ALEXANDRU MIHAIL
The theory of fractal sets is an old one, but it also is a modern domain of re- search. One of the main source of the development of the theory of fractal sets is the discovery of a new types of fractal sets. One of the well-known examples are the attractors of iterated function systems. There are several generalization of iterated functions systems: directed-graph iterated function systems, random iterated function systems and so on. The aim of the paper is to present a gener- alization of iterated function systems (IFS) and of their attractors, the recurrent iterated function systems.
AMS 2000 Subject Classification: 28A80.
Key words: recurrent iterated function system, fractal set, Lipschitz function, con- traction, autosimilar set.
1. INTRODUCTION
In this section we review concepts, notation and basic results concerning iterated function systems (IFS). Details can be found in [1], [2] or [3].
Let (X, dX) and (Y, dY) be two metric spaces andC(X, Y) the set of con- tinuous functions fromXin toY. OnC(X, Y) we will consider the generalized metric d:C(X, Y)×C(X, Y)→R+=R+∪ {+∞}defined by
d(f, g) = sup
x∈X
dY(f(x), g(x)).
For a sequence (fn)nof elements fromC(X, Y) andf ∈C(X, Y),fn→s f denotes pointwise convergence,fnu.c
→f uniform convergence on compact sets, and fn
→u f uniform convergence (that is, convergence in the generalized metric d).
For a function f : X → X we denote by Lip(f) ∈ R+ the Lipschitz constant associated with f, that is,
Lip(f) = sup
x,y∈X;x6=y
d(f(x), f(y)) d(x, y) .
A function f is said to be a Lipschitz function if Lip(f) < +∞ and a contraction if Lip(f)<1.
REV. ROUMAINE MATH. PURES APPL.,53(2008),1, 43–53
Let
Con(X) ={f :X→X |Lip(f)<1}.
Let (X, d) be a metric space andK(X) the set of nonvoid compact subsets of X. ThenK(X) with the Hausdorff-Pompeiu distanceh:K(X)×K(X)→ [0,∞) defined by
h(E, F) = max(d(E, F), d(F, E)) = min{r |E⊂B(F, r) and F ⊂B(E, r)}, where d(E, F) = sup
x∈E
d(x, F) = sup
x∈E
( inf
y∈F d(x, y)),is a metric space.
(K(X), h) is a complete metric space if (X, d) is a complete metric space.
(K(X), h) is compact if (X, d) is compact (see [4]).
The Hausdorff-Pompeiu distance can be extended to a semidistance on P∗(X) with values inR+ with the same definition, where P∗(X) is the set of nonvoid subsets of X.
Concerning the Hausdorff-Pompeiu distance we have the following im- portant properties.
Proposition 1.1. Let (X, d) be a metric space,f :X→X a Lipschitz function and K(X) the set of compact subsets of X. If (Hi)i∈I and (Ki)i∈I
are two families of subsets of X,then 1)h(S
i∈I
Hi, S
i∈I
Ki) =h(S
i∈I
Hi,S
i∈I
Ki)≤sup
i∈I
h(Hi, Ki);
2)h(f(K), f(H))≤Lip(f)h(K, H) for any K, H ∈K(X).
An iterated function system, (IFS for short) consists of a metric space (X, d) and a finite family of contractions fk : X → X, k ∈ {1,2, . . . , n}, n∈N∗.We denote such an IFS byS = (X,(fk)k=1,n). Its contraction constant is cS = max
k=1,n
Lip(fk).
There are three important mathematical results concerning contractions used in the IFS theory.
Theorem 1.1 (Banach Contraction Principle). Let (X, dX) be a com- plete metric space and f ∈Con(X). Then there exists a unique α ∈X such that f(α) =α. Moreover, lim
n→∞dX(f[n](x), α) = 0 for any x∈X.
Theorem 1.2(Continuity Theorem). Let (X, dX) be a complete metric space, f, g∈Con(X), α the fixed point of f and β the fixed point of g. Then
dX(α, β)≤d(f, g) 1
1−min{Lip(f),Lip(g)}.
Theorem 1.3 (Collage Theorem). Let (X, dX) be a complete metric space, f ∈ Con(X) and x ∈ X. Let α ∈ X be the unique fixed point of f.
Then
dX(x, α)≤ dX(f(x), x) 1−Lip(f).
With an IFS S = (X,(fk)k=1,n) with contraction constant cS = max
k=1,n
Lip(fk) one can associate the function FS : K(X) → K(X) defined by FS(B) =
n
S
k=1
fk(B), B ∈ K(X), which is a cS-contraction. Therefore, by the Banach contraction principle, we have
Theorem 1.4. Let (X, dX) be a complete metric space and S = (X,(fk)k=1,n)an IFS with cS = max
k=1,n
Lip(fk)<1. Then there exists a unique set A(S),such that FS(A(S)) =A(S), and for every H0∈K(X)the sequence (Hn)n, where Hn+1 =FS(Hn), n∈N, is convergent to A(S) in K(X). As for the speed of convergence we have
h(Hn, A(S))≤ cnS 1−cS
h(H0, H1), for all n∈N. In particular,
h(H0, A(S))≤ 1 1−cS
h(H0, H1).
We also have the following result concerning the continuity of the attrac- tor of an IFS.
Theorem 1.5. Let (X, dX) be a complete metric space and x∈X. Let S = (X,(fk)k=1,n) and S0 = (X,(gk)k=1,n) be two iterated function systems with contraction constants cS and cS0. Then
h(A(S), A(S0))≤ max
k=1,n
d(fk, gk) 1
1−min(cS, cS0).
We present now the idea of our generalization. Instead of considering contractions f from X in to X in the definition of an IFS, we shall con- sider contractions from X2 = X ×X in to X. We shall call recurrent it- erated function systems (RIFS for short) this kind of iterated function sys- tems. On X2 we consider the distance d2 : X2 ×X2 → [0,+∞) given by d2((x, x1),(y, y1)) = max{d(x, y), d(x1, y1)}.
Definition 1.1. Let (X, d) be a complete metric space. For a function f :X×X→Xthe number
Lip(f) = sup
x,y,x1,y1∈X;x6=yorx16=y1
d(f(x, x1), f(y, y1)) max{d(x, y), d(x1, y1)}
is called the Lipschitz constant of f. We have Lip(f)∈[0,+∞].
A functionf :X×X →Xis said to be a Lipschitz function if Lip(f)<∞ and a contraction if Lip(f)<1. Let
Con2(X) ={f :X×X→X |Lip(f)<1}.
Definition 1.2. Let (X, d) be a complete metric space. A recurrent iterated function systems (RIFS for short) on X consists of a finite family of contractions (fk)k=1,n,n∈N, with fk :X×X→ X, and is denoted byS = (X,(fk)k=1,n). The contraction constant of such an RIFS iscS = max
k=1,n
Lip(fk).
2. MAIN RESULT
In this section we prove the existence of the attractor of an RIFS and also present its main properties.
Definition 2.1. Let (X, d) be a metric space and f : X ×X → X a continuous function. The function Ff :K(X)×K(X)→K(X) defined by
Ff(K, H) =f(K, H) =f(K×H) ={f(x, y)|x∈K and y∈H}
is called the set function associated with f.
LetS= (X,(fk)k=1,n) be an RIFS. The functionFS :K(X)×K(X)→ K(X) defined byFS(K, H) =
n
S
k=1
fk(K, H) is called the set function associated with S.
The next result is obvious.
Lemma 2.1. If (X, dX) is a complete metric space fn, f : X×X → Xare continuous functions, n ∈ N, such that fn
→u f and H, K ∈ K(X), then lim
n→∞fn(H, K) =f(H, K) in K(X).
The next result is a consequence of Lemma 2.1 and Proposition 1.1.
Lemma 2.2. Let (X, dX) be a complete metric space. Let Sm = (X,(fkm)k=1,n) and S = (X,(fk)k=1,n), m ∈ N∗, be recurrent iterated func- tion systems. If fkm→u fk on a dense set in X as m→+∞,k∈ {1,2, . . . , n}, then lim
n→∞FSn(K, H) =FS(K, H) for every H, K ∈K(X).
Lemma 2.3.Let f :X×X →Xbe a Lipschitz function. Then Lip(Ff)≤ Lip(f).
Proof. Let K, H, K1, H1 ∈ K(X) and r = max(h(K, K1), h(H, H1)).
Then K ⊂ B(K1, r) and H ⊂ B(H1, r). Let x ∈ K and y ∈ H. There are
x1∈K1 and y1 ∈H1 such that d(x, x1)< rand d(y, y1)< r. Then d(f(x1, y1), f(x, y))≤Lip(f) max{d(x, x1), d(y, y1)}<Lip(f)r.
This means thatf(x, y)∈B(f(x1, y1),Lip(f)r)⊂B(f(K1, H1),Lip(f)r) which implies thatf(K, H)⊂B(f(K1, H1),Lip(f)r).
In a similar way, we obtain the inclusionf(K1, H1)⊂B(f(K, H),Lip(f)r).
This means that h(f(K, H), f(K1, H1))≤Lip(f)r.
The next result is a key one.
Lemma 2.4.Let S= (X,(fk)k=1,n)be an RIFS with cS= max
k=1,n
Lip(fk)<
<1.Then the set function FS :K(X)×K(X)→K(X) associated with S, is a cS contraction.
Proof. We have
h(FS(K, H), FS(K1, H1)) =h n
[
k=1
fk(K, H),
n
[
k=1
fk(K1, H1)
≤
≤ max
k=1,n
h(fk(K, H), fk(K1, H1))≤cSmax(h(K, K1), h(H, H1)).
For a contraction of typef :X×X →X we have results similar to those of type f :X →X. We shall sketch the proof and present some applications.
Theorem 2.1(Banach Contraction Principle). Let (X, d) be a complete metric space and f ∈Con2(X)with contractivity factor c∈[0,1). Then there exists a unique α ∈X such that f(α, α) = α. Moreover, for any x0, x1 ∈X the sequence (xn)n≥0 defined by xn+1=f(xn, xn−1),n∈N∗,is convergent to α. As for the speed of convergence of (xn)n≥0 we have
d(xn, α)≤ 2c[n/2]
1−c max{d(x0, x1), d(x1, x2)}.
In particular,
d(x0, α)≤ 2
1−cmax{d(x0, x1), d(x1, x2)}.
Proof. Letzn= max{d(xn+1, xn), d(xn, xn−1)}. We have d(xn+1, xn) =d(f(xn, xn−1), f(xn−1, xn−2))≤
≤cmax{d(xn−1, xn−2), d(xn, xn−1)}=czn−1
and
d(xn+2, xn+1) =d(f(xn+1, xn), f(xn, xn−1))≤
≤cmax{d(xn+1, xn), d(xn, xn−1)} ≤cmax{czn−1, zn−1}=czn−1. Then zn+1≤czn−1.It follows that
d(xn+1, xn)≤c[n/2]z1
and
d(xn+p, xn)≤d(xn+p, xn+p−1) +d(xn+p−1, xn+p−2) +· · ·+d(xn+1, xn)≤
≤c[(n+p−1)/2]
z1+c[(n+p−2)/2]
z1+· · ·+c[n/2]z1, hence
d(xn+p, xn)≤ 2c[n/2]z1
1−c for every p∈N∗.
The rest of the proof goes in the same manner as that for the Banach contraction principle.
Using Lemma 2.4 in the case of an RIFS, Theorem 2.1 becomes
Theorem 2.2. Let (X, dX) be a complete metric space and S = (X,(fk)k=1,n)an RIFS with contraction constant cS = max
k=1,n
Lip(fk) < 1.
Then there exists a unique A(S)∈K(X) such that FS(A(S), A(S)) =A(S).
Moreover, for any H0, H1 ∈ K(X), the sequence (Hn)n≥1 defined by Hn+1 = FS(Hn, Hn−1), n ∈ N∗, is convergent to A(S). As for the speed of convergence we have
h(Hn, A(S))≤ 2c[n/2]S 1−cS
max{h(H0, H1), h(H2, H1)}.
In particular (Collage Theorem), h(H0, A(S))≤ 2
1−cS
max{h(H0, H1), h(H2, H1)}.
We remark that A(S) =
n
[
k=1
fk(A(S), A(S)) =
n
[
k=1
[
x∈A(S)
fk({x}, A(S)) =
=
n
[
k=1
[
x∈A(S)
fk(A(S),{x}) =
n
[
k=1
[
x,y∈A(S)
fk({x},{y}).
Theorem 2.3(Continuity Theorem). Let (X, dX) be a complete metric space and f, g∈Con2(X)with fixed points α and β and contraction constants cf and cg. Then
dX(α, β)≤minn 1
1−cfdX(f(β, β), g(β, β)), 1
1−cgdX(f(α, α), g(α, α))o
≤
≤ 1
1−min(cf, cg)d(f, g).
Proof. We have
dX(α, β) =dX(f(α, α), g(β, β))≤dX(f(α, α), f(β, β)) +dX(f(β, β), g(β, β))≤
≤cfdX(α, β) +dX(f(β, β), g(β, β))≤cfdX(α, β) +d(f, g), whence
dX(α, β)≤ 1
1−cfdX(f(β, β), g(β, β))≤ 1
1−cf d(f, g).
Similarly, we have
dX(α, β)≤ 1
1−cgdX(f(α, α), g(α, α))≤ 1
1−cgd(f, g) and the conclusion follows.
Using Lemma 2.4 in the case of an RIFS, Theorem 2.3 becomes
Theorem 2.4(Continuity Theorem). Let (X, dX) be a complete metric space, and S = (X,(fk)k=1,n) and S0 = (X,(gk)k=1,n)be two RIFS with c=
max
k=1,n
Lip(fk)<1 and c0 = max
k=1,n
Lip(gk)<1. Then h(A(S), A(S0))≤ 1
1−min(c, c0) max
k=1,n
d(fk, gk).
Another version of the continuity theorem is
Theorem 2.5. Let (X, dX) be a complete metric space and a function f :X×X→X. Let us consider fn∈Con2(X), where n∈N, with fixed point αn. Assume c = sup
n≥1
cfn < 1, where cfn = Lip(fn), and fn
→s f on a dense set in X. Then f ∈ Con2(X), Lip(f) ≤ c and if α is the fixed point of f, then lim
n→∞ αn=α.
Proof. It follows from the hypotheses that fnu.c→ f on X×X and cf = Lip(f)≤c <1. Hencef is a contraction. By Theorem 2.3 we have
dX(α, αn)≤ 1
1−c dX(fn(α, α), f(α, α)).
Using Lemma 2.4 in the case of an RIFS, Theorem 2.5 becomes
Theorem 2.6. Let (X, dX) be a complete metric space, n ∈ N∗, and Sm = (X,(fkm)k=1,n), m ∈ N∗, and S = (X,(fk)k=1,n) recurrent iterated function systems with contraction constants cn and c0. If sup
n≥1
cn=c <1and fkm →s fk on a dense set in X×X, k∈ {1,2, . . . , n}, then lim
n→∞A(Sn) =A(S).
3. EXAMPLES AND REMARKS
Let (X, dX) be a complete metric space, S = (X,(fk)k=1,n) an RIFS, A(S)∈K(X) the attractor ofS and (Hn)n∈N∗ a sequence defined byHn+1= FS(Hn, Hn−1), n∈N∗, which is convergent to A(S). We make some remarks on the monotony of the sequence (Hn)n∈N∗.
IfH0 ⊂H1 ⊂H2, thenHn−1 ⊂Hn, for all n∈N∗, andA(S) = S
n≥1
Hn. In particular, if H⊂FS(H, H) then H⊂A(S).
Let A = {x0 | there exists k such that fk(x0, x0) = x0} be the set of fixed points of the functions (fk)k=1,n. ThenA⊂f(A, A), and so A⊂A(S).
In other words, every fixed point of a function from S is in A(S).
IfH0⊃H1 ⊃H2, thenHn−1 ⊃Hn, for all n∈N∗, and A(S) = T
n≥1
Hn. Proposition 3.1. Let (X, dX) be a complete metric space, S = (X,(fk)k=1,n) a recurrent iterated function system with contraction constant c,and H∈K(X). If H⊂B(FS(H, H), r) then H⊂B A(S),1−cr
.
Proof. Let (Hn)n∈N∗ be the sequence defined by Hn+1 = FS(Hn, Hn), n ∈ N∗, and H0 = H. As in Lemmas 2.3 and 2.4, we have by induction that Hn ⊂ B(Hn+1, rcn), so H ⊂ B
Hn, r
n−1
P
k=0
ck
. Since Hn → A(S) and
n→∞limr
n−1
P
k=0
ck = 1−cr , we get H⊂B(A(S),1−cr ).
Proposition 3.2. Let (X, dX) be a complete metric space, S = (X,(fk)k=1,n) an RIFS , x1, x2, . . . , xl, y1, y2, . . . , ym ∈A(S) and t:{1,2, . . . , l+m} → {1,2, . . . , n} a fixed function. Let S1 = (X,(gk)k=1,l+m)be the IFS defined by
gk(x) =
( ft(k)(xk, x) if k≤n ft(k)(x, yk−n) if k≥n+ 1.
Then A(S1)⊂A(S).
Proof. We have gk(A(S)) = ft(k)(xk, A(S)) ⊂ A(S) if k ≤ n and gk(A(S)) =ft(k)(A(S), yk−n)⊂A(S) if k≥n+ 1. Hence
FS1(A(S)) =
l+m
[
k=1
gk(A(S))⊂A(S).
It follows that
A(S1)⊂A(S).
Let (X, dX) be a metric space and a function. f :X×X →X. Let us consider the function ∆(f) : X → X defined by ∆(f)(x) = f(x, x), x ∈ X.
For an RIFS S = (X,(fk)k=1,n), let us define ∆(S) = (X,(∆(fk))k=1,n). IfS has contraction constantc, then ∆(S) has a contraction constant less than c.
Proposition 3.3. Let (X, dX) be a complete metric space and S = (X,(fk)k=1,n)an RIFS . Then A(∆(S))⊂A(S).
Proof. It is easy to see that F∆(S)(H)⊂FS(H, H) for everyH ∈K(X).
Indeed
F∆(S)(H) =
n
[
k=1
∆(fk)(H) =
n
[
k=1
{f(x, x)|x∈H}=
=
n
[
k=1
{f(x, y)|x∈H andy ∈H}=
n
[
k=1
fk(H, H) =FS(H, H).
Also, forH, K ∈K(X) such thatH ⊂Kwe haveF∆(S)(H)⊂FS(K, K) since F∆(S)(H)⊂F∆(S)(K)⊂FS(K, K).
LetH ∈K(X) be fixed. Consider the sequences (Hn)n∈N∗and (Kn)n∈N∗
defined by H0 =H, K0 =H, Hn+1 =FS(Hn, Hn),and F∆(S)(Kn) =Kn+1, where n∈N.
It is clear that K0 ⊂H0. IfKn⊂Hn then
Kn+1 =F∆(S)(Kn)⊂FS(Hn, Hn) =Hn+1, for all n∈N. Since lim
n→∞Kn=A(∆(S)) and lim
n→∞Hn=A(S), we have A(∆(S))⊂A(S).
Example 1. Every IFS can be seen as an RIFS. Indeed, let S = (X,(fk)k=1,n) be an IFS. Let S = (X,(fk)k=1,n) be the RIFS defined by fk(x, y) =fk(x). Then A(S) =A(S).
Example 2. LetX =Rand let f, g:R×R→Rbe defined byf(x, y) =
x
3 +y3 and g(x, y) = x3 +23.
Let the RIFS S = (R,(f, g)), and let FS : K(R)×K(R)→K(R) be de- fined by FS(K, H) = f(K, H)∪g(K, H).The attractor of S is [0,1]. Indeed f([0,1],[0,1]) = [0,2/3]⊂[0,1] andg([0,1],[0,1]) = [2/3,1]⊂[0,1],so that
FS([0,1],[0,1]) = [0,1].
It follows that A(S) = [0,1].
Example 3. LetX =Rand let f, g:R×R→Rbe defined byf(x, y) =
x
5 +y5 and g(x, y) = x5 +y5 +35.
Let the RIFSS = (R,(f, g)), and letF :K(R)×K(R)→K(R) be defined by FS(K, H) = f(K, H)∪g(K, H). The attractor of S is [0,2/5]∪[3/5,1].
Indeed f([0,1],[0,1]) = [0,2/5] ⊂ [0,1] and g([0,1],[0,1]) = [3/5,1] ⊂ [0,1], so that
FS([0,1],[0,1]) = [0,2/5]∪[3/5,1]⊂[0,1].
It follows thatA(S)⊂[0,1] andA(S) = T
n≥1
Hn, where (Hn)n≥1is the sequence defined by Hn+1=f(Hn, Hn), n∈N, and H0= [0,1].
We remark thatf(x, y) =f(y, x),g(x, y) =g(y, x),F(K, H) =F(H, K), g(x, y) = f(x, y) + 35 and g(K, H) = f(K, H) + 35. Let also t : R→Rbe defined by t(x) = −x+ 12. Then (t◦t)(x) = x, f(t(x), t(y)) = t(g(x, y)), g(t(x), t(y)) =t(f(x, y)) andt(A(S)) =A(S). In other words the attractor of the RIFS S = (R,(f, g)) is symmetric with respect to 1/2. Then
H1 =FS([0,1],[0,1]) = [0,2/5]∪[3/5,1], f([0,2/5]∪[3/5,1],[0,2/5]∪[3/5,1]) =
=f([0,2/5],[0,2/5])∪f([0,2/5],[3/5,1])∪f([3/5,1],[3/5,1]) = [0,4/25]∪[3/25,7/25]∪[6/25,2/5] = [0,2/5],
g([0,2/5]∪[3/5,1]) = [3/5,1],
H2 =F(H1, H1) =f(H1, H1)∪g(H1, H1) = [0,2/5]∪[3/5,1] =H1. It follows that
A(S) =H1= [0,2/5]∪[3/5,1].
We remark that A(S) cannot be the attractor of a classical IFS with two injective functions. Such an attractor, if disconnected, must be totally disconnected. However, A(S) is the attractor of a classical IFS with four functions, namely f0 :R→R defined by
f0(x) = min{|x|,2/5}
2 ,
f1 :R→R defined by
f1(x) = min{2/5− |x−2/5|,2/5}
2 +4
5, f2 :R→R defined by
f2(x) = min{3/5 +|x−3/5|,1}
2 +4
5 and f3 :R→R defined by
f3(x) = min{1− |x−1|,2/5}
2 +1
2.
Example 4. Let X = R and f0, f1 : R2→R2be defined by f0(x, y) = (x3,0) and f1(x, y) = (x3 +23,0).
Let the IFS S = (R,(f0, f1)) and A(S) = C the Cantor set. Let C0 = f0(C) and C1 = f1(C). Let us note that C0 ⊂ C, C1 ⊂ C, f0(C0) ⊂ C0 and f1(C1) ⊂ C1. Consider the RIFS S = (R,(f0, f1)), where f0 : R2×R2→R2andf1 :R2×R2→R2 are given byf0((x1, x2),(y1, y2)) = (x31,y31) and f1((x1, x2),(y1, y2)) = (x31 +23,0).Then A(S) =C0×C0∪C1× {0}.
Indeed, since
f0(C0×C0∪C1× {0}) =f0(C0×C0)∪f0(C1× {0}) =
=f0(C0)×f0(C0)∪f0(C1)× {0} ⊂C0×C0∪C1× {0}
and
f1(C0×C0∪C1× {0}) =f1(C0×C0)∪f1(C1× {0}) =
=f1(C0)× {0} ∪f1(C1)× {0}=C1× {0}, we have A(S)⊂C0×C0∪C1× {0}.
Since f0((x1, x2),(0,0)) =f0(x1, x2) and f1((x1, x2),(0,0)) =f1(x1, x2), we get C⊂A(S). Also,f0(C×C) =C0×C0 and f1(C×C) =C1× {0}. It follows that
C0×C0∪C1× {0} ⊂A(S), which completes the proof.
REFERENCES
[1] M.F. Barnsley,Fractals Everywhere. Academic Press, New York, 1998.
[2] K.J. Falconer,The Geometry of Fractal Sets. Cambridge Univ. Press, Cambridge, 1985.
[3] K.J. Falconer,Fractal Geometry-Foundations and Applications.Wiley, 1990.
[4] N Secelean,M˘asur˘a ¸si fractali. Ed. Univ. Sibiu, 2002.
Received 1st March 2006 University of Bucharest
Faculty of Mathematics and Computer Science Str. Academiei 14
010014 Bucharest, Romania mihail alex@yahoo.com