Mathematica Pannonica 7
/2 (1996), 223 { 232BOUNDED SOLUTIONS OF
SCHILLING'S PROBLEM
Janusz Morawiec
Instytut Matematyki, Uniwersytet Slaski, ul. Bankowa 14, PL{40{007 Katowice, Poland
Received: August 1995 MSC 1991: 39 B12, 39 B 22
Keywords: Schilling's problem, bounded and continuous functions.
Abstract: Let n be a positive integer, qn be the unique x 2 (1312) with
x n+1
;3x+ 1 = 0 and q 2 (0qn]. We found a set Anq of reals with the following property (P): Every solution f :R!Rof the functional equation
f(q x) = 14qf(x;1) +f(x+ 1) + 2f(x)]
which vanishes outside of ;1;qq 1;qq ] and is bounded in a neighbourhood of a point of that set vanishes everywhere. It is also observed that forq2(013] the setS1n=1Anq, which equals then
n 1
X
n=1
"(n)qn : "2f;101gNo is the largest one with property (P).
Following R. Schilling 9] we consider solutions f : R! Rof the functional equation
(1) f(qx) = 14q f(x;1) +f(x+ 1) + 2f(x) such that
(2) f(x) = 0 for jxj> Q
where q is a xed number from the open interval (01) and
Q= q 1;q:
In what follows any solution f : R! Rof (1) satisfying (2) will be called a solution of Schilling's problem.
(3) If 3q1;p32 +p34
then according to 7] the zero function is the only solution of Schilling's problem which is bounded in a neighbourhood of a point of the set (4) n"Xn
i=1qi : n2 Nf0+1g "2f;11go:
This generalizes in particular 1 Th. 1]. It is the aim of the present paper to obtain such a result with the set (4) replaced by a larger one.
However, we are not able to enlarge (4) for allq's satisfying (3) but, on the other hand, for q 13 we succeeded in nding even the largest set to be put in the place of (4) (cf. Cor. 1).
Given a positive integernand q 2(01) consider the setAnq of all the real numbers of the form
(5)
"XL
l=1(;1)lXKl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm)+M +"(;1)L XM
m=1qm where " 2 f;11g, M, L are non-negative integers, K1:::KL 2
2 f1:::ng, and :f1:::Lgf1:::ng!N. Evidently, the set (4) is a subset of clAnq. Let us observe also that for l1l2 2 f1:::Lg, k(6)K1 2f1:::Kl1g, k22 f1:::Kl2g, if (l1k1)6= (l2k2) then
l1
X
m=k1(l1m) + XL
j=l1+1 Kj
X
m=1(jm)6= KXl2
m=k2(l2m) + XL
j=l2+1 Kj
X
m=1(jm): The proof of the following fact is left to the reader (cf. also 6 Th. 21(a), (d)]).
Remark 1.
If q 2(0 13] then cl1n=1Anq =nX1
n=1"(n)qn : "2f;101gNo and if q 2131) then
n 1
X
n=1"(n)qn : "2f;101gNo= ;QQ]:
For every positive integer n let qn denote the unique x 2 (1312) with(7) xn+1;3x+ 1 = 0
and observe that ifq 2(0 12) then
qqn i qn+1;3q+ 1 0: Our main result reads.
Theorem 1.
If n is a positive integer and q 2 (0qn] then the zero function is the only solution of Schilling's problem which is bounded in a neighbourhood of a point of the set clAnq.The proof of this theorem is based on four lemmas. However, we start with the following simple remarks.
Remark 2.
If f is a solution of Schilling's problem then so is the func- tion g:R!Rdened by the formulag(x) =f(;x).Remark 3.
Assume f is a solution of Schilling's problem.If q 6= 14 then f(;Q) =f(Q) = 0. If q < 12 then f(0) = 0.
Lemma 1.
Assume q2(012). Iff is a solution of Schilling's problem then(8) f(qN+Mx+"XM
m=1qm) =1 2
M 1 2q
N+Mf(x)
for all x 2 (Q;11;Q) (for all x 2 Q;11;Q] if q 6= 14), for all
"2f;11g, and for all non-negative integers M and N.
Forx 2(Q;11;Q) this was proved in 7] as Lemma 2. In the case of the closed interval Q;11;Q] and q 6= 14 we argue similarly as in the proof of 7 Lemma 2] using also 7 Remarks 1 and 2(i)].
Lemma 2.
Let n2N,q 2(0qn] and y =qN+PLl=1PKlk=1(lk)x++XL
l=1(;1)lXKl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm)
whereN is a non-negative integer, Lis a positive integer, K1:::KL2
2f1:::ng, : f1:::Lgf1:::ng! N, and x 2 01;Q) (x 2
201;Q] if q 6= 14).
If L is even then y2(01;Q).
If L is odd then y 2Q;10] (y 2(Q;10] if q < 13).
Proof.
Since q qn< 12 we have(9) Q <1:
Moreover, asqn is a solution of (7), (10) Xn
i=1qi Xn
i=1qin= 1; qn
1;qn 1; q
1;q = 1;Q and
(11) if q < 13 then Xn
i=1qi < Q <1;Q:
Observe also that
(12)
y =q(LKL)qN+PLl=1PKlk=1(lk);(LKL)x+ +LX;1
l=1(;1)lXKl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm);(LKL)+ + (;1)L KXL;1
k=1 qPKLm=;1k (Lm) + (;1)L
(13)
y=qPLl=1PKlk=1(lk);qNx;1;
;
K1
X
k=2qPKm1=k(1m)+PLj=2PKjm=1(jm)+ +XL
l=2(;1)l KXl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm) (14) qPLl=1PKlk=1(lk);qNx;1<0
and
(15) XKL
k=1qPKLm=k(Lm) XKL
k=1qPKLm=k1 Xn
k=1qk:
Suppose rst L is even. Applying (13), (14), (6), (15), (9) and (10) we obtain
y <XL
l=2(;1)lXKl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm)
LX;2 l=2
Kl
X
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm);
;
KXL;1
k=1 qPKLm=;1k (L;1m)+PKLm=1(Lm)+XKL
k=1qPKLm=k(Lm)
1
X
i=1q(L;1KL;1)+PKLm=1(Lm)+i;
;q(L;1KL;1)+PKLm=1(Lm) +Xn
k=1qk =
=q(L;1KL;1)+PKLm=1(Lm)(Q;1) +Xn
k=1qk <Xn
k=1qk 1;Q whereas (12), (6) and (9) give
y q(LKL);X1
i=1qi+ 1=q(LKL)(;Q+ 1)>0:
Suppose now L is odd. If L = 1 then using the denition of y, (15) and (10) we see that
y ;XK1
k=1qPKm1=k(1m) ;Xn
k=1qk Q;1
with the last inequality being strict ifq < 13 (cf. (11)). If L 3 then on account of the denition of y, (6), (15), (9) and (10) we have
y ;LX;2
l=1 Kl
X
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm)+ +KXL;1
k=1 qPKLm=;1k (L;1m)+PKLm=1(Lm) ;XKL
k=1qPKLm=k(Lm)
; 1
X
i=1q(L;1KL;1)+PKLm=1(Lm)+i+
+q(L;1KL;1)+PKLm=1(Lm);XKL
k=1qk
>;Xn
k=1qk Q;1:
Finally, ifL is odd then taking into account (12) and (6) we obtain yq(LKL)x+X1
i=1qi;1q(LKL)(1;Q) +Q;1] = 0:
Lemma 3.
Assume n 2 N and q 2 (0qn]. If f is a solution of Schilling's problem then for every x201;Q), for every non-negative integers MLand N, for every K1:::KL 2 f1:::ng, and for every : f1:::Lgf1:::ng!Nwe have(16)
f(qN+PLl=1PKlk=1(lk)+Mx+ +XL
l=1(;1)lXKl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm)+M+ + (;1)L MX
m=1qm) =
=1 2
PLl=1Kl+M 1 2q
N+PLl=1PKlk=1(lk)+Mf(x):
Proof.
According to Lemma 1, (16) holds for L = 0. Assume L is a positive integer.Consider rst the case M = 0.
Let L= 1. Equality (16) takes then the form (17) f(qN+PKk=11 (1k)x;XK1
k=1qPKm1=k(1m)) =
=1 2
K1 1 2q
N+PKk=11 (1k)f(x)
and making use of Lemma 1 we see that if K1 = 0 then (17) holds for all x 2 (Q;11; Q) (for all x 2 Q; 11;Q] if q 6= 14) and for every non-negative integer N. Fix now a K1 2 f0:::n;1g and suppose that (17) is satised for every non-negative integerN, for every
: f1gf1:::ng!N, and for allx201;Q) (for allx201;Q] if q6= 14). Let N 2Nf0g, :f1gf1:::ng!Nand x201;Q) (x 201;Q] ifq 6= 14). Putting
z=qN+PKk=11 (1k)x;XK1
k=1qPKm1=k(1m) ;1 we have
(18) zx;1
and, according to Lemma 2, y :=qz 2 Q;10] (and y 2 (Q;10] if q < 13). This jointly with the denition of z, Lemma 1, (1), (17), (2), Remark 3 and (17) gives
f(qN+PKk=11+1(1k)x;KX1+1
k=1 qPKm1=+1k (1m)) =
= f(q(1K1+1);1y) == 1 2q
(1K1+1);1f(y) =
= 1 2q
(1K1+1);1 1
4qf(z;1) +f(z+ 1) + 2f(z)] =
= 1 2q
(1K1+1)1
2f(z+1) =
= 1 2q
(1K1+1)1 2
1 2
K1 1 2q
N+PKk=11 (1k)f(x) =
= 1 2
K1+1 1 2q
N+PKk=11+1(1k)f(x):
Hence (17) holds for every K1 2 f1:::ng, for every non-negative integerN, for every :f1gf1:::ng!N, and for allx 201;Q) (for all x 2 01 ;Q] if q 6= 14). Consequently, taking into account Remark 2 we have also
(19) f(qN+PKk=11 (1k)x+XK1
k=1qPKm1=k(1m)) =
=1 2
K1 1 2q
N+PKk=11 (1k)f(x)
for every K1 2 f1:::ng, for every non-negative integerN, : f1g
f1:::ng ! N, and for all x 2 (Q;10] (for all x 2 Q;10] if q 6= 14).
Fixnow a positiveintegerLand suppose that (16) holds withM =
= 0 for every K1:::KL 2 f1:::ng, for every non-negative integer N, : f1:::Lgf1:::ng ! N, and for all x 2 01;Q) (for all x 201;Q] if q 6= 14). Dening y as in Lemma 2 and making use of Lemma 2, (17) and (19) with x replaced byy, and (16) withM = 0 we obtain
f(qN+PLl=1+1PKlk=1(lk)x+ +LX+1
l=1(;1)lXKl
k=1qPKlm=k(lm)+PLj=+1l+1PKjm=1(jm)) =
=f(qPKLk=1+1(L+1k)y+ (;1)L+1KXL+1
k=1 qPKLm=+1k (L+1m)) =
=1 2
KL+1 1 2q
PKLk=1+1(L+1k)f(y) =
=1 2
KL+1 1 2q
PKLk=1+1(L+1k)1 2
PLl=1Kl
1
2q N
+
PLl=1PKlk=1(lk)
f(x) =
=1 2
PLl=1+1Kl 1 2q
N+PLl=1+1PKlk=1(lk)f(x): This ends the proof of (16) in the case where M = 0.
IfM is a positive integer then dening once moreyas in Lemma2 and making use of this lemma, (8) with N = 0 and x replaced by y, and (16) with M = 0 we get
f(qN+PLl=1PKlk=1(lk)+Mx+ +XL
l=1(;1)lXKl
k=1qPKlm=k(lm)+PLj=l+1PKjm=1(jm)+M + (;1)L MX
m=1qm) =
=fqMy+ (;1)L MX
m=1qm=1 2
M 1 2q
Mf(y) =
=1 2
PLl=1Kl+M 1 2q
N+PLl=1PKlk=1(lk)+Mf(x):
The fourth lemma is just 7 Lemma 1].
Lemma 4.
Assume q 2 (012). If a solution of Schilling's problem vanishes either on the interval (;q0) or on the interval (0q) then it vanishes everywhere.Proof of Theorem 1.
Suppose f is a solution of Schilling's problem bounded in a neighbourhood of a point x0 2 clAnq. We may (and we do) assume that x0 is of the form (5), where " 2 f;11g, ML are non-negative integers, K1:::KL 2 f1:::ng, and : f1:::Lgf1:::ng! N. Moreover, according to Remark 2, we may (and we do) assume"= 1.
If x201;Q) is xed then the left-hand side of (16) is bounded with respect to N whereas the right-hand side is bounded if(x) = 0.
This shows thatf vanishes on 01;Q). Hence and from (10) it follows that f vanishes, in particular, on 0q) which jointly with Lemma 4 proves that f vanishes everywhere.
To formulate a corollary accept the following denition.
Denition 1.
Let q 2 (01) and x 2 ;QQ]. We say that x 2 Bq(resp. x 2 Cq) if and only if the zero function is the only solution of Schilling's problem which is bounded in a neighbourhood of x (resp.
continuous at x).
We will use also the following result of W. Forg{Rob cf. 6 The- orems 20, 21, 23{26 and 28] and Remark 1.
If q 2(01) and f is a solution of Schilling's problem then suppf nX1
n=1"(n)qn : "2f;101gNo
and for every q2 (0 13]the Schilling's problem has a nonzero solution.
Corollary 1.
If q2(013] then Bq =Cq =nX1n=1"(n)qn : "2f;101gNo:
Proof.
Obviously Bq Cq, whereas the above quoted result of W.Forg{Rob gives Cq n
1
X
n=1"(n)qn : "2f;101gNo: Moreover, applying Remark 1 and Th. 1 we obtain that
(
1
X
n=1"(n)qn : "2f;101gN
)
Bq:
Applying Lemma 3 (formula (16) with x = 0 and Remark 2) and Remark 3 we obtain also the following result.
Theorem 2.
If n is a positive integer and q2(0qn]then any solution of Schilling's problem vanishes on the set Anq.The reader interested in further results on Schilling's problem is referred to 2] by K. Baron, A. Simon and P. Volkmann, 3] by K. Baron and P. Volkmann, 4] by J. M. Borwein and R. Girgensohn, 5] by G. Derfel and R. Schilling, 6] by W. Forg{Rob and 8].
Acknowledgement.
This research was supported by the Silesian Uni- versity Mathematics Department (Iterative Functional Equations pro- gram).References
1] BARON, K.: On a problem of R. Schilling,Berichte der Mathematisch-statis- tischen Sektion in der Forschungsgesellschaft Joanneum{Graz, Bericht Nr.
286(1988).
2] BARON, K., SIMON, A. et VOLKMANN, P.: Solutions d'une equation fonctionnelle dans l'espace des distributions temperees, Comptes Rendus des Seances de l'Academie des Sciences. Serie I. Mathematiques. (Paris) 319 (1994), 1249{1252.
3] BARON, K. et VOLKMANN, P.: Unicite pour une equation fonctionnelle, Rocznik Naukowo{Dydaktyczny WSP w Krakowie. Prace Matematyczne 13 (1993), 53{56.
4] BORWEIN, J. M. and GIRGENSOHN, R.: Functional equations and distri- bution functions,Results in Mathematics 26(1994), 229{237.
5] DERFEL, G. and SCHILLING, R.: Spatially chaotic congurations and func- tional equations with rescaling, Manuscript.
6] FORG-ROB, W.: On a problem of R. Schilling,Mathematica Pannonica 5/1 (1994), I 29{65, II: 145{168.
7] MORAWIEC, J.: On bounded solutions of a problem of R. Schilling, Annales Mathematicae Silesianae 8(1994), 97{101.
8] MORAWIEC, J.: On continuous solutions of a problem of R. Schilling,Results in Mathematics 27(1995), 381-386.
9] SCHILLING, R.: Spatially chaotic structures, in: H. Thomas (editor), Non- linear Dynamics in Solids, Springer, Berlin, 1992, 213-241.