Fixed points and infinite series

445 Views Asked by At

Consider the formula $1 + \frac{y}2$. This has a fixed point at $y = 2$. And if we use the equation $y = 1 + \frac{y}2$ to substitute for $y$ in our formula, we get $1 + \frac{1 + \frac{y}2}2$, or $1 + \frac12 + \frac{y}4$. If we keep repeating we get $1+\frac12 + \frac14 + \dots$, which does indeed equal our fixed point $2$.

But now consider the formula $1+2y$. This has a fixed point at $y=-1$. Here if we use the equation $y = 1+2y$ to substitute for $y$ in our formula, we get $1+2(1+2y)$, or $1+2+4y$. If we keep repeating we get $1+2+4+\dots$, which does not equal our fixed point $-1$.

Can someone explain in simple terms why this procedure works in the first case but not the second? Would the procedure work in every case where the generated series converged?

1

There are 1 best solutions below

3
On BEST ANSWER

Aside from the argument that $1+2+4+\cdots$ could equal $-1$, to answer your question, there are two classes of fixed points.

An attractive fixed point is a fixed point where within a certain range, a point fed into the formula recursively will eventually converge to the fixed point. Your first example is an attractive fixed point.

A fixed point that is not attractive has no points converging to the fixed point when fed into the formula recursively. Your second example is a non-attractive fixed point.

The property of attractive fixed points is that the derivative (i.e., slope) of the function in the formula has magnitude less than $1$ on the range of points that converges to the fixed point:

$$|f'(x)|<1$$

where $x_{n+1}=f(x_n)$ is the recursive formula.

See this (fixed point) and this (fixed-point iteration) for more information and a clearer explanation.