I have a smooth compact oriented manifold without boundary, $M$, with the volume form $\Omega$ and a Riemannian metric $g$.

Given a function $\phi$ is there any canonical way to obtain a divergence-free smooth vector field $V$ which is orthogonal to $\nabla _{g} \, \phi $?

Optimally, I would like $V$ to be globally defined and orthogonal to $\nabla _g \, \phi $, but eventually even locally could do it,

thanks in advance,


  • $\begingroup$ thank you for the answer. the emphasis was supposed to be on "canonical", though: is there a way of using the (rich) structure in order to obtain the vector field? $\endgroup$
    – jesus
    Nov 16 '16 at 14:56
  • $\begingroup$ Are you modelling yourself on the situation of $\mathbb{R}^2$ with the standard metric where given any function $\phi$, the vector field $(\partial_2 \phi, - \partial_1\phi)$ is orthogonal to $\phi$ and divergence free? This generalizes to any two dimensional Riemannian manifold if you take $V = *\mathrm{d}\phi$ where $*$ is the Hodge star and we freely raise/lower indices with $g$. In higher dimensions the analogue are not vector fields. $\endgroup$ Nov 16 '16 at 15:35
  • $\begingroup$ Alternatively, a higher dimensional structure that better captures what you want in this case is that of a symplectic structure. If you are given a even-dimensional manifold $M^{2n}$ with a symplectic form $\omega$ and associated volume form $\omega^n$, then a direct computation shows that the Hamiltonian vector field of $\phi$ is tangent to the level sets of $\phi$ (so "orthogonal to the gradient") and is divergence free. $\endgroup$ Nov 16 '16 at 15:50

Your problem is perhaps better stated as asking, given a manifold $M$ with a volume form $\Omega$, and a smooth function $\phi$, if there is a vector field $X$ whose flow preserves $\Omega$ and for which $\phi$ is constant along the flow lines of $X$. This problem has no Riemannian metric. Locally, near a point where $d\phi \ne 0$, we can find coordinates in which $\Omega$ is the Euclidean volume form, and in which $\phi$ is the first coordinate function. Why? Use the Moser homotopy lemma to arrange coordinates $x^1, \dots, x^n$ in which $\Omega$ is the Euclidean volume form. Then find a function $f$ so that $\partial f/\partial x^2 \partial \phi/\partial x^1-\partial f/\partial x^1 \partial \phi/\partial x^2=1$, locally, and then replace $x^1$ with $\phi$ and $x^2$ with $f$. Now your vector field can be any divergence free vector field in $x^2,\dots,x^n$, and it gives a vector field tangent to level sets of $\phi$, since it doesn't involve $x^1$.


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.