OctoInformBeta

OpenAI preprint, September 2026 · §3. Weighted Fourier grading · equations (4) and (6)

Block Parseval and the total energy: For an orthogonal projection family, the squared Hilbert-Schmidt norm is additive over blocks PsBPrP_sBP_r, and the centered grading’s is a weighted sum of traces

Block Parseval and the Hilbert-Schmidt norm of the centered grading.

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

Fix a weight w>0w > 0 on the cube Ω={−1,1}n\Omega = \{-1,1\}^n, which has N=2nN = 2^n points, and let DD be multiplication by w\sqrt{w} on the space ℂΩ\mathbb{C}^\Omega of complex functions on Ω\Omega. Let VkV_k be the span of the characters χS(x)=∏j∈Sxj\chi_S(x) = \prod_{j \in S} x_j with S⊆{1,…,n}S \subseteq \{1, \dots, n\} and |S|≤k|S| \le k. The subspaces Kk=DVkK_k = DV_k increase from K−1={0}K_{-1} = \{0\} to Kn=ℂΩK_n = \mathbb{C}^\Omega. The layer EkE_k is the part of KkK_k orthogonal to Kk−1K_{k-1}, that is Ek=Kk∩Kk−1⟂E_k = K_k \cap K_{k-1}^\perp, and ℂΩ\mathbb{C}^\Omega is the orthogonal direct sum of the layers E0,…,EnE_0, \dots, E_n. Write Πk\Pi_k for the orthogonal projection onto EkE_k.

Operators on ℂΩ\mathbb{C}^\Omega are matrices in the point basis, and an operator BB splits into the blocks ΠsBΠr\Pi_s B \Pi_r, each carrying ErE_r into EsE_s. The question is how the squared Hilbert-Schmidt norm of BB is shared among its blocks, and what it is for the centered grading M=L−(n/2)IdM = L - (n/2)\mathrm{Id}, where L=∑kkΠkL = \sum_k k\Pi_k is the grading, multiplication by kk on EkE_k.

The library answers both questions for every resolution of the identity into orthogonal projections on a space ℂI\mathbb{C}^I with II finite, and evaluates the second moment about n/2n/2 of the binomial coefficients (nk)\binom{n}{k}.

Projection Families and Gradings

For finite sets II and JJ, the squared Hilbert-Schmidt norm ∥A∥HS2\|A\|_{HS}^2 of a complex I×JI \times J matrix AA is the sum of the squared moduli of its entries: ∥A∥HS2=∑i∈I∑j∈J|Aij|2.\|A\|_{HS}^2 = \sum_{i \in I} \sum_{j \in J} |A_{ij}|^2 .(1)✦

An orthogonal projection family on ℂI\mathbb{C}^I, for a finite set II, is a family (Pk)k∈K(P_k)_{k \in K} of complex I×II \times I matrices, indexed by a finite set KK, whose members are mutually orthogonal Hermitian projections summing to the identity: Pk*=Pk,Pk2=Pk,PrPs=0(r≠s),∑k∈KPk=Id.P_k^* = P_k, \qquad P_k^2 = P_k, \qquad P_r P_s = 0 \ \ (r \neq s), \qquad \sum_{k \in K} P_k = \mathrm{Id}.(2)✦ The projections Π0,…,Πn\Pi_0, \dots, \Pi_n form such a family on ℂΩ\mathbb{C}^\Omega.

For a family (Pk)k∈K(P_k)_{k \in K} of complex I×II \times I matrices, with KK finite, and real weights (ak)k∈K(a_k)_{k \in K}, the grading of the family with weights (ak)(a_k) is the matrix ∑k∈KakPk\sum_{k \in K} a_k P_k.✦ The centered grading MPM_P of a family P=(Pk)k=0nP = (P_k)_{k=0}^{n} of complex I×II \times I matrices is its grading with weights k−n/2k - n/2: MP=∑k=0n(k−n2)Pk.M_P = \sum_{k=0}^{n} \Bigl(k - \frac{n}{2}\Bigr) P_k .(3)✦ The centered grading of the family (Πk)(\Pi_k) is the operator MM.

The Claim

Theorem 1. Let II be a finite set and (Pk)k∈K(P_k)_{k \in K} an orthogonal projection family on ℂI\mathbb{C}^I. Every complex I×II \times I matrix BB satisfies ∥B∥HS2=∑r∈K∑s∈K∥PsBPr∥HS2.\|B\|_{HS}^2 = \sum_{r \in K} \sum_{s \in K} \|P_s B P_r\|_{HS}^2 .(4)✦ If the family is indexed by {0,…,n}\{0, \dots, n\}, its centered grading satisfies ∥MP∥HS2=∑k=0n(k−n2)2RetrPk.\|M_P\|_{HS}^2 = \sum_{k=0}^{n} \Bigl(k - \frac{n}{2}\Bigr)^2 \, \mathrm{Re}\,\mathrm{tr}\, P_k .(5)✦ For every natural number nn, ∑k=0n(nk)(k−n2)2=n2n4.\sum_{k=0}^{n} \binom{n}{k} \Bigl(k - \frac{n}{2}\Bigr)^2 = \frac{n\,2^n}{4} .(6)✦

In the paper, the result reads: ∥B∥HS2=∑r,s∥ΠsBΠr∥HS2\|B\|_{HS}^2 = \sum_{r,s} \|\Pi_s B \Pi_r\|_{HS}^2 for every operator BB on ℂΩ\mathbb{C}^\Omega, and ∥M∥HS2=∑k(nk)(k−n/2)2=nN/4\|M\|_{HS}^2 = \sum_{k} \binom{n}{k} (k - n/2)^2 = nN/4.

For the family (Πk)(\Pi_k), where RetrΠk=dim⁡Ek=(nk)\mathrm{Re}\,\mathrm{tr}\,\Pi_k = \dim E_k = \binom{n}{k}, the first of these is (4), and the second is (5) followed by (6).

What Is Checked

The claim is stated over orthogonal projection families (2) of complex matrices indexed by a finite set and over the squared Hilbert-Schmidt norm (1); (5) is stated over the centered grading (3), the grading of a family indexed by {0,…,n}\{0, \dots, n\} with weights k−n/2k - n/2, and (6) over the binomial coefficients. For an orthogonal projection family (Pk)(P_k), the grading with weights (ak)(a_k) acts as multiplication by aka_k on the range of each PkP_k.