On hypoellipticity for a certain operator with double characteristic (Q1179725)

Consider an operator \(P=D_ 1^ 2+D_ 2^ 2+D_ 3^ 2+x_ 3^ 2 D_ 4^ 2+(f(x_ 1)-1)D_ 4\) under conditions: (A.1) \(f(0)>0\), \(f(x_ 1)>0\) for \(x_ 1\neq 0\), \(f(x_ 1)\) is monotone in \([0,\delta)\) and \((-\delta,0]\); (A.2) there exist positive numbers \(\delta_ 1\) and \(\varepsilon\) such that \(| x_ 1\log f(x_ 1)|\leq\varepsilon\) for \(0<x_ 1<\delta_ 1\); (A.3) \(| x_ 1\log f(x_ 1)|\) tends to 0 as \(x_ 1\) tends to 0. Then the author proves the following two theorems: Theorem 1: We assume (A.1) and (A.2). Then the operator \(P\) is not hypoelliptic in any neighborhood of the hypersurface \(x_ 1=0\). Theorem 2: We assume (A.1) and (A.3). Then the operator \(P\) is hypoelliptic in some neighborhood of the hypersurface \(x_ 1=0\). Generally speaking, one of the ideas of studying the hypoellipticity is the study of the propagation of singularities along the double characteristic manifold \(\Sigma\). For the operator \(P\), we note that Melin's invariant vanishes on \(\Lambda=\Sigma\cap\{x_ 1=0\), \(\xi_ 4>0\}\) and for any point in \(\Lambda\) there exists a vector \(\partial_{x_ 2}\) tangent to \(\Lambda\). Thus, the author uses the assumption (A.3) in order to show the hypoellipticity of the operator \(P\).