Analytic Solutions of Second Order Nonlinear Difference Equations all of whose Eigenvalues are 1
MAMISUZUKI(*)
ABSTRACT- For nonlinear difference equations, it is difficult to obtain analytic so- lutions. Especially, when all the absolute values of the equation are equal to 1, it is quite difficult.
We consider a second order nonlinear difference equation which can be transformed into the following simultaneous system of nonlinear difference equations,
x(t1)X(x(t);y(t));
y(t1)Y(x(t);y(t));
(
whereX(x;y)l1x P
ij^2cijxiyj,Y(x;y)l2y P
ij^2dijxiyjand we assume some conditions. For these equations, whenjl1j 61or jl2j 61, we have ob- tained analytic solutions in earlier studies. In the present paper, we will prove the existence of an analytic solution and obtain analytic solutions of the differ- ence equations for the casel1l21:
1. Introduction.
We start by considering the following second order nonlinear differ- ence equation,
u(t1)U(u(t);v(t));
v(t1)V(u(t);v(t));
( (1:1)
whereU(u;v) an dV(u;v) are entirefunctionsforuandv. Wesuppose thatthe
(*) Indirizzo dell'A.: Department of Mathematics and Science, College of Liberal Arts, J. F. OberlinUniversity. 3758 Tokiwa-cho, Machida-City, Tokyo, 194-0294, Japan.
E-mail: m-suzuki@obirin.ac.jp
2000Mathematics Subject Classification. 39A10, 39A11, 39B32.
equation(1.1) admits anequilibrium point (u;v): u v
U(u;v) V(u;v)
. We canassume, without loss of generality, that (u;v)(0;0):Furthermore we suppose thatUandVare writteninthe following form
u(t1) v(t1)
M u(t) v(t)
U1(u(t);v(t)) V1(u(t);v(t))
;
whereU1(u;v) an dV1(u;v) are higher order terms ofuandv, an dMis a constant matrix. Letl1,l2be the characteristic values of the matrixM. For some regular matrixPdetermined byM, put u
v P xy , thenwe can transform the system (1.1) into the following simultaneous system of first order difference equations (1.2):
x(t1)X(x(t);y(t));
y(t1)Y(x(t);y(t));
( (1:2)
whereX(x;y) an dY(x;y) are supposed to be holomorphic and expanded in a neighborhood of (0;0) inthe form
X(x;y)l1x X
ij^2
cijxiyjl1xX1(x;y);
Y(x;y)l2y X
ij^2
dijxiyjl2yY1(x;y);
8>
><
>>
(1:3) : or
X(x;y)lxy X
ij^2
c0ijxiyjlxX10(x;y);
Y(x;y)ly X
ij^2
d0ijxiyjlyY10(x;y);
8>
><
>>
(1:4) :
where, inthe second case,ll1l2.
In this paper we consider analytic solutions of difference system (1.2).
In [9] and [8], already, we have obtained general analytic solutions of (1.2) inthe case jl1j 61 or jl2j 61. However inthe case jl1j jl2j 1, it is difficult to prove the existence of analytic solution or obtain an analytic solution of the equation. For a long time we have not been able to treat equation (1.2) under the latter condition.
For analytic solutions of nonlinear first order difference equations, Kimura [2] has studied the cases inwhich one eigenvalue is equal to 1, furthermore Yanagihara [10] has studied the cases in which the absolute value of the eigenvalue is equal to 1. Here we shall study analytic solutions
of nonlinear second order difference equations in which the absolute values of the eigenvalues of the matrixMare both equal to 1.
As anexample of the equation(1.3), we consider the following ``popu- lationmodel''
u(t2)au(t1)bu(t1) au(t)
au(t) ; fort> 1;
wherea1r>0,b>0 are constants,ris the net (births minus death) endogenous population growth rate. This model was proposed by Prof. D.
Dendrios [1], and may be transformed into a system as in (1.2). In [6], we have investigated properties of a solution assuming its existence when a1, i.e.r0. However if both eigenvalues of the equation are equal to 1, i.e. under the conditionr0, the existence of a solution of the model is not established. Inthe next paper, we will show a solutionof the population model inthe caser0.
Moreover we have studied some economic models in the form (1.2), but we had to exclude the casejl1j jl2j 1. For example, in[7], we consider the following ``duopoly system''
x(t1)ay(t)(1 y(t)) y(t1)bx(t)(1 x(t));
(
whereaandbare constants. For this system, we havel1 l2. In[7], we consider the system under the condition such that 05l21l2251.
Therefore we will investigate the equation (1.2) in the case jl1j
jl2j 1.
In this present paper, making use of theorems in [2] and [5], we will prove the existence of a solution and obtain an analytic solution of (1.2) under the conditionsl1l21, inwhichX,Yare defined by (1.3), i.e., we suppose that
X(x;y)x X
ij^2;i^1
cijxiyjxX1(x;y);
Y(x;y)y X
ij^2;j^1
dijxiyjyY1(x;y):
8>
><
>>
(1:5) :
Here we suppose thatX1(x;y)60 orY1(x;y)60.
Next we consider a functional equation
C(X(x;C(x)))Y(x;C(x));
(1:6)
whereX(x;y) an dY(x;y) are holomorphic functions injxj5d1,jyj5d1. We assume thatX(x;y) an dY(x;y) are expanded there as in (1.5).
Now we will consider the meaning of equation (1.6).
Consider the simultaneous system of difference equations (1.2). Suppose (1.2) admits a solution(x(t);y(t)). Ifdx
dt 60 for somet0, thenwe canwrite tc(x) with a functioncina neighborhood ofx0x(t0), and we can write
yy(t)y(c(x))C(x);
(1:7)
there. Thenthe functionCsatisfies the equation(1.6).
Conversely we assume that a functionCis a solutionof the functional equation(1.6). If the following first order difference equation
x(t1)X(x(t);C(x(t)));
(1:8)
has a solutionx(t), we puty(t)C(x(t)). Thenthe (x(t);y(t)) is a solutionof (1.2). Hence if there is a solutionCof (1.6), thenwe canreduce the system (1.2) to a single equation (1.8).
This relation is important in order to derive analytic solutions of non- linear second order difference equations which are written as in (1.2). We have proved the existence of solutionsC of (1.6) in[3] (or [4]) and [8] for other conditions onXandY.
Hereafter we considertto be a complex variable, and concentrate on the difference system (1.2). We define a domainD1(k0;R0) by
D1(k0;R0) ft : jtj>R0;jarg[t]j5k0g;
(1:9)
wherek0 is any constant such that 05k0%p
4, an dR0is a sufficiently large numberwhichmaydependonXandY.FurtherwedefineadomainD(k;d) by
D(k;d) fx : jarg[x]j5k;05jxj5dg;
(1:10)
where d is a small constant, and k is a constant such that k2k0, i.e., 05k%p
2.
Our aim inthis paper is to prove the following Theorem 1.
THEOREM1. Suppose X(x;y)and Y(x;y)are expanded in the forms (1.5)such that X1(x;y)60or Y1(x;y)60,and put Ac20.
(1)Suppose that kc20 d11for some k2N, k^2,then we have a formal solution x(t)of(1.2)the following form
1
At 1 X
jk^1
^qjkt j logt
t
k! 1
; (1:11)
where^qjkare constants defined by X and Y.
(2) Suppose kc20d1150 for some k2N, k^2,and R1
max (R0;2=(jAjd)),then there is a solution x(t)of(1.2)such that (i)x(t)is holomorphic and x(t)2D(k;d)for t2D1(k0;R1), (ii)x(t)is expressible in the following form
x(t) 1 At 1b
t;logt
t
! 1
; (1:12)
where b(t;logt=t)is asymptotically expanded for t2D1(k0;R1)such that b
t;logt
t
X
jk^1
^qjkt j logt
t k
; as t! 1through D1(k0;R1).
2. Proof of Theorem1.
In [2], Kimura considered the following first order difference equation w(tl)F(w(t));
D1
whereFis represented in a neighborhood of1by a Laurent series F(z)z
1X1
jm
bjz j
; bml60:
(2:1)
He defined the following domains D(e;R) ft : jtj>R;jarg[t] uj5p
2 e;orIm(ei(u e)t)>R;
orIm(ei(ue)t)5 Rg;
(2:2)
D(e;^ R) ft : jtj>R;jarg[t] u pj5p
2 e orIm(e i(up e)t)>R
orIm(e i(upe)t)5 Rg;
(2:3)
where e is anarbitrarily small positive number,R is a sufficiently large number which may depend oneandF, an duis defined byuargl, (inthis present paper, we consider the casel1 in(D1)). He proved the following Theorems A and B.
THEOREMA. -Equation(D1)admits a formal solution of the form t 1 X
jk^1
^qjkt j logt
t k! (2:4)
containing an arbitrary constant,where^qjkare constants defined by F.
THEOREMB. -Given a formal solution of the form(2.4)of(D1),there exists a unique solution w(t)satisfying the following conditions:
(i) w(t)is holomorphic in D(e;R), (ii) w(t)is expressible in the form
w(t)t 1b
t;logt t
!
; (2:5)
where the domain D(e;R)is defined by(2.2),and b(t;h)is holomorphic for t2D(e;R),jhj51=R,and in the expansion
b(t;h)X1
k1
bk(t)hk: Here bk(t)is asymptotically developed into
bk(t) X1
jk^1
^ qjkt j;
as t! 1through D(e;R),where^qjkare constants which are defined by F.
Also there exists a unique solutionw which is holomorphic in^ D(e;^ R) and satisfies a condition analogous to (ii),where the domainD(e;^ R)is defined by(2.3).
In Theorems A and B, Kimura defined the functionFas in(2.1). Inour method, we do not have a Laurent series for the function F. Hence we derive the following Propositions.
Inthe following,Awill be a constant which is defined as in Theorem 1 such thatAc20, wherec20 is the coefficient in (1.5).
PROPOSITION2. -SupposeF(t)~ is formally expanded such that F(t)~ t 1X1
j1
bjt j
!
; b1l60:
(2:6)
Then the equation
c(F(t))~ c(t)l (2:7)
has a formal solution
c(t)t 1X1
j1
qjt jqlogt t
!
; (2:8)
where q1can be arbitrarily prescribed while other coefficients qj(j^2)and q are uniquely determined by bj,(j1;2; ),independently of q1.
PROPOSITION3. -SupposeF(t)~ is holomorphic and expanded asymp- totically inft : 1=(At)2D(k;d);A50gas
F(t)~ t 1X1
j1
bjt j
!
; b1l60;
where D(k;d)is defined in(1.10).Then the equation(2.7)has a solution wc(t);which is holomorphic inft : 1=(At)2D(k=2;d=2);A50gand has an asymptotic expansion
c(t)t 1X1
j1
qjt jqlogt t
!
;
there.
These Propositions are proved as in [2] pp. 212-222. SinceAc2050 and k0 k=2, we see that x 1=(At)2D(k=2;d=2) equivalent to t2D1(k=2;2=(jAjd))D1(k0;2=(jAjd)), whereD1(k0;R0) is defined in (1.9).
We define a functionfto be the inverse ofc, so thatwc 1(t)f(t).
Thenwe havefc(w)w;cf(t)t, furthermorefis a solutionof the following difference equation
w(tl)F(w(t));~ D
whereF~is defined as inProposition2 (see pp. 236 in[2]). Hereafter, we put l1. Sinceu0, we then have the following Propositions 4 and 5, ana- logous to Theorems A and B.
PROPOSITION4. -SupposeF(t)~ is formally expanded as in(2.6).Then the equation(D)has a formal solution
wf(t)t 1 X
jk^1
^qjkt j logt t k!
; (2:9)
where^qjkare constants which are defined byF as in Theorem A.~
PROPOSITION5. - Suppose a functionf is the inverse of c such that wc 1(t)f(t). Given a formal solution of the form(2.9)of (D),there exists a unique solution w(t)f(t)which is holomorphic and admits an asymptotic expansion for t2D1(k0;2=(jAjd))such that
wf(t)t 1b
t;logt t
!
; (2:10)
where
b t;logt t
X
jk^1
^qjkt j logt t k
:
This functionf(t)is a solution of difference equation of(D).
In[5], we proved the following Theorem C.
THEOREMC. -Suppose X(x;y)and Y(x;y)are defined in(1.5).Then we have following:
(1)If kc206d11for any k2N, k^2,the formal solutionC(x)of(1.6)of the following form
C(x)X1
m1
amxm; (2:11)
is identical to0,i.e.,a1a2 0.
(2)If kc20d11for some k2N, k^2,we have a formal solutionC(x)of (1.6)of the following form
C(x)X1
mk
amxm; (2:12)
i.e.,a1a2 ak 10.
(3)Suppose
kc20d1150 for some k2N; k^2:
(2:13)
For anyk,05k% p
2and smalld,we define the following domain D(k;d), D(k;d) fx : jarg[x]j5k;05jxj5dg:
(1:10)
There is a constantd>0and a solutionC(x)of(1.6),which is holomorphic
and can be expanded asymptotically in D(k;d)such that C(x)X1
jk
ajxj: (2:14)
PROOF OFTHEOREM 1. - We will first prove (1) of Theorem 1. From Theorem C (2), we have a formal solution C(x) of (1.6) which canbe for- mally expanded so that one has
C(x)X1
jk
ajxj: (2:15)
On the other hand puttingw(t) 1
Ax(t)in(1.8), sinceAc20, we have w(t1) 1
Ax(t1) 1
AX x(t);C(x(t))
1
AX 1
Aw(t);C
1 Aw(t)
!: (2:16)
From (1.5), we have X x(t);C(x(t))
x(t) X
ij^2;i^1
cijx(t)i(C(x(t))j
x(t)
1 X
ij^2;i^1
cijx(t)i 1(C(x(t))j :
Thus we have 1 X x(t);C(x(t))
1
x(t)n
1 P
ij^2;i^1 cijx(t)i 1(C(x(t))jo
1 x(t)
"
1
X
ij^2;i^1
cijx(t)i 1(C(x(t))j
X
ij^2;i^1
cijx(t)i 1(C(x(t))j2
X
ij^2;i^1
cijx(t)i 1(C(x(t))j3
# :
Sincew(t) 1
Ax(t), we have 1
X x(t);C(x(t)) Aw(t)
"
1 X
ij^2;i^1
cij
1 Aw(t)
i 1
C 1 Aw(t)
j!
X
ij^2;i^1
cij 1 Aw(t)
i 1
C 1
Aw(t) j!2
X
ij^2;i^1
cij 1 Aw(t)
i 1
C 1 Aw(t)
j3
# : Since C(x) is a formal solutionof (1.6) such that C(x)C 1
Aw
P1
mkaj 1
Aw m
, (k^2), we have 1
AX x;C(x)w
1c201
Aw 1X
k^2
~ck(w) k
; (2:17)
where~ckare defined bycijandak(ij^2,i^1,k^2). From (2.17) and the definition ofA, we canwrite (2.16) inthe following form
w(t1)F(w(t))~ w(t)
1w(t) 1X
k^2
~ck(w(t)) k
: (2:18)
On the other hand, puttingl1 an d m1 in(2.1), i.e.uarg[l]
arg[1]0, and making use of Proposition 4, we have the following for- mal solutionw(t) (2.19) of the first order difference equation (2.18),
w(t)t 1 X
jk^1
^qjkt j logt
t k!
; (2:19)
where^qjkare defined byF~in(2.18). From (2.17) and (1.6),F~is defined byX andY. Hence^qjkare defined byXandY.
Fromx(t) 1
Aw(t), we have a formal solutionof (1.2) such that x(t) 1
At 1 X
jk^1
^qjkt j logt
t
k! 1 : (2:20)
From the relationof (1.2) and (1.8) with (1.6), we have proved (1) of Theorem 1.
Next we will show (2). From the assumptionthatkc20 q1150 for some
k2N,k^2, we suppose thatR0>Randk05p
4 e. Sinceu0, we have D1(k0;R0)D(e;R):
(2:21)
For ax2D(k;d), making use of Theorem C (3), we have a solutionC(x) of (1.6) which is holomorphic and can be expanded asymptotically inD(k;d) such that
C(x)X1
jk
ajxj: (2;14)
SinceR1 max (R0;2=(jAjd)), making use of Proposition 5, we have a so- lutionw(t) of (2.18) which has anasymptotic expansion
w(t)t 1b
t;logt t
!
;
in t2D1(k0;R1). Therefore we have a solutionx(t) of (1.2) which has an asymptotic expansion
x(t) 1 At 1b
t;logt
t
! 1
;
there. At first we take a small d>0. For sufficiently large R, since R1^R0>R, we will have
jx(t)j 1
At 1b
t;logt
t
1
5 1
jAjR(11)5d;
(2:22)
fort2D1(k0;R1). SinceAc2050,
arg[x(t)]arg
"
1 At 1b
t;logt
t
! 1#
arg[t] arg
1b
t;logt t
# : For sufficiently largeR1, we thenhave
arg
"
1b
t;logt t
#5k0; fort2D1(k0;R1):
Hence we have
k0 k0arg[x(t)]k0k0:
From the assumptionofk2k0, we have jarg[x(t)]j5k% p
2 for t2D1(k0;R1):
(2:23)
From (2.22) and (2.23), we havex(t)2D(k;d) for a somek;
05k%p 2
. Hence we have aC(x(t)) which satisfies the equation(1.6).
Therefore from the existence of a solutionCof (1.6) and Proposition 5, we have a holomorphic solutionw(t) of first order difference equation (2.18) fort2D1(k0;R1). Thus we obtaina solutionx(t) of (1.2) fortthere, which satisfies the following conditions:
(i) x(t) is holomorphic inD1(k0;R1), (ii) x(t) is expressible inthe form
x(t) 1 At 1b
t;logt
t
! 1
: (2:24)
Hereb(t;logt=t) is asymptotically expanded fort2D1(k0;R1) such that b
t;logt
t
X
jk^1
^qjkt j logt
t k
;
ast! 1throughD1(k0;R1). p
Finally, we obtain a solution (u(t),v(t)) of (1.1) by the transformation u(t)v(t)
P x(t) C(x(t))
:
REFERENCES [1] D. DENDRINOS, Private communication.
[2] T. KIMURA,On the Iteration of Analytic Functions, Funkcialaj Ekvacioj,14 (1971), pp. 197-238.
[3] M. SUZUKI,Holomorphic solutions of some functional equations, Nihonkai Math. J.,5 (1994), pp. 109-114.
[4] M. SUZUKI,Holomorphic solutions of some system of n functional equations with n variables related to difference systems, Aequationes Mathematicae,57 (1999), pp. 21-36.
[5] M. SUZUKI,Holomorphic solutions of some functional equations II, South- east AsianBulletinof Mathematics,24(2000), pp. 85-94.
[6] M. SUZUKI,Difference Equation for a Population Model, Discrete Dynamics inNature and Society,5(2000), pp. 9-18.
[7] A. MATSUMOTO - M. SUZUKI, Analytic Solutions of Nonlinear Cournot- Duopoly Game, Discrete Dynamics in Nature and Society, 2005, issue 2 (2005), pp.119-133.
[8] M. SUZUKI, Holomorphic solutions of a functional equation and their applications to nonlinear second order difference equations, Aequationes mathematicae,74(2007), pp. 7-25.
[9] M. SUZUKI,Analytic General Solutions of Nonlinear Difference Equations, Annali di Matematica Pura ed Applicata,187(2008), pp. 353-368.
[10] N. YANAGIHARA,Meromorphic solutions of some difference equations, Funk- cialaj Ekvacioj,23(1980), pp. 309-326.
Manoscritto pervenuto in redazione il 7 dicembre 2007