Here, the conjugate pairs ϕi and ϕ̄i are conjugate with respect to ρ1 = (1, 0).

While f (ϕi )
and f (ϕ̄i ) are conjugate with respect to ρ2 = (0, 1).
We write f (Φ) to denote {f (ϕ1 ), f (ϕ2 )}. Then note that

Φσ = Φ ⇐⇒ f (Φσ) = f (Φ) ⇐⇒ f (Φ) ◦ f (σ) = f (Φ).

This shows that f (σ) fixes f (Φ) precisely when σ fixes Φ. This leads to the following:

Φ is induced by (K H , Φ′ ) =⇒ f (Φ) is induced by K f (H) .

This shows that f (Φ) is a primitive CM-type of the same field as the primitive CM-type Φ
up to relabeling the elements in Gal(K/Q). This idea is formally proven in Section 2.

1.3.1 Reflex fields

Let K be a CM-field and let Φ be a CM-type of K. Let ΦK cl denote a CM-type of K cl
induced by Φ. Since K cl is Galois over Q, the embeddings in the CM-type ΦK cl are each
automorphisms of K cl and thus each have inverses (Theorem 1.2.9). Denote the CM-type
containing this set of inverses with Φ−1 K cl
. The unique primitive CM-pair (K r , Φr ) that
−1
induces ΦK cl is called the reflex pair of (K, Φ).
Definition 1.3.20. Let K be a CM field with CM-type Φ. Then the reflex pair (K r , Φr )
is the unique primitive CM-pair that induces Φ−1
K cl
as defined above.
We can also take the reflex of the reflex.
Lemma 1.3.21. Let (K, Φ) be a CM-type, we have that K rr ⊂ K and that Φrr induces Φ.
If Φ is primitive, then K rr = K and Φrr = Φ.

Proof. Note that the extension of Φr to K cl is given by Φ−1 K cl

. Then taking the inverse
embeddings gives us ΦK cl . By definition, Φrr is now given by the primitive CM-type that
induces ΦK cl . This shows that Φrr also induces Φ and thus K rr ⊂ K.
If Φ is primitive, the only CM-type inducing Φ is given by Φ itself. In this case, we may
conclude Φrr = Φ and K rr = K.

Theorem 1.3.22. Let (K, Φ) be a CM-pair. The reflex field K r is the fixed field of

Gal(K cl /K r ) = {σ ∈ Gal(K cl /Q) | σΦK cl = ΦK cl }.

Proof. By definition of the reflex field, we know that K r is given by the unique primitive
pair (K r , Φr ) that induces Φ−1 K cl
. By Theorem 1.3.15, this field is fixed by τ ∈ Gal(K cl /Q)
such that ΦK cl τ = ΦK cl ; i.e. we have {ϕ−1
−1 −1 −1 −1 −1 −1 −1
1 ◦ τ, ϕ2 ◦ τ, . . . , ϕn ◦ τ } = {ϕ1 , ϕ2 , . . . , ϕn }.
−1 −1 −1 −1 −1 −1 −1 −1 −1
Write τ = (τ ) so that ϕi ◦ τ = ϕi ◦ (τ ) = (τ ◦ ϕ) . Setting σ = τ yields
that {(σ ◦ ϕ1 )−1 , . . . , (σ ◦ ϕn )−1 } = {(ϕ1 )−1 , . . . , (ϕn )−1 }. Lastly, make the observation
that (σϕi )−1 = (ϕj )−1 ⇐⇒ σϕi = ϕj which finishes the results as we may get rid of the
inverses.

14
Remark 1.3.23. If K is a Galois CM-field over Q, then Theorem 1.3.22 shows that the
reflex field of K is given by the fixed field of

Gal(K/K r ) = {σ ∈ Gal(K/Q) | σΦ = Φ}.

Corollary 1.3.24. Reflex fields are CM-fields.

Proof. From Theorem 1.3.22, we know that the reflex field K r of (K, Φ) is given by the
fixed field of σ ∈ Gal(K cl /Q) such that σΦK cl = ΦK cl . Note that ρ cannot fix ΦK cl as it
conjugates all embeddings, thus K r is not fixed by complex conjugation and hence cannot
be a totally real field. By Theorem 1.3.3, we conclude that K r is a CM-field.

1.3.2 Equivalent CM-types

Recall that two CM-types Φ1 and Φ2 of K are equivalent if there exists an automorphism
of K such that Φ1 σ = Φ2 . We will show that equivalent CM-types form a partition of the
set of CM-types of K.
Lemma 1.3.25. Let K be a CM-field that is Galois over Q with G := Gal(K/Q) and CM-
types X = {Φi }. Then G acts on X from both the left and the right via a1 , a2 : X ×G → X
defined as a1 (Φ, σ) = Φσ and a2 (Φ, σ) = σΦ respectively.

Proof. From Proposition 1.3.8, note that a1 (Φ, σ) ∈ X and a2 (Φ, σ) ∈ X. Furthermore, it
can be checked that a1 and a2 satisfy the definition of an action.

Lemma 1.3.26. Let K be a Galois CM-field and Φ a CM-type. The equivalent CM-types
are given by the orbit of Φ of the right action as defined in Lemma 1.3.25.

Proof. The equivalent CM-types of Φ are given by {Φσ | σ ∈ Gal(K/Q)}. Note that this
is the definition of the orbit of action a1 as defined in Lemma 1.3.25.

Remark 1.3.27. Since G defines an action on the set of CM-types X = {Φi }, we have
that the orbits give a partition on the set of CM-types. Thus being equivalent CM-types
defines an equivalence relation.
Theorem 1.3.28. Let K be a Galois CM-field with a CM-type Φ and |Gal(K/Q)| = 2n.
Then Φ is primitive if and only if Φ has 2n equivalent CM-types.

Proof. The CM-type Φ has 2n equivalent CM-types precisely when |Orb(Φ)| = 2n. By the
orbit stabilizer theorem, we have Stab(Φ) = |Gal(K/Q)|/|Orb(Φ)| = 1. Thus there is only
one element σ ∈ Gal(K/Q) such that Φσ = Φ, hence Theorem 1.3.15 tells us that Φ is
primitive.

Remark 1.3.29. Note that the above definitions and proofs also hold when K is not
Galois, but then we have to replace Φ by a CM-type ΦK cl where ΦK cl is induced by Φ.

15
Theorem 1.3.30. Let K be a CM-field that is Galois over Q with CM-types Φ1 and Φ2 ,
which are induced by (K1′ , Φ′1 ) and (K2′ , Φ′2 ) respectively. If Φ1 and Φ2 are equivalent,
then K1′ = K2′ .

Proof. Let Φ1 and Φ2 be equivalent CM-types and write Φ1 τ = Φ2 for some τ ∈ Aut(K).
By Theorem 1.3.15, we know that K1′ is the fixed field of {σ ∈ Gal(K cl /Q) | Φ1 σ = Φ1 }.
In particular, we have

{σ ∈ Gal(K cl /Q) | Φ1 σ = Φ1 } = {σ ∈ Gal(K cl /Q) | Φ1 στ = Φ1 τ }

= {σ ∈ Gal(K cl /Q) | Φ1 τ σ = Φ1 τ }
= {σ ∈ Gal(K cl /Q) | Φ2 σ = Φ2 }.

The second equality holds since {στ | σ ∈ Gal(K cl /Q)} = {τ σ | σ ∈ Gal(K cl /Q)}. This
shows that K1′ = K2′ as the fields are fixed by the same automorphisms.

Corollary 1.3.31. Let Φ and Φ′ be equivalent CM-types, then Φ is primitive if and only
if Φ′ is primitive.
Theorem 1.3.32. Let K be a CM-field with reflex pairs (K1r , Φ1 ) and (K2r , Φ2 ). If Φ1
and Φ2 are equivalent, then K1r = K2r .

Proof. Let Φ1 and Φ2 be equivalent CM-types and write Φ1 τ = Φ2 for some τ ∈ Aut(K).
By Theorem 1.3.22, we know that K1r is the fixed field of {σ ∈ Gal(K cl /Q) | σΦ1 = Φ1 }.

{σ ∈ Gal(K cl /Q) | σΦ1 = Φ1 } = {σ ∈ Gal(K cl /Q) | σΦ1 τ = Φ1 τ }

= {σ ∈ Gal(K cl /Q) | σΦ2 = Φ2 }.

This shows that K1r and K2r are fixed by the same automorphisms and thus K1r = K2r .

1.3.3 Reflex fields for abelian Galois groups

In the case that K is a Galois CM-field with abelian Galois group, we can relate the reflex
fields with the primitivity of a CM-type Φ. This will be highlighted in this section.
Theorem 1.3.33. Let K be a CM-field that is Galois over Q with abelian Galois group.
Assume that (K, Φ) is a CM-pair induced by (K ′ , Φ′ ) with reflex field K r . Then K ′ = K r .

Proof. From Theorems 1.3.15 and 1.3.22, we have

Gal(K/K ′ ) = {σ ∈ Gal(K/Q) | Φσ = Φ} = {σ ∈ Gal | σΦ = Φ} = (K/Q)} = Gal(K/K r ).

Here we use that σΦ = Φσ since Gal(K/Q) is abelian. Thus we conclude K r = K ′ .

Corollary 1.3.34. Let K be a CM-field that is Galois over Q with abelian Galois group,
together with primitive CM-type Φ which has reflex K r . Then K r = K.

16
2 Equivalent actions and ρ-structures
Our aim is to classify all primitive CM-types and reflex fields of Galois CM-fields of degree 8.
For this, we take the same course of action as in the example of the degree 4 Galois CM-field
in Example 1.3.19.
1. Fix a group G of order 8;
2. Fix an element of order 2 in the center of G to represent complex conjugation;
3. Determine the primitive CM-types and reflex fields using Theorems 1.3.15 and 1.3.22.
This process will be repeated over all possible groups of order 8 and all possible choices of
complex conjugation.
Theorem 2.0.1. There are five groups of order 8. The groups together with possible
representatives for ρ are given below.
• D4 = ⟨a, b | a4 = b2 = e, ab = ba−1 ⟩ where ρ = a2 ;
• Q8 = ⟨i, j, k | i2 = j 2 = ijk = −1⟩ where ρ = −1;
• Z8 where ρ = 4̄;
• Z2 × Z4 where ρ ∈ {(1, 0), (0, 2), (1, 2)};
• (Z2 )3 where ρ ∈ {(1, 0, 0), (0, 1, 0), (0, 0, 1), (1, 1, 0), (1, 0, 1), (0, 1, 1), (1, 1, 1)}.
As one can see, the analysis for the groups D4 , Q8 and Z8 can be done quick as there is a
unique element of order 2 in the center of the group. However, the groups Z2 ×Z4 and (Z2 )3
require more work. To reduce the casework, we introduce the notion of ρ-equivalence.
As explained in Example 1.3.19, ρ-equivalence encapsulates the idea that if 2 elements
of order 2 in the center of the group can be interchanged using a group automorphism,
then they must lead to the same primitive CM-types and reflex fields. This idea can be
formalized via the notion of equivalent orbits.
Definition 2.0.2. Let G1 ∼ = G2 via isomorphism f . Moreover, let X1 and X2 be sets
with bijection ψ : X1 → X2 . Assume that G1 , G2 act on X1 , X2 via a1 , a2 respectively.
If ψ(a1 (g, x)) = a2 (f (g), ψ(x)), then the a1 and a2 are called equivalent actions.
Equivalent actions have, up to relabeling the elements by automorphism f and bijection ψ,
the same orbits and stabilizer.
Theorem 2.0.3. Let G1 ∼ = G2 be groups acting on X1 , X2 via equivalent actions a1 , a2
respectively. Let x0 ∈ X1 be arbitrary. Then Orb(ψ(x0 )) = {ψ(x) | x ∈ Orb(x0 )}.

Proof. The orbit of ψ(x0 ) ∈ X2 is given by Orb(ψ(x0 )) = {a2 (g, ψ(x0 )) | g ∈ G2 }. Since
we have G1 ∼ = G2 via isomorphism f , we may write each g ∈ G2 as f (g ′ ) for some
element g ∈ G1 . Thus we find Orb(ψ(x0 )) = {a2 (f (g ′ ), ψ(x0 )) | g ′ ∈ G2 }. Using the
′

definition of equivalent orbits, we get Orb(ψ(x0 )) = {ψ(a1 (g ′ , x0 ) | g ′ ∈ G2 }, which shows

that Orb(ψ(x0 )) = {ψ(x) | x ∈ Orb(x0 )}.

17
Theorem 2.0.4. Let G1 ∼ = G2 be groups acting on X1 , X2 via equivalent actions a1 , a2
respectively. Let x0 ∈ X1 be arbitrary, then Stab(ψ(x0 )) = {f (g) | g ∈ Stab(x0 )}.

Proof. The stabilizer of ψ(x0 ) is given by Stab(ψ(x0 )) = {g ∈ G2 | a2 (g, ψ(x0 )) = ψ(x0 )}.
Since G1 ∼
= G2 via isomorphism ϕ, we have that for each g ∈ G2 , we may assign g ′ ∈ G1
such that f (g ′ ) = g. Thus we get Stab(ψ(x0 )) = {f (g ′ ) ∈ G2 | a2 (f (g ′ ), ψ(x0 )) = ψ(x0 )} =
{g ′ ∈ G1 | ψ(a1 (g ′ , x0 )) = ψ(x0 )} where the last equality follows from the definition of
equivalent actions. Lastly, note that we may get rid of ψ as it is a bijection to find
that Stab(ψ(x0 )) = {f (g ′ ) | g ′ ∈ Stab(x0 )} as desired.

Remark 2.0.5. Note that neither the definition of equivalent orbits nor the two results
following rely on specifying whether a1 and a2 are left or right orbits.
Definition 2.0.6. Let G be a group and ρ1 , ρ2 be elements of order 2 in the center
of G. The elements ρ1 and ρ2 are said to be ρ-equivalent if there exists f ∈ Aut(G) such
that f (ρ1 ) = ρ2 .
Equivalence classes formed under being ρ-equivalence are called a ρ-structures.
Theorem 2.0.7. Let G1 ∼ = G2 be Galois groups of Galois CM-fields K1 and K2 respec-
tively. Let ρ1 ∈ G1 and ρ2 ∈ G2 represent complex conjugation and assume f : G1 → G2 is
an isomorphism such that f (ρ1 ) = ρ2 . Since K1 is Galois, the embeddings in Φ are elements
in the Galois group G1 (Theorem 1.2.9). Write Φ = {g1 , g2 , . . . , gn } where gi ∈ G1 , then

f (Φ) := {f (g1 ), f (g2 ), . . . , f (gn )},

is a CM-type of the field K2 .

Proof. Note that Φ is a CM-type of K1 so that no two embeddings differ by complex

conjugation; i.e. we have gi ̸= ρ1 gj for all gi , gj ∈ Φ. Then f (Φ) = {f (g1 ), f (g2 ), . . . , f (gn )}
is a set of n embeddings of K2 . Furthermore, we have

gi ̸= ρ1 gj =⇒ f (gi ) ̸= f (ρ1 gj ) =⇒ f (gi ) ̸= ρ2 f (gj ),

so that no two embeddings of K2 differ by complex conjugation. This shows that f (Φ) is
a CM-type of K2 .

Corollary 2.0.8. With the same notation as in Theorem 2.0.7, we additionally write {Φi }
to be the collection of the 2n distinct CM-types of K1 . Then {f (Φi )} is the collection of 2n
distinct CM-types of K2 .

Proof. This follows from Theorem 2.0.7 along with the fact that f is an isomorphism.

Next, we show that equivalent CM-types give equivalent orbits when groups G1 , G2 are
taken to be the Galois groups of fields K1 , K2 and X1 , X2 are taken to be the set of CM-
types.

18
Theorem 2.0.9. Let G1 ∼ = G2 be Galois groups of CM-fields K1 and K2 respectively.
Let ρ1 ∈ G1 and ρ2 ∈ G2 represent complex conjugation and assume f : G1 → G2 is an
isomorphism such that f (ρ1 ) = ρ2 . Set X1 = {Φi } and X2 = {f (Φi )}. Define the left action
as a1 : G1 × X1 → X1 as a1 (σ, Φ) = σΦ and a2 : G2 × X2 → X2 as a2 (τ, f (Φ)) = τ (f (Φ)).
Then a1 and a2 are equivalent actions.

Proof. We note that f (a1 (g, Φ)) = f (g◦Φ) = f (g)◦f (Φ) = a2 (f (g), f (Φ)) so that a1 and a2
are equivalent actions by defining ψ : X1 → X2 as ψ(Φ) = f (Φ), where Theorem 2.0.7
guarantees that this is a bijection of sets.
Corollary 2.0.10. With notation as in Theorem 2.0.9, defining a1 , a2 as right actions
instead according to a1 (σ, Φ} = Φσ and a2 (τ, f (Φ)) = f (Φ)τ . Then a1 and a2 are also
equivalent actions.
By Theorems 1.3.15 and 1.3.22, primitivity and reflex fields of Φ are determined by the
fixed points of the right and left action of G on Φ respectively. Thus we can use the
newly obtained results on equivalence of these action of G on Φ to relate the CM-types,
primitivity, and reflex fields under choices of complex conjugation in the same ρ-structure.
Theorem 2.0.11. Let K1 , K2 be two Galois CM-fields with G ∼ = Gal(K1 /Q) ∼
= Gal(K2 /Q).
Assume that ρ1 ∈ G represents complex conjugation for K1 and ρ2 ∈ G represent complex
conjugation for K2 . If ρ1 and ρ2 are ρ-equivalent via an automorphism f of G, then the
following holds.
1. Let K1′ and K2′ be subfields of K1 , K2 respectively, with f (Gal(K/K1′ )) = Gal(K/K2′ ).
Then K1′ is a CM-subfield if and only if K2′ is a CM-subfield;
2. If {Φi } is the collection of CM-types for K1 , then {f (Φi )} is the collection of CM-
types for K2 ;
3. Let (K1 , Φ) be induced by primitive pair (K1H1 , Φ′ ), then (K2 , f (Φ)) is induced by
f (H )
primitive pair (K2 1 , f (Φ′ ));
4. If (K1 , Φ) has the reflex pair (K1r , Φr ) = (K1H1 , Φr ), then (K2 , f (Φ)) has the reflex
f (H )
pair (K2 1 , f (Φr )).

Proof. (1) Let K1′ and K2′ be subfields of K1 , K2 respectively, so that f (Gal(K/K1′ )) =
Gal(K/K2′ ). Recall that a subfield of a CM-field is either totally real or a CM-field so that
subfield K1′ is a CM-field ⇐⇒ ρ1 ∈ / Gal(K/K1′ ) ⇐⇒ f (ρ1 ) ∈ / f (Gal(K/K1′ )) ⇐⇒ ρ2 ∈ /
′ ′
Gal(K/K2 ) ⇐⇒ K2 is a CM-field.
(2) This is Theorem 2.0.8.
(3) We know (K1 , Φ) is induced by K1H1 where H1 = {σ ∈ G | Φσ = Φ}, thus H1 = Stab(Φ).
By Theorems 2.0.9 and 2.0.4, we know that Stab(f (Φ)) = {f (g) | g ∈ Stab(Φ)} = f (H1 )
f (H )
Thus (K2 , f (Φ)) is induced from K2 1 . Lastly, since Φ′ induces Φ, we have that f (Φ′ )
induces f (Φ).
(4) Same proof as (3), using the left action instead of the right action.

19
3 The results for the order 8 groups
3.1 The group Z2 × Z4
The first group of order 8 we consider is the group Z2 × Z4 . The analysis of this group is
particularly interesting as there are three representatives for ρ, which together give two ρ-
structures. In this section, we work out the primitive CM-types and reflex fields for Z2 ×Z4
for the two ρ-structures. Moreover, examples of fields attaining both ρ-structures are given.
We start with determining the elements that can represent complex conjugation. Note
that Z2 × Z4 contains three elements of order 2 in the center, these are

ρ1 := (1, 0), ρ2 := (0, 2) and ρ3 := (1, 2).

3.1.1 The subgroup lattice

In light of giving a complete overview of the problem, we compute the subgroup lattice for
each choice of complex conjugation. For this, observe that the subgroups of Z2 × Z4 are
given by
• Order 2: ⟨(1, 0)⟩, ⟨(0, 2)⟩ and ⟨(1, 2)⟩;
• Order 4: ⟨(0, 1)⟩, ⟨(1, 1)⟩ and ⟨(1, 0), (0, 2)⟩.
Then the subgroup lattice is given by
ρ2 = (0, 2)

{id}

⟨(1, 0)⟩ ⟨(1, 2)⟩ ⟨(0, 2)⟩

⟨(1, 0), (0, 2)⟩ ⟨(1, 1)⟩ ⟨(0, 1)⟩

Z2 × Z4

By the Galois correspondence, these subgroups correspond to intermediate fields fixed by

the elements in the Galois group. In particular, subfields of CM-fields are either totally
real or CM-fields (Theorem 1.3.3). This means that the subgroups containing complex
conjugation, given by ρ2 = (0, 2) correspond to totally real fields, while the subgroups not
containing complex conjugation correspond to CM-subfields, which are indicated by a box.

20
Drawing the subgroup lattice, but choosing ρ1 = (1, 0) and ρ3 = (1, 2) for complex con-
jugation gives the following diagram, where the CM-intermediate fields are in almost the
same location.
ρ = (1, 0) ρ = (1, 2)

{id} {id}

⟨(1, 0)⟩ ⟨(1, 2)⟩ ⟨(0, 2)⟩ ⟨(1, 0)⟩ ⟨(1, 2)⟩ ⟨(0, 2)⟩

∼
=
⟨(1, 0), (0, 2)⟩ ⟨(1, 1)⟩ ⟨(0, 1)⟩ ⟨(1, 0), (0, 2)⟩ ⟨(1, 1)⟩ ⟨(0, 1)⟩

Z2 × Z4 Z2 × Z4

However, recall that the way we draw the subgroup lattice is to some extent arbitrary.
In this case, we can swap the groups in the right lattice without affecting the subgroup
structure. This interchange of subgroups puts the intermediate CM-fields in exactly the
same location in the subgroup lattice. This intuitively justifies that the CM-subfields in
both cases are in some sense equivalent. We use ρ-structures to make this formal.
Lemma 3.1.1. The group Z2 × Z4 has two ρ-structures with representatives ρ1 = (1, 0)
and ρ2 = (0, 2). The element ρ3 = (1, 2) is in the same ρ-structure as ρ1 .

Proof. Define f : Z2 × Z4 → Z2 × Z4 by f (1, 0) = (1, 2) and f (0, 1) = (0, 1). It can be

shown that f is a group automorphism such that f (ρ1 ) = ρ3 . Hence, ρ1 and ρ3 lie in the
same ρ-structure.
We claim there is no automorphism g such that g(ρ1 ) = ρ2 . Indeed, assume such an
automorphism exists, then 2g −1 (0, 1) = (1, 0), but (1, 0) cannot be written this way.

Observe that applying automorphism f to the subgroups in the subgroup lattice places the
CM-group ⟨(1, 2)⟩ into the location f (⟨(1, 2)⟩) = ⟨(1, 0)⟩. It can be checked that f fixes
all other subgroups of Z2 × Z4 . This shows that f changes the subgroups according to the
red arrow in Diagram 3.1.1. This verifies (1) in Theorem 2.0.11.

3.1.2 The CM-types in the case ρ2 = (0, 2)

Next, we compute the CM-types of fields with Galois group Z2 × Z4 for both ρ-structures
and highlight which CM-types are primitive. We will also compute the reflex fields.

21
First, fix ρ2 = (0, 2). The embeddings are given by the elements in the Galois group.
ϕ1 = (0, 0), ϕ2 = (0, 1), ϕ3 = (1, 0), ϕ4 = (1, 1),
as none of these four embeddings differ by ρ2 = (0, 2). The conjugated embeddings are
ϕ̄1 = (0, 2), ϕ̄2 = (0, 3), ϕ̄3 = (1, 2), ϕ̄4 = (1, 3).
By definition of CM-types, the 24 = 16 CM-types are given by choosing one embedding
out of each pair {ϕi , ϕ̄i } for 1 ≤ i ≤ 4.
Next, we compute which CM-types are equivalent by computing for which σ ∈ Z2 × Z4 ,
we have Φσ = Φ. The calculations are explicitly shown for Φ = {ϕ1 , ϕ2 , ϕ3 , ϕ4 }.
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (0, 0) = {(0, 0), (0, 1), (1, 0), (1, 1)} = {ϕ1 , ϕ2 , ϕ3 , ϕ4 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (0, 1) = {(0, 1), (0, 2), (1, 1), (1, 2)} = {ϕ2 , ϕ̄1 , ϕ4 , ϕ̄3 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (0, 2) = {(0, 2), (0, 3), (1, 2), (1, 3)} = {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ̄4 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (0, 3) = {(0, 3), (0, 0), (1, 3), (1, 0)} = {ϕ̄2 , ϕ1 , ϕ̄4 , ϕ3 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (1, 0) = {(1, 0), (1, 1), (0, 0), (0, 1)} = {ϕ3 , ϕ4 , ϕ1 , ϕ2 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (1, 1) = {(1, 1), (1, 2), (0, 1), (0, 2)} = {ϕ4 , ϕ̄3 , ϕ2 , ϕ̄1 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (1, 2) = {(1, 2), (1, 3), (0, 2), (0, 3)} = {ϕ̄3 , ϕ̄4 , ϕ̄1 , ϕ̄2 };
{(0, 0), (0, 1), (1, 0), (1, 1)} ◦ (1, 3) = {(1, 3), (1, 0), (0, 3), (0, 0)} = {ϕ̄4 , ϕ3 , ϕ̄2 , ϕ1 }.
And thus we find that the CM-types equivalent to {ϕ1 , ϕ2 , ϕ3 , ϕ4 } are
[{ϕ1 , ϕ2 , ϕ3 , ϕ4 }] = {{ϕ1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ̄2 , ϕ3 , ϕ̄4 }}.
Similar computations show that we have the following equivalence classes:
[{ϕ1 , ϕ2 , ϕ3 , ϕ4 }] = {{ϕ1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ̄2 , ϕ3 , ϕ̄4 }};
[{ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 }] = {{ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 }, {ϕ1 , ϕ̄2 , ϕ̄3 , ϕ4 }, {ϕ1 , ϕ2 , ϕ̄3 , ϕ̄4 }, {ϕ̄1 , ϕ2 , ϕ3 , ϕ̄4 }};
[{ϕ̄1 , ϕ2 , ϕ3 , ϕ4 }] = {{ϕ̄1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ4 }, {ϕ1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ2 , ϕ3 , ϕ̄4 }
{ϕ1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ̄1 , ϕ2 , ϕ̄3 , ϕ̄4 }, {ϕ̄1 , ϕ̄2 , ϕ3 , ϕ̄4 }, {ϕ1 , ϕ̄2 , ϕ3 , ϕ4 }}.

3.1.3 The reflex fields in the case ρ2 = (0, 2)

We are interested in finding the reflex fields for all 16 CM-type. From Theorem 1.3.22,
we know that the fixed field of {σ ∈ Gal(K/Q) | σΦ = Φ} is the reflex field of (K, Φ).
Theorem 1.3.32 shows that each equivalence class leads to the same reflex field K r . Thus
we compute which σ ∈ Z2 ×Z4 fix Φ for the three representatives of the equivalence classes.
• For Φ1 = {ϕ1 , ϕ2 , ϕ3 , ϕ4 }, we compute that σΦ1 = Φ1 ⇐⇒ σ ∈ {(0, 0), (1, 0)}. Thus
we find that Φ1 has reflex K r = K ⟨(1,0)⟩ .
• For Φ2 = {ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 }, we compute that σΦ2 = Φ2 ⇐⇒ σ ∈ {(0, 0), (1, 2)}. Thus
we have that Φ2 has reflex K r = K ⟨(1,2)⟩ .
• Note that Φ3 = {ϕ̄1 , ϕ2 , ϕ3 , ϕ4 } has 8 equivalent CM-types and thus is primitive by
Theorem 1.3.28. From Theorem 1.3.34, we have that K r = K.
Comparing this result to the Diagram 3.1.1, we find that each CM-subfield is a reflex field
for some CM-type of K.

22
3.1.4 The primitive CM-types in the case ρ2 = (0, 2)
From Theorem 1.3.28, we know that Φ is primitive if and only if it has 8 equivalent
CM-types, which shows that {ϕ̄1 , ϕ2 , ϕ3 , ϕ4 } is primitive. Furthermore, Theorem 1.3.30,
shows that all 8 CM-types in the equivalence class of {ϕ̄1 , ϕ2 , ϕ3 , ϕ4 } are primitive. Lastly,
Theorem 1.3.33 shows that Φ1 is induced by a CM-type of the field K ⟨(1,0)⟩ and Φ2 is
induced by a CM-type of the field K ⟨(1,2)⟩ .

3.1.5 The CM-types in the case ρ1 = (1, 0)

Here, we study the second ρ-structure, that is the case for ρ1 = (1, 0) and ρ3 = (1, 2).
From Theorem 2.0.11, we get that the results for ρ3 = (1, 2) are identical (up to relabeling
of the Galois group) to the results of ρ1 = (1, 0). Thus only computations for ρ1 = (1, 0)
are shown.
When complex conjugation is given by (1, 0), the four embeddings that do not differ by
complex conjugation, together with their conjugate embeddings are given by

ϕi = (0, i) =⇒ ϕ̄i = (1, i) where 1 ≤ i ≤ 4.

Computing the equivalent CM-types in the same manner as in Section 3.1.2, we find that
the equivalent CM-types are given by

[{ϕ1 , ϕ2 , ϕ3 , ϕ4 }] = {{ϕ1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }};

[{ϕ̄1 , ϕ2 , ϕ̄3 , ϕ4 }] = {{ϕ̄1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ1 , ϕ̄2 , ϕ3 , ϕ̄4 }};
[{ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 }] = {{ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 }, {ϕ1 , ϕ̄2 , ϕ̄3 , ϕ4 }, {ϕ1 , ϕ2 , ϕ̄3 , ϕ̄4 }, {ϕ̄1 , ϕ2 , ϕ3 , ϕ̄4 }};
[{ϕ̄1 , ϕ2 , ϕ3 , ϕ4 }] = {{ϕ̄1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ1 , ϕ̄2 , ϕ3 , ϕ4 }, {ϕ1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ1 , ϕ2 , ϕ3 , ϕ̄4 }
{ϕ1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, {ϕ̄1 , ϕ2 , ϕ̄3 , ϕ̄4 }, {ϕ̄1 , ϕ̄2 , ϕ3 , ϕ̄4 }, {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ4 }}.

3.1.6 The reflex fields in the case ρ1 = (1, 0)

The reflex fields are given by the fixed points of the action of the Galois group on the left,
these can easily be computed in the same manner as in Section 3.1.3 to find that the above
equivalence classes have the following reflex fields
• Φ1 := {ϕ1 , ϕ2 , ϕ3 , ϕ4 } has associated reflex field K r = K ⟨(0,1)⟩ ;
• Φ2 := {ϕ̄1 , ϕ2 , ϕ̄3 , ϕ4 } has associated reflex field K r = K ⟨(1,1)⟩ ;
• Φ3 := {ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 } has associated reflex field K r = K ⟨(1,2)⟩ ;
• Φ4 := {ϕ̄1 , ϕ2 , ϕ3 , ϕ4 } has associated reflex field K r = K.
Comparing this result to the subgroup lattice of ρ1 = (1, 0), we find that not every CM-field
corresponds to a reflex field. In the lattice below, the CM-fields are circled in red if they
are reflex fields of some CM-type of K, and boxed in black if not.

23
 ρ = (1, 0)
{id}

⟨(1, 0)⟩ ⟨(1, 2)⟩ ⟨(0, 2)⟩

⟨(1, 0), (0, 2)⟩ ⟨(1, 1)⟩ ⟨(0, 1)⟩

Z2 × Z4

3.1.7 The primitive CM-types in the case ρ1 = (1, 0)

Theorem 1.3.28 tells us that the 8 CM-types in the equivalence class of {ϕ̄1 , ϕ2 , ϕ3 , ϕ4 } = Φ4
are primitive. Furthermore, Theorem 1.3.33 shows that Φ1 is induced from a CM-field
of K ⟨(0,1)⟩ , Φ2 is induced from K ⟨(1,1)⟩ and lastly Φ3 is induced from K ⟨(1,2)⟩ .

3.1.8 An example
In this section, we will find two Galois CM-fields K with Gal(K/Q) ∼ = Z2 × Z4 one of
which attains the results regarding CM-subfields, primitivity of CM-types and reflex fields
according to the case ρ2 = (0, 2) and the other one according to the case ρ1 = (1, 0).
The fields we will look at is the field K1 := Q(ζ16 ) and K2 the splitting field of the
polynomial p(x) := x8 + 8x6 + 20x4 + 16x2 + 1.
Note that K1 is a cyclotomic field and thus has Galois group (Z16 )× = Z2 × Z4 . Further-
−1
more K1 is a CM-field as Q(ζ16 + ζ16 ) ⊂ Q(ζ16 ) is a totally real subfield such that we
−1
have [Q(ζ16 ) : Q(ζ16 + ζ16 )] = 2.
Showing that K2 , the splitting field of p = x8 + 8x6 + 20x4 + 16x2 + 1, is a CM-field and
has Galois group Z2 × Z4 requires a bit more work and will be done via the following
observations.
1. Note that p is irreducible: Indeed, we have that p(x − 1) = x8 − 8x7 − 36x6 − 104x5 +
210x4 − 296x3 + 284x2 − 168x + 46 is an Eisenstein polynomial for the prime 2.
√
q p
2. The roots of f are given by x = ±i 2 ± 2 ± 3: To see this, we write the poly-
nomial p(x) = (x4 + 4x2 )2 + 4(x4 + 4x2 ) + 1. Repeated application of the quadratic
formula yields
r
√
q
p(x) = 0 ⇐⇒ x = ±i 2 ± 2 ± 3.

24
 √
q p
3. Set α := i 2 + 2 + 3, we have that K2 = Q(α) is the splitting field of p(x).
p √ √ √ p √ p √
Note that 2 + 3, 3, 2 ∈ Q(α) and 1/( 2 + 3) = 2 − 3 ∈ Q(α). Also
√ √
p q p
• 2 − 3/α = i 2 − 2 + 3 ∈ Q(α);
√ √ √
q p
• ( 2 + 3)/α = i 2 + 2 − 3 ∈ Q(α);
√ √
q p
• (1 + 2)/α = i 2 − 2 − 3 ∈ Q(α).
Thus Q(α) contains all roots of p(x) and hence is the splitting field of p(x).
4. We have G = Gal(Q(α)/Q) ∼
= Z2 × Z4 .
√ √ √
q p q p q p
Define β = i 2 − 2 + 3, γ = i 2 + 2 − 3, δ = i 2 − 2 − 3.
Next, define σ ∈ G to be the automorphism α → β. Then we have
p that√ sends p
2 2
√
that σ(α ) = β which is equivalent
√ to √σ(− 2 + √3) =√ 2 + 3. Rewriting the
√+ √ 3)/ − 2) = (1 + 3)/
nested square root gives σ((1 √ 2). √And thus√regarding
√σ
as an automorphism√ √ of Q( 2, 3), we see σ sends 3 → 3 and 2 → − 2.
Since αβ = (1 − 3)/ 2, we conclude that σ(αβ) = −αβ. From this, it follows
that σ(αβ) = σ(α)σ(β) = βσ(β) = −αβ =⇒ σ(β) = −α and thus σ has order 4.
Let us define τ ∈ G to be the automorphism√that√sends β → √γ. Then √ we have
2 2
that τ (β ) = γ pwhich is equivalent to τ ((1 + 3)/ 2) = (1 − 3)/ 2, so that τ
√ √ √ √ √
restricted to Q( 2,√ 3) can be√identified with sending 2 → 2 and 3 → − 3.
Now τ (βγ) = τ (1 − 2) = 1 − 2 = βγ. Thus √ we conclude
√ that τ (γ) √
= β and
√ τ has
order 2. Moreover, σ ̸= τ since τ (α ) = 2+τ (( 3+1)/ 2) = 2+(1− 3)/ 2 ̸= β 2 .
2 2

Lastly, we show στ√ = τ σ.√For this we √

first show
√ τ (β) = −δ and σ(−δ) = γ. Note
that τ (αβ) = τ (( 3 − 1)/ 2) = −(1 + 3)/ 2 = −γδ. From this, we can conclude
√
that√τ (α)τ (β) = τ (α)γ = −γδ =⇒ τ (α) = −δ. Furthermore, σ(αδ) = σ(1 + 2) =
1− 2 = −βγ. From this we get σ(αδ) = σ(α)σ(δ) = σ(δ)β = −βγ =⇒ σ(δ) = −γ.
Thus, (στ )(α) = σ(−δ) = γ and (τ σ)(α) = τ (β) = γ. Hence σ and τ commute.
By the previous observations, we have that |Gal(Q(α))| = 8. Since τ ̸= σ 2 , we
conclude that |⟨τ, σ⟩| = 8. Furthermore, σ and τ commute, giving us the following
representation: Gal(Q(α)/Q) = {σ, τ | σ 4 = τ 2 = id, τ σ = στ }, which is isomorphi
to Z2 × Z4 as desired.
p √
5. Lastly, we claim K2 is a CM-field. Indeed observe that Q( 2 + 3) ⊂ K2 is a totally
p √
real field and that [K2 : Q( 2 + 3)] = 2.
√
q p
We computed that both Q(ζ16 ) and Q( 2 + 2 + 3) are CM-fields and have the same
Galois group Z2 × Z4 over Q, we now compute the intermediate field lattices and indicate
the CM-subfields with boxes as before.

25
 4 2 −2
√
For K1 = Q(ζ16 ),√we observe that ζ16 = i, thus Q(i) is a √
subfield. Similarly
√ ζ16 + ζ16 = 2
2 6
and ζ16 + ζ16 = −2 give us the intermediate fields Q( 2) and Q( −2). As there are 3
subgroups of Z2 × Z4 with index 2, we know we have found all fields of degree 2 over Q.
Next, Q(ζ8 ) is the first intermediate field
p of √degree 4 over Q. Thepother√two degree 4
−1 3 5
extensions
p are√given by ζp 16 + ζ16 = 2 + 2 and ζ16 + ζ16 = i 2 + 2 to give us
√ √ √
that Q( 2 + 2) and Q(i 2 + 2) respectively. We have Q(i), Q( 2), Q( −2) ⊂ Q(ζ8 )
which shows what intermediate fields go on which places in the field lattice.
√ √ √
q p
Similar computations for K2 = Q(i 2 + 2 + 3) give the subfields Q( 2), Q( 3)
√
and Q( 6) as fields of degree 2 over Q.
√ √ p √ p √
The degree 4 fields are given by Q( 2, 3), Q(i 2 + 3) and Q(i 6 + 3 3). Construct-
ing the intermediate field lattice for both cases leads to
√
q p
Q(ζ16 ) Q(i 2 + 2 + 3)

p √ p √ p √ p √ √ √
Q( 2 + 2) Q(i 2 + 2) Q(ζ8 ) Q(i 2 + 3) Q(i 6 + 3 3) Q( 2, 3)

√ √ √ √ √
Q( 2) Q(i) Q( −2) Q( 2) Q( 3) Q( 6)

Q Q

Thus we see that the field K1 = Q(ζ16 ) corresponds to the case where ρ2 = (0, 2) is complex
conjugation whereas the field K2 corresponds to the case where ρ1 = (1, 0) (or ρ3 = (1, 2))
corresponds to complex conjugation.

3.2 The cyclic group Z8

Next, we look at the cyclic group Z8 . Since Z8 is cyclic, there can only be one element of
order 2, in this case this is the element ρ := 4 ∈ Z8 . Since there is only one element that
can represent complex conjugation, there will only be one ρ-structure, meaning that the
analysis for this case will go much quicker.

3.2.1 The subgroup lattice

Again, we first compute the subgroup lattice, and indicate which intermediate fields can
be CM-fields. Note that the only subgroups of Z8 are given by ⟨4⟩ of order 2 and ⟨2⟩ of
order 4. Thus the subgroup lattice looks as follows.

26
 {id}

Z2

Z4

Z8

Every subgroup of Z8 contains the element 4 ∈ Z8 , which corresponds to complex con-

jugation. Thus every subfield of a CM-field K with Galois group Z8 is fixed by complex
conjugation and therefore cannot be a CM-field. Thus K has no CM-subfields.

3.2.2 The CM-types

Recall that ρ = 4 ∈ Z8 . Thus the 4 embeddings that do not differ by complex conjugation
can be given by
ϕ1 = 0, ϕ2 = 1, ϕ3 = 2, ϕ4 = 3,
which gives that the conjugate embeddings must be

ϕ̄1 = 4, ϕ̄2 = 5, ϕ̄3 = 6, ϕ̄4 = 7.

Now, the 24 = 16 CM-types are given by choosing one embedding out of each pair ϕi , ϕ̄i
for 1 ≤ i ≤ 4. The equivalent CM-types can be computed in the same manner as Sec-
tion 1.3.9 to yield

[{ϕ1 , ϕ2 , ϕ3 , ϕ4 }] = {{ϕ1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ4 }

{ϕ̄1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ2 , ϕ3 , ϕ̄4 }};
[{ϕ1 , ϕ̄2 , ϕ3 , ϕ4 }] = {{ϕ1 , ϕ̄2 , ϕ3 , ϕ4 }, {ϕ̄1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ̄1 , ϕ̄2 , ϕ3 , ϕ̄4 }, {ϕ1 , ϕ̄2 , ϕ̄3 , ϕ4 }
{ϕ̄1 , ϕ2 , ϕ̄3 , ϕ̄4 }, {ϕ1 , ϕ̄2 , ϕ3 , ϕ̄4 }, {ϕ1 , ϕ2 , ϕ̄3 , ϕ4 }, {ϕ̄1 , ϕ2 , ϕ3 , ϕ̄4 }}.

3.2.3 The reflex fields

From Theorem 1.3.24, we know that reflex fields are CM-fields. From the definition of the
reflex fields, we also know that K r ⊆ K cl . Since K is Galois, we have K r ⊆ K. There are
no strict CM-subfields of K, and thus we have K r = K.

3.2.4 The primitive CM-types

All CM-types are primitive as there are no intermediate CM-fields that can induce any of
the CM-types.
This can also be reasoned through Theorem 1.3.33 as each reflex field is given by K r = K
and thus all CM-types are induced from the field K itself; i.e. they are primitive.

27