For a vector field $\vec{F}(x,y,z) = \langle F_1(x,y,z), F_2(x,y,z), F_3(x,y,z) \rangle$ in $\mathbb{R}^3$, how can I use mixed second-order partial derivatives of each of the components to determine whether it is conservative? Which partial derivatives should I compare?

  • 2
    $\begingroup$ Since the domain is $\mathbb R^3$, you should check that $\mathrm{curl}(\textbf{F}) = \textbf{0}$. $\endgroup$ – dannum Apr 25 '17 at 17:46
  • $\begingroup$ The key word in your question is “mixed.” That should give you a strong clue as to which derivatives to examine. $\endgroup$ – amd Apr 25 '17 at 18:19

You should check that

$$ \frac{\partial F_1}{\partial y} = \frac{\partial F_2}{\partial x}, \\ \frac{\partial F_1}{\partial z} = \frac{\partial F_3}{\partial x}, \\ \frac{\partial F_2}{\partial z} = \frac{\partial F_3}{\partial y}. $$

An easy way to remember this is to call your variables $x^1,x^2,x^3$ (instead of $x,y,z$). Then you need to check that

$$ \frac{\partial F_i}{\partial x^j} = \frac{\partial F_j}{\partial x^i} $$

for all $i \neq j$. This works in all dimensions.

  • $\begingroup$ I'd like to know how you came to these comparisons. $\endgroup$ – Ksquared Apr 25 '17 at 21:05
  • $\begingroup$ @Ksquared: This is the basic theorem about conservative vector spaces in $\mathbb{R}^n$... If $F(x^1,\dots,x^n) = (F_1(x^1,\dots,x^n), \dots, F_n(x^1,\dots,x^n))$ is a smooth enough vector field and $\frac{\partial F_i}{\partial x^j} = \frac{\partial F_j}{\partial x^i}$ for all $i \neq j$ then $F$ is locally conservative (and globally conservative if it is defined on a simply connected domain). $\endgroup$ – levap Apr 25 '17 at 21:12
  • $\begingroup$ It is easy to see that this is a necessary condition because if $F = \nabla f$ then $\frac{\partial f}{\partial x^i x^j} = \frac{\partial F_j}{\partial x^i} = \frac{\partial f}{\partial x^j x^i} = \frac{\partial F_i}{\partial x^j}$. $\endgroup$ – levap Apr 25 '17 at 21:14

Assume the derivatives of the components of $\vec F$ are continuous functions. If $\vec F = \left<P(x,y,z),Q(x,y,z),R(x,y,z)\right>$ were the gradient of a function $f(x,y,z)$, then $$ \frac{\partial f}{\partial x} = P \text{,}\qquad \frac{\partial f}{\partial y} = Q \text{,}\qquad\text{and}\qquad \frac{\partial f}{\partial z} = R $$ Taking the derivative of both sides of the first equation with respect to $y$, and the second equation with respect to $x$, we get $$ \frac{\partial^2 f}{\partial y\,\partial x} = \frac{\partial P}{\partial y} \qquad\text{and}\qquad \frac{\partial^2 f}{\partial x\,\partial y} = \frac{\partial Q}{\partial x} $$ By Clairaut's theorem the left-hand sides of the two equations are equal. Therefore $$ \frac{\partial P}{\partial y} = \frac{\partial Q}{\partial x} $$ Similarly, it can be shown that $$ \frac{\partial P}{\partial z} = \frac{\partial R}{\partial x} \qquad\text{and}\qquad \frac{\partial Q}{\partial z} = \frac{\partial R}{\partial y} $$ These are all necessary conditions for $\vec F$ to be conservative. They can be written in a form that doesn't assume only three variables, or, in the case of three variables, in a vector form. If these equations do not hold, $\vec F$ cannot be conservative.

The big question is when these conditions are sufficient as well. The answer is that as long as the domain of $\vec F$ is connected and simply connected, they are sufficient. This follows from Stokes's Theorem.


Hint: show that $$\nabla\times\vec F(x,y,z) = \vec 0$$

Then, it easily follows that $\vec F $ is conservative.


If $\nabla G = <G_x,G_y,G_z> \ = F = \ <F_1,F_2,F_3>$ for some $G(x,y,z)$, then $F$ is conservative.

$(F_1)_y = (G_x)_y = G_{xy} = G_{yx} = (G_y)_x = (F_2)_x$.

So, $\displaystyle \frac{\partial F_1}{\partial y} = \frac{\partial F_2}{\partial x}$ must hold.

The other two equations are similar, and given in other answers.


The criterion is $$ \DeclareMathOperator{curl}{curl} \curl F = \epsilon_{ijk} e_i \partial_j F_k = 0 $$ where we sum over same indices, and $\epsilon_{ijk}$ is the sign of the permutation $(ijk)$ or zero if index values repeat.


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.