Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Since $\cos (2x) = 2 \cos^2 (x) - 1$, I wonder about the iteration $(x,y) \mapsto (2x\mod 2\pi, 2y^2-1)$. Will it converge to the cosine graph? I tried it in mathematica and got quite a few points, if I cheat and give it an initial value.

x = {1, 0.540302306};
step[{a_, b_}] := {Mod[2 a, 2 Pi], 2 b^2 - 1}
Table[(x = step[x]; x), {k, 1, 30}];
ListPlot[%, AspectRatio -> Automatic]

Here's a variation based on the sum of angles formula gives nice convergenc to the graph.

x = {1, 0.540302306};
step[{a_, b_}] := (c = Mod[a + 1, 2 Pi]; {c , 
b Cos[1] - If[c < Pi, 1, -1] Sin[1] Sqrt[1 - b^2]})
Table[(x = step[x]; x), {k, 1, 500}];
ListPlot[%, AspectRatio -> Automatic]

These two algorithms have difference tolerances for error. In the first algorithm, if I start with a wrong initial value like {1, 0.4} I get a chaotic "cloud", but in the second case, it still converges to the cosine graph.

I guess I would like to know, does my original iteration have any other fixed points besides the cosine graph?

share|cite|improve this question
up vote 4 down vote accepted

Because the $x$ and $y$ coordinates do not interact in your iteration, it does not make sense to expect any relation between them in the limit.

For example, the line $\{y=1\}$ is also invariant under your map.

You could ask the following: What are the real-valued continuous functions $h$, defined on the circle, such that the graph of $h$ is invariant under the map you define?

Then you are asking about semiconjugacies between angle doubling and the Chebyshev polynomial $T(x) = 2x^2-1$. That is, your function $h$ must satisfy $h(2x) = T(h(x))$.

Note that the function $h(x) = T^{\circ n}(\cos(x))$, where $T^{\circ n}$ denotes the $n$-th iterate, provides additional examples, since $$ h(2x) = T^{\circ n}(\cos(2x)) = T^{\circ (n+1)}(\cos(x))=T(h(x)).$$

It seems possible that these (together with the constant map $h(x)=1$) are the only examples, but I have not checked this.

In any case, you cannot expect convergence to these curves unless you start exactly on them.

share|cite|improve this answer

I gave the first over 400 iterations to settle down starting from (1, 0.4) and still got a chaotic cloud. Even starting with (1,cos(1)), Excel had a chaotic cloud after 400. I don't think this is surprising as you have two separate iterations that don't talk to each other, and the first coordinate one is about the most unstable iteration known.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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