Skip to main content

Dynamical Systems: Course Works

This is a place where I store my notes and/or exercise attempts when studying MA4233 Dynamical Systems during the second semester of Academic Year 2024/2025, instructed by Dr. Daren Wei.

The main reference book used is A First Course in Dynamics, with a Panorama of Recent Developments by Boris Hasselblatt.

Chapter 1: Introduction

The famous Logistic equation is

f(x)=kx(1x),x[0,1].f(x)=kx(1-x), \quad x \in [0,1].
  1. Verify that if k1k \leqslant 1, fn(x)0f^n(x) \to 0 as n+n \to +\infty for all x[0,1]x \in [0,1].
  2. Starting with x=0.01x = 0.01, iterate 2x(1x)2x(1-x) until you get a clear pattern.
  3. Starting with x=0.01x = 0.01, iterate 3x(1x)3x(1-x) until you get a clear pattern.
  4. Starting with x=0.01x = 0.01, iterate 3.45x(1x)3.45x(1-x) and 3.5x(1x)3.5x(1-x) until you get a clear pattern.
  5. Starting with x=0.5x = 0.5, iterate 3.83x(1x)3.83x(1-x) until you get a clear pattern.

Please find the Sage implementations of the workings of parts 2. to 5. of this question here.

Let f(x,y)=(x+y2,2xyx+y)f(x,y) = \left(\frac{x+y}{2},\frac{2xy}{x+y}\right), check that if you start with (1,4)(1,4), the iterations will converge to (2,2)(2,2).

note

The context of this exercise is Heron's method of obtaining the square root of an integer zz that is not a perfect square, which is by taking a rectangle with sides xx and yy such that xy=zxy=z, and xx is an educated guess of the square root. Iteratively, one then replaces the sides xx and yy by (x+y)/2(x+y)/2 (the arithmetic mean) and 2xy/(x+y)2xy/(x+y) (harmonic mean, so that their product yields the same number zz) respectively. As the number of iteration increases, one then sees that the two sides tend to the same value, which gives a better approximation of z\sqrt{z}.

A Sage computation of this procedure can be found here.

If xn=n2p/qx_n=n^2p/q, where p,qNp,q \in \mathbb{N} and p,qp,q are coprime, prove that the last digits of xnx_n is periodic.

(Here, the 'last digit' refers to the digit before the decimal point, where applicable.)

A Sage computation that illustrates this assertion can be found here.

A direct argument allows us to see that the period is in fact at most 10q10q, because

xn+10qxn=(n+10q)2p/qn2p/q=10(2np+10pq)x_{n+10q}-x_n=(n+10q)^2p/q-n^2p/q=10(2np+10pq)

is an integer multiple of 1010, so the last digit before the decimal point of xn+10qx_{n+10q} and xnx_n are equal.