On arithmetic fundamental lemmas (Q421019)

The celebrated Gross-Zagier formula [\textit{B. H. Gross} and \textit{D. B. Zagier}, Invent. Math. 84, 225--320 (1986; Zbl 0608.14019)] relates the Néron-Tate height of Heegner points on modular curves to the central derivative of certain \(L\)-functions associated to modular forms. A parallel formula of \textit{J.-L. Waldspurger} [Compos. Math. 54, 173--242 (1985; Zbl 0567.10021)] relates certain toric periods to the central value of the same type of \(L\)-functions. Conjectural generalizations to some classical groups were proposed by \textit{B. H. Gross} and \textit{D. Prasad} [Can. J. Math. 44, No. 5, 974--1002 (1992; Zbl 0787.22018)] and \textit{W. T. Gan}, \textit{B. H. Gross} and \textit{D. Prasad} [``Symplectic local root numbers, central critical L-values, and restriction problems in the representation theory of classical groups'', preprint]. In the unitary case, they assert an identity relating certain period integrals for the groups \(U(n-1) \times U(n)\) (or the orthogonal counterparts) to the central values of some Rankin-Selberg \(L\)-functions. Their arithmetic versions are identities relating heights of certain algebraic cycles on Shimura varieties to the central derivative of some Rankin-Selberg \(L\)-functions. \smallskip\noindent In this paper, the author follows the relative trace formula (RTF for short) approach to the (global) Gross-Prasad conjecture for \(U(n-1) \times U(n)\) initiated by Jacquet-Rallis (preprint), which consists of a comparison of two relative trace formulas, one on the general linear groups and the other on the unitary groups. The spectral side of the former deals with central values of Rankin-Selberg \(L\)-functions via its integral representation, while the spectral side of the latter deals with the period integral on unitary groups. Then the Gross-Prasad conjecture should follow from the comparison of the geometric sides of the two relative trace formulas. A key ingredient is a conjectural fundamental lemma which takes the form of an identity where the LHS is an orbital integral on some symmetric space and the RHS is an orbital integral on some unitary group. Applying the same approach to the (global) Gan-Gross-Prasad conjecture for \(U(n-1) \times U(n),\) the author on one hand takes the first derivative of the RTF on the general linear groups, and on the other hand he formulates an ``arithmetic RTF'' on unitary groups which deals with the Beilinson-Bloch heights of certain algebraic cycles. The comparison of the two RTFs leads to a conjectural ``arithmetic fundamental lemma'' which is an identity where the LHS is the derivative of Jacquet-Rallis orbital integral and the RHS is an intersection number on unitary Rapoport-Zink space. The author then proves this arithmetic fundamental lemma for \(n = 2, 3\).