Can someone help me with this question? Is a qual exam question and I have no idea how to tackle it.

Prove or disprove: Let $f\in \mathcal{C}^\infty(S^n)$ be a smooth function and $x_1, x_2\in S^n$ be two points such that $f(x_1)<0<f(x_2)$. Let $\omega$ be a smooth nonvanishing $n$-form. Then there is a diffeomorphism $\phi\colon S^n \rightarrow S^n$ such that $\int_{S^n} (\phi^* f)\omega =0$.

  • 2
    $\begingroup$ Hint: find a one parameter family if diffeomorphisms and use intermediate value theorem. $\endgroup$ – Moishe Kohan Jun 3 '16 at 20:35
  • 1
    $\begingroup$ Is there any way you could make the title less vague? I don't know differential geometry so I can't, but surely you can give a better overview than "geometry question" $\endgroup$ – pjs36 Jun 3 '16 at 22:48

The short answer was given by studiosus in the comment already.

If $\int f \omega = 0$, set $\phi =\text{id}$ and we are done. Assume that $\int f \omega <0$ (If not, consider $-f$). Since $\omega$ is nonvanishing, We also assume that $\omega$ is positive, so $$\mu(A) := \int_A \omega >0$$ for all nonempty open sets $A$.

Let $x_0 \in \mathbb S^n$ such that $f$ is positive. Then by continuity there is an open ball $U$ centered at $x_0$ so that $f \ge c>0$ on $U$. Let $\{\phi_r\}_{r >0}$ be the one parameter family of diffeomorphisms so that $\phi_1 = \text{id}$, $\phi_s U \subset \phi_r U$ if $s<r$ and $$\bigcup_r \phi^{-1}_r U = \mathbb S^n\setminus \{-x_0\}.$$ (see the construction below) Note that $$g(r)= \int_{\mathbb S^n} (\phi_r^* f)\omega$$ is continuous and $g(1) <0$. Since $f$ is uniformly bounded below by $-m$, we have

$$\begin{split} \int_{\mathbb S^n} (\phi_r^* f)\omega &= \int_{\phi_r^{-1}\ U} (\phi_r^*f) \omega + \int_{\mathbb S^n\ \setminus \phi^{-1}_r\ U} (\phi_r^*f)\omega\\ &\ge c \mu(\phi^{-1}_r U) - m \mu( \mathbb S^n \setminus \phi^{-1}_r U) \\ &\to c \mu(\mathbb S^n) >0 \end{split}$$

as $r\to 0$, where $\mu(A) = \int_A \omega$. Thus there is $r_0$ so that

$$\int_{\mathbb S^n} (\phi_{r_0}^* f)\omega =0.$$

To construct the family of diffeomorphisms, consider the stereographic projection at $\psi : \mathbb S^n \setminus \{-x_0\} \to \mathbb R^n$ at $x_0$. Let $U = \psi^{-1}(B_1)$, where $B_1$ is the unit ball in $\mathbb R^n$. Then define $\phi_r : \mathbb S^n \to \mathbb S^n$ by

$$ \phi_r(x) =\begin{cases} \psi^{-1} (r\psi(x)) & \text{if }x\neq -x_0 \\ x_0 &\text{if }x = -x_0 .\end{cases}$$

Note that $\phi_r$ is smooth even at $-x_0$, since if you use the stereographic projection $\overline \psi$ at $-x_0$, you get (Since $\overline \psi \circ \psi (x) =\frac{x}{|x|^2}$)

$$\phi_r(x) = \begin{cases} \overline\psi^{-1} (\frac 1r \overline \psi(x)) &\text{if } x\neq x_0 \\ x_0 & \text{if }x = x_0.\end{cases}$$

Note that you can also construct a one parameter family of diffeomorphisms using the gradient vector (with the standard metric on $\mathbb S^n$) of the function $$ F(x) = x\cdot x_0,$$ where the dot product is the standard one on $\mathbb R^{n+1}$.

  • $\begingroup$ +1 well done. I interpreted the hint incorrectly; I thought about this idea (using the IVT on the integral), but I couldn't figure out how to do so. $\endgroup$ – Moya Jun 5 '16 at 4:48

Based on the hint:

Since $S^n$ is path connected, there is path $\gamma:[0,1]\to S^n$ such that $\gamma(0)=x_1,\gamma(1)=x_2$. We can take this path to be injective (think about $S^2$ using angular coordinates). Push the vector field $\dfrac{\partial}{\partial t}$ forward using $\gamma$ to a vector field $V$ on $\gamma([0,1])$. Since this is an embedded submanifold which is closed in $S^n$, we can extend this to a compactly supported vector field (also called $V$) defined on $S^n$ (see Lee, chapter 8).

Because $S^n$ is compact, $V$ admits a complete flow, i.e. a smooth 1-parameter family of diffeomorphisms $\Phi(t,\cdot)=\phi_t:S^n\to S^n$. Consider the map $$F=f\circ \Phi(\cdot,x_1):\mathbb{R}\to\mathbb{R}$$Then $F(0)=f(\Phi(0,x_1))=f(x_1)<0$ and $F(1)=f(\Phi(1,x_1))=f(x_2)>0$ (check that $\Phi(1,x_1)=x_2$). By the intermediate value theorem, there is a $t_0\in(0,1)$ with $F(t_0)=0$. Now consider the form $(\phi_{t_0}^{\ast}f)\omega$....

Still working on it.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.