The equation $f^{\circ k}(x) = \mathrm{Id}$ for $x\in E$ is called the Babbage equation and the general solution is given in the following way [M. Kuczma, Functional equations in a single variable]:
Let $1 = n_0 < \cdots < n_r = k$ be the complete set of divisors of $k$ and let $$ E = \cup_{i=0}^{r}\cup_{j=1}^{n_i}U^i_{j} $$ be an arbitrary decomposition of $E$ into disjoint sets such that for every $i$ the sets $U^i_{1},\ldots,U^i_{i_n}$ have the same cardinality. For $1\leq i\leq r$ and $1\leq j\leq n_i-1$, let $f_{ij}$ be an arbitrary one-to-one map of $U^i_j$ onto $U^i_{j+1}$. Then the general solution is given by:

  • $f(x) = x$ for $x\in U^0_1$;
  • $f(x) = f_{ij}(x)$ for $x\in U^i_j$, $j=1,\ldots,n_i-1$, $i\ge 1$;
  • $f(x) = f_{i1}^{-1}(\cdots(f_{i,n_i-1}^{-1}(x))\cdots)$ for $x\in U_{n_i}^i$, $i\geq 1$.

This solution is also valid when $f$ is a function of more that one variable.

In dimension 1, the solution $f$ is continuous if and only if $f$ is monotone, and in this case either $f(x)=x$ or $f$ is a decreasing function such that $f^{\circ 2} = \mathrm{Id}$.

Question: In dimension $n\ge 2$, assuming $f$ is continuous, what are the general solutions of the Babbage equation?

We can of course give some special solutions, such as when $n=2$:

  • $f(x,y)=(y,x)$ for even $k$.
  • $f(x,y)=(f_1(x),f_2(y))$ such that $(f_1)^{\circ k}(x)=x$ and $(f_2)^{\circ k}(y)=y$.
  • Of course, the question is what you expect $f$ to be. If you don't assume, say, continuity (regarding the domain as just a set), then certainly the answer below describes everything: the whole set splits into orbits, which should be of bounded length. – Alex Degtyarev Apr 3 '14 at 19:14
  • Can you tell us what's the "general solution" for a function of one variable? Just curious. – Alexandre Eremenko Apr 3 '14 at 20:59
  • I modified my question with a general solution. – Lucy Apr 4 '14 at 12:25
  • @AlexDegtyarev You are right, I assume the solutions to be continuous or differentiable. – Lucy Apr 4 '14 at 12:26

There are certainly many other solutions. Here is a recipe for constructing them. If $S$ is a set (say infinite to simplify the presentation). We split $S$ into a disjoint union of $k$-element subsets $S_\alpha$. On each such subset we consider a cyclic permutation. We then fit them together to a mapping on the whole space. This clearly satisfies your condition.

  • Yes. I considered permutations before. Maybe I should modify my question to the subclass of functions up to permutations rotations and reflections. – Lucy Apr 3 '14 at 18:12
  • Actually I think that all solutions arise essentially in this manner (I didn't include this fact in my response since I hadn't sat down to write it out). In this case, this kind of modification might not be very useful. – barcelos Apr 3 '14 at 18:20
  • I suspect that the article "On certain real solutions of Babbage's functional equation" by Ritt (Ann.Math. 79 (1972) 113-122) is being referred to. By the way, using his simple observation that a function which is conjugate to a solution is also a solution, it is easy to get a plethora of CONTINUOUS solutions on the plane. Just take a function which is conjugate to a Möbius function. Many of the latter are periodic in the above sense. – barcelos Apr 4 '14 at 8:59
  • Thank you, @barcelos. I read the paper. I am trying to characterize the "complete" continuous solutions and hope to find interesting results as in dimension 1, where continuity forces all solutions to be either trivial($f(x)=x$) or a involutory, i.e. $f^{\circ 2}(x)=x$. Moreover, a involutory on a real interval is nothing but a decreasing monotone function with some fixed points. – Lucy Apr 4 '14 at 15:07

Your Answer

 
discard

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

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