Nonexistence of arithmetic fake compact Hermitian symmetric spaces of type other than \(A_n\) \((n\leq 4)\) (Q455000)

Hermitian symmetric spaces are those among the symmetric spaces who have a complex structure which is invariant under their group of isometries. When of non-compact type, such a space has a compact dual: for example, the compact dual of the complex hyperbolic plane \({\mathbb H}_{\mathbb C}^2 = \operatorname{SU}(2,1)/\operatorname{U}(2)\) is the complex projective plane \({\mathbb P}_{\mathbb C}^2 = \operatorname{SU}(3)/\operatorname{U}(2)\). If \(X\) is such a space of non-compact type, \(X_u\) its compact dual, then a quotient \(X/\Pi\) of \(X\) by a torsion-free lattice \(\Pi\subset \operatorname{Isom}X\) is called a fake \(X_u\) if its characteristic zero cohomology is the same as that of \(X_u\). Since there is always an injection \(H^*(X_u,{\mathbb Q})\to H^*(X/\Pi,{\mathbb Q})\) such quotients have minimal Betti numbers among all manifolds locally isometric to \(X\).NEWLINENEWLINEIn [Invent. Math. 168, No. 2, 321--370 (2007; Zbl 1253.14034)], the authors have studied compact arithmetic fake projective planes (the complete classification of such has been completed by \textit{D. I. Cartwright} and \textit{T. Steger} [C. R., Math., Acad. Sci. Paris 348, No. 1--2, 11--13 (2010; Zbl 1180.14039)], and in [Am. J. Math. 131, No. 2, 379--407 (2009; Zbl 1168.14018)], the authors have extended their study to the case of fake \(X_u\)s when \(X\) has type \(A_n\), \(n\) even. This paper almost finishes the study of fake \(X_u\)s by showing that, in all other cases, except possibly for the type \(A_3\) (when \(X\) has to be isometric to \(\operatorname{SU}(2,2)/\operatorname{U}(3)\)), there are no compact arithmetic fakes (note that the existence of a nonarithmetic quotient in these cases is known to be impossible, except when \(X={\mathbb H}_{\mathbb C}^m\) for \(m\geq 4\) where it is a major open question). The method of proof is the same as in the first two papers of the series: using the volume formula due to the first author to limit the problem to a finite number of commensurability classes -- in this case none. In the case left open, when \(X\) is of type \(A_3\), the authors obtain reasonable bounds for the field of definition of an hypothetic fake.