OctoInformBeta

OpenAI preprint, September 2026 · §3. A strict spectral surplus

Lemma 3.3: Fixed Berger parameters give a degreewise mean bound

A finite selection with a strict spectral surplus.

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

For harmonic growth, an angular eigenvalue λ\lambda is associated with an indicial root dd satisfying d(d+m−1)=λd(d+m-1)=\lambda, where mm is the link dimension. A finite selection of angular modes has both a total cardinality and a mean indicial root in each degree. The comparison concerns these two quantities at a fixed growth order.

The Cartesian dimension nn, link dimension mm, Berger parameter q*q_* and metric scale a*a_* are fixed by (1); β\beta denotes the inverse squared scale and kk the growth order. n=16,m=15,q*=101100,a*=99853100000q*1/15,β=a*−2,k=50000.n=16,\qquad m=15,\qquad q_*=\frac{101}{100},\qquad a_*=\sqrt{\frac{99853}{100000}q_*^{1/15}},\qquad \beta=a_*^{-2},\qquad k=50000.(1)

The real polynomial space in nn variables is ℝ[x1,…,xn]\mathbb R[x_1,\ldots,x_n].✦ The Euclidean polynomial Laplacian Δpol\Delta_{\mathrm{pol}} on ℝ[x1,…,xn]\mathbb R[x_1,\ldots,x_n] is the differential operator given in (2). The homogeneous harmonic subspace VlV_l of ℝ[x1,…,xn]\mathbb R[x_1,\ldots,x_n] consists of the homogeneous polynomials of degree ll annihilated by this Laplacian, as in (2); NlN_l is its real dimension, VlV_l is equipped with the Fischer inner product, and BlB_l is the round eigenvalue. Δpol=∑ν=1n∂xν,2,Vl={P∈ℝ[x1,…,xn]:P is homogeneous of degree l,ΔpolP=0},Nl=dim⁡ℝVl,Bl=l(l+m−1).\Delta_{\mathrm{pol}}=\sum_{\nu=1}^{n}\partial_{x_\nu}^{,2},\qquad V_l=\{P\in\mathbb R[x_1,\ldots,x_n]: P\text{ is homogeneous of degree }l,\ \Delta_{\mathrm{pol}}P=0\}, \qquad N_l=\dim_{\mathbb R}V_l,\qquad B_l=l(l+m-1).(2)

The Spectral Data

An orthogonal complex structure JJ on ℝ16\mathbb R^{16} is a real matrix with J2=−IJ^2=-I and J𝖳J=IJ^{\mathsf T}J=I.✦ The standard structure J0J_0 on ℝ8⊕ℝ8\mathbb R^8\oplus\mathbb R^8 and the infinitesimal coordinate motion ℛM\mathcal R_M generated by a real matrix MM are given in (3). The Hopf derivative DJD_J on VlV_l is the restriction of ℛJ\mathcal R_J, regarded as a continuous linear operator; its square is DJ2D_J^2. J0=(0−I8I80),ℛMP=∑ν,η=116Mνηxη∂xνP,DJ=ℛJ|Vl.J_0=\begin{pmatrix}0&-I_8\\ I_8&0\end{pmatrix},\qquad \mathcal R_M P=\sum_{\nu,\eta=1}^{16}M_{\nu\eta}x_\eta \partial_{x_\nu}P,\qquad D_J=\mathcal R_J\big|_{V_l}.(3) The reindexed complex structure JeJ^e of JJ under a coordinate bijection ee is the matrix with entries (Je)uv=Je(u)e(v)(J^e)_{uv}=J_{e(u)e(v)}.✦

An orthonormal spectral decomposition (el,i,μl,i)i∈Il(e_{l,i},\mu_{l,i})_{i\in I_l} of an operator on VlV_l consists of an orthonormal basis (el,i)i∈Il(e_{l,i})_{i\in I_l} of VlV_l together with its eigenvalues (μl,i)i∈Il(\mu_{l,i})_{i\in I_l}, where Il={1,…,Nl}I_l=\{1,\ldots,N_l\}.✦ Choose the standard Hopf square spectrum (el,i,μl,i)i∈Il(e_{l,i},\mu_{l,i})_{i\in I_l} as in (4). DJ0,2el,i=μl,iel,i,λl,i=β(q*1/15Bl+(q*1/15−q*1/15−1)μl,i),dl,i=−7+49+λl,i.D_{J_0}^{,2}e_{l,i}=\mu_{l,i}e_{l,i},\qquad \lambda_{l,i}=\beta\bigl(q_*^{1/15}B_l+ (q_*^{1/15}-q_*^{1/15-1})\mu_{l,i}\bigr),\qquad d_{l,i}=-7+\sqrt{49+\lambda_{l,i}}.(4) Here λl,i\lambda_{l,i} is the Berger angular eigenvalue and dl,id_{l,i} its larger indicial root.

The Finite Counts

The construction uses Lb=49962+249=50211L_b=49962+249=50211 spectral blocks.✦ A block numbered jj has degree j+2j+2, so LbL_b blocks cover degrees 2,…,Lb+12,\ldots,L_b+1, namely 2,…,502122,\ldots,50212.✦

For a nonnegative integer rr, the binomial polynomial (xr)\binom{x}{r} is the rational polynomial x(x−1)⋯(x−r+1)/r!x(x-1)\cdots(x-r+1)/r!, with (x0)=1\binom{x}{0}=1.✦ The reflected Vandermonde coefficient cl,jc_{l,j} and the squared charge summation primitive Pl,j(x)P_{l,j}(x) are the polynomials in (5). cl,j=(−1)j(l+6−j6−j)(6+j6),0≤j≤6,Pl,j(x)=(l−2j)2(x+77+j)+4(7+j)(2j+1−l)(x+78+j)+4(7+j)(8+j)(x+79+j).\begin{aligned} c_{l,j}&=(-1)^j\binom{l+6-j}{6-j}\binom{6+j}{6}, &&0\leq j\leq6,\\ P_{l,j}(x)&=(l-2j)^2\binom{x+7}{7+j} +4(7+j)(2j+1-l)\binom{x+7}{8+j}\\ &\hspace{2em}+4(7+j)(8+j)\binom{x+7}{9+j}. \end{aligned}(5) The multiplicity prefix sum Cl(x)C_l(x) and squared charge prefix sum Ql(x)Q_l(x) for degree ll are the polynomials given in (6). Cl(x)=∑j=06cl,j(x+77+j),Ql(x)=∑j=06cl,jPl,j(x).C_l(x)=\sum_{j=0}^{6}c_{l,j}\binom{x+7}{7+j},\qquad Q_l(x)=\sum_{j=0}^{6}c_{l,j}P_{l,j}(x).(6)

A paired Hopf prefix certificate (l,b,c,z)(l,b,c,z) is a quadruple of natural numbers satisfying (7).✦ The construction specifies one such certificate (l,bl,clsel,zl)(l,b_l,c_l^{\mathrm{sel}},z_l) for each of the 249249 degrees from 4996449964 through 5021250212. 49964≤l≤50212,2b<l,c>0,c=2(l+7)7Cl(b),z=2(l+7)7Ql(b),100000(101l(l+14)c−z)<101⋅99853⋅50000(50000+14)c.\begin{gathered} 49964\leq l\leq50212,\qquad 2b<l,\qquad c>0,\\ c=\frac{2(l+7)}7 C_l(b),\qquad z=\frac{2(l+7)}7 Q_l(b),\\ 100000\bigl(101l(l+14)c-z\bigr) <101\cdot99853\cdot50000(50000+14)c. \end{gathered}(7) The prescribed selected mode count MlM_l is the natural number in (8); the Euclidean comparator E(n,k)E(n,k) is the binomial expression in (9). Ml={Nl,2≤l≤49963,clsel,49964≤l≤50212.M_l=\begin{cases} N_l,&2\leq l\leq49963,\\ c_l^{\mathrm{sel}},&49964\leq l\leq50212. \end{cases}(8)✦ E(n,k)=(n+k−1k)+(n+k−2k−1).E(n,k)=\binom{n+k-1}{k}+\binom{n+k-2}{k-1}.(9)✦

The Claim

Theorem 1. For the parameters in (1) and each degree 2≤l≤502122\leq l\leq50212, there is a nonempty finite subset s⊆Ils\subseteq I_l with |s|=Ml.|s|=M_l.(10)✦ For the same parameters and degree, this subset satisfies ∑i∈sdl,i<50000|s|.\sum_{i\in s}d_{l,i}<50000\,|s|.(11)✦ Let SlS_l be the selected set obtained by fixing one such choice in each degree.✦ The prescribed counts have the strict surplus E(16,50000)<∑l=250212Ml.E(16,50000)<\sum_{l=2}^{50212}M_l.(12)✦

In the paper, the result reads: there is an odd m=2s−1≥7m=2s-1\geq7 such that a*∈(0,1)a_*\in(0,1) and q*>1q_*>1 can be taken arbitrarily close to 11, satisfying the link curvature margin at a*a_* for every q∈[1,q*]q\in[1,q_*], and for these parameters there exist integers K,L≥2K,L\geq2 and counts 0≤M̃l≤Nl0\leq\widetilde M_l\leq N_l for 2≤l≤L2\leq l\leq L with M̃l−1∑i=1M̃ldl,i<K\widetilde M_l^{-1}\sum_{i=1}^{\widetilde M_l}d_{l,i}<K whenever M̃l>0\widetilde M_l>0 and ∑l=2LM̃l>∑l=0KNl\sum_{l=2}^{L}\widetilde M_l>\sum_{l=0}^{K}N_l.

What Is Checked

The claims (10), (11) and (12) are stated over finite subsets of the standard Hopf eigenbasis indices of the polynomial spaces VlV_l, with the fixed Berger parameters and the exact rational prefix data in (7).

For an orthonormal basis bb, the orthonormal matrix representation ℳb(T)\mathcal M_b(T) is the real matrix of an operator TT in bb; it depends continuously and linearly on TT.✦ The standard Hopf square is symmetric for the Fischer inner product.✦ For a symmetric real operator TT on VlV_l, there is an orthonormal basis b=(bi)b=(b_i) of eigenvectors with eigenvalues τi\tau_i for which (13) holds. ℳb(T)=diag⁡(τi),Tbi=τibi.\mathcal M_b(T)=\operatorname{diag}(\tau_i),\qquad T b_i=\tau_i b_i.(13)