Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I'm asked to prove the following identity, using index notation:

$(\nabla\times A)\times A=A \cdot\nabla A - \nabla(A \cdot A)$

However, when I work it out, I find that the actual solution should be:

$(\nabla\times A)\times A=A \cdot\nabla A - \frac{1}{2}\nabla(A \cdot A)$

Am I missing something, or is the book wrong?

share|cite|improve this question
Here, A is a vector, of course. – Grave Oct 9 '12 at 5:28

Expanding everything out componentwise, I get

$(\nabla\times A)\times A$ $$ \begin{align} &{\lower{1pt}{\Large(}}(\partial_1,\partial_2,\partial_3)\times(A_1,A_2,A_3){\lower{1pt}{\Large)}}\times(A_1,A_2,A_3)\\ &=(\partial_2A_3-\partial_3A_2,\partial_3A_1-\partial_1A_3,\partial_1A_2-\partial_2A_1)\times(A_1,A_2,A_3)\\ &=(A_3\partial_3A_1+A_2\partial_2A_1-A_3\partial_1A_3-A_2\partial_1A_2,\\ &\hphantom{=(}A_1\partial_1A_2+A_3\partial_3A_2-A_1\partial_2A_1-A_3\partial_2A_3,\\ &\hphantom{=(}A_2\partial_2A_3+A_1\partial_1A_3-A_2\partial_3A_2-A_1\partial_3A_1) \end{align} $$ $(A\cdot\nabla)A$ $$ \begin{align} &(A_1\partial_1+A_2\partial_2+A_3\partial_3)(A_1,A_2,A_3)\\ &=(A_1\partial_1A_1+A_2\partial_2A_1+A_3\partial_3A_1,\\ &\hphantom{=(}A_1\partial_1A_2+A_2\partial_2A_2+A_3\partial_3A_2,\\ &\hphantom{=(}A_1\partial_1A_3+A_2\partial_2A_3+A_3\partial_3A_3) \end{align} $$ $\frac12\nabla(A\cdot A)$ $$ \begin{align} &\tfrac12(\partial_1,\partial_2,\partial_3)(A_1A_1+A_2A_2+A_3A_3)\\ &=(A_1\partial_1 A_1+A_2\partial_1 A_2+A_3\partial_1 A_3,\\ &\hphantom{=(}A_1\partial_2 A_1+A_2\partial_2 A_2+A_3\partial_2 A_3,\\ &\hphantom{=(}A_1\partial_3 A_1+A_2\partial_3 A_2+A_3\partial_3 A_3) \end{align} $$ And from these, I get $$ (\nabla\times A)\times A=(A\cdot\nabla)A-\tfrac12\nabla(A\cdot A) $$

share|cite|improve this answer
I think you forgot to change the $3$s to $2$s in the middle column of the result for $\frac12\nabla(A\cdot A)$. – joriki Oct 9 '12 at 10:48
@joriki: fixed, thanks. – robjohn Oct 9 '12 at 10:50

A similar proof in spirit to Shuhao Cao's, using geometric calculus.

In GC it's easier not to deal with the cross product or curl, but with the more fundamental entities: the generalized dot and wedge products, and the corresponding interior and exterior derivatives.

The basic relation needed to relate curl/cross product to these other operations is $a \times b = -i (a \wedge b)$. The quantity $i$ is called the unit pseudoscalar; it might also be called a volume form, and using it finds for you the Hodge dual of a given object. Here, that duality is captured in the natural multiplication operation called the geometric product. I won't worry over too many of these details; just know that $i$ commutes with all objects in 3d, and $i^2 = -1$.

Hence we see that

$$\begin{align*}(\nabla \times A) \times A &= -i [(\nabla \times A) \wedge A] \\ &= -i [(-i \{\nabla \wedge A\}) \wedge A] \\&= i^2 (\nabla \wedge A) \cdot A\end{align*}$$

where in the last step we've used the duality of the dot and wedge: $(iX) \wedge Y = i (X \cdot Y)$.

To analyze $(\nabla \wedge A) \cdot A$, we can use an analogue of the BAC-CAB rule.

$$(\nabla \wedge A) \cdot A = \dot \nabla (\dot A \cdot A) - (A \cdot \nabla) A$$

But by symmetry, $\dot \nabla (\dot A \cdot A) = \frac{1}{2} \nabla (A \cdot A)$. This basically completes the proof; in fact, nothing really exotic has been done except changing notation and using the BAC-CAB rule! Certainly this was a surprise to me, or else I never would've bothered to change notation in the first place.

share|cite|improve this answer
Nice answer, +1. – Shuhao Cao Jun 1 '13 at 20:03

There is a nice identity for smooth vector fields: $\newcommand{\b}{\boldsymbol}$

$$ \nabla (\b{u}\cdot \b{v}) = (\b{u}\cdot \nabla) \b{v} + (\b{u}\cdot \nabla) \b{v} - (\nabla \times\b{u})\times\b{v} - (\nabla \times\b{u})\times\b{v} .\tag{1} $$

For a component wise proof like robjohn gave, you can check Balanis's book. A more physical and geometrical proof would be using Feynman subscript notation: $$ \nabla(\b{u} \cdot \b{v})= \nabla_{\b{u}}(\b{u} \cdot \b{v}) + \nabla_{\b{v}} (\b{u} \cdot \b{v}) , $$ where the notation $\nabla_{\b{u}}$ means the gradient is applied only on $\b{u}$ and we keep $\b{v}$ fixed.

Geometrically, $\nabla_{\b{u}}(\b{u} \cdot \b{v}) $ can be decomposed into the normal derivative and the tangential derivative along $\b{v}$: $$ \nabla_{\b{u}}(\b{u} \cdot \b{v}) = (\b{v}\cdot\nabla)\b{u}- (\nabla \times \b{u})\times \b{v} , $$ hence (1) holds.

Now in your case $\b{u} = \b{v} = \b{A}$: $$ \nabla (\b{A}\cdot \b{A}) = (\b{A}\cdot \nabla) \b{A} + (\b{A}\cdot \nabla) \b{A} - (\nabla \times\b{A})\times\b{A} - (\nabla \times\b{A})\times\b{A}, $$ thus $$ (\nabla \times\b{A})\times\b{A} = (\b{A}\cdot \nabla) \b{A} - \frac{1}{2}\nabla (\b{A}\cdot \b{A}) . $$

share|cite|improve this answer
There's also a similar notation using overdots: i.e. Feynman's $\nabla_u (u \cdot v) = \dot \nabla (\dot u \cdot v)$. – Muphrid Jun 1 '13 at 19:39
@Muphrid Thanks, good to know. – Shuhao Cao Jun 1 '13 at 19:41

Your Answer


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.