6
$\begingroup$

Let $(M,g)$ be a compact Riemannian manifold without boundary. Fix a point $x\in M$ and $N\ge 2$ large. Then there exists a metric $\tilde g$, conformal to $g$ such that $$ \det \tilde g=1+O(r^N)$$ for $r\le inj(x)$ the injectivity radius at $x$, where $r=|x|$ in the $\tilde g$-normal coordinates. References: Lee, Parker - the Yamabe Problem, THM 5.1 (existence of normal conformal coordinates)

Question: Does (and how) this imply $$ \partial_r^{(i)}\log(\sqrt{\det \tilde g})=O(r^{N-i}).$$

It seems to me that this could come from the Taylor expansion of the determinant of the metric but i cannot find a proper proof. Gursky and Malchiodi in this paper http://arxiv.org/pdf/1401.3216.pdf (Lemma 2.6) use the notation $$ \det \tilde g=1+O^{(3)}(r^N)$$ and they argue $$ \partial_r \log(\sqrt{\det \tilde g})=O^{''}(r^{N-1})$$ but it is not clear to me if i can push it to higher order derivatives and yet why they use that notation since in Lee and Parker nothing like this seems to be proved.

$\endgroup$

1 Answer 1

7
$\begingroup$

The answer to your question is yes. To see why, write $\det \tilde g = 1 + f$, where $f$ is a smooth function satisfying $f=O(r^N)$. Let $(u^i)$ be a smooth coordinate chart centered at $x$. Because $f$ is smooth, one version of Taylor's theorem says that near $x$, we can write $$ f(u) = \sum_{i_1,\dots,i_N} h_{i_1,\dots,i_N}(u) u^{i_1}\dots u^{i_N}, $$ where the functions $h_{i_1,\dots,i_N}(u)$ are all smooth. It follows that every $j$th partial derivative of $f$ is $O(r^{N-j})$ for $0\le j\le N$.

The problem with applying this directly to the question at hand is that $\partial_r$ is not a smooth vector field on a neighborhood of $x$. However, $r\partial_r = \sum_j x^j \partial_{x^j}$ is smooth, and a simple inductive argument shows that $(r\partial_r)^k f=O(r^N)$ for all $k$. Using the fact that $$(r\partial_r)^k = r^k(\partial_r)^k + a_{k-1} r^{k-1}(\partial_r)^{k-1} + \dots + a_1 r \partial_r$$ for some constants $a_1,\dots,a_{k-1}$, another simple induction implies that $r^k(\partial_r)^k f = O(r^N)$, which is equivalent to the result you want.

$\endgroup$
5
  • $\begingroup$ Jack, couldn't you also just use the 1-sided 1-variable Taylor series with respect to $r$ along each geodesic ray? $\endgroup$
    – Deane Yang
    Jun 6, 2014 at 1:29
  • 1
    $\begingroup$ Hi Deane. Yes, I guess you could, although in that form it might not be easy to see that the constant in $O(r^{N-i})$ can be taken to be the same in all directions. But if you pull back to $[0,\infty)\times S^{n-1}$ under spherical coordinates, then it should be an easy consequence of the $1$-variable Taylor expansion in $r$. $\endgroup$
    – Jack Lee
    Jun 6, 2014 at 4:55
  • $\begingroup$ @DeaneYang, what do you mean exactly with 1-sided 1-variable Taylor series? $\endgroup$
    – gin111
    Jun 6, 2014 at 9:43
  • $\begingroup$ gin111, I mean the Taylor series centered at $0$ for a smooth function on the half-closed interval $[0,\delta)$. $\endgroup$
    – Deane Yang
    Jun 6, 2014 at 12:40
  • $\begingroup$ Jack, using spherical co-ordinates is a better way to describe what I was trying to say. $\endgroup$
    – Deane Yang
    Jun 6, 2014 at 12:41

Your Answer

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

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