How to show the volume preserving mean curvature flow preserve volume?

As picture below, I want to show the volume enclosed by $M_t$ is constant. Because $$V(t)=\frac{1}{n}\int_{M_t} F(x,t)\cdot \nu(x,t) dS$$ And $\partial_t\sqrt g=-H(H-h)\sqrt g$, so $$\frac{dV(t)}{dt}=\frac{1}{n}\int_{M_t} [\partial_tF(x,t)\cdot \nu(x,t) +F(x,t)\cdot \partial_t\nu(x,t)-H(H-h)F(x,t)\cdot \nu(x,t)] dS \\ =\frac{1}{n}\int_{M_t} [(h(t)-H(x,t))\nu(x,t)\cdot \nu(x,t) +F(x,t)\cdot \partial_t\nu(x,t)-H(H-h)F(x,t)\cdot \nu(x,t)] dS \\ =\frac{1}{n}\int_{M_t}[F(x,t)\cdot \partial_t\nu(x,t)-H(H-h)F(x,t)\cdot \nu(x,t)] dS$$ Then, to show $\int_{M_t}[F(x,t)\cdot \partial_t\nu(x,t)-H(H-h)F(x,t)\cdot \nu(x,t)] dS =0$

Because $\nu(x,t)$ is outer unit normal vector. We have $\langle\nu(x,t),\nu(x,t)\rangle=1$. Then $$\langle \partial_t\nu(x,t),\nu(x,t) \rangle=0$$ Besides, $\{ \partial_iF \}$ is a basis of $T_xM$, so we can assume $$\partial_t\nu (x,t)=v^i\partial_iF$$ Then, we have $$\partial_t\nu= \langle \partial_t\nu, \partial_jF\rangle \partial_j F g^{ij}$$ Because $\langle \nu, \partial_jF\rangle = 0$, we have $\langle \partial_t\nu, \partial_jF\rangle =-\langle \nu, \partial_t\partial_jF\rangle$. Then $$\partial_t\nu=\langle \nu,\partial_i(H\nu) \rangle \partial_jF g^{ij}=\partial_iH \partial_jFg^{ij}=\nabla H ~~~~~\text{gradient on M}$$ So $$F(x,t)\cdot \partial_t\nu(x,t)-H(H-h)F(x,t)\cdot \nu(x,t) \\ =F\cdot\nabla H-H(H-h)F(x,t)\cdot \nu(x,t)$$ Then, I get stuck.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

• You're pretty much there. If the volume is constant in time then $V(0)=V(t)$ for any $t\in\mathbb{R}$. Differentiating this (since $V(0)$ is independent of $t$) gives $\frac{dV(t)}{dt} =0$. – AloneAndConfused Mar 13 '17 at 10:41
• @AloneAndConfused how to show $\frac{dV(t)}{dt} =0$ ? – lanse7pty Mar 19 '17 at 6:36
• It is a simple deduction from the first variation formula and divergence theorem, then the direct definition of $h(t)$. – AloneAndConfused Mar 20 '17 at 9:10
• @AloneAndConfused Could you detail explain it ? In my calculation, there always be $\nabla H$, $H$ is mean curvature, I don't know how to deal $\nabla H$. – lanse7pty Mar 22 '17 at 5:45

This is very simple, so I'm not going to try to read your long calculation. If you move a surface in the outwards normal direction at speed $f(x)$, then the enclosed volume changes by $dV/dt = \int f dA$. Since $h$ is defined as the average of $H$, $h-H$ has average $0$, and thus $dV/dt = \int (h - H) dA = 0$.
• Could you show $dV/dt = \int f dA$ ? I fail to show it. – lanse7pty Mar 19 '17 at 7:52