NAVA CHITRIK
Referenced heavily from Tate's thesis and from Goldfeld and Hundley (2011), Automorphic Representations and L-Functions for the General Linear Group, Cambridge University Press
Contents
Automorphic Forms for GL(1,AQ ) Dirichlet L-Functions Integration and Poisson Summation on 1 4
A -Function
4 5 7 9
The Zeta-Integral and its Functional Equation The Local Zeta Integral The L-Function and its -Factors
14
Q\A.
They
will turn out to be no more than the classical objects, the Dirichlet characters. First we will make some denitions for automorphic forms which are capable of generalization to
n>1
forms are associated to a unique Dirichlet character. In this way, we present Tate's thesis as the bottom rung in the theory of Automorphic forms. Many of the techniques encountered here will be used again in the theory of Automorphic forms for GL(2, A).
: A Q C such
that
| (x)| = 1
and
(q ) = 1
whenever
q Q.
homomorphism
1 : Q \A Q S .
1
GL(1,AQ )
- TATE'S THESIS
g = {g , g2 , ...},
t R.
be a xed unitary Hecke character. An
: Q \ AQ C
satisfying:
(1) (2)
(zg ) = (z )(g ) z, g AQ
has moderate growth (this is automatic)
While the above denition seems a bit silly- it implies that automorphic forms are simply scaled unitary Hecke characters - this denition is only to mimic the denitions for higher dimension when these functions can be more complicated.
mod q ).
: (Z/q Z)
and
we can lift
to all of
by
(n) = ( n)
where
nn ( mod q )
(n) = 0
if
(q, n) = 1.
be a character mod pf
idelic : Q \AQ
which is called
as follows:
idelic (g ) =
p
p (gp )
where
(g ) =
sign(g ) 1
if (1) = 1 if (1) = 1
and
GL(1,AQ )
- TATE'S THESIS
(v )m v (gv ) = (j )1
if v = p where gv v m Z v if v = p where gv = pk (j + pf Zp )
which a character of a nite
group must lie on the unit circle. The other properties can be checked easily. More generally, any Dirichlet character factors into characters mod case we dene
pf ,
and in this
idelic as
Theorem 6.
where c C, t R and idelic is the idelic lift of a Dirichlet character as dened above. Proof.
As we noted in the denition,
and
is a
of each local
eld. Also from continuity, for all but nitely many primes, the factors must satisfy
v (v ) = 1.
Characters of
|r|it
or
|r|it sign(r)
Q p
is
(u) = 1 Q p:
whenever
is a unit, i.e.
Z p . Then
form on
u pn ,
(u p ) = (p) (p).This
(Z p ) 1.
such that
(1 + pk Zp ) 1
must contain an
open subgroup of this form. We call such a character ramied with conductor Now, for an arbitrary eld element addition to specifying simpler, on
pk .
in
u pn ,
we have that
(p) we must also specify how acts on the group Z p , or even Z/pk Z.
So
k Z p /(1 + p Zp )
GL(1,AQ )
- TATE'S THESIS
character mod
pk .
(p)
(u)
is odd, i.e.
(u) = (u)
e.g.
To make the
We have left out some tedious calculation but it follows that You may wonder where the sign(g ) has gone. other components, indeed we have that
(1) = (1)(1).
Dirichlet L-Functions
Denition 7. (The L-function of an Automorphic form) Since any automorphic
form
L(s, ) =
p
(p) ps
(g ) = 1,
1
we get that
L(s, ) = (s) =
p
1 ps
=
n=1
ns
(s) > 1
via functional equation was solved in the 1800's by means of theta functions and other complicated mechanisms. In his thesis, Tate presented an elegant way to get these continuations using adelic integrals. remainder of these notes. This process will be described in the
1Zp
GL(1,AQ )
- TATE'S THESIS
integral of a function in such a space by using linearity and taking the product of the integral, for purely factorizable functions, i.e. if
( x ) =
p (xp )
then
(x)d x =
A Q p S
Q v
v (xv )d xv
d x =
where the
dx
x
dxv d x = 1
1 1p1
dxp |xp |p
at a nite place
(a + pn Zp ) = pn .
Zp
dxp = 1
and
Z p
Theorem 9.
1 |x|
h x
Proof.
omitted
(The Fourier transform is taken with respect to a self-dual Haar measure i.e. formula
(x) = f (x) f
holds)
-function
(g ) = eg
p<
1Zp
1.
g ), (g ) = (
and
= (0) = (0)
(g )|g |s d x =
A Q
Now, since
ex |x|s1 dx
p< Zp \{0}
|gp |s d gp = ()
Zp \{0} =
n=0
pn Z p
GL(1,AQ )
- TATE'S THESIS
|gp |s d gp =
Zp \{0}
So that
pns d x =
pn Z p
n=0
1 1 ps
() = 2
0
ex xs
() =
Q
(x)|x|s d x =
Q \A Q
Q \A Q
(x)|x| d x +
s
(x)|x|s d x
Q \A Q
|x|1
|x|1
Now we'll apply Poisson summation to the rst of these two integrals
(x)|x|s d x
Q \A
Q \A Q
|x|1
|x|1
Q \A Q
|x|s1 d x + 1
0
(|x|s |x|s1 )d x
p Z p
|x|1
|x|1
Q \A Q
(x) |x|
1s
d x+
(xs xs1 ) dx x
d x
|x|1
Q \A Q
(x) |x|s1 d x +
1 s
1 s1
|x|1
() =
Q \A Q
|x|1
, s 1 s.
Q \A Q = (0, )
p
Z p
GL(1,AQ ) s 2
- TATE'S THESIS
s/2
(s)
we have that
s/2
1s s (s) = 2 2
1s 2
(1 s)
You may have noticed a couple of thing about the above calculations. One major observation is that to get the functional equation we didn't even need to use the fact that on a
so that there is some reason to suspect that we will not need to rely =
function
particular
The remainder of these notes will be devoted to generalizing this process to obtain functional equations for more general Dirichlet L-functions.
Z (s, , ) =
A Q
This denition is by analogy with the Mellin Transform of real functions which is useful in classical analytic number theory.
(x) (x)|x|s d x
Theorem 12.
, Z (s, , ) = Z (1 s, )
Proof.
We compute:
GL(1,AQ )
- TATE'S THESIS
Z (s, , ) =
A Q
(x) (x)|x|s d x
=
Q \A Q
(x) (x)|x|s d x
Q
by strong approximation
Q \A Q =
Q \A Q
||s () = 1
(x) (x)|x|s d x
Q
for
|x| 1 +
(x) (x)|x|s d x
Q
for
|x| 1
= I + II
We we will now use Poisson summation on the rst integral. We rewrite the Poisson summation formula as
(x) =
Q Q
Then integral
II
turns into:
II =
Q \A |x|1 Q
(x) (x)|x|s d x
Q
=
Q \A |x|1 Q
(x)|x| d x + (0)
s
|x|1 Q
(x)|x|s1 d x
|x|1 Q \A Q
Q \A
=
Q \A |x|1 Q
(x)|x| d x + (0)
s
|x|1 Q \A Q
(x)|x|s1 d x
GL(1,AQ )
- TATE'S THESIS
(0.1)
Z (s, , ) = I + II =
Q \A |x|1 Q
(x)|x|s1 d x
(0)
If
Q \A Q = (0, )
is trivial on the
Z p Z p 0
to
and
Zp
then by standard tricks, the last two integrals will be zero. part and the last two pieces
1:
1 1 (0) + (0) s s1
In summary, by inspection of 0.1 and the above discussion, we have that
Z (s, , ) =
, Z (1 s, )
Summary of above proof: (1) Use strong approximation (2) Break into
|x| 1
and
|x| 1
|x| 1
in the big piece of the formula
x 1/x
(5) determine if ramication occurs Although the global zeta integral satises such a neat functional equation, the local pieces (dened below) require a slight modication for their functional equations.
s C
with
a Bruhat-
Schwartz function on the local eld (i.e. Locally constant and compactly supported on
Q p , or a Schwartz-function on R ). Zv (s, , ) =
Let
We dene
(x) (x)|x|s d x
Q v
GL(1,AQ )
- TATE'S THESIS
10
Does the above converge? Well we can use the standard trick of splitting it into two pieces. If we consider the integral when
|x| > 1
there since we are dealing with a (Bruhat-) Schwartz function. hand, for
|x| 1
On the other
Zp
|x| d x =
|x| d x =
0 pn Z p 0
pn
(s) > 0.
So the integral
Zv (s, , )
converges for
(s) > 0.
Theorem 14.
Proof.
and
on
Qv .
We compute:
) = Zv (s, , ) Zv (1 s, ,
By letting
y xy
=
Q p Qp
=
Qp Q p Qp
Since
dx d x = c and the point 0 is measureless |x| =c (x) (z )ev (xyz )|y |1s (y )dx d y dz
Qp Q Qp p
GL(1,AQ )
- TATE'S THESIS
11
and so
, therefore
Zv (s, , ) Zv (1
) = Zv (1 s, , ) ) Zv (1 s, , s, , Zv (s, , ) ) Zv (1 s, ,
which shows that
Zv (s, , ) ) Zv (1 s, ,
(s, )
(s, )
in each case.
All that
Zv (s,, ) is required is a particularly easy test function and the comparison of ) Zv (1s,,
for that easy function whether one on
Z p)
(x) = 1
for
x R.
(x) = ex
. Then,
Zv (s, , ) =
R
ex |x|s
s 2
dx |x|
some
= 2 dx Zv (1 s, ) = ex |x|1s |x|
R
s ( ) 2
u substitution
and recall
(s) =
0
e x x s
dx x
1s 2
1s ) 2 (s, )
v=
and
(x)
|x| x .
(x) = xex
so we
GL(1,AQ )
- TATE'S THESIS
12
have that
Zv (s, , ) =
R
ex x2 |x|s+1
s+1 2
dx |x|
some
s+1 ( ) 2
u substitution
and recall
(s) =
0
ex xs
dx x
= i
(1s)+1 2
1s+1 ) 2
v = p
is unramied.
This is the
easiest case and the one that comes up in proving the functional equation of the Riemann
function.
= 1Z p
transform.
Computing,
Zv (s, , ) =
Q p
1Zp (x)|x|s d x
(x)|x|s d x
=
Zp \{0}
=
n=0
(p)n pns
since
Zp \{0}=
n=0
pn Z p
1 1 (p)p(1s)
GL(1,AQ )
- TATE'S THESIS
13
(x) =
Qp
(y )e2i{xy} dy e2i({x(1y}) dy
pr Zp
This is only nonzero when
x 1 + pr Zp
= 11+pr Zp = pr 11+pr Zp
dy
pr Z
p
Recall that
e2i{x}
Qp
{x}
x,
that is if
x =
N
an pn
then
{x} =
an pn Q
N
GL(1,AQ )
- TATE'S THESIS
14
Zv (s, , ) =
Q p
=
pr Zp \{0}
pr
=
n=0
(p) p
n ns
since
Zp \{0}=
=1
p (j + pr Zp )
j =1 (j,p)=1
pr
e2i{x} (x)|x|s d x
p (j +pr Zp ) 2ij l
p
=
=1 j =1 r p
r
(p) (j ) p
pr s
=
=1 j =1
d x
p (j +pr Zp )
p r = p p1
p (p)
=1
j =1
2ij l
p
(j )
pr+1 rs = p (p)r p1 Zv (1 s, , ) =
Q p
pr
e2ijp (j )
j =1 (j,p)=1
11+pr Zp pr (x)|x|1s d x
pr (x)|x|1s d x d x |x|1 dx
=
1+pr Zp
= pr
1+pr Zp
p pr p1
1+pr Zp
p = p1
The L-Function and its -Factors
Great. So now we have functional equations and meromorphic continuations of the local Zeta-integrals to the entire complex plane. We now use this fact, together with
GL(1,AQ )
- TATE'S THESIS
15
the global-functional equation to obtain a functional equation and root numbers for a classical L-function. Let
Then
v be the unitary Hecke character associated to an automorphic form. v is ramied at only nitely many places. We dene the local L-function of s s ) 2 ( 2 s +1 2 ( s+1 ) 2 (1 p (p)ps )1 1
as follows if if if if
v= v= v=p v=p
is unramied is ramied
Lv (s, v )=
is unramied is ramied
pth
Euler product, so this is a reasonable denition of a local L-function. Also, in the unramied places notice that above examples.
Lv (s, v ) = Zv (s, v , )
with
chosen as in the
Remark
of
19
The local L-function can be thought of as the greatest-common-divisor That is, the local L-
Zv (s,, ) function is the least complicated function so that Lv (s, ) is always holomorphic.
For example, if
is ramied, then
Zv (s, , )
so that
Zv (s, , ) = Lv (s, )
, Zv (1 s, , ) = Lv (1 s, ).
v (s, ) is dened by
(0.2)
Zv (1 s, , ) = Lv (1 s, )
v (s, )
Zv (s, , ) Lv (s, )
was.
which is independent of
(s, )
is unramied at
Zv (1 s, , ) = Lv (1 s, )
Lemma 21.
v (s, )
1.
GL(1,AQ )
- TATE'S THESIS
16
Let us to back to the global functional equation and choose our test function,
=
v
where the and at that
v
At the unramied places
v are
chosen so that
Zv (s, v , ) = Lv (s, ).
we choose v
as in the examples and in the last case, its not hard to show
11+pr Zp
Lv (s, )
We think of the global L-function as a completed L-function. It looks a lot like the Dirichlet L-function, but now we also have a component at the innite place, making it more symmetric and complete. It is this L-function which will have a natural functional equation coming from the adelic factorization.
Theorem 23.
, Z (s, , ) = Z (1 s, )
to the product over
of equation 0.2
v
The
Zv (1 s, , ) = Lv (1 s, )
(s, )
v
Zv (s, , ) Lv (s, )
Zv 's
factor out entirely by the global functional equation and we get that
L(s, ) =
v
where
Lv (s, ) = (s, )
v
Lv (1 s, )
(s, ) =
v
v (s, )
=
v S
GL(1,AQ )
- TATE'S THESIS
17