🔢 mathematics

A Complete Classification of Complex Hadamard Matrices of Order Six

This paper provides a complete and exact finite-incidence classification of complex Hadamard matrices of order six up to standard equivalence by proving that every such matrix can be algebraically reconstructed from a $3 \times 3$ corner, thereby resolving a decades-old open problem and confirming Szöllősi's conjecture regarding the structure of these matrices.

Original authors: Mateo Cárdenes Wuttig, Joseph Tindall

Published 2026-08-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Mateo Cárdenes Wuttig, Joseph Tindall

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the quantum world, where particles can exist in multiple states at once, scientists rely on special mathematical tools to describe how these particles interact and change. One such tool is a grid of numbers called a complex Hadamard matrix. Imagine a square grid where every single number has a size of exactly one, but they can point in different directions, like arrows spinning on a clock face. The most important rule for these grids is that if you compare any two rows, they are perfectly balanced against each other, canceling out in a way that ensures no information is lost when the system changes. These grids are the backbone of many quantum technologies, from the way we measure particles without biasing the result to the design of optical circuits that split and recombine light beams.

For decades, scientists have been trying to map out every possible version of these grids for different sizes. For small grids, the map was already complete. But when the grid size reached six, the task hit a wall. This specific size is unique because it is the first time that two very different types of these grids appear together: some that form smooth, continuous families where you can slide from one to another, and one that stands completely alone, isolated from the rest. For over thirty years, this six-by-six case remained a mystery, with researchers unable to prove whether they had found every single possibility or if some were hiding in the gaps.

A team of researchers has now solved this decades-old puzzle. They have produced a complete and exact list of every possible six-by-six complex Hadam matrix, proving that no others exist. Their work does not just list them; it provides a specific, step-by-step recipe for how to build any one of them. The researchers discovered that every single one of these grids, no matter how complex or isolated, can be constructed by starting with a small three-by-three corner piece and following a finite set of rules to fill in the rest. This finding confirms a long-standing guess made by a mathematician named Szöllösi, who had proposed that such a construction method was possible but could not prove it worked for every case.

The team's method is rigorous and exhaustive. They showed that if you pick any six-by-six grid that follows the rules, you can always find a specific three-by-three corner within it that acts as a key. Once you have this key, the rest of the grid is forced into place by a finite number of mathematical choices. There are no infinite loops or missing pieces. The researchers proved that the only exceptions to their general construction rules are two specific, well-known cases: a famous isolated grid discovered by Tao and a large family of grids described by Karlsson. Even for these two exceptions, the team showed that they too fit into the broader picture, with the isolated grid being the only one that cannot be reached by the standard continuous families.

This breakthrough changes how scientists can approach problems in quantum information. Because the researchers have mapped the entire space of these grids, they have removed the need to search through an unclassified, unknown territory. For example, in the design of balanced optical networks that split light into six paths, engineers now have a complete design space to work with. They know exactly which configurations are possible and which are not. Similarly, for the study of mutually unbiased bases—a fundamental concept in quantum mechanics that deals with how to measure a system in different, non-overlapping ways—this work provides a solid framework. Instead of wondering if a new type of measurement exists, scientists can now focus on the specific, known structures that the paper has fully classified.

The solution was achieved by developing a new algorithm that avoids the mathematical pitfalls that had stumped previous attempts. Earlier methods relied on dividing by numbers that could sometimes be zero, which caused valid solutions to be lost or ignored. The new approach uses a division-free method that keeps every possible solution, even the tricky ones where numbers vanish. By carefully tracking these edge cases, the team was able to prove that their method captures every single valid grid. They also used computer verification to double-check their logic, ensuring that the mathematical arguments hold up under the strictest scrutiny.

The result is a definitive classification that separates the known from the unknown. It shows that the landscape of these six-by-six grids is finite and fully understood. While the question of how many mutually unbiased bases exist in a six-dimensional space remains open, this work has cleared the path by defining the exact set of building blocks available. The researchers have turned a chaotic, decades-long search into a structured, complete map, giving the scientific community a reliable foundation for future experiments in quantum optics, error correction, and the fundamental study of quantum states.

Technical Summary: A Complete Classification of Complex Hadamard Matrices of Order Six

Problem Statement
Complex Hadamard matrices (CHMs) of order $n$ are square matrices with unimodular entries and mutually orthogonal rows. They are fundamental to quantum information theory, underpinning mutually unbiased bases (MUBs), multiphoton interferometry, and quantum walks. While the classification of CHMs is complete for orders $n \le 5$ , the case of order $n=6$ has remained an open problem for decades. Order six is the first dimension where multiple inequivalent continuous families coexist with a genuinely isolated solution. Previous attempts, notably Szöllösi's dilation algorithm, provided a method to reconstruct matrices from a $3 \times 3$ corner but faced two critical gaps: (1) the algorithm relied on generic polynomial divisions that could lose valid solutions on vanishing-denominator branches, and (2) it was unproven whether every order-six CHM possesses a "finite-corner witness" (a $3 \times 3$ submatrix with only finitely many compatible completions).

Methodology
The authors introduce a branch-complete finite-dilation procedure that resolves the limitations of previous methods. The core methodology involves:

Division-Free Formulation: Instead of using generic quotient formulas that discard solutions when denominators vanish, the authors formulate the problem using uncancelled fixed-Gram equations. This ensures that all physical solutions, including exceptional finite branches, are retained.
Global Routing Argument: The proof proceeds by examining all 400 possible positional $3 \times 3$ corners of an arbitrary order-six CHM. The authors prove that if a matrix does not have a finite-corner witness, it must belong to specific structural sectors (Karlsson's family or Tao's isolated matrix).
Structural Inputs: The proof relies on two previously established structural results:
- Karlsson's classification of $H_2$ -reducible matrices (those containing a $2 \times 2$ Hadamard submatrix).
- Szöllösi's criterion regarding matrices with rows and columns composed entirely of cubic roots of unity.
Algebraic Reconstruction: For matrices admitting a finite-corner witness, the reconstruction is reduced to solving a system of equations over a four-phase seed torus. The authors demonstrate that, outside specific exceptions, this reconstruction involves solving one quadratic and one cubic equation in both horizontal and vertical directions.

Key Contributions and Results

Complete Classification: The paper provides a complete and exact finite-incidence classification of all order-six complex Hadamard matrices up to standard equivalence. The set of all such matrices, denoted $\mathcal{H}_6$ , is shown to be exactly equal to the output of the branch-complete dilation procedure ( $A^{fc}_6$ ).
Proof of Szöllösi's Conjecture: The authors prove Szöllösi's Conjecture 4.2, demonstrating that every order-six CHM class outside the Karlsson family and Tao's isolated matrix can be recovered algebraically from a suitable $3 \times 3$ corner using the dilation method.
Finite-Corner Witness Theorem: It is proven that every order-six CHM is equivalent to a dephased matrix possessing at least one $3 \times 3$ corner with a finite set of compatible normalized side blocks. This closes the gap between Szöllösi's conditional fixed-corner result and a global classification.
Geometric Structure of Reconstruction:
- Regular Domain: For all classes except Tao's isolated matrix and a single explicit Karlsson matrix ( $H^\times$ ), there exists a "product-regular" frame where the reconstruction is governed by a quadratic-cubic system.
- Physical Domain: The physical solutions correspond to the region where the discriminant of the product quadratic is non-positive ( $\omega_n \le 0$ ). The region $\omega_n < 0$ yields two distinct physical sheets, while $\omega_n = 0$ represents a ramification hypersurface where sheets coalesce.
- Exceptions: The only classes lacking a product-regular frame are Tao's isolated matrix and the specific Karlsson class represented by $H^\times$ .
Formal Verification: The classification argument is supported by a Lean 4 formalization that verifies the logical structure of the proof, retaining two published results as explicit hypotheses.

Significance and Claims
The paper claims to resolve the long-standing classification problem for order-six complex Hadamard matrices. By establishing that every such matrix has a finite-corner witness, the authors provide a rigorous, exhaustive framework for investigating these objects.

MUBs in $\mathbb{C}^6$ : While the classification does not directly solve the problem of determining the maximum number of mutually unbiased bases in dimension six, it removes the need to search over an unclassified set of Hadamard matrices. It provides a complete, finite-chart design space for transition matrices between unbiased bases.
Interferometry and Quantum Dynamics: The results supply a complete design space for balanced six-port unitary transformations used in multiphoton interferometry. This allows for the systematic study of suppression laws and phase-sensitive interference signatures beyond previously selected families.
Methodological Rigor: The work introduces a "branch-complete" algebraic approach that avoids the pitfalls of generic division, ensuring no valid solutions are lost. This methodology is presented as a framework that may extend to higher-order classifications.

The authors maintain a modest stance regarding future implications, noting that while the classification provides the necessary framework, the specific geometric questions regarding irreducible components and singularities of the class space remain open. The paper emphasizes that the classification is a structural result, providing the "Hadamard ingredient" for various quantum protocols, while device-specific constraints remain as explicit conditions on this complete space.

Technical Summary: A Complete Classification of Complex Hadamard Matrices of Order Six

More like this