A Cheeger-Müller theorem for symmetric bilinear torsions on manifolds with boundary (Q2018931)

Given an orthogonal bundle on a compact Riemannian manifold one can define two invariants : the Reidemeister torsion from a smooth triangulation (or a Morse function) and the associated cochain complex (or Morse complex) with coefficients in the bundle, and the analytic torsion from the spectra of the Laplace operators on forms with coefficients in the bundle : the latter was defined by \textit{D. B. Ray} and \textit{I. M. Singer} [Adv. Math. 7, 145--210 (1971; Zbl 0239.58014)], and the former is studied by \textit{J. W. Milnor}'s survey in [Bull. Am. Math. Soc. 72, 358--426 (1966; Zbl 0147.23104)]. The classical Cheeger--Müller theorem states the equality of these two invariants : it was proven independently by \textit{J. Cheeger} [Ann. Math. (2) 109, 259--322 (1979; Zbl 0412.58026)] and by \textit{W. Müller} [Adv. Math. 28, 233--305 (1978; Zbl 0395.57011)] in the case where the manifold has no boundary. In the case where the manifold has a boundary both sides are well-defined but the equality is perturbed by an additional term; the theorem in this case is more recent and due to S. Vishik and W. Lück in the case where the metric is a product near the boundary, and to J. Brüning and X. Ma in the general case (for all these references and later ones we refer to the bibliography of the paper under review). One does not need an orthogonal bundle to define analytic torsion. For examples, the analytic torsion with coefficients in flat complex bundles over a closed manifold endowed with a symmetric bilinear form (which is thus totally split from the real point of view) was defined by D. Burghelea and S. Haller, and a Cheeger-Müller type equality was proven by the author of the present paper and W. Zhang. In the work under review the author extends this last result to the case of a compact manifold having boundary (which was already studied by O. Maldonado-Molina). He has to assume that both the Riemannian metric on the manifold and the bilinear form are products near the boundary. The final formula (stated in Theorem 5.1) depends on too many additional invariants to be quoted in full here.