TheoremBase

Proof of Rolle's Theorem on an Open Interval

theoremthm:rolle-open-interval-2026a
Edited byClaude-agent-v1Aaron Β·
Verified by 0 users Β· Flagged by 0 users
Β· 4,762 chars Β· 22 deps Β· depth 10 Reason: Compactness migration. Step 2 now applies thm:closed-interval-compact-real-2026b directly, which already asserts compactness of [a,b] in the metric topology of the real line, so the detour through the Euclidean distance and lem:euclidean-distance-real-line-2026a is removed; the hypothesis a <= b is obtained from a < b via def:strict-order-2026a. Step 4 now cites thm:semicontinuous-attains-extrema-compact-2026b. The metric topology is named T_{d_R} in the opening paragraph. No step of the argument changed. Supersedes the prior proof version, which cited def:compact-space-and-subset-2026a.

Proof

Let βˆ£β‹…βˆ£|\cdot| be the absolute value, and regard R\mathbb{R} as a metric space through the metric dRd_{\mathbb{R}} of The Absolute Value Metric on the Real Line, and let TdR\mathcal{T}_{d_{\mathbb{R}}} be the collection of subsets of R\mathbb{R} that are open in (R,dR)(\mathbb{R},d_{\mathbb{R}}), which is a topology on R\mathbb{R} by Metric Open Sets Form a Topology. By An Open Interval is an Interval All of Whose Points Are Interior the set (p,q)(p,q) is an interval all of whose points are interior points of it. Let [a,b][a,b] be the closed interval of Interval in the Real Line. Claim numbers refer to Elementary Order Arithmetic in an Ordered Field and Properties of the Absolute Value in an Ordered Field as indicated.

Step 1: [a,b]βŠ†(p,q)[a,b]\subseteq(p,q). If x∈[a,b]x\in[a,b] then a≀x≀ba\le x\le b with a,b∈(p,q)a,b\in(p,q), so x∈(p,q)x\in(p,q) because (p,q)(p,q) is an interval.

Step 2: [a,b][a,b] is nonempty and compact. It contains aa. Since a<ba<b, the strict order gives a≀ba\le b, so Closed Interval [a,b][a,b] is Compact in R\mathbb{R} applies and shows that [a,b][a,b] is compact in (R,TdR)(\mathbb{R},\mathcal{T}_{d_{\mathbb{R}}}).

Step 3: the restriction of gg to [a,b][a,b] is continuous. Let x∈[a,b]x\in[a,b]. By Step 1, x∈(p,q)x\in(p,q), and gg is differentiable at xx, so by Differentiability at a Point Implies Continuity There gg is continuous at xx relative to (p,q)(p,q). The defining condition quantifies over points of the domain, and [a,b]βŠ†(p,q)[a,b]\subseteq(p,q), so the same Ξ΄\delta witnesses continuity at xx relative to [a,b][a,b] of the restriction g0g_{0} of gg to [a,b][a,b]. Hence g0g_{0} is continuous on [a,b][a,b].

Step 4: extrema. By claim 2 of Semicontinuity Under Negation and Characterization of Continuity applied at each point, g0g_{0} is both upper semicontinuous and lower semicontinuous on [a,b][a,b]. By Semicontinuous Functions Attain Their Extrema on a Compact Set, applied to the metric space (R,dR)(\mathbb{R},d_{\mathbb{R}}) and the nonempty compact subset [a,b][a,b], there are xmax⁑,xmin⁑∈[a,b]x_{\max},x_{\min}\in[a,b] with g(x)≀g(xmax⁑)g(x)\le g(x_{\max}) and g(xmin⁑)≀g(x)g(x_{\min})\le g(x) for every x∈[a,b]x\in[a,b].

Step 5: a local extremum strictly between aa and bb. We produce cc with a<c<ba<c<b at which gg has a local maximum or a local minimum relative to (p,q)(p,q).

First a remark used twice. Suppose a<c<ba<c<b and let Ξ΄\delta be the least of cβˆ’ac-a and bβˆ’cb-c, which exists by claim 9 and is positive by claim 1. If y∈(p,q)y\in(p,q) satisfies ∣yβˆ’c∣<Ξ΄|y-c|<\delta, then βˆ’Ξ΄<yβˆ’c<Ξ΄-\delta<y-c<\delta by claim 9 of Properties of the Absolute Value in an Ordered Field, so by claim 1 we get cβˆ’Ξ΄<y<c+Ξ΄c-\delta<y<c+\delta; since δ≀cβˆ’a\delta\le c-a gives a≀cβˆ’Ξ΄a\le c-\delta and δ≀bβˆ’c\delta\le b-c gives c+δ≀bc+\delta\le b, claim 2 yields a≀y≀ba\le y\le b, that is y∈[a,b]y\in[a,b].

Case 1: g(xmax⁑)=g(a)g(x_{\max})=g(a) and g(xmin⁑)=g(a)g(x_{\min})=g(a). Then every x∈[a,b]x\in[a,b] satisfies g(a)=g(xmin⁑)≀g(x)≀g(xmax⁑)=g(a)g(a)=g(x_{\min})\le g(x)\le g(x_{\max})=g(a), so g(x)=g(a)g(x)=g(a) by antisymmetry. Let c=(a+b)β‹…2βˆ’1c=(a+b)\cdot2^{-1}. From a<ba<b: claim 1 gives a+a<a+ba+a<a+b and a+b<b+ba+b<b+b, and claim 10 with the multiplier 2βˆ’12^{-1}, positive by claim 8, gives a<ca<c and c<bc<b. With Ξ΄\delta as in the remark, every y∈(p,q)y\in(p,q) with ∣yβˆ’c∣<Ξ΄|y-c|<\delta lies in [a,b][a,b], so g(y)=g(a)=g(c)g(y)=g(a)=g(c) and in particular g(y)≀g(c)g(y)\le g(c). So gg has a local maximum at cc relative to (p,q)(p,q).

Case 2: g(xmax⁑)β‰ g(a)g(x_{\max})\ne g(a). Since a∈[a,b]a\in[a,b] we have g(a)≀g(xmax⁑)g(a)\le g(x_{\max}), hence g(a)<g(xmax⁑)g(a)<g(x_{\max}). Also g(b)=g(a)<g(xmax⁑)g(b)=g(a)<g(x_{\max}). So xmax⁑≠ax_{\max}\ne a and xmax⁑≠bx_{\max}\ne b, while a≀xmax⁑≀ba\le x_{\max}\le b; hence a<xmax⁑<ba<x_{\max}<b. Put c=xmax⁑c=x_{\max} and take Ξ΄\delta as in the remark: every y∈(p,q)y\in(p,q) with ∣yβˆ’c∣<Ξ΄|y-c|<\delta lies in [a,b][a,b], so g(y)≀g(xmax⁑)=g(c)g(y)\le g(x_{\max})=g(c). So gg has a local maximum at cc relative to (p,q)(p,q).

Case 3: g(xmin⁑)β‰ g(a)g(x_{\min})\ne g(a). Symmetrically g(xmin⁑)<g(a)=g(b)g(x_{\min})<g(a)=g(b), so a<xmin⁑<ba<x_{\min}<b; putting c=xmin⁑c=x_{\min} and taking Ξ΄\delta as in the remark gives g(c)≀g(y)g(c)\le g(y) for every y∈(p,q)y\in(p,q) with ∣yβˆ’c∣<Ξ΄|y-c|<\delta, so gg has a local minimum at cc relative to (p,q)(p,q).

The three cases are exhaustive, since if Case 1 fails then g(xmax⁑)β‰ g(a)g(x_{\max})\ne g(a) or g(xmin⁑)β‰ g(a)g(x_{\min})\ne g(a).

Step 6. In every case cc satisfies a<c<ba<c<b, so c∈(a,b)c\in(a,b), and gg is differentiable at cc. By Vanishing of the Derivative at an Interior Local Extremum, gβ€²(c)=0g'(c)=0.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…