I've read that $\mathbb{S}^2 \times \mathbb{R}$ is one of the model geometries of Thurston which has non constant curvature. I took it to mean that the manifold $(\mathbb{S}^2 \times \mathbb{R}, g)$ has non constant sectional curvature (where $g$ is the standard product metric) and tried to compute its sectional curvature. Here's what I used:

Let $(M, g_1)$ and $(N, g_2)$ be two Riemannian manifolds with curvature tensors $R_1$ and $R_2$. Using that for each $(p, q) \in M \times N$, $T_pM \oplus T_q N \cong T_{(p, q)}(M \times N)$, for each $X, Y, Z \in \Gamma(T(M \times N))$, we have:

  • $R(X, Y)Z = R_1(X_1, Y_1)Z_1 + R_2(X_2, Y_2)Z_2$
  • $\text{Rm}(X, Y, Z, W) = \langle R(X,Y)Z, W \rangle$

where $X = (X_1, X_2)$, with $X_1 \in \Gamma(TM)$, $X_2 \in \Gamma(TN)$ and analogously for $Y$ and $Z$, and the metric on $M \times N$ is given by:

$$g^{M \times N}_{(p, q)}(X, Y) = g^{M}_{p}(X_1, Y_1) + g^{N}_{q}(X_2, Y_2)$$

Denoting by $R_1$ the curvature tensor for the sphere $\mathbb{S}^2$ and by $R_2$ the one for the real line, it's obvious that $R_2 \equiv 0$. Now, let $X, Y$ be an orthonormal basis for a $2$ plane contained in some tangent space of $\mathbb{S}^2 \times \mathbb{R}$. We have:

$$\begin{align}R(X,Y)Y &= R(X_1 + X_2, Y_1 + Y_2)( X_1 + X_2)\\ &= R(X_1, Y_1)X_1 + R(X_2, Y_1)X_1 + R(X_1, Y_2)X_1 + R(X_2, Y_2)X_1& \\ &+R(X_1, Y_1)X_2 + R(X_2, Y_1)X_2 + R(X_1, Y_2)X_2 + R(X_2, Y_2)X_2 \\ &= R_{1}(X_1, Y_1)X_1 \end{align}$$

since all the other terms disappear, where we're using that $R_2 \equiv 0$. Then:

$$\begin{align}K(X, Y) &= \langle R(X, Y)Y, X \rangle \\ &= \langle R_1(X_1, Y_1)Y_1, X_1 \rangle + \langle R_1(X_1, Y_1)Y_1, X_2 \rangle \\ &= \langle R_1(X_1, Y_1)Y_1, X_1 \rangle = 1 \end{align}$$

because $\mathbb{S}^2$ has constant sectional curvature equal to $1$. So we have that $\mathbb{S}^2 \times \mathbb{R}$ has constant sectional curvature as well.

Where did I make a mistake here? Or does $\mathbb{S}^2 \times \mathbb{R}$ actually have constant curvature?

(I also realized that if my computations are correct, it would imply that $\mathbb{S}^n \times \mathbb{R}$ has constant curvature for all $n \geq 1$...)


The fact that $X$ and $Y$ are orthonormal says nothing about whether or not $X_1$ and $X_2$ are orthonormal. Thus, the condition $\langle R(X_1,Y_1)Y_1, X_1\rangle =1$ does not need to hold. In fact, $X_1$ and $Y_1$ could be linearly dependent, in which case the curvature is $0$.

  • $\begingroup$ Thanks! I see. Can you tell me if everything I did before the "$= 1$" is right though? I think it isn't because it'd imply that the curvature depends only on tangent vectors of $\mathbb{S}^2$, which I'm not sure is correct. $\endgroup$ – Matheus Andrade Sep 18 '19 at 3:43
  • 1
    $\begingroup$ The rest of the argument looks fine to me. It is true in this case that the curvature of a $2$-plane only depends on its projection to $S^2$, but that occurs only because $\mathbb{R}^n$ is flat. $\endgroup$ – Jason DeVito Sep 18 '19 at 5:34

If you let $u, v$ be orthogonal unit tangent vectors to $S^2$ at a point $P\in S^2$, and $w$ be a tangent vector to $\Bbb R$ at $t \in \Bbb R$, then all three can be regarded as tangent vectors to $S^2 \times \Bbb R$ at $(P, t)$.

The sectional curvature in the plane defined by $(u,v)$ is $1$ (assuming you've got a unit $S^2$).

The sectional curvature in the plane defined by $(u, w)$ is the product of the curvature in the $u$-direction, which is $1$, with the curvature of $\Bbb R$ in the $w$ direction, which is $0$. So The sectional curvature in that plane is $0$. And $0 \ne 1$. :)

This doesn't answer your question of where you went wrong, because I'm too tired to read all you wrote. I'm sorry about this.

  • $\begingroup$ Thanks! But can you give me more details on why the curvature in those cases is equal to the product of the curvature in different directions? $\endgroup$ – Matheus Andrade Sep 18 '19 at 3:48
  • 1
    $\begingroup$ Because "sectional curvature" is designed to mean "look at the portion of my object that's locally produced by exponentiating curves in this 2-plane in the tangent space at $Q$ and then compute the ordinary curvature of that surface at the point $Q$." And ordinary curvature can be computed as a product of curvatures in the two principal-curvature directions. Of course, you have to believe that the directions I chose in fact represent principal curvature directions on $S^2$ and on $S^1 \times \Bbb R$...but that's in any ugrad diff'l geomerty text, e.g., ONeill. $\endgroup$ – John Hughes Sep 18 '19 at 11:08

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.