Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I sort of asked a version of this question before and it was unclear; try I will now to make an honest attempt to state everything clerly.

I am trying to evaluate the following, namely $\nabla w = \nabla |\overrightarrow{a} \times \overrightarrow{r}|^n$, where $\overrightarrow{a}$ is a constant vector and $\overrightarrow{r}$ is the vector $<x_1,x_2,\ldots x_n>$. Now say that I use the chain rule, first say by setting $\overrightarrow{u}$ to be equal to the cross product of $\overrightarrow{a}$ and $\overrightarrow{r}$.

Now here's the part that I'm confused. How do we extend the chain rule over when dealing with the gradient? Do I take $\nabla |\overrightarrow{a} \times \overrightarrow{r}|^n$ to be equal to $\nabla |\overrightarrow{u}|^n$ $\times$ $\nabla (\overrightarrow{a} \times \overrightarrow{r})$, where $\times$ denotes the cross product?

The first bit $\nabla |\overrightarrow{u}|^n$ is easy, it just evaluates to $n|\overrightarrow{a} \times \overrightarrow{r}|^{n-2} (\overrightarrow{a} \times \overrightarrow {r})$, remembering that $\overrightarrow{u} = \overrightarrow{a} \times \overrightarrow{r}$.

I am guessing that $\nabla |\overrightarrow{a} \times \overrightarrow{r}|^n$ $\neq$ $\nabla |\overrightarrow{u}|^n$ $\times$ $\nabla (\overrightarrow{a} \times \overrightarrow{r})$, as to even speak about $\nabla (\overrightarrow{a} \times \overrightarrow{r})$, i.e. the gradient of a vector we would have to talk about either the cross product or dot product of the gradient and $\nabla (\overrightarrow{a} \times \overrightarrow{r})$

By the way, I am told the answer given is $\nabla |\overrightarrow{a} \times \overrightarrow{r}|^n$ = $n|\overrightarrow{a} \times \overrightarrow{r}|^{n-2} \Big( \quad \overrightarrow{a} \times (\overrightarrow{r} \times \overrightarrow{a})\Big)$.

So let's say that I try a component wise approach, i.e. we look first at $\frac{\partial w}{\partial x_1}$. Then is it true (I could be wrong) that:

$\frac{\partial w}{\partial x_1} = n|\overrightarrow{a} \times \overrightarrow{r}|^{n-2} \quad \overrightarrow{u_1} \times \frac{\partial}{\partial x_1} \Big(\overrightarrow{a} \times \overrightarrow{r}\Big) = n|\overrightarrow{a} \times \overrightarrow{r}|^{n-2} \quad \overrightarrow{u_1} \times \Big(\overrightarrow{a} \times \frac{\partial\overrightarrow{r}}{\partial x_1}\Big)$, as $\overrightarrow{a}$ is a constant vector? Here $\overrightarrow{u_1}$ denotes the first component of the vector $\overrightarrow{a} \times \overrightarrow{r}$.

I would really aprreciate an interpretation of this, it is just that I am confused about what to take and the meanings of these operations.

Thanks,

Ben

share|improve this question
add comment

2 Answers

up vote 3 down vote accepted

Let $f(r)=|a \times r|$. (I'll be lazy and omit the vector arrows.)

Then, to begin with, $\nabla(f(r)^n) = n f(r)^{n-1} \nabla f(r)$ by the chain rule. So it only remains to compute $\nabla f(r)$, and if you can't think of any clever way, then you just do it by brute force in terms of coordinates: take $a=(a_1, a_2, a_3)$ and $r=(x_1,x_2,x_3)$ (the cross products only exists in three dimensions, so your $x_n$ doesn't make sense), compute the components of $u=a \times r=(u_1,u_2,u_3)$, then compute the length $f(r)=\sqrt{u_1^2 + u_2^2 + u_3^2}$, and compute the derivatives of that with respect to $x_1$, $x_2$ and $x_3$. Done!

share|improve this answer
3  
The clever way being to use $|a \times r|^2 = |a|^2 |r|^2 - (a . r)^2$ –  Robert Israel Apr 11 '11 at 16:44
    
@RobertIsrael @HansLundmark Hi guys thanks for your replies. My comment is too long so I've put it as an answer, please have a look at it. –  fpqc Apr 11 '11 at 22:44
add comment

Say I differentiate $|\mathbf{a} \times \mathbf{r}|^2$. Then for the second term it agrees with $\mathbf{a} \times (\mathbf{r} \times \mathbf{a})$, but for the first term there is some confusion. Say I look at the first term of $\nabla |\mathbf{a}|^2|\mathbf{r}|^2$, namely $|\mathbf{a}|^2|\mathbf{r}|^2_{x_1} \mathbf{e_1}$ = $|\mathbf{a}|^2|\mathbf{r}|\mathbf{r}_{x_1} $ $\bullet$ $\mathbf{e_1}$? If it is indeed the dot product then I'm done! Oh by the way $\mathbf{r}_{x_1}$ means the partial derivative of $\mathbf{r}$ with respect to $x_1$.

Ben

share|improve this answer
    
If you're not sure how to do it the clever way, you should fall back to the brute force method. ;) Anyway, your formula is not correct. The term you're asking about is $$\frac{\partial}{\partial x_1}(|a|^2 |r|^2) e_1 = 2 |a|^2 x_1 e_1 = 2 |a|^2 (r \cdot e_1) e_1.$$ And if you don't know why $\frac{\partial}{\partial x_1}(|r|^2) = 2 x_1$, then write it out in coordinates, $|r|^2=\sqrt{\dots}^2$, so that you can see what's going on. –  Hans Lundmark Apr 12 '11 at 6:00
    
@HansLundmark sorry my bad I went a step too quick. $|r|$^2 is just like you said $x_1^2 + x_2^2 \ldots x_n^2$ which when differentiated w.r.t. $x_1$ gives $2x_1$, thanks so much for your help. –  fpqc Apr 12 '11 at 6:06
    
You're welcome. :) –  Hans Lundmark Apr 12 '11 at 6:14
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.