On a two dimensional dynamical system generated by the floor function
Usmonov J.B
J. B. Usmonov
Institute of mathematics,
81, M.Ulugbek str., 100125, Tashkent, Uzbekistan.
[email protected]
Abstract.
In this paper we investigate the two dimensional dynamical system generated by the floor function with a parameter λ∈R.
We describe all limit points of the dynamical system depending on λ and on the initial point.
Key words and phrases:
Dynamical systems; floor function; fixed point.
2010 Mathematics Subject Classification:
37E05.
1. Introduction
Let X⊂R2 and A be an operator from X to itself. The main problem of the theory of dynamical systems is to study iterations of A : if An denotes the n-fold composition (iteration) of A with itself, then for a given point x one investigates the sequence x,Ax,A2x,A3x, and so on. This sequence is called two-dimensional discrete time dynamical system or the forward orbit of x, or just the orbit of x for short ([1], [3]).
Definition 1**.**
A point z∈R2 is called a fixed point of A operator if A(z)=z. The set of all fixed points is denoted by Fix(A).
Definition 2**.**
The point z∈R2 is a periodic point of period n if An(z)=z and An−1(z)=z. We denote the set of periodic points of period n by Pern(A). The set of all iterates of a periodic point form a periodic orbit.
For a given operator A:R2→R2 the ω-limit set of z∈R2, denoted by ω(z,A) or ω(z), is the set of limit points of the forward orbit {An(z)}n∈N. Hence, y∈ω(z) if and only if there is a strictly increasing sequence of natural numbers {nk}k∈N such that Ank(z)→y as k→∞.
In this paper we will study the dynamical system generated by operator
[TABLE]
defined by
[TABLE]
where λ∈R is parameter and ⌊x⌋ denotes the integer part of x.
In our case the dynamical system is
[TABLE]
The main problem is to investigate the following limit
[TABLE]
for any z∈R2.
2. The main results
2.1. Fixed points
The following lemma gives all fixed points of this operator.
Lemma 1**.**
For the set of fixed points the followings hold:
If λ<0(λ=−1), then Fix(A)={(0,0)};
If λ=−1, then Fix(A)={(m,−m)∣m∈Z};
If mm−1<λ≤m+1m for some m∈N, then
[TABLE]
If λ=1, then Fix(A)={(m,m)∣m∈Z};
If mm+1≤λ<m−1m for some m∈N, then
[TABLE]
Proof. For finding fixed points of the operator we need solve A(z)=z, i.e., that system \left\{\begin{array}[]{ll}x=\lfloor\lambda y\rfloor&\\
y=\lfloor\lambda x\rfloor.&\end{array}\right.
By expressing y we get the equation ⌊λ⌊λx⌋⌋=x. Roots of ⌊λx⌋=x are also roots of ⌊λ⌊λx⌋⌋=x. That’s why fixed points given in parts 1-5 of Lemma took by solving ⌊λx⌋=x (see e.g. [2]). But there may be some fixed points of g(x)=⌊λ⌊λx⌋⌋ those are roots of ⌊λ⌊λx⌋⌋=x and are not roots of ⌊λx⌋=x. We shall prove that fixed points of operator are just roots of ⌊λx⌋=x.
In [2] all limit points ω(x) of the floor function f(x)=⌊λx⌋ found for ∀λ,x∈R. It was proved that the limit of {fn(x)} converges: to fixed points; to ∞, or -∞(case λ=−1 exception). For the case λ=−1 it was proved that Per2(f)=Z.
By Per2(f)⊂Fix(g) we have part 2 of Lemma. Because of {gn(x)}={f2n(x)}, if {fn(x)} is convergent then {gn(x)} also is convergent, if {fn(x)} converges to ∞ then {gn(x)} also converges to ∞. Thus fixed points of g(x) for λ=−1 consist of roots of ⌊λx⌋=x only. Now we prove part 2, x=⌊−y⌋=⌊−⌊−x⌋⌋=x.
2.2. The limit points
Now we shall describe the set ω(z) for each given z∈R2.
2.2.1. The case λ≤0.
Theorem 1**.**
If λ<0, then the dynamical system generated by operator A has the following properties:
If −1<λ<0, then
[TABLE]
for all z∈R2.
If λ=−1, then each pairs of integer numbers has period two and
[TABLE]
If λ<−1, then A(z)=0 for z∈U∣λ∣1−(0)={(x,y)∣λ1<x≤0,λ1<y≤0} and
[TABLE]
where
R++2={(x,y)∣x,y∈R,x>0,y>0}, R−−2={(x,y) ∣ x,y∈R,x<0,y<0},
R+−2={(x,y) ∣ x,y∈R,x>0,y<0}, R−+2={(x,y) ∣ x,y∈R,x<0,y>0}.
Proof. (1) Let −1<λ<0. For all z=(x0,y0)∈R2 we have
[TABLE]
We can separate sequence (2.1) to two subsequences {un} and {vn}:
[TABLE]
[TABLE]
Let’s write several terms of those sequences,
[TABLE]
[TABLE]
In Theorem 2 of [2] was proved limn→∞un=0 for any λ∈(−1,0) and for all x0∈R.
That’s why we have
[TABLE]
(2) If λ=−1, then A2(z)=z and A(z)=z for all z∈Z×Z. Thus each pairs of integer numbers has period two.
If z∈R2∖Z×Z, then A(z)∈Z×Z. So we have
[TABLE]
(3) Let λ<−1 and z=(x0,y0)∈U∣λ∣1−(0). Then we have ⌊λx0⌋=0 and ⌊λy0⌋=0, i.e., A(z)=0. If x0≤λ1 then for sequence (2.2) we have
[TABLE]
for all z∈R2∖U∣λ∣1−(0) and, if x0>0 then
[TABLE]
Since {∣un∣}n≥1⊂N and limn→∞∣un∣=∞ the following hold
[TABLE]
[TABLE]
2.2.2. The case 0<λ<1.
Note that for each λ∈(0,1) there exists m∈N such that mm−1<λ≤m+1m.
Theorem 2**.**
Let mm−1<λ≤m+1m for some m∈N. Then the following hold:
If z∈{(x0,y0)∣x0≥0,y0≥0}, then
[TABLE]
If z∈{(x0,y0)∣λk≤x0<λk+1,y0≥0}⋃{(x0,y0)∣λk≤y0<λk+1,x0≥0}, then
[TABLE]
where k∈{−1,−2,...,−m}.
If z∈{(x0,y0)∣x0<λ−m,y0≥0}⋃{(x0,y0)∣y0<λ−m,x0≥0}, then
[TABLE]
If z∈{(x0,y0)∣x0<λ−m+1,y0<λ−m+1}, then
[TABLE]
If z∈{(x0,y0)∣λk≤x0<λk+1,λp≤y0<λp+1}, then
[TABLE]
where k∈{−1,−2,...,−m},p∈{−1,−2,...,−m}.
If z∈{(x0,y0)∣x0<λ−m+1,λk≤y0<λk+1}⋃{(x0,y0)∣λk≤x0<λk+1,y0<λ−m+1}, then
[TABLE]
where k∈{−1,−2,...,−m}.
Proof. (1) For subsequences (2.2) and (2.3) of (2.1) we have
[TABLE]
[TABLE]
for all z=(x0,y0)∈R++2, i.e. limn→∞un=limn→∞vn=0. Then limn→∞An(z)=(0,0).
(2) If z∈{(x0,y0) ∣ λk≤y0<λk+1,y0≥0} then ⌊λx0⌋=k, where k∈{−1,−2,...,−m}. Since k is a fixed point of f(x)=⌊λx⌋, then limn→∞un=k and by proof of 1st part we have limn→∞vn=0 for y0≥0.
In case ∀z∈{(x0,y0)∣λk≤y0<λk+1,x0≥0} we can write limn→∞un=0 and limn→∞vn=k as above. That means, ω(z)={(k,0),(0,k)}.
(3) u1<−m and u1>u0 are true for all z∈{(x0,y0)∣x0<λ−m,y0≥0} . For terms of (2) we see that un+1>un, i.e. {un} is an increasing sequence, which is bounded from above by −m. Since −m is the unique fixed point of f(x)=⌊λx⌋ in (−∞,−m], we have limn→∞un=−m. For {vn} we have limn→∞vn=0.
In case z∈{(x0,y0)∣y0<λ−m,x0≥0} we can write limn→∞un=0 and limn→∞vn=−m, then ω(z)={(−m,0),(0,−m)}.
(4) Like the proof of 3rd part we may write
[TABLE]
for all z∈{(x0,y0)∣x0<λ−m+1,y0<λ−m+1}.
(5) If z∈{(x0,y0)∣λk≤x0<λk+1,λp≤x0<λp+1} then ⌊λx0⌋=k,⌊λy0⌋=p. Since k and p are fixed points of f(x)=⌊λx⌋ we have
[TABLE]
where k∈{−1,−2,...,−m},p∈{−1,−2,...,−m}.
(6) The proof is based on parts 3-4.
2.2.3. The case λ≥1.
In case λ=1, the form of operator A is A(z)=\left\{\begin{array}[]{ll}x^{\prime}=\lfloor y\rfloor&\hbox{}\\
y^{\prime}=\lfloor x\rfloor&\hbox{}\end{array}\right. and
[TABLE]
for all z=(x,y)∈R2.
Theorem 3**.**
Let mm+1≤λ<m−1m for some m∈N. Then the following hold:
If z∈{(x0,y0)∣λk≤x0<λk+1,y0<0}⋃{(x0,y0)∣λk≤y0<λk+1,x0<0}, then
[TABLE]
where k∈{0,1,2,...,m−1}.
If z∈{(x0,y0)∣x0<0,y0≥λm}⋃{(x0,y0)∣y0<0,x0≥λm}, then
[TABLE]
If z∈{(x0,y0)∣λk≤x0<λk+1,λp≤x0<λp+1}, then
[TABLE]
where k∈{0,1,2,...,m−1},p∈{0,1,2,...,m−1}.
If z∈{(x0,y0)∣λk≤x0<λk+1,y0≥λm}⋃{(x0,y0)∣λk≤y0<λk+1,x0≥λm}, then
[TABLE]
where k∈{0,1,2,...,m−1}.
If z∈{(x0,y0)∣x0<0,y0<0}, then
[TABLE]
If z∈{(x0,y0)∣x0≥λm,y0≥λm}, then
[TABLE]
Proof. (1) Since z∈{(x0,y0)∣λk≤x0<λk+1,y0<0} then limn→∞un=k.
y0>λy0≥⌊λy0⌋=f(y0) because y0<0 and λy0<0 for λ>1. Using this inequality we get fn(y0)>fn+1(y0)(yn>yn+1). Due to lack of fixed points of f(x)=⌊λx⌋ in (−∞,0), we have limn→∞fn(y0)=limn→∞vn=−∞.
In this case also ∀z∈{(x0,y0)∣λk≤y0<λk+1,x0<0}, we may write limn→∞vn=k and limn→∞un=−∞ as above. Thus ω(z)={(k,−∞),(−∞,k)}.
(2) We showed that limn→∞un=−∞ is true for all z∈{(x0,y0)∣x0<0,y0≥λm} in part 2. If y0≥λm, we have vn<vn+1 and {vn} is an increasing sequence, that’s why
limn→∞vn=∞. In case when z∈{(x0,y0)∣y0<0,x0≥λm} we get limn→∞un=∞ and limn→∞vn=−∞, i.e. ω(z)={(−∞,∞),(∞,−∞)}.
(3) If z∈{(x0,y0)∣λk≤x0<λk+1,λp≤x0<λp+1} then ⌊λx0⌋=k and ⌊λy0⌋=p. Since k and p are fixed points of f(x)=⌊λx⌋ , we get limn→∞un=k,limn→∞vn=p⇒ω(z)={(k,p),(p,k)}.
Proofs of parts 4-6 are directly come from above results.