The Langlands program is a vast web of conjectures connecting number theory, representation theory, and geometry. A central theme is that automorphic forms, which are highly symmetric functions arising in number theory, on different groups should be deeply related to each other. This paper develops a technique called "twisted automorphic descent" that constructs automorphic forms on one type of group, called odd GSpin groups, starting from automorphic forms on a more familiar group called GL(2n). GSpin groups are algebraic groups related to spin geometry and quadratic forms, and the "odd" and "twisted" labels refer to specific structural variants that make the construction both more general and more technically challenging than previous work.
The authors use this construction to give a new, explicit proof of the Jacquet-Langlands correspondence for GL(2) and quaternion algebras. A quaternion algebra is a four-dimensional number system generalizing the familiar quaternions of Hamilton, and the Jacquet-Langlands correspondence says that automorphic forms on such an algebra are secretly the same as certain automorphic forms on the group GL(2), just with some local conditions. Having an explicit construction, rather than an existence proof, is valuable because it gives concrete control over how the correspondence works and opens the door to further calculations.
The paper also proves a new case of the Gan-Gross-Prasad conjecture, which predicts when a certain type of integral involving two automorphic forms, called a period integral, is nonzero. Nonvanishing of these integrals is connected to special values of L-functions, which are generalizations of the Riemann zeta function encoding deep arithmetic information. By establishing this case of the conjecture, the authors contribute to a broader program linking the analytic behavior of L-functions to the representation-theoretic structure of automorphic forms, with potential implications for arithmetic questions about number fields and their extensions.