On a problem of Yau regarding a higher dimensional generalization of the Cohn-Vossen inequality (Q1938068)

In this article, the author gives a negative answer to a natural analogue of the famous Cohn-Vossen inequality which gives a bound on the normalized total curvature of a complete Riemann surface. Namely, he provides counterexamples to a question by \textit{S.T. Yau} [AMS Proc. Symp. Pure Math. 54, Part 1, 1--28 (1993; Zbl 0801.53001)]: {`` Given a complete Riemannian manifold \((X^n,g)\) with nonnegative Ricci tensor, \(p\in X\), and \(k\in\{2,\dots,n\}\), are the normalized integrals \(r^{2k-n}\int_{B_p(r)}\sigma_k(\text{ric}^g)\,\text{vol}^g\) bounded as \(r\) goes to infinity?''}. The counterexample manifolds are elementary: they are the affine spaces \(\mathbb C^n\), \(n\geq1\). The candidates for counterexample metrics selected by the author are reasonably simple as well: they are taken in the set \(\overline{\mathcal{M}}_n\) of \(\mathbb U(n)\)-invariant Kähler metrics on \(\mathbb C^n\) of nonnegative bisectional curvature. Indeed, extending results by \textit{H. H. Wu} and \textit{F. Zheng} [Adv. Lect. Math. (ALM) 17, 517--542 (2011; Zbl 1262.53063)] (who dealt with the ``positive bisectional curvature'' case), the author establishes that each such metric can be completely described through a one-parameter function. He proves more precisely (Theorem 3.2) a one-to-one correspondence between \(\overline{\mathcal{M}}_n\) and the set \(\mathcal{F}\) of smooth, even, convex functions \(F\) defined on \((-x_0,x_0)\) for some \(x_0\in(0,\infty]\), vanishing at 0, and verifying precise limit conditions -- certain situations, one could think to appear by simply replacing strict inequalities by large ones in Wu and Zheng's result, not being allowed. This is one of the main technical results of the paper, which requires a careful one-dimensional analysis. Moreover, for a given metric, the associated even convex function \(F\) encodes some geometric data; for instance, the geodesic distance \(s\) of a point \(z\) to the origin is given by \(\int_0^{x}\sqrt{1+F'(\tau)^2}d\tau\), and \(\text{Vol}\big(B_0(s)\big)=c\big(\int_0^{x}\tau\sqrt{1+F'(\tau)^2}d\tau\big)^n\); here \(x=x(|z|)>0\) is the solution of the differential equation \(\tfrac{\partial \log x}{\partial r}=\tfrac{1}{2r[1+F'(x)^2]^{1/2}}\), and \(c\) a positive constant. Further elementary formulas are available for the relevant curvature quantities. This description enables the author to construct a metric \(g\) of \(\overline{\mathcal{M}}_n\), such that \(s^{2k-2n}\int_{B_0(s)}\sigma_k(\text{ric}^g)\,\text{vol}^g\), \(k=2,\dots n\), blows up as \(s\) goes to infinity, giving the desired counterexamples (Theorem 3.3); the use of metrics of nonnegative, instead of only positive, bisectional curvature, is here crucial. Among other auxiliary results, the author uses his description via the \(F\) function to give a positive answer for metrics of \(\overline{\mathcal{M}}_n\) to the following modification of Yau's question, natural in the Kähler setting: Given \(p\) in an \(n\)-complex-dimensional complete Kähler manifold \((X,g,\omega)\) of nonnegative Ricci form \(\rho\), are the normalized integrals \(s^{2k-2n}\int_{B_p(s)}\rho^k\wedge\omega^{n-k}\) bounded as \(s\) goes to infinity?'' (Theorem 3.6). Notice that the integrands of this statement are slightly different from those of Yau's problem, making here possible a positive answer.