Two questions here, both related to one another.

SOLVED Question 1. I am looking to transform a dual vector $p_a$ from spherical polar coordinates to Cartesian coordinates. My dual vector is given generally by $p_a=(p_r,p_\theta,p_\phi)$. Transforming this would usually be done using the transformation matrix \begin{equation} \Lambda^a_{\;a'}= \begin{pmatrix} \frac{\partial x}{\partial r} & \frac{\partial y}{\partial r} & \frac{\partial z}{\partial r} \\ \frac{\partial x}{\partial \theta} & \frac{\partial y}{\partial \theta} & \frac{\partial z}{\partial \theta} \\ \frac{\partial x}{\partial \phi} & \frac{\partial y}{\partial \phi} & \frac{\partial z}{\partial \phi} \end{pmatrix}, \end{equation} but the result of $\Lambda^a_{\;a'}p_a$ is: \begin{equation} \Lambda^a_{\;a'}p_a= \begin{pmatrix} p_\theta r \cos\theta \cos\phi + p_r \cos\phi \sin\theta - p_\phi r \sin\theta\sin\phi \\ p_\phi r \cos\phi \sin\theta + p_\theta r \cos\theta \sin\phi + p_r \sin\theta \sin\phi\\ p_r \cos\theta- p_\theta r \sin\theta \end{pmatrix}. \end{equation} This would equal $(p_x,p_y,p_z)$ if not for the factors of $r$ beside $p_\theta$ and $p_\phi$. I'm clearly missing something in this transformation, can someone point me in the right direction?

Question 2. I am also working on a transformation of a rank 3 tensor by the same means: $$T^{a'b'c'}=\Lambda^{a'}_{\;a}\Lambda^{b'}_{\;b}\Lambda^{c'}_{\;c}T^{abc}.$$ The $\Lambda$'s here should include the flipped derivatives of above (i.e. $\frac{\partial x}{\partial r}\rightarrow \frac{\partial r}{\partial x}$). When I do the final multiplication of $T^{a'b'c'}p_{b'}p_{c'}$ (where $p_{c'}$ is $(p_x,p_x,p_z)$) I get a result which seems to include factors such as $(p_x^2+p_y^2+z^2p_z^2)$, which obviously has some unit problems. Should the components of the $\Lambda$'s include any prefactors such as $r$ or $r\sin\theta$, i.e.

$\begin{pmatrix} \frac{\partial r}{\partial x} & \frac{\partial \theta}{\partial x} & \frac{\partial \phi}{\partial x} \\ \frac{\partial r}{\partial y} & \frac{\partial \theta}{\partial y} & \frac{\partial \phi}{\partial y} \\ \frac{\partial r}{\partial \phi} & \frac{\partial \theta}{\partial z} & \frac{\partial \phi}{\partial z} \end{pmatrix}$ or $\begin{pmatrix} \frac{\partial r}{\partial x} & r\frac{\partial \theta}{\partial x} & r \sin\theta\frac{\partial \phi}{\partial x} \\ \frac{\partial r}{\partial y} & r\frac{\partial \theta}{\partial y} & r\sin\theta\frac{\partial \phi}{\partial y} \\ \frac{\partial r}{\partial \phi} & r\frac{\partial \theta}{\partial z} & r\sin\theta\frac{\partial \phi}{\partial z} \end{pmatrix}$

  • $\begingroup$ Why do you think that the factors of $r$ shouldn't be there? Remember that because $r$ and $\theta$ have different units so do $p_r$ and $p_\theta$. $\endgroup$
    – mmeent
    Jan 17, 2018 at 8:02
  • $\begingroup$ This should be obvious, but somehow I did not know that. Thanks! I've edited the second part of the question, perhaps you have some insight for that part as well? @mmeent $\endgroup$
    – Karl
    Jan 17, 2018 at 15:49
  • $\begingroup$ It's not clear you need a $3$rd order tensor - it looks like a vector transforming from $(x,y,z) \rightarrow (r,\theta,\phi)$. Here's a $3$rd order tensor $T=T^{ijk}\frac{\partial}{\partial u^i} \otimes \frac{\partial}{\partial u^{j}}\otimes \frac{\partial}{\partial u^{k}}$ which would require one $3\times3$ matrix for each vector. What $3$ vectors are you tensoring together? $\endgroup$ Aug 21, 2019 at 9:14

1 Answer 1


If you transform a dual vector $p_a$ (covariant object) from spherical coordinates labelled $a$ to cartesian ones $a'$ the transformation law is:
$p_{a'} = \Lambda^a_{a'} p_a$
where $\Lambda^a_{a'}$ are the partial derivatives of the spherical coordinates (old coordinates) vs. the cartesian ones (new coordinates), that is $\frac{\partial r}{\partial x}$, etc.
The matrix in question 1. shows the opposite.
As for question 2. the comment is similar, remembering that now you transform a contravariant object, hence the $\Lambda^{a'}_a$ require the partial derivatives of the cartesian coordinates (new coordinates) vs. the spherical ones (old coordinates), that is $\frac{\partial x}{\partial r}$, etc.


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.