See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271771394

Route to Chaos in Generalized Logistic Map

Article in Acta Physica Polonica Series a · February 2015

DOI: 10.12693/APhysPolA.127.A-113 · Source: arXiv

CITATIONS READS

4 428

2 authors:

Rafał Rak Ewa Rak

UNIVERSITY OF RZESZÓW Rzeszów University
25 PUBLICATIONS 440 CITATIONS 31 PUBLICATIONS 178 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Intuitionistic fuzzy ideals and filters on ordered structures View project

Distributivity and modularity of aggregation functions View project

All content following this page was uploaded by Ewa Rak on 13 April 2015.

The user has requested enhancement of the downloaded file.

 Route to chaos in generalized logistic map

R. Rak1 , E. Rak1
arXiv:1502.00248v1 [nlin.CD] 1 Feb 2015

1
Faculty of Mathematics and Natural Sciences, University of Rzeszów,
PL–35-959 Rzeszów, Poland

Motivated by a possibility to optimize modelling of the population evo-

lution we postulate a generalization of the well-know logistic map. Gener-
alized difference equation reads:

xn+1 = rxpn (1 − xqn ), (1)

x ∈ [0, 1], (p, q) > 0, n = 0, 1, 2, ..., where the two new parameters p and
q may assume any positive values. The standard logistic map thus corre-
sponds to the case p = q = 1. For such a generalized equation we illustrate
the character of the transition from regularity to chaos as a function of r
for the whole spectrum of p and q parameters. As an example we consider
the case for p = 1 and q = 2 both in the periodic and chaotic regime. We
focus on the character of the corresponding bifurcation sequence and on
the quantitative nature of the resulting attractor as well as its universal
attribute (Feigenbaum constant).

1. Introduction
Since the dawn time the chaos is an indispensable part of human life.
One of the most famous models with chaotic behavior is the logistic map.
Its simple analytic form meant that it was used in many scientific disciplines
such as biology, cryptography, communication chemical physics and stock
market [1, 2, 3, 4]. One of the mathematical models of chaos was discovered
by Mitchell Feigenbaum (1978) [5, 6, 7, 8]. Feigenbaum considered ordinary
difference equations used for example in biology to describe the development
of population in its dependence on time. He discovered that population
oscillates in time between stable powers (fixed points), the number of which
doubles according to changes in the power of external parameter. The
generalization of the Feigenbaum model contains all first-order difference
equations f (xn ) = xn+1 . The condition for the existence of chaos is the
single maximum of a function f (xn ). Feigenbaum also proved that the
transition to chaos is described by two universal constans α and δ, later

(1)
2 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015

named the Feigenbaum numbers.

We quote here a few definitions which are basic in the development of this
exposition.
Definition 1. [cf. [9], p. 62]
Let I = [a, b] ⊂ ℜ, a < b, f (I) ⊂ I. Discrete iterations of a function
f : I → I are called functions f n : I → I defined inductively: f 0 = id,
f n+1 = f ◦ f n , n = 1, 2, 3, ....
Definition 2. [[9], p. 62]
The sequence {xn }, where xn = f n (x0 ) is called the sequence of iterates of
a function f generated by point x0 ∈ I.
Definition 3. [cf. [9], p. 65]
A point s ∈ I, with f (s) = s is called a first-order fixed point of f . A fixed
point s of a function f is locally stable if there exists a neighbourhood Us
such that all sequences of iterates are convergent to s for x0 ∈ Us .
Theorem1. [see [9], p. 68]
′
If function f is differentiable in fixed point s and |f (s)| < 1, then s is
attractive.
Definition 4. [cf. [9], p. 72]
Let f : I → I. We say that s is a periodic point of the function f , or that
s generates a cycle of order k ≥ 2, if f k (s) = s and ∀i<k f (i) (s) 6= s.
Theorem2. [10]
If function f has a cycle of order different to 2k , then f is chaotic.
Theorem3. [cf. [9], p. 79]
For arbitrary continuous function f : I → I existence cycles of order m
involve the existence cycles of order n in the following order:
3 >> 5 >> 7 >> 2 · 3 >> 2 · 5 >> 2 · 7... >> 2k · 3 >> 2k · 7 >> ... >>
2k >> ... >> 8 >> 4 >> 2 >> 1.

The conventional logistic map

The Logistic map [11, 12, 13],

f (xn ) = xn+1 = rxn (1 − xn ), x ∈ [0, 1], (p, q) > 0, n = 0, 1, 2, ... (2)

is one of the most simple forms of a chaotic process. Basically, this map, like
any one-dimensional map, is a rule for getting a number from a number.
The parameter r is fixed, but if one studies the map for different values
of r (up to 4, else the unit interval is no longer invariant) it is found that
r is the catalyst for chaos. After many iterations x reaches some values
independent of its starting value 3 regimes: r < 1 : x = 0 for large n,
1 < r < 3 : x = constant for large n, 3 < r < r∞ : cyclic behavior, where
r∞ ≈ 3.56, r∞ < r ≤ 4 : mostly chaotic,
An interesting thing happens if a value of r greater than 3 is chosen. The
 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015 3

map becomes unstable and we get a pitchfork bifurcation with two stable
orbits of period two corresponding to the two stable fixed points of the sec-
ond iteration of f . With r slightly bigger than 3.54, the population will
oscillate between 8 values, then 16, 32, etc. The lengths of the parame-
ter intervals which yield the same number of oscillations decrease rapidly;
the ratio between the lengths of two successive such bifurcation intervals
approaches the F eigenbaum constant δ = 4.669.... The period doubling
bifurcations come faster and faster ( 8, 16, 32, ...), then suddenly break off.
Beyond a certain point, known as the accumulation point r∞ , periodicity
gives way to chaos. In the middle of the complexity, a window suddenly
appears with a regular period like 3 or √ 7 as a result of mode locking. The
3-period bifurcation occurs at r = 1 + 2 2, and period doubling then begin
again with cycles of 6,12,... and 7,14,28,... and then once again break off to
chaos.

2. Generalized logistic map

One-dimensional representations are the simplest arrangements capa-
ble to producing a chaotic movement. Let us consider p, q > 0. We postulate
non-linear function:

fr (x) = rxp (1 − xq ), x ∈ [0, 1] (3)

set by difference equation:

xn+1 = rxpn (1 − xqn ), x ∈ [0, 1], n = 0, 1, 2, .... (4)

For the first time the function (3) has been presented in [14]. The max-
 1/q
p ′
imum of the function (3) is reached for x = p+q , because fr (x) =
 1/q
′ p
r(pxp−1 − (p + q)xp+q−1 ) and fr (x) = 0 ⇔ x = p+q , ℜ ∋ (p, q) > 0.
p
!
 1  p p+q q
p q ( ) (p+q)
= rqp p+q and for r ∈ 0, p q
q
It results that fr p+q where
(p+q) q
p
!
( p+q
p
) q (p+q)
0, q = rmax function fr is f ([0, 1]) ⊂ [0, 1].

Parameter rmax depending on p and q accepts the following limit values

(Fig.1):
lim rmax = 1, lim rmax = ∞
q→∞ q→0+

lim rmax = ∞, lim rmax = 1

p→∞ p→0+
4 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015

p 100 rmax

1 10
20
8
rmax 10 6
5 4
1
1 2
k
q 100 2 4 6 8

Fig. 1: Values of rmax for generalized logistic map depending on p and q

(left) and depending on k (right).

For the sake of parameters p and q the function (3) generates a lot of
new functions which can be analyzed using mathematical methods, just like
for a logistic map, thus obtaining very interesting results.
a) Let us assume that k = qp and k > 0. Then r is depend on the value k
k+1
 
(k+1) k
and r ∈ 0, k .
b) If k = 1, then p = q and rmax is independent on p, q and rmax = 4
always - for logistic map, we have p = q = 1 and rmax = 4.
k+1
(k+1) k
c) If k → ∞, then rmax → 1, because limk→∞ k = 1.
k+1
(k+1) k
d) If k → 0+ , then rmax → ∞, because limk→∞ k = +∞.
e) Now let us consider a case, when p = 1 and q > 0. Then we have

fr (x) = rx(1 − xq ), x ∈ [0, 1], n = 0, 1, 2, .... (5)

 q+1 
(q+1) q
and r ∈ 0, q .
Taking into account the above relationship we can observe change max-
imum of the function (5) depending on q (Fig.2).
 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015 5

x 1
0.8
1
€€€€ 0.5
ã
0.2

0.75
rmax xH1-xq L 0.5

0.25

10
20
q 30

Fig. 2: fr (x) = rx(1 − xq ) for r = rmax depending on q.

A) The property of xn+1 = rxpn (1 − xqn ) for p = 1, q = 2 in the

periodic region.

In the following part of this paper we will show that the Feigenbaum
model of transition to chaos is correct for representation (4) for given pos-
itive values of parameters p and q. Therefore, we will study the dynamics
of this map in the case of p = 1 and q = 2. Then the representation (5)
assumes the following form:

fr (x) = rx(1 − x2 ), x ∈ [0, 1], n = 0, 1, 2, ... (6)

√
3
A function fr in the point x = 3 has
√
a local
√
maximum and the value of
3 2r 3
function in this point is equal to r ( 3 ) = 9 . From this relationship, we
f√
3 3
have f ([0, 1]) ⊂ [0, 1] for r ∈ (0, 2 ]. The trajectories behavior generated
by fr depends on r and, therefore, we have to distinguish several intervals
of this value:
1) For r ∈ (0, 1] all the sequences of iterates x1 , x2 , x3 , ..., xn depend on
the initial value x0 tend to s = 0. This point is the fixed point of the map
for r ∈ (0, 1] and according to Definition 3 it is also the stable point of this
map. q
r−1
2) Let r = r1 > 1. If x = rx(1 − x2 ) then, x1 = 0, x2 = r , so for r > 1
′
fixed points. Using Theorem 1 we get: fr (x) = r−3rx2 ,
function (6) has two q
|fr (0)| = r and |fr ( r−1
′ ′
r )| = |3 − 2r|.
Therefore, we have that x1 = 0 is an unstable fixed point, whereas x2 =
6 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015

q
r−1
ris an attractive fixed point for r ∈ (1, 2).
Change
√ r from 1 to 2 has caused the migration of the attractive point from
0 to 22 .
Conclusion 1.
For 1 < r < 2 there exists
q a stable fixed point attracting all points. When
r = r1 = 2, then |fr ( r−1
′
r )| = |3 − 2r| = 1 and this point stops being an
attractive point.
(2)
3) For r = r2 ≥ 2 we will look into the stability of fixed points fr and fr
of the map (5) as a function of parameter r, because the Feigenbaum model
is generated by bifurcations connected with iterations of a function.
(2)
r 2 x − r 2 x3 −
Let us consider the second iteration of fr , namely fr (x) = q
r 4 x3 + 3r 4 x5 − 3r 4 x7 + r 4 x9 , with fixed points: x1 = 0, x2 = r−1 r , x3 =
q √ q √ ′
1 r 2 −4 1 r 2 −4 (2)
2 − 2r , x4 = 2 + 2r . Because fr (x) = r 2 − 3r 2 x2 − 3r 4 x2 +
′ (2) ′ (2)
15r 4 x4 − 21r 4 x6 + 9r 4 x8 we have fr (x1 ) = r 2 , fr (x2 ) = (3 − 2r)2 ,
′ (2) ′ (2) √
fr (x3 ) = fr (x4 ) = 9 − 2r 2 whereas for 2 < r2 < 5 fixed points x3
and x4 are attractive points according to condition in Theorem 1 .
Conclusion 2. q
(1) (2)
The fixed point x = r−1 r of fr is also the fixed point of fr as well as of
all higher iterations.
Conclusion 3.
(1)
If the fixed point of fr becomes unstable, then it is also the unstable fixed
(2)
point of fr and of all next iterations.
′ ′ ′ ′
From the inequality |f (s)| > 1 we have |f (2) (s)| = |f [f (s)]f (s)| =
′
|f (s)| > 1.
√ (2)
If 5 < r3 < r4 , then the fixed points of fr become repulsive at the same
time. Following this instability, the fourth iteration shows two new pitchfork
bifurcations giving cycle 22 order for four attractive fixed points, which is
called period duplication.
To generalize the above examples we get:
a) For rn−1 < r < rn there exists a stable cycle 2n−1 , whose elements
x0 , x1 , x2 , x3 , ..., x2n−1 −1 are defined by the following connections:
n−1
Y ′
fr (xi ) = xi+1 , fr2 (xi ) = xi , | fr (xi )| < 1
i

b) For rn all points of the cycle of 2n−1 become unstable simultane-

n
ously and the bifurcation creates a new stable cycle of 2n order for fr2 =
n−1 n−1
fr2 (fr2 ) existing for rn−1 < r < rn .
 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015 7

c) For a map (6) the√ value of parameter rn+1 we can describe

√ the following
connection: rn+1 ≈ 3 + rn , where r0 = 1, r1 = 2, r2 = 5, ..., r∞ .
d) There exists a point of accumulation of an infinite number of the bi-
furcation of period duplication
√ for the finite value r which we appointed as
1+ 13
r∞ : r∞ = limn→∞ rn = 2 ≈ 2.303.
√
Taking into√account the above, rn+1 ≈ rn , 3 + r = r, r 2 − r − 3 = 0,
1+ 13
thus r = 2 .
e) For the representation (6) there exists a constant, whose value corre-
sponds to the Feigenbaum constant:
rn − rn−1
δ = lim ≈ 4.61.
n→∞ rn+1 − rn

√ √
Proof. Let us assume that rn−1 = a. Then rn = 3 + rp n−1 = 3 + a,
√ p √ rn −rn−1 √
rn+1 = 3 + rn = 3 + 3 + a and δ = limn→∞ rn+1 −rn = 14 + 2 13 ≈
4.61.
A table of 2n cycles type and values of rn is given below. It is clearly vis-
ible that doubling bifurcations come faster and faster (8, 16, 32, ...) and
suddenly break off beyond a certain point known as the accumulation point
and periodicity gives way to chaos.
n 1 2 3 4 5 … 11 12 … ∞
n
2 cycle 2 4 8 16 32 … 2048 4096 … accum. point

rn 2 2.23606 2.28824 2.29961 2.30209 … 2.3027755 2.30277562 … 2.30277563…

… … … … … …

Table 1: The algebraic orders of the values of rn for n = 1, 2, ... are given
by 1, 2, 4, 8, ....

B) The property of xn+1 = rxpn (1 − xqn ) for p = 1, q = 2 in chaotic

region.
a) For r∞ the bifurcation sequence ends with a set of the infinitely nu-
merous points, which √ is called the set of Feigenbaum attraction.
b) If r∞ < r ≤ 3 2 3 then we observe the irregular dynamics of map (6)
with narrow ranges in which the set of attraction has a periodical character.
c) Numeric calculations for the representation (6) show that the largest
range is for r = rc ≈ 2.451. Then there exists 3-cycle (Fig.3), which accord-
ing to th.2 and th.3, implicates the chaos as well as the existence of cycles
of different orders. √
d) For the map (6) at r = rmax = 3 2 3 ,

√
3 3
xn+1 = xn (1 − x2n ) ≡ frmax (xn ), (7)
2
8 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015

fr HxL r=rc fr HxL r=rc -Dr

1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
n
20 40 60 80 20 40 60 80 100

Fig. 3: Iterates of map (6) strting from x0 = 0.99. Left: In the stable 3-cycle
region for r = rc ; Right: In the chaotic region for r = rc − ∆r, ∆r = 0.001
.

can be solved by the simple change of variables:

xn = cos (yn ) ≡ g(yn ). (8)

Then (7) can be converted into

√ √
3 3 2 3 3
cos (yn+1 ) = cos (yn )[1 − cos (yn )] = cos (yn ) sin 2 (yn ),
2 2
h √ i
3 3
which has one solution yn+1 = arccos 2 cos (yn ) sin 2 (yn ) and

 1
3
p
1 −y0 + −1 + y0 2
xn = arccos( 1 + √ ).
√  3
2 3
p
3 −y0 + −1 + y0

The invariant density function ρ 3√3 (x) of f 3√3 (x) can be calculated using
2 2
the following definition:

N −1 N −1
1 X 1 X
lim δ (x − xn ) = lim δ (x − g(yn )) (9)
N →∞ N N →∞ N
n=0 n=0

R1
Using ρ(y) = 1, (9) becomes: ρ 3√3 (x) = 0 dyρ(y)δ[x − h(y)]
2
 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015 9

i.e.
√ 2/3
x2 − 1 − x −1
ρ 3√3 (x) = s .
2 √ p
3
√ √ 2/3
1
3 x2 − 1 x2 − 1 − x − x2 − 1 − x − √ 2/3 +1
( x2 −1−x)
(10)
The above function is shown in Fig.4.

ΡHxL

x
0.1 0.3 0.5 0.7 0.9 1 0.1 0.3 0.5 0.7 0.9 1
√
3 3
Fig. 4: The invariant density function of map (6) for r = rmax = Left: 2 .
An analytical form represented by the equation (10). Right: Numerical
simulation - the histogram for 50000 iterations.

The invariant density function of the logistic map for r = rmax = 4 is

given by √ 1 .
π (1−x)x

C) Ordered chaos - construction √

of an attractor for
2 3 3
xn+1 = rxn (1 − xn ) and r = rmax = 2 .

At the end of this considerations we will present a method of the order-

ing of chaos. A lot of numeric experiments have led to the situation where
it is possible to rank chaos in such a way that the successive iterations gen-
erated by chaotic dynamic system would reach an attractor of a regular
shape. This method also has it, that from the sequence of several thousand √
iterations (for representation xn+1 = rxn (1 − x2n ) where r = rmax = 3 2 3
defined recursively) we draw up the following points (xn+1 , xn xn+1 ) on a
plane - we construct an attractor. In this way an attractor comes into be-
ing, which is visited irregularly by points of successive iterations (Fig.5).
10route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015

Fig. 5: The set of attraction for representation (6) [60000 iterations]: a) for
r = rmax (red), r = 2.56(blue), r = 2.53(green), r = 2.45(orange); b) for
r = rmax (red), r = 2.42(black), r = 2.4(blue).

D) Bifurcation diagram and Lyapunov exponent.

Nowadays the Lyapunov exponent is applicable in various scientific dis-

ciplines such as biology, engineering, bioengineering and informatics [15, 16,
17, 18, 19, 20]. A quantitative measure of the sensitive dependence to initial
conditions is the Lyapunov exponent, which is a measure of the exponential
separation of nearby orbits. For discrete time system xn+1 = f (xn ), for an
orbit starting with x0 the Lyapunov exponent λ(x0 ) can be calculated by

n−1
1X
λ(x0 ) = lim ln |f ′ (xi )|, (11)
n→∞ n
i=0

A positive Lyapunov exponent can be considered as an indicator of chaos,

whereas negative exponents are associated with regular and periodic be-
havior of the system. Figure 6 shows the bifurcation diagram and Lya-
punov exponent
√
for the equation (6). It is clear that even in chaotic areas
3 3
(r∞ < r ≤ 2 ) there are many periodic intervals (λ < 0).
 route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015 11

1.0

0.5
x,Λ

0.0

-0.5

2.2 r¥ 2.4 2.5 3 3

2
r
Fig. 6: The bifurcation
√ diagram and the Lyapunov exponent for the map
3 3
(6) and 2.2 < r ≤ 2 .

3. Summary
Currently, modeling real-world systems is very popular. A well-known
model that takes into account the deterministic chaos is the logistic map.
However, this one-dimensional map can be controlled only by one parameter.
In the present contribution, we postulate a generalization of the classical
logistic map which is generally controlled by three parameters. This could
allow to adapt to the conditions prevailing in the modeled system.
We focus on the rmax parameter (depending on p and q) which is re-
sponsible for the dynamics of the system - among others shown that if p = q
then rmax = 4.
As an example, we carry out analytical and quantitative analysis in the
case where p = 1 and q = 2 both in the periodic and chaotic regime where
the Lyapunov exponent has positive values. It turns out that the dynamics
of this equation is faster than in the case of the logistic map - the value of
the parameter rmax ≈ 2.3 is less than the value of rmax = 4 for the logistic
map. In the periodic regime we identified Feigenbaum constant δ which is
typical for all dissipative systems (also for the logistic map). For chaotic
area we have found an analytical form of the invariant density function. For
both regimes we have proposed specific form of data representation which
allowed to obtain a non-trivial structure of an attractor set.

Footnotes
This work was partially supported by the Centre for Innovation and Transfer
of Natural Sciences and Engineering Knowledge (University of Rzeszow).
 12route˙to˙cahaos˙fens2014˙revised˙rak printed on February 9, 2015

REFERENCES

[1] N.K. Pareeka, Vinod Patidara, K.K. Sud, Image encryption using chaotic lo-
gistic map, Image and Vision Computing, 24, 926-934, (2006).
[2] A. Ferrettia, N.K. Rahman, A study of coupled logistic map and its applications
in chemical physics, Chemical Physics, 119, 275-288, (1988).
[3] C. Pellicer-Lostao, R. Lpez-Ruiz A chaotic gas-like model for trading markets,
Journal of Computational Science, 1, 2432, (2010).
[4] M. Andrecut, Logistic map as a random number generator, Int. J. Modern Phys.
B, 12, 921930, (1998).
[5] M.J. Feigenbaum, The Metric Universal Properties of Period Doubling Bifurca-
tions and the Spectrum for a Route to Turbulence, NewYork. Acad. Sci., 330-336,
(1980).
[6] M.J. Feigenbaum, Quantitative Universality for a Class of Non-Linear Trans-
formations, Stat. Phys., 25-52, (1978).
[7] C.J. Thompson, J. B. McGuire, Asymptotic and Essentially Singular Solutions
of the Feigenbaum Equation, J. Stat. Phys. 51, 991-1007, (1988).
[8] J.W. Stephenson, Y. Wang, Numerical Solution of Feigenbaum’s Equation,
Appl. Math. Notes 15, 68-78, (1990).
[9] J. Smital, On Functions and Functional Equations, IOP Publishing Ltd, Bristol
1998.
[10] T. Li, J.A. Yorke, Period three implies chaos, Amer. Math., 985-992, (1975).
[11] Pierre-Franois Verhulst, Notice sur la loi que la population poursuit dans son
accroissement, Correspondance mathmatique et physique, 113121, (1838).
[12] E. Ott, Chaos in Dynamical Systems, Cambridge University Press, (1993).
[13] H.G. Schuster, Deterministic chaos: an introduction, Weinheim, (1988).
[14] R. Rak, S. Drożdż (supervisor), Modeling of population growth, masters thesis,
University of Rzeszów, (2003).
[15] A. Lyapunov, Problem general de la stabilite du mouvement, Ann Math Stud
17, (1949).
[16] A. Wolf, J.B. Swift, H.L. Swinney, J.A. Vastano, Determining Lyapunov Ex-
ponents from a Time Series, Physica D, 16, 285-317, (1985).
[17] M.A. Van Wzk, W. STEEB, Kluwer Academic Publishing, (1997).
[18] S.H. Strogatz, Nonlinear dynamics and chaos: with applications to physics,
biology, chemistry, and engineering, Addison-Wesley, New York, (1994).
[19] A.H. Nayfeh, B. Balachandran, Applied nonlinear dynamics: analytical, com-
putational, and experimental methods, John Wiley and Sons, New York (1995).
[20] P. Bryant, R. Brown, H. Abarbanel, Lyapunov exponents from observed time
series, Physical Review Letters 65, (1990).

View publication stats