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

It's very very often and in many books depending on author there is mostly interchange of gradient $D$ and divergence $\nabla$.

What should I be careful about while dealing with such cases?
For example say $\int_\Omega v\Delta u$ using integration by parts and assuming that the boundary term vanishes this becomes $\int_\Omega Dv\cdot Du$.

Very often I also see the expression $\int_\Omega\nabla v\cdot\nabla u $, so this leads me to confusion sometimes.

share|cite|improve this question
$\Delta=\nabla^2$ doesn't really fit in your pattern: So this leads me to confusion... – draks ... Jul 26 '12 at 19:25
I suggest writing $\mathrm{div}$ for divergence. It's annoying that \div is used for division in $\LaTeX$ but you can DeclareMathOperator. – user31373 Jul 26 '12 at 23:01
up vote 2 down vote accepted

These are cases when index notations becomes extremely helpful. Denote by $\partial_1 = \partial_x, \partial_2 = \partial_y, \partial_3 = \partial_z$ and for a vector write its components as $$ \vec{v} = v_1 \hat{x} + v_2 \hat{y} + v_3 \hat{z}$$ Then the divergence of a vector field is $$ \vec{\nabla}\cdot\vec{v} = \sum_{i = 1}^3 \partial_i v_i $$ The gradient vector of a scalar field has components $$ (\vec{\nabla f})_i = \partial_i f $$ and the Laplacian of a function is $$ \triangle f = \sum_i \partial_i \partial_i f $$

When doing computations, do them term by term (that is, expand out the sums if necessary) and in the end, when the $\partial_i$ hits on a scalar quantity, it is a gradient, and when the $\partial_i$ hits on a vector quantity indexed also by $i$ you get a divergence.

To illustrate: if $f$ be function a $\vec{v}$ a vector field, then the object $f \vec{v}$ is a vector field. Let us compute its divergence. We do it like so:

$$ \vec{\nabla}\cdot(f\vec{v}) = \sum_i \partial_i \left( f v_i\right)= \sum_i (\partial_i f)v_i + f \partial_i v_i \tag{1}$$

The first term then is interpreted as the dot product of the gradient vector $\vec{\nabla f}$ against the vector $\vec{v}$, so for this term "the divergence outside changed to a gradient inside". The second term contains the divergence of the vector field $\vec{v}$ it self, multiplied against the function $f$. So for this term the divergence remains a divergence.

The use of index notations is also extremely helpful when there are multiple sets of indices in play. Let $\vec{v}$ be a vector field, we can consider the gradient of its divergence $\vec{\nabla}(\vec{\nabla} \cdot \vec{v})$. Now, since partial derivatives commute, what happens when we try to interchange the derivatives? In index notation you have

$$ \underbrace{\partial_j}_{\text{the gradient}} \underbrace{\sum_i \partial_i v_i}_{\text{the divergence}} = \sum_i \partial_j \partial_i v_i = \sum_i \partial_i (\partial_j v_i)$$

which is now a divergence of an object with two indices (a tensor field, in fact). Using indices help enormously in keeping track where various operators hit: for example compare the above expression with the similar but different expression $$ (\triangle \vec{v})_j = \sum_i \partial_i \partial_i v_j $$

As an aside, (1) is basically the same as the integration by parts formula which you described.

share|cite|improve this answer

It can be confusing because the divergence and the gradient are two different things that happen to share the same symbol, $\nabla$. The notation is slightly idiosyncratic: the gradient of $u$ is denoted $\nabla u$, which is the same as your $Du$, while the divergence is written $\nabla\cdot u$. So you can always tell whether $\nabla$ means gradient or divergence depending on whether or not it is followed by a dot.

share|cite|improve this answer
"Even otherwise, you can't take the gradient of a vector field or the divergence of a scalar field" is one of those statements I'd be very careful about making: it may be a convenient mnemonic now, but one day the student may learn about the Navier-Stokes equations or about co-differentials in Riemannian geometry... – Willie Wong Jul 26 '12 at 20:11
@Willie, that's true, I was being sloppy. In fluid dynamics one does take the gradient of velocity, but does the divergence of a scalar field ever make sense? I'll update my answer once I get your response. – Rahul Jul 26 '12 at 20:35
Divergence can be defined (relative to a specified index) for any tensor field on a Riemannian manifold; and it can be defined for any multi-vectors (think dual of differential forms) on a orientable smooth manifold. Naturally this extends to the scalar field as a rank-0 contravariant tensor field. Of course, necessarily also its divergence is 0 (the same as the exterior derivative of a top form). With regards to the divergence of the scalar field, it is less important than the thing about gradients of vector fields, and I was just picking bones (because it is there to pick). – Willie Wong Jul 26 '12 at 21:32
Also, you may want to add a few words about why after integrating by parts, the divergence can become a gradient... – Willie Wong Jul 26 '12 at 21:35
@Willie, I just deleted the parts that weren't universally correct. Perhaps you could post an answer elaborating your last comment (though isn't it the Laplacian that's being integrated here?). – Rahul Jul 27 '12 at 4:16

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.