I was wondering if anyone had a simple example of a manifold which only has Ricci curvature in one direction ie. such that the Ricci tensor only has one non-zero component.

Intuitively, I would expect such a thing to be possible somehow just from the definition of Ric but could not think of an example.


Yes, this is possible. More precisely: In dimension $\geq 3$ there are metrics for which the Ricci curvature has rank $1$, that is, which can be written (locally, around any point) as $\phi \alpha \otimes \alpha$ for some smooth function $\phi$ and $1$-form $\alpha$. (In dimension $2$ the Ricci tensor is a (smooth) multiple of the metric, so this behavior is possible only in dimension $\geq 3$.)

Here's one way to construct an example. The warped product of the flat metric $\bar g := dx_1^2 + \cdots + dx_{n - 1}^2$ on $\Bbb R^{n - 1}$ with the flat metric $dy^2$ on $\Bbb R$ via $f(y)$ is $$g := \bar g \times_{f(y)} dy^2 = f(y) \bar g + dy^2 .$$

Computing gives that the Ricci curvature of $g$ is $$\operatorname{Ric} = - \frac{2 f''(y) f(y) + (n - 1) f'(y)^2}{4 f(y)} \bar g - \frac{(n - 1)(2 f''(y) f(y) - f'(y)^2)}{4 f(y)^2} dy^2$$

Demanding that the coefficient of $\bar g$ vanish defines a second-order differential equation in $f$ with general solution $$f(y) = \left[\frac{n - 1}{2} (A y + B)\right]^{2 / (n - 1)},$$ and substituting in $g$ gives (now restricting the metric to where $A y + B > 0$) that $$\operatorname{Ric} = \frac{(n - 2) A^2}{(n - 1) (A y + B)} dy^2 .$$ So, if we take $A \neq 0$, then for $n > 2$, $\operatorname{Ric}$ has rank $1$.

  • $\begingroup$ I have two questions: would the final expression be neater if you assumed $f^2$ in the warped product? Also, I don't quite follow why you get one differential equation, if you assume the coefficient of g bar is 0 and the coefficient of $dy^2$ is 0, does that not give you a system of two differential equations to solve? $\endgroup$ – Tom Mar 20 at 20:12
  • $\begingroup$ ie. to solve for $f$, would you not need to assume the coefficients of both metrics are $0$ and that would then give you a pair of coupled differential equations? $\endgroup$ – Tom Mar 20 at 20:47
  • 1
    $\begingroup$ If you replaced $f$ with $f^2$ then the solution would be replaced with its square root. This would change the exponent in the formula for $f$ in the answer, but I don't think that's much of an improvement. And no---if we ask for both coefficients to vanish, that's asking $\operatorname{Ric}$ to be zero (and for $n > 2$ this imposes exactly $A =0$), but here we're looking for a metric for which $\operatorname{Ric}$ has rank $1$. So, by demanding that the coefficient of $\bar g$ in $\operatorname{Ric}$ vanish, we're looking for a metric for which $\operatorname{Ric}=\phi\,dy^2$ for some $\phi$. $\endgroup$ – Travis Willse Mar 20 at 20:54
  • $\begingroup$ I'm not sure if I have explained myself properly. Basically, take a manifold $(M,g)$ with a scalar curvature (call it $R$), I am looking for an example of a manifold such that one of the components of Ric is equal to the value of $R$ at every point, or maybe on most points if that is not possible, and Ric is $0$ for all the other components. In a sense this must mean you are solving two equations since you want the g bar coefficient to vanish but you want the coefficient for $dy^2$ to equal $R$. $\endgroup$ – Tom Mar 20 at 22:00
  • $\begingroup$ Strictly speaking one can only talk about "components of $\operatorname{Ric}$" once you choose additional information like a local frame. (This is why I refer to "rank $1$", which is a coordinate-independent condition, in my answer.) In any case, is your problem that you have a specific scalar function $R$ and you want to find a metric $g$ and coordinates $(x_1, \ldots, x_{n - 1}, y)$ such that $\operatorname{Ric} g = R \,dy^2$? $\endgroup$ – Travis Willse Mar 20 at 22:13

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.