Topic 15

Optimization

Every optimization question is the same three moves: write one function of one variable, find where its derivative is zero, and prove that point is the global max or min. The proof is where most marks are lost — every recipe below ends with it.

5 kinds of question. Open the one you are stuck on.

1

Find the global max and min on a closed interval

e.g. “Find the global (absolute) maximum and minimum of ff on [a,b][a, b]”
  1. 1
    Check ff is continuous on [a,b][a, b]

    Then the Extreme Value Theorem guarantees a global max and min, and they sit at critical points or endpoints. Say this: it is the justification.

  2. 2
    Find f′(x)f'(x)
  3. 3
    Find the critical points inside (a,b)(a, b)

    Where f′(x)=0f'(x) = 0 or f′(x)f'(x) is undefined. Throw away any outside the interval.

  4. 4
    Evaluate ff at every critical point and both endpoints

    Put the values in a small table. Use ff, not f′f'.

  5. 5
    Biggest value is the global max; smallest is the global min

    Give the value and every xx where it happens — ties are allowed. Why global: the EVT puts them in this table.

Remember: Critical points plus endpoints, then compare.

See it in a problem4 worked examples

Find the global maximum and minimum of f(x)=x3−6x2+9x+2f(x) = x^3 - 6x^2 + 9x + 2 on [−1,4][-1, 4].

  1. 1
    Check ff is continuous on [a,b][a, b]

    ff is a polynomial, so it is continuous on [−1,4][-1, 4]; the EVT says a global max and min exist.

  2. 2
    Find f′(x)f'(x)

    f′(x)=3x2−12x+9=3(x−1)(x−3)f'(x) = 3x^2 - 12x + 9 = 3(x - 1)(x - 3).

  3. 3
    Find the critical points inside (a,b)(a, b)

    f′(x)=0f'(x) = 0 at x=1x = 1 and x=3x = 3, both inside (−1,4)(-1, 4). f′f' is never undefined.

  4. 4
    Evaluate ff at every critical point and both endpoints

    x−1134f(x)−14626\begin{array}{c|cccc} x & -1 & 1 & 3 & 4 \\ \hline f(x) & -14 & 6 & 2 & 6 \end{array}

  5. 5
    Biggest value is the global max; smallest is the global min

    Largest value 66 (at x=1x = 1 and at x=4x = 4); smallest value −14-14 (at x=−1x = -1).

Answer:

Global max 66 at x=1x = 1 and x=4x = 4; global min −14-14 at x=−1x = -1. (Global: ff is continuous on [−1,4][-1, 4], so by the EVT the extremes are in the table.)

Common mistakes4
  • Forgetting the endpoints. On a closed interval the max or min is often at an end.
  • Missing a critical point where f′f' is undefined (a cusp from x2/3x^{2/3}, a corner from ∣x∣|x|).
  • Keeping a critical point outside [a,b][a, b] in the table.
  • Comparing f′f' values instead of ff values. The table is of ff.
2

Find the global max or min on an open or infinite interval

e.g. “Find the global maximum of ff on (0,∞)(0, \infty) or on [0,∞)[0, \infty)”
  1. 1
    Check the ends: values at closed ends, limits at open ends

    If ff climbs above both ends somewhere, a global max must exist: the ends cannot beat it, so EVT on a large closed piece finds it. Flip it for a min.

  2. 2
    Find f′(x)f'(x) and the critical points in the domain
  3. 3
    Make a sign chart of f′f'

    Pick a test value on each side of every critical point.

  4. 4
    Conclude with a justification sentence

    f′>0f' > 0 everywhere left of a point and f′<0f' < 0 everywhere right of it: rising then falling, so it is the global max. −- then ++: global min. Several critical points: compare their ff values with the ends.

Remember: Ends first, then the one critical point, then the sentence.

See it in a problem3 worked examples

Let f(x)=3xx2+9f(x) = \dfrac{3x}{x^2 + 9} on [0,∞)[0, \infty). (a) Explain without differentiation why ff has a global maximum. (b) Find it.

  1. 1
    Check the ends: values at closed ends, limits at open ends

    (a) f(0)=0f(0) = 0, f(x)>0f(x) > 0 for x>0x > 0, and lim⁡x→∞f(x)=0\displaystyle\lim_{x\to\infty} f(x) = 0. Since f(1)=0.3>0f(1) = 0.3 > 0, past some RR we have f(x)<0.3f(x) < 0.3; on [0,R][0, R] the EVT gives a max ≥0.3\ge 0.3, and nothing past RR beats it. So a global max exists.

  2. 2
    Find f′(x)f'(x) and the critical points in the domain

    (b) f′(x)=3(x2+9)−3x(2x)(x2+9)2=3(9−x2)(x2+9)2f'(x) = \dfrac{3(x^2 + 9) - 3x(2x)}{(x^2 + 9)^2} = \dfrac{3(9 - x^2)}{(x^2 + 9)^2}, zero at x=3x = 3 (and x=−3x = -3, outside the domain).

  3. 3
    Make a sign chart of f′f'

    f′>0f' > 0 on (0,3)(0, 3) and f′<0f' < 0 on (3,∞)(3, \infty).

  4. 4
    Conclude with a justification sentence

    ff rises up to x=3x = 3 and falls after it, so f(3)=918=12f(3) = \dfrac{9}{18} = \dfrac{1}{2} is the global max.

Answer:

Global maximum 12\tfrac{1}{2} at x=3x = 3 (global: ff increases on (0,3)(0, 3) and decreases on (3,∞)(3, \infty)).

Common mistakes3
  • Saying "local max, so global max" with no reason. You need the sign chart on the whole domain, or one critical point and f′′f'' of one sign everywhere.
  • Using the EVT on [0,∞)[0, \infty) directly. It needs a closed bounded interval; argue with the limit at ∞\infty instead.
  • Forgetting that a global max may not exist: if f→∞f \to \infty at an end, say so.
3

Solve a geometry optimization word problem

e.g. “Cut squares from the corners of a sheet and fold up the sides … largest volume”
  1. 1
    Draw it, name the variables, write the quantity to optimize

    Volume, surface area, area: whatever the question says to make biggest or smallest.

  2. 2
    Use the constraint to get one variable

    The fixed number (volume, sheet size, perimeter) links the variables. Solve it for one and substitute.

  3. 3
    State the domain of that variable

    Lengths are positive; a cut cannot exceed half the shorter side.

  4. 4
    Differentiate and solve f′(x)=0f'(x) = 0 in the domain
  5. 5
    Justify it is the global max or min

    Closed domain: compare with endpoints. Open domain: sign chart of f′f', or f′′f'' of one sign on the whole domain.

  6. 6
    Answer the question asked, with units

Remember: Picture, objective, constraint, domain, derivative, proof, answer.

See it in a problem3 worked examples

A sheet of metal is 2424 cm by 99 cm. Equal squares of side xx are cut from the four corners and the sides are folded up to make an open box. Find the largest possible volume.

  1. 1
    Draw it, name the variables, write the quantity to optimize

    Base (24−2x)(24 - 2x) by (9−2x)(9 - 2x), height xx. Maximize V=x(24−2x)(9−2x)V = x(24 - 2x)(9 - 2x).

  2. 2
    Use the constraint to get one variable

    The sheet size is the constraint, already used: V(x)=4x3−66x2+216xV(x) = 4x^3 - 66x^2 + 216x.

  3. 3
    State the domain of that variable

    0≤x≤4.50 \le x \le 4.5 (the cut cannot pass half of 99).

  4. 4
    Differentiate and solve f′(x)=0f'(x) = 0 in the domain

    V′(x)=12x2−132x+216=12(x−2)(x−9)V'(x) = 12x^2 - 132x + 216 = 12(x - 2)(x - 9); only x=2x = 2 is in the domain.

  5. 5
    Justify it is the global max or min

    Closed interval: V(0)=0V(0) = 0, V(4.5)=0V(4.5) = 0, V(2)=2⋅20⋅5=200V(2) = 2 \cdot 20 \cdot 5 = 200. So x=2x = 2 gives the global max.

  6. 6
    Answer the question asked, with units

    Cut 22 cm squares; the box is 20×5×220 \times 5 \times 2 cm.

Answer:

Maximum volume 200 cm3200\ \text{cm}^3 (cut x=2x = 2 cm). Global max: on the closed domain [0,4.5][0, 4.5] it beats both endpoints, where V=0V = 0.

Common mistakes4
  • Differentiating a function of two variables. Substitute the constraint first.
  • Skipping the justification. "f′(x)=0f'(x) = 0 so it is a minimum" earns nothing on its own.
  • Answering with xx when the question asks for all the dimensions, or for the maximum volume.
  • Open-top box: only one base in the surface area. Closed: two.
4

Minimize a cost or spend a fixed budget

e.g. “The top and bottom cost $… per m², the sides cost $… — minimize the cost”
  1. 1
    Write cost as price times amount, added over the parts

    Budget question: the thing to maximize is the objective, and the budget is the constraint.

  2. 2
    Write the fixed quantity as an equation

    A volume, a budget, a distance: whatever cannot change.

  3. 3
    Substitute to get one variable, and state its domain
  4. 4
    Differentiate and solve C′(x)=0C'(x) = 0

    Budget question: differentiate the thing you maximize instead, such as A′(x)=0A'(x) = 0.

  5. 5
    Justify the max or min, then answer in context

    Closed domain: compare with endpoints. Open domain: sign chart, or second derivative of one sign everywhere.

Remember: Price times amount, fix the constraint, one variable.

See it in a problem3 worked examples

A closed cylindrical water tank must hold 40π m340\pi\ \text{m}^3. Material for the top and bottom costs $5 per m2\text{m}^2; the curved side costs $2 per m2\text{m}^2. Find the dimensions that minimize the cost.

  1. 1
    Write cost as price times amount, added over the parts

    C=5(2πr2)+2(2πrh)=10πr2+4πrhC = 5(2\pi r^2) + 2(2\pi r h) = 10\pi r^2 + 4\pi r h.

  2. 2
    Write the fixed quantity as an equation

    πr2h=40π⇒h=40r2\pi r^2 h = 40\pi \Rightarrow h = \dfrac{40}{r^2}.

  3. 3
    Substitute to get one variable, and state its domain

    C(r)=10πr2+160πrC(r) = 10\pi r^2 + \dfrac{160\pi}{r}, with r>0r > 0.

  4. 4
    Differentiate and solve C′(x)=0C'(x) = 0

    C′(r)=20πr−160πr2=0⇒r3=8⇒r=2C'(r) = 20\pi r - \dfrac{160\pi}{r^2} = 0 \Rightarrow r^3 = 8 \Rightarrow r = 2.

  5. 5
    Justify the max or min, then answer in context

    C′′=20π+320πr3>0C'' = 20\pi + \dfrac{320\pi}{r^3} > 0 for all r>0r > 0, so r=2r = 2 is the global min. Then h=404=10h = \dfrac{40}{4} = 10 and C=40π+80π=120πC = 40\pi + 80\pi = 120\pi.

Answer:

Radius 22 m, height 1010 m (cost 120π120\pi dollars, about $377). Global min: C′′>0C'' > 0 for every r>0r > 0.

Common mistakes3
  • Using one price for every part. Each surface or stretch gets its own price.
  • On a budget, the domain is closed (spend between none and all on each part): check the endpoints.
  • Reporting the minimum cost when the question asks for the dimensions, or the reverse.
5

Find the shortest distance from a point to a curve

e.g. “Find the point on the curve closest to (x0,y0)(x_0, y_0)”
  1. 1
    Take a general point (x, f(x))(x,\ f(x)) on the curve

    If the curve is easier as x=g(y)x = g(y) (like y2=2xy^2 = 2x), use (g(y), y)(g(y),\ y) and work in yy.

  2. 2
    Write D2=(x−x0)2+(f(x)−y0)2D^2 = (x - x_0)^2 + (f(x) - y_0)^2

    Minimize D2D^2, not DD: the square root is increasing, so both have the same minimizer, and D2D^2 has no root to differentiate.

  3. 3
    Differentiate D2D^2 and solve (D2)′=0(D^2)' = 0
  4. 4
    Justify it is the global min

    Sign chart of (D2)′(D^2)', or D2→∞D^2 \to \infty at the ends with one critical point.

  5. 5
    Answer: the point, and D=D2D = \sqrt{D^2}

Remember: Square it, differentiate, take the root at the end.

See it in a problem3 worked examples

Find the point on y=xy = \sqrt{x} closest to (4,0)(4, 0), and the shortest distance.

  1. 1
    Take a general point (x, f(x))(x,\ f(x)) on the curve

    A point on the curve: (x,x)(x, \sqrt{x}), x≥0x \ge 0.

  2. 2
    Write D2=(x−x0)2+(f(x)−y0)2D^2 = (x - x_0)^2 + (f(x) - y_0)^2

    D2=(x−4)2+(x−0)2=(x−4)2+xD^2 = (x - 4)^2 + (\sqrt{x} - 0)^2 = (x - 4)^2 + x.

  3. 3
    Differentiate D2D^2 and solve (D2)′=0(D^2)' = 0

    (D2)′=2(x−4)+1=0⇒x=72(D^2)' = 2(x - 4) + 1 = 0 \Rightarrow x = \tfrac{7}{2}.

  4. 4
    Justify it is the global min

    (D2)′′=2>0(D^2)'' = 2 > 0 on the whole domain, so x=72x = \tfrac{7}{2} is the global min (check: D2(0)=16D^2(0) = 16 is bigger).

  5. 5
    Answer: the point, and D=D2D = \sqrt{D^2}

    D2=14+72=154D^2 = \tfrac{1}{4} + \tfrac{7}{2} = \tfrac{15}{4}, so D=152D = \dfrac{\sqrt{15}}{2} at (72,72)\left(\tfrac{7}{2}, \sqrt{\tfrac{7}{2}}\right).

Answer:

Closest point (72,72)\left(\tfrac{7}{2}, \sqrt{\tfrac{7}{2}}\right); distance 152\dfrac{\sqrt{15}}{2}. Global min: (D2)′′=2>0(D^2)'' = 2 > 0 on the whole domain.

Common mistakes3
  • Forgetting the square root at the end: the answer is DD, not D2D^2.
  • Ignoring the curve's domain (x\sqrt{x} needs x≥0x \ge 0) when checking the ends.
  • Giving only the point when the question asks for the distance, or the reverse.

Need the formulas? Must memorize · Flashcards

Found this helpful? Leave us a review on Google ↗Spotted a mistake?

Which topics are you reviewing?

The topic list, Flashcards, Must memorize and Exam words will only show these. Change it any time with the Topics button above the topic list.

Limits & continuity
Derivatives
Applications
Theorems & approximation
10 of 17 topics selected