# Does the curvature determine the metric?

Here I asked the question whether the curvature deterined the metric. Since I am unfortunately completely new to Riemannian geometry, I wanted to ask, if somebody could give and explain a concrete example to me, as far as the following is concerned:

At the MO page (as cited above) I got the following answer to the question

Given a compact Riemannian manifold M, are there two metrics g1 and g2, which are not everywhere flat, such that they are not isometric to one another, but that there is a diffeomorphism which preserves the curvature? If the answer is yes: Can we chose M to be a compact 2-manifold?

On the positive side, if $M$ is compact of dimension $\ge 3$ and has nowhere constant sectional curvature, then combination of results of Kulkarni and Yau show that a diffeomorphism preserving sectional curvature is necessarily an isometry.

Concerning 2-dimensional counter-examples: First of all, every surface which admits an open subset where curvature is (nonzero) constant would obviously yield a counter-example. Thus, I will assume now that curvature is nowhere constant. Kulkarni refers to Kreyszig's "Introduction to Differential Geometry and Riemannian Geometry", p. 164, for a counter-example attributed to Stackel and Wangerin. You probably can get the book through interlibrary loan if you are in the US.

I looked up the example in Kreyszig's "Introduction to Differential Geometry and Riemannian Geometry", p. 164:

If we rotate the curve $x_3=\log x_1$ about the $x_3$-axis in space, we obtain the surface of revolution $X(u_1,u_2)=(u_2\cos(u_1), u_2\sin(u_1),\log(u_2))$, $u_2>0$. This is diffeomorphic to the helicoid $X(u_1,u_2) =(u_2\cos(u_1),u_2\sin(u_1),u_1)$.

I think, these manifolds are not compact (but I assumed compactness of the manifold in my question on MO). I don't understand, how to manipulate this example in order to get a compact manifold.

• @Zhen: Thank you for improving the layout of the question. Jun 23, 2012 at 22:28
• In order to not make things even more confusing than they are, I split the question into two parts. The second part is here. Jun 24, 2012 at 16:38

Here's another example.

First, imagine a short cylinder $S^1\times [0,1]$ and a long cylinder $S^1\times [0,10^{10}]$, but with the same radii. Smoothly cap off both ends of the cylinders in the same way using spaces homeomorphic to discs.

The resulting manifolds are both homeomorphic to $S^2$, are not isometric (since one has a much larger diameter than the other), but there is a curvature preserving diffeomorphism between them.

To see this, just convince yourself there is a diffeomorphism $f:S^1\times[0,1]\rightarrow S^1\times [0,10^{10}]$ with the property that $f$ is an isometry when restricted to $[0,\frac{1}{4}]$ and $[\frac{3}{4},1]$. This condition allows you to extend $f$ to a curvature preserving diffeo of both compact manifolds.

• +1 Indeed, I would give the same example but I was concerned that "not everywhere flat" might turn out to mean "nowhere flat / nowhere zero curvature". MSE questions are prone to mutations of this sort. :)
– user31373
Jun 24, 2012 at 0:07
• @Leonid: You're absolutely right that they're prone to morph. I should have also said that this example comes from Do Carmo's book "Riemannian Geometry". Jun 24, 2012 at 1:57
• Does the curvature determine the metric up to conformal class? Sep 3, 2020 at 8:54
• @C.F.G.: I have never studied conformal geometry, so I have no idea. Sep 3, 2020 at 11:08
• @C.F.G Concerning the question "does curvature determines the conformal class", I think the answer is no. There are infinitely many non-isometric hyperbolic metrics on surfaces of genus $\geqslant 2$ (see Teichmüller theory), and any Riemannian metric is conformal to exactly one of them (see Riemannian Geometry, Gallot-Hulin-Lafontaine, 2nd edition, Theorem 3.163 page 198) Mar 22, 2022 at 13:35

Imagine a sphere with several identical bumps. More precisely, the surface given in spherical coordinates by $\rho=F(\theta,\phi)$ where $F$ is equal to one except in some places that are far one from another. To be concrete, suppose there are four bumps uniformly distributed along the equator.

You can grab one of these bumps and slide it to the North Pole, or, if you wish, arrange them at the vertices of a tetrahedron. There is a diffeomorphism that maps the initial surface onto the new one, and preserves curvature. Namely, the diffeomorphism is the isometry of corresponding bumps, extended in an arbitrary way to the rest of the sphere (where the curvature is constant).

The surfaces will not be isometric in general. For example, the diameter of the surface is greater when you have two bumps exactly opposite each other on the sphere.