Covariant Quantum Geometry as a State Sum: Spin Foams, Vertex Amplitudes, and the Refinement Problem of Boundary Amplitudes

1 Introduction

1.1 The reconstruction thesis for covariant dynamics

The canonical picture of loop quantum gravity gives a kinematics: a state of quantum three-geometry is a spin network, a graph whose edges carry $\mathrm{SU}(2)$ representations and whose vertices carry intertwiners, and geometry appears as the discrete spectrum of area and volume operators built from those labels. That picture is the subject of a companion paper. What it does not give, on its own, is dynamics. A theory of gravity must say how one three-geometry evolves into another, and general covariance forbids doing so by a preferred time coordinate. The covariant answer is a sum over histories: instead of evolving a spin network by a Hamiltonian, one sums over all “quantum four-geometries” interpolating between a fixed incoming and outgoing spin network. A spin foam is precisely such a history, and the sum over them is the spin foam amplitude.

This paper works in that covariant layer, under the organizing principle shared across a modular library of twelve quantum-gravity regimes:

Quantum gravity is not primarily the quantization of material objects in spacetime. It is the reconstruction of spacetime, matter, causality, and measurement from invariant representation structures.

In the present regime the invariant structure is a colored two-complex. A history is a two-complex $\mathcal{F}$ whose faces are labeled by irreducible representations and whose edges are labeled by intertwiners; the amplitude of that history is a product of local recoupling invariants, one per face, edge, and vertex; and the physical content is the boundary amplitude obtained by summing over the bulk labels. Nowhere in this data is there a spacetime metric. A smooth four-geometry, with its Einstein–Hilbert or Regge action, is not put in by hand: it is recovered, if at all, only in an asymptotic (large-spin) limit of the representation-theoretic sum. Whether that recovery survives refinement of the two-complex into a genuine continuum theory is the central open problem of the field, and we treat it as open throughout.

1.2 What is settled and what is not

Honesty about status is the point of this series, so we state it at the outset. The following are theorems or standard constructions in a well-defined mathematical setting, and we tag them [S]:

• For a finite two-complex with a cutoff on the representation labels, the state sum is a finite sum of products of finite-dimensional recoupling invariants, hence a well-defined amplitude map between boundary Hilbert spaces (5).

• The topological $BF$ state sum built from the $\{15j\}$ symbol is invariant under Pachner moves once regularized by a quantum group, so it is a piecewise-linear invariant of the triangulated four-manifold (7).

• Amplitudes glue along shared boundary spin networks, giving the state sum the composition law of a functor on a cobordism-like category (6).

• The Engle–Pereira–Rovelli–Livine four-simplex vertex amplitude, evaluated on boundary data peaked on a nondegenerate Regge geometry, has a large-spin stationary-phase asymptotic whose leading term is proportional to $\cos(S_{\mathrm{Regge}})$ (12).

The following are physical hypotheses, well motivated but not proved, and we tag them [H]:

• That the Engle–Pereira–Rovelli–Livine or Freidel–Krasnov vertex is the correct quantum dynamics of gravity. The vertex is fixed by imposing a discretized simplicity constraint, but the choice of how to impose it (strongly, weakly, on which variables), the face amplitude, and the measure are physical inputs, not consequences of an axiom.

• That summing over refinements of the two-complex, or over two-complexes themselves, defines a continuum theory that reproduces general relativity. This is open. The known obstruction, the flatness problem (7.2), is that a naive large-spin limit at fixed complex is dominated by flat geometries.

The dictionary of the library assigns the entry spin foam two-complex with face/edge labels the bare status [H], the only bare-[H] entry among the core geometric regimes, precisely because the dynamics and its continuum limit remain a heuristic proposal rather than a settled structure. We honor that label: the abstract and this introduction state the continuum-limit status as unresolved, and every claim in the body carries its warrant explicitly.

1.3 Contributions and outline

We do not claim new physics. The contribution is a rigorous, status-labeled synthesis of the covariant-dynamics regime organized around one invariant — the boundary amplitude and its refinement behavior — together with a companion computation that exhibits that invariant exactly in the solvable topological case.

2 fixes the reconstruction pipeline and the status calculus. 3 defines colored two-complexes dual to triangulations, the face/edge/vertex data, and the state sum, and states the well-definedness and gluing results. 4 gives the constraint route from $BF$ theory to gravity: the topological $BF$ state sum and its invariance, the simplicity constraints, the Barrett–Crane model, and the $Y_{\gamma}$ map defining the Engle–Pereira–Rovelli–Livine vertex. 5 states the semiclassical asymptotics. 6 works the four-simplex example explicitly. 7 treats refinement, cylindrical consistency, the flatness problem, and the effective-spin-foam reframing honestly. 8 locates the regime among the others through the weakest-link calculus. 9 describes the Haskell model, whose core module is reproduced in 13. 10 lists the open problems, and [sec:discussion,sec:conclusion] conclude.

2 Mathematical framework: reconstruction from representation data

2.1 The realization pipeline

The library models every physical representation as a translation from a mathematical object $M$ to a physical target, factored through a realization: $$M \;\longmapsto\; \Phi(M) \;\longmapsto\; \mathrm{Real}_\alpha(\Phi(M)) \;\longmapsto\; \text{physical representation} \;\longmapsto\; \mathrm{Obs}.$$ For spin foams the mathematical object $M$ is a colored two-complex; the intermediate $\Phi(M)$ is its state sum, a number attached to boundary data; the realization $\mathrm{Real}_\alpha$ chooses a model (which group, which vertex amplitude, which face amplitude and measure); and the observable is the boundary transition amplitude $\mathcal{Z}(\mathcal{F}):\mathcal{H}_{\Gamma_{\mathrm{in}}}\to\mathcal{H}_{\Gamma_{\mathrm{out}}}$. The index $\alpha$ matters: Barrett–Crane, Engle–Pereira–Rovelli–Livine, and Freidel–Krasnov are different realizations of the same abstract state-sum object, and they differ in physically consequential ways. Keeping $\alpha$ explicit is what prevents the honest statement “the topological state sum is a triangulation invariant” from silently upgrading into the false statement “the gravitational state sum is a continuum theory.”

2.2 The status calculus

We use three ordered warrants $\mathsf{S} < \mathsf{H} < \mathsf{P}$: $\mathsf{S}$ for a standard mathematical or mathematical-physics correspondence, $\mathsf{H}$ for a strong heuristic physical representation, and $\mathsf{P}$ for a speculative ontological extension. A status is a tuple rendered [S], [H], [P], or a composite such as [S/H]; composition takes the worst component.

The rule is not decoration. It is the precise sense in which the library is modular rather than unified: chaining spin-network kinematics (status [S/H]) to spin-foam dynamics (status [H]) yields a composite bounded above by [H], no matter how many rigorous theorems live inside the kinematical factor. One cannot buy credibility for the dynamics with the rigor of the kinematics.

2.3 The relevant dictionary entries

Three entries of the twenty-six-row dictionary bear on this paper.

Spin network: graph labeled by $\mathrm{SU}(2)$ representations $=$ [S/H]. The boundary data. The kinematical Hilbert space and the area/volume spectra are [S] theorems; the physical identification with the geometry of our universe is [H].

Spin foam two-complex with face/edge labels $=$ [H]. The subject of this paper. No assumptions field is populated; the limitation clause reads “continuum/ refinement limit and phenomenology are active research problems.”

Cobordism category $=$ [S/H]. The categorical grammar in which a spin foam is a morphism from an incoming to an outgoing spin network, and gluing is composition.

The composite of the first two, under 1, is [H]: a complete covariant quantum-gravity dynamics built from spin-network kinematics plus spin-foam amplitudes inherits the bare-[H] label of the dynamics. This is exactly why the roadmap’s task for this module demands “record entries with explicit continuum-limit limitations” rather than asserting the package as settled.

2.4 Gauge and boundary as retained structure

A recurring structural fact across the library is that physical content is a groupoid, not a coarse quotient: gluing of local data is by equivalence, so automorphism and stabilizer data (gauge redundancy, symmetry) is retained rather than erased. In the spin-foam regime this appears twice. First, the edge intertwiners are exactly the gauge-invariant data at a node, the $\mathrm{SU}(2)$ (or $\mathrm{SL}(2,\mathbb{C})$) invariant subspace of a tensor product of representations; the state sum integrates over the group at each edge, which is the discrete avatar of imposing the Gauss constraint. Second, the boundary of a spin foam is a spin network taken modulo spatial diffeomorphisms, and the amplitude is a pairing on that quotient. The reconstruction never works with a metric; it works with representation labels and invariants, and the geometry is a derived quantity.

3 Colored two-complexes and the state sum

3.1 Two-complexes dual to triangulations

The spin foams relevant to four-dimensional gravity arise as the two-skeleton of the cell complex dual to a triangulation $\Delta$ of a four-manifold $\mathcal{M}$. Under the duality, $$\text{4-simplices of }\Delta \;\leftrightarrow\; \text{vertices }v,\quad \text{tetrahedra} \;\leftrightarrow\; \text{edges }e,\quad \text{triangles} \;\leftrightarrow\; \text{faces }f .$$ A four-simplex has five tetrahedra and ten triangles on its boundary; dually, each vertex $v$ is met by five edges and ten faces, and the local combinatorics at $v$ is the complete graph $K_5$ whose five nodes are the foam edges and whose ten links are the foam faces. This $K_5$ pattern is what fixes the vertex amplitude to be a $\{15j\}$-type recoupling invariant. When $\mathcal M$ has boundary, $\partial\mathcal{F}$ is a pair of graphs $\Gamma_{\mathrm{in}},\Gamma_{\mathrm{out}}$, the spin networks between which the foam interpolates.

3.2 Face, edge, and vertex labels

The invariant space is finite-dimensional for $\mathrm{SU}(2)$; for $\mathrm{SL}(2,\mathbb{C})$ it is infinite-dimensional, and the models handle this by an embedding (the $Y_{\gamma}$ map of 4.3) that lands the $\mathrm{SU}(2)$ intertwiner inside an $\mathrm{SL}(2,\mathbb{C})$ one. On the boundary, the face and edge labels restrict to the edge and vertex labels of the boundary spin networks $\Gamma_{\mathrm{in}},\Gamma_{\mathrm{out}}$: a boundary face is dual to a boundary spin-network edge, and a boundary edge is dual to a boundary node.

3.3 The state sum

The face amplitude is most often $A_f(j_f)=\mathop{\mathrm{dim}}V_{j_f}=2j_f+1$ (the $BF$ choice), though the correct power of the dimension is itself a model choice that affects the divergence structure and the semiclassical limit. The vertex amplitude $A_v$ is the crucial dynamical input; the entire content of a spin-foam model is the choice of $A_v$, which is fixed by imposing a discretized simplicity constraint on a topological $BF$ theory, as we now describe.

5 is deliberately modest. It says the amplitude is a well-defined number once a complex and a cutoff are fixed. It says nothing about removing the cutoff (the sum over bulk spins may diverge; see 7) or about refining the complex. Those are exactly the open questions, and separating the settled part from them is the discipline the status calculus enforces.

3.4 Gluing and the composition law

6 is the structural reason spin foams sit in the cobordism/TQFT column of the dictionary. A spin foam is a morphism $\Gamma_{\mathrm{in}}\to \Gamma_{\mathrm{out}}$; composition is gluing; the identity is a trivial cylinder. The qualifier “like” matters: a genuine TQFT would assign a finite-dimensional space to each boundary and be diffeomorphism (indeed triangulation) invariant, and the gravitational models are not triangulation invariant — if they were, they would be topological and contain no local degrees of freedom, hence no gravitons. The tension between wanting a triangulation invariant (for a background-independent continuum limit) and wanting local curvature degrees of freedom is the conceptual core of the refinement problem (7).

4 From $BF$ theory to gravity: simplicity constraints

4.1 $BF$ theory and its state sum

The starting point is a topological field theory. Let $G$ be a Lie group with Lie algebra $\mathfrak g$. $BF$ theory has fields a connection $A$ (with curvature $F(A)$) and a $\mathfrak g$-valued two-form $B$, and action $$\begin{equation} S_{BF}[A,B]=\int_{\mathcal M}\mathop{\mathrm{tr}}\big(B\wedge F(A)\big). \end{equation}$$ The equations of motion are $F(A)=0$ (flat connection) and $\mathrm{d}_A B=0$; the theory has no local degrees of freedom. Discretizing on the two-complex $\mathcal{F}$, one assigns a group element (holonomy) $g_e\in G$ to each edge and integrates the constraint $F(A)=0$ face by face. Expanding the group delta function on the holonomy $g_f=\prod_{e\subset\partial f} g_e$ around each face by the Peter–Weyl theorem, $$\begin{equation} \delta_G(g_f)=\sum_{j_f}\mathop{\mathrm{dim}}(V_{j_f})\,\chi_{j_f}(g_f), \end{equation}$$ and performing the edge integrals, one obtains a state sum in which each face carries an irreducible $j_f$, each edge carries an intertwiner produced by the Haar projection, and each vertex carries the recoupling invariant contracting the intertwiners on its edges. For $G=\mathrm{SU}(2)$ on a four-dimensional complex the vertex invariant is the $\{15j\}$ symbol, and the partition function is the Ooguri model, $$\begin{equation} \mathcal{Z}_{BF}(\mathcal{F})=\sum_{\{j_f\}}\;\prod_{f}\mathop{\mathrm{dim}}(V_{j_f})\;\prod_{v}\{15j\}(j_{f\ni v}). \end{equation}$$

7 is the sharpest rigorous statement in the vicinity: a state sum of representation labels reconstructs a topological invariant of the manifold, exactly. It is also a warning. A theory that is fully triangulation invariant is topological — it has no local degrees of freedom and cannot be gravity. Gravity must break the topological invariance in a controlled way, keeping local curvature while retaining enough structure to have a continuum limit. That controlled breaking is the job of the simplicity constraint.

4.2 Simplicity constraints

General relativity in the Plebanski (or Palatini–Cartan) formulation is $BF$ theory for $G=\mathrm{SO}(4)$ (Riemannian) or $\mathrm{SO}(3,1)$ (Lorentzian) supplemented by a constraint forcing the two-form $B$ to come from a tetrad: $B^{IJ}=\pm(e\wedge e)^{IJ}$ or its dual. This simplicity constraint is what turns the topological $BF$ theory into gravity with local degrees of freedom. In terms of the bivector $B^{IJ}_f$ attached to a face (a discretized triangle), simplicity requires $B_f$ to be a simple bivector, one of the form $u\wedge w$ for vectors $u,w$.

For $\mathrm{Spin}(4)=\mathrm{SU}(2)\times\mathrm{SU}(2)$, the universal (double) cover of $\mathrm{SO}(4)$, a bivector splits into self-dual and anti-self-dual parts, carrying spins $(j^+_f,j^-_f)$; the two $\mathrm{SU}(2)$ factors are exactly the self-dual and anti-self-dual sectors of the covered rotation group. Discrete simplicity, in its strong (Barrett–Crane) form, sets $$\begin{equation} j^+_f=j^-_f, \end{equation}$$ the balanced or simple representations. Geometrically, [eq:bcsimplicity] says the self-dual and anti-self-dual areas agree, which is the statement that the bivector is simple. The Barrett–Crane model imposes [eq:bcsimplicity] strongly, as an operator equation annihilating states, and builds the vertex from the $\mathrm{SO}(4)$-invariant contraction of balanced representations.

4.3 The Engle–Pereira–Rovelli–Livine vertex and the $Y_{\gamma}$ map

The Engle–Pereira–Rovelli–Livine (EPRL) model imposes simplicity weakly, in the sense of the expectation value of a linear constraint, and in doing so incorporates the Barbero–Immirzi parameter $\gamma$. The construction rests on a map embedding $\mathrm{SU}(2)$ representations into $\mathrm{SL}(2,\mathbb{C})$ (Lorentzian) or $\mathrm{Spin}(4)$ (Riemannian) ones.

In the Lorentzian case the relevant $\mathrm{SL}(2,\mathbb{C})$ representations are the unitary principal series $\mathcal H_{(\rho,k)}$, labeled by $\rho\in\mathbb R$ and $k\in\tfrac12\mathbb Z$. Each such representation decomposes under the $\mathrm{SU}(2)$ subgroup as $\mathcal H_{(\rho,k)}=\bigoplus_{j\ge k} V_j$. The linear simplicity constraint, imposed weakly, selects the lowest $\mathrm{SU}(2)$ component and ties the labels through $\gamma$: $$\begin{equation} \rho_f=\gamma\, j_f,\qquad k_f=j_f . \end{equation}$$

The vertex [eq:eprlvertex] is the recoupling invariant of the $K_5$ pattern of 1: five edges, ten faces, contracted along the complete graph. In the Riemannian case it reduces to a $\gamma$-weighted $\{15j\}$ symbol. The Freidel–Krasnov model reaches an amplitude of the same form through a coherent-state (rather than projector) route; for $\gamma<1$ the two agree, and we treat them together as the weak-simplicity vertex.

5 Semiclassical asymptotics

The strongest evidence that the EPRL vertex is a gravity amplitude is its large-spin limit. Uniformly rescale all ten spins $j_f\mapsto\lambda j_f$ and let $\lambda\to\infty$.

Two honest caveats accompany 12, both [H]. First, the two terms of the cosine are the two parity-related geometric reconstructions of the four-simplex from the boundary areas and normals: in the Lorentzian model these are the two time orientations, so the amplitude does not distinguish a geometry from its time-reverse, and extracting a causal, single-exponential propagator requires additional input. The Riemannian model is more delicate still: its stationary-phase locus contains, besides the two geometric (Regge) critical points, additional degenerate “vector geometry” configurations that contribute non-Regge terms unless the boundary coherent state is chosen to suppress them, so [eq:asymp] is the leading behavior only on suitably peaked geometric boundary data. Second, and more consequentially, [eq:asymp] is the asymptotics of a single vertex on fixed boundary data. It does not by itself say what happens when one sums over the bulk spins of a many-vertex complex, and it is exactly there that the flatness problem appears (7.2).

6 Worked example: the four-simplex amplitude

We make the smallest nontrivial spin foam explicit: the two-complex dual to a single four-simplex, which has one bulk vertex, five edges (dual to the five boundary tetrahedra), and ten faces (dual to the ten boundary triangles). Its boundary is the spin network on the pentagram graph $K_5$: five nodes (tetrahedra), ten links (triangles).

6.1 Boundary data and the amplitude

Label the ten faces by spins $j_{ab}=j_{ba}$ for $1\le a<b\le5$ (the triangle shared by tetrahedra $a$ and $b$), and the five edges by intertwiners $\iota_a\in\mathrm{Inv}\big(\bigotimes_{b\ne a}V_{j_{ab}}\big)$, each a four-valent intertwiner. The vertex amplitude [eq:eprlvertex] contracts the five intertwiners along $K_5$. In the Riemannian, $\gamma\to 0$ or topological limit it degenerates to the $\{15j\}$ symbol $$\begin{equation} A_v(j_{ab},\iota_a)=\{15j\}\big(\{j_{ab}\};\{\iota_a\}\big), \end{equation}$$ the unique (up to normalization) $\mathrm{SU}(2)$-invariant contraction of five four-valent intertwiners whose external legs are the ten $j_{ab}$. The full state sum for this one-vertex complex with fixed boundary is just the single vertex weighted by its face and edge amplitudes: $$\begin{equation} \mathcal{Z}(\mathcal{F}_{\Delta_4})[b] =\prod_{a<b}A_f(j_{ab})\;\prod_{a}A_e(\iota_a)\;A_v(j_{ab},\iota_a). \end{equation}$$ There is no bulk sum for a single four-simplex with fully specified boundary: every face and edge is a boundary object. The sum over bulk labels first appears when two or more four-simplices are glued and share an internal triangle or tetrahedron.

6.2 Equilateral boundary and the asymptotic angle

Take the symmetric boundary $j_{ab}=j$ for all ten faces, the quantum analog of an equilateral four-simplex. By 12, as $j\to\infty$, $$\begin{equation} A_v(\lambda j)\;\sim\;\frac{1}{\lambda^{12}}\,2\,|N|\, \cos\!\Big(\lambda\,S_{\mathrm{Regge}}(j)+\varphi\Big),\qquad S_{\mathrm{Regge}}(j)=10\,\gamma\, j\,\Theta, \end{equation}$$ where $\Theta=\arccos(1/4)$ is the interior dihedral angle of a regular Euclidean four-simplex (the angle between two of its tetrahedral faces), and the total is ten triangles times the area $\gamma j$ times that dihedral angle. The amplitude oscillates in $j$ with a frequency set by the Regge action; the appearance of the regular four-simplex dihedral angle is the sense in which “the representation labels know the geometry.”

6.3 A solvable topological shadow of the refinement invariant

The four-simplex amplitude [eq:15j] is not something we can sum in closed form over refinements; that is the open problem. To exhibit the refinement invariant exactly, we drop to the solvable case: a finite-group $BF$ (Dijkgraaf–Witten) state sum, the abelian, discrete-group avatar of the topological theory of 4.1. Let $G$ be a finite abelian group. On a closed oriented surface $\Sigma$ triangulated with $V$ vertices, $E$ edges, and $F$ triangles, assign a group element $h_e\in G$ to each edge and define $$\begin{equation} \mathcal{Z}_G(\Sigma)=\frac{1}{|G|^{V}} \sum_{\{h_e\}}\;\prod_{f}\big[\textstyle\sum_{e\subset\partial f}\pm h_e=0\big], \end{equation}$$ where $[\cdot]$ is $1$ if the oriented holonomy around the face is trivial and $0$ otherwise, and the normalization $|G|^{-V}$ divides out the gauge redundancy $h_e\mapsto h_e+(\phi_{t(e)}-\phi_{s(e)})$. The following is a clean, fully [S] statement.

13 is the exactly-solvable shadow of what the gravitational state sum is being asked to do: produce an amplitude that stabilizes under refinement. In the topological case it stabilizes perfectly, because the theory has no local degrees of freedom. Gravity must stabilize while carrying local curvature, and whether it can is open. The companion Haskell code (9, 13) computes [eq:dw] by brute force for $G=\mathbb Z_n$ on several triangulations and checks [eq:dwvalue] and refinement invariance, alongside the $\mathrm{SU}(2)$ recoupling bookkeeping that supports [eq:15j].

7 Refinement, the continuum limit, and the flatness problem

7.1 Refinement and cylindrical consistency

A single two-complex is a truncation of the degrees of freedom, like a single lattice in lattice gauge theory. Because gravity is background independent, there is no fixed lattice spacing to send to zero; refinement means subdividing the complex, and the continuum theory (if it exists) is a limit over the directed set of refinements.

7.2 The flatness problem

The most concrete obstruction to a naive continuum limit is the flatness (or “cosine,” or “accidental curvature”) problem. Consider a complex with internal faces and take the large-spin limit while summing over the bulk spins. On coherent boundary data each vertex contributes $e^{\pm i S^{(v)}}$ by 12, and the bulk amplitude becomes a stationary-phase integral over the group elements $g_e$ on the internal edges (together with the bulk spin sums). The critical-point equations of these group integrals are the geometric content of the flatness problem: at each internal face $f$ they force the product of edge holonomies around $f$ to be trivial, $$\begin{equation} H_f=\prod_{e\subset\partial f} g_e=\mathbf{1}, \end{equation}$$ and a trivial holonomy around every internal hinge is exactly the statement that the reconstructed geometry has vanishing deficit angle there, $\varepsilon_f=2\pi-\sum_{v\supset f}\theta_f^{(v)}=0$, with $\theta_f^{(v)}$ the interior dihedral angle contributed by the simplex at vertex $v$. That is, the dominant configurations in the summed large-spin limit are flat. It is the group-integral saddle [eq:flatness], not the variation of the phase with respect to the areas $A_f=\gamma j_f$, that carries the obstruction: for a generic Barbero–Immirzi parameter the naive area-variation condition holds only modulo $2\pi$ and does not by itself force $\varepsilon_f=0$, whereas the holonomy condition does. Curved geometries, which carry nonzero deficit angles and are exactly the physical content of general relativity, are therefore suppressed. This was flagged by Bonzom and given a partition-function-level (wavefront-set) derivation by Hellmann and Kaminski, and confirmed numerically in EPRL and $BF$ cases by Donà and collaborators.

7.3 Effective spin foams

A recent reframing (Asante, Dittrich, Haggard) confronts the flatness problem by building an effective spin foam directly from the discrete gravity (Area–Regge) action, taking the quantum area spectrum as fundamental and imposing the gluing (shape-matching) constraints as strongly as the uncertainty relations of the area variables allow. In these models curved Regge geometries are not suppressed; the flat dominance of [eq:flatness] is traced to over-imposing constraints that the area uncertainties forbid imposing sharply. This is real progress on the flatness problem, and it clarifies which choices in 11 drive it. It is not a resolution of the continuum-limit problem: effective spin foams are still defined on a fixed complex, and the limit over refinements remains open. We tag the effective-spin-foam resolution of flatness [H] and the continuum limit itself open.

7.4 Divergences and the sum over complexes

Two further honest points. First, at fixed complex the sum over bulk spins can diverge: internal “bubbles” (closed surfaces in the complex) produce factors resembling $\sum_j (\mathop{\mathrm{dim}}V_j)^{p}$ that diverge for the natural face amplitudes, the analog of loop divergences in lattice gauge theory. The quantum-group truncation of 7 regulates the topological case; for gravity the divergence structure depends on the face amplitude and is part of what a renormalization treatment must control. Second, summing over two-complexes (rather than refining a fixed one) is the group-field-theory strategy; its convergence and its independence of the chosen family of complexes are not established. We tag the sum over complexes [H/P]: heuristic where a concrete group field theory is specified, speculative as a claimed definition of the full nonperturbative theory.

8 Relation to the other regimes and the S/H/P calculus

The library treats twelve quantum-gravity regimes as a modular site, composed by the weakest-link calculus of 1 rather than unified. We locate spin foams among them.

8.1 Spin networks as boundary data (QG-III)

The most direct relation is internal to module QG-III. The boundary Hilbert spaces $\mathcal{H}_{\Gamma}$ on which spin-foam amplitudes act are the spin-network states of the companion regime: graphs with $\mathrm{SU}(2)$ edge labels and intertwiner vertex labels. The kinematics is settled (the kinematical Hilbert space and area/volume spectra are [S]; the representation is unique by the Lewandowski–Okołów–Sahlmann–Thiemann theorem). The dynamics — the spin-foam amplitude between these states — is the bare-[H] link. The composite “spin-network kinematics $\ast$ spin-foam dynamics” is therefore $$\textsf{[S/H]}\;\ast\;\textsf{[H]}\;=\;\textsf{[H]},$$ the precise statement that a complete covariant quantum-gravity dynamics inherits the heuristic status of its dynamical half. This is not pessimism; it is bookkeeping. The kinematics really is settled, and the calculus records exactly that the settled part does not settle the whole.

8.2 Classical geometry as the target (QG-I)

Module QG-I — Lorentzian manifolds and the moduli of Einstein solutions — is the target that the missing continuum limit must reproduce. The one rigorous rung of the bridge is the semiclassical asymptotics of 12: a single vertex, on Regge boundary data, reproduces $e^{\pm iS_{\mathrm{Regge}}}$, the discrete Einstein–Hilbert action. Extending this from one vertex on fixed data to a refined complex with summed bulk, and thence to smooth solutions, is exactly the continuum-limit problem, obstructed by flatness. The composite $$\text{(classical geometry, }\textsf{[S/H]})\;\ast\;\text{(spin foam, }\textsf{[H]})=\textsf{[H]},$$ and it is optimistic even at [H]: no theorem currently establishes that QG-I is the continuum limit of the state sum.

8.3 Asymptotic safety and renormalization (QG-VII)

The coarse-graining program for spin foams (15) is a renormalization group flow on amplitudes, and it is natural to ask whether a fixed point of that flow could coincide with the non-Gaussian fixed point of asymptotic safety (module QG-VII). Both are bare-[H] entries, so any composite is bounded by [H]; but here the weakest-link bound is an upper bound only. No rigorous bridge exists between a combinatorial refinement fixed point and a continuum theory-space fixed point, so the [H] of the composite should be read as “neither endpoint exceeds heuristic, hence no chain built from them can,” not as “the bridge is established.” The recent tensor-network and Monte Carlo methods make the refinement fixed-point question computationally accessible, which is where a genuine comparison might eventually be made.

8.4 Gauge, $BF$, and cobordism (QG-II, TQFT)

The route from $BF$ theory to gravity (4) ties this regime to the Palatini–Cartan and Plebanski formulations of classical gravity (module QG-I) and, through the constraint structure, to the gauge/BV–BRST regime (module QG-II): simplicity is a second-class constraint whose proper treatment is a homological question. The gluing law (6) places spin foams in the cobordism/TQFT column: a spin foam is a morphism between spin networks. The composite with the ([S/H]) cobordism entry stays [S/H] for the topological $BF$ models and degrades to [H] for the gravitational ones, whose triangulation dependence breaks the strict TQFT axioms.

8.5 Explicit non-identifications

To keep the modularity honest we record what spin foams are not. A spin foam is not a spacetime: it is a combinatorial object whose relation to a smooth four-manifold is, beyond the topological/asymptotic level, exactly the open problem. The state sum is not a proven path integral for quantum gravity: it is a proposal whose vertex is chosen, not derived. And the sum over refinements is not known to converge: cylindrical consistency fails generically, and the continuum theory is a program, not a theorem. The weakest-link calculus exists to prevent these non-identifications from being quietly elided.

9 A Haskell model of the state sum

To make the reconstruction concrete and checkable we accompany the paper with a small Haskell development, split into base-only modules (association lists, no external containers) so that it compiles with a bare ghc. It models the two layers of the paper that are exactly computable: the $\mathrm{SU}(2)$ recoupling bookkeeping underlying the vertex amplitude, and the finite-group $BF$ (Dijkgraaf–Witten) state sum that exhibits the refinement invariant exactly.

The recoupling layer (Core) implements $\mathrm{SU}(2)$ spins as half-integers, irreducible dimensions $2j+1$, the triangle admissibility condition, Clebsch–Gordan fusion, and the dimension of an intertwiner space $\mathrm{Inv}(\bigotimes_i V_{j_i})$ as the multiplicity of the trivial representation, by two independent routes (iterated fusion and, for four legs, the recoupling count). A checkable property asserts the two routes agree, and that a four-simplex boundary labeled by admissible spins has the expected nonzero intertwiner support — the data on which the $\{15j\}$ vertex [eq:15j] is defined.

The state-sum layer (StateSum, Complexes) implements the amplitude [eq:dw] for $G=\mathbb Z_n$ by brute-force enumeration of edge colorings with the face-holonomy constraints, on three explicit triangulations: the boundary of a tetrahedron and an octahedron (two triangulations of $S^2$) and a minimal two-triangle torus. A checkable property asserts 13: that both sphere triangulations give $\mathcal{Z}=|G|^{-1}$ (refinement invariance) and that the torus gives $\mathcal{Z}=|G|$, matching $|G|^{1-\chi}$. This is the exact, solvable version of the paper’s core invariant — the refinement behavior of a state-sum amplitude — in the one case where it is fully under control.

The Main module runs the demonstrations and the property suite and exits with a nonzero status if any property fails. The core module is reproduced in 13; the full sources (Core, StateSum, Complexes, Properties, Main) accompany the submission as ancillary files.

10 Limitations and open problems

We collect the honest status.

10.0.0.1 Continuum/refinement limit (open, [H]).

The central open problem: whether the family $\{\mathcal{Z}(\mathcal{F})\}$ over refinements is cylindrically consistent, or admits a renormalization-group fixed point, defining a continuum amplitude that reproduces general relativity. Coarse-graining/tensor-network and group-field-theory programs exist but have not closed this. No theorem asserts that module QG-I is the continuum limit of the state sum.

10.0.0.2 The flatness problem (open, [H]).

The naive summed large-spin limit is dominated by flat configurations [eq:flatness]. Effective spin foams reframe the amplitude so that curved configurations survive, which is progress on this specific obstruction, but not a resolution of the continuum limit.

10.0.0.3 Choice of vertex and measure ([H]).

The EPRL/FK vertex is fixed by a chosen (weak) imposition of simplicity, a chosen label identification [eq:eprllabels], and a chosen face amplitude. Different choices give different models; there is no derivation from first principles selecting one.

10.0.0.4 Divergences ([H]).

At fixed complex the bulk sum can diverge through internal bubbles; the divergence structure depends on the face amplitude and is part of the renormalization problem.

10.0.0.5 Lorentzian signature and causality ([H]).

The cosine in [eq:asymp] contains both time orientations; extracting a causal, single-exponential propagator, and controlling the noncompact $\mathrm{SL}(2,\mathbb{C})$ integrals, are technical and conceptual issues beyond the scope of the settled asymptotics.

10.0.0.6 Sum over topologies ([H/P]).

Whether and how to sum over the topology of the underlying manifold, and whether the group-field-theory generating function defines the nonperturbative theory, is speculative.

11 Discussion

The reconstruction thesis is sharp in this regime, and so is its limit. What is reconstructed, rigorously, is a family of amplitudes: a colored two-complex, through its state sum, defines a well-defined map between spin-network boundary Hilbert spaces (5), the map composes by gluing (6), the topological version is an exact manifold invariant (7), and a single gravitational vertex reproduces the discrete Einstein–Hilbert action in its large-spin asymptotics (12). At this level the slogan “a spacetime history is a pattern of representation labels on a two-complex, and its amplitude is a product of recoupling invariants” is not a metaphor: there is no metric in the data.

What the thesis has not earned, in this regime, is the continuum. A smooth curved spacetime with its Einstein dynamics is supposed to emerge as the two-complex is refined, and it is precisely there that the construction is unfinished: cylindrical consistency fails generically, the naive large-spin limit is flat, and the renormalization and group-field-theory programs that would fix this are open. The value of the status calculus is that it holds these two facts apart. The settled part is genuinely settled, and the calculus forbids it from lending its credibility to the unsettled continuum limit: a chain is as strong as its weakest warrant, and the weakest warrant here is bare [H].

Seen this way, spin foams are the covariant-dynamics counterpart of the spin-network kinematics: the kinematics gives the boundary states, the foams give the amplitudes between them, and together they would be a background-independent path integral for gravity — if the continuum limit existed. The neighboring regimes supply the rest of the picture: classical Lorentzian geometry is the target, asymptotic safety offers a continuum fixed point to compare against, and the $BF$/cobordism structure supplies the categorical grammar in which a history is a morphism.

12 Conclusion

We have presented covariant quantum geometry as a state sum on a colored two-complex and organized the regime around one invariant: the boundary amplitude and its refinement behavior. The proved statements — well-definedness of the truncated amplitude (5), the topological state-sum invariant (7), the gluing/composition law (6), the exact refinement invariance of the solvable finite-group model (13), and the Regge asymptotics of the EPRL four-simplex vertex (12) — are [S]: standard mathematics in well-defined settings. The physical readings — that the EPRL/FK vertex is the correct gravitational dynamics, and above all that the family of amplitudes has a continuum limit reproducing general relativity — are [H], and the continuum limit is openly unresolved, obstructed at the simplest level by the flatness problem.

The regime composes modularly with its neighbors through the weakest-link calculus, which bounds every composite by its bare-[H] spin-foam link, and the accompanying Haskell model turns the exactly-solvable part of the invariant — refinement invariance of a topological state sum — into a machine-checkable property. The honest headline is narrow and firm: a spin foam is a well-defined amplitude between representation-theoretic boundary states, and its single-vertex semiclassical limit is the Regge action; both are theorems. The broad headline — that this is a continuum theory of quantum gravity — remains a hypothesis, and we have been careful to say so.

13 The core Haskell module

For reproducibility we reproduce the recoupling core, which models $\mathrm{SU}(2)$ irreducibles, Clebsch–Gordan fusion, and intertwiner-space dimensions by two independent routes. It uses only base (association lists), compiles with a bare ghc, and underlies the checkable properties of 9. The remaining modules (StateSum, Complexes, Properties, Main) accompany the submission as ancillary files.

{-# LANGUAGE ScopedTypeVariables #-}
-- Module Core: SU(2) recoupling bookkeeping for spin-foam vertices.
-- All content here is the [S] (standard mathematical) layer.
module Core
  ( Spin, spin, halfInt, dimIrrep, casimir, triangle
  , fuse, fuseMany, intertwinerDim, intertwinerDim4
  ) where

-- A spin j is a nonnegative half-integer stored as twice its value (an Int),
-- so j = 1/2 is HalfInt 1, j = 1 is HalfInt 2. Base-only, no Data.Ratio needed.
newtype Spin = Spin Int deriving (Eq, Ord)   -- Spin k represents j = k/2

instance Show Spin where
  show (Spin k)
    | even k    = show (k `div` 2)
    | otherwise = show k ++ "/2"

-- Smart constructor from twice-the-spin (rejects negatives).
halfInt :: Int -> Spin
halfInt k | k < 0     = error ("halfInt: negative " ++ show k)
          | otherwise = Spin k

-- Convenience: integer spin.
spin :: Int -> Spin
spin j = halfInt (2 * j)

-- Dimension of the carrier V_j: 2j + 1 = k + 1.
dimIrrep :: Spin -> Int
dimIrrep (Spin k) = k + 1

-- Casimir j(j+1), returned as twice-numerator over 4 to stay in Int-space:
-- j(j+1) = (k/2)(k/2 + 1) = k(k+2)/4.
casimir :: Spin -> Rational'
casimir (Spin k) = R (k * (k + 2)) 4

-- A tiny rational to avoid Data.Ratio (containers/ratio kept base-simple).
data Rational' = R Int Int
instance Show Rational' where show (R a b) = show a ++ "/" ++ show b

-- Triangle admissibility for j1 (x) j2 -> j3: |j1-j2| <= j3 <= j1+j2 and
-- j1 + j2 + j3 integral (i.e. k1 + k2 + k3 even).
triangle :: Spin -> Spin -> Spin -> Bool
triangle (Spin a) (Spin b) (Spin c) =
  abs (a - b) <= c && c <= a + b && even (a + b + c)

-- Clebsch-Gordan fusion V_a (x) V_b = sum_{k=|a-b|}^{a+b} V_k, step 1 in j
-- (step 2 in the twice-spin integer). Returned as (spin, multiplicity) alist.
fuse :: Spin -> Spin -> [(Spin, Int)]
fuse (Spin a) (Spin b) =
  [ (Spin k, 1) | k <- [abs (a - b), abs (a - b) + 2 .. a + b] ]

-- Iterated fusion of a list of irreducibles, accumulating multiplicities in
-- a base-only association list.
fuseMany :: [Spin] -> [(Spin, Int)]
fuseMany []       = [(spin 0, 1)]
fuseMany (x : xs) = foldl step [(x, 1)] xs
  where
    step acc j = collect
      [ (k, m * mult) | (a, m) <- acc, (k, mult) <- fuse a j ]

-- Sum multiplicities of equal keys in an association list.
collect :: [(Spin, Int)] -> [(Spin, Int)]
collect = foldr ins []
  where ins (k, m) [] = [(k, m)]
        ins (k, m) ((k', m') : rest)
          | k == k'   = (k', m + m') : rest
          | otherwise = (k', m') : ins (k, m) rest

-- Intertwiner-space dimension Inv(V_{j1} (x) ... (x) V_{jn}) = multiplicity
-- of the trivial rep (spin 0) in the full tensor product.
intertwinerDim :: [Spin] -> Int
intertwinerDim js =
  maybe 0 id (lookup (spin 0) (fuseMany js))

-- Independent four-valent count via the recoupling channel:
-- #{ k : max(|j1-j2|,|j3-j4|) <= k <= min(j1+j2,j3+j4), parities match }.
intertwinerDim4 :: Spin -> Spin -> Spin -> Spin -> Int
intertwinerDim4 (Spin a) (Spin b) (Spin c) (Spin d) =
  length
    [ () | k <- [lo, lo + 2 .. hi]
         , even (a + b + k), even (c + d + k) ]
  where lo = max (abs (a - b)) (abs (c - d))
        hi = min (a + b) (c + d)