The Main Theorem of Complex Multiplication

Today we’ll state the Main Theorem of Complex Multiplication. We have our standing assumptions. Let be a quadratic imaginary field and the ring of integers. For a prime of define to be the completion at and the ring of integers. For a fractional ideal of define to be the fractional ideal .

We will need a few facts about modules over . We know is a Dedekind domain, so these will easily follow from the sequence of posts about the structure of modules over Dedekind domains. We will denote the -primary part of an -module, , to be . This is just the set of elements of the module annihilated by some power of .

Back in our posts on this topic we only talked about finitely generated modules, but some things are true in more generality. For example, if is a torsion -module, then the natural map is an isomorphism. This is essentially the part of the theorem that says if we attempt to decompose a module into a projective and torsion part, the torsion part has a nice primary decomposition. We also will need that for any fractional the natural map is an isomorphism. This comes from the inclusion .

Putting the above to facts together gives us that . Now let be an idele. The fractional ideal generated by this is . Given any fractional ideal of we define . Since we can define multiplication by on all of piece by piece on the -primary parts .

Let be an elliptic curve with CM by . Fix some . We were calling our surjection from idele class field theory . Choose some with the property that . Let be a fractional ideal and fix a complex analytic isomorphism .

The Main Theorem of Complex Multiplication tells us that with this setup there is a unique isomorphism which makes the following diagram commute:

This gives us the following simple way to translate between the analytic action of multiplying by and the algebraic action of by . The proof involves several reductions to a manageable case, and then a lot of work. As usual, the statements are all we are really interested in right now. The next two posts will be some applications of this, so we might prove a few more things there.

Complex Multiplication 2

I’ll start with thanking Google/Chrome. I sent in a request to fix the CSS rendering issue that happened with the blog and the next day it was fixed (maybe a coincidence?).

We start with our standing assumptions. Let be an elliptic curve with CM by for some quadratic imaginary (with normalized iso ). We will need the following fact. Suppose is defined over a number field . Let be a prime of and the reduction of mod . An element is in the image of the natural map if and only if commutes with every element in the image of the map.

The proof is a straightforward follow your nose type argument by cases. Caution: You must consider the case that is an order in a quaternion algebra because is defined over a field of positive characteristic. Recall that last time we said that is the Hilbert class field of . Thus without loss of generality we may assume from here on that is defined over (because ).

Here is something that seemed bizarre to me the first time I saw it. Let be a prime of degree of and a prime of lying over . Suppose has the property that the natural map is an isogeny of degree ( lies over ) and the reduction is purely inseparable. This happens for all but finitely many . Then there is an isogeny lifting the -th power Frobenius. This should feel funny because we’re lifting the Frobenius map to characteristic , i.e. the following diagram commutes:

The proof is to use that we already know there is some whose reduction is purely inseparable of degree and hence factors as where the second map of degree and hence an automorphism. We’re done if we can check that this automorphism lifts, but we can do this by checking that it commutes with everything in the image of by the above lemma. There’s quite a bit of work checking things element-wise to finish this off.

A special case is the following. For all but finitely many primes such that (i.e. is principal) the is unique such that the endomorphism descends to the -the power Frobenius.

We want to now think about generating abelian extensions of using the torsion points of . Fix a finite map defined over (called a Weber function for ). These are easy to come by. For example, if , then if our model is , then works. Here is another beautiful relation between CM theory and class field theory. Recall are the -torsion points with respect to the normalized map , i.e. the set of such that for all .

For any integral ideal of the field is the ray class field of of conductor . An immediate corollary is that . Hence we get the amazing fact that the maximal abelian extension of can be made by taking and adjoining to the numbers and the -coordinates of the torsion points of the Weierstrass model (as long as ). The proof of this is pretty long and involves all the stuff from earlier in the post, so we’ll skip it.

Last time we said that if we don’t just take -coordinates and instead adjoin all torsion coordinates we get which is an abelian extension of . It is interesting that doing this does not in general produce an abelian extension of . Another interesting corollary to this is that we get a situation in which we can adjoin all torsion points. Given any with class number we get a chain and hence .

Complex Multiplication 1

Today we’ll really get into the class field theory of CM elliptic curves. Let be a quadratic imaginary field. Let be an elliptic curve with the property that , e.g. take . At the end of the last post we constructed a group homomorphism which was defined by the property that .

Now we know that is an abelian group, so in fact factors through (a now relabelled) where is the maximal abelian extension of . For the rest of the post we’ll sketch why a few things are true:

1) It turns out that is the Hilbert class field of .

2) Last time we stated that , but given the previous statement, the proof is now easy from class field theory.

3) If are a complete set of representatives of , then form a complete set of Galois conjugates for .

4) For any prime of , we have . Extending to all fractional ideals we get .

Let . One can check directly that . This shows that is abelian. Let . By class field theory there is some associated conductor . Considering the composition with the Artin map one can see this is just the natural projection map. Now using the injectivity of one can extrapolate that showing is unramified and hence contained in the Hilbert class field.

But class field theory tells us that surjects onto and hence is an isomorphism which shows and hence is the Hilbert class field of . We’ve already discussed the second item. The third just follows from identifying with and using the fact that is a -torsor. The fourth just comes from the fact that consists of all fractional ideals because .

I know, this was pretty skimpy on details, but the point of this series should be to see some of the ideas and results from basic CM theory for elliptic curves and not full blown explanations. The results in this post are really cool in my mind because it takes these very classical purely field theory questions and converts them to geometric questions about elliptic curves and vice-versa. Given a quadratic imaginary , it is easy to cook up an elliptic curve with CM by . If you want to know the maximal unramified abelian extension of , we now know we need only figure out because .

Some Fields of Definition Results

Suppose is an elliptic curve. We will need some standard facts. First, if is a field automorphism, then will denote the elliptic curve formed by acting on the coefficients of a Weierstrass equation by . We have an isomorphism by . If has CM by the full ring of integers in a quadratic imaginary , then . This follows by exploiting the fact that to argue that lies in a finite extension of .

The natural map from isomorphism classes of curves over with CM by to isomorphism classes over is an isomorphism. This follows easily from the fact that . It turns out that if we use our normalized isomorphism , we get the nice relation for all and .

If is defined over and has CM by in , then every endomorphism of is defined over . It is useful to see why this is. Let fix . Then by definition , so using the normalized relation we get that for all . Thus if also fixes , then will be fixed by and hence by Galois descent every endomorphism descends to .

We can use this result to show that if has CM by the full ring of integers , then the class number of . As an example application, without knowing the Weierstrass equation for our curve we can argue that the unique curve with CM by must be defined over . This is because , so the degree of the extension is and hence . Of course, we already knew that because we said last time it is given by . Likewise, we could make the same argument for the curve with CM by where .

Now we get to our first real relation between CM and class field theory. Suppose has CM by . The field generated by the -invariant and the torsion points is an abelian extension of . The proof uses our field of definition result above about the endomorphisms. If , then we can reduce to showing that each is abelian over . Using the standard Galois representation, we see that injects into .

Since we have CM floating around we know every endomorphism is already defined over . Using the fact that the image of commutes with , we get that injects into an abelian group and hence is abelian. This result about being abelian I think is an exercise in Silverman’s first book.

Fix a quadratic imaginary field . Let’s return to the idea that we have a simply transitive action of on by . We also have a nice action of on given by . Fix some . This gives us a map by sending to the unique element with the property that . It turns out that is a group homomorphism that is independent of the choice of . This is actually not so easy to prove, and the underlying philosophy of why it is hard is that the action of is analytic in nature since it acts on the lattice , whereas the Galois action is purely algebraic. To prove the independence of requires translating between the two. Although it is sort of a fun proof, we’ll skip it in order to get to the meat of CM theory more quickly.

Elliptic Curves over C

Today we’ll start exploring the fascinating topic of complex multiplication. I’ll assume familiarity with basic elliptic curve theory. For today we’ll review some of the basic theory when our elliptic curve is defined over . In this case, recall that an elliptic curve has complex multiplication by if is an order in a quadratic imaginary field . A quick reminder about the terminology. Recall that for some lattice . Thus if we take a complex number with the property that , then “multiplication by ” descends to a well-defined map and hence we get the term “complex multiplication” by .

Define a normalized elliptic curve to be a pair where is the isomorphism preserving the invariant differential, i.e. . Now let and let be the ring of integers of . One important thing to be able to do is classify the elliptic curves with complex multiplication by . Given any non-zero ideal we have by our theory of projective modules over Dedekind rings that is a lattice in and hence is an elliptic curve.

Moreover, using the fact that is an ideal, we get that has complex multiplication by . Clearly altering to by a principal ideal will give a homothetic lattice and hence an isomorphic curve which leads to the natural guess that elements of the class group might give the different isomorphism classes of curves with CM by . Let’s denote the set of isomorphism classes of elliptic curves with CM by as . It is easily verified that the natural action by is well-defined and simply transitive.

Thus the number of curves over with CM by is the class number of . As an example we could ask how many elliptic curves have complex multiplication by the Gaussian integers . Since the class number is , there is a unique elliptic curve over with complex multiplication by . This curve is well-known to have Weierstrass equation , but from our above theory is also constructed as .

Recall that is the group of -torsion points. Given an ideal of we could also form with respect to the normalized . Now since for any lattice we have we get an isogenty . One can check that the kernel of this isogeny is exactly and that is a free -module of rank (less easy).

This allows us to compute that the degree of the isogeny is and for any the map multiplication by is an isogeny of degree . This allows us to recover the classical fact that multiplication by is a degree isogeny (even an exercise in Hartshorne IV). One interesting thing about CM is that there are endomorphisms that have degree other than a square. Going back to our example we see that multiplication by on has degree since .

This was a short review and introduction. Next time we’ll move onto some more subtle issues that arise when we don’t work over .

Local Class Field Theory

Today will probably be our last class field theory post. I want to end with a brief description of local class field theory. Let be a global field, and a place. We have our standard inclusion by putting the element in the component and ‘s everywhere else. Suppose is abelian. We have the Artin map .

What we learned two posts ago is that the image upon composing the maps is exactly the decomposition group where lies over . The image of the units under the map is the inertia group . This gives us two exact sequences that fit together:

This gives us a local Artin map . The main theorem of local class field theory is that the map is surjective with kernel and moreover the map can be defined independently of localizing the global fields. Just like global class field theory there is an “existence” part of the theorem as well.
This part says that every finite abelian extension of local fields arises as the localization of an extension of global fields and the local Artin maps give a bijection between

where the correspondence is .
Let’s do the simplest example. Let’s think about the quadratic extensions by looking at the bijection. The standard construction is that are the quadratic extensions where is not a square, and is the same extension if is a square. Thus there is a nice bijection between the quadratic extensions and the non-trivial elements of .

Local class field theory tells us that the quadradic extensions of are in bijection with the open index subgroups via . It turns out that any index subgroup at all must be open because it will contain .

Now it will be useful to think in different terms, which is actually a more standard modern reformulation of class field theory since it generalizes to “higher dimensions.” There is a bijection between the open index subgroups of and surjective characters just by taking the kernel of the character. Thus we have reformulated the problem of counting these subgroups to counting characters.

The unit groups have the form (if odd) and where is thought of via the Teichmuller lift. If has order , then it is trivial on .

Thus we get the result that factors through which is a finite abelian -group of order . Thus the number of non-trivial characters is . This gives us a nice alternate description to the classical Kummer description. It tells us there are exactly quadratic extensions up to isomorphism. These aren’t hard to figure out explicitly either. Just take some which is not a square. The three quadratic extensions are , , and . A similar description can be computed when , but you get in that case.

That is all for class field theory for now. We’ll move on to complex multiplication, and I think we’ve done enough of the basics that we can probably do what we need as we need it now.

Examples with the ideles

Last time was a bit of a mess, so let’s review what happened using some examples. We take and . The conductor is (assume for simplicity that odd). We saw this before when discussing the Artin map because now where the map is .

For the new stuff let’s think about the idele class group . Let be a finite place associated to . This means there is some number . Suppose is a place in over . This is ramified, so we’ll get something interesting for the inertia. There is a natural embedding given by (where the non is at the spot).

There is also the Artin map that factors through the class group . The composite is the map . This is picking out the inertia group. We get that .

Let be a uniformizer at . Then is an element of the Frobenius coset of in (the decomposition group). We could pick the . Now since we could multiply through by and not change the class in we can look at . This shows the Frobenius coset can have anything in the spot, but must have times something in all other spots.

Now let’s think about in general. It is pretty quick to check that and hence and hence since we get that . You can follow through the same maps as above to get an element-wise description.

Slightly more fun, we can try again with the function field case . The set of places here is where . To get a handle on the ideles, we can look at the degree map. We have by . This gives a sequence .

Strangely, the map is not surjective like the number field case. This is because we now produce a diagram

where the right vertical arrow is just . You get the Artin map is an injection, but not surjective. You get a dense subgroup, but miss the profinite powers. Now we get a very similar decomposition of the ideles: .

We can calculate some ray class groups using this stuff. Let be a conductor for . Note this is nothing more than an effective divisor say where . We have a the map given by (the alpha coming from the approximation theorem argument). We can compute exactly the kernel of this, so . With a bit of work we can put all this together to get the following results.

If we assume that , then we put together the sequence . In the other case , then we must divide out by the image of the scalars . Thus all this global idele class field theory tells us something about our ray class groups which in turn can be used to tell us about the abelian extensions of our field. Next time we’ll try to wrap this stuff up with some local class field theory.

Idele Class Field Theory 1

Today we’ll start reformulating class field theory in the idelic setting. Let be a global field (a number field, or a finite separable extension of ). Let be all the places and the normalized absolute value at . Define be to be the completion of at . Let the ring of integers if is non-archimedean and just itself otherwise.

We’re going to assume that a first algebraic number theory course covers some of the basics of the ideles. Define to be the ideles of . Recall this is by definition . This is a multiplicative group, but moreover there is a natural topology on it from the restricted direct product. We’ll recall properties of this as needed. We should quickly remind the reader that this is not the same as the product topology. It is the topology generated by the local basis around of the form where you take finitely many components to be open sets and the rest must be .

The ring of adeles is exactly analogous. We take . This is a locally compact topological ring. By construction as a set. But the subspace topology is weaker than the topology we give to . You do recover the right topology by taking the subspace of the product by .

We define now the idele class group to be . Fix a conductor . Our first theorem is: there is a surjection where is constructed as follows. Given use the approximation theorem to find a non-zero with the property that if and where ranges through for . Define to be the fractional ideal such that for all non-archimedean .

Note that if we do this, at very least we’ve constructed a fractional ideal that is relatively prime to . Thus we have a candidate map. We should check well-definedness, but that is more tedious than it’s worth. Let’s just take the theorem as true. Out of it we get the nice corollary that there is a surjection where is the maximal abelian extension of unramified outside . The equality comes from piecing together the Artin maps using class field theory. The kernel is the connected component of the identity of . This corollary is by no means obvious from the theorem.

The fact that the form a compatible system to get a map to the inverse limit is easy because from the construction of the that we built, we can just use the same for any compatible system. From a few posts ago we worked out what the kernel of the natural surjection was. Call that kernel . To figure out the kernel of let’s consider the diagram:

Now notice that going around the square is just taking and sending it to the ideal we got out of this , but then pushing forward to we see that the is changing things by some principal and hence is the map that builds the ideal from alone. So we want to know the kernel and hence when is making the ideal formed from looking at principal.

Suppose maps to , then there is some so that for all . Thus maps to , the so-called unit ideles. Since this map factors through the idele class group , we haven’t changed where goes by altering it to . This shows the part of the diagram except for the surjection. To finish the proof of this corollary just involves messing with the ord conditions and using the Mittag-Leffler condition, so we’ll omit it. It is the statements in class field theory together with some motivating examples that are of interest for this blog at this time.

Let’s wrap up today by stating the big theorem. Suppose is finite abelian with conductor and . We get maps . Call the composite . Then

1)

2) For all places of we get an inclusion by and and where is the decomposition group and is any place of over .

Now this gives us an early glimpse at why the idele formulation is going to give us more. We’ll get some control on inertia and decomposition groups.

Functoriality of the Artin Map

Instead of restating the functoriality of the Artin map, let’s just review the statement through an example. We’ll re-use our example from last time. Let be the splitting field of over . We get a non-abelian Galois group (to keep notation the same, we called this last time). Take the quadratic subextension . We have an abelian Galois group . We need the abelianization .

By Galois theory we know gives us a field extension sitting between and . Class field theory tells us that the conductor because we must pick up all ramification from . The general argument we gave a few posts ago shows us that . Now is the Hilbert class field, so via the Artin map. We take with no embeddings specified.

This gives us the diagram:

First, the right vertical arrow is clearly the zero map. The other important part of the diagram is that the norm map is taking a fractional ideal (class) that is relatively prime to and taking the norm of it which lands you in the units . Moreover, the map is the unique surjective one and the image of the norm map must land in the kernel of this by exactness. Interestingly, this tells us that the positive generator of for any prime to is a square mod 23.

Let’s wrap up today by stating another functoriality result. Given the same setup of where is abelian and is finite possibly non-abelian. Suppose now that and Galois with . Now acts on as follows. Let . Choose a lift . The action is given by . Call this action .

We can transfer this Galois action to the ray class group as follows:

where the vertical arrow is just the natural map on ideals. Commutativity of the diagram just comes from the standard fact that . There is another functoriality we could do, but it doesn’t seem worth it at this point because it is overly complicated and there isn’t a plan to use it anytime soon.

Class Field Theory 2

Today we’ll start by sketching an example of a Hilbert class field. Let be the splitting field of over . Thus where are the roots of the polynomial. Standard Galois theory shows us that . I know, we assumed abelian extensions last time, but we aren’t interested in this extension. It turns out that one can also argue that we have a subextension .

To see how useful this class field theory machinery is, we can check with some basic algebraic number theory (involves the Minkowski bound) that . But above we said that . Thus by the main theorem last time we see that must be the Hilbert class field of , and hence is the maximal unramified abelian extension of . That’s pretty interesting that it is so small.

This gives us an idea of what these Hilbert class fields look like. Note also that class field theory tells us that if the class number is , then there are no unramified abelian extensions, and hence fields like or have Hilbert class field equal to themselves. The word “maximal” makes these things sound big, but being unramified is pretty strict and we see that in order to have large Hilbert class fields the class group needs to be large.

Let’s try to go the other direction now. Recall a few posts ago we calculated that when we include the embedding in the conductor. We have an isomorphism , where is a primitive -th root of unity. We’ll suggestively call the isomorphism where . This comes from standard theory of cyclotomic fields.

We have that the ring of integers is . But now on the primes, we see that is just the Frobenius map of raising to the . This shows us that our standard isomorphism is actually the Artin map and hence class field theory tells us that the ray class field of of conductor is just the cyclotomic extension .

This is an incredibly important example, because now let’s apply the Galois correspondence. Let be any abelian extension at all. We know that corresponds to some conductor . This must be of the form or . Since the first one divides the second, we get two surjections . But the first term of this by the above is that . Therefore, . Out of class field theory we get the classical result called the Kronecker-Weber theorem that any abelian extension of must be contained inside a cyclotomic field.

To finish today let’s talk about one more topic. Suppose we have our abelian extension of number fields with Galois group . If we have a further subextension , then we get is possibly non-abelian. But we can take the abelianization and this corresponds via Galois theory to a maximal abelian subextension say between and , i.e. . By the universal property of the abelianization, we have a map since .

The point is that corresponds to an abelian extension and so class field theory over tells us something about it, and corresponds to an abelian extension , so class field theory over tells us something about it. We just produced a natural map , and hence there should be a corresponding statement in class field theory.

Here’s the theorem. Let so that divides and is unramified outside the places of under . Then we get a commutative diagram:

We could call this a functoriality property of the Artin map. The loose description of this is that inclusions of Galois groups go to the corresponding norm map. We’ll pick up here next time.