---
title: Triple-Point State-Sum in TQFT
url: https://www.emergentmind.com/topics/triple-point-state-sum
type: topic
---

# Triple-Point State-Sum in TQFT

Searching arXiv for the cited papers and closely related work on universal state sums and defect state-sum models.
First, I’ll retrieve the universal state-sum paper by its arXiv ID.
Now I’ll retrieve the defect state-sum paper by its arXiv ID.
Finally, I’ll check for closely related state-sum literature that the universal construction explicitly specializes to.
Triple-point state-sum denotes a class of local amplitudes in state-sum topological quantum field theory in which the fundamental local contribution is attached to a trivalent junction of codimension-1 or defect data. In the universal framework of Morrison and Walker, triple-point structure appears as a specialization of a general $(n+1)$-dimensional state sum built from a linear, $H$-pivotal, finite, weakly complete, semisimple $n$-category equipped with conjugation and a nondegenerate evaluation map [2104.02101]. In the Turaev–Viro–Barrett–Westbury framework with defects, triple points arise as point defects on an oriented embedded defect surface, labeled by bimodule natural transformations between composable bimodule functors; their amplitudes are encoded by cyclic evaluations of polygon diagrams and inserted into a triangulation-independent state sum [2205.06874]. Taken together, these constructions show that “triple-point” is not a separate axiomatic model but a recurring local feature of broader state-sum formalisms, realized in three dimensions by theta-type junctions and in four dimensions by spun-theta configurations.

## 1. Universal categorical setting

The universal state sum takes as input “a linear, H-pivotal, finite, weakly complete, semisimple n-category equipped with conjugation and a nondegenerate evaluation map” [2104.02101]. The $H$-pivotal structure means that morphisms are equipped with an action of automorphisms of $n$-balls carrying an $H$-structure, so that one can define string diagrams on $H$-manifolds. Conjugation implements orientation reversal: a $k$-morphism $x$ of shape $X$ is sent to a $k$-morphism of shape $\breve X$, and gluing $x$ to $\breve x$ along their common boundary produces a diagram on $S^k$ that can be evaluated. The nondegenerate evaluation map is $z = Z(B^{n+1}) \in Z(S^n)$, and it induces pairings
$$
A(B^n; c) \times A(B^n; c) \to \mathbb C
$$
by $x,y \mapsto \operatorname{ev}(\breve x \cup y)$; nondegeneracy is required for all boundary conditions [2104.02101].

The finiteness and semisimplicity hypotheses are formulated in terms of the spaces and categories assigned to bordisms. For existence of the $(n+1)$-dimensional part of the TQFT, the construction requires that $A(S^j\times B^{n-j}; c)$ be finite-dimensional for all $j$ and boundary conditions $c$, that $A(S^j\times B^{n-j-1}; c)$ be a semisimple $1$-category for all $j$ and $c$, and that there exist $z \in Z(S^n)$ inducing nondegenerate inner products on all $A(B^n;c)$ [2104.02101]. Weak completeness is imposed through minimal morphisms: a $k$-morphism $x$ is minimal if $\operatorname{End}(\operatorname{id}^{\,n-k-1}(x))$ is simple, and every $k$-morphism must be isomorphic to a sum of minimal ones. If this fails, the paper states that one completes to a Morita equivalent weakly complete $n$-category [2104.02101].

This categorical package is the source of all local amplitudes, including triple-point contributions. A plausible implication is that the triple-point sector is controlled less by ad hoc combinatorics than by the same duality, semisimplicity, and gluing structures that govern the entire state sum.

## 2. Universal state-sum formula and local weights

For a closed $(n+1)$-dimensional $H$-manifold $W$ with a cell or handle decomposition $\mathcal H$, the universal partition function is
$$
Z(W)=\sum_{\beta\in\mathcal L(\mathcal H)}\ \prod_{j=0}^{n+1}\ \prod_{h\in j\text{-handles}}\frac{\operatorname{ev}(\beta(h))}{N(\beta(h))},
$$
where $\mathcal L(\mathcal H)$ is the finite set of labelings of the $j$-cells or $j$-handles by minimal $(n+1-j)$-morphisms [2104.02101]. The boundary of each handle determines an unlabeled cell complex in the linking sphere $S^n$; a labeling $\beta$ turns this into a labeled string diagram $\beta(\partial h)$, and $\operatorname{ev}(\beta(\partial h))$ is the evaluation of that diagram.

Labelings are assigned sequentially, starting with the top-dimensional cells. The $(n+1)$-cells are labeled by minimal $0$-morphisms, the $n$-cells by minimal $1$-morphisms, the $(n-1)$-cells by minimal $2$-morphisms compatible with their determined boundaries, and so on [2104.02101]. At each stage, the higher-dimensional labels determine a string diagram on the linking sphere of the lower-dimensional cell, and the admissible labels are chosen from equivalence classes of minimal morphisms matching that boundary.

The normalization factor $N(\cdot)$ is defined inductively from the sphere-trace. For a $k$-morphism $x$,
$$
\operatorname{tr}_s(x):=\operatorname{ev}\!\big(x\times S^{n-k}\cup (\operatorname{trev}x)\times B^{n-k+1}\big),
$$
and
$$
N(x):=\sum_y \frac{\operatorname{tr}_s(y)^2}{N(y)},
$$
where $y$ ranges over minimal endomorphisms of $x$; the induction begins with $N(x)=\operatorname{tr}_s(x)^2$ when $x$ is an $n$-morphism [2104.02101]. The paper identifies $N(x)$ with the TQFT inner product $\langle x\times S^{n-k},\breve x\times S^{n-k}\rangle$, so the denominator is not auxiliary normalization but the gluing-theoretic dual pairing for the attaching region.

In this formulation, a triple-point amplitude is simply a particular local evaluation $\operatorname{ev}(\beta(h))$ associated to a handle whose linking sphere carries a trivalent configuration. The significance of the universal formalism is that such amplitudes are already built into the same $\operatorname{ev}/N$ mechanism as all other local terms.

## 3. Triple-point structures in low-dimensional specializations

In the universal construction, triple points are identified with local singular strata where three codimension-1 sheets meet along a codimension-2 locus [2104.02101]. Their most explicit manifestations occur in the specializations to Turaev–Viro and Douglas–Reutter type models.

For $n=2$, corresponding to three-dimensional Turaev–Viro, generic dual cell decompositions have $1$-cells adjacent to three $2$-cells. The boundary of a $1$-handle is then a $2$-sphere carrying a theta graph with three edges and two vertices. The paper states that this is the usual theta symbol of the fusion category, with edge labels given by the adjacent $2$-handle labels and vertex labels given by the chosen basis morphism and its conjugate [2104.02101]. This is the canonical three-dimensional triple junction.

For $n=3$, corresponding to four-dimensional theories, the triple-point analogue appears in a different codimension. In the Douglas–Reutter specialization, the boundary of a $2$-handle is a spun-theta $2$-complex “built out of a circle and three 2-cells,” representing three sheets meeting along an $S^1$ triple line [2104.02101]. By contrast, in the Crane–Yetter specialization with trivial $0$- and $1$-morphisms, the $1$-handle boundary is a mutant theta with four edges, so the most visible local singularity is a four-sheet junction rather than a trivalent one [2104.02101].

The relation among these cases may be summarized as follows.

| Model | Local triple-point feature | Handle carrying the weight |
|---|---|---|
| Turaev–Viro | Theta graph with three edges | $1$-handle |
| Crane–Yetter | Triple-point analogue suppressed; mutant theta with four edges appears | $1$-handle |
| Douglas–Reutter | Spun-theta $2$-complex with three sheets meeting along a circle | $2$-handle |

This suggests that “triple point” should be understood geometrically rather than dimension-specifically: in three dimensions it is a trivalent junction in a linking $S^2$, while in four dimensions it is realized as a triple-line neighborhood in a linking $S^3$.

## 4. Explicit local formulas and orientation conventions

In the Turaev–Viro specialization, for $n=2$, $H=SO(2)$, and $\mathcal C$ a pivotal fusion category, the local $1$-handle weight is
$$
\operatorname{ev}(\beta(h_1))=\lambda\,\Theta_{abc\alpha},
$$
and the norm of the $1$-handle label is
$$
N(\alpha)=\big(\lambda\,\Theta_{abc\alpha}\big)^2
$$
[2104.02101]. Here $a,b,c$ are the simple object labels on adjacent $2$-handles, and $\alpha$ is a basis morphism in $\operatorname{mor}(1\to abc)$, with duals inserted as required by orientations. For non-generic cell decompositions, the theta graph is replaced by a mutant theta graph with $k$ edges, and the local weight becomes $\operatorname{ev}(\Theta^m(h_1))$ with norm $\big[\operatorname{ev}(\Theta^m(\alpha))\big]^2$ [2104.02101].

In the Douglas–Reutter specialization, for $n=3$, $H=SO(3)$, and fusion $2$-categorical input, the explicit triple-junction term is attached to a $2$-handle:
$$
\operatorname{ev}(\beta(h_2))=\lambda\,\operatorname{ev}(\text{spun-}\Theta),
$$
while the norm of a $2$-handle label is $1$ in $\mathrm{Vec}$ [2104.02101]. The corresponding local geometry is a circle with three attached $2$-disks, and the weight is determined by the fusion $2$-category data. In the same specialization, $1$-handles carry $\lambda\,\operatorname{ev}(\mathrm{DCTet})$, and $3$-handle norms involve $\operatorname{ev}(S^2_m)^2$ multiplied by the global dimension of $\operatorname{End}(m)$ [2104.02101].

Orientation handling is part of the local definition. In the Turaev–Viro case, if the orientation of one normal fiber agrees with the orientation of a region, it disagrees with the orientation of the opposite region, so one side carries $\alpha$ while the other carries $\breve\alpha \in V_{c^*b^*a^*}$ [2104.02101]. The paper attributes the well-definedness of such junction evaluations to the pivotal and conjugation structures. This means that the triple-point amplitude is intrinsically orientation-sensitive, but the sensitivity is absorbed into the dualization and conjugation conventions rather than treated as an extra combinatorial rule.

## 5. Defect-surface triple points in TVBW-type models

A different realization of triple-point state sums appears in the Turaev–Viro–Barrett–Westbury model with defects. Here one begins with a spherical fusion category $\mathcal C$ over $\mathbb C$, with simple representatives $I\subset \operatorname{Ob}(\mathcal C)$, quantum dimensions $d_i=\operatorname{tr}(1_i)$, and global dimension
$$
D_{\mathcal C}:=\sqrt{\dim(\mathcal C)}, \qquad \dim(\mathcal C)=\sum_{i\in I} d_i^2
$$
[2205.06874]. A defect surface $\Sigma$ is an oriented embedded PL $2$-submanifold of an oriented compact PL $3$-manifold $M$, and each oriented defect area is labeled by a finite semisimple $\mathcal C$–$\mathcal C$ bimodule category $(\mathcal M,\theta)$ with coherence isomorphisms
$$
c_{x,y,m}:(xy)\triangleright m\to x\triangleright (y\triangleright m),\qquad
d_{m,x,y}:(m\triangleleft x)\triangleleft y\to m\triangleleft (xy),
$$
$$
b_{x,m,y}:(x\triangleright m)\triangleleft y\to x\triangleright (m\triangleleft y),
$$
together with a $\mathcal C$-bimodule trace $\theta$ [2205.06874]. Reversal of the defect-surface orientation replaces $\mathcal M$ by the opposite bimodule category $\mathcal M^\#$.

Defect lines on $\Sigma$ are oriented line segments labeled by $\mathcal C$–$\mathcal C$ bimodule functors $F:\mathcal M_a\to \mathcal M_b$, equipped with coherence maps
$$
s^F_{x,m}:F(x\triangleright m)\to x\triangleright F(m),\qquad
t^F_{m,y}:F(m)\triangleleft y\to F(m\triangleleft y),
$$
which satisfy the hexagon constraint of Equation (2.19) in the paper [2205.06874]. Triple points are point defects, i.e. vertices of the directed graph on $\Sigma$, labeled by bimodule natural transformations
$$
\alpha:F_1\star F_2\Rightarrow F_3,
$$
with source and target determined by orientation and cyclic ordering of the incident defect lines [2205.06874]. These transformations lie in $\operatorname{Hom}_{(\mathcal C,\mathcal C)}(F_1\star F_2,F_3)$ and satisfy module and right-module naturality.

The triangulation is required to be transversal and generic with respect to $\Sigma$, so a generic transversal tetrahedron intersects the defect surface either in three edges incident at a common vertex, in a quadrilateral, or not at all [2205.06874]. The state sum is
$$
Z(M,\Sigma,\Gamma; l_{\partial M}, b_{\partial M})=
\prod_{\text{regions }r}\dim(\mathcal C_r)^{-\sum_{v\in V_r}\epsilon(v)}
\cdot
\sum_{\Lambda}\sum_{\text{bases on internal faces}}
\left[
\prod_{t\in T_3}\{6j\}(t,\Lambda)\cdot \prod_{e\in E}\dim(l(e))^{\epsilon(e)}
\right],
$$
where $\epsilon(v)=1$ for internal vertices and $\epsilon(e)=1$ for internal edges, with value $1/2$ on boundary vertices and edges [2205.06874]. Bulk tetrahedra contribute the ordinary TVBW $6j$-symbols, while defect-intersecting tetrahedra contribute generalized $6j$-symbols defined as cyclic evaluations of polygon diagrams.

Within this formalism, the triple-point amplitude at a trivalent defect vertex $v$ is the cyclic evaluation on the small defect disc surrounding $v$:
$$
Z_v=\theta_{\mathcal M}(E_v),
$$
where $E_v$ is the endomorphism built from $\alpha$ and the coherence isomorphisms transporting between $(F_1\star F_2)(X)$ and $F_3(X)$ for the appropriate boundary-labeled object $X$ [2205.06874]. The paper gives the concrete form
$$
E_v=s^{F_1}_{c,-}\circ t^{F_2}_{-,d}\circ b_{c,-,d}\circ \alpha_{(-)}
\circ (b_{c,-,d})^{-1}\circ (t^{F_2}_{-,d})^{-1}\circ (s^{F_1}_{c,-})^{-1},
$$
and then expresses the amplitude through projection/inclusion bases as a finite sum over simple labels [2205.06874]. This is exactly the cyclic evaluation of the polygon diagram with internal vertex $\alpha$.

## 6. Invariance, coherence, and geometric sensitivity

In the universal state sum, invariance is derived from TQFT gluing rather than imposed as explicit combinatorial identities. For a decomposition $W=M\cup \breve M\cup N$, the gluing formula is
$$
Z(W_{\mathrm{gl}})(c_{\mathrm{gl}})
=
\sum_e
\frac{Z(W)(c\cup e\cup \operatorname{trev}e)}{\langle e,\operatorname{trev}e\rangle},
$$
where the sum is over an orthogonal basis $e$ of $A(M;b)$ [2104.02101]. Associativity of gluing is proved for decompositions $W=W_1\cup W_2\cup W_3$, showing that different orders of gluing yield the same answer [2104.02101]. Since the normalization $N(\beta(h))$ is exactly the relevant inner product, the $\operatorname{ev}/N$ ratios satisfy the identities required for handle slides and cancellations. The paper explicitly states that bubble moves and $2$–$3$-type local changes in the dual skeleton follow from gluing identities and semisimplicity [2104.02101]. Triple-point creation and annihilation are therefore absorbed into the same invariance mechanism.

In the defect model, the corresponding technology is formulated in terms of polygon diagrams. The generalized $6j$-symbols incorporate module pentagon and triangle identities, the bimodule-functor hexagon identity, naturality of the module constraints, and bimodule naturality [2205.06874]. The cutting and gluing identities for polygon diagrams include gluing sides, gluing around a vertex, gluing a $2$-gon, and insertion of diagrams from $\mathcal C$ [2205.06874]. These identities imply that the state sum for a $3$-ball bisected by a defect disc reduces to the cyclic evaluation of the projected dual graph on the disc, and then yield invariance under generalized Pachner moves and stellar subdivisions. The main theorem states that if two transversal and generic triangulations agree on the boundary, their state sums coincide [2205.06874].

The defect framework also supports geometric sensitivity. The paper states that defect surfaces can detect genus and can be sensitive to embedding, with examples in which the invariant depends on the complement fundamental group of an embedded torus in $S^3$ [2205.06874]. A plausible implication is that triple-point amplitudes in this setting are not merely formal local weights but constituents of invariants responsive to both intrinsic and extrinsic topology.

## 7. Relations to established models and interpretive scope

The universal state sum explicitly specializes to most previously known state sums, including Turaev–Viro, Dijkgraaf–Witten, Crane–Yetter, Douglas–Reutter, the Witten–Reshetikhin–Turaev surgery formula, and Brown–Arf [2104.02101]. In the Turaev–Viro specialization, the local data are organized by handle dimension: $3$-handle boundaries evaluate to $\lambda$, $2$-handle boundaries to $\lambda d_a$, $1$-handle boundaries to $\lambda\Theta_{abc\alpha}$, and $0$-handle boundaries to $\lambda\,\operatorname{ev}(\mathrm{Tet})$ [2104.02101]. For closed $3$-manifolds, the resulting formula has $\chi(W)=0$, so there is no $\lambda$-dependence [2104.02101]. The triple points are exactly the trivalent vertices on the boundaries of $1$-handles.

In Crane–Yetter, the local structure is shifted: the $1$-handle boundary is a mutant theta graph with four edges and the $0$-handle boundary is the $1$-skeleton of a $4$-simplex in $S^3$ [2104.02101]. The paper remarks that the $2$-handle spun-theta triple junction does not appear explicitly in the premodular $3$-category case because the $1$-morphisms are trivial, whereas in the Douglas–Reutter generalization the spun-theta triple junction is explicit [2104.02101]. This distinction addresses a common possible misconception: not every four-dimensional state sum displays triple-point data at the same combinatorial locus, even when all arise from the same universal construction.

The defect-surface theory is related to the Drinfeld center. When the defect surface is trivial and labeled by $\mathcal C$ as a bimodule category over itself, bimodule endofunctors correspond to objects of $Z(\mathcal C)$, and defect lines and point defects realize ribbon diagrams on the surface; the state sum then factorizes as the TVBW invariant of the defect-free manifold times the ribbon link evaluation in $Z(\mathcal C)$ [2205.06874]. In examples with $\mathcal C=\mathrm{Vec}_G$, indecomposable bimodule categories correspond to transitive $G$-sets with a class in $H^2(L,\mathbb C^\times)$, and a simple triple point with $\alpha=\operatorname{id}_{F_1\star F_2\to F_3}$ yields a weighted count of compatible boundary labelings [2205.06874].

Taken together, these results support a broad interpretation of triple-point state-sum structures. In one direction, the universal theory shows that theta and spun-theta amplitudes are local manifestations of a general $\operatorname{ev}/N$ TQFT formalism [2104.02101]. In the other, the defect TVBW theory shows that trivalent point defects can be encoded by bimodule natural transformations and computed through cyclic polygon evaluations, with triangulation independence proved by categorical coherence and cutting–gluing identities [2205.06874]. The common theme is that triple-point amplitudes are not isolated combinatorial gadgets; they are localized evaluations of higher-categorical or defect-categorical data within globally invariant state sums.

Source: https://www.emergentmind.com/topics/triple-point-state-sum