An error estimate for fixed point computation of b^x
#1
Ok, we want to compute the lower fixed point \( a \) of \( b^x \) for, as usual, \( 1<b<e^{1/e} \).

We do that by iterating:
\( x_{n+1} = b^{x_n} \)
\( \lim_{n\to\infty} x_n = a \)

while iterating we can not directly compute the difference \( a-x_n \) but we can compute \( \mu_{n+1} = x_{n+1}/x_n \) and we ask our self how close \( \mu_n \) must come to 1 such that

\( a-x_n<\epsilon \)

First we translate the difference into a quotient:
\(
\begin{align*}\frac{b^a}{b^{x_n}}&<b^\epsilon\\
y_{n+1}:=\frac{a}{x_{n+1}}&<b^\epsilon
\end{align*} \)

\( y_n>1 \) is decreasing. Now lets compute
\( \mu_{n+1} = \frac{x_{n+1}}{x_n}=\frac{b^{x_n}}{x_n}=\frac{b^{a/y_n}}{a/y_n}
=y_n a^{1/y_n-1} = y_n (y_n x_n)^{1/y_n-1} \)

To make an estimate we want to know whether \( f(x)=x(xc)^{1/x-1}=c^{-1}(xc)^{1/x} \), \( c<a \) is decreasing for \( x>1 \). To decide this we differentiate:
\( f'(x)=c^{-1}\left( xc \right) ^{1/x} {\frac {1-\ln \left( xc
\right) }{x^2}} \)

\( f'(x)>0 \) for
\( 0<1-\ln(xc) \)
\( xc<e \)

This is true for \( \epsilon<\log_b(e/c) \) because then \( x<b^\epsilon<\frac{e}{c} \).

So we know that \( f \) is strictly increasing for small enough \( \epsilon \). So we get
\( y_{n+1}<b^\epsilon \) iff \( \mu_{n+2}<b^\epsilon (b^\epsilon x_{n+1})^{b^{-\epsilon}-1} \)
for \( \epsilon<\log_b(e/x_{n+1}) \), equivalently:

\( \frac{a}{x_{n+1}}<b^\epsilon \) iff \( \frac{x_{n+2}}{x_{n+1}}<b^\epsilon (b^\epsilon x_{n+1})^{b^{-\epsilon}-1} \)
\( a - x_n <\epsilon \) iff \( x_{n+2}<(b^\epsilon x_{n+1})^{b^{-\epsilon}} \)
The right side can be further simplified:
\( b^{x_{n+1}} < b^{(\epsilon +x_n)b^{-\epsilon}} \)
\( x_{n+1}<(\epsilon + x_n)b^{-\epsilon} \)

The condition \( \epsilon<\log_b(e/x_{n+1}) \) is satisfied for \( \epsilon < \log_b(e/a) \), i.e. on the right side is a constant \( >0 \). So during the iteration we can decrease \( \epsilon \) according to \( x_n \) without fear that \( \epsilon \) becomes to small and the iteration does not stop.

Proposition. Let \( 1<b<e^{1/e} \), let \( a \) be the lower fixed point of \( b^x \), let \( x_0<a \) and \( x_{n+1}=b^{x_n} \) then for each \( 0<\epsilon < \log_b(e/x_n) \):
\( a - x_n <\epsilon \) iff \( x_{n+1}<(\epsilon + x_n)b^{-\epsilon} \).
Reply


Possibly Related Threads…
Thread Author Replies Views Last Post
  Writing Kneser's super logarithm using values of Kneser at a single point JmsNxn 1 1,333 04/21/2023, 04:26 AM
Last Post: JmsNxn
  fixed point formula sheldonison 6 26,083 05/23/2015, 04:32 AM
Last Post: mike3
  Attempt to find a limit point but each step needs doubling the precision... Gottfried 15 45,997 11/09/2014, 10:25 PM
Last Post: tommy1729
  Find all fixed points of exp[b] MorgothV8 10 36,427 10/07/2014, 11:00 AM
Last Post: Gottfried
  A note on computation of the slog Gottfried 6 21,359 07/12/2010, 10:24 AM
Last Post: Gottfried
  Single-exp series computation code mike3 0 5,888 04/20/2010, 08:59 PM
Last Post: mike3
  sexp(strip) is winding around the fixed points Kouznetsov 8 26,068 06/29/2009, 10:05 AM
Last Post: bo198214



Users browsing this thread: 1 Guest(s)