Calculus › Section D · Applications of derivatives › Topic D.7
Optimisation problems
This is the payoff of the whole of Section D: turning a messy word problem into a function and letting the calculus find its best value. Most of the work is the translation, so a fixed routine does the heavy lifting.
The payoff
What size box holds the most from a given sheet? Which production level earns the most? Which point on a curve is closest to a given point? These are optimisation problems, and they are a main reason the machinery of Sections C and D exists. The calculus part is by now routine: differentiate, find critical numbers, test. The hard part comes earlier: building a one-variable function out of a story. That translation is the real content of this topic.
The seven-step routine
One: read twice and draw a picture. Two: assign variables to everything that varies, and name the objective, the quantity to maximise or minimise. Three: write the objective as a formula. Four: use the constraint, the problem's fixed condition, to eliminate variables until the objective depends on one. Five: write the domain reality forces, lengths positive, cuts smaller than the sheet. Six: differentiate, find critical numbers, compare candidates and endpoints. Seven: answer the question asked, with units. The two worst errors, differentiating before reducing and skipping the domain, happen at steps four and five.
Worked example: the riverside pen
A farmer has 400 m of fencing and wants a rectangular pen against a straight river, which replaces one long side. Let the sides perpendicular to the river be \(x\) and the parallel side \(y\). Objective: area \(A = xy\). Constraint: \(2x + y = 400\), so \(y = 400 - 2x\). Then \(A(x) = 400x - 2x^2\) on \(0 \le x \le 200\). Then \(A'(x) = 400 - 4x = 0\) at \(x = 100\). Endpoints give zero area, so the maximum is \(x = 100\), \(y = 200\), enclosing \(20\,000\) m²: the side along the river is twice each end. Compare a pen fenced on all four sides: \(2x + 2y = 400\) gives \(A = x(200 - x)\), maximised by the square \(x = y = 100\), with only \(10\,000\) m². The river doubles the area.
Is it really the best?
Finding a critical number is not yet finding the best answer; judgement closes the gap. On a closed domain, compare against the endpoints, exactly as D.1 taught. On an open or unbounded interval, use the single-critical-number argument: if a continuous objective has only one critical number there, and the first or second derivative test shows it is a local maximum, then it is the absolute maximum (likewise for a minimum). The riverside pen used the endpoint check; the box coming next uses the same. Always write one sentence of justification: without it you have found a level point, not yet the best one.
Worked example: the open box
From a square sheet of side 30 cm, cut equal squares of side \(x\) from the corners and fold up an open box. The base then measures \(30 - 2x\) by \(30 - 2x\) and the height is \(x\), so \(V(x) = x(30 - 2x)^2\) for \(0 \le x \le 15\). Differentiate by the product and chain rules: \(V'(x) = (30 - 2x)^2 - 4x(30 - 2x) = (30 - 2x)(30 - 6x)\). The zeros are \(x = 15\), which folds the sheet away entirely and gives zero volume, and \(x = 5\). Comparing candidates, the volume at \(x = 5\) is \(5 \cdot 20^2 = 2000\) cm³, the maximum: cut 5 cm squares.
Worked example: the cheapest can
A cylindrical can must hold a fixed volume \(V\). Which shape uses least material? The surface area is \(S = 2\pi r^2 + 2\pi r h\), and the constraint \(\pi r^2 h = V\) gives \(h = \frac{V}{\pi r^2}\), so
Then \(S'(r) = 4\pi r - \frac{2V}{r^2} = 0\) means \(4\pi r^3 = 2V\), and substituting back into the constraint shows \(h = 2r\): the cheapest can is exactly as tall as it is wide. There are no endpoints on \(r > 0\), and \(S \to \infty\) as \(r \to 0^{+}\) or \(r \to \infty\), so this unique critical point is the absolute minimum.
Worked example: the closest point
Which point of the parabola \(y = x^2\) is closest to \((0, 3)\)? Minimising distance squared avoids the square root: \(D(x) = x^2 + (x^2 - 3)^2\). Then \(D'(x) = 2x + 4x(x^2 - 3) = 2x(2x^2 - 5)\), zero at \(x = 0\) and \(x = \pm\sqrt{5/2}\). Three candidates: compare. \(D(0) = 9\), while \(D\left(\pm\sqrt{5/2}\right) = \frac{5}{2} + \left(\frac{5}{2} - 3\right)^2 = \frac{5}{2} + \frac{1}{4} = \frac{11}{4}\), much smaller. Since \(D\) grows beyond all bounds as \(|x| \to \infty\), the winners are global: the closest points are \(\left(\pm\sqrt{\tfrac{5}{2}},\ \tfrac{5}{2}\right)\). The vertex, straight below \((0, 3)\), is not the nearest point.
Worked example: maximum profit
Profit is revenue minus cost, \(P = R - C\), so \(P'(x) = 0\) means \(R'(x) = C'(x)\): at maximum profit, marginal revenue equals marginal cost. See it concretely. Demand sets the price \(p = 100 - x\) for \(x\) units, so revenue is \(R(x) = 100x - x^2\); let cost be \(C(x) = 20x + 500\), a line with a fixed overhead, assuming constant per-unit cost. Then \(P(x) = 80x - x^2 - 500\), and \(P'(x) = 80 - 2x = 0\) at \(x = 40\). There \(R'(40) = 100 - 80 = 20 = C'(40)\), exactly as the rule says. Since \(P''(x) = -2 < 0\), this is the maximum, and the profit is \(80(40) - 1600 - 500 = 1\,100\) dollars.
A one-variable classic
Some optimisation results are worth carrying in memory. Which positive number plus its reciprocal is smallest? Let \(f(x) = x + \frac{1}{x}\) for \(x > 0\); then \(f'(x) = 1 - \frac{1}{x^2} = 0\) at \(x = 1\), and \(f''(1) = 2 > 0\) confirms a minimum, with value \(1 + 1 = 2\). Since it is the only critical number on \(x > 0\), it is the absolute minimum: \(x + \frac{1}{x} \ge 2\) for all positive \(x\), with equality only at \(1\). Similarly, of two positive numbers with sum 20, the largest product comes from \(P = x(20 - x)\): \(P' = 20 - 2x = 0\) at \(x = 10\), so the numbers are 10 and 10, with product 100.
Model honestly
Every answer above came from an idealisation, and saying so is part of the solution. The pen assumed a straight river and gate-free fencing; the can assumed the same material for top, bottom and side, with no seams or waste; the profit model assumed a linear demand curve and constant unit cost. Real fences have gates, cans have seams, demand curves bend. The standard method is what we did: solve the clean model exactly, state the assumptions out loud, then adjust for reality.
Common mistakes to avoid
The four killers: differentiating the objective before eliminating to one variable, which produces a meaningless derivative of a two-variable formula; skipping the domain, so answers like negative side lengths sneak through; reporting the critical number when the question asked for the maximum value, or the reverse, so read the final sentence again; and stopping at a critical number without comparing endpoints or arguing the shape. Lesser sins: dropping units; minimising the distance itself, which works but gives a much messier derivative than its square.
Key takeaways
Optimisation is a routine: translate the story into a one-variable objective using the constraint, respect the physical domain, then let the derivative find candidates and compare. The classics: the riverside pen's long side is twice each end, the 30 cm sheet gives up 5 cm corners, the cheapest can is as tall as it is wide, and maximum profit lies where marginal revenue equals marginal cost. State your assumptions, justify why the candidate is genuinely best, and answer the question with units.
Flashcards
Flashcard
The quantity to maximise or minimise
Test your understanding
Question 1 of 5
In the routine, what is the constraint used for?
Question 2 of 5
With 400 m of fencing and a river forming one long side, what is the largest possible rectangular pen area?
Question 3 of 5
From a 30 cm square sheet, which corner-cut size maximises the open box's volume?
Question 4 of 5
Profit is maximised at the production level where…
Question 5 of 5
For a fixed volume, the least-material can has which shape?
Standards: 5
