Abstract
For a finite-dimensional simple Lie algebra \(\mathfrak {g}\), let \(U^+_q(\mathfrak {g})\) be the positive part of the quantized universal enveloping algebra \(U_q(\mathfrak {g})\) with respect to the triangular decomposition. It has the Poincaré–Birkhoff–Witt (PBW) base labeled with the longest element of the Weyl group W of \(\mathfrak {g}\). Let \(A_q(\mathfrak {g})\) be the quantized coordinate ring of \(\mathfrak {g}\). In this chapter, the intertwiner of the irreducible \(A_q(\mathfrak {g})\) modules labeled with two different reduced expressions of W is identified with the transition matrix of the corresponding PBW bases of \(U^+_q(\mathfrak {g})\). It leads to an alternative proof of the tetrahedron and 3D reflection equations within \(U^+_q(\mathfrak {g})\). The boundary vectors in Sects. 3.6.1, 5.8.1 and 8.6.1 give rise to invariants of an anti-algebra involution in \(U^+_q(\mathfrak {g})\) in an infinite product form.
Access provided by Autonomous University of Puebla. Download chapter PDF
Similar content being viewed by others
10.1 Quantized Universal Enveloping Algebra \(U_q(\mathfrak {g})\)
10.1.1 Definition
In this chapter \(\mathfrak {g}\) stands for a finite-dimensional simple Lie algebra. Its simple roots, simple coroots, fundamental weights are denoted by \(\{\alpha _i\}_{i\in I}\),\(\{h_i\}_{i\in I}\), \(\{\varpi _i\}_{i\in I}\), where I is the index set of the Dynkin diagram of \(\mathfrak {g}\). The weight lattice is \(P = \oplus _{i \in I} {\mathbb Z}\varpi _i\) and the Cartan matrix \((a_{ij})_{i,j\in I}\) is given by \(a_{ij}=\langle {h_i,\alpha _j}\rangle =2(\alpha _i, \alpha _j)/(\alpha _i,\alpha _i)\).
The quantized universal enveloping algebra \(U_q(\mathfrak {g})\) is an associative algebra over \({\mathbb Q}(q)\) generated by \(\{e_i, f_i, k_i^{\pm 1} \mid i\in I\}\) satisfying the relations:
Here we use the following notations: \(q_i=q^{(\alpha _i,\alpha _i)/2},[m]_i=(q_i^m-q_i^{-m})/(q_i-q_i^{-1})\), \([n]_i!=\prod _{m=1}^n[m]_i\), \(e_i^{(n)}=e_i^n/[n]_i!,f_i^{(n)}=f_i^n/[n]_i!\). We normalize the simple roots so that \(q_i=q\) when \(\alpha _i\) is a short root. The relation (10.1) is called q-Serre relation. The algebra \(U_q(\mathfrak {g})\) is a Hopf algebra. For the comultiplication (or coproduct), we adopt the followingFootnote 1:
10.1.2 PBW Basis
Let W be the Weyl group of \(\mathfrak {g}\). It is generated by simple reflections \(\{s_i\mid i\in I\}\) obeying the relations: \(s_i^2=1\), \((s_is_j)^{m_{ij}}=1\) (\(i\ne j\)), where \(m_{ij}=2,3,4,6\) for \(\langle {h_i,\alpha _j}\rangle \langle {h_j,\alpha _i}\rangle =0,1,2,3\), respectively. Let \(w_0\) be the longest element of W and fix a reduced expression \(w_0=s_{i_1}s_{i_2}\cdots s_{i_l}\). Then every positive root occurs exactly once in
Correspondingly, define elements \(e_{\beta _r}\in U_q(\mathfrak {g})\) (\(r=1,\ldots ,l\)) by
Here \(T_i\) is the action of the braid group on \(U_q(\mathfrak {g})\). It is an algebra automorphism and is given on the generators \(\{e_j\}\) by
Let \(U_q^+(\mathfrak {g})\) be a subalgebra of \(U_q(\mathfrak {g})\) generated by \(\{e_i\mid i\in I\}\). The only relation among them is the q-Serre relation (10.1) for \(e_i\)’s. It is known that \(e_{\beta _r} \in U^+_q(\mathfrak {g})\) holds for any r. \(U^+_q(\mathfrak {g})\) has the PBW basis. It depends on the reduced expression \(s_{i_1}s_{i_2}\cdots s_{i_l}\) of \(w_0\). Set \(\textbf{i}=(i_1,i_2,\ldots ,i_l)\) and define for \(A=(a_1,a_2,\ldots ,a_l)\in ({\mathbb Z}_{\ge 0})^l\)
Then \(\{E_\textbf{i}^A\mid A\in ({\mathbb Z}_{\ge 0})^l\}\) forms a basis of \(U_q^+(\mathfrak {g})\). We warn that the notations \(e_{i_r}\) with \(i_r \in I\) and \(e_{\beta _r}\) with a positive root \(\beta _r\) should be distinguished properly from the context. In particular \(e^{(a_r)}_{\beta _r} = (e_{\beta _r})^{a_r}/\prod _{m=1}^{a_r}\frac{p_r^m-p^{-m}_r}{p_r-p_r^{-1}}\) with \(p_r = q^{(\beta _r, \beta _r)/2}\).
10.2 Quantized Coordinate Ring \(A_q(\mathfrak {g})\)
10.2.1 Definition
Let us give the definition of the quantized coordinate ring \(A_q(\mathfrak {g})\).Footnote 2 The relation to the concrete realization by generators and relations in earlier chapters will be explained later.
Let \(O_\textrm{int}(\mathfrak {g})\) be the category of integrable left \(U_q(\mathfrak {g})\) modules M such that, for any \(v\in M\), there exists \(l\ge 0\) satisfying \(e_{i_1}\cdots e_{i_l}v=0\) for any \(i_1,\ldots ,i_l\in I\). Then \(O_\textrm{int}(\mathfrak {g})\) is semisimple and any simple object is isomorphic to the irreducible module \(V(\lambda )\) with dominant integral highest weight \(\lambda \). Similarly, we can consider the category \(O_\textrm{int}(\mathfrak {g}^\textrm{op})\) of integrable right \(U_q(\mathfrak {g})\) modules \(M^r\) such that, for any \(u\in M^r\), there exists \(l\ge 0\) satisfying \(uf_{i_1}\cdots f_{i_l}=0\) for any \(i_1,\ldots ,i_l\in I\). The superscript \(\textrm{op}\) signifies “opposite”. \(O_\textrm{int}(\mathfrak {g}^\textrm{op})\) is also semisimple and any simple object is isomorphic to the irreducible module \(V^r(\lambda )\) with dominant integral highest weight \(\lambda \). Let \(v_\lambda \) (resp. \(u_\lambda \)) be a highest weight vector of \(V(\lambda )\) (resp. \(V^r(\lambda )\)). Then there exists a unique bilinear form \((\,,\,)\)
satisfying
Let \(U_q(\mathfrak {g})^*\) be \(\textrm{Hom}_{{\mathbb Q}(q)}(U_q(\mathfrak {g}),{\mathbb Q}(q))\) and \(\langle {\,,\,}\rangle \) be the canonical pairing between \(U_q(\mathfrak {g})^*\) and \(U_q(\mathfrak {g})\). The comultiplication \(\Delta \) of \(U_q(\mathfrak {g})\) induces a multiplication of \(U_q(\mathfrak {g})^*\) by
thereby giving \(U_q(\mathfrak {g})^*\) the structure of \({\mathbb Q}(q)\)-algebra. It also has a \(U_q(\mathfrak {g})\) bimodule structure by
We define the subalgebra \(A_q(\mathfrak {g})\) of \(U_q(\mathfrak {g})^*\) by
and call it the quantized coordinate ring.
The following theorem is the q-analogue of the Peter–Weyl theorem.
Theorem 10.1
As a \(U_q(\mathfrak {g})\) bimodule, \(A_q(\mathfrak {g})\) is isomorphic to \(\bigoplus _\lambda V^r(\lambda )\otimes V(\lambda )\), where \(\lambda \) runs over all dominant integral weights, by the homomorphisms
given by
for \(u\in V^r(\lambda ),v\in V(\lambda )\), and \(g\in U_q(\mathfrak {g})\).Footnote 3
In our case of a finite-dimensional simple Lie algebra \(\mathfrak {g}\), \(A_q(\mathfrak {g})\) turns out to be a Hopf algebra. See for example [66, Chap. 9]. Its comultiplication is also denoted by \(\Delta \).
Let \(\mathcal {R}\) be the universal R matrix for \(U_q(\mathfrak {g})\). For its explicit formula see [29, p. 273] for example. For our purpose it is enough to know that
where \(q^{(\textrm{wt}\,\cdot ,\textrm{wt}\,\cdot )}\) is an operator acting on the tenor product \(v_\lambda \otimes v_\mu \) of weight vectors \(v_\lambda ,v_\mu \) of weight \(\lambda ,\mu \) by \(q^{(\textrm{wt}\,\cdot ,\textrm{wt}\,\cdot )}(v_\lambda \otimes v_\mu )=q^{(\lambda ,\mu )} v_\lambda \otimes v_\mu \), \(Q_+=\bigoplus _i{\mathbb Z}_{\ge 0}\alpha _i\), and \((U_q^\pm )_{\pm \beta }\) is the subspace of \(U_q^\pm (\mathfrak {g})\) spanned by root vectors corresponding to \(\pm \beta \).
Fix \(\lambda \), let \(\{u^\lambda _j\}\) and \(\{v^\lambda _i\}\) be bases of \(V^r(\lambda )\) and \(V(\lambda )\) such that \((u^\lambda _i,v^\lambda _j)=\delta _{ij}\). Set
Let R be the so-called constant R matrix for \(V(\lambda )\otimes V(\mu )\). Denoting the homomorphism \(U_q(\mathfrak {g})\rightarrow \textrm{End}(V(\lambda ))\) by \(\rho _\lambda \), it is given as
where P stands for the exchange of the first and second components. The scalar multiple is determined appropriately depending on \(\mathfrak {g}\). The reason we apply P is to fit the so-called RTT relation in (10.15). The dependence of R on \(\lambda \) and \(\mu \) has been suppressed in the notation. R satisfies
where \(\Delta ^{\textrm{op}}=P \circ \Delta \circ P\). Define matrix elements \(R^{ij}_{kl}\) by
Define the right action of R on \(V^r(\lambda )\otimes V^r(\mu )\) in such a way that \(((u^\lambda _i\otimes u^\mu _j)R,v^\lambda _k\otimes v^\mu _l)=(u^\lambda _i\otimes u^\mu _j,R(v^\lambda _k\otimes v^\mu _l))\) holds. Then we have
Now for any \(x \in U_q(\mathfrak {g})\), we have
Thus we get
We call such a relation an RTT relation. It forms a large family containing conventional ones as the special case where \(\lambda = \mu = \varpi _r\) for some specific fundamental weight \(\varpi _r\).
Example 10.2
Consider the simplest case \(\mathfrak {g}=A_1\) with \(\lambda = \mu =\varpi _1\). We write \(u^{\varpi _1}_i, v^{\varpi _1}_i\) simply as \(u_i, v_i\, (i = 1,2)\). The \(U_q(sl_2)\) module structure is
The R matrix (3.3) acts as
Set \(t_{ij} = \Psi _{\omega _1}(u_i \otimes v_j) \in A_q(A_1)\). Then we have
which reproduces the relation \([t_{11}, t_{22}] = (q-q^{-1}) t_{21}t_{12}\) in (3.9). Similarly, we have
Suppose \(x = e^l_1k^m_1f^n_1 \in U_q(sl_2)\, (l,m,n \in {\mathbb Z}_{\ge 0})\) without loss of generality. Since \(v^0_1:=v_1\otimes v_2-qv_2\otimes v_1\) is a \(U_q(sl_2)\)-singlet annihilated either by \(\Delta (e_1)\) and \(\Delta (f_1)\), one has \(\Delta (x) v^0_1 = \delta _{l0}\delta _{n0}v^0_1\). Thus the RHS of the above calculation is equal to \(\delta _{l0}\delta _{n0}(u_1 \otimes u_2, v^0_1) = \delta _{l0}\delta _{n0} = \langle 1, x\rangle \). This yields \(t_{11}t_{22} - q t_{12}t_{21} = 1\) in (3.9).
Let us mention the relation to the formulation of \(A_q(\mathfrak {g})\) in earlier chapters using specific generators and relations. Suppose \(\varpi _l\) is a fundamental weight such that any \(V(\lambda )\) is included in the tensor power \(V(\varpi _l)^{\otimes m}\) for some m.Footnote 4 Denoting the base of \(V^r(\varpi _l)\) and \(V(\varpi _l)\) by \(u_i\) and \(v_i\), set
We know that \(t_{ij}\) satisfies the RTT relation (10.15) whose structure constant is the constant R matrix for \(\lambda = \mu = \varpi _l\). Any vectors \(u \in V^r(\lambda )\) and \(v \in V(\lambda )\) are expressed as linear combinations \(u=\sum C_{i_1,\ldots , i_m} u_{i_1} \otimes \cdots \otimes u_{im}\) and \(v = \sum D_{j_1, \ldots , j_m}v_{j_1} \otimes \cdots \otimes v_{j_m}\). Theorem 10.1 shows that an arbitrary element of \(A_q(\mathfrak {g})\) is constructed as \(\Psi _\lambda (u \otimes v)\). A calculation similar to Example 10.2 leads to \(\Psi _\lambda (u \otimes v) = \sum C_{i_1,\ldots , i_m} D_{j_1, \ldots , j_m} t_{i_1 j_1} \cdots t_{i_m j_m}\), which says that \(t_{ij}\)’s are certainly generators. They satisfy RTT and additional relations reflecting a fine structure of the Grothendieck ring of \(\mathfrak {g}\) like \(V(\varpi _l)^{\otimes m} \supset V(0)\) and \(V(\varpi _l)^{\otimes m} \supset V(\varpi _l)\), etc. Our individual treatment in the earlier chapters corresponds to the choice \(l=1\) for \(A_{n-1}, C_n, G_2\) and \(l=n\) for \(B_n\).Footnote 5
10.2.2 Right Quotient Ring \(A_q(\mathfrak {g})_{\mathcal {S}}\)
Here we prepare the necessary ingredients for the proof of Theorem 10.6. The point is to assure the well definedness of the division in (10.39).
Recall that \(w_0 \in W\) is the longest element of the Weyl group. For any \(l \in I\), let \(v_{w_0\varpi _l} \in V(\varpi _l)\) be a lowest weight vector. Similarly, let \(u_{\varpi _l} \in V^r(\varpi _l)\) be a highest weight vector. The following element will play a key role:
Example 10.3
For \(\mathfrak {g}= A_1\) treated in Example 10.2, one has \(\sigma _1 = \Psi _{\omega _1}(u_1 \otimes v_2) = t_{12}\).
Proposition 10.4
The commutativity \(\sigma _r\sigma _s = \sigma _s\sigma _r\) holds for any \(r,s \in I\).
Proof
From (10.9) and (10.11) we have
where \((w_0\varpi _r, w_0\varpi _s) = (\varpi _r,\varpi _s)\) has been used. Consider the RTT relation (10.15) with \(\lambda = \varpi _r\), \(\mu = \varpi _s\), and take the indices i, j, k, l so as to specify the following bases:
Then (10.24) and (10.25) indicate \(R^{ij}_{mp} = q^{(\varpi _r,\varpi _s)}\delta ^i_m\delta ^i_p\) and \(R^{mp}_{kl} = q^{(\varpi _r,\varpi _s)}\delta ^m_k\delta ^p_l\). Thus the RTT relation (10.15) reduces to
The proof is finished by noting \(\varphi ^{\varpi _r}_{ik} = \sigma _r\) and \(\varphi ^{\varpi _s}_{jl} = \sigma _s\) by comparing (10.10) and (10.23). \(\square \)
Since \(A_q(\mathfrak {g})\) is a right \(U_q(\mathfrak {g})\) module, we have an element \(\sigma _i e_i \in A_q(\mathfrak {g})\). Later in Sect. 10.3.2, we will need the division \((\sigma _i e_i)/\sigma _i\) for \(i \in I\). The following localization is known to be possible making sense of it.
Theorem 10.5
Let n be the rank of \(\mathfrak {g}\). For the multiplicatively closed subset \(\mathcal {S}=\{\sigma _1^{m_1}\cdots \sigma _n^{m_n}\mid m_1,\ldots ,m_n\in {\mathbb Z}_{\ge 0}\} \subset A_q(\mathfrak {g})\), the right quotient ring \(A_q(\mathfrak {g})_\mathcal {S}\) exists.
Elements of \(A_q(\mathfrak {g})_\mathcal {S}\) are expressed in the form r/s with \(r \in A_q(\mathfrak {g})\) and \(s \in \mathcal {S}\). Theorem 10.5 guarantees the well-defined ring structure, namely, the addition and the multiplication of \(r_1/s_1\) and \(r_2/s_2\) in \(A_q(\mathfrak {g})_\mathcal {S}\) as
where \(u,u',v,v'\) are so chosen that \(s_1u=s_2u'\) (\(u\in \mathcal {S},u'\in A_q(\mathfrak {g})\)), \(r_2v=s_1v'\) (\(v\in \mathcal {S},v'\in A_q(\mathfrak {g})\)).
10.3 Main Theorem
In this section we fix two reduced words \(\textbf{i}=(i_1,\ldots ,i_l),\, \textbf{j}=(j_1,\ldots ,j_l)\) of the longest element \(w_0 = s_{i_1}\cdots s_{i_l} = s_{j_1}\cdots s_{j_l} \in W\).
10.3.1 Definitions of \(\gamma ^A_B\) and \(\Phi ^A_B\)
In the \(U_q(\mathfrak {g})\) side, we defined the PBW bases \(E_\textbf{i}^A,E_\textbf{j}^B\) of \(U_q^+(\mathfrak {g})\) in Sect. 10.1.2. We define their transition coefficient \(\gamma ^A_B\) by
In the \(A_q(\mathfrak {g})\) side, we have the intertwiner \(\Phi : \mathcal {F}_{q_{i_1}}\otimes \cdots \otimes \mathcal {F}_{q_{i_l}} \rightarrow \mathcal {F}_{q_{j_1}}\otimes \cdots \otimes \mathcal {F}_{q_{j_l}}\) satisfying
We take the parameters \(\mu _i\) as in (3.21) and (5.19) to be 1. The intertwiner \(\Phi \) is normalized by \(\Phi (|0\rangle \otimes \cdots \otimes |0\rangle ) = |0\rangle \otimes \cdots \otimes |0\rangle \). Under these conditions a matrix element \(\Phi ^A_B\) of \(\Phi \) is uniquely specified by
where \(A=(a_1, \ldots , a_l) \in ({\mathbb Z}_{\ge 0})^l\) and \(|A\rangle = |a_1\rangle \otimes \cdots \otimes |a_l\rangle \in \mathcal {F}_{q_{j_1}}\otimes \cdots \otimes \mathcal {F}_{q_{j_l}}\) and similarly for \(|B\rangle \in \mathcal {F}_{q_{i_1}}\otimes \cdots \otimes \mathcal {F}_{q_{i_l}}\). The main result of this chapter is
Theorem 10.6
For any pair \((\textbf{i},\textbf{j})\), from \(\textbf{i}\) one can reach \(\textbf{j}\) by applying Coxeter relations (for indices of the simple reflections). In view of the uniqueness of \(\gamma \) and \(\Phi \) and the fact that the braid group action \(T_i\) is an algebra homomorphism, the proof of this theorem reduces to establishing the same equality for the rank 2 case \(\mathfrak {g}=A_2, C_2\) and \(G_2\).Footnote 6 This will be done in the sequel.
10.3.2 Proof of Theorem 10.6 for Rank 2 Cases
In the rank 2 cases, there are two reduced expressions \(s_{i_1}\cdots s_{i_l}\) for the longest element of the Weyl group. Denote the associated sequences \(\textbf{i} = (i_1,\ldots , i_l)\) by \(\textbf{1}, \textbf{2}\) and set \(\textbf{1}' = \textbf{2}, \textbf{2}' = \textbf{1}\). Concretely, we take them as
where \(q_i\) defined after (10.1) is also recalled. In order to simplify the formulas in Sect. 10.4, we use the PBW bases and the Fock states in yet another normalization as follows:
where \(A=(a_1,\ldots , a_l)\). See after (10.1) for the symbol \([a]_i!\). The root vector \(e_{\beta _r}\) is defined in (10.4). Accordingly, we introduce the matrix elements \({\tilde{\gamma }}^A_B\) and \({\tilde{\Phi }}^A_B\) by
It follows that \(\gamma ^A_B = {\tilde{\gamma }}^A_B\prod _{k=1}^l([b_k]_{i_k}!/[a_k]_{i_k}!)\) and \(\Phi ^A_B= {\tilde{\Phi }}^A_B\prod _{k=1}^l(d_{i_k,a_k}/d_{i_k,b_k})\) for \(B=(b_1,\ldots , b_l)\). On the other hand, we know \(\Phi ^A_B=\Phi ^B_A\prod _{k=1}^l((q_{i_k}^2)_{b_k}/(q_{i_k}^2)_{a_k})\) from (3.63), (5.75) and (8.30). Due to the identity \((q_i^2)_md_{i,m}=[m]_i!\), the assertion \(\gamma ^A_B=\Phi ^A_B\) of Theorem 10.6 is equivalent to
Let \(\rho _\textbf{i}(x)=(\rho _\textbf{i}(x)_{A B})\) be the matrix for the left multiplication of \(x \in U^+_q(\mathfrak {g})\):
Let further \(\pi _\textbf{i}(g)=(\pi _\textbf{i}(g)_{A B})\) be the representation matrix of \(g \in A_q(\mathfrak {g})\):
The following element in the right quotient ring \(A_q(\mathfrak {g})_\mathcal {S}\) (see Theorem 10.5) will play a key role in our proof:
We recall that the general definition of \(\sigma _i\) is (10.23). Its concrete form in the rank 2 case will be given in Lemmas 10.10, 10.12 and 10.14. In Sect. 10.4 we will check the following statement case by case. It says that the “conjugation” of \(e_i\) by \(\sigma _i\) on \(A_q(\mathfrak {g})\) modules \((\sigma _i e_i)/ \sigma _i\) corresponds to \((1-q_i^2)e_i\) in \(U_q^+(\mathfrak {g})\).
Proposition 10.7
For \(\mathfrak {g}\) of rank 2, \(\pi _\textbf{i}(\sigma _i)\) is invertible and the following equality is valid:
where the RHS means \(\lambda _i\pi _\textbf{i}(\sigma _ie_i) \pi _\textbf{i}(\sigma _i)^{-1}\).
Proof of Theorem 10.6 for rank 2 case. We write both sides of (10.40) as \(M^i_{AB}\) and the term for \(\textbf{i}'\) instead of \(\textbf{i}\) as \(M^{\prime i}_{AB}\). From
we have \(\sum _{B} M^{\prime i}_{CB}{\tilde{\gamma }}^A_B = \sum _{B}{\tilde{\gamma }}^B_CM^i_{BA}\). On the other hand, the actions of the two sides of (10.29) with \(g=\xi _i\) and \(\textbf{j}=\textbf{i}'\) are calculated as
and
Hence \(\sum _{B} M_{CB}^{\prime i}{\tilde{\Phi }}_A^B = \sum _{B}{\tilde{\Phi }}_B^C M_{BA}^i\). Thus \({\tilde{\gamma }}^A_B\) and \({\tilde{\Phi }}^B_A\) satisfy the same relation. Moreover, the maps \(\pi _\textbf{i}\) and \(\rho _\textbf{i}\) are both homomorphisms, i.e. \(\pi _\textbf{i}(gh)=\pi _\textbf{i}(g)\pi _\textbf{i}(h)\) and \(\rho _\textbf{i}(xy)=\rho _\textbf{i}(x)\rho _\textbf{i}(y)\). We know that \({\Phi }\) is the intertwiner of the irreducible \(A_q(\mathfrak {g})\) modules and (10.36) obviously holds as \(1=1\) at \(A=B=(0,\ldots ,0)\). Thus it is valid for arbitrary A and B. \(\square \)
Remark 10.8
The equality (10.40) is valid for any \(\mathfrak {g}\).
10.4 Proof of Proposition 10.7
Here we present the explicit formulas of (10.37) with \(x=e_i\) and (10.38) with \(g=\sigma _i, \sigma _ie_i\) that allow one to check Proposition 10.7. In each case, there are two \(\textbf{i}\)-sequences, \(\textbf{1}\) and \(\textbf{2}=\textbf{1}'\) corresponding to the two reduced words. Define
Then both \(E^A_\textbf{i}\) in (10.6) and \({\tilde{E}}^{A}_\textbf{i}\) in (10.33) satisfy
where \(A^\vee = (a_l,\ldots , a_2,a_1)\) denotes the reversal of \(A=(a_1,a_2,\ldots , a_l)\). Applying \(\chi \) to (10.37) with \(x=e_i\) yields the right multiplication formula \({\tilde{E}}^{A^\vee }_{\textbf{i}'}\cdot e_i = \sum _B {\tilde{E}}^{B^\vee }_{\textbf{i}'} \rho _\textbf{i}(e_i)_{BA}\) for the \(\textbf{i}'\)-sequence. In view of this fact, we shall present the left and right multiplication formulas for \(\textbf{i}=\textbf{2}\) only.
As for (10.38) with \(g=\xi _i\) in (10.39), explicit formulas for \(\sigma _i, \sigma _i e_i \in A_q(\mathfrak {g})\) and their image by both representations \(\pi _\textbf{1}\) and \(\pi _\textbf{2}\) will be given. We include an exposition on how to use these data to check (10.40) along the simplest \(A_2\) case. The \(C_2\) and \(G_2\) cases are similar.
Following (10.34), we write \(|m\rangle \!\rangle :=d_{i,m}|m\rangle \in {\mathcal {F}}_{q_i}\) for each component. From the choice (10.30)–(10.32), the action of the \(q_i\)-oscillator on \({\mathcal {F}}_{q_i}\,(i=1,2)\) takes the form
See (10.34) and (3.13). We also use the shorthand
10.4.1 Explicit Formulas for \(A_2\)
Consider \(\mathfrak {g}= A_2\).
The q-Serre relations are
where \([m]_1 = \langle m \rangle /\langle 1 \rangle \). For simplicity we write the positive root vectors \(e_{\beta _{i_r}}\) in (10.4) with \((i_1, i_2, i_3) = \textbf{2}\) (10.30) as
The corresponding positive roots are \((\beta _1, \beta _2, \beta _3)=(\alpha _2, \alpha _1+\alpha _2, \alpha _1)\). In particular, \(b_2 = T_2(e_1)\). Their commutation relations are
Lemma 10.9
For \({\tilde{E}}^{a,b,c}_\textbf{2}=b_1^a b_2^b b_3^c\), we have
Proof
By induction, we have
The lemma is a direct consequence of these formulas. \(\square \)
Set \({\tilde{E}}^{a,b,c}_\textbf{1} = \chi ({\tilde{E}}^{c,b,a}_\textbf{2}) = \chi (b^a_3)\chi (b^b_2)\chi (b^c_1) =b_3^ab_2^{\prime b}b_1^c\), where \(b'_2 := \chi (b_2) = e_2e_1-qe_1e_2\). By applying \(\chi \) to the first two relations in Lemma 10.9, we get
Thus we find \(\rho _{\textbf{i}'}(e_i) = \rho _\textbf{i}(e_{3-i})\). This property is only valid for \(A_2\) and not in \(C_2\) and \(G_2\).
Let \(u_i\,(i=1,2,3)\) be the bases of the right \(U_q(A_2)\) module \(V^r(\varpi _1)\) such that \(u_j = u_1e_1\cdots e_{j-1}e_j\). Similarly, let \(v_i\,(i=1,2,3)\) be the bases of the left \(U_q(A_2)\) module \(V(\varpi _1)\) such that \(v_j = f_jf_{j-1}\cdots f_1 v_1\).
The left two columns specify the weights for example as \(u_2k_1 = q^{-1}u_2\), \(k_1v_1=qv_1\). For the coproduct (10.2), the bases of \(V^r(\varpi _2)\) and \(V(\varpi _2)\) are similarly given as
Here \(g= k_i, e_i, f_i\) are to be understood as \(\Delta (g)\) in (10.2).
Following (10.22) with \(l=1\) we set
for \(1 \le i,j \le 3\). They satisfy the relations (3.5) and (3.2) of the earlier definition of \(A_q(A_2)\). The formula (10.23) reads
where \((1+q^2)^{-1}\) is the normalization factor.Footnote 7 Thus we see \(\sigma _1 = t_{13}\). On the other hand, from
we find \(\sigma _2 = (1+q^2)^{-1} (t_{12} t_{23} - q t_{13}t_{22} - qt_{22}t_{13} + q^2 t_{23}t_{12})\).Footnote 8 Using the relations \([t_{12},t_{23}]=(q-q^{-1})t_{22}t_{13}\) and \([t_{22},t_{13}]=0\) from (3.5), this is simplified into \(\sigma _2 = t_{12}t_{23}-qt_{22}t_{13}\), which is the (3, 1)-quantum minor of \((t_{ij})_{1\le i,j \le 3}\).
Let us turn to \(\sigma _ie_i\). First we note
They imply
Using this and the coproduct \(\Delta \) in (10.2), we see
In these calculations, one should distinctively recognize that \(t_{13}e_1\) for instance is an action of \(e_1 \in U_q(A_2)\) on \(t_{13} \in A_q(A_2)\) viewed as an element of a right \(U_q(A_2)\) module, whereas \(t_{12}t_{33}\) is just a multiplication within \(A_q(A_2)\). To summarize, we have shown:
Lemma 10.10
For \(A_q(A_2)\), the following relations are valid:
From (3.35) and Lemma 10.10, we find
where a notation like \(\textbf{k}_1\mathrm{\textbf{a}}^+_3 = \textbf{k}\otimes 1 \otimes \mathrm{\textbf{a}}^+\) has been used. Since \(\textbf{k}\in \textrm{End}({\mathcal {F}}_q)\) is invertible, so is \(\pi _\textbf{i}(\sigma _i)\) and we may write
where \(\lambda _1=\lambda _2 = (1-q^2)^{-1}\). Thus (10.43) leads to
These formulas agree with (10.48) proving (10.40) for \(\textbf{i}=\textbf{1}\). The other case \(\textbf{i}=\textbf{2}\) also holds due to the symmetry \(\pi _\textbf{2}(\xi _i) = \pi _\textbf{1}(\xi _{3-i})\). Thus Proposition 10.7 is established for \(A_2\).
In terms of the 3DR in Chap. 3, Theorem 10.6 implies
This is valid either for \((\textbf{i}, \textbf{i}') =(\textbf{1},\textbf{2})\) or \((\textbf{2}, \textbf{1})\) thanks to (3.62). The weight conservation (3.48) assures the equality of weights of the two sides.
10.4.2 Explicit Formulas for \(C_2\)
Consider \(\mathfrak {g}= C_2\).
/ The q-Serre relations are
where \([m]_1 = \langle m \rangle /\langle 1 \rangle \) and \([m]_2 = \langle 2m \rangle /\langle 2 \rangle \). For simplicity we write the positive root vectors \(e_{\beta _{i_r}}\) in (10.4) with \((i_1,\ldots , i_4) = \textbf{2}\) (10.31) as
Their commutation relations are
Lemma 10.11
For \({\tilde{E}}^{a,b,c,d}_\textbf{2}=b_1^a b_2^b b_3^c b_4^d\), we have
Proof
By induction, we have
The lemma is a direct consequence of these formulas. \(\square \)
Set \({\tilde{E}}_\textbf{1}^{a,b,c,d} = \chi ({\tilde{E}}^{d,c,b,a}_\textbf{2})\). The left multiplication formula for this basis is deduced from the above lemma by applying \(\chi \).
Let \(u_i\) and \(v_i\,(i=1,2,3,4)\) be bases of \(V^r(\varpi _1)\) and \(V(\varpi _1)\) such that \(u_j = u_1e_1\cdots e_{j-1}e_j\) and \(v_j = f_jf_{j-1}\cdots f_1 v_1\), where \(e_3=e_1, f_3=f_1\) just temporarily.
The left two columns specify the weights as in the \(A_2\) case. For the coproduct (10.2), the bases of \(V(\varpi _2)\) and \(V^r(\varpi _2)\) are similarly given as
Arrows here indicate the images only up to overall normalization.
We adopt the definition of \(t_{ij}\) in (10.22) with \(l=1\) for \(1 \le i,j \le 4\). Then \(t_{ij}\)’s satisfy the relations (5.1), (5.2) of the earlier definition of \(A_q(C_2)\). The formula (10.23) reads as
By a calculation similar to \(A_q(A_2)\) using the commutation relations
we get:
Lemma 10.12
For \(A_q(C_2)\), the following relations are valid:
Images of the generators \(t_{ij}\) by the representations \(\pi _\textbf{1}\) and \(\pi _\textbf{2}\) in (10.31) are available in Sect. 5.4 as \(\pi _\textbf{1}(t_{ij}) = P_{14}P_{23}\pi _{2121}(\tilde{\Delta }(t_{ij}))P_{14}P_{23}\) and \(\pi _\textbf{2}(t_{ij}) = \pi _{2121}(\Delta (t_{ij}))\), where the conjugation by \(P_{14}P_{23}\) reverses the order of the four-fold tensor product. See (5.39) and (5.40). From (5.37), the relations (5.41)–(5.56) are displaying the concrete form of \(\pi _\textbf{2}(t_{ij}) K = K (P_{14}P_{23}\pi _\textbf{1}(t_{ij})P_{14}P_{23})\). For convenience, we pick those generators appearing in Lemma 10.12:
From this and Lemma 10.12 we get
Note that \({\pi _\textbf{i}(\sigma _i)}\) is invertible. Comparing these formulas with Lemma 10.11 by using (10.43), the equality (10.40) is directly checked. Thus Proposition 10.7 is established for \({C_2}\).
In terms of the 3D K in Chap. 5, Theorem 10.6 implies
The weight conservation (5.65) assures the equality of weights of the two sides.
10.4.3 Explicit Formulas for \(G_2\)
Consider \(\mathfrak {g}= G_2\).
The q-Serre relations are
where \({[m]_1 = \langle m \rangle /\langle 1 \rangle }\) and \({[m]_2 = \langle 3m \rangle /\langle 3 \rangle }\). For simplicity we write the positive root vectors \({e_{\beta _{i_r}}}\) in (10.4) with \({(i_1,\ldots , i_6) = \textbf{2}}\) (10.32) as
Their commutation relations are
Lemma 10.13
For \({{\tilde{E}}^{a,b,c,d,e,f}_\textbf{2}=b_1^a b_2^b b_3^cb_4^db_5^e b_6^f}\), we have
Proof
By induction, we have
The lemma is a direct consequence of these formulas. \(\square \)
Let \({v_i\,(i=1,\ldots , 7)}\) be the basis of \({V(\varpi _1)}\) for which the representation matrix is given by (8.79)–(8.81). Its highest and lowest weight vectors are \({v_1}\) and \({v_7}\), respectively. Let \({u_i \in V^r(\varpi _1)}\) be the dual base of \({v_i}\).
The representation \({V(\varpi _2)}\) is the adjoint representation with dimension 14. Its lowest weight vector is \({v^{(14)}_{14}}\) in (8.84), which is \({v_6\otimes v_7-q v_7\otimes v_6}\) in the notation here. The highest weight vector of \({V^r(\varpi _2)}\) is \({u_1\otimes u_2 - q u_2 \otimes u_1}\). From these facts we have
We define \({t_{ij}}\) by the formula (10.22) with \({l=1}\) for \({1\le i,j \le 7}\). They satisfy the relations (8.3) and (8.4) of the earlier definition of \({A_q(G_2)}\). By a calculation similar to \({A_q(A_2)}\) using the commutation relations
we getFootnote 9
Lemma 10.14
For \({A_q(G_2)}\), the following relations are valid:
Images of the generators \({t_{ij}}\) by the representations \({\pi _\textbf{1}}\) and \({\pi _\textbf{2}}\) in (10.31) are available from (8.11) and (8.12). For convenience, we present explicit formulas for those appearing in Lemma 10.14:
From this and Lemma 10.14 we get
Note that \({\pi _\textbf{i}(\sigma _i)}\) is invertible. Comparing these formulas with Lemma 10.13 by using (10.43), the equality (10.40) is directly checked. Thus Proposition 10.7 is established for \({G_2}\).
In terms of the intertwiner F in Chap. 8, Theorem 10.6 implies
The weight conservation (8.29) assures the equality of weights of the two sides.
10.5 Tetrahedron and 3D Reflection Equations from PBW Bases
The relation (10.58) serves as an auxiliary linear system by which the tetrahedron equation (2.6) is established as the non-linear consistency condition. To see this, consider a PBW basis (10.6) of \({U_q^+(A_3)}\) having the form \({E^{a,b,c,d,e}_{1,2,3,1,2,1}}\). In addition to \(E^{...ab...}_{...13...} = E^{...ba...}_{...31...}\), we may apply (10.58) as
reflecting the \(U_q^+(A_2)\) subalgebra structure. Then we have
There is another route going from \(E^{a,b,c,d,e,f}_{1,2,3,1,2,1}\) to \(E^{f_2,e_2,d_2,c_2,b_2,a_2}_{3,2,3,1,2,3}\) as
Comparison of them leads to
for arbitrary a, b, c, d, e, f and \(a_2,b_2,c_2,d_2,e_2,f_2\). The sums are over \(a_1,b_1,c_1,d_1,e_1,f_1\in {\mathbb Z}_{\ge 0}\) on both sides. They are finite sums due to the weight conservation (3.48). The identity (10.98) reproduces the tetrahedron equation (2.9).
A similar proof of the 3D reflection equation (4.3) is possible based on (10.80). We now start from a PBW basis (10.6) of \(U_q^+(C_3)\) having the form \(E^{a,b,c,d,e,f,g,h,i}_{3,2,3,1,2,1,3,2,1}\) and apply (10.97) and \(E^{...abcd...}_{...3232...} = \sum K^{abcd}_{ijkl}E^{...lkji...}_{...2323...}\). The two routes are as follows:
Thus we get
for any a, b, c, d, e, f, g, h, i and \(a_2, b_3, c_2, d_3, e_3, f_3, g_2, h_3, i_3\). The sums are over \(a_1, b_1, b_2, c_1, d_1, d_2, e_1, e_2, f_1, f_2, g_1, h_1, h_2, i_1, i_2 \in {\mathbb Z}_{\ge 0}\) on both sides. They are finite sums due to the weight conservation (3.48) and (5.65). The identity (10.99) reproduces the 3D reflection equation (4.5). By a parallel argument for \(U^+_q(B_3)\), the 3D reflection equation of type B (6.31) can also be derived.
10.6 \(\chi \)-Invariants
Theorem 10.6 implies non-trivial identities in (a completion of) \(U^+_q(\mathfrak {g})\). They are stated as invariance of some infinite products under the anti-involution \(\chi \) introduced in (10.42). Here we illustrate the derivation along \(\mathfrak {g}=A_2\) and present the results for \(C_2\) and \(G_2\). The point is to translate the boundary vectors in Sects. 3.6.1, 5.8.1 and 8.6.1 in terms of the PBW basis.
Let us write the boundary vectors (3.132) as
By comparing the coefficient of \(|a\rangle \otimes |b\rangle \otimes |c\rangle \) on the two sides of (3.143) using (3.47), we get
In view of (3.63), this is equivalent to
Multiply this by \(E_\textbf{2}^{k,j,i}\) and sum over \(i,j,k\in {\mathbb Z}_{\ge 0}\). From (10.46) and (10.6), the RHS gives
As for the LHS, we have
The first equality is due to (10.58) which is the \(A_2\) case of the main theorem of this chapter. The second equality is (10.42). The quantity within \(\chi \) in (10.104) is equal to (10.103). Thus we find that (10.103) is \(\chi \)-invariant. To describe the result neatly we introduce a quantum-dilogarithm-type infinite product:
Then a direct calculation using (3.132) yields
Thus we get a corollary of Theorem 10.6 and Proposition 3.28.
Corollary 10.15
Set \(c_i = (1-q^2)b_i, c'_i = \chi (c_i) \in U^+_q(A_2)\, (i=1,2,3)\) using \(b_i\) in (10.46) and the anti-algebra involution \(\chi \) in (10.41). Then the following equalities are valid:
Remark 10.16
By the rescaling \(e_1\rightarrow x e_1, e_2 \rightarrow ye_2\) with parameters x, y, the identity (10.107) is seemingly generalized to
containing x, y in the same manner as spectral parameters in the Yang–Baxter equation. The same holds for (10.108). Similar remarks apply to the \(C_2\) and \(G_2\) cases in the sequel where the parameters arranged along the positive roots fit the spectral parameters in the reflection and the \(G_2\) reflection equations.
The product (10.107) is expanded as
where the q-Serre relation (10.45) has been used to make it manifestly invariant under \(\chi \). Similarly, (10.108) is expanded as
For \(C_2\), the relevant results are (10.80) and Proposition 5.21 concerning the boundary vectors in (5.118)–(5.120). There are three identities corresponding to the choices of (r, k) in (5.136).
Corollary 10.17
Set \(c_i = (1-q^4)b_i\,(i=1,3), c_i= (1-q^2)b_i\,(i=2,4)\) and \(c'_i = \chi (c_i) \in U^+_q(C_2)\, (i=1,2,3,4)\) using \(b_i\) in (10.60) and the anti-algebra involution \(\chi \) in (10.41). Then the following equalities are valid:
For \(G_2\), the relevant result is Conjecture 8.9 for the boundary vector (8.61) and (10.96).
Corollary 10.18
Set \(c_i = (1-q^6)b_i\,(i=1,3,5), c_i= (1-q^2)b_i\,(i=2,4,6)\) and \(c'_i = \chi (c_i) \in U^+_q(G_2)\, (i=1,\ldots , 6)\) using \(b_i\) in (10.82) and the anti-algebra involution \(\chi \) in (10.41). If Conjecture 8.9 holds, the following equality is valid:
10.7 Bibliographical Notes and Comments
This chapter is an extended exposition of [102]. The braid group action (10.5) is introduced in [111]. The formulation of quantized coordinate ring in this chapter follows [76, 139]. See also [43] and [29, Chap. 7]. For quantum cluster algebra structure of quantized coordinate rings, see [52].
The Peter–Weyl-type Theorem 10.1 is taken from [76, Proposition 7.2.2]. Proposition 10.4 is a special case of [66, Corollary 9.1.4]. In [149, Theorem 7], \(U^+_q(\mathfrak {g})\) has been identified with an explicit subalgebra of \(A_q(\mathfrak {g})_{\mathcal {S}}\). A proof of Theorem 10.5 adapted to the present setting has been given in [102, Sect. 3.2]. The main result, Theorem 10.6, is due to [102, Theorem 5]. The case \(\mathfrak {g}=A_2\) was obtained earlier in the pioneering work [131]. Remark 10.8 is due to [141], where a unified conceptual proof of Theorem 10.6 has been attained. See also [128] for yet another proof using the representation theory of q-boson algebra and the Drinfeld pairing of \(U_q(\mathfrak {g})\). The multiplication rule on the PBW bases like Lemmas 10.9, 10.11 and 10.13 plays an important role also in the study of the positive principal series representations and modular double [61]. For type \(C_2\), one can adjust the definition of \(E^A_\textbf{i}\) in (10.6) with that in [148] by setting \(v=q^{-1}\). Some of the results like Lemma 10.13 have also been obtained in [147]. An analogue of Sect. 10.5 for quantum superalgebras has been argued in [151].
Notes
- 1.
This convention will be kept throughout the book.
- 2.
The definition and Theorem 10.1 are valid for any symmetrizable Kac–Moody algebra.
- 3.
- 4.
For example, in type B, it is the spin representation that qualifies this postulate rather than the vector representation. For type D, the argument in the text needs a slight modification since the two kinds of spin representations \(V(\varpi _{n-1})\) and \(V(\varpi _n)\) are necessary, but it does not influence the results in the chapter.
- 5.
As for \(F_4\) we did not present specific generators and relations.
- 6.
The \(B_2\) case reduces to \(C_2\) by the interchange of indices \(1 \leftrightarrow 2 \in I\).
- 7.
The normalization of \(\sigma _i\) actually does not matter since only \(\sigma _ie_i/\sigma _i\) will be used.
- 8.
- 9.
\({\sigma _2}\) and \({\sigma _2e_2}\) in [102, Eq. (42)] are \((-q)\) times those in Lemma 10.14.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
Copyright information
© 2022 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
About this chapter
Cite this chapter
Kuniba, A. (2022). Connection to PBW Bases of Nilpotent Subalgebra of \(U_q\). In: Quantum Groups in Three-Dimensional Integrability. Theoretical and Mathematical Physics. Springer, Singapore. https://doi.org/10.1007/978-981-19-3262-5_10
Download citation
DOI: https://doi.org/10.1007/978-981-19-3262-5_10
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-19-3261-8
Online ISBN: 978-981-19-3262-5
eBook Packages: Physics and AstronomyPhysics and Astronomy (R0)