Solutions to Ireland, Rosen “A Classical Introduction to

Modern Number Theory”

Richard Ganaye

August 21, 2022

Chapter 4

Ex. 4.1 Show that 2 is a primitive root modulo 29.

Proof. Let p = 29 : p − 1 = 22 × 7.
24 = 16 6≡ 1[29]
214 = 47 = 4 × 163 = 64 × 256 ≡ 6 × (−34) = −204 ≡ 86 = 3 × 29 − 1 ≡ −1[29]
228 ≡ 1[29] and 2d 6≡ 1 if d | 28, d < 28, hence 2 is a primitive element modulo 29.

Ex. 4.2 Compute all primitive roots for p = 11, 13, 17, and 19.

Proof. • p = 11. Then p − 1 = 10 = 2 × 5.

22 = 4 6≡ 1 (mod 11), and 25 = 32 ≡ −1 6≡ 1 (mod 11), so 2 is a primitive element
modulo 11.
The other primitive elements modulo 11 are congruent to the powers 2i , i ∧ 10 =
1, 1 ≤ i < 10, namely 2, 23 , 27 , 29 .
27 ≡ 7 (mod 11), 29 ≡ 6 (mod 11), so
{2, 8, 7, 6} is the set of the generators of U (Z/11Z).
Similarly :
• p = 13 : {2, 6, 11, 7} is the set of the generators of U (Z/13Z).
• p = 17 : {3, 10, 5, 11, 14, 7, 12, 6} is the set of the generators of U (Z/17Z).
• p = 19 : {2, 13, 14, 15, 3, 10} is the set of the generators of U (Z/19Z).
I obtain these results with the direct orders in S.A.G.E. :
p = 19; Fp = GF(p); a = Fp.multiplicative_generator()
print([a^k for k in range(1,p) if gcd(k,p-1) == 1])

Ex. 4.3 Suppose that a is a primitive root modulo pn , p an odd prime. Show that a is
a primitive root modulo p.

Proof. Suppose that a is a primitive root modulo pn : then a is a generator of U (Z/pn Z).
If a was not a primitive root modulo p, a is not a generator of U (Z/pZ), so there
exists b ∈ Z, b∧p = 1 such that ak 6≡ b (mod p) for all k ∈ Z. A fortiori ak 6≡ b (mod pn ),
and b ∧ pn = 1, so b ∈ U (Z/pn Z) and b 6∈ hai in U (Z/pn Z), in contradiction with the
hypothesis. So a is a primitive root modulo p.
(the reasoning on the orders of a, modulo p and modulo pn , is possible, but not so
easy.)

1
Ex. 4.4 Consider a prime p of the form 4t + 1. Show that a is a primitive root modulo
p iff −a is a primitive root modulo p.

Proof. Solution 1.
Suppose that a is a primitive root modulo p. As p − 1 is even, (−a)p−1 = ap−1 ≡ 1
(mod p).
If (−a)n ≡ 1 (mod p), with n ∈ N, then an ≡ (−1)n (mod p).
Therefore a2n ≡ 1 (mod p). As a is a primitive root modulo p, p − 1 | 2n, 2t | n, so
n is even.
Hence an ≡ 1 (mod p), and p − 1 | n. So the least n ∈ N∗ such that (−a)n ≡ 1
(mod p) is p − 1 : the order of −a modulo p is p − 1, −a is a primitive root modulo p.
Conversely, if −a is a primitive root modulo p, we apply the previous result at −a to
to obtain that −(−a) = a is a primitive root.

Solution 2.
Let p − 1 = 2a0 pa11 · · · pakk the decomposition of p − 1 in prime factors.
As pi is odd for i = 1, 2, · · · k, (p − 1)/pi is even, and a is primitive, so

(−a)(p−1)/pi = a(p−1)/pi 6≡ 1 (mod p),

(−a)(p−1)/2 = (−a)2k = a2k = a(p−1)/2 6≡ 1 (mod p).

So the order of a is p − 1 modulo p (see Ex. 4.8) : a is a primitive element modulo p.

Ex. 4.5 Consider a prime p of the form 4t + 3. Show that a is a primitive root modulo
p iff −a has order (p − 1)/2.

Proof. Let a a primitive root modulo p.

As ap−1 ≡ 1 (mod p), p | (a(p−1)/2 − 1)(a(p−1)/2 + 1), so p | a(p−1)/2 − 1 or p |
a (p−1)/2 + 1. As a is a primitive root modulo p, a(p−1)/2 6≡ 1 (mod p), so

a(p−1)/2 ≡ −1 (mod p).

Hence (−a)(p−1)/2 = (−1)2t+1 a(p−1)/2 ≡ (−1) × (−1) = 1 (mod p).

Suppose that (−a)n ≡ 1 (mod p), with n ∈ N.
Then a2n = (−a)2n ≡ 1 (mod p), so p − 1 | 2n, p−1
2 | n.
So −a has order (p − 1)/2 modulo p.

Conversely, suppose that −a has order (p − 1)/2 = 2t + 1 modulo p. Let 2, p1 , . . . pk

the prime factors of p − 1, where pi are odd.
a(p−1)/2 = a2t+1 = −(−a)2t+1 = −(−a)(p−1)/2 ≡ −1, so a(p−1)/2 6≡ 1 (mod 2).
As p − 1 is even, (p − 1)/pi is even, so
a(p−1)/pi = (−a)(p−1)/pi 6≡ 1 (mod p) (since −a has order p − 1).
So the order of a is p − 1 (see Ex. 4.8) : a is a primitive root modulo p.

n
Ex. 4.6 If p = 22 + 1 is a Fermat prime, show that 3 is a primitive root modulo p.

Proof. Solution 1 (with quadratic reciprocity).

Write p = 2k + 1, with k = 2n .
We suppose that n > 0, so k ≥ 2, p ≥ 5. As p is prime, 3p−1 ≡ 1 (mod p).
k
In other words, 32 ≡ 1 (mod p) : the order of 3 is a divisor of 2k , a power of 2.

2
 k−1
 k−1 2
3 has order 2k modulo p iff 32 6≡ 1 (mod p). As 32 ≡ 1 (mod p), where p is
k−1
prime, this is equivalent to 32 ≡ −1 (mod p), which remains to prove.
2k−1 (p−1)/2 3


3 =3 ≡ p (mod p).
As the result is true for p = 5, we can suppose n ≥ 2. From the law of quadratic
reciprocity :   
3 p k−1
= (−1)(p−1)/2 = (−1)2 = 1.
p 3
So p3 = p3

n n
p = 22 + 1 ≡ (−1)2 + 1 (mod 3)
≡ 2 ≡ −1 (mod 3),
3

p

so p = 3 = −1, that is to say
k−1
32 ≡ −1 (mod p).
n n
The order of 3 modulo p = 22 + 1 is p − 1 = 22 : 3 is a primitive root modulo p.
(On the other hand, if 3 is of order p − 1 modulo p, then p is prime, so
n 2n −1
Fn = 22 + 1 is prime ⇐⇒ 3(Fn −1)/2 = 32 ≡ −1 (mod Fn ).)

Solution 2 (without quadratic reciprocity, with the hint of chapter 4).

n−1
As above, if we suppose that 3 is not a primitive root modulo p, then 32 ≡ 1
(p−1)/2 2n−1
(mod p), so n ≥ 2, and (−3) =3 ≡ 1 (mod p), so −3 is a square modulo p :
there exists a ∈ Z such that −3 ≡ a2 (mod p).
As 2 ∧ √
p = 1, there exists u ∈ Z such that 2u ≡ −1 + a (mod p) (u is similar to
ω = −1+i2
3
∈ C). Then

8u3 ≡ (−1 + a)3

≡ −1 + 3a − 3a2 + a3
≡ −1 + 3a + 9 − 3a
≡8 (mod p)

As p ∧ 2 = p ∧ 8 = 1, u3 ≡ 1 (mod p). Moreover, if u ≡ 1 (mod 3), then a ≡ 3 (mod p),

−3 ≡ 9 (mod p), p | 12, so p = 2 or p = 3, in contradiction with p ≥ 5. So the order of u
modulo p is 3 : (Z/pZ)∗ contains an element u of order 3. So 3 | p − 1, p ≡ 1 (mod 3),
n
but p ≡ (−1)2 + 1 ≡ 2 ≡ −1 (mod 3) : this is a contradiction, so 3 is a primitive root
n
modulo p = 22 + 1.

Ex. 4.7 Suppose that p is a prime of the form 8t + 3 and that q = (p − 1)/2 is also a
prime. Show that 2 is a primitive root modulo p.

Proof. The first examples of such couples (q, p) are (5, 11), (29, 59), (41, 83), (53, 107), (89, 179).
p = 2q + 1 = 8t + 3 and p, q are prime numbers.
From Fermat’s little theorem, 2p−1 ≡ 1 (mod p), so 22q ≡ 1 (mod p).
The order of 2 modulo p divides 2q : to prove that the order of 2 is 2q = p − 1, it is
suffisant to prove
22 6≡ 1 (mod p), 2q 6≡ 1 (mod p).

3
 If 22 ≡ 1 (mod p), then p | 3, p = 3 and q = 1 : q is not a prime, so 22 6≡ 1 (mod p).
If 2q = 2(p−1)/2 ≡ 1 (mod p), then 2 is a square modulo p (prop. 4.2.1) : there exists
a ∈ Z such that 2 ≡ a2 (mod p).
From the complementary case of the law of quadratic reciprocity (see next chapter,
prop. 5.1.3), 2 is a square modulo p iff
 
2 2
1= = (−1)(p −1)/8 .
p
2
Yet p ≡ 3 (mod 8), so p2 ≡ 9 (mod 16), p2 = (−1)(p −1)/8 = −1, so 2 is not a square

modulo p. This is a contradiction, so 2q 6≡ 1 (mod p) : 2 is a primitive root modulo

p.

Ex. 4.8 Let p be an odd prime. Show that a is a primitive root modulo p iff a(p−1)/q 6≡ 1
(mod p) for all prime divisors q of p − 1.
Proof. • If a is a primitive root, then ak 6≡ 1 for all k, 1 ≤ k < p − 1, so a(p−1)/q 6≡ 1
(mod p) for all prime divisors q of p − 1.
• In the other direction, suppose a(p−1)/q 6≡ 1 (mod p) for all prime divisors q of p−1.
Let δ the order of a, and p − 1 = q1a1 q2a2 · · · qkak the decomposition of p − 1 in prime
factors. As δ | p − 1, δ = q1b1 pb22 · · · qkbk , with bi ≤ ai , i = 1, 2, . . . , k. If bi < ai for some
index i, then δ | (p − 1)/qi , so a(p−1)/qi ≡ 1 (mod p), which is in contradiction with the
hypothesis. Thus bi = ai for all i, and δ = q − 1 : a is a primitive root modulo p.

Ex. 4.9 Show that the product of all the primitive roots modulo p is congruent to
(−1)φ(p−1) modulo p.
Proof. Here we suppose p prime, p > 2. Let g a primitive root modulo p. U (Z/pZ) is
cyclic, generated by g:
U (Z/pZ) = {1, g, g 2 , . . . , g p−2 }, g p−1 = 1.
g k is a primitive element iff k ∧ (p − 1) = 1, so the product of primitive elements in
U (Z/pZ) is Y
P = gk .
k∧(p−1)=1
1≤k<p−1

gS ,
P
so P = where S = k.
k∧(p−1)=1
1≤k<p−1
From Ex. 2.22, we know that for n ≥ 2,
X 1
k = nφ(n).
2
k∧n=1
1≤k<n

k = 12 (p − 1)φ(p − 1).
P
So S =
k∧(p−1)=1
1≤k<p−1
As p > 2, p − 1 is even. (g (p−1)/2 )2 = g p−1 = 1, and g (p−1)/2 6= 1. As Z/pZ is a field,
g (p−1)/2 = −1.
Thus P = (−1)φ(p−1) : so the product P of all the primitive roots modulo p is such
that
P ≡ (−1)φ(p−1) (mod p).

4
Ex. 4.10 Show that the sum of all the primitive roots modulo p is congruent to µ(p−1)
modulo p.

Proof. Notation : Fp = Z/pZ is the field with p elements, |x| the multiplicative order of
an element x ∈ F∗p , N∗ = {1, 2, 3, . . .}.
Let 
 N∗ → Fp P
ψ: n 7→ ψ(n) = d
d∈F∗ ,|d|=n

p

ψ(n) is the sum of the elements with order n in F∗p .

So ψ(n) = 0 if n - p − 1, and
S = ψ(p − 1) is the sought sum of all the primitive roots modulo p.
We compute for all n ∈ N∗ X
f (n) = ψ(d).
d|n

f (n) is the sum of elements whose order divides n, in other worlds the sum of the
roots of xn − 1. This sum is, up to the sign, the coefficient of xn−1 , so is null, except in
the case n = 1, where the sum of the unique root 1 of x − 1 is 1. So

f (1) = 1, ∀n > 1, f (n) = 0,

(f = χ{1} is the characteristic function of {1}).

From the Möbius inversion formula, for all n ∈ N∗ , ψ(n) = d|m µ n
P

d f (d), so
X p − 1
ψ(p − 1) = µ f (d) = µ(p − 1).
d
d|p−1

Conclusion : X
S= d = µ(p − 1) :
d∈F∗p ,|d|=p−1

the sum of all the primitive roots modulo p is congruent to µ(p − 1) modulo p.

Ex. 4.11 Prove that 1k + 2k + ... + (p − 1)k ≡ 0 (mod p) if p − 1 - k, and −1 (mod p)

if p − 1 | k.

Proof. Let Sk = 1k + 2k + · · · + (p − 1)k .

Let g a primitive root modulo p : g a generator of F∗p .
As (1, g, g 2 , . . . , g p−2 ) is a permutation of (1, 2, . . . , p − 1),
k k k
Sk = 1 + 2 + · · · + p − 1
p−2
(
X
ki
p − 1 = −1 if p − 1 | k
= g = g (p−1)k −1
g k −1
= 0 if p − 1 - k
i=0

since p − 1 | k ⇐⇒ g k = 1.
Conclusion :

1k + 2k + · · · + (p − 1)k ≡ 0 (mod p) if p − 1 - k
k k k
1 + 2 + · · · + (p − 1) ≡ −1 (mod p) if p − 1 | k

5
Ex. 4.12 Use the existence of a primitive root to give another proof of Wilson’s theorem
(p − 1)! ≡ −1 (mod p).

Proof. As the result is trivial if p = 2, we suppose that p is an odd prime.

Let g a primitive root modulo p : g a generator of F∗p .
As (g (p−1)/2 )2 = g p−1 = 1, and g (p−1)/2 6= 1 in the field F∗p , then g (p−1)/2 = −1, and
(1, g, g 2 , . . . , g p−2 ) is a permutation of (1, 2, . . . , p − 1), so
p−2
Y
(p − 1)! = gk
k=0
Pp−2
k
=g k=0

= g (p−2)(p−1)/2
 p−2
= g (p−1)/2
= (−1)p−2
= −1.

Hence (p − 1)! ≡ −1 (mod p) for each prime p.

Ex. 4.13 Let G be a finite cyclic group and g ∈ G a generator. Show that all the other
generators are of the form g k , where (k, n) = 1, n being the order of G.

Proof. Suppose G = hgi, with Card G = n, so the order of g is n.

Let x another generator of G, then x = g k , and g = xl , k, l ∈ Z, so g = g kl , g kl−1 =
e : n | kl − 1, then kl − 1 = qn, q ∈ Z, so n ∧ k = 1.
Conversely, if u ∧ k = 1, there exist u, v ∈ Z such that un + vk = 1, so g = g un+vk =
(g )u (g k )v = xv ∈ hxi, so G ⊂ hxi, G = hxi : x is a generator of G.
n

Conclusion : if g is a generator of G, all the other generators are the elements g k ,

where k ∧ n = 1, n = |G|.

Ex. 4.14 Let A be a finite abelian group and a, b ∈ A elements of order m and n,
respectively. If (m, n) = 1, prove that ab has order mn.

Proof. Suppose |a| = m, |b| = n, m ∧ n = 1.

• If (ab)k = e, then ak = b−k , so akn = b−kn = (bn )−k = e, so m | kn, with m ∧ n = 1,
so m | k.
Similarly, bkm = a−km = (am )−k = e, so n | km, n ∧ m = 1 : n | k.
As n | k, m | k, n ∧ m = 1, nm | k.
• Conversely, if nm | k, k = qnm, q ∈ Z, so (ab)k = ak bk = (am )qn (bn )qm = e.

∀k ∈ Z, (ab)k = e ⇐⇒ nm | k.

So |ab| = nm.

Ex. 4.15 Let K be a field and G ⊂ K ∗ a finite subgroup of the multiplicative group of
K. Extend the arguments used in the proof of Theorem 4.1 to show that G is cyclic.

Solution 1.

6
Proof. Let n = |G|. From Lagrange’s theorem, an = 1 for all a ∈ G, so the polynomial
xn − 1 ∈ K[x] has exactly n roots in G, and so

∀x ∈ K, x ∈ G ⇐⇒ xn = 1.

If d | n, the polynomial xd − 1 ∈ K[x] has exactly d roots in K otherwise xn − 1 =

(xd − 1)g(x), g(x) ∈ K[x], and deg(g) = n − d has at most n − d roots, so xn − 1 would
have less than n roots in K. As xd0 = 1 ⇒ xn0 = 1, all these roots are in G : xd − 1 has d
roots in G. P
Let ψ(d) the number of elements in G of order d ( ψ(d)
P = 0 if d - n). Then c|d ψ(c) =
d. Applying the Möbius inversion theorem, ψ(d) = c|d µ(c)d/c = φ(d) (Prop. 2.2.5),
in particular, ψ(n) = φ(n) ≥ 1. This proves the existence of an element of order n in G,
so G is cyclic.
(variation : ψ(d) = 0 if there exists no element
P of order d,Pand ψ(d) = φ(d) otherwise
: see Ex.4.13. So ψ(d) ≤ φ(d) for all d | n. As d|n ψ(d) = d|n φ(d) = n, ψ(d) = φ(d)
for all d | n. So there exists in G an element of order n, and G is cyclic.)

Solution 2.

Proof. Let n = |G| = pa11 · · · pakk . From Lagrange’s theorem, y n = 1 for all y ∈ G.
p(x) = xn/p1 − 1 ∈ K[x] has at most n/p1 < n roots in K ∗ , a fortiori in G, so there
exists a ∈ G such that an/p1 6= 1.
a1 a2 ak a1 a1 −1
p p
Let c1 = an/p1 = ap2 ···pk . Then c11 = 1 and c11 = an/p1 6= 1, so |c1 | = pa11 .
Similarly, there exist c2 , . . . , ck with respective orders |ci | = pai i .
From exercise 4.14, we obtain by induction that c = c1 · · · ck has order pa11 · · · pakk = n,
so G is cyclic.

Ex. 4.16 Calculate the solutions to x3 ≡ 1 (mod 19) and x4 ≡ 1 (mod 17).

Proof. Here we note a the class of x in Z/pZ.

Let a ∈ F19 . Then

a3 − 1 = 0 ⇐⇒ a − 1 = 0 or a2 + a + 1 = 0.

a2 + a + 1 = 0 ⇐⇒ (a + 10) − 99 = 0
⇐⇒ (a + 10)2 − 4 = 0
⇐⇒ (a + 8)(a + 12) = 0

So, for all x ∈ Z,

x3 ≡ 1 (mod 19) ⇐⇒ x ≡ 1, 7, 11 (mod 19).

Let a ∈ F17 .

a4 = 1 ⇐⇒ a2 = 1 or a2 = −1 = 42
⇐⇒ a = ±1 or a = ±4

So, for all x ∈ Z,

x4 ≡ 1 (mod 17) ⇐⇒ x ≡ −1, 1, −4, 4 (mod 17).

7
 Alternatively, we can take primitives roots modulo 19 and 17.
2 is a primitive root modulo 19, Let a = 2k ∈ F19 .

a3 = 1 ⇐⇒ 23k = 1
⇐⇒ 18 | 3k
⇐⇒ 6 | k
⇐⇒ a = 1, 26 = 7, 212 = 11

3 is a primitive root modulo 17. Let a = 3k ∈ F17 .

a4 = 1 ⇐⇒ 34k = 1
⇐⇒ 16 | 4k
⇐⇒ 4 | k
⇐⇒ a = 1, 34 = −4, 38 = −1, 312 = 4

Ex. 4.17 Use the fact that 2 is a primitive root modulo 29 to find the seven solutions
to x7 ≡ 1 (mod 29).

Proof. Let x ∈ Z, then x ≡ 2k (mod 29), k ∈ N.

x7 ≡ 1 (mod 29) ⇐⇒ 27k ≡ 1 (mod 29)

⇐⇒ 28 | 7k
⇐⇒ 4 | k

So the group cyclic S of the roots of x7 − 1 in F29 are

S = {1, 24 , 28 , 212 , 216 , 220 , 224 },

S = {1, 16, 24, 7, 25, 23, 20}.

Ex. 4.18 Solve the congruence 1 + x + · · · + x6 ≡ 0 (mod 29).

Proof. As (1 + x + · · · + x6 )(1 − x) = 1 − x7 ,

x7 ≡ 1

6 (mod 29)
1 + x + ··· + x ≡ 0 (mod 29) ⇐⇒
x 6≡ 1 (mod 29)

From Ex. 4.17, the solutions are congruent to 24 , 28 , 212 , 216 , 220 , 224 modulo 29.

Ex. 4.19 Determine the numbers a such that x3 ≡ a (mod p) is solvable for p =
7, 11, 13.

Proof. (a) If p = 7, then 3 | p − 1, d = 3 ∧ (p − 1) = 3. From Prop. 4.2.1,

∃x ∈ Z, a ≡ x3 (mod 7) ⇐⇒ a ≡ 0 (mod 7) or a(p−1)/3 = a2 ≡ 1 (mod 7).

So the numbers a such that x3 ≡ a (mod 7) is solvable are congruent at 0, 1, −1

modulo 7.

8
 (b) If p = 11, then d = 3 ∧ (p − 1) = 1. With the same proposition,

∃x ∈ Z, a ≡ x3 (mod 11) ⇐⇒ a ≡ 0 (mod 11) or ap−1 = a6 ≡ 1 (mod 11).

So all integers a are cube modulo 11, in only one way.

For an alternative proof, the application
 ∗
F11 → F∗11
f:
x 7→ x3

f is a bijection. Indeed,
• f is a group homomorphism,
• x3 = 1 ⇒ (x3 )7 = 1 ⇒ x = 1 so ker(f ) = {1},
• f : F∗11 → F∗11 is injective and F∗11 is finite, so f is bijective.
In F11 , 0 = 03 , 1 = 13 , 2 = 73 , 3 = 93 , 4 = 53 , 5 = 33 , 6 = 83 , 7 = 63 , 8 = 23 , 9 =
43 , 10 = 103 .

(c) If p = 13, then 3 | p − 1, 3 ∧ (p − 1) = 3, so

∃x ∈ Z, a ≡ x3 (mod 13) ⇐⇒ a ≡ 0 (mod 13) or a(p−1)/3 = a4 ≡ 1 (mod 13)

⇐⇒ a ≡ 0, 1, −1, 5, −5 (mod 13)

(5 ≡ 83 (mod 13).)

Ex. 4.20 Let p be a prime, and d a divisor of p − 1. Show that dth powers form a
subgroup of U (Z/pZ) of order (p − 1)/d. Calculate this subgroup for p = 11, d = 5, for
p = 17, d = 4, and for p = 19, d = 6.

Proof. Here p is a prime number, and d | p − 1. Let

 ∗
Fp → F∗p
f:
x → xd

Then f is a group homomorphism, and im(f ) is the set of dth powers, and consequently
is a subgroup of U (Fp ) = F∗p . ker(f ) is the group of the roots of xd − 1. As d | p − 1, the
polynomial xd − 1 has exactly d roots (Prop. 4.1.2), so | ker(f )| = d.
As im(f ) ' F∗p / ker(f ),

|im(f )| = |F∗p |/| ker(f )| = (p − 1)/d.

So there exist exactly (p − 1)/d dth powers in (Z/pZ)∗ .

From Prop. 4.2.1, as d | p − 1, d ∧ (p − 1) = d, for all x ∈ F∗p ,

x ∈ im(f ) ⇐⇒ x(p−1)/d = 1.
So the group of dth powers is the group of the roots of x(p−1)/d − 1.
• If p = 11, d = 5, im(f ) = {1, −1}.
• If p = 17, d = 4, x ∈ im(f ) ⇐⇒ x4 = 1 : im(f ) = {1, −1, 4, −4}.
• If p = 19, d = 6, x ∈ im(f ) ⇐⇒ x3 = 1 : im(f ) = {1, 7, 72 = 11},
where 7 ≡ 26 (mod 19).

9
Ex. 4.21 If g is a primitive root modulo p, and d|p − 1, show that g (p−1)/d has order
d. Show also that a is a dth power iff a ≡ g kd (mod p) for some k. Do Exercises 16-20
making use of those observations.

Proof. Let x = g (p−1)/d ∈ F∗p , where g is a primitive root modulo p. For all k ∈ Z,
p−1
xk = 1 ⇐⇒ g k =1
d

p−1
⇐⇒ p − 1 | k
d
⇐⇒ d | k

So the order of g (p−1)/d is d.

• If a = g kd , then a = xd , where x = g k , so a is a dth power.
• If a 6= 0 is a dth power, a = xd , x ∈ F∗p . As x ∈ hgi, x = g k , so a = g kd .
So, if a 6≡ 0 (mod p), a is a dth power iff a ≡ g kd (mod p) for some k.
By example (Ex. 4.20), 2 is a primitive root modulo 19, so the 6th powers modulo
19 are 20 = 1, 26 = 7, 212 = 11.

Ex. 4.22 If a has order 3 modulo p, show that 1 + a has order 6.

Proof. If a has order 3 modulo p, then 0 ≡ a3 − 1 = (a − 1)(a2 + a + 1) (mod p), with

a 6≡ 1 (mod p), so a2 + a + 1 ≡ 0 (mod p). Thus

(1 + a)3 ≡ 1 + 3a + 3a2 + a3
≡ 1 + 3a + 3(−1 − a) + 1
≡ −1 (mod p)

So (1 + a)6 ≡ 1 (mod p).

(1 + a)2 ≡ 1 + 2a + a2 = 1 + 2a + (−1 − a) ≡ a 6≡ 1 (mod p).
So (1 + a)6 ≡ 1, (1 + a)2 6≡ 1, (1 + a)3 6≡ 1 (mod p), so the order of 1 + a divides 6,
but doesn’t divides 2 or 3, so 1 + a has order 6 modulo p.

Ex. 4.23 Show that x2 ≡ −1 (mod p) has a solution iff p ≡ 1 (mod 4), and that
x4 ≡ −1 (mod p) has a solution iff p ≡ 1 (mod 8).

Proof. If x2 ≡ −1 (mod p), then x has order 4 in F∗p , hence from Lagrange’s theorem,
4 | p − 1.
Conversely, suppose 4 | p − 1, so p = 4k + 1, k ∈ N∗ . From proposition 4.2.1, as
2 | p − 1, −1 is a square modulo p iff (−1)(p−1)/2 ≡ 1 (mod p), which is true because
(−1)(p−1)/2 = (−1)2k = 1.
If x4 ≡ −1 (mod p), then x8 = 1 ∈ F∗p , and x4 6= 1, so x has order 8 in F∗p , so 8 | p − 1.
Conversely, if p ≡ 1 (mod 8), p = 8K + 1, K ∈ N∗ . From Prop.4.2.1, as 4 | p − 1,
there exists x ∈ Z such that −1 = x4 iff (−1)(p−1)/4 ≡ 1 (mod 8), which is true because
(−1)(p−1)/4 = (−1)2K = 1.
Conclusion :

∃x ∈ Z, x4 ≡ −1 (mod p) ⇐⇒ p ≡ 1 (mod 8).

10
Ex. 4.24 Show that axm + by n ≡ c (mod p) has the same number of solutions as
0 0
axm + by n ≡ c (mod p), where m0 = (m, p − 1) and n0 = (n, p − 1).

Proof. If a∧b - c, the two equations have no solution. So we can suppose a∧b | c, and after
division by δ = a ∧ b, we obtain an equation a0 xm + b0 y n = c0 , a0 = a/δ, b0 = bδ, c0 = cδ,
and a0 ∧ b0 = 1. So it remains to prove that axm + by n ≡ c (mod p) has the same number
0 0
of solutions as axm + by n ≡ c (mod p) when a ∧ b = 1.
In this case the equation au + bv = c has solutions. Let N the number of solutions
(x, y) of the equation a xm + b y n = c,N 0 the number of solutions (x, y) of the equation
0 0
a xm + b y n = c. Then

N = Card{(x, y) ∈ Fp × Fp | a xm + b y n = c}
X
= Card{(x, y) ∈ Fp × Fp | xm = u, y n = v}
au+bv=c
X
= Card{x ∈ Fp |xm = u} × Card{y ∈ Fp |y n = v}.
au+bv=c

The same is true for N 0 , so it is suffisiant to prove that

0
Card{x ∈ Fp | xm = u} = Card{x ∈ Fp | xm = u},

where m0 = m ∧ (p − 1), and a similar equality for the equation y n = v.

Let g a generator of F∗p . Write u = g r , r ∈ N.

∃x ∈ Fp , xm = u ⇐⇒ ∃k ∈ Z, g mk = g r
⇐⇒ ∃k ∈ Z, p − 1 | mk − r
⇐⇒ ∃k ∈ Z, ∃l ∈ Z, r = mk + l(p − 1)
⇐⇒ m ∧ (p − 1) | r

So
{x ∈ Fp | xm = u} =
6 ∅ ⇐⇒ m ∧ (p − 1) | r,
and similarly
0
6 ∅ ⇐⇒ m0 ∧ (p − 1) | r.
{x ∈ Fp | xm = u} =
Since m0 ∧ (p − 1) = (m ∧ (p − 1)) ∧ (p − 1) = m ∧ (p − 1), these two conditions are
equivalent, so these two sets are empty for the same values of u.
Let u is such that {x ∈ Fp | xm = u} = 6 ∅, and x0 a fixed solution of xm = u.
Write x = g , x0 = g 0 . Let d = m ∧ (p − 1)(= m0 ).
k k

xm = u ⇐⇒ xm = x0 m
⇐⇒ g mk = g mk0
⇐⇒ p − 1 | m(k − k0 )
p−1 m
⇐⇒ | (k − k0 )
d d
p−1
⇐⇒ | k − k0
d
p−1
⇐⇒ ∃j ∈ Z, k = k0 + j
d

11
 p−1 p−1 p−1
As g is a primitive root modulo p, the distinct solutions are x0 , x0 g d , . . . , x0 g k d , . . . x0 g (d−1) d ,
so in this case
Card{x ∈ Fp | xm = u} = d = m ∧ (p − 1).
As m0 ∧ (p − 1) = m ∧ (p − 1),
0
Card{x ∈ Fp | xm = u} = Card{x ∈ Fp | xm = u}.
0 0
So N = N 0 : axm + by n ≡ c (mod p) has the same number of solutions as axm + by n ≡ c
(mod p), where m0 = (m, p − 1) and n0 = (n, p − 1).

Ex. 4.25 Prove Propositions 4.2.2 and 4.2.4.

Proposition 4.2.2. Suppose that a is odd, e ≥ 3, and consider the congruence
xn ≡ a (mod 2e ). If n is odd, a solution always exists and it is unique.
e−2
If n is even, a solution exists iff a ≡ 1 (mod 4), a2 /d ≡ 1 (mod 2e ), where d =
(n, 2e−2 ). When a solution exists there are exactly 2d solutions.

Proof. We suppose that a is odd and e ≥ 3.

From Theorem 2’, we know that {(−1)a 5b | 0 ≤ a ≤ 1, 0 ≤ b ≤ 2e−2 } constitutes a
reduced residue system modulo 2e , so we can write

a ≡ (−1)s 5t (mod 2e ), 0 ≤ s ≤ 1, 0 ≤ t ≤ 2e−2 ,

x ≡ (−1)y 5z (mod 2e ), 0 ≤ y ≤ 1, 0 ≤ z ≤ 2e−2 .

For all x ∈ Z,

xn ≡ a (mod 2e ) ⇐⇒ (−1)ny 5nz ≡ (−1)s 5t (mod 2e )

Then (−1)ny ≡ (−1)s (mod 4), ny ≡ s (mod 2), (−1)ny = (−1)s , so 5nz ≡ 5t (mod 2e ).
Conversely, if ny ≡ s (mod 2) and 5nz ≡ 5t (mod 2e ), then xn ≡ a (mod 2e ), so
 
ny ≡ s (mod 2) ny ≡ s (mod 2)
xn ≡ a (mod 2e ) ⇐⇒ ⇐⇒
5nz ≡ 5t (mod 2e ) nz ≡ t (mod 2e−2 )

since the order of 5 modulo 2e is 2e−2 .

• Suppose that n is an odd integer. Then
 
ny ≡ s (mod 2) y ≡ s (mod 2)
⇐⇒
nz ≡ t (mod 2e−2 ) z ≡ n0 t (mod 2e−2 )

where n0 is an inverse of n modulo 2e−2 : nn0 ≡ 1 (mod 2e−2 ).

So xn ≡ a (mod 2e ) has an unique solution modulo 2e .
• Suppose
 that n is an even integer.
ny ≡ s (mod 2)
Then implies s ≡ 0 (mod 2) and d = n ∧ 2e−2 | t.
nz ≡ t (mod 2e−2 )
Then a ≡ (−1)s 5t ≡ 5t (mod 2e ), so a ≡ 1 (mod 4).
2e−2
 e−2  t
d
Hence a d ≡ 52 ≡ 1 (mod 2e ), since 5 has order 2e−2 , and d | t.
So, if n is even, and d = n ∧ 2e−2 ,
(
a ≡ 1 (mod 4),
∃x ∈ Z, xn ≡ a (mod 2e ) ⇒ 2e−2
a d ≡ 1 (mod 2e ).

12
 (
a ≡ 1 (mod 4),
Conversely, suppose that 2e−2
a ≡ 1 (mod 2e ).
d

Then a ≡ (−1)s 5t (mod 2e ) implies a ≡ (−1)s (mod 4), so s is even, and a ≡ 5t

(mod 2e ).
2e−2 e−2
Therefore 5t d ≡ 1 (mod 2e ), which implies 2e−2 | t 2 d , so d | t.

ny ≡ s (mod 2)
∃x ∈ Z, xn ≡ a (mod 2e ) ⇐⇒ ∃y ∈ Z, ∃z ∈ Z,
nz ≡ t (mod 2e−2 )
⇐⇒ ∃z ∈ Z, nz ≡ t (mod 2e−2 ) (since n, s even)
e−2
⇐⇒ ∃z ∈ Z, 2 | nz − t
n 2e−2 t
⇐⇒ ∃z ∈ Z, z− |
d d d
2e−2 n t
⇐⇒ ∃z ∈ Z, ∃q ∈ Z, q +z =
d d d
e−2 2e−2
As 2 d ∧ nd = 1, there exists a solution (q, z0 ) of this last equation, where 0 ≤ z0 < d ,
and so x0 = 5z0 is a particular solution of xn ≡ a (mod 2e ), therefore
(
n e
a ≡ 1 (mod 4)
∃x ∈ Z, x ≡ a (mod 2 ) ⇐⇒ 2e−2
a d ≡ 1 (mod 2e )

If there exists a particular solution x0 ≡ (−1)y0 5z0 , then

xn ≡ a (mod 2e ) ⇐⇒ xn ≡ xn0 (mod 2e )


ny ≡ ny0 (mod 2)
⇐⇒
nz ≡ nz0 (mod 2e−2 )
⇐⇒ n(z − z0 ) ≡ 0 (mod 2e−2 ) (since n even)
2e−2 n
⇐⇒ | (z − z0 )
d d
2 e−2 2e−2 n
⇐⇒ | z − z0 , (since ∧ = 1)
d d d
2 e−2
⇐⇒ ∃k ∈ Z, z = z0 + k
d
As the order of 5 modulo 2e is 2e−2 , the solutions of xn ≡ a (mod 2e ) are
2e−2
xk = (−1)y 5z0 +k d , 0 ≤ y < 2, 0 ≤ k < d,

so there are exactly 2d solutions modulo 2e .

Proposition 4.2.4. Let 2l be the highest power of 2 dividing n. Suppose that a is

odd and that xn ≡ a (mod 22l+1 ) is solvable. Then xn ≡ a (mod 2e ) is solvable for all
e ≥ 2l + 1 (and consequently for all e ≥ 1). Moreover, all these congruences have the
same number of solutions.

Proof. We suppose that a is odd, and that xn ≡ a (mod 22l+1 ) is solvable. l is such that
n = 2l n0 , where n0 is an odd integer.

13
 Let the induction hypothesis be, for a fixed integer m ≥ 2l + 1,
∃x0 ∈ Z, xn0 ≡ a (mod 2m ).
Let x1 = x0 + b2m−l : we show that for an appropriate choice of b ∈ {0, 1}, xn1 ≡ a
(mod 2m+1 ).
xn1 = xn0 + nb2m−l xn−1
0 + 22m−2l A, A ∈ Z.
Since m ≥ 2l + 1, 2m − 2l ≥ m + 1, so
xn1 ≡ xn0 + nb2m−l xn−1
0 (mod 2m+1 ).

xn1 ≡ a (mod 2m+1 ) ⇐⇒ (xn0 − a) + n0 bxn−1 0 2m ≡ 0 (mod 2m+1 )

xn − a
⇐⇒ 0 m + n0 bxn−1 0 ≡ 0 (mod 2)
2
As a is odd, and xn0 ≡ a (mod 2m ), m ≥ 1, x0 is odd, and n0 is odd, so there exists an
xn −a
unique b ∈ {0, 1} such that 02m + n0 bx0n−1 ≡ 0 (mod 2). So there exists x1 ∈ Z such
that xb1 ≡ a (mod 2m+1 ), and the induction is completed. Therefore, xn ≡ a (mod 2e )
is solvable for all e ≥ 2l + 1, and consequently for all e ≥ 1.

From the Proposition 4.2.2., with the hypothesis e ≥ 3, we know that the number of
solutions of the solvable equation xn ≡ a (mod 2e ), e ≥ 2l + 1, is 1 if n is odd, 2(n ∧ 2e−2 )
if n is even.
If n is even, l ≥ 1, e ≥ 2l + 1 ≥ 3. Since e ≥ 2l + 1, and n = 2l n0 for an odd n0 ,
l ≤ e−1 e−2 = n0 2l ∧ 2e−2 = 2l , and the number of solutions is 2l+1 ,
2 ≤ e − 2, so n ∧ 2
independent of e ≥ 2l + 1.
Conclusion : under the hypothesis xn ≡ a (mod 22l+1 ), where l = ord2 (n), then
x ≡ a (mod 2e ) is solvable for all e ≥ 1, and all these congruences have the same
n

number of solutions for e ≥ 2l + 1, e ≥ 3.

Chapter 5
5 3 6 −1





Ex. 5.1 Use Gauss’ lemma to determine 7 , 11 , 13 , p .
Proof. • a = 5, p = 7.
The array of values of the least residues modulo p = 7, for 1 ≤ k ≤ (p − 1)/2.
k mod 7 1 2 3
5k mod 7 −2 3 1
So the number of negative least residues is µ = 1, and 57 = (−1)µ = −1.

• a = 3, p = 11.
k mod 11 1 2 3 4 5
3k mod 11 3 −5 −2 1 4
3
= (−1)µ = 1.


So µ = 2, 11
• a = 6, p = 13.
k mod 13 1 2 3 4 5 6
6k mod 13 6 −1 5 −2 4 −3
6
= (−1)µ = −1.


So µ = 3, 13
• If a = −1, and p is an odd prime, the values of the least residues of −k modulo p
for k = 1, 2, . . . , (p − 1)/2
are −k, all negative. So the number of negative least residues
is µ = (p − 1)/2, and −1 p = (−1)
(p−1)/2 .

14
Ex. 5.2 Show that the number of solutions to x2 ≡ a (mod p) is equal to 1 + (a/p).

Proof. Let
N the number of solutions of x2 ≡ a (mod p).

• If ap = 0, then p | a, a ≡ 0 (mod p), so the unique solution of x2 ≡ a = 0 is x ≡ 0
(mod p), so N = 1 = 1 + ap .

• If ap = −1, then N = 0 = 1 + ap .

• If ap = 1, then x2 ≡ a (mod p) has a solution x0 , and x2 ≡ a (mod p) ⇐⇒ x2 ≡

x20 (mod p) ≡ p | (x − x0 )(x + x0 ) ≡ x ≡ ±x0 (mod p), so N = 2 = 1 + ap .

Ex. 5.3 Suppose p 6 | a. Show that the number of solutions to ax2 + bx + c ≡ 0 (mod p)
is equal to 1 + ((b2 − 4ac)/p).

Proof. Here p is an odd prime number, and p - a. Let N be the number of solutions of
ax2 + bx + c ≡ 0 (mod p)
For x ∈ Fp = Z/pZ,
 
2 2 b c
ax + bx + c = a x + x +
a a
 2 2
!
b b − 4ac
=a x+ −
2a 4a2
 2
b ∆
Let ∆ = b2 − 4ac. Then N is the number of solutions of x + 2a − 4a2
= 0 in Fp . As
in Ex.5.2, N = 1 if ∆ = 0, N = 0 if ∆ is not a square in F∗p , otherwise ∆ = δ 2 , δ ∈ F∗p ,
and the solutions are x = (−b ± δ)/2a, so N = 2. In the three cases, N = 1 + ∆


p .

Pp−1
Ex. 5.4 Prove that a=1 (a/p) = 0.

Proof. Here p is an odd prime (the result is false if p = 2). In the interval [1, p − 1],
there exist (p − 1)/2 residues, and (p − 1)/2 nonresidues (Prop. 5.1.2., Corollary 1), so
P p−1
a=1 (a/p) = 0.

p−1
a b
P


Proof. As an alternative proof, let S = p , and b a nonresidue modulo p : p = −1
a=1
(such a b exists if p 6= 2). As a 7→ ab is a bijection from F∗p to itself,
  p−1   p−1  
b X ab X c
S= = = S,
p p p
a=1 c=1

so −S = S, S = 0.

Ex. 5.5 Prove that p−1

P
x=1 ((ax + b)/p) = 0 provided that p - a.
There is a mistake
Pp−1 in the sentence : we must read
Prove that x=0 ((ax + b)/p) = 0 provided that p - a.
For instance,
5−1        
X x+1 2 3 4
= + + = −1 6= 0.
5 5 5 5
x=1

15
 0


Proof. From exercise 5.3, as p = 0, we know that

X  x Xp−1   p−1  
x X x
= = = 0.
p p p
x∈Fp x=0 x=1

x0
(This sum is well defined, since xp depends only of x : x ≡ x0 (mod p) ⇒ x




p = p .)

Fp → Fp
As a 6= 0 in Fp , f : is a bijection. Thus
x 7→ ax + b

p−1   X f (x)
X ax + b
=
p p
x=0 x∈Fp
X y 
= (y = f (x))
p
y∈Fp

=0

Ex. 5.6 Show that the number of solutions to x2 − y 2 ≡ a (mod p) is given by:
p−1   2 
X y +a
1+ .
p
y=0

Proof. Let S = {(x, y) ∈ F2p | x2 − y 2 = a}. From Ex.5.2,

X
|S| = Card {x ∈ Fp | x2 = y 2 + a}
y∈Fp
p−1   2 
X y +a
= 1+ .
p
y=0

Ex. 5.7 By calculating directly show that the number of solutions to x2 − y 2 ≡ a

(mod p) is p − 1 if p - a, and 2p − 1 if p | a. (Hint. Use the change of variables
u = x + y, v = x − y.)

Proof. Let S = {(x, y) ∈ F2p | x2 − y 2 = a}, and T = {(u, v) ∈ F2p | u v = a}. Then

S → T
f : is well defined (if (x, y) ∈ S, (x − y)(x + y) = a, so
(x, y) 7→ (x + y, x − y)
(x + y, x − y) ∈ T ). Moreover f is a bijection, with inverse (u, v) 7→ ((u + v)/2, (u − v)/2),
so |S| = |T |.
We compute |T |.
• Suppose that p - a, so a 6= 0. For v = 0, there is no solution, and for each v 6= 0,
we obtain the unique solution (a v −1 , v), so there exist p − 1 solutions.
• Suppose that p | a. The solutions of uv = 0 are (0, 0), (0, v) for each v 6= 0, (u, 0)
for each v = 0, that is to say N = 1 + (p − 1) + (p − 1) = 2p − 1 solutions.

16
 Conclusion :
Card {(x, y) ∈ F2p | x2 − y 2 = a} = p − 1 if p - a
= 2p − 1 if p | a

Ex. 5.8 Combining the results of Ex. 5.6 and 5.7 show that:
p−1  2  (
X y +a −1 if p - a
=
p p − 1 if p | a
y=0

Proof. Let S = {(x, y) ∈ F2p | x2 − y 2 = a}.

p−1
P 2

We obtain in Ex 5.6, |S| = 1 + y p+a , and in Ex. 5.7. , |S| = p − 1 if p - a,
y=0
|S| = 2p − 1 if p | a.
So
p−1  2  (
X y +a −1 if p - a
|S| − p = =
p p − 1 if p | a
y=0

Ex. 5.9 Prove that 12 32 · · · (p − 2)2 ≡ (−1)(p+1)/2 (mod p) using Wilson’s theorem.

Proof. Here p is an odd prime.

From Wilson’s theorem, as k(p − k) ≡ −k 2 (mod p) for k = 1, 2, . . . , p − 1,
−1 ≡ (p − 1)!
     
p−1 p+1
≡ 1 × 2 × ··· × k × ··· × × × · · · × (p − k) · · · × (p − 2) × (p − 1)
2 2
(p−1)/2
Y
≡ k(p − k)
k=1
(p−1)/2
Y
(p−1)/2
≡ (−1) k2
k=1
  2
(p−1)/2 p−1
≡ (−1) ! (mod p)
2
So  2

p−1
! ≡ (−1)(p+1)/2 (mod p).
2
Moreover, from Wilson’ theorem and Fermat’s little theorem,
12 22 32 · · · (p − 1)2 = [(p − 1)!]2 ≡ 1 (mod p)
  2   2
2 2 2 p−1 2

p−1 p−1
2 4 · · · (p − 1) = 2 ! ≡ ! (mod p)
2 2

17
Thus   2
p−1
12 32 · · · (p − 2)2 ! ≡1 (mod p),
2
which gives
12 32 · · · (p − 2)2 ≡ (−1)(p+1)/2 (mod p).

Ex. 5.10 Let r1 , r2 , . . . , r(p−1)/2 be the quadratic residues between 1 and p. Show that
their product is congruent to 1 (mod p) if p ≡ 3 (mod 4), and to −1 if p ≡ 1 (mod 4).

Proof. We proved in Ex. 5.9 that

  2
p−1
! ≡ (−1)(p+1)/2 (mod p).
2

{1, 2, . . . , (p − 1)/2} →
7 {r1 , r2 , . . . , r(p−1)/2 }
The application f : is a bijection, so
x 7 → x2

(p−1)/2   2
Y p−1
ri ≡ ! (mod p),
2
i=1

so
(p−1)/2
Y
ri ≡ (−1)(p+1)/2 (mod p).
i=1

That is to say, the product of the quadratic residues between 1 and p is congruent to 1
(mod p) if p ≡ 3 (mod 4), and to −1 if p ≡ 1 (mod 4).

Ex. 5.11 Suppose that p ≡ 3 (mod 4), and that q = 2p + 1 is also prime. Prove that
2p − 1 is not prime. (Hint : Use the quadratic character of 2 to show that q | 2p − 1)
One must assume that p > 3.

Proof. The result is false if p = 3, so we must suppose p > 3.

p = 4k + 3 for an integer k, so q = 2p + 1 = 8k + 7 ≡ −1 (mod 8). Thus
 
2 2
= (−1)(q −1)/8 = 1.
q

So 2(q−1)/2 ≡ 1 (mod q), 2p ≡ 1 (mod q), so q | 2p − 1.

Moreover, as p > 3, q = 2p + 1 < 2p − 1
(2p + 1 < 2p − 1 ⇐⇒ 2p < 2p − 2 ⇐⇒ p + 1 < 2p−1 .
4 + 1 < 24−1 and for all k ≥ 4, k + 1 < 2k−1 implies k + 2 < 2k−1 + 1 ≤ 2k , and
4 + 1 < 24−1 , so by induction k + 1 < 2k−1 for all k > 3.
So q | 2p − 1 with 1 < q < 2p − 1 : 2p − 1 is composite.
Conclusion : if p ≡ 3 (mod 4), p > 3 is prime, and q = 2p + 1 is also prime, then
p
2 − 1 is not a prime.
For instance, le Mersenne’s number 211 −1 = 2047 is not a prime : 2047 = 23×89.

18
Ex. 5.12 Let f (x) ∈ Z[x]. We say that a prime p divides f (x) if there’s an integer n
such that p | f (n). Describe the prime divisors of x2 + 1 and x2 − 2.
2 2

p divides x + 1 iff there exists a ∈ Z such that −1 ≡ a (mod p), iff p = 2 or
Proof.
−1
p = 1 iff p = 2 or p ≡ 1 (mod 4).
p divides x2 − 2 iff there exists a ∈ Z such that 2 ≡ a2 (mod p), iff p = 2 or p2 = 1

iff p = 2 or p ≡ ±1 (mod 8).

Ex. 5.13 Show that any prime divisor of x4 − x2 + 1 is congruent to 1 modulo 12.

Proof. • As a6 + 1 = (a2 + 1)(a4 − a2 + 1), p | a4 − a2 + 1 implies p | a6 + 1, so −1

p =1
and p ≡ 1 (mod 4).
• p | 4a4 − 4a2 + 4 = (2a − 1)2 + 3, so −3


p = 1.
−3

p
p

As −3 ≡ 1 (mod 4), p = 3 , so 3 = 1, thus p ≡ 1 (mod 3).
4 | p − 1 and 3 | p − 1, thus 12 | p − 1 :

p≡1 (mod 12).

Ex. 5.14 Use the fact that U (Z/pZ) is cyclic to give a direct proof that (−3/p) = 1
when p ≡ 1 (mod 3).[Hint : There is a ρ in U (Z/pZ) of order 3. Show that (2ρ + 1)2 =
−3.]

Proof. Suppose that p ≡ 1 (mod 3). Let g a generator of F∗p . Then g has order p − 1,
thus ρ = g (p−1)/3 has order 3. As ρ3 = 1, ρ 6= 1, then ρ2 + ρ + 1 = 0.

(2ρ + 1)2 = 4ρ2 + 4ρ + 1

= 4(ρ2 + ρ + 1) − 3
= −3.
−3


Thus p = 1.
−3
F∗p


The converse is also true for an odd prime p : if p = 1, then there exists a ∈
−1+a 2
such that −3 = a2 . ρ = 2 has order 3. Indeed ρ = 1+a4−2a = −2−2a
2
4 = −1−a
2 , so

−1 + a −1 − a
1 + ρ + ρ2 = 1 + +
2 2
=0

so ρ 6= 1, ρ3 = 1. The group F∗p contains an element of order 3, thus from Lagrange’s

theorem 3 | p − 1 : p ≡ 1 (mod 3).

Ex. 5.15 If p ≡ 1 (mod 5), show directly that (5/p) = 1 by the method of Ex. 5.14.
[Hint : Let ρ be an element of U (Z/pZ)) of order 5. Show that (ρ+ρ4 )2 +(ρ+ρ4 )−1 = 0,
etc.]

19
Proof. Let g a generator of F∗p . g has order p − 1, thus ρ = g (p−1)/5 has order 5.
Let
α = ρ + ρ4


β = ρ2 + ρ3
As 0 = ρ5 − 1 = (ρ − 1)(1 + ρ + ρ2 + ρ3 + ρ4 ) and ρ 6= 1, then 1 + ρ + ρ2 + ρ3 + ρ4 = 0,
thus

α + β = −1
αβ = ρ3 + ρ4 + ρ + ρ2 = −1

So α, β are the roots in Fp of x2 + x − 1 : α2 + α − 1 = 0.

Thus 4α2 + 4α − 4 = (2α + 1)2 − 5 = 0 : 5 is a square in F∗p and 5


p = 1.

Ex. 5.16 Using quadratic reciprocity find the primes for which 7 is quadratic residue.
Do the same for 15.

Proof. 7 is a quadratic residue for 2 and for the odd primes such that p7 = 1.

From the law of quadratic reciprocity,

   
7 (p−1)/2 p
= 1 ⇐⇒ (−1) =1
p 7

iff either p ≡ 1 (mod 4) and p7 = 1, or p ≡ −1 (mod 4) and p7 = −1.

In the first case , p ≡ 1 (mod 4), p ≡ 1, 4, 2 (mod 7), which gives p ≡ 1, −3, 9
(mod 28).
In the second case, p ≡ −1 (mod 4), p ≡ −1, −4, −2 (mod 7), which gives p ≡
−1, 3, −9 (mod 28).
Conclusion : the primes for which 7 is a quadratic residue are 2 and the odd primes
p such that  
7
= 1 ⇐⇒ p ≡ ±1, ±3, ±9 (mod 28).
p

15


15 is a quadratic residue for 2 and for the odd primes such that p = 1.
         
15 3 5 3 5
= 1 ⇐⇒ = = 1 or = = −1
p p p p p
From the examples of theorem 2, we know that
   
3 3
= 1 ⇐⇒ p ≡ 1, −1 (mod 12), = −1 ⇐⇒ p ≡ 5, −5 (mod 12),
p p
   
5 5
= 1 ⇐⇒ p ≡ 1, −1 (mod 5), = −1 ⇐⇒ p ≡ 2, −2 (mod 5).
p p
As 5 ∧ 12 = 1, there exist 8 cases, all possible, which give
 
15
= 1 ⇐⇒ p ≡ ±1, ±7, ±11, ±17 (mod 60).
p
For instance, the primes 2, 7, 11, 17, 43, 53, 59, 61, 67, 137, . . . are suitable.

20
Ex. 5.17 Supply the details to the proof of Proposition 5.2.1 and to the corollary to
the lemma following it.
Proposition 5.2.1

(a) (a1 /b) = (a2 /b) if a1 ≡ a2 (mod b).

(b) (a1 a2 /b) = (a1 /b)(a2 /b).

(c) (a/b1 b2 ) = (a/b1 )(a/b2 ).

Proof. (a) Let b = p1 p2 · · · pm , where the pi are not necessarily

Q distinct
Q primes. For
each prime pi , (a1 , pi ) = (a2 , pi ) (Prop. 5.1.2 (c)), so i (a1 , pi ) = i (a2 , pi ), thus
(a1 /b) = (a2 /b).

(b) From Prop. 5.1.2(b),

Q Q Q Q
(a1 a2 /b) = i (a1 a2 /pi ) = i (a1 /pi )(a2 /pi ) = i (a1 /pi ) i (a2 /pi ) = (a1 /b)(a2 /b).

(c) Let b1 = p1 p2 · · · pm , b2 = q1 q2 · · · ql . Then b1 b2 = p1 p2 · · · pm q1 q2 · · · ql = m+l

Q
i=1 ri ,
where ri = pi for i = 1, . . . , m, ri = qi−m for i = m + 1, . . . , m + l. Then
(a/b1 b2 ) = m+l
Q Qm Ql
i=1 (a/ri ) = i=1 (a/pi ) j=1 (a/qi ) = (a/b1 )(a/b2 ).

Lemma. Let r and s be odd integers. Then

(a) (rs − 1)/2 ≡ ((r − 1)/2) + ((s − 1)/2) (mod 2).

(b) (r2 s2 − 1)/8 ≡ ((r2 − 1)/8) + ((s2 − 1)/8) (mod 2).

(Proof in the book.)

Corollary. Let r1 , r2 , . . . , rm be odd integers. Then

Pm
(a) i=1 (ri − 1)/2 ≡ (r1 r2 · · · rm − 1)/2 (mod 2).
Pm 2 2 2 2
(b) i=1 (ri − 1)/8 ≡ (r1 r2 · · · rm − 1)/8 (mod 2).

Proof. Let P(m) the proposition defined by

m
X
P(m) ⇐⇒ (ri − 1)/2 ≡ (r1 r2 · · · rm − 1)/2 (mod 2).
i=1

Then P(1) ⇐⇒ (r1 − 1)/2 ≡ (r1 − 1)/2 (mod 2) is true, and P(2) is part (a) of the
lemma. If we make the induction hypothesis P(m), then
m+1
X m
X
(ri − 1)/2 = (ri − 1)/2 + (rm+1 − 1)/2
i=1 i=1
≡ (r1 r2 · · · rm − 1)/2 + (rm+1 − 1)/2 (mod 2)
≡ (r1 r2 · · · rm rm+1 − 1)/2 (mod 2),

where the last congruence is a consequence of the part (a) of the Lemma : the induction
is completed, and P(m) is true for all m ≥ 1.
The proof of part (b) is similar.

21
Ex. 5.18 Let D be a square-free integer that is also odd and positive. Show that there
is an integer b prime to D such that (b/D) = −1.

Proof. Let D = p1 p2 · · · pk , where the pi are distinct odd primes.

Let s a nonresidue modulo pk . From Chinese remainder theorem, as pi ∧ pj = 1 if
i 6= j, there exists an integer b such that

b≡1 (mod p1 ), b ≡ 1 (mod p2 ), · · · , b ≡ 1 (mod pk−1 ), b ≡ s (mod pk ).

Then (b/pi ) = 1, i = 1, 2, . . . k − 1, (b/pk ) = −1, so b ∧ pi = 1 for all i = 1, 2, . . . , k. Then

b ∧ D = b ∧ p1 · · · pk = 1 , and
  Yk    
b b b
= = = −1.
D pi pk
i=1

P
Ex. 5.19 Let D be as in Exercise 18. Show that (a/D) = 0, where the sum is over
a reduced residue system modulo D. Conclude that exactly one half of the elements in
U (Z/DZ) satisfy (a/D) = 1.

Proof. Let b such that (b/D) = −1 and b ∧ D = 1 : the existence of b comes from Ex
5.18. P
Let S = a∈A (a/D), where A is reduced residue system modulo D. As two reduced
system modulo D represent the same elements in U (Z/DZ), the sum is independent of
the reduced residue system A : we can write
X
S= (a/D).
a∈U (Z/DZ)

As b ∧ D = 1, we know from Ex. 3.6 that B = bA = {ba | a ∈ A} is also a reduced

system modulo D. In other words, the application U (Z/DZ) → U (Z/DZ), a 7→ ab is a
bijection, so
       X c
b X b a X ba
S= = = =S (c = ab).
D D D D D
a∈U (Z/DZ) a∈U (Z/DZ) c∈U (Z/DZ)

As (b/D) = −1, −S = S, so S = 0.
Since (a/D) = ±1, one half of the elements in U (Z/DZ) satisfy (a/D) = 1, and one
half of the elements in U (Z/DZ) satisfy (a/D) = −1.

Ex. 5.20 (continuation) Let a1 , a2 , . . . , aφ(D)/2 be integers between 1 and D such that
(ai , D) = 1 and (ai /D) = 1. Prove that D is a quadratic residue modulo a prime p 6 | D,
p ≡ 1 (mod 4) iff p ≡ ai (mod D) for some i.

Proof. From Ex. 5.19 we know that there exist exactly φ(D)/2 integers ai between 1
and D such that ai ∧ D = 1 and (ai /D) = 1. So {a1 , . . . , aφ(D)/2 } is the set of all
a ∈ U (Z/DZ) such that (a/D) = 1.
Let D = p1 p2 · · · pk , with distinct pi , and p a prime number, p ≡ 1 (mod 4), p 6∈
{p1 , . . . , pk } (so p = 4k + 1, k ∈ N).

22
 (⇐) Suppose that p ≡ ai for some i, 1 ≤ i ≤ φ(D)/2, then (p/D) = (ai /D) = 1, so
(Prop. 5.2.2)
       
D p−1 D−1 p 2k( D−1 ) p p
= (−1) 2 2 = (−1) 2 = = 1.
p D D D
(⇒) Suppose that D is a quadratic residue modulo p. Then (D/p) = 1, so
   
p p−1 D−1 D
= (−1) 2 2 = 1.
D p
Thus p ∈ {a1 , . . . , aφ(D)/2 } since {a1 , . . . , aφ(D)/2 } is the set of all a ∈ U (Z/DZ) such
that (a/D) = 1. Consequently p ≡ ai (mod D) for some i.

Ex. 5.21 Apply the method of Ex. 5.19 and 5.20 to find those primes for which 21 is
a quadratic residue.

Proof. Let D = 21 = 3 × 7 (D is positive, odd and square-free). We first search the

φ(D)/2 = 6 integers a, 1 ≤ a ≤ 21, such that (a/D) = 1.
         
a a a a a
= 1 ⇐⇒ = = 1 or = = −1.
21 3 7 3 7
The first case is equivalent to a ≡ 1 (mod 3), a ≡ 1, 2, 4 (mod 7), that is a ≡ 1, 16, 4
(mod 21).
The second case gives a ≡ −1 (mod 3), a ≡ −1, −2, −4 (mod 7), that is a ≡ −1, −16, −4
(mod 21), or equivalently a ≡ 20, 5, 17 (mod 21).
So A = {1, 4, 5, 16, 17, 20} is the set of the integers a such that 1 ≤ a ≤ 21, (a/D) = 1.
As (21/3) = (21/7) = 0, 21 is not a quadratic residue modulo 3 or 7.
• p ≡ 1 (mod 4).
From Ex.5.20, we know that D = 21 is a quadratic residue modulo an odd prime p,
p 6= 3, p 6= 7, p ≡ 1 (mod 4), iff p ≡ a (mod D) for some a ∈ A.
• p ≡ −1 (mod 4).

p
p−1 D−1
As D = 21 ≡ 1 (mod 4), D p D = (−1)
2 2 = 1, so the same reasoning as in Ex.
5.20 show that D is a quadratic residue modulo 21 iff p ≡ a, a ∈ A.
Conclusion : 21 is a quadratic residue for 2, and for the primes p such that
p ≡ 1, 4, 5, 16, 17, 20 (mod 21).

Ex. 5.22 Use the Jacobi symbol to determine (113/997), (215/761), (514/1093), and
(401/757).

Proof. (113/997) = (997/113) = (93/113) = (113/93) = (20/93) = (22 /93)(5/93) =

(5/93) = (93/5) = (3/5) = (5/3) = (2/3) = −1.
(215/761) = (761/215) = (116/215) = (22 /215)(29/215) = (29/215) = (215/29) =
(12/29) = (22 /29)(3/29) = (3/29) = (29/3) = (2/3) = −1.
(514/1093) = (2/1093)(257/1093) = −(257/1093) = −(1093/57) = −(65/257) =
−(257/65) = −(62/65) = −(2/65)(31/65) = −(31/65) = −(65/31) = −(3/31) =
(31/3) = (1/3) = 1.
(401/757) = (757/401) = (356/401) = (401/89) = (45/89) = (89/45) = (44/45) =
(22 /45)(11/45)
= (11/45) = (45/11) = (1/11) = 1.

23
Ex. 5.23 Suppose that p ≡ 1 (mod 4). Show that there exist integers s and t such
that pt = 1 + s2 . Conclude that p is not a prime in Z[i]. Remember that Z[i] has unique
factorization.
p−1
Proof. As p ≡ 1 (mod 4), then −1


p = (−1)
2 = 1 : −1 is a square modulo p.
So −1 ≡ s (mod p), s ∈ Z : there exist s ∈ Z, t ∈ Z such that pt = 1 + s2 .
2

In Z[i], p|(s + i)(s − i).

If p was a prime in Z[i], then p | s + i ou p | s − i.
This implies s ± i = (a + bi)p, (a, b) ∈ Z2 , thus ±1 = bp, p | 1 : it’s impossible.
Conclusion : if p ≡ 1 (mod 4), p is not a prime in Z[i].

Ex. 5.24 If p ≡ 1 (mod 4), show that p is a sum of two squares, i.e. p = a2 + b2 with
a, b ∈ Z.(Hint : p = αβ, with α and β being non units in Z[i]. Remember that Z[i] has
unique factorisation.)

Proof. Z[i] is a principal ideal domain, thus p prime is in Z[i] iff p is irreducible in Z[i].
If p ≡ 1 (mod 4), p is not a prime from Ex.5.23, so it is not irreducible :
p = αβ, α, β ∈ Z[i], N (α) > 1, N (β) > 1 (where N (a + bi) = a2 + b2 is the complex
norm).
N (p) = p2 = N (u)N (v), 1 < N (u) < p2
Thus N (u) = p, that is p = a2 + b2 , where u = a + bi.
Conclusion : if p is prime in N, p ≡ 1 (mod 4), then p = a2 + b2 , a, b ∈ Z, p is a sum
of two squares.

Ex. 5.25 An integer is called a biquadratic residue modulo p if it is congruent to a

fourth power. Using the identity x4 + 4 = ((x + 1)2 + 1)((x − 1)2 + 1) show that −4 is a
biquadratic residue modulo p iff p ≡ 1 (mod 4).

Proof. x4 + 4 = (x4 + 4x2 + 4) − 4x2 = (x2 + 2)2 − 4x2 = (x2 + 2 − 2x)(x2 + 2 + 2x), so

x4 + 4 = ((x − 1)2 + 1)((x + 1)2 + 1).

If −4 ≡ x4 [p], then p | (x + 1)2 + 1 or p | (x − 1)2 + 1
In the two cases, −1 is a quadratic residue modulo p, thus −1


p = 1 : p ≡ 1 [4].
−1
Conversely, if p ≡ 1 [4], p = 1, then it exists an integer a such that −1 ≡ a2 [p].

Let x = a − 1. Then p | (x + 1)2 + 1, thus p | x4 + 4 : −4 is a biquadratic residue

modulo p.
Conclusion :
∃x ∈ Z, x4 ≡ −4 [p] ⇐⇒ p ≡ 1 [4].

Ex. 5.26 This exercise and Ex. 5.27 and 5.28 give Dirichlet’s beautiful proof that 2
is a biquadratic residue modulo p iff p can be written in the form A2 + 64B 2 , where
A, B ∈ Z. Suppose that p ≡ 1 (mod 4). Then p = a2 + b2 by Ex. 5.24. Take a to be odd.
Prove the following statements:

(a) (a/p) = 1.
2 −1)/8
(b) ((a + b)/p) = (−1)((a+b) .

24
 (c) (a + b)2 ≡ 2ab (mod p)

(d) (a + b)(p−1)/2 ≡ (2ab)(p−1)/4 (mod p).

Proof. Let p a prime number, p ≡ 1 [4] : p = 4k + 1, k ∈ N∗ .

Then p = a2 + b2 (Ex. 5.24).
As a, b are not of the same parity, up to exchange a and b, we will suppose that a is
odd (then b is even).
(a)  
a p−1
≡ a 2 = a2k [p].
p
Using the law of quadratic reciprocity for Jacobi’s symbol (Proposition 5.2.2), where
a, p are odd numbers :
     
a p p−1 a−1 p
= (−1) 2 2 = ,
p a a
since p ≡ 1 [4].
If a = p1 p2 · · · pl is the decomposition of a in prime factors, with not necessarily
distinct primes , then
      
p p p p
= ··· .
a p1 p2 pl
Since p = a2 + b2 , p ≡ b2 [pi ], thus ppi = 1 for all i.

   
a p
= = 1.
p a

(b) a + b is odd, and p ≡ 1 [4], thus

     2    
a+b p 2 p 2 2p
= = = .
p a+b a+b a+b a+b
2p

If a + b = q1 q2 · · · ql , as 2p = (a + b)2 + (a − b)2 , 2p ≡ (a − b)2 [qi ], thus qi = 1.
     
2p 2p 2p
= ··· = 1.
a+b q1 ql
(a+b)2 −1
2


Moreover a+b = (−1) 8 , so
 
a+b (a+b)2 −1
= (−1) 8 .
p

(c) (a + b)2 = a2 + b2 + 2ab = p + 2ab ≡ 2ab [p]

p−1 p−1
(d)[(a + b)2 ] 4 ≡ (2ab) 4 [p], thus
p−1 p−1
(a + b) 2 ≡ (2ab) 4 [p].

25
Ex. 5.27 Suppose that f is such that b ≡ af (mod p). Show that f 2 ≡ −1 (mod p),
and that 2(p−1)/4 ≡ f ab/2 (mod p).

Proof. Let f such as b ≡ af [p].

This is equivalent to f = ba−1 in F∗p .
2 2
As a2 = −b , f = −1 : f 2 ≡ −1 [p].
We deduce from Ex. 5.26 (d) and (b) that
 
p−1 p−1 a+b
(2ab) 4 ≡ (a + b) 2 =
p
(a+b)2 −1
≡ (−1) 8

(a+b)2 −1
≡ (f 2 ) 8

(a+b)2 −1 a2 +b2 −1+2ab

≡f 4 =f 4

p−1 ab
≡f 4 f 2 (mod p)
p−1
a


Since a 2 = p = 1 from Ex. 5.26(a)), then
p−1 p−1 p−1 p−1 p−1
(ab) 4 ≡ (a2 f ) 4 ≡a 2 f 4 ≡f 4 [p],

so p−1 p−1 p−1

ab
2 4 f 4 ≡f 2 f 4 [p].
p−1
As f 4 6≡ 0 [p],
p−1 ab
2 4 ≡f 2 [p].

Ex. 5.28 Show that x4 ≡ 2 (mod p) has a solution for p ≡ 1 (mod 4) iff p is of the
form A2 + 64B 2 .

Proof. If p ≡ 1 [4] and if there exists x ∈ Z such that x4 ≡ 2 [p], then

p−1
2 4 ≡ xp−1 ≡ 1 [p].

From Ex. 5.27, where p = a2 + b2 , a odd, we know that

ab p−1
f 2 ≡2 4 ≡ 1 [p].

Since f 2 ≡ −1 [p], the order of f modulo p is 4, thus 4 | ab

2 , so 8 | ab.
As a is odd, 8|b, then p = A2 + 64B 2 (with A = a, B = b/8).

Conversely, if p = A2 + 64B 2 , then p ≡ A2 ≡ 1 [4].

Let a = A, b = 8B. Then
p−1 ab
2 4 ≡f 2 ≡ f 4AB ≡ (−1)2AB ≡ 1 [p].
p−1
As 2 4 ≡ 1 [p], x4 ≡ 2 [p] has a solution in Z (Prop. 4.2.1) : 2 is a biquadratic
residue modulo p.
Conclusion :

26
 ∃A ∈ Z, ∃B ∈ Z , p = A2 + 64B 2 ⇐⇒ (p ≡ 1 [4] and ∃x ∈ Z, x4 ≡ 2 [p]).

Note : the equation x4 ≡ 2 [p] has also solutions if p ≡ −1 [8].

p−1
Indeed, the equation x4 ≡ 2 [p] has a solution in Z iff 2 d = 1, where d = 4∧(p−1) =
p−1
2, thus iff 2 2 ≡ 1 [p], which is true since p2 = 1.

For instance, 84 ≡ 2 (mod 23), with 23 ≡ −1 (mod 8).

Ex. 5.29 Let (RR) be the number of pairs (n, n + 1) in the set 1, 2, 3, . . . , p − 1 such
that n and n + 1 are both quadratic residues modulo p. Let (N R) be the number of pairs
(n, n + 1) in the set 1, 2, 3, . . . , p − 1 such that n is a quadratic nonresidue and n + 1 is a
quadratic residue. Similarly, define (RN ) and (N N ). Determine the sums (RR)+(RN ),
(N R) + (N N ), (RR) + (N R), and (RN ) + (N N ).

Proof. Let E the set of pairs (n, n + 1) ∈ N2 , 1 ≤ n ≤ p − 2 : |E| = p − 2.

Write RR the set of pairs (n, n + 1) such that n and n + 1 are both a quadratic
residues, and (RR) = |RR| its cardinality, and similar definitions for RN, N R, N N .
As E = RR ∪ RN ∪ N R ∪ N N (disjoint union),

(RR) + (RN ) + (N R) + (N N ) = |E| = p − 2.

• RR ∪ RN is the set of pairs (n, n + 1) in E such that n is a residue. Its cardinality

is the number of residues in [1, p − 2], so the number of residues in [1, p − 1], minus s,
where s = 1 if p − 1 is a residue, s = 0 otherwise. In both cases p ≡ 1, 3 (mod 4),
p−1
1+(−1) 2
s= 2 , and the total number of residues is (p − 1)/2, so
p−1
p−1 p − 1 1 + (−1) 2 1 p−1
(RR) + (RN ) = −s= − = (p − 2 − (−1) 2 ).
2 2 2 2
• Similarly, (N R) + (N N ) is the number of nonresidues in [1, p − 1], minus t, where
p−1
1−(−1) 2
t = 1 if p − 1 is a nonresidue, t = 0 otherwise : t = 2 , so

1 p−1
(N R) + (N N ) = (p − 2 + (−1) 2 )
2
(the sum of these two results is indeed p − 2 = |E|).
• As 1 is a residue, (RR) + (N R) is the number of residues in [1, p − 1], minus 1 :
p−1
(RR) + (N R) = − 1.
2
• (RN ) + (N N ) is the number of nonresidues in [2, p − 1], equal to the number of
residues in [1, p − 1] :
p−1
(RN ) + (N N ) = .
2

27
Ex. 5.30 Show that (RR) + (N N ) − (RN ) − (N R) = p−1
P
n=1 (n(n + 1)/p). Evaluate
this sum and show that it is equal to −1. (Hint : The result of Exercise 8 is useful.)

Proof. Let χ be the characteristic function of RR ∪ N N : if 1 ≤ n ≤ p − 1, χ(n) = 1 if

n, n + 1 are both residues, or if n, n + 1 are both non residues. Then
   
1 n n+1
χ(n) = 1+
2 p p

(if χ(n) = 1, np n+1 = 1, and np n+1

p p = −1 otherwise.)
Similarly, let χ0 the characteristic function of RN ∪ N R : χ0 (n) = 1 if exactly one of
the integer n, n + 1 is a residue, 0 otherwise. Then
   
0 1 n n+1
χ (n) = 1− .
2 p p

As each integer n between 1 and p − 1 brings the contribution 1 if n ∈ RR ∪ N N , and

−1 if n ∈ RN ∪ N R, then
p−1
X
(RR) + (N N ) − (RN ) − (N R) = (χ(n) − χ0 (n))
n=1
p−1      
1 X n(n + 1) n(n + 1)
= 1+ − 1−
2 p p
n=1
p−1  
X n(n + 1)
=
p
n=1

To evaluate this sum S, note that 4n(n + 1) = (2n + 1)2 − 1, so

p−1  p−1  p−1 
(2n + 1)2 − 1
  
X n(n + 1) X 4n(n + 1) X
S= = = .
p p p
n=1 n=1 n=1

This sum can be written S = n∈F∗p ((2n + 1)2 − 1)/p) = n∈Fp ((2n + 1)2 − 1)/p), since
P P

(0/p) = 0. As f : Fp → Fp , n 7→ (2n + 1) is a bijection (2 is invertible in F∗p ),

X (2n + 1)2 − 1 X y 2 − 1
= (y = 2n + 1).
p p
n∈Fp y∈Fp

As p - 1, the evaluation of this last sum is given in Exercise 5.8 : S = −1, so

p−1  
X n(n + 1)
(RR) + (N N ) − (RN ) − (N R) = = −1.
p
n=1

Ex. 5.31 Use the results of Exercises 29 and 30 to show that (RR) = 14 (p − 4 − ε),
where ε = (−1)(p−1)/2

28
Proof. To summarize the results of the Ex. 5.29 and 5.30,

(a)(RR) + (RN ) + (N R) + (N N ) = p − 2
(b)(RR) + (N N ) − (RN ) − (N R) = −1

and
1 p−1

(c)(RR) + (RN ) = p − 2 − (−1) 2
2
p−1
(d)(RR) + (N R) = −1
2

The sum of (a) and (b) gives

p−3
(e)(RR) + (N N ) = .
2
The sum of (c),(d),(e) gives (using (a))
p−1
p−2 p−1 p−3 (−1) 2
2(RR) + p − 2 = + + −1− ,
2 2 2 2
so p−1 p−1
p−1 p−3 p−2 (−1) 2 p (−1) 2
2(RR) = + − −1− = −2− ,
2 2 2 2 2 2
1 p−1
(RR) = (p − 4 − ε), where ε = (−1) 2 .
4

Q(p−1)/2
Ex. 5.32 If p is an odd prime, show that (2/p) = j=1 2 cos(2πj/p). Use this to
give another proof to Proposition 5.1.3.

Proof. Let p an odd prime number, and ζ = e2iπ/p : then ζ p = 1.

Let
p−1
Y p−1
Y
P = (ζ j + ζ −j ) = 2 cos(2πj/p).
j=0 j=0

p−1
Y
P = ζ 0 ζ −1 · · · ζ −(p−1) (ζ 2j + 1)
j=0
p−1
Y
= (ζ p )−(p−1)/2 (ζ 2j + 1)
j=0
p−1
Y
= (ζ 2j + 1)
j=0

As ζ j depends only of the class j ∈ Fp , this product can be written

Y Y
P = (ζ 2j + 1) = (ζ k + 1) (k = 2j),
j∈Fp k∈Fp

29
since f : Fp → Fp , x 7→ 2x is a bijection. So
p−1
Y
P = (ζ k + 1).
k=0

Since ζ 0 = 1, ζ, · · · , ζ p−1 are the roots of the polynomial f (x) = xQ

p −1, then 1+ζ 0 , · · · , 1+

ζ p−1 are the roots of g(x) = (x − 1)p − 1 = f (x − 1), so g(x) = p−1 k

k=0 (x − (1 + ζ )).
p−1
As g(0) = (−1)p − 1 = −2 = (−1 − ζ 0 ) · · · (−1 − ζ p−1 ) = − k=0 (ζ k + 1), we obtain
Q

p−1
Y p−1
Y
P = 2 cos(2πj/p) = (ζ k + 1) = 2,
j=0 k=0

so
p−1
Y
2 cos(2πj/p) = 1.
j=1

p−1
Y
1= 2 cos(2πj/p)
j=1
(p−1)/2 p−1
Y Y
= 2 cos(2πj/p) 2 cos(2πj/p)
j=1 j=(p+1)/2
(p−1)/2 (p−1)/2
Y Y
= 2 cos(2πj/p) 2 cos(2π − 2πk/p) (k = p − j)
j=1 k=1

As cos(2π − α) = cos(α),
 2
(p−1)/2 (p−1)/2
Y Y
1= 2 cos(2πj/p) , so 2 cos(2πj/p) = ±1
j=1 j=1

Case 1 : if 1 ≤ j ≤ p/4, 0 ≤ 2πj/p < π/2, so cos(2πj/p) > 0,

case 2 : if p/4 < j ≤ (p − 1)/2, π/2 < 2πj/p < π, so cos(2πj/p) < 0.
In the first case, 2 ≤ 2j ≤ (p − 1)/2 : the least residue of 2j is positive. In the second
case p/2 < 2j ≤ p − 1 : the least residue of 2j is negative.
Let µ the number of negative least residues of the integer 2j, 1 ≤ j ≤ (p − 1)/2 : we
know from Gauss’ Lemma that (2/p) = (−1)µ . As µ is also the number of j, 1 ≤ j ≤
(p − 1)/2 such that cos(2πj/p) < 0,
(p−1)/2  
Y 2 µ
2 cos(2πj/p) = (−1) = .
p
j=1

If p ≡ 1 [8] , p = 8q + 1, q ∈ N. For 1 ≤ j ≤ (p − 1)/2,

cos(2πj/p) < 0 ⇐⇒ p/4 ≤ j ≤ (p − 1)/2 ⇐⇒ 2q + 1 ≤ j ≤ 4q,

so µ = 2q and (2/p) = (−1)µ = 1.

30
 If p ≡ −1 [8], p = 8q − 1, q ∈ N∗ .

cos(2πj/p) < 0 ⇐⇒ p/4 ≤ j ≤ (p − 1)/2 ⇐⇒ 2q ≤ j ≤ 4q − 1,

so µ = 2q and (2/p) = (−1)µ = 1.

If p ≡ 3 [8], p = 8q + 3, q ∈ N.

cos(2πj/p) < 0 ⇐⇒ p/4 ≤ j ≤ (p − 1)/2 ⇐⇒ 2q + 1 ≤ j ≤ 4q + 1,

so µ = 2q + 1 and (2/p) = (−1)µ = 1.

If p ≡ −3 [8], p = 8q − 3, q ∈ N∗ ,

cos(2πj/p) < 0 ⇐⇒ p/4 ≤ j ≤ (p − 1)/2 ⇐⇒ 2q ≤ j ≤ 4q − 2,

so µ = 2q − 1 and (2/p) = (−1)µ = 1.

Ex. 5.33 Use Proposition 5.3.2 to derive the quadratic character of −1.

Proof. Let f (z) = e2πiz − e−2πiz . If p is an odd prime, a ∈ Z, and p - a, we know from
Prop. 5.3.2 that
(p−1)/2     (p−1)/2  
Y la a Y l
f = f .
p p p
l=1 l=1

For a = −1, as f (−z) = −f (z),

  (p−1)/2   (p−1)/2  
−1 Y l Y −l
f = f
p p p
l=1 l=1
(p−1)/2  
Y l
= (−1)(p−1)/2 f
p
l=1

(z) = 0 ⇐⇒ e4πiz = 1 ⇐⇒ 4πiz = 2kiπ, k ∈ Z ⇐⇒ z = k/2, k ∈ Z, so, if

Moreover f 
l ∈ Z, f pl = 0 ⇐⇒ l/p = k/2, k ∈ Z ⇐⇒ p | 2l ⇐⇒ p | l. For 1 ≤ l < p, this is
Q(p−1)/2  l 
impossible, so l=1 f p 6= 0. Consequently,
 
−1
= (−1)(p−1)/2 .
p

Ex. 5.34 If p is an odd prime distinct from 3, show that

  (p−1)/2
Y   
3 2 2πj
= 3 − 4 sin .
p p
j=1

31
Proof. Let p an odd prime number, p 6= 3 and ζ = e2iπ/p .
2
ζ j − ζ −j
  
2 2πj
3 − 4 sin =3−4
p 2i
= 3 + ζ 2j + ζ −2j − 2
= 1 + ζ 2j + ζ −2j
 
4πj
= 1 + 2 cos
p

(As cos(2α) = 1 − 2 sin2 α, so 3 − 4 sin2 α = 1 + 2 cos α.)

Let
p−1
Y   Y  
2 2πj 2 2πj
P = 3 − 4 sin = 3 − 4 sin .
p p
j=1 ∗ j∈Fp

Then
Y
1 + ζ 2j + ζ −2j


P =
j∈F∗p
Y  
= 1 + ζ k + ζ −k (k = 2j)
k∈F∗p

since f : F∗p → F∗p , j 7→ 2j is a bijection. So

p−1
Y  
P = ζ −k 1 + ζ k + ζ 2k
k=1
p−1 Qp−1
Y
−k (1 − ζ 3k )
= ζ Qk=1
p−1 k
k=1 k=1 (1 − ζ )

= (ζ p )−(p−1)/2 = 1. Moreover, p−1

Qp−1 −k
Q 3k
Qp−1 k
k=1 ζ k=1 (1 − ζ ) = k=1 (1 − ζ ), since k 7→ 3k is
∗
a bijection in Fp , so P = 1, and consequently

p−1
Y  
2 2πj
1= 3 − 4 sin
p
j=1
(p−1)/2    p−1   
Y
2 2πj Y
2 2πj
= 3 − 4 sin 3 − 4 sin
p p
j=1 j=(p+1)/2
(p−1)/2   (p−1)/2

Y   
2πj 2 2π(k − j)
Y
2
= 3 − 4 sin 3 − 4 sin (k = p − j)
p p
j=1 k=1
 2
 
(p−1)/2  
Y 2πj
= 3 − 4 sin2 
p
j=1

Q(p−1)/2   
So j=1 3 − 4 sin2 2πj
p = ±1.
Let ν the number of negative factors in this product.

32
 If 1 ≤ j ≤ (p − 1)/2, then 0 < 4πj/p < 2π.

4πj 4πj 2π
1 + 2 cos < 0 ⇐⇒ cos < cos
p p 3
2π 4πj 4π
⇐⇒ < <
3 p 3
p p
⇐⇒ <j<
6 3
p
⇐⇒ < 3j < p
2
Let µ the number of integers j, 1 ≤ j ≤ (p − 1)/2 such that their least remainder is
negative. Since 3 ≤ 3j ≤ 3(p − 1)/2 and 3j 6= p/2, these j are such that p2 < 3j < p, so
µ = ν. Therefore
(p−1)/2     
Y
2 2πj ν µ 3
3 − 4 sin = (−1) = (−1) = .
p p
j=1

Ex. 5.35 Use the preceding exercise to show that 3 is a square modulo p iff p is
congruent to 1 or −1 modulo 12.

Proof. We know from Ex. 5.34 that ν = Card {j ∈ [1, (p − 1)/2] | p/2 ≤ 3j < p} = µ.
Therefore ν is the number of j such that p/6 ≤ j < p/3, so ν = bp/3c − bp/6c.
If p = 12k + 1, ν = bp/3c − bp/6c = 4k − 2k = 2k : (3/p) = (−1)ν = 1
If p = 12k + 5, ν = bp/3c − bp/6c = 4k + 1 − 2k = 2k + 1 : (3/p) = (−1)ν = −1
If p = 12k − 5, ν = bp/3c − bp/6c = 4k − 2 − (2k − 1) = 2k − 1 : (3/p) = (−1)ν = −1
If p = 12k − 1, ν = bp/3c − bp/6c = 4k − 1 − (2k − 1) = 2k : (3/p) = (−1)ν = −1
3 is a square modulo p (p 6= 2, p 6= 3) iff p is congruent to 1 or −1 modulo 12.

Ex. 5.36 Show that part (c) of Proposition 5.2.2 is true if a is negative and b is positive
(both still odd).

As said by Adam Michalik, the Jacobi symbol ab only defined for positive b, so the

question, which concerns ab ab , a < 0 makes no sense.

To give sense to this question, we must substitute the Kronecker symbol to the Jacobi
symbol. The Kronecker symbol (not defined in Ireland-Rosen) is the usual extension
of Jacobi symbol (see for instance [Henri Cohen] A course in computational algebraic
number theory, [Henri Cohen] Number theory (vol. 1), or [Harvey Cohn] Advanced
number theory).
We define Kronecker (or Kronecker-Jacobi) symbol ab for any a and b in Z in the

following way.

(1) If b = 0, then a0 = 1 if a = ±1, and a0 = 0 otherwise.

Q
(2) For b 6= 0, write b = p, where the p are not necessarily distinct primes (including
2), or p = −1 to take care of the sign. Then we set
  Y 
a a
= ,
b p

33
 a


where p is the Legendre symbol defined above for p > 2, and where we define
  (
a 0 if a is even
= 2
2 (−1)(a −1)/8 if a is odd,

and also   (
a 1 if a ≥ 0
=
−1 −1 if a < 0

Proof. Suppose a < 0, b > 0, both odd. Let a = −A, A > 0, A = p1 p2 · · · pk , where the
pi are not necessarily distinct primes. Then
     
a b −A b
=
b a b −A
      
−A −1 A (b−1)/2 A
= = (−1)
b b b b
        
b b b b b
= ··· =
−A −1 p1 pk A

so, from Prop. 5.2.2, as A, b are odd and positive,

     
a b b−1 A b
= (−1) 2
b a b A
b−1 A−1 b−1
= (−1) 2 (−1) 2 2

b−1
[1+ −a−1 ]
= (−1) 2 2

b−1 1−a
= (−1) 2 2

b−1 a−1
= (−1) 2 2

So the law of quadratic reciprocity remains valid for the Kronecker symbol when a is
negative (b > 0, a, b both odd).

Ex. 5.37 Show that if a is negative, then p ≡ q (mod 4a), p - a implies (a/p) = (a/q).

Proof. Write a = −A, A > 0. As p ≡ q (mod 4a), we know from Prop. 5.3.3. (b) that
(A/p) = (A/q).
Moreover,
     
a −A (p−1)/2 A
= = (−1)
p p p
     
a −A A
= = (−1(q−1)/2
q q q

As p ≡ q (mod 4a), p = q + 4ak, k ∈ Z, so

(−1)(p−1)/2 = (−1)(q+4ak−1)/2 = (−1)(q−1)/2 ,

so (a/p) = (a/q).

34
Ex. 5.38 Let p be an odd prime. Derive the quadratic character of 2 modulo p by
verifying the following steps, involving the Jacobi symbol:
         
2 8−p p 8 2
= = = = .
p p p−8 p−8 p−8
Generalize the argument to show that
   
a a
= , a > 0, p - a.
p p − 4a

(As in Ex. 5.36, since 8 − p or p − 8 is negative, we interpret (a/b) as the Kronecker

symbol : see definition in Ex. 5.36.)

Proof. As (22 /p) = 1 and 8 − p ≡ 8 (mod p),

   2      
2 2 2 8 8−p
= = = .
p p p p p

As p and 8 − p are odd numbers and p > 0, from the extension of the law of quadratic
reciprocity to a < 0 proved in Ex. 5.36, we obtain
   
8−p 7−p p−1 p
= (−1) 2 2 .
p 8−p
Moreover

(7 − p)(p − 1) ≡ (−1 − p)(p − 1) = 1 − p2 (mod 8)

As p = 2k + 1 is odd, p2 = 4k 2 + 4k + 1 = 8 k(k+1)
2 + 1 ≡ 1 (mod 8), so (7 − p)(p − 1) ≡ 0
(mod 8) and 7−p 2
p−1
2 is even, so
   
8−p p
= .
p 8−p
p
p
p
p
p

As p > 0, −1 = 1, thus 8−p = −1 p−8 = p−8 (with the same argument, this is
also true for the 3 odd primes such that 8 − p > 0), so
   
8−p p
= .
p p−8

p
8 8 2
, and since 8 = 22 × 2, p−8




As p ≡ 8 (mod p − 8), p−8 = p−8 = p−8 . We have
proved for all odd primes p that
         
2 8−p p 8 2
= = = = .
p p p−8 p−8 p−8

The preceding arguments remain valid if we replace the odd prime p by any odd
positive integer. So with an immediate induction, we see that for all k ∈ N,
   
2 2
= .
p p − 8k

35
So the quadratic character

of
2 modulo p depends only of the class of p modulo 8.
If p ≡ 1 (mod 8), p2 = 21 = 1.
If p ≡ −1 (mod 8), p2 = −1 2



= 1.
2 2



If p ≡ ±3 (mod 8), p = ±3 = −1.

Generalization : let a > 0 and p an odd positive integer such that p ∧ a = 1 (not
necessarily prime).
       
a 4ap 4a − p 4a−p−1 p−1 p
= = = (−1) 2 2 .
p p p 4a − p
(4a − p − 1)(p − 1) = 4a(p − 1) + 1 − p2 ≡ 0 (mod 8), so
   
a p
= .
p 4a − p
p

As −1 = 1,
   
p p
= .
4a − p p − 4a
Since p ≡ 4a (mod p) − 4a, and 4 is a square,
     
p 4a a
≡ = .
p − 4a p − 4a p − 4a

We have proved
         
a 4a − p p 4a a
= = = = .
p p p − 4a p − 4a p − 4a

By induction, for all k ≥ 0, ap = p−4a

a
, so ap depends only of the class of p modulo

4a.

Chapter 6
√ √
Ex. 6.1 Show that 2 + 3 is an algebraic integer.
√ √ √
Proof. Let x = 2√+ 3. Then x2 = 5 + 2 6.
(x2 − 5)2 = (2 6)2 = 24, so x4 − 10x2 + 1 = 0 : x is an algebraic integer.

Ex. 6.2 Let α be an algebraic number. Show that there’s an integer n such that nα is
an algebraic integer.
(0 is a valid answer to this sentence ! More seriously, we search a positive integer n.)

Proof. Let α an algebraic number. By definition, there exist a0 , a1 , · · · , an ∈ Z, an 6= 0,

such that
an αn + an−1 αn−1 + · · · + ak αk + · · · + a0 = 0.
(Up to multiply this equation by −1, we can suppose that an > 0).
Multiplying by an−1
n , we obtain

ann αn + an−1
n an−1 α
n−1
+ · · · + ann−1 ak αk + · · · + ann−1 a0 = 0.

36
So
(an α)n + an−1 (an α)n−1 + · · · + ann−k−1 ak (an α)k + · · · + an−1
n a0 = 0.
n−1
Soit p(x) = xn + an−k−1 ak xk . Then p(x) ∈ Z[x], p(x) is monic, and p(an α) = 0.
P
n
k=0
So an α is an algebraic integer, with m = an ∈ N∗ .
Conclusion : if α is an algebraic number, there exists an integer m > 0 such that mα
is an algebraic integer.

Ex. 6.3 If α and β are algebraic integers, prove that any solution to f (x) = x2 + αx +
β = 0 is an algebraic integer. Generalize this result.

Proof. Let γ a root of x2 + αx + β, where α, β verify :

αn + r1 αn−1 + · · · + rn = 0, ri ∈ Z,

β m + s1 β m−1 + · · · + sm = 0, sj ∈ Z.
Let V the set of linear combinations with integer coefficients of

αi β j γ k , 0 ≤ i < n, 0 ≤ j < m, 0 ≤ k < 2.

Then V if a finitely generated Z-module.

Moreover, for all δ ∈ V, γδ ∈ V . Indeed, every δ ∈ V is a linear combination with
coefficients in Z of αi β j , αi β j γ, and

γ(αi β j ) = αi β j γ ∈ V
γ(αi β j γ) = αi β j γ 2 = αi β j (−αγ − β) = −αi+1 β j γ − αi β j+1 ∈ V.
n−1
(if i + 1 = n, we replace αi+1 = αn by − rk αn−k , and a similar replacement if if
P
k=1
j + 1 = m.)
As for each x ∈ V , where V if a finitely generated Z-module, xγ ∈ V , so γ is an
algebraic integer (Proposition 6.1.4).
More generally, if γ n + α1 γ n−1 + · · · + αn = 0, where the αi are algebraic integers,
then x is an algebraic integer.

Ex. 6.4 A polynomial f (x) ∈ Z[x] is said to be primitive if the greatest common divisor
of its coefficients is 1. Prove that product of primitive polynomials is also primitive.

Solution 1

Proof. Let p(x) = ni=0 ai xi , q(x) = m j

P P
j=0 bj x two primitive polynomials, and p a prime
number. There exist a coefficient of p(x) (and of q(x)) not divisible by p. Let

i0 = min{i ∈ [0, n] | ai 6≡ 0 [p]}

j0 = min{j ∈ [0, m] | bj 6≡ 0 [p]}

Pn+m
ck xk . Then ck =
P
Let p(x)q(x) = k=0 ai bj , k = 0, . . . n + m. Then
i+j=k
X
ci0 +j0 = ai bj .
i+j=i0 +j0

37
 • If i < i0 , then ai ≡ 0 (mod p).
• If i > i0 , then j < j0 and bj ≡ 0 (mod p).
In the two cases ai bj ≡ 0 (mod p), so ci0 +j0 ≡ ai0 bj0 (mod p), so cj0 6≡ 0 (mod p) :
as it’s true for all primes p, the polynomial p(x)q(x) is primitive.

Solution 2

Proof. Let

Z[x] → Fp [x]
ϕ:
p(x) = a0 + · · · + an xn 7→ p(x) = a0 + · · · + an xn ,

where ai is the class of ai in Fp . ϕ is a ring homomorphism.

As Fp [x] is an integrity domain, if p(x), q(x) are both primitive,

p(x) 6= 0, q(x) 6= 0 ⇒ p(x)q(x) = p(x) q(x) 6= 0.

As p(x)q(x) 6= 0 in all fields Fp , p(x)q(x) is a primitive polynomial.

Ex. 6.5 Let α be an algebraic integer and f (x) ∈ Q[x] be the monic polynomial of least
degree such that f (α) = 0. Use Exercise 6.4 to show that f (x) ∈ Z[x].

Proof. As α is an algebraic integer, there exists a monic polynomial h(x) ∈ Z[x] such
that h(α) = 0. As f (x) ∈ Q[x] is the minimal polynomial of α, and h(α) = 0, f (x)
divides h(x) in Q[x].
(quick reminder : h(x) = q(x)f (x) + r(x), q(x), r(x) ∈ Q[x], deg(r(x)) < deg(f (x))
or r(x) = 0. As r(α) = 0 and f (x) ∈ Q[x] is the monic polynomial of least degree such
that f (α) = 0, r = 0 so f (x) | h(x)).
So there exists g(x) ∈ Q[x] such that h(x) = f (x)g(x). As h(x), f (x) are both monic,
g(x) is also monic.
Let d ∈ Z, d 6= 0 such that df (x) = m i
P
i=0 ai x ∈ Z[x], and c = a1 ∧ a2 ∧ · · · ∧ am ,
c
ai = cbi , with b1 ∧ b2 ∧ · · · ∧ bm = 1, so f (x) = d f1 (x), with f1 is primitive. Similarly
g(x) = st g1 (x), s, t ∈ Z, g1 (x) primitive.
Pr So h(x) = cs u
dt f1 (x)f2 (x) = v f1 (x)f2 (x), where u∧v = 1. The polynomial f1 (x)f2 (x) =
k
k=0 ck x is primitive (Ex. 6.4). As vh(x)(x) = uf1 (x)f2 (x), v | uck , and u ∧ v = 1,
thus v | ck , k = 0, 1, . . . , r. As c1 ∧ · · · ck = 1, v | 1, so v = ±1. h(x) = uf1 (x)f2 (x)
is monic, thus u = 1, and f1 , f2 are monic. From f (x) = dc f1 (x) we deduce dc = 1 and
f (x) = f1 (x) ∈ Z[x].
Conclusion : if f (x) is the minimal polynomial of an algebraic integer α, f ∈ Z[x].

Ex. 6.6 Let x2 + mx + n ∈ Z[x] be irreducible, and α be a root. Show that Q[α] =
{r + sα : r, s ∈ Q} is a ring (in√fact, it is a field). Let m2 − 4n = D02 D, where D is
square-free. Show that Q[α] = Q[ D].

Proof. By definition, for all z ∈ C, z ∈ Q[α] ⇐⇒ ∃P ∈ Q[x], z = P (α).

The Euclidean division gives P = Q1 (x2 + mx + n) + R, Q1 , R ∈ Q[x], deg(R) < 2,
so R = rx + s, r, s ∈ Q. So z = Q1 (α)(α2 + mα + n) + rα + s = rα + s :

Q[α] = {z ∈ C | ∃r ∈ Q, ∃s ∈ Q, z = r + sα}.

38
 • Q[α] ⊂ C, where (C, +, ×) is a field. 1 ∈ Q[α] (1 = P0 (α), where P0 is the constant
polynomial 1).
• Let β, γ ∈ Q[α] : β = P (α), γ = Q(α), where P, Q are in Q[x]. Then α − β =
P (α) − Q(α) = R(α), where R = P − Q ∈ Q[x], and αβ = P (α)Q(α) = S(α), where
S = P Q ∈ Q[x]. So α − β ∈ Q[α], αβ ∈ Q[α]. So Q[α] is a subring of (C, +, ×).
• Let β = P (α) ∈ Q[α], P ∈ Q[x] and β 6= 0. As β 6= 0, Q = x2 + mx + n - P .
Let D ∈ Q[x] such that D | P, D | Q. As Q is irreducible by hypothesis, D = λ or
D = λQ, λ ∈ C∗ (D is an associate of 1 or Q). If D = λQ, then Q | D, and D | P , so
Q | P . Since Q(α) = 0, this implies β = P (α) = 0, in contradiction with the definition
of β. So D = λ | 1. Therefore P ∧ Q = 1.
From Bézout’s theorem, there exist polynomials U, V ∈ Q[x] such that U P +V Q = 1.
As Q(α) = 0, U (α)P (α) = 1 and γ = U (α) ∈ Q[α] is such that γβ = 1. Therefore Q[α]
is a subfield of (C, +, ×) (and Q(α) = Q[α]).
As x2 + mx + n is irreducible, ∆ = m2 − 4n 6= 0 (if not, x2 + mx + n = (x +
m/2)2 − (m2 − 4n)/4 = (x + m/2)2 is not irreducible). So ∆ ∈ Z \ {0} can be written
∆ = m2 − 4n =√D02 D, where D is square-free (positive or negative), D 6= 0, D0 6= 0.
√ √ √
α = −m + ε ∆
, ε = ±1, so α = − m
+ εD 0 2
D
, thus α ∈ Q[ D] and Q[α] ⊂ Q[ D].
2 √2 2 √
As D0 6= 0, D = ε 2α+m D0 ∈ Q[α], so Q[ D] ⊂ Q[α)] :
√
Q[α] = Q[ D].

Ex.√ 6.7 (continuation) If√D ≡ 2, 3 (mod 4), show that all the algebraic integers in
Q[ D] have the form a√+ b D, where a, b ∈ Z. If D √ ≡ 1 (mod 4), show that all the
algebraic integers in Q[ D] have the form a + b((−1 + D)/2), where a, b ∈ Z.

Proof. (We write Z the ring of algebraic integers in C, and OK (or ZK ) the ring of
algebraic integers√in the field K.) √
If D = 1, Q[ D] = Q. If D 6= 1, as D is square-free, D is not a square, so D is
irrational. √ √ √
Let γ = r + s D ∈ Q[ D] (r, s ∈ Q) an algebraic integer of Q[ D] (D ∈ Z, D
square-free). (γ − r)2 = s2 D, so γ 2 − 2rγ + r2 − Ds2 = 0. γ is a root of

p(x) = x2 − 2rx + r2 − Ds2 .

If s = 0, then the minimal polynomial of γ is x − r. As r = γ is an algebraic integer

and r ∈ Q, then r ∈ Z. In this case r ∈ Z and s = 0.
If s 6= 0, γ 6∈ Q, so no polynom of degree d ≤ 1 has the root γ. Thus the minimal
polynomial of γ is p(x). From Exercise 6.5, p(x) ∈ Z[x], so (in the two cases s = 0, s 6= 0)

2r ∈ Z, r2 − Ds2 ∈ Z.

Conversely, if 2r ∈ Z, r2 − Ds2 ∈ Z, then p(x) ∈ Z[x] and p(γ) = 0, thus γ is an algebraic

integer.
If r, s ∈ Q, D 6= 1 square-free,
√
r + s D ∈ Z ⇐⇒ 2r ∈ Z, r2 − Ds2 ∈ Z.
√
Let γ = r + s D ∈ Z. We can write
a b
r= ,s = , a, b, d ∈ Z, d ≥ 1, d ∧ a ∧ b = 1.
d d

39
Then
2a a2 − Db2
n= ∈ Z, m= ∈ Z.
d d2
As D is square-free, D ≡
6 0 (mod 4).

• Case 1 : D ≡ 2, 3 (mod 4).

4Db2
n2 − 4m = d2
, so d | 2a, d2 | 4Db2 .
If 2 | d, 4 | a2 − Db2 , a2 ≡ Db2 (mod 4). As d ∧ a ∧ b = 1, and 2 | d, a or b is
odd, and a2 ≡ Db2 (mod 4), D 6≡ 0 (mod 4), implies that a and b are both odd.
Then a2 ≡ b2 ≡ 1 (mod 4), so D ≡ 1 (mod 4) : this is in contradiction with the
hypothesis D ≡ 2, 3 (mod 4). So d is an odd number.
Consequently, d | a, d2 | Dq 2 . If p ∈ N is a prime factor of d, p | d, p | a ,and
d ∧ a ∧ b = 1, so p - b, and since p2 | Db2 , p2 | D, in contradiction with D square-
free. So d√≥ 1 has no prime factor : d = 1 and r = a, s = b ∈ Z. Conversely, any
γ = a + b D, a, b ∈ Z is an algebraic integer, so
√ √
OQ[√D] = Z ∩ Q[ D] = {a + b D | a, b ∈ Z}.

• Case 2 : D ≡ 1 (mod 4).

Then r = n2 , n ∈ Z. Write s = uv , u ∧ v = 1, v ≥ 1.
2 2 2
m = r2 − Ds2 = n4 − D uv2 ∈ Z, 4D uv2 = n2 − 4m ∈ Z, so v 2 | 4Du2 . Since
u ∧ v = 1, u2 ∧ v 2 = 1, so v 2 | 4D. As D is square-free, v has no odd prime factor,
so v = 2k . Since D is odd, k ≤ 1 and v = 1 or v = 2. So r, s are both half-integers:
r = n/2, s = n0 /2, n, n0 ∈ Z.
4m = n2 − Dn02 , thus n2 ≡ n02 (mod 4), so n, n0 have the same parity. √ Let
0
a = n+n ∈ Z, b = n 0 ∈ Z. Then n = 2a − b, n0 = b and γ = n + n0 D =
2 2 2
b b
√  √ 
−1+ D
a− 2 + 2 D =a+b 2 .
√ 2
 √ 
Conversely, −1+2 D is a root of x2 + x + 1−D
4 ∈ Z[x], so every a + b −1+ D
2 is
an algebraic integer.
√ !
√ −1 + D
OQ[√D] = Z ∩ Q[ D] = {a + b | a, b ∈ Z}.
2

Ex. 6.8 Let ω = e2πi/3 , ω satisfies x3 − 1 = 0. Show that (2ω + 1)2 = −3, and use this
to determine (−3/p) by the method of section 2.

Proof. As ω 2 + ω + 1 = 0,(2ω + 1)2 = 4ω 2 + 4ω + 1 = −4 + 1 = −3. Let α = 2ω + 1, so

α2 = −3
 
−3
≡ (−3)(p−1)/2 (mod p)
p
≡ αp−1(mod p)
 
p −3
α = α.
p

40
From Prop. 6.1.6,

αp = (2ω + 1)p
≡ 2p ω p + 1 (mod p)
p
≡ 2ω + 1 (mod p)
−3


• If p ≡ 0 (mod 3), p = 0.

• If p ≡ 1 (mod 3), ω p = ω, so αp ≡ α (mod p).

−3 −3 −3


2 2


p
α ≡ α (mod p), thus p
α ≡ α (mod p), p 3 ≡ 3 (mod p). As p ∧ 3 = 1,
−3 −3 −3


p ≡ 1 (mod p). Since p = ±1, p = 1.

• If p ≡ −1 (mod 3),

αp ≡ 2ω p + 1 (mod p)
2
≡ 2ω + 1 = 2(−1 − ω) + 1 = −2ω − 1 = −α (mod p).

−3
α ≡ −α (mod p), thus −3 α ≡ −α2 (mod p), −3


2

p p p 3 ≡ −3 (mod p). As
−3 −3 −3




p ∧ 3 = 1, p ≡ −1 (mod p). Since p = ±1, p = −1.

Conclusion :
 
−3
p ≡ 0[3] ⇐⇒ =0
p
 
−3
p ≡ 1[3] ⇐⇒ =1
p
 
−3
p ≡ −1[3] ⇐⇒ = −1
p

p

In other words, −3


p = 3 .
Note : α = 2ω + 1 = ω − ω 2 = g, the quadratic Gauss sum for p = 3.

Ex. 6.9 Verify Proposition 6.3.2 explicitly for p = 3, 5, i.e., write out the Gauss sum
longhand and square.

• p=3. Let ω = e2iπ/3 . Let g = 2t=0 (t/3)ω t the quadratic Gauss sum. Then
P
Proof.
g = ω − ω2.
As 1 + ω + ω 2 = 0, g 2 = (ω − ω 2 )2 = ω 2 − 2ω 3 + ω 4 = ω 2 − 2 + ω = −3 :

g 2 = −3.

• p=5. Let ζ = e2iπ/5 .

g = 4t=0 (t/3)ζ t = ζ − ζ 2 − ζ 3 + ζ 4 .
P

Then g = α − β, where α = ζ + ζ 4 , β = ζ 2 + ζ 3 .
α + β = ζ + ζ 4 + ζ 2 + ζ 3 = −1.
αβ = ζ 3 + ζ 4 + ζ 6 + ζ 7 = ζ 3 + ζ 4 + ζ + ζ 2 = −1

41
 So α, β are the two roots of x2 + x − 1.

g 2 = (α − β)2
= α2 + β 2 − 2αβ
= (α + β)2 − 4αβ
= (−1)2 − 4(−1)
= 5.

Note : here we know explicitely g :

√
if p = 3, g = ω − ω 2 = i 3.
√ √ √
If p = 5, g = α − β = (−1 + 5)/2 − (−1 − 5)/2 = 5.

Pp−1
Ex. 6.10 What is a=1 ga ?

Proof. From Prop. 6.3.1 and Lemma 2,

p−1 p−1  
X X a
ga = g1 = 0.
p
a=1 a=1

+ (t/p))ζ t in two ways, prove that g = t2 .

P P
Ex. 6.11 By evaluating t (1 tζ

Proof. For a ∈ Fp , Write N [x2 = a] the number of solutions of the equation x2 = a in

Fp . We know from Ex. 5.2 that N [x2 = a] = 1 + (a/p). So
p−1
2 2
X X
ζt = ζt
t=0 t∈Fp
X
= N [x2 = t]ζ t
t∈Fp
X  
t
= 1+ ζt
p
t∈Fp
X X t
t
= ζ + ζt
p
t∈Fp t∈Fp
p−1  
X t t
= ζ
p
t=0
=g

42
Ex. 6.12 Write ψa (t) = ζ at . Show that

(a) ψa (t) = ψa (−t) = ψ−a (t)

P
(b) (1/p) a ψa (t − s) = δ(t, s)

Proof. (a) Let a ∈ Z. As ζ = ζ −1 ,

ψa (t) = ζ at = ζ −at
= ζ a(−t) = ζ (−a)t
= ψa (−t) = ψ−a (t)

ψa (t) = ψa (−t) = ψ−a (t)

(b) From Corollary of Lemma 1 :

p−1 p−1
1X 1 X a(t−s)
ψa (t − s) = ζ = δ(t, s)
p p
a=0 a=0

1X
ψa (t − s) = δ(t, s).
p a

Ex. 6.13 Let f be a function from Z to the complex numbers. Suppose that p is a
prime and that f (n + p) = f (n) for all n ∈ Z. Let fˆ(a) = p−1 t f (t)ψ−a (t). Prove that
P

f (t) = a fˆ(a)ψa (t). This result is directly analogous to a result in the theory of Fourier
P
series.

Proof. Let fˆ(a) = p−1 t f (t)ψ−a (t). Then

P

p−1
X p−1
X p−1
X
−1
fˆ(a)ψa (t) = p f (s)ψ−a (s)ψa (t)
a=0 a=0 s=0
p−1
X p−1
X
−1
=p f (s) ψ−a (s)ψa (t)
s=0 a=0
p−1
X p−1
X
= p−1 f (s) ψa (t − s)
s=0 a=0
p−1
X
= f (s)δ(s, t)
s=0
= f (t)

Ex. 6.14 In Ex. 13 take f to be the Legendre symbol and show that fˆ(a) = p−1 g−a .
Pp−1 t
−at
Proof. Here f (a) = ap . Then fˆ(a) = p−1 t=0 = p−1 g−a .


p ζ

43
Ex. 6.15 Show that  
m  
X t  √
 < p log p.

p 


t=n
The inequality holds for the sum over any range.
Lemma. If 0 ≤ x ≤ π2 , sin x ≥ π2 x.
Proof. As − sin is a convex function on [0, π/2], the graph of sin is above any chord, and
the chord between the points (0, 0) and (π/2, 1) has equation y = (2/π)x, we conclude
that sin x ≥ π2 x for 0 ≤ x ≤ π/2.
 
m √  Pm
t t 
P


Proof. Let S = p g with n ≤ m. Then |S| = p p . As (t/p)g = gt ,
t=n t=n
n
X
S= gt
t=m
n Xp−1  
X s ts
= ζ
t=m s=0
p
p−1   n
X s ms X (t−m)s
= ζ ζ
p t=m
s=0
p−1   n−m
X s ms X us
= ζ ζ (u = t − m)
p
s=0 u=0
p−1  
X s ms ζ (n−m+1)s − 1
= ζ
p ζs − 1
s=1

Pn−m s
ζ us = n − m + 1 and


(since for s = 0, the sum u=0 p = 0). So
p−1   (n+1)s
X s ζ − ζ ms
S=
p ζs − 1
s=1
p−1  n+m+1 n−m+1 −n+m−1
X s ζ 2 s ζ 2 s−ζ 2 s
= s s −s
p ζ2 ζ2 −ζ 2
s=1
 
p−1 π

X s n+m sin (n − m + 1)s p
= ζ 2 s  
p sin s π
s=1 p

As sin(x) ≥ π2 x for x ∈ [0, π2 ], for all s, 1 ≤ s < p2 , 0 ≤ sπ π

p ≤ 2 , so
  
 sin (n − m + 1)s π 
 ≤ 1  = p
p 

   (s = 1, 2, . . . , (p − 1)/2).

sin s πp
 2
s π 2s
π p
 

Since ps ζ ts depends only of the class of s, we can replace in the preceding calculation

the values s = 1, 2, . . . , p − 1 by s = −(p − 1)/2, . . . , −1, 1, . . . , (p − 1)/2, so

   
(p−1)/2   π −1 π
X s n+m s sin (n − m + 1)s p X  
s n+m s sin (n − m + 1)s p
S= ζ 2   + ζ 2   .
p sin s π p sin s π
s=1 p s=−(p−1)/2 p

44
As sin is an odd function,
   
(p−1)/2  π (p−1)/2  π
X 
s n+m s sin (n − m + 1)s p X 
−s − n+m s sin (n − m + 1)s p
S= ζ 2   + ζ 2   .
p sin s πp p sin s πp
s=1 s=1

Thus
(p−1)/2 (p−1)/2
X p X 1
|S| ≤ 2 =p .
2s s
s=1 s=1
m √
t
P

As S = p g and |g| = p,
t=n
(p−1)/2
 
m  
X t  √ X 1
≤ p .

p  s


t=n s=1
It remains to do a sufficient estimation of the harmonic sum. We prove by induction
that for all n ≥ 1,
1 1
1 + + · · · + ≤ log(2n + 1).
2 n
As 1 ≤ log(3), this proposition is true for n = 1. Suppose that is it true for n − 1 :
1 1
1 + + ··· + ≤ log(2n − 1).
2 n−1
Then
1 1 1
1 + + · · · + ≤ + log(2n − 1).
2 n n
If we prove that n1 + log(2n − 1) ≤ log(2n + 1), the induction is done.
Let u(x) = log(2x + 1) − log(2x − 1) − x1 , x > 12 .
2 2 1
u0 (x) = − + 2
2x + 1 2x − 1 x
−4 1
= 2 +
4x − 1 x2
−1
= <0
(4x − 1)x2
2
 
2x+1
As u(x) = log 2x−1 − x1 , lim u(x) = 0. Moreover u is a decreasing function, so for
x→+∞
all x > 1/2, u(x) > 0, and for all n ∈ N, n ≥ 1,
1
+ log(2n − 1) ≤ log(2n + 1).
n
We have proved by induction that for all n ≥ 1,
1 1
1 + + · · · + ≤ log(2n + 1).
2 n
If n = (p − 1)/2, where p is an odd prime (p ≥ 3),
(p−1)/2
X 1
≤ log p.
s
s=1
Conclusion :  
m  
X t  √
 < p log p.

p

 
t=n

45
Ex. 6.16 Let α be an algebraic number with minimal polynomial f (x). Show that f (x)
does not have repeated roots in C.

Proof. Let γ a repeated root of f (x). Then f (γ) = f 0 (γ) = 0, so x − γ is a common

factor of f and f 0 . Thus f ∧ f 0 6= 1 (deg(f ∧ f 0 ) ≥ 1). Since f ∧ f 0 | f and f is irreducible
(with f , f ∧ f 0 monic), we conclude f ∧ f 0 = f , so f | f 0 . In C, this is impossible since
deg(f ) ≥ 1, so f 0 6= 0, and deg(f 0 ) < deg(f ). f (x) does not have repeated roots in C.
√
Ex. 6.17 Show that the minimal polynomial for 3 2 is x3 − 2.
√
Proof. Let f (x) = x3 − 2. Then f ( 3 2) = 0. If f (x) was not irreducible, then f (x) =
u(x)v(x), with 1 ≤ deg(u) ≤ deg(v) ≤ 2, deg(u) + deg(v) = deg(f ) = 3, so deg(u) =
1, deg(v) = 2.
Then f (x) = (ax + b)(cx2 + dx + e), a, b, c, d, e ∈ Q. Let w = −b/a. Then f (w) =
3
w − 2 = 0 and w ∈ Q, so there exist p, q ∈ Z, such that w = p/q, p ∧ q = 1.
Thus p3 = 2q 3 , so p3 is even, thefore p is even : p = 2p0 , p0 ∈ Z.
8p03 = 2q 3 , 4p03 = q 3 , so q 3 is even, which implies that q is even. Then 2 | p ∧ q = 1 :
this is a contradiction.
√ √
So f ( 3 2) = 0, and f is monic, irreducible : f is the minimal polynomial of 3 2 on
Q.

Ex. 6.18 Show that there exist algebraic numbers of arbitrarily high degree.

Proof. As 1 + x + · · · + xp−1 is irreducible on Q[x] (Prop. 6.4.1), the numbers ζp = e2iπ/p ,

with p prime number, are algebraic numbers of arbitrary large degree.

Ex. 6.19 Find the conjugates of cos(2π/5).

Proof. Let γ = cos(2π/5), ζ = e2iπ/5 and α = ζ + ζ 4 , β = ζ 2 + ζ 3 .

−1 4
Then γ = ζ+ζ2 = ζ+ζ 2 = 2.
α
4 2
α + β = ζ + ζ + ζ + ζ = −1. 3

αβ = ζ 3 + ζ 4 + ζ 6 + ζ 7 = ζ 3 + ζ 4 + ζ + ζ 2 = −1
So α, β are the two roots of x2 + x − 1 :
α2 + α − 1 = 0, so 4(α/2)2 + 2(α/2) − 1 = 0 : γ = α/2 is a root of

f (x) = 4x2 + 2x − 1.

As ∆ = 4 × 5, the two roots of f are irrational. deg(f ) = 2 and f has no root in Q,

so f (x) is irreducible in Q[x]. Therefore the minimal polynomial of γ = cos(2π/5) is
f (x) = 4x2 + 2x − 1. The other root of f is β/2 = (ζ 2 + ζ 3 )/2 = cos(4π/5).
Conclusion : the conjugates of γ = cos(2π/5) are γ = cos(2π/5) and cos(4π/5).

Ex. 6.20 Let F be a subfield of C which is a finite-dimensional vector space over Q of

degree n. Show that every element of F is algebraic of degree at most n.

Proof. Let α ∈ F , with dimQ F = n. Any subset of n + 1 vectors in F is linearly

dependent, so {1, α, α2 , · · · , αn } is linearly dependent.
Thus there exists (a0 , . . . , an ) ∈ Qn+1 , (a0 , . . . , an ) 6= (0, 0, . . . , 0) such that a0 +a1 α+
· · · + an αn = 0.
Let f (x) = a0 + a1 x + · · · + an xn . Then f (x) ∈ Q[x], f (x) 6= 0 and f (α) =
0, deg(f (x)) ≤ n. So every element of F is algebraic of degree at most n.

46
Ex. 6.21 Let f (x) = ∞
P n
P∞ n
n=0 an x /n! and g(x) = n=0 bn x /n! be power series with
an and bn integers. If p is a prime such Pthat p|ani for i = 0, . . . , p − 1, show that each
coefficient ct of the product f (x)g(x) = ∞ n=0 cn x for t = 0, . . . , p − 1 may be written in
the form p(A/B), p 6 |B.

Proof. Let k ∈ N, 0 ≤ k ≤ p − 1.
X ai bj
ck =
i! j!
i+j=k
k
X ai bk−i
=
i! (k − i)!
i=0
k  
1 X k
= ai bk−i
k! i
i=0

As k! ∧ p = 1, and ki=0 i ai bk−i ≡ 0 (mod p) for k = 0, 1, . . . , p − 1,

k
P

ck = p(A/B), p ∧ B = 1.

Ex. 6.22 Show that the relation ε ≡ 1 (mod p) in Proposition 6.4.4 can also be
achieved by replacing x by 1 + t instead of ez .

Proof. (solution given by Mikomikon and A.Grounds (agrounds))

We know from the remark after Prop 6.4.3 that
(p−1)/2
Y
g(χ) = ε (ζ 2k−1 − ζ −(2k−1) ),
k=1

where ε = ±1. Let

p−1 (p−1)/2
X Y
j
f (x) = χ(j)x − ε (x2k−1 − xp−(2k−1) ).
j=1 k=1

Then f (0) = 0 and f (ζ) = 0, so (xp − 1) divides f (x). As f (x) ∈ Z[x] and xp − 1 ∈ Z[x]
is monic, f (x) = (xp − 1)h(x), h(x) ∈ Z[x]. If we replace x by 1 + t, we obtain

p−1 (p−1)/2  
X Y
j
f (1 + t) = χ(j)(1 + t) − ε (1 + t)2k−1 − (1 + t)p−(2k−1) .
j=1 k=1

We compute the coefficient of t(p−1)/2 in the polynomial f (1 + t) :

p−1 p−1 j  
X X X j i
χ(j)(1 + t)j = χ(j) t
i
j=1 j=1 i=1
p−1 X
p−1  
X j i
= χ(j) t
i
i=1 j=i

47
 p−1 p−1
j
So the coefficient of t(p−1)/2 in χ(j)(1 + t)j is
P P

χ(j) (p−1)/2 .
j=1 j=(p−1)/2

(p−1)/2 (p−1)/2
Y Y
2k−1 p−(2k−1)
(1 + (2k − 1)t) − (1 + (p − (2k − 1))t + t2 u(t)


((1 + t) − (1 + t) )=
k=1 k=1
(p−1)/2
Y
(4k − 2 − p)t + t2 v(t)


=
k=1
 
(p−1)/2
Y
= t(p−1)/2  (4k − 2 − p) + t(p+1)/2 w(t),
k=1

where u(t), v(t), w(t) are polynomials. So the coefficient of t(p−1)/2 in f (1 + t) is

p−1   (p−1)/2
X j Y
c(p−1)/2 = χ(j) −ε (4k − 2 − p).
(p − 1)/2
j=(p−1)/2 k=1

Furthermore,

f (1 + t) = ((1 + t)p − 1) h(1 + t)

" p   #
X p
= ti h(1 + t)
i
i=1
" p   #
X p ti
= i! h(1 + t)
i i!
i=1
" p #
X ti
= ai h(1 + t),
i!
i=0

p p!


where a0 = 0, ai = i! = (p−i)!
i , so p | ai , i = 0, . . . , p − 1 : the conditions of Ex.21 are
Pp−1 i
verified, so f (1 + t) = i=0 ci t is such that c(p−1)/2 = p(A/B), p - B. Equating these
two evaluations of c(p−1)/2 , we obtain

p−1   (p−1)/2
X j Y A
χ(j) −ε (4k − 2 − p) = p , p - B.
(p − 1)/2 B
j=(p−1)/2 k=1

Multiplying by B(p − 1)!/2, we obtain, as p - B,

p−1   (p−1)/2
(p − 1)! X j (p − 1)! Y
χ(j) ≡ε (4k − 2)
2 (p − 1)/2 2
j=(p−1)/2 k=1
(p−1)/2
Y
≡ ε(2 · 4 · 6 · · · (p − 1)) (2k − 1) ≡ ε(p − 1)!
k=1
≡ −ε (mod p)

48
To prove that ε = +1, it remains to prove
  p−1  
(p − 1) X j
! χ(j) ≡ −1 (mod p)
2 (p − 1)/2
j=(p−1)/2

j

The factor of ((p − 1)/2)! cancels the denominator of (p−1)/2 , which leaves

p−1
X p−1
χ(j) · j(j − 1) · · · (j − + 1)
2
j=(p−1)/2
p−1
X p−1
= χ(j) · j(j − 1) · · · (j − + 1).
2
j=1

The equality is justified because all terms for j < p−1 2 are zero. Collecting powers of j,
this is
p−1 (p−1)/2 (p−1)/2 p−1
X X X X p−1
χ(j) ak j k ≡ ak j k+ 2 (mod p)
j=1 k=0 k=0 j=1

for some integers ak . It’s important to note that a(p−1)/2 = 1.

Now, we compute p−1 n
P
j=1 j mod p. Let S denote this sum and let g be a generator
of Z/p× . Then
p−1
X p−1
X
n n
g S= (gj) ≡ j n = S (mod p).
j=1 j=1

Congruence holds because gj also runs over a complete system of nonzero residues mod
p. If g n 6≡ 1, then S ≡ 0. If g n ≡ 1, then j n ≡ 1 for all j, hence S ≡ p − 1 ≡ −1.
Returning to the previous sum, the only nonzero term modulo p is k = p−1 2 , so

(p−1)/2 p−1
X X p−1
≡ ak j k+ 2 ≡ a(p−1)/2 · (−1) = −1 (mod p)
k=0 j=1

as desired.

Ex. 6.23 If f (x) = xn + a1 xn−1 + . . . + an , ai ∈ Z, and p is prime such that p | ai for

i = 1, . . . , n, and p2 - an , show that f (x) is irreducible over Q (Eisenstein’s irreducibility
criterion).

Lemma. If f ∈ Z[x], deg(f ) ≥ 1, is not irreducible in Q[x], then there exist g, h ∈

Z[x], deg(g) ≥ 1, deg(h) ≥ 1 such that f = gh.
n
ak xk , ak
P
Proof. (lemma) Suppose that f (x) = ∈ Z, is not irreducible in Q[x].
k=0
Then f (x) = f1 (x)f2 (x), with f1 , f2 ∈ Q[X],
and deg(f1 ) ≥ 1, deg(f2 ) ≥ 1. As in
Ex. 6.5, we can write f1 (x) = λp(x), f2 (x) = µq(x) where λ, µ ∈ Q, and p, q ∈ Z[X] are
primitive. Let ν = λµ ∈ Q : write ν = u/v, u ∧ v = 1, v ≥ 1. Then r(x) = p(x)q(x) =
n
ck xk is primitive (Ex. 6.4), and f (x) = uv r(x) = uv p(x)q(x).
P
k=0
As vf (x) = ur(x), v | uci , i = 0, 1, . . . , n, with u ∧ v = 1, so u | ci for all i. The
polynomial r being primitive, v | 1, so v = ε = ±1.

49
 Let g(x) = εup(x), h(x) = q(x).Then g, h ∈ Z[x], deg(g) ≥ 1, deg(h) ≥ 1, and f = gh
is the product of two non constant polynomials in Z[x].

Proof. (Ex. 6.23)

Let 
Z[x] → Fp [x]
ϕ:
p(x) = a0 + · · · + an x 7→ p(x) = a0 + · · · + an xn ,
n

where ai is the class of ai in Fp . ϕ is a ring homomorphism.

We show that f (x) = g(x)h(x), g, h ∈ Z[x], deg(g) ≥ 1, deg(h) ≥ 1 is impossible.
Indeed in such a situation,
f (x) = xn = g(x)h(x).
As the only irreducible factor of xn is x, the unicity of the decomposition of a polynomial
in irreducible factors in Fp [x] gives

g(x) = λxi , h(x) = µxj , λ, µ ∈ Fp , i, j ∈ N.

As deg(g) ≤ deg(g), deg(h) ≤ deg(h) and deg(g) + deg(h) = n = deg(f ) + deg(g),

this implies that i = deg(f ) = deg(f ), j = deg(g) = deg(g), so i ≥ 1, j ≥ 1. Therefore
p | g(0), p | h(0), so p2 | an = g(0)h(0), which is in contradiction with the hypothesis.
From the lemma we deduce that f (x) is irreducible in Q[x].

50