Solutions to Ireland, Rosen “A Classical Introduction to

Modern Number Theory”

Richard Ganaye

October 29, 2022

Chapter 9

Ex. 9.1 If α ∈ Z[ω], show that α is congruent to either 0, 1, or −1 modulo 1 − ω.

Proof. Let λ = 1 − ω, and α = a + bω ∈ D = Z[ω], a, b ∈ Z.

ω ≡ 1 (mod λ), so α ≡ a + b (mod λ), α ≡ c with c = a + b ∈ Z.
c ≡ 0, 1, −1 (mod 3), and since λ | 3, z ≡ 0, 1, −1 (mod λ).
Every α ∈ D is congruent to either 0, 1, or −1 modulo λ = 1 − ω.
The classes of 0, 1, −1 in D/λD are distinct. Indeed, 1 6≡ −1 (mod λ), if not λ | 2,
so 2 = λλ0 , N (2) = N (λ)N (λ0 ), thus 4 = 3N (λ0 ), so 3 | 4, which is nonsense.
±1 ≡ 0 (mod λ) implies λ | 1, so λ would be a unit, in contradiction with λ prime.
So there exist exactly three classes modulo λ in D : |D/λD| = 3 = N (λ).

Ex. 9.2 From now on we shall set D = Z[ω] and λ = 1 − ω. For µ in D show that we
can write µ = (−1)a ω b λc π1a1 π2a2 · · · πtat , where a, b, c, and the ai are nonnegative integers
and the πi are primary primes.

Proof. Let S the set containing λ = 1 − ω and all primary primes.

We show that,

(a) every prime in D is associate to a prime in S,

(b) no two primes in S are associate.

Let π be a prime in D. There are three cases.

• If N(π) = 3, then π is associate to λ ∈ S, and no associate of λ is primary.

• If N(π) = q 2 , where q ≡ −1 (mod 3) is a rational prime, then π is associate to

q (Proposition 9.1.2), and q is a primary prime. The primes associate to q are
q, −q, ωq, −ωq, −q − ωq, q + ωq, so only q is primary.

• If N(π) = p, where p ≡ 1 (mod 4), then the proposition 9.1.4. shows among the
associates of π exactly one is primary.

Moreover, the norm of two primes belonging to two different cases are distinct, so two
such primes are not associate.

1
 By Theorem 3, Chapter 1, as D = Z[ω] is a principal ideal domain, every µ ∈ D is
of the form Y
µ=u λe(π) ,
π∈S

where u is a unit, so u = (−1)a ω b . Thus

µ = (−1)a ω b λc π1a1 π2a2 · · · πtat ,
where the π are primary primes, and a, b, c and the ai are nonnegative integers.

Ex. 9.3 Let γ a primary prime. To evaluate χγ (µ) we see, by Exercise 2, that it is
enough to evaluate χγ (−1), χγ (ω), χγ (λ), and χγ (π), where π is a primary prime. Since
−1 = (−1)3 we have χγ (−1) = 1. We now consider χγ (ω). Let γ = a + bω and set
a = 3m − 1 and b = 3n. Show that χγ (ω) = ω m+n .
N (γ)−1
Proof. Let γ = a + bω = 3m − 1 + 3nω. Then χγ (ω) = ω 3 (remark (b) of Theorem
1).
N (γ) − 1 = (3m − 1)2 + (3n)2 − 3n(3m − 1) − 1
= 9m2 − 6m + 9n2 − 9nm + 3n
N (γ) − 1
= 3m2 − 2m + 3n2 − 3nm + n ≡ n + m [3]
3
Thus, for γ = a + bω = 3m − 1 + 3nω,
N (γ)−1
χγ (ω) = ω 3 = ω n+m .

Ex. 9.4 (continuation) Show that χγ (ω) = 1, ω, or ω 2 according to whether γ is

congruent to 8,2, or 5 modulo 3λ. In particular, if q is a rational prime, q ≡ 2 (mod 3),
then χq (ω) = 1, ω, or ω 2 according to whether q ≡ 8, 2, or 5 (mod 9). [Hint : γ =
a + bω = −1 + 3(m + nω), and so γ ≡ −1 + 3(m + n) (mod 3λ).]

Proof. λ = 1 − ω, so ω ≡ 1 (mod λ). Thus

m + nω ≡ m + n (mod λ)
3(m + nω) ≡ 3(m + n) (mod 3λ)
γ = −1 + 3(m + nω) ≡ −1 + 3(m + n) (mod 3λ)
Moreover 9 = 3λλ̄ ≡ 0 (mod 3λ), thus γ is congruent modulo 3λ to an integer between
0 and 8 of the form 3k − 1 : γ ≡ 8, 2 or 5 (mod 3λ).
By Ex. 9.3, χγ (ω) = 1 ⇐⇒ m + n ≡ 0 [3], and m + n ≡ 0 [3] implies m + n =
3k, k ∈ Z, so γ ≡ −1 + 9k ≡ −1 ≡ 8 [3λ].
Conversely, if γ ≡ 8 ≡ −1 [3λ], then 3λ | 3(m + n), so λ | m + n, and N(λ) | N(m + n),
3 | (m + n)2 , thus 3 | m + n, m + n ≡ 0 [3], and so χγ (ω) = 1. The two other cases are
similar, so we obtain

χγ (ω) = 1 ⇐⇒ m + n ≡ 0 [3] ⇐⇒ γ ≡ 8 [3λ],

χγ (ω) = ω ⇐⇒ m + n ≡ 1 [3] ⇐⇒ γ ≡ 2 [3λ],
χγ (ω) = ω 2 ⇐⇒ m + n ≡ 2 [3] ⇐⇒ γ ≡ 5 [3λ].

2
 If γ = q is a rational prime, q ≡ 8 [9] implies q ≡ 8 [3λ], since 3λ | 9 = 3λλ̄, thus
χq (ω) = 1.
Conversely, if χq (ω) = 1, then q ≡ 8 [3λ], q − 8 = µ(3λ), µ ∈ D, therefore
(q − 8)2 = N (µ)33 , 33 | (q − 8)2 , thus 32 | q − 8 and so q ≡ 8 [9]. The two other cases
are similar.

χq (ω) = 1 ⇐⇒ q ≡ 8 [9],
χq (ω) = ω ⇐⇒ q ≡ 2 [9],
χq (ω) = ω 2 ⇐⇒ q ≡ 5 [9].

Ex. 9.5 In the text we stated Eisenstein’s result χγ (λ) = ω 2m . Show that χγ (3) = ω 2n .

Proof. Here γ = (3m − 1) + 3nω.

Note that (1 − ω)2 = −3ω, thus χγ ((1 − ω)2 ) = χγ (−1)χγ (3)χγ (ω).
Using Eisenstein’s result (see a proof in Ex.24-26),
χγ ((1 − ω)2 ) = χγ (λ2 ) = χγ (λ)2 = ω 4m = ω m .
As −1 = (−1)3 , χγ (−1) = 1. Finally χγ (ω) = ω m+n by Exercise 9.3. Thus

ω m = χγ (3)ω m+n , χγ (3) = ω −n = ω 2n .

Conclusion :
χγ (3) = ω 2n .

Ex. 9.6 Prove that

(a) χγ (λ) = 1 for γ ≡ 8, 8 + 3ω, 8 + 6ω [9].
(b) χγ (λ) = ω for γ ≡ 5, 5 + 3ω, 5 + 6ω [9].
(c) χγ (λ) = ω 2 for γ ≡ 2, 2 + 3ω, 2 + 6ω [9].
Proof. Here γ = −1 + 3(m + nω) is a primary prime, and χγ (λ) = ω 2m .
χγ (λ) = 1 ⇐⇒ m ≡ 0 [3] ⇒ γ ≡ 8 + 3nω [9] ⇒ γ ≡ 8, 8 + 3ω, 8 + 6ω [9]
χγ (λ) = ω ⇐⇒ m ≡ 2 [3] ⇒ γ ≡ 5 + 3nω [9] ⇒ γ ≡ 5, 5 + 3ω, 5 + 6ω [9]
χγ (λ) = ω 2 ⇐⇒ m ≡ 1 [3] ⇒ γ ≡ 2 + 3nω [9] ⇒ γ ≡ 2, 2 + 3ω, 2 + 6ω [9]

As χγ (λ) ∈ {1, ω, ω 2 }, these 9 cases are the only possibilities. Moreover these 9 cases
are mutually exclusive, since 9 doesn’t divide any difference. Thus the reciprocals are
true.
χγ (λ) = 1 ⇐⇒ γ ≡ 8, 8 + 3ω, 8 + 6ω [9]
χγ (λ) = ω ⇐⇒ γ ≡ 5, 5 + 3ω, 5 + 6ω [9]
χγ (λ) = ω 2 ⇐⇒ γ ≡ 2, 2 + 3ω, 2 + 6ω [9]

3
Ex. 9.7 Find primary primes associate to 1 − 2ω, −7 − 3ω, and 3 − ω.

Proof. :

• (1 − 2ω)ω = 2 + 3ω ≡ 2 (mod 3), so 2 + 3ω is primary, and associate to 1 − 2ω.

N(2 + 3ω) = 7 and 7 is a rational prime, thus 2 + 3ω is a primary prime.

• −7 − 3ω ≡ 2 (mod 3).
N(−7 − 3ω) = 37 and 37 is a rational prime, thus −7 − 3ω is a primary prime.

• (3 − ω)ω 2 = −4 − 3ω ≡ 2 (mod 3), so −4 − 3ω is primary, and associate to 3 − ω.

N(−4 − 3ω) = 13 and 13 is a rational prime, thus −4 − 3ω is a primary prime.

Ex. 9.8 Factor the following numbers into primes in D : 7, 21, 45, 22, and 143.

Proof. 7 = N(2 + 3ω), thus 7 = (2 + 3ω)(2 + 3ω 2 ) = (2 + 3ω)(−1 − 3ω), where 2 + 3ω

and −1 − 3ω are primes in D, since their norm is a prime integer. Since these primes are
primary, they are not associate.
21 = 3 × 7 = −ω 2 λ2 (2 + 3ω)(−1 − 3ω) since 3 = −ω 2 (1 − ω)2 .
45 = 32 × 5 = ωλ4 5, where 5 ≡ 2 (mod 3) is a primary prime in D.
22 = 2 × 11 , where 2 and 11 are primes in D.
143 = 11 × 13 = 11(−4 − 3ω)(−4 − 3ω 2 ) = 11(−4 − 3ω)(−1 + 3ω).

Ex. 9.9 Show that α 6= 0, the residue class of α, is a cube in the field D/πD iff
α(N π−1)/3 ≡ 1 (mod π). Conclude that there are (N π − 1)/3 cubes in (D/πD)∗ .

Solution 1 :

Proof. Let π be a prime in D, N π 6= 3, and α ∈ D, π - α.

α is a cube in (D/πD)∗
⇐⇒ x3 ≡ α (mod π) has a solution in D
⇐⇒ χπ (α) = 1 (by Prop. 9.3.3(a))
N π−1
⇐⇒ α 3 ≡ 1 (mod π)
N π−1
⇐⇒ α 3 = 1.
N π−1
The cubes in (D/πD)∗ are then the roots of the polynomial f (x) = x 3 − 1 in
D/πD.
Let q be the cardinal of the field D/πD. Since q = |D/πD| = N π, N π−1 3 | q − 1,
q−1 q N π−1
f (x) | x − 1 | x − x. By Corollary 2 of Proposition 8.1.1, f has deg(f ) = 3 roots.
Conclusion : there are exactly N π−1
3 cubes in (D/πD)∗ .

Solution 2 :

Proof. Let ϕ : (D/πD)∗ → (D/πD)∗ be the group homomorphism defined by ϕ(x) = x3 .

Then im(ϕ) is the set of cubes in (D/πD)∗ .
The equation x3 = 1 has three distinct solutions 1, ω, ω 2 in D/πD if N π 6= 3 (see the
demonstration of Proposition 9.3.1).
So ker(ϕ) = {1, ω, ω 2 } and | ker(ϕ)| = 3. Thus |im(ϕ)| = |(D/πD)∗ /| ker(ϕ)| =
(N π − 1)/3. There exist exactly N π−1
3 cubes in (D/πD)∗ .

4
 Note : if N π = 3, that is to say, if π is associate to 1 − ω, D/πD = {0, 1, 2}. As
3 3
1 = 1, 2 = 2, all the elements of (D/πD)∗ are cubes.

Ex. 9.10 What is the factorisation of x24 − 1 in D/5D.

Proof. |(D/5D)∗ | = N (5) − 1 = 24, thus x24 − 1 =

Q
(x − α).
α∈(D/5D)∗
(where the α ∈ (D/5D)∗ are of the form α = a + b [ω], 0 ≤ a < 5, 0 ≤ b < 5, (a, b) 6=
(0, 0)).

Ex. 9.11 How many cubes are there in D/5D ?

Proof. By Exercise 9.9, there exist (N (5) − 1)/3 = 8 cubes in (D/5D)∗ (and 0 = 03 is a
cube).

Ex. 9.12 Show that ωλ has order 8 in D/5D and that ω 2 λ has order 24. [Hint : Show
first that (ωλ)2 has order 4.]

Proof. If α = (ωλ)2 , then

α = (ωλ)2 = ω 2 (1 − ω)2 = ω 2 (1 + ω 2 − 2ω) = −3ω 3 = −3.

So α2 = 9 ≡ −1 (mod 5), α4 ≡ 1 (mod 5) and α2 6≡ 1 (mod 5), thus the class of

α = (ωλ)2 has order 4 in (D/5D)∗ , and this implies that ωλ has order 8.
Let β = ω 2 λ. |(D/5D)∗ | = 24, thus [β]24 = 1 (where [β] is the class of β in D/5D.)
To verify that [β] has order 24, it is sufficient to verify that [β]8 6= 1, [β]12 6= 1 :
β 8 = ω 16 λ8 = ωλ8 = (ωλ)8 ω 2 ≡ ω 2 6≡ 1 (mod 5).
β 12 = (ω 2 λ)12 = λ12 = (ωλ)12 ≡ (ωλ)4 ≡ −1 (mod 5) (since (ωλ) has order 8 in
D/5D).
Conclusion : ωλ has order 8, ωλ2 has order 24 in (D/5D)∗ .

Ex. 9.13 Show that π is a cube in D/5D iff π ≡ 1, 2, 3, 4, 1 + 2ω, 2 + 4ω, 3 + ω, or

4 + 3ω (mod 5).
2
Proof. Let π ∈ D, [π] 6= 0. Then [π] is a cube in D/5D iff [π](q −1)/3 = 1, with q = 5,
namely [π]8 = 1 (Prop. 7.1.2, where 3 | q 2 − 1 = 24 = |(D/5D)∗ |).
By Exercise 9.12, the class of γ = ωλ has order 8, thus the 8 elements [γ]k , 0 ≤ k ≤ 7
are distinct roots of the polynomial x8 − 1, which has at most 8 roots. Therefore the
subgroup of cubes in (D/5D)∗ is

{1, [γ], [γ]2 , . . . , [γ]7 }.

5
γ = ω(1 − ω) = ω + 1 + ω = 1 + 2ω, so

γ0 = 1
γ 1 = 1 + 2ω
γ 2 ≡ −3 ≡ 2 [5] (Ex. 9.12)
3
γ = −3 − 6ω ≡ 2 + 4ω [5]
γ 4 ≡ −1 ≡ 4 [5]
γ 5 ≡ −1 − 2ω ≡ 4 + 3ω [5]
γ 6 ≡ 3 [5]
γ 7 ≡ 3 + 6ω ≡ 3 + ω [5]

Conclusion : If π 6≡ 0 (mod 5), π ≡ α3 (mod 5), α ∈ D iff

π ≡ 1, 2, 3, 4, 1 + 2ω, 2 + 4ω, 3 + ω, 4 + 3ω [5].

Ex. 9.14 For which primes π ∈ D is x3 ≡ 5 (mod π) solvable ?

Proof. If π is associate to 5, then 53 ≡ 0 ≡ 5 (mod π), so x3 ≡ 5 (mod π) is solvable.

If π is a primary prime not associate to 5, the Law of Cubic Reciprocity gives

5 ≡ x3 [π], x ∈ D ⇐⇒ χπ (5) = 1
⇐⇒ χ5 (π) = 1
⇐⇒ π is a cube in D/5D
⇐⇒ π ≡ 1, 2, 3, 4, 1 + ω, 2 + 4ω, 3 + ω, 4 + 3ω [5]

(see Ex. 9.13)

Conclusion : the equation 5 ≡ x3 [π], x ∈ D is solvable iff the primary prime associate
to π is congruent modulo 5 to 1, 2, 3, 4, 1 + 2ω, 2 + 4ω, 3 + ω, 4 + 3ω.
Examples :
• q = 23 is a primary prime congruent to 3 modulo 5, thus the equation x3 ≡ 5
(mod 23) has a solution x ∈ D (x = 19).
• −4 − 3ω is the primary prime associate to the prime 3 − ω, and −4 − 3ω ≡ 1 + 2ω
(mod 5), thus the equation x3 ≡ 5 (mod 3 − ω) has a solution a + bω ∈ Z[ω].
Indeed , 73 ≡ 53 ≡ 113 ≡ 5 (mod 13), and 3 − ω | 13, so 73 ≡ 53 ≡ 113 ≡ 5
(mod 3 − ω).

Ex. 9.15 Suppose that p ≡ 1 (mod 3) and that p = ππ, where π is a primary prime in
D. Show that x3 ≡ a (mod p) is solvable in Z iff χπ (a) = 1 . We assume that a ∈ Z.

Proof. Since π | p, if x3 ≡ a (mod p), x ∈ Z, then x3 ≡ a (mod π), thus χπ (a) = 1.

Conversely, suppose that χπ (a) = 1. Then the equation y 3 ≡ a (mod π) has a
solution y = u + vω, u, v ∈ Z. Moreover, the class of y has a representative x ∈ Z
modulo π (see the proof of Proposition 9.2.1) :

y≡x (mod π), x ∈ Z.

6
 So x3 ≡ a (mod π) has a solution x ∈ Z.
Thus π | x3 −a, N (π) = p | (x3 −a)2 , therefore p | x3 −a in Z, and so x3 ≡ a (mod p).
Conclusion ; if p ≡ 1 (mod 3) , p = ππ, where π is is a primary prime and a ∈ Z,

∃x ∈ Z, x3 ≡ a (mod p) ⇐⇒ χπ (a) = 1.

In other words, x3 ≡ a (mod π) is solvable in D iff it is solvable in Z.

Ex. 9.16 Is x3 ≡ 2 − 3ω (mod 11) solvable ? Since D/11D has 121 elements this is
hard to resolve by straightforward checking. Fill in the details of the following proof that
it is not solvable. χπ (2 − 3ω) = χ2−3ω (11) and so we shall have a solution iff x3 ≡ 11
(mod 2 − 3ω) is solvable. This congruence is solvable iff x3 = 11 (mod 7) is solvable in
Z. However, x3 ≡ a (mod 7) is solvable in Z iff a ≡ 1 or 6 (mod 7).

Warning : false sentence, since

N (2 − 3ω) = (2 − 3ω)(2 − 3ω 2 ) = 4 + 9 − 6(ω + ω 2 ) = 4 + 9 + 6 = 19 (and not 7!).

Proof. Since 19 is a rational prime, and since π = 2 − 3ω and 11 are primary primes, by
the Law of Cubic Reciprocity, and by Exercise 9.15 (with p = 11 ≡ 1 (mod 3)),

∃x ∈ D, 2 − 3ω ≡ x3 [11] ⇐⇒ χ11 (2 − 3ω) = 1

⇐⇒ χ2−3ω (11) = 1
⇐⇒ ∃x ∈ D, x3 ≡ 11 [2 − 3ω]
⇐⇒ ∃x ∈ Z, x3 ≡ 11 [19]

Moreover, by Proposition 7.1.2 (with p = 19, d = (p − 1) ∧ 3 = 3, (p − 1)/d = 6),

∃x ∈ Z, x3 ≡ 11 [19] ⇐⇒ 116 ≡ 1 (mod 19),

which is true : 116 = 1213 = (19 × 6 + 7)3 ≡ 49 × 7 ≡ 11 × 7 ≡ 77 ≡ 1 [19].

Conclusion : there exists x ∈ D such that 2 − 3ω ≡ x3 (mod 11).
With some computer code, we find a solution x = 1 + 8ω (and its associates ω 2 x =
7 − ω, ωx = −8 − 7ω ≡ 3 + 4ω (mod 11)) :

x3 = (1 + 8ω)3 = 321 − 168ω ≡ 2 − 3ω (mod 11).

Note : The sentence becomes true if we replace 2 − 3ω by the primary prime 2 + 3ω.
Since N (2 + 3ω) = 7, with the same reasoning,

∃x ∈ D, 2 + 3ω ≡ x3 [11] ⇐⇒ χ2+3ω (11) = 1

⇐⇒ ∃x ∈ D, x3 ≡ 11 [2 + 3ω]
⇐⇒ ∃x ∈ Z, x3 ≡ 11 ≡ 4 [7]
⇐⇒ 42 ≡ 1 (mod 7)

but 42 ≡ 2 6≡ 1 (mod 7), so the equation x3 ≡ 2 + 3ω (mod 11) is not solvable.

7−1
(x3 ≡ a (mod 11) is solvable in Z iff a 3 = a2 ≡ 1 (mod 7) iff a ≡ ±1 (mod 7).)

7
Ex. 9.17 An element γ ∈ D is called primary if γ ≡ 2 (mod 3). If γ and ρ are
primary, show that −γρ is primary. If γ is primary, show that γ = ±γ1 γ2 . . . γt , where
the γi are (not necessarily distinct) primary primes.

Proof. If γ ≡ 2, ρ ≡ 2 (mod 3), then −γρ ≡ −2 × 2 ≡ 2 (mod 3), so −γρ is primary.

By Ex. 9.2, γ can be written

γ = (−1)a ω b λc π1a1 · · · πtat ,

where πi ≡ 2 (mod 3), a ∈ {0, 1}, b ∈ {0, 1, 2}.

As πi ≡ −1 (mod 3), and γ ≡ −1 (mod 3), we obtain ω b λc ≡ ±1 (mod 3). We prove
that b = c = 0.
Note that λ2 = (1 − ω)2 = −3ω ≡ 0 (mod 3). If c ≥ 2, we would obtain γ ≡ 0
(mod 3), in contradiction with the hypothesis, thus c = 0 or c = 1.
If c = 1,

ω b λc ∈ {1 − ω, ω(1 − ω), ω 2 (1 − ω)} = {1 − ω, 1 + 2ω, −2 − ω}.

Since 1 − ω 6≡ ±1, 1 + 2ω 6≡ ±1, −2 − ω 6≡ ±1 (mod 3), this is impossible, so c = 0.

Then ω b ≡ ±1 (mod 3), where ω b ∈ {1, ω, −1 − ω}. Since ω 6≡ ±1 (mod 3), and
−1 − ω 6≡ ±1 (mod 3), then ω b = 1, 0 ≤ b ≤ 2, thus b = 0.
Finally, γ = (−1)a π1a1 · · · πtat .
Conclusion : every primary γ ∈ D is under the form

γ = ±γ1 γ2 · · · γt ,

where the γi are primary primes.

Ex. 9.18 (continuation) If γ = ±γ1 γ2 · · · γt is a primary decomposition of the primary

element γ, define χγ (α) = χγ1 (α)χγ2 (α) · · · χγt (α). Prove that χγ (α) = χγ (β) if α ≡ β
(mod γ) and χγ (αβ) = χγ (α)χγ (β). If ρ is primary, show that χρ (α)χγ (α) = χ−ργ (α).

Proof. If α ≡ β [γ], then α ≡ β (mod γi ), 1 ≤ i ≤ t, so χγi (α) = χγi (β), thus χγ (α) =
χγ (β).
By Proposition 9.3.3,

χγ (αβ) = χγ1 (αβ)χγ2 (αβ) · · · χγt (αβ)

= χγ1 (α)χγ2 (α) · · · χγt (α)χγ1 (β)χγ2 (β) · · · χγt (β)
= χγ (α)χγ (β)

Finally, if ρ = ±ρ1 ρ2 · · · ρl is primary, then −ργ = ±ρ1 ρ2 · · · ρl γ1 γ2 · · · γt is primary by

Ex. 9.17, therefore

χ−ργ (α) = (χρ1 χρ2 · · · χρl χγ1 χγ2 · · · χγt )(α) = χρ (α)χγ (α).

Note : The unit −1 is primary by définition, and −1 is the opposite of the empty
product, so for all α in D, χ−1 (α) = 1 by definition. The result of the exercises remain
true if we accept the unit −1 as a primary element.

8
Ex. 9.19 Suppose that γ = A + Bω is primary and that A = 3M − 1 and B = 3N .
Prove that χγ (ω) = ω M +N and that χγ (λ) = ω 2M .

Proof. We verify first that if γ = −γ1 γ2 , with

γ = A + Bω, A = 3M − 1, B = 3N,
γ1 = A1 + B1 ω, A1 = 3M1 − 1, B1 = 3N1 ,
γ2 = A2 + B2 ω, A2 = 3M2 − 1, B2 = 3N2 ,

then M ≡ M1 + M2 (mod 3), N ≡ N1 + N2 (mod 3).

−γ1 γ2 = −A1 A2 + B1 B2 + (−A1 B2 − A2 B1 + B1 B2 )ω = A + Bω,

therefore

3M − 1 = A = −A1 A2 + B1 B2 ≡ 3(M1 + M2 ) − 1 (mod 9),

thus M ≡ M1 + M2 (mod 3).

3N = B = −A1 B2 − A2 B1 + B1 B2 ≡ 3(N1 + N2 ) (mod 9),

thus N ≡ N1 + N2 (mod 3).

By induction, if γ = ±γ1 γ2 · · · γt = (−1)t−1 γ1 γ2 · · · γt , where γi = Ai + Bi ω, Ai =
3Mi − 1, Bi = 3Ni , then

M ≡ M1 + · · · + M t (mod 3), N ≡ N1 + · · · + Nt (mod 3).

By Exercise 9.3,

χγ (ω) = χγ1 (ω) · · · χγt (ω)

= ω M1 +N1 · · · ω Mt +Nt
= ω (M1 +···+Mt )+(N1 +···+Nt )
= ω M +N ,

and by Eisenstein’s result,

χγ (λ) = χγ1 (λ) · · · χγt (λ)

= ω 2M1 · · · ω 2Mt
= ω 2(M1 +···+Mt )
= ω 2M .

Conclusion : if γ = 3M − 1 + 3N ω, then

χγ (ω) = ω M +N , χγ (λ) = ω 2M .

9
Ex. 9.20 If γ and ρ are primary, show that χγ (ρ) = χρ (γ).

Proof. ρ, γ are written

ρ = ±ρ1 ρ2 · · · ρl ,
γ = ±γ1 γ2 · · · γm ,

where ρi , γi are primary primes. By the law of Cubic Reciprocity, we obtain

m
Y
χγ (ρ) = χγj (ρ)
j=1
m Y
Y l
= χγj (ρi )
j=1 i=1
l Y
Y m
= χγj (ρi )
i=1 j=1
l Y
Y m
= χρi (γj )
i=1 j=1
l
Y
= χρi (γ)
i=1
= χρ (γ).

(if γ = −1, or ρ = −1, some products are empty, but the result remains true :
χ−1 (ρ) = 1 = χρ (−1).)

Ex. 9.21 If γ is primary, show that there are infinitely many primary primes π such
that x3 ≡ γ (mod π) is not solvable. Show also that there are infinitely many primary
primes π such that x3 ≡ ω (mod π) is not solvable and the same for x3 ≡ λ (mod π).
(Hint: Imitate the proof of Theorem 3 of Chapter 5.)

Proof. a) As some primary elements of D may be cubes, by example 53 + 36ω =

(−1 + 3ω)3 , we must of course suppose that γ is not the cube of some element of
D (in the contrary case x3 ≡ γ (mod π) is solvable for all prime π).
Note first that for all primes π in D, there exists σ ∈ D such that χπ (σ) = ω.
Indeed, there exist (N π − 1)/3 cubes in (D/πD)∗ , which has N π − 1 elements, so
there exists an element τ ∈ (D/πD)∗ which is not a cube, therefore there exists
τ ∈ D such that χπ (τ ) 6= 1. If χπ (τ ) = ω, we put σ = τ and if χπ (τ ) = ω 2 , we put
σ = τ 2 . In the two cases, χπ (σ) = ω.

n
Let γ ∈ D, where γ is primary. Then γ = ±γ1n1 γ2n2 · · · γp p , where the γi are distinct
primary primes. Write ni = 3qi +ri , ri ∈ {0, 1, 2}. Then grouping in γ 0 the γ ri such
n
that ri 6= 0, we can write γ = δ 3 γ 0 , γ 0 = γ1r1 γ2r2 · · · γlrl , ri ∈ {1, 2}, δ = ±γ1q1 · · · γp p ∈
D (−1 is a cube). Since by hypothesis γ is not a cube, l ≥ 1. Moreover the equation
x3 ≡ γ (mod π) is solvable iff x3 ≡ γ 0 (mod π) is solvable. We may then suppose
that
γ = γ1r1 γ2r2 · · · γlrl , 1 ≤ ri ≤ 2,

10
 without cubic factors.
Note that the γi are not associate to λ = 1 − ω (see Ex. 9.17).
Let A = {λ1 , λ2 , . . . , λk } a set (possibly empty) of distinct primary primes λi
(therefore they are not associate), and not associate neither to γi , 1 ≤ i ≤ l, nor to
λ = 1 − ω.
We will show that we can find a primary prime λk+1 distinct of the λi with the
same properties and such that the equation x3 ≡ λ (mod λk+1 ) is not solvable.
This will prove the existence of infinitely many primes π such that the equation
x3 ≡ λ (mod π) is not solvable.
Using the initial note, let σ ∈ D such that χγl (σ) = ω. As D is a principal ideal
domain, the Chinese Remainder Theorem is valid. Since 3 = λλ = −ω 2 λ2 is
relatively prime to γi , λi , there exists β ∈ D such that

β ≡ 2 [3]
β ≡ 1 [λi ] (1 ≤ i ≤ k)
β ≡ 1 [γi ] (1 ≤ i ≤ l − 1)
β ≡ σ [γl ]

The first equation show that β is primary, so β = (−1)m−1 β1 . . . βm , where the βi

are primary primes.
By Exercise 9.20,

χβ (γ) = χβ (γ1 )r1 · · · χβ (γl )rl = χγ1 (β)r1 · · · χγl (β)rl .

As χγi (1) = 1 (1 ≤ i ≤ l − 1), and χγl (β) = χγl (σ) = ω, we obtain χβ (γ) = ω rl 6= 1,
since rl = 1 or rl = 2.
By Exercise 9.18, χρ (α)χγ (α) = χ−ργ (α), with primary ρ, γ, so by induction, as
β = (−1)m−1 β1 · · · βm ,

χβ (γ) = χβ1 (γ) · · · χβm (γ) 6= 1.

Thus there exists a subscript j such that χβj (γ) 6= 1.

We can then take λk+1 = βj . Indeed, since β ≡ 1 [λi ] and β 6≡ 0 [γi ], βj is distinct
of the λi and γi , and βj is not associate to λ since β ≡ 2 (mod 3).
As χλk+1 (γ) 6= 1, the equation x3 ≡ γ [λk+1 ] is not solvable, so λk+1 is convenient.
Conclusion : if γ ∈ D is primary and is not a cube in D, there exist infinitely many
primes π ∈ D such that the equation x3 ≡ γ [π] is not solvable.

b) We show that x3 ≡ ω [π] has no solution for infinitely many primes π.

To initialize the induction, we display such a prime π, namely π = 2 + 3ω. Indeed,
N (π) = 4 + 9 − 6 = 7, 7 is a rational prime, so π is a primary prime in D, of
the form π = 3m − 1 + 3nω, with n = m = 1, so χπ (ω) = ω m+n = ω 2 6= 1 : the
equation x3 ≡ ω [π] is not solvable. Moreover π is not associate to λ = 1 − ω.
Suppose now the existence of a set A = {λ1 , λ2 , . . . , λl }, l ≥ 1, of distinct primary
primes λi , not associate to λ and such the equation x3 ≡ ω [λi ] is not solvable for

11
 each i, 1 ≤ i ≤ l. We will show that we can add a prime λl+1 to the set A with
the same properties.
Let
β = 3(−1)l−1 λ1 · · · λl − 1.

(−1)l−1 λ1 · · · λl is primary, so (−1)l−1 λ1 · · · λl = 3m − 1 + 3nω, m, n ∈ Z.

β = 3(3m − 1 + 3nω) − 1 = 3(3m − 1) − 1 + 9nω = 3M − 1 + 3N ω, where
M = 3m − 1, N = 3n. By Exercise 9.19,

χβ (ω) = ω M +N = ω 3m−1+3n = ω 2 6= 1.

As β = ±β1 · · · βm , where the βi are primary primes, χβ (ω) = χβ1 (ω) · · · χβm (ω) 6=
1, so there exists a subscript i such that χβi (ω) 6= 1.
Since β = 3(−1)l−1 λ1 · · · λl − 1, βi is associate neither to λi nor to λ. Moreover
χβi (ω) 6= 1, thus the equation x3 ≡ ω [βi ] is not solvable : λl+1 = βi is convenient.
Conclusion : the equation x3 ≡ ω [π] is not solvable for infinitely many primes π.

c) We show that x3 ≡ λ [π] has no solution for infinitely many primes π.

To initialize the induction, we display such a prime π, namely π = −4+3ω. Indeed,
N (π) = 16 + 9 + 12 = 37, 37 is a rational prime, so π is a primary prime in D, of
the form π = 3m − 1 + 3nω, with m = −1, n = 1, so χπ (λ) = ω 2m = ω 6= 1 : the
equation x3 ≡ λ [π] is not solvable.
Suppose now the existence of a set A = {λ1 , λ2 , . . . , λl }, l ≥ 1, of distinct primary
primes λi , not associate to λ and such the equation x3 ≡ λ [λi ] is not solvable. We
will show that we can add a prime λl+1 to the set A with the same properties.
Let
β = 3(−1)l−1 λ1 · · · λl − 1.
(−1)l−1 λ1 · · · λl is primary, so (−1)l−1 λ1 · · · λl = 3m − 1 + 3nω, m, n ∈ Z.
β = 3(3m − 1 + 3nω) − 1 = 3(3m − 1) − 1 + 9nω = 3M − 1 + 3N ω, where
M = 3m − 1, N = 3n. By Exercise 9.19,

χβ (λ) = ω 2M = ω 2(3m−1) = ω 6= 1.

As β = ±β1 · · · βm , where the βi are primary primes, χβ (ω) = χβ1 (ω) · · · χβm (ω) 6=
1, so there exists a subscript i such that χβi (λ) 6= 1.
Since β = 3(−1)l−1 λ1 · · · λl − 1, βi is associate neither to λi nor to λ. Moreover
χβi (λ) 6= 1, thus the equation x3 ≡ λ [βi ] is not solvable : λl+1 = βi is convenient.
Conclusion : the equation x3 ≡ λ [π] is not solvable for infinitely many primes π.

Ex. 9.22 (continuation) Show in general that if γ ∈ D and x3 ≡ γ (mod π) is solvable

for all but finitely finitely many primary primes π, then γ is a cube in D.

Proof. Let γ ∈ D and suppose that γ is not a cube in D. We will show that the equation
x3 ≡ γ [π] is not solvable for infinitely primes π ∈ D.
By Exercise 9.2, we can write
n
γ = (−1)u ω v λw γ1n1 · · · γp p ,

12
where the γi are distinct primary primes, not associate to λ. Let v = 3q + b, w =
3q 0 +c, ni = 3qi +ri , with the remainders b, c, ri in {0, 1, 2}. Grouping the factors with null
remainders, we obtain γ = δ 3 γ 0 , γ 0 = ω b λc γ1r1 · · · γlrl , with b, c, ri in {1, 2}, δ ∈ D, l ≥ 0
(−1 is a cube).
Moreover the equation x3 ≡ γ [π] is solvable iff the equation x3 ≡ γ 0 [π] is solvable.
So we may suppose that

γ = ω b λc γ1r1 · · · γlrl , b ∈ {1, 2}, c ∈ {1, 2}, ri ∈ {1, 2},

without cubic factors.

• Case 1 : l ≥ 1.
Let A = {λ1 , . . . , λk } a possibly empty set of distinct primary primes λi , distinct
of the γi , not associate to λ, and such that the equation x3 ≡ γ [λi ] is not solvable.
We will show that we can add a prime λk+1 with the same properties.
Suppose that l ≥ 1. We have proved in Ex. 9.21 that there exists σ ∈ D such that
χγl (σ) = ω. Since 9, λi , γj are relatively prime, there exists β ∈ D such that

β ≡ −1 [9]
β ≡ 1 [λi ], 1 ≤ i ≤ k
β ≡ 1 [γi ], 1 ≤ i ≤ l − 1
β ≡ σ [γl ]

β ≡ −1 [9], thus β ≡ −1 [3] : β is primary, of the form β = 3M − 1 + 3N ω.

β = 3M − 1 + 3N ω ≡ −1 [9], so 3M + 3N ω ≡ 0 [9], M + N ω ≡ 0 [3], thus
3 | M, 3 | N .
By Exercise 9.18,

χβ (ω) = ω M +N = 1
χβ (λ) = ω 2M = 1

As β and γi are primary, χβ (γi ) = χγi (β) = χγi (1) = 1 (1 ≤ i ≤ l − 1).

χβ (γ) = χβ (ω)b χβ (λ)c χβ (γ1 )r1 · · · χβ (γl )rl = χβ (γl )rl = χγl (β)rl = χγl (σ)rl =
ω rl 6= 1, since rl ∈ {1, 2}.
β = ±β1 · · · βm , with βi primary primes, therefore

χβ (γ) = (χβ1 · · · χβm )(γ) 6= 1.

Thus there exists a subscript i such that χβi (γ) 6= 1, so x3 ≡ γ [βi ] is not solvable.
Moreover β ≡ 1 [γi ], so βi is not associate to any γj . Similarly, βi is not associate
to any γj , and β ≡ −1 [9], therefore βi is not associate to λ. So λk+1 = βi is
convenient.
There exist infinitely many π such that x3 ≡ γ [π] is not solvable.

• Case 2 : l = 0, so γ = ω b λc , 1 ≤ b ≤ 2, 1 ≤ c ≤ 2.
π0 = 2 − 3ω is a primary prime (N (π0 ) = 19).
Let A = {λ1 , . . . , λk } a possibly empty set of distinct primary primes λi 6= π0 such
that the equation x3 ≡ γ [λi ] is not solvable. We will show that we can add a prime
λk+1 with the same properties.

13
 Let β = 9(−1)k−1 λ1 · · · λk + 2 − 3ω.
β ≡ 2 [3] : β is primary.
Moreover (−1)k−1 λ1 · · · λk is primary, so

(−1)k−1 λ1 · · · λk = 3m − 1 + 3nω, m ∈ Z, n ∈ Z.
Then

β = 9(3m − 1 + 3nω) + 2 − 3ω
= 27m − 7 + (27n − 3)ω
= 3(9m − 2) − 1 + 3(9n − 1)ω
= 3M − 1 + 3N ω,

where M = 9m − 2, N = 9n − 1. Therefore

χβ (ω) = ω M +N = ω 9m−2+9n−1 = 1
χβ (λ) = ω 2M = ω 2(9m−2) = ω 2 6= 1

β = ±β1 · · · βm , where the βi are primary primes.

χβ (γ) = χβ (ω)b χβ (λ)c = ω 2c 6= 1 since c = 1 or c = 2.

χβ (γ) = (χβ1 · · · χβm )(γ) 6= 1.

Thus there exists a subscript i such that χβi (γ) 6= 1, so x3 ≡ γ [βi ] is not solvable.
As βi | β = 9(−1)k−1 λ1 · · · λk + 2 − 3ω, if βi = λj for some subscript j, λj |
π0 = 2 − 3ω, so λj = π0 , which is a contradiction, thus βi 6∈ A. Similarly, if
βi = π0 = 2 − 3ω, then π0 | 9λ1 · · · λk , and π0 is relatively prime to λ, so π0 = λj
for some subscript j : this is a contradiction, thus βi 6= π0 . λk+1 = βi is convenient.
So there exist infinitely many π such that x3 ≡ γ [π] is not solvable.
• Conclusion :
if γ is not a cube in D, there exist infinitely many primes π such that x3 ≡ γ [π]
is not sovable.
By contraposition, if the equation x3 ≡ γ [π] is solvable for every prime π, at the
exception perhaps of the primes in a finite set, then γ is a cube in D.

Ex. 9.23 Suppose that p ≡ 1 (mod 3). Use Exercise 5 to show that x3 ≡ 3 (mod p) is
solvable in Z iff p is of the form 4p = C 2 + 243B 2 .

Proof. Let p be a rational prime, p ≡ 1 (mod 3), then p = ππ, where π ∈ D is a primary
prime : π = a + bω = 3m − 1 + 3nω.
• Suppose that there exists x ∈ Z such that x3 ≡ 3 (mod p). Then x3 ≡ 3 (mod π),
so χπ (3) = 1. By Exercise 9.5, ω 2n = χπ (3) = 1, thus 3 | n, therefore 9 | b = 3n,
namely b = 9B, B ∈ Z.
p = N π = a2 +b2 −ab, 4p = (2a−b)2 +3b2 = C 2 +243B 2 , where C = 2a−b, B = b/9.
So there exists C, B ∈ Z such that 4p = C 2 + 243B 2 .

14
 • Conversely, suppose that there exist C, B ∈ Z such that 4p = C 2 + 243B 2 .
As 4p = (2a − b)2 + 3b2 = C 2 + 3(9B)2 , from the unicity proved in Exercise 8.13,
we obtain b = ±9B, so 9 | b = 3n, 3 | n, and χπ (3) = ω 2n = 1.
Thus there exists x ∈ D such that x3 ≡ 3 (mod π). As p ≡ 1 (mod 3), D/πD =
{0, . . . , p − 1}, so there exists h ∈ Z such that x ≡ h (mod π), and h3 ≡ 3 (mod π).
Therefore p = N π | N (h3 − 3), namely p | (h3 − 3)2 , where p is a rational prime,
thus p | h3 − 3 : there exists x ∈ Z such that x3 ≡ 3 (mod p).
Moreover 4p = C 2 + 243B 2 implies p ≡ 1 (mod 3).

(p ≡ 1 [3] and ∃x ∈ Z, x3 ≡ 3 [p]) ⇐⇒ ∃C ∈ Z, ∃B ∈ Z, 4p = C 2 + 243B 2 .

Ex. 9.24 Let π = a + bω be a complex primary element of D = Z[ω]. Put a =

3m − 1, b = 3n, p = N (π).

(a) (p − 1)/3 ≡ −2m + n (mod 3).

(b) (a2 − 1)/3 ≡ m (mod 3).

(c) χπ (a) = ω m .

(d) χπ (a + b) = ω 2n χπ (1 − ω).

Lemma. Let a ∈ Z, a ≡ −1 (mod 3), and b ∈ Z such that a∧b = 1. Then χa (b) = 1.

Proof. (of Lemma.)

If q is a rational prime, q ≡ 2 (mod 3), and q ∧ b = 1, then χq (b) = 1 (Prop. 9.3.4,
Corollary).
If p is a rational prime, p ≡ 1 (mod 3) and p ∧ b = 1, then p = ππ, with π primary
prime in D (and also π), and by definition of χp , χp (b) = χπ (b)χπ (b).
As χπ (b) = χπ (b) = χπ (b) (Prop. 9.3.4(b)), so χp (b) = χπ (b)χπ (b) = χπ (b)χπ (b) = 1.
a has a decomposition in prime factors of the form :

a = ±q1 q2 · · · qk p1 p2 · · · pl = ±q1 q2 · · · qk π1 π1 π2 π2 · · · πl πl ,

where qi ≡ −1, pj ≡ 1 (mod 3), and the πk are primary primes (since all these elements
are primary, the symbol ± is (−1)k−1 ). Thus, by definition of χa ,

χa (b) = χq1 (b) · · · χqk (b) χπ1 (b)χπ1 (b) · · · χπl (b)χπl (b) = 1.

The result remains true if a = −1 : then, by definition, χa (b) = 1.

Proof. (of Ex 9.24.) By hypothesis, π is a primary element, so π = 3m − 1 + 3nπ, m, n ∈

Z. We don’t suppose in this proof that π is a prime element, so p = N (π) is not
necessarily prime.

15
 (a) p − 1 = (3m − 1)2 + (3n)2 − 3n(3m − 1) − 1 ≡ −6m + 3n (mod 9), thus
p−1
≡ −2m + n (mod 3).
3

(b) a2 − 1 = (3m − 1)2 − 1 ≡ −6m (mod 9), thus

a2 − 1
≡ m (mod 3).
3

(c) As π, a are primary, by Exercise 9.20, χπ (a) = χa (π).

Since π ≡ bω (mod a), χa (π) = χa (b)χa (ω).
By Exercise 9.18, as a = 3m − 1, χa (ω) = ω M +N , where M = m, N = 0, so

χa (ω) = ω m .

Here a is relatively prime to b in Z : if a rational prime r divides a, b, then r | π in

D, thus r | π, so r2 | ππ = p in D, thus r2 | p in Z, which is absurd. The Lemma
gives then χa (b) = 1.
We conclude that χa (b) = 1, χa (ω) = ω m , so χπ (a) = χa (π) = χa (b)χa (ω) = ω m .

χπ (a) = ω m .

(d)
a + b = [(a + b)ω]ω −1 ,
and
(a + b)ω = (a + bω) + aω − a ≡ a(ω − 1) (mod π),
thus
a + b ≡ −a(1 − ω)ω −1 [π],
χπ (a + b) = χπ (1 − ω)χπ (a)χπ (ω)−1 ,
χπ (a) = ω m by (c), and χπ (ω) = ω m+n (Ex. 9.3), thus

χπ (a + b) = ω 2n χπ (1 − ω).

Ex. 9.25 Show that χa+b (π) may be computed as follows.

(a) χa+b (π) = χa+b (1 − ω).

(b) χa+b (π) = ω 2(m+n) .

Proof. (a) π = a + bω and a ≡ −b (mod a + b), thus π ≡ −b(1 − ω) (mod a + b). So

χa+b (π) = χa+b (b)χa+b (1 − ω).

Since a ∧ b = 1, (a + b) ∧ b = 1 : as in Ex. 9.24, χa+b (b) = 1. So

χa+b (π) = χa+b (1 − ω).

16
 (b) Since the character χa+b has order 3,

χa+b (1 − ω) = (χa+b ((1 − ω)2 ))2

= (χa+b (−3ω))2
= [χa+b (3)χa+b (ω)]2

χa+b (3) = 1 car (a + b) ∧ 3 = (3(m + n) − 1) ∧ 3 = 1.

χa+b (ω) = ω m+n (Ex. 9.19).
Conclusion :
χa+b (1 − ω) = ω 2(m+n) .

Ex. 9.26 Combine the previous two exercises to conclude that χπ (1 − ω) = ω 2m .

Proof. Since π and a + b are primary elements of D, by Exercise 9.20,

χπ (a + b) = χa+b (π).

By Exercises 9.24 and 9.25,

χπ (a + b) = ω 2n χπ (1 − ω)
χa+b (π) = ω 2(m+n)

Thus ω 2n χπ (1 − ω) = ω 2(m+n) . Consequently

χπ (1 − ω) = ω 2m .

Ex. 9.27 Let π = a + bi be a primary irreducible in Z[i], b 6= 0. Show

(a) a ≡ (−1)(p−1)/4 (mod 4), p = N (π).

(b) b ≡ (−1)(p−1)/4 − 1 (mod 4).

(Wrong sentence for (b) in an older edition.)

Proof. Let π = a + bi a primary prime in Z[i], b 6= 0, such that p = N (π) :

p = ππ̄ = a2 + b2 ≡ 1 [4].

By Lemma 6, Section 7, a is odd, b even, and

(a ≡ 1 [4], b ≡ 0 [4]) or (a ≡ 3 [4], b ≡ 2 [4]).

(a) • Case 1 : a ≡ 1 [4], b ≡ 0[4]. Then a = 4A + 1, b = 4B, A, B ∈ Z, so

(a2 + b2 − 1)/4 = 4A2 + 4B 2 + 2A is even :
2 2
(−1)(p−1)/4 = (−1)(a +b −1)/4 = 1, and a ≡ 1[4], thus a ≡ (−1)(p−1)/4 [4].

17
 • Case 2 : a ≡ 3 [4], b ≡ 2 (mod 4).
a = 4A+3, b = 4B +2, a2 +b2 −1 = 16A2 +24A+9+16B 2 +16B +4−1 ≡ 4 [8],
2 2
so (a2 + b2 − 1)/4 ≡ 1 [2], (−1)(p−1)/4 = (−1)(a +b −1)/4 = −1, and a ≡ −1 [4],
thus a ≡ (−1)(p−1)/4 [4].
In both cases,
a ≡ (−1)(p−1)/4 [4].

(b) In every case, b ≡ a − 1 [4], thus

b ≡ (−1)(p−1)/4 − 1 [4].

In other words, for all primary primes π = a + bi such that N (π) = p,

p ≡ 1 [8] ⇐⇒ π ≡ 1 [4],
p ≡ 5 [8] ⇐⇒ π ≡ 3 + 2i [4].

Ex. 9.28 The notation being as in Exercise 27 show χπ (π) = χπ (2)χπ (a).

Proof. π = a + bi, π = a − bi = 2a − π ≡ 2a [π], thus, by Proposition 9.8.3 (e) :

χπ (π) = χπ (2a) = χπ (2)χπ (a).

Ex. 9.29 By Exercise 9.27, a(−1)(p−1)/4 is primary. Use biquadratic reciprocity to

2
show χπ (a(−1)(p−1)/4 ) = (−1)(a −1)/8 .

Proof. a ≡ (−1)(p−1)/4 [4] (Ex. 9.27(a)), a(−1)(p−1)/4 ≡ 1 [4], thus a(−1)(p−1)/4 is

primary (if a 6= ±1).
2
If a = ±1 is an unit, a(−1)(p−1)/4 = 1 and χπ (a(−1)(p−1)/4 ) = 1 = (−1)(a −1)/8 , so
we can suppose that a is not an unit.
As a(−1)(p−1)/4 ≡ 1 (mod 4), the Law of Biquadratic Reciprocity (Prop. 9.9.8) gives

χπ (a(−1)(p−1)/4 ) = χa(−1)(p−1)/4 (π)

= χa (π) (Prop.9.8.3(f))
= χa (a + bi)
= χa (bi)
= χa (b)χa (i).

As a ∧ b = 1 (since p = a2 + b2 ), χa (b) = 1 (Prop. 9.8.5, with a 6= 1), so

χπ (a(−1)(p−1)/4 ) = χa (i).

If a ≡ 1 [4], Proposition 8.9.6 gives χa (i) = (−1)(a−1)/4 . Write a = 4A + 1, A ∈ Z.

Then
2 2
(−1)(a −1)/8 = (−1)2A +A = (−1)A = (−1)(a−1)/4 = χa (i).

18
 If a ≡ −1 [4], then χa (i) = χ−a (i) = (−1)(−a−1)/4 by the same proposition. Write
a = 4A − 1, A ∈ Z. Then
2 −1)/8 2 −A
(−1)(a = (−1)2A = (−1)−A = (−1)(−a−1)/4 = χa (i).

So, for each odd a, a 6= ±1,

2 −1)/8
χa (i) = i(a .
Conclusion : if π = a + bi is a primary irreducible such that N (π) = p, then
2 −1)/8
χπ (a(−1)(p−1)/4 ) = (−1)(a .

2 −1)/8
Ex. 9.30 Use the preceding two exercises to show χπ (π) = χπ (2)(−1)(a .

Proof. By Exercises 9.28, 9.29, and χπ (−1) = (−1)(a−1)/2 (Prop. 9.8.3(d)),

χπ (π) = χπ (2)χπ (a)

= χπ (2)χπ (a(−1)(p−1)/4 )(χπ (−1))(p−1)/4
2 −1)/8
= χπ (2)(−1)(a ((−1)(a−1)/2 )(p−1)/4
2 −1)/8
= χπ (−2)(−1)(a ((−1)(a−1)/2 )(p+3)/4
2 −1)/8
= χπ (−2)(−1)(a (−1)((a−1)/2) ((p+3)/4) .

If a ≡ 1 (mod 4), then (−1)(a−1)/2 = 1.

If a ≡ 3 (mod 4), then b ≡ 2 [4] :

a = 4A + 3, b = 4B + 2, p + 3 = a2 + b2 + 3 = (4A + 3)2 + (4B + 2)2 + 3 ≡ 0 [8],

so (p + 3)/4 ≡ 0 [2].
In both cases (−1)((a−1)/2) ((p+3)/4) = 1, and so
2 −1)/8
χπ (π) = χπ (−2)(−1)(a .

Ex. 9.31 Let p be prime, p ≡ 1 (mod 4). Show that p = a2 + b2 where a and b are
uniquely determined by the conditions a ≡ 1 (mod 4), b ≡ −((p − 1)/2)!a (mod p).

Proof. Recall the following lemma :

Lemma : h  i2
p−1
Let p be a prime, p ≡ 1 [4], then 2 ! ≡ −1 [p].
By Wilson’s theorem (Prop. 4.1.1, Corollary), (p − 1)! ≡ −1 [p].

19
   
p−1 p+1
−1 ≡ (p − 1)! = 1.2. · · · . · · · (p − 2)(p − 1)
2 2
  
p−1 p−1
≡ 1.2. · · · − · · · (−2)(−1)
2 2
  2
(p−1)/2 p−1
≡ (−1) !
2
  2
p−1
≡ ! [p],
2

since p ≡ 1 [4].
• We show that there exists a pair a, b ∈ Z which verifies the sentence.
By lemma 5 section 7, as p ≡ 1 [4], there exists an irreducible π such that N (π) = p,
and we can choose π such that π = A + Bi is primary (lemma 7 section 7), so A is
odd.
If A ≡ 1 (mod 4), we take a = A, and if A ≡ 3 (mod 4), we take a = −A : then
a ≡ 1 (mod 4).
 
Let u = p−1
2 !. Then 0 ≡ p = A2 + B 2 (mod p), B 2 ≡ −A2 ≡ (uA)2 (mod p).
p | (B − uA)(B + uA), thus B ≡ ±uA (mod p).
Since a = ±A, B ≡ ±ua (mod p).
If B ≡ −ua (mod p), we take b = B, if not b = −B.
Then a, b are such that p = a2 + b2 , a ≡ 1 [4], b ≡ −((p − 1)/2)! a [p].

• Unicity of the pair (a, b) such that

p = a2 + b2 , a ≡ 1 [4], b ≡ −((p − 1)/2)! a [p].

Suppose that c, d are such that p = c2 + d2 , c ≡ 1 [4], d ≡ −((p − 1)/2)!c [p].

Let π = a + ib, λ = c + id. As p = N π = N λ is a rational prime, π and λ are
primes in D, and p = ππ = λλ, thus λ is associate to π or π. :

λ ∈ {π, −π, iπ, −iπ, π, −π, iπ, −iπ}.

As a, c are odd, and b, d even, it remains only the possibilities λ = ±π, λ = ±π,
thus c = ±a. Moreover a ≡ c ≡ 1 [4], thus a = c, and d ≡ −((p − 1)/2)!c ≡
−((p − 1)/2)!a ≡ b [p].
p = a2 + b2 = a2 + d2 , so d = ±b, and d ≡ b [p].
If d = −b, then p | 2b, thus p | b, and also p | a, so p2 | a2 + b2 = p: this is
impossible. So a = b, c = d. Unicity is proved.
Conclusion : if p ≡ 1 [4], there exists an unique pair a, b such that

p = a2 + b2 , a ≡ 1 (mod 4), b ≡ −((p − 1)/2)!a (mod p).

20
Ex. 9.32 Let p be a prime, p ≡ 1 (mod 4) and write p = ππ, π ∈ Z[i]. Show χp (1+i) =
i(p−1)/4 .

Proof.

χp (1 + i) = χπ (1 + i)χπ̄ (1 + i)
= χπ (1 + i)χπ (1 − i) (Prop. 9.8.3(c))
χπ (1 + i)
= = χπ (i) (since (1 − i)i = 1 + i)
χπ (1 − i)
p−1
=i 4 .
p−1
The last equality is a consequence of the definition of χπ : χπ (i) ≡ i 4 (mod π), and
the classes of 1, i, i2 , i3 modulo π are distinct.

Ex. 9.33 Let q be a positive prime, q ≡ 3 (mod 4). Show χq (1 + i) = i(q+1)/4 . [Hint :
(1 + i)q−1 ≡ −i (mod q).]

The sentence is false and must be replaced by

χq (1 + i) = (−i)(q+1)/4 = i−(q+1)/4 .

We verify this on the example q = 11 :

2 −1)/4
χq (1 + i) ≡ (1 + i)(q
≡ (1 + i)30
≡ −215 i ≡ −32i ≡ i (mod 11),

so χ11 (1 + i) = i, and i(−q−1)/4 = i−3 = i (but i(q+1)/4 = −i).

Proof. Write q = 4k + 3, k ∈ N.
As (1 + i)2 =
2i, (1 +
i)q−1 = (2i)(q−1)/2 .
2 2
2(q−1)/2 ≡ 2q [q] et 2q = (−1)(q −1)/8 = (−1)2k +3k+1 = (−1)k+1
i(q−1)/2 = i2k+1 = (−1)k i.
So
(1 + i)q−1 ≡ −i [q].
2 −1)/4
N (q) = q 2 , so χq (1 + i) ≡ (1 + i)(q = [(1 + i)q−1 ](q+1)/4 ≡ (−i)(q+1)/4 [q] :

χq (1 + i) = (−i)(q+1)/4 = i(−q−1)/4 .

Ex. 9.34 Let π = a + bi be a primary irreducible, (a, b) = 1. Show

(a) if π ≡ 1 (mod 4), then χπ (a) = i(a−1)/2 .

(b) if π ≡ 3 + 2i (mod 4), then χπ (a) = −i(−a−1)/2 .

Proof. Let π = a + bi be a primary irreducible, with a ∧ b = 1, so b 6= 0 : we can apply
the result of Exercise 9.29 :
2 −1)/8
χπ (a(−1)(p−1)/4 ) = (−1)(a .

21
 (a) Suppose that π ≡ 1 [4].
Then a ≡ 1 [4], b ≡ 0 [4], a = 4A + 1, b = 4B, A, B ∈ Z.
As χπ (−1) = (−1)(a−1)/2 ,
a−1 p−1 a2 −1
χπ (a) = (−1) 2 4 (−1) 8 ,

where
a2 −1 2
+ b4 2 +2A+4B 2
p = N π = a2 + b2 , (−1)(p−1)/4 = (−1) 4 = (−1)4A = 1,
a−1 p−1
thus (−1) 2 4 = 1.
2 −1)/8 2 +A
χπ (a) = (−1)(a = (−1)2A = (−1)A = (−1)(a−1)/4 = i(a−1)/2 .

Conclusion : if π ≡ 1 [4], χπ (a) = i(a−1)/2 .

(b) Suppose that π ≡ 3 + 2i [4].

Then a ≡ 3 [4], b ≡ 2 [4], a = 4A + 3, b = 4B + 2, A, B ∈ Z. As in (a),
a−1 p−1 a2 −1
χπ (a) = (−1) 2 4 (−1) 8 ,
a2 +b2 −1
where a2 + b2 − 1 = 16A2 + 24A + 16B 2 + 16B + 12 ≡ 4 [8], so 4 ≡ 1 [2],
2 2
thus (−1)(p−1)/4 = (−1)(a +b −1)/4 = −1.

a−1 p−1 a−1

(−1) 2 4 = (−1) 2 = (−1)2A+1 = −1,

a2 − 1 2
= 2A2 + 3A + 1, (−1)(a −1)/8 = (−1)3A+1 = (−1)A+1 = (−1)(a+1)/4 ,
8

χπ (a) = −(−1)(a+1)/4 = −i(a+1)/2 .

Moreover
a+1 −a − 1
≡ [4] ⇐⇒ a + 1 ≡ −a − 1 [8] ⇐⇒ 2a ≡ −2 [8] ⇐⇒ a ≡ 3 [4],
2 2

thus i(a+1)/2 = i(−a−1)/2 .

Conclusion : if π ≡ 3 + 2i [4], χπ (a) = −i(−a−1)/2 .

Ex. 9.35 If π = a + bi is as in Exercise 9.34 show χπ (a)χπ (1 + i) = i(3(a+b−1))/4 .

[Hint: a(1 + i) = a + b + i(a + bi). Generalize Exercises 32 and 33 to any integer ≡ 1
(mod 4) and use Proposition 9.9.8. Note a + b ≡ 1 (mod 4).]

22
Proof. We give a generalization of Exercises 9.32 and 9.33 : if n ≡ 1 [4], n 6= 1, then
χn (1 + i) = i(n−1)/4 .
By Exercises 9.32 and 9.33, we know that if p ≡ 1 [4] is a rational prime, then

χp (1 + i) = i(p−1)/4 ,

and if q ≡ 3 [4], in other words −q ≡ 1 [4], where q is a rational prime, then

χ−q (1 + i) = χq (1 + i) = i(−q−1)/4 .

Let n ∈ Z, n ≡ 1 [4], n 6= 1.
If n > 0, n = q1 q2 · · · qk p1 p2 · · · pl , where qi ≡ −1 [4], pi ≡ 1 [4], thus k is odd.
If n < 0, n = −q1 q2 · · · qk p1 p2 · · · pl , with k odd. In both cases,

n = (−q1 )(−q2 ) · · · (−qk )p1 p2 · · · pl ,

so we can write

n = s1 s2 · · · sN , where si = −qi , 1 ≤ i ≤ k, si = pi−k , k + 1 ≤ i ≤ k + l = N,

where si ≡ 1 [4], 1 ≤ i ≤ N .

χn (1 + i) = χ−q1 (1 + i) · · · χ−qk (1 + i)χp1 (1 + i) · · · χpl (1 + i)

= i(−q1 −1)/4 · · · i(−qk −1)/4 i(p1 −1)/4 · · · i(pl −1)/4
= i(s1 −1)/4 · · · i(sN −1)/4
PN si −1
=i i=1 4

= i(n−1)/4 ,

the last equality resulting of Exercise 9.44.

Conclusion : if n ∈ Z, n ≡ 1 [4], n 6= 1, then χn (1 + i) = i(n−1)/4 .

Let π = a + bi, a ∧ b = 1 a primary irreducible. As a(1 + i) = a + b + i(a + bi),

a(1 + i) ≡ a + b [π], so
χπ (a)χπ (1 + i) = χπ (a + b).
As π = a + bi is primary, a + b ≡ 1 [4].
If a + b = 1, then χπ (a)χπ (1 + i) = χπ (a + b) = 1 = i3(a+b−1)/4 . If not, the Law of
Biquadratic Reciprocity (Proposition 9.9.8) gives

χπ (a + b) = χa+b (π).

Now b ≡ −a (mod a + b), so a + bi ≡ a(1 − i) ≡ −ia(1 + i) (mod a + b). Therefore

χa+b (π) = χa+b (−1)χa+b (a)χa+b (i)χa+b (1 + i).

Since n ≡ 1 [4], χn (i) = (−1)(n−1)/4 (Prop.9.8.6), thus

n−1
χn (−1) = χn (i2 ) = (−1) 2 = 1.

Consequently, since a + b ≡ 1 [4], χa+b (−1) = 1 .

As a ∧ b = 1, (a + b) ∧ a = 1, thus χa+b (a) = 1 (Prop 9.8.5).
a + b ≡ 1 [4], thus χa+b (i) = (−1)(a+b−1)/4 (Prop. 9.8.6).

23
 From the first part of this proof, χa+b (1 + i) = i(a+b−1)/4 , so

χa+b (π) = χa+b (−1)χa+b (a)χa+b (i)χa+b (1 + i)

= (−1)(a+b−1)/4 i(a+b−1)/4
= i(a+b−1)/2 i(a+b−1)/4
= i3(a+b−1)/4

Conclusion : if π = a + bi is a primary irreducible, such that a ∧ b = 1, then

χπ (a)χπ (1 + i) = i3(a+b−1)/4

Ex. 9.36 Remove the restriction (a, b) = 1 in Exercise 9.34.

Proof. Suppose that q = a ∧ b > 1. Then a = qa0 , b = qb0 , a0 , b0 ∈ Z, so π = q(a0 + ib0 ).

As π is irreducible, and as q is not an unit, u = a0 + b0 i is an unit, and so π = uq
is associate to q : the rational integer q is then a prime in D, so a rational prime q ≡ 3
(mod 4).
If u = ±i, then π = ±q = a+bi is such that b is odd, in contradiction with π primary.
Thus u = ±1, and π = εq, ε = ±1. As π is primary, ε = −1, so π = −q.
Then χπ (a) = χ−q (−q) = 0, the result of Ex. 34 is false if b = 0.
Conclusion : if π = a + bi is a primary irreducible, and b 6= 0, then

(a) if π ≡ 1 [4], χπ (a) = i(a−1)/2 ,

(b) if π ≡ 3 + 2i [4], χπ (a) = −i(−a−1)/2 .

2
Ex. 9.37 Combine Exercises 32, 33, 34, and 35 to show χπ (1 + i) = i(a−b−b −1)/4 .
Show that this result implies Exercise 26 of Chapter 5 “the biquadratic character of 2”).

Lemma. If π = a + bi is a primary prime, then

−a+1
χπ (i) = i 2 .

Proof. (of Lemma.) Let π = a + bi a primary prime in Z[i].

• If π = −q, where q ≡ 3 (mod 4), q > 0 is a rational prime, then a = −q, b = 0. By

definition of the quartic character,
N (q)−1 q 2 −1
χq (i) = i 4 =i 4 .

Write −q = a = 4k + 1, k ∈ Z. Then

q2 − 1
= 4k 2 + 2k
4
a−1
≡ 2k = (mod 4).
2

24
 Therefore
  −a+1
q 2 −1 a−1 1 2 −a+1 −a+1 −a+1 −a+1
χ−q (i) = χq (i) = i 4 =i 2 = = (−i) 2 = (−1) 2 i 2 = i 2 ,
i
−a+1
since (−1) 2 = (−1)−2k = 1.
Suppose now that N (π) = p, where p ≡ 1 (mod 4) is a rational prime. Then
N (π)−1 p−1
χπ (i) = i 4 =i 4 .

Since π = a + bi is primary, there are two cases.

• If a ≡ 1 (mod 4), b ≡ 0 (mod 4), then a = 4A + 1, b = 4B, A, B ∈ Z.

p−1 a2 + b2 − 1
=
4 4
16A2 + 8A + 16B 2
=
4
= 4A2 + 2A + 4B 2
a−1
≡ 2A =
2
Therefore
  −a+1
p−1 a−1 1 2 −a+1 −a+1 −a+1 −a+1
χπ (i) = i 4 =i 2 = = (−i) 2 = (−1) 2 i 2 = i 2 ,
i
−a+1
since (−1) 2 = (−1)−2A = 1.

• If a ≡ 3 (mod 4), b ≡ 2 (mod 4), then a = 4A − 1, B = 4B + 2, A, B ∈ Z.

p−1 a2 + b2 − 1
=
4 4
16A − 8A + 16B 2 + 16B + 4
2
=
4
= 4A − 2A + 4B 2 4B + 1
2

−a + 1
≡ −2A + 1 = (mod 4)
2
−a+1
Therefore χπ (i) = (−1) 4 .
−a+1
The equality χπ (i) = (−1) 4 is verified for all primary primes π.

Proof. (of Ex.9.37) Let π = a + ib be a primary irreducible in Z[i].

• If b = 0, then π = a ∈ Z. As π is primary, π = −q, q ≡ 3 (mod 4), where q is a

rational prime, so a = −q, b = 0. By Ex. 9.32 (or its generalization 9.35),
2 −1)/4
χπ (1 + i) = χ−q (1 + i) = i(−q−1)/4 = i(a−b−b .

25
 • If b 6= 0, then a ∧ b = 1 (see Ex. 9.36), and by Ex. 9.35,

χπ (a)χπ (1 + i) = i3(a+b−1)/4 .

• If π ≡ 1 [4], a ≡ 1 [4], b ≡ 0 [4] : a = 4A + 1, b = 4B, A, B ∈ Z.

By Ex. 9.34(a),

χπ (a) = i(a−1)/2 , χπ (a)−1 = (−i)(a−1)/2 = i(a−1)/2 .

a+b−1
−2 a−1
χπ (1 + i) = i3 4 4

a+3b−1
=i 4

a−b−b2 −1
=i 4 ,
 
a+3b−1 a−b−b2 −1 b2
= 4B + 4B 2 ≡ 0 [4].


since 4 − 4 =b+ 4
• If π ≡ 3 + 2i [4], a ≡ 3[4], b ≡ 2 [4] : a = 4A − 1, b = 4B + 2, A, B ∈ Z.
By Ex. 9.34(b),

χπ (a) = −i(−a−1)/2 , χπ (a)−1 = −i(a+1)/2 = i(a−3)/2 ,

so
χπ (1 + i) = i(3a+3b−3+2a−6)/4 = i(5a+3b−9)/4 .
2
Now 14 [(a−b−b2 −1)−(5a+3b−9)] = 14 (−4a−4b−b2 +8) = −a−b+2− b4 =
−4A + 1 − 4B − 2 + 2 − (2B + 1)2 ≡ 0 [4],
2
thus χπ (1 + i) = i(a−b−b −1)/4 .
Conclusion : if π = a + ib is primary irreducible, then

2 −1)/4
χπ (1 + i) = i(a−b−b

Second part : the biquadratic character of 2 (see Ex. 5.25 to 5.28).

Let p ≡ 1 [4]. Then p = N (π), where π = a + bi is a primary prime.
ab
We show first that χπ (2) = i 2 .
Since 2 = i3 (1 + i)2 , the first part of the exercise, and the Lemma, give

χπ (2) = χπ (i)3 χπ (1 + i)2

3(−a+1) a−b−b2 −1
=i 2 i 2

b
= i1−a−(b+1) 2

Since π is primry, a ≡ b + 1 ≡ −b + 1 (mod 4), therefore

b b
1 − a − (b + 1) ≡ −b − (b + 1)
2 2
b
≡ (−b − 3)
2
b
≡ (−b + 1)
2
ab
≡ (mod 4),
2

26
 ab
so χπ (2) = i 2 .

Now we show that p is of the form p = A2 + 64b2 if and only if p ≡ 1 (mod 4) and if
x4 ≡ 2 has a solution x ∈ Z.
If p = A2 + 64B 2 = A2 + (8B)2 , then the prime number p is a sum of two squares,
and p 6= 2, therefore p ≡ 1 (mod 4). Since p = A2 + 64b2 , A is odd. Put b = 8B, and
a = A if A ≡ 1 (mod 4), a = −A if A ≡ −1 (mod 4). Then π = a + bi is such that
N (π) = a2 + b2 = p, and a ≡ 1, b ≡ 0 (mod 4), therefore π is a primary prime. Then
ab
χπ (2) = i 2 = i4aB = 1.

Therefore there exists α ∈ D such that 2 ≡ α4 (mod π). As D/πD is the set of classes
of 0, 1, · · · , p − 1, there exists x ∈ Z such that x ≡ α (mod π), so 2 ≡ x4 (mod π).
Then p = N (π) | N (x4 − 2) = (x4 − 2)2 , thus p | x4 − 2, in other words 2 ≡ x4
(mod p).
Conversely, suppose that p ≡ 1 (mod 4) and that 2 is a biquadratic residue modulo
p. As p ≡ 1 (mod 4), p = ππ, where π = a + bi is a primary prime. Since 2 ≡ x4
(mod p) for some x ∈ Z, then 2 ≡ x4 (mod π), so χπ (2) = 1. Moreover
ab
1 = χπ (2) = i 2 .

Since a is odd, 8 | b, therefore p = A2 + 64b2 , where A = a, B = b/8.

Conclusion :

∃(A, B) ∈ Z2 , p = A2 + 64B 2 ⇐⇒ (p ≡ 1 [4] and ∃x ∈ Z, x4 ≡ 2 [p]).

Ex. 9.38 Prove part (d) of Proposition 9.8.3.

Proposition 9.8.3(d) If π is a primary irreducible then χπ (−1) = (−1)(a−1)/2 ,

where π = a + bi.

Proof. Let π = a + bi a primary irreducible. Then a is odd, and b is even, and N (π) =
a2 + b2 . Then
N (π)−1 a2 −1 2 a+1 a−1 b2
+ b4
χπ (−1) = (−1) 4 = (−1) 4 = [(−1) 2 ]2 (−1) 4 .

By Lemma 6, section 7, a ≡ 1 [4], b ≡ 0 [4], or a ≡ 3[4], b ≡ 2[4].

a+1 b2
• If a ≡ 1 [4], b ≡ 0 [4], then (−1) 2 = −1, (−1) 4 = +1 , so
a−1
χπ (−1) = (−1) 2 .

a+1 b2
• If a ≡ 3 [4], b ≡ 2 [4], then (−1) 2 = 1, (−1) 4 = −1, so
a−1
χπ (−1) = −1 = (−1) 2 .

Conclusion : if π is a primary irreducible in Z[i], then

χπ (−1) = (−1)(a−1)/2 .

27
Ex. 9.39 Let p ≡ 1 (mod 6) and write 4p = A2 + 27B 2 , A ≡ 1 (mod 3). Put
3m


m = (p − 1)/6. Show m ≡ −1 (mod p) ⇐⇒ 2 | B.

Proof. Let p a rational prime, p ≡ 1 (mod 6). As p ≡ 1 (mod 3), we know from Theorem
2, Chapter 8, that there are integers A and B such that 4p = A2 + 27B 2 , A ≡ 1 (mod 3),
and that A is uniquely determined by these conditions.
Then A, B are of opposite parity. If we take a = A+3B
2 , b = 3B, then A = 2a − b, B =
b 2 2 2 2
3 , and 4p = (2a − b) + 3b , so p = a − ab + b . If π = a + bω, then N (π) = p. Since
A = 2a − b ≡ 1 [3], and b = 3B ≡ 0 [3], then a ≡ −1 [3], so π is a primary prime.

• Suppose that 2 | B. Since p = a2 − ab + b2 is odd, and b = 3B,

2 | B ⇐⇒ 2 | b ⇐⇒ (b ≡ 0 [2], a ≡ 1 [2]) ⇐⇒ π ≡ 1 [2].

By Proposition 9.6.1,

π ≡ 1 [2] ⇐⇒ x3 − 2 is solvable in D ⇐⇒ χπ (2) = 1.

Therefore
2 | B ⇐⇒ χπ (2) = 1.
Here χπ is of order 3, so χ2π 6= ε. By Exercise 8.6,

J(χπ , χπ ) = χπ (2)−2 J(χπ , ρ),

where ρ is the Legendre’s character.

In this case, 2 | B, χπ (2) = 1, so J(χπ , χπ ) = J(χπ , ρ), and by Lemma 1 section 4,
where p ≡ 1 [3] and p = N (π),

π = a + bω = J(χπ , χπ ) = J(χπ , ρ).

By Exercise 8.15,
N (y 2 = x3 + 1) = p + A,
and the Exercise 8.27(b) gives

N (y 2 = x3 + 1) = N (y 2 + x3 = 1) = p + 2 Re J(χπ , ρ).

thus
A = 2 Re J(χπ , ρ) = 2 Re π = 2a − b.
Moreover, since J(χπ , ρ) = π = a + bω, by Exercise 8.27(c),
 
(p − 1)/2
2a − b ≡ − .
(p − 1)/3

Therefore
       
(p − 1)/2 (p − 1)/2 (p − 1)/2 3m
−A ≡ = = = (mod p),
(p − 1)/3 (p − 1)/2 − (p − 1)/6 (p − 1)/6 m

where m = (p − 1)/6. Since A ≡ 1 (mod 3),

 
3m
≡ −1 (mod p).
m

28
 • Conversely, suppose that 3m 2a − b ≡ − 3m



m ≡ −1 (mod p). Then A = m (mod p).
Write J(χπ , ρ) = c + dω. By Exercise 8.27(c), 2c − d ≡ − 3m


m (mod p). thus

2a − b ≡ 2c − d (mod p).
√
Since |J(χπ , ρ)| = p,

4p = (2a − b)2 + 3b2 = (2c − d)2 + 3d2 ,

thus d ≡ ±b (mod p).

By Exercise 8.6,
π = J(χπ , χπ ) = χπ (2)−2 J(χπ , ρ),
Here χπ is of order 3, therefore χπ (2)−2 = χπ (2) ∈ {1, ω, ω 2 }, so

π = J(χπ , χπ ) = χπ (2)J(χπ , ρ).

If χπ (2) = ω, then a + bω = ω(c + dω) = −d + ω(c − d). Then a = −d ≡ ±b

(mod p). As a ≡ −bω (mod π), we would have −bω ≡ ±b (mod π). Here π - b, otherwise
p = N (π) | N (b) = b2 , so p | b, and p = a2 − ab + b2 , so p | a, and p2 | p, which is a
nonsense. Therefore π | ω ± 1, where π is a primary prime : it’s impossible : ω + 1 is
a unit and ω − 1 is prime, so π | ω − 1 = −λ implies that π and λ are associate, in
contradiction with N (π) = p 6= 3 = N (π).
If χπ (2) = ω 2 , then a + bω = ω 2 (c + dω) = (d − c) − ωc, so a = d − c, b = −c.
Reasoning modulo π = a + bω 2 = (a − b) + bω, where π | ππ = p, we obtain

d = a − b ≡ −bω (mod π),

where d ≡ ±b (mod π), so −bω ≡ ±b (mod π). Since N (π) = p, we obtain the same
contradiction as above.
So χπ (2) = 1, and the previously proved equivalence 2 | B ⇐⇒ χπ (2) = 1 show
that 2 | B.
Conclusion :  
(p − 1)/2
≡ −1 (mod p) ⇐⇒ 2 | B.
(p − 1)/6

Ex. 9.40 Let p ≡ 1 (mod 6), and put p = ππ where π is primary. Write π = a + bω
and show
3m


(a) If χπ (2) = ω then 2b − a ≡ − m (mod p).
3m
(b) If χπ (2) = ω 2 then a + b ≡


m (mod p).
3m


(c) If χπ (2) = ω put A = 2a − b, B = b/3. Show (A − 9B)/2 ≡ (mod p).
m

(d) If χπ (2) = ω 2 put 2a − b = A and B = −b/3. Show (A − 9B)/2 ≡ 3m

m (mod p).

(e) Show that the “normalization”” of B in (c) and (d) is equivalent to A ≡ B (mod 4).
[Recall χπ (2) ≡ π (mod 2) by cubic reciprocity.]

29
Proof. Here p = 6m + 1, m ∈ Z, and p = ππ, where π = a + bω is a primary prime.
We have proved in Exercise 39 that

π = J(χπ , χπ ) = χπ (2)J(χπ , ρ). (1)

Write J(χπ , ρ) = c + dω. The Exercise 8.27(c) shows that

 
3m
2c − d ≡ − (mod p). (2)
m

(a) If χπ (2) = ω, then (1) gives

a + bω = ω(c + dω) = −d + ω(c − d),

so a = −d, b = c − d, therefore the equality (2) gives

 
3m
2b − a = 2(c − d) + d = 2c − d ≡ − (mod p).
m

(b) If χπ (2) = ω 2 , then

a + bω = ω 2 (c + dω) = d − c − cω,
so a = d − c, b = −c, and
 
3m
a + b = d − 2c ≡ (mod p).
m

(c) Suppose that χπ (2) = ω, and put A = 2a − b, B = b/3, so

4p = A2 + 27B 2 , A ≡ 1 [3],

which shows that A, B have opposite parities. Then, by part (a),

A − 9B 2a − b − 3b
=
2 2
= a − 2b
 
3m
≡ (mod p)
m

(d) Suppose that χπ (2) = ω 2 , and put A = 2a − b, B = −b/3, so we have again

4p = A2 + 27B 2 , A ≡ 1 [3].

In this case, by part (b)

A − 9B 2a − b + 3b
=
2 2
=a+b
 
3m
≡ (mod p)
m

30
(e) The conditions 4p = A2 + 27B 2 , A ≡ 1 [3], determine A, B, except the sign of B.
So 4p = A2 + 27B 2 = (2a − b)2 + 3b2 , implies A = 2a − b and B = ± 3b .
By Exercise 39, since A, B have same parity, the condition A, B odd is equivalent
to χπ (2) ∈ {ω, ω 2 }. We choose this sign of B so that
 
A − 9B 3m
≡ (mod p).
2 m

By parts (d) and (e), where A, B are odd, this choice is given by B = b/3 if
χπ (2) = ω, and B = −b/3 if χπ (2) = ω 2 . We show that these conditions are
equivalent to A ≡ B (mod 4).
• If χπ (2) = ω, then A = 2a − b, B = b/3.
By cubic reciprocity, χπ (2) ≡ π (mod 2) (see section 6). Here χπ (2) = ω, so
ω ≡ a + bω (mod 2), therefore a ≡ 0 (mod 2), b ≡ 1 (mod 2),

b
A = 2a − b ≡ −b ≡ =B (mod 4),
3

so A ≡ B (mod 4).
• If χπ (2) = ω 2 , then A = 2a − b, B = −b/3. In this case,

ω 2 = −1 − ω ≡ a + bω (mod 2),

therefore a ≡ 1 ≡ b (mod 2), and

b
A = 2a − b ≡ 2 − b ≡ b ≡ − = B (mod 4).
3
In both cases, the choice of the sign of B implies that A ≡ B (mod 4).

Conversely, suppose that A ≡ B (mod 4). Write B = ε 3b , where ε = ±1. Then

A ≡ B (mod 4) gives
b
2a − b ≡ ε ≡ −εb (mod 4),
3
1−ε
thus a ≡ 2 b (mod 2). Then

χπ (2) ≡ π = a + bω
 
1−ε
≡b +ω (mod 2)
2

1−ε
If χπ (2) = ω, since b = 3B is odd, 2 ≡ 0 (mod 2), therefore ε = 1, and B = 3b .
1−ε
If χπ (2) = ω 2 = −1 − ω, 2 ≡ 1 (mod 2), therefore ε = −1, and B = − 3b .
The normalisation given in parts (c) and (d) for the choice of the sign of B is
equivalent to A ≡ B (mod 4) (where A, B are odd).

31
Ex. 9.41 Let p ≡ 1 (mod 6), 4p = A2 + 27B 2 , A ≡ 1 (mod 3), A and B odd. Put
π = a + bω, 2a − b = A, b = 3B. Let χπ be the cubic residue character.
(a) If χπ (2) = ω show N (x3 + 2y 3 = 1) = p + 1 + 2b − a ≡ 0 (mod 2).

(b) If χπ (2) = ω 2 show N (x3 + 2y 3 = 1) = p + 1 − a − b ≡ 0 (mod 2).

(c) Show that if A ≡ B (mod 4), then assuming χπ (2) 6= 1, one has χπ (2) = ω.

(d) If χπ (2) 6= 1, A ≡ B (mod 4) then

2(p−1)/3 ≡ (−A − 3B)/6B ≡ (A + 9B)/(A − 9B) (mod π).

(This generalization of Euler’s criterion is due to E.Lehmer [174]. See also K.Williams
[243].)
Proof. With the help of Theorem 1, Chapter 8, we obtain, writing χπ (2) = ω k ,
X
N (x3 + 2y 3 = 1) = N (x3 = a)N (y 3 = b)
a+2b=1
2
! 2

X X X
= χiπ (a)  χjπ (b)
a+2b=1 i=0 j=0
2 X
X 2 X
= χiπ (a)χjπ (b)
i=0 j=0 a+2b=1
2 X
X 2 X
= χiπ (a)χjπ (2−1 b0 )
i=0 j=0 a+b0 =1
2 X
X 2
= χπ (2)−j J(χiπ , χjπ )
i=0 j=0
2 X
X 2
= ω −kj J(χiπ , χjπ )
i=0 j=0

= p + ω −k J(χ2π , χπ ) + ω −2k J(χπ , χ2π )

+ ω −k J(χπ , χπ ) + ω −2k J(χ2π , χ2π )
= p − ω −k χπ (−1) − ω −2k χπ (−1)2 + 2 Re (ω −k J(χπ , χπ )
= p − ω −k − ω −2k + 2 Re (ω −k J(χπ , χπ )).

(a) If χπ (2) = ω, then k = 1. Using χ2π = χ−1

π = χπ , we obtain

N (x3 + 2y 3 = 1) = p + 1 + 2 Re (ω 2 J(χπ , χπ ))
= p + 1 + 2 Re (ω 2 π),

since J(χπ , χπ ) = π (Lemma 1, section 4).

ω 2 π = ω 2 (a + bω) = b − a − ωa,
2 Re (ω 2 π) = (b − a − ωa) + (b − a − ω 2 a) = 2b − 2a + a = 2b − a,

therefore
N (x3 + 2y 3 = 1) = p + 1 + 2b − a (if χπ (2) = ω).

32
 Since in then case χπ (2) = ω, then a ≡ 0 (mod 2) (see Ex. 39, part (e)), so
p + 1 + 2b − a ≡ 0 (mod 2).
(b) If χπ (2) = ω 2 = ω −1 , then k = −1, and

N (x3 + 2y 3 = 1) = p + 1 + 2 Re (ωπ),

with

ωπ = ω(a + bω) = −b + (a − b)ω,

2 Re (ωπ) = (−b + (a − b)ω) + (−b + (a − b)ω 2 ) = −2b − (a − b) = −a − b,

therefore
N (x3 + 2y 3 = 1) = p + 1 − a − b (if χπ (2) = ω 2 ).

(c) Suppose that A ≡ B (mod 4), and χπ (2) 6= 1. By hypothesis, b = 3B, and this
implies by Exercise 40 (e) that χπ (2) = ω (if not, χπ (2) = ω2 , and A ≡ B (mod 4)
gives B = −b/3).

(d) Suppose that χπ (2) 6= 1, A ≡ B (mod 4). By part (c), χπ (2) = ω.

A+3B
Since 2a − b = A, B = b/3, then a = 2 ,b = 3B.
Starting from a + bω ≡ 0 (mod π), we obtain

A + 3B
3Bω ≡ − (mod π).
2
Since pa = a2 −ab+b2 , a is relatively prime with p, therefore π ∧b = 1, so π ∧B = 1,
and π ∧ 6 = 1, since p ≡ 1 (mod 6), thus

−A − 3B
χπ (2) = ω ≡ (mod π),
6B
where we must read in this fraction the product of A + 3B by the inverse modulo
p of 6B. By definition, using N (π) = p,
p−1
χπ (2) ≡ 2 3 (mod π),

so
p−1 −A − 3B
2 3 ≡ (mod π).
6B
Moreover, since 4p = A2 + 27B 2 , A2 + 27B 2 ≡ 0 (mod p), therefore

6B(A + 9B) + (A + 3B)(A − 9B) ≡ 0 (mod p).

If p | A − 9B, since p - 6B, this equality implies that p | A + 9B, therefore

p | (A − 9B) + (A + 9B) = 2A, which is false. Therefore A − 9B 6≡ 0 (mod p), and
p−1 −A − 3B A + 9B
2 3 ≡ ≡ (mod π).
6B A − 9B

33
 p−1 p−1
Note : By a usual argument, if h ∈ Z, 2 3 ≡ h (mod π) ⇐⇒ 2 3 ≡ h (mod p).
Note that the hypothesis χπ (2) 6= 1 means that 2 is not a cubic residue modulo p, which
is equivalent to A, B odd by Exercise 39. We can conclude

Suppose that p ≡ 1 (mod 6), and let (A, B) be the unique solution of 4p = A2 + 27B 2
such that A ≡ 1 (mod 3), and B ≡ A (mod 4) if B odd, and B > 0 otherwise.
p−1
If B is even, then 2 is a cubic residue modulo p, and 2 3 = 1.
If B is odd, then 2 is not a cubic residue modulo p, and B satisfies B ≡ A (mod 4).
Writing a = A+3B2 , b = 3B, and π = a + bω, then χπ (2) = ω, and

p−1 A + 9B
2 3 ≡ (mod p).
A − 9B

The three roots of x3 − 1 in Fp are 1, A+9B A−9B

A−9B , A+9B . Here 2 is not a cubic residue
p−1 p−1
A±9B
modulo p, and 2 3 is also a cubic root of unity modulo p, so 2 3 ≡ A∓9B (mod p).
p−1
A+9B
The proposition explicits the choice of the sign of B which gives 2 3 ≡ A−9B (mod p).

Numerical example : Let p be the prime 967. If we decompose p on the form p = ππ,
we obtain π = a + bω = −34 − 27ω. To obtain these result without tries, I find k = 682
such that p | k 2 +3 with the Tonelli-Shanks algorithm, and I compute gcd(p, k +1+2ω) =
a + bω, where a + bω is primary, with a small Python program using the class of elements
in Z[ω] and the Euclid algorithm in Z[ω]. This gives the decompositions

967 = p = a2 − ab + b2 = 342 − 34 × 27 + 272 ,

and
3868 = 4p = (2a − b)2 + 3b2 = A2 + 27B 2 = 412 + 27 × 92 ,
where A ≡ 1 (mod 3), and I choose the sign of B such that B ≡ A (mod 4). We obtain
A = −41, B = −9, and a, b must verify A = 2a − b, B = b/3.
Then χπ (2) = ω, where π = −41 − 9ω. In F967 , the cubic roots of unity modulo p
are 1, 142, 824: 1423 ≡ 8243 ≡ 1 (mod 967).
Here (A + 9B)(A − 9B)−1 = 142, and we verify with a fast exponentiation that
p−1
2 3 = 2322 ≡ 142 (mod 967).
I give here an extract of a table obtained with this program, which for each p gives
A, B such that 4p = A2 + 27B 2 , A ≡ 1 (mod 3) and such that B ≡ A (mod 4) if A, B
odd, and π = a + bω satisfies χπ (2) = ω (or χπ (2) = 1 if A, B even, which corresponds

34
to the case a ≡ 1, b ≡ 0 (mod 2)).
p−1
p A B π = a + bω a%2 b%2 2 3 % p A−9B
A+9B χπ (2)
787 31 −9 2 − 27ω 0 1 379 379 ω
811 −56 −2 −31 − 6ω 1 0 1 130 1
823 −5 11 14 + 33ω 0 1 648 648 ω
829 7 11 20 + 33ω 0 1 125 125 ω
853 −35 9 −4 + 27ω 0 1 632 632 ω
859 13 −11 −10 − 33ω 0 1 260 260 ω
877 −59 1 −28 + 3ω 0 1 594 594 ω
883 −47 −7 −34 − 21ω 0 1 545 545 ω
907 19 11 26 + 33ω 0 1 384 384 ω
919 52 −6 17 − 18ω 1 0 1 52 1
937 61 1 32 + 3ω 0 1 614 614 ω
967 −41 −9 −34 − 27ω 0 1 142 142 ω
991 61 −3 26 − 9ω 0 1 113 113 ω
997 10 −12 −13 − 36ω 1 0 1 692 1
p−1
As a verification I compute χπ (2) with a fast exponentiation in Z[ω] : χπ (2) = 2 3
(mod π).
We obtain the primary prime µ such that N (µ) = p, χµ (2) = ω 2 by taking the
conjugate of π. For instance, with p = 787, π = 2 − 27ω satisfies χπ (2) = ω, therefore
χπ (2) = χ29+27ω = ω 2 .
The lines where χπ (2) = 1, corresponding to the case where A, B are even (or equiv-
alently a odd, b even), give the decomposition p = x2 + 27y 2 , (x = A/2, y = B/2). For
instance 997 = 52 + 27 × 62 . If p is prime,

∃x ∈ Z, ∃y ∈ Z, p = x2 + 27y 2 ⇐⇒ p ≡ 1 (mod 3) and ∃a ∈ Z, 2 ≡ a3 (mod p).

Ex. 9.42 The notation being as in Section 12 show that the minimal polynomial of
g(χπ ) is x3 − 3px − Ap.
Note : we must read “the minimal polynomial of G = g(χπ )+g(χπ ) is x3 −3px−Ap”.

Proof. Write f (x) = 3i=0 ai xi = x3 − 3px − Ap.

P
Then a3 = 1, p | a0 = Ap, p | a1 = −3p, p | a2 = 0.
Moreover, since 4p = A2 + 27B 2 , p - A, therefore p2 - a0 .
The Eisenstein’s Irreducibility Criterion (Ex. 6.23) shows that f (x) is irreducible
over Q. By section 12, G is a root of f , so f is the minimal polynomial of G.

Ex. 9.43 Find the local maxima and minima of x3 − 3px − Ap and show that each
√ √ √ √ √ √
of the intervals (−2 p, − p), (− p, p), ( p, 2 p) contains exactly one of the values
2Re(ω k g(χπ )), k = 0, 1, 2.

Proof. Write χ = χπ , and for k ∈ {0, 1, 2},

Gk = 2Re(ω k g(χ)) = ω k g(χ) + ω k g(χ),

35
so G = G0 . As in section 12, since g(χ)3 = pπ, and |g(χ)|2 = p,
3 2
G3k = g(χ)3 + g(χ) + 3ω 2k g(χ)2 g(χ) + 3ω k g(χ)ω 2k g(χ)
= pπ + pπ + 3g(χ)g(χ)(ω k g(χ) + ω k g(χ)
= 3pGk + p(2a − b)
= 3pGk + pA

So G0 , G1 , G2 are the three roots of f (x) = x3 − 3px − Ap.

√ √ √ √
f 0 (x) = 3(x2 −p) < 0 iff − p < x < p. f is decreasing on [− p, p], and increasing
√ √
on ] − ∞, − p[, and on [ p, +∞[.
√
Since 4p = A2 + 27B 2 , |A| < 2 p, therefore
√ √ √
f ( p) = p p − 3p p − Ap
√
= −p(2 p + A) < 0,

and
√ √ √
f (− p) = −p p + 3p p − Ap
√
= p(2 p − A) > 0.

Since lim f (x) = −∞ and lim f (x) = +∞, the intermediate value theorem
x→−∞ x→+∞
√ √ √ √
shows that f has a unique root in each of the intervals ]−∞, − p[, ]− p, p[, [ p, +∞[.
Moreover
√ √ √ √
f (2 p) = 8p p − 6p p − Ap = p(2 p − A) > 0,
√ √ √ √
f (−2 p) = −8p p + 6p p − Ap = p(−2 p − A) < 0,
√ √ √ √ √ √
therefore f has a unique root in each of the intervals ]−2 p, − p[, ]− p, p[, [ p, 2 p[.

Pt 9.44 Let n ∈ Z, n = s1 · · · st , n ≡ 1 (mod 4), i = 1, . . . , t. Show (n − 1)/4 ≡

Ex.
i=1 (si − 1)/4 (mod 4).

Proof. If n = st, s ≡ 1, t ≡ 1 [4], then s = 4k + 1, t = 4l + 1, k, t ∈ Z, so

n−1 s−1 l−1
n = (4k + 1)(4l + 1) = 16kl + 4k + 4l + 1, = 4kl + k + l ≡ k + l = + [4].
4 4 4
Reasoning by induction on t, suppose that every product of t factors n = s1 s2 · · · st ,
where si ≡ 1 [4] verifies
t
n − 1 X si − 1
≡ [4].
4 4
i=1

If n0 = s1 s2 · · · st st+1 = nst+1 , si ≡ 1[4], then n ≡ 1, st+1 ≡ 1 [4], so

t t+1
n0 − 1 n − 1 st+1 − 1 X si − 1 st+1 − 1 X si − 1
≡ + ≡ + ≡ [4].
4 4 4 4 4 4
i=1 i=1

t
n−1 P si −1
Conclusion : if n = s1 s2 · · · st , si ≡ 1[4], alors 4 ≡ 4 [4].
i=1

36
Ex. 9.45 Let π = a + bi ∈ Z[i] and q ≡ 3 [4] a rational prime. Show π q ≡ π [4].

Proof. Let
q

π = a + bi ∈ Z[i], and q ≡ 3 [4] a rational prime.
As k ≡ 0 (mod q) for 1 ≤ k ≤ q − 1, the Fermat’s Little Theorem gives

π q = (a + bi)q
≡ aq + bq iq [q]
≡ a + bi3 [q]
= a − bi
=π

Conclusion : π q ≡ π̄ [q] (π ∈ Z[i], and q ≡ 3 [4])

37