Boundary value problem of a third-order mixed type differential-difference equation

Boundary value problem of a third-order mixed type differential-difference equation

li NOg~- HOJAND B o u n d a r y V a l u e P r o b l e m of a Third-Order M i x e d T y p e Differential-Difference E q u a t i o n Wang Peiguang Dep...

370KB Sizes 0 Downloads 30 Views

li NOg~- HOJAND

B o u n d a r y V a l u e P r o b l e m of a Third-Order M i x e d T y p e Differential-Difference E q u a t i o n Wang Peiguang

Department of Mathematics Hebei University Baoding, 071002 People', 's Republic of China

ABSTRACT In tlhis paper, the existence and uniqueness of the solution for the boundary value problem of a third order mixed differential-difference equation is given by using Schauder's fLxed point theorem, and an approximate solution is given by using the Picard's iterative methods. © Elsevier Science Inc., 1996

1.

INTRODUCTION

Due to the applications of the differential-difference equations in the fields of biological mathematics, mechanics, and control theory, people are getting more and more interested in the boundary-value problem. Wang and Wang [1, 2] have discussed the existence and uniqueness of solutions for second-order equation and second-order mixed boundary-value problems, respectively, and have given the estimation of the error between the solution and the approximate solution. In this paper, by using Schander's fixed point theorem, the same problem is studied for third-order mixed differentialdifference equations. Wang and Luo [3] present a special case of the problem discussed in this paper.

APPLIED MATHEMATICS AND COMPUTATION 80:273-286 (1996) © Elsevier Science Inc., 1996 0096-3003//96/$15.00 655 Avenue of the Americas, New York, NY 10010 SSDI 0096-3003(95)00303-7

274

W. PEIGUANG

We consider the boundary value problem V"( t) - cl y ' ( t - ~ ) - ~ y ' ( t + ~ ) ---f(t;y(t),~l(t-~-l),~i(t+~-2) y( t) = ¢b( t)

a-

y( t) = ~ ( t)

b <.< t <. b + ~

)

(a
(1.1)

~'1 <~ t <~ a, y'( a + O) = m (1.2)

where ~(t)= (y(t), y'(t), y"(t)); ~(t + A ) = ( y ( t + A), y'(t + A), y"(t + h)), ( h = - ~'i, ~'2), ~'i, ~'2 are positive numbers, ¢ ~ C2([ a - ~'1, a], R), 0 ~ C~([ b, b + ~'2], R) and f is continuous in t, and assume

Ic~l DEFINITION 1.

+

Ic21 <

(1.3)

1.

A function y(t) is called a solution of (1.1)-(1.2), if (i)

y(t) ~ C [ a - vl, b + ~2] ~ C:[ a, b], (ii) y"(t) on [a,b] has only some limited discontinuous points of the first kind, (iii) y(t) satisfies the condition (1, 2), (iv) Equation (1.1) holds for t ~ [ a, b], and t ~ a + k~-t, t ~: b -

k~'2 (k ~ Z+). For convenience, we make a similar change of the variable as [3]

z( t) = a( t)dp( t) + ~( t)~( t) + ~/( t) + y( t).

(1.4)

Then the problem in transformed into

z'( t) - clz"( t - ~'1) - c2z"( t + "r2) --F(t;~(t),~(t-~'l),~.(t+~'2) ) z(t)=0

t~[a-Tl,

a]U[b,b+r2] ,

(a
(1.5)

z'(a+0)=0,

(1.6)

where ~.(t) = (z(t), z'(t), z"(t)); z(t + A) = ( z ( t + A), z'(t + A), z"(t + A)), (A

-~- - - T1 , '/'2) ,

275

Third-Order Differential-Difference Equation

Constructing the Green function of problems (1.5)-(1.6) gives - ( ( t - a)( b - s ) / ( b - a ) ) 2 / 2 + ( t -

c ( t, s) =

I

a
s)2/2

a))2/2

0

a <~ t <. s <~ b a-'q
0

a-71<~s<~aorb<.s
Using the Green function, we can obtain the following lemma.

LEMMA 1 [3].

Let x(t) ~ K = { h(t); h ' ( t) on [ a, b] has only some limited discontinuous points of the first, kind, h(t) = 0, t ~ [ a - ~'1, a] U [b, b + r2], h ' ( a + 0) ~ 0}. Then Ix(k)(t)l < C3, k(b - a ) 3 - k M ( k = 0,1,2) in which M = maxa~ t< b] Y(3)(t)l, C3,o = 2/81, C3,1 = 1/6, C3, 2 = 2/3.

LEMMA 2. Let Q = maxnlFI, (t; ~,(t), ~,(t - ~'1), ~(t + ~'2)) e 12 [ a, b] × R 9. If z( t) is a solution of (1.5)-(1.6) then max

Iz'( t)] ~ Q/(1 - t c 1] -]c2])

c

(1.7)

a~ t~ b

LEMMA 3 [4]. Let B be a Banach space and let r > 0, r ~ R, S(x o, r) = { x e B: 1 i x - xol] < r}. Map S( x o, r) into B and (i) for all x, y ~ S(x o, r): i l T x - TyOi <~ a t l x (ii) r o = ( 1 ko) -1 × l I T x o - x o l ] ~ < r.

yll, (0 ~< a < 1),

Then

(1) T has a fixed point x* in S( Xo, ro); (2) x* is the unique fixed point of T in S( x o, r); (3) the sequence{xm} defined by xm+ 1 = Tx m. m = O, 1 ... converges to x* with ILx* - xmll < k~ro. 2.

E X I S T E N C E AND UNIQUENESS

Now we give the existence and uniqueness of the solution to the problem (1.5)-(1.6).

276

W. PEIGUANG

THEOREM 1.

Suppose that

(i) k i > 0 (i = 0, 1, 2) are given constants, l-I is a closed and bounded subset of [a, b] × C9[a, b] and IFI <~ Q, in which Q = max iFI,

a = {a <~ t <~ b, z ( t ) ~ C[a - vl, b + ~'2] N C2[a, b], ]z(i)(t)l ~< ki, I z ( ° ( t - T1) [ <

ki,

]z(i)(t-[-

T2)I <

ki};

(ii) the following inequality holds b - a < min{(cko/QC3,o) i/s, ( ck~/QC a ,),/2, ( c k j Q C 3 , 2 ) } ( c = 1 - I c , I - Ic21)

Then the boundary value problems (1.5)-(1.6) have at least one solution.

PROOF. Denote E = {z(t): z(t) ~ C[a - ~'1, b + r2] A C2[a, b], z(t) = O, t ~ [ a - T1, a] [,.J [b, b + ~'2]; z'(a + O) = O, maxa~t~blz(3)(t) ] ~< kj, j = 0, 1, 2, z ' ( t ) on [a, b] has only some limited discontinuous points of the first kind}; then E is a closed convex subset of the Banach space C2[ a, b]. We define an operator T as follows.

Tz(t) --/b÷~2 G(t, s){F(s; ~(s), ~ ( s - ~,), ~(s + ~2)) a--~- 1

+ [ e l z ' ( s - ~,) + c2z'(s + ~2)1} ds (a-

T, ~< t~< b + r2)

(2.1)

By the condition (ii) and L e m m a 1 and L e m m a 2, we have

]( Tz)(')( t)l < ( QC3 ,( b -

a)S-')/c

~ k,.

It is easy to see T maps E onto itself and is completely continuous. It then follows from Schauder's fixed point theorem that T has a fixed point z*(t). The z*(t) is a solution of (1.5)-(1.6).

Third-Order Differential-Difference Equation

277

COROLLARY. Assume the function F(t; ~(t), ~(t - ~'1), ~(t + r2)) on [a, b] × R 9 satisfying, 2

[El ~ h --~ E {hj+l[ z(J)( t)l °t{j) "+- lj+l[ z(J)( t - T1)I fl(j) j=0 + m.,I z(J)(t + ~2)I V(j)

(2.2)

where h, hi+l, lj+l, mj+ 1 are nonnegative constants, 0 ~ a ( j ) <. 1,0 ~ ( j ) < 1, 0 ~< T(J) ~< 1. Then the boundary value problem (1.5)-(1.6) has at least one solution, provided that

{(

O=

E {hi+l+ /j+l +

j=O

PROOF.

mj+l}Ca,j(b-

a) a-j

)/

c

< 1

(2.3)

Let

Zj+,(.(j)) = { ohi+ 1 ~j+l(~(J)) = { 0lj+~ mj+l(f~(J)) ---- {

ol(j) = 1 o<.(j)
mj+ l

v( J)

0

0 < ~(j) <1.

=

1

Using Theorem 1, we have 2

Q ~ h-~ E {(hj+l --hj+l)ka(J) -[- (lj+l

-

-

-lj+l ) k~(J)

j=o

"I'-(mj+ 1 -- -mj+l)k "y(j)} -{- I( ko, kl, k2) (2,4)

278

W. PEIGUANG

in which 2

I(k0, kl, ks) = 2 {~j+l + ~j+l + ~+~)kj.

(2.5)

j=O

Taking

kj = c a , / ( q , o ( ~ -

a) j)

( j = o, 1,2),

then the following inequality holds

I( ko, k1, ks)

Similarly, we have

I( ko, kl, k2) .

-hi+ 1 +-Ij+ 1 +Nj+,}Ca, i ( b - a ) a-j k a ( 0 3 , 1 ( b - a ) 2)

I( ko, kl, k~)

when a ( j ) = fl(j) = y(j) = 1, the coefficient of k "(j), k t~(j) and k ~'(j) are equal to zero; then by choosing kj sufficiently large and using (2.3), we have

Q<~ cko/( C3,o( b - a)3), q<~ ck2/( ca 2( b -

Q<~ ckt/( Ca,,( b - a)2),

~) ).

(2.6)

Third- Order Differential-Difference Equation

279

From (2.0, it is easy to get

b-

a < min{(ck0/QC3,0)l/3), (ckl/QC3,1)1/2), (ck2/QC3,2))};

(2.7)

Theorem 1 conditions are satisfied, so the corollary holds.

REMARK. Theorem 1 is a local existence theorem, whereas a corollary does not require any condition on the length of the interval.

THEOREM 2.

Assume the function F on [ a, b] × R 9 satisfies the follow-

ing condition 2

IFI < h + E {hj+ 11 zCJ)(t)[ + zj÷ll z(J (t- 1)1 ÷ mj÷ll z'J'(t ÷ j=O

(2.8) Then problems (1.5)-(1.6) have one solution, provided that

PROOF. From (2.8), we can easily get that z(t) -~ 0 is a solution of the problem (1.5)-(1.6); in order to prove the theorem, we only need to prove problems (1.5)-(1.6) have no nontrivial solutions. Supposing the problems (1.5)-(1.6) have nontrivial solutions, using Lemma 1, we have

IZ"(t)l ~

[2

E {hj+l "~- lj+l "~- mj+l}C3, j( b - a) 3-j j=o

1

max b I z"( t) l × a<~
280

W. PEIGUANG

then

max I z"( t)l < o a ~ t<~ b

Iz"( t)l a ~ t<~ b

by 0 < 0 < 1, we can get m a x ~ t~ b ]Z"(t)l = 0, then z ' ( t ) = 0. But this means t h a t z(t) is a nonzero linear function. This is impossible because we have z(a) = z(b) = 0, z ' ( a + 0) = 0 from the condition (1.6); therefore,

z(t) - O.

THEOREM 3. Let f on [ a, b] × R 9 be differential with respect to y(J) ( j = 0, 1, 2), and satisfy

l a f / 3 y(J)( t)l <~ hi+ 1, ] ~ f / 3 y(J)( t - T1)I

< lj+l, IOf/OY(J)(t + ~2)1 <~ mj+l. Then the boundary value problems (1.1)-(1.2) have a unique solution, provided that

hi+ 1 + lj+ l + mj+ 1} C3, j( b - a) 3-j

0 =

c < 1.

PROOF. Suppose yl(t), y2(t) are solutions of (1.1)-(1.2), respectively. Let z(t) = Yl(t) - Y2(t),

IOf/c~Y(i)( t)] = fj, [Of/OY(J)( t - ~'1)1 = gj, [Of/c~Y(J)( t + ¢2)J = Pj; then z(t) satisfies zm(t)

-- C1 Z m ( t -- T1) -- C2 z m ( t "~- '1"2)

---- f ( t ; ~ l ( t ) , ~ll(t -- T1) , ~ l ( t + 72) ) - f ( t ; ~lz(t), ~12(t + T1) ,

~2(t + a'2) )

Third-Order Differential-Difference Equation

281

and z(t) = 0 t ~ [a - ~-, a] U [b, b + ~'2] z'(a + O) = O. Let

aj(t) = fol fj(t; ~2(t) + s z ( t ) , 7]2(t - "Q), ~2(t + ~'2)) ds bj( t )

= £ 1 gj(t; ~:(t), ~:(t- ~) + ~z(t- ~), ~ ( t + ~)) d~

e.( t) -~- fo~pj(t; Y2(t), y 2 ( t -

T1) , ~]2( t Jr" T2) + sz( t-~- 'T2) ) ds.

Then ~ " ( t ) - ~ z " ( t - ~-,) - c~ z " ( t + ~-~)

2 E {aj(t) z(2(t) + bj(t) z(J)( t - "Q) + ej(t) z(J)(t + "r2)}. j=O Noticing [aj(t)] <<.hi+l, Jbj(t)] <~ lj+ 1, lej(t)l ~< mj+l, then the Theorem 2 condition is satisfied; therefore z(t) -- 0. 3.

PICARD'S ITERATIONS

We know it is very difficult to obtain the solutions of differential-difference equations, so we now introduce the approximate solution and discuss an estimation of the error between the approximate solution and the solution. Because problems (1.1)-(1.2) are equivalent to (1.5)-(1.6), we can discuss problems (1.5)-(1.6).

DEFINTION 2. A function z(t) is called an approximate solution of problems (1.5)-(1.6), if there exist e(6 > 0), such that

max I z " ( t ) - c l z " ( t a<~t<~b

.Q) - c 2 z " ( t + 72)

- F ( t; -~( t), -z( t - ~,), -z( t + ~-~))] < ~

(a.1)

282

W. PEIGUANG

and z( t) ~ 0 t e [ a - 71, a] U [ b, b + ~'2] z'( a + 0) - 0. In fact, the approximate solution z(t) can be expressed as

z( t) = fab+'~ a( t, s){F(s; ~( s), ~( s - r~), ~( s + r2) "{'[ClZm(8 -

T1) +

in which rl(t) = z"(t) - q z " ( t T1) , 7z(t + 72)).

DEFINITION 3.

C 2 Z m ( 8 "["

72) ] + TI(8)}

d8

(3,2)

71) - c2z"(t + r~) + F(t; ~(t), ~ ( t -

The function f is said to be of Lipschitz class for all (t; ~ ( t ) , ~(t - r , ) , ~(t + T2)),

(t; ~( t), 7( t -

71), 7( t + ~'2)) e [ a, b] x D, ( D e

Rg).

The following inequality is satisfied f( t; ~ ( t ) , ~( t - T1) , "~( t "4" T2) ) -- f(t; 7 ( t ) , 7( t - r l ) , 7( t + r2))l

2 <~ E {hj+llu'J)( t) - v(J)( t)l + li+llu(J) ( t -

"rl) - v(J)( t -

71) ]

~'~)

7 )1}

j~ 0

+%+,lu(J)(t

+

-

v(')(t

+

m which hi+ 1, lj+ 1, mj+ 1 are nonnegative constants. Denote R,, = {z(t): z(t) ~ C[a - rl, b + 72 ] c~ C2[a, b]; z(t) = O, t [ a - rl, a ] U [ b , b + r 2], z ' ( a + 0 ) = 0 z ' ( t ) on [a,b] has only some limited discontinuous point of the first kind} For z(t) E R,,, we define the norm max Iz'J)(t)l, Hzll = ()
(C3,o(b-a)J/c3

'

j)max.z(')(t a,~t
It is easy to verify that R,, is a Banach space.

"Q)I, +'r2). }.

283

Third-Order Differential.Difference Equation

For Ra, we can obtain by simple calculus the following lemma. LEMMA 4.

For any x ~ R a, define an operator

(3.3)

To = f b + "2 G( t, s) x " ( s - vl) ds. a--T 1

T h e n l T o z ( t ) l < Aollxll, IT; z(t)l < Al(b - a)llxJl, T~' x(t)l <~ A2(b - a)2ll xll in which A o = 3, A 1 = 81/4, A 2 = 81/2.

LEMMA 5.

For any x ~ R a, define an operator

(3.4)

T1 = fb+ "2 G( t, s) x " ( s + "2) ds. Ca

-T

1

Then l Tl x( t)l • Aoli zll, IT~ x( t)l

~

AI( b -

a)ll xll, IT~' x( t)l

<<.A2( b

-

a)211xll

in which A o = 3, A 1 = 81/4, A 2 = 81/2.

THEOREM 4. With respect to problems (1.5)-(1.6), we assume that there exists an approximate solution z( t), and (i) the F is of Lipschitz class on [ a, b] × D, in which D = { ( ~ ( t ) , ~( t - ~'1), u( t + ~'2)): lu(J)(t) - z(J)( t)l <~ g c 3 , / C 3 , o ( b - a) j, lu(J)( t -

r~) - z(J)( t -

~'1)1

<<-NC3, J C a , o ( b -

a) j, ]u(J)( t + ~'2) - z(J)( t + ~'2)1

<<. NC3, j/C3,o( b -

a)J};

(ii) 2

0

-- ~: {h~+l ÷ lj+l ÷ mj+~}C3, j(b - a)3-¢÷ 3(Icll + Ic2i) < 1; j=0

(iii) ,e(1 -- 0)-1 Ca,o (b - a) 3 ~< N

(e ~ max I~(t)l). a~t.~b

284

W. PEIGUANG

Then (1) there exists a solution z*( t) of (1.5)-(1.6) in S( z, N); (2) z*(t) is the unique of (1.5)-(1.6) in S(z, N); (3) the sequence { zm( t)} converges to z*( t),

in which Zm+ l( t ) = fab+~ G( t, s){F( s; ~m( s), $m( S - 7.1), ~m(S + ~'2)) q-[ ClZ~( S -- T1) Jr C 2 4 ( S

Zo= Z(t )

"4- T2)]} d,8

m= 0,1,2...

and has

IIz*(t)

-

zmll ~< OmNo •

( N O = (1 --

o)-'llz,

- =o11).

PROOF. For mapping T, using Theorem 1 we can get T to map R a to itself, and let z(t) ~ S(z, N). By the norm of R~, we have (Y~(t), ~(t 7.1), Y~(t+ z2)) ~ D. Further, if z(t), y(t) ~ S( z, N), using Lemma 1 and Lemma 2, we have

I T(J)x(t) - T(J)y( t)l <<. C3, y( b - a) 3-j max IF(t; ~( t), ~c( t as~ t ~ b

~ ] ( t ) , ~](t - T1) , y ( t + 7./'2))1

-F(t;

+

le~l

+ Ic21

7.1), ~( t + 7.2))

fb+,2a_ .l

G( t, s)[ ~ttt( s - 7.1) - ~/t'( s - 7"1) ] d8

fb+'~G(t'a-~-i s)[ z"(s

+ 7.2) - y"(s + 7.2)]

dsI

2

<~ C3, y( b- a) a-j max

E {hj+llx(J)( t) - y(Y)(t)l

a
+ lj+ 11 x(J)( t +

+ lclllfb+'~ a-~, G( t' s)[ ~"( s -

xO)(t

7.1) +

-y"(s-7.1)1

Y(J)( t -

y('(t +

dsj

7.1)1

285

Third-Order Differential-Difference Equation

+ I~:t f"+'~ c(t, ,)[ ~"(, + ~-~) - ¢ ' ( , + ~-~)1 '~1 a--'r 1

2

max

E

{hj+l + lj+l -4- mj+l}C3, j ( b -

a
+ Aj(ICll +

a)3-J]]x - yH

Ic21)llz- yll.

Hence, ( C3,o( b - a)J/C3, j)l T(J)x(t) - T (j) y( t)l 2

-< E {hj+l + lj+l + mj+l}c~,j(b - a)~-Jllx - yll j=O

+ At(Ca,o(b - a)J/c3, j)Ocl] + Ic2[)llx- yll.

Similarly, we have (C3,0(b- a)J/c3, j)IT(J)x( t - ~1) - T(J)Y( t - ~'1)1 2

-< E {h~+~ + I~+1 + mj+l}c~ j(~ - a)3-Jll~-

Yll

j=O -~- Aj( C3,o( b - a) J// c3, j)(Ic11 + Ic21)ll x - yll

( C3,0( b - a)J/C3, j)LT(J)x( t + "r2) - T(J)Y( t + ~2)] 2 ~'< E {hj+l + lj+l -4- mj+l}C3, j(b - a)3-JHx - y]]

j=O

+ Aj(C3,o(b - a)J/c3, j)(]cl] + ]c2])Hx- yH.

W. PEIGUANG

286 Therefore,

IITx(~) - Ty(t)ll < O l l x - yll.

(4.3)

Further, from (3.2), we have

Tz( t) - z( t) = f b+~2 G( t, s)~?( s) ds. a-1"

(4.4)

1

Using Lemma 1, we obtain

[T(J)z( t) - z(J)( t)[ < 8C3, j( b - a)3-J; hence,

( C3,o( b - a)J/Ca, j)lT(J)z( t) - z(J)( t)[ <~ 8 C 3 , o ( b - a) 3. Similarly, we have

( Ca, o( b - a)J/Ca, i)tT(J)z( t - ~'1) - z(J)( t - T1)] ~ 8C3,o( b - a) 3 ( C3, o( b - a)J/C3, j)[T(i)z( t + v2) - z(i)( t + ~2)1 < eC3,o( b - a) 3 then

IITz( t) - z(t)ll < 6 C 3 , o ( b -

a) 3

by condition (iii) ( 1 - 0 ) - l l l T z ( t ) - z(t)[[ < N. Thus, the conditions of Lemma 3 are satisfied and the conclusions (1)-(3) follow. REFERENCES 1 H.Z. Wang, Iterative Methods for Differential-difference Equation, (in Chinese) J. Jilin University, 3:31-42 (1988). 2 L.L. Wang, On Boundary Value Problem and its Singular Perturbation Solutions for a Mixed Differential-Difference Equations, Chinese Anhui University, Master Thesis, 1990. 3 P.G. Wang and Y. L. Luo, Iterative Methods for the boundary value problem of a third order differential-difference equation, Ann. of Diff. Eqs. 11(1):88-94 (1995). 4 L. B. Rall, Computational Solutions of Nonlinear Operator, John Wiley, New York, 1969.