Journal of Algebra 522 (2019) 101–123

Contents lists available at ScienceDirect

Journal of Algebra
www.elsevier.com/locate/jalgebra

Lie polynomials in q-deformed Heisenberg algebras

Rafael Reno S. Cantuba
Mathematics and Statistics Department, De La Salle University, Manila,
Philippines

a r t i c l e i n f o a b s t r a c t

Article history: Let F be a field, and let q ∈ F. The q-deformed Heisenberg al-
Received 30 October 2017 gebra is the unital associative F-algebra H(q) with generators
Available online 21 December 2018 A, B and relation AB −qBA = I, where I is the multiplicative
Communicated by Alberto Elduque
identity in H(q). The set of all Lie polynomials in A, B is the
MSC: Lie subalgebra L(q) of H(q) generated by A, B. If q = 1 or the
17B60 characteristic of F is not 2, then the equation AB − qBA = I
16S15 cannot be expressed in terms of Lie algebra operations only,
81R50 yet this equation still has consequences on the Lie algebra
05A30 structure of L(q), which we investigate. We show that if q
is not a root of unity, then L(q) is a Lie ideal of H(q), and
Keywords: the resulting quotient Lie algebra is infinite-dimensional and
q-Deformed Heisenberg algebras one-step nilpotent.
Lie subalgebra © 2018 Elsevier Inc. All rights reserved.
Ideal of a Lie algebra

1. Introduction

The q-deformed Heisenberg algebras are said to be algebraic structures with rich
properties as mathematical objects and have many important applications in physics
and beyond [5, p. vii]. The q-deformed Heisenberg algebra H(q) is an algebraic formalism
of the so-called q-deformed canonical commutation relation. This is an equation AB −
qBA = I for some fixed scalar deformation parameter q. A common realization is that of
A and B being interpreted as mutually adjoint linear operators (with I as the identity

E-mail address: [email protected].

https://doi.org/10.1016/j.jalgebra.2018.12.008
0021-8693/© 2018 Elsevier Inc. All rights reserved.
102 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

map) on an infinite-dimensional separable Hilbert space [7, Section 1.3]. Algebraically,

we view H(q) as an associative algebra abstractly defined as having a presentation with
generators A, B and relation AB − qBA = I.
By a simple routine calculation based on a well-known result in the literature about
free Lie algebras [8, Lemma 1.7], the relation AB − qBA = I, in general, cannot be
expressed in terms of Lie algebra operations only. That is, in general, AB − qBA − I
is not a Lie polynomial in the free unital associative algebra in two generators A, B.
However, the associative algebra structure in H(q) allows for the computation of com-
mutators [X, Y ] := XY − Y X for X, Y ∈ H(q). Thus, we investigate the interplay of
the q-deformed and non-deformed commutator structures in H(q). That is, we investi-
gate the consequences of the deformed commutation relation AB − qBA = I on the Lie
subalgebra L(q) of H(q) generated by A, B.
A main idea that helps us accomplish our goals in this study is a spanning set for L(q)
that is obtained from the theory of free Lie algebras. There is a considerable amount of
literature for the basis of a free Lie algebra on a finite number of generators. These arose
from seminal works such as [4,9,10]. In this work, we use the formulation in [12, Section
2.8] of the concepts in [9,10]. This deals with so-called regular words on the generators.
We recall important results and properties about presentations of associative and Lie
algebras, about regular associative and nonassociative words, and about the algebra H(q)
in Sections 2 and 3. The case q = 1 involves the three-dimensional Lie algebra L(1) with
basis A, B, [B, A] that satisfy the commutation relations [[B, A] , A] = 0 = [B, [B, A]].
The Lie algebra L(1) is well-known. See for instance [5, Section 1.1]. In this work, we
consider the case when q is not a root of unity. Our goal is to show that, under the
said case, L(q) is an ideal of H(q) under the Lie algebra structure (i.e., a Lie ideal). We
further divide this into the case when q = 0 in Section 4, and the case when q is nonzero
and not a root of unity, which is in Section 5.

2. Preliminaries

Let F be a field. Throughout, by an F-algebra we mean a unital associative F-algebra,

with multiplicative identity I. Let A be an F-algebra. Recall that an anti-automorphism
of A is a bijective F-linear map ψ : A → A such that ψ(f g) = ψ(g)ψ(f ) for all f, g ∈ A.
We turn A into a Lie algebra with Lie bracket [f, g] = f g − gf for f, g ∈ A.
We denote the set of all nonnegative integers by N, and the set of all positive integers
by Z+ . Given n ∈ N, let X denote an n-element set. We shall refer to any element of X
as a letter. For t ∈ N, by a word of length t on X we mean a sequence of the form

X1 X2 · · · Xt , (1)

where Xi ∈ X for 1 ≤ i ≤ t. Given a word W on X , denote the length of W by |W |. The

word of length 0 is I. Let X  denote the set of all words on X . Given words X1 X2 · · · Xs
and Y1 Y2 · · · Yt on X , their concatenation product is
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 103

X1 X2 · · · Xs Y1 Y2 · · · Yt .

We now recall the free F-algebra F X . The F-vector space F X  has basis X .
Multiplication in the F-algebra F X  is the concatenation product.
Given n ∈ N, the subspace of F X  spanned by all the words of length n is the
n-homogeneous component of F X . Observe that F X  is the direct sum of all the
n-homogeneous components for all n. If f is an element of the m-homogeneous component
and g is an element of the n-homogeneous component, then f g is an element of the (m +
n)-homogeneous component. It follows that the set of all n-homogeneous components of
F X  for all n is a grading of F X .
The following notation will be useful. Let W = X1 X2 · · · Xt denote a word on X . We
define W ∗ to be the word Xt Xt−1 · · · X1 . Let θ denote the F-linear map

θ : F X  → F X  ,
W → (−1)|W | W ∗ , (2)

for any word W . By [8, p. 19], the map θ is the unique anti-automorphism of the F-algebra
F X  that sends X to −X for any letter X.
For the rest of the section, we fix an ordering < on the elements of X . The order-
ing < can be extended into the so-called lexicographic ordering on X  as follows. Let
W = X1 X2 · · · X|W | and V = Y1 Y2 · · · Y|V | denote nonempty words, where Xi , Yj are
letters for all i, j. We write W < V if for some positive integer t, that does not exceed
the minimum of {|W |, |V |}, we have Xt < Yt and Xi = Yi for all i < t. The term “lexico-
graphic” is suggestive of how the previously described process compares words as to how
they are to be “ordered in a dictionary.” However, the logical rules described have no
sensible conclusion when V = W R for some nonempty word R. There are several ways
of extending the lexicographic ordering to cover the said case. We note that described
in [12, p. 34]. Denote two elements of X  by W = X1 X2 · · · X|W | and V = Y1 Y2 · · · Y|V |
such that Xi , Yi ∈ X for any i. Then in the free algebra F X , we have W = V if and
only if |W | = |V | and Xi = Yi for any i. We define ∼ to be the relation on X  given
by W ∼ V if and only if W V = V W . An immediate consequence is that W ∼ V if and
only if W and V are powers of the same word, and this fact can be used to show that
∼ is an equivalence relation. We define the relation  on X  by V  W if and only
if V W > W V . Under this new relation on X , the words W, V do not belong to the
same equivalence class under ∼ if and only if either V  W or W  V . Thus,  is a
partial ordering on X . Also, the relation  may be interpreted as a well ordering of
the equivalence classes of ∼.
Write a nonempty word W as W = X1 X2 · · · X|W | where Xi is a letter for any i.
Consider positive integers s, t such that s ≤ t ≤ |W |. We call Xs Xs+1 · · · Xt a subword
of W . The word G = X1 X2 · · · Xt is called a beginning of W , and H = Xt Xt+1 · · · X|W |
an ending of W . If t < |W |, then G and H are a proper beginning and a proper ending
of W , respectively.
104 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

Definition 2.1. A nonempty word W is regular (or is an associative regular word) if for
any proper beginning G and proper ending H of W such that W = GH, we have G  H.

Proposition 2.2 ([12, p. 35]). Let f be a regular word. Then f is maximal among all
its cyclic permutations with respect to the ordering . Furthermore, if h is the length-
maximal nonempty proper ending of f that is also regular, and if g is the word such
that f = gh, then g is regular. As described, any regular word of length at least 2 can be
expressed uniquely as a product of two regular proper subwords.

We now recall the notion of a Lie monomial on X . The set of all Lie monomials on
X is the minimal subset of F X  that contains X and is closed under the Lie bracket.
Observe that 0 is a Lie monomial. Let U be a Lie monomial. Then U is an element of
some n-homogeneous component of F X . Given a Lie monomial U = 0, we find that U
belongs to exactly one n-homogeneous component. Let L denote the Lie subalgebra of
the Lie algebra F X  generated by X . Following [8, Theorem 0.5], we call L the free Lie
algebra on X . An element of L is called a Lie polynomial on X . The following gives us
a useful necessary condition for Lie polynomials on X .

Proposition 2.3 ([8, Lemma 1.7]). For f ∈ L, we have θ(f ) = −f .

We now recall a basis for L that consists of Lie monomials and is related to regular
words.

Definition 2.4. Let f be a regular word. If |f | = 1, then we define Jf K := f . If |f | ≥ 2,

let f = gh as described in Proposition 2.2. We define Jf K := [JgK , JhK]. We call the
Lie monomial Jf K a nonassociative regular word on X . We also say that Jf K is the
nonassociative regular word for the (associative) regular word f .

Proposition 2.5 ([12, p. 36]). The nonassociative regular words on X form a basis for L.

Throughout, by an ideal of an F-algebra A we mean a two-sided ideal of A. By a Lie

ideal of a Lie algebra L we mean an ideal of L under the Lie algebra structure. We now
recall the notion of algebras having generators and relations (i.e., having a presentation).
Denote the elements of X by G1 , G2 , . . . , Gn . Let f1 , f2 , . . . , fm ∈ F X  and let I be the
ideal of F X  generated by f1 , f2 , . . . , fm . We define F X  /I as the F-algebra with
generators G1 , G2 , . . . , Gn and relations f1 = 0, f2 = 0, . . . , fm = 0. The Lie subalgebra
of F X  /I generated by X is L/ (I ∩ L).
Let g1 , g2 , . . . , gm ∈ L and let J be the Lie ideal of L generated by g1 , g2 , . . . , gm .
We define L/J as the Lie algebra with generators G1 , G2 , . . . , Gn and relations g1 = 0,
g2 = 0, . . . , gm = 0.
Suppose L is a Lie algebra (over F) generated by X . Then there exists an ideal K of L
such that L = L/K. Let φ : L → L/K be the canonical Lie algebra homomorphism. Then
L has a spanning set consisting of all vectors of the form φ(U ) where U is a nonassociative
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 105

regular word. We call this spanning set the nonassociative regular words in L. Given a
Lie algebra L and a fixed x ∈ L, recall the adjoint linear map

ad x : L → L,
y → [x, y] .

Recall that given x, y, z ∈ L, the Jacobi identity [x, [y, z]] + [z, [x, y]] + [y, [z, x]] = 0 can
be rewritten in terms of adjoint maps as

(ad [x, y]) (z) = ((ad x) ◦ (ad y)) (z) − ((ad y) ◦ (ad x)) (z). (3)

Given subsets K1 and K2 of a Lie algebra L, we define [K1 , K2 ] as the span of all
vectors of the form [x, y] for some x ∈ K1 and some y ∈ K2 . The derived algebra of L is
 
the Lie algebra L(1) := [L, L]. Set L(0) := L, and we inductively define L(p) := L(p−1) , L
for all p ∈ Z+ . We say that L is p-step nilpotent whenever p is the smallest integer such
that L(p) is the zero Lie algebra. Observe that a Lie algebra is one-step nilpotent if and
only if the Lie bracket of any two basis vectors is zero.
From this point onward, suppose we only have two generators A, B for F X . Any
word W on two letters A, B can be written as a product of powers of A, B. That is,

W = B m1 An1 B m2 An2 · · · B mk Ank , (4)

for some k, m1 , n1 , m2 , n2 , . . . , mk , nk ∈ N. Without loss of generality, we further as-

sume that if k ≥ 2, then mi and nj are nonzero for any i ∈ {2, 3, . . . , k} and any
j ∈ {1, 2, . . . , k − 1}. By doing so, any such word W , if nonempty, is paired uniquely
with the integer k in (4). If W is the empty word I, we view k = 0. Thus, we define the
function κ : X  → Z given by the rule W → k whenever W is written as (4), with the
exponents as previously described. Throughout, we use the ordering A < B to construct
the regular words on A, B. From these facts, we have an immediate consequence that
if the word W on A, B is regular, then both m1 and nκ(W ) are nonzero. We note that
given a regular word W , the number κ(W ) is precisely the number of occurrences of the
word BA as a subword of W , which is usually denoted in the literature as degBA (W ).
i.e., κ(W ) = degBA (W ). This notation is from [12, p. 27]. The following immediate
consequence of Definition 2.4 on X = {A, B} shall be useful.

Lemma 2.6. Let m, n ∈ N. The following hold in the free Lie algebra over F generated by
A, B.
q y
[JBAn K , A] = BAn+1 , (5)
q y
[B, JB m An K] = B m+1 An , (m, n ≥ 1). (6)

Denote by Iq the ideal of F X  generated by AB − qBA − I. Thus, the q-deformed

Heisenberg algebra is H(q) = F X  /Iq . By contraposition of Proposition 2.3, it is routine
106 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

to show that if q = 1 or char F = 2, then AB − qBA − I ∈ / L. This implies that the

q-deformed commutation relation AB −qBA = I, under the said cases, is not expressible
in terms of Lie algebra operations only. Equivalently, L has, by definition, no Lie ideal
generated by AB − qBA − I. Our goal is to study the Lie algebra L/ (Iq ∩ L). We shall
refer to the elements of L/ (Iq ∩ L) as the Lie polynomials in H(q).

3. Some properties of H(q)

We recall the following expressions from the so-called q-special combinatorics [5, Sec-
tion C.1]. For a given n ∈ Z and z ∈ F,


n−1
{n}z := zl, (7)
l=0


n
{n}z ! := {l}z , (8)
l=1
 
n {n}z !
:= . (9)
i z {i}z !{n − i}z !

If n ≤ 0, then we interpret (7) as the empty sum 0 and (8) as the empty product 1. One
of the most important properties of H(q) is that the vectors B m An (m, n ≥ 0) form a
basis for H(q) [5, Theorem 3.1]. The proof is by the Diamond Lemma for Ring Theory
[1]. Products of these basis elements can be written as linear combinations of the same
using the following “reordering formulae” from [6, Proposition 4.1]. If n ∈ Z+ , then

AB n = q n B n A + {n}q B n−1 , (10)

A B = q BA + {n}q A
n n n n−1
, (11)
BAn = q −n An B − q −1 {n}q−1 An−1 , (q = 0), (12)
−n −1
n
B A=q AB − q
n
{n}q−1 B n−1
, (q = 0). (13)

For any subspaces H, K of H(q), define HK := Span {hk | h ∈ H, k ∈ K}. Recall that

a Z-grading of H(q) is a sequence {An }n∈Z of subspaces such that H(q) = n∈Z An
is a direct sum and that Am An ⊆ Am+n [3, p. 202]. For any n ∈ Z, define Hn :=
Span {B k Al | k, l ∈ N, k − l = n}. Then {Hn }n∈Z is a Z-grading of H(q) [6, Proposi-
tion 2.4].

Proposition 3.1 ([6, Corollary 4.5]). If q ∈

/ {0, 1}, then the vectors

k k k
[A, B] , B l [A, B] , [A, B] Al , k ∈ N, l ∈ Z+ ,
k
form a basis for H(q). Furthermore, the vectors [A, B] (k ∈ N) form a basis for the
k
subalgebra H0 of H(q). Given l ∈ Z+ , the vectors B l [A, B] (k ∈ N) form a basis for the
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 107

k
subspace Hl of H(q), while the vectors [A, B] Al (k ∈ N) form a basis for the subspace
H−l .

Some useful relations involving the basis vectors in the above proposition are in the
following.

Proposition 3.2 ([6, pp. 105, 111]). Let n ∈ N. If P is any polynomial in two non-
commuting variables and q = 0, then

P (B, A) [A, B] = [A, B] P (q −n B, q n A).

n n
(14)

Furthermore, if q ∈
/ {0, 1}, then


n  
n n−i n
B n An = q − (q − 1)−n
i
2 (−1)n−i q 2 [A, B] . (15)
i=0
i q

In the computations that will follow, we would also need a relation for expressing
the similar product An B n in terms of the basis vectors given in Proposition 3.1. We
exhibit such in the following proposition, with proof similar to how (15) was proven in
[6, pp. 111–112].

Proposition 3.3. If q ∈
/ {0, 1}, then for any n ∈ N,


n  
i+1 n
An B n = (q − 1)−n
i
(−1)n−i q 2 [A, B] . (16)
i=0
i q

Proof. We first claim that for any n ∈ N,

n 
n−1
q− 2 An B n = AB + q −1 {i}q−1 I . (17)
i=0

If n = 0, then the left side of (17) reduces to the identity I, while the right side is the
empty product, which is also I in the algebra H(q). Thus, (17) holds for n = 0. Suppose
(17) holds for n. We have the following computations.


n
n
AB + q −1 {i}q−1 I = q − 2 An B n (AB + q −1 {n}q−1 I),
i=0
n
= q− 2 An (B n AB + q −1 {n}q−1 B n ). (18)

We use the reordering formula (13) on the expression B n AB in the right side of (18).
n+1
The resulting right side is q − 2 An+1 B n+1 . By induction, (17) holds for any n. By the
defining relation of H(q), we have the equation AB = (q − 1)−1 (q [A, B] − I), which we
108 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

use to rewrite AB + q −1 {i}q−1 I in (17) as a linear combination of [A, B] and I. This

results to

n 
n−1
q− 2 An B n = (q − 1)−1 q [A, B] − q −i I ,
i=0


n−1
= (q − 1)−n q [A, B] − q −i I . (19)
i=0

q [A, B] − q −i I in (19) is the Gauss polynomial

n−1
The product i=0

Gn (q [A, B] ; q −1 ).

That is,
n
An B n = q 2 (q − 1)−n Gn (q [A, B] ; q −1 ). (20)

We recall that an important property of the Gauss polynomial Gn (x; z) on the indeter-
minate x and constant parameter z is the following relation [5, Theorem C.17].


n  
n−i n
Gn (x; z) = (−1)n−i z 2 xi . (21)
i=0
i z

Setting x = q [A, B] and z = q −1 in (21) and substituting in (20), we obtain


n  
n
−n n−i n
(−1)n−i q −
i
n
A B =q n 2 (q − 1) 2 q i [A, B] . (22)
i=0
i q−1

From [5, Theorem C.10], we have the identity n

i q −1 = q −i(n−i) n
i q, which we use to
n i
replace i q −1 in (22). The resulting coefficient of [A, B] in the summation is

 
n−i i− n−i
−i(n−i) n
(−1) q 2 ,
i q

where the exponent of q, by some routine computations, can be simplified into i+1
2 − n
2 .
By these observations, we get (16) from (22). 2

Lemma 3.4. Let m, n ∈ N. The following hold in H(q).

[JBAK , B m An ] = (q − 1)(q n − q m )B m+1 An+1 + (q n − q m )B m An , (23)

[B, B A ] = (1 − q )B
m n n m+1
A − {n}q B A
n m n−1
. (24)
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 109

Proof. We first prove (23). We consider the case in which both m, n are nonzero. The
proofs are similar for the cases in which one of m, n is zero. Observe that

[JBAK , B m An ] = [BA − AB, B m An ] ,

= B(AB m )An − B m (An B)A
−(AB m+1 )An + B m (An+1 B). (25)

We replace all instances of a product of the form AB j or Ai B in (25) by the appropriate

formula from (10) or (11). The result is a linear combination of only the words B m+1 An+1
and B m An , where the scalar coefficient of B m+1 An+1 is q m − q m+1 − q n + q n+1 while
that of B m An is {m}q − {m + 1}q − {n}q + {n + 1}q . By some routine calculations, these
can simplified into (q − 1)(q n − q m ) and q n − q m , respectively. Thus, (23) holds. The
proof for (24) is similar. 2

Our next goal is to derive some relations for expressing the regular nonassociatve
words JBAn K and JB n AK in terms of the basis {B i Aj | i, j ∈ N} of H(q).

Proposition 3.5. For any positive integer n, the following relations hold in H(q).

JBAn K = (1 − q)n BAn − (1 − q)n−1 An−1 , (26)

JB n AK = (1 − q)n B n A − (1 − q)n−1 B n−1 . (27)

Proof. We use induction on n ≥ 1. Using the relation AB − qBA = I, we have JBAK =

BA − AB = (1 − q)BA − I, which proves (26) and (27) for n = 1. Suppose that (26)
holds for some n. If we apply the map −ad A on both sides of the equation, by (5), the
q y
left side becomes BAn+1 while the right side becomes (1 − q)n BAn+1 − (AB n )A ,
in which we replace the expression AB n in the second term using (10). After combining
some terms, the resulting equation serves as proof that (26) holds for n + 1. By a similar
argument, it can be shown that (27) holds for n + 1 if (27) is assumed to hold for n. By
induction, we get the desired result. 2

4. The Lie algebra L(0)

In this section, we show a basis and some corresponding commutation relations for the
Lie algebra L(0). First, we mention some routine consequences of q = 0. The reordering
formulae are now

AB n = B n−1 , (28)
An B = An−1 . (29)

For nonzero integers m, n the above can be generalized into

110 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

An−m , n ≥ m,
An B m = (30)
B m−n , n < m.

From Lemma 3.4, for positive integers m, n, the images of the basis vectors B m An of
H(0) under the linear maps ad JBAK, ad B, and ad B m are given by the following.

[JBAK , B m An ] = 0, (31)
q y
[JBAK , B ] = −B
m m+1
=− B
A+B m
A , m+1
(32)
q y
[JBAK , An ] = BAn+1 − An = BAn+1 , (33)
[B, B m An ] = B m+1 An − B m An−1 , (34)
[B, A ] = BA − A
n n n−1
= JBA K ,
n
(35)
[B m , A] = B m A − B m−1 = JB m AK . (36)

Proposition 4.1. For any m, n ∈ Z+ ,

h  
h
JB m An K = (−1)i B m−i An−i , (37)
i=0
i

where h is the minimum of {m, n}.

Proof. For the case n ≥ m, use (26), (34), and induction on m. For the case n < m,
we shall use induction on m − n. First, use the previous case to obtain the formula for
JB n An K. Apply ad B on both sides of the equation. This leads to the proof for m −n = 1.
The rest follows by induction on m − n. 2

Corollary 4.2. The following relations hold in H(0) for any m, n ∈ Z+ .

⎧ q m−n+1 y
⎨− B
⎪ A , n < m,
[JBAK , JB A K] = 0,
m n
n = m, (38)
⎪
⎩q y
BAn−m+1 , n > m.

Proof. In view of (37), the only basis vector of the form B x Ay that appears in the linear
combination for JB m An K with xy = 0 is for the index i = h = min{m, n}, which is either
B m−n , I, or An−m . Then in view of (31), if we apply ad JBAK on both sides of (37), all
terms of the form [JBAK , B x Ay ] with xy = 0 will vanish. We are left with [JBAK , B m−n ]
if n < m, with 0 if n = m, or with [JBAK , An−m ] if n > m. To get the desired result,
use (32) and (33) on the first and third of the said cases, respectively. 2

Corollary 4.3. Let l ∈ N. The following vectors form a basis for the Z-grading subspace
Hl of H(0).
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 111

B l , JB m An K , m, n ∈ Z+ , m − n = l, (39)

while the following vectors form a basis for H−l ,

Al , JB m An K , m, n ∈ Z+ , m − n = −l. (40)

Proof. We show that the vectors in (39) form a basis for Hl . Since m − n = l, we rewrite
B m An as B n+l An . We claim that for any n, there exists a bijective linear transformation
of the span of

B l , B h+l Ah , h ≤ n, (41)

onto the span of

q y
Bl, B h+l Ah , h ≤ n. (42)

We prove this claim by induction on n. Consider n = 1. By (37),

q y
B 1+l A = B 1+l A − B l . (43)

Because of the above equation, we immediately find that there is a transition matrix,
q y
with determinant 1, of the span of B l , B 1+l A onto the span of B l , B 1+l A . Thus, the
claim holds for n = 1. Suppose the claim holds for some n. Our goal is to show that
there is a bijective linear transformation of the span of

B l , B h+l Ah , h ≤ n + 1, (44)

onto the span of

q y
Bl, B h+l Ah . h ≤ n + 1. (45)

Observe that the only vector that is in (44) that is not in (41) is B 1+n+l An+1 , while
q y
the only vector in (45) that is not in (42) is B 1+n+l An+1 . Also, by the inductive
hypothesis there exists an upper triangular transition matrix T from the span of (41)
onto the span of (42) with determinant c = 0. These observations imply that we need
q y
only one equation expressing B 1+n+l An+1 as a linear combination of (44). We get this
from (37), and this may be used to construct an upper triangular transition matrix T 
also with determinant c = 0 from the span of (44) onto the span of (45). By induction,
this proves the claim, and since it holds for all n, we find that there is a bijective linear
transformation from the span of (the basis) {B m An | m − n = l} of Hl onto the subspace
of Hl spanned by (39). Therefore, (39) is a basis for Hl . The proof for H−l is similar. 2
112 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

Corollary 4.4. The following vectors form a basis for H(0).

I, An , B m , JB m An K , m, n ≥ 1. (46)

Proof. Use Corollary 4.3 and the fact that {Hn }n∈Z is a Z-grading of H(0). 2

Corollary 4.5. The following relations hold in H(0) for any m, n ∈ Z+ with m, n ≥ 4.
q y q y
B 2 An − An−2 = B 2 An + 2 BAn−1 , (47)
q y q y
B m A2 − B m−2 = B m A2 + 2 B m−1 A . (48)

Proof. Use (26), (27), and (37). 2

Corollary 4.6. The following vectors form a basis for H(0).

I, A, B, B 2 Aj , B i A2 , JB m An K , m, n ≥ 1, i, j ≥ 4. (49)

Proof. An immediate consequence of the equations (47) and (48) is that for m, n ≥ 2,
if An−2 is replaced by B 2 An and B m−2 by B m A2 , we get a bijective linear map of the
span of the basis {B i Aj | i, j ∈ N} of H(0) onto the span of (49), and so H(0) has a
basis consisting of the vectors in (49). 2

Theorem 4.7. The Lie algebra L(0) has a basis consisting of the vectors

A, B, JB m An K , (m, n ∈ Z+ ). (50)

Proof. We first claim that the following Lie monomials are elements of the span of (50)
for all m, n, i, j ∈ Z+ .

[B, JB m An K] (51)
[JB A K , A]
m n
(52)
 m n q i j y
JB A K , B A . (53)
q y
The Lie monomial (51) is equal to B m+1 An ∈ L(0) by the definition of a nonassociative
regular word. We show (52) is an element of L(0) by induction on m. The case m = 1
follows immediately from the definition of JBAn K. Suppose (52) is in L(0) for some m.
Then by the properties of the Lie bracket,
q y 
B m+1 An , A = [[B, JB m An K] , A] ,
= [B, [JB m An K , A]] − [JB m An K , [B, A]] ,
= [B, [JB m An K , A]] + [JBAK , JB m An K] . (54)
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 113

By Corollary 4.2 the second term in (54) is in L(0). As for the remaining term, the
inductive hypothesis applies to [JB m An K , A], and so [B, [JB m An K , A]] is in L(0). By
induction, we find that (52) is in L(0). We now prove (53) is in L(0). Let f = JB m An K.
First, we consider the case i = 1, and use induction on j. For j = 1, we simply use (31)
and (37). Suppose (53) is in L(0) for i = 1 and some j. Then
 q y  q y 
f, BAj+1 = f, BAj , A
  q y q y 
= − A, f, BAj − BAj , [A, f ] ,
  q y  q y
= − A, f, BAj − [f, A] , BAj ,
  q y  m n q y
= − A, f, BAj − [JB A K , A] , BAj . (55)

We use (52) on the expression [JB m An K , A], which is in the second term in (55). This
gives us
 q y   q y q m n+1 y q y
f, BAj+1 = − A, f, BAj − B A , BAj , (56)
 q y
wherein the inductive hypothesis applies to both of the Lie monomials f, BAj and
q m n+1 y q y
B A , BAj . Then the right side of (56) is in L(0). By induction, (53) is in
L(0) for i = 1 and all j. We now proceed with induction on i. By the previous part of
the proof, we are done with i = 1. Suppose that for some i, (53) is in L(0) for all j. We
further have
 q i+1 j y   q i j y
f, B A = f, B, B A ,
q i j y   q y 
= − B A , [f, B] − B, B i Aj , f ,
 q y   q y
= − [B, f ] , B i Aj + B, f, B i Aj ,
 q i j y   q i j y
= − [B, JB A K] , B A
m n
+ B, f, B A ,
q m+1 n y q i j y   q i j y
=− B A , BA + B, f, B A , (57)
q y q y  q y
where the inductive hypothesis applies to B m+1 An , B i Aj and also to f, B i Aj .
Then the right side of (57) is in L(0). By induction, (53) is in L(0) for any i, j, and
our claim is proven. We now prove that the vectors in (50) form a basis for L. Linear
independence follows from Corollary 4.4. Also, observe that by construction, all the
vectors in (50) are in L(0), and so to prove spanning, we show that for any regular
word W on B > A, the commutator JW K is in the span of the vectors in (50). This is
because all the nonassociative regular words on A < B form a spanning set for L(0).
Given a regular word W , we use induction on κ(W ). The case κ(W ) = 1 is trivial. If
κ(W ) ≥ 2, then W = U V for some regular words U, V such that JW K = [JU K , JV K] and
that κ(U ), κ(V ) < κ(W ). By the inductive hypothesis, both JU K and JV K are in the
span of (50). Then in taking the Lie bracket [JU K , JV K], by the claim, the result is still a
linear combination of the vectors in (50). Then JW K is in the span of (50). By induction,
we get the desired result. 2
114 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

Lemma 4.8. The following relations hold in H(0) for any i, j, k, l ≥ 4.

  q y q y q y
B, B 2 Ak = B 3 Ak + 2 B 2 Ak−1 − 3 BAk−2 , (58)
 2 k  q 2 k+1 y q y
B A ,A = B A + BAk , (59)
 i 2  q i 3y q y q y
B A , A = B A + 2 B i−1 A2 − 3 B i−2 A , (60)
  q y q y
B, B l A2 = B l+1 A2 + B l A , (61)
 i 2 l 2  
B A , B A = 0 = B 2 Aj , B 2 Ak . (62)
 
Proof. We first prove (58). By (30), we have B, B 2 Ak = B 3 Ak − B 2 Ak−1 . Use (26)
and (37) to eliminate B 3 Ak and B 2 Ak−1 , and also BAk−2 , Ak−3 in succeeding steps so
that only Lie monomials remain. The result is (58). The proofs for (59), (60) (61) are
similar, while (62) is a simple consequence of (30). 2

Corollary 4.9. The following are elements of L(0) for any i, j, k, l ≥ 4.

q y 
B i Aj , B l A2 , (63)
q i j y 2 k 
B A ,B A . (64)

Proof. We first prove (63) by induction on i, j. If i = 1 = j, then by (31), the result

is 0 ∈ L(0). Assume that for i = 1, the relation (63) holds for some j. By the Jacobi
identity,
q y  q y     
BAj+1 , B l A2 = BAj , B l A2 , A − JBAn K , B l A2 , A , (65)
q y 
in which the inductive hypothesis applies to BAj , B l A2 in the first term, while in
 l 2  q y 
the second term, B A , A ∈ L(0) by (60). Then BAj+1 , B l A2 ∈ L(0), and (63)
holds for all j if i = 1. We further assume that for some i, (63) holds for any j. We have
q y   q y  q y  
B i+1 Aj , B l A2 = B, B i Aj , B l A2 − B i Aj , B, B l A2 . (66)
q y 
Using the inductive hypothesis and (61), we find that B i+1 Aj , B l A2 ∈ L(0). By
induction, we get the desired result. The proof for (64) is similar. 2

Theorem 4.10. The Lie subalgebra L(0) of H(0) is a Lie ideal of H(0), and the quotient
Lie algebra H(0)/L(0) is infinite-dimensional and one-step nilpotent.

Proof. In view of Corollary 4.6 to Corollary 4.9, L(0) is a Lie ideal of H(0) and the
quotient Lie algebra H(0)/L(0) is spanned by I, B i A2 , B 2 Aj (i, j ≥ 4). The said vectors
are linearly independent in H(0)/L(0). Otherwise, some linear combination of these
vectors in H(0) is an element of L(0), contradicting Corollary 4.6. Then I, B i A2 , B 2 Aj
(i, j ≥ 4) form a basis for H(0)/L(0), and hence, H(0)/L(0) is infinite-dimensional. We
show H(0)/L(0) is one-step nilpotent. Define
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 115

f1 := BA − I = JBAK , (67)
fr := B r Ar − I = B r−1 JBAK Ar−1 + fr−1 , r ≥ 2. (68)

Setting q = 0 in (27) gives us B r−1 JBAK Ar−1 = JB r AK Ar−1 . But because

 r 
JB AK , Ar−1 = JB r AK Ar−1 − Ar−1 JB r AK ,

we further have

 
B r−1 JBAK Ar−1 = JB r AK , Ar−1 + Ar−1 JB r AK ,
 
= JB r AK , Ar−1 + Ar−1 B r A − Ar−1 B r−1 ,
 
= JB r AK , Ar−1 + BA − I,
 
= JB r AK , Ar−1 + JBAK . (69)
 
Since L(0) is a Lie ideal, JB r AK , Ar−1 ∈ L(0), and so by (69), we deduce that

B r−1 JBAK Ar−1 ∈ L(0).

Then fr ∈ L(0) for any r. By a similar argument and similar computations, this can be
generalized into B x fr Ay ∈ L(0) for any x, r, y ∈ Z+ . Thus, given the relations,

  B i−k+2 (B k−2 Ak−2 − I)A2 , i ≥ k,

B i A2 , B 2 Ak = (70)
B (B2 i−2
A i−2
− I)A k−i+2
, i < k,
 
which are consequences of (30), we find that B i A2 , B 2 Ak ∈ L(0) for any i, k ≥ 4.
Then the images of B i A2 , B 2 Ak under the canonical map H(0) → H(0)/L(0) pairwise
commute. Therefore, H(0)/L(0) is one-step nilpotent. 2

5. The Lie algebra L(q) with q nonzero and not a root of unity

We assume throughout the section that q is nonzero, and is not a root of unity. To
remind us of this restriction on q, we denote H(q) and L(q) by H(q) and L(q), respectively.
For any k ∈ N and any l ∈ Z+ , we define the following commutators in L(q).

q y  k l+1

α(k, l) := (BA)k BAl+1 = (ad JBAK) ◦ (−ad A) (B) , (71)
q y   q 2 y
l−1 k
β(k, l) := B l+1 A(BA)k = (ad B) ◦ (−ad JBAK) B A , (72)
q y  k
 q y
γ(k) := B(BA)k BA2 = (ad B) ◦ (ad JBAK) BA2 . (73)
116 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

Lemma 5.1. For any k ∈ N and any l ∈ Z+ , the following relations hold in H(q).

k+1
α(k, l) = −(1 − q)l (q l − 1)k [A, B] Al , (74)
k+1
β(k, l) = (q − 1) (1 − q
k+1 k+1 l−1
) B [A, B] , l
(75)
 
γ(k) = q −k (q − 1)k+1 q{k}q [A, B]
k+1 k+2
− {k + 1}q [A, B] . (76)

Furthermore, the commutators γ(k) satisfy the following relation.


k
(q − 1)−(i+1) γ(i) = −{k + 1}q [A, B]
k+2
qk . (77)
i=0

Proof. Use induction on k for (76), (77), and induction on k, l for (74), (75). 2

Our next goal is to show that for f, g arbitrarily chosen among

JBAK , A, B, α(k, l), β(k, l), γ(k), (k ∈ N, l ∈ Z+ ), (78)

the commutator [f, g] is a linear combination of (78). From (74), (75), and (76), we have
some complications in view of the scalar coefficients. For simpler computations, we define
the following.

−(1 − q)−l (q l − 1)−k α(k, l)

k+1
α(k, l) := = [A, B] Al , (79)
−k−1 k+1
β(k, l) := (q − 1) (1 − q k+1 1−l
) l
β(k, l) = B [A, B] , (80)

qk k
(q − 1)−i γ(i)
k+2
γ(k) := = [A, B] . (81)
1 − q k+1 i=0

Observe that α(k, l) and β(k, l) are simply scalar multiples of α(k, l) and β(k, l), respec-
k+2
tively. Also, by some simple computation, [A, B] can be shown to be equal to the
expression to the right of γ(k) in (81) using (77). With these simpler expressions, we
only need (14), (15), (16) to compute for commutation relations.

Proposition 5.2. For any k, m ∈ N and any l, n ∈ Z+ , the following relations hold in
H(q).

[α(k, l), α(m, n)] = (q l(m+1) − q n(k+1) )α(k + m + 1, l + n), (82)

 
β(k, l), β(m, n) = (q n(k+1) − q l(m+1) )β(k + m + 1, l + n). (83)

Proof. Observe that

 
k+1 l m+1 n
[α(k, l), α(m, n)] = [A, B] A , [A, B] A ,
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 117

 
k+1 m+1
= [A, B] Al [A, B] An
 
m+1 k+1
− [A, B] An [A, B] Al . (84)

m+1 k+1
We simply use (14) to rewrite Al [A, B] and An [A, B] as scalar multiples of
m+1 l k+1 n
[A, B] A and [A, B] A , respectively. The result is (82). The relation (83) is sim-
ilarly verified. 2

Proposition 5.3. For any k ∈ N and any l ∈ Z+ ,

[α(k, l), B] = (1 − q −(k+2) ){l}q α(k + 2, l − 1), (l ≥ 2), (85)

[α(k, 1), B] = q −(k+1) ({k + 2}q γ(k) − {k + 1}q γ(k − 1)) , (k ≥ 1), (86)
−1
[α(0, 1), B] = q ((1 + q)γ(0) + JBAK) . (87)

Proof. We first prove (85). Observe that

 
k+2 l
[α(k, l), B] = [A, B] A ,B ,
k+2 k+2
= [A, B] Al B − B [A, B] Al . (88)

k+2
In the second term in (88), we use (14) to rewrite B [A, B] as a scalar multiple of
k+2
[A, B] B. This results to

[α(k, l), B] = (1 − q −(k+2) ) [A, B]

k+2
(Al−1 (AB) − (BA)Al−1 ),

in which we substitute for AB, BA using the following relations that are immediate from
AB − qBA = I.

AB = (q − 1)−1 (q [A, B] − I), (89)

BA = (q − 1)−1 ([A, B] − I). (90)

The result is

1 − q −(k+2) k+2
[α(k, l), B] = [A, B] (qAl−1 [A, B] − [A, B] Al−1 ),
q−1

on which we use (14) in order to rewrite Al−1 [A, B] as a scalar multiple of [A, B] Al−1 .
This yields (85). A similar pattern of computations may be made to verify (86) and (87).
We note in the following the important steps in the said computations.
 
k+1
[α(k, 1), B] = [A, B] A, B ,
k+1 k+1
= [A, B] AB − B [A, B] A, (91)
118 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

(AB − q −(k+1) BA),

k+1
= [A, B] (92)
= q −(k+1) {k}q [A, B]
k+2 k+1
− {k + 1}q [A, B] , (93)
−(k+1) k+1
=q {k}q γ(k) − {k + 1}q [A, B] . (94)

k+1
The transition from (91) to (92) involves the use of (14) to rewrite B [A, B] as a
k+1
scalar multiple of [A, B] B. Going from (92) to (93) involves the use of (89) and (90).
Finally, in (94), the Lie monomial in the second term is either γ(k − 1) if k ≥ 1, or
− JBAK if k = 0. From these we get (86) and (87). 2

Proposition 5.4. For any m ∈ N and any n ∈ Z+ ,

 
A, β(m, n) = q −(m+2) {m + n + 2}q β(m + 2, n − 1)
− q −(m+2) {m + 2}q β(m + 1, n − 1), (n ≥ 2), (95)
 
A, β(m, 1) = q −(m+1) {m + 2}q γ(m)
− q −(m+1) {m + 1}q γ(m − 1), (m ≥ 1), (96)
 
A, β(0, 1) = q −1 ((1 + q)γ(0) + JBAK) . (97)

Proof. The computations needed to verify relations (95) to (97) are similar to those
outlined in the proof of Proposition 5.3. 2

Proposition 5.5. Let k, m ∈ N and l, n ∈ Z+ . For any i ∈ N, define

   
n i+1 n−i
− q −n(k+m+
n+3
ci (k, l, m, n) := (−1)n−i q (l−n)(i+m+1)+ 2 2 )+ 2 .
i q

Then the following relations hold in H(q).

⎧ n
⎪
⎪
⎪
⎪
(q − 1)−n i=0 ci (k, l, m, n)α(i + k + m + 1, l − n), l > n,
⎪
⎪
  ⎨ l
α(k,l),β(m,n) = (q
⎪
− 1)−l i=0 ci (m, n, k, l)β(i + k + m + 1, n − l), l < n, (98)
⎪
⎪
⎪
⎪
⎪ l
⎩ (q − 1)−l i=0 ci (k, l, m, l)γ(i + k + m), l = n.

Proof. First, we consider the case l > n. Observe that

   k+1 l m+1

α(k, l), β(m, n) = [A, B] A , B n [A, B] , (99)
k+1 m+1
= [A, B] Al B n [A, B]
k+m+2
−B n [A, B] Al , (100)
(l−n)(m+1) k+m+2 l n
=q [A, B] AB
−q −n(k+m+2) [A, B]
k+m+2
B n Al , (101)
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 119

k+m+2
= q (l−n)(m+1) [A, B] Al−n (An B n )
−q −n(k+m+2) [A, B]
k+m+2
(B n An )Al−n . (102)

The step from (100) to (101) is because of the use of (14) in order to rewrite
m+1 k+m+2 m+1 l n
Al B n [A, B] and B n [A, B] as scalar multiples of [A, B] A B and
[A, B]k+m+2 B n , respectively. We note here that it is possible to go from (101) to (102)
because of the assumption l > n. We use (16) to replace An B n by a linear combina-
tion of powers of [A, B]. The result is that the Al−n (An B n ) in (102) becomes a linear
y
combination of vectors of the form Ax [A, B] , which can be replaced as scalar multiples
y x
of [A, B] A using (14). If we further use (15) to replace B n An in (102) as a linear
combination of powers of [A, B], we find that (102) is a linear combination of vectors of
u
the form [A, B] Av . It is routine to show that the exact scalar coefficients and exponents
u v
in each [A, B] A is as given in the statement of this proposition. To prove the other
cases, we need equations similar to (102), on which we can use (14), (15), (16) in order
to obtain the desired linear combinations in the second and third cases in (98). The said
equations are the following.
  k+m+2
α(k, l), β(m, n) = q (n−l)(k+1) (Al B l )B n−l [A, B]
− q −l(k+m+2) B n−l (B l Al ) [A, B]
k+m+2
, (l < n) (103)
  k+m+2 l l
α(k, l), β(m, n) = [A, B] AB
− q −l(k+m+2) [A, B]
k+m+2
B l Al , (l = n). (104)

The above equations can be derived from (100) using (14). As previously described, we
replace Al B l and B l Al in (103), (104) using (15), (16). In (104), we get the desired
relation after some routine computations involving the scalar coefficients. For (103), we
x
need to use (14) to rewrite any vector of the form [A, B] B y into a scalar multiple
x
of B y [A, B] , after which it is routine to show that the exact scalar coefficients and
x
exponents in all B y [A, B] are as given in the statement. 2

In Propositions 5.2 to 5.5, we have tackled the most complicated commutation rela-
tions. However, we still have several more cases for the commutators [f, g] where f, g are
among

α(k, l), A, JBAK , γ(k), B, β(k, l).

These other cases not covered in Propositions 5.2 to 5.5 can be obtained by simple
routine computations mostly involving (14). We summarize the results in the following
table.
For any f, g among α(k, l), A, JBAK , γ(k), B, β(k, l), all the possibilities for f are
listed in the first column, while all possibilities for g are listed in the first row. The
commutator [f, g] can be found on the corresponding cell. Because of the Lie algebra
 120
R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123
Table 1
Commutation relations for the vectors α(k, l), A, JBAK , γ(k), B, β(k, l).

α(m, n) A JBAK γ(m) B β(m, n)

α(k, l) (82) (1 − q)α(k, l + 1) (1 − q )α(k + 1, l)

l
(q l(m+2)
− 1)α(k + m + 2, l) (85), (86), (87) (98)
A 0 (1 − q)α(0, 1) (q m+2 − 1)α(m + 1, 1) − JBAK (95)
JBAK 0 0 (1 − q)β(0, 1) (1 − q n )β(m + 1, n)
γ(k) 0 (q k+2 − 1)β(k + 1, 1) (q n(k+2) − 1)β(k + m + 2, n)
B 0 (1 − q m+1 )β(m, n + 1)
β(k, l) (83)
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 121

property [v, u] = − [u, v] for any u, v ∈ L(q), some cells in the table are left blank. Our
most important observation here is the following.

Remark 5.6. With reference to Table 1, the span of

α(k, l), A, JBAK , γ(k), B, β(k, l)

is closed under the Lie bracket.

Lemma 5.7. The following vectors form a basis for H(q).

JBAK , Ak , B l , α(k, l), β(k, l), γ(k), (k ∈ N, l ∈ Z+ ). (105)

Proof. This immediately follows from the fact that JBAK = − [A, B], the equations for
α(k, l), β(k, l), γ(k) in (79) to (81), and Proposition 3.1. 2

Theorem 5.8. The following vectors form a basis for L(q).

JBAK , A, B, α(k, l), β(k, l), γ(k), (k ∈ N, l ∈ Z+ ). (106)

The commutation relations for the above basis are given in Table 1.

Proof. The proof that (106) is a basis for L(q) is similar to that in the proof of Theo-
rem 4.7. Linear independence follows from Lemma 5.7, and also observe that all vectors in
(106) are in L(q). Thus, to show spanning, we prove that all nonassociative regular words
on A < B, which form a spanning set for L(q), belong to the span of (106). To do this,
given a regular word W , we use induction on κ(W ). Observe that JBAn K is either JBAK
or α(0, n − 1), the latter being a scalar multiple of α(0, n − 1). Thus, in any case, JBAn K
is in the span of (106). By Remark 5.6 and induction on m, it is routine to show JB m An K
is in the span of (106) for all m. This serves as proof for κ(W ) = 1. Using Remark 5.6
again, the induction step is similar to that done in the proof of Theorem 4.7. 2

Lemma 5.9. Let k ∈ N and l, m, n ∈ Z+ with m, n ≥ 2. Then the following hold.

 
An , β(k, l) ∈ L(q), (107)
[B , α(k, l)] ∈ L(q).
m
(108)

Proof. Define f := An B l − q −n(k+1) B l An ∈ Hl−n . By some routine calculations,

 n  k+1
A , β(k, l) = f [A, B] . If l = n, then by Proposition 3.1, f is a linear combina-
k+1
tion of I, JBAK , γ(t) for 0 ≤ t ≤ s for some s. Since k + 1 ≥ 1, we find that f [A, B]
k+1
is a linear combination of the JBAK , γ(t), and so f [A, B] ∈ L(q). If l > n or l < n,
k+1
by similar reasoning, we find that f [A, B] is a linear combination of vectors of the
k+1
form α(x, y) or β(x, y), and so f [A, B] ∈ L(q). The proof for (108) is similar. 2
122 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123

Table 2
Lie products for L(q) as a Lie ideal of H(q).

JBAK α(k, l) β(k, l) γ(k)

An (1 − q n )α(0, n) (q n(k+1) − 1)α(k, l + m) (107) (q n(k+2) − 1)α(k + 1, n)

B m
(q m
− 1)β(0, m) (108) (1 − q m(k+1)
)β(k, l + m) (1 − q m(k+2) )β(k + 1, m)

Theorem 5.10. The Lie subalgebra L(q) of H(q) is a Lie ideal of H(q), and the resulting
quotient Lie algebra H(q)/L(q) is infinite-dimensional and one-step nilpotent.

Proof. By Lemma 5.7 and Theorem 5.8, it suffices to show that L(q) contains all Lie
products of the form [L, R] where L is one of the vectors in (105) not in (106), and R
is one of the vectors in (106). To this end, we summarize relevant information about all
such Lie products [L, R] in Table 2. Observe that L ∈ {An , B m | m, n ≥ 2} and these
are listed in the first column. The possibilities for R are listed in the first row.
Therefore, L(q) is a Lie ideal of H(q). By Lemma 5.7 and Theorem 5.8, a basis for
H(q)/L(q) consists of the images of An , B m , m, n ≥ 0 under the canonical Lie algebra
homomorphism H(q) → H(q)/L(q). Thus, H(q)/L(q) is infinite-dimensional. To show
H(q)/L(q) is one-step nilpotent, it suffices to show that in H(q), the Lie product [B m , An ]
(for any m, n ≥ 2) is an element of L(q). By routine computations it can be shown that
[B m , An ] is an element of the span of

{JBAK , β(i − 1, m − n) | i ∈ N},

{α(i − 1, n − m) | i ∈ N},
{JBAK , γ(k) | k ∈ N},

for the cases m > n, m < n, and m = n, respectively. This completes the proof. 2

6. Further directions

We invite the interested reader to explore the case when q is a root of unity. Also,
we note that investigation on commutators in associative algebras with q-deformed com-
mutation relations is a relatively new territory of exploration. The idea of exploring
the interplay between a presentation with q-deformed commutation relations and the
non-deformed commutators in the same algebraic structure was proposed in [11, Prob-
lem 12.14], but this was for an algebra that is not H(q), and has more complicated
q-deformed defining relations. In [2], it was shown that there exists an associative al-
gebra whose defining relations are q-deformed commutation relations that are not Lie
polynomials, but the underlying Lie subalgebra is not free. Based on the state of cur-
rent literature, if any, about our topic, we consider as open the question of whether a
Lie algebra presentation can be deduced from an associative algebra presentation whose
relations are q-deformed commutation relations that do not consist of Lie polynomials.
 R.R.S. Cantuba / Journal of Algebra 522 (2019) 101–123 123

Acknowledgments

This work was supported by a grant from De La Salle University, Manila, [grant
number DLSU-PHD-04.2017-2018.T1.COS] through the University Research and Coor-
dination Office (DLSU-URCO). About a year after the completion and initial submission
of this work, the author had the chance to meet with the authors of the main references
about q-deformed Heisenberg algebras [5,6] through a grant from the Commission for
Developing Countries of the International Mathematical Union (IMU-CDC), and further
support from Mälardalen University, Västerås, Sweden, and also from De La Salle Uni-
versity, Manila. It was an opportunity for introducing this line of research, which is about
characterizing Lie polynomials in algebras that are not free, and some related problems,
to which this study about the Lie polynomials in q-deformed Heisenberg algebras is of
central significance, and so the author would like to thank all those who expressed their
interest and appreciation.

References

[1] G. Bergman, The diamond lemma for ring theory, Adv. Math. 29 (1978) 178–218.
[2] R. Cantuba, A Lie algebra related to the universal Askey–Wilson algebra, Matimyás Matematika
38 (2015) 51–75, available at http://mathsociety.ph/matimyas/images/vol38/Cantuba2.pdf.
[3] R. Carter, Lie Algebras of Finite and Affine Type, Cambridge Studies in Advanced Mathematics,
vol. 96, Cambridge University Press, Cambridge, 2005.
[4] M. Hall, A basis for free Lie rings and higher commutators in free groups, Proc. Amer. Math. Soc.
1 (1950) 575–581.
[5] L. Hellström, S. Silvestrov, Commuting Elements in q-Deformed Heisenberg Algebras, World Sci-
entific, 2000.
[6] L. Hellström, S. Silvestrov, Two-sided ideals in q-deformed Heisenberg algebras, Expo. Math. 23
(2005) 99–125.
[7] A. Hora, N. Obata, Quantum Probability and Spectral Analysis of Graphs, Springer-Verlag, 2007.
[8] C. Reutenauer, Free Lie Algebras, Oxford Univ. Press, New York, 1993.
[9] A. Shirshov, Subalgebras of free Lie algebras, in: L. Bokut, et al. (Eds.), Selected Works of A.I.
Shirshov, Birkhäuser, Switzerland, 2009, pp. 3–13.
[10] A. Shirshov, On free Lie rings, in: L. Bokut, et al. (Eds.), Selected Works of A.I. Shirshov,
Birkhäuser, Switzerland, 2009, pp. 77–87.
[11] P. Terwilliger, The universal Askey–Wilson algebra, SIGMA 7 (2011) 069, arXiv:1104.2813.
[12] V. Ufnarovskij, Algebra VI: combinatorial and asymptotic methods of algebra, in: A. Kostrikin, I.
Shafarevich (Eds.), Non-Associative Structures, in: Encyclopaedia of Mathematical Sciences, vol. 57,
Springer, Berlin Heidelberg, 1995.