Joint numerical ranges of three 4 × 4 4\times 4 matrices

Piotr Pikul, Ilya Spitkovsky, Konrad Szymański, Stephan Weis, Karol Życzkowski · 2025

Read Joint numerical ranges of three 4 × 4 4\times 4 matrices by Piotr Pikul, Ilya Spitkovsky, Konrad Szymański, Stephan Weis, Karol Życzkowski (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2510-27670.txt.

Introduction

Piotr Pikul, Ilya Spitkovsky, Konrad Szymański, Stephan Weis, and Karol Życzkowski analyze the joint numerical range of three Hermitian matrices of size four in 2025. The range is the set of all triple averages a quantum state can give for three traits. In the usual case, this convex set, which contains all blends of its points, has a smooth edge.

The team sorts the odd cases by how many non-elliptic faces the edge holds, finding fifteen classes with an example for each. They also show that a shared point of three distinct line-like faces must be a corner. A split form of the space then defines a separable range used in entanglement studies, and its edge is compared with the full range.

Catalog record

CREATORPiotr Pikul, Ilya Spitkovsky, Konrad Szymański, Stephan Weis, Karol Życzkowski
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE97 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2510-27670

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Joint numerical ranges of three 4 × 4 4\times 4 matrices Piotr Pikul, Ilya Spitkovsky, Konrad Szymański, Stephan Weis, Karol Życzkowski Abstract . We analyze the joint numerical range W W of three complex hermitian matrices of order four. In the generic case this 3 ​ D 3D convex set has a smooth boundary. We analyze non-generic structures and investigate non-elliptic faces in the boundary ∂ W \partial W . Fifteen possible classes regarding the numbers of non-elliptic faces are identified and an explicit example is presented for each class. Secondly, it is shown that a nonempty intersection of three mutually distinct one-dimensional faces is a corner point. Thirdly, introducing a tensor product structure into ℂ 4 = ℂ 2 ⊗ ℂ 2 \mathbb{C}^{4}=\mathbb{C}^{2}\otimes\mathbb{C}^{2} , one defines the separable joint numerical range — a subset of W W useful in studies of quantum entanglement. The boundary of the separable joint numerical range is compared with that of W W . 1991 Mathematics Subject Classification 15A60, 52A15, 47L07, 52A20 1. Introduction The joint numerical range of k hermitian n\times n matrices A_{1},\dots,A_{k} is the compact convex subset W:=W(A_{1},\dots,A_{k}):=\big\{(\tr\rho A_{1},\ldots,\tr\rho A_{k}){}^{\top}\colon\rho\in\mathrm{M}_{n},\,\rho\geq 0,\,\tr\rho=1\big\} of \mathbb{R}^{k} , where \mathrm{M}_{n} is the set of n\times n matrices with complex entries, \tr(X) is the trace of X\in\mathrm{M}_{n} , and X\geq 0 means that X is positive semidefinite. The numerical range of B\in\mathrm{M}_{n} is the joint numerical range of the hermitian matrices \real B:=\frac{1}{2}(B+B^{\ast}) and \imaginary B:=\frac{1}{2\mathrm{i}}(B-B^{\ast}) understood as the subset \widehat{W}(B)=\{x_{1}+\mathrm{i}x_{2}\colon(x_{1},x_{2}){}^{\top}\in W(\real B,\imaginary B)\} of the complex plane \mathbb{C} . We tacitly identify \mathbb{R}^{2}\cong\mathbb{C} , see Remark 2.10 for more details. The numerical range of an operator is a well-known notion studied in operator theory and linear algebra [ 60 , 28 , 37 , 18 , 23 , 24 , 32 ] , while its generalizations find several applications in theoretical physics [ 56 , 38 , 16 , 35 ] . The joint numerical range is well known in matrix theory [ 1 , 8 , 9 , 12 , 25 , 39 , 45 , 51 , 59 , 61 ] and theoretical physics [ 26 , 65 , 58 ] . It is also known as the set of states [ 48 ] of a matrix system [ 13 , 4 ] , an algebraic object that has numerous applications in quantum information theory, see for example [ 29 , 49 ] . The subset W of \mathbb{R}^{k} has a clear interpretation in quantum physics: It is the set of possible expectation values of joint measurements of k observables A_{1},\dots,A_{k} . The case of k=3 hermitian matrices of size n=3 was studied in [ 59 ] , in which all possible shapes of W were divided into 10 classes with respect to the structure of its boundary. Here, we shall go beyond this case and analyze the problem of three hermitian matrices of size n=4 . This step forward has important consequences for quantum theory: As four is a composite number, one can introduce a tensor product structure, \mathbb{C}^{4}=\mathbb{C}^{2}\otimes\mathbb{C}^{2} , which corresponds to a two-qubit system. Such a splitting of a four-dimensional system into two subsystems allows for the introduction of product states of a physical systems, which convey no correlations between the two parts, and separable states , which encode classical correlations – all the other states are entangled , and their correlations can never be explained by classical probability theory. By restricting the range of states in the definition of numerical range one arrives at product [ 52 ] and separable numerical range [ 58 ] , which physically correspond to classically achievable sets of expectation values. The geometry of bipartite quantum systems is also studied outside the algebraic setting of numerical ranges, see for instance [ 27 , 40 , 44 , 54 ] . This work investigates the boundary of the joint numerical range W=W(A_{1},A_{2},A_{3}) of three hermitian matrices A_{1},A_{2},A_{3} of size n=4 . Since k\leq 3 , the von Neumann-Wigner non-crossing theorem implies that the boundary \partial W is generically smooth and strictly convex, i.e. the set of triples with a smooth joint numerical range is open and dense in the set of all triples of complex hermitian 4\times 4 matrices with the Euclidean topology, see [ 25 , Proposition 4.9] and [ 59 , Theorem 4.2] . In the analogous sense of being true for triples of real symmetric 4\times 4 matrices in an open and dense set, W has generically an even number of at most ten elliptic disks on the boundary and it has no other flat portions on \partial W . This follows from [ 47 , Theorem 1.1] , see also [ 30 ] , by duality as described in Remark 6.3 below. In this paper we focus on non-elliptic flat portions on \partial W . Any flat portion on \partial W corresponds to the numerical range of a 3\times 3 matrix, see Remark 2.3 (c), (d) for details. The classification of these numerical ranges was achieved by Kippenhahn [ 37 ] . See also [ 36 , 20 ] and, for more rigorous proofs, see [ 12 , 51 ] . Let (1.1) p_{B}(u_{0},u_{1},u_{2}):=\det(u_0\id_3+u_1\Re B+u_2\Im B),\qquad u_{0},u_{1},u_{2}\in\mathbb{C}, and let C(B) be the dual curve [ 19 ] to the curve p_{B}=0 that consists of all tangent lines to p_{B}=0 . Identifying the complex projective plane \mathbb{C}\mathbb{P}^{2} with its dual space of lines, we define C_{\mathbb{R}}(B):=\{(x_{1},x_{2}){}^{\top}\in\mathbb{R}^{2}\mid(1:x_{1}:x_{2})\in C(B)\} . The set C_{\mathbb{R}}(B) is known as the Kippenhahn curve [ 6 , 20 ] and \widehat{W}(B) is the convex hull of C_{\mathbb{R}}(B) . This result divides the shape of \widehat{W}(B) into four different classes [ 20 ] : (1) C_{\mathbb{R}}(B) consists of three points that are the eigenvalues of B , and \widehat{W}(B) is the closed triangular region whose vertices are these points. (2) C_{\mathbb{R}}(B) consists of an ellipse and a point. Depending on whether the point lies inside the ellipse or not, \widehat{W}(B) is an elliptic disk or it has a droplet shape . (3) C(B) is an irreducible curve of degree four that has a double tangent, C_{\mathbb{R}}(B) is of heart shape, and the boundary of \widehat{W}(B) contains a line segment but no corner. (4) C(B) is an irreducible curve of degree six, C_{\mathbb{R}}(B) consists of two parts, one inside the other. The outer part is an oval (a smooth and strictly convex curve), whose convex hull is \widehat{W}(B) . For simplicity (see Lemma 3.1 and Example 3.3 ) we focus on flat parts of the boundary \partial W different from elliptic disks. This prompts the following definitions. A face of a convex subset C of \mathbb{R}^{d} is a convex subset of C that contains the endpoints x,y\in C of every open segment \{(1-\lambda)x+\lambda y:\lambda\in(0,1)\} it intersects. An exposed face of C is the set of minimizers of any linear functional on C . For technical reasons, \emptyset is also considered an exposed face. It is well known that every exposed face of C is a face of C . Faces that are no exposed faces are called nonexposed faces , see [ 53 ] . Definition 1.1 (Non-elliptic faces) . Let A_{1},A_{2},A_{3} be hermitian 4\times 4 matrices such that W=W(A_{1},A_{2},A_{3}) has

Read the complete text