Metric and Hilbert Spaces

Arun Ram
Department of Mathematics and Statistics
University of Melbourne
Parkville, VIC 3010 Australia
aram@unimelb.edu.au

Last updated: 4 November 2014

Lecture 14: Examples of complete spaces

Completions

Let (X,d) be a metric space.

The completion of (X,d) is a metric space (Xˆ,dˆ) with an isometry φ:X→Xˆ such that (Xˆ,dˆ) is complete and φ(X)‾=Xˆ.

HW: (Uniqueness of completions). If (Xˆ1,dˆ1) with φ1:X→Xˆ1 and (Xˆ2,dˆ2) with φ2:X→Xˆ2 are completions of (X,d) then there exists f:Xˆ1→Xˆ2 such that

(a) f is an isometry,
(b) f is a bijection,
(c) f∘φ1=φ2.
Xˆ1 φ1↗ ↓f X ↘φ2 Xˆ2

Existence of completions

Let (X,d) be a metric space.

Let Xˆ be the set of Cauchy sequences x⇀:ℤ>0→X with x⇀=y⇀ if limn→∞d(xn,yn)=0, where x⇀: ℤ>0 ⟶ X n ⟼ xn and y⇀: ℤ>0 ⟶ X n ⟼ yn .

Define dˆ:Xˆ×Xˆ→ℝ≥0 by d(x⇀,y⇀)= limn→∞ d(xn,yn), where x⇀: ℤ>0 ⟶ X n ⟼ xn and y⇀: ℤ>0 ⟶ X n ⟼ yn .

Define φ:X→Xˆ by φ(x)=(x,x,…), i.e. φ(x): ℤ>0 ⟶ X n ⟼ x .

(Xˆ,dˆ) with φ:X→Xˆ is a completion of (X,d).

Proof.

To show:
(a) (Xˆ,dˆ) is a metric space.
(b) (Xˆ,dˆ) is complete.
(c) φ:X→Xˆ is an isometry.
(d) φ(X)‾=Xˆ.
(c) To show: If x,y∈X then d(φ(x),φ(y))=d(x,y).
Assume x,y∈X.
To show: d(φ(x),φ(y))=d(x,y). d(φ(x),φ(y)) = limn→∞d(φ(x)n,φ(y)n) = limn→∞d(x,y) = d(x,y). So φ is an isometry.
(a)
To show: (Xˆ,dˆ) is a metric space.
To show:
(aa) dˆ:Xˆ×Xˆ→ℝ≥0 given by dˆ(x⇀,y⇀)=limn→∞d(xn,yn), is a function.
(ab) If x⇀,y⇀∈Xˆ then dˆ(x⇀,y⇀)=dˆ(y⇀,x⇀).
(ac) If x⇀∈Xˆ then dˆ(x⇀,x⇀)=0.
(ad) If x⇀,y⇀∈Xˆ and dˆ(x⇀,y⇀)=0 then x⇀=y⇀.
(ae) If x⇀,y⇀,z⇀∈Xˆ then dˆ(x⇀,y⇀)≤dˆ(x⇀,z⇀)+dˆ(z⇀,y⇀).
(aa) To show: If x⇀,y⇀∈Xˆ then there exists a unique z∈ℝ≥0 such that z=limn→∞d(xn,yn).
Assume x⇀,y⇀∈Xˆ with x⇀=(x1,x2,…) and y⇀=(y1,y2,…).
Let d1,d2,… be the sequence in ℝ≥0 given by dn=d(xn,yn). To show: There exists z∈ℝ≥0 such that z=limn→∞dn.
Since limits in metric spaces are unique when they exist, z will be unique if it exists, see Rubinstein Notes Proposition 2.9.
To show: d1,d2,… is a Cauchy sequence in ℝ≥0.
This will show that z exists since Cauchy sequences in ℝ≥0 converge, since ℝ≥0 is complete, see Rubinstein notes Theorem 4.6.
To show: If ε∈ℝ>0 then there exists N∈ℤ>0 such that if m,n∈ℤ>0 and m>N and n>N then ∣dm-dn∣<ε.
Assume ε∈ℝ>0.
Let N=max(N1,N2), where N1 is such that if n,m>N1 then d(xm,xn)<ε2,
N2 is such that if n,m>N2 then d(ym,yn)<ε2.
(N1 and N2 exist since x⇀ and y⇀ are Cauchy sequences).
To show: If m,n∈ℤ>0 and m>N and N>N then ∣dm-dn∣<ε.
Assume m,n∈ℤ>0 and m>N and n>N.
To show: ∣dm-dn∣<ε. ∣dm-dn∣ = ∣ d(xm,ym)- d(xn,yn) ∣ < ∣ d(xn,xm)+ d(yn,ym) ∣ , since d(xn,yn)≤d(xn,xm)+d(xm,ym)+d(yn,ym).
So ∣dm-dn∣< ε2+ε2=ε. So d1,d2,… is a Cauchy sequence in ℝ≥0.
So z=limn→∞dn exists in ℝ≥0.

□

Notes and References

These are a typed copy of Lecture 14 from a series of handwritten lecture notes for the class Metric and Hilbert Spaces given on August 20, 2014.

page history