Combinatorial Anabelian Geometry

• Combinatorial anabelian geometry is the study of profinite fundamental groups through a functorial and algorithmic framework to reconstruct arithmetic, topological, and moduli-theoretic information.
• It utilizes group-theoretic invariants from moduli spaces of curves and braid group structures to achieve a recursive reconstruction of objects like the Grothendieck–Teichmüller group and absolute Galois groups.
• The approach extends to positive characteristic, enabling the recovery of inertia subgroup structures and field data, thereby unifying methods in arithmetic geometry and topology.

Combinatorial anabelian geometry is the study of profinite fundamental groups and their automorphism groups in a form that is purely group-theoretic, functorial, and algorithmic, without further reference to algebraic curves. In the setting of Galois–Teichmüller theory, it arises from the arithmetic homotopy of moduli spaces of curves and recasts constructions surrounding braid groups, mapping class groups, and the Grothendieck–Teichmüller group $GT$ in terms of towers of profinite groups, forgetting morphisms, cuspidal inertia, and reconstruction procedures. In this perspective, $GT$ appears as an anabelian object, and the absolute Galois group $G_{\mathbf Q}$ is recovered by a purely topological-group-theoretic algorithm; in positive characteristic, related methods recover the topological type of pointed stable curves and the field structures attached to inertia subgroups from admissible fundamental groups (Collas, 3 Mar 2026, Yang, 2021).

1. Foundational profinite and semi-graph structures

A basic input is the profinite fundamental group of the moduli space of curves. For each pair $(g,m)$ with $2g-2+m>0$,

$\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$

the profinite completion of the orbifold mapping-class group. In genus $0$, the classical Artin presentation is

$$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$

and

$$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$

with $Z_m$ the center. The notation $GT$0, $GT$1, and $GT$2 denotes profinite completions.

The defining geometric-combinatorial object is the semi-graph of anabelioids of PSC-type. If $GT$3 is a nodal stable curve of type $GT$4 over a complete DVR $GT$5, its dual graph $GT$6 carries, at each vertex $GT$7, the Galois category of finite étale covers of the normalization $GT$8, and, at each edge, the category of covers of a formal disk. The associated fundamental group $GT$9 comes equipped with canonical verticial subgroups $G_{\mathbf Q}$0 and edge, or inertia, subgroups isomorphic to $G_{\mathbf Q}$1. In the formulation developed by Hoshi and Mochizuki, combinatorial anabelian geometry is precisely the study of such $G_{\mathbf Q}$2 and their automorphism groups purely in profinite group theory (Collas, 3 Mar 2026).

This formulation shifts the basic ontology of the subject. Stable curves and their degenerations remain the source of intuition, but the reconstruction problem is transferred to the internal structure of profinite groups, their subgroup patterns, and the compatibility of automorphisms with functorial morphisms.

2. Fiber-admissibility, cuspidal inertia, and forgetting maps

A central genus-zero tower is

$G_{\mathbf Q}$3

There are natural surjections $G_{\mathbf Q}$4 obtained by forgetting one label, each with kernel $G_{\mathbf Q}$5. These kernels encode the fiber structure of the configuration tower.

The corresponding automorphism groups are defined group-theoretically. The fiber-admissible outer automorphisms are

$G_{\mathbf Q}$6

and the cuspidally fiber-admissible subgroup is

$G_{\mathbf Q}$7

The key theorem is that for $G_{\mathbf Q}$8,

$G_{\mathbf Q}$9

and the natural surjections

$(g,m)$0

are injective; for $(g,m)$1 they are bijective. This result means that the cuspidal structure is not an additional external decoration. Rather, within the relevant range, fiber-admissibility already captures it.

The theorem also supplies the recursive mechanism used later in reconstruction arguments. The tower of forgetting maps becomes a rigid combinatorial scaffold: automorphisms are constrained by their interaction with the kernels $(g,m)$2, and the resulting rigidity replaces explicit braid calculations by functorial control of subgroup data (Collas, 3 Mar 2026).

3. The Grothendieck–Teichmüller group as an anabelian object

The classical braid-theoretic description of $(g,m)$3, due to Drinfeld and Ihara, is expressed in terms of pairs

$(g,m)$4

satisfying the relations

$(g,m)$5

$(g,m)$6

together with the $(g,m)$7-term relation coming from $(g,m)$8. In this presentation, $(g,m)$9 acts on each profinite braid group $2g-2+m>0$0 by

$2g-2+m>0$1

Combinatorial anabelian geometry recasts $2g-2+m>0$2 without explicit generators, braid presentations, or chosen homotopies. One shows that

$2g-2+m>0$3

equivalently,

$2g-2+m>0$4

The significance of this characterization is conceptual as well as technical. The classical Drinfeld–Ihara definition is recovered a posteriori from an anabelian avatar. The object $2g-2+m>0$5 is no longer introduced through explicit relations in $2g-2+m>0$6 and actions on braid generators; it is singled out by functorial preservation of the full graph of forgetting maps in the tower $2g-2+m>0$7. A recurrent misunderstanding is therefore corrected at the level of definition: in the combinatorial anabelian formulation, no choice of generators $2g-2+m>0$8 and no explicit homotopies are required (Collas, 3 Mar 2026).

4. Functorial reconstruction of $2g-2+m>0$9 and $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$0

The algorithmic reconstruction of $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$1 proceeds by replacing non-canonical choices with functorial algorithms that are continuous in the profinite topology and functorial under morphisms. The distinction between mono-anabelian and arithmetic-holomorphic objects is part of this framework: a mono-anabelian object has a one-dimensional profinite fundamental group, whereas an arithmetic-holomorphic object has a higher-dimensional fundamental group.

The first step is to recover discrete invariants from the lower central series $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$2 and its associated graded Lie algebra. For instance,

$\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$3

and

$\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$4

This determines $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$5 and the level $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$6.

The second step is the reconstruction of inertia, or cuspidal, subgroups. A purely group-theoretic search for maximal procyclic subgroups of weight $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$7, characterized through the action of the adjoint on $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$8, recovers the finite set of cuspidal subgroups, and these generate each $\Gamma_{g,[m]}=\pi_1^\et\bigl(M_{g,[m]}\times_{\mathbf Q}\bar{\mathbf Q}\bigr)\cong\widehat{\mathrm{MCG}}_{g,m},$9. In the proper case of a curve of strict-Belyi type, one first constructs the diagram of all open complements $0$0 and then applies the same inertia reconstruction.

The third step identifies fiber-admissible automorphisms with $0$1: $0$2 This is a purely functorial and cohomological characterization. The braid-theoretic presentation is not used in the reconstruction itself.

The fourth step constructs a combinatorial model of $0$3 and $0$4. One forms a cofiltered system of arithmetic Belyi diagrams, after Tsujimura, indexed by open normal subgroups $0$5. By tripod-synchronization, the subgroup of $0$6-cusps in each $0$7 reconstructs uniquely the same $0$8. Passing to the direct limit over all $0$9 yields the combinatorial Belyi–Galois–Teichmüller model

$$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$0

equipped with functorial operations $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$1 and involutions $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$2 and $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$3, making $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$4 into a field isomorphic to $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$5. Consequently,

$$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$6

This suggests that the arithmetic content of the tower $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$7 is sufficiently rich not merely to detect automorphism groups of configuration spaces, but to support a full combinatorial reconstruction of the absolute Galois group of $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$8 (Collas, 3 Mar 2026).

5. Positive-characteristic invariants and admissible fundamental groups

A parallel development in positive characteristic begins with a connected pointed stable curve $$B_m=\Bigl\langle \sigma_1,\dots,\sigma_{m-1}\,\Big|\, \sigma_i\sigma_{i+1}\sigma_i=\sigma_{i+1}\sigma_i\sigma_{i+1},\; [\sigma_i,\sigma_j]=1\,(|i-j|>1) \Bigr\rangle,$$9 of topological type $$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$0 over an algebraically closed field $$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$1 of characteristic $$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$2. A finite morphism $$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$3 is an admissible Galois cover if it is generically étale, tamely ramified over nodes and marked points in the sense of Deligne–Mumford, and satisfies the local tameness condition at nodes. The associated profinite group

$$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$4

classifies finite admissible covers of $$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$5.

For a prime-to-$$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$6 cyclic admissible cover with deck group $$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$7, one decomposes

$$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$8

under the action of the deck group and defines

$$\Gamma_{0,[m]}\cong B_m/\langle H_m,Z_m\rangle,\qquad H_m=(\sigma_1\sigma_2\cdots\sigma_{m-1})^m,$$9

The tuple $Z_m$0 is the system of generalized Hasse–Witt invariants. These invariants measure the $Z_m$1-rank of the cover in each isotypical summand.

For a component-generic pointed stable curve of type $Z_m$2, Yang proves that the first generalized Hasse–Witt invariant attains its maximum possible value. The sharp bound is

$Z_m$3

and Theorem 4.12 identifies conditions under which equality holds for every non-zero cyclic character $Z_m$4 of the appropriate type. The proof proceeds by induction along degenerations, using reductions to components of type $Z_m$5 or $Z_m$6, Nakajima–Raynaud ordinariness results, and, in the reducible case, a minimal quasi-tree in the dual graph.

These maximum values feed into an anabelian formula for the topological type. If $Z_m$7, then with

$Z_m$8

with $Z_m$9 defined by existence or nonexistence of a continuous surjection onto

$GT$00

and with

$GT$01

one has

$GT$02

Thus $GT$03 is recovered group-theoretically from $GT$04.

A second application is the reconstruction of field structures associated to inertia subgroups of marked points. For a marked-point inertia group

$GT$05

one recovers, up to conjugacy, the maximal finite subquotients $GT$06 for all $GT$07, hence the tower of finite fields $GT$08. More precisely, Theorem 6.4 shows that a surjective open homomorphism

$GT$09

carries inertia to inertia and induces a bijection

$GT$10

thereby recovering $GT$11 and the full field $GT$12 at each puncture (Yang, 2021).

6. Conceptual significance, moduli, applications, and difficulties

The main conceptual shift is from explicit braid-theoretic computation to functorial and algorithmic anabelian reconstruction. In place of a tangential base-point, explicit paths in $GT$13, manual verification of the $GT$14- and $GT$15-term relations, and cohomological lifting in individual $GT$16, combinatorial anabelian geometry uses fiber-admissibility, cuspidal admissibility, and recursive injectivity along forgetting maps. The advantages stated for this framework are universality, canonicity, and algorithmicity: it does not refer to any particular presentation of braid or mapping class groups; constructions commute with morphisms of towers $GT$17; and finite-quotient implementations are, in principle, possible (Collas, 3 Mar 2026).

The approach also supports broader anabelian applications. These include the absolute Grothendieck conjecture for hyperbolic curves and configuration spaces over number fields and $GT$18-adic fields, canonical group-theoretic descriptions of $GT$19 for $GT$20 local or global, and prospects for an intrinsic “Galois-Teichmüller stack” whose automorphism group is $GT$21 and whose rational points recover $GT$22. At the same time, the stated challenges are substantial: reconstruction of discrete invariants still relies on $GT$23-adic Lie algebra computations of lower central series, and extensions to mixed characteristic or to towers beyond genus $GT$24 require delicate anabelian input, including Mochizuki’s higher-genus work over $GT$25-adic local fields (Collas, 3 Mar 2026).

In positive characteristic, admissible fundamental groups lead to a topological moduli space $GT$26 consisting of isomorphism classes of groups abstractly isomorphic to $GT$27. Yang constructs a natural continuous surjection

$GT$28

conjectured to be a homeomorphism. The maximum-value theorem for generalized Hasse–Witt invariants and the reconstruction of inertia-field structures serve as the building blocks for distinguishing strata in $GT$29, recovering the combinatorial dual graph $GT$30 from $GT$31, and identifying connected components of $GT$32 with components of $GT$33. In the low-dimensional cases $GT$34, $GT$35, and $GT$36, $GT$37 is established to be a homeomorphism (Yang, 2021).

Taken together, these developments indicate that combinatorial anabelian geometry is not merely a reformulation of known braid-group constructions. It is a framework in which arithmetic, topological, and moduli-theoretic data are reconstructed from profinite groups, subgroup configurations, and functorial compatibility conditions, with $GT$38, $GT$39, and positive-characteristic moduli all appearing as outputs of the same general anabelian strategy.