TheoremBase

Uniform convergence gives the uniform Cauchy condition by an epsilon/2 argument; conversely each sequence of values is Cauchy, hence convergent, and passing to the limit in the Cauchy estimate gives uniform convergence to the map of pointwise limits, which is unique by uniqueness of limits.

Proof

Two facts on R\mathbb{R} are used repeatedly. For real Ξ΅>0\varepsilon>0, the number Ξ΅/2\varepsilon/2 satisfies 0<Ξ΅/2<Ξ΅0<\varepsilon/2<\varepsilon and Ξ΅/2+Ξ΅/2=Ξ΅\varepsilon/2+\varepsilon/2=\varepsilon, by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares (vii). For s,t,u∈Rs,t,u\in\mathbb{R}, ∣sβˆ’tβˆ£β‰€βˆ£sβˆ’u∣+∣uβˆ’t∣|s-t|\le|s-u|+|u-t| by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares (xi), and ∣sβˆ’t∣=∣tβˆ’s∣|s-t|=|t-s|.

Step 1: a uniform limit is a pointwise limit. Suppose that (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly to a map g:Sβ†’Rg:S\to\mathbb{R} on SS. Let x∈Sx\in S and let Ξ΅>0\varepsilon>0 be real. By Pointwise and Uniform Convergence of a Sequence of Real-Valued Maps (ii), read with E=SE=S and gg in place of ff, there is K∈NK\in\mathbb{N} with ∣fk(y)βˆ’g(y)∣<Ξ΅|f_{k}(y)-g(y)|<\varepsilon for every k∈Nk\in\mathbb{N} with kβ‰₯Kk\ge K and every y∈Sy\in S; in particular ∣fk(x)βˆ’g(x)∣<Ξ΅|f_{k}(x)-g(x)|<\varepsilon for every kβ‰₯Kk\ge K. As Ξ΅\varepsilon was arbitrary, (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} converges to g(x)g(x) by Convergent Sequences of Real Numbers (i); as x∈Sx\in S was arbitrary, (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges pointwise to gg on SS by Pointwise and Uniform Convergence of a Sequence of Real-Valued Maps (i).

Step 2: the map of the limits. Suppose that (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} is convergent for every x∈Sx\in S. Then Sequences of Real-Valued Maps: the Map of Pointwise Limits on a Subset, and the Sequence of Pointwise Partial Sums (i), applied to the set DD, the subset SS and the sequence (fk)k∈N(f_{k})_{k\in\mathbb{N}} in RD\mathbb{R}^{D}, gives exactly one map L:Sβ†’RL:S\to\mathbb{R} with L(x)=lim⁑kβ†’βˆžfk(x)L(x)=\lim_{k\to\infty}f_{k}(x) for every x∈Sx\in S; this is the map x↦lim⁑kβ†’βˆžfk(x)x\mapsto\lim_{k\to\infty}f_{k}(x), and (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} converges to L(x)L(x) for every x∈Sx\in S by The Limit of a Convergent Sequence (i). If g:Sβ†’Rg:S\to\mathbb{R} and (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges pointwise to gg on SS, then for every x∈Sx\in S the sequence (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} converges to g(x)g(x) by Pointwise and Uniform Convergence of a Sequence of Real-Valued Maps (i) and to L(x)L(x), so g(x)=L(x)g(x)=L(x) by Limits of Sequences of Real Numbers: Uniqueness, Boundedness, Constants, Tails, Arithmetic, Quotients, Absolute Values, Finite Sums, Order, Squeezing, Domination and Subsequences (ii); hence g=Lg=L (Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (iv)).

Proof of (ii). Suppose that (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly to a map h:Sβ†’Rh:S\to\mathbb{R} on SS. By Step 1 it converges pointwise to hh on SS, so (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} converges to h(x)h(x), and is convergent, for every x∈Sx\in S. Step 2 now provides the map L=(x↦lim⁑kβ†’βˆžfk(x))L=(x\mapsto\lim_{k\to\infty}f_{k}(x)) and shows that every g:Sβ†’Rg:S\to\mathbb{R} to which (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges pointwise on SS equals LL; in particular h=Lh=L, so (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly to LL on SS. Finally, if (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly to g:Sβ†’Rg:S\to\mathbb{R} on SS, it converges pointwise to gg on SS by Step 1, so g=Lg=L.

Proof of (i), from left to right. Suppose that (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly to f:Sβ†’Rf:S\to\mathbb{R} on SS, and let Ξ΅>0\varepsilon>0 be real. Apply Pointwise and Uniform Convergence of a Sequence of Real-Valued Maps (ii), read with E=SE=S, to the real number Ξ΅/2>0\varepsilon/2>0: there is K∈NK\in\mathbb{N} with ∣fk(x)βˆ’f(x)∣<Ξ΅/2|f_{k}(x)-f(x)|<\varepsilon/2 for every k∈Nk\in\mathbb{N} with kβ‰₯Kk\ge K and every x∈Sx\in S. Let m,k∈Nm,k\in\mathbb{N} with mβ‰₯Km\ge K and kβ‰₯Kk\ge K, and let x∈Sx\in S. Then ∣fm(x)βˆ’f(x)∣<Ξ΅/2|f_{m}(x)-f(x)|<\varepsilon/2 and ∣f(x)βˆ’fk(x)∣=∣fk(x)βˆ’f(x)∣<Ξ΅/2|f(x)-f_{k}(x)|=|f_{k}(x)-f(x)|<\varepsilon/2, so

∣fm(x)βˆ’fk(x)βˆ£β‰€βˆ£fm(x)βˆ’f(x)∣+∣f(x)βˆ’fk(x)∣<Ξ΅/2+Ξ΅/2=Ξ΅,|f_{m}(x)-f_{k}(x)|\le|f_{m}(x)-f(x)|+|f(x)-f_{k}(x)|<\varepsilon/2+\varepsilon/2=\varepsilon ,

the strict inequality by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares (ii) and the two together by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares (i). So KK is as required.

Proof of (i), from right to left. Suppose that for every real Ξ΅>0\varepsilon>0 there is K∈NK\in\mathbb{N} with ∣fm(x)βˆ’fk(x)∣<Ξ΅|f_{m}(x)-f_{k}(x)|<\varepsilon for all m,k∈Nm,k\in\mathbb{N} with mβ‰₯Km\ge K and kβ‰₯Kk\ge K and every x∈Sx\in S.

First, for every x∈Sx\in S the sequence (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} is convergent. Indeed, for every real Ξ΅>0\varepsilon>0 the KK of the hypothesis gives ∣fm(x)βˆ’fk(x)∣<Ξ΅|f_{m}(x)-f_{k}(x)|<\varepsilon for all m,k∈Nm,k\in\mathbb{N} with m,kβ‰₯Km,k\ge K, so (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} is a Cauchy sequence by Cauchy Sequences of Real Numbers (i), hence convergent by Completeness of the Real Numbers for Sequences: Monotone Convergence, the Bolzano-Weierstrass Theorem and Cauchy Sequences (v). Let L:Sβ†’RL:S\to\mathbb{R}, L(x)=lim⁑kβ†’βˆžfk(x)L(x)=\lim_{k\to\infty}f_{k}(x), be the map of Step 2; (fk(x))k∈N(f_{k}(x))_{k\in\mathbb{N}} converges to L(x)L(x) for every x∈Sx\in S.

We show that (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly to LL on SS. The choices are made in this order: first a real Ξ΅>0\varepsilon>0 is given; then K∈NK\in\mathbb{N} is taken from the hypothesis applied to Ξ΅/2\varepsilon/2, once for all points of SS; then k∈Nk\in\mathbb{N} with kβ‰₯Kk\ge K and x∈Sx\in S are arbitrary. So ∣fm(x)βˆ’fk(x)∣<Ξ΅/2|f_{m}(x)-f_{k}(x)|<\varepsilon/2 for every m∈Nm\in\mathbb{N} with mβ‰₯Km\ge K, that is, βˆ’Ξ΅/2≀fm(x)βˆ’fk(x)≀Ρ/2-\varepsilon/2\le f_{m}(x)-f_{k}(x)\le\varepsilon/2 by Rules of Arithmetic and Order in an Ordered Field (x), and so fk(x)βˆ’Ξ΅/2≀fm(x)≀fk(x)+Ξ΅/2f_{k}(x)-\varepsilon/2\le f_{m}(x)\le f_{k}(x)+\varepsilon/2 by Rules of Arithmetic and Order in an Ordered Field (iv). Since (fm(x))m∈N(f_{m}(x))_{m\in\mathbb{N}} converges to L(x)L(x), Limits of Sequences of Real Numbers: Uniqueness, Boundedness, Constants, Tails, Arithmetic, Quotients, Absolute Values, Finite Sums, Order, Squeezing, Domination and Subsequences (x), applied with fk(x)+Ξ΅/2f_{k}(x)+\varepsilon/2 and fk(x)βˆ’Ξ΅/2f_{k}(x)-\varepsilon/2 in place of Ξ»\lambda, gives fk(x)βˆ’Ξ΅/2≀L(x)≀fk(x)+Ξ΅/2f_{k}(x)-\varepsilon/2\le L(x)\le f_{k}(x)+\varepsilon/2. Hence βˆ’Ξ΅/2≀fk(x)βˆ’L(x)≀Ρ/2-\varepsilon/2\le f_{k}(x)-L(x)\le\varepsilon/2 by Rules of Arithmetic and Order in an Ordered Field (iv), so ∣fk(x)βˆ’L(x)βˆ£β‰€Ξ΅/2|f_{k}(x)-L(x)|\le\varepsilon/2 by Rules of Arithmetic and Order in an Ordered Field (x), and therefore ∣fk(x)βˆ’L(x)∣<Ξ΅|f_{k}(x)-L(x)|<\varepsilon by Inequalities in an Ordered Field: Mixed Transitivity, Strict Sums, Signs, Products, Natural Numbers, Halving, Reciprocals, Absolute Values and Squares (i). This is the condition of Pointwise and Uniform Convergence of a Sequence of Real-Valued Maps (ii), read with E=SE=S and LL in place of ff; so (fk)k∈N(f_{k})_{k\in\mathbb{N}} converges uniformly on SS to the map L:Sβ†’RL:S\to\mathbb{R}.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…