On Wolfram Mathworld they give a unit vector in the $\textbf{x}_n $ direction as: $$ \hat{\textbf{x}_n} \equiv \dfrac{\frac{\partial \textbf{r}}{\partial x_n}}{\lvert \frac{\partial \textbf{r}}{\partial x_n} \rvert} $$

Can anyone give an intuitive interpretation of this? Alternatively, seeing a proof would help explain where this equation comes from, and help me out.

My current understanding is as follows. For an example in polar coordinates, $\dfrac{\partial \textbf{r}}{\partial \theta}$ is the change in $\textbf{r}$ generated by a change of $\theta \rightarrow \theta+d\theta$. That change is a movement along the $\theta$ direction, as shown in the picture below.

enter image description here

I'm not too sure, however, if this is a valid interpretation.

  • $\begingroup$ I think what Wolfram Alpha is doing is saying that dr/dxn is a small vector in the direction of xn, similar to your explanation but instead of changing with theta, it's changing with xn in some way. Then, once we have a vector in the direction of xn, we divide by its norm and get a unit vector. I don't know why they didn't just use xn to begin with, which is itself a vector in the direction of xn (of course). $\endgroup$
    – NicNic8
    Oct 28, 2014 at 2:46

2 Answers 2


That is overly complicated. A unit vector is a simple thing.

A unit vector is just a vector with size 1. You can get a unit vector just by dividing a vector by its norm:


Now $\hat{x}$ is a unit vector pointing in the same direction as $x$.

Wolfram Alpha is giving an example of that.


Your interpretation is correct. Consider the spherical coordinate system, $(r,\theta,\phi)$. Then the unit vectors $\hat{\mathbf{r}}$, $\hat{\boldsymbol{\theta}}$, and $\hat{\boldsymbol{\phi}}$ are defined as the direction that an arbitrary vector $\mathbf{R}(r,\theta,\phi)$ moves in when the coordinates $r$, $\theta$, and $\phi$ are infinitesimally tweaked.

To be concrete, we can write $\mathbf{R}(r,\theta,\phi)$ in Cartesian coordinates:

$$\mathbf{R}(r,\theta,\phi) = r\sin\phi\cos\theta \,\hat{\mathbf{i}} + r\sin\phi\sin\theta \,\hat{\mathbf{j}} + r\cos\phi\,\hat{\mathbf{k}}.$$

Then $\hat{\mathbf{r}}$ is defined to be along the direction that $r$ changes along: $$ \hat{\mathbf{r}}\propto \frac{d\mathbf{R}}{dr} = \sin\phi\cos\theta \,\hat{\mathbf{i}} + \sin\phi\sin\theta \,\hat{\mathbf{j}} + \cos\phi\, \hat{\mathbf{k}}, $$ and similarly $$ \hat{\boldsymbol{\theta}}\propto \frac{d\mathbf{R}}{d\theta} = -r\sin\phi\sin\theta\,\hat{\mathbf{i}} + r\sin\phi\cos\theta \,\hat{\mathbf{j}} $$ and $$ \hat{\boldsymbol{\phi}}\propto \frac{d\mathbf{R}}{d\phi} = r\cos\phi\cos\theta \,\hat{\mathbf{i}} + r\cos\phi\sin\theta \,\hat{\mathbf{j}} - r\sin\phi\, \hat{\mathbf{k}}. $$

We also have the condition that $|\hat{\mathbf{r}}| = |\hat{\boldsymbol{\theta}}| = |\hat{\boldsymbol{\phi}}| = 1$, which means we have to normalize the vectors: this where the division by the magnitude comes from. For example, $$\hat{\boldsymbol{\theta}} = \frac{\frac{d\mathbf{R}}{d\theta}}{|\frac{d\mathbf{R}}{d\theta}|} = \frac{-r\sin\phi\sin\theta\,\hat{\mathbf{i}} + r\sin\phi\cos\theta \,\hat{\mathbf{j}}}{\sqrt{(-r\sin\phi\sin\theta)^2 + (r\sin\phi\cos\theta)^2}} = -\sin\theta \hat{\mathbf{i}}+\cos\theta \hat{\mathbf{j}}.$$


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.