Question: Which primes are represented by a given quadratic form of discriminant ?
If , then we can see that the form is equivalent to a form of the form . To observe this, choose a matrix of determinant and apply the coordinate transformation to get a form with coefficient . We can always do that because are coprime (why?) –we just need The coefficient is always represented because , the above argument says that if we have a number that is represented primitively with , then we can transform the form to have to look like . That is the coefficient of are precisely the numbers primitively represented by the form.
Now if is represented, the form is equivalent to and hence we have the discriminant equation .
This means that
On the other hand, if we have a prime satisfying this equation, we can construct and hence a form of discriminant that represents .
In general, any number is properly representable by “some” form of discriminant if and only if is a square modulo If is odd it’s enough to ask for solution to .
But what about a given fixed form? How do we decide if a given number is represented by the form? Let’s assume we are given that is a square modulo . (Note this already cuts down the possibilities to half). By using the solutions to , we can construct one form which represent . If this form is equivalent to the given form, we are done. In general, for an odd prime we have
iff is represented by one of the forms (up to equivalence) of discriminant .
For instance, if the class group is trivial, that is if every two forms of discriminant are equivalent, numbers satisfying these congruence conditions are represented by every form of discriminant . Take the example of . Here the class group is trivial and any form is equivalent to . If we have prime such that , which is true for any prime , we have a form of discriminant . Now because the class group is trivial, this form has to be equivalent to and hence we have a solution to Thus we have the following.
or
Similarly we can get
or
or
because the class groups are trivial. The residues are determined by the property , which define a subgroup of index 2 inside
But the class numbers become large as becomes large. In fact we have
For instance a non-trivial factorization produces a form which is reduced if (This covers most cases except prime powers. The forms and cover the rest of the cases. Also note that the case of odd negative discriminant is much more difficult is the complete list with class number 1)
What about representation by ? Here is the class group has two reduced forms . Hence an odd prime which satisfies latex is represented by one of these two forms and we have
The residue classes are represented by one of the two reduced forms. But we can see that only half of these residues are represented by the principal form . And we have
We said that the number coprime to represented by forms of discriminant have square modulo , that is . This forms a subgroup of index 2 inside . Using the identity
the numbers represented by the principal form are a subgroup. In fact, each of the forms represent a coset of this particular subgroup. Like in the above case is subgroup represented by the principal form and the coset is represented by the other reduced form.
Let us the take case . We now have 4 reduced form
So we have
But the classes represented by are only which form a subgroup of
Let’s say we take a prime from one of these classes what can we say about the forms representing it. For , we had only one reduced form representing those classes, but in general the residues can be represented by the principal form and a set of other reduced forms. This collection of forms is a subgroup of the class group called the “genus”. For , we have
So the forms form a subgroup of class group, and form a genus. And the coset of this subgroup contains which form another genus.
Thus in addition to the condition , if we know the coset of the subgroup of numbers represented by principal form to which belongs, we can deduce that is represented by one of the reduced forms of the coset of a subgroup of the class group called genus.
For instance is represented by the genus of the principal form iff .
So using these congruence conditions, we can determine a genus of forms which could represent .
And It’s not possible to distinguish between the forms of a particular genus by congruence conditions alone! Every form in the class represents all the residue classes modulo . In general there are no other congruence (abelian conditions) that can distinguish primes represented by forms in a genus.
Question: How many genera exist?
Because the squares are represented by principal form, all the squares inside are represented. If is the subgroup represented by the principal genus, we see that every element of , where is the character satisfying is of order 2 modulo . Thus number of genera have to be of size for some
Theorem: If is the number of odd primes dividing , then the number genera is given by where
In general if we also take forms of discriminant , we have
Ambiguous classes: To explain this result, we first need to know the 2-torsion of the class group- that is forms of order at most called the ambiguous classes. A class is ambiguous precisely when it is equivalent to it’s inverse. Using the fact that positive definite forms are reduced only if and that is the inverse of , we see that ambiguous forms precisely corresponds to reduced forms with
and and .
These conditions are enough to see that there are ambiguous classes. For instance in the , and there are exactly factorisation with (each prime can either belong to or -out of which exactly half of them satisfy , so in total forms. Other cases the computations are similar.
Because the 2-torsion is of size , the number of squared-classes if exactly equal to . Look at the square map , whose kernel is the ambiguous classes. Hence the image which is the set of squares has size
Genus characters:
For any prime dividing the discriminant the character on the class group (primitive forms) by
is well defined. In particular we can assume
To see this note that
So the values (co-prime to discriminant) represented by the forms are all quadratic residues or non-residues, and the character is well defined.
Thus we have a list of characters for each odd prime dividing the discriminant.
For these is the complete list of “genus characters”. There are more characters for depending on other classes or . In total there are complete characters.
They are called genus characters precisely because all of the forms in a fixed genus have the same values for the complete list of genus characters. In fact, the principal genus corresponds to the forms that evaluate to on all these characters. But the number of possible value of this complete set of characters is (precisely half of all the possible values since the values modulo only belong to half the residue classes). Hence the number of genera is equal to
It’s easy to see that the classes of squares in the class group have all the genus characters equal to and hence belong to the principal genus. By the previous discussion on ambiguous classes, if the principal genus had more classes than these then the number of genera would be which would be smaller than , which is not true. Thus the principal genus is exactly the set of squares (Duplicated forms). This is the beautiful Gauss’s duplication Theorem!
Question: Can you describe the discriminants for which each genus contains just a single reduced form?
Such discriminants are nicer because we can precisely pin point one reduced form with the congruence conditions.
For example in each of the following cases, there is only class per genus.
From the previous discussion, this can happen only if there are no non-trivial square classes. That is if square of every form is the principal class, equivalently a class is equivalent to it’s inverse. In this case the class number if exactly equal to
Gauss listed 65 discriminants with this property:
(These numbers are first found by Euler- they are called convenient/idoneal numbers- any composite number represented by form has at least 2 solutions)
The list is known to be finite and conjectured to be be just this 65 numbers.
Other ways to define genus: We saw two ways to think of a genus–
- As the set of forms representing the same values in
- As the set of forms which have the same complete characters–same values on the genus characters.
In fact the genus can be seen as the
3. set of forms equivalent modulo any integer
4. set of form equivalent under $, where corresponds to rationals with denominators coprime to a given integer .
5. set of form equivalent under $, where corresponds to rationals with denominators coprime to a the integer .
6. set of form which represent the same values modulo any integer .
For fundamental discriminants, it’s also just the classes equivalent under