0
$\begingroup$

Consider $R: \mathbb{R}^2 \rightarrow \mathbb{R}^2$ given by $R(a,b) = (b,2.5b-a).$ Let $p_0 = (x_0, y_0)$ be arbitrary and $p_{i+1} = R(p_i).$ Most starting points $p_0$ give a divergent path. One can show that the set of $p_0$ which give a convergent path are $(2t, t),$ and these all converge to $(0,0).$

Next consider $S(a,b) = (b, 2.5b-a-a^2).$ Let $p_0 = (x_0, y_0)$ be arbitrary and $p_{i+1} = R(p_i).$ Let $T$ be the set of points $p_0$ in $\mathbb{R}^2$ or maybe $\mathbb{C}^2$ which give converge to $(0,0).$ (Note that $(0.5, 0.5)$ is another fixed point, but not what I'm interested in. oops, edited)

Edit: The comment about possibly being a fractal seems correct. Instead, let $U$ be the set of points $p_0$ such that the path converges to $(0,0)$ and also never goes outside the set $[-0.1,0.1] \times [-0.1,0.1].$

Intuitively, $U$ should be a curve or part of a curve which is locally given by $(2t, t)$ near $(0,0).$ Furthermore, one can assume that $U$ is locally given by $(t, P(t))$ where $P$ is a Taylor series, and then one can actually solve for the coefficients recursively, although with no obvious closed form. My question is what is $U$? Could it be part of an algebraic curve or solution set of a reasonable equation?

Note: I am actually more interested in the following more complicated recursion: $T(a,b,c,d) = (c,d,e,f),$ where $c = (fa)/(f+a+e), d = (ad+b-bc-af)/(ad+b+1+cf-c-bc-af-f)$ and the fixed point $(1-\sqrt{1/2}, 2-\sqrt{2}, 1-\sqrt{1/2}, 2-\sqrt{2}).$ I do not even know the asymptotic form near this fixed point.

$\endgroup$
3
  • $\begingroup$ Julia sets might suggest that your set can be far from the set of zeros of a reasonable equation. Since convergence is fast, you could at least draw a picture of your set: maybe it will clearly show off fractal behavior. $\endgroup$
    – YCor
    Commented Jan 10, 2017 at 22:58
  • $\begingroup$ I didn't check the equations in your question, but the curve you're describing is usually called a "local stable manifold". Assuming the fixed point at the origin is hyperbolic (the Jacobian matrix has one eigenvalue inside the unit circle and the other outside) then it will indeed be a curve; this is the Hadamard-Perron theorem. With those keywords you should be able to find some literature on how to compute it. $\endgroup$ Commented Jan 11, 2017 at 3:46
  • $\begingroup$ That definitely describes the recurrence S, thank you. Unfortunately, T has four eigenvalues of 1 and is likely to be much harder $\endgroup$
    – Bobby Shen
    Commented Jan 11, 2017 at 7:05

0

You must log in to answer this question.

Browse other questions tagged .