In the process of getting the formula of Christoffel symbols in the terms of metric tensors, we get

$$\partial_n g_{rm}+\partial_m g_{rn}-\partial_r g_{mn}=2\Gamma ^t_{mn} g_{rt}. $$

The book I have read goes ' multiply both sides of the equation by the inverse metric tensor' and yields

$$\Gamma ^t_{mn}={1\over 2} g^{rt}(\partial_n g_{rm}+\partial_m g_{rn}-\partial_r g_{mn}).$$

But I am confused with dummy variables. In the first equation, r was the real variable and t was a dummy variable, but in the second one, the opposite seems to be true. If we write the right hand side of the first equation without summation, it'll be

$$2(\Gamma ^0_{mn} g_{r0}+\Gamma ^1_{mn} g_{r1}+\Gamma ^2_{mn} g_{r2}+\Gamma ^3_{mn} g_{r3})$$

I will multiply this with $g^{rt}$. In this case it's for t=0, 1, 2, 3, so the left hand side is $$(\Gamma ^0_{mn} g_{r0}+\Gamma ^1_{mn} g_{r1}+\Gamma ^2_{mn} g_{r2}+\Gamma ^3_{mn} g_{r3})(g^{r0}+g^{r1}+g^{r2}+g^{r3}) \\ = \Gamma ^0_{mn}+\Gamma ^1_{mn}+\Gamma ^2_{mn}+\Gamma ^3_{mn}$$ since $g_{rm}g^{rn}=\delta ^n_m$. (Let's divide both sides with 2)

The left hand side will be $${1\over 2} (g^{r0}+g^{r1}+g^{r2}+g^{r3})(\partial_n g_{rm}+\partial_m g_{rn}-\partial_r g_{mn})$$ and it is not a summation with r because r is not a dummy variable.

So I don't know the way to get to the right result from the last equation. And how the real and dummy variables changed during the calculation.

  • $\begingroup$ Because t is dummy variable you can change it with any letter you want, for example change t in first equation with an other letter like f and try solution $\endgroup$
    – Sancol.
    Feb 27 at 14:55

2 Answers 2


No, it is not working like that.

As one disposes of ( as one has 3 different indices r, m, n which are not contracted of which each runs from 0 to 3 -- so 4 different values) $4\times 4\times 4$ equations (some of these might be identical due to symmetry relations) one can combines these equations upon multiplication with the metric tensor $g^{rs}$ --- i.e. the multiplication with the metric tensor is not just multiplication but also implicates summation:

$$g^{rs}(\partial_n g_{rm} + \partial_m g_{rn} -\partial_r g_{nm})= 2g^{rs}g_{rt}\Gamma^t_{mn}$$

which actually means: $$\sum\limits_{r=0,\ldots, 3} g^{rs}(\partial_n g_{rm} + \partial_m g_{rn} -\partial_r g_{nm})=\sum\limits_{r=0,\ldots, 3} 2 g^{rs}g_{rt} \Gamma^t_{mn}$$

This means that one takes from the $4\times 4\times 4$ equations $4\times 4$ eqns (for m and n) but with fixed $r=0$ multiplying it with $g^{0s}$, adds it to the next set of equations for $r=1$ multiplying that one with $g^{1s}$ and so forth, i.e. also adds the set of eqns for $r=2$ weighted by $g^{2s}$ and for $r=3$ weighted by $g^{3s}$. And of course one does it for each $r$ on LHS and RHS.

Then one gets: $$ \sum\limits_{r=0,\ldots, 3} g^{rs}\frac{1}{2}(\partial_n g_{rm} + \partial_m g_{rn} -\partial_r g_{nm})= \delta^s_t\Gamma^t_{mn} = \Gamma^s_{mn}$$


$$ \sum\limits_{r=0,\ldots, 3} g^{rs}g_{rt} = \delta^s_t$$

where $\delta^s_t$ is the Kronecker symbol. Or applying Einstein's summation convention the summation symbol can be omitted.

$$g^{rs}\frac{1}{2}(\partial_n g_{rm} + \partial_m g_{rn} -\partial_r g_{nm})= \delta^s_t\Gamma^t_{mn} = \Gamma^s_{mn}$$

As the very last step one can replace the index $s$ by $t$ again. But the name of an index can be chosen (almost) arbitrarily --- better do not use indices which are already used in the equation.


Act on your first equation with $g^{ar}$. Then the action on the right-hand side will produce $2\Gamma^t_{mn} \delta^a_t$. Contracting with the Kronecker delta and rearranging yields: $$ \Gamma^a_{mn} = \frac{1}{2} g^{ar} \left( \partial_n g_{rm} + \partial_m g_{rn} - \partial_r g_{mn}\right) $$ This is identical to the expression in your second line (with the free index $t$ replaced with $a$, and $g^{ab} = g^{ba}$).

You are right about what is free and what is dummy. When you contract with the "inverse metric" you should use a new free index, $a$, and the old free index $r$. The book then, confusingly, renames the new free index $t$ again, even though it was summed over in the calculation.


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.