Hypoellipticity and nonhypoellipticity for sums of squares of complex vector fields (Q371167)

The authors consider the following linear partial differential operator in two real variables \((x,t)\): \[ P=BB^{\ast }+B^{\ast }(t^{2l}+x^{2k})B , \quad B=D_{x}+ix^{q-1}D_{t}, \] where \(l,k\) and \(q\) are positive integers, and \(q\) is even, with the notation \( D_{x}=i^{-1}\partial _{x}\)\ and \(D_{t}=i^{-1}\partial _{t}\). The operator \(P\) is a model of sum of squares of complex vector fields in the plane with a characteristic variety the symplectic real analytic manifold \(\left\{ (x,t,\xi ,\tau ):x=\xi =0\right\}\). Using a suitable pseudodifferential operator calculus in the setting of \(C^{\infty } \)\ functions and Gevrey spaces \(G^{s},\)\ as well as asymptotic methods, the main result of the paper is the following theorem. Theorem. The following hold true. {\parindent=6mm \begin{itemize}\item[(i)] Suppose that \(l>\frac{k}{q}\); then \item [1.] \(P\) is \(C^{\infty}\)-hypoelliptic (in a neighborhood of the origin) with a loss of \(\frac{2(q-1+k)}{q}\)\ derivatives; \item [2.] \(P\) is \(G^{s}\)-hypoelliptic for any \(s\geq \frac{lq}{lq-k}\); \item [3.] \(P\) is not \(G^{s}\)-hypoelliptic for any \(1\leq s<\frac{lq}{lq-k}\); \item [(ii)] if \(l\leq \frac{k}{q},\) then \(P\) is not \(C^{\infty }\)-hypoelliptic. \end{itemize}} It is interesting to note that the operator \(P\)\ satisfies the complex Hörmander condition; but it is well-known that this condition in the case of sums of squares of real vector fields implies \(C^{\infty }\)-hypoellipticity, however the result (ii) of the theorem gives the following corollary. Corollary. The complex Hörmander condition does not imply \(C^{\infty }\)-hypoellipticity for sums of squares of complex vector fields. The following interesting consequence, linked with a conjecture due to \textit{F. Treves} [Proc. Symp. Pure Math. 65, 201--219 (1999; Zbl 0938.35038)] stating a necessary and sufficient for analytic hypoellipticity for sums of squares of real vector fields, is given by the result (i)-3. of the theorem. Corollary. Treves's conjecture does not carry over to sums of squares of complex vector fields.