Consider the sum
We want to prove lower bound on
Consider the sieve weights
which have the property
The contribution from prime powers is small and we get
On the other hand opening up the sum over the divisors and switching sums we get,
So we have
The contribution from is , therefore
and partial summation with some smooth functions, we get the following
(The factor is really important here, it cannot be improved with basic sieve estimates as above.)
Writing
and plugging in the above expression for , we get
So we have
Till now we did some elementary manipulations and estimates, and these equations are consistent with having primes concentrated in a few residue classes. In fact, as we will see later we have to somehow argue that they don’t all lie in the set for the quadratic character. So the main problem of vanishing is still not really addressed. It’s equivalent to showing lower bounds on , that is we don’t want to behave like for most primes.
Selberg achieves it uses the following lemma.
Main lemma: For every real, non-principal character , for large enough , we have
There has to be some Class Number Formula kind of ideas to related the and the arithmetic of (or binary quadratic forms of discriminant ) (In fact, all such ideas relate the information about point counting/size/archimedean behaviour of the forms to their multiplicative behaviour ie how they behave modulo primes.
Selberg consider the product
The product excludes term with . It’s easy to see that . This is the archimedean part.
On the other hand, the contribution of a prime (in factorization) to the product depends on whether or
If , we need and to be divisible by and the exponent is exponent. On the other hand for , each solution satisfies for any fixed solution. Thus we get for the contributions to the exponent of in this case.
Putting both the cases together we get,
and comparing with the lower bound, we attain
Finishing the proof:
Thus we have is large for a a lot of residues because we get all primes by adding over all the residues . It’s large on average over residues with from the main lemma.
We want to say is large. If it were small, then using , we get that there is a pair such that is large, and in fact each of the quantities is large because by we know neither of them can be too large alone. Using this information with can construct such that with large.
(Proof: If there are more than which are large, we can see that at least one pair of and should be equal. If there are exactly number of large, forms a subgroup of index 2 inside and is either for all or for all . The main lemma produced with and large, hence we see that all of the satisfy , and is large precisely for the subgroup of such that .This show that also lies in that subgroup and hence works)
Once we have using we get
is large.