2
$\begingroup$

I am following some lecture notes on Ricci flow and reached the section where we linearize the Ricci tensor and obtain the principal symbol for the resulting operator. We have $T \in \: \Gamma(Sym^2 T^{*}M)$ smooth, fixed and positive definite and then compute the time derivative for the divergence of $G(T)$:

$\bigg(\frac{\partial}{\partial t}\delta G(T) \bigg)Z = -T \bigg((\delta G(h))^{\#},Z\bigg) + \bigg<h,\nabla T(.,.,Z) - \frac{1}{2}\nabla_{Z}T \bigg>, $

where $\frac{\partial g}{\partial t}=h$, $Z$ an arbitrary vector field and $G(T)=T-\frac{1}{2}(tr \: T)h$.

The notes then say that this implies

$\frac{\partial}{\partial t}T^{-1}\delta G(T) = -\delta G(h)\: + \:\: ...$

where the dots indicate terms which don't have derivatives of $h$. I see that you can apply $T^{-1}$ as it doesn't depend on $t$ and then lose the $Z$ as it is arbitrary, but not sure exactly what happens to the right-hand side such that we end up with $-\delta G(h)$ or where the sharp goes.

$\endgroup$
5
  • $\begingroup$ I gotta admit that, despite my preference for nice abstract notation, I only like what I invent myself. So when faced with deciphering someone else's abstract notation like this, I start by converting it all to local coordinates and write the tensors with indices. Here, it's particularly confusing, at least for me, with the dependence on $t$. $\endgroup$
    – Deane Yang
    Feb 15, 2019 at 16:23
  • $\begingroup$ Yes, this is why I am confused, I have studied other treatments of the Ricci-DeTurck flow in Chow and others and it is completely explicit and done in indices, so I find this treatment difficult to decipher. $\endgroup$ Feb 15, 2019 at 18:05
  • 1
    $\begingroup$ I don't see why the sharp has been dropped, should I just convert to indices, do the calculation and then convert back to the notation of the lecture notes at the end? Not sure what else to do. $\endgroup$ Feb 15, 2019 at 18:18
  • 1
    $\begingroup$ I think there is a typo in the definition of $G(T)$. It should be $G(T) = T -\frac 12 tr(T)\cdot \mathbf g$. $\endgroup$ Feb 15, 2019 at 21:01
  • $\begingroup$ I think in Peter Topping, Lecture notes on Ricci flow, he refers to do computations in a coordinate-free way. Some people refer to this way. This avoids local coordinate sickness. Of course, one of the drawbacks is "a heavier notation" $\endgroup$
    – Truong
    Feb 16, 2019 at 3:05

1 Answer 1

3
$\begingroup$

It looks like this is from the Lecture notes on Ricci flow from Peter Topping. He mentions that he uses the symbol $T$ as the symmetric, positive definite bilinear form and also for the map $\Gamma(TM^\ast) \to \Gamma(TM^\ast)$ induced by $T$ and the metric $g$ in the following way $$ T(\alpha)(Z)=T(\alpha^\#,Z) $$ where $\alpha^\#$ is the dual vector field to the $1$-form $\alpha$ with respect to $g$. Thus looking at \begin{eqnarray*} \bigg(\frac{\partial}{\partial t}\delta G(T) \bigg)Z &=& -T \bigg((\delta G(h))^{\#},Z\bigg) + \bigg<h,\nabla T(.,.,Z) - \frac{1}{2}\nabla_{Z}T \bigg> \\ &=& -T(\delta G(h))(Z) + \ldots \end{eqnarray*} we just apply to the equation the inverse of $T$ to obtain your desired expression (note that $\delta G(T)$ is a $1$-form)

$\endgroup$
4
  • $\begingroup$ Hi this is correct, I was working from Topping's lecture notes on Ricci flow. Also, once we have the desired expression he says that this implies that $\frac{\partial}{\partial t} L_{(T^{-1} \delta G(T))^{\#}}g = -L_{(\delta G(T))^{\#}} + ... $ where $L$ is the Lie derivative. So, it has to be a sharp as it needs to be a vector field in the direction of which the Lie derivative acts, but why is it $\delta G(T)$ and not $\delta G(h)$? $\endgroup$ Feb 16, 2019 at 15:26
  • $\begingroup$ Also, in this expression with the Lie derivatives the term on the right-hand side involves a second derivative of $h$ and the dots indicate terms which contain $h$ and its first derivative. $\endgroup$ Feb 16, 2019 at 15:35
  • $\begingroup$ I've thought about it and I think that the fact that Topping has it with a $T$ rather than a $h$ must be a typo, as it changes back to $h$ when he linearizes the operator, then the implication comes from definition of the Lie derivative. $\endgroup$ Feb 17, 2019 at 16:41
  • $\begingroup$ That would have been my guess too. $\endgroup$ Feb 17, 2019 at 18:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.