Hang Lung Mathematics Awards

Vol. 5 (2012)

HANG LUNG MATHEMATICS AWARDS 2012

HONORABLE MENTION

Manipulating the Fermat’s Equation

Team members: Wing Man Chik, Ka Kit Ku, Ming Hong Lui,
Long Hin Sin
Teacher: Mr. Yan Ching Chan
School: Po Leung Kuk Centenary Li Shiu Chung
Memorial College
Hang Lung Mathematics Awards c 2012, IMS, CUHK
Vol. 5 (2012), pp. 259–274

MANIPULATING THE FERMAT’S EQUATION

TEAM MEMBERS
Wing Man Chik, Ka Kit Ku,
Ming Hong Lui, Long Hin Sin
TEACHER
Mr. Yan Ching Chan
SCHOOL
Po Leung Kuk Centenary Li Shiu Chung Memorial College

Abstract. In our report, we will manipulate the Fermat’s Equation by al-

lowing one of the exponents to be arbitrary. It turns out that if a prime
base is restricted, there are either no solutions or a unique primitive solution,
depending on the residue class that the prime belonging to modulo 4.

1. Motivation

Our problem is motivated from the celebrated Fermat’s Last Theorem. Solving the
equation xk + y k = z n is definitely a great challenge. However, if we restrict z to
be a prime number, then we may explicitly write down all the solutions by using no
more than Basic Valuation Theory and Number Theory. In our report, we prove
that the only solutions to xk + y k = pn are that p = 2, 3 or other primes p ≡ 1
(mod 4).

In this chapter, we will state the theorem we are investigating.

1.1. The Case p = 3

Theorem 1. For all positive integers n > 1, if there exists relatively prime positive
integers x, y, where x ≥ y and an integer k > 1 such that
xk + y k = 3n ,
then (x, y, k, n) = (2, 1, 3, 2).

Proof. Either both x, y are multiples of 3, which is rejected or 3 - xy. Therefore if k

is even, xk and y k are congruent to 1 mod 3 and their sum is congruent to 2 mod 3,
260 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

which is not a power of 3. If k is odd and k > 1, then 3n = (x + y)(xk−1 − xk−2 y +

· · · + y k−1 ). Thus x + y = 3m for some integers m ≥ 1. We will show that n ≥ 2m.
Moreover, since y ≡ −x (mod 3), xk−1 − xk−2 y + · · · + y k−1 ≡ xk−1 + · · · + xk−1 ≡
kxk−1 (mod 3), which yields 3 | k.

By putting x1 = xk/3 and y1 = y k/3 , we may assume that k = 3. Then x3 +y 3 = 3n

and x + y = 3m . To prove n ≥ 2m, it suffices to prove that x3 + y 3 ≥ (x + y)2
or x2 − xy + y 2 ≥ x + y. Since x ≥ y + 1, x2 − x = x(x − 1) ≥ xy, we have
(x2 − xy − x) + (y 2 − y) ≥ y(y − 1) ≥ 0 and n ≥ 2m is proved.

From the identity (x + y)3 − (x3 + y 3 ) = 3xy(x + y), it follows that

32m−1 − 3n−m−1 = xy
But 2m − 1 ≥ 1 and n − m − 1 ≥ n − 2m ≥ 0. If strict inequality holds in either
place in the last inequality, then 3n−m−1 contains factor 3 and 32m−1 − 3n−m−1 is
divisible by 3, but xy is not. Contradiction arises.

Therefore, m = 1 and n = 2. Substituting into the equation, (x, y, n, k) =

(2, 1, 2, 3).

After finishing the case p = 3, we want to further investigate the situation when p
is extended to all primes. So, let’s clearly state the statement of our thought.
Theorem 2. For an odd positive integer k, any positive integers x, y, n and a prime
p, where x ≥ y and k, n > 1 such that
xk + y k = p n ,
the only solutions are (x, y, p, k, n) = (2m , 2m , 2, k, mk + 1),(2(3t ), 3t , 3, 3, 2 + 3t),
where m is a positive integer and t is a non-negative integer.
Theorem 3. For an even positive integer k, any positive integers x, y, n and a
prime p, where x ≥ y and k, n > 1 such that
xk + y k = pn ,
it only has solutions when p = 2 or p ≡ 1 (mod 4). In particular, if p = 2,
(x, y, k, n) = (2m , 2m , k, mk + 1), where m is a positive integer. If p ≡ 1 (mod 4)
and (x, y) = 1, then x = |an |, y = |bn | with an , bn satisfying an + bn i = π n , where
π is a Gaussian prime dividing p and k = 2.

Our final goal is to verify these two theorems.

2. Essential Tools

In this chapter, we will briefly go over all the necessary knowledge and theorems
we need in solving the generalized equations. This involves the idea of quadratic
reciprocity and the p-adic valuation.
 MANIPULATING THE FERMAT’S EQUATION 261

2.1. Quadratic Residue

Definition 4. For all integers a such that (a, m) = 1, a is called a quadratic

residue modulo m if x2 ≡ a (mod m) has a solution. Otherwise, it is called a
quadratic nonresidue modulo m.
Example 5. 2 is a quadratic residue mod 7 as 32 ≡ 2 (mod 7), whereas 2 is a
nonresidue mod 5.
Definition
a 6 (Legengre’s Symbol). If p is an odd prime, then the Legengre Symbol
is defined as
p

 a   1, if a is a quadratic residue of p
= −1, if a is not a quadratic residue of p
p 
0, if p | a
2 2
Example 7. From Example 5, = 1 and = −1.
7 5

Here, we quote some of the propositions about Legengre’s Symbol. [See reviewer’s
comment (3)]
Proposition 8. If p is a prime, then
a
1) (Euler’s criterion) ≡ a(p−1)/2 (mod p).
p
 ab   a  b 
2) = .
p p p
a  b 
3) If a ≡ b (mod p), then = .
p p
 a2 
4) If (a, p) = 1, then = 1.
p

Proof. Refer to [1] Niven, Zuckerman.

Theorem 9. If p ≡ 3 (mod 4) and p | a2 + b2 , then p | a and p | b.

Proof. p | a2 + b2 implies a2 + b2 ≡ 0 (mod p). We have a2 ≡ −b2 (mod p) after

 −b2   −b2 
rearranging terms. We know that = 1 or 0 as a has a solution. If =
p p
 −1  b2   −1 
1, by Proposition 8, = 1. We find that = (−1)(p−1)/2 by
p p p
 −1 
Euler’s criterion. As p ≡ 3 (mod 4), (p − 1)/2 is odd and = −1. This
p
 −b2 
leads to a contradiction so = 0 and p | b.
p
As p | a2 + b2 with p | b, p | a2 and so p | a. We’re done.
262 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

2.2. p-adic Valuation

Definition 10. Let p be a prime number. For all positive integers n, there exists
a unique l such that n = pl m, where l, m are integers and p - m. Here we define
νp (n) = l.

The p-adic map is defined as n → νp (n) and it plays a very important role in solving
many number theory problems. Note that we define νp (0) = ∞.

There are some properties about the p-adic function, which are essential to solve
number theory problems.
Proposition 11. For non-negative integers a, b and p is a prime, [See reviewer’s
comment (4)]

1) νp (ab) = νp (a) + νp (b).

2) νp ((a, b)) = min(νp (a), νp (b)) and νp ([a, b]) = max(νp (a), νp (b)).
3) νp (a + b) ≥ min(νp (a), νp (b)), and equality holds if and only if νp (a) 6= νp (b).
Remark 12. (a, b) is the g.c.d. of a and b and [a, b] is the l.c.m. of a and b.

The first two are trivial results so we only show the third proposition.

Proof. Let a = pm k and b = pn l, where p - kl. When m = n, it becomes

a + b = pm k + pm l = pm (k + l)
p may be or may not be a factor of k + l, making that
νp (a + b) = νp (pm (k + l)) ≥ m = min(νp (a), νp (b))
We now prove the necessity part. When m < n (equivalent to νp (a) < νp (b)),
a + b = pm k + pn l = pm (k + pn−m l)
As p - k + pn−m l, we have
νp (a + b) = νp (pm (k + pn−m l)) = m = min(νp (a), νp (b))
For the suffiency part, we have
νp (a + b) = νp (pm k + pn l) = νp (pm (k + pn−m l)) = m + νp (k + pn−m l)
We know that min(νp (a), νp (b)) = m when m ≤ n, but equality is impossible as p
can be a factor of k + l, making that νp (a + b) 6= min(νp (a), νp (b)). Hence m < n
and νp (a) 6= νp (b). We’re done.

Here, we have one more definition.

a
Definition 13. For all non-negative integers a, b and b 6= 0, νp ( ) = νp (a) − νp (b).
b
 MANIPULATING THE FERMAT’S EQUATION 263

a a c
As rational numbers can be written as in many different ways, if = for some
b b d
c, d ∈ Z, then νp (a) − νp (b) = νp (c) − νp (d). Since ad = bc, by Proposition 11.1,
we have
νp (a) + νp (d) = νp (ad) = νp (bc) = νp (b) + νp (c)
Each rational number will output a unique value. As a result, we can extend the
function to νp : Q → Z ∪ {∞}.
Theorem 14 (Lifting Exponent Lemma). Let p be an odd prime and a, b are
integers such that p - ab and p | a − b. Then for all n ≥ 1, we have
νp (an − bn ) = νp (a − b) + νp (n)

Proof. Firstly, when νp (n) = 0, it means p - n. We want to show that νp (an − bn ) =

νp (a − b). Note that a ≡ b (mod p). We find out that an − bn = (a − b)(an−1 +
an−2 b + · · · + abn−2 + bn−1 ). Moreover,
(an−1 + an−2 b + · · · + abn−2 + bn−1 ) ≡ an−1 + an−1 + · · · an−1
= nan−1 (mod p)
a, n are not multiples of p, making that νp (an − bn ) = νp (a − b), as desired.

Secondly, we prove it for n = p. This time, we need to prove that p divides

an−1 + an−2 b + · · · + abn−2 + bn−1 exactly once. Let b = a + pk for some integers
k. By Binomial Theorem, we know that bi ≡ ai + ipkai−1 (mod p2 ), so

p−1
X p−1
X
ap − bp p − 1 p−2
= ap−1−i bi ≡ (ap−1 + ipkap−2 ) = pap−1 + p2 ka ≡ pap−1 (mod p2 )
a−b i=0 i=0
2

This shows that p divides an−1 + an−2 b + · · · + abn−2 + bn−1 exactly once.

Finally, to prove the general case νp (n) ≥ 1, we use induction.

Suppose the theorem holds when νp (n) = l.
Then for νp (n) = l+1, νp (n/p) = l. Then by the theorem, we have νp (an/p −bn/p ) =
νp (n/p) + νp (a − b). From the second case, together we have
νp (an − bn ) = νp (a(n/p)p − b(n/p)p ) = νp (p) + νp (a(n/p) − b(n/p) )
= 1 + νp (n/p) + νp (a − b)
= l + 1 + νp (a − b)
= νp (a − b) + νp (n)
By induction, we know the theorem holds.
Remark 15. Note that this theorem in NOT applicable to p = 2 as from the second
p−1
case of the proof, is not an integer. Therefore we have a spacial theorem for
2
the case p = 2.
264 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

Theorem 16 (Lifting Exponent Lemma for p = 2). Let x, y be odd integers and n
be an even positive integer. Then
x2 − y 2
ν2 (xn − y n ) = ν2 ( ) + ν2 (n)
2

Proof. Let n = 2k a for some odd integers a. Then

k−1 k−1
xn − y n = (xa − y a )(xa + y a )(x2a + y 2a ) · · · (x2 a
+ y2 a
)
2 2
We can observe that if u, v are odd integers, then u + v ≡ 2 (mod 4). Therefore,
ν2 (xn − y n ) = ν2 (x2a − y 2a ) + k − 1
So what is the relation between ν2 (x2a − y 2a ) and ν2 (x2 − y 2 )?

Actually, we found that

a−1 a−1
x2a − y 2a (xa + y a )(xa − y a ) X i a−1−i i
X
= = (−1) x y xa−1−i y i
x2 − y 2 (x + y)(x − y) i=0 i=0
All the terms are odd so we need to determine the number 2a
of terms.
−y 2a )
Both summation systems have a terms, which are odd, so (xx2 −y 2 is actually odd.
x2a − y 2a
ν2 ( )=0
x2 − y 2
and
ν2 (x2a − y 2a ) = ν2 (x2 − y 2 )
Finally,
ν2 (xn − y n ) = ν2 (x2 − y 2 ) + k − 1
x2 − y 2
= ν2 ( ) + ν2 (n)
2
and we’re done.

2.3. Gaussian Integers

Definition 17 (Gaussian Integers). The set of Gaussian Integers is defined to be

Z[i] = {a + bi | a, b ∈ Z}.
Definition 18 (Unit). An element α = a + bi ∈ Z[i] is a unit if there is another
element β = c + di ∈ Z[i] such that αβ = 1.

The units of Gaussian Integers are {1, −1, i, −i}.

So what is a prime under Z[i]? We have the following definition.
Definition 19 (Prime). For an element α ∈ Z[i], we call it a prime under Z[i] if
it cannot be written as α = βγ, where β, γ ∈ Z[i] and they are not units. If α is a
prime and it is also a prime in Z, we call it a rational prime.
 MANIPULATING THE FERMAT’S EQUATION 265

Definition 20 (Norm). The norm of α = a + bi is defined as N (α) = (a + bi)(a −

bi) = αᾱ = a2 + b2 . The norm is multiplicative, i.e. N (αβ) = N (α)N (β).

The arithmetic in Z is similar to that in Z[i], so all the theorem about division can
also be applied in Z[i]. Here, we quote one that is useful to us.
Theorem 21 (Unique Factorization Theorem). Every non-zero, non-unit α ∈ Z[i]
can be written as a unique product of primes up to order and multiplication of units.

With all these definitions, we can deduce a very essential theorem which will be at
utmost importance for us to obtain the solutions for the equation.
Theorem 22. There are infinite number of primes p that can be written into sum
of two squares if p ≡ 1 (mod 4). [See reviewer’s comment (5)]

Proof. Refer to [2] Niven, Zuckerman, Montgomery.

3. Success of Generalization

After going through all the necessary tools we will use, let’s see if these will help
us in attempting the general case.

3.1. The Case p = 3 by p-adic Valuation

Let’s recall what the theorem is.

Theorem 23. For all positive integers n, if there exist relatively prime positive
integers x, y, where x ≥ y and an integer k > 1 such that
xk + y k = 3n ,
then (x, y, k, n) = (2, 1, 3, 2).

Proof. We can observe that 3 | xk + y k . If k is even, by Theorem 9, 3 | x and

3 | y and contradict to the requirement of (x, y) = 1.

When k is odd, xk + y k = (x + y)(xk−1 − · · · + y k−1 ) = 3n , meaning x + y = 3m for

some m ≥ 1, or x + y = 1, which is absurd. [See reviewer’s comment (6)] By the
Lifting Exponent Lemma, we have
ν3 (xk + y k ) = ν3 (x + y) + ν3 (k)
Rearranging terms we have m = n − ν3 (k).

Case 1: m ≥ 2. We observe that 3a ≥ a+2 for a ≥ 1 since LHS grows exponentially

faster than RHS. Putting a = ν3 (k), we have ν3 (k) ≤ 3ν3 (k) − 2 ≤ k − 2.
266 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

Since x + y = 3m ≥ 9, we have M := max(x, y) > 3. Moreover,

x+y 3m
M≥ =
2 2
Hence,
3m k−1
xk + y k ≥ M k = M · M k−1 > ·3
2
3m k−1
> ·3
3
= 3m+k−2
≥ 3m+ν3 (k)
= 3n
It is a contradiction.

Case 2: m = 1. Then x + y = 3 and (x, y) = (1, 2) or (2, 1). Rearranging

the equation yields 3n − 2k = 1. By Catalan’s Theorem1 , we have n = 2 and
k = 3.

This method saves time to eliminate all possible values of k other than 3, unlike
the previous proof. It gives us a hope to tackle the generalized form by using p-adic
valuation.

3.2. The Case p = 2

We first look into the case where p = 2.

As x, y > 0, xk + y k ≥ 1 + 1 = 2, which means n ≥ 1. When n ≥ 1, 2n must be
even so x and y must be both odd or both even.

Case 1: x, y are odd.

Case 1.1: k is even.

Rearranging the terms, we have xk − y k = 2n − 2y k . Therefore ν2 (xk − y k ) =
ν2 (2n − 2y k ).
x2 − y 2
By Theorem 16, ν2 ( ) + ν2 (k) = ν2 (2(2n−1 − y k ).
2
Noticing 2n−1 − y k is odd, we have
ν2 (x2 − y 2 ) − 1 + ν2 (k) = 1
ν2 (k) = 2 − ν2 (x2 − y 2 )
= 2 − ν2 (x + y) − ν2 (x − y)
If x = y, it will become
xk + y k = 2xk = 2n
1 From [3] Mihǎilescu
 MANIPULATING THE FERMAT’S EQUATION 267

It has no solutions as LHS has odd factors where RHS is a perfect power of 2. [See
reviewer’s comment (7)]

Now suppose x 6= y. Since x, y are both odd, x + y and x − y are even. It comes
up with
ν2 (x + y), ν2 (x − y) ≥ 1
and we have
ν2 (k) ≤ 0
which is absurd.

Case 1.2: k is odd.

We can factorize the equation.
xk + y k = (x + y)(xk−1 − xk−2 y + · · · − xy k−2 + y k−1 ) = 2n
Recall x and y are odd, y ≡ −x (mod 2), so
xk−1 − xk−2 y + · · · − xy k−2 + y k−1 ≡ xk−1 + xk−1 + · · · + xk−1
= kxk−1 (mod 2).

Thus 2 - xk−1 − xk−2 y + · · · − xy k−2 + y k−1 .

Also, if x ≥ y > 1, then
xk−1 − xk−2 y + · · · − xy k−2 + y k−1
=xk−2 (x − y) + xk−4 (x − y) + · · · + xy k−3 (x − y) + y k−1
=(x − y)(xk−2 + xk−4 + · · · + xy k−3 ) + y k−1
>1
LHS consists of factors other than 2, but RHS is a perfect power of 2, which is a
contradiction.

Case 2: x, y are even

Let x = 2α a and y = 2β b, where a, b are odd positive integers and α, β > 0. Then
xk + y k = (2α a)k + (2β b)k = 2n
If α < β, the equation becomes
ak + 2(β−α)k bk = 2n−αk
We have LHS is odd but RHS is even. So it has no solutions.
Similarly, it has no solutions when α > β. When α = β, the equation becomes
ak + bk = 2n−αk
One trivial solution is a = b = 1 and n = αk + 1. In fact, by putting a = b = 2m ,
the general solution can be derived in the form of (x, y, k, n) = (2m , 2m , k, mk + 1),
where m is a positive integer. Note that the value of k can be arbitrary.
268 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

3.3. The Case for All Odd Primes

Then we consider the case for all odd primes, but before proceeding, we hope to
divide k into two cases.

3.3.1. When k is odd

After factorization, we have

xk + y k = (x + y)(xk−1 − xk−2 y + . . . − xy k−2 + y k−1 ) = pn
either p is a factor of x + y, or x + y = 1. The latter case is impossible so p must
divide x + y. Hence let x + y = pm .
Therefore, we may apply the Lifting Exponent Lemma.
νp (xk + y k ) = νp (x + y) + νp (k) = n
So we have m = n − νp (k).

Case 1: m ≥ 2.
We can see that pa ≥ a + 2 for p ≥ 3 and a ≥ 1 as LHS grows exponentially faster
that RHS. Substituting a = νp (k), we have νp (k) ≤ pνp (k) − 2 ≤ k − 2.

As x + y = pm ≥ p2 , let M : max(x, y) > p. Then we have

x+y pm
M≥ =
2 2
Hence,
pm k−1
xk + y k ≥ M k = M · M k−1 > ·p
2
m
p
> · pk−1
p
= pm+k−2
≥ pm+νp (k)
= pn
It is a contradiction.

Case 2: m = 1. In this case, we have 2 equations.


xk + y k = p n
x+y =p
Moreover, from the Lifting Exponent Lemma, we know that νp (k) = n − 1.
n−1 n−1
Therefore, xk + y k = xp l + y p l = pn , where p - l.
n−1 n−1 n−1 n−1
As xp + yp ≤ xp l + y p l for 1 ≤ l, we shall have to prove that either
n−1 n−1
xp or y p is greater than pn .
 MANIPULATING THE FERMAT’S EQUATION 269

l l
Case 2.1: n = 2. Then xp + y p = p2 .
We will prove the following inequality: xp > p2 for all p > 4 and x > 1.
Since xp ≥ 2p , it suffices to prove 2p > p2 .

For p = 5, 25 = 32 > 52 .
Assume the result holds when p = k, i.e. 2k > k 2 .
When p = k + 1, (k + 1)2 = k 2 + 2k + 1. As

k 2 − 2k − 1 = k 2 − 2k + 1 − 2 = (k − 1)2 − 2 > 0 ⇒ k 2 > 2k + 1

for k ≥ 5, we have

(k + 1)2 = k 2 + 2k + 1 < 2k 2 < 2(2k ) = 2k+1

By induction, we have 2p > p2 .

l l l
Since xp ≥ xp ≥ 2p > p2 , we have xp + y p ≥ xp + y p > p2 , and x, y, p, k have
no integral solutions. It doesn’t matter if the values of x and y are interchanged.
Since x + y = p with p > 4, either x or y must be greater than 2.

The case p = 3 is done in Theorem 1 and (x, y, k, p, n) = (1, 2, 3, 3, 2) or

(2, 1, 3, 3, 2). But to obtain a general solution, we can multiply both sides by 33
and have new sets of solutions.
So
(x, y, p, k, n) = (3t , 2(3t ), 3, 3, 2 + 3t), (2(3t ), 3t , 3, 3, 2 + 3t),

where t is a non-negative integer.

n−1
Case 2.2: n > 2. This time, we need to prove a stronger inequality: xp > pn
n−1 n−1 n−1
for x > 1 and p > 2. As xp ≥ 2p , we’ll check if 2p > pn .
2
p
When n = 3, LHS=2 . By Binomial Theorem, we have
 2  2  2 
2 2 p p p
2p = (1 + 1)p ≥ + + 2
1 2 p −2
= p4
> p3

Suppose it is true for some k ≥ 3. Then

k k−1
2p = (2p )p > (pk )p > p(pk ) = pk+1

We are done by induction.

n−1 n−1 n−1 n−1 n−1 n−1 n−1
As xp l ≥ xp ≥ 2p > pn , xp l + y p l
≥ xp + yp > pn , which
means x, y, p, k have no integral solutions.
270 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

3.3.2. When k is even

Case 1: p ≡ 3 (mod 4)
Let k = 2l, where l is a positive integer. The equation becomes
xk + y k = x2l + y 2l = (xl )2 + (y l )2 = pn
If such solution exists, we have p | ((xl )2 + (y l )2 ). By Theorem 9, we have p
divides both xl and y l , implying p | x and p | y.
Now let x = apc , y = bpd , where p - a, b.
The equation then becomes
(apc )k + (bpd )k = pn
Without loss of generality, assume d ≥ c. Dividing both sides by pck ,
ak + (bpd−c )k = pn−ck
We can see that d = c is a must, or else p - LHS. Then the equation can be reduced
to ak + bk = pn−ck .
By Theorem 9 again, we know that a, b have a solution only if p | a and p | b,
which is a contradiction here.
Therefore, there are no solutions for all p ≡ 3 (mod 4) and p > 3.

Case 2: p ≡ 1 (mod 4)

Case 2.1: When k = 6 2m , where m is a positive integers.

Let k = 2 h, where g is a positive integer and 2 - h. The equation becomes
g

g g
x2 h
+ y2 h
= pn
The equation is reduced to a new form where h becomes the corresponding k of the
original equation.
We can observe that the new equation can fit into the case where k is odd.
g g
(x2 , y 2 , p, h, n) = (2, 2m , 2m , k, mk + 1), (3, 3m , 2(3m ), 3, 2 + 3m) and
(3, 2(3m ), 3m , 3, 2 + 3m).
g g
There are no integral solutions for (x2 , y 2 ) = (2m , 2m ) or (3m , 2(3m )) or
(2(3m ), 3m ). Therefore the equation has no integral solutions when k 6= 2m , where
m is a positive integer.

Case 2.2: When k = 2.

Since p ≡ 1 (mod 4), we can factorize p = ππ̄, where π, π̄ are Gaussian integers. In
LHS, x2 + y 2 = (x + yi)(x − yi). In general, we have
(x + yi)(x − yi) = π n π̄ n
Here, we can put x + yi = π n and x − yi = π̄ n . Suppose π = a + bi, where a, b
are integers. Then we can get an + bn i by expanding π n . Similarly, x − yi = π̄ n =
π¯n = an − bn i. Hence we can deduce x = Re(π n ) = |an | and y = Im(π n ) = |bn |.
The key to find an and bn is to identify a, b from the beginning first. After that it
is all left to find them by brute force.
 MANIPULATING THE FERMAT’S EQUATION 271

Of course, you may not choose x, y in this way. Let x + yi = π n−1 π̄ and x − yi =
¯ . Then
π n−1
π n−1 π̄ = π n−2 ππ̄ = pπ n−2
You will see that x, y both contain factor p after expanding pπ n−2 . So we have
pθk ck + pθk dk = pn ,
for some c, d, θ. But dividing both sides by pθk , then we will have
ck + dk = pn−θk
Actually this is a set of solutions comes from the first method of choosing the
factors.

Case 2.3: When k = 2m , where m ≥ 1.

This time we factorize the LHS in a different way.
m m m−1 m−1
x2 + y 2 = x(2 )2
+ y (2 )2
= pn
The intermediate steps are the same but the only difference is to compare
m−1
x2 = Re(π n ) = |an |
and
m−1
y2 = Im(π n ) = |bn |
To have x, y to be integers, an and bn must be 2m−1 -th powers.

After investigation, we can conclude that there are solutions for the equation only
when p = 2, p = 3 and p ≡ 1 (mod 4). In particular, when p = 2, (x, y, k, n) =
(2m , 2m , k, mk + 1), where m is a positive integer.
When p = 3, (x, y, k, n) = (2(3t ), 3t , 3, 2 + 3t), where t is a non-negative integer.
When p ≡ 1 (mod 4), if (x, y) = 1, then (x, y, k, n) = (|α|, |β|, 2, n), where α, β
satisfy α + βi = π n , with π being a Gaussian prime dividing p.
If (x, y) 6= 1, then (x, y) = (pm α, pm β), where α, β satisfy α + βi = π n−m , with
π being a Gaussian prime dividing p and m < n. Based on these results, the two
theorems in Chapter 1 are proved.

4. Beyond

4.1. Obstacles for general case: Composite numbers z

After the generalization, we want to investigate further. We then modify our aspect.
Instead of a prime number p, we want to find out what happens when a natural
number z is replaced, i.e.
xk + y k = z n
However, the proof of generalization does not work with composite z. Let’s see
which part fails.
272 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

4.1.1. Attempt by using the same approach

Suppose we divide k into two case: k is odd and k is even.

If k is odd, then we factorize xk + y k = (x + y)(xk−1 − xk−2 y · · · + y k−1 ) = z n .

Since z consists of primes, there will be many different combinations for the values
in both terms in LHS. This means solving the equation xk + y k = z n by means
of Elementary Number Theory will be abysmally hopeless. In particular, the case
n = k was solved by Prof. Andrew Wiles in 1995 using advanced theory of Elliptic
Curve and Modular Form.

Yet, in here, we approach a further generalization in the following ways.

1. Fixing the index k.

2. Fixing the base value z.

4.1.2. The Case z = 14, k = 3

Let’s see if we can solve this equation.

x3 + y 3 = 14n

Factorizing both sides yields

(x + y)(x2 − xy + y 2 ) = 2n 7n

Supposes x + y = 2α 7β and x2 − xy + y 2 = 2n−α 7n−β , where α, β are non-negative

integers. From the first equation, we have x = 2α 7β − y. Put this in the latter
equation, we have [See reviewer’s comment (8)]

x2 − xy + y 2 = y 2 − y(2α 7β − y) + (2α 7β − y)2 = 2n−α 7n−β

Rearranging yields 3y 2 −3(2α 7β )y+22α 72β −2n−α 7n−β = 0. It becomes a quadratic

equation in y. We then try to find the discriminant ∆.

∆ = 9(22α 72β ) − 12(22α 72β − 2n−α 7n−β )

= −3(2α 7β )2 + 12(2n−α 7n−β )

Substitution yields ∆ = −3(x + y)2 + 12(x2 − xy + y 2 ) = [3(x − y)]2 . We can see

that the discriminant is a perfect square, which means that integer solutions exist.
Using the quadratic formula, we have

3(2α 7β ) ± 3(x − y)
y=
6
6y = 3(x + y) ± 3(x − y)
 MANIPULATING THE FERMAT’S EQUATION 273

If 6y = 3(x + y) − 3(x − y), it yields 0 = 0, which is useless for us. Therefore

6y = 3(x + y) + 3(x − y) is our only hope. Yet
6y = 3(x + y) + 3(x + y − 2y)
12y = 2 · 3(x + y) = 3(2α+1 7β )
y = 2α−1 7β
Substituting our result into x + y = 2α 7β , we have x = y = 2α−1 7β . However, if we
put this equation back into the beginning, we have 23α−2 73β = 2n 7n . Contradiction
arises when we compare the indices taking modulo 3. Therefore there are no integral
solutions for x3 + y 3 = 14n for all x, y, n.

4.2. Summary

In Chapter 3 of this report, we successfully find out that for equation xk + y k = pn ,

there exist solutions for primes 2, 3 and other primes p ≡ 1 (mod 4).
Let’s recall what the solutions are for p = 2, 3 (for the case p ≡ 1 (mod 4), the
existence is proved but the form of solutions cannot be explicitly written).
If p = 2, (x, y, k, n) = (2m , 2m , k, mk + 1), where m is a positive integer.
If p = 3, (x, y, k, n) = (3t , 2(3t ), 3, 2 + 3t), (2(3t ), 3t , 3, 2 + 3t), where t is a non-
negative integer. (If x ≥ y, then the set of solutions (3t , 2(3t ), 3, 2 + 3t) is rejected)

Then in Chapter 4, we try to extend prime numbers p into all natural numbers z,
i.e. xk + y k = z n .
Our theorem is not applicable to composite numbers so we try to attack it case by
case. The new equation is way far beyond the Fermat’s equation. We are not able
to solve it generally but to tackle it by fixing different indices k and the base value
z.

REFERENCES

[1] Amir H. Pavardi, Lifting Exponent Lemma,

http://s3.amazonaws.com/aops-cdn.artofproblemsolving.com/resources/articles/lifting-the-
exponent.pdf
[2] Ivan Niven, Herbert S. Zuckerman and Hugh L. Montgomery, An Introduction to the Theory
of Numbers, 5th ed., Courier Companies. Inc., 1991.
[3] Kenneth Ireland, Michael Rosen, A Classical Introduction to Modern Number Theory,
Springer-Verlag, New York, 2010.
[4] Preda Mihǎilescu, Primary cyclotomic units and a proof of Catalan’s conjecture, J. Reine
Angew. Math. 572 (2004), 167-195.
[5] Titu Andreescu, Dorin Andrica and Ion Cucurezeanu, An Introduction to Diophantine Equa-
tions, Birkhäuser Boston, 2010.
274 W.M. CHIK, K.K. KU, M.H. LUI, L.H. SIN

Reviewer’s Comments
The presentation of the paper is good. The following is an incomplete list of cor-
rections and stylistic suggestions.

1. The reviewer has comments on the wordings, which have been amended in
this paper.
2. Punctuation marks are missing in most of the formulas in this paper.
3. “of the propositions” should be rewritten as “results”.
4. “p is a prime” should be rewritten as “any prime”.
5. “number of” should be deleted.
6. “meaning” should be rewritten as “implying”.
7. “where” should be rewritten as “while”.
8. “Put” should be rewritten as “Substituting”.