3
$\begingroup$

Consider surface of revolution $$S(t,\theta)=(t,f(t)\cos\theta,f(t)\sin\theta)$$ If we want to calculate the principal curvatures, first we calculate the coordinate tangent vectors, then by cross product and normalisation to get unit normal vector $N$, finally using $S\partial_i=-D_{\partial_i}N$ to calculate the shape operator. This is how I do to find principal curvatures for two dimensional surface after parametrization. But now If I revolve the graph $(t,f(t))$ along $x$-axis in $R^{n+1}$, then we get $$S(t,w)=(t,f(t)w)=(t,f(t)\cos\theta_1,f(t)\sin\theta_1\cos\theta_2,\cdots)$$ where $w$ is the direction in $S^n$. Now I am kinda stuck in how to calculate the shape operator of this surface. But I have some intuition that the principal curvatures should be the same in all angular directions.
Could anyone give me how to proceed? Since finally I am interested in its sectional curvature and mean curvature.

$\endgroup$
3
$\begingroup$

You don't want to do this by brute force computation. Your intuition is correct: The principal directions are the tangent vector to the profile curve (meridian) and any basis vectors for the tangent space to the sphere.

You can verify this by looking at the covariant derivative of the unit normal in these directions. As you move along a profile curve, you are considering a plane curve and the covariant derivative of the normal must be orthogonal to the normal, hence tangent to the curve. Next, by rotational symmetry, as you move along a latitude sphere, you can see that the normal vector to the hypersurface has two components, one in the direction of the normal to the sphere and one pointing along the first axis; by symmetry, the magnitudes in those directions are constant. Thus, when you compute the covariant derivative in the direction of $V$, tangent to the latitude sphere, the first component gives you a multiple of the $V$ and the second component drops out.

Using the usual Meusnier's Theorem analysis that works for a surface in $\Bbb R^3$ [see, for example, pp. 51-2 of my text] you can figure out the principal curvatures. Up to sign, the principal curvature in the direction of the profile curve is the curvature of the curve. More interestingly, by Meusnier's Theorem, the normal curvature in any direction $V$ tangent to the sphere will be the $1/f(t)$ (the curvature of a great circle with tangent vector $V$) multiplied by $\cos\phi$, where $\phi$ is the angle between the normal to the hypersurface and the principal normal of that great circle. The principal normal points toward the center of the sphere, and so with a drawing you can check that (again up to supplements) $\phi$ is the angle between $(-f'(t),1)$ and $(0,1)$ in $\Bbb R^2$. That is, $\cos\phi = 1/\sqrt{1+f'(t)^2}$.

Finally, then, the principal curvatures (up to sign) of your hypersurface are $\dfrac{f''(t)}{(1+f'(t)^2)^{3/2}}$ and $\dfrac 1{f(t)\sqrt{1+f'(t)^2}}$ (with multiplicity $n-1$).

$\endgroup$
  • $\begingroup$ Nice explanation. Thank you, $\endgroup$ – STUDENT Aug 27 '18 at 22:05
  • $\begingroup$ You're welcome :) When you're satisfied that you have no other questions, please accept the answer, so that this question will not stay on the unanswered list. $\endgroup$ – Ted Shifrin Aug 28 '18 at 0:05

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.