ROCZNIKI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO Séria I : PRACE MATEMATYCZNE X V II (1974)
S
tefanC
zeetwtk(Katowice)
On some properties oî solutions oî a functional equation with parameter
1. In the present paper we are concerned with the functional equation
(1) <p [f ( x ), t] = g {x, t) <p {x, t) + F {x, t) ,
where <p(x, t) is an unknown function and f(x ), g(x, t), F (x , t) are known real functions of a real variables and ^ is a real parameter.
We shall prove that under some assumptions concerning the given functions (in case (C) — cf. [3], and p. 336) the solution cp(x, t) of equation (1) is of class Cr, 1 < r < oo, with respect to the parameter t.
2. We write f°(x) = x, f n+1(x) = f [ f n(x)], n — 0, 1 , . . . We assume the following hypotheses analogous to the hypotheses in [3] :
(I) The functions g(x, t), F (x , t) are continuous in A =
df<a , b )x T and g(x, t) Ф 0 in A0 = (а, Ъ) x T , T — interval.
dfThe function f(x ) is continuous and strictly increasing in an interval
<■a , b) and a < f(x ) < x in {a, b).
(П)
(11Г m7 d°g dvF
There exist the derivatives -^-(&, t), -~v- (x, t), v = 1, 2, . tinuous in A.
r, con-
(1УГ) For every v = 1 , . . . , r and for every closed interval <a, /?> с T there exist: interval {a, a + rjf) c: <a, b), r\v > 0, function B v(x, t) bounded in <a , a + rjvy X <a , /?> and a constant 0 < в < 1, such that the inequalities
dvg
di < B v(x, t), BV[ M , Ï] < 0Bv(x, t)
hold in <a , a + rjfy X <a, /?>.
(Vr) For every v = 1, ..., r and for every closed interval (a , /9) a T there exist: internal (a , a
qv}
cz(a , b),
q,„ > 0, a function B v(x, t) bounded in (a , a + Qvy X <a, /9) and constant 0 < в < 1, such that the inequalities
d v
~ Ж {ас’ г) < JDv(æ, t), A ,№ ) , t] < 6Dv{x, t) holds in <a, a -f
qv}
x<a, /?>.
(YI) There exists a function c(t) = c (c — constant) such that for every closed interval (a , /?> с T exist an interval {a, a + d) a (a , 6), d > 0 , a function В (x, t) continuous with respect to the variable t in <a, /9) and bounded in (a , a + x <a, /?) mid a constant 0 < 6 < 1 such that the inequalities
(2) \F{x, $)| < B (x , t), В [f(x ), f] < 6B{x, t)
hold in <a , a-\-d} x (a , /9>, wüere .F(
æ, t) = F (x , t) + c(t)[g(x, t) — 1].
We put
71 — 1
(3) On(x ,t) = f ] g [ f v(x ),t],
V = 0
(4) G(x, t) = limG„(æ, $).
П— >oo
There are the following three possibilities (i is fixed) :
(A) The limit (4) exists, G(x, t) is continuous and G{x, t) Ф 0 in (a , b).
(B) There exists an interval I <= (a, b) such that limGn(x, t) = 0 n— >oo
uniformly in I.
(C) Neither (A) nor (B) occurs.
3. Now we shall prove the following
T
heorem1. Suppose that hypotheses (I), (II), (III1)-(Y 1), (YI) are fulfilled. If, moreover, g(x, t ) ^ 1 in A and for every te T case (C) occurs, then there exists exactly one solution of equation (1) in A of class C1 with respect to the parameter t in A.
P ro o f. On account of [2] there exists exactly one function <p(x,t) satisfying equation (1) and continuous in A. Let te T be fixed and e > 0 be such that for \h\ < e is t + h t T . The function <p(x,t-\-h) satisfies the equation
(5)
+ = g{x, t-hh)<p{x, t + h) + F (x , t + h).
By (5) and (1) we obtain
<P U iæ)i t + h ] - ( p [f(x ), f\ = g {x, t + b) <p {x, t + h) — g (x, t) cp {x?t) + + F ( x ,t + Ti) — F (x ,t).
According to the Hadamard’s lemma ([
6], cf. also [4]) we have
<p[f{x), t + Ti]-<p[f{x), t] --= Ф2{х, Ti)[<p{x, t + Ti)-q>{x, «)] + Фх(ж, h)Ti + -\-уг{х, Ti)Ti, where
1
Ф1 = Ф1(х,Т1) --= f {<р(х, t) + 8[<p{x,t + h ) -<р{х, t)]}gt (x ,t + sh )d s, о
1 1
Ф2 = ф2(х,Ть) — f g (x, t + sh) ds, чрг = ip1(x ,h ) = J F t(x, t-\-sh)ds.
ô о
Let. ns write
(p(x, t-\-Ti) — (p{x, t) = ip(x, h), and then we obtain
У [/(я
0> ^) + (Фх + Ух)^?
or? for h Ф
0V [f(x ),h ] , у(я, Л)
Ti Ф,
Л ■+ Фх + Ух*
-г, ,,. w (x.h)
Putting y(x, Ti) = —- , we have fb
(
6) y [ f ( x ), Tl ] = фг{х, h)y(x, h) + Фг(х, ft) + yi(æ, Tl).
Now we shall consider the equation
(7) a[f{x), Ti] = Ф2{х, Ti)a{x, Т
1) + Фх(х, ft) + yi(® , Л).
Since <
7(
æ, i) ^ l , we have
П
—1and hence
(
8
) 103(æ ,*)= / 7
<
1, xe (a , b), \Ti\ < £, w- = 1, 2, ...
G»(®, Л) .
Thus case (B) for equation (7) does not occur for any \h\ < e.
Let {hn} be any sequence such that \hn\ < e and hn -> 0. We have (9) in <a, a + rjoy X <2 — s ,t + s}.
In view of (IV1), (V1) and (9), we obtain
i i
(10) |0>x|< j (L + 2Ls)\gt(x, t + shn)\ d s^ 3 L j B x{x, t + shn)d s= B x{x, hn),
о 0
xe < a , a + ^x),
l l
/ J ^I ds ^ J
Г(t^7j $ “j- sJbyi) ds —
d t •—>(*^j ~^n) ? *1? ^ (л, 6Ь I •
0 0
It is easily seen that
(12) B x[f(x ), hn] < 6B x{x, йя), 2>х[/(я?), hn] < 0£>x(^ hn) in <e, a+£>, where £ = min^x, £x).
According to (10), (11), (
1 2) and [1] for every hn equation (7) has exactly one continuous solution in <я, b) fulfilling condition a (a, hn) =
0. It is given by formula
(13) a (x , hn) = —
@v+l (*^J ^n)
If hn = 0, then Ф
2(а?, 0) = g(x, t) and for equation (7) case (C) occurs.
On account of (10) and (11) it is easily verify that (VI) hold if we assume e(t) — 0 and F{x, 0) = Фх(х, 0) + у х{х1 0). Consequently in view of [1], Theorem 9, the equation
« № ), 0] = g {x, t)a{x,
0) + Фх{x, 0) + tpx(x , 0) has exactly one continuous solution in <a , b) given by the formula
a(x ,
0) = - ^
v= o
Ф ЛГ(х),_0] + щ [Г (х ),
0]
@v+l{®1 ^)
Let de (a, b) be arbitrarily fixed and let N be chosen so that f N(x)e e (a , a + £> for xe {a, d}. Next, we have
Ф хГ Г И ^ З + ухСГИ, K l
^«+i(^? К ) <\ф1 [ Г ( я ) , К 1\ + \ п 1Г (х ),К ]\
< B x[ f ( x ) , h n] + Dx[ r ( x ) , hn]
< Bv- N(Bx[ f N(x), hn]+ D x[ f N(x), hn])
< 6V~N[ sup B x(x ,h )+ sup Dx(x ,h )].
<a,tZ> x <— е,б> <a, й>х<— e,e>
Thus series (13) uniformly converges in <a, d>x< — s, e) and we have
(14) lima (a?, hn) = a (x ,
0).
On the other hand, the function
(15) y(cc, hn) = cp{x, t + hn)-c p (x , t) К
is the continuous solution of equation (7) such that у (a, hn) =
0. Since equation (7) has exactly one continuous solution in (a , b) fulfilling the condition a (a, hn) =
0, we have
a (x f hn) = y(x , Jin)y 7i =
1,
2, ...,
and consequently by (14) and (15) for every t < ? T there exists the derivative
— (x , t) and is continuous in <a, b). dcp
Replacing h by hn in equation (7) and passing to the limit as n -> oo, we obtain
(16) dcp dcp
-ft *1 = 9 (®> *) ~ft *) + dg_
dt (x, t)cp(x, t) dF dt (x, t).
In view of (
1 0) and (
1 1) it is easily verify that for equation (16) (YI) holds if we assume c(t) = 0 and F (x , t) = gt(x, t)cp{x, t) + F t{x, t). Hence, according to the condition g{x, t) >
1and [
2] equation (16) has exactly one solution continuous in A. Therefore the function dcp (x, t) is continuous in A, which was to be proved. 01
T
heorem2. Let the assumptioTis of Theorem 1 be fulfilled (instead of (III'H Y1) we assume (ПГ)-(УГ), 1 < r < oo). Then there exists exactly one solution of equation (1) in A of class Cr with respect to the parameter t in A.
The proof is analogous to the proof of Theorem 1 and Theorem 5 in [4], and is therefore omitted.
References
[1] B. Choc zew ski and M. K uczm a,
On the “indeterm inate case” in the theory of a lin e a r fu n c tio n a l equation,Fund. Math. 58 (1966), p. 163-175.
[2] S. C zerw ik,
On the dependence on a p aram eter of solutions of a lin e a r fu n c tio n al equation,Prace Naukowe U.&L, Prace Mat. 3 (1973), p. 25-30.
[3] —
Solutions of class Gr w ith respect to the p aram eter of a lin e a r fu n ctio n al equation,this volume, p. 341-345.
[4] —
On the d iffere n tiab ility of solutions o f a lin e a r fu n c tio n al equation w ith respect to the p aram eter,Ann. Polon. Math. 24 (1971), p. 217-225.
[5] I. P ie tro w s k i,
O rd in ary d ifferen tial equations,Warszawa 1953.
3 ~~ Roczniki PTM — Prace M atem atyczne XVII.