Barycentric straightening and bounded cohomology (Q1725575)

It is well-known that the volume of straight simplices of dimension $\ge2$ in hyperbolic $n$-space is universally bounded, while there is no such bound for the volume of straight simplices in euclidean $n$-space. \par For straight simplices in general symmetric spaces of noncompact type, \textit{J.-F. Lafont} and \textit{B. Schmidt} [Acta Math. 197, No. 1, 129--143 (2006; Zbl 1111.57020)] proved that there is an upper bound for the volume of top-dimensional straight simplices, and \textit{I. Kim} and \textit{S. Kim} [``Simplicial volume, barycenter method, and bounded cohomology'', Preprint, \url{arXiv:1503.02381}] proved an upper bound for the volume of straight simplices in codimension one. \par On the other hand, the presence of flats in higher rank symmetric spaces implies that one can not have such a bound when the dimension of the simplices is too small. For example, $\mathrm{SL}(m,\mathbb{R})/\mathrm{SO}(m)$ contains a subspace isometric to the product $\mathbb{R}^1\times \mathrm{SL}(m-1,\mathbb{R})/\mathrm{SO}(m-1)$ and therefore can not have an upper bound for the volume of straight simplices of dimension $\le 1+\dim(\mathrm{SL}(m-1,\mathbb{R})/\mathrm{SO}(m-1))=\frac{1}{2}m(m-1)$. \par In the paper under review, the authors consider irreducible symmetric spaces $X=G/K$ of noncompact type, excluding $\mathrm{SL}(3,\mathbb{R})/\mathrm{SO}(3)$ and $\mathrm{SL}(4,\mathbb{R})/\mathrm{SO}(4)$, and prove an upper bound for the volume of straight simplices of dimension $\ge \dim(X)-rk(X)+2$. In the case of $X=\mathrm{SL}(m,\mathbb{R})/\mathrm{SO}(m)$ this means that they obtain an upper bound in dimensions $\ge \frac{1}{2}m(m-1)+2$, thus only the case of dimension $\frac{1}{2}m(m-1)+1$ remains open. \par The definition of straight simplices in higher-rank symmetric spaces is more involved than in the constant curvature case. Given a symmetric space of noncompact type $X$ with Furstenberg boundary $\partial_{Fur}X$, denote its Busemann functions by $B(x,\theta)$ and its family of Patterson-Sullivan measures by $\mu(x)$, for $x\in X$ and $\theta\in\partial_FX$. For any probability measure $\nu$ fully supported on $\partial_{Fur}X$, one has a strictly convex function $F: x\to\int_{\partial_FX}B(x,\theta)d\nu(\theta)$ on $X$, and one defines the barycenter of $\nu$ to be the unique minimum of this function. The straight simplex with given vertices $x_1,\ldots,x_{k+1}$ is now defined as the simplex $\Delta^k\to X$ which sends $a_1e_1+\ldots+a_{k+1}e_{k+1}$ to the barycenter of the measure $a_1^2\mu(x_1)+\ldots+a_{k+1}^2\mu(x_{k+1})$. \par The second derivative of the strictly convex function $F$ involves certain quadratic forms $Q_1(v,v)=\int_{\partial_FX}dB^2_{(x,\theta)}(v)d\mu(\theta)$ and $Q_2(v,v)=\int_{\partial_FX}DdB_{(x,\theta)}(v,v)d\mu(\theta)$ for $v\in T_xX$. The upper bound on the volume of straight $k$-simplices follows from an upper bound on the quantity $\frac{\det(Q_1\mid_S)^\frac{1}{2}}{\det(Q_2\mid_S)}$ for any $k$-dimensional subspace $S\subset T_xX$. The proof of an upper bound for the latter quantity occupies the bulk of the paper. Its main technical constituent is a ``weak eigenvalue matching'': for the small eigenvalues of $Q_2$ one can find twice as many comparably small eigenvalues of $Q_1$ and $r-2$ additional small eigenvalues of $Q_1$ that are bounded by a universal constant times the smallest eigenvalue of $Q_2$. This remarkably depends on a combinatorial result, namely a slight strengthening of Hall's marriage theorem. \par By a well-known straightening argument, the result implies a partial answer to Dupont's conjecture, namely one obtains surjectivity of the homomorphism $H_{c,b}^*(G)\to H_b^*(G)$ from continuous, bounded cohomology to continuous cohomology in degrees at least $\dim(X)-rk(X)+2$. The conjecture of \textit{J. L. Dupont} [Lect. Notes Math. 763, 109--119 (1979; Zbl 0511.57018)] from 1978 predicts that this homomorphism should be surjective in all degrees.