Ramanujan 4

20
4. Theta functions and the Ramanujan–Petersson conjecture

In the previous lecture, we defined a procedure for constructing a graph Gp (O)
for any prime p and order O ⊂ H. Actually, it makes sense to define a graph Gn (O)
for any positive integer n; it just happens that this graph tends to be Ramanujan
only when n is prime. Let An (O) be the adjacency matrix of Gn (O).
Let us return to the example where O ⊆ H is the order generated by
√ √ √
i − 3k √ 1 + 3i + 3j + 3k
1, , i − 3j, .
2 2
� � � �
1 1
In this case, the An (O) are 2 × 2 matrices. For each n, the vectors and
1 −1
� �
1
are eigenvectors. The eigenvalues of the An (O) acting on are
−1
1, −1, −1, −1, 1, 1, 0, 3, 1, −1, −4, . . . .
Now consider the power series
∞
�
g(q) = q (1 − q n )(1 − q 3n )(1 − q 5n )(1 − q 15n )
n=1
= q − q 2 − q 3 − q 4 + q 5 + q 6 + 3q 8 + q 9 − q 10 − 4q 11 + · · · .
The coefficients of this power series are the same as the eigenvalues of An (O). This
is not a coincidence!
So why would these two sequences equal? It is a bit of a mystery. In number
theory, sometimes you can prove something and still feel like you don’t really un-
derstand why it’s true. Some people might find this frustrating, but personally, I
think it’s really cool!
Anyway, the way that I know to prove that the above sequences are equal is to
prove that they are both modular forms. To motivate the definition of a modular
form, let’s look at symmetries of the complex plane again.
M. C. Escher, drawing E55
This drawing has rotational and translational symmetry. If we think of this drawing
as living in the complex plane, its symmetries are described by transformations of
the form
f (z) = eiθ z + c .
21
It is less obvious how to describe the symmetries of the following drawing in terms
of complex numbers.
M. C. Escher, Circle Limit III
But it can be done. Consider the set of transformations of the form

az + b
f (z) = ∗ , a, b ∈ C, |a| > |b| .
b z + a∗
These turn out to preserve the disc. We can see that they preserve its boundary,
the unit circle, as follows. If |z| = 1, then
|az + b| = |azz ∗ + bz ∗ | = |a + bz ∗ | = |a∗ + b∗ z|
so |b|az+b|
∗ z+a∗ | = 1. The condition |a| > |b| guarantees that the transformation pre-
serves the interior of the circle, rather than swapping the interior and exterior.
Similarly, we can consider the following drawing.
J. Leys, after M. C. Escher, drawing E69
The drawing is meant to be infinite, filling the upper half plane. The symmetries
of this drawing are of the form
az + b
f (z) = , a, b, c, d ∈ R, ad − bc > 0 .
cz + d
It is clear that such transformations preserve the real line. The condition ad−bc > 0
guarantees that the transformation preserves the upper half plane, rather than
swapping the upper and lower half planes.
Now let’s actually define modular forms.
Definition 4.1. Let k be a nonnegative integer, and let N be a positive integer.
A holomorphic function f on the upper half complex plane is a modular function
of weight k and level Γ0 (N ) if, for any integers a, b, c, d satisfying ad − bc = 1 and
N | c, � �
az + b
f = (cz + d)k f (z) .
cz + d
22
We say that f is a cuspidal modular form if, in addition, limz→t f (z) = 0 for any
t ∈ Q ∪ {∞}.
We write Sk (Γ0 (N )) for the space of all cuspidal modular forms of weight k and
level Γ0 (N ).
Proposition 4.2.
g(e2πiz ) ∈ S2 (Γ0 (15)) .
This is not at all obvious. There is a proof using Fourier analysis and a proof
using complex analysis. I will give a reference below when I state a more general
theorem.
We can also interpret the coefficients of f as eigenvalues of operators called
Hecke
� operators. � Let p be a prime number. Let Tp to be the operator defined by
Tp ( a n q n ) = bn qn where
�
apn + pan/p , p | n
bn =
apn , p � n.
Let’s compute T2 (g).
T2 (g) = −q + q 2 + q 3 + q 4 − q 5 + · · · . = −g .
�∞
Every element of Sk (Γ0 (N )) has an expansion of the form f (z) = n=1 an q n ,
where q = e2πiz . One can show that whenever p � N , Tp preserves Sk (Γ0 (N )).
Furthermore, S2 (Γ0 (15)) is one-dimensional. Therefore, g is an eigenvector of Tp
/ {3, 5}. The coefficient of q 1 in f is 1, while the coefficient of q 1 in Tp g is the
for p ∈
same as the coefficient of q p in g. Therefore, the Tp -eigenvalue is the coefficient of
qp .
Theorem 4.3 (Ramanujan–Petersson conjecture). Let k ≥ 2 and N be positive
integers. Let p be a prime not dividing N . Then the eigenvalues of Tp acting on
Sk (Γ0 (N )) satisfy
|λ| ≤ 2p(k−1)/2 .
The conjecture was proved by Eichler for k = 2 and by Deligne for all k. The
proof uses some techniques (in particular, algebraic geometry over Z/pZ) that I
will not be able to cover in this course.
√
In particular, when k = 2, we get |λ| ≤ 2 p. To prove that the graph Gp (O)
is Ramanujan, one constructs cuspidal modular forms whose Tp -eigenvalues are
the nontrivial eigenvalues of Ap (O). The Ramanujan–Petersson conjecture then
implies that Gp (O) is Ramanujan. Let us now explain how these modular forms
are constructed.
Proposition
  4.4. Let p be a prime, and let e be an eigenvector of Ap (O). Suppose
1
 
e �=  ... . For each n, let λn denote the eigenvalue of An (O) acting on e. Define
1
a function θe on the upper half complex plane by
∞
�
θe (z) = λn e2πinz .
n=1
Let
O∗ := {x ∈ H|tr xy ∈ Z for all y ∈ O} .
23
Let N be the smallest positive integer such that N O∗ ⊆ O. Then θe (z) is a cuspidal
modular form of weight 2 and level Γ0 (N ), and Tp θe (z) = λθe (z).
Proof. See [Shi73, Proposition 2.1]. �
The fact that Gp (O) is Ramanujan is then a consequence of the Ramanujan–
Petersson conjecture.
There is much more to the theory of modular forms than I can possibly say in
one lecture. Some books that cover modular forms in more detail are [Ser73] and
[DS05].
References
[DS05] F. Diamond and J. Shurman. A first course in modular forms, volume 228 of Graduate
Texts in Mathematics. Springer-Verlag, New York, 2005.
[Ser73] J.-P. Serre. A course in arithmetic. Springer-Verlag, New York-Heidelberg, 1973. Trans-
lated from the French, Graduate Texts in Mathematics, No. 7.
[Shi73] G. Shimura. On modular forms of half integral weight. Ann. of Math. (2), 97:440–481,
1973.

Ramanujan 4

Uploaded by

Copyright:

Available Formats

Ramanujan 4

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ramanujan 4

Uploaded by

Copyright:

Available Formats

20

4. Theta functions and the Ramanujan–Petersson conjecture

M. C. Escher, drawing E55

M. C. Escher, Circle Limit III

But it can be done. Consider the set of transformations of the form

J. Leys, after M. C. Escher, drawing E69

You might also like