Non-elliptic Laplace equations on nilpotent Lie groups (Q801424)

In what follows N is a (connected, simply connected) nilpotent Lie group, with Lie algebra \({\mathfrak n}\). Assume that there is a linear map \(J: {\mathfrak n}\to {\mathfrak n}\) such that \(J^ 2=-1\) and the \(\pm i\) eigenspaces of J in \({\mathfrak n}_{{\mathbb{C}}}\) are subalgebras. Assume further that n has a nondegenerate 2-form \(\Phi\), invariant under J, such that \(\Phi (X,[Y,Z])=\Phi ([X,Y],Z)+\Phi (Y,[X,Z])\) for all X,Y,Z in \({\mathfrak n}\). J induces a complex structure on \({\mathfrak n}\), and hence a complex structure on N for which left multiplication is holomorphic. If the form \(\omega (X,Y)=\Phi (JX,Y)\) is positive, then N has the structure of a Kaehler manifold; in general, we describe N as pseudo-Kaehlerian. We can use \(\Phi\) to identify the tangent and cotangent bundles of N; by a standard procedure, we get a Laplacian on q-forms. This the non-elliptic Laplacian of the title. If \(\Gamma\) is a discrete subgroup, we can use the same construction to define \(\Delta_ q\) on \(\Gamma\) \(\setminus N\). For \(q=0\) (the main case of interest in the paper), one can construct \(\Delta_ 0\) in another way. Let \(\{X_ 1,...,X_ n\}\) be a basis of \({\mathfrak n}\), and let \(\bar X_ 1,...,\bar X_ n\) be the dual basis under \(\omega\). Then \(\square =\sum^{n}_{j=1}\bar X_ JX_ J\in U({\mathfrak n})\) acts as \(\Delta_ 0.\) Say that pseudo-Kaehlerian structure is measured if there exists a sequence of ideals \({\mathfrak n}={\mathfrak n}_ 1\neq {\mathfrak n}_ 2\neq...\neq {\mathfrak n}_ d=(0)\), stable under J, with \({\mathfrak n}_ j^{\Phi}={\mathfrak n}_{d-j+1}\) and [\({\mathfrak n}_ 1,{\mathfrak n}_ j]={\mathfrak n}_{j+1}\). One fundamental fact for the paper is that if the structure is measured, then \({\mathfrak n}\cup\) \(\{\) \(\square \}\) generates a finite-dimensional Lie subalgebra \({\mathfrak t}_ C\) in U(\({\mathfrak n})\). Let \({\mathfrak t}\) be the Lie algebra of formally skew-adjoint elements in \({\mathfrak t}_{{\mathbb{C}}}\); it corresponds to a Lie group T, and the right regular representation R of N extends to a representation R' of T on \(L^ 2(N)\); moreover, R and R' have the same \(C^{\infty}\) vectors. The author shows how these results give detailed information on the spectrum of \(\square\) (or \(\Delta_ 0)\). For instance, in the case of \(\Gamma\) \(| N\), with \(\Gamma\) discrete and cocompact, he shows that if \(\pi\) is an irreducible representation in \(L^ 2(\Gamma /N)\), then either \(\pi\) (\(\square)\) is scalar or \(\pi\) (\(\square)\) has no discrete spectrum. In the case of N, \(\square\) has no discrete spectrum. The author applies these results in various ways. For instance, he shows that for certain \(\Gamma\) \(| N\) and a certain complex line bundle defined from \(\Phi\), one obtains a \({\bar \partial}\) complex with nontrivial cohomology such that every harmonic cocycle is a coboundary. Thus no Hodge theory exists in these cases. He also shows how to adapt his constructions to obtain results on CR manifolds; in particular, he generalizes theorems of Folland, Greiner, and Stein.