Non-Invertible Symmetries in 6d From Green-Schwarz Automorphisms

Non-Invertible Symmetries in 6d from
Green-Schwarz Automorphisms
Fabio Apruzzi⋆ , Sakura Schäfer-Nameki∗ , Alison Warman∗
arXiv:2411.09674v1 [hep-th] 14 Nov 2024
⋆ Dipartimento di Fisica e Astronomia “Galileo Galilei”, Università di Padova,

Via Marzolo 8, 35131 Padova, Italy
⋆ INFN, Sezione di Padova, Via Marzolo 8, 35131 Padova, Italy
∗ Mathematical Institute, University of Oxford,
Andrew Wiles Building, Woodstock Road, Oxford, OX2 6GG, UK
We construct non-invertible symmetries in 6d N = (2, 0) superconformal field theories that arise

from Green-Schwarz (GS) automorphisms, which form abelian or non-abelian groups. Applied
to Z2 , Z3 and S3 GS automorphisms, gives rise to non-invertible duality, triality and S3 -ality
defects, respectively, once combined with stacking symmetry protected topological phases (SPTs)
and gauging 2-form symmetries. We derive the defects and their fusion rules from two distinct
perspectives: from half-space gauging as well as from the Symmetry Topological Field Theory
(SymTFT). This is the first concrete construction of symmetry defects in 6d forming a fusion
5-category whose fusions are intrinsically non-invertible and non-abelian.
Contents
1 Introduction 3
2 6d N = (2, 0) Theories: Polarizations and GS Automorphisms 5

2.1 Absolute Theories with GS Automorphisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 6d N = (2, 0) so(8) Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.1 Defect Group and Polarizations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Green-Schwarz Automorphisms and Dualities . . . . . . . . . . . . . . . . . . . . . . . 7
3 S3 -ality Defects in 6d from Half-space Gauging 10

3.1 Non-invertible Symmetries from Half-space Gauging . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 Non-invertible Duality Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.1 Gauging 2-form Symmetries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2.2 Definition of Duality Interface and Defects . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.2.3 Fusion of Duality Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.3 Non-invertible Triality Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.1 Twisted Gauging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.3.2 Definition of Triality Interfaces and Defects . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Summary of the S3 -ality Defects and Fusions . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 SymTFT Perspective on S3 -ality Defects 24

4.1 SymTFT for the 6d N = (2, 0) so(8) theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.1 7d Chern-Simons Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1.2 GS-duality Web and Topological Manipulations . . . . . . . . . . . . . . . . . . . . . . 27
4.2 SymTFT Construction of Duality and Triality Defects . . . . . . . . . . . . . . . . . . . . . . 29
4.2.1 Invertible Topological Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.2.2 Duality Twist Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.3 Triality Twist Defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.4 Gauging the S3 Symmetry in the Bulk . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
A Computation of Fusion of Triality and Condensation Defects 40

A.1 Fusion of two triality defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
A.2 Fusions between duality and triality defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
A.3 Fusions with condensation defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2
B Details on SymTFT computations 50
B.1 Invertible GS defects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
B.2 Triality twist defect fusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
C Bosonic SymTFT and Interfaces 57
1 Introduction
One of the most surprising insights in the past few years is the existence of a new type of symmetry, called
non-invertible or higher-categorical symmetries [1–3] (for reviews of non-invertible symmetries see [4, 5]). As
the name indicates, these symmetries may not admit inverses, and are therefore strictly beyond the realm of
standard symmetries that form groups. One of the constructions of non-invertible or categorical symmetries
in higher dimensional quantum field theories (QFTs) is inspired by the Kramers-Wannier (KW) duality
symmetry in the 2d critical Ising model [6]. Namely, a theory T which has a duality to a theory T∨ , such
that another topological operation, such as stacking a TQFT and/or gauging maps back to T, will admit a
non-invertible symmetry. In the 2d case this is the map g → 1/g, in the transverse field Ising model
X X
HIsing = − σjz σj+1
z
−g σjx , (1.1)
j j
which becomes a symmetry at g = 1, i.e. the critical Ising conformal field theory (CFT). In d spacetime
dimensions we generically expect a fusion (d − 1)-category as the most general (internal, finite) symmetry
structure. In even d = 2n we expect that a generalization of the KW dualities exists, which is obtained
by combining dualities and gauging of (d/2 − 1)-form symmetries. Indeed, examples are the non-invertible
KW defects in 4d [1, 2, 7, 8]. This construction works in any even dimension, and generalizations of such
duality defects were subsequently studied in various contexts: in 2d, beyond the KW duality defects there
are generalizations to triality and G-ality defects for a finite group G [9, 10]. In 4d, using the Symmetry
Topological Field Theory (SymTFT) [11–13] or holography in [8,14], from branes and string compactifications
[15–17] and in class S-theories [18, 19]. In particular the latter have the feature that the duality symmetry
can form a non-abelian finite group. Using a more mathematical approach the possibilities of duality defects
and generalizations thereof was studied in the context of fusion 3-categories in [20].
Finally in 6d N = (2, 0) theories it was suggested that using Green-Schwarz dualities combined with
stacking TQFTs and/or gauging 2-form symmetries one can construct non-invertible duality and triality
defects in [21]. The purpose of the present paper is to concretely construct the full S3 -ality defects and
crucially to compute their fusion. Achieving this will act as a proper starting point for the exploration of
these fusion 5-categories.
More generally we expect that in any even dimension d = 2n such non-invertible defects will exist that
are G-ality defects for some duality symmetry G. With the huge proliferation of non-invertible symmetries,
3
a fair question is why these G-ality defects are particularly interesting. Non-invertible symmetries are
broadly characterized by either being group-theoretical or not. Group-theoretical (also referred to as “non-
intrinsically non-invertible”) means that the non-invertible symmetry is related by gauging to an invertible,
i.e. group-like, symmetry. This may be a p-form symmetry or a higher-group symmetry. A non-group-
theoretical one is such that there is no gauging that maps it to an invertible symmetry. We should emphasize
that whether or not a non-invertible symmetry is group-theoretical does not diminish or increase its appeal, or
the non-invertible specific physical properties. In fact non-invertible symmetries of any type have interesting
physical implications (e.g. predict new phases, new order parameters).
Nevertheless it is interesting – and challenging in general – to construct non-group-theoretical non-
invertible symmetries. From general considerations, such (non-group theoretical) duality defects will occur
only when the spacetime dimension d is even. On the other hand, a reasonable conjecture is that in d =
(2n + 1) generalized symmetries are (mostly) related by gauging to invertible symmetries – a fact that
seems to be true in 3d bosonic symmetries [22–24]. With duality defects very well known in 4d and 2d,
it is interesting in light of the existence of non-trivial 6d QFTs (and strongly-coupled UV fixed points in
particular) to determine non-group-theoretical non-invertible symmetries in 6d.
Concretely we will use two approaches to construct these non-invertible symmetries:
• Half-space gauging,
• Gauging an outer automorphism of the SymTFT.
The first approach was first introduced for duality and their Z3 generalization, triality, defects and appeared
in [7], which constructs the defects and fusion: pictorially we start with a half-space gauging interface (for a
d-dimensional theory, this can be the gauging of a (d/2 − 1) form symmetry) of codimension 1
D(Md−1 )
(1.2)
T|M <0 (T/G)|M ⩾0
d d
On the left hand side, in the half-space Md<0 we have the original theory T, on the right hand side, after the
insertion of a half-space gauging defect the theory T/G. Combining this with a duality transformation that
maps the theory T/G back to T we obtain a non-invertible duality defect. We will generalize this to triality
and S3 -ality defects in turn.
Our construction is applicable to a large class of 6d theories, but the richest setting, which has the largest
Green-Schwarz duality symmetry, is that of the 6d N = (2, 0) so(8) SCFT. The S3 -ality defects result from
(twisted) gauging of the 2-form symmetry and Green-Schwarz (GS) automorphisms.
We should note that compactification of these theories on a two-torus gives rise to 4d N = 4 Super-Yang
Mills, which only has duality or triality defects. The main reason for this is that the modulus τ of the torus
becomes part of the action of the duality symmetry, and there is no locus where both triality and duality
4
defects can coexist. This is in turn possible for class S constructions, which generically are non-Lagrangian
and can have G-ality defects.
We start in section 2 with the discussion of the 6d SCFTs and their Green-Schwarz dualities as well as the
possible choices of polarizations and consequently, absolute theories. Then we construct the non-invertible
defects in 6d from half-space gauging in section 3, and from the SymTFT point of view in section 4.
A summary of defects and their fusion for S3 -ality defects in 6d is given in subsection 3.4. Note that for
a special case, when the group is Z2 , i.e. duality defects, this is a generalization of the Tambara-Yamagami
construction that is well-known in 2d and 4d, now to 6d.
2 6d N = (2, 0) Theories: Polarizations and GS Automorphisms
6d N = (2, 0) SCFTs of ADE type are specified by an ADE Lie algebra g, whose weight lattice, Λ, describes
string charges under a two-form symmetry [25–28]. For 6d N = (1, 0) SCFTs the lattice of BPS string
charges can be more general. A necessary condition for the lattice is that its Dirac pairing must be positive
definite. These 6d SCFTs are in general relative theories [29, 30], coupled to a 7d bulk: to get an absolute
theory with a well-defined partition function a polarization must be chosen: this ensures mutual locality
(trivial Dirac pairing) of the operators in the absolute theory [21, 27, 29–34]. Generalized symmetries in this
context were studied in [21, 33, 35–38].
Before choosing a polarization, 6d SCFTs are coupled to a 7d bulk anomaly theory, whose action is that
of a 3-form Chern-Simons theory [27, 28, 31, 33]
Z
1 X (I) (J)
SCS(K) = KIJ c3 ∧ dc3 , I, J ∈ {1, ..., r} , (2.1)
4π M7
I,J
(I)
where c3 , I = 1, ..., r are 3-form U (1) gauge fields on M7 . The matrix K is integral, positive definite and
symmetric, and governs the braiding statistics for strings in the boundary 6d system. It can be obtained
geometrically from an F-theory compactification on an elliptically fibered Calabi-Yau threefold as the inter-
section pairing for the base [28, 37]. If the diagonal entries of K are not all even, the manifold M7 requires a
“Wuc structure”, as explained in appendix A of [33]. For 6d SCFTs with N = (2, 0) of ADE type, K is the
Cartan matrix of the ADE Lie algebra g [27, 33].
2.1 Absolute Theories with GS Automorphisms
Green-Schwarz (GS) automorphisms in 6d SCFTs were analyzed in [37]. They are automorphisms, η, of the
BPS string lattice, Λ,
η : Λ → Λ, s.t. η T K η = K. (2.2)
We recall that for the N = (2, 0) case K always corresponds to the Cartan matrix of g, CMg , whereas only
for certain (1, 0) theories this is true. These automorphisms decompose into two parts, O(GS) ⋉ W(GS):
5
1. The inner automorphisms, W(GS), which for the case of K = CMg , correspond to the Weyl group Wg
are redundancies of the tensor branch description. In particular they imply identifications of the tensor
branch moduli space into Weyl chambers,
MT = R(5)T /W , (2.3)
where the (5) stands for the (2, 0) case, since every (2, 0) tensor multiplet has 5 scalars.
2. The outer automorphisms, O(GS), instead, provide self-dualities1 of the 6d (relative) theory. We also
recall that 6d SCFTs do not have marginal deformations and no conformal manifold [39–41]. Therefore
the self-duality will not change any coupling constant. This could be enough to declare that the self-
duality is a symmetry, though 6d theories are generically relative theories. Indeed we will see that
when the 6d theory admits polarizations that lead to absolute theories, these dualities will in fact map
one absolute theory to another. In order to clarify the nomenclature, O(GS) are self-G-ality of the
relative theory, and a G-ality of absolute theories when such polarizations exist.
We recall that the GS automorphisms are read off from the low-energy tensor branch description where the
Dirac pairing appears in the coupling among the various tensor multiplets. Moreover, in the (1, 0) case K
appears also in coupling tensors to vector multiplets. The low-energy tensor branch description of 6d (1, 0)
theories consists of a quiver gauge theory coupled to tensor multiplets. The automorphism is preserved when
it leaves the quiver structure unchanged.
We discuss the types of G-alities that are possible in 6d SCFTs: in [37] a strategy was implemented to
classify automorphisms based on the F-theory classification of 6d SCFTs [42–44]. We skip the details of the
classification and refer to [37]. Non-trivial GS outer automoprhisms O(GS) fall into two main classes:
1. G = Z2 , i.e. duality, these are either reflection of the Dirac pairing K, or exchange of the auter
automorphisms of the K = Cso(2N ) (with N > 4) and K = Ce6 .
2. G = S3 , i.e. S3 -ality, these are the automorphisms of K = Cso(8) .
In this paper we will in particular construct the S3 -ality non-invertible symmetries for 6d SCFTs which
have K = Cso(8) . This example admits polarizations that lead to absolute theories. In addition, the outer au-
tomorphism group provides the most general G-ality that can appear, therefore the defects that we construct
are quite general and their construction can be directly applied to all the other examples that have G-ality
and polarizations leading to absolute theories. For example S3 -ality will contain duality defects. They will
also appear in 6d theories of AN −1 -type with N = n2 where n ∈ N.
1
We use the word duality even though we will soon discover that this is really a G-ality.
6
2.2 6d N = (2, 0) so(8) Theory
A particularly interesting case is the 6d N = (2, 0) SCFT associated to the Lie algebra so(8), as its auto-
morphism group is the non-abelian group of permutations on three elements: S3 . In particular we will show
that this implies that there are non-invertible S3 -ality defects. These are known to exist in 2d QFTs, but
so far not in higher dimensions (though in theory certain non-Lagrangian theories may have such S3 -ality
defects in 4d, there are no completely explicit constructions including fusions known to us).
2.2.1 Defect Group and Polarizations
The defect group for the 6d N = (2, 0) so(8) relative theory is [21, 26, 33, 37]
D = Z2 ⊕ Z2 . (2.4)
This is easily derived from the Cartan matrix of so(8), which identifies the Dirac pairing, and computing the
Smith normal form thereof. We can derive three polarizations L, which satisfy D = L ⊕ L. L can be chosen,
one for each of the three Z2 subgroups of the defect group (2.4), which we label by their generators:
Polarization: LS = (1, 0) LC = (0, 1) LV = (1, 1) (2.5)

Absolute theory: Ss(8) Sc(8) SO(8) (2.6)
However, as stressed in [21, 33], fixing a subgroup L of D in equation (2.4) is not sufficient to fully specify an
absolute theory: one ought to also specify L. This can also be seen by similar reasoning to the 4d discussion
of [45], with the difference that in 6d the Dirac pairing is symmetric and we must specify the weights not of
lines but of 2d operators. This implies the existence of six absolute 6d N = (2, 0) so(8) theories:

Ss(8)+ = LS , LC , Sc(8)+ = LC , LS , SO(8)+ = LV , LS ,

Ss(8)− = LS , LV , Sc(8)− = LC , LV , SO(8)− = LV , LC . (2.7)
These six theories are related by a web of topological manipulations (gauging, possibly after stacking with
an SPT). Furthermore, they are related by Green-Schwarz dualities. This combination of facts allows us to
construct non-invertible G-ality defects in these theories.
2.2.2 Green-Schwarz Automorphisms and Dualities
A key ingredient for the construction of non-invertible defects is the existence of dualities. In this case we
will see that there are dualities mapping between different polarizations. These dualities have closed orbits,
and combined form a group S3 = Z3 ⋊ Z2 .

Let us consider an absolute theory with polarization pair LJ , LI , I ̸= J ∈ {S, C, V }, defined on the
(I)
6-dimensional spacetime M6 , coupled to classical background fields for the global 2-form symmetry Z2 :
(I)
C3 ∈ H 3 (M6 , Z2 ) . (2.8)
7
SO(8)+ Ss(8)+

LV , LS LS , LC
SO(8)− Sc(8)+

LV , LC LC , LS
Ss(8)−

LS , LV
Sc(8)−

LC , LV
Figure 1: The Green-Schwarz automorphisms of the relative 6d (2, 0) SCFTs with algebra so(8). GS2 (whose
action is shown as black dashed arrows) is of order 2 and exchanges S ↔ C, whereas GS3 (shown in black
arrows) is of order 3 and cyclically permutes S, V, C. The operations shown in cyan and blue are gauging
the 2-form symmetry (cyan) and stacking a TQFT (blue).
(I)
We denote the partition function by ZTJ [C3 ], where the superscript I denotes the global 2-form symmetry
whereas the subscript J specifies the polarization, which, as we will recall shortly, corresponds to the dual
(J) (I)
2-form symmetry Z2 one obtains after gauging Z2 .
Green-Schwarz (GS) Dualities. The GS automorphisms act on the polarization pairs and fall into two
types: order 2 and order 3. The order 2 GS autorphisms act as follows:
GS2(V ) : S ↔ C, V ↔V
Z2 : GS2(S) : C ↔ V, S↔S (2.9)
GS2(C) : S ↔ V, C↔C
and those of order 3 are cyclic or anti-cyclic permutations
GS3 : V →C→S→V
Z3 : (2.10)
GS3 : V →S→C→V .
From this action on the polarizations we can infer the action on the absolute theories as shown in figure 1 in
terms of the black solid/dashed lines. It is important to note that the combination gives rise to an S3 group
8
(S) (V ) (C)
⊗ D5 D5 D5 T5 T5
(S) (id) (C) (V )
D5 D5 T5 T5 D5 D5
(V ) (id) (S) (C)
D5 T5 D5 T5 D5 D5
(C) (id) (V ) (S)
D5 T5 T5 D5 D5 D5
(V ) (C) (S) (id)
T5 D5 D5 D5 T5 D5
(C) (S) (V ) (id)
T5 D5 D5 D5 D5 T5
Table 1: Fusion table of the invertible GS interfaces: D denotes the order 2 duality interfaces, T the order
3 triality ones.
of automorphisms, where
S3 = {id, a, a2 , a3 , b, ab, a2 b} (2.11)
satisfying
a3 = b2 = id , ba2 = ab . (2.12)
We can identify GS2 with b and GS3 with a and check the third relation above:
GS2 GS3 GS3
Sc(8)+ −→ Ss(8)+ −→ SO(8)+ −→ Sc(8)− (2.13)
which equals
GS3 GS2
Sc(8)+ −→ Ss(8)− −→ Sc(8)− (2.14)
More precisely, the GS automorphpisms define interfaces between theories with different polarizations.
We denote the interfaces implementing the GS2(I) , GS3, GS3 automorphisms (where I ∈ {S, C, V }) defined
(I)
in equations (2.9)-(2.10) by D5 (M5 ), T5 (M5 ), T 5 (M5 ) respectively: they are supported on a 5-manifold M5
between two different absolute theories, which are related by the corresponding Green-Schwarz duality.
(I)
It follows from (2.9) that the corresponding interfaces D5 are invertible of order 2:
(V ) (V ) (id) (S) (S) (id) (C) (C) (id)
D5 ⊗ D5 = D5 , D5 ⊗ D5 = D5 , D5 ⊗ D5 = D5 , (2.15)
whereas the GS3 imply that the interfaces T5 , T 5 are invertible of order 3:
(id) (id)
T5 ⊗ T5 ⊗ T5 = D5 , T 5 ⊗ T 5 ⊗ T 5 = D5 (2.16)
The fusions of the GS defects follow the S3 group multiplication, which we summarize in table 1.
Gauging and stacking. The construction of duality/triality and in this case S3 -ality defects requires
further that we have operations that can be used to construct maps that “invert” the action of the GS
dualities. For this we define the topological operations:
(I)
σ: gauging of the global 2-form symmetry Z2
(2.17)
τ: stacking a 6d Z2 SPT phase .
9
More concretely: σ acts on the polarization pairs of the theory as

σ: LJ , LI → LI , LJ , (2.18)
(I)
whereas τ consists of stacking a 6d SPT phase with Z2 symmetry on the partition function of the LJ , LI

theory, giving a theory with LJ , LK , K ̸= J, I symmetry (stacking with an SPT does not change the
polarization but changes the second element of the polarization pair). Gauging after stacking with an SPT is
called “twisted” gauging. These 6d topological manipulations were analyzed in [33] and described in terms
of polarization pairs in [21].
Non-Invertible Symmetry Defects. Combining a Green-Schwarz automorphism and a topological ma-

nipulation (gauging, possibly with twist) gives rise to duality and triality defects, and in the present case
to S3 -ality ones, within a single absolute theory. We will derive them both from the “half-space” gauging
approach in section 3 and from the SymTFT in section 4.2. We will in particular compute the fusion rules of
the defects implementing these symmetries, which are an important step in identifying the symmetry fusion
5-category.
3 S3 -ality Defects in 6d from Half-space Gauging

3.1 Non-invertible Symmetries from Half-space Gauging
Non-invertible duality and triality symmetry defects satisfy fusion rules of order two or three respectively
with non-trivial condensation defects and TQFT coefficients. The half-space gauging approach to construct
non-invertible duality defects in 4d was used in [2, 7, 8, 46]. Non-invertible triality defects from half-space
twisted gauging (when the gauging occurs after stacking with a non-trivial SPT) were studied in 4d in [47].
Recently, this approach for non-invertible duality and triality symmetries has also been employed for 3d
theories [48].
In this section, we will apply the half-space (twisted) gauging method to construct and analyze in detail
duality, triality, and S3 -ality defects in the 6d N = (2, 0) so(8) SCFT. For duality defects, we will firstly
review the operation σ of gauging the 2-form global symmetry of an absolute 6d theory. Then, following
the above cited works in lower dimensions, we define a codimension-1 half-space gauging interface σ(M5 )
between a 6d absolute theory and the theory obtained from it by means of the σ-operation.
Denote by x the coordinate on the 6d spacetime M6 in the direction locally orthogonal to the 5-manifold
M5 supporting the half-space gauging interface σ(M5 ), and separate this into two regions x < 0 and x ⩾ 0.
Schematically this will be depicted as follows:
σ(M5 )
(3.1)
T|M <0 (T/Z2 )|M ⩾0
6 6
10
On M6<0 the theory is T, whereas on M6⩾0 lies the theory obtained from it by gauging its Z2 2-form global
symmetry. The half-space gauging interface σ(M5 ) separates them at M5 = ∂M6⩾0 : it is obtained by gauging
the Z2 2-form global symmetry of T in half of spacetime while imposing the Dirichlet boundary condition for
the dynamical gauge field on the interface M5 ; on the remaining half of spacetime we have the theory with
the gauged symmetry T/Z2 and M5 thus supports a codimension-1 interface σ(M5 ). Since we will concretely
perform computations in terms of partition functions, we will schematically depict the interfaces between
effective Lagrangian densities, like in reference [7], rather than abstract theories, e.g.
σ(M5 )
(I) (I) (I) (J) (3.2)
LTJ [C3 ] LTJ [c3 ] + π(c3 ∪ C3 )
(I)
c3 |M5 =0
The theory on the left-hand side has polarization LJ , with J ∈ {S, C, V }, so we denote it as TJ and its
effective Lagrangian density, that we use only to describe its Z2 2-form symmetry, as LTJ .2 The effective
(I) (I)
Lagrangian is a function of background fields C3 for the Z2 2-form global symmetry of the theory, where
(I) (I)
I ̸= J ∈ {S, C, V }. The interface σ(M5 ) gauges Z2 by making c3 dynamical with the Dirichlet boundary
(J) (J)
condition for it imposed on M5 and couples it to a background field C3 for the dual Z2 global symmetry.
The precise equation corresponding to the above schematic figure is (3.11).
Triality interfaces can instead be constructed by gauging in combination with stacking the partition
function with a Z2 TQFT. This requires a quadratic refinement of the intersection paring: we write it as
the integral of a 6-form q̄, which gives a Z2 -valued quadratic function of the 3-form background fields for
the 2-form Z2 symmetry. When these fields are classical backgrounds (i.e. the 2-form symmetry is global)
(I)
we denote them by an uppercase letter, e.g. C3 , I ∈ {S, C, V }, whereas when they are dynamical (i.e. the
symmetry is being gauged and the fields are summed-over in the path integral) we use lowercase. Therefore,
(I)
if we perform twisted gauging (where we fist stack and then gauge) of a Z2 symmetry, we will encounter
(I) (K) (K)
q̄(c3 ), whereas if we stack after gauging, then q̄(C3 ) will appear, where Z2 is the global symmetry of the
resulting theory. By performing these operations in half of spacetime we will define the interfaces τ σ(M5 )
and στ (M5 ) respectively.
Defects. To define duality/triality/S3 -ality defects in a given absolute theory, we combine:
• an order-2 Green–Schwarz duality interface with a half-space gauging interface σ(M5 ) to construct a
non-invertible duality defect D5 (M5 ),
• an order-3 Green–Schwarz duality interface with a half-space twisted gauging interface τ σ(M5 ) or
gauging and stacking interface στ (M5 ) to construct non-invertible triality defects T5 (M5 ) and T 5 (M5 ).
2
Note that 6d SCFTs are generically Non-Lagrangian theories, i.e. they do not admit a description in terms of a local
Lagrangian density. However, we can still use an effective Lagrangian to describe only their discrete 2-form symmetry.
11
Fusion. To compute the (parallel) fusion of two codimension-1 defects, following [8], we place the first at
x = 0 and the second at x = ϵ. Schematically this looks as follows (for the precise version of this equation
see (3.17)):
D5 (M5|0 ) D5 (M5|ϵ )
(J) (I) (J)
TJ |M <0 (TI /Z2 )|M [0,ϵ] (TJ /Z2 /Z2 )|M >ϵ (3.3)
6 6 6
x=0 x=ϵ
spacetime M6 is divided into three regions:
• M6<0 hosts the initial theory,

[0,ϵ]
• M6 constitutes the region between the two defects whose fusion we are computing. We denote
its boundaries by M5|0 and M5|ϵ to indicate that they are 5-manifolds located at x = 0 and x = ϵ
[0,ϵ]
respectively. The theory on M6 is obtained from the starting one by acting only with the first defect.
• M6>ϵ hosts the final theory, given by acting on the initial one with both defects sequentially.
For these computations, we will denote the theories by their effective Lagrangian densities, since the action
of the symmetry defects is given concretely on them. After performing appropriate manipulations3 we will
take ϵ → 0 and write the resulting fusion.
3.2 Non-invertible Duality Defects
We now consider the duality GS-dualities GS2, and combine them with gauging/stacking operations of order
two to obtain non-invertible duality symmetries. First we discuss the operation σ of gauging the 2-form
symmetry, which will be used in the construction.
3.2.1 Gauging 2-form Symmetries

If one starts with the absolute 6d N = (2, 0) so(8) theory with polarization pair LJ , LI the operation σ
(I)
of gauging its global 2-form Z2 symmetry without any twist gives rise to a theory with polarization pair
(J)
LI , LJ and a dual Z2 global 2-form symmetry by acting as follows on the partition function:
σ
LJ , LI −
→ LI , LJ
Z
(I) σ (I) (I) (J)
X
ZTJ [C3 ] 7−
→ N(M6 , Z2 ) ZTJ [c3 ] exp iπ c3 ∪ C3 . (3.4)
(I) M6
c3 ∈H 3 (M6 , Z2 )
3
For example, integrating out dynamical fields, using properties of the quadratic refinement, cohomology isomorphisms and
Poincaré duality.
12
(I)
On the right-hand side of equation (3.4), one makes the gauge field for the Z2 2-form symmetry dynamical
(I)
and sums over all field configurations weighted by a term coupling the dynamical gauge field c3 to the
(J) (J)
classical background C3 for the dual Z2 2-form symmetry. The normalization factor
|H 1 (M6 , Z2 )|
N(M6 , Z2 ) = (3.5)
|H 2 (M6 , Z2 )| |H 0 (M6 , Z2 )|
takes into account the volume of the gauge redundancy of the 2-form symmetry being gauged on M6 .
(J)
The resulting absolute theory, with polarization pair LI , LJ , has a global Z2 2-form symmetry: by

gauging it, we return to the original theory LJ , LI . We can check this by applying the σ operation (3.4)
twice:
Z
σ2

(I) (I) (I) (J) (J) (I)
X
2
ZTJ [C3 ] 7−→ N(M 6 , Z2 )
ZTJ [c3 ] exp iπ c3 ∪ c3 + c3 ∪ C3 . (3.6)
(I) (J) M6
c3 , c3 ∈H 3 (M6 , Z2 )
(J) (I) (I)

In the above expression, we can integrate out c3 : this enforces c3 = C3 which trivializes the exponential
and produces a factor of |H 3 (M6 , Z2 )| showing that gauging twice maps the partition function to itself, up
to an Euler counterterm:4
(I) σ2 (I)
ZTJ [C3 ] 7−→ χ−1
(M6 , Z2 ) ZTJ [C3 ] , (3.8)
which means that

σ 2 = χ−1
(M6 , Z2 ) . (3.9)
The normalization N(M6 , Z2 ) in the gauging operation σ could be re-defined by a power of the Euler coun-
1/2
terterm χ(M6 , Z2 ) : in particular multiplying the partition function in (3.4) by χ(M6 , Z2 ) would eliminate the
counterterm from (3.9) and give σ 2 = id. However, in the following, we will maintain the normalization of
equation (3.4), which enables us to keep track of the volume of gauge redundancies.
3.2.2 Definition of Duality Interface and Defects
Half-space gauging interface σ(M5 ). A codimension-1 half-space gauging interface σ(M5 ) between the
(I)
6d absolute theories with polarization pairs LJ , LI and LI , LJ can be constructed by gauging the Z2
2-form symmetry of the former theory, equation (3.4), in half of spacetime while imposing the Dirichlet
(I)
boundary condition for the dynamical gauge field c3 on the interface M5 . Schematically:
σ(M5 )
(I) (I) (I) (J) (3.10)
LTJ [C3 ] LTJ [c3 ] + π(c3 ∪ C3 )
(I)
c3 |M5 =0
4
For the normalization, recall the definition of N(M6 , Z2 ) in eq. (3.5), the property |H n (M6 , Z2 )| = |H 6−n (M6 , Z2 )| and the
definition of the Euler counterterm
|H 0 (M6 , Z2 )| |H 2 (X6 , Z2 )| |H 4 (M6 , Z2 )| |H 6 (X6 , Z2 )|
χ(M6 , Z2 ) = . (3.7)
|H 1 (X6 , Z2 )| |H 3 (M6 , Z2 )| |H 5 (X6 , Z2 )|
13
The above picture translates to the following equation:
Z !
(I) (I) (I) (J)
X
ZTJ [C3 ; σ(M5 )] = N(M ⩾0 , Z2 ) ZTJ [c3 ] exp iπ c3 ∪ C3 (3.11)
6
(I) M6⩾0
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
where on the left hand side we denote the partition function with the insertion of the σ(M5 ) defect and
|H 1 (M6⩾0 , M5|0 , Z2 )|
N(M ⩾0 , Z2 ) = (3.12)
6 |H 2 (M6⩾0 , M5|0 , Z2 )| |H 0 (M6⩾0 , M5|0 , Z2 )|
is the appropriate normalization factor.5

Duality defect D 5 (M5 ). In a fixed absolute theory with polarization pair LJ , LI , a non-invertible
duality defect D5 (M5 ) can be constructed by composing a GS2 duality and half-space gauging [21]:
(K)
D5 (M5 ) = D5 (M5 ) ⊗ σ(M5 ) , I ̸= J ̸= K ∈ {S, C, V } . (3.13)
(K)
First the GS2(K) duality (with K ̸= I, J), realized by D5 (M5 ), exchanges the labels I ↔ J and then the
(J)
σ(M5 ) interface implements gauging of the new global Z2 2-form symmetry:
(K)
D (M5 ) σ(M5 )
−−5−−−−→ LI , LJ

LJ , LI −−−−→ LJ , LI
(K)
(I) D (M5 ) (J) σ(M5 ) (I)
ZTJ [C3 ] 7−−5−−−−→ ZTI [C3 ] 7−−−−→ χ−1/2 ZTJ [C3 ] . (3.14)

D5 (M5 ) is thus a defect in a given absolute theory with polarization pair LJ , LI , and as we shall see below,
obeys non-invertible fusion rules. Recalling the definition of the σ(M5 ) duality interface (3.10) and the fact
(K)
that D5 (M5 ) exchanges I ↔ J, the action of the duality defect D5 (M5 ) can be depicted as follows:
D5 (M5 )
(I) (J) (J) (I) (3.15)
LTJ [C3 ] LTI [c3 ] + π(c3 ∪ C3 )
(J)
c3 |M5 =0
which is spelled out in terms of the partition function as

Z !
(I) (J) (J) (I)
X
ZTJ [C3 ; D5 (M5 )] = N(M ⩾0 , Z2 ) ZTI [c3 ] exp iπ c3 ∪ C3 . (3.16)
6
(J) M6⩾0
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
3.2.3 Fusion of Duality Defects
We now compute the fusion of two duality defects, defined in equations (3.15)-(3.16), located at x = 0 and
x = ϵ respectively, and then take ϵ → 0. As described in the summary, the two interfaces divide spacetime
5
Recall that, when considering a manifold with boundary, cohomology becomes relative but holomology is still absolute [8].
14
[0,ϵ]
M6 into three regions, M6<0 , M6 , M6>ϵ . Schematically this is depicted as:
D5 (M5|0 ) D5 (M5|ϵ )

(I) (J) (J) (I) (I) (I) (J) (J) (I)
LTJ [C3 ] LTI [c3 ] + π(c3 ∪ C3 ) LTJ [c3 ] + π c3 ∪ c3 + c3 ∪ C3 .
(J)
c(J) |M5|0 =0 c3 |M5|ϵ =0
(3.17)
From its definition in equation (3.13), the duality defect D5 (M5|ϵ ) first acts with a Green-Schwarz interface
(K) (J) (K) (I)
D5 at M5|ϵ : when c3 passes through D5 (M5|ϵ ), it gets converted to c3 ,
(J) (K) (K) (I)
c3 |M5|ϵ D5 (M5|ϵ ) = D5 (M5|ϵ ) c3 |M5|ϵ . (3.18)
(J)
We then impose the Dirichlet boundary condition on c3 when performing half-space gauging by means of
(J) [0,ϵ] (J)
σ(M5|ϵ ): we can therefore split it into c3 ∈ H 3 (M6 , M5|0 ∪ M5|ϵ , Z2 ) and c3 ∈ H 3 (M6⩾ϵ , M5|ϵ , Z2 ).
(K)
Since D5 is invertible, equation (3.18) implies that we must also impose the Dirichlet boundary condition
(I)
for c3 on M5|ϵ . The equation corresponding to (3.17) is thus:
Z !
(I) (J) (I)
X
ZTJ [C3 ; D5 (M5|0 ), D5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (J) (J) (I)
X
2
×N(M ⩾ϵ ZTJ [c3 ] exp iπ c3 ∪ c3 + c3 ∪ C3 . (3.19)
6 , Z2 ) M6⩾ϵ
(I) (J)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(J) (I) (I) (I)

Integrating out c3 on M6⩾ϵ enforces c3 = C3 and leaves the partition function ZTJ [C3 ] on M6⩾ϵ .6
Z !
(I) (J) (I)
X
ZTJ [C3 ; D5 (M5|0 ), D5 (M5|ϵ )] =N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
(I)
× χ−1
(M ⩾ϵ , Z ZTJ [C3 ] . (3.21)
6 2)
We now use the isomorphism (which is described for n-cohomologies in appendix B of [1])
[0,ϵ]
H 3 (M6 , M5|0 ∪ M5|ϵ , Z2 ) → H 2 (M5 , Z2 ) (3.22)
(J)
to map c3 7→ b2 ∈ H 2 (M5 , Z2 ) and write in the ϵ → 0 limit:7
Z
(I)
X
D5 (M5 ) ⊗ D5 (M5 ) =χ−1(M ⩾0 ,Z )
N(M 5 , Z2 ) exp iπ b2 ∪ C3 . (3.25)
6 2
M5
b2 ∈H 2 (M5 , Z2 )
6
For the normalization, we have defined
[0,ϵ]
|H 1 (M6 , M5|0 ∪ M5|ϵ , Z2 )|
N(M [0,ϵ] , Z = [0,ϵ] [0,ϵ]
. (3.20)
2)
6 |H 2 (M6 , M5|0 ∪ M5|ϵ , Z2 )| |H 0 (M6 , M5|0 ∪ M5|ϵ , Z2 )|
(J)
Integrating out c3 yields a factor of |H 3 (M6⩾ϵ , M5|ϵ , Z2 )|, which when combined with N(M
2
⩾ϵ
,Z )
, produces the Euler countert-
6 2
erm χ−1 ,
defined in equation (3.7).
(M6 ⩾ϵ , Z2 )
7
For the normalization we use the properties
[0,ϵ] [0,ϵ]
|H 0 (M6 , M5|0 ∪ M5|ϵ , Z2 )| = 1, |H n (M6 , M5|0 ∪ M5|ϵ , Z2 )| = |H n−1 (M5 , Z2 )| , (3.23)
15
We then use Poincaré duality on M5 to convert the sum over b2 ∈ H 2 (M5 , Z2 ) to one over M3 = PD(b2 ) ∈
H3 (M5 , Z2 ) and obtain:
Z
(I)
X
D5 (M5 ) ⊗ D5 (M5 ) = χ−1
(M6 ⩾0 , Z2 )
N(M5 , Z2 ) exp iπ C3 , (3.26)
M3 ∈H3 (M5 , Z2 ) M3
whose right hand side is a condensation defect (we include the Euler counterterm χ−1
(M ⩾0 , Z ) in its normal-
6 2
ization) Z
(0) (I)
X
C5 (M5 ) = χ−1
(M6 ⩾0 , Z2 )
N(M5 , Z2 ) exp iπ C3 (3.27)
M3 ∈H3 (M5 , Z2 ) M3
of the 2-form symmetry defect Z

(I) (I)
D3 (M3 ) = exp iπ C3 (3.28)
M3
(0)
without discrete torsion (hence the 0 superscript in C5 (M5 )). Its normalization is (up to the Euler coun-
terterm)
|H 0 (M5 , Z2 )|
N(M5 , Z2 ) = , (3.29)
|H 1 (M5 , Z2 )|
(I)
(derived in footnote 7) which correctly takes into account the gauge redundancy from condensing D3 (M3 )
on M5 , on which it generates a 1-form symmetry. The fusion of two duality defects is therefore:
(0)
D5 (M5 ) ⊗ D5 (M5 ) = C5 (M5 ) . (3.30)
3.3 Non-invertible Triality Defects
We now consider the triality GS-dualities GS3, and combine them with order three gauging/stacking opera-
tions to obtain non-invertible triality symmetries.
3.3.1 Twisted Gauging
A 3-form background gauge field C3 for a 2-form global symmetry has odd form degree, therefore:
C3 ∪ C3 = −C3 ∪ C3 ⇒ 2 C3 ∪ C3 = 0 . (3.31)
R
This means that M6 C3 ∪ C3 is zero unless one takes the coefficients to be in a ring without a multiplicative
inverse of 2, such as Z2 = {0, 1} [21]. A possible 6d counterterm one can stack onto the partition function
before gauging can be defined by means of a quadratic refinement of the intersection pairing, notions which
we will now review, following [33, 49]. Recall that, if M6 is an oriented 6-manifold, the intersection form:
Z
3 3
H (M6 , Z2 ) × H (M6 , Z2 ) → Z2 : (C3 , C̃3 ) 7→ C3 ∪ C̃3 (mod 2) (3.32)
M6
which imply
[0,ϵ]
|H 1 (M6 , M5|0 ∪ M5|ϵ , Z2 )| |H 0 (M5 , Z2 )|
N(M [0,ϵ] , Z = → N(M5 , Z2 ) = . (3.24)
6 2) [0,ϵ]
|H 2 (M6 , M5|0 ∪ M5|ϵ ,
[0,ϵ]
Z2 )| |H 0 (M6 , M5|0 ∪ M5|ϵ , Z2 )| |H 1 (M5 , Z2 )|
16
can have a quadratic refinement8 which we write as the integral of a density q̄
Z
q̄ : H 3 (M6 , Z2 ) → Z2 (3.33)
M6
such that Z Z
q̄(C3 + C̃3 ) = q̄(C3 ) + q̄(C̃3 ) + C3 ∪ C̃3 (mod 2) . (3.34)
M6 M6
Note that q̄ is a special case of a Z4 -valued quadratic refinement q that satisfies:
Z Z
q(C3 + C̃3 ) = q(C3 ) + q(C̃3 ) + 2 C3 ∪ C̃3 (mod 4) , (3.35)
M6 M6
by setting C̃3 = 0 one learns that q(0) = 0 and by then taking C̃3 = C3 , one has q(C3 ) = C3 ∪ C3 . The
relation between the Z4 -valued q and Z2 -valued q̄ is q = 2q̄.
From the Z2 -valued quadratic refinement one can define a 3-form Z2 gauge theory with the following
topological action [33]: Z
π q̄(C3 ) (3.36)
M6
and the Arf–Kervaire invariant - AK [51–54], that is valued in Z2 = {0, 1}, can be expressed as
R
q̄(c3 )
X
πAK = Arg (−1) M6 . (3.37)
c3 ∈H 3 (M 6 , Z2 )

If one starts with the 6d absolute theory specified by the polarization pair LJ , LI and corresponding
(I)
partition function ZTJ [C3 ], one can define the operation τ [7, 49] which acts on the partition function by
stacking with (3.36). This operation does not affect the polarization L but changes L (the second entry of
the polarization pair) and hence the global symmetry of the theory. This follows from appendix B of [21]
and the earlier discussion in section 4.4 of [33].
τ
LJ , LI −→ K ̸= I, J ∈ {S, C, V } ,
LJ , LK ,
Z
(I) τ (I) (I) (K)
ZTJ [C3 ] 7−→ ZTJ [C3 ] exp iπ q̄(C3 ) = ZTJ [C3 ] . (3.38)
M6
The combined “twisted gauging” operation τ σ of stacking (3.38) and then gauging (3.4) acts as follows:
τσ
LJ , LI −→ LK , LJ , K ̸= I, J ∈ {S, C, V } ,
Z
(I) τσ (I) (I) (I) (J)
X
ZTJ [C3 ] 7−→ N(M6 , Z2 ) ZTJ [c3 ] exp iπ q̄(c3 ) + c3 ∪ C3 (3.39)
(I) M6
c3 ∈H 3 (M6 , Z2 )
−1/2 (J)
= χ(M6 , Z2 ) ZTK [C3 ] . (3.40)
8
Note that the existence of a quadratic refinement requires specific dimension-dependent spacetime structures: on general
manifolds it is Z4 -valued but for orientable manifolds, it is Z2 -valued (see e.g. appendix B of [33]). We will follow the conventions
in e.g. section 4.4 and appendix B of [33] and section 2.4 of [21] by taking a Z2 -valued quadratic refinement. For the Z4 -valued
convention and the relation with the Z2 -valued one, see e.g. section 2.1 of [49] or appendix B of [50].
17
(I)
We have gauged the Z2 global symmetry of the starting theory after having stacked it with a non-trivial
(J) (K) (I)
SPT: this gives rise to a theory with a global Z2 symmetry and a dual Z2 symmetry (which is not Z2
due to the twisted gauging). N(M6 , Z2 ) is the normalization factor given in equation (3.5).
(I)
Starting from the theory with partition function ZTJ [C3 ] and gauging (3.4) then stacking with (3.36)
defines the operation στ :
στ
LJ , LI −→ LI , LK , K ̸= I, J ∈ {S, C, V } ,
Z
(I) στ (I) (I) (K) (K)
X
ZTJ [C3 ] 7−→ N(M6 , Z2 ) ZTJ [c3 ] exp iπ c3 ∪ C3 + q̄(C3 ) (3.41)
(I) M6
c3 ∈H 3 (M6 , Z2 )
−1/2 (K)
= χ(M6 , Z2 ) ZTI [C3 ]. (3.42)
We note that στ and τ σ are inverses of each other. This can be seen in terms of polarization pairs
στ τσ

LJ , LI −→ LI , LK −→
LJ , LI ,
τσ στ
LJ , LI −→ LK , LJ −→ LJ , LI , (3.43)
as well as from the partition functions

R (I) (K) (K) (K) (K) (I)
(I) (I) iπ c3 ∪c3 +q̄(c3 )+q̄(c3 )+c3 ∪C3
X
2
Z(στ τ σ) TJ [C3 ] = N(M 6 , Z2 )
ZTJ [c3 ] e M6
, (3.44)
(I) (K)
c3 , c3 ∈H 3 (M6 , Z2 )

R (I) (I) (J) (J) (I) (I)
(I) (I) iπ q̄(c3 )+c3 ∪c3 +c3 ∪C3 +q̄(C3 )
X
2
Z(τ σστ ) TJ [C3 ] = N(M 6 , Z2 )
ZTJ [c3 ] e M6
. (3.45)
(I) (J)
c3 , c3 ∈H 3 (M6 , Z2 )
R (K) (J)
Using the fact that M6 q̄ is Z2 -valued (3.33) and integrating out c3 in the first case and c3 in the second
(I) (I)
sets c3 = C3 giving:
(I) (I) (I)

Z(στ τ σ) TJ [C3 ] = Z(τ σστ ) TJ [C3 ] = χ−1
(M6 , Z2 ) ZTJ [C3 ] (3.46)
Both τ σ and στ are order-3 operations (with opposite cyclicity). This can be seen both in terms of
polarization pairs [21]:
τσ τσ τσ
LJ , LI −→ LK , LJ −→ LI , LK −→ LJ , LI (3.47)
στ στ στ
LJ , LI −→ LI , LK −→ LK , LJ −→ LJ , LI (3.48)
18
and from partition functions [49]. Indeed, the action of (τ σ)2 is:

R (I) (I) (J) (J) (J) (K)
(I) (I) iπ q̄(c3 )+c3 ∪c3 +q̄(c3 )+c3 ∪C3
X
2
Z(τ σ)2 TJ [C3 ] = N(M 6 , Z2 )
ZTJ [c3 ] e M6
=
(I) (J)
c3 , c3 ∈H 3 (M6 , Z2 )

R (I) (J) (K) (I) (K) (K)
(I) iπ q̄(c3 +c3 +C3 )+c3 ∪C3 +q̄(C3 )
X
2
= N(M 6 , Z2 )
ZTJ [c3 ] e M6
=
(I) (J)
c3 , c3 ∈H 3 (M6 , Z2 )
 

R (J) R (I) (K) (K)
iπ q̄(c̃3 )  (I) iπ c3 ∪C3 +q̄(C3 )
X
= N(M6 , Z2 ) e M6
 N(M6 , Z2 ) ZTJ [c3 ] e M6
, (3.49)

(J)
c̃3 ∈H 3 (M6 , Z2 )
where we used the property of q̄, eq. (3.34), the fact that the fields are Z2 cocycles (so their sign is irrelevant)
(J) (J) (I) (J) (K)
and we replaced the sum over c3 with one over c̃3 = c3 + c3 + C3 . In the second factor we recognize
the action of στ , therefore
(I) (I)
Z(τ σ)2 TJ [C3 ] = Z(Y,M6 ) Z(στ )TJ [C3 ] , (3.50)
where
R
iπ q̄(c3 )
X
Z(Y,M6 ) = N(M6 , Z2 ) e M6 (3.51)
c3 ∈H 3 (M6 , Z2 )
is (up to normalization) the exponential of the Arf–Kervaire invariant of the quadratic refinement of the
intersection form [33, 49].
Similarly, for (στ )2 we have:

R (I) (K) (K) (K) (J) (J)
(I) (I) iπ c3 ∪c3 + q̄(c3 )+c3 ∪C3 +q̄(C3 )
X
Z(στ )2 TJ [C3 ] = N(M6 , Z2 ) ZTJ [c3 ] e M6
=
(I) (K)
c3 , c3 ∈H 3 (M6 , Z2 )

R (I) (K) (J) (I) (I) (J)
(I) iπ q̄(c3 +c3 +C3 )+q̄(c3 )+c3 ∪C3 )
X
2
= N(M6 , Z2 )
ZTJ [c3 ] e M6
=
(I) (K)
c3 , c3 ∈H 3 (M6 , Z2 )
 

R (K) R (I) (I) (J)
iπ q̄(c̃3 ) (I) iπ q̄(c3 )+c3 ∪C3 )
X
= N(M6 , Z2 ) e M6
 N(M6 , Z2 ) ZTJ [c3 ] e M6
, (3.52)

(K)
c̃3 ∈H 3 (M6 , Z2 )
(K) (I) (K) (J)

where c̃3 = c3 + c3 + C3 . The second factor is the action on the partition function of τ σ, therefore:
(I) (I)
Z(στ )2 TJ [C3 ] = Z(Y,M6 ) Z(τ σ)TJ [C3 ] (3.53)
with Z(Y,M6 ) defined in equation (3.51).
3.3.2 Definition of Triality Interfaces and Defects
Half-space twisted gauging interface τ σ(M5 ). A codimension-1 interface τ σ(M5 ) between the 6d ab-

solute theories with polarization pairs LJ , LI and LK , LJ can be defined by performing twisted gauging,
19
equation (3.39), in half of spacetime and imposing the Dirichlet boundary condition for the dynamical gauge
(I)
field c3 on the interface M5 . Schematically:
τ σ(M5 )
(I) (I)

(I) (I) (J)
(3.54)
LTJ [C3 ] LTJ [c3 ] + π q̄(c3 ) + c3 ∪ C3
(I)
c3 |M5 =0
The above picture translates to the following equation:

Z !

(I) (I) (I) (I) (J)
X
ZTJ [C3 ; τ σ(M5 )] = N(M ⩾0 , Z2 ) ZTJ [c3 ] exp iπ q̄(c3 ) + c3 ∪ C3 (3.55)
6
(I) M6⩾0
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
with the normalization factor N(M ⩾0 , Z2 ) defined in equation (3.12).

6
Triality defect T 5 (M5 ). By combining a GS3 automorphism and a triality interface τ σ(M5 ) we can

construct a non-invertible triality defect T5 (M5 ) in a given absolute theory with polarization pair LJ , LI
T5 (M5 ) = T5 (M5 ) ⊗ τ σ(M5 ) . (3.56)
First the GS3 automorphism performs an anti-cyclic permutation of I, J, K and then the τ σ(M5 ) interface
(K)
implements twisted gauging of the new global Z2 2-form symmetry:
T5 (M5 ) τ σ(M5 )
LJ , LI −−−−→ LI , LK −−−−−→ LJ , LI
(I) T5 (M5 ) (K) τ σ(M5 ) (I)

ZTJ [C3 ] 7−−−−→ ZTI [C3 ] 7−−−−−→ χ−1/2 ZTJ [C3 ] (3.57)

T5 (M5 ) is thus a defect in a given absolute theory with polarization pair LJ , LI and, as we shall see
below, obeys non-invertible fusion rules. Recalling the fact that GS3 anti-cyclically permutes I, J, K and the
definition of the τ σ(M5 ) triality interface (3.54), the triality defect T5 (M5 ) can be depicted as follows:
T5 (M5 )
(I) (K)

(K) (K) (I)
(3.58)
LTJ [C3 ] LTI [c3 ] + π q̄(c3 ) + c3 ∪ C3
(K)
c3 |M5 =0
with corresponding equation:

Z !

(I) (K) (K) (K) (I)
X
ZTJ [C3 ; T5 (M5 )] = N(M ⩾0 , Z2 ) ZTI [c3 ] exp iπ q̄(c3 ) + c3 ∪ C3
6
(K) M6⩾0
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
(3.59)
with the normalization factor N(M ⩾0 , Z2 ) defined in equation (3.12).

6
20
Gauging and stacking interface στ (M5 ). A codimension-1 interface στ (M5 ) between the 6d absolute

theories with polarization pairs LJ , LI and LI , LK can be defined by performing the στ operation (3.41)
(I)
in half of spacetime and imposing the Dirichlet boundary condition for the dynamical gauge field c3 on the
interface M5 . We depict it as follows:
στ (M5 )
(I) (I)

(I) (K) (K)
(3.60)
LTJ [C3 ] LTJ [c3 ] + π c3 ∪ C3 + q̄(C3 )
(I)
c3 |M5 =0

Z !

(I) (I) (I) (K) (K)
X
ZTJ [C3 ; στ (M5 )] = N(M6 , Z2 ) ZTJ [c3 ] exp iπ c3 ∪ C3 + q̄(C3 ) . (3.61)
(I) M6⩾0
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
Triality defect T 5 (M5 ). By combining a GS3 automorphism and a triality interface στ (M5 ) we can
construct another non-invertible triality defect T 5 (M5 ) in a given absolute theory with polarization pair

LJ , LI
T 5 (M5 ) = T 5 (M5 ) ⊗ στ (M5 ) . (3.62)
First the GS3 automorphism performs a cyclic permutation of I, J, K and then the στ (M5 ) interface imple-
(J) (I)
ments gauging of Z2 and stacking with an SPT for the new Z2 global 2-form symmetry:
T 5 (M5 ) στ (M5 )
LJ , LI −−−−−→ LK , LJ −−−−−→ LJ , LI ,
(I) T 5 (M5 ) (J) στ (M5 ) (I)

ZTJ [C3 ] 7−−−−−→ ZTK [C3 ] 7−−−−−→ χ−1/2 ZTJ [C3 ] . (3.63)

T 5 (M5 ) is thus a defect in a given absolute theory with polarization pair LJ , LI and, as we shall see below,
obeys non-invertible fusion rules. Recalling the fact that GS3 cyclically permutes I, J, K and the definition
of the στ (M5 ) duality interface (3.60) the action of the triality defect T 5 (M5 ) can be depicted as:
T 5 (M5 )
(I) (J)

(J) (I) (I)
(3.64)
LTJ [C3 ] LTK [c3 ] + π c3 ∪ C3 + q̄(C3 )
(J)
c3 |M5 =0

Z !

(I) (J) (J) (I) (I)
X
ZTJ [C3 ; T 5 (M5 )] = N(M ⩾0 , Z2 ) ZTK [c3 ] exp iπ c3 ∪ C3 + q̄(C3 ) ,
6
(J) M6⩾0
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
(3.65)
where the normalization factor N(M ⩾0 , Z2 ) defined in equation (3.12).
6
The fusion of the triality defects is derived in appendix A, and summarized in the next subsection,
alongside the duality and condensation defect fusions.
21
3.4 Summary of the S3 -ality Defects and Fusions
Now that we defined the duality and triality defects, the remaining structure, in terms of the objects of this
fusion 5-category is to compute the fusion – at least on the level of objects. These are provided in detail in
appendix A. Here we simply summarize them.
The 5d topological defects of this fusion 5-category that form the simple objects are
• duality defect D5 (M5 ) defined in (3.15),
• triality defects T5 (M5 ) and T5 (M5 ), defined in (3.57) and (3.64).
The object in this higher category are obtained by taking fusions of these. The lower dimensional topological
(J)
defects are k-morphisms in terms of the category. In the present case we also have D3 , which generate a
(I) (J) (J) (id)
Z2 2-form symmetry and are invertible D3 ⊗ D3 = D3 . In terms of the fusion 5-category they are
2-morphisms. In particular we will need to include the condensation defects of these, which are denoted by
C5 .
Note that further topological defects – invertible and non-invertible – can be obtained completely analo-
gously by starting with a fixed absolute theory in figure 1, and following the arrows in a closed loop. If this
involves GS automorphisms and stacking, i.e. τ , only, the defect is invertible, if closing the loop involves also
gauging, then it is non-invertible. The computation of fusions is entirely analogous to the Rones we provided
(I)
(J) (I) iπ ⩾0 q̄(C3 )
M6
in detail. These are D5 τ and D5 τ στ . The former, invertible defect, appears as the e factor
in equations (3.74)-(3.75), whereas the latter, non-invertible defect is what appears on both sides of equation
(3.76): the concrete expression for its insertion in the partition function is (A.43).
Condensation Defects and TQFT Coefficients. A central type of defect in any fusion higher category
are condensation defects. For this particular theory we will have condensation defects of the 2-form symmetry
Z
(0) (I)
X
C5 (M5 ) = χ−1(M ⩾0 , Z )
N(M5 , Z2 ) exp iπ C 3 ,
6 2
M3 ∈H3 (M5 , Z2 ) M3
Z Z (3.66)
(1) (I)
X
C5 (M5 ) = χ−1 N
(M6 ⩾0 , Z2 ) (M5 , Z2 )
exp iπ Q(M5 , M3 ) + iπ C3 ,
M3 ∈H3 (M5 , Z2 ) M5 M3
where9 Z
Q(M5 , M3 ) = PD(M3 ) ∪ β(PD(M3 )) . (3.67)
M5
9
Here β : H 2 (M5 , Z2 ) → H 3 (M5 , Z2 ) is the Bockstein homomorphism associated to the short exact sequence
1 → Z2 → Z4 → Z2 → 1.
We denote the Poincaré dual of M3 ∈ H3 (M5 , Z2 ) by PD(M3 ) ∈ H 2 (M5 , Z2 ).
22
Furthermore, some fusions will have TQFT coefficients which we abbreviate by
R
X iπ ⩾0 q̄(c3 )
M6
Z(Y,M ⩾0 ) = χ(M6 ⩾0 , Z2 ) N(M ⩾0 , Z2 ) e
6 6
c3 ∈H 3 (M6⩾0 , M5 , Z2 )
Z
(I) (I) (I) (J) (I) (I)
X
(Z2 )2ℓ (M5 ) = χ−1
(M ⩾0 , Z N2
) (M5 , Z2 )
exp iπ ℓ b2 ∪ β(b2 ) + b2 ∪ (δ b2 − C3 ) .
6 2
(I) (J) M5
b2 , b2 ∈H 2 (M5 Z2 )
(3.68)
for ℓ ∈ {0, 1}.
We now summarize all the fusions of triality and duality defects. It is clear from these that the defects
are non-invertible, as e.g. the duality squares to a condensation defect (not the identity). The duality and
triality defects fused among themselves just generalize the known duality and triality fusions. They will be
relevant equally for the theories discussed in section 2.1.
Furthermore, both duality and triality defects are symmetries of the theory, so that we can also fuse them
and generate the full symmetry category, which then has generically S3 -ality defects – they arise from the
S3 = Z3 ⋊ Z2 invertible defects dressed by the stacking with SPT and gauging operation. Concretely we will
see the full structure of the S3 -ality defects from the fusion of duality with triality defects.
Duality Defect Fusions.
(0)
D5 (M5 ) ⊗ D5 (M5 ) = C5 (M5 ) . (3.69)
Triality Defect Fusions.
(0)
T 5 (M5 ) ⊗ T5 (M5 ) = C5 (M5 ) , (3.70)
(1)
T5 (M5 ) ⊗ T 5 (M5 ) = C5 (M5 ) , (3.71)
(1)
T5 (M5 ) ⊗ T5 (M5 ) = Z(Y,M ⩾0 ) C5 (M5 ) ⊗ T 5 (M5 ) , (3.72)
6
(0)
T 5 (M5 ) ⊗ T 5 (M5 ) = Z(Y,M ⩾0 ) C5 (M5 ) ⊗ T5 (M5 ) . (3.73)
6
Fusions between Duality and Triality Defects: S3 -ality Structure

R (I)
(0) iπ ⩾0 q̄(C3 )
M6
D5 (M5 ) ⊗ T 5 (M5 ) = C5 (M5 ) e , (3.74)
R (I)
(1) iπ ⩾0 q̄(C3 )
M6
T5 (M5 ) ⊗ D5 (M5 ) = C5 (M5 ) e , (3.75)
D5 (M5 ) ⊗ T5 (M5 ) = T 5 (M5 ) ⊗ D5 (M5 ) . (3.76)
The last equation is precisely the relation expected from the S3 -ality nature of these defects.
23
Condensation Defect Fusions. For completeness we compute the fusion of condensation defects as well.
Let X5 (M5 ) ∈ {D5 (M5 ), T5 (M5 ), T 5 (M5 )}, the resulting fusions are:
(ℓ) (I)
C5 (M5 ) ⊗ X5 (M5 ) = (Z2 )2ℓ (M5 ) X5 (M5 ) , (3.77)
(ℓ) (I)
X5 (M5 ) ⊗ C5 (M5 ) = (Z2 )2δ(X ,T ) (M5 ) X5 (M5 ) , (3.78)
(ℓ) (ℓ′ ) (I) (ℓ′ )
C5 (M5 ) ⊗ C5 (M5 ) = (Z2 )2ℓ (M5 ) C5 (M5 ) . (3.79)
We should note that a complete description of the full fusion 5-category is beyond the scope of this paper –
even for fusion 3-categories the full structure is currently not entirely worked out. It should be possible to
generalize the mathematical analysis on the level of [20] for fusion 3-categories to the present case.
4 SymTFT Perspective on S3 -ality Defects
An alternative way to construct duality and triality defects is using the SymTFT approach: we consider
the 7d 3-form Chern-Simons theory with Lie algebra g = so(8). The S3 GS automorphisms act as outer
automorphisms of the SymTFT. We construct the condensation defects implementing these invertible GS
transformations and then the non-invertible twist defects: combining them we obtain duality, triality, and
S3 -ality defects in the boundary 6d theory.
4.1 SymTFT for the 6d N = (2, 0) so(8) theory

4.1.1 7d Chern-Simons Theory
As reviewed in section 2, before specifying a polarization 6d SCFTs are in general relative theories, coupled
to a 7d bulk. The action for a 3-form Chern-Simons (CS) theory (in 7d) depends on a K-matrix and takes
the form [27, 28, 31, 33, 55]
Z
1 X (I) (J)
SCS(K) = KIJ c3 ∧ dc3 , I, J ∈ {1, ..., r} , (4.1)
4π M7 I,J
(I)
where c3 , for I ∈ {1, ..., r} are 3-form U (1) gauge fields on M7 and K is an integral, positive definite,
symmetric matrix. For 6d SCFTs with N = (2, 0) of ADE type, K is the Cartan matrix of the ADE Lie
algebra g.10
Topological operators. The 7d Chern-Simons theory (4.1) has topological operators of dimension 3 given
by the holonomy of the 3-form gauge fields [25, 31, 33].
Z
(I) (I)
U3 (M3 ) = exp i c3 (4.2)
M3
10
This is reminiscent of the 3d Chern-Simons theories describing Abelian topological phases characterized by ADE Lie algebras
[56, 57].
24
with fusions corresponding to the defect group of the 6d theory
D = Zr /KZr ∼
= Z(G̃) (4.3)
which is isomorphic to the center of the simply-connected group G̃ with ADE Lie algebra g whose Cartan
matrix is K. Properties of the operators
Z
(λ ) (I)
U3 I (M3 ) = exp i λI c3 λI ∈ D (4.4)
M3
can be determined from a quadratic function s, written in terms of the Cartan matrix K: the “spins” s(λ)
and braidings ⟨λ, λ′ ⟩ are [33]:
s(λ) = 12 λT K −1 λ , (4.5)
⟨λ, λ′ ⟩ = exp −2πi s(λ + λ′ ) − s(λ) − s(λ′ ) .

(4.6)
Kso(8) Chern-Simons Theory. The 7d CS theory of interest is the one with K-matrix given by the
Cartan matrix of g = so(8)  
2 −1 −1 −1
−1 2 0 0
Kso(8) = CMso(8) =
−1 0
 (4.7)
2 0
−1 0 0 2
The 3-dimensional operators (4.2) satisfy [33]:
(1) (id) (I) (J) (K)

U3 = U3 , U3 U3 = U3 , I, J, K ∈ {S, C, V } (4.8)
indeed the defect group for the 6d N = (2, 0) so(8) theory is D = Z2 × Z2 . Their “spins” follow from (4.5)
and are:
(1) (S) (C) (V ) 1
s(U3 ) = 0, s(U3 ) = s(U3 ) = s(U3 ) = . (4.9)
2
The braiding can be expressed by the commutation relation:
(I) (J) ′ (J) (I)

U3 (M3 )U3 (M3′ ) = (−1)⟨M3 ,M3 ⟩ U3 (M3′ ) U3 (M3 ), I ̸= J ∈ {S, C, V }. (4.10)
showing that in any basis the bilinear pairing on D is non-trivial only between the two different Z2 factors.
The 3-form Chern-Simons action (4.1) whose K-matrix is the Cartan matrix of so(8) is [33]:
Z
2 (1) (1) (S) (S) (C) (C) (V ) (V )
1 (1) (S) (C) (V )
SCS(Kso(8) ) = c3 ∧ dc3 + c3 ∧ dc3 + c3 ∧ dc3 + c3 ∧ dc3 − c d ∧ (c3 + c3 + c3 )
M7 4π 2π 3
(4.11)
(S) (C) (V )
which has an S3 symmetry permuting c3 , c3 , c3 . We note that, from equation (4.9), the operators
(S) (C) (V ) 1
U3 , U3 , U3 all have “spin” 2 so it is consistent to permute them.
25
The equations of motion, in R/2πZ, which follow from the action (4.11) are:

(1) (S) (C) (V )


 2 c3 − c3 − c3 − c3 = 0

 2 c(S) − c(1) = 0


3 3
(C) (1) (4.12)


 2 c3 − c3 =0

 (V ) (1)

 2 c3 − c3 =0
which imply (in R/2πZ) 

c(1)
3 =0
(4.13)
2 c(S) = 2 c(C) = 2 c(V ) = 0
3 3 3
in addition to
(I) (J) (K)
c3 = c3 + c3 , (4.14)
(S)
where (I, J, K) is any permutation of (S, C, V ). Using the above equation, one can eliminate either c3 or
(C) (V )
c3 or c3 when writing the on-shell actions: these are three equivalent choices as a consequence of the S3
permutation symmetry of (4.11).
There are three possible boundary conditions which are obtained by imposing Dirichlet boundary condi-
tions on a gauge field, i.e. fixing it to a classical background field on the boundary. They correspond to each
one of the three choices in (4.14) evaluated on a 6d boundary of the 7d bulk:
(I) (J) (K) (I)
B(I) : c3 |B(I) = (c3 + c3 )|B(I) = C3 , (4.15)
where, again, (I, J, K) is any permutation of (S, C, V ) and we use uppercase letters for non-dynamical
background fields.
Equation (4.13) implies that we can use Z2 -valued gauge fields, which are related to their U (1)-valued
counterparts by a factor of π as follows: c3,U (1) = π c3,Z2 . If we then use the boundary condition (4.15), we
can write the boundary action in terms of (Z2 -valued) gauge fields as:11
Z Z
π
(J) (I)
π
(K) (I)

S|B(I) = q̄(c3 ) + q̄(C3 ) or S|B(I) = q̄(c3 ) + q̄(C3 ) . (4.16)
2 B(I) 2 B(I)
(I)
The case in which C3 = 0 was discussed in [33].
Alternative presentation of the SymTFT. The Z2 3-form Dijkgraaf-Witten (DW) theory

Z
(I) (J)
π c3 ∪ δc3 , (4.17)
M7
(I) (J)
where c3 , c3 for I ̸= J ∈ {S, C, V } are Z2 -valued 3-forms, has 4 dimension-3 operators
Z
(id) (I) (I)
U3 (M3 ), U3 (M3 ) = exp iπ c3 ,
M3
Z Z (4.18)
(J) (J) (IJ) (I) (J)
U3 (M3 ) = exp iπ c3 , U3 (M3 ) = exp iπ (c3 + c3 )
M3 M3
11
Our conventions for the quadratic refinement of the intersection pairing are reviewed in subsection 3.3.1.
26
whose “spins” are
(id) (I) (J) (IJ) 1
s(U3 ) = s(U3 ) = s(U3 ) = 0, s(U3 )= , (4.19)
2
which differ from those in the 7d CS(Kso(8) ) theory (4.9). To obtain this latter theory starting from the Z2
(I) (J)
DW (4.17) one can add a coupling between (c3 + c3 ) and the the 4th Wu class v4 [33]:12
Z
(I,J) (I) (J) (I) (J)
S =π c3 ∪ δc3 + (c3 + c3 ) ∪ v4 . (4.20)
M7
(I) (J)
The coupling to v4 changes the “spins” of U3 and U3 to 21 :
(id) (I) (J) (IJ) 1

s(U3 ) = 0, s(U3 ) = s(U3 ) = s(U3 )= , (4.21)
2
matching those of the 7d CS(Kso(8) ), (4.9). The braidings and fusions also match those of the 7d CS(Kso(8) ),
so (4.20) is an alternative presentation of the theory, as discussed in Appendix F of [33].
4.1.2 GS-duality Web and Topological Manipulations
To construct the defects in the SymTFT picture we need to map the dualities and topological manipulations
into the language of background fields for the two-form symmetries.
In terms of the background fields the action of the GS dualities (2.9)-(2.10) is
(S) (C) (V ) (V )
GS2(V ) : c3 ↔ c3 , c3 ↔ c3 (4.22a)
(C) (V ) (S) (S)
GS2(S) : c3 ↔ c3 , c3 ↔ c3 (4.22b)
(S) (V ) (C) (V )
GS2(C) : c3 ↔ c3 , c3 ↔ c3 (4.22c)
(V ) (C) (S) (V )
GS3 : c3 → c3 → c3 → c3 (4.23a)
(V ) (S) (C) (V )
GS3 : c3 → c3 → c3 → c3 (4.23b)
of order 2 and 3. We show in figure 2 how the SymTFT actions (4.20) are related by the GS automorphisms.
Furthermore in order to construct the non-invertible symmetries we need to define topological manipu-
lations, which “undo” the action of the GS dualities. These map the SymTFT actions (4.20) as follows:
Z
(I,J) (I,J) (I) (J)
σ: S 7→ S +π δ c3 ∪ c3 = S (J,I) (4.24)
M7
Z
(I) (I) (I)
τ : S (I,J) 7→ S (I,J) + π c3 ∪ δc3 + c3 ∪ v4 = S (I,K) (4.25)
M7
(I) (J) (K)

where we used the fact that c3 + c3 = c3 for (I, J, K) any permutation of (S, C, V ). It is important to
(I)
note that, since we are considering Z2 gauge fields, 2 c3 = 0, ∀ I ∈ {S, C, V }. We show in figure 2 how the
SymTFT actions S (I,J) (eq. (4.20) for I ̸= J ∈ {S, C, V }) are related by the GS and σ, τ operations.
12
It can be written in terms of the Stiefel–Whitney classes wi of the tangent bundle on orientable manifolds as: v4 = w4 +w22 [33].
27
S (V,S) S (S,C)
S (V,C) S (C,S)
S (S,V )
S (C,V )
Figure 2: The Green-Schwarz Automorphisms of the SymTFT actions S (I,J) defined in equation (4.20),
for the 6d (2, 0) SCFTs with Lie algebra so(8): GS2 is of order 2 (black dashed arrows) and exchanges
(S) (C) (S) (V ) (C)
c3 ↔ c3 , whereas GS3 is of order 3 and cyclically permutes c3 , c3 , c3 (black arrows). The topological
manipulations are the σ operation, eq. (4.24) in cyan, and the τ operation, eq. (4.25) in blue arrows. This
reproduces, in the SymTFT formalism, the web of figure 1. The notation is such that the first and second
entries refer respectively to Neumann and to Dirichlet boundary conditions for the corresponding gauge
fields.
28
4.2 SymTFT Construction of Duality and Triality Defects
We now construct the topological defects as condensation defects in the SymTFT, which in our case is the 7d
3-form Chern-Simons theory CS(Kso(8) ), equation (4.11). Subsequently, we will construct the twist defects
and push them to the symmetry boundaries to obtain the symmetries of the absolute theories. It is also
possible to construct the codimension-one topological defect in the 7d Z2 DW theory, where the S3 symmetry
is not manifest, see appendix C for the details of this construction.
4.2.1 Invertible Topological Defects
Defects of dimension 3. There are 4 dimension-3 invertible topological defects, which we denote as13
Z
(id) (S) (S)
D3 (M3 ), D3 (M3 ) = exp iπ c3 ,
M3
Z Z (4.26)
(C) (C) (V ) (V )
D3 (M3 ) = exp iπ c3 , D3 (M3 ) = exp iπ c3 ,
M3 M3
which satisfy the fusions
(I) (I) (id)
D3 (M3 ) ⊗ D3 (M3 ) = D3 (M3 ), I ∈ {S, C, V } , (4.27)
(S) (C) (V ) (id)
D3 (M3 ) ⊗ D3 (M3 ) ⊗ D3 (M3 ) = D3 (M3 ), (4.28)
and exchange relations

(I) (J) ′ (J) (I)
D3 (M3 ) D3 (M3′ ) = (−1)⟨M3 ,M3 ⟩ D3 (M3′ ) D3 (M3 ), I ̸= J ∈ {S, C, V } , (4.29)
equivalent to (4.10). Furthermore,

(I) ′ (I) (I)
D3 (M3 + M3′ ) = (−1)⟨M3 ,M3 ⟩ D3 (M3 ) D3 (M3′ ), I ∈ {S, C, V } . (4.30)
Codimension-1 GS2 defects. Condensation defects are constructed by higher-gauging a subgroup of the
(S) (C)
higher-form symmetry on a submanifold [59]. By higher-gauging a Z2 subgroup of the Z2 × Z2 3-form
symmetry of the SymTFT on a codimension-1 manifold M6 , on which it is a 2-form symmetry, we obtain:
(I) (I)
X
D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ), I ∈ {S, C, V } , (4.31)
M3 ∈H3 (M6 , Z2 )
where the normalization factor N(M6 , Z2 ) was defined in (3.5).

(I) (I)
When crossing D6 (M6 ), the 3-dimensional defect D3 (M3′ ) is left unchanged:
(I) (I) (I) (I)
X
D3 (M3′ ) D6 (M6 ) = N(M6 , Z2 ) D3 (M3′ ) D3 (M3 ) =
M3 ∈H3 (M6 , Z2 )
(I) (I)
X
= N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ) =
M3 ∈H3 (M6 , Z2 )
(I) (I)
= D6 (M6 ) D3 (M3′ ) . (4.32)
13
These are reminiscent of the line defects in the 3d Spin(8) theory [3, 56–58].
29
(J)
D6 (M6 )
(J) (J)
D3 (M3′ ) D3 (M3′ )
(I) (K)
D3 (M3′ ) D3 (M3′ )
(K) (I)
D3 (M3′ ) D3 (M3′ )
(J)
Figure 3: Action of the invertible, codimension-1, duality defects D6 (M6 ), on the dimension-3 defects
(I)
D3 (M3′ ), as computed in eqs. (4.32) and (4.33). (I, J, K) is any permutation of (S, C, V ).
(J) (I)
The action of D6 (M6 ) on the dimension-3 defects D3 (M3′ ) for I ̸= J can be computed using (4.28)-
(4.30).
(I) (J) (I) (J)

X
D3 (M3′ ) D6 (M6 ) = N(M6 , Z2 ) D3 (M3′ ) D3 (M3 ) =
M3 ∈H3 (M6 , Z2 )
′ (J) (I)
X
= N(M6 , Z2 ) (−1)⟨M3 ,M3 ⟩ D3 (M3 ) D3 (M3′ ) =
M3 ∈H3 (M6 , Z2 )
′ (J) (J) (K)
X
= N(M6 , Z2 ) (−1)⟨M3 ,M3 ⟩ D3 (M3 ) D3 (M3′ ) D3 (M3′ ) =
M3 ∈H3 (M6 , Z2 )
(J) (K)
X
= N(M6 , Z2 ) D3 (M3 + M3′ ) D3 (M3′ ) =
M3 ∈H3 (M6 , Z2 )
(J) (K)
X
= N(M6 , Z2 ) D3 (M̃3 ) D3 (M3′ ) =
M̃3 ∈H3 (M6 , Z2 )
(J) (K)
= D6 (M6 ) D3 (M3′ ) (4.33)
where we used
(K) (J) (I)
D3 (M3′ ) = D3 (M3′ ) ⊗ D3 (M3′ ) . (4.34)
and replaced the sum over M3 with one over M̃3 ≡ M3 + M3′ .
Equations (4.32) and (4.33) are depicted in figure 3.
(I)
Fusion of two GS2 defects. We have thus determined that the D3 (M3 ) defects, where I ∈ {S, C, V }
implement the order-2 Green-Schwarz automorphisms GS2(I) , equations (4.22a)-(4.22c). Correspondingly,
30
(I)
let us show that D6 (M6 ) is an invertible defect of order 2.
(I) (I) 2 X (I) (I)

D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 ′ (I)
X
= N(M6 , Z2 ) (−1)⟨M3 ,M3 ⟩ D3 (M3 + M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 ′ (I)
X
= N(M6 , Z2 ) (−1)⟨M3 ,M̃3 ⟩ D3 (M̃3′ ) =
M3 ,M̃3′ ∈H3 (M6 , Z2 )
Z
2 X (I)
= N(M6 , Z2 ) exp iπ PD(M3 ) + c3 ∪ PD(M̃3′ ) (4.35)
M6
PD(M3 ),PD(M̃3′ )
∈H 3 (M6 , Z2 )
where we used (4.30) and then wrote the sum over M3 , M3′ as a sum over their Poincaré duals PD(M3 ), PD(M3′ ).
(I)
Integrating out PD(M̃3′ ) sets PD(M3 ) = −c3 (trivializing the integrand) and produces a factor of |H 3 (M6 , Z2 )|.
Recalling the definition of the normalization factor (3.5) and footnote 4, we have:
(I) (I)
D6 (M6 ) ⊗ D6 (M6 ) = χ−1
(M6 , Z2 ) . (4.36)
(I)
showing that D6 (M6 ) is an invertible defect of order 2 (the Euler counterterm could be removed by changing
(I)
the normalization of D6 (M6 ) in (4.31)).
Codimension-1 GS3 defects. When fusing
(I) (J)
D6 (M6 ) ⊗ D6 (M6 ), I ̸= J, (4.37)
we obtain new defects implementing the GS3 automorphisms, which we describe below. The defects obtained
(S) (C)
by higher-gauging the full Z2 × Z2 3-form symmetry of the SymTFT on a codimension-1 manifold M6
are:
(S) (C) 2 X (S) (C)

T6 (M6 ) = D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ), (4.38)
M3 ,M3′ ∈H3 (M6 , Z2 )
(C) (S) 2 X (C) (S)
T 6 (M6 ) = D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ) , (4.39)
M3 ,M3′ ∈H3 (M6 , Z2 )
where the normalization factor is the square of (3.5).
31
It is important to note that one can equivalently write:
(C) (V ) 2 X (C) (V )
T6 (M6 ) = D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
(V ) (S) 2 X (V ) (S)
= D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ), (4.40)
M3 ,M3′ ∈H3 (M6 , Z2 )
(S) (V ) 2 X (S) (V )
T 6 (M6 ) = D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
(V ) (C) 2 X (V ) (C)
= D6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (M3′ ) . (4.41)
M3 ,M3′ ∈H3 (M6 , Z2 )
This follows from
(S) (C) (C) (V ) (C) (C) (V )

D3 (M3 ) D3 (M3′ ) = D3 (M3 ) D3 (M3 ) D3 (M3′ ) = D3 (M3 + M3′ ) D3 (M3 ) , (4.42)
(C) (V ) (V ) (S) (V ) (V ) (S)
and similarly
(C) (S) (S) (V ) (S) (S) (V )

(S) (V ) (V ) (C) (V ) (V ) (C)
where we then redefine M3 + M3′ → M3′ and re-label the summed-over M3 ↔ M3′ .
(I)
Action on dimension-3 defects. The action of T6 (M6 ) on the dimension-3 defects D3 (M3 ) for I ∈
{S, C, V }, computed in appendix B.1,
(S) (V )
D3 (N3 ) T6 (M6 ) = T6 (M6 ) D3 (N3 ) ,
(C) (S)
D3 (N3 ) T6 (M6 ) = T6 (M6 ) D3 (N3 ) , (4.46)
(V ) (C)
D3 (N3 ) T6 (M6 ) = T6 (M6 ) D3 (N3 ) ,
shows that T6 (M6 ) implements the order-3 Green-Schwarz automorphism (4.23a), as shown on the left of
figure 4.
(I)
The action of T 6 (M6 ) on the dimension-3 defects D3 (M3 ) for I ∈ {S, C, V } can be obtained by similar
computations, or, equivalently, by simply by exchanging S ↔ C in the above equations, giving:
(C) (V )
D3 (N3 ) T 6 (M6 ) = T 6 (M6 ) D3 (N3 ) ,
(S) (C)
D3 (N3 ) T 6 (M6 ) = T 6 (M6 ) D3 (N3 ) , (4.47)
(V ) (S)
D3 (N3 ) T 6 (M6 ) = T 6 (M6 ) D3 (N3 ) .
T 6 (M6 ) thus implements the order-3 Green-Schwarz automorphism (4.23b), as shown on the right of figure
4.
32
T6 (M6 ) T 6 (M6 )
(S) (V ) (S) (C)

D3 (M3′ ) D3 (M3′ ) D3 (M3′ ) D3 (M3′ )
(C) (S) (C) (V )

D3 (M3′ ) D3 (M3′ ) D3 (M3′ ) D3 (M3′ )
(V ) (C) (V ) (S)
D3 (M3′ ) D3 (M3′ ) D3 (M3′ ) D3 (M3′ )
Figure 4: Action of the invertible, codimension-1, triality defects T6 (M6 ) and T 6 (M6 ), on the dimension-3
(I)
defects D3 (M3′ ) for I ∈ {S, C, V }, as written in eqs. (4.46) and (4.47).
Fusion of GS2 and GS3 defects. Using similar computations to the fusion of two duality GS2 defects,
we compute the fusion between GS2 and GS3 defects, which are non-abelian. The calculations can be found
in appendix B.1. We summarize the results below:
(J) (K)
D6 (M6 ) ⊗ T6 (M6 ) = χ−1
(M6 , Z2 ) D6 (M6 ) ,
(J) (I)
D6 (M6 ) ⊗ T 6 (M6 ) = χ−1
(M6 , Z2 ) D6 (M6 ) ,
(J) (I)
(4.48)
T6 (M6 ) ⊗ D6 (M6 ) = χ−1
(M6 , Z2 ) D6 (M6 ) ,
(J) (K)
T 6 (M6 ) ⊗ D6 (M6 ) = χ−1
(M6 , Z2 ) D6 (M6 ) .
where (I, J, K) is a cyclic permutation of (S, C, V ).
Fusion of two GS3 defects. The detailed computations of the fusion of two GS3 defects can also be
found in appendix B.1, the results are:
T6 (M6 ) ⊗ T 6 (M6 ) = χ−2

(M6 , Z2 ) ,
T 6 (M6 ) ⊗ T6 (M6 ) = χ−2

(M6 , Z2 ) ,
(4.49)
T6 (M6 ) ⊗ T6 (M6 ) = χ−1
(M6 , Z2 ) T 6 (M6 ) ,
T 6 (M6 ) ⊗ T 6 (M6 ) = χ−1

(M6 , Z2 ) T6 (M6 ) .
Codimension-1 defects and S3 . The fusions discussed above reveal that the defects
(id) (S) (C) (V )

D6 , D6 , D6 , D6 , T6 , T6 (4.50)
33
form (up to Euler terms) the non-abelian group S3
S3 : ⟨a, b | a3 = b2 = 1, bab = a2 ⟩ (4.51)
specifically, we have the following mapping
(id)
D6 ↔ 1, T6 ↔ a, T 6 ↔ a2 (4.52a)
(S) (C) (V )
D6 ↔ b, D6 ↔ a2 b, D6 ↔ ab, (4.52b)
with fusion corresponding to S3 group multiplication (with Euler terms on the right hand side).
4.2.2 Duality Twist Defects
By condensing a symmetry generator on a manifold with boundary one obtains a twist defect, as described
in [8] for duality twist interfaces in 3d and 5d SymTFTs. For the 7d SymTFT case, we define the duality
(I) (I)
twist defect D6 (M6⩾0 , M5|0 ) for I ∈ {S, C, V } by condensing D3 (M3 ) on a 6-manifold M6⩾0 with boundary
M5|0 = ∂M6⩾0 and imposing the Dirichlet boundary condition for M3 on M5|0 :
(I) (I)
X
D6 (M6⩾0 , M5|0 ) = N(M ⩾0 , Z2 ) D3 (M3 ), I ∈ {S, C, V },
6 (4.53)
M3 ∈H3 (M6⩾0 , Z2 )
where the normalization factor was defined in (3.12).
Fusion of Duality Twist defects. We will compute the fusion of two duality twist defects (4.53),
located at x = 0 and x = ϵ respectively, and then take ϵ → 0.
(I) (I) (I) (I)

X
D6 (M6⩾0 , M5|0 ) ⊗ D6 (M6⩾ϵ , M5|ϵ ) = N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) D3 (M3 ) D3 (M3′ ) =
6 6
M3 ∈H3 (M6⩾0 , Z2 )
M3′ ∈H3 (M6⩾ϵ , Z2 )
(I) ′
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) D3 (M3 + M3′ )(−1)⟨M3 ,M3 ⟩ =
6 6
M3 ∈H3 (M6⩾0 , Z2 )
M3′ ∈H3 (M6⩾ϵ , Z2 )
(4.54)
We now define M̃3′ = M3 + M3′ on M6⩾ϵ , write the sum over 3-cycles M3 , M̃3′ as a sum over their Lefschetz
[0,ϵ]
duals LD(M3 ), LD(M̃3′ ) and split the integral using M6⩾0 = M6 ∪ M6⩾ϵ :
Z !
(I)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) exp iπ c3 ∪ LD(M3 ) × (4.55)
6 6 [0,ϵ]
M6
LD(M3 )∈H 3 (M6⩾0 ,M5|0 , Z2 )
LD(M̃3′ )∈H 3 (M6⩾ϵ ,M5|ϵ , Z2 )
Z !

(I)
× exp iπ c3 + LD(M3 ) ∪ LD(M̃3′ ) = (4.56)
M6⩾ϵ
34
(I) [0,ϵ]
Integrating out LD(M̃3′ ) sets LD(M3 )|M ⩾ϵ = −c3 and leaves the dynamical LD(M3 ) only on M6 , with
6
Dirichlet boundary conditions on both boundaries:
Z
(I)
X
= N(M [0,ϵ] , Z ) χ−1
(M6 ⩾ϵ , Z2 )
exp iπ c3 ∪ LD(M3 ) = (4.57)
6 2
[0,ϵ] M6
LD(M3 )∈H 3 (M6 ,M5|0 ∪M5|ϵ , Z2 )
We now use the isomorphism (3.22) to map LD(M3 ) → b2 ∈ H 2 (M5 , Z2 ):14

Z
(I)
X
−1
= χ(M ⩾ϵ , Z ) N(M5 , Z2 ) exp iπ c3 ∪ b2 . (4.58)
6 2
M5
b2 ∈H 2 (M5 , Z2 )
Poincaré duality on M5 converts the sum over b2 ∈ H 2 (M5 , Z2 ) to one over M3 = PD(b2 ) ∈ H3 (M5 , Z2 ).
We thus obtain, in the ϵ → 0 limit:
(I) (I) (I)

X
D6 (M6⩾0 , M5|0 ) ⊗ D6 (M6⩾ϵ , M5|ϵ ) = χ−1
(M ⩾0 , Z ) N(M5 , Z2 ) D3 (M3 ) (4.59)
6 2
M3 ∈H3 (M5 , Z2 )
resulting in the condensation defect (with normalization factor given in eq. (3.29))
(I)
Z
(Z2 ) (I)
X
−1
C5 (M5 ) = χ(M ⩾0 , Z ) N(M5 , Z2 ) exp iπ c3 . (4.60)
6 2
M3 ∈H3 (M5 , Z2 ) M3
Therefore, the fusion is:

(I)
(I) (I) (Z2 )
D6 (M6 , M5 ) ⊗ D6 (M6 , M5 ) = C5 (M5 ) . (4.61)
4.2.3 Triality Twist Defects

(I) (J)
We define triality twist defects by condensing D3 (M3 ) D3 (N3 ) for I ̸= J ∈ {S, C, V } on a 6-manifold M6⩾0
with boundary M5|0 = ∂M6⩾0 and imposing Dirichlet boundary conditions for M3 , N3 on M5|0 :
2
(I) (J)
X
T6 (M6⩾0 , M5|0 ) = N(M ⩾0 , Z2 ) D3 (M3 ) D3 (N3 ),
6
M3 ,N3 ∈H3 (M6⩾0 , Z2 )
2 (4.62)
(J) (I)
X
T 6 (M6⩾0 , M5|0 ) = N(M ⩾0 , Z2 ) D3 (N3 ) D3 (M3 ) ,
6
M3 ,N3 ∈H3 (M6⩾0 , Z2 )
with normalization factor the square of (3.12).

14
For the normalization recall footnote 7.
35
Fusion of Triality Twist Defects. The detailed computations of triality twist defect fusions, obtained
by similar methods to those used above, can be found in appendix B.2. The results are:
(I) (J)
(I) (J) (Z2 ×Z2 )
X
T6 (M6 , M5 ) ⊗ T 6 (M6 , M5 ) = χ−2
(M ⩾0 , Z )
2
N(M 5 , Z2 )
D3 (M3 ) D3 (N3 ) = C5 (M5 ) ,
6 2
M3 ,N3 ∈H3 (M5 , Z2 )
(J) (I)
(J) (I) (Z2 ×Z2 )
X
T 6 (M6 , M5 ) ⊗ T6 (M6 , M5 ) = χ−2
(M ⩾0 , Z )
2
N(M 5 , Z2 )
D3 (M3 ) D3 (N3 ) = C5 (M5 ) ,
6 2
M3 ,N3 ∈H3 (M5 , Z2 )
(I) (J)
(Z2 ×Z2 )
T6 (M6 , M5 ) ⊗ T6 (M6 , M5 ) = C5 (M5 ) T 6 (M6 , M5 ) ,
(J) (I)
(Z2 ×Z2 )
T 6 (M6 , M5 ) ⊗ T 6 (M6 , M5 ) = C5 (M5 ) T6 (M6 , M5 ) , (4.63)
and the expression obtained from them by cyclically permuting (I, J, K).
Fusion between Duality and Triality Twist defects. Similarly:

(J)
(J) (Z2 ) (K)
D6 (M6 , M5 ) ⊗ T6 (M6 , M5 ) = C5 (M5 ) ⊗ D6 (M6 , M5 ) ,
(J)
(J) (Z2 ) (I)
D6 (M6 , M5 ) ⊗ T 6 (M6 , M5 ) = C5 (M5 ) ⊗ D6 (M6 , M5 ) ,
(J) (K)
(J) (Z2 ×Z2 ) (I)
T6 (M6 , M5 ) ⊗ D6 (M6 , M5 ) = C5 (M5 ) ⊗ D6 (M6 , M5 ) ,
(J) (I)
(J) (Z2 ×Z2 ) (K)
T 6 (M6 , M5 ) ⊗ D6 (M6 , M5 ) = C5 (M5 ) ⊗ D6 (M6 , M5 ) . (4.64)
Analogous equations follow by taking cyclic permutations of (I, J, K).
4.2.4 Gauging the S3 Symmetry in the Bulk
We now gauge the S3 symmetry in the bulk. In order to do so, we generalize the approach in [3, 8, 14, 19]
to 6d. The process is however the same: first we take invariant combinations of defects in the pre-gauged
theory, and then decorate with possible twist defects those that are invariant.
Topological operators from the ungauged theory: We first need to consider S3 invariant combinations
of the operators of the ungauged theory
(I)
• The S3 invariant combination of D3 , that is
(S) (V ) (C)
D3 ⊕ D3 ⊕ D3 . (4.65)
• Genuine codimension 2 defects are provided by twist defects discussed in the previous section. The
condensation of subgroups of Z2 × Z2 on M6 with ∂M6 = M5 becomes transparent. The topological
operators are labeled by conjugacy classes [g] of S3 , and representations of the stabilizer group of
36
g ∈ [g]. For instance the ones labeled by the trivial representation are directly related to the twist
defects presented in the previous section.
(S) (V ) (C)
D5 ⊕ D5 ⊕ D5 , T5 ⊕ T 5 . (4.66)
where
(I) (I)
D5 = D6 (M6 , M5 )|M5 ⊂X6 =∂X7 , T5 = T6 (M6 , M5 )|M5 ⊂X6 =∂X7 , T 5 = T 6 (M6 , M5 )|M5 ⊂X6 =∂X7 ,
(4.67)
where X7 is the bulk and X6 is the symmetry boundary. In order to gauge the 0-form symmetry, S3 ,
we need to couple the bulk theory to a ∈ H 1 (M7 , S3 ). The gauge field can have non-trivial holonomies
around a twist defect supported on M5 , such that δa = g. Twist defects are not gauge invariant alone,
but the combined configuration is gauge invariant. They are particularly important because they will
generate the full S3 symmetry when Dirichlet boundary conditions are specified for a [14, 19].
Boundary conditions. In the ungauged theory the b.c. come by specifying a Lagrangian subalgebra of
the original ungauged theory, L, or equivalently a polarization (LJ , LI ). When gauging S3 we need to specify
S3 invariant boundary conditions. Therefore we take the S3 orbit as follows
1 X
|B sym ⟩ → |Bgsym ⟩ , (4.68)
|Stab(B sym )|
g∈S3
where |Bgsym ⟩ is given by the action of g ∈ S3 on the polarization choice (LJ , LI ). We first need to understand
the action when the twist defects end on the boundary conditions of the ungauged theory.15 Since there is
a one-to-one correspondence between polarization of the relative theory and gapped boundary conditions of
the ungauged bulk theory, we can label the B sym as follows,
|B sym ⟩ = |LJ , LI ⟩ . (4.69)
The S3 orbit will span the full set of boundary conditions, therefore we have that
1 X
sym
|Bgsym ⟩ = |LV , LC ⟩ + |LC , LS ⟩ + |LS , LV ⟩ + |LV , LS ⟩ + |LC , LV ⟩ + |LS , LC ⟩ . (4.70)
|Stab(B )|
g∈S3
(I)
We need to evaluate now the defects D5 , T5 defined in (4.67) on the explicit symmetry boundaries, |LJ , LI ⟩ .
(I) (I)
Since they are defined as condensations of the D3 (M3 ) surfaces, the boundary condition on c3 will dictate
(I)
whether they trivialize or not on |LJ , LI ⟩ . In particular we have that, when c3 has Dirichlet boundary
(I)
conditions it is in relative cohomology: c3 ∈ H 3 (X6 , X7 , Z2 ). When M5 = ∂M6 ∈ X6 = ∂X7 the 3-cycle
(I)
M3 , where c3 with Dirichlet b.c. is integrated over, is such that
M3 ∈ H3 (M5 , M6 , Z2 ) = H3 (M5 , Z2 ) ∩ H3 (X6 , X7 , Z2 ). (4.71)

15
One can also study interfaces between two dual theories by looking at S3 symmetry defects that end and form an L-shape
on |Bsym ⟩. This will be analogous to the 4d case studied in [14].
37
Since all M3 ∈ M5 must be in relative homology, it means that this intersection is trivial. Equivalently
(I)
as stated in [8], the M3 cycle is relative to itself. Therefore the Dirichlet boundary conditions on c3 will
(I) (I)
trivialize any D3 on M5 = ∂M6 ∈ X6 = ∂X7 . Neumann boundary conditions for c3 will instead project
(I)
the D3 entirely on M5 = ∂M6 ∈ X6 = ∂X7 .
(I)
With this in hand let us evaluate D5 and T5 on |LJ , LI ⟩ ,
(K) (K)
D6 (M6 , M5 = ∂M6 ∈ X6 = ∂X7 )|(LJ ,LI ) = D5 |(LJ ,LI ) , ∀K ̸= J ̸= I ̸= K
T6 (M6 , M5 = ∂M6 ∈ X6 = ∂X7 )|(LJ ,LI ) = T5 |(LJ ,LI ) , ∀I ̸= J (4.72)
T 6 (M6 , M5 = ∂M6 ∈ X6 = ∂X7 )|(LJ ,LI ) = T 5 |(LJ ,LI ) , ∀I ̸= J
where the evaluation of T5 and T 5 is non-trivial for any (LJ , LI ) due to the identities (4.42) and (4.44).
Matching with Half-space Gauging Results. The fusion rules will follow from (4.59), (4.63) and (4.64).
However, there are subtleties which we now discuss for the polarization (LJ , LI ).
When performing different topological manipulations in the absolute 6d so(8), N = (2, 0) SCFTs involv-
ing gauging/stacking, as we did in section 3, one can obtain the same global variant up to stacking with the
Arf–Kervaire invariant of the manifold, which is a v4 -fermionic SPT phase, as explained in section 6 of [33],
and previously in [50]. Indeed, in our fusion results computed from half-space gauging and stacking, summa-
rized in section 3.4, the factor Z(Y,M ⩾0 ) , defined (up to normalization) in (3.68) is present on the right-hand
6
side of the T5 (M5 ) ⊗ T5 (M5 ) and T 5 (M5 ) ⊗ T 5 (M5 ) fusions. If we think in terms of the spins of the defects,
condensing fermionic surface defects will give rise to fermionic codimension-one defects. Stacking them with
a fermionic SPT provides codimension-one bosonic defects generating the S3 symmetry, which is a bosonic
symmetry as expected. For this reason we claim that the stacking is necessary. For instance, the stacking
comes naturally if one computes codimension-one defect correlators with (4.20), where q̄(C3 ) = C3 ∪ X, with
dX = v4 .16
The SymTFT computations we have performed in the current section, instead, do not give rise to this
factor and thus relate different global variants which do not differ by stacking with Z(Y,M ⩾0 ) , (indeed, the
6
S3 bulk symmetry is not anomalous and can be gauged). To match with the half-space gauging results, we
therefore define the symmetry defects on the symmetry boundary as given by the bulk twist defects stacked
with Z(Y,M ⩾0 ) : on the left-hand side of the fusion results it cancels since it is Z2 -valued, but remains on the
6
right-hand side when a non-trivial symmetry defect is present.
For each |LJ , LI ⟩ we will consider
sym (K)
D5B (M5 ) = Z(Y,M ⩾0 ) D5 (M5 )|(LJ ,LI ) ∀ I ̸= J . (4.73)
6
16
See [50] for a more detailed discussion on the activation of the space-time background parametrized by the Wu class
(3)
v4 ∈ H 4 (BO(d), Z2 ), and the 3-group given by 1 → Z2 → G(3) → O(d) → 1.
38
In addition the triality twist defect acts in the same way on each |LJ , LI ⟩, therefore we automatically have
sym Bsym
T5B = Z(Y,M ⩾0 ) T5 |(LJ ,LI ) , T5 = Z(Y,M ⩾0 ) T 5 |(LJ ,LI ) ∀ I ̸= J . (4.74)
6 6
(K)
The SymTFT duality twist defect fusion (4.61) for D5 , thus gives, on the symmetry boundary, the
(K)
(Z2 ) (0)
same result as (3.69) if we identify C5 |(LJ ,LI ) with C5 . For triality fusions, the SymTFT triality twist
fusions (4.63), give the boundary results (3.70)-(3.73) if we take into account the aforementioned Z(Y,M ⩾0 )
6
(I) (J) (J) (K) (J) (I)
(Z ×Z2 ) (Z ×Z2 (Z ×Z2 )
) (0) (1)
factor and identify C5 2 |(LJ ,LI )
= |(LJ ,LI ) with
C5 2 and C5 2 with C5 . Note that for
C5
(J)
these identifications, we choose the representative of the condensation defect with c3 on the left, since this
(K)
is the dynamical field on the left SymTFT boundary Bsym : the condensation of c3 in the first case gives
(I) (I)
the non-trivial twist whereas in the second case the c3 factor is frozen since c3 has Dirichlet boundary
conditions and thus the twist is trivial. Summarizing this we have
(K) (J) (I)
(0) (Z2 ) (Z2 ×Z2 )
C5 (M5 )|Bsym = C5 (M5 )|(LJ ,LI ) = C5 (M5 )|(LJ ,LI ) ∀ J ̸= I ,
(I) (J) (J) (K)
(4.75)
(1) (Z ×Z2 ) (Z ×Z2 )
C5 (M5 )|Bsym = C5 2 (M5 )|(LJ ,LI ) = C5 2 (M5 )|(LJ ,LI ) ∀ J ̸= I, K ̸= J ̸= I ̸= K .
Finally, the fusions between duality and triality defects follow from the SymTFT results (4.64) and match
the boundary fusions (3.74)-(3.76). In summary the fusions are as follows:
sym sym (0)
D5B (M5 ) ⊗ D5B (M5 ) = C5 (M5 )|Bsym
Bsym sym (0)
T5 (M5 ) ⊗ T5B (M5 ) = C5 (M5 )|Bsym
sym Bsym (1)
T5B (M5 ) ⊗ T 5 (M5 ) = C5 (M5 )|Bsym
sym sym (1)
T5B (M5 ) ⊗ T5B (M5 ) = Z(Y,M ⩾0 ) C5 (M5 )|Bsym ⊗ T 5 (M5 )
6
Bsym Bsym (0) sym (4.76)
T5 (M5 ) ⊗ T 5 (M5 ) = Z(Y,M ⩾0 ) C5 (M5 )|Bsym ⊗ T5B (M5 )
6
R (I)
sym Bsym (0) iπ ⩾0 q̄(C3 )
D5B (M5 ) ⊗ T 5 (M5 ) = C5 (M5 )|Bsym e M6
R (I)
sym sym (1) iπ ⩾0 q̄(C3 )
T5B (M5 ) ⊗ D5B (M5 ) = C5 (M5 )|Bsym e M6
Bsym Bsym Bsym sym
D5 (M5 ) ⊗ T5 (M5 ) = T 5 (M5 ) ⊗ D5B (M5 ) .
Comments on the S3 Anomaly. In order to compute the ’t Hooft anomaly for the S3 -ality symmetry we
follow the strategy of [60, 61]. These are conditions on the SymTFT that come from gauging the symmetry
G of the bulk theory (4.11). This is precisely our construction where G = S3 .
1. The first sufficient condition for a boundary theory to be anomalous is the absence of G-invariant
Lagrangian subalgebra. This is equivalent to the notion of non-intrinsic non-invertible symmetry, i.e.
there exist a Lagrangian subalgebra which leads to an invertible symmetry (with mixed anomaly)
[18, 46, 62].
39
2. The second sufficient condition for a boundary theory to be anomalous is having a non-trivial anomaly
functional for S3 symmetry. This is usually related to the higher-dimensional generalization of the
Frobenius-Schur indicator [60, 61].
In our cases it is easy to see that the first obstruction is already realized. There is no polarization or choice
of Lagrangian subalgebra such that the S3 -ality symmetry is invertible, and therefore it is always intrisically
non-invertible. This provides an obstruction to gauging it, as well as subgroup thereof. The observation
that the S3 -ality symmetry symmetry is anomalous, provides a top-down field theory explanation why 6d
N = (2, 0) theory of non-simply laced type (Bn , Cn , G2 ) cannot exist and matches the fact that it is not
possible to realize them from string theory constructions.
Acknowledgments
We thank Andrea Antinucci, Ho-Tat Lam, Craig Lawrie, Daniele Migliorati, Shu-Heng Shao, Xingyang Yu,
Yunquin Zheng for many insightful discussions. The work of FA is partially supported by the University of
Padua under the 2023 STARS Grants@Unipd programme (GENSYMSTR – Generalized Symmetries from
Strings and Branes) and in part by the Italian MUR Departments of Excellence grant 2023-2027 “Quantum
Frontiers”. The work of SSN and AW is supported by the UKRI Frontier Research Grant, underwriting the
ERC Advanced Grant “Generalized Symmetries in Quantum Field Theory and Quantum Gravity”.
A Computation of Fusion of Triality and Condensation Defects
In this appendix we provide the details for the fusions involving triality and condensation defects from
half-space gauging, summarized in subsection 3.4.
A.1 Fusion of two triality defects

R
Triality fusion computations often involve the term [0,ϵ]
M6
q̄(C3 ): since C3 is a classical background, it is
[0,ϵ] [0,ϵ]
smooth across the boundaries of M6 . In the ϵ → 0 limit, the volume of the 6-manifold M6 → 0 and,
since there are no divergences or discontinuities, so does the integral of a quadratic function of a classical
[0,ϵ]
background evaluated on M6 : Z
q̄(C3 ) → 0 as ϵ → 0. (A.1)
[0,ϵ]
M6
40
T 5 (M5 ) ⊗ T 5 (M5 ). To compute the fusion between two different triality defects T 5 (M5 ) and T5 (M5 ), we
write schematically:
T 5 (M5|0 )

(I) (J) (J) (I) (I)
LTJ [C3 ] LTK [c3 ] + π c3 ∪ C3 + q̄(C3 ) (A.2)
(J)
c3 |M5|0 =0
T5 (M5|ϵ )

(I) (I) (K) (K)
(K) (I)
LTJ [c3 ] + π c3 ∪ c3 + ∪

2 q̄(c
3 ) + c3 C 3 .
(K)
c3 |M5|ϵ =0
(A.3)
From its definition in equation (3.56), the defect T5 (M5|ϵ ) first acts with a Green-Schwarz interface T5 at
M5|ϵ , whose action on the gauge fields crossing it is:
(J) (I)
c3 |M5|ϵ T5 (M5|ϵ ) = T5 (M5|ϵ ) c3 |M5|ϵ , (A.4a)
(I) (K)
c3 |M5|ϵ T5 (M5|ϵ ) = T5 (M5|ϵ ) c3 |M5|ϵ . (A.4b)
(K)
We impose the Dirichlet boundary condition on M5|ϵ for c3 when performing half-space twisted gauging:
this, along with the invertibility of T5 , implies that we must also impose the Dirichlet boundary condition
(I) (J) R R (K)
on M5|ϵ also for c3 and c3 . In (A.2), we used the fact that q̄ is Z2 -valued (3.33) so 2 q̄(c3 ) = 0.
The above picture translates to the equation:
Z !

(I) (J) (I) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), T5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 + q̄(C3 ) ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (K) (K) (I)
X
2
×N(M ⩾ϵ ZTJ [c3 ] exp iπ c3 ∪ c3 + c3 ∪ C3 .
6 , Z2 ) M6⩾ϵ
(I) (K)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.5)
(K) (I) (I) (I)
Integrating out c3 enforces c3 = C3 , leaving the partition function ZTJ [C3 ] on M6⩾ϵ :
Z !

(I) (J) (I) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), T5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 + q̄(C3 ) ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
(I)
× χ−1
(M ⩾ϵ , Z ) ZTJ [C3 ] . (A.6)
6 2
As discussed before equation (A.1), in the ϵ → 0 limit the integral of a quadratic function of a classical
(J)
background field vanishes. We then use the isomorphism (3.22) and Poincaré duality on M5 to map c3 7→
M3 ∈ H3 (M5 , Z2 ) and obtain, in the ϵ → 0 limit (for the normalization, we refer to footnote 7):
Z
(I)
X
−1
T 5 (M5 ) ⊗ T5 (M5 ) = χ(M ⩾0 , Z ) N(M5 , Z2 ) exp iπ C3 , (A.7)
6 2
M3 ∈H3 (M5 , Z2 ) M3
41
recalling the definition of the condensation defect (3.27), we therefore have:
(0)
T 5 (M5 ) ⊗ T5 (M5 ) = C5 (M5 ) . (A.8)
T 5 (M5 ) ⊗ T 5 (M5 ). We depict the fusion in the opposite order to the one computed above as:
T5 (M5|0 )

(I) (K) (K) (K) (I)
LTJ [C3 ] LTI [c3 ] + π q̄(c3 ) + c3 ∪ C3 (A.9)
(K)
c3 |M5 =0
T 5 (M5|ϵ )

(I) (I) (I) (J) (J) (I) (I)
LTJ [c3 ] + π q̄(c3 ) + c3 ∪ c3 + c3 ∪ C3 + q̄(C3 ) .
(J)
c3 |M5|ϵ =0
(A.10)
From its definition in equation (3.62), the defect T 5 (M5|ϵ ) first acts with a Green-Schwarz interface T 5 at
M5|ϵ , whose action on the gauge fields crossing it is:
(K) (I)
c3 |M5|ϵ T 5 (M5|ϵ ) = T 5 (M5|ϵ ) c3 |M5|ϵ , (A.11a)
(I) (J)
c3 |M5|ϵ T 5 (M5|ϵ ) = T 5 (M5|ϵ ) c3 |M5|ϵ . (A.11b)
(J)
Similarly to above, the Dirichlet boundary condition on M5|ϵ for c3 and the invertibility of T 5 imply the
(I) (K)
same boundary condition on M5|ϵ also for c3 , c3 . The equation corresponding to the above figure is:
Z !

(I) (K) (K) (I)
X
ZTJ [C3 ; T5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ q̄(c3 ) + c3 ∪ C3 ×
2)
6
(K) [0,ϵ] M6⩾0
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (I) (J) (J) (I) (I)
X
2
×N(M ⩾ϵ ZTJ [c3 ] exp iπ q̄(c3 ) + c3 ∪ c3 + c3 ∪ C3 + q̄(C3 ) .
6 , Z2 ) M6⩾ϵ
(I) (J)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.12)
(J) (I) (I) (I)
Integrating out c3 enforces c3 = C3 , leaving the partition function ZTJ [C3 ] on M6⩾ϵ (note that for
(I) (I) R (I) (I) R (I) R
c3 = C3 we have (q̄(c3 ) + q̄(C3 )) = 2 q̄(C3 ) = 0 since q̄ is Z2 -valued), therefore:
Z !

(I) (K) (K) (I)
X
ZTJ [C3 ; T5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z ) exp iπ q̄(c3 ) + c3 ∪ C3 ×
6 2
(K) [0,ϵ] M6⩾0
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
(I)
× χ−1
(M ⩾ϵ , Z ZTJ [C3 ] . (A.13)
6 2)
(K)
We then use the isomorphism (3.22) to map c3 7→ b2 ∈ H 2 (M5 , Z2 ) which implies:
Z Z
(K)
q̄(c3 ) 7→ b2 ∪ β(b2 ) , (A.14)
[0,ϵ]
M6 M5
42
where β : H 2 (M5 , Z2 ) → H 3 (M5 , Z2 ) is the Bockstein homomorphism associated to the short exact sequence
1 → Z2 → Z4 → Z2 → 1. Therefore, as ϵ → 0 we have:
Z
(I)
X
−1
T5 (M5 ) ⊗ T 5 (M5 ) =χ(M6 ⩾0 , Z2 ) N(M5 , Z2 ) exp iπ b2 ∪ β(b2 ) + b2 ∪ C3 . (A.15)
M5
b2 ∈H 2 (M5 , Z2 )
Using Poincaré duality on M5 converts the sum over b2 ∈ H 2 (M5 , Z2 ) to one over M3 = PD(b2 ) ∈
H3 (M5 , Z2 ). We can thus define the discrete torsion term:
Z
Q(M5 , M3 ) = PD(M3 ) ∪ β(PD(M3 )) (A.16)
M5
and the resulting condensation defect with non-trivial discrete torsion:17

Z Z
(1) (I)
X
−1
C5 (M5 ) = χ(M ⩾0 , Z ) N(M5 , Z2 ) exp iπ Q(M5 , M3 ) + iπ C3 (A.17)
6 2
M3 ∈H3 (M5 , Z2 ) M5 M3
where N(M5 , Z2 ) is given by equation (3.29). The fusion is therefore:
(1)
T5 (M5 ) ⊗ T 5 (M5 ) = C5 (M5 ) . (A.18)
T 5 (M5 ) ⊗ T 5 (M5 ). Let us now depict the fusion between two triality defects T5 (M5 ):
T5 (M5|0 )

(I) (K) (K) (K) (I)
LTJ [C3 ] LTI [c3 ] + π q̄(c3 ) + c3 ∪ C3 (A.19)
(K)
c3 |M5 =0
T5 (M5|ϵ )

(J) (J) (J) (K) (K) (K) (I)
LTK [c3 ] + π q̄(c3 ) + c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 ) .
(K)
c3 |M5|ϵ =0
(A.20)
As before, T5 (M5|ϵ ) first acts with a Green-Schwarz interface T5 at M5|ϵ , recall equation (A.4), so we must
(J)
also impose the Dirichlet boundary condition on M5|ϵ for c3 . We therefore write the equation:
Z !

(I) (K) (K) (I)
X
ZTJ [C3 ; T5 (M5|0 ), T5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ q̄(c3 ) + c3 ∪ C3 ×
6 2) [0,ϵ]
(K) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(J) (J) (J) (K) (K) (K) (I)
X
2
×N(M ⩾ϵ ZTK [c3 ] exp iπ q̄(c3 ) + c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 .
6 , Z2 ) M6⩾ϵ
(J) (K)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.21)
17
We also include the Euler counterterm in its normalization.
43
Using the property of q̄ written in equation (3.34), we have:
Z !

(I) (K) (K) (I)
X
6 2) [0,ϵ]
(K) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(J) (J) (K) (I) (J) (I) (I)
X
2
×N(M ⩾ϵ ZTK [c3 ] exp iπ q̄(c3 + c3 + C3 ) + c3 ∪ C3 + q̄(C3 ) .
6 , Z2 ) M6⩾ϵ
(J) (K)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.22)
(K) (K) (J) (K) (I)
Replacing the sum over c3 on M6⩾ϵ with one over c̃3 = c3 + c3 + C3 gives:
Z !

(I) (K) (K) (I)
X
6 2) [0,ϵ]
(K) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(J) (K) (J) (I) (I)
X
2
×N(M ⩾ϵ ZTK [c3 ] exp iπ q̄(c̃3 ) + c3 ∪ C3 + q̄(C3 ) .
6 , Z2 ) M6⩾ϵ
(J) (K)
c3 , c̃3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.23)
[0,ϵ] (1)
In the ϵ → 0 limit, the term on M6 gives the condensation defect with non-trivial twist C5 (M5 ), as
(K)
discussed for the computation of T5 (M5 ) ⊗ T 5 (M5 ). The sum over c̃3 produces a decoupled partition
function R
X iπ ⩾0 q̄(c3 )
M6
Z(Y,M ⩾0 ) = χ(M6 ⩾0 , Z2 ) N(M ⩾0 , Z2 ) e (A.24)
6 6
c3 ∈H 3 (M6⩾0 , M5 , Z2 )
that is analogous to equation (3.51) but now defined on a manifold M6⩾0 with boundary M5 .18 Finally, we
recognize from (3.65) that the last terms are the insertion in the partition function of T (M5 ), therefore:
(1)
T5 (M5 ) ⊗ T5 (M5 ) = Z(Y,M ⩾0 ) C5 (M5 ) ⊗ T 5 (M5 ) . (A.25)
6
The fusion on the right-hand-side, involving the condensation defect, will be computed in appendix A.3.
T 5 (M5 ) ⊗ T 5 (M5 ). We depict the fusion between two T 5 (M5 ) defects as:
T 5 (M5|0 )

(I) (J) (J) (I) (I)
LTJ [C3 ] LTK [c3 ] + π c3 ∪ C3 + q̄(C3 ) (A.26)
(J)
c3 |M5|0 =0
T 5 (M5|ϵ )

(K) (K) (J) (J) (J) (I) (I)
LTI [c3 ] + π c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 + q̄(C3 ) .
(J)
c3 |M5|ϵ =0
(A.27)
18
Note that since we include a factor of χ−1 (M6 ⩾0 , Z2 )
in the normalization of the condensation defect, we have a compensating
factor of χ(M6 ⩾0 , Z2 ) in the normalization of (A.24).
44
As before, T 5 (M5|ϵ ) first acts with a Green-Schwarz interface T 5 at M5|ϵ , recall equation (A.11), so we must
(K)
also impose the Dirichlet boundary condition on M5|ϵ for c3 . The equation corresponding to the above
figure is:
Z !

(I) (J) (I) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 + q̄(C3 ) ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(K) (K) (J) (J) (J) (I) (I)
X
2
×N(M ⩾ϵ ZTI [c3 ] exp iπ c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 + q̄(C3 ) .
6 , Z2 ) M6⩾ϵ
(K) (J)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.28)
R (I)
We recall that, as ϵ → 0 we have [0,ϵ]
M6
q̄(C3 ) → 0 and use the property of q in equation (3.34), to write:
Z !
(I) (J) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(K) (K) (J) (I) (K) (K) (I)
X
2
×N(M ⩾ϵ
,Z
ZTI [c3 ] exp iπ q̄(c3 + c3 + C3 ) + q̄(c3 ) + c3 ∪ C3 .
2)
6
(K) (J) M6⩾ϵ
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.29)
(J) (J) (K) (J) (I)

We can replace the sum over c3 with one over c̃3 = c3 + c3 + C3 :
Z !
(I) (J) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(K) (J) (K) (K) (I)
X
2
×N(M ⩾ϵ ZTI [c3 ] exp iπ q̄(c̃3 ) + q̄(c3 ) + c3 ∪ C3 . (A.30)
6 , Z2 ) M6⩾ϵ
(K) (J)
c3 , c̃3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
[0,ϵ] (0)
In the ϵ → 0 limit, the term on M6 gives the condensation defect with trivial twist C5 (M5 ), as discussed
(J)
for the computation of T 5 (M5 )⊗T5 (M5 ). The sum over c̃3 produces decoupled Z(Y,M ⩾0 ) , defined in equation
6
(A.24). Finally, we recognize from (3.59) that the last terms are the insertion in the partition function of
T (M5 ), therefore:
(0)
T 5 (M5 ) ⊗ T 5 (M5 ) = Z(Y,M ⩾0 ) C5 (M5 ) ⊗ T5 (M5 ) . (A.31)
6
The fusion on the right-hand-side, involving the condensation defect, will be computed in appendix A.3.
45
A.2 Fusions between duality and triality defects
D 5 (M5 ) ⊗ T 5 (M5 ). We can schematically depict the fusion of a duality defect D5 (M5 ) with a triality
defect T 5 (M5 ) as:
D5 (M5|0 ) T 5 (M5|ϵ )

(I) (J) (J) (I) (K) (K) (J) (J) (I) (I)
LTJ [C3 ] LTI [c3 ] + π(c3 ∪ C3 ) LTJ [c3 ] + π c3 ∪ c3 + c3 ∪ C3 + q̄(C3 ) .
(J)
c(J) |M5|0 =0 c3 |M5|ϵ =0
(A.32)
The equation for the above picture is:

Z !
(I) (J) (I)
X
ZTJ [C3 ; D5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(K) (K) (J) (J) (I) (I)
X
2
×N(M ⩾ϵ
,Z
ZTJ [c3 ] exp iπ c3 ∪ c3 + c3 ∪ C3 + q̄(C3 ) . (A.33)
2)
6
(K) (J) M6⩾ϵ
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(J) (K) (I)

On M6⩾ϵ , integrating out c3 sets c3 = C3 :
Z !
(I) (J) (I)
X
ZTJ [C3 ; D5 (M5|0 ), T 5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !
(I) (I)
× χ−1 Z [C3 ]
(M6 ⩾ϵ , Z2 ) TJ
exp iπ q̄(C3 ) . (A.34)
M6⩾ϵ
As for the fusion of two duality defects, described in detail in section 3.2.3, in the ϵ → 0 limit the term on
[0,ϵ] (0)
M6 gives rise to a condensation defect C5 (M5 ). There is, however, an extra term on M6⩾0 : this is a 6d
(I)
Z2 -SPT on a manifold M6⩾ϵ with boundary M5 :
R (I)
(0) iπ ⩾0 q̄(C3 )
M6
D5 (M5 ) ⊗ T 5 (M5 ) = C5 (M5 ) e . (A.35)
T 5 (M5 ) ⊗ D 5 (M5 ). Fusing a triality defect T5 (M5 ) with a duality defect D5 (M5 ) is depicted as:
T5 (M5|0 ) D5 (M5|ϵ )

(I) (K) (K) (K) (I) (K) (K) (K) (J) (J) (I)
LTJ [C3 ] LTI [c3 ] + π q̄(c3 ) + c3 ∪ C3 LTJ [c3 ] + π q̄(c3 ) + c3 ∪ c3 + c3 ∪ C3 ) .
(K) (J)
c3 |M5 =0 c3 |M5|ϵ =0
(A.36)
46
Z !

(I) (K) (K) (I)
X
ZTJ [C3 ; T5 (M5|0 ), D5 (M5|ϵ )] = N(M ⩾0 , Z2 ) exp iπ q̄(c3 ) + c3 ∪ C3 ×
6 [0,ϵ]
(K) M6
c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
Z !

(K) (K) (K) (J) (J) (I)
X
×N(M ⩾ϵ , Z2 ) ZTJ [c3 ] exp iπ q̄(c3 ) + c3 ∪ c3 + c3 ∪ C3 .
6
(J) M6⩾ϵ
c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.37)
(J) (K) (I) (K) [0,ϵ]

Integrating out c3 sets c3 |M ⩾ϵ = C3 , reducing the dynamical c3 on M6 with Dirichlet boundary
6
conditions on both boundaries:
Z !

(I) (K) (K) (I)
X
ZTJ [C3 ; T5 (M5|0 ), D5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ q̄(c3 ) + c3 ∪ C3 ×
6 2) [0,ϵ]
(K) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !
(I) (I)
× χ−1 Z [C3 ]
(M6 ⩾ϵ , Z2 ) TJ
exp iπ q̄(C3 ) .
M6⩾ϵ
(A.38)
[0,ϵ]
As discussed for triality fusions in section A.1, the term on M6 gives rise to the condensation defect
(1)
C5 (M5 ), defined in (A.17). Like for the previous case, we also have a term on M6⩾0 , so the result is:
R (I)
(1) iπ ⩾0 q̄(C3 )
M6
T5 (M5 ) ⊗ D5 (M5 ) = C5 (M5 ) e . (A.39)
D 5 (M5 ) ⊗ T 5 (M5 ) and T 5 (M5 ) ⊗ D 5 (M5 ). For the fusion D5 (M5 ) ⊗ T5 (M5 ), we draw schematically:
D5 (M5|0 ) T5 (M5|ϵ )

(I) (J) (J) (I) (I) (I) (K) (K) (K) (I)
LTJ [C3 ] LTI [c3 ] + π(c3 ∪ C3 ) LTK [c3 ] + π c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 .
(K)
c(J) |M5|0 =0 c3 |M5|ϵ =0
(A.40)
The equation for the above picture is:

Z !
(I) (J) (I)
X
ZTJ [C3 ; D5 (M5|0 ), T5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (K) (K) (K) (I)
X
2
×N(M ⩾ϵ ZTK [c3 ] exp iπ c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 . (A.41)
6 , Z2 ) M6⩾ϵ
(K) (I)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
47
Using the properties of q̄, we can write:
Z !
(I) (J) (I)
X
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (K) (I) (I) (I) (I) (I)
X
2
× N(M ⩾ϵ
,Z
ZTK [c3 ] exp iπ q̄(c3 + c3 + C3 ) + q̄(c3 ) + c3 ∪ C3 + q̄(C3 ) .
2)
6
(K) (I) M6⩾ϵ
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.42)
(I) (K) (I) (K)

We can define c̃3 = c3 + c3 + C3 , to replace the sum over c3 :
Z !
(I) (J) (I)
X
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (I) (I) (I)
X
2
× N(M ⩾ϵ
,Z
ZTK [c3 ] exp iπ q̄(c̃3 ) + q̄(c3 ) + c3 ∪ C3 + q̄(C3 ) .
2)
6
(I) M6⩾ϵ
c̃3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.43)
For T 5 (M5 ) ⊗ D5 (M5 ), we instead draw:
T 5 (M5|0 ) D5 (M5|ϵ )

(I) (J) (J) (I) (I) (I) (I) (J) (J) (J) (I)
LTJ [C3 ] LTK [c3 ] + π c3 ∪ C3 + q̄(C3 ) LTK [c3 ] + π c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 .
(J) (J)
c3 |M5 =0 c3 |M5|ϵ =0
(A.44)

Z !

(I) (J) (I) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), D5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 + q̄(C3 ) ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (J) (J) (J) (I)
X
2
×N(M ⩾ϵ ZTK [c3 ] exp iπ c3 ∪ c3 + q̄(c3 ) + c3 ∪ C3 .
6 , Z2 ) M6⩾ϵ
(J) (I)
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.45)
(I)
q̄(C3 ) → 0 as ϵ → 0. On M6⩾ϵ we use the properties of q̄ and write:
R
Recall that M6
[0,ϵ]
Z !
(I) (J) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), D5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (J) (I) (I) (I) (I)
X
2
×N(M ⩾ϵ
,Z
ZTK [c3 ] exp iπ q̄(c3 + c3 + C3 ) + c3 ∪ C3 + q̄(C3 ) .
2)
6
(J) (I) M6⩾ϵ
c3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.46)
48
(J) (I) (J) (I)
If we replace the sum over c3 with one over c̃3 = c3 + c3 + C3 , we have:
Z !
(I) (J) (I)
X
ZTJ [C3 ; T 5 (M5|0 ), D5 (M5|ϵ )] = N(M [0,ϵ] , Z exp iπ c3 ∪ C3 ×
6 2) [0,ϵ]
(J) [0,ϵ] M6
c3 ∈H 3 (M6 , M5|0 ∪M5|ϵ , Z2 )
Z !

(I) (I) (I) (I) (I)
X
2
× N(M ⩾ϵ ZTK [c3 ] exp iπ q̄(c̃3 ) + q̄(c3 ) + c3 ∪ C3 + q̄(C3 ) .
6 , Z2 ) M6⩾ϵ
(I)
c̃3 , c3 ∈H 3 (M6⩾ϵ , M5|ϵ , Z2 )
(A.47)
The right-hand sides of equations (A.43) and (A.47) are the same, showing that:
D5 (M5 ) ⊗ T5 (M5 ) = T 5 (M5 ) ⊗ D5 (M5 ) . (A.48)
A.3 Fusions with condensation defects
Equations (3.17)-(3.30) and (A.9)-(A.18) imply that we can write the insertion of the condensation defect
(ℓ)
C5 , ℓ ∈ {0, 1} schematically as [7]
(ℓ)
C5 (M5 )

(I) (I) (I) (I) (J) (J) (I) (I)
LTJ [C3 ] (I)
c3 |M5 =0
LTJ [c3 ] + π ℓ q̄(c3 ) + c3 ∪ c3 + c3 ∪ C3 + ℓ q̄(C3 ) , (A.49)
(J)
c3 |M5 =0
(I) (J)
where both fields c3 , c3 are dynamical gauge fields with Dirichlet boundary conditions on the defect. The
figure in (A.49) corresponds to the equation:

R (I) (I) (J) (J) (I) (I)
(I) (ℓ)
X (I) iπ ⩾0 ℓ q̄(c3 )+c3 ∪c3 +c3 ∪C3 +ℓ q̄(C3 )
2 M6
ZTJ [C3 ; C5 (M5 )] =N(M ⩾0 ZTJ [c3 ] e .
6 , Z2 )
(I) (J)
c3 , c3 ∈H 3 (M6⩾0 , M5|0 , Z2 )
(A.50)
From the figures defining the defects symmetry D5 (M5 ), T5 (M5 ), T 5 (M5 ) in (3.15), (3.58), (3.64), we can
generally denote them as X5 (M5 ) and depict:
X5 (M5 )
(I) (I) (I) (A.51)
LTJ [C3 ] L[c3 ] + π(δ(X ,T ) q̄(c3 ) + c3 ∪ C3 + δ(X ,T ) q̄(C3 )) .
c3 |M5 =0
where δ(X ,Y) is 1 if X = Y and 0 otherwise.

(ℓ)
The fusion C5 (M5 ) ⊗ X5 (M5 ) gives a Dijkgraaf-Witten (DW) theory, (with twist if ℓ = 1) on M5 ,
[0,ϵ]
multiplying X5 (M5 ). Indeed using the expressions (A.49)-(A.50) on M6 and the isomorphism (3.22), we
obtain the following Z2 TQFT on M5 in the ϵ → 0 limit:
Z
(I) (I) (I) (J) (I) (I)
X
(Z2 )2ℓ (M5 ) = χ−1 N2
(M6 ⩾0 , Z2 ) (M5 , Z2 )
exp iπ ℓ b2 ∪ β(b2 ) + b2 ∪ (δ b2 − C3 ) .
(I) (J) M5
b2 , b2 ∈H 2 (M5 Z2 )
(A.52)
49
(J)
On M6⩾ϵ we have integrated out c3 appearing in the definition of the condensation defect (A.49)-(A.50):
(I) (I)
this sets c3 = C3 and leaves on M6⩾ϵ the insertion of the defect X5 :19
(ℓ) (I)
C5 (M5 ) ⊗ X5 (M5 ) = (Z2 )2ℓ (M5 ) X5 (M5 ) . (A.53)
(ℓ) (J)
For the fusion X5 (M5 ) ⊗ C5 (M5 ), we can again integrate out c3 appearing in the definition of the
[0,ϵ]
condensation defect, that leaves us with X5 on M6⩾ϵ . On M6 we use the isomorphism (3.22) and obtain
(δ )
the condensation defect on M5 C5 (X ,T ) : the twist is zero except when X = T (owing to the q̄(c3 ) term in
its definition). Therefore:
(ℓ) (δ(X ,T ) ) (I)

X5 (M5 ) ⊗ C5 (M5 ) = C5 (M5 ) ⊗ X5 (M5 ) = (Z2 )2δ(X ,T ) (M5 ) X5 (M5 ). (A.54)
Analogously, when fusing two condensation defects, we obtain:
(ℓ) (ℓ′ ) (I) (ℓ′ )

C5 (M5 ) ⊗ C5 (M5 ) = (Z2 )2ℓ (M5 ) C5 (M5 ) . (A.55)
B Details on SymTFT computations

B.1 Invertible GS defects
In this subsection, we present the derivations of results in section 4.2.
Action of T6 (M6 ) on dimension-3 defects. Below is the derivation of equation (4.46) using properties
(4.28)-(4.30):
(S) 2 X (S) (S) (C)

D3 (N3 ) T6 (M6 ) = N(M6 , Z2 ) D3 (N3 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 ′ (S) (C) (S)
X
= N(M6 , Z2 ) (−1)⟨N3 ,M3 ⟩ D3 (M3 ) D3 (M3′ ) D3 (N3 ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (C) (V )
= N(M6 , Z2 ) D3 (M3 ) D3 (M3′ + N3 ) D3 (N3 ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (C) (V )
= N(M6 , Z2 ) D3 (M3 ) D3 (M̃3′ ) D3 (N3 ) =
M3 ,M̃3′ ∈H3 (M6 , Z2 )
(V )
= T6 (M6 ) D3 (N3 ) , (B.1)
19
It also produces a factor of |H 3 (M6⩾0 , Z2 )| which combined with N(M
2
⩾0
,Z
gives the Euler term χ−1
(M ⩾0 , Z ) which we have
6 2) 6 2
included in the normalization of (A.52).
50
(C) 2 X (C) (S) (C)
D3 (N3 ) T6 (M6 ) = N(M6 , Z2 ) D3 (N3 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (V ) (S) (S) (C)
= N(M6 , Z2 ) D3 (N3 ) D3 (N3 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (S) (C) (C)
= N(M6 , Z2 ) D3 (M3 + N3 )D3 (N3 ) D3 (N3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (C) (S)
= N(M6 , Z2 ) D3 (M3 + N3 ) D3 (M3′ + N3 ) D3 (N3 ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (C) (S)
= N(M6 , Z2 ) D3 (M̃3 ) D3 (M̃3′ ) D3 (N3 ) =
M̃3 ,M̃3′ ∈H3 (M6 , Z2 )
(S)
= T6 (M6 ) D3 (N3 ) , (B.2)
(V ) 2 X (V ) (S) (C)
D3 (N3 ) T6 (M6 ) = N(M6 , Z2 ) D3 (N3 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (C) (S) (S) (C)
= N(M6 , Z2 ) D3 (N3 ) D3 (N3 ) D3 (M3 ) D3 (M3′ ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (C) (C)
= N(M6 , Z2 ) D3 (M3 + N3 ) D3 (M3′ ) D3 (N3 ) =
M3 ,M3′ ∈H3 (M6 , Z2 )
2 X (S) (C) (C)
= N(M6 , Z2 ) D3 (M̃3 ) D3 (M3′ ) D3 (N3 ) =
M̃3 ,M3′ ∈H3 (M6 , Z2 )
(C)
= T6 (M6 ) D3 (N3 ) , (B.3)
Fusion of GS2 and GS3 defects. Let (I, J, K) be a cyclic permutation of (S, C, V ), and recall definitions
(4.38)-(4.41) of T6 (M6 ) and T 6 (M6 ). We compute:
(J) 3 X (J) (J) (K)

D6 (M6 ) ⊗ T6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (N3′ ) D3 (M3′ ) =
M3 ,N3′ ,M3′ ∈H3 (M6 , Z2 )
3 ′ (J) (K)
X
= N(M6 , Z2 ) (−1)⟨M3 ,N3 ⟩ D3 (M3 + N3′ ) D3 (M3′ ) =
M3 ,N3′ ,M3′ ∈H3 (M6 , Z2 )
3 ′ (J) (K)
X
= N(M6 , Z2 ) (−1)⟨M3 ,Ñ3 ⟩ D3 (Ñ3′ ) D3 (M3′ ) =
M3 ,Ñ3′ ,M3′ ∈H3 (M6 , Z2 )
Z
2 X (J)
= N(M6 , Z2 ) exp iπ PD(M3 ) + c3 ∪ PD(Ñ3′ ) ×
M6
PD(M3 ),PD(Ñ3′ )
∈H 3 (M6 , Z2 )
(K)
X
× N(M6 , Z2 ) D3 (M3′ ) =
M3′ ∈H3 (M6 , Z2 )
(K)
= χ−1
(M6 , Z2 ) D6 (M6 ) (B.4)
51
After writing the sum over M3 , Ñ3′ as a sum over their Poincaré duals PD(M3 ), PD(Ñ3′ ), we integrated out
(K)
PD(Ñ3′ ), trivializing the first integrand20 and leaving (up to an Euler term) the defect D6 (M6 ). Similarly,
exchanging K ↔ I in the above calculation, we have:
(J) (I)
D6 (M6 ) ⊗ T 6 (M6 ) = χ−1
(M6 , Z2 ) D6 (M6 ) . (B.5)
The fusions are non-abelian as can be seen by computing
(J) 3 X (J) (K) (J)

T6 (M6 ) ⊗ D6 (M6 ) = N(M6 , Z2 ) D3 (N3 ) D3 (M3 ) D3 (M3′ ) =
N3 ,M3 ,M3′ ∈H3 (M6 , Z2 )
3 ′ (J) (J) (K)
X
= N(M6 , Z2 ) (−1)⟨M3 ,M3 ⟩ D3 (N3 )D3 (M3′ ) D3 (M3 ) =
N3 ,M3 ,M3′ ∈H3 (M6 , Z2 )
3 ′ (J) (K)
X
= N(M6 , Z2 ) (−1)⟨M3 +N3 ,M3 ⟩ D3 (N3 + M3′ ) D3 (M3 ) =
N3 ,M3 ,M3′ ∈H3 (M6 , Z2 )
3 ′ (J) (K)
X
= N(M6 , Z2 ) (−1)⟨M3 ,N3 ⟩+⟨M3 +N3 ,M̃3 ⟩ D3 (M̃3′ ) D3 (M3 ) =
N3 ,M3 ,M̃3′ ∈H3 (M6 , Z2 )
(B.6)
we can redefine M3 + N3 → N3 and use Poincaré duality

Z Z
3 X (J)
= N(M6 , Z2 ) exp iπ PD(M3 ) ∪ PD(N3 ) exp iπ PD(N3 ) + c3 ∪ PD(M̃3′ ) ×
M6 M6
PD(N3 ),PD(M3 ),PD(M̃3′ )
∈H 3 (M6 , Z2 )
Z
(K)
× exp iπ c3 ∪ PD(M3 ) (B.7)
M6
(J)
Integrating out PD(M̃3′ ) sets PD(N3 ) = −c3 , which, once replaced in the integral and using again Poincaré
duality, gives:
(J) (J) (K)

X
T6 (M6 ) ⊗ D6 (M6 ) = χ−1
(M6 , Z2 ) N(M6 , Z2 ) D3 (M3 ) D3 (M3 ) =
M3 ∈H3 (M6 , Z2 )
(I)
X
= χ−1
(M6 , Z2 ) N(M6 , Z2 ) D3 (M3 ) =
M3 ∈H3 (M6 , Z2 )
(I)
= χ−1
(M6 , Z2 ) D6 (M6 ) . (B.8)
Similarly, by exchanging I ↔ K in the above, we obtain:
(J) (K)
T 6 (M6 ) ⊗ D6 (M6 ) = χ−1
(M6 , Z2 ) D6 (M6 ) . (B.9)
We have therefore shown that equations (4.48) hold.

2
20
It also produces a factor of H 3 (M6 , Z2 ), which, combined with N(M6 , Z2 ) gives χ−1
(M6 , Z2 ) .
52
Fusion of two GS3 defects.
4 X (I) (J) (J) (I)
T6 (M6 ) ⊗ T 6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (N3 ) D3 (N3′ ) D3 (M3′ ) =
M3 ,N3 ,N3′ ,M3′
∈H3 (M6 , Z2 )
4 ′ ′ ′ (J) (I)
X
= N(M6 , Z2 ) (−1)⟨N3 ,N3 ⟩+⟨M3 ,N3 +N3 +M3 ⟩ D3 (N3 + N3′ ) D3 (M3 + M3′ ) =
M3 ,N3 ,N3′ ,M3′
∈H3 (M6 , Z2 )
4 ′ (J) ′ (I)
X
= N(M6 , Z2 ) (−1)⟨N3 +M3 ,Ñ3 ⟩ D3 (Ñ3′ ) (−1)⟨M3 ,M̃3 ⟩ D3 (M̃3′ ) = (B.10)
M3 ,N3 ,Ñ3′ ,M̃3′
∈H3 (M6 , Z2 )
we can redefine N3 + M3 → N3 and use Poincaré duality to write

Z
2 X (J) ′
= N(M6 , Z2 ) exp iπ PD(N3 ) + c3 ∪ PD(Ñ3 ) ×
M6
PD(N3 ),PD(Ñ3′ )
∈H 3 (M6 , Z2 )
Z
2 X (I)
× N(M6 , Z2 ) exp iπ PD(M3 ) + c3 ∪ PD(M̃3′ ) . (B.11)
M6
PD(M3 ),PD(M̃3′ )
∈H 3 (M6 , Z2 )
For each factor, we now proceed similarly to the fusion of two GS2 defects, i.e. we integrate out PD(M̃3′ ) and
PD(Ñ3′ ): this trivializes the integrands and produces a factor |H 3 (M6 , Z2 )|2 , which, when combined with
4
N(M6 , Z2 ) , produces an Euler counterterm χ−2(M6 , Z2 ) (this could be eliminated by redefining the normaliza-
tion of the GS3 defects) giving:
T6 (M6 ) ⊗ T 6 (M6 ) = χ−2

(M6 , Z2 ) . (B.12)
The fusion T 6 (M6 )⊗T6 (M6 ) can be obtained by simply exchanging the labels I ↔ J in the above computation
and yields the same result:
T 6 (M6 ) ⊗ T6 (M6 ) = χ−2

(M6 , Z2 ) . (B.13)
Let us now compute the fusion of T6 (M6 ) with itself:

4 X (I) (J) (I) (J)
T6 (M6 ) ⊗ T6 (M6 ) = N(M6 , Z2 ) D3 (M3 ) D3 (N3 ) D3 (M3′ ) D3 (N3′ ) =
M3 ,N3 ,M3′ ,N3′
∈H3 (M6 , Z2 )
4 ′ ′ (I) (J)
X
= N(M6 , Z2 ) (−1)⟨M3 +N3 ,M3 ⟩+⟨N3 ,N3 ⟩ D3 (M3 + M3′ ) D3 (N3 + N3′ ) =
M3 ,N3 ,M3′ ,N3′
∈H3 (M6 , Z2 )
4 ′ ′ (I) (J)
X
= N(M6 , Z2 ) (−1)⟨N3 ,M3 ⟩+⟨M3 +N3 ,M̃3 ⟩+⟨N3 ,Ñ3 ⟩ D3 (M̃3′ ) D3 (Ñ3′ ) = (B.14)
M3 ,N3 ,M̃3′ ,Ñ3′
∈H3 (M6 , Z2 )
53
redefining M3 + N3 → M3 and using Poincaré duality gives:
Z
4 X
= N(M6 , Z2 ) exp iπ PD(N3 ) ∪ PD(M3 ) ×
PD(M3 ),PD(N3 ), M6
PD(M̃3′ ),PD(Ñ3′ )∈H 3 (M6 , Z2 )
Z
(I) ′
× exp iπ PD(M3 ) + c3 ∪ PD(M̃3 ) ×
M6
Z
(J)
× exp iπ PD(N3 ) + c3 ∪ PD(Ñ3′ ) . (B.15)
M6
(I)
Integrating out PD(M̃3′ ) sets PD(M3 ) = −c3 . Replacing this into the above expression, and using again
Poincaré duality, we obtain:
2 (I) (J) ′
X
T6 (M6 ) ⊗ T6 (M6 ) = χ−1
(M6 , Z2 ) N(M6 , Z2 ) D3 (N3 ) D3 (Ñ3′ )(−1)⟨N3 ,Ñ3 ⟩ =
N3 ,Ñ3′ ∈H3 (M6 , Z2 )
2 X (J) (I)
= χ−1
(M6 , Z2 ) N(M6 , Z2 ) D3 (Ñ3′ ) D3 (N3 ) (B.16)
N3 ,Ñ3′ ∈H3 (M6 , Z2 )
showing that:
T6 (M6 ) ⊗ T6 (M6 ) = χ−1

(M6 , Z2 ) T 6 (M6 ) . (B.17)
The fusion of T 6 (M6 ) with itself can be obtained by exchanging I ↔ J in the above computation:
T 6 (M6 ) ⊗ T 6 (M6 ) = χ−1

(M6 , Z2 ) T6 (M6 ) . (B.18)
B.2 Triality twist defect fusions
We will compute the fusion of two triality twist defects (4.62), located at x = 0 and x = ϵ respectively,
and then take ϵ → 0. The results are summarized in equation (4.63). We first compute the fusions of two
different triality twist defects:
2
T6 (M6⩾0 , M5|0 ) ⊗ T 6 (M6⩾ϵ , M5|ϵ ) = N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) ×
6 6
(I) (J) (J) (I)

X
× D3 (M3 ) D3 (N3 ) D3 (N3′ ) D3 (M3′ ) =
M3 ,N3 ∈H3 (M6⩾0 , Z2 )
N3′ ,M3′ ∈H3 (M6⩾ϵ , Z2 )
2
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) ×
6 6
′ ′ ′ (J) (I)
X
× (−1)⟨N3 ,N3 ⟩+⟨M3 ,N3 +N3 +M3 ⟩ D3 (N3 + N3′ ) D3 (M3 + M3′ ) =
M3 ,N3 ∈H3 (M6⩾0 , Z2 )
N3′ ,M3′ ∈H3 (M6⩾ϵ , Z2 )
(B.19)
54
[0,ϵ] [0,ϵ]
We now set, on M6⩾ϵ , Ñ3′ = N3 + N3′ , M̃3′ = M3 + M3′ , and split the homologies using M6⩾0 = M6 ∪ M6 :
 
2
(J) (I)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 )  (−1)⟨M3 ,N3 ⟩ D3 (N3 ) D3 (M3 ) ×
 
6 6
[0,ϵ]
M3 ,N3 ∈H3 (M6 , Z2 )
′ (J) ′ (I)
X
× (−1)⟨N3 +M3 ,Ñ3 ⟩ D3 (Ñ3′ ) (−1)⟨M3 ,M̃3 ⟩ D3 (M̃3′ ) =
M3 ,N3 ,Ñ3′ ,M̃3′ ∈H3 (M6⩾ϵ , Z2 )
 
2
(I) (J)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) D3 (N3 ) D3 (M3 ) ×
 

6 6
[0,ϵ]
M3 ,N3 ∈H3 (M6 , Z2 )
′ (J) ′ (I)
X
× (−1)⟨N3 +M3 ,Ñ3 ⟩ D3 (Ñ3′ ) (−1)⟨M3 ,M̃3 ⟩ D3 (M̃3′ ) = (B.20)
M3 ,N3 ,Ñ3′ ,M̃3′ ∈H3 (M6⩾ϵ , Z2 )
We write the sum over 3-cycles as a sum over their Lefschetz duals then integrate out LD(M̃3′ ) and LD(Ñ3′ ),
[0,ϵ]
which reduce the dynamical LD(M3 ) and LD(N3 ) to M4 with Dirichlet boundary conditions on both
boundaries. Using then the isomorphism (3.22) and Poincaré duality on M5 , we obtain in the ϵ → 0 limit:
(I) (J)
X
T6 (M6 , M5 ) ⊗ T 6 (M6 , M5 ) = χ−2
(M ⩾0 , Z )
2
N(M 5 , Z2 )
D3 (M3 ) D3 (N3 ) . (B.21)
6 2
M3 ,N3 ∈H3 (M5 , Z2 )
By exchanging I ↔ J, we similarly have:
(J) (I)
X
T 6 (M6 , M5 ) ⊗ T6 (M6 , M5 ) = χ−2
(M ⩾0 , Z )
2
N(M 5 , Z2 )
D3 (M3 ) D3 (N3 ) . (B.22)
6 2
M3 ,N3 ∈H3 (M5 , Z2 )
We now consider the fusion of a triality twist defect with itself:

2
T6 (M6⩾0 , M5|0 ) ⊗ T6 (M6⩾ϵ , M5|ϵ ) = N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) ×
6 6
(I) (J) (I) (J)

X
× D3 (M3 ) D3 (N3 ) D3 (M3′ ) D3 (N3′ ) =
M3 ,N3 ∈H3 (M6⩾0 , Z2 )
M3′ ,N3′ ∈H3 (M6⩾ϵ , Z2 )
2
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) ×
6 6
′ ′ (I) (J)
X
× (−1)⟨M3 +N3 ,M3 ⟩+⟨N3 ,N3 ⟩ D3 (M3 + M3′ ) D3 (N3 + N3′ ) =
M3 ,N3 ∈H3 (M6⩾0 , Z2 )
M3′ ,N3′ ∈H3 (M6⩾ϵ , Z2 )
(B.23)
55
Like for the previous case, we now set, on M6⩾ϵ , Ñ3′ = N3 + N3′ , M̃3′ = M3 + M3′ , and split the homologies
[0,ϵ] [0,ϵ]
using M6⩾0 = M6 ∪ M6 :
 
2
(I) (J)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) D3 (N3 ) D3 (M3 ) ×
 

6 6
[0,ϵ]
M3 ,N3 ∈H3 (M6 , Z2 )
′ ′ (I) (J)
X
× (−1)⟨N3 ,M3 ⟩+⟨M3 +N3 ,M̃3 ⟩+⟨N3 ,Ñ3 ⟩ D3 (M̃3′ ) D3 (Ñ3′ ) =
M3 ,N3 ,Ñ3′ ,M̃3′ ∈H3 (M6⩾ϵ , Z2 )
 
2
(I) (J)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) D3 (N3 ) D3 (M3 ) ×
 

6 6
[0,ϵ]
M3 ,N3 ∈H3 (M6 , Z2 )
′ ′ (I) (J)
X
× (−1)⟨N3 ,M3 ⟩+⟨M3 ,M̃3 ⟩+⟨N3 ,Ñ3 ⟩ D3 (M̃3′ ) D3 (Ñ3′ ) = (B.24)
M3 ,N3 ,Ñ3′ ,M̃3′ ∈H3 (M6⩾ϵ , Z2 )
where, in the last step, we re-defined M3 + N3 → M3 on M6⩾ϵ . We write the sum over 3-cycles as a sum over
(I)
their Lefschetz duals, then integrate out LD(M̃3′ ), which sets LD(M3 )|M ⩾ϵ = −c3 , which we then replace in
6
the expression:
 
2
(I) (J)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 ) D3 (N3 ) D3 (M3 ) ×
 

6 6
[0,ϵ]
M3 ,N3 ∈H3 (M6 , Z2 )
′ (I) (J)
X
× (−1)⟨N3 ,Ñ3 ⟩ D3 (N3 ) D3 (Ñ3′ ) =
N3 ,Ñ3′ ,∈H3 (M6⩾ϵ , Z2 )
 
2
(I) (J)
X
= N(M ⩾0 , Z2 ) N(M ⩾ϵ , Z2 )  D3 (N3 ) D3 (M3 ) ×
 
6 6
[0,ϵ]
M3 ,N3 ∈H3 (M6 , Z2 )
(J) (I)
X
× D3 (Ñ3′ ) D3 (N3 ) = (B.25)
N3 ,Ñ3′ ,∈H3 (M6⩾ϵ , Z2 )
After using then the isomorphism (3.22) and Poincaré duality on M5 , and taking the ϵ → 0 limit, we
obtain:
 
(I) (J)
X
T6 (M6 , M5 ) ⊗ T6 (M6 , M5 ) = χ−2
(M ⩾0 , Z )
2
N(M 5 , Z2 )
D3 (M3 ) D3 (N3 ) T 6 (M6 , M5 ).
6 2
M3 ,N3 ∈H3 (M5 , Z2 )
(B.26)
Exchanging I ↔ J gives:
 
(J) (I)
X
T 6 (M6 , M5 ) ⊗ T 6 (M6 , M5 ) = χ−2
(M ⩾0 , Z
2
N(M5 , Z2 )
D3 (M3 ) D3 (N3 ) T6 (M6 , M5 ).
6 2)
M3 ,N3 ∈H3 (M5 , Z2 )
(B.27)
56
These results are written in a more concise manner in equation (4.63).
The fusions between duality and triality twist defects can be computed very similarly, following the
derivation for duality twist defect fusions in section 4.2.2, the fusions of GS2 and GS3 defects in appendix
B.1 and the above triality twist defect fusions. The results are provided in equation (4.64).
C Bosonic SymTFT and Interfaces
We have learnt that in the Kso (8) Chern-Simons theory the 3-dimensional surface operator are fermionic.
(I) (I)
We can start with an absolute theory A[Z2 ], where Z2 indicates the 2-form symmetry of the theory.
To describe the SymTFT of this theory the standard procedure entails the gauging in the bulk via the
Dijkgraaf-Witten (DW) discrete gauge theory for this symmetry,
Z
(I) (J)
S7d DW = π c3 ∪ δc3 (C.1)
M7
in this case the matrix K reads,

0 2
K= (C.2)
−2 0
and the spins of the operators (which are defined in d = 2k + 1 dimensions with odd k)
Z
(λI ) (I) (I) (J)
U3 (M3 ) = exp i λI c3 λI ∈ D = Z2 × Z2 (C.3)
M3
are determined by (4.5), and read

1
s((1, 0)) = 0, s((0, 1)) = 0, s((1, 1)) = . (C.4)
2
So we notice that U (I) , U (J) are bosonic whereas U (K) is fermionic. Due to different spins of the U operators,
the theory does not have manifest S3 symmetry. In addition it does not contain all possible topological
manipulation of the boundary A[ZI2 ]. For instance the topological manipulation, τ , that is stacking with an
SPT given by the quadratic refinement M6 q̄(cI3 ) changes the spins of the U operators, while gauging remains
R
a boundary manipulation captured by the DW theory. This is in contrast with the Kso (8) Chern-Simons
where the U defects are all fermionic and the S3 symmetry is manifest.
On the other hand we can still proceed with the theory (C.1) and construct the S3 bulk topological
operators by fusing interfaces. For instance the SPT stacking operation, τ , much like stacking in the 4d
57
Yang-Mills theory, provides interfaces, τ, στ, τ σ, τ στ ,
τ
(I) (J) (K) (J)
π c3 ∪ δc3 π c3 ∪ δc3
στ
(I) (J) (K) (I)
π c3 ∪ δc3 π c3 ∪ δc3
(C.5)
τσ
(I) (J) (J) (K)
π c3 ∪ δc3 π c3 ∪ δc3
τ στ
(I) (J) (I) (K)
π c3 ∪ δc3 π c3 ∪ δc3
(I) (I)
Other interfaces are also provided by the codimension-one condensation defects D6 , T6 , T 6 defined in
(4.31), (4.38) and (4.39). This can be argued from the action of the condensation defects on the U (I) surfaces
defects, where we denote by (I, J, K) a cyclic permutation of (S, C, V ),
(I)
D6
(I) (J) (K) (J)
π c3 ∪ δc3 π c3 ∪ δc3
(J)
D6
(I) (J) (I) (K)
π c3 ∪ δc3 π c3 ∪ δc3
(K)
D6
(C.6)
(I) (J) (J) (I)
π c3 ∪ δc3 π c3 ∪ δc3
T6
(I) (J) (K) (I)
π c3 ∪ δc3 π c3 ∪ δc3
T6
(I) (J) (J) (K)
π c3 ∪ δc3 π c3 ∪ δc3
(K)
where only D6 provides a Z2 0-form symmetry of the 7d DW theory, and therefore we can construct duality
defects of the boundary theory as twist defects, which become genuine after we gauge Z2 . The action of
58
(I)
these interfaces on gapped boundary conditions follows from the action on the c3 ,
τ: (LJ , LI ) 7→ (LJ , LK )
στ : (LJ , LI ) 7→ (LI , LK )
τσ : (LJ , LI ) 7→ (LK , LJ )
τ στ : (LJ , LI ) 7→ (LK , LI )
(I)
D6 : (LJ , LI ) 7→ (LJ , LK ) (C.7)
(J)
D6 : (LJ , LI ) 7→ (LK , LI )
(K)
D6 : (LJ , LI ) 7→ (LI , LJ )
T6 : (LJ , LI ) 7→ (LI , LK )
T6 : (LJ , LI ) 7→ (LK , LJ )
By combining these two type of interfaces we can now construct codimension one bulk symmetry defect
for the full S3 , that are
(K) (I) (J)
D6 , D6 τ , D6 τ στ , T6 τ σ , T 6 στ. (C.8)
The non-invertible S3 -ality boundary defect operators can now be constructed by considering twist defects
ending on the symmetry boundary (4.70).
References
[1] J. Kaidi, K. Ohmori, and Y. Zheng, “Kramers-Wannier-like Duality Defects in (3+1)D Gauge Theories,” Phys.
Rev. Lett. 128 no. 11, (2022) 111601, arXiv:2111.01141 [hep-th].
[2] Y. Choi, C. Cordova, P.-S. Hsin, H. T. Lam, and S.-H. Shao, “Noninvertible duality defects in 3+1 dimensions,”
Phys. Rev. D 105 no. 12, (2022) 125016, arXiv:2111.01139 [hep-th].
[3] L. Bhardwaj, L. E. Bottini, S. Schafer-Nameki, and A. Tiwari, “Non-invertible higher-categorical symmetries,”
SciPost Phys. 14 no. 1, (2023) 007, arXiv:2204.06564 [hep-th].
[4] S. Schafer-Nameki, “ICTP lectures on (non-)invertible generalized symmetries,” Phys. Rept. 1063 (2024) 1–55,
arXiv:2305.18296 [hep-th].
[5] S.-H. Shao, “What’s Done Cannot Be Undone: TASI Lectures on Non-Invertible Symmetries,”
arXiv:2308.00747 [hep-th].
[6] J. Frohlich, J. Fuchs, I. Runkel, and C. Schweigert, “Kramers-Wannier duality from conformal defects,” Phys.
Rev. Lett. 93 (2004) 070601, arXiv:cond-mat/0404051.
[7] Y. Choi, C. Cordova, P.-S. Hsin, H. T. Lam, and S.-H. Shao, “Non-invertible Condensation, Duality, and Triality
Defects in 3+1 Dimensions,” Commun. Math. Phys. 402 no. 1, (2023) 489–542, arXiv:2204.09025 [hep-th].
[8] J. Kaidi, K. Ohmori, and Y. Zheng, “Symmetry TFTs for Non-invertible Defects,” Commun. Math. Phys. 404
no. 2, (2023) 1021–1124, arXiv:2209.11062 [hep-th].
[9] D.-C. Lu and Z. Sun, “On triality defects in 2d CFT,” JHEP 02 (2023) 173, arXiv:2208.06077 [hep-th].
[10] D.-C. Lu, Z. Sun, and Z. Zhang, “Exploring G-ality defects in 2-dim QFTs,” arXiv:2406.12151 [hep-th].
[11] D. Gaiotto and J. Kulp, “Orbifold groupoids,” JHEP 02 (2021) 132, arXiv:2008.05960 [hep-th].
59
[12] F. Apruzzi, F. Bonetti, I. n. Garcı́a Etxebarria, S. S. Hosseini, and S. Schafer-Nameki, “Symmetry TFTs from
String Theory,” Commun. Math. Phys. 402 no. 1, (2023) 895–949, arXiv:2112.02092 [hep-th].
[13] D. S. Freed, G. W. Moore, and C. Teleman, “Topological symmetry in quantum field theory,”
arXiv:2209.07471 [hep-th].
[14] A. Antinucci, F. Benini, C. Copetti, G. Galati, and G. Rizi, “The holography of non-invertible self-duality
symmetries,” arXiv:2210.09146 [hep-th].
[15] F. Apruzzi, F. Bonetti, D. S. W. Gould, and S. Schafer-Nameki, “Aspects of categorical symmetries from branes:
SymTFTs and generalized charges,” SciPost Phys. 17 no. 1, (2024) 025, arXiv:2306.16405 [hep-th].
[16] F. Apruzzi, I. Bah, F. Bonetti, and S. Schafer-Nameki, “Noninvertible Symmetries from Holography and
Branes,” Phys. Rev. Lett. 130 no. 12, (2023) 121601, arXiv:2208.07373 [hep-th].
[17] J. J. Heckman, M. Hubner, E. Torres, X. Yu, and H. Y. Zhang, “Top down approach to topological duality
defects,” Phys. Rev. D 108 no. 4, (2023) 046015, arXiv:2212.09743 [hep-th].
[18] V. Bashmakov, M. Del Zotto, A. Hasan, and J. Kaidi, “Non-invertible symmetries of class S theories,” JHEP 05
(2023) 225, arXiv:2211.05138 [hep-th].
[19] A. Antinucci, C. Copetti, G. Galati, and G. Rizi, ““Zoology” of non-invertible duality defects: the view from
class S,” JHEP 04 (2024) 036, arXiv:2212.09549 [hep-th].
[20] L. Bhardwaj, T. Décoppet, S. Schafer-Nameki, and M. Yu, “Fusion 3-Categories for Duality Defects,”
arXiv:2408.13302 [math.CT].
[21] C. Lawrie, X. Yu, and H. Y. Zhang, “Intermediate defect groups, polarization pairs, and noninvertible duality
defects,” Phys. Rev. D 109 no. 2, (2024) 026005, arXiv:2306.11783 [hep-th].
[22] T. D. Décoppet, “Drinfeld centers and morita equivalence classes of fusion 2-categories,” arXiv:2211.04917
[math.QA].
[23] T. Lan and X.-G. Wen, “Classification of 3 + 1D bosonic topological orders (ii): The case when some pointlike
excitations are fermions,” Phys. Rev. X 9 (Apr, 2019) 021005.
[24] T. Lan, L. Kong, and X.-G. Wen, “Classification of (3 + 1)D bosonic topological orders: The case when pointlike
excitations are all bosons,” Phys. Rev. X 8 (Jun, 2018) 021074.
[25] D. Gaiotto, A. Kapustin, N. Seiberg, and B. Willett, “Generalized Global Symmetries,” JHEP 02 (2015) 172,
arXiv:1412.5148 [hep-th].
[26] M. Del Zotto, J. J. Heckman, D. S. Park, and T. Rudelius, “On the Defect Group of a 6D SCFT,” Lett. Math.
Phys. 106 no. 6, (2016) 765–786, arXiv:1503.04806 [hep-th].
[27] S. Monnier, “The anomaly field theories of six-dimensional (2,0) superconformal theories,” Adv. Theor. Math.
Phys. 22 (2018) 2035–2089, arXiv:1706.01903 [hep-th].
[28] J. J. Heckman and L. Tizzano, “6D Fractional Quantum Hall Effect,” JHEP 05 (2018) 120, arXiv:1708.02250
[hep-th].
[29] E. Witten, “Geometric Langlands From Six Dimensions,” arXiv:0905.2720 [hep-th].
[30] D. S. Freed and C. Teleman, “Relative quantum field theory,” Commun. Math. Phys. 326 (2014) 459–476,
[31] E. Witten, “AdS / CFT correspondence and topological field theory,” JHEP 12 (1998) 012,
arXiv:hep-th/9812012.
[32] S. Monnier, “Topological field theories on manifolds with Wu structures,” Rev. Math. Phys. 29 no. 05, (2017)
1750015, arXiv:1607.01396 [math-ph].
[33] S. Gukov, P.-S. Hsin, and D. Pei, “Generalized global symmetries of T [M ] theories. Part I,” JHEP 04 (2021)
232, arXiv:2010.15890 [hep-th].
60
[34] E. Witten, “Five-brane effective action in M theory,” J. Geom. Phys. 22 (1997) 103–133,
arXiv:hep-th/9610234.
[35] C. Cordova, T. T. Dumitrescu, and K. Intriligator, “2-Group Global Symmetries and Anomalies in
Six-Dimensional Quantum Field Theories,” JHEP 04 (2021) 252, arXiv:2009.00138 [hep-th].
[36] L. Bhardwaj and S. Schäfer-Nameki, “Higher-form symmetries of 6d and 5d theories,” JHEP 02 (2021) 159,
arXiv:2008.09600 [hep-th].
[37] F. Apruzzi, J. J. Heckman, and T. Rudelius, “Green-Schwarz Automorphisms and 6D SCFTs,” JHEP 02 (2018)
157, arXiv:1707.06242 [hep-th].
[38] F. Apruzzi, L. Bhardwaj, D. S. W. Gould, and S. Schafer-Nameki, “2-Group symmetries and their classification
in 6d,” SciPost Phys. 12 no. 3, (2022) 098, arXiv:2110.14647 [hep-th].
[39] C. Cordova, T. T. Dumitrescu, and K. Intriligator, “Deformations of Superconformal Theories,” JHEP 11 (2016)
135, arXiv:1602.01217 [hep-th].
[40] F. Apruzzi, M. Fazzi, D. Rosa, and A. Tomasiello, “All AdS7 solutions of type II supergravity,” JHEP 04 (2014)
064, arXiv:1309.2949 [hep-th].
[41] J. Louis and S. Lüst, “Supersymmetric AdS7 backgrounds in half-maximal supergravity and marginal operators
of (1, 0) SCFTs,” JHEP 10 (2015) 120, arXiv:1506.08040 [hep-th].
[42] J. J. Heckman, D. R. Morrison, and C. Vafa, “On the Classification of 6D SCFTs and Generalized ADE
Orbifolds,” JHEP 05 (2014) 028, arXiv:1312.5746 [hep-th]. [Erratum: JHEP 06, 017 (2015)].
[43] J. J. Heckman, D. R. Morrison, T. Rudelius, and C. Vafa, “Atomic Classification of 6D SCFTs,” Fortsch. Phys.
63 (2015) 468–530, arXiv:1502.05405 [hep-th].
[44] M. Del Zotto, J. J. Heckman, A. Tomasiello, and C. Vafa, “6d Conformal Matter,” JHEP 02 (2015) 054,
[45] O. Aharony, N. Seiberg, and Y. Tachikawa, “Reading between the lines of four-dimensional gauge theories,”
JHEP 08 (2013) 115, arXiv:1305.0318 [hep-th].
[46] J. Kaidi, G. Zafrir, and Y. Zheng, “Non-invertible symmetries of N = 4 SYM and twisted compactification,”
JHEP 08 (2022) 053, arXiv:2205.01104 [hep-th].
[47] Y. Choi, H. T. Lam, and S.-H. Shao, “Non-invertible Gauss law and axions,” JHEP 09 (2023) 067,
arXiv:2212.04499 [hep-th].
[48] W. Cui, B. Haghighat, and L. Ruggeri, “Non-Invertible Surface Defects in 2+1d QFTs from Half Spacetime
Gauging,” arXiv:2406.09261 [hep-th].
[49] L. Bhardwaj, Y. Lee, and Y. Tachikawa, “SL(2, Z) action on QFTs with Z2 symmetry and the Brown-Kervaire
invariants,” JHEP 11 (2020) 141, arXiv:2009.10099 [hep-th].
[50] P.-S. Hsin, W. Ji, and C.-M. Jian, “Exotic invertible phases with higher-group symmetries,” SciPost Phys. 12
no. 2, (2022) 052, arXiv:2105.09454 [cond-mat.str-el].
[51] C. Arf, “Untersuchungen über quadratische formen in körpern der charakteristik 2. (teil i.).,” Journal für die
reine und angewandte Mathematik 1941 no. 183, (1941) 148–167.
[52] M. A. Kervaire, “A manifold which does not admit any differentiable structure,” Commentarii Mathematici
Helvetici 34 (1960) 257–270.
[53] W. Browder, “The kervaire invariant of framed manifolds and its generalization,” Annals of Mathematics 90
no. 1, (1969) 157–186.
[54] E. H. Brown, “Generalizations of the Kervaire Invariant,” Annals of Mathematics 95 no. 2, (1972) 368–383.
[55] J. Tian and Y.-N. Wang, “A Tale of Bulk and Branes: Symmetry TFT of 6D SCFTs from IIB/F-theory,”
arXiv:2410.23076 [hep-th].
61
[56] M. N. Khan, J. C. Y. Teo, and T. L. Hughes, “Anyonic symmetries and topological defects in abelian topological
phases: An application to the ADE classification,” Phys. Rev. B 90 (Dec, 2014) 235149.
[57] J. C. Y. Teo, T. L. Hughes, and E. Fradkin, “Theory of Twist Liquids: Gauging an Anyonic Symmetry,” Annals
Phys. 360 (2015) 349–445, arXiv:1503.06812 [cond-mat.str-el].
[58] P. Boyle Smith and Y. Zheng, “Backfiring Bosonisation,” arXiv:2403.03953 [hep-th].
[59] K. Roumpedakis, S. Seifnashri, and S.-H. Shao, “Higher Gauging and Non-invertible Condensation Defects,”
Commun. Math. Phys. 401 no. 3, (2023) 3043–3107, arXiv:2204.02407 [hep-th].
[60] C. Cordova, P.-S. Hsin, and C. Zhang, “Anomalies of Non-Invertible Symmetries in (3+1)d,” arXiv:2308.11706
[hep-th].
[61] A. Antinucci, F. Benini, C. Copetti, G. Galati, and G. Rizi, “Anomalies of non-invertible self-duality
symmetries: fractionalization and gauging,” arXiv:2308.11707 [hep-th].
[62] Z. Sun and Y. Zheng, “When are Duality Defects Group-Theoretical?,” arXiv:2307.14428 [hep-th].
62

Non-Invertible Symmetries in 6d From Green-Schwarz Automorphisms

Uploaded by

Copyright:

Available Formats

Non-Invertible Symmetries in 6d From Green-Schwarz Automorphisms

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Non-Invertible Symmetries in 6d From Green-Schwarz Automorphisms

Uploaded by

Copyright:

Available Formats

Non-Invertible Symmetries in 6d from

⋆ Dipartimento di Fisica e Astronomia “Galileo Galilei”, Università di Padova,

We construct non-invertible symmetries in 6d N = (2, 0) superconformal field theories that arise

2 6d N = (2, 0) Theories: Polarizations and GS Automorphisms 5

3 S3 -ality Defects in 6d from Half-space Gauging 10

4 SymTFT Perspective on S3 -ality Defects 24

A Computation of Fusion of Triality and Condensation Defects 40

C Bosonic SymTFT and Interfaces 57

• Gauging an outer automorphism of the SymTFT.

2 6d N = (2, 0) Theories: Polarizations and GS Automorphisms

2.1 Absolute Theories with GS Automorphisms

2. G = S3 , i.e. S3 -ality, these are the automorphisms of K = Cso(8) .

2.2.1 Defect Group and Polarizations

Polarization: LS = (1, 0) LC = (0, 1) LV = (1, 1) (2.5)

2.2.2 Green-Schwarz Automorphisms and Dualities

Non-Invertible Symmetry Defects. Combining a Green-Schwarz automorphism and a topological ma-

3 S3 -ality Defects in 6d from Half-space Gauging

Defects. To define duality/triality/S3 -ality defects in a given absolute theory, we combine:

spacetime M6 is divided into three regions:

• M6<0 hosts the initial theory,

3.2 Non-invertible Duality Defects

3.2.1 Gauging 2-form Symmetries

(J) (I) (I)

which means that

3.2.2 Definition of Duality Interface and Defects

is the appropriate normalization factor.5

which is spelled out in terms of the partition function as

3.2.3 Fusion of Duality Defects

(J) (I) (I) (I)

of the 2-form symmetry defect  Z 

3.3 Non-invertible Triality Defects

3.3.1 Twisted Gauging

as well as from the partition functions

(I) (I) (I)

(K) (I) (K) (J)

with Z(Y,M6 ) defined in equation (3.51).

3.3.2 Definition of Triality Interfaces and Defects

The above picture translates to the following equation:

with the normalization factor N(M ⩾0 , Z2 ) defined in equation (3.12).

T5 (M5 ) = T5 (M5 ) ⊗ τ σ(M5 ) . (3.56)

(I) T5 (M5 ) (K) τ σ(M5 ) (I)

with corresponding equation:

with the normalization factor N(M ⩾0 , Z2 ) defined in equation (3.12).

with corresponding equation:

T 5 (M5 ) = T 5 (M5 ) ⊗ στ (M5 ) . (3.62)

(I) T 5 (M5 ) (J) στ (M5 ) (I)

with corresponding equation:

• duality defect D5 (M5 ) defined in (3.15),

• triality defects T5 (M5 ) and T5 (M5 ), defined in (3.57) and (3.64).

We denote the Poincaré dual of M3 ∈ H3 (M5 , Z2 ) by PD(M3 ) ∈ H 2 (M5 , Z2 ).

Duality Defect Fusions.

Triality Defect Fusions.

Fusions between Duality and Triality Defects: S3 -ality Structure

4 SymTFT Perspective on S3 -ality Defects

4.1 SymTFT for the 6d N = (2, 0) so(8) theory

(1) (id) (I) (J) (K)

(I) (J) ′ (J) (I)

which imply (in R/2πZ) 

Alternative presentation of the SymTFT. The Z2 3-form Dijkgraaf-Witten (DW) theory

of the 2-form symmetry defect Z