In order to obtain the upper bound of
\(|\nabla h|\) by using the maximum principle for (
2.2), we need to choose
ϵ,
δ such that the coefficient of
\(\frac{|\nabla h|^{4}}{h^{2}}\) in (
2.2) is positive. That is, we need
$$ \frac{(\epsilon-1)^{2}}{n\epsilon^{2}}-\frac{\epsilon-1}{\epsilon}+\frac {2\delta(\epsilon-1)}{n\epsilon}>0. $$
(2.11)
In particular, by choosing
\(\epsilon=\frac{4}{5}\) and letting
\(\delta\rightarrow\frac{1}{2}\), we find that the inequality (
2.1) holds and (
2.2) becomes
$$ \begin{aligned}[b] \frac{1}{2}\Delta_{f}| \nabla h|^{2}\geq{}&\frac{4n-3}{16n}\frac{|\nabla h|^{4}}{h^{2}}- \frac{1}{4}\frac {\nabla h}{h}\nabla\bigl(|\nabla h|^{2}\bigr) \\ &-\bigl[a+\tilde{b}\epsilon+(n-1)K+a\tilde{L}\bigr]|\nabla h|^{2}. \end{aligned} $$
(2.12)
As in [
8], we define a cut-off function
\(\psi\in C^{2}([0,+\infty))\) by
$$ \psi(t)= \textstyle\begin{cases} 1, & t\in[0,R];\\ 0, & t\in[2R,+\infty], \end{cases} $$
(2.13)
satisfying
\(\psi(t)\in[0,1]\) and
$$ -\frac{C}{R}\leq\frac{\psi'(t)}{\sqrt{\psi}}\leq0,\qquad \big| \psi''(t)\big|\leq\frac {C}{R^{2}}, $$
(2.14)
where
C is a positive constant. Let
$$\phi=\psi\bigl(d(x,p)\bigr). $$
Using Eq. (2.19) in [
8] (see Eq. (4.5) in [
5] or [
12, Theorem 3.1]), we obtain
$$ \Delta_{f}\phi\geq-\frac{C\beta}{R}- \frac{C(n-1)K(2R-1)}{R}-\frac{C}{R^{2}} $$
(2.15)
and
$$ \frac{|\nabla\phi|^{2}}{\phi}\leq\frac{C}{R^{2}}. $$
(2.16)
Denote by
\(B_{p}(R)\) the geodesic ball centered at
p with radius
R. Let
\(G=\phi|\nabla h|^{2}\). Assume
G achieves its maximum at the point
\(x_{0}\in B_{p}(2R)\) and assume
\(G(x_{0})>0\) (otherwise the proof is trivial). Then, at the point
\(x_{0}\),
$$\Delta_{f} G\leq0,\qquad \nabla\bigl(|\nabla h|^{2}\bigr)=- \frac{|\nabla h|^{2}}{\phi} \nabla\phi $$
and
$$ \begin{aligned} [b]0\geq{}& \Delta_{f}G \\ ={}&\phi\Delta_{f}\bigl(|\nabla h|^{2}\bigr)+|\nabla h|^{2}\Delta_{f}\phi+2\nabla\phi\nabla |\nabla h|^{2} \\ ={}&\phi\Delta_{f}\bigl(|\nabla h|^{2}\bigr)+ \frac{\Delta_{f}\phi}{\phi}G -2\frac{|\nabla\phi|^{2}}{\phi^{2}}G \\ \geq{}&\frac{\Delta_{f}\phi}{\phi}G -2\frac{|\nabla\phi|^{2}}{\phi^{2}}G+2\phi \biggl[\frac{4n-3}{16n} \frac{|\nabla h|^{4}}{h^{2}}-\frac{1}{4}\frac{\nabla h}{h}\nabla\bigl(|\nabla h|^{2}\bigr) \\ &-\bigl[a+\tilde{b}\epsilon+(n-1)K+a\tilde{L}\bigr]|\nabla h|^{2} \biggr] \\ ={}&\frac{\Delta_{f}\phi}{\phi}G -2\frac{|\nabla\phi|^{2}}{\phi^{2}}G+\frac {4n-3}{8n} \frac{G^{2}}{\phi h^{2}}+\frac{G}{2\phi}\nabla\phi\frac{\nabla h}{h} \\ &-2\bigl[a+\tilde{b}\epsilon+(n-1)K+a\tilde{L}\bigr]G, \end{aligned} $$
(2.17)
where in the second inequality, we used (
2.12). Multiplying both sides of (
2.17) by
\(\frac{\phi}{G}\), we obtain
$$ \begin{aligned}[b] \frac{4n-3}{8n}\frac{G}{h^{2}} \leq{}&{-}\frac{1}{2}\nabla\phi\frac{\nabla h}{h}+2\bigl[a+\tilde{b} \epsilon+(n-1)K+a\tilde{L}\bigr]\phi \\ &-\Delta_{f}\phi+2\frac{|\nabla\phi|^{2}}{\phi}. \end{aligned} $$
(2.18)
Substituting the Cauchy inequality
$$\begin{aligned}-\frac{1}{2}\nabla\phi\frac{\nabla h}{h}\leq{}& \frac{1}{2}|\nabla\phi|\frac{|\nabla h|}{h} \\ \leq{}&\frac{n}{4n-3}\frac{|\nabla\phi|^{2}}{\phi} +\frac{4n-3}{16n}\phi \frac{|\nabla h|^{2}}{h^{2}} \\ ={}&\frac{n}{4n-3}\frac{|\nabla\phi|^{2}}{\phi} +\frac{4n-3}{16n}\frac{G}{h^{2}} \end{aligned} $$
into (
2.18) gives
$$ \begin{aligned}[b] \frac{4n-3}{16n}\frac{G}{h^{2}} \leq{}&2\bigl[a+\tilde{b}\epsilon+(n-1)K+a\tilde {L}\bigr]\phi-\Delta_{f} \phi +\frac{9n-6}{4n-3}\frac{|\nabla\phi|^{2}}{\phi} \\ \leq{}&2\bigl[a+\tilde{b}\epsilon+(n-1)K+a\tilde{L}\bigr]+\frac {C_{1}[(n-1)K(2R-1)+\beta]}{R}+ \frac{C_{2}}{R^{2}}, \end{aligned} $$
(2.19)
where
\(C_{1}\),
\(C_{2}\) are two positive constants depending on
n. Hence, on
\(B_{p}(R)\) with
\(R>1\), it follows from (
2.19) that
$$ \begin{aligned}[b] \frac{4n-3}{16n}G(x)\leq{}& \frac{4n-3}{16n}G(x_{0}) \\ \leq{}&h^{2}(x_{0}) \biggl[2\bigl[a+\tilde{b} \epsilon+(n-1)K+a\tilde{L}\bigr] \\ &+\frac{C_{1}[(n-1)K(2R-1)+\beta]}{R}+\frac{C_{2}}{R^{2}} \biggr]. \end{aligned} $$
(2.20)
In particular, the estimate (
2.20) gives
$$ \begin{aligned} |\nabla u|^{2}\leq{}&C A^{2} \biggl[\max \biggl\{ \frac{4}{5}b+a \biggl(1+ \frac {4}{5}L \biggr),0 \biggr\} +K+\frac{|\beta|+1}{R} \biggr], \end{aligned} $$
(2.21)
which finishes the proof of Theorem
1.1.