UK University Integration Bee Basic

Techniques Guide
A guide covering the foundations of integration technique

Vishal Gupta & Alfie Jones

Z
Contents
1 Basic Techniques 3
1.1 Odd and Even functions . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Reflection Substitutions . . . . . . . . . . . . . . . . . . . . . . . 7
1.3 Recurrence relations . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4 Wallis Product* . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Integral of Inverse . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.6 The T-Substitution. . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.7 Floor Function Integrals . . . . . . . . . . . . . . . . . . . . . . . 17
1.8 Trigonometric Identities . . . . . . . . . . . . . . . . . . . . . . . 18
1.9 Introducing constants . . . . . . . . . . . . . . . . . . . . . . . . 21
1.10 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2 Hints on problems 23

3 Solutions 25

1
Introduction
This booklet is intended to reinforce the basic techniques of integration - it’s
assumed you know the integration covered at A level - integration by substi-
tution and by parts. Despite being quite basic, these techniques can take you
a long way and form the foundation of any more advanced method. They can
be used to evaluate a lot of interesting integrals, derive results of theoretical
importance and to us are interesting in themselves - it’s really satisfying being
able to manage a problem with clever substitutions and algebraic manipulations
over using a complicated and powerful technique like contour integration.
It’s also intended as a support to the UK University Integration Bee (website
can be found here) which is run across several universities in the UK - a team
integration competition. The techniques in here are sufficient to solve a lot of
the problems but not all - later this booklet will be developed into a book. This
book will cover the major methods of integration and theory such as special
functions which will be applied to derive results of importance in other areas of
mathematics. Particularly theoretical areas which aren’t of much importance if
you just want to learn techniques for integrating will be marked with a *.
We hope that this booklet and the competition will help convince you that
these integration techniques are more than just tricks but that these integrals
are interesting problems which are also useful!

2
1 Basic Techniques
Life’s difficulties do not come in increasing order.
– Neshan Wickramasekera

1.1 Odd and Even functions

An odd function is one such that f (x) = −f (−x) and an even function is one
for which f (x) = f (−x). Examples of odd functions are x, sin(x) and tan(x)
while examples of even functions are |x|, cos(x) and x2 . The symmetry they
have make evaluating certain integrals easier for example if you were confronted
with evaluating the following integral
Z 1
x3 dx (1.1)
−1

then the usual method would be to go ahead and directly calculate it:
1 1
x4 14 (−1)4
Z
x3 dx = = − =0
−1 4 −1 4 4

Or you could notice that, because f (x) = x3 is an odd function that the two
halves of the integral - from [−1, 0] and [0, 1] cancel out

−1 1x

Figure 1: x3 on the interval [−1, 1]. The two shaded areas are the same but
opposite sign.

3
so we can immediately see that the integral is 0. In general, we have that, for
an odd function f
Z a
f (x)dx = 0 ∀a ∈ R (1.2)
−a

The counterpart for even functions isn’t quite as dramatic - the symmetry in
this case is that on the two halves the integrals are equal so instead we have the
result
Z a Z a
f (x)dx = 2 f (x)dx (1.3)
−a 0

While this doesn’t seem as useful now, it can simplify integrals a lot.
For a more extreme example of these techniques, consider the integral
Z 2 x 1 p
3
x cos + 4 − x2 dx (1.4)
−2 2 2

This integral is also well known from some meme about solving this for a wifi
password

Figure 2: The integral you’re faced with when trying to access wifi at Nanjing
University of Aeronautics and Astronautics

It looks daunting at first - maybe not even worth the effort for the wifi - but
with the previous idea, it’s not bad:
Z 2
1 2p
x p Z
3 2
I1 = x cos 4 − x dx & I2 = 4 − x2 dx
−2 2 2 −2

The first integral is the nasty part but after squinting at it for long enough, you
can see that the integrand is an odd function - both the cosine and square root

4
are even while x3 is odd so I1 = 0. Squinting at I2 , you see that the integrand
is an even function so we can simplify it as
Z 2p
I2 = 4 − x2 dx
0

Now this integral (and before this simplification) can be done with a trigonomet-
ric substitution butpa neater and more geometric way of viewing the problem is
noting that if y = 4 − x2 then x2 + y 2 = 4 i.e the equation of a circle. Those
bounds in particular represent a quarter of a circle of radius 2 so that the final
answer is just π.
Sometimes this trick can really help save time like in the 2021 Shuttle where
being able to answer the question as fast as possible was really useful:
Z 1
sin(cot−1 x) + cos(tan−1 x)
I= dx (1.5)
−1 x2 + 1

While you can work out both of the expressions on the numerator without too
much difficulty, some time can be saved by noticing that the sine term is entirely
an odd function so its contribution to the integral is 0. So you can work with
only the cosine term. To work out cos(tan−1 (x)), it’s useful to use a triangle

p
x2 + 1
x

tan−1 (x)

Figure 3: A triangle with angle tan−1 (x)

From the picture, we see that cos(tan−1 x) = √ 1

x2 +1
so that
1
1 √
Z
dx x
I= 3 = √ = 2
−1 (x2 + 1) 2 x2 + 1 −1

In the competition, the team had to pass on the answer squared minus 2 which
in this case was 0. That was then used in the bounds of the next integral - it
was from 0 to b so it actually turned out to be 0 by default!

5
For the next example let’s do something fairly general to demonstrate the im-
portance of odd and even decomposition.
The crux of the idea is realising that any function f (x) can be written as a sum
of odd and even ”parts”
1
f (x) = ((f (x) + f (−x)) + (f (x) − f (−x)))
2 | {z } | {z }
Even Odd

(if this looks weird, don’t worry the main application of it involves just subbing
−x into an integral and then adding the two different forms)
Consider now the integral
a
x2
Z
dx (1.6)
−a esin x+1

We can break it down as above, but you might wonder how we could think
of this (if it wasn’t in the section named odd and even functions...). Clearly
there’s no right answer to this but the main indicators to me are the bounds of
integration, and the fact that sin x and x2 have not much to do with each other
except for being of opposite parity
Thus we decompose the function as above to get

1 a x2 (−x)2
Z
I= + dx
2 −a esin x + 1 esin(−x) + 1
1 a x2 (e− sin x + 1) + x2 (esin x + 1)
Z
= dx
2 −a (esin x + 1)(e− sin x + 1)
1 a x2 (e− sin x + esin x + 2)
Z
= dx
2 −a 1 + esin x + e− sin x + esin x e− sin x
| {z }
1
1 a 2
Z
= x dx
2 −a

As an exercise you could consider the properties of x2 and esin x that made this
work and try to generalise it!

6
1.2 Reflection Substitutions
The previous section shows how powerful the symmetry given by an odd or even
function is but that’s a very specific situation. The reflection substitution takes
advantage of similar symmetries but around points other than the origin.
The actual identity is
Z b Z b
f (x)dx = f (a + b − x)dx (1.7)
a a

which can be verified by the substitution u = a + b − x but can be viewed

geometrically as a reflection in the line x = a+b
2 . So the odd and even function
symmetry uses a reflection substitution too - with a = −b. Another name for
this type of substitution is King’s Substitution but I feel like reflection works
better as it highlights one of the key points - we preserve the endpoints by
flipping them.
A reflection substitution which turns out to be really useful for trigonometric
functions is u = π2 − x - taking advantage of the fact that sin( π2 − x) = cos(x),
cos( π2 − x) = sin(x) and tan( π2 − x) = cot(x) which can be seen from any right
angled triangle.
To demonstrate its use, consider this integral - Putnam 1980 A3:
Z π2
dx
I= (1.8)
0 1 + tanα (x)

where α is any real number. Since α can be anything, this integral looks really
daunting but it also suggests there isn’t really much we can actually do - it
limits us quite a lot. So a reflection substitution is all we can really do - with
u = π2 − x we get
π π
tanα (u)
Z Z
2 du 2
I= = du
0 1+ 1
tanα (u) 0 1 + tanα (u)

The integral has ended up in a very similar looking form - this is what usually
happens with reflection substitutions - we get something of a similar form which
we can then manipulate alongside our original integral. In this case, if we add
the two integrals together, we get
Z π2
π
2I = dx =
0 2
π
so that I = . Not so bad for a Putnam problem! In the original problem
√ 4
α = 2.
Another classic integral with a more involved use of the reflection substitution
is the log sine integral - one of my favourite integrals (there’ll be a section on it

7
in the full book!) Consider the integral
Z π2
I= log(sin x)dx (1.9)
0
π
With the reflection substitution u = 2 − x we get
Z π
2
I= log(cos(x))dx
0

A consequence of this is
Z π
2
log(tan x)dx = 0
0

Adding these two together, we get

Z π2
2I = log(sin(x) cos(x))dx
0
Z π
2
= log(sin(2x)) − log(2)dx
0
Z π
2 π log 2
= log(sin(2x))dx −
0 2
In the integral, substitute u = 2x. Then du = 2dx and the bounds become from
0 to π so the integral becomes
1 π
Z
log(sin(u))du
2 0
Thinking about the values sin(x) takes on the interval [ π2 , π], it’s the same as
the interval [0, π] so that
Z π2 Z π
log(sin(u))du = log(sin(u))du
π
0 2

Alternatively we could also use the substitution u = π − x and use the fact that
sin(π − x) = sin(x). This means that the above becomes
π log 2
2I = I −
2
π log 2
which gives us I = − .
2
Another problem using the reflection substitution on the way is this one from
JEE Advanced 2019:
Z π2 √
cos θ
I= √ √ dθ
0 ( sin θ + cos θ)5

8
Beginning with a reflection substitution we get
Z π2 √
sin θ
I= √ √ dθ (1.10)
0 ( sin θ + cos θ)5

Adding these two together, we get

Z π Z π
2 dθ 2 dθ
2I = √ √ = √
0 ( sin θ + cos θ)4 0 cos2 θ( tan θ + 1)4

where factoring out the cos2 θ term lets us substitute u = tan θ so that du =
sec2 θdθ. The bounds then become from 0 to ∞ and we get
Z ∞
du
2I = √ 4
0 (1 + u)
√ du du
Now we can substitute t = u which gives us dt = √ = . This gives
2 u 2t
Z ∞
2t
2I = dt
0 (1 + t)4

The substitution u = x1 pairs really well with functions for which f ( x1 ) and
f (x) have a close relation. Examples are log( x1 ) = − log(x) and arctan( x1 ) =
π 1
2 − arctan(x). In view of the actual substitution made, we get dx = − u2 du
1
which works well with an integrand involving x2 +1 because it becomes

1 1 1
1 × − 2 du = − 2 du
u2 +1 u u +1

which is the same form. A good example is

Z ∞
ln(x)
I= dx (1.11)
0 x2 + 1

Substituting u = x1 we get, using the above, noting that the boundaries flip
giving a minus sign
Z ∞
ln( 1 )
Z 0
− ln(u)
− 2 u du = du = −I
∞ u + 1 0 u2 + 1
Since I = −I, we have I = 0. In fact the same argument shows that
Z 1 Z ∞
ln(x) ln(x)
2
dx = − dx
0 x +1 1 x2 + 1
Splitting intervals into parts and identifying results like the above is fairly com-
mon. Later on, we’ll tackle these individual integrals too.

9
Another good rule of thumb is if the denominator has x2 + 1 in it, it’s a good
idea to substitute x = tan θ. This is because the denominator cancels because
dx = sec2 θdθ = (x2 + 1)dθ With the integral (1.11), putting this in, we get
π
Z ∞ Z
ln x 2
dx = ln(tan x)
0 x2 + 1 0

There’s two ways worth noting todo this integral. Either use a reflection sub-
π
stitution u = π2 − x and use tan − x = tan1 x to get I = −I. Or you can
2
split the integral into a sine and cosine one:
Z π Z π Z π
2 2 2
ln(tan x) = ln(sin x) − ln(cos x)
0 0 0

Now this is 0 because these two functions take the same values on this inter-
val. While they boil down to the same thing, I think it’s worth keeping this
interpretation in mind.
Another integral which combines all of the above functions is the following
Z ∞
arctan(x)
I= 2 + 1)
(1.12)
0 x(ln(x)
1
With the substitution u = x we get
∞ π ∞
0
arctan( u1 ) − arctan(u)
Z Z Z
1 2 π du
I= 1 − 2 du = du = −I
∞ u ((− ln(u)) 2 + 1)) u 0 u(ln(u)2 + 1) 2 0 u(ln(u)2 + 1)

Therefore we have
Z ∞
π du
2I =
2 0 u(ln(u)2 + 1)
d 1
Since dx (ln(x)) =
it makes sense to substitute v = ln(u). Doing this, the
x
du
bounds change to −∞ to ∞ and dv = so we get
u
π ∞ dv
Z
∞
2I = = [arctan(v)]∞ = π
2 ∞ v2 + 1
π
so I = .
2

10
1.3 Recurrence relations
For integrals indexed by some number, usually an integer, a recursion relation
can often be a good way to reduce it to something which is easier to handle.
The first example, which will be relevant later, is
Z ∞
In = xn e−x dx (1.13)
0

To form a recursion relation, we can integrate by parts with u = xn so that

du
dx = nx
n−1 dv
and dx = e−x so that v = −e−x . Then we get
Z ∞
∞
In = − −xn e−x 0 + n xn−1 e−x dx = nIn−1

0

This yields In = n!I0 and I0 = 1 can be checked quite easily so we get In = n!.
Note that this integral is defined for non integer values too - something we’ll
think about later.
Another way to form recurrence relations is to manipulate integrals algebraically,
consider
Z 1 n
x
In = dx
0 1 +x

for integer n. Adding together In and In−1 we get

Z 1
1
In + In−1 = xn−1 dx = (1.14)
0 n

Then rearranging this,

1 1
In = − In−1 = + In−2
n n(n − 1)

Repeatedly applying this, if n is odd then it will reduce to I1 and to I0 if n is

even. We have
Z 1 Z 1 Z 1
1 x 1+x−1
I0 = dx = ln 2I1 = dx = dx = 1 − ln 2
0 1+x 0 1+x 0 1+x

where the second is an example of the ’adding 1 taking away 1’ trick which is
quite useful. With these, we get
 n
 X 1

 + ln 2 n = 2k
2k(2k − 1)


k=1
In = X n
 1

 + 1 − ln 2 n = 2k + 1
2k(2k + 1)


k=1

11
Another class of recursion formulas is with powers of trigonometric integrals -
integration by parts is used but in combination with some trigonometric iden-
tities. Consider
Z π
In = sinn (x)dx (1.15)
0

Assume n > 2 - the reason for this will become clear. Using sin2 x = 1 − cos2 x
we can rewrite the integral as
Z π
In = sinn−2 (x)(1 − cos2 (x))dx
Z0 π Z π
= sinn−2 (x)dx − sinn−2 (x) cos2 (x)dx
0 0
Z π
= In−2 − sinn−2 (x) cos(x) cos(x)dx
0

Now we can perform integration by parts on this with u = cos(x) so we have

n−1
du dv
dx = − sin(x) and dx = sin
n−2
(x) cos(x) so that v = sinn−1(x) .
Then we get
π
Z π iπ Z
h
cos(x) sinn−1 (x) 1 2
sinn−2 (x) cos(x) cos(x)dx = n−1 + sinn (x)dx
0 0 n−1 0
1
= In
n−1
This then gives us
n−1
In = In−2
n
Repeatedly applying this, we obtain if n = 2k + 1 is odd

n−1 2k 2k(2k − 2)
In = In−2 = I2k−1 = I2k−3
n 2k + 1 (2k + 1)(2k − 1)
2k(2k − 2) · · · × 4 × 2
= I1
(2k + 1)(2k − 1) · · · × 3 × 1
22k k 2 (k − 1)2 (k − 2)2 · · · × 22 × 12
= I1
(2k + 1)!
22k k!2
= I1
(2k + 1)!

We can work out that I1 = 2. Similarly for even numbers n = 2k we get

n−1 2k − 1 (2k)!
In = In−2 = I2k−2 = · · · = 2k 2 I0
n 2k 2 k!

12
We have I0 = π so in summary we get

π(2k)!


 2k 2 n = 2k
In = 222k+1
k! 2
k!

 n = 2k + 1
(2k + 1)!

1.4 Wallis Product*

We can use this to calculate the Wallis Product, an interesting product for π.
Using the previous integral (1.15), observing that since | sin x| ≤ 1, we have

sin2n+1 (x) ≤ sin2n (x) ≤ sin2n−1 (x) =⇒ I2n+1 ≤ I2n ≤ I2n−1

Dividing through and using the recurrence relation we get

I2n I2n−1 2n + 1
1≤ ≤ =
I2n+1 I2n−1 2n

By the squeeze theorem, taking n to ∞

n
I2n π Y 2k − 1 2k + 1
lim =1= ·
n→∞ I2n+1 2 2k 2k
k=1

This gives us the Wallis Product

∞
Y 2k 2k π
· = (1.16)
2k − 1 2k + 1 2
k=1

The Wallis product can be used to get another interesting result. Using the
expression derived for In directly,
2k+1 2
2 k! 2k 2 2
I2k+1 (2k+1)! 24k+1 k!4 2 k! 2
= = = · →1
I2k π(2k)! π(2k)!(2k + 1)! (2k)! π(2k + 1)
22k k!2

as k → ∞. Rearranging this, we can write it as

1 2k 1
2k
∼√
2 k πk
This identity can be used in various proofs of Stirling’s formula and can actually
be made exact by generalising the notion of a factorial!

13
1.5 Integral of Inverse
This is a neat trick where spotting it can save you a lot of time. The result is
that
Z b Z f (b)
f (x)dx + f −1 (x)dx = bf (b) − af (a) (1.17)
a f (a)

which can be demonstrated by the following diagram

Z f (b)
f −1 (x)dx
f (a)

Z b
f (x)dx
a

a bx

Figure 4: f (x) = x2 with the area under it and its inverse in red and blue.

The result follows from seeing that the two regions together form a rectangle.
Make sure that the function is injective on [a, b] before using it!
A typical integral using this identity is
Z e
I= W (x)dx
0

where W (x) is the Lambert W function, the solution to W (x)eW (x) = x i.e the
inverse of xex . Checking xex injective on this interval is a good idea which can
be seen by noticing it’s a strictly increasing function on this interval. Using
(1.17) we get, with a = 0 and b = 1
Z 1 Z e
x
xe dx + W (x)dx = 1 × e1 − 0 × e0 = e
0 0

14
The other integral is doable with integration by parts
Z 1 Z 1
1
xex dx = [xex ]0 − ex dx
0 0
=e−e+1=1

so that we get
Z e
W (x)dx = e − 1
0

1.6 The T-Substitution.

When you have complicated trig integrals often a good thing to try is the t-sub
- especially if it’s a rational function of trig functions. The t sub is also known
as the Weierstrass substitution. The goal here is to change a trig integral into
a polynomial one which can be solved much easier.
So what is the t-sub? If you have an integral involving trig functions of a variable
say, x. The t sub is setting x
t = tan
2
Why? Because we can get a nice form for dx, sin x, cos x, tan x meaning any trig
function we have can be turned into a nice (rational) function of t.
So what are these functions then? Let’s use double angle formulas to derive
them.

p
1 + t2
t
x
2
1
x
Figure 5: Right angled triangle with t = tan
2

x t x 1
We see from the diagram above that sin =√ & cos = √
2 1 + t2 2 1 + t2

15
Thus we get
x x 2t
sin x = 2 sin cos =
2 2 1 + t2

x x 1 − t2
cos x = cos2 − sin2 =
2 2 1 + t2

2t
tan x =
1 − t2

x
Further since t = tan we get
2
1 x 1 + t2
dt = 1 + tan2 dx = dx
2 2 2
Thus we have
1 − t2
Z Z
2t 2t 2
f (sin x, cos x, tan x)dx = f , , · · dt
1 + t2 1 + t2 1 − t2 1 + t2

This is an example:
Z
1
I= dx (1.18)
a + cos x
x
We now substitute t = tan
2

Z Z
1 1 2
I= dx = 1−t2
· · dt
a + cos x a+ 1+t2
1 + t2
Z
2
= dt
(a + 1) + t2 (a − 1)
Z
2 1
= q dt
a+1 1 + (t a−1 2
a+1 )
r !
2 a−1
=√ arctan t +C
a2 − 1 a+1
!
2 x r a − 1
=√ arctan tan +C
a2 − 1 2 a+1

16
1.7 Floor Function Integrals
An interesting, and fairly common class of problems involve the floor function
⌊x⌋. These tend to look fairly daunting at the start but there are some standard
tricks to decompose them into much more manageable problems. Also Note that
this section will be closely linked to the series solution section, for reasons that
will hopefully become clear.
For completeness sake the definition of the floor function of x is the greatest
integer less than or equal to x denoted by ⌊x⌋. For example ⌊2.2⌋ = 2, ⌊−2.5⌋ =
−3, ⌊π⌋ = 3.
Consider the integral
∞
x − ⌊x⌋
Z
dx (1.19)
1 x4
Z ∞ ∞ Z
X k+1
The trick here is to rewrite (· · · ) = (· · · ) as we know that when
1 k=1 k
x ∈ [k, k + 1) that ⌊x⌋ = k.
Thus we have:
∞
Z ∞ !
∞ Z k+1
x − ⌊x⌋ ⌊x⌋
Z
dx X
dx = − dx
1 x4 1 x3 k x4
k=1
∞
∞ X Z k+1 !
−1 k
= − dx
2x2 1 k x4
k=1
X ∞ k+1 !
1 k
= 0− − − − 3
2 3x k
k=1
∞
!
1 X k k+1 1
= − − 3 + 3
2 3k 3 3 (k + 1) 3 (k + 1)
k=1
1
The first two terms telescope leaving only
3
∞
!
1 1 X 1
= − +
2 3 3(k + 1)3
k=1
1 1
= − · ζ(3)
2 3

As you may be able to see the main trick here is splitting the limits of the
integral into sections such that the floor function is known. Then we usually
get some kind of series which we are able to evaluate or quote.

17
An example of this is the next integral - I would recommend trying this one
yourself before checking the solution.
Z ∞
e−x ⌊x⌋dx (1.20)
0

Again we split the integral from 0 to ∞ with a sum.

Z ∞ ∞ Z k+1
X
e−x ⌊x⌋dx = ke−x dx
0 k=0 k
∞
X k+1
−k e−x k

=
k=0
∞
X 1
= −1 · ke−k −1
e
k=0
∞
X
1 d −ky
= 1− e−1 ·
e dy
y=1
k=0

1 d 1
= 1− · (−1) · −y

e dy 1 − e
y=1
1
=
e−1

1.8 Trigonometric Identities

A problem from JEE Main 2016 (I like to have JEE Problems) is
Z 1
1
I= arctan dx (1.21)
0 x2 − x + 1
For this, there doesn’t seem to be much which can be done because a substitution
will make things nasty outside the arctan. The only thing to do is appeal to
identities. A relevant one is the tan addition formula which implies the arctan
addition formula (as long as you’re careful with the restriction of arctan to the
range [− π2 , π2 ]:

a+b
arctan a + arctan b = arctan
1 − ab
Staring at the integrand above, if we want to use this we must have a + b = 1.
The denominator is 1 − ab = 1 − x(1 − x) so with a = x and b = 1 − x we can
apply the formula to get
Z 1 Z 1
arctan(x) + arctan(1 − x)dx = 2 arctan(x)dx
0 0

The use of the addition formula is fine because arctan(x) and arctan(1 − x) are
in the range [0, π4 ] on the domain [0, 1] so their sum is in the range [0, π2 ].

18
Now to integrate this, the integral of inverse method could be used. Another
way is to use the derivative of inverse formula and integration by parts if it
works out nicely, which it does here:
Z 1 Z 1
1 x π π
arctan(x)dx = [x arctan(x)]0 − 2
dx = − [ 12 ln(x2 + 1)]10 = − ln 2
0 0 x +1 4 4
So the original integral is
Z 1
1 π
arctan 2
dx = − 2 ln 2
0 x −x+1 2

Another important thing to keep in mind is using sec x and tan x - they have a
lot of nice relations together, namely:
d d
sec2 x = tan2 x + 1, (tan x) = sec2 x, (sec x) = sec x tan x
dx dx
It can be quite useful when handling other trigonometric functions like sines
and cosines to reduce them to sec and tan. This can be demonstrated by this
integral
Z π
x sin x
I= 2
dx (1.22)
0 3 + cos x

Performing a reflection substitution and adding them together yields

Z π
sin x
2I = π 2
dx
0 3 + cos (x)

As it is, I don’t really see how to handle it with sines and cosines. However, if
you divide through by cos2 (x) then the integral becomes
Z π
sec x tan x
2I = π dx
0 3 sec2 (x) + 1
Now the substitution u = sec x works well but we need to be careful with how
we go about it as sec x isn’t defined at x = π2 . So, splitting the integral into two
and applying the substitution u = sec x to both individually we get
Z π2 Z π
sec x tan x sec x tan x
2I = π 2
dx + π dx
3 sec (x) + 1 π 3 sec2 (x) + 1
0 2
Z ∞ Z −1
du du
=π 2+1
+π 2+1
1 3u −∞ 3u
Z ∞
du
= 2π 2+1
3u
"1 √ #∞ √
arctan( 3u) π2 3
= 2π √ =
3 9
1

19
Another thing to note is the reflection substitution u = π −x can be quite useful
for trigonometric functions too as cos(π − x) = − cos(x) and sin(π − x) = sin(x)
- they stay as the same function but the sign flips for cosines.
√

3π
A more advanced substitution is u = sin x − cos x = 2 cos x + . The
4
second form is useful for seeing on what intervals it’s injective so we know when
it can be used as a substitution. It can be useful when there’s at least two of
sin x, cos x and sin(2x) in an integral. An example of its use is in the integral
Z π2 Z π
cos x 1 2 sin x + cos x
I= p dx = p dx (1.23)
0 (1 + sin(2x))2 2 0 (1 + sin(2x))2

where the second line is by a reflection substitution. Now substituting u =

sin x − cos x (it is injective on this interval), we get du
dx = cos x + sin x taking
care of the numerator. For sin(2x), notice that
u2 = sin2 x − 2 sin x cos x + cos2 x = 1 − 2 sin x cos x =⇒ sin(2x) = 1 − u2
Then
Z 1 Z 1
1 du du
I= √ = √
2 −1 (1 + 1 − u2 )2 0 (1 + 1 − u2 )2
Now it’s in a form where putting a trigonometric sub in is good, with u = sin t
the integral becomes
Z π2
cos xdx
I=
0 (1 + cos x)2
This is a rational function of cosines so a t sub will work well, turning it into
2
1 1−t · 2 1
1 − t2
Z Z
1+t2 1+t2 1 1 1
I= 2 dt = = − =
2 2 6 3

0 2 0
1+t2

Another neat trick with trigonometric functions is using Euler’s formula, eix =
cos x + i sin x. This is used a lot in the contour integration section but can be
used outside of that. Consider the integral
Z 2π
cos2n (x)dx (1.24)
0

Expanding with Euler’s formula,

Z 2π 2π 2n
eix + e−ix
Z
2n
cos xdx = dx
0 0 2
2n Z 2π
1 X 2n
= 2n e(2k−2n)ix dx
2 k 0
k=0

20
Now for n ̸= k,
2π 2π
e(2k−2n)ix
Z
(2k−2n)ix
e dx = =0
0 i(2k − 2n) 0
as 2k − 2n is an integer. For n = k, the integrand is 1 so the integral is 2π which
gives us
Z 2π
2n π 2n
cos (x)dx = 2n−1
0 2 n
which is a neat way of going about it and a lot nicer than a recurrence method.

1.9 Introducing constants

The title of this section isn’t really telling because it covers both the more
common adding a constant and subtracting it (usually 1) and the less common
multiplying by a constant and dividing it out which isn’t often useful. These
techniques appear all over the place just like the reflection substitution so there
will be lots of examples throughout the book, generally combined with other
techniques. A direct example of the technique is
Z 2 Z 2 Z
t + 7t + 8 t + 5t + 3 2t + 5
I= 2
dt = 2
dt + 2
= t + ln(t2 + 5t + 3) + C
t + 5t + 3 t + 5t + 3 t + 5t + 3
Another one which helps us save a lot of work is
√ tan x2
Z Z
5 cos x + 7 cos x + 2 3
dx = 5 dx − dx = 5x − 2 3 arctan √ +C
2 + cos x 2 + cos x 2 + cos x 3
using the above result. Doing it directly with a t sub would be quite a lot messier
- I always try to avoid having trig terms in the numerator where possible when
using t subs.
For an example of multiplying by a constant, a really good example is Problem
21 from the 2021 Integration Bee Round 1 (by coincidence, 2020 and 2021’s
Problem 21 has been my favourite!)
Z a
x
I= dx (1.25)
0 cos(x) cos(a − x)
The first thing to do is the usual reflection substitution,
Z a
1
2I = a dx
0 cos(x) cos(a − x)

This integral can be done by using product to sum on the numerator but instead
a really clever way is to multiply and divide by the constant sin a
Z a Z a
a sin a sin x cos(a − x) + sin(a − x) cos x
2I = dx = dx
sin a 0 cos x cos(a − x) 0 cos x cos(a − x)
Z a Z a
a
= tan xdx + tan(a − x)dx
sin a 0 0

21
These two integrals are the same by the reflection substitution and
Z Z
sin x
tan xdx = dx = − ln | cos x| = ln | sec x|
cos x
So going back to the integral and dividing by two, we get

a ln | sec a|
I=
sin a
which is really nice!

1.10 Exercises
Z π
2
1. log(cos 2x + 1)dx
−π
2
Z 2π
2. sin2n xdx
0
Z 2π
cos x
3. dx
0 2 + cos x
Z π
2 cos x
4. dx
0 2 − sin(2x)
Z ∞
arctan(x)
5. dx.
0 1 + x2
Z 1
ln(1 + x)
6. dx
0 1 + x2
Z ∞
ln x
7. dx
0 x2 + 2x + 2
Z ∞
ln x
8.
0 x2 + a2
Z 1r
1
9. − 1dx
0 x
Z ∞ Z 1
r
dx 1
10. Calculate and use this to calculate − 1dx.
0 x4 + 1 0 x

22
2 Hints on problems
Problem 1
Hint 1: The most similar integral to this so far is the log sine integral (1.9).
Maybe it can be manipulated to turn into that or evaluated in a similar way.
Hint 2: Use the double angle formula

Problem 2
Hint 1: This looks quite similar to (1.24)!

Problem 3
Hint 1: This is almost exactly the same as (1.18) - you could use the same
technique or you could manipulate the integrand and possibly reuse that result!

Problem 4
Hint 1: It has a cos(x) and a sin(2x) together - there’s a technique which was
covered for situations like this!

Problem 5
Hint 1: Think about a substitution which works nicely with the limits and
doesn’t change the integrand much.

Problem 6
Hint 1: With a denominator like that, a tan substitution is good.
Hint 2: Try a reflection substitution - you’ll need the tan addition formula.
sin x
Alternative Hint 2: For an alternative method, use tan x = and see if you
cos x
can manipulate it to something like the log sine integral.

Problem 7
Hint 1: Complete the square and think about the denominator.
Hint 2: Try something like Problem 6!

Problem 8
Hint 1: This is the sort of integral where it’s good to mess around with it for a
bit to simplify it.
Hint 2: Try manipulate it into an integral which looks like (1.11)

23
Problem 9
Hint 1: Try the integral of inverse - it has an inverse as it’s monotonic.

Problem 10
Hint 1: Try a reflection substitution and adding the integrals together.
Hint 2: Factor a z 2 from the numerator and denominator.
Hint 3: Complete the square on the denominator.

24
3 Solutions
Problem 1
Following the hint, using the double angle formula gives us
Z π2 Z π2 Z π
2
log(cos 2x + 1)dx = log(2 cos2 x)dx = π log 2 + 2 log(cos x)dx
−π
2 −π
2 −π
2

Now thinking about the graph of cos x, it takes the same values on the interval
[− π2 , π2 ] as sin x does on [0, π] so we can write
Z π2 Z π
log(cos 2x + 1)dx = π log 2 + 2 log(sin x)dx = −π log 2
−π
2 0

Problem 2
The similarity with (1.24) suggests using the same method:
Z 2π Z 2π ix 2n
2n e − e−ix
sin xdx = dx
0 0 2i
2n Z 2π
X 2n
= 2n e(2k−2n)ix (−1)2n−k (i)2 ndx
2 k 0
k=0

π 2n
= 2n−1
2 n
which is exactly the same answer. That makes sense, thinking about the graph
of sin(x) and cos(x) on the interval [0, 2π], they both cover the same values so
they could be demonstrated to be the same with some interval splitting and
reflection substitutions.
You could also take the answer to (1.15) and just multiply it by two after
noticing that
Z 2π Z π
sin2n (x)dx = 2 sinn (x)dx
0 0

Problem 3
A similar integral done earlier was (1.18). This integral can be manipulated
into that form:
Z 2π Z 2π Z 2π
cos x 2 dx
= 1− dx = 2π − 2
0 2 + cos x 0 2 + cos x 0 2 + cos x

Now we know that

!
Z
dx 2 x r 1
= √ arctan tan +C
2 + cos x 3 2 3

25
We can’t substitute the limits 0 to 2π in though - this integral was worked out
with a t sub so we need to make sure that it’s done on an injective domain so
splitting into [0, π], we get
Z 2π Z π Z 2π Z π
dx dx dx dx 2π
= + =2 =√
0 2 + cos x 0 2 + cos x π 2 + cos x 0 2 + cos x 3
Putting this all together, we get
Z
dx 4
= 2π 1 − √
2 + cos x 3

Problem 4
This integral has a similar form to (1.23) so a similar approach could work.
Using a reflection substitution to put it in the right form,
Z π Z π
2 cos x 1 2 sin x + cos x
I= dx = dx
0 2 − sin 2x 2 0 2 − sin 2x

Now using the substitution u = sin x − cos x we get

1 1 du
Z
1 1 π
I= = [arctan(u)]−1 =
2 −1 1 + u2 2 4

Problem 5
1
This integrand is full of things which work nicely with a u = substitution; the
x
1
[0, ∞] limits, arctan(x) and the 1+x2 . Going ahead and using that substitution,
we get
∞ ∞ ∞ π
arctan( x1 ) − arctan(x) π2
Z Z Z
arctan(x) 2
I= dx = dx = dx = −I
0 1 + x2 0 1 + x2 0 1 + x2 4

Rearranging this, we get

π2
I=
8

Problem 6
This integral is famous - it’s known as Serret’s integral. Following the hint,
using a tan substitution, we get
π
Z 1 Z
ln(1 + x) 4
I= dx = ln(1 + tan x)dx
0 1 + x2 0

26
Now with a reflection substitution, this becomes
Z π
4
I= ln(1 + tan( π4 − x))dx
0

With the addition formula for tan, we have

tan( π4 ) − tan(x) 1 − tan(x)

tan( π4 − x) = π =
1 + tan( 4 ) tan(x) 1 + tan(x)

Putting this into the above,

π π Z π4
1 − tan(x)
Z Z
4 4 2 π
I= ln(1 + tan( π4 − x))dx = ln 1 + dx = ln = ln 2 − I
0 0 1 + tan(x) 0 1 + tan(x) 4

This gives us
Z 1
ln(1 + x) π
dx = ln 2
0 1 + x2 8

For the alternative method (which I prefer - seems more natural), writing the
tan integral in terms of sines and cosines
Z π Z π Z π
4 4 4
ln(1 + tan x)dx = ln(sin x + cos x)dx − ln(cos x)dx
0 0 0
π Z π4
Z √
4 π
= ln( 2 sin x − )dx − ln(cos x)dx
0 4 0
Z π4 Z π
π π 4 π
= ln 2 + ln(sin x − )− ln(cos x)dx = ln 2
8 0 4 0 8

where the second to last line can be seen by a reflection substitution or just
thinking about the values sine and cosine take on those intervals.

Problem 7
This one can be treated like Problem 6. Completing the square on the denomi-
nator gives us
Z ∞
ln x
I= dx
0 (x + 1)2 + 1

In this form, a tan substitution is good; x + 1 = tan θ which gives us

Z π
2
I= ln(tan θ − 1)dθ
π
4

27
This can now be treated like the alternative method of Problem 6.
Z π2 Z π2
I= ln(sin θ − cos θ)dθ − ln(cos θ)dθ
π π
4 4
π π
Z 2 √ π
Z 2 π
= ln( 2 sin θ − )dθ − ln(cos θ)dθ = ln 2
π
4
4 π
4
8

which is the same answer as the previous problem! I tried to manipulate them
to see if there’s a way to see if they’re equal but couldn’t do it.

Problem 8
There are two approaches to this - we could go straight with x = a tan u or just
manipulate it to turn it into an integral we’ve done before - it looks similar to
(1.11). To do that, putting in x = au gives
Z ∞
1 ∞ ln(au)
Z Z ∞ Z ∞
ln x 1 ln(a) ln(u)
dx = du = du + du
0 x2 + a2 a 0 u2 + 1 a 0 u2 + 1 0 u2 + 1
The second integral we’ve done before; its 0 and the first integral is fairly stan-
dard, giving us
Z ∞
ln x π ln a
2 + a2
dx =
0 x 2a

Problem 9
Following the hint for this problem, the inverse of the function is x21+1 whose
integral isn’t too difficult. When x21+1 = 0 is 1, x = 0 but for 0, it’s x → ∞.
However substituting this into the formula could cause problems so taking limits
instead,
Z b Z 1 r
dx 1 b
2+1
− − 1dx = 2
0 x 2
1 x b +1
b +1

Taking the limit as b → ∞ we get

Z 1r Z ∞
1 dx π
− 1dx = 2+1
=
0 x 0 x 2

Problem 10
This integral is fairly tough to do with these methods, it can be done with more
advanced techniques like contour integration but this method is much nicer.
First by using a reflection substitution, we get
Z ∞ Z ∞ 2
− dx
Z 0
dx x2 x dx
I= 4
= 1 =
0 x +1 ∞ x4 + 1 0 x4 + 1

28
Adding these two together, we get
∞
x2 + 1
Z
2I = dx
0 x4 + 1

Now the difficult step is to divide by x2 , then completing the square on this
gives us

1 ∞ 1 + x12
Z ∞
1 + x12
Z
1
I= dx = dx
2 0 x2 + x12 2 0 (x − x1 )2 + 2

The numerator is now actually the derivative of x − x1 so substituting u = x − x1

(which is injective as it’s always increasing; the derivative of u is always positive)
gives us

1 ∞ du
Z
π
I= = √
2 −∞ u2 + 2 2 2
by standard tan substitution stuff.
Another ’basic technique’ method would be to use partial fractions but I really
don’t like partial fractions and the factorisation of x4 + 1 isn’t that easy to find;
once you know it exists it’s a bit easier though!
For the other part, use the integral of inverse, it works out the same as Problem
9.

29