4
$\begingroup$

A Riemannian manifold $(M, g)$ is said to be an almost Ricci soliton if there exists a complete vector field $X \in \Gamma(TM)$ and a smooth function $\lambda: M \to \mathbb{R}$ such that $$\operatorname{Ric} + \frac{1}{2}\mathscr{L}_{X} g = \lambda g$$ When this vector field is the gradient of a smooth function $f: M \to \mathbb{R}$, we say $M$ is a gradient almost Ricci soliton, and this equation becomes: $$\operatorname{Ric} + \operatorname{Hess}(f) = \lambda g$$ Obviously, any Einstein manifold is a Ricci soliton and hence an almost Ricci soliton (gradient as well, trivially), so these are trivial examples.
If $M$ satisfies: $$\operatorname{div}({\operatorname{Rm}}) = 0$$ we then say $M$ has harmonic curvature (notice this happens if and only if $M$ has harmonic Weyl curvature and constant scalar curvature. I think that part of some work I've been doing with some other people shows that any gradient almost Ricci soliton with harmonic curvature satisfies the property that for any $p \in M$, there is a neighborhood $U_p \ni p$ such that $U_p$ has constant sectional curvature (and is therefore necessarily Einstein) (EDIT AT NOVEMBER 27: this supposes the dimension is $\geq 4$. Also, I've come to realize since the initial writing of this post that the Einstein examples might not be exhaustive).

As a sanity check, I'm looking for some explicit examples of nontrivial (i.e, not Einstein and with nonconstant $\lambda$) gradient almost Ricci solitons (preferably of dimension $\geq 5$) with harmonic curvature. Can anyone here provide some examples? I'd appreciate any help. Thanks in advance!

$\endgroup$
5
  • $\begingroup$ I'm curious why you define the condition of 'harmonic curvature' in one sentence and then don't refer to again. Is this a stray sentence that should be deleted or did you intend to add this as a hypothesis somewhere and forgot? $\endgroup$ Commented Nov 2, 2022 at 14:33
  • $\begingroup$ @RobertBryant sorry, I indeed forgot to add it as a hypothesis. I'll fix it, thanks for pointing it out. $\endgroup$ Commented Nov 2, 2022 at 16:41
  • $\begingroup$ I thought a little bit about this and did a few caculations. In dimension 3, at least, it is not true that a gradient almost Ricci soliton with harmonic curvature is locally Einstein. There is a $3$-parameter family of mutually non-isometric, nontrivial examples $(g,f,\lambda)$ in dimension $3$ with the function $\lambda$ and the sectional curvatures not being constant. I haven't looked at higher dimensions, but I don't see why it would necessarily fail there. $\endgroup$ Commented Nov 27, 2022 at 14:06
  • $\begingroup$ @RobertBryant thanks a lot! However, I must apologize again for not having included the hypothesis of dimension $\geq 4$ in my original post when I mentioned "part of some work that I've been doing with some other people...". But I am indeed very interested in knowing the explicit example of this $3$-parameter family you mentioned, I would very much appreciate it if you shared it. $\endgroup$ Commented Nov 27, 2022 at 18:43
  • $\begingroup$ OK. As I suspected, the three-dimensional examples easily generalize to all higher dimensions as a 3-parameter family of non-trivial gradient almost Ricci solitons with harmonic curvature. They are not completely explicit, though, because each equivalence class of solutions corresponds to an integral curve of a vector field in $\mathbb{R}^3$. That vector field depends on a parameter: the (constant) scalar curvature $S=n(n{-}1)c$ of the metric $g$. I don't know how to integrate the vector field in elementary terms, but in the case $c=0$ phase portrait methods give good qualitative information. $\endgroup$ Commented Nov 27, 2022 at 21:41

2 Answers 2

4
$\begingroup$

I'm revising my answer to shorten it, since there is a much simpler way to describe these solutions more fully.

Let $(N^n,h)$ be a metric of constant sectional curvature $k$ and consider the quadratic form $$ g = \frac{\mathrm{d}u^2}{k-a\,u^2+ b\,u^{1-n}} + u^2\,h $$ on $M^{n+1} = \mathbb{R}^+\times N$, where $a$ and $b$ are constants and $u>0$ is the coordinate on $\mathbb{R}^+$. If $I\subset\mathbb{R}^+$ is an interval on the $u$-line on which $k-a\,u^2+ b\,u^{1-n} >0$, then $g$ is a Riemannian metric on $I\times N$ that is conformally flat and has constant scalar curvature $S = n(n{+}1)a$. Hence it has harmonic curvature. The Ricci curvature is $$ \mathrm{Ric}(g) = \bigl(n\,a - \tfrac{1}{2}\,b\,u^{-n-1}\bigr)\,g + \frac{(n^2{-}1)b\,\mathrm{d}u^2}{2\bigl(b\,u^2+k\,u^{n+1}-a\,u^{n+3}\bigr)}, $$ so $g$ is Einstein if and only if $b=0$.

Moreover, it is now easy to construct (by quadrature) a function $f = f(u)$ on $I$ such that $\mathrm{Ric}(g) + \mathrm{Hess}_g(f) = \lambda\,g$ for some function $\lambda$. When $b\not=0$, $\lambda$ will not be constant. Thus, this gives a completely explicit $3$-parameter family of non-trivial almost Ricci solitons with harmonic curvature.

If $I = (r_1,r_2)$ where $r_2>r_1>0$ are simple roots of $k-a\,u^2+ b\,u^{1-n}=0$, then the curve $v^2 = k-a\,u^2+ b\,u^{1-n}$ in the $uv$-plane has a smooth circle component $C$ between the lines $u=r_1$ and $u=r_2$. In this case, the metric $g$ extends to a smooth complete metric on $C\times N$ (assuming that $(N,h)$ is complete). In this way, one can construct many complete or compact examples of such metrics. However, when $b\not=0$, the functions $f$ and $\lambda$ will only be locally defined unless one passes to the simply-connected cover of $C$, so that $M = \mathbb{R}\times N$. On this covering space, $f$ (and $\lambda$) can be globally defined.

Remark: In dimension $3$, it turns out that every conformally flat metric $(M^3,g)$ with constant scalar curvature (i.e., every metric in dimension $3$ with harmonic curvature) that admits a 'Ricci potential', i.e., a function $f$ such that $\mathrm{Ric}(g)+\mathrm{Hess}_g(f) = \lambda\,g$ for some function $\lambda$, is locally of the above form for some $(N^2,h)$ with constant curvature. It was after I worked out this fact, via an exterior differential system analysis, that I realized that the above construction could be used to produce examples in any dimension.

In dimensions above $3$, it is not likely that every metric with harmonic curvature that admits a Ricci potential in the above sense is of the form given above, but I don't know a classification, even when $n=4$.

$\endgroup$
6
  • $\begingroup$ Thanks very much! This confirmed a few suspicions of mine. When you have the time I'd love to see all the other examples you mentioned, but this is already pretty good :) $\endgroup$ Commented Nov 28, 2022 at 15:50
  • $\begingroup$ I just saw your addendum. What a wonderful answer! Thanks a lot. This completely explicit construction goes above and beyond what I expected. $\endgroup$ Commented Dec 1, 2022 at 19:36
  • $\begingroup$ Could you perhaps have made a typo when defining $g$? I calculated the scalar curvature of a metric $$g = \frac{\mathrm{d}u^2}{(f(u))^2} + u^2 h$$ and found it to be equal to $$\text{Scal}_{g} = \frac{n f'}{fu} - \frac{n f f'}{u} + \frac{n(n-1)}{u^2}\left(k - \frac{1}{f^2} \right) $$ Setting $f = \sqrt{k - au^2 + bu^{1-n}}$ as in your example, I found that $\text{Scal}_g$ is not constant. $\endgroup$ Commented Jan 12, 2023 at 13:35
  • $\begingroup$ @MatheusAndrade: Hmmm. You are right that something is wrong with my answer, but I'm not sure what. I'm traveling now and don't have those calculations with me so I can't check them. I'll have a look at them when I get back home next week, check them over, and see what I can do about correcting the formulae. $\endgroup$ Commented Jan 13, 2023 at 7:29
  • $\begingroup$ Alright, I'm looking forward to it, thanks for the reply! I think maybe $\text{Ric}(g)$ or the definition of $g$ is not right. $\endgroup$ Commented Jan 13, 2023 at 17:52
4
$\begingroup$

The Riemannian product $\mathbb{R}^m \times S^n$ is always of this type, where $\mathbb{R}^m$ is given the flat metric and $S^n$ the round metric of constant sectional curvature one. In this case $\lambda=n-1$ and you can take $f(x,\theta) = \frac{n-1}{2}\lvert x\rvert^2$.

$\endgroup$
1
  • $\begingroup$ Thanks very much for the answer, but I would prefer examples where $\lambda$ is not constant. I'm sorry I didn't specify that, I'll edit my question. $\endgroup$ Commented Nov 2, 2022 at 0:05

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .