11 Polyhedral degeneration and quantum volume asymptotics

11.1 Three different geometric quantities

The connection to the volume conjecture becomes precise only after separating three quantities: $$A=\operatorname{Area}(\partial P),\qquad W(P)=\operatorname{Vol}_{\mathbb H^3}(P),\qquad V_{\mathrm{WP}}=\int_{\mathcal M}\frac{\omega_{\mathrm{WP}}^d}{d!}.$$ The first is a surface area, the second an enclosed three-dimensional volume, and the third a measure of a space of shapes. The ordinary knot volume conjecture concerns three-dimensional volume. For a hyperbolic knot, in the reduced colored-Jones normalization $J_N(\text{unknot};q)=1$, its familiar form is $$\lim_{N\to\infty}\frac{2\pi}{N} \log\left|J_N(K;e^{2\pi i/N})\right| =\operatorname{Vol}(S^3\setminus K).$$(11.1) This remains a conjecture in general; see the colored-Jones formulation in [82] and the tetrahedral results in [52].

Suppose a family of compact convex polyhedra $P_A$ has hyperbolic boundary area $A\downarrow0$. Make the geometric convergence precise by choosing isometries into a Klein chart for which the rescaled convex bodies $A^{-1/2}P_A$ converge in Euclidean Hausdorff distance to a nondegenerate polyhedron $P_{\mathrm E}$ of surface area one. The rescaled bodies then lie in a common bounded set, the hyperbolic volume density tends uniformly to the Euclidean density, and Euclidean volume is continuous under this convex-body convergence. Hence $$W(P_A)=A^{3/2}W_{\mathrm E}(P_{\mathrm E})+o(A^{3/2}).$$(11.2) For the separate intrinsic path (8.1), with fixed $0<\mu_i<1$, $\sum_i\mu_i=2$, and $d=n-3$, the moduli volume satisfies $$V_{\mathrm{WP}}(A)=2^{-d}A^d V_{\mathrm{CH}}(\mu)+O(A^{d+1})$$ by (8.11) in its final chamber. The normalizations are those of Section 8. The different powers reflect different dimensions. Flattened Euclidean limits and changes of polyhedral combinatorics require additional care; they are not covered by the nondegeneracy hypothesis in (11.2).

11.2 An explicit small-polyhedron expansion

A controlled model gives more than the leading power. Let $Q$ be a bounded nondegenerate convex Euclidean polyhedron, and place $P_r=rQ$ in the curvature-$-1$ Klein ball. Its Euclidean planar faces are hyperbolic geodesic faces. The Klein metric is $$g_x=\frac{I}{1-|x|^2} +\frac{xx^{\mathsf T}}{(1-|x|^2)^2}.$$ Its two tangential eigenvalues are $(1-|x|^2)^{-1}$ and its radial eigenvalue is $(1-|x|^2)^{-2}$, so $$\mathop{}\!\mathrm d\operatorname{Vol}_{\mathbb H^3}=(1-|x|^2)^{-2}\,\mathop{}\!\mathrm d^3x.$$ Consequently, whenever $r\sup_Q|y|<1$, $$\begin{aligned} W(rQ) &=r^3\int_Q(1-r^2|y|^2)^{-2}\,\mathop{}\!\mathrm d^3y\\ &=\sum_{k=0}^{\infty}(k+1)r^{3+2k} \int_Q|y|^{2k}\,\mathop{}\!\mathrm d^3y\\ &=r^3 W_{\mathrm E}(Q) +2r^5\int_Q|y|^2\,\mathop{}\!\mathrm d^3y +3r^7\int_Q|y|^4\,\mathop{}\!\mathrm d^3y+\cdots. \end{aligned}$$ The series converges uniformly on $Q$ for each $rR<1$, where $R=\sup_Q|y|$. Its nonnegative remainder after $k=m$ is bounded by $$0\le W(rQ)-\sum_{k=0}^m(k+1)r^{3+2k} \int_Q|y|^{2k}\,\mathop{}\!\mathrm d^3y \le r^3W_{\mathrm E}(Q) \frac{t^{m+1}\{m+2-(m+1)t\}}{(1-t)^2},\qquad t=r^2R^2.$$ This follows by summing the scalar tail of $(1-t)^{-2}$. It bounds truncation of this geometric series, not a quantum asymptotic error. The expansion is exact for the specified homothetic Klein path. It is not generally the fixed-conformal angle path (8.1). Nor does it imply a Taylor expansion of this form at ideal, flattening, or otherwise uncontrolled degenerations. Unlike (8.10), it is an infinite series: the finite polynomial for the moduli volume comes from a quadratic cohomology class and a top intersection.

11.3 Schläfli, doubling, and degeneration/regeneration

For a compact hyperbolic polyhedron with interior dihedral angles $\alpha_e$ and edge lengths $\ell_e$, Schläfli’s formula is $$\mathop{}\!\mathrm dW=-\frac12\sum_e\ell_e\,\mathop{}\!\mathrm d\alpha_e.$$(11.4) For the Klein family above, write $$\ell_e(r)=rL_e+O(r^3),\qquad \alpha_e(r)=\alpha_e^{\mathrm E}+r^2b_e+O(r^4).$$ These parity expansions follow directly from the Klein model. For example, write two adjacent face planes of $Q$ as $n_f\cdot y=h_f$ and $n_g\cdot y=h_g$, with unit outward normals. Their unit Lorentzian normals at scale $r$ are $(rh_f,n_f)/\sqrt{1-r^2h_f^2}$ and $(rh_g,n_g)/\sqrt{1-r^2h_g^2}$, for signature $(-,+,+,+)$. Thus, with $c_e=\cos\alpha_e^{\mathrm E}=-n_f\cdot n_g$, $$\cos\alpha_e(r) =\frac{c_e+r^2h_fh_g} {\sqrt{(1-r^2h_f^2)(1-r^2h_g^2)}},\qquad b_e=-\frac{h_fh_g+\tfrac12c_e(h_f^2+h_g^2)} {\sin\alpha_e^{\mathrm E}}.$$ We take actual faces, so adjacent ones are not coplanar and $0<\alpha_e^{\mathrm E}<\pi$. The length expansion follows by integrating the Klein metric along the fixed Euclidean edge, which is a hyperbolic geodesic. The leading term of (11.3) and (11.4) imply $$\boxed{\quad\sum_eL_e b_e=-3W_{\mathrm E}(Q).\quad}$$(11.5) Indeed, $\mathop{}\!\mathrm dW/\mathop{}\!\mathrm dr=3r^2W_{\mathrm E}+O(r^4)$, while the right side of (11.4), divided by $\mathop{}\!\mathrm dr$, is $-r^2\sum_eL_e b_e+O(r^4)$. Thus the first curvature correction to the Euclidean dihedral angles already determines the Euclidean enclosed volume. The sign reverses if exterior dihedral angles are used. As a check, for $Q=[-1,1]^3$ every $L_e=2$ and $b_e=-1$, so the sum is $-24=-3\cdot8$. The same identity holds with $\ell_e=A^{1/2}L_e+O(A^{3/2})$ and $\alpha_e=\alpha_e^{\mathrm E}+A b_e+O(A^2)$ for a fixed-combinatorics unit-area limiting family whose expansions are differentiable in $\sqrt A$. It uses $W=A^{3/2}W_{\mathrm E}+O(A^{5/2})$ with the corresponding differentiated remainder; a bare $o(A^{3/2})$ estimate cannot be differentiated without control.

Doubling a convex polyhedron across its faces produces a cone three-manifold with singular graph given by its edges. The cone angle along an edge is $2\alpha_e$, and its volume is $2W(P)$. A shrinking polyhedron therefore supplies a direct example of degeneration to a Euclidean cone three-manifold after rescaling. This is the geometric setting suggested by Hodgson’s thesis, Degeneration and regeneration of geometric structures on 3-manifolds [62]. Later precise regeneration results include Porti–Weiss [86, Theorem 1.3], whose input is a closed orientable Euclidean cone three-manifold with cone angles at most $\pi$ which is not almost product. Those hypotheses are not automatic for a doubled convex polyhedron; in particular, $2\alpha_e^{\mathrm E}$ can exceed $\pi$. No unrestricted regeneration theorem is being asserted here. The edge cone angles $2\alpha_e$ are also distinct from the intrinsic vertex openings $\theta_i$ of the boundary surface.

Rivin’s Volumes of degenerating polyhedra—on a conjecture of J. W. Milnor is directly relevant [89]. It studies continuous extension of volume toward degenerate angle data, the Euclidean vanishing boundary, and quantitative control through Schläfli’s formula and polar geometry. Its estimates allow logarithmically growing volume derivatives at degenerations; continuity does not imply a smooth power series at every boundary point. The controlled small-polyhedron expansion above gives a particularly explicit local calculation on the Euclidean side of this broader degeneration picture.

11.4 The established tetrahedral quantum connection

Quantum $6j$-symbols provide a concrete bridge to three-dimensional geometry. Costantino’s hyperbolic-type theorem uses the evaluation $\operatorname{ev}_n([n]\{6j\})$ after removing a root-of-unity pole; its geometric value is the volume of the doubled truncated-tetrahedral block $D(\boldsymbol u)$ [52, Definitions 3.1–3.2 and Theorem 3.9]. It is not a formula for an unregularized value at that pole.

For clarity, distinguish two results of Chen–Murakami. Fix a hyperbolic tetrahedron $T$ with an ideal or ultraideal vertex and interior dihedral angles $\alpha_e$ in the convention of (11.4), and put $\xi_r=e^{2\pi i/r}$, $r$ odd. Their Theorem 1 gives $$\frac{2\pi}{r}\log|\{6j\}^{\mathrm{RW}}_{q=\xi_r^2}| \longrightarrow W(T)$$ for admissible half-integer labels with $j_{e,r}/r\to(\pi-\alpha_e)/(4\pi)$ and with the sum of the three labels at a chosen ideal or ultraideal vertex at least $(r-3)/2$. The prefactor requires more: Theorem 2 assumes that $\eta_e=(\pi-\alpha_e)/(2\pi)$ have odd numerators and odd denominators, takes $r$ through odd common multiples of those denominators, and uses exactly $j_{e,r}=r\eta_e/2-1/2$. Then $$|\{6j\}^{\mathrm{RW}}_{q=\xi_r^2}|\sim \sqrt2\pi\,r^{-3/2}(-\det G)^{-1/4} \exp\left(\frac{r}{2\pi}W(T)\right),$$(11.6) where $G_{ff}=1$ and $G_{fg}=-\cos\alpha_{fg}$ for the incident face dihedrals [51, Theorems 1–2, pp. 1–2]. The labels are in the source’s Racah–Wigner normalization, whose quantum integer uses $q^{1/2}=\xi_r$; hence $q=\xi_r^2$ here corresponds to $q_{\mathrm{alg}}=\xi_r$ in (10.2). Arbitrary convergent label sequences give the stated leading rate under Theorem 1, not automatically this refined equivalent.

At the Euclidean end, Roberts proves the Ponzano–Regge asymptotic for classical $6j$-symbols: Euclidean tetrahedral edge lengths and dihedral angles enter an oscillatory Regge phase, and Euclidean volume enters its amplitude [91]. This is an established appearance of Euclidean tetrahedral geometry in quantum asymptotics. It does not automatically identify the classical large-label limit with a chosen root-of-unity hyperbolic limit.

In particular, the stated Chen–Murakami theorem does not cover an arbitrary family of compact tetrahedra shrinking to Euclidean ones. At a Euclidean tetrahedron the angle Gram determinant vanishes, so the displayed fixed-shape prefactor itself warns of nonuniformity. A continuous classical volume limit, even with good quantitative estimates, does not by itself control the quantum Hessian, the prefactor, or the error term in a joint limit.

11.5 A proposed joint scale and a model integral

Suppose a suitable quantum object, in a specified geometric branch and choice of variables, has an exponential contribution $\exp(W(P_A)/\hbar)$. Equation (11.2) then singles out the dimensionless ratio $$\lambda=\frac{A^{3/2}}{\hbar},\qquad A\asymp\hbar^{2/3}.$$(11.7) This is the regime in which geometric degeneration competes with the semiclassical limit. At this stage it is a scaling prediction; Section 15 proves an additive uniform Airy statement for the pole-cleared regular $b$-$6j$ integral and fixed bounded real transition parameters. In that convention the volume-action parameter is $\hbar_{\mathrm{vol}}=\pi b^2$, whereas its fold parameter is $h=2\pi b^2$; see (14.24). No claim for fixed nonzero volume or unbounded transition parameter follows. A different polarization may involve a Regge or Legendre-transformed action, so the quantum object and phase must be specified first.

There is a useful classical model making the potential role of the WP–Thurston limit explicit. Fix the weights and intrinsic path (8.1). Let $K\subset\mathcal M_{0,n}$ be a compact measurable subset and assume the corresponding realizations obey $A^{-3/2}W(P_A(X))\to W_{\mathrm E}(P_{\mathrm E}(X))$ uniformly on $K$, with a continuous nondegenerate Euclidean limiting family. Define $$\mathcal Z_K(A,\hbar)=\int_K \exp\left(\frac{W(P_A(X))}{\hbar}\right) \frac{\omega_{\mathrm{WP},A}^d}{d!}.$$(11.8) This is a constructed geometric integral, not an identification with a colored-Jones or other known quantum invariant. Take positive $A,\hbar\to0$ with $A^{3/2}/\hbar\to\lambda_0\in(0,\infty)$. The exponential then converges uniformly on $K$. By (8.5), $2A^{-1}\omega_{\mathrm{WP},A}\to\omega_{\mathrm{CH},\mu}$ locally uniformly in form coefficients; taking its $d$th exterior power preserves this convergence. A finite coordinate cover of $K$ therefore gives a common integrable bound and proves $$2^d A^{-d}\mathcal Z_K(A,\hbar)\longrightarrow \int_K e^{\lambda_0 W_{\mathrm E}(P_{\mathrm E}(X))} \frac{\omega_{\mathrm{CH},\mu}^d}{d!}.$$(11.9) Extending this formula to the full noncompact space requires domination and control of collision and flattened strata. Identifying its measure and phase with those of an actual quantum state integral is a further mathematical problem.

This separates the two potentially compatible roles: enclosed three-dimensional volume supplies an action, while the WP form supplies a measure on boundary shapes. Their Euclidean limits are respectively Euclidean polyhedral volume and the Thurston–Deligne–Mostow measure. Establishing that both occur together in a particular quantum construction would provide the substantive connection to the volume conjecture.