Quantum magic squares have recently been introduced as a “magical” combination of quantum measurements. In contrast to quantum measurements, they cannot be purified (i.e., dilated to a quantum permutation matrix)—only the so-called semiclassical ones can. Purifying establishes a relation to an ideal world of fundamental theoretical and practical importance; the opposite of purifying is described by the matrix convex hull. In this paper, we prove that semiclassical magic squares can be purified to quantum Latin squares, which are “magical” combinations of orthonormal bases. Conversely, we prove that the matrix convex hull of quantum Latin squares is larger compared to the semiclassical ones. This tension is resolved by our third result: we prove that the quantum Latin squares that are semiclassical are precisely those constructed from a classical Latin square. Our work sheds light on the internal structure of quantum magic squares, on how this is affected by the matrix convex hull, and, more generally, on the nature of the “magical” composition rule, both at the semiclassical and at the quantum level.
Quantum permutation matrices and quantum magic squares are generalizations of permutation matrices and magic squares, where the entries are no longer numbers but elements from arbitrary (non-commutative) algebras. The famous Birkhoff–von Neumann theorem characterizes magic squares as convex combinations of permutation matrices. In the non-commutative case, the corresponding question is as follows: Does every quantum magic square belong to the matrix convex hull of quantum permutation matrices? That is, does every quantum magic square dilate to a quantum permutation matrix? Here, we show that this is false even in the simplest non-commutative case. We also classify the quantum magic squares that dilate to a quantum permutation matrix with commuting entries and prove a quantitative lower bound on the diameter of this set. Finally, we conclude that not all Arveson extreme points of the free spectrahedron of quantum magic squares are quantum permutation matrices.
Quantum Latin Squares have emerged as mathematical structures of interest in quantum information theory. This paper provides a comprehensive overview of quantum Latin squares. We aim to synthesize the current knowledge for the benefit of researchers in quantum information theory. We begin by establishing the theoretical foundation linking classical Latin squares with their quantum counterparts. This provides insight into their potential applications, particularly in the realm of quantum error correction. Through a detailed examination, we outline known constructions and elucidate the properties of quantum Latin squares, highlighting their role within quantum computing frameworks. We then explore the relationships of quantum Latin squares with mutually unbiased bases and unitary error bases. This underscores their importance in the construction of quantum error correcting codes using mutually orthogonal quantum Latin squares. We give significant focus to a comparative analysis between classical and quantum Latin squares, particularly in their applications to error correcting codes. The applications are analogous to the application of classical Latin squares to classical error correcting codes. This overview aims to not only compile existing research but also to inspire further exploration into Quantum Latin squares and their applications in quantum information theory.
A quantum Latin square is an
n×n array of unit vectors where each row and column forms an orthonormal basis of a fixed complex vector space. We introduce the notion of
(G,G′)-invariant quantum Latin squares for finite groups
G and
G′. These are quantum Latin squares with rows and columns indexed by
G and
G′ respectively such that the inner product of the
a,b-entry with the
c,d-entry depends only on
a−1c∈G and
b−1d∈G′. This definition is motivated by the notion of group invariant bijective correlations introduced in [Roberson \& Schmidt (2020)], and every group invariant quantum Latin square produces a group invariant bijective correlation, though the converse does not hold. In this work we investigate these group invariant quantum Latin squares and their corresponding correlations. Our main result is that, up to applying a global isometry to every vector in a
(G,G′)-invariant quantum Latin square, there is a natural bijection between these objects and trace and conjugate transpose preserving isomorphisms between the group algebras of
G and
G′. This in particular proves that a
(G,G′)-invariant quantum Latin square exists if and only if the multisets of degrees of irreducible representations are equal for
G and
G′. Another motivation for this line of work is that whenever Cayley graphs for groups
G and
G′ are quantum isomorphic, then there is a
(G,G′)-invariant quantum correlation witnessing this, and thus it suffices to consider such correlations when searching for quantum isomorphic Cayley graphs. Given a group invariant quantum correlation, we show how to construct all pairs of graphs for which it gives a quantum isomorphism.
Goyeneche et al recently proposed a notion of orthogonality for quantum Latin squares, and showed that orthogonal quantum Latin squares yield quantum codes. We give a simplified characterization of orthogonality for quantum Latin squares, which we show is equivalent to the existing notion. We use this simplified characterization to give an upper bound for the number of mutually orthogonal quantum Latin squares of a given size, and to give the first examples of orthogonal quantum Latin squares that do not arise from ordinary Latin squares. We then discuss quantum Latin isometry squares, generalizations of quantum Latin squares recently introduced by Benoist and Nechita, and define a new orthogonality property for these objects, showing that it also allows the construction of quantum codes. We give a new characterization of unitary error bases using these structures.
In this paper we introduce quantum Latin squares, combinatorial quantum objects which generalize classical Latin squares, and investigate their applications in quantum computer science. Our main results are on applications to unitary error bases (UEBs), basic structures in quantum information which lie at the heart of procedures such as teleportation, dense coding and error correction. We present a new method for constructing a UEB from a quantum Latin square equipped with extra data. Developing construction techniques for UEBs has been a major activity in quantum computation, with three primary methods proposed: shift-and-multiply, Hadamard, and group theoretic. We show that our new approach simultaneously generalizes the shift-and multiply and Hadamard methods. Furthermore, we explicitly construct a UEB using our technique which we prove cannot be obtained from any of these existing methods.
We introduce orthogonal quantum Latin squares, which restrict to traditional orthogonal Latin squares, and investigate their application in quantum information science. We use quantum Latin squares to build maximally entangled bases, and show how a pair of mutually unbiased maximally entangled bases can be constructed in square dimension from orthogonal quantum Latin squares. We also compare our construction to an existing construction due to Beth and Wocjan and show that ours is strictly more general.