Some remarks on microhypoelliptic operators of infinitely degenerate type (Q1318883)

The author proves a result on the microhypoellipticity of a linear operator of the form \[ L = a(x,y,D_x) + g(x) b(x,y,D_y)\quad\text{in}\quad\mathbb{R}_x^{n_1} \times \mathbb{R}^{n_2}_y\quad(n = n_1 + n_2).\tag{*} \] The conditions on \(L\) are \((A_1) : g(0) = 0\) and \(g(x) > 0\) for \(x \neq 0\), \((A_2) : a(x,y, D_x)\) is a differential operator of order \(2l\) verifying \(\text{Re} a (x,y, \xi) \geq C_1 |\xi |^{2l}\) for large \(|\xi |\), and \((A_3) : b(x,y,D_y)\) is a differential operator of order \(2m\), such that \(\text{Re} b(x,y, \eta) \geq C_2 |\eta |^{2m}\) for large \(|\eta |\). \(C_1\), \(C_2\) are positive constants \(l,m \in \mathbb{N}\) and the coefficients of \(a\) and \(b\) are supposed to be \(C^\infty\). The main result is the following: If \(z = (x^0, y^0, \xi^0, \eta^0) \in T^* (\mathbb{R}^n)\) (with \((x^0, \xi^0) \in \mathbb{R}^{n_1} \times \mathbb{R}^{n_1}\), \((y^0, \eta^0) \in \mathbb{R}^{n_2} \times \mathbb{R}^{n_2})\), if \(|\eta^0 |\neq 0\) and if \(l \geq m\), then \(L\) is microhypoelliptic at \(z\). If \(l < m\), if \((A_4)\): there exist \(C \geq 0\) and \(\tau \in (0, {1 \over 2m - 2l})\) such that \(|D^\alpha_x g(x) |\leq C g(x)^{1 - \tau |\alpha |}\) for any multiindex \(\alpha\) with \(|\alpha |\leq 2m - 2l\) near \(x = 0\), then \(L\) is still microhypoelliptic. The proof is a refinement of a previous work of \textit{T. Morioka} [Osaka J. Math. 28, No. 3, 563-578 (1991; Zbl 0763.35023)] which allows to prove that if \(L\) satisfies \((A_1)-(A_3)\) and \((A_5)\): \(L\) is microhypoelliptic at any \(\widetilde z = (\widetilde x, \widetilde y, \widetilde \xi, \widetilde \eta) \in T^* (\mathbb{R}^n)\) with \(|\widetilde \xi ||\widetilde \eta |\neq 0\), that \(L\) is also microhypoelliptic at any \(z = (z^0, y^0; 0, \eta^0)\) with \(|\eta^0 |= 1\). The author shows that in fact that \((A_5)\) follows from \((A_1)-(A_4)\) \((A_4\) is used only in the case \(l < m)\).