A necessary and sufficient condition for Melin's inequality for a class of systems (Q1604970)

The authors give necessary and sufficient conditions for the Melin inequality to hold: Given a fixed formally selfadjoint \(N\times N\) system \(A=A^{\text{ w}}(x,D_x)\) of Weyl pseudodifferential operators of order \(m\) on \({\mathbb R}^n\), for any given \(\varepsilon>0\) and any given compact \(K\subset{\mathbb R}^n\) there exists a constant \(C=C_{K,\varepsilon}\in{\mathbb R}\) such that \[ (Au,u)\geq-\varepsilon\|u\|^2_{(m-1)/2}-C\|u\|_{m/2-1}^2,\qquad \forall u\in C_0^\infty(K;{\mathbb C}^N). \] Following \textit{R. Brummelhuis} [C. R. Acad. Sci., Paris, Sér. I 315, No. 2, 149-152 (1992; Zbl 0754.35197); Commun. Partial Differ. Equations 26, No. 9-10, 1559-1606 (2001; Zbl 1007.35111), preceding review], these conditions are given at each characteristic point, in terms of spectral properties of families of matrix oscillators. The proof of the sufficiency, however, requires an extra condition which is roughly stated by saying that the square of each derivative of the principal symbol has to be dominated, in the sense of Hermitian matrices, by the principal symbol itself. (It is not clear whether this extra condition is technical or not.) On the other hand, when the principal symbol has locally on the characteristic set constant rank \(r\) and its determinant vanishes exactly to order \(2(N-r)\) there, then only one suitable matrix oscillator has to be considered and the extra condition mentioned above is automatically fulfilled. In this way, since no condition is imposed on the symplectic nature of the characteristic set, the authors generalize the version of Melin's inequality as proved by \textit{C. Parenti} and \textit{A. Parmeggiani} [J. Anal. Math. 86, 49-91 (2002; Zbl 1055.35153)] for systems with symplectic double characteristics. (However, the main result of that paper is an extension of \textit{L. Hörmander}'s inequality [J. Anal. Math. 32, 118-196 (1977; Zbl 0367.35054)] to the case of systems.) In the first section, after defining the crucial family of matrix oscillators, the authors state the main theorem, whose proof is given, by generalizing Melin's arguments, in the second section. In the third and last section they show how to derive, in the case of systems of constant rank and precise order of vanishing on their characteristic set, a condition in terms of a single suitable matrix oscillator which is both necessary and sufficient.