We extend this definition to rational functions in K al (C) in the following manner.

Definition 5.1.10. Let U ∈ K al (C) be a rational function and let us write R = G/H
for G, H ∈ K al [C]. The order of R at P , given as ordP (R), is defined as ordP (G)−ordP (H).
Rational functions R = G/H with polynomials G, H ∈ K al [C] are such that the polyno-
mials are of finite degree and thus have a finite number of zeros. In particular that means
that the order of R at P is nonzero at only finitely many places. This leads to the following
definition.
Definition 5.1.11. Let R ∈ K al (C), with R ̸= 0 be a rational function. Then the divisor
of R is X
div(R) = ordP (R)P.
P ∈C

Since ordP (R) is nonzero in only finitely many places, we have that div(R) ∈ D.
P
Divisors of the form div(R) = P ∈C ordP (R)P are called principal. The set of all principal
divisors is denoted by P.
Theorem 5.1.12.PLet R ∈ K al (C) be an arbitrary rational function. Then div(R) ∈ D0 ,
i.e. we have that P ∈C ordP (R) = 0. Furthermore, P is a subgroup of D0 .

Proof. This is Section II.3 of [Sil09].

Using Definition 5.1.10, the order at the point P∞ is also defined. However, in practice
it is easier to compute the order at infinity using the above theorem as the only pole of
polynomials is at infinity and thus the order of this pole must ‘cancel out’ the sum of the
orders of all the zeros. This is illustrated in the following example.
Examplep 5.1.13. Let C : y 2 = x17 + x, and P = (x0 , y0 ) where x0 ∈ K al and y0 is given
al
by y0 = + x17 0 + x0 . Note that the unique maximal ideal of K [C]P must be given by all
rational functions that are 0 at P (In a DVR R, the unique maximal ideal is given by R/R× ,
thus mP is the rational functions that are 0 at P ). We get mP = (x−x0 , y −y0 ) = (x−x0 ),
where the last equality follows from the fact that mP is generated by a single element.
Given that mP has only one factor of x − x0 , we conlcude that ordP (x − x0 ) = 1. Similarly,
it can be shown that P1 := (x0 , −y0 ) also gives ordP1 (x − x0 ) = 1. Lastly, if P2 = (x′ , y ′ )
where x′ ̸= x0 , then we have ordP2 (x − x0 ) = 0. Thus div(x − x0 ) = (P ) + (P1 ) − rP∞ . We
know x − x0 is a rational function and thus deg(x − x0 ) = 0, which gives us r = 2.
This concludes the discussion of the divisor of rational functions over function fields of
curves.

5.1.2 The regular differentials

Regular differentials form an important component of the study of curves. In this section,
we introduce regular differentials and show examples on how to compute them for various
curves.

50
Definition 5.1.14. Let C be a curve. The space of differentials on C is denoted by ΩC and
is a vector space over K al generated by symbols of the form dg for g ∈ K al (C). Elements
in the space ΩC are subject to the standard differentiation rules
1. d(f + g) = df + dg;
2. d(f g) = f dg + gdf ;
3. d(α) = 0 ∀α ∈ K al .
Example 5.1.15. Let C be given by y = x, so that dy = dx. Using these 2 relations, we
can eliminate y and dy from all elements f dg ∈ ΩC . Thus we get that all elements in ΩC
are of the form f (x)dx for some f ∈ K(x).
Next, we define the order of differentials. The standard definition of order/divisors of
differentials uses uniformizers much in the same way as the definition for order of the
rational functions. However, the order of the differentials can be related to divisors of
rational functions as is shown in [Sil09, Proposition 3.4]. We take this proposition as
definition.
Theorem 5.1.16. Let f, g ∈ K al (C) and P ∈ C such that g(P ) = 0. Then we have

ordP (f dg) = ordP (f ) + ordP (g) − 1.

Definition 5.1.17. Let ω ∈ ΩC be a differential of a curve C. Then the divisor of ω is

X
ordP (ω)(P ) ∈ div(C).
P ∈C

The differential ω is called regular if ordP (ω) ≥ 0 for all P ∈ C.

Example 5.1.18. Let C be given by y 2 = x17 + x. Write xi with 1 ≤ i ≤ 17 for the 17
roots of x17 + x, so that (xi , 0) ∈ C. We have that d(x − xi ) = dx − dxi = dx so that we
may use Theorem 5.1.16 to compute ord(xi ,0) (d(x − xi )) = ord(xi ,0) (x − xi ) − 1 = 2 − 1 = 1.
For points of the form P = (p1 , p2 ) with p2 ̸= 0, we have that ord(p1 ,p2 ) (d(x − p1 )) =
ord(p1 ,p2 ) (x − p1 ) − 1 = 1 − 1 = 0. Lastly, at the point P∞ , we must take 1/x as uniformizer.
Thus we find ordP∞ (dx) = ordP∞ (−1/x2 d(1/x)) = ordP∞ (−1/x2 ) + ordP∞ (1/x) − 1 = −3.
Similarly, we compute ord(xi ,0) (y) = 1 and ord(p1 ,p2 ) (y) = 0. Lastly, since y is a rational
function, we use Theorem 5.1.12 to find that the point P∞ is such that ordP∞ (y) = −17.
Thus we get that div(dx) = (x1 , 0) + (x2 , 0) + · · · + (x17 , 0) − 3(P∞ ) and div(y) = (x1 , 0) +
(x2 , 0) + · · · + (x17 , 0) − 17(P∞ ). In particular this means that div(dx/y) = 14(P∞ ) and
thus ω0 := dx/y is a regular differential.
Theorem 5.1.19. The space of regular differentials, {ω ∈ ΩC | ω is regular}, is a vector
space over K al . The dimension of this vector space is called the genus of C, denoted by g.

Proof. This can be found in [Sil09, Corollary 5.5 in Section II.5].

51
5.1.3 The Riemann-Hurwitz formula
In practice, one cannot compute the genus of a curve C using the definition, as it is
impractical to prove that a given set of differentials indeed spans the space of differentials.
The Riemann-Hurwitz formula gives the genus g in terms of the rammification points of a
curve C. This section covers the necessary theory to describe the theorem. Note that the
theory of this section is aimed at superelliptic curves.
Remark 5.1.20. Let C be a superelliptic curve given by y m = f (x). Denote with π a
projection of C onto P1 , that is, π(x, y) = x ∈ P1 . This yields a covering of P1 of degree m
in the sense that almost every point f (x) ∈ P1 has m distinct preimages. The finite points
where there are less than m associated values are exactly the points where f (x) = 0 as
here y m = f (x) = 0 =⇒ y = 0 and thus there is only one value of y associated to
this value of x. We call these points ramification points or branching points. The point
at infinity can also be a ramification point, the idea behind the ramification at infinity is
explained below.
Write C as y m = Πdi=1 (x − αi ), where we assume d < m. Projecting to P1 gives Π(x − αi ).
To see what happens at infinity, we make the change of coordinates given by x → 1/X.
i −1/X)
This gives Π (1/α(Xα i)
.
Next assume that d = cm − k where 1 ≤ k < m. We make a change of variables Y = yxc
to obtain Π(αi )Y m = X k Π(1/αi − X). Since we set x = 1/X and d < m, the point at
infinity is given by X = 0. At X = 0, the curve is ramified exactly when k ̸= 0 as this gives
us that at X = 0, the only value for Y is given by Y = 0. If k = 0, we find that at X = 0,
there are m choices for Y , and thus the curve is unramified at X = 0. Hence we find that
superelliptic curves of the form y m = f (x) are ramified at infinity exactly when d is not a
multiple of m. If m > d, a similar reasoning can be given to find that C is ramified at P∞
when m is not a multiple of d. Given that our definition of superelliptic curves requires
that gcd(d, m) = 1, all superelliptic curves are ramified at infinity.
We will take this intuition behind our definition of ramification. Please note that this is
not the standard definition of ramification points, but rather a corollary of theory that can
be found in [Koo91].
Definition 5.1.21. Let superelliptic curve C be given by y m = f (x) where f (x) has
degree d. Then the finite ramification points of C are given by the d (distinct) points (xi , 0)
where f (xi ) = 0. The point P∞ is always a ramification point.
The ramification index is an integer associated to each ramification point. The standard
definition is rather cumbersome, but the assumption that f (x) can have no repeated roots
and gcd(d, m) = 1, fixes the possible values for the ramification index, as is discussed
in [Koo91]. We take this result as definition.
Definition 5.1.22. Let C be a superelliptic curve given by y m = f (x). Then each ramifi-
cation point P has ramification index m, denoted by eP .
The ramification points and ramification index is used in the Riemann-Hurwitz theorem.

52
Theorem 5.1.23. Let y m = f (x) be a superelliptic curve. The genus g of C is given by
X
2g − 2 = −2m + (ep − 1).
P ∈C

Proof. See [Sil09, Theorem 5.9].

Example 5.1.24. From Definition 5.1.21, it follows that the ramification points of the
curve y 2 = x17 + x are given by the points (0, xi ) where 1 ≤ i ≤ 17 and xi ∈ K al is such
that x17
i +xi = 0. The point P∞ is also a ramification point and each ramification point has
ramification index m = 2 (see Definition 5.1.22). Thus by the Riemann-Hurwitz formula,
we find that 2g − 2 = −2 · 2 + 18 =⇒ 2g = 16 =⇒ g = 8.
Example 5.1.25. The ramification points of y 5 = x3 + 1 are given by the points (0, xi )
where 1 ≤ i ≤ 3 and xi is such that x3i +1 = 0. P∞ is also a ramification point and all points
have ramification index m = 5. The genus is then given by 2g − 2 = −2 · 5 + 16 =⇒ g = 4.
Remark 5.1.26. In Theorem 5.1.19, it was discussed that the regular differentials form
a vector space over K al of dimension g. Thus we can use the above result to note that
the curve C : y 2 = x17 + x has 8 linearly independent regular differentials and that the
curve C : y 5 = x3 + 1 has 4.
Example 5.1.27. This theorem allows us to find a basis for the regular differentials of
superelliptic curves. We return to the curve C given by y 2 = x17 +x. In Example 5.1.18, we
found that ω0 = dx/y is a regular differential. We claim that ωi = xi dx/y for 0 ≤ i ≤ g − 1
forms a basis for L(KC ). For this, note that div(x) = 2(0, 0) − 2(P∞ ). Hence we find
div(ωi ) = div(xi dx/y) = i(0, q1 ) + i(0, q2 ) + (14 − 2i)(P∞ ),
which is holomorphic exactly when i ≥ 0 and 14 − 2i ≥ 0 =⇒ 0 ≤ i ≤ g − 1 = 7. Thus
we found 8 regular differentials. To show that these 8 differentials are linearly independent
over K al , we note that
X X X X
ci ωi = 0 ⇐⇒ ci xi dx/y = 0 ⇐⇒ dx/y ci xi = 0 ⇐⇒ ci xi = 0.

Thus each ci must be 0 and we find that the 8 differentials are linearly independent and
thus must form a basis for L(KC ).
Example 5.1.28. Let C be given by the curve y 5 = x3 + 1, and let xi be the three zeros
of f (x) = x3 + 1. Denote by Qi the points (x, y) = (xi , 0) ∈ C. Then denote by Pi the five
points given by (0, ζ5i ) ∈ C.
First, we compute div(dx). For this, note that
ordPi (dx) = ordPi (d(x − xi )) = ordPi (x − xi ) − 1 = 5 − 1 = 4.
Similar computations show that for other finite points P , we have ordP (dx) = 0 and for
the point at infinity, we get ordP∞ (dx) = −6. Thus we find
div(dx) = 4(Q1 ) + 4(Q2 ) + 4(Q3 ) − 6(P∞ ).

53
Further computations show that
div(x) = (P1 ) + (P2 ) + (P3 ) + (P4 ) + (P5 ) − 5(P∞ );
div(y) = (Q1 ) + (Q2 ) + (Q3 ) − 3(P∞ ).

In Example 5.1.24, we computed that the genus of C was 4. Thus there must be 4 linearly
independent (over Qal ) regular differentials. We can compute these by making educated
guesses of combinations of the above divisors to find that the regular differentials are
• div(dx/y 2 ) = div(dx) − 2div(y) = 2(Q1 ) + 2(Q2 ) + 2(Q3 );
• div(dx/y 3 ) = div(dx) − 3div(y) = (Q1 ) + (Q2 ) + (Q3 ) + 3(P∞ );
• div(dx/y 4 ) = div(dx) − 4div(y) = 6(P∞ );
• div(xdx/y 4 ) = div(x) + div(dx) + 4div(y) = (P1 ) + (P2 ) + (P3 ) + (P4 ) + (P5 ) + (P∞ ).
Similar computations to Example 5.1.27 shows that these regular differentials are linearly
independent and thus must form a basis for L(KC ).

5.1.4 The Jacobian of curves

In this section, we introduce the Jacobian of a superelliptic curve. Recall that the Jacobian
is a generalization of the group structure for elliptic curves and thus is a group. However,
the Jacobian contains more structure, making it an abelian variety. The definition of
an abelian variety is out of the scope of this thesis, but an introduction can be found
in [Mil86a].
Definition 5.1.29. The Jacobian of a curve C, denoted by J(C), is an abelian variety
such that J(C) = D0 /P.
Theorem 5.1.30. [Mil86b, Proposition 2.1] The dimension of J(C) is equal to the genus
of C.
Definition 5.1.31. A Jacobian J(C) is called simple if it is not isogenous to the product
of abelian varieties over an algebraically closed field.
Remark 5.1.32. The Jacobian of a curve C is always an abelian variety. Hence if J(C) is
simple, then there does not exists a lower genus curve C ′ such that J(C ′ ) embeds in J(C).
However, keep in mind that not every abelian variety can be written as the Jacobian of a
curve.
Definition 5.1.33. An endomorphism of J(C) is a morphism of varieties respecting the
group structure of J(C). The set of all endomorphisms of J(C) is denoted by End(J(C))
and forms a ring.
RemarkP 5.1.34.P In particular, the endomorphism n : J(C) → J(C) which is defined
as n([ ai Pi ]) = [ n · ai (Pi )] is an endomorphism for all n ∈ Z and thus Z ⊆ End(J(C)).
Definition 5.1.35. The endomorphism algebra of J(C) is given by End(J(C)) ⊗ Q and is
denoted by End0 (J(C)).

54
We can relate an automorphism of C to a cyclotomic field present in the endomorphism
algebra of J(C) as follows.
Theorem 5.1.36. Let C be a curve with an automorphism ζ : C → C of order r. Then
we have that Q(ζr ) ⊂ End0 (J(C)).

Proof. We first relate the automorphisms of C to the endomorphisms of J(C). Define

G = Aut(C) to be the automorphism group of the curve C. Then an automorphism h ∈ G
induces an endomorphism [h] ∈ End(J(C)) by
X X
h : J(C) → J(C), h([ ai Pi ]) → [ ai h(Pi )].

Furthermore, if h1 , h2 ∈ G then m[h1 ] + n[h2 ] is an endomorphism on J(C) defined by

hX i h X X i
ai Pi 7→ m ai h1 (Pi ) + n ai h2 (Pi ) ,

and similarly [h1 ] · [h2 ] is an endomorphism on J(C) defined by

hX i h X i
ai Pi 7→ m ai h1 · h2 (Pi ) .
P
This gives an embedding from the group ring Z[G] := { mi [hi ] | hi ∈ G} into End(J(C));
i.e. we have Z[G] ,→ End(J(C)). Tensoring both Z-modules with Q yields the following
embedding Q[G] ,→ End0 (J(C)). Note that we assumed that there exists an automor-
phism ζ : C → C with order r. It can be easily checked that Q[ζ] = Q(ζr ) and thus

Q(ζr ) ∼
= Q[ζ] ⊂ Q[G] ,→ End0 (J(C)),

shows that the cyclotomic field Q(ζr ) embeds into End0 (J(C)).

The last notion we introduce is the notion of a CM-Jacobian.

Theorem 5.1.37. [Lan83, Theorem 3.1] Let J(C) be the Jacobian of some curve C with
genus g. If a field K of degree 2g embeds in End0 (J(C)), then K is a CM-field.
Definition 5.1.38. Let C be a curve of genus g. Let K be a CM-field of degree 2g over Q.
We say that J(C) has CM by K if there is an embedding K ,→ End0 (J(C)). In this case,
we call J(C) a CM-Jacobian.
Remark 5.1.39. It follows from Theorems 5.1.36 and 5.1.37, that a curve C with genus g
containing an automorphism of degree n such that |(Zn )× | = 2g, then J(C) has CM
by Q(ζn ).

5.2 Finding simple CM-Jacobians

In this section, we find Jacobians with simple CM-jacobians. In [TTV91], a method is
given to show that J(C) is simple.

55
Remark 5.2.1. As explained in Chapter 1 of [Lan83] if the Jacobian J(C) of a genus g
curve C has CM by a CM-field K then the complex representation of End0 (J(C)) is a C
vector space of dimension g and is isomorphic to the direct sum of a CM-type Φ of K. We
say that J(C) is of type (K, Φ).
Using the method in [TTV91] we read the CM-type of the CM Jacobians of some explicit
curves via the action of the automorphisms of C on the regular differentials of C.
Lastly, the main reason why we are interested in CM-Jacobians is that Jacobians of this
form have an easy to check equivalent criteria of being simple.
Theorem 5.2.2. [Lan83, Theorem 3.5] A Jacobian J(C) is of type (K, Φ) is simple if and
only if Φ is primitive.
Remark 5.2.3. Putting everything together, we show that the Jacobian of a curve C is
simple by means of the following steps.
1. Find (if possible) an automorphism ζ of C of order n such that [Q(ζn ) : Q] = 2g;
2. Compute the regular differentials {ω1 , . . . , ωg } of C;
3. Compute Φ := {ϕi : ζ2g 7→ ζ(ωi i )};
4. Check that Φ gives a CM-type of the field Q(ζ2g );
5. Check if Φ is primitive.

5.2.1 The first example

We will illustrate this method by applying it to the curve C given by y 5 = x3 + 1. Note
that ζ : C → C given by ζ(x, y) = (ζ3 x, ζ5 y) is an automorphism of C. Furthermore, note
that ⟨ζ⟩ ∼
= Z15 so that Q(ζ15 ) embeds in the endomorphism algebra of J(C). In particular,
note that [Q(ζ15 ) : Q] = 8 and recall that g = 4 is the genus of C (see Example 5.1.24)
so that we find a field of sufficiently large and thus J(C) is a CM-Jacobian. We will show
that J(C) is of primitive CM-type (Q(ζ15 ), Φ) and hence use Theorem 5.2.2 to conclude
that J(C) is simple.
The regular differentials of C were computed in Example 5.1.28, and are given by

ω1 = dx/y 2 , ω2 = dx/y 3 , ω3 = dx/y 4 and ω4 = xdx/y 4 .

The action of the automorphism ζ on the regular differentials is then given by

• ζ(ω1 ) = d(ζ3 x)/(ζ5 y)2 = ζ3 ζ53 ω1 = ζ15
14
ω1 ;
• ζ(ω2 ) = d(ζ3 x)/(ζ5 y)3 = ζ3 ζ52 ω2 = ζ15
11
ω2 ;
• ζ(ω3 ) = d(ζ3 x)/(ζ5 y)4 = ζ3 ζ5 ω3 = ζ15
8
ω3 ;
• ζ(ω4 ) = ζ3 xd(ζ3 x)/(ζ5 y)4 = ζ32 ζ5 ω4 = ζ15
13
ω4 .

56
As described in Remark 5.2.3, we define the embeddings ϕi (ζ15 ) = ζ(ωi ). This results
in the following four homomorphisms, which give embeddings of Q(ζ15 ). Furthermore,
since Q(ζ15 ) is Galois over Q, we may associate these embeddings with elements of the
Galois group, as follows from Theorem 1.2.9.
• ϕ1 (ζ15 ) = ζ15
14
=⇒ 14 ∈ (Z15 )× ;
• ϕ2 (ζ15 ) = ζ15
11
=⇒ 11 ∈ (Z15 )× ;
• ϕ3 (ζ15 ) = ζ15
8
=⇒ 8 ∈ (Z15 )× ;
• ϕ4 (ζ15 ) = ζ15
13
=⇒ 13 ∈ (Z15 )× .
This gives the set of embeddings Φ := {14, 11, 8, 13}. We check that Φ indeed gives a
CM-type. Note that |(Z15 )× | = 8, hence Φ is a CM-type when we show that Φ contains
no conjugate embeddings. Complex conjugation is given by the element 14 = −1 ∈ (Z15 )×
and none of the embeddings in {14, 11, 8, 13} differ by multiplication by −1, thus Φ is a
CM-type.
Lastly, we claim that Φ is primitive. For this, note that (Z15 )× ∼ = Z2 × Z4 via the isomor-
×
phism f : (Z15 ) → Z2 × Z4 such that f (14) = (1, 0) and f (2) = (0, 1). By applying the
automorphism to Φ, we find that Φ is given by the CM-type {(1, 0), (1, 2), (0, 3), (1, 1)} on
the CM-field Z2 × Z4 with ρ = (1, 0). By defining the embeddings as in Section 3.1.5, gives
that Φ = {ϕ1 , ϕ̄2 , ϕ̄3 , ϕ̄4 }, which is primitive as claimed in Section 3.1.7.
This shows that Jacobian of the curve y 5 = x3 + 1 is simple.

5.2.2 The second example

Next, we look at the curve C given by y 2 = x17 + x. As will be shown, the Jacobian of C
decomposes as it allows for an embedding of the Jacobian of a genus 4 curve. Moreover, we
will show that the Jacobian of the curve of genus 4 is simple. This example is more involved
than Example 5.2.1, and thus we first have to introduce some additional results about the
Jacobian. In particular, we find specific divisors that represent the equivalence classes of
Jacobians of hyperelliptic curves. We also introduce a corollary of Torelli’s theorem.
Definition 5.2.4. Let C be a hyperelliptic curve. A divisor D is called a reduced divisor
if the following holds
• D is of the form D =
P P
mi (Pi ) − ( mi ) (P∞ ) where each mi ≥ 0;
• Each point Pi ∈ Supp(D) is finite and Pi ∈ Supp(D) =⇒ P̃i ∈
/ Supp(D) unless we
have Pi = P̃i in which case mi = 1;
•
P
mi ≤ g where g is the genus of a curve C.
Here, Supp(D) denotes the points Pi ∈ C where ordPi (D) ̸= 0 and if P = (x0 , y0 ), then we
denote P̃i = (x0 , −y0 ).
Theorem 5.2.5. [MZW96, Theorem 47] Let C be a hyperelliptic curve, then each equiv-
alence class in J(C) is represented by a unique reduced divisor.

57
For this second example, we also need to relate the automorphisms of a curve C to the
automorphisms of the Jacobian J(C). This can be done through a corollary of Torelli’s
theorem.
Theorem 5.2.6. Let C be a superelliptic curve, then

Aut(J(C)) if C is hyperelliptic;
Aut(C) ∼
=
Aut(J(C)) × Z2 if C is not hyperelliptic.

Proof. This is a direct consequence of Torelli’s theorem, of which the proof is beyond the
scope of the thesis but can be found in [LS01, Theorem 3].

Definition 5.2.7. Let C be a curve and τ an automorphism of C. Then C/⟨τ ⟩ is the set
of points on the curve C that are fixed by the automorphisms in the group ⟨τ ⟩.
Example 5.2.8. Take y 2 = x17 + x and let τ : C → C be given by τ (x, y) = (1/x, y/x9 ) be
an automorphism of C. Since τ 2 = id, we have that C/⟨τ ⟩ is precisely given by the points
of C fixed under τ . We claim that C ′ = C/⟨τ ⟩ is also a hyperelliptic curve.
Lemma 5.2.9. Let C be given by y 2 = x17 + x and τ (x, y) = (1/x, y/x9 ) be an automor-
phism of C. Then C ′ = C/⟨τ ⟩ is a hyperelliptic curve.

Proof. To show this, note that τ (y/x4 ) = y/x5 and τ (y/x5 ) = y/x4 . Therefore we find
that τ fixes y/x5 + y/x4 = y(1 + 1/x)/x4 := η. Next, note that

η 2 = y 2 (1 + 2/x + 1/x2 )/x8

= (x17 + x)(1 + 2/x + 1/x2 )/x8
= x9 + x−9 + 2x8 + 2x−8 + x7 + x−7 .

Furthermore, note that τ fixes x + 1/x so that we can substitute s = x + 1/x to get

η 2 = (s + 2)(s8 − 8s6 + 20s4 − 16s2 + 2).

This can be readily verified to be a hyperelliptic curve C ′ with genus 4 and is fixed by τ .
Furthermore, we have that J(C ′ ) embeds into J(C). This follows directly from the fact
that every reduced divisor of J(C ′ ) also satisfies the criteria of being a reduced divisor
in J(C). In particular, this means that the Jacobian J(C) is not simple as it allows for an
embedding of the Jacobian of a smaller genus curve C ′ .

We find that the Jacobian of y 2 = x17 +x is not simple, however, we claim that the Jacobian
of C ′ defined by η 2 = (s + 2)(s8 − 8s6 + 20s4 − 16s2 + 2) is simple.
Theorem 5.2.10. Let C ′ be the curve given by η 2 = (s + 2)(s8 − 8s6 + 20s4 − 16s2 + 2),
then J(C ′ ) is simple.

58
 2
Proof. For this, observe that ζ : C → C given by ζ(x, y) = (ζ32 x, ζ32 y) is an automorphism
−1 ′
of C. We claim that ζ − ζ is an automorphism of C . To conclude this, we follow the
same argument as in [TTV91] and show that ζ − ζ −1 preserves the regular differentials
that are fixed under τ . For this, we first show that τ and ζ − ζ −1 commute.

2
(τ ◦ ζ)(x, y) = τ (ζ32 x, ζ32 y)
2 18 9
= (1/(ζ32 x), ζ32 y/(ζ32 x ))
30 15 9
= (ζ32 /x, ζ32 y/x )
= ζ 15 ◦ τ.

We conclude, τ ◦ ζ = ζ 15 ◦ τ . Therefore we find

τ ◦ (ζ − ζ −1 ) = τ ◦ ζ − τ ◦ ζ −1 = ζ 15 ◦ τ − ζ −15 τ = (−ζ −1 + ζ)τ,

and thus ζ − ζ −1 commutes with τ .

Next, we show that ζ − ζ −1 preserves the regular differentials fixed by τ . For this, let ωj
be a differential fixed by τ . Since τ and ζ − ζ −1 commute, we get

τ ◦ (ζ − ζ −1 )(ωj ) = (ζ − ζ −1 ) ◦ τ (ωj ) = (ζ − ζ −1 )(ωj ),

and thus τ fixes (ζ − ζ −1 )(ωj ).

From Example 5.1.28, it follows that a basis for the regular differentials for C was given
by ωi = xi dx/y for 0 ≤ i ≤ 7. We have that

x9−t d(1/x)
 
t dx dx
τ x = = −x7−t .
y y y

This shows that the invariant regular differentials under τ are given by
dx dx dx dx
ω1 = (1 − x7 ) , ω2 = (x − x6 ) , ω3 = (x2 − x5 ) , ω4 = (x3 − x4 ) .
y y y y

Thus we have found that these regular differentials are the regular differentials of C ′ .
Similar to Example 5.2.1, we compute the action of ζ − ζ −1 on these regular differentials
and check that we obtain a CM-type. If this CM-type is primitive, then J(C ′ ) is simple.
We compute the action of ζ − ζ −1 on the differential fixed by τ .

ζ 2t xt d(ζ32
2
 
t dx x) 2t+1 t dx
ζ x = 32 = ζ32 x .
y ζ32 y y

Using this, we find

• (ζ − ζ −1 )ω0 = (ζ32 − ζ32
−1
)ω0 ;

59
 • (ζ − ζ −1 )ω1 = (ζ32
3 −3
− ζ32 )ω1 ;
• (ζ − ζ −1 )ω2 = (ζ32
5 −5
− ζ32 )ω2 ;
• (ζ − ζ −1 )ω3 = (ζ32
7 −7
− ζ32 )ω3 .
−1 2i−1 −2i+1
We again define ϕi (ζ32 − ζ32 ) = ζ(ωi ) = ζ32 − ζ32 . We can compute that the minimal
−1
polynomial of ζ32 −ζ32 is given by p(x) = x −8x +20x4 −16x2 +2 and that the other roots
8 6
2i−1 −2i+1 2i−1 −2i+1
of this polynomial are given by ζ32 − ζ32 with complex conjugates −ζ32 + ζ32 .
−1
Thus, the functions ϕi define embeddings of the CM-field Q(ζ32 − ζ32 ). We concretely
write out the embeddings as follows
• ϕ1 (ζ32 − ζ32
−1 −1
) = ζ32 − ζ32 −1
=⇒ ϕ̄1 (ζ32 − ζ32 −1
) = −ζ32 + ζ32 ;
• ϕ2 (ζ32 − ζ32
−1 3
) = ζ32 −3
− ζ32 −1
=⇒ ϕ̄2 (ζ32 − ζ32 3
) = −ζ32 −3
+ ζ32 ;
• ϕ3 (ζ32 − ζ32
−1 5
) = ζ32 −5
− ζ32 −1
=⇒ ϕ̄3 (ζ32 − ζ32 5
) = −ζ32 −5
+ ζ32 ;
• ϕ4 (ζ32 − ζ32
−1 7
) = ζ32 −7
− ζ32 −1
=⇒ ϕ̄4 (ζ32 − ζ32 7
) = −ζ32 −7
+ ζ32 .
This shows that the action of ζ on the regular differentials yields the set of embed-
dings {ϕ1 , ϕ2 , ϕ3 , ϕ4 }, which is indeed a CM-type. Next we compute the Galois group
−1 −1 15 15
of the field Q(ζ32 − ζ32 ). For this, note that ζ32 − ζ32 = ζ32 + ζ32 is fixed by ζ32 and ζ32 .
−1 ∼ × ∼
This shows that Gal(Q(ζ32 − ζ32 )/Q) = (Z32 ) /⟨15⟩ = Z8 . From Section 3.2.4, we know
that all CM-types of a field with Galois group Z8 are primitive, in particular {ϕ1 , ϕ2 , ϕ3 , ϕ4 }
is primitive and thus J(C ′ ) is simple.

5.3 Generalizing this example

5.3.1 The curve y m = xd + 1
In this last section, we will generalize the proof of the fact that the Jacobian of the superel-
liptic curve y 5 = x3 + 1 is simple. The generalized curve we look at is given by Cd,m defined
as y m = xd + 1, where m > d are primes and m ≡ −1 (mod d). Note that y 5 = x3 + 1
is indeed of this form as 5 ≡ −1 (mod 3). We follow the same steps to prove that the
Jacobian of superelliptic curves of the form Cd,m is simple. For this, we first need to find
an automorphism of the curve.
Lemma 5.3.1. Let Cd,m be given as above, then ζ : C → C given by (x, y) → (ζd x, ζm y)
is an automorphism of C.
Lemma 5.3.2. The curve Cd,m has genus g = (d − 1)(m − 1)/2.

Proof. Note that projecting the x-coordinate of Cd,m to P1 yields an m-folded covering with
ramification points Qi = (xi , 0) (with 1 ≤ i ≤ d) where xi is such that xdi + 1 = 0. Further-
more, the projection is ramified at infinity as follows from Definition 5.1.21. At all d + 1
ramification points, the ramification index is m, which follows from Definition 5.1.22. Thus
by the Riemann-Hurwitz formula, Cd,m has genus

2g − 2 = −2m + (d + 1)(m − 1) =⇒ g = (d − 1)(m − 1)/2.

60
5.3.2 The regular differentials
In this section, we compute the regular differentials of Cd,m . The final result is given in
Theorem 5.3.12. In the following paragraphs, we follow the thought process which lead to
finding the regular differentials, for which we first introduce the following notation.
Definition 5.3.3. Let Cd,m be given by y m = xd + 1. Then Qi and Pi are the points

Qi = (xi , 0) where xdi + 1 = 0 for 1 ≤ i ≤ d, j

Pj = (0, ζm ) for 1 ≤ j ≤ m.

Lemma 5.3.4. The divisors of dx, x and y on the curve Cd,m are given by

div(dx) = −(m + 1)(P∞ ) + (m − 1)(DQ );

div(x) = −m(P∞ ) + (DP );
div(y) = −d(P∞ ) + (DQ ).

Here Q and P denote the sum of points

(DQ ) = Σdi=1 (Qi ) and (DP ) = Σm

i=1 (Pi ).

Proof. This follows from identical computations as seen in Example 5.1.28.

Remark 5.3.5. We claim that a basis for the space of regular differentials can be made
by combinations of the functions dx, x and y. An educated guess for the form of these
regular differentials, based on the regular differentials computed for y 5 = x3 + 1, is given
by ωs,t = xs dx/y t . We will show that the correct restrictions on the pairs (s, t) indeed give
that the ωs,t form a basis for the space of regular differentials.
Definition 5.3.6. Recall that m ≡ −1 (mod d), so that we can write m = dk − 1.
Throughout this section, the integer k is defined by k = (m + 1)/d. In particular, note
that k is the inverse of d modulo m.
Lemma 5.3.7. Let k be as above. Then dx/y k is a regular differential of Cd,m .

Proof. We have

div(dx/y k ) = div(dx) − kdiv(y)

= −(m + 1)(P∞ ) + (m − 1)(DQ ) + dk(P∞ ) − k(DQ )
= −(m + 1)(P∞ ) + (m − 1)(DQ ) + (m + 1)(P∞ ) − k(DQ )
= (m − k − 1)(DQ ).

Since k < m, we have that (m − k − 1) ≥ 0 and thus dx/y k is regular.

It turns out that we can slightly generalize the above computation to find that dx/y t is
also a regular differential for the correct choice of t.

61
Lemma 5.3.8. Let t ∈ Z with k ≤ t ≤ m − 1. Then dx/y t is a regular differential of Cd,m .

Proof. We compute that

div(dx/y t ) = div(dx) − tdiv(y)
= (−m − 1 + dt)(P∞ ) + (m − 1 − t)(DQ ).
Since we assumed k ≤ t, we have that m + 1 = dk ≤ dt. Thus −m − 1 + dt ≥ 0.
Furthermore, we assumed that t ≤ m − 1 so that m − 1 − t ≥ 0. Hence dx/y t is a regular
differential of Cd,m .

Remark 5.3.9. From Lemma 5.3.2, we have that the genus of Cd,m is g = (d−1)(m−1)/2.
We know that there are (d − 1)(m − 1)/2 linearly independent regular differentials. Note
that there are ’only’ m − k integer values t that satisfy k ≤ t ≤ m − 1. Thus, we are still
short by a considerable amount of regular differentials.
We will show that ωs,t = xs−1 dx/y t are regular differentials, as long as 1 ≤ s ≤ d − 1
and sk ≤ t ≤ m − 1. For this, first note that these bounds are well-defined as
1 ≤ s ≤ d − 1 =⇒ sk ≤ (d − 1)k ≤ m − k + 1 ≤ m − 1.
therefore, the values 1 ≤ s ≤ d − 1 allow us to bound t as sk ≤ t ≤ m − 1.
Lemma 5.3.10. Write ωs,t = xs−1 dx/y t where 1 ≤ s ≤ d − 1 and sk ≤ t ≤ m − 1.
Then ωs,t is a regular differential of Cd,m .

Proof. Again, we compute div(ωs,t ) to find

div(ωs,t ) = div(xs−1 dx/y t )
= (s − 1)div(x) + div(dx) − tdiv(y)
= (−m − 1 − (s − 1)m + dt)(P∞ ) + (m − 1 − t)(DQ ) + (s − 1)(DP )
= (dt − sm − 1)(P∞ ) + (m − 1 − t)(DQ ) + (s − 1)(DP ).
Note that dt−sm−1 = dt−s(dk−1)−1 = dt−skd+s−1 ≥ dt−dt+s−1 ≥ 0. Furthermore,
we assumed that m − 1 − t ≥ 0. This concludes that ωs,t is a regular differential.

We claim that the differentials ωs,t = xs−1 dx/y t under the restrictions 1 ≤ s ≤ d − 1
and sk ≤ t ≤ m − 1 form a basis for the space of regular differentials.
Lemma 5.3.11. The set {xs−1 dx/y t | 1 ≤ s ≤ d − 1 and sk ≤ t ≤ m − 1} forms a basis
for the space of regular differentials of Cd,m .

Proof. It suffices to show that there are g = (d − 1)(m − 1)/2 regular differentials in the
set {ωs,t } and that this set of regular differentials is linearly independent over K al . First
we show the linear independence. Take an arbitrary linear combination of the regular
differentials
X X X
ci ωs,t = 0 ⇐⇒ dx/y m ci xs y m−t = 0 ⇐⇒ ci xs y m−t = 0 ⇐⇒ ci = 0.

62
 P s m−t
Here, the last step follows from the fact ci x y is a polynomial in x, y where the
powers of x, y are such that 1 ≤ s ≤ d − 1 and 1 ≤ m − t ≤ m − 1. Therefore, we may
never replace y t with (a power of) xd + 1 or xs with (a power of) y m − 1. Since each term
in the sum has a unique power for x and y, we have that all ci must be 0.
We count the number of differentials by iterating over s. There are m − k differentials
when s = 1, since the corresponding restriction on t is given by k ≤ t ≤ m − 1. Similarly,
if we fix s = i, then t must satisfy ik ≤ t ≤ m − 1 and thus there are m − ik possibilities
for ωs,t whenever s = i. Given that s takes values 1 ≤ s ≤ d − 1, we have that the total
number of differentials equals

Σd−1
i=1 (m − ik) = (d − 1)m − k(d(d − 1)/2)
= (d − 1)m − (m + 1)(d − 1)/2)
= (2dm − 2m − dm − d + m + 1)/2
= (d − 1)(m − 1)/2.

This shows that we have found the g = (d − 1)(m − 1)/2 differentials that form a basis for
the space of regular differentials.

This finishes the proof of the final result on the regular differentials, which is stated below
for the sake of completeness.
Theorem 5.3.12. Let m > d be primes such that m ≡ −1 (mod d). Define k = (m + 1)/d
and let s ∈ Z such that 1 ≤ s ≤ d − 1. Define t ∈ Z such that sk ≤ t ≤ m − 1. Then the
differentials of the form xs−1 dx/y t form a basis for the space of regular differentials of the
curve y m = xd + 1.
Example 5.3.13. We use the above theory to find the regular differentials of the superel-
liptic curve y 5 = x3 + 1. Note that 5 = 2 · 3 − 1 so that we have k = 2, m = 5, d = 3. Thus
we have that 1 ≤ s ≤ d−1 =⇒ 1 ≤ s ≤ 2. Furthermore, sk ≤ t ≤ m−1 =⇒ 2s ≤ t ≤ 4.
Thus s = 1 =⇒ 2 ≤ t ≤ 4 and s = 2 =⇒ t = 4. This gives rise to the following pairs
for (s, t) : {(1, 2), (1, 3), (1, 4), (2, 4)}. Recall that ωs,t = xs−1 dx/y t so that the differentials
are given by dx/y 2 , dx/y 3 , dx/y 4 and xdx/y 4 , which agrees with Example 5.1.28.

5.3.3 The CM-type

From the method in [TTV91], we have that automorphism ζ applied to the regular dif-
ferentials ωs,t yields a CM-type of the CM-field K = Q(ζdm ) which has degree 2g over Q.
In this section, we will show that the action of automorphism ζ on the differentials ωs,t
gives a primitive CM-type of the field K. First note that K is Galois over Q with Galois
group Gal(K/Q) ∼ = (Zdm )× . Thus the embeddings of K are given by the elements in the
Galois group (Theorem 1.2.9), which will be used throughout this section.
Lemma 5.3.14. The action of ζ on the regular differentials ωs,t yields the set of embeddings

Φ := {ms − dt | 1 ≤ s ≤ d − 1, ks ≤ t ≤ m − 1} ⊂ (Zdm )× .

63
Proof. We compute ζ(ωs,t ) below:
−t s−1 −t
ζ(ωs,t ) = ζ(xs−1 dx/y t ) = ζds ζm x dx/y t = ζds ζm ωs,t
−t sm−dt
Next, note that ζds ζm = ζdm , which in turn corresponds to the element sm−dt ∈ (Zdm )× .
Thus, the set of embeddings is given by

{ms − dt | 1 ≤ s ≤ d − 1, ks ≤ t ≤ m − 1}.

Lemma 5.3.15. The set Φ = {ms − dt | 1 ≤ s ≤ d − 1, ks ≤ t ≤ m − 1} forms a CM-type

of the CM-field K = Q(ζdm ).

Proof. To show this, we first show that Φ ⊂ Gal(K/Q) ∼

= (Zdm )× . Note that

ms − dt ∈ (Zdm )× ⇐⇒ gcd(ms − dt, dm) = 1 ⇐⇒

 
gcd(ms − dt, d) = 1 s is not a multiple of d;
⇐⇒
gcd(ms − dt, m) = 1 t is not a multiple of m.

Given that 1 ≤ s ≤ d − 1 and k ≤ t ≤ m − 1, the last statement is trivially true, which

shows that ms − dt ∈ (ζdm )× .
Next, recall that there are (m − 1)(d − 1)/2 differentials ωs,t . We also claim that there
are (m − 1)(d − 1)/2 distinct values for ms − dt.

ms − dt = ms′ − dt′ ⇐⇒ m(s − s′ ) = d(t − t′ ).

Since d, m are primes, we must have that s−s′ is a multiple of d. Given that 1 ≤ s, s′ ≤ d−1,
this is only possible if s − s′ = 0 which shows s = s′ . A similar argument shows t = t′ .
Thus, distinct regular differentials correspond to distinct embeddings and this shows that
the set of (d − 1)(m − 1)/2 regular differentials correspond to (d − 1)(m − 1)/2 distinct
embeddings in (Zdm )× . Furthermore, we have |(Zdm )× | = (d − 1)(m − 1). Thus Φ contains
exactly half of the embeddings of K and hence is a CM-type upon showing that these
embeddings are non-conjugate.
Two embeddings in Φ being conjugate is equivalent to ms − dt ≡ −(ms′ − dt′ ) (mod dm).
Aiming for contradiction, we write sm − dt = cdm − s′ m + dt′ for some c ∈ Z. This gives
us that (s + s′ )m + cdm = d(t + t′ ). Since the right hand side of the equation is a multiple
of d, the left hand side must be so as well. Since d, m are primes, we have that s + s′ is
a multiple of d. Given that 1 ≤ s, s′ ≤ d − 1, we thus have that s + s′ = d. A similar
argument shows that t + t′ = m. Given that sk ≤ t ≤ m − 1 and s′ k ≤ t′ ≤ m − 1, we
have that m + 1 = dk = (s + s′ )k ≤ t + t′ . However, this contradicts that t + t′ = m. This
shows that the embeddings in Φ is non-conjugate. In particular, Φ is a CM-type of the
field Q(ζdm ).

64
5.3.4 CM-type Φ is primitive
In this section, we show that the CM-type Φ = {ms − dt | 1 ≤ s ≤ d − 1, ks ≤ t ≤ m − 1}
is a primitive CM-type of the field K = Q(ζdm ).
Remark 5.3.16. To show this efficiently, we first notice that K = Q(ζdm ) is Galois over Q
so that the embeddings are given by the automorphisms in the Galois group (Zdm )× . Then
we note that choosing (s, t) = (1, k) corresponds to ms − dt = m − dk = −1 ∈ Φ. We can
look at the conjugated CM-type Φ̄ = Φ ◦ (−1) so that 1 ∈ Φ̄. Any automorphism fixing Φ̄
must send 1 to some other embedding in Φ̄, thus this automorphism must be given by one
of the embeddings in Φ̄. This will be used to show that Φ̄ is primitive. Since Φ̄ and Φ are
equivalent CM-types, this will also show that Φ is primitive.
Lemma 5.3.17. The CM-type Φ̄ = {dt − ms | 1 ≤ s ≤ d − 1, ks ≤ t ≤ m − 1} is primitive.

Proof. Aiming for contradiction, we assume Φ̄ is not primitive, then Φ̄ ◦ σ = Φ̄ for some
non-trivial automorphism σ. By the explanation in Remark 5.3.16, we must have σ ∈ Φ̄.
Thus we write σ = (dt − sm). If σ fixes Φ̄, we must have for all embeddings (dt′ − s′ m) ∈ Φ̄
that (dt − sm)(dt′ − s′ m) ∈ Φ̄. i.e. we have (dt − sm)(dt′ − s′ m) ≡ (dt∗ − s∗ m) (mod dm)
for some dt∗ − s∗ m ∈ Φ̄. We first simplify this equation.

(dt − sm)(dt′ − s′ m) ≡ dt∗ − s∗ m (mod dm)

⇐⇒ d2 tt′ + ss′ m2 ≡ dt∗ − s∗ m (mod dm)
⇐⇒ ss′ ≡ s∗ (mod d) and dtt′ ≡ t∗ (mod m)

Recall that the last set of equations must hold for all embeddings in Φ̄, i.e. if Φ̄ were
not primitive, there must be a pair (s, t) such that for all (s′ , t′ ) we have that there is a
pair (s∗ , t∗ ) that solves  ′
ss ≡ s∗ (mod d)
dtt′ ≡ t∗ (mod m).
Moreover, recall that the following restrictions hold

1 ≤ s, s′ , s∗ ≤ d − 1 and sk ≤ t ≤ m − 1, s′ k ≤ t′ ≤ m − 1, s∗ k ≤ t∗ ≤ m − 1.

We will show that this is not possible.

For this, we split the proof in two cases. We first assume s ̸= 1. Since d is a prime, we
know that s ∈ (Zd )× and thus can choose s′ such s′ ≡ −s−1 (mod d). In particular, note
that s ̸= 1 =⇒ s′ ̸= d − 1. This choice of s′ gives that s∗ ≡ −1 (mod d). Since we
have 1 ≤ s∗ ≤ d − 1, we find that s∗ = d − 1.
From the bound s∗ k ≤ t∗ ≤ m − 1, we find that (d − 1)k = m − k + 1 ≤ t∗ ≤ m − 1. We
thus write t∗ = m − k + j where 1 ≤ j ≤ k − 1. Next, note that s′ k ≤ (d − 2)k = m − 2k + 1
implies that t′ = m − k is a valid choice for t′ (i.e. this choice does not violate the

65
bound s′ k ≤ t ≤ m − 1). Plugging these values into dtt′ ≡ t∗ (mod m) yields

dtt′ ≡ t∗ ⇐⇒ dt(m − k) ≡ m − k + j (mod m)

⇐⇒ −dkt ≡ −k + j (mod m)
⇐⇒ −(m + 1)t ≡ −k + j (mod m)
⇐⇒ t ≡ k − j (mod m)

This is a contradiction since 0 ≤ k − j < k and k ≤ t ≤ m − 1. Thus we conclude that

if Φ is primitive, then the automorphism fixing Φ̄ must be given by dt − sm where s = 1.
Next, we check the case when s = 1. Here we first set s′ = 1, so that ss′ ≡ s∗ (mod d)
implies that s∗ ≡ 1 (mod d). Given that 1 ≤ s∗ ≤ d − 1, we conclude s∗ = 1. In particular,
this means that k ≤ t, t′ , t∗ ≤ m − 1 since s = s′ = s∗ = 1. Given that m is a prime
and that d, t are not multiples of m, we see that dt ∈ (Zm )× . Let a ∈ (Zm )× be such
that (dt)a ≡ 1 (mod m). If k ≤ a ≤ m − 1, then set t′ = a so that dtt′ ≡ 1 (mod m).
Again, since k ≤ t∗ ≤ m − 1, this is a contradiction.
Next assume a = 2, 3, . . . , k − 1. In this case, we cannot set t′ = a as we must have
that k ≤ t′ ≤ m − 1. However, now set t′ = a · ⌈k/a⌉ (note: k ≤ a · ⌈k/a⌉ < 2k ≤ m − 1 so
that this value for t′ is allowed). Thus, we have dtt′ = dta·⌈k/a⌉ ≡ ⌈k/a⌉ (mod m). Hence,
in particular, dtt′ ≡ t∗ =⇒ t∗ ≡ ⌈k/a⌉ (mod m). Since a ≥ 2, we have that ⌈k/a⌉ < k
which is a contradiction as k ≤ t∗ ≤ m − 1.
Lastly, if a = 1, then dta ≡ 1 (mod m) ⇐⇒ dt ≡ 1 (mod m) ⇐⇒ t = k. This shows
that the only automorphism that can fix Φ̄ is given by (s, t) = (1, k), which corresponds
to dt − ms = dk − m = 1, which trivially fixes Φ̄. Thus we conclude that Φ̄ is primitive.
In particular, Φ is primitive and J(Cd,m ) is simple.

Corollary 5.3.18. The Jacobian of y m = xd +1 where m > d are primes such that m ≡ −1
(mod d) is simple.
Remark 5.3.19. It is expected that the condition that m ≡ −1 (mod d) is not necessary.
This condition was imposed to make the computations in the proof shorter and more
tangible, but is not necessary from any fundamental point of view. As an example, consider
the curve C given by y 7 = x3 + 1. Note that 7 ̸≡ −1 (mod 3), but the curve C has simple
Jacobian, as is shown below.
Note that ζ(x, y) = (ζ3 x, ζ7 y) is an automorphism of C, and that the genus of C is g = 6.
It can be computed that a basis for the space of regular differentials is given by

ω1 = dx/y 3 , ω2 = dx/y 4 , ω3 = dx/y 5 , ω4 = dx/y 6 , ω5 = xdx/y 5 , ω6 = x2 dx/y 6 .

Note that we have that Q(ζ21 ) is present in the endomorphism ring of J(C). Further-
more, ζ(ωi ) corresponds to a CM-type of Q(ζ21 ). One can compute that this CM-type is
given by Φ = {19, 16, 13, 10, 20, 17} ⊂ (Z21 )× . One can check that this CM-type is only
fixed by 1 ∈ (Z21 )× and is thus primitive. We conclude that J(C) is primitive.

66
Conclusion
We find that the Jacobian J(C) is simple for the hyperelliptic curve C that is given
by η 2 = (s + 2)(s8 − 8s6 + 20s4 − 16s2 + 2) and for the family of superelliptic curves C
given by y m = xd + 1 where m > d are primes and m ≡ −1 (mod d). It is expected that
further computations can show that the condition m ≡ −1 (mod d) is not necessary; i.e.
that the curve y m = xd + 1 has simple Jacobian for all primes m > d.
The simple Jacobian of the curve η 2 = (s + 2)(s8 − 8s6 + 20s4 − 16s2 + 2) has CM by the
−1
field Q(ζ32 − ζ32 −1
), which has Galois group Gal(ζ32 − ζ32 /Q) ∼
= Z8 . The Jacobian of the
curve y = x + 1 had CM via the field Q(ζ15 ), for which Gal(Q(ζ15 )/Q) ∼
5 3
= Z2 × Z4 . A
natural question that arises is: Can a simple Jacobian have CM by a Galois CM-field K
where Gal(K/Q) is any of the studied groups of order 8?
The answer is not always: In our classification, we found that CM-fields K with Galois
group D4 have no primitive CM-types, and thus a Jacobian with CM by K cannot be
simple. However, as is seen in the classification, fields with Galois group Q8 and (Z2 )3
both have primitive CM-types and thus could allow for simple Jacobians. It would be an
interesting exercise to find simple Jacobians that have CM by a field with Galois group Q8
or (Z2 )3 , or to find a decomposable Jacobian that has CM by a field with Galois group D4 .

67
6 References
[HR06] Christopher J. Hillar and Darren Rhea. Automorphisms of finite abelian groups,
2006.
[Koo91] Ja Kyung Koo. On holomorphic differentials of some algebraic function
field of one variable over c. Bulletin of the Australian Mathematical Society,
43(3):399–405, 1991.
[Lan83] S. Lang. Complex Multiplication. Grundleheren der math. Wiss,255. Springer-
Verlag, 1983.
[LS01] Kristin Lauter and Jean-Pierre Serre. Geometric methods for improving the
upper bounds on the number of rational points on algebraic curves over finite
fields, 2001.
[Mil] J.S. Milne. Complex multiplication.
[Mil86a] J. S. Milne. Abelian Varieties, pages 103–150. Springer New York, New York,
NY, 1986.
[Mil86b] James S. Milne. Jacobian varieties. In Arithmetic geometry (Storrs, Conn.,
1984), pages 167–212. Springer, New York, 1986.
[Mos23] O. Mos. Dikke dijen, 2023.
[MZW96] A. Menezes, R. Zuccherato, and Y.H. Wu. An Elementary Introduction to Hyper-
elliptic Curves. CORR Report. Faculty of Mathematics, University of Waterloo,
1996.
[Sil09] Joseph H Silverman. The Arithmetic of Elliptic Curves. Graduate texts in
mathematics. Springer, Dordrecht, 2009.
[Tow93] C.W. Towse. Weierstrass Points on Cyclic Covers of the Projective Line. Brown
University, 1993.
[TTV91] Walter Tautz, Jaap Top, and Alain Verberkmoes. Explicit hyperelliptic curves
with real multiplication and permutation polynomials. Canadian Journal of
Mathematics, 43(5):1055–1064, 1991.

68