Conservative vector fields

The previous post discussed the fact that the curl of a divergence is zero. The converse is also true (given one more requirement): a vector field F is the gradient of some potential φ function if ∇×F = 0. In that case we say F is a conservative vector field.

It’s clear that having zero curl is a necessary condition for a vector field to be the gradient of a potential. It’s not as clear that having zero curl is sufficient. And in fact, it’s not quite sufficient: it’s sufficient over a simply connected domain. This means a domain with no holes: any loop you draw in the domain can be continuously deformed to a point without leaving the domain. If you’re working with vector fields defined everywhere on ℝ³ then you don’t have to worry about this condition because ℝ³ is simply connected.

Aside on topology

For a calculus student, the requirement that a domain be simply connected is a footnote. For a topologist, it’s the main thing.

If a domain is not simply connected, then a vector field might have curl zero, but not be the gradient of a potential. So in general we have two kinds of vector fields with zero curl: those that are gradients and those that are not. We could look at the space of all vector fields that have zero curl and mod out by the vector fields that are gradients. The dimension of this space tells us something about how the domain is connected. This is the start of de Rham cohomology.

Calculating the potential function

If a vector field is conservative, i.e. if it is the gradient of a potential function φ, then you can find φ by integration.

The partial derivative of φ with respect to x is the first component of your vector field, so you can integrate to find φ as a function of x (and y and z). This integral will only be unique up to a constant, and functions of y and z alone are constants as far as partial derivatives with respect to x are concerned.

Now the partial derivative of φ with respect to y has to equal the second component of your vector field. So taking the derivative of what you found above determines your potential, up to a function of z. So then you differentiate again, this time with respect to z, and set this equal to the third component of your vector field, you’ve determined your potential function up to an constant. And that’s as far as it can be determined: any constant term goes away when you take the gradient.

Exact differential equations

A differential equation of the form

M(x, y) + N(x, y) y‘(x) = 0

is said to be exact if the partial derivative of M with respect to y equals the partial of N with respect to x. In that case you can find a function φ such that the partial of φ with respect to x is M and the partial of φ with respect to y is N (assuming you’re working in a simply connected domain). This function φ is a potential, though differential equation texts don’t call it that, and you find it just as you found φ above. The solution to your differential equation is given (implicitly) by

φ(x, y) = c

where c is a constant to be determined by initial conditions.

Poincaré’s lemma

For a vector field over a simply connected domain, having zero curl is necessary and sufficient for the existence of a potential function φ. This is a case of Poincaré’s lemma. The next post will look at another case of Poincaré’s lemma, finding a vector potential.