Classical Optimisation

Problem 1. Let f be a non-negative function satisfying the following conditions:

  • f is continuous on \mathbb R,
  • f is differentiable on \mathbb R \backslash \{0\},
  • f is strictly increasing on (-\infty, 0),
  • f is strictly decreasing on (0, \infty), and
  • f(|x|) \to 0 as x \to \infty.

Let A(x) denote the rectangle under the curve y=f(x) whose base that intersects the positive x-axis has length x. If A has a minimum prove that A is minimised when

\begin{aligned}  \left( 1 - \frac { f'(x) }{(g' \circ g^{-1} \circ f)(x)} \right) \cdot f(x) + (x - (g^{-1} \circ f)(x)) \cdot f'(x)  &= 0, \end{aligned}

where g = f|_{(-\infty, 0)}.

(Click for Solution)

Solution. Fix c > 0. If f(c) = 0, then f being strictly decreasing implies f(c+1) < 0, a contradiction. Therefore, f(0) > f(c) > 0. Since f(-x) \to 0 as x \to \infty, there exists b \in (-\infty, 0) such that f(b) < f(c) < f(0). Use the intermediate value theorem to obtain a \in (b, 0) such that f(a) = f(c). It is not hard to verify that f(x) > f(c) for x \in [a, c]. Therefore, the rectangle has an area A(c) of A(c) = (c - a) \cdot f(c).

In fact, more is true. Since g := f|_{(-\infty, 0)} is differentiable and strictly decreasing,

g(a) = f(a) = f(c)\quad \Rightarrow \quad a = g^{-1}(f(c)).

Hence, A(x) = (x - (g^{-1} \circ f)(x) ) \cdot f(x). Differentiating,

\begin{aligned} A'(x) &= (1 - (g^{-1})'(f(x)) \cdot f'(x)) \cdot f(x) + (x - (g^{-1} \circ f)(x) ) \cdot f'(x) \\ &= \left( 1 - \frac { f'(x) }{g'((g^{-1} \circ f)(x))} \right) \cdot f(x) + (x - (g^{-1} \circ f)(x)) \cdot f'(x).\end{aligned}

Setting A' yields the desired result.

Problem 2. Suppose further in Problem 1 that f is even. Prove that the x-value x_0 that maximises A satisfies the equation

f(x_0) + x_0 \cdot f'(x_0) = 0

and the inequality

2f'(x_0) + x_0 \cdot f''(x_0) < 0.

Furthermore, the maximum area is given by A(x_0) = 2x_0 \cdot f(x_0).

(Click for Solution)

Solution. If f is even, then for any x > 0,

\displaystyle (g^{-1} \circ f)(x) = g^{-1}(f(x)) = f^{-1}(f(-x)) = -x,

Furthermore, f' is odd:

\displaystyle f'(-x)\cdot(-1) = \frac{\mathrm d}{\mathrm dx} (f(-x)) = \frac{\mathrm d}{\mathrm dx}(f(x)) = f'(x),

so that f'(-x) = -f'(x). Therefore, for any x < 0, g(x) = f(x) = f(-x) and

g'(x) = f'(-x) \cdot (-1) = f'(x).

Hence, for any x > 0,

\begin{aligned} (g' \circ g^{-1} \circ f)(x) &= f' ((g^{-1} \circ f)(x)) \\ &= f'(-x) = -f'(x) \neq 0. \end{aligned}

Substituting into the result in Problem 1,

\begin{aligned}  \left( 1 - \frac { f'(x) }{-f'(x)} \right) \cdot f(x) + (x - (-x)) \cdot f'(x)  &= 0, \end{aligned}

yielding f(x) + x \cdot f'(x) = 0. For the inequality, apply the second derivative test.

Problem 3. Evaluate the maximum area of the rectangle defined in Problem 1 given that f(x) = 1/( 1+x^2 ).

(Click for Solution)

Solution. For x > 0, f(x) = 1/( 1+x^2 ) so that

\displaystyle f'(x) = -\frac 1{( 1+x^2 )^2} \cdot 2x = -\frac{2x}{( 1+x^2 )^2}.

Since f is even, by Problem 2,

\displaystyle \begin{aligned}  \frac 1{1+x_0^2 } + x_0 \cdot \left( -\frac{2x_0}{ (1+x_0^2)2 } \right) &= 0 \\ 1+x_0^2  + x_0 \cdot (-2x_0) &= 0\\ 1 - x_0^2 &= 0\\ x_0 &= 1, \end{aligned}

since x_0 = 1. To verify that this value gives the maximum, we evaluate the second derivative:

\displaystyle f''(x) = 2 \cdot \frac{x^2 -1 }{(1+x^2)^4}\quad \Rightarrow \quad f''(x_0) = 0.

Then f'(x_0) = -1 and f''(x_0) = 0, so that

2 \cdot f'(x_0) + f''(x_0) = -2 < 0,

as required. Therefore, f(x_0) =1/2 and the maximum area is given by

A(x_0) = 2 \cdot 1 \cdot f(1) = 1.

—Joel Kindiak. 9 Aug 25, 1510H

,

Published by


Leave a comment