First-hand information for everyone
This paper proves a concrete matching result for a specific case of the Jacquet–Rallis relative trace formula. In plain terms: when GL(2) acts on 3-by-3 matrices (the Lie algebra gl(3)), every polynomial-type Schwartz test function on the real side has a matching function on the corresponding unitary side, and vice versa. Matching here means that certain orbital integrals — averages of a test function along group orbits — agree under the natural bijection of regular semisimple orbits.
The authors work in the real (archimedean) setting and use Lie algebra methods and (g,K)-modules. A (g,K)-module is a standard way to package smooth representation data: g is the Lie algebra and K is a maximal compact subgroup. The “polynomial-type Schwartz” functions they treat are the usual rapidly decreasing smooth functions multiplied by polynomials; concretely these arise from the standard Gaussian times polynomials. Their main theorem covers the case n = 2, i.e. GL(2) acting on gl(3), and the analogous unitary groups coming from two-dimensional hermitian spaces.
The proof mixes algebra and analysis. Algebraically, the authors use the infinitesimal Weil representation — a classical tool that links two groups via a shared representation — and they study rings of invariant differential operators. They form certain coinvariant modules (quotients that record how the orthogonal group action factors out) and show these modules are generated by explicit elements. On the GL(2) side they write down three explicit generators; on the unitary side the orbital integral of the Siegel Gaussian generates the module. They then show these generators transfer to each other, and the ring of invariant differential operators lets them extend the transfer to the whole polynomial-type space. Part of the analytic work uses a differential equation satisfied by the orbital integrals and a uniqueness argument; a short computer calculation was used to check some lengthy analytic identities.
This result matters because smooth transfer of orbital integrals is a key local ingredient in comparisons of relative trace formulas. Such comparisons underpin important global arithmetic results, for example work toward the Gan–Gross–Prasad conjectures. Previously, the strongest archimedean result was that functions admitting transfers are dense (work of Hang Xue); in non-archimedean (p-adic) settings a full transfer theorem is known (Wei Zhang). The present paper fills in a concrete archimedean case (n = 2) at the level of polynomial-type test functions and suggests a representation-theoretic viewpoint that may help with derivatives of orbital integrals and related arithmetic intersection problems.
There are important caveats. The theorem is proved only for n = 2 and for polynomial-type Schwartz functions, not for all Schwartz functions; the authors note polynomial-type functions are dense, but extending the result to all Schwartz functions is left as a conjecture. Some analytic regularity questions about orbital integrals are only partially resolved; the argument for one side required a computer-assisted verification of many terms. The authors also formulate broader conjectures about the coinvariant modules and expect the same generation phenomenon to hold in higher rank, but those generalizations remain open.