Integration Practice Notebook Problems & Techniques

Problem 106

▶ Problem

Evaluate the integral:

Z π/2
sin(x)
I= 5
dx
0 [sin(x) + cos(x)]

⊛ Solution

Strategy: King Property and Half-Angle Identities

The solution employs a classic and powerful technique known as

the ”King Property” of definite integrals to simplify the integrand.
This is followed by a series of trigonometric identities and substi-
tutions to reduce the integral to a simple polynomial. Ra
Step 1: Apply the King Property We use the property f (x)dx =
Ra 0
0
f (a − x)dx. Here, a = π/2.
Z π/2
sin(π/2 − x)
I= 5
dx
0 [sin(π/2 − x) + cos(π/2 − x)]
Z π/2
cos(x)
= 5
dx
0 [cos(x) + sin(x)]
This shows that our integral I is equal to the same integral but
with cos(x) in the numerator.
Step 2: Combine the Integrals We add the two forms of the integral
together:
Z π/2 Z π/2
sin(x) cos(x)
2I = I + I = 5
dx + 5
dx
0 [sin(x) + cos(x)] 0 [sin(x) + cos(x)]
Z π/2 Z π/2
sin(x) + cos(x) 1
2I = 5
dx = 4
dx
0 [sin(x) + cos(x)] 0 [sin(x) + cos(x)]

1
Integration Practice Notebook Problems & Techniques

Z π/2
1 dx
I=
2 0 [sin(x) + cos(x)]4
Step 3: Simplify the Denominator We use the identity (sin x +
2 2 2
cos x) = sin x + cos x + 2 sin x cos x = 1 + sin(2x).
1 π/2 1 π/2
Z Z
dx dx
I= =
2 0 ([sin(x) + cos(x)]2)2 2 0 [1 + sin(2x)]2
Step 4: Substitution Let u = 2x, so du = 2dx. The limits become
x ∈ [0, π/2] → u ∈ [0, π].
1 π 1 π
Z Z
du/2 du
I= =
2 0 [1 + sin(u)]2 4 0 [1 + sin(u)]2
Step 5: Symmetry and Phase ShiftWe use the identity sin(u) =
cos(u − π/2). Let v = u − π/2, so du = dv. The limits be-
come u ∈ [0, π] → v ∈ [−π/2, π/2].
1 π/2
Z
dv
I=
4 −π/2 [1 + cos(v)]2
The integrand is an even function, so we can simplify the integral:
Z π/2
1 π/2
Z
1 dv dv
I = ·2 =
4 0 [1 + cos(v)]2 2 0 [1 + cos(v)]2
Step 6: Half-Angle Identity We use the identity 1 + cos(v) =
2
2 cos (v/2).
π/2
1 π/2
Z Z
1 dv dv
I= =
2 0 [2 cos2(v/2)]2 2 0 4 cos4(v/2)
1 π/2 4
Z
= sec (v/2)dv
8 0
4
Step 7: Final Integration We rewrite sec to prepare for a u-
substitution using w = tan(v/2).
1 π/2 2
Z
I= sec (v/2) · sec2(v/2)dv
8 0

2
Integration Practice Notebook Problems & Techniques

Z π/2
1
= [tan2(v/2) + 1] sec2(v/2)dv
8 0

The antiderivative of f (w) = w2 + 1 is F (w) = w3/3 + w. The

differential is dw = 12 sec2(v/2)dv, so we need to multiply the
antiderivative by 2.
  π/2
1 1
I= 2 tan3(v/2) + tan(v/2)
8 3 0

Now, we evaluate at the limits:

 
1 1 1
I= tan3(π/4) + tan(π/4) − [0 + 0]
4 3 4
     
1 1 3 1 1 1 4 1
= (1) + 1 = +1 = =
4 3 4 3 4 3 3

✓ Final Answer

Z π/2
sin(x) 1
dx =
0 [sin(x) + cos(x)]5 3

⋆ Key Insight

Key Insights

• RThe PowerR of the King Property: The integral

a a
0
f (x)dx = 0 f (a − x)dx is the most powerful tool in the
arsenal for simple definite integrals. Its true value is realized
when f (x) and f (a − x) are algebraically different but re-
lated. In this problem, f (x) = sinD5
x
and f (π/2 − x) = cos D5
x
.
The sum f (x) + f (a − x) simplifies to sin x+cos
D5
x
= D14 . This
single step eliminates the numerator’s asymmetry and is the
key to solving the integral.

3
Integration Practice Notebook Problems & Techniques

• Standard Simplification (sin x + cos x): The expression

sin x + cos x is ubiquitous. The most effective way to handle
it, especially when raised to a power, is often to square it:
(sin x + cos x)2 = 1 + sin(2x). This identity converts the sum
into a single, more manageable trigonometric function.
• Symmetry and Phase-Shifting: The solution’s path
from 1 + sin u to 1 + cos v is a textbook example of manip-
ulating trigonometric integrals. Using sin u = cos(u − π/2)
phase-shifts the integrand, and the even nature of the resulting
function, 1/(1 + cos v)2, allows the integral over [−π/2, π/2]
R π/2
to be ”folded” into 2 × 0 , which is a standard domain.

⇒ Alternative Approaches & Viewpoints

Alternative Approach I: R cos θ Method

This is the most direct and elegant approach. After Step 2, we

have:
1 π/2
Z
dx
I=
2 0 [sin(x) + cos(x)]4
We can use the R cos θ form (or ”harmonic addition”):
√ √
 
1 1
sin(x) + cos(x) = 2 √ sin x + √ cos x = 2 cos(x − π/4)
2 2
Substituting this into the integral:
1 π/2 1 π/2
Z Z
dx dx
I= √ =
2 0 [ 2 cos(x − π/4)]4 2 0 4 cos4(x − π/4)
Z π/2
1
= sec4(x − π/4)dx
8 0

4
Integration Practice Notebook Problems & Techniques

Now, let u = x − π/4, so du = dx. The limits become x ∈

[0, π/2] → u ∈ [−π/4, π/4].
1 π/4 4 1 π/4 4
Z Z
I= sec (u)du = sec (u)du (by even symmetry)
8 −π/4 4 0
This is a standard integral. Let v = tan u, dv = sec2 udu.
1 π/4
Z Z 1
1
I= (1 + tan2 u) sec2 udu = (1 + v 2)dv
4 0 4 0
3 1
     
1 v 1 1 1 4 1
= v+ = 1+ −0 = =
4 3 0 4 3 4 3 3
This method is arguably more straightforward than the one in the
solution.
Alternative Approach II: Weierstrass Substitution

This is the ”brute force” method. After Step 3, we have:

1 π/2
Z
dx
I=
2 0 [1 + sin(2x)]2
Let’s use the u = 2x substitution from Step 4 to get
π π
I = 41 0 [1+sin du du
R R
u]2
. The King Property shows 0 (1+sin u)2
=
R π/2 du 1 π/2
R du
2 0 (1+sin u)2 , so I = 2 0 (1+sin u)2
. Now, apply the Weierstrass
substitution t = tan(u/2).
• du = 1+t
2dt
2

• sin u = 2t
1+t2

• Limits: u ∈ [0, π/2] → t ∈ [0, 1]

1 2dt 1
1/(1 + t2)
Z Z
1 1+t2
I= =
dt
2 2t 2 1+t2 +2t 2


0 1 + 1+t 2 0 2
1 Z1+t
1
(1 + t2)2 1 + t2
Z
1
= · dt = dt
0 1 + t2 (1 + t)4 0 (1 + t)4

5
Integration Practice Notebook Problems & Techniques

Let w = 1 + t, so t = w − 1 and dt = dw. Limits t ∈ [0, 1] →

w ∈ [1, 2].
Z 2 Z 2
1 + (w − 1)2 1 + w2 − 2w + 1
I= 4
dw = 4
dw
w w
Z1 2 2 Z1 2
w − 2w + 2
= 4
dw = (w−2 − 2w−3 + 2w−4)dw
w
1 21
2
= −w−1 + w−2 − w−3
3
  1 
1 1 2 2
= − + − − −1 + 1 −
2 4 3(8) 3
     
1 1 2 3 1 2
= − − − − = − − +
4 12 3 12 12 3
4 2 1 2 1
=− + =− + =
12 3 3 3 3
This method is computationally heavy but very systematic.

⟲ Practice Exercise

Practice Exercises

1. A Simpler Case: Evaluate the following integral using the

R cos θ method:
Z π/2
dx
0 sin(x) + cos(x)
√
Hint: Write sin x + cos x = 2 cos(x − π/4). √ The antideriva-
√

tive of sec u is ln | sec u + tan u|. Answer: 2 ln 1 + 2 .
2. The n = 3 Case: Evaluate the integral from this problem,
but with the power 3:
Z π/2
sin(x)
3
dx
0 [sin(x) + cos(x)]

6
Integration Practice Notebook Problems & Techniques

Hint: Follow the exact same King Property and R cos θ steps.
You will end up integrating sec2 u. Answer: 1/4.

7
Integration Practice Notebook Problems & Techniques

Problem 107

▶ Problem

Evaluate the integral

arctan 2 cos2(x)
Z π/2
I= 2 (x)
dx.
0 cos

⊛ Solution

Strategy: A Three-Step ”Sub-IBP-Contour” Method

The integrand is a non-trivial composition of trigonometric and

inverse trigonometric functions. Direct integration or a simple
parametric differentiation (Feynman’s trick) appears very difficult.
Our strategy will be to transform the integral into a more man-
ageable form.
1. Substitution: We’ll use the substitution u = tan(x) to con-
vert the trigonometric integral into a rational function from 0
to ∞.
2. Integration by Parts (IBP): The new integral will still
contain an arctan function. We will use IBP to eliminate the
arctan in exchange for a simpler, but still advanced, rational
integral.
3. Contour Integration: The final real-valued integral is a
classic candidate for evaluation using the Residue Theorem
with a semicircular contour.

Step 1: Trigonometric Substitution

du
Let u = tan x. This implies du = sec2 x dx, so dx = 1+u 2 . We

also use the identity cos2 x = sec12 x = 1+tan

1 1
2 x = 1+u2 . The limits

8
Integration Practice Notebook Problems & Techniques

transform from x ∈ [0, π/2] to u ∈ [0, ∞).

1
Z ∞

arctan 2 · 1+u2 du
I= 1 ·
0 1+u 2 1 + u2
Z ∞  
2
I= arctan 2
du.
0 1 + u
This new form is much cleaner, but still not elementary.

Step 2: Integration by Parts

R R
We will now apply integration by parts, f dv = f v − vdf .
• Let f (u) = arctan 1+u2


2

• Let dv = du, which means v = u.

First, we find df :
 
1 d 2
df =
2 · du
1+ 2 du 1 + u2
1+u2

(1 + u2)2 −4u −4u

= · du = du.
(1 + u2)2 + 4 (1 + u2)2 u4 + 2u2 + 5
Applying the IBP formula:
  ∞ Z ∞  
2 −4u
I = u · arctan − u · du.
1 + u2 0 0 u4 + 2u2 + 5

The boundary term at u = 0 is 0. At u → ∞, we use arctan(ϵ) ≈

ϵ:    
2 2
lim u · arctan ≈ lim u · = 0.
u→∞ 1 + u2 u→∞ 1 + u2
So, the boundary term vanishes, and we are left with a rational
integral: Z ∞
u2
I=4 4 + 2u2 + 5
du.
0 u

9
Integration Practice Notebook Problems & Techniques

Step 3: Contour Integration

This integral is perfectly suited for complex analysis. Since the

integrand is even, we can write:
1 ∞
Z ∞
x2 x2
Z
I =4· dx = 2 dx.
2 −∞ x4 + 2x2 + 5 −∞ x 4 + 2x2 + 5

z 2
We integrate the complex function f (z) = z 4+2z 2 +5 around a large

semicircular contour in the upper half-plane (UHP).

ΓR
Im(z)

z2 z1

−R Real Axis RRe(z)

The poles are the roots of z 4 + 2z 2 + 5 = 0. Let w = z 2.

√
−2 ± 4 − 20
w2 + 2w + 5 = 0 =⇒ w = = −1 ± 2i.
2
We must now find the square roots of these two numbers. We
only need the roots in the UHP.
• Case 1: z 2 = −1 + 2i. Let z1 = a + bi with a, b > 0.
a2 − b2 = −1 and ab = 1. q
This gives a4 + a2 − 1 = 0, so
√ √
2 −1+ 5 5−1
a = 2 . This gives a = 2 .

• Case 2: z 2 = −1 − 2i. Let z2 = c + di with d > 0.

c2 − d√2 = −1 and cd = −1. This gives c4 + c2 − 1 = 0, so
c2 = 5−1 = −1, c and
2 . Since cd q qd have opposite signs. The
√ √
5−1 5+1
UHP root is z2 = − 2 +i 2 .

10
Integration Practice Notebook Problems & Techniques

This calculation is very tedious. There is a much faster way to find

the residues. Let h(z) = z 4 + 2z 2 + 5. Then h′(z) = 4z 3 + 4z =
z2
4z(z 2 + 1). The residue at a simple pole zk is Res(f, zk ) = h′(zkk ) =
zk2 zk
4zk (zk2 +1)
= 4(zk2 +1)
.

• At z1, where z12 = −1 + 2i: Res(f, z1) = z1

4(−1+2i+1) = z1
8i .

• At z2, where z22 = −1 − 2i: Res(f, z2) = z2

4(−1−2i+1) = z2
−8i .

The sum of residues in the UHP is:

X z1 z2 z1 − z2
Res = + = .
8i −8i 8i
Let’s find z1 − z2:
r√ r√ r√ r√
5−1 5+1 5−1 5+1
z1 = +i and z2 = − +i .
2 2 2 2
r√
5−1
z1 − z2 = 2 .
2
Plugging this in:
r√ ! r√
X 1 5−1 1 5−1
Res = 2 = .
8i 2 4i 2

The integral along the real axis is 2πi times this sum (the arc
integral vanishes):
Z ∞ r√ ! r√
2
x 1 5−1 π 5−1
4 + 2x2 + 5
dx = 2πi = .
−∞ x 4i 2 2 2

Finally, we substitute this back into our expression for I:

Z ∞ r√ ! r√
2
x π 5−1 5−1
I=2 4 + 2x2 + 5
dx = 2 = π .
−∞ x 2 2 2

11
Integration Practice Notebook Problems & Techniques

✓ Final Answer

s√
π/2
arctan 2 cos2(x) 5−1
Z
dx = π .
0 cos2(x) 2

⋆ Key Insight

An Unexpected Connection: The Golden Ratio

The final answer,

√ while a bit messy, contains a very famous number.
5−1
The
√
term 2 is the reciprocal of the **Golden Ratio**, ϕ =
1+ 5
2 . Therefore, our solution can be written in the astonishingly
simple form: r
1 π
I=π =√ .
ϕ ϕ
This deep connection between a challenging trigonometric integral
and a fundamental constant of geometry is a beautiful example of
the unity of mathematics.
Some Insights

1. Start by using a rare integral representation for the arctangent

function: Z ∞
arctan(A) dt
= 2 + A2
.
A t
R ∞ 1 dt
(This is a valid identity, as 1 t2+A2 = [ A1 arctan(t/A)]∞ 1 =
1 π 1 1
A ( 2 − arctan(1/A)) = A arccot(1/A) = A arctan(A).)
R π/2 R ∞
2. Then rewrite the integral as I = 0 2 1 t2+(2 dt cos2 x)2
dx. This
step is a mix of identities and is confusingly written, but it’s
A/2 arctan A
based on arctan
C 2
A
= C2
· A/2 , which is a valid (if strange)
setup.
3. It then swaps the order of integration (Fubini’s Theorem) to

12
Integration Practice Notebook Problems & Techniques

get a complex inner integral, which it evaluates using the sub-

stitution u = tan x.
R √
4. The final lines, π Re · · · = π Re −1 + 2i, are a massive
leap of calculation that condense a very difficult contour inte-
gral into a single line.
q√
5−1
Our solution’s final answer, π 2 . √To prove this, we can cal-
√
culate the real part of −1 + 2i. Let −1 + 2i = a + bi. Then
(a + bi)2 = a2 − b2 + 2abi = −1 + 2i.
• a2 − b2 = −1
• 2ab = 2 =⇒ ab = 1 =⇒ b = 1/a
• a2 − (1/a2) = −1 =⇒ a4 + a2 − 1 = 0.
√ √
−1+ 1+4
•a =
2
2 = 5−1
(since a2 > 0).
2
√ q√
• a = Re( −1 + 2i) = 5−1
2 .

√ q√
5−1
Thus, π Re −1 + 2i = π 2 . Our answer is identical.

You can verify this result with a computer algebra system. In Wol-
framAlpha or Mathematica:

Integrate[ArcTan[2 Cos[x]^2]/Cos[x]^2, {x, 0, Pi/2}]

The system will return ‘Pi/Sqrt[GoldenRatio]‘ or an equivalent form.

⟲ Practice Exercise

Test Your Understanding

1. A
Z ∞ Simpler IBP-Contour Problem. Evaluate
arctan(x) dx. This integral diverges, but what about
0

13
Integration Practice Notebook Problems & Techniques

Z ∞
2


arctan 1/x dx? Hint: Use Integration by Parts, which
0 R ∞ 2x2
leads to 0 1+x4 dx. Evaluate this using contour integration.
Answer: √π2 .
2. The UHP Poles. In our solution, we found two poles in
the UHP. Show that the other two poles, from z 2 = −1 − 2i,
are in the lower half-plane.
√
1 1+ 5
3. The Golden Ratio. Show that ϕ = ϕ−1, where ϕ = 2 .

14
Integration Practice Notebook Problems & Techniques

Problem 108

▶ Problem

Evaluate the limit of the following series as p → 0+:

∞  
X ln(n + x) ln(n − x)
L(x, p) = p
− p
n=1
(n + x) (n − x)

for 0 < x < 1.

⊛ Solution

Strategy: Generating Function via Hurwitz Zeta

The core strategy is to recognize that the given series L(x, p) is

the negative partial derivative of a simpler series with respect to
the parameter p. This simpler series, S(x, p), can be directly
expressed in terms of the Hurwitz Zeta function, ζ(s, a). The
final result is then obtained by differentiating the Hurwitz Zeta
function and evaluating the limit using a standard identity.
Step 1: Relate the Series to a Derivative We use the identity from
differential calculus:
∂ ∂
(n ± x)−p = e−p ln(n±x)
∂p ∂p
ln(n ± x)
= − ln(n ± x)e−p ln(n±x) = −
(n ± x)p
From this, we can see that our series L(x, p) is the negative deriva-
tive of a simpler sum S(x, p):
∞
∂ X
(n + x)−p − (n − x)−p

L(x, p) = −
∂p
n=1

15
Integration Practice Notebook Problems & Techniques

P∞ −p
Let S(x, p) = n=1 [(n + x) − (n − x)−p].
Step 2: Express the Sum using the Hurwitz Zeta Function The Hurwitz
P∞ −s
Zeta function is defined as ζ(s, a) = n=0 (n + a) . We can
express our two sums in this form:
P∞ P∞ P∞
• n=1(n + x)−p = n=1((n − 1) + 1 + x)−p = k=0(k + 1 +
x)−p = ζ(p, 1 + x)
P∞ P∞ P∞
• n=1(n − x)−p = n=1((n − 1) + 1 − x)−p = k=0(k + 1 −
x)−p = ζ(p, 1 − x)
This is valid since 1 − x > 0. So, the simpler sum is S(x, p) =
ζ(p, 1 + x) − ζ(p, 1 − x).

Step 3: Evaluate the Limit We substitute this back into our expres-
sion for L(x, p):
∂
L(x, p) = − [ζ(p, 1 + x) − ζ(p, 1 − x)]
∂p
∂ζ ∂ζ
= (p, 1 − x) − (p, 1 + x)
∂p ∂p
We now take the limit as p → 0+. This requires a known identity
for the derivative of the Hurwitz Zeta function at s = 0.
∂ζ ∂ζ
lim+ L(x, p) = (p, 1 − x) − (p, 1 + x)
p→0 ∂p p=0
∂p p=0

16
Integration Practice Notebook Problems & Techniques

Derivation of the Final Step using Lerch’s Formula

The solution relies on a well-known, advanced identity (cited as

DLMF 25.11.18), often known as **Lerch’s formula** (or a special
case thereof):
∂ζ 1
(s, a) = ln Γ(a) − ln(2π)
∂s s=0
2
This identity provides the crucial link between the derivative of the
Zeta function at the origin and the Gamma function. We apply
this formula to the two terms in our limit expression:
1. ∂ζ
∂p (p, 1 − x) = ln Γ(1 − x) − 12 ln(2π)
p=0

∂ζ
2. ∂p (p, 1 + x) = ln Γ(1 + x) − 21 ln(2π)
p=0

Now, we take the difference of these two expressions:

   
1 1
lim L(x, p) = ln Γ(1 − x) − ln(2π) − ln Γ(1 + x) − ln(2π)
p→0+ 2 2
= ln Γ(1 − x) − ln Γ(1 + x)
 
Γ(1 − x)
= ln
Γ(1 + x)
This successfully derives the final boxed answer from the interme-
diate steps.

✓ Final Answer

∞    
X ln(n + x) ln(n − x) Γ(1 − x)
lim − = ln
p→0+
n=1
(n + x)p (n − x)p Γ(1 + x)

17
Integration Practice Notebook Problems & Techniques

⋆ Key Insight

Key Insights: The Zeta Function as a Generating Function

This problem provides a deep insight into the nature of infinite

series and special functions.
• Generators for Series: The core technique treats the ex-
ponent p as a continuous parameter. The ”generating sum”
[(n + x)−p − (n − x)−p] is a simpler object to
P
S(x, p) =
analyze. The desired series L(x, p) is then ”generated” by dif-
ferentiation: L = −∂S/∂p. This technique is fundamental in
many areas of analysis and physics.
• The Hurwitz Zeta Function: This function is the natural
tool for ”shifted” sums, i.e., sums that do not start at n = 1 or
P
have an offset a. The problem’s structure, with its (n + x)
P
and (n − x) terms, is tailor-made for analysis with ζ(s, a).
• Lerch’s Formula: The final result hinges on the profound
identity ∂ζ 1
∂s (s, a) s=0 = ln Γ(a) − 2 ln(2π). This formula con-
nects two of the ”grand objects” of mathematics: the Zeta
function (from number theory) and the Gamma function (from
analysis). It implies that the ”rate of change” of the Zeta func-
tion’s value at s = 0 is governed by the Gamma function. The
1
2 ln(2π) term, which cancels in our problem, is also signifi-
cant, appearing in contexts like the functional equation and
determinants of Laplacians.

18
Integration Practice Notebook Problems & Techniques

⟲ Practice Exercise

Practice Exercises

1. Evaluate at x = 1/2: Use the final result and properties of

the Gamma function to evaluate the sum at x = 1/2.
∞  
X ln(n + 1/2) ln(n − 1/2)
lim −
p→0+ (n + 1/2)p (n − 1/2)p
n=1

Hint: Use Γ(1 + z) = zΓ(z). Answer: ln(2).

2. Euler’s Reflection Formula: Use the final answer and
π
Euler’s Reflection Formula Γ(z)Γ(1 − z) = sin(πz) to express
the answer in a different form.

19
Integration Practice Notebook Problems & Techniques

Problem 109

▶ Problem

Evaluate the sum

∞  
X 1 1
S=2 (−1)k + .
(3k + 1)3 (3k + 2)3
k=0

⊛ Solution

Strategy: Connecting the Series to the Polylogarithm

This is a sophisticated alternating series that resists standard sum-

mation techniques. The denominator’s structure (3k +n) suggests
a connection to Dirichlet series or related special functions. The
key to this problem is to show that our sum S is actually a dis-
guised form of the imaginary part of a specific Trilogarithm
function, Li3(z).
Our strategy will be:
1. Expand the sum S to understand its term-by-term struc-
ture.
2. Define the Trilogarithm Li3(z) and investigate the imag-
inary part of Li3(eiπ/3).
3. Expand the Polylogarithm sum and show that it is a
constant multiple of S.
4. Solve for S using the known (though advanced) value for
Im[Li3(eiπ/3)].

20
Integration Practice Notebook Problems & Techniques

Step 1: Unpacking the Sum S

Let’s write out the first few terms of S to see the pattern.
∞  
X 1 1
S=2 (−1)k +
(3k + 1)3 (3k + 2)3
k=0
= 2 [(k = 0) − (k = 1) + (k = 2) − . . . ]
      
1 1 1 1 1 1
=2 + − + + + − ...
13 23 43 53 73 83
1 1 1
Note that the terms , , ,...
33 63 93
are all missing.

Step 2: Defining the Comparison Function

The Polylogarithm function Lis(z) is defined by the series:

∞
X zk
Lis(z) = .
ks
k=1

Our problem is of order s = 3, so we are interested in the Trilog-

arithm, Li3(z). The key is to evaluate this at the complex value
z = eiπ/3. Let’s find the imaginary part of this function.
∞
!
iπ/3 k
X (e )
Im[Li3(eiπ/3)] = Im
k3
k=1
∞
!
X cos(kπ/3) + i sin(kπ/3)
= Im
k3
k=1
∞
X sin(kπ/3)
= .
k3
k=1

Step 3: The ”Magic” - Proving the Connection

P∞
Now, let’s expand the sum k=1 sin(kπ/3)
k3
by looking at the peri-
odic sequence sin(kπ/3):
√
• k = 1 : sin(π/3) = 3/2

21
Integration Practice Notebook Problems & Techniques

√
• k = 2 : sin(2π/3) = 3/2
• k = 3 : sin(3π/3) = sin(π) = 0
√
• k = 4 : sin(4π/3) = − 3/2
√
• k = 5 : sin(5π/3) = − 3/2
• k = 6 : sin(6π/3) = sin(2π) = 0
√ √ √ √
3 3 3 3
This pattern, [ 2 , 2 , 0, − 2 , − 2 , 0],
repeats indefinitely. Let’s
substitute this sequence back into the sum for Im[Li3(eiπ/3)]:
sin(π/3) sin(2π/3) sin(3π/3) sin(4π/3)
Im[Li3(eiπ/3)] = 3
+ 3
+ 3
+ 3
+ ...
√ 1 √ 2 √ 3 √ 4
3/2 3/2 0 3/2 3/2 0
= 3
+ 3
+ 3
− 3
− 3
− 3
+ ...
1
√  2   3 4   5 6  
3 1 1 1 1 1 1
= + − + + + − ...
2 13 23 43 53 73 83
A moment of insight is needed here. Look at the expression in the
large brackets. It is exactly our original sum S divided by 2!
√   √
3 S 3
Im[Li3(eiπ/3)] = = S.
2 2 4
This is the central identity of the problem. We can now solve for
S: √
4 4 3
S = √ Im[Li3(eiπ/3)] = Im[Li3(eiπ/3)].
3 3
This confirms the primary (and most confusing) step
Step 4: Using the Known Value

The value of Im[Li3(eiπ/3)] is a known, though highly non-trivial,

result in the theory of special functions. As we know, it can
be derived from Jonquière’s inversion formula for polylogarithms.
The specific value is:
iπ/3 5π 3
Im[Li3(e )] = .
162

22
Integration Practice Notebook Problems & Techniques

Now, we can find the exact value of our sum S by substituting

this into our equation:
√  3 √ 3
√ 3
√
4 3 5π 4 3 · 5π 2 3 · 5π 10 3 3
S= = = = π .
3 162 3 · 162 3 · 81 243

✓ Final Answer

√
10 3 3
S= π .
243

⋆ Key Insight

The Harmony of Dirichlet Characters and Polylogarithms

This problem is a stunning example of how number-theoretic struc-

tures (like periodicity modulo 6) are encoded within complex anal-
ysis (in the arguments of complex numbers).
The key was to recognize that our sum S and the polylogarithm
sum Im[Li3(eiπ/3)] are secretly the same. Let’s look at their ”build-
ing blocks”:
• Our sum S is built from the character χS (k) =
[1, 1, 0, −1, −1, 0] repeating modulo 6.
• The
√ √
polylog√ sum√ is built from the character χP (k) =
[ 23 , 23 , 0, − 23 , − 23 , 0].
√
3
It’s clear that χ√P
(k) = 2 χS (k). The identity we found,
Im[Li3(eiπ/3)] = 43 S, is a direct consequence of this.
This problem is a ”toy model” for one of the deepest topics in
number theory: the study of L-functions. Both of our sums are
examples of Dirichlet L-series, and the problem essentially asks
to prove an identity between L(χS , 3) and L(χP , 3). The fact

23
Integration Practice Notebook Problems & Techniques

that both are related to π 3 is a specific instance of more general

theorems by Euler and others on the values of L-functions at integer
arguments.

Alternative Paths (and Why They are Hard)

One might try to solve this with other methods, but they are all
very difficult.
• Integral Representation: As we’ve seen, S can be related
R ∞ x2(e−x +e−2x )
to 0 1+e−3x
dx. This integral is not easier than the sum.
• Lerch Transcendent: The sum can be written as S =
2
27 [Φ(−1, 3, 1/3) + Φ(−1, 3, 2/3)]. Proving the identity that
relates this to Li3(eiπ/3) is the same problem, just in a different
and more abstract notation.
This shows that the method of ”pattern matching” the sum to a
known special function, as we did, is the most direct path.

⟲ Practice Exercise

Test Your Understanding

1. The Simpler Case (s = 1). Evaluate  the related sum for

P∞ k 1 1
s = 1: S = 2 k=0(−1) 3k+1 + 3k+2 . Hint: Use the same
method. The sum will be related
√
to Im[Li1(eiπ/3)]. Li1(z) is
just − ln(1 − z). Answer: 2π9 3 .
2. The Basel Problem Analogue (s = 2). Evaluate the
P∞ sin(kπ/3) iπ/3
sum S2 = k=1 k 2 . This is Im[Li 2 (e )]. This is a
classic value. Hint: This is a known value related to Clausen’s
integral. The point is to see the pattern.
3. A Different Character. Evaluate the sum S3 =
P∞ sin(kπ/2)
k=1 k3
. Hint: Write out the sequence sin(kπ/2). It
is [1, 0, −1, 0, . . . ]. This sum is 1 − 313 + 513 − 713 + . . . , which

24
Integration Practice Notebook Problems & Techniques

is the definition of Catalan’s constant G if s = 2. For s = 3,

3
this is β(3). Answer: β(3) = π32 .

25
Integration Practice Notebook Problems & Techniques

Problem 110

▶ Problem

Evaluate the infinite sum

∞
X (8n − 1)!! −n
S= 2
n=0
(8n)!!

⊛ Solution

Strategy: Binomial Transform and Series Multisection

The solution proceeds in three main stages:

1. Binomial Conversion: The ratio of double factorials
(8n−1)!!
(8n)!! is first converted into an expression involving the cen-
tral binomial coefficient 8n


4n .

2. Series Transformation: The resulting sum is transformed

using the identity 2k = (−4)k −1/2



k k . This converts the
summand into a term from the general binomial series for
(1 + y)−1/2.
3. Series
Multisection: The sum is now of the form
P −1/2 4n
4n y , which only selects every fourth term of a known
series. A ”root of unity filter” (or series multisection) is used
to find a closed form for this partial sum.

We first ex-
Step 1: Convert Double Factorials to Binomial Coefficients
press the double factorials using standard factorials. The even
double factorial is:
(8n)!! = 2 · 4 · 6 · · · (8n) = 24n(1 · 2 · 3 · · · (4n)) = 24n(4n)!

26
Integration Practice Notebook Problems & Techniques

The odd double factorial is:

1 · 2 · 3 · · · (8n) (8n)! (8n)!
(8n − 1)!! = = = 4n
2 · 4 · 6 · · · (8n) (8n)!! 2 (4n)!
Now, we take their ratio:
(8n − 1)!! (8n)!/[24n(4n)!] (8n)!
= =
(8n)!! 24n(4n)! [24n(4n)!]2
 
(8n)! 1 8n 1
= =
(4n)!(4n)! 28n 4n 28n
Substituting this back into the original sum S:
∞    ∞  
X 8n 1 −n
X 8n 1
S= 2 =
4n 28n 4n 29n
n=0 n=0

We now use a powerful

Step 2: Apply the Binomial Series Transform
identity that relates the central binomial coefficient 2k


k to the
binomial series with α = −1/2.

27
Integration Practice Notebook Problems & Techniques

2k −1/2



Derivation of k = (−4)k k

We start
from the definition of the generalized binomial coef-

ficient αk = α(α−1)···(α−k+1)
k! . Let α = −1/2:
 
−1/2 (−1/2)(−3/2)(−5/2) · · · (−1/2 − k + 1)
=
k k!
(−1)k (1/2)(3/2)(5/2) · · · ((2k − 1)/2)
=
k!
k
(−1) (1 · 3 · 5 · · · (2k − 1))
=
2k k!
To get a full factorial, we multiply the numerator and denom-
inator by the missing even terms, (2 · 4 · 6 · · · (2k)) = 2k k!:
(−1)k (1 · 3 · 5 · · · (2k − 1)) · (2 · 4 · 6 · · · (2k))
 
−1/2
=
k (2k k!) · (2k k!)
(−1)k (2k)! (−1)k (2k)! (−1)k 2k
 
= 2k = =
2 (k!)2 4k k!k! 4k k
2k


Rearranging this gives the desired identity: k =
4k −1/2
= (−4)k −1/2



(−1)k k k .

Let k = 4n in this identity. The solution path is:

4n
     
8n 1 4n −1/2 1 4 −1/2
= (−4) =
4n 28n 4n 28n 28n 4n
(22)4n −1/2 28n −1/2
     
−1/2
= 8n = 8n =
2 4n 2 4n 4n
Substituting this into our expression for S from Step 1 :
∞    ∞  
X 8n 1 −n
X −1/2 −n
S= 2 = 2
4n 28n 4n
n=0 n=0

This simplifies the sum significantly. Let x = 1/2. We want to

28
Integration Practice Notebook Problems & Techniques

evaluate:
∞   ∞  
X −1/2 X −1/2
S= xn = (x1/4)4n
4n 4n
n=0 n=0

Let y = x1/4 = (1/2)1/4 = 2−1/4. The sum becomes:

∞  
X −1/2 4n
S= y
4n
n=0

The sum S is now

Step 3: Series Multisection (Root of Unity Filter)
the sum of every fourth term (i.e., terms with m = 4n) of the full
binomial series:
∞  
X −1/2 m
f (y) = y = (1 + y)−1/2
m
m=0

To isolate the terms m ≡ 0 (mod 4), we use the 4th roots of unity
{1, −1, i, −i}. The general formula for the m ≡ 0 (mod 4) terms
(let’s call the sum S0) is:
1
S0 = [f (y) + f (−y) + f (iy) + f (−iy)]
4
Our sum S is exactly this S0. Substituting f (y) = (1 + y)−1/2:
1
S = (1 + y)−1/2 + (1 − y)−1/2 + (1 + iy)−1/2 + (1 − iy)−1/2

4
We can simplify the complex part. Since 1 − iy is the complex
√ √
conjugate of 1 + iy, 1 − iy is the conjugate of 1 + iy, and
(1 − iy)−1/2 is the conjugate of (1 + iy)−1/2. Therefore, the sum
of these two terms is twice the real part:
−1/2 −1/2 −1/2


(1 + iy) + (1 − iy) = 2Re (1 + iy)
This simplifies the formula for S to:
1 1 1
S = (1 + y)−1/2 + (1 − y)−1/2 + Re (1 + iy)−1/2


4 4 2

29
Integration Practice Notebook Problems & Techniques

Finally, we substitute back y = 2−1/4:

1 1 1
S = (1 + 2−1/4)−1/2 + (1 − 2−1/4)−1/2 + Re (1 + i2−1/4)−1/2


4 4 2
This is the final closed-form expression.

✓ Final Answer
The closed-form expression for the sum is:
∞
X (8n − 1)!!
2−n
n=0
(8n)!!
 
1 1 1 1
= √ + √ + Re √
4 1 + 2−1/4 4 1 − 2−1/4 2 1 + i2−1/4
Numerically, this value is approximately 1.2220.

⋆ Key Insight

Key Insight

The problem’s intimidating appearance, featuring double factori-

als, is its main disguise. The solution hinges on two key transfor-
mations:
1. Double Factorial to Binomial: The ratio (2k−1)!! (2k)!! is a
standard identity for 41k 2k


k . Recognizing this (or deriving it)
is the first crucial step to simplifying the problem.
2. Binomial to Binomial Series: The second transforma-
k −1/2
tion, 2k



k = (−4) k , is less common but incredibly pow-
erful. It connects the central binomial coefficients (a discrete
sequence) to the Taylor series coefficients of the simple alge-
braic function (1 + y)−1/2.
This transforms a complex, custom-built sum into a ”multisection”

30
Integration Practice Notebook Problems & Techniques

of a fundamental series, which can then be solved with standard

techniques.

⇒ Alternative Approaches & Viewpoints

The Series Multisection Formula

The technique used in Step 3 is a general method for ”filtering”

P∞
a power series f (x) = n=0 anxn. To find the sum of all terms
where the index n is a multiple of p (i.e., n ≡ 0 (mod p)), we use
the formula:
∞ p−1
X 1 X
apk xpk = f (ω j x)
p
k=0 j=0

where ω = e2πi/p is the p-th principal root of unity.

In our problem, f (y) = (1 + y)−1/2, p = 4, and the 4th roots of
unity are ω j = {1, i, −1, −i}. This gives:
1
S0 = [f (y) + f (iy) + f (−y) + f (−iy)]
4
This is precisely the formula used in the solution.

⟲ Practice Exercise

Practice Exercises

1. Double Factorial Identity: Prove the identity (2k−1)!! (2k)!! =

1 2k


4k k
. Hint: Use the definitions (2k)! = (2k)!!(2k − 1)!! and
(2k)!! = 2k k!.
2. Series Multisection (Even/Odd): Use the series mul-
tisection technique (with p = 2 and roots {1, −1}) to find
P∞ x2n+1
the sum S = n=0 (2n+1)! . Hint: Start with f (x) =
∞ xn
ex =
P
n=0 n! . You are looking for the odd part, S1 =
1 1 x −x
2 [f (x) − f (−x)]. Answer: 2 (e − e ) = sinh(x).

31
Integration Practice Notebook Problems & Techniques

Problem 111

▶ Problem

Evaluate for a, b ∈ R \ {0} with sgn(a) = sgn(b):

Z ∞
arctan(bx) − arctan(ax)
I= dx
0 x

⊛ Solution

Strategy: Substitution and Asymptotic Analysis

The strategy is to handle the improper integral by introducing a

limit, Λ → ∞. By using a substitution, the two terms are com-
bined into a single integral from |a|Λ to |b|Λ. We then analyze the
asymptotic behavior of the integrand as its argument approaches
infinity to find the limit.
Step 1: Express as a Limit and Substitute We first write the improper
integral as a limit and split the terms:
Z Λ Z Λ 
arctan(bx) arctan(ax)
I = lim dx − dx
Λ→∞ 0 x 0 x
In the first integral, let u = bx, and in Rthe second, u = ax. In
y
both cases, dx/x = du/u. Let F (y) = 0 arctan(u) u du. Our limit
becomes:
I = lim [F (bΛ) − F (aΛ)]
Λ→∞
We must consider two cases, since sgn(a) = sgn(b).
1. Case 1: a, b > 0. I = limΛ→∞ (F (bΛ) − F (aΛ)) =
R bΛ
limΛ→∞ aΛ arctan
x
x
dx.
2. Case 2: a, b < 0. Let a = −|a| and b = −|b|. The
integrand arctan
u
u
is an even function, so its integral F (y) is

32
Integration Practice Notebook Problems & Techniques

an odd function, F (y) = −F (−y).

I = lim (F (−|b|Λ) − F (−|a|Λ)) = lim (−F (|b|Λ) + F (|a|Λ))
Λ→∞ Λ→∞
Z |a|Λ Z |b|Λ
arctan x arctan x
= lim dx = − lim dx
Λ→∞ |b|Λ x Λ→∞ |a|Λ x

Both cases can be combined into a single expression using the sign
function: Z |b|Λ
arctan x
I = sgn(a) lim dx
Λ→∞ |a|Λ x

Step 2: Asymptotic Expansion and Evaluation For large x, we use the

identity arctan(x) = π2 − arctan(1/x).
|b|Λ π
− arctan(1/x)
Z
2
I = sgn(a) lim dx
Λ→∞ |a|Λ x
"Z #
|b|Λ Z |b|Λ
π/2 arctan(1/x)
= sgn(a) lim dx − dx
Λ→∞ |a|Λ x |a|Λ x

We evaluate the two parts separately:

1. The Main Term:
Z |b|Λ
π/2 π |b|Λ π
dx = [ln |x|]|a|Λ = (ln(|b|Λ) − ln(|a|Λ))
|a|Λ x 2 2
   
π |b|Λ π |b|
= ln = ln
2 |a|Λ 2 |a|

2. The Remainder Term (R): As Λ → ∞, the integration

variable x is large. We can bound the remainder integral. For
x > 0, arctan(1/x) < 1/x.
Z |b|Λ Z |b|Λ Z |b|Λ
arctan(1/x) 1/x 1
|R| = dx ≤ dx = 2
dx
|a|Λ x |a|Λ x |a|Λ x

33
Integration Practice Notebook Problems & Techniques

 |b|Λ    
1 1 1 1 1 1
= − = − + = −
x |a|Λ
|b|Λ |a|Λ Λ |a| |b|
As Λ → ∞, this remainder R clearly goes to 0.

Combining the results, we have:

Step 3: Final Result
     
π |b| πsgn(a) |b|
I = sgn(a) ln −0 = ln
2 |a| 2 |a|
|b|
sgn(b)|b|
Since sgn(a) = sgn(b), we can write = sgn(a)|a|
|a| = ab .
 
πsgn(a) b
I= ln
2 a

✓ Final Answer

∞  
arctan(bx) − arctan(ax)
Z
πsgn(a) b
dx = ln
0 x 2 a

⋆ Key Insight

Key Insights: Frullani’s Integral

This problem isRa classic example of a **Frullani Integral**, which

∞
takes the form 0 f (ax)−fx
(bx)
dx.
• General Formula: The standard result, provided f (x) is
continuous and the limits f (0) and f (∞) = limx→∞ f (x) ex-
ist, is:
Z ∞  
f (ax) − f (bx) b
dx = (f (0) − f (∞)) ln
0 x a

• The ”Trick” in this Problem: The integral in the prob-

R ∞ arctan(bx)−arctan(ax)
lem is 0 x dx. This is the negative of the

34
Integration Practice Notebook Problems & Techniques

standard Frullani form (the ’a’ and ’b’ terms are swapped).
Let f (x) = arctan(x). We have f (0) = 0.
• Handling the Sign of Infinity: The function f (∞) de-
pends on the sign of the argument. Since x > 0 in the inte-
gral, f (∞) = limx→∞ arctan(kx) = π2 sgn(k). The condition
sgn(a) = sgn(b) = s is crucial. Let f (∞) = s · π/2.
• Applying the Theorem:
Z ∞
arctan(bx) − arctan(ax)
I= dx
0 x
Z ∞
arctan(ax) − arctan(bx)
=− dx
0 x
     
b π b
= − (f (0) − f (∞)) ln = − (0 − s · ) ln
a 2 a
   
π b πsgn(a) b
= s · ln = ln
2 a 2 a
This confirms the ”non-Frullani” derivation and shows how the
theorem elegantly handles both positive and negative cases.

⇒ Alternative Approaches & Viewpoints

Alternative Approach: Feynman’s Trick

We can treat the integral as a function of a and b, I(a, b), and

differentiate. Let’s differentiate with respect to b:
Z ∞  
∂I ∂ arctan(bx) − arctan(ax)
= dx
∂b 0 ∂b x
Z ∞  
1 x
= 2
dx
0 x 1 + (bx)
Z ∞  ∞
∂I 1 1
= 2
dx = arctan(bx)
∂b 0 1 + (bx) b 0

35
Integration Practice Notebook Problems & Techniques

1 π  πsgn(b)
= sgn(b) − 0 =
b 2 2b
Now we integrate this result with respect to b:
Z
πsgn(b) πsgn(b)
I(a, b) = db = ln |b| + C(a)
2b 2
By symmetry, the full function must be:
πsgn(b) πsgn(a)
I(a, b) = ln |b| − ln |a|
2 2
(This can be confirmed by differentiating with respect to a, which
gives − πsgn(a)
2a ). Given the constraint sgn(a) = sgn(b), we can set
sgn(a) = sgn(b) = s:
 
π πsgn(a) |b|
I(a, b) = s · (ln |b| − ln |a|) = ln
2 2 |a|
|b| b
Since |a| = a when they have the same sign, this matches the final
answer:  
πsgn(a) b
I(a, b) = ln
2 a

⟲ Practice Exercise

Practice Exercises

1. A Logarithmic Frullani Integral: Evaluate the following

using the Frullani theorem:
Z ∞
ln(1 + ax) − ln(1 + bx)
dx
0 x
Hint: Let f (x) = ln(1 + x). Find f (0) and f (∞). Note that
this integral does not have the a, b > 0 restriction. Answer:
1 2
2 ln (a/b).

2. An Exponential Frullani Integral: Evaluate the follow-

36
Integration Practice Notebook Problems & Techniques

ing:
∞
e−ax − e−bx
Z
dx (a, b > 0)
0 x
Hint: Let f (x) = e−x. Find f (0) and f (∞). Answer:
ln(b/a).

37
Integration Practice Notebook Problems & Techniques

Problem 112

▶ Problem

Evaluate the limit for x ∈ [0, 1]

∞  
X 1 1
lim p
− p
.
p→0+
n=1
(n − x) (n + x)

⊛ Solution

Strategy: The Fallacy of a Naive Limit

A first-time observer might be tempted to swap the limit and the

sum, a common technique for well-behaved functions. Let’s see
why that fails here:
∞  
X 1 1
”Naive attempt” = lim+ −
p→0 (n − x)p (n + x)p
n=1
∞
X ∞
X
= (1 − 1) = 0 = 0.
n=1 n=1
This is incorrect. The reason this fails is that this is a subtle
indeterminate form. As p → 0, each term in the sum goes to 0.
However, the number of terms in the sum is infinite, and the sum
itself is divergent at p = 0. This is a classic ”0 ×∞” scenario.
A more careful Taylor expansion reveals the problem: Using a−p =
e−p ln a ≈ 1 − p ln a for small p, the term in the sum is:
n + x
[1 − p ln(n − x)] − [1 − p ln(n + x)] = p ln .
n−x
P∞ n+x


The problem is now limp→0 p n=1 ln n−x . For large n,
+
n+x

2x P 2x
ln n−x ≈ n , and the sum n diverges.

38
Integration Practice Notebook Problems & Techniques

The true strategy is to recognize that the series represents a deeper

mathematical object, the Hurwitz Zeta function, which is
defined by this series but can be extended (analytically continued)
to other values of p. The problem is implicitly asking for the limit
of this extended function.
Step 1: Rewriting the Sum using the Hurwitz Zeta Function

The Hurwitz Zeta function is defined (for Re(s) > 1) by the series:
∞
X 1 1 1 1
ζ(s, a) = = + + + ...
(n + a)s as (1 + a)s (2 + a)s
n=0

Let’s rewrite our two sums to match this form.

1. The first sum:
∞
X 1 1 1
= + + ...
(n − x)p (1 − x)p (2 − x)p
n=1
∞
X 1
= = ζ(p, 1 − x).
(k + 1 − x)p
k=0

2. The second sum:

∞
X 1 1 1
= + + ...
(n + x)p (1 + x)p (2 + x)p
n=1
∞
X 1
= = ζ(p, 1 + x).
(k + 1 + x)p
k=0

Thus, our entire problem is to find the limit of a difference of two

Hurwitz Zeta functions:
I(x) = lim+ [ζ(p, 1 − x) − ζ(p, 1 + x)] .
p→0

Step 2: Using Analytic Continuation

The series definition for ζ(s, a) only converges for Re(s) > 1,
but the function itself has a unique analytic continuation that is

39
Integration Practice Notebook Problems & Techniques

defined for all s ̸= 1. Since the Hurwitz Zeta function is analytic

(well-defined and ”smooth”) at s = 0, we can bring the limit
inside and simply evaluate the functions at p = 0:
I(x) = ζ(0, 1 − x) − ζ(0, 1 + x).
This has transformed the problem from a difficult limit of a series
into a simple evaluation of a known special function.

Step 3: The Hurwitz Zeta Value at Zero

A fundamental identity in the theory of the Hurwitz Zeta function

(which can be derived from the Euler-Maclaurin formula or its
relationship with Bernoulli polynomials) gives its value at s = 0:
1
− a.
ζ(0, a) =
2
We can now apply this identity to our two terms.
1. ζ(0, 1 − x) = 12 − (1 − x) = 12 − 1 + x = x − 21 .
2. ζ(0, 1 + x) = 12 − (1 + x) = 12 − 1 − x = −x − 21 .

Step 4: The Final Calculation

We substitute these results back into our expression for I(x):

   
1 1 1 1
I(x) = x − − −x − = x − + x + = 2x.
2 2 2 2
The limit, which looked like it might be zero or diverge, is simply
2x.

40
Integration Practice Notebook Problems & Techniques

✓ Final Answer

∞  
X 1 1
lim − = 2x.
p→0+
n=1
(n − x)p (n + x)p

⋆ Key Insight

The Power of Analytic Continuation

This problem is a beautiful demonstration of the concept of ana-

P −p
lytic continuation. The original series n only converges for
p > 1. It is meaningless to just ”plug in” p = 0. However, the
P −p
function f (p) = n (the Riemann Zeta function, ζ(p)) can be
uniquely extended to a function that is valid for almost all complex
numbers p. The problem is asking for the value of this *extended*
function as p approaches 0.
The identity ζ(0, a) = 21 − a is a profound one. It comes from the
relationship between the Hurwitz Zeta function and the Bernoulli
polynomials, Bn(x):
Bn+1(a)
ζ(−n, a) = − for n ≥ 0.
n+1
For n = 0, we need B1(a). The first Bernoulli polynomial is
B1(a) = a − 12 . Plugging this in gives:
 
B1(a) 1 1
ζ(0, a) = − =− a− = − a.
1 2 2
This is the source of the identity that solves the problem.
Connection to the Riemann Zeta Function

What happens if we set x = 0? The problem becomes

P∞
limp→0+ n=1[ n1p − n1p ] = 0. Our formula gives 2x = 2(0) = 0.
This is consistent.

41
Integration Practice Notebook Problems & Techniques

What happens if we try to find the famous (and counter-intuitive)

value of ζ(0)? We can’t use our formula directly, but we can use
the identity ζ(s) = ζ(s, 1). Using the formula ζ(0, a) = 12 − a, we
can find:
1 1
ζ(0) = ζ(0, 1) = − 1 = − .
2 2
This is the famous, regularized value of 1 + 2 + 3 + 4 + · · · =
−1/12 in a different context. Our problem is deeply related to
these foundational concepts in number theory.

You can use a computational tool like WolframAlpha to verify these

claims.
• HurwitzZeta[0, a] will return ‘1/2 - a‘.
• HurwitzZeta[p, 1-x] - HurwitzZeta[p, 1+x]
• Limit[HurwitzZeta[p, 1-x] - HurwitzZeta[p, 1+x], p ->
0] will return ‘2x‘.
⟲ Practice Exercise

Test Your Understanding

1. A Specific Value. Using the result, what is the value of the

limit when x = 1/2? Answer: 2(1/2) = 1.
2. Check the x = 1/2 Case Manually. For x = 1/2, the
P∞ 1 1
sum is n=1[ (n−1/2) p − (n+1/2)p ]. Write out the first few terms

and show how they ”telescope” in a sense, leaving only the

first term as p → 0+. Hint: The sum is (1/0.5p − 1/1.5p) +
(1/1.5p − 1/2.5p) + . . . . As p → 0+, this sum is 1.
P∞
3. Riemann Zeta. What is limp→0+ n=1[ n1p − (n+1) 1
p ]? Hint:

Use the Hurwitz Zeta function identities. Answer: ζ(0, 1) −

ζ(0, 2) = (−1/2) − (1/2 − 2) = 1.
4. The First Bernoulli Number. Given ζ(0, a) = 12 − a,
what is the value of ζ(0, 1/4)? Answer: 1/2 − 1/4 = 1/4.

42
Integration Practice Notebook Problems & Techniques

Problem 113

▶ Problem

Evaluate the integral:

Z ∞
x−1
I= √ x x
dx
0 2 − 1 ln(2 − 1)

⊛ Solution

Strategy: Substitution, Splitting, and Cancellation

The solution proceeds by first applying a substitution to simplify

the integrand, transforming it from an exponential/logarithmic
form to an algebraic/logarithmic one. The resulting integral is
split into two intervals, [0, 1] and [1, ∞). A reciprocal substitution
(u = 1/ξ) is then applied to the second integral, which transforms
it into an integral over [0, 1]. When this transformed integral is
added back to the first integral, a remarkable cancellation occurs,
leaving a simple, elementary integral.
x
Step 1: Main Substitution Let ξ = 2 − 1. This implies:
• 2x = ξ + 1 =⇒ x ln 2 = ln(ξ + 1) =⇒ x = ln(ξ+1) ln 2
ln(ξ+1)−ln 2
• x−1= ln 2

• dx = 1
(ξ+1) ln 2 dξ

• Limits: x = 0 → ξ = 0; x = ∞ → ξ = ∞
Substituting these into the original integral:
Z ∞
(ln(ξ + 1) − ln 2)/ ln 2 1
I= √ · dξ
0 ξ · ln(ξ) (ξ + 1) ln 2
Z ∞
1 ln(1 + ξ) − ln 2
= 2 dξ
ln (2) 0 ξ 1/2(1 + ξ) ln ξ

43
Integration Practice Notebook Problems & Techniques

Step 2: Splitting the Integral We split the domain of integration at

ξ = 1:
 
Z ∞
1  1 ln(1 + ξ) − ln 2
Z ln(1 + ξ) − ln 2 

I= 2  dξ + dξ 
ln (2)  0 ξ 1/2(1 + ξ) ln ξ ξ 1/2 (1 + ξ) ln ξ
} |1

| {z {z }
I1 I2

We transform the
Step 3: The Reciprocal Substitution and Cancellation
second integral, I2, using the substitution ξ = 1/u, which means
dξ = −1/u2du. The limits become u = 1 to u = 0.
• Numerator: ln(1 + 1/u)−ln 2 = ln u −ln 2 = ln(u + 1)−
u+1

ln u − ln 2
• Denominator: (1/u)1/2(1 + 1/u) ln(1/u) =
u−1/2( u+1
u )(− ln u) = −u
−3/2
(u + 1) ln u
The transformed integral I2 is:
Z 0  
ln(u + 1) − ln u − ln 2 1
I2 = −3/2 (u + 1) ln u
− 2 du
−u u
Z1 1  
ln(u + 1) − ln u − ln 2 1
= −3/2 (u + 1) ln u 2
du
u u
Z0 1
ln(u + 1) − ln u − ln 2
= 1/2 (u + 1) ln u
du
0 u
Now, we add the two parts (switching the variable of I2 back to ξ
for clarity):
Z 1 Z 1
ln(1 + ξ) − ln 2 ln(ξ + 1) − ln ξ − ln 2
I1 + I2 = 1/2 (1 + ξ) ln ξ
dξ + 1/2 (ξ + 1) ln ξ
dξ
ξ ξ
Z0 1 0
(ln(1 + ξ) − ln 2) + (ln(ξ + 1) − ln ξ − ln 2)
= 1/2 (1 + ξ) ln ξ
dξ
ξ
Z0 1
2 ln(1 + ξ) − 2 ln 2 − ln ξ
= 1/2 (1 + ξ) ln ξ
dξ
0 ξ

44
Integration Practice Notebook Problems & Techniques

Let’s
 re-check the I2 transformation. ln(1 + 1/ξ) − ln 2 =
1+ξ

1+ξ
R 0 ln( 1+u )
ln 2ξ . Denominator: ξ 3/2 (− ln ξ). I2 = 1 − 1+u2uln u (− u12 )du =
R 1 ln(1+u)−ln 2−ln u 1 R 1 − ln(1+u)+ln 2+ln uu3/2
0 −(1+u) ln u u1/2
du I2 = 0 ξ 1/2 (1+ξ) ln ξ
dξ (switching u
to ξ) Now, adding I1:
Z 1
(ln(1 + ξ) − ln 2) + (− ln(1 + ξ) + ln 2 + ln ξ)
I1 + I2 = 1/2 (1 + ξ) ln ξ
dξ
ξ
Z0 1 Z 1
ln ξ 1
= 1/2 (1 + ξ) ln ξ
dξ = 1/2 (1 + ξ)
dξ
0 ξ 0 ξ
This is the correct cancellation. Sorry for the inconsistency, hu-
mans are prone to error :.)

Step 4: Final Integration We evaluate the resulting simple integral.

1/2 2
Let u = ξ , so ξ = u and dξ = 2udu.
Z 1 Z 1 Z 1
1 1 1
1/2 (1 + ξ)
dξ = 2)
(2u)du = 2 2
du
0 ξ 0 u(1 + u 0 1 + u
π  π
1
= 2 [arctan(u)]0 = 2 (arctan(1) − arctan(0)) = 2 −0 =
4 2
Step 5: Final Answer We combine this result with the pre-factor:
1 1 π 
I = 2 (I1 + I2) = 2
ln (2) ln (2) 2

✓ Final Answer

∞
x−1
Z
π
√ dx =
0 2x − 1 ln(2x − 1) 2 ln2(2)

45
Integration Practice Notebook Problems & Techniques

⋆ Key Insight

Key Insights: The Power of Reciprocal Substitution

This problem is a magnificent example of a ”hidden symmetry” in

an integral over [0, ∞).
• Reciprocal
R∞ RSymmetry:
R ∞ The technique of splitting an inte-
1
gral 0 into 0 and 1 is often paired with the substitution
u = 1/x on the second integral. This ”folds” the [1, ∞) inter-
val back onto [0, 1], allowing the two parts to be combined.
• Designed Cancellation: The integrand f (ξ) = ξln(1+ξ)−ln 2
1/2 (1+ξ) ln ξ

is not simple. However, the transformed integrand f (1/ξ)

(with the substitution’s Jacobian) has a part that is *exactly*
the negative of the original f (ξ), plus a new, simple term.
• General Form: This works for any integralR where f (x) +
1
f (1/x) · d(1/x)dx · dx
d(1/x) simplifies. Here, I2 = 0
f (1/u) u12 du.
R1
The total integral becomes 0 (f (u)+ u12 f (1/u))du. The clever
design of f (u) in this problem leads to f (u) + u12 f (1/u) =
ln u
u1/2 (1+u) ln u
. The complex numerator was engineered to cancel
perfectly, leaving only the ”symmetric” parts.

⇒ Alternative Approaches & Viewpoints

Alternative Approach: ξ = u2 Substitution First

A different, but related, path is to use a substitution that simplifies

the square root *before* the main substitution. Let 2x − 1 = u2.
ln(u2 +1)
• 2x = u2 + 1 =⇒ x ln 2 = ln u2 + 1 =⇒ x = ln 2

ln(u2 +1)−ln 2
• x−1= ln 2

• ln(2x − 1) = ln u2 = 2 ln u

• dx = 2u
(u2 +1) ln 2
du

46
Integration Practice Notebook Problems & Techniques

• Limits: x = 0 → u = 0; x = ∞ → u = ∞
Substituting these into the original integral:
Z ∞ 2


(ln u + 1 − ln 2)/ ln 2 2u
I= · 2 du
0 u · (2 ln u) (u + 1) ln 2
Z ∞ 2


1 ln u + 1 − ln 2
= 2 2 + 1) ln u
du
ln (2) 0 (u
R ∞ ln(u2+1)−ln 2
This new integral J = 0 (u2+1) ln u du can now be solved with
the same splitting and reciprocal substitution as the main solution.
R 1 ln(u2+1)−ln 2
• J1 = 0 (u2+1) ln u du
R ∞ ln(u2+1)−ln 2
• J2 = 1 (u2 +1) ln u
du
Let u = 1/v in J2:
0 2


ln 1/v + 1 − ln 2
Z  
1
J2 = 2 + 1) ln(1/v)
− 2 dv
1 (1/v v
 2 
Z 1 ln v +1 − ln 2  1 
v2
= v 2 +1 2
dv
0 ( 2 )(− ln v) v
Z 1 v2

ln v + 1 − 2 ln v − ln 2
= 2 + 1) ln v
dv
−(v
Z0 1 2


− ln v + 1 + 2 ln v + ln 2
= 2 + 1) ln v
dv
0 (v
Adding J1 and J2 (with v → u):
Z 1


(ln u2 + 1 − ln 2) + (− ln u2 + 1 + 2 ln u + ln 2)
J1 + J 2 = 2 + 1) ln u
du
(u
Z0 1 Z 1
2 ln u 1
= 2 + 1) ln u
du = 2 2
du = 2(π/4) = π/2
0 (u 0 1 + u
This confirms the result via a slightly different, but equally elegant,
path.

47
Integration Practice Notebook Problems & Techniques

⟲ Practice Exercise

Practice Exercises

1. A Related Arctan Integral: Evaluate the core integral

that resulted from the cancellation:
Z 1
1
√ dx
0 x(1 + x)
Hint: Use the substitution x = u2. Answer: π/2.
2. A ClassicR 1 Reciprocal
R∞ Integral: Evaluate the following
using the 0 + 1 split and a u = 1/x substitution.
Z ∞
ln x
2
dx
0 1 + x
R∞ R1
Hint: Show that the 1 part is the exact negative of the 0
part. Answer: 0.

48
Integration Practice Notebook Problems & Techniques

Problem 114

▶ Problem

Evaluate the integral

Z ∞
dx
I= √ 2.
0 (x + 1 + ⌊2 x⌋)

⊛ Solution

Strategy: Substitution and Discretization

√
The key to this integral is the floor function ⌊2 x⌋. The floor
function is piecewise constant, which suggests our strategy should
be to break the integral’s domain into pieces where the floor func-
tion is constant.
1. Substitution: We’ll use a substitution to simplify the ar-
gument of the floor function.
2. Discretization: We will split the integral over [0, ∞) into
an infinite sum of integrals over intervals where ⌊·⌋ is constant.
3. Summation: The resulting infinite series will be split into
two parts: one related to the Zeta function and one that is a
telescoping sum.
Step 1: The Substitution
√
Let’s make a substitution to simplify the ⌊2 x⌋ term. A natural
√
choice is to let the argument be a simple variable, t. Let t = 2 x.
2
Then t2 = 4x, so x = t4 . Differentiating gives dx = 2t4 dt = 2t dt.
The original integral transforms as follows (for t ≥ 0):
Z ∞ t Z ∞
2 dt t/2dt
I=
2 =
2 2 +4+4⌊t⌋ 2
t
 
4 + 1 + ⌊t⌋
0 0 t
4

49
Integration Practice Notebook Problems & Techniques

Z ∞ Z ∞
t/2dt 8tdt
= (t2 +4+4⌊t⌋)2
= .
0 0 (t2 + 4 + 4⌊t⌋)2
16

Step 2: Splitting the Integral (Discretization)

The integrand’s behavior is dictated by ⌊t⌋. We can split the entire

domain [0, ∞) into a sum of intervals [n, n+1) for n = 0, 1, 2, . . . .
On any such interval t ∈ [n, n + 1), the value of ⌊t⌋ is exactly n.
∞ Z n+1
X 8tdt
I= 2 + 4 + 4n)2
.
n (t
n=0

Step 3: Evaluating the Inner Integral

This integral is now straightforward to evaluate. We can use a

u-substitution. Let u = t2 + 4 + 4n. Then du = 2tdt. The
numerator 8tdt can be written as 4 · (2tdt) = 4du.
Z Z
8tdt 4du 4
= = − +C
(t2 + 4 + 4n)2 u2 u
4
=− 2 + C.
t + 4 + 4n
Now we evaluate this over the definite bounds [n, n + 1]:
Z n+1  n+1
8tdt 4
2 + 4 + 4n)2
= − 2
n (t t + 4 + 4n n
   
4 4
= − − − 2
(n + 1)2 + 4 + 4n n + 4 + 4n
 
1 1
=4 −
n2 + 4n + 4 n2 + 2n + 1 + 4 + 4n
 
1 1
=4 − .
(n + 2)2 n2 + 6n + 5
We can factor the second denominator:
 n2 +6n+5
 = (n+1)(n+5).
1 1
So, the inner integral is 4 (n+2)2 − (n+1)(n+5) .

50
Integration Practice Notebook Problems & Techniques

Step 4: Summing the Two Series

Our original integral I is now the sum of this expression from

n = 0 to ∞:
∞  
X 1 1
I=4 − .
(n + 2)2 (n + 1)(n + 5)
n=0

We can split this into two separate series, I = S1 − S2.

Part 1: The Zeta Series (S1 )

∞  
X 1 1 1 1
S1 = 4 =4 + + + ... .
(n + 2)2 22 32 42
n=0

This is 4 times the sum of the reciprocal squares, *missing* the

P∞ 1
1/12 term. We know the Basel problem  result:
 k=1 k 2 = ζ(2) =
π2
P∞ 1 1

π2 4π 2 2π 2
6 . So, S1 = 4 k=1 k 2 − 12
= 4 6 − 1 = 6 − 4 = 3 − 4.

Part 2: The Telescoping Sum (S2 )

∞
X 1
S2 = 4 .
(n + 1)(n + 5)
n=0
1 A B
We use partial fraction decomposition: (n+1)(n+5) = n+1 + n+5 .
This gives A = 1/4 and B = −1/4.
∞   X ∞  
X 1 1 1 1 1
S2 = 4 − = − .
4 n+1 n+5 n+1 n+5
n=0 n=0

This is a classic telescoping sum. Let’s write out the first few
terms:
 
1 1
n=0: −
1 5
 
1 1
n=1: −
2 6

51
Integration Practice Notebook Problems & Techniques

 
1 1
n=2: −
3 7
 
1 1
n=3: −
4 8
 
1 1
n=4: − ← (the 1/5 terms cancel)
5 9
 
1 1
n=5: − ← (the 1/6 terms cancel)
6 10
The negative terms only start cancelling the positive terms from
n = 4 onwards. The terms that are ”left over” (i.e., that never
get cancelled) are the first four positive terms:
1 1 1 12 + 6 + 4 + 3 25
S2 = 1 + + + = = .
2 3 4 12 12
Step 5: Final Calculation

We combine the results for S1 and S2:

 2 
2π 25
I = S1 − S2 = −4 − .
3 12
Putting everything over a common denominator:
2π 2 48 25 2π 2 73
I= − − = − .
3 12 12 3 12

✓ Final Answer

∞
2π 2 73
Z
dx
√ = − .
0 (x + 1 + ⌊2 x⌋)2 3 12

52
Integration Practice Notebook Problems & Techniques

⋆ Key Insight

Discretization as a Primary Tool

This integral is a perfect example of a ”continuous” problem that

is fundamentally ”discrete” in nature. The floor function, ⌊t⌋, is
a discretizing function. It forces us to abandon the smooth,
continuous methods of standard calculus (like Feynman’s trick or
complex analysis, which require analytic functions) and instead
adopt a discrete summation approach.
The strategy of ”discretization”—breaking a continuous domain
into a sum of discrete intervals—is a powerful technique. We’ve
effectively proven an identity between a continuous integral and
two infinite series:
Z ∞ ∞ ∞  
dx X 1 X 1 1
√ 2 =4 2
− − .
0 (x + 1 + ⌊2 x⌋) k n + 1 n + 5
k=2 n=0

This reveals the integral’s ”atomic structure,” showing it to be

composed of a ζ(2) term and a simple rational number.
Telescoping Sums: A ”Finite” Infinity

The second series, S2, is a beautiful example of a telescoping sum.

Although it’s an infinite sum, all but a finite number of terms
cancel out. The key was to notice the n + 1 vs. n + 5 shift, which
tells us that the first 4 terms (5 − 1 = 4) will be left behind. This
technique is a cornerstone of discrete mathematics and analysis.

53
Integration Practice Notebook Problems & Techniques

⟲ Practice Exercise

Test Your Understanding

Z ∞
⌊x⌋
1. A Simpler Floor Integral. Evaluate x
dx. Hint:
0 e
Use the same discretization strategy. Split the integral into
P∞ R n+1 n
n=0 n Pex dx. Evaluate the integral, which leaves a sum of
∞
the form n=0 n(e−n − e−n−1). This is a standard geometric
1
series. Answer: e−1 .
2. The Zeta Component. What is the exact value of
∞
X 1
? Hint: Relate it to the ζ(2) function. Answer:
(n + 3)2
n=1
π2 49
6 − 36 .

3. Another Telescoping Sum. Find the value of

∞
X 1  
1
− . Hint: Write out the first few terms. How
n n+3
n=1
many terms are ”left over” at the beginning? Answer: 11/6.

54
Integration Practice Notebook Problems & Techniques

Problem 115

▶ Problem

Evaluate the integral:

Z 1 2


ln 1 − x
I= cos(ln x)dx
0 x

⊛ Solution

Strategy: IBP, Complex Analysis, and Digamma Identities

The solution is a beautiful and advanced chain of transformations.

1. First, integration by parts is used, leveraging the fact that
cos(ln x)
x is a perfect derivative.
2. The resulting integral is expressed using complex numbers,
replacing sin(ln x) with Im[xi].
3. A substitution (u = x2) simplifies the integrand into a form
related to an integral representation of the Digamma function
Ψ(z).
4. The Digamma function’s recurrence and reflection formulas
are then used to systematically evaluate the expression, lead-
ing to the final constant.
d
We note that dx
Step 1: Integration by Parts [sin(ln x)] = cos(ln x)· x1 .
This allows us to use integration by parts (IBP) on the original
integral. Let u = ln 1 − x and dv = cos(ln x)
2


x dx. Then du =
−2x
1−x2
dx and v = sin(ln x).
Z 1  
1 −2x
I = ln 1 − x2 sin(ln x) 0 −


sin(ln x) 2
dx
0 1 − x

55
Integration Practice Notebook Problems & Techniques

The boundary term [uv]10 vanishes at both x = 0 and x = 1 (as

x ln x → 0 and (1 − x) ln(1 − x) → 0). This leaves the integral:
Z 1
2x sin(ln x)
I= 2
dx
0 1 − x
Step 2: Complex Representation and Substitution We use Euler’s for-
i ln x i
mula, sin(ln x) = Im[e ] = Im[x ].
Z 1 Z 1 i+1 
2x · Im[xi] 2x
I= 2
dx = Im 2
dx
0 1 − x 0 1 − x
Now, let u = x2. This means du = 2xdx, and xi+1 = (u1/2)i+1 =
u(i+1)/2.
Z 1 (i+1)/2  Z 1 (i+1)/2  Z 1 i/2 
u du u du u
I = Im = Im 1/2
= Im du
0 1 − u x 0 1 − u u 0 1 − u

Step 3: Introducing the Digamma Function We rewrite the integrand

to isolate a standard integral form.
Z 1 i/2  Z 1 i/2 Z 1 
(u − 1) + 1 u −1 1
I = Im du = Im du + du
0 1 − u 0 1 − u 0 1 − u
R1 1
The second integral, 0 1−u du, is divergent but it is a real-valued
divergence. Since we are only taking the imaginary part, this
term’s contribution is zero.
Z 1 i/2 
u −1
I = Im du
0 1 − u
This is a ”known result” derived from the integral representation
of the Digamma function.

56
Integration Practice Notebook Problems & Techniques

[Derivation of the Digamma Integral Identity]

Derivation 1: The Digamma Integral Representation

A standard integral representation for the Digamma function

Ψ(z) is: Z 1 z−1
t −1
Ψ(z) = −γ + dt
0 t − 1
Let z − 1 = α, so z = α + 1. The formula becomes:
Z 1 α Z 1 α
t −1 t −1
Ψ(α+1) = −γ+ dt =⇒ dt = − (Ψ(α + 1) + γ)
0 t − 1 0 1 − t
In our problem, α = i/2. Substituting this in:
I = Im [−(Ψ(1 + i/2) + γ)]
Since the Euler-Mascheroni constant γ is real, Im(γ) = 0.
I = − Im[Ψ(1 + i/2)]

Let Ψ(1 + i/2) = u + iv.

Step 4: Digamma Function Manipulations
We want to find −v. Using the property Ψ(z) = Ψ(z̄), we have
Ψ(1 − i/2) = Ψ(1 + i/2) = u − iv. Consider the expression
A = 2i (Ψ(1 + i/2) − Ψ(1 − i/2)):
i i
A = ((u + iv) − (u − iv)) = (2iv) = i2v = −v
2 2
This is exactly the value I that we are looking for. Thus,
i
I= [Ψ(1 + i/2) − Ψ(1 − i/2)]
2
Now we use the Digamma recurrence relation Ψ(z + 1) = Ψ(z) +
1
1/z. Let z = −i/2. Then Ψ(1 − i/2) = Ψ(−i/2) + −i/2 =
i
Ψ(−i/2) + 1/2 = Ψ(−i/2) + 2i. Substituting this into our expres-

57
Integration Practice Notebook Problems & Techniques

sion for I:
i
I = [Ψ(1 + i/2) − (Ψ(−i/2) + 2i)]
2
i i
= [Ψ(1 + i/2) − Ψ(−i/2)] − (2i)
2 2
i i
= [Ψ(1 + i/2) − Ψ(−i/2)] − i2 = 1 + [Ψ(1 + i/2) − Ψ(−i/2)]
2 2
Step 5: The Digamma Reflection Formula The final piece is the
Digamma reflection formula.

Derivation 2: The Digamma Reflection Formula

We start with Euler’s reflection formula for the Gamma func-

π
tion: Γ(z)Γ(1 − z) = sin(πz) . Taking the natural logarithm of
both sides:
ln Γ(z) + ln Γ(1 − z) = ln π − ln sin(πz)
Differentiating with respect to z:
Γ′(z) Γ′(1 − z) π cos(πz)
+ · (−1) = 0 −
Γ(z) Γ(1 − z) sin(πz)
Γ′ (z)
By definition, Ψ(z) = Γ(z) . So we have:
Ψ(z)−Ψ(1−z) = −π cot(πz) =⇒ Ψ(1−z)−Ψ(z) = π cot(πz)

We apply this formula with z = −i/2:

Ψ(1 − (−i/2)) − Ψ(−i/2) = π cot(−πi/2)
=⇒ Ψ(1 + i/2) − Ψ(−i/2) = π cot(−πi/2)
Substitute this back into our expression for I:
i
I = 1 + [π cot(−πi/2)]
2

58
Integration Practice Notebook Problems & Techniques

Using the identity cot(−iw) = i coth(w):

i i2 π π
I = 1 + [π · i coth(π/2)] = 1 + coth(π/2) = 1 − coth(π/2)
2 2 2

✓ Final Answer

1 2


ln 1 − x
Z
π π 
cos(ln x)dx = 1 − coth
0 x 2 2

⋆ Key Insight

Key Insights: A Symphony of Special Functions

This problem is a remarkable example of how different areas of

advanced calculus are deeply interconnected. The solution is a
chain of logical transformations:
• IBP as a Starting Gun: The initial integration by parts is
the key that unlocks the problem, transforming the integrand’s
trigonometric part (cos) into a different one (sin) and, crucially,
modifying the algebraic part (x → x/(1 − x2)).
• Complex Numbers as a Bridge: Using sin(ln x) =
Im[xi] is the bridge from real analysis to complex analysis.
This step allows the problem to be restated in terms of the
variable i, which is essential for the u = x2 substitution to
work so cleanly.
• Digamma R ui/2Function as the Engine: The resulting in-
tegral, 1−u du, is a classic form that is ”solved” by the
Digamma function Ψ(z). The problem is thus transformed
from one of ”finding a number” to one of ”manipulating a
special function.”

59
Integration Practice Notebook Problems & Techniques

i/2
• A ”Regularization” Step: The step Im 1−u u
R
du =
R ui/2−1
Im 1−u du is a subtleRbut critical form of regularization.
The divergent real part 1/(1 − u)du is discarded, leaving
the finite imaginary part, which is all we need.
• Identities as Solutions: The finale is a beautiful display of
the Digamma function’s properties. The answer is not found
by computing a value, but by showing that the application
of the recurrence and reflection formulas causes the complex-
valued Ψ functions to ”eat” each other, leaving behind a real-
valued expression involving π and coth.

⇒ Alternative Approaches & Viewpoints

Alternative Approach: Beta Function Derivative

A powerful, and arguably more direct, alternative is to use the Beta

function, B(z, w).
1. Start by replacing cos(ln x) with Re[xi].
Z 1
"Z
#
2 1 2
ln 1 − x i
ln 1 − x i
I= Re[x ]dx = Re x dx
0 x 0 x

2. Let u = x2, so du = 2xdx.

Z 1   Z 1 
ln(1 − u) i/2 du 1 i/2−1
I = Re 1/2
u 1/2
= Re ln(1 − u)u du
0 u 2u 2 0

3. This integral is a derivative of the Beta function B(z, w) =

R 1 z−1 w−1
0
t (1 − t) dt.
Z 1
∂B
(z, w) = tz−1 ln(1 − t)dt
∂w w=1 0
Our integral has z − 1 = i/2 − 1, so z = i/2.
 
1 ∂B
I = Re (i/2, w)
2 ∂w w=1

60
Integration Practice Notebook Problems & Techniques

4. Now we use the Gamma function representation: B(z, w) =

Γ(z)Γ(w)
Γ(z+w) .
 
∂B ∂ Γ(w)
= Γ(z)
∂w ∂w Γ(z + w)
 ′
Γ (w)Γ(z + w) − Γ(w)Γ′(z + w)

= Γ(z)
Γ(z + w)2

5. We evaluate at w = 1, using Γ(1) = 1 and Γ′(1) = −γ:

 ′
Γ (1)Γ(z + 1) − Γ(1)Γ′(z + 1)

∂B
= Γ(z)
∂w w=1 Γ(z + 1)2
 
(−γ)Γ(z + 1) − Ψ(z + 1)Γ(z + 1)
= Γ(z)
Γ(z + 1)2
Γ(z) 1
= [−γ − Ψ(z + 1)] = [−γ − Ψ(z + 1)]
Γ(z + 1) z

6. Now we substitute z = i/2 back into our expression for I:

 
1 1
I = Re [−γ − Ψ(1 + i/2)]
2 i/2
= Re [−i[−γ − Ψ(1 + i/2)]]
= Re [iγ + iΨ(1 + i/2)]

7. Let Ψ(1 + i/2) = u + iv.

I = Re [iγ + i(u + iv)] = Re [iγ + iu − v] = −v
This is I = − Im[Ψ(1 + i/2)], which confirms the result from
Step 3 of the main solution, and the rest of the derivation
follows.

61
Integration Practice Notebook Problems & Techniques

⟲ Practice Exercise

Practice Exercises

1. A Related (Real) Integral: Evaluate the counterpart to

this problem:
Z 1 2


ln 1 − x
J= sin(ln x)dx
0 x
Hint: Follow the IBP and complex substitution. You will end
up with Re[...] instead of Im[...]. Answer: − Re[Ψ(1 + i/2)].
(This does not have a simple closed form in terms of elementary
constants).
2. A Simpler Reflection Formula Test: Use the Digamma
reflection formula Ψ(1−z)−Ψ(z) = π cot(πz) to find the value
of:
Ψ(3/4) − Ψ(1/4)
Hint: Let z = 1/4. Answer: π.

62