Take the defining equation of Christoffel symbols: $$\nabla_{\frac{\partial}{\partial x^\mu}}\frac{\partial}{\partial x^\nu}=\Gamma^{\sigma}_{\mu\nu}\frac{\partial}{\partial x^{\sigma}}$$ Both sides of the above definition are vector fields, in fact, the right side being a linear combination of coordinate vector fields with the coefficients of the combination being precisely the Christoffel symbols $\Gamma^{\sigma}_{\mu\nu}$ that do not transform as tensor. The above fact motivates my following question: If one has a vector field $X$ written out in a chart $x^\mu$ as $X=X^{\mu}\frac{\partial}{\partial x^\mu}$, is it the case that the smooth functions $X^\mu$ on the manifold should always transform as a vector? The right side of the definition above for Christoffel symbols suggest that this claim is not true.

  • 1
    $\begingroup$ What makes you say that? The Christoffel symbols are defined at the level of the chart, not the level of the manifold - that's largely the point, that different choices in chart lead to non-trivial differences in the connection coefficient functions. $\endgroup$
    – J. Murray
    Oct 10, 2017 at 4:26
  • $\begingroup$ If I write a vector field $X=f_{i}\frac{\partial}{\partial x^{i}}$, do the smooth functions $f_i$'s on the manifold necessarily transform as vector? $\endgroup$
    – Hasib
    Oct 10, 2017 at 4:34
  • $\begingroup$ No, but that's because you've defined it using a chart already. You would have to make sure that it had the appropriate transformation properties once you expressed it in a different chart. The Christoffel symbols don't. $\endgroup$
    – J. Murray
    Oct 10, 2017 at 4:39
  • $\begingroup$ So, in an appropriate chart I can always make it transform as a vector but in general the $f_i$'s need not transform as a vector, right? Could you please give a reference (Chapter, page etc.) to delve further into it. $\endgroup$
    – Hasib
    Oct 10, 2017 at 4:43
  • $\begingroup$ It should be noted that you haven't actually defined a vector field by expressing it in a chart. A vector field is defined at the manifold level as a "directional derivative" along a smooth curve on the manifold. It can then be expressed in whatever chart you'd like. I don't have any good references on hand, but perhaps somebody else will. $\endgroup$
    – J. Murray
    Oct 10, 2017 at 4:46

2 Answers 2


If you have some $\{f^\mu\}\subset C^\infty(U)$ where $U$ is some coordinate domain, then $$\tag{$1$}X=f^\mu \partial_\mu$$ is indeed a vector field in $U$. If with respect to some other coordinate system, we have $$X=g^{\mu'}\partial_{\mu'},$$ then we will of course have $$\tag{$2$}g^{\mu'}=\frac{\partial x^{\mu'}}{\partial x^\mu} f^\mu.$$ But if $\{f^\mu\}$ has a transformation law other than (2), it does not have to be that $$g^{\mu'}=f^{\mu'}.$$ So basically, if you define a vector field by (1) and then transform, you have to forget whatever auxiliary transformation law the multiplet $\{f^\mu\}$ has.


A geometric object can be defined by its local expression in a particular coordinate system combined with rules governing coordinate change.

So if you treated the lower indices as mere labels, for each choice $n,k$ of $\nu, \kappa$ you could take the right-hand side as the definition of a local vector field

$$ Y_{nk} = \Gamma^{\mu}_{nk}\frac{\partial}{\partial x^{\mu}} $$

However, these fields would need to transform under coordinate change according to

$$ {\tilde Y}^\mu_{nk} = \frac{\partial {\tilde x}^\mu}{\partial x^{\alpha}} \Gamma^{\alpha}_{nk} $$

which in general no longer agrees with the Christoffel symbols expressed in the new coordinates, which are given by

$$ \tilde \Gamma^{\mu}_{\nu\kappa} = {\partial \tilde x^\mu \over \partial x^\alpha} \left [ \Gamma^\alpha_{\beta \gamma}{\partial x^\beta \over \partial \tilde x^\nu}{\partial x^\gamma \over \partial \tilde x^\kappa} + {\partial ^2 x^\alpha \over \partial \tilde x^\nu \partial \tilde x^\kappa} \right ] $$


Because the rules governing coordinate change are different, you're dealing with different geometric objects, whose coordinate expression may just coincidentally agree in a particular coordinate system.

  • $\begingroup$ In $Y = \Gamma^{\mu}_{~\nu\kappa}\partial_{\mu}$ the indices are not balanced, so the extra 2 indices should be ignored. Therefore, you cannot use the transformation formula to conclude anything, because this is made to an object with 3 "active" indices, not to one with 2 dummy ones. $\endgroup$
    – DanielC
    Oct 10, 2017 at 14:55
  • $\begingroup$ The motivation behind posing the Christoffel symbol definition was as follows: If in a chart $(U,x^{i})$ one writes a vector field $$Y\vert_{U}=f_{i}\vert_{U}\frac{\partial}{\partial x^{i}}|_{U}$$, does the smooth functions $f_{i}$'s on $U$, necessarily transform as a contravariant vector? We see from the definition of the Christoffel symbols that the coefficients $f_{i}$'s don't even transform as a tensor, yet the linear combination rendering a vector field in that chart. Am I missing anything? $\endgroup$
    – Hasib
    Oct 10, 2017 at 15:48
  • $\begingroup$ @ Christoph, in the second line above, please correct the typo: should be $\tilde{\Gamma}^{\mu}_{nk}$ instead of $\tilde{Y}^{\mu}_{nk}$. So, you are saying that the object $Y_{nk}$ is not any geometric object, local vector field, in particular, as it behaves inconsistently under change of charts. So, the defining equation of Christoffel symbols only applies to a chart, right? $\endgroup$
    – Hasib
    Oct 11, 2017 at 4:27

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.