Feigenbaum Constants
Let
, where
is fixed.
The 1-cycles (i.e., fixed points) of f are intersections between the
red curve and the green line. Solving x = f(x), we obtain
x=0 (which attracts for a<a0=1)
and x=(a-1)/a (which attracts for
a<a1=3).
The 2-cycles of f are intersections between the blue curve (graph
of the iterate
) and the green line that are not already accounted for
above:
Solving, we obtain
(which attracts for a<a2=
). For
, an
attracting 4-cycle emerges:
The sequence of period-doubling bifurcations converges at a point
considerably less than 4. Letting
and
so forth, it can be proved that
This
point marks the division between the "periodic regime" and the "chaotic
regime" for this family of quadratic functions. Two "universal" Feigenbaum
constants are associated with the above process:
and
where
is the width
of the unique "fork" in the
which crosses
the line
.
If we replace the logistic maps f by, for example,
, then
interestingly the same constants
and
occur. Both
functions f and g are unimodal and have quadratic maximum points. See the
references for rigorous proofs of universality.
The Mathcad PLUS 6.0 file feigen.mcd
discusses a simple method for computing
and
due to Keith
Briggs. A method developed later by Briggs is faster and more precise. (Click here if you have 6.0
and don't know how to view web-based Mathcad files). Plouffe gave accurate
approximations of both Feigenbaum constants in the Inverse Symbolic
Calculator web pages.
Here also is a recent communication from David Broadhurst containing
the Feigenbaum constants to apparently even higher
precision than Briggs' estimates.
In a January, 2000 math-fun posting regarding the (unitless) inverse fine
structure constant from physics, Gene Salamin wrote:
The NIST value is 137.03599976(50), where the standard deviation
uncertainty in parentheses applies to the last decimal places. Our state of
knowledge of this physical constant is given by a Gaussian with this mean and
standard deviation...
I will conjecture that in a better theory than we have today, where the fine
structure constant can be calculated, it will be related to a Feigenbaum style
constant.
Salamin elaborated on his conjecture in a subsequent
e-message.
More
details and references (contact Steven Finch).
Return to
the MathSoft Constants page.
Return to the MathSoft
Unsolved Problems page.
Return to the MathSoft home
page.
Copyright © 1995-2001 Steven Finch
MathSoft Engineering & Education, Inc.
All rights
reserved.