Let $g$ be a Riemannian metric on $M$ and let $\tilde{g}=f^{2}g$ where $f$ is a smooth function that is never zero. let $\nabla$ and $\nabla'$ be the Riemannain connections of $g$ and $\tilde{g}$ on $M$, give the relation between $\nabla$ and $\nabla'$.

So I'm not really sure what to do here but I tryed to express each connection in terms of its Christoffel symbols and locall expression for a connection

What I get when I do that is the following let $\Gamma^{k}_{i,j}$ and $'\Gamma^{k}_{i,j}$ denote the Christoffel of each connections then I get that $$'\Gamma^{k}_{i,j}=\Gamma^{k}_{i,j}+\dfrac{1}{f}g^{k,l}(g_{i,j}\partial_{i}f+g_{i,l}\partial_{j}f-g_{i,j}\partial_{l}f) $$

Now when I plug this into the local expression for the connection of $\nabla'$ I get

$$ \nabla'_{X}Y=\nabla_{X}Y + big ~mess$$

Is there a better way of doing this?

  • 1
    $\begingroup$ You can get the Levi-Civita connection from the metric using the Koszul formula, and I actually think that would be a bit cleaner. Have you tried using this? $\endgroup$ – Elchanan Solomon Feb 3 '14 at 2:03
  • $\begingroup$ No I haven't I'm not that familiar with the Koszul formula but thanks for the tip I will try this out. $\endgroup$ – lance wellton Feb 3 '14 at 2:09
  • 2
    $\begingroup$ I assume you're not used to the differential forms/moving frames approach. BTW, such conformal changes of metric are usually written $\tilde g = e^{2\phi}g$ to avoid the logarithmic derivatives that show up this way. $\endgroup$ – Ted Shifrin Feb 3 '14 at 2:46
  • $\begingroup$ I do know what differential forms and local frame are but I don't know what moving frames are, sorry. But what is this moving frames approach? $\endgroup$ – lance wellton Feb 3 '14 at 3:43

The formula for the conformal rescaling of the Levi-Civita connection is an essential tool in Riemannian geometry, and its derivation is given in many sources.

As Isaac Solomon and Ted Shifrin have mentioned in the comments, a slick way to derive it is to consider $f = e^{2 \omega}$ and use the Koszul formula. The result will be in the form: $$ \nabla' _X Y = \nabla _X Y + (X \omega )Y + (Y \omega )X - g(X,Y) \operatorname{grad}\omega \tag{1} $$

Proof. The Koszul formula (see e.g. here) gives the following expression for the Levi-Civita connection $\nabla$ of the metric $g$: $$ \begin{align} 2 g(\nabla_X Y, Z) & = X \, g(Y,Z) + Y \, g(Z,X) - Z \, g(X,Y) \\ \tag{2} &- g(X,[Y,Z]) + g(Y,[Z,X]) + g(Z,[X,Y]) \end{align} $$

Let $\nabla'$ be the Levi-Civita connection for the metric $g' = e^{2\omega}g$. Substituting these objects into (2) $$ \begin{align} 2 e^{2 \omega} g(\nabla'_X Y, Z) & = X \left( e^{2 \omega} g(Y,Z) \right) + Y \left( e^{2 \omega} g(Z,X) \right) - Z \left( e^{2 \omega} g(X,Y) \right) \\ &- e^{2 \omega} g(X,[Y,Z]) + e^{2 \omega} g(Y,[Z,X]) + e^{2 \omega} g(Z,[X,Y]) \end{align} $$ and computing the derivatives using the product rule, we obtain $$ \begin{align} 2 e^{2 \omega} g(\nabla'_X Y, Z) & = e^{2 \omega} X g(Y,Z) + e^{2 \omega} Y g(Z,X) - e^{2 \omega} Z g(X,Y) \\ & + 2 e^{2 \omega} g(Y,Z) \, X \omega + 2 e^{2 \omega} g(Z,X) \, Y \omega - 2 e^{2 \omega} g(X,Y) \, Z \omega \\ &- e^{2 \omega} g(X,[Y,Z]) + e^{2 \omega} g(Y,[Z,X]) + e^{2 \omega} g(Z,[X,Y]) \end{align} $$

In the last display we can divide both sides of the equation by $e^{2 \omega}$, which is a strictly positive function, to get $$ \begin{align} 2 g(\nabla'_X Y, Z) & = X g(Y,Z) + Y g(Z,X) -Z g(X,Y) \\ & + 2 g(Y,Z) \, X \omega + 2 g(Z,X) \, Y \omega - 2 g(X,Y) \, Z \omega \\ &- g(X,[Y,Z]) + g(Y,[Z,X]) + g(Z,[X,Y]) \end{align} $$

Using the Koszul formula (2) again we rewrite the above expression as $$ \begin{align} 2 g(\nabla'_X Y, Z) & = 2 g(\nabla_X Y, Z) + 2 g(Y,Z) \, X \omega + 2 g(Z,X) \, Y \omega - 2 g(X,Y) \, Z \omega \end{align} $$ which is equivalent to (1) because vector field $Z$ is arbitrary, $g$ is non-degenerate, and $Z \omega = \mathrm{d} \omega (Z)$. Recall also that $\operatorname{grad} \omega = (\mathrm{d} \omega)^{\sharp}$.

The version of this formula in terms of coordinates and the Christoffel symbols is obtained by a similar calculation, the result will be $$ '\Gamma^{k}_{ij}=\Gamma^{k}_{ij} + \delta_{i}^{k} \partial_j \omega + \delta_{j}^{k} \partial_i \omega - g_{i j} g^{k l} \partial_{l} \omega $$

This can be also obtained as a consequence of (1).

  • $\begingroup$ Thanks for this it was very helpful. $\endgroup$ – lance wellton Feb 4 '14 at 21:16
  • $\begingroup$ Those who prefer Christoffel symbols may take a look at this answer. $\endgroup$ – Yuri Vyatkin Mar 15 at 9:36

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.