OctoInformBeta

OpenAI preprint, September 2026 · §3. Hyperbolicity and two cone slices

§3: Rank-one identity (3.1), positivity, and scaling (3.2): Order preservation on ℝ4×4\mathbb{R}^{4 \times 4} and positive semidefiniteness on real and complex matrices for y∈ℝ3y \in \mathbb{R}^3

Rank-one identity and quadratic scaling of Φy\Phi_y..

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

Consider 4×44 \times 4 matrices over the real and complex numbers under the positive semidefinite order. A family of maps on 4×44 \times 4 matrices, parameterized by three real variables via a 3×33 \times 3 quadratic matrix alongside the Choi-Lam biquadratic form, acts on both rank-one and general matrices. The question is how the quadratic form of each map on a rank-one matrix evaluates as a Choi-Lam form. One also asks whether the maps preserve positive semidefiniteness and matrix order, and how they scale with the parameters.

Definitions

For a natural number nn, Euclidean space ℝn\mathbb{R}^n is the space of nn-tuples of real numbers equipped with the standard inner product, and complex Euclidean space ℂn\mathbb{C}^n is the space of nn-tuples of complex numbers.✦ The space Rn×nR^{n \times n} of n×nn \times n square matrices over a set RR consists of the n×nn \times n matrices with entries in RR, and for a matrix M∈Rn×nM \in R^{n \times n}, the trace of MM is denoted by tr⁡(M)\operatorname{tr}(M).✦ For a commutative ring RR, the space R3R^3 consists of the triples with entries in RR indexed cyclically by {1,2,3}\{1, 2, 3\} (so 3+1=13 + 1 = 1), and the space S4S^4 is the subspace of real symmetric 4×44 \times 4 matrices in ℝ4×4\mathbb{R}^{4 \times 4}.

A real matrix X∈ℝn×nX \in \mathbb{R}^{n \times n} is positive semidefinite, written X≽0X \succeq 0, if X⊤=XX^\top = X and w⊤Xw≥0w^\top X w \ge 0 for every w∈ℝnw \in \mathbb{R}^n. The Loewner order ≼\preceq on ℝn×n\mathbb{R}^{n \times n} is the relation defined by A≼BA \preceq B when B−A≽0B - A \succeq 0 for matrices A,B∈ℝn×nA, B \in \mathbb{R}^{n \times n}. A complex matrix X∈ℂn×nX \in \mathbb{C}^{n \times n} is positive semidefinite, written X≽0X \succeq 0, if X*=XX^* = X and w*Xw≥0w^* X w \ge 0 for every w∈ℂnw \in \mathbb{C}^n, where X*X^* and w*w^* denote conjugate transposes.

Given a natural number nn and a vector v∈ℝnv \in \mathbb{R}^n, the outer product vv⊤∈ℝn×nv v^\top \in \mathbb{R}^{n \times n} is the n×nn \times n real matrix whose entries are given by (1): (vv⊤)ij=vivj.(v v^\top)_{ij} = v_i v_j.(1)✦

For a vector u∈ℝ4u \in \mathbb{R}^4, the block decomposition u=(u0,u′)u = (u_0, u') consists of the initial entry u0∈ℝu_0 \in \mathbb{R} and trailing vector u′=(u1,u2,u3)∈ℝ3u' = (u_1, u_2, u_3) \in \mathbb{R}^3. For vectors v=(v0,v′)v = (v_0, v') and u=(u0,u′)u = (u_0, u') in ℝ4\mathbb{R}^4, the wedge coordinates of vv and uu are the vector v0u′−u0v′∈ℝ3v_0 u' - u_0 v' \in \mathbb{R}^3 whose ii-th entry for i∈{1,2,3}i \in \{1, 2, 3\} is defined by (2): (v0u′−u0v′)i=v0ui−u0vi.(v_0 u' - u_0 v')_i = v_0 u_i - u_0 v_i.(2)✦

When RR is a commutative ring and y=(y1,y2,y3)∈R3y = (y_1, y_2, y_3) \in R^3, the quadratic matrix at yy is the 3×33 \times 3 matrix Q(y)∈R3×3Q(y) \in R^{3 \times 3} defined by (3): Q(y)=(y12+y22−y1y2−y1y3−y2y1y22+y32−y2y3−y3y1−y3y2y32+y12).Q(y) = \begin{pmatrix} y_1^2 + y_2^2 & -y_1 y_2 & -y_1 y_3 \\ -y_2 y_1 & y_2^2 + y_3^2 & -y_2 y_3 \\ -y_3 y_1 & -y_3 y_2 & y_3^2 + y_1^2 \end{pmatrix}.(3)✦

The Choi-Lam form for triples z,y∈R3z, y \in R^3 over a commutative ring RR is the biquadratic polynomial b(z,y)∈Rb(z, y) \in R given by (4): b(z,y)=∑i=13zi2yi2−2(z1y1z2y2+z1y1z3y3+z2y2z3y3)+∑i=13zi2yi+12.b(z, y) = \sum_{i=1}^3 z_i^2 y_i^2 - 2(z_1 y_1 z_2 y_2 + z_1 y_1 z_3 y_3 + z_2 y_2 z_3 y_3) + \sum_{i=1}^3 z_i^2 y_{i+1}^2.(4)✦

For a commutative ring RR and y∈R3y \in R^3, the matrix map Φy:R4×4→R4×4\Phi_y : R^{4 \times 4} \to R^{4 \times 4} sends any 4×44 \times 4 matrix X∈R4×4X \in R^{4 \times 4}, written in (1+3)×(1+3)(1 + 3) \times (1 + 3) block form as X=(ar⊤sT)X = \bigl(\begin{smallmatrix} a & r^\top \\ s & T \end{smallmatrix}\bigr) with a∈Ra \in R, column vectors r,s∈R3r, s \in R^3, and 3×33 \times 3 lower-right submatrix T∈R3×3T \in R^{3 \times 3} on the trailing indices {1,2,3}\{1, 2, 3\}, to the 4×44 \times 4 matrix defined by (5): Φy(ar⊤sT)=(tr⁡(Q(y)T)−r⊤Q(y)−Q(y)saQ(y)).\Phi_y \begin{pmatrix} a & r^\top \\ s & T \end{pmatrix} = \begin{pmatrix} \operatorname{tr}(Q(y) T) & -r^\top Q(y) \\ -Q(y) s & a Q(y) \end{pmatrix}.(5)✦

The Claim

Theorem 1. Let y∈ℝ3y \in \mathbb{R}^3.

  1. Vectors u=(u0,u′)u = (u_0, u') and v=(v0,v′)v = (v_0, v') in ℝ4\mathbb{R}^4 satisfy (6): u⊤Φy(vv⊤)u=b(v0u′−u0v′,y).u^\top \Phi_y(v v^\top) u = b(v_0 u' - u_0 v', y).(6)✦

  2. If X∈ℝ4×4X \in \mathbb{R}^{4 \times 4} satisfies X≽0X \succeq 0, then (7) gives Φy(X)≽0.\Phi_y(X) \succeq 0.(7)✦

  3. For A,B∈ℝ4×4A, B \in \mathbb{R}^{4 \times 4} with A≼BA \preceq B, the Loewner comparison (8) holds: Φy(A)≼Φy(B).\Phi_y(A) \preceq \Phi_y(B).(8)✦

  4. Rescaling yy by s∈ℝs \in \mathbb{R} acts on any X∈ℝ4×4X \in \mathbb{R}^{4 \times 4} via (9): Φsy(X)=s2Φy(X).\Phi_{s y}(X) = s^2 \Phi_y(X).(9)✦

  5. Whenever X∈ℂ4×4X \in \mathbb{C}^{4 \times 4} satisfies X≽0X \succeq 0, the complex matrix image obeys (10): Φy(X)≽0.\Phi_y(X) \succeq 0.(10)✦

In the paper, the result reads: for y∈ℝ3y \in \mathbb{R}^3 and vectors u=(u0,u′)u = (u_0, u') and v=(v0,v′)v = (v_0, v') in ℝ4\mathbb{R}^4, u⊤Φy(vv⊤)u=b(v0u′−u0v′,y)u^\top \Phi_y(v v^\top) u = b(v_0 u' - u_0 v', y); hence Φy\Phi_y is a positive linear map on S4S^4 (preserving ≽\succeq) and Φsy=s2Φy\Phi_{s y} = s^2 \Phi_y for s∈ℝs \in \mathbb{R}. In the theorem, (7), (8), and (9) hold on all of ℝ4×4\mathbb{R}^{4 \times 4} rather than only on S4S^4, while (10) applies to ℂ4×4\mathbb{C}^{4 \times 4}.

What Is Checked

The claims are stated for y∈ℝ3y \in \mathbb{R}^3, s∈ℝs \in \mathbb{R}, and u,v∈ℝ4u, v \in \mathbb{R}^4 in terms of the quadratic matrix Q(y)Q(y), the matrix map Φy\Phi_y, the Choi-Lam form bb, the outer product vv⊤v v^\top, and the wedge coordinates v0u′−u0v′v_0 u' - u_0 v', over positive semidefinite matrices X∈ℝ4×4X \in \mathbb{R}^{4 \times 4} and X∈ℂ4×4X \in \mathbb{C}^{4 \times 4} as well as pairs A≼BA \preceq B in ℝ4×4\mathbb{R}^{4 \times 4}.