1
$\begingroup$

Let $M$ be a compact orientable smooth surface in $\mathbb R^3$ with $g>0$, and $K$ be its gaussian curvature. Prove that $$\int_M |K|\geq 8\pi.$$

One finds that, if $M$ is a $g$-torus, the two "extremal" regions with positive curvature are such that the Gauss map $N$ is a diffeomorphism from each of them onto $S^2\setminus\{N,S\}$. Then, since $K=\det\ dN$, the change of variables theorem gives us that $\int_M|K|\geq 4\pi$. Furthermore, applying the same argument to the most extremal negative-curvature regions (where you "put your fingers" when you grab the $g$-torus) we obtain another $4\pi$-term.

Therefore, in a $g$-torus $$\int_M |K|\geq 8\pi.$$

Now, we know that every compact, orientable surface of genus $g$ is diffeomorphic to a $g$ torus. Can this assure us the same property on the integral?

Thank you in advance.

$\endgroup$
2
  • $\begingroup$ If "compact orientable surface" really means "compact smooth surface embedded in $\mathbb R^3$ (as the $N$ in your proposed proof seems to suggest), then this is true. See e.g. math.stackexchange.com/questions/1316830/… $\endgroup$ – Anthony Carapetis Sep 4 '16 at 12:39
  • $\begingroup$ What is a $g$-torus? $\endgroup$ – user99914 Sep 5 '16 at 6:16
4
$\begingroup$

Here is some heuristic: By the Gauss-Bonnet Theorem the total curvature of such a surface $S$ is $$\int_SK\>{\rm d}\omega=4\pi(1-g)\ .\tag{1}$$ Consider now the convex hull $C$ of this surface. Its boundary $\partial C$ is homeomorphic to a sphere and is lying on $S$ where it is not flat. Let $S':=S\cap \partial C$. Then $K\geq0$ on $S'$, and $$\int_{S'} |K|\>{\rm d}\omega=\int_{S'} K\>{\rm d}\omega=\int_{\partial C}K\>{\rm d}\omega=4\pi\ .\tag{2}$$ From $(1)$ and $(2)$ we therefore get $$\int_{S\setminus S'}|K|\>{\rm d}\omega\geq - \int_{S\setminus S'}K\>{\rm d}\omega=-\left(\int_S K\>{\rm d}\omega-\int_{S'}K\>{\rm d}\omega\right)=-\bigl(4\pi(1-g)-4\pi\bigl)=4\pi g\ .$$ It follows that $$\int_S|K|\>{\rm d}\omega\geq4\pi (g+1)\geq8\pi\ .$$

$\endgroup$
3
$\begingroup$

You can deduce this for smooth surfaces, using basic Morse theory, from Proposition 2.14 in Langevin-Shifrin, Polar Varieties and Integral Geometry, Am. J. Math. 104, No. 3 (1982), pp. 553-605. This is an integro-geometric version of the Gauss-Bonnet Theorem, both with and without absolute values, which says (in your case) that $$\int_M |K|\,dA = \int_{\Bbb RP^2} |\mu|(M,L)\,dL,$$ where $|\mu|(M,L)$ is the number of critical points of the orthogonal projection of $M$ onto the line $L$ through the origin in $\Bbb R^3$. (This number is finite for almost all $L$.) Since Morse theory tells us that for $g\ge 1$ we must have at least $4$ critical points, we get $\int_M |K|\,dA \ge 4\cdot 2\pi = 8\pi$.

Even better, it follows from Theorem 1 of Chern-Lashof's On the Total Curvature of Immersed Manifolds, Am. J. Math., 79 (1957), pp. 306-318, which as a simple corollary implies that for a smooth compact surface $M$ immersed in $\Bbb R^3$ the total absolute curvature is at least $2\cdot 4\pi = 8\pi$.

Both theorems apply in much greater generality (both to higher codimensions, and the former to compact submanifolds with boundary).

$\endgroup$
2
$\begingroup$

This is not true for abstract surfaces (i.e. $2$-dimensional manifolds). Indeed the flat torus satisfies all the assumptions and has zero curvature everywhere.

For surfaces in $\mathbb{R}^3$, it is still not true if you allow less than perfectly smooth surfaces: the flat torus can be isometrically $C^1$-embedded in $\mathbb{R}^3$. Because of this, I would expect the statement to be false also for $C^\infty$-manifolds, but I don't have a proof.

$\endgroup$
4
  • $\begingroup$ Oh. So the statement "compact+orientable+genus>0$\Longrightarrow$ integral$\geq8\pi$" is false? $\endgroup$ – W. Rether Sep 4 '16 at 10:25
  • 1
    $\begingroup$ @W.Rether Yes, it is false. (Modulo mistakes on my part, of course.) $\endgroup$ – Daniel Robert-Nicoud Sep 4 '16 at 11:21
  • 1
    $\begingroup$ If you're assuming an embedding then this doesn't settle the question in the smooth case, since the flat torus can only be $C^1$ embedded. $\endgroup$ – Anthony Carapetis Sep 4 '16 at 11:47
  • $\begingroup$ @AnthonyCarapetis Ah, right. I'll edit my answer a bit. $\endgroup$ – Daniel Robert-Nicoud Sep 4 '16 at 12:15

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.