OctoInformBeta

OpenAI preprint, September 2026 · §3. The dimension-drop criterion without stable rank one

3.2 Dimension-drop map from commuting order-zero matrix maps: For N≥1N \ge 1, commuting weighted matrix-unit systems of sizes NN and N+1N+1 whose weights sum to 1A1_A induce a unital **-homomorphism I(N,N+1)→AI(N, N+1) \to A

Constructing a unital **-homomorphism from the dimension-drop algebra.

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 positive integer N≥1N \ge 1 and a complex unital C*-algebra AA. Given two commuting systems of weighted matrix units in AA of ranks NN and N+1N+1 whose weights sum to 1A1_A, the question is how to build a unital **-homomorphism from the dimension-drop algebra I(N,N+1)I(N, N+1) into AA by combining matrix maps at the endpoints 00 and 11 with an interior continuous field representation on [0,1][0, 1]. These notes go through the definitions this construction rests on and the theorem itself.

The Dimension-drop Algebra and Drop Sections

For the integer N≥1N \ge 1, let MN=MN(ℂ)M_N = M_N(\mathbb{C}) and MN+1=MN+1(ℂ)M_{N+1} = M_{N+1}(\mathbb{C}) be the C*-algebras of complex matrices indexed by {1,…,N}\{1, \dots, N\} and {1,…,N+1}\{1, \dots, N+1\} (with identity matrices 1N1_N and 1N+11_{N+1}), and let MN(N+1)=MN(N+1)(ℂ)M_{N(N+1)} = M_{N(N+1)}(\mathbb{C}) be the C*-algebra of complex matrices indexed by {1,…,N}×{1,…,N+1}\{1, \dots, N\} \times \{1, \dots, N+1\} (recovering M2M_2, M3M_3, and M6M_6 when N=2N = 2). The left matrix inclusion λleft:MN→MN(N+1)\lambda_{\mathrm{left}} : M_N \to M_{N(N+1)} is the unital **-homomorphism a↦a⊗1N+1a \mapsto a \otimes 1_{N+1}✦, and the right matrix inclusion λright:MN+1→MN(N+1)\lambda_{\mathrm{right}} : M_{N+1} \to M_{N(N+1)} is the unital **-homomorphism b↦1N⊗bb \mapsto 1_N \otimes b, formed via the Kronecker product on {1,…,N}×{1,…,N+1}\{1, \dots, N\} \times \{1, \dots, N+1\}.✦

Write C([0,1],ℂ)C([0, 1], \mathbb{C}) for the C*-algebra of continuous complex-valued functions on [0,1][0, 1]✦, and write C([0,1],MN(N+1))C([0, 1], M_{N(N+1)}) for the C*-algebra of continuous MN(N+1)M_{N(N+1)}-valued paths on [0,1][0, 1], identified entrywise via the unital **-homomorphism f↦(s↦f(s)uv)u,vf \mapsto (s \mapsto f(s)_{uv})_{u,v} with the matrix path algebra MN(N+1)(C([0,1],ℂ))M_{N(N+1)}(C([0, 1], \mathbb{C})) indexed by {1,…,N}×{1,…,N+1}\{1, \dots, N\} \times \{1, \dots, N+1\}. For each s∈[0,1]s \in [0, 1], point evaluation ev⁡s:C([0,1],MN(N+1))→MN(N+1)\operatorname{ev}_s : C([0, 1], M_{N(N+1)}) \to M_{N(N+1)} is the unital **-homomorphism given by ev⁡s(f)=f(s)\operatorname{ev}_s(f) = f(s).✦

The dimension-drop algebra I(N,N+1)I(N, N+1) (recovering I(2,3)I(2, 3) when N=2N = 2) is the unital **-subalgebra of C([0,1],MN(N+1))C([0, 1], M_{N(N+1)}) defined by I(N,N+1)={f∈C([0,1],MN(N+1)):ev⁡0(f)∈λleft(MN),ev⁡1(f)∈λright(MN+1)}.I(N, N+1) = \bigl\{ f \in C([0, 1], M_{N(N+1)}) : \operatorname{ev}_0(f) \in \lambda_{\mathrm{left}}(M_N),\; \operatorname{ev}_1(f) \in \lambda_{\mathrm{right}}(M_{N+1}) \bigr\}.(1)✦

When SS is a subset of a topological space XX, the vanishing subalgebra C0(X,S)C_0(X, S) is the non-unital **-subalgebra of C(X,ℂ)C(X, \mathbb{C}) consisting of all continuous complex-valued functions g∈C(X,ℂ)g \in C(X, \mathbb{C}) such that g(s)=0g(s) = 0 for every s∈Ss \in S.✦ For a topological space XX and a continuous function p∈C(X,ℂ)p \in C(X, \mathbb{C}) with zero set Z(p)={s∈X:p(s)=0}Z(p) = \{s \in X : p(s) = 0\}, the multiplication map Mp:C(X,ℂ)→C0(X,Z(p))M_p : C(X, \mathbb{C}) \to C_0(X, Z(p)) is the ℂ\mathbb{C}-linear map defined by Mp(f)=pfM_p(f) = pf.✦

On [0,1][0, 1], the interval coordinate function t∈C([0,1],ℂ)t \in C([0, 1], \mathbb{C}) is the map s↦ss \mapsto s✦, and the interval weight function w∈C([0,1],ℂ)w \in C([0, 1], \mathbb{C}) is the continuous function (1−t)t(1 - t)t. The algebra of endpoint-vanishing scalar functions C0([0,1],Z(w))C_0([0, 1], Z(w)) on [0,1][0, 1] is the non-unital **-subalgebra C0([0,1],{0,1})C_0([0, 1], \{0, 1\}) of C([0,1],ℂ)C([0, 1], \mathbb{C}), and for an integer N≥1N \ge 1, the interior drop-section algebra ℐN\mathcal{I}_N is the matrix algebra MN(N+1)(C0([0,1],{0,1}))M_{N(N+1)}(C_0([0, 1], \{0, 1\})) indexed by {1,…,N}×{1,…,N+1}\{1, \dots, N\} \times \{1, \dots, N+1\}.

On I(N,N+1)I(N, N+1) from (1), the left endpoint evaluation map eleft:I(N,N+1)→MNe_{\mathrm{left}} : I(N, N+1) \to M_N and the right endpoint evaluation map eright:I(N,N+1)→MN+1e_{\mathrm{right}} : I(N, N+1) \to M_{N+1} are the ℂ\mathbb{C}-linear maps defined by eleft(f)ij=f(0)(i,1),(j,1)(1≤i,j≤N),eright(f)kl=f(1)(1,k),(1,l)(1≤k,l≤N+1).e_{\mathrm{left}}(f)_{ij} = f(0)_{(i,1),(j,1)} \quad (1 \le i, j \le N), \qquad e_{\mathrm{right}}(f)_{kl} = f(1)_{(1,k),(1,l)} \quad (1 \le k, l \le N+1).(2)

The constant matrix path embedding c:MN(N+1)→MN(N+1)(C([0,1],ℂ))c : M_{N(N+1)} \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the unital **-homomorphism sending each matrix aa to the constant path s↦as \mapsto a✦, the left constant path embedding cleft:MN→MN(N+1)(C([0,1],ℂ))c_{\mathrm{left}} : M_N \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the unital **-homomorphism c∘λleftc \circ \lambda_{\mathrm{left}}, and the right constant path embedding cright:MN+1→MN(N+1)(C([0,1],ℂ))c_{\mathrm{right}} : M_{N+1} \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the unital **-homomorphism c∘λrightc \circ \lambda_{\mathrm{right}}. For each p∈C([0,1],ℂ)p \in C([0, 1], \mathbb{C}), entrywise scalar multiplication Tp:MN(N+1)(C([0,1],ℂ))→MN(N+1)(C([0,1],ℂ))T_p : M_{N(N+1)}(C([0, 1], \mathbb{C})) \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the ℂ\mathbb{C}-linear endomorphism (muv)↦(pmuv)(m_{uv}) \mapsto (p\, m_{uv})✦, the left interpolated boundary map Lleft:MN→MN(N+1)(C([0,1],ℂ))L_{\mathrm{left}} : M_N \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the ℂ\mathbb{C}-linear map T1−t∘cleftT_{1-t} \circ c_{\mathrm{left}}, and the right interpolated boundary map Lright:MN+1→MN(N+1)(C([0,1],ℂ))L_{\mathrm{right}} : M_{N+1} \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the ℂ\mathbb{C}-linear map Tt∘crightT_t \circ c_{\mathrm{right}}.

Using (2), the drop-section residual map r:I(N,N+1)→MN(N+1)(C([0,1],ℂ))r : I(N, N+1) \to M_{N(N+1)}(C([0, 1], \mathbb{C})) is the ℂ\mathbb{C}-linear map defined by r(f)=f−Lleft(eleft(f))−Lright(eright(f)),r(f) = f - L_{\mathrm{left}}(e_{\mathrm{left}}(f)) - L_{\mathrm{right}}(e_{\mathrm{right}}(f)),(3)✦ where ff is identified entrywise in MN(N+1)(C([0,1],ℂ))M_{N(N+1)}(C([0, 1], \mathbb{C})). The interior drop-section map rint:I(N,N+1)→ℐNr_{\mathrm{int}} : I(N, N+1) \to \mathcal{I}_N is the ℂ\mathbb{C}-linear map assigning to each f∈I(N,N+1)f \in I(N, N+1) the matrix rint(f)r_{\mathrm{int}}(f) whose (u,v)(u, v)-entry is r(f)uv∈C0([0,1],{0,1})r(f)_{uv} \in C_0([0, 1], \{0, 1\}) from (3).✦

Weighted Matrix Units and Commuting Pairs

A system of weighted matrix units WW in a C*-algebra AA indexed by a finite set ι\iota consists of a weight element h∈Ah \in A satisfying 0≤h0 \le h and ∥h∥≤1\|h\| \le 1, together with a family of elements (xij)i,j∈ι(x_{ij})_{i,j \in \iota} in AA satisfying, for all i,j,k,l∈ιi, j, k, l \in \iota, xij*=xji,xijxkl=δjkhxil,∑m∈ιxmm=h.x_{ij}^* = x_{ji}, \qquad x_{ij} x_{kl} = \delta_{jk}\, h x_{il}, \qquad \sum_{m \in \iota} x_{mm} = h.(4)✦

Letting Mι(ℂ)M_\iota(\mathbb{C}) denote the C*-algebra of complex matrices indexed by a finite set ι\iota, the matrix linear map ϕW:Mι(ℂ)→A\phi_W : M_\iota(\mathbb{C}) \to A of a system of weighted matrix units W=(h,(xij)i,j∈ι)W = (h, (x_{ij})_{i,j \in \iota}) in AA satisfying (4) is the ℂ\mathbb{C}-linear map defined by ϕW(m)=∑i,j∈ιmijxij.\phi_W(m) = \sum_{i, j \in \iota} m_{ij}\, x_{ij}.(5)✦

A commuting weighted unit pair PP in a C*-algebra AA over finite index sets ι\iota and κ\kappa consists of a left system of weighted matrix units Wleft=(hleft,(xijleft)i,j∈ι)W_{\mathrm{left}} = (h_{\mathrm{left}}, (x^{\mathrm{left}}_{ij})_{i,j \in \iota}) indexed by ι\iota and a right system of weighted matrix units Wright=(hright,(xklright)k,l∈κ)W_{\mathrm{right}} = (h_{\mathrm{right}}, (x^{\mathrm{right}}_{kl})_{k,l \in \kappa}) indexed by κ\kappa in AA such that xijleftxklright=xklrightxijleftx^{\mathrm{left}}_{ij} x^{\mathrm{right}}_{kl} = x^{\mathrm{right}}_{kl} x^{\mathrm{left}}_{ij} for all i,j∈ιi, j \in \iota and k,l∈κk, l \in \kappa, and hleft+hright=1A.h_{\mathrm{left}} + h_{\mathrm{right}} = 1_A.(6)

Continuous Field Representations and the Linear Lift

When h∈Ah \in A satisfies 0≤h0 \le h and ∥h∥≤1\|h\| \le 1, the continuous map Re⁡:σ(h)→[0,1]\operatorname{Re} : \sigma(h) \to [0, 1] on the spectrum σ(h)\sigma(h) of hh is given by z↦Re⁡(z)z \mapsto \operatorname{Re}(z), and the interval functional calculus homomorphism θh:C([0,1],ℂ)→A\theta_h : C([0, 1], \mathbb{C}) \to A is the unital **-homomorphism obtained by composing the continuous functional calculus of hh with pullback along Re\operatorname{Re}.

Continuous field representation data DD on a compact Hausdorff space XX over a finite index set ι\iota in a C*-algebra AA consists of a self-adjoint function p∈C(X,ℂ)p \in C(X, \mathbb{C}), a unital **-homomorphism eval:C(X,ℂ)→A\mathrm{eval} : C(X, \mathbb{C}) \to A, and a family (xij)i,j∈ι(x_{ij})_{i,j \in \iota} in AA satisfying, for all i,j,k,l∈ιi, j, k, l \in \iota and all f∈C(X,ℂ)f \in C(X, \mathbb{C}), xij*=xji,xijxkl=δjkeval(p)xil,xijeval(f)=eval(f)xij,∥xijeval(f)∥≤∥pf∥.x_{ij}^* = x_{ji}, \quad x_{ij} x_{kl} = \delta_{jk}\,\mathrm{eval}(p)\,x_{il}, \quad x_{ij}\,\mathrm{eval}(f) = \mathrm{eval}(f)\,x_{ij}, \quad \|x_{ij}\,\mathrm{eval}(f)\| \le \|p f\|.(7)✦ For continuous field representation data D=(p,eval,(xij)i,j∈ι)D = (p, \mathrm{eval}, (x_{ij})_{i,j \in \iota}) on XX over ι\iota in AA, the vanishing domain of DD is the non-unital **-subalgebra C0(X,Z(p))C_0(X, Z(p)) of C(X,ℂ)C(X, \mathbb{C}), and the coefficient matrix algebra of DD is the matrix algebra Mι(C0(X,Z(p)))M_\iota(C_0(X, Z(p))).

The elementary field coefficient map αij:C(X,ℂ)→A\alpha_{ij} : C(X, \mathbb{C}) \to A for indices i,j∈ιi, j \in \iota and continuous field representation data D=(p,eval,(xij)i,j∈ι)D = (p, \mathrm{eval}, (x_{ij})_{i,j \in \iota}) satisfying (7) is the ℂ\mathbb{C}-linear map f↦xijeval(f)f \mapsto x_{ij}\,\mathrm{eval}(f)✦, and the continuous field coefficient map cij:C0(X,Z(p))→Ac_{ij} : C_0(X, Z(p)) \to A is the continuous ℂ\mathbb{C}-linear map obtained as the bounded linear extension of αij\alpha_{ij} along the multiplier map MpM_p.✦ The field matrix representation map ΨD:Mι(C0(X,Z(p)))→A\Psi_D : M_\iota(C_0(X, Z(p))) \to A of continuous field representation data D=(p,eval,(xij)i,j∈ι)D = (p, \mathrm{eval}, (x_{ij})_{i,j \in \iota}) on XX over ι\iota in AA is the ℂ\mathbb{C}-linear map defined by ΨD(m)=∑i,j∈ιcij(mij).\Psi_D(m) = \sum_{i, j \in \iota} c_{ij}(m_{ij}).(8)✦

For a commuting weighted unit pair PP over ι\iota and κ\kappa in AA as in (6), the pair weight function wP∈C([0,1],ℂ)w_P \in C([0, 1], \mathbb{C}) is the continuous function ww, and the pair evaluation homomorphism evalP:C([0,1],ℂ)→A\mathrm{eval}_P : C([0, 1], \mathbb{C}) \to A is the unital **-homomorphism θhright\theta_{h_{\mathrm{right}}}. The continuous field representation data DPD_P of a commuting weighted unit pair PP over ι\iota and κ\kappa in AA on [0,1][0, 1] over ι×κ\iota \times \kappa is the continuous field representation data with weight function wPw_P, evaluation homomorphism evalP\mathrm{eval}_P, and generators x(i,k),(j,l)=xijleftxklrightx_{(i,k),(j,l)} = x^{\mathrm{left}}_{ij}\,x^{\mathrm{right}}_{kl} for (i,k),(j,l)∈ι×κ(i, k), (j, l) \in \iota \times \kappa.✦

For a commuting weighted unit pair PP in AA of ranks NN and N+1N+1 with ι={1,…,N}\iota = \{1, \dots, N\} and κ={1,…,N+1}\kappa = \{1, \dots, N+1\}, the coefficient matrix algebra of DPD_P is ℐN\mathcal{I}_N, the left matrix linear map ϕleft:MN→A\phi_{\mathrm{left}} : M_N \to A is the ℂ\mathbb{C}-linear map ϕWleft\phi_{W_{\mathrm{left}}}, and the right matrix linear map ϕright:MN+1→A\phi_{\mathrm{right}} : M_{N+1} \to A is the ℂ\mathbb{C}-linear map ϕWright\phi_{W_{\mathrm{right}}}. The linear lift ΦP:I(N,N+1)→A\Phi_P : I(N, N+1) \to A of PP is the ℂ\mathbb{C}-linear map defined using (2), (3), and (8) by ΦP(f)=ϕleft(eleft(f))+ϕright(eright(f))+ΨDP(rint(f)).\Phi_P(f) = \phi_{\mathrm{left}}(e_{\mathrm{left}}(f)) + \phi_{\mathrm{right}}(e_{\mathrm{right}}(f)) + \Psi_{D_P}(r_{\mathrm{int}}(f)).(9)✦

The Claim

Theorem 1. Let N≥1N \ge 1 be a positive integer, let AA be a complex unital C*-algebra, and let PP be a commuting weighted unit pair of ranks NN and N+1N+1 in AA. Then the linear lift ΦP\Phi_P from (9) defines a unital **-homomorphism πP:I(N,N+1)→A,πP(f)=ΦP(f),\pi_P : I(N, N+1) \to A, \qquad \pi_P(f) = \Phi_P(f),(10)✦ and for all f,g∈I(N,N+1)f, g \in I(N, N+1), ΦP(fg)=ΦP(f)ΦP(g).\Phi_P(fg) = \Phi_P(f)\Phi_P(g).(11)✦

In the paper, the result reads: if α:M3→B\alpha : M_3 \to B and β:M2→B\beta : M_2 \to B are completely positive contractive order-zero maps with commuting ranges satisfying α(13)+β(12)=1B\alpha(1_3) + \beta(1_2) = 1_B, where BB stands for the C*-algebra AA, then there is a unital **-homomorphism I(2,3)→BI(2, 3) \to B. In (10) and (11), the theorem generalizes (2,3)(2, 3) to (N,N+1)(N, N+1) for all N≥1N \ge 1 and starts directly from a commuting weighted unit pair of ranks NN and N+1N+1 in AA, with the left system corresponding to β\beta and the right system to α\alpha when N=2N = 2.

What Is Checked

The theorem in (10) and (11) is stated over a positive integer N≥1N \ge 1, a complex unital C*-algebra AA, the dimension-drop algebra I(N,N+1)I(N, N+1), and a commuting weighted unit pair PP of ranks NN and N+1N+1 in AA, with linear lift ΦP\Phi_P built from the endpoint evaluations elefte_{\mathrm{left}} and erighte_{\mathrm{right}}, the interior drop-section map rintr_{\mathrm{int}}, and the continuous field representation DPD_P. For these constructions, the library also states that the weight function wPw_P of any commuting weighted unit pair PP is self-adjoint in C([0,1],ℂ)C([0, 1], \mathbb{C}).✦