The Hodge conjecture is one of the most famous unsolved problems in mathematics. It asks whether certain natural mathematical objects called Hodge classes, which arise from the geometry of complex algebraic varieties, are always represented by algebraic cycles, meaning they can be built from polynomial equations. Proving this in general has resisted decades of effort, but researchers can make progress by tackling specific families of shapes. This paper focuses on a family called Fermat fourfolds, which are four-dimensional geometric objects defined by a simple equation where each variable is raised to the same power. The question is whether every Hodge class on these shapes is algebraic.
The authors prove the Hodge conjecture for all Fermat fourfolds of odd degree up to 199, using a combination of theoretical criteria and large-scale computer verification. They develop three geometric rules for certifying that a Hodge class is algebraic. The first rule works when the relevant mathematical data can be split into two balanced groups, allowing the class to be traced back to simpler geometry on products of curves. The second handles cases where a known structural pattern, studied by a mathematician named Aoki, applies after a small adjustment. The third deals with a single stubborn exception at degree 33 by lifting the problem to a related shape at degree 66, proving algebraicity there, and then pulling the result back down. Together these rules cover all cases after a systematic computer census of roughly 78,000 representative objects.
The census identified thirteen especially difficult Hodge classes that resisted the standard methods, and the paper resolves each one through a careful combination of the new criteria and independent computational verification. Seven of these thirteen cases are genuinely new, falling outside all previously known theoretical frameworks. The authors also perform an arithmetic check using a technique called Jacobi sum computation to rule out a particular shortcut approach for the degree-33 case, showing the geometry there is subtler than it might appear. All results come with machine-checkable certificates and were confirmed by an independent implementation, making the proof unusually transparent and verifiable for work of this kind.