You are on page 1of 10

ALGEBRAIC NUMBER THEORY LECTURE NOTES

ABHINAV NEGI

Abstract. This is a short introductory text to help the reader get introduced to “Algebraic
Number Theory”. It assumes a fair amount of knowledge in calculus, and in basic number
theory. The text includes the lecture notes from “Algebraic Number Theory Course” by Prof.
Paul Pollack in Ross Mathematical Program 2020.

1. Introduction
Algebraic Number Theory, as its name suggests can be viewed in two ways;
“Number Theory using Algebra.”
“Theory of Algebraic Numbers.”
We can think of it through ways. But what is an Algebraic Number?

2. Algebraic Numbers
2.1. Analytic Questions. So, how do we quantitatively analyse arithmetic objects. One way
to to ask “How many such number exist?”, “Another way to ask what portion of such number
satisfy this property?”. These are all interesting questions. Lets answer them.
(1) How many squares are there?
The answer is countable infinite. But this isn’t a hard question to answer, nor does it
provide us with any new insight into the squares.
(2) How many even integers are there? Countably infinite.
Again, the answer is countable infinite. This question is still very easy to answer.
(3) What proportion of {1, 2, . . . , 10} are even?
This is a better question that tells us about the distribution of the even integers in the
integers. The answer is 12 , since odd and even integers alternate in the set {1, 2, . . . , 10},
so exactly half are even.
(4) What proportion of {1, 2, . . . , n} are even?
The answer is always either 12 or a little bit less. When n is even, the number of even
and odd numbers is the same, so the answer is 12 . When n is odd, the number of odd
numbers is one more than the number of evens. So let n = 2k + 1, there are k evens in
k
the set. So the proportion of the set that is even is 2k+1 = n−1
2n . So our answer is;
(
1
2 if 2 | n
n−1
2n if 2 ∤ n
(5) What proportion of positive integers are evens?
Now, our idea is to let δ(n) denote the proportion of the elements of {1, 2, . . . , n} that
are even (
1
if 2 | n
δ(n) = 2n−1
2n if 2 ∤ n
1
2 ABHINAV NEGI

and now let n → ∞. We get;


1
lim δ(n) = .
n→∞ 2
So, we can say that half of all positive integers are even. Or the density of evens in
the positive integers is 12 .
We have developed a the idea of density now.
Definition 2.1 (Density). Let A ⊂ Z+ . We define the density of A as;
#A ∩ [1, x]
lim .
x→∞ x
We may also define the Relative Density of a subset of of a set of positive integers as follows;
Definition 2.2 (Relative Density). Let A ⊂ Z+ and B ⊂ A. We define the relative density of B
in A as;
#B ∩ [1, x]
lim .
x→∞ #A ∩ [1, x]

2.2. Problems. Here are some problems related to the concept of density that one may attempt.
These will prepare you for the harder problems we will discuss in the next section.
(1) What is the density of the set of all squares? Cubes? Generalize.
(2) What is the density of the set of all positive integers that can be expressed in the form
n2 + n + 1?
(3) Let integers a, b with b ̸= 0 be given. What is the density of the set of all numbers that
are a (mod b)?
(4) Let a be a given integer. What is the density of the set of all positive integers divisible
by a? What is the density of the set of all positive integers coprime to a?
(5) What is the density of the set of prime numbers? (You may use the Prime Number
Theorem)
2.3. Solutions.
Problem 1. What is the density of the set of all squares? Cubes? Generalize.
Solution. Let A be the set of all positive perfect squares. The density of A is given by
#A ∩ [1, x]
lim .
x→∞ x
√ #A ∩ [1, x] is the number of positive perfect squares not greater than x. This is equal to
But
⌊ x⌋. So, the density is √
⌊ x⌋
lim .

x→∞ x
√ √
⌊ x⌋
But x−1x ≤ x ≤ xx . So, by the Squeeze Theorem, we get
√ √ √
x−1 ⌊ x⌋ x
lim ≤ lim ≤ lim .
x→∞ x x→∞ x x→∞ x
Now, we will show that both limits on the Left and Right side are 0. We have;
√ √ √
x−1 x 1 x 1
lim = lim − lim lim = lim √
x→∞ x x→∞ x x→∞ x x→∞ x x→∞ x
1 1
= lim √ − lim = 0.
x→∞ x x→∞ x
= 0,
ALGEBRAIC NUMBER THEORY LECTURE NOTES 3

So, clearly the density of the set of all positive squares is 0. Similarly, the density of the set of all
positive cubes is j 1k
x3
lim .
x→∞ x
We can show this limit is 0 using a similar argument. In general, the density of the set of all
positive nth powers is j k 1
xn
lim = 0.
x→∞ x

Problem 2. What is the density of the set of all positive integers that can be expressed in the
form n2 + n + 1?
Solution. Using an argument similar to before, we can show that the number of such positive
integers in the interval [1, x] for x > 1 is
⌊largest positive root of α2 + α + 1 = x⌋.
Using the quadratic formula, the largest positive root of α2 + α + 1 = x is
p √
−1 + 12 − 4(1)(1 − x) −1 + 4x − 3
α= = .
2 2
So the density is j √
k
−1+ 4x−3
2
lim .

x→∞ x
4x
We can bound this above by 2x = √1x . Since x ≥ 1, we get

√ √
j k
−1+ 4x−3
4x − 3 1 −1 + 4x − 3 2
> =⇒ > 0 =⇒ > 0.
2 2 2 x
So, we can take 0 as the lower bound. We have
j √ k
−1+ 4x−3
2 1
lim 0 ≤ lim ≤ lim √ .
x→∞ x→∞ x x→∞ x
Clearly, the left and right limits are both 0. So, the required density is also 0. □
̸ 0 be given. What is the density of the set of all numbers
Problem 3. Let integers a, b with b =
that are a (mod b)?
Solution. Let A denote this set. Then the density of A is given by
#A ∩ [1, x]
lim .
x→∞ x
We know that α ≡ a (mod b) ⇐⇒ b | α − a. So A is the set of all positive integer solutions α
such that β = α − a is divisible by b. Let x > 1, we will find #A ∩ [1, x]. We need to find the
number of α ∈ [1, x] satisfying b | α − a. Since β = α − a, clearly β ∈ [1 − a, x − a]. We need to
find the number of β ∈ [1 − a, x − a] such that b | β. This is the number of integers from 1−a
b to
x−a x−a
b . There are b + 1 such numbers.
So, we now have to evaluate the limit
 x−a 
b +1
lim .
x→∞ x
4 ABHINAV NEGI

x−a x
a/b b +1
We can let the lower bound be xb = x−a 1
bx = b − x , and the upper bound be x = 1
b + x1 .
Clearly, both limits converse to b . So, the density of A is 1b .
1

Problem 4. Let a be a given integer. What is the density of the set of all positive integers
divisible by a? What is the density of the set of all positive integers coprime to a?
Solution. The density of the first set is a1 using the result from the last problem. To find the
density of the second set, we note that every x coprime to a is among one of φ(a) equivalence
classes (mod a). So the density of the second set is φ(a)
a . □
Problem 5. What is the density of the set of prime numbers? (You may use the Prime Number
Theorem)
Solution. We know that  
x x
π(x) = +O .
ln x ln2 (x)
So, the density of the set of prime numbers is
 
x x
π(x) ln x + O 2
ln (x)

1

1

lim = lim = lim +O .
x→∞ x x→∞ x x→∞ ln x ln2 x
As x → ∞, we get ln x → ∞. So the limit is 0. □

3. Sum of two squares


If you solved the problems from last section, you should have found that the density of the set
of perfect squares is √
⌊ x⌋
lim = 0.
x∈∞ x
This was a very easy question to answer. Lets make it harder. We will now talk about the set
of all integers that can be expressed as the sum of two perfect squares. Let
A = {1, 2, 4, 5, 8, 9, 10, 13, 16, 17, 18, . . . }.
What is the density of A?
Lets first prove a famous result on the multiplicative closure of sums of squares.
3.1. Lemmas.
Theorem 3.1 (Diophantus identity). For all a, b, c, d ∈ C
2 2
a2 + b2 c2 + d2 = (ac − bd) + (ad + bc) .
 

Proof. This is just a matter of expanding and rearranging;


a2 + b2 c2 + d2 = a2 c2 + a2 d2 + b2 c2 + b2 d2
 

= (a2 c2 + b2 d2 ) + (a2 d2 + b2 c2 )
= (a2 c2 + b2 d2 − 2abcd) + (a2 d2 + b2 c2 + 2abcd)
= (ac − bd)2 + (ad + bc)2 .

Using this, it is easy to see that the set A is clused usng multiplication. Now, we prove a
famous theorem related to prime numbers that are a sum of two squares.
Theorem 3.2 (Fermat’s Chrismas Theorem). Given an odd prime p, there exist a, b ∈ Z such
that p = a2 + b2 if and only if p ≡ 1 (mod 4).
ALGEBRAIC NUMBER THEORY LECTURE NOTES 5

Proof. Let p = a2 + b2 be an odd prime. Since p is odd, a and b can’t both be even, nor can they
both be odd. So, one of them is odd, and one is even. Without Loss of Generality, let a = 2k + 1
be odd, and b = 2l be even. Then
p = a2 + b2 = (2k + 1)2 + (2l)2 = 4(k 2 + k + l2 ) + 1.
So clearly, p ≡ 1 (mod 4). Now, suppose p ≡ 1 (mod 4) is prime, we will show that there exist
integers a and b such that p = a2 + b2 . Since p ≡ 1 (mod 4), we have p = 4k + 1. First, we claim
that their exists some α ∈ Z such that p | α2 + 1. Note that
 
−1 p−1
= (−1) 2 = (−1)2k = 1.
p
So −1 is a quadratic residue (mod p), and there exists α such that α2 ≡ −1 (mod p) or p | α2 + 1.
Clearly p ∤ a, otherwise a2 + 1 ≡ 1 (mod p). Now, let x and y range over all the numbers
√  √  2
0, 1, 2, . . . p . There are a total of p + 1 > p choices of x and y. But for each of them,
there are p possible values √of y − xα in (mod p). So by the Pigeonhole Principle, there must be
some 0 ≤ x1 , x2 , y1 , y2 ≤ p such that
y1 − x1 α ≡ y2 − x2 α (mod p)
=⇒ y1 − y2 ≡ α(x1 − x2 ) (mod p)
2 2 2
=⇒ (y1 − y2 ) ≡ α (x1 − x2 ) (mod p)
2
≡ −(x1 − x2 ) (mod p).
√  √
Now, let a = |x1 − x2 | and a = |y1 − y2 |. We know that p | a2 + b2 , and 0 ≤ a, b ≤ p < p.

So 0 < a2 + b2 < 2( p)2 = 2p. Clearly a2 + b2 = p as required.
Having shown correspondence both sides, we have established the result. □
We want a way to easily characterize the integers in A, so we can easily tell whether a given
integer is in the set or not. Using the previous two theorems, we will prove a useful result due to
Euler.
3.2. Sum of two Squares Theorem.
Theorem 3.3 (Sum of two squares - Euler). Given a positive integer n, n ∈ A if and only if its
prime decomposition contains no factor pk , where p ≡ 3 (mod 4) and k is odd.
Proof. First, suppose that n is a positive integer with a prime decomposition that doesn’t contain
a factor pk , where p ≡ 3 (mod 4) and k is odd.
Let n = k 2 l, for k, l ∈ Z such that l is squarefree. We know that l = p1 p2 · pm where all
pi are mutually distinct and none of them are 3 (mod 4) by assumption. We know that all of
these primes are either 2 = 12 + 12 or they are 1 (mod 4), so all of them are in A. And thus,
l = p1 p2 · · · pm is also a sum of two squares. Since m2 = 02 + m2 , it follows that n = m2 l is also
a sum of two squares.
Now, suppose n = a2 + b2 for some a, b ∈ Z. We let n = k 2 l again for k, l ∈ Z where l is
squarefree. If we show that l has no prime factor that is 3 (mod 4) then we are done, as all the
prime factors of n that are 3 (mod 4) would then divide k 2 , and thus their exponents will be even.
We suppose that d = gcd(a, b). Let a = dx and b = dy. Clearly, x and y are coprime. We
have that d2 (x2 + y 2 ) = n = k 2 l. Since l is squarefree, k 2 is the square that divides n, so
d2 | n =⇒ d2 | k 2 . So k 2 = md2 , and
d2 (x2 + y 2 ) = k 2 l = d2 lm =⇒ x2 + y 2 = lm =⇒ l | x2 + y 2 .
So, for any prime p ̸= 2 dividing l, we have p | l | x2 + y 2 , so x2 ≡ −y 2 (mod p). We have
x2 y 2p−4 ≡ −y 2 · y 2p−4 (mod p) ≡ −y 2p−2 (mod p) ≡ −(y p−1 )2 (mod p).
6 ABHINAV NEGI

By Fermat’s Little Theorem, y p−1 ≡ 1 (mod p), so this is −1 (mod p). So (xy p−2 )2 ≡ −1
(mod p). But then
 
−1 p−1 p−1
= 1 =⇒ (−1) 2 = 1 =⇒ 2 | =⇒ 4 | p − 1.
p 2
So p ≡ 1 (mod 4). We have shown that l has no prime divisors that are 3 (mod 4). So we are
done. □
3.3. Problems. The result above can help us understand the equivalence classes that are occupied
by elements in A under various moduli. Here are a few problems you can solve using the obtained
result.
(1) What equivalence classes are occupied by elements of A under (mod 4)? (mod 9)? (mod
36)? (mod 81)?
(2) What equivalence classes are occupied by elements of A under (mod 49)? What about
(mod 2401)? Generalize for (mod 72k ) for all k ≥ 1.
(3) Given a prime p ≡ 3 (mod 4), what equivalence classes are occupied by elements of A
under (mod p2k ) for k ≥ 1?
3.4. Finding the Density. If you completed the problems in the last section, you will have
come to the conclusion that
Proposition 3.4. For each prime number p ≡ 3 (mod 4), and positive integers k, l such that
1 ≤ l ≤ p − 1. We have that
a ∈ A =⇒ a ̸≡ lp2k−1 (mod p2k ).
In simpler words, given any prime p ≡ 3 (mod 4) and any a ∈ A, we have;
a ̸≡ p, 2p, 3p, . . . , (p − 1)p (mod p2 )
a ̸≡ p3 , 2p3 , 3p3 , . . . , (p − 1)p3 (mod p4 )
a ̸≡ p5 , 2p5 , 3p5 , . . . , (p − 1)p5 (mod p6 )
..
.
Now that we have this fact, how do we approach to find the density of A? The key observation
is the following idea;
Idea. If for some m, when we look at the elements of a set A in (mod m), there are only k
k
residue classes the elements occupy. Then the density of A is at most m .
So maybe, lets work in mods. Notice that if n ∈ A then n ̸≡ 3 (mod 4). This implies using
the above idea that δ ≤ 34 . Can we find more such mods? Notice that n ̸≡ 3, 6 (mod 9)?
More generally n ̸≡ p, 2p, 3p, . . . , (p − 1)p (mod p2 ) for all primes p ≡ 3 (mod 4). So, now we
define sequence qi of primes 3 (mod 4) with q1 = 3, and we define the sequence ri as;
r1 = q12 , and ri = ri−1 qi2 for each n > 1.
So, we can use induction and CRT to show that the number of residue classes a (mod rk ) where
a ∈ A is at most;
k k  
Y Y qi − 1
(qi2 − qi + 1) = rk 1− for each k > 0.
i=1 i=1
qi2
So, for each k > 0, we get that;
k  
Y qi − 1
δ≤ 1− .
i=1
qi2
ALGEBRAIC NUMBER THEORY LECTURE NOTES 7

Taking k to infinity, gives us that;


 
Y p−1
δ≤ 1− .
p2
p prime
p≡3 (mod 4)

Our aim now is to show that the product on the right is 0. We will prove a lemma first;
Lemma 3.5. If t1 , t2 , . . . is a sequence of real numbers in the interval [0, 1), and if the series
t1 + t2 + · · · diverges, then (1 − t1 )(1 − t2 ) · · · = 0.
Proof. Suppose for contradiction that (1 − t1 )(1 − t2 ) · · · ̸= 0. Let this product be t, we have that
1 − t1 < e−t1 , so t < e−t1 e−t2 · · · , thus t < e−(t1 +t2 +··· ) . We get that;
 
t1 +t2 +··· 1 1
e < =⇒ t1 + t2 + · · · < ln .
t t
But since t1 + t2 + · · · is supposed to diverge, this is absurd. □

We can use this fact, so if we show that;


X p−1
p2
p prime
p≡3 (mod 4)

diverges, we are done. This sum can be split into two parts;
X p−1 X 1 X 1
2
= − .
p p p2
p prime p prime p prime
p≡3 (mod 4) p≡3 (mod 4) p≡3 (mod 4)

π2
P∞ 1
The second part is clearly finite, because i=1 i2 converges to 6 . So it suffices to show that the
first part diverges. We want to show that;
X 1
̸∈ R.
p
p prime
p≡3 (mod 4)

3.5. The Normal Dirichlet Series. Now, we are going to drop the 3 (mod 4) requirement for
a moment and just look at the Dirichlet series, and prove it diverges. We let;
X 1
S= .
p
p prime

Lets take a look at the series 1 + S + S 2 + S 3 + · · · . Expanding this series will give us a lot of
different fractions added up together. All the fractions of the form n1 for n ∈ Z+ are all included
in this series (figure out yourself why). So this must diverge. Because this is a geometric series,
we get S ≥ 1.
Now, we will use this argument to show that any tail P of this series is always greater than 1. For
an arbitrary positive integer k > 0, we define Sk = p prime p1 . We have that 1 + Sk + Sk2 + · · ·
p>k
1
contains all the terms of the form n for n ≡ 1 (mod k!) (figure out why). So

X 1
1 + Sk + Sk2 + · · · ≥ .
i=0
k! · i + 1
8 ABHINAV NEGI

It is easy to show that the sum of the right side diverges, implying that 1 + Sk + Sk2 + · · · must
also diverge. Again, this implies that Sk ≥ 1. Now, note that k chosen was arbitrary, so we have
shown that any tail of the series;
X 1
p
p prime
is always at least 1. Clearly, our sum diverges.

3.6. The 3 (mod 4) Dirichlet Series. We want to use a similar argument to show that the
series;
X 1
p
p prime
p≡3 (mod 4)
also diverges. Let’s try some ideas.
Let X = {n ∈ Z+ : n ≡ 3 (mod 4)}. If A ⊂ X, we define the relative density of A with respect
to X by
#A ∩ [1, x]
δ ′ (A) = lim = 4δ(A)
x→∞ x/4
Clearly, δ ′ (X) = 1. Now we will define Xp = {n ≡ 3 (mod 4) : p | n} for each prime p ≡ 3
(mod 4). We observe that everything that is 3 (mod 4) must have a prime factor which is 3 (mod
4). Thus:
[
X= Xp
p≡3 (4)
Hence technically, we should have;
[  X X 1
1 = δ ′ (X) = δ ′ Xp ≤ δ ′ (Xp ) = .
p
p prime p prime
p≡3 (4) p≡3 (4)

Note that δ ′ (Xp ) = 4δ(Xp ), and δ(Xp ) = 4p


1
because the elements of Xp all satisfy a congruence
(mod 4p). So, we have shown that p prime p1 ≥ 1. This is definitely useful, as now our aim
P
p≡3 (4)
should be to use the same argumentP
to show that any tail of this series is greater than 1.
1
But have we really shown that p prime p > 1? We haven’t. There is a mistake in this
p≡3 (4)
argument. We assumed that; [  X
δ′ Xp ≤ δ ′ (Xp ).
p prime
p≡3 (4)
It turns out that this property doesn’t hold. It isn’t true that the asymptotic density of union of a
family of sets isn’t always bounded by the sum of the individual asymptotic densities of the sets.
For example, consider the set of positive integers, which can be written as the union of the
sets {n} for all positive integers n. The density of Z is 1, while the sum of the densities of the
sets {n} for positive integers n is 0.
This can be taken care of, we will bound our set X and all its subsets in the range [1, x].
Instead of considering the densities, in this case we can the compare cardinalities of the sets,
divide both sides by x4 , and then take x to infinity. Let X = {n ≡ 3 (mod 4) : n ≤ x}, the
number of elements in this set is x4 + ϵ where e ∈ [−1, 1].
Now define Xp for all primes p ∈ X, as the set of Xp = {n : n ≤ x, n ≡ 3 (mod 4), p | n}.
x
Now, we can see that for each p satisfying the required conditions, the cardinality of xp is 4p + ϵp
for ϵp ∈ [−1, 1].
ALGEBRAIC NUMBER THEORY LECTURE NOTES 9

Now, we know that if n ∈ X, then n ≤ x, n ≡ 3 (mod 4), so there must be some prime divisor
of n with p ≡ 3 (mod 4). So n ∈ Xp . This way, we can see that
[
X= Xp .
Now, lets compare the cardinality of X with the sum of cardinalities of all the Xp ’s. We get;
x X x 
+ϵ≤ + ϵp
4 4p
But we have that;
x x X x  X
x

− 1 ≤ + ϵ and + ϵp ≤ +1 .
4 4 4p 4p
Combining these gives;
x Xx 
−1≤ +1
4 4p
p∈X
x X x X X x
=⇒ −1≤ + 1≤ + π(x)
4 4p 4p
p∈X p∈X p∈X
x xX1
=⇒ −1≤ + π(x)
4 4 p
p∈X
x 1
x
P
4 − 1 4 p∈X p π(x)
=⇒ ≤ +
x/4 x/4 x/4
4 X 1 4π(x)
=⇒ 1 − ≤ +
x p x
p∈X
X 1 4π(x) + 4
=⇒ 1 ≤ +
p x
p∈X

Now, we take x → ∞. The second term on the Right-Hand Side will become negligible, and we
get;
X 1
1≤ .
p
p prime
p≡3 (4)

So, we have been able to show that the sum of the series is greater than 1. Now, can we tweak
our argument somehow to show instead that any tail of this series is at least 1.

3.6.1. Final Argument. To do this, we introduce an arbitrary constant y ∈ R+ . We will focus


on the primes 3 (mod 4) that are greater than y. For this we need to make sure that all the
elements of our new X are all divisible by some prime p ≡ 3 (mod 4) with p > y. One way to do
this is by letting;
Y
P =4 p
p prime
p≤y

X = {n : n ≤ x, n ≡ −1 (mod P )}.
Now, all elements in X are −1 (mod p) for any prime p ≤ y, and are all 3 (mod 4) as well. So,
they clearly have a prime divisor p > y and p ≡ 3 (mod 4).
10 ABHINAV NEGI

So, for each p ∈ (y, x], let Xp be the set of all n ∈ X with p | n. It is clear that each n ∈ X
belongs to at least one of the Xp ’s. So, we can do the cardinality comparison again. We have
that;
x
#X = +ϵ
P
where ϵ ∈ [−1, 1]. For each p ≡ 3 (mod 4) and p ∈ (y, x], we have that;
x
#Xp = + ϵp
pP
for ϵp ∈ [−1, 1]. Now, lets compare. We get;
x X x 
+ϵ≤ + ϵp .
P pP
The left side is at least Px − 1, while the right side is at most Px p≡3 (mod 4) p1 + π(x). This gives
P
p∈(y,x]
us;
x x X 1
−1≤ + π(x)
P P p
p≡3 (4)
p∈(y,x]
x x X 1
=⇒ ≤ + π(x) + 1
P P p
p≡3 (4)
p∈(y,x]
X 1 π(x) + 1
=⇒ 1 ≤ +
p x/4
p≡3 (4)
p∈(y,x]
X 1 4π(x) + 4
=⇒ 1 ≤ + .
p x
p≡3 (4)
p∈(y,x]

Now, we can take x to infinity. We get that;


X 1
1≤ .
p>y
p
p≡3 (4)

So, we have shown that the tail of the series;


X 1
p
p≡3 (4)

is always greater than 1. So clearly, the series must diverge.

You might also like