2
$\begingroup$

A physicist colleague asks me the following question. I have no idea to answer him. Your answer is very appreciated.

Is there a global first integral on $\mathbb{R}^3$ for the following vector field?

$$\begin{cases}x'=sin(y) \\ y'=sin(z)\\ z'=sin(x) \end{cases}$$

$\endgroup$
5
  • $\begingroup$ Is this to be read as $x$, $y$, $z$ depending on one parameter, say, "$t$" and the prime denotes the derivative w.r.t. $t$? And what is sought is the solution of this set of first-order differential equations? $\endgroup$ Sep 7, 2019 at 23:24
  • 1
    $\begingroup$ @MichaelEngelhardt yes $x'=dx/dt$. $\endgroup$ Sep 7, 2019 at 23:41
  • 1
    $\begingroup$ @MichaelEngelhardt by the second part of your comment do you mean "what is the motivation for consideration of this system"? I have no idea of such a motovation. I can ask him. $\endgroup$ Sep 7, 2019 at 23:46
  • 1
    $\begingroup$ No no, the second part of my comment was simply to make sure I understood what form of answer was expected - I suppose that was a bit redundant once the first part was clarified. The second part was simply reinforcing the first part. Still, it would be interesting where this arises - but don't go to any length investigating, it's not essential of course. $\endgroup$ Sep 8, 2019 at 1:42
  • 2
    $\begingroup$ Now, for small $x$, $y$, $z$, the solutions of course are of the form $x=\exp (t)$, $\exp (\exp (i 2\pi /3) t)$ or $\exp (\exp (i 4\pi /3) t)$, with $y$, $z$ obtained by differentiation, but this is probably obvious to your colleague, since he explicitly asks for a global answer ... $\endgroup$ Sep 8, 2019 at 1:53

1 Answer 1

3
$\begingroup$

One can find a solution of the form $x=y=z$, namely, $x=2$arccot$(\exp (-t-a))$ with the free parameter $a$. Of course, there should be more. Note also the symmetries of the problem: For any solution $(x,y,z)$, also $(-x,-y,-z)$ is a solution, as is $(x+2k\pi,y+2l\pi,z+2m\pi)$, with arbitrary integers $k,l,m$.

$\endgroup$
5
  • 1
    $\begingroup$ The latter symmetries presumably provide useful guardrails if one ultimately wants to resort to a numerical solution - one only has to interpolate between two adjacent analytic solutions bounding the direction of the vector $(x,y,z)$. $\endgroup$ Sep 8, 2019 at 14:11
  • 1
    $\begingroup$ In particular, for large $x$, $y$, $z$, one is always very close to one of the analytic solutions. In combination with the form of solutions for small $x$, $y$, $z$ in comments above, there isn't that much wiggle room left. $\endgroup$ Sep 8, 2019 at 14:20
  • $\begingroup$ Thank you very much for your answer. I will forward your answer to him. by global first integral I mean a function $f:\mathbb{R}63 \to \mathbb{R}$ with $f_x sin(y)+f_y sin(z)+f_z sin(x0=0$. may be your answer can help to find this globaly defined first integral. Thanks again for consideration of the question. $\endgroup$ Sep 9, 2019 at 21:17
  • 1
    $\begingroup$ Oh - looks like we talked past each other then ... this is what I was trying to clarify with my initial question, but it seems we still didn't quite match minds. Still, maybe this helps anyway. I'll update if I think of anything else. $\endgroup$ Sep 9, 2019 at 23:47
  • 1
    $\begingroup$ Indeed, the trajectories $(x(t),y(t),z(t))$ I describe are by construction equipotential lines, i.e., lines of constant $f$, to the extent that $f$ exists. So they at least provide a scaffolding for a more complete construction of $f$. $\endgroup$ Sep 10, 2019 at 19:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.