I'm currently reading through Schoen & Yau's 1979 proof of the positive-mass theorem and am trying to understand the following statement on p. 55 of the publication (page 11 of the proof / PDF):

Remark 2.1: The Cohn-Vossen inequality says that $\int_S K \leq 2\pi \chi(S)$, where [$K$ is the Gauss curvature and] $\chi(S)$ is the Euler characteristic of $S$. Combining this with (2.18) we see immediately that $S$ is homeomorphic to $\mathbb{R}^2$

Here, (2.18) refers to the inequality $\int_S K > 0$ derived just before.

How can I see that the surface $S$ is homeomorphic to $\mathbb{R}^2$? The immediate implication of (2.18) obviously is that $\chi(S) > 0$, so $\chi(S) \geq 1$. Apart from that, I know that $S$ is non-compact but I'm unsure about other properties$^\dagger$ of $S$ as my understanding of its construction is somewhat limited so far. Is it those properties that imply that the Betti numbers $b_i$ vanish for $i \geq 1$?


$^\dagger$ For instance, I suspect that $S$ is boundaryless, connected & simply connected and also a closed subset of the ambient manifold $N$ but since I only have a rough idea of how $S$ is constructed as a limit of surfaces $S_\sigma$ (namely by representing the minimal surfaces $S_\sigma$ locally as graphs in the tangent space (using normal coordinates) and using Arzelà-Ascoli for finding the limit), I'm having trouble coming up with rigorous proofs of any properties of $S$ that could help prove the claim.


A noncompact surface has no homology groups above dimension 1. So $\chi(S) = 1 - \dim H_1(S)$. If $\chi(S) = 1$, then in fact $H_1(S) = 0$.

A noncompact surface with finitely generated homology groups is diffeomorphic to some $S_{g,n}$, the genus $g$ surface with $n>0$ points removed. This is proved using the similar classification of compact surfaces, and using a compact exhaustion of your manifold. I wrote something about the simplest case in the second part of the post here but it's not hard to extrapolate to the general case.

Now $H_1(\Sigma_{g,n}) = \Bbb Z^{2g+n-1}$ so long as $n > 0$. If this is zero (so that the Euler characteristic is 1), then necessarily $g=0$ and $n=1$. Puncturing the 2-sphere leaves you with $\Bbb R^2$, as desired.

  • $\begingroup$ Months later, I'm revisiting this question because I just realized the following: 1) I don't think I can assume $S$ to be simply connected (which, if I understand you correctly, you used in your classification). 2) The Cohn-Vossen inequality used by Schoen & Yau already requires the surface to have finite Euler characteristic. But we need the Cohn-Vossen inequality to derive $χ(S) \geq 1$ and $H_1(S) = 0 < \infty$ in the first place. Can the finiteness of $H_1(S)$ and $χ$ be deduced in a different way? $\endgroup$ – balu Nov 7 '19 at 17:14
  • $\begingroup$ A quick update: It turns out that Huber's theorem (used later in Schoen & Yau's proof) says that finite total curvature of $S$ (which was shown earlier in the paper) implies that it has finite Euler characteristic. This can then be used to justify using the Cohn-Vossen inequality. My question regarding simple-connectedness is still open, though. $\endgroup$ – balu Nov 14 '19 at 12:37

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.