5

Given sectional curvature as a constant, i.e. $\dfrac{R_m(X,Y,Y,X)}{|X|^2|Y|^2-<X,Y>^2} = C$, I want to compute the curvature tensors $R(X,Y)Z$ and $R_m(X,Y,Z,W)$.

I believe I need to use the identity $$-6R_m(X,Y,Z,W) = \partial_t\partial_s\{R_m(X+sZ, Y+tW, Y+tW, X+sZ) - R_m(X+sW, Y+tZ, Y+tZ, X+sW)\}$$

I can almost see what to do for $R_m(X,Y,Z,W)$ and how this identity and the defintion of sectional curvature are related but can't really see how to finish it off. I don't really even know where to begin with $R(X,Y)Z$.

  • 1
    Do you know that if two curvature tensors have the same sectional curvature, then they are equal? Can you write down the curvature tensor of a constant curvature metric? – Gunnar Þór Magnússon Apr 07 '17 at 15:04
  • I'm not sure if I do know that, would I be able to get a reference to this? The book I've been using is Do Carmo. – user291678 Apr 07 '17 at 18:03
  • It must be somewhere in Do Carmo, it's a very basic fact. It's really just the same as a quadratic form determining a unique bilinear form by polarization. I know it's in Kobayashi-Nomizu vol 1, but probably somewhere in wikipedia as well. (Of course, you could prove it. :) – Gunnar Þór Magnússon Apr 07 '17 at 18:20
  • Btw, this is a purely linear algebra fact. You don't have to take the derivative of anything to prove it. – Gunnar Þór Magnússon Apr 07 '17 at 18:22
  • For what it's worth, you're not the first to have problems with this: http://math.stackexchange.com/questions/1932735/constant-sectional-curvature-and-the-riemann-tensor – Gunnar Þór Magnússon Apr 07 '17 at 18:24
  • Okay I can see/accept the uniqueness of the curvature tensor. I'm not really too sure how I would write down the curvature tensor of a constant curvature metric though?

    I did in fact see that thread when trying to work out how to solve this issue. But felt that the fact there was no concrete answer there and that it was 6 months old a new thread was fine!

    – user291678 Apr 08 '17 at 09:57

2 Answers2

11

To fix notation, we define the Riemann tensor as $$ \operatorname{Rm}(x,y,z,w) = \langle R(x,y)w, z \rangle, $$ and the sectional curvature as $K(x,y) = \operatorname{Rm}(x,y,x,y) / (|x|^2|y|^2 - \langle x,y \rangle^2)$.

The symmetries of the curvature tensor $\operatorname{Rm}$ mean that the bilinear form $b$ on $\bigwedge^2 V$ defined by $$ b(x \wedge y, z \wedge w) = \operatorname{Rm}(x,y,z,w) $$ is actually well defined, and we note that the sectional curvature of the metric is just the quadratic form defined by $b$ (divided by the square of the norm of $x \wedge y$). Thus the sectional curvature determines the curvature tensor, because a quadratic form determines a unique bilinear form by polarization.

We now pull out of our hat the tensor $$ \operatorname{Rm}'(x,y,z,w) = \langle x, z \rangle \langle y, w \rangle - \langle x, w \rangle \langle y, z \rangle $$ (or recall where the idea of constant sectional curvature comes from and calculate the curvature tensor of the sphere). This tensor has sectional curvature $$ K'(x,y) = \frac{|x|^2 |y|^2 - \langle x, y \rangle^2}{|x|^2|y|^2 - \langle x, y \rangle^2} = 1. $$ It thus follows that if our curvature tensor $R$ has constant sectional curvature $C$, then $R = C R'$.

We can now stare at $\operatorname{Rm}'$ for a couple of minutes and see that we must have $R'(x,y)z = \langle y, z \rangle \, x - \langle x, z \rangle \, y$, because that is a tensor such that $$ \langle R'(x,y)w, z \rangle = \operatorname{Rm}'(x,y,z,w) $$ for all $w$.

1

For $Rm(X, Y, Z, W)$, you can use equations (5) and (6) in Karcher's 1970 paper "A short proof of Berger's curvature tensor estimates". These formulas give the result for $X, Y, Z, W$ orthonormal. Then, some work is needed to generalize to arbitrary vectors, along the same lines as in the proof of the second theorem in that (3 page) paper. Theorems in this paper provide bounds on $Rm$ in terms of sectional curvature in the general case, which may be useful to go beyond constant sectional curvature.