1
$\begingroup$

Denote by $S^n$ the unit sphere in $\mathbb{R}^{n+1}$, and consider the coordinate function $x_{n+1}$ on it, i.e. the function $(x_1, \ldots, x_{n+1}) \mapsto x_{n+1}$. Denoting by $\mathrm{Hess}(x_{n+1})$ the Hessian of $x_{n+1}$ with respect to the round metric $g$ on $S^n$, it is claimed in the paper that I am reading that $\mathrm{Hess}(x_{n+1}) = -x_{n+1} \cdot g$.

I have to admit that I have trouble varifying this identity. How can one prove it?

edit: As Deane suggests, I am posting my computation. I am parametrizing the upper hemisphere as $$ \phi(x_1, \ldots, x_n) := (x_1, \ldots, x_n, \sqrt{1-|x|^2}), $$ where $|x| := |(x_1, \ldots, x_n)|$ is the usual Euclidean norm. The coordinate function $f := x_{n+1}$ is then given by $f(x_1, \ldots, x_n) := \sqrt{1-|x|^2}$ and therefore $$ \partial_i \partial_j f = \frac{-x_i x_j}{(1-|x|^2)^{3/2}} $$ for $i \not= j$ (let me just treat this case here). On the other hand, to compute $g_{ij}$ we first compute $\partial_i \phi = \frac{-x_i}{(1-|x|^2)^{1/2}}$ and therefore $$ g_{ij} = g(\partial_i \phi,\partial_j \phi) = \frac{x_i x_j}{1-|x|^2} $$ for $i \not= j$. For $-f\cdot g$ we finally get, in the case $i \not= j$, $$ (-f \cdot g)_{ij} = \frac{-x_i x_j}{(1-|x|^2)^{1/2}}. $$ This is, as I said in the comments, off by the factor $\frac{1}{1-|x|^2}$ from the claimed answer.

edit 2: I think I found my error. I forgot to evaluate the matrix $(\partial_i\partial_j f)_{ij}$ on the basis vectors. And indeed, I get $$(\partial_i\partial_j f)_{ij}(\partial_i \phi,\partial_j \phi) = \frac{-x_i x_j}{(1-|x|^2)^{1/2}}$$ (for $i \not= j$) as claimed.

$\endgroup$
3
  • 1
    $\begingroup$ Do it for the $2$-sphere first. Write the upper half of the sphere as a graph, and compute the Hessian of $z$ with respect to the coordinates $x, y$. The same calculation also works for the lower half. From here, it should be clear how to do it on an $n$-sphere. $\endgroup$
    – Deane
    Commented Apr 24 at 13:53
  • 1
    $\begingroup$ Thanks. I tried to do the computation as you suggest and I almost did it ... I am just off by a factor of 1/(1-|x|^2) between the two expressions, but this is probably just due to a missing normalization that I forgot to do somewhere. I will try again ... $\endgroup$
    – AlexE
    Commented Apr 25 at 19:54
  • $\begingroup$ Post your calculation. Someone will help you figure out where the error is. $\endgroup$
    – Deane
    Commented Apr 25 at 20:16

0

You must log in to answer this question.

Browse other questions tagged .