2
$\begingroup$

Let M be an n-dimensional Riemannian manifold without boundary, with sectional curvature $\geqslant -1$. For a point $p\in M$, suppose there exist $l, \delta>0$, $x,y \in M$ with $d(p,x),d(p,y)>l$ and a geodesic $px$ and $py$ with angle $\angle xpy>\frac{\pi}{2}+\delta$. Let $q$ be a point on geodesic $px$ or $py$, Question: is there $r>0$, which depends only on $n,l,\delta$ such that $B_q(r)$ is homeomorphic to $B_p(r)$?

Equivalently, we can state the question in the following way:

Let $M_i$ be a sequence of Riemannian manifolds with $sec \geqslant -1$ and diameter $\leqslant D$. Suppose $(M_i,p_i)$ Gromov-Hausdorff converge (possibly collapse) to $(X,p)$ (we know it's an Alexandrov space). Suppose there exist $l>0, \delta>0$, $x,y\in X$ with $\angle xpy> \frac{pi}{2}+\delta$. lift $x,y$ to $M_i$, we get $x_i,y_i\in M_i$. with$\angle x_i p_i y_i >\frac{\pi}{2}+\delta$. Let $q_i$ be a point on geodesic $p_ix_i$ or $p_iy_i$. Question: Is there $r>0$, such that such that $B_{q_i}(r)$ is homeomorphic to $B_{p_i}(r)$?

$\endgroup$

1 Answer 1

1
$\begingroup$

There is a simple counterexample. Consider warped product $M=\mathbb{R}{\times}_{\exp}\mathbb{S}^1$; that is, $\mathbb{R}{\times}\mathbb{S}^1$ with the metric $(\begin{smallmatrix}1&0\\ 0&e^{2\cdot x}\end{smallmatrix})$ at $(x,y)\in\mathbb{R}{\times}\mathbb{S}^1$. Note that for any $r>0$ there are arbitrary close points $p,q$ on a line $\mathbb{R}{\times}\{y\}$ such that $B_q(r)$ is homeomorphic to a disc and $B_p(r)$ is homeomorphic to a cylinder.

However by Morse lemma, there is a neighborhood of $p$ with size of order $\delta\cdot l$ is a homeomorphic to a product $(a,b)\times L$, where the first coordinates is the distance function $\mathrm{dist}_x$. (So morally your are right.)

$\endgroup$
2
  • $\begingroup$ Thank you. For my first question, you give a counter example, which is a non-compact manifold. What about my second question? i.e. when $M_i$ have uniformly bounded diameter collapse to X. For $r=\frac{\delta l}{1000}$, $|q_ip_i|=10r$, then $B_r(p_i),B_r(p_i)$ are in the neighborhood you mentioned, do we have $B_r{q_i}$ is homemorphic to $B_r(p_i)$? It seems to me it's not easy to construct a homeo. $\endgroup$ Jun 16, 2019 at 13:22
  • $\begingroup$ @mathmetricgeometry one can make it compact if needed and a similar construction gives a counterexample to the second question. $\endgroup$ Jun 16, 2019 at 14:43

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.