In one of my lecture, it is said:

Let us use the freedom of the choice of parametrization to demand that the variation of $\lambda$ after a small displacement along the curve is proportional to the length of the displacement. Such parameter $\lambda$ is called affine. One has thus $$ d\lambda^2 = K\,g_{ab}\,dx^a dx^b \quad (\mbox{with $K$ constant})\;\rightarrow\;g_{ab}{x'}^a{x'}^b = K^{-1} $$

The previous equation is then reduced to $$ \boxed{ \frac{d}{d\lambda}(g_{ac}{x'}^a) = \tfrac12 {x'}^a{x'}^b \,\partial_cg_{ab} }$$

This is the first form of the equations of the geodesic curves, or curve of extremal length, in the affine parametrization.

Remark $\rightarrow$ For such a definition of geodesics to make sense, it is necessary that a metric exists (only case considered here), and it has to be positive definite.

The geodesic equation can take another form by expanding the left hand side: $$ \begin{aligned} g_{ac}{x''}^a + &\underbrace{\frac{dg_{ac}}{d\lambda} {x'}^a} = \tfrac12 {x'}^a {x'}^b \,\partial_c g_{ab}\\ &=\partial_b g_{ac}\frac{dx^b}{d\lambda} {x'}^a = {x'}^a{x'}^b \partial_b g_{ac} = \tfrac12 (\partial_a g_{bc} + \partial_c g_{ac}){x'}^a{x'}^b \end{aligned} $$ $$ g_{ac}{x''}^a+\tfrac12(\partial_a g_{bc} + \partial_c g_{ac}-\partial_cg_{ab}){x'}^a{x'}^b = 0 $$ multiply by $g^{cd}$ $$ \boxed{ {x''}^d + \tfrac12 g^{cd}(\partial_a g_{bc} + \partial_c g_{ac}-\partial_cg_{ab}) {x'}^a{x'}^b = 0 } $$

How to prove at the bottom of page that:

$$x'^{a}\,x'^{b}\,\partial_{b}g_{ac} = \dfrac{1}{2}(\partial_{a}g_{bc}+\partial_{b}g_{ac})\,x'^{a}\,x'^{b}$$

that would imply that: $\partial_{b}g_{ac}=\partial_{a}g_{bc}$ , is it right?


Perhaps $ x^a x^b \partial_b g_{ac} =\\ =1/2\cdot x^a x^b \partial_b g_{ac}+1/2\cdot x^a x^b \partial_b g_{ac}\\ =1/2\cdot x^a x^b \partial_b g_{ac}+1/2\cdot x^b x^a \partial_b g_{ac}\\ =1/2\cdot x^a x^b \partial_b g_{ac}+1/2\cdot x^a x^b \partial_a g_{bc} $

I am not entirely confident that substitutions are allowed on subformulas like that but if they are this seems to do the trick.

So the answer is for most instances no; the equality is for the fully contracted term, not the partial derivatives of the tensor components in isolation.

  • $\begingroup$ thanks a lot ! that was the trick $\endgroup$ – youpilat13 May 4 '19 at 16:48

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.