Bessel periods, seesaw and separation of functionals
To read this chapter you need the global theta lift, Bessel and Fourier-Jacobi periods, and local multiplicity one for Bessel functionals.
The Bessel seesaw chapter proves FunctionalsSeparate (defined below) for the unitary hypotheses chapter. It says that a Bessel period of an iterated theta lift is not zero, and that one real test datum works for all inputs . The proof climbs a tower of Bessel and Fourier-Jacobi periods, then uses local multiplicity one at the real place.
Read first
- Hermitian and skew-Hermitian forms: the grid of Hermitian spaces over .
- Haar measure. Each automorphic integral runs over a fundamental domain.
- Unitary groups: the groups of the grid and the real groups .
- Representations of real groups and discrete series, for Harish-Chandra modules at the real place.
- Constant terms and cusp forms. The rapid decay of cusp forms is an input of .
- The Weil representation sits on each edge of the grid, with the splitting characters of the Weil representation made from Hecke characters.
- The Fock model gives the vacuum vectors.
- Theta kernels and the global theta lift, for each step of the iterated lift.
- Seesaw pairs. Here the seesaw identity is an exchange of two integrals.
- Generic characters and Bessel and Fourier-Jacobi periods: the character , the Gross-Prasad period and the periods of the tower.
- The descent of Ginzburg, Rallis and Soudry is the method the ascent follows.
- Local multiplicity one bounds the space of real Bessel functionals.
- Interlacing and the branching law for discrete series is the source of Interlaced (below).
- The Gan-Gross-Prasad conjectures, the deepest: the code states a case of the local Gan-Gross-Prasad conjecture but never uses it.
The new objects
- The iterated lift runs along a row of the grid. The Bessel period pairs a form on the top node with a form on . The number studied is .
- FunctionalsSeparate says: for large families and each nonzero input there is a pair such that for each nonzero some , give . The pair is the same for all .
- PairEx gives a real intertwiner for one theta step. Its theorem is named after a seesaw, but states no seesaw identity.
- The bottom period is an integral over the smallest unitary group of the tower. PeriodNe says that it is nonzero for some core vectors. One input is a real matrix coefficient identity, proved in Lean.
- Ascent says: if along a chain of forms, the top Bessel period is not zero. WeakSep is FunctionalsSeparate with a pair that may depend on .
- BurGapPos bounds by the rank of the real Bessel functionals of a discrete series with a large gap. Interlaced is a sign condition on two Harish-Chandra parameters.
The path through the chapter
- By definition, is a global Bessel period of the iterated lift.
- The real place gives a nonzero local period, so .
- The ascent climbs one rung at a time. In the first rung, the change of the order of integration is the seesaw identity.
- With the surjectivity of the core lifts, the ascent gives WeakSep.
- Local multiplicity one turns WeakSep into one datum for all .
The source of this page. Markdown. Something is wrong? Report it.