Dirichlet INFORME

DIRICHLETS THEOREM ON PRIMES IN PROGRESSIONS, I
K. Soundararajan
Over the next several lectures we shall develop a proof of Dirichlets famous theorem
on primes in arithmetic progressions.
Dirichlets theorem. If (a, q) = 1 then there are infinitely many primes p with p a
(mod q).
We now describe Dirichlets proof in the case that q = 4. Although in this case we may
find elementary proofs, along the lines of Euclids argument, the proof we now describe
is the one that generalizes to other moduli. Dirichlets proof builds
Pon the ideas behind
Eulers proof of the infinitude of primes (which in fact shows that p 1/p diverges). Let
us now recall Eulers proof.
For a real number s > 1 we considered there the Riemann zeta-function

X 1 Y 1 1 Y 1 1
(1) (s) = s
= 1 + s
+ 2s
+ . . . = 1 s
.
n=1
n p
p p p
p
Note that in the domain s > 1 the sum and the product converge. It may be helpful for
you to recall that the convergence of a product
Y
(1 + an )
n
P
is equivalent to the convergence of n an ; to see this note that log(1+an ) is approximately
an if an is small. Eulers proof is then to note that if there are only finitely many primes
then the RHS of (1) must be bounded as s 1+ . On the other hand we may see that the
LHS of (1) must be unbounded as s 1+ by appealing to the divergence of the harmonic
series. Lets make this last observation a little more precise.
Lemma 1. For real numbers s > 1 we have
1 1
< (s) < + 1.
s1 s1
Proof. If f is a monotone decreasing function on R+ then note that

Z n+1 Z n
f (t)dt f (n) f (t)dt.
n n1
Typeset by AMS-TEX
1
2 K. SOUNDARARAJAN
Therefore
Z n+1 Z
X 1 X 1 dt 1
s
s
dt = s
= ,
n=1
n n=1 n
t 1 t s1
and also
Z n Z
X 1 X 1 X dt dt 1
s
=1+ s
1+ s
=1+ s
=1+ .
n=1
n n=2
n n=2 n1 t 1 t s 1
This proves our Lemma.

For future use, we introduce now the O (Big Oh) notation. We say that a function f (x)
is O(g(x)) as x a if there is a constant C such that as x a (that is in some small
neighborhood of a) we have |f (x)| Cg(x). Sometimes the condition x a will be clear
from the context, in which case well omit it and simply say f (x) = O(g(x)). Examples:
as x we have sin x = O(1), x100 = O(2x ) (why?), log x = O(x0.01 ) etc. As another
example, we could rewrite Lemma 1 as
1
(s) = + O(1),
s1
as s 1+ ; the above statement means that there is some constant C such that as s 1+
we have |(s) 1/(s 1)| C, and Lemma 1 tells us that in fact we could choose C = 1.
Heres a final important example on using O-notation. I claim that for small values of x
(e.g. |x| 1/2) we have
log(1 + x) = x + O(x2 ).
To see this we may use the Taylor series for log(1 + x) getting
x2 x3
log(1 + x) = x + ... ,
2 3
so that for |x| 1/2
x2 |x|3 x2
| log(1 + x) x| + + ... 1 + |x| + |x| + . . . x2
2
2 3 2
which justifies our claim.
Returning to primes, we see from (1) that
X 1 1
log (s) = log 1 s ,
p
p
and using our approximation log(1 + x) = x + O(x2 ) (so that log(1 x) = x + O(x2 )
also), we have
X 1 1
(2) log (s) = s
+ O 2s
.
p
p p
DIRICHLETS THEOREM ON PRIMES IN PROGRESSIONS, I 3
Now for s > 1 we have that

X 1 X 1
X 1
2s
2
= O(1),
p
p p
p n=1
n2
and so (2) gives that for s > 1

X 1
(3) log (s) = + O(1).
p
ps
Now we use Lemma 1 and see that the LHS of (3) is

1 1
log (s) = log + O(1) = log + O(1),
s1 s1
as s 1+ ; make sure you understand the last step, its simple but will help you getting
familiar with the O-notation. We conclude as follows:
Corollary 2. As s 1+ we have
X 1 1
s
= log + O(1).
p
p s1
P
In particular, p 1/p diverges and there are infinitely many primes.
With this in the background, let us proceed to Dirichlets theorem in the case q = 4.
Here we construct a cousin of the Riemann zeta-function namely

1 1 1 1 X 4 (n)
L(s, 4 ) = + + . . . = ,
1s 3s 5s 7s n=1
n s
where
1
if n 1 (mod 4)
4 (n) = 0 if n 0, 2 (mod 4)

1 if n 3 (mod 4).

Note that this series converges (absolutely) in s > 1, and by the alternating series test, it
converges (conditionally) in s > 0. Moreover, just like it can be written as a product
over primes (such products are called Euler products)
Y 4 (p) 4 (p2 ) Y 4 (p) 1
L(s, 4 ) = 1+ s
+ 2s
+ . . . = 1 s
.
p
p p p
p
Notice that the key property responsible for the Euler product is that the 4 is multi-
plicative: 4 (mn) = 4 (m)4 (n) for all m, and n, which reduces the determination of
4 to its behavior on primes. Taking logarithms we find that for s > 1
X 4 (p) X 4 (p) 1
log L(s, 4 ) = log 1 = + O
p
ps p
ps p2s
X 4 (p)
(4) = + O(1),
p
ps
4 K. SOUNDARARAJAN
arguing exactly as in (2) and (3).

What does all this have to do with primes (mod 4)? Note that 4 (p) = 1 if p 1
(mod 4), and 4 (p) = 1 if p 3 (mod 4). Therefore, adding the equations (3) and (4)
we find that
X 1
(5,) log (s) + log L(s, 4 ) = 2 + O(1),
p ps
p1 (mod 4)
and subtracting (4) from (3) we have

X 1
(6) log (s) log L(s, 4 ) = 2 + O(1).
p ps
p3 (mod 4)
Now consider (5) as s 1+ . We already know that log (s) = log(1/(s1))+O(1) .

If we knew that log L(s, 4 ) does not go to as s 1+ we would be able to conclude
that there are infinitely many primes that are 1 (mod 4), and indeed that the sum of their
reciprocals diverges. Similarly if we consider (6) as s 1+ then we find that so long as
log L(s, 4 ) does not go to + we would be able to conclude that there are infinitely
many primes 3 (mod ).
Thus we have boiled our problem down to determining the limiting behavior of L(s, 4 )
as s 1+ . But as s 1+ we see that
1 1 1
L(s, 4 ) 1 + + . . . = L(1, 4 )
3 5 7
We already noted that by the alternating series test the series above converges, and so
L(s, 4 ) certainly stays bounded as s 1+ . This shows that there are infinitely many
primes 3 (mod 4). It remains lastly to explain why L(s, 4 ) doesnt go to zero as
s 1+ (and so log L(s, 4 ) doesnt go to ). There are two ways in which we can see
this: First 1 1 1
L(1, 4 ) = 1 ) + + ...
3 5 7
is evidently positive. Second we may note that L(1, 4 ) = /4 (known to Gregory and
Leibniz) which follows from
Z 1 Z 1 1
2 4 6 dt 1
L(1, 4 ) = (1 t + t t + . . . )dt = = tan (t) = .

2
0 0 1+t 0 4
We have established Dirichlets theorem in the case q = 4!
Corollary 3. There are infinitely many primes p 1 (mod 4), and infinitely many primes
p 3 (mod 4). Moreover, as s 1+ ,
X 1 1 1
s
= log + O(1),
p p 2 s1
p1 (mod 4)
and X 1 1 1
s
= log + O(1).
p p 2 s1
p3 (mod 4)

Dirichlet INFORME

Uploaded by

Copyright:

Available Formats

Dirichlet INFORME

Uploaded by

Document Information

Original Description:

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

Dirichlet INFORME

Uploaded by

Copyright:

Available Formats

DIRICHLETS THEOREM ON PRIMES IN PROGRESSIONS, I

Proof. If f is a monotone decreasing function on R+ then note that

This proves our Lemma.

Now for s > 1 we have that

and so (2) gives that for s > 1

Now we use Lemma 1 and see that the LHS of (3) is

arguing exactly as in (2) and (3).

and subtracting (4) from (3) we have

Now consider (5) as s 1+ . We already know that log (s) = log(1/(s1))+O(1) .

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.