OctoInformBeta

OpenAI preprint, September 2026 · §3. Comparison with an arbitrary aspherical presentation

Lemma 3.2: Normal generation and full boundary injectivity persist, with no cardinality bounds on the added alphabets

Stabilization aligns presentations with prescribed cycle boundaries.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

The comparison concerns a prescribed family of Cayley cycles and the relators of a presentation of a group Γ\Gamma, possibly on a different alphabet. Stabilization adjoins letters whose value in Γ\Gamma is the identity.

Put Λ=ℤ[Γ]\Lambda=\mathbb Z[\Gamma]. For an alphabet SS and a value map v:S→Γv:S\to\Gamma, word evaluation εv:F(S)→Γ\varepsilon_v:F(S)\to\Gamma is the homomorphism extending vv from the free group F(S)F(S).✦ The based integral chain group C(S)C(S) consists of finitely supported integral sums of edges (g,s)(g,s) from gg to gv(s)gv(s); letters with the same value remain distinct.✦ Write ese_s for the chain consisting of the single edge (1,s)(1,s). The edge boundary ∂v\partial_v of C(S)C(S) is the homomorphism to Λ\Lambda given in (1), and the cycle module Z1(v)=ker⁡∂vZ_1(v)=\ker\partial_v is its kernel, with Λ\Lambda acting on the left: C(S)=ℤ(Γ×S)=Λ(S),∂v(ges)=g(v(s)−1),Z1(v)=ker⁡∂v.C(S)=\mathbb Z^{(\Gamma\times S)}=\Lambda^{(S)},\qquad \partial_v(ge_s)=g(v(s)-1),\qquad Z_1(v)=\ker\partial_v.(1)

For edge labels ℓ(g,s)\ell(g,s) in a group KK, the semidirect label homomorphism Lℓ:F(S)→KΓ⋊ΓL_\ell:F(S)\to K^\Gamma\rtimes\Gamma extends s↦(ℓ(⋅,s),v(s))s\mapsto(\ell(\mathord\cdot,s),v(s)), using right translation on functions.✦ The path holonomy hℓ(w,g)h_\ell(w,g) of a word ww in F(S)F(S) at a vertex gg in Γ\Gamma is the element of KK obtained by evaluating the function component of Lℓ(w)L_\ell(w) at gg; its factors follow traversal order, and reverse traversal takes inverses.✦ The signed path chain [w]g∈C(S)[w]_g\in C(S) uses each edge as its own additive label, with reverse traversal contributing its negative; write [w]=[w]1[w]=[w]_1.✦

For a word family q:J→F(S)q:J\to F(S), the relator boundary Dv,qD_{v,q} is the integral linear map from C(J)C(J) to C(S)C(S) given in (2). A boundary basis B:C(J)→Z1(v)B:C(J)\to Z_1(v) for qq is an integral linear isomorphism whose composite with inclusion into C(S)C(S) is Dv,qD_{v,q}. A genuine presentation ⟨S∣qj(j∈J)⟩\langle S\mid q_j\ (j\in J)\rangle of Γ\Gamma is an alphabet and relator family for which word evaluation is surjective and (3) holds. A null word ww in F(S)F(S) is a word whose evaluation is 11. Dv,q:C(J)→C(S),Dv,q(gej)=[qj]g.D_{v,q}:C(J)\longrightarrow C(S),\qquad D_{v,q}(ge_j)=[q_j]_g.(2)✦ εv(qj)=1(j∈J),ker⁡εv=⟨⟨qj:j∈J⟩⟩.\varepsilon_v(q_j)=1\quad(j\in J),\qquad \ker\varepsilon_v=\langle\!\langle q_j:j\in J\rangle\!\rangle.(3)

Stabilized Chain Groups

Adjoin an alphabet TT, and let v⊔1Tv\sqcup1_T be the value map on S⊔TS\sqcup T extending vv and taking value 11 on TT. The chain splitting and cycle splitting in (4) separate the two alphabets. The extra edge projection πT\pi_T retains only TT-coordinates; the extra relator boundary Db,qTD^T_{b,q} is its composite with the full relator boundary for a value map b:S⊔T→Γb:S\sqcup T\to\Gamma. C(S⊔T)≅C(S)⊕C(T),Z1(v⊔1T)≅Z1(v)⊕C(T),πT(x,y)=y,Db,qT=πTDb,q.\begin{gathered} C(S\sqcup T)\cong C(S)\oplus C(T),\qquad Z_1(v\sqcup1_T)\cong Z_1(v)\oplus C(T),\\ \pi_T(x,y)=y,\qquad D^T_{b,q}=\pi_TD_{b,q}. \end{gathered}(4)

Let ZZ be an alphabet. Given an integral linear isomorphism B:C(Z)≅Z1(v)B:C(Z)\cong Z_1(v), its stabilized boundary basis UB:C(Z⊔T)≅Z1(v⊔1T)U_B:C(Z\sqcup T)\cong Z_1(v\sqcup1_T) is the map in (5). Let RR and II be alphabets. For an integral linear isomorphism M:C(R)≅C(I)M:C(R)\cong C(I), the ordinary block OM:C(T⊔R)≅C(T⊔I)O_M:C(T\sqcup R)\cong C(T\sqcup I) is defined there too. All coordinates use the splittings in (4); inclusions of words from F(S)F(S) into enlarged alphabets are denoted by ι\iota. UB(c,d)=(Bc,d),OM(x,y)=(x,−My).U_B(c,d)=(Bc,d),\qquad O_M(x,y)=(x,-My).(5)

The Claim

Theorem 1. For presentation basis alignment (6), let v:S→Γv:S\to\Gamma and a:A→Γa:A\to\Gamma have surjective word evaluations. Let z:Z→F(S)z:Z\to F(S) and r:R→F(A)r:R\to F(A) be null word families whose boundaries induce integral linear isomorphisms B:C(Z)≅Z1(v)B:C(Z)\cong Z_1(v) and E:C(R)≅Z1(a)E:C(R)\cong Z_1(a). Suppose rr normally generates ker⁡εa\ker\varepsilon_a. Put T0=A⊔(Z⊔A)T_0=A\sqcup(Z\sqcup A), R0=A⊔(R⊔S)R_0=A\sqcup(R\sqcup S) and b0=v⊔1T0b_0=v\sqcup1_{T_0}. There are a family q:Z⊔R0→F(S⊔T0)q:Z\sqcup R_0\to F(S\sqcup T_0) and an integral linear isomorphism Q:C(R0)≅C(T0)Q:C(R_0)\cong C(T_0) satisfying εb0(qj)=1(j∈Z⊔R0),ker⁡Db0,q=0,ker⁡εb0=⟨⟨qj:j∈Z⊔R0⟩⟩,[qi]=[ιzi](i∈Z),Db0,q|R0T0=Q.\begin{gathered} \varepsilon_{b_0}(q_j)=1\quad(j\in Z\sqcup R_0),\qquad \ker D_{b_0,q}=0,\\ \ker\varepsilon_{b_0}=\langle\!\langle q_j:j\in Z\sqcup R_0\rangle\!\rangle,\\ [q_i]=[\iota z_i]\quad(i\in Z),\qquad D^{T_0}_{b_0,q|_{R_0}}=Q. \end{gathered}(6)✦

For stabilized boundary alignment (7), let v:S→Γv:S\to\Gamma have surjective evaluation, and let r:R→F(S)r:R\to F(S) and z:I→F(S)z:I\to F(S) be null word families with boundary isomorphisms E:C(R)≅Z1(v)E:C(R)\cong Z_1(v) and B:C(I)≅Z1(v)B:C(I)\cong Z_1(v). Suppose rr normally generates ker⁡εv\ker\varepsilon_v. Put b1=v⊔1Ib_1=v\sqcup1_I and M=B−1EM=B^{-1}E. There are families q:R→F(S⊔I)q:R\to F(S\sqcup I) and u:I→F(S⊔I)u:I\to F(S\sqcup I) satisfying, for c∈C(R)c\in C(R) and d∈C(I)d\in C(I), εb1(qj)=1(j∈R),εb1(ui)=1(i∈I),ker⁡εb1=⟨⟨q(R)∪u(I)⟩⟩,ker⁡Db1,q⊔u=0,Db1,q(c)=(0,−Mc),Db1,u(d)=(Bd,0).\begin{gathered} \varepsilon_{b_1}(q_j)=1\quad(j\in R),\qquad \varepsilon_{b_1}(u_i)=1\quad(i\in I),\\ \ker\varepsilon_{b_1} =\langle\!\langle q(R)\cup u(I)\rangle\!\rangle, \qquad \ker D_{b_1,q\sqcup u}=0,\\ D_{b_1,q}(c)=(0,-Mc),\qquad D_{b_1,u}(d)=(Bd,0). \end{gathered}(7)✦

For relator block alignment (8), let v:S→Γv:S\to\Gamma, let z:Z→F(S)z:Z\to F(S) induce B:C(Z)≅Z1(v)B:C(Z)\cong Z_1(v), and let M:C(R)≅C(Z⊔T)M:C(R)\cong C(Z\sqcup T) be an integral linear isomorphism. Put b=(v⊔1T)⊔1Z⊔Tb=(v\sqcup1_T)\sqcup1_{Z\sqcup T}. Suppose null word families r:R→F((S⊔T)⊔(Z⊔T))r:R\to F((S\sqcup T)\sqcup(Z\sqcup T)) and u:Z⊔T→F((S⊔T)⊔(Z⊔T))u:Z\sqcup T\to F((S\sqcup T)\sqcup(Z\sqcup T)) together normally generate ker⁡εb\ker\varepsilon_b, with Db,r⊔uD_{b,r\sqcup u} injective, and satisfy Db,r(c)=(0,−Mc)D_{b,r}(c)=(0,-Mc) and Db,u(d)=(UBd,0)D_{b,u}(d)=(U_Bd,0) for all cc and dd in their chain groups. Set T2=T⊔(Z⊔T)T_2=T\sqcup(Z\sqcup T) and b2=v⊔1T2b_2=v\sqcup1_{T_2}. There is a family q:Z⊔(T⊔R)→F(S⊔T2)q:Z\sqcup(T\sqcup R)\to F(S\sqcup T_2) satisfying εb2(qj)=1(j∈Z⊔(T⊔R)),ker⁡Db2,q=0,ker⁡εb2=⟨⟨qj:j∈Z⊔(T⊔R)⟩⟩,[qi]=[ιzi](i∈Z),Db2,q|T⊔RT2=OM.\begin{gathered} \varepsilon_{b_2}(q_j)=1\quad(j\in Z\sqcup(T\sqcup R)),\qquad \ker D_{b_2,q}=0,\\ \ker\varepsilon_{b_2} =\langle\!\langle q_j:j\in Z\sqcup(T\sqcup R)\rangle\!\rangle,\\ [q_i]=[\iota z_i]\quad(i\in Z),\qquad D^{T_2}_{b_2,q|_{T\sqcup R}}=O_M. \end{gathered}(8)✦

In the paper, the result reads: under the hypotheses of Theorem 3.1, including a two-dimensional K(Γ,1)K(\Gamma,1), there is a genuine presentation ⟨S,T∣(ρz)z∈Z,(r)r∈R⟩\langle S,T\mid(\rho_z)_{z\in Z},(r)_{r\in R}\rangle with all TT-letters of value 11, the complete relator boundaries freely forming the cycle module, [ρz]=[z][\rho_z]=[z], where [z][z] denotes the prescribed cycle indexed by zz, and projection of the ordinary relator boundaries to TT-coordinates an isomorphism Λ(R)→Λ(T)\Lambda^{(R)}\to\Lambda^{(T)}, with no cardinality restriction on TT or RR.

What Is Checked

The alignments (6), (7) and (8) are stated over arbitrary groups and alphabets, using integral chain isomorphisms and the stated normal-generation hypotheses, with no cardinality bounds.