New monotonicity formulas for Ricci curvature and applications. I (Q1932329)

The author obtains three new monotonicity formulas for manifolds with a lower Ricci curvature bound and shows that they are connected to rate of convergence to tangent cones. Let \(M^n\) be a smooth complete \(n\)-dimensional manifold with \(n\geq 3\), and assume that it is non-parabolic, i.e. it admits a positive Green function. Let \(G\) denote the minimal positive Green function on \(M\) and set \(b=G^{1/(2-n)}\). The author defines two non-negative functions, the `area' \(A(r)\) and the `volume' \(V(r)\), by \[ A(r)=r^{1-n}\int_{b=r}|\nabla b|^3\,d\text{Area}\qquad\text{and}\qquad V(r)=r^{-n}\int_{b\leq r}|\nabla b|^4\,d\text{Vol}. \] The first monotonicity formula is \[ (A-2(n-1)V)'={r^{-1-n}\over2}\int_{b\leq r}\Big(\Big|\text{Hess}_{b^2}-{\Delta b^2\over n}g\Big|^2+\text{Ric}(\nabla b^2,\nabla b^2)\Big)\,d\text{Vol}. \] From here, two consequences are deduced if the Ricci curvature is non-negative: (1) \(A(r)-\text{Vol}(\partial B_1(0))\geq 2(n-1)V(r)-\text{Vol}(B_1(0))\). If for some \(r>0\) we have equality then the set \(\{x:b(x)\leq r\}\) is isometric to a ball of radius \(r\) in \({\mathbb R}^n\). (2) \(A(r_2)-2(n-1)V(r_2)\geq A(r_1)-2(n-1)V(r_1)\), for \(r_2>r_1>0\), and equality holds if and only if the set \(\{x:b(x)\leq r_2\}\) is isometric to a ball of radius \(r_2\) in \({\mathbb R}^n\). The second monotonicity formula is \[ (r^{2-n}[A-\text{Vol}(\partial B_1(0))])'={r^{1-n}\over2}\int_{b\leq r}\Big(\Big|\text{Hess}_{b^2}-{\Delta b^2\over n}g\Big|^2+\text{Ric}(\nabla b^2,\nabla b^2)\Big)b^{-n}\,d\text{Vol}, \] which is equivalent to \[ (A-(n-2)V_\infty)'={1\over 2r}\int_{b\leq r}\Big(\Big|\text{Hess}_{b^2}-{\Delta b^2\over n}g\Big|^2+\text{Ric}(\nabla b^2,\nabla b^2)\Big)b^{-n}\,d\text{Vol}, \] where \(V_\infty\) stands for \[ V_\infty(r)=\int_{1\leq b\leq r}(|\nabla b|^2-1)|\nabla b|^2b^{-n}d\text{Vol}. \] If the Ricci curvature is non-negative, then the author gives the following consequence: \[ A(r_2)-(n-2)V_\infty(r_2)\geq A(r_1)-(n-2)V_\infty(r_1),\quad r_2>r_1>0, \] and equality holds if and only if the set \(\{x:b(x)\leq r_2\}\) is isometric to a ball of radius \(r_2\) in \({\mathbb R}^n\). The third monotonicity formula says: \[ r_2^{3-n}A'(r_2)-r_1^{3-n}A'(r_1)={1\over2}\int_{r_1\leq b\leq r_2}\Big(\Big|\text{Hess}_{b^2}-{\Delta b^2\over n}g\Big|^2+\text{Ric}(\nabla b^2,\nabla b^2)\Big)b^{2-2n}\,d\text{Vol}. \] As a consequence, if the Ricci curvature is non-negative then \[ A'(r)=-{r^{n-3}\over2}\int_{b\geq r}\Big(\Big|\text{Hess}_{b^2}-{\Delta b^2\over n}g\Big|^2+\text{Ric}(\nabla b^2,\nabla b^2)\Big)b^{2-2n}\,d\text{Vol}. \] It is well known that Perelman's monotonicity formula for the Ricci flow is closely related to the sharp gradient estimate of Li-Yau for the heat kernel; the author obtains a similar result by showing that the new monotonicity formulas are related to a sharp gradient estimate for the Green function. In fact, he proves that if \(M\) has non-negative curvature with \(n\geq 3\), then \(|\nabla b|\leq 1\). Moreover, if equality holds at any point of \(M\), then \(M\) is flat Euclidean space \({\mathbb R}^n\). The author also obtains the following sharp asymptotic gradient estimates for the Green function on manifolds with non-negative Ricci curvature and \(n\geq3\): \[ \lim_{r\to\infty}\sup_{M\backslash B_r(x)}|\nabla b|=\Big({V_M\over\text{Vol}(B_1(0))}\Big)^{1/(n-2)}, \] or equivalently \[ (n-2)\lim_{r\to\infty}\sup_{M\backslash B_r(x)}{|\nabla G|\over G^{(n-1)/(n-2)}}=\Big({V_M\over\text{Vol}(B_1(0))}\Big)^{1/(n-2)}. \] In Section 4 the author defines the scale-invariant distance from a metric ball \(B_r(x)\) in a metric space \((X,d_X)\) to the space of cones, and gives some results on the uniqueness of tangent cones. For example, if \(\alpha>1\) and \[ \int_1^\infty{\Theta_r^2\over r|\log r|^{-\alpha}}dr<\infty, \] then the tangent cone at infinity is unique, likewise for tangent cones at a point. Here, \(\Theta_r(x)>0\) stands for the infimum of all \(\Theta>0\) such that \[ d_{GH}(B_r(x),B_r(v))<\Theta r, \] where \(B_r(v)\) is a ball in the cone \(C(Y)\) and \(v\) is the vertex of the cone. In the last section the author defines the weighted distance to the space of cones.