Prescribing Ricci curvature on a product of spheres (Q2118004)

Given a Riemannian manifold \(M\), the problem of prescribing the Ricci curvature consists of finding a Riemannian metric \(g\) on \(M\) so that \[\text{Ric}(g) = T\tag{1} \] for a given symmetric \((0,2)\)-tensor field \(T\). This equation is a second-order nonlinear PDE in the components of the metric. \textit{D. M. DeTurck} [Invent. Math. 65, 179--207 (1981; Zbl 0489.53014)] proved that (1) always has a solution locally around each point \(p \in M\) where \(T\) is non-degenerate. On the other hand, the global solvability of (1) is a much more subtle problem. \textit{D. M. DeTurck} and \textit{N. Koiso} [Ann. Inst. Henri Poincaré, Anal. Non Linéaire 1, 351--359 (1984; Zbl 0556.53026)] and \textit{R. Hamilton} [Publ., Math. Sci. Res. Inst. 2, 47--72 (1984; Zbl 0557.53018)] showed that for a \emph{compact} manifold \(M\) and for every given \(T\), there exists \(c_0 = c_0(T)\) so that for all \(c > c_0\), the tensor \(c \cdot T\) cannot be a Ricci tensor of any Riemannian metric on \(M\). Because of this negative result, a more reasonable global problem is to find both the metric \(g\) and the scaling \(c\) so that \[\text{Ric}(g) = c \cdot T. \tag{2}\] This global problem was solved on certain manifolds in a perturbative setting or under additional symmetry assumptions (e.g., by assuming the invariance under certain groups acting on the manifold). A result of the latter kind, that was a motivation for the paper under review, was obtained by Hamilton. In [loc. cit.], for \(M = \mathbb S^3\) he showed that if \(T\) is itself a metric of positive curvature on \(M\) equivariant under the action of \(\mathrm{SO}(3) \times \mathbb Z_2\) (i.e., rotationally symmetric with an additional reflection symmetry about the equatorial sphere), then there exists a unique \(c > 0\) and a metric \(g\) on the \(3\)-sphere satisfying (2). In the paper under review, the authors solve a similar problem when \(M\) is a product of two spheres. Their main result is the following theorem: Theorem. Let \(T\) be a \(\mathrm{SO}(d_1 + 1) \times \mathrm{SO}(d_2 + 1) \times \mathbb Z_2\)-invariant metric on \(M = \mathbb S^{d_1 + 1} \times \mathbb S^{d_2}\) of the form \[ T = dt^2 + T_1(t) \cdot \Omega_1^2 + T_2(t) \cdot \Omega_2^2, \quad t \in [0,1], \] with \(\Omega_i^2\) being the standard round metric on \(\mathbb S^{d_i}\). Assume that the \(T_i\)'s additionally satisfy \begin{align*} &T_i'(t) \geqslant 0 \text{ for }t \in [0,1/2], i \in \{1,2\},\\ &T_1'(t) > 0 \text{ for }t \in (0,1/2),\\ &T_1''(1/2) < 0. \end{align*} Then there exist positive constants \(c_1, c_2\) and a smooth invariant metric \(g\) on \(M\) satisfying the equation \[ \text{Ric}(g) = c_1\left(dt^2 + T_1(t) \cdot \Omega_1^2\right) + c_2 T_2(t) \cdot\Omega_2^2. \]