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 was asked the following vector calculus problem:

Let $D$ be the unit ball and let $S$ be the unit sphere in $\mathbb{R}^3$. Suppose that $F:\mathbb{R}^3\rightarrow \mathbb{R}^3$ is a $C^1$ vector field on some open neighborhood of $D$ which satisfies:

$(i) \nabla\times F=0$

$(ii) \nabla\cdot F=0$

$(iii)$ On $S$, $F$ is orthogonal to the radial vector.

Prove that $F=0$ on all of $D$.

Conditions $(i)$ and $(ii)$ imply that $F=\nabla g$ for some $g:\mathbb{R}^3\rightarrow \mathbb{R}$ where $g$ must be harmonic as well.

I know one solution (see end), however my initial instinct was to try to use the max/min property of harmonic functions, and I couldn't get it to work. Since the gradient is always orthogonal to the sphere, there must be a point on the sphere where it is $0$. (Hairy ball) If that was a local max or min in $\mathbb{R}^3$ we would be done, by taking a small neighborhood around it. If it is a saddle point this doesn't work. (We know that it must be a local max/min on $S$ since it is harmonic)

My question is: Is there any way to modify this approach, and solve the problem?


Other Solution: Here is one solution that first uses the fact that the radial vector is orthogonal, and then applies Gauss's Divergence theorem to the function $gF$. ($\nabla g=F$) That is $$0=\iint_S (gF\cdot n)dS=\iiint_D \nabla\cdot (gF)dV=\iiint_D \|F\|^2dV,$$ and since the integrand on the right hand side is non-negative, continuous and integrates to give zero, it must be zero.

share|cite|improve this question
The problem you have is not with saddle points. By assumption $F|_S = \nabla (g|_S)$, and $g$ being a continuous function on $S$ must attain a max/min. – Willie Wong Apr 27 '11 at 17:49
@Willie Wong: I edited to make it clearer. I meant meant local max/min in $\mathbb{R}^3$ as opposed to in $S$. We have a critical point, but maybe it is some sort of saddle. – Eric Naslund Apr 27 '11 at 18:04
ah, then you should use the Strong Maximum Principle / Hopf's Lemma as Hans suggests. – Willie Wong Apr 27 '11 at 18:41
up vote 3 down vote accepted

Your second proof is a good one and entirely appropriate within vector calculus. The first attempt has a gap (I think).

A third proof relies on Hopf's Lemma (commonly taught in graduate level classes on partial differential equations) which implies here that if a function $u$ satisfying $\Delta u \le 0$ in $D$ attains a strict minimum at $z \in \partial D = S$, then the outer normal derivative at that point satisfies $\nu \cdot \nabla u(z) < 0$. Applying this with $g = u$, it follows that $g$ must attain its minimum in the interior of $D$ and hence must be constant.

share|cite|improve this answer
This is very interesting, thank you. Quick question: what is $\nu$? I assume its the vector going from $0$ to $z$. – Eric Naslund Apr 27 '11 at 18:13
$\nu$ is the unit outward normal vector (to $S$). So yes, the vector going from $0$ to $z$. – Willie Wong Apr 27 '11 at 18:42
A bit late, but I forgot to accept your answer! Thanks! – Eric Naslund Jun 7 '11 at 16:34

Eric, I am simply rephrasing your second proof to give you a more PDE style approach to this problem, let $\vec{F} = \nabla u$, the problem you gave was equivalent to solve the following boundary value problem: Find $u\in C^2(D)$ such that $$ \left\{\begin{eqnarray} -\Delta u &=& 0 \text{ in } D \\ \nabla u \cdot n &=& 0 \text{ on } \partial D \end{eqnarray} \right.$$ Now multiply both sides of the equation by a test function in some proper space(as large as possible), doing integration by parts to use the boundary condtion, and you will get the variational formulation of this problem: Find $u\in V = W^{1,2}(D)$ such that $$ \int_D \nabla u\cdot \nabla v = 0 \quad \forall v\in V $$ Notice there is no Dirichlet type boundary condition enforced for the test function $v$, which makes this problem yield non-unique solutions; In order to make this problem well-posed, instead in the space $V$, consider a new space $V_0 = \{v\in V | \int_D v dx = 0\} = V/R$ this variational problem is equivalent to the following functional minimization problem: If $u$ solves the original pde, then $u$ minimizes $$ J(u) = \min_{v\in V_0}\int_D |\nabla v|^2\,dx $$ Now assume $u$ is the solution, then $$ J(u) = \int_D |\nabla u|^2\,dx \leq \sup_{v\in V_0} \int_D \nabla u\cdot \nabla v = 0 $$ which proves for the solution $u\in V_0$: $\nabla u = 0$(other solutions are different with some specific $u$ from a constant by the definition of that quotient space, hence the gradient does not change)

share|cite|improve this answer

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.