Introduction
The tensor product of modules over a vertex operator algebra (VOA) V V V and its braiding isomorphism play an essential role in the representation theory of VOAs and in conformal field theory (CFT). Using the restriction technique of conformal blocks [25 , 26 , 51 ] , we obtain an explicit description of the braiding isomorphism on the bottom degrees of V V V -modules, which are modules over the Zhu algebra A ( V ) A(V) A ( V ) [56 ] . As a consequence, for a class of VOAs that are not necessarily rational or C 2 C_2 C 2 -cofinite, the category Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) of finite-dimensional left A ( V ) A(V) A ( V ) -modules is a braided tensor category with an explicit braiding isomorphism, and it is braided equivalent to the corresponding category of V V V -modules.
In the VOA tensor product theory, the module M 1 ⊠ P ( z ) M 2 M^1\boxtimes_{P(z)}M^2 M 1 ⊠ P ( z ) M 2 is defined by a universal property, and the braiding isomorphism is given by the commutativity isomorphism and the parallel transport of the P ( z ) P(z) P ( z ) -tensor products [39 ] , which depend on the analytic properties of intertwining operators among infinite-dimensional modules. On the other hand, the fusion rules can be computed through the bottom degrees M 2 ( 0 ) M^2(0) M 2 ( 0 ) and M 3 ( 0 ) ∗ M^3(0)^* M 3 ( 0 ) ∗ , and the bimodule A ( M 1 ) A(M^1) A ( M 1 ) over the Zhu algebra A ( V ) A(V) A ( V ) . This is the well-known fusion rules theorem claimed in [24 ] , which was revised by Li [48 ] , and a variant of it was proved by the second-named author [50 ] ; see also [31 , 32 ] . It is natural to expect that the tensor product and the braiding isomorphism can also be described using the bottom degrees, that is, on the category of finite-dimensional A ( V ) A(V) A ( V ) -modules.
Recently, Gao, the second-named author, and Zhu introduced (twisted) restricted conformal blocks in [25 , 26 ] . By restricting the module ( M 3 ) ′ (M^3)' ( M 3 ) ′ attached to ∞ \infty ∞ in the three-point conformal blocks on ( P 1 , ∞ , 1 , 0 ) (\mathbb{P}^1,\infty,1,0) ( P 1 , ∞ , 1 , 0 ) to its bottom degree M 3 ( 0 ) ∗ M^3(0)^* M 3 ( 0 ) ∗ , the second-named author introduced a left A ( V ) A(V) A ( V ) -module
M 1 ⊙ M 2 = M 1 ⊗ C M 2 L P 1 ∖ { 1 , 0 } ( V ) < 0 ⋅ ( M 1 ⊗ C M 2 ) M^1\odot M^2=\frac{M^1\otimes_{\mathbb{C}}M^2}{\mathcal{L}_{\mathbb{P}^1\setminus\{1,0\}}(V)_{<0}\cdot(M^1\otimes_{\mathbb{C}}M^2)} M 1 ⊙ M 2 = L P 1 ∖ { 1 , 0 } ( V ) < 0 ⋅ ( M 1 ⊗ C M 2 ) M 1 ⊗ C M 2
where L P 1 ∖ { 1 , 0 } ( V ) < 0 \mathcal{L}_{\mathbb{P}^1\setminus\{1,0\}}(V)_{<0} L P 1 ∖ { 1 , 0 } ( V ) < 0 is an ideal of the ∞ \infty ∞ -restricted chiral Lie algebra [51 , 26 ] . This quotient space also occurred in a construction of fusion tensor product of modules over C 2 C_2 C 2 -cofinite VOAs by Tsuchiya and Wood (see [54 ] , (2.84)]). If V V V is strongly rational, then M 1 ⊙ M 2 M^1\odot M^2 M 1 ⊙ M 2 is isomorphic to the bottom degree of Huang-Lepowsky’s P ( 1 ) P(1) P ( 1 ) -tensor product module M 1 ⊠ P ( 1 ) M 2 M^{1} \boxtimes_{P(1)} M^{2} M 1 ⊠ P ( 1 ) M 2 , and Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) is a tensor category with respect to the fusion tensor product S ⊠ T = Φ L ( S ) ⊙ Φ L ( T ) S \boxtimes T = \Phi^{L}(S) \odot\Phi^{L}(T) S ⊠ T = Φ L ( S ) ⊙ Φ L ( T ) [51 Theorem 6.8] , where Φ L \Phi^{L} Φ L is the generalized Verma module functor [15 , 13 ] . In this paper, we give an explicit formula for the braiding on Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) and extend this tensor product construction to VOAs that need not be rational or C 2 C_{2} C 2 -cofinite.
Main theorems
Let C 1 ( V ) C_{1}(V) C 1 ( V ) denote the category of C 1 C_{1} C 1 -cofinite grading-restricted generalized V V V -modules, and write Ω \Omega Ω for the bottom-degree functor. Throughout most of the paper, we assume that
C 1 ( V ) is semisimple, and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) . (1) C_{1}(V)\text{ is semisimple, and }\Phi^{L}(\operatorname{Mod}_{\mathrm{fd}}(A(V))) \subseteq C_{1}(V). \tag*{(1)} C 1 ( V ) is semisimple, and Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) . (1)
Then ( Φ L ⊣ Ω ) : Mod f d ( A ( V ) ) ⇄ C 1 ( V ) (\Phi^{L} \dashv\Omega) : \operatorname{Mod}_{\mathrm{fd}}(A(V)) \rightleftarrows C_{1}(V) ( Φ L ⊣ Ω ) : Mod fd ( A ( V )) ⇄ C 1 ( V ) is an equivalence of categories, see Proposition 3.11 and [1 Theorem 3.2] .
For w ∈ C × w \in\mathbb{C}^{\times} w ∈ C × , the same restriction at ∞ \infty ∞ gives a left A ( V ) A(V) A ( V ) -module M 1 ⊙ w , 0 M 2 M^{1} \odot_{w,0} M^{2} M 1 ⊙ w , 0 M 2 , with M 1 ⊙ 1 , 0 M 2 = M 1 ⊙ M 2 M^{1} \odot_{1,0} M^{2} = M^{1} \odot M^{2} M 1 ⊙ 1 , 0 M 2 = M 1 ⊙ M 2 , see Definition 3.7 . Our first main theorem identifies this module with the bottom degree of the P ( w ) P(w) P ( w ) -tensor product (see Theorem 3.14 ).
Theorem A. Assume C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^{L}(\operatorname{Mod}_{\mathrm{fd}}(A(V))) \subseteq C_{1}(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) . Let M 1 , M 2 ∈ C 1 ( V ) M^{1}, M^{2} \in C_{1}(V) M 1 , M 2 ∈ C 1 ( V ) . Then the canonical map
η w : M 1 ⊙ w , 0 M 2 ⟶ Ω ( M 1 ⊠ P ( w ) M 2 ) , v 1 ⊙ w v 2 ⟼ π 0 ( v 1 ⊠ P ( w ) v 2 ) , \eta_{w} : M^{1} \odot_{w,0} M^{2} \longrightarrow\Omega\left(M^{1} \boxtimes_{P(w)} M^{2}\right), \qquad v^{1} \odot_{w} v^{2} \longmapsto\pi_{0}\left(v^{1} \boxtimes_{P(w)} v^{2}\right), η w : M 1 ⊙ w , 0 M 2 ⟶ Ω ( M 1 ⊠ P ( w ) M 2 ) , v 1 ⊙ w v 2 ⟼ π 0 ( v 1 ⊠ P ( w ) v 2 ) , is a well-defined isomorphism of left A ( V ) A(V) A ( V ) -modules, which is natural in M 1 M^{1} M 1 and M 2 M^{2} M 2 , where π 0 \pi_{0} π 0 is the projection onto the bottom degree.
By shifting the marked points and using parallel transport of the restricted conformal blocks, we obtain the following isomorphisms (see Theorem 4.12 ).
Theorem B. Assume C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^{L}(\operatorname{Mod}_{\mathrm{fd}}(A(V))) \subseteq C_{1}(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) . Let M 1 , M 2 ∈ C 1 ( V ) M^{1}, M^{2} \in C_{1}(V) M 1 , M 2 ∈ C 1 ( V ) . Then there exist isomorphisms of left A ( V ) A(V) A ( V ) -modules which are natural in M 1 , M 2 M^{1}, M^{2} M 1 , M 2 :
R ± M 2 , M 1 : M 2 ⊙ M 1 ⟶ M 1 ⊙ M 2 , R_{\pm}^{M^{2},M^{1}} : M^{2} \odot M^{1} \longrightarrow M^{1} \odot M^{2}, R ± M 2 , M 1 : M 2 ⊙ M 1 ⟶ M 1 ⊙ M 2 , R ± M 2 , M 1 ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) , R_{\pm}^{M^{2},M^{1}}(v^{2} \odot v^{1}) = e^{\mp\pi i L(0)}\left(e^{\pm\pi i L(0)}v^{1} \odot e^{\pm\pi i L(0)}v^{2}\right), R ± M 2 , M 1 ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) , where the outer factor e ∓ π i L ( 0 ) e^{\mp\pi i L(0)} e ∓ π i L ( 0 ) acts on M 1 ⊙ M 2 M^{1} \odot M^{2} M 1 ⊙ M 2 through e ∓ π i [ ω ] e^{\mp\pi i[\omega]} e ∓ π i [ ω ] . Moreover,
( R − M 1 , M 2 ) − 1 = R + M 2 , M 1 and ( R + M 1 , M 2 ) − 1 = R − M 2 , M 1 . \left(R_{-}^{M^{1},M^{2}}\right)^{-1} = R_{+}^{M^{2},M^{1}} \qquad\text{and}\qquad\left(R_{+}^{M^{1},M^{2}}\right)^{-1} = R_{-}^{M^{2},M^{1}}. ( R − M 1 , M 2 ) − 1 = R + M 2 , M 1 and ( R + M 1 , M 2 ) − 1 = R − M 2 , M 1 .
Let R M 1 , M 2 = R − M 1 , M 2 R_{M^{1},M^{2}} = R_{-}^{M^{1},M^{2}} R M 1 , M 2 = R − M 1 , M 2 . Using Theorems A and B, we prove that R R R is the restriction of the braiding isomorphism c M 1 , M 2 c_{M^{1},M^{2}} c M 1 , M 2 of Huang–Lepowsky–Zhang [39 , 36 ] to the bottom degrees, namely, η 1 ∘ R M 1 , M 2 = Ω ( c M 1 , M 2 ) ∘ η 1 \eta_{1} \circ R_{M^{1},M^{2}} = \Omega(c_{M^{1},M^{2}}) \circ\eta_{1} η 1 ∘ R M 1 , M 2 = Ω ( c M 1 , M 2 ) ∘ η 1 , see Theorem 4.14 . Hence Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) is a braided tensor category with respect to the fusion tensor product S ⊠ T = Φ L ( S ) ⊙ Φ L ( T ) S \boxtimes T = \Phi^{L}(S) \odot\Phi^{L}(T) S ⊠ T = Φ L ( S ) ⊙ Φ L ( T ) and the braiding isomorphism
R M 1 , M 2 ( v 1 ⊙ v 2 ) = e π i [ ω ] ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) . (2) R_{M^{1},M^{2}}(v^{1} \odot v^{2}) = e^{\pi i[\omega]}\left(e^{-\pi iL(0)}v^{2} \odot e^{-\pi iL(0)}v^{1}\right). \tag*{(2)} R M 1 , M 2 ( v 1 ⊙ v 2 ) = e π i [ ω ] ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) . (2)
and Φ L : Mod f d ( A ( V ) ) ⟶ C 1 ( V ) \Phi^{L} : \operatorname{Mod}_{\mathrm{fd}}(A(V)) \longrightarrow C_{1}(V) Φ L : Mod fd ( A ( V )) ⟶ C 1 ( V ) is a braided tensor equivalence. In particular, if V V V is strongly rational, then Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) is a modular tensor category with the twist e 2 π i [ ω ] e^{2\pi i[\omega]} e 2 π i [ ω ] . Note that the associativity constraints on Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) are transported from C 1 ( V ) C_{1}(V) C 1 ( V ) through Φ L \Phi^{L} Φ L , while the braiding isomorphism (2) is given by an explicit formula on finite-dimensional A ( V ) A(V) A ( V ) -modules, which only involves the action of L ( 0 ) L(0) L ( 0 ) and the central element [ ω ] ∈ A ( V ) [\omega] \in A(V) [ ω ] ∈ A ( V ) . As far as we know, such a formula for the braiding isomorphism on the bottom degrees did not appear in the literature before, even in the strongly rational case.
Besides the strongly rational VOAs, condition (1) is satisfied by the affine VOAs at admissible levels, the affine VOAs V k ( g ) V^{k}(\mathfrak{g}) V k ( g ) at generic levels k + h ∨ ∉ Q k + h^{\vee} \notin\mathbb{Q} k + h ∨ ∈ / Q , the affine VOAs V − 2 + 1 p ( s l 2 ) V^{-2+\frac{1}{p} }(\mathfrak{sl}_{2}) V − 2 + p 1 ( sl 2 ) , and some affine VOAs at negative levels in [10 ] , see Example 3.12. Note that the second condition in (1) does not follow from the first one. For instance, it fails for the Virasoro VOA with generic central charge.
Relation to earlier work and proofs of the main theorems
The braided tensor structure on module categories has been studied extensively in the theory of both VOAs and CFTs. For instance, on the CFT side, Tsuchiya and Kanie proved that the conformal blocks of the WZNW-model on P 1 \mathbb{P}^{1} P 1 give rise to monodromy representations of the braid groups [53 ] , and Moore and Seiberg described the braiding and fusion matrices of rational CFTs by a system of polynomial equations [19 ] . For affine Lie algebras, these conformal blocks satisfy the Knizhnik-Zamolodchikov (KZ) equations [46 ] , whose monodromy was studied by Kohno [47 ] and Drinfeld [17 ] . Using the KZ equations, Drinfeld introduced a braided tensor category of modules over the quasi-triangular quasi-Hopf algebra U ( g ) [ [ ϖ ] ] U(\mathfrak{g})[[\varpi]] U ( g ) [[ ϖ ]] , whose commutativity isomorphism is given by e π i ϖ t e^{\pi i\varpi t} e π iϖ t , where t t t is a multiple of the Casimir tensor of g \mathfrak{g} g [17 ] ; and Kazhdan and Lusztig constructed a braided tensor structure on the category K L k ( g ) KL^{k}(\mathfrak{g}) K L k ( g ) of modules over the affine Lie algebra g ^ \widehat{\mathfrak{g}} g of level k k k , with k + h ∨ ∉ Q ≥ 0 k+h^{\vee}\notin\mathbb{Q}_{\geq0} k + h ∨ ∈ / Q ≥ 0 , and proved that it is equivalent to the module category of a quantum group [40 , 41 , 42 , 44 ] . On the VOA side, Huang and Lepowsky developed the tensor product theory for module categories of VOAs [33 , 34 , 35 ] , and Huang proved the convergence of products of intertwining operators [29 ] and the modularity of the module category of a strongly rational VOA [30 ] . This theory was generalized to the logarithmic case by Huang, Lepowsky, and Zhang [37 , 38 , 39 ] , and it was applied to many irrational VOAs, such as the affine VOAs at admissible levels [7 , 9 MY25] , the affine VOAs beyond admissible levels [10 ] , and the Virasoro VOAs [8 ] . Recently, Huang proved that the category C 1 ( V ) C_{1}(V) C 1 ( V ) admits a natural vertex tensor category structure for any VOA V V V of CFT-type [36 ] .
We prove Theorem A by relating the universal property of Huang-Lepowsky’s P ( w ) P(w) P ( w ) -tensor product [33 , 34 ] with the isomorphisms among the spaces of P ( w ) P(w) P ( w ) -intertwining maps, three-point conformal blocks, and ∞ \infty ∞ -restricted conformal blocks on ( P 1 , ∞ , w , 0 ) (\mathbb{P}^{1},\infty,w,0) ( P 1 , ∞ , w , 0 ) , see the diagram (48) . In comparison with [51 ] , one advantage of Theorem A is that we do not need V V V to be rational or C 2 C_{2} C 2 -cofinite. Moreover, η w \eta_{w} η w is given by an explicit formula, which will be used to compare our braiding isomorphism with the one of Huang-Lepowsky-Zhang.
To prove Theorem B, we first study the effect of shifting the marked points on the restricted conformal blocks. Recall that there is a linear isomorphism Ω 0 : I ( M 3 M 1 M 2 ) ≅ I ( M 3 M 2 M 1 ) \Omega_{0}:I\binom{M^{3}}{M^{1}\ M^{2}}\cong I\binom{M^{3}}{M^{2}\ M^{1}} Ω 0 : I ( M 1 M 2 M 3 ) ≅ I ( M 2 M 1 M 3 ) given by Huang and Lepowsky [34 ] . It turns out that the geometric realization of Ω 0 \Omega_{0} Ω 0 on the space of three-point conformal blocks is given by shifting the marked points ( w , 0 ) (w,0) ( w , 0 ) uniformly to ( 0 , − w ) (0,-w) ( 0 , − w ) , with the attached modules unchanged, see Proposition 4.2 and Remark 4.3. Since the factor e − w L ( 1 ) e^{-wL(1)} e − w L ( 1 ) in Ω 0 \Omega_{0} Ω 0 acts trivially on M 3 ( 0 ) ∗ M^{3}(0)^{*} M 3 ( 0 ) ∗ , Ω 0 \Omega_{0} Ω 0 restricts to the ∞ \infty ∞ -restricted conformal blocks, and it gives rise to a canonical isomorphism of left A ( V ) A(V) A ( V ) -modules
R w : M 2 ⊠ − w , 0 M 1 ⟶ M 1 ⊠ w , 0 M 2 , R w ( v 2 ⊠ − w v 1 ) = v 1 ⊠ w v 2 , R_{w}:M^{2}\mathbin{\boxtimes_{-w,0}}M^{1}\longrightarrow M^{1}\mathbin{\boxtimes_{w,0}}M^{2},\qquad R_{w}(v^{2}\mathbin{\boxtimes_{-w}}v^{1})=v^{1}\mathbin{\boxtimes_{w}}v^{2}, R w : M 2 ⊠ − w , 0 M 1 ⟶ M 1 ⊠ w , 0 M 2 , R w ( v 2 ⊠ − w v 1 ) = v 1 ⊠ w v 2 ,
see Theorem 4.6 . Then we move the point − 1 -1 − 1 back to 1 1 1 by the parallel transport of the three-point conformal blocks along the upper and lower half unit circles, see Subsection 4.3. Combining these maps gives the isomorphisms in Theorem B.
Affine examples
Our main example is the affine VOA L g ( k ) L_{\mathfrak{g}}(k) L g ( k ) at a positive integral level k k k , where everything in (1.2) can be written in terms of finite-dimensional g \mathfrak{g} g -modules. Let g \mathfrak{g} g be a finite-dimensional simple Lie algebra, κ = k + h ∨ \kappa=k+h^{\vee} κ = k + h ∨ , and Ω = ∑ a x a ⊗ x a \Omega=\sum_{a}x_{a}\otimes x^{a} Ω = ∑ a x a ⊗ x a the Casimir tensor of g \mathfrak{g} g . Recall that A ( L k ( g ) ) ≅ U ( g ) / I k A(L_{k}(\mathfrak{g}))\cong U(\mathfrak{g})/I_{k} A ( L k ( g )) ≅ U ( g ) / I k , with I k = ⟨ e θ k + 1 ⟩ I_{k}=\langle e_{\theta}^{k+1}\rangle I k = ⟨ e θ k + 1 ⟩ [24 ] . For λ , μ ∈ P + k \lambda,\mu\in P_{+}^{k} λ , μ ∈ P + k , we first show that
L ( λ ) ⊗ L ( μ ) ≅ T λ , μ = L ( λ ) ⊗ C L ( μ ) U ( g ) ⋅ J λ , μ , J λ , μ = v λ ⊗ e θ k − ( λ ∣ θ ) + 1 L ( μ ) + e θ k − ( μ ∣ θ ) + 1 L ( λ ) ⊗ v μ , L(\lambda)\otimes L(\mu)\cong T_{\lambda,\mu}=\frac{L(\lambda)\otimes_{\mathbb{C}}L(\mu)}{U(\mathfrak{g})\cdot J_{\lambda,\mu}},\qquad J_{\lambda,\mu}=v_{\lambda}\otimes e_{\theta}^{k-(\lambda\mid\theta)+1}L(\mu)+e_{\theta}^{k-(\mu\mid\theta)+1}L(\lambda)\otimes v_{\mu}, L ( λ ) ⊗ L ( μ ) ≅ T λ , μ = U ( g ) ⋅ J λ , μ L ( λ ) ⊗ C L ( μ ) , J λ , μ = v λ ⊗ e θ k − ( λ ∣ θ ) + 1 L ( μ ) + e θ k − ( μ ∣ θ ) + 1 L ( λ ) ⊗ v μ ,
where U ( g ) U(\mathfrak{g}) U ( g ) acts on L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) diagonally, see Theorem 5.1 , which is a variation of [1 Proposition 6.5] . Using the Segal-Sugawara construction and the relations in the ⊠ \boxtimes ⊠ -tensor product, we show that [ ω ] [\omega] [ ω ] acts on T λ , μ T_{\lambda,\mu} T λ , μ by h λ + h μ + 1 κ Ω λ , μ h_{\lambda}+h_{\mu}+\frac{1}{\kappa}\Omega_{\lambda,\mu} h λ + h μ + κ 1 Ω λ , μ , see Lemma 5.3 . Then Theorem B gives the following explicit braiding isomorphism on the modular tensor category Mod f d ( U ( g ) / I k ) \operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})/I_{k}) Mod fd ( U ( g ) / I k ) (see Theorem 5.4):
R λ , μ : T λ , μ → ∼ T μ , λ , R λ , μ ( π λ , μ ( u ⊗ v ) ) = π μ , λ ( ( τ ∘ exp ( π i κ Ω λ , μ ) ) ( u ⊗ v ) ) , (3) \mathcal{R}_{\lambda,\mu}:T_{\lambda,\mu}\xrightarrow{\sim}T_{\mu,\lambda},\qquad\mathcal{R}_{\lambda,\mu}\left(\pi_{\lambda,\mu}(u\otimes v)\right)=\pi_{\mu,\lambda}\left(\left(\tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega_{\lambda,\mu}\right)\right)(u\otimes v)\right), \tag*{(3)} R λ , μ : T λ , μ ∼ T μ , λ , R λ , μ ( π λ , μ ( u ⊗ v ) ) = π μ , λ ( ( τ ∘ exp ( κ π i Ω λ , μ ) ) ( u ⊗ v ) ) , (3)
where π λ , μ : L ( λ ) ⊗ C L ( μ ) → T λ , μ \pi_{\lambda,\mu}:L(\lambda)\otimes_{\mathbb{C}}L(\mu)\to T_{\lambda,\mu} π λ , μ : L ( λ ) ⊗ C L ( μ ) → T λ , μ is the quotient map, and τ ( u ⊗ v ) = v ⊗ u \tau(u\otimes v)=v\otimes u τ ( u ⊗ v ) = v ⊗ u . Note that the ideal I k I_k I k is not a Hopf ideal of U ( g ) U(\mathfrak{g}) U ( g ) , and L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) is in general not a U ( g ) / I k U(\mathfrak{g})/I_k U ( g ) / I k -module. It turns out that the operator τ ∘ exp ( π i κ Ω ) \tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega\right) τ ∘ exp ( κ π i Ω ) on L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) descends to the quotients T λ , μ T_{\lambda,\mu} T λ , μ , and gives the braiding isomorphism of the modular tensor category of L k ( g ) L_k(\mathfrak{g}) L k ( g ) -modules. Thus the fusion tensor product and the braiding for L k ( g ) L_k(\mathfrak{g}) L k ( g ) -modules are described by finite-dimensional g \mathfrak{g} g -modules, their quotients T λ , μ T_{\lambda,\mu} T λ , μ , and the Casimir tensor Ω \Omega Ω .
At generic levels, there is no quotient. If k + h ∨ ∉ Q k+h^{\vee}\notin\mathbb{Q} k + h ∨ ∈ / Q , or g = s l 2 \mathfrak{g}=\mathfrak{sl}_2 g = sl 2 and k = − 2 ± 1 p k=-2\pm\frac{1}{p} k = − 2 ± p 1 , then A ( V g ) ≅ U ( g ) A(V_{\mathfrak{g}})\cong U(\mathfrak{g}) A ( V g ) ≅ U ( g ) , and we show that S ⊠ T = S ⊗ C T S\boxtimes T=S\otimes_{\mathbb{C}}T S ⊠ T = S ⊗ C T with the diagonal action of U ( g ) U(\mathfrak{g}) U ( g ) , and the braiding isomorphism is τ ∘ exp ( π i κ Ω S , T ) \tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega_{S,T}\right) τ ∘ exp ( κ π i Ω S , T ) , see Proposition 6.2 and Corollary 6.3. This is Drinfeld’s commutativity isomorphism e π i ϑ t e^{\pi i\vartheta t} e π i ϑt , with t = d Ω t=d\Omega t = d Ω , at ϑ = 1 / ( d κ ) \vartheta=1/(d\kappa) ϑ = 1/ ( d κ ) [17 , 42 ] . In other words, the braiding isomorphism of the Drinfeld category is recovered from the restricted conformal blocks. However, the associativity constraints are not trivial, see Remark 6.6, and their description using the KZ connection will be addressed in a subsequent work.
This paper is organized as follows: we first recall the basics of module categories of VOAs, some examples of C 1 ( V ) C_1(V) C 1 ( V ) , and the three-point conformal blocks on P 1 \mathbb{P}^1 P 1 in Section 2. In Section 3, we introduce the category C A ( V ) \mathcal{C}_{A(V)} C A ( V ) and the left A ( V ) A(V) A ( V ) -module M 1 ⊙ w , 0 M 2 M^1\odot_{w,0}M^2 M 1 ⊙ w , 0 M 2 using the restricted conformal blocks, and prove Theorem A. In Section 4, we construct the isomorphisms R w R_w R w and T γ T_\gamma T γ and prove Theorem B. Then we compare our braiding isomorphism with the one of Huang-Lepowsky-Zhang, and show that Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) is a braided tensor category. In Section 5, we determine the fusion tensor product and the braiding isomorphism for the affine VOAs at positive integral levels. Finally, in Section 6, we discuss the affine VOAs at generic levels.
Preliminaries
Module categories of VOAs
For the definitions and basic properties of vertex operator algebras (VOAs), their modules, Borcherds Lie algebras, and Zhu algebras, we refer to [23 , 22 , 56 , 49 ] . We recall the module categories used in this paper and fix our conventions. All vector spaces are over C \mathbb{C} C , and N = { 0 , 1 , 2 , … } \mathbb{N}=\{0,1,2,\ldots\} N = { 0 , 1 , 2 , … } .
Throughout this paper, V V V is a VOA of CFT-type , that is,
V = ⨁ n ∈ N V n = C 1 ⊕ V + , V + = ⨁ n ≥ 1 V n . V=\bigoplus_{n\in\mathbb{N}}V_n=\mathbb{C}\mathbf{1}\oplus V_+,\qquad V_+=\bigoplus_{n\geq1}V_n. V = n ∈ N ⨁ V n = C 1 ⊕ V + , V + = n ≥ 1 ⨁ V n .
We denote its vacuum element and conformal element by 1 ∈ V 0 \mathbf{1}\in V_0 1 ∈ V 0 and ω ∈ V 2 \omega\in V_2 ω ∈ V 2 , respectively. For a homogeneous vector a ∈ V n a\in V_n a ∈ V n , we write wt a = n \operatorname{wt}a=n wt a = n . Recall the following definition for various module categories over a VOA V V V .
Definition 2.1. [22 , 33 , 56 , 15 , 49 , 30 H2S] A weak V V V -module is a vector space M M M equipped with a linear map Y M : V → End ( M ) [ [ z , z − 1 ] ] Y_M:V\to\operatorname{End}(M)[[z,z^{-1}]] Y M : V → End ( M ) [[ z , z − 1 ]] , denoted by
Y M ( a , z ) = ∑ n ∈ Z a ( n ) z − n − 1 , a ∈ V , Y_M(a,z)=\sum_{n\in\mathbb{Z}}a_{(n)}z^{-n-1},\qquad a\in V, Y M ( a , z ) = n ∈ Z ∑ a ( n ) z − n − 1 , a ∈ V ,
satisfying the lower truncation, vacuum property, and Jacobi identity. We write
Y M ( ω , z ) = ∑ n ∈ Z L ( n ) z − n − 2 . Y_M(\omega,z)=\sum_{n\in\mathbb{Z}}L(n)z^{-n-2}. Y M ( ω , z ) = n ∈ Z ∑ L ( n ) z − n − 2 .
Morphism in all module categories considered below are V V V -module homomorphisms.
An admissible or N \mathbb{N} N -gradable [56 , 15 ] V V V -module is a weak V V V -module M M M admitting a decomposition M = ⨁ n = 0 ∞ M ( n ) M=\bigoplus_{n=0}^{\infty}M(n) M = ⨁ n = 0 ∞ M ( n ) such that
a ( m ) M ( n ) ⊆ M ( n + wt a − m − 1 ) (4) a_{(m)}M(n)\subseteq M(n+\operatorname{wt}a-m-1) \tag*{(4)} a ( m ) M ( n ) ⊆ M ( n + wt a − m − 1 ) (4)
for every homogeneous a ∈ V a \in V a ∈ V , m ∈ Z m \in\mathbb{Z} m ∈ Z , and n ∈ N n \in\mathbb{N} n ∈ N , where M ( r ) = 0 M(r) = 0 M ( r ) = 0 for r < 0 r < 0 r < 0 . The category of admissible V V V -modules is denote by Adm ( V ) \operatorname{Adm}(V) Adm ( V ) . A VOA is called rational if the category Adm ( V ) \operatorname{Adm}(V) Adm ( V ) is semisimple.
(2) A generalized V V V -module is a weak V V V -module admitting a decomposition
M = ⨁ α ∈ C M [ α ] , M [ α ] = { v ∈ M : ( L ( 0 ) − α ) N v = 0 for some N ≥ 1 } . M = \bigoplus_{\alpha\in\mathbb{C}} M_{[\alpha]}, \qquad M_{[\alpha]} = \{v \in M : (L(0)-\alpha)^N v = 0 \text{ for some } N \geq1\}. M = α ∈ C ⨁ M [ α ] , M [ α ] = { v ∈ M : ( L ( 0 ) − α ) N v = 0 for some N ≥ 1 } .
A generalized module is called grading-restricted [30 , 36 ] if every M [ α ] M_{[\alpha]} M [ α ] is finite-dimensional and M [ α ] = 0 M_{[\alpha]} = 0 M [ α ] = 0 when Re α ≪ 0 \operatorname{Re}\alpha\ll0 Re α ≪ 0 .
Note that for homogeneous a ∈ V a \in V a ∈ V and α ∈ C \alpha\in\mathbb{C} α ∈ C , one has a ( m ) M [ α ] ⊆ M [ α + wt a − m − 1 ] a(m)M_{[\alpha]} \subseteq M_{[\alpha+\operatorname{wt}a-m-1]} a ( m ) M [ α ] ⊆ M [ α + wt a − m − 1 ] . Hence every grading-restricted generalized module is admissible:
M = ⨁ n = 0 ∞ M ( n ) , M = \bigoplus_{n=0}^{\infty} M(n), M = n = 0 ⨁ ∞ M ( n ) ,
where
M ( n ) = ⨁ β ∈ C / Z M [ β + n ] M(n) = \bigoplus_{\beta\in\mathbb{C}/\mathbb{Z}} M_{[\beta+n]} M ( n ) = β ∈ C / Z ⨁ M [ β + n ]
satisfying (4) .
(3) We denote by Ord ( V ) \operatorname{Ord}(V) Ord ( V ) the category of ordinary V V V -modules [22 ] . With our lower-bound convention, these are grading-restricted generalized modules on which L ( 0 ) L(0) L ( 0 ) acts semisimply:
M = ⨁ α ∈ C M α , M α = { v ∈ M : L ( 0 ) v = α v } . M = \bigoplus_{\alpha\in\mathbb{C}} M_{\alpha}, \qquad M_{\alpha} = \{v \in M : L(0)v = \alpha v\}. M = α ∈ C ⨁ M α , M α = { v ∈ M : L ( 0 ) v = α v } .
An irreducible ordinary module has the form M = ⨁ n ∈ N M h + n M = \bigoplus_{n \in\mathbb{N}} M_{h+n} M = ⨁ n ∈ N M h + n , where M h ≠ 0 M_h \ne0 M h = 0 . The number h ∈ C h \in\mathbb{C} h ∈ C is called the conformal weight of M M M .
(4) For a weak V V V -module M M M , set
C 1 ( M ) = span C { a ( − 1 ) v : a ∈ V + , v ∈ M } . C_1(M) = \operatorname{span}_{\mathbb{C}}\{a_{(-1)}v : a \in V_+,\, v \in M\}. C 1 ( M ) = span C { a ( − 1 ) v : a ∈ V + , v ∈ M } .
The module M M M is called C 1 C_1 C 1 -cofinite if dim C M / C 1 ( M ) < ∞ \dim_{\mathbb{C}} M/C_1(M) < \infty dim C M / C 1 ( M ) < ∞ [29 ] . Introduce the following category [36 ] :
C 1 ( V ) = { C 1 -cofinite grading-restricted generalized V -modules } . (5) C_1(V) = \{C_1\text{-cofinite grading-restricted generalized }V\text{-modules}\}. \tag*{(5)} C 1 ( V ) = { C 1 -cofinite grading-restricted generalized V -modules } . (5)
In particular, C 1 ( V ) C_1(V) C 1 ( V ) is a full-subcategory of Adm ( V ) \operatorname{Adm}(V) Adm ( V ) .
For a grading-restricted generalized module M M M , its contragredient module is the restricted dual
M ′ = ⨁ α ∈ C M α ∗ , M' = \bigoplus_{\alpha\in\mathbb{C}} M_\alpha^*, M ′ = α ∈ C ⨁ M α ∗ ,
equipped with Y M ′ : V → End ( M ′ ) [ [ z , z − 1 ] ] Y_{M'}: V \to\operatorname{End}(M')[[z,z^{-1}]] Y M ′ : V → End ( M ′ ) [[ z , z − 1 ]] determined by
⟨ Y M ′ ( a , z ) f , v ⟩ = ⟨ f , Y M ( e z L ( 1 ) ( − z − 2 ) L ( 0 ) a , z − 1 ) v ⟩ (6) \langle Y_{M'}(a,z)f,v\rangle= \langle f,Y_M(e^{zL(1)}(-z^{-2})^{L(0)}a,z^{-1})v\rangle \tag*{(6)} ⟨ Y M ′ ( a , z ) f , v ⟩ = ⟨ f , Y M ( e z L ( 1 ) ( − z − 2 ) L ( 0 ) a , z − 1 ) v ⟩ (6)
for a ∈ V a \in V a ∈ V , f ∈ M ′ f \in M' f ∈ M ′ , and v ∈ M v \in M v ∈ M [22 , 39 ] . In particular, as an adjoint module, V V V itself has a contragredient module V ′ V' V ′ . Set
C 2 ( V ) = span C { a ( − 2 ) b : a , b ∈ V } . C_2(V) = \operatorname{span}_{\mathbb{C}}\{a_{(-2)}b : a,b \in V\}. C 2 ( V ) = span C { a ( − 2 ) b : a , b ∈ V } .
The VOA V V V is C 2 C_2 C 2 -cofinite [56 ] if dim C V / C 2 ( V ) < ∞ \dim_{\mathbb{C}} V/C_2(V) < \infty dim C V / C 2 ( V ) < ∞ .
Definition 2.2. [15 , 16 ] A VOA V V V is called strongly rational if it is of CFT-type, simple, rational, C 2 C_2 C 2 -cofinite, and self-dual V ≅ V ′ V \cong V' V ≅ V ′ . In this case, Any irreducible admissible V V V -module is ordinary; there are finitely many irreducible ordinary modules up to isomorphism, denoted by
W = { V = M 0 , M 1 , … , M N } ; \mathcal{W} = \{V = M^0, M^1, \ldots, M^N\}; W = { V = M 0 , M 1 , … , M N } ; and the conformal weight h M i h_{M^i} h M i of an irreducible ordinary module M i M^i M i is a rational number.
The following theorem was proved by Huang:
Theorem 2.3. [36 Theorem 1.2] Let V V V be a CFT-type VOA. Then the category C 1 ( V ) C_1(V) C 1 ( V ) of C 1 C_1 C 1 -cofinite grading-restricted generalized V V V -modules admits a natural vertex tensor category structure. In particular, C 1 ( V ) C_1(V) C 1 ( V ) is a braided tensor category with a twist.
Examples of C 1 ( V ) C_1(V) C 1 ( V )
The category C 1 ( V ) C_1(V) C 1 ( V ) will be the principal VOA module category considered in this paper. We list some examples of VOA and this category.
Strongly rational VOAs
If V V V is strongly rational, then any grading-restricted generalized V V V -module is ordinary and C 1 C_1 C 1 -cofinite [2 ] . In particular,
C 1 ( V ) = Ord ( V ) ⊊ Adm ( V ) , (7) C_1(V) = \operatorname{Ord}(V) \subsetneq\operatorname{Adm}(V), \tag*{(7)} C 1 ( V ) = Ord ( V ) ⊊ Adm ( V ) , (7)
which is semisimple. C 1 ( V ) C_1(V) C 1 ( V ) consist precisely of finite direct sums of irreducible ordinary modules. Moreover, Adm ( V ) \operatorname{Adm}(V) Adm ( V ) is generated by C 1 ( V ) C_1(V) C 1 ( V ) by arbitrary direct sums. Furthermore, Ord ( V ) \operatorname{Ord}(V) Ord ( V ) is a modular tensor category, with fusion tensor product ⊠ = ⊠ P ( 1 ) \boxtimes= \boxtimes_{P(1)} ⊠ = ⊠ P ( 1 ) , tensor unit V V V , and twist θ M = e 2 π i L ( 0 ) \theta_M = e^{2\pi iL(0)} θ M = e 2 π i L ( 0 ) [30 , 39 ] .
Examples include positive-definite even lattice VOAs V L V_L V L [11 ] , Virasoro minimal model VOAs L ( c p , q , 0 ) L(c_{p,q}, 0) L ( c p , q , 0 ) [55 , 18 ] , affine VOAs at positive integral levels L k ( g ) L_k(\mathfrak{g}) L k ( g ) [14 ] , and the principal series W \mathcal{W} W -algebras W k ( g ) W_k(\mathfrak{g}) W k ( g ) [3 , 6 ] , etc.
The examples listed below are all irrational:
Admissible level affine VOAs
Let g \mathfrak{g} g be a finite-dimensional simple Lie algebra, and let p , q ∈ Z ≥ 1 p,q \in\mathbb{Z}_{\geq1} p , q ∈ Z ≥ 1 be coprime integers. Define k ∈ Q k \in\mathbb{Q} k ∈ Q by [45 ] :
k + h ∨ = p q , p ≥ { h ∨ if ( r ∨ , q ) = 1 , h if ( r ∨ , q ) = r ∨ . k+h^\vee=\frac{p}{q},\qquad p\geq\begin{cases}h^\vee& \text{if }(r^\vee,q)=1,\\ h & \text{if }(r^\vee,q)=r^\vee.\end{cases} k + h ∨ = q p , p ≥ { h ∨ h if ( r ∨ , q ) = 1 , if ( r ∨ , q ) = r ∨ .
The simple affine VOA L k ( g ) L_k(\mathfrak{g}) L k ( g ) at level k k k is called an admissible level affine VOA . The category O k , Ord \mathcal{O}_{k,\operatorname{Ord}} O k , Ord of ordinary L k ( g ) L_k(\mathfrak{g}) L k ( g ) -modules in the category O k \mathcal{O}_k O k of g ^ \widehat{\mathfrak{g}} g -modules is semisimple [4 ] . Moreover, it was proved in [10 Section 3.4] that
C 1 ( L k ( g ) ) = O k , Ord , (8) C_1(L_k(\mathfrak{g}))=\mathcal{O}_{k,\operatorname{Ord}}, \tag*{(8)} C 1 ( L k ( g )) = O k , Ord , (8)
which is also semisimple. Prior to Theorem 2.3 , it was proved that O k , Ord \mathcal{O}_{k,\operatorname{Ord}} O k , Ord is a braided tensor category with a twist [7 ] , and it is rigid when g \mathfrak{g} g is simply-laced [9 ] .
Generic level affine VOAs
Let g \mathfrak{g} g be a finite-dimensional simple Lie algebra, and let k ∈ C k \in\mathbb{C} k ∈ C be such that k + h ∨ ∉ Q ≥ 0 k+h^\vee\notin\mathbb{Q}_{\geq0} k + h ∨ ∈ / Q ≥ 0 . Such a level k k k is called generic . In this case, the vacuum module VOA V k ( g ) V^k(\mathfrak{g}) V k ( g ) is simple [40 ] , and so it is the affine VOA V k ( g ) = L k ( g ) V^k(\mathfrak{g})=L_k(\mathfrak{g}) V k ( g ) = L k ( g ) . It was proved in [10 Corollary 3.4.5] that
C 1 ( V k ( g ) ) = K L k ( g ) (9) C_1(V^k(\mathfrak{g}))=KL^k(\mathfrak{g}) \tag*{(9)} C 1 ( V k ( g )) = K L k ( g ) (9)
is the Kazhdan–Lusztig’s category for the affine Lie algebra g \mathfrak{g} g , which contains finite-length grading-restricted generalized V k ( g ) V^k(\mathfrak{g}) V k ( g ) -modules. Note that K L k ( g ) KL^k(\mathfrak{g}) K L k ( g ) was proved to admit a monoidal structure in [40 , 41 ] . Note that this category is semisimple if k + h ∨ ∉ Q k+h^\vee\notin\mathbb{Q} k + h ∨ ∈ / Q .
However, if k + h ∨ ∈ Q < 0 k+h^\vee\in\mathbb{Q}_{<0} k + h ∨ ∈ Q < 0 , then K L k ( g ) KL^k(\mathfrak{g}) K L k ( g ) may or may not be semisimple. For instance, let
V = V − 2 − 1 p ( s l 2 ) , p ∈ Z > 0 . V=V^{-2-\frac{1}{p}}(\mathfrak{sl}_2),\qquad p\in\mathbb{Z}_{>0}. V = V − 2 − p 1 ( sl 2 ) , p ∈ Z > 0 .
Then the category C 1 ( V ) = K L − 2 − 1 p ( s l 2 ) C_1(V)=KL_{-2-\frac{1}{p}}(\mathfrak{sl}_2) C 1 ( V ) = K L − 2 − p 1 ( sl 2 ) is semisimple. Its objects are finite direct sums of the simple s l ^ 2 \widehat{\mathfrak{sl}}_2 sl 2 modules whose top spaces are finite-dimensional s l 2 \mathfrak{sl}_2 sl 2 -modules [10 , 5 ] .
But if we take
V = V − 2 − a b ( s l 2 ) , a ≥ 2 , b ≥ 1 , gcd ( a , b ) = 1. V=V^{-2-\frac{a}{b}}(\mathfrak{sl}_2),\qquad a\geq2,\ b\geq1,\ \gcd(a,b)=1. V = V − 2 − b a ( sl 2 ) , a ≥ 2 , b ≥ 1 , g cd( a , b ) = 1.
Then it follows from Kazhdan–Lusztig equivalence [42 , 44 ] that
C 1 ( V − 2 − a b ( s l 2 ) ) = K L − 2 − a b ( s l 2 ) ≃ C ( q , s l 2 ) , q = e π i k + 2 = e − π i b a , (10) C_1\left(V^{-2-\frac{a}{b}}(\mathfrak{sl}_2)\right)=KL_{-2-\frac{a}{b}}(\mathfrak{sl}_2)\simeq\mathcal{C}(q,\mathfrak{sl}_2),\qquad q=e^{\frac{\pi i}{k+2}}=e^{-\frac{\pi i b}{a}}, \tag*{(10)} C 1 ( V − 2 − b a ( sl 2 ) ) = K L − 2 − b a ( sl 2 ) ≃ C ( q , sl 2 ) , q = e k + 2 π i = e − a π ib , (10)
where C ( q , s l 2 ) \mathcal{C}(q,\mathfrak{sl}_2) C ( q , sl 2 ) is the module category of the quantum group U q ( s l 2 ) U_q(\mathfrak{sl}_2) U q ( sl 2 ) , which is not semisimple.
The Virasoro VOAs at generic central charge
Let V = L ( c , 0 ) V=L(c,0) V = L ( c , 0 ) be the Virasoro VOA [24 ] , with central charge
c = 13 − 6 ( t + t − 1 ) , t ∉ Q . c=13-6(t+t^{-1}),\qquad t\notin\mathbb{Q}. c = 13 − 6 ( t + t − 1 ) , t ∈ / Q .
It was proved in [8 Theorems 3.1.4 and 5.1.2] that C 1 ( L ( c , 0 ) ) C_1(L(c,0)) C 1 ( L ( c , 0 )) is semisimple, with infinitely many simple objects
L ( c , h r , s ( t ) ) , r , s ∈ Z > 0 , where h r , s ( t ) = ( r t − s ) 2 − ( t − 1 ) 2 4 t . L(c,h_{r,s}(t)),\qquad r,s\in\mathbb{Z}_{>0},\qquad\text{where}\quad h_{r,s}(t)=\frac{(rt-s)^2-(t-1)^2}{4t}. L ( c , h r , s ( t )) , r , s ∈ Z > 0 , where h r , s ( t ) = 4 t ( r t − s ) 2 − ( t − 1 ) 2 .
Every object in C 1 ( L ( c , 0 ) ) C_1(L(c,0)) C 1 ( L ( c , 0 )) is a finite direct sum of these modules.
Formal intertwining operators and P ( w ) P(w) P ( w ) -intertwining operators
In this paper, we primarily consider C 1 ( V ) C_1(V) C 1 ( V ) , or a full tensor subcategory, whose objects are ordinary V V V -modules. For any three such objects, every logarithmic intertwining operator among them contains no logarithmic terms [38 Remark 3.23] , and hence is an intertwining operator in the usual sense [22 ] .
Lemma 2.4. If C 1 ( V ) C_1(V) C 1 ( V ) is semisimple, then every object of C 1 ( V ) C_1(V) C 1 ( V ) is an ordinary V V V -module.
Proof. Let W W W be an irreducible grading-restricted generalized module in C 1 ( V ) C_1(V) C 1 ( V ) . Choose α ∈ C \alpha\in\mathbb{C} α ∈ C such that W [ α ] ≠ 0 W_{[\alpha]}\ne0 W [ α ] = 0 , and let 0 ≠ v ∈ W [ α ] 0\ne v\in W_{[\alpha]} 0 = v ∈ W [ α ] . There exists a positive integer N N N such that u = ( L ( 0 ) − α ) N − 1 v ≠ 0 u=(L(0)-\alpha)^{N-1}v\ne0 u = ( L ( 0 ) − α ) N − 1 v = 0 , ( L ( 0 ) − α ) N v = 0 (L(0)-\alpha)^N v=0 ( L ( 0 ) − α ) N v = 0 . Then 0 ≠ u ∈ W α 0\ne u\in W_\alpha 0 = u ∈ W α , where W α W_\alpha W α denotes the eigenspace of L ( 0 ) L(0) L ( 0 ) with eigenvalue α \alpha α . Set W ′ = ⨁ α ∈ C W α W'=\bigoplus_{\alpha\in\mathbb{C}}W_\alpha W ′ = ⨁ α ∈ C W α . Then W ′ W' W ′ is a nonzero submodule of W W W , so irreducibility implies that W = W ′ W=W' W = W ′ . Hence W W W is ordinary.
For any object M ∈ C 1 ( V ) M\in C_1(V) M ∈ C 1 ( V ) , semisimplicity gives a decomposition M = ⨁ i W i M=\bigoplus_i W^i M = ⨁ i W i , where each W i W^i W i is irreducible. Since each W i W^i W i is ordinary and M M M is grading-restricted, M M M is also ordinary.
Examples of semisimple C 1 ( V ) C_1(V) C 1 ( V ) are given in the previous subsection.
Definition 2.5. [22 ] Let M 1 , M 2 , M 3 M^1,M^2,M^3 M 1 , M 2 , M 3 be ordinary V V V -modules. A (formal) intertwining operator of type ( M 3 M 1 M 2 ) \binom{M^3}{M^1\ M^2} ( M 1 M 2 M 3 ) is a linear map
I ( ⋅ , x ) : M 1 → Hom ( M 2 , M 3 ) { x , x − 1 } , I ( v 1 , x ) v 2 = ∑ n ∈ C v ( n ) 1 v 2 x − n − 1 , I(\cdot,x):M^1\to\operatorname{Hom}(M^2,M^3)\{x,x^{-1}\},\qquad I(v^1,x)v^2=\sum_{n\in\mathbb{C}}v^1_{(n)}v^2x^{-n-1}, I ( ⋅ , x ) : M 1 → Hom ( M 2 , M 3 ) { x , x − 1 } , I ( v 1 , x ) v 2 = n ∈ C ∑ v ( n ) 1 v 2 x − n − 1 , where x x x is a formal variable, satisfying the following axioms:
(1) (truncation property) For any v 1 ∈ M 1 v^1\in M^1 v 1 ∈ M 1 and v 2 ∈ M 2 v^2\in M^2 v 2 ∈ M 2 , v ( n ) 1 v 2 = 0 v^1_{(n)}v^2=0 v ( n ) 1 v 2 = 0 when the real part of n n n is sufficiently large.
(2) (L ( − 1 ) L(-1) L ( − 1 ) -derivative property) For any v 1 ∈ M 1 v^1\in M^1 v 1 ∈ M 1 , I ( L ( − 1 ) v 1 , x ) = d d x I ( v 1 , x ) I(L(-1)v^1,x)=\frac{d}{dx}I(v^1,x) I ( L ( − 1 ) v 1 , x ) = d x d I ( v 1 , x ) .
(3) (Formal Jacobi identity) For any v 1 ∈ M 1 v^1\in M^1 v 1 ∈ M 1 , v 2 ∈ M 2 v^2\in M^2 v 2 ∈ M 2 , and a ∈ V a\in V a ∈ V , one has
x 0 − 1 δ ( x 1 − x 2 x 0 ) Y M 3 ( a , x 1 ) I ( v 1 , x 2 ) v 2 − x 0 − 1 δ ( − x 2 + x 1 x 0 ) I ( v 1 , x 2 ) Y M 2 ( a , x 1 ) v 2 = x 2 − 1 δ ( x 1 − x 0 x 2 ) I ( Y M 1 ( a , x 0 ) v 1 , x 2 ) v 2 . (11) \begin{aligned}
x_0^{-1}\delta\left(\frac{x_1-x_2}{x_0}\right)Y_{M^3}(a,x_1)I(v^1,x_2)v^2
&-x_0^{-1}\delta\left(\frac{-x_2+x_1}{x_0}\right)I(v^1,x_2)Y_{M^2}(a,x_1)v^2\\
&=x_2^{-1}\delta\left(\frac{x_1-x_0}{x_2}\right)I(Y_{M^1}(a,x_0)v^1,x_2)v^2. \tag*{(11)}
\end{aligned} x 0 − 1 δ ( x 0 x 1 − x 2 ) Y M 3 ( a , x 1 ) I ( v 1 , x 2 ) v 2 − x 0 − 1 δ ( x 0 − x 2 + x 1 ) I ( v 1 , x 2 ) Y M 2 ( a , x 1 ) v 2 = x 2 − 1 δ ( x 2 x 1 − x 0 ) I ( Y M 1 ( a , x 0 ) v 1 , x 2 ) v 2 . (11) In particular, v ( n ) 1 M α 2 ⊂ M wt ( v 1 ) − n − 1 + α 3 v^1_{(n)}M^2_\alpha\subset M^3_{\operatorname{wt}(v^1)-n-1+\alpha} v ( n ) 1 M α 2 ⊂ M wt ( v 1 ) − n − 1 + α 3 for all n , α ∈ C n,\alpha\in\mathbb{C} n , α ∈ C .
The vector space of intertwining operators of type ( M 3 M 1 M 2 ) \binom{M^3}{M^1\ M^2} ( M 1 M 2 M 3 ) is denoted by I ( M 3 M 1 M 2 ) \mathcal{I}\binom{M^3}{M^1\ M^2} I ( M 1 M 2 M 3 ) . Its dimension, denoted by N ( M 3 M 1 M 2 ) N\binom{M^3}{M^1\ M^2} N ( M 1 M 2 M 3 ) , is called the fusion rule among M 1 M^1 M 1 , M 2 M^2 M 2 , and M 3 M^3 M 3 .
Remark 2.6. Assume M 1 , M 2 , M 3 M^1,M^2,M^3 M 1 , M 2 , M 3 are irreducible ordinary V V V -modules of conformal weights h 1 , h 2 , h 3 h_1,h_2,h_3 h 1 , h 2 , h 3 , respectively. Let M i = ⨁ n = 0 ∞ M i ( n ) M^i=\bigoplus_{n=0}^{\infty}M^i(n) M i = ⨁ n = 0 ∞ M i ( n ) , where M i ( n ) = M h i + n i M^i(n)=M^i_{h_i+n} M i ( n ) = M h i + n i is the eigenspace of L ( 0 ) L(0) L ( 0 ) of eigenvalue h i + n h_i+n h i + n . Denote deg ( v i ) = n \deg(v^i)=n deg ( v i ) = n for v i ∈ M i ( n ) v^i\in M^i(n) v i ∈ M i ( n ) , i = 1 , 2 , 3 i=1,2,3 i = 1 , 2 , 3 , and write h = h 1 + h 2 − h 3 h=h_1+h_2-h_3 h = h 1 + h 2 − h 3 . Then the intertwining operator can be written as
I ( v 1 , x ) v 2 = ∑ n ∈ Z v ( n ) 1 v 2 ⋅ x − n − 1 − h . I(v^1,x)v^2=\sum_{n\in\mathbb{Z}}v^1_{(n)}v^2\cdot x^{-n-1-h}. I ( v 1 , x ) v 2 = n ∈ Z ∑ v ( n ) 1 v 2 ⋅ x − n − 1 − h .
The components satisfy v ( n ) 1 M 2 ( m ) ⊂ M 3 ( deg v 1 − n − 1 + m ) v^1_{(n)}M^2(m)\subset M^3(\deg v^1-n-1+m) v ( n ) 1 M 2 ( m ) ⊂ M 3 ( deg v 1 − n − 1 + m ) , m ∈ N m\in\mathbb{N} m ∈ N , n ∈ Z n\in\mathbb{Z} n ∈ Z . However, if the modules M i M^i M i do not have a uniform conformal weight h i h_i h i , one cannot make such a simplification.
Let w ∈ P 1 ∖ { ∞ , 0 } w\in\mathbb{P}^1\setminus\{\infty,0\} w ∈ P 1 ∖ { ∞ , 0 } . Fix an integer p ∈ Z p\in\mathbb{Z} p ∈ Z for the branch of logarithm of w w w :
l p ( w ) = Log ( w ) + 2 p π i , where Log ( w ) = log ( ∣ w ∣ ) + i Arg ( w ) (12) l_p(w)=\operatorname{Log}(w)+2p\pi i,\qquad\text{where }\operatorname{Log}(w)=\log(|w|)+i\operatorname{Arg}(w) \tag*{(12)} l p ( w ) = Log ( w ) + 2 p π i , where Log ( w ) = log ( ∣ w ∣ ) + i Arg ( w ) (12)
is the principal branch of log ( w ) \log(w) log ( w ) , with the branch cut along the negative real axis. i.e., − π < Arg ( w ) ≤ π -\pi<\operatorname{Arg}(w)\leq\pi − π < Arg ( w ) ≤ π . Then w n = e n l p ( w ) w^n=e^{nl_p(w)} w n = e n l p ( w ) for any n ∈ C n\in\mathbb{C} n ∈ C .
The notion of P ( w ) P(w) P ( w ) -intertwining operators was introduced by Huang-Lepowsky. Given a graded vector space W = ⨁ n ∈ C W ( n ) W=\bigoplus_{n\in\mathbb{C}}W(n) W = ⨁ n ∈ C W ( n ) , with dim W ( n ) < ∞ \dim W(n)<\infty dim W ( n ) < ∞ for all n n n , let its graded dual space be W ′ = ⨁ n ∈ C W ( n ) ∗ W'=\bigoplus_{n\in\mathbb{C}}W(n)^* W ′ = ⨁ n ∈ C W ( n ) ∗ . Write
W ‾ = Hom C ( W ′ , C ) = ∏ n ∈ C W ( n ) . \overline{W}=\operatorname{Hom}_{\mathbb{C}}(W',\mathbb{C})=\prod_{n\in\mathbb{C}}W(n). W = Hom C ( W ′ , C ) = n ∈ C ∏ W ( n ) .
If ( W , Y W ) (W,Y_W) ( W , Y W ) is an ordinary V V V -module, then the action of a ( n ) = Res z z n Y W ( a , z ) a_{(n)}=\operatorname{Res}_{z}z^nY_W(a,z) a ( n ) = Res z z n Y W ( a , z ) on W W W extends to an action on W ‾ \overline{W} W , for all a ∈ V a\in V a ∈ V , n ∈ Z n\in\mathbb{Z} n ∈ Z .
Definition 2.7. [33 , 35 ] Let M 1 M^1 M 1 , M 2 M^2 M 2 , M 3 M^3 M 3 be three ordinary V V V -modules. A P ( w ) P(w) P ( w ) -intertwining operator is a linear map
F : M 1 ⊗ C M 2 ⟶ M 3 ‾ = ∏ α ∈ C M α 3 F:M^1\otimes_{\mathbb{C}}M^2\longrightarrow\overline{M^3}=\prod_{\alpha\in\mathbb{C}}M^3_\alpha F : M 1 ⊗ C M 2 ⟶ M 3 = α ∈ C ∏ M α 3 satisfying the formal Jacobi identity
x 0 − 1 δ ( x 1 − w x 0 ) Y M 3 ( a , x 1 ) F ( v 1 ⊗ v 2 ) = w − 1 δ ( x 1 − x 0 w ) F ( Y M 1 ( a , x 0 ) v 1 ⊗ v 2 ) + x 0 − 1 δ ( w − x 1 − x 0 ) F ( v 1 ⊗ Y M 2 ( a , x 1 ) v 2 ) . (13) x_0^{-1}\delta\left(\frac{x_1-w}{x_0}\right)Y_{M^3}(a,x_1)F(v_1\otimes v_2)
=
w^{-1}\delta\left(\frac{x_1-x_0}{w}\right)F(Y_{M^1}(a,x_0)v_1\otimes v_2)
+
x_0^{-1}\delta\left(\frac{w-x_1}{-x_0}\right)F(v_1\otimes Y_{M^2}(a,x_1)v_2). \tag*{(13)} x 0 − 1 δ ( x 0 x 1 − w ) Y M 3 ( a , x 1 ) F ( v 1 ⊗ v 2 ) = w − 1 δ ( w x 1 − x 0 ) F ( Y M 1 ( a , x 0 ) v 1 ⊗ v 2 ) + x 0 − 1 δ ( − x 0 w − x 1 ) F ( v 1 ⊗ Y M 2 ( a , x 1 ) v 2 ) . (13)
The vector space of P ( w ) P(w) P ( w ) -intertwining operators is denoted by M [ P ( w ) ] M 1 M 2 M 3 \mathcal{M}[P(w)]^{M^3}_{M^1M^2} M [ P ( w ) ] M 1 M 2 M 3 .
Note that the formal Jacobi identity above also has an equivalent residue form:
Res x 1 Y M 3 ( a , x 1 ) F ( v 1 ⊗ v 2 ) ⋅ ι ∣ x 1 ∣ > ∣ w ∣ ( x 1 n ( x 1 − w ) l ) − Res x 1 F ( v 1 ⊗ Y M 2 ( a , x 1 ) v 2 ) ⋅ ι ∣ w ∣ > ∣ x 1 ∣ ( x 1 n ( x 1 − w ) l ) = Res x 1 − w F ( Y M 1 ( a , x 1 − w ) v 1 ⊗ v 2 ) ⋅ ι ∣ w ∣ > ∣ x 1 − w ∣ ( x 1 n ( x 1 − w ) l ) , (14) \begin{aligned}
&\operatorname{Res}_{x_1}Y_{M^3}(a,x_1)F(v^1\otimes v^2)\cdot\iota_{|x_1|>|w|}\left(x_1^n(x_1-w)^l\right)\\
&\quad-\operatorname{Res}_{x_1}F(v^1\otimes Y_{M^2}(a,x_1)v^2)\cdot\iota_{|w|>|x_1|}\left(x_1^n(x_1-w)^l\right)\\
&=\operatorname{Res}_{x_1-w}F(Y_{M^1}(a,x_1-w)v^1\otimes v^2)\cdot\iota_{|w|>|x_1-w|}\left(x_1^n(x_1-w)^l\right), \tag*{(14)}
\end{aligned} Res x 1 Y M 3 ( a , x 1 ) F ( v 1 ⊗ v 2 ) ⋅ ι ∣ x 1 ∣ > ∣ w ∣ ( x 1 n ( x 1 − w ) l ) − Res x 1 F ( v 1 ⊗ Y M 2 ( a , x 1 ) v 2 ) ⋅ ι ∣ w ∣ > ∣ x 1 ∣ ( x 1 n ( x 1 − w ) l ) = Res x 1 − w F ( Y M 1 ( a , x 1 − w ) v 1 ⊗ v 2 ) ⋅ ι ∣ w ∣ > ∣ x 1 − w ∣ ( x 1 n ( x 1 − w ) l ) , (14)
for all n , l ∈ Z n,l\in\mathbb{Z} n , l ∈ Z . Indeed, by applying Res x 0 Res x 1 x 0 l x 1 n \operatorname{Res}_{x_0}\operatorname{Res}_{x_1}x_0^lx_1^n Res x 0 Res x 1 x 0 l x 1 n to (13) , we obtain (14) , with the indicated expansions of the rational functions. Conversely, (14) gives the coefficient of x 0 − l − 1 x 1 − n − 1 x_0^{-l-1}x_1^{-n-1} x 0 − l − 1 x 1 − n − 1 in (13) for every l , n ∈ Z l,n\in\mathbb{Z} l , n ∈ Z . Hence it implies (13) .
We briefly recall the argument in [35 Section 12] showing that
I ( M 3 M 1 M 2 ) ≅ M [ P ( w ) ] M 1 M 2 M 3 . I\left(\frac{M^3}{M^1\ M^2}\right)\cong\mathcal{M}[P(w)]^{M^3}_{M^1M^2}. I ( M 1 M 2 M 3 ) ≅ M [ P ( w ) ] M 1 M 2 M 3 .
Let I ∈ I ( M 3 M 1 M 2 ) I\in I\left(\frac{M^3}{M^1M^2}\right) I ∈ I ( M 1 M 2 M 3 ) . Fix a branch l p ( w ) l_p(w) l p ( w ) for log ( w ) \log(w) log ( w ) (12) , define F I , p P ( w ) ∈ M [ P ( w ) ] M 1 M 2 M 3 F^{P(w)}_{I,p}\in\mathcal{M}[P(w)]^{M^3}_{M^1M^2} F I , p P ( w ) ∈ M [ P ( w ) ] M 1 M 2 M 3 by
F I , p P ( w ) ( v 1 ⊗ v 2 ) : = I ( v 1 , e l p ( w ) ) v 2 = ∑ n ∈ C v ( n ) 1 v 2 ⋅ e ( − n − 1 ) l p ( w ) ∈ M 3 ‾ . (15) F^{P(w)}_{I,p}(v^1\otimes v^2):=I(v^1,e^{l_p(w)})v^2=\sum_{n\in\mathbb{C}}v^1_{(n)}v^2\cdot e^{(-n-1)l_p(w)}\in\overline{M^3}. \tag*{(15)} F I , p P ( w ) ( v 1 ⊗ v 2 ) := I ( v 1 , e l p ( w ) ) v 2 = n ∈ C ∑ v ( n ) 1 v 2 ⋅ e ( − n − 1 ) l p ( w ) ∈ M 3 . (15)
Conversely, given F ∈ M [ P ( w ) ] M 1 M 2 M 3 F \in\mathcal{M}[P(w)]_{M^1M^2}^{M^3} F ∈ M [ P ( w ) ] M 1 M 2 M 3 , define I F , p ∈ I ( M 3 M 1 M 2 ) I_{F,p} \in I\binom{M^3}{M^1\ M^2} I F , p ∈ I ( M 1 M 2 M 3 ) by
I F , p ( v 1 , x ) v 2 : = ∑ n ∈ C v ( n ) 1 v 2 ⋅ x − n − 1 ∈ M 3 { x } , (16) I_{F,p}(v^1,x)v^2:=\sum_{n\in\mathbb{C}}v^1_{(n)}v^2\cdot x^{-n-1}\in M^3\{x\}, \tag*{(16)} I F , p ( v 1 , x ) v 2 := n ∈ C ∑ v ( n ) 1 v 2 ⋅ x − n − 1 ∈ M 3 { x } , (16)
where v ( n ) 1 v 2 ∈ M wt ( v 1 ) − n − 1 + wt ( v 2 ) 3 v^1_{(n)}v^2\in M^3_{\operatorname{wt}(v^1)-n-1+\operatorname{wt}(v^2)} v ( n ) 1 v 2 ∈ M wt ( v 1 ) − n − 1 + wt ( v 2 ) 3 is the projection of F ( v 1 ⊗ v 2 ) ∈ M 3 F(v^1\otimes v^2)\in M^3 F ( v 1 ⊗ v 2 ) ∈ M 3 onto this homogeneous graded subspace multiplied by e ( n + 1 ) l p ( w ) ∈ C e^{(n+1)l_p(w)}\in\mathbb{C} e ( n + 1 ) l p ( w ) ∈ C .
Lemma 2.8. [[35 Proposition 12.2] ]. Fix a branch l p ( w ) l_p(w) l p ( w ) for log ( w ) \log(w) log ( w ) (12) , so that w m = e m l p ( w ) w^m=e^{ml_p(w)} w m = e m l p ( w ) for any m ∈ C m\in\mathbb{C} m ∈ C . The linear maps I ↦ F I , p I\mapsto F_{I,p} I ↦ F I , p (15) and F ↦ I F , p F\mapsto I_{F,p} F ↦ I F , p (16) induce mutually inverse isomorphism between the vector spaces I ( M 3 M 1 M 2 ) I\binom{M^3}{M^1\ M^2} I ( M 1 M 2 M 3 ) and M [ P ( w ) ] M 1 M 2 M 3 \mathcal{M}[P(w)]_{M^1M^2}^{M^3} M [ P ( w ) ] M 1 M 2 M 3 .
Furthermore, let z = l p ( w ) ∈ C z=l_p(w)\in\mathbb{C} z = l p ( w ) ∈ C , so that e z = w e^z=w e z = w . Then there is an isomorphism of vector spaces
M [ P ( w ) ] M 1 M 2 M 3 → ≅ M [ P ( 1 ) ] M 1 M 2 M 3 , F P ( w ) ↦ F P ( 1 ) , \mathcal{M}[P(w)]_{M^1M^2}^{M^3}\xrightarrow{\cong}\mathcal{M}[P(1)]_{M^1M^2}^{M^3},\qquad F^{P(w)}\mapsto F^{P(1)}, M [ P ( w ) ] M 1 M 2 M 3 ≅ M [ P ( 1 ) ] M 1 M 2 M 3 , F P ( w ) ↦ F P ( 1 ) ,
where
F P ( 1 ) ( v 1 ⊗ v 2 ) = e − z L 3 ( 0 ) F P ( w ) ( e z L 1 ( 0 ) v 1 ⊗ e z L 2 ( 0 ) v 2 ) , F^{P(1)}(v^1\otimes v^2)=e^{-zL_3(0)}F^{P(w)}(e^{zL_1(0)}v^1\otimes e^{zL_2(0)}v^2), F P ( 1 ) ( v 1 ⊗ v 2 ) = e − z L 3 ( 0 ) F P ( w ) ( e z L 1 ( 0 ) v 1 ⊗ e z L 2 ( 0 ) v 2 ) ,
where L i ( 0 ) L_i(0) L i ( 0 ) acts on M i M^i M i , for $i=1,2,3.
In the residue formulas below, Res z o u t \operatorname{Res}^{\mathrm{out}}_z Res z out denotes the coefficient of z − 1 z^{-1} z − 1 in the Laurent expansion for ∣ z ∣ > ∣ w ∣ |z|>|w| ∣ z ∣ > ∣ w ∣ . Thus it is the negative of the geometric residue at z = ∞ z=\infty z = ∞ :
Res z o u t F ( z ) d z = − Res z = ∞ F ( z ) d z . \operatorname{Res}^{\mathrm{out}}_z F(z)\,dz=-\operatorname{Res}_{z=\infty}F(z)\,dz. Res z out F ( z ) d z = − Res z = ∞ F ( z ) d z .
We use the usual local residues at the finite marked points. Under the correspondence (15) , we can rewrite the residue form Jacobi identity for the P ( w ) P(w) P ( w ) -intertwining operators (14) as a residue form Jacobi identity for I ( v 1 , e l p ( w ) ) I(v^1,e^{l_p(w)}) I ( v 1 , e l p ( w ) ) , which is useful in our later discussion.
Lemma 2.9. Let I ∈ I ( M 3 M 1 M 2 ) I\in I\binom{M^3}{M^1\ M^2} I ∈ I ( M 1 M 2 M 3 ) . Fix a branch l p ( w ) l_p(w) l p ( w ) for log ( w ) \log(w) log ( w ) (12) , so that w m = e m l p ( w ) w^m=e^{ml_p(w)} w m = e m l p ( w ) for any m ∈ C m\in\mathbb{C} m ∈ C . The formal Jacobi identity (11) has the following equivalent residue form in terms of the corresponding P ( w ) P(w) P ( w ) -intertwining operator I ( ⋅ , e l p ( w ) ) I(\cdot,e^{l_p(w)}) I ( ⋅ , e l p ( w ) ) :
Res z o u t ⟨ ( v 3 ) ′ | Y M 3 ( a , z ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ⋅ ι ∣ z ∣ > ∣ w ∣ ( z n ( z − w ) m ) d z − Res z = 0 ⟨ ( v 3 ) ′ | I ( v 1 , e l p ( w ) ) Y M 2 ( a , z ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z ∣ ( z n ( z − w ) m ) d z = Res z − w = 0 ⟨ ( v 3 ) ′ | I ( Y M 1 ( a , z − w ) v 1 , e l p ( w ) ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z − w ∣ ( z n ( z − w ) m ) d z , (17) \begin{aligned}
&\operatorname{Res}^{\mathrm{out}}_z\left\langle(v^3)'\middle|Y_{M^3}(a,z)I(v^1,e^{l_p(w)})v^2\right\rangle\cdot\iota_{|z|>|w|}\left(\frac{z^n}{(z-w)^m}\right)\,dz\\
&\quad-\operatorname{Res}_{z=0}\left\langle(v^3)'\middle|I(v^1,e^{l_p(w)})Y_{M^2}(a,z)v^2\right\rangle\cdot\iota_{|w|>|z|}\left(\frac{z^n}{(z-w)^m}\right)\,dz\\
&\quad=\operatorname{Res}_{z-w=0}\left\langle(v^3)'\middle|I(Y_{M^1}(a,z-w)v^1,e^{l_p(w)})v^2\right\rangle\cdot\iota_{|w|>|z-w|}\left(\frac{z^n}{(z-w)^m}\right)\,dz, \tag*{(17)}
\end{aligned} Res z out ⟨ ( v 3 ) ′ Y M 3 ( a , z ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ⋅ ι ∣ z ∣ > ∣ w ∣ ( ( z − w ) m z n ) d z − Res z = 0 ⟨ ( v 3 ) ′ I ( v 1 , e l p ( w ) ) Y M 2 ( a , z ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z ∣ ( ( z − w ) m z n ) d z = Res z − w = 0 ⟨ ( v 3 ) ′ I ( Y M 1 ( a , z − w ) v 1 , e l p ( w ) ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z − w ∣ ( ( z − w ) m z n ) d z , (17)
where z n / ( z − w ) m z^n/(z-w)^m z n / ( z − w ) m , with m , n ∈ Z m,n\in\mathbb{Z} m , n ∈ Z , is a standard rational function on P 1 \mathbb{P}^1 P 1 in the affine coordinate z z z , with possible poles at z = 0 , ∞ , w z=0,\infty,w z = 0 , ∞ , w , and a ∈ V a\in V a ∈ V , ( v 3 ) ′ ∈ ( M 3 ) ′ (v^3)'\in(M^3)' ( v 3 ) ′ ∈ ( M 3 ) ′ , v 1 ∈ M 1 v^1\in M^1 v 1 ∈ M 1 , v 2 ∈ M 2 v^2\in M^2 v 2 ∈ M 2 are homogeneous elements.
Following the notation in [[52 , 51 ] ], spanning elements of the chiral Lie algebra L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) can be represented by
X w ( z ) = a ⊗ f ( z ) ( d z ) 1 − wt a = a ⊗ z n ( z − w ) m ( d z ) 1 − wt a , a ∈ V , m , n ∈ Z , (18) X_w(z)=a\otimes f(z)(dz)^{1-\operatorname{wt}a}=a\otimes\frac{z^n}{(z-w)^m}(dz)^{1-\operatorname{wt}a},\qquad a\in V,\ m,n\in\mathbb{Z}, \tag*{(18)} X w ( z ) = a ⊗ f ( z ) ( d z ) 1 − wt a = a ⊗ ( z − w ) m z n ( d z ) 1 − wt a , a ∈ V , m , n ∈ Z , (18)
where z z z is the affine coordinate on P 1 ∖ { ∞ , w , 0 } \mathbb{P}^1\setminus\{\infty,w,0\} P 1 ∖ { ∞ , w , 0 } . The representatives (18) satisfy the following relation:
L ( − 1 ) a ⊗ f ( z ) ( d z ) − wt a = − a ⊗ d f ( z ) d z ( d z ) 1 − wt a , a ∈ V . L(-1)a\otimes f(z)(dz)^{-\operatorname{wt}a}=-a\otimes\frac{df(z)}{dz}(dz)^{1-\operatorname{wt}a},\qquad a\in V. L ( − 1 ) a ⊗ f ( z ) ( d z ) − wt a = − a ⊗ d z df ( z ) ( d z ) 1 − wt a , a ∈ V .
The Lie bracket on L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) is given by
[ a ⊗ f ( z ) ( d z ) 1 − wt a , b ⊗ g ( z ) ( d z ) 1 − wt b ] : = ∑ j ≥ 0 1 j ! a ( j ) b ⊗ d j f ( z ) d z j ⋅ g ( z ) ( d z ) 2 − wt a − wt b + j . (19) \left[a\otimes f(z)(dz)^{1-\operatorname{wt}a},b\otimes g(z)(dz)^{1-\operatorname{wt}b}\right]:=\sum_{j\geq0}\frac{1}{j!}a_{(j)}b\otimes\frac{d^jf(z)}{dz^j}\cdot g(z)(dz)^{2-\operatorname{wt}a-\operatorname{wt}b+j}. \tag*{(19)} [ a ⊗ f ( z ) ( d z ) 1 − wt a , b ⊗ g ( z ) ( d z ) 1 − wt b ] := j ≥ 0 ∑ j ! 1 a ( j ) b ⊗ d z j d j f ( z ) ⋅ g ( z ) ( d z ) 2 − wt a − wt b + j . (19)
We omit the symbol of the differential forms for simplicity of our notation.
Definition 2.10. [52 , 12 , 51 ] Let V V V be a VOA, M 1 M^1 M 1 , M 2 M^2 M 2 , M 3 ∈ Ord ( V ) M^3 \in\operatorname{Ord}(V) M 3 ∈ Ord ( V ) , and let ( M 3 ) ′ (M^3)' ( M 3 ) ′ be the contragredient module of M 3 M^3 M 3 . Then
(1) ( M 3 ) ′ (M^3)' ( M 3 ) ′ is a module over the Lie algebra L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) via ρ ∞ : L P 1 ∖ { ∞ , w , 0 } ( V ) → g l ( ( M 3 ) ′ ) \rho_\infty: \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) \to\mathfrak{gl}((M^3)') ρ ∞ : L P 1 ∖ { ∞ , w , 0 } ( V ) → gl (( M 3 ) ′ ) ,
ρ ∞ ( a ⊗ z n ( z − w ) m ) ( ( v 3 ) ′ ) = Res z Y ( M 3 ) ′ ( ϑ ( a ) , z − 1 ) ( v 3 ) ′ ⋅ ι ∣ z ∣ > ∣ w ∣ ( z n ( z − w ) m ) = − ∑ j ≥ 0 ( − m j ) ( − w ) j a ( n − m − j ) ′ ( v 3 ) ′ , ( v 3 ) ′ ∈ ( M 3 ) ′ , (20) \begin{aligned}
\rho_\infty\left(a \otimes\frac{z^n}{(z-w)^m}\right)((v^3)') &=
\operatorname{Res}_{z}Y_{(M^3)'}(\vartheta(a),z^{-1})(v^3)'\cdot\iota_{\lvert z\rvert>\lvert w\rvert}\left(\frac{z^n}{(z-w)^m}\right) \\
&= -\sum_{j\geq0}\binom{-m}{j}(-w)^j a'_{(n-m-j)}(v^3)',\qquad(v^3)'\in(M^3)', \tag*{(20)}
\end{aligned} ρ ∞ ( a ⊗ ( z − w ) m z n ) (( v 3 ) ′ ) = Res z Y ( M 3 ) ′ ( ϑ ( a ) , z − 1 ) ( v 3 ) ′ ⋅ ι ∣ z ∣ > ∣ w ∣ ( ( z − w ) m z n ) = − j ≥ 0 ∑ ( j − m ) ( − w ) j a ( n − m − j ) ′ ( v 3 ) ′ , ( v 3 ) ′ ∈ ( M 3 ) ′ , (20) where ϑ ( a ) = − e z L ( 1 ) ( − z − 2 ) L ( 0 ) ( a ) \vartheta(a)=-e^{zL(1)}(-z^{-2})^{L(0)}(a) ϑ ( a ) = − e z L ( 1 ) ( − z − 2 ) L ( 0 ) ( a ) and a ( k ) ′ = ∑ i ≥ 0 ( − 1 ) wt a i ! ( L ( 1 ) i a ) ( 2 wt a − k − i − 2 ) a'_{(k)}=\sum_{i\geq0}\frac{(-1)^{\operatorname{wt}a}}{i!}(L(1)^ia)_{(2\operatorname{wt}a-k-i-2)} a ( k ) ′ = ∑ i ≥ 0 i ! ( − 1 ) wt a ( L ( 1 ) i a ) ( 2 wt a − k − i − 2 ) .
(2) M 1 M^1 M 1 is a module over the Lie algebra L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) via ρ w : L P 1 ∖ { ∞ , w , 0 } ( V ) → g l ( M 1 ) \rho_w:\mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V)\to\mathfrak{gl}(M^1) ρ w : L P 1 ∖ { ∞ , w , 0 } ( V ) → gl ( M 1 ) ,
ρ w ( a ⊗ z n ( z − w ) m ) ( v 1 ) = Res z = w Y M 1 ( a , z − w ) v 1 ⋅ ι ∣ w ∣ > ∣ z − w ∣ ( z n ( z − w ) m ) = ∑ j ≥ 0 ( n j ) w n − j a ( j − m ) v 1 , v 1 ∈ M 1 . (21) \begin{aligned}
\rho_w\left(a\otimes\frac{z^n}{(z-w)^m}\right)(v^1)
&=\operatorname{Res}_{z=w}Y_{M^1}(a,z-w)v^1\cdot\iota_{\lvert w\rvert>\lvert z-w\rvert}\left(\frac{z^n}{(z-w)^m}\right)\\
&=\sum_{j\geq0}\binom{n}{j}w^{n-j}a_{(j-m)}v^1,\qquad v^1\in M^1. \tag*{(21)}
\end{aligned} ρ w ( a ⊗ ( z − w ) m z n ) ( v 1 ) = Res z = w Y M 1 ( a , z − w ) v 1 ⋅ ι ∣ w ∣ > ∣ z − w ∣ ( ( z − w ) m z n ) = j ≥ 0 ∑ ( j n ) w n − j a ( j − m ) v 1 , v 1 ∈ M 1 . (21) (3) M 2 M^2 M 2 is a module over the Lie algebra L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) via ρ 0 : L P 1 ∖ { ∞ , w , 0 } ( V ) → g l ( M 2 ) \rho_0:\mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V)\to\mathfrak{gl}(M^2) ρ 0 : L P 1 ∖ { ∞ , w , 0 } ( V ) → gl ( M 2 ) ,
ρ 0 ( a ⊗ z n ( z − w ) m ) ( v 2 ) = Res z = 0 Y M 2 ( a , z ) v 2 ⋅ ι ∣ w ∣ > ∣ z ∣ ( z n ( z − w ) m ) = ∑ j ≥ 0 ( − m j ) ( − w ) − m − j a ( n + j ) v 2 , v 2 ∈ M 2 . (22) \begin{aligned}
\rho_0\left(a\otimes\frac{z^n}{(z-w)^m}\right)(v^2)
&=\operatorname{Res}_{z=0}Y_{M^2}(a,z)v^2\cdot\iota_{\lvert w\rvert>\lvert z\rvert}\left(\frac{z^n}{(z-w)^m}\right)\\
&=\sum_{j\geq0}\binom{-m}{j}(-w)^{-m-j}a_{(n+j)}v^2,\qquad v^2\in M^2. \tag*{(22)}
\end{aligned} ρ 0 ( a ⊗ ( z − w ) m z n ) ( v 2 ) = Res z = 0 Y M 2 ( a , z ) v 2 ⋅ ι ∣ w ∣ > ∣ z ∣ ( ( z − w ) m z n ) = j ≥ 0 ∑ ( j − m ) ( − w ) − m − j a ( n + j ) v 2 , v 2 ∈ M 2 . (22) In particular, ( M 3 ) ′ ⊗ C M 1 ⊗ C M 2 (M^3)'\otimes_{\mathbb{C}}M^1\otimes_{\mathbb{C}}M^2 ( M 3 ) ′ ⊗ C M 1 ⊗ C M 2 is a tensor product module over the chiral Lie algebra L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) . The action of the Lie algebra can be expressed explicitly:
ρ ∞ , w , 0 ( a ⊗ z n ( z − w ) m ) ⋅ ( ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) = − ∑ j ≥ 0 ( − m j ) ( − w ) j a ( n − m − j ) ′ ( ( v 3 ) ′ ) ⊗ v 1 ⊗ v 2 + ∑ j ≥ 0 ( n j ) w n − j ( v 3 ) ′ ⊗ a ( j − m ) ( v 1 ) ⊗ v 2 + ∑ j ≥ 0 ( − m j ) ( − w ) − m − j ( v 3 ) ′ ⊗ v 1 ⊗ a ( n + j ) ( v 2 ) . (23) \begin{aligned}
\rho_{\infty,w,0}\left(a\otimes\frac{z^n}{(z-w)^m}\right)\cdot((v^3)'\otimes v^1\otimes v^2)
={}&-\sum_{j\geq0}\binom{-m}{j}(-w)^j a'_{(n-m-j)}((v^3)')\otimes v^1\otimes v^2\\
&+\sum_{j\geq0}\binom{n}{j}w^{n-j}(v^3)'\otimes a_{(j-m)}(v^1)\otimes v^2\\
&+\sum_{j\geq0}\binom{-m}{j}(-w)^{-m-j}(v^3)'\otimes v^1\otimes a_{(n+j)}(v^2). \tag*{(23)}
\end{aligned} ρ ∞ , w , 0 ( a ⊗ ( z − w ) m z n ) ⋅ (( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) = − j ≥ 0 ∑ ( j − m ) ( − w ) j a ( n − m − j ) ′ (( v 3 ) ′ ) ⊗ v 1 ⊗ v 2 + j ≥ 0 ∑ ( j n ) w n − j ( v 3 ) ′ ⊗ a ( j − m ) ( v 1 ) ⊗ v 2 + j ≥ 0 ∑ ( j − m ) ( − w ) − m − j ( v 3 ) ′ ⊗ v 1 ⊗ a ( n + j ) ( v 2 ) . (23) where ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ∈ ( M 3 ) ′ ⊗ M 1 ⊗ M 2 (v^3)'\otimes v^1\otimes v^2\in(M^3)'\otimes M^1\otimes M^2 ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ∈ ( M 3 ) ′ ⊗ M 1 ⊗ M 2 .
The space of coinvariants associated to ( P 1 , ( ∞ , w , 0 ) , ( ( M 3 ) ′ , M 1 , M 2 ) ) (\mathbb{P}^1,(\infty,w,0),((M^3)',M^1,M^2)) ( P 1 , ( ∞ , w , 0 ) , (( M 3 ) ′ , M 1 , M 2 )) :
[ ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ] ( P 1 , ∞ , w , 0 ) = ( M 3 ) ′ ⊗ M 1 ⊗ M 2 L P 1 ∖ { ∞ , w , 0 } ( V ) ⋅ ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) . [(M^3)'\otimes M^1\otimes M^2]_{(\mathbb{P}^1,\infty,w,0)}
=
\frac{(M^3)'\otimes M^1\otimes M^2}
{\mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V)\cdot((M^3)'\otimes M^1\otimes M^2)}. [( M 3 ) ′ ⊗ M 1 ⊗ M 2 ] ( P 1 , ∞ , w , 0 ) = L P 1 ∖ { ∞ , w , 0 } ( V ) ⋅ (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( M 3 ) ′ ⊗ M 1 ⊗ M 2 . The space of conformal blocks associated to ( P 1 , ( ∞ , w , 0 ) , ( ( M 3 ) ′ , M 1 , M 2 ) ) (\mathbb{P}^1,(\infty,w,0),((M^3)',M^1,M^2)) ( P 1 , ( ∞ , w , 0 ) , (( M 3 ) ′ , M 1 , M 2 )) :
C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) = ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 L P 1 ∖ { ∞ , w , 0 } ( V ) ⋅ ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ) ∗ . \mathcal{C}((M^3)'\otimes M^1\otimes M^2)_{(\mathbb{P}^1,\infty,w,0)}
=
\left(
\frac{(M^3)'\otimes M^1\otimes M^2}
{\mathcal{L}_{\mathbb{P}^1\setminus\{\infty,w,0\}}(V)\cdot((M^3)'\otimes M^1\otimes M^2)}
\right)^*. C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) = ( L P 1 ∖ { ∞ , w , 0 } ( V ) ⋅ (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ∗ . When M 1 M^1 M 1 , M 2 M^2 M 2 , M 3 M^3 M 3 are irreducible ordinary modules, the space of three-point conformal blocks on P 1 \mathbb{P}^1 P 1 can be identified with the intertwining operators I ( M 3 M 1 M 2 ) I\binom{M^3}{M^1\ M^2} I ( M 1 M 2 M 3 ) [56 , 26 ] . We generalize this fact to arbitrary ordinary V V V -modules in Ord ( V ) \operatorname{Ord}(V) Ord ( V ) and to the P ( w ) P(w) P ( w ) -intertwining operators.
Proposition 2.11. Let M 1 M^1 M 1 , M 2 M^2 M 2 , M 3 M^3 M 3 be three ordinary V V V -modules. Fix a branch l p ( w ) l_p(w) l p ( w ) for log ( w ) \log(w) log ( w ) (12) , so that w m = e m l p ( w ) w^m=e^{ml_p(w)} w m = e m l p ( w ) for any m ∈ C m\in\mathbb{C} m ∈ C . Then we have the following isomorphism of vector
spaces:
I ( M 3 M 1 M 2 ) ≅ M [ P ( w ) ] M 1 M 2 M 3 ≅ C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) , (24) I\left(\frac{M^{3}}{M^{1}\ M^{2}}\right)\cong\mathcal{M}[P(w)]_{M^{1}M^{2}}^{M^{3}}\cong\mathcal{C}\left((M^{3})' \otimes M^{1}\otimes M^{2}\right)_{(\mathbb{P}^{1},\infty,w,0)}, \tag*{(24)} I ( M 1 M 2 M 3 ) ≅ M [ P ( w ) ] M 1 M 2 M 3 ≅ C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) , (24)
I ( ⋅ , x ) ↦ F = I ( ⋅ , e l p ( w ) ) ↦ φ I , p , I(\cdot,x)\mapsto F=I(\cdot,e^{l_{p}(w)})\mapsto\varphi_{I,p}, I ( ⋅ , x ) ↦ F = I ( ⋅ , e l p ( w ) ) ↦ φ I , p ,
where
⟨ φ I , p ∣ ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ : = ⟨ ( v 3 ) ′ ∣ F ( v 1 ⊗ v 2 ) ⟩ = ⟨ ( v 3 ) ′ ∣ I ( v 1 , e l p ( w ) ) v 2 ⟩ , (25) \left\langle\varphi_{I,p}\mid(v^{3})'\otimes v^{1}\otimes v^{2}\right\rangle:=\left\langle(v^{3})'\mid F(v^{1}\otimes v^{2})\right\rangle=\left\langle(v^{3})'\mid I(v^{1},e^{l_{p}(w)})v^{2}\right\rangle, \tag*{(25)} ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ := ⟨ ( v 3 ) ′ ∣ F ( v 1 ⊗ v 2 ) ⟩ = ⟨ ( v 3 ) ′ ∣ I ( v 1 , e l p ( w ) ) v 2 ⟩ , (25)
where ( v 3 ) ′ (v^{3})' ( v 3 ) ′ , v 1 v^{1} v 1 , v 2 v^{2} v 2 are homogeneous elements in ( M 3 ) ′ (M^{3})' ( M 3 ) ′ , M 1 M^{1} M 1 , M 2 M^{2} M 2 , respectively.
Proof. By Lemma 2.8, I ( ⋅ , x ) ↦ F = I ( ⋅ , e l p ( w ) ) I(\cdot,x)\mapsto F=I(\cdot,e^{l_{p}(w)}) I ( ⋅ , x ) ↦ F = I ( ⋅ , e l p ( w ) ) is a linear isomorphism. It remains to show that F ↦ φ I , p F\mapsto\varphi_{I,p} F ↦ φ I , p is an isomorphism. Given an element a ⊗ z n ( z − w ) l ∈ L P 1 \ { ∞ , w , 0 } ( V ) a\otimes z^{n}(z-w)^{l}\in\mathcal{L}_{P^{1}\backslash\{\infty,w,0\}}(V) a ⊗ z n ( z − w ) l ∈ L P 1 \ { ∞ , w , 0 } ( V ) , with n , l ∈ Z n,l\in\mathbb{Z} n , l ∈ Z , it follows from (23) that
⟨ φ I , p ∣ ( a ⊗ z n ( z − w ) l ) ⋅ ( ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) ⟩ = − ∑ j ≥ 0 ( l j ) ( − w ) j ⟨ φ I , p ∣ a ( n + l − j ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ + ∑ j ≥ 0 ( n j ) w n − j ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ a ( j + 1 ) v 1 ⊗ v 2 ⟩ + ∑ j ≥ 0 ( l j ) ( − w ) l − j ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ v 1 ⊗ a ( n + j ) v 2 ⟩ = − ∑ j ≥ 0 ( l j ) ( − w ) j ⟨ ( v 3 ) ′ ∣ a ( n + l − j ) F ( v 1 ⊗ v 2 ) ⟩ + ∑ j ≥ 0 ( n j ) w n − j ⟨ ( v 3 ) ′ ∣ F ( a ( j + 1 ) v 1 ⊗ v 2 ) ⟩ + ∑ j ≥ 0 ( l j ) ( − w ) l − j ⟨ ( v 3 ) ′ ∣ F ( v 1 ⊗ a ( n + j ) v 2 ) ⟩ = − Res x 1 ( v 3 ) ′ ∣ Y M 3 ( a , x 1 ) F ( v 1 ⊗ v 2 ) ∣ ∣ x 1 ∣ > ∣ w ∣ ( x 1 n ( x 1 − w ) l ) + Res x 1 − w ⟨ ( v 3 ) ′ ∣ F ( Y M 1 ( a , x 1 − w ) v 1 ⊗ v 2 ) ⟩ ∣ ∣ w ∣ > ∣ x 1 − w ∣ ( x 1 n ( x 1 − w ) l ) + Res x 1 ⟨ ( v 3 ) ′ ∣ F ( v 1 ⊗ Y M 2 ( a , x 1 ) v 2 ) ⟩ ∣ ∣ w ∣ > ∣ x 1 ∣ ( x 1 n ( x 1 − w ) l ) = 0 , (26) \begin{aligned}
\left\langle\varphi_{I,p}\mid\left(a\otimes z^{n}(z-w)^{l}\right)\cdot\left((v^{3})'\otimes v^{1}\otimes v^{2}\right)\right\rangle
={}&-\sum_{j\geq0}\binom{l}{j}(-w)^{j}\left\langle\varphi_{I,p}\mid a_{(n+l-j)}(v^{3})'\otimes v^{1}\otimes v^{2}\right\rangle\\
&+\sum_{j\geq0}\binom{n}{j}w^{n-j}\left\langle\varphi_{I,p}\mid(v^{3})'\otimes a_{(j+1)}v^{1}\otimes v^{2}\right\rangle\\
&+\sum_{j\geq0}\binom{l}{j}(-w)^{l-j}\left\langle\varphi_{I,p}\mid(v^{3})'\otimes v^{1}\otimes a_{(n+j)}v^{2}\right\rangle\\
={}&-\sum_{j\geq0}\binom{l}{j}(-w)^{j}\left\langle(v^{3})'\mid a_{(n+l-j)}F(v^{1}\otimes v^{2})\right\rangle\\
&+\sum_{j\geq0}\binom{n}{j}w^{n-j}\left\langle(v^{3})'\mid F(a_{(j+1)}v^{1}\otimes v^{2})\right\rangle\\
&+\sum_{j\geq0}\binom{l}{j}(-w)^{l-j}\left\langle(v^{3})'\mid F(v^{1}\otimes a_{(n+j)}v^{2})\right\rangle\\
={}&-\operatorname{Res}_{x_{1}}(v^{3})'\mid Y_{M^{3}}(a,x_{1})F(v^{1}\otimes v^{2})\mid_{|x_{1}|>|w|}(x_{1}^{n}(x_{1}-w)^{l})\\
&+\operatorname{Res}_{x_{1}-w}\left\langle(v^{3})'\mid F(Y_{M^{1}}(a,x_{1}-w)v^{1}\otimes v^{2})\right\rangle\mid_{|w|>|x_{1}-w|}(x_{1}^{n}(x_{1}-w)^{l})\\
&+\operatorname{Res}_{x_{1}}\left\langle(v^{3})'\mid F(v^{1}\otimes Y_{M^{2}}(a,x_{1})v^{2})\right\rangle\mid_{|w|>|x_{1}|}(x_{1}^{n}(x_{1}-w)^{l})\\
={}&0, \tag*{(26)}
\end{aligned} ⟨ φ I , p ∣ ( a ⊗ z n ( z − w ) l ) ⋅ ( ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) ⟩ = = = = − j ≥ 0 ∑ ( j l ) ( − w ) j ⟨ φ I , p ∣ a ( n + l − j ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ + j ≥ 0 ∑ ( j n ) w n − j ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ a ( j + 1 ) v 1 ⊗ v 2 ⟩ + j ≥ 0 ∑ ( j l ) ( − w ) l − j ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ v 1 ⊗ a ( n + j ) v 2 ⟩ − j ≥ 0 ∑ ( j l ) ( − w ) j ⟨ ( v 3 ) ′ ∣ a ( n + l − j ) F ( v 1 ⊗ v 2 ) ⟩ + j ≥ 0 ∑ ( j n ) w n − j ⟨ ( v 3 ) ′ ∣ F ( a ( j + 1 ) v 1 ⊗ v 2 ) ⟩ + j ≥ 0 ∑ ( j l ) ( − w ) l − j ⟨ ( v 3 ) ′ ∣ F ( v 1 ⊗ a ( n + j ) v 2 ) ⟩ − Res x 1 ( v 3 ) ′ ∣ Y M 3 ( a , x 1 ) F ( v 1 ⊗ v 2 ) ∣ ∣ x 1 ∣ > ∣ w ∣ ( x 1 n ( x 1 − w ) l ) + Res x 1 − w ⟨ ( v 3 ) ′ ∣ F ( Y M 1 ( a , x 1 − w ) v 1 ⊗ v 2 ) ⟩ ∣ ∣ w ∣ > ∣ x 1 − w ∣ ( x 1 n ( x 1 − w ) l ) + Res x 1 ⟨ ( v 3 ) ′ ∣ F ( v 1 ⊗ Y M 2 ( a , x 1 ) v 2 ) ⟩ ∣ ∣ w ∣ > ∣ x 1 ∣ ( x 1 n ( x 1 − w ) l ) 0 , (26)
where the last equality follows from the fact that F F F satisfies Jacobi identity (14) . Hence φ I , p \varphi_{I,p} φ I , p is a well-defined element in C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) \mathcal{C}((M^{3})'\otimes M^{1}\otimes M^{2})_{(\mathbb{P}^{1},\infty,w,0)} C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) .
Conversely, we define
C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ⟶ M [ P ( w ) ] M 1 M 2 M 3 , φ ↦ F φ , \mathcal{C}((M^{3})'\otimes M^{1}\otimes M^{2})_{(\mathbb{P}^{1},\infty,w,0)}\longrightarrow\mathcal{M}[P(w)]_{M^{1}M^{2}}^{M^{3}},\qquad\varphi\mapsto F_{\varphi}, C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ⟶ M [ P ( w ) ] M 1 M 2 M 3 , φ ↦ F φ ,
⟨ ( v 3 ) ′ ∣ F φ ( v 1 ⊗ v 2 ) ⟩ : = ⟨ φ ∣ ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ , \left\langle(v^{3})'\mid F_{\varphi}(v^{1}\otimes v^{2})\right\rangle:=\left\langle\varphi\mid(v^{3})'\otimes v^{1}\otimes v^{2}\right\rangle, ⟨ ( v 3 ) ′ ∣ F φ ( v 1 ⊗ v 2 ) ⟩ := ⟨ φ ∣ ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ ,
The reverse calculation of (26) shows that F φ F_{\varphi} F φ is a well-defined P ( w ) P(w) P ( w ) -intertwining operator. Clearly, φ ↦ F φ \varphi\mapsto F_{\varphi} φ ↦ F φ is the inverse of the second map of (24) . □
The following observation which immediately follows from Proposition 2.11 and Definition 2.10 is useful in our later discussion. It is convenient for us to use the notation I ( ⋅ , e l p ( w ) ) I(\cdot,e^{l_{p}(w)}) I ( ⋅ , e l p ( w ) ) instead of F F F for the P ( w ) P(w) P ( w ) -intertwining operators.
Corollary 2.12. Let X w ( z ) = a ⊗ z n ( z − w ) m ∈ L P 1 \ { ∞ , w , 0 } ( V ) X_{w}(z)=a\otimes\frac{z^{n}}{(z-w)^{m}}\in\mathcal{L}_{P^{1}\backslash\{\infty,w,0\}}(V) X w ( z ) = a ⊗ ( z − w ) m z n ∈ L P 1 \ { ∞ , w , 0 } ( V ) . Under the identification (24) , we have
⟨ φ I , p ∣ ρ ∞ ( X w ( z ) ) ( ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) ⟩ = − Res z o u t ⟨ ( v 3 ) ′ ∣ Y M 3 ( a , z ) I ( v 1 , e l p ( w ) ) v 2 ⟩ z n ( z − w ) m d z ∣ ∣ z ∣ > ∣ w ∣ , \left\langle\varphi_{I,p}\mid\rho_{\infty}(X_{w}(z))((v^{3})'\otimes v^{1}\otimes v^{2})\right\rangle
=-\operatorname{Res}_{z}^{\mathrm{out}}\left\langle(v^{3})'\mid Y_{M^{3}}(a,z)I(v^{1},e^{l_{p}(w)})v^{2}\right\rangle\left.\frac{z^{n}}{(z-w)^{m}}\,dz\right|_{|z|>|w|}, ⟨ φ I , p ∣ ρ ∞ ( X w ( z )) (( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) ⟩ = − Res z out ⟨ ( v 3 ) ′ ∣ Y M 3 ( a , z ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ( z − w ) m z n d z ∣ z ∣ > ∣ w ∣ , ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ ρ w ( X w ( z ) ) ( v 1 ⊗ v 2 ) ⟩ \left\langle\varphi_{I,p}\mid(v^{3})'\otimes\rho_{w}(X_{w}(z))(v^{1}\otimes v^{2})\right\rangle ⟨ φ I , p ∣ ( v 3 ) ′ ⊗ ρ w ( X w ( z )) ( v 1 ⊗ v 2 ) ⟩
= Res z − w = 0 ⟨ ( V 3 ) ′ | I ( Y M 1 ( a , z − w ) v 1 , e ℓ p ( w ) ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z − w ∣ ( z n ( z − w ) m ) d z , ⟨ φ I , p | ( V 3 ) ′ ⊗ v 1 ⊗ ρ 0 ( X w ( z ) ) ( v 2 ) ⟩ = Res z = 0 ⟨ ( V 3 ) ′ | I ( v 1 , e ℓ p ( w ) ) Y M 2 ( a , z ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z ∣ ( z n ( z − w ) m ) d z . \begin{aligned}
&= \operatorname{Res}_{z-w=0}\left\langle(V^{3})' \middle| I\left(Y_{M^{1}}(a,z-w)v^{1},e^{\ell_{p}(w)}\right)v^{2}\right\rangle\cdot\iota_{|w|>|z-w|}\left(\frac{z^{n}}{(z-w)^{m}}\right)\,dz,\\
&\qquad\left\langle\varphi_{I,p}\middle| (V^{3})'\otimes v^{1}\otimes\rho_{0}(X_{w}(z))(v^{2})\right\rangle\\
&= \operatorname{Res}_{z=0}\left\langle(V^{3})'\middle| I\left(v^{1},e^{\ell_{p}(w)}\right)Y_{M^{2}}(a,z)v^{2}\right\rangle\cdot\iota_{|w|>|z|}\left(\frac{z^{n}}{(z-w)^{m}}\right)\,dz.
\end{aligned} = Res z − w = 0 ⟨ ( V 3 ) ′ I ( Y M 1 ( a , z − w ) v 1 , e ℓ p ( w ) ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z − w ∣ ( ( z − w ) m z n ) d z , ⟨ φ I , p ( V 3 ) ′ ⊗ v 1 ⊗ ρ 0 ( X w ( z )) ( v 2 ) ⟩ = Res z = 0 ⟨ ( V 3 ) ′ I ( v 1 , e ℓ p ( w ) ) Y M 2 ( a , z ) v 2 ⟩ ⋅ ι ∣ w ∣ > ∣ z ∣ ( ( z − w ) m z n ) d z .
In particular, the fact that φ I , p \varphi_{I,p} φ I , p is invariant under the action of X w ( z ) X_{w}(z) X w ( z ) (23) is equivalent to the residue form Jacobi identity (17) of the P ( w ) P(w) P ( w ) -intertwining operator I ( ⋅ , e ℓ p ( w ) ) I(\cdot,e^{\ell_{p}(w)}) I ( ⋅ , e ℓ p ( w ) ) .
The fusion tensor structure on C A ( V ) C_{A(V)} C A ( V )
In this section, we generalize the contracted tensor product M 1 ⊙ M 2 M^{1}\odot M^{2} M 1 ⊙ M 2 in [51 ] to an arbitrary point w ∈ C × w\in\mathbb{C}^{\times} w ∈ C × , and show that if C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^{\mathrm{L}}(\operatorname{Mod}_{\mathrm{fd}}(A(V)))\subseteq C_{1}(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) , then M 1 ⊙ w , 0 M 2 M^{1}\odot_{w,0}M^{2} M 1 ⊙ w , 0 M 2 is isomorphic to the bottom degree of Huang–Lepowsky’s P ( w ) P(w) P ( w ) -tensor product module M 1 ⊠ P ( w ) M 2 M^{1}\boxtimes_{P(w)}M^{2} M 1 ⊠ P ( w ) M 2 , see Theorem 3.14. This isomorphism will be used to define the fusion tensor structure on C A ( V ) = Mod f d ( A ( V ) ) C_{A(V)}=\operatorname{Mod}_{\mathrm{fd}}(A(V)) C A ( V ) = Mod fd ( A ( V )) in Subsection 4.5.
Zhu algebra and the category C A ( V ) C_{A(V)} C A ( V )
We first recall properties of the Zhu algebra A ( V ) A(V) A ( V ) [56 ] and functors between the categories of A ( V ) A(V) A ( V ) -modules and V V V -modules.
The adjoint functors ( Φ L ⊣ Ω ) : Mod ( A ( V ) ) ⇄ Adm ( V ) (\Phi^{\mathrm{L}}\dashv\Omega):\operatorname{Mod}(A(V))\rightleftarrows\operatorname{Adm}(V) ( Φ L ⊣ Ω ) : Mod ( A ( V )) ⇄ Adm ( V )
Let A ( V ) A(V) A ( V ) be the Zhu algebra associated to the VOA V V V [56 ] , and let Mod ( A ( V ) ) \operatorname{Mod}(A(V)) Mod ( A ( V )) denote the category of left A ( V ) A(V) A ( V ) -modules. The bottom-degree M ( 0 ) M(0) M ( 0 ) of an admissible V V V -module M M M , or more generally, the space of highest-weight vectors Ω ( M ) = { v ∈ M : a ( m ) v = 0 , wt a − m − 1 < 0 } \Omega(M)=\{v\in M:a_{(m)}v=0,\ \operatorname{wt}a-m-1<0\} Ω ( M ) = { v ∈ M : a ( m ) v = 0 , wt a − m − 1 < 0 } of a weak V V V -module M M M , is a left A ( V ) A(V) A ( V ) -module via
A ( V ) ⟶ End ( Ω ( M ) ) , [ a ] ⟼ o ( a ) = a ( wt a − 1 ) , a ∈ V . A(V)\longrightarrow\operatorname{End}(\Omega(M)),\qquad[a]\longmapsto o(a)=a_{(\operatorname{wt}a-1)},\qquad a\in V. A ( V ) ⟶ End ( Ω ( M )) , [ a ] ⟼ o ( a ) = a ( wt a − 1 ) , a ∈ V .
We can view Ω \Omega Ω as a functor Ω : Adm ( V ) ⟶ Mod ( A ( V ) ) \Omega:\operatorname{Adm}(V)\longrightarrow\operatorname{Mod}(A(V)) Ω : Adm ( V ) ⟶ Mod ( A ( V )) .
For a left A ( V ) A(V) A ( V ) -module S S S , the linear dual space S ∗ = Hom C ( S , C ) S^{*}=\operatorname{Hom}_{\mathbb{C}}(S,\mathbb{C}) S ∗ = Hom C ( S , C ) is a natural right A ( V ) A(V) A ( V ) -module. We give it a left A ( V ) A(V) A ( V ) -module via the contragredient action A ( V ) × S ∗ → S ∗ A(V)\times S^{*}\to S^{*} A ( V ) × S ∗ → S ∗ , ( [ a ] , f ) ↦ [ a ] . f ([a],f)\mapsto[a].f ([ a ] , f ) ↦ [ a ] . f , where
⟨ [ a ] . f , s ⟩ = ⟨ f , θ A ( V ) ( [ a ] ) s ⟩ , θ A ( V ) ( [ a ] ) = [ e L ( 1 ) ( − 1 ) L ( 0 ) a ] . (27) \langle[a].f,s\rangle=\langle f,\theta_{A(V)}([a])s\rangle,\qquad\theta_{A(V)}([a])=[e^{L(1)}(-1)^{L(0)}a]. \tag*{(27)} ⟨[ a ] . f , s ⟩ = ⟨ f , θ A ( V ) ([ a ]) s ⟩ , θ A ( V ) ([ a ]) = [ e L ( 1 ) ( − 1 ) L ( 0 ) a ] . (27)
for f ∈ S ∗ f\in S^{*} f ∈ S ∗ , s ∈ S s\in S s ∈ S . Now let M ∈ Ord ( V ) M\in\operatorname{Ord}(V) M ∈ Ord ( V ) , and M ′ M' M ′ be its contragredient module via (6) . Then Ω ( M ′ ) \Omega(M') Ω ( M ′ ) is also a natural right A ( V ) A(V) A ( V ) -module:
⟨ f . [ a ] , v ⟩ = ⟨ f , o ( a ) v ⟩ , f ∈ Ω ( M ′ ) , [ a ] ∈ A ( V ) , v ∈ M . (28) \langle f.[a],v\rangle=\langle f,o(a)v\rangle,\qquad f\in\Omega(M'),\ [a]\in A(V),\ v\in M. \tag*{(28)} ⟨ f . [ a ] , v ⟩ = ⟨ f , o ( a ) v ⟩ , f ∈ Ω ( M ′ ) , [ a ] ∈ A ( V ) , v ∈ M . (28)
which can be viewed as a left A ( V ) A(V) A ( V ) -module via (27) . Note that this left A ( V ) A(V) A ( V ) -module structure agrees with the left A ( V ) A(V) A ( V ) -module structure inherited from the contragredient module M ′ M' M ′ .
Let U = U ( V ) \mathcal{U}=\mathcal{U}(V) U = U ( V ) be the universal enveloping algebra associated to the VOA V V V [24 , 21 ] , which is a complete Z \mathbb{Z} Z -graded topological associative algebra with neighborhood bases
N U n = U ⋅ U ≤ − n , n ≥ 0. \mathcal{N}^{n}_{\mathcal{U}}=\mathcal{U}\cdot\mathcal{U}_{\leq-n},\qquad n\geq0. N U n = U ⋅ U ≤ − n , n ≥ 0.
The generalized Verma module [15 , 13 GL26] associated to a left A ( V ) A(V) A ( V ) -module S S S is an admissible V V V -module defined by
Φ L ( S ) = ( U / N U 1 ) ⊗ U 0 S = ⨁ n = 0 ∞ Φ L ( S ) ( n ) . (29) \Phi^{\mathrm{L}}(S)=\left(\mathcal{U}/\mathcal{N}^{1}_{\mathcal{U}}\right)\otimes_{\mathcal{U}_{0}}S=\bigoplus_{n=0}^{\infty}\Phi^{\mathrm{L}}(S)(n). \tag*{(29)} Φ L ( S ) = ( U / N U 1 ) ⊗ U 0 S = n = 0 ⨁ ∞ Φ L ( S ) ( n ) . (29)
Here S S S is a U 0 \mathcal{U}_{0} U 0 -module via U 0 ↠ A ( V ) ≅ U 0 / N U 0 1 \mathcal{U}_{0}\twoheadrightarrow A(V)\cong\mathcal{U}_{0}/\mathcal{N}^{1}_{\mathcal{U}_{0}} U 0 ↠ A ( V ) ≅ U 0 / N U 0 1 , and
Φ L ( S ) ( n ) = span { a [ n 1 ] 1 ⋯ a [ n r ] r v : a i ∈ V , v ∈ S , ∑ i = 1 r ( wt a i − n i − 1 ) = n } . \Phi^{\mathrm{L}}(S)(n)=\operatorname{span}\left\{a^{1}_{[n_{1}]}\cdots a^{r}_{[n_{r}]}v:a^{i}\in V,\ v\in S,\ \sum_{i=1}^{r}(\operatorname{wt}a^{i}-n_{i}-1)=n\right\}. Φ L ( S ) ( n ) = span { a [ n 1 ] 1 ⋯ a [ n r ] r v : a i ∈ V , v ∈ S , i = 1 ∑ r ( wt a i − n i − 1 ) = n } .
Φ L \Phi^{L} Φ L is also a functor Φ L : Mod ( A ( V ) ) → Adm ( V ) \Phi^{L}: \operatorname{Mod}(A(V)) \to\operatorname{Adm}(V) Φ L : Mod ( A ( V )) → Adm ( V ) . Moreover, ( Φ L ⊣ Ω ) : Mod ( A ( V ) ) ⇄ Adm ( V ) (\Phi^{L} \dashv\Omega): \operatorname{Mod}(A(V)) \rightleftarrows\operatorname{Adm}(V) ( Φ L ⊣ Ω ) : Mod ( A ( V )) ⇄ Adm ( V ) is an adjoint pair of functors. Recall the following fact in [15 , 13 GL26] :
Lemma 3.1. If the VOA V V V satisfies the strong identity condition, then the pair of functors
( Φ L ⊣ Ω ) : Mod ( A ( V ) ) ⇄ Adm ( V ) (\Phi^{L} \dashv\Omega): \operatorname{Mod}(A(V)) \rightleftarrows\operatorname{Adm}(V) ( Φ L ⊣ Ω ) : Mod ( A ( V )) ⇄ Adm ( V ) is an adjoint equivalence. In particular, it is an adjoint equivalence if V V V is rational or if V V V is the Heisenberg VOA M ( 1 ) M(1) M ( 1 ) .
The subcategory C A ( V ) C_{A(V)} C A ( V ) of A ( V ) A(V) A ( V ) -modules.
Definition 3.2. Note that C 1 ( V ) ⊂ Adm ( V ) C_{1}(V) \subset\operatorname{Adm}(V) C 1 ( V ) ⊂ Adm ( V ) is a full subcategory, see Definition 2.1. Then we have a restricted functor
Ω : C 1 ( V ) → Mod ( A ( V ) ) . (30) \Omega: C_{1}(V) \to\operatorname{Mod}(A(V)). \tag*{(30)} Ω : C 1 ( V ) → Mod ( A ( V )) . (30)
We denote its essential image by C A ( V ) C_{A(V)} C A ( V ) , which is the full subcategory of Mod ( A ( V ) ) \operatorname{Mod}(A(V)) Mod ( A ( V )) consisting of modules isomorphic to Ω ( M ) \Omega(M) Ω ( M ) for some M ∈ C 1 ( V ) M \in C_{1}(V) M ∈ C 1 ( V ) .
Whenever the functor (30) is fully faithful, it induces an equivalence of categories Ω : C 1 ( V ) ⟶ ~ C A ( V ) \Omega: C_{1}(V) \widetilde{\longrightarrow} C_{A(V)} Ω : C 1 ( V ) ⟶ C A ( V ) . We give some sufficient conditions for the fully faithfulness of Ω \Omega Ω (30) .
Proposition 3.3. The functor Ω \Omega Ω (30) is fully faithful if either one of the following condition holds:
V V V satisfies the strong identity condition. In this case, the quasi-inverse of Ω \Omega Ω is the generalized Verma module functor Φ L \Phi^{L} Φ L (29) , and so
( Φ L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) (\Phi^{L} \dashv\Omega): C_{A(V)} \rightleftarrows C_{1}(V) ( Φ L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) are equivalence of categories.
C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple. In this case, the quasi-inverse of Ω \Omega Ω is Dong-Li-Mason’s quotient module functor L L L [15 Section 7] , and
( L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) (L \dashv\Omega): C_{A(V)} \rightleftarrows C_{1}(V) ( L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) are equivalence of categories. Moreover, in this case, if Φ L ( Ω ( M ) ) ∈ C 1 ( V ) \Phi^{L}(\Omega(M)) \in C_{1}(V) Φ L ( Ω ( M )) ∈ C 1 ( V ) for any object M M M in C 1 ( V ) C_{1}(V) C 1 ( V ) , then there is a natural isomorphism Φ L ∣ C A ( V ) ≅ L ∣ C A ( V ) \Phi^{L}|_{C_{A(V)}} \cong L|_{C_{A(V)}} Φ L ∣ C A ( V ) ≅ L ∣ C A ( V ) . In particular, ( Φ L ⊣ Ω ) (\Phi^{L} \dashv\Omega) ( Φ L ⊣ Ω ) also restricts to an adjoint equivalence.
In particular, when V V V is strongly rational, then both conditions hold, and Φ L ∣ C A ( V ) ≅ L ∣ C A ( V ) \Phi^{L}|_{C_{A(V)}} \cong L|_{C_{A(V)}} Φ L ∣ C A ( V ) ≅ L ∣ C A ( V ) .
Proof. If V V V satisfies the strong identity condition, then ( Φ L ⊣ Ω ) : Mod ( A ( V ) ) ⇄ Adm ( V ) (\Phi^{L} \dashv\Omega): \operatorname{Mod}(A(V)) \rightleftarrows\operatorname{Adm}(V) ( Φ L ⊣ Ω ) : Mod ( A ( V )) ⇄ Adm ( V ) is an adjoint equivalence by Lemma 3.1 . Hence the restriction Ω ∣ C 1 ( V ) \Omega|_{C_{1}(V)} Ω ∣ C 1 ( V ) is fully faithful, and the restriction of functors give rise to an adjoint equivalence ( Φ L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) (\Phi^{L} \dashv\Omega): C_{A(V)} \rightleftarrows C_{1}(V) ( Φ L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) .
Now assume C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple. By Lemma 2.4 any object M M M in C 1 ( V ) C_{1}(V) C 1 ( V ) is an ordinary V V V -module. Then by [15 Theorem 7.2] , for any irreducible V V V -module W W W in C 1 ( V ) C_{1}(V) C 1 ( V ) , one has L ( Ω ( W ) ) ≅ W L(\Omega(W)) \cong W L ( Ω ( W )) ≅ W . It follows from the semisimplicity of C 1 ( V ) C_{1}(V) C 1 ( V ) that L ( Ω ( M ) ) ≅ M L(\Omega(M)) \cong M L ( Ω ( M )) ≅ M for any object M M M in C 1 ( V ) C_{1}(V) C 1 ( V ) . i.e., ( L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) (L \dashv\Omega): C_{A(V)} \rightleftarrows C_{1}(V) ( L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) are adjoint equivalence.
Finally, if Φ L ( Ω ( M ) ) ∈ C 1 ( V ) \Phi^{L}(\Omega(M)) \in C_{1}(V) Φ L ( Ω ( M )) ∈ C 1 ( V ) for all the objects M M M in C 1 ( V ) C_{1}(V) C 1 ( V ) , then both Φ L ∣ C A ( V ) \Phi^{L}|_{C_{A(V)}} Φ L ∣ C A ( V ) and L ∣ C A ( V ) L|_{C_{A(V)}} L ∣ C A ( V ) are therefore left adjoints to Ω ∣ C 1 ( V ) \Omega|_{C_{1}(V)} Ω ∣ C 1 ( V ) . By uniqueness of left adjoints, there is a natural isomorphism Φ L ∣ C A ( V ) ≅ L ∣ C A ( V ) \Phi^{L}|_{C_{A(V)}} \cong L|_{C_{A(V)}} Φ L ∣ C A ( V ) ≅ L ∣ C A ( V ) .
The left A ( V ) A(V) A ( V ) -module M 1 ⊙ w , 0 M 2 M^{1} \odot_{w,0} M^{2} M 1 ⊙ w , 0 M 2 from one-point restricted conformal blocks
We recall the definition of (one-point) restricted conformal blocks on smooth genus-zero curves [25 , 26 ] , then mildly generalize the construction left A ( V ) A(V) A ( V ) -module M 1 ⊙ M 2 M^{1} \odot M^{2} M 1 ⊙ M 2 by the second-named author in [51 ] . In this subsection, we fix a point w ∈ P 1 ∖ { ∞ , 0 } = C × w \in\mathbb{P}^{1} \setminus\{\infty, 0\} = \mathbb{C}^{\times} w ∈ P 1 ∖ { ∞ , 0 } = C × .
Definition 3.4. [26 , 51 ] Consider the following subspace of the chiral Lie algebra L P 1 ∖ { ∞ , w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^{1}\setminus\{\infty,w,0\}}(V) L P 1 ∖ { ∞ , w , 0 } ( V ) :
L P 1 ∖ { w , 0 } ( V ) ≺ 0 = span { a ⊗ f ( z ) ∈ L P 1 ∖ { ∞ , w , 0 } ( V ) : a ∈ V , ord ∞ ( f ( z ) ) ≻ 1 − wt a } \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{\prec0} = \operatorname{span}\{a \otimes f(z) \in\mathcal{L}_{\mathbb{P}^{1}\setminus\{\infty,w,0\}}(V) : a \in V,\ \operatorname{ord}_{\infty}(f(z)) \succ1-\operatorname{wt}a\} L P 1 ∖ { w , 0 } ( V ) ≺ 0 = span { a ⊗ f ( z ) ∈ L P 1 ∖ { ∞ , w , 0 } ( V ) : a ∈ V , ord ∞ ( f ( z )) ≻ 1 − wt a } = span { a ⊗ z n ( z − w ) m ∈ L P 1 ∖ { ∞ , w , 0 } ( V ) : a ∈ V , n − m ≺ wt a − 1 } , where ≺ represents ≤ or < . (31) = \operatorname{span}\left\{a \otimes\frac{z^n}{(z-w)^m} \in\mathcal{L}_{\mathbb{P}^{1}\setminus\{\infty,w,0\}}(V) : a \in V,\ n-m \prec\operatorname{wt}a-1\right\}, \qquad\text{where } \prec\text{ represents } \leq\text{ or } <. \tag*{(31)} = span { a ⊗ ( z − w ) m z n ∈ L P 1 ∖ { ∞ , w , 0 } ( V ) : a ∈ V , n − m ≺ wt a − 1 } , where ≺ represents ≤ or < . (31) L P 1 ∖ { w , 0 } ( V ) ≤ 0 \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{\leq0} L P 1 ∖ { w , 0 } ( V ) ≤ 0 is a Lie subalgebra of L P 1 ∖ { w , 0 } ( V ) \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V) L P 1 ∖ { w , 0 } ( V ) , called the the ∞ \infty ∞ -restricted chiral Lie algebra. L P 1 ∖ { w , 0 } ( V ) < 0 \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{<0} L P 1 ∖ { w , 0 } ( V ) < 0 is an ideal of the Lie algebra L P 1 ∖ { w , 0 } ( V ) ≤ 0 \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{\leq0} L P 1 ∖ { w , 0 } ( V ) ≤ 0 .
The quotient Lie algebra L P 1 ∖ { w , 0 } ( V ) ≤ 0 / L P 1 ∖ { w , 0 } ( V ) < 0 \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{\leq0}/\mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{<0} L P 1 ∖ { w , 0 } ( V ) ≤ 0 / L P 1 ∖ { w , 0 } ( V ) < 0 is isomorphic to the degree-zero part of Borcherds’ Lie algebra L ( V ) 0 = span { a [ wt a − 1 ] : a ∈ V } L(V)_0 = \operatorname{span}\{a_{[\operatorname{wt}a-1]} : a \in V\} L ( V ) 0 = span { a [ wt a − 1 ] : a ∈ V } .
Definition 3.5. [26 , 51 ] Let M 1 , M 2 , M 3 ∈ Ord ( V ) M^1, M^2, M^3 \in\operatorname{Ord}(V) M 1 , M 2 , M 3 ∈ Ord ( V ) . Then Ω ( ( M 3 ) ′ ) \Omega((M^3)') Ω (( M 3 ) ′ ) is a left A ( V ) A(V) A ( V ) -module via θ A ( V ) \theta_{A(V)} θ A ( V ) (27) , which is also module over the Lie algebra L P 1 ∖ { w , 0 } ( V ) ≤ 0 \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{\leq0} L P 1 ∖ { w , 0 } ( V ) ≤ 0 via ρ ∞ \rho_{\infty} ρ ∞ (20) . The space of ∞ \infty ∞ -restricted conformal blocks associated to the data ( P 1 , ( ∞ , w , 0 ) , ( Ω ( ( M 3 ) ′ ) , M 1 , M 2 ) ) (\mathbb{P}^{1},(\infty,w,0),(\Omega((M^3)'),M^1,M^2)) ( P 1 , ( ∞ , w , 0 ) , ( Ω (( M 3 ) ′ ) , M 1 , M 2 )) is defined by
C r e s ( ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 ) ) ) = ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 L P 1 ∖ { w , 0 } ( V ) ≤ 0 ⋅ ( Ω ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ) ∗ \mathcal{C}_{\mathrm{res}}\left(\left(\Omega((M^3)')\otimes M^1\otimes M^2\right)_{(\mathbb{P}^{1},(\infty,w,0))}\right)
=
\left(
\frac{\Omega((M^3)')\otimes M^1\otimes M^2}
{\mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{\leq0}\cdot\left(\Omega((M^3)'\otimes M^1\otimes M^2\right)}
\right)^{*} C res ( ( Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 )) ) = ( L P 1 ∖ { w , 0 } ( V ) ≤ 0 ⋅ ( Ω (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ∗ ≅ Hom L ( V ) 0 ( M 1 ⊗ M 2 L P 1 ∖ { w , 0 } ( V ) < 0 ⋅ ( M 1 ⊗ M 2 ) , Ω ( ( M 3 ) ′ ) ∗ ) . (32) \cong\operatorname{Hom}_{L(V)_0}\left(
\frac{M^1\otimes M^2}{\mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{<0}\cdot(M^1\otimes M^2)},
\Omega((M^3)')^{*}
\right). \tag*{(32)} ≅ Hom L ( V ) 0 ( L P 1 ∖ { w , 0 } ( V ) < 0 ⋅ ( M 1 ⊗ M 2 ) M 1 ⊗ M 2 , Ω (( M 3 ) ′ ) ∗ ) . (32) The action of L P 1 ∖ { w , 0 } ( V ) < 0 \mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{<0} L P 1 ∖ { w , 0 } ( V ) < 0 on M 1 ⊗ M 2 M^1\otimes M^2 M 1 ⊗ M 2 is given by the tensor action of ρ w \rho_w ρ w and ρ 0 \rho_0 ρ 0 as below:
ρ w , 0 : L P 1 ∖ { w , 0 } ( V ) < 0 ⟶ g l ( M 1 ⊗ M 2 ) , \rho_{w,0}:\mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{<0}\longrightarrow\mathfrak{gl}(M^1\otimes M^2), ρ w , 0 : L P 1 ∖ { w , 0 } ( V ) < 0 ⟶ gl ( M 1 ⊗ M 2 ) , ρ w , 0 ( a ⊗ z n ( z − w ) m ) ( v 1 ⊗ v 2 ) = ρ w ( a ⊗ z n ( z − w ) m ) v 1 ⊗ v 2 + v 1 ⊗ ρ 0 ( a ⊗ z n ( z − w ) m ) v 2 , (33) \rho_{w,0}\left(a\otimes\frac{z^n}{(z-w)^m}\right)(v^1\otimes v^2)
=
\rho_w\left(a\otimes\frac{z^n}{(z-w)^m}\right)v^1\otimes v^2
+
v^1\otimes\rho_0\left(a\otimes\frac{z^n}{(z-w)^m}\right)v^2, \tag*{(33)} ρ w , 0 ( a ⊗ ( z − w ) m z n ) ( v 1 ⊗ v 2 ) = ρ w ( a ⊗ ( z − w ) m z n ) v 1 ⊗ v 2 + v 1 ⊗ ρ 0 ( a ⊗ ( z − w ) m z n ) v 2 , (33) where ρ w \rho_w ρ w and ρ 0 \rho_0 ρ 0 are given by (21) and (22) , respectively.
Restricted conformal blocks extend uniquely to (unrestricted) conformal blocks after replacing the restricted A ( V ) A(V) A ( V ) -module by its associated generalized Verma module; see [51 Theorem 5.5] and [26 Theorems 5.18 and 5.19] .
Lemma 3.6. Let M 1 M^1 M 1 , M 2 M^2 M 2 , M 3 M^3 M 3 be ordinary V V V -modules, and let S = Ω ( ( M 3 ) ′ ) S=\Omega((M^3)') S = Ω (( M 3 ) ′ ) , equipped with the natural left A ( V ) A(V) A ( V ) -module structure inherited from the contragredient module ( M 3 ) ′ (M^3)' ( M 3 ) ′ (27) . Then the restriction of conformal blocks induces an isomorphism
C ( Φ L ( S ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 ) ) ≅ C r e s ( S ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 ) ) , φ ↦ φ ∣ S ⊗ M 1 ⊗ M 2 . \mathcal{C}\left(\Phi^{L}(S)\otimes M^1\otimes M^2\right)_{(\mathbb{P}^{1},(\infty,w,0))}
\cong
\mathcal{C}_{\mathrm{res}}\left(S\otimes M^1\otimes M^2\right)_{(\mathbb{P}^{1},(\infty,w,0))},
\qquad\varphi\mapsto\varphi|_{S\otimes M^1\otimes M^2}. C ( Φ L ( S ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 )) ≅ C res ( S ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 )) , φ ↦ φ ∣ S ⊗ M 1 ⊗ M 2 . In particular, if the counit ε ( M 3 ) ′ : Φ L ( Ω ( ( M 3 ) ′ ) ) → ( M 3 ) ′ \varepsilon_{(M^3)'}:\Phi^{L}(\Omega((M^3)'))\to(M^3)' ε ( M 3 ) ′ : Φ L ( Ω (( M 3 ) ′ )) → ( M 3 ) ′ is an isomorphism, then the restriction also induces an isomorphism
C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 ) ) ≅ C r e s ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 ) ) , φ ↦ φ ∣ Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 . (34) \mathcal{C}\left((M^3)'\otimes M^1\otimes M^2\right)_{(\mathbb{P}^{1},(\infty,w,0))}
\cong
\mathcal{C}_{\mathrm{res}}\left(\Omega((M^3)')\otimes M^1\otimes M^2\right)_{(\mathbb{P}^{1},(\infty,w,0))},
\qquad\varphi\mapsto\varphi|_{\Omega((M^3)')\otimes M^1\otimes M^2}. \tag*{(34)} C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 )) ≅ C res ( Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ( ∞ , w , 0 )) , φ ↦ φ ∣ Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 . (34)
Definition 3.7. With the notations as above, define
M 1 ⊙ w , 0 M 2 : = M 1 ⊗ M 2 L P 1 ∖ { w , 0 } ( V ) < 0 ⋅ ( M 1 ⊗ M 2 ) = span { v 1 ⊙ w v 2 : v 1 ∈ M 1 , v 2 ∈ M 2 } , M^1\mathbin{\odot_{w,0}}M^2:=
\frac{M^1\otimes M^2}{\mathcal{L}_{\mathbb{P}^{1}\setminus\{w,0\}}(V)_{<0}\cdot(M^1\otimes M^2)}
=
\operatorname{span}\{v^1\mathbin{\odot_w}v^2:v^1\in M^1,\ v^2\in M^2\}, M 1 ⊙ w , 0 M 2 := L P 1 ∖ { w , 0 } ( V ) < 0 ⋅ ( M 1 ⊗ M 2 ) M 1 ⊗ M 2 = span { v 1 ⊙ w v 2 : v 1 ∈ M 1 , v 2 ∈ M 2 } , where v 1 ⊙ w v 2 v^1\mathbin{\odot_w}v^2 v 1 ⊙ w v 2 denotes v 1 ⊗ v 2 v^1\otimes v^2 v 1 ⊗ v 2 .
M 1 ⊙ w , 0 M 2 M^1\mathbin{\odot_{w,0}}M^2 M 1 ⊙ w , 0 M 2 is spanned by symbols v 1 ⊙ w v 2 v^1\mathbin{\odot_w}v^2 v 1 ⊙ w v 2 , subject to the following relations:
ρ w ( a ⊗ z wt a − s ( z − w ) t ) v 1 ⊙ w v 2 = − v 1 ⊙ w ρ 0 ( a ⊗ z wt a − s ( z − w ) t ) v 2 , equivalently, \rho_w\left(a\otimes\frac{z^{\operatorname{wt}a-s}}{(z-w)^t}\right)v^1\mathbin{\odot_w}v^2
=
-v^1\mathbin{\odot_w}\rho_0\left(a\otimes\frac{z^{\operatorname{wt}a-s}}{(z-w)^t}\right)v^2,
\qquad\text{equivalently,} ρ w ( a ⊗ ( z − w ) t z wt a − s ) v 1 ⊙ w v 2 = − v 1 ⊙ w ρ 0 ( a ⊗ ( z − w ) t z wt a − s ) v 2 , equivalently, ∑ j ≥ 0 ( wt a − s j ) w wt a − s − j a ( j − t ) v 1 ⊙ w v 2 = − v 1 ⊙ w ∑ j ≥ 0 ( − t j ) ( − w ) − t − j a ( wt a − s + j ) v 2 . (35) \sum_{j\geq0}\binom{\operatorname{wt}a-s}{j}w^{\operatorname{wt}a-s-j}a_{(j-t)}v^1\mathbin{\odot_w}v^2
=
-v^1\mathbin{\odot_w}\sum_{j\geq0}\binom{-t}{j}(-w)^{-t-j}a_{(\operatorname{wt}a-s+j)}v^2. \tag*{(35)} j ≥ 0 ∑ ( j wt a − s ) w wt a − s − j a ( j − t ) v 1 ⊙ w v 2 = − v 1 ⊙ w j ≥ 0 ∑ ( j − t ) ( − w ) − t − j a ( wt a − s + j ) v 2 . (35)
where s , t ∈ Z s,t \in\mathbb{Z} s , t ∈ Z such that s + t > 1 s+t>1 s + t > 1 . Moreover, M 1 ⊙ w , 0 M 2 M^{1}\odot_{w,0}M^{2} M 1 ⊙ w , 0 M 2 is a L ( V ) 0 L(V)_{0} L ( V ) 0 -module with respect to the action
a [ w wt a − 1 ] . ( v 1 ⊙ w v 2 ) = ∑ j ≥ 0 ( w wt a − 1 j ) w wt a − 1 − j a ( j ) v 1 ⊙ w v 2 + v 1 ⊙ w o ( a ) v 2 , (36) a_{[w\operatorname{wt}a-1]}.(v^{1}\odot_{w}v^{2})=\sum_{j\geq0}\binom{w\operatorname{wt}a-1}{j}w^{\operatorname{wt}a-1-j}a_{(j)}v^{1}\odot_{w}v^{2}+v^{1}\odot_{w}o(a)v^{2}, \tag*{(36)} a [ w wt a − 1 ] . ( v 1 ⊙ w v 2 ) = j ≥ 0 ∑ ( j w wt a − 1 ) w wt a − 1 − j a ( j ) v 1 ⊙ w v 2 + v 1 ⊙ w o ( a ) v 2 , (36)
where o ( a ) = a ( w wt a − 1 ) o(a)=a_{(w\operatorname{wt}a-1)} o ( a ) = a ( w wt a − 1 ) . When w = 1 w=1 w = 1 , M 1 ⊙ 1 , 0 M 2 M^{1}\odot_{1,0}M^{2} M 1 ⊙ 1 , 0 M 2 is the contracted tensor product M 1 ⊙ M 2 M^{1}\odot M^{2} M 1 ⊙ M 2 in [51 Definition 4.3] .
The following proposition follows from [51 Theorem 4.8] by the change of coordinate z = w x z=wx z = w x . The proof is similar, we omit the details.
Proposition 3.8. Assume either M 1 M^{1} M 1 or M 2 M^{2} M 2 is generated by its bottom degrees Ω ( M 1 ) \Omega(M^{1}) Ω ( M 1 ) or Ω ( M 2 ) \Omega(M^{2}) Ω ( M 2 ) , respectively. Then the L ( V ) 0 L(V)_{0} L ( V ) 0 -action (36) factors through L ( V ) 0 ↠ A ( V ) L(V)_{0}\twoheadrightarrow A(V) L ( V ) 0 ↠ A ( V ) , a [ w wt a − 1 ] ↦ [ a ] a_{[w\operatorname{wt}a-1]}\mapsto[a] a [ w wt a − 1 ] ↦ [ a ] , which makes M 1 ⊙ w , 0 M 2 M^{1}\odot_{w,0}M^{2} M 1 ⊙ w , 0 M 2 a left A ( V ) A(V) A ( V ) -module with respect to the following action:
A ( V ) × ( M 1 ⊙ w , 0 M 2 ) ⟶ M 1 ⊙ w , 0 M 2 , A(V)\times(M^{1}\odot_{w,0}M^{2})\longrightarrow M^{1}\odot_{w,0}M^{2}, A ( V ) × ( M 1 ⊙ w , 0 M 2 ) ⟶ M 1 ⊙ w , 0 M 2 , [ a ] . ( v 1 ⊙ w v 2 ) = ∑ j ≥ 0 ( w wt a − 1 j ) w wt a − 1 − j a ( j ) v 1 ⊙ w v 2 + v 1 ⊙ w o ( a ) v 2 [a].(v^{1}\odot_{w}v^{2})=\sum_{j\geq0}\binom{w\operatorname{wt}a-1}{j}w^{\operatorname{wt}a-1-j}a_{(j)}v^{1}\odot_{w}v^{2}+v^{1}\odot_{w}o(a)v^{2} [ a ] . ( v 1 ⊙ w v 2 ) = j ≥ 0 ∑ ( j w wt a − 1 ) w wt a − 1 − j a ( j ) v 1 ⊙ w v 2 + v 1 ⊙ w o ( a ) v 2 = ρ w , 0 ( a ⊗ z w wt a − 1 ) . ( v 1 ⊙ w v 2 ) . (37) =\rho_{w,0}\left(a\otimes z^{w\operatorname{wt}a-1}\right).(v^{1}\odot_{w}v^{2}). \tag*{(37)} = ρ w , 0 ( a ⊗ z w wt a − 1 ) . ( v 1 ⊙ w v 2 ) . (37) Furthermore, it follows from [51 Remark 4.7] that
[ a ] . ( v 1 ⊙ w v 2 ) = ρ w , 0 ( a ⊗ z w wt a − 1 + q ( z − w ) q ) . ( v 1 ⊙ w v 2 ) , q ∈ Z . (38) [a].(v^{1}\odot_{w}v^{2})=\rho_{w,0}\left(a\otimes\frac{z^{w\operatorname{wt}a-1+q}}{(z-w)^{q}}\right).(v^{1}\odot_{w}v^{2}),\qquad q\in\mathbb{Z}. \tag*{(38)} [ a ] . ( v 1 ⊙ w v 2 ) = ρ w , 0 ( a ⊗ ( z − w ) q z w wt a − 1 + q ) . ( v 1 ⊙ w v 2 ) , q ∈ Z . (38)
With the definition of M 1 ⊙ w , 0 M 2 M^{1}\odot_{w,0}M^{2} M 1 ⊙ w , 0 M 2 given by Definition 3.7, it follows from (32) that
C r e s ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( ( M 3 ) ′ ) ∗ ) . \mathcal{C}_{\mathrm{res}}\left(\Omega((M^{3})')\otimes M^{1}\otimes M^{2}\right)_{(\mathsf{P}_{1,\infty,w,0})}\cong\operatorname{Hom}_{A(V)}\left(M^{1}\odot_{w,0}M^{2},\Omega((M^{3})')^{*}\right). C res ( Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω (( M 3 ) ′ ) ∗ ) .
Recall that Ω ( ( M 3 ) ′ ) \Omega((M^{3})') Ω (( M 3 ) ′ ) is a natural right A ( V ) A(V) A ( V ) -module via (28) . Hence Ω ( ( M 3 ) ′ ) ∗ \Omega((M^{3})')^{*} Ω (( M 3 ) ′ ) ∗ is a natural left A ( V ) A(V) A ( V ) -module via A ( V ) × Ω ( ( M 3 ) ′ ) ∗ → Ω ( ( M 3 ) ′ ) ∗ A(V)\times\Omega((M^{3})')^{*}\to\Omega((M^{3})')^{*} A ( V ) × Ω (( M 3 ) ′ ) ∗ → Ω (( M 3 ) ′ ) ∗ , ( [ a ] , u ) ↦ [ a ] . u ([a],u)\mapsto[a].u ([ a ] , u ) ↦ [ a ] . u , where
⟨ [ a ] . u , f ⟩ : = ⟨ u , f . [ a ] ⟩ , u ∈ Ω ( ( M 3 ) ′ ) ∗ , f ∈ Ω ( ( M 3 ) ′ ) , [ a ] ∈ A ( V ) . \langle[a].u,f\rangle:=\langle u,f.[a]\rangle,\qquad u\in\Omega((M^{3})')^{*},\ f\in\Omega((M^{3})'),\ [a]\in A(V). ⟨[ a ] . u , f ⟩ := ⟨ u , f . [ a ]⟩ , u ∈ Ω (( M 3 ) ′ ) ∗ , f ∈ Ω (( M 3 ) ′ ) , [ a ] ∈ A ( V ) .
Since we do not assume V V V is strongly rational, it is a subtle question whether Ω ( ( M 3 ) ′ ) ∗ \Omega((M^{3})')^{*} Ω (( M 3 ) ′ ) ∗ is isomorphic to Ω ( M 3 ) \Omega(M^{3}) Ω ( M 3 ) as a left A ( V ) A(V) A ( V ) -module. Fortunately, the sufficient conditions in Proposition 3.3 are also sufficient here.
Lemma 3.9. Let M ∈ C 1 ( V ) M\in C_{1}(V) M ∈ C 1 ( V ) be an ordinary module. Then the natural left A ( V ) A(V) A ( V ) -module Ω ( M ′ ) ∗ \Omega(M')^{*} Ω ( M ′ ) ∗ is isomorphic to the bottom-degree left A ( V ) A(V) A ( V ) -module Ω ( M ) \Omega(M) Ω ( M ) if one of the following conditions hold:
C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple.
V V V satisfies the strong identity condition.
In particular, when V V V is strongly rational, then both condition are satisfied.
Proof. Assume C 1 ( V ) C_{1}(V) C 1 ( V ) is semisimple, then M = ⨁ i W i M=\bigoplus_{i}W^{i} M = ⨁ i W i with W i W^{i} W i being irreducible C 1 C_{1} C 1 -cofinite ordinary modules. Since Ω ( M ) = ⨁ i Ω ( W i ) \Omega(M)=\bigoplus_{i}\Omega(W^{i}) Ω ( M ) = ⨁ i Ω ( W i ) , it suffices to show Ω ( W ) ≅ Ω ( W ′ ) ∗ \Omega(W)\cong\Omega(W')^{*} Ω ( W ) ≅ Ω ( W ′ ) ∗ for irreducible ordinary module W ∈ C 1 ( V ) W\in C_{1}(V) W ∈ C 1 ( V ) . By Definition 2.1, W W W is admissible and Ω ( W ) = W ( 0 ) \Omega(W)=W(0) Ω ( W ) = W ( 0 ) , which is finite dimensional. Then W ′ = ⨁ n = 0 ∞ W ( n ) ∗ W'=\bigoplus_{n=0}^{\infty}W(n)^{*} W ′ = ⨁ n = 0 ∞ W ( n ) ∗ is also irreducible [22 ] , with Ω ( W ′ ) = W ( 0 ) ∗ \Omega(W')=W(0)^{*} Ω ( W ′ ) = W ( 0 ) ∗ . Hence there is a natural isomorphism Ω ( W ′ ) ∗ = ( W ( 0 ) ∗ ) ∗ ≅ W ( 0 ) = Ω ( W ) \Omega(W')^{*}=(W(0)^{*})^{*}\cong W(0)=\Omega(W) Ω ( W ′ ) ∗ = ( W ( 0 ) ∗ ) ∗ ≅ W ( 0 ) = Ω ( W ) .
Now assume V V V satisfies the strong identity condition. Then by Lemma 3.1 and Proposition 3.3, there is an adjoint equivalence ( Φ L ⊣ Ω ) : Mod ( A ( V ) ) ⇄ Adm ( V ) (\Phi^{L}\dashv\Omega):\operatorname{Mod}(A(V))\rightleftarrows\operatorname{Adm}(V) ( Φ L ⊣ Ω ) : Mod ( A ( V )) ⇄ Adm ( V ) . For M ∈ C 1 ( V ) ⊂ Adm ( V ) M\in C_{1}(V)\subset\operatorname{Adm}(V) M ∈ C 1 ( V ) ⊂ Adm ( V ) , since M ≅ Φ L ( Ω ( M ) ) M\cong\Phi^{L}(\Omega(M)) M ≅ Φ L ( Ω ( M )) , M ′ ≅ Φ L ( Ω ( M ′ ) ) M'\cong\Phi^{L}(\Omega(M')) M ′ ≅ Φ L ( Ω ( M ′ )) , and M ′ M' M ′ is also ordinary, the natural pairing ⟨ , ⟩ : M × M ′ → C \langle\ ,\ \rangle:M\times M'\to\mathbb{C} ⟨ , ⟩ : M × M ′ → C restricts to a non-degenerate pairing ⟨ , ⟩ : Ω ( M ) × Ω ( M ′ ) → C \langle\ ,\ \rangle:\Omega(M)\times\Omega(M')\to\mathbb{C} ⟨ , ⟩ : Ω ( M ) × Ω ( M ′ ) → C . On the other hand, since
⟨ f , a ( − 1 ) v ⟩ = ⟨ θ ( a ( − 1 ) ) f , v ⟩ = 0 , a ∈ V + , v ∈ M , f ∈ Ω ( M ′ ) , (39) \langle f,a_{(-1)}v\rangle=\langle\theta(a_{(-1)})f,v\rangle=0,\qquad a\in V_{+},\ v\in M,\ f\in\Omega(M'), \tag*{(39)} ⟨ f , a ( − 1 ) v ⟩ = ⟨ θ ( a ( − 1 ) ) f , v ⟩ = 0 , a ∈ V + , v ∈ M , f ∈ Ω ( M ′ ) , (39)
we have Ω ( M ′ ) ⊆ ( M / C 1 ( M ) ) ∗ \Omega(M') \subseteq(M/C_1(M))^* Ω ( M ′ ) ⊆ ( M / C 1 ( M ) ) ∗ , which is finite dimensional. Hence Ω ( M ) ≅ Ω ( M ′ ) ∗ \Omega(M) \cong\Omega(M')^* Ω ( M ) ≅ Ω ( M ′ ) ∗ via the perfect pairing.
Corollary 3.10. Assume C 1 ( V ) C_1(V) C 1 ( V ) is semisimple or V V V satisfies the strong identity condition, then for any ordinary modules M 1 M^1 M 1 , M 2 M^2 M 2 , M 3 ∈ C 1 ( V ) M^3 \in C_1(V) M 3 ∈ C 1 ( V ) , we have
C r e s ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) . (40) \mathcal{C}_{\mathrm{res}}\left(\Omega((M^3)') \otimes M^1 \otimes M^2\right)_{(\mathbb{P}^1,\infty,w,0)} \cong\operatorname{Hom}_{A(V)}\left(M^1 \mathbin{\odot_{w,0}} M^2,\Omega(M^3)\right). \tag*{(40)} C res ( Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) . (40)
The corresponding elements φ r e s \varphi_{\mathrm{res}} φ res and f f f satisfy
⟨ ( u 3 ) ′ ∣ f ( v 1 ⊙ w v 2 ) ⟩ = ⟨ φ r e s ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ , ( u 3 ) ′ ∈ Ω ( ( M 3 ) ′ ) . (41) \left\langle(u^3)' \mid f(v^1 \mathbin{\odot_w} v^2)\right\rangle= \left\langle\varphi_{\mathrm{res}} \mid(u^3)' \otimes v^1 \otimes v^2\right\rangle,\qquad(u^3)' \in\Omega((M^3)'). \tag*{(41)} ⟨ ( u 3 ) ′ ∣ f ( v 1 ⊙ w v 2 ) ⟩ = ⟨ φ res ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ , ( u 3 ) ′ ∈ Ω (( M 3 ) ′ ) . (41)
The isomorphism η w : M 1 ⊙ w , 0 M 2 → Ω ( M 1 ⊠ P ( w ) M 2 ) \eta_w: M^1 \mathbin{\odot_{w,0}} M^2 \to\Omega(M^1 \boxtimes_{P(w)} M^2) η w : M 1 ⊙ w , 0 M 2 → Ω ( M 1 ⊠ P ( w ) M 2 )
In this subsection, we fix a point w ∈ P 1 \ { ∞ , 0 } = C × w \in\mathbb{P}^1\backslash\{\infty,0\}=\mathbb{C}^{\times} w ∈ P 1 \ { ∞ , 0 } = C × . We show that under certain conditions for the category C 1 ( V ) C_1(V) C 1 ( V ) , the ⊙ \odot ⊙ -tensor product in Definition 3.7 captures all the information of the bottom degree Zhu algebra module of Huang–Lepowsky’s P ( w ) P(w) P ( w ) -tensor product.
Key assumptions for C 1 ( V ) C_1(V) C 1 ( V )
In order to describe the top degree Zhu algebra module of M 1 ⊠ P ( w ) M 2 M^1 \boxtimes_{P(w)} M^2 M 1 ⊠ P ( w ) M 2 , we need some further conditions on C 1 ( V ) C_1(V) C 1 ( V ) .
Proposition 3.11. Assume C 1 ( V ) C_1(V) C 1 ( V ) is semisimple, and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^{\mathrm{L}}(\operatorname{Mod}_{\mathrm{fd}}(A(V))) \subseteq C_1(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) , where Φ L \Phi^{\mathrm{L}} Φ L is the generalized Verma module functor (29) . Then
(1) C A ( V ) = Mod f d ( A ( V ) ) C_{A(V)}=\operatorname{Mod}_{\mathrm{fd}}(A(V)) C A ( V ) = Mod fd ( A ( V )) , and
( Φ L ⊣ Ω ) : Mod f d ( A ( V ) ) ⇄ C 1 ( V ) (\Phi^{\mathrm{L} } \dashv\Omega):\operatorname{Mod}_{\mathrm{fd}}(A(V)) \rightleftarrows C_1(V) ( Φ L ⊣ Ω ) : Mod fd ( A ( V )) ⇄ C 1 ( V ) are equivalence of categories;
(2) C 1 ( V ) ⊆ Ord ( V ) C_1(V) \subseteq\operatorname{Ord}(V) C 1 ( V ) ⊆ Ord ( V ) is closed under contragredient: M ′ ∈ C 1 ( V ) M' \in C_1(V) M ′ ∈ C 1 ( V ) when M ∈ C 1 ( V ) M \in C_1(V) M ∈ C 1 ( V ) ;
(3) For M 1 , M 2 ∈ C 1 ( V ) M^1,M^2 \in C_1(V) M 1 , M 2 ∈ C 1 ( V ) , we have M 1 ⊙ w , 0 M 2 ∈ C A ( V ) M^1 \mathbin{\odot_{w,0}} M^2 \in C_{A(V)} M 1 ⊙ w , 0 M 2 ∈ C A ( V ) .
Proof. (1). Since C 1 ( V ) C_1(V) C 1 ( V ) is semisimple, by Proposition 3.3 and Lemma 2.4, Ω : C 1 ( V ) → Mod ( A ( V ) ) \Omega:C_1(V)\to\operatorname{Mod}(A(V)) Ω : C 1 ( V ) → Mod ( A ( V )) is fully faithful and any object in C 1 ( V ) C_1(V) C 1 ( V ) is an ordinary V V V -module. Let M ∈ C 1 ( V ) M\in C_1(V) M ∈ C 1 ( V ) . Then M = ⨁ i W i M=\bigoplus_i W^i M = ⨁ i W i , with W i W^i W i being irreducible for all i i i and M / C 1 ( M ) ≅ ⨁ i W i / C 1 ( W ) M/C_1(M)\cong\bigoplus_i W^i/C_1(W) M / C 1 ( M ) ≅ ⨁ i W i / C 1 ( W ) is finite dimensional. Therefore, the sum is finite and Ω ( M ) = ⨁ i = 0 n Ω ( W i ) ∈ Mod f d ( A ( V ) ) \Omega(M)=\bigoplus_{i=0}^{n}\Omega(W^i)\in\operatorname{Mod}_{\mathrm{fd}}(A(V)) Ω ( M ) = ⨁ i = 0 n Ω ( W i ) ∈ Mod fd ( A ( V )) . Since Φ L ( S ) ∈ C 1 ( V ) \Phi^{\mathrm{L}}(S)\in C_1(V) Φ L ( S ) ∈ C 1 ( V ) for any S ∈ Mod f d ( A ( V ) ) S\in\operatorname{Mod}_{\mathrm{fd}}(A(V)) S ∈ Mod fd ( A ( V )) by the assumption, and we have equivalence of categories
( Φ L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) , C A ( V ) ⊆ Mod f d ( A ( V ) ) . (\Phi^{\mathrm{L}}\dashv\Omega):C_{A(V)}\rightleftarrows C_1(V),\qquad C_{A(V)}\subseteq\operatorname{Mod}_{\mathrm{fd}}(A(V)). ( Φ L ⊣ Ω ) : C A ( V ) ⇄ C 1 ( V ) , C A ( V ) ⊆ Mod fd ( A ( V )) .
It follows that S ≅ Ω ( Φ L ( S ) ) ∈ C A ( V ) S\cong\Omega(\Phi^{\mathrm{L}}(S))\in C_{A(V)} S ≅ Ω ( Φ L ( S )) ∈ C A ( V ) , and so C A ( V ) = Mod f d ( A ( V ) ) C_{A(V)}=\operatorname{Mod}_{\mathrm{fd}}(A(V)) C A ( V ) = Mod fd ( A ( V )) .
(2). Given M ∈ C 1 ( V ) M\in C_1(V) M ∈ C 1 ( V ) , it follows from (39) that Ω ( M ′ ) ⊆ ( M / C 1 ( M ) ) ∗ \Omega(M')\subseteq(M/C_1(M))^* Ω ( M ′ ) ⊆ ( M / C 1 ( M ) ) ∗ is finite dimensional. Then Ω ( M ′ ) ∈ Mod f d ( A ( V ) ) = C A ( V ) \Omega(M')\in\operatorname{Mod}_{\mathrm{fd}}(A(V))=C_{A(V)} Ω ( M ′ ) ∈ Mod fd ( A ( V )) = C A ( V ) and M ′ ≅ Φ L ( Ω ( M ′ ) ) ∈ C 1 ( V ) M'\cong\Phi^{\mathrm{L}}(\Omega(M'))\in C_1(V) M ′ ≅ Φ L ( Ω ( M ′ )) ∈ C 1 ( V ) .
(3). Let M 1 , M 2 ∈ C 1 ( V ) M^1,M^2\in C_1(V) M 1 , M 2 ∈ C 1 ( V ) be irreducible ordinary V V V -modules. Choose homogeneous subspaces U 1 , U 2 U^1,U^2 U 1 , U 2 such that M i = U i ⊕ C 1 ( M i ) M^i=U^i\oplus C_1(M^i) M i = U i ⊕ C 1 ( M i ) and M i ( 0 ) ⊆ U i M^i(0)\subseteq U^i M i ( 0 ) ⊆ U i , i = 1 , 2 i=1,2 i = 1 , 2 . In order to show (3), we claim that the canonical map
U 1 ⊗ U 2 ⟶ M 1 ⊙ w , 0 M 2 , u 1 ⊗ u 2 ⟼ u 1 ⊙ w u 2 (42) U^1\otimes U^2\longrightarrow M^1\mathbin{\odot_{w,0}}M^2,\qquad u^1\otimes u^2\longmapsto u^1\mathbin{\odot_w}u^2 \tag*{(42)} U 1 ⊗ U 2 ⟶ M 1 ⊙ w , 0 M 2 , u 1 ⊗ u 2 ⟼ u 1 ⊙ w u 2 (42)
is surjective. In other words, M 1 ⊙ w , 0 M 2 = U 1 ⊙ w , 0 U 2 = span { u 1 ⊙ w u 2 : u i ∈ U i , i = 1 , 2 } M^1\mathbin{\odot_{w,0}}M^2=U^1\mathbin{\odot_{w,0}}U^2=\operatorname{span}\{u^1\mathbin{\odot_w}u^2:u^i\in U^i,\ i=1,2\} M 1 ⊙ w , 0 M 2 = U 1 ⊙ w , 0 U 2 = span { u 1 ⊙ w u 2 : u i ∈ U i , i = 1 , 2 } .
Let u ∈ M 1 ( p ) u\in M^1(p) u ∈ M 1 ( p ) , v ∈ M 2 ( q ) v\in M^2(q) v ∈ M 2 ( q ) , and a ∈ V d a\in V_d a ∈ V d , with p , q ≥ 0 p,q\geq0 p , q ≥ 0 and d ≥ 1 d\geq1 d ≥ 1 . For homogeneous elements u i ∈ M i u^i\in M^i u i ∈ M i , we define deg ( u 1 ⊙ w u 2 ) : = deg ( u 1 ) + deg ( u 2 ) \deg(u^1\mathbin{\odot_w}u^2):=\deg(u^1)+\deg(u^2) deg ( u 1 ⊙ w u 2 ) := deg ( u 1 ) + deg ( u 2 ) . It follows from (35) , with ( s , t ) = ( 1 , 1 ) (s,t)=(1,1) ( s , t ) = ( 1 , 1 ) that
( a ( − 1 ) u ) ⊙ w v = − ∑ j = 1 d − 1 ( d − 1 j ) w − j ( a ( j − 1 ) u ) ⊙ w v + ∑ j ≥ 0 w − d − j u ⊙ w a ( d − 1 + j ) v . (a_{(-1)}u)\mathbin{\odot_w}v=-\sum_{j=1}^{d-1}\binom{d-1}{j}w^{-j}(a_{(j-1)}u)\mathbin{\odot_w}v+\sum_{j\geq0}w^{-d-j}u\mathbin{\odot_w}a_{(d-1+j)}v. ( a ( − 1 ) u ) ⊙ w v = − j = 1 ∑ d − 1 ( j d − 1 ) w − j ( a ( j − 1 ) u ) ⊙ w v + j ≥ 0 ∑ w − d − j u ⊙ w a ( d − 1 + j ) v .
Note that deg ( ( a ( − 1 ) u ) ⊙ w v ) = d + p + q \deg((a_{(-1)}u)\mathbin{\odot_w}v)=d+p+q deg (( a ( − 1 ) u ) ⊙ w v ) = d + p + q , and deg ( ( a ( j − 1 ) u ) ⊙ w v ) = d − j + p + q < d + p + q \deg((a_{(j-1)}u)\mathbin{\odot_w}v)=d-j+p+q<d+p+q deg (( a ( j − 1 ) u ) ⊙ w v ) = d − j + p + q < d + p + q for all j ≥ 1 j\geq1 j ≥ 1 , and deg ( u ⊙ w a ( d − 1 + j ) v ) = − j + p + q < d + p + q \deg(u\mathbin{\odot_w}a_{(d-1+j)}v)=-j+p+q<d+p+q deg ( u ⊙ w a ( d − 1 + j ) v ) = − j + p + q < d + p + q for all j ≥ 0 j\geq0 j ≥ 0 . Then by induction on the degree deg ( ( a ( − 1 ) u ) ⊙ w v ) \deg((a_{(-1)}u)\mathbin{\odot_w}v) deg (( a ( − 1 ) u ) ⊙ w v ) , we have C 1 ( M 1 ) ⊙ w , 0 M 2 ⊆ U 1 ⊙ w , 0 U 2 C_1(M^1)\mathbin{\odot_{w,0}}M^2\subseteq U^1\mathbin{\odot_{w,0}}U^2 C 1 ( M 1 ) ⊙ w , 0 M 2 ⊆ U 1 ⊙ w , 0 U 2 . On the other hand, apply (35) with ( s , t ) = ( d + 1 , 0 ) (s,t)=(d+1,0) ( s , t ) = ( d + 1 , 0 ) , we have
u ⊠ w ( a ( − 1 ) v ) = − ∑ j ≥ 0 ( − 1 ) j w − 1 − j ( a ( j ) u ) ⊠ w v , u\boxtimes_w\left(a_{(-1)}v\right)=-\sum_{j\geq0}(-1)^j w^{-1-j}\left(a_{(j)}u\right)\boxtimes_w v, u ⊠ w ( a ( − 1 ) v ) = − j ≥ 0 ∑ ( − 1 ) j w − 1 − j ( a ( j ) u ) ⊠ w v ,
where deg ( ( a ( j ) u ) ⊠ w v ) = d − j − 1 + p + q < d + p + q = deg ( u ⊠ w ( a ( − 1 ) v ) ) \operatorname{deg}\left((a_{(j)}u)\boxtimes_w v\right)=d-j-1+p+q<d+p+q=\operatorname{deg}\left(u\boxtimes_w(a_{(-1)}v)\right) deg ( ( a ( j ) u ) ⊠ w v ) = d − j − 1 + p + q < d + p + q = deg ( u ⊠ w ( a ( − 1 ) v ) ) , for all j ≥ 0 j\geq0 j ≥ 0 . Then by induction on degree again, we have M 1 ⊠ w , 0 C 1 ( M 2 ) ⊆ U 1 ⊠ w , 0 U 2 M^1\boxtimes_{w,0}C_1(M^2)\subseteq U^1\boxtimes_{w,0}U^2 M 1 ⊠ w , 0 C 1 ( M 2 ) ⊆ U 1 ⊠ w , 0 U 2 . This proves the surjectivity of (42) . Since M 1 , M 2 M^1,M^2 M 1 , M 2 are C 1 C_1 C 1 -cofinite, we may choose U 1 , U 2 U^1,U^2 U 1 , U 2 to be finite dimensional. Then dim ( M 1 ⊠ w , 0 M 2 ) ≤ dim U 1 ⋅ dim U 2 < ∞ \dim(M^1\boxtimes_{w,0}M^2)\leq\dim U^1\cdot\dim U^2<\infty dim ( M 1 ⊠ w , 0 M 2 ) ≤ dim U 1 ⋅ dim U 2 < ∞ . Hence M 1 ⊠ w , 0 M 2 ∈ Mod f d ( A ( V ) ) = C A ( V ) M^1\boxtimes_{w,0}M^2\in\operatorname{Mod}_{\mathrm{fd}}(A(V))=C_{A(V)} M 1 ⊠ w , 0 M 2 ∈ Mod fd ( A ( V )) = C A ( V ) . The general case follows since ⊠ w , 0 \boxtimes_{w,0} ⊠ w , 0 commutes with finite direct sums.
Example 3.12. The following VOAs satisfy the conditions in Proposition 3.11, namely, C 1 ( V ) C_1(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^L(\operatorname{Mod}_{\mathrm{fd}}(A(V)))\subseteq C_1(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) .
Any strongly rational VOA V V V , in which case C 1 ( V ) = Ord ( V ) C_1(V)=\operatorname{Ord}(V) C 1 ( V ) = Ord ( V ) . In particular, V = L k ( g ) V=L_k(\mathfrak{g}) V = L k ( g ) an affine VOA at positive integral level k ∈ Z > 0 k\in\mathbb{Z}_{>0} k ∈ Z > 0 [24 , 14 ] .
V = L k ( g ) V=L_k(\mathfrak{g}) V = L k ( g ) an admissible level affine VOA, where
k + h ∨ = p q , p ≥ { h ∨ if ( r ∨ , q ) = 1 , h if ( r ∨ , q ) = r ∨ . k+h^\vee=\frac{p}{q},\qquad p\geq
\begin{cases}
h^\vee& \text{if }(r^\vee,q)=1,\\
h & \text{if }(r^\vee,q)=r^\vee.
\end{cases} k + h ∨ = q p , p ≥ { h ∨ h if ( r ∨ , q ) = 1 , if ( r ∨ , q ) = r ∨ .
See subsection 2.2.2.
V = V k ( g ) = L k ( g ) V=V^k(\mathfrak{g})=L_k(\mathfrak{g}) V = V k ( g ) = L k ( g ) a generic level affine VOA, where k + h ∨ ∉ Q k+h^\vee\notin\mathbb{Q} k + h ∨ ∈ / Q . See subsection 2.2.3.
V = V k ( s l 2 ) = L k ( s l 2 ) V=V^k(\mathfrak{sl}_2)=L_k(\mathfrak{sl}_2) V = V k ( sl 2 ) = L k ( sl 2 ) , where
k = − 2 ± 1 p , p ∈ Z > 0 ; k=-2\pm\frac{1}{p},\qquad p\in\mathbb{Z}_{>0}; k = − 2 ± p 1 , p ∈ Z > 0 ;
see [5 Proposition 3.1] .
V = L k ( g ) V=L_k(\mathfrak{g}) V = L k ( g ) , where ( g , k ) (\mathfrak{g},k) ( g , k ) is one of the following pairs:
Table 5.
See [10 Theorems 4.2.1 and 4.2.3] .
Remark 3.13. Even if C 1 ( V ) C_1(V) C 1 ( V ) is semisimple, we do not have Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^L(\operatorname{Mod}_{\mathrm{fd}}(A(V)))\subseteq C_1(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) in general. For example, let V = L ( c , 0 ) V=L(c,0) V = L ( c , 0 ) be the Virasoro VOA with generic central charge in the subsection 2.2.4. Then A ( V ) ≅ C [ x ] A(V)\cong\mathbb{C}[x] A ( V ) ≅ C [ x ] [55 , 18 ] . For S = C v c , 0 ∈ Mod f d ( A ( V ) ) S=\mathbb{C}v_{c,0}\in\operatorname{Mod}_{\mathrm{fd}}(A(V)) S = C v c , 0 ∈ Mod fd ( A ( V )) , we have Φ L ( S ) ≅ M ( c , 0 ) \Phi^L(S)\cong M(c,0) Φ L ( S ) ≅ M ( c , 0 ) , the Verma module over the Virasoro algebra. It is not C 1 C_1 C 1 -cofinite since L ( − 1 ) m v c , 0 = ω ( 0 ) v c , 0 ≠ 0 L(-1)^m v_{c,0}=\omega^{(0)}v_{c,0}\neq0 L ( − 1 ) m v c , 0 = ω ( 0 ) v c , 0 = 0 in M ( c , 0 ) M(c,0) M ( c , 0 ) for all m ≥ 1 m\geq1 m ≥ 1 .
Universal property of ⊠ P ( w ) \boxtimes_{P(w)} ⊠ P ( w )
We recall the definition of P ( w ) P(w) P ( w ) -tensor product of ordinary modules by Huang-Lepowsky [33 , 34 ] . Let M 1 , M 2 ∈ C 1 ( V ) M^1,M^2\in C_1(V) M 1 , M 2 ∈ C 1 ( V ) . A P ( w ) P(w) P ( w ) -tensor product is a pair ( M 1 ⊠ P ( w ) M 2 , ⊠ P ( w ) ) (M^1\boxtimes_{P(w)}M^2,\boxtimes_{P(w)}) ( M 1 ⊠ P ( w ) M 2 , ⊠ P ( w ) ) , consisting of a V V V -module M 1 ⊠ P ( w ) M 2 ∈ C 1 ( V ) M^1\boxtimes_{P(w)}M^2\in C_1(V) M 1 ⊠ P ( w ) M 2 ∈ C 1 ( V ) together with a P ( w ) P(w) P ( w ) -intertwining map, see Definition 2.7,
⊠ P ( w ) : M 1 ⊗ C M 2 ⟶ M 1 ⊠ P ( w ) M 2 , v 1 ⊗ v 2 ⟼ v 1 ⊠ P ( w ) v 2 , \boxtimes_{P(w)}:M^1\otimes_{\mathbb{C}}M^2\longrightarrow M^1\boxtimes_{P(w)}M^2,\qquad v^1\otimes v^2\longmapsto v^1\boxtimes_{P(w)}v^2, ⊠ P ( w ) : M 1 ⊗ C M 2 ⟶ M 1 ⊠ P ( w ) M 2 , v 1 ⊗ v 2 ⟼ v 1 ⊠ P ( w ) v 2 ,
satisfying the following universal property: for any M 3 ∈ C 1 ( V ) M^3\in C_1(V) M 3 ∈ C 1 ( V ) , and any P ( w ) P(w) P ( w ) -intertwining map F ∈ M [ P ( w ) ] M 1 M 2 M 3 F\in\mathcal{M}[P(w)]^{M^3}_{M^1M^2} F ∈ M [ P ( w ) ] M 1 M 2 M 3 , there exists a unique homomorphism of V V V -modules g : M 1 ⊠ P ( w ) M 2 ⟶ M 3 g:M^1\boxtimes_{P(w)}M^2\longrightarrow M^3 g : M 1 ⊠ P ( w ) M 2 ⟶ M 3 such that g ‾ ∘ ⊠ P ( w ) = F \overline{g}\circ\boxtimes_{P(w)}=F g ∘ ⊠ P ( w ) = F .
It was proved by Huang [36 Theorem 1.2] that the category C 1 ( V ) C_1(V) C 1 ( V ) admits a natural vertex tensor category structure. In particular, M 1 ⊠ P ( w ) M 2 ∈ C 1 ( V ) M^1\boxtimes_{P(w)}M^2\in C_1(V) M 1 ⊠ P ( w ) M 2 ∈ C 1 ( V ) for M 1 , M 2 ∈ C 1 ( V ) M^1,M^2\in C_1(V) M 1 , M 2 ∈ C 1 ( V ) . Then the universal property of the P ( w ) P(w) P ( w ) -tensor product states that, for every M 3 ∈ C 1 ( V ) M^3\in C_1(V) M 3 ∈ C 1 ( V ) , there is a natural isomorphism
M [ P ( w ) ] M 1 M 2 M 3 → ≅ Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) , F ↦ g , (43) \mathcal{M}[P(w)]^{M^3}_{M^1M^2}\xrightarrow{\ \cong\ }\operatorname{Hom}_{C_1(V)}(M^1\boxtimes_{P(w)}M^2,M^3),\qquad F\mapsto g, \tag*{(43)} M [ P ( w ) ] M 1 M 2 M 3 ≅ Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) , F ↦ g , (43)
where g g g is the unique V V V -module homomorphism satisfying g ‾ ∘ ⊠ P ( w ) = F \overline{g} \circ\boxtimes_{P(w)} = F g ∘ ⊠ P ( w ) = F .
Moreover, ⊠ P ( w ) \boxtimes_{P(w)} ⊠ P ( w ) is a bifunctor on C 1 ( V ) C_1(V) C 1 ( V ) . Given homomorphisms g i ∈ Hom V ( M i , N i ) g_i \in\operatorname{Hom}_V(M^i,N^i) g i ∈ Hom V ( M i , N i ) , i = 1 , 2 i=1,2 i = 1 , 2 , then g 1 ⊠ P ( w ) g 2 g_1 \boxtimes_{P(w)} g_2 g 1 ⊠ P ( w ) g 2 is the unique homomorphism in Hom V ( M 1 ⊠ P ( w ) M 2 , N 1 ⊠ P ( w ) N 2 ) \operatorname{Hom}_V(M^1 \boxtimes_{P(w)} M^2,N^1 \boxtimes_{P(w)} N^2) Hom V ( M 1 ⊠ P ( w ) M 2 , N 1 ⊠ P ( w ) N 2 ) satisfying
( g 1 ⊠ P ( w ) g 2 ) ( v 1 ⊠ P ( w ) v 2 ) = g 1 v 1 ⊠ P ( w ) g 2 v 2 , v 1 ∈ M 1 , v 2 ∈ M 2 . (g_1 \boxtimes_{P(w)} g_2)(v^1 \boxtimes_{P(w)} v^2)=g_1v^1 \boxtimes_{P(w)} g_2v^2,\qquad v^1\in M^1,\ v^2\in M^2. ( g 1 ⊠ P ( w ) g 2 ) ( v 1 ⊠ P ( w ) v 2 ) = g 1 v 1 ⊠ P ( w ) g 2 v 2 , v 1 ∈ M 1 , v 2 ∈ M 2 .
The top degree of M 1 ⊠ P ( w ) M 2 M^1 \boxtimes_{P(w)} M^2 M 1 ⊠ P ( w ) M 2
Now assume the category C 1 ( V ) C_1(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^{\mathrm L}(\operatorname{Mod}_{\mathrm{fd}}(A(V)))\subseteq C_1(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) . We fix a set W \mathcal{W} W of representatives of the irreducible objects in C 1 ( V ) C_1(V) C 1 ( V ) . From the proof of Proposition 3.11 , every M ∈ C 1 ( V ) M\in C_1(V) M ∈ C 1 ( V ) is a finite direct sum of irreducible ordinary modules. Since an irreducible module W ∈ W W\in\mathcal{W} W ∈ W is admissible W = ⨁ n = 0 ∞ W ( n ) W=\bigoplus_{n=0}^{\infty}W(n) W = ⨁ n = 0 ∞ W ( n ) , with Ω ( W ) = W ( 0 ) \Omega(W)=W(0) Ω ( W ) = W ( 0 ) , we may write
M = ⨁ W ∈ W n W W , M ( n ) = ⨁ W ∈ W n W W ( n ) , n ≥ 0. (44) M=\bigoplus_{W\in\mathcal{W}}n_WW,\qquad M(n)=\bigoplus_{W\in\mathcal{W}}n_WW(n),\qquad n\geq0. \tag*{(44)} M = W ∈ W ⨁ n W W , M ( n ) = W ∈ W ⨁ n W W ( n ) , n ≥ 0. (44)
Then M M M is admissible, with Ω ( M ) = ⨁ W ∈ W n W Ω ( W ) = M ( 0 ) \Omega(M)=\bigoplus_{W\in\mathcal{W}}n_W\Omega(W)=M(0) Ω ( M ) = ⨁ W ∈ W n W Ω ( W ) = M ( 0 ) . Write
π 0 : M ‾ = ∏ n = 0 ∞ M ( n ) ⟶ Ω ( M ) = M ( 0 ) (45) \pi_0:\overline{M}=\prod_{n=0}^{\infty}M(n)\longrightarrow\Omega(M)=M(0) \tag*{(45)} π 0 : M = n = 0 ∏ ∞ M ( n ) ⟶ Ω ( M ) = M ( 0 ) (45)
for the bottom-degree projection. For a V V V -module homomorphism f : M → N f:M\to N f : M → N and y ∈ C y\in\mathbb{C} y ∈ C , it satisfies
π 0 f ‾ = Ω ( f ) π 0 , π 0 e y L ( 0 ) = e y [ ω ] π 0 , π 0 e y L ( − 1 ) = π 0 . (46) \pi_0\overline{f}=\Omega(f)\pi_0,\qquad\pi_0e^{yL(0)}=e^{y[\omega]}\pi_0,\qquad\pi_0e^{yL(-1)}=\pi_0. \tag*{(46)} π 0 f = Ω ( f ) π 0 , π 0 e y L ( 0 ) = e y [ ω ] π 0 , π 0 e y L ( − 1 ) = π 0 . (46)
Indeed, module homomorphisms preserve this grading, L ( 0 ) L(0) L ( 0 ) acts on Ω ( M ) \Omega(M) Ω ( M ) by [ ω ] [\omega] [ ω ] , and L ( − 1 ) L(-1) L ( − 1 ) raises degree.
Theorem 3.14. Assume C 1 ( V ) C_1(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊆ C 1 ( V ) \Phi^{\mathrm L}(\operatorname{Mod}_{\mathrm{fd}}(A(V)))\subseteq C_1(V) Φ L ( Mod fd ( A ( V ))) ⊆ C 1 ( V ) . Let M 1 , M 2 ∈ C 1 ( V ) M^1,M^2\in C_1(V) M 1 , M 2 ∈ C 1 ( V ) . Then the canonical map
η w = η w M 1 , M 2 : M 1 ⊙ w , 0 M 2 ⟶ Ω ( M 1 ⊠ P ( w ) M 2 ) , v 1 ⊙ w v 2 ⟼ π 0 ( v 1 ⊠ P ( w ) v 2 ) (47) \eta_w=\eta_w^{M^1,M^2}:M^1\odot_{w,0}M^2\longrightarrow\Omega(M^1\boxtimes_{P(w)}M^2),\qquad v^1\odot_w v^2\longmapsto\pi_0(v^1\boxtimes_{P(w)}v^2) \tag*{(47)} η w = η w M 1 , M 2 : M 1 ⊙ w , 0 M 2 ⟶ Ω ( M 1 ⊠ P ( w ) M 2 ) , v 1 ⊙ w v 2 ⟼ π 0 ( v 1 ⊠ P ( w ) v 2 ) (47) is a well-defined isomorphism of left A ( V ) A(V) A ( V ) -modules, which is natural in M 1 M^1 M 1 and M 2 M^2 M 2 .
Proof. By Proposition 3.11 , ( Φ L ⊣ Ω ) : C A ( V ) = Mod f d ( A ( V ) ) ⇄ C 1 ( V ) (\Phi^{\mathrm L}\dashv\Omega):C_{A(V)}=\operatorname{Mod}_{\mathrm{fd}}(A(V))\rightleftarrows C_1(V) ( Φ L ⊣ Ω ) : C A ( V ) = Mod fd ( A ( V )) ⇄ C 1 ( V ) are equivalence of categories, and M 1 ⊙ w , 0 M 2 ∈ C A ( V ) M^1\odot_{w,0}M^2\in C_{A(V)} M 1 ⊙ w , 0 M 2 ∈ C A ( V ) for any M 1 , M 2 ∈ C 1 ( V ) M^1,M^2\in C_1(V) M 1 , M 2 ∈ C 1 ( V ) . Let M 3 ∈ C 1 ( V ) M^3\in C_1(V) M 3 ∈ C 1 ( V ) , then by our previous results, there is a commutative diagram of isomorphisms
M [ P ( w ) ] M 3 ( M 1 , M 2 ) → ≃ Prop. x r e f ‘ 2.11 ‘ C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ↓ ≃ ( 3.17 ) ↓ ≃ Lemma x r e f ‘ 3.6 ‘ Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) → ≃ β M 3 C r e s ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ↓ ≃ Prop. x r e f ‘ 3.11 ‘ ↓ ≃ Corollary x r e f ‘ 3.10 ‘ Hom A ( V ) ( Ω ( M 1 ⊠ P ( w ) M 2 ) , Ω ( M 3 ) ) → ≃ β M 3 r e s Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) . (48) \begin{aligned}
\mathcal{M}_{[P(w)]}^{M^3}(M^1,M^2)
&\xrightarrow[\simeq]{\text{Prop. }{xref}`2.11`}
\mathcal{C}((M^3)'\otimes M^1\otimes M^2)_{(P^1,\infty,w,0)}\\
\downarrow{\scriptstyle\simeq\ (3.17)}\qquad
&\qquad\downarrow{\scriptstyle\simeq\ \text{Lemma }{xref}`3.6`}\\
\operatorname{Hom}_{C_1(V)}(M^1\boxtimes_{P(w)}M^2,M^3)
&\xrightarrow[\simeq]{\beta_{M^3}}
\mathcal{C}_{\mathrm{res}}(\Omega((M^3)')\otimes M^1\otimes M^2)_{(P^1,\infty,w,0)}\\
\downarrow{\scriptstyle\simeq\ \text{Prop. }{xref}`3.11`}
&\qquad\downarrow{\scriptstyle\simeq\ \text{Corollary }{xref}`3.10`}\\
\operatorname{Hom}_{A(V)}(\Omega(M^1\boxtimes_{P(w)}M^2),\Omega(M^3))
&\xrightarrow[\simeq]{\beta^{\mathrm{res}}_{M^3}}
\operatorname{Hom}_{A(V)}(M^1\odot_{w,0}M^2,\Omega(M^3)). \tag*{(48)}
\end{aligned} M [ P ( w )] M 3 ( M 1 , M 2 ) ↓ ≃ ( 3.17 ) Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) ↓ ≃ Prop. x r e f ‘3.11‘ Hom A ( V ) ( Ω ( M 1 ⊠ P ( w ) M 2 ) , Ω ( M 3 )) Prop. x r e f ‘2.11‘ ≃ C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ↓ ≃ Lemma x r e f ‘3.6‘ β M 3 ≃ C res ( Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ↓ ≃ Corollary x r e f ‘3.10‘ β M 3 res ≃ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 )) . (48)
Note that all the isomorphisms in diagram (48) are natural in the input M 3 M^3 M 3 . Denote the middle horizontal isomorphism by
β M 3 : Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) → ≃ C r e s ( Ω ( ( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) . \beta_{M^3}:\operatorname{Hom}_{C_1(V)}(M^1\boxtimes_{P(w)}M^2,M^3)\xrightarrow{\simeq}\mathcal{C}_{\mathrm{res}}(\Omega((M^3)')\otimes M^1\otimes M^2)_{(P^1,\infty,w,0)}. β M 3 : Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) ≃ C res ( Ω (( M 3 ) ′ ) ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) .
Then by (43) , Proposition 2.11 , and Lemma 3.6 , given g ∈ Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) g\in\operatorname{Hom}_{C_1(V)}(M^1\boxtimes_{P(w)}M^2,M^3) g ∈ Hom C 1 ( V ) ( M 1 ⊠ P ( w ) M 2 , M 3 ) and v 1 ∈ M 1 v^1\in M^1 v 1 ∈ M 1 , v 2 ∈ M 2 v^2\in M^2 v 2 ∈ M 2 , and ( u 3 ) ′ ∈ Ω ( ( M 3 ) ′ ) (u^3)'\in\Omega((M^3)') ( u 3 ) ′ ∈ Ω (( M 3 ) ′ ) , we have
⟨ β M 3 ( g ) ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ φ I , p ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ I ( v 1 , e l p ( w ) ) v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ F ( v 1 ⊗ v 2 ) ⟩ = ⟨ ( u 3 ) ′ ∣ g ‾ ( v 1 ⊠ P ( w ) v 2 ) ⟩ = ⟨ ( u 3 ) ′ ∣ Ω ( g ) ( π 0 ( v 1 ⊠ P ( w ) v 2 ) ) ⟩ . (49) \begin{aligned}
\left\langle\beta_{M^3}(g)\mid(u^3)'\otimes v^1\otimes v^2\right\rangle
&=\left\langle\varphi_{I,p}\mid(u^3)'\otimes v^1\otimes v^2\right\rangle\\
&=\left\langle(u^3)'\mid I(v^1,e^{l_p(w)} )v^2\right\rangle\\
&=\left\langle(u^3)'\mid F(v^1\otimes v^2)\right\rangle\\
&=\left\langle(u^3)'\mid\overline{g}(v^1\boxtimes_{P(w)}v^2)\right\rangle\\
&=\left\langle(u^3)'\mid\Omega(g)(\pi_0(v^1\boxtimes_{P(w)}v^2))\right\rangle. \tag*{(49)}
\end{aligned} ⟨ β M 3 ( g ) ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ φ I , p ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ I ( v 1 , e l p ( w ) ) v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ F ( v 1 ⊗ v 2 ) ⟩ = ⟨ ( u 3 ) ′ ∣ g ( v 1 ⊠ P ( w ) v 2 ) ⟩ = ⟨ ( u 3 ) ′ ∣ Ω ( g ) ( π 0 ( v 1 ⊠ P ( w ) v 2 )) ⟩ . (49)
where the last equality follows from the fact that g g g preserves the grading subspaces. Denote the bottom horizontal isomorphism by
β M 3 r e s : Hom A ( V ) ( Ω ( M 1 ⊠ P ( w ) M 2 ) , Ω ( M 3 ) ) → ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) . \beta^{\mathrm{res}}_{M^3}:\operatorname{Hom}_{A(V)}\left(\Omega\left(M^1\boxtimes_{P(w)}M^2\right),\Omega(M^3)\right)\xrightarrow{\cong}\operatorname{Hom}_{A(V)}\left(M^1\odot_{w,0}M^2,\Omega(M^3)\right). β M 3 res : Hom A ( V ) ( Ω ( M 1 ⊠ P ( w ) M 2 ) , Ω ( M 3 ) ) ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) .
Let M 3 = M 1 ⊠ P ( w ) M 2 ∈ C 1 ( V ) M^3=M^1\boxtimes_{P(w)}M^2\in C_1(V) M 3 = M 1 ⊠ P ( w ) M 2 ∈ C 1 ( V ) , define
η w : = β M 1 ⊠ P ( w ) M 2 r e s ( Id Ω ( M 1 ⊠ P ( w ) M 2 ) ) ∈ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 1 ⊠ P ( w ) M 2 ) ) . \eta_w:=\beta^{\mathrm{res}}_{M^1\boxtimes_{P(w)}M^2}\left(\operatorname{Id}_{\Omega(M^1\boxtimes_{P(w)}M^2)}\right)\in\operatorname{Hom}_{A(V)}\left(M^1\odot_{w,0}M^2,\Omega\left(M^1\boxtimes_{P(w)}M^2\right)\right). η w := β M 1 ⊠ P ( w ) M 2 res ( Id Ω ( M 1 ⊠ P ( w ) M 2 ) ) ∈ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 1 ⊠ P ( w ) M 2 ) ) .
Since Ω \Omega Ω is an equivalence of category, then by (49) we have
⟨ ( u 3 ) ′ ∣ η w ( v 1 ⊙ w v 2 ) ⟩ = ⟨ ( u 3 ) ′ ∣ β M 1 ⊠ P ( w ) M 2 r e s ( Ω ( Id M 1 ⊠ P ( w ) M 2 ) ) ( v 1 ⊙ w v 2 ) ⟩ = ⟨ β M 1 ⊠ P ( w ) M 2 ( Id M 1 ⊠ P ( w ) M 2 ) ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ π 0 ( v 1 ⊠ P ( w ) v 2 ) ⟩ , \begin{aligned}
\left\langle(u^3)'\mid\eta_w(v^1\odot_w v^2)\right\rangle
&=\left\langle(u^3)'\mid\beta^{\mathrm{res}}_{M^1\boxtimes_{P(w)}M^2}\left(\Omega\left(\operatorname{Id}_{M^1\boxtimes_{P(w)}M^2}\right)\right)(v^1\odot_w v^2)\right\rangle\\
&=\left\langle\beta_{M^1\boxtimes_{P(w)}M^2}\left(\operatorname{Id}_{M^1\boxtimes_{P(w)}M^2}\right)\mid(u^3)'\otimes v^1\otimes v^2\right\rangle\\
&=\left\langle(u^3)'\mid\pi_0(v^1\boxtimes_{P(w)}v^2)\right\rangle,
\end{aligned} ⟨ ( u 3 ) ′ ∣ η w ( v 1 ⊙ w v 2 ) ⟩ = ⟨ ( u 3 ) ′ ∣ β M 1 ⊠ P ( w ) M 2 res ( Ω ( Id M 1 ⊠ P ( w ) M 2 ) ) ( v 1 ⊙ w v 2 ) ⟩ = ⟨ β M 1 ⊠ P ( w ) M 2 ( Id M 1 ⊠ P ( w ) M 2 ) ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ π 0 ( v 1 ⊠ P ( w ) v 2 ) ⟩ ,
where ( u 3 ) ′ ∈ Ω ( ( M 3 ) ′ ) (u^3)'\in\Omega((M^3)') ( u 3 ) ′ ∈ Ω (( M 3 ) ′ ) is arbitrary. This shows
η w ( v 1 ⊙ w v 2 ) = π 0 ( v 1 ⊠ P ( w ) v 2 ) , v 1 ∈ M 1 , v 2 ∈ M 2 . \eta_w(v^1\odot_w v^2)=\pi_0(v^1\boxtimes_{P(w)}v^2),\qquad v^1\in M^1,\ v^2\in M^2. η w ( v 1 ⊙ w v 2 ) = π 0 ( v 1 ⊠ P ( w ) v 2 ) , v 1 ∈ M 1 , v 2 ∈ M 2 .
Thus, (47) is a well-defined homomorphism of left A ( V ) A(V) A ( V ) -modules.
Furthermore, for any f ∈ Hom A ( V ) ( Ω ( M 1 ⊠ P ( w ) M 2 ) , Ω ( M 3 ) ) f\in\operatorname{Hom}_{A(V)}(\Omega(M^1\boxtimes_{P(w)}M^2),\Omega(M^3)) f ∈ Hom A ( V ) ( Ω ( M 1 ⊠ P ( w ) M 2 ) , Ω ( M 3 )) , by the full faithfulness of Ω \Omega Ω , there exists a unique V V V -module homomorphism g : M 1 ⊠ P ( w ) M 2 → M 3 g:M^1\boxtimes_{P(w)}M^2\to M^3 g : M 1 ⊠ P ( w ) M 2 → M 3 such that Ω ( g ) = f \Omega(g)=f Ω ( g ) = f . By (49) , we have
⟨ ( u 3 ) ′ ∣ β M 3 r e s ( f ) ( v 1 ⊙ w v 2 ) ⟩ = ⟨ β M 3 ( g ) ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ Ω ( g ) ( π 0 ( v 1 ⊠ P ( w ) v 2 ) ) ⟩ = ⟨ ( u 3 ) ′ ∣ ( f ∘ η w ) ( v 1 ⊙ w v 2 ) ⟩ . \begin{aligned}
\left\langle(u^3)'\mid\beta^{\mathrm{res}}_{M^3}(f)(v^1\odot_w v^2)\right\rangle
&=\left\langle\beta_{M^3}(g)\mid(u^3)'\otimes v^1\otimes v^2\right\rangle\\
&=\left\langle(u^3)'\mid\Omega(g)\left(\pi_0(v^1\boxtimes_{P(w)}v^2)\right)\right\rangle\\
&=\left\langle(u^3)'\mid(f\circ\eta_w)(v^1\odot_w v^2)\right\rangle.
\end{aligned} ⟨ ( u 3 ) ′ ∣ β M 3 res ( f ) ( v 1 ⊙ w v 2 ) ⟩ = ⟨ β M 3 ( g ) ∣ ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ ∣ Ω ( g ) ( π 0 ( v 1 ⊠ P ( w ) v 2 ) ) ⟩ = ⟨ ( u 3 ) ′ ∣ ( f ∘ η w ) ( v 1 ⊙ w v 2 ) ⟩ .
This shows β M 3 r e s ( f ) = f ∘ η w \beta^{\mathrm{res}}_{M^3}(f)=f\circ\eta_w β M 3 res ( f ) = f ∘ η w . In other words, β M 3 r e s = η w ∗ \beta^{\mathrm{res}}_{M^3}=\eta_w^* β M 3 res = η w ∗ is an isomorphism for every object S = Ω ( M 3 ) S=\Omega(M^3) S = Ω ( M 3 ) in the category C A ( V ) C_{A(V)} C A ( V ) . Then by the Yoneda lemma, η w \eta_w η w is also an isomorphism.
Finally, let f i : M i → N i f_i:M^i\to N^i f i : M i → N i be morphisms in C 1 ( V ) C_1(V) C 1 ( V ) , i = 1 , 2 i=1,2 i = 1 , 2 . Since both − ⊠ P ( w ) − -\boxtimes_{P(w)}- − ⊠ P ( w ) − and − ⊙ w , 0 − -\odot_{w,0}- − ⊙ w , 0 − are natural in the two arguments, by adopting a similar argument as above, we have
Ω ( f 1 ⊠ P ( w ) f 2 ) ∘ η w M 1 , M 2 = η w N 1 , N 2 ∘ ( f 1 ⊙ w , 0 f 2 ) . \Omega(f_1\boxtimes_{P(w)}f_2)\circ\eta_w^{M^1,M^2}=\eta_w^{N^1,N^2}\circ(f_1\odot_{w,0}f_2). Ω ( f 1 ⊠ P ( w ) f 2 ) ∘ η w M 1 , M 2 = η w N 1 , N 2 ∘ ( f 1 ⊙ w , 0 f 2 ) .
Hence η w = η w M 1 , M 2 \eta_w=\eta_w^{M^1,M^2} η w = η w M 1 , M 2 is natural in M 1 M^1 M 1 and M 2 M^2 M 2 . □
It follows from Lemma 2.8 that M [ P ( w ) ] M 1 M 2 M 3 ≅ M [ P ( 1 ) ] M 1 M 2 M 3 \mathcal{M}[P(w)]_{M^1M^2}^{M^3}\cong\mathcal{M}[P(1)]_{M^1M^2}^{M^3} M [ P ( w ) ] M 1 M 2 M 3 ≅ M [ P ( 1 ) ] M 1 M 2 M 3 . Moreover, by [39 Section 12.2] , one can take ⊠ = ⊠ P ( 1 ) \boxtimes=\boxtimes_{P(1)} ⊠ = ⊠ P ( 1 ) as the tensor product bifunctor of the braided tensor category C 1 ( V ) C_1(V) C 1 ( V ) .
In particular, Theorem 3.14 and Proposition 3.11 give the natural isomorphism
Hom V ( M 1 ⊠ P ( w ) M 2 , M 3 ) → ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) , g ⟼ Ω ( g ) ∘ η w , (50) \operatorname{Hom}_V\left(M^1\boxtimes_{P(w)}M^2,M^3\right)\xrightarrow{\cong}\operatorname{Hom}_{A(V)}\left(M^1\odot_{w,0}M^2,\Omega(M^3)\right),
\qquad g\longmapsto\Omega(g)\circ\eta_w, \tag*{(50)} Hom V ( M 1 ⊠ P ( w ) M 2 , M 3 ) ≅ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , Ω ( M 3 ) ) , g ⟼ Ω ( g ) ∘ η w , (50)
for M 3 ∈ C 1 ( V ) M^3\in C_1(V) M 3 ∈ C 1 ( V ) . This isomorphism will be used later.
Moreover, semisimplicity gives the decomposition
M 1 ⊙ w , 0 M 2 ≅ ⨁ W ∈ W W ( 0 ) ⊗ C I ( W M 1 M 2 ) ∗ , (51) M^1\odot_{w,0}M^2\cong\bigoplus_{W\in\mathcal{W}}W(0)\otimes_{\mathbb{C}}I\left(\frac{W}{M^1\ M^2}\right)^*, \tag*{(51)} M 1 ⊙ w , 0 M 2 ≅ W ∈ W ⨁ W ( 0 ) ⊗ C I ( M 1 M 2 W ) ∗ , (51)
where only finitely many summands are nonzero; see also [51 Section 6] . For simplicity, we denote M 1 ⊙ 1 , 0 M 2 M^1\odot_{1,0}M^2 M 1 ⊙ 1 , 0 M 2 by M 1 ⊙ M 2 M^1\odot M^2 M 1 ⊙ M 2 , and v 1 ⊙ 1 v 2 v^1\odot_1v^2 v 1 ⊙ 1 v 2 by v 1 ⊙ v 2 v^1\odot v^2 v 1 ⊙ v 2 as in [51 Section 4] . Then
M 1 ⊙ M 2 = span { v 1 ⊙ v 2 : v i ∈ M i , i = 1 , 2 } , M^1\odot M^2=\operatorname{span}\{v^1\odot v^2:v^i\in M^i,\ i=1,2\}, M 1 ⊙ M 2 = span { v 1 ⊙ v 2 : v i ∈ M i , i = 1 , 2 } ,
subjects to the following relations for s + t > 1 s+t>1 s + t > 1 and a ∈ V a\in V a ∈ V :
∑ j ≥ 0 ( w t a − s j ) ( a ( j − t ) v 1 ) ⊗ v 2 = − ∑ j ≥ 0 ( − t j ) ( − 1 ) − t − j v 1 ⊗ ( a ( w t a − s + j ) v 2 ) . (52) \sum_{j\geq0}\binom{wta-s}{j}\left(a_{(j-t)}v^{1}\right)\otimes v^{2}
=
-\sum_{j\geq0}\binom{-t}{j}(-1)^{-t-j}v^{1}\otimes\left(a_{(wta-s+j)}v^{2}\right). \tag*{(52)} j ≥ 0 ∑ ( j w t a − s ) ( a ( j − t ) v 1 ) ⊗ v 2 = − j ≥ 0 ∑ ( j − t ) ( − 1 ) − t − j v 1 ⊗ ( a ( w t a − s + j ) v 2 ) . (52)
Throughout this section, we assume that C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) is semisimple and Φ L ( Mod f d ( A ( V ) ) ) ⊂ C 1 ( V ) \Phi^{L}(\operatorname{Mod}_{\mathrm{fd}}(A(V)))\subset\mathcal{C}_{1}(V) Φ L ( Mod fd ( A ( V ))) ⊂ C 1 ( V ) . Using the point-shifting isomorphism R w R_{w} R w and the parallel transport of the restricted conformal blocks, we construct the isomorphisms R ± R_{\pm} R ± in Theorem 4.12 , and show that R − R_{-} R − is the restriction of the braiding isomorphism of Huang-Lepowsky-Zhang to the bottom degrees, see Theorem 4.14 .
The isomorphism I ( M 3 M 1 M 2 ) ≅ I ( M 3 M 2 M 1 ) I\left(\begin{smallmatrix}M^{3}\\M^{1}\ M^{2}\end{smallmatrix}\right)\cong I\left(\begin{smallmatrix}M^{3}\\M^{2}\ M^{1}\end{smallmatrix}\right) I ( M 3 M 1 M 2 ) ≅ I ( M 3 M 2 M 1 ) and the points shifting ( w , 0 ) ↦ ( 0 , − w ) (w,0)\mapsto(0,-w) ( w , 0 ) ↦ ( 0 , − w )
We first recall some of the constructions in [34 Section 7] . Let M 1 M^{1} M 1 , M 2 M^{2} M 2 , M 3 M^{3} M 3 be ordinary V V V -modules, and r ∈ Z r\in\mathbb{Z} r ∈ Z . There is a linear isomorphism
Ω r : I ( M 3 M 1 M 2 ) ≅ I ( M 3 M 2 M 1 ) , I ↦ Ω r ( I ) , \Omega_{r}:I\left(\begin{matrix}M^{3}\\M^{1}\ M^{2}\end{matrix}\right)\cong I\left(\begin{matrix}M^{3}\\M^{2}\ M^{1}\end{matrix}\right),\qquad I\mapsto\Omega_{r}(I), Ω r : I ( M 3 M 1 M 2 ) ≅ I ( M 3 M 2 M 1 ) , I ↦ Ω r ( I ) ,
Ω r ( I ) ( v 2 , e π i x ) v 1 = e − x L ( − 1 ) I ( v 1 , e 2 r π i x ) v 2 , \Omega_{r}(I)(v^{2},e^{\pi i}x)v^{1}=e^{-xL(-1)}I(v^{1},e^{2r\pi i}x)v^{2}, Ω r ( I ) ( v 2 , e π i x ) v 1 = e − xL ( − 1 ) I ( v 1 , e 2 r π i x ) v 2 ,
such that Ω 1 − r ( Ω r ( I ) ) = Ω r ( Ω 1 − r ( I ) ) = I \Omega_{1-r}(\Omega_{r}(I))=\Omega_{r}(\Omega_{1-r}(I))=I Ω 1 − r ( Ω r ( I )) = Ω r ( Ω 1 − r ( I )) = I , see [34 Proposition 7.1] . Our Ω r \Omega_{r} Ω r is the operator denoted by Ω r − 1 \Omega_{r-1} Ω r − 1 in [34 ] . We denote
Ω 0 ( I ) = I ~ , I ~ ( v 2 , e π i x ) v 1 = e − x L ( − 1 ) I ( v 1 , x ) v 2 . (53) \Omega_{0}(I)=\widetilde{I},\qquad\widetilde{I}(v^{2},e^{\pi i}x)v^{1}=e^{-xL(-1)}I(v^{1},x)v^{2}. \tag*{(53)} Ω 0 ( I ) = I , I ( v 2 , e π i x ) v 1 = e − xL ( − 1 ) I ( v 1 , x ) v 2 . (53)
Fix a determination l p ( w ) l_{p}(w) l p ( w ) of log ( w ) \log(w) log ( w ) as in (12) . We use l p ( w ) + π i l_{p}(w)+\pi i l p ( w ) + π i as the determination of log ( − w ) \log(-w) log ( − w ) . This isomorphism (53) leads to the following commutative diagram of linear isomorphisms:
I ( M 3 M 1 M 2 ) → Ω 0 I ( M 3 M 2 M 1 ) ≃ ↓ ↓ ≃ M [ P ( w ) ] M 1 M 2 M 3 → Ω 0 M [ P ( − w ) ] M 2 M 1 M 3 ≃ ↓ ↓ ≃ C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) → Ω 0 C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , − w , 0 ) (54) \begin{CD}
I\left(\begin{matrix}M^{3}\\M^{1}\ M^{2}\end{matrix}\right) @>{\Omega_{0}}>> I\left(\begin{matrix}M^{3}\\M^{2}\ M^{1}\end{matrix}\right)\\
@V{\simeq}VV @VV{\simeq}V\\
\mathcal{M}[P(w)]^{M^{3}}_{M^{1}M^{2}} @>{\Omega_{0}}>> \mathcal{M}[P(-w)]^{M^{3}}_{M^{2}M^{1}}\\
@V{\simeq}VV @VV{\simeq}V\\
\mathcal{C}((M^{3})'\otimes M^{1}\otimes M^{2})_{(\mathbb{P}^{1},\infty,w,0)}
@>{\Omega_{0}}>>
\mathcal{C}((M^{3})'\otimes M^{2}\otimes M^{1})_{(\mathbb{P}^{1},\infty,-w,0)} \tag*{(54)}
\end{CD} I ( M 3 M 1 M 2 ) ≃ ↓ ⏐ M [ P ( w ) ] M 1 M 2 M 3 ≃ ↓ ⏐ C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) Ω 0 Ω 0 Ω 0 I ( M 3 M 2 M 1 ) ↓ ⏐ ≃ M [ P ( − w ) ] M 2 M 1 M 3 ↓ ⏐ ≃ C (( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , − w , 0 ) (54)
where the vertical maps are given by the identification (24) . The correspondence of elements in these vector spaces are given by the following diagram:
\begin{CD}
I @>>> \widetilde{I}\\
@VVV @VVV\\
I(\cdot,e^{l_{p}(w)}) @>>> \widetilde{I}(\cdot,e^{\pi i+l_{p}(w)})\\
@VVV @VVV\\
\varphi@>>> \widetilde{\varphi}
\end{CD}
Then by (25) , (53) , and the fact that ⟨ v ′ ∣ L ( − 1 ) v ⟩ = ⟨ L ( 1 ) v ′ ∣ v ⟩ \langle v'|L(-1)v\rangle=\langle L(1)v'|v\rangle ⟨ v ′ ∣ L ( − 1 ) v ⟩ = ⟨ L ( 1 ) v ′ ∣ v ⟩ , we have
⟨ φ ~ ∣ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ ( v 3 ) ′ ∣ I ~ ( v 2 , e π i + l p ( w ) ) v 1 ⟩ = ⟨ ( v 3 ) ′ ∣ e − w L ( − 1 ) I ( v 1 , e l p ( w ) ) v 2 ⟩ = ⟨ e − w L ( 1 ) ( v 3 ) ′ ∣ I ( v 1 , e l p ( w ) ) v 2 ⟩ = ⟨ φ ∣ e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ . (55) \begin{aligned}
\langle\widetilde{\varphi}|(v^{3})'\otimes v^{2}\otimes v^{1}\rangle
=
\langle(v^{3})'|\widetilde{I}(v^{2},e^{\pi i+l_{p}(w)})v^{1}\rangle
=
\langle(v^{3})'|e^{-wL(-1)}I(v^{1},e^{l_{p}(w)})v^{2}\rangle\\
&=
\langle e^{-wL(1)}(v^{3})'|I(v^{1},e^{l_{p}(w)})v^{2}\rangle
=
\langle\varphi|e^{-wL(1)}(v^{3})'\otimes v^{1}\otimes v^{2}\rangle. \tag*{(55)}
\end{aligned} ⟨ φ ∣ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨( v 3 ) ′ ∣ I ( v 2 , e π i + l p ( w ) ) v 1 ⟩ = ⟨( v 3 ) ′ ∣ e − w L ( − 1 ) I ( v 1 , e l p ( w ) ) v 2 ⟩ = ⟨ e − w L ( 1 ) ( v 3 ) ′ ∣ I ( v 1 , e l p ( w ) ) v 2 ⟩ = ⟨ φ ∣ e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ . (55)
Note that e − w L ( 1 ) ( v 3 ) ′ e^{-wL(1)}(v^{3})' e − w L ( 1 ) ( v 3 ) ′ is a well-defined element in ( M 3 ) ′ (M^{3})' ( M 3 ) ′ , since L ( 1 ) . M ( n ) ⊂ M ( n − 1 ) L(1).M(n)\subset M(n-1) L ( 1 ) . M ( n ) ⊂ M ( n − 1 ) .
Remark 4.1. Note that the element ω ⊗ 1 = ω ⊗ z 0 ( z − w ) 0 ∈ L P 1 \ { ∞ , w , 0 } ( V ) \omega\otimes1=\omega\otimes\frac{z^{0}}{(z-w)^{0}} \in\mathcal{L}_{\mathbb{P}^{1}\backslash\{\infty,w,0\}}(V) ω ⊗ 1 = ω ⊗ ( z − w ) 0 z 0 ∈ L P 1 \ { ∞ , w , 0 } ( V ) . Then by (23) ,
ρ ∞ , w , 0 ( ω ⊗ 1 ) . ( ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) = − L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 + ( v 3 ) ′ ⊗ L ( − 1 ) v 1 ⊗ v 2 + ( v 3 ) ′ ⊗ v 1 ⊗ L ( − 1 ) v 2 . \rho_{\infty,w,0}(\omega\otimes1).((v^{3})' \otimes v^{1}\otimes v^{2})
=-L(1)(v^{3})'\otimes v^{1}\otimes v^{2}+(v^{3})'\otimes L(-1)v^{1}\otimes v^{2}+(v^{3})'\otimes v^{1}\otimes L(-1)v^{2}. ρ ∞ , w , 0 ( ω ⊗ 1 ) . (( v 3 ) ′ ⊗ v 1 ⊗ v 2 ) = − L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 + ( v 3 ) ′ ⊗ L ( − 1 ) v 1 ⊗ v 2 + ( v 3 ) ′ ⊗ v 1 ⊗ L ( − 1 ) v 2 .
In particular, in the space of three-point coinvariants, we have
[ L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ] = [ ( v 3 ) ′ ⊗ L ( − 1 ) v 1 ⊗ v 2 ] + [ ( v 3 ) ′ ⊗ v 1 ⊗ L ( − 1 ) v 2 ] , (56) [L(1)(v^{3})'\otimes v^{1}\otimes v^{2}]=[(v^{3})'\otimes L(-1)v^{1}\otimes v^{2}]+[(v^{3})'\otimes v^{1}\otimes L(-1)v^{2}], \tag*{(56)} [ L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ] = [( v 3 ) ′ ⊗ L ( − 1 ) v 1 ⊗ v 2 ] + [( v 3 ) ′ ⊗ v 1 ⊗ L ( − 1 ) v 2 ] , (56)
where [ α ] [\alpha] [ α ] is the equivalence class of α ∈ ( M 3 ) ′ ⊗ M 1 ⊗ M 2 \alpha\in(M^{3})'\otimes M^{1}\otimes M^{2} α ∈ ( M 3 ) ′ ⊗ M 1 ⊗ M 2 in the quotient space, see Definition 2.10. The operator e − w L ( 1 ) e^{-wL(1)} e − w L ( 1 ) acts on the restricted dual, whereas e − w L ( − 1 ) v i e^{-wL(-1)}v^{i} e − w L ( − 1 ) v i in general lies only in the completion of M i M^{i} M i . We therefore use the formula on the restricted dual in the following proposition.
Proposition 4.2. Let M 1 M^{1} M 1 , M 2 M^{2} M 2 , M 3 M^{3} M 3 be ordinary V V V -modules. Then there exists a linear isomorphism between the space of conformal blocks
Ω 0 : C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) P 1 , ∞ , w , 0 ≅ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) P 1 , ∞ , − w , 0 , φ ↦ φ ~ , \Omega_{0}:\mathcal{C}((M^{3})'\otimes M^{1}\otimes M^{2})_{\mathbb{P}^{1},\infty,w,0}\cong\mathcal{C}((M^{3})'\otimes M^{2}\otimes M^{1})_{\mathbb{P}^{1},\infty,-w,0},\qquad\varphi\mapsto\widetilde{\varphi}, Ω 0 : C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) P 1 , ∞ , w , 0 ≅ C (( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) P 1 , ∞ , − w , 0 , φ ↦ φ , ⟨ φ ~ | ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ | e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ . (57) \left\langle\widetilde{\varphi}\middle|(v^{3})'\otimes v^{2}\otimes v^{1}\right\rangle
=
\left\langle\varphi\middle|e^{-wL(1)}(v^{3})'\otimes v^{1}\otimes v^{2}\right\rangle. \tag*{(57)} ⟨ φ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ . (57) More precisely, the following identities hold
⟨ φ | ρ ∞ ( a ⊗ z n ( z − w ) m ) e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ φ ~ | ρ ∞ ( a ⊗ u − m ( u + w ) − n ) . ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ , \left\langle\varphi\middle|\rho_{\infty}\left(a\otimes\frac{z^{n}}{(z-w)^{m}}\right)e^{-wL(1)}(v^{3})'\otimes v^{1}\otimes v^{2}\right\rangle
=
\left\langle\widetilde{\varphi}\middle|\rho_{\infty}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right).(v^{3})'\otimes v^{2}\otimes v^{1}\right\rangle, ⟨ φ ρ ∞ ( a ⊗ ( z − w ) m z n ) e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ φ ρ ∞ ( a ⊗ ( u + w ) − n u − m ) . ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ , ⟨ φ | e − w L ( 1 ) ( v 3 ) ′ ⊗ ρ w ( a ⊗ z − n ( z − w ) m ) . v 1 ⊗ v 2 ⟩ = ⟨ φ ~ | ( v 3 ) ′ ⊗ v 2 ⊗ ρ 0 ( a ⊗ u − m ( u + w ) − n ) . v 1 ⟩ , \left\langle\varphi\middle|e^{-wL(1)}(v^{3})'\otimes\rho_{w}\left(a\otimes\frac{z^{-n}}{(z-w)^{m}}\right).v^{1}\otimes v^{2}\right\rangle
=
\left\langle\widetilde{\varphi}\middle|(v^{3})'\otimes v^{2}\otimes\rho_{0}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right).v^{1}\right\rangle, ⟨ φ e − w L ( 1 ) ( v 3 ) ′ ⊗ ρ w ( a ⊗ ( z − w ) m z − n ) . v 1 ⊗ v 2 ⟩ = ⟨ φ ( v 3 ) ′ ⊗ v 2 ⊗ ρ 0 ( a ⊗ ( u + w ) − n u − m ) . v 1 ⟩ , ⟨ φ | e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ ρ 0 ( a ⊗ z n ( z − w ) m ) . v 2 ⟩ = ⟨ φ ~ | ( v 3 ) ′ ⊗ ρ − w ( a ⊗ u − m ( u + w ) − n ) . v 2 ⊗ v 1 ⟩ , (58) \left\langle\varphi\middle|e^{-wL(1)}(v^{3})'\otimes v^{1}\otimes\rho_{0}\left(a\otimes\frac{z^{n}}{(z-w)^{m}}\right).v^{2}\right\rangle
=
\left\langle\widetilde{\varphi}\middle|(v^{3})'\otimes\rho_{-w}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right).v^{2}\otimes v^{1}\right\rangle, \tag*{(58)} ⟨ φ e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ ρ 0 ( a ⊗ ( z − w ) m z n ) . v 2 ⟩ = ⟨ φ ( v 3 ) ′ ⊗ ρ − w ( a ⊗ ( u + w ) − n u − m ) . v 2 ⊗ v 1 ⟩ , (58) for any m , n ∈ Z m,n\in\mathbb{Z} m , n ∈ Z , where u u u is the affine coordinate on P 1 \ { ∞ , − w , 0 } \mathbb{P}^{1}\backslash\{\infty,-w,0\} P 1 \ { ∞ , − w , 0 } .
Proof. The isomorphism (57) follows directly from the commutative diagram (54) and our calculation (55) . It also follows from the identities (58) , which gives a more precise relation between φ \varphi φ and φ ~ \widetilde{\varphi} φ .
We use the language of contour integration for residues to show (58) . All contours are positively oriented, and ∮ C \oint_{C} ∮ C denotes the normalized integral ( 2 π i ) − 1 ∮ C (2\pi i)^{-1}\oint_{C} ( 2 π i ) − 1 ∮ C . For the fixed complex number w ∈ P 1 \ { 0 , ∞ } w\in\mathbb{P}^{1}\backslash\{0,\infty\} w ∈ P 1 \ { 0 , ∞ } , consider the following contours in the affine coordinate z z z :
C ∞ ( z ) C_{\infty}(z) C ∞ ( z ) a circle in z z z centered at 0 0 0 such that ∣ z ∣ > 2 ∣ w ∣ |z|>2|w| ∣ z ∣ > 2∣ w ∣ ,
C 0 ( z ) C_{0}(z) C 0 ( z ) a circle in z z z centered at 0 0 0 such that ∣ z ∣ < ∣ w ∣ |z|<|w| ∣ z ∣ < ∣ w ∣ ,
C w ( z ) C_{w}(z) C w ( z ) a circle in z z z centered at w w w such that ∣ z − w ∣ < ∣ w ∣ |z-w|<|w| ∣ z − w ∣ < ∣ w ∣ .
Choose the determinations l p ( w ) l_{p}(w) l p ( w ) and l p ( w ) + π i l_{p}(w)+\pi i l p ( w ) + π i of log ( w ) \log(w) log ( w ) and log ( − w ) \log(-w) log ( − w ) , respectively. Let I ( ⋅ , e l p ( w ) ) I(\mathord{\cdot},e^{l_{p}(w)}) I ( ⋅ , e l p ( w ) ) and I ~ ( ⋅ , e l p ( w ) + π i ) \widetilde{I}(\mathord{\cdot},e^{l_{p}(w)+\pi i}) I ( ⋅ , e l p ( w ) + π i ) be the P ( w ) P(w) P ( w ) and P ( − w ) P(-w) P ( − w ) -intertwining operators corresponding to φ \varphi φ and φ ~ \widetilde{\varphi} φ , respectively. Then by (20) and Corollary 2.12, we have
⟨ φ | ρ ∞ ( a ⊗ z n ( z − w ) m ) e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = Res z o u t ⟨ Y M 3 ( ϑ ( a ) , z ) − 1 e − w L ( 1 ) ( v 3 ) ′ | I ( v 1 , e l p ( w ) ) v 2 ⟩ ∣ ∣ z ∣ > ∣ w ∣ ( z n ( z − w ) m ) d z = − Res z o u t ⟨ ( v 3 ) ′ | e − w L ( − 1 ) Y M 3 ( a , z ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ∣ ∣ z ∣ > ∣ w ∣ ( z n ( z − w ) m ) d z = − ∮ C ∞ ( z ) ⟨ ( v 3 ) ′ | Y M 3 ( a , z − w ) e − w L ( − 1 ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ∣ ∣ z ∣ > ∣ w ∣ ( z n ( z − w ) m ) d z \begin{aligned}
\left\langle\varphi\middle|\rho_{\infty}\left(a\otimes\frac{z^{n}}{(z-w)^{m}}\right)e^{-wL(1)}(v^{3})'\otimes v^{1}\otimes v^{2}\right\rangle
=
\operatorname{Res}^{\mathrm{out}}_{z}\left\langle Y_{M^{3}}\left(\vartheta(a),z\right)^{-1}e^{-wL(1)}(v^{3})'\middle|I\left(v^{1},e^{l_{p}(w)}\right)v^{2}\right\rangle\bigg|_{|z|>|w|}\left(\frac{z^{n}}{(z-w)^{m}}\right)\,dz \\
&=-\operatorname{Res}^{\mathrm{out}}_{z}\left\langle(v^{3})'\middle|e^{-wL(-1)}Y_{M^{3}}(a,z)I\left(v^{1},e^{l_{p}(w)}\right)v^{2}\right\rangle\bigg|_{|z|>|w|}\left(\frac{z^{n}}{(z-w)^{m}}\right)\,dz \\
&=-\oint_{C_{\infty}(z)}\left\langle(v^{3})'\middle|Y_{M^{3}}(a,z-w)e^{-wL(-1)}I\left(v^{1},e^{l_{p}(w)}\right)v^{2}\right\rangle\bigg|_{|z|>|w|}\left(\frac{z^{n}}{(z-w)^{m}}\right)\,dz
\end{aligned} ⟨ φ ρ ∞ ( a ⊗ ( z − w ) m z n ) e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = Res z out ⟨ Y M 3 ( ϑ ( a ) , z ) − 1 e − w L ( 1 ) ( v 3 ) ′ I ( v 1 , e l p ( w ) ) v 2 ⟩ ∣ z ∣ > ∣ w ∣ ( ( z − w ) m z n ) d z = − Res z out ⟨ ( v 3 ) ′ e − w L ( − 1 ) Y M 3 ( a , z ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ∣ z ∣ > ∣ w ∣ ( ( z − w ) m z n ) d z = − ∮ C ∞ ( z ) ⟨ ( v 3 ) ′ Y M 3 ( a , z − w ) e − w L ( − 1 ) I ( v 1 , e l p ( w ) ) v 2 ⟩ ∣ z ∣ > ∣ w ∣ ( ( z − w ) m z n ) d z
Figure 1. Integration contours for the residue computation
= − ∮ C ∞ ( z ) ⟨ ( v 3 ) ′ | Y M 3 ( a , z − w ) I ~ ( v 2 , e l p ( w ) + π i ) v 1 ⟩ ∣ z ∣ > ∣ w ∣ ( z n ( z − w ) m ) d z = − ∮ C ∞ ( u ) ⟨ ( v 3 ) ′ | Y M 3 ( a , u ) I ~ ( v 2 , e l p ( w ) + π i ) v 1 ⟩ ∣ u ∣ > ∣ − w ∣ ( ( u + w ) n u m ) d u ( z − w ) ↦ u = ⟨ φ ~ | ρ ∞ ( a ⊗ u − m ( u + w ) − n ) ⋅ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ , = -\oint_{C_{\infty}(z)} \left\langle(v^{3})' \middle| Y_{M^{3}}(a,z-w)\widetilde{I}(v^{2},e^{l_{p}(w)+\pi i})v^{1}\right\rangle_{\lvert z\rvert>\lvert w\rvert}\left(\frac{z^{n}}{(z-w)^{m}}\right) dz
\\
= -\oint_{C_{\infty}(u)} \left\langle(v^{3})' \middle| Y_{M^{3}}(a,u)\widetilde{I}(v^{2},e^{l_{p}(w)+\pi i})v^{1}\right\rangle_{\lvert u\rvert>\lvert-w\rvert}\left(\frac{(u+w)^{n}}{u^{m}}\right)du \qquad(z-w)\mapsto u
\\
= \left\langle\widetilde{\varphi}\middle|\rho_{\infty}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot(v^{3})'\otimes v^{2}\otimes v^{1}\right\rangle, = − ∮ C ∞ ( z ) ⟨ ( v 3 ) ′ Y M 3 ( a , z − w ) I ( v 2 , e l p ( w ) + π i ) v 1 ⟩ ∣ z ∣ > ∣ w ∣ ( ( z − w ) m z n ) d z = − ∮ C ∞ ( u ) ⟨ ( v 3 ) ′ Y M 3 ( a , u ) I ( v 2 , e l p ( w ) + π i ) v 1 ⟩ ∣ u ∣ > ∣ − w ∣ ( u m ( u + w ) n ) d u ( z − w ) ↦ u = ⟨ φ ρ ∞ ( a ⊗ ( u + w ) − n u − m ) ⋅ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ ,
where C ∞ ( u ) C_{\infty}(u) C ∞ ( u ) is a circle of u u u centered at 0 such that ∣ u ∣ > ∣ − w ∣ \lvert u\rvert>\lvert-w\rvert ∣ u ∣ > ∣ − w ∣ . Indeed, the translated contour lies in ∣ u ∣ > ∣ w ∣ \lvert u\rvert>\lvert w\rvert ∣ u ∣ > ∣ w ∣ , since the radius of C ∞ ( z ) C_{\infty}(z) C ∞ ( z ) is larger than 2 ∣ w ∣ 2\lvert w\rvert 2 ∣ w ∣ . The matrix coefficient is rational in u u u , with possible poles only at 0 and − w -w − w , so the translated contour can be deformed to C ∞ ( u ) C_{\infty}(u) C ∞ ( u ) without changing the integral. Similarly, by (2.18) and Corollary 2.12 , we have
⟨ φ | e − w L ( 1 ) ( v 3 ) ′ ⊗ ρ w ( a ⊗ z n ( z − w ) m ) ⋅ v 1 ⊗ v 2 ⟩ = Res z = w ⟨ e − w L ( 1 ) ( v 3 ) ′ | I ( Y M 1 ( a , z − w ) v 1 , e l p ( w ) ) v 2 ⟩ ∣ w ∣ > ∣ z − w ∣ ( z n ( z − w ) m ) d z = ∮ C w ( z ) ⟨ ( v 3 ) ′ | I ~ ( v 2 , e l p ( w ) + π i ) Y M 1 ( a , z − w ) v 1 ⟩ ∣ w ∣ > ∣ z − w ∣ ( z n ( z − w ) m ) d z = ∮ C 0 ( u ) ⟨ ( v 3 ) ′ | I ~ ( v 2 , e l p ( w ) + π i ) Y M 1 ( a , u ) v 1 ⟩ ∣ − w ∣ > ∣ u ∣ ( ( u + w ) n u m ) d u ( z − w ) ↦ u = ⟨ φ ~ | ( v 3 ) ′ ⊗ v 2 ⊗ ρ 0 ( a ⊗ u − m ( u + w ) − n ) ⋅ v 1 ⟩ , \begin{aligned}
\left\langle\varphi\middle|e^{-wL(1)}(v^{3})'\otimes\rho_{w}\left(a\otimes\frac{z^{n}}{(z-w)^{m}}\right)\cdot v^{1}\otimes v^{2}\right\rangle\\
&= \operatorname{Res}_{z=w}\left\langle e^{-wL(1)}(v^{3})'\middle|I\left(Y_{M^{1}}(a,z-w)v^{1},e^{l_{p}(w)}\right)v^{2}\right\rangle_{\lvert w\rvert>\lvert z-w\rvert}\left(\frac{z^{n}}{(z-w)^{m}}\right)dz \\
&= \oint_{C_{w}(z)}\left\langle(v^{3})'\middle|\widetilde{I}(v^{2},e^{l_{p}(w)+\pi i})Y_{M^{1}}(a,z-w)v^{1}\right\rangle_{\lvert w\rvert>\lvert z-w\rvert}\left(\frac{z^{n}}{(z-w)^{m}}\right)dz \\
&= \oint_{C_{0}(u)}\left\langle(v^{3})'\middle|\widetilde{I}(v^{2},e^{l_{p}(w)+\pi i})Y_{M^{1}}(a,u)v^{1}\right\rangle_{\lvert-w\rvert>\lvert u\rvert}\left(\frac{(u+w)^{n}}{u^{m}}\right)du \qquad(z-w)\mapsto u \\
&= \left\langle\widetilde{\varphi}\middle|(v^{3})'\otimes v^{2}\otimes\rho_{0}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^{1}\right\rangle,
\end{aligned} ⟨ φ e − w L ( 1 ) ( v 3 ) ′ ⊗ ρ w ( a ⊗ ( z − w ) m z n ) ⋅ v 1 ⊗ v 2 ⟩ = Res z = w ⟨ e − w L ( 1 ) ( v 3 ) ′ I ( Y M 1 ( a , z − w ) v 1 , e l p ( w ) ) v 2 ⟩ ∣ w ∣ > ∣ z − w ∣ ( ( z − w ) m z n ) d z = ∮ C w ( z ) ⟨ ( v 3 ) ′ I ( v 2 , e l p ( w ) + π i ) Y M 1 ( a , z − w ) v 1 ⟩ ∣ w ∣ > ∣ z − w ∣ ( ( z − w ) m z n ) d z = ∮ C 0 ( u ) ⟨ ( v 3 ) ′ I ( v 2 , e l p ( w ) + π i ) Y M 1 ( a , u ) v 1 ⟩ ∣ − w ∣ > ∣ u ∣ ( u m ( u + w ) n ) d u ( z − w ) ↦ u = ⟨ φ ( v 3 ) ′ ⊗ v 2 ⊗ ρ 0 ( a ⊗ ( u + w ) − n u − m ) ⋅ v 1 ⟩ ,
where C 0 ( u ) C_{0}(u) C 0 ( u ) is a circle of u u u centered at 0 such that ∣ u ∣ < ∣ − w ∣ \lvert u\rvert<\lvert-w\rvert ∣ u ∣ < ∣ − w ∣ . Finally, by (2.19) and Corollary 2.12 , we have
⟨ φ | e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ ρ 0 ( a ⊗ z n ( z − w ) m ) ⋅ v 2 ⟩ = Res z = 0 ⟨ e − w L ( 1 ) ( v 3 ) ′ | I ( v 1 , e l p ( w ) ) Y M 2 ( a , z ) v 2 ⟩ ∣ z ∣ < ∣ w ∣ ( z n ( z − w ) m ) d z = ∮ C 0 ( z ) ⟨ ( v 3 ) ′ | I ~ ( Y M 2 ( a , z ) v 2 , e l p ( w ) + π i ) v 1 ⟩ ∣ z ∣ < ∣ w ∣ ( z n ( z − w ) m ) d z = ∫ C − w ( u ) ⟨ ( v 3 ) ′ | I ~ ( Y M 2 ( a , u − ( − w ) ) v 2 , e l p ( w ) + π i v 1 ) ⟩ ∣ ∣ u + w ∣ < ∣ w ∣ ( ( u + w ) n u m ) d u ( z − w ) ↦ u = ⟨ φ ~ | ( v 3 ) ′ ⊗ ρ − w ( a ⊗ u − m ( u + w ) − n ) ⋅ v 2 ⊗ v 1 ⟩ , \begin{aligned}
\left\langle\varphi\middle|e^{-wL(1)}(v^{3})'\otimes v^{1}\otimes\rho_{0}\left(a\otimes\frac{z^{n}}{(z-w)^{m}}\right)\cdot v^{2}\right\rangle\\
&= \operatorname{Res}_{z=0}\left\langle e^{-wL(1)}(v^{3})'\middle|I(v^{1},e^{l_{p}(w)})Y_{M^{2}}(a,z)v^{2}\right\rangle_{\lvert z\rvert<\lvert w\rvert}\left(\frac{z^{n}}{(z-w)^{m}}\right)dz \\
&= \oint_{C_{0}(z)}\left\langle(v^{3})'\middle|\widetilde{I}\left(Y_{M^{2}}(a,z)v^{2},e^{l_{p}(w)+\pi i}\right)v^{1}\right\rangle_{\lvert z\rvert<\lvert w\rvert}\left(\frac{z^{n}}{(z-w)^{m}}\right)dz \\
&= \int_{C_{-w}(u)} \left\langle(v^3)' \middle| \widetilde{I}\left(Y_{M^2}(a,u-(-w))v^2,e^{l_p(w)+\pi i}v^1\right)\right\rangle\bigg|_{|u+w|<|w|}\left(\frac{(u+w)^n}{u^m}\right)\,du \qquad(z-w)\mapsto u \\
&= \left\langle\widetilde{\varphi}\middle| (v^3)'\otimes\rho_{-w}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^2\otimes v^1\right\rangle,
\end{aligned} ⟨ φ e − w L ( 1 ) ( v 3 ) ′ ⊗ v 1 ⊗ ρ 0 ( a ⊗ ( z − w ) m z n ) ⋅ v 2 ⟩ = Res z = 0 ⟨ e − w L ( 1 ) ( v 3 ) ′ I ( v 1 , e l p ( w ) ) Y M 2 ( a , z ) v 2 ⟩ ∣ z ∣ < ∣ w ∣ ( ( z − w ) m z n ) d z = ∮ C 0 ( z ) ⟨ ( v 3 ) ′ I ( Y M 2 ( a , z ) v 2 , e l p ( w ) + π i ) v 1 ⟩ ∣ z ∣ < ∣ w ∣ ( ( z − w ) m z n ) d z = ∫ C − w ( u ) ⟨ ( v 3 ) ′ I ( Y M 2 ( a , u − ( − w )) v 2 , e l p ( w ) + π i v 1 ) ⟩ ∣ u + w ∣ < ∣ w ∣ ( u m ( u + w ) n ) d u ( z − w ) ↦ u = ⟨ φ ( v 3 ) ′ ⊗ ρ − w ( a ⊗ ( u + w ) − n u − m ) ⋅ v 2 ⊗ v 1 ⟩ ,
This proves (58) . □
Remark 4.3. In view of (58) , the geometric effect of Ω 0 \Omega_0 Ω 0 -action on the space of three-point conformal blocks is given by shifting the marked points ( w , 0 ) (w,0) ( w , 0 ) uniformly to ( 0 , − w ) (0,-w) ( 0 , − w ) , with attached modules unchanged. i.e.,
( ∞ , w , 0 , ( M 3 ) ′ , M 1 , M 2 ) ↦ ( ∞ , 0 , − w , ( M 3 ) ′ , M 1 , M 2 ) . (\infty,w,0,(M^3)',M^1,M^2)\mapsto(\infty,0,-w,(M^3)',M^1,M^2). ( ∞ , w , 0 , ( M 3 ) ′ , M 1 , M 2 ) ↦ ( ∞ , 0 , − w , ( M 3 ) ′ , M 1 , M 2 ) .
See Figure 2 for an illustration. In fact, with the point at ∞ \infty ∞ fixed, these three-point conformal
Figure 2. Translation of the marked points ( w , 0 ) (w, 0) ( w , 0 ) to $(0, −w)
blocks form a sheaf C \mathcal{C} C over the configuration space Conf 2 ( A 1 ) = { ( z 1 , z 2 ) ∈ A 1 × A 1 : z 1 ≠ z 2 } \operatorname{Conf}_2(\mathbb{A}^1)=\{(z_1,z_2)\in\mathbb{A}^1\times\mathbb{A}^1:z_1\ne z_2\} Conf 2 ( A 1 ) = {( z 1 , z 2 ) ∈ A 1 × A 1 : z 1 = z 2 } , equipped with its standard flat connection [21 , 12 ] . The isomorphism Ω 0 \Omega_0 Ω 0 is given by parallel transport along the straight line path ( w , 0 ) ↦ ( 0 , − w ) (w,0)\mapsto(0,-w) ( w , 0 ) ↦ ( 0 , − w ) , followed by reordering the two finite marked points and their attached modules.
The A ( V ) A(V) A ( V ) -module isomorphism R w : M 2 ⊙ − w , 0 M 1 → M 1 ⊙ w , 0 M 2 R_w:M^2\mathbin{\odot}_{-w,0}M^1\to M^1\mathbin{\odot}_{w,0}M^2 R w : M 2 ⊙ − w , 0 M 1 → M 1 ⊙ w , 0 M 2
Proposition 4.4. Let M 1 , M 2 , M 3 ∈ C 1 ( V ) M^1,M^2,M^3\in\mathcal{C}_1(V) M 1 , M 2 , M 3 ∈ C 1 ( V ) . Then there exists a linear isomorphism which is natural in M 1 , M 2 M^1,M^2 M 1 , M 2 :
Ω 0 : Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 3 ( 0 ) ) ≅ Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 3 ( 0 ) ) , f ↦ f ~ , \Omega_0:\operatorname{Hom}_{A(V)}(M^1\mathbin{\odot}_{w,0}M^2,M^3(0))\cong\operatorname{Hom}_{A(V)}(M^2\mathbin{\odot}_{-w,0}M^1,M^3(0)),\qquad f\mapsto\widetilde{f}, Ω 0 : Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 3 ( 0 )) ≅ Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 3 ( 0 )) , f ↦ f ,
⟨ ( u 3 ) ′ | f ~ ( v 2 ⊙ − w v 1 ) ⟩ = ⟨ ( u 3 ) ′ | f ( v 1 ⊙ w v 2 ) ⟩ , \left\langle(u^3)'\middle|\widetilde{f}(v^2\mathbin{\odot}_{-w}v^1)\right\rangle=\left\langle(u^3)'\middle|f(v^1\mathbin{\odot}_{w}v^2)\right\rangle, ⟨ ( u 3 ) ′ f ( v 2 ⊙ − w v 1 ) ⟩ = ⟨ ( u 3 ) ′ f ( v 1 ⊙ w v 2 ) ⟩ ,
for all ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^3)'\in M^3(0)^* ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ , v 1 ∈ M 1 v^1\in M^1 v 1 ∈ M 1 , and v 2 ∈ M 2 v^2\in M^2 v 2 ∈ M 2 .
Proof. Using Lemma 3.6 and Corollary 3.10, we can extend the diagram (54) to the following commutative diagram:
I ( M 3 M 1 M 2 ) → Ω 0 I ( M 3 M 2 M 1 ) ≃ ↓ ↓ ≃ M [ P ( w ) ] M 1 M 2 M 3 → Ω 0 M [ P ( − w ) ] M 2 M 1 M 3 ≃ ↓ ↓ ≃ C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) → Ω 0 C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , − w , 0 ) ( 3.8 ) ↓ ≃ ↓ ( 3.8 ) C r e s ( M 3 ( 0 ) ∗ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) → Ω 0 C r e s ( M 3 ( 0 ) ∗ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , − w , 0 ) ( 3.14 ) ↓ ≃ ↓ ( 3.14 ) Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 3 ( 0 ) ) → Ω 0 Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 3 ( 0 ) ) (59) \begin{CD}
I\left(\frac{M^{3}}{M^{1}\ M^{2}}\right) @>{\Omega_{0}}>> I\left(\frac{M^{3}}{M^{2}\ M^{1}}\right) \\
@V{\simeq}VV @VV{\simeq}V \\
\mathcal{M}[P(w)]^{M^{3}}_{M^{1}M^{2}} @>{\Omega_{0}}>> \mathcal{M}[P(-w)]^{M^{3}}_{M^{2}M^{1}} \\
@V{\simeq}VV @VV{\simeq}V \\
\mathcal{C}((M^{3})' \otimes M^{1} \otimes M^{2})_{(\mathbb{P}^{1},\infty,w,0)} @>{\Omega_{0}}>> \mathcal{C}((M^{3})' \otimes M^{2} \otimes M^{1})_{(\mathbb{P}^{1},\infty,-w,0)} \\
@V{(3.8)}V{\simeq}V @VV{(3.8)}V \\
\mathcal{C}_{\mathrm{res}}(M^{3}(0)^{*}\otimes M^{1}\otimes M^{2})_{(\mathbb{P}^{1},\infty,w,0)} @>{\Omega_{0}}>> \mathcal{C}_{\mathrm{res}}(M^{3}(0)^{*}\otimes M^{2}\otimes M^{1})_{(\mathbb{P}^{1},\infty,-w,0)} \\
@V{(3.14)}V{\simeq}V @VV{(3.14)}V \\
\operatorname{Hom}_{A(V)}(M^{1}\mathbin{\odot_{w,0}}M^{2},M^{3}(0)) @>{\Omega_{0}}>> \operatorname{Hom}_{A(V)}(M^{2}\mathbin{\odot_{-w,0}}M^{1},M^{3}(0)) \tag*{(59)}
\end{CD} I ( M 1 M 2 M 3 ) ≃ ↓ ⏐ M [ P ( w ) ] M 1 M 2 M 3 ≃ ↓ ⏐ C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ( 3.8 ) ↓ ⏐ ≃ C res ( M 3 ( 0 ) ∗ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) ( 3.14 ) ↓ ⏐ ≃ Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 3 ( 0 )) Ω 0 Ω 0 Ω 0 Ω 0 Ω 0 I ( M 2 M 1 M 3 ) ↓ ⏐ ≃ M [ P ( − w ) ] M 2 M 1 M 3 ↓ ⏐ ≃ C (( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , − w , 0 ) ↓ ⏐ ( 3.8 ) C res ( M 3 ( 0 ) ∗ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , − w , 0 ) ↓ ⏐ ( 3.14 ) Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 3 ( 0 )) (59)
Given φ ∈ C ( ( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) \varphi\in\mathcal{C}((M^{3})'\otimes M^{1}\otimes M^{2})_{(\mathbb{P}^{1},\infty,w,0)} φ ∈ C (( M 3 ) ′ ⊗ M 1 ⊗ M 2 ) ( P 1 , ∞ , w , 0 ) , we denote the correspondence of elements in the vector spaces of (59) by
\begin{CD}
\varphi@>{\Omega_{0}}>> \widetilde{\varphi}\\
@V{(3.8)}VV @VV{(3.8)}V\\
\varphi_{\mathrm{res}} @>{\Omega_{0}}>> \widetilde{\varphi}_{\mathrm{res}}\\
@V{(3.14)}VV @VV{(3.14)}V\\
f @>{\Omega_{0}}>> \widetilde{f}.
\end{CD}
Note that for ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^{3})'\in M^{3}(0)^{*} ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ , we have L ( 1 ) . ( u 3 ) ′ = 0 L(1).(u^{3})'=0 L ( 1 ) . ( u 3 ) ′ = 0 . Then by {(xref)}(4.5) and (41) , we have
⟨ ( u 3 ) ′ | f ~ ( v 2 ⊙ − w v 1 ) ⟩ = ⟨ φ ~ r e s | ( u 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ r e s | e − w L ( 1 ) ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ φ r e s | ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ | f ( v 1 ⊙ w v 2 ) ⟩ . (60) \begin{aligned}
\left\langle(u^{3})'\middle| \widetilde{f}(v^{2}\mathbin{\odot_{-w}}v^{1})\right\rangle
=
\left\langle\widetilde{\varphi}_{\mathrm{res}}\middle| (u^{3})'\otimes v^{2}\otimes v^{1}\right\rangle
=
\left\langle\varphi_{\mathrm{res}}\middle|e^{-wL(1)}(u^{3})'\otimes v^{1}\otimes v^{2}\right\rangle\\
&=
\left\langle\varphi_{\mathrm{res}}\middle| (u^{3})'\otimes v^{1}\otimes v^{2}\right\rangle
=
\left\langle(u^{3})'\middle|f(v^{1}\mathbin{\odot_{w}}v^{2})\right\rangle. \tag*{(60)}
\end{aligned} ⟨ ( u 3 ) ′ f ( v 2 ⊙ − w v 1 ) ⟩ = ⟨ φ res ( u 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ res e − w L ( 1 ) ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ φ res ( u 3 ) ′ ⊗ v 1 ⊗ v 2 ⟩ = ⟨ ( u 3 ) ′ f ( v 1 ⊙ w v 2 ) ⟩ . (60)
where v 1 ∈ M 1 v^{1}\in M^{1} v 1 ∈ M 1 , v 2 ∈ M 2 v^{2}\in M^{2} v 2 ∈ M 2 , and ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^{3})'\in M^{3}(0)^{*} ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ . Finally, given morphisms α : N 1 → M 1 \alpha:N^{1}\to M^{1} α : N 1 → M 1 and β : N 2 → M 2 \beta:N^{2}\to M^{2} β : N 2 → M 2 , we claim that
Ω 0 N 1 , N 2 ( f ∘ ( α ⊙ w , 0 β ) ) = Ω 0 M 1 , M 2 ( f ) ∘ ( β ⊙ − w , 0 α ) , (61) \Omega_{0}^{N^{1},N^{2}}(f\circ(\alpha\mathbin{\odot_{w,0}}\beta))
=
\Omega_{0}^{M^{1},M^{2}}(f)\circ(\beta\mathbin{\odot_{-w,0}}\alpha), \tag*{(61)} Ω 0 N 1 , N 2 ( f ∘ ( α ⊙ w , 0 β )) = Ω 0 M 1 , M 2 ( f ) ∘ ( β ⊙ − w , 0 α ) , (61)
for any f : M 1 ⊙ w , 0 M 2 → M 3 ( 0 ) f:M^{1}\mathbin{\odot_{w,0}}M^{2}\to M^{3}(0) f : M 1 ⊙ w , 0 M 2 → M 3 ( 0 ) an A ( V ) A(V) A ( V ) -homomorphism. Indeed, by the definition of Ω 0 ( f ) \Omega_{0}(f) Ω 0 ( f ) , we have for any ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^{3})'\in M^{3}(0)^{*} ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ ,
⟨ ( u 3 ) ′ | ( Ω 0 N 1 , N 2 ( f ∘ ( α ⊙ w , 0 β ) ) ( n 2 ⊙ − w n 1 ) ) ⟩ = ⟨ ( u 3 ) ′ | f ( α ( n 1 ) ⊙ w β ( n 2 ) ) ⟩ = ⟨ ( u 3 ) ′ | f ~ ( β ( n 2 ) ⊙ − w α ( n 1 ) ) ⟩ = ⟨ ( u 3 ) ′ | ( Ω 0 M 1 , M 2 ( f ) ∘ ( β ⊙ − w , 0 α ) ) ( n 2 ⊙ − w n 1 ) ⟩ . \begin{aligned}
\left\langle(u^{3})'\middle|\left(\Omega_{0}^{N^{1},N^{2}}(f\circ(\alpha\mathbin{\odot_{w,0}}\beta))(n^{2}\mathbin{\odot_{-w}}n^{1})\right)\right\rangle
=
\left\langle(u^{3})'\middle|f(\alpha(n^{1})\mathbin{\odot_{w}}\beta(n^{2}))\right\rangle\\
&=
\left\langle(u^{3})'\middle|\widetilde{f}(\beta(n^{2})\mathbin{\odot_{-w}}\alpha(n^{1}))\right\rangle
=
\left\langle(u^{3})'\middle|\left(\Omega_{0}^{M^{1},M^{2}}(f)\circ(\beta\mathbin{\odot_{-w,0}}\alpha)\right)(n^{2}\mathbin{\odot_{-w}}n^{1})\right\rangle.
\end{aligned} ⟨ ( u 3 ) ′ ( Ω 0 N 1 , N 2 ( f ∘ ( α ⊙ w , 0 β )) ( n 2 ⊙ − w n 1 ) ) ⟩ = ⟨ ( u 3 ) ′ f ( α ( n 1 ) ⊙ w β ( n 2 )) ⟩ = ⟨ ( u 3 ) ′ f ( β ( n 2 ) ⊙ − w α ( n 1 )) ⟩ = ⟨ ( u 3 ) ′ ( Ω 0 M 1 , M 2 ( f ) ∘ ( β ⊙ − w , 0 α ) ) ( n 2 ⊙ − w n 1 ) ⟩ .
This shows (61) . Hence Ω 0 \Omega_{0} Ω 0 is natural in M 1 M^{1} M 1 and M 2 M^{2} M 2 .
Lemma 4.5. Let M 1 , M 2 , M 3 ∈ C 1 ( V ) M^{1},M^{2},M^{3}\in\mathcal{C}_{1}(V) M 1 , M 2 , M 3 ∈ C 1 ( V ) , and
Ω 0 : Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 3 ( 0 ) ) ≃ Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 3 ( 0 ) ) , f ↦ f ~ \Omega_{0}:\operatorname{Hom}_{A(V)}(M^{1}\mathbin{\odot_{w,0}}M^{2},M^{3}(0))
\simeq
\operatorname{Hom}_{A(V)}(M^{2}\mathbin{\odot_{-w,0}}M^{1},M^{3}(0)),
\qquad f\mapsto\widetilde{f} Ω 0 : Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 3 ( 0 )) ≃ Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 3 ( 0 )) , f ↦ f
be the isomorphism given by (59) . Let a ⊗ ( z n / ( z − w ) m ) ∈ L P 1 \ { w , 0 } ( V ) ≤ 0 a\otimes\left(z^{n}/(z-w)^{m}\right)\in\mathcal{L}_{\mathbb{P}^{1}\backslash\{w,0\}}(V)_{\leq0} a ⊗ ( z n / ( z − w ) m ) ∈ L P 1 \ { w , 0 } ( V ) ≤ 0 , with n − m ≤ wt a − 1 n-m\leq\operatorname{wt}a-1 n − m ≤ wt a − 1 , and a ⊗ ( u − m / ( u + w ) − n ) ∈ L P 1 \ { − w , 0 } ( V ) ≤ 0 a\otimes\left(u^{-m}/(u+w)^{-n}\right)\in\mathcal{L}_{\mathbb{P}^{1}\backslash\{-w,0\}}(V)_{\leq0} a ⊗ ( u − m / ( u + w ) − n ) ∈ L P 1 \ { − w , 0 } ( V ) ≤ 0 . Then
f ( ρ w ( a ⊗ z n ( z − w ) m ) . v 1 ⊙ w v 2 ) = f ~ ( v 2 ⊙ − w ρ 0 ( a ⊗ u − m ( u + w ) − n ) . v 1 ) . (62) f\left(\rho_{w}\left(a\otimes\frac{z^{n}}{(z-w)^{m}}\right).v^{1}\mathbin{\odot_{w}}v^{2}\right)
=
\widetilde{f}\left(v^{2}\mathbin{\odot_{-w}}\rho_{0}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right).v^{1}\right). \tag*{(62)} f ( ρ w ( a ⊗ ( z − w ) m z n ) . v 1 ⊙ w v 2 ) = f ( v 2 ⊙ − w ρ 0 ( a ⊗ ( u + w ) − n u − m ) . v 1 ) . (62)
f ( v 1 ⊙ w ρ 0 ( a ⊗ z n ( z − w ) m ) ⋅ v 2 ) = f ~ ( ρ − w ( a ⊗ u − m ( u + w ) − n ) ⋅ v 2 ⊙ − w v 1 ) . (63) f\left(v^{1}\mathbin{\odot_w}\rho_0\left(a\otimes\frac{z^n}{(z-w)^m}\right)\cdot v^{2}\right)=\widetilde{f}\left(\rho_{-w}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^{2}\mathbin{\odot_{-w}}v^{1}\right). \tag*{(63)} f ( v 1 ⊙ w ρ 0 ( a ⊗ ( z − w ) m z n ) ⋅ v 2 ) = f ( ρ − w ( a ⊗ ( u + w ) − n u − m ) ⋅ v 2 ⊙ − w v 1 ) . (63)
Proof. By (58) and (41) , and the fact that e − w L ( 1 ) ( u 3 ) ′ = ( u 3 ) ′ e^{-wL^{(1)}(u^3)'=(u^3)'} e − w L ( 1 ) ( u 3 ) ′ = ( u 3 ) ′ for all ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^3)'\in M^3(0)^* ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ , we have
\begin{aligned}
\left\langle(u^3)'\mathrel{\left|}f\left(\rho_w\left(a\otimes\frac{z^n}{(z-w)^m}\right)\cdot v^1\mathbin{\odot_w}v^2\right)\right.\right\rangle\\
&=\left\langle\widetilde{\varphi}_{\mathrm{res}}\left|(u^3)'\mathbin{\otimes}\rho_w\left(a\otimes\frac{z^n}{(z-w)^m}\right)\cdot v^1\otimes v^2\right.\right\rangle
=\left\langle\widetilde{\varphi}_{\mathrm{res}}\left|(u^3)'\otimes v^2\otimes\rho_0\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^1\right.\right\rangle\\
&=\left\langle(u^3)'\mathrel{\left|}\widetilde{f}\left(v^2\mathbin{\odot_{-w}}\rho_0\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^1\right)\right.\right\rangle.
\end{aligned}
This proves (62) . Similarly, (63) follows from the third equality in (58) and (41) , we omit the details. □
The following is our main theorem in this Section, which is essential for the definition of braiding isomorphism on Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) .
Theorem 4.6. Let M 1 , M 2 ∈ C 1 ( V ) M^1,M^2\in\mathcal{C}_1(V) M 1 , M 2 ∈ C 1 ( V ) , and w ∈ P 1 ∖ { ∞ , 0 } w\in\mathbb{P}^1\setminus\{\infty,0\} w ∈ P 1 ∖ { ∞ , 0 } . Then
R w : M 2 ⊙ − w , 0 M 1 ⟶ M 1 ⊙ w , 0 M 2 , R w ( v 2 ⊙ − w v 1 ) = v 1 ⊙ w v 2 (64) R_w:M^2\mathbin{\odot_{-w,0}}M^1\longrightarrow M^1\mathbin{\odot_{w,0}}M^2,\qquad R_w\left(v^2\mathbin{\odot_{-w}}v^1\right)=v^1\mathbin{\odot_w}v^2 \tag*{(64)} R w : M 2 ⊙ − w , 0 M 1 ⟶ M 1 ⊙ w , 0 M 2 , R w ( v 2 ⊙ − w v 1 ) = v 1 ⊙ w v 2 (64) is a canonical isomorphism of left A ( V ) A(V) A ( V ) -modules which is natural in M 1 M^1 M 1 and M 2 M^2 M 2 . It satisfies
R w ( v 2 ⊙ − w ρ 0 ( a ⊗ u − m ( u + w ) − n ) ⋅ v 1 ) = ρ w ( a ⊗ z n ( z − w ) m ) ⋅ v 1 ⊙ w v 2 , (65) R_w\left(v^2\mathbin{\odot_{-w}}\rho_0\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^1\right)=\rho_w\left(a\otimes\frac{z^n}{(z-w)^m}\right)\cdot v^1\mathbin{\odot_w}v^2, \tag*{(65)} R w ( v 2 ⊙ − w ρ 0 ( a ⊗ ( u + w ) − n u − m ) ⋅ v 1 ) = ρ w ( a ⊗ ( z − w ) m z n ) ⋅ v 1 ⊙ w v 2 , (65) R w ( ρ − w ( a ⊗ u − m ( u + w ) − n ) ⋅ v 2 ⊙ − w v 1 ) = v 1 ⊙ w ρ 0 ( a ⊗ z n ( z − w ) m ) ⋅ v 2 , (66) R_w\left(\rho_{-w}\left(a\otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\cdot v^2\mathbin{\odot_{-w}}v^1\right)=v^1\mathbin{\odot_w}\rho_0\left(a\otimes\frac{z^n}{(z-w)^m}\right)\cdot v^2, \tag*{(66)} R w ( ρ − w ( a ⊗ ( u + w ) − n u − m ) ⋅ v 2 ⊙ − w v 1 ) = v 1 ⊙ w ρ 0 ( a ⊗ ( z − w ) m z n ) ⋅ v 2 , (66) where n − m ≤ wt a − 1 n-m\leq\operatorname{wt}a-1 n − m ≤ wt a − 1 .
Proof. By Proposition 3.11, M 2 ⊙ − w , 0 M 1 M^2\mathbin{\odot_{-w,0}}M^1 M 2 ⊙ − w , 0 M 1 and M 1 ⊙ w , 0 M 2 M^1\mathbin{\odot_{w,0}}M^2 M 1 ⊙ w , 0 M 2 are finite-dimensional left A ( V ) A(V) A ( V ) -modules. Moreover, ( Φ L ⊣ Ω ) : Mod f d ( A ( V ) ) ⇄ C 1 ( V ) (\Phi^L\dashv\Omega):\operatorname{Mod}_{\mathrm{fd}}(A(V))\rightleftarrows\mathcal{C}_1(V) ( Φ L ⊣ Ω ) : Mod fd ( A ( V )) ⇄ C 1 ( V ) is an equivalence of categories. Let M 3 = Φ L ( M 1 ⊙ w , 0 M 2 ) M^3=\Phi^L(M^1\mathbin{\odot_{w,0}}M^2) M 3 = Φ L ( M 1 ⊙ w , 0 M 2 ) . Then M 3 ∈ C 1 ( V ) M^3\in\mathcal{C}_1(V) M 3 ∈ C 1 ( V ) is a finite direct sum of irreducible ordinary V V V -modules. Hence
M 3 ( 0 ) = Ω ( Φ L ( M 1 ⊙ w , 0 M 2 ) ) = M 1 ⊙ w , 0 M 2 . M^3(0)=\Omega\left(\Phi^L(M^1\mathbin{\odot_{w,0}}M^2)\right)=M^1\mathbin{\odot_{w,0}}M^2. M 3 ( 0 ) = Ω ( Φ L ( M 1 ⊙ w , 0 M 2 ) ) = M 1 ⊙ w , 0 M 2 . By Proposition 3.11, ( M 3 ) ′ ∈ C 1 ( V ) (M^3)'\in\mathcal{C}_1(V) ( M 3 ) ′ ∈ C 1 ( V ) , and so ( M 3 ) ′ ≅ Φ L ( Ω ( ( M 3 ) ′ ) ) = Φ L ( M 3 ( 0 ) ∗ ) (M^3)'\cong\Phi^L(\Omega((M^3)'))=\Phi^L(M^3(0)^*) ( M 3 ) ′ ≅ Φ L ( Ω (( M 3 ) ′ )) = Φ L ( M 3 ( 0 ) ∗ ) . Thus the hypotheses of Lemma 3.6 hold. By Proposition 4.4 there exists a canonical isomorphism
Ω 0 : Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 1 ⊙ w , 0 M 2 ) ≅ Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 1 ⊙ w , 0 M 2 ) , f ↦ f ~ . \Omega_0:\operatorname{Hom}_{A(V)}(M^1\mathbin{\odot_{w,0}}M^2,M^1\mathbin{\odot_{w,0}}M^2)\cong\operatorname{Hom}_{A(V)}(M^2\mathbin{\odot_{-w,0}}M^1,M^1\mathbin{\odot_{w,0}}M^2),
\qquad f\mapsto\widetilde{f}. Ω 0 : Hom A ( V ) ( M 1 ⊙ w , 0 M 2 , M 1 ⊙ w , 0 M 2 ) ≅ Hom A ( V ) ( M 2 ⊙ − w , 0 M 1 , M 1 ⊙ w , 0 M 2 ) , f ↦ f . Let f = Id f=\operatorname{Id} f = Id . Then Id : M 2 ⊙ − w , 0 M 1 → M 1 ⊙ w , 0 M 2 \operatorname{Id}:M^2\mathbin{\odot_{-w,0}}M^1\to M^1\mathbin{\odot_{w,0}}M^2 Id : M 2 ⊙ − w , 0 M 1 → M 1 ⊙ w , 0 M 2 is an A ( V ) A(V) A ( V ) -homomorphism. Moreover, for any ( u 3 ) ′ ∈ ( M 3 ( 0 ) ) ∗ = ( M 1 ⊙ w , 0 M 2 ) ∗ (u^3)'\in(M^3(0))^*=(M^1\mathbin{\odot_{w,0}}M^2)^* ( u 3 ) ′ ∈ ( M 3 ( 0 ) ) ∗ = ( M 1 ⊙ w , 0 M 2 ) ∗ , it follows from (60) that
\left\langle(u^3)'\mathrel{\left|}\widetilde{\operatorname{Id}}(v^2\mathbin{\odot_{-w}}v^1)\right.\right\rangle
=\left\langle(u^3)'\mathrel{\left|}\operatorname{Id}(v^1\mathbin{\odot_w}v^2)\right.\right\rangle
=\left\langle(u^3)'\mathrel{\left|}v^1\mathbin{\odot_w}v^2\right.\right\rangle.
Thus, Id ~ ( v 2 ⊙ − w v 1 ) = v 1 ⊙ w v 2 \widetilde{\operatorname{Id}}(v^2\mathbin{\odot_{-w}}v^1)=v^1\mathbin{\odot_w}v^2 Id ( v 2 ⊙ − w v 1 ) = v 1 ⊙ w v 2 , and R w = Id ~ R_w=\widetilde{\operatorname{Id}} R w = Id (64) is a homomorphism of left A ( V ) A(V) A ( V ) -modules. Finally, we may take M 3 = Φ L ( M 2 ⊙ − w , 0 M 1 ) M^3=\Phi^L(M^2\mathbin{\odot_{-w,0}}M^1) M 3 = Φ L ( M 2 ⊙ − w , 0 M 1 ) in diagram (59) and construct the inverse of R w R_w R w using the inverse of Ω 0 \Omega_0 Ω 0 . Hence R w R_w R w is an isomorphism of A ( V ) A(V) A ( V ) -modules.
Finally, to show (65) and (66) , we let M 3 = Φ L ( M 1 ⊙ w , 0 M 2 ) M^3=\Phi^L(M^1\mathbin{\odot_{w,0}}M^2) M 3 = Φ L ( M 1 ⊙ w , 0 M 2 ) and f = Id f=\operatorname{Id} f = Id in Lemma 4.5. Then (65) follows from (62) , (66) follows from (63) . Naturality follows directly from the defining formula (64) . Indeed, for V V V -module homomorphisms α : M 1 → N 1 \alpha:M^1\to N^1 α : M 1 → N 1 and β : M 2 → N 2 \beta:M^2\to N^2 β : M 2 → N 2 , we have
( α ⊙ w , 0 β ) ∘ R w M 1 , M 2 = R w N 1 , N 2 ∘ ( β ⊙ − w , 0 α ) , (\alpha\mathbin{\odot_{w,0}}\beta)\circ R_w^{M^1,M^2}=R_w^{N^1,N^2}\circ(\beta\mathbin{\odot_{-w,0}}\alpha), ( α ⊙ w , 0 β ) ∘ R w M 1 , M 2 = R w N 1 , N 2 ∘ ( β ⊙ − w , 0 α ) , since both sides send v 2 ⊙ − w v 1 v^2\mathbin{\odot_{-w}}v^1 v 2 ⊙ − w v 1 to α ( v 1 ) ⊙ w β ( v 2 ) \alpha(v^1)\mathbin{\odot_w}\beta(v^2) α ( v 1 ) ⊙ w β ( v 2 ) . □
Remark 4.7. In particular, by (33) , we have
R w ( ρ − w , 0 ( a ⊗ u − m ( u + w ) − n ) ( v 2 ⊙ − w v 1 ) ) = ρ w , 0 ( a ⊗ z n ( z − w ) m ) ( v 1 ⊙ w v 2 ) . (67) R_w\left(\rho_{-w,0}\left(a \otimes\frac{u^{-m}}{(u+w)^{-n}}\right)\left(v^2 \odot_{-w} v^1\right)\right)=\rho_{w,0}\left(a \otimes\frac{z^n}{(z-w)^m}\right)\left(v^1 \odot_w v^2\right). \tag*{(67)} R w ( ρ − w , 0 ( a ⊗ ( u + w ) − n u − m ) ( v 2 ⊙ − w v 1 ) ) = ρ w , 0 ( a ⊗ ( z − w ) m z n ) ( v 1 ⊙ w v 2 ) . (67)
If we take n = w t a − 1 n = \mathrm{wt}a - 1 n = wt a − 1 and m = 0 m = 0 m = 0 in (67) , then by (38) we have
R w ( [ a ] . ( v 2 ⊙ − w v 1 ) ) = R w ( ρ − w , 0 ( a ⊗ 1 ( u + w ) − w t a + 1 ) . ( v 2 ⊙ − w v 1 ) ) = ρ w , 0 ( a ⊗ z w t a − 1 ) . ( v 1 ⊙ w v 2 ) = [ a ] . R w ( v 2 ⊙ − w v 1 ) , \begin{aligned}
R_w([a].(v^2 \odot_{-w} v^1)) &= R_w\left(\rho_{-w,0}\left(a \otimes\frac{1}{(u+w)^{-\mathrm{wt}a+1}}\right).(v^2 \odot_{-w} v^1)\right) \\
&= \rho_{w,0}\left(a \otimes z^{\mathrm{wt}a-1}\right).(v^1 \odot_w v^2) \\
&= [a].R_w(v^2 \odot_{-w} v^1),
\end{aligned} R w ([ a ] . ( v 2 ⊙ − w v 1 )) = R w ( ρ − w , 0 ( a ⊗ ( u + w ) − wt a + 1 1 ) . ( v 2 ⊙ − w v 1 ) ) = ρ w , 0 ( a ⊗ z wt a − 1 ) . ( v 1 ⊙ w v 2 ) = [ a ] . R w ( v 2 ⊙ − w v 1 ) ,
for any [ a ] ∈ A ( V ) [a] \in A(V) [ a ] ∈ A ( V ) . This gives an alternative way of showing the A ( V ) A(V) A ( V ) -linearity of R w R_w R w (64) .
Parallel transport of three-point conformal blocks
Given two distinct points x , y ∈ P 1 ∖ { ∞ , 0 } x,y \in\mathbb{P}^1 \setminus\{\infty,0\} x , y ∈ P 1 ∖ { ∞ , 0 } , there are associated spaces of three-point conformal blocks
C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) and C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , y , 0 ) . \mathcal{C}\left((M^3)^\prime\otimes M^2 \otimes M^1\right)_{(\mathbb{P}^1,\infty,x,0)}
\quad\text{and}\quad
\mathcal{C}\left((M^3)^\prime\otimes M^2 \otimes M^1\right)_{(\mathbb{P}^1,\infty,y,0)}. C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) and C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , y , 0 ) .
We first prove that they are isomorphic as vector spaces using the idea of parallel transport isomorphism by Huang-Lepowsky-Zhang [39 ] .
Without loss of generality, we assume y = 1 y = 1 y = 1 . Note that fixing a determination of log ( x ) \log(x) log ( x ) with the normalization log ( 1 ) = 0 \log(1) = 0 log ( 1 ) = 0 amounts to choosing a homotopy class of paths with fixed endpoints:
γ : [ 0 , 1 ] → C × , γ ( 0 ) = 1 , γ ( 1 ) = x . \gamma: [0,1] \to\mathbb{C}^{\times}, \qquad\gamma(0) = 1,\ \gamma(1) = x. γ : [ 0 , 1 ] → C × , γ ( 0 ) = 1 , γ ( 1 ) = x .
Denote by l γ ( x ) l_\gamma(x) l γ ( x ) the logarithm of x x x determined by γ \gamma γ , with l γ ( 1 ) = 0 l_\gamma(1) = 0 l γ ( 1 ) = 0 .
Example 4.8. Let x = − 1 x = -1 x = − 1 . There are two standard paths from 1 1 1 to − 1 -1 − 1 . One along the upper-half unit circle, the other along the lower-half unit circle, see Figure 3 . The branch cut for the logarithm defined by γ + \gamma_+ γ + could be chosen to the negative y y y -axis; while the branch cut for the logarithm defined by γ − \gamma_- γ − could be chosen to the positive y y y -axis. Then l γ + ( − 1 ) = π i l_{\gamma_+}(-1) = \pi i l γ + ( − 1 ) = π i and l γ − ( − 1 ) = − π i l_{\gamma_-}(-1) = -\pi i l γ − ( − 1 ) = − π i .
Figure 3. Paths γ + \gamma_+ γ + and γ − \gamma_- γ − from 1 to − 1 -1 − 1
Similar to Remark 4.3, by fixing the marked points at ∞ \infty ∞ and 0 0 0 , we obtain a sheaf of conformal blocks C 0 \mathcal{C}_0 C 0 over C × \mathbb{C}^{\times} C × , whose fiber at z z z is
C 0 ∣ z = C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , z , 0 ) . \mathcal{C}_0|_z = \mathcal{C}\left((M^3)^\prime\otimes M^2 \otimes M^1\right)_{(\mathbb{P}^1,\infty,z,0)}. C 0 ∣ z = C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , z , 0 ) .
Here C × \mathbb{C}^{\times} C × is identified with the slice { ( z , 0 ) : z ≠ 0 } ⊂ Conf 2 ( A 1 ) \{(z,0):z\ne0\}\subset\operatorname{Conf}_{2}(\mathbb{A}^{1}) {( z , 0 ) : z = 0 } ⊂ Conf 2 ( A 1 ) . The isomorphisms T γ ± T_{\gamma_{\pm}} T γ ± in (69) are parallel transport for the standard flat connection along the two reversed semicircular paths γ ± − 1 \gamma_{\pm}^{-1} γ ± − 1 from − 1 -1 − 1 to 1 1 1 in Figure 3 .
Lemma 4.9. Fix a path γ : 1 ∼ x \gamma:1\sim x γ : 1 ∼ x in C × \mathbb{C}^{\times} C × . Let l γ ( x ) l_{\gamma}(x) l γ ( x ) be the logarithm of x x x uniquely determined by γ \gamma γ and l γ ( 1 ) = 0 l_{\gamma}(1)=0 l γ ( 1 ) = 0 . Let M 1 , M 2 , M 3 M^{1},M^{2},M^{3} M 1 , M 2 , M 3 be ordinary V V V -modules. With respect to the action (23) of the chiral Lie algebras on ( M 3 ) ′ ⊗ M 2 ⊗ M 1 (M^{3})^{\prime}\otimes M^{2}\otimes M^{1} ( M 3 ) ′ ⊗ M 2 ⊗ M 1 , we have
x wt a − n + m − 1 ⋅ ( ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ∘ ρ ∞ , x , 0 ( a ⊗ z n ( z − x ) m ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) = ( ρ ∞ , 1 , 0 ( a ⊗ z n ( z − 1 ) m ) ∘ ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) . (68) x^{\operatorname{wt}a-n+m-1}\cdot\left(\left(x^{L(0)}\otimes x^{-L(0)}\otimes x^{-L(0)}\right)\circ\rho_{\infty,x,0}\left(a\otimes\frac{z^{n}}{(z-x)^{m}}\right)\right)\cdot\left((v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right)
=
\left(\rho_{\infty,1,0}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\circ\left(x^{L(0)}\otimes x^{-L(0)}\otimes x^{-L(0)}\right)\right)\cdot\left((v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right). \tag*{(68)} x wt a − n + m − 1 ⋅ ( ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ∘ ρ ∞ , x , 0 ( a ⊗ ( z − x ) m z n ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) = ( ρ ∞ , 1 , 0 ( a ⊗ ( z − 1 ) m z n ) ∘ ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) . (68) where x L ( 0 ) = e l γ ( x ) L ( 0 ) = ∑ j ≥ 0 1 j ! ( l γ ( x ) ) j L ( 0 ) j x^{L(0)}=e^{l_{\gamma}(x)L(0)}=\sum_{j\geq0}\frac{1}{j!}\left(l_{\gamma}(x)\right)^{j}L(0)^{j} x L ( 0 ) = e l γ ( x ) L ( 0 ) = ∑ j ≥ 0 j ! 1 ( l γ ( x ) ) j L ( 0 ) j as an operator on ordinary modules, and ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ∈ ( M 3 ) ′ ⊗ M 2 ⊗ M 1 (v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\in(M^{3})^{\prime}\otimes M^{2}\otimes M^{1} ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ∈ ( M 3 ) ′ ⊗ M 2 ⊗ M 1 .
Proof. First, we note that x L ( 0 ) x^{L(0)} x L ( 0 ) is a well-defined operator on an ordinary module M M M . Indeed, given any v ∈ M λ v\in M_{\lambda} v ∈ M λ , with λ ∈ C \lambda\in\mathbb{C} λ ∈ C , we have x L ( 0 ) ⋅ v = ∑ j ≥ 0 1 j ! ( l γ ( x ) λ ) j ⋅ v x^{L(0)}\cdot v=\sum_{j\geq0}\frac{1}{j!}\left(l_{\gamma}(x)\lambda\right)^{j}\cdot v x L ( 0 ) ⋅ v = ∑ j ≥ 0 j ! 1 ( l γ ( x ) λ ) j ⋅ v , wherein the coefficient is a convergent series. Since M M M is a direct sum of M λ M_{\lambda} M λ ’s. Hence x L ( 0 ) x^{L(0)} x L ( 0 ) is well-defined. Furthermore, the operator x L ( 0 ) x^{L(0)} x L ( 0 ) is invertible, with inverse x − L ( 0 ) x^{-L(0)} x − L ( 0 ) .
With the notation (20) , we have [ L ( 0 ) , a ( n − m − j ) ] = ( − wt a + n − m − j + 1 ) a ( n − m − j ) [L(0),a_{(n-m-j)}]=(-\operatorname{wt}a+n-m-j+1)a_{(n-m-j)} [ L ( 0 ) , a ( n − m − j ) ] = ( − wt a + n − m − j + 1 ) a ( n − m − j ) . Since M i M^{i} M i are ordinary V V V -modules, we may choose ( v 3 ) ′ ∈ ( M 3 ) ′ (v^{3})^{\prime}\in(M^{3})^{\prime} ( v 3 ) ′ ∈ ( M 3 ) ′ , v 1 ∈ M 1 v^{1}\in M^{1} v 1 ∈ M 1 , and v 2 ∈ M 2 v^{2}\in M^{2} v 2 ∈ M 2 to be L ( 0 ) L(0) L ( 0 ) -eigenvectors. It follows from (23) that
x wt a − n + m − 1 ⋅ ( ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ∘ ρ ∞ , x , 0 ( a ⊗ z n ( z − x ) m ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) = − ∑ j ≥ 0 ( − m j ) ( − 1 ) j x wt a − n + m − 1 + j x L ( 0 ) ⋅ ( a ( n − m − j ) ′ ( v 3 ) ′ ) ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 + ∑ j ≥ 0 ( n j ) x wt a − n + m − 1 + n − j x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ ( a ( j − m ) v 2 ) ⊗ x − L ( 0 ) ⋅ v 1 + ∑ j ≥ 0 ( − m j ) ( − 1 ) − m − j x wt a − n + m − 1 − m − j x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ ( a ( n + j ) v 1 ) = − ∑ j ≥ 0 ( − m j ) ( − 1 ) j x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ a ( n − m − j ) ′ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 + ∑ j ≥ 0 ( n j ) x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ ( v 3 ) ′ ⊗ a ( j − m ) v 2 ⊗ v 1 + ∑ j ≥ 0 ( − m j ) ( − 1 ) − m − j x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ ( v 3 ) ′ ⊗ v 2 ⊗ a ( n + j ) v 1 = x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ ( ρ ∞ , 1 , 0 ( a ⊗ z n ( z − 1 ) m ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) ) = ( ρ ∞ , 1 , 0 ( a ⊗ z n ( z − 1 ) m ) ∘ ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) , \begin{aligned}
&x^{\operatorname{wt}a-n+m-1}\cdot\left(\left(x^{L(0)}\otimes x^{-L(0)}\otimes x^{-L(0)}\right)\circ\rho_{\infty,x,0}\left(a\otimes\frac{z^{n}}{(z-x)^{m}}\right)\right)\cdot\left((v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right)\\
&=-\sum_{j\geq0}\binom{-m}{j}(-1)^{j}x^{\operatorname{wt}a-n+m-1+j}x^{L(0)}\cdot\left(a'_{(n-m-j)}(v^{3})^{\prime}\right)\otimes x^{-L(0)}\cdot v^{2}\otimes x^{-L(0)}\cdot v^{1}\\
&\quad+\sum_{j\geq0}\binom{n}{j}x^{\operatorname{wt}a-n+m-1+n-j}x^{L(0)}\cdot(v^{3})^{\prime}\otimes x^{-L(0)}\cdot\left(a_{(j-m)}v^{2}\right)\otimes x^{-L(0)}\cdot v^{1}\\
&\quad+\sum_{j\geq0}\binom{-m}{j}(-1)^{-m-j}x^{\operatorname{wt}a-n+m-1-m-j}x^{L(0)}\cdot(v^{3})^{\prime}\otimes x^{-L(0)}\cdot v^{2}\otimes x^{-L(0)}\cdot\left(a_{(n+j)}v^{1}\right)\\
&=-\sum_{j\geq0}\binom{-m}{j}(-1)^{j}x^{\operatorname{wt}(v^{3})^{\prime}-\operatorname{wt}v^{2}-\operatorname{wt}v^{1}}\cdot a'_{(n-m-j)}(v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\\
&\quad+\sum_{j\geq0}\binom{n}{j}x^{\operatorname{wt}(v^{3})^{\prime}-\operatorname{wt}v^{2}-\operatorname{wt}v^{1}}\cdot(v^{3})^{\prime}\otimes a_{(j-m)}v^{2}\otimes v^{1}\\
&\quad+\sum_{j\geq0}\binom{-m}{j}(-1)^{-m-j}x^{\operatorname{wt}(v^{3})^{\prime}-\operatorname{wt}v^{2}-\operatorname{wt}v^{1}}\cdot(v^{3})^{\prime}\otimes v^{2}\otimes a_{(n+j)}v^{1}\\
&=x^{\operatorname{wt}(v^{3})^{\prime}-\operatorname{wt}v^{2}-\operatorname{wt}v^{1}}\cdot\left(\rho_{\infty,1,0}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\cdot\left((v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right)\right)\\
&=\left(\rho_{\infty,1,0}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\circ\left(x^{L(0)}\otimes x^{-L(0)}\otimes x^{-L(0)}\right)\right)\cdot\left((v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right),
\end{aligned} x wt a − n + m − 1 ⋅ ( ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ∘ ρ ∞ , x , 0 ( a ⊗ ( z − x ) m z n ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) = − j ≥ 0 ∑ ( j − m ) ( − 1 ) j x wt a − n + m − 1 + j x L ( 0 ) ⋅ ( a ( n − m − j ) ′ ( v 3 ) ′ ) ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 + j ≥ 0 ∑ ( j n ) x wt a − n + m − 1 + n − j x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ ( a ( j − m ) v 2 ) ⊗ x − L ( 0 ) ⋅ v 1 + j ≥ 0 ∑ ( j − m ) ( − 1 ) − m − j x wt a − n + m − 1 − m − j x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ ( a ( n + j ) v 1 ) = − j ≥ 0 ∑ ( j − m ) ( − 1 ) j x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ a ( n − m − j ) ′ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 + j ≥ 0 ∑ ( j n ) x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ ( v 3 ) ′ ⊗ a ( j − m ) v 2 ⊗ v 1 + j ≥ 0 ∑ ( j − m ) ( − 1 ) − m − j x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ ( v 3 ) ′ ⊗ v 2 ⊗ a ( n + j ) v 1 = x wt ( v 3 ) ′ − wt v 2 − wt v 1 ⋅ ( ρ ∞ , 1 , 0 ( a ⊗ ( z − 1 ) m z n ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) ) = ( ρ ∞ , 1 , 0 ( a ⊗ ( z − 1 ) m z n ) ∘ ( x L ( 0 ) ⊗ x − L ( 0 ) ⊗ x − L ( 0 ) ) ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) ,
where x α x^{\alpha} x α is defined as e α l γ ( x ) e^{\alpha l_{\gamma}(x)} e α l γ ( x ) for any complex number α ∈ C \alpha\in\mathbb{C} α ∈ C . □
Lemma 4.10. Fix a path γ : 1 ∼ x \gamma:1\sim x γ : 1 ∼ x in C × \mathbb{C}^{\times} C × . Let l γ ( x ) l_{\gamma}(x) l γ ( x ) be the logarithm of x x x uniquely determined by γ \gamma γ and l γ ( 1 ) = 0 l_{\gamma}(1)=0 l γ ( 1 ) = 0 . Then there exists an isomorphism of conformal blocks
T γ : C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) ⟶ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) , φ ↦ φ γ , T_{\gamma}:\mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,x,0)}\longrightarrow\mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,1,0)},\quad\varphi\mapsto\varphi_{\gamma}, T γ : C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) ⟶ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) , φ ↦ φ γ , ⟨ φ γ ∣ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ ∣ x − L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x L ( 0 ) ⋅ v 2 ⊗ x L ( 0 ) ⋅ v 1 ⟩ (69) \left\langle\varphi_{\gamma}\mid(v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right\rangle=\left\langle\varphi\mid x^{-L(0)}\cdot(v^{3})^{\prime}\otimes x^{L(0)}\cdot v^{2}\otimes x^{L(0)}\cdot v^{1}\right\rangle \tag*{(69)} ⟨ φ γ ∣ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ ∣ x − L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x L ( 0 ) ⋅ v 2 ⊗ x L ( 0 ) ⋅ v 1 ⟩ (69) is an isomorphism of three-point conformal blocks on P 1 \mathbb{P}^{1} P 1 , where x L ( 0 ) = e l γ ( x ) L ( 0 ) x^{L(0)}=e^{l_{\gamma}(x)L(0)} x L ( 0 ) = e l γ ( x ) L ( 0 ) .
Proof. First, we show that φ γ \varphi_{\gamma} φ γ is invariant under the action of L P 1 \ { ∞ , 1 , 0 } ( V ) \mathcal{L}_{\mathbb{P}^{1}\backslash\{\infty,1,0\}}(V) L P 1 \ { ∞ , 1 , 0 } ( V ) . i.e., T γ T_{\gamma} T γ is well-defined. It follows from (68) and (69) that
⟨ φ γ | ρ ∞ , 1 , 0 ( a ⊗ z n ( z − 1 ) m ) ⋅ ( x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 ) ⟩ = ⟨ φ | ( ( x − L ( 0 ) ⊗ x L ( 0 ) ⊗ x L ( 0 ) ) ∘ ρ ∞ , 1 , 0 ( a ⊗ z n ( z − 1 ) m ) ) ⋅ ( x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 ) ⟩ = x w t a − n + m − 1 ⟨ φ | ρ ∞ , x , 0 ( a ⊗ z n ( z − x ) m ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) ⟩ = 0 , \begin{aligned}
\left\langle\varphi_{\gamma}\middle|\rho_{\infty,1,0}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\cdot\left(x^{L(0)}\cdot(v^{3})^{\prime}\otimes x^{-L(0)}\cdot v^{2}\otimes x^{-L(0)}\cdot v^{1}\right)\right\rangle\\
&=\left\langle\varphi\middle|\left(\left(x^{-L(0)}\otimes x^{L(0)}\otimes x^{L(0)}\right)\circ\rho_{\infty,1,0}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\right)\cdot\left(x^{L(0)}\cdot(v^{3})^{\prime}\otimes x^{-L(0)}\cdot v^{2}\otimes x^{-L(0)}\cdot v^{1}\right)\right\rangle\\
&=x^{\mathrm{wt}a-n+m-1}\left\langle\varphi\middle|\rho_{\infty,x,0}\left(a\otimes\frac{z^{n}}{(z-x)^{m}}\right)\cdot\left((v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right)\right\rangle\\
&=0,
\end{aligned} ⟨ φ γ ρ ∞ , 1 , 0 ( a ⊗ ( z − 1 ) m z n ) ⋅ ( x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 ) ⟩ = ⟨ φ ( ( x − L ( 0 ) ⊗ x L ( 0 ) ⊗ x L ( 0 ) ) ∘ ρ ∞ , 1 , 0 ( a ⊗ ( z − 1 ) m z n ) ) ⋅ ( x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 ) ⟩ = x wt a − n + m − 1 ⟨ φ ρ ∞ , x , 0 ( a ⊗ ( z − x ) m z n ) ⋅ ( ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ) ⟩ = 0 ,
as φ ∈ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) \varphi\in\mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,x,0)} φ ∈ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) is invariant under L P 1 \ { ∞ , x , 0 } ( V ) \mathcal{L}_{\mathbb{P}^{1}\backslash\{\infty,x,0\}}(V) L P 1 \ { ∞ , x , 0 } ( V ) . Moreover, since x L ( 0 ) x^{L(0)} x L ( 0 ) is an invertible operator on M i M^{i} M i , this shows φ γ \varphi_{\gamma} φ γ vanishes on L P 1 \ { ∞ , 1 , 0 } ( V ) ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) \mathcal{L}_{\mathbb{P}^{1}\backslash\{\infty,1,0\}}(V)\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right) L P 1 \ { ∞ , 1 , 0 } ( V ) ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) .
It is clear that the inverse of T γ T_{\gamma} T γ is given by
T γ − 1 : C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) ⟶ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) , φ ↦ φ γ − 1 , T_{\gamma}^{-1}:\mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,1,0)}\longrightarrow\mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,x,0)},\quad\varphi\mapsto\varphi_{\gamma}^{-1}, T γ − 1 : C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) ⟶ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) , φ ↦ φ γ − 1 ,
⟨ φ γ − 1 ∣ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ ∣ x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 ⟩ (70) \left\langle\varphi_{\gamma}^{-1}\mid(v^{3})^{\prime}\otimes v^{2}\otimes v^{1}\right\rangle=\left\langle\varphi\mid x^{L(0)}\cdot(v^{3})^{\prime}\otimes x^{-L(0)}\cdot v^{2}\otimes x^{-L(0)}\cdot v^{1}\right\rangle \tag*{(70)} ⟨ φ γ − 1 ∣ ( v 3 ) ′ ⊗ v 2 ⊗ v 1 ⟩ = ⟨ φ ∣ x L ( 0 ) ⋅ ( v 3 ) ′ ⊗ x − L ( 0 ) ⋅ v 2 ⊗ x − L ( 0 ) ⋅ v 1 ⟩ (70)
The well-definedness of T γ − 1 T_{\gamma}^{-1} T γ − 1 can be proved similarly using (68) . □
Proposition 4.11. Let M 1 , M 2 , M 3 ∈ C 1 ( V ) M^{1},M^{2},M^{3}\in\mathcal{C}_{1}(V) M 1 , M 2 , M 3 ∈ C 1 ( V ) . Then there exists a linear isomorphism which is natural in M 1 , M 2 M^{1},M^{2} M 1 , M 2 :
T γ : Hom A ( V ) ( M 2 ⊙ x , 0 M 1 , M 3 ( 0 ) ) ≅ Hom A ( V ) ( M 2 ⊙ 1 , 0 M 1 , M 3 ( 0 ) ) , f ↦ f γ , T_{\gamma}:\operatorname{Hom}_{A(V)}(M^{2}\mathbin{\odot_{x,0}}M^{1},M^{3}(0))\cong\operatorname{Hom}_{A(V)}(M^{2}\mathbin{\odot_{1,0}}M^{1},M^{3}(0)),\quad f\mapsto f_{\gamma}, T γ : Hom A ( V ) ( M 2 ⊙ x , 0 M 1 , M 3 ( 0 )) ≅ Hom A ( V ) ( M 2 ⊙ 1 , 0 M 1 , M 3 ( 0 )) , f ↦ f γ , f γ ( v 2 ⊙ 1 v 1 ) = x − L ( 0 ) ⋅ f ( x L ( 0 ) ⋅ v 2 ⊙ x x L ( 0 ) ⋅ v 1 ) , (71) f_{\gamma}(v^{2}\mathbin{\odot_{1}}v^{1})=x^{-L(0)}\cdot f\left(x^{L(0)}\cdot v^{2}\mathbin{\odot_{x}}x^{L(0)}\cdot v^{1}\right), \tag*{(71)} f γ ( v 2 ⊙ 1 v 1 ) = x − L ( 0 ) ⋅ f ( x L ( 0 ) ⋅ v 2 ⊙ x x L ( 0 ) ⋅ v 1 ) , (71) for all ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^{3})^{\prime}\in M^{3}(0)^{*} ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ , v 1 ∈ M 1 v^{1}\in M^{1} v 1 ∈ M 1 , and v 2 ∈ M 2 v^{2}\in M^{2} v 2 ∈ M 2 .
Proof. Similar to Proposition 4.4, using Lemma 3.6 and Corollary 3.10 again, we can extend the isomorphism T γ T_{\gamma} T γ to the following commutative diagram:
C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) → T γ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) ( 3.8 ) ↓ ≅ ( 3.8 ) ↓ ≅ C r e s ( M 3 ( 0 ) ∗ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) → T γ C r e s ( M 3 ( 0 ) ∗ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) ( 3.14 ) ↓ ≅ ( 3.14 ) ↓ ≅ Hom A ( V ) ( M 2 ⊙ x , 0 M 1 , M 3 ( 0 ) ) → T γ Hom A ( V ) ( M 2 ⊙ 1 , 0 M 1 , M 3 ( 0 ) ) , (72) \begin{CD}
\mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,x,0)} @>{T_{\gamma}}>> \mathcal{C}\left((M^{3})^{\prime}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,1,0)} \\
@V{(3.8)}V{\cong}V @V{(3.8)}V{\cong}V \\
\mathcal{C}_{\mathrm{res}}\left(M^{3}(0)^{*}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,x,0)} @>{T_{\gamma}}>> \mathcal{C}_{\mathrm{res}}\left(M^{3}(0)^{*}\otimes M^{2}\otimes M^{1}\right)_{(\mathbb{P}^{1},\infty,1,0)} \\
@V{(3.14)}V{\cong}V @V{(3.14)}V{\cong}V \\
\operatorname{Hom}_{A(V)}(M^{2}\mathbin{\odot_{x,0}}M^{1},M^{3}(0)) @>{T_{\gamma}}>> \operatorname{Hom}_{A(V)}(M^{2}\mathbin{\odot_{1,0}}M^{1},M^{3}(0)), \tag*{(72)}
\end{CD} C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) ( 3.8 ) ↓ ⏐ ≅ C res ( M 3 ( 0 ) ∗ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) ( 3.14 ) ↓ ⏐ ≅ Hom A ( V ) ( M 2 ⊙ x , 0 M 1 , M 3 ( 0 )) T γ T γ T γ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) ( 3.8 ) ↓ ⏐ ≅ C res ( M 3 ( 0 ) ∗ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , 1 , 0 ) ( 3.14 ) ↓ ⏐ ≅ Hom A ( V ) ( M 2 ⊙ 1 , 0 M 1 , M 3 ( 0 )) , (72)
where the vertical maps are given by (34) and (40) . Given φ ∈ C ( ( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) \varphi\in\mathcal{C}((M^{3})' \otimes M^{2} \otimes M^{1})_{(P_{1},\infty,x,0)} φ ∈ C (( M 3 ) ′ ⊗ M 2 ⊗ M 1 ) ( P 1 , ∞ , x , 0 ) , the corresponding elements under the isomorphisms in (54) are denoted by
\begin{CD}
\varphi@>{T_{\gamma}}>> \varphi_{\gamma}\\
@V{(3.8)}VV @VV{(3.8)}V\\
\varphi_{\mathrm{res}} @>{T_{\gamma}}>> \varphi_{\mathrm{res},\gamma}\\
@V{(3.14)}VV @VV{(3.14)}V\\
f @>{T_{\gamma}}>> f_{\gamma}
\end{CD}
For ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ (u^{3})' \in M^{3}(0)^{*} ( u 3 ) ′ ∈ M 3 ( 0 ) ∗ and u 3 ∈ M 3 ( 0 ) u^{3} \in M^{3}(0) u 3 ∈ M 3 ( 0 ) , we have ⟨ L ( 0 ) ( u 3 ) ′ ∣ u 3 ⟩ = ⟨ ( u 3 ) ′ ∣ L ( 0 ) u 3 ⟩ \langle L(0)(u^{3})' \mid u^{3}\rangle= \langle(u^{3})' \mid L(0)u^{3}\rangle ⟨ L ( 0 ) ( u 3 ) ′ ∣ u 3 ⟩ = ⟨( u 3 ) ′ ∣ L ( 0 ) u 3 ⟩ . It follows from (69) that given any A ( V ) A(V) A ( V ) -homomorphism f : M 2 ⊙ x , 0 M 1 → M 3 ( 0 ) f:M^{2}\mathbin{\odot_{x,0}}M^{1}\to M^{3}(0) f : M 2 ⊙ x , 0 M 1 → M 3 ( 0 ) , the corresponding A ( V ) A(V) A ( V ) -homomorphism f γ = T γ ( f ) f_{\gamma}=T_{\gamma}(f) f γ = T γ ( f ) has the following expression:
⟨ ( u 3 ) ′ ∣ f γ ( v 2 ⊙ 1 v 1 ) ⟩ = ⟨ x − L ( 0 ) ( u 3 ) ′ ∣ f ( x L ( 0 ) v 2 ⊙ x x L ( 0 ) v 1 ) ⟩ = ⟨ ( u 3 ) ′ ∣ x − L ( 0 ) ⋅ f ( x L ( 0 ) v 2 ⊙ x x L ( 0 ) v 1 ) ⟩ , \begin{aligned}
\left\langle(u^{3})' \mid f_{\gamma}(v^{2}\mathbin{\odot_{1}}v^{1})\right\rangle
=
\left\langle x^{-L(0)}(u^{3})' \mid f\left(x^{L(0)}v^{2}\mathbin{\odot_{x}}x^{L(0)}v^{1}\right)\right\rangle\\
&=
\left\langle(u^{3})' \mid x^{-L(0)}\cdot f\left(x^{L(0)}v^{2}\mathbin{\odot_{x}}x^{L(0)}v^{1}\right)\right\rangle,
\end{aligned} ⟨ ( u 3 ) ′ ∣ f γ ( v 2 ⊙ 1 v 1 ) ⟩ = ⟨ x − L ( 0 ) ( u 3 ) ′ ∣ f ( x L ( 0 ) v 2 ⊙ x x L ( 0 ) v 1 ) ⟩ = ⟨ ( u 3 ) ′ ∣ x − L ( 0 ) ⋅ f ( x L ( 0 ) v 2 ⊙ x x L ( 0 ) v 1 ) ⟩ ,
where x L ( 0 ) = e l γ ( x ) L ( 0 ) x^{L(0)}=e^{l_{\gamma}(x)L(0)} x L ( 0 ) = e l γ ( x ) L ( 0 ) . This proves (71) . Finally, for α : N 1 → M 1 \alpha:N^{1}\to M^{1} α : N 1 → M 1 and β : N 2 → M 2 \beta:N^{2}\to M^{2} β : N 2 → M 2 , it follows immediately from (71) that
T γ N 1 , N 2 ( f ∘ ( β ⊙ x , 0 α ) ) = T γ M 1 , M 2 ( f ) ∘ ( β ⊙ 1 , 0 α ) . T_{\gamma}^{N^{1},N^{2}}\left(f\circ(\beta\mathbin{\odot_{x,0}}\alpha)\right)
=
T_{\gamma}^{M^{1},M^{2}}(f)\circ(\beta\mathbin{\odot_{1,0}}\alpha). T γ N 1 , N 2 ( f ∘ ( β ⊙ x , 0 α ) ) = T γ M 1 , M 2 ( f ) ∘ ( β ⊙ 1 , 0 α ) .
The naturality of T γ T_{\gamma} T γ is clear. □ \square □
Construction of R + M 2 , M 1 R^{M^{2},M^{1}}_{+} R + M 2 , M 1 and R − M 2 , M 1 R^{M^{2},M^{1}}_{-} R − M 2 , M 1 for the ⊙ \odot ⊙ -tensor product
We adopted the following notation in Definition 3.7:
M ⊙ N = M ⊙ 1 , 0 N , and u ⊙ v = u ⊙ 1 v ∈ M ⊙ N . M\odot N=M\mathbin{\odot_{1,0}}N,\qquad\text{and}\qquad u\odot v=u\mathbin{\odot_{1}}v\in M\odot N. M ⊙ N = M ⊙ 1 , 0 N , and u ⊙ v = u ⊙ 1 v ∈ M ⊙ N .
By Proposition 3.11, every object S S S of Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) is a finite direct sum of irreducible modules S = ⨁ i S i S=\bigoplus_i S^i S = ⨁ i S i , with S i ≅ W i ( 0 ) S^i\cong W^i(0) S i ≅ W i ( 0 ) for some irreducible W i ∈ C 1 ( V ) W^i\in\mathcal{C}_{1}(V) W i ∈ C 1 ( V ) . Note that [ ω ] [\omega] [ ω ] is a central element of A ( V ) A(V) A ( V ) [56 ] and acts on W i ( 0 ) W^i(0) W i ( 0 ) by h i ⋅ Id W i ( 0 ) h^i\cdot\operatorname{Id}_{W^i(0)} h i ⋅ Id W i ( 0 ) , where h i ∈ C h^i\in\mathbb{C} h i ∈ C is the conformal weight of W i W^i W i . Hence [ ω ] [\omega] [ ω ] acts semisimply on S S S , and we can write
e ± π i [ ω ] ∣ S = ∑ i e ± π i h i Id S i . (73) e^{\pm\pi i[\omega]}\big|_{S}=\sum_i e^{\pm\pi i h^i}\operatorname{Id}_{S^i}. \tag*{(73)} e ± π i [ ω ] S = i ∑ e ± π i h i Id S i . (73)
Since [ ω ] [\omega] [ ω ] is central, these are mutually inverse A ( V ) A(V) A ( V ) -linear automorphisms of S S S , which are natural in S S S .
Theorem 4.12. Let M 1 , M 2 ∈ C 1 ( V ) M^{1},M^{2}\in\mathcal{C}_{1}(V) M 1 , M 2 ∈ C 1 ( V ) . Then there exists isomorphisms of left A ( V ) A(V) A ( V ) -modules which are natural in M 1 , M 2 M^{1},M^{2} M 1 , M 2 :
R ± M 2 , M 1 : M 2 ⊙ M 1 ⟶ M 1 ⊙ M 2 , R_{\pm}^{M^{2},M^{1}}:M^{2}\odot M^{1}\longrightarrow M^{1}\odot M^{2}, R ± M 2 , M 1 : M 2 ⊙ M 1 ⟶ M 1 ⊙ M 2 , R ± M 2 , M 1 ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ⋅ ( e ∓ π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) , (74) R_{\pm}^{M^{2},M^{1}}(v^{2}\odot v^{1})
=
e^{\mp\pi iL(0)}\cdot\left(e^{\mp\pi iL(0)}v^{1}\odot e^{\pm\pi iL(0)}v^{2}\right), \tag*{(74)} R ± M 2 , M 1 ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ⋅ ( e ∓ π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) , (74) where the outer factor e ∓ π i L ( 0 ) e^{\mp\pi iL(0)} e ∓ π i L ( 0 ) acts on M 1 ⊙ M 2 M^{1}\odot M^{2} M 1 ⊙ M 2 through e ∓ π i [ ω ] e^{\mp\pi i[\omega]} e ∓ π i [ ω ] (73) . Moreover,
( R − M 1 , M 2 ) − 1 = R + M 2 , M 1 , ( R + M 1 , M 2 ) − 1 = R − M 2 , M 1 . (75) \left(R_{-}^{M^{1},M^{2}}\right)^{-1}=R_{+}^{M^{2},M^{1}},\qquad
\left(R_{+}^{M^{1},M^{2}}\right)^{-1}=R_{-}^{M^{2},M^{1}}. \tag*{(75)} ( R − M 1 , M 2 ) − 1 = R + M 2 , M 1 , ( R + M 1 , M 2 ) − 1 = R − M 2 , M 1 . (75)
First Proof of Theorem 4.12: using the natural identifications of Hom-spaces. Let γ ± : 1 ∼ − 1 \gamma_{\pm}:1\sim-1 γ ± : 1 ∼ − 1 be the paths in Example 4.8, so that l γ ± ( − 1 ) = ± π i l_{\gamma_{\pm}}(-1)=\pm\pi i l γ ± ( − 1 ) = ± π i . By Proposition 4.4 and Proposition 4.11, for any M 3 ∈ C 1 ( V ) M^{3}\in\mathcal{C}_{1}(V) M 3 ∈ C 1 ( V ) , we have a composition of isomorphisms
Hom A ( V ) ( M 1 ⊙ 1 , 0 M 2 , M 3 ( 0 ) ) → Ω 0 Hom A ( V ) ( M 2 ⊙ − 1 , 0 M 1 , M 3 ( 0 ) ) \operatorname{Hom}_{A(V)}\left(M^{1}\mathbin{\odot_{1,0}}M^{2},M^{3}(0)\right)
\xrightarrow{\ \Omega_{0}\ }
\operatorname{Hom}_{A(V)}\left(M^{2}\mathbin{\odot_{-1,0}}M^{1},M^{3}(0)\right) Hom A ( V ) ( M 1 ⊙ 1 , 0 M 2 , M 3 ( 0 ) ) Ω 0 Hom A ( V ) ( M 2 ⊙ − 1 , 0 M 1 , M 3 ( 0 ) )
→ T γ ± Hom A ( V ) ( M 2 ⊙ 1 , 0 M 1 , M 3 ( 0 ) ) , f ⟼ f ~ ⟼ f ~ γ ± . \xrightarrow{\ T_{\gamma_{\pm}}\ }
\operatorname{Hom}_{A(V)}\left(M^{2}\mathbin{\odot_{1,0}}M^{1},M^{3}(0)\right),
\qquad
f\longmapsto\widetilde{f}\longmapsto\widetilde{f}_{\gamma_{\pm}}. T γ ± Hom A ( V ) ( M 2 ⊙ 1 , 0 M 1 , M 3 ( 0 ) ) , f ⟼ f ⟼ f γ ± .
It follows from (60) and (71) that
f ~ γ ± ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ⋅ f ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) . (76) \widetilde{f}_{\gamma_{\pm}}(v^{2}\mathbin{\odot}v^{1})=e^{\mp\pi iL(0)}\cdot f\left(e^{\pm\pi iL(0)}v^{1}\mathbin{\odot}e^{\pm\pi iL(0)}v^{2}\right). \tag*{(76)} f γ ± ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ⋅ f ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) . (76)
Similar to the proof of Theorem 4.6 , since M 1 ⊙ 1 , 0 M 2 M^{1}\mathbin{\odot}_{1,0}M^{2} M 1 ⊙ 1 , 0 M 2 is a finite-dimensional object in the semisimple category Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) , by letting
M 3 = Φ L ( M 1 ⊙ 1 , 0 M 2 ) , with M 3 ( 0 ) = M 1 ⊙ 1 , 0 M 2 , M^{3}=\Phi^{L}(M^{1}\mathbin{\odot}_{1,0}M^{2}),\qquad\text{with}\quad M^{3}(0)=M^{1}\mathbin{\odot}_{1,0}M^{2}, M 3 = Φ L ( M 1 ⊙ 1 , 0 M 2 ) , with M 3 ( 0 ) = M 1 ⊙ 1 , 0 M 2 ,
the element f = Id ∈ Hom A ( V ) ( M 1 ⊙ 1 , 0 M 2 , M 1 ⊙ 1 , 0 M 2 ) f=\operatorname{Id}\in\operatorname{Hom}_{A(V)}(M^{1}\mathbin{\odot}_{1,0}M^{2},M^{1}\mathbin{\odot}_{1,0}M^{2}) f = Id ∈ Hom A ( V ) ( M 1 ⊙ 1 , 0 M 2 , M 1 ⊙ 1 , 0 M 2 ) gives rise to an isomorphism of A ( V ) A(V) A ( V ) -modules under Ω 0 \Omega_{0} Ω 0 :
R 1 = Id ~ : M 2 ⊙ − 1 , 0 M 1 ⟶ M 1 ⊙ 1 , 0 M 2 , R 1 ( v 2 ⊙ − 1 v 1 ) = v 1 ⊙ v 2 . R_{1}=\widetilde{\operatorname{Id}}:M^{2}\mathbin{\odot}_{-1,0}M^{1}\longrightarrow M^{1}\mathbin{\odot}_{1,0}M^{2},\qquad R_{1}(v^{2}\mathbin{\odot}_{-1}v^{1})=v^{1}\mathbin{\odot}v^{2}. R 1 = Id : M 2 ⊙ − 1 , 0 M 1 ⟶ M 1 ⊙ 1 , 0 M 2 , R 1 ( v 2 ⊙ − 1 v 1 ) = v 1 ⊙ v 2 .
Then by Proposition 4.11, together with (76) for f = Id f=\operatorname{Id} f = Id , we have homomorphisms of A ( V ) A(V) A ( V ) -modules by applying T γ ± T_{\gamma_{\pm}} T γ ±
R ± M 2 , M 1 : = T γ ± ( R 1 ) = Id ~ γ ± : M 2 ⊙ 1 , 0 M 1 ⟶ M 1 ⊙ 1 , 0 M 2 , R_{\pm}^{M^{2},M^{1}}:=T_{\gamma_{\pm}}(R_{1})=\widetilde{\operatorname{Id}}_{\gamma_{\pm}}:M^{2}\mathbin{\odot}_{1,0}M^{1}\longrightarrow M^{1}\mathbin{\odot}_{1,0}M^{2}, R ± M 2 , M 1 := T γ ± ( R 1 ) = Id γ ± : M 2 ⊙ 1 , 0 M 1 ⟶ M 1 ⊙ 1 , 0 M 2 ,
R ± M 2 , M 1 ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ⋅ ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) , R_{\pm}^{M^{2},M^{1}}(v^{2}\mathbin{\odot}v^{1})=e^{\mp\pi iL(0)}\cdot\left(e^{\pm\pi iL(0)}v^{1}\mathbin{\odot}e^{\pm\pi iL(0)}v^{2}\right), R ± M 2 , M 1 ( v 2 ⊙ v 1 ) = e ∓ π i L ( 0 ) ⋅ ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) ,
which is the formula (74) .
Let α : M 1 → N 1 \alpha:M^{1}\to N^{1} α : M 1 → N 1 and β : M 2 → N 2 \beta:M^{2}\to N^{2} β : M 2 → N 2 be V V V -module homomorphisms. Since α \alpha α and β \beta β commute with L ( 0 ) L(0) L ( 0 ) , and α ⊙ β \alpha\mathbin{\odot}\beta α ⊙ β is an A ( V ) A(V) A ( V ) -homomorphism, (74) gives
( α ⊙ β ) ∘ R ± M 2 , M 1 = R ± N 2 , N 1 ∘ ( β ⊙ α ) . (\alpha\mathbin{\odot}\beta)\circ R_{\pm}^{M^{2},M^{1}}=R_{\pm}^{N^{2},N^{1}}\circ(\beta\mathbin{\odot}\alpha). ( α ⊙ β ) ∘ R ± M 2 , M 1 = R ± N 2 , N 1 ∘ ( β ⊙ α ) .
Hence R ± M 2 , M 1 R_{\pm}^{M^{2},M^{1}} R ± M 2 , M 1 are natural in M 1 M^{1} M 1 and M 2 M^{2} M 2 .
Finally, we may interchange M 2 , M 1 M^{2},M^{1} M 2 , M 1 in (74) and obtain A ( V ) A(V) A ( V ) -homomorphisms R ± M 1 , M 2 : M 1 ⊙ M 2 → M 2 ⊙ M 1 R_{\pm}^{M^{1},M^{2}}:M^{1}\mathbin{\odot}M^{2}\to M^{2}\mathbin{\odot}M^{1} R ± M 1 , M 2 : M 1 ⊙ M 2 → M 2 ⊙ M 1 . Since R − M 1 , M 2 R_{-}^{M^{1},M^{2}} R − M 1 , M 2 commutes with the action of e − π i [ ω ] e^{-\pi i[\omega]} e − π i [ ω ] , we have
R − M 1 , M 2 ( R + M 2 , M 1 ( v 2 ⊙ v 1 ) ) = R − M 1 , M 2 ( e − π i [ ω ] ( e π i L ( 0 ) v 1 ⊙ e π i L ( 0 ) v 2 ) ) = e − π i [ ω ] R − M 1 , M 2 ( e π i L ( 0 ) v 1 ⊙ e π i L ( 0 ) v 2 ) = e − π i [ ω ] e π i [ ω ] ( e − π i L ( 0 ) e π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) e π i L ( 0 ) v 1 ) = v 2 ⊙ v 1 . \begin{aligned}
R_{-}^{M^{1},M^{2}}\left(R_{+}^{M^{2},M^{1}}(v^{2}\mathbin{\odot}v^{1})\right)
=R_{-}^{M^{1},M^{2}}\left(e^{-\pi i[\omega]}\left(e^{\pi iL(0)}v^{1}\mathbin{\odot}e^{\pi iL(0)}v^{2}\right)\right) \\
&=e^{-\pi i[\omega]}R_{-}^{M^{1},M^{2}}\left(e^{\pi iL(0)}v^{1}\mathbin{\odot}e^{\pi iL(0)}v^{2}\right) \\
&=e^{-\pi i[\omega]}e^{\pi i[\omega]}\left(e^{-\pi iL(0)}e^{\pi iL(0)}v^{2}\mathbin{\odot}e^{-\pi iL(0)}e^{\pi iL(0)}v^{1}\right) \\
&=v^{2}\mathbin{\odot}v^{1}.
\end{aligned} R − M 1 , M 2 ( R + M 2 , M 1 ( v 2 ⊙ v 1 ) ) = R − M 1 , M 2 ( e − π i [ ω ] ( e π i L ( 0 ) v 1 ⊙ e π i L ( 0 ) v 2 ) ) = e − π i [ ω ] R − M 1 , M 2 ( e π i L ( 0 ) v 1 ⊙ e π i L ( 0 ) v 2 ) = e − π i [ ω ] e π i [ ω ] ( e − π i L ( 0 ) e π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) e π i L ( 0 ) v 1 ) = v 2 ⊙ v 1 .
Similarly, R + M 2 , M 1 R − M 1 , M 2 ( v 1 ⊙ v 2 ) = v 1 ⊙ v 2 R_{+}^{M^{2},M^{1}}R_{-}^{M^{1},M^{2}}(v^{1}\mathbin{\odot}v^{2})=v^{1}\mathbin{\odot}v^{2} R + M 2 , M 1 R − M 1 , M 2 ( v 1 ⊙ v 2 ) = v 1 ⊙ v 2 . Therefore, R + M 2 , M 1 R_{+}^{M^{2},M^{1}} R + M 2 , M 1 and R − M 1 , M 2 R_{-}^{M^{1},M^{2}} R − M 1 , M 2 are mutually inverse isomorphisms of A ( V ) A(V) A ( V ) -modules. Interchanging M 1 M^{1} M 1 and M 2 M^{2} M 2 shows that R − M 2 , M 1 R_{-}^{M^{2},M^{1}} R − M 2 , M 1 and R + M 1 , M 2 R_{+}^{M^{1},M^{2}} R + M 1 , M 2 are also mutually inverse. This proves (75) and completes the proof. □
Second Proof of Theorem 4.12: using relations in restricted conformal blocks.
First, we show well-definedness. Using the isomorphism R 1 R_{1} R 1 in Theorem 4.6 , define
R ~ ± : M 2 ⊗ C M 1 ⟶ M 1 ⊙ M 2 , R ~ ± ( v 2 ⊗ v 1 ) = e ∓ π i [ ω ] R 1 ( e ± π i L ( 0 ) v 2 ⊙ − 1 e ± π i L ( 0 ) v 1 ) . \widetilde{R}_{\pm}:M^{2}\otimes_{\mathbb{C}}M^{1}\longrightarrow M^{1}\mathbin{\odot}M^{2},\qquad\widetilde{R}_{\pm}(v^{2}\otimes v^{1})=e^{\mp\pi i[\omega]}R_{1}\left(e^{\pm\pi iL(0)}v^{2}\mathbin{\odot}_{-1}e^{\pm\pi iL(0)}v^{1}\right). R ± : M 2 ⊗ C M 1 ⟶ M 1 ⊙ M 2 , R ± ( v 2 ⊗ v 1 ) = e ∓ π i [ ω ] R 1 ( e ± π i L ( 0 ) v 2 ⊙ − 1 e ± π i L ( 0 ) v 1 ) .
By (64) , this is precisely
R ~ ± ( v 2 ⊗ v 1 ) = e ∓ π i [ ω ] ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) . \widetilde{R}_{\pm}(v^{2}\otimes v^{1})=e^{\mp\pi i[\omega]}\left(e^{\pm\pi iL(0)}v^{1}\mathbin{\odot}e^{\pm\pi iL(0)}v^{2}\right). R ± ( v 2 ⊗ v 1 ) = e ∓ π i [ ω ] ( e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) .
Let a ∈ V a\in V a ∈ V be homogeneous and m , n ∈ Z m,n\in\mathbb{Z} m , n ∈ Z satisfy n − m < wt a − 1 n-m<\operatorname{wt}a-1 n − m < wt a − 1 . Using the identity
e ± π i L ( 0 ) a ( j ) e ∓ π i L ( 0 ) = ( − 1 ) wt a − j − 1 a ( j ) e^{\pm\pi iL(0)}a_{(j)}e^{\mp\pi iL(0)}=(-1)^{\operatorname{wt}a-j-1}a_{(j)} e ± π i L ( 0 ) a ( j ) e ∓ π i L ( 0 ) = ( − 1 ) wt a − j − 1 a ( j )
and the local-action formulas (21) , (22) , we can show that
e ± π i L ( 0 ) ρ 1 ( a ⊗ t − m ( t − 1 ) − n ) = ( − 1 ) wt a + n − m − 1 ρ − 1 ( a ⊗ u − m ( u + 1 ) − n ) e ± π i L ( 0 ) , e^{\pm\pi iL(0)}\rho_{1}\left(a\mathbin{\otimes}\frac{t^{-m}}{(t-1)^{-n}}\right)
=(-1)^{\operatorname{wt}a+n-m-1}\rho_{-1}\left(a\mathbin{\otimes}\frac{u^{-m}}{(u+1)^{-n}}\right)e^{\pm\pi iL(0)}, e ± π i L ( 0 ) ρ 1 ( a ⊗ ( t − 1 ) − n t − m ) = ( − 1 ) wt a + n − m − 1 ρ − 1 ( a ⊗ ( u + 1 ) − n u − m ) e ± π i L ( 0 ) ,
e ± π i L ( 0 ) ρ 0 ( a ⊗ t − m ( t − 1 ) − n ) = ( − 1 ) wt a + n − m − 1 ρ 0 ( a ⊗ u − m ( u + 1 ) − n ) e ± π i L ( 0 ) . e^{\pm\pi iL(0)}\rho_{0}\left(a\mathbin{\otimes}\frac{t^{-m}}{(t-1)^{-n}}\right)
=(-1)^{\operatorname{wt}a+n-m-1}\rho_{0}\left(a\mathbin{\otimes}\frac{u^{-m}}{(u+1)^{-n}}\right)e^{\pm\pi iL(0)}. e ± π i L ( 0 ) ρ 0 ( a ⊗ ( t − 1 ) − n t − m ) = ( − 1 ) wt a + n − m − 1 ρ 0 ( a ⊗ ( u + 1 ) − n u − m ) e ± π i L ( 0 ) .
Apply (66) with w = 1 w=1 w = 1 , using the formulas above, we have
R ~ ± ( ρ 1 ( a ⊗ t − m ( t − 1 ) − n ) ⋅ v 2 ⊗ v 1 ) = e ± π i [ ω ] ⋅ R 1 ( e ± π i L ( 0 ) ρ 1 ( a ⊗ t − m ( t − 1 ) − n ) ⋅ v 2 ⊙ − 1 e ± π i L ( 0 ) v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ R 1 ( e ± π i L ( 0 ) ρ − 1 ( a ⊗ u − m ( u + 1 ) − n ) ⋅ e ± π i L ( 0 ) v 2 ⊙ − 1 e ± π i L ( 0 ) v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ ( e ± π i L ( 0 ) v 1 ⊙ ρ 0 ( a ⊗ z n ( z − 1 ) m ) ⋅ e ± π i L ( 0 ) v 2 ) . \begin{aligned}
\widetilde{R}_{\pm}\left(\rho_{1}\left(a\otimes\frac{t^{-m}}{(t-1)^{-n}}\right)\cdot v^{2}\otimes v^{1}\right)
&=e^{\pm\pi i[\omega]}\cdot R_{1}\left(e^{\pm\pi iL(0)}\rho_{1}\left(a\otimes\frac{t^{-m}}{(t-1)^{-n}}\right)\cdot v^{2}\mathbin{\odot_{-1}}e^{\pm\pi iL(0)}v^{1}\right)\\
&=(-1)^{\operatorname{wt}a+n-m-1}e^{\mp\pi i[\omega]}\cdot R_{1}\left(e^{\pm\pi iL(0)}\rho_{-1}\left(a\otimes\frac{u^{-m}}{(u+1)^{-n}}\right)\cdot e^{\pm\pi iL(0)}v^{2}\mathbin{\odot_{-1}}e^{\pm\pi iL(0)}v^{1}\right)\\
&=(-1)^{\operatorname{wt}a+n-m-1}e^{\mp\pi i[\omega]}\cdot\left(e^{\pm\pi iL(0)}v^{1}\mathbin{\odot}\rho_{0}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\cdot e^{\pm\pi iL(0)}v^{2}\right).
\end{aligned} R ± ( ρ 1 ( a ⊗ ( t − 1 ) − n t − m ) ⋅ v 2 ⊗ v 1 ) = e ± π i [ ω ] ⋅ R 1 ( e ± π i L ( 0 ) ρ 1 ( a ⊗ ( t − 1 ) − n t − m ) ⋅ v 2 ⊙ − 1 e ± π i L ( 0 ) v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ R 1 ( e ± π i L ( 0 ) ρ − 1 ( a ⊗ ( u + 1 ) − n u − m ) ⋅ e ± π i L ( 0 ) v 2 ⊙ − 1 e ± π i L ( 0 ) v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ ( e ± π i L ( 0 ) v 1 ⊙ ρ 0 ( a ⊗ ( z − 1 ) m z n ) ⋅ e ± π i L ( 0 ) v 2 ) .
Similarly, apply (65) with w = 1 w=1 w = 1 , we have
R ~ ± ( v 2 ⊗ ρ 0 ( a ⊗ t − m ( t − 1 ) − n ) ⋅ v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ R 1 ( e ± π i L ( 0 ) v 2 ⊙ − 1 ρ 0 ( a ⊗ u − m ( u + 1 ) − n ) ⋅ e ± π i L ( 0 ) v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ ( ρ 1 ( a ⊗ z n ( z − 1 ) m ) ⋅ e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) . \begin{aligned}
\widetilde{R}_{\pm}\left(v^{2}\otimes\rho_{0}\left(a\otimes\frac{t^{-m}}{(t-1)^{-n}}\right)\cdot v^{1}\right)
&=(-1)^{\operatorname{wt}a+n-m-1}e^{\mp\pi i[\omega]}\cdot R_{1}\left(e^{\pm\pi iL(0)}v^{2}\mathbin{\odot_{-1}}\rho_{0}\left(a\otimes\frac{u^{-m}}{(u+1)^{-n}}\right)\cdot e^{\pm\pi iL(0)}v^{1}\right)\\
&=(-1)^{\operatorname{wt}a+n-m-1}e^{\mp\pi i[\omega]}\cdot\left(\rho_{1}\left(a\otimes\frac{z^{n}}{(z-1)^{m}}\right)\cdot e^{\pm\pi iL(0)}v^{1}\mathbin{\odot}e^{\pm\pi iL(0)}v^{2}\right).
\end{aligned} R ± ( v 2 ⊗ ρ 0 ( a ⊗ ( t − 1 ) − n t − m ) ⋅ v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ R 1 ( e ± π i L ( 0 ) v 2 ⊙ − 1 ρ 0 ( a ⊗ ( u + 1 ) − n u − m ) ⋅ e ± π i L ( 0 ) v 1 ) = ( − 1 ) wt a + n − m − 1 e ∓ π i [ ω ] ⋅ ( ρ 1 ( a ⊗ ( z − 1 ) m z n ) ⋅ e ± π i L ( 0 ) v 1 ⊙ e ± π i L ( 0 ) v 2 ) .
Adding these two identities and applying (35) , we obtain
R ~ ± ( ρ 1 ( a ⊗ t − m ( t − 1 ) − n ) ⋅ v 2 ⊗ v 1 + v 2 ⊗ ρ 0 ( a ⊗ t − m ( t − 1 ) − n ) ⋅ v 1 ) = 0 \widetilde{R}_{\pm}\left(\rho_{1}\left(a\otimes\frac{t^{-m}}{(t-1)^{-n}}\right)\cdot v^{2}\otimes v^{1}+v^{2}\otimes\rho_{0}\left(a\otimes\frac{t^{-m}}{(t-1)^{-n}}\right)\cdot v^{1}\right)=0 R ± ( ρ 1 ( a ⊗ ( t − 1 ) − n t − m ) ⋅ v 2 ⊗ v 1 + v 2 ⊗ ρ 0 ( a ⊗ ( t − 1 ) − n t − m ) ⋅ v 1 ) = 0
in M 1 ⊙ M 2 M^{1}\mathbin{\odot}M^{2} M 1 ⊙ M 2 , since n − m < wt a − 1 n-m<\operatorname{wt}a-1 n − m < wt a − 1 .
Note that the elements a ⊗ t − m / ( t − 1 ) − n a\otimes t^{-m}/(t-1)^{-n} a ⊗ t − m / ( t − 1 ) − n span restricted chiral ideal L P 1 ∖ { 1 , 0 } ( V ) < 0 \mathcal{L}_{P^{1}\setminus\{1,0\}}(V)_{<0} L P 1 ∖ { 1 , 0 } ( V ) < 0 as a a a varies in V V V and m , n m,n m , n varies in Z \mathbb{Z} Z , with n − m < wt a − 1 n-m<\operatorname{wt}a-1 n − m < wt a − 1 , see (31) . Hence R ~ ± \widetilde{R}_{\pm} R ± annihilates the subspace L P 1 ∖ { 1 , 0 } ( V ) < 0 ⋅ ( M 2 ⊗ C M 1 ) \mathcal{L}_{P^{1}\setminus\{1,0\}}(V)_{<0}\cdot(M^{2}\otimes_{\mathbb{C}}M^{1}) L P 1 ∖ { 1 , 0 } ( V ) < 0 ⋅ ( M 2 ⊗ C M 1 ) and induces a well-defined linear map
R ± M 2 , M 1 : M 2 ⊙ M 1 ⟶ M 1 ⊙ M 2 . R_{\pm}^{M^{2},M^{1}}:M^{2}\mathbin{\odot}M^{1}\longrightarrow M^{1}\mathbin{\odot}M^{2}. R ± M 2 , M 1 : M 2 ⊙ M 1 ⟶ M 1 ⊙ M 2 .
given by (74) . Using (67) with w = 1 w=1 w = 1 , together with a similar argument as above, we can show that R ± M 2 , M 1 R_{\pm}^{M^{2},M^{1}} R ± M 2 , M 1 are A ( V ) A(V) A ( V ) -homomorphisms. The naturality and (75) follow from a similar arguments as the first proof, we omit the details. □
Definition 4.13. In what follows, we write
R M 1 , M 2 : = R − M 1 , M 2 : M 1 ⊙ M 2 ⟶ M 2 ⊙ M 1 , \mathcal{R}_{M^{1},M^{2}}:=R_{-}^{M^{1},M^{2}}:M^{1}\mathbin{\odot}M^{2}\longrightarrow M^{2}\mathbin{\odot}M^{1}, R M 1 , M 2 := R − M 1 , M 2 : M 1 ⊙ M 2 ⟶ M 2 ⊙ M 1 , R M 1 , M 2 ( v 1 ⊙ v 2 ) = e π i [ ω ] ⋅ ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) . (77) \mathcal{R}_{M^{1},M^{2}}(v^{1}\mathbin{\odot}v^{2})=e^{\pi i[\omega]}\cdot\left(e^{-\pi iL(0)}v^{2}\mathbin{\odot}e^{-\pi iL(0)}v^{1}\right). \tag*{(77)} R M 1 , M 2 ( v 1 ⊙ v 2 ) = e π i [ ω ] ⋅ ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) . (77) Here [ ω ] [\omega] [ ω ] acts on M 2 ⊙ M 1 M^{2}\mathbin{\odot}M^{1} M 2 ⊙ M 1 , and the outer factor e π i [ ω ] e^{\pi i[\omega]} e π i [ ω ] may also be written as e π i L ( 0 ) e^{\pi iL(0)} e π i L ( 0 ) .
In the next Subsection, we show that this family of isomorphisms defines a braiding on Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) with respect to the tensor structure transported from C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) through the equivalence ( Φ L , Ω ) (\Phi^{L},\Omega) ( Φ L , Ω ) , using the canonical identification η 1 \eta_{1} η 1 in (47) .
Comparison with the braiding isomorphism of Huang–Lepowsky–Zhang
By [36 Theorem 1.2] and [39 Section 12] , C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) has a natural braided tensor category structure with a twist. As in Section 3.3, we take ⊠ = ⊠ P ( 1 ) \boxtimes=\boxtimes_{P(1)} ⊠ = ⊠ P ( 1 ) and fix the branch l ( 1 ) = 0 l(1)=0 l ( 1 ) = 0 . We take the braiding associated to the path γ + 1 \gamma_{+}^{1} γ + 1 from − 1 -1 − 1 to 1 1 1 in the upper half-plane, see Example 4.8. Thus the continued logarithm at − 1 -1 − 1 is l γ + 1 ( − 1 ) = π i l_{\gamma_{+}^{1}}(-1)=\pi i l γ + 1 ( − 1 ) = π i .
Let Y M 1 , M 2 \mathcal{Y}_{M^{1},M^{2}} Y M 1 , M 2 be the universal intertwining operator satisfying
Y M 1 , M 2 ( v 1 , e 0 ) v 2 = v 1 ⊠ v 2 . \mathcal{Y}_{M^{1},M^{2}}(v^{1},e^{0})v^{2}=v^{1}\boxtimes v^{2}. Y M 1 , M 2 ( v 1 , e 0 ) v 2 = v 1 ⊠ v 2 .
With the above choice of branch, the HLZ braiding is characterized by
C M 1 , M 2 ( v 1 ⊠ v 2 ) = e L ( − 1 ) Y M 2 , M 1 ( v 2 , e π i ) v 1 . (78) \mathcal{C}_{M^{1},M^{2}}(v^{1}\boxtimes v^{2})=e^{L(-1)}\mathcal{Y}_{M^{2},M^{1}}(v^{2},e^{\pi i})v^{1}. \tag*{(78)} C M 1 , M 2 ( v 1 ⊠ v 2 ) = e L ( − 1 ) Y M 2 , M 1 ( v 2 , e π i ) v 1 . (78)
Indeed, the commutativity isomorphism first sends v 1 ⊠ v 2 v^{1}\boxtimes v^{2} v 1 ⊠ v 2 to e L ( − 1 ) ( v 2 ⊠ P ( − 1 ) v 1 ) e^{L(-1)}(v^{2}\boxtimes_{P(-1)}v^{1}) e L ( − 1 ) ( v 2 ⊠ P ( − 1 ) v 1 ) , and parallel transport from − 1 -1 − 1 to 1 1 1 in the upper half-plane sends v 2 ⊠ P ( − 1 ) v 1 v^{2}\boxtimes_{P(-1)}v^{1} v 2 ⊠ P ( − 1 ) v 1 to Y M 2 , M 1 ( v 2 , e π i ) v 1 \mathcal{Y}_{M^{2},M^{1}}(v^{2},e^{\pi i})v^{1} Y M 2 , M 1 ( v 2 , e π i ) v 1 ; see [39 Section 12.2] .
Following [51 Theorem 6.8] , we define
S ⊠ T : = Φ L ( S ) ⊙ Φ L ( T ) , f ⊠ g : = Φ L ( f ) ⊙ Φ L ( g ) , S , T ∈ Mod f d ( A ( V ) ) . (79) S\boxtimes T:=\Phi^{L}(S)\odot\Phi^{L}(T),\qquad f\boxtimes g:=\Phi^{L}(f)\odot\Phi^{L}(g),\qquad S,T\in\operatorname{Mod}_{\mathrm{fd}}(A(V)). \tag*{(79)} S ⊠ T := Φ L ( S ) ⊙ Φ L ( T ) , f ⊠ g := Φ L ( f ) ⊙ Φ L ( g ) , S , T ∈ Mod fd ( A ( V )) . (79)
By Proposition 3.11 , (79) is a bifunctor on Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) . We use ⊠ \boxtimes ⊠ for the tensor products in both categories, with the intended product determined by its arguments. We take the unit object to be 1 A ( V ) : = Ω ( V ) = C 1 \mathbf{1}_{A(V)}:=\Omega(V)=\mathbb{C}\mathbf{1} 1 A ( V ) := Ω ( V ) = C 1 . Denote the counit of the adjoint equivalence ( Φ L ⊣ Ω ) (\Phi^{L}\dashv\Omega) ( Φ L ⊣ Ω ) by ε : Φ L Ω ⇒ Id C 1 ( V ) \varepsilon:\Phi^{L}\Omega\Rightarrow\operatorname{Id}_{\mathcal{C}_{1}(V)} ε : Φ L Ω ⇒ Id C 1 ( V ) , and define
J S , T : = ε Φ L ( S ) ⊠ Φ L ( T ) ∘ Φ L ( η 1 Φ L ( S ) , Φ L ( T ) ) : Φ L ( S ⊠ T ) → ≅ Φ L ( S ) ⊠ Φ L ( T ) , J_{S,T}:=\varepsilon_{\Phi^{L}(S)\boxtimes\Phi^{L}(T)}\circ\Phi^{L}\left(\eta^{\Phi^{L}(S),\Phi^{L}(T)}_{1}\right):\Phi^{L}(S\boxtimes T)\xrightarrow{\ \cong\ }\Phi^{L}(S)\boxtimes\Phi^{L}(T), J S , T := ε Φ L ( S ) ⊠ Φ L ( T ) ∘ Φ L ( η 1 Φ L ( S ) , Φ L ( T ) ) : Φ L ( S ⊠ T ) ≅ Φ L ( S ) ⊠ Φ L ( T ) ,
J 0 : = ε V : Φ L ( 1 A ( V ) ) → ≅ V . (80) J_{0}:=\varepsilon_{V}:\Phi^{L}(\mathbf{1}_{A(V)})\xrightarrow{\ \cong\ }V. \tag*{(80)} J 0 := ε V : Φ L ( 1 A ( V ) ) ≅ V . (80)
By Theorem 3.14 and the naturality of ε \varepsilon ε , these maps are natural isomorphisms. Let α \alpha α , λ \lambda λ , and ρ \rho ρ be the associativity and unit constraints of C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) , and put J S , T , U L = ( J S , T ⊠ Id Φ L ( U ) ) ∘ J S ⊠ T , U J^{L}_{S,T,U}=(J_{S,T}\boxtimes\operatorname{Id}_{\Phi^{L}(U)})\circ J_{S\boxtimes T,U} J S , T , U L = ( J S , T ⊠ Id Φ L ( U ) ) ∘ J S ⊠ T , U and J S , T , U R = ( Id Φ L ( S ) ⊠ J T , U ) ∘ J S , T ⊠ U J^{R}_{S,T,U}=(\operatorname{Id}_{\Phi^{L}(S)}\boxtimes J_{T,U})\circ J_{S,T\boxtimes U} J S , T , U R = ( Id Φ L ( S ) ⊠ J T , U ) ∘ J S , T ⊠ U . Since Φ L \Phi^{L} Φ L is fully faithful, there exist unique natural isomorphisms α A ( V ) \alpha^{A(V)} α A ( V ) , λ A ( V ) \lambda^{A(V)} λ A ( V ) , and ρ A ( V ) \rho^{A(V)} ρ A ( V ) satisfying
Φ L ( α S , T , U A ( V ) ) = ( J S , T , U R ) − 1 ∘ α Φ L ( S ) , Φ L ( T ) , Φ L ( U ) ∘ J S , T , U L , \Phi^{L}\left(\alpha^{A(V)}_{S,T,U}\right)=\left(J^{R}_{S,T,U}\right)^{-1}\circ\alpha_{\Phi^{L}(S),\Phi^{L}(T),\Phi^{L}(U)}\circ J^{L}_{S,T,U}, Φ L ( α S , T , U A ( V ) ) = ( J S , T , U R ) − 1 ∘ α Φ L ( S ) , Φ L ( T ) , Φ L ( U ) ∘ J S , T , U L ,
Φ L ( λ S A ( V ) ) = λ Φ L ( S ) ∘ ( J 0 ⊠ Id Φ L ( S ) ) ∘ J 1 A ( V ) , S L , \Phi^{L}\left(\lambda^{A(V)}_{S}\right)=\lambda_{\Phi^{L}(S)}\circ\left(J_{0}\boxtimes\operatorname{Id}_{\Phi^{L}(S)}\right)\circ J^{L}_{\mathbf{1}_{A(V)},S}, Φ L ( λ S A ( V ) ) = λ Φ L ( S ) ∘ ( J 0 ⊠ Id Φ L ( S ) ) ∘ J 1 A ( V ) , S L ,
Φ L ( ρ S A ( V ) ) = ρ Φ L ( S ) ∘ ( Id Φ L ( S ) ⊠ J 0 ) ∘ J S , 1 A ( V ) R . (81) \Phi^{L}\left(\rho^{A(V)}_{S}\right)=\rho_{\Phi^{L}(S)}\circ\left(\operatorname{Id}_{\Phi^{L}(S)}\boxtimes J_{0}\right)\circ J^{R}_{S,\mathbf{1}_{A(V)}}. \tag*{(81)} Φ L ( ρ S A ( V ) ) = ρ Φ L ( S ) ∘ ( Id Φ L ( S ) ⊠ J 0 ) ∘ J S , 1 A ( V ) R . (81)
Finally, we define R S , T : = R Φ L ( S ) , Φ L ( T ) : S ⊠ T → ≅ T ⊠ S R_{S,T}:=R_{\Phi^{L}(S),\Phi^{L}(T)}:S\boxtimes T\xrightarrow{\ \cong\ }T\boxtimes S R S , T := R Φ L ( S ) , Φ L ( T ) : S ⊠ T ≅ T ⊠ S , see (77) .
Theorem 4.14. Under the assumptions of this section, we have the following results.
(1) For M 1 , M 2 ∈ C 1 ( V ) M^{1},M^{2}\in\mathcal{C}_{1}(V) M 1 , M 2 ∈ C 1 ( V ) , we have
η 1 M 2 , M 1 ∘ R M 1 , M 2 = Ω ( c M 1 , M 2 ) ∘ η 1 M 1 , M 2 . (82) \eta^{M^{2},M^{1}}_{1}\circ R_{M^{1},M^{2}}=\Omega(c_{M^{1},M^{2}})\circ\eta^{M^{1},M^{2}}_{1}. \tag*{(82)} η 1 M 2 , M 1 ∘ R M 1 , M 2 = Ω ( c M 1 , M 2 ) ∘ η 1 M 1 , M 2 . (82) In other words, R M 1 , M 2 R_{M^{1},M^{2}} R M 1 , M 2 is the restriction of the HLZ braiding c M 1 , M 2 c_{M^{1},M^{2}} c M 1 , M 2 to bottom degrees, and it is the unique homomorphism of left A ( V ) A(V) A ( V ) -modules satisfying (82) .
(2) The tensor product (79) , the unit 1 A ( V ) \mathbf{1}_{A(V)} 1 A ( V ) , the constraints (81) , and the isomorphisms R S , T R_{S,T} R S , T give Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) a braided tensor category structure. In particular, R R R satisfies the hexagon identities, and ( Φ L , J − 1 , J 0 − 1 ) (\Phi^{L},J^{-1},J_{0}^{-1}) ( Φ L , J − 1 , J 0 − 1 ) is a braided tensor equivalence:
J T , S ∘ Φ L ( R S , T ) = c Φ L ( S ) , Φ L ( T ) ∘ J S , T . (83) J_{T,S}\circ\Phi^{L}(R_{S,T})=c_{\Phi^{L}(S),\Phi^{L}(T)}\circ J_{S,T}. \tag*{(83)} J T , S ∘ Φ L ( R S , T ) = c Φ L ( S ) , Φ L ( T ) ∘ J S , T . (83) (3) The twist θ M = e 2 π i L ( 0 ) \theta_{M}=e^{2\pi iL(0)} θ M = e 2 π i L ( 0 ) on C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) corresponds to θ S = e 2 π i [ ω ] \theta_{S}=e^{2\pi i[\omega]} θ S = e 2 π i [ ω ] . If C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) is rigid, ribbon, or modular, then Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) has the same structure with the transported duals and twist. In particular, if V V V is strongly rational, then Φ L \Phi^{L} Φ L is an equivalence of modular tensor categories.
Proof. (1) Write Y = Y M 2 , M 1 \mathcal{Y}=\mathcal{Y}_{M^{2},M^{1}} Y = Y M 2 , M 1 . We first determine the bottom-degree component of Y ( v 2 , e π i ) v 1 \mathcal{Y}(v^{2},e^{\pi i})v^{1} Y ( v 2 , e π i ) v 1 . By linearity, we may assume that v 1 v^{1} v 1 and v 2 v^{2} v 2 are L ( 0 ) L(0) L ( 0 ) -eigenvectors. Write
Y ( v 2 , z ) v 1 = ∑ n ∈ C v ( n ) 2 v 1 z − n − 1 . \mathcal{Y}(v^{2},z)v^{1}=\sum_{n\in\mathbb{C}}v^{2}_{(n)}v^{1}z^{-n-1}. Y ( v 2 , z ) v 1 = n ∈ C ∑ v ( n ) 2 v 1 z − n − 1 .
By the L ( 0 ) L(0) L ( 0 ) -bracket identity for intertwining operators, we have
L ( 0 ) ( v ( n ) 2 v 1 ) = ( wt v 1 + wt v 2 − n − 1 ) v ( n ) 2 v 1 . L(0)(v^{2}_{(n)}v^{1})=(\operatorname{wt}v^{1}+\operatorname{wt}v^{2}-n-1)v^{2}_{(n)}v^{1}. L ( 0 ) ( v ( n ) 2 v 1 ) = ( wt v 1 + wt v 2 − n − 1 ) v ( n ) 2 v 1 .
Let p h p_{h} p h be the projection of Ω ( M 2 ⊠ M 1 ) \Omega(M^{2}\boxtimes M^{1}) Ω ( M 2 ⊠ M 1 ) onto the h h h -eigenspace of [ ω ] [\omega] [ ω ] . It follows that only the term with n = wt v 1 + wt v 2 − h − 1 n=\operatorname{wt}v^{1}+\operatorname{wt}v^{2}-h-1 n = wt v 1 + wt v 2 − h − 1 can contribute to p h π 0 Y ( v 2 , z ) v 1 p_{h}\pi_{0}\mathcal{Y}(v^{2},z)v^{1} p h π 0 Y ( v 2 , z ) v 1 . Thus we have
p h π 0 Y ( v 2 , e π i ) v 1 = e π i ( h − wt v 1 − wt v 2 ) p h π 0 Y ( v 2 , e 0 ) v 1 . p_{h}\pi_{0}\mathcal{Y}(v^{2},e^{\pi i})v^{1}=e^{\pi i(h-\operatorname{wt}v^{1}-\operatorname{wt}v^{2})}p_{h}\pi_{0}\mathcal{Y}(v^{2},e^{0})v^{1}. p h π 0 Y ( v 2 , e π i ) v 1 = e π i ( h − wt v 1 − wt v 2 ) p h π 0 Y ( v 2 , e 0 ) v 1 .
Since [ ω ] [\omega] [ ω ] acts semisimply on the finite-dimensional space Ω ( M 2 ⊠ M 1 ) \Omega(M^{2}\boxtimes M^{1}) Ω ( M 2 ⊠ M 1 ) , we sum over its eigenvalues. Since π 0 Y ( v 2 , e 0 ) v 1 = π 0 ( v 2 ⊠ v 1 ) = η 1 ( v 2 ⊙ v 1 ) \pi_{0}\mathcal{Y}(v^{2},e^{0})v^{1}=\pi_{0}(v^{2}\boxtimes v^{1})=\eta_{1}(v^{2}\odot v^{1}) π 0 Y ( v 2 , e 0 ) v 1 = π 0 ( v 2 ⊠ v 1 ) = η 1 ( v 2 ⊙ v 1 ) by Theorem 3.14 , we obtain
π 0 Y ( v 2 , e π i ) v 1 = e π i [ ω ] η 1 ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) . (84) \pi_{0}\mathcal{Y}(v^{2},e^{\pi i})v^{1}=e^{\pi i[\omega]}\eta_{1}\left(e^{-\pi iL(0)}v^{2}\odot e^{-\pi iL(0)}v^{1}\right). \tag*{(84)} π 0 Y ( v 2 , e π i ) v 1 = e π i [ ω ] η 1 ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) . (84)
By linearity, the same identity holds for arbitrary v 1 , v 2 v^{1},v^{2} v 1 , v 2 . Then by (46) , (78) , (84) , and the A ( V ) A(V) A ( V ) -linearity of η 1 \eta_{1} η 1 , we have
Ω ( c M 1 , M 2 ) η 1 ( v 1 ⊙ v 2 ) = π 0 c ‾ M 1 , M 2 ( v 1 ⊠ v 2 ) = π 0 e L ( − 1 ) Y ( v 2 , e π i ) v 1 = π 0 Y ( v 2 , e π i ) v 1 = η 1 ( e π i [ ω ] ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) ) = η 1 R M 1 , M 2 ( v 1 ⊙ v 2 ) . \begin{aligned}
\Omega(c_{M^{1},M^{2}})\eta_{1}(v^{1}\odot v^{2})=\pi_{0}\overline{c}_{M^{1},M^{2}}(v^{1}\boxtimes v^{2})=\pi_{0}e^{L(-1)}\mathcal{Y}(v^{2},e^{\pi i})v^{1}=\pi_{0}\mathcal{Y}(v^{2},e^{\pi i})v^{1} \\
&=\eta_{1}\left(e^{\pi i[\omega]}\left(e^{-\pi iL(0)}v^{2}\odot e^{-\pi iL(0)}v^{1}\right)\right)=\eta_{1}\mathcal{R}_{M^{1},M^{2}}(v^{1}\odot v^{2}).
\end{aligned} Ω ( c M 1 , M 2 ) η 1 ( v 1 ⊙ v 2 ) = π 0 c M 1 , M 2 ( v 1 ⊠ v 2 ) = π 0 e L ( − 1 ) Y ( v 2 , e π i ) v 1 = π 0 Y ( v 2 , e π i ) v 1 = η 1 ( e π i [ ω ] ( e − π i L ( 0 ) v 2 ⊙ e − π i L ( 0 ) v 1 ) ) = η 1 R M 1 , M 2 ( v 1 ⊙ v 2 ) .
This proves (82) . The uniqueness follows since η 1 M 1 , M 2 \eta^{M^{1},M^{2}}_{1} η 1 M 1 , M 2 and η 1 M 2 , M 1 \eta^{M^{2},M^{1}}_{1} η 1 M 2 , M 1 are isomorphisms.
(2) Applying (1) to Φ L ( S ) \Phi^{L}(S) Φ L ( S ) , Φ L ( T ) \Phi^{L}(T) Φ L ( T ) and using the naturality of the counit, we have
J T , S ∘ Φ L ( R S , T ) = ε Φ L ( T ) ⊠ Φ L ( S ) ∘ Φ L ( η Φ L ( T ) , Φ L ( S ) ) ∘ R Φ L ( S ) , Φ L ( T ) = ε Φ L ( T ) ⊠ Φ L ( S ) ∘ Φ L Ω ( c Φ L ( S ) , Φ L ( T ) ) ∘ Φ L ( η 1 Φ L ( S ) , Φ L ( T ) ) = c Φ L ( S ) , Φ L ( T ) ∘ ε Φ L ( S ) ⊠ Φ L ( T ) ∘ Φ L ( η 1 Φ L ( S ) , Φ L ( T ) ) = c Φ L ( S ) , Φ L ( T ) ∘ J S , T . \begin{aligned}
J_{T,S}\circ\Phi^{L}(\mathcal{R}_{S,T})=\varepsilon_{\Phi^{L}(T)\boxtimes\Phi^{L}(S)}\circ\Phi^{L}(\eta_{\Phi^{L}(T),\Phi^{L}(S)})\circ\mathcal{R}_{\Phi^{L}(S),\Phi^{L}(T)} \\
&=\varepsilon_{\Phi^{L}(T)\boxtimes\Phi^{L}(S)}\circ\Phi^{L}\Omega(c_{\Phi^{L}(S),\Phi^{L}(T)})\circ\Phi^{L}(\eta^{\Phi^{L}(S),\Phi^{L}(T)}_{1}) \\
&=c_{\Phi^{L}(S),\Phi^{L}(T)}\circ\varepsilon_{\Phi^{L}(S)\boxtimes\Phi^{L}(T)}\circ\Phi^{L}(\eta^{\Phi^{L}(S),\Phi^{L}(T)}_{1})=c_{\Phi^{L}(S),\Phi^{L}(T)}\circ J_{S,T}.
\end{aligned} J T , S ∘ Φ L ( R S , T ) = ε Φ L ( T ) ⊠ Φ L ( S ) ∘ Φ L ( η Φ L ( T ) , Φ L ( S ) ) ∘ R Φ L ( S ) , Φ L ( T ) = ε Φ L ( T ) ⊠ Φ L ( S ) ∘ Φ L Ω ( c Φ L ( S ) , Φ L ( T ) ) ∘ Φ L ( η 1 Φ L ( S ) , Φ L ( T ) ) = c Φ L ( S ) , Φ L ( T ) ∘ ε Φ L ( S ) ⊠ Φ L ( T ) ∘ Φ L ( η 1 Φ L ( S ) , Φ L ( T ) ) = c Φ L ( S ) , Φ L ( T ) ∘ J S , T .
This proves (83) . By (81) , (83) , and the naturality of J J J , the maps J J J identify the images under Φ L \Phi^{L} Φ L of the pentagon, triangle, and hexagon diagrams in Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) with the corresponding diagrams in C 1 ( V ) \mathcal{C}_{1}(V) C 1 ( V ) , which commute. Since Φ L \Phi^{L} Φ L is faithful, the original diagrams also commute.
(3) For homogeneous a ∈ V a\in V a ∈ V and n ∈ Z n\in\mathbb{Z} n ∈ Z , the identity [ L ( 0 ) , a ( n ) ] = ( wt a − n − 1 ) a ( n ) [L(0),a(n)]=(\operatorname{wt}a-n-1)a(n) [ L ( 0 ) , a ( n )] = ( wt a − n − 1 ) a ( n ) implies that e 2 π i L ( 0 ) e^{2\pi iL(0)} e 2 π i L ( 0 ) commutes with all modes of V V V . Thus it is a natural V V V -module automorphism, whose restriction to Ω ( Φ L ( S ) ) = S \Omega(\Phi^{L}(S))=S Ω ( Φ L ( S )) = S is e 2 π i [ ω ] ∣ S e^{2\pi i[\omega]}|_{S} e 2 π i [ ω ] ∣ S . Since Φ L ( S ) \Phi^{L}(S) Φ L ( S ) is generated by S S S , we have Φ L ( e 2 π i [ ω ] ∣ S ) = e 2 π i L ( 0 ) ∣ Φ L ( S ) \Phi^{L}(e^{2\pi i[\omega]}|_{S})=e^{2\pi iL(0)}|_{\Phi^{L}(S)} Φ L ( e 2 π i [ ω ] ∣ S ) = e 2 π i L ( 0 ) ∣ Φ L ( S ) . Moreover, by (78) and the L ( 0 ) L(0) L ( 0 ) -conjugation property,
c ‾ M 2 , M 1 ∘ c M 1 , M 2 Y M 1 , M 2 ( v 1 , x ) v 2 = Y M 1 , M 2 ( v 1 , e 2 π i x ) v 2 = e 2 π i L ( 0 ) Y M 1 , M 2 ( e − 2 π i L ( 0 ) v 1 , x ) e − 2 π i L ( 0 ) v 2 , \begin{aligned}
\overline{c}_{M^{2},M^{1}}\circ c_{M^{1},M^{2}}\mathcal{Y}_{M^{1},M^{2}}(v^{1},x)v^{2}=\mathcal{Y}_{M^{1},M^{2}}(v^{1},e^{2\pi i}x)v^{2} \\
&=e^{2\pi iL(0)}\mathcal{Y}_{M^{1},M^{2}}(e^{-2\pi iL(0)}v^{1},x)e^{-2\pi iL(0)}v^{2},
\end{aligned} c M 2 , M 1 ∘ c M 1 , M 2 Y M 1 , M 2 ( v 1 , x ) v 2 = Y M 1 , M 2 ( v 1 , e 2 π i x ) v 2 = e 2 π i L ( 0 ) Y M 1 , M 2 ( e − 2 π i L ( 0 ) v 1 , x ) e − 2 π i L ( 0 ) v 2 ,
which is the balancing identity for θ = e 2 π i L ( 0 ) \theta=e^{2\pi iL(0)} θ = e 2 π i L ( 0 ) , see also [30 ] . The duals, and the ribbon and modular structures, are transported along the braided tensor equivalence Φ L \Phi^{L} Φ L in the same way. Finally, if V V V is strongly rational, then C 1 ( V ) = Ord ( V ) \mathcal{C}_{1}(V)=\operatorname{Ord}(V) C 1 ( V ) = Ord ( V ) with these structures is modular by [30 ] . □
If we use the lower half-plane, then c M 1 , M 2 c_{M^{1},M^{2}} c M 1 , M 2 is replaced by c ‾ M 2 , M 1 \overline{c}_{M^{2},M^{1}} c M 2 , M 1 and R M 1 , M 2 \mathcal{R}_{M^{1},M^{2}} R M 1 , M 2 is replaced by R M 1 , M 2 + = ( R M 2 , M 1 ) − 1 \mathcal{R}^{+}_{M^{1},M^{2}}=(\mathcal{R}_{M^{2},M^{1}})^{-1} R M 1 , M 2 + = ( R M 2 , M 1 ) − 1 .
Explicit braiding isomorphism for affine VOAs at positive integral levels
In this section, we describe the fusion tensor product and the braiding isomorphism on Mod f d ( A ( L k ( g ) ) ) \operatorname{Mod}_{\mathrm{fd}}(A(L_{k}(\mathfrak{g}))) Mod fd ( A ( L k ( g ))) for k ∈ Z > 0 k\in\mathbb{Z}_{>0} k ∈ Z > 0 explicitly in terms of finite-dimensional g \mathfrak{g} g -modules and the Casimir tensor of g \mathfrak{g} g , see Theorems 5.1 and 5.4 .
Description of the ⊙ \odot ⊙ -tensor product for the affine VOA L k ( g ) L_{k}(\mathfrak{g}) L k ( g )
Throughout this section, let g \mathfrak{g} g be a finite-dimensional simple Lie algebra over C \mathbb{C} C , and let ( ⋅ ∣ ⋅ ) (\cdot\mid\cdot) ( ⋅ ∣ ⋅ ) be the invariant bilinear form on g \mathfrak{g} g normalized by ( θ ∣ θ ) = 2 (\theta\mid\theta)=2 ( θ ∣ θ ) = 2 , where θ \theta θ is the highest root. Fix root vectors e θ ∈ g θ e_{\theta}\in\mathfrak{g}_{\theta} e θ ∈ g θ and f θ ∈ g − θ f_{\theta}\in\mathfrak{g}_{-\theta} f θ ∈ g − θ . Let h ∨ h^{\vee} h ∨ be the dual Coxeter number of g \mathfrak{g} g , and k ∈ Z > 0 k\in\mathbb{Z}_{>0} k ∈ Z > 0 . Put
κ = k + h ∨ , P + k = { λ ∈ P + : ( λ ∣ θ ) ≤ k } . (85) \kappa=k+h^{\vee},\qquad P^{k}_{+}=\{\lambda\in P_{+}:(\lambda\mid\theta)\leq k\}. \tag*{(85)} κ = k + h ∨ , P + k = { λ ∈ P + : ( λ ∣ θ ) ≤ k } . (85)
Let V k ( g ) V^{k}(\mathfrak{g}) V k ( g ) be the universal affine VOA at level k k k , and let
g ^ = g ⊗ C [ t , t − 1 ] ⊕ C K , with a ( n ) = a ⊗ t n , \widehat{\mathfrak{g}}=\mathfrak{g}\otimes\mathbb{C}[t,t^{-1}]\oplus\mathbb{C}K,\qquad\text{with }a(n)=a\otimes t^{n}, g = g ⊗ C [ t , t − 1 ] ⊕ C K , with a ( n ) = a ⊗ t n ,
for a ∈ g a\in\mathfrak{g} a ∈ g and n ∈ Z n\in\mathbb{Z} n ∈ Z . For λ ∈ P + k \lambda\in P^{k}_{+} λ ∈ P + k , denote by
V k ( g , λ ) = U ( g ^ ) ⊗ U ( g [ t ] ⊕ C K ) L ( λ ) V^{k}(\mathfrak{g},\lambda)=U(\widehat{\mathfrak{g}})\otimes_{U(\mathfrak{g}[t]\oplus\mathbb{C}K)}L(\lambda) V k ( g , λ ) = U ( g ) ⊗ U ( g [ t ] ⊕ C K ) L ( λ )
the Weyl module associated to L ( λ ) L(\lambda) L ( λ ) , where g ⊗ t C [ t ] \mathfrak{g} \otimes t\mathbb{C}[t] g ⊗ t C [ t ] acts trivially on L ( λ ) L(\lambda) L ( λ ) and K K K acts as k k k . V k ( g , λ ) V^k(\mathfrak{g},\lambda) V k ( g , λ ) is a V k ( g ) V^k(\mathfrak{g}) V k ( g ) -module with a unique maximal submodule W k ( g , λ ) W^k(\mathfrak{g},\lambda) W k ( g , λ ) .
Since k ∈ Z > 0 k \in\mathbb{Z}_{>0} k ∈ Z > 0 and λ ∈ P + k \lambda\in P^k_+ λ ∈ P + k , W k ( g , λ ) W^k(\mathfrak{g},\lambda) W k ( g , λ ) is generated by one singular vector e θ ( − 1 ) k − ⟨ λ ∣ θ ⟩ + 1 v λ e_\theta(-1)^{k-\langle\lambda|\theta\rangle+1}v_\lambda e θ ( − 1 ) k − ⟨ λ ∣ θ ⟩ + 1 v λ [43 ] . Denote the irreducible quotient g ^ \widehat{\mathfrak{g}} g or V k ( g ) V^k(\mathfrak{g}) V k ( g ) -module by
L ^ k ( λ ) : = V k ( g , λ ) W k ( g , λ ) = V k ( g , λ ) U ( g ^ ) ⋅ ( e θ ( − 1 ) k − ⟨ λ ∣ θ ⟩ + 1 v λ ) . (86) \widehat{L}_k(\lambda):=\frac{V^k(\mathfrak{g},\lambda)}{W^k(\mathfrak{g},\lambda)}=\frac{V^k(\mathfrak{g},\lambda)}{U(\widehat{\mathfrak{g}})\cdot(e_\theta(-1)^{k-\langle\lambda|\theta\rangle+1}v_\lambda)}. \tag*{(86)} L k ( λ ) := W k ( g , λ ) V k ( g , λ ) = U ( g ) ⋅ ( e θ ( − 1 ) k − ⟨ λ ∣ θ ⟩ + 1 v λ ) V k ( g , λ ) . (86)
Let L k ( g ) = V k ( g ) / W k ( g ) L_k(\mathfrak{g})=V^k(\mathfrak{g})/W^k(\mathfrak{g}) L k ( g ) = V k ( g ) / W k ( g ) be the affine VOA at level k k k [24 ] , which is a strongly rational VOA [14 ] , with irreducible modules
W = { L ^ k ( λ ) : λ ∈ P + k } , where Ω ( L ^ k ( λ ) ) = L ^ k ( λ ) ( 0 ) = L ( λ ) . \mathcal{W}=\{\widehat{L}_k(\lambda):\lambda\in P^k_+\},\qquad\text{where}\quad\Omega(\widehat{L}_k(\lambda))=\widehat{L}_k(\lambda)(0)=L(\lambda). W = { L k ( λ ) : λ ∈ P + k } , where Ω ( L k ( λ )) = L k ( λ ) ( 0 ) = L ( λ ) .
Note that the Zhu algebras of V k ( g ) V^k(\mathfrak{g}) V k ( g ) and L k ( g ) L_k(\mathfrak{g}) L k ( g ) have the following identifications
A ( V k ( g ) ) → ∼ U ( g ) , [ a 1 ( − n 1 − 1 ) … a r ( − n r − 1 ) 1 ] ↦ ( − 1 ) n 1 + ⋯ + n r a r ⋯ a 1 , A(V^k(\mathfrak{g}))\xrightarrow{\sim}U(\mathfrak{g}),\qquad[a^1(-n_1-1)\ldots a^r(-n_r-1)\mathbf{1}]\mapsto(-1)^{n_1+\cdots+n_r}a^r\cdots a^1, A ( V k ( g )) ∼ U ( g ) , [ a 1 ( − n 1 − 1 ) … a r ( − n r − 1 ) 1 ] ↦ ( − 1 ) n 1 + ⋯ + n r a r ⋯ a 1 ,
A ( L k ( g ) ) ≅ U ( g ) / I k , I k = ⟨ e θ k + 1 ⟩ = U ( g ) e θ k + 1 U ( g ) . (87) A(L_k(\mathfrak{g}))\cong U(\mathfrak{g})/I_k,\qquad I_k=\langle e_\theta^{k+1}\rangle=U(\mathfrak{g})e_\theta^{k+1}U(\mathfrak{g}). \tag*{(87)} A ( L k ( g )) ≅ U ( g ) / I k , I k = ⟨ e θ k + 1 ⟩ = U ( g ) e θ k + 1 U ( g ) . (87)
In order to describe the braiding isomorphism on Mod f d ( A ( L k ( g ) ) ) \operatorname{Mod}_{\mathrm{fd}}(A(L_k(\mathfrak{g}))) Mod fd ( A ( L k ( g ))) , we first give a more concrete description of the ⊙ \odot ⊙ -tensor product L ^ k ( λ ) ⊙ L ^ k ( μ ) \widehat{L}_k(\lambda)\odot\widehat{L}_k(\mu) L k ( λ ) ⊙ L k ( μ ) using (86) and its defining relations (35) .
The following Theorem is a variation of [1 Proposition 6.5] .
Theorem 5.1. For λ , μ ∈ P + k \lambda,\mu\in P^k_+ λ , μ ∈ P + k , let J λ , μ J_{\lambda,\mu} J λ , μ be the subspace of L ( λ ) ⊗ L ( μ ) L(\lambda)\otimes L(\mu) L ( λ ) ⊗ L ( μ ) defined by
J λ , μ = v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 L ( μ ) + e θ k − ⟨ μ ∣ θ ⟩ + 1 L ( λ ) ⊗ v μ , (88) J_{\lambda,\mu}=v_\lambda\otimes e_\theta^{k-\langle\lambda|\theta\rangle+1}L(\mu)+e_\theta^{k-\langle\mu|\theta\rangle+1}L(\lambda)\otimes v_\mu, \tag*{(88)} J λ , μ = v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 L ( μ ) + e θ k − ⟨ μ ∣ θ ⟩ + 1 L ( λ ) ⊗ v μ , (88) and let U ( g ) . J λ , μ U(\mathfrak{g}).J_{\lambda,\mu} U ( g ) . J λ , μ be the g \mathfrak{g} g -submodule generated by J λ , μ J_{\lambda,\mu} J λ , μ with respect to the diagonal action. Then we have an isomorphism of A ( L k ( g ) ) A(L_k(\mathfrak{g})) A ( L k ( g )) -modules
φ ~ : L ( λ ) ⊗ C L ( μ ) U ( g ) . J λ , μ → ∼ L ^ k ( λ ) ⊙ L ^ k ( μ ) , u ⊗ v ‾ ↦ u ⊙ v . (89) \widetilde{\varphi}:\frac{L(\lambda)\otimes_{\mathbb{C}}L(\mu)}{U(\mathfrak{g}).J_{\lambda,\mu}}\xrightarrow{\sim}\widehat{L}_k(\lambda)\odot\widehat{L}_k(\mu),\qquad\overline{u\otimes v}\mapsto u\odot v. \tag*{(89)} φ : U ( g ) . J λ , μ L ( λ ) ⊗ C L ( μ ) ∼ L k ( λ ) ⊙ L k ( μ ) , u ⊗ v ↦ u ⊙ v . (89) Here the diagonal g \mathfrak{g} g -action on L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) descends to an action of A ( L k ( g ) ) ≅ U ( g ) / I k A(L_k(\mathfrak{g}))\cong U(\mathfrak{g})/I_k A ( L k ( g )) ≅ U ( g ) / I k on the quotient.
Proof. Since the affine VOA L k ( g ) L_k(\mathfrak{g}) L k ( g ) is strongly rational, it was proved in [51 Corollary 6.3] that the following map is an isomorphism of left A ( L k ( g ) ) A(L_k(\mathfrak{g})) A ( L k ( g )) -modules
A ( L ^ k ( λ ) ) ⊗ A ( L k ( g ) ) L ( μ ) → ∼ L ^ k ( λ ) ⊙ L ^ k ( μ ) , [ α ] ⊗ v ↦ α ⊙ v , (90) A(\widehat{L}_k(\lambda))\otimes_{A(L_k(\mathfrak{g}))}L(\mu)\xrightarrow{\sim}\widehat{L}_k(\lambda)\odot\widehat{L}_k(\mu),\qquad[\alpha]\otimes v\mapsto\alpha\odot v, \tag*{(90)} A ( L k ( λ )) ⊗ A ( L k ( g )) L ( μ ) ∼ L k ( λ ) ⊙ L k ( μ ) , [ α ] ⊗ v ↦ α ⊙ v , (90)
where α ∈ L ^ k ( λ ) \alpha\in\widehat{L}_k(\lambda) α ∈ L k ( λ ) , v ∈ L ( μ ) v\in L(\mu) v ∈ L ( μ ) , and A ( L ^ k ( λ ) ) A(\widehat{L}_k(\lambda)) A ( L k ( λ )) is Frenkel-Zhu’s bimodule. On the other hand, it was proved in [24 Theorem 3.2.2] that the following map is an isomorphism
L ( λ ) ⊗ C L ( μ ) U ( g ) . ( v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 L ( μ ) ) → ∼ A ( L ^ k ( λ ) ) ⊗ A ( L k ( g ) ) L ( μ ) , u ⊗ v ‾ ↦ [ u ] ⊗ v . (91) \frac{L(\lambda)\otimes_{\mathbb{C}}L(\mu)}{U(\mathfrak{g}).(v_\lambda\otimes e_\theta^{k-\langle\lambda|\theta\rangle+1}L(\mu))}\xrightarrow{\sim}A(\widehat{L}_k(\lambda))\otimes_{A(L_k(\mathfrak{g}))}L(\mu),\qquad\overline{u\otimes v}\mapsto[u]\otimes v. \tag*{(91)} U ( g ) . ( v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 L ( μ )) L ( λ ) ⊗ C L ( μ ) ∼ A ( L k ( λ )) ⊗ A ( L k ( g )) L ( μ ) , u ⊗ v ↦ [ u ] ⊗ v . (91)
By composing (90) and (91) , we see that there exists a surjective map
φ : L ( λ ) ⊗ C L ( μ ) ↠ L ^ k ( λ ) ⊙ L ^ k ( μ ) , u ⊗ v ↦ u ⊙ v , \varphi:L(\lambda)\otimes_{\mathbb{C}}L(\mu)\twoheadrightarrow\widehat{L}_k(\lambda)\odot\widehat{L}_k(\mu),\qquad u\otimes v\mapsto u\odot v, φ : L ( λ ) ⊗ C L ( μ ) ↠ L k ( λ ) ⊙ L k ( μ ) , u ⊗ v ↦ u ⊙ v ,
with ker ( φ ) = U ( g ) . ( v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 L ( μ ) ) \operatorname{ker}(\varphi)=U(\mathfrak{g}).(v_\lambda\otimes e_\theta^{k-\langle\lambda|\theta\rangle+1}L(\mu)) ker ( φ ) = U ( g ) . ( v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 L ( μ )) . We claim that ker ( φ ) = U ( g ) . J λ , μ \operatorname{ker}(\varphi)=U(\mathfrak{g}).J_{\lambda,\mu} ker ( φ ) = U ( g ) . J λ , μ .
Indeed, by the definition of J λ , μ J_{\lambda,\mu} J λ , μ (88) , it is clear that ker ( φ ) ⊆ U ( g ) . J λ , μ \operatorname{ker}(\varphi)\subseteq U(\mathfrak{g}).J_{\lambda,\mu} ker ( φ ) ⊆ U ( g ) . J λ , μ . On the other hand, it is clear that v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 v ∈ ker ( φ ) v_\lambda\otimes e_\theta^{k-\langle\lambda|\theta\rangle+1}v\in\operatorname{ker}(\varphi) v λ ⊗ e θ k − ⟨ λ ∣ θ ⟩ + 1 v ∈ ker ( φ ) for all v ∈ L ( μ ) v\in L(\mu) v ∈ L ( μ ) . Now apply (35) to a = e θ a=e_\theta a = e θ , t = 0 t=0 t = 0 , s = 2 s=2 s = 2 , and wt e θ = 1 \operatorname{wt}e_\theta=1 wt e θ = 1 , we have the following relation in L ^ k ( λ ) ⊙ L ^ k ( μ ) \widehat{L}_k(\lambda)\odot\widehat{L}_k(\mu) L k ( λ ) ⊙ L k ( μ )
u ⊙ e θ ( − 1 ) v = − e θ ( 0 ) u ⊙ v = − e θ u ⊙ v . u\odot e_\theta(-1)v=-e_\theta(0)u\odot v=-e_\theta u\odot v. u ⊙ e θ ( − 1 ) v = − e θ ( 0 ) u ⊙ v = − e θ u ⊙ v .
Since e θ ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 v μ = 0 e_\theta(-1)^{k-\langle\mu|\theta\rangle+1}v_\mu=0 e θ ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 v μ = 0 in L ^ k ( μ ) \widehat{L}_k(\mu) L k ( μ ) (86) , it follows that for any u ∈ L ( λ ) u\in L(\lambda) u ∈ L ( λ ) , we have
e θ k − ⟨ μ ∣ θ ⟩ + 1 u ⊙ v μ = ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 u ⊙ e θ ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 v μ = ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 u ⊙ 0 = 0. e_\theta^{k-\langle\mu|\theta\rangle+1}u\odot v_\mu=(-1)^{k-\langle\mu|\theta\rangle+1}u\odot e_\theta(-1)^{k-\langle\mu|\theta\rangle+1}v_\mu=(-1)^{k-\langle\mu|\theta\rangle+1}u\odot0=0. e θ k − ⟨ μ ∣ θ ⟩ + 1 u ⊙ v μ = ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 u ⊙ e θ ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 v μ = ( − 1 ) k − ⟨ μ ∣ θ ⟩ + 1 u ⊙ 0 = 0.
In other words, φ ( e θ k − ( μ ∣ θ ) + 1 u ⊗ v μ ) = 0 \varphi(e_{\theta}^{k-(\mu\mid\theta)+1}u\otimes v_{\mu})=0 φ ( e θ k − ( μ ∣ θ ) + 1 u ⊗ v μ ) = 0 , and so e θ k − ( μ ∣ θ ) + 1 u ⊗ v μ ∈ ker ( φ ) e_{\theta}^{k-(\mu\mid\theta)+1}u\otimes v_{\mu}\in\ker(\varphi) e θ k − ( μ ∣ θ ) + 1 u ⊗ v μ ∈ ker ( φ ) . Thus, J λ , μ = v λ ⊗ e θ k − ( λ ∣ θ ) + 1 L ( μ ) + e θ k − ( μ ∣ θ ) + 1 L ( λ ) ⊗ v μ ⊆ ker ( φ ) J_{\lambda,\mu}=v_{\lambda}\otimes e_{\theta}^{k-(\lambda\mid\theta)+1}L(\mu)+e_{\theta}^{k-(\mu\mid\theta)+1}L(\lambda)\otimes v_{\mu}\subseteq\ker(\varphi) J λ , μ = v λ ⊗ e θ k − ( λ ∣ θ ) + 1 L ( μ ) + e θ k − ( μ ∣ θ ) + 1 L ( λ ) ⊗ v μ ⊆ ker ( φ ) . Since ker ( φ ) \ker(\varphi) ker ( φ ) is a U ( g ) U(\mathfrak{g}) U ( g ) -submodule of L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) , we have U ( g ) . J λ , μ ⊆ ker ( φ ) U(\mathfrak{g}).J_{\lambda,\mu}\subseteq\ker(\varphi) U ( g ) . J λ , μ ⊆ ker ( φ ) . This proves the isomorphism (89) .
For the simplicity of notation, we write
T λ , μ : = L ( λ ) ⊗ C L ( μ ) U ( g ) . J λ , μ , (92) T_{\lambda,\mu}:=\frac{L(\lambda)\otimes_{\mathbb{C}}L(\mu)}{U(\mathfrak{g}).J_{\lambda,\mu}}, \tag*{(92)} T λ , μ := U ( g ) . J λ , μ L ( λ ) ⊗ C L ( μ ) , (92)
and denote the quotient map by
π λ , μ : L ( λ ) ⊗ C L ( μ ) ↠ T λ , μ , u ⊗ v ↦ u ⊗ v ‾ . \pi_{\lambda,\mu}:L(\lambda)\otimes_{\mathbb{C}}L(\mu)\twoheadrightarrow T_{\lambda,\mu},\quad u\otimes v\mapsto\overline{u\otimes v}. π λ , μ : L ( λ ) ⊗ C L ( μ ) ↠ T λ , μ , u ⊗ v ↦ u ⊗ v .
The following Corollary follows immediately from (79) , Theorem 4.14 , and Theorem 5.1 .
Corollary 5.2. Let k ∈ Z > 0 k\in\mathbb{Z}_{>0} k ∈ Z > 0 . The category Mod f d ( U ( g ) / I k ) \operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})/I_k) Mod fd ( U ( g ) / I k ) (87) is a modular tensor category, with fusion tensor product given by
L ( λ ) ⊠ L ( μ ) = T λ , μ = L ( λ ) ⊗ C L ( μ ) U ( g ) . J λ , μ , λ , μ ∈ P + k . (93) L(\lambda)\boxtimes L(\mu)=T_{\lambda,\mu}=\frac{L(\lambda)\otimes_{\mathbb{C}}L(\mu)}{U(\mathfrak{g}).J_{\lambda,\mu}},\quad\lambda,\mu\in P_{+}^{k}. \tag*{(93)} L ( λ ) ⊠ L ( μ ) = T λ , μ = U ( g ) . J λ , μ L ( λ ) ⊗ C L ( μ ) , λ , μ ∈ P + k . (93) Moreover, the action of U ( g ) / I k U(\mathfrak{g})/I_k U ( g ) / I k on T λ , μ T_{\lambda,\mu} T λ , μ is induced by the diagonal g \mathfrak{g} g -action on L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) .
The braiding isomorphism on Mod_fd(A(L_k(g)))
Under the isomorphism (89) , the braiding isomorphism R = R L k ( λ ) ^ , L k ( μ ) ^ R=R_{\widehat{L_k(\lambda)},\widehat{L_k(\mu)}} R = R L k ( λ ) , L k ( μ ) translates to an isomorphism R λ , μ : T λ , μ → ∼ T μ , λ R_{\lambda,\mu}:T_{\lambda,\mu}\xrightarrow{\sim}T_{\mu,\lambda} R λ , μ : T λ , μ ∼ T μ , λ that fits into the following diagram:
L k ( λ ) ^ ⊙ L k ( μ ) ^ → R L k ( μ ) ^ ⊙ L k ( λ ) ^ ≃ ↓ ↓ ≃ T λ , μ → R λ , μ T μ , λ (94) \begin{CD}
\widehat{L_k(\lambda)}\mathbin{\odot}\widehat{L_k(\mu)} @>R>> \widehat{L_k(\mu)}\mathbin{\odot}\widehat{L_k(\lambda)} \\
@V{\simeq}VV @VV{\simeq}V \\
T_{\lambda,\mu} @>R_{\lambda,\mu}>> T_{\mu,\lambda} \tag*{(94)}
\end{CD} L k ( λ ) ⊙ L k ( μ ) ≃ ↓ ⏐ T λ , μ R R λ , μ L k ( μ ) ⊙ L k ( λ ) ↓ ⏐ ≃ T μ , λ (94)
In order to give an explicit description of R λ , μ R_{\lambda,\mu} R λ , μ , in view of (77) , we first need to determine the action of [ ω ] ∈ A ( L k ( g ) ) [\omega]\in A(L_k(\mathfrak{g})) [ ω ] ∈ A ( L k ( g )) on T λ , μ T_{\lambda,\mu} T λ , μ . The Virasoro element ω \omega ω of the affine VOA L k ( g ) L_k(\mathfrak{g}) L k ( g ) is given by the Segal-Sugawara construction. More precisely, choose dual bases { x a } a = 1 dim g \{x^a\}_{a=1}^{\dim\mathfrak{g}} { x a } a = 1 d i m g and { x a } a = 1 dim g \{x_a\}_{a=1}^{\dim\mathfrak{g}} { x a } a = 1 d i m g of g \mathfrak{g} g with respect to ( ⋅ ∣ ⋅ ) (\cdot\mid\cdot) ( ⋅ ∣ ⋅ ) , recall that
ω = 1 2 k ∑ a x a ( − 1 ) x a ( − 1 ) 1 ∈ L k ( g ) 2 . \omega=\frac{1}{2k}\sum_a x^a(-1)x_a(-1)\mathbf{1}\in L_k(\mathfrak{g})_2. ω = 2 k 1 a ∑ x a ( − 1 ) x a ( − 1 ) 1 ∈ L k ( g ) 2 .
Consider the following two elements
C = ∑ a x a x a ∈ U ( g ) , Ω = ∑ a x a ⊗ x a ∈ U ( g ) ⊗ U ( g ) . (95) C=\sum_a x_a x^a\in U(\mathfrak{g}),\qquad\Omega=\sum_a x_a\otimes x^a\in U(\mathfrak{g})\otimes U(\mathfrak{g}). \tag*{(95)} C = a ∑ x a x a ∈ U ( g ) , Ω = a ∑ x a ⊗ x a ∈ U ( g ) ⊗ U ( g ) . (95)
Let Δ : U ( g ) → U ( g ) ⊗ C U ( g ) \Delta:U(\mathfrak{g})\to U(\mathfrak{g})\otimes_{\mathbb{C}}U(\mathfrak{g}) Δ : U ( g ) → U ( g ) ⊗ C U ( g ) be the standard coproduct, then
Δ ( C ) = C ⊗ 1 + 1 ⊗ C + 2 Ω . (96) \Delta(C)=C\otimes1+1\otimes C+2\Omega. \tag*{(96)} Δ ( C ) = C ⊗ 1 + 1 ⊗ C + 2Ω. (96)
Note that C C C is the Casimir element, which acts on the irreducible g \mathfrak{g} g -module L ( λ ) L(\lambda) L ( λ ) by the scalar ( λ ∣ λ + 2 ρ ) (\lambda\mid\lambda+2\rho) ( λ ∣ λ + 2 ρ ) . Then
1 2 k C ∣ L ( λ ) = ( λ ∣ λ + 2 ρ ) 2 k ⋅ Id L ( λ ) = h λ ⋅ Id L ( λ ) , λ ∈ P + k . \frac{1}{2k}C\big|_{L(\lambda)}=\frac{(\lambda\mid\lambda+2\rho)}{2k}\cdot\operatorname{Id}_{L(\lambda)}=h_\lambda\cdot\operatorname{Id}_{L(\lambda)},\quad\lambda\in P_{+}^{k}. 2 k 1 C L ( λ ) = 2 k ( λ ∣ λ + 2 ρ ) ⋅ Id L ( λ ) = h λ ⋅ Id L ( λ ) , λ ∈ P + k .
Under the isomorphism of Zhu algebra (87) , we have
[ ω ] = 1 2 k ∑ a x a x a + I k = 1 2 k C ‾ ∈ U ( g ) / I k . (97) [\omega]=\frac{1}{2k}\sum_a x_a x^a+I_k=\frac{1}{2k}\overline{C}\in U(\mathfrak{g})/I_k. \tag*{(97)} [ ω ] = 2 k 1 a ∑ x a x a + I k = 2 k 1 C ∈ U ( g ) / I k . (97)
Lemma 5.3. Under the quotient identification (89) , the action of [ ω ] ∈ U ( g ) / I k [\omega] \in U(\mathfrak{g})/I_k [ ω ] ∈ U ( g ) / I k on T λ , μ T_{\lambda,\mu} T λ , μ (92) is given by
[ ω ] . π λ , μ ( u ⊗ v ) = π λ , μ ( ( h λ + h μ ) u ⊗ v + 1 κ Ω λ , μ ( u ⊗ v ) ) , (98) [\omega].\pi_{\lambda,\mu}(u \otimes v)=\pi_{\lambda,\mu}\left((h_\lambda+h_\mu)u \otimes v+\frac{1}{\kappa}\Omega_{\lambda,\mu}(u \otimes v)\right), \tag*{(98)} [ ω ] . π λ , μ ( u ⊗ v ) = π λ , μ ( ( h λ + h μ ) u ⊗ v + κ 1 Ω λ , μ ( u ⊗ v ) ) , (98) where u ∈ L ( λ ) u \in L(\lambda) u ∈ L ( λ ) , v ∈ L ( μ ) v \in L(\mu) v ∈ L ( μ ) , and Ω λ , μ \Omega_{\lambda,\mu} Ω λ , μ is defined by
U ( g ) ⊗ C U ( g ) ⟶ End ( L ( λ ) ⊗ C L ( μ ) ) , Ω ↦ Ω λ , μ . (99) U(\mathfrak{g})\otimes_{\mathbb{C}} U(\mathfrak{g})\longrightarrow\operatorname{End}(L(\lambda)\otimes_{\mathbb{C}}L(\mu)),\qquad\Omega\mapsto\Omega_{\lambda,\mu}. \tag*{(99)} U ( g ) ⊗ C U ( g ) ⟶ End ( L ( λ ) ⊗ C L ( μ )) , Ω ↦ Ω λ , μ . (99) In particular, U ( g ) . J λ , μ ⊂ L ( λ ) ⊗ C L ( μ ) U(\mathfrak{g}).J_{\lambda,\mu}\subset L(\lambda)\otimes_{\mathbb{C}}L(\mu) U ( g ) . J λ , μ ⊂ L ( λ ) ⊗ C L ( μ ) is invariant under Ω λ , μ \Omega_{\lambda,\mu} Ω λ , μ .
Proof. By Theorem 5.1 , the U ( g ) / I k U(\mathfrak{g})/I_k U ( g ) / I k -action on T λ , μ T_{\lambda,\mu} T λ , μ is induced by the diagonal g \mathfrak{g} g -action. Then for any u ⊗ v ∈ L ( λ ) ⊗ C L ( μ ) u\otimes v\in L(\lambda)\otimes_{\mathbb{C}}L(\mu) u ⊗ v ∈ L ( λ ) ⊗ C L ( μ ) , it follows from (96) and (97) that
[ ω ] . π λ , μ ( u ⊗ v ) = [ ω ] . u ⊗ v = 1 2 κ Δ ( C ) . ( u ⊗ v ) = 1 2 κ π λ , μ ( C u ⊗ v + u ⊗ C v + 2 Ω λ , μ ( u ⊗ v ) ) = π λ , μ ( ( h λ + h μ ) u ⊗ v + 1 κ Ω λ , μ ( u ⊗ v ) ) , \begin{aligned}
[\omega].\pi_{\lambda,\mu}(u\otimes v)&=[\omega].u\otimes v=\frac{1}{2\kappa}\Delta(C).(u\otimes v)\\
&=\frac{1}{2\kappa}\pi_{\lambda,\mu}\left(Cu\otimes v+u\otimes Cv+2\Omega_{\lambda,\mu}(u\otimes v)\right)\\
&=\pi_{\lambda,\mu}\left((h_\lambda+h_\mu)u\otimes v+\frac{1}{\kappa}\Omega_{\lambda,\mu}(u\otimes v)\right),
\end{aligned} [ ω ] . π λ , μ ( u ⊗ v ) = [ ω ] . u ⊗ v = 2 κ 1 Δ ( C ) . ( u ⊗ v ) = 2 κ 1 π λ , μ ( C u ⊗ v + u ⊗ C v + 2 Ω λ , μ ( u ⊗ v ) ) = π λ , μ ( ( h λ + h μ ) u ⊗ v + κ 1 Ω λ , μ ( u ⊗ v ) ) ,
which proves (98) . Since U ( g ) . J λ , μ U(\mathfrak{g}).J_{\lambda,\mu} U ( g ) . J λ , μ is a diagonal g \mathfrak{g} g -submodule, it is invariant under Δ ( C ) \Delta(C) Δ ( C ) . On L ( λ ) ⊗ L ( μ ) L(\lambda)\otimes L(\mu) L ( λ ) ⊗ L ( μ ) , the operator C ⊗ 1 + 1 ⊗ C C\otimes1+1\otimes C C ⊗ 1 + 1 ⊗ C is the scalar 2 κ ( h λ + h μ ) ⋅ Id 2\kappa(h_\lambda+h_\mu)\cdot\operatorname{Id} 2 κ ( h λ + h μ ) ⋅ Id . Hence (96) implies that U ( g ) . J λ , μ U(\mathfrak{g}).J_{\lambda,\mu} U ( g ) . J λ , μ is invariant under Ω λ , μ \Omega_{\lambda,\mu} Ω λ , μ . □
Second proof of (98) . There is an alternative way to prove (98) using the relations in ⊙ \odot ⊙ -tensor product only, without invoking the identification of Zhu algebras (87) .
Indeed, by (37) , the action of [ ω ] ∈ A ( L k ( g ) ) [\omega]\in A(L_k(\mathfrak{g})) [ ω ] ∈ A ( L k ( g )) on L ^ k ( λ ) ⊙ L ^ k ( μ ) \widehat{L}_k(\lambda)\odot\widehat{L}_k(\mu) L k ( λ ) ⊙ L k ( μ ) is given by
[ ω ] . ( u ⊙ v ) = ∑ j ≥ 0 ( 1 j ) ω ( j ) u ⊙ v + u ⊙ o ( ω ) v = L ( 0 ) u ⊙ v + L ( − 1 ) u ⊙ v + u ⊙ L ( 0 ) v , \begin{aligned}
[\omega].(u\odot v)&=\sum_{j\geq0}\binom{1}{j}\omega_{(j)}u\odot v+u\odot o(\omega)v\\
&=L(0)u\odot v+L(-1)u\odot v+u\odot L(0)v,
\end{aligned} [ ω ] . ( u ⊙ v ) = j ≥ 0 ∑ ( j 1 ) ω ( j ) u ⊙ v + u ⊙ o ( ω ) v = L ( 0 ) u ⊙ v + L ( − 1 ) u ⊙ v + u ⊙ L ( 0 ) v ,
for u ∈ L ( λ ) = L ^ k ( λ ) ( 0 ) u\in L(\lambda)=\widehat{L}_k(\lambda)(0) u ∈ L ( λ ) = L k ( λ ) ( 0 ) , v ∈ L ( μ ) = L ^ k ( μ ) ( 0 ) v\in L(\mu)=\widehat{L}_k(\mu)(0) v ∈ L ( μ ) = L k ( μ ) ( 0 ) . Since L ( 0 ) L(0) L ( 0 ) acts semisimply on the irreducible ordinary module L ^ k ( λ ) \widehat{L}_k(\lambda) L k ( λ ) , we have L ( 0 ) u = h λ u L(0)u=h_\lambda u L ( 0 ) u = h λ u and L ( 0 ) v = h μ v L(0)v=h_\mu v L ( 0 ) v = h μ v . Moreover, it follows from the Sugawara formula [43 ] and x a ( i ) u = 0 x^a(i)u=0 x a ( i ) u = 0 for i > 0 i>0 i > 0 that
L ( − 1 ) u = 1 κ ∑ a ∑ i ≥ 0 x a ( − 1 − i ) x a ( i ) u = 1 κ ∑ a x a ( − 1 ) x a ( 0 ) u . L(-1)u=\frac{1}{\kappa}\sum_a\sum_{i\geq0}x_a(-1-i)x^a(i)u=\frac{1}{\kappa}\sum_a x_a(-1)x^a(0)u. L ( − 1 ) u = κ 1 a ∑ i ≥ 0 ∑ x a ( − 1 − i ) x a ( i ) u = κ 1 a ∑ x a ( − 1 ) x a ( 0 ) u .
Apply (35) to x a ∈ L k ( g ) 1 x^a\in L_k(\mathfrak{g})_1 x a ∈ L k ( g ) 1 , s = 1 s=1 s = 1 , and t = 1 t=1 t = 1 , we have
L ( − 1 ) u ⊙ v = − 1 κ ∑ a ∑ j ≥ 0 ( − 1 j ) ( − 1 ) − 1 − j x a ( 0 ) u ⊙ x a ( j ) v = 1 κ ∑ a x a ( 0 ) u ⊙ x a ( 0 ) v = 1 κ ∑ a x a u ⊙ x a v . \begin{aligned}
L(-1)u\odot v&=-\frac{1}{\kappa}\sum_a\sum_{j\geq0}\binom{-1}{j}(-1)^{-1-j}x^a(0)u\odot x_a(j)v\\
&=\frac{1}{\kappa}\sum_a x^a(0)u\odot x_a(0)v=\frac{1}{\kappa}\sum_a x^a u\odot x_a v.
\end{aligned} L ( − 1 ) u ⊙ v = − κ 1 a ∑ j ≥ 0 ∑ ( j − 1 ) ( − 1 ) − 1 − j x a ( 0 ) u ⊙ x a ( j ) v = κ 1 a ∑ x a ( 0 ) u ⊙ x a ( 0 ) v = κ 1 a ∑ x a u ⊙ x a v .
Since φ ~ \widetilde{\varphi} φ is an A ( V ) A(V) A ( V ) -module isomorphism, it follows from (95) that
[ ω ] . π λ , μ ( u ⊗ v ) = φ ~ − 1 ( [ ω ] . ( u ⊙ v ) ) = φ ~ − 1 ( ( h λ + h μ ) u ⊙ v + 1 κ ∑ a x a u ⊙ x a v ) = π λ , μ ( ( h λ + h μ ) u ⊗ v + 1 κ Ω λ , μ ( u ⊗ v ) ) , \begin{aligned}
[\omega].\pi_{\lambda,\mu}(u\otimes v)&=\widetilde{\varphi}^{-1}([\omega].(u\odot v))\\
&=\widetilde{\varphi}^{-1}\left((h_\lambda+h_\mu)u\odot v+\frac{1}{\kappa}\sum_a x^a u\odot x_a v\right)\\
&=\pi_{\lambda,\mu}\left((h_\lambda+h_\mu)u\otimes v+\frac{1}{\kappa}\Omega_{\lambda,\mu}(u\otimes v)\right),
\end{aligned} [ ω ] . π λ , μ ( u ⊗ v ) = φ − 1 ([ ω ] . ( u ⊙ v )) = φ − 1 ( ( h λ + h μ ) u ⊙ v + κ 1 a ∑ x a u ⊙ x a v ) = π λ , μ ( ( h λ + h μ ) u ⊗ v + κ 1 Ω λ , μ ( u ⊗ v ) ) ,
where we used ∑ a x a u ⊗ x a v = Ω λ , μ ( u ⊗ v ) \sum_a x^a u\otimes x_a v=\Omega_{\lambda,\mu}(u\otimes v) ∑ a x a u ⊗ x a v = Ω λ , μ ( u ⊗ v ) . This proves (98) . □
Since ( h λ + h μ ) ⋅ Id (h_\lambda+h_\mu)\cdot\operatorname{Id} ( h λ + h μ ) ⋅ Id and 1 κ Ω λ , μ \frac{1}{\kappa}\Omega_{\lambda,\mu} κ 1 Ω λ , μ are commutative when viewed as operators on L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) , it follows from (98) that
e π i [ ω ] π λ , μ ( u ⊗ v ) = π λ , μ ( e π i ( h λ + h μ ) exp ( π i κ Ω λ , μ ) ( u ⊗ v ) ) . (100) e^{\pi i[\omega]}\pi_{\lambda,\mu}(u\otimes v)=\pi_{\lambda,\mu}\left(e^{\pi i(h_\lambda+h_\mu)}\exp\left(\frac{\pi i}{\kappa}\Omega_{\lambda,\mu}\right)(u\otimes v)\right). \tag*{(100)} e π i [ ω ] π λ , μ ( u ⊗ v ) = π λ , μ ( e π i ( h λ + h μ ) exp ( κ π i Ω λ , μ ) ( u ⊗ v ) ) . (100)
Now we define a tensor-swapping operator. Let
τ : L ( λ ) ⊗ C L ( μ ) ⟶ L ( μ ) ⊗ C L ( λ ) , τ ( u ⊗ v ) = v ⊗ u . \tau:L(\lambda)\otimes_{\mathbb{C}}L(\mu)\longrightarrow L(\mu)\otimes_{\mathbb{C}}L(\lambda),\qquad\tau(u\otimes v)=v\otimes u. τ : L ( λ ) ⊗ C L ( μ ) ⟶ L ( μ ) ⊗ C L ( λ ) , τ ( u ⊗ v ) = v ⊗ u .
It is easy to see from (95) that τ ∘ Ω λ , μ = Ω μ , λ ∘ τ \tau\circ\Omega_{\lambda,\mu}=\Omega_{\mu,\lambda}\circ\tau τ ∘ Ω λ , μ = Ω μ , λ ∘ τ . Then by (77) , the corresponding elements in vector spaces of the diagram (94) are given by
u ⊙ v → R e π i [ ω ] ⋅ ( e − π i L ( 0 ) v ⊙ e − π i L ( 0 ) u ) ( 5.5 ) ↓ ↓ ( 5.5 ) π λ , μ ( u ⊗ v ) → R λ , μ φ ~ − 1 ( e π i [ ω ] ⋅ ( e − π i L ( 0 ) v ⊙ e − π i L ( 0 ) u ) ) . \begin{CD}
u\mathbin{\odot}v @>{\mathcal{R}}>> e^{\pi i[\omega]}\cdot\left(e^{-\pi iL(0)}v\mathbin{\odot}e^{-\pi iL(0)}u\right)\\
@V{(5.5)}VV @VV{(5.5)}V\\
\pi_{\lambda,\mu}(u\otimes v) @>{\mathcal{R}_{\lambda,\mu}}>> \widetilde{\varphi}^{-1}\left(e^{\pi i[\omega]}\cdot\left(e^{-\pi iL(0)}v\mathbin{\odot}e^{-\pi iL(0)}u\right)\right).
\end{CD} u ⊙ v ( 5.5 ) ↓ ⏐ π λ , μ ( u ⊗ v ) R R λ , μ e π i [ ω ] ⋅ ( e − π i L ( 0 ) v ⊙ e − π i L ( 0 ) u ) ↓ ⏐ ( 5.5 ) φ − 1 ( e π i [ ω ] ⋅ ( e − π i L ( 0 ) v ⊙ e − π i L ( 0 ) u ) ) .
It follows from (89) and (100) that
φ ~ − 1 ( e π i [ ω ] ⋅ ( e − π i L ( 0 ) v ⊙ e − π i L ( 0 ) u ) ) = e π i [ ω ] π μ , λ ( e − π i h μ v ⊗ e − π i h λ u ) = π μ , λ ( exp ( π i κ Ω μ , λ ) ( v ⊗ u ) ) = π μ , λ ( exp ( π i κ Ω μ , λ ) τ ( u ⊗ v ) ) = π μ , λ ( ( τ ∘ exp ( π i κ Ω λ , μ ) ) ( u ⊗ v ) ) . \begin{aligned}
\widetilde{\varphi}^{-1}\left(e^{\pi i[\omega]}\cdot\left(e^{-\pi iL(0)}v\mathbin{\odot}e^{-\pi iL(0)}u\right)\right)=e^{\pi i[\omega]}\pi_{\mu,\lambda}\left(e^{-\pi ih_\mu}v\otimes e^{-\pi ih_\lambda}u\right) \\
&=\pi_{\mu,\lambda}\left(\exp\left(\frac{\pi i}{\kappa}\Omega_{\mu,\lambda}\right)(v\otimes u)\right)=\pi_{\mu,\lambda}\left(\exp\left(\frac{\pi i}{\kappa}\Omega_{\mu,\lambda}\right)\tau(u\otimes v)\right) \\
&=\pi_{\mu,\lambda}\left(\left(\tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega_{\lambda,\mu}\right)\right)(u\otimes v)\right).
\end{aligned} φ − 1 ( e π i [ ω ] ⋅ ( e − π i L ( 0 ) v ⊙ e − π i L ( 0 ) u ) ) = e π i [ ω ] π μ , λ ( e − π i h μ v ⊗ e − π i h λ u ) = π μ , λ ( exp ( κ π i Ω μ , λ ) ( v ⊗ u ) ) = π μ , λ ( exp ( κ π i Ω μ , λ ) τ ( u ⊗ v ) ) = π μ , λ ( ( τ ∘ exp ( κ π i Ω λ , μ ) ) ( u ⊗ v ) ) .
Therefore, we proved the following Theorem.
Theorem 5.4. Under the identifications L ^ k ( λ ) ⊙ L ^ k ( μ ) ≃ T λ , μ = ( L ( λ ) ⊗ C L ( μ ) ) / U ( g ) ⋅ J λ , μ \widehat{L}_k(\lambda)\mathbin{\odot}\widehat{L}_k(\mu)\simeq T_{\lambda,\mu}=(L(\lambda)\otimes_{\mathbb{C}}L(\mu))/U(\mathfrak{g})\cdot J_{\lambda,\mu} L k ( λ ) ⊙ L k ( μ ) ≃ T λ , μ = ( L ( λ ) ⊗ C L ( μ )) / U ( g ) ⋅ J λ , μ in Theorem 5.1 , the braiding isomorphism on the modular tensor category Mod f d ( A ( L k ( g ) ) ) ≃ Mod f d ( U ( g ) / I k ) \operatorname{Mod}_{\mathrm{fd}}(A(L_k(\mathfrak{g})))\simeq\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})/I_k) Mod fd ( A ( L k ( g ))) ≃ Mod fd ( U ( g ) / I k ) is given by
R λ , μ : T λ , μ → ∼ T μ , λ , R λ , μ ( π λ , μ ( u ⊗ v ) ) = π μ , λ ( ( τ ∘ exp ( π i κ Ω λ , μ ) ) ( u ⊗ v ) ) . (101) \mathcal{R}_{\lambda,\mu}:T_{\lambda,\mu}\xrightarrow{\sim}T_{\mu,\lambda},\qquad\mathcal{R}_{\lambda,\mu}\left(\pi_{\lambda,\mu}(u\otimes v)\right)=\pi_{\mu,\lambda}\left(\left(\tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega_{\lambda,\mu}\right)\right)(u\otimes v)\right). \tag*{(101)} R λ , μ : T λ , μ ∼ T μ , λ , R λ , μ ( π λ , μ ( u ⊗ v ) ) = π μ , λ ( ( τ ∘ exp ( κ π i Ω λ , μ ) ) ( u ⊗ v ) ) . (101) Equivalently, denote q = e π i / κ ∈ C × q=e^{\pi i/\kappa}\in\mathbb{C}^{\times} q = e π i / κ ∈ C × , then we can write
R λ , μ ( u ⊗ v ) = q Ω μ , λ ( v ⊗ u ) , \mathcal{R}_{\lambda,\mu}(u\otimes v)=q^{\Omega_{\mu,\lambda}}(v\otimes u), R λ , μ ( u ⊗ v ) = q Ω μ , λ ( v ⊗ u ) , where λ , μ ∈ P + κ \lambda,\mu\in P^{\kappa}_{+} λ , μ ∈ P + κ , u ∈ L ( λ ) u\in L(\lambda) u ∈ L ( λ ) , v ∈ L ( μ ) v\in L(\mu) v ∈ L ( μ ) , and Ω λ , μ \Omega_{\lambda,\mu} Ω λ , μ is defined by (99) .
Braiding isomorphism on K L k ( g ) KL^k(\mathfrak{g}) K L k ( g ) at generic levels
Let g \mathfrak{g} g , ( ⋅ ∣ ⋅ ) (\cdot\mid\cdot) ( ⋅ ∣ ⋅ ) , h ∨ h^{\vee} h ∨ , and g ^ \widehat{\mathfrak{g}} g be as in § 5 , and let k ∈ C k\in\mathbb{C} k ∈ C with κ = k + h ∨ ≠ 0 \kappa=k+h^{\vee}\ne0 κ = k + h ∨ = 0 . For λ ∈ P + \lambda\in P_{+} λ ∈ P + , we denote by V k ( g , λ ) V^{k}(\mathfrak{g},\lambda) V k ( g , λ ) the Weyl module associated to L ( λ ) L(\lambda) L ( λ ) , as in § 5 .
Throughout this section, we assume that either k + h ∨ ∉ Q k+h^{\vee}\notin\mathbb{Q} k + h ∨ ∈ / Q , or g = s l 2 \mathfrak{g}=\mathfrak{sl}_{2} g = sl 2 and k = − 2 ± 1 p k=-2\pm\frac{1}{p} k = − 2 ± p 1 with p ∈ Z > 0 p\in\mathbb{Z}_{>0} p ∈ Z > 0 . In both cases, there is no singular vector in the vacuum module, that is, L k ( g ) = V k ( g ) L_k(\mathfrak{g})=V^{k}(\mathfrak{g}) L k ( g ) = V k ( g ) [28 ] and L − 2 ± 1 p ( s l 2 ) = V − 2 ± 1 p ( s l 2 ) L_{-2\pm\frac{1}{p}}(\mathfrak{sl}_{2})=V^{-2\pm\frac{1}{p}}(\mathfrak{sl}_{2}) L − 2 ± p 1 ( sl 2 ) = V − 2 ± p 1 ( sl 2 ) [5 ] . We write V = L k ( g ) = V k ( g ) V=L_k(\mathfrak{g})=V^{k}(\mathfrak{g}) V = L k ( g ) = V k ( g ) . Then
A ( V ) ≃ U ( g ) , [ a 1 ( − n 1 − 1 ) ⋯ a r ( − n r − 1 ) 1 ] ⟼ ( − 1 ) n 1 + ⋯ + n r a r ⋯ a 1 , A(V)\simeq U(\mathfrak{g}),\qquad[a^{1}(-n_{1}-1)\cdots a^{r}(-n_{r}-1)\mathbf{1}]\longmapsto(-1)^{n_{1}+\cdots+n_{r}}a^{r}\cdots a^{1}, A ( V ) ≃ U ( g ) , [ a 1 ( − n 1 − 1 ) ⋯ a r ( − n r − 1 ) 1 ] ⟼ ( − 1 ) n 1 + ⋯ + n r a r ⋯ a 1 ,
where a i ∈ g a^{i}\in\mathfrak{g} a i ∈ g for all i i i , and n 1 ≥ ⋯ ≥ n r ≥ 0 n_{1}\ge\cdots\ge n_{r}\ge0 n 1 ≥ ⋯ ≥ n r ≥ 0 . Hence Mod f d ( A ( V ) ) \operatorname{Mod}_{\mathrm{fd}}(A(V)) Mod fd ( A ( V )) is the category of finite-dimensional g \mathfrak{g} g -modules, which is semisimple. Moreover, the category C 1 ( V ) = K L k ( g ) C_{1}(V)=KL^{k}(\mathfrak{g}) C 1 ( V ) = K L k ( g ) is semisimple and Φ L ( Mod f d ( U ( g ) ) ) ⊆ K L k ( g ) \Phi^{L}(\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})))\subseteq KL^{k}(\mathfrak{g}) Φ L ( Mod fd ( U ( g ))) ⊆ K L k ( g ) , see Subsection 2.2.3 and Example 3.12. Hence Theorems 3.14 , 4.12 , and 4.14 apply to V V V .
Description of the ⊙ \odot ⊙ -tensor product at generic levels
Lemma 6.1. Let S , T ∈ Mod f d ( A ( L k ( g ) ) ) S,T \in\operatorname{Mod}_{\mathrm{fd}}(A(L_k(\mathfrak{g}))) S , T ∈ Mod fd ( A ( L k ( g ))) . Then Φ L ( S ) ⊙ Φ L ( T ) = span { s ⊙ t : s ∈ S , t ∈ T } \Phi^L(S) \odot\Phi^L(T) = \operatorname{span}\{s \odot t : s \in S, t \in T\} Φ L ( S ) ⊙ Φ L ( T ) = span { s ⊙ t : s ∈ S , t ∈ T } . In particular, the canonical map
S ⊗ C T ⟶ Φ L ( S ) ⊙ Φ L ( T ) = S ⊠ T , s ⊗ t ↦ s ⊙ t (102) S \otimes_{\mathbb{C}} T \longrightarrow\Phi^L(S) \odot\Phi^L(T) = S \boxtimes T,\qquad s \otimes t \mapsto s \odot t \tag*{(102)} S ⊗ C T ⟶ Φ L ( S ) ⊙ Φ L ( T ) = S ⊠ T , s ⊗ t ↦ s ⊙ t (102) is surjective. Furthermore, the action of A ( L k ( g ) ) A(L_k(\mathfrak{g})) A ( L k ( g )) on Φ L ( S ) ⊙ Φ L ( T ) \Phi^L(S) \odot\Phi^L(T) Φ L ( S ) ⊙ Φ L ( T ) is given as follows: for any a ∈ g a \in\mathfrak{g} a ∈ g , with [ a ] = a ( − 1 ) 1 + O ( V ) ∈ A ( L k ( g ) ) = U ( g ) [a] = a(-1)\mathbf{1} + O(V) \in A(L_k(\mathfrak{g})) = U(\mathfrak{g}) [ a ] = a ( − 1 ) 1 + O ( V ) ∈ A ( L k ( g )) = U ( g ) ,
[ a ] . ( s ⊙ t ) = a ( 0 ) s ⊙ t + s ⊙ a ( 0 ) t , s ∈ S , t ∈ T . (103) [a].(s \odot t) = a(0)s \odot t + s \odot a(0)t,\qquad s \in S,\ t \in T. \tag*{(103)} [ a ] . ( s ⊙ t ) = a ( 0 ) s ⊙ t + s ⊙ a ( 0 ) t , s ∈ S , t ∈ T . (103) Proof. It follows from [51 Lemma 4.5] that Φ L ( S ) ⊙ Φ L ( T ) = span { v ⊙ t : v ∈ Φ L ( S ) , t ∈ T } \Phi^L(S) \odot\Phi^L(T) = \operatorname{span}\{v \odot t : v \in\Phi^L(S), t \in T\} Φ L ( S ) ⊙ Φ L ( T ) = span { v ⊙ t : v ∈ Φ L ( S ) , t ∈ T } . Moreover, since U ( L k ( g ) ) ≅ U ^ k ( g ^ ) / I ^ k \mathcal{U}(L_k(\mathfrak{g})) \cong\widehat{U}_k(\widehat{\mathfrak{g}})/\widehat{I}_k U ( L k ( g )) ≅ U k ( g ) / I k , where I ^ k \widehat{I}_k I k is the closed two-sided ideal generated by the image of the singular vectors of V k ( g ) V^k(\mathfrak{g}) V k ( g ) in U ^ k ( g ^ ) \widehat{U}_k(\widehat{\mathfrak{g}}) U k ( g ) [24 , 21 ] , then we have
Φ L ( S ) = span { a 1 ( − n 1 ) ⋯ a r ( − n r ) s : a i ∈ g , n 1 ≥ ⋯ ≥ n r ≥ 1 , s ∈ S } . \Phi^L(S) = \operatorname{span}\{a^1(-n_1)\cdots a^r(-n_r)s : a^i \in\mathfrak{g}, n_1 \ge\cdots\ge n_r \ge1, s \in S\}. Φ L ( S ) = span { a 1 ( − n 1 ) ⋯ a r ( − n r ) s : a i ∈ g , n 1 ≥ ⋯ ≥ n r ≥ 1 , s ∈ S } . Note that a i ∈ g ≅ V 1 a^i \in\mathfrak{g} \cong V_1 a i ∈ g ≅ V 1 . Then by (52) with w t a 1 = 1 wta^1 = 1 w t a 1 = 1 , s = 1 s = 1 s = 1 , and t = n 1 t = n_1 t = n 1 , we have
a 1 ( − n 1 ) ⋯ a r ( − n r ) s ⊙ t = − ∑ j ≥ 0 ( − n 1 j ) ( − 1 ) − n 1 − j ( a 2 ( − n 2 ) ⋯ a r ( − n r ) s ) ⊙ a 1 ( j ) t a^1(-n_1)\cdots a^r(-n_r)s \odot t
= -\sum_{j\ge0}\binom{-n_1}{j}(-1)^{-n_1-j}(a^2(-n_2)\cdots a^r(-n_r)s)\odot a^1(j)t a 1 ( − n 1 ) ⋯ a r ( − n r ) s ⊙ t = − j ≥ 0 ∑ ( j − n 1 ) ( − 1 ) − n 1 − j ( a 2 ( − n 2 ) ⋯ a r ( − n r ) s ) ⊙ a 1 ( j ) t = ( − 1 ) − n 1 − 1 ( a 2 ( − n 2 ) ⋯ a r ( − n r ) s ) ⊙ a 1 ( 0 ) t , = (-1)^{-n_1-1}(a^2(-n_2)\cdots a^r(-n_r)s)\odot a^1(0)t, = ( − 1 ) − n 1 − 1 ( a 2 ( − n 2 ) ⋯ a r ( − n r ) s ) ⊙ a 1 ( 0 ) t , since t ∈ T = Ω ( Φ L ( T ) ) t \in T = \Omega(\Phi^L(T)) t ∈ T = Ω ( Φ L ( T )) and a 1 ( j ) t = 0 a^1(j)t = 0 a 1 ( j ) t = 0 for j ≥ 1 j \ge1 j ≥ 1 . Proceed inductively, we have
a 1 ( − n 1 ) ⋯ a r ( − n r ) s ⊙ t = ( − 1 ) − n 1 − ⋯ − n r − r s ⊙ ( a r ( 0 ) ⋯ a 1 ( 0 ) t ) , a^1(-n_1)\cdots a^r(-n_r)s \odot t
= (-1)^{-n_1-\cdots-n_r-r}s \odot(a^r(0)\cdots a^1(0)t), a 1 ( − n 1 ) ⋯ a r ( − n r ) s ⊙ t = ( − 1 ) − n 1 − ⋯ − n r − r s ⊙ ( a r ( 0 ) ⋯ a 1 ( 0 ) t ) , where a r ( 0 ) ⋯ a 1 ( 0 ) t ∈ T a^r(0)\cdots a^1(0)t \in T a r ( 0 ) ⋯ a 1 ( 0 ) t ∈ T . Hence Φ L ( S ) ⊙ Φ L ( T ) = S ⊙ T \Phi^L(S) \odot\Phi^L(T) = S \odot T Φ L ( S ) ⊙ Φ L ( T ) = S ⊙ T . Finally, the action formula (103) of A ( L k ( g ) ) A(L_k(\mathfrak{g})) A ( L k ( g )) on s ⊙ t s \odot t s ⊙ t is given by (37) . □ \square □
The following Proposition generalizes [1 Lemma 6.3] .
Proposition 6.2. Assume V = V k ( g ) V = V^k(\mathfrak{g}) V = V k ( g ) , where either k + h ∨ ∉ Q k + h^\vee\notin\mathbb{Q} k + h ∨ ∈ / Q , or g = s l 2 \mathfrak{g} = \mathfrak{sl}_2 g = sl 2 and k = − 2 ± 1 p k = -2 \pm\frac{1}{p} k = − 2 ± p 1 with p ∈ Z > 0 p \in\mathbb{Z}_{>0} p ∈ Z > 0 . Then we have
Φ L ( L ( λ ) ) ≅ V k ( g , λ ) , ∀ L ( λ ) ∈ Mod f d ( A ( L k ( g ) ) ) . \Phi^L(L(\lambda)) \cong V^k(\mathfrak{g},\lambda),\qquad\forall L(\lambda) \in\operatorname{Mod}_{\mathrm{fd}}(A(L_k(\mathfrak{g}))). Φ L ( L ( λ )) ≅ V k ( g , λ ) , ∀ L ( λ ) ∈ Mod fd ( A ( L k ( g ))) . In particular, for any simple modules L ( λ ) , L ( μ ) ∈ Mod f d ( A ( L k ( g ) ) ) L(\lambda), L(\mu) \in\operatorname{Mod}_{\mathrm{fd}}(A(L_k(\mathfrak{g}))) L ( λ ) , L ( μ ) ∈ Mod fd ( A ( L k ( g ))) the canonical map (102) is a natural isomorphism of A ( L k ( g ) ) ≅ U ( g ) A(L_k(\mathfrak{g})) \cong U(\mathfrak{g}) A ( L k ( g )) ≅ U ( g ) -modules
L ( λ ) ⊗ C L ( μ ) ≅ L ( λ ) ⊠ L ( μ ) , s ⊗ t ↦ s ⊙ t , (104) L(\lambda) \otimes_{\mathbb{C}} L(\mu) \cong L(\lambda) \boxtimes L(\mu),\qquad s \otimes t \mapsto s \odot t, \tag*{(104)} L ( λ ) ⊗ C L ( μ ) ≅ L ( λ ) ⊠ L ( μ ) , s ⊗ t ↦ s ⊙ t , (104) where the left-hand side carries the usual diagonal g \mathfrak{g} g -action.
Proof. Since U ( V k ( g ) ) ≅ U ^ k ( g ^ ) \mathcal{U}(V^k(\mathfrak{g})) \cong\widehat{U}_k(\widehat{\mathfrak{g}}) U ( V k ( g )) ≅ U k ( g ) and deg ( a ( n ) ) = − n \deg(a(n))=-n deg ( a ( n )) = − n , for all a ∈ g = V 1 a \in\mathfrak{g}=V_1 a ∈ g = V 1 and n ∈ Z n \in\mathbb{Z} n ∈ Z , then by (29) ,
Φ L ( L ( λ ) ) = ( U ^ k ( g ^ ) U ^ k ( g ^ ) U ^ k ( g ^ < 0 ) ) ⊗ U ^ k ( g ^ 0 ) L ( λ ) ≅ Ind g ^ ≥ 0 g ^ L ( λ ) = V k ( g , λ ) . \Phi^L(L(\lambda))
=
\left(
\frac{\widehat{U}_k(\widehat{\mathfrak{g}})}
{\widehat{U}_k(\widehat{\mathfrak{g}})\widehat{U}_k(\widehat{\mathfrak{g}}_{<0})}
\right)
\otimes_{\widehat{U}_k(\widehat{\mathfrak{g}}_0)} L(\lambda)
\cong\operatorname{Ind}_{\widehat{\mathfrak{g}}_{\ge0}}^{\widehat{\mathfrak{g}}} L(\lambda)
= V^k(\mathfrak{g},\lambda). Φ L ( L ( λ )) = ( U k ( g ) U k ( g < 0 ) U k ( g ) ) ⊗ U k ( g 0 ) L ( λ ) ≅ Ind g ≥ 0 g L ( λ ) = V k ( g , λ ) . Similar to the proof of [1 Lemma 6.3] , we define a linear map
φ ~ : V k ( g , λ ) ⊗ C V k ( g , μ ) ⟶ L ( λ ) ⊗ C L ( μ ) \widetilde{\varphi}: V^k(\mathfrak{g},\lambda) \otimes_{\mathbb{C}} V^k(\mathfrak{g},\mu) \longrightarrow L(\lambda) \otimes_{\mathbb{C}} L(\mu) φ : V k ( g , λ ) ⊗ C V k ( g , μ ) ⟶ L ( λ ) ⊗ C L ( μ ) on the tensor products of PBW basis elements of V k ( g , λ ) V^k(\mathfrak{g},\lambda) V k ( g , λ ) and V k ( g , μ ) V^k(\mathfrak{g},\mu) V k ( g , μ ) by induction on the total degree, as follows.
For u ∈ L ( λ ) u \in L(\lambda) u ∈ L ( λ ) and v ∈ L ( μ ) v \in L(\mu) v ∈ L ( μ ) , define φ ~ ( u ⊗ v ) : = u ⊗ v ∈ L ( λ ) ⊗ C L ( μ ) \widetilde{\varphi}(u \otimes v) := u \otimes v \in L(\lambda) \otimes_{\mathbb{C}} L(\mu) φ ( u ⊗ v ) := u ⊗ v ∈ L ( λ ) ⊗ C L ( μ ) . Choose an ordered basis { a 1 , … , a n } \{a^1,\ldots,a^n\} { a 1 , … , a n } of g \mathfrak{g} g . A PBW basis element of V k ( g , λ ) V^k(\mathfrak{g},\lambda) V k ( g , λ ) of positive degree can be uniquely written as a ( − r ) u a(-r)u a ( − r ) u for some a = a i a=a^i a = a i , r ≥ 1 r \ge1 r ≥ 1 , and u u u a PBW basis element of lower degree. Define
φ ~ ( a ( − r ) u ⊗ v ) : = − ∑ j ≥ 0 ( − r j ) ( − 1 ) − r − j φ ~ ( u ⊗ a ( j ) v ) . (105) \widetilde{\varphi}(a(-r)u \otimes v)
:=
-\sum_{j\ge0}\binom{-r}{j}(-1)^{-r-j}\widetilde{\varphi}(u \otimes a(j)v). \tag*{(105)} φ ( a ( − r ) u ⊗ v ) := − j ≥ 0 ∑ ( j − r ) ( − 1 ) − r − j φ ( u ⊗ a ( j ) v ) . (105)
Similarly, given a PBW basis element a ( − r ) v a(-r)v a ( − r ) v of V k ( g , μ ) V^k(\mathfrak{g},\mu) V k ( g , μ ) , with r ≥ 1 r \ge1 r ≥ 1 , define
φ ~ ( u ⊗ a ( − r ) v ) : = − ∑ j ≥ 0 ( − r j ) φ ~ ( a ( j ) u ⊗ v ) . (106) \widetilde{\varphi}(u \otimes a(-r)v) := -\sum_{j\geq0}\binom{-r}{j}\widetilde{\varphi}(a(j)u \otimes v). \tag*{(106)} φ ( u ⊗ a ( − r ) v ) := − j ≥ 0 ∑ ( j − r ) φ ( a ( j ) u ⊗ v ) . (106)
Note that
deg ( u ⊗ v ) < deg ( a ( − r ) u ⊗ v ) , deg ( a ( j ) u ⊗ v ) = − j + deg ( u ⊗ v ) < deg ( u ⊗ a ( − r ) v ) , \operatorname{deg}(u \otimes v) < \operatorname{deg}(a(-r)u \otimes v),\qquad\operatorname{deg}(a(j)u \otimes v)=-j+\operatorname{deg}(u \otimes v)<\operatorname{deg}(u\otimes a(-r)v), deg ( u ⊗ v ) < deg ( a ( − r ) u ⊗ v ) , deg ( a ( j ) u ⊗ v ) = − j + deg ( u ⊗ v ) < deg ( u ⊗ a ( − r ) v ) ,
for all j ≥ 0 j \geq0 j ≥ 0 , we can apply the induction hypothesis after rewriting a ( j ) u a(j)u a ( j ) u and a ( j ) v a(j)v a ( j ) v as linear combinations of PBW basis elements for all j ≥ 0 j \geq0 j ≥ 0 . Hence φ ~ \widetilde{\varphi} φ is well-defined.
Moreover, using the commutation relations in g ^ \widehat{\mathfrak{g}} g and induction on the total degree, one can show that (105) and (106) are compatible with PBW reordering and hold for arbitrary pairs of PBW basis elements. By bilinearity, (105) and (106) hold for all u ∈ V k ( g , λ ) u\in V^k(\mathfrak{g},\lambda) u ∈ V k ( g , λ ) and v ∈ V k ( g , μ ) v\in V^k(\mathfrak{g},\mu) v ∈ V k ( g , μ ) .
Now, we claim that given b = a 1 ( − n 1 ) ⋯ a r ( − n r ) 1 ∈ V k ( g ) b=a^1(-n_1)\cdots a^r(-n_r)\mathbf{1}\in V^k(\mathfrak{g}) b = a 1 ( − n 1 ) ⋯ a r ( − n r ) 1 ∈ V k ( g ) , there exists a rational function F b : V k ( g , λ ) ⊗ C V k ( g , μ ) → ( L ( λ ) ⊗ C L ( μ ) ) [ z , z − 1 , ( z − 1 ) − 1 ] F_b:V^k(\mathfrak{g},\lambda)\otimes_{\mathbb{C}}V^k(\mathfrak{g},\mu)\to(L(\lambda)\otimes_{\mathbb{C}}L(\mu))[z,z^{-1},(z-1)^{-1}] F b : V k ( g , λ ) ⊗ C V k ( g , μ ) → ( L ( λ ) ⊗ C L ( μ )) [ z , z − 1 , ( z − 1 ) − 1 ] , u ⊗ v ↦ F b ( u ⊗ v ) ( z ) u\otimes v\mapsto F_b(u\otimes v)(z) u ⊗ v ↦ F b ( u ⊗ v ) ( z ) , with poles only at z = 0 z=0 z = 0 and z = 1 z=1 z = 1 , such that
ι z = 1 F b ( u ⊗ v ) ( z ) = φ ~ ( Y ( b , z − 1 ) u ⊗ v ) , \iota_{z=1}F_b(u\otimes v)(z)=\widetilde{\varphi}(Y(b,z-1)u\otimes v), ι z = 1 F b ( u ⊗ v ) ( z ) = φ ( Y ( b , z − 1 ) u ⊗ v ) ,
ι z = 0 F b ( u ⊗ v ) ( z ) = φ ~ ( u ⊗ Y ( b , z ) v ) , \iota_{z=0}F_b(u\otimes v)(z)=\widetilde{\varphi}(u\otimes Y(b,z)v), ι z = 0 F b ( u ⊗ v ) ( z ) = φ ( u ⊗ Y ( b , z ) v ) ,
ι z = ∞ F b ( u ⊗ v ) ( z ) ∈ z − wt b ( L ( λ ) ⊗ C L ( μ ) ) [ z − 1 ] . (107) \iota_{z=\infty}F_b(u\otimes v)(z)\in z^{-\operatorname{wt}b}(L(\lambda)\otimes_{\mathbb{C}}L(\mu))[z^{-1}]. \tag*{(107)} ι z = ∞ F b ( u ⊗ v ) ( z ) ∈ z − wt b ( L ( λ ) ⊗ C L ( μ )) [ z − 1 ] . (107)
We use induction on the length r r r of b b b to construct F b F_b F b (107) . When r = 0 r=0 r = 0 and b = 1 b=1 b = 1 , we let F 1 ( u ⊗ v ) ( z ) = φ ~ ( u ⊗ v ) F_1(u\otimes v)(z)=\widetilde{\varphi}(u\otimes v) F 1 ( u ⊗ v ) ( z ) = φ ( u ⊗ v ) . For b = a ( − 1 ) 1 b=a(-1)1 b = a ( − 1 ) 1 , let
F a ( − 1 ) 1 ( u ⊗ v ) ( z ) : = ∑ j ≥ 0 φ ~ ( a ( j ) u ⊗ v ) ( z − 1 ) j + 1 + ∑ j ≥ 0 φ ~ ( u ⊗ a ( j ) v ) z j + 1 . F_{a(-1)1}(u\otimes v)(z):=\sum_{j\geq0}\frac{\widetilde{\varphi}(a(j)u\otimes v)}{(z-1)^{j+1}}+\sum_{j\geq0}\frac{\widetilde{\varphi}(u\otimes a(j)v)}{z^{j+1}}. F a ( − 1 ) 1 ( u ⊗ v ) ( z ) := j ≥ 0 ∑ ( z − 1 ) j + 1 φ ( a ( j ) u ⊗ v ) + j ≥ 0 ∑ z j + 1 φ ( u ⊗ a ( j ) v ) .
In view of (52) , φ ~ \widetilde{\varphi} φ is defined by the ⊙ \odot ⊙ tensor product relation (52) with particular choices ( s , t ) = ( 1 , r ) (s,t)=(1,r) ( s , t ) = ( 1 , r ) and ( s , t ) = ( r + 1 , 0 ) (s,t)=(r+1,0) ( s , t ) = ( r + 1 , 0 ) . Therefore, it is easy to show by induction that φ ~ \widetilde{\varphi} φ preserves the relations (52) . In particular, φ ~ \widetilde{\varphi} φ induces a well-defined linear map
φ : V k ( g , λ ) ⊙ V k ( g , μ ) ⟶ L ( λ ) ⊗ C L ( μ ) \varphi:V^k(\mathfrak{g},\lambda)\odot V^k(\mathfrak{g},\mu)\longrightarrow L(\lambda)\otimes_{\mathbb{C}}L(\mu) φ : V k ( g , λ ) ⊙ V k ( g , μ ) ⟶ L ( λ ) ⊗ C L ( μ )
which is clearly the inverse of the surjective map (102) . This shows (104) is an isomorphism. Furthermore, it follows from (103) that (104) is an isomorphism of g \mathfrak{g} g -modules, where the g \mathfrak{g} g -act on L ( λ ) ⊗ C L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu) L ( λ ) ⊗ C L ( μ ) diagonally. □
The braiding isomorphism on Mod_fd(U ( g ) U(\mathfrak{g}) U ( g ) )
For λ ∈ P + \lambda\in P_+ λ ∈ P + , let h λ = ( λ ∣ λ + 2 ρ ) / 2 κ h_\lambda=(\lambda|\lambda+2\rho)/2\kappa h λ = ( λ ∣ λ + 2 ρ ) /2 κ be the conformal weight of V k ( g , λ ) V^k(\mathfrak{g},\lambda) V k ( g , λ ) . The second proof of (98) only uses the relations (35) and the Segal-Sugawara construction. Hence the same argument shows that the action of [ ω ] ∈ A ( V ) ≅ U ( g ) [\omega]\in A(V)\cong U(\mathfrak{g}) [ ω ] ∈ A ( V ) ≅ U ( g ) on L ( λ ) ⊗ C L ( μ ) ≅ L ( λ ) ⊠ L ( μ ) L(\lambda)\otimes_{\mathbb{C}}L(\mu)\cong L(\lambda)\boxtimes L(\mu) L ( λ ) ⊗ C L ( μ ) ≅ L ( λ ) ⊠ L ( μ ) (104) is given by
[ ω ] . ( u ⊗ v ) = ( h λ + h μ ) u ⊗ v + 1 κ Ω λ , μ ( u ⊗ v ) , u ∈ L ( λ ) , v ∈ L ( μ ) . [\omega].(u\otimes v)=(h_\lambda+h_\mu)u\otimes v+\frac{1}{\kappa}\Omega_{\lambda,\mu}(u\otimes v),\qquad u\in L(\lambda),\ v\in L(\mu). [ ω ] . ( u ⊗ v ) = ( h λ + h μ ) u ⊗ v + κ 1 Ω λ , μ ( u ⊗ v ) , u ∈ L ( λ ) , v ∈ L ( μ ) .
Since Mod f d ( U ( g ) ) \operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) Mod fd ( U ( g )) is semisimple and the canonical map (102) is natural, the isomorphism (104) holds for all objects S , T S,T S , T of Mod f d ( U ( g ) ) \operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) Mod fd ( U ( g )) , i.e., S ⊠ T = S ⊗ C T S\boxtimes T=S\otimes_{\mathbb{C}}T S ⊠ T = S ⊗ C T with the diagonal action of U ( g ) U(\mathfrak{g}) U ( g ) . Then by (77) , the same calculation as in the proof of Theorem 5.4, and the naturality of R \mathcal{R} R , we have the following Corollary.
Corollary 6.3. Let V = V k ( g ) V=V^k(\mathfrak{g}) V = V k ( g ) be as above. The category Mod f d ( U ( g ) ) \operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) Mod fd ( U ( g )) is a braided tensor category, with fusion tensor product given by
S ⊠ T = S ⊗ C T , S\boxtimes T=S\otimes_{\mathbb{C}}T, S ⊠ T = S ⊗ C T ,
on which U ( g ) U(\mathfrak{g}) U ( g ) acts diagonally, with the associativity and unit constraints (81) , and the braiding isomorphism is given by
R S , T : S ⊗ C T → ≅ T ⊗ C S , R S , T ( u ⊗ v ) = ( τ ∘ exp ( π i κ Ω S , T ) ) ( u ⊗ v ) = q Ω T , S ( v ⊗ u ) , (108) \mathcal{R}_{S,T}: S \otimes_{\mathbb{C}} T \xrightarrow{\cong} T \otimes_{\mathbb{C}} S,\qquad\mathcal{R}_{S,T}(u \otimes v)=\left(\tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega_{S,T}\right)\right)(u \otimes v)=q^{\Omega_{T,S}}(v \otimes u), \tag*{(108)} R S , T : S ⊗ C T ≅ T ⊗ C S , R S , T ( u ⊗ v ) = ( τ ∘ exp ( κ π i Ω S , T ) ) ( u ⊗ v ) = q Ω T , S ( v ⊗ u ) , (108)
where q = e π i / κ q=e^{\pi i/\kappa} q = e π i / κ , u ∈ S u\in S u ∈ S , v ∈ T v\in T v ∈ T , and Ω S , T \Omega_{S,T} Ω S , T is the action of Ω \Omega Ω (95) on S ⊗ C T S\otimes_{\mathbb{C}}T S ⊗ C T . Moreover, Φ L : Mod f d ( U ( g ) ) → K L k ( g ) \Phi^{L}:\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g}))\to KL^{k}(\mathfrak{g}) Φ L : Mod fd ( U ( g )) → K L k ( g ) is a braided tensor equivalence.
In this case, there is no quotient in (108) since S ⊠ T = S ⊗ C T S\boxtimes T=S\otimes_{\mathbb{C}}T S ⊠ T = S ⊗ C T , and the braiding isomorphism is given by an R R R -matrix. Indeed, let
R = exp ( π i κ Ω ) = ∑ n ≥ 0 1 n ! ( π i κ ) n Ω n ∈ U ( g ) ⊗ ^ U ( g ) , (109) \mathcal{R}=\exp\left(\frac{\pi i}{\kappa}\Omega\right)=\sum_{n\geq0}\frac{1}{n!}\left(\frac{\pi i}{\kappa}\right)^{n}\Omega^{n}\in U(\mathfrak{g})\widehat{\otimes}U(\mathfrak{g}), \tag*{(109)} R = exp ( κ π i Ω ) = n ≥ 0 ∑ n ! 1 ( κ π i ) n Ω n ∈ U ( g ) ⊗ U ( g ) , (109)
which converges on S ⊗ C T S\otimes_{\mathbb{C}}T S ⊗ C T for all S , T ∈ Mod f d ( U ( g ) ) S,T\in\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) S , T ∈ Mod fd ( U ( g )) , since S ⊗ C T S\otimes_{\mathbb{C}}T S ⊗ C T is finite-dimensional. Then (108) can be written as
R S , T = τ ∘ R ∣ S ⊗ C T , S , T ∈ Mod f d ( U ( g ) ) . \mathcal{R}_{S,T}=\tau\circ\mathcal{R}|_{S\otimes_{\mathbb{C}}T},\qquad S,T\in\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})). R S , T = τ ∘ R ∣ S ⊗ C T , S , T ∈ Mod fd ( U ( g )) .
Relations with Drinfeld’s R R R -matrix
R \mathcal{R} R (109) is a specialization of Drinfeld’s R R R -matrix [17 ] . We first recall some notation and constructions in [42 ] , Section 19. Let ϖ \varpi ϖ be a formal variable, and let A = C [ [ ϖ ] ] A=\mathbb{C}[[\varpi]] A = C [[ ϖ ]] . Let ( a i j ) i , j ∈ I (a_{ij})_{i,j\in I} ( a ij ) i , j ∈ I be the Cartan matrix of g \mathfrak{g} g , and choose integers δ i > 0 \delta_i>0 δ i > 0 such that δ i a i j = δ j a j i \delta_i a_{ij}=\delta_j a_{ji} δ i a ij = δ j a j i for all i , j ∈ I i,j\in I i , j ∈ I . Under our normalization ( θ ∣ θ ) = 2 (\theta|\theta)=2 ( θ ∣ θ ) = 2 , we have
δ i = d 2 ( α i ∣ α i ) , where d = max i ∈ I δ i . \delta_i=\frac{d}{2}(\alpha_i|\alpha_i),\qquad\text{where }d=\max_{i\in I}\delta_i. δ i = 2 d ( α i ∣ α i ) , where d = i ∈ I max δ i .
Let g A = A ⊗ C g \mathfrak{g}_A=A\otimes_{\mathbb{C}}\mathfrak{g} g A = A ⊗ C g , and let U = A ⊗ C U ( g ) U=A\otimes_{\mathbb{C}}U(\mathfrak{g}) U = A ⊗ C U ( g ) be its universal enveloping algebra over A A A , with the coproduct Δ : U → U ⊗ A U \Delta:U\to U\otimes_A U Δ : U → U ⊗ A U given by Δ ( x ) = x ⊗ 1 + 1 ⊗ x \Delta(x)=x\otimes1+1\otimes x Δ ( x ) = x ⊗ 1 + 1 ⊗ x for x ∈ g A x\in\mathfrak{g}_A x ∈ g A . The Casimir element C C C and the element Ω \Omega Ω in (95) can be viewed as elements C A ∈ U C_A\in U C A ∈ U and Ω A ∈ U ⊗ A U \Omega_A\in U\otimes_A U Ω A ∈ U ⊗ A U . The following element is introduced in [42 ] , Section 19.2:
t = d 2 ( Δ ( C A ) − 1 ⊗ C A − C A ⊗ 1 ) = d Ω A . (110) t=\frac{d}{2}\left(\Delta(C_A)-1\otimes C_A-C_A\otimes1\right)=d\Omega_A. \tag*{(110)} t = 2 d ( Δ ( C A ) − 1 ⊗ C A − C A ⊗ 1 ) = d Ω A . (110)
It satisfies Δ ( x ) t = t Δ ( x ) \Delta(x)t=t\Delta(x) Δ ( x ) t = t Δ ( x ) in U ⊗ A U U\otimes_A U U ⊗ A U for all x ∈ g A x\in\mathfrak{g}_A x ∈ g A .
Let D \mathcal{D} D be the category of U U U -modules which are free of finite rank as A A A -modules. It is referred to as the Drinfeld category in [42 ] . In particular, A ⊗ C S A\otimes_{\mathbb{C}}S A ⊗ C S is an object of D \mathcal{D} D for any S ∈ Mod f d ( U ( g ) ) S\in\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) S ∈ Mod fd ( U ( g )) . For V 1 , V 2 ∈ D V_1,V_2\in\mathcal{D} V 1 , V 2 ∈ D , V 1 ⊗ A V 2 V_1\otimes_A V_2 V 1 ⊗ A V 2 is a U U U -module via Δ \Delta Δ , and the series
e π i ϖ t = ∑ n ≥ 0 ϖ n ( π i t ) n n ! e^{\pi i\varpi t}=\sum_{n\geq0}\varpi^{n}\frac{(\pi i t)^{n}}{n!} e π iϖ t = n ≥ 0 ∑ ϖ n n ! ( π i t ) n
converges ϖ \varpi ϖ -adically on V 1 ⊗ A V 2 V_1\otimes_A V_2 V 1 ⊗ A V 2 . Following [42 ] , Section 19.12, the commutativity isomorphism is defined by
C V 1 , V 2 : V 1 ⊗ A V 2 → ≅ V 2 ⊗ A V 1 , x 1 ⊗ x 2 ↦ e π i ϖ t ( x 2 ⊗ x 1 ) . (111) C_{V_1,V_2}:V_1\otimes_A V_2\xrightarrow{\cong}V_2\otimes_A V_1,\qquad x_1\otimes x_2\mapsto e^{\pi i\varpi t}(x_2\otimes x_1). \tag*{(111)} C V 1 , V 2 : V 1 ⊗ A V 2 ≅ V 2 ⊗ A V 1 , x 1 ⊗ x 2 ↦ e π iϖ t ( x 2 ⊗ x 1 ) . (111)
Since t t t commutes with Δ ( g A ) \Delta(\mathfrak{g}_A) Δ ( g A ) , C V 1 , V 2 C_{V_1,V_2} C V 1 , V 2 is a U U U -module isomorphism.
Lemma 6.4. [17 , 42 ] The category D \mathcal{D} D carries an A A A -linear braided monoidal structure. Its tensor product is V 1 ⊗ A V 2 V_1\otimes_A V_2 V 1 ⊗ A V 2 , with the U U U -action induced by Δ \Delta Δ , and its commutativity isomorphism is given by (111) . The associator is defined by comparing asymptotically normalized solutions of the KZ equation at 0 and 1.
Moreover, D \mathcal{D} D is equivalent, as an A A A -linear braided monoidal category, to the category of type-1 modules over the formal Drinfeld–Jimbo quantum group U q ( g ) U_q(\mathfrak{g}) U q ( g ) that are free of finite rank over A A A , where q = e π i ϖ q=e^{\pi i\varpi} q = e π iϖ .
Now we compare C V 1 , V 2 C_{V_1,V_2} C V 1 , V 2 (111) with our braiding R S , T R_{S,T} R S , T (108) . For finite-dimensional U ( g ) U(\mathfrak{g}) U ( g ) -modules S , T S,T S , T , we let V 1 = A ⊗ C S V_1=\mathbb{A}\otimes_{\mathbb{C}}S V 1 = A ⊗ C S and V 2 = A ⊗ C T V_2=\mathbb{A}\otimes_{\mathbb{C}}T V 2 = A ⊗ C T . Then V 1 ⊗ A V 2 = A ⊗ C ( S ⊗ C T ) V_1\otimes_{\mathbb{A}}V_2=\mathbb{A}\otimes_{\mathbb{C}}(S\otimes_{\mathbb{C}}T) V 1 ⊗ A V 2 = A ⊗ C ( S ⊗ C T ) , on which Ω A \Omega_{\mathbb{A}} Ω A acts by Id A ⊗ Ω S , T \operatorname{Id}_{\mathbb{A}}\otimes\Omega_{S,T} Id A ⊗ Ω S , T . For any z ∈ C z\in\mathbb{C} z ∈ C , define
C S , T ( z ) = τ ∘ exp ( z ⋅ π i ( d Ω S , T ) ) : S ⊗ C T ⟶ T ⊗ C S . (112) C_{S,T}(z)=\tau\circ\exp\left(z\cdot\pi i\left(d\Omega_{S,T}\right)\right):S\otimes_{\mathbb{C}}T\longrightarrow T\otimes_{\mathbb{C}}S. \tag*{(112)} C S , T ( z ) = τ ∘ exp ( z ⋅ π i ( d Ω S , T ) ) : S ⊗ C T ⟶ T ⊗ C S . (112)
Since S ⊗ C T S\otimes_{\mathbb{C}}T S ⊗ C T is finite-dimensional, this gives a family of well-defined U ( g ) U(\mathfrak{g}) U ( g ) -module isomorphisms, and the matrix coefficients of C S , T ( z ) C_{S,T}(z) C S , T ( z ) with respect to any bases of S ⊗ C T S\otimes_{\mathbb{C}}T S ⊗ C T and T ⊗ C S T\otimes_{\mathbb{C}}S T ⊗ C S are holomorphic functions of z z z . Let Hol ( C ) \operatorname{Hol}(\mathbb{C}) Hol ( C ) be the ring of holomorphic functions on C \mathbb{C} C , and define the formal Taylor expansion
ι 0 : Hol ( C ) ⟶ A = C [ [ ϖ ] ] , f ⟼ ∑ n ≥ 0 f ( n ) ( 0 ) n ! ϖ n . \iota_0:\operatorname{Hol}(\mathbb{C})\longrightarrow\mathbb{A}=\mathbb{C}[[\varpi]],\qquad f\longmapsto\sum_{n\geq0}\frac{f^{(n)}(0)}{n!}\varpi^n. ι 0 : Hol ( C ) ⟶ A = C [[ ϖ ]] , f ⟼ n ≥ 0 ∑ n ! f ( n ) ( 0 ) ϖ n .
It is an injective ring homomorphism. Applying ι 0 \iota_0 ι 0 to the matrix coefficients of C S , T ( z ) C_{S,T}(z) C S , T ( z ) , we have
ι 0 ( C S , T ) = τ A ∘ exp ( π i ϖ ( d Ω A ) ) = C V 1 , V 2 (113) \iota_0(C_{S,T})=\tau_{\mathbb{A}}\circ\exp\left(\pi i\varpi\left(d\Omega_{\mathbb{A}}\right)\right)=C_{V_1,V_2} \tag*{(113)} ι 0 ( C S , T ) = τ A ∘ exp ( π iϖ ( d Ω A ) ) = C V 1 , V 2 (113)
in view of (110) , (111) , and τ ∘ Ω S , T = Ω T , S ∘ τ \tau\circ\Omega_{S,T}=\Omega_{T,S}\circ\tau τ ∘ Ω S , T = Ω T , S ∘ τ , where τ A = Id A ⊗ τ \tau_{\mathbb{A}}=\operatorname{Id}_{\mathbb{A}}\otimes\tau τ A = Id A ⊗ τ . On the other hand, let z = 1 / ( d κ ) z=1/(d\kappa) z = 1/ ( d κ ) , then by (108) we have
C S , T ( 1 d κ ) = τ ∘ exp ( π i κ Ω S , T ) = R S , T . C_{S,T}\left(\frac{1}{d\kappa}\right)=\tau\circ\exp\left(\frac{\pi i}{\kappa}\Omega_{S,T}\right)=R_{S,T}. C S , T ( d κ 1 ) = τ ∘ exp ( κ π i Ω S , T ) = R S , T .
Therefore, we have the following comparison between our braiding isomorphism and Drinfeld’s commutativity isomorphism.
Proposition 6.5. Let S , T ∈ Mod f d ( U ( g ) ) S,T\in\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) S , T ∈ Mod fd ( U ( g )) , and let V 1 = A ⊗ C S V_1=\mathbb{A}\otimes_{\mathbb{C}}S V 1 = A ⊗ C S and V 2 = A ⊗ C T V_2=\mathbb{A}\otimes_{\mathbb{C}}T V 2 = A ⊗ C T . The family of U ( g ) U(\mathfrak{g}) U ( g ) -module isomorphisms C S , T ( z ) : S ⊗ C T ⟶ T ⊗ C S C_{S,T}(z):S\otimes_{\mathbb{C}}T\longrightarrow T\otimes_{\mathbb{C}}S C S , T ( z ) : S ⊗ C T ⟶ T ⊗ C S in (112) is holomorphic in z ∈ C z\in\mathbb{C} z ∈ C , its formal Taylor expansion at z = 0 z=0 z = 0 is Drinfeld’s commutativity isomorphism C V 1 , V 2 C_{V_1,V_2} C V 1 , V 2 (111) , and its value at z = 1 / ( d κ ) z=1/(d\kappa) z = 1/ ( d κ ) is the braiding isomorphism (108) :
Tay 0 ( C S , T ) = C V 1 , V 2 , C S , T ( 1 d κ ) = R S , T . \operatorname{Tay}_0(C_{S,T})=C_{V_1,V_2},\qquad C_{S,T}\left(\frac{1}{d\kappa}\right)=R_{S,T}. Tay 0 ( C S , T ) = C V 1 , V 2 , C S , T ( d κ 1 ) = R S , T .
In other words , with respect to any bases of S ⊗ C T S\otimes_{\mathbb{C}}T S ⊗ C T and T ⊗ C S T\otimes_{\mathbb{C}}S T ⊗ C S , the matrix coefficients of C V 1 , V 2 C_{V_1,V_2} C V 1 , V 2 are power series in ϖ \varpi ϖ with infinite radius of convergence, and their values at ϖ = 1 / ( d κ ) \varpi=1/(d\kappa) ϖ = 1/ ( d κ ) give R S , T R_{S,T} R S , T . In particular, R \mathcal{R} R (109) is Drinfeld’s R R R -matrix e π i ϖ t e^{\pi i\varpi t} e π iϖ t at ϖ = 1 / ( d κ ) \varpi=1/(d\kappa) ϖ = 1/ ( d κ ) .
Remark 6.6. The associators α \alpha α in the category Mod f d ( U ( g ) ) \operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) Mod fd ( U ( g )) in Corollary 6.3 are transported from K L k ( g ) KL^k(\mathfrak{g}) K L k ( g ) through Φ L \Phi^L Φ L , and R \mathcal{R} R satisfies the hexagon identities with respect to them by Theorem 4.14. Under (104) , they are natural U ( g ) U(\mathfrak{g}) U ( g ) -module automorphisms of S ⊗ C T ⊗ C U S\otimes_{\mathbb{C}}T\otimes_{\mathbb{C}}U S ⊗ C T ⊗ C U for S , T , U ∈ Mod f d ( U ( g ) ) S,T,U\in\operatorname{Mod}_{\mathrm{fd}}(U(\mathfrak{g})) S , T , U ∈ Mod fd ( U ( g )) , but they are not the trivial ones in general. Indeed, for g = s l 2 \mathfrak{g}=\mathfrak{sl}_2 g = sl 2 and S = T = U = L ( ω 1 ) S=T=U=L(\omega_1) S = T = U = L ( ω 1 ) , a direct computation shows that R \mathcal{R} R does not satisfy the hexagon identities with respect to the trivial associativity constraint unless 1 / κ ∈ 2 Z 1/\kappa\in2\mathbb{Z} 1/ κ ∈ 2 Z . This is consistent with Lemma 6.4, where the associativity isomorphism of D \mathcal{D} D is defined by the KZ equation.
References
School of Mathematics, Sichuan University, Chengdu, China,
Email address: zgcui@scu.edu.cn
Department of Mathematics, New Uzbekistan University, Movarounnahr Street 1, Tashkent 100001, Uzbekistan
Email address: j.liu@newuu.uz