So I have read that components of vectors transform linearly between frames like $$ A'^\mu = \frac{\partial x'^\mu}{\partial x^\nu} A^\nu .$$

I have also read that coordinates in general do not transform linearly between coordinate systems, however their differentials do:

$$ dx'^\mu = \frac{\partial x'^\mu}{\partial x^\nu} dx^\nu .$$

What I cannot grasp is how on earth can vector components transform between coordinate systems linearly when the coordinates do not? What if the vector in question is a position vector, then the components are the coordinates and then we come to a contradiction? At this point I become confused.

Also, in deriving the vector component transformation, I have always seen the example of transforming from a Cartesian $x$,$y$ plane to another one rotated at an angle $\theta$ relative to the first. Well in this example the components of the vector transform linearly and I can completely see this, but so do the coordinates!

Any help would be much appreciated!

  • $\begingroup$ What do you mean by coordinates do not transform linearly? The transformation between spacetime coordinates is just a Lorentz transformation $\Lambda_{\mu}^{\nu}$. $\endgroup$ – InertialObserver Jan 2 '17 at 19:42
  • 1
    $\begingroup$ Transforming from Cartesian to spherical polar is not linear $\endgroup$ – Matt0410 Jan 2 '17 at 19:44
  • $\begingroup$ Oh, so you are saying the transformation between different coordinates in that sense, not in the sense of "different reference frame". $\endgroup$ – InertialObserver Jan 2 '17 at 19:45
  • 1
    $\begingroup$ Do you mind my asking where you heard that components of a vector transform linearly? That is not quite correct. $\endgroup$ – InertialObserver Jan 2 '17 at 20:01
  • 1
    $\begingroup$ Well the first equation above is in all textbooks I have read on General Relativity. Doesn't that imply a linear sum of the unprimed components? $\endgroup$ – Matt0410 Jan 2 '17 at 20:11

In a general curved spacetime, vectors are only defined in the tangent space of individual points, and are by definition part of a vector space. This is the only thing that makes sense since one cannot assign a consistent rule to map vectors at one point to another in a curved spacetime. This is related to the fact that there is a non-trivial holonomy when there is non-zero curvature. Coordinates on the other hand are defined on larger patches of the spacetime.

General transformations of coordinates could be non-linear, but they would always induce a local transformation on vectors within their own tangent spaces. This would then result in a linear transformation on the vectors.

The example of the Cartesian plane is confusing because it has zero curvature. Try this with some other non-trivial manifold like a sphere.


For clarification you have to distinguish between coordinate systems and sets of base vectors. The linearity in transforming the vector components is a consequence of the properties of vector spaces. Coordinate transformations are somewhat different from base changes.


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.