WorksPhysics and Astronomy
Stabilizer models of achiral abelian topological orders
Abstract
We construct continuous-variable stabilizer codes realizing every achiral 2D abelian topological order, both bosonic and fermionic. Starting from an arbitrary signature-zero integer -matrix, we show that the corresponding topological order can be obtained by condensing a lattice of dyons with trivial pairwise mutual braiding in an gauge theory. We implement this condensation microscopically by gauging a non-on-site 1-form symmetry of the toric code.
Introduction
Topological stabilizer codes provide exactly solvable models of abelian topological phases, allowing their universal properties to be understood directly at the microscopic level. The 2D toric code is the paradigmatic example: the fusion rules and braiding statistics of its anyonic excitations can be derived explicitly from the underlying lattice model [1]. At the same time, topological stabilizer codes are a promising avenue toward quantum error correction, since they encode quantum information nonlocally and are therefore intrinsically robust against local errors [2].
In Ref. [3], it was shown that every gappable 2D bosonic abelian topological order—i.e., one that admits a gapped boundary [4]—can be realized by a topological stabilizer code on a system of qudits. For example, the double semion theory [5] admits a stabilizer code realization using qudits [3]. This result has been extended to the fermionic setting in Ref. [6], which constructed Majorana-qudit stabilizer code models of all gappable fermionic abelian topological orders. Conversely, it is believed that 2D qudit or Majorana-qudit stabilizer codes can only realize gappable abelian topological orders; this has been proven in the translation-invariant bosonic setting [7]. Thus, gappability is conjectured to precisely characterize the class of 2D abelian topological orders realizable by stabilizer codes on finite-dimensional local Hilbert spaces.
The authors of Ref. [8] constructed 2D continuous-variable (CV) stabilizer codes [9] realizing certain achiral topological orders that are nevertheless ungappable, such as that of Chern-Simons theory [4]. These examples demonstrate that CV stabilizer codes can realize topological orders beyond those accessible to qudit stabilizer codes, raising a natural question: which 2D topological orders can be realized by CV stabilizer codes? In this paper, we construct stabilizer code realizations of all achiral abelian topological orders, both bosonic and fermionic. We further conjecture that achirality precisely characterizes the class of 2D abelian topological orders realizable by CV stabilizer codes.
Building on the insights of Ref. [8], our construction begins from the observation that every achiral bosonic abelian topological order can be obtained by condensing a -dimensional lattice of mutually bosonic dyons in an gauge theory. To implement this condensation in a quantum lattice model, we start from the toric code and gauge the anomaly-free 1-form symmetry generated by closed string operators of the mutually bosonic dyons. This yields a CV stabilizer code realizing the desired topological order. We further generalize this approach to achiral fermionic topological orders, which can be obtained by condensing fermionic lattices in an gauge theory. This requires the introduction of fermionic modes that trivialize the anomaly of the corresponding 1-form symmetry [10]. Gauging the 1-form symmetry then yields a Majorana-CV stabilizer code realizing the achiral fermionic topological order.
Dyon condensation in gauge theory
We begin by reviewing the formalism of dyon condensation in gauge theory [8], allowing both electric charges and magnetic fluxes as dynamical excitations. The dyons of the theory are labeled by their flux and charge,
with fusion given by addition in . We use a normalization in which represents flux of the th gauge field. The topological spins and mutual braiding statistics are encoded by the symmetric bilinear form
In particular, the topological spin and braiding are
for any pair of dyons . Consider a -dimensional lattice of dyons . Assume that the dyons in have trivial pairwise braiding, that is,
In particular, , so every dyon is either a boson or a fermion. We call bosonic if every dyon in is a boson, and fermionic otherwise.
A bosonic (resp. fermionic) lattice can be condensed in a bosonic (resp. fermionic) system [10, 11]. Condensing has two effects. First, dyons that braid nontrivially with any become confined; the dyons that survive as deconfined excitations are those which braid trivially with , and therefore form the dual lattice
Note that . Second, deconfined dyons that differ by fusion with an element of become identified. Thus, the fusion group of the resulting topological order, modulo local excitations, is the discriminant group [12]
We illustrate the formalism with four examples obtained by condensing two-dimensional dyon lattices in gauge theory. The condensate lattice is generated by and , and the dual lattice is generated by and . These lattices are depicted in Figure 1.
Figure 1. Condensate lattice (large dots) and dual lattice (small dots) for (a) toric code, (b) double semion, (c) , and (d) the achiral theory.
toric code:
is generated by charge 2 and flux. The unit charge and flux descend, respectively, to the and anyons of the toric code. Equivalently, we may take , , as shown in Figure 1(a).
Double semion:
Here, and are a semion and antisemion, respectively. This condensation can be viewed as a two-step process: first condense and to obtain the toric code, then condense its boson, corresponding to . This example admits the following generalization:
:
The classes and generate the and factors, respectively. The theory is bosonic when both and are even, and fermionic otherwise.
Achiral theory, :
The condensate lattice is bosonic, with discriminant group . It is generated by , with , and the generator has topological spin
We note that these examples include minimal representatives of all nontrivial classes in the achiral subgroups of the prime components of the pointed Witt group [13]. Let denote the subgroup represented by achiral abelian bosonic topological orders, and define
Minimal representatives of the nontrivial classes are given by for and the achiral theory for each prime .
Chern-Simons theory description
Condensing the dyon lattice amounts to gauging the 1-form symmetry generated by the corresponding Wilson lines, since gauging has the effect of proliferating symmetry defects [14]—which are precisely the dyons in [10, 14]. Here we describe this gauging in the continuum theory. As we will see, this gives rise to a Chern–Simons theory with -matrix equal to the Gram matrix of :
where
is a basis of .
First, recall the BF theory [15] description of the topological sector of gauge theory. The theory has noncompact gauge fields and noncompact Lagrange multiplier fields , with action
Here we have defined the noncompact gauge field
The equations of motion impose the flatness condition . The action is invariant under gauge transformations
Since the gauge fields are noncompact, is a genuine -valued function. Wilson lines are labeled by arbitrary dyons as
The 1-form symmetry acts as for closed ; its action on Wilson lines is
where . Note that is unquantized, making this a 1-form symmetry transformation. We are interested in gauging subgroups of this 1-form symmetry.
Let be the matrix of condensed dyons, such that the condensate lattice . Consider the 1-form subgroup consisting of transformations
where are closed () and obey the quantization . Gauging this 1-form symmetry identifies the gauge field configurations which differ by (16). In other words, these transformations become gauge transformations. Requiring invariance of the Wilson lines imposes the quantization condition
We may therefore label these deconfined Wilson lines by an integer vector . Defining , the deconfined Wilson lines are
Hence are properly quantized gauge fields. Substituting into gives
We thus obtain a Chern–Simons theory with -matrix
The condition (4) follows from the requirement that the theory is properly quantized, , and therefore gauge invariant. The theory thus describes a 2D abelian topological order [16]. Moreover, is even, meaning that for all , precisely when is bosonic. (When is odd, the theory is a spin Chern–Simons theory, whose definition requires a choice of spin structure on spacetime [17].)
The anyons of the theory are labeled by integer vectors . Vectors in represent local excitations, which are all bosonic when is even, but otherwise include fermions. The fusion group is thus
in agreement with the dyon condensation picture. To see the second isomorphism, observe that a dyon belongs to precisely when , and hence . Under the isomorphism given by , the sublattice maps to . This gives the isomorphism .
The topological spins and mutual braiding statistics are [16, 17]
For odd , the topological spin depends on the representative of the equivalence class: attaching a local fermion reverses its sign. These expressions also agree with the dyon condensation picture, since the Gram matrix of is . Indeed, defining , where is the th basis vector of , gives
For instance, the lattice of the above example 4 gives
Under the isomorphism in (21), the anyon corresponds to .
Finally, the chiral central charge is given by the signature of [18],
where denote the numbers of positive and negative eigenvalues of . Since is congruent to , it has the same signature by Sylvester’s law of inertia. As , the resulting topological order is achiral, with . Conversely, every achiral Chern-Simons theory arises from dyon condensation in gauge theory. To see this, diagonalize as
where is orthogonal and are positive diagonal matrices. Define
Since
the lattice satisfies . As every achiral 2D abelian topological order can be described by a zero-signature -matrix [16, 17], we conclude that dyon condensation in gauge theory realizes precisely the class of achiral 2D abelian topological orders.
Lattice implementation
We now implement the gauging of the 1-form symmetry on the toric code, which serves as a lattice realization of gauge theory [8].
The RN toric code
The Hilbert space consists of continuous-variable modes on each edge of the square lattice. Each mode has position and momentum operators and satisfying . We define displacement operators
which obey the commutation relation
Our normalization is chosen so that shifts momentum by , mirroring the convention that denotes a flux. We denote the displacement operators acting on the th CV mode of edge by and , with .
The Hamiltonian of the toric code is
where
Violations of the vertex constraint are electric gauge charges, whereas violations of the plaquette constraint are magnetic gauge fluxes. In particular, an electric charge at vertex corresponds to eigenvalue of for all , and a magnetic flux through plaquette corresponds to eigenvalue of for all .
The closed string operators associated with the dyons in form a non-on-site 1-form symmetry of the toric code. Let denote the th generator, corresponding to the dyon . There is some freedom in the choice of the microscopic representative of the symmetry transformations associated to . We consider the symmetry generated by operators
where
The 1-form symmetry generated by is anomaly-free if is a boson, but anomalous if is a fermion [10, 14]. We denote the fermion parity of the dyon as
For each such that is fermionic, we introduce a physical fermionic mode at each vertex described by Majorana operators and . The Majorana modes form a Hilbert space . These ancillary modes trivialize the possible anomalies of the 1-form symmetry [19]. We will gauge the augmented toric code Hamiltonian,
Here and below, a fermionic factor raised to denotes the identity.
Gauging the 1-form symmetry
Adapting the defect-based gauging approach of Ref. [20], our procedure consists of three steps:
Introduce a 2-form gauge field living on the vertices of the lattice, forming a Hilbert space consisting of rotor degrees of freedom on each vertex. The total Hilbert space is thus . Each rotor has angular coordinate and angular momentum satisfying . The eigenvalues of the angular momentum operators on vertex encode the symmetry defect at . The Hilbert space is thus spanned by defect configurations . We define displacement operators
obeying the commutation relation
In particular, for . We denote the displacement operators acting on the th rotor on vertex by and , for .
Impose Gauss law constraints for each edge , oriented from to , and , where the Gauss law operators satisfy the following properties:
I. for some unitary operator diagonal in the basis and supported near .
II. in the bosonic case, and in general.
III. .
Taken together, properties I and II imply that can be interpreted as a short string operator that hops a symmetry defect from to . In the fermionic case, the additional Majorana factors in property II implement fermionic gauging by attaching a physical fermion to each fermionic symmetry defect. Property III ensures that the constraints can be imposed simultaneously, either directly or energetically. To obtain a stabilizer code model, we will impose the constraints energetically.
Define the gauged Hamiltonian,
in terms of defect insertion operators defined for each defect configuration . Specifically,
in terms of elementary defect insertion operators
Here, is an arbitrary choice of semi-infinite oriented path terminating at . For concreteness, we take to be a straight line from the west. Thus creates a symmetry defect at vertex . Note that these operators are well-defined due to commutativity of the Gauss law operators. By construction, commutes with the Gauss law operators.
We carry out this procedure explicitly by choosing the Gauss law operators
where
It is straightforward to verify that these operators satisfy properties I–III. The Majorana dressing is closely related to exact two-dimensional bosonization [22]; products of Majorana hopping operators around a plaquette yield fermion parities at two opposite corners, explaining the parity factors in property II.
The resulting defect insertion operator has the form
where denotes the edge situated to the northeast of . This operator commutes with unless , with unless is the southwest corner of , and with unless and . Imposing the Gauss law constraints energetically, the gauged Hamiltonian is therefore
where [23]
Here we have introduced the notation
Topological order of the gauged model
We now examine how the dyon condensation rules arise microscopically in the gauged model. String operators for generic dyons of the original toric code fail to commute with the Gauss law operators along their length, reflecting their confinement. The dyons in , however, admit string operators that commute with the Hamiltonian away from their endpoints.
To construct these strings, consider the product
where and is the southwest corner of . Its action on the th rotor at is . For , this action is trivial, leaving the closed string
Multiplying these elementary loops over a region cancels their action on the interior edges, giving a string supported near the boundary. For , the same construction instead leaves a nontrivial action on rotors throughout the interior. The corresponding open string along an oriented path is
Here is the edge situated to the southwest of , for eastward or northward steps and for westward or southward steps, and for horizontal edges and for vertical edges. An example is shown in Figure 2(b). Each elementary segment commutes with every Gauss law operator when for all . The string therefore creates excitations only at its endpoints: a dyon at the initial endpoint and at the final endpoint, with the flux at each endpoint residing on the plaquette southwest of its charge. These are the deconfined anyons predicted by the dyon condensation picture.

Figure 2. (a) The defect insertion operator , which acts as on solid blue edges and on solid red edges. The black dot represents vertex . The Majorana factors are suppressed. (b) An open string operator for the deconfined dyon , acting as on solid blue edges and on solid/dashed red edges. excites the plaquette and vertex term at each endpoint.
A pair of deconfined anyons belong to the same superselection sector, modulo local excitations including physical fermions, whenever . To see why, consider the local operator
The rotor shift in creates charge at the southwest corner of and flux through , while the two shifts move the charge to . Moreover, commutes with every Gauss law operator and every . It therefore creates precisely the endpoint excitation of a string at : a flux through the plaquette southwest of a charge. Thus, applying at a string endpoint transforms into , identifying dyon labels modulo . We conclude that the anyons of the gauged model have fusion group , with braiding statistics inherited from the original dyons.
We compute the ground state degeneracy of the purely bosonic model on a torus in Appendix B. We find that , with logical operators given by noncontractible strings associated with dyons in . Together with the fusion rules and braiding statistics, this calculation confirms that the model realizes the topological order described by the input -matrix.
Discussion
We have constructed CV stabilizer code realizations of all achiral 2D abelian topological orders. Our construction implements dyon condensation by gauging a non-on-site 1-form symmetry; a general formalism for gauging anomaly-free 1-form symmetries will be discussed in future work [24]. Conversely, we believe that chiral topological orders cannot be realized by CV stabilizer codes purely in two spatial dimensions. On the other hand, we do expect it is possible to realize any abelian topological order at the surface of a 3D CV stabilizer code with no deconfined bulk excitations. We conjecture that such 3D stabilizer codes can be disentangled by CV Clifford quantum cellular automata (QCAs), which are nontrivial in the Clifford category precisely when the surface topological order is chiral. Related connections between surface topological order and QCAs have been established for qudit systems [25, 26, 27, 28].
The Hamiltonians in this paper should be understood as ideal CV stabilizer models: exact codewords involve non-normalizable states, as in multimode GKP codes [9]. An additional ingredient here is the presence of continuous stabilizers, whose integrals formally impose exact constraints. We expect that these constraints can be softened into finite-strength energy penalties by cutting off the domain of integration. As in Ref. [8], we expect that adding suitable quadratic terms in the CV variables and rotor angular momenta will produce normalizable states with a finite energy gap while preserving the desired topological order.
**Acknowledgments—**We are grateful to Clay Cordova, Michael Levin, and Valerie Wu for insightful discussions. The authors used various versions of Chat-GPT and Claude to assist with drafting and revising exposition, as well as exploring, developing, and checking mathematical arguments. The work of MQ is supported in part by the Simons Collaboration on Ultra-Quantum Matter, which is a grant from the Simons Foundation (Grant No. 651442, D.T.S.). The work of WS is supported by U.S. Department of Energy Grant 5-29073. The authors of this paper were ordered alphabetically.
Stabilizer models of 2 by 2 even -matrices
In this appendix, we specialize the construction of the main text to the bosonic case, then describe two alternative CV stabilizer code constructions of bosonic topological orders described by a even -matrix. Let , , generate the condensate lattice , with . The Hilbert space is composed of one CV mode per edge and two gauge rotors per vertex.
In this case, the gauged Hamiltonian (44) is
The gauged star and plaquette operators are
The Gauss law operators are
where
The first alternative construction uses a simplified choice of Gauss law operators on . In particular, we define
With this choice, the toric code operators and have gauged counterparts
The gauged Hamiltonian, with energetically imposed Gauss law constraints, is thus
The second alternative construction implements condensation of and directly in , without introducing gauge rotors. Choose a basis , , of the dual lattice . The Hamiltonian is composed of elementary hopping operators for the condensed dyons and elementary loop operators for the surviving dyons:
where
and
Ground state degeneracy on the torus
In this appendix, we compute the logical operators and ground state degeneracy of the gauged model (44) in the case of bosonic , corresponding to a topological order described by an even signature-zero matrix. Logical operators are closed string operators for deconfined dyons in acting on noncontractible cycles, and closed (contractible and noncontractible) string operators for condensed dyons in are stabilizers. Hence
and the ground state degeneracy on the torus is . To obtain these results, we use the polynomial formalism for translation-invariant stabilizer codes developed by Haah [29], which we adapt to CV degrees of freedom (real variables and rotors) in App. B 1. We then apply the formalism to our model and derive Eq. (61) in App. B 2.
Polynomial formalism for CVs
We begin with a review of the polynomial formalism for translation-invariant stabilizer codes, appropriately modified to accommodate real variable and rotor degrees of freedom. For an abelian group , let be the module of Laurent polynomials in variables and with coefficients in . In the -dimensional qudit setting, is typically taken to be . For us, will be when considering real variables, and or when considering rotors. Periodic boundary conditions are implemented by the quotient . We will often abuse notation and use for both as well as its quotient by , , and the meaning of depends on whether the model is on the infinite plane or the torus.
Consider a translation-invariant system with real variable degrees of freedom and rotors in each unit cell. The Pauli group of the system is
We adopt the following notation for products of Pauli operators at a single site:
for , , and . The operators act on the real variable degrees of freedom, while the operators act on the rotor degrees of freedom. A general Pauli operator is a tensor product of , , , operators over different sites . Pauli operators are represented as polynomials via
To distinguish between Pauli operators and their polynomial representation, we use capital letters (, , etc.) to denote Pauli operators and lowercase letters (, , etc.) to refer to their polynomial representation. The phase obtained from commuting Pauli operators through one another is
Here denotes the constant coefficient, taken after imposing the periodic identifications on the torus. The overline denotes the antipode map and is the symplectic matrix
A translation-invariant stabilizer code is specified by a translation-invariant, mutually commuting subgroup of the Pauli group that does not contain any nontrivial scalar multiples of the identity. Its code space is the mutual eigenspace of all . A stabilizer may be either continuous or discrete: if is a continuous stabilizer, then for are also stabilizers; if is a discrete stabilizer, then is a stabilizer for .
We may pick a translation-invariant set of generators of these stabilizers. With continuous and discrete stabilizer generators per unit cell,
The stabilizer subgroup can be thought of as the image of a map . This map sends each stabilizer generator to its corresponding Pauli operator; its kernel encodes relations between the stabilizers. The stabilizers commute when
where the dagger operation indicates transposition followed by the antipode. This condition is understood to mean exactly in the block involving continuous generators, and mod in the block of discrete generators.
An excitation of a stabilizer code is an assignment of a phase to each stabilizer generator. The space of excitations forms . There is an excitation map which maps a Pauli operator to the excitations it creates, defined as
In this language, the stabilizer condition (B8) can then be written as ; any Pauli operator in the image of the stabilizer map creates no nontrivial excitations. We may summarize our discussion so far by saying that the maps and form a chain complex
The logical operators on a stabilizer code are Pauli operators which commute with the stabilizers (and hence create no excitation), which are themselves not stabilizers. The logical group is therefore the homology of this complex at :
As an illustration, let us analyze the instructive example of the toric code (31) on the infinite plane and the torus using this formalism. The purpose of studying this familiar example is to interpret the chain complex of (B10) as a cellular chain complex of the spatial manifold, so that computing the logical group reduces to computing cellular homology. This interpretation is naturally expressed in terms of dyon ribbon operators. The intuition involving ribbon operators generalizes readily to the general bosonic -matrix case, which we treat in the next subsection.
The toric code has degrees of freedom per unit cell ( on each edge), and consists of continuous stabilizer generators per unit cell, given by the and stabilizers for . We therefore have , , and . We label plaquettes by the vertex at their southwest corner, and horizontal (vertical) edges by the vertex at their west (south) endpoint. With these conventions, the continuous stabilizers are represented in the polynomial formalism as
Here is the -component unit vector with 1 in the -th component and 0 in all others. The stabilizers can be organized into the stabilizer map taking the form
where each entry should be understood as a scalar multiple of the identity matrix. The excitation map is
We now explain the geometric picture that will allow us to compute the logical group as the cellular homology group of the spatial manifold or . The cellular chain groups , , and are formal sums with coefficients of faces, edges, and vertices, respectively. They can be described in the polynomial language as
With the same labeling conventions for faces and edges as before, the boundary maps and are given by
Each entry above should be understood as a scalar multiple of the identity matrix. The homology group vanishes for and is for .
Notice that, as abelian groups, , , and . Importantly, these isomorphisms can be chosen so that and become the boundary maps and . We identify by the map in . This assigns the plaquette stabilizer to its corresponding face, and the vertex stabilizer to its northeast face. We identify using ribbon operators. To an edge with coefficient , we assign the Pauli operator
which is the elementary ribbon operator (see Eq. (49)) for the dyon associated to the edge $e. It creates the dyon at the west (south) endpoint of and at the east (north), with each flux in the face southwest of the charge. Every Pauli operator on the real variable degrees of freedom can be uniquely written as a product of these elementary ribbon operators, which identifies with 1-chains:
Since the elementary ribbon operator excites only its endpoints, we identify by associating the charge and flux of each endpoint excitation to its corresponding endpoint vertex. This gives the isomorphism
Under this identification, the flux excitation at a plaquette lives at the vertex at its northeast corner. One can check that, using the identifications above,
and
We therefore identify with , and with . Thus the logical group is . On the torus, the logical group is continuous, and the code space is infinite-dimensional. The number of encoded real variable degrees of freedom follows from a generalization of the usual counting argument: on an torus, there are degrees of freedom, and independent continuous stabilizers which each remove one degree of freedom. Thus the toric code encodes degrees of freedom.
Application to the lattice model
We will now use the polynomial approach to analyze the Hamiltonian (44) in the bosonic case, and show that it realizes the topological order described by the corresponding -matrix Chern-Simons theory. In particular, we show that the group of logical operators on the torus is
Our method will be similar in spirit to the one used above in the analysis of the toric code: we will show that the stabilizer and excitation maps are related to boundary maps of an appropriate cellular chain complex. The additional ingredients are the discrete Gauss law stabilizers, which have the effect of confining dyons outside of , and turning closed (contractible or noncontractible) loops of ribbon operators of into stabilizers.
The Hilbert space of our model consists of real variable degrees of freedom on each edge, and rotors on each site, for a total of CV degrees of freedom per unit cell. The Pauli group and symplectic form are
Let us also denote by the subgroup of Pauli operators which only act on the degrees of freedom
The Hamiltonian consists of stabilizer generators per unit cell: star terms, plaquette terms, horizontal Gauss laws, and vertical Gauss laws. The star and plaquette stabilizers are continuous stabilizers, while the Gauss law stabilizers are discrete stabilizers. The space of stabilizer generators is therefore
The space of excitations is
The stabilizer map consists of rows and columns, and takes the form
We have defined the matrices
where and signify the strictly lower triangular and upper triangular parts of the matrix . The excitation map is
It can be verified explicitly that , with the Gauss law excitation rows understood modulo integers. Our aim now is to compute the group of logical operators on the infinite plane and the torus. Recall that is given by the homology of the chain complex (B10) at . We will compute this homology group in two steps. First, we show that the support of any operator that creates no excitation (equivalently, any operator in ) can be “cleaned” so that it only acts on the real variable degrees of freedom and commutes with the Gauss laws. This allows us to consider instead the complex
Here, we have defined as the subspace of which maps to stabilizers supported on the real variable degrees of freedom, and as the subspace of excitations which violate only the star and plaquette stabilizers. Next, we will show that this chain complex admits a simple geometric interpretation in terms of homology on the plane and the torus. We show, for the square lattice cellulation of and , that
and provide identifications such that and can be interpreted as geometric boundary maps. It then follows that the homology of this complex is canonically isomorphic to . This verifies the natural expectation that there are no nontrivial logical operators on the plane, and that the logical operators on the torus are represented by string operators supported on noncontractible cycles of the torus.
The first step is to show that any element of has a representative in . Let
be an arbitrary element of . The first rows of are
Evaluating both sides at (in other words, summing over all coefficients) gives
Since , we have . This means that can be written as
We can then add Gauss law stabilizers
to so that the component of vanishes. Having shown this, let us now take an arbitrary element with vanishing , but not necessarily vanishing . To this , we can add and stabilizers
for satisfying
Here, in an abuse of notation, we have used the same symbol to denote its lift into . For a given lift, there is a unique satisfying the above since . After adding this combination of star and plaquette stabilizers, acts nontrivially only on the real variable degrees of freedom. This allows us to restrict to the subcomplex
without changing the homology . Finally, writing , we can further restrict to to get (B30). Since logical operators belong to the kernel of , we lose nothing by making this further restriction to . This also allows us to restrict to , the subspace of excitations of the vertex and plaquette stabilizers. This completes the reduction to the chain complex (B30).
We have shown that the logical operators are given by the homology of (B30) at the middle term . We now provide the geometric interpretation of the chain complex in terms of cellular homology, beginning with the map . Recall from our analysis of the toric code that
where a ribbon operator (B17) in is mapped to its corresponding edge in via (B18). Similarly, is the set of excitations of the and stabilizers. Just as in the toric code, we associate to a vertex with coefficient the dyon excitation with at and at the plaquette southwest of , giving .
Restricting to restricts the coefficients of to . To see this, let be given by (B32), with . Elements of obey
These conditions imply that the ribbon operators associated to
create dyons in . Therefore . The map acts on in precisely the same way as the excitation map of the toric code. We can therefore identify with .
Let us now understand and . First, observe that is isomorphic to : given an element , we can think of
as an element of and
as an element of . For to be an element of , must have vanishing support on the rotor degrees of freedom. This requirement imposes
The first equation above states that is closed, so . To make sense of the second equation, define . The second equation becomes
This step makes use of the fact that is closed. Therefore, is an element of . This gives the isomorphism
between and . Under this isomorphism, the map acts on as
where is the matrix of condensed dyons. Therefore acts via the boundary map on the summand, and by the map of coefficients on the summand. We can now identify our desired homology as
We therefore recover the fact that the logical group is trivial on the infinite plane, and on the torus. The logical operators are given by the closed noncontractible ribbon operators for dyons in .
It follows the ground state degeneracy on the torus is . This is due to the following fact: for a GKP code on a finite number of modes with finite, the dimension of the code space is given by . This is a standard fact in the theory of lattice GKP codes [9, 30], which consist of only discrete stabilizers; it also holds in our context, where some of the stabilizers are continuous. To apply this argument, we first embed each rotor in a real variable degree of freedom stabilized by . The key point is that, since all of the continuous stabilizer commute, and each independent continuous stabilizer removes one mode, they can be completely disentangled via Gaussian unitaries without affecting the structure of the logical group. Since is finite, the discrete stabilizers act on the remaining modes as a full-rank lattice (otherwise there would be a continuous family of logical operators). Thus the code on the remaining modes is an ordinary multimode GKP code, which satisfies .
References
- [1]A. Yu. Kitaev, Fault-tolerant quantum computation by anyons, Ann. Phys. 303, 2 (2003).DOI
- [2]E. Dennis, A. Kitaev, A. Landahl, and J. Preskill, Topological quantum memory, J. Math. Phys. 43, 4452 (2002).DOI
- [3]T. D. Ellison, Y.-A. Chen, A. Dua, W. Shirley, N. Tantivasadakarn, and D. J. Williamson, Topological stabilizer models of twisted quantum doubles, PRX Quantum 3, 010353 (2022).DOI
- [4]M. Levin, Protected edge modes without symmetry, Phys. Rev. X 3, 021009 (2013).
- [5]M. A. Levin and X.-G. Wen, String-net condensation: A physical mechanism for topological phases, Phys. Rev. B 71, 045110 (2005).
- [6]M. Sun, Z. Wang, N. Tantivasadakarn, and Y.-A. Chen, Majorana-Pauli stabilizer codes and duality webs of fermionic topological phases (2026), arXiv:2606.25048.
- [7]B. Ruba and B. Yang, Witt groups and bulk-boundary correspondence for stabilizer states (2025), arXiv:2509.10418.arxiv.org/abs/2509.10418
- [8]J. C. Magdalena de la Fuente, T. D. Ellison, M. Cheng, and D. J. Williamson, Topological stabilizer models on continuous variables, Phys. Rev. X 16, 011054 (2026).
- [9]D. Gottesman, A. Kitaev, and J. Preskill, Encoding a qubit in an oscillator, Phys. Rev. A 64, 012310 (2001).DOI
- [10]L. Bhardwaj, D. Gaiotto, and A. Kapustin, State sum constructions of spin-TFTs and string net constructions of fermionic phases of matter, J. High Energy Phys. 04, 096.
- [11]F. A. Bais and J. K. Slingerland, Condensate-induced transitions between topologically ordered phases, Phys. Rev. B 79, 045316 (2009).
- [12]V. V. Nikulin, Integral symmetric bilinear forms and some of their applications, Math. USSR-Izv. 14, 103 (1980).
- [13]A. Davydov, M. Müger, D. Nikshych, and V. Ostrik, The Witt group of non-degenerate braided fusion categories, J. Reine Angew. Math. 677, 135 (2013).arxiv.org/abs/1009.2117
- [14]D. Gaiotto, A. Kapustin, N. Seiberg, and B. Willett, Generalized global symmetries, J. High Energy Phys. 02, 172.
- [15]G. T. Horowitz, Exactly soluble diffeomorphism invariant theories, Commun. Math. Phys. 125, 417 (1989).
- [16]X.-G. Wen and A. Zee, Classification of Abelian quantum Hall states and matrix formulation of topological fluids, Phys. Rev. B 46, 2290 (1992).
- [17]D. Belov and G. W. Moore, Classification of abelian spin Chern-Simons theories (2005), arXiv:hep-th/0505235.DOI
- [18]X.-G. Wen, Topological orders and edge excitations in fractional quantum Hall states, Adv. Phys. 44, 405 (1995).DOI
- [19]In principle, it would suffice to introduce a single mode on each vertex. We introduce a mode for each fermionic generator as a matter of convenience.
- [20]S. Seifnashri, Lieb-Schultz-Mattis anomalies as obstructions to gauging (non-on-site) symmetries, SciPost Phys. 16, 098 (2024).
- [22]Y.-A. Chen, A. Kapustin, and D. Radicević, Exact bosonization in two spatial dimensions and a new class of lattice gauge theories, Ann. Phys. 393, 234 (2018).
- [23]The powers Ãᵖ and B̃ᵠ are defined by multiplying every Pauli exponent in the expressions for à and B̃ by ϕ and q, respectively, before simplifying any factors that equal the identity. For instance, a rotor factor Zᵐ with m ∈ ℤ contributes Zᵐᵠ to Ãᵖ.
- [24]M. Qi and W. Shirley, In preparation.
- [25]J. Haah, L. Fidkowski, and M. B. Hastings, Nontrivial quantum cellular automata in higher dimensions, Commun. Math. Phys. 398, 469 (2023).
- [26]J. Haah, Clifford quantum cellular automata: Trivial group in 2D and Witt group in 3D, J. Math. Phys. 62, 092202 (2021), erratum: J. Math. Phys. 63, 079901 (2022).DOI
- [27]W. Shirley, Y.-A. Chen, A. Dua, T. D. Ellison, N. Tantivasadakarn, and D. J. Williamson, Three-dimensional quantum cellular automata from chiral semion surface topological order and beyond, PRX Quantum 3, 030326 (2022).
- [28]B. Yang, Categorifying Clifford QCA, Commun. Math. Phys. 407, 77 (2026).arxiv.org/abs/2504.14811
- [29]J. Haah, Commuting Pauli Hamiltonians as maps between free modules, Commun. Math. Phys. 324, 351 (2013).
- [30]S. Chakraborty and V. V. Albert, Stabilizer codes over general phase spaces (2026), arXiv:2610.04694.arxiv.org/abs/2610.04694
