I am doing this problem:

Let $g$ be a Riemannian metric on manifold $M$ and $\tilde{g}=e^{2f}g$ another metric conformal to $g$, where $f$ is a smooth function on $M$. Give the relation between the Levi-Civita connection $\nabla$ of $g$ and the Levi-Civita connection $\tilde{\nabla}$ of $\tilde{g}$.

I am currently at this step, but don't know how to continue next:


Any help is appreciated!

  • $\begingroup$ I posted this calculation a while ago here $\endgroup$ – Yuri Vyatkin Sep 14 '18 at 13:53

A connection of the conformal metric is determined, as is any connection, by its Christoffel symbols. We have the following formula for Levi-Civita connections:

$$\Gamma_{ij}^k = \frac{1}{2}g^{kl}(\partial_i g_{jl} + \partial_j g_{il} - \partial_l g_{ij})$$

Another pretty obvious fact is that $\hat g^{ij} = e^{-2f}g^{ij}$. So for the conformal metric $\hat g$ we can compute the Christoffel symbols: $$ \hat \Gamma_{ij}^k = \frac{1}{2}e^{-2f}g^{kl}(\partial_i (e^{2f}g_{jl}) + \partial_j (e^{2f}g_{il}) - \partial_l (e^{2f}g_{ij})) $$ Now it's just a matter of applying the product rule and using $e^{-2f}e^{2f} = 1$, $g_{ab}g^{bc} = \delta_a^c$, and $g^{ab}\partial_b f = \nabla^a f$ to get:

$$ \hat \Gamma_{ij}^k = \Gamma_{ij}^k + (\partial_i f) \delta_j^k + (\partial_j f)\delta_i^k - (\nabla^k f)g_{ij} $$

If we apply $\hat \nabla$ to any tensor $H^{b_1 ...b_k}_{c_1 ... c_l}$ we have:

$$ \hat \nabla_a H^{b_1 ...b_k}_{c_1 ... c_l} = \partial_a H^{b_1 ...b_k}_{c_1 ... c_l} + \sum_{i=1}^{k} \hat \Gamma_{a p}^{b_i} H^{b_1 ... p ...b_k}_{c_1 ... c_l} - \sum_{j=1}^{l} \hat \Gamma_{a c_j}^{q} H^{b_1 ...b_k}_{c_1 ... q ... c_l} $$

Where $p$ and $q$ are in the i-th and j-th positions, respectively. Hopefully that helps.

  • $\begingroup$ Welcome to MathSX. Thank you for this answer. It may help some users. $\endgroup$ – Yuri Vyatkin Mar 15 '19 at 9:33

I'll register a shorter coordinate-free proof. Let the Koszul formula be written as $$2g(\nabla_XY,Z) = A(X,Y,Z) + B(X,Y,Z),$$where $A$ is the part of the Koszul formula containing directional derivatives and $B$ is the part containing Lie brackets. We know that any two connections differ by a tensor, so write $\widetilde{\nabla}_XY = \nabla_XY + T_XY$ -- the goal is to find $T$, and we know that $$2\widetilde{g}(\widetilde{\nabla}_XY,Z) = \widetilde{A}(X,Y,Z) + \widetilde{B}(X,Y,Z).$$Clearly $\widetilde{B}(X,Y,Z) = {\rm e}^{2f}B(X,Y,Z)$, while $$ X\widetilde{g}(Y,Z) = X({\rm e}^{2f})g(Y,Z) + {\rm e}^{2f}X(g(Y,Z))$$says that $\widetilde{A}(X,Y,Z) = X({\rm e}^{2f})g(Y,Z) + Y({\rm e}^{2f})g(X,Z) - Z({\rm e}^{2f})g(X,Y) + {\rm e}^{2f}A(X,Y,Z)$. Thus $$2\widetilde{g}(\widetilde{\nabla}_XY,Z) = X({\rm e}^{2f})g(Y,Z) + Y({\rm e}^{2f})g(X,Z) - Z({\rm e}^{2f})g(X,Y) + 2{\rm e}^{2f}g(\nabla_XY,Z).$$Evaluating $X({\rm e}^{2f}) = 2{\rm e}^{2f}\,X(f)$, etc., and simplyfying $2{\rm e}^{2f}$ on everything, we get $$g(\nabla_XY + T_XY,Z) = X(f)g(Y,Z) + Y(f)g(X,Z) - Z(f)g(X,Y) + g(\nabla_XY,Z).$$Eliminate $\nabla_XY$ from the above and use the definition of $g$-gradient to write the right side in the form $g({\rm something}, Z)$, obtaining $$g(T_XY,Z) = g(X(f)Y + Y(f)X - g(X,Y){\rm grad}(f), Z).$$This means that $$T_XY = X(f)Y + Y(f)X - g(X,Y){\rm grad}(f)$$and hence $$\widetilde{\nabla}_XY = \nabla_XY +X(f)Y + Y(f)X - g(X,Y){\rm grad}(f).$$


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.