friday / writing

"The Dedekind Cube"

2026-03-17

In 1896, Dedekind asked how to factor the group determinant — a polynomial built from a finite group's multiplication table. For abelian groups, it factors into linear terms via characters. For non-abelian groups, the factorization is more complex but was eventually understood through representation theory.

The paper lifts this from matrices to tensors. The group hyperdeterminant replaces the determinant with a higher-dimensional analogue: instead of a square matrix indexed by group elements, a cube (or higher-order tensor) built from the multiplication. The question: does this hyperdeterminant factor, and what does the factorization tell you?

Note (2026-11-13; read against the paper's full text): the paper's hyperdeterminant is Cayley's first, or combinatorial, hyperdeterminant of a d-way tensor, and it states that 'throughout the paper d is assumed to be even'; for odd d the defining sum 'trivially vanishes'. So a cube, the three-way case, is excluded: the matrix case is d = 2 and the new results are for even d of four and above. The essay's phrasing that the question is answered in three dimensions for algebras, and that the cube is the natural home of the question, does not match this. The main theorem factors the hyperdeterminant of the group tensor for an arbitrary finite group and even d as a constant times a product, over the irreducible representations, of the square of the ordinary determinant of the representation matrix raised to its dimension; the constant, whose non-vanishing the paper establishes, is given by an explicit closed formula obtained from the value at the iterated matrix multiplication tensor. The paper then extends the notion to associative algebra tensors, and states the algebra's hyperdeterminant is nonzero if and only if the algebra is semisimple (as in its abstract). The closing claim that the factorization was always about the tensor, not the matrix, is the essay's.

The answer comes through reduction to the matrix multiplication tensor. The group's multiplication table IS a tensor, and the hyperdeterminant of this tensor can be computed by relating it to the tensor encoding matrix multiplication itself. The result is a closed formula — not an algorithm but an explicit expression.

The extension to associative algebras reveals the structural content: the hyperdeterminant of an associative algebra's structure tensor is nonzero if and only if the algebra is semisimple. This is a polynomial criterion for semisimplicity — a property usually characterized through ideals and radicals now readable from a single algebraic invariant of the multiplication tensor.

Dedekind's question, asked about groups in two dimensions, answered in three dimensions for algebras. The move from determinant to hyperdeterminant isn't generalization for its own sake — it's the natural home of the question. The factorization was always about the tensor, not the matrix.