) W_e(\Gamma).$$

If the $e$ particle winds once around the $m$ particle, the many-body state acquires

$$e^{i\theta_{em}} = -1, \qquad \theta_{em} = \pi \pmod{2\pi}.$$

Thus $e$ and $m$ are individually bosonic in the fixed-point theory but have nontrivial mutual statistics. Their composite $\epsilon=e\times m$ is fermionic.

Ground-sector operators

Close the strings around noncontractible cycles. The loops create no excitations because they have no endpoints, so they commute with the Hamiltonian while acting within the ground space:

$$\begin{gathered} [W_e(\alpha),H_{\mathrm{TC}}] = 0, \\ [W_m(\beta),H_{\mathrm{TC}}] = 0. \end{gathered}$$

Loops of the same type along homologous paths differ by products of local stabilizers. Their action on the ground space depends only on the homology class.

Together, these results connect the four torus ground states to local indistinguishability, the loop algebra, and the anyon value $\mathcal D=2$.

What the fixed point hides

The exactly solvable Hamiltonian is unusually clean. A generic realization can have:

• dispersive quasiparticles;
• nonzero correlation length;
• exponentially split ground sectors at finite size;
• dressed, quasi-local string operators;
• interactions among excitations;
• nontrivial boundary dynamics.

These changes do not by themselves destroy the phase. The invariant statement concerns a gapped region of Hamiltonian space, not exact solvability.

The Surface Code develops code-family conventions, planar boundaries, decoding, thresholds, and lattice surgery. Those engineering questions are separate from the intrinsic-phase structure developed here.

Long-Range Entanglement

Circuit definition

At a fixed-point representative, a short-range-entangled phase can be related to a product state by a finite-depth local unitary circuit, with allowed local product ancillas:

$$\lvert\psi_{\mathrm{SRE}}\rangle = U_{\mathrm{FD}} \lvert\mathrm{product}\rangle.$$

The circuit depth remains bounded as $L\to\infty$, and every gate acts on a bounded-diameter region. For a generic state within the phase, this is an equivalence at the level of gapped paths or controlled quasi-local approximation, not usually an exact finite-circuit identity. A strict finite-depth circuit on a product state has exactly vanishing connected correlations beyond its light cone, whereas a generic trivial gapped ground state has small exponential tails.

An intrinsically topologically ordered state is long-range entangled in the sense that no such finite-depth local circuit removes its nonlocal structure:

$$\lvert\psi_{\mathrm{TO}}\rangle \ne U_{\mathrm{FD}} \lvert\mathrm{product}\rangle$$

under the declared equivalence rules.

The phrase does not mean that a two-point correlator remains large at arbitrary separation. It means that the global entanglement pattern cannot be disentangled by bounded-depth local operations.

Why finite depth matters

A circuit of depth $D$ with gate range $r$ spreads the support of a local operator by at most a distance of order

$$\ell_{\mathrm{circuit}} \sim Dr.$$

If $D$ and $r$ remain fixed as $L$ grows, the circuit cannot turn a contractible local operator into one that wraps a macroscopic cycle. This explains why a local equivalence of entire ground spaces preserves the distinction between local operators and logical loops. A proof for preparing one specified ground state requires the corresponding state-preparation or entanglement argument as well; support growth alone is not that full proof.

A circuit whose depth scales with $L$ can move information across the system and may disentangle the state. Such a circuit is not a phase-equivalence transformation of bounded depth.

Not the same as long-range order

In a gapped topological phase, connected correlations of local operators can decay exponentially:

$$\left| \langle O_X O_Y\rangle - \langle O_X\rangle \langle O_Y\rangle \right| \lesssim C e^{-d(X,Y)/\xi}.$$

At the same time, the state has long-range entanglement and nontrivial loop operators. Long-range entanglement is not long-range local correlation.

This distinction is central. Long-range order concerns the different, correlation-based notion.

Entanglement entropy

For a smooth simply connected region $A$ in a suitable two-dimensional gapped topological phase,

$$S(A) = \alpha \frac{ \lvert\partial A\rvert }{ a } - \gamma + \cdots.$$

The leading coefficient $\alpha$ is nonuniversal. The constant

$$\gamma = \ln\mathcal D$$

is the topological entanglement entropy under the standard assumptions and vacuum-sector convention.

For the toric-code phase,

$$\mathcal D = 2, \qquad \gamma = \ln2.$$

This formula is powerful but easy to misuse. Corners, finite correlation length, symmetry breaking, Goldstone modes, boundaries, and small-region fits can contaminate constant terms. Reliable extraction uses combinations of regions designed to cancel boundary contributions; see topological entanglement entropy for the cancellation geometries.

Entanglement is richer than one number

Topological entanglement entropy determines $\mathcal D$, not the entire anyon theory. Distinct topological orders can share the same total quantum dimension. A fuller diagnosis can require:

• minimally entangled states on a cylinder or torus;
• entanglement spectra;
• modular transformations;
• topological spins;
• fusion coefficients;
• braiding matrices;
• chiral central charge or thermal response.

Topological order in quantum matter develops modular data, Abelian $K$ matrices, and chiral response.

Quantum Error-Correction Viewpoint

Approximate Knill–Laflamme structure

For a code projector $P_{\mathcal C}$ and a set of correctable errors $E_i$, the Knill–Laflamme condition is

$$P_{\mathcal C} E_i^\dagger E_j P_{\mathcal C} = c_{ij}P_{\mathcal C}.$$

Compare this with local indistinguishability:

$$P_0 O_X P_0 \approx c_XP_0.$$

If products of sufficiently local errors remain supported in a correctable region, the topological ground space behaves as a quantum code against those local errors.

Local information and global information

The ground-sector label is stored globally. A local perturbation can:

• create nearby quasiparticle pairs;
• dress the ground states;
• shift their common energy;
• deform a logical string locally.

But it cannot distinguish or enact a logical transformation at leading order unless a process spans a noncontractible distance.

This explains both robustness and finite-size splitting. A logical error requires an extended process:

$$\begin{aligned} \text{create a pair} &\longrightarrow \text{separate the anyons}, \\ &\longrightarrow \text{wind around a cycle}, \\ &\longrightarrow \text{annihilate the pair}. \end{aligned}$$

Protection is not absolute

Topological encoding does not mean immunity to all noise. Protection depends on:

• a locality structure;
• a nonzero energy gap;
• system size;
• temperature and noise dynamics;
• active error correction when required;
• boundaries and defects;
• measurement and control errors.

The two-dimensional toric code at nonzero temperature is not a self-correcting quantum memory in the thermodynamic limit. Pointlike anyons can be thermally created and diffuse around noncontractible cycles at a finite energetic cost. Static ground-state robustness and long-time thermal memory are different claims.

Neighboring Notions

A comparison table

Structure	Local bulk order parameter	Needs protecting symmetry	Intrinsic anyons	Long-range entangled in the standard non-invertible sense
Symmetry-breaking phase	usually yes	symmetry defines the broken pattern	no	no
Trivial symmetric phase	no	no	no	no
Bosonic SPT phase	no	yes	no intrinsic bulk anyons	no after symmetry is forgotten
Free-fermion band topology	no	sometimes	no intrinsic fractional anyons	classification depends on symmetry and fermionic conventions
Intrinsic topological order	no local classifier	not required	yes in two dimensions	yes
Stable gapless phase	not necessarily	model dependent	possible fractionalization	circuit classification requires extra care

The rows describe the separate structures in isolation; symmetry breaking can coexist with intrinsic topological order. A pure symmetry-broken branch is used here, not its finite-volume cat superposition. The final column uses a common condensed-matter convention. Terminology around invertible topological order, chiral phases, and fermionic systems varies in the literature. State the convention rather than arguing from the word topological alone.

Symmetry-protected phases

An SPT phase can be nontrivial only while a specified symmetry is preserved. Once that symmetry constraint is removed, a bosonic SPT ground state can be connected to a trivial short-range-entangled state without closing the bulk gap.

Intrinsic topological order does not require an ordinary global protecting symmetry. Weak local perturbations that break microscopic symmetries need not destroy it, provided locality and the bulk gap survive.

This distinction does not make SPT phases less physical. It says that their equivalence relation includes a symmetry constraint. See Symmetry-Protected Structure Preview.

Band topology and invertible phases

An integer quantum Hall state or Chern insulator can have:

• a unique ground state on a closed manifold;
• no fractional anyons;
• a quantized Hall response;
• protected chiral edge structure.

Such phases are topological, but they are not the same as non-invertible intrinsic topological order exemplified by a fractional quantum Hall fluid or the toric code. Invertible phases possess an inverse under stacking and require their own classification language.

String order

A nonlocal string expectation value can diagnose some one-dimensional phases:

$$\left\langle O_i \exp\!\left( i\sum_{k=i+1}^{j-1}Q_k \right) O_j \right\rangle.$$

The presence of a string order parameter does not by itself imply two-dimensional intrinsic topological order. Some string orders diagnose hidden symmetry breaking or SPT structure.

Fracton and gapless extensions

Fracton phases can have restricted quasiparticle mobility and ground-state degeneracy that depends on system size or lattice geometry, not merely manifold topology. Gapless spin liquids can have fractionalized excitations without a fully gapped topological field theory.

These are important extensions, but importing the ordinary two-dimensional package without modification can be misleading.

Robustness Under Perturbations

Admissible perturbations

Consider

$$H(\lambda) = H_0 + \lambda V, \qquad V = \sum_X v_X.$$

Topological stability concerns perturbations that are:

• local or sufficiently rapidly decaying;
• small in local interaction strength relative to the gap, rather than small in the extensive total operator norm;
• applied without changing the thermodynamic setting;
• small enough that the relevant bulk gap remains open.

Under suitable topological-order conditions, weak local perturbations preserve a separated low-energy band whose width vanishes rapidly with size. Exact decay bounds depend on the stability theorem; a simple exponential is a common physical estimate, not the conclusion of every theorem.

A sufficiently strong perturbation can close the gap, condense an anyon, confine excitations, or drive a first-order transition.

Dressed operators

Away from an exactly solvable point, bare strings may no longer commute with the Hamiltonian. Quasi-adiabatic continuation dresses them into quasi-local operators:

$$\widetilde W_a(C;\lambda) = U(\lambda) W_a(C;0) U(\lambda)^\dagger.$$

The operators acquire rapidly decaying quasi-local tails, but the ground-sector algebra remains stable along the gapped path. Rigorous exact spectral-flow constructions commonly establish decay faster than any power or almost exponential decay; a strictly exponential bound should not be assumed without the appropriate theorem.

Exact microscopic formulas are therefore not phase invariants. Fusion, braiding, topology-dependent sector structure, and robust response data are the infrared content.

Ways a topological phase can end

A phase transition can occur through:

• closure of the bulk quasiparticle gap;
• condensation of a bosonic anyon;
• confinement of formerly deconfined excitations;
• proliferation of defects;
• a first-order level crossing between distinct phases;
• loss of the locality or dimensional assumptions.

At a continuous transition, the correlation length can diverge:

$$\xi \longrightarrow \infty.$$

The uniform locality and gap estimates underlying the phase equivalence then fail. Quantum phase transitions explains the broader transition framework.

Boundaries, Defects, and Geometry

Closed versus open systems

Topology-dependent ground-state degeneracy is cleanest on a closed surface. Boundaries introduce additional choices:

• which anyons can terminate or condense at a boundary;
• whether gapless edge modes are required;
• whether boundary conditions create logical sectors;
• how corners and defects bind localized modes.

A planar toric-code patch can encode information even though the underlying surface is topologically a disk. Different boundary types allow different strings to end, creating nontrivial logical paths between boundaries.

Edge degeneracy is not bulk degeneracy

Low-energy edge states can be distinguished by operators near the boundary. They may be protected by symmetry, chirality, or an anomaly. This differs from bulk topological ground states that are indistinguishable by every contractible bulk probe on a closed manifold.

When reporting degeneracy, specify:

$$\begin{gathered} \text{manifold and boundary conditions}, \\ \text{location of distinguishing operators}, \\ \text{bulk gap and edge gap}, \\ \text{finite-size scaling}. \end{gathered}$$

Punctures and defects

Punctures, twist defects, and boundaries can enlarge a protected state space. Their degeneracy may depend on defect type and separation. Such structures can emulate aspects of non-Abelian braiding even when the underlying bulk anyons are Abelian.

Defect statistics and intrinsic quasiparticle statistics must therefore be labeled separately.

Finite Temperature

Ground-state order versus thermal order

This page has used a pure ground-state notion. At nonzero temperature,

$$\rho_\beta = \frac{ e^{-\beta H} }{ Z(\beta) }$$

contains a thermal density of excitations. Questions about mixed-state topological order, thermal stability, and memory time require definitions beyond the ground-state projector.

Two-dimensional toric-code warning

Pointlike $e$ and $m$ excitations cost finite energy. At any nonzero temperature, sufficiently large systems contain thermally activated anyons. Their diffusion can implement a noncontractible logical process.

The equilibrium density is schematically

$$n_a(T) \propto e^{-\Delta_a/(k_{\mathrm B}T)},$$

where the single-anyon energies at the fixed point are $\Delta_e=2J_e$ and $\Delta_m=2J_m$, half the corresponding pair-creation energies. This is the bulk dilute-density estimate; the even-particle constraint affects very small finite systems. The density is small at low temperature but nonzero. Increasing system size supplies more places for errors and does not create a growing energy barrier for separating pointlike anyons.

Thus the statements

$$\begin{gathered} \text{stable topological ground phase}, \\ \text{self-correcting thermal memory} \end{gathered}$$

are not equivalent.

Higher-dimensional codes, constrained dynamics, long-range interactions, or active correction can change the memory problem. Those mechanisms lie beyond this preview.

Identifying a phase in calculations and measurements

A useful numerical comparison follows the ground-band width and bulk gap across sizes, then tests projected local matrix elements, loop operators, and entanglement with converged region sizes. A short correlation length or several low eigenvalues alone is insufficient. In particular, a finite-bond-dimension tensor-network state can have a finite transfer correlation length even when its target Hamiltonian is gapless.

Experiments can combine fractional charge, quantized response, interferometry, and fusion-sensitive measurements. These observables probe different aspects of the phase. Distinguish intrinsic quasiparticles from engineered defects or operations in a code subspace; a verified braid operation does not by itself establish an equilibrium topological phase.

Exercises

Exercise 1: Local distinguishability

Two finite systems each have two nearly degenerate low-energy states. In system A, there is a fixed local operator $M_X$ such that

$$\begin{aligned} \Delta M_X(L) &:= \langle\psi_1\lvert M_X\rvert\psi_1\rangle - \langle\psi_2\lvert M_X\rvert\psi_2\rangle , \\ \lim_{L\to\infty} \Delta M_X(L) &= 2m_0 \ne 0. \end{aligned}$$

In system B, every fixed contractible $X$ obeys

$$P_0O_XP_0 = c_XP_0 + O(e^{-L/\xi}).$$

Which system is compatible with ordinary symmetry breaking, and which is compatible with a topological ground sector?

Solution

System A is compatible with ordinary symmetry breaking. The two states remain distinguishable by the fixed local observable $M_X$, which can serve as an order parameter or a local proxy for one.

System B has the local-indistinguishability structure expected of a topological ground sector. Every local operator acts approximately as a scalar on the low-energy subspace.

Neither condition alone proves the full interpretation. For A one should identify the symmetry, thermodynamic branches, and source response. For B one should additionally establish a bulk gap, topology dependence, noncontractible operators, and stability over a parameter region.

Exercise 2: Count the toric-code ground states

On a square lattice with periodic boundary conditions, let

$$N_e = 2L_xL_y, \qquad N_s = N_p = L_xL_y.$$

Assume all star and plaquette stabilizers commute and that the two global products are the only independent stabilizer relations. Compute the ground-space dimension.

Solution

The number of independent stabilizers is

$$r = N_s+N_p-2 = 2L_xL_y-2.$$

For $N_e$ qubits, a rank-$r$ stabilizer group leaves

$$k = N_e-r$$

encoded qubits. Hence

$$\begin{aligned} k &= 2L_xL_y - \left( 2L_xL_y-2 \right) \\ &= 2. \end{aligned}$$

Therefore

$$\dim\mathcal H_0 = 2^k = 4.$$

The topology enters through the global relations and the two independent noncontractible cycles.

Exercise 3: Crossing strings

An $e$ string is a product of $Z$ operators and an $m$ string is a product of $X$ operators. Show that strings crossing once anticommute, while strings crossing twice commute.

Solution

At each crossing edge,

$$ZX = -XZ.$$

Operators on distinct edges commute. If the number of crossings is $I$, moving one full string past the other produces

$$W_eW_m = (-1)^I W_mW_e.$$

For one crossing,

$$W_eW_m = -W_mW_e.$$

For two crossings,

$$W_eW_m = W_mW_e.$$

Only the parity of the intersection number matters in the $\mathbb Z_2$ model. The odd-crossing minus sign is the mutual braiding phase between $e$ and $m$.

Exercise 4: Quantum dimensions

The toric-code fusion sectors are $\{1,e,m,\epsilon\}$, all Abelian. Compute $\mathcal D$ and $\gamma$.

Then consider a Fibonacci theory with sectors $\{1,\tau\}$ and fusion rule

$$\tau\times\tau = 1+\tau.$$

Find the positive quantum dimension $d_\tau$ and the total quantum dimension.

Solution

For the toric code, every sector has quantum dimension one:

$$\mathcal D_{\mathrm{TC}} = \sqrt{ 1+1+1+1 } = 2.$$

Therefore

$$\gamma_{\mathrm{TC}} = \ln2.$$

For Fibonacci fusion, the dimension equation is

$$d_\tau^2 = 1+d_\tau.$$

The positive solution is the golden ratio

$$d_\tau = \varphi = \frac{1+\sqrt5}{2}.$$

Thus

$$\mathcal D_{\mathrm{Fib}} = \sqrt{ 1+\varphi^2 }.$$

Because $d_\tau>1$, $\tau$ is non-Abelian: fusion spaces grow with the number of $\tau$ anyons.

Exercise 5: Exponentially small splitting

Suppose a virtual process that changes a torus sector requires a quasiparticle to traverse $L$ links. In perturbation theory, each step contributes a factor of order $\lambda/\Delta$, with $0<\lambda<\Delta$. Estimate the scaling of the sector splitting.

Solution

The first nontrivial process appears at an order proportional to $L$, so

$$\delta E_{\mathrm{top}}(L) \sim \Delta \left( \frac{\lambda}{\Delta} \right)^{cL}$$

for a geometry-dependent positive constant $c$. Writing

$$\left( \frac{\lambda}{\Delta} \right)^{cL} = \exp\!\left[ -cL \ln\!\left( \frac{\Delta}{\lambda} \right) \right]$$

gives

$$\delta E_{\mathrm{top}}(L) \sim \Delta e^{-L/\xi_{\mathrm{tun}}},$$

with

$$\xi_{\mathrm{tun}}^{-1} = c\ln\!\left( \frac{\Delta}{\lambda} \right).$$

This is a scaling estimate, not a universal prefactor. This estimates a representative process. Bounding the sum of all virtual paths also requires accounting for their number, which can grow exponentially with length; $\lambda/\Delta<1$ alone is not a general convergence proof. In a controlled weak-perturbation regime, those factors change the decay length without removing the suppression.

Exercise 6: Circuit-depth obstruction

A circuit has gate range $r$ and depth $D$. Explain why a depth bounded independently of $L$ cannot map a local operator to a loop winding around a torus of circumference $L$.

Solution

Conjugation by one layer enlarges an operator’s support only by a distance of order $r$. After $D$ layers, the support expands by at most order

$$Dr.$$

If $D$ and $r$ stay fixed while $L\to\infty$, then

$$Dr \ll L.$$

The dressed operator remains supported in a contractible neighborhood and cannot wrap a noncontractible cycle. A bounded-depth circuit therefore preserves the distinction between local and logical loop operators.

To build or erase a winding loop by local gates, the depth must grow with system size or the allowed gates must become nonlocal.

Exercise 7: Classify the protection mechanism

Classify each observation as most naturally suggesting symmetry breaking, SPT structure, invertible topology, intrinsic topological order, or an inconclusive result.

1. Two ground states have opposite local magnetization.
2. A one-dimensional gapped chain has edge spin-$1/2$ modes protected by spin rotation, but a unique bulk ground state on a ring.
3. A two-dimensional insulator has a unique torus ground state and integer Hall conductance.
4. A two-dimensional gapped system has four locally indistinguishable torus ground states and deconfined $e$ and $m$ excitations with mutual phase $\pi$.
5. Exact diagonalization on one cluster shows four equal lowest eigenvalues.

Solution

1. Symmetry breaking. A local observable distinguishes the branches.
2. SPT structure. The edge modes require the protecting symmetry, while the closed bulk is short-range entangled.
3. Invertible topology. The integer Hall response and unique closed-manifold ground state are characteristic of an integer quantum Hall or Chern phase, not non-invertible intrinsic order.
4. Intrinsic topological order. Local indistinguishability, topology-dependent sectors, and anyonic braiding form a coherent package.
5. Inconclusive. One finite-size degeneracy could be accidental, symmetry generated, edge generated, or a solvable-point artifact.

The classification assumes the usual locality and gap conditions. Additional symmetries or interactions can refine each entry.

Exercise 8: Test a numerical interpretation

A tensor-network calculation on cylinders of circumference $6$, $8$, and $10$ finds exponentially decaying spin correlations and fits

$$S(L_y) = \alpha L_y-\gamma$$

with $\gamma\approx\ln2$. The authors conclude that the phase is definitively the toric-code phase.

List what the data support and what remains to be established.

Solution

The data support:

• absence of obvious conventional spin long-range order in the measured channel;
• an area-law entanglement pattern over the fitted sizes;
• a constant compatible with the toric-code value.

They do not yet uniquely identify the phase. The fit may suffer from short circumferences, corners or cylinder conventions, finite correlation length, restricted variational sectors, or symmetry-breaking contributions.

A stronger identification should include:

• stability across bond dimension and circumference;
• multiple region or cylinder constructions;
• a bulk-gap estimate together with converged correlations; a finite transfer-matrix correlation length alone does not prove a Hamiltonian gap;
• minimally entangled sectors and their local indistinguishability;
• noncontractible loop or flux-insertion algebra;
• sector counting;
• modular or braiding data compatible with $\{1,e,m,\epsilon\}$;
• exclusion of other phases with the same $\mathcal D=2$.

The value $\gamma=\ln2$ determines the total quantum dimension under suitable assumptions; it does not determine the complete anyon theory.

Further reading

Symmetry-protected phases explains symmetry constraints on otherwise trivial phases. Surface codes develops planar boundaries, repeated measurements, and decoding. Entanglement spectra gives complementary diagnostics and their limits.

References

• D. Arovas, J. R. Schrieffer, and F. Wilczek, “Fractional Statistics and the Quantum Hall Effect”, Physical Review Letters 53, 722–723, 1984.
• S. Bachmann, S. Michalakis, B. Nachtergaele, and R. Sims, “Automorphic Equivalence within Gapped Phases of Quantum Lattice Systems”, Communications in Mathematical Physics 309, 835–871, 2012.
• S. Bravyi, M. B. Hastings, and S. Michalakis, “Topological Quantum Order: Stability under Local Perturbations”, Journal of Mathematical Physics 51, 093512, 2010.
• C. Castelnovo and C. Chamon, “Entanglement and Topological Entropy of the Toric Code at Finite Temperature”, Physical Review B 76, 184442, 2007.
• X. Chen, Z.-C. Gu, and X.-G. Wen, “Local Unitary Transformation, Long-Range Quantum Entanglement, Wave Function Renormalization, and Topological Order”, Physical Review B 82, 155138, 2010.
• E. Dennis, A. Kitaev, A. Landahl, and J. Preskill, “Topological Quantum Memory”, Journal of Mathematical Physics 43, 4452–4505, 2002.
• M. B. Hastings and X.-G. Wen, “Quasi-Adiabatic Continuation of Quantum States: The Stability of Topological Ground-State Degeneracy and Emergent Gauge Invariance”, Physical Review B 72, 045141, 2005.
• A. Y. Kitaev, “Fault-Tolerant Quantum Computation by Anyons”, Annals of Physics 303, 2–30, 2003.
• A. Kitaev and J. Preskill, “Topological Entanglement Entropy”, Physical Review Letters 96, 110404, 2006.
• A. Kitaev, “Anyons in an Exactly Solved Model and Beyond”, Annals of Physics 321, 2–111, 2006.
• R. B. Laughlin, “Anomalous Quantum Hall Effect: An Incompressible Quantum Fluid with Fractionally Charged Excitations”, Physical Review Letters 50, 1395–1398, 1983.
• J. M. Leinaas and J. Myrheim, “On the Theory of Identical Particles”, Il Nuovo Cimento B 37, 1–23, 1977.
• M. Levin and X.-G. Wen, “Detecting Topological Order in a Ground State Wave Function”, Physical Review Letters 96, 110405, 2006.
• M. A. Levin and X.-G. Wen, “String-Net Condensation: A Physical Mechanism for Topological Phases”, Physical Review B 71, 045110, 2005.
• C. Nayak, S. H. Simon, A. Stern, M. Freedman, and S. Das Sarma, “Non-Abelian Anyons and Topological Quantum Computation”, Reviews of Modern Physics 80, 1083–1159, 2008.
• D. C. Tsui, H. L. Störmer, and A. C. Gossard, “Two-Dimensional Magnetotransport in the Extreme Quantum Limit”, Physical Review Letters 48, 1559–1562, 1982.
• X.-G. Wen and Q. Niu, “Ground-State Degeneracy of the Fractional Quantum Hall States in the Presence of a Random Potential and on High-Genus Riemann Surfaces”, Physical Review B 41, 9377–9396, 1990.
• X.-G. Wen, Quantum Field Theory of Many-Body Systems, Oxford University Press, 2004.
• F. Wilczek, “Quantum Mechanics of Fractional-Spin Particles”, Physical Review Letters 49, 957–959, 1982.