This is the corrected exercise set for the chain rule in MATH 140, Calculus 1, at McGill University, section 3.4 of Stewart. The chain rule is the rule that every later chapter of the course leans on: implicit differentiation, related rates, the derivative of a logarithm and every optimization problem are chain rules in disguise. Every number here is exact and chosen to be done by hand, and each solution names the inner function before differentiating, because that sentence is where the method marks are.
The thread running through the whole set: name the inner function u first, differentiate the outer function AT u, leaving u untouched inside it, then multiply by u′, one factor per layer. The same two errors are hunted from the first exercise to the last: the missing inner factor, and the outer derivative evaluated at x instead of at g(x).
The traps named in the solutions: cos(x2) as the derivative of sin(x2), cos(2x) placed inside the cosine, f′(1)g′(1) read from the wrong row of a table, the power rule applied to 2x, ex2 as its own derivative, a chain rule used at a corner where it does not apply, the second derivative written f′′(g)g′′, an extraneous root created by squaring, a cost derivative evaluated at a number of weeks, and a rate multiplied by 60 instead of divided.
Self-checking setType your answers under each question: the page tells you right or wrong before you open the solution. With an account, every answer right on the first try earns points.
•Change of variable: dxdf(kx+c)=kf′(kx+c); the tangent at a is y=F(a)+F′(a)(x−a).
Part A: the basics (/50)
Exercise 1: Name the inner function first: one layer
The chain rule: if g is differentiable at x and f is differentiable at g(x), then h=f∘g is differentiable at x and h′(x)=f′(g(x))g′(x). In Leibniz notation, with u=g(x) and y=f(u): dxdy=dudy⋅dxdu.
The write-up a marker expects starts with a sentence that NAMES the inner function: let u=…, so y=f(u). Then the outer function is differentiated at u, with u left untouched inside it, and the result is multiplied by u′. In every part below, write that sentence before any derivative.
a) Differentiate y=(3x2−5)7 and give y′(1).
b) Differentiate y=x3+1.
c) The three functions y1=sin(x2), y2=sin2x and y3=x2sinx use the same two pieces. Say which operation is done LAST in each, then differentiate all three.
d) Differentiate y=e−x2/2 and y=e1/x.
e) A student writes dxd(x2+4)−3=−3(x2+4)−4. Correct the answer, then explain why the quotient rule on (x2+4)31 gives the same result with more work.
e)dxd(x2+4)−3=−(x2+4)46x: the factor 2x was missing.
a) Let u=3x2−5, so y=u7. The outer function is the seventh power, whose derivative is 7u6; the inner derivative is u′=6x. By the chain rule, y′=7(3x2−5)6⋅6x=42x(3x2−5)6. At x=1: u=−2 and (−2)6=64, so y′(1)=42⋅64=2688. Expanding (3x2−5)7 first would take a page and invite ten sign errors: the chain rule exists precisely so that the bracket is never opened. The inside stays as it is in the outer derivative, (3x2−5)6, and never becomes (6x)6.
b) Let u=x3+1, so y=u=u1/2 and dudy=2u1. With u′=3x2: y′=2x3+11⋅3x2=2x3+13x2. The rule (u)′=2uu′ is worth knowing by heart, with its u′ on top: the answer 2x3+11, without the 3x2, is the single most common error of the chapter.
c) Read the order of operations as a calculator would perform it. In sin(x2) you square first and take the sine LAST: the outer function is sin, the inner one u=x2, so y1′=cos(x2)⋅2x. In sin2x=(sinx)2 you take the sine first and square LAST: the outer function is the square, u=sinx, so y2′=2sinx⋅cosx. In x2sinx the last operation is a PRODUCT, not a composition, so the product rule applies: y3′=2xsinx+x2cosx. Three expressions built from the same pieces, three different rules: the last operation decides, and a student who reads sin2x as sin(x2) loses the whole question.
d) For y=e−x2/2, let u=−2x2, so y=eu and dudy=eu; with u′=−x, y′=−xe−x2/2. For y=e1/x, let u=x1=x−1, so u′=−x21 and y′=−x2e1/x. The exponential is its own derivative only when its exponent is x itself: with any other exponent, eu comes back unchanged AND multiplied by u′.
e) Let u=x2+4, so y=u−3 and y′=−3u−4⋅u′=−3(x2+4)−4⋅2x=−(x2+4)46x. The student differentiated the outer layer correctly and stopped: the factor 2x is missing, and the answer is wrong for every x=21. By the quotient rule, (v1)′=−v2v′ with v=(x2+4)3, and v′ itself needs the chain rule: v′=3(x2+4)2⋅2x. So −(x2+4)66x(x2+4)2=−(x2+4)46x, the same answer after a simplification. Rewriting a constant over a power as a negative power is the faster road, and it removes one place to make a mistake.
Tick the exercises you have done or want to review: a free account, no password, keeps your ticks from one visit to the next and tells you which chapter to tackle next. Create your space, an email is enough.
Exercise 2: Three layers and more: one factor per layer
When a function is built in three layers, y=f(g(k(x))), the chain rule is applied twice and produces THREE factors: y′=f′(g(k(x)))⋅g′(k(x))⋅k′(x). In Leibniz notation, with v=k(x) and u=g(v): dxdy=dudy⋅dvdu⋅dxdv.
The safe method peels the function from the outside in, like an onion: write the layers down, outermost first, then write one factor per layer, each outer derivative keeping everything inside it untouched. Count your factors at the end: as many factors as layers.
a) Differentiate y=sin3(4x), listing its three layers first.
b) Differentiate y=1+cos2x.
c) Differentiate y=esin(x2).
d) Let h(x)=(1+sin6πx)5. Compute h′(1) exactly.
e) A student writes dxdsin(cos(x2))=cos(−sin(2x)). Find every error, and give the correct derivative.
Show the solution
Answers
a)y′=12sin2(4x)cos(4x)
b)y′=−1+cos2xsinxcosx
c)y′=2xcos(x2)esin(x2)
d)h′(1)=641353π
e)dxdsin(cos(x2))=−2xsin(x2)cos(cos(x2))
a) sin3(4x)=(sin(4x))3. Layers, from the outside in: the cube, the sine, the multiplication by 4. One factor per layer: 3(sin4x)2 for the cube, cos(4x) for the sine, 4 for the inner layer. So y′=3sin2(4x)⋅cos(4x)⋅4=12sin2(4x)cos(4x). Three layers, three factors. The usual loss is the last one: 3sin2(4x)cos(4x) looks finished and is off by a factor 4 everywhere.
b) Layers: the square root, then 1+w2, then w=cosx. Factors: 21+cos2x1, then 2cosx, then −sinx. So y′=21+cos2x1⋅2cosx⋅(−sinx)=−1+cos2xsinxcosx. The derivative of the constant 1 is 0, so the middle layer contributes only 2cosx; the minus sign comes from the innermost layer and must survive to the end.
c) Layers: e(⋅), then sin, then x2. Factors: esin(x2) (the exponential comes back unchanged, with its whole exponent), cos(x2), 2x. So y′=esin(x2)⋅cos(x2)⋅2x. Check the middle factor: the sine was applied to x2, so its derivative is evaluated at x2 too, cos(x2) and not cosx. Each outer derivative is evaluated at exactly what that layer received.
d) Layers: the fifth power, then 1+sinw, then w=6πx. So h′(x)=5(1+sin6πx)4⋅cos6πx⋅6π. At x=1: sin6π=21 and cos6π=23, so h′(1)=5⋅(23)4⋅23⋅6π=5⋅1681⋅123π=1924053π=641353π. Differentiate in general FIRST, then substitute x=1: substituting first turns the function into the constant (23)5, whose derivative is 0.
e) Three errors. First, the derivative of the outer sine is cos evaluated at what the sine received, cos(x2), untouched: the student differentiated INSIDE the cosine instead of multiplying by the inner derivative. Second, the inner layers were differentiated as if cos(x2) had derivative −sin(2x): the cosine layer gives −sin(x2), and the square gives a separate factor 2x, never sin(2x). Third, a product of factors became a single nested expression. Correctly, with layers sin, cos, x2: y′=cos(cos(x2))⋅(−sin(x2))⋅2x=−2xsin(x2)cos(cos(x2)). The chain rule MULTIPLIES factors side by side; it never places one derivative inside another.
Exercise 3: The chain rule from a table and from a graph
On a McGill exam, the chain rule is often tested with functions you cannot write down: only their values and slopes are given, in a table or on a graph. The rule is then pure bookkeeping, (f∘g)′(a)=f′(g(a))g′(a): first read g(a), THEN go to the row or the point g(a) to read f′ there.
The table gives values of f, f′, g and g′. The figure shows two other functions, also called f (blue) and g (orange) in parts d) and e), made of straight pieces through the marked points.
x
f(x)
f′(x)
g(x)
g′(x)
1
3
2
2
4
2
1
5
3
−1
3
4
−2
1
6
a) Table. Let h(x)=f(g(x)). Find h′(1). A student answers 8: what did the student compute?
b) Table. Let k(x)=g(f(x)) and G(x)=[g(x)]3. Find k′(1) and G′(2).
c) Table. Let F(x)=f(x2−1) and H(x)=f(f(x)). Find F′(2) and H′(1).
d) Graph. Let u(x)=f(g(x)). Find u′(1) and u′(5).
e) Graph. Let v(x)=g(f(x)) and w(x)=g(g(x)). Does v′(1) exist? Find w′(4).
Show the solution
Answers
a)h′(1)=f′(2)g′(1)=20; 8=f′(1)g′(1) is the wrong point.
b)k′(1)=g′(3)f′(1)=12; G′(2)=3[g(2)]2g′(2)=−27
c)F′(2)=f′(3)⋅4=−8; H′(1)=f′(3)f′(1)=−4
d)u′(1)=f′(3)g′(1)=21; u′(5)=f′(3)g′(5)=−21
e)v′(1) does not exist (f(1)=3 is a corner of g; one-sided slopes −2 and 2); w′(4)=−1
a) h′(1)=f′(g(1))⋅g′(1). Read g(1)=2 first, then f′ at 2: f′(2)=5. With g′(1)=4, h′(1)=5⋅4=20. The student's 8 is f′(1)⋅g′(1)=2⋅4: the outer derivative was read at x=1 instead of at g(1)=2. The outer function never sees x, only what g hands it, so its slope is read at g(1). That is the whole point of the table question, and a wrong row gives 0 out of the method marks.
b) k′(1)=g′(f(1))⋅f′(1). Read f(1)=3, then g′(3)=6; with f′(1)=2, k′(1)=12. For G=g3, the outer function is the cube and the inner one g: G′(x)=3[g(x)]2g′(x), so G′(2)=3⋅32⋅(−1)=−27. The power of a function is a composition, and 3[g(2)]2 alone, without g′(2), is the missing-factor error again.
c) F′(x)=f′(x2−1)⋅2x. At x=2: x2−1=3, f′(3)=−2, and the inner derivative is 2⋅2=4, so F′(2)=−8. The trap is to read f′(2) because the question says x=2: the table is read at the inside, 3. For H=f∘f: H′(1)=f′(f(1))⋅f′(1)=f′(3)⋅2=−4. A function composed with itself is still two layers, and its derivative is still two factors read at two different points.
d) On the graph, f has slope 2−05−1=2 on (0,2) and 6−23−5=−21 on (2,6); g has slope −1 on (0,3) and 1 on (3,6). Now u′(1)=f′(g(1))g′(1): g(1)=3, which lies in (2,6) where f′=−21, and g′(1)=−1, so u′(1)=21. Similarly g(5)=3 and g′(5)=1, so u′(5)=−21. The same inner VALUE, 3, gives the same outer slope; the inner SLOPE changes sign, and so does u′.
e) v′(1)=g′(f(1))f′(1) requires g′(3), and g has a corner at 3: the chain rule cannot be applied, since its hypothesis, g differentiable at f(1), fails. Look at the one-sided behaviour instead. Just left of x=1, f(x) is slightly below 3 (because f increases with slope 2), where g has slope −1: v has slope −1⋅2=−2. Just right of 1, f(x)>3 and g has slope 1: v has slope 2. The one-sided slopes differ, so v′(1) does not exist; v has a corner at 1. For w: g(4)=2, g′(2)=−1 and g′(4)=1, so w′(4)=−1. Checking the hypotheses of the chain rule is part of the answer.
Exercise 4: Exponentials in any base: b to the x is e to the x ln b
For b>0, b=elnb, so bx=exlnb. The number lnb is a CONSTANT: nothing about logarithms is differentiated in this exercise. The chain rule, with inner function u=xlnb, is all that is needed to differentiate an exponential whose base is not e.
The figure shows y=2x, y=ex and y=3x, and the dashed line y=1+x through their common point (0,1).
a) Prove that dxdbx=bxlnb, naming the inner function.
b) Differentiate 3x2, 5sinx and 10−x.
c) The line y=1+x is tangent to exactly one of the three curves at (0,1). Which one? Give the slopes of the other two at 0, and explain what the figure shows near (0,1).
d) Find the point where the tangent to y=4x has slope 8ln2. Then differentiate 8x in two ways, as 8x and as 23x, and reconcile the answers.
e) Differentiate xπ, πx and ππ. A student writes (2x)′=x2x−1: show with the figure that this is impossible at x=0.
Show the solution
Answers
a)bx=eu with u=xlnb, u′=lnb: (bx)′=exlnblnb=bxlnb
b)2xln3⋅3x2; ln5cosx⋅5sinx; −ln10⋅10−x
c)y=ex (slope 1); slopes ln2<1 for 2x and ln3>1 for 3x: the line crosses those two.
d)x=1, point (1,4); 8xln8=23x⋅3ln2 since ln8=3ln2
e)πxπ−1; πxlnπ; 0. The false formula gives slope 0 at x=0 for an increasing curve.
a) Since b=elnb, bx=(elnb)x=exlnb. Let u=xlnb, so bx=eu. The outer function is eu, its own derivative; the inner function is a constant times x, so u′=lnb. By the chain rule, dxdbx=exlnb⋅lnb=bxlnb. For b=e, lne=1 and the familiar (ex)′=ex comes back: e is the only base with no extra factor. Writing lnb as a constant, and saying so, is what the marker checks here.
b) 3x2: the outer function is 3u, of derivative 3uln3, and u=x2: dxd3x2=3x2ln3⋅2x. 5sinx: 5sinxln5⋅cosx. 10−x: 10−xln10⋅(−1)=−ln10⋅10−x. Each answer has the exponential unchanged, the constant lnb, and the derivative of the exponent: three factors, and the most often forgotten is the last.
c) The slope at 0 of y=bx is b0lnb=lnb. For ex it is lne=1, exactly the slope of y=1+x, which also passes through (0,1): the line is the tangent to y=ex. For 2x the slope is ln2≈0.69<1, and for 3x it is ln3≈1.10>1. On the figure, the dashed line therefore CROSSES y=2x and y=3x at (0,1), passing from one side to the other, while it stays below y=ex on both sides. This is Stewart's definition of e read on a graph: the base whose exponential has slope exactly 1 at 0.
d) dxd4x=4xln4=4x⋅2ln2. Setting 4x⋅2ln2=8ln2 gives 4x=4, so x=1 and the point is (1,4). For 8x: directly, (8x)′=8xln8. As 23x, the outer function is 2u with u=3x: (23x)′=23xln2⋅3. The two agree because 23x=8x and 3ln2=ln23=ln8, a law of logarithms, not a derivative. Two correct routes must meet, and checking that they do is the cheapest verification there is.
e) xπ is a power of x with a constant exponent: power rule, πxπ−1. πx is an exponential with a constant base: πxlnπ. ππ is a constant: its derivative is 0. The student applied the power rule to an exponential. At x=0 the false formula gives 0⋅2−1=0, a horizontal tangent, while the figure shows y=2x strictly increasing through (0,1) with slope ln2. Before differentiating, ask where the variable is: in the base (power rule) or in the exponent (exponential rule). A variable in both, like xx, needs another tool, met later in the course.
Exercise 5: Powers and roots of a quotient: rewrite, then one chain
A power or a root of a whole expression is a composition: [q(x)]n has outer function un and inner function u=q(x), whatever q is. When q is a quotient, the inner derivative needs the quotient rule, and the order is always: outer layer first, then the inner derivative as a separate factor, then simplify.
Rewriting before differentiating saves work: w=w1/2, wk1=w−k. And the final answer is simplified into a FACTORED form, because that is the form from which the zeros of y′, the horizontal tangents, can be read.
a) Differentiate y=(x−32x+1)5 and simplify.
b) Differentiate y=x+1x−1 for x>1, simplify, and give y′(3).
c) Differentiate y=9−x21 and give y′(5).
d) Differentiate y=(x2+1)3(2x−5)4 and write y′ as a product of factors.
e) Using d), find every x at which the curve y=(x2+1)3(2x−5)4 has a horizontal tangent.
Show the solution
Answers
a)y′=−(x−3)635(2x+1)4
b)y′=(x+1)x2−11, y′(3)=162
c)y′=(9−x2)3/2x, y′(5)=85
d)y′=2(x2+1)2(2x−5)3(10x2−15x+4)
e)x=25 and x=2015±65
a) Let u=x−32x+1, so y=u5 and y′=5u4u′. By the quotient rule, u′=(x−3)22(x−3)−(2x+1)⋅1=(x−3)2−7. So y′=5(x−3)4(2x+1)4⋅(x−3)2−7=−(x−3)635(2x+1)4. The exponents add in the denominator, 4+2=6. The frequent error is to apply the quotient rule to the whole fifth power with numerator (2x+1)5 and denominator (x−3)5: it is legitimate, but it needs the chain rule twice more and ends in a fraction that nobody simplifies correctly under exam time.
b) Let u=x+1x−1, so y=u1/2 and y′=2uu′. Quotient rule: u′=(x+1)2(x+1)−(x−1)=(x+1)22. So y′=(x+1)22⋅21x−1x+1=(x+1)21⋅x−1x+1=(x+1)3/2(x−1)1/21=(x+1)x2−11. Splitting x−1x+1 into x−1x+1 is allowed here because x>1 makes both positive; for x<−1 it would be false, and that is why the question fixes the domain. At x=3: 481=821=162.
c) Rewrite: y=(9−x2)−1/2. With u=9−x2 and u′=−2x: y′=−21(9−x2)−3/2⋅(−2x)=(9−x2)3/2x. Two minus signs meet and cancel: one from the exponent, one from the inner derivative. Losing one of them changes the sign of every slope, and a check is available: y is smallest at x=0 and grows as ∣x∣ approaches 3, so y′ must be positive for x>0. At x=5: 9−5=4, 43/2=8, so y′(5)=85.
d) The last operation is a product, so the product rule comes first, and each factor needs the chain rule: [(x2+1)3]′=3(x2+1)2⋅2x and [(2x−5)4]′=4(2x−5)3⋅2. So y′=6x(x2+1)2(2x−5)4+8(x2+1)3(2x−5)3. Factor out the common powers, the LOWEST power of each bracket: y′=(x2+1)2(2x−5)3[6x(2x−5)+8(x2+1)]=(x2+1)2(2x−5)3(20x2−30x+8)=2(x2+1)2(2x−5)3(10x2−15x+4). Expanding everything into a polynomial of degree 13 is correct and useless: the factored form is the one the next question needs.
e) Horizontal tangent where y′=0. A product is zero when one factor is: x2+1=0 has no real solution; 2x−5=0 gives x=25; 10x2−15x+4=0 has discriminant 225−160=65>0, so x=2015±65. Three horizontal tangents, at x=25 and at the two roots, which lie between 0 and 25 since 8<65<9. From the expanded form of degree 13, this question would be impossible by hand: factoring the derivative is part of the chain rule, not an optional cleanup.
Part B: problems and reasoning (/50)
Exercise 6: Second derivatives: the chain inside the product
The first derivative of a composition is a PRODUCT, f′(g(x))⋅g′(x). Differentiating it again therefore needs the product rule first, and the chain rule again inside the first factor. Nothing new is required, only the discipline to see the product that the first derivative has created.
Write the first derivative in its simplest factored form before differentiating a second time: every factor you leave in costs one more application of the product rule.
a) Let f(x)=e−x2. Find f′(x) and f′′(x), and give f′′(0).
b) Find the second derivative of g(x)=sin(x2).
c) Find the second derivative of h(x)=1+x2 and simplify it to a single power.
d) Prove that for twice differentiable f and g, (f(g(x)))′′=f′′(g(x))[g′(x)]2+f′(g(x))g′′(x). Check the formula on a).
e) We know g(1)=2, g′(1)=3, g′′(1)=−1, f′(2)=4 and f′′(2)=5. Find (f∘g)′′(1). A student answers −5: what formula did the student use?
Show the solution
Answers
a)f′(x)=−2xe−x2, f′′(x)=(4x2−2)e−x2, f′′(0)=−2
b)g′′(x)=2cos(x2)−4x2sin(x2)
c)h′′(x)=(1+x2)−3/2
d)Product rule on f′(g(x))⋅g′(x), chain rule on f′(g(x)); on a) it gives 4x2e−x2−2e−x2.
e)(f∘g)′′(1)=5⋅32+4⋅(−1)=41; the student wrote f′′(g(1))g′′(1).
a) Let u=−x2: f′(x)=e−x2⋅(−2x)=−2xe−x2. This is a PRODUCT of −2x and e−x2, so the product rule applies, with the chain rule on the second factor: f′′(x)=−2⋅e−x2+(−2x)⋅(−2xe−x2)=(4x2−2)e−x2. At 0: f′′(0)=−2. The classic loss is to differentiate −2xe−x2 as −2e−x2, treating the x in front as a constant, or as 4x2e−x2, forgetting the derivative of the first factor. Each gives half of the right answer.
b) g′(x)=cos(x2)⋅2x=2xcos(x2), a product. Product rule: (2x)′cos(x2)+2x(cos(x2))′, and the second derivative needs the chain rule again, (cos(x2))′=−sin(x2)⋅2x. So g′′(x)=2cos(x2)+2x⋅(−2xsin(x2))=2cos(x2)−4x2sin(x2). The inner derivative 2x appears twice in the second term, once from each differentiation: this x2 is the trace of the square [g′(x)]2 of the general formula in d).
c) h=(1+x2)1/2, so h′(x)=21(1+x2)−1/2⋅2x=x(1+x2)−1/2. Product rule: h′′(x)=(1+x2)−1/2+x⋅(−21)(1+x2)−3/2⋅2x=(1+x2)−3/2[(1+x2)−x2]=(1+x2)−3/2. Factoring out the LOWEST power, (1+x2)−3/2, makes the bracket collapse to 1. Writing h′ as a product with a negative exponent, rather than as the quotient 1+x2x, avoids a quotient rule with a chain rule inside its numerator.
d) By the chain rule, (f∘g)′(x)=f′(g(x))⋅g′(x), a product of A(x)=f′(g(x)) and B(x)=g′(x). By the product rule, (f∘g)′′=A′B+AB′. Now A=f′∘g is itself a composition, so A′(x)=f′′(g(x))⋅g′(x) by the chain rule, while B′=g′′. Hence (f∘g)′′(x)=f′′(g(x))[g′(x)]2+f′(g(x))g′′(x). Check on a): f=exp outside, g(x)=−x2, g′=−2x, g′′=−2, so e−x2⋅4x2+e−x2⋅(−2)=(4x2−2)e−x2, as found.
e) By d), (f∘g)′′(1)=f′′(g(1))[g′(1)]2+f′(g(1))g′′(1)=f′′(2)⋅9+f′(2)⋅(−1)=45−4=41. The student's −5=f′′(2)⋅g′′(1) comes from the false rule (f∘g)′′=f′′∘g⋅g′′, a chain rule for second derivatives that does not exist. It fails even for g(x)=2x, f(x)=x2: (f∘g)(x)=4x2 has second derivative 8, while f′′(2x)⋅g′′(x)=2⋅0=0. Note also that f′′ and f′ are read at g(1)=2, never at 1.
Exercise 7: Tangent lines to composite curves
The tangent line to y=F(x) at x=a is y=F(a)+F′(a)(x−a). With a composite F, the only new work is F′(a), and the usual order is: differentiate in general, factor, THEN substitute x=a. A horizontal tangent is a zero of F′, and the factored form is what makes it readable.
The figure shows y=esinx on [0,2π], with its points at x=0 and x=π.
a) Find the tangent and the normal lines to y=(x2−3)4 at x=2, then every x where this curve has a horizontal tangent.
b) Find the tangent lines to y=esinx at x=0 and at x=π, and the point where they meet.
c) Find the points of y=esinx, 0≤x≤2π, with a horizontal tangent. Compare the highest one with the meeting point of b).
d) Find the points of y=sin2x+2sinx, 0≤x≤2π, with a horizontal tangent.
e) Find the point of y=x2+9 where the tangent is parallel to the line y=2x.
Show the solution
Answers
a)Tangent y=16x−31, normal y=−16x+89; horizontal at x=0, x=±3
b)y=1+x and y=1+π−x, meeting at (2π,1+2π)
c)(2π,e) and (23π,e1); the meeting point lies below the top, 1+2π<e
d)(3π,233), (π,0), (35π,−233)
e)(3,23); the candidate x=−3 is rejected (slope −21).
a) With u=x2−3: y′=4(x2−3)3⋅2x=8x(x2−3)3. At x=2: u=1, so y(2)=1 and y′(2)=16. Tangent: y=1+16(x−2)=16x−31. The normal is perpendicular, slope −161: y=1−161(x−2)=−16x+89. Horizontal tangents where 8x(x2−3)3=0: x=0 or x2=3, so x=0 and x=±3. Forgetting the inner factor 2x would give the slope 4 at x=2 and would lose the horizontal tangent at x=0 entirely: the inner derivative carries zeros of its own.
b) y′=esinxcosx (outer eu, inner u=sinx). At 0: y(0)=e0=1 and y′(0)=1⋅cos0=1, so the tangent is y=1+x. At π: y(π)=e0=1 and y′(π)=1⋅cosπ=−1, so the tangent is y=1−(x−π)=1+π−x. They meet where 1+x=1+π−x, that is x=2π, at the point (2π,1+2π). The two points have the same height, 1, and opposite slopes, so the meeting point sits exactly halfway between them, as the figure of the solution shows.
c) y′=esinxcosx=0. The factor esinx is never 0, so cosx=0: x=2π, where y=e1=e, and x=23π, where y=e−1=e1. Saying why the exponential factor cannot vanish is part of the answer. The highest point, (2π,e), lies directly above the meeting point of b), since 1+2π≈2.57<e≈2.72: the two tangents cross UNDER the top of the arch, where the curve bends down.
d) dxdsin2x=cos2x⋅2 by the chain rule, with inner function 2x. So y′=2cos2x+2cosx. With cos2x=2cos2x−1: y′=2(2cos2x+cosx−1)=2(2cosx−1)(cosx+1). Zeros: cosx=21, so x=3π or 35π, and cosx=−1, so x=π. Values: y(3π)=23+3=233, y(π)=0, y(35π)=−23−3=−233. The factor 2 of the inner derivative is what makes the identity factor neatly; without it, the equation cos2x+2cosx=0 has different, wrong solutions.
e) y=(x2+9)1/2, so y′=2x2+92x=x2+9x. Parallel to y=2x means slope 21: x2+9x=21, so 2x=x2+9, which forces x≥0. Squaring: 4x2=x2+9, x2=3, x=±3. The candidate −3 was created by the squaring: there the slope is −123=−21, so it is rejected. The point is (3,12)=(3,23). Every squaring of an equation must be followed by this check.
Exercise 8: Five statements to correct
Each statement below was written by a student on a MATH 140 assignment, and each is false. Say what is wrong, give a counterexample with numbers, and write the correct statement.
a) dxdsin(x2)=cos(2x).
b) If h(x)=f(g(x)), then h′(x)=f′(x)g′(x).
c) Since ex is its own derivative, dxdex2=ex2.
d) The derivative of an even function is even.
e) If g(x)=f(2x), then g′(1)=f′(2).
Show the solution
Answers
a)False: dxdsin(x2)=2xcos(x2); at x=0 the claim gives 1 instead of 0.
b)False: h′(x)=f′(g(x))g′(x); with f(x)=x2, g(x)=x+1, h′(0)=2, not 0.
c)False: dxdex2=2xex2; at x=0 the claim gives 1 instead of 0.
d)False: x2 is even and 2x is odd. The derivative of an even function is ODD.
e)False: g′(1)=2f′(2); with f(x)=x2, g′(1)=8 and f′(2)=4.
a) FALSE. The student differentiated the inside and put it INSIDE the cosine. The sine is the outer layer and receives x2, so its derivative is read at x2, and the inner derivative 2x multiplies: dxdsin(x2)=cos(x2)⋅2x. Counterexample at x=0: the true slope is 2⋅0⋅cos0=0 (the function sin(x2) is even, its graph is flat at 0), while the claim gives cos0=1. Correct statement: (sinu)′=cosu⋅u′, the factor outside, never inside.
b) FALSE. The outer derivative must be evaluated at g(x), not at x. Take f(x)=x2 and g(x)=x+1: h(x)=(x+1)2, so h′(x)=2(x+1) and h′(0)=2; the claim gives f′(0)g′(0)=0⋅1=0. Correct statement: h′(x)=f′(g(x))g′(x), where f′ is read at the point that g hands to f. This is the error the table questions are built to catch.
c) FALSE. ex is its own derivative with respect to its OWN exponent; with the exponent u=x2, the chain rule adds the factor u′=2x: dxdex2=2xex2. At x=0 the true slope is 0, and indeed ex2 is even with its minimum at 0, while the claim gives e0=1. At x=1 the true slope is 2e, twice the claimed e. Correct statement: (eu)′=euu′.
d) FALSE. f(x)=x2 is even, and f′(x)=2x is odd. The chain rule proves the correct statement: if f(−x)=f(x) for all x, differentiate both sides; the left side is a composition with inner function −x, so −f′(−x)=f′(x), that is f′(−x)=−f′(x). Correct statement: the derivative of a differentiable even function is ODD, and by the same argument the derivative of an odd function is even (sin and cos).
e) FALSE. g(x)=f(2x) is a composition with inner function 2x, so g′(x)=f′(2x)⋅2 and g′(1)=2f′(2). Counterexample: f(x)=x2 gives g(x)=4x2, g′(1)=8, while f′(2)=4. Geometrically, y=f(2x) is the graph of f squeezed horizontally by a factor 2, so every slope doubles. Correct statement: dxdf(kx)=kf′(kx).
Exercise 9: A cost in cascade: dollars per unit, units per hour, hours per week
A new workshop increases its working hours as it gains customers. Week t after opening (0≤t≤16), it works h=36+4t hours. Working h hours in a week produces q=30h units, and producing q units in a week costs C=800+5q+100q2 dollars.
The weekly cost depends on the production, which depends on the hours, which depend on the week: a chain of three links, drawn in the figure with the derivative of each link. The chain rule multiplies along the arrows, each derivative read at the value its own variable takes at that moment.
a) Give the unit of dtdh, dhdq and dqdC, and show that the units of their product are those of dtdC.
b) Find dtdC at t=7, with a sentence of interpretation.
c) A student computes C′(7)⋅q′(7)⋅h′(7). Explain why this makes no sense, using units.
d) Write C directly as a function of t, differentiate, and check b).
e) Find dtdC at t=16. The cost of each extra unit is higher than at t=7, yet the weekly cost grows more slowly. Which link of the chain explains it?
Show the solution
Answers
a)hours per week, units per hour, dollars per unit; the product is in dollars per week
b)dtdC=9.8⋅815⋅4=73.5 dollars per week
c)C′ must be read at q=240 units and q′ at h=64 hours, not at 7 weeks.
e)dtdC=11⋅23⋅4=66 dollars per week: dhdq falls from 815 to 23.
a) dtdh is a change of hours per change of weeks: hours per week. dhdq: units per hour. dqdC: dollars per unit. In Leibniz notation, dtdC=dqdC⋅dhdq⋅dtdh, and the units cancel like fractions: unitdollars⋅hourunits⋅weekhours=weekdollars. Leibniz notation is built so that the chain rule looks like a cancellation; it is not one, but the units check is real, and a product whose units do not reduce to those of the answer is wrong.
b) At t=7: h=36+28=64 hours and q=3064=240 units. The three derivatives: dtdh=4; dhdq=h15=815 at h=64; dqdC=5+50q=5+4.8=9.8 at q=240. So dtdC=9.8⋅815⋅4=9.8⋅7.5=73.5. At week 7, the weekly cost is rising at a rate of 73.5 dollars per week: next week's cost should be about 73.5 dollars higher. Each derivative is read at ITS variable's current value: 64 hours, 240 units, week 7.
c) C′(7)=5+507 is the cost of an extra unit when the workshop makes 7 units, a situation that never occurs at week 7; q′(7)=715 is the output of an extra hour when the workshop works 7 hours. The number 7 is a number of WEEKS, and it was fed to functions whose input is a number of units or of hours. This is exactly the error f′(x)g′(x) instead of f′(g(x))g′(x): the outer derivative must be evaluated at the inner function's value. Units expose it: C′ takes units, not weeks.
d) q=3036+4t and q2=900(36+4t), so C(t)=800+15036+4t+9(36+4t)=1124+36t+15036+4t. By the chain rule on the root, with inner function 36+4t: C′(t)=36+150⋅236+4t4=36+36+4t300. At t=7: 36+8300=36+37.5=73.5, as in b). Composing first and differentiating once is often shorter; the chain of b) is what an exam asks when the links are given separately, or only by a table.
e) At t=16: h=100, q=300, dqdC=5+6=11, dhdq=1015=23, dtdh=4, so dtdC=11⋅23⋅4=66 dollars per week; d) agrees, 36+10300=66. The first link grew, from 9.8 to 11 dollars per unit, and the last one is constant. The middle link fell, from 815 to 23 units per hour: each extra hour produces fewer units, so fewer extra units are paid for. A product of rates can decrease while one factor increases, and only the chain shows which factor is responsible.
Exercise 10: Unit conversions: every change of unit is an inner function
Changing the unit of the variable is a composition. If a quantity is known as a function of kelvins and you measure in degrees Fahrenheit, or of hours and you measure in minutes, the conversion formula is an inner function, and the chain rule multiplies every rate by the derivative of the conversion. That is where the factors 95, 601 and 180π of this exercise come from.
In dry air, the speed of sound is well modelled by v=20T metres per second, where T is the absolute temperature in kelvins. The conversions are T=C+273 and C=95(F−32), with C in degrees Celsius and F in degrees Fahrenheit.
a) Find v and dTdv at T=289 K, with units.
b) At 16 degrees Celsius, find the rate of change of v per degree Celsius, then per degree Fahrenheit.
c) One morning the air warms at 3 degrees Celsius per hour at the moment it reaches 16 degrees Celsius. How fast is the speed of sound changing, in metres per second per hour, then per minute?
d) Let S(x)=sin(x∘), the sine of an angle of x DEGREES. Find S′(x), S′(0) and S′(60).
e) Prove that if g(x)=f(kx+c) then g′(x)=kf′(kx+c), and identify k in b), c) and d). Explain why calculus measures angles in radians.
Show the solution
Answers
a)v=340 m/s, dTdv=1710 m/s per kelvin
b)1710 m/s per degree Celsius; 15350 m/s per degree Fahrenheit
c)1730 m/s per hour, that is 341 m/s per minute
d)S′(x)=180πcos(x∘), S′(0)=180π, S′(60)=360π
e)k=1 then 95 in b), 601 in c), 180π in d); only in radians is k=1 and (sinx)′=cosx.
a) v(289)=20289=20⋅17=340 m/s. With v=20T1/2: dTdv=10T−1/2=T10, so at T=289, dTdv=1710 m/s per kelvin, a little more than half a metre per second for each kelvin. The unit of a derivative is always the unit of the output divided by the unit of the input: here Km/s.
b) 16 degrees Celsius is T=289 K. As a function of C, v=20C+273, a composition with inner function C+273, whose derivative is 1: dCdv=dTdv⋅dCdT=1710⋅1=1710 m/s per degree Celsius. A degree Celsius and a kelvin have the same size, only the zero moves, so the rate does not change. In Fahrenheit, dFdT=dCdT⋅dFdC=1⋅95, so dFdv=1710⋅95=15350 m/s per degree Fahrenheit. The rate is smaller because a degree Fahrenheit is smaller: only 95 of a kelvin. The +273 and the −32 shift the zero and disappear on differentiation; only the slope of the conversion survives.
c) The chain is t→C→v: dtdv=dCdv⋅dtdC=1710⋅3=1730 m/s per hour, at the moment when C=16. Per minute: an hour is 60 minutes, so the time in hours is 60m for m minutes, a conversion of derivative 601, and 1730⋅601=341 m/s per minute. Check the units: degreem/s⋅hourdegrees⋅minutehours reduces to metres per second per minute. Multiplying by 60 instead of dividing is the usual slip, and the units catch it.
d) An angle of x degrees measures 180πx radians, and the sine of calculus takes radians: S(x)=sin180πx. With inner function u=180πx and u′=180π: S′(x)=180πcos180πx=180πcos(x∘). So S′(0)=180π and S′(60)=180π⋅21=360π. In degrees, the sine curve is stretched over 360 units instead of 2π, so its slope at 0 is 180π, about 0.017, and not 1: the figure of the solution shows how gentle that tangent is.
e) g=f∘ℓ with inner function ℓ(x)=kx+c and ℓ′(x)=k, so g′(x)=f′(kx+c)⋅k by the chain rule: the constant c vanishes and the factor k stays. In b), k=1 for T=C+273 and k=95 for C=95F−9160; in c), k=601 from hours to minutes; in d), k=180π from degrees to radians. The formula (sinx)′=cosx holds only when x is in radians: it is the one unit in which the conversion factor is 1, so no extra factor appears. In degrees every derivative of a trigonometric function would carry 180π, which is why calculus, and every formula of MATH 140, works in radians.