TheoremBase

Density supplies sequences from the subset; uniform continuity makes their images Cauchy, hence convergent, and the limit is independent of the sequence, which defines the extension; the modulus and the bound pass to the limit.

Proof

Each result cited is universally quantified over the data in its own statement.

Step 0 (two facts used repeatedly). First, for every x∈Xx\in X there is a sequence in SS converging to xx in (X,d)(X,d): since SS is dense, the closure of SS in (X,Td)(X,\mathcal{T}_{d}) is XX, so xx belongs to that closure, and Sequential Characterization of the Closure in a Metric Space supplies the sequence. Secondly, if a sequence (am)m∈N(a_{m})_{m\in\mathbb{N}} of real numbers converges to aa and c∈Rc\in\mathbb{R} satisfies am≀ca_{m}\le c for every m∈Nm\in\mathbb{N}, then a≀ca\le c: otherwise Ξ·=aβˆ’c\eta=a-c is positive, and convergence gives an mm with ∣amβˆ’a∣<Ξ·|a_{m}-a|<\eta, hence am>aβˆ’Ξ·=ca_{m}>a-\eta=c by claim 9 of Properties of the Absolute Value in an Ordered Field, contradicting am≀ca_{m}\le c.

Claim 2. Let g1,g2g_{1},g_{2} be as stated and let x∈Xx\in X. By Step 0 choose a sequence (xm)m∈N(x_{m})_{m\in\mathbb{N}} in SS converging to xx. By Continuity Between Metric Spaces is Equivalent to Sequential Continuity the sequences (g1(xm))m∈N(g_{1}(x_{m}))_{m\in\mathbb{N}} and (g2(xm))m∈N(g_{2}(x_{m}))_{m\in\mathbb{N}} converge to g1(x)g_{1}(x) and to g2(x)g_{2}(x) respectively. These are the same sequence of real numbers, because g1g_{1} and g2g_{2} agree on SS and every xmx_{m} lies in SS; since limits are unique by The Real Numbers: Standing Notation and Background (iii), g1(x)=g2(x)g_{1}(x)=g_{2}(x).

Claim 1. Let f:Sβ†’Rf:S\to\mathbb{R} be uniformly continuous on SS, let x∈Xx\in X and let (xm)m∈N(x_{m})_{m\in\mathbb{N}} be a sequence in SS converging to xx.

(i) The sequence (f(xm))m∈N(f(x_{m}))_{m\in\mathbb{N}} converges. Let Ρ∈R\varepsilon\in\mathbb{R} be positive. By uniform continuity of ff on SS there is a positive ρ\rho such that all a,b∈Sa,b\in S with d(a,b)<ρd(a,b)<\rho satisfy ∣f(a)βˆ’f(b)∣<Ξ΅|f(a)-f(b)|<\varepsilon. By convergence there is N∈NN\in\mathbb{N} with d(xm,x)<ρ2d(x_{m},x)<\tfrac{\rho}{2} for every mm with N≀mN\le m, and then d(xm,xn)≀d(xm,x)+d(x,xn)<ρd(x_{m},x_{n})\le d(x_{m},x)+d(x,x_{n})<\rho for all such m,nm,n, whence ∣f(xm)βˆ’f(xn)∣<Ξ΅|f(x_{m})-f(x_{n})|<\varepsilon. Thus (f(xm))m∈N(f(x_{m}))_{m\in\mathbb{N}} is a Cauchy sequence, and it converges by The Real Numbers: Standing Notation and Background (iii).

(ii) The limit does not depend on the sequence. Let (ym)m∈N(y_{m})_{m\in\mathbb{N}} be a sequence in SS converging to xx, let aa be the limit of (f(xm))m∈N(f(x_{m}))_{m\in\mathbb{N}} and bb that of (f(ym))m∈N(f(y_{m}))_{m\in\mathbb{N}}, and let Ξ΅\varepsilon be positive. With ρ\rho as in (i) there is NN such that d(xm,x)<ρ2d(x_{m},x)<\tfrac{\rho}{2} and d(ym,x)<ρ2d(y_{m},x)<\tfrac{\rho}{2}, hence d(xm,ym)<ρd(x_{m},y_{m})<\rho and ∣f(xm)βˆ’f(ym)∣<Ξ΅|f(x_{m})-f(y_{m})|<\varepsilon, for every mm with N≀mN\le m; enlarging NN we may also assume ∣f(xN)βˆ’a∣<Ξ΅|f(x_{N})-a|<\varepsilon and ∣f(yN)βˆ’b∣<Ξ΅|f(y_{N})-b|<\varepsilon. Then, by the triangle inequality of claim 5 of Properties of the Absolute Value in an Ordered Field,

∣aβˆ’bβˆ£β‰€βˆ£aβˆ’f(xN)∣+∣f(xN)βˆ’f(yN)∣+∣f(yN)βˆ’b∣<3Ξ΅.|a-b|\le|a-f(x_{N})|+|f(x_{N})-f(y_{N})|+|f(y_{N})-b|<3\varepsilon .

As Ξ΅\varepsilon was an arbitrary positive real and 3Ξ΅3\varepsilon ranges over all positive reals with it, Comparison of Real Numbers with Arbitrary Positive Slack gives ∣aβˆ’bβˆ£β‰€0|a-b|\le0, and since 0β‰€βˆ£aβˆ’b∣0\le|a-b| we get a=ba=b.

By (i) and (ii), for each x∈Xx\in X there is exactly one real number LL such that (f(xm))m∈N(f(x_{m}))_{m\in\mathbb{N}} converges to LL for every sequence (xm)m∈N(x_{m})_{m\in\mathbb{N}} in SS converging to xx: such an LL exists because at least one such sequence exists by Step 0 and (ii) identifies the limits of all of them, and it is unique for the same reason. Let g:Xβ†’Rg:X\to\mathbb{R} be the function whose value at xx is that number.

(iii) gg agrees with ff on SS. For x∈Sx\in S the sequence all of whose terms are xx lies in SS and converges to xx in (X,d)(X,d), because d(x,x)=0d(x,x)=0 is smaller than every positive real, and the sequence of its values, all equal to f(x)f(x), converges to f(x)f(x) by Constant Sequences and Index-Shifted Sequences of Real Numbers (i); hence g(x)=f(x)g(x)=f(x) by the definition of gg.

(iv) gg is uniformly continuous, hence continuous, on XX. Let Ξ΅\varepsilon be positive and choose a positive ρ\rho such that all a,b∈Sa,b\in S with d(a,b)<ρd(a,b)<\rho satisfy ∣f(a)βˆ’f(b)∣<Ξ΅2|f(a)-f(b)|<\tfrac{\varepsilon}{2}. Let x,y∈Xx,y\in X satisfy d(x,y)<ρ3d(x,y)<\tfrac{\rho}{3} and choose sequences (xm)m∈N(x_{m})_{m\in\mathbb{N}} and (ym)m∈N(y_{m})_{m\in\mathbb{N}} in SS converging to xx and to yy. There is NN with d(xm,x)<ρ3d(x_{m},x)<\tfrac{\rho}{3} and d(ym,y)<ρ3d(y_{m},y)<\tfrac{\rho}{3} for every mm with N≀mN\le m, so that d(xm,ym)≀d(xm,x)+d(x,y)+d(y,ym)<ρd(x_{m},y_{m})\le d(x_{m},x)+d(x,y)+d(y,y_{m})<\rho and hence ∣f(xm)βˆ’f(ym)βˆ£β‰€Ξ΅2|f(x_{m})-f(y_{m})|\le\tfrac{\varepsilon}{2} for all such mm. Replacing the two sequences by the subsequences with terms xm+Nx_{m+N} and ym+Ny_{m+N}, which converge to the same limits by A Subsequence of a Convergent Sequence Has the Same Limit, we may assume that this bound holds for every m∈Nm\in\mathbb{N}. The sequences (f(xm))m∈N(f(x_{m}))_{m\in\mathbb{N}} and (f(ym))m∈N(f(y_{m}))_{m\in\mathbb{N}} converge to g(x)g(x) and g(y)g(y) by the definition of gg, so by claim 3 of Arithmetic of Limits of Real Sequences and claim 4 of Order Properties of Limits of Real Sequences the sequence with terms ∣f(xm)βˆ’f(ym)∣|f(x_{m})-f(y_{m})| converges to ∣g(x)βˆ’g(y)∣|g(x)-g(y)|. By Step 0, ∣g(x)βˆ’g(y)βˆ£β‰€Ξ΅2<Ξ΅|g(x)-g(y)|\le\tfrac{\varepsilon}{2}<\varepsilon. Thus the condition of Uniformly Continuous Map Between Metric Spaces holds with ρ3\tfrac{\rho}{3}, so gg is uniformly continuous on XX; taking, for a given point and a given Ξ΅\varepsilon, the same number ρ3\tfrac{\rho}{3} shows that gg is continuous on XX.

Claim 3. Let gg, Ξ΅\varepsilon and ΞΈ\theta be as stated and let x,y∈Xx,y\in X satisfy d(x,y)<ΞΈd(x,y)<\theta; put ΞΌ=ΞΈβˆ’d(x,y)\mu=\theta-d(x,y), a positive real. Choose sequences (xm)m∈N(x_{m})_{m\in\mathbb{N}} and (ym)m∈N(y_{m})_{m\in\mathbb{N}} in SS converging to xx and to yy, and NN with d(xm,x)<ΞΌ2d(x_{m},x)<\tfrac{\mu}{2} and d(ym,y)<ΞΌ2d(y_{m},y)<\tfrac{\mu}{2} for N≀mN\le m. For such mm,

d(xm,ym)≀d(xm,x)+d(x,y)+d(y,ym)<d(x,y)+ΞΌ=ΞΈ,d(x_{m},y_{m})\le d(x_{m},x)+d(x,y)+d(y,y_{m})<d(x,y)+\mu=\theta ,

so d(xm,ym)≀θd(x_{m},y_{m})\le\theta and the hypothesis gives ∣g(xm)βˆ’g(ym)βˆ£β‰€Ξ΅|g(x_{m})-g(y_{m})|\le\varepsilon, both points lying in SS. Passing to the subsequences with terms xm+Nx_{m+N}, ym+Ny_{m+N} as in (iv), this holds for every index. By Continuity Between Metric Spaces is Equivalent to Sequential Continuity the sequences (g(xm))m∈N(g(x_{m}))_{m\in\mathbb{N}} and (g(ym))m∈N(g(y_{m}))_{m\in\mathbb{N}} converge to g(x)g(x) and g(y)g(y), so by claim 3 of Arithmetic of Limits of Real Sequences and claim 4 of Order Properties of Limits of Real Sequences the sequence with terms ∣g(xm)βˆ’g(ym)∣|g(x_{m})-g(y_{m})| converges to ∣g(x)βˆ’g(y)∣|g(x)-g(y)|; Step 0 gives ∣g(x)βˆ’g(y)βˆ£β‰€Ξ΅|g(x)-g(y)|\le\varepsilon.

Claim 4. Let gg and CC be as stated and let x∈Xx\in X. Choose a sequence (xm)m∈N(x_{m})_{m\in\mathbb{N}} in SS converging to xx. Then ∣g(xm)βˆ£β‰€C|g(x_{m})|\le C for every mm, the points xmx_{m} lying in SS, and by Continuity Between Metric Spaces is Equivalent to Sequential Continuity and claim 4 of Order Properties of Limits of Real Sequences the sequence with terms ∣g(xm)∣|g(x_{m})| converges to ∣g(x)∣|g(x)|. By Step 0, ∣g(x)βˆ£β‰€C|g(x)|\le C.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…