Introduction
Filippov introduced δ \delta δ -derivations in the adjoint case [9 ] and studied them for prime Lie algebras [10 ] . Derivations with δ = − 1 \delta= -1 δ = − 1 , called antiderivations, had been considered earlier by Hopkins [15 ] . Related results on generalized derivations and δ \delta δ -derivations appear in [19 , 23 ] . Zohrabi and Zusmanovich [22 ] classified δ \delta δ -derivations of finite-dimensional semisimple Lie algebras with values in finite-dimensional modules and suggested studying the case of infinite-dimensional coefficient modules.
For a complex Lie algebra g \mathfrak{g} g , a g \mathfrak{g} g -module M M M , and δ ∈ C \delta\in\mathbb{C} δ ∈ C , write Der δ ( g , M ) \operatorname{Der}_{\delta}(\mathfrak{g}, M) Der δ ( g , M ) for the space of δ \delta δ -derivations g → M \mathfrak{g} \to M g → M (Definition 2.1). We determine these spaces when g \mathfrak{g} g is finite-dimensional and simple, with no restriction on M M M . We assume δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C} \setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } : the case δ = 0 \delta= 0 δ = 0 is trivial since g \mathfrak{g} g is perfect, and δ = 1 \delta= 1 δ = 1 gives ordinary derivations. In rank at least two, we prove the following theorem.
Theorem A (Theorem 2.7 below). Let g \mathfrak{g} g be a finite-dimensional complex simple Lie algebra of rank at least two, let M M M be any g \mathfrak{g} g -module, and let δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C} \setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } . Then
Der δ ( g , M ) = { Hom g ( g , M ) , δ = 1 2 , 0 , δ ≠ 1 2 . (1) \operatorname{Der}_{\delta}(\mathfrak{g}, M) =
\begin{cases}
\operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g}, M), & \delta= \frac{1}{2},\\
0, & \delta\ne\frac{1}{2}.
\end{cases} \tag*{(1)} Der δ ( g , M ) = { Hom g ( g , M ) , 0 , δ = 2 1 , δ = 2 1 . (1) The action (5) on this space is trivial. If M M M is simple, the space is nonzero precisely when δ = 1 2 \delta= \frac{1}{2} δ = 2 1 and M ≅ g M \cong\mathfrak{g} M ≅ g . In that case, fixing a module isomorphism ψ : g → M \psi: \mathfrak{g} \to M ψ : g → M , one has Der 1 2 ( g , M ) = C ψ \operatorname{Der}_{\frac{1}{2}}(\mathfrak{g}, M) = \mathbb{C}\psi Der 2 1 ( g , M ) = C ψ .
For finite-dimensional M M M , Theorem A is contained in [22 Main Theorem] . For arbitrary M M M , we show that every δ \delta δ -derivation D D D defines the commutative 2 2 2 -cocycle Φ D ( x , y ) = 1 2 ( x D ( y ) + y D ( x ) ) \Phi_D(x,y) = \frac{1}{2}(xD(y) + yD(x)) Φ D ( x , y ) = 2 1 ( x D ( y ) + y D ( x )) (Lemma 2.5 ). Combined with the known vanishing Z c o m m 2 ( g ) = 0 Z^2_{\mathrm{comm}}(\mathfrak{g}) = 0 Z comm 2 ( g ) = 0 in rank at least two [7 Theorem 1.1] , [8 Corollary 2.5] , this yields Theorem A. We also give a short root space proof of this vanishing in Proposition 2.6, making the proof of the theorem self-contained.
In rank one, g ≅ s l 2 \mathfrak{g} \cong\mathfrak{sl}_2 g ≅ sl 2 . Let V ( n ) V(n) V ( n ) denote the simple s l 2 \mathfrak{sl}_2 sl 2 -module of highest weight n ∈ Z ≥ 0 n \in\mathbb{Z}_{\ge0} n ∈ Z ≥ 0 , and normalize the Casimir element Ω \Omega Ω to act on V ( n ) V(n) V ( n ) by n ( n + 2 ) n(n+2) n ( n + 2 ) . For an s l 2 \mathfrak{sl}_2 sl 2 -module X X X and a ∈ C a \in\mathbb{C} a ∈ C , write X ( a ) = { v ∈ X : Ω v = a v } X^{(a)} = \{v \in X : \Omega v = av\} X ( a ) = { v ∈ X : Ω v = a v } , and set
c δ = 4 ( 1 − δ ) δ 2 . c_{\delta} = \frac{4(1-\delta)}{\delta^2}. c δ = δ 2 4 ( 1 − δ ) .
2020 Mathematics Subject Classification . 17B10, 17B20, 17B40.
Key words and phrases. δ \delta δ -derivations, simple Lie algebras, simple modules, Casimir element, translation functors.
Theorem B (Theorem 3.3 and Corollary 3.4 below). Let δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C} \setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } , and let M M M be an s l 2 \mathfrak{sl}_2 sl 2 -module.
(i) There is a natural isomorphism of s l 2 \mathfrak{sl}_2 sl 2 -modules
Der δ ( s l 2 , M ) ≅ ( V ( 2 ) ⊗ M ( c δ ) ) ( ( 1 − 2 δ ) c δ ) (2) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M) \cong\left(V(2) \otimes M^{(c_{\delta})}\right)^{((1-2\delta)c_{\delta})} \tag*{(2)} Der δ ( sl 2 , M ) ≅ ( V ( 2 ) ⊗ M ( c δ ) ) (( 1 − 2 δ ) c δ ) (2) where the tensor product carries the diagonal action.
(ii) If δ ≠ 2 3 \delta\ne\frac{2}{3} δ = 3 2 , then Der δ ( s l 2 , M ) = 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=0 Der δ ( sl 2 , M ) = 0 if and only if M ( c δ ) = 0 M^{(c_{\delta})}=0 M ( c δ ) = 0 . If δ = 2 3 \delta=\frac{2}{3} δ = 3 2 , then Der δ ( s l 2 , M ) = 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=0 Der δ ( sl 2 , M ) = 0 if and only if M ( c δ ) M^{(c_{\delta})} M ( c δ ) is a possibly zero direct sum of copies of V ( 1 ) V(1) V ( 1 ) .
(iii) If M M M is simple with Casimir scalar κ \kappa κ , then the parameters δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C} \setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } for which Der δ ( s l 2 , M ) ≠ 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)\ne0 Der δ ( sl 2 , M ) = 0 are exactly the roots of
κ δ 2 + 4 δ − 4 = 0 (3) \kappa\delta^2+4\delta-4=0 \tag*{(3)} κ δ 2 + 4 δ − 4 = 0 (3) except that δ = 2 3 \delta=\frac{2}{3} δ = 3 2 is omitted when M ≅ V ( 1 ) M\cong V(1) M ≅ V ( 1 ) .
For s l 2 \mathfrak{sl}_2 sl 2 , we write the δ \delta δ -derivation identity as an eigenvalue equation for a linear operator on Hom C ( s l 2 , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M) Hom C ( sl 2 , M ) . A cubic identity for this operator (Lemma 3.1) and the associated spectral projections yield Theorem B.
For simple M M M , we use translation functors to determine the structure of every nonzero derivation module (Theorem 3.7). The parameters δ = 2 3 \delta=\frac{2}{3} δ = 3 2 and δ = 2 \delta=2 δ = 2 correspond to translation onto and out of the singular block, respectively. At δ = 2 \delta=2 δ = 2 , every nonzero derivation module has a simple essential socle, and its quotient by the socle is a finite direct sum of copies of V ( 1 ) V(1) V ( 1 ) . Such a module need not be simple, as shown in Proposition 4.5 and Example 4.11.
We then derive explicit formulas for δ \delta δ -derivations with values in simple s l 2 \mathfrak{sl}_2 sl 2 -modules, using Block’s classification [1 , 4 ] and the notation of [20 ] . These formulas give bases of the derivation spaces for weight modules and a parametrization by evaluation at h h h for non-weight modules. For finite-dimensional modules, we recover the results of Zohrabi and Zusmanovich [22 ] . Proposition 4.6 and Theorem 4.7 apply in particular to the finite-rank C [ h ] \mathbb{C}[h] C [ h ] -torsion-free simple modules described in [12 , 13 , 14 , 21 ] .
Section 2 collects the basic properties of δ \delta δ -derivations and proves Theorem A. Section 3 proves Theorem B and describes the resulting modules by translation functors. Section 4 gives explicit formulas for derivations with values in simple weight and non-weight modules.
The higher rank case
Throughout this paper, all vector spaces, Lie algebras, modules, and tensor products are over C \mathbb{C} C , and scalars in operator identities denote scalar multiples of the identity operator. Let g \mathfrak{g} g be a Lie algebra, let M M M be a g \mathfrak{g} g -module, and let δ ∈ C \delta\in\mathbb{C} δ ∈ C .
Definition 2.1. A linear map D : g → M D:\mathfrak{g}\to M D : g → M is called a δ \delta δ -derivation if
D ( [ x , y ] ) = δ ( x D ( y ) − y D ( x ) ) , ( x , y ∈ g ) . (4) D([x,y])=\delta\left(xD(y)-yD(x)\right),\qquad(x,y\in\mathfrak{g}). \tag*{(4)} D ([ x , y ]) = δ ( x D ( y ) − y D ( x ) ) , ( x , y ∈ g ) . (4)
We denote the space of all δ \delta δ -derivations from g \mathfrak{g} g to M M M by Der δ ( g , M ) \operatorname{Der}_{\delta}(\mathfrak{g},M) Der δ ( g , M ) .
The natural g \mathfrak{g} g -action on Hom C ( g , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{g},M) Hom C ( g , M ) is given by
( z ⋅ D ) ( x ) = z D ( x ) − D ( [ z , x ] ) . (5) (z\cdot D)(x)=zD(x)-D([z,x]). \tag*{(5)} ( z ⋅ D ) ( x ) = z D ( x ) − D ([ z , x ]) . (5)
Proposition 2.2. The space Der δ ( g , M ) \operatorname{Der}_{\delta}(\mathfrak{g},M) Der δ ( g , M ) is a g \mathfrak{g} g -submodule of Hom C ( g , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{g},M) Hom C ( g , M ) . Every g \mathfrak{g} g -module homomorphism F : M → N F:M\to N F : M → N induces a g \mathfrak{g} g -module homomorphism
Der δ ( g , M ) ⟶ Der δ ( g , N ) , D ⟼ F ∘ D . \operatorname{Der}_{\delta}(\mathfrak{g},M)\longrightarrow\operatorname{Der}_{\delta}(\mathfrak{g},N),\qquad D\longmapsto F\circ D. Der δ ( g , M ) ⟶ Der δ ( g , N ) , D ⟼ F ∘ D .
Moreover, Hom g ( g , M ) ⊆ Der 1 2 ( g , M ) \operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M)\subseteq\operatorname{Der}_{\frac{1}{2}}(\mathfrak{g},M) Hom g ( g , M ) ⊆ Der 2 1 ( g , M ) , and the natural g \mathfrak{g} g -action on this subspace is trivial.
Proof. Define ∂ δ : Hom C ( g , M ) → Hom C ( ⋀ 2 g , M ) \partial_{\delta}:\operatorname{Hom}_{\mathbb{C}}(\mathfrak{g},M)\to\operatorname{Hom}_{\mathbb{C}}(\bigwedge^2\mathfrak{g},M) ∂ δ : Hom C ( g , M ) → Hom C ( ⋀ 2 g , M ) by
( ∂ δ D ) ( x , y ) = D ( [ x , y ] ) − δ ( x D ( y ) − y D ( x ) ) . (\partial_{\delta}D)(x,y)=D([x,y])-\delta\left(xD(y)-yD(x)\right). ( ∂ δ D ) ( x , y ) = D ([ x , y ]) − δ ( x D ( y ) − y D ( x ) ) .
The g \mathfrak{g} g -action on Hom C ( ⋀ 2 g , M ) \operatorname{Hom}_{\mathbb{C}}(\bigwedge^2\mathfrak{g},M) Hom C ( ⋀ 2 g , M ) is given by
( z ⋅ φ ) ( x , y ) = z φ ( x , y ) − φ ( [ z , x ] , y ) − φ ( x , [ z , y ] ) . (z\cdot\varphi)(x,y)=z\varphi(x,y)-\varphi([z,x],y)-\varphi(x,[z,y]). ( z ⋅ φ ) ( x , y ) = z φ ( x , y ) − φ ([ z , x ] , y ) − φ ( x , [ z , y ]) .
The Jacobi identity and the module identity imply that
( ∂ δ ( z ⋅ D ) ) ( x , y ) = z ( ∂ δ D ) ( x , y ) − ( ∂ δ D ) ( [ z , x ] , y ) − ( ∂ δ D ) ( x , [ z , y ] ) = ( z ⋅ ∂ δ D ) ( x , y ) . (\partial_{\delta}(z \cdot D))(x,y) = z(\partial_{\delta}D)(x,y) - (\partial_{\delta}D)([z,x],y) - (\partial_{\delta}D)(x,[z,y]) = (z \cdot\partial_{\delta}D)(x,y). ( ∂ δ ( z ⋅ D )) ( x , y ) = z ( ∂ δ D ) ( x , y ) − ( ∂ δ D ) ([ z , x ] , y ) − ( ∂ δ D ) ( x , [ z , y ]) = ( z ⋅ ∂ δ D ) ( x , y ) .
Thus ∂ δ \partial_{\delta} ∂ δ is g \mathfrak{g} g -equivariant, and its kernel Der δ ( g , M ) \operatorname{Der}_{\delta}(\mathfrak{g},M) Der δ ( g , M ) is a g \mathfrak{g} g -submodule. For a module homomorphism F : M → N F:M \to N F : M → N , one has
∂ δ ( F ∘ D ) = F ∘ ∂ δ D , z ⋅ ( F ∘ D ) = F ∘ ( z ⋅ D ) , \partial_{\delta}(F \circ D) = F \circ\partial_{\delta}D, \qquad z \cdot(F \circ D) = F \circ(z \cdot D), ∂ δ ( F ∘ D ) = F ∘ ∂ δ D , z ⋅ ( F ∘ D ) = F ∘ ( z ⋅ D ) ,
so composition with F F F induces the stated homomorphism. Finally, for D ∈ Hom g ( g , M ) D \in\operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M) D ∈ Hom g ( g , M ) , equivariance gives
x D ( y ) − y D ( x ) = 2 D ( [ x , y ] ) , z ⋅ D = 0. xD(y) - yD(x) = 2D([x,y]), \qquad z \cdot D = 0. x D ( y ) − y D ( x ) = 2 D ([ x , y ]) , z ⋅ D = 0.
For δ = 0 \delta= 0 δ = 0 and δ = 1 \delta= 1 δ = 1 , the defining identity (4) gives
Der 0 ( g , M ) = Hom C ( g / [ g , g ] , M ) , Der 1 ( g , M ) = Z 1 ( g , M ) , \operatorname{Der}_{0}(\mathfrak{g},M) = \operatorname{Hom}_{\mathbb{C}}(\mathfrak{g}/[\mathfrak{g},\mathfrak{g}],M), \qquad\operatorname{Der}_{1}(\mathfrak{g},M) = Z^{1}(\mathfrak{g},M), Der 0 ( g , M ) = Hom C ( g / [ g , g ] , M ) , Der 1 ( g , M ) = Z 1 ( g , M ) ,
where Z 1 ( g , M ) Z^{1}(\mathfrak{g},M) Z 1 ( g , M ) is the space of Lie algebra 1-cocycles. Thus Der 0 ( g , M ) = 0 \operatorname{Der}_{0}(\mathfrak{g},M) = 0 Der 0 ( g , M ) = 0 whenever g \mathfrak{g} g is perfect. For ordinary derivations, see [11 , 17 ] ; for derivations of s l 2 \mathfrak{sl}_{2} sl 2 with values in simple modules, see [2 ] . Henceforth, we assume δ ∉ { 0 , 1 } \delta\notin\{0,1\} δ ∈ / { 0 , 1 } .
Lemma 2.3. Let g \mathfrak{g} g be a Lie algebra, let M M M be a g \mathfrak{g} g -module, and suppose δ ∉ { 0 , 1 } \delta\notin\{0,1\} δ ∈ / { 0 , 1 } . Every D ∈ Der δ ( g , M ) D \in\operatorname{Der}_{\delta}(\mathfrak{g},M) D ∈ Der δ ( g , M ) satisfies
[ x , y ] D ( z ) + [ y , z ] D ( x ) + [ z , x ] D ( y ) = 0 ( x , y , z ∈ g ) . (6) [x,y]D(z) + [y,z]D(x) + [z,x]D(y) = 0 \qquad(x,y,z \in\mathfrak{g}). \tag*{(6)} [ x , y ] D ( z ) + [ y , z ] D ( x ) + [ z , x ] D ( y ) = 0 ( x , y , z ∈ g ) . (6)
Proof. Write ∑ c y c \sum_{\mathrm{cyc}} ∑ cyc for the cyclic sum over x , y , z x,y,z x , y , z . The δ \delta δ -derivation and module identities give
∑ c y c z D ( [ x , y ] ) = δ ∑ c y c ( z x D ( y ) − z y D ( x ) ) = δ ∑ c y c [ x , y ] D ( z ) . (7) \sum_{\mathrm{cyc}} zD([x,y]) = \delta\sum_{\mathrm{cyc}} (zxD(y) - zyD(x)) = \delta\sum_{\mathrm{cyc}} [x,y]D(z). \tag*{(7)} cyc ∑ z D ([ x , y ]) = δ cyc ∑ ( z x D ( y ) − z y D ( x )) = δ cyc ∑ [ x , y ] D ( z ) . (7)
Applying D D D to the Jacobi identity, we obtain
0 = ∑ c y c D ( [ [ x , y ] , z ] ) = δ ∑ c y c ( [ x , y ] D ( z ) − z D ( [ x , y ] ) ) = δ ( 1 − δ ) ∑ c y c [ x , y ] D ( z ) . \begin{aligned}
0 &= \sum_{\mathrm{cyc}} D([[x,y],z]) \\
&= \delta\sum_{\mathrm{cyc}} ([x,y]D(z) - zD([x,y])) \\
&= \delta(1-\delta) \sum_{\mathrm{cyc}} [x,y]D(z).
\end{aligned} 0 = cyc ∑ D ([[ x , y ] , z ]) = δ cyc ∑ ([ x , y ] D ( z ) − z D ([ x , y ])) = δ ( 1 − δ ) cyc ∑ [ x , y ] D ( z ) .
Since δ ≠ 0 , 1 \delta\ne0,1 δ = 0 , 1 , this establishes (6) .
We prove Theorem 2.7 using the vanishing theorem for commutative 2-cocycles proved below.
Definition 2.4 ([8 ] ). Let W W W be a vector space. A symmetric bilinear map φ : g × g → W \varphi:\mathfrak{g} \times\mathfrak{g} \to W φ : g × g → W is called a commutative 2-cocycle if
φ ( [ x , y ] , z ) + φ ( [ y , z ] , x ) + φ ( [ z , x ] , y ) = 0 ( x , y , z ∈ g ) . \varphi([x,y],z) + \varphi([y,z],x) + \varphi([z,x],y) = 0 \qquad(x,y,z \in\mathfrak{g}). φ ([ x , y ] , z ) + φ ([ y , z ] , x ) + φ ([ z , x ] , y ) = 0 ( x , y , z ∈ g ) . We write Z c o m m 2 ( g ) Z^{2}_{\mathrm{comm}}(\mathfrak{g}) Z comm 2 ( g ) for the space of C \mathbb{C} C -valued commutative 2-cocycles on g \mathfrak{g} g .
Lemma 2.5. Let g \mathfrak{g} g be a Lie algebra, let M M M be a g \mathfrak{g} g -module, and suppose δ ∉ { 0 , 1 } \delta\notin\{0,1\} δ ∈ / { 0 , 1 } . For any D ∈ Der δ ( g , M ) D \in\operatorname{Der}_{\delta}(\mathfrak{g},M) D ∈ Der δ ( g , M ) , the symmetric bilinear map Φ D : g × g → M \Phi_D:\mathfrak{g} \times\mathfrak{g} \to M Φ D : g × g → M defined by
Φ D ( x , y ) = 1 2 ( x D ( y ) + y D ( x ) ) (8) \Phi_D(x,y) = \frac{1}{2}(xD(y) + yD(x)) \tag*{(8)} Φ D ( x , y ) = 2 1 ( x D ( y ) + y D ( x )) (8) is a commutative 2-cocycle.
Proof. By (7) and Lemma 2.3,
∑ c y c Φ D ( [ x , y ] , z ) = 1 2 ∑ c y c ( [ x , y ] D ( z ) + z D ( [ x , y ] ) ) = 1 + δ 2 ∑ c y c [ x , y ] D ( z ) = 0. \sum_{\mathrm{cyc}} \Phi_D([x,y],z) = \frac{1}{2}\sum_{\mathrm{cyc}}([x,y]D(z) + zD([x,y])) = \frac{1+\delta}{2}\sum_{\mathrm{cyc}}[x,y]D(z) = 0. cyc ∑ Φ D ([ x , y ] , z ) = 2 1 cyc ∑ ([ x , y ] D ( z ) + z D ([ x , y ])) = 2 1 + δ cyc ∑ [ x , y ] D ( z ) = 0.
Proposition 2.6 ([8 Corollary 2.5] ). Let g \mathfrak{g} g be a finite-dimensional complex simple Lie algebra of rank at least two. Then Z c o m m 2 ( g ) = 0 Z^2_{\mathrm{comm}}(\mathfrak{g})=0 Z comm 2 ( g ) = 0 .
Proof. The proof in [8 ] uses the classification of simple Lie algebras with a subalgebra of codimension one. For the reader’s convenience, we give a direct root space proof in the present setting. We use the standard root space facts in [16 ] . Fix a Cartan subalgebra h \mathfrak{h} h with root system R R R and a Weyl-invariant inner product on its real span. Let φ ∈ Z c o m m 2 ( g ) \varphi\in Z^2_{\mathrm{comm}}(\mathfrak{g}) φ ∈ Z comm 2 ( g ) . For x ∈ g α x\in\mathfrak{g}_{\alpha} x ∈ g α , y ∈ g β y\in\mathfrak{g}_{\beta} y ∈ g β , and h , h ′ ∈ h h,h'\in\mathfrak{h} h , h ′ ∈ h , the cocycle identity gives
α ( h ′ ) φ ( h , x ) = α ( h ) φ ( h ′ , x ) , \alpha(h')\varphi(h,x)=\alpha(h)\varphi(h',x), α ( h ′ ) φ ( h , x ) = α ( h ) φ ( h ′ , x ) ,
( α − β ) ( h ) φ ( x , y ) = − φ ( h , [ x , y ] ) . (9) (\alpha-\beta)(h)\varphi(x,y)=-\varphi(h,[x,y]). \tag*{(9)} ( α − β ) ( h ) φ ( x , y ) = − φ ( h , [ x , y ]) . (9)
Every root γ \gamma γ is a sum of two linearly independent roots: by irreducibility and the rank assumption, choose α \alpha α not proportional to γ \gamma γ with ( γ , α ) > 0 (\gamma,\alpha)>0 ( γ , α ) > 0 , and put β = γ − α ∈ R \beta=\gamma-\alpha\in R β = γ − α ∈ R by the root-string property. For this pair, (9) shows that h ↦ φ ( h , [ x , y ] ) h\mapsto\varphi(h,[x,y]) h ↦ φ ( h , [ x , y ]) is proportional to both α + β \alpha+\beta α + β and α − β \alpha-\beta α − β . These linear forms are independent, so the functional vanishes. Since [ g α , g β ] = g γ [\mathfrak{g}_{\alpha},\mathfrak{g}_{\beta}]=\mathfrak{g}_{\gamma} [ g α , g β ] = g γ , we obtain
φ ( h , g γ ) = 0 ( γ ∈ R ) . (10) \varphi(h,\mathfrak{g}_{\gamma})=0 \qquad(\gamma\in R). \tag*{(10)} φ ( h , g γ ) = 0 ( γ ∈ R ) . (10)
The natural action ( z ⋅ φ ) ( u , v ) = − φ ( [ z , u ] , v ) − φ ( u , [ z , v ] ) (z\cdot\varphi)(u,v)=-\varphi([z,u],v)-\varphi(u,[z,v]) ( z ⋅ φ ) ( u , v ) = − φ ([ z , u ] , v ) − φ ( u , [ z , v ]) preserves Z c o m m 2 ( g ) Z^2_{\mathrm{comm}}(\mathfrak{g}) Z comm 2 ( g ) . Indeed, z ⋅ φ z\cdot\varphi z ⋅ φ is symmetric, and the Jacobi identity gives
∑ c y c ( z ⋅ φ ) ( [ u , v ] , w ) = − ∑ c y c ( φ ( [ [ z , u ] , v ] , w ) + φ ( [ v , w ] , [ z , u ] ) + φ ( [ w , [ z , u ] ] , v ) ) = 0. \sum_{\mathrm{cyc}}(z\cdot\varphi)([u,v],w)
=-\sum_{\mathrm{cyc}}\bigl(\varphi([[z,u],v],w)+\varphi([v,w],[z,u])+\varphi([w,[z,u]],v)\bigr)=0. cyc ∑ ( z ⋅ φ ) ([ u , v ] , w ) = − cyc ∑ ( φ ([[ z , u ] , v ] , w ) + φ ([ v , w ] , [ z , u ]) + φ ([ w , [ z , u ]] , v ) ) = 0.
The sums run over cyclic permutations of u , v , w u,v,w u , v , w , and each expression in parentheses is zero by the cocycle identity. For arbitrary x ∈ g α x\in\mathfrak{g}_{\alpha} x ∈ g α and y ∈ g β y\in\mathfrak{g}_{\beta} y ∈ g β , applying (10) to x ⋅ φ x\cdot\varphi x ⋅ φ gives
0 = ( x ⋅ φ ) ( h , y ) = α ( h ) φ ( x , y ) − φ ( h , [ x , y ] ) = ( 2 α − β ) ( h ) φ ( x , y ) , 0=(x\cdot\varphi)(h,y)=\alpha(h)\varphi(x,y)-\varphi(h,[x,y])=(2\alpha-\beta)(h)\varphi(x,y), 0 = ( x ⋅ φ ) ( h , y ) = α ( h ) φ ( x , y ) − φ ( h , [ x , y ]) = ( 2 α − β ) ( h ) φ ( x , y ) ,
where the last equality is (9) . Since R R R is reduced, 2 α − β ≠ 0 2\alpha-\beta\ne0 2 α − β = 0 , and hence φ ( g α , g β ) = 0 \varphi(\mathfrak{g}_{\alpha},\mathfrak{g}_{\beta})=0 φ ( g α , g β ) = 0 for all roots. Equation (9) also gives φ ( h , h ) = 0 \varphi(h,h)=0 φ ( h , h ) = 0 , because h \mathfrak{h} h is spanned by [ g α , g − α ] [\mathfrak{g}_{\alpha},\mathfrak{g}_{-\alpha}] [ g α , g − α ] . Together with (10) , this proves φ = 0 \varphi=0 φ = 0 . □
Theorem 2.7. Let g \mathfrak{g} g be a finite-dimensional complex simple Lie algebra of rank at least two, let M M M be any g \mathfrak{g} g -module, and let δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C}\setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } . Then
Der δ ( g , M ) = { Hom g ( g , M ) , δ = 1 2 , 0 , δ ≠ 1 2 . \operatorname{Der}_{\delta}(\mathfrak{g},M)=
\begin{cases}
\operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M), & \delta=\frac{1}{2},\\
0, & \delta\ne\frac{1}{2}.
\end{cases} Der δ ( g , M ) = { Hom g ( g , M ) , 0 , δ = 2 1 , δ = 2 1 . The action (5) on this space is trivial. If M M M is simple, the space is nonzero precisely when δ = 1 2 \delta=\frac{1}{2} δ = 2 1 and M ≅ g M\cong\mathfrak{g} M ≅ g . In that case, fixing a module isomorphism ψ : g → M \psi:\mathfrak{g}\to M ψ : g → M , one has Der 1 2 ( g , M ) = C ψ \operatorname{Der}_{\frac{1}{2}}(\mathfrak{g},M)=\mathbb{C}\psi Der 2 1 ( g , M ) = C ψ .
Proof. Let D ∈ Der δ ( g , M ) D\in\operatorname{Der}_{\delta}(\mathfrak{g},M) D ∈ Der δ ( g , M ) . For every ℓ ∈ M ∗ \ell\in M^* ℓ ∈ M ∗ , Lemma 2.5 and Proposition 2.6 imply that ℓ ∘ Φ D = 0 \ell\circ\Phi_D=0 ℓ ∘ Φ D = 0 . Since the algebraic dual M ∗ M^* M ∗ separates points, we have Φ D = 0 \Phi_D=0 Φ D = 0 . Thus x D ( y ) = − y D ( x ) xD(y)=-yD(x) x D ( y ) = − y D ( x ) , and (4) yields
x D ( y ) = 1 2 δ D ( [ x , y ] ) ( x , y ∈ g ) . (11) xD(y)=\frac{1}{2\delta}D([x,y]) \qquad(x,y\in\mathfrak{g}). \tag*{(11)} x D ( y ) = 2 δ 1 D ([ x , y ]) ( x , y ∈ g ) . (11) When δ = 1 2 \delta=\frac{1}{2} δ = 2 1 , equation (11) says that D D D is a g \mathfrak{g} g -module homomorphism. The reverse inclusion and the triviality of the natural action follow from Proposition 2.2.
Suppose now that δ ≠ 1 2 \delta\ne\frac{1}{2} δ = 2 1 . Using (11) and the Jacobi identity, we obtain
[ x , y ] D ( z ) = x ( y D ( z ) ) − y ( x D ( z ) ) = 1 2 δ ( x D ( [ y , z ] ) − y D ( [ x , z ] ) ) = 1 4 δ 2 D ( [ x , [ y , z ] ] − [ y , [ x , z ] ] ) = 1 4 δ 2 D ( [ [ x , y ] , z ] ) . \begin{aligned}
[x,y]D(z)&=x(yD(z))-y(xD(z))\\
&=\frac{1}{2\delta}\bigl(xD([y,z])-yD([x,z])\bigr)\\
&=\frac{1}{4\delta^2}D\bigl([x,[y,z]]-[y,[x,z]]\bigr)\\
&=\frac{1}{4\delta^2}D([[x,y],z]).
\end{aligned} [ x , y ] D ( z ) = x ( y D ( z )) − y ( x D ( z )) = 2 δ 1 ( x D ([ y , z ]) − y D ([ x , z ]) ) = 4 δ 2 1 D ( [ x , [ y , z ]] − [ y , [ x , z ]] ) = 4 δ 2 1 D ([[ x , y ] , z ]) .
On the other hand, (11) also gives [ x , y ] D ( z ) = ( 2 δ ) − 1 D ( [ [ x , y ] , z ] ) [x,y]D(z)=(2\delta)^{-1}D([[x,y],z]) [ x , y ] D ( z ) = ( 2 δ ) − 1 D ([[ x , y ] , z ]) . Hence
( 2 δ − 1 ) D ( [ [ x , y ] , z ] ) = 0 ( x , y , z ∈ g ) . (2\delta-1)D([[x,y],z])=0 \qquad(x,y,z\in\mathfrak{g}). ( 2 δ − 1 ) D ([[ x , y ] , z ]) = 0 ( x , y , z ∈ g ) .
Since g \mathfrak{g} g is perfect, [ [ g , g ] , g ] = g [[\mathfrak{g},\mathfrak{g}],\mathfrak{g}]=\mathfrak{g} [[ g , g ] , g ] = g , and hence D = 0 D=0 D = 0 .
Finally, if M M M is simple, any nonzero homomorphism g → M \mathfrak{g}\to M g → M is an isomorphism because the adjoint module g \mathfrak{g} g is simple. For a fixed isomorphism ψ : g → M \psi:\mathfrak{g}\to M ψ : g → M , Schur’s lemma gives
Hom g ( g , M ) = ψ ∘ End g ( g ) = C ψ . \operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M)=\psi\circ\operatorname{End}_{\mathfrak{g}}(\mathfrak{g})=\mathbb{C}\psi. Hom g ( g , M ) = ψ ∘ End g ( g ) = C ψ .
Remark 2.8. For finite-dimensional M M M , Theorem 2.7 is contained in [22 Main Theorem] , including the case δ = 1 2 \delta=\frac{1}{2} δ = 2 1 . For arbitrary M M M , the vanishing for δ ∉ { 0 , 1 , 1 2 } \delta\notin\{0,1,\frac{1}{2}\} δ ∈ / { 0 , 1 , 2 1 } also follows from [23 Lemma 4.4] and [8 proof of Corollary 2.5] . This argument does not cover δ = 1 2 \delta=\frac{1}{2} δ = 2 1 , which is excluded in the cited lemma.
The commutative 2-cocycle argument above also treats δ = 1 2 \delta=\frac{1}{2} δ = 2 1 for arbitrary M M M . Lemma 2.5, combined with the known vanishing Z c o m m 2 ( g ) = 0 Z^{2}_{\mathrm{comm}}(\mathfrak{g})=0 Z comm 2 ( g ) = 0 [7 Theorem 1.1] , [8 Corollary 2.5] , gives Der 1 2 ( g , M ) = Hom g ( g , M ) \operatorname{Der}_{\frac{1}{2}}(\mathfrak{g},M)=\operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M) Der 2 1 ( g , M ) = Hom g ( g , M ) . Proposition 2.6 supplies a direct root space proof of this vanishing, so the proof is self-contained.
Remark 2.9. For a geometric interpretation, let G G G be a connected complex Lie group with finite-dimensional Lie algebra g \mathfrak{g} g and left Maurer–Cartan form θ \theta θ . For a representation ρ : g → End C M \rho:\mathfrak{g}\to\operatorname{End}_{\mathbb{C}}M ρ : g → End C M , set E = O G ⊗ C M \mathcal{E}=\mathcal{O}_{G}\otimes_{\mathbb{C}}M E = O G ⊗ C M , where O G \mathcal{O}_{G} O G is the sheaf of holomorphic functions on G G G . Equip T G TG T G with the left-invariant connection ∇ x G y = 1 2 [ x , y ] \nabla^{G}_{x}y=\frac{1}{2}[x,y] ∇ x G y = 2 1 [ x , y ] for left-invariant vector fields x , y ∈ g x,y\in\mathfrak{g} x , y ∈ g , and endow E \mathcal{E} E with the connection ∇ δ = d + δ ρ ( θ ) \nabla^{\delta}=d+\delta\rho(\theta) ∇ δ = d + δ ρ ( θ ) . Let ∇ ^ δ \widehat{\nabla}^{\delta} ∇ δ be the induced connection on T ∗ G ⊗ E T^{*}G\otimes\mathcal{E} T ∗ G ⊗ E , and let d ∇ δ d^{\nabla^{\delta}} d ∇ δ denote the covariant exterior derivative. For a linear map D : g → M D:\mathfrak{g}\to M D : g → M , define the left-invariant E \mathcal{E} E -valued 1-form α D = D ∘ θ \alpha_{D}=D\circ\theta α D = D ∘ θ . Evaluating on left-invariant vector fields x , y ∈ g x,y\in\mathfrak{g} x , y ∈ g yields
( d ∇ δ α D ) ( x , y ) = δ ( x D ( y ) − y D ( x ) ) − D ( [ x , y ] ) , (d^{\nabla^{\delta}}\alpha_{D})(x,y)=\delta(xD(y)-yD(x))-D([x,y]), ( d ∇ δ α D ) ( x , y ) = δ ( x D ( y ) − y D ( x )) − D ([ x , y ]) ,
( ∇ ^ x δ α D ) ( y ) = δ x D ( y ) − 1 2 D ( [ x , y ] ) . (\widehat{\nabla}^{\delta}_{x}\alpha_{D})(y)=\delta xD(y)-\frac{1}{2}D([x,y]). ( ∇ x δ α D ) ( y ) = δ x D ( y ) − 2 1 D ([ x , y ]) .
Thus D D D is a δ \delta δ -derivation precisely when α D \alpha_{D} α D is covariantly closed, or equivalently when it satisfies the Codazzi equation
( ∇ ^ x δ α D ) ( y ) = ( ∇ ^ y δ α D ) ( x ) . (\widehat{\nabla}^{\delta}_{x}\alpha_{D})(y)=(\widehat{\nabla}^{\delta}_{y}\alpha_{D})(x). ( ∇ x δ α D ) ( y ) = ( ∇ y δ α D ) ( x ) .
At δ = 1 2 \delta=\frac{1}{2} δ = 2 1 , the form α D \alpha_{D} α D is parallel if and only if D ∈ Hom g ( g , M ) D\in\operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M) D ∈ Hom g ( g , M ) . By Theorem 2.7, every such left-invariant Codazzi form is parallel when g \mathfrak{g} g is simple of rank at least two.
The rank one case
For s l 2 \mathfrak{sl}_{2} sl 2 , the derivation identity is an eigenvalue equation on Hom C ( s l 2 , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_{2},M) Hom C ( sl 2 , M ) . We use this equation to determine when the derivation space is nonzero and to describe its module structure via translation functors.
Casimir identities
Fix a basis e , h , f e,h,f e , h , f of s l 2 \mathfrak{sl}_{2} sl 2 with relations
[ h , e ] = 2 e , [ h , f ] = − 2 f , [ e , f ] = h . (12) [h,e]=2e,\qquad[h,f]=-2f,\qquad[e,f]=h. \tag*{(12)} [ h , e ] = 2 e , [ h , f ] = − 2 f , [ e , f ] = h . (12)
We use the Casimir element
Ω = 2 ( e f + f e ) + h 2 . \Omega=2(ef+fe)+h^{2}. Ω = 2 ( e f + f e ) + h 2 .
The Casimir element used in [20 ] is Ω + 1 \Omega+1 Ω + 1 . For the remainder of the paper, set
t = 2 δ − 1 , c = t 2 − 1 , c ′ = ( t − 2 ) 2 − 1. (13) t=\frac{2}{\delta}-1,\qquad c=t^{2}-1,\qquad c'=(t-2)^{2}-1. \tag*{(13)} t = δ 2 − 1 , c = t 2 − 1 , c ′ = ( t − 2 ) 2 − 1. (13)
Thus δ = 2 t + 1 \delta=\frac{2}{t+1} δ = t + 1 2 , c = c θ c=c_{\theta} c = c θ , and c ′ = ( 1 − 2 δ ) c θ c'=(1-2\delta)c_{\theta} c ′ = ( 1 − 2 δ ) c θ . The values δ = 2 , 2 3 \delta=2,\frac{2}{3} δ = 2 , 3 2 correspond to t = 0 , 2 t=0,2 t = 0 , 2 , respectively. The assumption δ ≠ 0 , 1 \delta\ne0,1 δ = 0 , 1 gives t ∉ { − 1 , 1 } t\notin\{-1,1\} t ∈ / { − 1 , 1 } .
For an s l 2 \mathfrak{sl}_{2} sl 2 -module X X X and a ∈ C a\in\mathbb{C} a ∈ C , write X ( a ) = { v ∈ X : Ω v = a v } X^{(a)}=\{v\in X:\Omega v=av\} X ( a ) = { v ∈ X : Ω v = a v } . As above, V ( n ) V(n) V ( n ) denotes the simple module of highest weight n ∈ Z ≥ 0 n\in\mathbb{Z}_{\geq0} n ∈ Z ≥ 0 .
For an arbitrary s l 2 \mathfrak{sl}_{2} sl 2 -module M M M , put H = Hom C ( s l 2 , M ) H=\operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_{2},M) H = Hom C ( sl 2 , M ) . Denote by Ω M \Omega_{M} Ω M the Casimir operator on M M M , and also the operator on H H H given by ( Ω M D ) ( x ) = Ω D ( x ) (\Omega_{M}D)(x)=\Omega D(x) ( Ω M D ) ( x ) = Ω D ( x ) . We reserve Ω \Omega Ω on H H H for the Casimir operator associated with the natural action (5) . The linear map Λ 2 s l 2 → s l 2 \Lambda^{2}\mathfrak{sl}_{2}\to\mathfrak{sl}_{2} Λ 2 sl 2 → sl 2 , x ∧ y ↦ [ x , y ] x \wedge y \mapsto[x,y] x ∧ y ↦ [ x , y ] , is an isomorphism: it is surjective because [ s l 2 , s l 2 ] = s l 2 [\mathfrak{sl}_2,\mathfrak{sl}_2]=\mathfrak{sl}_2 [ sl 2 , sl 2 ] = sl 2 , and both spaces have dimension three. Thus the identity
( S D ) ( [ x , y ] ) = x D ( y ) − y D ( x ) (14) (SD)([x,y])=xD(y)-yD(x) \tag*{(14)} ( S D ) ([ x , y ]) = x D ( y ) − y D ( x ) (14)
defines a linear operator S S S on H H H . By definition, D D D is a δ \delta δ -derivation if and only if S D = δ − 1 D SD=\delta^{-1}D S D = δ − 1 D , or equivalently, 2 S D = ( t + 1 ) D 2SD=(t+1)D 2 S D = ( t + 1 ) D . The bracket and the module action are s l 2 \mathfrak{sl}_2 sl 2 -equivariant, so S S S is also equivariant. In the coordinates D ↦ ( D ( e ) , D ( h ) , D ( f ) ) T D\mapsto(D(e),D(h),D(f))^{\mathsf{T}} D ↦ ( D ( e ) , D ( h ) , D ( f ) ) T , its matrix is
S = ( h 2 − e 2 0 − f 0 e 0 f 2 − h 2 ) . (15) S=\begin{pmatrix}
\frac{h}{2} & -\frac{e}{2} & 0 \\
-f & 0 & e \\
0 & \frac{f}{2} & -\frac{h}{2}
\end{pmatrix}. \tag*{(15)} S = 2 h − f 0 − 2 e 0 2 f 0 e − 2 h . (15)
Each entry is the action of an element of U ( s l 2 ) U(\mathfrak{sl}_2) U ( sl 2 ) , so centrality of the Casimir element implies Ω M S = S Ω M \Omega_M S=S\Omega_M Ω M S = S Ω M . The derivation identity is bilinear and alternating, so it suffices to check it on ( h , e ) , ( h , f ) , ( e , f ) (h,e),(h,f),(e,f) ( h , e ) , ( h , f ) , ( e , f ) , which gives
( h − t − 1 ) D ( e ) = e D ( h ) , ( h + t + 1 ) D ( f ) = f D ( h ) , ( t + 1 ) D ( h ) = 2 e D ( f ) − 2 f D ( e ) . (16) \begin{aligned}
(h-t-1)D(e)&=eD(h),\\
(h+t+1)D(f)&=fD(h),\\
(t+1)D(h)&=2eD(f)-2fD(e). \tag*{(16)}
\end{aligned} ( h − t − 1 ) D ( e ) ( h + t + 1 ) D ( f ) ( t + 1 ) D ( h ) = eD ( h ) , = f D ( h ) , = 2 eD ( f ) − 2 f D ( e ) . (16)
Lemma 3.1. On Hom C ( s l 2 , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M) Hom C ( sl 2 , M ) , one has
( S − 1 ) ( 4 S 2 − 4 S − Ω M ) = 0 , Ω = Ω M + 8 − 8 S . (17) (S-1)(4S^2-4S-\Omega_M)=0,\qquad\Omega=\Omega_M+8-8S. \tag*{(17)} ( S − 1 ) ( 4 S 2 − 4 S − Ω M ) = 0 , Ω = Ω M + 8 − 8 S . (17) As operators on M M M , the sums of the diagonal entries in the matrices of S S S and S 2 S^2 S 2 are 0 0 0 and Ω M 2 \frac{\Omega_M}{2} 2 Ω M , respectively.
Proof. To factor 4 S 2 − 4 S − Ω M 4S^2-4S-\Omega_M 4 S 2 − 4 S − Ω M through M M M , define ι : M → H \iota:M\to H ι : M → H and π : H → M \pi:H\to M π : H → M by
( ι m ) ( x ) = x m , π ( D ) = 2 f D ( e ) + h D ( h ) + 2 e D ( f ) . (\iota m)(x)=xm,\qquad\pi(D)=2fD(e)+hD(h)+2eD(f). ( ι m ) ( x ) = x m , π ( D ) = 2 f D ( e ) + h D ( h ) + 2 eD ( f ) . The matrices of ι \iota ι and π \pi π are the column ( e , h , f ) T (e,h,f)^{\mathsf{T}} ( e , h , f ) T and the row ( 2 f , h , 2 e ) (2f,h,2e) ( 2 f , h , 2 e ) . The module identity gives ( S ι m ) ( [ x , y ] ) = ( x y − y x ) m = [ x , y ] m (S\iota m)([x,y])=(xy-yx)m=[x,y]m ( S ι m ) ([ x , y ]) = ( x y − y x ) m = [ x , y ] m , so S ι = ι S\iota=\iota S ι = ι . For the other two compositions, the Lie relations give
π S = ( f h − h f , e f − f e , h e − e h ) = ( 2 f , h , 2 e ) = π , π ι = 2 f e + h 2 + 2 e f = Ω M . \pi S=(fh-hf,ef-fe,he-eh)=(2f,h,2e)=\pi,\qquad\pi\iota=2fe+h^2+2ef=\Omega_M. π S = ( f h − h f , e f − f e , h e − e h ) = ( 2 f , h , 2 e ) = π , π ι = 2 f e + h 2 + 2 e f = Ω M . Matrix multiplication in (15) gives
4 S 2 − 4 S − Ω M = − ( 2 e f e h 2 e 2 2 h f h 2 2 h e 2 f 2 f h 2 f e ) . 4S^2-4S-\Omega_M
=-\begin{pmatrix}
2ef & eh & 2e^2\\
2hf & h^2 & 2he\\
2f^2 & fh & 2fe
\end{pmatrix}. 4 S 2 − 4 S − Ω M = − 2 e f 2 h f 2 f 2 e h h 2 f h 2 e 2 2 h e 2 f e . For example, the ( 1 , 1 ) (1,1) ( 1 , 1 ) -entry is h 2 + 2 e f − 2 h − Ω M = − 2 e f h^2+2ef-2h-\Omega_M=-2ef h 2 + 2 e f − 2 h − Ω M = − 2 e f , and the ( 1 , 2 ) (1,2) ( 1 , 2 ) -entry is − h e + 2 e = − e h -he+2e=-eh − h e + 2 e = − e h . Thus
4 S 2 − 4 S − Ω M = − ι π . (18) 4S^2-4S-\Omega_M=-\iota\pi. \tag*{(18)} 4 S 2 − 4 S − Ω M = − ι π . (18) Since ( S − 1 ) ι = 0 (S-1)\iota=0 ( S − 1 ) ι = 0 , multiplying (18) on the left by S − 1 S-1 S − 1 proves the first identity. The natural action on H H H is represented by
h ↦ ( h − 2 0 0 0 h 0 0 0 h + 2 ) , e ↦ ( e 0 0 2 e 0 0 − 1 e ) , f ↦ ( f 1 0 0 f − 2 0 0 f ) . h\mapsto\begin{pmatrix}
h-2&0&0\\
0&h&0\\
0&0&h+2
\end{pmatrix},\qquad
e\mapsto\begin{pmatrix}
e&0&0\\
2&e&0\\
0&-1&e
\end{pmatrix},\qquad
f\mapsto\begin{pmatrix}
f&1&0\\
0&f&-2\\
0&0&f
\end{pmatrix}. h ↦ h − 2 0 0 0 h 0 0 0 h + 2 , e ↦ e 2 0 0 e − 1 0 0 e , f ↦ f 0 0 1 f 0 0 − 2 f . Substitution into Ω = 4 f e + h 2 + 2 h \Omega=4fe+h^2+2h Ω = 4 f e + h 2 + 2 h gives the matrix
( Ω M + 8 − 4 h 4 e 0 8 f Ω M + 8 − 8 e 0 − 4 f Ω M + 8 + 4 h ) = Ω M + 8 − 8 S , \begin{pmatrix}
\Omega_M+8-4h&4e&0\\
8f&\Omega_M+8&-8e\\
0&-4f&\Omega_M+8+4h
\end{pmatrix}
=\Omega_M+8-8S, Ω M + 8 − 4 h 8 f 0 4 e Ω M + 8 − 4 f 0 − 8 e Ω M + 8 + 4 h = Ω M + 8 − 8 S , which proves the second identity. The diagonal entries of S 2 S^2 S 2 are
h 2 4 + e f 2 , f e + e f 2 , f e 2 + h 2 4 , \frac{h^2}{4}+\frac{ef}{2},\qquad\frac{fe+ef}{2},\qquad\frac{fe}{2}+\frac{h^2}{4}, 4 h 2 + 2 e f , 2 f e + e f , 2 f e + 4 h 2 , whose sum is Ω M 2 \frac{\Omega_M}{2} 2 Ω M . The sum of the diagonal entries of S S S is zero by (15) . □
The invariant form normalized by ( e , f ) = 1 (e,f)=1 ( e , f ) = 1 and ( h , h ) = 2 (h,h)=2 ( h , h ) = 2 identifies H H H with s l 2 ∗ ⊗ M ≅ V ( 2 ) ⊗ M \mathfrak{sl}_2^* \otimes M \cong V(2)\otimes M sl 2 ∗ ⊗ M ≅ V ( 2 ) ⊗ M .
Remark 3.2. For dual bases ( x i ) (x_i) ( x i ) and ( x i ) (x^i) ( x i ) with respect to this form, the split Casimir satisfies
− 2 S = ∑ i ad ( x i ) ⊗ x i . -2S=\sum_i \operatorname{ad}(x_i)\otimes x^i. − 2 S = i ∑ ad ( x i ) ⊗ x i .
Indeed, the operator on the right sends D D D to the map x ↦ − ∑ i x i D ( [ x i , x ] ) x\mapsto-\sum_i x^iD([x_i,x]) x ↦ − ∑ i x i D ([ x i , x ]) , so the identity follows from (15) . Characteristic identities for finite-dimensional modules and for modules with a central character are treated in [5 ] and [18 ] , respectively. The calculation above applies to arbitrary modules.
Reduction to Casimir eigenspaces
Theorem 3.3. Let δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C}\setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } , let M M M be an s l 2 \mathfrak{sl}_2 sl 2 -module, and put c δ = 4 ( 1 − δ ) δ 2 c_\delta=\frac{4(1-\delta)}{\delta^2} c δ = δ 2 4 ( 1 − δ ) . There is a natural isomorphism of s l 2 \mathfrak{sl}_2 sl 2 -modules
Der δ ( s l 2 , M ) ≅ ( V ( 2 ) ⊗ M ( c δ ) ) ( ( 1 − 2 δ ) c δ ) . (19) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)\cong\left(V(2)\otimes M^{(c_\delta)}\right)^{((1-2\delta)c_\delta)}. \tag*{(19)} Der δ ( sl 2 , M ) ≅ ( V ( 2 ) ⊗ M ( c δ ) ) (( 1 − 2 δ ) c δ ) . (19) where the tensor product carries the diagonal action. Moreover:
(i) If δ ≠ 2 3 \delta\ne\frac{2}{3} δ = 3 2 , then Der δ ( s l 2 , M ) = 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=0 Der δ ( sl 2 , M ) = 0 if and only if M ( c δ ) = 0 M^{(c_\delta)}=0 M ( c δ ) = 0 .
(ii) If δ = 2 3 \delta=\frac{2}{3} δ = 3 2 , then Der δ ( s l 2 , M ) = 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=0 Der δ ( sl 2 , M ) = 0 if and only if M ( c δ ) M^{(c_\delta)} M ( c δ ) is a possibly zero direct sum of copies of V ( 1 ) V(1) V ( 1 ) .
Proof. Put N = M ( c δ ) N=M^{(c_\delta)} N = M ( c δ ) . In the notation of (13) , c = c δ c=c_\delta c = c δ and c ′ = ( 1 − 2 δ ) c δ c'=(1-2\delta)c_\delta c ′ = ( 1 − 2 δ ) c δ . Since the Casimir element is central, N N N is a submodule. For D ∈ Der δ ( s l 2 , M ) D\in\operatorname{Der}_{\delta}(\mathfrak{sl}_2,M) D ∈ Der δ ( sl 2 , M ) , use S D = t + 1 2 D SD=\frac{t+1}{2}D S D = 2 t + 1 D and Ω M S = S Ω M \Omega_MS=S\Omega_M Ω M S = S Ω M in (17) to obtain
0 = 2 ( S − 1 ) ( 4 S 2 − 4 S − Ω M ) D = ( t − 1 ) ( ( t + 1 ) 2 − 2 ( t + 1 ) − Ω M ) D = ( t − 1 ) ( c − Ω M ) D . \begin{aligned}
0=2(S-1)(4S^2-4S-\Omega_M)D \\
&=(t-1)((t+1)^2-2(t+1)-\Omega_M)D \\
&=(t-1)(c-\Omega_M)D.
\end{aligned} 0 = 2 ( S − 1 ) ( 4 S 2 − 4 S − Ω M ) D = ( t − 1 ) (( t + 1 ) 2 − 2 ( t + 1 ) − Ω M ) D = ( t − 1 ) ( c − Ω M ) D .
Since t ≠ 1 t\ne1 t = 1 , every value of D D D lies in N N N , so Der δ ( s l 2 , M ) = Der δ ( s l 2 , N ) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=\operatorname{Der}_{\delta}(\mathfrak{sl}_2,N) Der δ ( sl 2 , M ) = Der δ ( sl 2 , N ) . On Hom C ( s l 2 , N ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,N) Hom C ( sl 2 , N ) , the second identity becomes
Ω − c ′ = − 8 ( S − t + 1 2 ) , c + 8 − c ′ = 4 ( t + 1 ) . \Omega-c'=-8\left(S-\frac{t+1}{2}\right),\qquad c+8-c'=4(t+1). Ω − c ′ = − 8 ( S − 2 t + 1 ) , c + 8 − c ′ = 4 ( t + 1 ) .
Thus Ω − c ′ \Omega-c' Ω − c ′ and S − t + 1 2 S-\frac{t+1}{2} S − 2 t + 1 have the same kernel. The invariant form fixed above gives Hom C ( s l 2 , N ) ≅ s l 2 ∗ ⊗ N ≅ V ( 2 ) ⊗ N \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,N)\cong\mathfrak{sl}_2^*\otimes N\cong V(2)\otimes N Hom C ( sl 2 , N ) ≅ sl 2 ∗ ⊗ N ≅ V ( 2 ) ⊗ N . Restricting this identification to the c ′ c' c ′ -eigenspace of Ω \Omega Ω gives
Der δ ( s l 2 , M ) = Der δ ( s l 2 , N ) ≅ ( V ( 2 ) ⊗ N ) ( c ′ ) . (20) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=\operatorname{Der}_{\delta}(\mathfrak{sl}_2,N)\cong(V(2)\otimes N)^{(c')}. \tag*{(20)} Der δ ( sl 2 , M ) = Der δ ( sl 2 , N ) ≅ ( V ( 2 ) ⊗ N ) ( c ′ ) . (20)
The basis dual to e , h , f e,h,f e , h , f is f , h 2 , e f,\frac{h}{2},e f , 2 h , e , so this map is
D ⟼ f ⊗ D ( e ) + 1 2 h ⊗ D ( h ) + e ⊗ D ( f ) . (21) D\longmapsto f\otimes D(e)+\frac{1}{2}h\otimes D(h)+e\otimes D(f). \tag*{(21)} D ⟼ f ⊗ D ( e ) + 2 1 h ⊗ D ( h ) + e ⊗ D ( f ) . (21)
The inverse sends f ⊗ u + 1 2 h ⊗ v + e ⊗ w f\otimes u+\frac{1}{2}h\otimes v+e\otimes w f ⊗ u + 2 1 h ⊗ v + e ⊗ w to the map with values D ( e ) = u D(e)=u D ( e ) = u , D ( h ) = v D(h)=v D ( h ) = v , and D ( f ) = w D(f)=w D ( f ) = w . These identifications are s l 2 \mathfrak{sl}_2 sl 2 -equivariant and commute with homomorphisms in the second factor, so the isomorphism is natural.
It remains to determine when ker ( S − t + 1 2 ) \ker\left(S-\frac{t+1}{2}\right) ker ( S − 2 t + 1 ) is nonzero. On Hom C ( s l 2 , N ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,N) Hom C ( sl 2 , N ) , the first identity in (17) factors as
( S − 1 ) ( S − t + 1 2 ) ( S − 1 − t 2 ) = 0. (22) (S-1)\left(S-\frac{t+1}{2}\right)\left(S-\frac{1-t}{2}\right)=0. \tag*{(22)} ( S − 1 ) ( S − 2 t + 1 ) ( S − 2 1 − t ) = 0. (22)
Suppose first that t ≠ 0 t\ne0 t = 0 , and put
P = ( S − 1 ) ( 2 S + t − 1 ) t ( t − 1 ) . (23) P=\frac{(S-1)(2S+t-1)}{t(t-1)}. \tag*{(23)} P = t ( t − 1 ) ( S − 1 ) ( 2 S + t − 1 ) . (23)
The three roots in (22) are distinct because t ∉ { − 1 , 0 , 1 } t\notin\{-1,0,1\} t ∈ / { − 1 , 0 , 1 } . The polynomial used in (23) takes the values 0 , 1 , 0 0,1,0 0 , 1 , 0 at 1 , t + 1 2 , 1 − t 2 1,\frac{t+1}{2},\frac{1-t}{2} 1 , 2 t + 1 , 2 1 − t , respectively. Hence
P 2 = P , ( S − t + 1 2 ) P = 0 , P D = D ( D ∈ ker ( S − t + 1 2 ) ) . P^2=P,\qquad\left(S-\frac{t+1}{2}\right)P=0,\qquad PD=D\quad\left(D\in\ker\left(S-\frac{t+1}{2}\right)\right). P 2 = P , ( S − 2 t + 1 ) P = 0 , P D = D ( D ∈ ker ( S − 2 t + 1 ) ) .
Thus im P = ker ( S − t + 1 2 ) \operatorname{im} P = \ker\left(S-\frac{t+1}{2}\right) im P = ker ( S − 2 t + 1 ) . Write P = ( P i j ) P=(P_{ij}) P = ( P ij ) as a matrix of endomorphisms of N N N . By Lemma 3.1 ,
P 11 + P 22 + P 33 = c − 3 ( t − 1 ) t ( t − 1 ) id N = t − 2 t id N . (24) P_{11}+P_{22}+P_{33}=\frac{c-3(t-1)}{t(t-1)}\operatorname{id}_{N}=\frac{t-2}{t}\operatorname{id}_{N}. \tag*{(24)} P 11 + P 22 + P 33 = t ( t − 1 ) c − 3 ( t − 1 ) id N = t t − 2 id N . (24)
If N ≠ 0 N\ne0 N = 0 and t ∉ { 0 , 2 } t\notin\{0,2\} t ∈ / { 0 , 2 } , this sum is nonzero. Hence P ≠ 0 P\ne0 P = 0 and Der δ ( s l 2 , N ) ≠ 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_{2},N)\ne0 Der δ ( sl 2 , N ) = 0 .
Suppose t = 2 t=2 t = 2 , that is, δ = 2 3 \delta=\frac{2}{3} δ = 3 2 , so c = 3 c=3 c = 3 . Here P = S 2 − 1 2 S − 1 2 P=S^{2}-\frac{1}{2}S-\frac{1}{2} P = S 2 − 2 1 S − 2 1 , and the derivation space is zero if and only if P = 0 P=0 P = 0 . If P = 0 P=0 P = 0 , its ( 1 , 3 ) (1,3) ( 1 , 3 ) - and ( 3 , 1 ) (3,1) ( 3 , 1 ) -entries, − e 2 2 -\frac{e^{2}}{2} − 2 e 2 and − f 2 2 -\frac{f^{2}}{2} − 2 f 2 , give e 2 = f 2 = 0 e^{2}=f^{2}=0 e 2 = f 2 = 0 on N N N . The identities
0 = [ f , e 2 ] = − h e − e h , 0 = [ e , f 2 ] = h f + f h 0=[f,e^{2}]=-he-eh,\qquad0=[e,f^{2}]=hf+fh 0 = [ f , e 2 ] = − h e − e h , 0 = [ e , f 2 ] = h f + f h
together with (12) give h e = e he=e h e = e and h f = − f hf=-f h f = − f . Multiplying e f − f e = h ef-fe=h e f − f e = h on the right by e e e and f f f then gives
e f e = e , f e f = f . efe=e,\qquad fef=f. e f e = e , f e f = f .
Since e 2 = f 2 = 0 e^{2}=f^{2}=0 e 2 = f 2 = 0 , it follows that
h 2 = ( e f − f e ) 2 = e f e f + f e f e = e f + f e . h^{2}=(ef-fe)^{2}=efef+fefe=ef+fe. h 2 = ( e f − f e ) 2 = e f e f + f e f e = e f + f e .
Thus Ω M = 3 h 2 \Omega_{M}=3h^{2} Ω M = 3 h 2 , and Ω M = 3 \Omega_{M}=3 Ω M = 3 gives h 2 = 1 h^{2}=1 h 2 = 1 . Consequently,
e f = 1 + h 2 , f e = 1 − h 2 , N = N − 1 ⊕ N 1 , ef=\frac{1+h}{2},\qquad fe=\frac{1-h}{2},\qquad N=N_{-1}\oplus N_{1}, e f = 2 1 + h , f e = 2 1 − h , N = N − 1 ⊕ N 1 ,
where N ± 1 = { v ∈ N : h v = ± v } N_{\pm1}=\{v\in N:hv=\pm v\} N ± 1 = { v ∈ N : h v = ± v } . The maps e : N − 1 → N 1 e:N_{-1}\to N_{1} e : N − 1 → N 1 and f : N 1 → N − 1 f:N_{1}\to N_{-1} f : N 1 → N − 1 are inverse. Choose a basis { v i : i ∈ I } \{v_i:i\in I\} { v i : i ∈ I } of N − 1 N_{-1} N − 1 . Since
f v i = f 2 e v i = 0 , f e v i = v i , e 2 v i = 0 , fv_i=f^{2}ev_i=0,\qquad fev_i=v_i,\qquad e^{2}v_i=0, f v i = f 2 e v i = 0 , f e v i = v i , e 2 v i = 0 ,
each C v i ⊕ C e v i \mathbb{C}v_i\oplus\mathbb{C}ev_i C v i ⊕ C e v i is a submodule isomorphic to V ( 1 ) V(1) V ( 1 ) , and
N = ⨁ i ∈ I ( C v i ⊕ C e v i ) ≅ ⨁ i ∈ I V ( 1 ) . N=\bigoplus_{i\in I}(\mathbb{C}v_i\oplus\mathbb{C}ev_i)\cong\bigoplus_{i\in I}V(1). N = i ∈ I ⨁ ( C v i ⊕ C e v i ) ≅ i ∈ I ⨁ V ( 1 ) .
Conversely, V ( 2 ) ⊗ V ( 1 ) ≅ V ( 3 ) ⊕ V ( 1 ) V(2)\otimes V(1)\cong V(3)\oplus V(1) V ( 2 ) ⊗ V ( 1 ) ≅ V ( 3 ) ⊕ V ( 1 ) has no Casimir eigenvalue − 1 = c ′ -1=c' − 1 = c ′ . Tensor products and Casimir eigenspaces commute with direct sums, so Der 2 3 ( s l 2 , N ) = 0 \operatorname{Der}_{\frac{2}{3}}(\mathfrak{sl}_{2},N)=0 Der 3 2 ( sl 2 , N ) = 0 for every direct sum of copies of V ( 1 ) V(1) V ( 1 ) .
It remains to consider t = 0 t=0 t = 0 , that is, δ = 2 \delta=2 δ = 2 . Here c = − 1 c=-1 c = − 1 , and (22) reads ( S − 1 ) ( S − 1 2 ) 2 = 0 (S-1)(S-\frac{1}{2})^{2}=0 ( S − 1 ) ( S − 2 1 ) 2 = 0 . Since 1 2 \frac{1}{2} 2 1 is a repeated root, we replace the projection by an operator whose image lies in ker ( S − 1 2 ) \ker(S-\frac{1}{2}) ker ( S − 2 1 ) . Set
J = − 2 S 2 + 3 S − 1 = − 2 ( S − 1 ) ( S − 1 2 ) . (25) J=-2S^{2}+3S-1=-2(S-1)\left(S-\frac{1}{2}\right). \tag*{(25)} J = − 2 S 2 + 3 S − 1 = − 2 ( S − 1 ) ( S − 2 1 ) . (25)
The factorization gives
J 2 = 4 ( S − 1 ) 2 ( S − 1 2 ) 2 = 0 , ( S − 1 2 ) J = − 2 ( S − 1 ) ( S − 1 2 ) 2 = 0. J^{2}=4(S-1)^{2}\left(S-\frac{1}{2}\right)^{2}=0,\qquad\left(S-\frac{1}{2}\right)J=-2(S-1)\left(S-\frac{1}{2}\right)^{2}=0. J 2 = 4 ( S − 1 ) 2 ( S − 2 1 ) 2 = 0 , ( S − 2 1 ) J = − 2 ( S − 1 ) ( S − 2 1 ) 2 = 0.
By Lemma 3.1 , the sum of the diagonal entries of J J J is
( − c − 3 ) id N = − 2 id N . (-c-3)\operatorname{id}_{N}=-2\operatorname{id}_{N}. ( − c − 3 ) id N = − 2 id N .
If N ≠ 0 N\ne0 N = 0 , then J ≠ 0 J\ne0 J = 0 and
0 ≠ im J ⊆ ker ( S − 1 2 ) = Der 2 ( s l 2 , N ) . 0\ne\operatorname{im}J\subseteq\ker\left(S-\frac{1}{2}\right)=\operatorname{Der}_{2}(\mathfrak{sl}_{2},N). 0 = im J ⊆ ker ( S − 2 1 ) = Der 2 ( sl 2 , N ) .
Since U ( s l 2 ) U(\mathfrak{sl}_{2}) U ( sl 2 ) has countable dimension, Dixmier’s lemma gives End s l 2 M = C \operatorname{End}_{\mathfrak{sl}_{2}}M=\mathbb{C} End sl 2 M = C for every simple s l 2 \mathfrak{sl}_{2} sl 2 -module M M M ; in particular, Ω \Omega Ω acts by a scalar [20 ] .
Corollary 3.4. Let M M M be a simple s l 2 \mathfrak{sl}_{2} sl 2 -module with Casimir scalar κ \kappa κ . The parameters δ \delta δ for which Der δ ( s l 2 , M ) ≠ 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_{2},M)\ne0 Der δ ( sl 2 , M ) = 0 are exactly the roots in C ∖ { 0 , 1 } \mathbb{C}\setminus\{0,1\} C ∖ { 0 , 1 } of
κ δ 2 + 4 δ − 4 = 0 , (26) \kappa\delta^{2}+4\delta-4=0, \tag*{(26)} κ δ 2 + 4 δ − 4 = 0 , (26)
except that δ = 2 3 \delta=\frac{2}{3} δ = 3 2 is omitted when M ≅ V ( 1 ) M\cong V(1) M ≅ V ( 1 ) . There are no such parameters for κ = 0 \kappa=0 κ = 0 , and only δ = 2 \delta=2 δ = 2 for κ = − 1 \kappa=-1 κ = − 1 . For κ ≠ 0 \kappa\ne0 κ = 0 , the two roots, counted with multiplicity, are interchanged by
δ ⟼ δ δ − 1 . (27) \delta\longmapsto\frac{\delta}{\delta-1}. \tag*{(27)} δ ⟼ δ − 1 δ . (27)
Proof. The space M ( c δ ) M^{(c_\delta)} M ( c δ ) is M M M when κ = c δ = 4 ( 1 − δ ) δ 2 \kappa=c_\delta=\frac{4(1-\delta)}{\delta^2} κ = c δ = δ 2 4 ( 1 − δ ) , and is zero otherwise. Theorem 3.3 gives (26) and the stated exception. For κ = 0 \kappa=0 κ = 0 , its only root is the excluded value 1 1 1 ; for κ = − 1 \kappa=-1 κ = − 1 , it has the double root 2 2 2 . For κ ≠ 0 \kappa\ne0 κ = 0 , Vieta’s formulas show that the reciprocals of the roots sum to 1 1 1 , giving (27) . □
Corollary 3.5. Let δ ∉ { 0 , 1 , 2 } \delta\notin\{0,1,2\} δ ∈ / { 0 , 1 , 2 } , and let M M M be an s l 2 \mathfrak{sl}_2 sl 2 -module on which Ω \Omega Ω acts by 4 ( 1 − δ ) δ 2 \frac{4(1-\delta)}{\delta^2} δ 2 4 ( 1 − δ ) . There is a natural direct sum decomposition of s l 2 \mathfrak{sl}_2 sl 2 -modules
Hom C ( s l 2 , M ) = Der 1 ( s l 2 , M ) ⊕ Der δ ( s l 2 , M ) ⊕ Der δ δ − 1 ( s l 2 , M ) . (28) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M)=\operatorname{Der}_1(\mathfrak{sl}_2,M)\oplus\operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)\oplus\operatorname{Der}_{\frac{\delta}{\delta-1}}(\mathfrak{sl}_2,M). \tag*{(28)} Hom C ( sl 2 , M ) = Der 1 ( sl 2 , M ) ⊕ Der δ ( sl 2 , M ) ⊕ Der δ − 1 δ ( sl 2 , M ) . (28)
Proof. On Hom C ( s l 2 , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M) Hom C ( sl 2 , M ) , the first identity in (17) becomes
( S − 1 ) ( S − δ − 1 ) ( S − ( 1 − δ − 1 ) ) = 0. (S-1)(S-\delta^{-1})(S-(1-\delta^{-1}))=0. ( S − 1 ) ( S − δ − 1 ) ( S − ( 1 − δ − 1 )) = 0.
The three roots are distinct under the stated assumptions. Since ker ( S − ε − 1 ) = Der ε ( s l 2 , M ) \ker(S-\varepsilon^{-1})=\operatorname{Der}_{\varepsilon}(\mathfrak{sl}_2,M) ker ( S − ε − 1 ) = Der ε ( sl 2 , M ) for every ε ≠ 0 \varepsilon\ne0 ε = 0 , the eigenspace decomposition gives (28) . The operator S S S is s l 2 \mathfrak{sl}_2 sl 2 -equivariant and commutes with maps induced by module homomorphisms, so this decomposition is natural. □
Translation functors
We use translation functors to determine the structure of the derivation module for simple coefficient modules.
Definition 3.6. For κ ∈ C \kappa\in\mathbb{C} κ ∈ C , put A κ = U ( s l 2 ) / ( Ω − κ ) A_\kappa=U(\mathfrak{sl}_2)/(\Omega-\kappa) A κ = U ( sl 2 ) / ( Ω − κ ) . For an s l 2 \mathfrak{sl}_2 sl 2 -module X X X , write
X [ κ ] = ⋃ r ≥ 1 { v ∈ X : ( Ω − κ ) r v = 0 } . X^{[\kappa]}=\bigcup_{r\ge1}\{v\in X:(\Omega-\kappa)^r v=0\}. X [ κ ] = r ≥ 1 ⋃ { v ∈ X : ( Ω − κ ) r v = 0 } . The full subcategory of s l 2 \mathfrak{sl}_2 sl 2 -modules satisfying X = X [ κ ] X=X^{[\kappa]} X = X [ κ ] is called the generalized central-character block at κ \kappa κ . The A κ A_\kappa A κ -modules are those annihilated by Ω − κ \Omega-\kappa Ω − κ . With our normalization, a shifted Harish-Chandra parameter for this block is a choice of square root of κ + 1 \kappa+1 κ + 1 . The block is singular if this parameter is zero, and regular otherwise.
The translation functors used below are obtained by tensoring with V ( 2 ) V(2) V ( 2 ) and taking the appropriate generalized Casimir summand. The source and target parameters can be chosen as t t t and t − 2 t-2 t − 2 , whose difference is an extremal weight of V ( 2 ) V(2) V ( 2 ) . The Weyl group changes the sign of a shifted parameter; on δ \delta δ , this is the involution in (27) .
For an A c A_c A c -module N N N , equation (20) identifies Der δ ( s l 2 , N ) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,N) Der δ ( sl 2 , N ) with the c ′ c' c ′ -eigenspace in V ( 2 ) ⊗ N V(2)\otimes N V ( 2 ) ⊗ N . Thus the derivation functor is a translation functor whenever this eigenspace equals the corresponding generalized eigenspace. We use the identification (21) throughout this subsection.
Theorem 3.7. Let δ ∈ C ∖ { 0 , 1 } \delta\in\mathbb{C}\setminus\{0,1\} δ ∈ C ∖ { 0 , 1 } , and put
t = 2 δ − 1 , c = t 2 − 1 , c ′ = ( t − 2 ) 2 − 1. t=\frac{2}{\delta}-1,\qquad c=t^2-1,\qquad c'=(t-2)^2-1. t = δ 2 − 1 , c = t 2 − 1 , c ′ = ( t − 2 ) 2 − 1. If δ ∉ { 2 , 2 3 } \delta\notin\{2,\frac{2}{3}\} δ ∈ / { 2 , 3 2 } , then Der δ ( s l 2 , − ) : A c -Mod → A c ′ -Mod \operatorname{Der}_{\delta}(\mathfrak{sl}_2,-):A_c\text{-Mod}\to A_{c'}\text{-Mod} Der δ ( sl 2 , − ) : A c -Mod → A c ′ -Mod is an equivalence of categories given by a translation functor. In particular, Der δ ( s l 2 , M ) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M) Der δ ( sl 2 , M ) is simple for every simple A c A_c A c -module M M M .
If δ = 2 3 \delta=\frac{2}{3} δ = 3 2 , then Der 2 3 ( s l 2 , − ) \operatorname{Der}_{\frac{2}{3}}(\mathfrak{sl}_2,-) Der 3 2 ( sl 2 , − ) is translation from c = 3 c=3 c = 3 onto c ′ = − 1 c'=-1 c ′ = − 1 . It induces a bijection between the isomorphism classes of simple A 3 A_3 A 3 -modules other than V ( 1 ) V(1) V ( 1 ) and those of simple A − 1 A_{-1} A − 1 -modules.
If δ = 2 \delta=2 δ = 2 and M M M is a simple A − 1 A_{-1} A − 1 -module, put
T ( M ) = { v ∈ V ( 2 ) ⊗ M : ( Ω − 3 ) 2 v = 0 } . (29) \mathcal{T}(M)=\{v\in V(2)\otimes M:(\Omega-3)^2v=0\}. \tag*{(29)} T ( M ) = { v ∈ V ( 2 ) ⊗ M : ( Ω − 3 ) 2 v = 0 } . (29) Under the identification (21) , one has
Der 2 ( s l 2 , M ) = rad T ( M ) . (30) \operatorname{Der}_2(\mathfrak{sl}_2,M)=\operatorname{rad}\mathcal{T}(M). \tag*{(30)} Der 2 ( sl 2 , M ) = rad T ( M ) . (30) Moreover, soc T ( M ) \operatorname{soc}\mathcal{T}(M) soc T ( M ) is the simple essential socle of Der 2 ( s l 2 , M ) \operatorname{Der}_2(\mathfrak{sl}_2,M) Der 2 ( sl 2 , M ) , and
Der 2 ( s l 2 , M ) / soc T ( M ) ≅ V ( 1 ) ⊕ m for some m ∈ Z ≥ 0 . \operatorname{Der}_2(\mathfrak{sl}_2,M)/\operatorname{soc}\mathcal{T}(M)\cong V(1)^{\oplus m}\quad\text{for some }m\in\mathbb{Z}_{\ge0}. Der 2 ( sl 2 , M ) / soc T ( M ) ≅ V ( 1 ) ⊕ m for some m ∈ Z ≥ 0 .
Proof. For (i), assume t ∉ { 0 , 1 , 2 } t \notin\{0,1,2\} t ∈ / { 0 , 1 , 2 } and first work with finitely generated modules. On V ( 2 ) ⊗ N V(2) \otimes N V ( 2 ) ⊗ N , where N N N is an A c A_c A c -module, equations (22) and Ω = c + 8 − 8 S \Omega= c + 8 - 8S Ω = c + 8 − 8 S give
( Ω − c ) ( Ω − c ′ ) ( Ω − ( ( t + 2 ) 2 − 1 ) ) = 0. (\Omega-c)(\Omega-c')(\Omega-((t+2)^2-1))=0. ( Ω − c ) ( Ω − c ′ ) ( Ω − (( t + 2 ) 2 − 1 )) = 0.
Since
c ′ − c = 4 ( 1 − t ) ≠ 0 , c ′ − ( ( t + 2 ) 2 − 1 ) = − 8 t ≠ 0 , c'-c=4(1-t)\ne0,\qquad c'-((t+2)^2-1)=-8t\ne0, c ′ − c = 4 ( 1 − t ) = 0 , c ′ − (( t + 2 ) 2 − 1 ) = − 8 t = 0 ,
Ω − c ′ \Omega-c' Ω − c ′ is a simple factor of this annihilating polynomial, so
( V ( 2 ) ⊗ N ) [ c ′ ] = ( V ( 2 ) ⊗ N ) ( c ′ ) = Der δ ( s l 2 , N ) . (V(2)\otimes N)^{[c']}=(V(2)\otimes N)^{(c')}=\operatorname{Der}_{\delta}(\mathfrak{sl}_2,N). ( V ( 2 ) ⊗ N ) [ c ′ ] = ( V ( 2 ) ⊗ N ) ( c ′ ) = Der δ ( sl 2 , N ) .
This summand corresponds to the extremal shift t ↦ t − 2 t\mapsto t-2 t ↦ t − 2 .
If t ∉ Z t\notin\mathbb{Z} t ∈ / Z , both shifted parameters are dominant in the sense of the translation principle, since neither is a negative integer. If t ∈ Z t\in\mathbb{Z} t ∈ Z , then t ≥ 3 t\ge3 t ≥ 3 or t ≤ − 2 t\le-2 t ≤ − 2 ; in the latter case replace ( t , t − 2 ) (t,t-2) ( t , t − 2 ) by ( − t , 2 − t ) (-t,2-t) ( − t , 2 − t ) . In either case the parameters are dominant, have trivial stabilizers, and differ by an extremal weight 2 2 2 or − 2 -2 − 2 of V ( 2 ) V(2) V ( 2 ) . The translation principle [3 Theorem 4.1] , in the form recalled in [6 Lemma 12(i)] , gives inverse equivalences between the two generalized central-character blocks.
We next check that these equivalences restrict to modules on which the Casimir acts by a scalar. The equality of the eigenspace and generalized eigenspace shows that the forward functor takes A c A_c A c -modules to A c ′ A_{c'} A c ′ -modules. For an A c ′ A_{c'} A c ′ -module N N N , the reverse functor selects the generalized S S S -eigenspace for 3 − t 2 \frac{3-t}{2} 2 3 − t , since
c ′ + 8 − c 8 = 3 − t 2 , ( S − 1 ) ( S − t − 1 2 ) ( S − 3 − t 2 ) = 0. \frac{c'+8-c}{8}=\frac{3-t}{2},\qquad(S-1)\left(S-\frac{t-1}{2}\right)\left(S-\frac{3-t}{2}\right)=0. 8 c ′ + 8 − c = 2 3 − t , ( S − 1 ) ( S − 2 t − 1 ) ( S − 2 3 − t ) = 0.
This root differs from 1 1 1 and t − 1 2 \frac{t-1}{2} 2 t − 1 because t ≠ 1 , 2 t\ne1,2 t = 1 , 2 . Hence the generalized c c c -eigenspace selected by the reverse functor is also an eigenspace:
N ⟼ ( V ( 2 ) ⊗ N ) [ c ] = ( V ( 2 ) ⊗ N ) ( c ) . N\longmapsto(V(2)\otimes N)^{[c]}=(V(2)\otimes N)^{(c)}. N ⟼ ( V ( 2 ) ⊗ N ) [ c ] = ( V ( 2 ) ⊗ N ) ( c ) .
Tensoring with the finite-dimensional module V ( 2 ) V(2) V ( 2 ) and taking direct summands preserve finite generation. The unit and counit of the block equivalences therefore restrict to finitely generated A c A_c A c - and A c ′ A_{c'} A c ′ -modules.
For an arbitrary module, write N = ⋃ i N i N=\bigcup_i N_i N = ⋃ i N i as the directed union of its finitely generated submodules. Tensoring with V ( 2 ) V(2) V ( 2 ) commutes with this union, and each vector in a Casimir eigenspace already belongs to one of the tensor products. Thus
( V ( 2 ) ⊗ ⋃ i N i ) ( κ ) = ⋃ i ( V ( 2 ) ⊗ N i ) ( κ ) ( κ = c , c ′ ) . \left(V(2)\otimes\bigcup_iN_i\right)^{(\kappa)}
=\bigcup_i(V(2)\otimes N_i)^{(\kappa)}
\qquad(\kappa=c,c'). ( V ( 2 ) ⊗ i ⋃ N i ) ( κ ) = i ⋃ ( V ( 2 ) ⊗ N i ) ( κ ) ( κ = c , c ′ ) .
Both functors therefore commute with these unions. By naturality, the unit and counit extend to isomorphisms on the unions, proving (i) for arbitrary modules.
For (ii), t = 2 t=2 t = 2 and ( c , c ′ ) = ( 3 , − 1 ) (c,c')=(3,-1) ( c , c ′ ) = ( 3 , − 1 ) . The factor S − 3 2 S-\frac{3}{2} S − 2 3 occurs only once in (22) , so
Der 2 3 ( s l 2 , N ) = ( V ( 2 ) ⊗ N ) ( − 1 ) = ( V ( 2 ) ⊗ N ) [ − 1 ] ( N ∈ A 3 -Mod ) . \operatorname{Der}_{\frac{2}{3}}(\mathfrak{sl}_2,N)
=(V(2)\otimes N)^{(-1)}
=(V(2)\otimes N)^{[-1]}
\qquad(N\in A_3\text{-Mod}). Der 3 2 ( sl 2 , N ) = ( V ( 2 ) ⊗ N ) ( − 1 ) = ( V ( 2 ) ⊗ N ) [ − 1 ] ( N ∈ A 3 -Mod ) .
This is translation from the regular parameter 2 2 2 onto the singular parameter 0 0 0 . It sends a simple module to zero or a simple module and gives a bijection between the isomorphism classes of simple modules with nonzero image and those of simple modules in the singular block [6 Theorem 26] . For simple A 3 A_3 A 3 -modules, Corollary 3.4 gives
Der 2 3 ( s l 2 , M ) = 0 ⟺ M ≅ V ( 1 ) , \operatorname{Der}_{\frac{2}{3}}(\mathfrak{sl}_2,M)=0
\Longleftrightarrow
M\cong V(1), Der 3 2 ( sl 2 , M ) = 0 ⟺ M ≅ V ( 1 ) ,
which proves (ii).
For (iii), equations (17) and (25) give
Ω = 7 − 8 S , T ( M ) = ker ( ( S − 1 2 ) 2 ) , (31) \Omega=7-8S,\qquad\mathcal{T}(M)=\ker\left(\left(S-\frac{1}{2}\right)^2\right), \tag*{(31)} Ω = 7 − 8 S , T ( M ) = ker ( ( S − 2 1 ) 2 ) , (31)
Der 2 ( s l 2 , M ) = { D ∈ T ( M ) : S D = 1 2 D } . \operatorname{Der}_{2}(\mathfrak{sl}_2,M)
=\left\{D\in\mathcal{T}(M):SD=\frac{1}{2}D\right\}. Der 2 ( sl 2 , M ) = { D ∈ T ( M ) : S D = 2 1 D } .
The characteristic identity is ( S − 1 ) ( S − 1 2 ) 2 = 0 (S-1)(S-\frac{1}{2})^2=0 ( S − 1 ) ( S − 2 1 ) 2 = 0 . Its two factors are coprime, so
V ( 2 ) ⊗ M = ker ( S − 1 ) ⊕ T ( M ) . V(2)\otimes M=\ker(S-1)\oplus T(M). V ( 2 ) ⊗ M = ker ( S − 1 ) ⊕ T ( M ) .
Thus T ( M ) T(M) T ( M ) is the generalized Casimir summand at 3. The shifts ± 2 \pm2 ± 2 from the singular parameter 0 give the same regular target character, whereas the zero shift has Casimir scalar − 1 -1 − 1 . Hence T T T is translation out of the singular block. By [6 Theorem 33] , V ( 2 ) ⊗ M V(2)\otimes M V ( 2 ) ⊗ M , and hence its direct summand T ( M ) T(M) T ( M ) , has finite length. By [6 Theorem 26(i)–(iii)] , its socle L = soc T ( M ) L=\operatorname{soc}T(M) L = soc T ( M ) is simple and essential, and
T ( M ) / rad T ( M ) ≅ L , [ T ( M ) : L ] = 2. T(M)/\operatorname{rad}T(M)\cong L,\qquad[T(M):L]=2. T ( M ) / rad T ( M ) ≅ L , [ T ( M ) : L ] = 2.
where the brackets denote composition multiplicities. In particular, rad T ( M ) \operatorname{rad}T(M) rad T ( M ) is its unique maximal submodule. To identify the derivation space with the radical, we compute the endomorphism algebra of T ( M ) T(M) T ( M ) and determine J ∣ T ( M ) J|_{T(M)} J ∣ T ( M ) .
For Y Y Y in the generalized block at 3, put
Θ ( Y ) = ( V ( 2 ) ⊗ Y ) [ − 1 ] . \Theta(Y)=(V(2)\otimes Y)^{[-1]}. Θ ( Y ) = ( V ( 2 ) ⊗ Y ) [ − 1 ] .
This is translation back onto the singular block. By [6 Lemma 12(ii) and Theorem 26(i)] ,
Θ T ( M ) ≅ M ⊕ 2 , Θ L ≅ M . \Theta T(M)\cong M^{\oplus2},\qquad\Theta L\cong M. Θ T ( M ) ≅ M ⊕ 2 , Θ L ≅ M .
Biadjointness of the translation functors and End s l 2 M = C \operatorname{End}_{\mathfrak{sl}_2}M=\mathbb{C} End sl 2 M = C give
dim End s l 2 T ( M ) = dim Hom s l 2 ( M , Θ T ( M ) ) = dim Hom s l 2 ( M , M ⊕ 2 ) = 2. \dim\operatorname{End}_{\mathfrak{sl}_2}T(M)=\dim\operatorname{Hom}_{\mathfrak{sl}_2}(M,\Theta T(M))
=\dim\operatorname{Hom}_{\mathfrak{sl}_2}(M,M^{\oplus2})=2. dim End sl 2 T ( M ) = dim Hom sl 2 ( M , Θ T ( M )) = dim Hom sl 2 ( M , M ⊕ 2 ) = 2.
Let i : L ↪ T ( M ) i:L\hookrightarrow T(M) i : L ↪ T ( M ) be the socle inclusion and choose q : T ( M ) ↠ L q:T(M)\twoheadrightarrow L q : T ( M ) ↠ L with ker q = rad T ( M ) \ker q=\operatorname{rad}T(M) ker q = rad T ( M ) . If q i ≠ 0 qi\ne0 q i = 0 , it is an automorphism of the simple module L L L . Replacing i i i by i ( q i ) − 1 i(qi)^{-1} i ( q i ) − 1 would give q i = id L qi=\operatorname{id}_L q i = id L , hence T ( M ) = i ( L ) ⊕ ker q T(M)=i(L)\oplus\ker q T ( M ) = i ( L ) ⊕ ker q . Essentiality of i ( L ) i(L) i ( L ) would force ker q = 0 \ker q=0 ker q = 0 , contrary to [ T ( M ) : L ] = 2 [T(M):L]=2 [ T ( M ) : L ] = 2 . Thus q i = 0 qi=0 q i = 0 . Since q q q is surjective and i i i is injective, i q ≠ 0 iq\ne0 i q = 0 , and
( i q ) 2 = i ( q i ) q = 0 , End s l 2 T ( M ) = C id ⊕ C i q . (iq)^2=i(qi)q=0,\qquad\operatorname{End}_{\mathfrak{sl}_2}T(M)=\mathbb{C}\operatorname{id}\oplus\mathbb{C}iq. ( i q ) 2 = i ( q i ) q = 0 , End sl 2 T ( M ) = C id ⊕ C i q .
By the proof of Theorem 3.3 , J ≠ 0 J\ne0 J = 0 on V ( 2 ) ⊗ M V(2)\otimes M V ( 2 ) ⊗ M . By (25) , J J J vanishes on ker ( S − 1 ) \ker(S-1) ker ( S − 1 ) . On T ( M ) T(M) T ( M ) , the same factorization gives
J = − 2 [ ( S − 1 2 ) − 1 2 ] ( S − 1 2 ) = − 2 ( S − 1 2 ) 2 + ( S − 1 2 ) = S − 1 2 . \begin{aligned}
J=-2\left[\left(S-\frac{1}{2}\right)-\frac{1}{2}\right]\left(S-\frac{1}{2}\right) \\
&=-2\left(S-\frac{1}{2}\right)^2+\left(S-\frac{1}{2}\right)=S-\frac{1}{2}.
\end{aligned} J = − 2 [ ( S − 2 1 ) − 2 1 ] ( S − 2 1 ) = − 2 ( S − 2 1 ) 2 + ( S − 2 1 ) = S − 2 1 .
Thus J ∣ T ( M ) ≠ 0 J|_{T(M)}\ne0 J ∣ T ( M ) = 0 and J 2 = 0 J^2=0 J 2 = 0 . Write J ∣ T ( M ) = α id + β i q J|_{T(M)}=\alpha\operatorname{id}+\beta iq J ∣ T ( M ) = α id + β i q . Since q i = 0 qi=0 q i = 0 ,
q J = α q , 0 = q J 2 = α 2 q . qJ=\alpha q,\qquad0=qJ^2=\alpha^2q. q J = α q , 0 = q J 2 = α 2 q .
Surjectivity of q q q gives α = 0 \alpha=0 α = 0 , and nonvanishing of J ∣ T ( M ) J|_{T(M)} J ∣ T ( M ) gives β ≠ 0 \beta\ne0 β = 0 . Hence
im J = i ( L ) = soc T ( M ) , ker ( J ∣ T ( M ) ) = ker q = rad T ( M ) . \operatorname{im}J=i(L)=\operatorname{soc}T(M),\qquad\ker(J|_{T(M)})=\ker q=\operatorname{rad}T(M). im J = i ( L ) = soc T ( M ) , ker ( J ∣ T ( M ) ) = ker q = rad T ( M ) .
On T ( M ) T(M) T ( M ) , one has J = S − 1 2 = − 1 8 ( Ω − 3 ) J=S-\frac{1}{2}=-\frac{1}{8}(\Omega-3) J = S − 2 1 = − 8 1 ( Ω − 3 ) . Together with (31) , this proves (30) and gives the exact sequence
0 ⟶ Der 2 ( s l 2 , M ) ⟶ T ( M ) → Ω − 3 soc T ( M ) ⟶ 0. (32) 0\longrightarrow\operatorname{Der}_2(\mathfrak{sl}_2,M)\longrightarrow T(M)\xrightarrow{\Omega-3}\operatorname{soc}T(M)\longrightarrow0. \tag*{(32)} 0 ⟶ Der 2 ( sl 2 , M ) ⟶ T ( M ) Ω − 3 soc T ( M ) ⟶ 0. (32)
The first map is the inclusion under the identification (21) . Since J 2 = 0 J^2=0 J 2 = 0 , the simple socle of T ( M ) T(M) T ( M ) lies in Der 2 ( s l 2 , M ) \operatorname{Der}_2(\mathfrak{sl}_2,M) Der 2 ( sl 2 , M ) . Being essential in T ( M ) T(M) T ( M ) , it is also essential in Der 2 ( s l 2 , M ) \operatorname{Der}_2(\mathfrak{sl}_2,M) Der 2 ( sl 2 , M ) .
Finally, put Q = Der 2 ( s l 2 , M ) / L Q=\operatorname{Der}_2(\mathfrak{sl}_2,M)/L Q = Der 2 ( sl 2 , M ) / L . The functor Θ \Theta Θ is exact, so applying it to (32) gives
0 ⟶ Θ Der 2 ( s l 2 , M ) ⟶ M ⊕ 2 ⟶ M ⟶ 0. 0\longrightarrow\Theta\operatorname{Der}_2(\mathfrak{sl}_2,M)\longrightarrow M^{\oplus2}\longrightarrow M\longrightarrow0. 0 ⟶ Θ Der 2 ( sl 2 , M ) ⟶ M ⊕ 2 ⟶ M ⟶ 0.
Hence Θ Der 2 ( s l 2 , M ) ≅ M \Theta\operatorname{Der}_2(\mathfrak{sl}_2,M)\cong M Θ Der 2 ( sl 2 , M ) ≅ M . The inclusion L ↪ Der 2 ( s l 2 , M ) L\hookrightarrow\operatorname{Der}_2(\mathfrak{sl}_2,M) L ↪ Der 2 ( sl 2 , M ) therefore induces an isomorphism Θ L → ∼ Θ Der 2 ( s l 2 , M ) \Theta L\xrightarrow{\sim}\Theta\operatorname{Der}_2(\mathfrak{sl}_2,M) Θ L ∼ Θ Der 2 ( sl 2 , M ) . Exactness now gives Θ Q = 0 \Theta Q=0 Θ Q = 0 . Since Ω \Omega Ω acts by 3 on Q Q Q , part (ii) identifies Θ Q \Theta Q Θ Q with Der 3 ( s l 2 , Q ) \operatorname{Der}_3(\mathfrak{sl}_2,Q) Der 3 ( sl 2 , Q ) . Theorem 3.3 (ii) implies that Q Q Q is a direct sum of copies of V ( 1 ) V(1) V ( 1 ) . As Q Q Q is a subquotient of the finite-length module T ( M ) T(M) T ( M ) , only finitely many copies occur.
Derivations with values in simple modules
We give explicit formulas for δ \delta δ -derivations with values in simple s l 2 \mathfrak{sl}_2 sl 2 -modules. Recall that t = 2 / δ − 1 t=2/\delta-1 t = 2/ δ − 1 . By Corollary 3.4, we need only consider modules with Casimir scalar t 2 − 1 t^2-1 t 2 − 1 .
Weight modules
The Verma module M ( λ ) M(\lambda) M ( λ ) has a basis ( v n ) n ≥ 0 (v_n)_{n\geq0} ( v n ) n ≥ 0 on which
h v n = ( λ − 2 n ) v n , f v n = v n + 1 , e v n = n ( λ − n + 1 ) v n − 1 . (33) hv_n=(\lambda-2n)v_n,\qquad fv_n=v_{n+1},\qquad ev_n=n(\lambda-n+1)v_{n-1}. \tag*{(33)} h v n = ( λ − 2 n ) v n , f v n = v n + 1 , e v n = n ( λ − n + 1 ) v n − 1 . (33)
It is simple exactly when λ ∉ Z ≥ 0 \lambda\notin\mathbb{Z}_{\geq0} λ ∈ / Z ≥ 0 . Let θ \theta θ be the involution
θ ( e ) = f , θ ( f ) = e , θ ( h ) = − h , \theta(e)=f,\qquad\theta(f)=e,\qquad\theta(h)=-h, θ ( e ) = f , θ ( f ) = e , θ ( h ) = − h ,
and put M ′ ( μ ) = M ( − μ ) θ M'(\mu)=M(-\mu)^\theta M ′ ( μ ) = M ( − μ ) θ , where x ⋅ θ m = θ ( x ) m x\cdot_\theta m=\theta(x)m x ⋅ θ m = θ ( x ) m . The finite-dimensional module V ( n ) V(n) V ( n ) has basis v 0 , … , v n v_0,\ldots,v_n v 0 , … , v n and action (33) with λ = n \lambda=n λ = n . A basis vector outside the specified index range is understood to be zero.
For ξ ∈ C / 2 Z \xi\in\mathbb{C}/2\mathbb{Z} ξ ∈ C /2 Z and τ ∈ C \tau\in\mathbb{C} τ ∈ C , the dense module V ( ξ , τ ) V(\xi,\tau) V ( ξ , τ ) has basis ( v ν ) ν ∈ ξ (v_\nu)_{\nu\in\xi} ( v ν ) ν ∈ ξ with action
h v ν = ν v ν , f v ν = v ν − 2 , e v ν = τ − ( ν + 1 ) 2 4 v ν + 2 ( ν ∈ ξ ) . (34) hv_\nu=\nu v_\nu,\qquad fv_\nu=v_{\nu-2},\qquad ev_\nu=\frac{\tau-(\nu+1)^2}{4}v_{\nu+2}\quad(\nu\in\xi). \tag*{(34)} h v ν = ν v ν , f v ν = v ν − 2 , e v ν = 4 τ − ( ν + 1 ) 2 v ν + 2 ( ν ∈ ξ ) . (34)
Its Casimir scalar is τ − 1 \tau-1 τ − 1 , and it is simple if and only if − 1 ± τ ∉ ξ -1\pm\sqrt{\tau}\notin\xi − 1 ± τ ∈ / ξ . For these simplicity criteria, see [20 ] .
For a weight module M M M , write
Hom C ( s l 2 , M ) μ = { D : D ( e ) ∈ M μ + 2 , D ( h ) ∈ M μ , D ( f ) ∈ M μ − 2 } . \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M)_\mu=\{D:D(e)\in M_{\mu+2},\ D(h)\in M_\mu,\ D(f)\in M_{\mu-2}\}. Hom C ( sl 2 , M ) μ = { D : D ( e ) ∈ M μ + 2 , D ( h ) ∈ M μ , D ( f ) ∈ M μ − 2 } .
Since s l 2 \mathfrak{sl}_2 sl 2 is finite-dimensional, each map has only finitely many nonzero homogeneous components. Equation (16) holds separately in each degree, so
Der δ ( s l 2 , M ) = ⨁ μ ∈ C ( Der δ ( s l 2 , M ) ∩ Hom C ( s l 2 , M ) μ ) . \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)=\bigoplus_{\mu\in\mathbb{C}}\left(\operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)\cap\operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M)_\mu\right). Der δ ( sl 2 , M ) = μ ∈ C ⨁ ( Der δ ( sl 2 , M ) ∩ Hom C ( sl 2 , M ) μ ) .
To obtain bases from highest weight derivations, we record the powers of the lowering operator. Its action is
f ⋅ ( D ( e ) , D ( h ) , D ( f ) ) = ( f D ( e ) + D ( h ) , f D ( h ) − 2 D ( f ) , f D ( f ) ) . (35) f\cdot(D(e),D(h),D(f))=(fD(e)+D(h),\ fD(h)-2D(f),\ fD(f)). \tag*{(35)} f ⋅ ( D ( e ) , D ( h ) , D ( f )) = ( f D ( e ) + D ( h ) , f D ( h ) − 2 D ( f ) , f D ( f )) . (35)
The off-diagonal part of this action has scalar entries, its cube is zero, and it commutes with the componentwise action of f f f . The binomial formula therefore gives, for k ∈ Z ≥ 0 k\in\mathbb{Z}_{\geq0} k ∈ Z ≥ 0 ,
( f k ⋅ D ) ( e ) = f k D ( e ) + k f k − 1 D ( h ) − k ( k − 1 ) f k − 2 D ( f ) , ( f k ⋅ D ) ( h ) = f k D ( h ) − 2 k f k − 1 D ( f ) , ( f k ⋅ D ) ( f ) = f k D ( f ) . (36) \begin{aligned}
(f^k\cdot D)(e)&=f^kD(e)+kf^{k-1}D(h)-k(k-1)f^{k-2}D(f),\\
(f^k\cdot D)(h)&=f^kD(h)-2kf^{k-1}D(f),\\
(f^k\cdot D)(f)&=f^kD(f). \tag*{(36)}
\end{aligned} ( f k ⋅ D ) ( e ) ( f k ⋅ D ) ( h ) ( f k ⋅ D ) ( f ) = f k D ( e ) + k f k − 1 D ( h ) − k ( k − 1 ) f k − 2 D ( f ) , = f k D ( h ) − 2 k f k − 1 D ( f ) , = f k D ( f ) . (36)
Here the powers on the right act on M M M , and terms with negative powers are omitted.
For finite-dimensional simple modules, Theorem 3.3 and the Clebsch–Gordan decomposition recover the result of Zohrabi and Zusmanovich below. The basis formulas follow by applying (36) to a highest weight vector.
Proposition 4.1 ([22 Lemma 9(ii),(iii)] ). The nonzero derivation spaces with values in V ( n ) V(n) V ( n ) have the following bases.
(i) If n ≥ 1 n\geq1 n ≥ 1 and δ = − 2 n \delta=-\frac{2}{n} δ = − n 2 , a basis is D 0 , … , D n + 2 D_0,\ldots,D_{n+2} D 0 , … , D n + 2 , where
D k ( e ) = − k ( k − 1 ) v k − 2 , D k ( h ) = − 2 k v k − 1 , D k ( f ) = v k . (37) D_k(e)=-k(k-1)v_{k-2},\qquad D_k(h)=-2kv_{k-1},\qquad D_k(f)=v_k. \tag*{(37)} D k ( e ) = − k ( k − 1 ) v k − 2 , D k ( h ) = − 2 k v k − 1 , D k ( f ) = v k . (37) (ii) If n ≥ 2 n\geq2 n ≥ 2 and δ = 2 n + 2 \delta=\frac{2}{n+2} δ = n + 2 2 , a basis is D 0 , … , D n − 2 D_0,\ldots,D_{n-2} D 0 , … , D n − 2 , where
D k ( e ) = − ( n − k ) ( n − k − 1 ) v k , D_k(e)=-(n-k)(n-k-1)v_k, D k ( e ) = − ( n − k ) ( n − k − 1 ) v k , D k ( h ) = 2 ( n − k − 1 ) v k + 1 , D k ( f ) = v k + 2 . (38) D_k(h)=2(n-k-1)v_{k+1},\qquad D_k(f)=v_{k+2}. \tag*{(38)} D k ( h ) = 2 ( n − k − 1 ) v k + 1 , D k ( f ) = v k + 2 . (38) All other spaces vanish, including those with values in V ( 0 ) V(0) V ( 0 ) . Under the natural action,
Der − 2 n ( s l 2 , V ( n ) ) ≅ V ( n + 2 ) ( n ≥ 1 ) , \operatorname{Der}_{-\frac{2}{n}}(\mathfrak{sl}_2,V(n))\cong V(n+2)\qquad(n\geq1), Der − n 2 ( sl 2 , V ( n )) ≅ V ( n + 2 ) ( n ≥ 1 ) , Der 2 n + 2 ( s l 2 , V ( n ) ) ≅ V ( n − 2 ) ( n ≥ 2 ) . (39) \operatorname{Der}_{\frac{2}{n+2}}(\mathfrak{sl}_2,V(n))\cong V(n-2)\qquad(n\geq2). \tag*{(39)} Der n + 2 2 ( sl 2 , V ( n )) ≅ V ( n − 2 ) ( n ≥ 2 ) . (39)
Proof. The Casimir condition gives t 2 − 1 = n ( n + 2 ) t^2-1=n(n+2) t 2 − 1 = n ( n + 2 ) , hence t = ± ( n + 1 ) t=\pm(n+1) t = ± ( n + 1 ) . For n ≥ 2 n\ge2 n ≥ 2 ,
V ( 2 ) ⊗ V ( n ) ≅ V ( n + 2 ) ⊕ V ( n ) ⊕ V ( n − 2 ) , V(2)\otimes V(n)\cong V(n+2)\oplus V(n)\oplus V(n-2), V ( 2 ) ⊗ V ( n ) ≅ V ( n + 2 ) ⊕ V ( n ) ⊕ V ( n − 2 ) ,
with respective S S S -eigenvalues − n 2 -\frac{n}{2} − 2 n , 1 1 1 , n + 2 2 \frac{n+2}{2} 2 n + 2 by (17) . For n = 1 n=1 n = 1 , the last summand is absent. Taking the δ − 1 \delta^{-1} δ − 1 -eigenspace gives (39) . For V ( 0 ) V(0) V ( 0 ) , the derivation space is zero because s l 2 \mathfrak{sl}_2 sl 2 is perfect.
In the first family, D 0 ( e ) = D 0 ( h ) = 0 D_0(e)=D_0(h)=0 D 0 ( e ) = D 0 ( h ) = 0 and D 0 ( f ) = v 0 D_0(f)=v_0 D 0 ( f ) = v 0 . Since t = − n − 1 t=-n-1 t = − n − 1 , equations (16) hold by ( h + t + 1 ) v 0 = 0 (h+t+1)v_0=0 ( h + t + 1 ) v 0 = 0 and e v 0 = 0 ev_0=0 e v 0 = 0 . In the second family, t = n + 1 t=n+1 t = n + 1 , and (38) at k = 0 k=0 k = 0 gives
( h − t − 1 ) D 0 ( e ) = 2 n ( n − 1 ) v 0 = e D 0 ( h ) , (h-t-1)D_0(e)=2n(n-1)v_0=eD_0(h), ( h − t − 1 ) D 0 ( e ) = 2 n ( n − 1 ) v 0 = e D 0 ( h ) ,
( h + t + 1 ) D 0 ( f ) = 2 ( n − 1 ) v 2 = f D 0 ( h ) , (h+t+1)D_0(f)=2(n-1)v_2=fD_0(h), ( h + t + 1 ) D 0 ( f ) = 2 ( n − 1 ) v 2 = f D 0 ( h ) ,
e D 0 ( f ) − f D 0 ( e ) = ( n + 2 ) ( n − 1 ) v 1 = t + 1 2 D 0 ( h ) . eD_0(f)-fD_0(e)=(n+2)(n-1)v_1=\frac{t+1}{2}D_0(h). e D 0 ( f ) − f D 0 ( e ) = ( n + 2 ) ( n − 1 ) v 1 = 2 t + 1 D 0 ( h ) .
Thus D 0 D_0 D 0 is a derivation in both families. Formula (36) gives D k = f k ⋅ D 0 D_k=f^k\cdot D_0 D k = f k ⋅ D 0 . The first coordinate in (38) follows from
− n ( n − 1 ) + 2 k ( n − 1 ) − k ( k − 1 ) = − ( n − k ) ( n − k − 1 ) . -n(n-1)+2k(n-1)-k(k-1)=-(n-k)(n-k-1). − n ( n − 1 ) + 2 k ( n − 1 ) − k ( k − 1 ) = − ( n − k ) ( n − k − 1 ) .
The weights are n + 2 − 2 k n+2-2k n + 2 − 2 k and n − 2 − 2 k n-2-2k n − 2 − 2 k , respectively. In the first family, nonvanishing follows from
D k ( f ) = v k ( 0 ≤ k ≤ n ) , D n + 1 ( h ) = − 2 ( n + 1 ) v n , D n + 2 ( e ) = − ( n + 2 ) ( n + 1 ) v n . D_k(f)=v_k\ (0\le k\le n),\qquad D_{n+1}(h)=-2(n+1)v_n,\qquad D_{n+2}(e)=-(n+2)(n+1)v_n. D k ( f ) = v k ( 0 ≤ k ≤ n ) , D n + 1 ( h ) = − 2 ( n + 1 ) v n , D n + 2 ( e ) = − ( n + 2 ) ( n + 1 ) v n .
In the second family, D k ( f ) = v k + 2 ≠ 0 D_k(f)=v_{k+2}\ne0 D k ( f ) = v k + 2 = 0 for every index in the stated range. The displayed maps are thus nonzero and have distinct weights. By (39) , they form bases of the two derivation spaces. □
Corollary 4.2 ([22 ] , Main Theorem). Let g \mathfrak{g} g be a finite-dimensional complex simple Lie algebra, let M = ⨁ i M i M=\bigoplus_i M_i M = ⨁ i M i be a finite-dimensional g \mathfrak{g} g -module with simple summands, and let δ ∉ { 0 , 1 } \delta\notin\{0,1\} δ ∈ / { 0 , 1 } . Then
Der δ ( g , M ) = ⨁ i Der δ ( g , M i ) , Der 1 2 ( g , M ) = Hom g ( g , M ) . (40) \operatorname{Der}_{\delta}(\mathfrak{g},M)=\bigoplus_i\operatorname{Der}_{\delta}(\mathfrak{g},M_i),\qquad\operatorname{Der}_{\frac{1}{2}}(\mathfrak{g},M)=\operatorname{Hom}_{\mathfrak{g}}(\mathfrak{g},M). \tag*{(40)} Der δ ( g , M ) = i ⨁ Der δ ( g , M i ) , Der 2 1 ( g , M ) = Hom g ( g , M ) . (40) If δ ≠ 1 2 \delta\ne\frac{1}{2} δ = 2 1 , the space is nonzero precisely when g ≅ s l 2 \mathfrak{g}\cong\mathfrak{sl}_2 g ≅ sl 2 and M M M contains a simple summand V ( n ) V(n) V ( n ) with either δ = − 2 n \delta=-\frac{2}{n} δ = − n 2 , n ≥ 1 n\ge1 n ≥ 1 , or δ = 2 n + 2 \delta=\frac{2}{n+2} δ = n + 2 2 , n ≥ 3 n\ge3 n ≥ 3 . Every derivation is then a sum of maps into these summands, with bases given in Proposition 4.1.
Proof. Projection onto the simple summands gives the first equality. By Theorem 2.7 and Proposition 4.1, only adjoint summands contribute at δ = 1 2 \delta=\frac{1}{2} δ = 2 1 , and on each such summand the derivations are the module homomorphisms from g \mathfrak{g} g . The same results give the assertion for δ ≠ 1 2 \delta\ne\frac{1}{2} δ = 2 1 . □
For infinite-dimensional simple weight modules and δ ≠ 2 \delta\ne2 δ = 2 , Theorem 3.7 shows that every nonzero derivation module is simple.
Proposition 4.3. Suppose δ ≠ 2 \delta\ne2 δ = 2 .
(i) For derivations with values in infinite-dimensional simple highest weight modules, the nonzero spaces are
Der δ ( s l 2 , M ( t − 1 ) ) ≅ M ( t − 3 ) , t − 1 ∉ Z ≥ 0 , \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M(t-1))\cong M(t-3),\qquad t-1\notin\mathbb{Z}_{\ge0}, Der δ ( sl 2 , M ( t − 1 )) ≅ M ( t − 3 ) , t − 1 ∈ / Z ≥ 0 , Der δ ( s l 2 , M ( − t − 1 ) ) ≅ M ( 1 − t ) , − t − 1 ∉ Z ≥ 0 . (41) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M(-t-1))\cong M(1-t),\qquad-t-1\notin\mathbb{Z}_{\ge0}. \tag*{(41)} Der δ ( sl 2 , M ( − t − 1 )) ≅ M ( 1 − t ) , − t − 1 ∈ / Z ≥ 0 . (41) A basis ( D k ) k ≥ 0 (D_k)_{k\ge0} ( D k ) k ≥ 0 is given by (38) with n = t − 1 n=t-1 n = t − 1 for M ( t − 1 ) M(t-1) M ( t − 1 ) , and by (37) for M ( − t − 1 ) M(-t-1) M ( − t − 1 ) .
(ii) For derivations with values in infinite-dimensional simple lowest weight modules, the nonzero spaces are
Der δ ( s l 2 , M ′ ( 1 − t ) ) ≅ M ′ ( 3 − t ) , 1 − t ∉ Z ≤ 0 , \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M'(1-t))\cong M'(3-t),\qquad1-t\notin\mathbb{Z}_{\le0}, Der δ ( sl 2 , M ′ ( 1 − t )) ≅ M ′ ( 3 − t ) , 1 − t ∈ / Z ≤ 0 , Der δ ( s l 2 , M ′ ( t + 1 ) ) ≅ M ′ ( t − 1 ) , t + 1 ∉ Z ≤ 0 . (42) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M'(t+1))\cong M'(t-1),\qquad t+1\notin\mathbb{Z}_{\le0}. \tag*{(42)} Der δ ( sl 2 , M ′ ( t + 1 )) ≅ M ′ ( t − 1 ) , t + 1 ∈ / Z ≤ 0 . (42) For each M M M in (i), a basis is ( D k ∘ θ : s l 2 → M θ ) k ≥ 0 (D_k\circ\theta:\mathfrak{sl}_2\to M^\theta)_{k\ge0} ( D k ∘ θ : sl 2 → M θ ) k ≥ 0 .
(iii) The simple dense modules admitting nonzero derivations are V ( ξ , t 2 ) V(\xi,t^2) V ( ξ , t 2 ) , with t − 1 , − t − 1 ∉ ξ t-1,-t-1\notin\xi t − 1 , − t − 1 ∈ / ξ . Under the natural action,
Der δ ( s l 2 , V ( ξ , t 2 ) ) ≅ V ( ξ , ( t − 2 ) 2 ) . (43) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,V(\xi,t^2))\cong V(\xi,(t-2)^2). \tag*{(43)} Der δ ( sl 2 , V ( ξ , t 2 )) ≅ V ( ξ , ( t − 2 ) 2 ) . (43)
A basis ( D μ ) μ ∈ ξ (D_\mu)_{\mu\in\xi} ( D μ ) μ ∈ ξ of the derivation space is given by
D μ ( e ) = − ( μ + t + 1 ) ( μ + t − 1 ) 4 v μ + 2 , (44) D_\mu(e)=-\frac{(\mu+t+1)(\mu+t-1)}{4}v_{\mu+2}, \tag*{(44)} D μ ( e ) = − 4 ( μ + t + 1 ) ( μ + t − 1 ) v μ + 2 , (44)
D μ ( h ) = ( μ + t − 1 ) v μ , D μ ( f ) = v μ − 2 . D_\mu(h)=(\mu+t-1)v_\mu,\qquad D_\mu(f)=v_{\mu-2}. D μ ( h ) = ( μ + t − 1 ) v μ , D μ ( f ) = v μ − 2 .
Proof. For M ( λ ) M(\lambda) M ( λ ) , the Casimir equation λ ( λ + 2 ) = t 2 − 1 \lambda(\lambda+2)=t^2-1 λ ( λ + 2 ) = t 2 − 1 gives λ = t − 1 \lambda=t-1 λ = t − 1 or λ = − t − 1 \lambda=-t-1 λ = − t − 1 . For λ = t − 1 \lambda=t-1 λ = t − 1 , set
D 0 ( e ) = − λ ( λ − 1 ) v 0 , D 0 ( h ) = 2 ( λ − 1 ) v 1 , D 0 ( f ) = v 2 . D_0(e)=-\lambda(\lambda-1)v_0,\qquad D_0(h)=2(\lambda-1)v_1,\qquad D_0(f)=v_2. D 0 ( e ) = − λ ( λ − 1 ) v 0 , D 0 ( h ) = 2 ( λ − 1 ) v 1 , D 0 ( f ) = v 2 .
For λ = − t − 1 \lambda=-t-1 λ = − t − 1 , set
D 0 ( e ) = D 0 ( h ) = 0 , D 0 ( f ) = v 0 . D_0(e)=D_0(h)=0,\qquad D_0(f)=v_0. D 0 ( e ) = D 0 ( h ) = 0 , D 0 ( f ) = v 0 .
The verification of (16) in the proof of Proposition 4.1 uses only (33) and remains valid with n n n replaced by λ \lambda λ . The two maps have weights λ − 2 \lambda-2 λ − 2 and λ + 2 \lambda+2 λ + 2 , respectively, and in both cases
e ⋅ D 0 = ( e D 0 ( e ) , e D 0 ( h ) + 2 D 0 ( e ) , e D 0 ( f ) − D 0 ( h ) ) = 0. e\cdot D_0=(eD_0(e),eD_0(h)+2D_0(e),eD_0(f)-D_0(h))=0. e ⋅ D 0 = ( e D 0 ( e ) , e D 0 ( h ) + 2 D 0 ( e ) , e D 0 ( f ) − D 0 ( h )) = 0.
Theorem 3.7 (i),(ii) implies that the derivation space is simple, so it is generated by D 0 D_0 D 0 . The corresponding Verma modules have highest weights λ − 2 \lambda-2 λ − 2 and λ + 2 \lambda+2 λ + 2 , respectively, and are simple. Indeed, λ ∉ Z ≥ 0 \lambda\notin\mathbb{Z}_{\geq0} λ ∈ / Z ≥ 0 implies λ − 2 ∉ Z ≥ 0 \lambda-2\notin\mathbb{Z}_{\geq0} λ − 2 ∈ / Z ≥ 0 , whereas λ + 2 ∈ Z ≥ 0 \lambda+2\in\mathbb{Z}_{\geq0} λ + 2 ∈ Z ≥ 0 would force λ = − 1 , − 2 \lambda=-1,-2 λ = − 1 , − 2 , which give the excluded values δ = 2 , 1 \delta=2,1 δ = 2 , 1 . The homomorphism from the corresponding Verma module that sends its highest weight vector to D 0 D_0 D 0 is therefore an isomorphism. Formula (36) gives the stated basis D k = f k ⋅ D 0 D_k=f^k\cdot D_0 D k = f k ⋅ D 0 .
Since θ \theta θ is a Lie involution, twisting gives
Der δ ( s l 2 , M ) θ ⟶ ∼ Der δ ( s l 2 , M θ ) , D ⟼ D ∘ θ , \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)^\theta\overset{\sim}{\longrightarrow}\operatorname{Der}_{\delta}(\mathfrak{sl}_2,M^\theta),\qquad D\longmapsto D\circ\theta, Der δ ( sl 2 , M ) θ ⟶ ∼ Der δ ( sl 2 , M θ ) , D ⟼ D ∘ θ ,
with z ⋅ ( D ∘ θ ) = ( θ ( z ) ⋅ D ) ∘ θ z\cdot(D\circ\theta)=(\theta(z)\cdot D)\circ\theta z ⋅ ( D ∘ θ ) = ( θ ( z ) ⋅ D ) ∘ θ . Applying this isomorphism to the two highest weight modules proves (42) and the corresponding basis formulas.
A homogeneous map with values in a dense module has the form
D ( e ) = α v μ + 2 , D ( h ) = β v μ , D ( f ) = γ v μ − 2 . D(e)=\alpha v_{\mu+2},\qquad D(h)=\beta v_\mu,\qquad D(f)=\gamma v_{\mu-2}. D ( e ) = α v μ + 2 , D ( h ) = β v μ , D ( f ) = γ v μ − 2 .
The second and third equations of (16) give
β = ( μ + t − 1 ) γ , \beta=(\mu+t-1)\gamma, β = ( μ + t − 1 ) γ ,
α = t 2 − ( μ − 1 ) 2 4 γ − t + 1 2 β = − ( μ + t + 1 ) ( μ + t − 1 ) 4 γ . \alpha=\frac{t^2-(\mu-1)^2}{4}\gamma-\frac{t+1}{2}\beta=-\frac{(\mu+t+1)(\mu+t-1)}{4}\gamma. α = 4 t 2 − ( μ − 1 ) 2 γ − 2 t + 1 β = − 4 ( μ + t + 1 ) ( μ + t − 1 ) γ .
Substitution into the first equation gives
( μ − t + 1 ) α = − ( μ − t + 1 ) ( μ + t + 1 ) ( μ + t − 1 ) 4 γ = t 2 − ( μ + 1 ) 2 4 β . \begin{aligned}
(\mu-t+1)\alpha=-\frac{(\mu-t+1)(\mu+t+1)(\mu+t-1)}{4}\gamma\\
&=\frac{t^2-(\mu+1)^2}{4}\beta.
\end{aligned} ( μ − t + 1 ) α = − 4 ( μ − t + 1 ) ( μ + t + 1 ) ( μ + t − 1 ) γ = 4 t 2 − ( μ + 1 ) 2 β .
Hence the weight-μ \mu μ solution space is exactly C D μ \mathbb{C}D_\mu C D μ . Evaluation at f f f is injective on each homogeneous component of the derivation space, and
( e ⋅ D μ ) ( f ) = ( t 2 − ( μ − 1 ) 2 4 − ( μ + t − 1 ) ) v μ = ( t − 2 ) 2 − ( μ + 1 ) 2 4 v μ . (e\cdot D_\mu)(f)=\left(\frac{t^2-(\mu-1)^2}{4}-(\mu+t-1)\right)v_\mu=\frac{(t-2)^2-(\mu+1)^2}{4}v_\mu. ( e ⋅ D μ ) ( f ) = ( 4 t 2 − ( μ − 1 ) 2 − ( μ + t − 1 ) ) v μ = 4 ( t − 2 ) 2 − ( μ + 1 ) 2 v μ .
Together with the h h h - and f f f -actions, this gives
h ⋅ D μ = μ D μ , f ⋅ D μ = D μ − 2 , e ⋅ D μ = ( t − 2 ) 2 − ( μ + 1 ) 2 4 D μ + 2 . h\cdot D_\mu=\mu D_\mu,\qquad f\cdot D_\mu=D_{\mu-2},\qquad e\cdot D_\mu=\frac{(t-2)^2-(\mu+1)^2}{4}D_{\mu+2}. h ⋅ D μ = μ D μ , f ⋅ D μ = D μ − 2 , e ⋅ D μ = 4 ( t − 2 ) 2 − ( μ + 1 ) 2 D μ + 2 .
This proves (43) . The target module satisfies the same simplicity criterion, since t − 3 ≡ t − 1 t-3\equiv t-1 t − 3 ≡ t − 1 and 1 − t ≡ − t − 1 ( m o d 2 Z ) 1-t\equiv-t-1\pmod{2\mathbb{Z}} 1 − t ≡ − t − 1 ( mod 2 Z ) . □ \square □
Proposition 4.5. The nonzero 2 2 2 -derivation spaces with values in simple weight modules are
Der 2 ( s l 2 , M ( − 1 ) ) ≅ M ( 1 ) , Der 2 ( s l 2 , M ′ ( 1 ) ) ≅ M ′ ( − 1 ) , Der 2 ( s l 2 , V ( ξ , 0 ) ) ≅ V ( ξ , 4 ) , ξ ≠ 1 + 2 Z . (45) \begin{aligned}
\operatorname{Der}_2(\mathfrak{sl}_2,M(-1))&\cong M(1),\\
\operatorname{Der}_2(\mathfrak{sl}_2,M'(1))&\cong M'(-1),\\
\operatorname{Der}_2(\mathfrak{sl}_2,V(\xi,0))&\cong V(\xi,4),\qquad\xi\ne1+2\mathbb{Z}. \tag*{(45)}
\end{aligned} Der 2 ( sl 2 , M ( − 1 )) Der 2 ( sl 2 , M ′ ( 1 )) Der 2 ( sl 2 , V ( ξ , 0 )) ≅ M ( 1 ) , ≅ M ′ ( − 1 ) , ≅ V ( ξ , 4 ) , ξ = 1 + 2 Z . (45)
Bases for these spaces are given, respectively, by (4.5) with k ≥ 0 k\ge0 k ≥ 0 , by composing those maps with θ \theta θ , and by (4.12) at t = 0 t=0 t = 0 . The first two derivation modules have simple socles M ( − 3 ) M(-3) M ( − 3 ) and M ′ ( 3 ) M'(3) M ′ ( 3 ) , respectively. In both cases, the quotient by the socle is V ( 1 ) V(1) V ( 1 ) . The derivation module in the dense case is simple.
Proof. By Corollary 3.4, only simple weight modules with Casimir scalar − 1 -1 − 1 can occur; these are the modules listed in (4.13). Every nonzero homogeneous map s l 2 → M ( − 1 ) \mathfrak{sl}_2\to M(-1) sl 2 → M ( − 1 ) has weight 1 − 2 k 1-2k 1 − 2 k for some k ≥ 0 k\ge0 k ≥ 0 , and has the form
D ( e ) = α v k − 2 , D ( h ) = β v k − 1 , D ( f ) = γ v k . D(e)=\alpha v_{k-2},\qquad D(h)=\beta v_{k-1},\qquad D(f)=\gamma v_k. D ( e ) = α v k − 2 , D ( h ) = β v k − 1 , D ( f ) = γ v k .
For k ≥ 2 k\ge2 k ≥ 2 , the last two equations of (3.5) and e v k = − k 2 v k − 1 e v_k=-k^2v_{k-1} e v k = − k 2 v k − 1 give
β = − 2 k γ , α = − k 2 γ − 1 2 β = − k ( k − 1 ) γ . \beta=-2k\gamma,\qquad\alpha=-k^2\gamma-\frac{1}{2}\beta=-k(k-1)\gamma. β = − 2 k γ , α = − k 2 γ − 2 1 β = − k ( k − 1 ) γ .
The first equation is then automatically satisfied, since
( 2 − 2 k ) α = 2 k ( k − 1 ) 2 γ = − ( k − 1 ) 2 β . (2-2k)\alpha=2k(k-1)^2\gamma=-(k-1)^2\beta. ( 2 − 2 k ) α = 2 k ( k − 1 ) 2 γ = − ( k − 1 ) 2 β .
For k = 0 k=0 k = 0 , only D ( f ) D(f) D ( f ) can be nonzero; for k = 1 k=1 k = 1 , D ( e ) = 0 D(e)=0 D ( e ) = 0 and β = − 2 γ \beta=-2\gamma β = − 2 γ . Thus each homogeneous solution is γ D k \gamma D_k γ D k . Since
( e ⋅ D k ) ( f ) = e D k ( f ) − D k ( h ) = ( − k 2 + 2 k ) v k − 1 , (e\mathbin{\cdot}D_k)(f)=eD_k(f)-D_k(h)=(-k^2+2k)v_{k-1}, ( e ⋅ D k ) ( f ) = e D k ( f ) − D k ( h ) = ( − k 2 + 2 k ) v k − 1 ,
the action is
h ⋅ D k = ( 1 − 2 k ) D k , f ⋅ D k = D k + 1 , e ⋅ D k = k ( 2 − k ) D k − 1 . h\mathbin{\cdot}D_k=(1-2k)D_k,\qquad f\mathbin{\cdot}D_k=D_{k+1},\qquad e\mathbin{\cdot}D_k=k(2-k)D_{k-1}. h ⋅ D k = ( 1 − 2 k ) D k , f ⋅ D k = D k + 1 , e ⋅ D k = k ( 2 − k ) D k − 1 .
Consequently the space is isomorphic to M ( 1 ) M(1) M ( 1 ) , with
soc Der 2 ( s l 2 , M ( − 1 ) ) = ⨁ k ≥ 2 C D k ≅ M ( − 3 ) , M ( 1 ) / M ( − 3 ) ≅ V ( 1 ) . \operatorname{soc}\operatorname{Der}_2(\mathfrak{sl}_2,M(-1))=\bigoplus_{k\ge2}\mathbb{C}D_k\cong M(-3),\qquad M(1)/M(-3)\cong V(1). soc Der 2 ( sl 2 , M ( − 1 )) = k ≥ 2 ⨁ C D k ≅ M ( − 3 ) , M ( 1 ) / M ( − 3 ) ≅ V ( 1 ) .
Twisting by θ \theta θ gives the lowest weight case. The calculation for dense modules remains valid at t = 0 t=0 t = 0 . Its target V ( ξ , 4 ) V(\xi,4) V ( ξ , 4 ) is simple because 1 1 1 and − 3 -3 − 3 , the two weights excluded by the simplicity criterion, lie in 1 + 2 Z 1+2\mathbb{Z} 1 + 2 Z . □
Non-weight modules
Put R = C [ h ] R=\mathbb{C}[h] R = C [ h ] , K = C ( h ) K=\mathbb{C}(h) K = C ( h ) , and rank R M = dim K ( K ⊗ R M ) \operatorname{rank}_R M=\dim_K(K\otimes_R M) rank R M = dim K ( K ⊗ R M ) . Every simple non-weight s l 2 \mathfrak{sl}_2 sl 2 -module is R R R -torsion-free [20 Theorem 6.1.3] . We first treat arbitrary R R R -torsion-free modules with scalar Casimir action; finite generation over R R R is not required.
For an R R R -torsion-free s l 2 \mathfrak{sl}_2 sl 2 -module M M M , the action extends to K ⊗ R M K\otimes_R M K ⊗ R M by
e ( r ( h ) m ) = r ( h − 2 ) e m , f ( r ( h ) m ) = r ( h + 2 ) f m , h ( r ( h ) m ) = h r ( h ) m . (46) e(r(h)m)=r(h-2)em,\qquad f(r(h)m)=r(h+2)fm,\qquad h(r(h)m)=hr(h)m. \tag*{(46)} e ( r ( h ) m ) = r ( h − 2 ) e m , f ( r ( h ) m ) = r ( h + 2 ) f m , h ( r ( h ) m ) = h r ( h ) m . (46)
Here r ( h ) ∈ K r(h)\in K r ( h ) ∈ K and m ∈ M m\in M m ∈ M . We identify M M M with its image in K ⊗ R M K\otimes_R M K ⊗ R M . Rational factors in the formulas below act on the left in this localization. We specify the conditions under which the resulting vectors belong to M M M .
Proposition 4.6. Let M ≠ 0 M \ne0 M = 0 be an R R R -torsion-free s l 2 \mathfrak{sl}_2 sl 2 -module on which Ω \Omega Ω acts by a scalar κ \kappa κ , and put t = 2 / δ − 1 t = 2/\delta- 1 t = 2/ δ − 1 . Then Der δ ( s l 2 , M ) ≠ 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M) \ne0 Der δ ( sl 2 , M ) = 0 if and only if κ = t 2 − 1 \kappa= t^2 - 1 κ = t 2 − 1 . When κ = t 2 − 1 \kappa= t^2 - 1 κ = t 2 − 1 , put
Λ = { u ∈ M : e u ∈ ( h − t − 1 ) M , f u ∈ ( h + t + 1 ) M } . (47) \Lambda= \{u \in M : eu \in(h-t-1)M,\ fu \in(h+t+1)M\}. \tag*{(47)} Λ = { u ∈ M : e u ∈ ( h − t − 1 ) M , f u ∈ ( h + t + 1 ) M } . (47) Evaluation D ↦ D ( h ) D \mapsto D(h) D ↦ D ( h ) is a vector space isomorphism Der δ ( s l 2 , M ) → Λ \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M) \to\Lambda Der δ ( sl 2 , M ) → Λ , whose inverse is
D u ( e ) = ( h − t − 1 ) − 1 e u , D u ( h ) = u , D u ( f ) = ( h + t + 1 ) − 1 f u . (48) D_u(e) = (h-t-1)^{-1}eu,\qquad D_u(h) = u,\qquad D_u(f) = (h+t+1)^{-1}fu. \tag*{(48)} D u ( e ) = ( h − t − 1 ) − 1 e u , D u ( h ) = u , D u ( f ) = ( h + t + 1 ) − 1 f u . (48) Moreover, rank R Λ = rank R M \operatorname{rank}_{R}\Lambda= \operatorname{rank}_{R}M rank R Λ = rank R M . The natural action (5) induces the following action on Λ \Lambda Λ :
h ∗ u = h u , e ∗ u = h − t + 1 h − t − 1 e u , f ∗ u = h + t − 1 h + t + 1 f u . (49) h * u = hu,\qquad e * u = \frac{h-t+1}{h-t-1}eu,\qquad f * u = \frac{h+t-1}{h+t+1}fu. \tag*{(49)} h ∗ u = h u , e ∗ u = h − t − 1 h − t + 1 e u , f ∗ u = h + t + 1 h + t − 1 f u . (49) For this action, Ω \Omega Ω acts by ( t − 2 ) 2 − 1 (t-2)^2 - 1 ( t − 2 ) 2 − 1 .
Proof. Let D ∈ Der δ ( s l 2 , M ) D \in\operatorname{Der}_{\delta}(\mathfrak{sl}_2,M) D ∈ Der δ ( sl 2 , M ) and put u = D ( h ) u = D(h) u = D ( h ) . The first two equations of (16) give
( h − t − 1 ) D ( e ) = e u , ( h + t + 1 ) D ( f ) = f u . (h-t-1)D(e) = eu,\qquad(h+t+1)D(f) = fu. ( h − t − 1 ) D ( e ) = e u , ( h + t + 1 ) D ( f ) = f u .
Since M M M is torsion-free, these equations determine D ( e ) D(e) D ( e ) and D ( f ) D(f) D ( f ) uniquely, as in (48) . Hence evaluation at h h h is injective. To obtain the restriction on κ \kappa κ , use the identities
f e = κ − h 2 − 2 h 4 , e f = κ − h 2 + 2 h 4 . fe = \frac{\kappa-h^2-2h}{4},\qquad ef = \frac{\kappa-h^2+2h}{4}. f e = 4 κ − h 2 − 2 h , e f = 4 κ − h 2 + 2 h .
and the shift relations
e D ( f ) = ( h + t − 1 ) − 1 e f u , f D ( e ) = ( h − t + 1 ) − 1 f e u . eD(f) = (h+t-1)^{-1}efu,\qquad fD(e) = (h-t+1)^{-1}feu. eD ( f ) = ( h + t − 1 ) − 1 e f u , f D ( e ) = ( h − t + 1 ) − 1 f e u .
Consequently,
2 ( e D ( f ) − f D ( e ) ) = ( κ − h 2 + 2 h 2 ( h + t − 1 ) − κ − h 2 − 2 h 2 ( h − t + 1 ) ) u = ( t + 1 ) h 2 − ( t − 1 ) κ ( h − t + 1 ) ( h + t − 1 ) u . \begin{aligned}
2(eD(f)-fD(e)) =
\left(\frac{\kappa-h^2+2h}{2(h+t-1)}-\frac{\kappa-h^2-2h}{2(h-t+1)}\right)u \\
&= \frac{(t+1)h^2-(t-1)\kappa}{(h-t+1)(h+t-1)}u.
\end{aligned} 2 ( eD ( f ) − f D ( e )) = ( 2 ( h + t − 1 ) κ − h 2 + 2 h − 2 ( h − t + 1 ) κ − h 2 − 2 h ) u = ( h − t + 1 ) ( h + t − 1 ) ( t + 1 ) h 2 − ( t − 1 ) κ u .
Subtracting ( t + 1 ) u (t+1)u ( t + 1 ) u and using the third equation of (16) , we obtain
2 ( e D ( f ) − f D ( e ) ) − ( t + 1 ) u = ( t − 1 ) ( t 2 − 1 − κ ) ( h − t + 1 ) ( h + t − 1 ) u = 0. (50) 2(eD(f)-fD(e))-(t+1)u =
\frac{(t-1)(t^2-1-\kappa)}{(h-t+1)(h+t-1)}u = 0. \tag*{(50)} 2 ( eD ( f ) − f D ( e )) − ( t + 1 ) u = ( h − t + 1 ) ( h + t − 1 ) ( t − 1 ) ( t 2 − 1 − κ ) u = 0. (50)
If D ≠ 0 D \ne0 D = 0 , injectivity of evaluation gives u ≠ 0 u \ne0 u = 0 , so torsion-freeness and t ≠ 1 t \ne1 t = 1 imply κ = t 2 − 1 \kappa= t^2-1 κ = t 2 − 1 . Now suppose κ = t 2 − 1 \kappa= t^2-1 κ = t 2 − 1 . The first two equations show that evaluation takes values in Λ \Lambda Λ . Conversely, for u ∈ Λ u \in\Lambda u ∈ Λ , formula (48) defines a map D u : s l 2 → M D_u:\mathfrak{sl}_2 \to M D u : sl 2 → M satisfying the first two equations of (16) . The calculation in (50) shows that it also satisfies the third. Thus evaluation is an isomorphism onto Λ \Lambda Λ .
The shift relations make Λ \Lambda Λ an R R R -submodule. For m ∈ M m \in M m ∈ M , they give
e ( ( h − t + 1 ) ( h + t − 1 ) m ) = ( h − t − 1 ) ( h + t − 3 ) e m , e((h-t+1)(h+t-1)m) = (h-t-1)(h+t-3)em, e (( h − t + 1 ) ( h + t − 1 ) m ) = ( h − t − 1 ) ( h + t − 3 ) e m ,
f ( ( h − t + 1 ) ( h + t − 1 ) m ) = ( h − t + 3 ) ( h + t + 1 ) f m . f((h-t+1)(h+t-1)m) = (h-t+3)(h+t+1)fm. f (( h − t + 1 ) ( h + t − 1 ) m ) = ( h − t + 3 ) ( h + t + 1 ) f m .
Thus
( h − t + 1 ) ( h + t − 1 ) M ⊆ Λ ⊆ M . (51) (h-t+1)(h+t-1)M \subseteq\Lambda\subseteq M. \tag*{(51)} ( h − t + 1 ) ( h + t − 1 ) M ⊆ Λ ⊆ M . (51)
Since M ≠ 0 M \ne0 M = 0 is torsion-free, the leftmost submodule in (51) is nonzero. This proves Λ ≠ 0 \Lambda\ne0 Λ = 0 and completes the nonvanishing criterion. Localizing (51) gives K ⊗ R Λ = K ⊗ R M K \otimes_R \Lambda= K \otimes_R M K ⊗ R Λ = K ⊗ R M , hence the rank equality.
We now compute the transported action. The identities
( e ⋅ D u ) ( h ) = e u + 2 D u ( e ) , ( f ⋅ D u ) ( h ) = f u − 2 D u ( f ) , ( h ⋅ D u ) ( h ) = h u (e \mathbin{\cdot} D_u)(h) = eu + 2D_u(e),\qquad(f \mathbin{\cdot} D_u)(h) = fu - 2D_u(f),\qquad(h \mathbin{\cdot} D_u)(h) = hu ( e ⋅ D u ) ( h ) = e u + 2 D u ( e ) , ( f ⋅ D u ) ( h ) = f u − 2 D u ( f ) , ( h ⋅ D u ) ( h ) = h u
give (49) . Since f e = − 1 4 ( h − t + 1 ) ( h + t + 1 ) fe = -\frac{1}{4}(h-t+1)(h+t+1) f e = − 4 1 ( h − t + 1 ) ( h + t + 1 ) on M M M , we obtain
f ∗ ( e ∗ u ) = ( h + t − 1 ) ( h − t + 3 ) ( h + t + 1 ) ( h − t + 1 ) f e u . f * (e * u) = \frac{(h+t-1)(h-t+3)}{(h+t+1)(h-t+1)}feu. f ∗ ( e ∗ u ) = ( h + t + 1 ) ( h − t + 1 ) ( h + t − 1 ) ( h − t + 3 ) f e u .
It follows that
f ∗ ( e ∗ u ) = − ( h + t − 1 ) ( h − t + 3 ) 4 u , Ω ∗ u = ( ( t − 2 ) 2 − 1 ) u , (52) f * (e * u)=-\frac{(h+t-1)(h-t+3)}{4}u,\qquad\Omega* u=((t-2)^2-1)u, \tag*{(52)} f ∗ ( e ∗ u ) = − 4 ( h + t − 1 ) ( h − t + 3 ) u , Ω ∗ u = (( t − 2 ) 2 − 1 ) u , (52)
where the second identity follows from Ω ∗ u = h 2 u + 2 h u + 4 f ∗ ( e ∗ u ) \Omega* u=h^2u+2hu+4f*(e*u) Ω ∗ u = h 2 u + 2 h u + 4 f ∗ ( e ∗ u ) .
Theorem 4.7. Let t = 2 / δ − 1 t=2/\delta-1 t = 2/ δ − 1 , and let M M M be a simple non-weight s l 2 \mathfrak{sl}_2 sl 2 -module with Casimir scalar t 2 − 1 t^2-1 t 2 − 1 . Let Λ \Lambda Λ be given by (47) , with the action (49) .
(i) If δ ≠ 2 \delta\ne2 δ = 2 , then Der δ ( s l 2 , M ) ≅ Λ \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)\cong\Lambda Der δ ( sl 2 , M ) ≅ Λ is a simple non-weight module.
(ii) If δ = 2 \delta=2 δ = 2 , put
Λ 0 = ( h 2 − 1 ) M + ( h − 1 ) f M + ( h + 1 ) e M . (53) \Lambda_0=(h^2-1)M+(h-1)fM+(h+1)eM. \tag*{(53)} Λ 0 = ( h 2 − 1 ) M + ( h − 1 ) f M + ( h + 1 ) e M . (53) Then Λ 0 \Lambda_0 Λ 0 is the simple essential socle of Λ \Lambda Λ , and there is an exact sequence
0 ⟶ Λ 0 ⟶ Λ ⟶ V ( 1 ) ⊕ m ⟶ 0 , m ∈ Z ≥ 0 . (54) 0\longrightarrow\Lambda_0\longrightarrow\Lambda\longrightarrow V(1)^{\oplus m}\longrightarrow0,\qquad m\in\mathbb{Z}_{\ge0}. \tag*{(54)} 0 ⟶ Λ 0 ⟶ Λ ⟶ V ( 1 ) ⊕ m ⟶ 0 , m ∈ Z ≥ 0 . (54) The maps
e ˉ : M / ( h + 1 ) M ⟶ M / ( h − 1 ) M , f ˉ : M / ( h − 1 ) M ⟶ M / ( h + 1 ) M (55) \bar e:M/(h+1)M\longrightarrow M/(h-1)M,\qquad\bar f:M/(h-1)M\longrightarrow M/(h+1)M \tag*{(55)} e ˉ : M / ( h + 1 ) M ⟶ M / ( h − 1 ) M , f ˉ : M / ( h − 1 ) M ⟶ M / ( h + 1 ) M (55) induced by the original actions satisfy e ˉ f ˉ = f ˉ e ˉ = 0 \bar e\bar f=\bar f\bar e=0 e ˉ f ˉ = f ˉ e ˉ = 0 , and
m = dim C ker e ˉ im f ˉ = dim C ker f ˉ im e ˉ , 0 ≤ m ≤ rank R M . (56) m=\dim_{\mathbb{C}}\frac{\ker\bar e}{\operatorname{im}\bar f}
=\dim_{\mathbb{C}}\frac{\ker\bar f}{\operatorname{im}\bar e},\qquad0\le m\le\operatorname{rank}_{\mathbb{R}}M. \tag*{(56)} m = dim C im f ˉ ker e ˉ = dim C im e ˉ ker f ˉ , 0 ≤ m ≤ rank R M . (56) The module Λ \Lambda Λ is indecomposable and is simple exactly when m = 0 m=0 m = 0 . The sequence (54) is nonsplit when m > 0 m>0 m > 0 .
Proof. By Proposition 4.6, Der δ ( s l 2 , M ) ≅ Λ ≠ 0 \operatorname{Der}_{\delta}(\mathfrak{sl}_2,M)\cong\Lambda\ne0 Der δ ( sl 2 , M ) ≅ Λ = 0 . For δ ≠ 2 \delta\ne2 δ = 2 , Theorem 3.7(i),(ii) gives simplicity, since M ≇ V ( 1 ) M\not\cong V(1) M ≅ V ( 1 ) . The R \mathbb{R} R -module Λ ⊆ M \Lambda\subseteq M Λ ⊆ M is torsion-free, so it has no nonzero weight vectors. This proves (i).
For (ii), let t = 0 t=0 t = 0 , so Ω \Omega Ω acts by − 1 -1 − 1 on M M M . Equations (47) and (49) become
Λ = { u ∈ M : e u ∈ ( h − 1 ) M , f u ∈ ( h + 1 ) M } , \Lambda=\{u\in M:eu\in(h-1)M,\ fu\in(h+1)M\}, Λ = { u ∈ M : e u ∈ ( h − 1 ) M , f u ∈ ( h + 1 ) M } ,
h ∗ u = h u , e ∗ u = h + 1 h − 1 e u , f ∗ u = h − 1 h + 1 f u . (57) h*u=hu,\qquad e*u=\frac{h+1}{h-1}eu,\qquad f*u=\frac{h-1}{h+1}fu. \tag*{(57)} h ∗ u = h u , e ∗ u = h − 1 h + 1 e u , f ∗ u = h + 1 h − 1 f u . (57)
By the proof of Theorem 3.7(iii), the operator J J J on Hom C ( s l 2 , M ) \operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M) Hom C ( sl 2 , M ) satisfies
im J = soc Der 2 ( s l 2 , M ) , \operatorname{im}J=\operatorname{soc}\operatorname{Der}_2(\mathfrak{sl}_2,M), im J = soc Der 2 ( sl 2 , M ) ,
and this socle is simple and essential. Its image under evaluation at h h h can be computed from J = − 2 S 2 + 3 S − 1 J=-2S^2+3S-1 J = − 2 S 2 + 3 S − 1 in (25) : for F ∈ Hom C ( s l 2 , M ) F\in\operatorname{Hom}_{\mathbb{C}}(\mathfrak{sl}_2,M) F ∈ Hom C ( sl 2 , M ) ,
( J F ) ( h ) = ( f h − 3 f ) F ( e ) − ( f e + e f + 1 ) F ( h ) + ( e h + 3 e ) F ( f ) = ( h − 1 ) f F ( e ) + h 2 − 1 2 F ( h ) + ( h + 1 ) e F ( f ) , \begin{aligned}
(JF)(h)=(fh-3f)F(e)-(fe+ef+1)F(h)+(eh+3e)F(f) \\
&=(h-1)fF(e)+\frac{h^2-1}{2}F(h)+(h+1)eF(f),
\end{aligned} ( J F ) ( h ) = ( f h − 3 f ) F ( e ) − ( f e + e f + 1 ) F ( h ) + ( e h + 3 e ) F ( f ) = ( h − 1 ) f F ( e ) + 2 h 2 − 1 F ( h ) + ( h + 1 ) e F ( f ) ,
where the second equality uses 2 ( e f + f e ) + h 2 = − 1 2(ef+fe)+h^2=-1 2 ( e f + f e ) + h 2 = − 1 on M M M . The three values F ( e ) F(e) F ( e ) , F ( h ) F(h) F ( h ) , F ( f ) F(f) F ( f ) vary independently over M M M . Thus evaluation maps im J \operatorname{im}J im J onto Λ 0 \Lambda_0 Λ 0 , proving the socle assertion.
To compute Λ / Λ 0 \Lambda/\Lambda_0 Λ/ Λ 0 , consider the maps in (55) . They are well defined because the shift relations give
e ( ( h + 1 ) M ) ⊆ ( h − 1 ) M , f ( ( h − 1 ) M ) ⊆ ( h + 1 ) M . e((h+1)M)\subseteq(h-1)M,\qquad f((h-1)M)\subseteq(h+1)M. e (( h + 1 ) M ) ⊆ ( h − 1 ) M , f (( h − 1 ) M ) ⊆ ( h + 1 ) M .
The identities
f e = − 1 4 ( h + 1 ) 2 , e f = − 1 4 ( h − 1 ) 2 fe=-\frac{1}{4}(h+1)^2,\qquad ef=-\frac{1}{4}(h-1)^2 f e = − 4 1 ( h + 1 ) 2 , e f = − 4 1 ( h − 1 ) 2
give f ˉ e ˉ = e ˉ f ˉ = 0 \bar f\bar e=\bar e\bar f=0 f ˉ e ˉ = e ˉ f ˉ = 0 . Put P = ( h 2 − 1 ) M P=(h^2-1)M P = ( h 2 − 1 ) M . Under the Chinese remainder isomorphism
M / P ≅ M / ( h + 1 ) M ⊕ M / ( h − 1 ) M , M/P\cong M/(h+1)M\oplus M/(h-1)M, M / P ≅ M / ( h + 1 ) M ⊕ M / ( h − 1 ) M ,
the two conditions in (57) restrict the first coordinate to ker e ˉ \ker\bar e ker e ˉ and the second to ker f ˉ \ker\bar f ker f ˉ . For v ∈ M v\in M v ∈ M , the generators of Λ 0 / P \Lambda_0/P Λ 0 / P have images
( h − 1 ) f v + P ⟼ ( − 2 ( f v + ( h + 1 ) M ) , 0 ) , (h-1)fv+P\longmapsto\bigl(-2(fv+(h+1)M),0\bigr), ( h − 1 ) f v + P ⟼ ( − 2 ( f v + ( h + 1 ) M ) , 0 ) ,
( h + 1 ) e v + P ⟼ ( 0 , 2 ( e v + ( h − 1 ) M ) ) . (h+1)ev+P\longmapsto\bigl(0,2(ev+(h-1)M)\bigr). ( h + 1 ) e v + P ⟼ ( 0 , 2 ( e v + ( h − 1 ) M ) ) .
Together with (53) , this gives
Λ / P ≅ ker e ˉ ⊕ ker f ˉ , Λ 0 / P ≅ im f ˉ ⊕ im e ˉ . \Lambda/P \cong\ker\bar{e} \oplus\ker\bar{f}, \qquad\Lambda_0/P \cong\operatorname{im} \bar{f} \oplus\operatorname{im} \bar{e}. Λ/ P ≅ ker e ˉ ⊕ ker f ˉ , Λ 0 / P ≅ im f ˉ ⊕ im e ˉ .
Since h ∗ u = h u h*u=hu h ∗ u = h u , the quotient Q = Λ / Λ 0 Q=\Lambda/\Lambda_0 Q = Λ/ Λ 0 has weight spaces
Q − 1 ≅ ker e ˉ im f ˉ , Q 1 ≅ ker f ˉ im e ˉ , Q ν = 0 ( ν ≠ ± 1 ) . Q_{-1}\cong\frac{\ker\bar{e}}{\operatorname{im}\bar{f}}, \qquad Q_1\cong\frac{\ker\bar{f}}{\operatorname{im}\bar{e}}, \qquad Q_\nu=0\quad(\nu\ne\pm1). Q − 1 ≅ im f ˉ ker e ˉ , Q 1 ≅ im e ˉ ker f ˉ , Q ν = 0 ( ν = ± 1 ) .
By Theorem 3.7 (iii) and the socle assertion above, Q ≅ V ( 1 ) ⊕ m Q\cong V(1)^{\oplus m} Q ≅ V ( 1 ) ⊕ m for some finite m m m . Thus m = dim C Q − 1 = dim C Q 1 m=\dim_{\mathbb{C}}Q_{-1}=\dim_{\mathbb{C}}Q_1 m = dim C Q − 1 = dim C Q 1 , proving (54) and both expressions for m m m .
To bound m m m , fix α ∈ C \alpha\in\mathbb{C} α ∈ C and choose v 1 , … , v r ∈ M v_1,\ldots,v_r\in M v 1 , … , v r ∈ M whose classes in M / ( h − α ) M M/(h-\alpha)M M / ( h − α ) M are linearly independent. If these lifts were K K K -linearly dependent, clearing denominators would give
∑ i = 1 r a i ( h ) v i = 0 , a i ( h ) ∈ R , \sum_{i=1}^{r}a_i(h)v_i=0,\qquad a_i(h)\in R, i = 1 ∑ r a i ( h ) v i = 0 , a i ( h ) ∈ R ,
with the coefficients not all zero. Since M M M is torsion-free, we may cancel their largest common power of h − α h-\alpha h − α and assume that at least one a i ( α ) a_i(\alpha) a i ( α ) is nonzero. Reduction modulo ( h − α ) M (h-\alpha)M ( h − α ) M then gives
∑ i = 1 r a i ( α ) ( v i + ( h − α ) M ) = 0 , \sum_{i=1}^{r}a_i(\alpha)\bigl(v_i+(h-\alpha)M\bigr)=0, i = 1 ∑ r a i ( α ) ( v i + ( h − α ) M ) = 0 ,
contrary to the choice of the classes. The lifts are therefore K K K -linearly independent, so dim C M / ( h − α ) M ≤ rank R M \dim_{\mathbb{C}}M/(h-\alpha)M\le\operatorname{rank}_R M dim C M / ( h − α ) M ≤ rank R M . In particular,
m ≤ dim C M / ( h + 1 ) M ≤ rank R M , m\le\dim_{\mathbb{C}}M/(h+1)M\le\operatorname{rank}_R M, m ≤ dim C M / ( h + 1 ) M ≤ rank R M ,
which proves the bound in (56) . Every nonzero submodule of Λ \Lambda Λ meets the simple socle Λ 0 \Lambda_0 Λ 0 , so Λ \Lambda Λ is indecomposable. If m = 0 m=0 m = 0 , then Λ = Λ 0 \Lambda=\Lambda_0 Λ = Λ 0 is simple; if m > 0 m>0 m > 0 , then Λ 0 \Lambda_0 Λ 0 is proper, and its essentiality rules out a splitting of (54) . □
Block's classification
To identify Λ \Lambda Λ and its socle in Block’s classification [1 , 4 ] , we use the notation of [20 ] . Set
B = K [ X , X − 1 ; σ ] , σ ( h ) = h − 2 , q d ( h ) = d − h 2 − 2 h 4 ( d ∈ C ) . (58) B=K[X,X^{-1};\sigma],\qquad\sigma(h)=h-2,\qquad q_d(h)=\frac{d-h^2-2h}{4}\quad(d\in\mathbb{C}). \tag*{(58)} B = K [ X , X − 1 ; σ ] , σ ( h ) = h − 2 , q d ( h ) = 4 d − h 2 − 2 h ( d ∈ C ) . (58)
Thus X r ( h ) = r ( h − 2 ) X Xr(h)=r(h-2)X X r ( h ) = r ( h − 2 ) X . The embedding
A d ↪ B , e ↦ X , f ↦ q d ( h ) X − 1 , h ↦ h (59) A_d\hookrightarrow B,\qquad e\mapsto X,\qquad f\mapsto q_d(h)X^{-1},\qquad h\mapsto h \tag*{(59)} A d ↪ B , e ↦ X , f ↦ q d ( h ) X − 1 , h ↦ h (59)
identifies B B B with the localization of A d A_d A d at R ∖ { 0 } R\setminus\{0\} R ∖ { 0 } , since X − 1 = q d ( h ) − 1 f X^{-1}=q_d(h)^{-1}f X − 1 = q d ( h ) − 1 f . With our Casimir normalization, the parameter in [20 ] satisfies ( λ + 1 ) 2 = d + 1 (\lambda+1)^2=d+1 ( λ + 1 ) 2 = d + 1 . For an irreducible nonunit b ∈ B b\in B b ∈ B , put
W b = B / B b , L d ( b ) = soc A d W b . (60) W_b=B/Bb,\qquad L_d(b)=\operatorname{soc}_{A_d}W_b. \tag*{(60)} W b = B / B b , L d ( b ) = soc A d W b . (60)
For each d ∈ C d\in\mathbb{C} d ∈ C , we regard W b W_b W b as an A d A_d A d -module through the embedding (59) . By [20 Proposition 6.2.8 and Theorem 6.3.6] , L d ( b ) L_d(b) L d ( b ) is simple, K ⊗ R L d ( b ) = W b K\otimes_R L_d(b)=W_b K ⊗ R L d ( b ) = W b , and every simple non-weight A d A_d A d -module is obtained in this way. The socle is essential. Indeed, for each nonzero v ∈ W b v\in W_b v ∈ W b , the equality K ⊗ R L d ( b ) = W b K\otimes_R L_d(b)=W_b K ⊗ R L d ( b ) = W b gives a nonzero r ( h ) ∈ R r(h)\in R r ( h ) ∈ R such that r ( h ) v ∈ L d ( b ) r(h)v\in L_d(b) r ( h ) v ∈ L d ( b ) . Since W b W_b W b is a K K K -vector space, r ( h ) v ≠ 0 r(h)v\ne0 r ( h ) v = 0 . Thus every nonzero A d A_d A d -submodule of W b W_b W b meets L d ( b ) L_d(b) L d ( b ) . Write b = ∑ r = i j a r ( h ) X r b=\sum_{r=i}^{j}a_r(h)X^r b = ∑ r = i j a r ( h ) X r , with i < j i<j i < j and a i a j ≠ 0 a_i a_j\ne0 a i a j = 0 . For every ℓ ∈ Z \ell\in\mathbb{Z} ℓ ∈ Z , we have
0 = X ℓ b ( 1 + B b ) = ∑ r = i j a r ( h − 2 ℓ ) ( X r + ℓ + B b ) . 0=X^\ell b(1+Bb)=\sum_{r=i}^{j}a_r(h-2\ell)(X^{r+\ell}+Bb). 0 = X ℓ b ( 1 + B b ) = r = i ∑ j a r ( h − 2 ℓ ) ( X r + ℓ + B b ) .
Since the two end coefficients are invertible in K K K , recursion in both directions expresses every X k + B b X^k+Bb X k + B b in the K K K -span of $X^i+Bb,\ldots,X^{j-1}+Bb. Hence
rank R L d ( b ) = dim K W b ≤ j − i < ∞ . \operatorname{rank}_R L_d(b)=\dim_K W_b\le j-i<\infty. rank R L d ( b ) = dim K W b ≤ j − i < ∞.
Corollary 4.8. Let t = 2 / δ − 1 t = 2/\delta- 1 t = 2/ δ − 1 , let M = L t 2 − 1 ( b ) ⊂ W b M = L_{t^2-1}(b) \subset W_b M = L t 2 − 1 ( b ) ⊂ W b , and let Λ \Lambda Λ be given by (47) . The δ \delta δ -derivations with values in M M M are parametrized bijectively by
w ∈ ( h − t + 1 ) − 1 Λ ⊂ W b (61) w \in(h-t+1)^{-1}\Lambda\subset W_b \tag*{(61)} w ∈ ( h − t + 1 ) − 1 Λ ⊂ W b (61) via the formulas
D w ( e ) = X w , D w ( h ) = ( h − t + 1 ) w , D w ( f ) = − 1 4 ( h − t + 3 ) ( h − t + 1 ) X − 1 w . (62) \begin{aligned}
D_w(e) &= Xw,\\
D_w(h) &= (h-t+1)w,\\
D_w(f) &= -\frac{1}{4}(h-t+3)(h-t+1)X^{-1}w. \tag*{(62)}
\end{aligned} D w ( e ) D w ( h ) D w ( f ) = X w , = ( h − t + 1 ) w , = − 4 1 ( h − t + 3 ) ( h − t + 1 ) X − 1 w . (62) Equivalently, the three values in (62) must lie in M M M .
(i) If δ ≠ 2 \delta\ne2 δ = 2 , the parameter w w w ranges over L ( t − 2 ) 2 − 1 ( b ) L_{(t-2)^2-1}(b) L ( t − 2 ) 2 − 1 ( b ) , and
Der δ ( s l 2 , L t 2 − 1 ( b ) ) ≅ L ( t − 2 ) 2 − 1 ( b ) . (63) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,L_{t^2-1}(b)) \cong L_{(t-2)^2-1}(b). \tag*{(63)} Der δ ( sl 2 , L t 2 − 1 ( b )) ≅ L ( t − 2 ) 2 − 1 ( b ) . (63) (ii) If δ = 2 \delta= 2 δ = 2 , there is an exact sequence
0 ⟶ L 3 ( b ) ⟶ Der 2 ( s l 2 , L − 1 ( b ) ) ⟶ V ( 1 ) ⊕ m b ⟶ 0 , (64) 0 \longrightarrow L_3(b) \longrightarrow\operatorname{Der}_2(\mathfrak{sl}_2,L_{-1}(b)) \longrightarrow V(1)^{\oplus m_b} \longrightarrow0, \tag*{(64)} 0 ⟶ L 3 ( b ) ⟶ Der 2 ( sl 2 , L − 1 ( b )) ⟶ V ( 1 ) ⊕ m b ⟶ 0 , (64) where m b m_b m b is the multiplicity in (56) for M = L − 1 ( b ) M = L_{-1}(b) M = L − 1 ( b ) . The inclusion is w ↦ D w w \mapsto D_w w ↦ D w , and its image is the simple essential socle.
Proof. Write c = t 2 − 1 c=t^2-1 c = t 2 − 1 and c ′ = ( t − 2 ) 2 − 1 c'=(t-2)^2-1 c ′ = ( t − 2 ) 2 − 1 , as in (13) . Substituting u = ( h − t + 1 ) w u=(h-t+1)w u = ( h − t + 1 ) w in (48) and using
e ( ( h − t + 1 ) w ) = ( h − t − 1 ) X w , q c ( h ) = − 1 4 ( h − t + 1 ) ( h + t + 1 ) e((h-t+1)w)=(h-t-1)Xw,\qquad q_c(h)=-\frac{1}{4}(h-t+1)(h+t+1) e (( h − t + 1 ) w ) = ( h − t − 1 ) X w , q c ( h ) = − 4 1 ( h − t + 1 ) ( h + t + 1 )
gives (62) . For the f f f -action in (49) , we obtain
( h − t + 1 ) − 1 h + t − 1 h + t + 1 q c ( h ) X − 1 ( h − t + 1 ) = ( h + t − 1 ) ( h − t + 3 ) ( h − t + 1 ) ( h + t + 1 ) q c ( h ) X − 1 = q c ′ ( h ) X − 1 . \begin{aligned}
(h-t+1)^{-1}\frac{h+t-1}{h+t+1}q_c(h)X^{-1}(h-t+1)
&=\frac{(h+t-1)(h-t+3)}{(h-t+1)(h+t+1)}q_c(h)X^{-1}\\
&=q_{c'}(h)X^{-1}.
\end{aligned} ( h − t + 1 ) − 1 h + t + 1 h + t − 1 q c ( h ) X − 1 ( h − t + 1 ) = ( h − t + 1 ) ( h + t + 1 ) ( h + t − 1 ) ( h − t + 3 ) q c ( h ) X − 1 = q c ′ ( h ) X − 1 .
Together with the h h h - and e e e -actions, this gives
h ∗ w = h w , e ∗ w = X w , f ∗ w = q c ′ ( h ) X − 1 w . (65) h * w=hw,\qquad e * w=Xw,\qquad f * w=q_{c'}(h)X^{-1}w. \tag*{(65)} h ∗ w = h w , e ∗ w = X w , f ∗ w = q c ′ ( h ) X − 1 w . (65)
This is the A c ′ A_{c'} A c ′ -module structure on W b W_b W b . If δ ≠ 2 \delta\ne2 δ = 2 , Theorem 4.7(i) shows that ( h − t + 1 ) − 1 Λ (h-t+1)^{-1}\Lambda ( h − t + 1 ) − 1 Λ is a simple A c ′ A_{c'} A c ′ -submodule of W b W_b W b . Since L c ′ ( b ) L_{c'}(b) L c ′ ( b ) is the simple essential socle of W b W_b W b , it follows that ( h − t + 1 ) − 1 Λ = L c ′ ( b ) (h-t+1)^{-1}\Lambda=L_{c'}(b) ( h − t + 1 ) − 1 Λ = L c ′ ( b ) . If δ = 2 \delta=2 δ = 2 , the change of variables u = ( h + 1 ) w u=(h+1)w u = ( h + 1 ) w makes ( h + 1 ) − 1 Λ 0 (h+1)^{-1}\Lambda_0 ( h + 1 ) − 1 Λ 0 a simple A 3 A_3 A 3 -submodule of W b W_b W b . It therefore equals L 3 ( b ) L_3(b) L 3 ( b ) , so
Λ 0 = ( h + 1 ) L 3 ( b ) . (66) \Lambda_0=(h+1)L_3(b). \tag*{(66)} Λ 0 = ( h + 1 ) L 3 ( b ) . (66)
Now (54) gives (64) . □
Remark 4.9. For M = L − 1 ( b ) M=L_{-1}(b) M = L − 1 ( b ) , the exact sequences (32) and (64) give
[ T ( M ) : L 3 ( b ) ] = 2 , [ T ( M ) : V ( 1 ) ] = m b . [T(M):L_3(b)]=2,\qquad[T(M):V(1)]=m_b. [ T ( M ) : L 3 ( b )] = 2 , [ T ( M ) : V ( 1 )] = m b .
For M = L − 1 ( b ) M=L_{-1}(b) M = L − 1 ( b ) of R R R -rank one, (56) gives m b ∈ { 0 , 1 } m_b\in\{0,1\} m b ∈ { 0 , 1 } . The following examples realize both values.
Example 4.10. For η ∈ C × \eta\in\mathbb{C}^{\times} η ∈ C × and κ ∈ C \kappa\in\mathbb{C} κ ∈ C , the simple Whittaker module W η , κ = C [ h ] W_{\eta,\kappa}=\mathbb{C}[h] W η , κ = C [ h ] [20 ] has action
h ⋅ p = h p , e ⋅ p = η p ( h − 2 ) , f ⋅ p = κ − h 2 − 2 h 4 η p ( h + 2 ) . (67) h\cdot p=hp,\qquad e\cdot p=\eta p(h-2),\qquad f\cdot p=\frac{\kappa-h^2-2h}{4\eta}p(h+2). \tag*{(67)} h ⋅ p = h p , e ⋅ p = η p ( h − 2 ) , f ⋅ p = 4 η κ − h 2 − 2 h p ( h + 2 ) . (67) This module is non-weight and R R R -free of rank one, with e ⋅ 1 = η e\cdot1=\eta e ⋅ 1 = η and Ω = κ \Omega=\kappa Ω = κ . The parameters in the cited model are α = η − 1 \alpha=\eta^{-1} α = η − 1 and ( λ + 1 ) 2 = κ + 1 (\lambda+1)^2=\kappa+1 ( λ + 1 ) 2 = κ + 1 .
By Proposition 4.6, nonzero derivations exist exactly when κ = t 2 − 1 \kappa=t^2-1 κ = t 2 − 1 . In this case, the condition f u ∈ ( h + t + 1 ) W η , κ fu\in(h+t+1)W_{\eta,\kappa} f u ∈ ( h + t + 1 ) W η , κ is automatic, whereas e u ∈ ( h − t − 1 ) W η , κ eu\in(h-t-1)W_{\eta,\kappa} e u ∈ ( h − t − 1 ) W η , κ is equivalent to u ∈ ( h − t + 1 ) C [ h ] u\in(h-t+1)\mathbb{C}[h] u ∈ ( h − t + 1 ) C [ h ] . Writing u = ( h − t + 1 ) p ( h ) u=(h-t+1)p(h) u = ( h − t + 1 ) p ( h ) , we obtain all derivations in the form
\begin{aligned}
D_p(e)&=\eta p(h-2),\\
D_p(h)&=(h-t+1)p(h),\\
D_p(f)&=-\frac{(h-t+1)(h-t+3)}{4\eta}p(h+2),
\end{aligned}
\qquad p\in\mathbb{C}[h]. \tag*{(68)}\end{aligned}
Under this parametrization, (49) gives
h ∗ p = h p , e ∗ p = η p ( h − 2 ) , f ∗ p = ( t − 2 ) 2 − 1 − h 2 − 2 h 4 η p ( h + 2 ) . h*p=hp,\qquad e*p=\eta p(h-2),\qquad f*p=\frac{(t-2)^2-1-h^2-2h}{4\eta}p(h+2). h ∗ p = h p , e ∗ p = η p ( h − 2 ) , f ∗ p = 4 η ( t − 2 ) 2 − 1 − h 2 − 2 h p ( h + 2 ) .
Consequently
Der δ ( s l 2 , W η , t 2 − 1 ) ≅ W η , ( t − 2 ) 2 − 1 . (69) \operatorname{Der}_{\delta}(\mathfrak{sl}_2,W_{\eta,t^2-1})\cong W_{\eta,(t-2)^2-1}. \tag*{(69)} Der δ ( sl 2 , W η , t 2 − 1 ) ≅ W η , ( t − 2 ) 2 − 1 . (69)
At δ = 2 \delta=2 δ = 2 , the target is the simple Whittaker module W η , 3 W_{\eta,3} W η , 3 . Hence the quotient in (54) is zero.
Example 4.11. On M = C [ h ] M=\mathbb{C}[h] M = C [ h ] , put
h ⋅ p = h p , e ⋅ p = η ( h − 1 ) p ( h − 2 ) , f ⋅ p = − h + 1 4 η p ( h + 2 ) , η ≠ 0. (70) h\cdot p=hp,\qquad e\cdot p=\eta(h-1)p(h-2),\qquad f\cdot p=-\frac{h+1}{4\eta}p(h+2),\qquad\eta\ne0. \tag*{(70)} h ⋅ p = h p , e ⋅ p = η ( h − 1 ) p ( h − 2 ) , f ⋅ p = − 4 η h + 1 p ( h + 2 ) , η = 0. (70) These operators satisfy the relations in (12) , and Ω \Omega Ω acts by − 1 -1 − 1 . Since h h h acts by multiplication, every nonzero submodule is an ideal ( p ( h ) ) (p(h)) ( p ( h )) of C [ h ] \mathbb{C}[h] C [ h ] , with p ≠ 0 p\ne0 p = 0 . Suppose p p p is nonconstant and let α \alpha α be a root. Stability under the action forces p ( α − 2 ) = 0 p(\alpha-2)=0 p ( α − 2 ) = 0 unless α = 1 \alpha=1 α = 1 , and p ( α + 2 ) = 0 p(\alpha+2)=0 p ( α + 2 ) = 0 unless α = − 1 \alpha=-1 α = − 1 . The descending sequence of roots must terminate at 1 1 1 , and the ascending sequence at − 1 -1 − 1 , since p p p has only finitely many roots. Thus α = 1 + 2 k = − 1 − 2 ℓ \alpha=1+2k=-1-2\ell α = 1 + 2 k = − 1 − 2 ℓ for some k , ℓ ∈ Z ≥ 0 k,\ell\in\mathbb{Z}_{\geq0} k , ℓ ∈ Z ≥ 0 , a contradiction. Hence p p p is a nonzero constant and M M M is simple.
Here Λ = C [ h ] \Lambda=\mathbb{C}[h] Λ = C [ h ] , so all 2 2 2 -derivations are
D p ( e ) = η p ( h − 2 ) , D p ( h ) = p ( h ) , D p ( f ) = − 1 4 η p ( h + 2 ) . D_p(e)=\eta p(h-2),\qquad D_p(h)=p(h),\qquad D_p(f)=-\frac{1}{4\eta}p(h+2). D p ( e ) = η p ( h − 2 ) , D p ( h ) = p ( h ) , D p ( f ) = − 4 η 1 p ( h + 2 ) . The induced action on the parameter p p p is
h ∗ p = h p , e ∗ p = η ( h + 1 ) p ( h − 2 ) , f ∗ p = − h − 1 4 η p ( h + 2 ) . (71) h*p=hp,\qquad e*p=\eta(h+1)p(h-2),\qquad f*p=-\frac{h-1}{4\eta}p(h+2). \tag*{(71)} h ∗ p = h p , e ∗ p = η ( h + 1 ) p ( h − 2 ) , f ∗ p = − 4 η h − 1 p ( h + 2 ) . (71) Formula (53) gives Λ 0 = ( h 2 − 1 ) C [ h ] \Lambda_0=(h^2-1)\mathbb{C}[h] Λ 0 = ( h 2 − 1 ) C [ h ] . Both spaces in (55) have dimension one, and e ˉ = f ˉ = 0 \bar e=\bar f=0 e ˉ = f ˉ = 0 . Hence (56) gives m = 1 m=1 m = 1 . Let v + , v − v_+,v_- v + , v − be the standard basis of V ( 1 ) V(1) V ( 1 ) , with h v ± = ± v ± hv_\pm=\pm v_\pm h v ± = ± v ± , e v − = v + ev_-=v_+ e v − = v + , and f v + = v − fv_+=v_- f v + = v − . The map
Der 2 ( s l 2 , M ) ⟶ V ( 1 ) , D p ⟼ p ( 1 ) v + + 2 η p ( − 1 ) v − \operatorname{Der}_2(\mathfrak{sl}_2,M)\longrightarrow V(1),\qquad D_p\longmapsto p(1)v_++2\eta p(-1)v_- Der 2 ( sl 2 , M ) ⟶ V ( 1 ) , D p ⟼ p ( 1 ) v + + 2 η p ( − 1 ) v − is surjective. By (71) ,
( e ∗ p ) ( 1 ) = 2 η p ( − 1 ) , ( e ∗ p ) ( − 1 ) = 0 , (e*p)(1)=2\eta p(-1),\qquad(e*p)(-1)=0, ( e ∗ p ) ( 1 ) = 2 η p ( − 1 ) , ( e ∗ p ) ( − 1 ) = 0 , ( f ∗ p ) ( 1 ) = 0 , ( f ∗ p ) ( − 1 ) = p ( 1 ) 2 η . (f*p)(1)=0,\qquad(f*p)(-1)=\frac{p(1)}{2\eta}. ( f ∗ p ) ( 1 ) = 0 , ( f ∗ p ) ( − 1 ) = 2 η p ( 1 ) . Together with h ∗ p = h p h*p=hp h ∗ p = h p , these identities show that the map is s l 2 \mathfrak{sl}_2 sl 2 -equivariant. Its kernel consists of the D p D_p D p with p ( 1 ) = p ( − 1 ) = 0 p(1)=p(-1)=0 p ( 1 ) = p ( − 1 ) = 0 , or equivalently p ∈ ( h 2 − 1 ) C [ h ] p\in(h^2-1)\mathbb{C}[h] p ∈ ( h 2 − 1 ) C [ h ] . Thus we obtain the nonsplit sequence of Theorem 4.7, with first map p ↦ D p p\mapsto D_p p ↦ D p :
0 ⟶ ( h 2 − 1 ) C [ h ] ⟶ Der 2 ( s l 2 , M ) ⟶ V ( 1 ) ⟶ 0. 0\longrightarrow(h^2-1)\mathbb{C}[h]\longrightarrow\operatorname{Der}_2(\mathfrak{sl}_2,M)\longrightarrow V(1)\longrightarrow0. 0 ⟶ ( h 2 − 1 ) C [ h ] ⟶ Der 2 ( sl 2 , M ) ⟶ V ( 1 ) ⟶ 0.
References
School of Mathematics, Jilin University, Changchun 130012, Jilin, China Email address: jiangjlu@163.com
Department of Mathematics, Huizhou Normal University, Huizhou 313000, Zhejiang, China Email address: liudong@hznu.edu.cn
Department of Mathematics, Huizhou Normal University, Huizhou 313000, Zhejiang, China Email address: pei@hznu.edu.cn