A matrix can be built from a collection of distinct real numbers by taking their pairwise differences and raising each difference to a fixed integer power. Specifically, if you have N distinct real numbers and an integer D, you arrange the N-by-N grid so that the entry in row i and column j is the difference between the j-th and i-th number, raised to the power D. A central question in linear algebra is whether such a matrix is invertible, which is equivalent to asking whether its determinant is nonzero. This question was posed by Colombo in 1928 and remained only partially resolved for nearly a century.
The paper gives a complete and clean answer: the determinant is nonzero if and only if two conditions both hold. First, the power D must be at least N minus 1. Second, either N or D must be even. So, for example, if you have five numbers and raise differences to the fifth power, both conditions fail on the second requirement since both five and five are odd, and the determinant will be zero. The even-exponent half of this result was already established in work from the 1980s, but the odd-exponent case had remained open. The authors resolve it using a tool called Pfaffians, which are algebraic objects associated with antisymmetric matrices, and they prove a strict sign theorem about these Pfaffians to clinch the result.
An additional notable feature of the work is that the authors formally verified their proof using a computer proof assistant called Lean 4. This means every logical step in the odd-exponent argument was checked mechanically, leaving no room for hidden errors. This kind of formal verification is still relatively rare in mathematics and represents a growing trend toward making proofs fully machine-checkable. Together, the human mathematical insight and the formal verification give a thorough and trustworthy resolution to a problem that sat open for nearly a hundred years.