A Beginning Approach to The Theory of

Generalized Modular Forms – ”A-Castle Forms”,

Automorphic Factors, etc.
Philipp Harland
May 2025

Email: [email protected]

Abstract
In this paper, we will present a few new extensions of the theory of
modular forms, a departure from the typical theory – i.e. the theory of
modular forms over the matrix group SL2 (Z) with inputs in HC .

Contents
1 Introduction 1
1.1 Introductory Theory – Clarification . . . . . . . . . . . . . . . . . 1

2 Results of Research 2
2.1 Modular Forms in SLk (Z) . . . . . . . . . . . . . . . . . . . . . . 2
2.2 ”Generalized Automorphic Factors” . . . . . . . . . . . . . . . . 3
2.3 Eisenstein Series . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.4 Generalizations of the congruence subgroup Γ0 (N ) . . . . . . . . 4
2.5 ”Hypercomplex Modular Forms” . . . . . . . . . . . . . . . . . . 5

3 Works Referenced/Utilized/Further Reading 5

1 Introduction
1.1 Introductory Theory – Clarification
A modular form, in the standard literature, M, of weight k and level Γ is a
holomorphic function from the complex upper half plane, H, to the complex
numbers, C, which satisfies the following conditions:
1: ∀γ ∈ Γ ⊂ SL2 (Z), f (γ(z)) = (cz + d)k f (z).
2: ∀γ ∈ SL2 (Z), (cz + d)−k f (γ(z)) is bounded across the limit limz→i∞ .

1
2 Results of Research
2.1 Modular Forms in SLk (Z)
The first topic we are to go over is defining modular forms w.r.t. ”generalized
Möbius transformations” emerging from matrices in SLk (Z) for a general k
∈ Z+ . We introduce the notion of an ”A-castle form” for a subgroup χ of
SLk (z), where an A-castle form of weight w and level χ is a ”modular form-
like” function based off of a subset of Z+ × Z+ , A.
For our A-castle forms, they follow an analogous set of conditions:
Tg1 T (z)
1: For all ∈ χ ⊂ SLk (Z), f ( Tgg1 (z) ) = Tgw2 f (z).
Tg2 2
2: For all
Tg1
Tg2 ∈ χ ⊂ SLk (Z), Tg−w
2
f (γ(z)) is bounded across the limit limz→i∞ .
Pk j−1
For a matrix, M, in SLk (Z), we define Ti (z) = j=0 aij z . We use the
term ”generalized Möbius transformation” to refer to the rational function of
T
two such T, Tll1 for l1 , l2 ∈ {1, ..., k}.
2
For ”standard” modular forms where k=2, we obtain our usual equations as
discussed in 1.1 when we consider a so-called A-castle form using the singleton
set A = {(1, 2)}, constituting the transformation TT21 (z)
(z)
, where T1 = az + b, T2 =
 
a b
cz + d, given a matrix . This is trivial to verify by simple substitution.
c d
There are 2 properties that modular forms satisfy, namely those of addition and
multiplication:
(1):M + N is a modular form whenever M and N are;
(2):M N is a modular form, also whenever M and N are.
We will now prove that the 2 identities above hold for A-castle forms as well.
We will construct these 2 proofs similarly to how one would construct the proofs
of (1) and (2).
Proof of (1).
I.a: G1 , G2 are A-castle forms, for the same A, χ and weight w.
II.a: This means G1 (γ(z)) = D(γ)w G1 (z), and G2 (γ(z)) = D(γ(z))w G2 (z).
III.a: Invoking the distributive property, we get (G1 +G2 )(γ(z)) =
D(γ)w (G1 + G2 )(z).
IV.a: Thus, the first transformation property is satisfied.
I.b: Similar to subproof a, we have G1 + G2 satisfying the second
condition -- the boundedness condition -- as noted in 1.1. This is
since a sum of 2 bounded functions is a bounded function itself. This
is trivial to prove.
Conclusion: The sum of 2 A-castle forms of the same weight is an A-castle
form.
Proof of (2).
I.a: Similar to before, G1 and G2 both satisfy the first condition.
Additionally, G1 has a weight of k, and G2 has a weight of l.
II.a: (G1 G2 )(γ(z)) = D(γ)k D(γ)l · z = Dk+l (γ) · z.

2
I.b: The product of 2 bounded functions is necessarily a bounded function
itself. This is trivial to prove.
Conclusion: The product of 2 A-castle forms is an A-castle form.
Additionally, just like ”regular” modular forms, A-castle forms create a monoid
under addition and multiplication. This is because they also satisfy the other 2
monoid axioms, which are associativity and identity.

2.2 ”Generalized Automorphic Factors”

Previously, we discussed A-castle forms. Now, we will move onto the theory of
”automorphic factors”, or, functions of the type:

ε(a, b, c, d)(cz + d)k

defined w.r.t. SL2 (R). Which, are proposed to generalize the condition of mod-
ularity. Here, we call ε the ”nebentypus” of the automorphic factor. We can
replace the (cz + d)k factor in the original condition to give:

∀γ ∈ SL2 (Z), f (γ(z)) = ε(a, b, c, d)(cz + d)k f (z)

Now, when we have an A-castle form of weight w and level χ , G, in SLn (Z)
we can create a similar condition with a ”generalized” nebentypus, ψ, which is
the following:
T
∀ Tgg1 = Ta ∈ Hχ ⊂ SLk (Z), G(T (z)) = ψ(a11 ...akk )Tgw2 G(z)
2

An automorphic factor itself, is a function ν which is analytic and can be written

in the form of a product of a function υ and an integer power of the denominator
of a transformation γ:
ν(γ, z) = υ(γ)(cz + d)k
where |υ(γ)| = 1.
The way above is one possible way to generalize the theory of automorphic fac-
tors, and is what it would seem the ”most intuitive” for the theory we’ve built
up.

2.3 Eisenstein Series

The Eisenstein series for a positive even integer, d, is a special, interesting case
of a family of modular forms, in which this paper, deserves its P own section. 1
The typical definition of the Eisenstein series, Gd is Gd (τ ) = m,n∈Z−{0} (m+nτ )d
,τ ∈
H0CD . Keep in mind, this is for d2 > 1.
We will now cover the first way we can generalize this
Pparticular family of modu-
lar forms. We can rewrite the definition as Gd (s) = mω1 +nω2 =λ∈Zω1 +Zω2 (mω +1mω2 s )d .
1 i
When Λ = Z + Zi, or, Λ = Z[i], it reduces to the usual definition of the Eisen-
stein series of ”weight” d shown above. We can easily verify this by having
ω1 = 1, ω2 = i in our definition, and noting ωi2 = 1. The second way we can

3
generalize this family of modular forms is by considering larger lattices, e.g.
lattices over H or O. We define the following ”Generalized Eisenstein series”,
for a lattice, Φ, in CDm (C) which is generated by the 2m+1 -ions ω1 , ..., ω2m+1 ,
which the function will be of several τi ∈ CDm (C)–
GΦ  1
P
d (τ1 ...τ2m+1 −1 ) = p1 ,...,p2m+1 ∈Z−{0}  τj−1
d
j ≥ 2

P2m+1 Φ
pj ωj im+1,j 
 j=1 
j

=1 1

Here, analogously it reduces to a ”normal” Eisenstein function when our lattice

is equal to Z+Zim+1,1 +...+Zim+1,2m −1 , and it further reduces to the Eisenstein
series we first defined in this section if m=0, which is the typical Eisenstein
series known in the literature.
To quickly clarify our notation, we use the convention in,k as the nth imaginary
unit of CDk−1 (C).
We are aware, in the standard literature in the theory of elliptic functions and
modular forms of the mapping for certain lattices, Zω1 +Zω2 → Z+Zτ, τ = ω ω1 .
2

However, a direct analogy of this convenient mapping is not possible for larger
lattices, as there is no way currently known of to condense ω1 , ..., ω2m into a
single parameter that can describe any lattice (of a certain type) in the way τ
for (certain) complex lattices does.

2.4 Generalizations of the congruence subgroup Γ0 (N )

The Hecke congruence subgroup,
 Γ0 (N  ) is usuallyconstructed
 as the subgroup
a b a b
of SL2 (Z), where Γ0 (N ) = ∈ SL2 (Z)| %N = I2×2 , where
c d c d
M %k ij = Mij %k, % being our notation for ”reduced modulo”. We then have
the congruence subgroup Γ1 (N ), which we similarly define by the constraint
where a, b%N = 1, c%N = 0. We see here the constraint has been relaxed on b.
We can create a ”generalized” congruence group that is ”like” Γ0 (N ) through
the following construction: We have a positive integer, d. That is, an element
of Z+ . Now, we consider the multiplicative group of rank d tensors, or, ten-
sors with d indices used to specify their elements. For example, matrices have
2 indices used to specify each element, so we consider them ”rank 2 tensors”.
We introduce the notation Γd0 (Z), which denotes the group: Γd0,m (Z) = {A ∈
SLm × ... × m (Z)|A i...i %B = 1, Ai1 ,...,id %N = 0 given ¬(i1 = ... = id )}. This
| {z } |{z}
d d
works as well as our initial family of subgroups of SL2 (Z), Γ0 (N ) if we con-
sider the concepts of hyperdeterminant and tensor multiplication, but we cannot
be easily assured that the group axioms are even satisfied given a product of 2
hypermatrices, or, higher dimensional tensors.
We can also note the existence of the family of groups

4
    

 a11 . . ar1 

r×r
  . . 

Γ0 (N ) = A =   ∈ SLr (Z) : A%N = Ir×r , r ∈ Z+
  . .  
 
a1r . . arr
 

, for which we can construct ”congruence” A-castle forms.

These are easy to construct, but it’s unclear how we would construct the Γr×r
1 (N )
counterpart to this family. For now, we will construct the group by relaxing the
constraint on ar1 , as we did in the case of the ”original” Γ1 (N ) group, where
we relaxed the constraint on the ”top right” element b.

2.5 ”Hypercomplex Modular Forms”

Here, basing our insight off the typical definition of a modular form, we can
generalize the domain of modular forms to any upper half-plane in any hyper-
complex number system, i.e. a Cayley-Dickson construction.
We let HnCD be the ”upper half plane” of the nth Cayley-Dickson construction
△
from C onwards, where H0CD = HC . That is, the subset, H, of CDn−1 (C) of
that such construction s.t. ∀p ∈ H, sgn(Im2n −1 (p)) = 1. We can obtain this set
from CDn−1 (C) by applying the transformation:
(
a0 + ... + a2n −1 i2n −1 sgn(a2n −1 ) = 1
a0 + a1 i1 + ... + a2n −1 i2n −1 →
a0 + ... − a2n −1 i2n −1 sgn(a2n −1 ) = −1

. This acts as a ”quotient map” which maps the entire space onto its upper half
n
plane. We can similarly obtain such a half plane from R2 , which is naturally
isomorphic to CDn−1 (C). We have the standard q-series, where q = e2πinz .
We can then substitute z for any p in HkCD , to obtain a generalized ”q-series”
which will help us define modular forms in this manner.
Keep in mind, we will have to compute the terms with the Taylor series of ex ,
which, for computerized applications, may be cause for very volatile computa-
tional complexity, especially if we want a decent amount of precision ( 10 or
so terms). Since hypercomplex numbers transform similarly to C, it is expected
that they too can generate modular forms, yet we have yet to prove/disprove the
affirmative. A formal treatment of this conjecture will be given in a later paper.

3 Works Referenced/Utilized/Further Reading

Robert Rankin, Modular Forms and Functions (1977) – Section 3, pp. 70-88
Enrico De Ronche – Matrix valued Hilbert modular forms (further reading: re-
lated work)