OctoInformBeta

OpenAI preprint, September 2026 · §4. A planar surface cannot have one nontrivial boundary

§4 (proof of Theorem 4.1): mixing and the uniform phase upper bound: A scalar mixer for positive and negative word disks with a unimodular sign assignment

Mixing gives a uniform boundary phase bound.

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 boundary operator for word disks over a group AA combines a disk successor, a band pairing and inverse geometric corner labels. Let II and KK be finite nonempty sets, with II indexing coefficient positions and KK indexing disks of each sign. Write ρ\rho for a permutation of II, a:I→Aa:I\to A for the coefficient family and s:I→ℤs:I\to\mathbb Z for the exponent family. The word disk slots form the set DD in (1); the subscripts distinguish two disjoint copies of KK. D=(K+⊔K−)×I,n=|I|,k=|K|,m=nk,|D|=2m.D=(K_+\sqcup K_-)\times I,\qquad n=|I|,\quad k=|K|,\quad m=nk,\qquad |D|=2m.(1)✦

For q∈Kq\in K and i∈Ii\in I, the signed disk successor σ\sigma, signed incidence coefficients cc and traversal sign vv are the functions on slots specified in (2). The values of cc are inverse geometric corner labels. σ(+,q,i)=(+,q,ρi),σ(−,q,i)=(−,q,ρ−1i),c(+,q,i)=ai−1,c(−,q,i)=aρ−1i,v(+,q,i)=si,v(−,q,i)=−si.\begin{aligned} \sigma(+,q,i)&=(+,q,\rho i),& \sigma(-,q,i)&=(-,q,\rho^{-1}i),\\ c(+,q,i)&=a_i^{-1},&c(-,q,i)&=a_{\rho^{-1}i},\\ v(+,q,i)&=s_i,&v(-,q,i)&=-s_i. \end{aligned}(2)

The Operators and Their Phase

For a finite coordinate set EE, the matrix Hilbert space ℋE=ℓ2(E×A,ℂ)\mathcal H_E=\ell^2(E\times A,\mathbb C) consists of square-summable complex families with the usual inner product.✦ A bounded operator on ℋE\mathcal H_E is a continuous complex-linear endomorphism of ℋE\mathcal H_E.✦ For d∈Ed\in E and x∈Ax\in A, the delta vector δ(d,x)\delta_{(d,x)} in ℋE\mathcal H_E is the family that is 11 at (d,x)(d,x) and 00 elsewhere.✦ The matrix unit EijE_{ij} on ℋE\mathcal H_E, for i,j∈Ei,j\in E, transfers coordinate jj to coordinate ii, leaving the AA coordinate unchanged; EiiE_{ii} is the coordinate projection. For a complex E×EE\times E matrix FF and a scalar family u:E→ℂu:E\to\mathbb C, the scalar embedding ιE(F)\iota_E(F) and the scalar diagonal operator Δu\Delta_u on ℋE\mathcal H_E are defined in (3). Eijδ(d,x)={δ(i,x)d=j,0d≠j,ιE(F)=∑i,j∈EFijEij,Δu=ιE(diag⁡(u)).E_{ij}\delta_{(d,x)}= \begin{cases}\delta_{(i,x)}&d=j,\\0&d\ne j,\end{cases} \qquad \iota_E(F)=\sum_{i,j\in E}F_{ij}E_{ij},\qquad \Delta_u=\iota_E(\operatorname{diag}(u)).(3)

For a bijection ee of coordinate sets, the linear coordinate permutation ReR_e sends a family ff to f∘e−1f\circ e^{-1}.✦ Given a permutation π\pi of EE and a family b:E→Ab:E\to A, the group-labelled coordinate permutation eπ,be_{\pi,b}, its labelled shift Tπ,bT_{\pi,b} and the weighted labelled shift Wπ,u,bW_{\pi,u,b} are given by (4), where u:E→ℂu:E\to\mathbb C is a scalar weight family. eπ,b(d,x)=(πd,bdx),Tπ,b=Reπ,b,Wπ,u,b=Tπ,bΔu,Wπ,u,bδ(d,x)=udδ(πd,bdx).\begin{aligned} e_{\pi,b}(d,x)&=(\pi d,b_d x),& T_{\pi,b}&=R_{e_{\pi,b}},\\ W_{\pi,u,b}&=T_{\pi,b}\Delta_u,& W_{\pi,u,b}\delta_{(d,x)}&=u_d\delta_{(\pi d,b_d x)}. \end{aligned}(4)

The scalar rotation r(x)r(x) and fractional argument hh have the conventions in (5). On ℂ\mathbb C, h(z)h(z) is the fractional part of the principal argument of zz divided by 2π2\pi. r(x)=exp⁡(2πix),h(r(x))=x−⌊x⌋(x∈ℝ),h(1)=0.r(x)=\exp(2\pi i x),\qquad h(r(x))=x-\lfloor x\rfloor\quad(x\in\mathbb R),\qquad h(1)=0.(5)

For a permutation θ\theta of DD and a real family γ:D→ℝ\gamma:D\to\mathbb R, set ud=r(γd)u_d=r(\gamma_d). The weighted permutation matrix Θ0\Theta_0 on ℂD\mathbb C^D, its scalar operator Θ\Theta on ℋD\mathcal H_D and the boundary operator BB on ℋD\mathcal H_D are defined by (6), with matrix indices d,e∈Dd,e\in D; the permutation σθ\sigma\theta applies θ\theta first. (Θ0)ed={ude=θd,0e≠θd,Θ=ιD(Θ0),B=Wσθ,u,c∘θ.(\Theta_0)_{ed}= \begin{cases}u_d&e=\theta d,\\0&e\ne\theta d,\end{cases} \qquad \Theta=\iota_D(\Theta_0),\qquad B=W_{\sigma\theta,u,c\circ\theta}.(6)

For a complex K×KK\times K matrix XX, the coefficient shift SS on ℋD\mathcal H_D, scalar mixing matrix CXC_X on ℂD\mathbb C^D and scalar mixer JXJ_X on ℋD\mathcal H_D are specified in (7). Each block of CXC_X acts on the positive and negative KK coordinates at one position ii. S=Tσ,c,CX=⨁i∈I(0X*X0),JX=ιD(CX).S=T_{\sigma,c},\qquad C_X=\bigoplus_{i\in I} \begin{pmatrix}0&X^*\\X&0\end{pmatrix},\qquad J_X=\iota_D(C_X).(7)

Let UU be a normal bounded operator on a complex Hilbert space and ξ\xi a vector. For a compactly supported continuous real function ff on the spectrum of UU, its complexification fℂf_{\mathbb C} is the same function viewed as complex-valued.✦ The spectral vector functional ΛU,ξ\Lambda_{U,\xi} and its finite representing measure μU,ξ\mu_{U,\xi} satisfy (8); ΛU,ξ\Lambda_{U,\xi} is positive and real-linear. ΛU,ξ(f)=Re⁡⟨ξ,fℂ(U)ξ⟩=∫spec⁡(U)f(z)dμU,ξ(z).\Lambda_{U,\xi}(f) =\operatorname{Re}\langle\xi,f_{\mathbb C}(U)\xi\rangle =\int_{\operatorname{spec}(U)}f(z)\,d\mu_{U,\xi}(z).(8)

For a normal bounded operator UU on a complex Hilbert space and a vector ξ\xi in that space, the vector spectral phase vU(ξ)v_U(\xi) is the integral in (9).✦ For a normal bounded operator UU on a complex Hilbert space, the summed vector phase of a finite family of vectors in that space is the sum of their vector spectral phases for UU.✦ For a normal bounded operator UU on ℋE\mathcal H_E, the unnormalized phase trace ℓ(U)\ell(U) is the sum in (9) of the vector spectral phases of the delta vectors at the identity 1∈A1\in A. vU(ξ)=∫spec⁡(U)h(z)dμU,ξ(z),ℓ(U)=∑d∈EvU(δ(d,1)).v_U(\xi)=\int_{\operatorname{spec}(U)}h(z)\,d\mu_{U,\xi}(z),\qquad \ell(U)=\sum_{d\in E}v_U(\delta_{(d,1)}).(9)

The fixed space Fix⁡(F)\operatorname{Fix}(F) of a complex matrix FF on ℂE\mathbb C^E is the subspace in (10), where IEI_E is the identity matrix on ℂE\mathbb C^E; its fixed space dimension is its complex dimension. Fix⁡(F)=ker⁡(F−IE).\operatorname{Fix}(F)=\ker(F-I_E).(10)

The Claim

Theorem 1. For every unitary K×KK\times K matrix XX, the phase of the product of the coefficient shift and the scalar mixer is given by (11). ℓ(SJX)=nk2.\ell(SJ_X)=\frac{nk}{2}.(11)✦ Let each sis_i be 11 or −1-1, with sum 11 or −1-1. Suppose that θ\theta is an involution reversing traversal signs, so that v(θd)=−v(d)v(\theta d)=-v(d), and that γ\gamma is antisymmetric, so that γθd=−γd\gamma_{\theta d}=-\gamma_d. There is a unitary K×KK\times K matrix XX satisfying (12). dim⁡ℂFix⁡(CXΘ0)≥2k.\dim_{\mathbb C}\operatorname{Fix}(C_X\Theta_0)\ge 2k.(12)✦ For any finite set EE and unitary complex E×EE\times E matrices F,QF,Q with QFQ*=F*QFQ^*=F^*, the scalar phase estimate is (13). ℓ(ιE(F))≤|E|−dim⁡ℂFix⁡(F)2.\ell(\iota_E(F))\le \frac{|E|-\dim_{\mathbb C}\operatorname{Fix}(F)}{2}.(13)✦ Under the stated conditions on ss, θ\theta and γ\gamma, the boundary operator satisfies (14). ℓ(B)≤3nk2−k.\ell(B)\le\frac{3nk}{2}-k.(14)✦

In the paper, the result reads: for a connected disk-and-band assembly, every antisymmetric choice of γ\gamma defines SS and Θ\Theta with Θ2=I\Theta^2=I, where II here denotes the identity operator, and a scalar mixing involution JJ can be chosen so that JΘJ\Theta has at least 2k2k fixed directions, ℓ(SJ)=nk/2\ell(SJ)=nk/2, ℓ(JΘ)≤nk−k\ell(J\Theta)\le nk-k, and ℓ(SΘ)≤3nk/2−k\ell(S\Theta)\le 3nk/2-k.

What Is Checked

The mixing and boundary claims (11), (12) and (14) are stated over arbitrary coefficient groups, finite nonempty position and disk index sets, and arbitrary position permutations, with the indicated conditions on signs, pairing and weights. The scalar estimate (13) concerns finite complex matrices and a unitary conjugacy between a matrix and its adjoint.

Coordinate permutations associated with bijections preserve square-summability✦ and inner products, and their linear reindexing maps are unitary. Scalar embedding preserves unitarity✦, and every scalar rotation r(x)r(x) is unitary.✦ A weighted group-labelled shift is unitary when all its scalar weights have modulus one.✦ The coefficient shift SS is unitary✦, as is JXJ_X for every unitary XX.✦