Introduction
Anyon condensation is one of the few mechanisms that relate distinct topological orders in ( 2 + 1 ) (2+1) ( 2 + 1 ) dimensions. When a bosonic anyon condenses, some channels become part of the new vacuum, the remaining anyons may be identified, split, or confined, and only the deconfined excitations survive at the endpoint [1 , 2 , 3 ] . The same mechanism produces gapped boundaries and interfaces and underlies transitions between phases with non-abelian braid statistics [4 , 5 ] . It is not, however, a prescription for condensing any boson at will: the condensing channels must assemble into a consistent algebra of local operators, and the possible endpoints are constrained by the parent braiding [2 , 6 ] .
In the modular-category description this algebra is explicit. Let C \mathcal{C} C be the parent modular tensor category and let B ∈ C B \in\mathcal{C} B ∈ C be a condensable algebra: commutative, separable, and of trivial twist. The category C B \mathcal{C}_B C B of right B B B -modules describes excitations on the condensation wall, and its local subcategory C B l o c \mathcal{C}_B^{\mathrm{loc}} C B loc describes the deconfined bulk anyons of the condensed phase [2 , 4 ] . For the toric code with B = 1 ⊕ e B = \mathbf{1} \oplus e B = 1 ⊕ e , where e e e is the electric charge, the endpoint is trivial and the wall excitations form the representation category of Z 2 \mathbb{Z}_2 Z 2 . Data of this kind are specified throughout: the condensation algebra and its interface are taken as given. The algebraic framework constrains which endpoints are consistent; it does not by itself show that a microscopic transition is continuous, nor does it determine a critical universality class.
Ordinary global symmetries refine this picture. In a symmetry-enriched topological order, a finite group G G G acts on anyon labels, fusion spaces, symmetry fractionalization, and defect junctions [7 ] . Invariance of a set of anyon labels is not invariance of a condensation, because an invariant condensate must carry coherent junction data. The appropriate condition is that B B B admits a G G G -equivariant algebra structure, and this structure is additional data: even when g ( B ) ≅ B g(B) \cong B g ( B ) ≅ B for every g ∈ G g \in G g ∈ G , the isomorphisms need not satisfy the coherence imposed by the group multiplication. Bischoff, Jones, Lu, and Penneys showed that the equivariant algebra structure controls when the symmetry survives, that its absence forces symmetry breaking, and that in its presence the condensed local modules inherit the symmetry-enriched data and the condensation commutes with gauging [8 ] . That commutativity is the ordinary-group benchmark for the present work.
Gauging makes such comparisons tractable for three reasons. It turns symmetry defects into a dynamical network, reorganizing charge, flux, and defect sectors into a theory with different intrinsic anyon content rather than merely forgetting a symmetry. It converts symmetry-enrichment data into topological data against which a condensate, its symmetry action, its residual phase, and its interfaces can be tested. And it makes two specified orders of operations comparable: condense first and then gauge the induced operation, or gauge first and then condense the transported algebra. Two caveats belong here. Gauging requires gaugeability conditions and is not available unconditionally, and the operation below the wall cannot in general be chosen independently and then declared to agree with the operation above it. It is also worth separating two statements that are often conflated: the disappearance of intrinsic order under condensation does not imply that gauging an endpoint leaves a trivial vacuum. Complete Higgs behavior, where the final phase has a single intrinsically trivial vacuum, is a sharp test of this kind, and it appears below as an application rather than as the organizing question.
The modern language for these operations is that of topological defects. An ordinary group symmetry is represented by invertible defects whose fusion follows the group multiplication; generalized symmetry widens the dimensions and the organization of charged operators and of the defects acting on them [9 ] . Non-invertibility is a further, logically independent property: the fusion of defects need not admit a group-like inverse, and a defect fused with its dual may produce several sectors rather than the transparent defect alone. A dual defect is therefore not a fusion inverse, and generalized symmetry should not be identified with non-invertible symmetry, since higher-form symmetries are typically invertible. In finite ( 2 + 1 ) (2+1) ( 2 + 1 ) dimensions these defects form surfaces, line junctions, and point junctions, whose composition requires higher-categorical data and is not captured by a fusion-rule table [10 , 11 ] .
Recent work on non-invertible symmetries in ( 2 + 1 ) (2+1) ( 2 + 1 ) dimensions develops three related threads. The first constructs non-invertible defects by gauging, or by gauging on higher-codimensional loci, producing condensation defects with calculable fusion [12 , 13 , 14 ] . The second studies concrete gaugings, coset symmetries, and their fractionalization in topological orders, where gauging a non-normal subgroup is a standard source of non-invertible structure and leaves no ordinary quotient group [15 , 16 ] . The third characterizes gapped phases with categorical symmetry via symmetry topological field theory and realizes them on the lattice and in tensor networks [17 , 18 , 19 , 20 , 21 , 22 ] . Two distinctions matter here. An abstract symmetry fusion 2 2 2 -category, a realization of it in a parent theory, and a selected gauging algebra K K K are different objects: we work with a realized separable operation K K K , and compatibility with one such K K K is not automatically compatibility with every defect of an abstract symmetry category. Correspondingly, the fusion category of wall excitations of K K K is not the full symmetry 2 2 2 -category, and we never treat it as such.
These developments leave open a question that endpoint comparison does not answer. Given a prescribed condensate with its full algebra and interface, and a prescribed separable operation, when does the whole algebra network cross the operation compatibly, retaining the actual interfaces and allowing the two composite orders to be compared? Four sub-questions guide the answer: how the ordinary-group equivariant condition is expressed when the operation is not group-like; why label invariance and permutation of vacuum labels fail to describe charged condensation channels and defect junctions; what distinguishes an equivalence of endpoints from an equivalence of specified composite walls; and whether ordinary parent vacuum insertions still agree with the operators of the final vacuum when the specified unit itself absorbs nontrivial parent anyons. The inadequacy of a vacuum-label description is already visible in a small example: for the non-invertible subgroup gauging of the toric code associated with S 3 / Z 2 S_3/\mathbb{Z}_2 S 3 / Z 2 , obtained by gauging the Z 2 \mathbb{Z}_2 Z 2 subgroup generated by a transposition, a mixed endpoint carries two vacua of different intrinsic order, one retaining toric-code order and one trivial [15 , 16 ] , and the relation between them involves a charged channel that is local only when attached to an electric line, so no invertible permutation of vacuum labels can express it. We return to this example in § 5 .
The input of this paper is a tuple ( C , K , B , τ ) (\mathcal{C},K,B,\tau) ( C , K , B , τ ) : a parent anyon category C \mathcal{C} C , a condensable algebra B B B that may be disconnected, a specified separable operation K K K realized by an E 1 E_1 E 1 -algebra in the suspended defect category E = Σ C \mathcal{E}=\Sigma\mathcal{C} E = Σ C , and a coherent reversible pull-through τ \tau τ of the whole algebra B B B across K K K . Equivalently the input is ( C , K , A = Σ B , λ ) (\mathcal{C},K,A=\Sigma B,\lambda) ( C , K , A = Σ B , λ ) , with A A A the marked suspension of B B B and λ \lambda λ the corresponding marked pseudodistributive law (30) (see Proposition A.1). The pull-through preserves the multiplication and unit of the condensate, the multiplication and unit of K K K , the junction coherences of the defects, the charged channels, and the specified condensation interface together with its marking: the free interface η A \eta_A η A , its adjunction η A ⊣ η A R \eta_A \dashv\eta_A^R η A ⊣ η A R , and the monad identification η A R η A ≃ B \eta_A^R\eta_A \simeq B η A R η A ≃ B are part of the data rather than a convention that may be dropped when phases are compared. Allowing B B B to be disconnected keeps several locally distinguishable vacua in the parent phase; it is not a device for stacking independent topological orders. Two points of scope should be stated at once. Since separable operations of pure condensation type are included, we do not claim that every admissible K K K is the full gauging of a global symmetry. And a connected K K K means only that η K \eta_K η K is a simple 1 1 1 -morphism; strong connectedness is the additional transparent-unit condition d K ≃ 1 C d_K\simeq1_{\mathcal{C}} d K ≃ 1 C (the absorption algebra d K d_K d K is defined in § 4.1 ). The general framework does not require strong connectedness, while the fusion-category and classification statements are proved in the connected fusion setting; nothing is claimed for nonsimple tensor units.
The full input generates its own outputs. The A A A - and K K K -networks combine into a composite T λ = A ⊗ K T_{\lambda}=A\otimes K T λ = A ⊗ K whose relative algebras yield the transported condensate B K B_K B K and the induced operation K B K_B K B , while the specified walls and their comparison are collected in the 2-natural wall comparison Ξ λ \Xi_{\lambda} Ξ λ (§ 3 ). Both outputs come from the same composite network, and the comparison concerns the actual composite walls rather than final phases alone; it takes the form of a commuting square of marked walls,
W B K ∘ G K ≃ G K B ∘ W B , (1) \mathcal{W}_{B_K}\circ\mathcal{G}_K\simeq\mathcal{G}_{K_B}\circ\mathcal{W}_B, \tag*{(1)} W B K ∘ G K ≃ G K B ∘ W B , (1)
with G \mathcal{G} G a gauging wall and W \mathcal{W} W a condensation wall. Theorem 3.3 constructs B K B_K B K and shows that it is condensable, and Theorem 3.4 establishes the square; neither asserts that an arbitrary gauging commutes with an arbitrary condensation.
Compatibility admits one further presentation. Central matching states that compatibility is equivalent to a condensable algebra L L L in the gauged phase together with an identification of algebras in the wall category,
ξ : q ( L ) ⟶ ∼ ι ( B ) , (2) \xi:q(L)\overset{\sim}{\longrightarrow}\iota(B), \tag*{(2)} ξ : q ( L ) ⟶ ∼ ι ( B ) , (2)
where ι \iota ι brings parent anyons onto the specified wall and q q q restricts gauged anyons to it (Theorem 3.2); after reconstruction L L L is B K B_K B K . For a general specified operation, L L L alone does not determine the full compatible input ( B , τ ) (B,\tau) ( B , τ ) ; the parent algebra and its identification on the wall must also be retained. The specified unit may absorb a nontrivial parent anyon algebra. The two absorption algebras d K = ι R ( 1 K ) d_K=\iota^{R}(1_K) d K = ι R ( 1 K ) and a K = q R ( 1 K ) a_K=q^{R}(1_K) a K = q R ( 1 K ) , together with the local reduction lemma (Lemma 4.1), are the tools used in § 4 : the proofs below work directly with this reduction and prove the complete Higgs criterion in general form.
Vacuum data distinguish what the parent phase provides from what the final phase exhibits. Write r = dim Hom C ( 1 , B ) r=\dim\operatorname{Hom}_{\mathcal{C}}(1,B) r = dim Hom C ( 1 , B ) for the parent vacuum insertions and r c o m p r_{\mathrm{comp}} r comp for those compatible with the crossing, and let r f i n r_{\mathrm{fin}} r fin denote the dimension of the final vacuum algebra. In general r c o m p = r f i n r_{\mathrm{comp}}=r_{\mathrm{fin}} r comp = r fin does not hold: unit absorption can make the final vacuum algebra strictly larger than the compatible parent one. In the transparent-unit case the absorption dimensions also yield a residual-order identity for the final phase (Lemma B.2). § 5 illustrates these statements through three controlled comparisons: a fixed operation with different compatible condensations, a fixed parent condensate with different crossings, and a fixed transported condensate with different parent condensates when the specified unit absorbs a nontrivial parent algebra. For the fixed S 3 / Z 2 S_3/\mathbb{Z}_2 S 3 / Z 2 gauging of the toric code, the compatible data are built from four basic classes with connected transported condensate by taking orthogonal sums (Theorem 5.2), and a crossing relates distinct parent vacua only through charged channels (Proposition 5.1). One parent algebra admits two inequivalent crossings that induce different operations and leave different final phases (§ 5.4 ). The pure condensation operation with a nontransparent unit illustrates several distinct effects (§ 5.7 ). One compatible choice gives r c o m p < r f i n r_{\mathrm{comp}}<r_{\mathrm{fin}} r comp < r fin , and two non-isomorphic parent condensates give the same transported algebra. A different compatible choice realizes complete Higgs behavior with a non-Lagrangian parent.
The transparent-unit specialization recovers the familiar picture. When d K ≃ 1 C d_K\simeq1_{\mathcal{C}} d K ≃ 1 C the operation is strongly connected, the gauging wall is an orientation-reversed condensation wall with reverse condensate a K a_K a K , which need not be trivial, inverse transport is determined by L L L alone, and the compatible parent insertions and the final vacuum operators agree, r c o m p = r f i n r_{\mathrm{comp}}=r_{\mathrm{fin}} r comp = r fin . The complete Higgs criterion is proved in § 4 : after the specified unit absorbs d K d_K d K , it is stated for the transported condensate B ‾ = F d K ( B ) \overline{B}=F_{d_K}(B) B = F d K ( B ) in the reduced bulk M K \mathcal{M}_K M K (Lemma 4.1 and Theorem 4.3). In the transparent-unit specialization it reduces to the three conditions — B K B_K B K connected, r = κ K r=\kappa_K r = κ K , and every connected component B i B_i B i of B B B Lagrangian — which appear as Corollary 4.5; the existence criterion in terms of the Witt condition and the Tannakian property of the subcategory generated by a K a_K a K (Appendix § 8 ) belongs to the same specialization. Endpoints of this kind are also of dynamical interest, since self-dual Higgs transitions between toric-code phases and their relatives have been discussed from that perspective [23 ] , and the algebraic statements here are compatible with such a scenario without establishing it. Finally, the scope deserves a plain statement: the construction concerns specified gapped operations in a finite semisimple unitary defect realization, it does not prove that any microscopic transition is continuous, it does not classify gapped phases with categorical symmetry, and its quantitative statements hold in the connected fusion setting. § 2 sets up marked condensates and specified operations, § 3 develops compatibility, transport, and the central matching description, § 4 proves the complete Higgs criterion together with its transparent-unit specialization, and § 5 collects the examples. Appendix § 10 contains the coordinate classification, the proofs, and the explicit computations for the subgroup-gauging example.
Notation. Subscripts on categories denote modules (C B \mathcal{C}_B C B , Z L \mathcal{Z}_L Z L ); subscripts on algebras and operations denote the phase to which they are transported (B K B_K B K , A K A_K A K , K B K_B K B ); a subscript K K K on a number denotes an invariant of the operation (δ K \delta_K δ K , κ K \kappa_K κ K ). We write 2 V e c 2\mathrm{Vec} 2 Vec for the 2-category of finite semisimple C \mathbb{C} C -linear categories, so that 2 V e c = Σ V e c 2\mathrm{Vec}=\Sigma\mathrm{Vec} 2 Vec = Σ Vec . The superscript r e v \mathrm{rev} rev reverses the braiding and the superscript m o p \mathrm{mop} mop reverses the tensor product. The symbol [ C ] [\mathcal{C}] [ C ] denotes the Witt class of C \mathcal{C} C ; the condition [ C ] = 0 [\mathcal{C}]=0 [ C ] = 0 is the Witt condition. We omit the subscript K K K in ⊗ K \otimes_K ⊗ K when no confusion can arise. The notation Mod ( C ) \operatorname{Mod}(\mathcal{C}) Mod ( C ) denotes the 2-category of finite semisimple C \mathcal{C} C -module categories, while C B \mathcal{C}_B C B denotes the category of right B B B -modules in C \mathcal{C} C . A superscript R R R denotes a right adjoint or a right mate.
Table 1.
Table 2.
Marked condensates and specified separable operations
Condensates, local vacua, and marked interfaces
The examples of § 5 will show that condensations may have several intrinsic vacuum sectors together with charged junction data, and that both must be retained to describe their interaction with a specified operation K K K . The present section first establishes the categorical description of these data, without assuming that the symmetry originates from an ordinary group.
We begin with the standard description of the anyon sector of a bosonic ( 2 + 1 ) (2+1) ( 2 + 1 ) -dimensional topological order by a unitary modular tensor category C \mathcal{C} C [24 ] . Its simple objects label anyon types, its tensor product describes fusion, and its braiding describes exchange. Morphisms describe local operators at junctions of anyon worldlines. Throughout, we use C \mathcal{C} C to denote the anyon category corresponding to the topological order of the initial phase. All categories below are finite semisimple over C \mathbb{C} C , and tensor and defect realizations are taken to be unitary. We write c X , Y c_{X,Y} c X , Y for the braiding, θ X \theta_X θ X for the twist, and X ∨ X^\vee X ∨ for a chosen rigid dual. The symbol ⊠ \boxtimes ⊠ denotes the Deligne tensor product. Our phase comparisons concern the intrinsic defect theories, with the invertible gravitational sector held fixed. For the fusion-category and central-matching statements we assume that the specified separable algebra K K K is connected, meaning that its unit η K \eta_K η K is a simple 1-morphism. We do not impose strong connectedness. The latter appears only as the transparent-unit specialization of § 4.3 . The composite-network and relative-module constructions are algebraic; their applicability beyond the fusion setting is distinguished from the quantitative statements proved here. We write Vec \operatorname{Vec} Vec for the (tensor) category whose objects are finite-dimensional complex vector spaces and whose morphisms are linear maps.
In this description, an anyon condensate is specified by a commutative separable algebra B ∈ C B \in\mathcal{C} B ∈ C with θ B = id B \theta_B = \operatorname{id}_B θ B = id B . Its multiplication m B : B ⊗ B → B m_B : B \otimes B \to B m B : B ⊗ B → B specifies how condensate lines join, and its unit u B : 1 → B u_B : \mathbb{1} \to B u B : 1 → B inserts a condensate line from the vacuum. Separability means that m B m_B m B admits a B B B -bimodule section. In the following we choose by default a normalized dagger-Frobenius representative, with
m B m B † = id B , Δ B = m B † , ϵ B = u B † . m_Bm_B^\dagger= \operatorname{id}_B,\qquad\Delta_B = m_B^\dagger,\qquad\epsilon_B = u_B^\dagger. m B m B † = id B , Δ B = m B † , ϵ B = u B † .
Such representatives exist for separable algebras in the unitary finite semisimple setting [25 ] . (Equivalently, for any given condensation algebra B ∈ C B \in\mathcal{C} B ∈ C , there always exists a “unitarization” isomorphism such that the transformed multiplication and unit satisfy the above conditions.) This coisometric normalization is fixed before the other data are specified and is not an additional parameter. It does not also impose u B † u B = id 1 u_B^\dagger u_B = \operatorname{id}_{\mathbb{1}} u B † u B = id 1 . We retain the chosen unitary separable module-object realizations, free interfaces, adjunctions, and separability splittings. Comparisons preserve these markings and use unitary coherence maps [26 , 27 ] .
Throughout this paper, condensable algebras are assumed nonzero.
A right B B B -module is an object equipped with an action ρ M : M ⊗ B → M \rho_M : M \otimes B \to M ρ M : M ⊗ B → M , allowing condensate lines to end on it. We write C B \mathcal{C}_B C B for their category. For a commutative algebra D D D in a braided category M \mathcal{M} M , we use the left action
D ⊗ M → c M , D − 1 M ⊗ D → ρ M M D \otimes M \xrightarrow{c_{M,D}^{-1}} M \otimes D \xrightarrow{\rho_M} M D ⊗ M c M , D − 1 M ⊗ D ρ M M
to define the tensor product of right D D D -modules. Thus the free-module functor is
F D : M ⟶ M D , X ⟼ X ⊗ D , F_D : \mathcal{M} \longrightarrow\mathcal{M}_D,\qquad X \longmapsto X \otimes D, F D : M ⟶ M D , X ⟼ X ⊗ D ,
with the corresponding positive central lift [6 Secs. 3.3–3.4] . We use this convention throughout.
A module is local when its action is unchanged by the monodromy with the condensate:
ρ M c B , M c M , B = ρ M . \rho_M c_{B,M}c_{M,B} = \rho_M. ρ M c B , M c M , B = ρ M .
We denote the full category of local modules by C B l o c \mathcal{C}_B^{\mathrm{loc}} C B loc . For connected B B B , meaning Hom C ( 1 , B ) ≅ C \operatorname{Hom}_{\mathcal{C}}(\mathbb{1},B) \cong\mathbb{C} Hom C ( 1 , B ) ≅ C , this is the modular category of deconfined anyons after condensation, whereas C B \mathcal{C}_B C B describes the condensation-wall excitations [2 , 6 ] .
We also allow disconnected condensates:
B = ⨁ i = 1 r B i , V B = Hom C ( 1 , B ) ≅ C r , (3) B = \bigoplus_{i=1}^{r} B_i,\qquad\mathcal{V}_B = \operatorname{Hom}_{\mathcal{C}}(\mathbb{1},B) \cong\mathbb{C}^{r}, \tag*{(3)} B = i = 1 ⨁ r B i , V B = Hom C ( 1 , B ) ≅ C r , (3)
where each B i B_i B i is connected and multiplication between distinct summands vanishes. The multiplication of V B \mathcal{V}_B V B is v ⋅ w = m B ( v ⊗ w ) v \cdot w = m_B(v \otimes w) v ⋅ w = m B ( v ⊗ w ) , with unit u B u_B u B . Its primitive idempotents are the projectors onto the individual local vacua.
The module and local-module categories decompose as
C B ≃ ⨁ i C B i , C B l o c ≃ ⨁ i C B i l o c . (4) \mathcal{C}_B \simeq\bigoplus_i \mathcal{C}_{B_i}, \qquad\mathcal{C}_B^{\mathrm{loc}} \simeq\bigoplus_i \mathcal{C}_{B_i}^{\mathrm{loc}}. \tag*{(4)} C B ≃ i ⨁ C B i , C B loc ≃ i ⨁ C B i loc . (4)
Their tensor units are the regular module B B B , decomposing into the units B i B_i B i . Evaluation on u B u_B u B gives
End C B l o c ( B ) ≃ Hom C ( 1 , B ) ≃ C r (5) \operatorname{End}_{\mathcal{C}_B^{\mathrm{loc}}}(B) \simeq\operatorname{Hom}_{\mathcal{C}}(1,B) \simeq\mathbb{C}^{r} \tag*{(5)} End C B loc ( B ) ≃ Hom C ( 1 , B ) ≃ C r (5)
as algebras. Thus each connected component contributes one local vacuum. The outer direct sum records alternative vacua, not stacked topological orders.
For a connected condensable algebra D D D in a modular category M \mathcal{M} M ,
FPdim ( M D l o c ) = FPdim ( M ) FPdim ( D ) 2 . (6) \operatorname{FPdim}(\mathcal{M}^{\mathrm{loc}}_D) = \frac{\operatorname{FPdim}(\mathcal{M})}{\operatorname{FPdim}(D)^2}. \tag*{(6)} FPdim ( M D loc ) = FPdim ( D ) 2 FPdim ( M ) . (6)
This follows from [6 Corollary 3.32] . Such a D D D is Lagrangian when FPdim ( D ) 2 = FPdim ( M ) \operatorname{FPdim}(D)^2 = \operatorname{FPdim}(\mathcal{M}) FPdim ( D ) 2 = FPdim ( M ) , equivalently when M D l o c ≃ V e c \mathcal{M}^{\mathrm{loc}}_D \simeq\mathrm{Vec} M D loc ≃ Vec . For disconnected condensates, the formula is applied separately to the connected components.
The vacuum sectors are locally distinguishable and must not be confused with topology-dependent ground-state degeneracy. For example, for the toric code with electric charge e e e , write b = 1 ⊕ e b = 1 \oplus e b = 1 ⊕ e for the toric-code electric condensate; then B = 1 ⊕ b B = 1 \oplus b B = 1 ⊕ b gives the two-vacuum mixed example of § 5.1 , and B = b ⊕ 3 B = b^{\oplus3} B = b ⊕ 3 gives the three-vacuum comparison at fixed parent condensate in § 5.4 . Throughout, “condensable algebra” allows this possibly disconnected form.
To formulate compatibility with a codimension-one topological defect K K K , we first recall the categorical description of surface defects and their junctions [4 , 28 , 29 ] . To include them, we first introduce the delooping B C \mathbf{B}\mathcal{C} B C of C \mathcal{C} C : the one-object 2-category whose endomorphism category is C \mathcal{C} C , with composition supplied by anyon fusion. Its single object represents the transparent surface defect. We define the suspension of C \mathcal{C} C by its condensation completion:
C = Σ C : = Kar ( B C ) ≃ Mod ( C ) , \mathcal{C} = \Sigma\mathcal{C} := \operatorname{Kar}(\mathbf{B}\mathcal{C}) \simeq\operatorname{Mod}(\mathcal{C}), C = Σ C := Kar ( B C ) ≃ Mod ( C ) ,
Ω C : = End C ( 1 C ) ≃ C . (7) \Omega\mathcal{C} := \operatorname{End}_{\mathcal{C}}(1_{\mathcal{C}}) \simeq\mathcal{C}. \tag*{(7)} Ω C := End C ( 1 C ) ≃ C . (7)
Here Kar \operatorname{Kar} Kar denotes the finite semisimple completion that adjoins splittings of separable monads. In the one-object presentation, such monads are separable algebras in C \mathcal{C} C (not necessarily commutative). Physically, for an anomaly-free topological order described by the anyon category C \mathcal{C} C , the suspension Σ C \Sigma\mathcal{C} Σ C contains all its possible codimension-one surface defects as objects, together with all possible junctions between them as 1-morphisms [28 , 29 ] . An anyon c ∈ C c \in\mathcal{C} c ∈ C is therefore also a 1-endomorphism of the transparent surface defect, as expressed by Ω C ≃ C \Omega\mathcal{C} \simeq\mathcal{C} Ω C ≃ C . Figure 1 illustrates this change of categorical level. This does not involve extra spatial dimensions or a change of the bulk phase.
Figure 1. Fixed-time spatial slices. (a) An anyon c c c is a point excitation on the transparent defect, indicated by the dashed line. (b) Objects X , Y X,Y X , Y of Σ C \Sigma\mathcal{C} Σ C are spatial line defects, joined at f f f . Restoring time turns points into worldlines and spatial lines into surfaces. Defects extend to the boundaries of the displayed patches; these boundaries merely delimit the drawing [24 ] .
The module-category realization C ≃ Mod ( C ) \mathcal{C} \simeq\operatorname{Mod}(\mathcal{C}) C ≃ Mod ( C ) makes this construction concrete1 : its objects are finite semisimple C \mathcal{C} C -module categories, its 1-morphisms are module functors, and its 2-morphisms are module natural transformations. The transparent surface defect 1 C 1_{\mathcal{C}} 1 C is represented by the regular module category C \mathcal{C} C . Note also that the braiding of C \mathcal{C} C induces a right C \mathcal{C} C -module structure on any left C \mathcal{C} C -module category, and conversely; that is, every C \mathcal{C} C -module category is naturally equipped with a canonical C \mathcal{C} C -bimodule structure (which is also why we need not distinguish left from right in the notation Mod ( C ) \operatorname{Mod}(\mathcal{C}) Mod ( C ) ), so that for any two C \mathcal{C} C -modules one can define their balanced product ⊠ C : Mod ( C ) × Mod ( C ) → Mod ( C ) \boxtimes_{\mathcal{C}}:\operatorname{Mod}(\mathcal{C})\times\operatorname{Mod}(\mathcal{C})\to\operatorname{Mod}(\mathcal{C}) ⊠ C : Mod ( C ) × Mod ( C ) → Mod ( C ) . This gives the monoidal structure on C ≃ Mod ( C ) \mathcal{C}\simeq\operatorname{Mod}(\mathcal{C}) C ≃ Mod ( C ) . Concretely, this equivalence can be understood through the following picture:
For a surface defect X X X , the category Hom C ( 1 C , X ) \operatorname{Hom}_{\mathcal{C}}(1_{\mathcal{C}},X) Hom C ( 1 C , X ) describes its boundary conditions on the trivial defect. Attaching anyons of the bulk C \mathcal{C} C gives it a C \mathcal{C} C -module structure, and dragging a junction f : X → Y f:X\to Y f : X → Y onto the boundary gives a module functor between the corresponding categories. Conversely, a finite semisimple C \mathcal{C} C -module category has a presentation by modules over a separable algebra in C \mathcal{C} C , and hence defines an object of the chosen condensation completion [30 Corollary 7.10.5] . We therefore identify X X X with its module-category presentation when no confusion arises. The defect object X X X must nevertheless be distinguished from End C ( X ) \operatorname{End}_{\mathcal{C}}(X) End C ( X ) , the category of line defects supported on it [24 , 29 ] .
The suspension operation can also be generalized to separable algebras. For the chosen condensate B B B , condensing its network only along a line in space produces a line defect whose worldsheet we denote by A = Σ B A=\Sigma B A = Σ B , while the parent phase remains on both sides; see Figure 2 a. In the module-category model, this surface defect is represented by C B : = Mod C ( B ) \mathcal{C}_B:=\operatorname{Mod}_{\mathcal{C}}(B) C B := Mod C ( B ) as a C \mathcal{C} C -module category. Note that here Σ B \Sigma B Σ B denotes the condensation of the algebra B B B ; it and B B B itself are objects at two different levels. This is neither merely an operation at the anyon level, nor an operation that can be performed on an arbitrary anyon [29 ] .
Figure 2. Spatial pictures of the suspended condensate and its marking. (a) Condensing B B B along a line produces A = Σ B A = \Sigma B A = Σ B ; the surrounding bulk remains unchanged. The algebra and Frobenius junctions in spacetime are suppressed. (b) The specified junctions η A , η A R \eta_A,\eta_A^R η A , η A R connect the A A A segment to the transparent defect. Shrinking this segment recovers B B B , including its algebra structure through the adjunction. No defect has an unmarked endpoint in the bulk [26 , 29 ] .
Again, it is important to distinguish the categories appearing in these descriptions:
Hom C ( 1 C , A ) ≃ C B , End C ( A ) ≃ B C B . (8) \operatorname{Hom}_{\mathcal{C}}(1_{\mathcal{C}},A)\simeq\mathcal{C}_B,\qquad\operatorname{End}_{\mathcal{C}}(A)\simeq{}_B\mathcal{C}_B. \tag*{(8)} Hom C ( 1 C , A ) ≃ C B , End C ( A ) ≃ B C B . (8)
The first describes junctions ending the A A A surface defect, also regarded as the category representation of A A A ; the second, consisting of B B B -B B B bimodules, describes line defects supported within it. Neither should be confused with C B l o c \mathcal{C}^{\mathrm{loc}}_B C B loc , which describes the deconfined anyons after bulk condensation. In particular, the junction description retains modules that need not survive as deconfined bulk excitations.
We can define a special boundary for it, realized by the free-module functor:
η A : C ≃ 1 C ⟶ A ≃ C B , \eta_A:\mathcal{C}\simeq1_{\mathcal{C}}\longrightarrow A\simeq\mathcal{C}_B, η A : C ≃ 1 C ⟶ A ≃ C B ,
η A ( c ) = c ⊠ B . (9) \eta_A(c)=c\boxtimes B. \tag*{(9)} η A ( c ) = c ⊠ B . (9)
Geometrically, η A \eta_A η A is the line along which an A A A surface defect ends on the transparent surface defect, hence a point in a spatial slice. Its right adjoint η A R \eta_A^R η A R as a module functor is represented by the forgetful functor C B → C \mathcal{C}_B \to\mathcal{C} C B → C . Their composite satisfies
( η A R η A ) ( c ) ≃ c ⊗ B , η A R η A ≃ B as separable monads on 1 C . (10) (\eta_A^R\eta_A)(c) \simeq c \otimes B,
\qquad
\eta_A^R\eta_A \simeq B \quad\text{as separable monads on } \mathbbm{1}_{\mathcal{C}}. \tag*{(10)} ( η A R η A ) ( c ) ≃ c ⊗ B , η A R η A ≃ B as separable monads on 1 C . (10)
Definition 2.1 (Marked condensate). A marked condensate is a quadruple ( A , η A , η A ⊣ η A R , η A R η A ≃ B ) (A,\eta_A,\eta_A \dashv\eta_A^R,\eta_A^R\eta_A \simeq B) ( A , η A , η A ⊣ η A R , η A R η A ≃ B ) , consisting of a condensate interface A = Σ B A = \Sigma B A = Σ B , its free-module unit η A \eta_A η A , a chosen adjunction, and the monad identification of (10) . A marked equivalence is an equivalence of the underlying interfaces that preserves η A \eta_A η A , the adjunction, and the monad identification.
The monad identification includes the unit and multiplication of B B B , induced by the adjunction. Physically, this says that we can recover B B B by shrinking an A A A segment with the specified endpoint junctions; in other words, A A A and B B B give equivalent descriptions of the condensation data, as shown in Figure 2 (b) [26 , 29 ] . This is the marking that we keep when passing from an anyon condensate to its surface-defect description. An equivalence E : A ≃ A ′ E : A \simeq A' E : A ≃ A ′ of underlying surface-defect objects need not identify E η A E\eta_A E η A with the specified interface η A ′ \eta_{A'} η A ′ . When comparing condensation data, we therefore require preservation of the marked interface, its adjunction, and the monad identification (10) , rather than merely an equivalence of unmarked surface-defect objects.
The condensation-completed defect theory of the bulk-condensed phase is
Y B : = Bimod C ( A ) , A = Σ B . \mathcal{Y}_B := \operatorname{Bimod}_{\mathcal{C}}(A), \qquad A = \Sigma B. Y B := Bimod C ( A ) , A = Σ B .
Its tensor unit is the regular A A A –A A A bimodule. The local-module description gives
Ω Y B ≃ C B l o c ≃ ⨁ i C B i l o c . (11) \Omega\mathcal{Y}_B \simeq\mathcal{C}_B^{\mathrm{loc}} \simeq\bigoplus_i \mathcal{C}_{B_i}^{\mathrm{loc}}. \tag*{(11)} Ω Y B ≃ C B loc ≃ i ⨁ C B i loc . (11)
The surrounding defect theory also retains the junction data relating the sectors [2 , 29 ] . In particular, the vacuum-resolved anyon category does not by itself specify all the marked interfaces.
Specified separable operations
The preceding construction turns the anyon condensate B B B into the marked defect A = Σ B A=\Sigma B A = Σ B : a spatial line in the parent bulk, or a surface defect after restoring time. We now choose a second specified separable operation K K K in the same parent defect realization. We describe its multiplication, unit, and junction coherences, and the separability needed to promote its network to part of the vacuum. The specified unit need not be transparent; the transparent-unit specialization is discussed in § 4.3 .
Algebra data for a specified separable operation
An ordinary anomaly-free finite-group symmetry provides a useful starting point. Its labelled walls U g U_g U g obey U g ⊗ U h ≃ U g h U_g\otimes U_h\simeq U_{gh} U g ⊗ U h ≃ U g h , with U e ≃ 1 C U_e\simeq\mathbf{1}_{\mathcal C} U e ≃ 1 C . The direct sum ⨁ g ∈ G U g \bigoplus_{g\in G}U_g ⨁ g ∈ G U g , together with the chosen fusion junctions and their coherence data, packages the chosen gauging network into a single algebra object. For a more general specified operation, the constituent defects need not be invertible and their fusion need not be group-like. Nevertheless, we must still specify how the walls join and how the neutral operation is inserted.
The junctions of the chosen network are encoded by a multiplication and a unit,
μ K : K ⊗ K ⟶ K , η K : 1 C ⟶ K . (12) \mu_K:K\otimes K\longrightarrow K,\qquad\eta_K:\mathbf{1}_{\mathcal C}\longrightarrow K. \tag*{(12)} μ K : K ⊗ K ⟶ K , η K : 1 C ⟶ K . (12)
Together with coherent associativity and unit isomorphisms, these are the data of an E 1 E_1 E 1 -algebra in C \mathcal C C . Here E 1 E_1 E 1 means associative and unital; it does not impose commutativity. Specifying how the constituent defects fuse is not yet the same as specifying the gauging operation. To gauge this operation, we must also allow networks of these defects and their junctions to become part of the new vacuum, rather than remain fixed insertions. Their fusion data must therefore support a consistent condensation within the finite semisimple framework used here.
We require the chosen algebra network to admit a separable realization. Besides the coherent fusion operations, this requires bimodule-compatible adjunction and splitting data for the multiplication junction.2 These data ensure that the module and bimodule 2-categories describing the gauging interface and the gauged phase remain finite semisimple [31 ] . Thus the specified separable operation is encoded by
K ∈ Alg E 1 s e p ( C ) , (13) K\in\operatorname{Alg}_{E_1}^{\mathrm{sep}}(\mathcal C), \tag*{(13)} K ∈ Alg E 1 sep ( C ) , (13)
including its multiplication, unit, and junction coherence data.
Recall that C ≃ Σ C ≃ Mod ( C ) \mathcal C\simeq\Sigma\mathcal C\simeq\operatorname{Mod}(\mathcal C) C ≃ Σ C ≃ Mod ( C ) . In this presentation, the underlying object K K K is represented by a finite semisimple C \mathcal C C -module category. We write K \mathcal K K for this module category equipped with the additional monoidal structure induced by the specified algebra structure on K K K ; more explicitly, the underlying module category is
K ≃ Hom C ( 1 C , K ) as C -module categories . (14) \mathcal K\simeq\operatorname{Hom}_{\mathcal C}(\mathbf{1}_{\mathcal C},K)\qquad\text{as }\mathcal C\text{-module categories}. \tag*{(14)} K ≃ Hom C ( 1 C , K ) as C -module categories . (14)
For objects x , y : 1 C → K x,y:\mathbf{1}_{\mathcal C}\to K x , y : 1 C → K in this Hom-category, the multiplication and unit of K K K define a monoidal structure
x ⊗ K y : = μ K ∘ ( x ⊗ y ) , 1 K : = η K , (15) x\mathbin{\otimes_{\mathcal K}}y:=\mu_K\circ(x\otimes y),\qquad\mathbf{1}_{\mathcal K}:=\eta_K, \tag*{(15)} x ⊗ K y := μ K ∘ ( x ⊗ y ) , 1 K := η K , (15)
using the canonical identification 1 C ⊗ 1 C ≃ 1 C \mathbf{1}_{\mathcal C}\otimes\mathbf{1}_{\mathcal C}\simeq\mathbf{1}_{\mathcal C} 1 C ⊗ 1 C ≃ 1 C . The associativity and unit constraints are inherited from those of the algebra K K K . Equivalently, in module coordinates, μ K \mu_K μ K is represented by a C \mathcal C C -balanced multiplication K ⊠ C K → K \mathcal K\boxtimes_{\mathcal C}\mathcal K\to\mathcal K K ⊠ C K → K .
The suspended condensate A = Σ B A=\Sigma B A = Σ B also carries a separable E 1 E_1 E 1 -algebra structure: its multiplication is induced by ⊗ B \otimes_B ⊗ B , and its unit is the marked free-module interface (9) [29 ] .
Compatibility, transport, and marked wall interchange
We now specify what it means for a condensate to be compatible with a specified separable operation, and we show that the full input generates a transported condensate together with a commuting square of marked walls. The formulation is purely algebraic in the composite network and the relative modules; it does not use strong connectedness, and no transparency of the specified unit is assumed.
The gauged phase and its wall
Gauging promotes networks of the specified K K K defects and their junctions to part of the vacuum configuration. A defect in the gauged phase must consequently carry compatible left and right K K K actions. This gives the description
Z = C / / K : = Bimod C ( K ) , Ω Z ≃ Z = End Z ( K ) . (16) \mathcal{Z}=\mathcal{C}//K:=\operatorname{Bimod}_{\mathcal{C}}(K),\qquad\Omega\mathcal{Z}\simeq\mathcal{Z}=\operatorname{End}_{\mathcal{Z}}(K). \tag*{(16)} Z = C // K := Bimod C ( K ) , Ω Z ≃ Z = End Z ( K ) . (16)
The tensor product in Z \mathcal{Z} Z is the relative tensor product over K K K , and its tensor unit is the regular K K K –K K K bimodule. Accordingly, the endomorphisms of this regular bimodule are the anyons of the gauged phase.
Let G K : C → Z G_K:\mathcal{C}\to\mathcal{Z} G K : C → Z denote the canonical gauging wall. Its module description is RMod K ( C ) \operatorname{RMod}_K(\mathcal{C}) RMod K ( C ) together with the distinguished regular right module K K K . The category of excitations on this wall is
End RMod K ( C ) ( K ) ≃ Hom C ( 1 C , K ) ≃ K . (17) \operatorname{End}_{\operatorname{RMod}_K(\mathcal{C})}(K)\simeq\operatorname{Hom}_{\mathcal{C}}(\mathbbm{1}_{\mathcal{C}},K)\simeq\mathcal{K}. \tag*{(17)} End RMod K ( C ) ( K ) ≃ Hom C ( 1 C , K ) ≃ K . (17)
The first equivalence stems from the adjunction between the free module functor and the forgetful functor. Indeed, a right K K K -module endomorphism F : K → K F:K\to K F : K → K is determined by its restriction F ∘ η K : 1 C → K F\circ\eta_K:\mathbbm{1}_{\mathcal{C}}\to K F ∘ η K : 1 C → K . Conversely, a junction x : 1 C → K x:\mathbbm{1}_{\mathcal{C}}\to K x : 1 C → K extends to the right-module endomorphism μ K ∘ ( x ⊗ id K ) \mu_K\circ(x\otimes\operatorname{id}_K) μ K ∘ ( x ⊗ id K ) . Under this correspondence, composition of module endomorphisms becomes exactly the tensor product (2.13). This identification explains why the same K K K appears both in the module presentation of K K K and as the fusion category of wall excitations [29 ] .
Thus the bulk–boundary relation of an anomaly-free topological order gives
Z r e v ≃ Z ^ = j C ( C ) Z ( K ) ′ , Z ( K ) ≃ C ⊠ Z ^ r e v . (18) \mathcal{Z}^{\mathrm{rev}}\simeq\widehat{\mathcal{Z}}=\mathcal{j}_{\mathcal{C}}(\mathcal{C})'_{\mathcal{Z}(K)},\qquad\mathcal{Z}(K)\simeq\mathcal{C}\boxtimes\widehat{\mathcal{Z}}^{\mathrm{rev}}. \tag*{(18)} Z rev ≃ Z = j C ( C ) Z ( K ) ′ , Z ( K ) ≃ C ⊠ Z rev . (18)
We fix the functors used throughout. Let j C : C → Z ( K ) j_{\mathcal{C}}:\mathcal{C}\to\mathcal{Z}(K) j C : C → Z ( K ) be the braided lift of the parent anyons to the center of the wall category, and let
ι = U j C : C ⟶ K (19) \iota=Uj_{\mathcal{C}}:\mathcal{C}\longrightarrow\mathcal{K} \tag*{(19)} ι = U j C : C ⟶ K (19)
be its underlying central tensor functor, where U U U is the forgetful functor of the center. Likewise let j Z : Z ^ ↪ Z ( K ) j_{\mathcal{Z}}:\widehat{\mathcal{Z}}\hookrightarrow\mathcal{Z}(K) j Z : Z ↪ Z ( K ) be the braided lift of the gauged phase, which is the inclusion of the full Müger centralizer of j C ( C ) j_{\mathcal{C}}(\mathcal{C}) j C ( C ) , and let q = U j Z q=Uj_{\mathcal{Z}} q = U j Z be its underlying functor. The center then decomposes as in (18) , and we use the resulting identification Z r e v ≃ Z ^ \mathcal{Z}^{\mathrm{rev}}\simeq\widehat{\mathcal{Z}} Z rev ≃ Z throughout. The functor j Z j_{\mathcal{Z}} j Z is fully faithful by definition, while neither ι \iota ι nor q q q is assumed to be fully faithful or dominant.
Here j C ( C ) j_{\mathcal{C}}(\mathcal{C}) j C ( C ) is the image of the braided tensor functor introduced in (19) . The prime denotes the Müger centralizer: the objects whose double braiding with every object of j C ( C ) j_{\mathcal{C}}(\mathcal{C}) j C ( C ) is trivial. The reversed braiding records the opposite orientations of the two bulks adjacent to the wall.
With this orientation convention, q q q describes bringing an anyon from the gauged bulk onto the wall. We also write q ( L ) q(L) q ( L ) for L ∈ Z ^ L\in\widehat{\mathcal{Z}} L ∈ Z , since reversing the braiding does not change the underlying tensor category.
The two braided lifts supply the half-braidings used below:
β C , X : ι ( c ) ⊗ X → X ⊗ ι ( c ) , γ L , X : q ( L ) ⊗ X → X ⊗ q ( L ) , (20) \beta_{\mathcal{C},X}:\iota(c)\otimes X\to X\otimes\iota(c),\qquad\gamma_{L,X}:q(L)\otimes X\to X\otimes q(L), \tag*{(20)} β C , X : ι ( c ) ⊗ X → X ⊗ ι ( c ) , γ L , X : q ( L ) ⊗ X → X ⊗ q ( L ) , (20)
for c ∈ C c\in\mathcal{C} c ∈ C , L ∈ Z ^ L\in\widehat{\mathcal{Z}} L ∈ Z , and X ∈ K X\in\mathcal{K} X ∈ K . Here β \beta β is the half-braiding of j C ( c ) j_{\mathcal{C}}(c) j C ( c ) and γ \gamma γ that of $j_{\mathcal{Z}}(L).
The full pull-through condition
Intuitively, a condensate compatible with K K K must pass reversibly through the specified K K K defect network. Regarding B B B as a 1-endomorphism of the tensor unit, its crossing can be written as a unitary 2-isomorphism
τ : B ⊗ id K ⇒ ≃ id K ⊗ B . (21) \tau: B \otimes\operatorname{id}_{K} \xRightarrow{\simeq} \operatorname{id}_{K} \otimes B. \tag*{(21)} τ : B ⊗ id K ≃ id K ⊗ B . (21)
Merely requiring it to be an isomorphism between isolated line configurations is not sufficient: the crossing must respect the multiplication and unit of both K K K and B B B .
To gain a more computable description, we can turn this condition into a module representation criterion. Set
B ~ = ι ( B ) , m B ~ = ι ( m B ) , u B ~ = ι ( u B ) . \widetilde{B}=\iota(B), \qquad m_{\widetilde{B}}=\iota(m_{B}), \qquad u_{\widetilde{B}}=\iota(u_{B}). B = ι ( B ) , m B = ι ( m B ) , u B = ι ( u B ) .
Here ι \iota ι is the central functor defined in (19) . In the module presentation of K K K , the two sides of (21) are represented by the C \mathcal{C} C -module endofunctors of K \mathcal{K} K
L B ~ ( X ) = B ~ ⊗ K X , R B ~ ( X ) = X ⊗ K B ~ , X ∈ K . L_{\widetilde{B}}(X)=\widetilde{B}\otimes_{\mathcal{K}}X,\qquad R_{\widetilde{B}}(X)=X\otimes_{\mathcal{K}}\widetilde{B},\qquad X\in\mathcal{K}. L B ( X ) = B ⊗ K X , R B ( X ) = X ⊗ K B , X ∈ K .
Their module structures are naturally inherited from the ambient C \mathcal{C} C -action. The reversible crossing of B B B can then equivalently be written as a natural isomorphism between these two.
The half-braiding β \beta β was introduced in (20) .
Definition 3.1 (Compatible condensation). A K K K -compatible condensation consists of a condensable algebra B B B together with a module natural isomorphism τ : L B ~ ⇒ ≃ R B ~ \tau:L_{\widetilde{B}}\xRightarrow{\simeq}R_{\widetilde{B}} τ : L B ≃ R B . At the level of objects, this is equivalent to specifying a family of unitary isomorphisms
τ X : B ~ ⊗ K X → ≃ X ⊗ K B ~ , X ∈ K , (22) \tau_X:\widetilde{B}\otimes_{\mathcal{K}}X\xrightarrow{\simeq}X\otimes_{\mathcal{K}}\widetilde{B},\qquad X\in\mathcal{K}, \tag*{(22)} τ X : B ⊗ K X ≃ X ⊗ K B , X ∈ K , (22) satisfying
τ X ⊗ K Y = ( id X ⊗ K τ Y ) ( τ X ⊗ K id Y ) , (23) \tau_{X\otimes_{\mathcal{K}}Y}=(\operatorname{id}_X\otimes_{\mathcal{K}}\tau_Y)(\tau_X\otimes_{\mathcal{K}}\operatorname{id}_Y), \tag*{(23)} τ X ⊗ K Y = ( id X ⊗ K τ Y ) ( τ X ⊗ K id Y ) , (23) τ 1 K = id B ~ , (24) \tau_{1_{\mathcal{K}}}=\operatorname{id}_{\widetilde{B}}, \tag*{(24)} τ 1 K = id B , (24) τ X ( m B ~ ⊗ K id X ) = ( id X ⊗ K m B ~ ) ( τ X ⊗ K id B ~ ) ( id B ~ ⊗ K τ X ) , (25) \tau_X(m_{\widetilde{B}}\otimes_{\mathcal{K}}\operatorname{id}_X)=(\operatorname{id}_X\otimes_{\mathcal{K}}m_{\widetilde{B}})(\tau_X\otimes_{\mathcal{K}}\operatorname{id}_{\widetilde{B}})(\operatorname{id}_{\widetilde{B}}\otimes_{\mathcal{K}}\tau_X), \tag*{(25)} τ X ( m B ⊗ K id X ) = ( id X ⊗ K m B ) ( τ X ⊗ K id B ) ( id B ⊗ K τ X ) , (25) τ X ( u B ~ ⊗ K id X ) = id X ⊗ K u B ~ , (26) \tau_X(u_{\widetilde{B}}\otimes_{\mathcal{K}}\operatorname{id}_X)=\operatorname{id}_X\otimes_{\mathcal{K}}u_{\widetilde{B}}, \tag*{(26)} τ X ( u B ⊗ K id X ) = id X ⊗ K u B , (26) τ ι ( c ) β c , B ~ = id ι ( c ) ⊗ K B ~ , c ∈ C . (27) \tau_{\iota(c)}\beta_{c,\widetilde{B}}=\operatorname{id}_{\iota(c)\otimes_{\mathcal{K}}\widetilde{B}},\qquad c\in\mathcal{C}. \tag*{(27)} τ ι ( c ) β c , B = id ι ( c ) ⊗ K B , c ∈ C . (27) An isomorphism ( B , τ ) → ( B ′ , τ ′ ) (B,\tau)\to(B',\tau') ( B , τ ) → ( B ′ , τ ′ ) is a unitary algebra isomorphism f : B → B ′ f:B\to B' f : B → B ′ such that
( id X ⊗ K ι ( f ) ) τ X = τ X ′ ( ι ( f ) ⊗ K id X ) , ∀ X ∈ K . (\operatorname{id}_X\otimes_{\mathcal{K}}\iota(f))\tau_X=\tau'_X(\iota(f)\otimes_{\mathcal{K}}\operatorname{id}_X),\qquad\forall X\in\mathcal{K}. ( id X ⊗ K ι ( f )) τ X = τ X ′ ( ι ( f ) ⊗ K id X ) , ∀ X ∈ K . The associated marked surface-defect data are compared using the induced identifications of their free-module interfaces and adjunctions.
The tensor identities express compatibility with the multiplication and unit of K K K . The next two identities preserve the algebra structure of the condensate. Together with the tensor identities, the last identity is the required C \mathcal{C} C -module naturality. Compatibility with the specified K K K -unit is encoded through the entire C \mathcal{C} C -module functor ι : C → K \iota:\mathcal{C}\to\mathcal{K} ι : C → K , not merely through its value ι ( 1 C ) = 1 K \iota(1_{\mathcal{C}})=1_{\mathcal{K}} ι ( 1 C ) = 1 K .
Evidently, an arbitrary half-braiding on ι ( B ) \iota(B) ι ( B ) does not necessarily define a compatible condensation, since the ambient action at least fixes
τ ι ( c ) = β c , B ~ − 1 = ι ( c c , B − 1 ) . (28) \tau_{\iota(c)}=\beta^{-1}_{c,\widetilde{B}}=\iota(c^{-1}_{c,B}). \tag*{(28)} τ ι ( c ) = β c , B − 1 = ι ( c c , B − 1 ) . (28)
Figure 3. Two descriptions of compatible pull-through. (a) The condensate line B B B passes through K K K by the 2-isomorphism τ : B ⊗ id K ⇒ id K ⊗ B \tau: B \otimes\operatorname{id}_{K} \Rightarrow\operatorname{id}_{K} \otimes B τ : B ⊗ id K ⇒ id K ⊗ B , shown as a reversible change between junction configurations. (b) The marked suspension A = Σ B A = \Sigma B A = Σ B crosses K K K at the 1-equivalence junction λ : K ⊗ A → A ⊗ K \lambda: K \otimes A \to A \otimes K λ : K ⊗ A → A ⊗ K . The right diagram is read from bottom to top; its central point is a junction, not an over/under crossing. The teal point represents B B B , and solid purple and teal lines represent K K K and A A A . Dashed lines denote the transparent surface.
The fixed ambient action determines the crossing on objects in the image of ι \iota ι . Its extension to the remaining wall excitations is part of the compatibility datum. For a general operation, ι \iota ι may itself be dominant; neither dominance nor full faithfulness is assumed here.
For a compatible condensation there is also an equivalent surface-defect formulation. Let A = Σ B A = \Sigma B A = Σ B be the chosen separable module object, retaining its suspension multiplication, free interface, and adjunction η A ⊣ η A R \eta_A \dashv\eta_A^R η A ⊣ η A R . The marking includes
η A R η A ≃ B as separable monads. (29) \eta_A^R \eta_A \simeq B \qquad\text{as separable monads.} \tag*{(29)} η A R η A ≃ B as separable monads. (29)
For this realization, the line crossing is equivalent to a unitary marked pseudodistributive law
λ : K ⊗ A → ≃ A ⊗ K . (30) \lambda: K \otimes A \xrightarrow{\simeq} A \otimes K. \tag*{(30)} λ : K ⊗ A ≃ A ⊗ K . (30)
Geometrically, λ \lambda λ is a reversible junction identifying the two orderings of the specified K K K - and A A A -surfaces. Its accompanying coherence maps compare their algebra junctions and marked interfaces. Of course, one must also specify several coherent 2-morphisms concerning the unit and multiplication to obtain the complete data of a pseudodistributive law, such as the specified unit comparison
α λ : λ ( id K ⊗ η A ) ⇒ ≃ η A ⊗ id K . (31) \alpha_\lambda: \lambda(\operatorname{id}_{K} \otimes\eta_A) \xRightarrow{\simeq} \eta_A \otimes\operatorname{id}_{K}. \tag*{(31)} α λ : λ ( id K ⊗ η A ) ≃ η A ⊗ id K . (31)
It specifies how the unit of A A A passes reversibly through the K K K surface. Appendix § 7 .§ 1 gives the complete coherence data of the surface-defect pseudodistributive law and proves that it is equivalent to the line-crossing data. We consequently use ( B , τ ) (B, \tau) ( B , τ ) or ( B , λ ) (B, \lambda) ( B , λ ) for the corresponding marked data, these three being understood as equivalent formulations. The pseudodistributive-law conventions follow Refs. [32 , 33 , 34 ] .
Central matching
The compatibility datum of Definition 3.1 is a coherent reversible crossing of the whole parent condensate through the specified operation. In this section we show that, in the connected fusion setting, such a datum is equivalent to a condensable algebra L L L in the gauged phase together with an algebra identification ξ : q ( L ) → ∼ ι ( B ) \xi:q(L)\xrightarrow{\sim}\iota(B) ξ : q ( L ) ∼ ι ( B ) in the wall category; this is the central matching description of compatibility, and it retains both the transported condensate and the specified interface.
Throughout this section the specified separable algebra K \mathcal{K} K is assumed to be connected, so that its unit η K \eta_{\mathcal{K}} η K is a simple 1-morphism and 1 K = η K \mathbf{1}_{\mathcal{K}}=\eta_{\mathcal{K}} 1 K = η K is a simple object. Under the unitary assumptions, separability supplies rigidity, and K \mathcal{K} K is a unitary fusion category. We do not assume strong connectedness, and we do not assume that K \mathcal{K} K is invertible. The transparent-unit specialization is treated separately later in the paper.
Compatibility can be expressed entirely through condensable algebras on the two sides and one algebra identification in the wall. This is the general form of inverse transport; it retains both the transported condensate and the identification.
Theorem 3.2 (Central matching). Fix K \mathcal{K} K and its central presentation (19) . The groupoid of compatible condensation data ( B , τ ) (B,\tau) ( B , τ ) of Definition 3.1 is equivalent to the groupoid whose objects are the triples ( B , L , ξ ) (B,L,\xi) ( B , L , ξ ) consisting of
(i) a condensable algebra B ∈ C B\in C B ∈ C ;
(ii) a condensable algebra L ∈ Z ^ L\in\widehat{\mathcal{Z}} L ∈ Z , regarded in Z ^ = Z r e v \widehat{\mathcal{Z}}=\mathcal{Z}^{\mathrm{rev}} Z = Z rev through the canonical identification of underlying tensor categories;
(iii) a unitary algebra isomorphism
ξ : q ( L ) → ≅ ι ( B ) in K . \xi:q(L)\xrightarrow{\cong}\iota(B)\qquad\text{in }\mathcal{K}. ξ : q ( L ) ≅ ι ( B ) in K . The corresponding crossing is reconstructed by conjugation with ξ \xi ξ :
τ X = ( id X ⊗ ξ ) γ L , X ( ξ − 1 ⊗ id X ) , X ∈ K , (32) \tau_X=(\operatorname{id}_X\otimes\xi)\gamma_{L,X}(\xi^{-1}\otimes\operatorname{id}_X),\qquad X\in\mathcal{K}, \tag*{(32)} τ X = ( id X ⊗ ξ ) γ L , X ( ξ − 1 ⊗ id X ) , X ∈ K , (32) where the tensor product is that of K \mathcal{K} K and γ L , ∙ \gamma_{L,\bullet} γ L , ∙ is the central half-braiding of q ( L ) q(L) q ( L ) supplied by the central functor q q q . An isomorphism ( B , L , ξ ) → ( B ′ , L ′ , ξ ′ ) (B,L,\xi)\to(B',L',\xi') ( B , L , ξ ) → ( B ′ , L ′ , ξ ′ ) is a pair of unitary algebra isomorphisms f : B → B ′ f:B\to B' f : B → B ′ , g : L → L ′ g:L\to L' g : L → L ′ satisfying
ξ ′ q ( g ) = ι ( f ) ξ . (33) \xi' q(g)=\iota(f)\xi. \tag*{(33)} ξ ′ q ( g ) = ι ( f ) ξ . (33) The equivalence is a restatement of (32) and (33) : it does not determine L L L from C C C , K \mathcal{K} K and B B B alone.
Proof of Theorem 3.2. Objects. Let ( B , τ ) (B,\tau) ( B , τ ) be compatible. By Lemma A.2 the pair ( ι ( B ) , τ ) (\iota(B),\tau) ( ι ( B ) , τ ) is a condensable central algebra in K \mathcal{K} K , and its central structure makes it an algebra in the full centralizer Z ^ \widehat{\mathcal{Z}} Z ; choose a presentation L L L of that algebra together with the underlying algebra identification ξ : q ( L ) → ∼ ι ( B ) \xi:q(L)\xrightarrow{\sim}\iota(B) ξ : q ( L ) ∼ ι ( B ) . This gives ( B , τ ) ↦ ( B , L , ξ ) (B,\tau)\mapsto(B,L,\xi) ( B , τ ) ↦ ( B , L , ξ ) , and Theorem 3.3 identifies L L L with B K B_{\mathcal{K}} B K a posteriori.
Crossings. Conversely, let ( B , L , ξ ) (B,L,\xi) ( B , L , ξ ) be a triple with L L L condensable in Z ^ \widehat{\mathcal{Z}} Z and ξ \xi ξ an algebra isomorphism. Transporting the central half-braiding γ L , ∙ \gamma_{L,\bullet} γ L , ∙ of j Z ( L ) j_{\mathcal{Z}}(L) j Z ( L ) along ξ \xi ξ gives
τ X = ( id X ⊗ ξ ) γ L , X ( ξ − 1 ⊗ id X ) , X ∈ K , (34) \tau_X=(\operatorname{id}_X\otimes\xi)\gamma_{L,X}(\xi^{-1}\otimes\operatorname{id}_X),\qquad X\in\mathcal{K}, \tag*{(34)} τ X = ( id X ⊗ ξ ) γ L , X ( ξ − 1 ⊗ id X ) , X ∈ K , (34)
and centrality of the multiplication and unit of L L L gives (25) and (26) , while the half-braiding axioms give (23) and (24) ; the relative condition (27) follows from γ L , ι ( c ) = β C , ι ( B ) − 1 \gamma_{L,\iota(c)}=\beta^{-1}_{\mathcal{C},\iota(B)} γ L , ι ( c ) = β C , ι ( B ) − 1 . Thus τ \tau τ is a compatible crossing, and the two constructions are inverse.
Morphism. Let f : B → B ′ f:B\to B' f : B → B ′ be an algebra isomorphism intertwining τ \tau τ with τ ′ \tau' τ ′ . Then h : = ( ξ ′ ) − 1 ι ( f ) ξ : q ( L ) → q ( L ′ ) h:=(\xi')^{-1}\iota(f)\xi:q(L)\to q(L') h := ( ξ ′ ) − 1 ι ( f ) ξ : q ( L ) → q ( L ′ ) is a morphism of central objects,
( id X ⊗ h ) γ L , X = γ L ′ , X ( h ⊗ id X ) , (35) (\operatorname{id}_X\otimes h)\gamma_{L,X}=\gamma_{L',X}(h\otimes\operatorname{id}_X), \tag*{(35)} ( id X ⊗ h ) γ L , X = γ L ′ , X ( h ⊗ id X ) , (35)
by the intertwining identity, so the fullness and faithfulness of j Z j_{\mathcal Z} j Z provide a unique g : L → L ′ g:L\to L' g : L → L ′ with q ( g ) = h q(g)=h q ( g ) = h ,
q ( g ) = ( ξ ′ ) − 1 ι ( f ) ξ , equivalently ξ ′ q ( g ) = ι ( f ) ξ , (36) q(g)=(\xi')^{-1}\iota(f)\xi,\qquad\text{equivalently}\qquad\xi' q(g)=\iota(f)\xi, \tag*{(36)} q ( g ) = ( ξ ′ ) − 1 ι ( f ) ξ , equivalently ξ ′ q ( g ) = ι ( f ) ξ , (36)
and g g g is a unitary algebra isomorphism because j Z j_{\mathcal Z} j Z is a fully faithful unitary braided lift; centrality of g g g then gives the transport of the crossing, τ X ( ι ( f ) ⊗ id X ) = ( id X ⊗ ι ( f ) ) τ X \tau_X(\iota(f)\otimes\operatorname{id}_X)=(\operatorname{id}_X\otimes\iota(f))\tau_X τ X ( ι ( f ) ⊗ id X ) = ( id X ⊗ ι ( f )) τ X , which is (33) lifted to the wall. Conversely such a g g g makes f f f intertwine the crossings, so the two constructions are inverse on morphisms, functorial, and preserve the free interfaces, the chosen adjunctions and the monad markings; with the convention of Definition 3.1 this is the stated equivalence of groupoids, where a morphism with f = id B f=\operatorname{id}_B f = id B is precisely a morphism of the marked fibre.
Prescribed parent condensate. In applications one usually fixes the parent algebra B B B and asks which compatible crossings it admits. The matching statement is then used with B B B prescribed as in Proposition 4.4 : the triple retains ξ \xi ξ , and its isomorphisms are the pairs ( f , g ) (f,g) ( f , g ) of (33) with f f f an algebra automorphism of B B B . Forgetting ξ \xi ξ discards the specified algebra matching on the wall and, in general, yields coarser data than the marked compatibility datum.
Assume that K K K is connected, so that K K K is a fusion category. Central matching has a geometric interpretation. In the specified gauging-wall realization, the parent defect K K K can be resolved into a narrow strip of the Z \mathcal{Z} Z phase bounded by G K \mathcal{G}_K G K and its orientation reverse [29 Sec. 3.3.1] . A compatible condensate network continues through this strip as an algebra L ∈ Z L\in\mathcal{Z} L ∈ Z . The identification ξ : q ( L ) ≃ ι ( B ) \xi:q(L)\simeq\iota(B) ξ : q ( L ) ≃ ι ( B ) joins the two algebra networks on the wall, including their multiplication and unit. Figure 4 illustrates this resolved picture of the pull-through.
Figure 4. Central matching in the resolved picture of the marked defect K K K . The left diagram shows the pull-through of the pointlike condensate insertion B B B across K K K . Resolving K K K gives the Z \mathcal{Z} Z -phase strip on the right, bounded by the specified gauging wall and its orientation reverse. Inside the strip the corresponding insertion is L = B K L=B_K L = B K . White squares carry the local 2-morphism comparisons on the walls. The short shafts and arrows extending from their sides indicate the direction of comparison; they do not connect the point insertions or represent particle trajectories. The matching uses ξ − 1 \xi^{-1} ξ − 1 upon entering the strip and ξ \xi ξ upon leaving it. The cap retains the specified unit η K \eta_K η K . The symbol † \dagger † denotes orientation reversal, not a fusion inverse.
Transport and the commuting square
Fix a K K K -compatible datum ( B , λ ) (B,\lambda) ( B , λ ) with marked suspension A = Σ B A=\Sigma B A = Σ B . The crossing allows the two specified algebra networks to be combined coherently. We will construct a transported condensate B K B_K B K in the gauged phase and an induced separable operation K B K_B K B in the condensed phase. Neither is chosen independently: both are determined by the common composite algebra T λ T_\lambda T λ . We then compare the two orders of operations as specified marked walls, rather than merely identify their final phases. Relative tensor products and their coherences are understood in the framework of [27 , 31 ] .
Combining the two networks
The pseudodistributive law combines the A A A - and K K K -networks into a single algebra. Its underlying object is A ⊗ K A \otimes K A ⊗ K , but its multiplication depends on the chosen crossing. We therefore keep λ \lambda λ in its notation:
T λ : = A ⊗ K , T_\lambda:= A \otimes K, T λ := A ⊗ K ,
μ T λ : = ( μ A ⊗ μ K ) ( id A ⊗ λ ⊗ id K ) , \mu_{T_\lambda} := (\mu_A \otimes\mu_K)(\operatorname{id}_A \otimes\lambda\otimes\operatorname{id}_K), μ T λ := ( μ A ⊗ μ K ) ( id A ⊗ λ ⊗ id K ) ,
η T λ : = η A ⊗ η K . (37) \eta_{T_\lambda} := \eta_A \otimes\eta_K. \tag*{(37)} η T λ := η A ⊗ η K . (37)
Associators and tensorators are understood. In the multiplication, λ \lambda λ moves the middle K K K -factor past the middle A A A -factor before the two multiplications are performed.
There are two specified algebra maps into this composite:
i K = η A ⊗ id K : K → T λ , i A = id A ⊗ η K : A → T λ . (38) i_K = \eta_A \otimes\operatorname{id}_K : K \to T_\lambda,\qquad i_A = \operatorname{id}_A \otimes\eta_K : A \to T_\lambda. \tag*{(38)} i K = η A ⊗ id K : K → T λ , i A = id A ⊗ η K : A → T λ . (38)
They insert the condensation unit and the gauging unit, respectively. Their compatibility as algebra maps is supplied by the unit comparison (31) of the pseudodistributive law.
Recall that Z = Bimod C ( K ) \mathcal{Z} = \operatorname{Bimod}_{\mathcal{C}}(K) Z = Bimod C ( K ) is the defect category of the gauged phase, while Y B \mathcal{Y}_B Y B is the condensed defect theory introduced in § 2.1 . The same composite T λ T_\lambda T λ can be viewed relative to either intermediate phase. Restricting its two actions along i K i_K i K or i A i_A i A gives the transported E 1 E_1 E 1 -algebras
A K = K ( T λ ) K ∈ Alg E 1 s e p ( Z ) , A_K = {}_K(T_\lambda)_K \in\operatorname{Alg}^{\mathrm{sep}}_{E_1}(\mathcal{Z}), A K = K ( T λ ) K ∈ Alg E 1 sep ( Z ) ,
K B = A ( T λ ) A ∈ Alg E 1 s e p ( Y B ) . (39) K_B = {}_A(T_\lambda)_A \in\operatorname{Alg}^{\mathrm{sep}}_{E_1}(\mathcal{Y}_B). \tag*{(39)} K B = A ( T λ ) A ∈ Alg E 1 sep ( Y B ) . (39)
The first can be physically identified with the candidate condensation surface defect after gauging K K K . The second can be physically identified with the induced gauging algebra after condensing B B B .
Recall that the tensor unit of Z \mathcal{Z} Z is the regular K K K –K K K bimodule, and likewise the tensor unit of Y B \mathcal{Y}_B Y B is the regular A A A –A A A bimodule. Consequently, their relative multiplications and units can be defined as
μ A K : T λ ⊗ K T λ ⟶ T λ , η A K = i K , \mu_{A_K} : T_\lambda\otimes_K T_\lambda\longrightarrow T_\lambda,\qquad\eta_{A_K} = i_K, μ A K : T λ ⊗ K T λ ⟶ T λ , η A K = i K ,
μ K B : T λ ⊗ A T λ ⟶ T λ , η K B = i A . (40) \mu_{K_B} : T_\lambda\otimes_A T_\lambda\longrightarrow T_\lambda,\qquad\eta_{K_B} = i_A. \tag*{(40)} μ K B : T λ ⊗ A T λ ⟶ T λ , η K B = i A . (40)
Both multiplications descend from μ T λ \mu_{T_\lambda} μ T λ . If
π R : T λ ⊗ T λ ⟶ T λ ⊗ R T λ , R = K , A , \pi_R : T_\lambda\otimes T_\lambda\longrightarrow T_\lambda\otimes_R T_\lambda,\qquad R = K,A, π R : T λ ⊗ T λ ⟶ T λ ⊗ R T λ , R = K , A ,
is the balancing map, the corresponding coherent comparisons are
μ A K π K ⟹ ≃ μ T λ , μ K B π A ⟹ ≃ μ T λ . \mu_{A_K}\pi_K \overset{\simeq}{\Longrightarrow} \mu_{T_\lambda},\qquad\mu_{K_B}\pi_A \overset{\simeq}{\Longrightarrow} \mu_{T_\lambda}. μ A K π K ⟹ ≃ μ T λ , μ K B π A ⟹ ≃ μ T λ .
The dependence on the fixed λ \lambda λ is suppressed in A K A_K A K and K B K_B K B , but remains part of their algebra structures.
Recovering the transported anyon condensate
In the module realization, the unit η A K : K → A K \eta_{A_K}: K \to A_K η A K : K → A K can be regarded as a K K K –K K K bimodule functor. We may therefore take its right adjoint to define the adjunction algebra
B K : = η A K R η A K ∈ Z , B_K := \eta_{A_K}^{R}\eta_{A_K} \in\mathcal{Z}, B K := η A K R η A K ∈ Z ,
m B K : = η A K R ϵ K η A K , m_{B_K} := \eta_{A_K}^{R}\epsilon^{K}\eta_{A_K}, m B K := η A K R ϵ K η A K ,
u B K : id K ⇒ B K . (41) u_{B_K} : \operatorname{id}_K \Rightarrow B_K. \tag*{(41)} u B K : id K ⇒ B K . (41)
Here
u B K : id K ⇒ η A K R η A K , ε K : η A K η A K R ⇒ id A K u_{B_K}:\operatorname{id}_K \Rightarrow\eta_{A_K}^R\eta_{A_K}, \qquad\varepsilon^K:\eta_{A_K}\eta_{A_K}^R \Rightarrow\operatorname{id}_{A_K} u B K : id K ⇒ η A K R η A K , ε K : η A K η A K R ⇒ id A K
are the unit and counit of the specified adjunction η A K ⊣ η A K R \eta_{A_K} \dashv\eta_{A_K}^R η A K ⊣ η A K R inside Z \mathcal{Z} Z . In particular, ε K \varepsilon^K ε K is a 2-morphism compatible with both K K K -actions. Thus B K B_K B K is an anyon algebra, while A K A_K A K is a surface-defect algebra. Unlike the reverse condensate a K a_K a K (defined in (53) ), the algebra B K B_K B K depends on the chosen physical condensate B B B and its pull-through.
Theorem 3.3 (Marked condensate transport). B K B_K B K is condensable. Its central presentation is
B K ⟷ ( ι ( B ) , τ ) ∈ Z ^ , (42) B_K \longleftrightarrow(\iota(B),\tau)\in\widehat{\mathcal{Z}}, \tag*{(42)} B K ⟷ ( ι ( B ) , τ ) ∈ Z , (42) Here the arrow means that, under the identification Z ≃ Z ^ r e v \mathcal{Z}\simeq\widehat{\mathcal{Z}}^{\mathrm{rev}} Z ≃ Z rev , the central realization of B K B_K B K is the algebra ( ι ( B ) , τ ) (\iota(B),\tau) ( ι ( B ) , τ ) . Moreover,
( A K , η A K ) ≃ ( Σ B K , η Σ B K ) (43) (A_K,\eta_{A_K})\simeq(\Sigma B_K,\eta_{\Sigma B_K}) \tag*{(43)} ( A K , η A K ) ≃ ( Σ B K , η Σ B K ) (43) as marked E 1 E_1 E 1 -algebras. In particular,
q ( B K ) ≃ ι ( B ) as algebras in K , FPdim ( B K ) = FPdim ( B ) . (44) q(B_K)\simeq\iota(B)\quad\text{as algebras in }\mathcal{K},\qquad\operatorname{FPdim}(B_K)=\operatorname{FPdim}(B). \tag*{(44)} q ( B K ) ≃ ι ( B ) as algebras in K , FPdim ( B K ) = FPdim ( B ) . (44)
Proof. Lemma A.2 makes ( ι ( B ) , τ ) (\iota(B),\tau) ( ι ( B ) , τ ) a condensable algebra in Z ^ \widehat{\mathcal{Z}} Z , and hence also in Z \mathcal{Z} Z after reversing the braiding.
Under the right-regular-module identification (17) , the underlying adjunction monad is represented by ι ( B ) \iota(B) ι ( B ) . Its multiplication and unit are ι ( m B ) \iota(m_B) ι ( m B ) and ι ( u B ) \iota(u_B) ι ( u B ) , while its central structure is the marked crossing τ \tau τ . Appendix § 7 .§ 3 establishes this identification inside the K K K –K K K bimodule 2-category and proves that A K A_K A K is its separable module object, with the specified multiplication and free interface. This gives (42) –(43) .
Finally, q q q and ι \iota ι are tensor functors between fusion categories, so they preserve Frobenius–Perron dimensions [30 Proposition 3.3.13] . The algebra identification q ( B K ) ≃ ι ( B ) q(B_K)\simeq\iota(B) q ( B K ) ≃ ι ( B ) gives (44) .
Comparing the specified domain walls
We retain the notation G \mathcal{G} G for gauging walls and W \mathcal{W} W for condensation walls. The four walls to be compared are therefore
G K , W B , W B K , G K B . \mathcal{G}_K,\qquad\mathcal{W}_B,\qquad\mathcal{W}_{B_K},\qquad\mathcal{G}_{K_B}. G K , W B , W B K , G K B .
They represent, respectively, gauging K K K in the parent phase, condensing B B B in the parent phase, condensing the gauged condensate B K B_K B K in the gauged phase, and applying the induced gauging operation K B K_B K B in the condensed phase.
To compare the specified walls, we use their common module presentations over the initial phase C \mathcal{C} C . For a separable algebra R ∈ C R\in\mathcal{C} R ∈ C , RMod R ( C ) \operatorname{RMod}_R(\mathcal{C}) RMod R ( C ) carries the left C \mathcal{C} C -action and the right Bimod C ( R ) \operatorname{Bimod}_{\mathcal{C}}(R) Bimod C ( R ) -action. The distinguished regular module R R R is encoded by the free-module 2-functor X ↦ X ⊗ R X\mapsto X\otimes R X ↦ X ⊗ R . These module presentations describe walls from the parent phase, not the intermediate phases’ own defect categories. Their relation to the bimodule presentation is provided by separable Morita theory [27 Theorem 5.1.2] .
Concretely, the wall symbols denote these marked free-module functors:
G K : C ⟶ RMod K ( C ) , X ⟼ X ⊗ K , W B : C ⟶ RMod A ( C ) , X ⟼ X ⊗ A . (45) \begin{aligned}
\mathcal{G}_K &: \mathcal{C}\longrightarrow\operatorname{RMod}_K(\mathcal{C}),\qquad X\longmapsto X\otimes K,\\
\mathcal{W}_B &: \mathcal{C}\longrightarrow\operatorname{RMod}_A(\mathcal{C}),\qquad X\longmapsto X\otimes A. \tag*{(45)}
\end{aligned} G K W B : C ⟶ RMod K ( C ) , X ⟼ X ⊗ K , : C ⟶ RMod A ( C ) , X ⟼ X ⊗ A . (45)
Note that these functors should be understood as 2-functors that naturally carry left-module compatibility, not as mere tensor functors.
Similarly, the second wall acts on the intermediate phases by the free-module functors over T λ T_\lambda T λ :
W B K : RMod K ( C ) ⟶ RMod T λ ( C ) , M ↦ M ⊗ K T λ , G K B : RMod A ( C ) ⟶ RMod T λ ( C ) , N ↦ N ⊗ A T λ . (46) \begin{aligned}
\mathcal{W}_{B_K} &: \operatorname{RMod}_{K}(\mathcal{C}) \longrightarrow\operatorname{RMod}_{T_\lambda}(\mathcal{C}), \quad M \mapsto M \otimes_{K} T_\lambda,\\
\mathcal{G}_{K_B} &: \operatorname{RMod}_{A}(\mathcal{C}) \longrightarrow\operatorname{RMod}_{T_\lambda}(\mathcal{C}), \quad N \mapsto N \otimes_{A} T_\lambda. \tag*{(46)}
\end{aligned} W B K G K B : RMod K ( C ) ⟶ RMod T λ ( C ) , M ↦ M ⊗ K T λ , : RMod A ( C ) ⟶ RMod T λ ( C ) , N ↦ N ⊗ A T λ . (46)
Theorem 3.3 identifies the first with the specified ordinary condensation wall for B K B_K B K , and the second is the specified gauging wall for K B K_B K B . Comparing the composites of the two groups of functors, we have the following theorem:
Theorem 3.4 (Commuting square of marked walls). Set
T = Bimod C ( T λ ) . \mathcal{T} = \operatorname{Bimod}_{\mathcal{C}}(T_\lambda). T = Bimod C ( T λ ) . There is a unitary equivalence of the composite specified domain walls,
Ξ λ : W B K ∘ G K ⟹ ≃ G K B ∘ W B . (47) \Xi_\lambda: \mathcal{W}_{B_K} \circ\mathcal{G}_K \overset{\simeq}{\Longrightarrow} \mathcal{G}_{K_B} \circ\mathcal{W}_B. \tag*{(47)} Ξ λ : W B K ∘ G K ⟹ ≃ G K B ∘ W B . (47) In the common C \mathcal{C} C -module presentation ((3.30)–(46) ), this is a left C \mathcal{C} C -module 2-natural equivalence between 2-functors
C ⟶ RMod T λ ( C ) . \mathcal{C} \longrightarrow\operatorname{RMod}_{T_\lambda}(\mathcal{C}). C ⟶ RMod T λ ( C ) . Both composites are canonically identified with X ↦ X ⊗ T λ X \mapsto X \otimes T_\lambda X ↦ X ⊗ T λ . The corresponding endpoint identifications are monoidal equivalences
Bimod Bimod C ( K ) ( A K ) ≃ Bimod C ( T λ ) ≃ Bimod Bimod C ( A ) ( K B ) . (48) \operatorname{Bimod}_{\operatorname{Bimod}_{\mathcal{C}}(K)}(A_K) \simeq\operatorname{Bimod}_{\mathcal{C}}(T_\lambda) \simeq\operatorname{Bimod}_{\operatorname{Bimod}_{\mathcal{C}}(A)}(K_B). \tag*{(48)} Bimod Bimod C ( K ) ( A K ) ≃ Bimod C ( T λ ) ≃ Bimod Bimod C ( A ) ( K B ) . (48) Thus the square of specified gapped operations
C → gauge K Z condense B ↓ ↓ condense B K D B → gauge K B T (49) \begin{CD}
\mathcal{C} @>{\text{gauge }K}>> \mathcal{Z}\\
@V{\text{condense }B}VV @VV{\text{condense }B_K}V\\
\mathcal{D}_B @>{\text{gauge }K_B}>> \mathcal{T} \tag*{(49)}
\end{CD} C condense B ↓ ⏐ D B gauge K gauge K B Z ↓ ⏐ condense B K T (49) commutes together with the wall comparison Ξ λ \Xi_\lambda Ξ λ .
Proof. At the level of objects, for X ∈ C X \in\mathcal{C} X ∈ C , the two composite module functors can be compared through the same free right T λ T_\lambda T λ -module:
( X ⊗ K ) ⊗ K T λ ⟶ ≃ X ⊗ T λ ⟵ ≃ ( X ⊗ A ) ⊗ A T λ . (50) (X \otimes K) \otimes_K T_\lambda\overset{\simeq}{\longrightarrow} X \otimes T_\lambda\overset{\simeq}{\longleftarrow} (X \otimes A) \otimes_A T_\lambda. \tag*{(50)} ( X ⊗ K ) ⊗ K T λ ⟶ ≃ X ⊗ T λ ⟵ ≃ ( X ⊗ A ) ⊗ A T λ . (50)
These equivalences preserve the right T λ T_\lambda T λ -action and are compatible with the left C \mathcal{C} C -action. Appendix § 7 .§ 4 verifies the remaining complete identification data using the module realization of separable Morita theory [27 , 35 ] , including the endpoint actions and the composite with the regular marked wall, and establishes (48) .
Figure 5 displays the wall-level content. Reading from left to right above the junction gives gauging followed by condensation; reading below it gives condensation followed by gauging. The junction implements the 2-natural equivalence Ξ λ \Xi_\lambda Ξ λ .
Figure 5. The commuting square of specified domain walls. Reading from left to right above the junction gives C → Z → T \mathcal{C} \to\mathcal{Z} \to\mathcal{T} C → Z → T : gauging K K K followed by condensing B K B_K B K . Reading below it gives C → Y B → T \mathcal{C} \to\mathcal{Y}_B \to\mathcal{T} C → Y B → T : condensing B B B followed by gauging K B K_B K B . The central junction implements the unitary comparison Ξ λ \Xi_\lambda Ξ λ between these composite marked walls. Purple branches denote gauging walls and teal branches denote condensation walls. The dashed arrows indicate the reading direction and are not additional defects. Invertibility of the comparison does not imply invertibility of the individual walls.
Neither the transport theorem nor the wall comparison requires connectedness of B B B or B K B_K B K . Applying the local-module description of § 2.1 to the transported algebra gives
Ω T ≃ Z B K l o c . (51) \Omega_{\mathcal{T}} \simeq\mathcal{Z}^{\mathrm{loc}}_{B_K}. \tag*{(51)} Ω T ≃ Z B K loc . (51)
For disconnected B K B_K B K , this is understood componentwise as in (4) .
If B K B_K B K is connected, the condensation-completed phase realization gives
Y B / / K B ≃ T ≃ Σ ( Z B K l o c ) . (52) \mathcal{Y}_B // K_B \simeq\mathcal{T} \simeq\Sigma\left(\mathcal{Z}_{B_K}^{\mathrm{loc}}\right). \tag*{(52)} Y B // K B ≃ T ≃ Σ ( Z B K loc ) . (52)
The intermediate vacuum-resolved anyons remain those of (11) . Selecting a parent vacuum before transporting the full network can discard part of K B K_B K B and its marked gauging wall. Strong connectedness of K B K_B K B is not guaranteed.
Note that the square involves the compatibility data of the two specified networks, and therefore it does not assert that arbitrary condensation commutes with arbitrary gauging. For the standard full gauging of an ordinary group, K K K -compatibility recovers equivariant anyon condensation, as explained in Appendix § 9 [8 ] . The general statement concerns the two specified algebra networks and retains the comparison of their composite marked walls.
Complete Higgs behavior with unit absorption
Complete Higgs behavior means that the marked square ends in a single intrinsically trivial vacuum. The specified unit of K K K may already absorb a parent anyon algebra. The appropriate criterion therefore concerns the condensate after this absorption, together with the number of branches it produces.
Throughout this section K K K is connected, so that K \mathcal{K} K is a unitary fusion category. We fix compatible data ( B , τ ) (B,\tau) ( B , τ ) and write L = B K L = B_K L = B K for the transported condensate.
Absorption and local reduction
The two central functors of Section 3.1 determine the absorption algebras
d K : = ι R ( 1 K ) ∈ C , a K : = q R ( 1 K ) ∈ Z . (53) d_K := \iota^R(1_{\mathcal{K}}) \in\mathcal{C}, \qquad a_K := q^R(1_{\mathcal{K}}) \in\mathcal{Z}. \tag*{(53)} d K := ι R ( 1 K ) ∈ C , a K := q R ( 1 K ) ∈ Z . (53)
They carry the connected condensable algebra structures induced by the adjunctions [6 Lemma 3.5] , with the unitary Frobenius presentations fixed in § 2 . Set
δ K : = FPdim C ( d K ) , κ K : = FPdim Z ( a K ) , M K : = C d K l o c \delta_K := \operatorname{FPdim}_{\mathcal{C}}(d_K), \qquad\kappa_K := \operatorname{FPdim}_{\mathcal{Z}}(a_K), \qquad\mathcal{M}_K := \mathcal{C}_{d_K}^{\mathrm{loc}} δ K := FPdim C ( d K ) , κ K := FPdim Z ( a K ) , M K := C d K loc
The category M K \mathcal{M}_K M K is the parent bulk after condensation of the algebra absorbed by the specified unit.
Lemma 4.1 (Local reduction). For compatible data ( B , τ ) (B,\tau) ( B , τ ) , the algebra B B B centralizes d K d_K d K and L = B K L = B_K L = B K centralizes a K a_K a K . Consequently,
B ‾ : = F d K ( B ) (54) \overline{B} := F_{d_K}(B) \tag*{(54)} B := F d K ( B ) (54) is a condensable algebra in M K \mathcal{M}_K M K .
The dimensions satisfy
FPdim ( C ) = δ K 2 FPdim ( M K ) , FPdim ( Z ) = κ K 2 FPdim ( M K ) . (55) \operatorname{FPdim}(\mathcal{C}) = \delta_K^2 \operatorname{FPdim}(\mathcal{M}_K), \qquad\operatorname{FPdim}(\mathcal{Z}) = \kappa_K^2 \operatorname{FPdim}(\mathcal{M}_K). \tag*{(55)} FPdim ( C ) = δ K 2 FPdim ( M K ) , FPdim ( Z ) = κ K 2 FPdim ( M K ) . (55) Moreover,
n K ( B ) : = dim Hom C ( d K , B ) = dim Hom M K ( 1 M K , B ‾ ) = dim Hom Z ( a K , L ) . (56) n_K(B) := \dim\operatorname{Hom}_{\mathcal{C}}(d_K,B)
= \dim\operatorname{Hom}_{\mathcal{M}_K}(1_{\mathcal{M}_K},\overline{B}) = \dim\operatorname{Hom}_{\mathcal{Z}}(a_K,L). \tag*{(56)} n K ( B ) := dim Hom C ( d K , B ) = dim Hom M K ( 1 M K , B ) = dim Hom Z ( a K , L ) . (56) Thus B ‾ \overline{B} B has n K ( B ) n_K(B) n K ( B ) connected components. If L L L is connected, then n K ( B ) ≤ κ K n_K(B) \le\kappa_K n K ( B ) ≤ κ K .
Proof. Let U : Z ( K ) → K U : \mathcal{Z}(\mathcal{K}) \to\mathcal{K} U : Z ( K ) → K be the forgetful functor, let I = U R I = U^R I = U R , and put R = I ( 1 K ) R = I(1_{\mathcal{K}}) R = I ( 1 K ) . Adjunction identifies the intersections of R R R with the two factors of (18) as the subalgebras j C ( d K ) j_{\mathcal{C}}(d_K) j C ( d K ) and j Z ( a K ) j_{\mathcal{Z}}(a_K) j Z ( a K ) . Under the reconstruction K ≃ Z ( K ) R \mathcal{K} \simeq\mathcal{Z}(\mathcal{K})_R K ≃ Z ( K ) R , the centralizer-image theorem [36 Theorem 3.15] identifies Im ( q ) \operatorname{Im}(q) Im ( q ) with the objects X X X for which I ( X ) I(X) I ( X ) is local as a j C ( d K ) j_{\mathcal{C}}(d_K) j C ( d K ) -module. Central matching places X = ι ( B ) X = \iota(B) X = ι ( B ) in this image, and the projection formula gives
I ( ι ( B ) ) ≃ j ( B ) ⊗ R I(\iota(B)) \simeq j(B) \otimes R I ( ι ( B )) ≃ j ( B ) ⊗ R
as right R R R -modules. Restrict its locality identity along id j ( B ) ⊗ u R ⊗ id j C ( d K ) \operatorname{id}_{j(B)} \otimes u_R \otimes\operatorname{id}_{j_{\mathcal{C}}(d_K)} id j ( B ) ⊗ u R ⊗ id j C ( d K ) . Since j C ( d K ) ↪ R j_{\mathcal{C}}(d_K) \hookrightarrow R j C ( d K ) ↪ R is a monomorphism, the resulting identity is
c j C ( d K ) , j ( B ) c j ( B ) , j C ( d K ) = id j ( B ) ⊗ j C ( d K ) . c_{j_{\mathcal{C}}(d_K),j(B)}c_{j(B),j_{\mathcal{C}}(d_K)}
= \operatorname{id}_{j(B)}\otimes j_{\mathcal{C}}(d_K). c j C ( d K ) , j ( B ) c j ( B ) , j C ( d K ) = id j ( B ) ⊗ j C ( d K ) .
Hence B B B centralizes d K d_K d K . Applying the same argument with the two factors interchanged shows that L L L centralizes a K a_K a K . The free-module locality criterion [6 Lemma 3.15] now places F d K ( B ) F_{d_K}(B) F d K ( B ) in M K \mathcal{M}_K M K . On the centralizer of d K d_K d K , the free-module functor is braided, so it transports the condensable algebra structure of B B B to B ‾ \overline{B} B .
The centralizer-image theorem and central-functor reconstruction also give
FPdim ( Im ( q ) ) = FPdim ( K ) δ K = FPdim ( Z ) κ K . \operatorname{FPdim}(\operatorname{Im}(q)) =
\frac{\operatorname{FPdim}(\mathcal{K})}{\delta_K}
=
\frac{\operatorname{FPdim}(\mathcal{Z})}{\kappa_K}. FPdim ( Im ( q )) = δ K FPdim ( K ) = κ K FPdim ( Z ) .
Combining this with FPdim ( K ) 2 = FPdim ( C ) FPdim ( Z ) \operatorname{FPdim}(\mathcal{K})^2 = \operatorname{FPdim}(\mathcal{C})\operatorname{FPdim}(\mathcal{Z}) FPdim ( K ) 2 = FPdim ( C ) FPdim ( Z ) yields
FPdim ( Z ) FPdim ( C ) = κ K δ K . \frac{\sqrt{\operatorname{FPdim}(\mathcal{Z})}}{\sqrt{\operatorname{FPdim}(\mathcal{C})}}
= \frac{\kappa_K}{\delta_K}. FPdim ( C ) FPdim ( Z ) = δ K κ K .
The local-module dimension formula (6) then gives (55) . Central-functor reconstruction identifies C d K \mathcal{C}_{d_K} C d K with the full image of ι \iota ι , carrying F d K F_{d_K} F d K to ι \iota ι and the regular module d K d_K d K to 1 K 1_{\mathcal{K}} 1 K . Therefore
Hom M K ( 1 M K , B ‾ ) ≃ Hom K ( 1 K , ι ( B ) ) . \operatorname{Hom}_{\mathcal{M}_K}(1_{\mathcal{M}_K},\overline{B})
\simeq\operatorname{Hom}_{\mathcal{K}}(1_{\mathcal{K}},\iota(B)). Hom M K ( 1 M K , B ) ≃ Hom K ( 1 K , ι ( B )) .
Adjunction and symmetry of Hom-space dimensions in the semisimple setting give
dim Hom K ( 1 K , ι ( B ) ) = dim Hom C ( d K , B ) , dim Hom K ( 1 K , q ( L ) ) = dim Hom Z ( a K , L ) . \begin{aligned}
\dim\operatorname{Hom}_{\mathcal{K}}(\mathbf{1}_{\mathcal{K}},\iota(B)) &= \dim\operatorname{Hom}_{\mathcal{C}}(d_K,B),\\
\dim\operatorname{Hom}_{\mathcal{K}}(\mathbf{1}_{\mathcal{K}},q(L)) &= \dim\operatorname{Hom}_{\mathcal{Z}}(a_K,L).
\end{aligned} dim Hom K ( 1 K , ι ( B )) dim Hom K ( 1 K , q ( L )) = dim Hom C ( d K , B ) , = dim Hom Z ( a K , L ) .
The identification q ( L ) ≃ ι ( B ) q(L)\simeq\iota(B) q ( L ) ≃ ι ( B ) proves (56) . If L L L is connected, adjunction identifies Hom Z ( a K , L ) \operatorname{Hom}_{\mathcal{Z}}(a_K,L) Hom Z ( a K , L ) with Hom Z L ( F L ( a K ) , L ) \operatorname{Hom}_{\mathcal{Z}_L}(F_L(a_K),L) Hom Z L ( F L ( a K ) , L ) . Its dimension is the multiplicity of the tensor unit in F L ( a K ) F_L(a_K) F L ( a K ) , whose Frobenius–Perron dimension is κ K \kappa_K κ K . This proves the bound.
The integer n K ( B ) n_K(B) n K ( B ) counts the condensation branches after absorption of the specified unit. Each original parent vacuum can contribute several such branches.
The general complete Higgs criterion
Definition 4.2 (Complete Higgs type). A compatible condensation ( B , τ ) (B,\tau) ( B , τ ) is of complete Higgs type if its final intrinsic defect theory satisfies
D B / / K B = T ≃ 2 V e c . (57) \mathcal{D}_{B//K_B}=\mathcal{T}\simeq2\mathrm{Vec}. \tag*{(57)} D B // K B = T ≃ 2 Vec . (57) We retain the gravitational-sector convention of § 2.1 . Write the connected algebra decompositions as
B = ⨁ i = 1 r B i , B ‾ = ⨁ α = 1 n K ( B ) B ‾ α . B=\bigoplus_{i=1}^{r}B_i,\qquad\overline{B}=\bigoplus_{\alpha=1}^{n_K(B)}\overline{B}_{\alpha}. B = i = 1 ⨁ r B i , B = α = 1 ⨁ n K ( B ) B α .
Theorem 4.3 (Complete Higgs criterion). For a connected specified operation K K K , the following conditions on compatible data ( B , τ ) (B,\tau) ( B , τ ) are equivalent:
(i) ( B , τ ) (B,\tau) ( B , τ ) is of complete Higgs type;
(ii) L = B K L=B_K L = B K is Lagrangian in Z \mathcal{Z} Z ;
(iii) L L L is connected,
n K ( B ) = κ K , n_K(B)=\kappa_K, n K ( B ) = κ K , and every connected component B ‾ α \overline{B}_{\alpha} B α is Lagrangian in M K \mathcal{M}_K M K .
When these conditions hold, define
n i : = dim Hom C ( d K , B i ) , D i : = FPdim ( C B i l o c ) . n_i:=\dim\operatorname{Hom}_{\mathcal{C}}(d_K,B_i),\qquad D_i:=\sqrt{\operatorname{FPdim}(C_{B_i}^{\mathrm{loc}})}. n i := dim Hom C ( d K , B i ) , D i := FPdim ( C B i loc ) . Then
n i ∈ Z > 0 , D i = δ K n i , ∑ i = 1 r n i = κ K . (58) n_i\in\mathbb{Z}_{>0},\qquad D_i=\frac{\delta_K}{n_i},\qquad\sum_{i=1}^{r}n_i=\kappa_K. \tag*{(58)} n i ∈ Z > 0 , D i = n i δ K , i = 1 ∑ r n i = κ K . (58) In particular, κ K \kappa_K κ K is an integer and
κ K δ K ≤ r ≤ κ K . (59) \frac{\kappa_K}{\delta_K}\le r\le\kappa_K. \tag*{(59)} δ K κ K ≤ r ≤ κ K . (59) Proof. The equivalence of (i) and (ii) follows from (51) –(52) : a single trivial final vacuum is equivalent to L L L being connected with Z L l o c ≃ V e c \mathcal{Z}^{\mathrm{loc}}_L\simeq\mathrm{Vec} Z L loc ≃ Vec .
Suppose (ii) holds. By Lemma 4.1, a K a_K a K centralizes L L L , so F L ( a K ) F_L(a_K) F L ( a K ) is local. It therefore belongs to Z L l o c ≃ V e c \mathcal{Z}^{\mathrm{loc}}_L\simeq\mathrm{Vec} Z L loc ≃ Vec . Its dimension is κ K \kappa_K κ K , and its tensor-unit multiplicity is n K ( B ) n_K(B) n K ( B ) by adjunction. Hence
n K ( B ) = dim Hom Z L ( F L ( a K ) , L ) = FPdim Z L ( F L ( a K ) ) = κ K . (60) n_K(B)=\dim\operatorname{Hom}_{\mathcal{Z}_L}(F_L(a_K),L)=\operatorname{FPdim}_{\mathcal{Z}_L}(F_L(a_K))=\kappa_K. \tag*{(60)} n K ( B ) = dim Hom Z L ( F L ( a K ) , L ) = FPdim Z L ( F L ( a K )) = κ K . (60)
The free-module functor preserves Frobenius–Perron dimensions [30 Proposition 3.3.13] , so
FPdim M K ( B ‾ ) = FPdim C ( B ) = FPdim Z ( L ) = κ K FPdim ( M K ) = FPdim ( Z ) . \operatorname{FPdim}_{\mathcal{M}_K}(\overline{B})=\operatorname{FPdim}_{\mathcal{C}}(B)=\operatorname{FPdim}_{\mathcal{Z}}(L)
=\kappa_K\sqrt{\operatorname{FPdim}(\mathcal{M}_K)}
=\sqrt{\operatorname{FPdim}(\mathcal{Z})}. FPdim M K ( B ) = FPdim C ( B ) = FPdim Z ( L ) = κ K FPdim ( M K ) = FPdim ( Z ) .
Every connected condensable algebra in M K \mathcal{M}_K M K has dimension at most FPdim ( M K ) \sqrt{\operatorname{FPdim}(\mathcal{M}_K)} FPdim ( M K ) by (6) . There are exactly κ K \kappa_K κ K connected components of B ‾ \overline{B} B , and their total dimension reaches the sum of these upper bounds. Each B ‾ α \overline{B}_\alpha B α therefore reaches the bound and is Lagrangian. This proves (iii).
Conversely, under (iii),
FPdim Z ( L ) = FPdim M K ( B ‾ ) = κ K FPdim ( M K ) = FPdim ( Z ) . \operatorname{FPdim}_{\mathcal{Z}}(L)=\operatorname{FPdim}_{\mathcal{M}_K}(\overline{B})
=\kappa_K\sqrt{\operatorname{FPdim}(\mathcal{M}_K)}
=\sqrt{\operatorname{FPdim}(\mathcal{Z})}. FPdim Z ( L ) = FPdim M K ( B ) = κ K FPdim ( M K ) = FPdim ( Z ) .
Together with connectedness, this makes L L L Lagrangian.
For the remaining assertions, F d K ( B i ) F d_K(B_i) F d K ( B i ) is an orthogonal algebra summand of B ‾ \overline{B} B . The free–forgetful adjunction and self-duality of d K d_K d K show that it has n i n_i n i connected components. Each of these is Lagrangian, so
FPdim C ( B i ) = n i FPdim ( M K ) . \operatorname{FPdim}_{\mathcal{C}}(B_i)=n_i\sqrt{\operatorname{FPdim}(\mathcal{M}_K)}. FPdim C ( B i ) = n i FPdim ( M K ) .
On the other hand,
FPdim C ( B i ) = FPdim ( C ) D i = δ K FPdim ( M K ) D i . \operatorname{FPdim}_{\mathcal{C}}(B_i)
=\frac{\sqrt{\operatorname{FPdim}(\mathcal{C})}}{D_i}
=\frac{\delta_K\sqrt{\operatorname{FPdim}(\mathcal{M}_K)}}{D_i}. FPdim C ( B i ) = D i FPdim ( C ) = D i δ K FPdim ( M K ) .
Comparison gives D i = δ K / n i D_i=\delta_K/n_i D i = δ K / n i . Additivity gives ∑ i n i = n K ( B ) = κ K \sum_i n_i=n_K(B)=\kappa_K ∑ i n i = n K ( B ) = κ K . The unit of each F d K ( B i ) F d_K(B_i) F d K ( B i ) ensures n i ≥ 1 n_i\geq1 n i ≥ 1 , while D i ≥ 1 D_i\geq1 D i ≥ 1 gives n i ≤ δ K n_i\leq\delta_K n i ≤ δ K . Summing these inequalities proves (59) .
Thus complete Higgs behavior fixes the branch number after unit absorption and trivializes every resulting branch. Equation (58) also fixes the total quantum dimension of the residual intrinsic order in each original parent vacuum.
For example, take the toric-code operation K = Σ ( 1 ⊕ e ) K=\Sigma(1\oplus e) K = Σ ( 1 ⊕ e ) and B = 1 B=1 B = 1 , as in § 5.7 . Here
δ K = 2 , κ K = n K ( B ) = 1 , M K ≃ Vec . \delta_K=2,\qquad\kappa_K=n_K(B)=1,\qquad\mathcal{M}_K\simeq\operatorname{Vec}. δ K = 2 , κ K = n K ( B ) = 1 , M K ≃ Vec .
The final phase is trivial, while the original parent vacuum has D 1 = 2 D_1=2 D 1 = 2 . In this case the specified unit supplies the condensation that removes the parent topological order.
The transparent-unit specialization
A connected operation is strongly connected when its unit η K \eta_K η K includes the transparent direct summand of K K K . In the module presentation, this means that ι \iota ι includes the regular C \mathcal{C} C -module as a full summand. The equivalent central-functor conditions are
d K ≃ 1 C ⟺ ι is fully faithful ⟺ q is dominant . (61) d_K\simeq1_{\mathcal{C}}\Longleftrightarrow\iota\text{ is fully faithful}\Longleftrightarrow q\text{ is dominant}. \tag*{(61)} d K ≃ 1 C ⟺ ι is fully faithful ⟺ q is dominant . (61)
Indeed, the projection formula gives R ι ( c ) ≃ c ⊗ d K \mathcal{R}\iota(c)\simeq c\otimes d_K R ι ( c ) ≃ c ⊗ d K , which identifies the first two conditions. A fully faithful ι \iota ι splits off the regular module summand by semisimplicity and rigidity. The equivalence with dominance of q q q is the centralizer theorem [36 Theorem 3.12] .
Proposition 4.4 (Reverse condensation and inverse transport). Suppose d K ≃ 1 C d_K\simeq1_{\mathcal{C}} d K ≃ 1 C . There are equivalences retaining the specified central actions,
C ≃ Z a K l o c , K ≃ Z ^ a K ≃ ( Z a K ) m o p . (62) \mathcal{C}\simeq\mathcal{Z}^{\mathrm{loc}}_{a_K},\qquad
\mathcal{K}\simeq\widehat{\mathcal{Z}}_{a_K}\simeq(\mathcal{Z}_{a_K})^{\mathrm{mop}}. \tag*{(62)} C ≃ Z a K loc , K ≃ Z a K ≃ ( Z a K ) mop . (62)
They identify the gauging-wall presentation with the orientation reverse of the a K a_K a K -condensation wall. Compatible condensation data are classified, up to isomorphism, by condensable algebras L ∈ Z L \in\mathcal{Z} L ∈ Z satisfying
c a K , L c L , a K = id L ⊗ a K . (63) c_{a_K,L}c_{L,a_K}=\operatorname{id}_{L\otimes a_K}. \tag*{(63)} c a K , L c L , a K = id L ⊗ a K . (63)
The parent algebra and its transport are
B = F a K ( L ) ∈ Z a K l o c ≃ C , B K ≃ L . (64) B=F_{a_K}(L)\in\mathcal{Z}^{\mathrm{loc}}_{a_K}\simeq\mathcal{C},\qquad B_K\simeq L. \tag*{(64)} B = F a K ( L ) ∈ Z a K loc ≃ C , B K ≃ L . (64)
For a prescribed parent algebra B B B , the datum retains a unitary algebra identification F a K ( L ) ≃ B F_{a_K}(L)\simeq B F a K ( L ) ≃ B .
Proof. Dominance of q q q and central-functor reconstruction give K ≃ Z ^ a K \mathcal{K}\simeq\widehat{\mathcal{Z}}_{a_K} K ≃ Z a K [6 Lemma 3.5] . The center decomposition [6 Corollary 3.30] , compared with (18) , identifies the centralizer of the specified Z ^ \widehat{\mathcal{Z}} Z action with C \mathcal{C} C . This gives the local-module equivalence in (62) . The physical orientation produces the displayed mop for wall fusion; the regular modules and free interfaces correspond under these identifications.
If L L L satisfies (63) , its free a K a_K a K -module is local and carries the transported condensable algebra structure. Under (62) , it supplies the algebra identification q ( L ) ≃ ι ( B ) q(L)\simeq\iota(B) q ( L ) ≃ ι ( B ) . Theorem 3.2 then reconstructs the crossing.
Conversely, central matching identifies q ( L ) q(L) q ( L ) with the local module ι ( B ) \iota(B) ι ( B ) . The free-module locality criterion gives (63) . Full faithfulness of ι \iota ι , together with the morphism condition (33) , identifies the isomorphisms in the two descriptions. □
For a compatible datum, define the algebra of compatible parent vacuum operators by
V B τ : = { v ∈ Hom C ( 1 , B ) ∣ τ X ( ι ( v ) ⊗ id X ) = id X ⊗ ι ( v ) for all X ∈ K } . (65) \mathcal{V}^{\tau}_{B}:=\{v\in\operatorname{Hom}_{\mathcal{C}}(\mathbf{1},B)\mid\tau_X(\iota(v)\otimes\operatorname{id}_X)=\operatorname{id}_X\otimes\iota(v)\text{ for all }X\in\mathcal{K}\}. \tag*{(65)} V B τ := { v ∈ Hom C ( 1 , B ) ∣ τ X ( ι ( v ) ⊗ id X ) = id X ⊗ ι ( v ) for all X ∈ K } . (65)
Its product and unit are inherited from B B B . Write r c o m p : = dim V B τ r_{\mathrm{comp}}:=\dim\mathcal{V}^{\tau}_{B} r comp := dim V B τ .
Corollary 4.5 (Transparent-unit vacua and complete Higgs). Suppose d K ≃ 1 C d_K\simeq\mathbf{1}_{\mathcal{C}} d K ≃ 1 C . Then
r = dim Hom C ( 1 , B ) = dim Hom Z ( a K , B K ) , (66) r=\dim\operatorname{Hom}_{\mathcal{C}}(\mathbf{1},B)=\dim\operatorname{Hom}_{\mathcal{Z}}(a_K,B_K), \tag*{(66)} r = dim Hom C ( 1 , B ) = dim Hom Z ( a K , B K ) , (66) and there is an isomorphism of unital algebras
V B τ ≃ Hom Z ( 1 , B K ) . (67) \mathcal{V}^{\tau}_{B}\simeq\operatorname{Hom}_{\mathcal{Z}}(\mathbf{1},B_K). \tag*{(67)} V B τ ≃ Hom Z ( 1 , B K ) . (67) In particular, B K B_K B K is connected precisely when V B τ = C u B \mathcal{V}^{\tau}_{B}=\mathbb{C}u_B V B τ = C u B .
Complete Higgs behavior is equivalent to the three conditions
B K is connected , r = κ K , B i is Lagrangian in C for every i . B_K\text{ is connected},\qquad r=\kappa_K,\qquad B_i\text{ is Lagrangian in }\mathcal{C}\text{ for every }i. B K is connected , r = κ K , B i is Lagrangian in C for every i .
Proof. With d K ≃ 1 C d_K\simeq\mathbf{1}_{\mathcal{C}} d K ≃ 1 C , equation (56) gives n K ( B ) = r n_K(B)=r n K ( B ) = r and proves (66) . Central matching identifies Hom Z ( 1 , B K ) \operatorname{Hom}_{\mathcal{Z}}(\mathbf{1},B_K) Hom Z ( 1 , B K ) with the central morphisms from 1 K \mathbf{1}_{\mathcal{K}} 1 K to ( ι ( B ) , τ ) (\iota(B),\tau) ( ι ( B ) , τ ) . Full faithfulness of ι \iota ι identifies these with the parent morphisms satisfying (65) . This correspondence preserves multiplication and unit, proving (67) .
Finally, M K = C \mathcal{M}_K=\mathcal{C} M K = C , F d K F_{d_K} F d K is the identity, and n K ( B ) = r n_K(B)=r n K ( B ) = r . Theorem 4.3 therefore specializes to the stated three conditions. □
For fixed strongly connected K \mathcal{K} K , the existence criterion of Appendix § 8 determines whether these conditions can be realized by some parent condensate.
Examples
Three controlled comparisons organize the examples. First, the specified operation is fixed and the parent condensate varies (§ 5.3 ). Second, the parent condensate is fixed and the crossing varies (§ 5.4 ). Third, the transported condensate is fixed and the parent condensate varies (§ 5.7 ); this requires a specified unit that absorbs a nontrivial parent algebra, since for a transparent unit Proposition 4.4 determines the parent condensate from the transported one. The comparisons are not separate variations of one entry of a triple: changing the crossing while the parent condensate is fixed generally changes both the transported condensate and the algebra identification on the wall.
Each example follows the same pattern. The specified input is given, the crossing is displayed on the neutral and charged channels, the transported condensate and the induced operation are identified, the final phase is read off, and the general result it illustrates is named. The first three subsections treat the S 3 / Z 2 S_3/\mathbb{Z}_2 S 3 / Z 2 operation on the toric code; the pure electric condensation comes last because it exercises the general criterion of § 4 rather than its transparent-unit specialization.
Running example: the specified operation and mixed condensation
The specified operation
Condensing the electric charge e e e of the toric code with the connected electric condensate
identifies it with the vacuum, confines the magnetic flux, and leaves no intrinsic topological order [2 ] . We ask how this Higgs picture changes when the toric code carries a non-invertible symmetry, and whether gauging that symmetry can reveal residual topological order after the toric-code anyons have been removed.
Start with an untwisted trivial bosonic paramagnet with on-site symmetry S 3 S_3 S 3 , and gauge, without twist, the subgroup generated by a transposition. A transposition exchanges two of the three colours, so the subgroup it generates is not normal in S 3 S_3 S 3 . The resulting intrinsic topological order is the toric code. The remaining symmetry is not an ordinary quotient group: the original S 3 S_3 S 3 defects, dressed by the subgroup-gauging interfaces, give non-invertible surface defects [16 ] . Let K K K be the specified separable operation obtained in this way. Applying it completes the gauging to S 3 S_3 S 3 [15 ] ; we call K K K the S 3 / Z 2 S_3/\mathbb{Z}_2 S 3 / Z 2 operation. The notation names the specified operation rather than a quotient group. We write D ( F ) \mathcal{D}(F) D ( F ) for the anyon category of untwisted F F F -gauge theory. The parent phase and the specified operation are
C = D ( Z 2 ) , Z = D ( S 3 ) , d K ≃ 1 C , κ K = 3. (68) \mathcal{C}=\mathcal{D}(\mathbb{Z}_2),\qquad\mathcal{Z}=\mathcal{D}(S_3),\qquad d_K\simeq1_{\mathcal{C}},\qquad\kappa_K=3. \tag*{(68)} C = D ( Z 2 ) , Z = D ( S 3 ) , d K ≃ 1 C , κ K = 3. (68)
The first equality identifies the toric-code phase and the second the phase reached by completing the gauging. The third equality says that the specified unit is transparent, and the last records the dimension of the reverse condensate. The algebra α K \alpha_K α K is the electric reverse condensate of the subgroup-gauging interface [15 ] , Sec. 6.3: it is a connected electric algebra, and condensing it returns the toric code.
The toric code has an electric charge e e e and a magnetic flux m m m . Their composite is also a simple anyon, and we write 1 1 1 for the vacuum. Besides the parent anyons, the gauging wall carries exactly two further simple line excitations, which we call W W W and W m W_m W m . Both have Frobenius–Perron dimension two. They are not anyons of the toric code, and neither is a complete symmetry surface defect: they are simple line excitations on the specified gauging wall. The second is obtained from the first by fusing the magnetic anyon, W m : = ι ( m ) ⊗ K W W_m := \iota(m) \otimes_{\mathcal K} W W m := ι ( m ) ⊗ K W . A coordinate model of the wall category, and the identification of W W W and W m W_m W m with the two remaining wall simples, are given in Appendix § 10 .§ 1 .
As a surface defect, the specified operation decomposes as
K ≃ 1 C ⊕ Σ b . K \simeq1_{\mathbb{C}} \oplus\Sigma_b. K ≃ 1 C ⊕ Σ b .
This is a decomposition of the underlying surface defect, not a choice of two independent algebra operations; the multiplication, unit and junctions of K K K remain part of the specified operation. The first summand is the transparent surface defect, and the second is the nontransparent summand studied on its own in § 5.7 .
The mixed condensation and its charged crossing
To construct compatible condensations, return to the S 3 S_3 S 3 -symmetric system before the subgroup is gauged and let it order. The order parameter takes the three colours as values; the gauged transposition fixes one colour and interchanges the other two. We keep the full multiplet of ordered states and its symmetry defects, and assume no additional intrinsic topological order or symmetry-protected topological decoration. Gauging the subgroup in the paramagnet gives the parent toric code, and gauging it in the ordered phase gives the parent condensate: the fixed colour keeps the toric-code order, while the pair of exchanged colours undergoes the electric condensation. The parent condensate therefore has two orthogonal algebra components, and its two vacua are the fixed-colour vacuum and the electric-condensation vacuum. Keeping both branches gives the parent algebra and the intermediate phase
B = 1 ⊕ b , Ω Y B ≃ D ( Z 2 ) ⊕ V e c . (69) B = 1 \oplus b,\qquad\Omega_{\mathcal{Y}}B \simeq D(\mathbb{Z}_2) \oplus\mathrm{Vec}. \tag*{(69)} B = 1 ⊕ b , Ω Y B ≃ D ( Z 2 ) ⊕ Vec . (69)
The first summand is the vacuum that keeps the toric-code order; the second is the intrinsically trivial vacuum produced by the electric condensation. We label phases here by their vacuum-resolved anyon categories; the outer direct sum records alternative local vacua, not stacked topological orders, and the display relates gapped endpoints without assuming a continuous transition. The two vacua descend from one order parameter multiplet, so the inherited non-invertible defects must retain their relation.
The two summands of B B B carry the following channels. The neutral vacuum operator of the toric-code branch is the unit of the 1 1 1 summand, written p T C p_{\mathrm{TC}} p TC . The electric-condensation branch is b = 1 ⊕ e b = 1 \oplus e b = 1 ⊕ e . Its unit is p t r i v p_{\mathrm{triv}} p triv , and its electric channel is written ϕ \phi ϕ . Thus
p T C , p t r i v ∈ Hom C ( 1 , B ) , ϕ ∈ Hom C ( e , B ) . p_{\mathrm{TC}}, p_{\mathrm{triv}} \in\operatorname{Hom}_{\mathcal C}(1,B),\qquad\phi\in\operatorname{Hom}_{\mathcal C}(e,B). p TC , p triv ∈ Hom C ( 1 , B ) , ϕ ∈ Hom C ( e , B ) .
In the algebra B B B these channels multiply as
p T C 2 = p T C , p t r i v 2 = p t r i v , p T C p t r i v = 0 , p_{\mathrm{TC}}^2 = p_{\mathrm{TC}},\qquad p_{\mathrm{triv}}^2 = p_{\mathrm{triv}},\qquad p_{\mathrm{TC}}p_{\mathrm{triv}} = 0, p TC 2 = p TC , p triv 2 = p triv , p TC p triv = 0 ,
ϕ 2 = p t r i v , p t r i v ϕ = ϕ , p T C ϕ = 0 , u B = p T C + p t r i v . (70) \phi^2 = p_{\mathrm{triv}},\qquad p_{\mathrm{triv}}\phi= \phi,\qquad p_{\mathrm{TC}}\phi= 0,\qquad u_B = p_{\mathrm{TC}} + p_{\mathrm{triv}}. \tag*{(70)} ϕ 2 = p triv , p triv ϕ = ϕ , p TC ϕ = 0 , u B = p TC + p triv . (70)
Thus p T C p_{\mathrm{TC}} p TC and p t r i v p_{\mathrm{triv}} p triv are the two neutral parent vacuum projectors. The channel ϕ \phi ϕ is not a third vacuum operator: it is electrically charged, and an insertion of ϕ \phi ϕ on its own must be attached to an electric line or absorbed at a defect junction.
To display the crossing, choose orthonormal coordinate vectors w + , w − w_+, w_- w + , w − in a concrete two-dimensional realization of the wall excitation W W W . The transposition exchanges the two coordinates. The coordinates lie inside one simple wall excitation, not two wall simples and not two vacua. In formulas for the walls we write the images under ι \iota ι of the condensate channels with the same symbols. When W W W crosses the toric-code vacuum projector, the two coordinate components give
τ W ( p T C ⊗ w + ) = w + ⊗ 1 2 ( p t r i v − ϕ ) , \tau_W(p_{\mathrm{TC}} \otimes w_+) = w_+ \otimes\frac{1}{2}(p_{\mathrm{triv}}-\phi), τ W ( p TC ⊗ w + ) = w + ⊗ 2 1 ( p triv − ϕ ) ,
τ W ( p T C ⊗ w − ) = w − ⊗ 1 2 ( p t r i v + ϕ ) . (71) \tau_W(p_{\mathrm{TC}} \otimes w_-) = w_- \otimes\frac{1}{2}(p_{\mathrm{triv}}+\phi). \tag*{(71)} τ W ( p TC ⊗ w − ) = w − ⊗ 2 1 ( p triv + ϕ ) . (71)
The complete crossing, including its values on the remaining channels, is computed in Appendix § 10 .§ 4 .
The two equalities are the coordinate expressions for the two components of a single reversible crossing; they are not two separate crossings. The crossing is not an automorphism of the parent vacuum algebra: it sends a neutral vacuum projector to half the trivial projector, plus or minus half the electric charged channel. The charged term is exactly the information that a description by permuting vacuum labels loses: the labels would record only the neutral parts, whereas the crossing also transports the electric line that must accompany the insertion. An invertible symmetry cannot exchange the two vacua, because they have different intrinsic order; the non-invertible crossing replaces the exchange by a superposition with the charged channel. This is the sense in which no permutation of vacuum labels can express the relation between the two vacua.
Proposition 5.1 states the general consequence.
Proposition 5.1 (Crossings do not permute parent vacua). Fix the operation (68) , let ( B , τ ) (B,\tau) ( B , τ ) be compatible with it, and let f ∈ V B f \in\mathcal{V}_B f ∈ V B be a parent vacuum insertion. In each coordinate component of W W W and W m W_m W m , a parent vacuum insertion crosses without an electric charged contribution only if that component returns it unchanged. Consequently, every parent vacuum insertion not in the compatible vacuum algebra acquires a nonzero electric charged contribution in at least one coordinate component of the two wall excitations, and no crossing of this operation permutes the parent vacuum projectors nontrivially.
The proof is in Appendix § 10 .§ 3 . For this specified operation, the non-permutation statement follows from the particular S 3 S_3 S 3 action and the relations used there. In the mixed datum, the different intrinsic orders additionally exclude an ordinary invertible exchange of the two vacua. The proposition is stronger than that intrinsic-order obstruction: it also applies to b ⊕ 3 b^{\oplus3} b ⊕ 3 , whose parent vacua all have the same intrinsic order. The charged-channel conclusion should therefore not be inferred from differing intrinsic orders, or from non-normality alone.
Finally, the mixed datum is an instance of the commuting square (49) :
Σ D ( Z 2 ) → g K Σ D ( S 3 ) W 1 ⊕ b ↓ ↓ W a K Y 1 ⊕ b → g K 1 ⊕ b Σ D ( Z 2 ) (72) \begin{CD}
\Sigma\mathcal{D}(\mathbb{Z}_2) @>{\mathfrak{g}_K}>> \Sigma\mathcal{D}(S_3) \\
@V{W_{1\oplus b}}VV @VV{W_{a_K}}V \\
\mathcal{Y}_{1\oplus b} @>{\mathfrak{g}_{K_{1\oplus b}}}>> \Sigma\mathcal{D}(\mathbb{Z}_2) \tag*{(72)}
\end{CD} Σ D ( Z 2 ) W 1 ⊕ b ↓ ⏐ Y 1 ⊕ b g K g K 1 ⊕ b Σ D ( S 3 ) ↓ ⏐ W a K Σ D ( Z 2 ) (72)
The right-hand wall is the condensation wall of the reverse condensate. For the mixed crossing displayed in (71) , the coordinate realization S = G / H S=G/H S = G / H in Appendix § 10 .§ 2 identifies the transported condensate with
B K ≃ Fun ( G / H ) = a K . B_K \simeq\operatorname{Fun}(G/H)=a_K. B K ≃ Fun ( G / H ) = a K .
Proposition 4.4 then shows that condensing this algebra returns the toric-code phase:
Z B K l o c ≃ D ( Z 2 ) . \mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(\mathbb{Z}_2). Z B K loc ≃ D ( Z 2 ) .
The induced operation therefore merges the two intermediate vacua, the toric-code vacuum and the trivial one, into this single final vacuum, and the square closes on it.
One operation, many condensations
Classification for the fixed operation
We fix the operation (68) and classify the compatible data it allows, in the language of Sections 3–4.
Theorem 5.2 (Compatible condensations for the fixed operation). Exactly four isomorphism classes of compatible data for the operation (68) have connected transported condensate. For them the parent condensate and final phase are, up to isomorphism,
B = 1 : Z B K l o c ≃ D ( S 3 ) , B = b : Z B K l o c ≃ D ( Z 3 ) , B = 1 ⊕ b : Z B K l o c ≃ D ( Z 2 ) , B = b ⊕ 3 : Z B K l o c ≃ V e c . (73) \begin{aligned}
B &= 1: && \mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(S_3),\\
B &= b: && \mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(\mathbb{Z}_3),\\
B &= 1 \oplus b: && \mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(\mathbb{Z}_2),\\
B &= b^{\oplus3}: && \mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathrm{Vec}. \tag*{(73)}
\end{aligned} B B B B = 1 : = b : = 1 ⊕ b : = b ⊕ 3 : Z B K loc ≃ D ( S 3 ) , Z B K loc ≃ D ( Z 3 ) , Z B K loc ≃ D ( Z 2 ) , Z B K loc ≃ Vec . (73) Each row is realized by exactly one isomorphism class with connected transported condensate. Every other compatible datum is an orthogonal algebra sum of one or more basic data, each isomorphic to one of these four classes. The decomposition into basic data follows the connected components of the transported condensate, and it is unique up to reordering; the connected components of the parent condensate are not to be treated independently in advance.
The four entries are classes of full compatible data, not four parent objects or four endpoint phases: each class retains its crossing and its marked interfaces. The complete coordinate description of the crossing, and the proof of the theorem, are given in Appendix § 10 .§ 2 ; the coordinate classification there is Theorem D.1 .
Residual order from a single parent vacuum
Suppose the specified unit is transparent and the transported condensate B K B_K B K is connected. Write
B = ⨁ i = 1 r B i B = \bigoplus_{i=1}^{r} B_i B = i = 1 ⨁ r B i
for the decomposition into connected components, the parent vacua, and set
D i : = FPdim ( C B i l o c ) , D f i n : = FPdim ( Z B K l o c ) . (74) \mathcal{D}_i := \sqrt{\operatorname{FPdim}\left(\mathcal{C}^{\mathrm{loc}}_{B_i}\right)}, \qquad
\mathcal{D}_{\mathrm{fin}} := \sqrt{\operatorname{FPdim}\left(\mathcal{Z}^{\mathrm{loc}}_{B_K}\right)}. \tag*{(74)} D i := FPdim ( C B i loc ) , D fin := FPdim ( Z B K loc ) . (74)
Here D i \mathcal{D}_i D i is the total quantum dimension of the residual intrinsic order in the i i i -th parent vacuum. Similarly, D f i n \mathcal{D}_{\mathrm{fin}} D fin is the total quantum dimension of the final intrinsic order. Lemma B.2 gives
D f i n = κ K ∑ i = 1 r D i − 1 ≥ κ K r , (75) \mathcal{D}_{\mathrm{fin}} = \frac{\kappa_K}{\sum_{i=1}^{r} \mathcal{D}_i^{-1}} \geq\frac{\kappa_K}{r}, \tag*{(75)} D fin = ∑ i = 1 r D i − 1 κ K ≥ r κ K , (75)
with equality exactly when every parent vacuum is intrinsically trivial.
If the parent has a single vacuum, its vacuum algebra is one-dimensional and spanned by the unit. The unit is compatible with every crossing, so the compatible vacuum algebra is the same one-dimensional algebra, and Corollary 4.5 gives that B K B_K B K is connected. For the present operation, every compatible condensation with a single parent vacuum therefore ends in a phase of total quantum dimension at least κ K = 3 \kappa_K = 3 κ K = 3 , however completely it removes the toric-code anyons.
The bound is saturated by the unique compatible electric condensation. Take B = b B=b B = b . The intermediate phase is intrinsically trivial, the final phase is the untwisted Z 3 \mathbb{Z}_3 Z 3 gauge theory, and the vacuum data and final dimension are
r = r c o m p = 1 , Ω Y b ≃ V e c , Z B K l o c ≃ D ( Z 3 ) , D f i n = 3. (76) r=r_{\mathrm{comp}}=1,\qquad\Omega_{\mathcal{Y}b}\simeq\mathrm{Vec},\qquad\mathcal{Z}^{\mathrm{loc}}_{B_K}\simeq\mathcal{D}(\mathbb{Z}_3),\qquad D_{\mathrm{fin}}=3. \tag*{(76)} r = r comp = 1 , Ω Y b ≃ Vec , Z B K loc ≃ D ( Z 3 ) , D fin = 3. (76)
Here the wall excitation W W W crosses the unit and the electric channel of b b b without changing coefficients. The crossing through W m W_m W m contains the fixed magnetic crossing, which reverses the sign of the electric channel and carries no compatibility information. The unbroken Z 3 \mathbb{Z}_3 Z 3 becomes a deconfined gauge field once the gauging is completed, which is why the final phase has dimension three. Inequality (75) turns the observation that a trivial endpoint before gauging need not remain trivial afterwards into a bound.
The complete Higgs class
There is exactly one complete-Higgs class for the fixed operation. Its parent condensate is b ⊕ 3 b^{\oplus3} b ⊕ 3 : three parent vacua, each of which is an intrinsically trivial electric-condensation vacuum. The crossing connects the three parent vacuum sectors through their charged channels; the explicit mixing is displayed in § 5.4 . The transported condensate is connected, and the parent vacuum count equals κ K \kappa_K κ K . Each parent component is the electric condensate b b b , which is Lagrangian in C \mathcal{C} C ; by Corollary 4.5 the datum is of complete Higgs type, and its final phase is the single intrinsically trivial vacuum. Conversely, the classification theorem shows that no other class is of complete Higgs type: the remaining classes end in nontrivial gauge theories, not in a trivial vacuum.
The three complete-Higgs conditions are independent. The first two failures are realized for this operation, and the third is realized after a change of parent phase and operation:
The electric condensation B = b B=b B = b has a connected transported condensate and a Lagrangian parent component, but only one parent vacuum, so the parent vacuum count fails.
The crossing of b ⊕ 3 b^{\oplus3} b ⊕ 3 that preserves its three parent branches has three Lagrangian parent components and the correct parent vacuum count, but its transported condensate is disconnected.
Tensoring a complete-Higgs datum with a nontrivial spectator through (C.9) preserves connectedness and both vacuum counts and preserves the crossing, but every parent component ceases to be Lagrangian; this changes the parent phase and the operation, so it is not a third counterexample inside the fixed operation.
One condensate, two crossings
We now fix the parent condensate, the marked condensation wall, the intermediate defect theory, and the crossing of every parent anyon, which is fixed by the ambient action (28) . The two data of a pair share all of this common data. They can differ only on the two wall excitations W W W and W m W_m W m , which are not parent anyons. The crossing of W m W_m W m follows from that of W W W by the tensor identity and the fixed magnetic crossing, so it is not selected independently.
The possible crossings are classified as follows.
Proposition 5.3 (Crossings of a fixed parent condensate). The condensates 1 1 1 and b b b admit exactly one compatible crossing class each. The condensate 1 ⊕ b 1\oplus b 1 ⊕ b admits exactly two compatible crossing classes, and b ⊕ 3 b^{\oplus3} b ⊕ 3 admits exactly two.
The proof is in Appendix § 10 .§ 3 .
Two crossings of the mixed condensate
The first crossing is the charged-mixing crossing of § 5.1 : the two coordinate components of W W W send the toric-code vacuum projector as in (71) . In this crossing the unit is the only compatible parent vacuum operator, the transported condensate is connected, and the final phase is a single toric-code vacuum. The second crossing leaves the coefficients of every channel unchanged: crossing W W W only interchanges the two tensor factors, so each channel crosses by the flip
τ W ( z ⊗ w ± ) = w ± ⊗ z . \tau_W(z \otimes w_\pm) = w_\pm\otimes z. τ W ( z ⊗ w ± ) = w ± ⊗ z .
This is the crossing of W W W . The crossing of W m W_m W m is the fixed magnetic sign crossing, which fixes the neutral channels and reverses the sign of the electric one. It therefore preserves both parent vacuum projectors. The transported condensate is the orthogonal sum of the transported data of the two isolated branches 1 1 1 and b b b : the first branch carries the toric-code vacuum and the second the trivial vacuum, and after transport they remain separate. The two final phases are
first crossing : Z B K l o c ≃ D ( Z 2 ) , second crossing : Z B K l o c ≃ D ( S 3 ) ⊕ D ( Z 3 ) . (77) \begin{aligned}
\text{first crossing} &: \quad\mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(\mathbb{Z}_2),\\
\text{second crossing} &: \quad\mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(S_3) \oplus\mathcal{D}(\mathbb{Z}_3). \tag*{(77)}
\end{aligned} first crossing second crossing : Z B K loc ≃ D ( Z 2 ) , : Z B K loc ≃ D ( S 3 ) ⊕ D ( Z 3 ) . (77)
Because the second crossing preserves both parent vacuum projectors, the two intermediate vacua are not related by it: their coexistence is not enforced by the symmetry, they remain separate final vacua, and the compatible vacuum count is two.
Two crossings of three electric-condensation vacua
Consider the parent condensate B = b ⊕ 3 B = b^{\oplus3} B = b ⊕ 3 of the complete-Higgs class. Write p a p_a p a , ϕ a \phi_a ϕ a for the unit and electric channel of the a a a -th connected component, with the index taken modulo three. Products between channels in different components vanish, and within each component the multiplication is that of b b b . The first crossing mixes the three parent branches through the charged channels. In the coordinate component w + w_+ w + of W W W it acts by
τ W ( p a ⊗ w + ) = w + ⊗ 1 2 ( p a − 1 + p a + 1 + ϕ a − 1 − ϕ a + 1 ) . (78) \tau_W(p_a \otimes w_+) = w_+ \otimes\frac{1}{2}\left(p_{a-1} + p_{a+1} + \phi_{a-1} - \phi_{a+1}\right). \tag*{(78)} τ W ( p a ⊗ w + ) = w + ⊗ 2 1 ( p a − 1 + p a + 1 + ϕ a − 1 − ϕ a + 1 ) . (78)
The companion formula in the w − w_- w − component has the opposite signs on the two charged channels. Each parent vacuum projector is sent to half the sum of the other two, plus half the difference of their charged channels; hence the crossing connects the three parent sectors, and the unit is the only compatible parent vacuum operator. The transported condensate is connected, and the final phase is a single intrinsically trivial vacuum.
In the second crossing the coefficients in each component are unchanged when W W W is crossed. Crossing W W W interchanges the two tensor factors without mixing components, so each channel crosses by the flip
τ W ( z ⊗ w ± ) = w ± ⊗ z . \tau_W(z \otimes w_\pm) = w_\pm\otimes z. τ W ( z ⊗ w ± ) = w ± ⊗ z .
On W m W_m W m the crossing is the fixed magnetic sign crossing in each component. The transported condensate is the orthogonal sum of the three transported components, each of which is the electric condensation, and the final phase has three gauge vacua:
first crossing : Z B K l o c ≃ V e c , second crossing : Z B K l o c ≃ D ( Z 3 ) ⊕ 3 . (79) \begin{aligned}
\text{first crossing} &: \quad\mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathrm{Vec},\\
\text{second crossing} &: \quad\mathcal{Z}^{\mathrm{loc}}_{B_K} \simeq\mathcal{D}(\mathbb{Z}_3)^{\oplus3}. \tag*{(79)}
\end{aligned} first crossing second crossing : Z B K loc ≃ Vec , : Z B K loc ≃ D ( Z 3 ) ⊕ 3 . (79)
In both data the electric charge condenses in each of the three intrinsically trivial parent vacua, yet only the first crossing is of complete Higgs type. The combined comparison of the two pairs is
B = 1 ⊕ b : D ( Z 2 ) versus D ( S 3 ) ⊕ D ( Z 3 ) , B = b ⊕ 3 : Vec versus D ( Z 3 ) ⊕ 3 . (80) \begin{aligned}
B &= 1 \oplus b : \mathcal{D}(\mathbb{Z}_2) \quad\text{versus} \quad\mathcal{D}(S_3) \oplus\mathcal{D}(\mathbb{Z}_3),\\
B &= b^{\oplus3} : \operatorname{Vec} \quad\text{versus} \quad\mathcal{D}(\mathbb{Z}_3)^{\oplus3}. \tag*{(80)}
\end{aligned} B B = 1 ⊕ b : D ( Z 2 ) versus D ( S 3 ) ⊕ D ( Z 3 ) , = b ⊕ 3 : Vec versus D ( Z 3 ) ⊕ 3 . (80)
Whether two parent vacua are related is thus a property of the crossing, not of the parent condensate; and complete Higgs behavior is decided neither by whether the electric charge condenses nor by the parent condensate alone, but by how the full condensate crosses the specified network.
In the language of Theorem 3.4 , the two squares of a pair share the left edge and the lower-left corner: the parent phase, the condensation wall, and the intermediate defect theory are the same. They differ in the crossing, hence in the composite algebra T λ T_\lambda T λ , and therefore in the transported condensate and in the induced operation. Their right and lower walls, and their lower-right endpoint phases, differ. The upper-right phase remains fixed, since the specified operation K K K is unchanged. The induced operation is therefore not determined by the intermediate endpoint, and it is not chosen independently at the lower-left corner. This is the distinction, raised in the introduction, between an equivalence of endpoints and an equivalence of specified composite walls.
One transported condensate, two parent condensates
Absorption by the specified unit
The specified unit need not be transparent. The operation used here is the pure electric condensation of the toric code,
C = D ( Z 2 ) , b = 1 ⊕ e , K = Σ b , (81) C = \mathcal{D}(\mathbb{Z}_2), \qquad b = 1 \oplus e, \qquad K = \Sigma_b, \tag*{(81)} C = D ( Z 2 ) , b = 1 ⊕ e , K = Σ b , (81)
with K K K given its free-module unit. Then
K = C b , ι = F b , Z = C b l o c ≃ Vec , q : C b l o c ↪ C b , \mathcal{K} = C_b, \qquad\iota= F_b, \qquad\mathcal{Z} = C_b^{\mathrm{loc}} \simeq\operatorname{Vec}, \qquad q : C_b^{\mathrm{loc}} \hookrightarrow C_b, K = C b , ι = F b , Z = C b loc ≃ Vec , q : C b loc ↪ C b ,
d K = b , δ K = 2 , a K = 1 , κ K = 1 , M K = C b l o c ≃ Vec . (82) d_K = b, \qquad\delta_K = 2, \qquad a_K = 1, \qquad\kappa_K = 1, \qquad\mathcal{M}_K = C_b^{\mathrm{loc}} \simeq\operatorname{Vec}. \tag*{(82)} d K = b , δ K = 2 , a K = 1 , κ K = 1 , M K = C b loc ≃ Vec . (82)
The wall category has two simple wall objects: the regular unit b b b and the magnetic object m ⊗ b m \otimes b m ⊗ b . The unit is simple, so K K K is connected. The algebra it absorbs is d K = b d_K = b d K = b , so the unit is not transparent and the operation is not strongly connected.
Physically, proliferating the electric-condensation surface Σ b \Sigma_b Σ b condenses the electric charge in the bulk. The operation K K K is therefore the electric bulk-condensation operation itself: applying it is the condensation wall of the electric condensate. The example is related to the running one. The S 3 / Z 2 S_3/\mathbb{Z}_2 S 3 / Z 2 operation decomposes as the transparent surface defect plus the nontransparent summand Σ b \Sigma_b Σ b ; the present operation isolates that summand, equipped with its own free-module unit. The algebra structures of the two operations are related but not identified.
The three smallest cases below exercise statements that the transparent-unit examples cannot.
No parent condensation: a complete Higgs endpoint
For B = 1 B = 1 B = 1 there is no parent anyon condensation. The parent vacuum keeps the full toric-code order, while the specified operation itself supplies the condensation: the final phase is a single intrinsically trivial vacuum. The vacuum data are
r = r c o m p = 1 , n K ( 1 ) = r f i n = 1 , D 1 = δ K = 2. (83) r = r_{\mathrm{comp}} = 1, \qquad n_K(1) = r_{\mathrm{fin}} = 1, \qquad\mathcal{D}_1 = \delta_K = 2. \tag*{(83)} r = r comp = 1 , n K ( 1 ) = r fin = 1 , D 1 = δ K = 2. (83)
The datum is of complete Higgs type: the transported condensate is connected, the branch number equals κ K = 1 \kappa_K = 1 κ K = 1 , and the reduced condensate B ‾ = F b ( 1 ) \overline{B} = F_b(1) B = F b ( 1 ) is the unit of the reduced bulk, hence Lagrangian; condition (iii) of Theorem 4.3 therefore holds. The original parent condensate is not Lagrangian: nothing condenses in the parent phase, and the single parent vacuum has residual order D 1 = 2 \mathcal{D}_1 = 2 D 1 = 2 . The object tested by Theorem 4.3 is the reduced condensate after unit absorption, not the original parent algebra, and the Lagrangian condition of Corollary 4.5 fails while the general criterion holds.
One parent vacuum, two final vacua
For B = b B=b B = b the final vacuum count is strictly larger than the parent one:
r = r c o m p = 1 , n K ( b ) = r f i n = 2. (84) r = r_{\mathrm{comp}} = 1,\qquad n_K(b) = r_{\mathrm{fin}} = 2. \tag*{(84)} r = r comp = 1 , n K ( b ) = r fin = 2. (84)
A basis of the wall junctions Hom K ( 1 K , ι ( b ) ) \operatorname{Hom}_K(1_K,\iota(b)) Hom K ( 1 K , ι ( b )) consists of two elements. The first is the image u u u of the parent condensate unit. The second is the neutral channel z z z that pairs the electric channel of the condensate with the electric channel absorbed by the unit. They are normalized so that
u † u = z † z = 2 , z 2 = u . (85) u^\dagger u = z^\dagger z = 2,\qquad z^2 = u. \tag*{(85)} u † u = z † z = 2 , z 2 = u . (85)
The transported condensate is L = C 2 L = \mathbb{C}^2 L = C 2 , and the final vacuum algebra is the pointwise product algebra on these two points. Its primitive idempotents δ + , δ − \delta_+,\delta_- δ + , δ − are the two final vacua, written before any matching is chosen. The matching identifies them with the two spectral projectors of the neutral channel,
ξ ( δ ± ) = 1 2 ( u ± z ) . (86) \xi(\delta_\pm) = \frac{1}{2}(u \pm z). \tag*{(86)} ξ ( δ ± ) = 2 1 ( u ± z ) . (86)
The operators ξ ( δ ± ) \xi(\delta_\pm) ξ ( δ ± ) are projectors onto the two values of z z z , not eigenvalues. Equivalently,
z ξ ( δ ± ) = ± ξ ( δ ± ) . z\,\xi(\delta_\pm) = \pm\xi(\delta_\pm). z ξ ( δ ± ) = ± ξ ( δ ± ) .
The two values are opposite, so z z z records a relative sign. Physically, the parent condensate and the algebra absorbed by the unit each contain an electric condensate. Neither is neutral on its own, but their product is: z z z is a neutral order parameter that neither condensate sees by itself. On the final side z z z is a neutral vacuum operator that distinguishes the two final vacua. It is not an insertion of an original parent vacuum, which is why the parent and final vacuum counts differ. The crossing is reconstructed from ξ \xi ξ by (32) .
In general, each parent vacuum in which the electric charge has condensed splits into two final vacua, and each toric-code vacuum contributes one final vacuum. This splitting is the concrete content of the branch number n K ( B ) n_K(B) n K ( B ) of Lemma 4.1 .
The same transported algebra from two parent condensates
The data B = b B=b B = b and B = 1 ⊕ 1 B=1\oplus1 B = 1 ⊕ 1 share their transported condensate and their final phase. Both have transported condensate L = C 2 L=\mathbb{C}^2 L = C 2 , final phase V e c ⊕ 2 \mathrm{Vec}^{\oplus2} Vec ⊕ 2 , and the same final vacuum count. They have different parent condensates and different condensation walls: the intermediate phase is V e c \mathrm{Vec} Vec in the first case and D ( Z 2 ) \mathcal{D}(\mathbb{Z}_2) D ( Z 2 ) in the second. Their matchings also differ: charged-pair matching for the first, identity matching for the second. In the first, the two final vacua are distinguished by the product of two charged channels; in the second, by a neutral order already present in the parent phase. Each of the two toric-code branches of the second parent condensate contributes one intrinsically trivial final vacuum, because the specified unit absorbs its whole electric condensate; no toric-code vacuum survives the operation. The comparison is
B = b : Ω Y B ≃ V e c , Z L l o c ≃ V e c ⊕ 2 , B=b:\quad\Omega_{\mathcal{Y}B}\simeq\mathrm{Vec},\qquad\mathcal{Z}^{\mathrm{loc}}_{L}\simeq\mathrm{Vec}^{\oplus2}, B = b : Ω Y B ≃ Vec , Z L loc ≃ Vec ⊕ 2 ,
B = 1 ⊕ 1 : Ω Y B ≃ D ( Z 2 ) ⊕ 2 , Z L l o c ≃ V e c ⊕ 2 . (87) B=1\oplus1:\quad\Omega_{\mathcal{Y}B}\simeq\mathcal{D}(\mathbb{Z}_{2})^{\oplus2},\qquad\mathcal{Z}^{\mathrm{loc}}_{L}\simeq\mathrm{Vec}^{\oplus2}. \tag*{(87)} B = 1 ⊕ 1 : Ω Y B ≃ D ( Z 2 ) ⊕ 2 , Z L loc ≃ Vec ⊕ 2 . (87)
For a transparent unit this cannot happen, since Proposition 4.4 gives B = F a K ( L ) B=F_{a_K}(L) B = F a K ( L ) . This is why central matching retains the parent algebra and its identification on the wall, and not the transported condensate alone.
Classification and vacuum counts
Proposition 5.4 (Compatible data for a pure condensation). Isomorphism classes of compatible data for (81) are in bijection with pairs ( k , l ) ∈ Z ≥ 0 2 (k,l)\in\mathbb{Z}_{\geq0}^{2} ( k , l ) ∈ Z ≥ 0 2 satisfying k + l > 0 k+l>0 k + l > 0 . The corresponding datum has
B = 1 ⊕ k ⊕ b ⊕ l , B K ≃ C k + 2 l , B=1^{\oplus k}\oplus b^{\oplus l},\qquad B_K\simeq\mathbb{C}^{k+2l}, B = 1 ⊕ k ⊕ b ⊕ l , B K ≃ C k + 2 l , where C k + 2 l \mathbb{C}^{k+2l} C k + 2 l is the pointwise product algebra on k + 2 l k+2l k + 2 l points, and its vacuum numbers are
r = r c o m p = k + l , n K ( B ) = r f i n = k + 2 l . (88) r=r_{\mathrm{comp}}=k+l,\qquad n_K(B)=r_{\mathrm{fin}}=k+2l. \tag*{(88)} r = r comp = k + l , n K ( B ) = r fin = k + 2 l . (88)
Proof. Since Z ≃ V e c \mathcal{Z}\simeq\mathrm{Vec} Z ≃ Vec , the object q ( L ) q(L) q ( L ) is a direct sum of copies of 1 K 1_K 1 K . Theorem 3.2 therefore requires ι ( B ) = F b ( B ) \iota(B)=F_b(B) ι ( B ) = F b ( B ) to have the same property. The free-module functor satisfies
F b ( 1 ) ≃ F b ( e ) ≃ b , F b ( m ) ≃ F b ( e ⊗ m ) ≃ m ⊗ b . F_b(1)\simeq F_b(e)\simeq b,\qquad F_b(m)\simeq F_b(e\otimes m)\simeq m\otimes b. F b ( 1 ) ≃ F b ( e ) ≃ b , F b ( m ) ≃ F b ( e ⊗ m ) ≃ m ⊗ b .
The magnetic wall object m ⊗ b m\otimes b m ⊗ b is not the tensor unit. Semisimplicity consequently forces B B B to lie in the electric subcategory ⟨ e ⟩ ≃ Rep ( Z 2 ) \langle e\rangle\simeq\operatorname{Rep}(\mathbb{Z}_2) ⟨ e ⟩ ≃ Rep ( Z 2 ) .
Under its ordinary vector-space realization, B B B is a finite-dimensional commutative separable algebra equipped with a Z 2 \mathbb{Z}_2 Z 2 -action by algebra automorphisms. Equivalently, the generator acts as + 1 +1 + 1 on the neutral part and as − 1 -1 − 1 on the electric part; equivariance of the multiplication and unit makes this an algebra involution. It permutes the primitive idempotents of the underlying ordinary algebra. A fixed idempotent contributes the algebra 1 1 1 , while a pair exchanged by the involution contributes the electric condensate b b b . Thus
B ≃ 1 ⊕ k ⊕ b ⊕ l , k , l ≥ 0 , k + l > 0. B\simeq1^{\oplus k}\oplus b^{\oplus l},\qquad k,l\geq0,\qquad k+l>0. B ≃ 1 ⊕ k ⊕ b ⊕ l , k , l ≥ 0 , k + l > 0.
These primitive idempotents are those of the underlying ordinary algebra; only the indicators of their Z 2 \mathbb{Z}_2 Z 2 -orbits are neutral parent vacuum operators.
Conversely, the calculation in (86) gives F b ( b ) ≃ q ( C 2 ) F_b(b)\simeq q(\mathbb{C}^{2}) F b ( b ) ≃ q ( C 2 ) as algebras, while F b ( 1 ) ≃ q ( C ) F_b(1)\simeq q(\mathbb{C}) F b ( 1 ) ≃ q ( C ) . The additive free-module functor preserves orthogonal algebra sums, so
F b ( B ) ≃ q ( C k + 2 l ) F_b(B)\simeq q(\mathbb{C}^{k+2l}) F b ( B ) ≃ q ( C k + 2 l )
as algebras. This supplies the matching required by Theorem 3.2. For a fixed parent algebra, two unitary matchings differ by a unitary algebra automorphism of the final pointwise algebra. Hence each parent algebra carries exactly one isomorphism class of compatible data.
The parent vacuum count is r = k + l r=k+l r = k + l , and
n K ( B ) = dim Hom C ( 1 ⊕ e , B ) = k + 2 l . n_K(B)=\dim\operatorname{Hom}_{\mathcal{C}}(1\oplus e,B)=k+2l. n K ( B ) = dim Hom C ( 1 ⊕ e , B ) = k + 2 l .
Since a K = 1 a_K=1 a K = 1 , equation (56) identifies this number with r f i n r_{\mathrm{fin}} r fin .
Finally, the central lift j Z ( L ) j_{\mathcal{Z}}(L) j Z ( L ) is a direct sum of copies of the central tensor unit. Every morphism from 1 K \mathbf{1}_K 1 K to q ( L ) q(L) q ( L ) is therefore central. Transporting this structure through ξ \xi ξ makes every ι ( v ) \iota(v) ι ( v ) , with v ∈ V B v \in V_B v ∈ V B , a central morphism for the crossing. Thus every parent vacuum insertion is compatible, and r c o m p = r = k + l r_{\mathrm{comp}}=r=k+l r comp = r = k + l .
Discussion
The results of this work concern condensation compatible with a specified gauging operation. This choice is important: compatibility retains the full condensate network and its marked interface, while complete Higgs behavior tests the intrinsic phase obtained after gauging the induced network. The higher-categorical perspective clarifies both the scope of these statements and several directions in which the framework can be extended.
Symmetry categories and gauging algebras. A finite generalized symmetry in ( 2 + 1 ) (2+1) ( 2 + 1 ) dimensions is more generally described by a fusion 2-category S \mathcal{S} S of symmetry defects and their junctions, together with a monoidal 2-functor
ρ : S ⟶ C \rho:\mathcal{S}\longrightarrow\mathcal{C} ρ : S ⟶ C
realizing those defects in the chosen parent theory. An appropriate relative defect theory can be used instead when additional bulk data must be retained. The distinction between an abstract symmetry and its realization is essential: fusion rules alone do not determine the junction coherences or the action on the physical theory [20 , 21 ] .
A gauging requires a further choice. In the separable algebra presentation, one selects
k ∈ Alg E 1 s e p ( S ) , K = ρ ( k ) . k \in\operatorname{Alg}^{\mathrm{sep}}_{\mathcal{E}_1}(\mathcal{S}), \qquad K = \rho(k). k ∈ Alg E 1 sep ( S ) , K = ρ ( k ) .
The algebra specifies which defect networks are proliferated and how their junctions are weighted and composed. The associated dual defect category is described by Bimod S ( k ) \operatorname{Bimod}_{\mathcal{S}}(k) Bimod S ( k ) , whereas the realized gauged phase is described by Bimod C ( K ) \operatorname{Bimod}_{\mathcal{C}}(K) Bimod C ( K ) [20 , 21 ] . For an ordinary untwisted group gauging, the underlying object of K K K is the sum of the labelled symmetry walls, but its algebra structure also retains their specified fusion junctions.
We have worked directly with the realized algebra K K K . It is a specified gauging datum, not an unconditional replacement for the entire symmetry category S \mathcal{S} S . Different algebras in the same symmetry category can specify different gaugings, and compatibility with one selected algebra need not imply compatibility with every symmetry defect. Likewise, the fusion category K = Hom C ( 1 C , K ) \mathcal{K} = \operatorname{Hom}_{\mathcal{C}}(\mathbbm{1}_{\mathcal{C}}, K) K = Hom C ( 1 C , K ) describes line excitations on the selected gauging wall; it should not be confused with the symmetry fusion 2-category. The markings retained throughout this work distinguish the specified operation from a comparison of unmarked endpoint theories alone.
The existence of a separable gauging algebra should not by itself be interpreted as gaugeability of the entire symmetry: the chosen operation may gauge only a proper sub-symmetry. Here, full gauging means that the associated module 2-category is equivalent to 2 V e c 2\mathrm{Vec} 2 Vec , so that the gauging data realize a fiber 2-functor S → 2 V e c \mathcal{S} \to2\mathrm{Vec} S → 2 Vec . The absence of a full gauging does not exclude partial gaugings.
Compatibility with an entire symmetry category. The absence of an algebra implementing a desired gauging does not prevent one from asking whether a prescribed condensation is compatible with the symmetry action. Here absence means the absence of the required gauging datum, not the absence of all separable algebras: the tensor unit still gives a trivial operation. For example, in the ordinary-group symmetry category 2 V e c G α 2\mathrm{Vec}^{\alpha}_{G} 2 Vec G α —the fusion 2-category of finite semisimple G G G -graded unitary 2-vector spaces whose pentagonator is twisted by α \alpha α —with α ∈ Z 4 ( G , U ( 1 ) ) \alpha\in Z^{4}(G, U(1)) α ∈ Z 4 ( G , U ( 1 )) , the standard full-group gauging algebra requires a 3-cochain γ \gamma γ trivializing the associator obstruction, δ γ = α \delta\gamma= \alpha δ γ = α . A nontrivial obstruction excludes this gauging, although a suitable subgroup may still admit one [31 Sec. 3.3] .
A natural gauging-independent extension of our reversible notion of compatibility is an equivariant structure on the marked condensation itself. For A = Σ B A = \Sigma B A = Σ B , one would require coherent equivalences
Λ s : ρ ( s ) ⊗ A → ∼ A ⊗ ρ ( s ) , s ∈ S , \Lambda_s \colon\rho(s) \otimes A \xrightarrow{\sim} A \otimes\rho(s), \qquad s \in\mathcal{S}, Λ s : ρ ( s ) ⊗ A ∼ A ⊗ ρ ( s ) , s ∈ S ,
natural with respect to all symmetry junctions and higher morphisms. Suppressing the tensorators of ρ \rho ρ , compatibility with symmetry fusion has the form
Λ s ⊗ t ≃ ( Λ s ⊗ id ρ ( t ) ) ( id ρ ( s ) ⊗ Λ t ) . \Lambda_{s\otimes t} \simeq(\Lambda_s \otimes\operatorname{id}_{\rho(t)})(\operatorname{id}_{\rho(s)} \otimes\Lambda_t). Λ s ⊗ t ≃ ( Λ s ⊗ id ρ ( t ) ) ( id ρ ( s ) ⊗ Λ t ) .
The family must also preserve the multiplication of A A A and include the marked unit comparisons
Λ s ( id ρ ( s ) ⊗ η A ) ≃ η A ⊗ id ρ ( s ) , \Lambda_s(\operatorname{id}_{\rho(s)} \otimes\eta_A) \simeq\eta_A \otimes\operatorname{id}_{\rho(s)}, Λ s ( id ρ ( s ) ⊗ η A ) ≃ η A ⊗ id ρ ( s ) ,
together with their higher coherences. As in the present definition, reversibility of these comparisons does not require the symmetry defects themselves to be invertible.
Categorically, such data ask for the symmetry action to lift to the module object of the condensation monad. The relationship between distributive laws and liftings is already present in Beck’s theory and has a precise pseudomonad formulation [32 , 33 , 34 ] . For ordinary groups, its anyon-level realization is a G G G -equivariant algebra structure on B B B , rather than mere invariance of its object decomposition [8 ] .
When a gauging algebra k k k is available, restricting the full symmetry-level data to k k k yields the marked law for K K K considered here; the converse need not hold for an arbitrary choice of k k k .
The restriction should not, however, be confused with an unconditional equivalence between compatibility with S S S and compatibility with one chosen gauging algebra K K K . The full symmetry-level data restrict to the marked K K K -compatible data studied here, whereas the converse requires an extension. Extending a K K K -compatible datum to the entire symmetry category is a separate reconstruction problem. Such a reconstruction must control the generating junctions and their higher coherences, not merely the underlying defect objects. For the standard full-group gauging, Appendix § 9 supplies the correspondence with G G G -equivariant condensation. For a subgroup gauging, extension from the subgroup to the full group remains an additional question. Thus completeness of the gauging and extendability of a particular marked compatibility datum are distinct questions.
This extension would not require a symmetry-gauging algebra, although the condensate must still possess its own condensation structure. Nor does it guarantee compatible data for a prescribed B B B : an anomaly can enter the coherence problem, and a relative formulation must retain the corresponding inflow data. Gaugeability and compatibility with a unique symmetric trivially gapped phase must also be distinguished for non-invertible symmetries [22 ] . Without a specified gauging, the symmetry-level question remains meaningful, but the gauged condensate B K B_K B K , the reverse condensate a K a_K a K , and the complete Higgs test of this work are not defined by that question alone.
Non-simple units and multifusion extensions. The central matching description of § 3.3 assumes that 1 K \mathbb{1}_K 1 K is simple, so that K \mathcal{K} K is a fusion category. For a nontransparent specified unit, d K d_K d K need not be 1 C \mathbb{1}_{\mathcal{C}} 1 C , and the local reduction of § 4 records its effect. What is not covered is a nonsimple tensor unit. Then K \mathcal{K} K is only multifusion: its unit decomposes into several simple objects, and condensation data must be compared with the resulting block structure and its off-diagonal sectors retained. The fusion-category formulas of this paper therefore cannot be applied vacuum by vacuum, since that would discard those sectors and the coherence data relating them. The composite network and the marked wall square remain meaningful once the separability hypotheses hold, but the dimension and vacuum counts require a blockwise treatment not given here.
Higher dimensions and categorical levels. The structural mechanism behind compatible condensation is not specific to anyons or to ( 2 + 1 ) (2+1) ( 2 + 1 ) dimensions. Higher condensation theory describes condensations by condensation monads and their module objects; under appropriate adjointness and completeness hypotheses, these admit separable unital presentations [28 , 29 ] . A homotopy-coherent distributive law is then the appropriate interchange datum between the relevant condensation structures. For condensations at another defect level, the same question is posed in the corresponding endomorphism higher category, with the required monoidal structures and higher coherences retained.
Where the module objects, relative products, and separability conditions are available, the two orders of condensation can be compared through their common composite algebra. This is the higher-categorical content of the marked wall square. Higher Morita theory provides constructions of algebras and iterated bimodules at arbitrary categorical height, with explicit hypotheses ensuring the relative tensor products needed for composition [37 ] . A physical realization also requires suitable local topological consistency. For example, generalized orbifolds of defect TQFTs can be constructed in arbitrary spacetime dimension once the orbifold data satisfy the corresponding Pachner-move constraints [38 ] .
Thus the dimension-independent statement concerns the framework, not an unchanged list of low-dimensional equations. At higher categorical levels, coherence equalities are replaced by higher comparison morphisms and their own consistency conditions. An isolated crossing equivalence, or the ordinary distributive-law axioms without their higher completion, is not sufficient. In this work the anyon-condensation realization makes the data explicit: the algebra B B B , its charged channels, the crossing τ \tau τ , the marked suspension, and the comparison of specified walls. The vacuum-overlap formula, residual-order identity, and Lagrangian criterion additionally use the modular-category dimension theory of ( 2 + 1 ) (2+1) ( 2 + 1 ) -dimensional topological orders. Their analogues in higher dimensions require further input beyond the formal interchange construction.
These extensions separate several questions for future work: compatibility with an entire symmetry category rather than one gauging algebra; relative formulations retaining anomalies; condensations with a nonsimple specified unit; and higher-dimensional quantitative constraints. The present framework remains a statement about specified gapped operations in a finite semisimple defect realization. It does not by itself establish a microscopic implementation, the continuity of a transition, or a critical universality class. Those dynamical questions require information not contained in the marked condensation networks alone.
Marked crossings and transport proofs
Line pull-through and the marked surface-defect law
We use the notation
B ~ = ι ( B ) , m ~ B = ι ( m B ) , u ~ B = ι ( u B ) . \widetilde{B}=\iota(B),\qquad\widetilde{m}_{B}=\iota(m_{B}),\qquad\widetilde{u}_{B}=\iota(u_{B}). B = ι ( B ) , m B = ι ( m B ) , u B = ι ( u B ) .
We suppress ambient associators and unitors; ∗ \ast ∗ denotes the indicated whiskering of 2-morphisms.
Proposition A.1 (Line–surface-defect correspondence). Fix the unitary separable module-object realization A = Σ B A=\Sigma B A = Σ B , including its suspension multiplication, the adjunction η A ⊣ η A R \eta_A\dashv\eta_A^{R} η A ⊣ η A R , and the monad identification (29) . Full line pull-through data of Definition 3.1 are equivalent to unitary marked pseudodistributive laws λ : K ⊗ A → A ⊗ K \lambda:K\otimes A\to A\otimes K λ : K ⊗ A → A ⊗ K . The equivalence retains the specified free interfaces and their adjunctions.
Surface-defect coherence data
The structural 2-isomorphisms are
χ K : λ ( μ K ⊗ id A ) ⇒ ≅ ( id A ⊗ μ K ) ( λ ⊗ id K ) ( id K ⊗ λ ) , (93) \chi_K:\lambda(\mu_K\otimes\operatorname{id}_A)\xRightarrow{\cong}(\operatorname{id}_A\otimes\mu_K)(\lambda\otimes\operatorname{id}_K)(\operatorname{id}_K\otimes\lambda), \tag*{(93)} χ K : λ ( μ K ⊗ id A ) ≅ ( id A ⊗ μ K ) ( λ ⊗ id K ) ( id K ⊗ λ ) , (93)
χ A : λ ( id K ⊗ μ A ) ⇒ ≅ ( μ A ⊗ id K ) ( id A ⊗ λ ) ( λ ⊗ id A ) , (94) \chi_A:\lambda(\operatorname{id}_K\otimes\mu_A)\xRightarrow{\cong}(\mu_A\otimes\operatorname{id}_K)(\operatorname{id}_A\otimes\lambda)(\lambda\otimes\operatorname{id}_A), \tag*{(94)} χ A : λ ( id K ⊗ μ A ) ≅ ( μ A ⊗ id K ) ( id A ⊗ λ ) ( λ ⊗ id A ) , (94)
ν K : λ ( η K ⊗ id A ) ⇒ ≅ id A ⊗ η K , (95) \nu_K:\lambda(\eta_K\otimes\operatorname{id}_A)\xRightarrow{\cong}\operatorname{id}_A\otimes\eta_K, \tag*{(95)} ν K : λ ( η K ⊗ id A ) ≅ id A ⊗ η K , (95)
α λ : λ ( id K ⊗ η A ) ⇒ ≅ η A ⊗ id K . (96) \alpha_\lambda:\lambda(\operatorname{id}_K\otimes\eta_A)\xRightarrow{\cong}\eta_A\otimes\operatorname{id}_K. \tag*{(96)} α λ : λ ( id K ⊗ η A ) ≅ η A ⊗ id K . (96)
The notation χ K , χ A \chi_K,\chi_A χ K , χ A is reserved for these coherence maps.
We impose the following pseudodistributive-law coherences, with ambient associators, unitors, and tensorators inserted.
The two expansions using χ K \chi_K χ K on K ⊗ 3 ⊗ A K^{\otimes3}\otimes A K ⊗ 3 ⊗ A agree after the associator of K K K is inserted. The analogous condition holds for χ A \chi_A χ A on K ⊗ A ⊗ 3 K\otimes A^{\otimes3} K ⊗ A ⊗ 3 .
Set
C 2 , 2 = ( id A ⊗ λ ⊗ id K ) ( λ ⊗ λ ) ( id K ⊗ λ ⊗ id A ) . (97) C_{2,2}=(\operatorname{id}_A\otimes\lambda\otimes\operatorname{id}_K)(\lambda\otimes\lambda)(\operatorname{id}_K\otimes\lambda\otimes\operatorname{id}_A). \tag*{(97)} C 2 , 2 = ( id A ⊗ λ ⊗ id K ) ( λ ⊗ λ ) ( id K ⊗ λ ⊗ id A ) . (97)
The two expansions, first through χ K \chi_K χ K and then through χ A \chi_A χ A , or in the opposite order, agree:
P K = P A : λ ( μ K ⊗ μ A ) ⟹ ( μ A ⊗ μ K ) C 2 , 2 . (98) P_K=P_A:\lambda(\mu_K\otimes\mu_A)\Longrightarrow(\mu_A\otimes\mu_K)C_{2,2}. \tag*{(98)} P K = P A : λ ( μ K ⊗ μ A ) ⟹ ( μ A ⊗ μ K ) C 2 , 2 . (98)
Figure 6. The marked mate construction of the line crossing. Vertical composition is read from bottom to top. Expand B B B as the short suspension segment η A R η A \eta^{R}_{A}\eta_{A} η A R η A . The A A A -unit constraint α λ : λ F R ⇒ F L \alpha_{\lambda}:\lambda F_{R}\Rightarrow F_{L} α λ : λ F R ⇒ F L and its right mate identify this diagram with F R R λ R λ F R F^{R}_{R}\lambda^{R}\lambda F_{R} F R R λ R λ F R . Cancel the adjacent crossing pair using u λ − 1 : λ R λ ⇒ i d u^{-1}_{\lambda}:\lambda^{R}\lambda\Rightarrow\mathrm{id} u λ − 1 : λ R λ ⇒ id , then contract the remaining A A A segment. All horizontal moves are invertible 2-morphisms. The condensation interface η A \eta_{A} η A itself is not inverted.
Restricting either multiplication comparison along a unit agrees with the corresponding pasting of the unit comparisons and the remaining multiplication. This includes both K K K -unit restrictions and the A A A -unit restriction of χ K \chi_{K} χ K , and the corresponding three restrictions of χ A \chi_{A} χ A .
The two reductions at both units coincide:
ν K ∗ η A = α λ ∗ η K . (99) \nu_{K}\ast\eta_{A}=\alpha_{\lambda}\ast\eta_{K}. \tag*{(99)} ν K ∗ η A = α λ ∗ η K . (99)
This is a redundant form of the standard coherence requirements [33 , 34 ] .
Proof of Proposition A.1. Define the marked free interfaces
F R = i d K ⊗ η A , F L = η A ⊗ i d K . (100) F_{R}=\mathrm{id}_{K}\otimes\eta_{A},\qquad F_{L}=\eta_{A}\otimes\mathrm{id}_{K}. \tag*{(100)} F R = id K ⊗ η A , F L = η A ⊗ id K . (100)
Their induced monads on K K K are
T L = F L R F L ≃ B ⊗ i d K , T R = F R R F R ≃ i d K ⊗ B . (101) T_{L}=F^{R}_{L}F_{L}\simeq B\otimes\mathrm{id}_{K},\qquad T_{R}=F^{R}_{R}F_{R}\simeq\mathrm{id}_{K}\otimes B. \tag*{(101)} T L = F L R F L ≃ B ⊗ id K , T R = F R R F R ≃ id K ⊗ B . (101)
In module coordinates these are L B ~ L_{\widetilde{B}} L B and R B ~ R_{\widetilde{B}} R B . The objects A ⊗ K A\otimes K A ⊗ K and K ⊗ A K\otimes A K ⊗ A are their chosen separable module objects [39 Appendix A] .
From the surface-defect law to the line crossing. The marking is the comparison α λ : λ F R ⇒ F L \alpha_{\lambda}:\lambda F_{R}\Rightarrow F_{L} α λ : λ F R ⇒ F L . Choose a unitary adjoint equivalence λ ⊣ λ R \lambda\dashv\lambda^{R} λ ⊣ λ R , with unit u λ : i d ⇒ λ R λ u_{\lambda}:\mathrm{id}\Rightarrow\lambda^{R}\lambda u λ : id ⇒ λ R λ . Taking the right mate of α λ \alpha_{\lambda} α λ gives
B ⊗ i d K ≃ F L R F L ⇒ α λ R ∗ α λ − 1 F R R λ R λ F R ⇒ F R R u λ − 1 F R F R R F R ≃ i d K ⊗ B . (102) \begin{aligned}
B\otimes\mathrm{id}_{K}&\simeq F^{R}_{L}F_{L}
\xRightarrow{\alpha^{R}_{\lambda}\ast\alpha^{-1}_{\lambda}}
F^{R}_{R}\lambda^{R}\lambda F_{R}\\
&\xRightarrow{F^{R}_{R}u^{-1}_{\lambda}F_{R}}
F^{R}_{R}F_{R}\simeq\mathrm{id}_{K}\otimes B. \tag*{(102)}
\end{aligned} B ⊗ id K ≃ F L R F L α λ R ∗ α λ − 1 F R R λ R λ F R F R R u λ − 1 F R F R R F R ≃ id K ⊗ B . (102)
This construction identifies the adjunctions and therefore their monads, including multiplication and unit.
In module coordinates the result is a C C C -module natural isomorphism τ : L B ~ ⇒ R B ~ \tau:L_{\widetilde{B}}\Rightarrow R_{\widetilde{B}} τ : L B ⇒ R B . For the right action X ◃ c = X ⊗ ι ( c ) X\triangleleft c=X\otimes\iota(c) X ◃ c = X ⊗ ι ( c ) , module naturality is
( i d X ⊗ β c , B ~ ) τ X ⊗ ι ( c ) = τ X ⊗ i d ι ( c ) . (103) (\mathrm{id}_{X}\otimes\beta_{c,\widetilde{B}})\tau_{X\otimes\iota(c)}=\tau_{X}\otimes\mathrm{id}_{\iota(c)}. \tag*{(103)} ( id X ⊗ β c , B ) τ X ⊗ ι ( c ) = τ X ⊗ id ι ( c ) . (103)
The K K K -multiplication and unit comparisons give (23) . Setting X = 1 K X=1_K X = 1 K in (A.8) gives (27) ; conversely, these two identities recover the full module-naturality equation. The monad unit gives (26) . For multiplication, the tensor identity expands the component at B ~ ⊗ X \widetilde{B}\otimes X B ⊗ X . The resulting monad-morphism identity is equivalent to (25) , using
m B ~ τ ~ B ~ = ι ( m B C B , B − 1 ) = m B ~ . m_{\widetilde{B}}\widetilde{\tau}_{\widetilde{B}}=\iota(m_{B\mathsf{C}^{-1}_{B,B}})=m_{\widetilde{B}}. m B τ B = ι ( m B C B , B − 1 ) = m B .
Thus the surface-defect law gives all the conditions of Definition 3.1.
From the line crossing to the surface-defect law. Conversely, those conditions make τ : T L ⇒ T R \tau:T_L\Rightarrow T_R τ : T L ⇒ T R an isomorphism of separable monads in the C C C -module realization. Restriction of module structure defines
λ τ : ( X , ρ R : T R X → X ) ⟼ ( X , ρ R τ X : T L X → X ) . (104) \lambda_\tau:(X,\rho_R:T_RX\to X)\longmapsto(X,\rho_R\tau_X:T_LX\to X). \tag*{(104)} λ τ : ( X , ρ R : T R X → X ) ⟼ ( X , ρ R τ X : T L X → X ) . (104)
The inverse uses τ − 1 \tau^{-1} τ − 1 . On free modules the marked comparison λ τ F R ⇒ F L \lambda_\tau F_R\Rightarrow F_L λ τ F R ⇒ F L has underlying map τ X − 1 : T R X → T L X \tau_X^{-1}:T_RX\to T_LX τ X − 1 : T R X → T L X .
The extension from free modules is controlled by separability. For a right B ~ \widetilde{B} B -module ( M , ρ M ) (M,\rho_M) ( M , ρ M ) , set
s B ~ = m B ~ † , e B ~ = s B ~ u B ~ . s_{\widetilde{B}}=m_{\widetilde{B}}^\dagger,\qquad e_{\widetilde{B}}=s_{\widetilde{B}}u_{\widetilde{B}}. s B = m B † , e B = s B u B .
Then
σ M = ( ρ M ⊗ id B ~ ) ( id M ⊗ e B ~ ) : M ⟶ M ⊗ B ~ (105) \sigma_M=(\rho_M\otimes\operatorname{id}_{\widetilde{B}})(\operatorname{id}_M\otimes e_{\widetilde{B}}):M\longrightarrow M\otimes\widetilde{B} \tag*{(105)} σ M = ( ρ M ⊗ id B ) ( id M ⊗ e B ) : M ⟶ M ⊗ B (105)
is a module section of ρ M \rho_M ρ M . Every module is therefore a retract of a free module.
This reduction includes all maps between free modules. Such a map is represented by an arbitrary map f : X → Y ⊗ B ~ f:X\to Y\otimes\widetilde{B} f : X → Y ⊗ B through
f ‾ = ( id Y ⊗ m B ~ ) ( f ⊗ id B ~ ) . \overline{f}=(\operatorname{id}_Y\otimes m_{\widetilde{B}})(f\otimes\operatorname{id}_{\widetilde{B}}). f = ( id Y ⊗ m B ) ( f ⊗ id B ) .
Thus naturality together with the multiplication and unit identities controls all free-module morphisms, not only maps induced by morphisms X → Y X\to Y X → Y .
The tensor identity for τ \tau τ induces χ K \chi_K χ K . Its algebra identities make the crossing compatible with the balancing that defines μ A \mu_A μ A , inducing χ A \chi_A χ A . The unit and relative identities induce α λ τ \alpha_{\lambda_\tau} α λ τ and ν K \nu_K ν K . These maps preserve the separability projectors and hence descend to the chosen retracts.
On free modules, the K K K -associativity pasting compares the two expansions of τ X ⊗ Y ⊗ Z \tau_{X\otimes Y\otimes Z} τ X ⊗ Y ⊗ Z . The A A A -associativity pasting reduces to associativity of m B ~ m_{\widetilde{B}} m B and (25) . The mixed pasting compares crossing before or after multiplication and reduces to the same naturality and algebra identities. The unit pastings reduce to (26) , τ 1 K = id B ~ \tau_{1_K}=\operatorname{id}_{\widetilde{B}} τ 1 K = id B , and (27) . The equalities hold on all free-module maps and therefore on their retracts.
Finally, taking the marked mate of λ τ \lambda_\tau λ τ recovers τ \tau τ . Starting with a surface-defect law, the reconstructed law has the same action on free modules and their maps, together with the same free-interface comparison. The separable module-object universal property identifies it with the original marked law.
Central lift and separable realization
Lemma A.2 (Separable realization of a compatible datum). For every compatible datum ( B , τ ) (B,\tau) ( B , τ ) , the central algebra
L τ = ( B ~ , τ ) ∈ Z ^ L_\tau=(\widetilde{B},\tau)\in\widehat{\mathcal{Z}} L τ = ( B , τ ) ∈ Z is condensable. The composite algebra T λ T_\lambda T λ and the relative algebras A K A_K A K and K B K_B K B are separable.
Proof. The tensor identity makes τ \tau τ a half-braiding. The algebra identities make m B ~ m_{\widetilde{B}} m B and u B ~ u_{\widetilde{B}} u B central morphisms, and the relative identity places the lift in Z ^ \widehat{Z} Z . The neutral crossing gives
m_{\widetilde{B}}\widetilde{T}_{\widetilde{B}}=\iota(m_{B\mathcal{C}}^{-1}_{B,B})=m_{\widetilde{B}}. \tag*{(106)}
Hence the central algebra is commutative.
Because τ \tau τ is unitary, adjoints of central morphisms are central. The normalized dagger-Frobenius structure therefore lifts, with separability section m B ~ † m_{\widetilde{B}}^{\dagger} m B † . Each connected component is a haploid Q-system and hence standard; its commutativity therefore gives trivial twist [40 Remark 2.18 and Proposition 2.22] . This proves condensability, also in Z = Z ^ r e v Z=\widehat{Z}^{\mathrm{rev}} Z = Z rev .
To prove separability of T λ T_{\lambda} T λ , consider the category K B ~ \mathcal{K}_{\widetilde{B}} K B of right B ~ \widetilde{B} B -modules in K \mathcal{K} K . Use τ \tau τ to give a right module N N N the left action
B ~ ⊗ N → τ N N ⊗ B ~ → ρ N N . \widetilde{B}\otimes N \xrightarrow{\tau_N} N\otimes\widetilde{B} \xrightarrow{\rho_N} N. B ⊗ N τ N N ⊗ B ρ N N .
The relative tensor product then makes K B ~ \mathcal{K}_{\widetilde{B}} K B a finite semisimple monoidal category. On free modules its product is induced by
X ⊗ B ~ ⊗ Y ⊗ B ~ → id X ⊗ τ Y ⊗ id B ~ X ⊗ Y ⊗ B ~ ⊗ B ~ → id X ⊗ Y ⊗ m B ~ X ⊗ Y ⊗ B ~ . (107) \begin{aligned}
X\otimes\widetilde{B}\otimes Y\otimes\widetilde{B}
&\xrightarrow{\operatorname{id}_{X}\otimes\tau_Y\otimes\operatorname{id}_{\widetilde{B}}}
X\otimes Y\otimes\widetilde{B}\otimes\widetilde{B}\\
&\xrightarrow{\operatorname{id}_{X\otimes Y}\otimes m_{\widetilde{B}}}
X\otimes Y\otimes\widetilde{B}. \tag*{(107)}
\end{aligned} X ⊗ B ⊗ Y ⊗ B id X ⊗ τ Y ⊗ id B X ⊗ Y ⊗ B ⊗ B id X ⊗ Y ⊗ m B X ⊗ Y ⊗ B . (107)
giving
( X ⊗ B ~ ) ⊗ B ~ ( Y ⊗ B ~ ) ≃ ( X ⊗ Y ) ⊗ B ~ . (108) (X\otimes\widetilde{B})\otimes_{\widetilde{B}}(Y\otimes\widetilde{B})\simeq(X\otimes Y)\otimes\widetilde{B}. \tag*{(108)} ( X ⊗ B ) ⊗ B ( Y ⊗ B ) ≃ ( X ⊗ Y ) ⊗ B . (108)
Thus the free-module functor from K \mathcal{K} K is strong monoidal. Its free objects are dualizable, and every module is a retract of a free object by (A.10). Dualizability is preserved by retracts in this idempotent-complete setting. Hence K B ~ \mathcal{K}_{\widetilde{B}} K B is multifusion.
The relative condition says that L τ L_{\tau} L τ centralizes j C ( C ) j_{\mathcal{C}}(\mathcal{C}) j C ( C ) . The central C \mathcal{C} C -action therefore descends to K B ~ \mathcal{K}_{\widetilde{B}} K B . Under the module-object presentation of A ⊗ K A\otimes\mathcal{K} A ⊗ K , the multiplication of T λ T_{\lambda} T λ is precisely (A.12). Naturality and the algebra identity for τ \tau τ identify the products on arbitrary free-module maps f f f , and separability extends the identification to all modules. Consequently, T λ T_{\lambda} T λ is represented by this C \mathcal{C} C -central multifusion category.
It is therefore a rigid algebra in Mod ( C ) \operatorname{Mod}(\mathcal{C}) Mod ( C ) [31 Lemma 2.1.4] . Its category of regular bimodule endomorphisms is the centralizer of the descended C \mathcal{C} C -action in Z ( K B ~ ) Z(\mathcal{K}_{\widetilde{B}}) Z ( K B ) . It is finite semisimple, using semisimplicity of centers of unitary multifusion categories and closure under direct summands [30 Secs. 7.16 and 9.3] . The separability criterion for rigid algebras now makes T λ T_{\lambda} T λ separable [31 Theorem 3.1.6] .
Finally, an algebra in Bimod C ( R ) \operatorname{Bimod}_{\mathcal{C}}(R) Bimod C ( R ) corresponding to an algebra map R → T R\to T R → T is separable if and only if the underlying algebra T T T is separable [27 Corollary 3.2.12] . Apply this with ( R , T ) = ( K , T λ ) (R,T)=(\mathcal{K},T_{\lambda}) ( R , T ) = ( K , T λ ) and ( A , T λ ) (A,T_{\lambda}) ( A , T λ ) . It proves separability of A K A_{\mathcal{K}} A K and K B \mathcal{K}_{B} K B .
Why the transported surface defect is the marked suspension
Write
F = η A K : K → A K , U = F R , B K = U F . F=\eta_{A_{\mathcal{K}}}:\mathcal{K}\to A_{\mathcal{K}},\qquad U=F^{R},\qquad B_{\mathcal{K}}=UF. F = η A K : K → A K , U = F R , B K = U F .
After forgetting the K \mathcal{K} K -actions, A ⊗ K A\otimes\mathcal{K} A ⊗ K is the separable module object of B ⊗ id K B\otimes\operatorname{id}_{\mathcal{K}} B ⊗ id K . The two K \mathcal{K} K -actions on its underlying object are
K ⊗ A ⊗ K → λ ⊗ id K A ⊗ K ⊗ K → id A ⊗ μ K A ⊗ K , A ⊗ K ⊗ K → id A ⊗ μ K A ⊗ K . (109) \begin{aligned}
\mathcal{K}\otimes A\otimes\mathcal{K}
&\xrightarrow{\lambda\otimes\operatorname{id}_{\mathcal{K}}}
A\otimes\mathcal{K}\otimes\mathcal{K}
\xrightarrow{\operatorname{id}_{A}\otimes\mu_{\mathcal{K}}}
A\otimes\mathcal{K},\\
A\otimes\mathcal{K}\otimes\mathcal{K}
&\xrightarrow{\operatorname{id}_{A}\otimes\mu_{\mathcal{K}}}
A\otimes\mathcal{K}. \tag*{(109)}
\end{aligned} K ⊗ A ⊗ K A ⊗ K ⊗ K λ ⊗ id K A ⊗ K ⊗ K id A ⊗ μ K A ⊗ K , id A ⊗ μ K A ⊗ K . (109)
They are the restrictions of the multiplication of T λ T_\lambda T λ along i K i_K i K .
In module coordinates, F F F is the free-module functor for B ~ \widetilde{B} B and U U U is the forgetful functor. The left/right comparison in (A.14) is τ \tau τ . Evaluation on 1 K 1_K 1 K identifies the adjunction multiplication and unit with m B ~ m_{\widetilde{B}} m B and u B ~ u_{\widetilde{B}} u B . This identifies B K B_K B K with L τ L_\tau L τ as a central algebra.
For a K K K –K K K bimodule object M M M , consider
C M : Hom Z ( M , A K ) ⟶ EM B K ∘ − ( Hom Z ( M , K ) ) , \mathcal{C}_M:\operatorname{Hom}_{\mathcal{Z}}(M,A_K)\longrightarrow\operatorname{EM}_{B_K\circ-}(\operatorname{Hom}_{\mathcal{Z}}(M,K)), C M : Hom Z ( M , A K ) ⟶ EM B K ∘− ( Hom Z ( M , K )) ,
h ⟼ ( U h , U ε K h ) . (110) h\longmapsto(Uh,U\varepsilon^K h). \tag*{(110)} h ⟼ ( U h , U ε K h ) . (110)
Here EM B K ∘ − \operatorname{EM}_{B_K\circ-} EM B K ∘− denotes the category of algebras for the indicated monad on the Hom-category.
We must establish the module-object property inside the bimodule 2-category, not merely after forgetting its actions. The separability element is central:
( τ X ⊗ id B ~ ) ( id B ~ ⊗ τ X ) ( e B ~ ⊗ id X ) = id X ⊗ e B ~ . (111) (\tau_X\otimes\operatorname{id}_{\widetilde{B}})(\operatorname{id}_{\widetilde{B}}\otimes\tau_X)(e_{\widetilde{B}}\otimes\operatorname{id}_X)=\operatorname{id}_X\otimes e_{\widetilde{B}}. \tag*{(111)} ( τ X ⊗ id B ) ( id B ⊗ τ X ) ( e B ⊗ id X ) = id X ⊗ e B . (111)
Consequently, the section (A.10) respects both actions in (A.14). For the uncrossed action this follows from tensor naturality; for the crossed action it follows from (A.16). Applied pointwise to h : M → A K h:M\to A_K h : M → A K , it gives a retract
h ⟶ F U h → ε h K h h\longrightarrow FUh\xrightarrow{\varepsilon^K_h}h h ⟶ F U h ε h K h
inside Hom Z ( M , A K ) \operatorname{Hom}_{\mathcal{Z}}(M,A_K) Hom Z ( M , A K ) .
For f , g : M → K f,g:M\to K f , g : M → K , the adjunction gives
Hom ( F f , F g ) ≃ Hom ( f , B K g ) ≃ Hom B K ( B K f , B K g ) . (112) \operatorname{Hom}(Ff,Fg)\simeq\operatorname{Hom}(f,B_Kg)\simeq\operatorname{Hom}_{B_K}(B_Kf,B_Kg). \tag*{(112)} Hom ( F f , F g ) ≃ Hom ( f , B K g ) ≃ Hom B K ( B K f , B K g ) . (112)
These are the bijections induced by C M \mathcal{C}_M C M . Thus the comparison is fully faithful on free objects, including all morphisms between them. Every source object and every target module is a retract of a free object. Idempotent completeness therefore extends full faithfulness and gives essential surjectivity. The construction is natural under precomposition in M M M .
Hence A K A_K A K is the separable module object of B K B_K B K , with F F F identified with its specified free interface [26 , 39 ] .
It remains to compare multiplications. In the above coordinates, T λ ⊗ K T λ → T λ T_\lambda\otimes_K T_\lambda\to T_\lambda T λ ⊗ K T λ → T λ is the product (A.12). The suspension product on free modules is the same map. The identification holds on all free-module morphisms, and both products preserve the separability projectors. It therefore extends to arbitrary modules. Associativity and unit coherence agree on free modules by the marked pseudodistributive law and hence on their retracts.
The resulting equivalence identifies the multiplication, the unit F F F , its adjunction, and its induced monad:
( A K , η A K ) ≃ ( Σ B K , η Σ B K ) . (A_K,\eta_{A_K})\simeq(\Sigma B_K,\eta_{\Sigma B_K}). ( A K , η A K ) ≃ ( Σ B K , η Σ B K ) .
It is an equivalence of marked E 1 E_1 E 1 -algebras.
The commuting square at the level of marked walls
We use right internal modules throughout. For a separable algebra R ∈ C R\in\mathcal{C} R ∈ C , write
D R = Bimod C ( R ) . \mathcal{D}_R=\operatorname{Bimod}_{\mathcal{C}}(R). D R = Bimod C ( R ) .
The module 2-category RMod R ( C ) \operatorname{RMod}_R(\mathcal{C}) RMod R ( C ) has a left C \mathcal{C} C -action and a right D R \mathcal{D}_R D R -action. The latter is ( M , P ) ↦ M ⊗ R P (M,P)\mapsto M\otimes_R P ( M , P ) ↦ M ⊗ R P .
Proof of Theorem 3.4 . Let R R R be K K K or A A A , let i R : R → T λ i_R:R\to T_\lambda i R : R → T λ be the specified algebra map, and put
S R = R ( T λ ) R . S_R=R(T_\lambda)_R. S R = R ( T λ ) R .
All these algebras are separable by Lemma A.2 .
Endpoint categories. Restriction along i R i_R i R gives
R i R : Bimod C ( T λ ) ⟶ Bimod D R ( S R ) . (113) \mathcal{R}_{i_R}:\operatorname{Bimod}_{\mathcal{C}}(T_\lambda)\longrightarrow\operatorname{Bimod}_{\mathcal{D}_R}(S_R). \tag*{(113)} R i R : Bimod C ( T λ ) ⟶ Bimod D R ( S R ) . (113)
Restriction along i R i_R i R and the universal property of R R R -balancing give inverse descriptions of T λ T_\lambda T λ –T λ T_\lambda T λ bimodules and S R S_R S R –S R S_R S R bimodules internal to D R \mathcal{D}_R D R . They agree on both levels of morphisms, relative tensor products, and the regular unit [27 Corollary 3.2.12] . For R = K , A R=K,A R = K , A , this gives (48) .
The actual composite walls. The relative module-category description of wall composition gives
RMod R ( C ) ⊠ D R RMod S R ( D R ) ≃ RMod T λ ( C ) . (114) \operatorname{RMod}_R(\mathcal{C})\boxtimes_{\mathcal{D}_R}\operatorname{RMod}_{S_R}(\mathcal{D}_R)\simeq\operatorname{RMod}_{T_\lambda}(\mathcal{C}). \tag*{(114)} RMod R ( C ) ⊠ D R RMod S R ( D R ) ≃ RMod T λ ( C ) . (114)
Here ⊠ \boxtimes ⊠ is the relative tensor product of module 2-categories. The balanced functor is induced by
( M , N ) ⟼ M ⊗ R N . (M,N)\longmapsto M\otimes_R N. ( M , N ) ⟼ M ⊗ R N .
Equivalently, a right S R S_R S R -module internal to RMod R ( C ) \operatorname{RMod}_R(\mathcal{C}) RMod R ( C ) is a right R R R -module M M M with an action
M ⊗ R T λ ⟶ M . M\otimes_R T_\lambda\longrightarrow M. M ⊗ R T λ ⟶ M .
This is precisely a right T λ T_\lambda T λ -module whose restricted R R R -action is the given one. This is the extension-of-scalars description of composition in separable Morita theory [27 , 35 ] .
The equivalence retains the left C \mathcal{C} C -action and the right Bimod C ( T λ ) \operatorname{Bimod}_{\mathcal{C}}(T_\lambda) Bimod C ( T λ ) -action. The distinguished pair of regular objects is sent to
R ⊗ R T λ ≃ T λ . R\otimes_R T_\lambda\simeq T_\lambda. R ⊗ R T λ ≃ T λ .
The relative free-module functor is exactly M ↦ M ⊗ R T λ M\mapsto M\otimes_R T_\lambda M ↦ M ⊗ R T λ . For R = K R=K R = K , the marked suspension result identifies it with W B K W_{BK} W B K . For R = A R=A R = A , it is the specified wall G K B G_{KB} G K B .
Comparison of the markings. For X ∈ C X\in\mathcal{C} X ∈ C , let
a R , X : ( X ⊗ R ) ⊗ R T λ → ≃ X ⊗ ( R ⊗ R T λ ) → ≃ X ⊗ T λ (115) a_{R,X}:(X\otimes R)\otimes_R T_\lambda
\xrightarrow{\ \simeq\ }X\otimes(R\otimes_R T_\lambda)
\xrightarrow{\ \simeq\ }X\otimes T_\lambda \tag*{(115)} a R , X : ( X ⊗ R ) ⊗ R T λ ≃ X ⊗ ( R ⊗ R T λ ) ≃ X ⊗ T λ (115)
be the relative associator followed by the regular-module unitor. Define
Ξ λ , X = a A , X a K , X − 1 . (116) \Xi_{\lambda,X}=a_{A,X}a_{K,X}^{-1}. \tag*{(116)} Ξ λ , X = a A , X a K , X − 1 . (116)
These are right T λ T_\lambda T λ -module maps, natural in the objects and both levels of morphisms of C \mathcal{C} C . Their compatibility with tensoring on the left gives the left C \mathcal{C} C -module 2-naturality required in the main theorem.
For an algebra map f : R → S f:R\to S f : R → S , write
Res f : RMod S ( C ) ⟶ RMod R ( C ) \operatorname{Res}_f:\operatorname{RMod}_S(\mathcal{C})\longrightarrow\operatorname{RMod}_R(\mathcal{C}) Res f : RMod S ( C ) ⟶ RMod R ( C )
for the restriction-of-action 2-functor. It retains the underlying object and restricts a right S S S -action ρ M \rho_M ρ M to the right R R R -action
M ⊗ R → id M ⊗ f M ⊗ S → ρ M M , M\otimes R\xrightarrow{\operatorname{id}_M\otimes f}M\otimes S\xrightarrow{\rho_M}M, M ⊗ R id M ⊗ f M ⊗ S ρ M M ,
with the induced coherence maps. In particular, restriction along η R \eta_R η R takes values in RMod 1 C ( C ) ≃ C \operatorname{RMod}_{1_{\mathcal{C}}}(\mathcal{C})\simeq\mathcal{C} RMod 1 C ( C ) ≃ C .
The unit comparisons give
i K η K ≃ η T λ ≃ i A η A . i_K\eta_K \simeq\eta_T\lambda\simeq i_A\eta_A. i K η K ≃ η T λ ≃ i A η A .
Accordingly,
Res η K Res i K ≃ Res η T λ ≃ Res η A Res i A . (117) \operatorname{Res}_{\eta_K}\operatorname{Res}_{i_K}\simeq\operatorname{Res}_{\eta_T}\lambda\simeq\operatorname{Res}_{\eta_A}\operatorname{Res}_{i_A}. \tag*{(117)} Res η K Res i K ≃ Res η T λ ≃ Res η A Res i A . (117)
The maps a R , X a_{R,X} a R , X are the mates of these restriction identifications under the marked free–forgetful adjunctions. They therefore preserve the adjunction units and counits. Together with (A.19), this proves the comparison of actual composite walls, their endpoint actions, and their markings. The common target retains the multiplication of T λ T_\lambda T λ and hence the chosen crossing λ \lambda λ . The argument does not identify walls associated to different compatible data.
Applications: obstructions, existence, and vacuum estimates
Throughout this appendix, K K K is strongly connected. We study the existence of some K K K -compatible condensation of complete Higgs type, without prescribing the parent algebra.
In this transparent-unit setting, Corollary 4.5 forces every connected parent component to be Lagrangian. Consequently, [ C ] = 0 [\mathcal{C}]=0 [ C ] = 0 , equivalently C \mathcal{C} C is braided equivalent to the center of a fusion category [6 ] . We call this the Witt condition. If κ K > 1 \kappa_K>1 κ K > 1 , the equality r = κ K r=\kappa_K r = κ K also forces the parent condensate to be disconnected.
The following existence restriction imposed by the complete Higgs property is less obvious.
Theorem B.1 (Existence criterion). Fix a strongly connected gauging datum K K K . Some compatible condensation of complete Higgs type exists if and only if
[ C ] = 0 and E K : = ⟨ a K ⟩ Z is Tannakian . (118) [\mathcal{C}]=0 \qquad\text{and} \qquad\mathcal{E}_K:=\langle a_K\rangle_{\mathcal{Z}}\text{ is Tannakian}. \tag*{(118)} [ C ] = 0 and E K := ⟨ a K ⟩ Z is Tannakian . (118) In this case, for a finite group Q Q Q and a subgroup P ≤ Q P\leq Q P ≤ Q , one has E K ≃ Rep ( Q ) \mathcal{E}_K\simeq\operatorname{Rep}(Q) E K ≃ Rep ( Q ) and a K ≃ Fun ( Q / P ) a_K\simeq\operatorname{Fun}(Q/P) a K ≃ Fun ( Q / P ) as an algebra. The parent condensate is not prescribed.
Proof. Necessity:
By Corollary 4.5, every parent component is Lagrangian, so [ C ] = 0 [\mathcal{C}]=0 [ C ] = 0 .
Let L = B K L=\mathcal{B}_K L = B K be Lagrangian. Proposition 4.4 implies that a K a_K a K centralizes L L L , and therefore every object of E K \mathcal{E}_K E K centralizes L L L . The free-module functor restricts to a braided tensor functor
F L ∣ E K : E K ⟶ Z L l o c ≃ Vec . (119) F_L|_{\mathcal{E}_K}:\mathcal{E}_K\longrightarrow\mathcal{Z}^{\mathrm{loc}}_L\simeq\operatorname{Vec}. \tag*{(119)} F L ∣ E K : E K ⟶ Z L loc ≃ Vec . (119)
This tensor functor is faithful [30 Remark 4.3.10] , and hence is a braided fiber functor. Therefore E K \mathcal{E}_K E K is Tannakian [30 Sec. 9.9] .
In Rep ( Q ) \operatorname{Rep}(Q) Rep ( Q ) , a connected commutative separable algebra is the function algebra of a transitive finite Q Q Q -set. It therefore has the form Fun ( Q / P ) \operatorname{Fun}(Q/P) Fun ( Q / P ) [41 Lemma 2.10] .
Sufficiency:
Suppose that E K ≃ Rep ( Q ) \mathcal{E}_K\simeq\operatorname{Rep}(Q) E K ≃ Rep ( Q ) is Tannakian and that [ C ] = 0 [\mathcal{C}]=0 [ C ] = 0 . Let R = Fun ( Q ) ∈ E K R=\operatorname{Fun}(Q)\in\mathcal{E}_K R = Fun ( Q ) ∈ E K be its regular condensable algebra and set D = Z R l o c \mathcal{D}=\mathcal{Z}^{\mathrm{loc}}_R D = Z R loc . The Tannakian de-equivariantization description gives
D ≃ Mod ( E K ) Z ′ ( R ) \mathcal{D}\simeq\operatorname{Mod}_{(\mathcal{E}_K)'_{\mathcal{Z}}}(R) D ≃ Mod ( E K ) Z ′ ( R )
as braided categories [42 Proposition 4.56(i)] . Here ( E K ) Z ′ (\mathcal{E}_K)'_{\mathcal{Z}} ( E K ) Z ′ is the Müger centralizer of E K \mathcal{E}_K E K in Z \mathcal{Z} Z . The algebra R R R is transparent in this centralizer, so every right R R R -module there is local. In particular, the underlying objects belong to ( E K ) Z ′ (\mathcal{E}_K)'_{\mathcal{Z}} ( E K ) Z ′ . Condensation preserves the Witt class, so
[ D ] = [ Z ] = [ C ] = 0. [\mathcal{D}] = [\mathcal{Z}] = [\mathcal{C}] = 0. [ D ] = [ Z ] = [ C ] = 0.
The middle equality uses C ≃ Z a K l o c \mathcal{C} \simeq\mathcal{Z}^{\mathrm{loc}}_{a_K} C ≃ Z a K loc . Consequently, D \mathcal{D} D contains at least one connected Lagrangian algebra M M M [6 ] .
View M M M as an algebra in local R R R -modules. Forgetting to Z \mathcal{Z} Z produces a connected condensable algebra L L L together with R → L R \to L R → L [6 Proposition 3.16] . Concretely, the multiplication of L L L is the composite
L ⊗ L ⟶ L ⊗ R L → m M L , L \otimes L \longrightarrow L \otimes_R L \xrightarrow{m_M} L, L ⊗ L ⟶ L ⊗ R L m M L ,
and its unit is 1 → u R R → u M L 1 \xrightarrow{u_R} R \xrightarrow{u_M} L 1 u R R u M L .
Because the underlying objects of D \mathcal{D} D belong to ( E K ′ ) Z (\mathcal{E}_K')_{\mathcal{Z}} ( E K ′ ) Z , L L L therefore centralizes a K ∈ E K a_K \in\mathcal{E}_K a K ∈ E K . Moreover,
FPdim Z ( L ) = FPdim Z ( R ) FPdim D ( M ) = ∣ Q ∣ FPdim ( D ) = ∣ Q ∣ FPdim ( Z ) ∣ Q ∣ 2 = FPdim ( Z ) . \begin{aligned}
\operatorname{FPdim}_{\mathcal{Z}}(L)
= \operatorname{FPdim}_{\mathcal{Z}}(R)\operatorname{FPdim}_{\mathcal{D}}(M) \\
&= |Q|\sqrt{\operatorname{FPdim}(\mathcal{D})} \\
&= |Q|\sqrt{\frac{\operatorname{FPdim}(\mathcal{Z})}{|Q|^2}}
= \sqrt{\operatorname{FPdim}(\mathcal{Z})}.
\end{aligned} FPdim Z ( L ) = FPdim Z ( R ) FPdim D ( M ) = ∣ Q ∣ FPdim ( D ) = ∣ Q ∣ ∣ Q ∣ 2 FPdim ( Z ) = FPdim ( Z ) .
Hence L L L is Lagrangian in Z \mathcal{Z} Z . Proposition 4.4 now supplies compatible parent data whose transported algebra is L L L . Theorem 4.3 makes these data of complete Higgs type.
Theorem B.1 concerns the existence of some compatible condensate for the fixed gauging. It does not replace Theorem 4.3 , which tests an already specified datum, nor does it guarantee a crossing for a prescribed parent algebra B B B . In particular, κ K = ∣ Q / P ∣ \kappa_K = |Q/P| κ K = ∣ Q / P ∣ is an integer.
The Tannakian condition is not implied by the preceding Witt condition. Let I \mathcal{I} I be a unitary Ising modular category, with simple objects 1 , ψ , σ 1,\psi,\sigma 1 , ψ , σ , ψ ⊗ 2 ≃ 1 \psi^{\otimes2} \simeq1 ψ ⊗ 2 ≃ 1 , ψ ⊗ σ ≃ σ \psi\otimes\sigma\simeq\sigma ψ ⊗ σ ≃ σ , σ ⊗ 2 ≃ 1 ⊕ ψ \sigma^{\otimes2} \simeq1 \oplus\psi σ ⊗ 2 ≃ 1 ⊕ ψ , and θ ψ = − 1 \theta_\psi= -1 θ ψ = − 1 . We use Z ( I ) ≃ I ⊠ I r e v \mathcal{Z}(\mathcal{I}) \simeq\mathcal{I} \boxtimes\mathcal{I}^{\mathrm{rev}} Z ( I ) ≃ I ⊠ I rev . For example, consider the reverse condensation from Z ( I ) r e v \mathcal{Z}(\mathcal{I})^{\mathrm{rev}} Z ( I ) rev to C = V e c \mathcal{C} = \mathrm{Vec} C = Vec , whose reverse condensate is the diagonal algebra
a K = ( 1 ⊠ 1 ) ⊕ ( ψ ⊠ ψ ) ⊕ ( σ ⊠ σ ) . (120) a_K = (1 \boxtimes1) \oplus(\psi\boxtimes\psi) \oplus(\sigma\boxtimes\sigma). \tag*{(120)} a K = ( 1 ⊠ 1 ) ⊕ ( ψ ⊠ ψ ) ⊕ ( σ ⊠ σ ) . (120)
The tensor square of σ ⊠ σ \sigma\boxtimes\sigma σ ⊠ σ contains ψ ⊠ 1 \psi\boxtimes1 ψ ⊠ 1 , so this object belongs to ⟨ a K ⟩ \langle a_K\rangle ⟨ a K ⟩ . Its monodromy with σ ⊠ σ \sigma\boxtimes\sigma σ ⊠ σ is − id -\operatorname{id} − id , by the balancing identity and ψ ⊗ σ ≃ σ \psi\otimes\sigma\simeq\sigma ψ ⊗ σ ≃ σ . Thus ⟨ a K ⟩ \langle a_K\rangle ⟨ a K ⟩ is not symmetric and cannot be Tannakian. No complete Higgs condensation therefore exists for this input within Definition 3.1 , despite satisfaction of the Witt condition.
Vacuum estimates in the transparent-unit case
We record two vacuum estimates for the transparent-unit setting.
Lemma B.2. Assume that K K K is strongly connected and that L = B K L = B_K L = B K is connected, and let r = dim Hom C ( 1 , B ) r = \dim\operatorname{Hom}_{\mathcal{C}}(1,B) r = dim Hom C ( 1 , B ) and κ K = FPdim Z ( a K ) \kappa_K = \operatorname{FPdim}_{\mathcal{Z}}(a_K) κ K = FPdim Z ( a K ) . Then
r ≤ κ K . (121) r \le\kappa_K. \tag*{(121)} r ≤ κ K . (121) Write D C : = FPdim ( C ) \mathcal{D}_{\mathcal{C}} := \sqrt{\operatorname{FPdim}(\mathcal{C})} D C := FPdim ( C ) , D i : = FPdim ( C B i l o c ) \mathcal{D}_i := \sqrt{\operatorname{FPdim}(\mathcal{C}^{\mathrm{loc}}_{B_i})} D i := FPdim ( C B i loc ) and D f i n : = FPdim ( Z L l o c ) \mathcal{D}_{\mathrm{fin}} := \sqrt{\operatorname{FPdim}(\mathcal{Z}^{\mathrm{loc}}_L)} D fin := FPdim ( Z L loc ) . Then
D f i n = κ K D C − 1 ∑ i D C D i − 1 = κ K ∑ i D i − 1 . (122) \mathcal{D}_{\mathrm{fin}}
=
\frac{\kappa_K}{\mathcal{D}_{\mathcal{C}}^{-1}\sum_i \mathcal{D}_{\mathcal{C}}\mathcal{D}_i^{-1}}
=
\frac{\kappa_K}{\sum_i \mathcal{D}_i^{-1}}. \tag*{(122)} D fin = D C − 1 ∑ i D C D i − 1 κ K = ∑ i D i − 1 κ K . (122)
Proof. For (B.4), Corollary 4.5 gives r = dim Hom Z ( α K , L ) r = \dim\operatorname{Hom}_{\mathcal{Z}}(\alpha_K,L) r = dim Hom Z ( α K , L ) in the strongly connected case, and the free–forgetful adjunction identifies this with dim Hom Z L ( F L ( α K ) , L ) \dim\operatorname{Hom}_{\mathcal{Z}_L}(F_L(\alpha_K),L) dim Hom Z L ( F L ( α K ) , L ) ; a connected L L L is the simple tensor unit of Z L \mathcal{Z}_L Z L , so this is at most FPdim Z L ( F L ( α K ) ) = κ K \operatorname{FPdim}_{\mathcal{Z}_L}(F_L(\alpha_K))=\kappa_K FPdim Z L ( F L ( α K )) = κ K . This gives (B.4).
For (B.5), in the transparent-unit case the local-module dimension formula (6) applied to the connected algebra L ∈ Z L\in\mathcal{Z} L ∈ Z gives D f i n = FPdim ( Z ) / FPdim ( L ) \mathcal{D}_{\mathrm{fin}}=\sqrt{\operatorname{FPdim}(\mathcal{Z})/\operatorname{FPdim}(L)} D fin = FPdim ( Z ) / FPdim ( L ) , while FPdim ( L ) = FPdim ( B ) = ∑ i FPdim ( B i ) = FPdim ( C ) ∑ i D i − 1 \operatorname{FPdim}(L)=\operatorname{FPdim}(B)=\sum_i\operatorname{FPdim}(B_i)=\sqrt{\operatorname{FPdim}(\mathcal{C})}\sum_i\mathcal{D}_i^{-1} FPdim ( L ) = FPdim ( B ) = ∑ i FPdim ( B i ) = FPdim ( C ) ∑ i D i − 1 by the same formula applied to the connected components B i B_i B i ; the dimension ratio FPdim ( Z ) / FPdim ( C ) = κ K \sqrt{\operatorname{FPdim}(\mathcal{Z})}/\sqrt{\operatorname{FPdim}(\mathcal{C})}=\kappa_K FPdim ( Z ) / FPdim ( C ) = κ K of the transparent-unit case then gives (B.5). □
Ordinary group symmetries
For the chosen gaugeable symmetry G G G , use a faithfully G G G -graded braided G G G -crossed presentation of the defect category,
K = ⨁ g ∈ G K g , K 1 = C , Z ^ ≃ ( K G ) r e v , Z ≃ K G . (123) \mathcal{K}=\bigoplus_{g\in G}\mathcal{K}_g,\qquad\mathcal{K}_1=\mathcal{C},\qquad\widehat{\mathcal{Z}}\simeq(\mathcal{K}^G)^{\mathrm{rev}},\qquad\mathcal{Z}\simeq\mathcal{K}^G. \tag*{(123)} K = g ∈ G ⨁ K g , K 1 = C , Z ≃ ( K G ) rev , Z ≃ K G . (123)
The full crossed data are fixed as part of the symmetry; a permutation of anyon types alone does not specify them [7 ] . The regular algebra R G = Fun ( G ) ∈ Rep ( G ) ⊂ Z R_G=\operatorname{Fun}(G)\in\operatorname{Rep}(G)\subset\mathcal{Z} R G = Fun ( G ) ∈ Rep ( G ) ⊂ Z has pointwise multiplication and the left translation action. With the conventions of the main text, de-equivariantization gives [42 ]
α K ( G ) = R G , C ≃ Z R G l o c , K ≃ Z ^ R G ≃ ( Z R G ) o p . (124) \alpha_{\mathcal{K}(G)}=R_G,\qquad\mathcal{C}\simeq\mathcal{Z}^{\mathrm{loc}}_{R_G},\qquad\mathcal{K}\simeq\widehat{\mathcal{Z}}_{R_G}\simeq(\mathcal{Z}_{R_G})^{\mathrm{op}}. \tag*{(124)} α K ( G ) = R G , C ≃ Z R G loc , K ≃ Z R G ≃ ( Z R G ) op . (124)
Let ρ g \rho_g ρ g denote the categorical action and ω g , h : ρ g ρ h ⇒ ρ g h \omega_{g,h}:\rho_g\rho_h\Rightarrow\rho_{gh} ω g , h : ρ g ρ h ⇒ ρ g h its composition constraint. Compatibility is equivalent to algebra isomorphisms
u g : ρ g ( B ) → ≃ B , u g ∘ ρ g ( u h ) = u g h ∘ ω g , h , B , (125) u_g:\rho_g(B)\xrightarrow{\simeq}B,\qquad u_g\circ\rho_g(u_h)=u_{gh}\circ\omega_{g,h,B}, \tag*{(125)} u g : ρ g ( B ) ≃ B , u g ∘ ρ g ( u h ) = u g h ∘ ω g , h , B , (125)
with the corresponding unit condition [8 ] . For X g ∈ K g X_g\in\mathcal{K}_g X g ∈ K g , write the crossed braiding as c X g , B × : X g ⊗ B → ρ g ( B ) ⊗ X g c^{\times}_{X_g,B}:X_g\otimes B\to\rho_g(B)\otimes X_g c X g , B × : X g ⊗ B → ρ g ( B ) ⊗ X g . The associated line crossing is
τ X g = ( c X g , B × ) − 1 ∘ ( u g − 1 ⊗ id X g ) : B ⊗ X g ⟶ X g ⊗ B . (126) \tau_{X_g}=(c^{\times}_{X_g,B})^{-1}\circ(u_g^{-1}\otimes\operatorname{id}_{X_g}):B\otimes X_g\longrightarrow X_g\otimes B. \tag*{(126)} τ X g = ( c X g , B × ) − 1 ∘ ( u g − 1 ⊗ id X g ) : B ⊗ X g ⟶ X g ⊗ B . (126)
Together with the crossed-braiding axioms, equivariance gives the tensor condition for τ \tau τ , while the algebra-map property gives its multiplication and unit conditions. For X 1 ∈ C X_1\in\mathcal{C} X 1 ∈ C , this reduces to τ X 1 = c X 1 , B × \tau_{X_1}=c^{\times}_{X_1,B} τ X 1 = c X 1 , B × , as required by (28) . The crossed-braiding factor must be retained: the action on vacuum labels alone does not specify the charged crossings.
The transported algebra is
B K ( G ) = ( B , { u g } ) ∈ C G ⊂ Z . (127) B_{\mathcal{K}(G)}=(B,\{u_g\})\in\mathcal{C}^G\subset\mathcal{Z}. \tag*{(127)} B K ( G ) = ( B , { u g }) ∈ C G ⊂ Z . (127)
Taking morphisms from the unit gives
Hom Z ( 1 , B K ( G ) ) ≃ Hom C ( 1 , B ) G . \operatorname{Hom}_{\mathcal{Z}}(1,B_{\mathcal{K}(G)})\simeq\operatorname{Hom}_{\mathcal{C}}(1,B)^G. Hom Z ( 1 , B K ( G ) ) ≃ Hom C ( 1 , B ) G .
Equivalently, (C.4) identifies compatibility of a vacuum operator with G G G -invariance, giving (C.8).
For a concrete example, take the untwisted theory C = Vec \mathcal{C}=\operatorname{Vec} C = Vec , K = Vec G \mathcal{K}=\operatorname{Vec}_G K = Vec G , and B = Fun ( G / H ) B=\operatorname{Fun}(G/H) B = Fun ( G / H ) for a subgroup H ≤ G H\leq G H ≤ G , with the left translation action. Then Z ≃ D ( G ) \mathcal{Z}\simeq\mathcal{D}(G) Z ≃ D ( G ) and B K ( G ) = Fun ( G / H ) B_{\mathcal{K}(G)}=\operatorname{Fun}(G/H) B K ( G ) = Fun ( G / H ) as an algebra in Rep ( G ) ⊂ Z \operatorname{Rep}(G)\subset\mathcal{Z} Rep ( G ) ⊂ Z , with
r = ∣ G / H ∣ , r c o m p = 1 , Z B K ( G ) l o c ≃ D ( H ) . (128) r=|G/H|,\qquad r_{\mathrm{comp}}=1,\qquad\mathcal{Z}^{\mathrm{loc}}_{B_{\mathcal{K}(G)}}\simeq\mathcal{D}(H). \tag*{(128)} r = ∣ G / H ∣ , r comp = 1 , Z B K ( G ) loc ≃ D ( H ) . (128)
Thus ( B , λ ) (B,\lambda) ( B , λ ) is of complete Higgs type if and only if H = { 1 } H=\{1\} H = { 1 } .
Ordinary groups and the non-invertible distinction
As a simple special case of the general framework, consider a gaugeable ordinary finite unitary group symmetry G G G . The vacuum counting then admits a direct interpretation in terms of group orbits. The reverse condensate is the regular Tannakian algebra
a K ( G ) = R G = Fun ( G ) , κ K ( G ) = ∣ G ∣ . (129) a_{K(G)} = R_G = \operatorname{Fun}(G), \qquad\kappa_{K(G)} = |G|. \tag*{(129)} a K ( G ) = R G = Fun ( G ) , κ K ( G ) = ∣ G ∣. (129)
Compatibility is a G G G -equivariant algebra structure on B B B , and
V B T = V B G . (130) \mathcal{V}_{B}^{T} = \mathcal{V}_{B}^{G}. \tag*{(130)} V B T = V B G . (130)
Thus r c o m p r_{\mathrm{comp}} r comp counts the G G G -orbits of the r r r vacuum idempotents, and B K ( G ) B_{K(G)} B K ( G ) is connected precisely when the action is transitive. Condensates in the same orbit are related by tensor autoequivalences and have equal Frobenius–Perron dimensions. If ( B , λ ) (B,\lambda) ( B , λ ) is of complete Higgs type, then B K ( G ) B_{K(G)} B K ( G ) is connected, and Theorem 4.3 gives r = ∣ G ∣ r=|G| r = ∣ G ∣ with every B i B_i B i Lagrangian. Thus G G G acts freely and transitively on the vacua of B B \mathcal{B}_B B B before gauging, and the final phase is trivial. Equation (C.4) gives the corresponding line-crossing formula.
For a general specified operation K K K with transparent unit, connectedness of B K B_K B K is equivalent to dim V B T = 1 \dim\mathcal{V}_{B}^{T}=1 dim V B T = 1 , but there need not be a permutation action relating the individual vacua. For a general nontransparent unit, complete Higgs behavior fixes the absorbed branch number n K ( B ) = κ K n_K(B)=\kappa_K n K ( B ) = κ K , rather than the original parent vacuum number r r r alone (Theorem 4.3 ). Compatible parent insertions need not exhaust the final vacuum algebra.
However, these vacuum conditions do not exclude spectator topological order. For a unitary modular tensor category T \mathcal{T} T with trivial symmetry action, the replacement
( C , K , B ) ⟼ ( C ⊠ T , K ⊠ T , B ⊠ 1 T ) (131) (\mathcal{C},K,B)\longmapsto(\mathcal{C}\boxtimes\mathcal{T},K\boxtimes\mathcal{T},B\boxtimes1_{\mathcal{T}}) \tag*{(131)} ( C , K , B ) ⟼ ( C ⊠ T , K ⊠ T , B ⊠ 1 T ) (131)
preserves κ K \kappa_K κ K and both vacuum counts, and tensors the crossing with the identity. Applied to data of complete Higgs type, this replacement makes T \mathcal{T} T the final intrinsic phase.
Data and computations for the S 3 / Z 2 S_3/\mathbb{Z}_2 S 3 / Z 2 gauging
This appendix supplies the coordinate model behind the computations of § 5.1 –§ 5.4 . The surface decomposition also explains the relation to the pure-condensation operation of § 5.7 . Throughout,
G = S 3 = ⟨ s , t ∣ s 2 = t 3 = 1 , s t s = t − 1 ⟩ , H = ⟨ s ⟩ ≃ Z 2 , G=S_3=\langle s,t\mid s^2=t^3=1,\ sts=t^{-1}\rangle,\qquad H=\langle s\rangle\simeq\mathbb{Z}_2, G = S 3 = ⟨ s , t ∣ s 2 = t 3 = 1 , s t s = t − 1 ⟩ , H = ⟨ s ⟩ ≃ Z 2 ,
where s s s is a transposition and t t t is a 3-cycle: in the three-colour description of § 5.1 , s s s fixes one colour and exchanges the other two, while t t t cycles the three colours. Since t s t − 1 = s t ∉ H tst^{-1}=st\notin H t s t − 1 = s t ∈ / H , the subgroup H H H is not normal in G G G . Consequently, the left-coset set G / H G/H G / H does not inherit a quotient-group structure. The gauge theory is untwisted: no discrete torsion or other twisted coherence datum enters, and every coherence condition below is the ordinary identity for a genuine group action.
The specified wall and its central actions
The grading category. For a finite group F F F , let Vec F \operatorname{Vec}_F Vec F denote the category of finite-dimensional F F F -graded complex vector spaces V = ⨁ g ∈ F V g V=\bigoplus_{g\in F}V_g V = ⨁ g ∈ F V g , with grade-preserving linear maps as morphisms and with tensor product
( V ⊗ W ) k = ⨁ g h = k V g ⊗ W h , 1 Vec F = C in grade 1. (V\otimes W)_k=\bigoplus_{gh=k}V_g\otimes W_h,\qquad1_{\operatorname{Vec}_F}=\mathbb{C}\text{ in grade }1. ( V ⊗ W ) k = g h = k ⨁ V g ⊗ W h , 1 Vec F = C in grade 1.
For F = G = S 3 F=G=S_3 F = G = S 3 this is the untwisted graded category Vec G \operatorname{Vec}_{G} Vec G ; it is a unitary fusion category.
Equivariant objects over conjugation. Conjugation makes H H H act on the grades: it sends a grade g g g to h g h − 1 hgh^{-1} h g h − 1 . An H H H -equivariant object of V e c G \mathrm{Vec}_G Vec G is a graded vector space V = ⨁ g ∈ G V g V = \bigoplus_{g\in G} V_g V = ⨁ g ∈ G V g together with linear maps
ρ h : V g ⟶ V h g h − 1 , g ∈ G , h ∈ H , \rho_h : V_g \longrightarrow V_{hgh^{-1}}, \qquad g \in G,\ h \in H, ρ h : V g ⟶ V h g h − 1 , g ∈ G , h ∈ H ,
satisfying
ρ 1 = id V , ρ h 1 ρ h 2 = ρ h 1 h 2 for all h 1 , h 2 ∈ H . \rho_1 = \operatorname{id}_V,\qquad\rho_{h_1}\rho_{h_2} = \rho_{h_1h_2}\quad\text{for all } h_1,h_2 \in H. ρ 1 = id V , ρ h 1 ρ h 2 = ρ h 1 h 2 for all h 1 , h 2 ∈ H .
These are ordinary untwisted group-action identities: the coherence of the action is exactly the composition law of H H H together with its compatibility with the grading, and no projective factor or higher coherence datum is permitted. For H = ⟨ s ⟩ H = \langle s\rangle H = ⟨ s ⟩ the data reduce to a single linear involution
ρ s : V g ⟶ V s g s , ρ s 2 = id V , \rho_s : V_g \longrightarrow V_{sgs},\qquad\rho_s^2 = \operatorname{id}_V, ρ s : V g ⟶ V s g s , ρ s 2 = id V ,
and a morphism of equivariant objects is a grade-preserving linear map that intertwines the ρ h \rho_h ρ h . The tensor product of equivariant objects carries the diagonal action
ρ h ( v ⊗ w ) = ρ h v ⊗ ρ h w ∈ V h g h − 1 ⊗ W h k h − 1 ⊆ ( V ⊗ W ) h g k h − 1 , \rho_h(v \otimes w) = \rho_hv \otimes\rho_hw \in V_{hgh^{-1}} \otimes W_{hkh^{-1}} \subseteq(V \otimes W)_{hgkh^{-1}}, ρ h ( v ⊗ w ) = ρ h v ⊗ ρ h w ∈ V h g h − 1 ⊗ W hk h − 1 ⊆ ( V ⊗ W ) h g k h − 1 ,
and the associator and unit constraints are induced from those of V e c \mathrm{Vec} Vec . We denote the resulting fusion category of equivariant objects by ( V e c G ) H (\mathrm{Vec}_G)^H ( Vec G ) H . Explicitly, the wall category of the specified operation has the presentation
K = ( V e c G ) H . (132) \mathcal{K} = (\mathrm{Vec}_G)^H . \tag*{(132)} K = ( Vec G ) H . (132)
This is the coordinate presentation of the subgroup-gauging wall of [15 Sec. 6.3] . An object of K \mathcal{K} K is exactly a G G G -graded vector space with the H H H -action described above, and its monoidal structure is the one just displayed. The superscript H H H in (D.1) means that we pass to H H H -equivariant objects; it is not the operation of taking the H H H -invariant vectors of a fixed object.
Wall simples and fusion. The H H H -conjugacy classes in G G G are
{ 1 } , { s } , { t , t 2 } , { s t , s t 2 } . \{1\},\qquad\{s\},\qquad\{t,t^2\},\qquad\{st,st^2\}. { 1 } , { s } , { t , t 2 } , { s t , s t 2 } .
The classes are H H H -stable, and the projection onto a union of classes is an H H H -equivariant idempotent, so a simple object of K \mathcal{K} K is supported on a single class. On the classes { 1 } \{1\} { 1 } and { s } \{s\} { s } the conjugation action of H H H is trivial, and an equivariant object supported there is simply a representation of H H H ; on the classes of t t t and s t st s t the generator s s s exchanges the two grades, and the smallest nonzero such object has one dimension in each grade. This gives exactly six simple objects:
Table D.2.
Here + 1 +1 + 1 and − 1 -1 − 1 denote the trivial and the sign action of H H H on a one-dimensional space, and “exchange” means that ρ s \rho_s ρ s interchanges the two one-dimensional graded pieces. Each graded piece of W [ t ] W_{[t]} W [ t ] or W [ s t ] W_{[st]} W [ s t ] is one-dimensional with a basis vector v g v_g v g satisfying ρ s v g = v s g s \rho_s v_g = v_{sgs} ρ s v g = v s g s . The last column gives the Frobenius–Perron dimensions; for these graded simples the Frobenius–Perron dimension equals the total dimension of the underlying graded space. We write ε = e ⊗ m \varepsilon= e \otimes m ε = e ⊗ m for the composite toric-code anyon and use the same notation for its wall image. The four objects 1 , e , m , ε 1,e,m,\varepsilon 1 , e , m , ε obey the toric-code fusion rules, and
e ⊗ W [ t ] ≃ W [ t ] , e ⊗ W [ s t ] ≃ W [ s t ] , m ⊗ W [ t ] ≃ W [ s t ] , W [ t ] ⊗ W [ t ] ≃ 1 ⊕ e ⊕ W [ t ] , W [ s t ] ⊗ W [ s t ] ≃ 1 ⊕ e ⊕ W [ t ] , W [ t ] ⊗ W [ s t ] ≃ m ⊕ ε ⊕ W [ s t ] . (133) \begin{aligned}
e \otimes W_{[t]} &\simeq W_{[t]}, & e \otimes W_{[st]} &\simeq W_{[st]}, & m \otimes W_{[t]} &\simeq W_{[st]},\\
W_{[t]} \otimes W_{[t]} &\simeq1 \oplus e \oplus W_{[t]}, & W_{[st]} \otimes W_{[st]} &\simeq1 \oplus e \oplus W_{[t]}, & W_{[t]} \otimes W_{[st]} &\simeq m \oplus\varepsilon\oplus W_{[st]}. \tag*{(133)}
\end{aligned} e ⊗ W [ t ] W [ t ] ⊗ W [ t ] ≃ W [ t ] , ≃ 1 ⊕ e ⊕ W [ t ] , e ⊗ W [ s t ] W [ s t ] ⊗ W [ s t ] ≃ W [ s t ] , ≃ 1 ⊕ e ⊕ W [ t ] , m ⊗ W [ t ] W [ t ] ⊗ W [ s t ] ≃ W [ s t ] , ≃ m ⊕ ε ⊕ W [ s t ] . (133)
For example, the grade-1 part of W [ t ] ⊗ W [ t ] W_{[t]} \otimes W_{[t]} W [ t ] ⊗ W [ t ] is spanned by v t ⊗ v t 2 v_t \otimes v_{t^2} v t ⊗ v t 2 and v t 2 ⊗ v t v_{t^2} \otimes v_t v t 2 ⊗ v t , which ρ s \rho_s ρ s exchanges; their symmetric and antisymmetric combinations give 1 1 1 and e e e , and the remaining two vectors give W [ t ] W_{[t]} W [ t ] . The products in the opposite order are isomorphic, and the fusion rules give
FPdim ( K ) = 4 ⋅ 1 + 2 ⋅ 4 = 12. \operatorname{FPdim}(\mathcal{K}) = 4 \cdot1 + 2 \cdot4 = 12. FPdim ( K ) = 4 ⋅ 1 + 2 ⋅ 4 = 12.
In the notation of § 5.1 , the two simple line excitations of the wall that are not parent anyons are
W : = W [ t ] , W m : = W [ s t ] ; W := W_{[t]}, \qquad W_m := W_{[st]}; W := W [ t ] , W m := W [ s t ] ;
both have FPdim = 2 \operatorname{FPdim}=2 FPdim = 2 and neither lies in the image of ι \iota ι , because their grades t ± 1 t^{\pm1} t ± 1 and s t ± 1 st^{\pm1} s t ± 1 lie outside H H H . The second is the one used in § 5.1 under the name W m W_m W m : as an object of the wall,
W m ≃ ι ( m ) ⊗ W , W_m \simeq\iota(m) \otimes W, W m ≃ ι ( m ) ⊗ W ,
because m ⊗ W [ t ] ≃ W [ s t ] m \otimes W_{[t]} \simeq W_{[st]} m ⊗ W [ t ] ≃ W [ s t ] by (D.3).
The parent central action. The fully faithful functor ι : C → K \iota:\mathcal{C}\to\mathcal{K} ι : C → K sends the four toric-code anyons to the four wall simples supported in grades of H H H :
1 ⟼ 1 , e ⟼ e , m ⟼ m , ε ⟼ ε . 1 \longmapsto1,\qquad e \longmapsto e,\qquad m \longmapsto m,\qquad\varepsilon\longmapsto\varepsilon. 1 ⟼ 1 , e ⟼ e , m ⟼ m , ε ⟼ ε .
Hence d K ≃ 1 C d_{\mathcal{K}}\simeq1_{\mathcal{C}} d K ≃ 1 C by (61) . For c 0 ∈ C c_0 \in\mathcal{C} c 0 ∈ C write ι ( c 0 ) = ⨁ h ∈ H c h \iota(c_0)=\bigoplus_{h\in H}c_h ι ( c 0 ) = ⨁ h ∈ H c h . The half-braiding of j C ( c 0 ) j_{\mathcal{C}}(c_0) j C ( c 0 ) is
β c 0 , V ( w h ⊗ v g ) = h ( v g ) ⊗ w h , w h ∈ c h , v g ∈ V g , h ∈ H . (134) \beta_{c_0,V}(w_h \otimes v_g)=h(v_g)\otimes w_h,\qquad w_h\in c_h,\ v_g\in V_g,\ h\in H. \tag*{(134)} β c 0 , V ( w h ⊗ v g ) = h ( v g ) ⊗ w h , w h ∈ c h , v g ∈ V g , h ∈ H . (134)
Here h ( v g ) h(v_g) h ( v g ) abbreviates ρ h ( v g ) \rho_h(v_g) ρ h ( v g ) . For example, ι ( m ) \iota(m) ι ( m ) is supported in grade s s s , so its half-braiding exchanges the two graded components of W [ t ] W_{[t]} W [ t ] . This half-braiding fixes the crossing of the parent anyons across the wall; the extension of the crossing to the wall simples W W W and W m W_m W m is part of the compatible datum analysed in Section D.3.
Under the C \mathcal{C} C -action, W [ t ] W_{[t]} W [ t ] and W [ s t ] W_{[st]} W [ s t ] form a single module-category component. The internal endomorphism algebra of W [ t ] W_{[t]} W [ t ] is 1 ⊕ e = b 1\oplus e=b 1 ⊕ e = b , since ι ( c 0 ) ⊗ W [ t ] \iota(c_0)\otimes W_{[t]} ι ( c 0 ) ⊗ W [ t ] contains W [ t ] W_{[t]} W [ t ] exactly for c 0 ∈ { 1 , e } c_0\in\{1,e\} c 0 ∈ { 1 , e } . The other component is the regular module. As a surface defect, therefore,
K ≃ 1 C ⊕ Σ b . (135) \mathcal{K}\simeq1_{\mathcal{C}}\oplus\Sigma b. \tag*{(135)} K ≃ 1 C ⊕ Σ b . (135)
The unit η K \eta_{\mathcal{K}} η K selects the first summand, and the second is supported on the double coset H t H = { t , t 2 , s t , s t 2 } HtH=\{t,t^2,st,st^2\} H t H = { t , t 2 , s t , s t 2 } . The decomposition (D.5) describes the underlying surface object only: the multiplication, unit and associativity constraints of K \mathcal{K} K are those of the monoidal structure (D.1), and they are not determined by (D.5) alone. § 5.7 studies the summand Σ b \Sigma b Σ b on its own.
The restriction functor and the reverse condensate. The gauged phase is Z = D ( S 3 ) = D ( G ) \mathcal{Z} = \mathcal{D}(S_3) = \mathcal{D}(G) Z = D ( S 3 ) = D ( G ) , whose electric subcategory is Rep ( G ) \operatorname{Rep}(G) Rep ( G ) . We use the usual untwisted gauge-theory presentation of D ( G ) \mathcal{D}(G) D ( G ) by G G G -graded vector spaces equipped with a G G G -action covering conjugation of grades. We write sgn \operatorname{sgn} sgn for the one-dimensional sign representation of S 3 S_3 S 3 , on which s s s acts by − 1 -1 − 1 and t t t acts trivially, and 2 \mathbf{2} 2 for the two-dimensional irreducible representation, the standard representation on the sum-zero plane in C 3 \mathbb{C}^3 C 3 , on which s s s acts as a transposition and t t t as a 3-cycle. On an object of Z \mathcal{Z} Z presented by its grading and equivariance data, the functor q : Z → K q : \mathcal{Z} \to\mathcal{K} q : Z → K retains the grading and restricts the G G G -equivariance to H H H . In particular, q q q is not the operation of taking H H H -invariant vectors. On the electric subcategory Rep ( G ) \operatorname{Rep}(G) Rep ( G ) , whose objects sit in grade 1 1 1 , this is simply the restriction of the G G G -action to H H H , so
q ( 1 ) = 1 , q ( sgn ) = e , q ( 2 ) = 1 ⊕ e . (136) q(\mathbf{1}) = \mathbf{1}, \qquad q(\operatorname{sgn}) = e, \qquad q(\mathbf{2}) = \mathbf{1} \oplus e. \tag*{(136)} q ( 1 ) = 1 , q ( sgn ) = e , q ( 2 ) = 1 ⊕ e . (136)
Indeed e e e is the sign representation of H H H , viewed as the nontrivial character of Z 2 \mathbb{Z}_2 Z 2 , and 2 ∣ H ≃ 1 ⊕ e \mathbf{2}|_H \simeq\mathbf{1} \oplus e 2 ∣ H ≃ 1 ⊕ e . The central lift j Z j_{\mathcal{Z}} j Z and the identification Z r e v ≃ Z ~ \mathcal{Z}^{\mathrm{rev}} \simeq\widetilde{\mathcal{Z}} Z rev ≃ Z used here are those fixed by the reversed-braiding convention (18) of § 3.1 . The wall vacuum lies in grade 1 1 1 with trivial H H H -action. Only electric anyons with an H H H -invariant vector contribute to q R ( 1 K ) q^R(\mathbf{1}_{\mathcal{K}}) q R ( 1 K ) , namely 1 \mathbf{1} 1 and 2 \mathbf{2} 2 each once, so the algebra absorbed by the specified unit is
a K = q R ( 1 K ) = Fun ( G / H ) ≃ 1 ⊕ 2 , Z a K l o c ≃ D ( Z 2 ) , (137) a_{\mathcal{K}} = q^R(\mathbf{1}_{\mathcal{K}}) = \operatorname{Fun}(G/H) \simeq\mathbf{1} \oplus\mathbf{2}, \qquad\mathcal{Z}^{\mathrm{loc}}_{a_{\mathcal{K}}} \simeq\mathcal{D}(\mathbb{Z}_2), \tag*{(137)} a K = q R ( 1 K ) = Fun ( G / H ) ≃ 1 ⊕ 2 , Z a K loc ≃ D ( Z 2 ) , (137)
where Fun ( G / H ) \operatorname{Fun}(G/H) Fun ( G / H ) is the pointwise function algebra on the left-coset set G / H G/H G / H , with pointwise multiplication and the left-translation action of G G G on the cosets [15 Sec. 6.3] . By Proposition 4.4 , C ≃ Z a K l o c \mathcal{C} \simeq\mathcal{Z}^{\mathrm{loc}}_{a_{\mathcal{K}}} C ≃ Z a K loc , with C = D ( Z 2 ) \mathcal{C} = \mathcal{D}(\mathbb{Z}_2) C = D ( Z 2 ) this is the displayed equivalence. In Z = D ( G ) \mathcal{Z} = \mathcal{D}(G) Z = D ( G ) an electric anyon has categorical dimension equal to the dimension of its representation, so FPdim Z ( a K ) = 1 + 2 = ∣ G / H ∣ = 3 \operatorname{FPdim}_{\mathcal{Z}}(a_{\mathcal{K}}) = 1 + 2 = |G/H| = 3 FPdim Z ( a K ) = 1 + 2 = ∣ G / H ∣ = 3 . The dimensions are
FPdim ( C ) = 4 , FPdim ( K ) = 12 , FPdim ( Z ) = 36 , κ K = 3. (138) \operatorname{FPdim}(\mathcal{C}) = 4, \qquad\operatorname{FPdim}(\mathcal{K}) = 12, \qquad\operatorname{FPdim}(\mathcal{Z}) = 36, \qquad\kappa_{\mathcal{K}} = 3. \tag*{(138)} FPdim ( C ) = 4 , FPdim ( K ) = 12 , FPdim ( Z ) = 36 , κ K = 3. (138)
Thus the specified unit is transparent, d K ≃ 1 C d_{\mathcal{K}} \simeq\mathbf{1}_{\mathcal{C}} d K ≃ 1 C , but the reverse condensate is not trivial: a K ≃ 1 ⊕ 2 a_{\mathcal{K}} \simeq\mathbf{1} \oplus\mathbf{2} a K ≃ 1 ⊕ 2 and κ K = [ G : H ] \kappa_{\mathcal{K}} = [G : H] κ K = [ G : H ] .
Central matching and the group-coordinate classification
The operation K \mathcal{K} K is connected, with simple unit η K \eta_{\mathcal{K}} η K selecting the transparent summand of (D.5). Its unit is transparent: by the coordinate model of Section D.1 the functor ι \iota ι is fully faithful, so d K ≃ 1 C d_{\mathcal{K}} \simeq\mathbf{1}_{\mathcal{C}} d K ≃ 1 C by (61) . Thus central matching and reverse condensation reduce compatible data to condensable algebras in the gauged phase. For the present operation the result is expressed by finite G G G -sets; this subsection states and proves the result, first in coordinates and then in the form used in § 5.3 .
Finite G G G -sets and function algebras. All finite G G G -sets considered below are assumed nonempty. A finite left G G G -set is a finite set S S S with a map G × S → S G \times S \to S G × S → S , ( g , x ) ↦ g x (g,x) \mapsto gx ( g , x ) ↦ g x , satisfying 1 x = x 1x = x 1 x = x and g ( h x ) = ( g h ) x g(hx) = (gh)x g ( h x ) = ( g h ) x for all g , h ∈ G g,h \in G g , h ∈ G and x ∈ S x \in S x ∈ S . A morphism of G G G -sets is a G G G -equivariant map, and an isomorphism is a G G G -equivariant bijection. For x ∈ S x \in S x ∈ S , the orbit is G x = { g x : g ∈ G } Gx = \{gx : g \in G\} G x = { g x : g ∈ G } and the stabilizer is G x = { g ∈ G : g x = x } G_x = \{g \in G : gx = x\} G x = { g ∈ G : g x = x } ; the orbits partition S S S , and the orbit–stabilizer relation gives ∣ G x ∣ ∣ G x ∣ = ∣ G ∣ |Gx||G_x| = |G| ∣ G x ∣∣ G x ∣ = ∣ G ∣ . The G G G -set S S S is transitive when it has a single orbit. For a subgroup J ⊂ G J \subset G J ⊂ G , the set of left cosets
G / J = { g J : g ∈ G } G/J = \{gJ : g \in G\} G / J = { g J : g ∈ G }
is a transitive left G G G -set of cardinality [ G : J ] [G : J] [ G : J ] . When J J J is not normal, G / J G/J G / J is a set of cosets and does not inherit a quotient-group structure; this is the case for H H H , and G / H G/H G / H is the three-element set of colours of § 5.1 . For J = ⟨ t ⟩ J=\langle t\rangle J = ⟨ t ⟩ the two left cosets form a two-element G G G -set, which we write G / ⟨ t ⟩ G/\langle t\rangle G / ⟨ t ⟩ without implicitly using the quotient group structure. Conversely every transitive G G G -set is isomorphic to G / J G/J G / J for some subgroup J J J , determined up to conjugacy. We write H \ S H\backslash S H \ S and G \ S G\backslash S G \ S for the sets of H H H -orbits and G G G -orbits of S S S ; they are sets of subsets of S S S , not quotient groups.
We write S ⊔ T S\sqcup T S ⊔ T for the disjoint union of sets, on which G G G acts componentwise, and X ⊕ Y X\oplus Y X ⊕ Y for the direct sum of algebras or of objects. The direct sum is outer: it records orthogonal algebra components and alternative vacua, as in (3) , and is not the Deligne tensor product ⊠ \boxtimes ⊠ of tensor categories, which combines two categories without identifying or summing objects across them. Thus S = ⨆ O ∈ G \ S O S=\bigsqcup_{O\in G\backslash S}O S = ⨆ O ∈ G \ S O for the orbit decomposition, and Fun ( S 1 ⊔ S 2 ) ≃ Fun ( S 1 ) ⊕ Fun ( S 2 ) \operatorname{Fun}(S_1\sqcup S_2)\simeq\operatorname{Fun}(S_1)\oplus\operatorname{Fun}(S_2) Fun ( S 1 ⊔ S 2 ) ≃ Fun ( S 1 ) ⊕ Fun ( S 2 ) as algebras.
For a finite G G G -set S S S , let Fun ( S ) \operatorname{Fun}(S) Fun ( S ) be the algebra of complex-valued functions on S S S with pointwise multiplication and unit:
( z z ′ ) ( x ) = z ( x ) z ′ ( x ) , u L ( x ) = 1 , u L = ∑ x ∈ S δ x . (zz')(x)=z(x)z'(x),\qquad u_L(x)=1,\qquad u_L=\sum_{x\in S}\delta_x. ( z z ′ ) ( x ) = z ( x ) z ′ ( x ) , u L ( x ) = 1 , u L = x ∈ S ∑ δ x .
where δ x ( y ) = δ x , y \delta_x(y)=\delta_{x,y} δ x ( y ) = δ x , y are the orthonormal point functions, satisfying δ x δ y = δ x , y δ x \delta_x\delta_y=\delta_{x,y}\delta_x δ x δ y = δ x , y δ x , and for a subset O ⊂ S O\subset S O ⊂ S we write δ O = ∑ x ∈ O δ x \delta_O=\sum_{x\in O}\delta_x δ O = ∑ x ∈ O δ x . The group acts by
( g ⋅ z ) ( x ) = z ( g − 1 x ) . (g\mathbin{\cdot}z)(x)=z(g^{-1}x). ( g ⋅ z ) ( x ) = z ( g − 1 x ) .
so Fun ( S ) \operatorname{Fun}(S) Fun ( S ) is a commutative separable algebra in Rep ( G ) \operatorname{Rep}(G) Rep ( G ) , with the normalized dagger-Frobenius structure Δ ( δ x ) = δ x ⊗ δ x \Delta(\delta_x)=\delta_x\otimes\delta_x Δ ( δ x ) = δ x ⊗ δ x and ϵ ( δ x ) = 1 \epsilon(\delta_x)=1 ϵ ( δ x ) = 1 of § 2 . It is connected exactly when S S S is transitive. We identify Rep ( G ) \operatorname{Rep}(G) Rep ( G ) with the electric subcategory of Z = D ( G ) \mathcal{Z}=\mathcal{D}(G) Z = D ( G ) .
We also write g ∗ g_* g ∗ for the induced action on point functions, the same action as above in the other notation:
g ∗ δ x = δ g x , ( g ∗ z ) ( x ) = z ( g − 1 x ) . g_*\delta_x=\delta_{gx},\qquad(g_*z)(x)=z(g^{-1}x). g ∗ δ x = δ g x , ( g ∗ z ) ( x ) = z ( g − 1 x ) .
For F = H F=H F = H or G G G , Fun ( S ) F \operatorname{Fun}(S)^F Fun ( S ) F denotes the subalgebra of F F F -invariant functions.
Restriction and the functor F a K F_{a_K} F a K . For a G G G -set S S S , we write Res H G S \operatorname{Res}^{G}_{H}S Res H G S for the same set S S S with the H H H -action obtained by restriction. On function algebras, Res H G \operatorname{Res}^{G}_{H} Res H G keeps the same underlying vector space, the same pointwise multiplication and the same unit, and restricts only the group action; it does not pass to H H H -invariants, does not replace S S S by its H H H -orbits, and does not change the algebra. This is why the same function algebra occurs in Theorem D.1 (i) with the G G G -action for the gauged phase and with the H H H -action for the parent phase.
Under the equivalence C ≃ Z a K l o c \mathcal{C}\simeq\mathcal{Z}^{\mathrm{loc}}_{a_K} C ≃ Z a K loc of Proposition 4.4 , the free-module functor F a K F_{a_K} F a K restricted to Rep ( G ) ⊂ Z \operatorname{Rep}(G)\subset\mathcal{Z} Rep ( G ) ⊂ Z is Res H G \operatorname{Res}^{G}_{H} Res H G . Indeed, taking the fibre at the base point H ∈ G / H H\in G/H H ∈ G / H identifies Rep ( G ) Fun ( G / H ) \operatorname{Rep}(G)_{\operatorname{Fun}(G/H)} Rep ( G ) Fun ( G / H ) with Rep ( H ) \operatorname{Rep}(H) Rep ( H ) and sends V ⊗ Fun ( G / H ) V\otimes\operatorname{Fun}(G/H) V ⊗ Fun ( G / H ) to V ∣ H V|_H V ∣ H . On an algebra this retains the underlying vector space, multiplication and unit and restricts only the group action. Thus the parent algebra associated with a finite G G G -set S S S is B = Res H G Fun ( S ) B=\operatorname{Res}^{G}_{H}\operatorname{Fun}(S) B = Res H G Fun ( S ) .
We can now state the complete classification in coordinates.
Theorem D.1 (Coordinate classification of compatible data). Let G = S 3 G=S_3 G = S 3 and H = ⟨ s ⟩ H=\langle s\rangle H = ⟨ s ⟩ . Isomorphism classes of compatible data ( B , τ ) (B,\tau) ( B , τ ) for the operation (68) are in bijection with isomorphism classes of nonempty finite G G G -sets S S S . For the datum associated with S S S : (i) B K ≃ Fun ( S ) ∈ Rep ( G ) ⊂ Z B_K\simeq\operatorname{Fun}(S)\in\operatorname{Rep}(G)\subset\mathcal{Z} B K ≃ Fun ( S ) ∈ Rep ( G ) ⊂ Z and B ≃ Res H G Fun ( S ) ∈ Rep ( H ) ⊂ C B\simeq\operatorname{Res}^{G}_{H}\operatorname{Fun}(S)\in\operatorname{Rep}(H)\subset\mathcal{C} B ≃ Res H G Fun ( S ) ∈ Rep ( H ) ⊂ C , with pointwise multiplication and unit; (ii) the primitive idempotents of V B \mathcal{V}_B V B are the indicators δ O \delta_O δ O of the H H H -orbits O ⊂ S O\subset S O ⊂ S , and the parent vacuum indexed by O O O carries C O l o c ≃ D ( Stab H x ) \mathcal{C}^{\mathrm{loc}}_O\simeq\mathcal{D}(\operatorname{Stab}_H x) C O loc ≃ D ( Stab H x ) for any x ∈ O x\in O x ∈ O ; (iii) V B G ≃ Fun ( S ) G \mathcal{V}^{G}_B\simeq\operatorname{Fun}(S)^G V B G ≃ Fun ( S ) G , whose primitive idempotents are the indicators of the G G G -orbits, and Z B K l o c ≃ ⨁ G x ∈ G \ S D ( Stab G x ) \mathcal{Z}^{\mathrm{loc}}_{B_K}\simeq\bigoplus_{Gx\in G\backslash S}\mathcal{D}(\operatorname{Stab}_G x) Z B K loc ≃ ⨁ G x ∈ G \ S D ( Stab G x ) ;
(iv) on the g g g -graded part V g V_g V g of a wall excitation, τ \tau τ acts on Fun ( S ) \operatorname{Fun}(S) Fun ( S ) by the permutation P g = ( g − 1 ) ∗ P_g=(g^{-1})_* P g = ( g − 1 ) ∗ ; this is a formula on the full function algebra, not only on its invariant part. In particular r = ∣ H \ S ∣ r=|H\backslash S| r = ∣ H \ S ∣ and r c o m p = r f i n = ∣ G \ S ∣ r_{\mathrm{comp}}=r_{\mathrm{fin}}=|G\backslash S| r comp = r fin = ∣ G \ S ∣ . The transported condensate B K B_K B K is connected exactly when S S S is transitive. Up to isomorphism the transitive G G G -sets are G / G G/G G / G , G / ⟨ t ⟩ G/\langle t\rangle G / ⟨ t ⟩ , G / H G/H G / H and G G G .
Proof. From data to a G G G -set. Since d K ≃ 1 C d_K\simeq\mathbbm{1}_{\mathcal{C}} d K ≃ 1 C by (61) and the dimensions (D.8), the specified unit is transparent. Proposition 4.4 identifies compatible data, up to isomorphism, with condensable algebras L ∈ Z L\in\mathcal{Z} L ∈ Z satisfying the locality condition (63) , with B = F a K ( L ) ∈ Z a K l o c ≃ C B=F_{a_K}(L)\in\mathcal{Z}^{\mathrm{loc}}_{a_K}\simeq\mathcal{C} B = F a K ( L ) ∈ Z a K loc ≃ C and B K ≃ L B_K\simeq L B K ≃ L . Here Z = D ( G ) \mathcal{Z}=\mathcal{D}(G) Z = D ( G ) and a K = Fun ( G / H ) ≃ 1 ⊕ 2 a_K=\operatorname{Fun}(G/H)\simeq\mathbbm{1}\oplus\mathbf{2} a K = Fun ( G / H ) ≃ 1 ⊕ 2 by (D.7).
The constituents of a K a_K a K generate the electric subcategory E : = Rep ( G ) ⊂ Z \mathcal{E}:=\operatorname{Rep}(G)\subset\mathcal{Z} E := Rep ( G ) ⊂ Z : the object 2 \mathbf{2} 2 lies in a K a_K a K , and 2 ⊗ 2 ≃ 1 ⊕ sgn ⊕ 2 \mathbf{2}^{\otimes2}\simeq\mathbbm{1}\oplus\operatorname{sgn}\oplus\mathbf{2} 2 ⊗ 2 ≃ 1 ⊕ sgn ⊕ 2 places sgn \operatorname{sgn} sgn in the tensor subcategory generated by 2 \mathbf{2} 2 , which is therefore all of E \mathcal{E} E . Since L L L centralizes a K a_K a K it centralizes 2 \mathbf{2} 2 and hence every object of E \mathcal{E} E , so L ∈ E ′ L\in\mathcal{E}' L ∈ E ′ . The subcategory E \mathcal{E} E is symmetric, hence contained in its own centralizer E ′ \mathcal{E}' E ′ . Since Z \mathcal{Z} Z is nondegenerate, the Müger centralizer dimension formula gives
FPdim ( E ′ ) = FPdim ( Z ) FPdim ( E ) . \operatorname{FPdim}(\mathcal{E}')=\frac{\operatorname{FPdim}(\mathcal{Z})}{\operatorname{FPdim}(\mathcal{E})}. FPdim ( E ′ ) = FPdim ( E ) FPdim ( Z ) .
With FPdim ( Z ) = 36 \operatorname{FPdim}(\mathcal{Z})=36 FPdim ( Z ) = 36 and FPdim ( E ) = 1 + 1 + 4 = 6 \operatorname{FPdim}(\mathcal{E})=1+1+4=6 FPdim ( E ) = 1 + 1 + 4 = 6 this gives FPdim ( E ′ ) = FPdim ( E ) \operatorname{FPdim}(\mathcal{E}')=\operatorname{FPdim}(\mathcal{E}) FPdim ( E ′ ) = FPdim ( E ) , so E ′ = E \mathcal{E}'=\mathcal{E} E ′ = E and L ∈ Rep ( G ) L\in\operatorname{Rep}(G) L ∈ Rep ( G ) .
In Rep ( G ) \operatorname{Rep}(G) Rep ( G ) the commutative separable algebras are exactly the function algebras Fun ( S ) \operatorname{Fun}(S) Fun ( S ) of finite G G G -sets, their algebra isomorphisms are the G G G -equivariant bijections, Fun ( S ) \operatorname{Fun}(S) Fun ( S ) is connected exactly when S S S is transitive, and all of them have trivial twist [41 Lemma 2.10] . Hence L ≃ Fun ( S ) L\simeq\operatorname{Fun}(S) L ≃ Fun ( S ) for a finite G G G -set S S S , unique up to isomorphism, and B = F a K ( L ) ≃ Res H G Fun ( S ) B=F_{a_K}(L)\simeq\operatorname{Res}^{G}_{H}\operatorname{Fun}(S) B = F a K ( L ) ≃ Res H G Fun ( S ) with the pointwise multiplication and unit. This proves (i).
From a G G G -set to data. Conversely, for a nonempty finite G G G -set S S S put L = Fun ( S ) ∈ Rep ( G ) ⊂ Z L=\operatorname{Fun}(S)\in\operatorname{Rep}(G)\subset\mathcal{Z} L = Fun ( S ) ∈ Rep ( G ) ⊂ Z . Every object of Rep ( G ) \operatorname{Rep}(G) Rep ( G ) centralizes a K ∈ Rep ( G ) a_K\in\operatorname{Rep}(G) a K ∈ Rep ( G ) because Rep ( G ) \operatorname{Rep}(G) Rep ( G ) is symmetric, so (63) holds; L L L is commutative separable with trivial twist, hence condensable in Z \mathcal{Z} Z . Its image B = Res H G Fun ( S ) ∈ Rep ( H ) ⊂ C B=\operatorname{Res}^{G}_{H}\operatorname{Fun}(S)\in\operatorname{Rep}(H)\subset\mathcal{C} B = Res H G Fun ( S ) ∈ Rep ( H ) ⊂ C is again commutative separable with trivial twist, hence condensable in C \mathcal{C} C . Proposition 4.4 turns the algebra identification ξ = id \xi=\operatorname{id} ξ = id of the common underlying algebra into a compatible crossing τ \tau τ of the whole wall. Thus every finite G G G -set gives a compatible datum. Isomorphic G G G -sets give isomorphic data. If the data are isomorphic then Fun ( S ) ≃ B K ≃ Fun ( S ′ ) \operatorname{Fun}(S)\simeq B_K\simeq\operatorname{Fun}(S') Fun ( S ) ≃ B K ≃ Fun ( S ′ ) as algebras in Rep ( G ) \operatorname{Rep}(G) Rep ( G ) , hence S ≃ S ′ S\simeq S' S ≃ S ′ by the classification of function algebras just quoted. Since Proposition 4.4 classifies the data up to isomorphism by the condensable algebra L L L , the construction is well defined on isomorphism classes. This gives the asserted bijection.
Vacuum algebras. For (ii), the parent algebra B = Res H G Fun ( S ) B=\operatorname{Res}^{G}_{H}\operatorname{Fun}(S) B = Res H G Fun ( S ) is Fun ( S ) \operatorname{Fun}(S) Fun ( S ) with the restricted H H H -action, so V B = Hom C ( 1 , B ) = Fun ( S ) H \mathcal{V}_B=\operatorname{Hom}_{\mathcal{C}}(\mathbbm{1},B)=\operatorname{Fun}(S)^H V B = Hom C ( 1 , B ) = Fun ( S ) H , whose primitive idempotents are the indicators δ O \delta_O δ O of the H H H -orbits O ⊂ S O\subset S O ⊂ S . The summand B O = Fun ( O ) ≃ Fun ( H / Stab H x ) B_O=\operatorname{Fun}(O)\simeq\operatorname{Fun}(H/\operatorname{Stab}_H x) B O = Fun ( O ) ≃ Fun ( H / Stab H x ) for x ∈ O x\in O x ∈ O , and condensing the function algebra of H / K H/K H / K in D ( H ) \mathcal{D}(H) D ( H ) gives D ( H ) Fun ( H / K ) l o c ≃ D ( K ) \mathcal{D}(H)^{\mathrm{loc}}_{\operatorname{Fun}(H/K)}\simeq\mathcal{D}(K) D ( H ) Fun ( H / K ) loc ≃ D ( K ) [41 Theorem 2.11] , with K = Stab H x K=\operatorname{Stab}_H x K = Stab H x this is the assertion. Since r = dim V B r=\dim\mathcal{V}_B r = dim V B , this also gives r = ∣ H \ S ∣ r=|H\backslash S| r = ∣ H \ S ∣ .
For (iii), the explicit crossing (D.10) of Section D.3 fixes z ∈ Fun ( S ) z\in\operatorname{Fun}(S) z ∈ Fun ( S ) exactly when P g z = z P_g z=z P g z = z for every g ∈ G g\in G g ∈ G ; since g ↦ P g − 1 g\mapsto P_{g^{-1}} g ↦ P g − 1 is the left action of G G G on S S S , this is exactly G G G -invariance. Hence V B T = Fun ( S ) G \mathcal{V}_B^T=\operatorname{Fun}(S)^G V B T = Fun ( S ) G , whose primitive idempotents are the indicators of the G G G -orbits. The connected components of B K = Fun ( S ) = ⨁ G x ∈ G \ S Fun ( G x ) B_K=\operatorname{Fun}(S)=\bigoplus_{Gx\in G\backslash S}\operatorname{Fun}(Gx) B K = Fun ( S ) = ⨁ G x ∈ G \ S Fun ( G x ) are the function algebras Fun ( G x ) ≃ Fun ( G / Stab G x ) \operatorname{Fun}(Gx)\simeq\operatorname{Fun}(G/\operatorname{Stab}_Gx) Fun ( G x ) ≃ Fun ( G / Stab G x ) , and the same local-module computation in D ( G ) \mathcal{D}(G) D ( G ) gives D ( G ) Fun ( G x ) l o c ≃ D ( Stab G x ) \mathcal{D}(G)^{\mathrm{loc}}_{\operatorname{Fun}(Gx)}\simeq\mathcal{D}(\operatorname{Stab}_Gx) D ( G ) Fun ( G x ) loc ≃ D ( Stab G x ) ; summing over the G G G -orbits gives the stated decomposition. By Corollary 4.5 and (67) , V B T ≃ Hom Z ( 1 , B K ) \mathcal{V}_B^T\simeq\operatorname{Hom}_{\mathcal{Z}}(\mathbbm{1},B_K) V B T ≃ Hom Z ( 1 , B K ) as algebras, so r c o m p = r f i n = dim Fun ( S ) G = ∣ G \ S ∣ r_{\mathrm{comp}}=r_{\mathrm{fin}}=\dim\operatorname{Fun}(S)^G=|G\backslash S| r comp = r fin = dim Fun ( S ) G = ∣ G \ S ∣ .
For (iv), the crossing reconstructed by (32) from the central half-braiding is computed in Section D.3: the reverse braiding of D ( G ) \mathcal{D}(G) D ( G ) gives the half-braiding (D.9), and with ξ = id \xi=\operatorname{id} ξ = id the reconstruction (32) gives the formula (D.10), namely τ V ( z ⊗ v g ) = v g ⊗ P g z \tau_V(z\otimes v_g)=v_g\otimes P_g z τ V ( z ⊗ v g ) = v g ⊗ P g z with P g = ( g − 1 ) ∗ P_g=(g^{-1})_* P g = ( g − 1 ) ∗ . The formula is stated for all z ∈ Fun ( S ) z \in\operatorname{Fun}(S) z ∈ Fun ( S ) . The s s s -odd part of Fun ( S ) \operatorname{Fun}(S) Fun ( S ) is the e e e -multiplicity space of B B B , so the crossing acts on the charged channels and not only on vacuum labels.
Transitive G G G -sets. By (i), B K B_K B K is connected exactly when S S S is transitive. Up to isomorphism the subgroups of S 3 S_3 S 3 are S 3 S_3 S 3 , ⟨ t ⟩ ≃ Z 3 \langle t\rangle\simeq\mathbb{Z}_3 ⟨ t ⟩ ≃ Z 3 , ⟨ s ⟩ = H \langle s\rangle= H ⟨ s ⟩ = H and the trivial subgroup, so the transitive G G G -sets are G / G G/G G / G , G / ⟨ t ⟩ G/\langle t\rangle G / ⟨ t ⟩ , G / H G/H G / H and G G G .
Proof of Theorem 5.2. By Theorem D.1 , isomorphism classes of compatible data for (68) are in bijection with isomorphism classes of finite G G G -sets, with B K = Fun ( S ) B_K = \operatorname{Fun}(S) B K = Fun ( S ) ; in particular B K B_K B K is connected exactly when S S S is transitive. A transitive G G G -set is isomorphic to G / J G/J G / J for a subgroup J ⊂ G J \subset G J ⊂ G determined up to conjugacy, and the subgroups of S 3 S_3 S 3 up to conjugacy are G G G , ⟨ t ⟩ \langle t\rangle ⟨ t ⟩ , H = ⟨ s ⟩ H = \langle s\rangle H = ⟨ s ⟩ and the trivial subgroup. Thus the transitive G G G -sets are G / G G/G G / G , G / ⟨ t ⟩ G/\langle t\rangle G / ⟨ t ⟩ , G / H G/H G / H and G G G , and Theorem D.1 (i)–(iii) gives, in this order,
Table 1 (PDF p. 52).
These are exactly the four parent-algebra and final-phase pairs (73) . Thus there are exactly four isomorphism classes with connected transported condensate, one for each transitive G G G -set, and the bijection of Theorem D.1 says that each row is realized by exactly one of them. The four classes are pairwise non-isomorphic: the underlying objects of B K B_K B K have dimensions 1 , 2 , 3 , 6 1,2,3,6 1 , 2 , 3 , 6 respectively, so the algebras B K B_K B K are pairwise non-isomorphic.
Now let S = ⨆ α O α S=\bigsqcup_\alpha O_\alpha S = ⨆ α O α be a general finite G G G -set, decomposed into its G G G -orbits O α ≃ G / J α O_\alpha\simeq G/J_\alpha O α ≃ G / J α ; the decomposition is unique up to reordering. Then
B K = Fun ( S ) = ⨁ α Fun ( O α ) , B = Res H G Fun ( S ) = ⨁ α Res H G Fun ( O α ) B_K=\operatorname{Fun}(S)=\bigoplus_\alpha\operatorname{Fun}(O_\alpha),\qquad B=\operatorname{Res}^{G}_{H}\operatorname{Fun}(S)=\bigoplus_\alpha\operatorname{Res}^{G}_{H}\operatorname{Fun}(O_\alpha) B K = Fun ( S ) = α ⨁ Fun ( O α ) , B = Res H G Fun ( S ) = α ⨁ Res H G Fun ( O α )
as algebras, and the crossing (D.10) preserves each summand, because P g P_g P g maps the support of Fun ( O α ) \operatorname{Fun}(O_\alpha) Fun ( O α ) to itself for g ∈ G g\in G g ∈ G . Hence the compatible datum is the orthogonal algebra sum of the basic data attached to the orbits O α O_\alpha O α , each of which is one of the four classes; the connected components of B K B_K B K are exactly the summands Fun ( O α ) \operatorname{Fun}(O_\alpha) Fun ( O α ) , so the decomposition follows the connected components of B K B_K B K . Conversely, an orthogonal sum of one or more basic data is a compatible datum for the disjoint union of the corresponding G G G -sets. Uniqueness up to reordering holds because S S S is recovered from B K B_K B K as its G G G -orbit decomposition: two finite G G G -sets with isomorphic function algebras are isomorphic, so the multiset of orbits is determined by B K B_K B K . The decomposition is not read off from an a priori decomposition of the parent algebra: the mixed class S = G / H S=G/H S = G / H has disconnected B = 1 ⊕ b B=1\oplus b B = 1 ⊕ b but connected B K = Fun ( G / H ) B_K=\operatorname{Fun}(G/H) B K = Fun ( G / H ) .
Crossings and the fixed-parent comparisons
In this subsection we reconstruct the crossing in coordinates, verify that it satisfies Definition 3.1 , formulate the fixed-parent extension problem, and prove Propositions 5.1 and 5.3 . Fix a finite G G G -set S S S and L = Fun ( S ) ∈ Rep ( G ) L=\operatorname{Fun}(S)\in\operatorname{Rep}(G) L = Fun ( S ) ∈ Rep ( G ) , with orthonormal point functions. On the basis of point functions the multiplication and unit are
δ x δ y = δ x , y δ x , u L = ∑ x ∈ S δ x , Δ ( δ x ) = δ x ⊗ δ x , ϵ ( δ x ) = 1 ; \delta_x\delta_y=\delta_{x,y}\delta_x,\qquad u_L=\sum_{x\in S}\delta_x,\qquad\Delta(\delta_x)=\delta_x\otimes\delta_x,\qquad\epsilon(\delta_x)=1; δ x δ y = δ x , y δ x , u L = x ∈ S ∑ δ x , Δ ( δ x ) = δ x ⊗ δ x , ϵ ( δ x ) = 1 ;
this is the normalized dagger-Frobenius structure of § 2 , and permutations of S S S preserve it. As recalled in Section D.2, the permutation action on point functions is g ∗ δ x = δ g x g_*\delta_x=\delta_{gx} g ∗ δ x = δ g x , with ( g ∗ z ) ( x ) = z ( g − 1 x ) (g_*z)(x)=z(g^{-1}x) ( g ∗ z ) ( x ) = z ( g − 1 x ) .
Reconstruction of the crossing. In D ( G ) \mathcal{D}(G) D ( G ) , the braiding of a g g g -graded vector past the electric anyon L L L is
c V , L ( v g ⊗ z ) = g ∗ z ⊗ v g , z ∈ L , v g ∈ V g . c_{V,L}(v_g \otimes z)=g_*z\otimes v_g,\qquad z\in L,\ v_g\in V_g. c V , L ( v g ⊗ z ) = g ∗ z ⊗ v g , z ∈ L , v g ∈ V g .
The reversed braiding of Z ^ = Z r e v \widehat{Z}=Z^{\mathrm{rev}} Z = Z rev is c ^ V , L − 1 \widehat{c}_{V,L}^{-1} c V , L − 1 , so the half-braiding of j Z ( L ) j_Z(L) j Z ( L ) is
γ L , V ( z ⊗ v g ) = v g ⊗ ( g − 1 ) ∗ z , z ∈ L , v g ∈ V g . (139) \gamma_{L,V}(z\otimes v_g)=v_g\otimes(g^{-1})_*z,\qquad z\in L,\ v_g\in V_g. \tag*{(139)} γ L , V ( z ⊗ v g ) = v g ⊗ ( g − 1 ) ∗ z , z ∈ L , v g ∈ V g . (139)
By (32) , with ξ \xi ξ the identity of the underlying vector algebra, the crossing is
τ V ( z ⊗ v g ) = v g ⊗ P g z , P g : = ( g − 1 ) ∗ , P g δ x = δ g − 1 x , (140) \tau_V(z\otimes v_g)=v_g\otimes P_gz,\qquad P_g:=(g^{-1})_*,\qquad P_g\delta_x=\delta_{g^{-1}x}, \tag*{(140)} τ V ( z ⊗ v g ) = v g ⊗ P g z , P g := ( g − 1 ) ∗ , P g δ x = δ g − 1 x , (140)
where τ V \tau_V τ V denotes the action of τ \tau τ on L ⊗ V → V ⊗ L L\otimes V\to V\otimes L L ⊗ V → V ⊗ L . Here P g P_g P g is the permutation of S S S induced on point functions by left multiplication by g − 1 g^{-1} g − 1 ; this is the convention used throughout, and it is the same permutation that appears on the g g g -graded part of every wall excitation.
The tensor identity (23) fixes the composition convention for these permutations. If v g ∈ V g v_g\in V_g v g ∈ V g and w h ∈ W h w_h\in W_h w h ∈ W h , then crossing V V V and then W W W gives the coefficient P h P g P_hP_g P h P g on the grade-g h gh g h vector v g ⊗ w h v_g\otimes w_h v g ⊗ w h , and this must equal P g h P_{gh} P g h . This is the composition convention:
P g h = P h P g , P_{gh}=P_hP_g, P g h = P h P g ,
so g ↦ P g g\mapsto P_g g ↦ P g is an anti-homomorphism G → Perm ( S ) G\to\operatorname{Perm}(S) G → Perm ( S ) ; equivalently, g ↦ P g − 1 g\mapsto P_{g^{-1}} g ↦ P g − 1 is an ordinary homomorphism, i.e. a left action of G G G on S S S . The convention is fixed by (D.10) and must be kept when composing wall excitations. Explicitly,
P 1 = I , P t 2 = P t 2 , P s t = P t P s , P s t 2 = P t 2 P s . (141) P_1=I,\qquad P_{t^2}=P_t^2,\qquad P_{st}=P_tP_s,\qquad P_{st^2}=P_t^2P_s. \tag*{(141)} P 1 = I , P t 2 = P t 2 , P s t = P t P s , P s t 2 = P t 2 P s . (141)
The relative identity (27) fixes P s P_s P s as the H H H -action on B B B : on a grade-h h h vector of ι ( c 0 ) \iota(c_0) ι ( c 0 ) , the half-braiding with ι ( B ) \iota(B) ι ( B ) acts on the function-algebra factor by h ∗ h_* h ∗ , whereas the crossing acts by P h = ( h − 1 ) ∗ P_h=(h^{-1})_* P h = ( h − 1 ) ∗ . Hence a crossing of a prescribed parent algebra is determined by a single permutation P t P_t P t , and only the grades t ± 1 t^{\pm1} t ± 1 and s t ± 1 st^{\pm1} s t ± 1 of the wall, that is, the wall excitations W W W and W m W_m W m , can carry information beyond the parent anyons. Since g ↦ P g − 1 g\mapsto P_{g^{-1}} g ↦ P g − 1 must extend the H H H -action, a permutation matrix P t P_t P t defines a crossing exactly when
P t 3 = 1 , P s P t P s = P t − 1 . (142) P_t^3=1,\qquad P_sP_tP_s=P_t^{-1}. \tag*{(142)} P t 3 = 1 , P s P t P s = P t − 1 . (142)
Indeed, with Q t : = P t − 1 Q_t:=P_t^{-1} Q t := P t − 1 and Q s : = P s Q_s:=P_s Q s := P s , the relations (D.12) are exactly Q t 3 = 1 Q_t^3=1 Q t 3 = 1 and Q s Q t Q s = Q t − 1 Q_sQ_tQ_s=Q_t^{-1} Q s Q t Q s = Q t − 1 , together with Q s 2 = 1 Q_s^2=1 Q s 2 = 1 , which holds because P s 2 = 1 P_s^2=1 P s 2 = 1 , these are exactly the relations of the left action of G G G on S S S .
Coherence verification. We check Definition 3.1 for (D.10).
(1) For v g ∈ V g v_g\in V_g v g ∈ V g and w h ∈ W h w_h\in W_h w h ∈ W h , crossing V V V and then W W W gives the coefficient P h P g = P g h P_hP_g=P_{gh} P h P g = P g h by the anti-homomorphism convention on the grade-g h gh g h vector v g ⊗ w h v_g\otimes w_h v g ⊗ w h , which is (23) ; P 1 = I P_1=I P 1 = I gives (24) .
(2) Each P g P_g P g permutes point functions, so it preserves the pointwise product and the unit. This gives (25) and (26) .
(3) The identity P s g s P s = P s P g P_{sgs}P_s=P_sP_g P s g s P s = P s P g makes (D.10) commute with the H H H -actions on ι ( B ) ⊗ V \iota(B)\otimes V ι ( B ) ⊗ V and V ⊗ ι ( B ) V\otimes\iota(B) V ⊗ ι ( B ) , so each τ V \tau_V τ V is a morphism of K \mathcal{K} K .
(4) Naturality in V V V holds because morphisms of K \mathcal{K} K preserve grades.
(5) For c 0 ∈ C c_0\in C c 0 ∈ C and a grade-h h h vector of ι ( c 0 ) \iota(c_0) ι ( c 0 ) , the half-braiding (D.4) with ι ( B ) \iota(B) ι ( B ) acts on the function-algebra factor by h ∗ h_* h ∗ , whereas the crossing acts by P h = ( h − 1 ) ∗ P_h=(h^{-1})_* P h = ( h − 1 ) ∗ . Hence τ ι ( c 0 ) β c 0 , ι ( B ) = i d \tau_{\iota(c_0)}\beta_{c_0,\iota(B)}=\mathrm{id} τ ι ( c 0 ) β c 0 , ι ( B ) = id , which is (27) .
All P g P_g P g are unitary in the chosen basis, because they permute orthonormal point functions. The marked surface crossing is reconstructed by Proposition A.1.
The fixed-parent extension problem. Now let B = Fun ( S 0 ) ∈ Rep ( H ) B=\operatorname{Fun}(S_0)\in\operatorname{Rep}(H) B = Fun ( S 0 ) ∈ Rep ( H ) be a prescribed parent algebra, where S 0 S_0 S 0 is a finite H H H -set, and put P s = s ∗ P_s=s_* P s = s ∗ . By the preceding paragraph, a crossing of B B B is exactly a permutation matrix P t P_t P t satisfying (D.12), and by (D.10) this is the same as a G G G -action on S 0 S_0 S 0 extending the given H H H -action, with t t t acting through P t − 1 P_t^{-1} P t − 1 . Conversely every G G G -action extending the given H H H -action arises from exactly one permutation matrix, namely the inverse of the action of t t t on point functions. An isomorphism between compatible data with the same B B B is an algebra automorphism of B B B intertwining the two values of P t P_t P t , that is, an H H H -equivariant permutation of S 0 S_0 S 0 . Summarizing:
isomorphism classes of compatible crossings of B = Fun ( S 0 ) B=\operatorname{Fun}(S_0) B = Fun ( S 0 ) are in bijection with isomorphism classes of finite G G G -sets S S S with Res H G S ≃ S 0 \operatorname{Res}^{G}_{H}S\simeq S_0 Res H G S ≃ S 0 .
This is the fixed-parent extension problem : the parent algebra fixes the H H H -set S 0 S_0 S 0 , and a crossing is a choice of extension of the H H H -action to a G G G -action. Different extensions can have different transported condensates and different final phases; the two propositions below are proved by classifying these extensions.
Vacuum algebras. In these coordinates,
V B ≃ Fun ( S ) H , V B τ ≃ { z ∈ Fun ( S ) ∣ P g z = z for all g ∈ G } = Fun ( S ) G . (143) \mathcal{V}_B\simeq\operatorname{Fun}(S)^H,\qquad\mathcal{V}^{\tau}_B\simeq\{z\in\operatorname{Fun}(S)\mid P_gz=z\text{ for all }g\in G\}=\operatorname{Fun}(S)^G. \tag*{(143)} V B ≃ Fun ( S ) H , V B τ ≃ { z ∈ Fun ( S ) ∣ P g z = z for all g ∈ G } = Fun ( S ) G . (143)
The primitive idempotents are the orbit indicators δ O \delta_O δ O , for O ∈ H \ S O\in H\backslash S O ∈ H \ S and O ∈ G \ S O\in G\backslash S O ∈ G \ S respectively. Thus the parent vacuum operators are the indicators of the H H H -orbits, while the operators compatible with the crossing are the indicators of the G G G -orbits.
Proof of Proposition 5.1. We prove the statement in the coordinates above. In these coordinates it reads: for f ∈ V B = Fun ( S ) H f\in\mathcal{V}_B=\operatorname{Fun}(S)^H f ∈ V B = Fun ( S ) H and g ∈ G g\in G g ∈ G , if P g f ∈ V B P_gf\in\mathcal{V}_B P g f ∈ V B then P g f = f P_gf=f P g f = f ; consequently, if f ∉ V B τ f\notin\mathcal{V}^{\tau}_B f ∈ / V B τ , then some P g f P_gf P g f has a nonzero component in the e e e -multiplicity space of B B B . The last conclusion is exactly the statement that no nontrivial permutation of the parent vacuum projectors is induced by the crossing: if P g δ O P_g\delta_O P g δ O were again a parent vacuum projector, say δ O ′ \delta_{O'} δ O ′ , then the claim would give δ O ′ = P g δ O = δ O \delta_{O'}=P_g\delta_O=\delta_O δ O ′ = P g δ O = δ O and hence O ′ = O O'=O O ′ = O .
Elements of V B \mathcal{V}_B V B are H H H -invariant, so the claim holds for g ∈ H g\in H g ∈ H . Every element of G G G not in H H H is t k t^k t k or s t k st^k s t k with k ∈ { 1 , 2 } k\in\{1,2\} k ∈ { 1 , 2 } . Since P s t k = P t k P s P_{st^k}=P_{t^k}P_s P s t k = P t k P s and P s f = f P_sf=f P s f = f , we have P s t k f = P t k f P_{st^k}f=P_{t^k}f P s t k f = P t k f , so it suffices to take g = t k g=t^k g = t k . Because s t k s = t − k st^ks=t^{-k} s t k s = t − k , the anti-homomorphism property gives P s P t k P s = P t − k = P t k − 1 P_sP_{t^k}P_s=P_{t^{-k}}=P_{t^k}^{-1} P s P t k P s = P t − k = P t k − 1 . Suppose P s f = f P_sf=f P s f = f and P s ( P t k f ) = P t k f P_s(P_{t^k}f)=P_{t^k}f P s ( P t k f ) = P t k f , the second being the assumption P t k f ∈ V B P_{t^k}f\in\mathcal{V}_B P t k f ∈ V B . Then
P t k − 1 f = P s P t k P s f = P s P t k ( P s f ) = P s P t k f = P t k f , P_{t^k}^{-1}f=P_sP_{t^k}P_sf=P_sP_{t^k}(P_sf)=P_sP_{t^k}f=P_{t^k}f, P t k − 1 f = P s P t k P s f = P s P t k ( P s f ) = P s P t k f = P t k f ,
so P t k 2 f = f P_{t^k}^2f=f P t k 2 f = f , and P t k 3 = 1 P_{t^k}^3=1 P t k 3 = 1 gives P t k f = f P_{t^k}f=f P t k f = f .
For the last assertion, Fun ( S ) \operatorname{Fun}(S) Fun ( S ) splits as the direct sum of its s s s -even and s s s -odd parts, which are the multiplicity spaces of 1 1 1 and e e e in B B B . If f ∉ V B τ f\notin\mathcal{V}^{\tau}_B f ∈ / V B τ , then by (D.13) some P g f ≠ f P_gf\ne f P g f = f . This g g g is not in H H H , because elements of H H H fix f f f ; hence g g g is one of t , t 2 , s t , s t 2 t,t^2,st,st^2 t , t 2 , s t , s t 2 , that is, a coordinate component of W W W or of W m W_m W m . By the claim just proved P g f ∉ V B P_gf\notin\mathcal{V}_B P g f ∈ / V B , so its s s s -odd component is nonzero. That is, P g f P_gf P g f has a nonzero component in the e e e -multiplicity space of B B B . □ \square □
Proof of Proposition 5.3. By the fixed-parent extension problem, crossing classes of B = Fun ( S 0 ) B=\operatorname{Fun}(S_0) B = Fun ( S 0 ) are isomorphism classes of finite G G G -sets S S S with Res H G S ≃ S 0 \operatorname{Res}^{G}_{H}S\simeq S_0 Res H G S ≃ S 0 . Since H ≅ Z 2 H\cong\mathbb{Z}_2 H ≅ Z 2 , an H H H -set is determined up to isomorphism by its cardinality and its number of H H H -fixed points, the remaining points being paired into free orbits; for S S S these two numbers are ∣ S ∣ |S| ∣ S ∣ and the number of H H H -fixed points of S S S , because Res H G S \operatorname{Res}^{G}_{H}S Res H G S has the same underlying set. A finite G G G -set is a disjoint union of transitive G G G -sets, and each transitive G G G -set is one of
G / G , G / H , G / ⟨ t ⟩ , G . G/G,\qquad G/H,\qquad G/\langle t\rangle,\qquad G. G / G , G / H , G / ⟨ t ⟩ , G .
Their cardinalities and numbers of H H H -fixed points are
Table 3.
(here Fix H \operatorname{Fix}_{H} Fix H denotes the set of points fixed by H H H ). The corresponding H H H -sets determine the four parent algebras Fun ( S 0 ) \operatorname{Fun}(S_{0}) Fun ( S 0 ) , namely 1 1 1 , 1 ⊕ b 1 \oplus b 1 ⊕ b , b b b and b ⊕ 3 b^{\oplus3} b ⊕ 3 .
For B = 1 B=1 B = 1 we have ∣ S 0 ∣ = 1 |S_{0}|=1 ∣ S 0 ∣ = 1 and one H H H -fixed point, so ∣ S ∣ = 1 |S|=1 ∣ S ∣ = 1 and # Fix H ( S ) = 1 \#\operatorname{Fix}_{H}(S)=1 # Fix H ( S ) = 1 ; the only possibility is S = G / G S=G/G S = G / G . For B = b B=b B = b we have ∣ S 0 ∣ = 2 |S_{0}|=2 ∣ S 0 ∣ = 2 and no fixed point, so ∣ S ∣ = 2 |S|=2 ∣ S ∣ = 2 and # Fix H ( S ) = 0 \#\operatorname{Fix}_{H}(S)=0 # Fix H ( S ) = 0 ; the only possibility is S = G / ⟨ t ⟩ S=G/\langle t\rangle S = G / ⟨ t ⟩ .
For B = 1 ⊕ b B=1\oplus b B = 1 ⊕ b we have ∣ S 0 ∣ = 3 |S_{0}|=3 ∣ S 0 ∣ = 3 with one H H H -fixed point, hence ∣ S ∣ = 3 |S|=3 ∣ S ∣ = 3 and # Fix H ( S ) = 1 \#\operatorname{Fix}_{H}(S)=1 # Fix H ( S ) = 1 . A partition of 3 into cardinalities of the four transitive G G G -sets can only be 3, 2 + 1 or 1 + 1 + 1. If one orbit has 3 points it is G / H G/H G / H with one fixed point, giving S = G / H S=G/H S = G / H . If the partition is 2 + 1, the two-point orbit is G / ⟨ t ⟩ G/\langle t\rangle G / ⟨ t ⟩ with no fixed point and the one-point orbit is G / G G/G G / G with one fixed point, giving S = G / G ⊔ G / ⟨ t ⟩ S=G/G\sqcup G/\langle t\rangle S = G / G ⊔ G / ⟨ t ⟩ . The partition 1 + 1 + 1 gives three fixed points. Hence there are exactly two classes.
For B = b ⊕ 3 B=b^{\oplus3} B = b ⊕ 3 we have ∣ S ∣ = 6 |S|=6 ∣ S ∣ = 6 and # Fix H ( S ) = 0 \#\operatorname{Fix}_{H}(S)=0 # Fix H ( S ) = 0 . A part 1 in an orbit-size partition contributes the one-point orbit G / G G/G G / G with one fixed point, and a part 3 contributes the orbit G / H G/H G / H with one fixed point. Since # Fix H ( S ) = 0 \#\operatorname{Fix}_{H}(S)=0 # Fix H ( S ) = 0 , every orbit must have size 2 or 6; the only partitions of 6 into 2’s and 6’s are 6 and 2 + 2 + 2. These give S = G S=G S = G and S = 3 G / ⟨ t ⟩ S=3G/\langle t\rangle S = 3 G / ⟨ t ⟩ , and neither has an H H H -fixed point. Hence there are exactly two classes.
Together these enumerations give exactly one crossing class each for 1 1 1 and b b b , and exactly two crossing classes each for 1 ⊕ b 1\oplus b 1 ⊕ b and b ⊕ 3 b^{\oplus3} b ⊕ 3 .
Explicit channel bases and crossings
In this subsection the images under ι \iota ι of the channels of the parent condensate are denoted by the same symbols as the channels themselves. We record the coordinate matrices and then give the full crossing for the mixed, single-vacuum and complete-Higgs classes, together with the two fixed-parent comparisons. The permutation matrices are
C 3 = ( 0 1 0 0 0 1 1 0 0 ) , σ = ( 0 1 1 0 ) . C_{3}=
\begin{pmatrix}
0 & 1 & 0\\
0 & 0 & 1\\
1 & 0 & 0
\end{pmatrix},
\qquad
\sigma=
\begin{pmatrix}
0 & 1\\
1 & 0
\end{pmatrix}. C 3 = 0 0 1 1 0 0 0 1 0 , σ = ( 0 1 1 0 ) .
where C 3 C_{3} C 3 acts on the standard column basis by e a ↦ e a − 1 e_{a}\mapsto e_{a-1} e a ↦ e a − 1 for a ∈ Z 3 a\in\mathbb{Z}_{3} a ∈ Z 3 , and σ \sigma σ exchanges the two coordinates.
Ordered bases. The ordered bases for the six coordinates are the following, in the order of the rows of Table 1.
S = G / G S=G/G S = G / G : ( δ ∗ ) (\delta_{*}) ( δ ∗ ) , the single point function.
S = G / ⟨ t ⟩ S=G/\langle t\rangle S = G / ⟨ t ⟩ , with J = ⟨ t ⟩ J=\langle t\rangle J = ⟨ t ⟩ : ( δ J , δ s J ) (\delta_{J},\delta_{sJ}) ( δ J , δ s J ) .
S = G / H S=G/H S = G / H : ( δ 0 , δ 1 , δ 2 ) : = ( δ H , δ t H , δ t 2 H ) (\delta_{0},\delta_{1},\delta_{2}):=(\delta_{H},\delta_{tH},\delta_{t^{2}H}) ( δ 0 , δ 1 , δ 2 ) := ( δ H , δ t H , δ t 2 H ) .
S = G / G ⊔ G / ⟨ t ⟩ S=G/G\sqcup G/\langle t\rangle S = G / G ⊔ G / ⟨ t ⟩ : ( δ ∗ , δ J , δ s J ) (\delta_{*},\delta_{J},\delta_{sJ}) ( δ ∗ , δ J , δ s J ) .
S = G S=G S = G : ( δ t 0 , δ t 1 , δ t 2 , δ s , δ s t , δ s t 2 ) (\delta_{t_{0}},\delta_{t_{1}},\delta_{t_{2}},\delta_{s},\delta_{st},\delta_{st^{2}}) ( δ t 0 , δ t 1 , δ t 2 , δ s , δ s t , δ s t 2 ) .
(6) S = 3 G / ⟨ t ⟩ S = 3G/\langle t\rangle S = 3 G / ⟨ t ⟩ , where J a J_a J a is the base point of the a a a -th copy of G / ⟨ t ⟩ G/\langle t\rangle G / ⟨ t ⟩ : ( δ J 0 , δ J 1 , δ J 2 , δ s J 0 , δ s J 1 , δ s J 2 ) (\delta_{J_0}, \delta_{J_1}, \delta_{J_2}, \delta_{sJ_0}, \delta_{sJ_1}, \delta_{sJ_2}) ( δ J 0 , δ J 1 , δ J 2 , δ s J 0 , δ s J 1 , δ s J 2 ) . In each row the displayed matrices P s P_s P s and P t P_t P t are written relative to this ordered basis and act on the column vectors of coefficients.
Table 1 (PDF p. 56). The permutations P s P_s P s and P t P_t P t of (D.10) in the ordered bases listed in the text immediately above. The two fixed-parent comparisons of § 5.4 are the pairs of rows with the same parent algebra.
Mixed endpoint. For S = G / H S=G/H S = G / H , put δ 0 = δ H \delta_0=\delta_H δ 0 = δ H , δ 1 = δ t H \delta_1=\delta_{tH} δ 1 = δ t H and δ 2 = δ t 2 H \delta_2=\delta_{t^2H} δ 2 = δ t 2 H , and
p T C = δ 0 , p t r i v = δ 1 + δ 2 , ϕ = δ 1 − δ 2 . (144) p_{\mathrm{TC}}=\delta_0,\qquad p_{\mathrm{triv}}=\delta_1+\delta_2,\qquad\phi=\delta_1-\delta_2. \tag*{(144)} p TC = δ 0 , p triv = δ 1 + δ 2 , ϕ = δ 1 − δ 2 . (144)
The first two span the multiplicity space of 1 1 1 in B B B , and ϕ \phi ϕ spans that of e e e . The multiplication is
p T C 2 = p T C , p t r i v 2 = p t r i v , ϕ 2 = p t r i v , p t r i v ϕ = ϕ , p_{\mathrm{TC}}^2=p_{\mathrm{TC}},\qquad p_{\mathrm{triv}}^2=p_{\mathrm{triv}},\qquad\phi^2=p_{\mathrm{triv}},\qquad p_{\mathrm{triv}}\phi=\phi, p TC 2 = p TC , p triv 2 = p triv , ϕ 2 = p triv , p triv ϕ = ϕ ,
p T C p t r i v = p T C ϕ = 0 , u B = p T C + p t r i v . (145) p_{\mathrm{TC}}p_{\mathrm{triv}}=p_{\mathrm{TC}}\phi=0,\qquad u_B=p_{\mathrm{TC}}+p_{\mathrm{triv}}. \tag*{(145)} p TC p triv = p TC ϕ = 0 , u B = p TC + p triv . (145)
This is the parent algebra B = 1 ⊕ b B=1\oplus b B = 1 ⊕ b of § 5.1 , with p T C p_{\mathrm{TC}} p TC the unit of the 1 1 1 component, p t r i v p_{\mathrm{triv}} p triv the unit of the b b b component, and ϕ \phi ϕ the e e e -charged channel.
In the coordinates w + = v t w_+=v_t w + = v t and w − = v t 2 w_-=v_{t^2} w − = v t 2 of § 5.1 , the two identities (71) are the p T C p_{\mathrm{TC}} p TC lines of the grade-t t t and grade-t 2 t^2 t 2 crossings. In full, on the grade-t t t component,
τ ( p T C ⊗ v t ) = v t ⊗ 1 2 ( p t r i v − ϕ ) , \tau(p_{\mathrm{TC}}\otimes v_t)=v_t\otimes\tfrac{1}{2}(p_{\mathrm{triv}}-\phi), τ ( p TC ⊗ v t ) = v t ⊗ 2 1 ( p triv − ϕ ) ,
τ ( p t r i v ⊗ v t ) = v t ⊗ ( p T C + 1 2 ( p t r i v + ϕ ) ) , \tau(p_{\mathrm{triv}}\otimes v_t)=v_t\otimes\left(p_{\mathrm{TC}}+\tfrac{1}{2}(p_{\mathrm{triv}}+\phi)\right), τ ( p triv ⊗ v t ) = v t ⊗ ( p TC + 2 1 ( p triv + ϕ ) ) ,
τ ( ϕ ⊗ v t ) = v t ⊗ ( p T C − 1 2 ( p t r i v + ϕ ) ) . (146) \tau(\phi\otimes v_t)=v_t\otimes\left(p_{\mathrm{TC}}-\tfrac{1}{2}(p_{\mathrm{triv}}+\phi)\right). \tag*{(146)} τ ( ϕ ⊗ v t ) = v t ⊗ ( p TC − 2 1 ( p triv + ϕ ) ) . (146)
The grade-t 2 t^2 t 2 component is obtained by conjugating with P s P_s P s , since P t 2 = P t − 1 = P s P t P s P_{t^2}=P_t^{-1}=P_sP_tP_s P t 2 = P t − 1 = P s P t P s , and P s P_s P s fixes p T C p_{\mathrm{TC}} p TC and p t r i v p_{\mathrm{triv}} p triv and changes the sign of ϕ \phi ϕ .
The second wall excitation is W m = ι ( m ) ⊗ W W_m=\iota(m)\otimes W W m = ι ( m ) ⊗ W . Its crossing is not independent data: the crossing of the parent anyon ι ( m ) \iota(m) ι ( m ) is fixed by (27) , and the tensor identity (23) determines the crossing of W m W_m W m from the crossings of ι ( m ) \iota(m) ι ( m ) and W W W . On the grade-s t st s t component it is the permutation P s t = P t P s P_{st}=P_tP_s P s t = P t P s , so
τ ( p T C ⊗ v s t ) = v s t ⊗ 1 2 ( p t r i v − ϕ ) , \tau(p_{\mathrm{TC}}\otimes v_{st})=v_{st}\otimes\tfrac{1}{2}(p_{\mathrm{triv}}-\phi), τ ( p TC ⊗ v s t ) = v s t ⊗ 2 1 ( p triv − ϕ ) ,
τ ( p t r i v ⊗ v s t ) = v s t ⊗ ( p T C + 1 2 ( p t r i v + ϕ ) ) , \tau(p_{\mathrm{triv}}\otimes v_{st})=v_{st}\otimes\left(p_{\mathrm{TC}}+\tfrac{1}{2}(p_{\mathrm{triv}}+\phi)\right), τ ( p triv ⊗ v s t ) = v s t ⊗ ( p TC + 2 1 ( p triv + ϕ ) ) ,
τ ( ϕ ⊗ v s t ) = v s t ⊗ ( − p T C + 1 2 ( p t r i v + ϕ ) ) . \tau(\phi\otimes v_{st})=v_{st}\otimes\left(-p_{\mathrm{TC}}+\tfrac{1}{2}(p_{\mathrm{triv}}+\phi)\right). τ ( ϕ ⊗ v s t ) = v s t ⊗ ( − p TC + 2 1 ( p triv + ϕ ) ) .
and the grade-s t 2 st^2 s t 2 component is again obtained by conjugating with P s P_s P s , using P s t 2 = P t 2 P s P_{st^2}=P_t^2P_s P s t 2 = P t 2 P s .
The vacuum algebras are V B = C p T C ⊕ C p t r i v V_B=\mathbb{C}p_{\mathrm{TC}}\oplus\mathbb{C}p_{\mathrm{triv}} V B = C p TC ⊕ C p triv and V B = C u B V^B=\mathbb{C}u_B V B = C u B . The parent vacua have total quantum dimensions 2 2 2 and 1 1 1 , so (B.5) gives D f i n = 3 / ( 1 2 + 1 ) = 2 D_{\mathrm{fin}}=3/(\frac{1}{2}+1)=2 D fin = 3/ ( 2 1 + 1 ) = 2 , the total quantum dimension of the final toric code.
Single-vacuum endpoint. For S = G / ⟨ t ⟩ S=G/\langle t\rangle S = G / ⟨ t ⟩ , put δ S = δ J + δ s J \delta_S=\delta_J+\delta_{sJ} δ S = δ J + δ s J and ϕ = δ J − δ s J \phi=\delta_J-\delta_{sJ} ϕ = δ J − δ s J , with J = ⟨ t ⟩ J=\langle t\rangle J = ⟨ t ⟩ , so that ϕ 2 = δ S = u B \phi^2=\delta_S=u_B ϕ 2 = δ S = u B and B ≃ b B\simeq b B ≃ b as algebras, with δ S \delta_S δ S the unit and ϕ \phi ϕ the e e e -channel. Since P t = I 2 P_t=I_2 P t = I 2 , the crossing through W W W is the flip:
τ ( z ⊗ v t ± 1 ) = v t ± 1 ⊗ z , z ∈ { δ S , ϕ } . \tau(z\otimes v_{t^{\pm1}})=v_{t^{\pm1}}\otimes z,\qquad z\in\{\delta_S,\phi\}. τ ( z ⊗ v t ± 1 ) = v t ± 1 ⊗ z , z ∈ { δ S , ϕ } .
Since P s t = P t P s = P s P_{st}=P_tP_s=P_s P s t = P t P s = P s , the crossing through W m ≃ m ⊗ W W_m\simeq m\otimes W W m ≃ m ⊗ W coincides with that through m m m , namely ϕ ↦ − ϕ \phi\mapsto-\phi ϕ ↦ − ϕ . The latter is fixed by (28) and carries no compatibility information: it is the crossing of the parent anyon m m m . The content of this datum is that W W W crosses trivially, which is why Z 3 \mathbb Z_3 Z 3 survives the condensation. Both vacuum algebras are C u B Cu_B C u B ; the parent vacuum is intrinsically trivial, D f i n = 3 \mathcal D_{\mathrm{fin}}=3 D fin = 3 , and FPdim ( Z B l o c ) = 9 \operatorname{FPdim}(\mathcal Z_B^{\mathrm{loc}})=9 FPdim ( Z B loc ) = 9 .
Complete Higgs endpoint. For S = G S=G S = G , the H H H -orbits are O a = { t a , s t a } O_a=\{t^a,st^a\} O a = { t a , s t a } with a ∈ Z 3 a\in\mathbb Z_3 a ∈ Z 3 . Put
δ O a = δ t a + δ s t a , ϕ a = δ t a − δ s t a . (147) \delta_{O_a}=\delta_{t^a}+\delta_{st^a},\qquad\phi_a=\delta_{t^a}-\delta_{st^a}. \tag*{(147)} δ O a = δ t a + δ s t a , ϕ a = δ t a − δ s t a . (147)
Products between different a a a vanish, and
δ O a 2 = δ O a , δ O a ϕ a = ϕ a , ϕ a 2 = δ O a , u B = ∑ a δ O a . (148) \delta_{O_a}^2=\delta_{O_a},\qquad\delta_{O_a}\phi_a=\phi_a,\qquad\phi_a^2=\delta_{O_a},\qquad u_B=\sum_a\delta_{O_a}. \tag*{(148)} δ O a 2 = δ O a , δ O a ϕ a = ϕ a , ϕ a 2 = δ O a , u B = a ∑ δ O a . (148)
In the notation of § 5.4 the three δ O a \delta_{O_a} δ O a are the parent vacuum idempotents p a p_a p a , and ϕ a \phi_a ϕ a is the e e e -channel in the a a a -th vacuum. Since P t δ t a = δ t a − 1 P_t\delta_{t^a}=\delta_{t^{a-1}} P t δ t a = δ t a − 1 and P t δ s t a = δ s t a + 1 P_t\delta_{st^a}=\delta_{st^{a+1}} P t δ s t a = δ s t a + 1 , the mixing formula (78) of § 5.4 reads, in these coordinates,
τ ( δ O a ⊗ v t ) = v t ⊗ 1 2 ( δ O a − 1 + δ O a + 1 + ϕ a − 1 − ϕ a + 1 ) . \tau(\delta_{O_a}\otimes v_t)=v_t\otimes\frac{1}{2}\left(\delta_{O_{a-1}}+\delta_{O_{a+1}}+\phi_{a-1}-\phi_{a+1}\right). τ ( δ O a ⊗ v t ) = v t ⊗ 2 1 ( δ O a − 1 + δ O a + 1 + ϕ a − 1 − ϕ a + 1 ) .
which is (78) under p a = δ O a p_a=\delta_{O_a} p a = δ O a . The charged channels transform by
τ ( ϕ a ⊗ v t ) = v t ⊗ 1 2 ( δ O a − 1 − δ O a + 1 + ϕ a − 1 + ϕ a + 1 ) . \tau(\phi_a\otimes v_t)=v_t\otimes\frac{1}{2}\left(\delta_{O_{a-1}}-\delta_{O_{a+1}}+\phi_{a-1}+\phi_{a+1}\right). τ ( ϕ a ⊗ v t ) = v t ⊗ 2 1 ( δ O a − 1 − δ O a + 1 + ϕ a − 1 + ϕ a + 1 ) .
with indices modulo 3 3 3 . The w − w_- w − companion is the grade-t 2 t^2 t 2 line, obtained by conjugating with P s P_s P s ; explicitly,
τ ( δ O a ⊗ v t 2 ) = v t 2 ⊗ 1 2 ( δ O a − 1 + δ O a + 1 − ϕ a − 1 + ϕ a + 1 ) . \tau(\delta_{O_a}\otimes v_{t^2})=v_{t^2}\otimes\frac{1}{2}\left(\delta_{O_{a-1}}+\delta_{O_{a+1}}-\phi_{a-1}+\phi_{a+1}\right). τ ( δ O a ⊗ v t 2 ) = v t 2 ⊗ 2 1 ( δ O a − 1 + δ O a + 1 − ϕ a − 1 + ϕ a + 1 ) .
which is the companion formula of § 5.4 . Thus each vacuum idempotent is sent to the vacua and charged channels of the other two, and V B R = C u B \mathcal V_B^{\mathrm R}=Cu_B V B R = C u B . All three parent vacua are intrinsically trivial, so D f i n = 3 / ( 1 + 1 + 1 ) = 1 \mathcal D_{\mathrm{fin}}=3/(1+1+1)=1 D fin = 3/ ( 1 + 1 + 1 ) = 1 .
Two crossings of one parent algebra. For B = 1 ⊕ b B=1\oplus b B = 1 ⊕ b , both data use the basis ( δ 0 , δ 1 , δ 2 ) (\delta_0,\delta_1,\delta_2) ( δ 0 , δ 1 , δ 2 ) and P s = ( 1 ) ⊕ σ P_s=(1)\oplus\sigma P s = ( 1 ) ⊕ σ . For the transitive datum δ 0 = δ H \delta_0=\delta_H δ 0 = δ H ; for the split datum δ 0 = δ ∗ \delta_0=\delta_* δ 0 = δ ∗ and ( δ 1 , δ 2 ) = ( δ J , δ s J ) (\delta_1,\delta_2)=(\delta_J,\delta_{sJ}) ( δ 1 , δ 2 ) = ( δ J , δ s J ) . The two data differ only in P t = C 3 P_t=C_3 P t = C 3 versus P t = I 3 P_t=I_3 P t = I 3 . In the transitive datum the crossing on W W W is the charged mixing (D.16) and the crossing on W m W_m W m is through P s t = P t P s P_{st}=P_tP_s P s t = P t P s . In the split datum P t = I 3 P_t=I_3 P t = I 3 , so on W W W the crossing is the flip z ⊗ v ↦ v ⊗ z z\otimes v\mapsto v\otimes z z ⊗ v ↦ v ⊗ z for every channel z ∈ { p T C , p t r i v , ϕ } z\in\{p_{\mathrm{TC}},p_{\mathrm{triv}},\phi\} z ∈ { p TC , p triv , ϕ } , while on W m W_m W m it is the fixed magnetic crossing, which fixes p T C p_{\mathrm{TC}} p TC and p t r i v p_{\mathrm{triv}} p triv and reverses the sign of ϕ \phi ϕ . For B = b Z 3 B=b^{\mathbb Z_3} B = b Z 3 , identify t a t^a t a with J a J_a J a and s t a st^a s t a with s J a sJ_a s J a . Both data then have P s = σ ⊗ I 3 P_s=\sigma\otimes I_3 P s = σ ⊗ I 3 , and they differ only in P t = C 3 ⊕ C 3 − 1 P_t=C_3\oplus C_3^{-1} P t = C 3 ⊕ C 3 − 1 versus P t = I 6 P_t=I_6 P t = I 6 . In the transitive datum the crossing on W W W is the charged mixing (78) ; in the split datum it is the flip in each component, and on W m W_m W m it is the fixed magnetic crossing in each component. In each pair the transitive datum has V B T = C u B \mathcal V_B^{\mathsf T}=Cu_B V B T = C u B , while the other has V B T = V B \mathcal V_B^{\mathsf T}=\mathcal V_B V B T = V B . By Proposition 5.3 these are all the crossing classes of the two parent algebras.
For completeness we record the vacuum algebras of the remaining rows. For S = G / G S=G/G S = G / G , V B = V B T = C δ ∗ \mathcal V_B=\mathcal V_B^{\mathsf T}=C\delta_* V B = V B T = C δ ∗ . For S = G / G ∪ G / ⟨ t ⟩ S=G/G\cup G/\langle t\rangle S = G / G ∪ G / ⟨ t ⟩ , V B = V B T = C p T C ⊕ C p t r i v \mathcal V_B=\mathcal V_B^{\mathsf T}=Cp_{\mathrm{TC}}\oplus Cp_{\mathrm{triv}} V B = V B T = C p TC ⊕ C p triv , where p T C = δ ∗ p_{\mathrm{TC}}=\delta_* p TC = δ ∗ and p t r i v = δ J + δ s J p_{\mathrm{triv}}=\delta_J+\delta_{sJ} p triv = δ J + δ s J . For S = 3 G / ⟨ t ⟩ S=3G/\langle t\rangle S = 3 G / ⟨ t ⟩ , V B = V B T = ⨁ a C δ O a \mathcal V_B=\mathcal V_B^{\mathsf T}=\bigoplus_a C\delta_{O_a} V B = V B T = ⨁ a C δ O a , where O a = { J a , s J a } O_a=\{J_a,sJ_a\} O a = { J a , s J a } is the a a a -th H H H -orbit.
Algebra decompositions and vacuum counting
By Theorem D.1(i), the transported condensate of the datum associated with a nonempty finite G G G -set S S S is B K = Fun ( S ) ∈ Rep ( G ) ⊂ Z B_K=\operatorname{Fun}(S)\in\operatorname{Rep}(G)\subset\mathcal{Z} B K = Fun ( S ) ∈ Rep ( G ) ⊂ Z and its parent algebra is B = Res H G Fun ( S ) ∈ Rep ( H ) ⊂ C B=\operatorname{Res}^{G}_{H}\operatorname{Fun}(S)\in\operatorname{Rep}(H)\subset\mathcal{C} B = Res H G Fun ( S ) ∈ Rep ( H ) ⊂ C . The decompositions below describe the underlying objects only; the algebra structure in each case is the pointwise multiplication and unit supplied by Fun ( S ) \operatorname{Fun}(S) Fun ( S ) , restricted along Res H G \operatorname{Res}^{G}_{H} Res H G for B B B , and it is not implied by the decomposition into simple objects. Throughout, whenever B K B_K B K or B B B is named, the algebra meant is this function algebra.
Using Res H G sgn ≃ e \operatorname{Res}^{G}_{H}\operatorname{sgn}\simeq e Res H G sgn ≃ e and Res H G 2 ≃ 1 ⊕ e \operatorname{Res}^{G}_{H}2\simeq1\oplus e Res H G 2 ≃ 1 ⊕ e , the transported and parent condensates of the six coordinates decompose as
S B K = Fun ( S ) ∈ Z B = Res H G Fun ( S ) ∈ C G / G 1 1 G / ⟨ t ⟩ 1 ⊕ sgn b G / H 1 ⊕ 2 1 ⊕ b G / G ⊔ G / ⟨ t ⟩ 1 ⊕ 2 ⊕ sgn 1 ⊕ b G 1 ⊕ sgn ⊕ 2 ⊕ 2 b ⊕ 3 3 G / ⟨ t ⟩ ( 1 ⊕ sgn ) ⊕ 3 b ⊕ 3 (149) \begin{array}{c|c|c}
S & B_K=\operatorname{Fun}(S)\in\mathcal{Z} & B=\operatorname{Res}^{G}_{H}\operatorname{Fun}(S)\in\mathcal{C}\\
G/G & 1 & 1\\
G/\langle t\rangle& 1\oplus\operatorname{sgn} & b\\
G/H & 1\oplus2 & 1\oplus b\\
G/G\sqcup G/\langle t\rangle& 1^{\oplus2}\oplus\operatorname{sgn} & 1\oplus b\\
G & 1\oplus\operatorname{sgn}\oplus2^{\oplus2} & b^{\oplus3}\\
3G/\langle t\rangle& (1\oplus\operatorname{sgn})^{\oplus3} & b^{\oplus3} \tag*{(149)}
\end{array} S G / G G / ⟨ t ⟩ G / H G / G ⊔ G / ⟨ t ⟩ G 3 G / ⟨ t ⟩ B K = Fun ( S ) ∈ Z 1 1 ⊕ sgn 1 ⊕ 2 1 ⊕ 2 ⊕ sgn 1 ⊕ sgn ⊕ 2 ⊕ 2 ( 1 ⊕ sgn ) ⊕ 3 B = Res H G Fun ( S ) ∈ C 1 b 1 ⊕ b 1 ⊕ b b ⊕ 3 b ⊕ 3 (149)
The middle column records the decomposition of the underlying object of B K B_K B K into simple objects of Rep ( G ) \operatorname{Rep}(G) Rep ( G ) . The right column records the decomposition of the parent algebra B B B into its connected components, each of which is the function algebra of a single H H H -orbit; for instance b = Fun ( H ) b=\operatorname{Fun}(H) b = Fun ( H ) is the algebra of a free two-point orbit and 1 1 1 that of a one-point orbit. As objects of C \mathcal{C} C , 1 ⊕ b ≃ 1 ⊕ 2 ⊕ e 1\oplus b\simeq1^{\oplus2}\oplus e 1 ⊕ b ≃ 1 ⊕ 2 ⊕ e and b ⊕ 3 ≃ 1 ⊕ 3 ⊕ e ⊕ 3 b^{\oplus3}\simeq1^{\oplus3}\oplus e^{\oplus3} b ⊕ 3 ≃ 1 ⊕ 3 ⊕ e ⊕ 3 . In every row FPdim ( B K ) = FPdim ( B ) = ∣ S ∣ \operatorname{FPdim}(B_K)=\operatorname{FPdim}(B)=|S| FPdim ( B K ) = FPdim ( B ) = ∣ S ∣ , as required by (44) . Rows with the same B B B have different B K B_K B K , in accordance with Proposition 5.3.
Intermediate and final vacuum categories. For transitive S = G / J S=G/J S = G / J , the H H H -orbits on S S S are the double cosets H g J HgJ H g J , and the stabilizer of g J gJ g J in H H H is H ∩ g J g − 1 H\cap gJg^{-1} H ∩ g J g − 1 . Theorem D.1(ii) therefore gives
Ω B Y ≃ ⨁ H g J ∈ H \ G / J D ( H ∩ g J g − 1 ) , (150) \Omega_{B}^{Y}\simeq\bigoplus_{HgJ\in H\backslash G/J}\mathcal{D}(H\cap gJg^{-1}), \tag*{(150)} Ω B Y ≃ H g J ∈ H \ G / J ⨁ D ( H ∩ g J g − 1 ) , (150)
and a general S S S contributes one such sum for each G G G -orbit. For J = H J=H J = H the double cosets are H H H and H t H HtH H t H , with stabilizers H H H and { 1 } \{1\} { 1 } ; this gives the mixed endpoint D ( Z 2 ) ⊕ Vec \mathcal{D}(\mathbb{Z}_2)\oplus\operatorname{Vec} D ( Z 2 ) ⊕ Vec . The final phase is described by Theorem D.1(iii): for transitive S = G / J S=G/J S = G / J ,
Z B K l o c ≃ D ( J ) , \mathcal{Z}^{\mathrm{loc}}_{B_K}\simeq\mathcal{D}(J), Z B K loc ≃ D ( J ) ,
and a general S S S contributes one such factor for each G G G -orbit, so that r c o m p = r f i n = ∣ G \ S ∣ r_{\mathrm{comp}}=r_{\mathrm{fin}=|G\backslash S|} r comp = r fin = ∣ G \ S ∣ . For the mixed endpoint the intermediate phase has two local vacua, D ( Z 2 ) \mathcal{D}(\mathbb{Z}_2) D ( Z 2 ) and Vec \operatorname{Vec} Vec , while the final phase is the connected D ( Z 2 ) \mathcal{D}(\mathbb{Z}_2) D ( Z 2 ) ; the outer direct sum records alternative vacua, not stacked topological orders.
Vacuum overlap as Frobenius reciprocity. Both a K = Fun ( G / H ) ≃ Ind H G 1 a_K=\operatorname{Fun}(G/H)\simeq\operatorname{Ind}^{G}_{H}1 a K = Fun ( G / H ) ≃ Ind H G 1 and B K = Fun ( S ) B_K=\operatorname{Fun}(S) B K = Fun ( S ) lie in the electric subcategory Rep ( G ) ⊂ Z \operatorname{Rep}(G)\subset\mathcal{Z} Rep ( G ) ⊂ Z . Frobenius reciprocity gives
Hom Z ( a K , B K ) ≃ Hom G ( Ind H G 1 , Fun ( S ) ) ≃ Hom H ( 1 , Res H G Fun ( S ) ) = Fun ( S ) H . (151) \operatorname{Hom}_{\mathcal{Z}}(a_K,B_K)\simeq\operatorname{Hom}_{G}(\operatorname{Ind}^{G}_{H}1,\operatorname{Fun}(S))\simeq\operatorname{Hom}_{H}(1,\operatorname{Res}^{G}_{H}\operatorname{Fun}(S))=\operatorname{Fun}(S)^{H}. \tag*{(151)} Hom Z ( a K , B K ) ≃ Hom G ( Ind H G 1 , Fun ( S )) ≃ Hom H ( 1 , Res H G Fun ( S )) = Fun ( S ) H . (151)
Its dimension is ∣ H \ S ∣ = r |H\backslash S|=r ∣ H \ S ∣ = r , which is the vacuum-overlap formula (66) . Likewise Hom Z ( 1 , B K ) = Fun ( S ) G \operatorname{Hom}_{\mathcal{Z}}(1,B_K)=\operatorname{Fun}(S)^{G} Hom Z ( 1 , B K ) = Fun ( S ) G has dimension ∣ G \ S ∣ |G\backslash S| ∣ G \ S ∣ , which is (67) . For S = G S=G S = G , the decompositions (D.19) give
r = dim Hom Z ( 1 ⊕ 2 , 1 ⊕ sgn ⊕ 2 ⊕ 2 ) = 1 + 2 = 3 , r c o m p = dim Hom Z ( 1 , Fun ( G ) ) = 1. r=\dim\operatorname{Hom}_{\mathcal{Z}}(1\oplus2,1\oplus\operatorname{sgn}\oplus2^{\oplus2})=1+2=3,\qquad r_{\mathrm{comp}}=\dim\operatorname{Hom}_{\mathcal{Z}}(1,\operatorname{Fun}(G))=1. r = dim Hom Z ( 1 ⊕ 2 , 1 ⊕ sgn ⊕ 2 ⊕ 2 ) = 1 + 2 = 3 , r comp = dim Hom Z ( 1 , Fun ( G )) = 1.
Vacuum estimates as orbit counting. For x ∈ S x \in S x ∈ S , the orbit–stabilizer relation gives ∣ H x ∣ ∣ Stab H x ∣ = ∣ H ∣ |Hx|\,|\operatorname{Stab}_{H}x|=|H| ∣ H x ∣ ∣ Stab H x ∣ = ∣ H ∣ . Since FPdim D ( F ) = ∣ F ∣ 2 \operatorname{FPdim}\mathcal{D}(F)=|F|^2 FPdim D ( F ) = ∣ F ∣ 2 , the parent vacuum on an H H H -orbit O O O has D O = ∣ Stab H x ∣ = ∣ H ∣ / ∣ O ∣ \mathcal{D}_O=|\operatorname{Stab}_{H}x|=|H|/|O| D O = ∣ Stab H x ∣ = ∣ H ∣/∣ O ∣ , and hence
∑ O ∈ H \ S D O − 1 = 1 ∣ H ∣ ∑ O ∈ H \ S ∣ O ∣ = ∣ S ∣ ∣ H ∣ . (152) \sum_{O\in H\backslash S}\mathcal{D}_O^{-1}=\frac{1}{|H|}\sum_{O\in H\backslash S}|O|=\frac{|S|}{|H|}. \tag*{(152)} O ∈ H \ S ∑ D O − 1 = ∣ H ∣ 1 O ∈ H \ S ∑ ∣ O ∣ = ∣ H ∣ ∣ S ∣ . (152)
For transitive S = G / J S=G/J S = G / J the final phase is D ( J ) \mathcal{D}(J) D ( J ) , so D f i n = ∣ J ∣ = ∣ G ∣ / ∣ S ∣ \mathcal{D}_{\mathrm{fin}}=|J|=|G|/|S| D fin = ∣ J ∣ = ∣ G ∣/∣ S ∣ . With κ K = [ G : H ] \kappa_K=[G:H] κ K = [ G : H ] , Eq. (D.22) gives
κ K ∑ O D O − 1 = [ G : H ] ∣ H ∣ ∣ S ∣ = ∣ G ∣ ∣ S ∣ = D f i n , \frac{\kappa_K}{\sum_O\mathcal{D}_O^{-1}}=\frac{[G:H]|H|}{|S|}=\frac{|G|}{|S|}=\mathcal{D}_{\mathrm{fin}}, ∑ O D O − 1 κ K = ∣ S ∣ [ G : H ] ∣ H ∣ = ∣ S ∣ ∣ G ∣ = D fin ,
which is the residual-order identity (B.5). For the four transitive coordinates the parent dimensions are ( 2 ) (2) ( 2 ) , ( 1 ) (1) ( 1 ) , ( 2 , 1 ) (2,1) ( 2 , 1 ) and ( 1 , 1 , 1 ) (1,1,1) ( 1 , 1 , 1 ) , and the final dimensions are 6 6 6 , 3 3 3 , 2 2 2 and 1 1 1 .
For transitive S = G / J S=G/J S = G / J , the number r = ∣ H \ G / J ∣ = ∣ J \ G / H ∣ r=|H\backslash G/J|=|J\backslash G/H| r = ∣ H \ G / J ∣ = ∣ J \ G / H ∣ counts the J J J -orbits on the [ G : H ] [G:H] [ G : H ] points of G / H G/H G / H . Hence r ≤ κ K r\leq\kappa_K r ≤ κ K , which is (B.4). Equality holds exactly when J J J acts trivially on G / H G/H G / H , that is, when J ⊆ ⋂ g ∈ G g H g − 1 = { 1 } J\subseteq\bigcap_{g\in G}gHg^{-1}=\{1\} J ⊆ ⋂ g ∈ G g H g − 1 = { 1 } . This is the step used in § 5.3 for the complete Higgs class. In the same language the single-vacuum bound (75) reads ∣ J ∣ ⋅ ∣ J \ G / H ∣ ≥ [ G : H ] |J|\cdot|J\backslash G/H|\geq[G:H] ∣ J ∣ ⋅ ∣ J \ G / H ∣ ≥ [ G : H ] . It says that every J J J -orbit on G / H G/H G / H has at most ∣ J ∣ |J| ∣ J ∣ points. Equality holds exactly when J J J acts freely on G / H G/H G / H , that is, when H ∩ g J g − 1 = { 1 } H\cap gJg^{-1}=\{1\} H ∩ g J g − 1 = { 1 } for all g g g , which by (D.20) means that every parent vacuum is intrinsically trivial.
This completes the identification of the two descriptions for ( G , H ) = ( S 3 , Z 2 ) (G,H)=(S_3,\mathbb{Z}_2) ( G , H ) = ( S 3 , Z 2 ) . One step uses the specific pair: the classification in Theorem D.1 needs a K a_K a K to generate Rep ( S 3 ) \operatorname{Rep}(S_3) Rep ( S 3 ) , as shown in its proof. The counting arguments of this subsection use only the orbit–stabilizer relation and Frobenius reciprocity.