Real classes of finite special unitary groups

Amanda A. Schaeffer Fry; C. Ryan Vinroot

doi:10.1515/jgth-2016-0009

Article Publicly Available

Real classes of finite special unitary groups

Amanda A. Schaeffer Fry and C. Ryan Vinroot

Published/Copyright: March 10, 2016

Published by

Become an author with De Gruyter Brill

Submit Manuscript Author Information Explore this Subject

From the journal Journal of Group Theory Volume 19 Issue 5

Abstract

We classify all real and strongly real classes of the finite special unitary group SUn⁢(q). Unless q≡3⁢(mod⁡4) and n|4, the classification of real classes is similar to that of the finite special linear group SLn⁢(q). We relate strong reality in SUn⁢(q) to strong reality in the finite special orthogonal groups SOn±⁢(q), and we classify real and strongly real classes in the last case for the group SOn±⁢(q), when q is odd and n≡2⁢(mod⁡4).

1 Introduction

An element g in a group G is real in G if there is an element h∈G such that h⁢g⁢h-1=g-1, and such an element h will be called a reversing element. If the reversing element h further satisfies h2=1, then we say that the element g is strongly real in G (this includes the possibility that h=1). We remark that an element of order two is always real and strongly real, so we may restrict our investigation in what follows to g∈G with g2≠1. Further, an element g∈G is real or strongly real if and only if each element in its G-conjugacy class is as well, and so we may speak of real and strongly real classes of G.

In the case that G is a finite group, the number of real classes of G is equal to the number of complex irreducible representations of G which have real-valued characters. Brauer [1, Problem 14] asked for a group-theoretical characterization for the number of complex irreducible representations of G which can be realized over the field of real numbers. While there is still no satisfactory solution to this problem, there are many examples which point to the notion of strong reality being an important concept in an eventual solution, see [10] for example. This apparent connection has spurred the classification of real and strongly real classes in various families of finite groups.

For example, Vdovin and Gal’t [19] finished the final cases in the proof that if G is a finite simple group such that all of the elements of G are real, then all of the elements of G are strongly real. Tiep and Zalesski [18] have classified all finite simple and quasi-simple groups with the property that all of their classes are real. Gill and Singh [9, 8] have classified all of the real and strongly real classes of the finite special and projective linear groups, as well as all quasi-simple covers of the finite projective special linear groups. Gates, Singh, and the second-named author of this paper [7] classified the strongly real classes of the finite unitary group.

In this paper, we classify all of the real and strongly real classes of the finite special unitary groups. Apart from continuing the program already in place for such classifications, we are also motivated by a result of Gow [11] which gives a natural bijection between the real classes of the finite general linear groups and the finite unitary groups. Since the classification of real classes of the finite special linear groups has been completed, it is a natural question to ask what happens with the bijection described by Gow when considering the real classes of the finite special linear and special unitary groups. As we see, in most cases the description of real classes of the finite special unitary groups SUn⁢(q) is essentially the same as the finite special linear groups SLn⁢(q), but in the case that q≡3⁢(mod⁡4) and 4|n, the description is quite different.

This paper is organized as follows. In Section 2, notation and previous results on real and strongly real classes for the relevant groups are given. In Section 3, we discuss the Jordan decomposition of elements, and the description of centralizers of unipotent elements, which is key to our main arguments. We prove some initial results in Section 4, several of which are direct adaptations from [9]. In Section 5, we classify the real unipotent classes in the finite special unitary groups, and in Section 6 we prove our main result in Theorem 6.3 where we classify all real elements in these groups. Finally, in Section 7, we classify the strongly real elements in the finite special unitary groups in Theorem 7.1, and relate this classification to strong reality in the special orthogonal groups in Corollary 7.3.

2 Preliminaries and notation

For any positive integer n, we let (n)2 denote the 2-part of n, that is, the largest power of 2 which divides n, and we let (n)2′=n/(n)2 denote the odd part of n, or the largest odd divisor of n.

2.1 The groups

Let 𝔽q be a finite field with q elements of characteristic p, and let 𝔽¯q be a fixed algebraic closure. Let GLn denote the group of n-by-n invertible matrices over 𝔽¯q, and let F denote the standard Frobenius morphism of GLn defined by

F⁢((ai⁢j))=(ai⁢jq).

We denote the group of F-fixed points of GLn, the finite general linear group, by

GLn⁢(q)=GLnF.

The finite special linear group SLn⁢(q) is then defined as the set of determinant 1 elements of GLn⁢(q).

Now let η be some invertible n-by-n Hermitian matrix over 𝔽q2, so that if η=(ηi⁢j) then η⊤=(ηj⁢i)=(ηi⁢jq). Define another Frobenius morphism Fη of GLn (dependent on η) by

Fη⁢(g)=η-1⁢(F⊤⁢(g)-1)⁢η,

which is the standard Frobenius twisted by a graph automorphism of GLn. Define the unitary group GUn⁢(q) to be

GUn⁢(q)=GLnFη,

and the special unitary group SUn⁢(q) to be the set of determinant 1 elements of GUn⁢(q).

It will often be convenient to vary the choice of the Hermitian matrix η. While this choice changes the matrices which are elements of GUn⁢(q) or SUn⁢(q), any two groups obtained by varying η are isomorphic, and are in fact conjugate subgroups of GLn. It will also be useful to consider the fact that we have natural embeddings GUn⁢(q)⊂GLn⁢(q2) and SUn⁢(q)⊂SLn⁢(q2).

We identify GL1 with 𝔽¯q×, so GL1⁢(q) is identified with 𝔽q×, and GU1⁢(q) is identified with a cyclic subgroup of order q+1 inside of 𝔽q2×. We let Cq+1 denote this cyclic subgroup of order q+1.

For odd q we let On±⁢(q) denote the split (+) or non-split (-) orthogonal group over the field 𝔽q. That is, On±⁢(q) is the subgroup of elements of GLn⁢(q) which stabilize a non-degenerate (split or non-split) symmetric form on 𝔽qn, and when n is odd we have On⁢(q)=On+⁢(q)=On-⁢(q). We let SOn±⁢(q) denote the special orthogonal group, or the subgroup of determinant 1 elements of On±⁢(q).

2.2 Conjugacy classes

We let 𝒫 denote the set of all integer partitions. If μ∈𝒫, then we either write the parts of μ as μ=(μ1,μ2,…) with μi≥μi+1, or μ=(1m1⁢2m2⁢⋯), where mk is the multiplicity of the part k in μ, and we also write mk=mk⁢(μ). We write |μ| as the size of μ, so |μ|=∑iμi=∑kk⁢mk⁢(μ). The only partition of size zero is written as ∅, the empty partition.

The conjugacy classes of GLn⁢(q) are determined by elementary divisors. In particular, if ℐ is the set of non-constant monic irreducible polynomials in 𝔽q⁢[t] with nonzero constant, then conjugacy classes of GLn⁢(q) are in bijection with functions 𝝁:ℐ→𝒫 which satisfy

∑f∈ℐdeg⁢(f)⁢|𝝁⁢(f)|=n,

where |𝝁⁢(f)| is the size of the partition 𝝁⁢(f). If the partition 𝝁⁢(f) has parts 𝝁⁢(f)=(𝝁⁢(f)1,𝝁⁢(f)2,…), then the conjugacy class of GLn⁢(q) corresponding to 𝝁 has elementary divisors f𝝁⁢(f)1,f𝝁⁢(f)2,…. The unipotent classes of GLn⁢(q) are thus given by functions 𝝁 such that 𝝁⁢(f)=∅ unless f⁢(t)=t-1. In this way, we may parameterize unipotent classes of GLn⁢(q) by partitions μ of n. The semisimple classes of GLn⁢(q) then correspond to those functions 𝝁 such that for each f∈ℐ, 𝝁⁢(f) has parts that are at most size 1, so 𝝁⁢(f)=(1|𝝁⁢(f)|). This is equivalent to an element in the conjugacy class over 𝔽¯q having all Jordan blocks of size 1.

Note that for any irreducible polynomial f∈ℐ, there is an element α∈𝔽¯q× such that

f⁢(t)=(t-α)⁢(t-αq)⁢⋯⁢(t-αqd-1),

where d=deg⁢(f). In this way, the elements of ℐ are in bijection with the F-orbits of the elements of 𝔽¯q×. We may also consider the Fη-orbits of 𝔽¯q×, where

Fη⁢(α)=α-q.

For each such orbit, we associate a polynomial f,

f⁢(t)=(t-α)⁢(t-α-q)⁢⋯⁢(t-α(-q)d-1),

where d is the size of the Fη-orbit of α, and is also the degree of f. We denote this set of polynomials by 𝒰. This set of polynomials takes the place of ℐ in description of conjugacy classes of GUn⁢(q), as determined by Ennola [5] and Wall [20]. We in fact have 𝒰⊂𝔽q2⁢[t], and another way of describing 𝒰 is the set of monic non-constant polynomials with nonzero constant in 𝔽q2⁢[t] which satisfy f=f✓, where if f has roots α1,…,αd in 𝔽¯q×, then f✓ is the polynomial with roots α1-q,…,αd-q, and such that f has no proper factors in 𝔽q2⁢[t] with the same property. Then f∈𝒰 is also irreducible in 𝔽q2⁢[t] if and only if d=deg⁢(f) is odd, and every root α of f satisfies αqd+1=1. Otherwise, f=g⁢g✓, where g is irreducible in 𝔽q2⁢[t].

The conjugacy classes of the group GUn⁢(q) may be parameterized by functions 𝝁:𝒰→𝒫 which satisfy

∑f∈𝒰deg⁢(f)⁢|𝝁⁢(f)|=n.

In this case, we take f𝝁⁢(f)1,f𝝁⁢(f)2,… to be the elementary divisors of elements of GUn⁢(q) in the class corresponding to 𝝁. Like in GLn⁢(q), the unipotent classes of GUn⁢(q) may be parameterized by partitions μ of n, and the semisimple classes of GUn⁢(q) correspond to those 𝝁 such that 𝝁⁢(f)=(1|𝝁⁢(f)|) for each f∈𝒰.

If 𝝁 parameterizes a conjugacy class of GLn⁢(q) or GUn⁢(q) as above, we let 𝒞𝝁 denote this class. Given the class 𝒞𝝁, for each positive integer i, define the polynomial wi⁢(t) by

wi⁢(t)=∏ff⁢(t)mi⁢(𝝁⁢(f)),

where the product is over either ℐ or 𝒰, in the cases GLn⁢(q) or GUn⁢(q), respectively. That is, wi⁢(t) is the product, over all f, with a factor of f for each time the integer i is the part of the partition 𝝁⁢(f). Then let

ni=deg⁢(wi⁢(t))=∑fdeg⁢(f)⁢mi⁢(𝝁⁢(f)).

We define the type of the class 𝒞𝝁, following Macdonald [14], to be the partition ν of n such that ni=mi⁢(ν), so ν=(1n1⁢2n2⁢⋯).

The conjugacy classes of SLn⁢(q) and SUn⁢(q) were described (and enumerated) by Macdonald [14] (see also [2]). Given a class 𝒞𝝁 of GLn⁢(q) or GUn⁢(q), then it follows from the definition that 𝒞𝝁 is contained in SLn⁢(q) or SUn⁢(q) exactly when the polynomial ∏f,if𝝁⁢(f)i has constant term (-1)n, so that the corresponding elements have determinant 1. The question is then how such a class 𝒞𝝁 splits in GLn⁢(q) or GUn⁢(q). Macdonald [14, Section 3] showed that if 𝒞𝝁 is a GLn⁢(q)-class contained in SLn⁢(q) which is of type ν=(ν1,ν2,…,νs), then 𝒞𝝁 splits into exactly gcd⁡(q-1,ν1,…,νs) classes in SLn⁢(q).

The result for SUn⁢(q) is similar, and is implied in [14, Section 6]. If 𝒞𝝁 is a GUn⁢(q)-class of type ν=(ν1,ν2,…,νs) which is contained in SUn⁢(q), then 𝒞𝝁 splits into exactly gcd⁡(q+1,ν1,…,νs) classes in SUn⁢(q). This follows by applying [14, (3.2)] in a proof analogous to [14, (3.3)], where in [14, (3,3), equation (5)], in the case SUn⁢(q), we have π⁢(x,y)=π⁢(x,y-q) with y1-q∈Cq+1, so that 𝔽q× is replaced by the subgroup Cq+1 of 𝔽q2×. The resulting number is also the index of CGUn⁢(q)⁢(g)⋅SUn⁢(q) in GUn⁢(q) for g∈𝒞𝝁.

2.3 Real and strongly real classes

In this subsection we discuss some of the known results in the classification of real and strongly real classes in finite linear and unitary groups.

For any monic polynomial f∈𝔽¯q⁢[t] with nonzero constant, if f has roots α1,…,αd∈𝔽¯q with multiplicities, so f⁢(t)=∏i=1d(t-αi), then we define the monic polynomial with nonzero constant f~∈𝔽¯q⁢[t] by f~⁢(t)=∏i=1d(t-αi-1). If f=f~, we say that f is self-reciprocal.

If f∈ℐ or 𝒰, it follows that f~∈ℐ or 𝒰, respectively. If 𝒞𝝁 is a conjugacy class in GLn⁢(q) or GUn⁢(q), then 𝒞𝝁 is a real class in that group if and only if 𝝁⁢(f)=𝝁⁢(f~) for every f∈ℐ or 𝒰, respectively. This follows from considering the elementary divisor theory, and is discussed in [11], for example. An element f∈𝒰 which is self-reciprocal, other than t±1, must be of even degree. Such elements turn out to be elements of ℐ as well. If f∈𝒰 is not self-reciprocal, then there is a unique pair f1,f1~∈ℐ such that f⁢f~=f1⁢f1~, which follows by considering the corresponding sets of roots as Frobenius orbits. These facts give a natural bijection between the real classes of GLn⁢(q) and the real classes of GUn⁢(q), first described by Gow [11].

It is known that every real class of GLn⁢(q) is also strongly real, which was first shown for odd q by Wonenburger [21] and in general by Djoković [4] (these statements are also proved independently in [9]). This statement is far from true in the group GUn⁢(q), as given by the following main result of [7].

Theorem 2.1

Theorem 2.1 (Gates, Singh, and Vinroot)

Suppose that q is odd and g∈GUn⁢(q) is a real element of GUn⁢(q) in the class C𝛍. Then g is strongly real in GUn⁢(q) if and only if every elementary divisor of the form (t±1)2⁢k of g has even multiplicity. That is, the real class C𝛍 of GUn⁢(q) is strongly real if and only if when f⁢(t)=t±1, then m2⁢k⁢(𝛍⁢(f)) is always even. Equivalently, g∈GUn⁢(q) is strongly real in GUn⁢(q) if and only if g is an element of some embedded orthogonal group On±⁢(q).

It is a natural question to ask which real or strongly real elements of GLn⁢(q) or GUn⁢(q) that are contained in SLn⁢(q) or SUn⁢(q) remain real or strongly real in SLn⁢(q) or SUn⁢(q). Gill and Singh answer the question of which elements of SLn⁢(q) that are real in GLn⁢(q) are also real in SLn⁢(q) in [9, Propositions 4.4 and 5.5], with the following classification.

Theorem 2.2

Theorem 2.2 (Gill and Singh)

Suppose that g∈SLn⁢(q) such that g is a real element of GLn⁢(q), and g is in the GLn⁢(q)-class C𝛍.

If n≢2⁢(mod⁡4) or q≢3⁢(mod⁡4), then g is also real in SLn⁢(q), and 𝒞𝝁 is a union of real SLn⁢(q)-classes.
If n≡2⁢(mod⁡4) and q≡3⁢(mod⁡4), then g is real in SLn⁢(q) if and only if g has some elementary divisor of the form f⁢(t)k for k odd. That is, 𝒞𝝁 is a union of real SLn⁢(q)-classes if and only if there is some f∈ℐ such that 𝝁⁢(f) has some odd part.

Gill and Singh also answer the question of which elements are strongly real in SLn⁢(q) in [9, Theorem 6.1], which we state below.

Theorem 2.3

Theorem 2.3 (Gill and Singh)

Suppose that g∈SLn⁢(q) such that g is a real element of SLn⁢(q), and g is in the GLn⁢(q)-class C𝛍.

If n≢2⁢(mod⁡4) or q is even, then g is also strongly real in SLn⁢(q), and 𝒞𝝁 is a union of strongly real SLn⁢(q)-classes.
If n≡2⁢(mod⁡4) and q is odd, then g is strongly real in SLn⁢(q) if and only if g has an elementary divisor of the form (t±1)k with k odd. That is, 𝒞𝝁 is a union of strongly real SLn⁢(q)-classes if and only if at least one of 𝝁⁢(t±1) has an odd part.

The main results of this paper are the statements on real and strongly real classes for SUn⁢(q) analogous to Theorems 2.2 and 2.3, which are given in Theorems 6.3 and 7.1.

3 Jordan decomposition and centralizers of unipotents

3.1 Jordan decomposition

By the Jordan decomposition of elements, for any linear algebraic group 𝐆 over 𝔽¯q, any g∈𝐆 can be written uniquely as g=s⁢u for commuting elements s,u∈𝐆, where s is semisimple and u is unipotent. If G is the group of 𝔽q-points of 𝐆 under some Frobenius map and g∈G, then in the decomposition g=s⁢u, we also have s,u∈G, where s is the p′-part of g and u is the p-part of g.

Now suppose that G=GLn⁢(q) or GUn⁢(q), and g∈G is in the G-conjugacy class 𝒞𝝁. If we consider the Jordan decomposition g=s⁢u, then we can say exactly in which G-classes s and u lie in the next result, which follows directly from considering the uniqueness of the decomposition of g in GLn.

Proposition 3.1

Let g∈G with G=GLn⁢(q) or GUn⁢(q), respectively, and suppose that g∈C𝛍. Let g=s⁢u be the Jordan decomposition of g. Define 𝛍s by 𝛍s⁢(f)=(1|𝛍⁢(f)|) for f∈I or U, respectively, and define μu to be the partition such that the multiplicity of each positive integer i in μu is given by

mi⁢(μu)=∑fdeg⁢(f)⁢mi⁢(𝝁⁢(f)).

Then s is in the semisimple class C𝛍s, and u is in the unipotent class corresponding to the partition μu.

For example, suppose that g∈GL15⁢(q) and g∈𝒞𝝁, where 𝝁⁢(f1)=(2,1), 𝝁⁢(f2)=(2), and 𝝁⁢(f3)=(3,1) with deg⁢(f1)=1 and deg⁢(f2)=deg⁢(f3)=2, and 𝝁⁢(f)=∅ for all other f∈ℐ. Then we have g=s⁢u, where s∈𝒞𝝁s with 𝝁s⁢(f1)=(13), 𝝁s⁢(f2)=(12), and 𝝁s⁢(f3)=(14), and u is in the unipotent class corresponding to the partition μu=(32⁢23⁢13).

3.2 Centralizers of unipotent elements

We first consider unipotent elements in GLn, and we let Jk denote the unipotent Jordan block of size k:

Jk=(1111⋱⋱111).

If a unipotent element has mk Jordan blocks of type k, then we denote this unipotent element by ⊕kJkmk, where the sum is over only those k such that mk≠0. If μ=(1m1⁢2m2⁢⋯) is a partition of n, then ⊕kJkmk is a representative element of the unipotent class of GLn corresponding to the partition μ. Taking u=⊕kJkmk, consider the centralizer CGLn⁢(u) of u in GLn. By [13, Theorem 3.1 (iv)] (which the authors point out can be concluded from results in [16, 20]), we have

CGLn⁢(u)=U⁢R,

where U=Ru⁢(CGLn⁢(u)) is the unipotent radical of the centralizer, and the other factor is reductive given by

R≅∏kGLmk.

We may further see precisely how the GLmk factor embeds in the centralizer. In particular, given any element a=(ai⁢j)∈GLmk, we see that the element (ai⁢j⁢Ik), consisting of k-by-k scalar blocks replacing the entries of a, is in the centralizer of the element Jkmk. We note that this also holds when we replace each Jordan block Jk with some GLk-conjugate. One observation that we need is the fact that

(3.1)det⁢((ai⁢j⁢Ik))=(det⁢(a))k.

Now consider the finite group GLn⁢(q), where we still consider the element u=⊕kJkmk∈GLn⁢(q) as a representative of the unipotent class corresponding to μ=(1m1⁢2m2⁢⋯). By [13, Theorem 7.1 (ii)], the centralizer of u in GLn⁢(q) is given by CGLn⁢(q)⁢(u)=UF⁢RF, where UF is a unipotent factor and RF is given by

RF≅∏kGLmk⁢(q),

given exactly by the F-fixed points of R, so that the GLmk⁢(q) factors are embedded the same as described above.

In the case of the unitary group, the unipotent element u=⊕kJkmk of GLn is no longer necessarily in GUn⁢(q). However, we can choose our Frobenius map Fη in such a way that there is a GLk-conjugate of Jk, J~k∈GUk⁢(q), such that the unipotent element u~=⊕kJ~kmk is a representative in GUn⁢(q) of the unipotent class of type μ. As explained in [13, p. 114], the Frobenius morphism Fη can be chosen simultaneously to stabilize each GLmk factor in R of the centralizer CGLn⁢(u~). The centralizer CGUn⁢(q)⁢(u~) is then given by [13, Theorem 7.1 (ii)] as CGUn⁢(q)⁢(u~)=UFη⁢RFη with UFη again a unipotent factor, and

RFη≅∏kGUmk⁢(q),

given by the Fη-fixed points of the R factor in CGLn⁢(u~), again embedded as before so that in particular (3.1) holds.

The following result will be important in the main arguments of this paper.

Lemma 3.2

Let u be a unipotent element in either G=GLn⁢(q) or G=GUn⁢(q) in the class corresponding to the partition μ=(1m1⁢2m2⁢⋯). For any δk∈Fq× or Cq+1, respectively, for all k such that mk≠0, there exists some g∈CG⁢(u) such that det⁢(g)=∏kδkk.

Suppose that mk=0 when k is odd, that is, u has no elementary divisors of the form (t-1)k with k odd. Then for any element g∈CG⁢(u), we have det⁢(g) is a square for any g∈CG⁢(u).

Proof.

From the discussion above, any element g∈CG⁢(u) is of the form g=v⁢r, where v is unipotent, and r is in the factor isomorphic to either ∏kGLmk⁢(q) or ∏kGUmk⁢(q), respectively. Since v is unipotent, we have det⁢(g)=det⁢(r). Now write r=∏kak with ak in the embedded factor of GLmk⁢(q) or GUmk⁢(q). Then we have from (3.1) that det⁢(r)=∏k(det⁢(ak))k, where the product is only over k with mk≠0. Since any choice of ak∈GLmk⁢(q) or GUmk⁢(q) will yield an element g∈CG⁢(u), then given any prescribed elements δk∈𝔽q× or Cq+1, we can choose ak such that det⁢(ak)=δk. Then

det⁢(g)=det⁢(r)=∏kδkk.

The assumption that mk=0 when k is odd implies that for any choice of the ak, and so for any g∈CG⁢(u), we must have det⁢(g) is a square. ∎

3.3 An important example

We briefly return to Theorem 2.2 of Gill and Singh. In [9], the proofs of Theorem 2.2(i) and the existence direction of (ii) are obtained by constructing explicit reversing elements h∈SLn⁢(q) of the elements in question. The more subtle argument is the converse statement of (ii), that is, if n≡2⁢(mod⁡4) and q≡3⁢(mod⁡4), and g∈SLn⁢(q) is real in GLn⁢(q) and has no elementary divisor of the form f⁢(t)k with k odd, then g is not real in SLn⁢(q). In [9, Proof of Proposition 5.5], it is proved that such an element has a reversing element h∈GLn⁢(q) such that det⁢(h)=-1. It is then stated that also there cannot exist reversing elements of g in GLn⁢(q) which have determinant 1, although some of the details of the proof of this claim are omitted. As similar arguments are crucial to the main results in the present paper, we fill in the details of this proof in this section.

We first consider the case that g is unipotent, and so g has no elementary divisor of the form (t-1)k with k odd. That is, g is in the unipotent class corresponding to a partition μ=(1m1⁢2m2⁢⋯) such that mk=0 whenever k is odd. Given the reversing element h∈GLn⁢(q) of g, if h1∈GLn⁢(q) is any other reversing element of g, then we must have h1=h⁢a for some a∈CGLn⁢(q)⁢(g). By Lemma 3.2, we have det⁢(a) is a square in 𝔽q×, and det⁢(h)=-1 is a nonsquare in 𝔽q× since q≡3⁢(mod⁡4), and so det⁢(h1) is a nonsquare. In particular, we have det⁢(h1)≠1 and h1∉SLn⁢(q).

In the case of general g, we consider the Jordan decomposition g=s⁢u with u unipotent. We are assuming that g has no elementary divisor of the form f⁢(t)k with k odd, so if g∈𝒞𝝁, then for every f∈ℐ the partition 𝝁⁢(f) has no odd parts. It follows from Proposition 3.1 that u has no elementary divisor of the form (t-1)k with k odd, since the corresponding partition μu has no odd parts. From the previous case, we know that u is not real in SLn⁢(q). Now g-1=s-1⁢u-1, and if y⁢g⁢y-1=g-1 for some y∈SLn⁢(q), then g-1=(y⁢s⁢y-1)⁢(y⁢u⁢y-1) with y⁢s⁢y-1 semisimple and y⁢u⁢y-1 unipotent. By the uniqueness of the Jordan decomposition, we would then have y⁢u⁢y-1=u-1, contradicting the fact that u is not real in SLn⁢(q). Thus g is not real in SLn⁢(q).

4 Initial observations

We begin with the following observation.

Lemma 4.1

Let G be a group and let g∈G be a real element. Then there is a reversing element for g whose order is a power of 2.

Proof.

Indeed, if h∈G is an arbitrary reversing element for g, then h2 commutes with g. Hence if (|h|)2′=m, then hm also reverses g and has order a power of 2. ∎

Proposition 4.2

Let q be even and let g∈SUn⁢(q) be real in GUn⁢(q). Then g is real in SUn⁢(q).

Proof.

This follows immediately from Lemma 4.1, since when q is even, any element whose order is a power of 2 necessarily has determinant 1. ∎

Proposition 4.3

Let n be odd and let g∈SUn⁢(q) be real in GUn⁢(q). Then g is real in SUn⁢(q).

Proof.

Note that since n is odd, the determinant is an isomorphism from the group [Z⁢(GUn⁢(q))]2 of scalar matrices with 2-power order to the subgroup C(q+1)2 of 𝔽q2× consisting of elements in Cq+1 whose orders are a power of 2. Let h be a reversing element of g, whose order we may take to be a power of 2, by Lemma 4.1. Then det⁡(h) is a member of C(q+1)2 and we may choose an element z∈Z⁢(GUn⁢(q)) with det⁡(z)=det⁡(h)-1. Hence hz is a reversing element of g with determinant 1, and so g is real in SUn⁢(q). ∎

In classifying the real classes of SUn⁢(q), Propositions 4.2 and 4.3 allow us to now restrict our attention to the case that q is odd and n is even.

We continue by recalling and adapting some of the results of Gill and Singh [9] to use in our situation. The following is a generalization of [9, Lemma 4.2] which has essentially the same proof, so we omit it here.

Lemma 4.4

Let G be a group and N⊲G a normal subgroup, with g∈N in the G-conjugacy class C. Suppose C is the disjoint union C=C1∪⋯∪Cs, where each Ci is an N-conjugacy class. Let xi∈G such that xi⁢g⁢xi-1∈Ci for each i. Then g is real (or strongly real) in N if and only if xi⁢g⁢xi-1 is real (or strongly real) in N for each i.

Taking G=GUn⁢(q) and N=SUn⁢(q) in Lemma 4.4, we see that for an element g∈SUn⁢(q), to determine whether the SUn⁢(q)-class of g is real or strongly real in SUn⁢(q), it suffices to consider any GUn⁢(q)-conjugate of g.

Given a group G and g∈G, define the subgroup RG⁢(g) by

RG⁢(g)={h∈G:h⁢g⁢h-1=g⁢ or ⁢g-1}.

Unless g2=1, we have [RG(g):CG(g)]=2. We may use this to obtain more information about real classes of SUn⁢(q).

Proposition 4.5

Suppose that q is odd, and that g∈SUn⁢(q) is real in GUn⁢(q). If g has any elementary divisor of the form f⁢(t)k with k odd, then g is real in SUn⁢(q).

Proof.

Let N=SUn⁢(q) and G=GUn⁢(q), and suppose that g is not real in N. Then we have RG⁢(g)∩N≤CG⁢(g), and [RG(g)N:CG(g)N]=2. Suppose that g is in a G-conjugacy class of type ν=(ν1,ν2,…,νs) as defined in Section 2.2, and let tν=gcd⁡(q+1,ν1,…,νs). As discussed at the end of Section 2.2, we have tν=[G:CG(g)N], which must be even since [RG(g)N:CG(g)N]=2. Now every part νi of ν must be even, so writing ν=(1n1⁢2n2⁢⋯), we have that nk=0 whenever k is odd. By definition of ν, this implies g has no elementary divisor of the form f⁢(t)k with k odd. ∎

Recall that we let Cq+1 denote the cyclic subgroup of 𝔽q2× of order q+1. The following is our analogue of [9, Lemma 5.1].

Lemma 4.6

Let q be odd and g∈SUn⁢(q). Then g is real in SUn⁢(q) if and only if there is a reversing element h∈GUn⁢(q) such that det⁢(h) has odd multiplicative order in Cq+1. In particular, if q≡1⁢(mod⁡4), then g is real in SUn⁢(q) if and only if there is a reversing element h∈GUn⁢(q) such that det⁢(h)=λ2 with λ∈Cq+1.

Proof.

If g is real in SUn⁢(q), then there is a reversing element with det⁢(h)=1 and we are done. Conversely, suppose h⁢g⁢h-1=g-1 with h∈GUn⁢(q) such that det⁢(h) has odd multiplicative order c. Then we have det⁢(hc)=det⁢(h)c=1, so hc∈SUn⁢(q). Since c is odd, we also have hc⁢g⁢(hc)-1=h⁢g⁢h-1=g-1, and so g is real in SUn⁢(q).

The last statement follows from the fact that when q≡1⁢(mod⁡4), we have (q+1)2=2, and so the elements of odd order in Cq+1 are exactly the squares. ∎

We now return to a result of Wonenburger [21] which states that for any vector space V of finite dimension over a field of characteristic not 2, any element of any orthogonal group O⁢(V) over V is strongly real in O⁢(V). We need a slight refinement of this statement in the case of finite fields. Recall that in the finite orthogonal group On±⁢(q), with q odd, that for g∈On±⁢(q), the elementary divisors of g are such that g is real in GLn⁢(q). Further, every elementary divisor of g∈On±⁢(q) of the form (t±1)2⁢k appears with even multiplicity [20, Section 2.6, Case (C)].

Lemma 4.7

Let q be odd and let g∈On±⁢(q).

If g has an elementary divisor of the form (t±1)k with k odd, then there exist reversing elements h,h′∈On±⁢(q) such that
h2=h′⁣2=1 𝑎𝑛𝑑 det⁢(h)=-det⁢(h′).
If g has no elementary divisor of the form (t±1)k with k odd, then n is even, and g has a reversing element h∈On±⁢(q) such that
h2=1 𝑎𝑛𝑑 det⁢(h)=(-1)n/2.

Proof.

We let V=𝔽qn with some underlying symmetric form (⋅,⋅). Then by [21, Remark I after Lemma 5], V can be decomposed as a direct sum of mutually orthogonal g-invariant subspaces, V=⊕iVi, where we write gi=g|Vi, such that either Vi is cyclic with respect to g (and so with respect to gi), or Vi=Ui⁢1⊕Ui⁢2, where Ui⁢1 and Ui⁢2 are each cyclic with respect to gi, and the minimal and characteristic polynomials of gi|Ui⁢1 and gi|Ui⁢2 are of the form (t±1)2⁢ni. In each case we have gi∈O⁢(Vi), where O⁢(Vi) is the orthogonal group with respect to (⋅,⋅) restricted to Vi, and in the first case the characteristic polynomial of gi is self-reciprocal and not of the form (t±1)2⁢ni. Wonenburger proves [21, Lemmas 2 and 5] that each gi is reversed by some hi∈O⁢(Vi) such that hi2=1, so that h=⊕ihi∈O⁢(V) reverses g and satisfies h2=1. We have det⁢(h)=∏idet⁢(hi).

In the case that Vi is cyclic, either gi has characteristic polynomial of the form (f⁢f~)k, where f≠f~ and f⁢(t) is irreducible in 𝔽q⁢[t] so that (f⁢f~)k has even degree, or of the form fk where f=f~ is self-reciprocal and irreducible in 𝔽q⁢[t], in which case either deg⁢(f) is even or f⁢(t)=t±1 and k is odd. Let ki be the degree of the characteristic polynomial of gi, and let vi be a cyclic generator for Vi so that a basis for Vi is given by {vi,gi⁢vi,…,giki-1⁢vi}. By [21, proof of Lemma 2], gi is reversed by an element hi∈O⁢(Vi) such that hi2=1 defined by hi=1Pi⊕-1Qi, where Pi is generated by vectors of the form (gil+gi-l)⁢vi and Qi is generated by vectors of the form (gil-gi-l)⁢vi. In the case that Vi=Ui⁢1⊕Ui⁢2 with Ui⁢1 and Ui⁢2 cyclic, let ui⁢1 and ui⁢2 be cyclic generators. Then hi=1Pi⊕-1Qi, where Pi is generated by vectors of the form (gil+gi-l)⁢ui⁢1 and (gil+gi-l)⁢ui⁢2 and Qi is generated by vectors of the form (gil-gi-l)⁢ui⁢1 and (gil-gi-l)⁢ui⁢2.

If g has an elementary divisor of the form (t±1)k with k odd, then this corresponds to some gj with characteristic polynomial of the form (t±1)kj with kj=2⁢nj-1 for some nj. In this case, Pj has dimension nj while Qi has dimension nj-1, and hj=1Pj⊕-1Qj has determinant (-1)nj-1. However, -hj also reverses gj and satisfies (-hj)2=1, and det⁢(-hj)=-det⁢(hj). For any hi chosen for i≠j, we can take h=⊕ihi and h′=-hj⊕⊕i≠jhi, so that one of h or h′ has determinant 1 and the other -1.

Now assume that g has no elementary divisor of the form (t±1)k with k odd. In the case that Vi is not cyclic, Ui⁢1 and Ui⁢2 are each cyclic (although degenerate with respect to the underlying form), with gi|Ui⁢1 and gi|Ui⁢2 each having characteristic polynomial (t±1)2⁢ni. Then Pi and Qi each have dimension 2⁢ni, so that hi=1Pi⊕-1Qi has determinant (-1)2⁢ni=1, and we note that the characteristic polynomial of gi has degree 4⁢ni.

Now consider the case that Vi is cyclic. As discussed above, either gi has characteristic polynomial (fi⁢fi~)ki with fi≠fi~ or fiki with fi=fi~ and fi⁢(t)≠t±1. In both cases, dim⁢Vi is even, so that n=∑idim⁢Vi is now seen to be even. In the first case, if deg⁢(fi)=di, then let ni=di⁢ki, so hi has determinant (-1)ni and ni is the dimension of Qi. In the second case, we must have deg⁢(fi)=2⁢di even, and again we have det⁢(hi)=(-1)ni, where ni=di⁢ki is the dimension of Qi. It now follows that det⁢(h)=∏idet⁢(hi)=(-1)n/2. ∎

We note that Lemma 4.7 implies that if either n is odd or 4|n, every element of SOn±⁢(q) is strongly real in SOn±⁢(q), which was also mentioned by Gow in [12, p. 250]. We address the case that n≡2⁢(mod⁡4) in Section 7.

Consider a conjugacy class 𝒞𝝁 of GUn⁢(q). Let 𝒰𝝁 be the set of f∈𝒰 such that 𝝁⁢(f)≠∅. For each f∈𝒰𝝁, define nf=deg⁢(f)⁢|𝝁⁢(f)| and 𝝁f:𝒰→𝒫 by 𝝁f⁢(f)=𝝁⁢(f) and 𝝁f⁢(f′)=∅ if f′≠f. We may take an element g∈𝒞𝝁 which is in block diagonal form, g=⊕gf, where each gf∈GUnf⁢(q) is in the class 𝒞𝝁f of GUnf⁢(q).

In the case that 𝒞𝝁 is a real class of GUn⁢(q), so that 𝝁⁢(f)=𝝁⁢(f~) for each f∈𝒰, as in Section 2.3, write g as above in the form g=⊕gf*, where we define gf*=gf and nf*=nf if f=f~, and gf*=gf⊕gf~ and nf*=nf+nf~ if f≠f~. Then each gf* is a real element of GUnf*⁢(q). Also note that when f≠t±1, then nf* is always even, since deg⁢(f) is even whenever f=f~ and f≠t±1, and in general deg⁢(f)=deg⁢(f~).

We may now apply Lemma 4.7 in the following situation.

Lemma 4.8

Let q be odd, and C𝛍 a real class in GUn⁢(q). Let g∈C𝛍 such that g=⊕gf* as above, and let g*=⊕f≠t±1gf* with n*=∑f≠t±1nf*. Then g*∈GUn*⁢(q) is reversed by an element in SUn*⁢(q).

Proof.

Since g is real in GUn⁢(q), we see that g* is real in GUn*⁢(q). Also, we have g*∈SUn*⁢(q) since g* is real and has no elementary divisor of the form (t+1)k. Now, if g* has an elementary divisor of the form f⁢(t)k with k odd, then the statement follows by Proposition 4.5, so we may assume g* has no such elementary divisors.

If f⁢(t)k is an elementary divisor of g* and f is self-reciprocal, then deg⁢(f) is even, while if f is not self-reciprocal, then f~⁢(t)k is also an elementary divisor of g*. So 4|n*. Since g* is real as an element of GUn*⁢(q) and has no powers of t±1 as elementary divisors, it follows from the work of Wall [20, Section 2.6, Cases (A) and (C)] that g* is an element of some orthogonal group On*±⁢(q) embedded in GUn*⁢(q). Since 4|n*, it follows from Lemma 4.7 that g* is reversed by an element of the embedded SOn*±⁢(q), and so g* is real in SUn*⁢(q). ∎

In light of Lemma 4.8, we are particularly interested in reversing elements of gt-1 or gt+1. We therefore turn our attention to unipotent elements.

5 Real unipotent elements

First we consider regular unipotent elements in GUn⁢(q) (and so in SUn⁢(q)), where q is odd. Because of Proposition 4.3, we may restrict our considerations to the case that n is even.

If we take our Hermitian matrix η which defines GUn⁢(q) to be

then we may find a regular unipotent element in GUn⁢(q) which is upper unitriangular with all super-diagonal entries nonzero. We fix u to be such an element.

Lemma 5.1

Let q be odd, n=2⁢m be even, and let u∈GUn⁢(q) be the regular unipotent element just described. Then any reversing element h∈GUn⁢(q) of u must be upper triangular with diagonal entries (β,-β,…,β,-β) such that β∈Fq2× with βq+1=-1, and such that det⁢(h)=(-1)m⁢βn.

Proof.

The fact that h must be upper triangular follows from the fact that u is upper unitriangular, and all super-diagonal entries of u are nonzero. Since u-1 is also upper unitriangular and the super-diagonal entries are the negatives of those of u, it follows that the diagonal of h must be of the form (β,-β,…,β,-β) for some β∈𝔽q2×. It immediately follows that det⁢(h)=(-1)m⁢βn. Finally, since h∈GUn⁢(q), we have

ηn-1⁢(F⊤⁢(h)-1)⁢ηn=h.

By considering the first entry in the diagonal on each side, we have -β-q=β, so βq+1=-1. ∎

We first deal with the case q≡1⁢(mod⁡4). We remark that the results of Lemmas 5.2 and 5.4 are already known from [18, Lemma 2.2], although we include the proofs here for the sake of self-containment.

Lemma 5.2

Let n=2⁢m be even and q≡1⁢(mod⁡4). If u∈GUn⁢(q) is the regular unipotent element fixed above, then u is real in SUn⁢(q).

Proof.

Since u is real in GUn⁢(q), there is some reversing element h∈GUn⁢(q). By Lemma 5.1, h is upper triangular with diagonal entries (β,-β,…,β,-β) for some β∈𝔽q2× such that βq+1=-1, and det⁢(h)=(-1)m⁢βn. In particular, β2∈Cq+1.

If n≡0⁢(mod⁡4), then write n=4⁢k. Then we have det⁢(h)=βn=(β2⁢k)2 with β2⁢k∈Cq+1. By Lemma 4.6, u is real in SUn⁢(q).

Now assume that n≡2⁢(mod⁡4), and write n=4⁢k+2=2⁢m. Then we have det⁢(h)=-β2⁢m. Let γ be a multiplicative generator for Cq+1. Since (q+1)2=2, it follows that -1=γ(q+1)/2 is not the square of an element in Cq+1. We have β2∈Cq+1, so β2=γs for some integer s. If s is even, then β∈Cq+1, which contradicts βq+1=-1. Thus since m=2⁢k+1 is odd, we see β2⁢m is some odd power of γ, which implies -β2⁢m is the square of some element in Cq+1. By Lemma 4.6, u is real in SUn⁢(q). ∎

The case q≡3⁢(mod⁡4) is slightly different.

Lemma 5.3

Let n=2⁢m be even and q≡3⁢(mod⁡4). If u∈GUn⁢(q) is the regular unipotent fixed above, then u is real in SUn⁢(q) if and only if n is divisible by (q2-1)2.

Proof.

Again, the element u is real in the group GUn⁢(q), and so by Lemma 5.1 any reversing element h∈GUn⁢(q) of u must be upper triangular with diagonal entries (β,-β,…,β,-β) for some β∈𝔽q2× such that βq+1=-1, and

det⁢(h)=(-1)m⁢βn.

If |β| is the multiplicative order of β, then βq+1=-1 implies that

(|β|)2=2⁢(q+1)2=(q2-1)2.

By Lemma 4.6, we know u is real if and only if a reversing element h∈GUn⁢(q) exists such that det⁢(h)=(-1)n⁢β2⁢m has odd order in 𝔽q2×. However, the fact that (|β|)2=2⁢(q+1)2 means that this happens if and only if m is divisible by (q+1)2. Thus u is real in SUn⁢(q) if and only if n is divisible by (q2-1)2. ∎

We can now describe which unipotent elements are real in SUn⁢(q) in general.

Lemma 5.4

Let u be any unipotent element in GUn⁢(q). Suppose q≡1⁢(mod⁡4). Then u is real in SUn⁢(q).

Proof.

If u has any elementary divisor of the form (t-1)k with k odd, then u is real in SUn⁢(q) by Proposition 4.5, so we may assume otherwise. Suppose u is in the unipotent class of GUn⁢(q) corresponding to the partition μ of n, so we are assuming μ has no odd parts, and so μ=(μ1,μ2,…)=(2m2⁢4m4⁢⋯).

Define GUn⁢(q) by the Hermitian matrix in block diagonal form ⊕iημi, where each ημi has 1’s on the antidiagonal and 0’s elsewhere as above. Then there is an element in the conjugacy class corresponding to μ of the block diagonal form ⊕kJ~kmk, where each J~k is regular unipotent in GUk⁢(q) of the form considered in Lemma 5.2. Since each J~k is real in SUk⁢(q), it follows that ⊕kJ~kmk is real in SUn⁢(q), as is u. ∎

Again, the case q≡3⁢(mod⁡4) is more complicated. We consider elements which are slightly more general than unipotents.

Lemma 5.5

Let g∈SUn⁢(q) such that g only has elementary divisors which are powers of t±1, and suppose q≡3⁢(mod⁡4). Then g is real in SUn⁢(q) if and only if one of the following holds:

g has an elementary divisor of the form (t±1)k with k odd.
(q2-1)2 divides k for all elementary divisors (t±1)k of g.
Write 2r for the smallest power of 2 such that 2≤2r<(q2-1)2 and (k)2=2r for some elementary divisor (t±1)k of g. Then g has an even number of elementary divisors of the form (t±1)e such that (e)2=2r.

Proof.

Note that if (1) holds, then g is real in SUn⁢(q) by Proposition 4.5. So we now assume (1) does not hold. Then g is in the GUn⁢(q)-conjugacy class corresponding to 𝝁, where we let

𝝁⁢(t-1)=τ=(τ1,τ2,…)=(2m2⁢(τ)⁢4m4⁢(τ)⁢⋯),

and

𝝁⁢(t+1)=ν=(ν1,ν2,…)=(2m2⁢(ν)⁢4m4⁢(ν)⁢⋯),

so that each τi and each νi is even.

By Lemma 4.4, we may assume that g is of the form g=u⊕v, where u=gt-1 and u=gt+1 as in Section 4 before Lemma 4.8. Note that -v is unipotent.

Similar to the proof of Lemma 5.4, we define GUn⁢(q) by the Hermitian matrix in block diagonal form ⊕iητi⊕⊕iηνi. Then there is an element in the conjugacy class corresponding to 𝝁 in the block diagonal form

\text{{(}$*$)}⊕kJ~kmk⁢(τ)⊕⊕k(-J~k)mk⁢(ν),

where each J~k is regular unipotent of the form considered in Lemma 5.3. If we assume (2) holds, then each J~k and -J~k is real in SUk⁢(q), which implies (*) and thus g, is real in SUn⁢(q).

We now also assume that (2) does not hold, and we show that g is real in SUn⁢(q) if and only if (3) holds. As in the previous paragraph, we consider the element which is GUn⁢(q)-conjugate to g defined by

g~=u~⊕v~=⊕kJ~kmk⁢(τ)⊕⊕k(-J~k)mk⁢(ν).

Let mk=mk⁢(τ)+mk⁢(ν) and for each k such that mk≠0, fix hk∈GUk⁢(q) to be a reversing element of J~k, so that by Lemma 5.1, each hk is upper triangular with diagonal entries (βk,-βk,…,βk,-βk) for some element βk∈𝔽q2× satisfying βkq+1=-1. It follows that hk is also a reversing element for -J~k. The block diagonal element h′=⊕k⊕i=1mkhk∈GUn⁢(q) is a reversing element for g~ in GUn⁢(q), and we have

det⁢(h′)=(-1)n/2⁢∏kβkk⁢mk.

Every reversing element of g~ in GUn⁢(q) is of the form h=h′⁢c, where c can be any element of CGUn⁢(q)⁢(g~). Now, CGUn⁢(q)⁢(g~) is given by the direct sum

CGUn⁢(q)⁢(g~)=CGU|τ|⁢(q)⁢(u~)⊕CGU|ν|⁢(q)⁢(v~).

Since -v~ is unipotent, and has the same centralizer as v~, we may apply Lemma 3.2 to both u~ and v~. In particular, the possible determinants of elements in CGUn⁢(q)⁢(g~) are of the form ∏kδkk, where the product is over all k which appear as parts of either τ or ν, and δk∈Cq+1 are arbitrary. It follows that all possible determinants of reversing elements h of g~ in GUn⁢(q) are of the form

det⁢(h)=(-1)n/2⁢∏kβkk⁢mk⁢δkk.

Let 2r be the smallest power of 2 such that 2≤2r<(q2-1)2 and (k)2=2r for some elementary divisor (t±1)k of g~ (and so of g). Noting that all of the parts of τ and ν are assumed even, we re-label all of the distinct part sizes of τ or ν as 2⁢ki such that 2r-1=(k1)2≤(k2)2≤⋯. We also let s be the total number of distinct part sizes of τ or ν with 2-part 2r, so (k1)2=(k2)2=⋯=(ks)2=2r-1. We show that the element g is real in SUn⁢(q) if and only if ∑i=1sm2⁢ki is even, where m2⁢ki=m2⁢ki⁢(τ)+m2⁢ki⁢(ν).

Let α∈Cq+1 with multiplicative order |α|=(q+1)2. Since for each β2⁢ki we have |β2⁢ki| is divisible by (q2-1)2, we can write β2⁢ki2=ασi⁢γi for some odd integer σi and some γi∈Cq+1 with odd multiplicative order. We also write each δ2⁢ki=αρi⁢γi′, where γi′∈Cq+1 has odd multiplicative order and ρi is some integer (δ2⁢ki∈Cq+1 is still arbitrary with freedom to choose ρi and γi′). Now

det⁢(h)=(-1)n/2⁢α∑iki⁢(m2⁢ki⁢σi+2⁢ρi)⁢∏iγi′′,

where γi′′=γim2⁢ki⁢ki⁢γi′⁣2⁢ki has odd multiplicative order, and so ∏iγi′′=γ does as well.

Now write

x=∑i=1s(ki)2′⁢(m2⁢ki⁢σi+2⁢ρi),

y=∑i>ski(k1)2⁢(m2⁢ki⁢σi+2⁢ρi),

so that we have

(5.1)det⁢(h)=(-1)n/2⁢α(k1)2⁢(x+y)⁢γ,

where y is necessarily even.

First assume that ∑i=1sm2⁢ki is odd. Since each (ki)2′ and each σi is odd, this implies that x is odd, and so x+y is odd. Suppose that there is some h as above such that det⁢(h) has odd multiplicative order. If k1 is even, note first that this means 4|n. Since γ has odd multiplicative order, the only way det⁢(h) can as well is if α(k1)2⁢(x+y) does. Since x+y is odd, this would mean α(k1)2 has odd multiplicative order. But |α|=(q+1)2, while (k1)2<(q+1)2, a contradiction. So suppose k1 is odd, so 2⁢ki≡2⁢(mod⁡4) for i≤s. Since ∑i=1sm2⁢ki is odd, it follows that n/2 is odd. The only way det⁢(h) can have odd multiplicative order is if (-1)n/2⁢α(k1)2⁢(x+y)=-αx+y does. But x+y is odd, while |α|=(q+1)2≥4, and so this is impossible. Thus no such h exists, and g cannot be real in SUn⁢(q) by Lemma 4.6.

Now assume that ∑i=1sm2⁢ki is even. In this case, we must have 4|n. We show that ρi and γi′ can be chosen (that is, c∈CGUn⁢(q)⁢(g~) can be chosen) so that det⁢(h) has odd multiplicative order in Cq+1. We fix γi′ to be some elements of odd multiplicative order in Cq+1, and we take ρi=0 for i>1. We will choose ρ1 such that

x+y=(k1)2′⁢(m2⁢k1⁢σ1+2⁢ρ1)+∑i>1ki(k1)2⁢m2⁢ki⁢σi

is divisible by (q+1)2, which will give α(k1)2⁢(x+y)=1. Note that for i≤s we have ki/(k1)2 is odd while ki/(k1)2 is even for i>s. Since (k1)2′ and all σi are odd, and ∑i=1sm2⁢ki is assumed even, we have

(k1)2′⁢m2⁢k1⁢σ1+∑i>1ki(k1)2⁢m2⁢ki⁢σi

must be even. Noting that (k1)2′ is invertible modulo (q+1)2, we now choose ρ1 such that

ρ1≡12⁢(-(k1)2′⁢m2⁢k1⁢σ1-∑i>1ki(k1)2⁢m2⁢ki⁢σi)⁢(k1)2′-1⁢(mod ⁢(q+1)2).

This gives α(k1)2⁢(x+y)=1 as claimed, which makes det⁢(h)=γ have odd multiplicative order in Cq+1. By Lemma 4.6, we now have g is real in SUn⁢(q). ∎

We restate the conditions in Lemma 5.5 for the unipotent case in a slightly refined manner as follows.

Lemma 5.6

Let u∈GUn⁢(q) be unipotent, and suppose q≡3⁢(mod⁡4). Then u is real in SUn⁢(q) if and only if the following holds:

Write 2r for the smallest power of 2 such that 2≤2r<(q2-1)2 and (k)2=2r for some elementary divisor (t-1)k of u. Then u has an even number of elementary divisors of the form (t-1)e such that (e)2=2r (where the number of these is 0 if the smallest power 2r dividing the exponent of some elementary divisor (t-1)k is either 1 or is at least (q2-1)2).

The conditions implied by Lemma 5.5 for the case n≡2⁢(mod⁡4) are a bit simpler.

Corollary 5.7

Let u∈GUn⁢(q) be unipotent and suppose that q≡3⁢(mod⁡4) and n≡2⁢(mod⁡4). Then u is real in SUn⁢(q) if and only u has an elementary divisor of the form (t-1)k with k odd.

Proof.

If u has such an elementary divisor, then we know u is real in SUn⁢(q). Suppose u has no such elementary divisor. Then all elementary divisors of u are of the form (t-1)k with k even, and there must exist some k with (k)2=2, since n≡2⁢(mod⁡4). Since (q2-1)2≥4, we know (q2-1)2 does not divide such k. Now, 2 is the smallest 2-part of a k such that (t-1)k is an elementary divisor of u. There must be an odd number of such elementary divisors since n≡2⁢(mod⁡4). Since conditions (1), (2), and (3) of Lemma 5.5 all fail, it follows u is not real in SUn⁢(q). ∎

6 Classification of real elements

The following is a key result to understand reality in SUn⁢(q) in the most complicated case.

Lemma 6.1

Suppose that q≡3⁢(mod⁡4) and n≡0⁢(mod⁡4). Let g∈SUn⁢(q) with Jordan decomposition g=s⁢u. Then g is real in SUn⁢(q) if and only if u is real in SUn⁢(q).

Proof.

First, if g is real in SUn⁢(q) and is reversed by h, then g-1=s-1⁢u-1=(h⁢s⁢h-1)⁢(h⁢u⁢h-1), and u-1=h⁢u⁢h-1 by uniqueness of the Jordan decomposition, so u is real in SUn⁢(q). Therefore we now assume that u is real in SUn⁢(q), and we prove that g is real in SUn⁢(q).

We may suppose that g is in the form g=⊕gf*, as in Section 4 before Lemma 4.8. Write

g*=⊕f≠t±1gf*,so g=g*⊕gt+1⊕gt-1,

where, as in Lemma 4.8, g*∈SUn*⁢(q) with n*=∑f≠t±1nf*. By Lemma 4.8, g* is real in SUn*⁢(q). Write g±=gt+1⊕gt-1 and n±=nt+1+nt-1, so we have g±∈SUn±⁢(q). Since g* is real in SUn*⁢(q), we may assume that g± is not real in SUn±⁢(q), as otherwise we have g is real in SUn⁢(q).

First note that if u has an elementary divisor of the form (t-1)k with k odd, then by Proposition 3.1, g must have an elementary divisor of the form f⁢(t)k with k odd, and so by Proposition 4.5g is real in SUn⁢(q). Also, if (q2-1)2 divides k for all elementary divisors of the form (t-1)k of u, then by Proposition 3.1, it follows that (q2-1)2 divides k for all elementary divisors f⁢(t)k of g. In particular, (q2-1)2 divides k for all elementary divisors (t±1)k of gt+1⊕gt-1. By Lemma 5.5, we have g±=gt+1⊕gt-1 is real in SUn±⁢(q), and since g* is real in SUn*⁢(q), it follows that g=g*⊕g± is real in SUn⁢(q). So now let 2r be the smallest power of 2 such that 2≤2r<(q2-1)2 and (k)2=2r for some elementary divisor (t-1)k of u. By Lemma 5.6, we may now assume that u has a positive even number of elementary divisors of the form (t-1)e satisfying (e)2=2r.

By Proposition 3.1, 2r is also the smallest power of 2 such that 2≤2r<(q2-1)2 and (k)2=2r for some elementary divisor f⁢(t)k of g. Suppose that g has an elementary divisor of the form (t±1)k such that (k)2=2r. As noted in Section 4, if f⁢(t) is any elementary divisor of g of the form f⁢(t)e such that (e)2=2r and f⁢(t)≠t±1, then deg⁢(f) is even if f=f~, while if f≠f~, then f~⁢(t)e is an elementary divisor of g with the same multiplicity as f⁢(t)e. Since the total number of elementary divisors of u of the form (t-1)k with (k)2=2r is even, Lemma 3.1 then implies that the total number of elementary divisors of g of the form (t±1)k with (k)2=2r is also even. But now g± is real in SUn±⁢(q) by Lemma 5.5, and then g is real in SUn⁢(q). Thus, we can assume that g has no elementary divisor of the form (t±1)k with (k)2=2r, but there is some elementary divisor of g of the form f⁢(t)k with (k)2=2r and f⁢(t)≠t±1.

Let (k1)2 be the smallest 2-part of an exponent of an elementary divisor of g±. From the previous paragraph, we are assuming 2r<(k1)2. Since we are assuming that g± is not real in SUn±⁢(q), then the number of elementary divisors of g± with exponent having 2-part equal to (k1)2 is odd, by Lemma 5.5. Note also that as in the proof of Lemma 4.8, we have 4|n*, and since 4|n and n=n*+n± then 4|n±. We know that g± is real in GUn±⁢(q), so let h±∈GUn±⁢(q) be any reversing element of g±. If g is in the GUn⁢(q)-class corresponding to 𝝁, let 𝝁⁢(t-1)=τ and 𝝁⁢(t+1)=ν. Let α∈Cq+1 have multiplicative order |α|=(q+1)2. Using exactly the notation as in the proof of Lemma 5.5, we know from equation (5.1) that the determinant of h± must be of the form

det⁢(h±)=(-1)n±/2⁢α(k1)2⁢(x+y)⁢γ,

where γ has odd multiplicative order. Since we are assuming that g± is not real in SUn±⁢(q), we know from the proof of Lemma 5.5 that x+y is odd. Since |α|=(q+1)2, it follows that the element α1=αx+y also has multiplicative order |α1|=(q+1)2. Since 4|n±, we have

det⁢(h±)=α1(k1)2⁢γ.

Now let h*∈SUn*⁢(q) be a reversing element for g*. If c∈CGUn*⁢(q)⁢(g*), then h*⁢c is a reversing element for g* in GUn*⁢(q). Then h*⁢c⊕h± is a reversing element in GUn⁢(q) for g=g*⊕g±, and

det⁢(h*⁢c⊕h±)=det⁢(h*⁢c)⁢det⁢(h±)=det⁢(c)⁢α1(k1)2⁢γ.

Since γ has odd multiplicative order, if we can find c∈CGUn*⁢(q)⁢(g*) such that det⁢(c)⁢α1(k1)2 has odd multiplicative order in Cq+1, then by Lemma 4.6 we have g is real in SUn⁢(q).

It follows from the work of Wall [20, Section 2.6, Case (A)] (see also [7, Proposition 2.1]), that for g*=⊕f≠t±1gf, we have

CGUn*⁢(q)⁢(g*)=⊕f≠t±1CGUnf⁢(q)⁢(gf).

So, for any c∈CGUn*⁢(q)⁢(g*), we write

c=⊕f≠t±1cf,cf∈CGUnf⁢(q)⁢(gf).

Now let f1 be the polynomial such that f1⁢(t)≠t±1 and g has as an elementary divisor f1⁢(t)k with (k)2=2r<(k1)2.

Recall that (k)2′ is invertible modulo (q+1)2 and that (k)2 divides (k1)2, so that we may define

δ=α1-(k1)2/k∈Cq+1.

We will take c such that cf=1 when f≠f1 and show that we may find an element cf1∈CGUnf1⁢(q)⁢(gf1) such that det⁡(cf1)=δk, so that

det⁢(c)⁢α1(k1)2=det⁢(cf1)⁢α1(k1)2=δk⁢α1(k1)2

has odd multiplicative order in Cq+1.

Write gf1=s1⁢u1 in terms of its Jordan decomposition. In the following, we make use of the structure of the centralizer of semisimple elements, details of which can be found in [6, Proposition (1A)], [17, Part (B1), proof of Theorem 4.1], and [15, Theorem 2.4.1], for example.

Write d for the degree of f1 and C⁢(s1) for the centralizer CGUnf1⁢(q)⁢(s1). From Lemma 3.1, s1 is in the class corresponding to (1|𝝁⁢(f1)|). Then

C⁢(s1)≅{GU|𝝁⁢(f1)|⁢(qd)if f1=f1✓,GL|𝝁⁢(f1)|⁢(qd)if f1=f2⁢f2✓.

for some irreducible f2 with f2≠f2✓, where d=2⁢deg⁡(f2). As in [9, discussion before Lemma 5.2], in identifying C⁢(s1) with GU|𝝁⁢(f1)|⁢(qd), respectively GL|𝝁⁢(f1)|⁢(qd), we will need to distinguish between the determinant of an element of C⁢(s1) when viewed over 𝔽q2⁢d, respectively 𝔽qd, and when viewed over 𝔽q2. In the case f1=f1✓, if z∈C⁢(s1) has determinant detq2⁢d⁡(z) as an element of GU|𝝁⁢(f1)|⁢(qd), then z has determinant

detq2⁡(z)=Nq2⁢d/q2⁢(detq2⁢d⁡(z))

as an element of GUnf1⁢(q). In the case f1=f2⁢f2✓, if z∈C⁢(s1) has determinant detqd⁡(z) as an element of GL|𝝁⁢(f1)|⁢(qd), then z has determinant

detq2(z)=Nqd/q2(detqd(z)1-q)

as an element of GUnf1⁢(q). Here Nq2⁢d/q2 and Nqd/q2 are the standard norm maps for finite fields.

Now, recalling that the norm maps are surjective, we may find δ~ such that in the case f1=f1✓,

δ~∈Cqd+1≤𝔽q2⁢d× and Nq2⁢d/q2⁢(δ~)=δ,

and in the case f1=f2⁢f2✓,

δ~∈𝔽qd× and Nqd/q2⁢(δ~)1-q=δ.

Note that an element is a member of CGUnf1⁢(q)⁢(gf1) if and only if it is a member of C⁢(s1) which commutes with u1. Write

𝝁⁢(f1)=ρ=(ρ1,ρ2,…)

such that ρℓ=k, so (ρℓ)2=2r. Then u1 is conjugate in C⁢(s1) (identified with GU|𝝁⁢(f1)|⁢(qd), respectively GL|𝝁⁢(f1)|⁢(qd)) to an element u1~ in block diagonal form corresponding to its elementary divisors (t-1)ρi. The element

cf1~=⊕i<ℓIρi⊕δ~⁢Iρℓ⊕⊕i>ℓIρi,

where Iρi is the ρi-by-ρi identity, is then an element of GU|𝝁⁢(f1)|⁢(qd), respectively GL|𝝁⁢(f1)|⁢(qd), which commutes with u1~. Hence there is a conjugate in C⁢(s1), say cf1, of cf1~ which commutes with u1, and we see that detq2⁢d⁡(cf1)=δ~ρℓ, respectively detqd⁡(cf1)=δ~ρℓ, so that detq2⁡(cf1)=δρℓ by our definition of δ~. Hence we have cf1∈CGUnf1⁢(q)⁢(gf1) and detq2⁡(cf1)=δk as desired. ∎

Remark 6.2

We note that there are relatively few situations for other finite reductive groups in which the corresponding statement to Lemma 6.1 holds. In fact, for the remaining cases q≢3mod4 or n≢0mod4, Theorem 6.3 below yields that the statement is certainly not true for SUn⁢(q). Indeed, in case (1) of Theorem 6.3, u is real in SUn⁢(q) if and only if u is real in GUn⁢(q), which means every unipotent element is real in SUn⁢(q). Since in general there are elements g∈SUn⁢(q) which are not real in GUn⁢(q) (and therefore not real in SUn⁢(q)), the statement would fail in this case. Similarly, in case (2) of Theorem 6.3, note that u has an elementary divisor (t-1)k with k odd if and only if g has an elementary divisor f⁢(t)k with k odd (by Proposition 3.1), but that there can certainly be g∈SUn⁢(q) with such an elementary divisor which is not real in GUn⁢(q). For the same reasoning, observing Gill and Singh’s result (Theorem 2.2), the corresponding statement to Lemma 6.1 also fails for SLn⁢(q). From these observations, it also follows immediately that the statement fails for GLn⁢(q) and GUn⁢(q). However, the statement holds in all cases for On±⁢(q), as well as for SOn±⁢(q) when n≢2(mod4) and q is odd, since in these situations every element is real. It is exactly this reason, and the work of Wonenburger applied to Lemma 4.8, which makes the statement go through in Lemma 6.1. However, we see from Theorem 7.2 below that the statement of Lemma 6.1 will fail for SOn±⁢(q) when n≡2(mod4) and q is odd.

Finally, we arrive at our main results.

Theorem 6.3

Let g∈SUn⁢(q) and suppose that g is real as an element of GUn⁢(q).

If q≢3⁢(mod⁡4) or n is odd, then g is real in SUn⁢(q).
If q≡3⁢(mod⁡4) and n≡2⁢(mod⁡4), then g is real in SUn⁢(q) if and only if g has an elementary divisor of the form f⁢(t)k with k odd.
If q≡3⁢(mod⁡4) and n≡0⁢(mod⁡4), then g is real in SUn⁢(q) if and only if the following holds: Write 2r for the smallest power of 2 such that 2≤2r<(q2-1)2 and (k)2=2r for some elementary divisor f⁢(t)k of g. Then g has an even number of elementary divisors of the form (t±1)e such that (e)2=2r (where the number of these is 0 if the smallest power 2r dividing the exponent of some elementary divisor is either 1 or is at least (q2-1)2).

Proof.

If q is even or n is odd, then we know g is real in SUn⁢(q) by Propositions 4.2 and 4.3, so we assume q is odd and n is even. Replace g by an element in its conjugacy class of the form ⊕gf*, as in Section 4 before Lemma 4.8. As mentioned there, we have n=∑nf*, where nf* is even when f⁢(t)≠t±1.

If q≡1⁢(mod⁡4), then g*=⊕f≠t±1gf* is real in SUn*⁢(q) by Lemma 4.8. By Lemma 5.4 we have gt-1 is real in SUnt-1⁢(q). Since -gt+1 is unipotent, it is real, and so gt+1 is real in SUnt+1⁢(q). Thus ⊕gf* is real in SUn⁢(q), and so g is as well.

Now assume that q≡3⁢(mod⁡4), and suppose first that n≡2⁢(mod⁡4). If g has an elementary divisor of the form f⁢(t)k with k odd, then g is real in SUn⁢(q) by Proposition 4.5. Suppose g has no such elementary divisor, and write g=s⁢u in terms of its Jordan decomposition. If g has no elementary divisor of the form f⁢(t)k with k odd, then it follows from Proposition 3.1 that u has no elementary divisors of the form (t-1)k with k odd. By Corollary 5.7, it follows that u is not real in SUn⁢(q). But if g is then real in SUn⁢(q) with reversing element h, we would have

h⁢g⁢h-1=g-1=s-1⁢u-1=(h⁢s⁢h-1)⁢(h⁢u⁢h-1),and so u-1=h⁢u⁢h-1

by the uniqueness of the Jordan decomposition, a contradiction. Thus g is not real in SUn⁢(q).

Finally, assume that q≡3⁢(mod⁡4) and 4|n. Write g=s⁢u in its Jordan decomposition. The main point is that the condition in statement (3) holds if and only if the conditions in Lemma 5.6 hold, by applying Proposition 3.1. Statement (3) then follows immediately from Lemma 6.1. ∎

7 Strongly real elements and special orthogonal groups

We finally address the question of strong reality in SUn⁢(q). Just as embedded orthogonal subgroups On±⁢(q) play a key role in the strong reality of GUn⁢(q), the embedded subgroups SOn±⁢(q) are key in the strong reality of SUn⁢(q).

Theorem 7.1

Let g∈SUn⁢(q), and suppose g is strongly real in GUn⁢(q).

If n≢2⁢(mod⁡4) or q is even, then g is strongly real in SUn⁢(q).
If n≡2⁢(mod⁡4) and q is odd, then g is strongly real in SUn⁢(q) if and only if g has some elementary divisor of the form (t±1)k with k odd.

Proof.

Let h∈GUn⁢(q) be a reversing element of g with h2=1, so that det⁢(h) equals ±1. If q is even, then necessarily det⁢(h)=1 and h∈SUn⁢(q), so g is strongly real in SUn⁢(q). So we now assume that q is odd.

By Theorem 2.1, if q is odd and g∈GUn⁢(q) is strongly real, then every elementary divisor of g of the form (t±1)2⁢k appears with even multiplicity, and in particular g is the element of some embedded orthogonal group On±⁢(q). If n is odd or divisible by 4, it follows from Lemma 4.7 that g is reversed by an element that squares to the identity from SOn±⁢(q). In particular, g is reversed by an element that squares to the identity from SUn⁢(q), and so is strongly real in SUn⁢(q). If n≡2⁢(mod⁡4), and if g has some elementary divisor of the form (t±1)k with k odd, then it also follows from Lemma 4.7 that g is strongly real in SUn⁢(q). If g has no such elementary divisor when n≡2⁢(mod⁡4), then consider g as an element of SLn⁢(q2). It follows from Theorem 2.3 that g is not strongly real in SLn⁢(q2), so it cannot be strongly real as an element of the subgroup SUn⁢(q). ∎

As we mentioned after Lemma 4.7, it was noticed by Gow that if q is odd and n is odd or 4|n, then every element of SOn±⁢(q) is strongly real. We can complete the classification of real and strongly real classes of SOn±⁢(q) for q odd as follows.

Theorem 7.2

Let q be odd, n≡2⁢(mod⁡4), and g∈SOn±⁢(q). Then g is real if and only if g is strongly real, and g is real if and only it has an elementary divisor of the form (t±1)k with k odd.

Proof.

We first show that g is strongly real in SOn±⁢(q) if and only if g has an elementary divisor of the form (t±1)k with k odd. By Lemma 4.7, if g has an elementary divisor of this form, then a reversing element that squares to the identity can be found in On±⁢(q) with determinant 1, so g is strongly real in SOn±⁢(q). Conversely, if g has no elementary divisor of the form (t±1)k with k odd, then g cannot be strongly real as an element of SLn⁢(q) by Theorem 2.3 (ii), so g cannot be strongly real in the subgroup SOn±⁢(q).

We now show g is real in SOn±⁢(q) if and only if g has an elementary divisor of the form (t±1)k. If g has such an elementary divisor, then g is strongly real from the above, and so g is real in SOn±⁢(q). So assume that g has no such elementary divisor. Since n≡2⁢(mod⁡4), it follows from Lemma 4.7 that there is an inverting element h of g satisfying h2=1 such that h∈On±⁢(q) and det⁢(h)=-1. Any other inverting element of g in On±⁢(q) must be of the form ha for some a∈COn±⁢(q)⁢(g). By [3, Proposition 16.24], a conjugacy class of On±⁢(q) which is contained in SOn±⁢(q) splits into two conjugacy classes of SOn±⁢(q) if and only if the elements of that class have no elementary divisors of the form (t±1)k with k odd. At the same time, by considering corresponding orbits and stabilizers, a class of On±⁢(q) contained in SOn±⁢(q) splits into two conjugacy classes of SOn±⁢(q) if and only if the centralizer of those elements in On±⁢(q) are completely contained in SOn±⁢(q) (as mentioned in [3, proof of Proposition 16.23]). Thus, any element a∈COn±⁢(q)⁢(g) satisfies det⁢(a)=1, and so for any reversing element ha of g in On±⁢(q), we have det⁢(h⁢a)=-1. Thus g cannot be real in SOn±⁢(q). ∎

In combining the statements in Theorems 7.1 and 7.2, we obtain the following result on strong reality in SUn⁢(q) for q odd, which pleasingly resembles strong reality in GUn⁢(q) for q odd as in the last part of Theorem 2.1.

Corollary 7.3

Let q be odd, and let g∈SUn⁢(q). Then g is strongly real in SUn⁢(q) if and only if g is strongly real as an element of some embedded SOn±⁢(q).

Communicated by Robert M. Guralnick

Funding source: Simons Foundation

Award Identifier / Grant number: #351233

Award Identifier / Grant number: #280496

Funding statement: Schaeffer Fry and Vinroot were each supported in part by a grant from the Simons Foundation (Awards #351233 and #280496, respectively).

Acknowledgements

The authors thank Bob Guralnick, Nick Gill, and Anupam Singh for helpful conversation and communication, and the anonymous referee for helpful suggestions.

References

[1] Brauer R., Representations of finite groups, Lectures on Modern Mathematics. Vol. I, Wiley, New York (1963), 133–175. Search in Google Scholar

[2] Britnell J. R. and Wildon M., On types and classes of commuting matrices over finite fields, J. Lond. Math. Soc. (2) 83 (2011), no. 2, 470–492. 10.1112/jlms/jdq088Search in Google Scholar

[3] Cabanes M. and Enguehard M., Representation Theory of Finite Reductive Groups, New Math. Monogr. 1, Cambridge University Press, Cambridge, 2004. 10.1017/CBO9780511542763Search in Google Scholar

[4] Djoković D. Ž., Product of two involutions, Arch. Math. (Basel) 18 (1967), 582–584. 10.1007/BF01898863Search in Google Scholar

[5] Ennola V., On the conjugacy classes of the finite unitary groups, Ann. Acad. Sci. Fenn. Ser. A I 313 (1962). 10.5186/aasfm.1962.313Search in Google Scholar

[6] Fong P. and Srinivasan B., The blocks of finite general linear and unitary groups, Invent. Math. 69 (1982), no. 1, 109–153. 10.1007/BF01389188Search in Google Scholar

[7] Gates Z., Singh A. and Vinroot C. R., Strongly real classes in finite unitary groups of odd characteristic, J. Group Theory 17 (2014), no. 14, 589–617. 10.1515/jgt-2014-0010Search in Google Scholar

[8] Gill N. and Singh A., Real and strongly real classes in PGLn⁢(q) and quasi-simple covers of PSLn⁢(q), J. Group Theory 14 (2011), no. 3, 461–489. 10.1515/jgt.2010.055Search in Google Scholar

[9] Gill N. and Singh A., Real and strongly real classes in SLn⁢(q), J. Group Theory 14 (2011), no. 3, 437–459. 10.1515/jgt.2010.054Search in Google Scholar

[10] Gow R., Real-valued and 2-rational group characters, J. Algebra 61 (1979), no. 2, 388–413. 10.1016/0021-8693(79)90288-6Search in Google Scholar

[11] Gow R., Two multiplicity-free permutation representations of the general linear group GL⁢(n,q2), Math. Z. 188 (1984), 45–54. 10.1007/BF01163871Search in Google Scholar

[12] Gow R., Real representations of the finite orthogonal and symplectic groups of odd characteristic, J. Algebra 96 (1985), no. 1, 249–274. 10.1016/0021-8693(85)90049-3Search in Google Scholar

[13] Liebeck M. W. and Seitz G. M., Unipotent and Nilpotent Classes in Simple Algebraic Groups and Lie Algebras, Math. Surveys Monogr. 180, American Mathematical Society, Providence, 2012. 10.1090/surv/180Search in Google Scholar

[14] Macdonald I. G., Numbers of conjugacy classes in some finite classical groups, Bull. Aust. Math. Soc. 23 (1981), no. 1, 23–48. 10.1017/S0004972700006882Search in Google Scholar

[15] Schaeffer Fry A. A., Irreducible Representations of finite groups of Lie type: On the irreducible restriction problem and some local-global conjectures, Ph.D. thesis, University of Arizona, 2013. Search in Google Scholar

[16] Springer T. A. and Steinberg R., Conjugacy classes, Seminar on Algebraic Groups and Related Finite Groups (Princeton 1968/69), Lecture Notes in Math. 131, Springer, Berlin (1970), 167–266. 10.1007/BFb0081546Search in Google Scholar

[17] Tiep P. H. and Zalesski A. E., Minimal characters of the finite classical groups, Comm. Algebra 24 (1996), no. 6, 2093–2167. 10.1080/00927879608825690Search in Google Scholar

[18] Tiep P. H. and Zalesski A. E., Real conjugacy classes in algebraic groups and finite groups of Lie type, J. Group Theory 8 (2005), no. 3, 291–315. 10.1515/jgth.2005.8.3.291Search in Google Scholar

[19] Vdovin E. P. and Gal’t A. A., Strong reality of finite simple groups, Sib. Math. J. 51 (2010), no. 4, 610–615. 10.1007/s11202-010-0062-zSearch in Google Scholar

[20] Wall G. E., On the conjugacy classes in the unitary, orthogonal and symplectic groups, J. Aust. Math. Soc. 3 (1962), 1–62. 10.1017/S1446788700027622Search in Google Scholar

[21] Wonenburger M., Transformations which are products of two involutions, J. Math. Mech. 16 (1966), 327–338. 10.1512/iumj.1967.16.16023Search in Google Scholar

Received: 2015-12-12

Revised: 2016-1-28

Published Online: 2016-3-10

Published in Print: 2016-9-1

Articles in the same Issue

https://doi.org/10.1515/jgth-2016-0009