No label defined (Q3972650)

[For the entire collection see Zbl 0713.00009.] In recent years, problems of geometric origin have attracted the attention of many analysts. In some instances, geometry was pointing to the most subtle case of an analytic problem, e.g., the limiting case of Sobolev inequalities in the Yamabe problem or in Yang-Mills theory. Moreover, phenomena of a geometric nature were appearing in a P.D.E. context like the ``bubbling off'' phenomenon in the harmonic map problem. In many of the geometric situations considered, the problem could be reduced to solving an elliptic scalar equation, most often a nonlinear one. One of our main points in writing this survey is to show evidence that, although legitimate, such an emphasis on scalar equations is indeed a reduction, i.e., that by restricting the considerations to that special case some crucial information is lost. The theme of this talk will therefore be to make propaganda for studying systems of partial differential equations, and their relationships to scalar equations. The (may be strange sounding) title of this talk aims at suggesting two different, and partly contradictory, ideas. In the study of ``harmonic'' objects, Weitzenböck formulas are indeed a powerful tool which can be developed quite systematically, but the reason why they work so well is not yet completely clear. Another aim for this report may also be to convince more analysts that, as soon as they are dealing with operators with variable coefficients, they should not be afraid of using geometric tools, e.g., shift from open sets of Euclidean spaces to manifolds. This article is organized as follows: In Section I, we recall some standard situations where analysis and geometry have come together. In Section II, after having recalled the first use of Weitzenböck formulas as made by S. Bochner more than fifty years ago we give a fairly general definition of them, and explain their main uses. In Section III, we show a number of different instances where passing to a system, and using a Weitzenböck formula proves useful. Examples that we work out include deriving eigenvalue estimates for scalar operators from an appropriate Weitzenböck formula, and subelliptic estimates for the norms of sections satisfying a system of equations. In Section IV, we present two recent developments related to Weitzenböck formulas. The first one gives improved estimates for the norm of the gradient of the length of a harmonic section of a bundle (a so-called Kato inequality) using again elementary representation theory. The second one, due to \textit{A. Polombo} [C. R. Acad. Sci., Paris, Ser. I. 307, No. 12, 667-670 (1988; Zbl 0647.53038)], deals with sections of an endomorphism bundle for which Weitzenböck formulas appear when computing the Laplacian of its eigenvalues.