Alessandro Palliccia

Real and Functional Analysis

Cardinality, topology and measurable spaces

1,069 words,

The lecture recalls set theory and topology, and then begins real analysis with measurable spaces.

Recap of set theory

The recap follows Hrbacek and Jech. Its aim is to distinguish between sets with infinitely many elements.

Notation. A⊂BA \subset B means that AA is contained in BB or equal to it, while A⊊BA \subsetneq B means that AA is strictly contained in B.B.

Cardinality

Definition 1.1 (Comparison of cardinalities). Let XX and YY be sets. We write #X≤#Y,{\#X \le \#Y,} and say that the cardinality of XX is less than or equal to the cardinality of Y,Y, if there exists an injective map f ⁣:X→Y.{f \colon X \to Y.} We write #X=#Y{\#X = \#Y} if there exists a bijective map f ⁣:X→Y.{f \colon X \to Y.} Finally, #X<#Y{\#X < \#Y} means that #X≤#Y{\#X \le \#Y} and #X≠#Y.{\#X \neq \#Y.}

Remark 1.2. Let XX and YY be sets.
(i) If XX is nonempty, then #X≤#Y{\#X \le \#Y} if and only if there exists a surjective map f ⁣:Y→X.{f \colon Y \to X.}
(ii) (Cantor–Bernstein) If #X≤#Y{\#X \le \#Y} and #Y≤#X,{\#Y \le \#X,} then #X=#Y.{\#X = \#Y.}

Definition 1.3 (Countable and uncountable sets). Let XX be a set.
(i) XX is countable if XX is finite or #X=#N.{\#X = \#\mathbb{N}.}
(ii) XX is uncountable if XX is infinite but #N<#X.{\#\mathbb{N} < \#X.}

Example 1.4. The natural numbers, the integers and the rationals have the same cardinality, #N=#Z=#Q,{\#\mathbb{N} = \#\mathbb{Z} = \#\mathbb{Q},} so Z\mathbb{Z} and Q\mathbb{Q} are countable. The real line is uncountable: #R>#N.{\#\mathbb{R} > \#\mathbb{N}.}

Definition 1.5. Define the set

R‾=[−∞,+∞]:=R∪{+∞}∪{−∞},\overline{\mathbb{R}} = [-\infty, +\infty] := \mathbb{R} \cup \{+\infty\} \cup \{-\infty\},

with the following operations in R‾:\overline{\mathbb{R}}{:}

a∈R⟹a+∞=+∞,a−∞=−∞,a>0⟹a⋅(+∞)=+∞,a⋅(−∞)=−∞,a<0⟹a⋅(+∞)=−∞,a⋅(−∞)=+∞,⟹0⋅(+∞)=0⋅(−∞)=0.\begin{aligned} &a \in \mathbb{R} & &\Longrightarrow \quad a + \infty = +\infty, \quad a - \infty = -\infty, \\ &a > 0 & &\Longrightarrow \quad a \cdot (+\infty) = +\infty, \quad a \cdot (-\infty) = -\infty, \\ &a < 0 & &\Longrightarrow \quad a \cdot (+\infty) = -\infty, \quad a \cdot (-\infty) = +\infty, \\ & & &\phantom{\Longrightarrow \quad} 0 \cdot (+\infty) = 0 \cdot (-\infty) = 0. \end{aligned}

The power set

Definition 1.6 (Power set). Let XX be a set. The power set of XX is the family of all subsets of X,X,

P(X)={A:A⊂X}.\mathcal{P}(X) = \{ A : A \subset X \}.

Remark 1.7. If #X=n∈N,{\#X = n \in \mathbb{N},} then #P(X)=2n.{\#\mathcal{P}(X) = 2^n.}

Notation. For a set X,X, 2X2^X is the set of all functions from XX to the set with 2 elements {0,1}:\{0, 1\}{:}

2X:={f ⁣:X→{0,1}}.2^X := \{ f \colon X \to \{0, 1\} \}.

Proposition 1.8. Let XX be a set.
(i) #P(X)=#2X.{\#\mathcal{P}(X) = \#2^X.}
(ii) #P(X)>#X.{\#\mathcal{P}(X) > \#X.} In particular, there exists no biggest infinity.

Proof (complete). (i) Define

ϕ ⁣:P(X)→2X,A↦fA ⁣:X→{0,1},fA(x):={1x∈A,0x∉A.\begin{aligned} \phi \colon \mathcal{P}(X) &\to 2^X, \\ A &\mapsto f_A \colon X \to \{0, 1\}, \\ f_A(x) &:= \begin{cases} 1 & x \in A, \\ 0 & x \notin A. \end{cases} \end{aligned}

The map ϕ\phi is a bijection.

(ii) Suppose, by contradiction, that #P(X)≤#X.{\#\mathcal{P}(X) \le \#X.} By Remark 1.2 (i), applied to the nonempty set P(X),\mathcal{P}(X), there exists a surjective map f ⁣:X→P(X).{f \colon X \to \mathcal{P}(X).} Define

U:={x∈X:x∉f(x)}⊂X.U := \{ x \in X : x \notin f(x) \} \subset X.

Since ff is surjective, there exists y∈X{y \in X} with f(y)=U.{f(y) = U.} Is y∈U?{y \in U?}
Case 1. If y∈U,{y \in U,} then y∉f(y){y \notin f(y)} by the definition of U,U, and f(y)=U{f(y) = U} gives y∉U,{y \notin U,} which is impossible.
Case 2. If y∉U=f(y),{y \notin U = f(y),} then y∈U{y \in U} by the definition of U,U, which is impossible.

Hence #P(X)≤#X{\#\mathcal{P}(X) \le \#X} is false. To see that this gives #P(X)>#X,{\#\mathcal{P}(X) > \#X,} note that x↦{x}{x \mapsto \{x\}} is injective, so #X≤#P(X),{\#X \le \#\mathcal{P}(X),} and #X=#P(X){\#X = \#\mathcal{P}(X)} would give #P(X)≤#X.{\#\mathcal{P}(X) \le \#X.} ■\blacksquare

Countable sets

Notation. A countable union is written ⋃n=0∞Xn,\bigcup_{n=0}^{\infty} X_n, also ⋃nXn.\bigcup_n X_n. A sequence is written (xn),(x_n), and a family of sets is written F={Xα}α∈A.{\mathcal{F} = \{ X_\alpha \}_{\alpha \in A}.}

Proposition 1.9. (i) #R=#P(N)>#N.{\#\mathbb{R} = \#\mathcal{P}(\mathbb{N}) > \#\mathbb{N}.}
(ii) If XX and YY are countable, then X×Y={(x,y):x∈X,y∈Y}{X \times Y = \{ (x, y) : x \in X, y \in Y \}} is countable.
(iii) If X0,X1,X2,…{X_0, X_1, X_2, \dots} is a sequence of sets such that XnX_n is countable for every n,n, then ⋃n=0∞Xn\bigcup_{n=0}^{\infty} X_n is countable: a countable union of countable sets is countable.

Recap of topology

The recap follows Manetti. Topology gives an abstract theory of the open, closed and related subsets of some ambient space.

Topological spaces

Definition 1.10 (Topology). Let XX be a set. A topology on XX is a family of subsets τ⊂P(X){\tau \subset \mathcal{P}(X)} such that:
(i) ∅,X∈τ;{\emptyset, X \in \tau;}
(ii) for every F={Uα}α∈A⊂τ,{\mathcal{F} = \{ U_\alpha \}_{\alpha \in A} \subset \tau,} the union ⋃α∈AUα\bigcup_{\alpha \in A} U_\alpha belongs to τ;\tau;
(iii) if U1,U2,…,Uk∈τ,{U_1, U_2, \dots, U_k \in \tau,} then ⋂n=1kUn∈τ.{\bigcap_{n=1}^{k} U_n \in \tau.}
The sets in τ\tau are called open, and the couple (X,τ)(X, \tau) is called a topological space.

Example 1.11. On Rn\mathbb{R}^n the Euclidean topology τE\tau_E is defined as follows. We say that U∈τE{U \in \tau_E} if for every x∈U{x \in U} there exists r>0{r > 0} with Br(x)⊂U.{B_r(x) \subset U.} Here Br(x)B_r(x) is the open ball of centre xx and radius r.r. Two other topologies on Rn\mathbb{R}^n are τ={∅,Rn}{\tau = \{ \emptyset, \mathbb{R}^n \}} and τ=P(Rn).{\tau = \mathcal{P}(\mathbb{R}^n).}

Definition 1.12 (Closed set, closure, interior, boundary). Let (X,τ)(X, \tau) be a topological space. A set V⊂X{V \subset X} is closed if X∖V{X \setminus V} is open. For V⊂X:{V \subset X{:}}
(i) the closure V‾\overline{V} of VV is the smallest closed set containing V;V;
(ii) the interior int⁡(V)=V˚{\operatorname{int}(V) = \mathring{V}} of VV is the biggest open set contained in V;V;
(iii) the boundary of VV is ∂V=V‾∩X∖V‾.{\partial V = \overline{V} \cap \overline{X \setminus V}.}

Definition 1.13 (Compact set). Let (X,τ)(X, \tau) be a topological space and K⊂X.{K \subset X.} The set KK is compact if for every F={Uα}α∈A⊂τ{\mathcal{F} = \{ U_\alpha \}_{\alpha \in A} \subset \tau} such that ⋃α∈AUα⊃K{\bigcup_{\alpha \in A} U_\alpha \supset K} there exists a finite subfamily F~={V1,…,Vm}⊂F{\tilde{\mathcal{F}} = \{ V_1, \dots, V_m \} \subset \mathcal{F}} such that ⋃n=1mVn⊃K.{\bigcup_{n=1}^{m} V_n \supset K.}

Figure 1.1. KK is a triangle, the disks are open sets whose union contains K,K, and a finite subfamily is then highlighted whose union still contains K.K. The button draws a new family.

Definition 1.14 (Precompact set). A set V⊂X{V \subset X} is precompact if V‾\overline{V} is compact.

Theorem 1.15 (Heine–Borel). A set K⊂Rn{K \subset \mathbb{R}^n} is compact if and only if KK is closed and bounded.

Example 1.16 (Topology on R‾\overline{\mathbb{R}}). The following subsets of R‾\overline{\mathbb{R}} are open in R‾:\overline{\mathbb{R}}{:} the intervals (a,b)(a, b) with a,b∈R;{a, b \in \mathbb{R};} every set that is open in R,\mathbb{R}, which is also defined to be open in R‾;\overline{\mathbb{R}}; and the sets [−∞,a)[-\infty, a) and (a,+∞](a, +\infty] for every a∈R,{a \in \mathbb{R},} which are also defined to be open.

Definition 1.17 (Continuous map). Let (X,τX)(X, \tau_X) and (Y,τY)(Y, \tau_Y) be two topological spaces and f ⁣:X→Y.{f \colon X \to Y.} The map ff is continuous if f−1(A)∈τX{f^{-1}(A) \in \tau_X} for every A∈τY,{A \in \tau_Y,} that is, if the preimage of any open set is open.

Metric spaces

Definition 1.18 (Metric space). A metric space is a couple (X,d)(X, d) where XX is a set and dd is a distance on X,X, that is, a function d ⁣:X×X→[0,+∞){d \colon X \times X \to [0, +\infty)} such that, for all x,y,z∈X,{x, y, z \in X,}

d(x,y)=0  ⟺  x=y,d(x,y)=d(y,x),d(x,y)≤d(x,z)+d(z,y).\begin{aligned} &d(x, y) = 0 \iff x = y, \\ &d(x, y) = d(y, x), \\ &d(x, y) \le d(x, z) + d(z, y). \end{aligned}

Definition 1.19 (Topology induced by a distance). Let (X,d)(X, d) be a metric space. For x∈X{x \in X} and r>0,{r > 0,} define

Br(x):={y∈X:d(x,y)<r}.B_r(x) := \{ y \in X : d(x, y) < r \}.

A set UU in XX is open if for every x∈U{x \in U} there exists r>0{r > 0} such that Br(x)⊂U.{B_r(x) \subset U.} These open sets form the topology induced by the distance d.d.

Figure 1.2. An open set UU of the plane with the Euclidean distance, its edge dashed because it does not belong to U,U, and a ball Br(x)B_r(x) contained in UU around a moving point x.x.

A sequence (xn)⊂X{(x_n) \subset X} converges to x0∈X,{x_0 \in X,} written xn→x0,{x_n \to x_0,} if lim⁡n→∞d(xn,x0)=0.{\lim_{n \to \infty} d(x_n, x_0) = 0.}

Lemma 1.20. Let (X,d)(X, d) be a metric space.
(i) A set V⊂X{V \subset X} is closed if and only if VV is sequentially closed: if (xn)⊂V{(x_n) \subset V} and xn→x0,{x_n \to x_0,} then x0∈V.{x_0 \in V.}
(ii) A set K⊂X{K \subset X} is compact if and only if KK is sequentially compact: every (xn)⊂K{(x_n) \subset K} has a subsequence (xnk)(x_{n_k}) such that xnk→x0∈K.{x_{n_k} \to x_0 \in K.}
(iii) If VV is closed, KK is compact and V⊂K,{V \subset K,} then VV is compact.

Figure 1.3. Lemma 1.20 (i): a closed set V,V, its edge solid because it belongs to V,V, and a sequence of points of VV that converges to x0.x_0. The limit x0x_0 lies in V.V.

Lemma 1.21. Let (X,dX)(X, d_X) and (Y,dY)(Y, d_Y) be two metric spaces and f ⁣:X→Y.{f \colon X \to Y.} The following are equivalent.
(i) ff is continuous.
(ii) For every x0∈X{x_0 \in X} and every ε>0{\varepsilon > 0} there exists δ>0{\delta > 0} such that dX(x,x0)<δ{d_X(x, x_0) < \delta} implies dY(f(x),f(x0))<ε.{d_Y\big(f(x), f(x_0)\big) < \varepsilon.}
(iii) For every x0∈X{x_0 \in X} and every (xn)⊂X{(x_n) \subset X} with xn→x0,{x_n \to x_0,} that is, dX(xn,x0)→0,{d_X(x_n, x_0) \to 0,} the sequence (f(xn))⊂Y{\big(f(x_n)\big) \subset Y} is such that f(xn)→f(x0),{f(x_n) \to f(x_0),} that is, dY(f(xn),f(x0))→0.{d_Y\big(f(x_n), f(x_0)\big) \to 0.}

Measurable spaces

Real analysis begins here, following Folland.

σ-algebras

Definition 1.22 (σ\sigma-algebra, measurable space). Let XX be a set. A measurable space is a couple (X,M)(X, \mathcal{M}) where M\mathcal{M} is a σ\sigma-algebra, that is:
(i) M⊂P(X);{\mathcal{M} \subset \mathcal{P}(X);}
(ii) ∅∈M;{\emptyset \in \mathcal{M};}
(iii) for every A∈M,{A \in \mathcal{M},} X∖A∈M;{X \setminus A \in \mathcal{M};}
(iv) for every (An)⊂M,{(A_n) \subset \mathcal{M},} ⋃nAn∈M.{\bigcup_n A_n \in \mathcal{M}.}

Example 1.23. For every set X,X, P(X)\mathcal{P}(X) and M={∅,X}{\mathcal{M} = \{ \emptyset, X \}} are σ\sigma-algebras. Most of the times a topology is not a σ\sigma-algebra, for example the Euclidean topology.

Remark 1.24. Let (X,M)(X, \mathcal{M}) be a measurable space.
(i) If E,F∈M,{E, F \in \mathcal{M},} then E∩F∈M{E \cap F \in \mathcal{M}} and E∖F∈M.{E \setminus F \in \mathcal{M}.}
(ii) If (En)⊂M,{(E_n) \subset \mathcal{M},} then ⋂nEn∈M.{\bigcap_n E_n \in \mathcal{M}.}

Proof (sketch). Intersections are written through complements and unions:

E∩F=X∖(X∖(E∩F))=X∖((X∖E)∪(X∖F)).■E \cap F = X \setminus \big( X \setminus (E \cap F) \big) = X \setminus \big( (X \setminus E) \cup (X \setminus F) \big). \tag*{$\blacksquare$}

Definition 1.25 (Restricted σ\sigma-algebra). Let (X,M)(X, \mathcal{M}) be a measurable space and Ω⊂X.{\Omega \subset X.} The restricted σ\sigma-algebra is

MΩ=M∣Ω:={E∩Ω:E∈M}.\mathcal{M}_\Omega = \mathcal{M}|_\Omega := \{ E \cap \Omega : E \in \mathcal{M} \}.

One checks that (Ω,MΩ)(\Omega, \mathcal{M}_\Omega) is a measurable space. Indeed, ∅=∅∩Ω,{\emptyset = \emptyset \cap \Omega,} Ω∖(E∩Ω)=(X∖E)∩Ω{\Omega \setminus (E \cap \Omega) = (X \setminus E) \cap \Omega} and ⋃n(En∩Ω)=(⋃nEn)∩Ω.{\bigcup_n (E_n \cap \Omega) = \big(\bigcup_n E_n\big) \cap \Omega.}

Generated and Borel σ-algebras

Theorem 1.26 (Generation of σ\sigma-algebras). Let XX be a set and F⊂P(X).{\mathcal{F} \subset \mathcal{P}(X).} There exists a σ\sigma-algebra M\mathcal{M} on XX such that:
(i) F⊂M;{\mathcal{F} \subset \mathcal{M};}
(ii) for every σ\sigma-algebra Q\mathcal{Q} on XX with F⊂Q,{\mathcal{F} \subset \mathcal{Q},} M⊂Q.{\mathcal{M} \subset \mathcal{Q}.}
In words, M\mathcal{M} is the smallest σ\sigma-algebra containing F.\mathcal{F}. It is denoted σ0(F).\sigma_0(\mathcal{F}).

Proof (sketch). Consider

W:={Q⊂P(X):Q is a σ-algebra, F⊂Q}.W := \{ \mathcal{Q} \subset \mathcal{P}(X) : \mathcal{Q} \text{ is a } \sigma\text{-algebra}, \ \mathcal{F} \subset \mathcal{Q} \}.

Then W≠∅,{W \neq \emptyset,} as P(X)∈W.{\mathcal{P}(X) \in W.} Define

M:=⋂{Q:Q∈W}.■\mathcal{M} := \bigcap \{ \mathcal{Q} : \mathcal{Q} \in W \}. \tag*{$\blacksquare$}

Definition 1.27 (Borel σ\sigma-algebra). Let (X,τ)(X, \tau) be a topological space. The Borel σ\sigma-algebra is σ0(τ),\sigma_0(\tau), denoted B(X).\mathcal{B}(X).

Remark 1.28. For R\mathbb{R} with the Euclidean topology,

B(R)=σ0({open sets})=σ0({(a,b):a<b})=σ0({[a,b]:a<b})=σ0({(a,+∞):a∈R}).\begin{aligned} \mathcal{B}(\mathbb{R}) &= \sigma_0\big( \{ \text{open sets} \} \big) \\ &= \sigma_0\big( \{ (a, b) : a < b \} \big) \\ &= \sigma_0\big( \{ [a, b] : a < b \} \big) \\ &= \sigma_0\big( \{ (a, +\infty) : a \in \mathbb{R} \} \big). \end{aligned}

Similarly,

B(R‾)=σ0({(a,+∞]:a∈R}),B(RN)=σ0({open rectangles}),\begin{aligned} \mathcal{B}(\overline{\mathbb{R}}) &= \sigma_0\big( \{ (a, +\infty] : a \in \mathbb{R} \} \big), \\ \mathcal{B}(\mathbb{R}^N) &= \sigma_0\big( \{ \text{open rectangles} \} \big), \end{aligned}

where an open rectangle is a set (a1,b1)×(a2,b2)×⋯×(aN,bN).{(a_1, b_1) \times (a_2, b_2) \times \dots \times (a_N, b_N).}