This sheet is not a summary of section 5.5 of Stewart: you already have the course notes. It answers one question only, what makes students lose marks on the substitution rule in MATH 141 at McGill University, and which precise gesture avoids each loss.
Everything below is done by hand, as on the exam, and every antiderivative quoted has been checked the only way that settles it: by differentiating it back to the integrand.
The thread of the chapter
A substitution is finished only when the integral contains NO x at all: not in the integrand, not in the differential, not in the bounds. Every mark lost on this chapter is an x left somewhere: a variable factor divided out of du, a stray x treated as a constant, a bound that stayed in x.
•∫f(g(x))g′(x)dx=∫f(u)du with u=g(x) and du=g′(x)dx. Read backwards, it is the chain rule dxdF(g(x))=F′(g(x))g′(x).
•Choose u = the INSIDE function whose derivative is also a factor of the integrand: x3+1 in x2(x3+1)5, lnx in x(lnx)3, x in xcosx.
•A CONSTANT missing from du is supplied: u=1+x2 gives xdx=21du. A factor containing x is NEVER supplied by dividing: then this u is the wrong one.
•An x still present after du is used is rewritten through u: u=x−1 gives x=u+1; u=x2+1 gives x2=u−1.
•Indefinite integral: end in x, with +C, and differentiate the answer. The chain rule must give back the integrand exactly.
Before writing anything, ask: which function, differentiated, gives the other factor up to a constant? If none does, substitution is not the tool, and forcing it is where the marks go.
Definite integrals: the bounds travel through u
•Route 1: antiderivative in x, then the ORIGINAL bounds. Route 2: new bounds u=g(a) and u=g(b), and never come back to x.
•The mixture, an integrand in u with bounds in x, is always wrong. It is an unfinished substitution: an x is still hiding in the bounds.
•New bounds in decreasing order are normal (as with u=cosx on [0,2π]). Keep them; the minus sign of du and the reversed order cancel together.
•No one-to-one condition: g′ continuous on [a,b] is enough. u=sinx on [0,π] is legal.
•f continuous on [−a,a]: odd gives ∫−aaf=0, even gives 2∫0af. Both are proved by u=−x.
The bounds are read on the vertical axis: for u=1+x2, the interval 0≤x≤3 becomes 1≤u≤4. Neither bound stays where it was.
Route 2 is shorter, and the line when x=a, u=…; when x=b, u=… is exactly what the marker looks for.
The rules in table form
Each row reads from left to right: the assumptions, then the result. A red cell is not an answer, it is the finding that the form settles nothing and the instruction to rewrite it. Every case is followed by a worked example.
Which u, and which antiderivative comes out
Read a line as: for this integrand, this choice of u gives this antiderivative. Each example is a definite integral computed by hand. The red lines are rules students invent.
Integrand
Choice of u
Antiderivative
x1+x2
u=1+x2, xdx=21du
31(1+x2)3/2+C
Example: ∫03x1+x2dx=31(8−1)=37.
1+x2x
u=1+x2
21ln(1+x2)+C
Example: ∫011+x2xdx=21ln2.
1+x4x
u=x2
21arctan(x2)+C
Example: ∫011+x4xdx=21⋅4π=8π.
tanx
u=cosx
−ln∣cosx∣+C
Example: ∫0π/3tanxdx=−ln21=ln2.
x2+91
u=3x, dx=3du
31arctan3x+C
Example: ∫03x2+9dx=31⋅4π=12π.
xx−1
u=x−1, x=u+1
52(x−1)5/2+32(x−1)3/2+C
Example: ∫12xx−1dx=52+32=1516.
1+x21
u=1+x2
ln(1+x2)no such rule
Example: The derivative of ln(1+x2) is 1+x22x; and ∫011+x2dx=4π≈0.785, not ln2≈0.693.
What to do: No x on top, no du: it is the table, ∫1+x2dx=arctanx+C.
(x3+1)5
u=x3+1
18x2(x3+1)6no such rule
Example: Its derivative at x=1 is 32−964, not the integrand's 25=32.
What to do: No x2 in the integrand: expand (x3+1)5 by the binomial theorem and integrate term by term.
Every blue line is proved in one line by differentiating the last column. If the derivative of your answer carries an extra term, the rule you used does not exist.
The mistakes that cost marks
These are the errors I correct most often in session. Each one costs marks on a paper, even when the reasoning behind it is right.
1.Dividing by a variable to manufacture du
the whole question
What not to write
“u=x3+1, dx=3x2du, so ∫(x3+1)5dx=18x2(x3+1)6+C.”
What to write
“There is no x2 to form du, so this u does not apply: (x3+1)5=x15+5x12+10x9+10x6+5x3+1, integrated term by term.”
Why: Once dx=3x2du is written, the integral in u still contains x, and pulling 3x21 out treats a variable as a constant. Differentiating the answer produces the extra term −9x3(x3+1)6: that term is the error made visible.
2.Keeping the bounds of x in an integral in u
2 to 3 marks, the whole evaluation
What not to write
“u=x2: ∫02xex2dx=21∫02eudu=2e2−1.”
What to write
“When x=0, u=0; when x=2, u=4: ∫02xex2dx=21∫04eudu=2e4−1.”
Why: The numbers 0 and 2 are values of x, and the new integral is in u. The bound line is part of the substitution, like du: an integral with u inside and x on the bounds is still half in x.
3.Reordering reversed bounds and keeping the minus sign
1 to 2 marks, and a negative value for a positive integrand
“When x=0, u=1; when x=1, u=0: −21∫10udu=21∫01udu=31.”
The integrand x1−x2 is positive on (0,1), so its integral is the shaded area, 31. A negative answer is impossible, whatever the substitution did to the bounds.
Why: The reversed order and the minus of du come as a pair: either keep both, −21∫10, or drop both, 21∫01. Using one without the other flips the sign, and the figure shows it cannot be negative.
4.Treating a stray x as a constant
the whole question
What not to write
“u=x−1: ∫xx−1dx=x∫udu=32x(x−1)3/2+C.”
What to write
“x=u+1: ∫(u+1)udu=52(x−1)5/2+32(x−1)3/2+C.”
Why: Only a number passes through the integral sign. The derivative of the wrong answer is xx−1+32(x−1)3/2: at x=2 it gives 38 where the integrand is 2. Solve u=g(x) for x and replace it.
5.Losing the minus sign in the integral of tan x
1 to 2 marks, and a negative value for every definite integral of tan on the first quarter turn
tanx is positive, so its antiderivative must climb: −ln(cosx) does, ln(cosx) falls below zero. The sign error is visible without a single computation.
Why: The derivative of cosx is −sinx, and the minus sign goes into the answer. Quick test: tanx>0 on (0,2π), so its antiderivative must INCREASE there; ln(cosx) decreases.
6.Dropping the absolute value, then calling the integral undefined
2 marks, the whole value
What not to write
“∫−10x2−42xdx=[ln(x2−4)]−10, and ln(−4) does not exist.”
What to write
“[ln∣x2−4∣]−10=ln4−ln3=ln34.”
Why: ln∣u∣ is an antiderivative of u1 where u<0 too. The integrand is continuous on [−1,0], so the integral exists. Drop the bars only after proving the argument positive, as for x2+3x+5.
7.Quoting symmetry on an interval that is not symmetric
the whole question
What not to write
“x3 is odd, so ∫−12x3dx=0.”
What to write
“Only [−1,1] is symmetric: ∫−12x3dx=0+∫12x3dx=4−41=415.”
Why: The theorem has two hypotheses, an interval [−a,a] and continuity on it. On [−1,1], x31 fails the second: it is not a zero integral but an improper one.
8.Forgetting the factor of a linear substitution
1 to 2 marks
What not to write
“∫x2+9dx=91arctanx+C.”
What to write
“x2+9=9(1+(3x)2); with u=3x, dx=3du: ∫x2+9dx=31arctan3x+C.”
Why: 91⋅1+x21=9+9x21, which is not x2+91. Factoring the 9 leaves (3x)2, not x2, and dx=3du turns 91 into 31.
Which method to choose
Which u, decided by the FORM of the integrand
Look at the integrand and at the bounds before choosing anything
1+x2sinx is odd: on [−1,1] the two shaded lobes are mirror images of opposite sign, so the integral is 0 even though no antiderivative can be written.
If the interval is symmetric, [minus a, a], and the integrand is continuous on it → check parity first: an odd integrand gives 0 with no computation, an even one gives twice the integral over [0, a]
Example: ∫−111+x2sinxdx=0
If a function sits inside a power, a root, an exponential or a trig function, and its derivative is a factor up to a constant → u = the inside function; supply the constant
Example: x2(x3+1)5: u=x3+1, x2dx=31du
If a fraction whose numerator is the derivative of its denominator, up to a constant → u = the denominator; the answer is ln∣u∣
Example: x2+3x+52x+3 gives ln(x2+3x+5)+C
If x on top, and x4 in 1+x4 or in a2−x4 → u=x2, then the table: inverse tangent or inverse sine
Example: 1+x4x gives 21arctan(x2)+C
If the inside function is found, but an x is left over after du → solve u=g(x) for x (or for x2) and replace it
Example: x3x2+1: x2=u−1, xdx=21du
If a constant a2 added to x2, or a function of ax+b → linear substitution u=ax or u=ax+b, and keep its factor
Example: x2+91 gives 31arctan3x+C
If no inside function has its derivative present, and nothing can be rewritten → substitution is not the tool: expand, use the table, or wait for the later techniques
Example: (x3+1)5: expand into six terms
The test that decides every branch is the same: which function, differentiated, gives the other factor up to a constant? The same denominator 1+x2 gives a logarithm with x on top and an inverse tangent with 1 on top.
How the answer is expected to be written
A marker ticks steps. Here they are in order, with the concluding sentence expected word for word.
Writing a definite substitution the marker can follow
When to use it: Any integral where the substitution is not a one-line mental step, and every definite integral
1Name u and compute du on its own line, with the constant solved for: u=1+x2, du=2xdx, xdx=21du.
2Write the bound line: when x=a, u=g(a); when x=b, u=g(b). Keep them in that order, even if decreasing.
3Rewrite EVERY piece of the integral in u, then check that no x is left: not in the integrand, not in dx, not in the bounds. A leftover x is rewritten through u.
4Integrate in u and evaluate at the new bounds. Do not return to x.
5Check: sign of the integrand against the sign of the answer, and a rough size (width times a typical height).
Concluding sentence
“Let u=1+x2, du=2xdx; when x=0, u=1 and when x=3, u=4. Then ∫03x1+x2dx=21∫14u1/2du=31[u3/2]14=37.”
The trap: Writing ∫x1+x2dx=∫u with no du and no bound line: the chain of equalities is false as written, and markers treat it as such.
Marking: Typically 1 mark for u and du, 1 for the new bounds, 1 for the antiderivative, 1 for the value. The bound line is the one most often missing.
Check before you hand in
Five minutes of checking recover more marks than one more problem started in a hurry.
Differentiate the answer
For an indefinite integral, the chain rule applied to your answer must return the integrand exactly. An extra term means a variable was treated as a constant.
253/2=125 and 163/2=64: the numbers 16 and 25 were chosen as perfect squares, so the evaluation is done by hand. Evaluated at the new bounds, never at 0 and 3.
Step 5
Check by route 1: F(x)=31(x2+16)3/2−16(x2+16)1/2 gives F(3)−F(0)=(3125−80)−(364−64)=313. Size: the integrand rises from 0 to 527, so 0<313<3⋅527.
Why
Two independent routes and a size bracket confirm the value. On an exam, the size check alone takes ten seconds and catches a bound left in x.
The conclusion, written out
“∫03x2+16x3dx=313, with u=x2+16, x2=u−16 and bounds 16 and 25.”
The classic mistake on this problem: Pulling the leftover x2 out of the integral as a constant, or evaluating 31u3/2−16u1/2 at u=0 and u=3: both leave an x in the computation, the first in the integrand, the second in the bounds.
Learn by heart
•u = the inside function whose derivative is a factor, up to a constant. Constants are supplied; variables never are.
•No x left: not in the integrand, not in dx, not in the bounds. A stray x is rewritten through u.
•Definite: new bounds g(a), g(b), kept in their order; or antiderivative in x with the old bounds. Never half of each.
•Symmetric [−a,a], continuous: odd gives 0, even gives 2∫0a. Even times odd is odd.
•∫0af(a−x)dx=∫0af(x)dx.
•Always differentiate an indefinite answer, and check the sign of a definite one.
Frequently asked questions
How do I choose u for a substitution in calculus?
Look for a function sitting inside another one, a power, a root, an exponential, whose derivative also appears as a factor of the integrand, possibly up to a constant. That inside function is u. If the derivative is missing by a constant, supply the constant; if it is missing by a factor containing x, this u is the wrong choice.
Do I have to change the limits of integration when I substitute?
Either change them, or go back to x before evaluating, but never half of each. If the integral is written in u, the limits must be values of u: plug the old limits into u equals g of x. If you prefer to keep the old limits, substitute back to x first. An integral in u with limits in x gives a wrong number.
What happens when the new lower limit is bigger than the upper limit?
Nothing goes wrong: keep the limits in the order the substitution gives them. The minus sign that usually comes with du and the reversed order cancel each other. You may also swap the limits and drop the minus sign at the same time, but never do only one of the two, or the sign of the answer flips.
Why is the integral of tan x equal to minus ln of cos x?
Write tan x as sine over cosine and let u be cos x. Then du is minus sine of x dx, so the integral becomes minus the integral of du over u, which is minus ln of the absolute value of cos x, plus C. The same answer is often written as ln of the absolute value of sec x. The minus sign comes from the derivative of cosine.
What do I do if an x is left over after the substitution?
Solve the substitution for x and replace it. With u equals x minus 1, write x as u plus 1; with u equals x squared plus 1, write x squared as u minus 1. The integral in u must contain no x at all. Never pull the leftover x out of the integral as if it were a constant: only numbers can be moved through the integral sign.
Practise it
Corrected exercises: The substitution rule, MATH 141 at McGill
A method is proved on a paper, not on a sheet. The set for the same chapter takes each of these traps into a problem, with the solution written out step by step.
Get in touch for a first session. Substitution is the first technique of integration in the course, and every later one, from parts to partial fractions, leans on it.