Search the corpus

    Derivative zero on an interval implies equal endpoint values

    P-BERK83SU-04

    Area
    Prelim
    Seen at
    BERKELEY
    Years
    1983
    Status
    draft

    Outline a proof, starting from basic properties of the real numbers, of the following theorem.

    Let \(f:[a,b]\to\mathbb R\) be continuous and differentiable on \((a,b)\), with \[ f'(x)=0\qquad(x\in(a,b)). \] Then \(f(b)=f(a)\).

    Solution

    We use the least-upper-bound property of \(\RR\) as the basic completeness property.

    1. Every sequence in a closed bounded interval \([a,b]\) has aconvergent subsequence with limit in \([a,b]\).

    1.1. From any infinite sequence \((x_n)\) in \([a,b]\), one can constructnested closed intervals\[ I_1\supseteq I_2\supseteq\cdots \]such that \(I_k\) contains infinitely many terms of the sequence and\[ \operatorname{length}(I_k)=\frac{b-a}{2^k}. \]

    Bisect \([a,b]\). At least one half contains infinitely many terms of the sequence; call such a half \(I_1\). Bisect \(I_1\) and choose a half containing infinitely many terms; call it \(I_2\). Continue inductively.

    1.2. The nested intervals of step 1 have exactly one common point\(c\in[a,b]\).

    Write \[ I_k=[\alpha_k,\beta_k]. \] The sequence \((\alpha_k)\) is increasing and bounded above, so by the least-upper-bound property it has a supremum \[ c=\sup_k\alpha_k. \] For every \(k\) and every \(j\geq k\), \[ \alpha_j\leq\beta_k, \] so \(c\leq\beta_k\). Also \(\alpha_k\leq c\). Hence \[ c\in I_k \] for every \(k\). If \(c'\) were another common point, then \[ \abs{c-c'}\leq\operatorname{length}(I_k) \longrightarrow0, \] so \(c'=c\).

    1.3. The original sequence has a subsequence converging to \(c\).

    Choose inductively indices \[ n_1<n_2<\cdots \] with \(x_{n_k}\in I_k\), possible because every \(I_k\) contains infinitely many terms. Since both \(x_{n_k}\) and \(c\) lie in \(I_k\), \[ \abs{x_{n_k}-c} \leq \operatorname{length}(I_k) \longrightarrow0. \] Thus \(x_{n_k}\to c\).

    1.4. Q.E.D.

    Step 3 proves step 1.

    2. Every continuous function\[ g:[a,b]\longrightarrow\RR \]is bounded.

    Suppose \(g\) were unbounded. Then for every positive integer \(n\) one could choose \(x_n\in[a,b]\) with \[ \abs{g(x_n)}>n. \] By step 1, some subsequence satisfies \[ x_{n_k}\longrightarrow c\in[a,b]. \] Continuity gives \[ g(x_{n_k})\longrightarrow g(c), \] so the subsequence \((g(x_{n_k}))\) is bounded, contradicting \[ \abs{g(x_{n_k})}>n_k\longrightarrow\infty. \]

    3. Every continuous function\[ g:[a,b]\longrightarrow\RR \]attains both its maximum and its minimum.

    3.1. The function \(g\) attains its maximum.

    By step 2, the set \[ g([a,b]) \] is nonempty and bounded above. Let \[ M=\sup g([a,b]). \] For every positive integer \(n\), choose \(x_n\in[a,b]\) such that \[ M-\frac1n<g(x_n)\leq M. \] By step 1, a subsequence satisfies \[ x_{n_k}\longrightarrow c\in[a,b]. \] Continuity gives \[ g(x_{n_k})\longrightarrow g(c), \] while the defining inequalities imply \[ g(x_{n_k})\longrightarrow M. \] Therefore \(g(c)=M\).

    3.2. The function \(g\) attains its minimum.

    Apply step 1 to the continuous function \(-g\).

    3.3. Q.E.D.

    Steps 1 and 2 prove step 3.

    4. If \(g\) is differentiable at an interior point \(c\) and has alocal maximum or local minimum at \(c\), then\[ g'(c)=0. \]

    Suppose first that \(c\) is a local maximum. For all sufficiently small \(h>0\), \[ \frac{g(c+h)-g(c)}{h}\leq0, \] whereas for all sufficiently small \(h<0\), \[ \frac{g(c+h)-g(c)}{h}\geq0. \] If the derivative exists, both one-sided limits equal \(g'(c)\), so \[ g'(c)\leq0 \qquad\text{and}\qquad g'(c)\geq0. \] Hence \(g'(c)=0\). For a local minimum, apply the same argument to \(-g\).

    5. Rolle’s theorem follows: if \(g\) is continuous on \([a,b]\),differentiable on \((a,b)\), and\[ g(a)=g(b), \]then there is \(c\in(a,b)\) such that\[ g'(c)=0. \]

    By step 3, \(g\) attains a maximum and a minimum on \([a,b]\). If these values are equal, then \(g\) is constant and every \(c\in(a,b)\) satisfies \(g'(c)=0\). Otherwise, because \(g(a)=g(b)\), at least one of the maximum or minimum is attained at an interior point \(c\in(a,b)\). Step 4 then gives \(g'(c)=0\).

    6. The mean value theorem follows: if \(g\) is continuous on \([a,b]\)and differentiable on \((a,b)\), then for some \(c\in(a,b)\),\[ g(b)-g(a)=g'(c)(b-a). \]

    Define \[ h(x) = g(x) - \frac{g(b)-g(a)}{b-a}(x-a). \] Then \(h\) is continuous on \([a,b]\), differentiable on \((a,b)\), and \[ h(a)=g(a)=h(b). \] By step 5, there is \(c\in(a,b)\) with \(h'(c)=0\). Thus \[ 0 = g'(c) - \frac{g(b)-g(a)}{b-a}, \] which rearranges to the stated identity.

    7. For the function in the problem,\[ \boxed{f(b)=f(a)}. \]

    Apply step 6 to \(f\). There is \(c\in(a,b)\) such that \[ f(b)-f(a) = f'(c)(b-a). \] The hypothesis gives \(f'(c)=0\), hence \[ f(b)-f(a)=0. \]

    8. Q.E.D.

    Step 7 is the required conclusion.