4

I have to prove the following:

Let $\Omega \subseteq \mathbb{R}^d$ be open and $g$ a metric field on $\Omega$. For every $\phi \in \mathrm{Diff}(\Omega)$ let $\Xi^i_{jk}[\phi]$ be functions on $\Omega$ that transform in the same way as the Christoffel symbols $\Gamma^i_{jk}[\phi]$ $$ \Xi^i_{jk}[\phi](y) = \frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) \Xi^r_{pq}[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y)) + \frac{\partial y^i}{\partial x^m} (\phi^{-1}(y)) \frac{\partial^2 x^m}{\partial y^j \partial y^k}(y). $$ Assume that for each $y_0 \in \Omega$ there is a $\phi_0 \in \mathrm{Diff}(\Omega)$ such that $\Xi^i_{jk}[\phi_0](y_0) = 0$ and $(\partial_a g_{bc})[\phi_0](y_0)=0$. Show $\Xi^i_{jk}[\phi] =\Gamma^i_{jk}[\phi]$ for all $\phi \in \mathrm{Diff}(\Omega)$.

First thank you for all your answers.

I know now, that $T^{i}_{\ jk}$ is a Tensor field and that $T^{i}_{\ jk}[\phi_0](y_0) = 0$. But I'm struggling with the last step. In the second answer,there is a $x_0$ defind by $x_0 = \phi^{-1}(y_0)$. I don't think that one can reason that $\forall x_0 \in \Omega \ \exists \phi_0: T^{i}_{\ jk}[\phi_0](\phi_0(x_0))$, because not every point $x$ in $\Omega$ has to fulfill $x = \phi_0^{-1}(y_0)$. Every $y_0$ can have a different diffeomorphism $\phi_0$, with which it fulfills $\Xi^i_{jk}[\phi_0](y_0) = 0$. So $x_0 = \phi_0(y_0)$ is not bijective.

I've already done different attempts, but i don't get rid of the dependency of $y_0$ and $\phi_0$.

michael
  • 201

3 Answers3

3

I tried to solve the exercise by following the hint, could someone please tell me if it is correct? Thank you

We define:

$T^{i}_{\,jk}[\phi](y) := \Xi^i_{\,jk}[\phi](y) - \Gamma^i_{\,jk}[\phi](y)$

Thus we see that this transforms as a Tensor:

$T^{i}_{\,jk}[\phi](y)= \Xi^i_{jk}[\phi](y) - \Gamma^i_{\,jk}[\phi](y) = \bigg(\frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) \Xi^r_{pq}[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y)) + \frac{\partial y^i}{\partial x^m} (\phi^{-1}(y)) \frac{\partial^2 x^m}{\partial y^j \partial y^k}(y) \bigg)- \bigg(\frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) \Gamma^r_{pq}[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y)) + \frac{\partial y^i}{\partial x^m} (\phi^{-1}(y)) \frac{\partial^2 x^m}{\partial y^j \partial y^k}(y)\bigg) = \frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) (\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y))$

AFTER THIS POINT THE PROOF IS WRONG, SEE ANSWER ON BOTTOM FOR THE CONTINUATION

I still leave it here so that you can see what my original idea was (before they changed the exercise guidelines), not more:


Now we have that for our given $y_0$ and $\phi_0$:

$\Gamma^i_{\,jk}[\phi_0](y_0)=\frac{1}{2} g^{il}(\partial_k g_{lk} + \partial_j g_{lk} - \partial_l g_{jk})[\phi_0](y_0)=\frac{1}{2} g^{il}(\partial_k g_{lk}[\phi_0](y_0) + \partial_j g_{lk}[\phi_0](y_0) - \partial_l g_{jk}[\phi_0](y_0))=0$

So we get for our given $y_0$ and $\phi_0$ :

$T^{i}_{\,jk}[\phi_0](y_0) := \Xi^i_{\,jk}[\phi_0](y_0) - \Gamma^i_{\,jk}[\phi_0](y_0)=0$

Since $T$ is a tensor this Implies:

$(\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}](\phi_0^{-1}(y_0))\overset{!}{=}0$

Now we notice that since $\phi \in \mathrm{Diff}(\Omega)$ we have that $y_0\in\Omega$ this implies that for each $y_0$ we can find an $x_0\in\Omega$ s.t $y_0=\phi_0(x_0)$ so the constraint is equal to:

$\forall x_0\in\Omega$ $\exists$ $\phi_0$ s.t $T^{i}_{\,jk}[\phi_0](\phi_0(x_0))=0$

We can now take any $\phi \in \mathrm{Diff}(\Omega)$ and have $\phi=\phi\circ\phi_0^{-1}\circ\phi_0:=\tilde{\phi}\circ\phi_0$

So eventually we get: $\forall x_0\in\Omega$

$T^{i}_{\,jk}[\phi](\phi(x_0))=T^{i}_{\,jk}[\phi](\phi\circ\phi_0^{-1}(y_0))=T^{i}_{\,jk}[\tilde{\phi}\circ\phi_0](\tilde{\phi}(y_0))= \frac{\partial x^p}{\partial y^j}(\tilde{\phi}(y_0)) \frac{\partial x^q}{\partial y^k}(\tilde{\phi}(y_0) ) (\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}]((\tilde{\phi}\circ\phi_0)^{-1}(\tilde{\phi}(y_0)))\frac{\partial y^i}{\partial x^r}((\tilde{\phi}\circ\phi_0)^{-1}(\tilde{\phi}(y_0)))=\frac{\partial x^p}{\partial y^j}(\tilde{\phi}(y_0)) \frac{\partial x^q}{\partial y^k}(\tilde{\phi}(y_0) ) (\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}]((\phi_0^{-1})(\tilde{\phi}^{-1}\circ\tilde{\phi}(y_0)))\frac{\partial y^i}{\partial x^r}((\phi_0^{-1})(\tilde{\phi}^{-1}\circ\tilde{\phi}(y_0)))= \frac{\partial x^p}{\partial y^j}(\tilde{\phi}(y_0)) \frac{\partial x^q}{\partial y^k}(\tilde{\phi}(y_0) ) (\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}](\phi_0^{-1}(y_0))\frac{\partial y^i}{\partial x^r}(\phi_0^{-1}(y_0))=0$

$\forall \phi\in Diff(\Omega)$

This implies:

$\forall \phi\in Diff(\Omega)$ => $T^{i}_{\,jk}[\phi] = \Xi^i_{\,jk}[\phi] - \Gamma^i_{\,jk}[\phi]=0$

$\Xi^i_{\,jk}[\phi] = \Gamma^i_{\,jk}[\phi]$

$\blacksquare$

b00n heT
  • 16,360
  • 1
  • 36
  • 46
2

For the beginning you could define the following:

$$ T^{i}_{\,jk}[\phi](y) := \Xi^i_{\,jk}[\phi](y) - \Gamma^i_{\,jk}[\phi](y) $$ From the premises you know that there exists $\phi_0 \in$ Diff$(\Omega)$ and $y_0 \in \Omega$ so that $\Xi^i_{\,jk}[\phi_0](y_0) = 0$ and you can also show, that the Christoffel-Symbol is zero in those coordinates.

Then show that $T$ transforms the same way as a tensor, i.e. is a Tensor.

Now you only have to prove the invariance under diffeomorphisms and you are done!

stebu92
  • 378
2

The final/right solution should be the following:

We define:

$T^{i}_{\,jk}[\phi](y) := \Xi^i_{\,jk}[\phi](y) - \Gamma^i_{\,jk}[\phi](y)$

Thus we see that this transforms as a Tensor, hence is a Tensor:

$T^{i}_{\,jk}[\phi](y)= \Xi^i_{jk}[\phi](y) - \Gamma^i_{\,jk}[\phi](y) = \bigg(\frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) \Xi^r_{pq}[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y)) + \frac{\partial y^i}{\partial x^m} (\phi^{-1}(y)) \frac{\partial^2 x^m}{\partial y^j \partial y^k}(y) \bigg)- \bigg(\frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) \Gamma^r_{pq}[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y)) + \frac{\partial y^i}{\partial x^m} (\phi^{-1}(y)) \frac{\partial^2 x^m}{\partial y^j \partial y^k}(y)\bigg) = \frac{\partial x^p}{\partial y^j}(y) \frac{\partial x^q}{\partial y^k}(y) (\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}](\phi^{-1}(y))\frac{\partial y^i}{\partial x^r}(\phi^{-1}(y))$

Now we have that for our given $y_0$ and $\phi_0$:

$\Gamma^i_{\,jk}[\phi_0](\phi_0(y_0))=\frac{1}{2} g^{il}(\partial_k g_{lk} + \partial_j g_{lk} - \partial_l g_{jk})[\phi_0](\phi_0(y_0))= \frac{1}{2} g^{il}\left(\partial_k g_{lk}[\phi_0](\phi_0(y_0)) + \partial_j g_{lk}[\phi_0](\phi_0(y_0)) - \partial_l g_{jk}[\phi_0](\phi_0(y_0))\right)=0$

Now take any $y_0$ and any diffeomorphism $\phi\in Diff(\Omega)$ :

$T^{i}_{\,jk}[\phi_0](\phi(y_0))=\frac{\partial x^p}{\partial y^j}(\phi(y_0))\frac{\partial x^q}{\partial y^k}(\phi(y_0)) {\left({(\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}](y_0)}\right)}\frac{\partial y^i}{\partial x^r}(y_0)$

By the statement of task we can always find a $\phi_0$ such that:

$T^{i}_{\,jk}[\phi_0](\phi_0(y_0))=0=\frac{\partial x^p}{\partial y^j}(\phi_0(y_0))\frac{\partial x^q}{\partial y^k}(\phi_0(y_0)) {\left({(\Xi^r_{pq}-\Gamma^r_{\,pq})[\mathrm{id}](y_0)}\right)}\frac{\partial y^i}{\partial x^r}(y_0)$

for each $y_0\in\Omega$.

Since for any diffeomorphism we have that

$\frac{\partial x^p}{\partial y^j}(\phi(y))\frac{\partial x^q}{\partial y^k}(\phi(y))\neq 0$

We see that the transformed part which is invariant under any $\phi\in Diff(\Omega)$ is the one which maps to zero, we conclude for all of $\Omega$:

$T^i_{jk}[\phi]=\Xi^i_{jk}[\phi]-\Gamma^i_{jk}[\phi]\overset{\forall \phi}{=}0$ $$\Rightarrow \Xi^i_{jk}[\phi]=\Gamma^i_{jk}[\phi]$$

$\blacksquare$

b00n heT
  • 16,360
  • 1
  • 36
  • 46
  • a nice gathering of mmp students here :P unfortunately we aren't sure about it either. we have a similar solution but it still seems suspiciously easy. have you been succesful in proving ex9? did you explicitly show that the identity holds under diffeomorphisms? if so, could you please write to me to stephbur @ standard ethz address, maybe we can share some experience about that exercise – stebu92 Jan 19 '14 at 17:13
  • haha :D Yes, we have been able to solve it, I'll send you the solution in a "few minutes" okay? [if not, till tomorrow you'll have it] exactly... The main problem is that now that they have corrected the exercise the $\phi^{-1}(y_0)$ has changed in a $y_0$ (which of course makes much more sense, and that's why I originally introduced the $x_0$) and now I'm not sure anymore what to do... – b00n heT Jan 19 '14 at 17:18
  • That would be awesome. Thank you very much! Ill keep you updated on the progress of our discussion on ex8! – stebu92 Jan 19 '14 at 17:40