I'm trying to prove or disprove that if $\vec n(x,y,z)$ is a unit vector, then $(\vec n\cdot\nabla)\vec n$ is orthogonal to $\vec n$. For this I first tried to compute $\vec n\cdot((\vec n\cdot\nabla)\vec n)$ and show that it's zero. I wrote it out in components, but the resulting expression contained only additions, no subtractions at all, so I couldn't simply cancel out something.

Then I wrote $\vec n$ as

$$\vec n=\vec n_0/|\vec n_0|$$

and tried to do the same with its components. But the expression appeared too large to handle manually. I then used Wolfram Mathematica to check that $\vec n\cdot((\vec n\cdot\nabla)\vec n)=0$ symbolically, and it confirmed it.

But I still wonder, are there any ways of proving this without needing to handle large intermediate expressions?

  • $\begingroup$ Not sure what you mean by $(\vec n \nabla) \vec n$; could it possibly be the same as $\nabla_{\vec n} \vec n$, the (covariant?) derivative of $\vec n$ with respect to $\vec n$ itself? That is, should we take $\vec n \nabla$ to be $\vec n \cdot \nabla$? $\endgroup$ – Robert Lewis May 16 '15 at 20:17
  • $\begingroup$ @RobertLewis It means dot product of $\vec n$ and $\nabla$, acting on $\vec n$, i.e. $(n_x\partial_x+n_y\partial_y+n_z\partial_z)\vec n$. $\endgroup$ – Ruslan May 16 '15 at 20:18
  • $\begingroup$ Right; thanks; Cheers! $\endgroup$ – Robert Lewis May 16 '15 at 20:18

A demonstration sans coordinates:

We have

$\langle \vec n, \vec n \rangle = 1; \tag{1}$

thus, for any vector field $\vec v$,

$(\vec v \cdot \nabla) \langle \vec n, \vec n\ \rangle = 0; \tag{2}$


$(\vec v \cdot \nabla) \langle \vec n, \vec n \rangle = \langle (\vec v \cdot \nabla) \vec n, \vec n\ \rangle + \langle \vec n,(\vec v \cdot \nabla)\vec n \rangle$ $= 2\langle \vec n,(\vec v \cdot \nabla) \vec n \rangle; \tag{3}$

combining (2) with (3) yields

$\langle \vec n,(\vec v \cdot \nabla)\vec n \rangle = 0; \tag{4}$

we now take $\vec v = \vec n$, whence

$\langle \vec n,(\vec n \cdot \nabla)\vec n \rangle = 0, \tag{5}$

showing, as required, that $\vec n$ is orthogonal to $(\vec n \cdot \nabla) \vec n$.

Note we have in fact established the much more general (4); that is, $\vec n$ is orthogonal to $(\vec v \cdot \nabla)\vec n$ for any $\vec v$.


Let $(n\cdot\nabla) n=(n_1 \partial_1+n_2\partial_2 + n_3\partial_3) n$. Then $n\cdot ((n\cdot\nabla)n)$ equals $\sum_{i,j} n_i (n_j \partial_j) n_i=\sum_{i,j} n_i n_j (\partial_j n_i)$.

Applying the gradient to $n\cdot n=const$ gives $\sum_j (\partial_i n_j) n_j=0$ for all $i$. Multiplying the last vector by $n$, we get that $$ \sum_{i,j} n_i (\partial_i n_j) n_j=0. $$


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.