WorksMathematics
Non-Induced Representations in Characteristic Zero
Abstract
This paper explores the failure of induction from all proper subgroups to generate the full representation ring, measured by a quotient group . We extend existing results for cyclic groups to all finite abelian groups and calculate this quotient for all finite groups over fields of characteristic 0 by reducing to the abelian case. We briefly discuss extensions to positive characteristics via modular representation theory. Finally, we detail our result for certain well-known families of groups.
Introduction
We study representations of a finite group via the representation ring , where the elements are are integral combinations of isomorphism classes of -modules with addition and multiplication corresponding to the direct sum and tensor product. That is, for , we have;
Formally, this is the split Grothendieck ring for the representations of the group . This is different to the Grothendieck ring where we would have whenever
is a short exact sequence. The split Grothendieck ring has a more restricted definition as it only accounts for exact sequences that split, and is thus better adapted for studying indecomposable representations. The elements of the representation ring are called virtual representations, and the image of these under the character map to the class group of given by
are called virtual characters.
We seek to understand using the fundamental tool of induction. Induction from a subgroup gives rise to a well-defined additive homomorphism
Following from Corollary 4.3.8(4), the image of this homomorphism is an ideal, and therefore it is natural to consider the quotients of by these images for certain subgroups of . We study objects of the form
for some set of subgroups of . A significant result in this area is Brauer’s Induction Theorem, which asserts that the quotient is 0 provided contains all elementary subgroups of , with elementary being defined as follows:
Definition 1. A group is elementary if it is the direct product of a -group and a cyclic group with order coprime to .
For a thorough introduction to Brauer’s Theorem and other relevant results see chapter 10 in [5] or Section 5.6 in [1]. Moreover, J.A. Green showed in [2] in 1955 that no collection of subgroups smaller than the set of elementary subgroups can be used to make the quotient vanish.
In particular, in this paper we consider the following construction:
Definition 2. We define the ring of non-induced representations to be the quotient ring
where
is known as the full induced ideal of , with
being the ideal induced by .
:::
It measures the failure of the proper subgroups of in inducing the virtual representations of . Put another way, it describes the structure of the virtual representation classes of modulo induction. One use of this is to say if and only if every representation of is additively generated by induced representations.
By Brauer’s Induction Theorem and Green’s result, is elementary if and only if , since we are considering only proper subgroups, and so . In Section 2, we will see this consequence of Brauer’s Theorem in our results, but will not need to appeal to it in our proofs. In Section 3 we will assume this result.
To aid in understanding the representations of , we want to understand the structure of , hoping to relate it to the structure of . In the following sections, we will work over characteristic 0 unless otherwise stated. In Section 2, we will investigate the ring of non-induced representations for the finite abelian groups, extending the existing literature. In Section 3, we will extend this to all finite groups by reducing to the abelian case. Finally, in Section 4, we will provide some extra calculations of for certain families of groups that are particularly interesting, providing alternative family-specific proofs.
I would like to thank Dr Robert Spencer for his advice and Professor Simon Wadsley for his help in verifying and improving this paper.
Determining for Abelian Groups
In this section we will determine for all finite abelian groups.
Cyclic Groups
In 2021, Srihari considered this problem for the cyclic groups in the case of an algebraically closed field of characteristic 0 in [6], in which they proved:
Theorem 3 (Theorem 1.1 in [6]). Let be a cyclic group of order and an algebraically closed field of characteristic 0. Then
where is the cyclotomic polynomial. In particular, the rank of the ring of non-induced representations is .
Later that year, Jolliffe and Spencer extended this result using modular representation theory in [3] to the case of an algebraically closed field of positive characteristic not dividing the order of the group. Much of the notation used in this paper is continued from [3].
Corollary 4 (Corollary 3 in [3]). Let be a cyclic group of order and an algebraically closed field of characteristic such that . Then
In particular, .
Firstly, that it agrees with the characteristic 0 case is expected, but reassures us that we are dealing with a sensible object. Secondly, that it is independent of characteristic suggests we are seeing some information about the group. Therefore, we hope that in this and other cases, the ring of non-induced representations may be a helpful tool in understanding the group and its representations.
Finite Abelian Groups
We will extend the work of the previous section to finite abelian groups via the Structure Theorem.
A finite abelian group can be written as a product of the -primary components over primes . Explicitly, where each is a -group containing all elements of with order a power of . If, for a fixed , we have
we say has -primary component given by . We say that has a non-cyclic -primary component if , and we call the -rank of . We partition the set of finite abelian groups into the following three categories:
The group has no non-cyclic components ( for all primes ).
The group has exactly one non-cyclic component ( for only one prime ).
The group has more than one non-cyclic component ( for more than one prime ).
By the Chinese Remainder Theorem, the first case is a cyclic group, so Section 2.1 applies. For the other cases, we can prove the following result:
Lemma 5. Let be a finite abelian group with non-cyclic -primary component and an algebraically closed field of characteristic 0. Then is an -algebra.
Proof. Let be the character group of the abelian group . The integral group ring has basis elements satisfying . Therefore, there is a ring homomorphism defined as
where is the unique irreducible representation with character , and extending linearly. This is naturally surjective because for abelian groups each representation is a direct sum of characters in , and injective because the irreducible representations give a basis for , so there are no non-trivial relations of them.
Since has non-cyclic -primary component, we know that the -torsion subgroup
contains a subgroup such that .
For any subgroup of order , let
Since , we have , so .
Fix some irreducible character . By Frobenius reciprocity, for each we have
Therefore, the irreducible components of are precisely the characters of extending , namely . Therefore, by considering dimension,
Hence, this sum is 0 in . There are subgroups of order in , which we can view as lines in an -vector space. The identity element must be in all the lines, while every non-identity element of must lie on precisely one line. Above, we demonstrated that each line-sum is 0, so the sum over all lines is also 0 and thus,
However, the same Frobenius reciprocity argument gives
which must vanish in . Combining these results gives . This holds for any irreducible character , and therefore
because the irreducible characters generate . We know is a unital -algebra, and so implies that it is in fact an -algebra.
This result has two immediate corollaries. Firstly, when has more than one non-cyclic -primary component we must have since it cannot be an algebra over for two distinct primes simultaneously. This is expected because is not elementary. Secondly, with exactly one non-cyclic -primary component we see that as an additive group for some .
We next find the full ring presentation of . The remaining non-trivial case is when has exactly one non-cyclic -primary component. To reach this, we first want to consider abelian non-cyclic -groups.
Proposition 6. Let be a non-cyclic abelian -group and an algebraically closed field of characteristic 0. Then
where is the character group of , and the ideal is the image of in by reducing mod .
Proof. By Definition 2, to find , we want to find the full induced ideal . Let and define the annihilator of to be
Let be a 1-dimensional representation of , and an extension of , which exists by the Structure Theorem. By Frobenius reciprocity, we find
and hence
In the ring we define the subgroup-sums
for each subgroup . Then we can rewrite equation (1) as
Therefore, we deduce that the full induced ideal can be generated by these subgroup-sums
So far, this holds for any abelian group since we are only using the fact that all representations are 1-dimensional. We can improve the situation by also using the fact that is a -group. Each non-trivial necessarily contains a subgroup of order , and therefore can be covered by cosets of . The sum of the elements of a coset of within is a multiple of the subgroup-sum . So, from our observation above, can be written as a linear combination of subgroup sums for subgroups of size . Hence,
In Lemma 5 we showed that . Therefore, , and hence
where is the image of in . The isomorphism is induced by the ring isomorphism in the proof of Lemma 5.
Now the task is to make more explicit. This reformulation allows us to make the following reduction.
Lemma 7. Let be an abelian non-cyclic -group with -rank and an algebraically closed field of characteristic 0. Then where is the character group of and .
Proof. Let be generators for each cyclic factor of , with . Then is generated by and each has order . Every subgroup of order in must lie inside . Therefore, every generator of belongs to and the result follows.
So we only need to consider the case of . We will now solve this case explicitly.
Proposition 8. Let for a prime and integer . Then
as -algebras where , and are the generators of each cyclic component of for .
Proof. We know from Lemma 5 that where is the character group. If are generators of each cyclic component of then we can write
which reduces in characteristic to
where . Now, by Proposition 6, we wish to find . Let . We will show that .
For a non-zero vector , define . Then is an order subgroup of , and each subgroup of order can be written in this form. It has subgroup-sum given by
which in characteristic becomes
These subgroup-sums generate since they cover all order subgroups of , and thus
But, by changing variables as before, we have
which implies and thus . Since we found that these forms generate , this tells us that .
For the other direction, we begin by expanding this product:
Define the linear form , and the remainder so that and . Therefore,
Since has degree at least , the lowest degree term of the subgroup-sum is . In other words, .
Next, we will see that the -powers of all linear forms span the set of all homogeneous degree- polynomials in variables. To see why, suppose we have a linear functional on the space of homogeneous polynomials of degree that vanishes on for every linear form . This gives a homogeneous polynomial of degree which vanishes at all . This can only be true if is identically zero. Therefore, the elements do indeed span. This means
But the reverse inclusion is obvious from , so
Then the desired result follows by a simple application of Nakayama’s Lemma.
This is the crucial step in solving the general finite abelian group case, and we are not far from being able to write down a full statement. This particular result is also interesting as it reveals a rich structure.
First, we want to relate back to our original representation theory context. The ideal is naturally the kernel of the augmentation map
and therefore we will call it the augmentation ideal of . The proof of proposition 8 shows that . Hence
Consider the -adic filtration
The associated graded ring is
Since is obtained by quotienting out , its dimension is the sum of the dimensions of the graded pieces. The associated graded ring is
There are no restrictions for degree , so the degree- piece has dimension equal to the number of monomials of total degree in variables, namely . Therefore,
Example 1. Let . Proposition 8 says
and has dimension 3 as an -vector space. Therefore, if and then we have .
To finish dealing with finite abelian groups, we need to extend Proposition 8 to the general non-cyclic abelian -group. We then want to understand how direct products behave in to consider the cyclic factor in case 2.
Lemma 9. Let be a non-cyclic abelian -group with -primary component given by and be an algebraically closed field of characteristic 0. Then
as -algebras.
Proof. By Proposition 6 we know where . Let be generators for the cyclic factors of . Then we change variables to as before. We find
synonymously to the change of variables in Proposition 8.
By Lemma 7 we have where . The group has generators for , and each has order . Using variables , Proposition 8 can be applied to get
Now, in characteristic we have
Then the desired result follows by the result of Lemma 7 as stated above.
We need one final result to be able to combine the results for each -primary components in the second case.
Lemma 10. Let and be groups with coprime order and an algebraically closed field of characteristic 0. Then
as rings.
Proof. Since and have coprime order, each subgroup of can be split
for subgroups , . Also, irreducible characters of are precisely products of irreducible characters from each factor. Hence,
Moreover, induction factors as
Hence the full induced ideal is
Therefore
and the desired result follows.
We have proven everything needed to classify for all finite abelian groups.
Theorem 11. Let be a finite abelian group and an algebraically closed field of characteristic 0. Then
if is cyclic of order we have:
and has rank .
if has exactly one non-cyclic -primary component, we can write for and -primary component being given by , and we have:
and has -dimension
if has more than one non-cyclic -primary component we have:
Determining for a General Finite Group
In this section, we will compute for all finite groups over characteristic 0 by reducing to the abelian case. By Brauer’s Induction Theorem, we know all non-elementary groups have a trivial ring of non-induced representations. Therefore, the only finite groups that were not covered in Section 2 and are not trivial are the elementary groups with non-abelian -primary component. That is, with and a non-abelian -group. To find in this case, we can apply Lemma 10 to decompose the product as long as we can find for any -group .
Lemma 12 (Corollary 5.3.4 of [7]). Let be a -group and an algebraically closed field of characteristic 0. Then each non-linear irreducible representation of is of the form for some and an irreducible linear character of .
From which immediately follows:
Corollary 13. The quotient is generated by classes of linear characters of .
Thus the generators for are the same as for where is the commutator subgroup of . We can then ask whether can see any of the non-abelian structure of . Put another way, do the non-linear representations introduce any other relations that don’t come from the linear representations? Consider the inflation map on representations defined as
where here we are thinking of a representation as a homomorphism
and is the quotient map. Inflation naturally preserves direct sums and tensor products, and hence induces a ring homomorphism
Our question is then whether this map descends to the quotient ring of non-induced representations as an isomorphism. To break this down, we define a map
as Inf composed with the quotient map . Then Corollary 13 asserts that is surjective. Therefore, to answer our question, we need to find the kernel of .
Lemma 14. Let be a -group and be defined as above. Then
with being the full induced ideal as in Definition 2.
Proof. Let be the abelianisation of , so it is an abelian -group, and let
be as above. We can identify with the group of linear characters of . Due to corollary 13, we want to consider the additive projection onto the subgroup generated by the linear characters of
Let and let be an irreducible character of . By Frobenius reciprocity, we have
Therefore, if is not the restriction of a linear character of , then this inner product is 0 for all . This implies
since can have only non-linear irreducible components. So we assume for some linear character of . Then, using Frobenius reciprocity, we deduce
where
is the subgroup-sum defined in the previous section. This agrees with our results in the abelian case, except in that case we would have no non-linear characters.
The linear characters of will factor through , so we are interested in
because then . Therefore, by equation (2) every element of is the inflation of an element of . Hence,
For the converse, let and consider where is the quotient map. Then, by definition, , so
The subgroup-sums generate , so
and hence
Now, let . This means . But is a combination of linear characters, so applying and equation (3) gives
But is injective, so and hence
The converse is clear, since every subgroup-sum in must map to zero in . Therefore,
::: Corollary 15. If is a -group then .
:::
Proof. Apply the First Isomorphism Theorem to and use Lemma 14.
Therefore, we are only interested in the abelian group and so we can apply Section 2.
Theorem 16. Let be a finite group and an algebraically closed field of characteristic . If is not elementary then . If is cyclic of order then
Otherwise, we can write for some prime and non-trivial -group , and some integer such that . Suppose for and . Then
Proof. Corollary 15 reduces the problem to the abelian case. Then we can apply Theorem 11 and Lemma 10 to complete the calculation.
Extra Calculations of
In this section, we compute for two families of groups; symmetric and dihedral. We will find the result that follows from Theorem 16, and then find an elementary proof using our knowledge of the representations in each case.
Symmetric Groups
Here, we will consider the ring of non-induced representations for the symmetric groups for . An elementary group will have since is normal and central. However, for we have . It obviously can never be a -group, so is not elementary for all . Therefore, Theorem 16 in this case says:
Theorem 17. Let be an algebraically closed field of characteristic . Then
for every integer $n \ge 3.
We are able to provide a direct proof of this using some theory for the representations of symmetric groups. To do this, we will use the lecture notes for the course Representation Theory of the Symmetric Groups taught by Stacey Law [4].
Proof. Let be the Young subgroup for a partition . Let act on the cosets of this Young subgroup. If (so ) then its permutation representation is induced from the trivial representation:
and therefore it vanishes in . The standard decomposition of Young permutation modules is:
where is the irreducible Specht module and is a Kostka number. Crucially, and , so if we order by dominance then is unitriangular. Therefore, since the permutation modules vanish, this gives
for every by induction on dominance order.
Now, it suffices to check the case . Notice that for , we have
which translates in to
But we have already shown the second term vanishes. Thus, , and therefore
for .
This result is no surprise for the symmetric groups since it has many subgroups. More to the point:
The problem with the symmetric groups is that there is too much symmetry and not enough group.
G.D. James
Therefore, it was not a stretch to find a simple proof for this case by exploiting the symmetry.
Dihedral Groups
Next, we will consider the ring of non-induced representations for the dihedral groups . It is well-known that
Supposing , we know from § 4.1 that . So, if is odd then implies that, to be elementary, must be a -group, which is not possible. Therefore, must be even. Moreover, the center of a non-trivial -group is non-trivial, which means by Lagrange’s Theorem. Therefore, for to be elementary we must have being a power of so that is a -group. So, in this case, Brauer’s Induction Theorem implies
unless is a power of . In this case, we have
and hence Corollary 15 gives
whenever is a power of .
Therefore, our theorem has given us the full result for the dihedral groups. We will now provide an alternative elementary proof using the representation theory of the dihedral groups.
Theorem 18. Let be the dihedral group of order and be an algebraically closed field of characteristic 0. Then
Proof. The irreducible representations of are:
One-dimensional representations coming from and with .
Two-dimensional representations for given by
where is a primitive -th root of unity. Write characters as .
A first important observation is the following. Consider the linear character of defined by . Then, we have
for each . Therefore, the two-dimensional representations vanish in , so we are only interested in the one-dimensional representations.
If is odd then there are only 2 one-dimensional representations. Namely, and where
In this case, we have
which implies in , and thus .
If is even then there are 4 one-dimensional representations: where
Now, we have
which implies
By reducing these equations we find that all classes are equal in , and . Therefore, or . To decide which is true, we need to determine when .
Suppose is not a power of 2. Write for odd, and consider the proper subgroup . We will show that the only one-dimensional component of is . Notice, and , so none of restrict to on . Thus, by Frobenius reciprocity,
and therefore
for some integers . This implies
Finally, suppose is a power of 2. To establish that we consider the following dimension-mod-2 map:
It is a ring homomorphism by properties of dimension over direct sums and products. Let be a subgroup and a representation of . Then
Since is a 2-group, for every proper subgroup we have that is nonzero and even. Therefore,
for all , and thus
This means that factors through the quotient and reduces to a homomorphism
But we know , so the map is nonzero, and hence the domain is also nonzero. Therefore, we must have .
References
- [1]D. J. Benson. Representations and cohomology volume 1. basic representation theory of finite groups and associative algebras. / D. J. Benson. eng. Cambridge Studies in Advanced Mathematics no. 30. Cambridge: Cambridge University Press, 1998. ISBN: 9780511623615.
- [2]J. A. Green. “On the converse to a theorem of R. Brauer”. In: Mathematical Proceedings of the Cambridge Philosophical Society 51.1 (1955), pp. 237–239. DOI: 10.1017/S0305004100030103.DOI
- [3]Liam Jolliffe and Robert A. Spencer. “Non-Induced Modular Representations of Cyclic Groups”. In: Communications in Algebra 52.6 (2021), pp. 2346–2357. DOI: 10.1080/00927872.2023.2301057.
- [4]Stacey Law. Representation Theory of Symmetric Groups. https://math.berkeley.edu/~ltomczak/notes/Mich2022/RepSn_Notes.pdf. Notes taken by Leonard Tomczak. 2022.
- [5]Jean-Pierre Serre. Linear Representations of Finite Groups. Vol. 42. Graduate Texts in Mathematics. New York: Springer, 1977. DOI: 10.1007/978-1-4684-9458-7.DOI
- [6]Ramanujan Srihari. Non-Induced Representations of Finite Cyclic Groups. 2021. arXiv: arXiv:2110.07987.arxiv.org/abs/2110.07987
- [7]Peter Webb. A Course in Finite Group Representation Theory. Vol. 161. Cambridge Studies in Advanced Mathematics. Cambridge: Cambridge University Press, 2016.DOI