OctoInformBeta

OpenAI preprint, September 2026 · Section 4. Signed multilinear forms and conductor decay

Lemma 4.3: Real exact-support sums under prime tensor laws and full or good tuple products, with a squared slot product estimate

Conductor decay for conditional tensor energy.

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 coordinate support in a product of prime fields has a conductor given by the product of its primes. Conditional energy measures a real observable at one vertex against the configuration at the remaining vertices. The conductor estimates concern sums of complete exact coordinate supports.

Let JJ and VV be finite sets, let XjX_j be finite nonempty state spaces, and put X=∏j∈JXjX=\prod_{j\in J}X_j. All averages below are uniform probability averages; sums are counting sums. For a real linear functional LL on functions on a finite set Ω\Omega, its point mass mLm_L and its uniform relative joint density WL,eW_{L,e} for a bijection e:Ω→A×Be:\Omega\to A\times B are defined by (1). mL(z)=L(𝟏{z}),WL,e(a,b)=|Ω|mL(e−1(a,b)).m_L(z)=L(\mathbf1_{\{z\}}),\qquad W_{L,e}(a,b)=|\Omega|\,m_L\bigl(e^{-1}(a,b)\bigr).(1)

For a real kernel WW on A×BA\times B, the conditional energy ℰW(f)\mathcal E_W(f) of a real function f:A→ℝf:A\to\mathbb R is the quantity in (2), with a zero denominator contributing zero. ℰW(f)=𝔼b∈B(𝔼a∈Af(a)W(a,b))2𝔼a∈AW(a,b).\mathcal E_W(f)= \mathbb E_{b\in B} \frac{\bigl(\mathbb E_{a\in A}f(a)W(a,b)\bigr)^2} {\mathbb E_{a\in A}W(a,b)}.(2)✦

For configurations on VV, fix v∈Vv\in V and write Vv=V\{v}V_v=V\setminus\{v\}. The vertex decomposition separates {v}\{v\} from VvV_v, and gives the bijection ev(z)=(zv,z|Vv)e_v(z)=(z_v,z|_{V_v}). Write ℰL,v(f)=ℰWL,ev(f)\mathcal E_{L,v}(f)=\mathcal E_{W_{L,e_v}}(f).

For functionals LjL_j on real functions on XjVX_j^V, their tensor product Λ=⨂j∈JLj\Lambda=\bigotimes_{j\in J}L_j is the real linear functional on XVX^V defined by (3); the configuration assembled from (zj)j∈J(z_j)_{j\in J} has value (zj(w))j∈J(z_j(w))_{j\in J} at ww. Λ(F)=∑(zj)∈∏j∈JXjV(∏j∈JmLj(zj))F(w↦(zj(w))j∈J).\Lambda(F)= \sum_{(z_j)\in\prod_{j\in J}X_j^V} \left(\prod_{j\in J}m_{L_j}(z_j)\right) F\bigl(w\mapsto(z_j(w))_{j\in J}\bigr).(3)✦

An exact-support function ff on XX with support U⊆JU\subseteq J is a real function whose coordinate replacements satisfy (4), where xj←tx_{j\leftarrow t} is obtained from xx by replacing its jjth coordinate by tt. 𝔼t∈Xjf(xj←t)=0(j∈U),f(xj←t)=f(x)(j∉U),\mathbb E_{t\in X_j}f(x_{j\leftarrow t})=0 \quad(j\in U),\qquad f(x_{j\leftarrow t})=f(x) \quad(j\notin U),(4)✦ Here the conditions hold for every x∈Xx\in X, and the second for every t∈Xjt\in X_j. For prime coordinates Xj=𝔽pjX_j=\mathbb F_{p_j}, define the support conductor by q(U)=∏j∈Upjq(U)=\prod_{j\in U}p_j.

The Tuple Laws

The tuple construction uses height h=24h=24, block cutoff t0=80t_0=80, and b0=2t0+1=161b_0=2t_0+1=161 blocks. Put B0={0,…,b0−1}B_0=\{0,\ldots,b_0-1\}, r=h!/2r=h!/2, and δ=1/(10h2)\delta=1/(10h^2). Write 𝔖h\mathfrak S_h for the group of permutations of {0,…,h−1}\{0,\ldots,h-1\}. A permutation word uu on a block set BB is a function assigning such a permutation to every block✦; the tuple vertex set is 𝒱=(𝔖h)B0\mathcal V=(\mathfrak S_h)^{B_0}, of cardinality dd.✦ The cyclic shift c∈𝔖hc\in\mathfrak S_h sends each position to the next, returning the last to zero.✦ For a finite block set BB, a group GG, and group-valued words u,w∈GBu,w\in G^B, threshold shift adjacency with shift c∈Gc\in G and threshold kk means that |{b∈B:wb=ubc}|≥k|\{b\in B:w_b=u_bc\}|\ge k.✦ The list base is r=h!/2r=h!/2.✦ The interaction parameter is δ=1/(10h2)\delta=1/(10h^2).✦

An ordered position pair consists of two integers α\alpha and β\beta satisfying 0≤α<β<h0\le\alpha<\beta<h.✦ For a tuple vertex vv and a block bb, the position transposition vb;αβv^{b;\alpha\beta} replaces vbv_b by vb(αβ)v_b(\alpha\ \beta) and leaves the other blocks unchanged.✦ For active blocks S⊆B0S\subseteq B_0, the transposition interaction HSH_S is the matrix in (5). HS(v,w)=∑b∈S∑0≤α<β<hδβ−α𝟏{w=vb;αβ}.H_S(v,w)=\sum_{b\in S}\sum_{0\le\alpha<\beta<h} \delta^{\beta-\alpha}\mathbf1_{\{w=v^{b;\alpha\beta}\}}.(5)✦

For a real matrix HH on a finite set II and labelings ℓ,k:I→A\ell,k:I\to A, the label coincidence pairing KHK_H, index energy EHE_H, and index Boltzmann weight wHw_H are defined by (6). KH(ℓ,k)=∑u,w∈IH(u,w)𝟏{ℓ(u)=k(w)},EH(ℓ)=12KH(ℓ,ℓ),wH(ℓ)=exp⁡(EH(ℓ)).\begin{aligned} K_H(\ell,k)&=\sum_{u,w\in I}H(u,w)\mathbf1_{\{\ell(u)=k(w)\}},\\ E_H(\ell)&=\tfrac12K_H(\ell,\ell),\qquad w_H(\ell)=\exp(E_H(\ell)). \end{aligned}(6)

For active blocks S⊆B0S\subseteq B_0, an inactive-block address is an element of the set AS=(𝔖h)B0\SA_S=(\mathfrak S_h)^{B_0\setminus S}.✦ The addressed list indices are IS={0,…,r|S|−1}I_S=\{0,\ldots,r^{|S|}-1\}✦, and addressed list entries form ℒS=AS×IS\mathcal L_S=A_S\times I_S.✦ The inactive-block projection πS:𝒱→AS\pi_S:\mathcal V\to A_S is restriction to B0\SB_0\setminus S.✦ For a prime pp, list values Y:ℒS→𝔽pY:\mathcal L_S\to\mathbb F_p, and labels ℓ:𝒱→IS\ell:\mathcal V\to I_S, the sampled tuple configuration is zY,ℓ(v)=Y(πS(v),ℓ(v))z_{Y,\ell}(v)=Y(\pi_S(v),\ell(v)).✦ Let ES⊆ℒS2E_S\subseteq\mathcal L_S^2 be the list constraint edges specifying the square-difference constraints. The nonzero square indicator sps_p is one on nonzero squares in 𝔽p\mathbb F_p and zero elsewhere.✦ The square-constraint indicator aSa_S and tuple Boltzmann weight wSw_S are those in (7), which defines the tuple component average Cp,SC_{p,S} as a real linear functional on observables G:𝔽p𝒱→ℝG:\mathbb F_p^{\mathcal V}\to\mathbb R. aS(Y)=∏(u,w)∈ESsp(Y(w)−Y(u)),wS(ℓ)=wHS(ℓ),Cp,S(G)=𝔼Y:ℒS→𝔽p𝔼ℓ:𝒱→ISaS(Y)wS(ℓ)G(zY,ℓ).\begin{aligned} a_S(Y)&=\prod_{(u,w)\in E_S}s_p(Y(w)-Y(u)), &w_S(\ell)&=w_{H_S}(\ell),\\ C_{p,S}(G)&=\mathbb E_{Y:\mathcal L_S\to\mathbb F_p} \mathbb E_{\ell:\mathcal V\to I_S} a_S(Y)w_S(\ell)G(z_{Y,\ell}). \end{aligned}(7)

For a family DD of block subsets, the restricted tuple mixture Mp,DM_{p,D} is the weighted sum in (8), with block subset mixture weight p−|S|/16p^{-|S|/16}. For a truncation level nn, the truncated family DnD_n consists of the subsets of B0B_0 of cardinality at most nn.✦ The unnormalized full mixture MpM_p, normalized restricted law μp,D\mu_{p,D}, full tuple law μp\mu_p, and good tuple law μp∘\mu_p^\circ are defined in the same display. Dn={S⊆B0:|S|≤n},Mp,D(G)=∑S∈Dp−|S|/16Cp,S(G),Mp=Mp,Dt0+1,μp,D(G)=Mp,D(G)Mp,D(1),μp(G)=Mp(G)Mp(1),μp∘=μp,Dt0.\begin{aligned} D_n&=\{S\subseteq B_0:|S|\le n\}, &M_{p,D}(G)&=\sum_{S\in D}p^{-|S|/16}C_{p,S}(G),\\ M_p&=M_{p,D_{t_0+1}}, &\mu_{p,D}(G)&=\frac{M_{p,D}(G)}{M_{p,D}(1)},\\ \mu_p(G)&=\frac{M_p(G)}{M_p(1)}, &\mu_p^\circ&=\mu_{p,D_{t_0}}. \end{aligned}(8) The product tuple law on X𝒱X^{\mathcal V} is 𝝁=⨂j∈Jμpj\boldsymbol\mu=\bigotimes_{j\in J}\mu_{p_j}.✦

The Prime Threshold

For a bijection e:I→Ke:I\to K and an edge set E⊆K2E\subseteq K^2, the graph pullback is (e×e)−1(E)(e\times e)^{-1}(E).✦ The crossing edges of an edge set on a disjoint union I⊔KI\sqcup K are the pairs in I×KI\times K present in either orientation.✦ For an edge set EE on VV, let mE(v)m_E(v) count the crossing edges between {v}\{v\} and VvV_v. The graph conditional energy constant ΓE(v)\Gamma_E(v) of an edge set EE on VV at a vertex vv is the quantity in (9). ΓE(v)=(2(1/2)mE(v)+4((1/2)mE(v))2)((32mE(v))1/4)2.\Gamma_E(v)= \left(\frac{2}{(1/2)^{m_E(v)}}+ \frac{4}{((1/2)^{m_E(v)})^2}\right) \left((32m_E(v))^{1/4}\right)^2.(9)✦

The empty-block list index is the unique index 0∈I⌀0\in I_\varnothing.✦ The leading list identification sends vv to (π⌀(v),0)(\pi_\varnothing(v),0); its inverse reads the permutation at each block from the address.✦ Let E*⊆𝒱2E_*\subseteq\mathcal V^2 be the leading constraint graph, the pullback of E⌀E_\varnothing along this identification✦, and let e=|E⌀|e=|E_\varnothing| be the leading graph edge count.✦ The leading mass lower bound a*a_* and Boltzmann weight upper bound W*W_*, together with the conditional error numerator N*N_* and conditional decay constant C*C_*, are defined by (10). a*=2−(e+1),W*=exp⁡(d(t0+1)(h2)δ2),N*=∑v∈𝒱ΓE*(v)+2b0W*,C*=N*/a*.\begin{aligned} a_*&=2^{-(e+1)}, &W_*&=\exp\left(\frac{d(t_0+1)\binom h2\delta}{2}\right),\\ N_*&=\sum_{v\in\mathcal V}\Gamma_{E_*}(v)+2^{b_0}W_*, &C_*&=N_*/a_*. \end{aligned}(10)

The mass positivity threshold PmassP_{\mathrm{mass}}, conditional decay threshold PcondP_{\mathrm{cond}}, and their common prime cutoff P*P_* are given by (11). Pmass=max⁡{3,(2e+1e)2},Pcond=max⁡{Pmass,3,C*32},P*=max⁡{Pmass,Pcond}.\begin{aligned} P_{\mathrm{mass}}&=\max\{3,(2^{e+1}e)^2\},\\ P_{\mathrm{cond}}&=\max\{P_{\mathrm{mass}},3,C_*^{32}\},\qquad P_* = \max\{P_{\mathrm{mass}},P_{\mathrm{cond}}\}. \end{aligned}(11)

The Claim

Theorem 1. Let J,VJ,V be finite, let pjp_j be prime for every j∈Jj\in J, and put X=∏j∈J𝔽pjX=\prod_{j\in J}\mathbb F_{p_j}. Fix v∈Vv\in V and T>0T>0. Let 𝒰\mathscr U be a finite family of subsets of JJ, and let fU:X→ℝf_U:X\to\mathbb R have exact support UU for each U∈𝒰U\in\mathscr U. Assume q(U)≥Tq(U)\ge T throughout this family, and put g=∑U∈𝒰fUg=\sum_{U\in\mathscr U}f_U.

Suppose LjL_j are positive real linear functionals on 𝔽pjV\mathbb F_{p_j}^V with uniform marginals at every vertex. Assume that every mean-zero real function a:𝔽pj→ℝa:\mathbb F_{p_j}\to\mathbb R satisfies (12). ℰLj,v(a)≤pj−1/32𝔼x∈𝔽pja(x)2.\mathcal E_{L_j,v}(a)\le p_j^{-1/32}\mathbb E_{x\in\mathbb F_{p_j}}a(x)^2.(12) The tensor conductor estimate is (13). ℰ⨂jLj,v(g)≤T−1/32𝔼x∈Xg(x)2.\mathcal E_{\bigotimes_jL_j,v}(g) \le T^{-1/32}\mathbb E_{x\in X}g(x)^2.(13)✦

For V=𝒱V=\mathcal V and pj≥P*p_j\ge P_* at every coordinate, the good tuple product satisfies (14). ℰ⨂jμpj∘,v(g)≤T−1/32𝔼x∈Xg(x)2.\mathcal E_{\bigotimes_j\mu_{p_j}^\circ,v}(g) \le T^{-1/32}\mathbb E_{x\in X}g(x)^2.(14)✦ Under the same prime threshold, the full tuple product satisfies (15). ℰ𝝁,v(g)≤T−1/32𝔼x∈Xg(x)2.\mathcal E_{\boldsymbol\mu,v}(g) \le T^{-1/32}\mathbb E_{x\in X}g(x)^2.(15)✦

For the slot estimate, let AA be any finite nonempty state space, let VV be finite, and let v∈Vv\in V have m=|Vv|≥1m=|V_v|\ge1. Let LL be a positive real linear functional on AVA^V, let f:A→ℝf:A\to\mathbb R, and let Fw:A→ℝF_w:A\to\mathbb R for each w∈Vvw\in V_v. If a real number MM satisfies L(z↦|Fw(zw)|2m)≤ML(z\mapsto|F_w(z_w)|^{2m})\le M for every w∈Vvw\in V_v, then (16) holds. [L(z↦f(zv)∏w∈VvFw(zw))]2≤ℰL,v(f)M.\left[L\left(z\mapsto f(z_v)\prod_{w\in V_v}F_w(z_w)\right)\right]^2 \le \mathcal E_{L,v}(f)\,M.(16)✦

Suppose the coordinate laws satisfy (12). Let ss be a finite index set with assigned supports σi⊆J\sigma_i\subseteq J, allowing repetitions. Let fi:X→ℝf_i:X\to\mathbb R have exact support σi\sigma_i, and suppose q(σi)≥Tq(\sigma_i)\ge T for every i∈si\in s. The grouped conductor estimate is (17). ℰ⨂jLj,v(∑i∈sfi)≤T−1/32𝔼x∈X(∑i∈sfi(x))2.\mathcal E_{\bigotimes_jL_j,v}\left(\sum_{i\in s}f_i\right) \le T^{-1/32}\mathbb E_{x\in X} \left(\sum_{i\in s}f_i(x)\right)^2.(17)✦

In the paper, the result reads: for a finite set 𝒫\mathcal P of primes at least a cutoff PP, independently chosen laws ρp∈{μp,μp∘}\rho_p\in\{\mu_p,\mu_p^\circ\}, R≥1R\ge1, and complex functions gw:∏p∈𝒫𝔽p→ℂg_w:\prod_{p\in\mathcal P}\mathbb F_p\to\mathbb C such that one slot vv has gˆv(ξ)=0\widehat g_v(\xi)=0 whenever a character ξ\xi has denominator q(ξ)≤Rq(\xi)\le R, one has |∫∏w∈𝒱gw(zw)d⨂p∈𝒫ρp|≤R−1/64∥gv∥2∏w≠v∥gw∥2(d−1)\left|\int\prod_{w\in\mathcal V}g_w(z_w)\,d\bigotimes_{p\in\mathcal P}\rho_p\right| \le R^{-1/64}\|g_v\|_2\prod_{w\ne v}\|g_w\|_{2(d-1)} with constant one and uniform norms.

What Is Checked

The conductor claims concern real functions with complete exact coordinate supports in a fixed finite prime family, under positive coordinate laws with uniform marginals and (12), or under the specified tuple laws above (11); the slot estimate (16) concerns any positive real linear functional on a finite nonempty configuration space with at least two vertices and the stated moments. The tensor product law, tuple component averages, and full and restricted tuple mixtures preserve sums of observables and multiplication by real scalars.