← back to MATH 101
16 worked examples25 exercises5 figuresbölüm 14 / 14

14Trig substitution, partial fractions, numerical and improper integrals

$\int x\sqrt{9-x^{2}}\,dx$ takes ten seconds: the $x$ standing in front is exactly what $u=9-x^{2}$ needs. Rub that single $x$ out and the survivor, $\int \sqrt{9-x^{2}}\,dx$, stops every rule we own. It is not that the answer is hidden — the graph of $y=\sqrt{9-x^{2}}$ is a quarter circle of radius $3$ over $[0,3]$, so the number is $9\pi/4$ and we can read it off the picture. We simply have no machinery that produces it.

By the end of this section you can take an integral you have never seen, name in under a minute which of four tools it wants, carry that tool out, and check your own answer without an answer key.

In 60 seconds

Three shapes, three tools: a root of a quadratic wants a triangle, a ratio of polynomials wants a template of simpler fractions, and an integrand with no elementary antiderivative wants either a table entry or a handful of samples.

$\sqrt{a^{2}-x^{2}}\to x=a\sin\theta,\quad \sqrt{a^{2}+x^{2}}\to x=a\tan\theta,\quad \sqrt{x^{2}-a^{2}}\to x=a\sec\theta$

a root of a quadratic, and no spare $x$ in the numerator to spend on $du$

Partial fraction template
$\frac{A_{1}}{x-a}+\cdots+\frac{A_{k}}{(x-a)^{k}}+\frac{B_{1}x+C_{1}}{x^{2}+bx+c}+\cdots$

a ratio of polynomials with the numerator of lower degree

Simpson's rule
$S_{n}=\frac{\Delta x}{3}\left[f(x_{0})+4f(x_{1})+2f(x_{2})+\cdots+4f(x_{n-1})+f(x_{n})\right]$

no antiderivative in sight and $n$ is even

$\int_{a}^{\infty}f(x)\,dx=\lim_{t\to\infty}\int_{a}^{t}f(x)\,dx$

an endpoint is infinite, or the integrand blows up at one

Three most common mistakes
  1. Stopping at $\theta$. $\tfrac12\theta+\tfrac12\sin\theta\cos\theta$ answers a question nobody asked; the question was in $x$, and the is what finishes the job.

  2. Putting a single constant over an . Above $x^{2}+4$ the numerator is $Bx+C$ — two unknowns, always, no matter how simple the fraction looks.

  3. Writing $\bigl[F(x)\bigr]_{1}^{\infty}$. The upper endpoint of that bracket is a limit to be taken, not a number to be substituted.

Midterm 1, Midterm 2 and the Final carry 28 percent each; quizzes 10 and homework 6. The weights say nothing about which of the four tools a given question wants, and that decision is made in the first ten seconds — so practise the choosing, not only the executing.

How much time do you have?
10 minutes

You leave able to spot which of the three structures you are looking at and to run the two that appear most often.

In 60 seconds card, Trading a root for a triangle, Splitting a ratio of polynomials into pieces you already know, Formula card
45 minutes

Add the two rewrites that make the templates apply at all, plus the ordered strategy — enough for a full exam question of this type.

everything in the 10 minute path, When the quadratic is not a template yet, Choosing the technique before you start writing, Full exam-style question, Practice C
full read

The parts that transfer: reading an entry out of a table without being fooled, what to do when no antiderivative exists, and the interleaved set where the type of the question is hidden from you.

all blocks in order, Reading an answer out of a table, When there is no antiderivative: sampling instead of solving, Integrals that run to infinity, Scaffolding comes off, Practice A to D, Mistake ledger
By the end of this section
  1. Choose the substitution a root of a quadratic asks for, carry it out, and convert the answer back to $x$ with a reference triangle.

  2. Rewrite a quadratic with a linear term by , so that one of the three templates applies to it.

  3. Decompose a into partial fractions and integrate each piece.

  4. Decide, before writing anything, which technique is cheapest for a given integrand and say why the others cost more.

  5. Match an integrand to a table entry by a linear substitution, and check the result by differentiating it.

  6. Approximate a definite integral with the midpoint, trapezoidal or Simpson rule and bound the error you have committed.

  7. Settle whether an improper integral converges, by evaluating a limit or by naming a comparator whose verdict is already known.

Syllabus coverage
7.3

Trigonometric substitution

The three templates, the reference triangle that converts the answer back, and the rewrite that has to happen first when the quadratic carries a linear term.

covered
7.4

Integration of rational functions by partial fractions

Properness check and long division, the four template cases, solving for the unknowns, and the standard antiderivative of every piece.

covered
7.5

Strategy for integration

The ordered list of questions to ask an unfamiliar integrand, and the honest last branch: some integrands have no elementary antiderivative at all.

covered
7.6

Integration using tables and technology, together with the numerical rules that take over when no entry fits

Matching an integrand to a table entry through a linear substitution is the first half; the midpoint, trapezoidal and Simpson rules, taught in the block on sampling, are the second half.

covered
improper integrals

Improper integrals, revisited where the techniques of this week are what make them computable at all

The syllabus lists improper integrals with the previous week's sections; they come back here because a or a completed square is usually the missing antiderivative in the limit.

covered
Recall first
Substitution

$\int f(g(x))\,g'(x)\,dx=\int f(u)\,du$ with $u=g(x)$.

Half of the decisions in this section are the single question: is a $du$ already sitting in the numerator? If it is, none of the heavy machinery below is needed.

The three Pythagorean identities

$1-\sin^{2}\theta=\cos^{2}\theta$, $1+\tan^{2}\theta=\sec^{2}\theta$, $\sec^{2}\theta-1=\tan^{2}\theta$.

One per template. The identity, not the algebra, is what removes the root.

Two standard antiderivatives

$\int\frac{du}{a^{2}+u^{2}}=\frac{1}{a}\arctan\frac{u}{a}+C$ and $\int\frac{du}{\sqrt{a^{2}-u^{2}}}=\arcsin\frac{u}{a}+C$.

After a decomposition or a completed square, every surviving piece is one of these two or a logarithm. Note the $\frac{1}{a}$ in the first and its absence in the second.

The secant integral

$\int\sec\theta\,d\theta=\ln\lvert\sec\theta+\tan\theta\rvert+C$.

The tangent substitution lands on it again and again; without it, half of the $\sqrt{a^{2}+x^{2}}$ problems stall one line from the end.

An improper integral is a limit

$\int_{a}^{\infty}f=\lim_{t\to\infty}\int_{a}^{t}f$, and $\int_{a}^{b}f=\lim_{t\to b^{-}}\int_{a}^{t}f$ when $f$ blows up at $b$.

The definition was set up in the previous section. What was missing there was a way to compute the inner integral; that is what this section supplies.

Polynomial long division

For polynomials $P$ and $Q$ there are unique $S$ and $R$ with $\frac{P(x)}{Q(x)}=S(x)+\frac{R(x)}{Q(x)}$ and $\deg R<\deg Q$.

A partial fraction template is only valid on the remainder part, so the division has to happen first whenever the numerator degree is not lower.

Try it yourself first (3 questions)
1§14.1 — a root is not automatically the variable●○○○○

Before any new machinery, one habit has to be checked. Decide true or false and be ready to defend it with a number, not with a rule.

Given
  • Claim: $\sqrt{x^{2}}=x$ for every real number $x$.

Find
  1. True or false, with a reason or a counterexample.

Hint 1/4

Do not reason about it in general. Try to break it with one number.

Hint 2/4

$\sqrt{\;\cdot\;}$ always returns the non-negative root, so the output can never be negative.

Hint 3/4

Take $x=-3$: the claim says $\sqrt{(-3)^{2}}=-3$, while $\sqrt{9}=3$.

Hint 4/4

One number breaks it, so the claim is false and the correct identity is $\sqrt{x^{2}}=\lvert x\rvert$.

Show solution
Test the claim
$\sqrt{(-3)^{2}}=\sqrt{9}=3$

the root symbol returns the non-negative root

$3\neq -3$

so the claimed equality fails at this one number

$\sqrt{x^{2}}=\lvert x\rvert$

the identity that is true for every real $x$

Answer $$\text{false}$$
Check

Check the corrected identity on both signs: at $x=-3$ it gives $\lvert -3\rvert=3$ and at $x=3$ it gives $3$, matching $\sqrt{9}=3$ in both cases.

Every time a root of a square appears in this section, the bars are the default and dropping them has to be earned.

2§14.3 — the two standard forms that keep appearing●●○○○

The pieces produced by every decomposition in this section are logarithms, arctangents and powers. This one checks that the arctangent form is stored with its constant attached.

Given
  • $\displaystyle\int\frac{dx}{x^{2}+9}$

Find
  1. Which antiderivative is correct?

Hint 1/4

You are not being asked to invent anything: this is one of the two standard forms, and the only question is what the constant out front is.

Hint 2/4

$\int\frac{du}{a^{2}+u^{2}}=\frac{1}{a}\arctan\frac{u}{a}+C$.

Hint 3/4

Here $a^{2}=9$, so $a=3$: the form gives $\frac{1}{3}\arctan\frac{x}{3}$.

Hint 4/4

So the antiderivative is $\frac{1}{3}\arctan\frac{x}{3}+C$.

Show solution
Match the form
$a^{2}=9\;\Rightarrow\;a=3$

the constant of the form is the root of $9$, not $9$

$\int\frac{dx}{x^{2}+3^{2}}=\frac{1}{3}\arctan\frac{x}{3}+C$

the standard arctangent form with $a=3$

Answer $$\frac{1}{3}\arctan\frac{x}{3}+C$$
Check

Differentiate: $\frac{1}{3}\cdot\frac{1/3}{1+x^{2}/9}=\frac{1}{3}\cdot\frac{3}{9+x^{2}}=\frac{1}{x^{2}+9}$, which is the integrand.

3§14.3 — what has to happen before a template is written●●○○○

A trap that costs marks every year: the template for partial fractions is only valid for certain rational functions, and this one is not among them yet.

Given
  • $\displaystyle\int\frac{x^{3}+1}{x^{2}-4}\,dx$

Find
  1. What is the correct first move?

Hint 1/4

Compare the two degrees before you look at anything else.

Hint 2/4

A partial fraction template requires the numerator degree to be strictly lower than the denominator degree.

Hint 3/4

Here the numerator has degree $3$ and the denominator degree $2$, so the requirement fails.

Hint 4/4

Divide first: $\frac{x^{3}+1}{x^{2}-4}=x+\frac{4x+1}{x^{2}-4}$, and the template goes on the remainder.

Show solution
Compare degrees
$\deg(x^{3}+1)=3\;\ge\;\deg(x^{2}-4)=2$

the template is not licensed while this holds

Divide
$x^{3}+1=(x^{2}-4)\cdot x+(4x+1)$

one step of long division

$\frac{x^{3}+1}{x^{2}-4}=x+\frac{4x+1}{x^{2}-4}$

the remainder now has degree $1$, lower than $2$

Answer $$x+\frac{4x+1}{x^{2}-4}$$
Check

Multiply back: $(x^{2}-4)x+4x+1=x^{3}-4x+4x+1=x^{3}+1$, the original numerator.

If the degrees do not fall the right way, nothing else you do to the fraction is legal yet.

Notation
symbolreads asmeanswatch out
$\theta$

theta

the new variable a trigonometric substitution introduces; $x$ is written in terms of it, not the other way round

the final answer may not contain $\theta$

$a$

a

the positive constant inside the root, so $a=\sqrt{9}=3$ in $\sqrt{9-x^{2}}$

$a$ is the square root of the constant, not the constant

$u$

u

the shifted variable after completing the square, usually $u=x+\tfrac{b}{2}$

a definite integral in $u$ needs shifted limits as well

$\Delta x=\frac{b-a}{n}$

delta x equals b minus a over n

the common width of the $n$ subintervals of $[a,b]$

$n$ counts subintervals, so there are $n+1$ sample points

$x_{i}=a+i\,\Delta x$

x sub i

the $i$-th sample point, $i=0,1,\ldots,n$

the midpoint rule does not use these points but the midpoints between them

$M_{n},\;T_{n},\;S_{n}$

M sub n, T sub n, S sub n

the midpoint, trapezoidal and Simpson approximations with $n$ subintervals

$S_{n}$ exists only for even $n$

$K,\;L$

K and L

any numbers with $\lvert f''\rvert\le K$ and $\lvert f^{(4)}\rvert\le L$ on $[a,b]$

they are bounds, not maxima; a larger safe number is allowed and only weakens the estimate

$t$

t

the temporary endpoint that replaces the bad one in an improper integral

the answer is the limit as $t$ moves, not the value at any $t$

Conventions used here
Roots and signs

$\sqrt{u^{2}}=\lvert u\rvert$, never $u$. Each substitution below comes with a range of $\theta$ chosen so that the trig factor it produces is non-negative on that range; that is the licence — and the only licence — for writing $\sqrt{a^{2}\cos^{2}\theta}=a\cos\theta$ without bars.

A dropped absolute value is the single most common way a correct method produces a wrong sign.

Constants of integration

Every indefinite integral ends in $+C$. When a step produces a fixed constant, such as the $-\ln 2$ that falls out of a reference triangle, we absorb it into $C$ and say so in the line where it happens.

Silently dropping a constant and silently absorbing it look identical on paper; only one of them is legitimate.

Logarithms

$\int \frac{du}{u}=\ln\lvert u\rvert+C$ carries bars. Over an irreducible quadratic the argument $x^{2}+bx+c$ is positive for every real $x$, so $\ln(x^{2}+bx+c)$ needs none.

Bars written everywhere look careful but hide whether you know why they are there.

Angles

Every trigonometric function here takes radians, and the ranges of $\theta$ are quoted in radians.

The derivative and antiderivative formulas used throughout are false in degrees.

Reporting an improper integral

A verdict is either "converges to $L$" with the number $L$, or "diverges". We write "diverges to $\infty$" when the limit runs off in one direction, because it carries more information than "the limit does not exist" — but $\infty$ is never treated as a number.

Treating $\infty$ as a number is how a divergent integral acquires a value.

Reporting an approximation

A numerical answer is reported with three things: which rule, which $n$, and an . A decimal on its own is not an answer to "approximate".

Without a bound there is no way to know whether the digits you wrote down mean anything.

Trading a root for a triangle

Everything so far has removed a root by cancelling it against something the integrand already contained. Here there is nothing to cancel against.

Solvable with what we have
  • $\int x\sqrt{9-x^{2}}\,dx$ — the spare $x$ is $-\tfrac12\,du$ for $u=9-x^{2}$

  • $\int\frac{x\,dx}{x^{2}+4}$ — again the numerator is half a $du$

  • $\int\frac{dx}{x^{2}+4}$ — a standard arctangent form, no root involved

Not solvable yet
  • $\int\sqrt{9-x^{2}}\,dx$

  • $\int\frac{dx}{x^{2}\sqrt{x^{2}+9}}$

  • $\int\frac{dx}{\sqrt{x^{2}-4}}$

Push $u=9-x^{2}$ through the first one anyway. Then $du=-2x\,dx$, so $dx=-\frac{du}{2x}$ and the integral becomes $-\frac12\int\frac{\sqrt{u}}{x}\,du$ — with an $x$ still sitting in it, which is $\sqrt{9-u}$. Nothing cancelled.

Why it fails

A substitution built from the integrand can only remove a factor the integrand already has. This one has no spare $x$ to spend, so no such substitution exists. The move that works comes from outside: choose $x$ so that an identity, rather than a cancellation, kills the root.

RuleTrigonometric substitution
Conditions
  • $a>0$, and the integrand contains exactly one of the three roots below

  • $\theta$ is restricted to the range listed, which is what makes the trig factor non-negative and lets the absolute value bars go

  • the answer is converted back to $x$ before it counts as an answer

$$\boxed{\begin{aligned}&\sqrt{a^{2}-x^{2}}:\quad x=a\sin\theta,\ -\tfrac{\pi}{2}\le\theta\le\tfrac{\pi}{2}\ \Rightarrow\ \sqrt{a^{2}-x^{2}}=a\cos\theta\\&\sqrt{a^{2}+x^{2}}:\quad x=a\tan\theta,\ -\tfrac{\pi}{2}<\theta<\tfrac{\pi}{2}\ \Rightarrow\ \sqrt{a^{2}+x^{2}}=a\sec\theta\\&\sqrt{x^{2}-a^{2}}:\quad x=a\sec\theta,\ 0\le\theta<\tfrac{\pi}{2}\ \Rightarrow\ \sqrt{x^{2}-a^{2}}=a\tan\theta\end{aligned}}$$

A constant minus a square asks for sine, a constant plus a square asks for tangent, and a square with the constant taken away asks for secant. In each case the matching Pythagorean identity turns the root into $a$ times a single trigonometric function, and $dx$ travels with the substitution.

Looks like this, but is not

$\displaystyle\int\frac{x\,dx}{\sqrt{9-x^{2}}}$ carries the same root, so it looks like a job for $x=3\sin\theta$.

It is not. The numerator already carries an $x$, so $u=9-x^{2}$ finishes it in one line: $-\frac12\int u^{-1/2}\,du=-\sqrt{9-x^{2}}+C$. The trigonometric route reaches the same place three steps later. The root alone does not select the tool — the root together with the absence of a spare $x$ does.

An x² in the denominator: ∫ dx/(x²√(x²+9))

A root of a sum, and no $x$ anywhere to help. This is the tangent template.

Given
  • $\displaystyle\int\frac{dx}{x^{2}\sqrt{x^{2}+9}}$

Find

the antiderivative, written in $x$

Solution
Pick the template and convert everything
$x=3\tan\theta,\quad -\tfrac{\pi}{2}<\theta<\tfrac{\pi}{2}$

the root is $\sqrt{a^{2}+x^{2}}$ with $a=3$, so tangent is the one that fits; $a=\sqrt{9}=3$, not $9$

$dx=3\sec^{2}\theta\,d\theta$

the differential has to be converted too, or the integral is no longer the same integral

$\sqrt{x^{2}+9}=3\sqrt{\tan^{2}\theta+1}=3\sec\theta$

on this range $\sec\theta>0$, so no bars survive

Simplify the trigonometric integral
$\int\frac{3\sec^{2}\theta\,d\theta}{9\tan^{2}\theta\cdot 3\sec\theta}=\frac19\int\frac{\sec\theta}{\tan^{2}\theta}\,d\theta$

the three factors of $3$ collect into $\frac19$

$\frac{\sec\theta}{\tan^{2}\theta}=\frac{1}{\cos\theta}\cdot\frac{\cos^{2}\theta}{\sin^{2}\theta}=\frac{\cos\theta}{\sin^{2}\theta}$

writing everything in sine and cosine exposes a $du$: the numerator is the derivative of $\sin\theta$

$\frac19\int\frac{\cos\theta}{\sin^{2}\theta}\,d\theta=\frac19\int u^{-2}\,du=-\frac{1}{9\sin\theta}+C$

with $u=\sin\theta$

Go back to x with the triangle
$\tan\theta=\frac{x}{3}\ \Rightarrow\ \text{opposite }x,\ \text{adjacent }3,\ \text{hypotenuse }\sqrt{x^{2}+9}$

the substitution itself is the triangle; nothing new is being assumed

$\sin\theta=\frac{x}{\sqrt{x^{2}+9}}$

opposite over hypotenuse, read straight off the triangle

$-\frac{1}{9\sin\theta}=-\frac{\sqrt{x^{2}+9}}{9x}$

substituting that value

Answer $$-\frac{\sqrt{x^{2}+9}}{9x}+C$$
Check

Differentiate the answer instead of redoing the integral: $\frac{d}{dx}\left[\frac{\sqrt{x^{2}+9}}{x}\right]=\frac{x^{2}/\sqrt{x^{2}+9}-\sqrt{x^{2}+9}}{x^{2}}=\frac{-9}{x^{2}\sqrt{x^{2}+9}}$, so the derivative of $-\frac19$ times it is $\frac{1}{x^{2}\sqrt{x^{2}+9}}$ — the integrand.

Three conversions ($x$, $dx$, the root), one $u$-substitution inside, one triangle on the way out.

The triangle is not a memory aid; it is the substitution itself, drawn.

The integral from the opening: ∫ √(9 − x²) dx

The one that defeated us on the first line of this section.

Given
  • $\displaystyle\int\sqrt{9-x^{2}}\,dx$

Find

the antiderivative, and the area under the quarter circle as a check

Solution
Convert
$x=3\sin\theta,\quad dx=3\cos\theta\,d\theta$

a constant minus a square is the sine template, with $a=3$

$\sqrt{9-x^{2}}=3\sqrt{1-\sin^{2}\theta}=3\cos\theta$

$\cos\theta\ge0$ on $[-\pi/2,\pi/2]$, which is why the bars go

Integrate the square of a cosine
$\int 3\cos\theta\cdot 3\cos\theta\,d\theta=9\int\cos^{2}\theta\,d\theta$

the root and $dx$ each contributed a cosine

$\cos^{2}\theta=\frac{1+\cos 2\theta}{2}$

the half-angle identity is the only way to integrate an even power of cosine directly

$9\int\frac{1+\cos2\theta}{2}\,d\theta=\frac92\theta+\frac94\sin2\theta+C$

term by term

Undo the double angle and the substitution
$\frac94\sin2\theta=\frac94\cdot 2\sin\theta\cos\theta=\frac92\sin\theta\cos\theta$

$\sin 2\theta$ has to be broken up before the triangle can be used, since the triangle only knows $\theta$

$\sin\theta=\frac{x}{3},\qquad\cos\theta=\frac{\sqrt{9-x^{2}}}{3}$

read off the triangle for $x=3\sin\theta$

$\frac92\theta+\frac92\cdot\frac{x}{3}\cdot\frac{\sqrt{9-x^{2}}}{3}=\frac92\arcsin\frac{x}{3}+\frac{x\sqrt{9-x^{2}}}{2}$

and $\theta=\arcsin(x/3)$ because $\theta$ was restricted to the principal range

Answer $$\frac92\arcsin\frac{x}{3}+\frac{x\sqrt{9-x^{2}}}{2}+C$$
Check

Independent check against geometry: over $[-3,3]$ the graph is the upper half of a circle of radius $3$, so the area must be $\frac12\pi(3)^{2}=\frac{9\pi}{2}$. The antiderivative gives $\left(\frac92\cdot\frac{\pi}{2}+0\right)-\left(\frac92\cdot\left(-\frac{\pi}{2}\right)+0\right)=\frac{9\pi}{2}$. It matches.

The hook is closed: the missing $x$ in front of the root is exactly what $dx=3\cos\theta\,d\theta$ supplies.

Checkpoint
§14.1 — reading the template off the root●●○○○

Thirty seconds, no calculation. Only the choice of substitution is being asked for.

Given
  • $\displaystyle\int\frac{dx}{x^{2}\sqrt{x^{2}-25}}$

Find
  1. Which substitution does this integrand ask for?

Hint 1/4

Do not start integrating. Ask only which of the three roots this is.

Hint 2/4

A square with a constant subtracted from it, $\sqrt{x^{2}-a^{2}}$, is the secant template.

Hint 3/4

Here $a^{2}=25$, so $a=5$ and the substitution is $x=5\sec\theta$.

Hint 4/4

The substitution is $x=5\sec\theta$.

Show solution
Classify the root
$\sqrt{x^{2}-a^{2}}\ \text{with}\ a^{2}=25$

the square comes first and the constant is taken away

$a=5$

the template constant is the square root of $25$

$x=5\sec\theta$

the third line of the rule

Answer $$x=5\sec\theta$$
Check

Sanity check on the identity: $25\sec^{2}\theta-25=25\tan^{2}\theta$, a perfect square, so the root does collapse. Trying $5\sin\theta$ instead would give $25\sin^{2}\theta-25=-25\cos^{2}\theta$, a negative number under a root.

⚠ Leaving the answer in θ

the trigonometric integral is the hard part, so finishing it feels like finishing the problem

wrong$\int\sqrt{9-x^{2}}\,dx=\frac92\theta+\frac92\sin\theta\cos\theta+C$
right$\int\sqrt{9-x^{2}}\,dx=\frac92\arcsin\frac{x}{3}+\frac{x\sqrt{9-x^{2}}}{2}+C$
⚠ Substituting x but not dx

$x$ appears in several visible places and $dx$ in only one, at the end of the line

wrong$\int\frac{dx}{\sqrt{9-x^{2}}}\to\int\frac{dx}{3\cos\theta}$
right$\int\frac{dx}{\sqrt{9-x^{2}}}\to\int\frac{3\cos\theta\,d\theta}{3\cos\theta}=\theta+C$
⚠ Dropping the bars without the range

$\sqrt{a^{2}\cos^{2}\theta}=a\cos\theta$ is true on the standard range and false off it, and the range is the part nobody writes down

wrong$\sqrt{a^{2}\cos^{2}\theta}=a\cos\theta\ \text{for every }\theta$
right$\sqrt{a^{2}\cos^{2}\theta}=a\lvert\cos\theta\rvert=a\cos\theta\ \text{only for}\ -\tfrac{\pi}{2}\le\theta\le\tfrac{\pi}{2}$

When the quadratic is not a template yet

All three templates want a bare square. A quadratic with a linear term in it is one algebraic step away from being one.

MethodCompleting the square before substituting
Conditions
  • the expression under the root, or in the denominator, is a quadratic with a non-zero linear term

$$\boxed{x^{2}+bx+c=\left(x+\tfrac{b}{2}\right)^{2}+\left(c-\tfrac{b^{2}}{4}\right),\qquad u=x+\tfrac{b}{2}}$$

Halve the coefficient of $x$, square it, add and subtract it: the quadratic becomes a perfect square plus whatever is left over, and shifting the variable by that half turns it into one of the three templates.

Looks like this, but is not

$\displaystyle\int\frac{dx}{\sqrt{3-2x-x^{2}}}$ has a minus sign in front of $x^{2}$, so it looks like the secant case $\sqrt{x^{2}-a^{2}}$.

Complete the square first and the shape changes: $3-2x-x^{2}=4-(x+1)^{2}$, which is a constant minus a square, so this is the sine template with $a=2$ and $u=x+1$. Read the shape after the rewrite, never before.

A root that hides a template: ∫ dx/√(x² + 2x + 5)

There is no bare square here yet, so no template applies yet.

Given
  • $\displaystyle\int\frac{dx}{\sqrt{x^{2}+2x+5}}$

Find

the antiderivative in $x$

Solution
Make a square appear
$x^{2}+2x+5=(x+1)^{2}+4$

half of $2$ is $1$; $1^{2}=1$ is added inside the square and removed from the constant, $5-1=4$

$u=x+1,\quad du=dx$

the shift costs nothing: the differential is unchanged

Now a template fits
$\int\frac{du}{\sqrt{u^{2}+4}},\quad u=2\tan\theta,\ du=2\sec^{2}\theta\,d\theta$

a constant plus a square is the tangent template, with $a=2$

$\sqrt{u^{2}+4}=2\sec\theta$

the identity $1+\tan^{2}\theta=\sec^{2}\theta$

$\int\frac{2\sec^{2}\theta\,d\theta}{2\sec\theta}=\int\sec\theta\,d\theta=\ln\lvert\sec\theta+\tan\theta\rvert+C$

the recalled secant integral

Back to u, then to x
$\sec\theta=\frac{\sqrt{u^{2}+4}}{2},\qquad\tan\theta=\frac{u}{2}$

the triangle for $u=2\tan\theta$: opposite $u$, adjacent $2$, hypotenuse $\sqrt{u^{2}+4}$

$\ln\left\lvert\frac{\sqrt{u^{2}+4}+u}{2}\right\rvert=\ln\left\lvert\sqrt{u^{2}+4}+u\right\rvert-\ln 2$

the $-\ln 2$ is a constant, so it is absorbed into $C$

$=\ln\left\lvert x+1+\sqrt{x^{2}+2x+5}\right\rvert+C$

undoing $u=x+1$ in both places

Answer $$\ln\left\lvert x+1+\sqrt{x^{2}+2x+5}\right\rvert+C$$
Check

Differentiate: with $R=\sqrt{x^{2}+2x+5}$ the derivative is $\frac{1+(x+1)/R}{x+1+R}=\frac{(R+x+1)/R}{x+1+R}=\frac{1}{R}$, which is the integrand.

Completing the square never solves the integral; it only makes the integral belong to a family you already know.

Checkpoint
§14.2 — which template after the rewrite●●●○○

Thirty seconds. Complete the square in your head first, then classify.

Given
  • $\displaystyle\int\frac{dx}{\sqrt{3-2x-x^{2}}}$

Find
  1. After completing the square, which substitution applies?

Hint 1/4

Rewrite what is under the root before you classify it. The minus sign in front of $x^{2}$ is the whole trap.

Hint 2/4

Factor $-1$ out of the $x$ terms first: $-(x^{2}+2x)+3$, then complete the square inside the bracket.

Hint 3/4

$-(x^{2}+2x+1)+1+3=4-(x+1)^{2}$, so with $u=x+1$ the root is $\sqrt{4-u^{2}}$.

Hint 4/4

That is a constant minus a square with $a=2$, so $u=2\sin\theta$.

Show solution
Rewrite
$3-2x-x^{2}=-(x^{2}+2x)+3$

pull the minus out of the $x$ terms so the square can be completed in the usual direction

$=-\left[(x+1)^{2}-1\right]+3=4-(x+1)^{2}$

half of $2$ is $1$, and the $-1$ inside the bracket comes back out as $+1$

Classify
$\sqrt{4-u^{2}},\quad u=x+1$

a constant minus a square

$u=2\sin\theta$

the sine template with $a=2$

Answer $$u=x+1,\quad u=2\sin\theta$$
Check

Domain check, which is independent of the algebra: $3-2x-x^{2}\ge0$ holds exactly for $-3\le x\le1$, an interval of length $4$ centred at $-1$ — precisely what $4-(x+1)^{2}\ge0$ describes.

⚠ Completing the square without balancing

the square is added inside and easy to forget to remove outside

wrong$x^{2}+2x+5=(x+1)^{2}+5$
right$x^{2}+2x+5=(x+1)^{2}+4$
⚠ Forgetting the minus sign in front of x²

the rewrite is done on autopilot, in the direction practised most often

wrong$3-2x-x^{2}=(x-1)^{2}+2$
right$3-2x-x^{2}=4-(x+1)^{2}$
⚠ Shifting the variable but not the limits

in an indefinite integral the shift really is free, and the habit carries over to definite ones

wrong$\int_{0}^{2}\frac{dx}{\sqrt{x^{2}+2x+5}}=\int_{0}^{2}\frac{du}{\sqrt{u^{2}+4}}$
right$\int_{0}^{2}\frac{dx}{\sqrt{x^{2}+2x+5}}=\int_{1}^{3}\frac{du}{\sqrt{u^{2}+4}}$

Splitting a ratio of polynomials into pieces you already know

Roots are handled. The other shape that stops a substitution is a ratio of two polynomials with nothing in the numerator to spend.

RulePartial fraction template
Conditions
  • $\deg P<\deg Q$; if not, divide first and template only the remainder

  • $Q$ is factored completely over the real numbers, into linear factors and quadratics with $b^{2}-4c<0$

  • every power of every factor gets its own slot

$$\boxed{\begin{aligned}&(x-a)^{k}\ \longrightarrow\ \frac{A_{1}}{x-a}+\frac{A_{2}}{(x-a)^{2}}+\cdots+\frac{A_{k}}{(x-a)^{k}}\\&(x^{2}+bx+c)^{m}\ \longrightarrow\ \frac{B_{1}x+C_{1}}{x^{2}+bx+c}+\cdots+\frac{B_{m}x+C_{m}}{(x^{2}+bx+c)^{m}}\end{aligned}}$$

Each factor of the bottom contributes one fraction for each of its powers; a linear factor contributes a plain constant on top, an irreducible quadratic contributes a linear expression on top.

Looks like this, but is not

$\dfrac{3x+1}{x^{2}+1}$ is a ratio of polynomials with the numerator of lower degree, so it looks like a candidate for a decomposition.

There is nothing to decompose: $x^{2}+1$ has $b^{2}-4c=-4<0$, so it does not factor over the reals and the template gives back the same fraction. The move here is different — split the numerator into the part that is a multiple of $2x$ and the constant left over: $\frac{3x}{x^{2}+1}+\frac{1}{x^{2}+1}$, a logarithm plus an arctangent.

One linear factor, one irreducible quadratic

The mixed case: a plain constant over one factor, a linear numerator over the other.

Given
  • $\displaystyle\int\frac{3x+1}{(x-1)(x^{2}+1)}\,dx$

Find

the antiderivative

Solution
Write the template and clear it
$\frac{3x+1}{(x-1)(x^{2}+1)}=\frac{A}{x-1}+\frac{Bx+C}{x^{2}+1}$

$x^{2}+1$ is irreducible, so its numerator is $Bx+C$; three unknowns for a degree three denominator

$3x+1=A(x^{2}+1)+(Bx+C)(x-1)$

multiplying both sides by the whole denominator

Solve for the unknowns cheaply
$x=1:\quad 4=2A\ \Rightarrow\ A=2$

plugging a root of a linear factor kills the other term entirely, which is why roots are tried before coefficient matching

$A+B=0\ \Rightarrow\ B=-2$

matching the coefficient of $x^{2}$ on both sides

$A-C=1\ \Rightarrow\ C=1$

matching the constant terms; the coefficient of $x$ then reads $-B+C=2+1=3$, which is a free consistency check

Integrate the pieces
$\int\frac{2}{x-1}\,dx=2\ln\lvert x-1\rvert$

the linear piece is always a logarithm with bars

$\int\frac{-2x}{x^{2}+1}\,dx=-\ln(x^{2}+1)$

the numerator is exactly $-du$ for $u=x^{2}+1$; no bars, since $x^{2}+1>0$

$\int\frac{1}{x^{2}+1}\,dx=\arctan x$

the leftover constant piece is the arctangent form with $a=1$

Answer $$2\ln\lvert x-1\rvert-\ln(x^{2}+1)+\arctan x+C$$
Check

Independent numerical check at a single point. The integrand at $x=0$ is $\frac{1}{(-1)(1)}=-1$. The derivative of the answer at $x=0$ is $\frac{2}{-1}-\frac{0}{1}+\frac{1}{1}=-1$. They agree.

Three unknowns, one root plugged in, two coefficients matched, three standard antiderivatives.

Split the quadratic piece into its $2x$ part and its constant part before integrating: the first is a logarithm, the second an arctangent, and they never mix.

A repeated linear factor: three slots, not two

The factor $(x+1)^{2}$ contributes two fractions on its own.

Given
  • $\displaystyle\int\frac{x^{2}+2x+3}{(x-1)(x+1)^{2}}\,dx$

Find

the antiderivative

Solution
Count the slots
$\frac{x^{2}+2x+3}{(x-1)(x+1)^{2}}=\frac{A}{x-1}+\frac{B}{x+1}+\frac{C}{(x+1)^{2}}$

one slot for $(x-1)$ and one for each power of $(x+1)$; the denominator has degree three, so three unknowns is the right count

$x^{2}+2x+3=A(x+1)^{2}+B(x-1)(x+1)+C(x-1)$

clearing denominators

Use the two roots first
$x=1:\quad 6=4A\ \Rightarrow\ A=\tfrac32$

the root of the unrepeated factor

$x=-1:\quad 2=-2C\ \Rightarrow\ C=-1$

the repeated root reaches only the highest power slot, which is exactly why it is worth plugging in

$x^{2}:\ A+B=1\ \Rightarrow\ B=-\tfrac12$

one coefficient match finishes the middle slot; the constant terms then read $A-B-C=\tfrac32+\tfrac12+1=3$ as a check

Integrate
$\int\frac{3/2}{x-1}\,dx=\tfrac32\ln\lvert x-1\rvert$

logarithm

$\int\frac{-1/2}{x+1}\,dx=-\tfrac12\ln\lvert x+1\rvert$

logarithm again

$\int\frac{-1}{(x+1)^{2}}\,dx=\frac{1}{x+1}$

a power, not a logarithm: the squared slot integrates to a rational function

Answer $$\tfrac32\ln\lvert x-1\rvert-\tfrac12\ln\lvert x+1\rvert+\frac{1}{x+1}+C$$
Check

Point check at $x=0$: the integrand is $\frac{3}{(-1)(1)}=-3$, and the derivative of the answer is $\frac{3/2}{-1}-\frac{1/2}{1}-\frac{1}{1}=-\tfrac32-\tfrac12-1=-3$.

Only the first power of a repeated factor produces a logarithm; every higher power produces a rational function.

Checkpoint
§14.3 — counting the slots before solving●●●○○

Thirty seconds, no unknowns solved. Only the shape of the template is asked for.

Given
  • $\displaystyle\frac{2x+5}{(x+3)^{2}(x^{2}+4)}$

Find
  1. Which template is the correct one?

Hint 1/4

Count the powers, do not solve for anything. Each power of each factor gets a slot.

Hint 2/4

A linear factor to the power $k$ gives $k$ slots with constant numerators; an irreducible quadratic gives a slot with a $Bx+C$ numerator for each of its powers.

Hint 3/4

Here the factors are $(x+3)^{2}$, a linear factor squared, and $x^{2}+4$, irreducible and to the first power only.

Hint 4/4

So the template is $\frac{A}{x+3}+\frac{B}{(x+3)^{2}}+\frac{Cx+D}{x^{2}+4}$.

Show solution
Slot by slot
$(x+3)^{2}\ \longrightarrow\ \frac{A}{x+3}+\frac{B}{(x+3)^{2}}$

one slot per power of the linear factor

$x^{2}+4\ \longrightarrow\ \frac{Cx+D}{x^{2}+4}$

irreducible since $b^{2}-4c=-16<0$, and it appears to the first power only

Answer $$\frac{A}{x+3}+\frac{B}{(x+3)^{2}}+\frac{Cx+D}{x^{2}+4}$$
Check

Count check, independent of the template rules: the denominator has degree $4$, so the decomposition must carry exactly $4$ unknowns. $A,B,C,D$ is four.

⚠ A single constant over an irreducible quadratic

the linear factors all take one constant, and the pattern is applied one factor too far

wrong$\frac{2x+5}{(x+3)(x^{2}+4)}=\frac{A}{x+3}+\frac{B}{x^{2}+4}$
right$\frac{2x+5}{(x+3)(x^{2}+4)}=\frac{A}{x+3}+\frac{Bx+C}{x^{2}+4}$
⚠ Writing a template for an improper fraction

the factored denominator is the eye-catching part, and the degrees are never looked at

wrong$\frac{x^{3}+1}{x^{2}-4}=\frac{A}{x-2}+\frac{B}{x+2}$
right$\frac{x^{3}+1}{x^{2}-4}=x+\frac{A}{x-2}+\frac{B}{x+2}$
⚠ Losing the 1/a in the arctangent piece

the logarithm piece has no constant out front, so the arctangent piece is written the same way

wrong$\int\frac{dx}{x^{2}+4}=\arctan\frac{x}{2}+C$
right$\int\frac{dx}{x^{2}+4}=\frac12\arctan\frac{x}{2}+C$

Choosing the technique before you start writing

Two structural tools are now on the shelf next to substitution and parts, and the expensive mistake is no longer executing badly but choosing badly.

MethodStrategy for integration
Conditions
  • the questions are asked in this order, and the first yes wins

  • a yes does not forbid a second look: many integrands admit several routes, and the order is about cost, not legality

$$\boxed{\text{simplify}\ \to\ \text{is a }du\text{ present?}\ \to\ \text{classify the shape}\ \to\ \text{table entry}\ \to\ \text{samples}}$$

Clean the integrand up first, then look for a substitution that is already there, then classify by shape into root, ratio, product or trigonometric power, and only when all of that fails reach for a table or for numerical sampling.

Looks like this, but is not

$\displaystyle\int\frac{x\,dx}{\sqrt{x^{2}-16}}$ contains $\sqrt{x^{2}-a^{2}}$, which the classification step sends straight to $x=4\sec\theta$.

The classification step is not the first step. The question above it — is a $du$ already present? — is answered yes here, since $u=x^{2}-16$ has $du=2x\,dx$ and the numerator is $\tfrac12\,du$. One line: $\sqrt{x^{2}-16}+C$. Skipping the cheap question and landing on the expensive one is the most common way to lose time in an exam.

Triage: three integrals, three different answers to "which tool"

Nothing here is hard once the tool is chosen. The exercise is the choosing.

Given
  • $\displaystyle I_{1}=\int\frac{x\,dx}{\sqrt{x^{2}-16}}$

  • $\displaystyle I_{2}=\int\frac{x^{3}+1}{x^{2}-4}\,dx$

  • $\displaystyle I_{3}=\int e^{x^{2}}\,dx$

Find

for each one, the technique and, where possible, the antiderivative

Solution
I₁: the cheap question is answered yes
$u=x^{2}-16,\quad du=2x\,dx$

the numerator is half a $du$, so the first question in the strategy already stops the search

$\tfrac12\int u^{-1/2}\,du=u^{1/2}+C=\sqrt{x^{2}-16}+C$

the root never had to be removed at all

I₂: degrees first, then a template
$x^{3}+1=(x^{2}-4)x+(4x+1)$

the numerator degree is not lower, so long division comes before anything else

$\frac{4x+1}{(x-2)(x+2)}=\frac{A}{x-2}+\frac{B}{x+2}$

now the remainder is proper and both factors are linear

$x=2:\ 9=4A\Rightarrow A=\tfrac94;\qquad x=-2:\ -7=-4B\Rightarrow B=\tfrac74$

the two roots give both unknowns immediately

$\int\left(x+\frac{9/4}{x-2}+\frac{7/4}{x+2}\right)dx=\frac{x^{2}}{2}+\tfrac94\ln\lvert x-2\rvert+\tfrac74\ln\lvert x+2\rvert+C$

the quotient integrates on its own

I₃: the honest answer
$\int e^{x^{2}}\,dx\ \text{has no elementary antiderivative}$

no substitution, template or triangle will produce one, and this is a theorem rather than a gap in our technique

$\int_{0}^{1}e^{x^{2}}\,dx\approx S_{4}$

a definite version can still be answered, by sampling; that is the last branch of the strategy

Answer $$\sqrt{x^{2}-16}+C,\qquad \frac{x^{2}}{2}+\tfrac94\ln\lvert x-2\rvert+\tfrac74\ln\lvert x+2\rvert+C,\qquad \text{no elementary form}$$
Check

Check $I_{2}$ at $x=0$ without redoing it: the integrand is $\frac{1}{-4}=-\tfrac14$, and the derivative of the answer is $0+\frac{9/4}{-2}+\frac{7/4}{2}=-\tfrac98+\tfrac78=-\tfrac14$.

Two of the three took one line each once the right question was asked first; the third is not a failure of effort.

An integral that resists everything is not always a hard integral. Sometimes it is a definite integral in disguise, and the answer is a number rather than a formula.

Checkpoint
§14.4 — the cheapest route, not the first one recognised●●○○○

Thirty seconds. All four listed techniques would eventually work on this integrand; only one of them is cheap.

Given
  • $\displaystyle\int\frac{x\,dx}{x^{2}+9}$

Find
  1. Which technique gets there fastest?

Hint 1/4

Before classifying the shape, ask the cheaper question: is the numerator related to the derivative of the denominator?

Hint 2/4

If $u$ is the denominator and the numerator is a constant multiple of $du$, the integral is a logarithm and nothing else is needed.

Hint 3/4

Here $u=x^{2}+9$ gives $du=2x\,dx$, and the numerator $x\,dx$ is $\tfrac12\,du$.

Hint 4/4

So it is a substitution: the answer is $\tfrac12\ln(x^{2}+9)+C$.

Show solution
Answer the cheap question
$u=x^{2}+9,\ du=2x\,dx\ \Rightarrow\ x\,dx=\tfrac12\,du$

the numerator is already a multiple of $du$

$\tfrac12\int\frac{du}{u}=\tfrac12\ln\lvert u\rvert+C=\tfrac12\ln(x^{2}+9)+C$

and $x^{2}+9>0$, so the bars are unnecessary

Answer $$\tfrac12\ln(x^{2}+9)+C$$
Check

Contrast as the check: remove the $x$ and the same denominator gives $\frac13\arctan\frac{x}{3}+C$ instead. One factor changed the entire answer, which is what makes the cheap question worth asking first.

⚠ Classifying the shape before looking for a du

the new tools are the memorable ones, so the eye jumps to the root or the ratio and skips the cheap test

wrong$\int\frac{x\,dx}{\sqrt{9-x^{2}}}\ \xrightarrow{x=3\sin\theta}\ 3\int\sin\theta\,d\theta$
right$\int\frac{x\,dx}{\sqrt{9-x^{2}}}\ \xrightarrow{u=9-x^{2}}\ -\sqrt{9-x^{2}}+C$
⚠ Assuming every integrand has an elementary antiderivative

every integral met so far in the course had one, so failure feels like a personal shortcoming rather than a fact about the function

wrong$\int e^{-x^{2}}\,dx=\text{(keep trying substitutions)}$
right$\int_{0}^{1}e^{-x^{2}}\,dx\approx S_{4}=0.7469$

Reading an answer out of a table without being fooled

Some integrals are solved rather than solvable: somebody has already done the work and printed the result, and the only skill left is matching.

MethodUsing a table entry
Conditions
  • the entry is written in its own variable, usually $u$, and in its own constant, usually $a$

  • your integrand has to be brought into that exact shape by a substitution before the entry may be copied

  • the entry's own restriction — here $\lvert u\rvert

$$\boxed{\int\frac{u^{2}\,du}{\sqrt{a^{2}-u^{2}}}=-\frac{u}{2}\sqrt{a^{2}-u^{2}}+\frac{a^{2}}{2}\arcsin\frac{u}{a}+C}$$

A representative entry: an integrand of this exact shape, in whatever variable, has this antiderivative, with the same letter put everywhere the entry writes $u$ and the same constant everywhere it writes $a$.

Looks like this, but is not

A table gives $\int\frac{dx}{x^{2}-1}=\frac12\ln\left\lvert\frac{x-1}{x+1}\right\rvert+C$, while your own partial fractions gave $\frac12\ln\lvert x-1\rvert-\frac12\ln\lvert x+1\rvert+C$. Two different answers, so one of them looks wrong.

Neither is wrong: $\ln A-\ln B=\ln\frac{A}{B}$, so the two expressions are the same function. Antiderivatives are only determined up to shape and up to $+C$, and a printed answer that differs from yours is an invitation to check, not evidence of an error. Differentiating both is the fastest way to settle it.

Bending an integrand to fit an entry: ∫ x² dx / √(5 − 4x²)

The entry is in $u$ with a bare $u^{2}$ under the root; the integrand has $4x^{2}$.

Given
  • $\displaystyle\int\frac{x^{2}\,dx}{\sqrt{5-4x^{2}}}$

  • Table entry: $\displaystyle\int\frac{u^{2}\,du}{\sqrt{a^{2}-u^{2}}}=-\frac{u}{2}\sqrt{a^{2}-u^{2}}+\frac{a^{2}}{2}\arcsin\frac{u}{a}+C$

Find

the antiderivative in $x$

Solution
Force the integrand into the entry's shape
$u=2x\ \Rightarrow\ 4x^{2}=u^{2},\quad a=\sqrt5$

the entry needs the square of the new variable under the root, which fixes $u$; then $5-4x^{2}=a^{2}-u^{2}$ forces $a^{2}=5$

$du=2\,dx\ \Rightarrow\ dx=\tfrac12\,du,\qquad x^{2}=\tfrac{u^{2}}{4}$

the differential is the part most often left behind; it carries a factor of its own

$\int\frac{x^{2}\,dx}{\sqrt{5-4x^{2}}}=\int\frac{(u^{2}/4)}{\sqrt{5-u^{2}}}\cdot\frac{du}{2}=\frac18\int\frac{u^{2}\,du}{\sqrt{5-u^{2}}}$

two separate factors of the substitution collected into $\tfrac18$

Copy the entry and undo the substitution
$\frac18\left[-\frac{u}{2}\sqrt{5-u^{2}}+\frac{5}{2}\arcsin\frac{u}{\sqrt5}\right]$

the entry with $a^{2}=5$ put in every place it writes $a$

$=-\frac{u}{16}\sqrt{5-u^{2}}+\frac{5}{16}\arcsin\frac{u}{\sqrt5}$

distributing the $\tfrac18$

$=-\frac{x}{8}\sqrt{5-4x^{2}}+\frac{5}{16}\arcsin\frac{2x}{\sqrt5}+C$

replacing $u$ by $2x$; note $-\frac{2x}{16}=-\frac{x}{8}$

Answer $$-\frac{x}{8}\sqrt{5-4x^{2}}+\frac{5}{16}\arcsin\frac{2x}{\sqrt5}+C$$
Check

Differentiate the answer. The first term gives $-\frac18\sqrt{5-4x^{2}}+\frac{x^{2}}{2\sqrt{5-4x^{2}}}$ and the second gives $\frac{5}{8\sqrt{5-4x^{2}}}$; over the common denominator the constants cancel, $-\frac{5-4x^{2}}{8}+\frac{x^{2}}{2}+\frac58=x^{2}$, leaving $\frac{x^{2}}{\sqrt{5-4x^{2}}}$.

One substitution, one copied line, one back-substitution — and the check took longer than the lookup.

A table entry is a template, exactly like a partial fraction template: it is matched, not solved.

Checkpoint
§14.5 — what else has to change when the variable does●●●○○

Thirty seconds. A student rewrites an integrand to match a table entry written in $u$, and changes only the visible occurrences of the variable.

Given
  • Entry: $\displaystyle\int\frac{du}{u^{2}+a^{2}}=\frac1a\arctan\frac{u}{a}+C$

  • Integrand: $\displaystyle\int\frac{dx}{9x^{2}+4}$, rewritten with $u=3x$

Find
  1. What does the rewrite give?

Hint 1/4

The entry is fixed. The only question is what your integral turns into when $x$ is replaced by $u/3$ everywhere it appears — including in $dx$.

Hint 2/4

A substitution changes the differential too: $u=3x$ gives $du=3\,dx$, so $dx=\frac{du}{3}$.

Hint 3/4

With $u=3x$: $9x^{2}+4=u^{2}+4$, $a=2$, and $dx=\frac{du}{3}$, so the integral is $\frac13\int\frac{du}{u^{2}+2^{2}}$.

Hint 4/4

That is $\frac13\cdot\frac12\arctan\frac{u}{2}+C=\frac16\arctan\frac{3x}{2}+C$.

Show solution
Substitute completely
$u=3x,\quad du=3\,dx,\quad dx=\tfrac13\,du$

changing the variable without changing the differential changes the value of the integral

$\int\frac{dx}{9x^{2}+4}=\frac13\int\frac{du}{u^{2}+4}$

and $a=2$ because $a^{2}=4$

Copy the entry
$\frac13\cdot\frac12\arctan\frac{u}{2}+C$

the entry's own $\frac1a$ is a second, independent constant

$=\frac16\arctan\frac{3x}{2}+C$

undoing $u=3x$

Answer $$\frac16\arctan\frac{3x}{2}+C$$
Check

Differentiate: $\frac16\cdot\frac{3/2}{1+9x^{2}/4}=\frac16\cdot\frac{6}{4+9x^{2}}=\frac{1}{9x^{2}+4}$.

⚠ Copying an entry without converting the differential

the substitution is made mentally, and $dx$ is the only part of the integrand that does not look like it contains the variable

wrong$\int\frac{dx}{9x^{2}+4}=\int\frac{du}{u^{2}+4}$
right$\int\frac{dx}{9x^{2}+4}=\frac13\int\frac{du}{u^{2}+4}$
⚠ Matching the constant instead of its square root

the entry writes $a^{2}$ and the integrand shows the number, so the number gets copied into the place where $a$ belongs

wrong$\int\frac{du}{u^{2}+4}=\frac14\arctan\frac{u}{4}+C$
right$\int\frac{du}{u^{2}+4}=\frac12\arctan\frac{u}{2}+C$
⚠ Ignoring the restriction printed with the entry

the formula is the part that gets copied; the line of small print under it is not

wrong$\int_{0}^{3}\frac{x^{2}\,dx}{\sqrt{4-x^{2}}}\ \text{via the entry with}\ a=2$
right$\text{the entry needs}\ \lvert x\rvert<2,\ \text{so}\ \int_{0}^{3}\ \text{is not even defined here}$

When there is no antiderivative: sampling instead of solving

The last branch of the strategy said: sample the function. That branch is not a defeat, and it has its own rules and its own error control.

RuleThree approximation rules
Conditions
  • $\Delta x=\frac{b-a}{n}$ and $x_{i}=a+i\,\Delta x$ for $i=0,1,\ldots,n$

  • $\bar x_{i}$ is the midpoint of the $i$-th subinterval

  • $n$ must be even for $S_{n}$, because the parabolas are fitted two subintervals at a time

$$\boxed{\begin{aligned}T_{n}&=\frac{\Delta x}{2}\left[f(x_{0})+2f(x_{1})+\cdots+2f(x_{n-1})+f(x_{n})\right]\\M_{n}&=\Delta x\left[f(\bar x_{1})+f(\bar x_{2})+\cdots+f(\bar x_{n})\right]\\S_{n}&=\frac{\Delta x}{3}\left[f(x_{0})+4f(x_{1})+2f(x_{2})+\cdots+4f(x_{n-1})+f(x_{n})\right]\end{aligned}}$$

Join the sample points with straight lines and add the trapezoids; or use the height at the middle of each strip and add the rectangles; or fit a parabola through every consecutive three points, which is where the weights one, four, two, four, one come from.

Looks like this, but is not

More sample points always means a better answer, so $T_{100}$ must beat $S_{4}$.

Not necessarily. For $\int_{1}^{3}\frac{dx}{x}$ the trapezoidal error falls like $n^{-2}$ and Simpson's like $n^{-4}$: $T_{100}$ is off by about $3\times10^{-5}$ while $S_{4}$, on five function values instead of a hundred and one, is off by about $1.4\times10^{-3}$ — and $S_{20}$ already beats $T_{100}$ with a fifth of the work. Which rule you choose matters as much as how many points you spend.

rule, $n=4$valueerrorwhat the sign says

$T_{4}$

$1.1166667$

$+0.0180544$

chords lie above a curve that bends up

$M_{4}$

$1.0897547$

$-0.0088576$

the midpoint rectangle undershoots, and by about half as much

$S_{4}$

$1.1000000$

$+0.0013877$

parabolas follow the bend, so the error is an order smaller

$T_{4}$ and $S_{4}$ use exactly the same five heights — only the weights differ — and Simpson's is about thirteen times more accurate for it. That is the whole argument for learning the weights $1,4,2,4,1$.

Simpson with four strips on ∫₀¹ e^(−x²) dx, with an error bound

The standard example of an integrand with no elementary antiderivative.

Given
  • $f(x)=e^{-x^{2}}$ on $[0,1]$, $n=4$

  • you may use $\lvert f^{(4)}(x)\rvert\le 12$ on $[0,1]$

Find

the approximation $S_{4}$ and a bound on its error

Solution
Set up the sample points
$\Delta x=\frac{1-0}{4}=0.25$

four subintervals, so five points

$x_{0}=0,\ x_{1}=0.25,\ x_{2}=0.5,\ x_{3}=0.75,\ x_{4}=1$

and the weights that go with them are $1,4,2,4,1$

Assemble the sum
$S_{4}=\frac{0.25}{3}\left[f(0)+4f(0.25)+2f(0.5)+4f(0.75)+f(1)\right]$

keeping it symbolic until the last line keeps the arithmetic checkable

$f(0)=1,\ f(0.25)=e^{-1/16},\ f(0.5)=e^{-1/4},\ f(0.75)=e^{-9/16},\ f(1)=e^{-1}$

the exponents are $-x^{2}$, not $-x$; this is where most of the arithmetic errors happen

$\approx\frac{0.25}{3}\left[1+3.757652+1.557602+2.279131+0.367879\right]=\frac{0.25}{3}(8.962265)$

four decimal places kept in the intermediate values

$S_{4}\approx 0.7468554$

dividing by $12$

Bound the error
$\lvert E_{S_{4}}\rvert\le\frac{L(b-a)^{5}}{180n^{4}}$

the Simpson bound, with $L$ any number that dominates $\lvert f^{(4)}\rvert$

$\le\frac{12\cdot 1^{5}}{180\cdot 256}=\frac{12}{46080}\approx 2.6\times10^{-4}$

$b-a=1$ and $n^{4}=256$

$2.6\times10^{-4}<10^{-3}$

so every digit up to the third decimal place is trustworthy

Answer $$S_{4}\approx 0.7468554,\qquad \lvert E_{S_{4}}\rvert\le 2.6\times10^{-4}$$
Check

Independent check of the size, without recomputing the sum: $e^{-x^{2}}$ falls from $1$ to $0.3679$ across $[0,1]$ and is concave near the right end, so the integral must sit between the two crude bounds $0.3679$ and $1$, and nearer the top than the bottom. $0.747$ does. The true value is $0.7468241$, so the actual error is $3.1\times10^{-5}$ — comfortably inside the bound, as a bound should be.

A bound is a promise about the worst case, not a prediction: the real error here is eight times smaller than the guarantee.

Checkpoint
§14.6 — the one restriction Simpson's rule carries●●○○○

Thirty seconds. A calculation is set up with five subintervals and the weights are being written out.

Given
  • $\displaystyle\int_{0}^{1}e^{-x^{2}}\,dx$ with $n=5$

Find
  1. What is wrong, and what should be done?

Hint 1/4

Look at the number of subintervals against the way the parabolas are fitted.

Hint 2/4

Simpson's rule fits one parabola to every consecutive three sample points, which uses up two subintervals at a time — so $n$ has to be even.

Hint 3/4

With $n=5$ the last parabola would have only one subinterval left to sit on, and the weight pattern $1,4,2,4,\ldots,4,1$ does not close.

Hint 4/4

Use an even $n$ — $n=4$ or $n=6$ — or switch to the trapezoidal or midpoint rule, which accept any $n$.

Show solution
Count the subintervals per parabola
$\text{one parabola}\ \longleftrightarrow\ \text{three points}\ \longleftrightarrow\ \text{two subintervals}$

a parabola is determined by three points, and three consecutive sample points span two subintervals

$n=5\ \text{is odd}$

so the subintervals cannot be paired off and one is left over

$\text{use}\ n=4\ \text{or}\ n=6$

or a rule that works one subinterval at a time

Answer $$n\ \text{must be even}$$
Check

Check the weight pattern instead of the geometry: $1,4,2,4,\ldots,4,1$ must start and end with $1$ and alternate $4,2$ in between. With five subintervals the pattern would read $1,4,2,4,2,1$, ending on the wrong weight.

⚠ Running Simpson's rule with an odd n

the formula can be written down for any $n$, and nothing in the arithmetic complains

wrong$S_{5}=\frac{\Delta x}{3}\left[f_{0}+4f_{1}+2f_{2}+4f_{3}+2f_{4}+f_{5}\right]$
right$S_{4}=\frac{\Delta x}{3}\left[f_{0}+4f_{1}+2f_{2}+4f_{3}+f_{4}\right]$
⚠ Using the wrong divisor

three rules, three different front factors, and they are memorised as one blur

wrong$T_{n}=\frac{\Delta x}{3}\left[f_{0}+2f_{1}+\cdots+f_{n}\right]$
right$T_{n}=\frac{\Delta x}{2}\left[f_{0}+2f_{1}+\cdots+f_{n}\right]$
⚠ Reporting the bound as the error

the bound is the number the formula produces, so it is the number that gets written down

wrong$\int_{0}^{1}e^{-x^{2}}\,dx=0.7469\pm 2.6\times10^{-4}\ \text{(actual error)}$
right$\lvert E_{S_{4}}\rvert\le 2.6\times10^{-4},\ \text{actual error}\ 3.1\times10^{-5}$

Integrals that run to infinity

The definition of an improper integral was settled last time; what was missing was a way to compute the inner integral, and that is what the last three blocks have supplied.

MethodThe two step protocol
Conditions
  • name why the integral is improper, and at which endpoint, before anything else

  • if both endpoints are bad, split at any convenient interior point and require both halves to converge

  • a limit that runs to $\pm\infty$, or fails to settle, means divergence

$$\boxed{\int_{a}^{\infty}f=\lim_{t\to\infty}\int_{a}^{t}f,\qquad\int_{a}^{b}f=\lim_{t\to b^{-}}\int_{a}^{t}f}$$

Replace the bad endpoint with a letter, do an ordinary definite integral with the techniques of this section, and only then let the letter move to where it was not allowed to be.

Looks like this, but is not

$\displaystyle\int_{0}^{1}\frac{dx}{x^{1/2}}$ has an integrand that blows up at $0$, and $\int_{1}^{\infty}\frac{dx}{x^{1/2}}$ diverges, so this one should diverge too.

It converges, to $2$. The two $p$ conditions point in opposite directions: $\int_{1}^{\infty}x^{-p}\,dx$ converges when $p>1$, while $\int_{0}^{1}x^{-p}\,dx$ converges when $p<1$. Near infinity a large $p$ means fast decay; near zero a small $p$ means a mild blow-up. Same family, opposite danger, opposite condition.

A partial fraction that makes an infinite tail computable

The decomposition is what turns this into a limit anyone can take.

Given
  • $\displaystyle\int_{1}^{\infty}\frac{dx}{x(x+1)}$

Find

whether it converges, and its value if it does

Solution
Name the problem and cut it off
$\int_{1}^{\infty}\frac{dx}{x(x+1)}=\lim_{t\to\infty}\int_{1}^{t}\frac{dx}{x(x+1)}$

the upper endpoint is infinite; the integrand itself is perfectly well behaved on $[1,\infty)$

Decompose and integrate
$\frac{1}{x(x+1)}=\frac{1}{x}-\frac{1}{x+1}$

the template with $A=1$ and $B=-1$, found by putting $x=0$ and $x=-1$

$\int_{1}^{t}\left(\frac1x-\frac{1}{x+1}\right)dx=\Bigl[\ln\frac{x}{x+1}\Bigr]_{1}^{t}$

the two logarithms are combined before the limit is taken, which is the whole trick: separately they would each run to infinity

$=\ln\frac{t}{t+1}-\ln\frac12$

evaluating at both ends

Let t move
$\lim_{t\to\infty}\ln\frac{t}{t+1}=\ln 1=0$

the ratio tends to $1$ and the logarithm is continuous there

$\text{value}=0+\ln 2$

so the tail converges

Answer $$\ln 2$$
Check

An independent upper bound: for $x\ge1$ we have $x+1>x$, so $\frac{1}{x(x+1)}<\frac{1}{x^{2}}$, and $\int_{1}^{\infty}x^{-2}\,dx=1$. The answer must therefore be a positive number below $1$, and $\ln 2\approx0.693$ is.

One decomposition, one combination of logarithms, one limit.

Combine the logarithms before taking the limit. Two divergent pieces can hide a convergent difference, and splitting them destroys the information.

A verdict without an antiderivative

Nothing here integrates in closed form, and nothing needs to.

Given
  • $\displaystyle\int_{2}^{\infty}\frac{dx}{\sqrt{x^{3}-1}}$

Find

whether it converges, with the comparator named

Solution
Guess the comparator from the dominant term
$x^{3}-1\approx x^{3}\ \text{for large}\ x$

the constant is negligible far out, so the integrand behaves like $x^{-3/2}$

$\int_{2}^{\infty}x^{-3/2}\,dx\ \text{converges},\ p=\tfrac32>1$

the $p$ test at infinity; this is the verdict we hope to inherit

Turn the guess into an inequality
$x\ge2\ \Rightarrow\ x^{3}\ge8\ \Rightarrow\ x^{3}-1\ge x^{3}-\tfrac{x^{3}}{8}\ge\tfrac{x^{3}}{2}$

a comparison needs a genuine inequality on the whole interval, not an approximation

$0<\frac{1}{\sqrt{x^{3}-1}}\le\frac{\sqrt2}{x^{3/2}}$

taking square roots reverses nothing, since both sides are positive

$\sqrt2\int_{2}^{\infty}x^{-3/2}\,dx=\sqrt2\Bigl[-2x^{-1/2}\Bigr]_{2}^{\infty}=\sqrt2\cdot\frac{2}{\sqrt2}=2$

a finite dominating integral, so the applies

Answer $$\text{converges, and its value is at most }2$$
Check

Sanity check on the size: the integrand at $x=2$ is $1/\sqrt7\approx0.378$ and it decays faster than $x^{-3/2}$, so a total of at most $2$ is the right order. A verdict that had come out larger than the dominating integral would have been self-contradictory.

A comparison proves convergence but does not produce a value; when the question says decide, that is enough, and when it says evaluate, it is not.

Checkpoint
§14.7 — which p condition applies at which end●●○○○

Thirty seconds. The singularity is at the left endpoint, not at infinity.

Given
  • $\displaystyle\int_{0}^{1}\frac{dx}{x^{2/3}}$

Find
  1. Does it converge, and if so to what?

Hint 1/4

Ask first where the trouble is. That decides which of the two $p$ conditions you are allowed to use.

Hint 2/4

$\int_{0}^{1}x^{-p}\,dx$ converges exactly when $p<1$; the condition at infinity is the opposite one.

Hint 3/4

Here $p=\tfrac23<1$, so it converges, and $\int_{t}^{1}x^{-2/3}\,dx=\bigl[3x^{1/3}\bigr]_{t}^{1}=3-3t^{1/3}$.

Hint 4/4

Letting $t\to0^{+}$ gives $3$.

Show solution
Locate the trouble and cut it off
$x^{-2/3}\to\infty\ \text{as}\ x\to0^{+}$

the integrand, not the interval, is what is unbounded here

$\int_{0}^{1}x^{-2/3}\,dx=\lim_{t\to0^{+}}\int_{t}^{1}x^{-2/3}\,dx$

the bad endpoint is replaced by a letter

Integrate and take the limit
$\int_{t}^{1}x^{-2/3}\,dx=\Bigl[3x^{1/3}\Bigr]_{t}^{1}=3-3t^{1/3}$

the power rule with exponent $-\tfrac23+1=\tfrac13$

$\lim_{t\to0^{+}}\left(3-3t^{1/3}\right)=3$

the cube root is continuous at $0$

Answer $$3$$
Check

Cross-check with the rule rather than the calculation: $p=\tfrac23<1$, and the condition for convergence at $0$ is exactly $p<1$. The two agree, and $\int_{1}^{\infty}x^{-2/3}\,dx$ with the same $p$ diverges — same function, other end, opposite verdict.

⚠ Substituting infinity into a bracket

the bracket notation makes both ends look like numbers to be plugged in

wrong$\int_{1}^{\infty}\frac{dx}{x^{2}}=\Bigl[-\frac1x\Bigr]_{1}^{\infty}=0+1$
right$\int_{1}^{\infty}\frac{dx}{x^{2}}=\lim_{t\to\infty}\Bigl[-\frac1x\Bigr]_{1}^{t}=\lim_{t\to\infty}\left(1-\frac1t\right)=1$
⚠ Using the wrong p condition for the end you are at

one condition is memorised and the other is assumed to be the same

wrong$\int_{0}^{1}\frac{dx}{\sqrt{x}}\ \text{diverges since}\ p=\tfrac12<1$
right$\int_{0}^{1}\frac{dx}{\sqrt{x}}=2\ \text{converges, since}\ p=\tfrac12<1\ \text{is the condition at}\ 0$
⚠ Missing a singularity inside the interval

the endpoints are inspected because they are written down; the interior is not

wrong$\int_{-1}^{1}\frac{dx}{x^{2}}=\Bigl[-\frac1x\Bigr]_{-1}^{1}=-2$
right$\int_{-1}^{1}\frac{dx}{x^{2}}=\int_{-1}^{0}+\int_{0}^{1}\ \text{and both halves diverge}$
Running a trigonometric substitution

a root of a quadratic is present and the numerator has no spare factor to be absorbed into $du$

  1. Name the pattern

    Write the expression under the root as $a^{2}-x^{2}$, $a^{2}+x^{2}$ or $x^{2}-a^{2}$, completing the square first if there is a linear term. $a$ is the square root of the constant.

  2. Substitute three things

    Replace $x$, replace $dx$, and simplify the root by the matching identity. Leaving $dx$ behind is the most common single error.

  3. Do the trigonometric integral

    What is left is a trigonometric integral of the kind handled in the previous section: peel a factor, use an identity, or recognise a standard form such as $\int\sec\theta\,d\theta$.

  4. Draw the triangle

    The substitution itself gives the triangle: $x=a\sin\theta$ means opposite $x$, hypotenuse $a$. Read every trigonometric function of $\theta$ off it.

  5. Convert and check

    Rewrite the answer in $x$, add $+C$, and differentiate it once to confirm you get the integrand back.

Where it goes wrong
  • stopping while the answer still contains $\theta$

  • converting $x$ but not $dx$

  • dropping absolute value bars without naming the range of $\theta$

  • reaching for a triangle when a plain substitution was available

Integrating the four kinds of partial fraction piece

after a decomposition, when each fraction has to be turned into an antiderivative

  1. Plain linear

    $\displaystyle\int\frac{A}{x-a}\,dx=A\ln\lvert x-a\rvert+C$. Bars, always: the argument changes sign across $a$.

  2. Repeated linear

    $\displaystyle\int\frac{A}{(x-a)^{k}}\,dx=\frac{-A}{(k-1)(x-a)^{k-1}}+C$ for $k\ge2$. A power, not a logarithm.

  3. Irreducible quadratic — split the numerator

    $\displaystyle\frac{Bx+C}{x^{2}+a^{2}}$ splits into the part that is a multiple of the derivative $2x$, giving $\frac{B}{2}\ln(x^{2}+a^{2})$, and the constant left over, giving $\frac{C}{a}\arctan\frac{x}{a}$.

  4. Irreducible quadratic with a linear term

    Complete the square first, shift with $u=x+\frac{b}{2}$, then use step 3 on the shifted expression.

Where it goes wrong
  • a logarithm written for a repeated factor's higher power

  • the $\frac{1}{a}$ dropped from the arctangent

  • bars written around $x^{2}+a^{2}$, which is always positive

Deciding an improper integral

an endpoint is infinite, or the integrand blows up somewhere on the closed interval

  1. Locate the trouble

    Say out loud which endpoint is bad and why. If a blow-up sits strictly inside the interval, split there first; both halves have to converge.

  2. Try for an antiderivative

    Replace the bad endpoint by $t$ and integrate with the techniques of this section. Combine logarithms before taking the limit.

  3. If no antiderivative comes, compare

    Find $g$ with $0\le f\le g$ and a known verdict for $\int g$; a convergent $g$ drags $f$ with it. In the other direction, a divergent minorant forces divergence. The comparator is usually a $p$ integral.

  4. State the verdict properly

    Either "converges to $L$" with the number, or "diverges". A comparison gives the first word only, never the number.

Where it goes wrong
  • evaluating a bracket at $\infty$

  • using the $p>1$ condition at a singularity at $0$

  • comparing with an approximation instead of a genuine inequality

  • answering "converges to $2$" when only a bound of $2$ was proved

A sign that lets the denominator factor: ∫ dx/(x² − 4)

Minus four, so the bottom factors and the template applies.

Given
  • $\displaystyle\int\frac{dx}{x^{2}-4}$

Find

the antiderivative

Solution
Factor and decompose
$\frac{1}{(x-2)(x+2)}=\frac{A}{x-2}+\frac{B}{x+2}$

two distinct linear factors, one constant each

$x=2:\ 1=4A;\quad x=-2:\ 1=-4B$

each root kills the other term

Integrate
$\tfrac14\ln\lvert x-2\rvert-\tfrac14\ln\lvert x+2\rvert+C$

two logarithms, bars on both

Answer $$\tfrac14\ln\left\lvert\frac{x-2}{x+2}\right\rvert+C$$
Check

At $x=0$ the integrand is $-\tfrac14$; the derivative of the answer is $\tfrac14\left(\frac{1}{-2}-\frac{1}{2}\right)=-\tfrac14$.

One sign later: ∫ dx/(x² + 4)

Plus four, so nothing factors and the template has nothing to do.

Given
  • $\displaystyle\int\frac{dx}{x^{2}+4}$

Find

the antiderivative

Solution
Check reducibility first
$b^{2}-4c=0-16<0$

irreducible over the reals, so no decomposition exists

$\int\frac{dx}{x^{2}+2^{2}}=\frac12\arctan\frac{x}{2}+C$

the standard arctangent form with $a=2$

Answer $$\frac12\arctan\frac{x}{2}+C$$
Check

Differentiate: $\frac12\cdot\frac{1/2}{1+x^{2}/4}=\frac{1}{x^{2}+4}$.

The same two symbols with one sign changed: one integral is two logarithms, the other is a single arctangent, and no amount of algebra turns either into the other.

How to tell them apart

Compute $b^{2}-4c$ before writing anything. Negative means irreducible, which means arctangent and logarithm of the whole quadratic; non-negative means it factors, which means a partial fraction template.

With an x on top: ∫ x dx/√(x² − 16)

The numerator is half a $du$, so nothing else is needed.

Given
  • $\displaystyle\int\frac{x\,dx}{\sqrt{x^{2}-16}}$

Find

the antiderivative

Solution
Spot the du
$u=x^{2}-16,\quad du=2x\,dx$

the numerator is exactly $\tfrac12\,du$

$\tfrac12\int u^{-1/2}\,du=u^{1/2}+C$

power rule

Answer $$\sqrt{x^{2}-16}+C$$
Check

Differentiate: $\frac{x}{\sqrt{x^{2}-16}}$, the integrand.

Without it: ∫ dx/√(x² − 16)

The same root, no $x$ to spend, so the triangle is unavoidable.

Given
  • $\displaystyle\int\frac{dx}{\sqrt{x^{2}-16}}$

Find

the antiderivative

Solution
Substitute
$x=4\sec\theta,\quad dx=4\sec\theta\tan\theta\,d\theta$

the third template, with $a=4$

$\sqrt{x^{2}-16}=4\tan\theta$

the identity $\sec^{2}\theta-1=\tan^{2}\theta$

$\int\frac{4\sec\theta\tan\theta}{4\tan\theta}\,d\theta=\int\sec\theta\,d\theta=\ln\lvert\sec\theta+\tan\theta\rvert+C$

the recalled secant integral

Back to x
$\sec\theta=\frac{x}{4},\quad\tan\theta=\frac{\sqrt{x^{2}-16}}{4}$

read off the triangle for $x=4\sec\theta$

$\ln\left\lvert\frac{x+\sqrt{x^{2}-16}}{4}\right\rvert=\ln\lvert x+\sqrt{x^{2}-16}\rvert+C$

the $-\ln4$ is constant and joins $C$

Answer $$\ln\left\lvert x+\sqrt{x^{2}-16}\right\rvert+C$$
Check

Differentiate: with $R=\sqrt{x^{2}-16}$, $\frac{1+x/R}{x+R}=\frac{(R+x)/R}{x+R}=\frac1R$.

One factor of $x$ separates a one line substitution from a full trigonometric substitution with a triangle at the end.

How to tell them apart

Before classifying the root, ask whether the numerator is a constant multiple of the derivative of what is under it. If it is, the root never needs removing at all.

Scaffolding comes off
The common skeleton
  1. Check properness: is the numerator degree lower than the denominator degree? If not, divide.

  2. Factor the denominator completely over the real numbers.

  3. Write one slot for every power of every factor, with $Bx+C$ on top of any irreducible quadratic.

  4. Clear denominators, then plug in the roots of the linear factors before matching any coefficients.

  5. Integrate slot by slot: logarithm, power, or logarithm plus arctangent.

  6. Check by putting one convenient number into both the integrand and the derivative of your answer.

1 · fully worked

Two distinct linear factors, fully worked

Every step of the skeleton written out, with its reason.

Given
  • $\displaystyle\int\frac{5x-4}{x^{2}-x-2}\,dx$

Find

the antiderivative

Solution
Properness and factoring
$\deg(5x-4)=1<2=\deg(x^{2}-x-2)$

proper already, so no division

$x^{2}-x-2=(x-2)(x+1)$

two numbers multiplying to $-2$ and adding to $-1$

Template and unknowns
$\frac{5x-4}{(x-2)(x+1)}=\frac{A}{x-2}+\frac{B}{x+1}$

distinct linear factors, one constant each

$5x-4=A(x+1)+B(x-2)$

clearing denominators

$x=2:\ 6=3A\Rightarrow A=2$

the root of the first factor kills the $B$ term

$x=-1:\ -9=-3B\Rightarrow B=3$

and the other root kills the $A$ term

Integrate and check
$\int\left(\frac{2}{x-2}+\frac{3}{x+1}\right)dx=2\ln\lvert x-2\rvert+3\ln\lvert x+1\rvert+C$

two plain linear slots, so two logarithms

Answer $$2\ln\lvert x-2\rvert+3\ln\lvert x+1\rvert+C$$
Check

At $x=0$: the integrand is $\frac{-4}{-2}=2$, and the derivative of the answer is $\frac{2}{-2}+\frac{3}{1}=-1+3=2$.

Roots first, coefficient matching only for what the roots cannot reach.

2 · you write the reasoning

Same skeleton, easier numbers, and this time the reasons are yours to supply. Work out why each line is allowed before opening the model answers: $\displaystyle\int\frac{3\,dx}{x(x+3)}$.

  1. reasoning

    The denominator is already factored into two distinct linear pieces, and the numerator has degree $0<2$, so the template is legal with one constant per factor.

  2. reasoning

    Multiplying both sides by $x(x+3)$ clears every denominator at once and leaves an identity between polynomials, true for every $x$.

  3. reasoning

    Putting $x=0$ annihilates the $Bx$ term, so $A$ falls out on its own — cheaper than expanding and matching coefficients.

  4. reasoning

    Putting $x=-3$ annihilates the $A(x+3)$ term for the same reason, and gives $B$ in one line.

  5. reasoning

    Each slot is a plain linear factor, so each integrates to a logarithm with bars; the answer can also be written as $\ln\left\lvert\frac{x}{x+3}\right\rvert+C$.

3 · find the buried error

Harder than the last one: an irreducible quadratic joins in. Below is a student's full solution to $\displaystyle\int\frac{2x^{2}+x+4}{x(x^{2}+4)}\,dx$. The template and the unknowns are right. Two of the numbered lines are not.

the two buried errors (2)
⚠ step 4

the factor $\tfrac12$ is missing: the numerator is $\tfrac12\,du$, not $du$, since $du=2x\,dx$

the numerator is recognised as "the derivative of the denominator" and the constant that makes that statement exact is skipped

right

$\int\frac{x}{x^{2}+4}\,dx=\tfrac12\ln(x^{2}+4)+C$.

⚠ step 5

the factor $\tfrac1a=\tfrac12$ is missing from the arctangent

the argument $\frac{x}{2}$ is remembered because it is visible inside the function, while the identical constant in front is not

right

$\int\frac{dx}{x^{2}+4}=\tfrac12\arctan\frac{x}{2}+C$.

4 · the bare problem
§14.3 — no scaffolding left●●●●○

Same skeleton, no steps written out, one irreducible quadratic in the denominator. Hints are free and cost nothing.

Given
  • $\displaystyle\int\frac{4x}{(x-1)(x^{2}+1)}\,dx$

Find
  1. Evaluate the integral, and check your answer at one value of $x$.

Hint 1/4

The denominator is already factored, and the numerator degree is lower than the denominator degree, so you can go straight to a template. Decide first how many unknowns it must carry.

Hint 2/4

One slot per power: $\frac{A}{x-1}+\frac{Bx+C}{x^{2}+1}$. Clear denominators, put $x=1$ first, then match the coefficient of $x^{2}$ and the constant.

Hint 3/4

With $4x=A(x^{2}+1)+(Bx+C)(x-1)$: at $x=1$, $4=2A$ so $A=2$; the $x^{2}$ coefficients give $A+B=0$; the constants give $A-C=0$.

Hint 4/4

$A=2$, $B=-2$, $C=2$, so the integral is $2\ln\lvert x-1\rvert-\ln(x^{2}+1)+2\arctan x+C$.

Show solution
Template and unknowns
$\frac{4x}{(x-1)(x^{2}+1)}=\frac{A}{x-1}+\frac{Bx+C}{x^{2}+1}$

three unknowns for a denominator of degree three

$4x=A(x^{2}+1)+(Bx+C)(x-1)$

clearing denominators

$x=1:\ 4=2A\Rightarrow A=2$

the only real root available

$x^{2}:\ A+B=0\Rightarrow B=-2;\qquad \text{const}:\ A-C=0\Rightarrow C=2$

and the coefficient of $x$ then reads $-B+C=2+2=4$, which matches the left side

Integrate the three pieces
$\int\frac{2}{x-1}\,dx=2\ln\lvert x-1\rvert$

logarithm

$\int\frac{-2x}{x^{2}+1}\,dx=-\ln(x^{2}+1)$

the numerator is $-du$ exactly, so no extra constant appears

$\int\frac{2}{x^{2}+1}\,dx=2\arctan x$

the arctangent form with $a=1$, where $\frac1a=1$

Answer $$2\ln\lvert x-1\rvert-\ln(x^{2}+1)+2\arctan x+C$$
Check

At $x=0$ the integrand is $\frac{0}{(-1)(1)}=0$, and the derivative of the answer is $\frac{2}{-1}-0+\frac{2}{1}=0$. A check at a point where the answer is $0$ is weak on its own, so check $x=2$ too: the integrand is $\frac{8}{(1)(5)}=1.6$, and the derivative is $\frac{2}{1}-\frac{4}{5}+\frac{2}{5}=2-0.8+0.4=1.6$.

Full exam-style question

Final-style question: four parts, four different decisionsexam format

Each part stands on its own and each one is worth the same. The marks are in naming the technique and finishing the conversion, not in heroic algebra.

Given
  • (a) $\displaystyle\int\frac{dx}{x^{2}\sqrt{4-x^{2}}}$

  • (b) $\displaystyle\int\frac{x^{2}+x+1}{(x+1)(x^{2}+1)}\,dx$

  • (c) $\displaystyle\int_{3}^{\infty}\frac{dx}{x^{2}-4}$

  • (d) $\displaystyle\int_{0}^{1}e^{-x^{2}}\,dx$ by Simpson's rule with $n=2$, with an error bound, given $\lvert f^{(4)}\rvert\le12$ on $[0,1]$

Find

an antiderivative for (a) and (b), a verdict with a value for (c), an approximation with a bound for (d)

Solution
(a) A root and no spare x: the triangle
$x=2\sin\theta,\ dx=2\cos\theta\,d\theta,\ \sqrt{4-x^{2}}=2\cos\theta$

a constant minus a square, $a=2$

$\int\frac{2\cos\theta\,d\theta}{4\sin^{2}\theta\cdot 2\cos\theta}=\frac14\int\csc^{2}\theta\,d\theta=-\frac14\cot\theta+C$

the cosines cancel and what is left is a standard trigonometric antiderivative

$\cot\theta=\frac{\sqrt{4-x^{2}}}{x}\ \Rightarrow\ -\frac{\sqrt{4-x^{2}}}{4x}+C$

adjacent over opposite, read off the triangle

(b) A ratio of polynomials: the template
$\frac{x^{2}+x+1}{(x+1)(x^{2}+1)}=\frac{A}{x+1}+\frac{Bx+C}{x^{2}+1}$

proper already, one linear factor and one irreducible quadratic

$x=-1:\ 1=2A\Rightarrow A=\tfrac12;\quad x^{2}:\ A+B=1\Rightarrow B=\tfrac12;\quad \text{const}:\ A+C=1\Rightarrow C=\tfrac12$

root first, then two coefficient matches; the coefficient of $x$ reads $B+C=1$ as a check

$\tfrac12\ln\lvert x+1\rvert+\tfrac14\ln(x^{2}+1)+\tfrac12\arctan x+C$

the quadratic slot splits into its $2x$ part, giving $\tfrac12\cdot\tfrac12\ln(x^{2}+1)$, and its constant part, giving $\tfrac12\arctan x$

(c) An infinite endpoint: decompose, then take the limit
$\int_{3}^{\infty}\frac{dx}{x^{2}-4}=\lim_{t\to\infty}\int_{3}^{t}\frac{dx}{(x-2)(x+2)}$

name the improperness before computing anything

$\frac{1}{(x-2)(x+2)}=\frac{1/4}{x-2}-\frac{1/4}{x+2}$

the two roots give $A=\tfrac14$ and $B=-\tfrac14$

$\Bigl[\tfrac14\ln\frac{x-2}{x+2}\Bigr]_{3}^{t}=\tfrac14\ln\frac{t-2}{t+2}-\tfrac14\ln\frac15$

combine the logarithms before the limit, or two infinities will appear where there are none

$\lim_{t\to\infty}\tfrac14\ln\frac{t-2}{t+2}=0\ \Rightarrow\ \text{value}=\tfrac14\ln 5$

the ratio tends to $1$

(d) No antiderivative: sample it
$\Delta x=\tfrac12,\quad x_{0}=0,\ x_{1}=\tfrac12,\ x_{2}=1$

two subintervals, which is even, so Simpson's rule is available

$S_{2}=\frac{1/2}{3}\left[1+4e^{-1/4}+e^{-1}\right]\approx\frac{1}{6}(1+3.115203+0.367879)=0.7471804$

weights $1,4,1$

$\lvert E_{S_{2}}\rvert\le\frac{12\cdot1^{5}}{180\cdot 2^{4}}=\frac{12}{2880}\approx 4.2\times10^{-3}$

the Simpson bound with $n=2$; four subintervals would cut it by a factor of sixteen

Answer $$\text{(a)}\ -\frac{\sqrt{4-x^{2}}}{4x}+C;\quad\text{(b)}\ \tfrac12\ln\lvert x+1\rvert+\tfrac14\ln(x^{2}+1)+\tfrac12\arctan x+C;\quad\text{(c)}\ \tfrac14\ln5;\quad\text{(d)}\ 0.7472\pm 4.2\times10^{-3}$$
Check

Four independent checks, one per part. (a) differentiating $-\frac{\sqrt{4-x^{2}}}{4x}$ returns $\frac{1}{x^{2}\sqrt{4-x^{2}}}$. (b) at $x=0$ the integrand is $1$ and the derivative of the answer is $\tfrac12+0+\tfrac12=1$. (c) for $x\ge3$, $x^{2}-4\ge\tfrac59x^{2}$, so the value is at most $\tfrac95\int_{3}^{\infty}x^{-2}\,dx=0.6$, and $\tfrac14\ln5=0.402$ sits below it. (d) the true value is $0.7468241$, so the actual error is $3.6\times10^{-4}$, inside the bound.

Roughly twenty lines in total, and every one of the four decisions was made in the first line of its part.

Notice that (c) is a partial fraction problem wearing an improper integral's clothes. The improper part costs one extra line at the start and one at the end.

Practice

A · concept 3 questions
1§14.1 — a root does not by itself pick the tool●●○○○

One sentence of the kind that opens a quiz. Decide, and be ready to justify with a single integral rather than with a rule.

Given
  • Claim: "An integrand containing $\sqrt{x^{2}+a^{2}}$ can only be handled by a trigonometric substitution."

Find
  1. True or false, with a counterexample if false.

Hint 1/4

Try to break the claim before believing it: can you build one integral with that root that falls to something cheaper?

Hint 2/4

A plain substitution works whenever the numerator is a constant multiple of the derivative of what is under the root.

Hint 3/4

Take $\int\frac{x\,dx}{\sqrt{x^{2}+9}}$: with $u=x^{2}+9$, $du=2x\,dx$, so the integral is $\sqrt{x^{2}+9}+C$ in one line.

Hint 4/4

One counterexample settles it: the claim is false.

Show solution
Produce the counterexample
$u=x^{2}+9,\quad du=2x\,dx$

the numerator is $\tfrac12\,du$

$\tfrac12\int u^{-1/2}\,du=u^{1/2}+C=\sqrt{x^{2}+9}+C$

the root is carried along untouched

Answer $$\text{false}$$
Check

Differentiate the result: $\frac{x}{\sqrt{x^{2}+9}}$, the integrand. So the one line answer really is an answer.

What selects a trigonometric substitution is the absence of a usable numerator, not the presence of a root.

2§14.3 — which fraction is not ready for a template●●○○○

Four rational functions. Three of them can go straight to a partial fraction template; one cannot, and needs an algebraic step first.

Given
  • $\dfrac{x^{3}+2}{x^{2}-1}$, $\dfrac{2x+1}{(x-1)(x+3)}$, $\dfrac{x+5}{(x^{2}+4)^{2}}$, $\dfrac{3x^{2}+1}{(x-2)(x^{2}+9)}$

Find
  1. Which one is not ready as written?

Hint 1/4

You are comparing two numbers for each fraction, and nothing else.

Hint 2/4

The template is licensed only when the degree of the numerator is strictly lower than the degree of the denominator.

Hint 3/4

The degrees are $3$ against $2$; $1$ against $2$; $1$ against $4$; $2$ against $3$.

Hint 4/4

Only the first fails the test, so it needs long division first.

Show solution
Compare the degrees
$\frac{x^{3}+2}{x^{2}-1}:\ 3\ge 2$

improper, so a template is not yet legal

$\frac{2x+1}{(x-1)(x+3)}:\ 1<2$

proper

$\frac{x+5}{(x^{2}+4)^{2}}:\ 1<4$

proper

$\frac{3x^{2}+1}{(x-2)(x^{2}+9)}:\ 2<3$

proper

Fix the improper one
$x^{3}+2=(x^{2}-1)x+(x+2)$

one step of long division

$\frac{x^{3}+2}{x^{2}-1}=x+\frac{x+2}{x^{2}-1}$

and now the remainder is proper

Answer $$\frac{x^{3}+2}{x^{2}-1}$$
Check

Multiply the division back out: $(x^{2}-1)x+x+2=x^{3}-x+x+2=x^{3}+2$.

3§14.6 — how much a parabola can capture●●●○○

A statement about Simpson's rule that sounds too strong to be true. The error bound is the whole argument.

Given
  • Claim: "Simpson's rule with $n=4$ gives the exact value of $\int_{a}^{b}f$ whenever $f$ is a cubic polynomial."

Find
  1. True or false, with the reason.

Hint 1/4

Do not test it on an example first. Look at what controls the Simpson error.

Hint 2/4

$\lvert E_{S_{n}}\rvert\le\frac{L(b-a)^{5}}{180n^{4}}$, where $L$ bounds $\lvert f^{(4)}\rvert$ on the interval.

Hint 3/4

For a cubic, $f^{(4)}\equiv 0$, so $L=0$ is a legitimate choice and the bound becomes $0$.

Hint 4/4

An error bounded by zero is zero, so the claim is true — parabolic interpolation captures cubics exactly.

Show solution
Read the bound
$\lvert E_{S_{n}}\rvert\le\frac{L(b-a)^{5}}{180n^{4}},\quad \lvert f^{(4)}\rvert\le L$

the bound is driven by the fourth derivative, not by how wiggly the graph looks

$f(x)=px^{3}+qx^{2}+rx+s\ \Rightarrow\ f^{(4)}(x)=0$

four differentiations kill a cubic

$L=0\ \Rightarrow\ \lvert E\rvert\le 0$

so the error cannot be anything but zero

Answer $$\text{true}$$
Check

Test it on one case rather than trusting the algebra: for $f(x)=x^{3}$ on $[0,1]$, $S_{2}=\frac{1/2}{3}\left[0+4(0.125)+1\right]=\frac{1.5}{6}=0.25$, and $\int_{0}^{1}x^{3}\,dx=\tfrac14$. Exact, as promised.

The same argument shows the trapezoidal rule is exact for straight lines, since there the second derivative vanishes.

B · computation 5 questions
1§14.1 — secant template, all the way back to x●●●○○

A trigonometric substitution with the third template. The marks are split between the conversion in and the conversion out.

Given
  • $\displaystyle\int\frac{dx}{x^{2}\sqrt{x^{2}-9}}$, for $x>3$

Find
  1. (a) Name the substitution and write $dx$ and $\sqrt{x^{2}-9}$ in terms of $\theta$.

  2. (b) Reduce the integrand and integrate it.

  3. (c) Convert back to $x$ and state the antiderivative.

Hint 1/4

Classify the root first. Which of the three templates has the square first and the constant subtracted?

Hint 2/4

$\sqrt{x^{2}-a^{2}}$ goes with $x=a\sec\theta$, and then $\sec^{2}\theta-1=\tan^{2}\theta$ collapses the root.

Hint 3/4

With $a=3$: $x=3\sec\theta$, $dx=3\sec\theta\tan\theta\,d\theta$, $\sqrt{x^{2}-9}=3\tan\theta$, and $x^{2}=9\sec^{2}\theta$.

Hint 4/4

Everything collapses to $\frac19\int\cos\theta\,d\theta=\frac19\sin\theta+C=\frac{\sqrt{x^{2}-9}}{9x}+C$.

Show solution
Convert
$x=3\sec\theta,\quad 0\le\theta<\tfrac{\pi}{2}$

the root is $\sqrt{x^{2}-a^{2}}$ with $a=3$, and $x>3$ keeps $\theta$ in the first quadrant where $\tan\theta\ge0$

$dx=3\sec\theta\tan\theta\,d\theta$

the differential of the secant

$\sqrt{x^{2}-9}=3\sqrt{\sec^{2}\theta-1}=3\tan\theta$

the third Pythagorean identity

Collapse and integrate
$\int\frac{3\sec\theta\tan\theta\,d\theta}{9\sec^{2}\theta\cdot3\tan\theta}=\frac19\int\frac{d\theta}{\sec\theta}$

the tangents cancel outright and one secant survives in the denominator

$=\frac19\int\cos\theta\,d\theta=\frac19\sin\theta+C$

one over secant is cosine

Back to x
$\sec\theta=\frac{x}{3}\ \Rightarrow\ \text{adjacent }3,\ \text{hypotenuse }x,\ \text{opposite }\sqrt{x^{2}-9}$

the triangle is the substitution drawn

$\sin\theta=\frac{\sqrt{x^{2}-9}}{x}$

opposite over hypotenuse

Answer $$\frac{\sqrt{x^{2}-9}}{9x}+C$$
Check

Differentiate: $\frac19\cdot\frac{(x/\sqrt{x^{2}-9})x-\sqrt{x^{2}-9}}{x^{2}}=\frac19\cdot\frac{9}{x^{2}\sqrt{x^{2}-9}}=\frac{1}{x^{2}\sqrt{x^{2}-9}}$.

Compare with the first worked example of this section: same shape of answer, opposite sign, because there the root was a sum.

2§14.3 — a repeated factor with an empty slot●●●○○

A decomposition with three unknowns, one of which turns out to be zero. That is allowed, and it is not a sign that you made a mistake.

Given
  • $\displaystyle\int\frac{x^{2}+1}{x(x-1)^{2}}\,dx$

Find
  1. (a) Write the template and solve for the unknowns.

  2. (b) Integrate and check your answer at $x=2$.

Hint 1/4

Count the slots before solving anything: how many powers does each factor carry?

Hint 2/4

A linear factor to the power two gives two slots; the template is $\frac{A}{x}+\frac{B}{x-1}+\frac{C}{(x-1)^{2}}$.

Hint 3/4

Clearing gives $x^{2}+1=A(x-1)^{2}+Bx(x-1)+Cx$; at $x=0$, $1=A$; at $x=1$, $2=C$; matching $x^{2}$ gives $A+B=1$.

Hint 4/4

$A=1$, $B=0$, $C=2$, so the integral is $\ln\lvert x\rvert-\frac{2}{x-1}+C$.

Show solution
Template and unknowns
$\frac{x^{2}+1}{x(x-1)^{2}}=\frac{A}{x}+\frac{B}{x-1}+\frac{C}{(x-1)^{2}}$

one slot for $x$, two for the squared factor

$x^{2}+1=A(x-1)^{2}+Bx(x-1)+Cx$

clearing denominators

$x=0:\ 1=A;\qquad x=1:\ 2=C$

both roots are available, so two unknowns come free

$x^{2}:\ A+B=1\Rightarrow B=0$

the middle slot is genuinely empty; the coefficient of $x$ then reads $-2A-B+C=-2-0+2=0$, matching the left side

Integrate
$\int\frac{dx}{x}=\ln\lvert x\rvert$

the first slot

$\int\frac{2\,dx}{(x-1)^{2}}=-\frac{2}{x-1}$

the squared slot is a power, so a rational function and not a logarithm

Answer $$\ln\lvert x\rvert-\frac{2}{x-1}+C$$
Check

At $x=2$: the integrand is $\frac{5}{2\cdot1}=2.5$, and the derivative of the answer is $\frac12+\frac{2}{(2-1)^{2}}=0.5+2=2.5$.

A zero coefficient is information, not an error: it says the graph has no logarithmic behaviour at that factor's first power.

3§14.2 — complete the square, then substitute●●●○○

A root that matches none of the three templates until the quadratic is rewritten.

Given
  • $\displaystyle\int\frac{dx}{\sqrt{x^{2}-4x+13}}$

Find
  1. (a) Complete the square and name the shift.

  2. (b) Carry out the trigonometric substitution and integrate.

  3. (c) Write the antiderivative in $x$.

Hint 1/4

Nothing under that root is a template yet. Rewrite it before you classify it.

Hint 2/4

$x^{2}+bx+c=(x+\frac{b}{2})^{2}+(c-\frac{b^{2}}{4})$; here $b=-4$, so the shift is $u=x-2$.

Hint 3/4

$x^{2}-4x+13=(x-2)^{2}+9$, so with $u=x-2$ the integral is $\int\frac{du}{\sqrt{u^{2}+9}}$, a tangent template with $a=3$.

Hint 4/4

It gives $\ln\lvert u+\sqrt{u^{2}+9}\rvert+C=\ln\lvert x-2+\sqrt{x^{2}-4x+13}\rvert+C$.

Show solution
Rewrite the quadratic
$x^{2}-4x+13=(x-2)^{2}+9$

half of $-4$ is $-2$; $(-2)^{2}=4$ is taken out of the constant, $13-4=9$

$u=x-2,\quad du=dx$

a shift never changes the differential

Substitute and integrate
$u=3\tan\theta,\quad du=3\sec^{2}\theta\,d\theta,\quad \sqrt{u^{2}+9}=3\sec\theta$

a constant plus a square is the tangent template with $a=3$

$\int\frac{3\sec^{2}\theta}{3\sec\theta}\,d\theta=\int\sec\theta\,d\theta=\ln\lvert\sec\theta+\tan\theta\rvert+C$

the standard secant integral

Two conversions back
$\sec\theta=\frac{\sqrt{u^{2}+9}}{3},\quad\tan\theta=\frac{u}{3}$

from the triangle for $u=3\tan\theta$

$\ln\left\lvert\frac{u+\sqrt{u^{2}+9}}{3}\right\rvert=\ln\lvert u+\sqrt{u^{2}+9}\rvert+C$

the $-\ln3$ joins $C$

$=\ln\left\lvert x-2+\sqrt{x^{2}-4x+13}\right\rvert+C$

undoing the shift in both places

Answer $$\ln\left\lvert x-2+\sqrt{x^{2}-4x+13}\right\rvert+C$$
Check

Differentiate with $R=\sqrt{x^{2}-4x+13}$: $\frac{1+(x-2)/R}{x-2+R}=\frac{(R+x-2)/R}{x-2+R}=\frac{1}{R}$, the integrand.

Two undo steps, not one: first the trigonometric substitution, then the shift. Forgetting the second is the usual place marks are lost here.

4§14.6 — trapezoidal rule with its error bound●●●○○

An integral whose exact value you already know, used here to see how honest the error bound is.

Given
  • $\displaystyle\int_{0}^{1}\frac{dx}{1+x^{2}}$ with $n=4$

  • $f''(x)=\dfrac{6x^{2}-2}{(1+x^{2})^{3}}$, and $\lvert f''\rvert\le2$ on $[0,1]$

Find
  1. (a) Compute $T_{4}$ to seven decimal places.

  2. (b) Bound $\lvert E_{T_{4}}\rvert$.

  3. (c) Compare with the exact value and say whether the bound was honest.

Hint 1/4

Write down the five sample points and their weights before evaluating anything.

Hint 2/4

$T_{n}=\frac{\Delta x}{2}\left[f(x_{0})+2f(x_{1})+2f(x_{2})+2f(x_{3})+f(x_{4})\right]$ and $\lvert E_{T_{n}}\rvert\le\frac{K(b-a)^{3}}{12n^{2}}$.

Hint 3/4

$\Delta x=0.25$ and the heights are $f(0)=1$, $f(0.25)=0.9411765$, $f(0.5)=0.8$, $f(0.75)=0.64$, $f(1)=0.5$, with weights $1,2,2,2,1$.

Hint 4/4

$T_{4}=0.7827941$, the bound is $\frac{2\cdot1}{12\cdot16}=0.0104167$, and the exact value is $\frac{\pi}{4}=0.7853982$.

Show solution
Sample points and heights
$\Delta x=\frac{1-0}{4}=0.25$

four strips, five heights

$f(0)=1,\ f(0.25)=\tfrac{1}{1.0625}=0.9411765,\ f(0.5)=0.8,\ f(0.75)=0.64,\ f(1)=0.5$

each height is one division; keeping seven decimals stops rounding from swamping the comparison later

Assemble T₄
$T_{4}=\frac{0.25}{2}\left[1+2(0.9411765)+2(0.8)+2(0.64)+0.5\right]$

the interior weights are all $2$, the two ends are $1$

$=0.125\,(6.2623529)=0.7827941$

one multiplication

Bound and compare
$\lvert E_{T_{4}}\rvert\le\frac{K(b-a)^{3}}{12n^{2}}=\frac{2\cdot1}{12\cdot16}=0.0104167$

with $K=2$, the largest $\lvert f''\rvert$ on the interval, attained at $x=0$

$\int_{0}^{1}\frac{dx}{1+x^{2}}=\arctan1=\frac{\pi}{4}=0.7853982$

the exact value, which is available here only because the antiderivative happens to be known

$\lvert 0.7827941-0.7853982\rvert=0.0026041<0.0104167$

so the bound was honest, and about four times too pessimistic

Answer $$T_{4}=0.7827941,\quad\lvert E\rvert\le0.0104167,\quad\text{true error }0.0026041$$
Check

A second, independent check on the sign: $f(x)=\frac{1}{1+x^{2}}$ has $f''(0)=-2<0$, so the graph bends downwards over most of $[0,1]$ and chords lie below it — the trapezoidal value should therefore be an underestimate, and $0.7827941<0.7853982$ confirms it.

Multiplying $T_{4}$ by $4$ gives $3.1312$, an approximation of $\pi$ from five divisions and no calculator.

5§14.7 — an infinite tail after completing the square●●●○○

Two ideas from this section in one integral: a quadratic that has to be rewritten, and an endpoint that has to be replaced by a letter.

Given
  • $\displaystyle\int_{0}^{\infty}\frac{dx}{x^{2}+2x+2}$

Find
  1. (a) Say why the integral is improper and rewrite it as a limit.

  2. (b) Complete the square and integrate.

  3. (c) Take the limit and state the verdict.

Hint 1/4

Check both endpoints before anything else: which one is the problem, and is the integrand well behaved everywhere in between?

Hint 2/4

$\int_{0}^{\infty}f=\lim_{t\to\infty}\int_{0}^{t}f$, and after completing the square the standard arctangent form applies.

Hint 3/4

$x^{2}+2x+2=(x+1)^{2}+1$, so $\int_{0}^{t}\frac{dx}{(x+1)^{2}+1}=\arctan(t+1)-\arctan1$.

Hint 4/4

As $t\to\infty$, $\arctan(t+1)\to\frac{\pi}{2}$, so the value is $\frac{\pi}{2}-\frac{\pi}{4}=\frac{\pi}{4}$.

Show solution
Name the improperness
$x^{2}+2x+2=(x+1)^{2}+1\ge1>0$

so the integrand is continuous everywhere and the only trouble is the infinite endpoint

$\int_{0}^{\infty}=\lim_{t\to\infty}\int_{0}^{t}$

one limit, at the top end only

Integrate
$\int_{0}^{t}\frac{dx}{(x+1)^{2}+1}=\Bigl[\arctan(x+1)\Bigr]_{0}^{t}$

the shifted arctangent form, with $a=1$ so no constant appears out front

$=\arctan(t+1)-\arctan 1=\arctan(t+1)-\frac{\pi}{4}$

the lower endpoint contributes $\arctan1$, not $\arctan0$, because of the shift

Take the limit
$\lim_{t\to\infty}\arctan(t+1)=\frac{\pi}{2}$

the horizontal asymptote of the arctangent

$\frac{\pi}{2}-\frac{\pi}{4}=\frac{\pi}{4}$

so the tail converges

Answer $$\frac{\pi}{4}$$
Check

Independent bound: for $x\ge1$, $x^{2}+2x+2>x^{2}$, so the tail beyond $1$ is less than $\int_{1}^{\infty}x^{-2}\,dx=1$, and on $[0,1]$ the integrand is at most $\tfrac12$. The total is therefore below $1.5$, and $\frac{\pi}{4}\approx0.785$ sits under it.

Shifting inside an arctangent moves the endpoint values too; that is where the $\frac{\pi}{4}$ came from, not from the upper limit.

C · exam level 3 questions
1§14.4 — an odd power on top changes the route●●●●○

Exam level, and the trap is the strategy rather than the algebra: the root looks like a trigonometric substitution, and it is not.

Given
  • $\displaystyle\int\frac{x^{3}\,dx}{\sqrt{x^{2}+9}}$

Find
  1. Which antiderivative is correct?

Hint 1/4

Split the odd power: $x^{3}\,dx=x^{2}\cdot x\,dx$, and ask what $x\,dx$ is worth.

Hint 2/4

With $u=x^{2}+9$ we get $du=2x\,dx$ and, crucially, $x^{2}=u-9$: everything in the integrand can be written in $u$.

Hint 3/4

The integral becomes $\frac12\int\frac{u-9}{\sqrt u}\,du=\frac12\int\left(u^{1/2}-9u^{-1/2}\right)du$.

Hint 4/4

That is $\frac13(x^{2}+9)^{3/2}-9\sqrt{x^{2}+9}+C$.

Show solution
Spend one x on du, rewrite the rest
$u=x^{2}+9,\quad du=2x\,dx,\quad x^{2}=u-9$

an odd power always leaves one factor for $du$ and an even power behind, and an even power of $x$ is a polynomial in $u$

$\int\frac{x^{2}\cdot x\,dx}{\sqrt{x^{2}+9}}=\frac12\int\frac{u-9}{\sqrt{u}}\,du$

no trace of $x$ survives, which is the test that the substitution was complete

Integrate two powers
$\frac12\int\left(u^{1/2}-9u^{-1/2}\right)du=\frac12\left(\frac23u^{3/2}-18u^{1/2}\right)$

$\int u^{-1/2}\,du=2u^{1/2}$, so $9$ becomes $18$ — the factor of two that is easiest to lose

$=\frac13u^{3/2}-9u^{1/2}$

distributing the $\frac12$

$=\frac13(x^{2}+9)^{3/2}-9\sqrt{x^{2}+9}+C$

undoing the substitution

Answer $$\frac13(x^{2}+9)^{3/2}-9\sqrt{x^{2}+9}+C$$
Check

Differentiate: $\frac13\cdot\frac32(x^{2}+9)^{1/2}(2x)-9\cdot\frac{x}{\sqrt{x^{2}+9}}=x\sqrt{x^{2}+9}-\frac{9x}{\sqrt{x^{2}+9}}=\frac{x(x^{2}+9)-9x}{\sqrt{x^{2}+9}}=\frac{x^{3}}{\sqrt{x^{2}+9}}$.

Odd power over a root: substitute. Even power over a root: triangle. That one line decides most of the questions in this family.

2§14.3 — a definite integral through a decomposition●●●●○

Exam level. The decomposition is the work; the evaluation at the endpoints is where the arithmetic marks are.

Given
  • $\displaystyle\int_{0}^{1}\frac{2x+3}{(x+1)(x^{2}+1)}\,dx$

Find
  1. (a) Decompose the integrand.

  2. (b) Find the antiderivative.

  3. (c) Evaluate at the endpoints and give an exact answer.

Hint 1/4

Look at the two factors of the denominator and decide how many unknowns the template must carry before touching the numerator.

Hint 2/4

One linear factor and one irreducible quadratic: $\frac{A}{x+1}+\frac{Bx+C}{x^{2}+1}$, and the quadratic slot splits into a logarithm part and an arctangent part.

Hint 3/4

Clearing gives $2x+3=A(x^{2}+1)+(Bx+C)(x+1)$; at $x=-1$, $1=2A$; matching $x^{2}$ gives $A+B=0$; matching constants gives $A+C=3$.

Hint 4/4

$A=\frac12$, $B=-\frac12$, $C=\frac52$, and the antiderivative is $\frac12\ln\lvert x+1\rvert-\frac14\ln(x^{2}+1)+\frac52\arctan x$.

Show solution
Decompose
$\frac{2x+3}{(x+1)(x^{2}+1)}=\frac{A}{x+1}+\frac{Bx+C}{x^{2}+1}$

$x^{2}+1$ is irreducible, so its numerator carries two unknowns

$2x+3=A(x^{2}+1)+(Bx+C)(x+1)$

clearing denominators

$x=-1:\ 1=2A\Rightarrow A=\tfrac12$

the only real root of the denominator

$x^{2}:\ A+B=0\Rightarrow B=-\tfrac12;\quad\text{const}:\ A+C=3\Rightarrow C=\tfrac52$

and the coefficient of $x$ then reads $B+C=2$, matching the left side

Antiderivative
$\int\frac{1/2}{x+1}\,dx=\tfrac12\ln\lvert x+1\rvert$

logarithm

$\int\frac{-x/2}{x^{2}+1}\,dx=-\tfrac14\ln(x^{2}+1)$

the numerator is $-\tfrac14\,du$ for $u=x^{2}+1$; this is where the $\tfrac12$ from $du=2x\,dx$ appears

$\int\frac{5/2}{x^{2}+1}\,dx=\tfrac52\arctan x$

the arctangent form with $a=1$

Evaluate
$\Bigl[\tfrac12\ln\lvert x+1\rvert-\tfrac14\ln(x^{2}+1)+\tfrac52\arctan x\Bigr]_{0}^{1}$

all three pieces are continuous on $[0,1]$, so no splitting is needed

$x=1:\ \tfrac12\ln2-\tfrac14\ln2+\tfrac52\cdot\tfrac{\pi}{4}=\tfrac14\ln2+\tfrac{5\pi}{8}$

$\arctan1=\tfrac{\pi}{4}$

$x=0:\ 0-0+0=0$

every piece vanishes at the lower endpoint

Answer $$\tfrac14\ln 2+\tfrac{5\pi}{8}$$
Check

Independent numerical check with the other half of this section: Simpson's rule with $n=2$ on the same integral uses the heights $f(0)=3$, $f(0.5)=2.1333333$ and $f(1)=1.25$, giving $S_{2}=\frac{0.5}{3}(12.7833333)=2.1305556$. The exact answer is $2.1367822$, so the two agree to within $0.3$ percent — as they should, since the integrand is smooth and $n$ is small.

An exact answer and a numerical one are not rivals: the second is the cheapest way to catch an algebra slip in the first.

3§14.6 — how many strips does the promise need●●●●○

Exam level. The bound is run backwards: instead of computing an error, you are asked how much work buys a guarantee.

Given
  • $\displaystyle\int_{1}^{3}\frac{dx}{x}$ by the trapezoidal rule

  • $f''(x)=\dfrac{2}{x^{3}}$, so $\lvert f''\rvert\le2$ on $[1,3]$

  • required: $\lvert E_{T_{n}}\rvert<10^{-3}$

Find
  1. What is the smallest $n$ that the bound guarantees?

Hint 1/4

You are solving an inequality for $n$, not approximating anything. Write the bound first with the numbers in it.

Hint 2/4

$\lvert E_{T_{n}}\rvert\le\frac{K(b-a)^{3}}{12n^{2}}$ with $K=2$ and $b-a=2$.

Hint 3/4

$\frac{2\cdot 2^{3}}{12n^{2}}=\frac{16}{12n^{2}}=\frac{4}{3n^{2}}<10^{-3}$, so $n^{2}>\frac{4000}{3}=1333.3$.

Hint 4/4

$n>36.5$, and $n$ is a whole number, so $n=37$.

Show solution
Put the numbers into the bound
$\frac{K(b-a)^{3}}{12n^{2}}=\frac{2\cdot 8}{12n^{2}}=\frac{4}{3n^{2}}$

the interval length is cubed, which is where most slips happen: $2^{3}=8$, not $2$

Solve for n
$\frac{4}{3n^{2}}<10^{-3}\iff n^{2}>\frac{4000}{3}=1333.33$

the inequality flips direction because $n^{2}$ moves to the other side

$n>36.51$

taking the positive square root

$n=37$

the smallest whole number above it; any larger $n$ also works

Answer $$n=37$$
Check

Check the promise at $n=37$: $\frac{4}{3(37)^{2}}=\frac{4}{4107}=9.74\times10^{-4}<10^{-3}$, and at $n=36$ it is $\frac{4}{3888}=1.03\times10^{-3}$, which fails. So $37$ really is the smallest.

A guaranteed $n$ is almost always far larger than the $n$ you actually need: at $n=4$ the true error here is already $0.018$, and the bound claimed $0.083$.

D · interleaved 3 questions
1§14 — mixed practice, type not given●●●○○

From here on the type of the question is not announced. Read the integrand and decide for yourself which tool it wants.

Given
  • $\displaystyle\int_{0}^{\pi/2}\sin^{3}x\cos^{2}x\,dx$

Find
  1. Evaluate the integral.

Hint 1/4

Nothing here is a root of a quadratic or a ratio of polynomials. Ask instead which factor can be spared as a $du$.

Hint 2/4

With an odd power of sine, peel one factor of $\sin x$ for $du$ and convert the rest with $\sin^{2}x=1-\cos^{2}x$; then $u=\cos x$.

Hint 3/4

$\int_{0}^{\pi/2}(1-\cos^{2}x)\cos^{2}x\sin x\,dx$ with $u=\cos x$, $du=-\sin x\,dx$, and the limits $x=0\mapsto u=1$, $x=\frac{\pi}{2}\mapsto u=0$.

Hint 4/4

It becomes $\int_{0}^{1}(u^{2}-u^{4})\,du=\frac13-\frac15=\frac{2}{15}$.

Show solution
Choose the factor to spare
$\sin^{3}x\cos^{2}x=(1-\cos^{2}x)\cos^{2}x\cdot\sin x$

the odd power is the one to peel, because what is left is even and converts cleanly

$u=\cos x,\quad du=-\sin x\,dx$

the peeled factor is exactly $-du$

Change the limits with the variable
$x=0\Rightarrow u=1,\qquad x=\tfrac{\pi}{2}\Rightarrow u=0$

a definite integral in $u$ needs $u$ limits; converting back to $x$ is then unnecessary

$\int_{1}^{0}(1-u^{2})u^{2}(-du)=\int_{0}^{1}(u^{2}-u^{4})\,du$

the minus sign is absorbed by swapping the limits

Integrate
$\left[\frac{u^{3}}{3}-\frac{u^{5}}{5}\right]_{0}^{1}=\frac13-\frac15=\frac{2}{15}$

two powers

Answer $$\frac{2}{15}$$
Check

Size check without redoing it: on $[0,\pi/2]$ the integrand is non-negative and never exceeds $0.186$, its largest value, reached near $x\approx0.89$; so the integral is below $0.186\cdot\frac{\pi}{2}\approx0.29$. And $\frac{2}{15}=0.133$ sits comfortably under that.

Peeling works whenever one of the two powers is odd; when both are even, the half-angle identities are the only way in.

2§14 — mixed practice, type not given●●●●○

Two sections meet in this one: something has to be integrated, and then something has to be taken to a limit.

Given
  • $\displaystyle\int_{1}^{\infty}\frac{\ln x}{x^{2}}\,dx$

Find
  1. Decide whether the integral converges, and evaluate it if it does.

Hint 1/4

Deal with the two difficulties separately: first the antiderivative, then the endpoint. Which technique produces an antiderivative of a logarithm times a power?

Hint 2/4

Integration by parts with $u=\ln x$ and $dv=x^{-2}\,dx$, so $v=-\frac1x$; then the infinite endpoint is handled by $\lim_{t\to\infty}$.

Hint 3/4

$\int_{1}^{t}\frac{\ln x}{x^{2}}\,dx=\left[-\frac{\ln x}{x}\right]_{1}^{t}+\int_{1}^{t}\frac{dx}{x^{2}}=-\frac{\ln t}{t}+1-\frac1t$.

Hint 4/4

$\frac{\ln t}{t}\to0$, so the value is $1$.

Show solution
Cut off the bad endpoint
$\int_{1}^{\infty}\frac{\ln x}{x^{2}}\,dx=\lim_{t\to\infty}\int_{1}^{t}\frac{\ln x}{x^{2}}\,dx$

the integrand is continuous on $[1,\infty)$, and $\ln 1=0$, so the lower endpoint is harmless

Integrate by parts
$u=\ln x,\ dv=x^{-2}dx\ \Rightarrow\ du=\frac{dx}{x},\ v=-\frac1x$

the logarithm is chosen as $u$ because differentiating it simplifies it, while integrating it would not

$\int_{1}^{t}\frac{\ln x}{x^{2}}\,dx=\left[-\frac{\ln x}{x}\right]_{1}^{t}+\int_{1}^{t}\frac{dx}{x^{2}}$

the minus of $v$ turns the remaining integral positive

$=-\frac{\ln t}{t}+0+\left[-\frac1x\right]_{1}^{t}=-\frac{\ln t}{t}+1-\frac1t$

collecting both endpoint contributions

Take the limit
$\lim_{t\to\infty}\frac{\ln t}{t}=0$

the logarithm is beaten by any positive power of $t$

$\lim_{t\to\infty}\left(-\frac{\ln t}{t}+1-\frac1t\right)=1$

so the integral converges

Answer $$1$$
Check

Independent bound: for $x\ge e$ we have $\ln x\le x^{1/2}$, so the tail beyond $e$ is at most $\int_{e}^{\infty}x^{-3/2}\,dx=\frac{2}{\sqrt e}\approx1.21$, and since $\ln x\le1$ on $[1,e]$ the piece there is at most $\int_{1}^{e}x^{-2}\,dx=1-\frac1e\approx0.63$. A finite total below $1.9$, consistent with the value $1$.

When an improper integral needs parts, take the limit only after the boundary term and the remaining integral have both been written down.

3§14 — mixed practice, type not given●●●●○

A geometry question that turns into one of this week's integrals halfway through.

Given
  • The region under $y=\dfrac{1}{\sqrt{x^{2}+9}}$ from $x=0$ to $x=4$ is rotated about the $x$ axis.

Find
  1. (a) Set up the volume as an integral.

  2. (b) Evaluate it exactly.

Hint 1/4

Write the volume before thinking about techniques at all: what is the radius of the disc at a given $x$?

Hint 2/4

Discs perpendicular to the axis give $V=\pi\int_{a}^{b}\left[R(x)\right]^{2}dx$, and squaring the radius here removes the root entirely.

Hint 3/4

$R(x)=\frac{1}{\sqrt{x^{2}+9}}$, so $\left[R(x)\right]^{2}=\frac{1}{x^{2}+9}$ and $V=\pi\int_{0}^{4}\frac{dx}{x^{2}+9}$.

Hint 4/4

That is the arctangent form with $a=3$: $V=\frac{\pi}{3}\arctan\frac43$.

Show solution
Set up
$V=\pi\int_{0}^{4}\left[R(x)\right]^{2}dx$

the region touches the axis of rotation, so the cross sections are discs and not washers

$\left[R(x)\right]^{2}=\left(\frac{1}{\sqrt{x^{2}+9}}\right)^{2}=\frac{1}{x^{2}+9}$

the square undoes the root, which is why no triangle is needed here

Integrate
$V=\pi\int_{0}^{4}\frac{dx}{x^{2}+3^{2}}=\pi\cdot\frac13\left[\arctan\frac{x}{3}\right]_{0}^{4}$

the arctangent form; the $\frac13$ is $\frac1a$ and is not optional

$=\frac{\pi}{3}\left(\arctan\frac43-0\right)=\frac{\pi}{3}\arctan\frac43$

evaluating

Answer $$\frac{\pi}{3}\arctan\frac43$$
Check

Bracket the answer with two cylinders instead of recomputing: the radius falls from $\frac13$ at $x=0$ to $\frac15$ at $x=4$, so the volume lies between $\pi(\tfrac15)^{2}(4)=0.503$ and $\pi(\tfrac13)^{2}(4)=1.396$. The answer $0.9711$ sits between them.

Whenever a radius carries a root, square it before choosing a technique — half of these problems stop being root problems at that moment.

Mistake ledger (20 entries)
⚠ Leaving the answer in θ

the trigonometric integral is the hard part, so finishing it feels like finishing the problem

wrong$\int\sqrt{9-x^{2}}\,dx=\frac92\theta+\frac92\sin\theta\cos\theta+C$
right$\int\sqrt{9-x^{2}}\,dx=\frac92\arcsin\frac{x}{3}+\frac{x\sqrt{9-x^{2}}}{2}+C$
⚠ Substituting x but not dx

$x$ appears in several visible places and $dx$ in only one, at the end of the line

wrong$\int\frac{dx}{\sqrt{9-x^{2}}}\to\int\frac{dx}{3\cos\theta}$
right$\int\frac{dx}{\sqrt{9-x^{2}}}\to\int\frac{3\cos\theta\,d\theta}{3\cos\theta}=\theta+C$
⚠ Dropping the bars without the range

$\sqrt{a^{2}\cos^{2}\theta}=a\cos\theta$ is true on the standard range and false off it, and the range is the part nobody writes down

wrong$\sqrt{a^{2}\cos^{2}\theta}=a\cos\theta\ \text{for every }\theta$
right$\sqrt{a^{2}\cos^{2}\theta}=a\lvert\cos\theta\rvert=a\cos\theta\ \text{only for}\ -\tfrac{\pi}{2}\le\theta\le\tfrac{\pi}{2}$
⚠ Completing the square without balancing

the square is added inside and easy to forget to remove outside

wrong$x^{2}+2x+5=(x+1)^{2}+5$
right$x^{2}+2x+5=(x+1)^{2}+4$
⚠ Forgetting the minus sign in front of x²

the rewrite is done on autopilot, in the direction practised most often

wrong$3-2x-x^{2}=(x-1)^{2}+2$
right$3-2x-x^{2}=4-(x+1)^{2}$
⚠ Shifting the variable but not the limits

in an indefinite integral the shift really is free, and the habit carries over to definite ones

wrong$\int_{0}^{2}\frac{dx}{\sqrt{x^{2}+2x+5}}=\int_{0}^{2}\frac{du}{\sqrt{u^{2}+4}}$
right$\int_{0}^{2}\frac{dx}{\sqrt{x^{2}+2x+5}}=\int_{1}^{3}\frac{du}{\sqrt{u^{2}+4}}$
⚠ A single constant over an irreducible quadratic

the linear factors all take one constant, and the pattern is applied one factor too far

wrong$\frac{2x+5}{(x+3)(x^{2}+4)}=\frac{A}{x+3}+\frac{B}{x^{2}+4}$
right$\frac{2x+5}{(x+3)(x^{2}+4)}=\frac{A}{x+3}+\frac{Bx+C}{x^{2}+4}$
⚠ Writing a template for an improper fraction

the factored denominator is the eye-catching part, and the degrees are never looked at

wrong$\frac{x^{3}+1}{x^{2}-4}=\frac{A}{x-2}+\frac{B}{x+2}$
right$\frac{x^{3}+1}{x^{2}-4}=x+\frac{A}{x-2}+\frac{B}{x+2}$
⚠ Losing the 1/a in the arctangent piece

the logarithm piece has no constant out front, so the arctangent piece is written the same way

wrong$\int\frac{dx}{x^{2}+4}=\arctan\frac{x}{2}+C$
right$\int\frac{dx}{x^{2}+4}=\frac12\arctan\frac{x}{2}+C$
⚠ Classifying the shape before looking for a du

the new tools are the memorable ones, so the eye jumps to the root or the ratio and skips the cheap test

wrong$\int\frac{x\,dx}{\sqrt{9-x^{2}}}\ \xrightarrow{x=3\sin\theta}\ 3\int\sin\theta\,d\theta$
right$\int\frac{x\,dx}{\sqrt{9-x^{2}}}\ \xrightarrow{u=9-x^{2}}\ -\sqrt{9-x^{2}}+C$
⚠ Assuming every integrand has an elementary antiderivative

every integral met so far in the course had one, so failure feels like a personal shortcoming rather than a fact about the function

wrong$\int e^{-x^{2}}\,dx=\text{(keep trying substitutions)}$
right$\int_{0}^{1}e^{-x^{2}}\,dx\approx S_{4}=0.7469$
⚠ Copying an entry without converting the differential

the substitution is made mentally, and $dx$ is the only part of the integrand that does not look like it contains the variable

wrong$\int\frac{dx}{9x^{2}+4}=\int\frac{du}{u^{2}+4}$
right$\int\frac{dx}{9x^{2}+4}=\frac13\int\frac{du}{u^{2}+4}$
⚠ Matching the constant instead of its square root

the entry writes $a^{2}$ and the integrand shows the number, so the number gets copied into the place where $a$ belongs

wrong$\int\frac{du}{u^{2}+4}=\frac14\arctan\frac{u}{4}+C$
right$\int\frac{du}{u^{2}+4}=\frac12\arctan\frac{u}{2}+C$
⚠ Ignoring the restriction printed with the entry

the formula is the part that gets copied; the line of small print under it is not

wrong$\int_{0}^{3}\frac{x^{2}\,dx}{\sqrt{4-x^{2}}}\ \text{via the entry with}\ a=2$
right$\text{the entry needs}\ \lvert x\rvert<2,\ \text{so}\ \int_{0}^{3}\ \text{is not even defined here}$
⚠ Running Simpson's rule with an odd n

the formula can be written down for any $n$, and nothing in the arithmetic complains

wrong$S_{5}=\frac{\Delta x}{3}\left[f_{0}+4f_{1}+2f_{2}+4f_{3}+2f_{4}+f_{5}\right]$
right$S_{4}=\frac{\Delta x}{3}\left[f_{0}+4f_{1}+2f_{2}+4f_{3}+f_{4}\right]$
⚠ Using the wrong divisor

three rules, three different front factors, and they are memorised as one blur

wrong$T_{n}=\frac{\Delta x}{3}\left[f_{0}+2f_{1}+\cdots+f_{n}\right]$
right$T_{n}=\frac{\Delta x}{2}\left[f_{0}+2f_{1}+\cdots+f_{n}\right]$
⚠ Reporting the bound as the error

the bound is the number the formula produces, so it is the number that gets written down

wrong$\int_{0}^{1}e^{-x^{2}}\,dx=0.7469\pm 2.6\times10^{-4}\ \text{(actual error)}$
right$\lvert E_{S_{4}}\rvert\le 2.6\times10^{-4},\ \text{actual error}\ 3.1\times10^{-5}$
⚠ Substituting infinity into a bracket

the bracket notation makes both ends look like numbers to be plugged in

wrong$\int_{1}^{\infty}\frac{dx}{x^{2}}=\Bigl[-\frac1x\Bigr]_{1}^{\infty}=0+1$
right$\int_{1}^{\infty}\frac{dx}{x^{2}}=\lim_{t\to\infty}\Bigl[-\frac1x\Bigr]_{1}^{t}=\lim_{t\to\infty}\left(1-\frac1t\right)=1$
⚠ Using the wrong p condition for the end you are at

one condition is memorised and the other is assumed to be the same

wrong$\int_{0}^{1}\frac{dx}{\sqrt{x}}\ \text{diverges since}\ p=\tfrac12<1$
right$\int_{0}^{1}\frac{dx}{\sqrt{x}}=2\ \text{converges, since}\ p=\tfrac12<1\ \text{is the condition at}\ 0$
⚠ Missing a singularity inside the interval

the endpoints are inspected because they are written down; the interior is not

wrong$\int_{-1}^{1}\frac{dx}{x^{2}}=\Bigl[-\frac1x\Bigr]_{-1}^{1}=-2$
right$\int_{-1}^{1}\frac{dx}{x^{2}}=\int_{-1}^{0}+\int_{0}^{1}\ \text{and both halves diverge}$
Formula card
Trigonometric substitution
$\boxed{\begin{aligned}&\sqrt{a^{2}-x^{2}}:\quad x=a\sin\theta,\ -\tfrac{\pi}{2}\le\theta\le\tfrac{\pi}{2}\ \Rightarrow\ \sqrt{a^{2}-x^{2}}=a\cos\theta\\&\sqrt{a^{2}+x^{2}}:\quad x=a\tan\theta,\ -\tfrac{\pi}{2}<\theta<\tfrac{\pi}{2}\ \Rightarrow\ \sqrt{a^{2}+x^{2}}=a\sec\theta\\&\sqrt{x^{2}-a^{2}}:\quad x=a\sec\theta,\ 0\le\theta<\tfrac{\pi}{2}\ \Rightarrow\ \sqrt{x^{2}-a^{2}}=a\tan\theta\end{aligned}}$

$a>0$, and the integrand contains exactly one of the three roots below; $\theta$ is restricted to the range listed, which is what makes the trig factor non-negative and lets the absolute value bars go; the answer is converted back to $x$ before it counts as an answer

Completing the square before substituting
$\boxed{x^{2}+bx+c=\left(x+\tfrac{b}{2}\right)^{2}+\left(c-\tfrac{b^{2}}{4}\right),\qquad u=x+\tfrac{b}{2}}$

the expression under the root, or in the denominator, is a quadratic with a non-zero linear term

Partial fraction template
$\boxed{\begin{aligned}&(x-a)^{k}\ \longrightarrow\ \frac{A_{1}}{x-a}+\frac{A_{2}}{(x-a)^{2}}+\cdots+\frac{A_{k}}{(x-a)^{k}}\\&(x^{2}+bx+c)^{m}\ \longrightarrow\ \frac{B_{1}x+C_{1}}{x^{2}+bx+c}+\cdots+\frac{B_{m}x+C_{m}}{(x^{2}+bx+c)^{m}}\end{aligned}}$

$\deg P<\deg Q$; if not, divide first and template only the remainder; $Q$ is factored completely over the real numbers, into linear factors and quadratics with $b^{2}-4c<0$; every power of every factor gets its own slot

Strategy for integration
$\boxed{\text{simplify}\ \to\ \text{is a }du\text{ present?}\ \to\ \text{classify the shape}\ \to\ \text{table entry}\ \to\ \text{samples}}$

the questions are asked in this order, and the first yes wins; a yes does not forbid a second look: many integrands admit several routes, and the order is about cost, not legality

Using a table entry
$\boxed{\int\frac{u^{2}\,du}{\sqrt{a^{2}-u^{2}}}=-\frac{u}{2}\sqrt{a^{2}-u^{2}}+\frac{a^{2}}{2}\arcsin\frac{u}{a}+C}$

the entry is written in its own variable, usually $u$, and in its own constant, usually $a$; your integrand has to be brought into that exact shape by a substitution before the entry may be copied; the entry's own restriction — here $\lvert u\rvert

Three approximation rules
$\boxed{\begin{aligned}T_{n}&=\frac{\Delta x}{2}\left[f(x_{0})+2f(x_{1})+\cdots+2f(x_{n-1})+f(x_{n})\right]\\M_{n}&=\Delta x\left[f(\bar x_{1})+f(\bar x_{2})+\cdots+f(\bar x_{n})\right]\\S_{n}&=\frac{\Delta x}{3}\left[f(x_{0})+4f(x_{1})+2f(x_{2})+\cdots+4f(x_{n-1})+f(x_{n})\right]\end{aligned}}$

$\Delta x=\frac{b-a}{n}$ and $x_{i}=a+i\,\Delta x$ for $i=0,1,\ldots,n$; $\bar x_{i}$ is the midpoint of the $i$-th subinterval; $n$ must be even for $S_{n}$, because the parabolas are fitted two subintervals at a time

The two step protocol
$\boxed{\int_{a}^{\infty}f=\lim_{t\to\infty}\int_{a}^{t}f,\qquad\int_{a}^{b}f=\lim_{t\to b^{-}}\int_{a}^{t}f}$

name why the integral is improper, and at which endpoint, before anything else; if both endpoints are bad, split at any convenient interior point and require both halves to converge; a limit that runs to $\pm\infty$, or fails to settle, means divergence

Error bounds for the three rules
$\lvert E_{T_{n}}\rvert\le\frac{K(b-a)^{3}}{12n^{2}},\quad\lvert E_{M_{n}}\rvert\le\frac{K(b-a)^{3}}{24n^{2}},\quad\lvert E_{S_{n}}\rvert\le\frac{L(b-a)^{5}}{180n^{4}}$

$\lvert f''\rvert\le K$ and $\lvert f^{(4)}\rvert\le L$ on $[a,b]$

The two
$\int_{1}^{\infty}\frac{dx}{x^{p}}\ \text{converges}\iff p>1,\qquad\int_{0}^{1}\frac{dx}{x^{p}}\ \text{converges}\iff p<1$

the conditions are opposite because the danger sits at opposite ends

Comparison test
$0\le f\le g\ \text{on}\ [a,\infty):\quad\int_{a}^{\infty}g\ \text{converges}\Rightarrow\int_{a}^{\infty}f\ \text{converges};\quad\int_{a}^{\infty}f\ \text{diverges}\Rightarrow\int_{a}^{\infty}g\ \text{diverges}$

a genuine inequality on the whole interval, not an approximation for large $x$

The three antiderivatives every piece lands on
$\int\frac{du}{u}=\ln\lvert u\rvert,\quad\int\frac{du}{u^{2}+a^{2}}=\frac1a\arctan\frac{u}{a},\quad\int\sec\theta\,d\theta=\ln\lvert\sec\theta+\tan\theta\rvert$

each carries $+C$; the middle one carries $\frac1a$ and the first carries bars

Check yourself

Close the page and write down, from memory: the three roots and the substitution each one asks for, the template slot an irreducible quadratic gets, the one restriction Simpson's rule carries, and the two $p$ conditions with the end of the interval each belongs to.

  • Say which of the three templates $\sqrt{16-9x^{2}}$ asks for, and what $a$ is.

    c-trig-substitution

  • Rewrite $5-4x-x^{2}$ as a constant minus a square, and name the shift.

    c-complete-square

  • Write the template for $\frac{1}{(x-2)^{3}(x^{2}+1)}$ without solving for anything, and say how many unknowns it must carry.

    c-partial-fractions

  • Explain in one sentence why $\int\frac{x\,dx}{\sqrt{9-x^{2}}}$ and $\int\frac{dx}{\sqrt{9-x^{2}}}$ take different routes.

    c-strategy

  • State the two separate constants that appear when a table entry written in $u$ is applied to an integrand written in $3x$.

    c-tables

  • Write the Simpson weights for $n=6$ and say why $n=5$ is not allowed.

    c-numerical

  • Decide $\int_{0}^{1}x^{-3/2}\,dx$ and $\int_{1}^{\infty}x^{-3/2}\,dx$ in your head, and say why the answers differ.

    c-improper

Glossary (15 terms)
trigonometric substitutiontrigonometrik yerleştirme

Replacing $x$ by $a\sin\theta$, $a\tan\theta$ or $a\sec\theta$ so that a Pythagorean identity collapses a root of a quadratic to a single trigonometric term.

reference trianglereferans üçgen

The right triangle read straight off the substitution, used to turn an answer written in $\theta$ back into one written in $x$.

completing the squarekareye tamamlama

Rewriting $x^{2}+bx+c$ as $\left(x+\frac{b}{2}\right)^{2}+\left(c-\frac{b^{2}}{4}\right)$, which exposes one of the three substitution templates.

proper rational functiondüzgün rasyonel fonksiyon

A ratio of polynomials whose numerator has strictly lower degree than its denominator; only these may be given a partial fraction template.

partial fraction decompositionkısmi kesirlere ayırma

Writing a proper rational function as a sum of fractions whose denominators are powers of the linear and irreducible quadratic factors of the original denominator.

irreducible quadraticindirgenemez kuadratik

A quadratic $x^{2}+bx+c$ with $b^{2}-4c<0$, so it has no real roots and cannot be factored further over the real numbers.

integral tablosu

A printed list of antiderivatives written in a generic variable $u$ and a generic constant $a$, applied by matching an integrand to an entry through a substitution.

indirgeme formülü

A table entry that expresses an integral in terms of the same integral with a smaller exponent, applied repeatedly until a basic form is left.

sayısal integrasyon

Approximating $\int_{a}^{b}f$ by evaluating $f$ at finitely many points; the midpoint, trapezoidal and Simpson rules.

trapezoidal ruleyamuk kuralı

Joining consecutive sample points with straight lines: weights $1,2,2,\ldots,2,1$ over $\frac{\Delta x}{2}$, with error falling like $n^{-2}$.

Simpson's ruleSimpson kuralı

Fitting a parabola through each consecutive three sample points: weights $1,4,2,4,\ldots,4,1$ over $\frac{\Delta x}{3}$, valid only for even $n$, with error falling like $n^{-4}$.

error boundhata sınırı

A guaranteed ceiling on how far an approximation can be from the true value; the actual error is usually far smaller.

improper integralgenelleştirilmiş integral

A definite integral with an infinite endpoint or an unbounded integrand, defined as the limit of ordinary definite integrals.

p integralp integrali

The two anchor families $\int_{1}^{\infty}x^{-p}\,dx$, convergent exactly when $p>1$, and $\int_{0}^{1}x^{-p}\,dx$, convergent exactly when $p<1$.

comparison testkarşılaştırma testi

Inheriting a convergence verdict from a larger or smaller function whose integral is already known; it gives a verdict, never a value.

What comes next

That is the last block of the course. What is left is the exam, and the useful thing to do with these four tools now is to practise the choosing rather than the executing: shuffle problems from every section of the term together, cover the labels, and give yourself ten seconds per integral to name the technique before you write a single line.

Sources
  • James Stewart, Calculus, Metric Version, Ninth Edition — sections 7.3, 7.4, 7.5 and 7.6 The templates and rules are the ones this book states; every worked function here is a different one.
  • Course syllabus, week 14: Techniques of Integration 7.3, 7.4, 7.5, 7.6 The assessment weights quoted on the card come from the same syllabus.
  • Every numerical value on this page Recomputed independently before publication, including the approximation tables, the error bounds and the values of the definite integrals.

Spotted something missing or wrong? tell us · share your own notes or an old exam.