The Langlands program is a vast web of conjectures connecting number theory, geometry, and representation theory. A central object of study is the cohomology of Shimura varieties, which are certain geometric spaces that encode deep arithmetic information. Understanding how symmetry groups act on this cohomology is a key step toward proving parts of the Langlands program, but this becomes especially difficult at "bad primes," where the geometry of these spaces behaves irregularly.
This paper extends a powerful technique, developed by Langlands, Kottwitz, and Scholze, for computing a kind of accounting formula (called a trace formula) that tracks how these symmetry actions behave. The method works by constructing special functions, called local test functions, that capture the local arithmetic data at bad primes. The authors build these test functions in full generality using recent advances in the foundations of Shimura varieties, and they verify that the functions have the right mathematical properties to plug into the trace formula machinery. This gives a more complete and unconditional version of a conjecture by Scholze and Shin, which predicts a precise relationship between the cohomology of Shimura varieties and certain objects from representation theory called automorphic forms.
Along the way, the authors study some intermediate geometric structures called apertures, which are local models that approximate the behavior of Shimura varieties near bad primes. They show that these apertures, despite their abstract definition, carry exactly the same information as a more classical and concrete tool called Fontaine-Laffaille theory, at least in nice settings. They also give explicit descriptions of how these apertures deform, which makes the theory much more concrete and workable. Together, these results push the Langlands-Kottwitz-Scholze method closer to full generality and lay groundwork for future progress on the Langlands program.