arXiv:1809.04834v1 [math.GR] 13 Sep 2018
A Note on Commuting
Involution Graphs in
Affine Coxeter Groups
By
Sarah Hart and Amal Sbeiti Clarke
Birkbeck Mathematical Sciences Preprint Series
Preprint Number 43
www.bbk.ac.uk/ems/research/pure/preprints
A Note on Commuting Involution Graphs in Affine
Coxeter Groups
Sarah Hart and Amal Sbeiti Clarke
1
Introduction
Let G be a group and X a set of involutions of G. The commuting involution graph
C(G, X) is the graph whose vertex set is X, with vertices x, y joined by an edge whenever
x and y commute. These graphs have been studied for a wide variety of groups, usually
with X being either a conjugacy class or the set of all involutions. Perhaps the most
well-known example of their use was in the work of Fischer on 3-transposition groups [5].
In a series of papers, Bates et al looked at connectedness and diameters of commuting
involution graphs in the case where X is a conjugacy class of involutions and G is the
symmetric group [1], a finite Coxeter group [2], a linear group [3], or a sporadic simple
group [4]. In particular, for finite Coxeter groups they gave necessary and sufficient conditions under which the commuting involution graphs are connected, along with bounds
for the diameters in the connected cases. For the symmetric group, where C(G, X) is
connected the diameter is at most 4. For finite Coxeter groups this bound increases to
5. In [9], affine Coxeter groups of type Ãn were considered. Here, the diameter of any
connected commuting involution graph is at most 6.
More recently, Hart and Sbeiti Clarke considered the remaining classical Weyl groups
(see [7], [6] and [11]). They showed that if G is a classical affine Weyl group, then if
C(G, X) is connected, its diameter exceeds the rank of G by at most 1. The obvious next
question is: what happens in the exceptional affine groups? The purpose of this short
note is to establish some general results for commuting involution graphs in affine Coxeter
groups, and to deal with types F̃4 and G̃2 . Types Ẽ6 Ẽ7 and Ẽ8 are more substantial and
these will be addressed in a forthcoming paper.
Section 2 contains preliminaries and general results. Section 3 deals with F̃4 and G̃2 .
2
Preliminaries and General Results
Let W be a finite Weyl group with root system Φ in a Euclidean vector space V ∼
= Rn , and
2α
, for α ∈ Φ. The affine Weyl group W̃ is the semidirect
Φ∨ the set of coroots α∨ = hα,αi
1
product of W with translation group Z of the coroot lattice L(Φ∨ ) of W . We often
express roots and coroots in terms of the standard basis {e1 , . . . , en }, and we will, by a
slight abuse of notation, write elements of Z simply as vectors – in other words we will
identify Z with L(Φ∨ ). For any w ∈ W̃ , w is written in the form (a, u) where a ∈ W and
u ∈ Z. See, for example, [8, Chapter 4] for more detail.
For a, b ∈ W and u, v ∈ Z we have
(a, u)(b, v) = (ab, ub + v).
We have (a, u)−1 = (a−1 , −ua ). In W̃ , the element (a, u) is conjugate to (b, v) via some
(g, w) if and only if
−1
(b, v) = (a, u)(g,w) = (g −1 ag, ug + w − wg
−1 ag
).
The reflections of W̃ are the affine reflections sα,k (α ∈ Φ, k ∈ Z). Recall that, for v
in V ,
hα, vi
α = v − hα, viα∨
hα, αi
(hα, vi − k)
sα,k (v) = v − 2
α = sα (v) + kα∨ .
hα, αi
sα (v) = v − 2
If R is a set of simple reflections for W and α̃ is the highest root (that is, the root with
the highest coefficient sum when expressed as a linear combination of simple roots), then
it can be shown that R ∪ {sα̃,1 } is a set of simple reflections for W̃ .
Finally, we write Diam C(G, X) for the diameter of C(G, X) when C(G, X) is a connected graph, in other words the maximum distance d(x, y) between any x, y ∈ X in the
graph.
We begin with an observation about connectedness. For an element x = (a, u) in a
conjugacy class X of W̃ , we define x̂ = a. Then let X̂ be the conjugacy class of x̂ in W .
Clearly if x, y ∈ X, then x̂, ŷ ∈ X̂.
Lemma 2.1. Suppose x, y ∈ X. If d(x̂, ŷ) = k, then d(x, y) ≥ k. If C(W, X̂) is disconnected, then C(W̃ , X) is disconnected.
Proof. The result follows immediately from the observation that if x commutes with y in
W̃ , then x̂ commutes with ŷ in W .
Definition 2.2. Let W be an arbitrary Coxeter group, with R the set of simple reflections.
Two subsets I and J of R are W -equivalent if there exists w ∈ W such that I w = J.
In the next result, we use the notation wI for the longest element of a finite standard
parabolic subgroup WI .
2
Theorem 2.3 (Richardson [10]). Let W be an arbitrary Coxeter group, with R the set of
simple reflections. Let g ∈ W be an involution. Then there exists I ⊆ R such that wI is
central in WI , and g is conjugate to wI . In addition, for I, J ⊆ R, wI is conjugate to wJ
if and only if I and J are W -equivalent.
For the rest of this paper, W will always denote a finite Weyl group with root system
Φ, with W̃ its corresponding affine Weyl group. We will use the convention that a Weyl
group of type Γ will be denoted W (Γ), where Γ is the associated Coxeter graph. The next
two lemmas give conditions under which reflections w and w′ of W̃ commute.
Lemma 2.4. Let α, β ∈ Φ. Then sα commutes with sβ if and only if hα, βi = 0 or
α = ±β.
Proof Consider v ∈ V and α, β ∈ Φ. A quick calculation shows that
2hv, αi
α−
hα, αi
2hv, αi
α−
sα sβ (v) = v −
hα, αi
sβ sα (v) = v −
2hv, βi
β+
hβ, βi
2hv, βi
β+
hβ, βi
4hv, αihα, βi
β;
hα, αihβ, βi
4hv, βihα, βi
α.
hβ, βihα, αi
Hence sα sβ = sβ sα if and only if either hα, βi = 0 or hv, βiα = hv, αiβ for all v ∈ V .
One of the properties of root systems is that for any root γ we have hγi ∩ Φ = {±γ}.
Therefore sα sβ = sβ sα if and only if hα, βi = 0 or α = ±β.
Lemma 2.5. For all positive roots α and β, and all integers k and l, the affine reflections
sα,k and sβ,l commute if and only if either hα, βi = 0, or α = β and k = l.
Proof. Let α and β be positive roots, with k and l integers. Then
sα,k sβ,l (v) = sα,k (sβ (v) + lβ ∨ ) = sα sβ (v) + sα (lβ ∨ ) + kα∨
= sα sβ (v) + kα∨ + lβ ∨ − lhα, β ∨ iα∨ ;
sβ,l sα,k (v) = sβ sα (v) + kα∨ + lβ ∨ − khα∨ , βiβ ∨ .
Consequently sα,k sβ,l = sβ,l sα,k precisely when, for all v ∈ V , we have
sα sβ (v) − lhα, β ∨ iα∨ = sβ sα (v) − khα∨ , βiβ ∨ .
In particular, setting v = 0 we get lhα, β ∨ iα∨ = khα∨ , βiβ ∨ , which implies that either
α = β or hα, βi = 0. If α = β, we get sα,k sα,l (v) = v + (l − k)α∨ , whereas sα,l sα,k (v) =
v + (k − l)α∨ . Therefore, sα,k commutes with sα,l if and only if k = l. On the other hand,
if α 6= β, then hα, βi = 0 and so sα sβ = sβ sα . Thus sα,k sβ,l = sβ,l sα,k .
There is one case we can deal with that occurs in several groups.
Lemma 2.6. Suppose W is a finite Weyl group that has a central involution w̃. Now
suppose x is an involution in the corresponding affine Weyl group W̃ such that x = (w̃, u)
for some u ∈ Z, and write X = xW̃ , the conjugacy class in W̃ of x. Then C(W̃ , X) is
disconnected.
3
Proof. Since w̃ is central in W , every conjugate of x in W̃ has the form y = (w̃, v) for
some u in Z. Moreover w̃ acts as −1 on the root system, and hence on Z. Now
xy = (w̃, u)(w̃, v) = (w̃2 , uw̃ + v) = (1, v − u).
Thus xy is translation through v − u. Now x commutes with y precisely when (xy)2 = 1,
and hence x commutes with y if and only if u = v, meaning x does not commute with any
other member of its conjugacy class. Therefore C(W̃ , X) is completely disconnected.
Next we have a result which proves connectedness in certain circumstances. We write
x ↔ y to mean xy = yx.
Proposition 2.7. Let X be a conjugacy class of involutions in W̃ that contains (a, 0) for
some involution a of W . Suppose that C(W, X̂) is connected with diameter d, and further
that there is an integer k such that whenever (a, u) ∈ X, there is some x̂ ∈ X̂ such that
d((a, u), (x̂, 0)) ≤ k in C(W̃ , X). Then C(W̃ , X) is connected with diameter at most d + k.
Proof. Let (b, v) ∈ X. Then b is conjugate to a in W . That is, there is some g ∈ W
−1
with b = ag , meaning that (b, v)g is of the form (a, u) for an appropriate u. By
hypothesis then, there is a path (a, u) = x0 ↔ x1 · · · ↔ xm = (x̂, 0) of elements xi of
X, with m ≤ k. But this implies that there is a path xg0 ↔ xg1 ↔ · · · ↔ xgm in the
commuting involution graph. Writing y = x̂g , and noting that xg0 = (b, v), we obtain
d((b, v), (y, 0)) ≤ k. Now C(W, X̂) is connected with diameter d. Thus d(y, a) ≤ d in
C(W, X̂). Hence d((y, 0), (a, 0)) ≤ d in C(W̃ , X). Therefore d((b, v), (a, 0)) ≤ d + k and
since this holds for all (b, v) in X, we deduce that C(W̃ , X) is connected with diameter
at most d + k.
Lemma
2.8. Suppose a ∈ WI for some I ⊆ R. If (a, u) is an involution, with u =
P
u
α
/PI. Moreover, let J be the
r∈R r r , then a is an involution and ur = 0 whenever r ∈
set of reflections s of R that commute with all r in I. If v = s∈J vs αs , then vb = v for
all b ∈ WI .
Proof. We have (a, u)(a, u) = (a2 , ua +u). If (a, u) is an involution, then a is an involution
and ua + u = 0. Now a is a product of elements of I. We have vr = v − hv, αr iαr∨ for
all u ∈ Z and r ∈ R. Hence, inductively, ua = u − x for some x ∈ hαr : r ∈ Ii. So
ua + u = 2u − x. For this to equal zero, clearly u ∈ hαr : r ∈ Ii. That is, ur = 0 whenever
r∈
/ I. For the second part, observe that for any r ∈ I and s ∈ J we have hαr , αs i = 0.
Therefore vr = v for all r ∈ I. Hence vb = u for all b ∈ WI .
Lemma 2.9. Let (a, u) be an involution in W̃ , with X = (a, u)W̃ . Suppose b commutes
with a in W , where b ∈ aW . Then there is (b, v) ∈ X such that (a, u) ↔ (b, v).
Proof. Since b is conjugate in W to a, there is g ∈ W with b = g −1 ag. Let w = 21 (u − ug ).
Then
(a, u)(g,w) = (g −1 ag, ug + w − wg
4
−1 ag
) = (b, ug + w − wb ).
Set v = ug + w − wb . We claim that (a, u) ↔ (b, v). Certainly (b, v) ∈ X. Note that
since (a, u) is an involution, ua = −u; also since g −1 ag = b we have gb = ag and so
ugb = uag = −ug . Thus
b
v = ug + w − wb = ug + 12 (u − ug ) − 21 (ub − ug )
= ug + 12 u − 12 ug − 21 ub − 12 ug
= 12 (u − ub ).
Now
(a, u)(b, v) = (ab, ub + v) = (ab, ub + 12 (u − ub )) = (ab, 21 (u + ub ))
(b, v)(a, u) = (ba, va + u) = (ab, 12 ua − 21 uba + u) = (ab, 21 (u − uab )) = (ab, 21 (u + ub )).
Thus (a, u) ↔ (b, v), as required.
Proposition 2.10. Let X be a conjugacy class of involutions in W̃ containing (a, u).
Suppose there is an integer k such that whenever (a, u′ ) ∈ X, we have d((a, u′ ), (a, u)) ≤ k,
and also that C(W, X̂) is connected with diameter d. Then C(W̃ , X) is connected with
diameter at most d + k.
Proof. Let (b, v) ∈ X. By hypothesis C(W, X̂) is connected with diameter d, meaning
d(a, b) ≤ d. Hence, by Lemma 2.9, there is a corresponding path (b, v) ↔ · · · ↔ (a, u′ ),
for appropriate u′ , in C(W̃ , X) of length at most d. By hypothesis d((a, u′ ), (a, u)) ≤ k.
Therefore, d((b, v), (a, u)) ≤ d + k.
3
Types F̃4 and G̃2
Let W be of type F4 , with associated root system Φ in R4 and simple roots {α1 , α2 , α3 , α4 }.
The root system Φ consists of the 24 long roots ±ei ± ej (1 ≤ i < j ≤ 4) and 24 short
roots: eight of the form ±ei and sixteen of the form 21 (±e1 ± e2 ± e3 ± e4 ). We may set
α1 = 12 (e1 − e2 − e3 − e4 ), α2 = e4 , α3 = e3 − e4 and α4 = e2 − e3 . The highest root α̃ is
then e1 + e2 . Then W = hsα1 , sα2 , sα3 , sα4 i and for the simple reflections of W̃ we can take
ri = (sαi , 0) for i ∈ {1, 2, 3, 4} and r5 = (sα̃ , α̃∨ ). The Coxeter graph for W̃ is as shown
in Figure 1. (The subgroup hr1 , r2 , r3 , r4 i is of course isomorphic to W and we may for
convenience identify it with W on occasion.)
t
r1
t
r2
t
r3
t
r4
t
r5
Figure 1: Coxeter graph for type F̃4
Information about the commuting involution graphs of W was obtained in [2]. There
are seven conjugacy classes of involutions in W . Let a = (sα2 sα3 )2 . Then C(W, X) is
connected with diameter 2. Apart from this, and the graph consisting of just the central
5
Graph
A1
A1
A21
A21
B2
B3
B3
A31
B2 × A1
F4
B4
B3 × A1
Representative I
{r1 }
{r3 }
{r1 , r3 }
{r3 , r5 }
{r2 , r3 }
{r1 , r2 , r3 }
{r2 , r3 , r4 }
{r1 , r3 , r5 }
{r2 , r3 , r5 }
{r1 , r2 , r3 , r4 }
{r2 , r3 , r4 , r5 }
{r1 , r2 , r3 , r5 }
Underlying class in W
A1
A1
A21
B2
B2
B3
B3
B3
B3
F4
F4
F4
Table 1: Conjugacy classes in F̃4
involution, all the other commuting involution graphs are disconnected. See [2] for more
details.
Now let us consider the affine group W̃ . By Theorem 2.3, every involution conjugacy
class corresponds to a standard parabolic subgroup WI with a central involution wI . We
can identify the possibilities by finding subgraphs of the Coxeter graph for F̃4 which
correspond to Coxeter groups having nontrivial centres. Table 1 shows, for each subgraph
giving rise to a distinct involution conjugacy class X, a representative I ⊆ {r1 , . . . , r5 }
for which wI ∈ X, along with the name, in the third column, of the Coxeter graph for the
underlying class X̂ in W . To determine the Coxeter graph corresponding to X̂, note that
if I ⊆ W , then X̂ has the same graph as X. If x = (a, u) is a product of k reflections in W̃ ,
then a must be a product of k reflections in W , so, for example, the conjugacy class of W̃
corresponding to the B4 graph must have the class corresponding to F4 as its underlying
class in W . The only instance where this does not immediately tell us the type of the
underlying class is the case of A21 when I = {r3 , r5 }. Here, the underlying class might be
type B2 or type A21 . Note that r3 = se3 −e4 and r5 = se1 +e2 ,1 . So the underlying class is the
conjugacy class of se3 −e4 se1 +e2 in W . One can check that r1 r4 r3 r2 r3 r4 (e1 +e2 ) = e3 +e4 and
r1 r4 r3 r2 r3 r4 (e3 − e4 ) = e3 − e4 . Hence (se3 −e4 se1 +e2 )r4 r3 r2 r3 r4 r1 = se3 −e4 se3 +e4 = se3 se4 =
r3 r2 r3 r2 = (r3 r2 )2 . Consequently the underlying conjugacy class in this case is of type
B2 .
Theorem 3.1. Let W̃ be of type F̃4 , with graph shown in Figure 1. If X is the conjugacy class of (r2 r3 )2 or r3 r5 in F̃4 , then C(W̃ , X) is connected with diameter at most 4.
Otherwise, C(W̃ , X) is disconnected.
Proof. Let X be a conjugacy class in W̃ . If the underlying class in W is anything other
than type B2 or F4 , then C(W̃ , X) is disconnected, by Lemma 2.1. If the underlying class is
6
type F4 then C(W̃ , X) is disconnected by Lemma 2.6. So we are reduced to the case where
the underlying class in W is type B2 . There are two classes in W̃ where this happens,
one containing (r2 r3 )2 = ((sα2 sα3 )2 , 0), and one containing r3 r5 = ((sα2 sα3 )2 , α̃∨ ).
Let X be the conjugacy class of (r2 r3 )2 in W̃ ; its underlying class X̂ in W is of type
B2 . Note that X̂ contains ((sα2 sα3 )2 = se3 se4 and thus also se1 se2 (for example via the
conjugating element se1 −e3 se2 −e4 ). In Proposition 2.7, set a = se1 se2 and x̂ = se3 se4 .
Suppose (a, u) ∈ X. Now (a, u)(x̂, 0) = (se1 se2 se3 se4 , v) for the appropriate v. This is
clearly an involution because se1 se2 se3 se4 acts as −1 on Z. Thus (x̂, 0) commutes with
(a bu). Therefore we can apply Proposition 2.7 with k = 1 and d = 2, to see that C(W̃ , X)
is connected with diameter at most 3.
Now let X be the conjugacy class of r3 r5 in W̃ ; again its underlying class in W is of type
B2 . Let (a, u), (a, u′ ) ∈ X, where again a = se3 se4 . By Lemma 2.9 there is some v ∈ Z
for which (a, u) ↔ (se1 se2 , v). But now (a, u′ )(se1 se2 , v) = (se1 se2 se3 se4 , v′ ) for some v′ ,
which is an involution. Thus in Proposition 2.10 we have k = 2 and d = 2, meaning that
C(W̃ , X) is connected with diameter at most 4.
The Coxeter graph of type G̃2 is as follows.
r1
r2
6 ✉
✉
G̃2
r3
✉
The subgraphs corresponding to parabolic subgroups WI for which wI is central are
of types A1 , A1 × A1 or G2 . It turns out that in all these cases, the commuting involution
graphs are disconnected.
Proposition 3.2. Let X be a conjugacy class in the affine Coxeter group W̃ of type G̃2 .
Then C(W̃ , X) is disconnected.
Proof. When W̃ is of type G̃2 , the underlying Weyl group W is dihedral of order 12. It has
three conjugacy classes of involutions: two classes of reflections each with three elements,
and one class consisting of the unique central involution w̃. The commuting involution
graphs for the classes of reflections in W are completely disconnected. Now let x ∈ X. If
x̂ is a reflection, then C(G, X) is disconnected by Lemma 2.1. Otherwise, x̂ = w̃, and so
the disconnectedness of C(G, X) follows from Proposition 2.6.
References
[1] C. Bates, D. Bundy, P. Rowley and S. Perkins. Commuting involution graphs for
symmetric groups, J. Algebra, 266(1) (2003), 133–153.
[2] C. Bates, D. Bundy, P. Rowley and S. Perkins. Commuting Involution Graphs for
Finite Coxeter Groups, J. Group Theory 6 (2003), 461–476.
[3] C. Bates, D. Bundy, P. Rowley and S. Perkins. Commuting involution graphs in
special linear groups. Comm. in Algebra, 32(11) (2004), 4179-4196.
7
[4] C. Bates, D. Bundy, P. Rowley and S. Hart. Commuting involution graphs for sporadic
simple groups. J. Algebra, 316(2) (2007), 849-868.
[5] B. Fischer. Finite groups generated by 3-transpositions, I. Invent. Math. 13 (1971),
232–246.
[6] S. Hart and A. Sbeiti Clarke. Commuting Involution Graphs for C̃n , Comm. in Algebra 46 (9) (2018), 3965–3985.
[7] S. Hart and A. Sbeiti Clarke. Commuting Involution Graphs in Classical
Affine Groups, Birkbeck Mathematical Sciences Preprint Series No. 42 (2018).
http://www.bbk.ac.uk/ems/research/pure/preprints
[8] J.E. Humphreys. Reflection Groups and Coxeter Groups, Cambridge Studies in Advanced Mathematics, 29 (1990).
[9] S. Perkins. Commuting Involution Graphs in the affine Weyl group Ã, Arch. Math.
86 (2006), no. 1,16-25.
[10] R. W. Richardson. Conjugacy Classes of Involutions in Coxeter Groups, Bull. Austral. Math. Soc. 26 (1982), 1–15.
[11] A. Sbeiti Clarke. Affine Coxeter Groups, Conjugacy Classes and Commuting Involution Graphs, Ph.D. Thesis, Birkbeck (University of London), 2018.
8