TheoremBase

For f(a) <= y <= f(b), the supremum c of the points where f is at most y satisfies f(c) = y by epsilon-delta continuity at c on both sides; the other case passes to -f. The image of an interval follows by applying this to the restriction of f to a closed subinterval.

Proof

Throughout, dR(s,t)=∣s−t∣d_{\mathbb{R}}(s,t)=|s-t| for s,t∈Rs,t\in\mathbb{R}, by The Distance on the Real Numbers (i); for a≤ba\le b, t∈[a,b]t\in[a,b] if and only if t∈Rt\in\mathbb{R} and a≤t≤ba\le t\le b, by Intervals of the Real Line (ii); and for u,v∈Ru,v\in\mathbb{R}, Rules of Arithmetic and Order in an Ordered Field (x) gives −∣u∣≤u≤∣u∣-|u|\le u\le|u|, ∣−u∣=∣u∣|-u|=|u|, and ∣u∣≤v|u|\le v if and only if −v≤u≤v-v\le u\le v. In particular, if 0≤u0\le u then −u≤0≤u-u\le0\le u by Rules of Arithmetic and Order in an Ordered Field (v), and so ∣u∣≤u|u|\le u.

Proof of (i).

Case 1: f(a)≤y≤f(b)f(a)\le y\le f(b). This case is proved for every map f:[a,b]→Rf:[a,b]\to\mathbb{R} continuous on [a,b][a,b] and every such yy. Let S={x∈[a,b]:f(x)≤y}S=\{x\in[a,b]:f(x)\le y\} (Typed Terms under a Predicative Guard: Variables, Parameters, Values of Maps and Ordered Pairs, and Predicative Equations and Memberships between Them (vii)), a subset of R\mathbb{R}. As a≤a≤ba\le a\le b and f(a)≤yf(a)\le y, a∈Sa\in S, so S≠∅S\neq\emptyset; and bb is an upper bound of SS, so SS is bounded above. By The Real Numbers, with the Natural Numbers, Integers and Rationals Identified with Subsets of the Reals, and Completeness (vii), SS has a supremum, unique by Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order (iv); let c=sup⁡Sc=\sup S. By Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order (iv) and Bounds, Least and Greatest Elements, Suprema and Infima for a Partial Order (iii), cc is an upper bound of SS, and c≤uc\le u for every upper bound uu of SS. Hence a≤ca\le c, as a∈Sa\in S, and c≤bc\le b, as bb is an upper bound of SS; so c∈[a,b]c\in[a,b]. By Continuous Map Between Metric Spaces (ii), ff is continuous at cc relative to [a,b][a,b], so by Continuous Map Between Metric Spaces (i), for every real ε>0\varepsilon>0 there is a real δ>0\delta>0 such that every x∈[a,b]x\in[a,b] with ∣c−x∣<δ|c-x|<\delta satisfies ∣f(x)−f(c)∣<ε|f(x)-f(c)|<\varepsilon. In each of Steps 1 and 2, ε\varepsilon is fixed first, from cc and yy; then δ\delta is taken for this ε\varepsilon; and then the points tested against δ\delta.

Step 1: f(c)≤yf(c)\le y. Suppose instead that y<f(c)y<f(c). Put ε=f(c)−y\varepsilon=f(c)-y, so 0<ε0<\varepsilon, and take δ\delta for this ε\varepsilon as above. For x∈[a,b]x\in[a,b] with ∣c−x∣<δ|c-x|<\delta we get −(f(x)−f(c))≤∣f(x)−f(c)∣<ε-(f(x)-f(c))\le|f(x)-f(c)|<\varepsilon, and hence y=f(c)−ε<f(x)y=f(c)-\varepsilon<f(x). Let s∈Ss\in S; then s≤cs\le c, as cc is an upper bound of SS. If c−δ<sc-\delta<s, then 0≤c−s<δ0\le c-s<\delta, so ∣c−s∣≤c−s<δ|c-s|\le c-s<\delta, and since s∈[a,b]s\in[a,b] this gives y<f(s)y<f(s), contradicting f(s)≤yf(s)\le y. Hence s≤c−δs\le c-\delta, the order of R\mathbb{R} being total. So c−δc-\delta is an upper bound of SS, whence c≤c−δc\le c-\delta and δ≤0\delta\le0, contradicting 0<δ0<\delta. Therefore f(c)≤yf(c)\le y, again by totality.

Step 2: y≤f(c)y\le f(c). Suppose instead that f(c)<yf(c)<y. Then c≠bc\neq b, for otherwise f(b)=f(c)<y≤f(b)f(b)=f(c)<y\le f(b); with c≤bc\le b this gives c<bc<b. Put ε=y−f(c)\varepsilon=y-f(c), so 0<ε0<\varepsilon, and take δ\delta for this ε\varepsilon as above. Put h=δ/(1+1)h=\delta/(1+1); by Rules of Arithmetic and Order in an Ordered Field (ix), applied to 0<δ0<\delta, 0<h<δ0<h<\delta. Choose x0x_{0} as follows: if c+h≤bc+h\le b, let x0=c+hx_{0}=c+h, so that x0−c=hx_{0}-c=h; otherwise b<c+hb<c+h, and let x0=bx_{0}=b, so that x0−c<hx_{0}-c<h. In both cases c<x0≤bc<x_{0}\le b and 0<x0−c<δ0<x_{0}-c<\delta; hence a≤c<x0a\le c<x_{0} gives x0∈[a,b]x_{0}\in[a,b], and ∣c−x0∣=∣−(x0−c)∣=∣x0−c∣≤x0−c<δ|c-x_{0}|=|-(x_{0}-c)|=|x_{0}-c|\le x_{0}-c<\delta. So f(x0)−f(c)≤∣f(x0)−f(c)∣<ε=y−f(c)f(x_{0})-f(c)\le|f(x_{0})-f(c)|<\varepsilon=y-f(c), that is f(x0)<yf(x_{0})<y. Thus x0∈Sx_{0}\in S, and x0≤cx_{0}\le c as cc is an upper bound of SS, contradicting c<x0c<x_{0}. Therefore y≤f(c)y\le f(c).

Steps 1 and 2 give f(c)=yf(c)=y, which proves Case 1.

Case 2: f(b)≤y≤f(a)f(b)\le y\le f(a). Let k:[a,b]→Rk:[a,b]\to\mathbb{R} be the map x↦(−1)⋅f(x)x\mapsto(-1)\cdot f(x) (Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (v)), so k(x)=−f(x)k(x)=-f(x) by Rules of Arithmetic and Order in an Ordered Field (ii). Apply Continuity of Sums and Products of Real-Valued Functions on a Metric Space (v) with (X,d)=(R,dR)(X,d)=(\mathbb{R},d_{\mathbb{R}}), A=[a,b]A=[a,b], the point x=ax=a, both of its maps ff and gg taken to be our ff (continuous at aa relative to [a,b][a,b], being continuous on [a,b][a,b]), and its constant cc equal to −1-1; its map cfcf is then kk, so kk is continuous on [a,b][a,b]. By Rules of Arithmetic and Order in an Ordered Field (v), k(a)=−f(a)≤−y≤−f(b)=k(b)k(a)=-f(a)\le-y\le-f(b)=k(b). Case 1, applied to kk and −y-y, gives c∈[a,b]c\in[a,b] with −f(c)=k(c)=−y-f(c)=k(c)=-y, and hence f(c)=yf(c)=y.

Proof of (ii). The image f(I)f(I) is a subset of R\mathbb{R}, and for every vv, v∈f(I)v\in f(I) if and only if v=f(x)v=f(x) for some x∈Ix\in I, by Sets and Maps: Ordinary Notation (vii), The Image and the Preimage of a Class under a Class (i) and Values, Restriction and Equality of a Map, and the Map u↦t(u)u\mapsto t(u) as a Defined Set Symbol (i). By Intervals of the Real Line (i) we must show: if u,w∈f(I)u,w\in f(I) and v∈Rv\in\mathbb{R} satisfy u≤v≤wu\le v\le w, then v∈f(I)v\in f(I). Let x,z∈Ix,z\in I with u=f(x)u=f(x) and w=f(z)w=f(z).

If x=zx=z, then u=wu=w, so v=uv=u and v∈f(I)v\in f(I). Otherwise let ss be the lesser and tt the greater of xx and zz, so s<ts<t and s,t∈Is,t\in I. Every r∈[s,t]r\in[s,t] satisfies s≤r≤ts\le r\le t, hence r∈Ir\in I by Intervals of the Real Line (i); so [s,t]⊆I[s,t]\subseteq I. Let g=f∣[s,t]g=f|_{[s,t]} be the restriction of ff to [s,t][s,t]; it is a map g:[s,t]→Rg:[s,t]\to\mathbb{R} with g(r)=f(r)g(r)=f(r) for every r∈[s,t]r\in[s,t], by Domains and Restrictions of Maps (ii) and Domains and Restrictions of Maps (iii).

As ff is continuous on II, Restricting a Continuous Map to a Smaller Subset Preserves Continuity (ii), applied with (X,dX)=(Y,dY)=(R,dR)(X,d_X)=(Y,d_Y)=(\mathbb{R},d_{\mathbb{R}}), A=IA=I, B=[s,t]B=[s,t] and our ff, shows that gg is continuous on [s,t][s,t].

If x<zx<z, then s=xs=x, t=zt=z, and g(s)=u≤v≤w=g(t)g(s)=u\le v\le w=g(t); if z<xz<x, then s=zs=z, t=xt=x, and g(t)=u≤v≤w=g(s)g(t)=u\le v\le w=g(s). In either case (i), applied to s<ts<t, gg and vv, gives c∈[s,t]c\in[s,t] with g(c)=vg(c)=v. Then c∈Ic\in I and f(c)=g(c)=vf(c)=g(c)=v, so v∈f(I)v\in f(I). Hence f(I)f(I) is an interval.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…