A factorisation theorem for the coinvariant algebra of a unitary reflection group
G. I. Lehrer

TL;DR
This paper proves a new factorisation theorem for the coinvariant algebra of a complex unitary reflection group, generalising previous results and revealing a structured isomorphism involving harmonic polynomials and reflection subgroups.
Contribution
It introduces a degree-preserving isomorphism between tensor products of harmonic polynomial spaces for a reflection group and its subgroup, extending prior work to complex unitary groups.
Findings
Established a graded isomorphism for harmonic polynomial spaces
Generalised a known result from real to complex reflection groups
Enhanced understanding of the algebraic structure of coinvariant algebras
Abstract
We prove the following theorem. Let be a finite group generated by unitary reflections in a complex Hermitian space and let be any reflection subgroup of . Let be the space of -harmonic polynomials on . There is a degree preserving isomorphism of graded -modules, where and is the space of -fixed points of . This generalises a result of Douglass and Dyer for parabolic subgroups of real reflection groups.
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Algebra and Geometry · Advanced Topics in Algebra · Finite Group Theory Research
A factorisation theorem for the coinvariant algebra of a unitary reflection group
G.I. Lehrer
School of Mathematics and Statistics
University of Sydney
N.S.W. 2006, Australia
Fax: +61 2 9351 4534
Abstract.
We prove the following theorem. Let be a finite group generated by unitary reflections in a complex Hermitian space and let be any reflection subgroup of . Let be the space of -harmonic polynomials on . There is a degree preserving isomorphism of graded -modules, where and is the space of -fixed points of . This generalises a result of Douglass and Dyer for parabolic subgroups of real reflection groups. An application is given to counting rational conjugates of reductive groups over .
Key words and phrases:
Unitary reflection group; coinvariant algebra; reductive group
2010 Mathematics Subject Classification:
Primary 20F55; Secondary 14G05, 20G40, 51F15
1. Background and notation
Much of the background material in this section may be found in [8, Ch. 9]. Let be a finite group generated by (pseudo)reflections in a complex vector space of dimension . It is well known that if denotes the coordinate ring of (identified with the symmetric algebra on the dual ) then acts contragrediently on , and hence on , and the ring of polynomial invariants of on is free; if is a set of homogeneous free generators of , then the degrees () are determined by , and are called the invariant degrees of .
If is the ideal of generated by the elements of which vanish at , then realises the regular representation of . The space is called the coinvariant algebra of . Since the ideal is graded, is clearly graded, and for each , the graded component of degree is a module. By a classical result of Chevalley [8, Cor 3.31, p. 52], taking products of polynomials defines an isomorphism of -modules
[TABLE]
For any -graded -vector space where the are finite dimensional, we write
[TABLE]
for its Poincaré series. If then is a polynomial. For properties of these series see [8, Ch. 4].
1.1. The space of harmonic polynomials
If is any graded -stable complement of in , then evidently as graded -module. The space of -harmonic polynomials is a canonical such complement, which we now describe. Let be the set of reflecting hyperplanes of and for each hyperplane , let be a linear form such that and let , where is the (cyclic) group of reflections in . Define
[TABLE]
Then [8, Lemma 9.9] asserts that is skew, that is, for , , and further if is any skew polynomial, then
[TABLE]
Now let be the symmetric algebra on . For , we have the derivation of defined by
[TABLE]
It is explained in [8, §9.5] how the map extends to an algebra homomorphism from to the algebra of differential operators on . Thus for we have ; for homogeneous , this operator has degree . The properties of this algebra of operators are summarised in the next statement, whose proof may be found in [8, Ch. 9].
Theorem 1.5**.**
Let , , , etc be as above.
- (i)
Let , and . Then . 2. (ii)
Define a bilinear pairing by for , . This pairing is non-degenerate in both variables. 3. (iii)
The pairing in (ii) is respected by the action of ; i.e., for , and , we have .
Theorem 1.5 shows that puts the spaces and in -equivariant duality. Now let be the algebra of invariants of on . Denote by the analogue of for . That is, is the ideal of generated by the elements of with no constant term.
Definition 1.6**.**
Define the space of -harmonic polynomials in by . That is, if and only if for all .
The main facts concerning the space which we shall require are summarised in the next statement.
Theorem 1.7**.**
Maintain the above notation.
- (i)
We have if and only if for all . That is, is the space of functions which are annihilated by the invariant differential operators. 2. (ii)
The space coincides with , where is as defined in (1.3). 3. (iii)
The space is a -stable complement of in . In particular, we have the equivariant decomposition
[TABLE]
2. Reflection subgroups–the main theorem.
A reflection subgroup of the reflection group in is a subgroup of which is generated by some of the reflections in . For background concerning such groups see [4, 12]. They include the parabolic subgroups of [10, 7], but many other subgroups as well. Let be such a subgroup and write for its space of harmonic polynomials , and so on. For any -module , write for its subspace of -fixed elements. Our objective is to prove the following theorem.
Theorem 2.1**.**
Let be a finite group generated by unitary reflections in a complex Hermitian space and let be any reflection subgroup of . Let be the space of -harmonic polynomials on . There is a degree preserving isomorphism
[TABLE]
of graded -modules, where .
Proof.
We begin with the observation that using (1.1) applied to the reflection group acting on , we have the linear isomorphism
[TABLE]
given by multiplying the polynomials in the tensor factors.
Next, applying (1.8), we have the equivariant decomposition
[TABLE]
Since both summands on the right side of (2.3) are stable under and hence a fortiori under , it follows that we have a graded linear isomorphism
[TABLE]
Substituting (2.4) into (2.2) we obtain a linear isomorphism
[TABLE]
Now evidently the summand is mapped in the multiplication isomorphism (2.5) to a subspace of , since is an ideal of . It follows by restricting the map in (2.5) to the summand that we have a surjective degree preserving linear map
[TABLE]
But since realises the regular representation of , we have while evidently , so that the dimension of the left side of (2.6) is . It follows that is a graded isomorphism, and since as graded -modules, it follows that is a graded linear isomorphism.
It remains only to observe that all homomorphisms above evidently respect the action of , and the proof is complete. ∎
Remark 2.7*.*
One of the key ingredients of the proof is (1.1), which asserts that is free as module over . Notice that taking -invariants in (1.1) yields further that is free over . This was first noticed by Dyer [3].
Remark 2.8*.*
The special case of Theorem 2.1 where is a finite Coxeter group and is a parabolic subgroup was treated in [2, Thm. 2.1], where the author acknowledges input from M. Dyer, who pointed out that the result is connected to the discussion of the cohomology of the flag variety in [1]. We believe that our proof is significantly simpler and our result is more general than that in op. cit..
3. Complements
3.1. Poincaré polynomials
Since Poincaré polynomials are multiplicative on tensor products, the following statement is clear. In the statement we use the convention that for any unitary reflection group , (cf. (1.2)).
Corollary 3.1**.**
With notation as in Theorem 2.1, we have
[TABLE]
We can be a little more explicit about the second factor in the right side above. Recall that for any finite dimensional -module , the fake degree
[TABLE]
where is the intertwining number of with the degree graded component of .
Denote by the set of equivalence classes of irreducible -modules. For define integers by
[TABLE]
Proposition 3.2**.**
We have
[TABLE]
where is the fake degree of defined above.
Proof.
For any , . For by Frobenius reciprocity. It follows that .
Multiplying this last relation by and summing over and gives the stated formula. ∎
We give two examples where is a non-parabolic reflection subgroup.
Example 3.4*.*
Let be the group of roots of unity acting on in the obvious way. Let be the subgroup of roots of unity, where divides . Then , , , and , so that is given here by simple multiplication of the basis elements.
Example 3.5*.*
Let be the Weyl group of type . Let have orthonormal basis , with , where is the reflection in and similarly for . Let have dual basis , so that in the above notation and .
Further, we have
[TABLE]
Moreover and .
Using this data, it is straightforward to compute that has a basis
[TABLE]
and that has basis . Evidently has basis
Now consider , given by multiplication of polynomials, followed by projection to . Clearly . Now , where is the projection with respect to the decomposition . But . Since , with the second summand being in , . Similarly and . The group is generated by the interchange of and , and the -equivariance is easily checked.
Remark 3.6*.*
Our main result implies some constraints on which groups could occur as reflection subgroups. Here is one superficial one.
Corollary 3.7**.**
Let the degrees of be and those of be . Then
- (i)
* divides .* 2. (ii)
For each we have .
The relation (i) is evident because and similarly for , and (ii) follows using the fact that , where is the cyclotomic polynomial.
4. Duality.
The main result may be formulated without recourse to the projection by using the dual reflection structure. We briefly indicate how this is done.
The key point is the fact that also acts as a reflection group on . We denote the -analogues of , and for on by , and and write etc for their analogues. All the statements in §1.1 remain true with , and replaced by , and respectively. The following Lemma is an easy consequence of the basic facts outlined in §1.
Lemma 4.1**.**
- (i)
There is a linear isomorphism defined for by . If , then for homogeneous , . 2. (ii)
There is an isomorphism defined, for , by .
The maps and of Lemma 4.1 are evidently -equivariant, and our main result may now be formulated as follows.
Proposition 4.2**.**
The isomorphism of Theorem 2.1 may be explicitly realised as follows. Let and where and are uniquely defined as in Lemma 4.1. Then .
Equivalently, the projection of onto is .
5. An application to reductive groups.
Let be a connected reductive algebraic group defined over the finite field of elements, and let be the corresponding Frobenius endomorphism, as in [6], whose notation we adopt here. For any -stable subset we write for the (finite) set of -fixed points of . Let be an -stable maximally -split maximal torus of and be a Borel subgroup. This data determines the Weyl group , together with its reflection representation in , where is the cocharacter group of and its root system as well as a positive subsystem and its corresponding simple system .
Now take to be any -stable connected reductive subgroup of which has maximal rank. It is well known that such are characterised as the connected centralisers of semisimple elements of . They include Levi components of parabolic subgroups. We shall be concerned with the set of -conjugates of and in particular the set of -stable conjugates of . We may therefore assume, without loss of generality, that . It is always the case that such is the centraliser of an element of , but we shall not use this fact. Let be the root system of with respect to . Then is a positive system in and there is a unique corresponding simple system . Note that unlike in the case treated in [2], it is not generally the case that . Let be the Weyl group of ; this is a reflection subgroup of , but not necessarily a parabolic subgroup. Let . Then it is well known that . The methods of [5] show that has a large reflection component, but is not generally a reflection group. It is also well known that if denotes the connected component of the group ,
[TABLE]
where is as above. Now the conjugate is -stable precisely when . We therefore have a map obtained by taking the image in of in (cf. (5.1)). It is easily checked that this image is uniquely determined up to -conjugacy, where the -conjugate of by is . We denote this image by .
The following facts are standard and may be found, e.g. in [9], [11], [6] or [2].
Lemma 5.2**.**
Maintain the above notation.
- (i)
The map -conjugacy classes of described above induces a bijection from the set of -orbits on to the set of -conjugacy classes of . 2. (ii)
Let and let be the -centraliser of . Then is the semidirect product of with .
Next, recall that acts on as , where fixes both and setwise, since both and are -stable. If is the group generated by and , then , and hence we have an action of on the space of Theorem 2.1. We note also that the conjugacy class in of is of the form , where is an -conjugacy class of .
Let be a function on which is constant on -conjugacy classes. Define the corresponding class function on by .
In view of Lemma 5.2 (i), it is clear that also defines a function on , which is constant on -conjugacy classes. This function is also denoted by .
With this notation we may state the following result.
Theorem 5.3**.**
Let be any -stable connected reductive subgroup of maximal rank of and let etc. be as above.
Let be a function on which is constant on -conjugacy classes. Then
[TABLE]
where is the graded component of , denotes the usual inner product of class functions on , and .
The proof of this result is exactly as in [6], where the result is proved for equal to a torus, and [2] where the result is proved for a Levi factor. The crucial difference is that our Theorem 2.1 was not available for arbitrary reflection subgroups of in [2]. We remark finally that the usual corollaries concerning the number of -stable conjugates of are now available in the wider generality of our result. As an example, we have the following result.
Corollary 5.4**.**
Maintaining the above notation, the number of -stable conjugates of is equal to
[TABLE]
where is the characteristic function of the coset in .
In particular, if is split, and this number is just the Poincaré polynomial
[TABLE]
We close with two examples where is not a Levi factor.
Example 5.6*.*
Take and let be the reductive subgroup of maximal rank corresponding to the unique subsystem of the root system of type . Thus has semisimple part of type . One sees easily that in this case is normal in , so that . Further, and . We may therefore apply (5.5) to conclude that the number of -stable conjugates of is .
Example 5.7*.*
Take and let be a reductive subgroup of maximal rank with semisimple part isomorphic to . Thus the corresponding root subsystem is of type . In this case we have , and , so that . To compute , observe that for any -module and subgroup , we have by Frobenius reciprocity.
In our case, it is easily verified that , where is the two dimensional representation of obtained by pulling back the two dimensional irreducible representation of via the map . Further it is well known and easily verified that
[TABLE]
We may therefore apply (5.5) to deduce that the number of -stable conjugates of is equal to .
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] I.N. Berstein, I.M. Gelfand, and S.I. Gelfand, “Schubert cells and cohomology of the spaces G / P 𝐺 𝑃 G/P , Russian Mathematical Surveys 28 (1973), 1–25.
- 2[2] Douglass, J. Matthew, “A formula for the number of F 𝐹 F -stable Levi factors in a finite reductive group”, Comm. Algebra 22 (1994), no. 13, 5447–5455.
- 3[3] Dyer, M. J., “Modules for the dual nil Hecke ring”, unpublished ms. (1996), available at https://www 3.nd.edu/ dyer/papers/nilhecke.pdf
- 4[4] Dyer, M. J. and Lehrer, G. I., “Reflection subgroups of finite and affine Weyl groups”, Trans. Amer. Math. Soc. 363 (2011), no. 11, 5971–6005.
- 5[5] Howlett, R. B. and Lehrer, G. I., “ Induced cuspidal representations and generalised Hecke rings”, Invent. Math. 58 (1980), no. 1, 37–64.
- 6[6] G.I. Lehrer, “Rational tori, semisimple orbits and the topology of hyperplane complements”, Comment. Math. Helv. 67 (1992), no. 2, 226–251.
- 7[7] G.I. Lehrer, “A new proof of Steinberg’s fixed-point theorem”, Int. Math. Res. Not. 28 (2004),1407–1411.
- 8[8] Lehrer, Gustav I. and Taylor, Donald E. “Unitary reflection groups”, Australian Mathematical Society Lecture Series, 20 , Cambridge University Press, Cambridge, 2009. viii+294 pp. ISBN: 978-0-521-74989-3
