TheoremBase

The metric axioms follow pointwise from those of dYd_Y by evaluating at a point where the maximum is attained; constant maps are continuous; for completeness, a positive lower bound of w makes a dw−Cauchyd_w-Cauchy sequence uniformly Cauchy, its pointwise limit is a uniform limit, hence continuous by an epsilon/3 argument, and an upper bound of w gives dw−convergenced_w-convergence.

Proof

Write C=C([a,b],Y)C=C([a,b],Y). By The Space of Continuous Maps on a Closed Interval and Its Weighted Uniform Distance (i), the elements of CC are exactly the maps [a,b]→Y[a,b]\to Y continuous on [a,b][a,b], and by The Space of Continuous Maps on a Closed Interval and Its Weighted Uniform Distance (ii), dwd_{w} is a map C×C→RC\times C\to\mathbb{R}. Throughout, The Space of Continuous Maps on a Closed Interval and Its Weighted Uniform Distance (iii) is used in the form: for u,v∈Cu,v\in C, w(t) dY(u(t),v(t))≤dw(u,v)w(t)\,d_{Y}(u(t),v(t))\le d_{w}(u,v) for every t∈[a,b]t\in[a,b], with equality at some t0∈[a,b]t_{0}\in[a,b].

The metric ((i)). Let x,y,z∈Cx,y,z\in C.

Nonnegative. a∈[a,b]a\in[a,b] by Basic Facts about Intervals of the Real Line (v), w(a)>0w(a)>0 and dY(x(a),y(a))≥0d_{Y}(x(a),y(a))\ge0 by Metric Space (ii); so 0≤w(a) dY(x(a),y(a))≤dw(x,y)0\le w(a)\,d_{Y}(x(a),y(a))\le d_{w}(x,y).

Zero. If x=yx=y, take t0t_{0} with dw(x,x)=w(t0) dY(x(t0),x(t0))d_{w}(x,x)=w(t_{0})\,d_{Y}(x(t_{0}),x(t_{0})); this is 00, as dY(x(t0),x(t0))=0d_{Y}(x(t_{0}),x(t_{0}))=0 by Metric Space (iii). Conversely, let dw(x,y)=0d_{w}(x,y)=0 and t∈[a,b]t\in[a,b]. Then 0≤w(t) dY(x(t),y(t))≤dw(x,y)=00\le w(t)\,d_{Y}(x(t),y(t))\le d_{w}(x,y)=0, so w(t) dY(x(t),y(t))=0w(t)\,d_{Y}(x(t),y(t))=0; as w(t)>0w(t)>0, dY(x(t),y(t))=0d_{Y}(x(t),y(t))=0, and x(t)=y(t)x(t)=y(t) by Metric Space (iii). Hence x=yx=y (Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (iv)).

Symmetric. Take t0t_{0} with dw(x,y)=w(t0) dY(x(t0),y(t0))d_{w}(x,y)=w(t_{0})\,d_{Y}(x(t_{0}),y(t_{0})). By Metric Space (iv) this equals w(t0) dY(y(t0),x(t0))≤dw(y,x)w(t_{0})\,d_{Y}(y(t_{0}),x(t_{0}))\le d_{w}(y,x). Exchanging xx and yy gives dw(y,x)≤dw(x,y)d_{w}(y,x)\le d_{w}(x,y), so dw(x,y)=dw(y,x)d_{w}(x,y)=d_{w}(y,x).

Triangle. Take t0t_{0} with dw(x,z)=w(t0) dY(x(t0),z(t0))d_{w}(x,z)=w(t_{0})\,d_{Y}(x(t_{0}),z(t_{0})). By Metric Space (v) and w(t0)>0w(t_{0})>0,

dw(x,z)≤w(t0) dY(x(t0),y(t0))+w(t0) dY(y(t0),z(t0))≤dw(x,y)+dw(y,z).d_{w}(x,z)\le w(t_{0})\,d_{Y}(x(t_{0}),y(t_{0}))+w(t_{0})\,d_{Y}(y(t_{0}),z(t_{0}))\le d_{w}(x,y)+d_{w}(y,z).

So dwd_{w} is a metric on CC by Metric Space (i), and (C,dw)(C,d_{w}) is a metric space by Metric Space (vi).

Nonempty ((ii)). Let c∈Yc\in Y and let h:[a,b]→Yh:[a,b]\to Y be the map with h(s)=ch(s)=c for every s∈[a,b]s\in[a,b], which exists by Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (v). For all s,t∈[a,b]s,t\in[a,b], dY(h(s),h(t))=dY(c,c)=0d_{Y}(h(s),h(t))=d_{Y}(c,c)=0 by Metric Space (iii). So for every t∈[a,b]t\in[a,b] and every real ε>0\varepsilon>0, δ=1\delta=1 has the property required in Continuous Map Between Metric Spaces (i); hence hh is continuous on [a,b][a,b] by Continuous Map Between Metric Spaces (ii), and h∈Ch\in C.

Completeness ((iii)). Let (Y,dY)(Y,d_{Y}) be complete, and let (xk)k∈N(x_{k})_{k\in\mathbb{N}} be a Cauchy sequence in (C,dw)(C,d_{w}).

Step 1: bounds for the weight. By Extreme Value Theorem on a Closed Real Interval (ii) applied to ww, there are p,q∈[a,b]p,q\in[a,b] with w(p)≤w(t)≤w(q)w(p)\le w(t)\le w(q) for every t∈[a,b]t\in[a,b]. Put m=w(p)m=w(p) and M=w(q)M=w(q); then m>0m>0 and M>0M>0. For u,v∈Cu,v\in C and t∈[a,b]t\in[a,b], dY(u(t),v(t))≥0d_{Y}(u(t),v(t))\ge0 by Metric Space (ii), so m dY(u(t),v(t))≤w(t) dY(u(t),v(t))≤dw(u,v)m\,d_{Y}(u(t),v(t))\le w(t)\,d_{Y}(u(t),v(t))\le d_{w}(u,v), that is,

dY(u(t),v(t))≤dw(u,v)/m.(1)d_{Y}(u(t),v(t))\le d_{w}(u,v)/m.\qquad(1)

Step 2: the pointwise limit. As C⊆Y[a,b]C\subseteq Y^{[a,b]}, the set of maps from [a,b][a,b] to YY, (xk)(x_{k}) is a sequence in Y[a,b]Y^{[a,b]}. Let t∈[a,b]t\in[a,b]. By Sections of a Sequence of Maps into a Metric Space, and the Map of Their Limits (i), with [a,b][a,b] for its DD, the map k↦xk(t)k\mapsto x_{k}(t) is a sequence (xk(t))k(x_{k}(t))_{k} in YY. It is a Cauchy sequence in (Y,dY)(Y,d_{Y}): given a real ε>0\varepsilon>0, the Cauchy property of (xk)(x_{k}) applied to mε>0m\varepsilon>0 gives N∈NN\in\mathbb{N} with dw(xk,xl)<mεd_{w}(x_{k},x_{l})<m\varepsilon for all k,l∈Nk,l\in\mathbb{N} with k,l≥Nk,l\ge N, and then dY(xk(t),xl(t))<εd_{Y}(x_{k}(t),x_{l}(t))<\varepsilon by (1). As (Y,dY)(Y,d_{Y}) is complete, (xk(t))k(x_{k}(t))_{k} is convergent in (Y,dY)(Y,d_{Y}) by Complete Metric Space (i). By Sections of a Sequence of Maps into a Metric Space, and the Map of Their Limits (ii), with [a,b][a,b] for its DD and SS, there is exactly one map x:[a,b]→Yx:[a,b]\to Y such that (xk(t))k(x_{k}(t))_{k} converges to x(t)x(t) in (Y,dY)(Y,d_{Y}) for every t∈[a,b]t\in[a,b], and x(t)x(t) is the limit lim⁡k→∞xk(t)\lim_{k\to\infty}x_{k}(t).

Step 3: the convergence is uniform. We show: for every real ε>0\varepsilon>0 there is N∈NN\in\mathbb{N} with

dY(xk(t),x(t))<εfor all k∈N with k≥N and all t∈[a,b].(2)d_{Y}(x_{k}(t),x(t))<\varepsilon\qquad\text{for all }k\in\mathbb{N}\text{ with }k\ge N\text{ and all }t\in[a,b].\qquad(2)

The choices are made in this order. First ε\varepsilon is given. (Here and in Steps 4 and 5, the quotients mε/2m\varepsilon/2, ε/2\varepsilon/2, ε/3\varepsilon/3 and ε/(2M)\varepsilon/(2M) are real numbers, positive as m,M>0m,M>0, and are computed in the type Real\mathrm{Real} of The Real Numbers as a Type (i), by Differences, Quotients and Absolute Values of Real Numbers, Read in the Type Real (ii).) Then the Cauchy property of (xk)(x_{k}) applied to mε/2>0m\varepsilon/2>0 gives N∈NN\in\mathbb{N} with dw(xk,xl)<mε/2d_{w}(x_{k},x_{l})<m\varepsilon/2 for all k,l≥Nk,l\ge N, so that, by (1), dY(xk(t),xl(t))<ε/2d_{Y}(x_{k}(t),x_{l}(t))<\varepsilon/2 for all k,l≥Nk,l\ge N and all t∈[a,b]t\in[a,b]; this NN does not depend on tt. Now let k≥Nk\ge N and t∈[a,b]t\in[a,b]. By Step 2 and Convergent Sequence in a Metric Space (i) there is N′∈NN'\in\mathbb{N}, depending on tt, with dY(xl(t),x(t))<ε/2d_{Y}(x_{l}(t),x(t))<\varepsilon/2 for all l≥N′l\ge N'. Let ll be the larger of NN and N′N', which exists by Arithmetic and Order of the Natural Numbers (vii); then l∈Nl\in\mathbb{N}, N≤lN\le l and N′≤lN'\le l by Arithmetic and Order of the Natural Numbers (vi). By Metric Space (v),

dY(xk(t),x(t))≤dY(xk(t),xl(t))+dY(xl(t),x(t))<ε/2+ε/2=ε.d_{Y}(x_{k}(t),x(t))\le d_{Y}(x_{k}(t),x_{l}(t))+d_{Y}(x_{l}(t),x(t))<\varepsilon/2+\varepsilon/2=\varepsilon.

Step 4: the limit is continuous. Let t∈[a,b]t\in[a,b] and let a real ε>0\varepsilon>0 be given. First, (2) applied to ε/3\varepsilon/3 gives N∈NN\in\mathbb{N}, and, with k=Nk=N (as N≤NN\le N by Arithmetic and Order of the Natural Numbers (vi)), dY(xN(s),x(s))<ε/3d_{Y}(x_{N}(s),x(s))<\varepsilon/3 for every s∈[a,b]s\in[a,b]. Then, as xN∈Cx_{N}\in C is continuous on [a,b][a,b], it is continuous at tt relative to [a,b][a,b] by Continuous Map Between Metric Spaces (ii), and Continuous Map Between Metric Spaces (i) gives a real δ>0\delta>0 with dY(xN(s),xN(t))<ε/3d_{Y}(x_{N}(s),x_{N}(t))<\varepsilon/3 for every s∈[a,b]s\in[a,b] with dR(t,s)<δd_{\mathbb{R}}(t,s)<\delta. For such ss, by Metric Space (v) applied twice and Metric Space (iv),

dY(x(s),x(t))≤dY(xN(s),x(s))+dY(xN(s),xN(t))+dY(xN(t),x(t))<ε.d_{Y}(x(s),x(t))\le d_{Y}(x_{N}(s),x(s))+d_{Y}(x_{N}(s),x_{N}(t))+d_{Y}(x_{N}(t),x(t))<\varepsilon.

So xx is continuous at tt relative to [a,b][a,b]; as tt was arbitrary, xx is continuous on [a,b][a,b] by Continuous Map Between Metric Spaces (ii), and x∈Cx\in C.

Step 5: convergence in dwd_{w}. Let a real ε>0\varepsilon>0 be given, and let NN be given by (2) for ε/(2M)\varepsilon/(2M). Let k≥Nk\ge N. For every t∈[a,b]t\in[a,b], as w(t)≤Mw(t)\le M and dY(xk(t),x(t))≥0d_{Y}(x_{k}(t),x(t))\ge0,

w(t) dY(xk(t),x(t))≤M dY(xk(t),x(t))≤M⋅ε2M=ε2.w(t)\,d_{Y}(x_{k}(t),x(t))\le M\,d_{Y}(x_{k}(t),x(t))\le M\cdot\frac{\varepsilon}{2M}=\frac{\varepsilon}{2}.

Taking t0t_{0} with dw(xk,x)=w(t0) dY(xk(t0),x(t0))d_{w}(x_{k},x)=w(t_{0})\,d_{Y}(x_{k}(t_{0}),x(t_{0})) gives dw(xk,x)≤ε/2<εd_{w}(x_{k},x)\le\varepsilon/2<\varepsilon. Hence (xk)(x_{k}) converges to xx in (C,dw)(C,d_{w}) by Convergent Sequence in a Metric Space (i), and is convergent by Convergent Sequence in a Metric Space (ii).

As (xk)(x_{k}) was an arbitrary Cauchy sequence in the metric space (C,dw)(C,d_{w}) of (i), (C,dw)(C,d_{w}) is complete by Complete Metric Space (i).

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…