1
$\begingroup$

I'm reading Eschenburg's paper Local convexity and nonnegative curvature — Gromov's proof of the sphere theorem recently. And I meet a little question: Given a point $p\in M$, $N\in T_pM$, we want to construct a hypersurface $S$ around $p$ such that $D_XN=aX$ for all $X\in T_pS$, where $a>0$. He gives the construction as follows: $$ S=\exp_p(\partial B_{-\frac{1}{a}N}(\frac{1}{a})\cap V), $$ where $V$ is a neighborhood such that $\exp\rvert V$ is a diffeomorphism. But I can't verify this conclusion.

$\endgroup$

2 Answers 2

2
$\begingroup$

Think that you have two Riemannian metrics in a neighborhood of the origin in $\mathrm{T}_p$, the first is standard Euclidean and the second is induced by $\mathrm{exp}_p$ from $M$.

These two metrics coincide at the origin up to first order. It is sufficient to conclude that principle curvatures at the origin calculated in both metrics are the same.

$\endgroup$
2
  • 2
    $\begingroup$ Thank you for your excellent answer. And I have one more question: it seems that we just need these two metrics concide at the origin up to the first order because to calculate the principal curvature we only need the Christoffel symbols. $\endgroup$
    – eulershi
    Aug 27, 2022 at 16:12
  • $\begingroup$ @eulershi Right, I will correct it. $\endgroup$ Aug 28, 2022 at 10:06
1
$\begingroup$

Eschenburg's example is overkill. Here is a less elegant version of Anton's answer.

First, choose local coordinates $x=(x^1, \dots, x^n)$ near $p$ so that $x(p) = 0$, $g_{ij}(p) = \delta_{ij}$ and $\partial_kg_{ij}(p) = 0$. Using exponential coordinates for this is overkill. It's easy to prove this directly.

Give a unit $N \in T_pM$, the coordinates can be chosen so that at $p$, $N = \partial_n$.

If $$ S = \{ x^n = f(x^1, \dots, x^{n-1}) \}, $$ where $f(0) = 0$ and $\partial_kf(0)=0$, $1 \le k \le n-1$, then you can show that the second fundamental form of $S$ at $p$ is the Hessian of $f$, $$ \nabla^2_{ij}f(p) = \partial^2_{ij}f(0) + \Gamma^k_{ij}\partial_kf(0). $$

Here, $\Gamma^k_{ij}(0) = 0$ and therefore it suffices to let $$ f(x^1, \dots, x^{n-1}) = \frac{a}{2}((x^1)^2+\cdots (x^{n-1})^2. $$

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.