This paper was converted on www.awesomepapers.org from LaTeX by an anonymous user.
Want to know more? Visit the Converter page.

\usepackage

[pdftex]graphicx \usepackageascmac \usepackageamsmath,graphicx,amssymb,amscd \setlength\oddsidemargin10mm \setlength\evensidemargin10mm \setlength\topmargin-7mm \setlength\textheight225mm \setlength\textwidth150mm

\newtheorem

defineDefinition \newtheoremlemmaLemma \newtheoremtheoremTheorem

On a parametrized difference equation connecting chaotic and integrable mappings

Tomoko Nagai1, Atsushi Nagai2, Hiroko Yamaki2, and Kana Yanuma2
1 Academic Support Center, Kogakuin University,
2665-1 Nakano-cho, Hachioji, Tokyo 192-0015, Japan
2Department of Computer Sciences, Tsuda University,
2-1-1 Tsuda-cho, Kodaira, Tokyo 187-8577, Japan
[email protected]
Abstract

We present a new difference equation with two parameters c[0,1]c\in[0,1] and A[1,4]A\in[1,4]. This equation is equivalent to the logistic mapping if c=1c=1 and the Morishita mapping if c=0c=0, which are the well-known chaotic and integrable mappings, respectively. We first consider the case A=4A=4 and investigate the time evolution by changing the parameter c[0,1]c\in[0,1]. We next change both two parameters A[3,4]A\in[3,4] and c[0,1]c\in[0,1] and present the corresponding 3D bifurcation diagram.

1 Introduction

The logistic equation, {align} dudt = a u(1-u)  u(0)=u_0, where aa is a positive constant and u=u(t)u=u(t) is an unknown function, is a model equation describing population dynamics and possesses a solution {align} u(t)=u0eatu0eat+1-u0 Concerning its discrete version, the following two difference equations are well-known: {align} & u_n+1= A u_n(1-u_n),
u_n+1 = A u_n(1-u_n+1) ⇔u_n+1=A un1+A un, where A(1,4]A\in(1,4] is a given constant.

Eq. (1)(\ref{eq:logistic}) is the well-known logistic mapping and exhibits a chaotic behavior if AA exceeds the value A=3.5699456A=3.5699456\cdots.(See [1], for example.)

On the other hand, eq. (1)(\ref{eq:morisita}), which is called the Morishita mapping, is an integrable mapping [2, 3]. In other words, eq. (1)(\ref{eq:morisita}) is linearized by taking its reciprocal and putting vn=1/unv_{n}=1/u_{n} as {align*} v_n+1=1Av_n+1. This is solved as {align*} v_n=AA-1+1An(1u0-AA-1) and therefore unu_{n} is given by {align} u_n=An(A-1)u0A(An-1)u0+A-1, which converges to \dfracA1A\dfrac{A-1}{A} as nn tends to \infty.

The purpose of this paper is to present a new difference equation connecting the above two different mappings. We also investigate the time evolution of this new equation and calculate the Lyapunov exponent. The bifurcation diagram is also presented.

2 A new difference equation connecting the logistic and Morishita mappings

In this section, we put A=4A=4 and consider a new mapping including a variable parameter c[0,1]c\in[0,1], {align} & u_n+1=4u_n(1-cu_n-(1-c)u_n+1)
⇔  u_n+1=4un(1-cun)1+4(1-c)un=f(c,u_n), where f(c,x)=\dfrac4x(1cx)1+4(1c)xf(c,x)=\dfrac{4x(1-cx)}{1+4(1-c)x}. It is easy to observe that the mapping (2)(\ref{eq:a}) is equivalent to the Morishita mapping if c=0c=0 and to the logistic mapping if c=1c=1.

Concerning the properties of f(c,x)f(c,x), we have the following theorem. {theorem} The function f(c,x)(0x1,0c1)f(c,x)\quad(0\leq x\leq 1,~{}0\leq c\leq 1) satisfies the following properties. {align} & f(c,34)=34
0≤f(c,x) ≤1 The relation (2)(\ref{eq:th1}) means that un=34u_{n}=\frac{3}{4} is an equilibrium point of the mapping (2)(\ref{eq:a}) and (2)(\ref{eq:th2}) means that if u0[0,1]u_{0}\in[0,1] we have un[0,1]u_{n}\in[0,1] for any n=1,2,n=1,2,\cdots.
Proof :  The relation (2)(\ref{eq:th1}) is easily shown through direct calculation. In order to prove (2)(\ref{eq:th2}), we take a derivative of f(c,x)f(c,x) with respect to cc, which is given by {align*} ∂∂c f(c,x)=4x2(3-4x)(1+4x-4cx)2. Hence we have {align*} f(0,x) ≤f(c,x) ≤f(1,x) &  (0 ≤x ≤34),
f(1,x) ≤f(c,x) ≤f(0,x)   (34 ≤x ≤1). Together with the facts, {align*} 0≤f(0,x)=4x1+4x¡1,   0≤f(1,x)=4x(1-x) ≤1, the relation (2)(\ref{eq:th2}) follows. \blacksquare\blacksquare

By changing cc in an interval [0,1][0,1], we calculate the time evolutions of {un}\{u_{n}\}, which are given in Fig. 2. Numerical calculations are performed by Python 3.9.

\includegraphics

[scale=0.5]period1.png \includegraphics[scale=0.5]period2.png \includegraphics[scale=0.5]period4.png \includegraphics[scale=0.5]period8.png \includegraphics[scale=0.5]chaos1.png \includegraphics[scale=0.5]period3.png

Figure \thefigure: Time evolutions of the mapping (2)(\ref{eq:a}) for c=0.7c=0.7c=0.9c=0.9(period 2), c=0.93c=0.93(period 4),  c=0.94c=0.94(period 8), c=0.97c=0.97(chaotic), c=0.98c=0.98(period 3)

We here investigate the Fig. 2 in a detailed manner. If 0c<560\leq c<\frac{5}{6}, we can observe that unu_{n} converges to 34\frac{3}{4}. This is confirmed as follows. We put un=34+εu_{n}=\frac{3}{4}+\varepsilon in eq. (2)(\ref{eq:a}) and have the Taylor expansion of un+1=f(c,un)u_{n+1}=f(c,u_{n}) around ε=0\varepsilon=0 as follows. {align*} u_n+1&=(3+4ε)(1-c(34+ε))1+(1-c)(3+4ε) =34+-1+3c-4+3cε+⋯. If 0c<560\leq c<\frac{5}{6} we have |1+3c4+3c|<1\left|\frac{-1+3c}{-4+3c}\right|<1 and therefore the equilibrium point un=34u_{n}=\frac{3}{4} is stable.

If cc exceeds 56\frac{5}{6}, we have |1+3c4+3c|>1\left|\frac{-1+3c}{-4+3c}\right|>1 and therefore un=34u_{n}=\frac{3}{4} is unstable. It should be noted that un=0u_{n}=0 is another equilibrium point, which is, however, unstable for c[0,1]c\in[0,1].

Through straightforward calucations, if 56<c\frac{5}{6}<c, unu_{n} converges to a periodic orbit with period 2. That is, unu_{n} takes alternately two values {align*} u_2,±=-5+10c±55-16c+12c28(-c+2c2), which are solutions of the equation f(c,f(c,x))=xf(c,f(c,x))=x except x=0,34x=0,~{}\frac{3}{4}, as nn tends to \infty. In other words, we have {align*} f(c,u_2,±)=u_2,∓. The Taylor expansion of f(c,f(c,u2,±+ε))f(c,f(c,u_{2,\pm}+\varepsilon)) around ε=0\varepsilon=0 is given by {align*} f(c,f(c,u_2,±+ε))=u_2,±+63c2-84c+253c2-4cε+⋯. Hence if cc exceeds the value 44+28666=0.922902\frac{44+\sqrt{286}}{66}=0.922902\cdots, which is a solution to |63c284c+253c24c|=1\left|\frac{63c^{2}-84c+25}{3c^{2}-4c}\right|=1, the period of the sequence {un}\{u_{n}\} becomes 4.

If we further increase the parameter cc, the period of the sequence {un}\{u_{n}\} is doubled as 23,24,25,2^{3},2^{4},2^{5},\cdots and finally a chaotic behavior appears at c=0.942c=0.942\cdots. We next calculate the value cnc_{n} where unu_{n} is 2n2^{n}-periodic if cn<c<cn+1c_{n}<c<c_{n+1}. We further calculate {align} F_n=cn+1-cncn+2-cn+1, which is expected to converge to the Feigenbaum constant 4.6692014.669201\cdots as is shown in Table 2.

Table \thetable: nn dependence of FnF_{n} in Eq. (2)(\ref{eq:a})
nn 1 2 3 4 5 6 7 8
cnc_{n} 0.833333 0.922902 0.93803 0.941167 0.941833 0.9419755 0.94200603 0.94201257
FnF_{n} 5.920743 4.822442 4.710210 4.657343 4.761904 4.668196

It is interesting to note that the period 3 appears around c3(1)<c<c3(2)c_{3}^{(1)}<c<c_{3}^{(2)}, which reminds us of Li and York’s famous result [4]. The approximated critical values are given by c3(1)=0.978471c_{3}^{(1)}=0.978471\cdots and c3(2)=0.980113c_{3}^{(2)}=0.980113\cdots.

In Fig. 2, we give a Lyapnov exponent λ=λ(c)\lambda=\lambda(c) defined by {align} λ=lim_n→∞ 1n ∑_j=0^n-1 log ∂∂x f(c,u_j) . We can observe that λ\lambda takes negative value for c<0.942c<0.942\cdots and then takes positive value.

\includegraphics

[scale=0.5]lyapnov.png

Figure \thefigure: Lyapunov exponent
\includegraphics

[scale=0.5]figure1.png \includegraphics[scale=0.5]figure2.png

Figure \thefigure: Bifurcation diagram of the mapping (2)(\ref{eq:a})  (0.8c1)(0.8\leq c\leq 1)
\includegraphics

[scale=0.5]figure3.png \includegraphics[scale=0.5]figure4.png

Figure \thefigure: Enlarged bifurcation diagram. (Left : 0.92c0.950.92\leq c\leq 0.95, Right : 0.938c0.9450.938\leq c\leq 0.945)

The corresponding bifurcation diagrams, where (c,un)(n=200,,400)(c,u_{n})~{}~{}(n=200,\cdots,400) are plotted, are given in Fig. 2. If we expand a certain region of the diagram, we can find a self-similar diagram, as is observed in Fig. 2.

3 The mapping with two parameters and 3D bifurcation diagram

Next we extend Eq. (2)(\ref{eq:a}) to a mapping with two parameters A[3,4]A\in[3,4] and c[0,1]c\in[0,1], given by {align} & u_n+1=Au_n(1-cu_n-(1-c)u_n+1)
⇔  u_n+1=Aun(1-cun)1+A(1-c)un=f(A,c,u_n), where f(A,c,x)=\dfracAx(1cx)1+A(1c)xf(A,c,x)=\dfrac{Ax(1-cx)}{1+A(1-c)x}.

Changing the two parameters AA and cc, we obtain the corresponding bifurcation diagrams shown in Fig. 3, in which the upper and lower figures stand for the same diagram looked from different viewpoints. It should be noted that if A\leqq3A\leqq 3 unu_{n} converges to the value 11A1-\frac{1}{A} so we omitted.

We next calculate Fn=Fn(A)F_{n}=F_{n}(A) of Eq. (2)(\ref{eq:Fn}), which is given as Table 3:

\includegraphics

[scale=0.5]fig0103v1.png
\includegraphics[scale=0.5]fig0103v2.png

Figure \thefigure: 3D bifurcation diagram corresponding to Eq. (3)(\ref{eq:twopara}) from different viewpoints
Table \thetable: FnF_{n} in Eq. (3)(\ref{eq:twopara})
nn=1 nn=2 nn=3 nn=4
AA=3.80 5.89888 4.82824 4.70431 4.67570
AA=3.85 5.90474 4.82648 4.70576 4.67548
AA=3.90 5.90988 4.82597 4.70410 4.67736
AA=3.95 5.91525 4.82493 4.70450 4.67634

4 Concluding Remarks

We have investigated a mapping with two parameters connecting the logistic mapping and the Morishita mapping, which are famous chaotic and integrable difference equations, respectively. It is very interesting to note that a peridic orbit of period 3 appears and a self-similarity of the bifurcation diagram is observed. These phenomena are also observed in the logistic mapping un+1=Aun(1un)(0<A4)u_{n+1}=Au_{n}(1-u_{n})\quad(0<A\leq 4). From the obtained results, we may conclude that the difference equation (3)(\ref{eq:twopara}) is a chaotic mapping.

Acknowledgment

One of the authors (A. N.) is supported by JSPS KAKENHI Grant Number 18K03347.

  • [1] R. M. May, Nature 261, 457 (1976).
  • [2] M. Morishita, Res. Popul. Ecol. 7, 52 (1965).
  • [3] R. Hirota and D. Takahashi, Discrete and ultra-discrete, (Kyoritsu Pub. Tokyo, 2003) [in Japanese].
  • [4] T. Y. Li and J. A. Yorke, Amer. Math. Monthly 82, 985(1975).