Chapter X

Iterations of Interval Transformations

[79/94%]
Show
LaTeX
Problem X.1.1

Suppose that for a function f:N→Nf: \mathbb {N} \rightarrow \mathbb {N}, the inequality f(n+1)>f(f(n))f(n+1)>f(f(n)) holds for every n∈Nn \in \mathbb {N}. Prove that f(n)=nf(n)=n for every n∈Nn \in \mathbb {N}.

?
Problem X.1.2

Does there exist a continuous function f:R→Rf: \mathbb {R} \rightarrow \mathbb {R} satisfying the equation f∘f=Ff \circ f=F in the following cases:

?
(a)

F(x)=−xF(x)=-x;

(b)

F(x)=exF(x)=e^{x};

(c)

F(x)=x2−2F(x)=x^{2}-2 ?

Problem X.1.3

Let us denote by IdX\mathrm{Id}_{X} (or simply Id) the identity map of a set XX onto itself (i.e., Id⁡(x)=x\operatorname {Id}(x)=x for every xx in XX). A map h:X→Xh: X \rightarrow X will be called an involution if h∘h=IdXh \circ h=\mathrm{Id}_{X}.

Let f(x)=x+1,g(x)=x−1(x∈R)f(x)=x+1, g(x)=x-1(x \in \mathbb {R}). Find two involutions hh and kk such that the relations h∘k=f,k∘h=gh \circ k=f, k \circ h=g hold.

?
Problem X.1.4

Let us denote by IdX\mathrm{Id}_{X} (or simply Id) the identity map of a set XX onto itself (i.e., Id⁡(x)=x\operatorname {Id}(x)=x for every xx in XX). A map h:X→Xh: X \rightarrow X will be called an involution if h∘h=IdXh \circ h=\mathrm{Id}_{X}.

Prove that every bijection of a set onto itself is the composition of two involutions.

?
Problem X.1.5

A map h:X→Xh: X \rightarrow X is called an involution if h∘h=IdXh \circ h=\mathrm{Id}_{X}, where IdX\mathrm{Id}_{X} is the identity map of XX. A fixed point of a map ff is a point xx in its domain of definition such that f(x)=xf(x)=x.

Prove that a continuous map f:Δ→Rf: \Delta \rightarrow \mathbb {R}, where Δ\Delta is some interval, has a fixed point in each of the following cases:

?
(a)

Δ=[a,b],f(Δ)⊂Δ\Delta =\left[a, b\right], f(\Delta ) \subset \Delta;

(b)

Δ=[a,b],f(Δ)⊃Δ\Delta =\left[a, b\right], f(\Delta ) \supset \Delta;

(c)

Δ=⟨a,b⟩,f\Delta =\langle a, b\rangle , f is an involution.

Problem X.1.6

A fixed point of a map ff is a point xx in its domain of definition such that f(x)=xf(x)=x. Let a function f∈C([a,b])f \in C(\left[a, b\right]) be increasing, f([a,b])⊂[a,b]f(\left[a, b\right]) \subset [a, b]. Prove that for every x0∈[a,b]x_{0} \in [a, b] the sequence xn=f(xn−1)x_{n}=f\left(x_{n-1}\right) (n∈Nn \in \mathbb {N}) converges to a fixed point. Does this hold if ff is a decreasing function?

?
Problem X.1.7

A fixed point of a map ff is a point xx in its domain of definition such that f(x)=xf(x)=x. Let f∈C(R)f \in C(\mathbb {R}). For x0∈Rx_{0} \in \mathbb {R}, set x1=f(x0),xk=f(xk−1)(k∈N)x_{1}=f\left(x_{0}\right), x_{k}=f\left(x_{k-1}\right) (k \in \mathbb {N}). Prove that if for some x0∈Rx_{0} \in \mathbb {R} the sequence {1n∑0⩽k<nf(xk)}\left\{ \frac{1}{n} \sum_{0 \leqslant k<n} f\left(x_{k}\right)\right\} is bounded, then ff has a fixed point. Does this hold for continuous maps f:C→Cf: \mathbb {C} \rightarrow \mathbb {C}?

?
Problem X.1.8

A fixed point of a map ff is a point xx in its domain of definition such that f(x)=xf(x)=x.

Prove that a finite group of affine transformations of the plane has a fixed point (common to all transformations of the group).

?
Problem X.1.9

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map. It is assumed that the maps are continuous, defined on intervals, and map them into themselves.

Prove that the relation of topological conjugacy is an equivalence relation, i.e.

?
(a)

f∼ff \sim f;

(b)

f∼gf \sim g if and only if g∼fg \sim f;

(c)

if f1∼f2,f2∼f3f_{1} \sim f_{2}, f_{2} \sim f_{3}, then f1∼f3f_{1} \sim f_{3}.

Problem X.1.10

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map. It is assumed that the maps are continuous, defined on intervals, and map them into themselves.

Prove that topologically conjugate maps simultaneously possess or fail to possess the following properties:

?
(a)

injectivity;

(b)

monotonicity of a definite type;

(c)

one-sided boundedness (without specifying its type);

(d)

boundedness.

Problem X.1.11

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map. It is assumed that the maps are continuous, defined on intervals, and map them into themselves.

Prove that topologically conjugate maps have

?
(a)

the same number of intervals of monotonicity;

(b)

the same number of fixed points.

Problem X.1.12

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map. It is assumed that the maps are continuous, defined on intervals, and map them into themselves.

Prove that if f∼gf \sim g, then the intervals on which the inequality f(x)>xf(x)>x holds are carried by the conjugating map hh into intervals on which (depending on the character of the monotonicity of hh) the same or the opposite strict inequality holds for the map gg. Prove that if the maps ff and gg are defined on R\mathbb {R} and one of them is odd, then the conjugating bijection hh can be chosen to be increasing.

?
Problem X.1.13

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map.

Determine whether the maps ff and gg defined on R\mathbb {R} are topologically conjugate in the following cases:

?
(a)

f(x)=x2−af(x)=x^{2}-a, g(x)=1−ax2(a≠0);g(x)=1-a x^{2} \quad (a \neq 0) ;

(b)

f(x)=cos⁡xf(x)=\cos x, g(x)=sin⁡xg(x)=\sin x;

(c)

f(x)=cos⁡xf(x)=\cos x, g(x)=−cos⁡xg(x)=-\cos x;

(d)

f(x)=sin⁡xf(x)=\sin x, g(x)=−sin⁡xg(x)=-\sin x;

(e)

f(x)=x+sin⁡xf(x)=x+\sin x, g(x)=x+cos⁡xg(x)=x+\cos x.

Problem X.1.14

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map.

Let α>0,α≠1,fα(x)=xα,gα(x)=x−α(x>0)\alpha >0, \alpha \neq 1, f_{\alpha }(x)=x^{\alpha }, g_{\alpha }(x)=x^{-\alpha }(x>0). For which α\alpha and β\beta are the functions topologically conjugate

?
(a)

fαf_{\alpha } and fβf_{\beta };

(b)

gαg_{\alpha } and gβg_{\beta };

(c)

fαf_{\alpha } and gβg_{\beta } ?

Problem X.1.15

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map.

?
(a)

Let f(x)=x2,g(x)=x2+ax+b(x∈R)f(x)=x^{2}, g(x)=x^{2}+a x+b(x \in \mathbb {R}). Describe the set of pairs (a,b)(a, b) for which f∼gf \sim g.

(b)

For which a∈Ra \in \mathbb {R} can one find b∈Rb \in \mathbb {R} such that the maps f(x)=1−ax2f(x)=1-a x^{2} and g(x)=bx(1−x)(x∈R)g(x)=b x(1-x)(x \in \mathbb {R}) are topologically conjugate? Describe the set BB of all numbers bb that arise in this way.

(c)

Prove that the maps gg and g~:[0,1]→[0,1],g(x)=2g~(x)==3x(1−x)\widetilde{g}:\left[0,1\right] \rightarrow [0,1], g(x)=2 \widetilde{g}(x)= =3 x(1-x), are not topologically conjugate.

(d)

For which a∈Ra \in \mathbb {R} and which interval Δ=[p,q]⊂R\Delta =\left[p, q\right] \subset \mathbb {R} is the map f:Δ→Δ,f(x)=1−ax2f: \Delta \rightarrow \Delta , f(x)=1-a x^{2}, topologically conjugate to the map g:[0,1]→[0,1]g:\left[0,1\right] \rightarrow [0,1] of the form g(x)=bx(1−x)g(x)=b x(1-x)?

Problem X.1.16

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map.

Prove the topological conjugacy of the continuous maps f,g:[−1,1]→[−1,1]f, g:\left[-1,1\right] \rightarrow [-1,1], given as follows:

?
(a)

f(x)=1−2∣x∣,g(x)=1−2x2(x∈[−1,1])f(x)=1-2\left|x\right|, g(x)=1-2 x^{2} \quad (x \in [-1,1]);

(b)

f=fn(n∈N)f=f_{n}(n \in \mathbb {N}) is the ``sawtooth'' map taking values (−1)k+n(-1)^{k+n} at the points −1+2kn,0⩽k⩽n-1+\frac{2 k}{n}, 0 \leqslant k \leqslant n, and linear on the intervals between them (Fig. 14), and g=Tng=T_{n} is the nn-th Chebyshev polynomial, defined by the equality Tn(x)=cos⁡(narccos⁡x)T_{n}(x)=\cos (n \arccos x) (see Problem I.2.20 and Fig. 15);

Fig. 14Fig. 14

Fig. 15Fig. 15

(c)

prove that the functions from part a) are conjugate as maps from R\mathbb {R} to R\mathbb {R}.

Problem X.1.17

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map.

Let f:[a1,b1]→[a1,b1],g:[a2,b2]→[a2,b2]f:\left[a_{1}, b_{1}\right] \rightarrow \left[a_{1}, b_{1}\right], g:\left[a_{2}, b_{2}\right] \rightarrow \left[a_{2}, b_{2}\right] be homeomorphisms of intervals having no fixed interior points. Prove that f∼gf \sim g. Verify that the statement remains valid if the closed intervals are replaced by open intervals.

?
Problem X.1.18

Maps f:Δ1→Δ1f: \Delta_{1} \rightarrow \Delta_{1} and g:Δ2→Δ2g: \Delta_{2} \rightarrow \Delta_{2} (Δ1,Δ2\Delta_{1}, \Delta_{2} being intervals on the line) are called topologically conjugate if there exists a continuous bijection (homeomorphism) h:Δ2→Δ1h: \Delta_{2} \rightarrow \Delta_{1} such that f=h∘g∘h−1f=h \circ g \circ h^{-1} (i.e., h∘g=f∘hh \circ g=f \circ h). In this case we write f∼gf \sim g and call hh the conjugating map.

Classify, up to topological conjugacy, the homeomorphisms of an interval onto itself having a finite set of fixed points, among which are the endpoints of the interval.

?
Problem X.1.19

Let ff be a map of an interval Δ⊂R\Delta \subset \mathbb {R} into itself, f0(x)=xf^{0}(x)=x, f1(x)=f(x)f^{1}(x)=f(x), fn(x)=f(fn−1(x))f^{n}(x)=f\left(f^{n-1}(x)\right) for n>1n>1. We say that the method of successive approximations with initial point x0∈Δx_{0} \in \Delta converges if the limit lim⁡xn\lim x_{n} exists, where xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right).

Let the function f∈C([a,b])f \in C(\left[a, b\right]) be increasing. Prove that the method of successive approximations converges for any initial point x0∈[a,b]x_{0} \in [a, b], if

?
(a)

f(x)<xf(x)<x for x>a,f(a)=ax>a, f(a)=a;

(b)

f(x)>xf(x)>x for x<b,f(b)=bx<b, f(b)=b.

Problem X.1.20

Let ff be a map of an interval Δ⊂R\Delta \subset \mathbb {R} into itself, f0(x)=xf^{0}(x)=x, f1(x)=f(x)f^{1}(x)=f(x), fn(x)=f(fn−1(x))f^{n}(x)=f\left(f^{n-1}(x)\right) for n>1n>1. We say that the method of successive approximations with initial point x0∈Δx_{0} \in \Delta converges if the limit lim⁡xn\lim x_{n} exists, where xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right).

Prove the convergence of the method of successive approximations for any starting point x0∈Δx_{0} \in \Delta and find lim⁡xn\lim x_{n} for the following maps ff:

?
(a)

f(x)=arctg⁡x,Δ=Rf(x)=\operatorname {arctg} x, \quad \Delta =\mathbb {R};

(b)

f(x)=x+2,Δ=[−2,+∞)f(x)=\sqrt{x+2}, \quad \Delta =[-2,+\infty );

(c)

f(x)=12x−1,Δ=Rf(x)=\frac{1}{2} x-1, \quad \Delta =\mathbb {R};

(d)

f(x)=x(x2+3)3x2+1,Δ=Rf(x)=\frac{x\left(x^{2}+3\right)}{3 x^{2}+1}, \quad \Delta =\mathbb {R};

(e)

f(x)=12(x+1x),Δ=(0,+∞)f(x)=\frac{1}{2}\left(x+\frac{1}{x}\right), \quad \Delta =(0,+\infty );

(f)

f(x)=2πarcsin⁡2xx2+1,Δ=(0,+∞)f(x)=\frac{2}{\pi } \arcsin \frac{2 x}{x^{2}+1}, \quad \Delta =(0,+\infty );

(g)

f(x)=42+x,Δ=[0,+∞)f(x)=\frac{4}{2+x}, \quad \Delta =[0,+\infty ).

Problem X.1.21

Let ff be a map of an interval Δ⊂R\Delta \subset \mathbb {R} into itself, f0(x)=xf^{0}(x)=x, f1(x)=f(x)f^{1}(x)=f(x), fn(x)=f(fn−1(x))f^{n}(x)=f\left(f^{n-1}(x)\right) for n>1n>1. We say that the method of successive approximations with initial point x0∈Δx_{0} \in \Delta converges if the limit lim⁡xn\lim x_{n} exists, where xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right).

Determine for which starting points x0x_{0} in the domain of definition of the map ff the method of successive approximations converges, and find lim⁡xn\lim x_{n}, if

?
(a)

f(x)=x+sin⁡xf(x)=x+\sin x;

(b)

f(x)=x(x−1)f(x)=x(x-1);

(c)

f(x)=1+ax2(0<a⩽14);f(x)=1+a x^{2} \left(0<a \leqslant \frac{1}{4}\right) ;

(d)

f(x)=1−ax2(0<a⩽34);f(x)=1-a x^{2} \left(0<a \leqslant \frac{3}{4}\right) ;

(e)

f(x)=x2+c(∣c∣⩽14);f(x)=x^{2}+c \left(\left|c\right| \leqslant \frac{1}{4}\right) ;

(f)

f(x)=x2+(1−2c)x+c2(c∈R)f(x)=x^{2}+(1-2 c) x+c^{2}(c \in \mathbb {R});

(g)

f(x)={14−3x, for x≠4313, for x=43f(x)=\begin{cases} \frac{1}{4-3 x}, & \text{ for } x \neq \frac{4}{3} \\ \frac{1}{3}, & \text{ for } x=\frac{4}{3} \end{cases};

(h)

f(x)={2x33x2−1, if ∣x∣≠13,±1, if x=±13.f(x)=\begin{cases} \frac{2 x^{3}}{3 x^{2}-1}, & \text{ if }\left|x\right| \neq \frac{1}{\sqrt{3}}, \\ \pm 1, & \text{ if } x= \pm \frac{1}{\sqrt{3}} . \end{cases}

Problem X.1.22

We call a fixed point x∗x^{*} of the map ff attracting if for all x0x_{0} taken from some neighborhood of the point x∗x^{*}, the sequence xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converges to the point x∗x^{*}. If for some neighborhood UU of the fixed point x∗x^{*} and an arbitrary point x0∈U,x0≠x∗x_{0} \in U, x_{0} \neq x^{*}, there is an index nn such that xn∉Ux_{n} \notin U, then the point x∗x^{*} is called repelling.

Let x∗x^{*} be a fixed point of the map ff, and suppose ff is differentiable at x∗x^{*}. Prove that

?
(a)

if ∣f′(x∗)∣<1\left|f^{\prime }\left(x^{*}\right)\right|<1, then x∗x^{*} is attracting;

(b)

if ∣f′(x∗)∣>1\left|f^{\prime }\left(x^{*}\right)\right|>1, then x∗x^{*} is repelling;

(c)

show by examples that a fixed point may be neither attracting nor repelling.

Problem X.1.23

For which values of the parameter aa and for which x0∈Rx_{0} \in \mathbb {R} does the sequence of iterates xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converge, if

?
(a)

f(x)=1+axf(x)=1+a x;

(b)

f(x)=1−a∣x∣(a>0)f(x)=1-a\left|x\right|(a>0);

(c)

f(x)=1−ax2(a>0)f(x)=1-a x^{2}(a>0);

(d)

f(x)=ax(a>0)f(x)=a^{x}(a>0).

Problem X.1.24

Let φ\varphi be a differentiable function defined on an interval Δ\Delta, and NφN_{\varphi } the map defined at those points x∈Δx \in \Delta where φ′(x)≠0\varphi^{\prime }(x) \neq 0 by Nφ(x)=x−φ(x)/φ′(x)N_{\varphi }(x)=x-\varphi (x) / \varphi^{\prime }(x). The fixed points of NφN_{\varphi } coincide with the roots of φ\varphi. We say that Newton's iterative process with initial point x0x_{0} converges to cc if xk=Nφk(x0)→cx_{k}=N_{\varphi }^{k}\left(x_{0}\right) \rightarrow c.

Let φ∈C2(Δ),c∈Δ,φ(c)=0,∣φ′(x)∣>m>0\varphi \in C^{2}(\Delta ), c \in \Delta , \varphi (c)=0,\left|\varphi^{\prime }(x)\right|>m>0 for x∈Δx \in \Delta. Prove that if the point x0x_{0} is sufficiently close to cc, then

?
(a)

Newton's iterative process, started at x0x_{0}, converges to cc;

(b)

"superconvergence" holds: ∣xk−c∣⩽Mq2k\left|x_{k}-c\right| \leqslant M q^{2^{k}}, where M>0M>0, 0<q<1,k∈N0<q<1, k \in \mathbb {N} (cf. problem II.3.10).

Problem X.1.25

Let φ\varphi be a differentiable function defined on an interval Δ\Delta, and NφN_{\varphi } the map defined at those points x∈Δx \in \Delta where φ′(x)≠0\varphi^{\prime }(x) \neq 0 by Nφ(x)=x−φ(x)/φ′(x)N_{\varphi }(x)=x-\varphi (x) / \varphi^{\prime }(x). The fixed points of NφN_{\varphi } coincide with the roots of φ\varphi. We say that Newton's iterative process with initial point x0x_{0} converges to cc if xk=Nφk(x0)→cx_{k}=N_{\varphi }^{k}\left(x_{0}\right) \rightarrow c.

Prove that if φ∈C(n)(Δ),c∈Δ,φ(c)=φ′(c)=…==φ(n−1)(c)=0,φ(n)(c)≠0\varphi \in C^{(n)}(\Delta ), c \in \Delta , \varphi (c)=\varphi^{\prime }(c)=\ldots = =\varphi^{(n-1)}(c)=0, \varphi^{(n)}(c) \neq 0, then for initial points x0x_{0} sufficiently close to cc, Newton's iterative process converges to cc, and moreover ∣xk−c∣⩽Mqk(M>0,0<q<1)\left|x_{k}-c\right| \leqslant M q^{k}(M>0,0<q<1).

?
Problem X.1.26

Let φ\varphi be a differentiable function defined on an interval Δ\Delta, and NφN_{\varphi } the map defined at those points x∈Δx \in \Delta where φ′(x)≠0\varphi^{\prime }(x) \neq 0 by Nφ(x)=x−φ(x)/φ′(x)N_{\varphi }(x)=x-\varphi (x) / \varphi^{\prime }(x). The fixed points of NφN_{\varphi } coincide with the roots of φ\varphi. We say that Newton's iterative process with initial point x0x_{0} converges to cc if xk=Nφk(x0)→cx_{k}=N_{\varphi }^{k}\left(x_{0}\right) \rightarrow c.

Let φ∈C2([a,b])\varphi \in C^{2}(\left[a, b\right]) be a function such that φφ′′>0\varphi \varphi^{\prime \prime }>0 on (a,b)(a, b) and φ(a)=φ′′(b)=0\varphi (a)=\varphi^{\prime \prime }(b)=0 (or φ(b)=φ′′(a)=0\varphi (b)=\varphi^{\prime \prime }(a)=0). Verify that for any initial point from [a,b]\left[a, b\right], Newton's iterative process converges to a root of φ\varphi.

?
Problem X.1.27

Let φ\varphi be a differentiable function defined on an interval Δ\Delta, and NφN_{\varphi } the map defined at those points x∈Δx \in \Delta where φ′(x)≠0\varphi^{\prime }(x) \neq 0 by Nφ(x)=x−φ(x)/φ′(x)N_{\varphi }(x)=x-\varphi (x) / \varphi^{\prime }(x). We say that Newton's iterative process with initial point x0x_{0} converges to cc if xk=Nφk(x0)→cx_{k}=N_{\varphi }^{k}\left(x_{0}\right) \rightarrow c.

Let PP be a polynomial having only simple real roots. Prove that if x0x_{0} is an inflection point of PP (i.e. P′′(x0)=0P^{\prime \prime }\left(x_{0}\right)=0), then Newton's iterative process, started at the point x0x_{0}, converges to a root of PP. In a similar way, by varying the inflection points, one can find all the roots except the minimal and maximal ones.

?
Problem X.1.28

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Let (x0∗,x1∗,…,xn−1∗)−n\left(x_{0}^{*}, x_{1}^{*}, \ldots , x_{n-1}^{*}\right)-n-cycle of the map ff be given. Prove that

?
(a)

if ∣(fn)′(x0∗)∣<1\left|\left(f^{n}\right)^{\prime }\left(x_{0}^{*}\right)\right|<1, then for x0x_{0} taken from some neighborhood of x0∗x_{0}^{*}, the relation

fmn+k(x0)⟶m→∞xk∗,k=0,…,n−1; f^{m n+k}\left(x_{0}\right) \underset {m \rightarrow \infty }{\longrightarrow } x_{k}^{*}, \quad k=0, \ldots , n-1 ;

holds;

(b)

if ∣(fn)′(x0∗)∣>1\left|\left(f^{n}\right)^{\prime }\left(x_{0}^{*}\right)\right|>1, then for some ε>0\varepsilon >0 and any x0∈(x0∗−ε,x0∗+ε),x0≠x0∗\quad x_{0} \in \left(x_{0}^{*}-\varepsilon , x_{0}^{*}+\varepsilon \right), \quad x_{0} \neq x_{0}^{*}, there is an index mm such that ∣fm(x0)−xk∗∣⩾ε\left|f^{m}\left(x_{0}\right)-x_{k}^{*}\right| \geqslant \varepsilon for all k=0,…,n−1k=0, \ldots , n-1.

Problem X.1.29

Let ff be a continuous map of a segment Δ\Delta into itself, and let I0,I1,…I_{0}, I_{1}, \ldots be a finite or infinite sequence of segments, with I0⊂Δ,In+1⊂f(In)I_{0} \subset \Delta , I_{n+1} \subset f\left(I_{n}\right) for every n⩾0n \geqslant 0. Prove that

?
(a)

there exists a sequence of segments K0⊃K1⊃K2⊃⊃K3⊃…,K0⊂I0K_{0} \supset K_{1} \supset K_{2} \supset \supset K_{3} \supset \ldots , K_{0} \subset I_{0}, such that fn(Kn)=In(n⩾0)f^{n}\left(K_{n}\right)=I_{n}(n \geqslant 0);

(b)

there exists a point x∈Δx \in \Delta for which xn=fn(x)∈Inx_{n}=f^{n}(x) \in I_{n} for every nn.

Problem X.1.30

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Prove that if a continuous map of a segment into itself has a 4-cycle, then it also has a 2-cycle.

?
Problem X.1.31

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Let f:Δ→Δf: \Delta \rightarrow \Delta be a continuous map of a segment into itself. Prove that the following statements are equivalent:

  1. there exists a point a∈Δa \in \Delta such that the points b=f(a),c=f2(a)b=f(a), c=f^{2}(a), d=f3(a)d=f^{3}(a) satisfy the inequalities d⩽a<b<cd \leqslant a<b<c or c<b<a⩽dc<b<a \leqslant d;

  2. in Δ\Delta there exists a point having period three;

  3. for any m∈Nm \in \mathbb {N}, in Δ\Delta there exists a point having period mm.

?
Remark.
?

The implication b) ⇒ c) and problem 1.30 are special cases of A.N. Sharkovsky's theorem [Sh]. Let f:Δ→Δf: \Delta \rightarrow \Delta be a continuous map (Δ=[a,b])(\Delta =\left[a, b\right]). Consider the natural numbers ordered as follows:

1,2,4,8,…,2n,…,…,2m⋅7,2m⋅5,2m⋅3,…,2⋅7,2⋅5,2⋅3,…,7,5,3. 1,2,4,8, \ldots , 2^{n}, \ldots , \ldots , 2^{m} \cdot 7,2^{m} \cdot 5,2^{m} \cdot 3, \ldots , 2 \cdot 7,2 \cdot 5,2 \cdot 3, \ldots , 7,5,3 .

If pp stands to the left of qq in this sequence and ff has a qq-cycle, then ff also has a pp-cycle.

Problem X.1.32

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Prove that for n⩾2n \geqslant 2 the Chebyshev polynomial TnT_{n} (see problems I.2.20 and 1.16 b)), defined on [−1,1]\left[-1,1\right], has cycles of arbitrary order.

?
Problem X.1.33

A fixed point x∗x^{*} of a map ff is called attracting if for all x0x_{0} from some neighborhood of x∗x^{*} the sequence xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converges to x∗x^{*}, and repelling if for some neighborhood UU of it and every point x0∈Ux_{0} \in U, x0≠x∗x_{0} \neq x^{*}, there is an index nn such that xn∉Ux_{n} \notin U. If the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1}, where xk=fk(x0)x_{k}=f^{k}\left(x_{0}\right), are pairwise distinct while xn=x0x_{n}=x_{0}, then the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) is called an nn-cycle and nn the period of x0x_{0}; the nn-cycle is called attracting if x0x_{0} is an attracting fixed point of fnf^{n}, and repelling if x0x_{0} is a repelling fixed point of fnf^{n}.

Find the values of the parameter aa for which the map f(x)=1−a∣x∣(x∈R)f(x)=1-a\left|x\right|(x \in \mathbb {R}) has

?
(a)

a 2-cycle;

(b)

a 4-cycle;

(c)

a 3-cycle,

and determine when these cycles are attracting.

Problem X.1.34

A fixed point x∗x^{*} of a map ff is called attracting if for all x0x_{0} from some neighborhood of x∗x^{*} the sequence xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converges to x∗x^{*}, and repelling if for some neighborhood UU of it and every point x0∈Ux_{0} \in U, x0≠x∗x_{0} \neq x^{*}, there is an index nn such that xn∉Ux_{n} \notin U. If the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1}, where xk=fk(x0)x_{k}=f^{k}\left(x_{0}\right), are pairwise distinct while xn=x0x_{n}=x_{0}, then the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) is called an nn-cycle and nn the period of x0x_{0}; the nn-cycle is called attracting if x0x_{0} is an attracting fixed point of fnf^{n}, and repelling if x0x_{0} is a repelling fixed point of fnf^{n}.

Let a>0,f(x)=1−ax2(x∈R)a>0, f(x)=1-a x^{2}(x \in \mathbb {R}). Determine for which values of the parameter aa an attracting fixed point and an attracting 2-cycle exist. Find the "birth moments" of the 2-, 4- and 3-cycles (i.e., the lower bounds on the values of aa for which these cycles exist).

?
Problem X.1.35

A fixed point x∗x^{*} of a map ff is called attracting if for all x0x_{0} from some neighborhood of x∗x^{*} the sequence xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converges to x∗x^{*}, and repelling if for some neighborhood UU of it and every point x0∈Ux_{0} \in U, x0≠x∗x_{0} \neq x^{*}, there is an index nn such that xn∉Ux_{n} \notin U. If the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1}, where xk=fk(x0)x_{k}=f^{k}\left(x_{0}\right), are pairwise distinct while xn=x0x_{n}=x_{0}, then the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) is called an nn-cycle and nn the period of x0x_{0}; the nn-cycle is called attracting if x0x_{0} is an attracting fixed point of fnf^{n}, and repelling if x0x_{0} is a repelling fixed point of fnf^{n}.

Let f(x)=bx(1−x)(x∈[0,1]),0<b⩽4f(x)=b x(1-x)(x \in [0,1]), 0<b \leqslant 4. Find the lower bound of the values of the parameter bb for which ff has

?
(a)

an attracting 2-cycle;

(b)

a repelling 2-cycle;

(c)

a 4-cycle;

(d)

a 3-cycle.

Problem X.1.36

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Let x0∈(0,1)x_{0} \in (0,1) be a fixed point of a continuous map f:[0,1]→[0,1]f:\left[0,1\right] \rightarrow [0,1]. Prove that if the one-sided derivatives f±′(x0)f_{ \pm }^{\prime }\left(x_{0}\right) are negative and f+′(x0)f−′(x0)>1f_{+}^{\prime }\left(x_{0}\right) f_{-}^{\prime }\left(x_{0}\right)>1, then ff has a 2-cycle. In particular, a 2-cycle exists if f′(x0)<−1f^{\prime }\left(x_{0}\right)<-1.

?
Problem X.1.37

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Let gg be a piecewise smooth map of the interval [0,1]\left[0,1\right] into itself. Prove that if ∣g′(x)∣⩾a>1\left|g^{\prime }(x)\right| \geqslant a>1 at all points where the derivative exists, then gg has a 2n2^{n}-cycle for every n∈Nn \in \mathbb {N}. Show by an example that this fails if the smoothness of the map is violated on a countable set of points.

?
Problem X.1.38

We consider maps ff of an interval Δ=[a,b]\Delta =\left[a, b\right] into itself such that f∈C3(Δ)f \in C^{3}(\Delta ) and the Schwarzian derivative SfS f (see Problems VII.2.25, VII.2.26 and the definition preceding them) is negative at those points xx that are not critical points of ff (i.e., where f′(x)≠0f^{\prime }(x) \neq 0). We denote the class of such maps by the symbol S(Δ)S(\Delta ). If the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1}, where xk=fk(x0)x_{k}=f^{k}\left(x_{0}\right), are pairwise distinct while xn=x0x_{n}=x_{0}, then the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) is called an nn-cycle.

Let f∈S(Δ)f \in S(\Delta ) and let the set of critical points of ff be finite. Prove that for every n∈Nn \in \mathbb {N} the transformation ff can have only finitely many nn-cycles.

?
Problem X.1.39

We consider maps ff of an interval Δ=[a,b]\Delta =\left[a, b\right] into itself such that f∈C3(Δ)f \in C^{3}(\Delta ) and the Schwarzian derivative SfS f (see Problems VII.2.25, VII.2.26 and the definition preceding them) is negative at those points xx that are not critical points of ff (i.e., where f′(x)≠0f^{\prime }(x) \neq 0). We denote the class of such maps by the symbol S(Δ)S(\Delta ).

Let f∈S(Δ),g=fn(n⩾1),x<y<zf \in S(\Delta ), g=f^{n}(n \geqslant 1), x<y<z be consecutive fixed points of the map gg. Prove that if the interval [x,y]\left[x, y\right] contains no critical points of gg, then g′(y)>1g^{\prime }(y)>1.

?
Problem X.1.40

We consider maps ff of an interval Δ=[a,b]\Delta =\left[a, b\right] into itself such that f∈C3(Δ)f \in C^{3}(\Delta ) and the Schwarzian derivative SfS f (see Problems VII.2.25, VII.2.26 and the definition preceding them) is negative at those points xx that are not critical points of ff (i.e., where f′(x)≠0f^{\prime }(x) \neq 0). We denote the class of such maps by the symbol S(Δ)S(\Delta ). A fixed point x∗x^{*} of a map ff is called attracting if for all x0x_{0} from some neighborhood of x∗x^{*} the sequence xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converges to x∗x^{*}. If the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1}, where xk=fk(x0)x_{k}=f^{k}\left(x_{0}\right), are pairwise distinct while xn=x0x_{n}=x_{0}, then the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) is called an nn-cycle; the nn-cycle is called attracting if x0x_{0} is an attracting fixed point of fnf^{n}. The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}.

Let the map f∈S(Δ)f \in S(\Delta ) have a unique critical point c∈(a,b)c \in (a, b). Prove that every attracting cycle of the map ff attracts the trajectory of one of the points a,c,ba, c, b, and consequently ff can have at most three attracting cycles.

?
Problem X.1.41

A fixed point x∗x^{*} of a map ff is called attracting if for all x0x_{0} from some neighborhood of x∗x^{*} the sequence xn=fn(x0)x_{n}=f^{n}\left(x_{0}\right) converges to x∗x^{*}, and repelling if for some neighborhood UU of it and every point x0∈Ux_{0} \in U, x0≠x∗x_{0} \neq x^{*}, there is an index nn such that xn∉Ux_{n} \notin U. If the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1}, where xk=fk(x0)x_{k}=f^{k}\left(x_{0}\right), are pairwise distinct while xn=x0x_{n}=x_{0}, then the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) is called an nn-cycle and nn the period of x0x_{0}; the nn-cycle is called attracting if x0x_{0} is an attracting fixed point of fnf^{n}, and repelling if x0x_{0} is a repelling fixed point of fnf^{n}.

Prove that a map of the form f(x)=1−ax2(x∈R)f(x)=1-a x^{2}(x \in \mathbb {R}), for a⩽2a \leqslant 2, cannot have more than one attracting cycle, while for a=2a=2 it has an nn-cycle for every n∈Nn \in \mathbb {N}, but none of them is attracting.

?
Problem X.1.42

The set {xn=fn(x0)∣n∈N}\left\{ x_{n}=f^{n}\left(x_{0}\right) \mid n \in \mathbb {N}\right\} is called the trajectory of the point x0x_{0}. If the trajectory is finite, then there exists n∈Nn \in \mathbb {N} such that the points x0,x1,…,xn−1x_{0}, x_{1}, \ldots , x_{n-1} are pairwise distinct, while xn=x0,xn+1=x1,…x_{n}=x_{0}, x_{n+1}=x_{1}, \ldots; in this case x0x_{0} is called a periodic point, the number nn its period, and the tuple (x0,…,xn−1)\left(x_{0}, \ldots , x_{n-1}\right) an nn-cycle.

Let φ\varphi be a continuous map of the circle into itself having a fixed point and a 3-cycle.

?
(a)

Prove that φ\varphi also has nn-cycles for all n>3n>3.

(b)

Must φ\varphi have a 2-cycle?

Problem X.1.43

Let ff be a continuous increasing function on the line satisfying the condition f(x+1)=f(x)+1f(x+1)=f(x)+1. Prove that

?
(a)

the map z=e2πix↦φ(z)=e2πif(x)z=e^{2 \pi i x} \mapsto \varphi (z)=e^{2 \pi i f(x)} defines a homeomorphism of the circle S1={z∣∣z∣=1}S^{1}=\left\{ z \mid \left|z\right|=1\right\}, preserving the cyclic order of any three points, and that an arbitrary orientation-preserving homeomorphism of the circle can be defined in this way;

(b)

if the limit lim⁡n→∞fn(x)n=α\lim_{n \rightarrow \infty } \frac{f^{n}(x)}{n}=\alpha exists for some x0∈Rx_{0} \in \mathbb {R}, then it also exists for every x∈Rx \in \mathbb {R};

(c)

if the map φm\varphi^{m} has a fixed point for some m∈Nm \in \mathbb {N}, then the limit α\alpha exists and is a rational number;

(d)

if no power of φ\varphi has a fixed point, then the limit α\alpha exists and is an irrational number;

(e)

if two distinct functions f1f_{1} and f2f_{2} define the same homeomorphism φ\varphi, then the corresponding limits α1\alpha_{1} and α2\alpha_{2} differ by an integer.

The number α mod 1\alpha \bmod 1 defined in this way for an arbitrary orientation-preserving homeomorphism φ:S1→S1\varphi : S^{1} \rightarrow S^{1} is called the rotation number of this homeomorphism. This notion was introduced by H. Poincare.

Problem X.2.1

Check that the following transformations preserve Lebesgue measure:

?
(a)

T(x)=(2x) mod 1(x∈[0,1])T(x)=(2 x) \bmod 1 \quad (x \in [0,1]);

(b)

T(z)=zn(z∈C,∣z∣=1),n∈NT(z)=z^{n} \quad (z \in \mathbb {C},\left|z\right|=1), n \in \mathbb {N};

(c)

T(x)=x−1x(x∈R\{0})T(x)=x-\frac{1}{x}(x \in \mathbb {R} \backslash \left\{ 0\right\} ).

Problem X.2.2

Check that the transformation

T(x,y)=((2x) mod 1,12(y+[2x])),(x,y)∈[0,1)×[0,1) T(x, y)=\left((2 x) \bmod 1, \frac{1}{2}(y+\left[2 x\right])\right), \quad (x, y) \in [0,1) \times [0,1)

(the ``baker's transformation'') preserves Lebesgue measure λ2\lambda_{2}.

?
Problem X.2.3

Let X=[0,1)×[0,1),A=(abcd)X=[0,1) \times [0,1), A=\left(\begin{array}{ll}a & b \\ c & d\end{array}\right) be an integer matrix with det⁡A=±1\operatorname {det} A= \pm 1. Define the transformation TA:X→XT_{A}: X \rightarrow X by the formulas TA(x1,x2)=(y1,y2)T_{A}\left(x_{1}, x_{2}\right)=\left(y_{1}, y_{2}\right) for (x1,x2)∈X\left(x_{1}, x_{2}\right) \in X, where

y1=(ax1+bx2) mod 1,y2=(cx1+dx2) mod 1. y_{1}=\left(a x_{1}+b x_{2}\right) \bmod 1, \quad y_{2}=\left(c x_{1}+d x_{2}\right) \bmod 1 .

Prove that

?
(a)

TAT_{A} is one-to-one;

(b)

TAT_{A} preserves Lebesgue measure.

The map (x1,x2)↦(e2πix1,e2πix2)\left(x_{1}, x_{2}\right) \mapsto \left(e^{2 \pi i x_{1}}, e^{2 \pi i x_{2}}\right) lets us identify XX with the torus T2=S1×S1T^{2}=S^{1} \times S^{1} and TAT_{A} with an automorphism of T2T^{2} as a group. This gives us the right to call the transformation TAT_{A} an automorphism of the torus.

Problem X.2.4

Let the map TA:X→XT_{A}: X \rightarrow X be the same as in the previous problem, but with det⁡A∈Z,∣det⁡A∣>1\operatorname {det} A \in \mathbb {Z},\left|\operatorname {det} A\right|>1. Such a map (no longer one-to-one) is called an endomorphism of the torus. Prove that it preserves Lebesgue measure.

?
Remark.
?

Recall the fundamental result: on every compact group GG there exists a nontrivial finite Borel measure μ\mu invariant under translations:

μ(gA)=μ(Ag)=μ(A), \mu (g A)=\mu (A g)=\mu (A),

where A⊂G,g∈G,gA={gx∣x∈A},Ag={xg∣x∈A}A \subset G, g \in G, g A=\left\{ g x \mid x \in A\right\} , A g=\left\{ x g \mid x \in A\right\}. In the case of a metrizable group this measure is uniquely determined (up to a constant factor) and is called Haar measure. Its uniqueness can be used in solving this problem.

Problem X.2.5

Check that the map T(x)=1x mod 1(x∈(0,1])T(x)=\frac{1}{x} \bmod 1(x \in (0,1]) preserves the measure dμ=C1+xdxd \mu =\frac{C}{1+x} d x.

?
Problem X.2.6

Find an absolutely continuous invariant measure

?
(a)

for the map Tn:[−1,1]→[−1,1]T_{n}:\left[-1,1\right] \rightarrow [-1,1], where TnT_{n} is the Chebyshev polynomial (see problems I.2.20 and 1.16 b));

(b)

for the map f:R→Rf: \mathbb {R} \rightarrow \mathbb {R}, where f(x)=1−2x2f(x)=1-2 x^{2}.

Problem X.2.7

Prove that a transformation TT has an invariant (respectively, finite invariant, absolutely continuous invariant) measure if and only if some power Tn(n∈N)T^{n}(n \in \mathbb {N}) of it has this property.

?
Problem X.2.8

Let T:[a,b]→[a,b]T:\left[a, b\right] \rightarrow [a, b] be a strictly monotone continuous transformation. Prove that it has an absolutely continuous invariant measure (in general, an infinite one).

?
Problem X.2.9

Let the transformation T:[a,b]→[a,b]T:\left[a, b\right] \rightarrow [a, b] be piecewise monotone and smooth. Prove that if it has an attracting cycle (see 1), then TT does not preserve any finite measure equivalent to Lebesgue measure.

?
Problem X.2.10

Does there exist a nonzero measure on R\mathbb {R} invariant with respect to all

?
(a)

homotheties Ha:x↦ax(a≠0)H_{a}: x \mapsto a x(a \neq 0);

(b)

affine transformations Aa,b:x↦ax+b(a,b∈R,a≠0)A_{a, b}: x \mapsto a x+b(a, b \in \mathbb {R}, a \neq 0)?

In the case of an affirmative answer, give an example.

Problem X.2.11

Does there exist a nontrivial measure on (0,+∞)(0,+\infty ) invariant with respect to all transformations of the form

?
(a)

Tp:x→xp(p≠0)T_{p}: x \rightarrow x^{p}(p \neq 0);

(b)

Sb,p:x→bxp(p≠0,b>0)?S_{b, p}: x \rightarrow b x^{p}(p \neq 0, b>0)?

In the case of an affirmative answer, give an example.

Problem X.2.12

A measure μ\mu on Rn\mathbb {R}^{n} is called quasi-invariant with respect to shifts if for every a∈Rna \in \mathbb {R}^{n} the condition μ(A)=0\mu (A)=0 is equivalent to the condition μ(A+a)=0\mu (A+a)=0.

Prove that

?
(a)

if a measure μ\mu on Rn\mathbb {R}^{n} is equivalent to Lebesgue measure λn\lambda_{n}, then μ\mu is quasi-invariant with respect to shifts;

(b)

if μ\mu is quasi-invariant with respect to shifts, and the measure of every bounded set is finite, then μ\mu is equivalent to λn\lambda_{n}.

Problem X.2.13

Let TT be a transformation of a set X⊂RnX \subset \mathbb {R}^{n} into itself such that the image TλnT \lambda_{n} of the measure λn\lambda_{n} is absolutely continuous with respect to λn\lambda_{n}, and let ff be an arbitrary summable function on XX. Prove the following statements:

?
(a)

There exists a function f∗∈L1(X)f^{*} \in \mathscr {L}^{1}(X) such that

∫T−1(A)f(x)dx=∫Af∗(x)dx(A⊂X)(1) \int _{T^{-1}(A)} f(x) d x=\int _{A} f^{*}(x) d x \quad (A \subset X) \tag {1}

and f∗f^{*} is uniquely determined up to values on a set of measure zero.

The correspondence P:f↦f∗P: f \mapsto f^{*} is called the Perron-Frobenius operator associated with TT.

(b)

If X⊂RnX \subset \mathbb {R}^{n} is a domain, TT is a smooth one-to-one mapping, with det⁡T′(x)≠0(x∈X)\operatorname {det} T^{\prime }(x) \neq 0(x \in X), then

Pf(Tx)=f(x)∣det⁡T′(x)∣(x∈X). P f(T x)=\frac{f(x)}{\left|\operatorname {det} T^{\prime }(x)\right|} \quad (x \in X) .
(c)

If X=[0,1],TX=\left[0,1\right], T is piecewise differentiable and the set T−1(x)T^{-1}(x) is finite (x∈Xx \in X), then

Pf(y)=∑x∈T−1(y)f(x)∣T′(x)∣ P f(y)=\sum _{x \in T^{-1}(y)} \frac{f(x)}{\left|T^{\prime }(x)\right|}

at all points at which the right-hand side of the equality is defined.

(d)

If X=[0,1],f∈C([0,1])X=\left[0,1\right], f \in C(\left[0,1\right]) and TT is piecewise monotone, then

Pf(x)=ddx∫T−1([0,x])f(t)dt P f(x)=\frac{d}{d x} \int _{T^{-1}(\left[0, x\right])} f(t) d t
(e)

In order for the measure dμ=fdxd \mu =f d x to be invariant with respect to TT, it is necessary and sufficient that for almost all x∈Xx \in X the equality Pf(x)=f(x)P f(x)=f(x) hold.

Problem X.2.14

Let n∈N,Δi=[i−1n,in)(i=1,…,n)n \in \mathbb {N}, \Delta_{i}=\left[\frac{i-1}{n}, \frac{i}{n}\right)(i=1, \ldots , n). Consider the nonnegative functions on [0,1)[0,1) that are constant on the intervals Δi\Delta_{i}. Each such function is given by a vector c=(c1,…,cn)c=\left(c_{1}, \ldots , c_{n}\right) of the values it takes on Δ1,…,Δn\Delta_{1}, \ldots , \Delta_{n}. For clarity, let us denote such a function by the symbol fcf_{c}. Denote the set of nonnegative vectors in Rn\mathbb {R}^{n} by R+n\mathbb {R}_{+}^{n}.

Let T:[0,1)→[0,1)T:[0,1) \rightarrow [0,1) be an arbitrary transformation.

?
(a)

Prove that for every function of the form fc(c∈R+n)f_{c}\left(c \in \mathbb {R}_{+}^{n}\right) there exists a unique function fd(d∈R+n)f_{d}\left(d \in \mathbb {R}_{+}^{n}\right) satisfying the equalities

∫T−1(Δi)fc(x)dx=∫Δifd(x)dx,i=1,…,n \int _{T^{-1}\left(\Delta _{i}\right)} f_{c}(x) d x=\int _{\Delta _{i}} f_{d}(x) d x, \quad i=1, \ldots , n

and moreover the vector dd has the form d=πcd=\pi c, where π=(pij)i,j=1n\pi =\left(p_{i j}\right)_{i, j=1}^{n} is a certain matrix determined by the mapping TT.

The mapping c↦πcc \mapsto \pi c is the discrete (finite-dimensional) analogue of the Perron-Frobenius operator.

(b)

Find the matrix π\pi and prove that pij⩾0p_{i j} \geqslant 0 and ∑1⩽i⩽npij=1\sum_{1 \leqslant i \leqslant n} p_{i j}=1.

(c)

Prove that there exists a vector v∈R+n,v≠0v \in \mathbb {R}_{+}^{n}, v \neq 0, such that πv=v\pi v=v (to this end, consider the sequence of vectors v(m)=1m∑0⩽k<mπku,u∈R+n)\left.v^{(m)}=\frac{1}{m} \sum_{0 \leqslant k<m} \pi^{k} u, u \in \mathbb {R}_{+}^{n}\right).

Problem X.2.15

Let n∈N,Δi=[i−1n,in)(i=1,…,n)n \in \mathbb {N}, \Delta_{i}=\left[\frac{i-1}{n}, \frac{i}{n}\right)(i=1, \ldots , n). A nonnegative function on [0,1)[0,1) that is constant on the intervals Δi\Delta_{i} and takes the values c1,…,cnc_{1}, \ldots , c_{n} on Δ1,…,Δn\Delta_{1}, \ldots , \Delta_{n} is denoted by fcf_{c}, where c=(c1,…,cn)c=\left(c_{1}, \ldots , c_{n}\right).

Let T:[0,1]→[0,1]T:\left[0,1\right] \rightarrow [0,1].

?
(a)

Find the general form of an absolutely continuous measure invariant with respect to TT, if

T(x)={2x for x⩽12,32−x for x>12. T(x)=\begin{cases} 2 x & \text{ for } x \leqslant \frac{1}{2}, \\ \frac{3}{2}-x & \text{ for } x>\frac{1}{2} . \end{cases}
(b)

Indicate some absolutely continuous measure invariant with respect to TT, if

T(x)={x+13 for x⩽23,2−2x for x>23. T(x)=\begin{cases} x+\frac{1}{3} & \text{ for } x \leqslant \frac{2}{3}, \\ 2-2 x & \text{ for } x>\frac{2}{3} . \end{cases}
(c)

Let T(x)=x2T(x)=\frac{x}{2}. Find the fixed point vnv_{n} for the discrete Perron-Frobenius operator (for even nn) and prove the absence of an invariant finite absolutely continuous measure for TT. What do the measures dμn=fvndxd \mu_{n}=f_{v_{n}} d x converge to as n→∞n \rightarrow \infty, if they are normalized by the condition μn([0,1))=1\mu_{n}([0,1))=1?

Problem X.2.16

Let Ω={z∈C∣Im⁡z>0}\Omega =\left\{ z \in \mathbb {C} \mid \operatorname {Im} z>0\right\}. Find an absolutely continuous measure invariant under all conformal automorphisms of Ω\Omega, i.e., transformations T(z)=az+bcz+dT(z)=\frac{a z+b}{c z+d}, where a,b,c,d∈Ra, b, c, d \in \mathbb {R} and det⁡(abcd)>0\operatorname {det}\left(\begin{array}{ll}a & b \\ c & d\end{array}\right)>0.

?
Problem X.2.17

Let P={(x,y)∈R2∣x>0}P=\left\{ (x, y) \in \mathbb {R}^{2} \mid x>0\right\}. For each point (a,b)∈P(a, b) \in P define the transformations La,bL_{a, b} and Ra,bR_{a, b} of the half-plane PP into itself:

La,b(x,y)=(ax,ay+b),Ra,b(x,y)=(ax,bx+y). L_{a, b}(x, y)=(a x, a y+b), \quad R_{a, b}(x, y)=(a x, b x+y) .

Find absolutely continuous measures μL\mu_{L} and μR\mu_{R} on PP invariant under all transformations La,bL_{a, b} and Ra,bR_{a, b}, respectively.

Note that the transformations in question correspond to the left and right shift transformations on the group (under multiplication) of matrices

Mx,y=[xy01],(x,y)∈P, M_{x, y}=\left[\begin{smallmatrix} x & y \\ 0 & 1 \end{smallmatrix}\right],(x, y) \in P,

i.e., to the transformations Mx,y↦Ma,bMx,y;Mx,y↦Mx,yMa,bM_{x, y} \mapsto M_{a, b} M_{x, y} ; M_{x, y} \mapsto M_{x, y} M_{a, b}. The measures μL\mu_{L} and μR\mu_{R} are called the Haar measures (left and right) for this group.

?
Problem X.2.18

Prove that if a transformation T:[a,b]→[a,b]T:\left[a, b\right] \rightarrow [a, b] preserves some finite absolutely continuous measure, then TT is topologically conjugate to some transformation of the same interval preserving Lebesgue measure.

?
Problem X.2.19

Prove that if TT is a homeomorphism of the circle S1S^{1} with irrational rotation number α\alpha (see problem 1.43), then TT is related to the rotation TαT_{\alpha } of the circle by the angle α\alpha (i.e., Tα(e2πix)=e2πi(x+α))\left.T_{\alpha }\left(e^{2 \pi i x}\right)=e^{2 \pi i(x+\alpha )}\right) as follows: h∘T=Tα∘hh \circ T=T_{\alpha } \circ h for some continuous mapping h:S1→S1h: S^{1} \rightarrow S^{1}.

?
Remark.
?

The existence of an invariant measure (not necessarily absolutely continuous) is often established by means of the following Bogolyubov-Krylov theorem [KSF]: for every continuous mapping of a compact set into itself there exists a finite invariant measure.

Problem X.2.20

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

A measurable function ff defined on a set XX with measure μ\mu is called invariant under a transformation T:X→XT: X \rightarrow X if f(Tx)=f(x)f(T x)=f(x) for almost all xx. Prove that for TT to be ergodic it is necessary and sufficient that every invariant measurable function coincide with a constant almost everywhere.

?
Problem X.2.21

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

Are the following transformations ergodic:

?
(a)

T(x)=(x+α) mod 1,x∈X=[0,1),μ=λT(x)=(x+\alpha ) \bmod 1, x \in X=[0,1), \mu =\lambda;

(b)

T(x)=(2x) mod 1,x∈X=[0,1),μ=λT(x)=(2 x) \bmod 1, x \in X=[0,1), \mu =\lambda;

(c)

T(z)=zn,z∈X=S1,μT(z)=z^{n}, z \in X=S^{1}, \mu is Lebesgue measure on the circle (n∈N)(n \in \mathbb {N});

(d)

TT is an automorphism of the torus (see problem 2.3);

(e)

TT is an endomorphism of the torus (see problem 2.4)?

Problem X.2.22

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

Let XX be a set in Rn\mathbb {R}^{n} that is the closure of the set of its interior points, TT a transformation of XX preserving Lebesgue measure. Prove that if TT is ergodic, then for almost every x∈Xx \in X the trajectory of the point xx is everywhere dense.

?
Problem X.2.23

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

Let X=[0,1)nX=[0,1)^{n} be the nn-dimensional torus (n∈N),S(n \in \mathbb {N}), S the shift x↦y=S(x),x \mapsto y=S(x), \quad given by the formula yk=(xk+αk) mod 1\quad y_{k}=\left(x_{k}+\alpha_{k}\right) \bmod 1, k=1,…,nk=1, \ldots , n. For which values of α1,…,αn\alpha_{1}, \ldots , \alpha_{n} is the shift ergodic?

?
Problem X.2.24

The measures are assumed to be finite and normalized (probability measures): μ(X)=1\mu (X)=1.

Prove that the maps of the segment [0,1]\left[0,1\right] into itself given by the formulas

T(x)=(2x) mod 1,S(x)=1−∣2x−1∣, T(x)=(2 x) \bmod 1, \quad S(x)=1-\left|2 x-1\right|,

are metrically conjugate. (Maps T:X→XT: X \rightarrow X and S:Y→YS: Y \rightarrow Y preserving measures μ\mu and vv respectively are called metrically conjugate if there exists a measurable map h:X→Yh: X \rightarrow Y such that hμ=vh \mu =v and S∘h(x)=h∘T(x)S \circ h(x)=h \circ T(x) for almost all x∈Xx \in X.)

?
Problem X.2.25

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

Prove that the map TT of the segment [−1,1]\left[-1,1\right] into itself, preserving the measure μ\mu, is ergodic if

?
(a)

T(x)=1−2∣x∣,μ=12dxT(x)=1-2\left|x\right|, \quad \mu =\frac{1}{2} d x;

(b)

T(x)=1−2x2,μ=1π1−x2dxT(x)=1-2 x^{2}, \quad \mu =\frac{1}{\pi \sqrt{1-x^{2}}} d x;

(c)

T=TnT=T_{n} is the Chebyshev polynomial (see 1.16b)), dμ=1π1−x2dxd \mu =\frac{1}{\pi \sqrt{1-x^{2}}} d x.

Problem X.2.26

A measure μ\mu is called a probability measure if it is finite and μ(X)=1\mu (X)=1.

Prove that if T:X→XT: X \rightarrow X is a transformation preserving a probability measure μ\mu, and ff is a positive function, then ∑kf(Tk(x))=+∞\sum_{k} f\left(T^{k}(x)\right)=+\infty for almost all (with respect to μ\mu) x∈Xx \in X.

?
Problem X.2.27

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

Let T:X→XT: X \rightarrow X be an ergodic transformation preserving a probability measure μ,A\mu , A a measurable subset of XX.

Let εn=0\varepsilon_{n}=0 if Tn(x)∉AT^{n}(x) \notin A, and εn=1\varepsilon_{n}=1 if Tn(x)∈A(n⩾0)T^{n}(x) \in A(n \geqslant 0). Prove that for almost all x∈Xx \in X the limit lim⁡1n∑0⩽k<nεk\lim \frac{1}{n} \sum_{0 \leqslant k<n} \varepsilon_{k} exists (the average number of hits of the trajectory into the set AA), and find it.

?
Remark.
?

The Birkhoff--Khinchin ergodic theorem (see [Bi], [KSF]): let TT be a transformation of a set XX preserving a probability measure μ\mu, and let ff be a measurable function such that ∫X∣f(x)∣dμ(x)<+∞\int_{X}\left|f(x)\right| d \mu (x)<+\infty. Then for almost every (with respect to the measure μ\mu) x∈Xx \in X the limit

lim⁡1n∑0⩽k<nf(Tk(x))=fˉ(x) \lim \frac{1}{n} \sum _{0 \leqslant k<n} f\left(T^{k}(x)\right)=\bar{f}(x)

exists, called the mean value of the function ff along the trajectory of the point xx. Moreover the function fˉ\bar{f} is invariant and ∫Xfˉ(x)dμ(x)=∫Xf(x)dμ(x)\int_{X} \bar{f}(x) d \mu (x)=\int_{X} f(x) d \mu (x). If TT is ergodic, then fˉ\bar{f} turns out to be a constant (see problem 2.20), equal to ∫Xf(x)dμ(x)\int_{X} f(x) d \mu (x) (i.e. the mean value of ff on XX).

Problem X.2.28

A measure μ\mu is called a probability measure if it is finite and μ(X)=1\mu (X)=1. Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0. Probability measures μ1,μ2\mu_{1}, \mu_{2} on XX are called mutually singular if there exists a set A⊂XA \subset X such that μ1(A)=1\mu_{1}(A)=1 and μ2(A)=0\mu_{2}(A)=0.

Let μ1\mu_{1} and μ2\mu_{2} be two probability measures on X,TX, T a transformation ergodic with respect to both μ1\mu_{1} and μ2\mu_{2}. Prove that the measures μ1\mu_{1} and μ2\mu_{2} either coincide or are mutually singular.

?
Problem X.2.29

The measure μ\mu is assumed to be finite and to satisfy the normalization condition μ(X)=1\mu (X)=1 (such measures are called probability measures). Let TT be a transformation of the set XX. A set A⊂XA \subset X is called invariant if T−1(A)=AT^{-1}(A)=A. A transformation TT preserving the measure μ\mu is called ergodic if for every invariant set A⊂XA \subset X one of the following two holds: either μ(A)=0\mu (A)=0, or μ(X\A)=0\mu (X \backslash A)=0.

Consider the irrational shift

S(x)=(x+α) mod 1(x∈[0,1)=X,α∈R\Q). S(x)=(x+\alpha ) \bmod 1 \quad (x \in [0,1)=X, \alpha \in \mathbb {R} \backslash \mathbb {Q}) .

It is ergodic by 2.21a). By the ergodic theorem, for any summable function ff on XX and for almost all x∈Xx \in X the equality

lim⁡1n∑0⩽k<nf(Sk(x))=∫01f(t)dt(1) \lim \frac{1}{n} \sum _{0 \leqslant k<n} f\left(S^{k}(x)\right)=\int _{0}^{1} f(t) d t \tag {1}

holds.

?
(a)

Prove that if f∈C([0,1])f \in C(\left[0,1\right]), then for the shift SS one can assert more, namely that equality (1) holds for every x∈[0,1)x \in [0,1). (First assume that f(0)=f(1)f(0)=f(1), and then consider the general case.)

(b)

Prove that if Δ=⟨a,b⟩⊂[0,1),χΔ(x)=1\Delta =\langle a, b\rangle \subset [0,1), \chi_{\Delta }(x)=1 for x∈Δx \in \Delta, χΔ(x)=0\chi_{\Delta }(x)=0 for x∉Δx \notin \Delta, then

lim⁡1n∑0⩽k<nχΔ(Sk(x))=b−a(2) \lim \frac{1}{n} \sum _{0 \leqslant k<n} \chi _{\Delta }\left(S^{k}(x)\right)=b-a \tag {2}

for all x∈[0,1)x \in [0,1).

If {Sk(x)}\left\{ S^{k}(x)\right\} is replaced by an arbitrary sequence {xk}\left\{ x_{k}\right\}, then property b) is called uniform distribution of this sequence (see [PS], [KN]).

For a uniformly distributed sequence {xk}\left\{ x_{k}\right\} and a function ff continuous on [0,1]\left[0,1\right], a relation analogous to (1) holds:

1n∑k=1nf(xk)⟶n→∞∫01f(t)dt \frac{1}{n} \sum _{k=1}^{n} f\left(x_{k}\right) \underset {n \rightarrow \infty }{\longrightarrow } \int _{0}^{1} f(t) d t

In particular,

1n∑k=1ne2πimxk⟶n→∞0 for any integer m≠0.  \frac{1}{n} \sum _{k=1}^{n} e^{2 \pi i m x_{k}} \underset {n \rightarrow \infty }{\longrightarrow } 0 \text{ for any integer } m \neq 0 \text{. }

This condition is not only necessary but also sufficient for uniform distribution. The latter statement is called Weyl's criterion (cf. problem I.3.9).

Remark.
?

The Birkhoff--Khinchin ergodic theorem (see [Bi], [KSF]): let TT be a transformation of a set XX preserving a probability measure μ\mu, and let ff be a measurable function such that ∫X∣f(x)∣dμ(x)<+∞\int_{X}\left|f(x)\right| d \mu (x)<+\infty. Then for almost every (with respect to the measure μ\mu) x∈Xx \in X the limit

lim⁡1n∑0⩽k<nf(Tk(x))=fˉ(x) \lim \frac{1}{n} \sum _{0 \leqslant k<n} f\left(T^{k}(x)\right)=\bar{f}(x)

exists, called the mean value of the function ff along the trajectory of the point xx. Moreover the function fˉ\bar{f} is invariant and ∫Xfˉ(x)dμ(x)=∫Xf(x)dμ(x)\int_{X} \bar{f}(x) d \mu (x)=\int_{X} f(x) d \mu (x). If TT is ergodic, then fˉ\bar{f} turns out to be a constant (see problem 2.20), equal to ∫Xf(x)dμ(x)\int_{X} f(x) d \mu (x) (i.e. the mean value of ff on XX).

Problem X.2.30
?
(a)

Two points move along a circle, jumping once a second by the same angle, which is incommensurable with π\pi. What fraction of a sufficiently long time interval will both points simultaneously lie on some (fixed in advance) semicircle, if at the initial moment the angle between the points was equal to α\alpha?

(b)

Answer the same question if every second one point moves by an angle β1\beta_{1}, and the other by an angle β2\beta_{2}, where the numbers β1,β2\beta_{1}, \beta_{2}, π\pi are linearly independent over Q\mathbb {Q} (see the definition at the beginning of 4 of Chapter VII).

Problem X.2.31

Let x=∑xk2k,xk∈{0,1}x=\sum \frac{x_{k}}{2^{k}}, x_{k} \in \left\{ 0,1\right\} be the binary expansion of a number x∈[0,1)x \in [0,1). Using the ergodicity of the doubling transformation (see problem 2.21b)), prove that for almost all x∈[0,1)x \in [0,1) the limit lim⁡1n∑0⩽k<nxk\lim \frac{1}{n} \sum_{0 \leqslant k<n} x_{k} exists (the frequency of occurrence of the digit one), and find it. Solve the analogous problem for expansions in a base-pp fraction (cf. problems IX.2.14, IX.2.16).

?
Problem X.2.32

Prove that if a transformation T:X→XT: X \rightarrow X preserving a probability measure μ\mu has the property

lim⁡nμ(A∩T−n(B))=μ(A)μ(B)(A,B⊂X)(*) \lim _{n} \mu \left(A \cap T^{-n}(B)\right)=\mu (A) \mu (B) \quad (A, B \subset X) \tag {*}

(the mixing property with respect to the measure μ\mu), then TT is ergodic.

?
Problem X.2.33

Prove that the transformation

T(x)=(2x) mod 1(x∈[0,1)) T(x)=(2 x) \bmod 1(x \in [0,1))

is mixing, while the irrational rotation (see problem 2.29) is not mixing with respect to Lebesgue measure.

?
Problem X.2.34

Let p∈(0,1),q=1−p,μpp \in (0,1), q=1-p, \mu_{p} be the measure on the interval [0,1]\left[0,1\right] defined as follows: if ε1,…,εn∈{0,1}\varepsilon_{1}, \ldots , \varepsilon_{n} \in \left\{ 0,1\right\} and

Δε1…εn={x∈[0,1]∣xk=εk(k=1,…,n)} \Delta _{\varepsilon _{1} \ldots \varepsilon _{n}}=\left\{ x \in [0,1] \mid x_{k}=\varepsilon _{k}(k=1, \ldots , n)\right\}

(here x=∑xk2kx=\sum \frac{x_{k}}{2^{k}} is the binary expansion of the number xx), then

μp(Δε1…εn)=∏1⩽k⩽np1−εkqεk \mu _{p}\left(\Delta _{\varepsilon _{1} \ldots \varepsilon _{n}}\right)=\prod _{1 \leqslant k \leqslant n} p^{1-\varepsilon _{k}} q^{\varepsilon _{k}}

and the measure is extended to the remaining intervals by additivity. Verify that

  1. the transformation T(x)=(2x) mod 1T(x)=(2 x) \bmod 1 is mixing with respect to the measure μp\mu_{p} (see problem 2.32);

  2. μ1/2=λ\mu_{1 / 2}=\lambda;

  3. the measures μp\mu_{p} and λ\lambda are mutually singular for p≠12p \neq \frac{1}{2};

  4. μp1\mu_{p_{1}} is mutually singular with μp2\mu_{p_{2}} for p1≠p2p_{1} \neq p_{2}.

?
Problem X.2.35

Let aka_{k} be the first digit of the decimal representation of the number 2k(k=0,1,…)2^{k} (k=0,1, \ldots ), i.e., a0=1,a1=2,a2=4,a3=8,a4=1,…a_{0}=1, a_{1}=2, a_{2}=4, a_{3}=8, a_{4}=1, \ldots What is the frequency of occurrence of the digit p∈{1,…,9}p \in \left\{ 1, \ldots , 9\right\} in this sequence?

As is well known, every number xx from the interval (0,1](0,1] can be uniquely represented as a continued fraction

x=1a1+1a2+⋯ x=\frac{1}{a_{1}+\frac{1}{a_{2}+\cdots }}

which is also written as x=[a1,a2,…]x=\left[a_{1}, a_{2}, \ldots \right], where a1=a1(x),a2=a2(x)a_{1}=a_{1}(x), a_{2}=a_{2}(x), a3=a3(x),…a_{3}=a_{3}(x), \ldots are natural numbers. Such a fraction contains finitely many terms only when xx is rational. The finite continued fraction [a1,…,an]\left[a_{1}, \ldots , a_{n}\right] is uniquely written as an irreducible fraction pnqn\frac{p_{n}}{q_{n}}. The value of an infinite continued fraction is understood as the limit of the finite ones:

[a1,a2,…]=lim⁡[a1,…,an]=lim⁡pnqn, \left[a_{1}, a_{2}, \ldots \right]=\lim \left[a_{1}, \ldots , a_{n}\right]=\lim \frac{p_{n}}{q_{n}},

the number pnqn=pn(x)qn(x)\frac{p_{n}}{q_{n}}=\frac{p_{n}(x)}{q_{n}(x)} is called the convergent. It is easily verified that

∣x−pn(x)qn(x)∣⩽1qn(x)qn+1(x) \left|x-\frac{p_{n}(x)}{q_{n}(x)}\right| \leqslant \frac{1}{q_{n}(x) q_{n+1}(x)}

A variety of information about continued fractions can be gleaned from the book [X][\mathrm{X}]. Associated with the expansion of numbers into continued fractions is the Gauss transformation

T(x)=1x mod 1(x∈(0,1]). T(x)=\frac{1}{x} \bmod 1(x \in (0,1]) .

Indeed, let x=[a1,a2,…],y=[a2,a3,…]x=\left[a_{1}, a_{2}, \ldots \right], y=\left[a_{2}, a_{3}, \ldots \right]. Then x=1a1+yx=\frac{1}{a_{1}+y}, from which it follows that y=T(x)y=T(x). Using TT, the numbers ana_{n} are written as follows:

a1(x)=[1x],a2(x)=a1(T(x)),…,an(x)=a1(Tn−1(x)). a_{1}(x)=\left[\frac{1}{x}\right], a_{2}(x)=a_{1}(T(x)), \ldots , a_{n}(x)=a_{1}\left(T^{n-1}(x)\right) .

The Gauss transformation preserves the measure dμ=c1+xdxd \mu =\frac{c}{1+x} d x (see problem 2.5). The measure μ\mu is a probability measure for c=1ln⁡2c=\frac{1}{\ln 2}. The reader can learn about interesting applications of ergodic theory to number theory and continued fractions in [Po].

?
Problem X.2.36
?
(a)

For the Gauss transformation TT, establish the validity, for some C>0C>0, of the inequality

1Cμ(A)μ(B)⩽μ(T−n(A)∩B)⩽Cμ(A)μ(B), \frac{1}{C} \mu (A) \mu (B) \leqslant \mu \left(T^{-n}(A) \cap B\right) \leqslant C \mu (A) \mu (B),

where A,B⊂(0,1],n∈NA, B \subset (0,1], n \in \mathbb {N}. Deduce from this the ergodicity of the transformation TT.

(b)

Prove that for almost all x=[a1,a2,…]x=\left[a_{1}, a_{2}, \ldots \right] the following four statements hold:

  1. the relative frequency with which the number k∈Nk \in \mathbb {N} occurs among the numbers a1,a2,…a_{1}, a_{2}, \ldots equals log⁡2(k+1)2k(k+2)\log_{2} \frac{(k+1)^{2}}{k(k+2)};

  2. lim⁡a1+…+ann=+∞\lim \frac{a_{1}+\ldots +a_{n}}{n}=+\infty;

  3. lim⁡a1…ann=∏(1+1k2+2k)log⁡2k\lim \sqrt[n]{a_{1} \ldots a_{n}}=\prod \left(1+\frac{1}{k^{2}+2 k}\right)^{\log_{2} k};

  4. lim⁡qnn=exp⁡(π212ln⁡2)\lim \sqrt[n]{q_{n}}=\exp \left(\frac{\pi^{2}}{12 \ln 2}\right), where qnq_{n} is the denominator of the nn-th convergent.