OctoInformBeta

OpenAI preprint, September 2026 · §5. From a cone map with a full corner to a dimension-drop map

5.1 Tensor powers of the cone algebra C: Isometric lifting for cube functions whose generalized-Pauli coefficients vanish on zeros of the cone cube coefficients, under a rank N≥2N \ge 2 Pauli system

Faithful cube representations of matrix cone tensor powers.

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 a matrix dimension N≥2N \ge 2 and an index n∈ℕn \in \mathbb{N}, consider the algebra of N×NN \times N complex matrix-valued continuous paths on [0,1][0, 1] taking scalar values at 00. Coordinatewise evaluation carries its (n+1)(n+1)-fold maximal tensor power into continuous functions on the (n+1)(n+1)-dimensional unit cube [0,1]n+1[0, 1]^{n+1} taking values in the (n+1)(n+1)-fold matrix tensor power. The question is whether this representation is injective and how its image is characterized by vanishing generalized-Pauli coefficients at zero coordinates.

Cones, Tensor Powers, and the Cube Representation

A positive dimension NN is a natural number N>0N > 0.✦ The matrix C*C^*-algebra MNM_N is the unital complex C*C^*-algebra of complex matrices whose rows and columns are indexed by {0,…,N−1}\{0, \dots, N - 1\} (recovering M2M_2 when N=2N = 2).✦

For each t∈[0,1]t \in [0, 1], path evaluation ev⁡t:C([0,1],MN)→MN\operatorname{ev}_t : C([0, 1], M_N) \to M_N is the unital **-homomorphism sending each continuous path ff to f(t)f(t).✦ The matrix cone C*C^*-algebra CC (the cone subalgebra of C([0,1],MN)C([0, 1], M_N)) is the preimage ev⁡0−1(ℂ⋅1)\operatorname{ev}_0^{-1}(\mathbb{C} \cdot 1) of the scalar subalgebra ℂ⋅1\mathbb{C} \cdot 1 in MNM_N, a unital C*C^*-subalgebra of C([0,1],MN)C([0, 1], M_N) recovering the M2M_2-cone when N=2N = 2.✦

For a unital complex C*C^*-algebra DD and n∈ℕn \in \mathbb{N}, the (n+1)(n+1)-fold right-nested maximal tensor power D⊗max(n+1)D^{\otimes_{\max}(n+1)} is the unital complex C*C^*-algebra defined recursively by D⊗max1=D,D⊗max(n+2)=D⊗maxD⊗max(n+1).D^{\otimes_{\max} 1} = D, \qquad D^{\otimes_{\max}(n+2)} = D \otimes_{\max} D^{\otimes_{\max}(n+1)}.(1)✦ For n∈ℕn \in \mathbb{N}, the cone tensor power C⊗max(n+1)C^{\otimes_{\max}(n+1)} is the unital complex C*C^*-algebra obtained by setting D=CD = C in (1).✦

The unit cube [0,1]n+1[0, 1]^{n+1} for n∈ℕn \in \mathbb{N} is the space of (n+1)(n+1)-tuples t=(t0,…,tn)t = (t_0, \dots, t_n) with tj∈[0,1]t_j \in [0, 1] for all j∈{0,…,n}j \in \{0, \dots, n\}.✦ For n∈ℕn \in \mathbb{N} and an integer N≥2N \ge 2, the iterated row index set RnR_n is the finite set defined recursively by R0={0,…,N−1}R_0 = \{0, \dots, N - 1\} and Rn+1={0,…,N−1}×RnR_{n+1} = \{0, \dots, N - 1\} \times R_n, and the matrix C*C^*-algebra MN⊗(n+1)M_N^{\otimes(n+1)} is the unital complex C*C^*-algebra of matrices whose rows and columns are indexed by RnR_n.✦

For each n∈ℕn \in \mathbb{N}, the left and right tensor-factor inclusions λn:MN→MN⊗(n+2)\lambda_n : M_N \to M_N^{\otimes(n+2)} and ρn:MN⊗(n+1)→MN⊗(n+2)\rho_n : M_N^{\otimes(n+1)} \to M_N^{\otimes(n+2)} are commuting unital **-homomorphisms. The matrix leg inclusions ιn,j:MN→MN⊗(n+1)\iota_{n,j} : M_N \to M_N^{\otimes(n+1)} for j∈{0,…,n}j \in \{0, \dots, n\} are the unital **-homomorphisms defined recursively by ι0,0=id⁡MN\iota_{0,0} = \operatorname{id}_{M_N}, ιn+1,0=λn\iota_{n+1,0} = \lambda_n, and ιn+1,j+1=ρn∘ιn,j\iota_{n+1,j+1} = \rho_n \circ \iota_{n,j} for j∈{0,…,n}j \in \{0, \dots, n\}.✦

The cone tensor representation πn:C⊗max(n+1)→C([0,1]n+1,MN⊗(n+1))\pi_n : C^{\otimes_{\max}(n+1)} \to C([0, 1]^{n+1}, M_N^{\otimes(n+1)}) is the unital **-homomorphism defined recursively for each n∈ℕn \in \mathbb{N}. At n=0n = 0, the cone tensor representation π0:C→C([0,1]1,MN)\pi_0 : C \to C([0, 1]^1, M_N) is the unital **-homomorphism sending each g∈Cg \in C to the path t↦g(t0)t \mapsto g(t_0) on [0,1]1[0, 1]^1. At step n+1n + 1, writing (t0,t′)∈[0,1]×[0,1]n+1(t_0, t') \in [0, 1] \times [0, 1]^{n+1} for a point of [0,1]n+2[0, 1]^{n+2}, the commuting unital **-homomorphisms ℓn:C→C([0,1]n+2,MN⊗(n+2))\ell_n : C \to C([0, 1]^{n+2}, M_N^{\otimes(n+2)}) and rn:C⊗max(n+1)→C([0,1]n+2,MN⊗(n+2))r_n : C^{\otimes_{\max}(n+1)} \to C([0, 1]^{n+2}, M_N^{\otimes(n+2)}) defined for g∈Cg \in C and b∈C⊗max(n+1)b \in C^{\otimes_{\max}(n+1)} by ℓn(g)(t0,t′)=λn(g(t0)),rn(b)(t0,t′)=ρn(πn(b)(t′)),\ell_n(g)(t_0, t') = \lambda_n(g(t_0)), \qquad r_n(b)(t_0, t') = \rho_n(\pi_n(b)(t')),(2) induce πn+1:C⊗maxC⊗max(n+1)→C([0,1]n+2,MN⊗(n+2))\pi_{n+1} : C \otimes_{\max} C^{\otimes_{\max}(n+1)} \to C([0, 1]^{n+2}, M_N^{\otimes(n+2)}) via the universal property of the maximal tensor product.

Generalized Pauli Systems and Coefficient Maps

For N≥2N \ge 2, a Pauli matrix system on MNM_N consists of a finite index set LL of cardinality |L|=N2|L| = N^2, a distinguished zero label 0∈L0 \in L, and a family of matrices {eα:α∈L}⊆MN\{e_\alpha : \alpha \in L\} \subseteq M_N satisfying (3), such that for every α∈L\alpha \in L with α≠0\alpha \ne 0 there exists u∈MNu \in M_N satisfying u2=1u^2 = 1 and ueα=−eαuu e_\alpha = -e_\alpha u.✦ e0=1,eα*=eα,eα2=1,Tr⁡(eαeβ)=δαβN,e_0 = 1, \qquad e_\alpha^* = e_\alpha, \qquad e_\alpha^2 = 1, \qquad \operatorname{Tr}(e_\alpha e_\beta) = \delta_{\alpha\beta} N,(3)

Given a Pauli matrix system on MNM_N, for each n∈ℕn \in \mathbb{N} the iterated Pauli index set LnL_n is the finite set defined recursively by L0=LL_0 = L and Ln+1=L×LnL_{n+1} = L \times L_n.✦ For n∈ℕn \in \mathbb{N} and a∈Lna \in L_n, the Pauli label coordinates aj∈La_j \in L for j∈{0,…,n}j \in \{0, \dots, n\} are the elements defined recursively by a0=aa_0 = a at n=0n = 0, and at step n+1n + 1 for a=(α,a′)∈L×Lna = (\alpha, a') \in L \times L_n by a0=αa_0 = \alpha and aj+1=aj′a_{j+1} = a'_j for 0≤j≤n0 \le j \le n.✦ The generalized-Pauli basis matrix ea∈MN⊗(n+1)e_a \in M_N^{\otimes(n+1)} is defined recursively for a∈Lna \in L_n by setting ea=eαe_a = e_\alpha for a=α∈La = \alpha \in L at n=0n = 0, and e(α,a′)=eα⊗ea′e_{(\alpha, a')} = e_\alpha \otimes e_{a'} for a=(α,a′)∈L×Lna = (\alpha, a') \in L \times L_n at step n+1n + 1.✦

For each n∈ℕn \in \mathbb{N} and a∈Lna \in L_n, the continuous generalized-Pauli coefficient ca:MN⊗(n+1)→ℂc_a : M_N^{\otimes(n+1)} \to \mathbb{C} is the bounded linear functional defined for x∈MN⊗(n+1)x \in M_N^{\otimes(n+1)} by ca(x)=1Nn+1Tr⁡(eax).c_a(x) = \frac{1}{N^{n+1}} \operatorname{Tr}(e_a x).(4)✦

For each α∈L\alpha \in L, the single-cone coefficient function pα(1)∈C([0,1],ℂ)p^{(1)}_\alpha \in C([0, 1], \mathbb{C}) is defined for t∈[0,1]t \in [0, 1] by pα(1)(t)=1p^{(1)}_\alpha(t) = 1 if α=0\alpha = 0 and pα(1)(t)=tp^{(1)}_\alpha(t) = t otherwise. For each n∈ℕn \in \mathbb{N} and a∈Lna \in L_n, the cone cube coefficient map pa∈C([0,1]n+1,ℂ)p_a \in C([0, 1]^{n+1}, \mathbb{C}) is defined recursively by pa(t)=pa(1)(t0)(n=0),p(α,a′)(t0,t′)=pα(1)(t0)pa′(t′)((α,a′)∈L×Ln),p_a(t) = p^{(1)}_a(t_0) \quad (n = 0), \qquad p_{(\alpha, a')}(t_0, t') = p^{(1)}_\alpha(t_0)\, p_{a'}(t') \quad ((\alpha, a') \in L \times L_n),(5)✦ for t=(t0)∈[0,1]1t = (t_0) \in [0, 1]^1 at n=0n = 0 and t=(t0,t′)∈[0,1]×[0,1]n+1t = (t_0, t') \in [0, 1] \times [0, 1]^{n+1} at step n+1n + 1.

The Claim

Theorem 1. Let N≥2N \ge 2 be an integer equipped with a Pauli matrix system, and let n∈ℕn \in \mathbb{N}.

  1. If b1,b2∈C⊗max(n+1)b_1, b_2 \in C^{\otimes_{\max}(n+1)} satisfy πn(b1)=πn(b2)\pi_n(b_1) = \pi_n(b_2), then b1=b2.b_1 = b_2.(6)✦

  2. Let f∈C([0,1]n+1,MN⊗(n+1))f \in C([0, 1]^{n+1}, M_N^{\otimes(n+1)}) satisfy ca(f(t))=0c_a(f(t)) = 0 for all a∈Lna \in L_n and t∈[0,1]n+1t \in [0, 1]^{n+1} with pa(t)=0p_a(t) = 0. Then there exists a unique b∈C⊗max(n+1)b \in C^{\otimes_{\max}(n+1)} such that πn(b)=fand∥b∥=∥f∥.\pi_n(b) = f \qquad \text{and} \qquad \|b\| = \|f\|.(7)✦

  3. For any a∈Lna \in L_n and t∈[0,1]n+1t \in [0, 1]^{n+1} satisfying pa(t)=0p_a(t) = 0, there exists j∈{0,…,n}j \in \{0, \dots, n\} such that tj=0andaj≠0.t_j = 0 \qquad \text{and} \qquad a_j \ne 0.(8)✦

  4. For any a∈Lna \in L_n, if an element x∈MN⊗(n+1)x \in M_N^{\otimes(n+1)} commutes with ιn,j(MN)\iota_{n,j}(M_N) for some j∈{0,…,n}j \in \{0, \dots, n\} with aj≠0a_j \ne 0, then ca(x)=0.c_a(x) = 0.(9)✦

In the paper, the result reads: for k≥1k \ge 1 (corresponding to k=n+1k = n + 1 and N=2N = 2), C⊗maxkC^{\otimes_{\max} k} identifies with {f∈C([0,1]k,M2⊗k):f(t)∈⨂i=1kEi(t)}\{f \in C([0, 1]^k, M_2^{\otimes k}) : f(t) \in \bigotimes_{i=1}^k E_i(t)\}, where Ei(t)=M2E_i(t) = M_2 if ti>0t_i > 0 and Ei(t)=ℂ⋅1E_i(t) = \mathbb{C} \cdot 1 if ti=0t_i = 0.

What Is Checked

The theorem is stated for any positive dimension N≥2N \ge 2 equipped with a Pauli matrix system and any n∈ℕn \in \mathbb{N}, over the cone tensor power C⊗max(n+1)C^{\otimes_{\max}(n+1)} of the matrix cone C*C^*-algebra CC, the cone tensor representation πn:C⊗max(n+1)→C([0,1]n+1,MN⊗(n+1))\pi_n : C^{\otimes_{\max}(n+1)} \to C([0, 1]^{n+1}, M_N^{\otimes(n+1)}), the cone cube coefficient maps pap_a, the continuous generalized-Pauli coefficients cac_a, the Pauli label coordinates aja_j, and the matrix leg inclusions ιn,j\iota_{n,j}, for functions f∈C([0,1]n+1,MN⊗(n+1))f \in C([0, 1]^{n+1}, M_N^{\otimes(n+1)}) with ca(f(t))=0c_a(f(t)) = 0 where pa(t)=0p_a(t) = 0 and elements x∈MN⊗(n+1)x \in M_N^{\otimes(n+1)} commuting with ιn,j(MN)\iota_{n,j}(M_N) where aj≠0a_j \ne 0.