OctoInformBeta

OpenAI preprint, September 2026 · §5. Reducing the rank of a surface diagram

§5.3 (proof of Proposition 5.1): the subspace and quotient of a scalar disk: The induced submodule and quotient maps have the prescribed scalar form, including the closing edge

Scalar disk transport preserves every submodule.

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

A submodule chosen in one boundary fiber of a punctured disk can be transported successively along its old boundary. The closing side concerns both the transported submodule and the quotient of the ambient fiber by it.

Let RR be a commutative ring and AA a commutative RR-algebra, with structure map φ:R→A\varphi:R\to A. A ranked surface diagram DD has facets FF, seams SS and vertices 𝒱\mathcal V. A side occurrence at a seam ss is either its parent (s,*)(s,*) or its child (s,j)(s,j), where 0≤j<νs0\leq j<\nu_s and νs\nu_s is the seam arity.✦ Distinct occurrences remain distinct when their facet labels coincide. Write nan_a for the rank of the facet labelling a side occurrence aa. The total seam rank is ns=n(s,*)=∑j=0νs−1n(s,j).n_s=n_{(s,*)}=\sum_{j=0}^{\nu_s-1}n_{(s,j)}.(1)✦ Each child contributes separately to (1), in its specified order.

An ordered port assembly 𝒜\mathcal A for DD assigns an allowed type to every vertex, attaches its whole local ports bijectively to seam endpoints while preserving endpoint directions and ordered child counts, and labels side occurrences by facets so that locally paired occurrences have the same facet label. Its local vertex tables permit continuation, adjacent interchange, and splitting or merging.✦ At a splitting vertex the child lists are a prefix followed by one middle child and a suffix before the split, the same prefix followed by the ordered subblocks and the suffix after it, and those subblocks on the branch.✦ The corner pairing matches the before and after parents and their outer children; it matches the old middle child with the branch parent and each new subblock with its branch child. Reversing the endpoint directions gives the merging table.✦ At a splitting vertex, the before port and the branch are terminal seam endpoints, and the after port is an initial seam endpoint.

A permutation passage table for a permutation σ\sigma of a child label set has an incoming and an outgoing port with that same child set, pairs the two parent occurrences, and pairs an incoming child labelled jj with the outgoing child labelled σ(j)\sigma(j).✦ At a continuation the permutation is the identity, and at an adjacent interchange it exchanges the chosen consecutive positions.✦

A side germ η\eta of DD is a side occurrence of DD together with a choice of one of its two endpoints.✦ Positive boundary orientation follows the seam direction on the parent and opposes it on each child; this specifies the initial and terminal germs. The vertex incidence assembly identifies local endpoint occurrences bijectively with side germs, preserving their seam endpoints and their parent or child orientations, with equal facet labels on locally paired occurrences.✦ Under this identification, the local corner pairings give an involution on side germs that preserves facet labels and reverses boundary orientation.✦ The boundary successor sends a side to the side paired with its terminal germ.✦ For every facet ff, the diagram specifies its genus and enumerates its boundary circles by cyclic lists βf,b(i)\beta_{f,b}(i) of positive lengths Lf,bL_{f,b}. Together these lists contain every side occurrence exactly once and follow the boundary successor.

Scalar puncture data PP for DD are an assignment to each facet ff of a puncture count pfp_f and units λf,j∈R×\lambda_{f,j}\in R^\times, for 0≤j<pf0\leq j<p_f.✦

Identified Flags and Surface Solutions

For a finite coordinate set α\alpha with grading a:α→ℕa:\alpha\to\mathbb N, let Ea,k=Aa−1(k)E_{a,k}=A^{a^{-1}(k)} be the named module of grade kk. The coordinate prefix Ca,<k⊆AαC_{a,<k}\subseteq A^\alpha and the coordinate grade projection πa,k:Aα→Ea,k\pi_{a,k}:A^\alpha\to E_{a,k} are given by (2): Ca,<k={x:xt=0 whenever a(t)≥k},πa,k(x)=(xt)a(t)=k.C_{a,<k}=\{x:x_t=0\text{ whenever }a(t)\geq k\}, \qquad \pi_{a,k}(x)=(x_t)_{a(t)=k}.(2) Let WW be an AA-module; a frame e:Aα→∼We:A^\alpha\xrightarrow{\sim}W gives the framed prefix and grade map in (3): Fe,<k=e(Ca,<k),γe,k=πa,k∘e−1:W→Ea,k.F_{e,<k}=e(C_{a,<k}),\qquad \gamma_{e,k}=\pi_{a,k}\circ e^{-1}:W\longrightarrow E_{a,k}.(3) The restriction of the map in (3) identifies Fe,<k+1/Fe,<kF_{e,<k+1}/F_{e,<k} with Ea,kE_{a,k}. Two frames ee and e′e' from AαA^\alpha to WW, with coordinate grading aa, have the same identified flag when their prefixes agree at every level and their grade maps agree on the next prefix at every level.✦ A framed flag comparison cc on WW is a pair of coordinate isomorphisms to WW with a common source grading✦; agreement means equality of their identified flags.✦

Surface transport generators consist of a side transport HaH_a for every side occurrence, a seam frame EsεE_s^\varepsilon at each seam endpoint, where ε∈{0,1}\varepsilon\in\{0,1\} labels the two endpoints, and two handle transports Af,k,Bf,kA_{f,k},B_{f,k} for each handle.✦ Their sizes are respectively nan_a, nsn_s and the facet rank nfn_f.✦ An invertible coordinate matrix acts on column vectors, with its inverse acting by inverse matrix multiplication.✦ A surface transport word ww for DD is a matrix expression built from the surface transport generators of DD and invertible constant matrices by products, inverses and ordered direct sums.✦ Evaluation uses φ\varphi on constants and the chosen transport matrices on generators; the empty product is the identity. For a unit r∈R×r\in R^\times and a nonnegative integer dd, the invertible scalar matrix on RdR^d is the matrix of multiplication by rr.✦

The boundary monodromy word Mf,bM_{f,b} and surface relation word WfW_f have the order specified in (4), where [X,Y]=XYX−1Y−1[X,Y]=XYX^{-1}Y^{-1}: Mf,b=Hβf,b(Lf,b−1)⋯Hβf,b(0),Wf=(∏k[Af,k,Bf,k])(∏bMf,b).M_{f,b}=H_{\beta_{f,b}(L_{f,b}-1)}\cdots H_{\beta_{f,b}(0)}, \qquad W_f=\left(\prod_k[A_{f,k},B_{f,k}]\right)\left(\prod_bM_{f,b}\right).(4) The products over handles and boundaries in (4) use their specified order. The scalar surface target is the constant surface transport word that evaluates over AA to φ((∏jλf,j)−1)Inf\varphi((\prod_j\lambda_{f,j})^{-1})I_{n_f}.✦

A flag incidence circuit 𝒞\mathcal C for DD is a family of constraints, each specifying a pair of transport words of the same dimension and a coordinate grading.✦ A circuit solution gg of 𝒞\mathcal C is an assignment of invertible transport matrices for which the evaluated pair of words at each constraint has the same identified flag.✦ The surface and seam circuit compares WfW_f with its scalar surface target in constant grade zero at each facet. At a seam ss it compares the framed parent transport and framed inverse child transport in (5): H(s,*)Es0,Es1(⨁j=0νs−1H(s,j)−1).H_{(s,*)}E_s^0, \qquad E_s^1\left(\bigoplus_{j=0}^{\nu_s-1}H_{(s,j)}^{-1}\right).(5) The coordinate grading in (5) assigns grade jj to the coordinates of its jjth child block, retaining the named module An(s,j)A^{n_{(s,j)}} and its quotient identification.✦

The induced local occurrence ranks at a vertex are its facet ranks, equal across corner pairs and satisfying the parent sum at each port.✦ Reindexing a seam frame by the ordered child coordinates and the parent coordinates gives the port frame.✦ At a continuation the permitted vertex comparison compares the full ordered flags. The comparison for an adjacent interchange uses the coarsened flag whose exchanged pair is one named direct sum module. At a split it compares the after frame with the before frame refined in its middle block by the branch frame, retaining the prefix and suffix grade maps and the ordered subblock grade maps; a merge uses the same comparison.✦ A surface incidence solution gg is a solution of the surface and seam circuit for which every permitted vertex comparison has agreement of identified flags.✦ The splitting refinement is the isomorphism from the after child sum to the before child sum that applies the branch frame within the old middle block and uses the corner identifications on every prefix and suffix coordinate.

The Transported Submodule and Quotient

Fix a facet ff with genus zero and one boundary circle. Put n=nfn=n_f, V=AnV=A^n and m=Lf,0>0m=L_{f,0}>0. Its evaluated disk boundary transports are hi=Hβf,0(i)∈GL⁡n(A)h_i=H_{\beta_{f,0}(i)}\in\operatorname{GL}_n(A), for 0≤i<m0\leq i<m.✦ The disk monodromy scalar is the unit uu in (6): u=φ((∏j=0pf−1λf,j)−1)∈A×.u=\varphi\left(\left(\prod_{j=0}^{p_f-1}\lambda_{f,j}\right)^{-1}\right) \in A^\times.(6)✦ For a boundary position ii, the cyclic successor σ(i)\sigma(i) and closing edge scalar cic_i are defined by (7), using uu from (6): σ(i)=(i+1)mod⁡m,ci={1,i+1<m,u,i+1=m.\sigma(i)=(i+1)\bmod m,\qquad c_i=\begin{cases}1,&i+1<m,\\u,&i+1=m.\end{cases}(7)

Let U⊆VU\subseteq V be any AA-submodule. Initial segment transport gives the ambient frame TiT_i and the parallel submodule UiU_i in (8): T0=In,Ti=hi−1⋯h0(1≤i<m),Ui=Ti(U).T_0=I_n,\qquad T_i=h_{i-1}\cdots h_0\quad(1\leq i<m),\qquad U_i=T_i(U).(8) The transported submodule frame si:U→∼Uis_i:U\xrightarrow{\sim}U_i is the restriction of TiT_i, and the transported quotient frame s‾i:V/U→∼V/Ui\bar s_i:V/U\xrightarrow{\sim}V/U_i sends [v][v] to [Tiv][T_iv], with TiT_i as in (8). Write hiU:Ui→∼Uσ(i)h_i^U:U_i\xrightarrow{\sim}U_{\sigma(i)} for the restricted disk submodule transport and hiQ:V/Ui→∼V/Uσ(i)h_i^Q:V/U_i\xrightarrow{\sim}V/U_{\sigma(i)} for the induced disk quotient transport.

The Claim

Theorem 1. Let gg be a surface incidence solution over AA, let ff have genus zero and one boundary circle, and let U⊆AnfU\subseteq A^{n_f} be an AA-submodule. With the transported frames above and the scalars of (7), every boundary position ii satisfies (9): hi(Ui)=Uσ(i),hiU(si(v))=sσ(i)(civ)(v∈U),hiQ(s‾i(w))=s‾σ(i)(ciw)(w∈V/U).\begin{aligned} h_i(U_i)&=U_{\sigma(i)},\\ h_i^U(s_i(v))&=s_{\sigma(i)}(c_i v) &&(v\in U),\\ h_i^Q(\bar s_i(w))&=\bar s_{\sigma(i)}(c_i w) &&(w\in V/U). \end{aligned}(9)✦

In the paper, the result reads: for a rank-nn disk with scalar puncture monodromies, a chosen nonzero proper prefix of a starting seam flag extends to a parallel subsystem UU throughout the punctured disk, and Q=V/UQ=V/U has the same scalar monodromies, with both ranks smaller than nn.

What Is Checked

The claim is stated over commutative RR-algebras AA, surface incidence solutions of ranked diagrams, and arbitrary AA-submodules of the selected disk fiber. The evaluated boundary monodromy of a genus-zero facet with one boundary circle satisfies (10): hm−1⋯h0=uIn.h_{m-1}\cdots h_0=uI_n.(10)✦ The edge preservation theorem also applies to arbitrary cyclic families of invertible matrices with scalar total monodromy as in (10).