---
title: 'X-Cube Model: Fracton Topological Order'
url: https://www.emergentmind.com/topics/x-cube-model
type: topic
---

# X-Cube Model: Fracton Topological Order

The X-cube model is a canonical exactly solvable lattice model of type-I fracton topological order. In its standard form, it is defined on a three-dimensional cubic lattice with qubits on links and commuting stabilizer terms attached to cubes and planar vertex “crosses.” Its defining features are subextensive ground-state degeneracy, immobile fractons, lineons constrained to one-dimensional motion, and planons mobile only within planes. Unlike conventional topological phases, its long-distance structure depends not only on topology but also on geometry, lattice foliation, and boundary conditions [2203.13274] [1708.04619].

## 1. Lattice definition and stabilizer structure

A common lattice presentation places qubits on the edges of a cubic lattice and uses the Hamiltonian
$$
H=-\sum_{v\in V}\left(A_v^x+A_v^y+A_v^z\right)-\sum_{c\in C} B_c.
$$
Here \(A_v^i\) applies Pauli \(Z\) on the four edges perpendicular to the \(i\)-axis and incident to \(v\), while \(B_c\) applies Pauli \(X\) on the twelve edges of cube \(c\). These operators satisfy
$$
(A_v^i)^2=B_c^2=1,\qquad [A_v^i,B_c]=0,
$$
so the model is an exactly solvable commuting-projector stabilizer code. The ground state may be written as
$$
|GS\rangle=\prod_c \frac{1+B_c}{2}\,|\phi_0\rangle, \qquad |\phi_0\rangle=|00\cdots 0\rangle,
$$
with the vertex projectors omitted because \(A_v^i\) acts trivially on \(|\phi_0\rangle\) in this convention [2210.01682].

A \( \mathbb{Z}_N \) formulation uses generalized clock operators \(X\) and \(Z\) obeying
$$
X_i Z_j=\omega^{\delta_{ij}} Z_j X_i,\qquad \omega=e^{2\pi i/N},
$$
and writes the lattice Hamiltonian as
$$
H_{\text{X-cube}}=-\sum_x(\hat B_x+\hat B_x^\dagger)-\sum_{x,a}(\hat A_x^{(a)}+\hat A_x^{(a)\dagger}).
$$
In that notation, \(\hat B_x\) is a cube operator and \(\hat A_x^{(a)}\) is a vertex operator orthogonal to direction \(a\). The local relation
$$
A_i^x A_i^y A_i^z = 1
$$
encodes the redundancy among the three vertex terms at a site and is central to the excitation and constraint structure of the model [1708.04619] [2107.09073].

The model is often described as the representative exactly solvable realization of type-I fracton order because the commuting stabilizers encode restricted-mobility excitations without invoking conventional symmetry breaking. This suggests that the stabilizer algebra, rather than a continuum order parameter, is the primary organizing principle of the phase [2203.13274].

## 2. Excitations and mobility constraints

The elementary excitations fall into three standard classes. Fractons are cube-term violations and are completely immobile as isolated excitations. Lineons arise from violations of vertex-cross terms and come in three species, each mobile only along one Cartesian axis. Planons are bound states, typically of two fractons separated along a fixed axis, and can move within the plane perpendicular to that axis [2203.13274].

The mobility constraints are tied to the geometry of the creation operators. A membrane operator
$$
W(M)=\prod_{i\in M}\sigma_i^z
$$
creates four fractons at the corners of a square membrane, while a straight string operator
$$
W(S)=\prod_{i\in S}\sigma_i^x
$$
creates lineons at its endpoints. If the string bends, extra lineons appear at the turning points. The absence of spatially deformable string-like operators is therefore the direct origin of subdimensional motion: moving a fracton or lineon in a forbidden direction necessarily creates additional excitations and costs finite energy [2203.13274].

In the \( \mathbb{Z}_N \) description, lineons may be denoted \(e^x,e^y,e^z\), with motion constrained respectively to the \(x\), \(y\), and \(z\) directions. Their fusion rule is
$$
e_i^x e_i^y e_i^z = 1.
$$
Lineon dipoles \(d_e^a\) and fracton dipoles \(d_m^a\) are planons: they move freely in planes but not in the direction of their dipole moment. The continuum field-theory analysis encodes the same kinematics through higher-derivative charge constraints, with fracton density \(i^0\) and lineon density \(j^{0;a}\) obeying generalized continuity equations rather than ordinary particle-number conservation [2107.09073] [1708.04619].

These mobility constraints are robust to arbitrary local perturbations in the sense emphasized by the continuum description. A plausible implication is that the defining content of the X-cube phase is not merely the existence of unusual excitations, but the higher-rank kinematic structure that forbids generic transport processes [1708.04619].

## 3. Ground-state degeneracy, foliation, and geometric order

On an \(L_1\times L_2\times L_3\) periodic lattice, the \( \mathbb{Z}_N \) X-cube model has ground-state degeneracy
$$
\mathrm{degen}=N^{2L_1+2L_2+2L_3-3}.
$$
For \(N=2\), this becomes
$$
\mathrm{GSD}=2^{2(L_1+L_2+L_3)-3}.
$$
The \(-3\) reflects three global redundancies among the nonlocal operators. This subextensive scaling is one of the model’s defining features and distinguishes it sharply from conventional three-dimensional topological orders such as the 3D toric code, whose degeneracy is finite and topology-only [1708.04619] [2106.05749].

A central conceptual result is that this degeneracy is not purely topological. The continuum analysis shows that even on a manifold with trivial topology, spatial curvature can induce a robust degeneracy. In a curved cubic lattice with angular defects and “straight loops,” the simplest \(N=2\) construction yields a \(2^2\) degeneracy, and with \(\ell\) layers this becomes \(2^{2\ell}\). The splitting under local perturbations is exponentially small in the size of the curved region because connecting different sectors requires a large nonlocal loop operator. Geometry, not topology alone, therefore protects part of the degeneracy [1708.04619].

The generic-lattice construction makes this dependence explicit. Starting from intersecting two-dimensional “i-surfaces” in three dimensions, one defines vertices where three i-surfaces intersect, links where two intersect, and 3-cells as enclosed volumes. The resulting generalized X-cube models inherit their mobility constraints from the geometry of these surfaces: dimension-1 particles move along intersection lines, and dimension-2 particles move on the i-surfaces themselves. The degeneracy scales as
$$
\log_2 \text{GSD} \sim \sum_s 2 g_s,
$$
where \(g_s\) is the genus of an orientable i-surface \(s\). On this basis, the authors propose that fracton orders should be regarded as a geometric order [1712.04511].

The foliation perspective refines this picture. The X-cube model contains hidden layers of 2+1D topological states and can be viewed through a coupled-layer construction of stacked toric codes. On the 3-torus,
$$
\log_2 \mathrm{GSD} = 2L_x + 2L_y + 2L_z - 3,
$$
with the \(-3\) arising because Wilson loops of the minimal intersecting toric-code layers are tied together along their intersections. Screw dislocations provide a direct probe of this foliation structure: they change \(\log_2 \mathrm{GSD}\) by a finite amount and also induce tunneling of subdimensional excitations along the defect line, revealing structure that is not reducible to a simple stack of 2D layers [2012.07263].

## 4. Continuum field theory and boundary structure

The X-cube model admits a continuum field-theory description, but this theory is not and cannot be a topological quantum field theory. The reason is that X-cube order is not invariant under arbitrary smooth spacetime deformations; instead, the continuum action is invariant only under the restricted “subconformal” transformations
$$
t\to \tilde t(t),\qquad x\to \tilde x(x),\qquad y\to \tilde y(y),\qquad z\to \tilde z(z).
$$
This restricted invariance reflects the anisotropic locality of the lattice model and the fact that bending coordinate axes changes the allowed mobility of excitations [1708.04619].

In one continuum formulation, the Lagrangian takes the form
$$
L_{\text{X-cube}} =\frac{N}{2\pi}\,|\epsilon^{0abc}|\,\frac12\,B_{ab}\,\partial_0 A_c + B_0\,I^0 + A_{0;a}\,J^{0;a} - A_{0;a}J^{0;a}-A_aJ^a-B_0 I^0-\frac12 B_{ab}I^{ab},
$$
with the constraint
$$
\sum_a A_{0;a}=0.
$$
The gauge transformations are
$$
B_{ab}\to B_{ab}+\partial_a\partial_b\chi,\qquad A_a\to A_a-\epsilon^{0abc}\partial_b\zeta_c,\qquad \sum_a \zeta_a=0,
$$
together with temporal-component shifts. These relations reproduce the lattice commutation algebra, braiding-like statistical phases, and the system-size-dependent degeneracy [1708.04619].

Boundary physics is comparably rich. For the \( \mathbb{Z}_N \) X-cube model on a \(z=\) constant surface, the low-energy boundary theory reduces to a generalized two-component \(K\)-matrix theory,
$$
\mathcal{L}_0 =\frac{i}{4\pi}K_{IJ}\,\partial_0 \Phi_I\,\partial_x\partial_y \Phi_J, \qquad K=-iN\sigma^y,
$$
with a momentum subsystem symmetry
$$
\Phi_I(t,x,y)\rightarrow \Phi_I(t,x,y)+f_I(x,y).
$$
The double spatial derivative in the kinetic term is the boundary signature of subsystem structure. The theory reproduces a subset of the bulk exchange statistics, particularly those involving mobile composites such as fracton dipoles and lineons [2206.14829].

The boundary admits multiple gapped terminations. On \(T^2\times I\), smooth boundaries on both ends give
$$
\log_N \mathrm{GSD}^{(mm)\times(mm)}=l_x+l_y+2l_z-2,
$$
rough boundaries on both ends give
$$
\log_N \mathrm{GSD}^{(ee)\times(ee)}=l_x+l_y+2l_z-1,
$$
and one smooth plus one rough boundary gives
$$
\log_N \mathrm{GSD}^{(mm)\times(ee)}=2l_z.
$$
Anisotropic and dyonic boundaries also exist, and the anomaly inflow analysis shows that the boundary ’t Hooft anomaly is not canceled uniquely by the X-cube bulk. This suggests that the boundary theory probes only part of the three-dimensional fractonic data [2206.14829].

## 5. Dynamics, finite-temperature behavior, and phase transitions

Perturbing the model by Zeeman fields exposes distinctive dynamical signatures of subdimensional motion. For the Hamiltonian
$$
H=-K\sum_i A_{c,i}-\Gamma\sum_{i,v}B_{v,i}-h_x\sum_i \sigma_i^x-h_z\sum_i \sigma_i^z,
$$
large-scale quantum Monte Carlo and stochastic analytic continuation show that the fracton phase exhibits strong anisotropy in both real-space correlations and dynamical structure factors. For example, in the fracton phase at \(h_x=0.8\), the nearest-neighbor lineon correlation satisfies
$$
C_{O_x}(x)\approx 0.421(3),\qquad C_{O_x}(y)\approx 0.0003(6),\qquad C_{O_x}(z)\approx 0.00002(6).
$$
The lineon spectral function disperses only along the allowed direction, whereas the fracton spectral function is essentially flat in momentum and remains gapped, with a peak energy around \(\Delta\approx 8\). A first-order transition into a trivial paramagnetic phase occurs near
$$
h_x \approx 0.9 \quad\text{and}\quad h_z \approx 0.3,
$$
depending on the perturbation channel [2203.13274].

At finite temperature, the equilibrium thermodynamics are unexpectedly simple. Exact partition-function calculations and bond-algebraic dualities show that the X-Cube model and its \( \mathbb{Z}_p \) generalization have no finite-temperature phase transitions. For open boundaries, the partition function is
$$
\mathcal Z_{\text{Open}} =2^{3L^3+6L^2+3L}\,C^{L^3}\,[C^3+S^3]^{(L-1)^3},
$$
with \(C=\cosh(\beta a)\) and \(S=\sinh(\beta a)\), and the thermodynamic free-energy density agrees with the cylindrical and periodic cases. Under a solvable Glauber-type bath, autocorrelations decay exponentially with finite rates at all nonzero temperatures, so the model is thermally fragile rather than glassy in the sense of system-size-dependent positive-temperature memory [1812.04561].

Quantum phase transitions out of the X-cube phase depend strongly on which excitations condense. Condensing fracton dipoles or lineon dipoles produces stacked deconfined gauge theories; condensing one lineon species yields a stack of 2D \( \mathbb{Z}_N \) gauge theories; condensing two or more lineon species yields a trivial paramagnet. For \(N>4\), the dipole-condensation transitions can be governed by stable critical points described by decoupled stacks of \(d=2\) conformal field theories, while lineon condensation can produce a gapless intermediate phase described as an array of \(d=1\) conformal field theories [2107.09073].

A complementary route through the phase diagram arises when the X-cube model competes with the 3D toric code. In the combined Hamiltonian, the exact commuting decomposition
$$
\mathcal{H}=\mathcal{H}_A+\mathcal{H}_B,\qquad [\mathcal{H}_A,\mathcal{H}_B]=0,
$$
allows a detailed analysis. Besides the toric-code and X-cube phases, two further phases appear, both adiabatically connected to classical limits with nontrivial sub-extensive degeneracies. All phase transitions are found to be first order [2106.05749].

## 6. Realizations, algorithms, defects, and extensions

The X-cube ground state can be prepared exactly by a Clifford-only quantum circuit. On an \(L\times L\times L\) cubic lattice on a 3D torus, the stated preparation depth is
$$
12L+11.
$$
The construction reformulates the projector application problem geometrically, using Hadamard and CNOT gates in a layered pattern. A related gluing method extends the preparation strategy via measurements and membrane-based correction operators, reflecting the fact that X-cube excitations are fractonic rather than freely mobile [2210.01682].

A dynamical measurement-based realization is provided by the X-Cube Floquet code. Built from intersecting 2D 4.8.8 Floquet-code layers in the \(xy\), \(yz\), and \(xz\) directions, it uses a period-six sequence of local weight-two measurements:
1. yellow checks and on-site checks,
2. blue checks,
3. green checks,
4. yellow checks,
5. blue checks,
6. green checks.
Within one Floquet period, the codespace alternates between X-cube fracton order and layers of entangled 2D toric codes. The encoded logical-qubit count is
$$
6L-3,
$$
and the model is argued to have a non-zero error threshold. The same work also gives a two-body Hamiltonian realization of the coupled-layer X-cube limit [2211.05784].

Defects and non-Euclidean geometries further enlarge the family of X-cube-like phenomena. Screw dislocations change \(\log_2 \mathrm{GSD}\) by a finite amount and can enable fractons or lineons to tunnel along the defect line, exposing both the foliated structure and genuinely fractonic mobility effects [2012.07263]. On hyperbolic lattices embedded in \(H_2\times S^1\), the Y-cube generalization replaces some X-shaped vertex terms by Y-shaped ones and supports treeons, excitations confined to a fractal tree rather than a line. In flat limits, these treeons reduce to either lineons or planeons, showing explicitly that subdimensional mobility can itself be geometry-dependent [2211.15829].

Taken together, these constructions show that the X-cube model functions both as a fixed exactly solvable phase and as a template for a broader class of geometry-sensitive fracton systems. This suggests that its most durable significance lies in unifying stabilizer-code realizations, continuum higher-rank gauge structure, and lattice-geometric mechanisms of restricted mobility within a single paradigm [1712.04511].

Source: https://www.emergentmind.com/topics/x-cube-model