MATH 1024 – Honors Calculus II

Homework #07 Section 3.Ω

Due date: April 23, 2023

Here comes a long HW for you to enjoy over the Easter Holiday
Graded Problems:
1. (30 points) In this question, we calculate the Gaussian integral.
The usual way is to apply multivariable calculus techniques (MATH2023) by
using a double integral with polar coordinates, which is outside the scope of
this course. So instead we present a single-variable proof.
(a) (i) Show that for any positive integer n:
Z 1 Z 1 Z 1
2 n −nx2 dx
(1 − x ) dx ≤ e dx ≤ .
0 0 0 (1 + x2 )n

Z π
2
(ii) Let In = cosn θdθ. Show by change of variables that
0
√
√
Z n √
2
nI2n+1 ≤ e−x dx ≤ nI2n−2 .
0

(b) Recall from Lecture that


 (n − 1)!! π , if n is even,

In = n!! 2
 (n − 1)!! ,

if n is odd.
n!!
In
Show that lim = 1, and hence prove the Wallis formula
n→∞ In−1
π
lim nIn2 = .
n→∞ 2

(c) Hence conclude that √

Z ∞
−x2 π
e dx =
0 2
as well as the more familiar
Z ∞ √
2
e−x dx = π.
−∞
2. (30 points) In this question, we prove the Stirling’s formula for approximat-
ing n!:
√  n n
n! ∼ 2πn .
e
(a) Let
n!
an = √ n −n , bn = log an .
nn e
Show that (compare functions by differentiating twice)
1
0 < bn − bn+1 < .
12n2
Conclude by MCT that bn converges, and hence an converges.

(b) Using (a) and the Wallis formula, prove Stirling’s formula, in the sense
that
n!
lim √ = 1.
n→∞ 2πnnn e−n
√
n
n!
Using Stirling’s formula, we can easily get lim = e−1 as in HW6.
n→∞ n

(c) Derive the asymptotic formula for the double factorial:

 √ n n
 πn( e ) if all n are even,
 2

n!! ∼
√
 2n( n ) n2 if all n are odd.


e

3. (30 points) In this question we study the Gamma function Γ(x) defined by
Z ∞
Γ(x) = tx−1 e−t dt, x > 0.
0

(a) (i) Show that the integral converges for x > 0.

(ii) Show that
xΓ(x) = Γ(x + 1), x>0
and hence conclude that for integer n ≥ 0,

1 √ (2n)!
Γ(n + 1) = n! and Γ(n + ) = π n .
2 4 n!

Page 2
 (b) Define the Beta function by
Z 1
B(a, b) = ta−1 (1 − t)b−1 dt, a, b > 0
0

and for each integer n ≥ 1, let

Z n  n
x−1 t
In (x) := t 1− dt, x > 0.
0 n

(i) Show that

In (x) = nx B(n + 1, x).

(ii) Hence using the reduction formula from Worksheet #05, show that
n−1
n!nx 1 n Y (1 + k1 )x
In (x) = = .
x(x + 1) · · · (x + n) xx+n 1 + xk
k=1

(c) (i) Show that for integer n ≥ 1 and any 0 ≤ t ≤ n,

n
t2
  
t
1− e−t ≤ 1 − ≤ e−t .
n n

(ii) Conclude that for x > 0

Z n
1 n x+1 −t
Z
Γ(x + 2)
tx−1 e−t dt − In (x) ≤ t e dt <
0 n 0 n

and hence
∞
1 Y (1 + n1 )x
Γ(x) = lim In (x) = .
n→∞ x n=1 1 + nx

This is known as Euler’s product representation of Γ(x),

Since the right hand side actually converges for all real numbers x ̸= 0, −1, −2, ...
(see HW8), this representation extends the domain of Γ(x) to all x ̸= 0, −1, −2, ....
4. (30 points) The Euler’s product representation for sin πx:
∞ 
x2
Y 
sin πx = πx 1− 2 ,
n=1
n

is usually proved using the Weierstrass Factorization Theorem in complex

analysis. Here we give an elementary proof with real analysis only.

Page 3
(a) (i) Show that for non-negative integer n ≥ 0,
Z π2
1
cos2n x dx ≥ √
0 2n +1
using Q1(b) and MATH1023 HW1.
(ii) By MVT we have | cos at−1| ≤ |at|, for any a ∈ R. Using this together
with (a)(i), conclude that
Z π2
cos2n t cos at dt
0
lim Z π2 = 1.
n→∞
2n
cos t dt
0 Z π2
(b) (i) For any x ∈ R\Z and non-negative integer n ≥ 0, let In = cos2n t cos 2xt dt.
0
Prove the reduction formula
−1
x2

2n − 1
In = 1− 2 In−1 , n > 0.
2n n

(ii) Using (a)(ii), conclude that

∞ 
x2
Y 
sin πx = πx 1− 2
n=1
n

for all x ∈ R.
(c) Recall the Euler–Mascheroni’s constant defined by
n
!
X 1
γ = lim − log n ≈ 0.577216...
n→∞ k
k=1

which exists by Monotone Convergence Theorem, see Exercise 1.7.15.

(i) Derive the Weierstrass product representation for Γ(x)
∞
e−γx Y x  x −1
Γ(x) = en 1 + , x ̸= 0, −1, −2, ...
x n=1 n

(ii) Show that

xΓ(x) = Γ(x + 1), x ̸= 0, −1, −2, ...
still holds using the extended definition Γ(x) = lim In (x).
n→∞

Hence also derive the reflection formula for the Gamma function
for any x ∈
/ Z:
π
Γ(x)Γ(1 − x) = .
sin πx

Page 4
Optional Problems.
Let R(x, y) be a rational function in x, y. We know that
Z p
R(x, ax2 + bx + c)dx, a, b, c ∈ R

can always be integrated in closed form in terms of elementary functions, e.g. by

trigonometric substitution or Euler’s substitution (HW6) which turns the integral
into that of a rational function.
The situation for the square root of cubic (degree 3) or quartic (degree 4) is more
unfortunate, and in general it is a non-elementary integral. However, it is known
that in this situation (cubic when a4 = 0), the integral
Z p
R(x, a4 x4 + a3 x3 + a2 x2 + a1 x + a0 )dx ai ∈ R

can always be expressed in terms of elementary functions and a family of special

functions known as the Legendre forms of Elliptic Integrals. There are three
of them.
Let 0 < k < 1.
ˆ First kind:
Z ϕ Z x
dθ dt
F (ϕ, k) := p =(t=sin θ) p =: F (x; k)
0 1 − k 2 sin2 θ 0 (1 − t2 )(1 − k 2 t2 )

This is the original elliptic integral, which calculates the arc length of an
x2 y2
q
2
ellipse 2 + 2 = 1 for 0 < a < b with eccentricity k = 1 − ab2 .
a b
Note that F (x; k) is strictly increasing on [0, 1]. The inverse function is known
as the Jacobian elliptic function sn(u, k).
ˆ Second kind:
ϕ x
√
1 − k 2 t2
Z p Z
2
E(ϕ, k) := 1− k2 sin θdθ =(t=sin θ) √ dt =: E(x; k)
0 0 1 − t2

ˆ Third kind:
Z ϕ
2 dθ
Π(ϕ, α , k) := p
0 (1 − sin θ) 1 − k 2 sin2 θ
α2 2
Z x
dt
= p =: Π(x; α2 , k)
0 (1 − α2 t2 ) (1 − t2 )(1 − k 2 t2 )

where α2 ∈ C \ {1, k 2 } is complex.

Page 5
 √
Let y = a4 x4 + a3 x3 + a2 x2 + a1 x + a0 where ai ∈ R and assume y 2 = 0 has no
multiple roots (which reduces to the linear or quadratic case.)
(a) Show that we can always rationalize the integration such that
Z Z Z
R1 (x)
R(x, y)dx = R0 (x)dx + dx
y
for some rational function R0 (x) and R1 (x). In particular we know how to
integrate the first term, so we only need to focus on the second term.
(b) Fix some constant c ∈ R and assume
y 2 = a4 x4 + a3 x3 + a2 x2 + a1 x + a0
= b4 (x − c)4 + b3 (x − c)3 + b2 (x − c)2 + b1 (x − c) + b0 .
Let
xs
Z Z
dx
Is = dx, Js = .
y (x − c)s y
Show that for s = 0, 1, ..., we have the reduction formulas
a3 a1
xs y = (s + 2)a4 Is+3 + (2s + 3)Is+2 + a2 (s + 1)Is+1 + (2s + 1)Is + sa0 Is−1
2 2
y b3 b1
= (3 − s)b4 Js−4 + (5 − 2s)Js−3 + b2 (2 − s)Js−2 + (3 − 2s)Js−1 − (s − 1)b0 Js .
(x − c)s−1 2 2
(Hint: You don’t need to do any integration to prove these formulas!)
Therefore by taking full partial fraction of R1 (x) over C (so that all factors
are linear) we are reduced to study only the 4 “basic integrals”
Z Z Z 2 Z
dx xdx x dx dx
, , , .
y y y (x − c)y
Formally here we extend our integrals to include c ∈ C (so that bi ∈ C too),
and indefinite integration is linear over the complex field.
p
(c) There are 21 formulas that turns y into the normal form z = (1 − t2 )(1 − k 2 t2 )
where 0 < k < 1, depending on the range of integrations and the roots of
y 2 = 0. Here we just investigate two of them.
Consider
(i) y 2 = 0 is cubic with 3 real roots
y 2 = (x − α)(x − β)(x − γ), α<β<γ
q
x−γ
the substitution t = x−β for x > γ;

(ii) y 2 = 0 is quartic with 4 real roots

y 2 = (x − α)(x − β)(x − γ)(x − δ), α<β<γ<δ
q
the substitution t = (γ−α)(x−δ)
(δ−α)(x−γ) for x > δ.

Page 6
 (Note: the indicated range for x above requires the leading coefficient of y 2 > 0
so we can just factor it out and ignore it.)
Show that the substitution above turns the integration into normal form
Z Z
R1 (x) R2 (t)
dx = dt
y z

for some other rational function R2 (t).

Therefore by (b), we can assume y is in normal form for the 4 “basic integrals”.
(d) Evaluate the four basic integrals in terms of Legendre forms of elliptic
integrals
p (using indefinite integral and ignoring integration constant) where
y = (1 − x2 )(1 − k 2 x2 ) is in normal form.
Z
dx
ˆ is just F (x; k).
y
Z
x
ˆ dx can be evaluated in closed form by substitution.
y
Z 2
x
ˆ dx in terms of E(x; k) and F (x; k).
y
Z
dx
ˆ in terms of Π(x; c12 , k).
(x − c)y
Z
Hence we have completely described the integration R(x, y)dx where y 2 is cubic
or quartic, in terms of the elliptic integrals.
Note. The use of complex numbers is sometimes necessary, as in Cardano’s formula
for the roots of cubic or quartic polynomials. However, when a definite integral is
evaluated, the imaginary part will become zero since c and its complex conjugate c
above will always come in pairs.

Page 7