gpt 6 astra pro attempt on Math-70

Cannon's Conjecture

Definitions and mathematical statement

A finitely generated group is Gromov-hyperbolic if a Cayley graph has uniformly thin geodesic triangles. Its boundary consists of asymptotic classes of geodesic rays, with the standard visual topology. The conjecture is \[ \partial G\cong S^2\quad\Longrightarrow\quad G\curvearrowright\mathbb H^3\text{ properly, cocompactly, and isometrically}. \] Properness allows finite stabilizers in the presence of torsion. Equivalently the group is virtually a cocompact Kleinian group in the source's convention. The hypothesis is about the boundary's topology, not merely its dimension.

Formulation and concept references: [S1].

Short English statement

Must a hyperbolic group with a two-sphere boundary come from a cocompact hyperbolic three-dimensional geometry?

Research attempt

Current frontier.

The analytic route asks whether a visual boundary metric is quasisymmetric to the round sphere. Bonk--Kleiner explain the relation to Cannon's geometric action conclusion and prove uniformization for Ahlfors $2$-regular, linearly locally contractible metric spheres. Topological dimension two is not Ahlfors $2$-regularity. Their graph-modulus criteria offer another route without assuming that regularity at the outset. [E1] Marković's alternative criterion uses enough quasiconvex surface subgroups to separate boundary point pairs; their existence is an extra requirement, not a consequence proved by the criterion. [E2] The geometric-measure reformulation of Cheetham-West--Nolte is likewise not used here as an automatic regularity theorem. [E3]

Attack route.

Transfer quadratic energies from controlled planar model graphs to finer boundary approximations. The aim is to replace "approximately round" by verifiable path lengths and congestion bounds. Hyperbolicity supplies multiscale structure, but it does not by itself provide the scale-uniform comparison established conditionally below. The alternative surface-subgroup route would require an actual construction of separating subgroups.

Attempt: a finite network-transfer lemma.

For a finite connected undirected graph $\mathcal G$ and nonempty disjoint terminal sets $A,B$, define \[ \operatorname{cap}_{\mathcal G}(A,B) =\min_{f|_A=0,\ f|_B=1}\sum_{\{v,w\}\in E(\mathcal G)}(f(v)-f(w))^2. \] Clipping to $[0,1]$ cannot increase energy; a minimum thus exists on a compact feasible set. Suppose $\varphi:V(\mathcal H)\to V(\mathcal G)$ takes terminals to corresponding terminals and represents every edge $e=\{v,w\}$ of $\mathcal H$ by a path $P_e$ joining $\varphi(v),\varphi(w)$. If paths have length at most $L$ and each edge of $\mathcal G$ belongs to at most $M$ paths, counting multiplicity, then \begin{equation}\label{wsref-0-33-30-30-30-30-30-33-30-3a-74-72-61-6e-73-66-65-72} \operatorname{cap}_{\mathcal H}\le LM\operatorname{cap}_{\mathcal G}. \end{equation} For any admissible $f$ on $\mathcal G$, Cauchy--Schwarz gives \[ |f(\varphi(v))-f(\varphi(w))|^2 \le |P_e|\sum_{a\in P_e}|df(a)|^2. \] Sum over $e$, use congestion, then minimize. A reverse map with parameters $L',M'$ gives \[ (LM)^{-1}\operatorname{cap}_{\mathcal H} \le\operatorname{cap}_{\mathcal G} \le L'M'\operatorname{cap}_{\mathcal H}. \] More precisely $LM$ can be replaced by $\max_a\sum_{e:a\in P_e}|P_e|$, with multiplicities retained. This weighted congestion may be smaller when few paths are long. No limiting argument is involved in this lemma.

An exactly soluble model.

Let $\mathcal R_{m,n}$ have vertices $(i,j)$, $0\le i\le m$, $1\le j\le n$, nearest-neighbor edges, and terminals $i=0,m$. Each of its $n$ horizontal rows satisfies \[ \sum_{i=0}^{m-1}(f(i+1,j)-f(i,j))^2\ge\frac1m. \] Hence capacity is at least $n/m$. The potential $f(i,j)=i/m$ attains this bound and has zero vertical energy. Thus \begin{equation}\label{wsref-0-33-30-30-30-30-30-33-30-3a-67-72-69-64} \operatorname{cap}_{\mathcal R_{m,n}}=n/m. \end{equation} All graph degrees are at most four, yet capacity can tend to zero or infinity. Planarity and bounded valence do not supply uniform conformal control.

The issue has a metric realization. On a coordinate square, let \[ d_\alpha((x,y),(x',y'))=|x-x'|+|y-y'|^\alpha, \qquad 0<\alpha<1. \] This has the usual topology; balls contract under coordinate contraction. A comparable-mesh grid at scale $\epsilon$ has horizontal spacing $\asymp\epsilon$ and vertical spacing $\asymp\epsilon^{1/\alpha}$. Thus $m\asymp\epsilon^{-1}$ and $n\asymp\epsilon^{-1/\alpha}$, giving \[ \operatorname{cap}_{\rm left,right}\asymp \epsilon^{1-1/\alpha}\longrightarrow\infty, \] while transverse capacity tends to zero. This illustrates the anisotropic local obstruction discussed in Bonk--Kleiner, not a constructed boundary of a hyperbolic group. [E1]

Attempt to close the multiscale step.

If boundary approximation graphs could be compared in both directions to round-sphere approximations as above, uniformly in scale and with controlled terminal continua and relative distances, they would inherit round capacity bounds. To invoke the actual uniformization criteria one must additionally compare this edge energy to the source's graph modulus and verify its approximation hypotheses. Bounded valence allows local energy comparisons; it does not supply terminal geometry or compatible boundary maps. The relevant criteria are in Sections 10--11 of the cited manuscript. [E1]

A naïve level-by-level construction is insufficient. Composing two path maps multiplies available length and congestion bounds: $L_{12}\le L_1L_2$, $M_{12}\le M_1M_2$. A cost $K>1$ per level can become $K^j$, not an absolute constant. An energy-aware construction would need to redistribute paths and bound weighted congestion directly across scales. No such routing is obtained from the group axioms here. Finite checks of many subdivisions do not establish its uniform bound at all scales.

Stress test.

Good topology and bounded graph degrees coexist with the divergent capacity example. Conversely a globally snowflaked round metric is still quasisymmetric to the round metric, so Hausdorff dimension greater than two alone is not a counterexample: the issue is the conformal gauge, not a raw dimension computation. A continuous surjective limit of approximate maps need not be a boundary homeomorphism; both directions require control. Surface-subgroup criteria cannot be used before constructing the separating subgroups. The exact finite-network computations in the companion script check the lemma's normalization but supply none of these group-theoretic inputs.

Outcome.

Outcome: Exploratory approach with unresolved gap. A quantitative energy-transfer lemma, exact capacity models and a multiscale routing obstruction have been established. The missing step is a uniform, geometrically compatible low-congestion construction for arbitrary sphere boundaries, with every hypothesis of the uniformization criterion checked. No new class of groups satisfying Cannon's conjecture or counterexample group is claimed.

Sources

• S1 Expanding Thurston Maps. https://www.ams.org/bookstore/pspdf/surv-225-prev.pdf. Catalogue formulation source.
• E1 M. Bonk and B. Kleiner, Quasisymmetric parametrizations of two-dimensional metric spheres, Inventiones Mathematicae 150 $2002$, 127--183; manuscript arXiv:math/0107171, Theorems 1.1--1.2, Conjecture 1.3, Sections 10--11. https://arxiv.org/pdf/math/0107171.
• E2 V. Marković, Criterion for Cannon's conjecture, Geometric and Functional Analysis 23 $2013$, 1035--1061; arXiv:1205.5747. https://arxiv.org/abs/1205.5747.
• E3 T. Cheetham-West and A. Nolte, Characterizing candidates for Cannon's conjecture from geometric measure theory, Bulletin of the London Mathematical Society 55 $2023$, 1718--1725; arXiv:2207.01720v2. https://arxiv.org/html/2207.01720v2.