3

Let $\nabla$ be a Riemannian connection, $\Delta$ be the Laplacian defined as $\text{tr}_g \nabla^2$. The equation I want to show is $$ \Delta\nabla_i f = \nabla_i\Delta f + \sum_{j} Ric_{ij} \nabla_jf $$ where $Ric$ is the Ricci curvature tensor, $f$ is a smooth function. The proof my text gives is (they use the convention that basically $g_{ij} = \delta_{ij}$ which I am not using)

$$ \Delta \nabla_i f = \nabla_j\nabla_i\nabla_j f = \nabla_i\nabla_j\nabla_j f - R_{jijk} \nabla_k f $$

...which I understand 0% about it. My guess is that the first interchange $$ \begin{align*} \Delta\nabla_i f &= \sum_{j,k}g^{jk}\nabla_j\nabla_k\nabla_i f = \sum_{j,k}g^{jk}\nabla_j\nabla_i\nabla_k f \end{align*} $$ is free since $\nabla_i\nabla_j f = \nabla_j \nabla_if$ and $\nabla$ has no torsion. But what about the second one (i.e. how do we interchange $\nabla_j$ and $\nabla_i$?)

Edit: so I've found and use the Ricci identity for 1-forms

$$ (\nabla_i \nabla_j - \nabla_j \nabla_i) w(\partial_k) = w(R(\partial_j,\partial_i)\partial_k) = \sum_{l} w^l R_{jik}^l $$ put $w = \nabla f$ says $$ (\nabla_i \nabla_j - \nabla_j \nabla_i) \nabla f(\partial_k) = (\nabla_i \nabla_j - \nabla_j \nabla_i) \nabla_k f = \sum_{l} (\nabla f)^l R_{jik}^l $$ and I don't know how to proceed to get $Ric$.

Travis Willse
  • 99,363
macton
  • 659
  • The second equality follows from the definition of curvature $R(X, Y) Z = \nabla_X \nabla_Y Z - \nabla_Y \nabla_X Z - \nabla_{[X, Y]} Z$ for vector fields $X, Y, Z$. – Travis Willse Apr 08 '22 at 18:28
  • @TravisWillse I'm not sure why that directly follows. I can take $X = \partial_i, Y = \partial_j$, but what should $Z$ be? – macton Apr 09 '22 at 09:07
  • You can take $Z = (\nabla f)^\sharp$ and use the symmetries of the Riemannian curvature tensor. In any case, from your last line you can use the metric to raise and lower indices to raise an unpaired index, then use the symmetries of the Riemann curvature tensor and form the appropriate trace to compute an expression in terms of the Ricci tensor. – Travis Willse Apr 10 '22 at 01:45
  • @TravisWillse My idea is that $$\sum_{j,k,l}g^{jk}\nabla_lf R_{jik}^l = \sum_{j,k,l,m}g^{jk}g^{lm}\nabla_lf R_{jikm} = -\sum_{j,k,l,m}g^{jk}g^{lm}\nabla_lf R_{jimk} = -\sum_{l,m}g^{lm}\nabla_lf Ric_{im} \overset{???}{=} -\sum_{l}\nabla_lf Ric_{il}$$ but I don't think the last equality holds in any possible cases. Are there something that I am missing? – macton Apr 10 '22 at 12:44
  • https://math.stackexchange.com/q/4399574/272127 – C.F.G Apr 14 '22 at 01:44

1 Answers1

3

You're almost there. Using the Einstein summation convention, so that we sum over repeated indices, we have

$$\Delta \nabla_i f = \nabla^j \nabla_j \nabla_i f = \nabla^j \nabla_i \nabla_j f = - R^j{}_i{}^k{}_j \nabla_k f +\nabla_i \nabla^j \nabla_j f = g^{jk} \operatorname{Ric}_{ij} \nabla_k f + \nabla_i \Delta f.$$

The third equality follows from applying the Ricci identity for $1$-forms that you mention in your edit to $\nabla_j f$, and the fourth from the symmetries of the curvature tensor $R$ and the definitions of $\operatorname{Ric}$ and $\Delta$.

Travis Willse
  • 99,363