The Universal Askey-Wilson Algebra and the Equitable Presentation of $U_q(\mathfrak{sl}_2)$

Around 1992 A. Zhedanov introduced the Askey-Wilson algebra AW(3). Recently we introduced a central extension $\Delta$ of AW(3) called the universal Askey-Wilson algebra. In this paper we discuss a connection between $\Delta$ and the quantum algebra $U_q(\mathfrak{sl}_2)$. Our main result is an algebra injection from $\Delta$ into a relative of $U_q(\mathfrak{sl}_2)$; the relative is obtained from $U_q(\mathfrak{sl}_2)$ by adjoining three mutually commuting indeterminates. We describe the injection using the equitable presentation of $U_q(\mathfrak{sl}_2)$


Introduction
The Askey-Wilson polynomials were introduced in [4] and soon became renown for their algebraic, analytic, and combinatorial properties [7,11].In his study [20] of the "hidden symmetry" of these polynomials, A. Zhedanov introduced the Askey-Wilson algebra AW (3).This algebra is defined by generators and relations.The relations involve a nonzero parameter q and 5 additional parameters.The algebra is infinite dimensional and noncommutative.Zhedanov's original presentation involves three generators; however one generator is a q-commutator of the other two and is sometimes eliminated.The remaining two generators satisfy a pair of relations often called the Askey-Wilson relations [18].These relations describe the Askey-Wilson polynomials in the following way.Let {p n } ∞ n=0 denote a sequence of Askey-Wilson polynomials in a variable λ.These polynomials are eigenvectors for a certain q-difference operator, known as the Askey-Wilson operator and denoted for the moment by A. Let B denote the linear operator that sends p(λ) → λp(λ) for all polynomials p(λ).The operator B acts on the basis {p n } ∞ n=0 in an irreducible tridiagonal fashion, reflecting the fact that {p n } ∞ n=0 satisfy a three-term recurrence.In [20] Zhedanov showed that A, B satisfy a pair of Askey-Wilson relations.
Although the Askey-Wilson relations are slightly complicated, over time their significance became clear as they found applications to integrable systems [5,13,19], quantum groups [6], linear algebra [15,18], quantum mechanics [14], and the double affine Hecke algebra [12,8].We now describe the two applications most relevant to the present paper.
Our first application concerns a linear algebraic object called a Leonard pair.This is a pair of diagonalizable linear transformations on a finite-dimensional vector space, each of which acts in an irreducible tridiagonal fashion on an eigenbasis for the other one [15,Definition 1.1].In [18] Vidunas and the present author showed that a Leonard pair satisfies a pair of Askey-Wilson relations.This example is closely related to the one involving the Askey-Wilson polynomials.By [16,Section 5] the Leonard pairs correspond to a family of orthogonal polynomials consisting of the q-Racah polynomials and their relatives.The polynomials in this family are special or limiting cases of the Askey-Wilson polynomials [11].
Our second application is about quantum groups.Consider the algebra U q (sl 2 ) with the usual Chevalley generators e, f , k ±1 [10].Let A denote an arbitrary linear combination of ek −1 , f , k −1 and let B denote an arbitrary linear combination of e, f k, k.Then according to Granovskiȋ and Zhedanov [6] the elements A, B satisfy a pair of Askey-Wilson relations.In [19] Wiegmann and Zabrodin extended this result by displaying an element C in U q (sl 2 ) such that where g A , g B , g C and h A , h B , h C are scalars in the underlying field.The equations (1.1)- (1.3) are often called the Z 3 -symmetric Askey-Wilson relations [8].Upon eliminating C in (1.2), (1.3) using (1.1) we obtain the Askey-Wilson relations in the variables A, B. Upon substituting 3) we recover the original presentation for AW (3) in the variables A, B, C ′ .We now recall the universal Askey-Wilson algebra ∆ [17].To motivate this algebra consider the relations (1.1)- (1.3).They are attractive but one might object that there are too many parameters.To accomodate this objection we will eliminate all the parameters besides q.We will do this without significantly reducing the generality of the algebra (although we allow a minor technical assumption).We first eliminate g A , g B , g C with the following change of variables.Assume that each of g A g B , g B g C , g C g A is a nonzero square in the underlying field, and that q 4 = 1.Now in (1.1)-(1.3)replace A, B, C by respectively, where The resulting equations assert that each of is a scalar in the underlying field.We have eliminated g A , g B , g C and are now down to the three scalar parameters (1.4).To eliminate these we reinterpret them as central elements in the algebra generated by A, B, C. The resulting algebra is denoted ∆ and called the universal Askey-Wilson algebra [17].The formal definition of ∆ is given in Definition 2.1 below.
In [17] we investigated ∆ from a ring theoretic point of view.Our results include the following.We displayed a faithful action of the modular group PSL 2 (Z) on ∆ as a group of automorphisms [17,Theorems 3.1,3.13].We found several linear bases for ∆ [17,Theorems 4.1,7.5].We described the center Z(∆) under the assumption that q is not a root of unity.For such q we found that Z(∆) is generated by the three central elements (1.4) together with an element Ω called the Casimir element [17,Corollary 8.3].
We now discuss the equitable presentation for U q (sl 2 ) [9].This presentation involves generators x, y ±1 , z and relations yy −1 = y −1 y = 1, In [1] H. AlNajjar investigated Leonard pairs using the equitable presentation of U q (sl 2 ).His approach is summarized as follows.Let V denote a finite-dimensional irreducible U q (sl 2 )-module.
Let A denote an arbitrary linear combination of 1, x, y, xy and let B denote an arbitrary linear combination of 1, y, z, yz.Consider the coefficients.Alnajjar found necessary and sufficient conditions on the coefficients for A, B to act on V as a Leonard pair [1,Theorem 6.2].In [2] Alnajjar described the class of Leonard pairs that result from his construction.He showed that this class corresponds to a family of orthogonal polynomials consisting of the q-Racah, q-Hahn, dual q-Hahn, q-Krawtchouk, dual q-Krawtchouk, affine q-Krawtchouk, and quantum q-Krawtchouk polynomials.For the Leonard pairs A, B in the above class consider the corresponding Askey-Wilson relations.We use the Z 3 -symmetric version in view of the Z 3 -symmetric nature of the equitable presentation.In the style of Wiegmann and Zabrodin let C denote an arbitrary linear combination of 1, z, x, zx and consider when A, B, C satisfy some Z 3 -symmetric Askey-Wilson relations.Extending the work of Alnajjar one finds that the "most general" solution is described as follows.
Proposition 1.1.Let F denote a field, and fix a nonzero q ∈ F such that q 4 = 1.Consider the F-algebra U q (sl 2 ) with equitable generators x, y ±1 , z.Let a, b, c denote nonzero scalars in F and define Here Λ denotes the normalized Casimir element of U q (sl 2 ) from Lemma 2.15 below.
Let q, a, b, c be from Proposition 1.1.By that proposition and since Λ is central in U q (sl 2 ), there exists an algebra homomorphism ∆ → U q (sl 2 ) that acts on the ∆-generators A, B, C in the following way.It sends It turns out that this homomorphism is not injective.In order to shrink the kernel we reinterpret a, b, c as mutually commuting indeterminates, and view the above construction as giving an algebra homomorphism ∆ → U q (sl 2 ) A main result of the present paper is that this homomorphism is injective.In another main result we compute the image of Ω under the injection.We also use the injection to show that ∆ contains no zero divisors.The paper is organized as follows.In Section 2 we recall some basic facts and then state our main results, which are Theorems 2.16-2.18and Corollary 2.19.In Section 3 we establish some identities involving the equitable generators of U q (sl 2 ), which will be used repeatedly.In Section 4 we prove Theorem 2.16.In Section 5 we prove Theorem 2.17.In Sections 6-8 we establish some slightly technical facts about U q (sl 2 ), which will be used in Section 9 to prove Theorem 2.18.In Section 10 we discuss some issues concerning the PSL 2 (Z) action on ∆ that we mentioned earlier.
Our proofs for Theorems 2.16-2.18are essentially self contained and do not assume Proposition 1.1.We remark that Proposition 1.1 follows from Theorem 2.16.
For the rest of this paper a, b, c denote mutually commuting indeterminates.

Statement of results
Our conventions for the paper are as follows.An algebra is meant to be associative and have a 1.A subalgebra has the same 1 as the parent algebra.We fix a field F. All unadorned tensor products are meant to be over F. We fix q ∈ F such that q 4 = 1.Recall the natural numbers N = {0, 1, 2, . ..} and integers Z = {0, ±1, ±2, . ..}.
Definition 2.1 ([17, Definition 1.2]).Define an F-algebra ∆ = ∆ q by generators and relations in the following way.The generators are A, B, C. The relations assert that each of is central in ∆.The algebra ∆ is called the universal Askey-Wilson algebra.
By [17,Theorem 4.1] the following is a basis for the F-vector space ∆: There is a related basis [17,Theorem 7.5] that we will use in Section 9 below.This related basis involves a central element Ω known as the Casimir element [17,Lemma 6.1].This element is defined as follows.
Lemma 2.4 ([17, Theorem 6.2, Corollary 8.3]).The Casimir element Ω is contained in the center Z(∆).Moreover {Ω i α r β s γ t | i, r, s, t ∈ N} is a basis for the F-vector space Z(∆), provided that q is not a root of unity.
We will be discussing how ∆ is related to the quantum universal enveloping algebra U q (sl 2 ).For this algebra there are two presentations of interest to us; the Chevalley presentation [10, Section 1.1] and the equitable presentation [9].We now recall the Chevalley presentation.Definition 2.6 ([10, Section 1.1]).The F-algebra U = U q (sl 2 ) is defined by generators e, f , k ±1 and relations We call e, f , k ±1 the Chevalley generators for U .
We now briefly discuss some finite-dimensional U -modules.Strictly speaking we will not use this information; it is included in order to clarify the nature of the Casimir element for U described below.
Recall the notation [n] q = q n − q −n q − q −1 , n ∈ N.
Lemma 2.7 ([10, Section 2]).For all integers n ≥ 0 and ε ∈ {1, −1} there exists a U -module L(n, ε) with the following properties.L(n, ε) has a basis {v i } n i=0 such that The U -module L(n, ε) is irreducible provided that q is not a root of unity.
In Definition 2.3 we gave the Casimir element for ∆.We now recall the Casimir element for U .Definition 2.8 ([10, Section 2.7]).Define Φ ∈ U as follows: We call Φ the Casimir element of U .
Lemma 2.9 ([10, Lemma 2.7, Proposition 2.18]).The element Φ is contained in the center Z(U ).Moreover {Φ i } i∈N is a basis for the F-vector space Z(U ), provided that q is not a root of unity.
For notational convenience we now adjust the normalization for Φ.
Note that on L(n, ε), We call Λ the normalized Casimir element for U .
We now recall the equitable presentation for U [9].
Proposition 2.12 ([9, Theorem 2.1]).The algebra U is isomorphic to the F-algebra defined by generators x, y ±1 , z and relations ) qyz − q −1 zy q − q −1 = 1, (2.8) (2.9) An isomorphism with the presentation in Definition 2.6 is given by The inverse of this isomorphism is given by Definition 2.13 ([9, Definition 2.2]).By the equitable presentation of U we mean the presentation given in Proposition 2.12.We call x, y ±1 , z the equitable generators for U .

Note 2.14.
In what follows we identify the copy of U given in Definition 2.6 with the copy given in Proposition 2.12, via the isomorphism given in Proposition 2.12.
In the equitable presentation of U the normalized Casimir element Λ looks as follows.
Lemma 2.15.The normalized Casimir element Λ is equal to each of the following: ) ) Proof .For the data (2.10)-(2.12)let Λ − y , Λ − z , Λ − x denote the expressions in the first column and let Λ + y , Λ + z , Λ + x denote the expressions in the second column.Consider the expression for Λ given in (2.5).Writing this expression in terms of x, y, z using the isomorphism in Proposition 2.12 and Note 2.14, we obtain Λ The expression (2.13) is zero by (2.8) so Recall that a, b, c are mutually commuting indeterminates.Let F[a ±1 , b ±1 , c ±1 ] denote the F-algebra of Laurent polynomials in a, b, c that have all coefficients in F.
We now state our main results.
Theorem 2.16.There exists a unique F-algebra homomorphism where x, y, z denote the equitable generators for U .The homomorphism ♮ sends ) where Λ denotes the normalized Casimir element of U .
Theorem 2.17.Under the homomorphism ♮ from Theorem 2.16, the image of Ω is

18)
Here Λ denotes the normalized Casimir element of U .
We mention a corollary to Theorem 2.18.For an F-algebra A, an element u ∈ A is called a zero divisor whenever u = 0 and there exists 0 = v ∈ A such that uv = 0.By [10,Proposition 1.8] the algebra U contains no zero divisors.For an F-algebra A and indeterminate λ consider the contains no zero divisors.Applying this comment three times we see that the algebra U ⊗ F[a ±1 , b ±1 , c ±1 ] contains no zero divisors.By this and Theorem 2.18 we obtain the following result.
3 The elements ν x , ν y , ν z In this section we record a number of identities involving the equitable generators for U .These identities will be used in our proof of Theorems 2.16-2.18.
Lemma 3.4.The following relations hold in U : Proof .For each equation evaluate the left-hand side using Lemma 3.3.
Lemma 3.5.The following relations hold in U : ) ) By these comments xν y = q 2 ν y x.The remaining relations are similarly obtained.
Lemma 3.6.The following relations hold in U : Proof .To verify the equation on the left in (3.13), eliminate ν x using ν x = q(1 − yz), and eliminate Λ using the fact that Λ is equal to the expression on the left in (2.11).The remaining equations are similarly verified.
Lemma 3.7.The following relations hold in U : Proof .For each equation evaluate the left-hand side using Lemma 3.6.
Lemma 3.8.The normalized Casimir element Λ is equal to each of the following: ) ) Proof .Evaluate each of the displayed expressions using Lemma 3.6.
Lemma 3.9.The following relations hold in U : Proof .To get the equation on the left in (3.19), observe The remaining equations are similarly verified.
Lemma 3.10.The following relations hold in U : Proof .For each equation evaluate the left-hand side using Lemma 3.9.
4 The proof of Theorem 2.16 In this section we prove Theorem 2.16.
For notational convenience we define some elements in U ⊗ F[a ±1 , b ±1 , c ±1 ]: and ) ) Proof of Theorem 2.16.We first establish the existence of the homomorphism in the theorem statement.To do this it suffices to show that (4.9) We verify (4.9).Let P denote the left-hand side of (4.9) minus the right-hand side of (4.9).We show that P = 0. View P as a Laurent polynomial in a, b, c that has coefficients in U .We will show that in this polynomial each coefficient is zero.To this end, evaluate P using (4.1)-(4.3),(4.6) and then collect terms.We list below the terms for which the coefficient in P is potentially nonzero: For each of these terms the coefficient in P is listed in the table below, along with a reason why that coefficient is zero.term coefficient in P why the coefficient is 0 We have shown that for each term in (4.10) the coefficient in P is zero.Therefore P = 0 and the equation (4.9) holds.The equations (4.7), (4.8) are similarly verified.We have shown that the homomorphism exists.The homomorphism is unique since A, B, C generate ∆.The homomorphism satisfies (2.15)-(2.17)by (4.4)-(4.6)and (4.7)-(4.9).
5 The proof of Theorem 2.17 In this section we prove Theorem 2.17.Recall the Casimir element Ω from Definition 2.3.
Proof of Theorem 2.17.By Theorem 2.16 the image of Ω is where A ♮ , B ♮ , C ♮ , α ♮ , β ♮ , γ ♮ are from (4.1)-(4.6).We show that (5.1) is equal to (2.18).Define Q to be (5.1)minus (2.18).We show that Q = 0. To do this we proceed as in the proof of Theorem 2.16.View Q as a Laurent polynomial in a, b, c that has all coefficients in U .We will show that for this polynomial each coefficient is zero.To this end, evaluate Q using (4.1)-(4.6)and then collect terms.Below we list the terms for which the coefficient in Q is potentially nonzero: ) We show that for each term in (5.2)-(5.5) the coefficient in To see that (5.6) is zero, eliminate xy using the equation on the left in (3.4), and evaluate the result using the equation on the right in (3.20).The coefficient of a −2 in Q is qyν x x + q 2 y 2 − qΛy − q −1 ν x + 1. (5.7) To see that (5.7) is zero, first eliminate yν x using the equation on the right in (3.11).Evaluate the result using the equation on the right in (3.4) followed by the equation on the right in (3.21).
The coefficient of The expression (5.8) is zero by the equation on the right in (3.6).The coefficient of abc −1 in Q is To see that (5.9) is zero, eliminate yx using the equation on the right in (3.4), and eliminate yν y using the equation on the right in (3.14).Simplify the result using the equation on the left in (3.15).The coefficient of ab To see that (5.10) is zero, eliminate xz using the equation on the right in (3.6), and eliminate ν z z using the equation on the left in (3.15).Simplify the result using the equation on the right in (3.14).We have shown that for each term in (5.2) the coefficient in Q is zero.By Lemma 2.5 Ω is fixed by the automorphism of ∆ that sends (A, B, C) to (B, C, A).Combining this with the Z 3 -symmetric nature of (4.1)-(4.6),we see that for each term in (5.3), (5.4) the coefficient in Q is also zero.We now consider the terms in (5.5).The coefficient of abc in Q is qxyz − qx − q −1 y − qz + Λ. (5.11) The expression (5.11) is zero using the left side of (2.10).The coefficient of a (5.12) The expression (5.12) is zero using the left side of (2.11).The constant term in Q is q ν z z 2 + xν x x + y 2 ν y + ν z ν x ν y + q 2 (xy + yx) + q −2 (yz + zy) + q 2 (zx + xz) − q(ν z + Λy + Λx) − q −1 (ν x + Λz + Λy) − q(ν y + Λx + Λz) − q + q −1 2 + Λ 2 + 6.
We show that this constant term is equal to zero.Using Lemma 3.3 and Lemma 3.6 we find ) (5.15) Using the equation on the left in (3.19), followed by (5.13) and the equation on the left in (3.15), we find (5.16) ) zx + xz = 2 − q + q −1 ν y . ( Simplifying the constant term of Q using (5.13)-(5.19) we find that this constant term is equal to zero.We have shown that for each term in (5.2)-(5.5) the coefficient in Q is zero.Therefore Q = 0 and the result follows.

A Z-grading of U
Our next general goal is to prove Theorem 2.18.To prepare for this proof we obtain some results about U .In this section we discuss a certain Z-grading of U .In the next section we will use this will be used in our proof of Theorem 2.18.
Let A denote an F-algebra.By a Z-grading of A we mean a sequence {A n } n∈Z consisting of subspaces of A such that and A m A n ⊆ A m+n for all m, n ∈ Z.Let {A n } n∈Z denote a Z-grading of A. For n ∈ Z we call A n the n-homogeneous component of A. We refer to n as the degree of A n .An element of A is said to be homogeneous with degree n whenever it is contained in A n .Pick ξ ∈ A and write ξ = n∈Z ξ n with ξ n ∈ A n for n ∈ Z.We call the elements {ξ n } n∈Z the homogeneous components of ξ.
Lemma 6.1 ([10, Theorem 1.5]).The following is a basis for the F-vector space U : For n ∈ Z let U n denote the subspace of U spanned by those elements e r k s f t from (6.1) that satisfy r − t = n.By [10, Section 1.9] the sequence {U n } n∈Z is a Z-grading of U .With respect to this Z-grading the elements e, k, f are homogeneous with degrees 1, 0, −1 respectively.Moreover the normalized Casimir element Λ from (2.5) is homogeneous with degree 0.
By construction, for n ∈ Z the n-homogeneous component U n has a basis consisting of the elements e r k s f t from (6.1) that satisfy r − t = n.There is another basis for U n that is better suited to our purpose; this basis involves Λ and will be displayed shortly.Lemma 6.2.For an integer t ≥ 0, Proof .Assume t ≥ 1; otherwise the result is trivial.Using (2.5) and ke = q 2 ek, The result follows by induction on t.
(i) The F-vector space U n has a basis (ii) The F-vector space U −n has a basis Proof .(i) The elements {k s } s∈Z are linearly independent by Lemma 6.1, so they form a basis for a subalgebra of U which we denote by K.By Lemma 6.1 the sum U n = ∞ ℓ=0 e n+ℓ Kf ℓ is direct.We have Ke = eK since ke = q 2 ek, and similarly Kf = f K. Pick an integer t ≥ 0. By Lemma 6.2 and induction on t we find Λ t ∈ t ℓ=0 e ℓ Kf ℓ and Λ t − (q − q −1 ) 2t e t f t ∈ t−1 ℓ=0 e ℓ Kf ℓ .For the above t and all s ∈ Z we have e n k s Λ t ∈ t ℓ=0 e n+ℓ Kf ℓ and The result follows from these comments and the fact that {e n+t k s f t | s ∈ Z, t ∈ N} is a basis for U n .
(ii).Similar to the proof of (i) above.
We now consider the Z-grading {U n } n∈Z from the point of view of the equitable presentation.
Lemma 6.4.The F-algebra U is generated by ν x , y ±1 , ν z .Moreover In the previous section we discussed a Z-grading of U .In the present section we extend this to a Z-grading of U ⊗ F[a ±1 , b ±1 , c ±1 ].We obtain some results about the Z-grading of U ⊗ F[a ±1 , b ±1 , c ±1 ] that will be used to prove Theorem 2.18.By Lemma 6.4 the elements ν x , y ±1 , ν z form a generating set for the F-algebra U .Therefore the following is a generating set for the F-algebra U ⊗ F[a ±1 , b ±1 , c ±1 ]: Consider the Z-grading of U from below Lemma 6.1.This Z-grading of U induces a Z-grading of U ⊗ F[a ±1 , b ±1 , c ±1 ] whose homogeneous components are described as follows.For n ∈ Z the n-homogeneous component is ] the generators (7.1) are homogeneous with the following degrees: Proof .By Lemma 6.6 and the construction.
We comment on the homogeneous component The following is a basis for the F-vector space U 0 ⊗ F[a ±1 , b ±1 , c ±1 ]: Proof .By Lemma 6.7 and the construction.

Corollary 7.3.
There exists an F-algebra isomorphism Proof .Immediate from Lemma 7.2.
Definition 7.4.For n ∈ Z consider the map The map π n ⊗ 1 acts as the identity on the n-homogeneous component of U ⊗ F[a ±1 , b ±1 , c ±1 ] and zero on all other homogeneous components of is the homogeneous component of v that has degree n.
Lemma 7.5.In the table below we list some elements For each element v we display the homogeneous component πn (v) for −1 ≤ n ≤ 1.All other homogeneous components of v are zero.
Moreover Ω ♮ is from (2.18).Evaluate these lines using Lemma 6.10 and the fact that each of The following definition is for notational convenience.
We now give two lemmas of a slightly technical nature.
Lemma 7.8.For an integer i ≥ 0 the homogeneous components of (A ♮ ) i , (B ♮ ) i , (C ♮ ) i are described as follows.
(A ♮ ) i : The homogeneous component of degree n is zero unless 0 ≤ n ≤ i.The homogeneous component of degree 0 is (θ + θ −1 ) i and the homogeneous component of degree i is R i .
(B ♮ ) i : The homogeneous component of degree n is zero unless −i ≤ n ≤ 0. The homogeneous component of degree −i is L i and the homogeneous component of degree 0 is (ϑ + ϑ −1 ) i .
(C ♮ ) i : The homogeneous component of degree n is zero unless −i ≤ n ≤ i.The homogeneous component of degree −i is (−1) i q i 2 L i θ i and the homogeneous component of degree i is 9 The proof of Theorem 2.18 In this section we prove Theorem 2.18.Lemma 9.1 ([17, Theorem 7.5]).The following is a basis for the F-vector space ∆: We will be discussing the coefficients when an element of ∆ is written as a linear combination of the basis elements (9.1).To facilitate this discussion we define a bilinear form , : ∆×∆ → F such that u, v = δ u,v for all elements u, v in the basis (9.1).The bilinear form , is symmetric, and the basis (9.1) is orthonormal with respect to , .For v ∈ ∆, where the sum is over all elements A i B j C k Ω ℓ α r β s γ t in the basis (9.1).In this sum there are finitely many nonzero summands.
Proof of Theorem 2.18.Let J ⊆ ∆ denote the kernel of ♮.We show that J = 0. To do this we assume J = 0 and get a contradiction.Fix 0 = v ∈ J. Let S = S(v) denote the set of 7-tuples (i, j, k, ℓ, r, s, t) of nonnegative integers such that ijk = 0 and v, In this equation we apply ♮ to both sides and get 0 = For an element (i, j, k, ℓ, r, s, t) ∈ S define its height to be i + k and its depth to be j + k.For all integers n ≥ 0 let S + n (resp.S − n ) denote the set of elements in S that have height n (resp.depth n).By construction ) are not all empty.By construction S has finite cardinality, so finitely many of {S + n } ∞ n=0 (resp.{S − n } ∞ n=0 ) are nonempty.Define N = max{n|S + n = ∅} and M = max{n|S − n = ∅}.By construction S + N and S − M are nonempty.We now split the argument into two cases.
For the above values of (i, j, k) the corresponding values of (q The above line contains a sequence of Laurent polynomials in θ. (9.4) below contains a sequence of Laurent polynomials in θ.These two sequences are bases for the same vector space.
With the above comments in mind, equation ( 9.3) gives a nontrivial F-linear dependency among In the above line we multiply each term by θ M and obtain a nontrivial F-linear dependency among The above linear dependency contradicts Proposition 8.2, for the present case N ≤ M .Case M ≤ N : The argument is similar to the previous case.However the details are slightly different so we will show them.Apply πN to each side of (9.2).Pick (i, j, k, ℓ, r, s, t) ∈ S and consider the corresponding summand in (9.2).The image of this summand under πN is computed using Lemma 7.9, and found to be zero unless (i, j, k, ℓ, r, s, t) ∈ S + N .The result of the computation is that where R, ϑ are from Definition 7.6.By Lemma 7.7 R = 0. We have seen that U ⊗F[a ±1 , b ±1 , c ±1 ] contains no zero divisors.Therefore (9.5) Consider the above equation.We noted earlier that S + N is nonempty.By construction the scalar v, A i B j C k Ω ℓ α r β s γ t is nonzero for all (i, j, k, ℓ, r, s, t) ∈ S + N .For (i, j, k, ℓ, r, s, t) ∈ S + N , at least one of i, j, k is zero since ijk = 0. Moreover i + k = N and j + k ≤ N .For these constraints on i, j, k the possible solutions for (i, j, k) are (0, 0, N ), (1, 0, N − 1), . . ., (N − 1, 0, 1), (N, 0, 0), (N, 1, 0), . . ., (N, N − 1, 0), (N, N, 0).
For the above values of (i, j, k) the corresponding values of (q The above line contains a sequence of Laurent polynomials in ϑ. (9.6) below contains a sequence of Laurent polynomials in ϑ.These two sequences are bases for the same vector space.
(9.6)By these comments the equation (9.5) gives a nontrivial F-linear dependency among In the above line we multiply each term by ϑ N and obtain a nontrivial F-linear dependency among This linear dependency contradicts Proposition 8.3, for the present case M ≤ N .Both cases yield a contradiction under the assumption that J = 0. Therefore J = 0 so ♮ is injective.
10 Comments on the PSL 2 (Z) action Recall the generators ρ, σ of PSL 2 (Z) from below Definition 2.2.We will first consider the mathematics around σ.
One can readily check using Definition 2.6 that there exists an automorphism of U that sends More generally, for any nonzero ξ ∈ F there exists an automorphism of U that sends The above automorphism swaps U n and U −n for all n ∈ Z, where {U n } n∈Z is the Z-grading of U from below Lemma 6.1.With these comments in mind we now consider the algebra Lemma 10.1.There exists an automorphism Moreover σ2 = 1.
Lemma 10.2.The automorphism σ of U ⊗ F[a ±1 , b ±1 , c ±1 ] has the following effect on the Zgrading.For n ∈ Z, σ swaps the homogeneous components with degree n, −n.
Proof .In (10.1), (10.2) we gave the action of σ on some homogeneous generators for U ⊗ F[a ±1 , b ±1 , c ±1 ].For each generator its image under σ is homogeneous.Moreover the generator and its image have opposite degree.The result follows.
We now consider the automorphism σ of U ⊗ F[a ±1 , b ±1 , c ±1 ] from the point of view of the equitable presentation.
Lemma 10.3.In the table below we display some elements v of U ⊗ F[a ±1 , b ±1 , c ±1 ].For each element v we display the image σ(v) under the map σ from Lemma 10.1.
Proof .The images of y ⊗ 1, ν x ⊗ 1, ν z ⊗ 1 are obtained from Lemma 10.1, using Note 3.2 and y = k.The images of x ⊗ 1 and z ⊗ 1 are now obtained using (6.2).The image of Λ ⊗ 1 is found using Lemma 10.1 and (2.5).The image of ν y ⊗ 1 is found using row ν y in the table of Lemma 6.10.

.4)
Proof .To verify (10.3), in the equations (4.4)-(4.6)apply σ to each side, and evaluate the result using Lemma 10.1 and the fact that σ fixes Λ ⊗ 1.To verify (10.4) we refer to rows A ♮ and B ♮ of the table in Lemma 7.5.To each term in those rows, apply σ and evaluate the result using Lemma 10.3.
Proposition 10.5.The following diagram commutes: Proof .The F-algebra ∆ is generated by A, B, C. By this and (2.4) the F-algebra ∆ is generated by A, B, γ.Recall that σ swap A, B and fixes γ.By Lemma 10.4 σ swaps A ♮ , B ♮ and fixes γ ♮ .The result follows.
Proposition 10.5 shows that the action of σ on ∆ ♮ extends to an automorphism σ of U ⊗ F[a ±1 , b ±1 , c ±1 ] that has order 2. So far so good.We now turn to the mathematics around ρ.
Consider the following variation on U .
Definition 10.6.Define the F-algebra U ′ by generators X, Y , Z and relations The above presentation of U ′ resembles the equitable presentation of U , except that the generator Y −1 is missing.
By construction there exists an F-algebra homomorphism ι : We will need the fact that ι is injective.We will supply a proof shortly.
Lemma 10.7.The following is a basis for the F-vector space U : Proof .For all n ∈ N let V n denote the subspace of U spanned by those elements e r k s f t from (6.1) that satisfy r + t = n.By Lemma 6.1 the sum U = ∞ n=0 V n is direct.For all h, j ∈ N and i ∈ Z let us write x h y i z j in terms of e, k, f .By Proposition 2.12 and Note 2.14, Using this together with ke = q 2 ek and kf = q −2 f k, we find x h y i z j ∈ h+j n=0 V n and The result follows since {e h k i−h f j | h, j ∈ N, i ∈ Z} is a basis for U .
Lemma 10.8.The following is a basis for the F-vector space U ′ : Proof .Using the relations in Definition 10.6 we routinely find that the elements (10.5) span U ′ .The elements (10.5) are linearly independent, since their images under ι are linearly independent by Lemma 10.7.The result follows.
Proof .For the basis vectors (10.5) their images under ι are linearly independent by Lemma 10.7.
Consider the subalgebra of U generated by x, y, z.This subalgebra is the image of U ′ under ι.Invoking Lemma 10.9 we identify this subalgebra with U ′ via ι.The elements ν x , ν y , ν z are contained in U ′ by Definition 3.1, and Λ ∈ U ′ by Lemma 2.15.The algebra Lemma 10.10.The following is a basis for the F-vector space U ⊗ F[a ±1 , b ±1 , c ±1 ]: x h y i z j ⊗ a r b s c t , h, j ∈ N, i, r, s, t ∈ Z.
The following is a basis for the F-vector space U ′ ⊗ F[a ±1 , b ±1 , c ±1 ]: x h y i z j ⊗ a r b s c t , h, i, j ∈ N, r, s, t ∈ Z.
Consider the injection♮ : ∆ → U ⊗ F[a ±1 , b ±1 , c±1 ] from Theorem 2.16 and Theorem 2.18.Below Definition 2.2 we showed how PSL 2 (Z) acts on ∆ as a group of automorphisms.This PSL 2 (Z) action induces a PSL 2 (Z) action on the image ∆ ♮ .It is reasonable to ask whether this action extends to a PSL 2 (Z) action on U ⊗ F[a ±1 , b ±1 , c ±1 ] as a group of automorphisms.This extension does not quite work; let us examine what goes wrong.