Introduction
In the study of algebraic structures, there are two fundamental problems. The first is the extension problem, and the second is the deformation problem. To be more precise, the extension problem aims to classify all possible isomorphism classes of extensions of two given algebraic structures, and the deformation problem aims to understand how these structures vary in families. However, these two apparently unrelated problems can be treated in tandem with the help of the notion of global extensions. The main aim of this article is to provide a differential graded Lie algebraic (d g L a dgLa d g L a for short) approach to understand several problems involving natural interactions between the extension problem and the deformation problem of Lie algebras.
Let us first briefly recall these two basic problems.
Deformation problem: Formal deformations of associative algebras and the related cohomology theory were first studied systematically by Gerstenhaber in a series of works [17 , 18 , 19 , 20 ] . Later, these techniques were applied by Nijenhuis and Richardson to study the deformation problem for Lie algebras [35 , 36 , 37 ] . There is a well-known slogan, often attributed to Deligne, Drinfeld and Kontsevich: every reasonable formal deformation problem is controlled by a dgLa, determined up to quasi-isomorphism. This slogan has been made into a rigorous theorem by Lurie and Pridham [30 , 40 ] in the derived setting.
Parallel to the developments from the mathematical aspects, similar constructions appear in the world of physics. The notion of contractions of Lie algebras dates back to the work of Segal [41 ] and Inönü–Wigner [27 , 28 ] , aiming to understand the limiting structure of symmetries under small perturbations. Meanwhile, contractions are closely related to several well-known contexts in geometry: deformation (understood via a family of Lie algebra structures over a certain base space) and degeneration (described in terms of orbits in the variety of Lie brackets). Beyond their various incarnations in geometry, contractions are found to play a prominent role in the representation theory of Lie groups and p p p -adic groups. A typical example is their appearance in the proof of the Mackey bijection, relating the tempered dual of a reductive Lie group to the full unitary dual of the corresponding Cartan motion group [1 , 2 , 3 ] and in the constructions of algebraic families of Harish-Chandra modules [10 , 9 , 23 ] . Meanwhile, a relative theory, i.e., a contraction of subalgebraic groups in an ambient algebraic group as well as the corresponding sub-Lie algebras may also provide an algebraic framework for comparison of different periods.
0 → g ′ → g ′ ′ → g → 0 0 \to\mathfrak{g}' \to\mathfrak{g}'' \to\mathfrak{g} \to0 0 → g ′ → g ′′ → g → 0
up to isomorphism. If g ′ \mathfrak{g}' g ′ is abelian, this problem is well understood and described with the help of the ordinary Lie algebra cohomology with abelian coefficients. However, it is a complicated task to develop a non-abelian cohomology theory due to the lack of constructions of cochain complexes. Eilenberg and MacLane [14 ] introduced the notion of non-abelian group cohomology to understand extensions of abstract groups with non-abelian kernels (see also [25 , 34 ] ). Subsequently, the non-abelian cohomology theory was generalized to the context of Lie algebras and used to classify corresponding non-abelian extensions [7 , 8 , 16 , 26 , 33 ] . Using a different approach, Frégier constructed a dgLa and used its Maurer–Cartan elements as well as the corresponding Deligne groupoid to classify isomorphism classes of non-abelian extensions in [16 ] , where the comparison with non-abelian cohomology is also treated carefully.
Later, Agore and Militaru initiated the study of the global extension problem in several algebraic settings [4 , 5 , 6 ] . Specializing to Lie algebras, let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) be a Lie algebra, V ′ ′ V^{\prime\prime} V ′′ a vector space, and p : V ′ ′ → V p:V^{\prime\prime}\to V p : V ′′ → V a surjective linear map. The global extension problem seeks to describe and classify all Lie algebra structures ν ′ ′ \nu^{\prime\prime} ν ′′ on V ′ ′ V^{\prime\prime} V ′′ such that p : ( V ′ ′ , ν ′ ′ ) → g p:(V^{\prime\prime},\nu^{\prime\prime})\to\mathfrak{g} p : ( V ′′ , ν ′′ ) → g is a homomorphism of Lie algebras. To understand the global extension problem, they introduced the concept of global cohomology. The main feature of the global extension problem is that the Lie algebra structure on ker p \ker p ker p is not fixed, of which the variation is related to the deformation problem.
The purpose of this paper is to develop a dgLa description of the global extension problem and to use it to study the interaction between global extensions, non-abelian extensions, and formal deformations. More precisely, we stress the following problems:
We describe isomorphism classes of global extensions by Maurer–Cartan formulism and identify the twisted dgLas controlling the formal deformations of given extensions.
We obtain a geometric interpretation of decomposing global extensions as a union of nonabelian extensions by varying the Lie algebra structure on the kernel, in terms of fibration of Deligne groupoids.
We classify infinitesimal deformations of a global extension up to equivalence (unframed equivalence) in terms of cohomology of the controlling dgLas. We also take care of the obstruction theory.
We realize a particular family of Inönü–Wigner contractions from g ⋊ ad g \mathfrak{g}\rtimes_{\operatorname{ad}}\mathfrak{g} g ⋊ ad g to V a b ⋊ ad g V_{\mathrm{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ⋊ ad g as a flat family of global extensions over A 1 \mathbb{A}^{1} A 1 . Moreover, we also present a Maurer–Cartan characterization of this contraction based on the machinery developed by us.
For these purposes, we define dgLas L V , V ′ L_{V,V^{\prime}} L V , V ′ and L V , V ′ u n L^{\mathrm{un}}_{V,V^{\prime}} L V , V ′ un (see Proposition 4.2 ). Our first main result connects these dgLas with the moduli problem of global extensions.
Theorem 1.1. (Theorem 4.3 , Proposition 3.8 , Theorem 3.9 , Theorem 4.8 , Theorem 4.10 , Proposition 4.7 ) Let ( V ′ ′ , p , g ) (V^{\prime\prime},p,\mathfrak{g}) ( V ′′ , p , g ) be a global extension datum, g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) be a Lie algebra, and V ′ V^{\prime} V ′ denote ker p \ker p ker p .
(1) The correspondence
( π V ′ , ϱ , ω ) ⟼ π V ′ + ϱ + ω (\pi_{V^{\prime}},\varrho,\omega)\longmapsto\pi_{V^{\prime}}+\varrho+\omega ( π V ′ , ϱ , ω ) ⟼ π V ′ + ϱ + ω induces a bijection
GZ 2 ( g , V ′ ) ≅ MC ( L V , V ′ ) . \operatorname{GZ}^{2}(\mathfrak{g},V^{\prime})\cong\operatorname{MC}(L_{V,V^{\prime}}). GZ 2 ( g , V ′ ) ≅ MC ( L V , V ′ ) . (2) In the framed case, the action of Hom ( V , V ′ ) \operatorname{Hom}(V,V^{\prime}) Hom ( V , V ′ ) agrees with the gauge action. In the unframed case, the chosen section identifies transport of structure with the action of H V , V ′ H_{V,V^{\prime}} H V , V ′ . There are isomorphisms of algebraic groupoids
Gext ( g , V ′ ) / / G p ≅ GZ 2 ( g , V ′ ) / / Hom ( V , V ′ ) , \operatorname{Gext}(\mathfrak{g},V^{\prime})//G_{p}\cong\operatorname{GZ}^{2}(\mathfrak{g},V^{\prime})//\operatorname{Hom}(V,V^{\prime}), Gext ( g , V ′ ) // G p ≅ GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) , Gext ( g , V ′ ) / / G p u n ≅ GZ 2 ( g , V ′ ) / / H V , V ′ . \operatorname{Gext}(\mathfrak{g},V^{\prime})//G_{p}^{\mathrm{un}}\cong\operatorname{GZ}^{2}(\mathfrak{g},V^{\prime})//H_{V,V^{\prime}}. Gext ( g , V ′ ) // G p un ≅ GZ 2 ( g , V ′ ) // H V , V ′ . Their sets of connected components give the bijections
GExt ( g , V ′ ) ≅ GH 2 ( g , V ′ ) , GExt u n ( g , V ′ ) ≅ GH u n , 2 ( g , V ′ ) . \operatorname{GExt}(\mathfrak{g},V^{\prime})\cong\operatorname{GH}^{2}(\mathfrak{g},V^{\prime}),\qquad\operatorname{GExt}^{\mathrm{un}}(\mathfrak{g},V^{\prime})\cong\operatorname{GH}^{\mathrm{un},2}(\mathfrak{g},V^{\prime}). GExt ( g , V ′ ) ≅ GH 2 ( g , V ′ ) , GExt un ( g , V ′ ) ≅ GH un , 2 ( g , V ′ ) .
(3) Fix a global extension ν ′ ′ \nu^{\prime\prime} ν ′′ with the associated Maurer–Cartan element α s = π + ϱ + ω \alpha_s=\pi+\varrho+\omega α s = π + ϱ + ω . Write L s = L α s L^s=L^{\alpha_s} L s = L α s and L u n , s = ( L u n ) α s L^{\mathrm{un},s}=(L^{\mathrm{un}})^{\alpha_s} L un , s = ( L un ) α s for the twists, both with differential d s = d + [ α s , − ] d^s=d+[\alpha_s,-] d s = d + [ α s , − ] . For every local Artinian k k k -algebra A A A with the residue field k k k and the maximal ideal m A \mathfrak{m}_A m A , there are equivalences of based deformation groupoids
Def E p , [ ν ′ ′ ] ( A ) ≃ Del ( L s ⊗ m A ) , Def E , [ ν ′ ′ ] ( A ) ≃ Del ( L u n , s ⊗ m A ) . \operatorname{Def}_{E_p,[\nu^{\prime\prime}]}(A)\simeq\operatorname{Del}(L^s\otimes\mathfrak{m}_A),\qquad\operatorname{Def}_{E,[\nu^{\prime\prime}]}(A)\simeq\operatorname{Del}(L^{\mathrm{un},s}\otimes\mathfrak{m}_A). Def E p , [ ν ′′ ] ( A ) ≃ Del ( L s ⊗ m A ) , Def E , [ ν ′′ ] ( A ) ≃ Del ( L un , s ⊗ m A ) .
natural in A A A . A different choice of s s s produces a transformation in (19) and induces isomorphisms of the corresponding twisted dgLas.
Our second main result provides a geometric description of the relation between global and non-abelian extensions in terms of Deligne groupoids. More precisely, let L V ′ L_{V'} L V ′ denote the graded Lie algebra whose Maurer–Cartan elements are Lie algebra structures on V ′ V' V ′ . There is a short exact sequence of dgLas
0 ⟶ L ~ V , V ′ → ι L V , V ′ → pr L V ′ ⟶ 0. 0\longrightarrow\widetilde{L}_{V,V'}\xrightarrow{\iota}L_{V,V'}\xrightarrow{\operatorname{pr}}L_{V'}\longrightarrow0. 0 ⟶ L V , V ′ ι L V , V ′ pr L V ′ ⟶ 0.
Theorem 1.2. (Proposition 5.3, Theorem 5.5, Corollary 5.6) The morphisms in the above short exact sequence induce functors
Del ( L ~ V , V ′ ) → ι ∗ Del ( L V , V ′ ) → pr ∗ Del ( L V ′ ) . \operatorname{Del}(\widetilde{L}_{V,V'})\xrightarrow{\iota_*}\operatorname{Del}(L_{V,V'})\xrightarrow{\operatorname{pr}_*}\operatorname{Del}(L_{V'}). Del ( L V , V ′ ) ι ∗ Del ( L V , V ′ ) pr ∗ Del ( L V ′ ) . The first functor is an inclusion of groupoids, and the second is a fibration of groupoids. For every π ∈ MC ( L V ′ ) \pi\in\operatorname{MC}(L_{V'}) π ∈ MC ( L V ′ ) , put α π = ( 0 , 0 , π ) \alpha_\pi=(0,0,\pi) α π = ( 0 , 0 , π ) . Then the fiber of pr ∗ \operatorname{pr}_* pr ∗ over π \pi π is naturally isomorphic to the Deligne groupoid of the twisted dgLa L ~ V , V ′ α π \widetilde{L}_{V,V'}^{\alpha_\pi} L V , V ′ α π , i.e.,
pr ∗ − 1 ( π ) ≃ Del ( L ~ V , V ′ α π ) . \operatorname{pr}_*^{-1}(\pi)\simeq\operatorname{Del}(\widetilde{L}_{V,V'}^{\alpha_\pi}). pr ∗ − 1 ( π ) ≃ Del ( L V , V ′ α π ) . This twisted dgLa governs non-abelian extensions of g \mathfrak{g} g by the fixed Lie algebra ( V ′ , π ) (V',\pi) ( V ′ , π ) . Consequently, there is a decomposition
Del ( L V , V ′ ) ≃ ∐ π ∈ MC ( L V ′ ) Del ( L ~ V , V ′ α π ) . \operatorname{Del}(L_{V,V'})\simeq\coprod_{\pi\in\operatorname{MC}(L_{V'})}\operatorname{Del}(\widetilde{L}_{V,V'}^{\alpha_\pi}). Del ( L V , V ′ ) ≃ π ∈ MC ( L V ′ ) ∐ Del ( L V , V ′ α π ) .
The dgLa description further provides cohomological tools for studying infinitesimal and higher-order deformations. Fix a global extension structure μ ′ ′ \mu^{\prime\prime} μ ′′ and a section s s s of p p p . Let ζ s = π V ′ s + ϱ s + ω s \zeta_s=\pi_{V'}^s+\varrho^s+\omega^s ζ s = π V ′ s + ϱ s + ω s be its associated global 2-cocycle. The short exact sequence above can be twisted by ζ s \zeta_s ζ s to induce a long exact sequence in cohomology, from which information of infinitesimal deformations and obstructions can be read.
Theorem 1.3. (Proposition 6.3, Proposition 6.5, Proposition 6.4) With the above notations, the following statements hold.
(1) Framed equivalence (resp. unframed equivalence) classes of infinitesimal deformations of μ ′ ′ \mu^{\prime\prime} μ ′′ are classified by H 1 ( L V , V ′ ) δ H^1(L_{V,V'})^\delta H 1 ( L V , V ′ ) δ (resp. H 1 ( L V , V ′ u n ) δ H^1(L_{V,V'}^{\mathrm{un}})^\delta H 1 ( L V , V ′ un ) δ ).
(2) Let π V ′ s + t π V ′ ′ \pi_{V'}^s+t\pi_{V'}' π V ′ s + t π V ′ ′ be an infinitesimal deformation of the kernel bracket over k [ t ] / ( t 2 ) k[t]/(t^2) k [ t ] / ( t 2 ) . Its obstruction to being lifted to an infinitesimal deformation of the global extension is
Ob π V ′ s ( π V ′ ′ , ϱ s , ω s ) = [ [ ϱ s + ω s , π V ′ ′ ] ] ∈ H 2 ( L ~ V , V ′ δ ) . \operatorname{Ob}_{\pi_{V'}^s}(\pi_{V'}',\varrho^s,\omega^s)=\left[[\varrho^s+\omega^s,\pi_{V'}']\right]\in H^2(\widetilde{L}_{V,V'}^\delta). Ob π V ′ s ( π V ′ ′ , ϱ s , ω s ) = [ [ ϱ s + ω s , π V ′ ′ ] ] ∈ H 2 ( L V , V ′ δ ) . The kernel deformation admits such a lift if and only if this class vanishes.
(3) Suppose that an n n n -th order deformation of μ ′ ′ \mu^{\prime\prime} μ ′′ corresponds to
α s + ∑ i = 1 n t i η i , η i = π V , i ′ + ϱ i + ω i . \alpha_s+\sum_{i=1}^{n}t^i\eta_i,\qquad\eta_i=\pi V'_{,i}+\varrho_i+\omega_i. α s + i = 1 ∑ n t i η i , η i = π V , i ′ + ϱ i + ω i .
Then it extends to an ( n + 1 ) (n+1) ( n + 1 ) -st order deformation if and only if
Ob n = [ 1 2 ∑ i + j = n + 1 [ η i , η j ] ] ∈ H 2 ( L V , V ′ ) δ \operatorname{Ob}_n=\left[\frac{1}{2}\sum_{i+j=n+1}[\eta_i,\eta_j]\right]\in H^2(L_{V,V'})^\delta Ob n = [ 2 1 i + j = n + 1 ∑ [ η i , η j ] ] ∈ H 2 ( L V , V ′ ) δ
vanishes.
These obstruction classes make two kinds of variation present in the global extension problem explicit. The first records whether a variation of the Lie algebra structure on the kernel is compatible with the remaining extension datum, while the second is the usual obstruction to prolonging a finite-order Maurer–Cartan deformation.
Finally, we revisit the theory of contraction of Lie algebras. In particular, we show how a particular family of contractions fits into our framework of global extensions. Furthermore, we describe the relation between contractions and unframed moduli problems.
Theorem 1.4. (Proposition 7.4, Proposition 7.5, Proposition 7.6, Proposition 7.7) Fix a copy V ′ V' V ′ of the underlying vector space of g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) , and write μ V ′ \mu_{V'} μ V ′ for the induced Lie bracket. On ( V ′ ⊕ V ) ⊗ k [ t ] (V'\oplus V)\otimes k[t] ( V ′ ⊕ V ) ⊗ k [ t ] , the bracket
[ ( u , x ) , ( v , y ) ] t = ( ad x v − ad y u + t μ V ′ ( u , v ) , μ ( x , y ) ) [(u,x),(v,y)]_t=(\operatorname{ad}_xv-\operatorname{ad}_yu+t\mu_{V'}(u,v),\mu(x,y)) [( u , x ) , ( v , y ) ] t = ( ad x v − ad y u + t μ V ′ ( u , v ) , μ ( x , y )) defines a flat family of global extensions of g ⊗ k [ t ] \mathfrak{g}\otimes k[t] g ⊗ k [ t ] with projection p ⊗ 1 p\otimes1 p ⊗ 1 . Its fiber at t = 1 t=1 t = 1 (resp. t = 0 t=0 t = 0 ) are g ⋊ ad g \mathfrak{g}\rtimes_{\operatorname{ad}}\mathfrak{g} g ⋊ ad g (resp. V a b ′ ⋊ ad g V'_{\mathrm{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ′ ⋊ ad g ). Over G m \mathbb{G}_m G m , the map ( u , x ) ↦ ( t u , x ) (u,x)\mapsto(tu,x) ( u , x ) ↦ ( t u , x ) trivializes the family as a kernel-unframed extension.
Suppose that g \mathfrak{g} g is nonabelian. Let s a b = ad s_{\mathrm{ab}}=\operatorname{ad} s ab = ad denote the cocycle of the special fiber. Write c ( t ) = s a b + t μ V ′ c(t)=s_{\mathrm{ab}}+t\mu_{V'} c ( t ) = s ab + t μ V ′ .
The Lie bracket pr ( c ( t ) ) = t μ V ′ \operatorname{pr}(c(t))=t\mu_{V'} pr ( c ( t )) = t μ V ′ defines a family of Lie algebra structures on the kernel. In G p u n G_{\mathrm{p}}^{\mathrm{un}} G p un , the kernel rescalings Θ λ ( u , x ) = ( λ − 1 u , x ) \Theta_\lambda(u,x)=(\lambda^{-1}u,x) Θ λ ( u , x ) = ( λ − 1 u , x ) form a one-dimensional torus acting by Θ λ ⋅ c ( t ) = c ( λ t ) \Theta_\lambda\cdot c(t)=c(\lambda t) Θ λ ⋅ c ( t ) = c ( λ t ) . The orbit of c ( 1 ) c(1) c ( 1 ) is therefore c ( G m ) c(\mathbb{G}_m) c ( G m ) , and its closure in MC ( L V , V ′ ) \operatorname{MC}(L_{V,V'}) MC ( L V , V ′ ) is the affine line c ( A 1 ) c(\mathbb{A}^1) c ( A 1 ) . Its unique boundary point c ( 0 ) c(0) c ( 0 ) has abelian kernel and hence represents a distinct unframed isomorphism class.
In the twist at s a b s_{\mathrm{ab}} s ab , the formal germ of the family at t = 0 t=0 t = 0 is represented by the Maurer–Cartan element t μ V ′ t\mu_{V'} t μ V ′ . Its tangent class [ μ V ′ ] [\mu_{V'}] [ μ V ′ ] is nonzero in both
H 1 ( L V , V ′ s a b ) and H 1 ( L V , V ′ u n , s a b ) . H^1(L_{V,V'}^{s_{\mathrm{ab}}})\qquad\text{and}\qquad H^1(L_{V,V'}^{\mathrm{un},s_{\mathrm{ab}}}). H 1 ( L V , V ′ s ab ) and H 1 ( L V , V ′ un , s ab ) . Thus, the contraction defines a nontrivial based formal deformation of the special fiber in both moduli problems.
The paper is organized as follows. In Section 2, we recall basic concepts of differential graded Lie algebras, Deligne groupoids, quotient stacks and how they are related to the formal deformation problem via Maurer–Cartan formulation. In Section 3, we review the notion of global and non-abelian extensions of Lie algebras and introduce the concept of unframed equivalence for global extensions. Moreover, we obtain classification results in terms of action groupoids. In Section 4, we construct various dgLas and identify the corresponding Deligne groupoids with classification groupoids of global extensions. We also prove that certain twists of these dgLas control the corresponding formal deformation problems. In Section 5, we reinterpret the global extension problem in terms of fibrations of Deligne groupoids with the help of the natural short exact sequence of dgLas. In Section 6, we revisit infinitesimal deformations from the perspective of our machinery and construct obstruction classes for lifting and extending deformations. Finally, in Section 7, we will revisit the classical problem of contractions of Lie algebras in terms of unframed global extensions.
All vector spaces are finite-dimensional over a fixed field k k k of characteristic zero. All tensor and exterior products are taken over k k k .
Acknowledgements. This work was initiated to understand the special example of Inönü-Wigner contraction from g ⊕ g \mathfrak{g} \oplus\mathfrak{g} g ⊕ g to V a b ⋊ a d g V_{\mathrm{ab}} \rtimes_{\mathrm{ad}} \mathfrak{g} V ab ⋊ ad g from the point of view of deformation functors with additional structures and the corresponding dgLas. Unfortunately, we do not have a satisfying answer regarding that particular example. However, during our study, we found that the particular example of contraction from g ⋊ a d g \mathfrak{g} \rtimes_{\mathrm{ad}} \mathfrak{g} g ⋊ ad g to V a b ⋊ a d g V_{\mathrm{ab}} \rtimes_{\mathrm{ad}} \mathfrak{g} V ab ⋊ ad g fits well into the theory of unframed global extensions. Moreover, this approach enjoys a nice geometric description and does not seem to appear in the literature. We believe it is interesting and meaningful to record these results. During the preparation of this work, we thank Yunhe Sheng for enlightening discussions. The second author thanks Rui Chen, Caihua Luo and Jiandi Zou for their interest in this topic. Some results in this paper have been announced by the first author in a talk given at Guangzhou University in June 2026. He thanks Yunnan Li for the invitation and hospitality. The first author is supported by NSFC (12401076). The second author is supported by the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No. 17305223, 17308324).
AI declaration. We acknowledge the use of ChatGPT for finding typos, revising texts, proofreading and correcting mistakes after we finished the first draft of the paper. All mathematical work are done by the authors, who take full responsibility for the manuscript.
Differential graded Lie algebras and deformation of Lie algebra structures
In this section, we briefly review the basic concepts of dgLas and their role in studying the deformation problem of Lie algebras. These results are well known to experts and will be organized in a suitable way for our application.
Differential graded Lie algebras and Deligne groupoids
In this subsection, we collect some basic facts about dgLas and Deligne groupoids.
Definition 2.1. ([36 §1.3] ) A differential graded Lie algebra (dgLa) is a Z \mathbb{Z} Z -graded vector space L = ⨁ n ∈ Z L n L = \bigoplus_{n \in\mathbb{Z}} L^n L = ⨁ n ∈ Z L n equipped with a bilinear bracket [ ⋅ , ⋅ ] L : L p × L q ⟶ L p + q [\cdot,\cdot]_L : L^p \times L^q \longrightarrow L^{p+q} [ ⋅ , ⋅ ] L : L p × L q ⟶ L p + q and a differential d : L n ⟶ L n + 1 d : L^n \longrightarrow L^{n+1} d : L n ⟶ L n + 1 with the property that, for all homogeneous elements a , b , c ∈ L a,b,c \in L a , b , c ∈ L , we have d 2 = 0 d^2 = 0 d 2 = 0 and
d ( [ a , b ] L ) = [ d ( a ) , b ] L + ( − 1 ) ∣ a ∣ [ a , d ( b ) ] L , d([a,b]_L) = [d(a),b]_L + (-1)^{|a|}[a,d(b)]_L, d ([ a , b ] L ) = [ d ( a ) , b ] L + ( − 1 ) ∣ a ∣ [ a , d ( b ) ] L , [ a , b ] L = − ( − 1 ) ∣ a ∣ ∣ b ∣ [ b , a ] L , [a,b]_L = -(-1)^{|a||b|}[b,a]_L, [ a , b ] L = − ( − 1 ) ∣ a ∣∣ b ∣ [ b , a ] L , [ a , [ b , c ] L ] L = [ [ a , b ] L , c ] L + ( − 1 ) ∣ a ∣ ∣ b ∣ [ b , [ a , c ] L ] L , [a,[b,c]_L]_L = [[a,b]_L,c]_L + (-1)^{|a||b|}[b,[a,c]_L]_L, [ a , [ b , c ] L ] L = [[ a , b ] L , c ] L + ( − 1 ) ∣ a ∣∣ b ∣ [ b , [ a , c ] L ] L , where ∣ a ∣ |a| ∣ a ∣ denotes the degree of a a a .
If ( L , [ ⋅ , ⋅ ] L ) (L,[\cdot,\cdot]_L) ( L , [ ⋅ , ⋅ ] L ) only satisfies the last two conditions, it is usually called a graded Lie algebra (gLa) . We will always regard a gLa as a dgLa with the differential d = 0 d=0 d = 0 .
Definition 2.2. ([36 p. 15] ) A Maurer–Cartan element of a d g L a dgLa d g L a ( L , [ ⋅ , ⋅ ] L , d ) (L,[\cdot,\cdot]_L,d) ( L , [ ⋅ , ⋅ ] L , d ) is an element x ∈ L 1 x\in L^1 x ∈ L 1 satisfying the Maurer–Cartan equation
d x + 1 2 [ x , x ] L = 0. dx+\frac{1}{2}[x,x]_L=0. d x + 2 1 [ x , x ] L = 0.
Denote by MC ( L ) \operatorname{MC}(L) MC ( L ) the set of Maurer–Cartan elements of ( L , [ ⋅ , ⋅ ] L , d ) (L,[\cdot,\cdot]_L,d) ( L , [ ⋅ , ⋅ ] L , d ) .
For any d g L a dgLa d g L a ( L , [ ⋅ , ⋅ ] L , d ) (L,[\cdot,\cdot]_L,d) ( L , [ ⋅ , ⋅ ] L , d ) and α ∈ MC ( L ) \alpha\in\operatorname{MC}(L) α ∈ MC ( L ) , one can twist the differential d d d by α \alpha α to obtain a new differential d α : = d + [ α , ⋅ ] L d^\alpha:=d+[\alpha,\cdot]_L d α := d + [ α , ⋅ ] L .
Proposition 2.3. ([13 Proposition 1.43, Definition 1.44] [21 Proposition 4.4] ) With the above notations, L α : = ( L , [ ⋅ , ⋅ ] L , d α ) L^\alpha:=(L,[\cdot,\cdot]_L,d^\alpha) L α := ( L , [ ⋅ , ⋅ ] L , d α ) is again a d g L a dgLa d g L a , called the d g L a dgLa d g L a twisted by α \alpha α .
There is an equivalence relation on MC ( L ) \operatorname{MC}(L) MC ( L ) given by the action of exp ( L 0 ) \exp(L^0) exp ( L 0 ) and known as the gauge equivalence . More precisely, two Maurer–Cartan elements x x x and y y y of ( L , [ ⋅ , ⋅ ] L , d ) (L,[\cdot,\cdot]_L,d) ( L , [ ⋅ , ⋅ ] L , d ) are called equivalent , if there exists an element τ ∈ L 0 \tau\in L^0 τ ∈ L 0 with ad τ \operatorname{ad}\tau ad τ nilpotent, such that
y = e τ ⋆ x : = e ad τ x + Id − e ad τ ad τ ( d τ ) . y=e^\tau\star x:=e^{\operatorname{ad}\tau}x+\frac{\operatorname{Id}-e^{\operatorname{ad}\tau}}{\operatorname{ad}\tau}(d\tau). y = e τ ⋆ x := e ad τ x + ad τ Id − e ad τ ( d τ ) .
The element τ \tau τ is called a gauge transformation . Denote by GT ( x , y ) \operatorname{GT}(x,y) GT ( x , y ) the set of gauge transformations between x x x and y y y .
Definition 2.4. ([16 Definition 1.7] ) Let ( L , [ ⋅ , ⋅ ] L , d ) (L,[\cdot,\cdot]_L,d) ( L , [ ⋅ , ⋅ ] L , d ) be a nilpotent d g L a dgLa d g L a . The Deligne groupoid Del ( L ) \operatorname{Del}(L) Del ( L ) of L L L is the action groupoid associated to the action of exp ( L 0 ) \exp(L^0) exp ( L 0 ) on MC ( L ) \operatorname{MC}(L) MC ( L ) with
the set of objects: elements of MC ( L ) \operatorname{MC}(L) MC ( L ) .
∀ x , y ∈ MC ( L ) \forall x,y\in\operatorname{MC}(L) ∀ x , y ∈ MC ( L ) , the set of morphisms between x x x and y y y :
Hom Del ( L ) ( x , y ) = { GT ( x , y ) if x and y are gauge-equivalent , ∅ otherwise. \operatorname{Hom}_{\operatorname{Del}(L)}(x,y)=\begin{cases}\operatorname{GT}(x,y)&\text{if }x\text{ and }y\text{ are gauge-equivalent},\\ \varnothing&\text{otherwise.}\end{cases} Hom Del ( L ) ( x , y ) = { GT ( x , y ) ∅ if x and y are gauge-equivalent , otherwise. Its set of connected components π 0 ( Del ( L ) ) \pi_0(\operatorname{Del}(L)) π 0 ( Del ( L )) is defined by
π 0 ( Del ( L ) ) : = MC ( L ) / exp ( L 0 ) . \pi_0(\operatorname{Del}(L)):=\operatorname{MC}(L)/\exp(L^0). π 0 ( Del ( L )) := MC ( L ) / exp ( L 0 ) .
We remark that Deligne groupoids enjoy the following nice functorial properties.
Proposition 2.5. Let ( L , [ ⋅ , ⋅ ] L , d ) (L,[\cdot,\cdot]_L,d) ( L , [ ⋅ , ⋅ ] L , d ) and ( L ′ , [ ⋅ , ⋅ ] L ′ , d ′ ) (L',[\cdot,\cdot]_{L'},d') ( L ′ , [ ⋅ , ⋅ ] L ′ , d ′ ) be nilpotent d g L a dgLa d g L a s. Let f : L → L ′ f:L\to L' f : L → L ′ be a morphism of d g L a dgLa d g L a s. Then f f f induces a morphism f ∗ : Del ( L ) → Del ( L ′ ) f_*:\operatorname{Del}(L)\to\operatorname{Del}(L') f ∗ : Del ( L ) → Del ( L ′ ) of the corresponding Deligne groupoids, such that f ∗ f_* f ∗ as a functor can be described as
on objects: for x ∈ MC ( L ) x\in\operatorname{MC}(L) x ∈ MC ( L ) , f ∗ ( x ) : = f ( x ) f_*(x):=f(x) f ∗ ( x ) := f ( x ) ,
on morphisms: for φ ∈ Hom Del ( L ) ( x , y ) \varphi\in\operatorname{Hom}_{\operatorname{Del}(L)}(x,y) φ ∈ Hom Del ( L ) ( x , y ) with φ = e α \varphi=e^\alpha φ = e α , α ∈ L 0 \alpha\in L^0 α ∈ L 0 , i.e., y = e α ⋆ x y=e^\alpha\star x y = e α ⋆ x , then f ∗ ( φ ) : = e f ( α ) ∈ Hom Del ( L ′ ) ( f ( x ) , f ( y ) ) f_*(\varphi):=e^{f(\alpha)}\in\operatorname{Hom}_{\operatorname{Del}(L')}(f(x),f(y)) f ∗ ( φ ) := e f ( α ) ∈ Hom Del ( L ′ ) ( f ( x ) , f ( y )) .
Proof. For x ∈ MC ( L ) x\in\operatorname{MC}(L) x ∈ MC ( L ) , we have
d ′ ( f ( x ) ) + 1 2 [ f ( x ) , f ( x ) ] L ′ = f ( d x + 1 2 [ x , x ] L ) = 0. d'(f(x))+\frac{1}{2}[f(x),f(x)]_{L'}=f\left(dx+\frac{1}{2}[x,x]_L\right)=0. d ′ ( f ( x )) + 2 1 [ f ( x ) , f ( x ) ] L ′ = f ( d x + 2 1 [ x , x ] L ) = 0.
which implies f ( MC ( L ) ) ⊂ MC ( L ′ ) f(\operatorname{MC}(L)) \subset\operatorname{MC}(L') f ( MC ( L )) ⊂ MC ( L ′ ) . If there exists a morphism in Del ( L ) \operatorname{Del}(L) Del ( L ) from x x x to y y y , i.e., there exists α ∈ L 0 \alpha\in L^{0} α ∈ L 0 such that y = e α ⋆ x = e ad α x + Id − e ad α ad α ( d α ) y=e^{\alpha}\star x=e^{\operatorname{ad}\alpha}x+\frac{\operatorname{Id}-e^{\operatorname{ad}\alpha}}{\operatorname{ad}\alpha}(d\alpha) y = e α ⋆ x = e ad α x + ad α Id − e ad α ( d α ) , then we have
f ( y ) = f ( e ad α x + Id − e ad α ad α ( d α ) ) = ∑ n ≥ 0 1 n ! ad n f ( α ) ( f ( x ) ) − ∑ n ≥ 0 1 ( n + 1 ) ! ad n f ( α ) ( d ′ f ( α ) ) = e ad f ( α ) f ( x ) + Id − e ad f ( α ) ad f ( α ) ( d ′ f ( α ) ) , \begin{aligned}
f(y)&=f\left(e^{\operatorname{ad}\alpha}x+\frac{\operatorname{Id}-e^{\operatorname{ad}\alpha}}{\operatorname{ad}\alpha}(d\alpha)\right)\\
&=\sum_{n\geq0}\frac{1}{n!}\operatorname{ad}^{n}f(\alpha)(f(x))-\sum_{n\geq0}\frac{1}{(n+1)!}\operatorname{ad}^{n}f(\alpha)(d'f(\alpha))\\
&=e^{\operatorname{ad}f(\alpha)}f(x)+\frac{\operatorname{Id}-e^{\operatorname{ad}f(\alpha)}}{\operatorname{ad}f(\alpha)}(d'f(\alpha)),
\end{aligned} f ( y ) = f ( e ad α x + ad α Id − e ad α ( d α ) ) = n ≥ 0 ∑ n ! 1 ad n f ( α ) ( f ( x )) − n ≥ 0 ∑ ( n + 1 )! 1 ad n f ( α ) ( d ′ f ( α )) = e ad f ( α ) f ( x ) + ad f ( α ) Id − e ad f ( α ) ( d ′ f ( α )) ,
where the summations are finite because L L L and L ′ L' L ′ are nilpotent. Hence, there is a morphism in Del ( L ′ ) \operatorname{Del}(L') Del ( L ′ ) from f ( x ) f(x) f ( x ) to f ( y ) f(y) f ( y ) . Notice that f f f preserves the Baker–Campbell–Hausdorff product, hence the composition of arrows.
In the following example, we will review how to characterize Lie algebra structures on a fixed vector space V V V using a dgLa.
Example 2.6. Let V V V be a vector space. Define a graded vector space L V = ⨁ n ∈ Z L V n L_V=\bigoplus_{n\in\mathbb{Z}}L_V^n L V = ⨁ n ∈ Z L V n as
L V n = { 0 , n ≤ 0 ; Hom ( ⋀ n + 1 V , V ) , n ≥ 1. L_V^n=
\begin{cases}
0, & n\leq0;\\
\operatorname{Hom}(\bigwedge^{n+1}V,V), & n\geq1.
\end{cases} L V n = { 0 , Hom ( ⋀ n + 1 V , V ) , n ≤ 0 ; n ≥ 1. Then, ( L V , [ ⋅ , ⋅ ] N R ) (L_V,[\cdot,\cdot]_{\mathrm{NR}}) ( L V , [ ⋅ , ⋅ ] NR ) is a gLa [38 p. 50] , where the Lie bracket is given by the Nijenhuis-Richardson bracket [ ⋅ , ⋅ ] N R [\cdot,\cdot]_{\mathrm{NR}} [ ⋅ , ⋅ ] NR
[ α , β ] N R : = α ∘ N R β − ( − 1 ) m n β ∘ N R α , ∀ α ∈ L V m = Hom ( ⋀ m + 1 V , V ) , β ∈ L V n = Hom ( ⋀ n + 1 V , V ) [\alpha,\beta]_{\mathrm{NR}}:=\alpha\circ_{\mathrm{NR}}\beta-(-1)^{mn}\beta\circ_{\mathrm{NR}}\alpha,\qquad\forall\alpha\in L_V^m=\operatorname{Hom}(\bigwedge^{m+1}V,V),\ \beta\in L_V^n=\operatorname{Hom}(\bigwedge^{n+1}V,V) [ α , β ] NR := α ∘ NR β − ( − 1 ) mn β ∘ NR α , ∀ α ∈ L V m = Hom ( ⋀ m + 1 V , V ) , β ∈ L V n = Hom ( ⋀ n + 1 V , V ) with α ∘ N R β ∈ L V m + n = Hom ( ⋀ m + n + 1 V , V ) \alpha\circ_{\mathrm{NR}}\beta\in L_V^{m+n}=\operatorname{Hom}(\bigwedge^{m+n+1}V,V) α ∘ NR β ∈ L V m + n = Hom ( ⋀ m + n + 1 V , V ) defined by
( α ∘ N R β ) ( x 1 , … , x m + n + 1 ) : = ∑ σ ∈ S ( n + 1 , m ) ( − 1 ) σ α ( β ( x σ ( 1 ) , … , x σ ( n + 1 ) ) , x σ ( n + 2 ) , … , x σ ( m + n + 1 ) ) , (\alpha\circ_{\mathrm{NR}}\beta)(x_1,\ldots,x_{m+n+1}):=\sum_{\sigma\in S(n+1,m)}(-1)^{\sigma}\alpha(\beta(x_{\sigma(1)},\ldots,x_{\sigma(n+1)}),x_{\sigma(n+2)},\ldots,x_{\sigma(m+n+1)}), ( α ∘ NR β ) ( x 1 , … , x m + n + 1 ) := σ ∈ S ( n + 1 , m ) ∑ ( − 1 ) σ α ( β ( x σ ( 1 ) , … , x σ ( n + 1 ) ) , x σ ( n + 2 ) , … , x σ ( m + n + 1 ) ) , for all x i ∈ V x_i\in V x i ∈ V . Here, S ( n + 1 , m ) S(n+1,m) S ( n + 1 , m ) is the set of ( n + 1 , m ) (n+1,m) ( n + 1 , m ) -shuffles.
Indeed, in this case the Maurer–Cartan equation is the Jacobi identity, thus the set MC ( L V ) \operatorname{MC}(L_V) MC ( L V ) is the set of all Lie algebra structures on the vector space V V V .
By the fact that L V 0 = 0 L_V^0=0 L V 0 = 0 , the Deligne groupoid of L V L_V L V is discrete. Hence, the set of connected components of the Deligne groupoid is its set of objects, i.e.,
π 0 ( Del ( L V ) ) ≅ Obj ( Del ( L V ) ) = MC ( L V ) . \pi_0(\operatorname{Del}(L_V))\cong\operatorname{Obj}(\operatorname{Del}(L_V))=\operatorname{MC}(L_V). π 0 ( Del ( L V )) ≅ Obj ( Del ( L V )) = MC ( L V ) .
As for general groupoids, Deligne groupoids also carry nice geometric features which appear naturally in the global extension problem. To see this, we will first recall the notion of fibration for groupoids. More precisely, let G \mathcal{G} G be a groupoid. The set of objects and the collection of morphisms of G \mathcal{G} G are denoted as Obj ( G ) \operatorname{Obj}(\mathcal{G}) Obj ( G ) and Mor ( G ) \operatorname{Mor}(\mathcal{G}) Mor ( G ) respectively. For x , y ∈ Obj ( G ) x,y\in\operatorname{Obj}(\mathcal{G}) x , y ∈ Obj ( G ) , denote by Hom G ( x , y ) \operatorname{Hom}_{\mathcal{G}}(x,y) Hom G ( x , y ) the set of morphisms between x x x and y y y .
Definition 2.7. ([11 p. 105] ) For x ∈ Obj ( G ) x\in\operatorname{Obj}(\mathcal{G}) x ∈ Obj ( G ) , the star at x x x , denoted by St G ( x ) \operatorname{St}_{\mathcal{G}}(x) St G ( x ) , is defined as the union of the morphism sets Hom G ( x , y ) \operatorname{Hom}_{\mathcal{G}}(x,y) Hom G ( x , y ) for all y ∈ Obj ( G ) y\in\operatorname{Obj}(\mathcal{G}) y ∈ Obj ( G ) , that is,
St G ( x ) = ⋃ y ∈ Obj ( G ) Hom G ( x , y ) . \operatorname{St}_{\mathcal{G}}(x)=\bigcup_{y\in\operatorname{Obj}(\mathcal{G})}\operatorname{Hom}_{\mathcal{G}}(x,y). St G ( x ) = y ∈ Obj ( G ) ⋃ Hom G ( x , y ) . Equivalently, the star at x x x consists of all morphisms in G \mathcal{G} G whose source is x x x .
If F : G → H F:\mathcal{G}\to\mathcal{H} F : G → H is a morphism of groupoids, then F F F induces maps on the stars of G \mathcal{G} G in a natural way by
F : St G ( x ) → St H ( F ( x ) ) , φ ↦ F ( φ ) , ∀ φ ∈ St G ( x ) . F:\operatorname{St}_{\mathcal{G}}(x)\to\operatorname{St}_{\mathcal{H}}(F(x)),\qquad\varphi\mapsto F(\varphi),\qquad\forall\varphi\in\operatorname{St}_{\mathcal{G}}(x). F : St G ( x ) → St H ( F ( x )) , φ ↦ F ( φ ) , ∀ φ ∈ St G ( x ) .
Definition 2.8. ([11 p. 109] )
A morphism F : G → H F:\mathcal{G}\to\mathcal{H} F : G → H of groupoids is called a fibration if for all x ∈ Obj ( G ) x\in\operatorname{Obj}(\mathcal{G}) x ∈ Obj ( G ) the induced map on St G ( x ) \operatorname{St}_{\mathcal{G}}(x) St G ( x ) is surjective.
Let F : G → H F:\mathcal{G}\to\mathcal{H} F : G → H be a fibration of groupoids, and a ∈ Im ( F ∣ Obj ( G ) ) a\in\operatorname{Im}(F|_{\operatorname{Obj}(\mathcal{G})}) a ∈ Im ( F ∣ Obj ( G ) ) . Then the fiber F − 1 ( a ) F^{-1}(a) F − 1 ( a ) of F F F at object a a a is defined as the subcategory of G \mathcal{G} G , whose set of objects is
Obj ( F − 1 ( a ) ) = { x ∈ Obj ( G ) : F ( x ) = a } , \operatorname{Obj}(F^{-1}(a))=\{x\in\operatorname{Obj}(\mathcal{G}):F(x)=a\}, Obj ( F − 1 ( a )) = { x ∈ Obj ( G ) : F ( x ) = a } , and the collection of morphisms is Mor ( F − 1 ( a ) ) = { φ ∈ Mor ( G ) : F ( φ ) = Id a } \operatorname{Mor}(F^{-1}(a))=\{\varphi\in\operatorname{Mor}(\mathcal{G}):F(\varphi)=\operatorname{Id}_{a}\} Mor ( F − 1 ( a )) = { φ ∈ Mor ( G ) : F ( φ ) = Id a } .
Deformations in terms of Maurer–Cartan elements of dgLas
Throughout this subsection, k k k is a field of characteristic zero. Let Art k \operatorname{Art}_{k} Art k denote the category of local Artinian commutative k k k -algebras with residue field k k k . For A ∈ Art k A\in\operatorname{Art}_{k} A ∈ Art k , write m A = ker ( A → k ) \mathfrak{m}_{A}=\ker(A\to k) m A = ker ( A → k ) for the unique maximal ideal of $A.
Definition 2.9. ([24 §1.1][22, §2] ) The groupoid-valued deformation functor associated with a dgLa L L L is
Def L : Art k ⟶ Gpd , Def L ( A ) = Del ( L ⊗ m A ) . \operatorname{Def}_{L}:\operatorname{Art}_{k}\longrightarrow\operatorname{Gpd},\qquad\operatorname{Def}_{L}(A)=\operatorname{Del}(L\otimes\mathfrak{m}_{A}). Def L : Art k ⟶ Gpd , Def L ( A ) = Del ( L ⊗ m A ) . A morphism A → B A\to B A → B induces the corresponding functor by extension of scalars.
Since m A \mathfrak{m}_{A} m A is nilpotent, the dgLa L ⊗ m A L\otimes\mathfrak{m}_{A} L ⊗ m A is nilpotent, so its gauge action is well defined even when L L L itself is not nilpotent.
Let G G G be a smooth affine algebraic group acting on a k k k -scheme X X X .
Definition 2.10. ([42 Definition 78.20.1] ) The quotient stack [ X / G ] [X/G] [ X / G ] is the fppf stack associated with the action groupoid G × X ⇉ X G\times X\rightrightarrows X G × X ⇉ X . For a k k k -scheme S S S , its objects are pairs ( P , f ) (P,f) ( P , f ) consisting of a right G G G -torsor P → S P\to S P → S and a morphism f : P → X f:P\to X f : P → X satisfying
f ( p g ) = g − 1 ⋅ f ( p ) . f(pg)=g^{-1}\cdot f(p). f ( p g ) = g − 1 ⋅ f ( p ) . An arrow ( P , f ) → ( P ′ , f ′ ) (P,f)\to(P',f') ( P , f ) → ( P ′ , f ′ ) is a G G G -equivariant isomorphism φ : P → P ′ \varphi:P\to P' φ : P → P ′ over S S S such that f ′ ∘ φ = f f'\circ\varphi=f f ′ ∘ φ = f . Pullback defines the functoriality in S S S .
To define deformations at a fixed object, we use the homotopy pullback of groupoids. For functors X → F Z ← H Y \mathcal{X}\xrightarrow{F}\mathcal{Z}\xleftarrow{H}\mathcal{Y} X F Z H Y , the groupoid X × Z h Y \mathcal{X}\mathbin{\times^{h}_{\mathcal{Z}}}\mathcal{Y} X × Z h Y has objects ( x , y , η ) (x,y,\eta) ( x , y , η ) , where η : F ( x ) → H ( y ) \eta:F(x)\to H(y) η : F ( x ) → H ( y ) is an isomorphism, and arrows ( f , g ) : ( x , y , η ) → ( x ′ , y ′ , η ′ ) (f,g):(x,y,\eta)\to(x',y',\eta') ( f , g ) : ( x , y , η ) → ( x ′ , y ′ , η ′ ) satisfying
H ( g ) ∘ η = η ′ ∘ F ( f ) . H(g)\circ\eta=\eta'\circ F(f). H ( g ) ∘ η = η ′ ∘ F ( f ) .
See [12 §10.7, Exercise 1] for more detail.
Definition 2.12. ((30, Remark 1.1.18)) Let X \mathcal{X} X be a stack over k k k and x 0 ∈ X ( k ) x_0 \in\mathcal{X}(k) x 0 ∈ X ( k ) . Its formal deformation functor at x 0 x_0 x 0 is
Def X , x 0 ( A ) = X ( A ) × X ( k ) { x 0 } , A ∈ Art k , \operatorname{Def}_{\mathcal{X},x_0}(A)=\mathcal{X}(A)\mathbin{\times}_{\mathcal{X}(k)}\{x_0\},\qquad A\in\operatorname{Art}_k, Def X , x 0 ( A ) = X ( A ) × X ( k ) { x 0 } , A ∈ Art k , where { x 0 } \{x_0\} { x 0 } denotes the terminal groupoid mapping to X ( k ) \mathcal{X}(k) X ( k ) through x 0 x_0 x 0 .
Thus, an object is a pair ( ξ , η ) (\xi,\eta) ( ξ , η ) with ξ ∈ X ( A ) \xi\in\mathcal{X}(A) ξ ∈ X ( A ) and an isomorphism η : ξ ∣ k → x 0 \eta:\xi|_k\to x_0 η : ξ ∣ k → x 0 . Arrows are isomorphisms compatible with these identifications.
Proposition 2.13. Let x ∈ X ( k ) x\in X(k) x ∈ X ( k ) , and let [ x ] ∈ [ X / G ] ( k ) [x]\in[X/G](k) [ x ] ∈ [ X / G ] ( k ) be the corresponding object. Set
X x ( A ) = { x ~ ∈ X ( A ) : x ~ ∣ k = x } , G 1 ( A ) = ker ( G ( A ) → G ( k ) ) . X_x(A)=\{\widetilde{x}\in X(A):\widetilde{x}|_k=x\},\qquad G_1(A)=\ker(G(A)\to G(k)). X x ( A ) = { x ∈ X ( A ) : x ∣ k = x } , G 1 ( A ) = ker ( G ( A ) → G ( k )) . There is a natural equivalence
Def [ X / G ] , [ x ] ( A ) ≃ X x ( A ) / / G 1 ( A ) . \operatorname{Def}_{[X/G],[x]}(A)\simeq X_x(A)//G_1(A). Def [ X / G ] , [ x ] ( A ) ≃ X x ( A ) // G 1 ( A ) . Proof. The canonical functor from the right-hand side sends x ~ \widetilde{x} x to the corresponding trivial torsor with its standard identification over k k k .
Conversely, a deformation of [ x ] [x] [ x ] consists of a torsor P → Spec A P\to\operatorname{Spec}A P → Spec A , an equivariant map P → X P\to X P → X , and an identification of its special fiber with the object represented by x x x . This identification determines a point p 0 ∈ P ( k ) p_0\in P(k) p 0 ∈ P ( k ) . Since P → Spec A P\to\operatorname{Spec}A P → Spec A is smooth, p 0 p_0 p 0 lifts to a point of P ( A ) P(A) P ( A ) , giving a trivialization compatible with the prescribed special fiber. The resulting map corresponds to an element of X x ( A ) X_x(A) X x ( A ) . Arrows between two such objects are exactly the elements of G 1 ( A ) G_1(A) G 1 ( A ) carrying one lift to the other. Thus, the canonical functor is fully faithful and essentially surjective. Moreover, it is natural in A A A . □
Let L L L be a finite-dimensional nilpotent dgLa. Its Maurer–Cartan scheme is defined by
MC ( L ) ( A ) = { γ ∈ L 1 ⊗ A : d γ + 1 2 [ γ , γ ] = 0 } , \operatorname{MC}(L)(A)=\left\{\gamma\in L^1\otimes A:d\gamma+\frac{1}{2}[\gamma,\gamma]=0\right\}, MC ( L ) ( A ) = { γ ∈ L 1 ⊗ A : d γ + 2 1 [ γ , γ ] = 0 } ,
for every commutative k k k -algebra A A A . The unipotent algebraic group exp ( L 0 ) \exp(L^0) exp ( L 0 ) acts by gauge transformations, defining the quotient stack
M C ( L ) = [ MC ( L ) / exp ( L 0 ) ] . \mathfrak{MC}(L)=[\operatorname{MC}(L)/\exp(L^0)]. MC ( L ) = [ MC ( L ) / exp ( L 0 )] .
For α ∈ MC ( L ) ( k ) \alpha\in\operatorname{MC}(L)(k) α ∈ MC ( L ) ( k ) , let
L α = ( L , d α , [ ⋅ , ⋅ ] ) , d α = d + [ α , − ] , L^\alpha=(L,d^\alpha,[\cdot,\cdot]),\qquad d^\alpha=d+[\alpha,-], L α = ( L , d α , [ ⋅ , ⋅ ]) , d α = d + [ α , − ] ,
be the corresponding twisted dgLa.
Theorem 2.14. For every α ∈ MC ( L ) ( k ) \alpha\in\operatorname{MC}(L)(k) α ∈ MC ( L ) ( k ) , there is a natural equivalence
Def M C ( L ) , [ α ] ≃ Def L α . \operatorname{Def}_{\mathfrak{MC}(L),[\alpha]}\simeq\operatorname{Def}_{L^\alpha}. Def MC ( L ) , [ α ] ≃ Def L α . Thus L α L^\alpha L α controls the formal deformations of M C ( L ) \mathfrak{MC}(L) MC ( L ) at [ α ] [\alpha] [ α ] .
Proof. Every lift of α \alpha α over A A A has the form α + η \alpha+\eta α + η , with η ∈ L 1 ⊗ m A \eta\in L^1\otimes\mathfrak{m}_A η ∈ L 1 ⊗ m A . Since α \alpha α satisfies the Maurer–Cartan equation, we have
d ( α + η ) + 1 2 [ α + η , α + η ] = d α η + 1 2 [ η , η ] . d(\alpha+\eta)+\frac{1}{2}[\alpha+\eta,\alpha+\eta]=d^\alpha\eta+\frac{1}{2}[\eta,\eta]. d ( α + η ) + 2 1 [ α + η , α + η ] = d α η + 2 1 [ η , η ] . Translation by α \alpha α therefore gives a bijection
MC ( L α ⊗ m A ) ⟶ MC ( L ) α ( A ) . \operatorname{MC}(L^\alpha\otimes\mathfrak{m}_A)\longrightarrow\operatorname{MC}(L)_\alpha(A). MC ( L α ⊗ m A ) ⟶ MC ( L ) α ( A ) .
Moreover, we have
ker ( exp ( L 0 ) ( A ) → exp ( L 0 ) ( k ) ) = exp ( L 0 ⊗ m A ) . \ker\left(\exp(L^0)(A)\to\exp(L^0)(k)\right)=\exp(L^0\otimes\mathfrak{m}_A). ker ( exp ( L 0 ) ( A ) → exp ( L 0 ) ( k ) ) = exp ( L 0 ⊗ m A ) .
Notice that for a ∈ L 0 ⊗ m A a\in L^0\otimes\mathfrak{m}_A a ∈ L 0 ⊗ m A ,
d α a = d a − [ a , α ] , e a ⋆ d ( α + η ) = α + e a ⋆ d α η , d^\alpha a=da-[a,\alpha],\qquad e^a\star_d(\alpha+\eta)=\alpha+e^a\star_{d^\alpha}\eta, d α a = d a − [ a , α ] , e a ⋆ d ( α + η ) = α + e a ⋆ d α η ,
which means that translation intertwines the gauge actions. Proposition 2.13 now yields
Def MC ( L ) , [ α ] ( A ) ≃ MC ( L ) α ( A ) / / exp ( L 0 ⊗ m A ) ≃ Del ( L α ⊗ m A ) = Def L α ( A ) . \begin{aligned}
\operatorname{Def}_{\operatorname{MC}(L),[\alpha]}(A)\simeq\operatorname{MC}(L)_\alpha(A)//\exp(L^0\otimes\mathfrak{m}_A) \\
&\simeq\operatorname{Del}(L^\alpha\otimes\mathfrak{m}_A)=\operatorname{Def}_{L^\alpha}(A).
\end{aligned} Def MC ( L ) , [ α ] ( A ) ≃ MC ( L ) α ( A ) // exp ( L 0 ⊗ m A ) ≃ Del ( L α ⊗ m A ) = Def L α ( A ) .
Furthermore, the above proof is functorial in A A A . □
Example 2.15. Let V V V be an n n n -dimensional vector space with basis { e 1 , … , e n } \{e_1,\ldots,e_n\} { e 1 , … , e n } , and let ν : ∧ 2 V → V \nu:\wedge^2V\to V ν : ∧ 2 V → V be a linear map. For 1 ≤ i < j ≤ n 1\leq i<j\leq n 1 ≤ i < j ≤ n , write
ν ( e i , e j ) = ∑ a = 1 n c i j a e a . \nu(e_i,e_j)=\sum_{a=1}^{n}c_{ij}^{a}e_a. ν ( e i , e j ) = a = 1 ∑ n c ij a e a . where { c i j a } \{c_{ij}^{a}\} { c ij a } denote the structure constants of ν \nu ν . The scheme of Lie brackets on V V V is
Lie ( V ) = Spec ( k [ c i j a ∣ 1 ≤ i < j ≤ n ] / ( J i j k l ) ) , \operatorname{Lie}(V)=\operatorname{Spec}\left(k[c_{ij}^{a}\mid1\leq i<j\leq n]/(J_{ijk}^{l})\right), Lie ( V ) = Spec ( k [ c ij a ∣ 1 ≤ i < j ≤ n ] / ( J ij k l ) ) , where, for 1 ≤ l ≤ n 1\leq l\leq n 1 ≤ l ≤ n , 1 ≤ i < j < k ≤ n 1\leq i<j<k\leq n 1 ≤ i < j < k ≤ n ,
J i j k l = ∑ a = 1 n ( c i j a c a k l + c j k a c a i l + c k i a c a j l ) . J_{ijk}^{l}=\sum_{a=1}^{n}\left(c_{ij}^{a}c_{ak}^{l}+c_{jk}^{a}c_{ai}^{l}+c_{ki}^{a}c_{aj}^{l}\right). J ij k l = a = 1 ∑ n ( c ij a c ak l + c j k a c ai l + c k i a c aj l ) . For every commutative k k k -algebra A A A ,
Lie ( V ) ( A ) = { A -linear Lie brackets on V ⊗ A } . \operatorname{Lie}(V)(A)=\{A\text{-linear Lie brackets on }V\otimes A\}. Lie ( V ) ( A ) = { A -linear Lie brackets on V ⊗ A } . The group GL ( V ) \operatorname{GL}(V) GL ( V ) acts by transport of structure, and
L i e n = [ Lie ( V ) / GL ( V ) ] \mathfrak{Lie}_n=[\operatorname{Lie}(V)/\operatorname{GL}(V)] Lie n = [ Lie ( V ) / GL ( V )] is the moduli stack of Lie algebra structures on vector bundles of rank n n n . For A ∈ A r t k A\in\mathbf{Art}_k A ∈ Art k , every such vector bundle is free, so L i e n ( A ) \mathfrak{Lie}_n(A) Lie n ( A ) is equivalent to the groupoid of Lie brackets on V ⊗ A V\otimes A V ⊗ A and A A A -linear Lie algebra isomorphisms.
Fix ν ∈ Lie ( V ) ( k ) \nu\in\operatorname{Lie}(V)(k) ν ∈ Lie ( V ) ( k ) . By Proposition 2.13, Def L i e n , [ ν ] ( A ) \operatorname{Def}_{\mathfrak{Lie}_n,[\nu]}(A) Def Lie n , [ ν ] ( A ) is equivalent to the groupoid whose objects are brackets ν A \nu_A ν A satisfying
ν A m o d m A = ν , \nu_A\bmod\mathfrak{m}_A=\nu, ν A mod m A = ν , and whose arrows ν A → ν A ′ \nu_A\to\nu'_A ν A → ν A ′ are elements
g ∈ GL ( V ⊗ A ) , g m o d m A = Id V , g ⋅ ν A = ν A ′ . g\in\operatorname{GL}(V\otimes A),\qquad g\bmod\mathfrak{m}_A=\operatorname{Id}_V,\qquad g\cdot\nu_A=\nu'_A. g ∈ GL ( V ⊗ A ) , g mod m A = Id V , g ⋅ ν A = ν A ′ . Let L N R L_{\mathrm{NR}} L NR be the graded Lie algebra
L N R , V q = Hom k ( ∧ q + 1 V , V ) , q ≥ 0 , L_{\mathrm{NR},V}^{q}=\operatorname{Hom}_k(\wedge^{q+1}V,V),\qquad q\geq0, L NR , V q = Hom k ( ∧ q + 1 V , V ) , q ≥ 0 , equipped with the Nijenhuis–Richardson bracket in Example 2.6 and zero differential. The Jacobi identity is equivalent to [ ν , ν ] N R = 0 [\nu,\nu]_{\mathrm{NR}}=0 [ ν , ν ] NR = 0 , so ν \nu ν is a Maurer–Cartan element. Its twist is
L N R , V ν = ( L N R , V , [ ν , − ] N R , [ − , − ] N R ) . L_{\mathrm{NR},V}^{\nu}=\left(L_{\mathrm{NR},V},[\nu,-]_{\mathrm{NR}},[-,-]_{\mathrm{NR}}\right). L NR , V ν = ( L NR , V , [ ν , − ] NR , [ − , − ] NR ) .
Writing ν A = ν + η \nu_A = \nu+ \eta ν A = ν + η identifies deformed brackets with MC ( L N R , V ⊗ m A ) \operatorname{MC}(L_{\mathrm{NR},V} \otimes\mathfrak{m}_A) MC ( L NR , V ⊗ m A ) . Furthermore, exponential and logarithm identify
ker ( GL ( V ⊗ A ) → GL ( V ) ) = exp ( End ( V ) ⊗ m A ) , \operatorname{ker}(\operatorname{GL}(V \otimes A) \to\operatorname{GL}(V)) = \exp(\operatorname{End}(V) \otimes\mathfrak{m}_A), ker ( GL ( V ⊗ A ) → GL ( V )) = exp ( End ( V ) ⊗ m A ) ,
and transport of structure agrees with the gauge action. Consequently,
D e f L i e n , [ ν ] ≃ D e f L N R , V . \mathop{\mathrm{Def}}\nolimits_{\mathrm{Lie}_n,[\nu]} \simeq\mathop{\mathrm{Def}}\nolimits_{L_{\mathrm{NR},V}}. Def Lie n , [ ν ] ≃ Def L NR , V .
Thus the twisted Nijenhuis-Richardson dgLa controls the formal deformations of the Lie algebra ( V , ν ) (V,\nu) ( V , ν ) .
Global extensions and non-abelian extensions of Lie algebras
In this section, we recall the classification of global extensions of Lie algebras by global cohomology. Moreover, we will describe the classification result from the point of view of action groupoids.
Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) be a Lie algebra, let V ′ ′ V^{\prime\prime} V ′′ be a vector space, and let p : V ′ ′ → V p:V^{\prime\prime}\to V p : V ′′ → V be a surjective linear map. Denote by V ′ = ker p V^{\prime}=\operatorname{ker}p V ′ = ker p .
Definition 3.1. ([5 pp. 3–4] ) A global extension structure on V ′ ′ V^{\prime\prime} V ′′ is a Lie bracket ν ′ ′ \nu^{\prime\prime} ν ′′ for which
p : ( V ′ ′ , ν ′ ′ ) ⟶ ( V , μ ) p:(V^{\prime\prime},\nu^{\prime\prime})\longrightarrow(V,\mu) p : ( V ′′ , ν ′′ ) ⟶ ( V , μ ) is a Lie algebra homomorphism. The set of global extension structures on V ′ ′ V^{\prime\prime} V ′′ with respect to p p p and g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) is denoted by Gext ( g , V ′ ) \operatorname{Gext}(\mathfrak{g},V^{\prime}) Gext ( g , V ′ ) .
Two such structures ν 1 ′ ′ \nu^{\prime\prime}_1 ν 1 ′′ and ν 2 ′ ′ \nu^{\prime\prime}_2 ν 2 ′′ are equivalent (framed equivalent), if there is a Lie algebra isomorphism θ : ( V ′ ′ , ν 1 ′ ′ ) → ( V ′ ′ , ν 2 ′ ′ ) \theta:(V^{\prime\prime},\nu^{\prime\prime}_1)\to(V^{\prime\prime},\nu^{\prime\prime}_2) θ : ( V ′′ , ν 1 ′′ ) → ( V ′′ , ν 2 ′′ ) such that
θ ∣ V ′ = id V ′ , p ∘ θ = p . \left.\theta\right|_{V^{\prime}}=\operatorname{id}_{V^{\prime}},\qquad p\circ\theta=p. θ ∣ V ′ = id V ′ , p ∘ θ = p . The set of equivalence classes is denoted by GExt ( g , V ′ ) \operatorname{GExt}(\mathfrak{g},V^{\prime}) GExt ( g , V ′ ) .
However, in various interactions with deformation theory, the first condition is so strong, which will lead to additional rigidity results. To avoid this, we introduce a weaker notion of an equivalence relation, which we call the unframed equivalence .
Definition 3.2. Two such structures ν 1 ′ ′ \nu^{\prime\prime}_1 ν 1 ′′ and ν 2 ′ ′ \nu^{\prime\prime}_2 ν 2 ′′ are unframed equivalent if only the condition p ∘ θ = p p\circ\theta=p p ∘ θ = p is imposed for the Lie algebra isomorphism θ \theta θ . The set of unframed equivalence classes is denoted by GExt u n ( g , V ′ ) \operatorname{GExt}^{\mathrm{un}}(\mathfrak{g},V^{\prime}) GExt un ( g , V ′ ) .
Remark 3.3. A direct consequence of p ∘ θ = p p\circ\theta=p p ∘ θ = p is that θ ( V ′ ) = V ′ \theta(V^{\prime})=V^{\prime} θ ( V ′ ) = V ′ and θ ∣ V ′ \left.\theta\right|_{V^{\prime}} θ ∣ V ′ is automatically a Lie algebra isomorphism from ( V ′ , ν 1 ′ ′ ∣ V ′ ) (V^{\prime},\left.\nu^{\prime\prime}_1\right|_{V^{\prime}}) ( V ′ , ν 1 ′′ ∣ V ′ ) to the one induced by ( V ′ , ν 2 ′ ′ ∣ V ′ ) (V^{\prime},\left.\nu^{\prime\prime}_2\right|_{V^{\prime}}) ( V ′ , ν 2 ′′ ∣ V ′ ) .
Throughout, a triple ( V ′ ′ , p , g ) (V^{\prime\prime},p,\mathfrak{g}) ( V ′′ , p , g ) , where g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) is a Lie algebra, V ′ ′ V^{\prime\prime} V ′′ is a vector space, and p : V ′ ′ → V p:V^{\prime\prime}\to V p : V ′′ → V is a surjective linear map, will be called a global extension datum . We write V ′ = ker ( p ) V^{\prime}=\operatorname{ker}(p) V ′ = ker ( p ) .
Let m = dim V ′ ′ m=\dim V^{\prime\prime} m = dim V ′′ and n = dim g n=\dim\mathfrak{g} n = dim g , and choose bases { e 1 , … , e m } \{e_1,\ldots,e_m\} { e 1 , … , e m } of V ′ ′ V^{\prime\prime} V ′′ and { x 1 , … , x n } \{x_1,\ldots,x_n\} { x 1 , … , x n } of g \mathfrak{g} g . Then Gext ( g , V ′ ) \operatorname{Gext}(\mathfrak{g},V^{\prime}) Gext ( g , V ′ ) is an affine scheme with coordinate ring
O Gext ( g , V ′ ) = k [ t 1 , … , t N ] / ⟨ J , G ⟩ , N = m 2 ( m − 1 ) 2 , \mathcal{O}_{\operatorname{Gext}(\mathfrak{g},V^{\prime})}=k[t_1,\ldots,t_N]/\langle J,G\rangle,\qquad N=\frac{m^2(m-1)}{2}, O Gext ( g , V ′ ) = k [ t 1 , … , t N ] / ⟨ J , G ⟩ , N = 2 m 2 ( m − 1 ) ,
where J J J and G G G encode the Jacobi identity and the condition that p p p is a Lie algebra homomorphism, respectively.
Consider the algebraic group
G p u n = { θ ∈ GL ( V ′ ′ ) ∣ p ∘ θ = p } , G_p^{\mathrm{un}}=\{\theta\in\operatorname{GL}(V^{\prime\prime})\mid p\circ\theta=p\}, G p un = { θ ∈ GL ( V ′′ ) ∣ p ∘ θ = p } ,
and its closed subgroup
G p = { θ ∈ GL ( V ′ ′ ) ∣ θ ∣ V ′ = Id V ′ , p ∘ θ = p } , G_p=\{\theta\in\operatorname{GL}(V^{\prime\prime})\mid\theta|_{V^{\prime}}=\operatorname{Id}_{V^{\prime}},\ p\circ\theta=p\}, G p = { θ ∈ GL ( V ′′ ) ∣ θ ∣ V ′ = Id V ′ , p ∘ θ = p } ,
which is the pointwise stabilizer of V ′ V^{\prime} V ′ in G p u n G_p^{\mathrm{un}} G p un . To describe these groups more explicitly, we fix a linear section s : V → V ′ ′ s:V\to V^{\prime\prime} s : V → V ′′ of p p p .
Proposition 3.4. The group G p u n G_p^{\mathrm{un}} G p un is a closed algebraic subgroup of GL ( V ′ ′ ) \operatorname{GL}(V^{\prime\prime}) GL ( V ′′ ) . Moreover, the map
Φ u n : Hom ( V , V ′ ) ⋊ GL ( V ′ ) ⟶ G p u n , ( r , B ) ⟼ Φ ( r , B ) u n \Phi^{\mathrm{un}}:\operatorname{Hom}(V,V^{\prime})\rtimes\operatorname{GL}(V^{\prime})\longrightarrow G_p^{\mathrm{un}},\qquad(r,B)\longmapsto\Phi_{(r,B)}^{\mathrm{un}} Φ un : Hom ( V , V ′ ) ⋊ GL ( V ′ ) ⟶ G p un , ( r , B ) ⟼ Φ ( r , B ) un defined by
Φ ( r , B ) u n ( u + s ( x ) ) = B ( u ) + s ( x ) + r ( x ) , x ∈ V , u ∈ V ′ (1) \Phi_{(r,B)}^{\mathrm{un}}(u+s(x))=B(u)+s(x)+r(x),\qquad x\in V,\quad u\in V^{\prime} \tag*{(1)} Φ ( r , B ) un ( u + s ( x )) = B ( u ) + s ( x ) + r ( x ) , x ∈ V , u ∈ V ′ (1) is an isomorphism of algebraic groups. Here Hom ( V , V ′ ) \operatorname{Hom}(V,V^{\prime}) Hom ( V , V ′ ) is viewed as an additive vector group and the multiplication of Hom ( V , V ′ ) ⋊ GL ( V ′ ) \operatorname{Hom}(V,V^{\prime})\rtimes\operatorname{GL}(V^{\prime}) Hom ( V , V ′ ) ⋊ GL ( V ′ ) is given by
( r , B ) ⋅ ( r ′ , B ′ ) = ( r + B r ′ , B B ′ ) . (r,B)\cdot(r^{\prime},B^{\prime})=(r+Br^{\prime},BB^{\prime}). ( r , B ) ⋅ ( r ′ , B ′ ) = ( r + B r ′ , B B ′ ) .
Proof. Under the identification V ′ ′ ≅ V ′ ⊕ V V^{\prime\prime}\cong V^{\prime}\oplus V V ′′ ≅ V ′ ⊕ V induced by s s s , we have the following realization
G p u n = { ( B r 0 I V ) ∣ r ∈ Hom ( V , V ′ ) } . G_p^{\mathrm{un}}=\left\{\left.
\begin{pmatrix}
B&r\\
0&I_V
\end{pmatrix}
\right|r\in\operatorname{Hom}(V,V^{\prime})\right\}. G p un = { ( B 0 r I V ) r ∈ Hom ( V , V ′ ) } .
Thus, G p u n G_p^{\mathrm{un}} G p un is a closed subgroup of GL ( V ′ ′ ) \operatorname{GL}(V^{\prime\prime}) GL ( V ′′ ) . Moreover, Φ ( r , B ) u n Φ ( r ′ , B ′ ) u n = Φ ( r + B r ′ , B B ′ ) u n \Phi_{(r,B)}^{\mathrm{un}}\Phi_{(r^{\prime},B^{\prime})}^{\mathrm{un}}=\Phi_{(r+Br^{\prime},BB^{\prime})}^{\mathrm{un}} Φ ( r , B ) un Φ ( r ′ , B ′ ) un = Φ ( r + B r ′ , B B ′ ) un , and both Φ u n \Phi^{\mathrm{un}} Φ un and its inverse, given by extracting the upper blocks, are regular. Hence, Φ u n \Phi^{\mathrm{un}} Φ un is an isomorphism of algebraic groups.
Notice that G p G_p G p is the pointwise stabilizer of V ′ V^{\prime} V ′ in G p u n G_p^{\mathrm{un}} G p un , i.e.,
G p = { ( I V ′ r 0 I V ) ∣ r ∈ Hom ( V , V ′ ) } , G_p=\left\{\left.
\begin{pmatrix}
I_{V^{\prime}}&r\\
0&I_V
\end{pmatrix}
\right|r\in\operatorname{Hom}(V,V^{\prime})\right\}, G p = { ( I V ′ 0 r I V ) r ∈ Hom ( V , V ′ ) } ,
which implies that G p G_p G p is a unipotent group and ( Φ u n ) − 1 ( G p ) = Hom ( V , V ′ ) × { I V ′ } (\Phi^{\mathrm{un}})^{-1}(G_p)=\operatorname{Hom}(V,V^{\prime})\times\{I_{V^{\prime}}\} ( Φ un ) − 1 ( G p ) = Hom ( V , V ′ ) × { I V ′ } . Indeed, the map Φ u n ∣ Hom ( V , V ′ ) × { I V ′ } \left.\Phi^{\mathrm{un}}\right|_{\operatorname{Hom}(V,V^{\prime})\times\{I_{V^{\prime}}\}} Φ un ∣ Hom ( V , V ′ ) × { I V ′ } does not depend on the choice of s s s . □ \square □
For θ ∈ G p u n \theta\in G_p^{\mathrm{un}} θ ∈ G p un and ν ′ ′ ∈ Gext ( g , V ′ ) \nu^{\prime\prime}\in\operatorname{Gext}(\mathfrak{g},V^{\prime}) ν ′′ ∈ Gext ( g , V ′ ) , define the transported bracket by
( θ ⋅ ν ′ ′ ) ( a , b ) = θ ( ν ′ ′ ( θ − 1 a , θ − 1 b ) ) , ∀ a , b ∈ V ′ ′ . (2) (\theta\cdot\nu^{\prime\prime})(a,b)=\theta\bigl(\nu^{\prime\prime}(\theta^{-1}a,\theta^{-1}b)\bigr),\qquad\forall a,b\in V^{\prime\prime}. \tag*{(2)} ( θ ⋅ ν ′′ ) ( a , b ) = θ ( ν ′′ ( θ − 1 a , θ − 1 b ) ) , ∀ a , b ∈ V ′′ . (2)
Since p ∘ θ = p = p ∘ θ − 1 p\circ\theta=p=p\circ\theta^{-1} p ∘ θ = p = p ∘ θ − 1 , the map p p p remains a homomorphism to g \mathfrak{g} g with respect to the transported bracket. Thus, transport of structure defines an action of G p u n G_p^{\mathrm{un}} G p un on Gext ( g , V ′ ) \operatorname{Gext}(\mathfrak{g},V^{\prime}) Gext ( g , V ′ ) and, by restriction, an action of its subgroup G p G_p G p . Denote the corresponding action groupoid by Gext ( g , V ′ ) / / G p \operatorname{Gext}(\mathfrak{g},V^{\prime})//G_p Gext ( g , V ′ ) // G p and Gext ( g , V ′ ) / / G p u n \operatorname{Gext}(\mathfrak{g},V^{\prime})//G_p^{\mathrm{un}} Gext ( g , V ′ ) // G p un respectively. The sets GExt ( g , V ′ ) \operatorname{GExt}(\mathfrak{g},V^{\prime}) GExt ( g , V ′ ) (resp. GExt u n ( g , V ′ ) \operatorname{GExt}^{\mathrm{un}}(\mathfrak{g},V^{\prime}) GExt un ( g , V ′ ) ) classify global extensions up to equivalence (resp. unframed equivalence). Hence, we obtain natural bijections
GExt ( g , V ′ ) ≅ π 0 ( Gext ( g , V ′ ) / / G p ) , GExt u n ( g , V ′ ) ≅ π 0 ( Gext ( g , V ′ ) / / G p u n ) . \operatorname{GExt}(\mathfrak{g},V^{\prime})\cong\pi_0\bigl(\operatorname{Gext}(\mathfrak{g},V^{\prime})//G_p\bigr),\qquad\operatorname{GExt}^{\mathrm{un}}(\mathfrak{g},V^{\prime})\cong\pi_0\bigl(\operatorname{Gext}(\mathfrak{g},V^{\prime})//G_p^{\mathrm{un}}\bigr). GExt ( g , V ′ ) ≅ π 0 ( Gext ( g , V ′ ) // G p ) , GExt un ( g , V ′ ) ≅ π 0 ( Gext ( g , V ′ ) // G p un ) .
Next, we recall the concept of global cohomology of Lie algebras and use it to classify isomorphism classes of global extensions of Lie algebras. For more detail, we refer the reader to [4 , 5 ] .
Definition 3.5. Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) be a Lie algebra and V ′ V' V ′ a vector space. Let π V ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) \pi_{V'}\in\operatorname{Hom}(\wedge^{2}V',V') π V ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , ϱ ∈ Hom ( V ⊗ V ′ , V ′ ) \varrho\in\operatorname{Hom}(V\otimes V',V') ϱ ∈ Hom ( V ⊗ V ′ , V ′ ) , and ω ∈ Hom ( ∧ 2 V , V ′ ) \omega\in\operatorname{Hom}(\wedge^{2}V,V') ω ∈ Hom ( ∧ 2 V , V ′ ) . The triple ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) is said to be a global 2-cocycle of g \mathfrak{g} g with values in V ′ V' V ′ if the following equations
∑ c y c π V ′ ( π V ′ ( u , v ) , w ) = 0 , (3) \sum_{\mathrm{cyc}}\pi_{V'}(\pi_{V'}(u,v),w)=0, \tag*{(3)} cyc ∑ π V ′ ( π V ′ ( u , v ) , w ) = 0 , (3) ϱ ( x ) π V ′ ( u , v ) = π V ′ ( ϱ ( x ) u , v ) + π V ′ ( u , ϱ ( x ) v ) , (4) \varrho(x)\pi_{V'}(u,v)=\pi_{V'}(\varrho(x)u,v)+\pi_{V'}(u,\varrho(x)v), \tag*{(4)} ϱ ( x ) π V ′ ( u , v ) = π V ′ ( ϱ ( x ) u , v ) + π V ′ ( u , ϱ ( x ) v ) , (4) [ ϱ ( x ) , ϱ ( y ) ] u − ϱ ( μ ( x , y ) ) u = π V ′ ( ω ( x , y ) , u ) , (5) [\varrho(x),\varrho(y)]u-\varrho(\mu(x,y))u=\pi_{V'}(\omega(x,y),u), \tag*{(5)} [ ϱ ( x ) , ϱ ( y )] u − ϱ ( μ ( x , y )) u = π V ′ ( ω ( x , y ) , u ) , (5) ∑ c y c ϱ ( x ) ω ( y , z ) = ∑ c y c ω ( μ ( x , y ) , z ) , (6) \sum_{\mathrm{cyc}}\varrho(x)\omega(y,z)=\sum_{\mathrm{cyc}}\omega(\mu(x,y),z), \tag*{(6)} cyc ∑ ϱ ( x ) ω ( y , z ) = cyc ∑ ω ( μ ( x , y ) , z ) , (6) hold for all x , y , z ∈ V x,y,z\in V x , y , z ∈ V and u , v , w ∈ V ′ u,v,w\in V' u , v , w ∈ V ′ . The set of these triples is denoted by G Z 2 ( g , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) .
Notice that (3) shows that π V ′ \pi_{V'} π V ′ defines a Lie algebra structure on V ′ V' V ′ , and (4) shows that ϱ ( x ) \varrho(x) ϱ ( x ) is a derivation of the Lie algebra ( V ′ , π V ′ ) (V',\pi_{V'}) ( V ′ , π V ′ ) for all x ∈ V x\in V x ∈ V .
Definition 3.6. Two global 2-cocycles ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) and ( π V ′ ′ , ϱ ′ , ω ′ ) (\pi'_{V'},\varrho',\omega') ( π V ′ ′ , ϱ ′ , ω ′ ) are said to be equivalent (framed equivalent) if there exists a linear map τ : V → V ′ \tau:V\to V' τ : V → V ′ such that
π V ′ ′ = π V ′ , (7) \pi_{V'}'=\pi_{V'}, \tag*{(7)} π V ′ ′ = π V ′ , (7) ϱ ( x ) u − ϱ ′ ( x ) u = π V ′ ( τ ( x ) , u ) , (8) \varrho(x)u-\varrho'(x)u=\pi_{V'}(\tau(x),u), \tag*{(8)} ϱ ( x ) u − ϱ ′ ( x ) u = π V ′ ( τ ( x ) , u ) , (8) ω ( x , y ) − ω ′ ( x , y ) = ϱ ′ ( x ) τ ( y ) − ϱ ′ ( y ) τ ( x ) + π V ′ ( τ ( x ) , τ ( y ) ) − τ ( μ ( x , y ) ) , (9) \omega(x,y)-\omega'(x,y)=\varrho'(x)\tau(y)-\varrho'(y)\tau(x)+\pi_{V'}(\tau(x),\tau(y))-\tau(\mu(x,y)), \tag*{(9)} ω ( x , y ) − ω ′ ( x , y ) = ϱ ′ ( x ) τ ( y ) − ϱ ′ ( y ) τ ( x ) + π V ′ ( τ ( x ) , τ ( y )) − τ ( μ ( x , y )) , (9) for all x , y ∈ V x,y\in V x , y ∈ V and u ∈ V ′ u\in V' u ∈ V ′ . The set of equivalence classes is denoted by G H 2 ( g , V ′ ) \mathrm{GH}^{2}(\mathfrak{g},V') GH 2 ( g , V ′ ) , called the second global cohomology of g \mathfrak{g} g with values in V ′ V' V ′ .
The set G Z 2 ( g , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) of global 2-cocycles is a closed affine subscheme of
Hom ( ∧ 2 V ′ , V ′ ) ⊕ Hom ( V ⊗ V ′ , V ′ ) ⊕ Hom ( ∧ 2 V , V ′ ) . \operatorname{Hom}(\wedge^{2}V',V')\oplus\operatorname{Hom}(V\otimes V',V')\oplus\operatorname{Hom}(\wedge^{2}V,V'). Hom ( ∧ 2 V ′ , V ′ ) ⊕ Hom ( V ⊗ V ′ , V ′ ) ⊕ Hom ( ∧ 2 V , V ′ ) .
The additive group Hom ( V , V ′ ) \operatorname{Hom}(V,V') Hom ( V , V ′ ) acts on G Z 2 ( g , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) by
τ ⋅ ( π V ′ , ϱ , ω ) = ( π V ′ , ϱ τ , ω τ ) , \tau\cdot(\pi_{V'},\varrho,\omega)=(\pi_{V'},\varrho^{\tau},\omega^{\tau}), τ ⋅ ( π V ′ , ϱ , ω ) = ( π V ′ , ϱ τ , ω τ ) ,
where
ϱ τ ( x ) u = ϱ ( x ) u − π V ′ ( τ ( x ) , u ) , \varrho^{\tau}(x)u=\varrho(x)u-\pi_{V'}(\tau(x),u), ϱ τ ( x ) u = ϱ ( x ) u − π V ′ ( τ ( x ) , u ) ,
ω τ ( x , y ) = ω ( x , y ) − ϱ ( x ) τ ( y ) + ϱ ( y ) τ ( x ) + π V ′ ( τ ( x ) , τ ( y ) ) + τ ( μ ( x , y ) ) . \omega^{\tau}(x,y)=\omega(x,y)-\varrho(x)\tau(y)+\varrho(y)\tau(x)+\pi_{V'}(\tau(x),\tau(y))+\tau(\mu(x,y)). ω τ ( x , y ) = ω ( x , y ) − ϱ ( x ) τ ( y ) + ϱ ( y ) τ ( x ) + π V ′ ( τ ( x ) , τ ( y )) + τ ( μ ( x , y )) .
Its orbits are precisely the equivalence classes of global 2-cocycles. Thus,
G H 2 ( g , V ′ ) ≅ π 0 ( G Z 2 ( g , V ′ ) / / Hom ( V , V ′ ) ) , \mathrm{GH}^{2}(\mathfrak{g},V')\cong\pi_{0}\left(\mathrm{GZ}^{2}(\mathfrak{g},V')//\operatorname{Hom}(V,V')\right), GH 2 ( g , V ′ ) ≅ π 0 ( GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) ) ,
where G Z 2 ( g , V ′ ) / / Hom ( V , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V')//\operatorname{Hom}(V,V') GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) denotes the action groupoid.
We now introduce a notion of unframed equivalence for global 2-cocycles, analogous to that for global extensions.
Definition 3.7. Two global 2-cocycles ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) and ( π V ′ ′ , ϱ ′ , ω ′ ) (\pi'_{V'},\varrho',\omega') ( π V ′ ′ , ϱ ′ , ω ′ ) are said to be unframed equivalent if there exists a linear map τ : V → V ′ \tau:V\to V' τ : V → V ′ and B ∈ GL ( V ′ ) B\in\operatorname{GL}(V') B ∈ GL ( V ′ ) such that
π V ′ ′ ( u , v ) = B π V ′ ( B − 1 u , B − 1 v ) , \pi'_{V'}(u,v)=B\pi_{V'}(B^{-1}u,B^{-1}v), π V ′ ′ ( u , v ) = B π V ′ ( B − 1 u , B − 1 v ) , ϱ ′ ( x ) u = B ϱ ( x ) B − 1 ( u ) − π V ′ ′ ( τ ( x ) , u ) , \varrho'(x)u=B\varrho(x)B^{-1}(u)-\pi'_{V'}(\tau(x),u), ϱ ′ ( x ) u = B ϱ ( x ) B − 1 ( u ) − π V ′ ′ ( τ ( x ) , u ) , ω ′ ( x , y ) = B ω ( x , y ) − B ϱ ( x ) B − 1 τ ( y ) + B ϱ ( y ) B − 1 τ ( x ) + π V ′ ′ ( τ ( x ) , τ ( y ) ) + τ ( μ ( x , y ) ) , \omega'(x,y)=B\omega(x,y)-B\varrho(x)B^{-1}\tau(y)+B\varrho(y)B^{-1}\tau(x)+\pi'_{V'}(\tau(x),\tau(y))+\tau(\mu(x,y)), ω ′ ( x , y ) = B ω ( x , y ) − B ϱ ( x ) B − 1 τ ( y ) + B ϱ ( y ) B − 1 τ ( x ) + π V ′ ′ ( τ ( x ) , τ ( y )) + τ ( μ ( x , y )) ,
for all x , y ∈ V x,y \in V x , y ∈ V and u ∈ V ′ u \in V' u ∈ V ′ . The set of unframed equivalence classes is denoted by G H u n , 2 ( g , V ′ ) \mathrm{GH}^{\mathrm{un},2}(\mathfrak{g},V') GH un , 2 ( g , V ′ ) , called the unframed second global cohomology of g \mathfrak{g} g with values in V ′ V' V ′ .
Denote by H V , V ′ : = Hom ( V , V ′ ) ⋊ GL ( V ′ ) H_{V,V'} := \operatorname{Hom}(V,V') \rtimes\operatorname{GL}(V') H V , V ′ := Hom ( V , V ′ ) ⋊ GL ( V ′ ) with the group law
( τ , B ) ( η , C ) = ( τ + B η , B C ) , ∀ B , C ∈ GL ( V ′ ) , τ , η ∈ Hom ( V , V ′ ) . (\tau,B)(\eta,C)=(\tau+B\eta,BC),\qquad\forall B,C\in\operatorname{GL}(V'),\ \tau,\eta\in\operatorname{Hom}(V,V'). ( τ , B ) ( η , C ) = ( τ + B η , B C ) , ∀ B , C ∈ GL ( V ′ ) , τ , η ∈ Hom ( V , V ′ ) .
There is a natural action of H V , V ′ H_{V,V'} H V , V ′ on the set G Z 2 ( g , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) by
( τ , B ) ⋅ ( π , ϱ , ω ) = ( π ′ , ϱ ′ , ω ′ ) , ∀ ( τ , B ) ∈ H V , V ′ (\tau,B)\cdot(\pi,\varrho,\omega)=(\pi',\varrho',\omega'),\qquad\forall(\tau,B)\in H_{V,V'} ( τ , B ) ⋅ ( π , ϱ , ω ) = ( π ′ , ϱ ′ , ω ′ ) , ∀ ( τ , B ) ∈ H V , V ′
such that for x , y ∈ V x,y\in V x , y ∈ V , u , v ∈ V ′ u,v\in V' u , v ∈ V ′ ,
π ′ ( u , v ) = B π ( B − 1 u , B − 1 v ) , \pi'(u,v)=B\pi(B^{-1}u,B^{-1}v), π ′ ( u , v ) = B π ( B − 1 u , B − 1 v ) ,
ϱ ′ ( x ) u = B ϱ ( x ) B − 1 ( u ) − π ′ ( τ ( x ) , u ) , \varrho'(x)u=B\varrho(x)B^{-1}(u)-\pi'(\tau(x),u), ϱ ′ ( x ) u = B ϱ ( x ) B − 1 ( u ) − π ′ ( τ ( x ) , u ) ,
ω ′ ( x , y ) = B ω ( x , y ) − B ϱ ( x ) B − 1 τ ( y ) + B ϱ ( y ) B − 1 τ ( x ) + π ′ ( τ ( x ) , τ ( y ) ) + τ ( μ ( x , y ) ) . \omega'(x,y)=B\omega(x,y)-B\varrho(x)B^{-1}\tau(y)+B\varrho(y)B^{-1}\tau(x)+\pi'(\tau(x),\tau(y))+\tau(\mu(x,y)). ω ′ ( x , y ) = B ω ( x , y ) − B ϱ ( x ) B − 1 τ ( y ) + B ϱ ( y ) B − 1 τ ( x ) + π ′ ( τ ( x ) , τ ( y )) + τ ( μ ( x , y )) .
The orbits are precisely the unframed equivalence classes of global 2 2 2 -cocycles, i.e.,
G H u n , 2 ( g , V ′ ) ≅ π 0 ( G Z 2 ( g , V ′ ) / / H V , V ′ ) , \mathrm{GH}^{\mathrm{un},2}(\mathfrak{g},V')\cong\pi_0\left(\mathrm{GZ}^{2}(\mathfrak{g},V')//H_{V,V'}\right), GH un , 2 ( g , V ′ ) ≅ π 0 ( GZ 2 ( g , V ′ ) // H V , V ′ ) ,
where G Z 2 ( g , V ′ ) / / H V , V ′ \mathrm{GZ}^{2}(\mathfrak{g},V')//H_{V,V'} GZ 2 ( g , V ′ ) // H V , V ′ denotes the action groupoid.
Now fix the global extension datum ( V ′ ′ , p , g ) (V'',p,\mathfrak{g}) ( V ′′ , p , g ) , write V ′ = ker p V'=\ker p V ′ = ker p , and choose a linear section s : V → V ′ ′ s:V\to V'' s : V → V ′′ of p p p . The section induces an isomorphism of affine schemes
Ψ s : Gext ( g , V ′ ) ⟶ G Z 2 ( g , V ′ ) , ν ′ ′ ⟼ ( π V ′ , ϱ , ω ) , (10) \Psi_s:\operatorname{Gext}(\mathfrak{g},V')\longrightarrow\mathrm{GZ}^{2}(\mathfrak{g},V'),\qquad\nu''\longmapsto(\pi_{V'},\varrho,\omega), \tag*{(10)} Ψ s : Gext ( g , V ′ ) ⟶ GZ 2 ( g , V ′ ) , ν ′′ ⟼ ( π V ′ , ϱ , ω ) , (10)
defined by
π V ′ ( u , v ) = ν ′ ′ ( u , v ) , \pi_{V'}(u,v)=\nu''(u,v), π V ′ ( u , v ) = ν ′′ ( u , v ) ,
ϱ ( x ) u = ν ′ ′ ( s ( x ) , u ) , \varrho(x)u=\nu''(s(x),u), ϱ ( x ) u = ν ′′ ( s ( x ) , u ) ,
ω ( x , y ) = ν ′ ′ ( s ( x ) , s ( y ) ) − s ( μ ( x , y ) ) . (11) \omega(x,y)=\nu''(s(x),s(y))-s(\mu(x,y)). \tag*{(11)} ω ( x , y ) = ν ′′ ( s ( x ) , s ( y )) − s ( μ ( x , y )) . (11)
Its inverse assigns to ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) the bracket on V ′ ′ ≅ V ′ ⊕ V V''\cong V'\oplus V V ′′ ≅ V ′ ⊕ V given by
[ ( u , x ) , ( v , y ) ] = ( π V ′ ( u , v ) + ϱ ( x ) v − ϱ ( y ) u + ω ( x , y ) , μ ( x , y ) ) . [(u,x),(v,y)]=\left(\pi_{V'}(u,v)+\varrho(x)v-\varrho(y)u+\omega(x,y),\mu(x,y)\right). [( u , x ) , ( v , y )] = ( π V ′ ( u , v ) + ϱ ( x ) v − ϱ ( y ) u + ω ( x , y ) , μ ( x , y ) ) .
Proposition 3.8. Let s : V → V ′ ′ s:V\to V'' s : V → V ′′ be a linear section of p p p , and let H V , V ′ = Hom ( V , V ′ ) ⋊ GL ( V ′ ) H_{V,V'}=\operatorname{Hom}(V,V')\rtimes\operatorname{GL}(V') H V , V ′ = Hom ( V , V ′ ) ⋊ GL ( V ′ ) . The pair ( ( Φ u n ) − 1 , Ψ s ) ((\Phi^{\mathrm{un}})^{-1},\Psi_s) (( Φ un ) − 1 , Ψ s ) induces an isomorphism of algebraic action groupoids
Gext ( g , V ′ ) / / G p u n ⟶ G Z 2 ( g , V ′ ) / / H V , V ′ . \operatorname{Gext}(\mathfrak{g},V')//G_p^{\mathrm{un}}\longrightarrow\mathrm{GZ}^{2}(\mathfrak{g},V')//H_{V,V'}. Gext ( g , V ′ ) // G p un ⟶ GZ 2 ( g , V ′ ) // H V , V ′ . It sends an object ν ′ ′ \nu'' ν ′′ to Ψ s ( ν ′ ′ ) \Psi_s(\nu'') Ψ s ( ν ′′ ) and an arrow ( θ , ν ′ ′ ) (\theta,\nu'') ( θ , ν ′′ ) to ( ( Φ u n ) − 1 ( θ ) , Ψ s ( ν ′ ′ ) ) ((\Phi^{\mathrm{un}})^{-1}(\theta),\Psi_s(\nu'')) (( Φ un ) − 1 ( θ ) , Ψ s ( ν ′′ )) . Restricting to the action of G p G_p G p gives an isomorphism
Gext ( g , V ′ ) / / G p ⟶ G Z 2 ( g , V ′ ) / / Hom ( V , V ′ ) . \operatorname{Gext}(\mathfrak{g},V')//G_p\longrightarrow\mathrm{GZ}^{2}(\mathfrak{g},V')//\operatorname{Hom}(V,V'). Gext ( g , V ′ ) // G p ⟶ GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) .
Proof. Since Φ u n \Phi^{\mathrm{un}} Φ un and Ψ s \Psi_s Ψ s are algebraic isomorphisms, it suffices to verify equivariance. Let h = ( r , B ) ∈ H V , V ′ h=(r,B)\in H_{V,V'} h = ( r , B ) ∈ H V , V ′ and ν ′ ′ ∈ Gext ( g , V ′ ) \nu''\in\operatorname{Gext}(\mathfrak{g},V') ν ′′ ∈ Gext ( g , V ′ ) . Write
Ψ s ( ν ′ ′ ) = ( π V ′ , ϱ , ω ) , Ψ s ( Φ h u n ⋅ ν ′ ′ ) = ( π V ′ h , ϱ h , ω h ) . \Psi_s(\nu'')=(\pi_{V'},\varrho,\omega),\qquad\Psi_s(\Phi_h^{\mathrm{un}}\cdot\nu'')=(\pi_{V'}^h,\varrho^h,\omega^h). Ψ s ( ν ′′ ) = ( π V ′ , ϱ , ω ) , Ψ s ( Φ h un ⋅ ν ′′ ) = ( π V ′ h , ϱ h , ω h ) .
By (1) , the inverse of Φ h u n \Phi_h^{\mathrm{un}} Φ h un is given by
( Φ h u n ) − 1 ( u + s ( x ) ) = B − 1 ( u − r ( x ) ) + s ( x ) . (\Phi_h^{\mathrm{un}})^{-1}(u+s(x))=B^{-1}(u-r(x))+s(x). ( Φ h un ) − 1 ( u + s ( x )) = B − 1 ( u − r ( x )) + s ( x ) .
Substituting this expression into the transport formula (2) gives, for x , y ∈ V x,y\in V x , y ∈ V and u , v ∈ V ′ u,v\in V' u , v ∈ V ′ ,
π V ′ h ( u , v ) = B ( π V ′ ( B − 1 u , B − 1 v ) ) , \pi_{V'}^h(u,v)=B\left(\pi_{V'}(B^{-1}u,B^{-1}v)\right), π V ′ h ( u , v ) = B ( π V ′ ( B − 1 u , B − 1 v ) ) ,
ρ h ( x ) u = ( B ϱ ( x ) B − 1 ) u − π V ′ ( r ( x ) , u ) , ω h ( x , y ) = B ω ( x , y ) + r ( μ ( x , y ) ) − ( B ϱ ( x ) B − 1 ) r ( y ) + ( B ϱ ( y ) B − 1 ) r ( x ) + π V ′ ( r ( x ) , r ( y ) ) . \begin{aligned}
\rho^{h}(x)u &= (B\varrho(x)B^{-1})u-\pi_{V'}(r(x),u),\\
\omega^{h}(x,y) &= B\omega(x,y)+r(\mu(x,y))-(B\varrho(x)B^{-1})r(y)+(B\varrho(y)B^{-1})r(x)+\pi_{V'}(r(x),r(y)).
\end{aligned} ρ h ( x ) u ω h ( x , y ) = ( B ϱ ( x ) B − 1 ) u − π V ′ ( r ( x ) , u ) , = B ω ( x , y ) + r ( μ ( x , y )) − ( B ϱ ( x ) B − 1 ) r ( y ) + ( B ϱ ( y ) B − 1 ) r ( x ) + π V ′ ( r ( x ) , r ( y )) .
These are precisely the components of h ⋅ ( π V ′ , ϱ , ω ) h\cdot(\pi_{V'},\varrho,\omega) h ⋅ ( π V ′ , ϱ , ω ) . Hence
Ψ s ( Φ s u n ⋅ ν ′ ′ ) = h ⋅ Ψ s ( ν ′ ′ ) . (12) \Psi_s(\Phi_s^{\mathrm{un}}\cdot\nu^{\prime\prime})=h\cdot\Psi_s(\nu^{\prime\prime}). \tag*{(12)} Ψ s ( Φ s un ⋅ ν ′′ ) = h ⋅ Ψ s ( ν ′′ ) . (12)
This proves equivariance and yields the first groupoid isomorphism, whose inverse is induced by ( Φ u n , Ψ s − 1 ) (\Phi^{\mathrm{un}},\Psi_s^{-1}) ( Φ un , Ψ s − 1 ) .
Finally, Φ u n \Phi^{\mathrm{un}} Φ un identifies the subgroup { ( r , I V ′ ) ∣ r ∈ Hom ( V , V ′ ) } \{(r,I_{V'})\mid r\in\operatorname{Hom}(V,V')\} {( r , I V ′ ) ∣ r ∈ Hom ( V , V ′ )} with G p G_p G p . Restricting (12) to this subgroup and identifying ( r , I V ′ ) (r,I_{V'}) ( r , I V ′ ) with r r r gives the second isomorphism.
Passing to connected components yields the following classifications.
Theorem 3.9. Let ( V ′ ′ , p , g ) (V^{\prime\prime},p,\mathfrak{g}) ( V ′′ , p , g ) be a global extension datum with V ′ = ker p V'=\ker p V ′ = ker p .
(1) There is a bijection
GExt ( g , V ′ ) ≅ GH 2 ( g , V ′ ) . \operatorname{GExt}(\mathfrak{g},V')\cong\operatorname{GH}^{2}(\mathfrak{g},V'). GExt ( g , V ′ ) ≅ GH 2 ( g , V ′ ) . (2) Moreover, there is another bijection
GExt u n ( g , V ′ ) ≅ GH u n , 2 ( g , V ′ ) . \operatorname{GExt}^{\mathrm{un}}(\mathfrak{g},V')\cong\operatorname{GH}^{\mathrm{un},2}(\mathfrak{g},V'). GExt un ( g , V ′ ) ≅ GH un , 2 ( g , V ′ ) .
Remark 3.10. The first part of the above theorem recovers Theorem 2.9 in [5 ] . Furthermore, the above theorem holds regardless of the choice of the section s : V → V ′ ′ s:V\to V^{\prime\prime} s : V → V ′′ .
Indeed, any global extension structure ν ′ ′ \nu^{\prime\prime} ν ′′ on V ′ ′ V^{\prime\prime} V ′′ induces a non-abelian extension of g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) by ( ker p , ν ′ ′ ∣ ker p ) (\ker p,\nu^{\prime\prime}|_{\ker p}) ( ker p , ν ′′ ∣ k e r p ) , i.e.,
0 ⟶ ( ker p , ν ′ ′ ∣ ker p ) → ι ( V ′ ′ , ν ′ ′ ) → p g ⟶ 0. 0\longrightarrow(\ker p,\nu^{\prime\prime}|_{\ker p})\xrightarrow{\ \iota\ }(V^{\prime\prime},\nu^{\prime\prime})\xrightarrow{\ p\ }\mathfrak{g}\longrightarrow0. 0 ⟶ ( ker p , ν ′′ ∣ k e r p ) ι ( V ′′ , ν ′′ ) p g ⟶ 0.
There is an obvious relation between global extensions and non-abelian extensions of Lie algebras. Indeed, such relation also has a cohomological shadow.
Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) and g ′ = ( V ′ , μ V ′ ) \mathfrak{g}'=(V',\mu_{V'}) g ′ = ( V ′ , μ V ′ ) be Lie algebras. Let 0 → g ′ → ι g ′ ′ → j g → 0 0\to\mathfrak{g}'\xrightarrow{\iota}\mathfrak{g}^{\prime\prime}\xrightarrow{j}\mathfrak{g}\to0 0 → g ′ ι g ′′ j g → 0 be a non-abelian extension of g \mathfrak{g} g by g ′ \mathfrak{g}' g ′ .
Definition 3.11. ([16 Definition 1.10] ) A non-abelian 2-cocycle on g \mathfrak{g} g with values in g ′ \mathfrak{g}' g ′ is a pair ( χ , ψ ) (\chi,\psi) ( χ , ψ ) of linear maps χ : ∧ 2 V → V ′ \chi:\wedge^2V\to V' χ : ∧ 2 V → V ′ and ψ : V → Der ( V ′ , μ V ′ ) \psi:V\to\operatorname{Der}(V',\mu_{V'}) ψ : V → Der ( V ′ , μ V ′ ) such that for all x , y ∈ V x,y\in V x , y ∈ V ,
[ ψ ( x ) , ψ ( y ) ] − ψ ( μ ( x , y ) ) = ad χ ( x , y ) , [\psi(x),\psi(y)]-\psi(\mu(x,y))=\operatorname{ad}_{\chi(x,y)}, [ ψ ( x ) , ψ ( y )] − ψ ( μ ( x , y )) = ad χ ( x , y ) , ∑ c y c ψ ( x ) χ ( y , z ) = ∑ c y c χ ( μ ( x , y ) , z ) . \sum_{\mathrm{cyc}}\psi(x)\chi(y,z)=\sum_{\mathrm{cyc}}\chi(\mu(x,y),z). cyc ∑ ψ ( x ) χ ( y , z ) = cyc ∑ χ ( μ ( x , y ) , z ) . Two cocycles ( χ , ψ ) (\chi,\psi) ( χ , ψ ) and ( χ ′ , ψ ′ ) (\chi',\psi') ( χ ′ , ψ ′ ) are equivalent if there exists β ∈ Hom ( V , V ′ ) \beta\in\operatorname{Hom}(V,V') β ∈ Hom ( V , V ′ ) such that
ψ ′ ( x ) = ψ ( x ) + ad β ( x ) , \psi'(x)=\psi(x)+\operatorname{ad}_{\beta(x)}, ψ ′ ( x ) = ψ ( x ) + ad β ( x ) , χ ′ ( x , y ) = χ ( x , y ) + ψ ( x ) β ( y ) − ψ ( y ) β ( x ) − β ( μ ( x , y ) ) + μ V ′ ( β ( x ) , β ( y ) ) . \chi'(x,y)=\chi(x,y)+\psi(x)\beta(y)-\psi(y)\beta(x)-\beta(\mu(x,y))+\mu_{V'}(\beta(x),\beta(y)). χ ′ ( x , y ) = χ ( x , y ) + ψ ( x ) β ( y ) − ψ ( y ) β ( x ) − β ( μ ( x , y )) + μ V ′ ( β ( x ) , β ( y )) . The set of cocycles is denoted by Z n a b 2 ( V , μ ; V ′ , μ V ′ ) Z_{\mathrm{nab}}^2(V,\mu;V',\mu_{V'}) Z nab 2 ( V , μ ; V ′ , μ V ′ ) . Its quotient by the above equivalence relation is the second non-abelian cohomology set H n a b 2 ( V , μ ; V ′ , μ V ′ ) H_{\mathrm{nab}}^2(V,\mu;V',\mu_{V'}) H nab 2 ( V , μ ; V ′ , μ V ′ ) .
The map ( χ , ψ ) ↦ ( μ V ′ , ψ , χ ) (\chi,\psi)\mapsto(\mu_{V'},\psi,\chi) ( χ , ψ ) ↦ ( μ V ′ , ψ , χ ) identifies non-abelian 2-cocycles with global 2-cocycles having Lie bracket μ V ′ \mu_{V'} μ V ′ on the kernel. Moreover, this map is equivariant with respect to gauge actions on both sides. Since the global action preserves the Lie bracket on the kernel, we obtain the following decomposition.
Theorem 3.12. ([5 Corollary 2.10] ) Let Lie ( V ′ ) ( k ) \operatorname{Lie}(V')(k) Lie ( V ′ ) ( k ) be the set of Lie algebra structures on V ′ V' V ′ . Then
G H 2 ( g , V ′ ) ≅ ∐ μ V ′ ∈ Lie ( V ′ ) ( k ) H n a b 2 ( g , ( V ′ , μ V ′ ) ) . \mathrm{GH}^{2}(\mathfrak{g},V') \cong\coprod_{\mu_{V'}\in\operatorname{Lie}(V')(k)} \mathrm{H}^{2}_{\mathrm{nab}}(\mathfrak{g},(V',\mu_{V'})). GH 2 ( g , V ′ ) ≅ μ V ′ ∈ Lie ( V ′ ) ( k ) ∐ H nab 2 ( g , ( V ′ , μ V ′ )) .
Global extensions of Lie algebras and Deligne groupoids
In this section, we construct a dgLa whose Maurer–Cartan scheme parametrizes global extension structures and whose gauge action recovers their equivalence relation. This identifies the corresponding Deligne groupoid with the classification groupoid of global extensions. Passing to quotient stacks, we show that the formal deformations of a fixed extension are controlled by the associated twisted dgLa, independently of the choice of section.
We first recall a construction of gLas that will be used below. Let U U U and U ′ U' U ′ be vector spaces, and let ( U , U ′ ) k , l = ⋀ k U ⊗ ⋀ l U ′ (U,U')^{k,l}=\bigwedge^{k}U\otimes\bigwedge^{l}U' ( U , U ′ ) k , l = ⋀ k U ⊗ ⋀ l U ′ . For W = U W=U W = U or U ′ U' U ′ , let ι W : W ↪ U ⊕ U ′ \iota_W:W\hookrightarrow U\oplus U' ι W : W ↪ U ⊕ U ′ be the canonical inclusion. The lift of a map f ∈ Hom ( ( U , U ′ ) k , l , W ) f\in\operatorname{Hom}((U,U')^{k,l},W) f ∈ Hom (( U , U ′ ) k , l , W ) is the element f † ∈ Hom ( ⋀ k + l ( U ⊕ U ′ ) , U ⊕ U ′ ) f^\dagger\in\operatorname{Hom}(\bigwedge^{k+l}(U\oplus U'),U\oplus U') f † ∈ Hom ( ⋀ k + l ( U ⊕ U ′ ) , U ⊕ U ′ ) defined by
f † ( u 1 + u 1 ′ , … , u k + l + u k + l ′ ) = ∑ σ ∈ S ( k , l ) ( − 1 ) σ ι W ( f ( u σ ( 1 ) , … , u σ ( k ) , u σ ( k + 1 ) ′ , … , u σ ( k + l ) ′ ) ) . f^\dagger(u_1+u'_1,\ldots,u_{k+l}+u'_{k+l})=\sum_{\sigma\in S(k,l)}(-1)^\sigma\iota_W\left(f(u_{\sigma(1)},\ldots,u_{\sigma(k)},u'_{\sigma(k+1)},\ldots,u'_{\sigma(k+l)})\right). f † ( u 1 + u 1 ′ , … , u k + l + u k + l ′ ) = σ ∈ S ( k , l ) ∑ ( − 1 ) σ ι W ( f ( u σ ( 1 ) , … , u σ ( k ) , u σ ( k + 1 ) ′ , … , u σ ( k + l ) ′ ) ) .
where S ( k , l ) S(k,l) S ( k , l ) denotes the set of ( k , l ) (k,l) ( k , l ) -shuffles and ( − 1 ) σ (-1)^\sigma ( − 1 ) σ is the signature of σ \sigma σ .
Together with the canonical decomposition ⋀ n ( U ⊕ U ′ ) ≅ ⨁ k + l = n ( U , U ′ ) k , l \bigwedge^n(U\oplus U')\cong\bigoplus_{k+l=n}(U,U')^{k,l} ⋀ n ( U ⊕ U ′ ) ≅ ⨁ k + l = n ( U , U ′ ) k , l , these lifts identify L N R , U ⊕ U ′ L_{NR,U\oplus U'} L N R , U ⊕ U ′ with
⨁ n ≥ 1 ( ⨁ k + l = n ( Hom ( ( U , U ′ ) k , l , U ) ⊕ Hom ( ( U , U ′ ) k , l , U ′ ) ) ) , \bigoplus_{n\geq1}\left(\bigoplus_{k+l=n}\left(\operatorname{Hom}((U,U')^{k,l},U)\oplus\operatorname{Hom}((U,U')^{k,l},U')\right)\right), n ≥ 1 ⨁ ( k + l = n ⨁ ( Hom (( U , U ′ ) k , l , U ) ⊕ Hom (( U , U ′ ) k , l , U ′ ) ) ) ,
where the summands with k + l = n k+l=n k + l = n have degree n − 1 n-1 n − 1 . We equip this space with the graded Lie bracket transported from the Nijenhuis–Richardson bracket under this identification.
Lemma 4.1. With the above notations,
( ⨁ n = 1 + ∞ ( ( ⨁ k + l = n Hom ( ( U , U ′ ) k , l , U ) ) ⊕ ( ⨁ k + l = n Hom ( ( U , U ′ ) k , l , U ′ ) ) ) , [ ⋅ , ⋅ ] ) \left(\bigoplus_{n=1}^{+\infty}\left(\left(\bigoplus_{k+l=n}\operatorname{Hom}((U,U')^{k,l},U)\right)\oplus\left(\bigoplus_{k+l=n}\operatorname{Hom}((U,U')^{k,l},U')\right)\right),[\cdot,\cdot]\right) ( n = 1 ⨁ + ∞ ( ( k + l = n ⨁ Hom (( U , U ′ ) k , l , U ) ) ⊕ ( k + l = n ⨁ Hom (( U , U ′ ) k , l , U ′ ) ) ) , [ ⋅ , ⋅ ] ) is a gLa, where [ ⋅ , ⋅ ] [\cdot,\cdot] [ ⋅ , ⋅ ] is induced by the bracket [ ⋅ , ⋅ ] N R [\cdot,\cdot]_{\mathrm{NR}} [ ⋅ , ⋅ ] NR on L N R , U ⊕ U ′ L_{NR,U\oplus U'} L N R , U ⊕ U ′ and satisfies
[ f , g ] † = [ f † , g † ] N R . (13) [f,g]^\dagger=[f^\dagger,g^\dagger]_{\mathrm{NR}}. \tag*{(13)} [ f , g ] † = [ f † , g † ] NR . (13)
A dgLa for global extensions
Let ( V ′ ′ , p , g ) (V'',p,\mathfrak{g}) ( V ′′ , p , g ) be a global extension datum and V ′ V' V ′ denote ker p \ker p ker p , where g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) . Define two graded vector spaces
L V , V ′ = ⨁ n ∈ Z L V , V ′ n , L V , V ′ u n = ⨁ n ∈ Z L V , V ′ u n , n . L_{V,V'}=\bigoplus_{n\in\mathbb{Z}}L^n_{V,V'},\qquad L^{\mathrm{un}}_{V,V'}=\bigoplus_{n\in\mathbb{Z}}L^{\mathrm{un},n}_{V,V'}. L V , V ′ = n ∈ Z ⨁ L V , V ′ n , L V , V ′ un = n ∈ Z ⨁ L V , V ′ un , n .
Both spaces vanish in negative degrees. In degree zero, set
L V , V ′ 0 = Hom ( V , V ′ ) , L^0_{V,V'}=\operatorname{Hom}(V,V'), L V , V ′ 0 = Hom ( V , V ′ ) ,
L V , V ′ u n , 0 = Hom ( V , V ′ ) ⊕ End ( V ′ ) . L^{\mathrm{un},0}_{V,V'}=\operatorname{Hom}(V,V')\oplus\operatorname{End}(V'). L V , V ′ un , 0 = Hom ( V , V ′ ) ⊕ End ( V ′ ) .
For each n ≥ 1 n \ge1 n ≥ 1 , set
L V , V ′ n = L V , V ′ u n , n = ⨁ i , j ≥ 0 i + j = n + 1 Hom ( ∧ i V ⊗ ∧ j V ′ , V ′ ) . L_{V,V'}^{n}=L_{V,V'}^{\mathrm{un},n}=\bigoplus_{\substack{i,j\ge0\\i+j=n+1}}\operatorname{Hom}(\wedge^{i}V\otimes\wedge^{j}V',V'). L V , V ′ n = L V , V ′ un , n = i , j ≥ 0 i + j = n + 1 ⨁ Hom ( ∧ i V ⊗ ∧ j V ′ , V ′ ) .
Define the Lie bracket and differential on L V , V ′ L_{V,V'} L V , V ′ and L V , V ′ u n L_{V,V'}^{\mathrm{un}} L V , V ′ un as follows
[ ⋅ , ⋅ ] L V , V ′ : L V , V ′ × L V , V ′ ⟶ L V , V ′ , d : L V , V ′ ⟶ L V , V ′ , [\cdot,\cdot]_{L_{V,V'}}:L_{V,V'}\times L_{V,V'}\longrightarrow L_{V,V'},\qquad d:L_{V,V'}\longrightarrow L_{V,V'}, [ ⋅ , ⋅ ] L V , V ′ : L V , V ′ × L V , V ′ ⟶ L V , V ′ , d : L V , V ′ ⟶ L V , V ′ ,
( f , g ) ⟼ [ f , g ] L V , V ′ = [ f , g ] , f ⟼ d ( f ) = [ μ , f ] , (f,g)\longmapsto[f,g]_{L_{V,V'}}=[f,g],\qquad f\longmapsto d(f)=[\mu,f], ( f , g ) ⟼ [ f , g ] L V , V ′ = [ f , g ] , f ⟼ d ( f ) = [ μ , f ] ,
for any f ∈ L V , V ′ k f\in L_{V,V'}^{k} f ∈ L V , V ′ k , g ∈ L V , V ′ l g\in L_{V,V'}^{l} g ∈ L V , V ′ l , where [ ⋅ , ⋅ ] [\cdot,\cdot] [ ⋅ , ⋅ ] is given in Lemma 4.1.
Proposition 4.2. With these operations, L V , V ′ L_{V,V'} L V , V ′ and L V , V ′ u n L_{V,V'}^{\mathrm{un}} L V , V ′ un are dgLas.
Proof. Since μ \mu μ is a Maurer–Cartan element of L N R , V ⊕ V ′ L_{\mathrm{NR},V\oplus V'} L NR , V ⊕ V ′ , by Proposition 2.3 there is a differential graded Lie algebra structure on L N R , V ⊕ V ′ L_{\mathrm{NR},V\oplus V'} L NR , V ⊕ V ′ with differential d = [ μ , ⋅ ] d=[\mu,\cdot] d = [ μ , ⋅ ] . By (13) , we conclude that ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) and ( L V , V ′ u n , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'}^{\mathrm{un}},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ un , [ ⋅ , ⋅ ] L V , V ′ , d ) are dgLas. □
With these preparations in place, we can now characterize G Z 2 ( g , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) in terms of Maurer–Cartan elements of ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) .
Theorem 4.3. The set of Maurer–Cartan elements of ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) is in bijection with G Z 2 ( g , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) , i.e.,
G Z 2 ( g , V ′ ) ≅ M C ( L V , V ′ ) , ( π V ′ , ϱ , ω ) ⟼ ω + ϱ + π V ′ , ∀ ( π V ′ , ϱ , ω ) ∈ G Z 2 ( g , V ′ ) . \mathrm{GZ}^{2}(\mathfrak{g},V')\cong\mathrm{MC}(L_{V,V'}),\qquad(\pi_{V'},\varrho,\omega)\longmapsto\omega+\varrho+\pi_{V'},\qquad\forall(\pi_{V'},\varrho,\omega)\in\mathrm{GZ}^{2}(\mathfrak{g},V'). GZ 2 ( g , V ′ ) ≅ MC ( L V , V ′ ) , ( π V ′ , ϱ , ω ) ⟼ ω + ϱ + π V ′ , ∀ ( π V ′ , ϱ , ω ) ∈ GZ 2 ( g , V ′ ) . Proof. Suppose ω + ϱ + π V ′ \omega+\varrho+\pi_{V'} ω + ϱ + π V ′ satisfies the Maurer–Cartan equation, for ω ∈ Hom ( ∧ 2 V , V ′ ) \omega\in\operatorname{Hom}(\wedge^{2}V,V') ω ∈ Hom ( ∧ 2 V , V ′ ) , ϱ ∈ Hom ( V ⊗ V ′ , V ′ ) \varrho\in\operatorname{Hom}(V\otimes V',V') ϱ ∈ Hom ( V ⊗ V ′ , V ′ ) and π V ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) \pi_{V'}\in\operatorname{Hom}(\wedge^{2}V',V') π V ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) . By (13) , we have [ ω , ω ] = 0 [\omega,\omega]=0 [ ω , ω ] = 0 , [ μ , π V ′ ] = 0 [\mu,\pi_{V'}]=0 [ μ , π V ′ ] = 0 . Then, by the expansion of the Maurer–Cartan equation for ω + ϱ + π V ′ \omega+\varrho+\pi_{V'} ω + ϱ + π V ′ in L V , V ′ L_{V,V'} L V , V ′ , we get
Hom ( ∧ 3 V , V ′ ) : [ μ , ω ] + [ ω , ϱ ] = 0 , (14) \operatorname{Hom}(\wedge^{3}V,V'):[\mu,\omega]+[\omega,\varrho]=0, \tag*{(14)} Hom ( ∧ 3 V , V ′ ) : [ μ , ω ] + [ ω , ϱ ] = 0 , (14) Hom ( ∧ 2 V ⊗ V ′ , V ′ ) : 2 ( [ μ , ϱ ] + [ ω , π V ′ ] ) + [ ϱ , ϱ ] = 0 , (15) \operatorname{Hom}(\wedge^{2}V\otimes V',V'):2([\mu,\varrho]+[\omega,\pi_{V'}])+[\varrho,\varrho]=0, \tag*{(15)} Hom ( ∧ 2 V ⊗ V ′ , V ′ ) : 2 ([ μ , ϱ ] + [ ω , π V ′ ]) + [ ϱ , ϱ ] = 0 , (15) Hom ( V ⊗ ∧ 2 V ′ , V ′ ) : [ ϱ , π V ′ ] = 0 , (16) \operatorname{Hom}(V\otimes\wedge^{2}V',V'):[\varrho,\pi_{V'}]=0, \tag*{(16)} Hom ( V ⊗ ∧ 2 V ′ , V ′ ) : [ ϱ , π V ′ ] = 0 , (16) Hom ( ∧ 3 V ′ , V ′ ) : [ π V ′ , π V ′ ] = 0. (17) \operatorname{Hom}(\wedge^{3}V',V'):[\pi_{V'},\pi_{V'}]=0. \tag*{(17)} Hom ( ∧ 3 V ′ , V ′ ) : [ π V ′ , π V ′ ] = 0. (17) Notice that (14) –(17) are equivalent to (6) –(3) . Therefore, the triple ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) is a global 2-cocycle.
Conversely, let ( π V ′ , ϱ , ω ) ∈ G Z 2 ( g , V ′ ) (\pi_{V'},\varrho,\omega)\in\mathrm{GZ}^{2}(\mathfrak{g},V') ( π V ′ , ϱ , ω ) ∈ GZ 2 ( g , V ′ ) be a global 2-cocycle of g \mathfrak{g} g with values in V ′ V' V ′ . Then define ζ = ω + ϱ + π V ′ ∈ L V , V ′ 1 \zeta=\omega+\varrho+\pi_{V'}\in L_{V,V'}^{1} ζ = ω + ϱ + π V ′ ∈ L V , V ′ 1 . By direct calculation, one obtains that ζ \zeta ζ is a Maurer–Cartan element of the dgLa ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) . □
We can characterize the action groupoid G Z 2 ( g , V ′ ) / / Hom ( V , V ′ ) \mathrm{GZ}^{2}(\mathfrak{g},V')//\operatorname{Hom}(V,V') GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) introduced in § 3 in terms of the Deligne groupoid of the corresponding nilpotent dgLa, as follows.
Theorem 4.4. (1) The action groupoid of the global 2-cocycles is the same as the Deligne groupoid of the nilpotent dgLa L V , V ′ L_{V,V'} L V , V ′ , i.e.,
G Z 2 ( g , V ′ ) / / Hom ( V , V ′ ) ≅ Del ( L V , V ′ ) . \mathrm{GZ}^{2}(\mathfrak{g},V')//\operatorname{Hom}(V,V')\cong\operatorname{Del}(L_{V,V'}). GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) ≅ Del ( L V , V ′ ) . (2) Moreover, the set of connected components of Del ( L V , V ′ ) \operatorname{Del}(L_{V,V'}) Del ( L V , V ′ ) is parameterized by the second global cohomology G H 2 ( g , V ′ ) \mathrm{GH}^{2}(\mathfrak{g},V') GH 2 ( g , V ′ ) , i.e.,
π 0 ( Del ( L V , V ′ ) ) ≅ G H 2 ( g , V ′ ) . \pi_{0}(\operatorname{Del}(L_{V,V'}))\cong\mathrm{GH}^{2}(\mathfrak{g},V'). π 0 ( Del ( L V , V ′ )) ≅ GH 2 ( g , V ′ ) .
Proof. By Theorem 4.3, the set of objects of Del ( L V , V ′ , V ′ ) \operatorname{Del}(L_{V,V'},V') Del ( L V , V ′ , V ′ ) is GZ 2 ( g , V ′ ) \operatorname{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) . Then, we only need to prove that Hom Del ( L V , V ′ , V ′ ) ( ζ , ζ ′ ) \operatorname{Hom}_{\operatorname{Del}(L_{V,V'},V')}(\zeta,\zeta') Hom Del ( L V , V ′ , V ′ ) ( ζ , ζ ′ ) is in one-to-one correspondence with the Hom set between ζ = ω + ϱ + π V ′ \zeta=\omega+\varrho+\pi_{V'} ζ = ω + ϱ + π V ′ and ζ ′ = ω ′ + ϱ ′ + π V ′ ′ ∈ GZ 2 ( g , V ′ ) \zeta'=\omega'+\varrho'+\pi'_{V'}\in\operatorname{GZ}^{2}(\mathfrak{g},V') ζ ′ = ω ′ + ϱ ′ + π V ′ ′ ∈ GZ 2 ( g , V ′ ) .
Let τ ∈ Hom Del ( L V , V ′ , V ′ ) ( ζ , ζ ′ ) \tau\in\operatorname{Hom}_{\operatorname{Del}(L_{V,V'},V')}(\zeta,\zeta') τ ∈ Hom Del ( L V , V ′ , V ′ ) ( ζ , ζ ′ ) , such that
ζ ′ = e ad τ ζ + Id − e ad τ ad τ d ( τ ) . (18) \zeta'=e^{\operatorname{ad}\tau}\zeta+\frac{\operatorname{Id}-e^{\operatorname{ad}\tau}}{\operatorname{ad}\tau}d(\tau). \tag*{(18)} ζ ′ = e ad τ ζ + ad τ Id − e ad τ d ( τ ) . (18)
By (13) , we obtain [ τ , ω ] = 0 [\tau,\omega]=0 [ τ , ω ] = 0 and [ τ , [ τ , ϱ ] ] = 0 [\tau,[\tau,\varrho]]=0 [ τ , [ τ , ϱ ]] = 0 , which implies that
e ad τ ζ = ω + ϱ + π V ′ + [ τ , ϱ ] + [ τ , π V ′ ] + 1 2 [ τ , [ τ , π V ′ ] ] . e^{\operatorname{ad}\tau}\zeta=\omega+\varrho+\pi_{V'}+[\tau,\varrho]+[\tau,\pi_{V'}]+\frac{1}{2}[\tau,[\tau,\pi_{V'}]]. e ad τ ζ = ω + ϱ + π V ′ + [ τ , ϱ ] + [ τ , π V ′ ] + 2 1 [ τ , [ τ , π V ′ ]] .
By (13) , we also obtain [ τ , [ μ , τ ] ] = 0 [\tau,[\mu,\tau]]=0 [ τ , [ μ , τ ]] = 0 , which implies that
Id − e ad τ ad τ d ( τ ) = − ∑ n ≥ 0 1 ( n + 1 ) ! ad τ n ( d ( τ ) ) = − ( d ( τ ) + 1 2 [ τ , d ( τ ) ] + ⋯ ) = − [ μ , τ ] , \frac{\operatorname{Id}-e^{\operatorname{ad}\tau}}{\operatorname{ad}\tau}d(\tau)=-\sum_{n\geq0}\frac{1}{(n+1)!}\operatorname{ad}_{\tau}^{n}(d(\tau))=-\left(d(\tau)+\frac{1}{2}[\tau,d(\tau)]+\cdots\right)=-[\mu,\tau], ad τ Id − e ad τ d ( τ ) = − n ≥ 0 ∑ ( n + 1 )! 1 ad τ n ( d ( τ )) = − ( d ( τ ) + 2 1 [ τ , d ( τ )] + ⋯ ) = − [ μ , τ ] ,
where the summation is finite because L V , V ′ L_{V,V'} L V , V ′ is nilpotent. We can rewrite equation (18) and rearrange the result according to the corresponding components. Then we obtain the following equations
Hom ( ∧ 2 V , V ′ ) : ω ′ = ω + [ τ , ϱ ] − [ μ , τ ] + 1 2 [ τ , [ τ , π V ′ ] ] , \operatorname{Hom}(\wedge^{2}V,V'):\quad\omega'=\omega+[\tau,\varrho]-[\mu,\tau]+\frac{1}{2}[\tau,[\tau,\pi_{V'}]], Hom ( ∧ 2 V , V ′ ) : ω ′ = ω + [ τ , ϱ ] − [ μ , τ ] + 2 1 [ τ , [ τ , π V ′ ]] ,
Hom ( V ⊗ V ′ , V ′ ) : ϱ ′ = ϱ + [ τ , π V ′ ] , \operatorname{Hom}(V\otimes V',V'):\quad\varrho'=\varrho+[\tau,\pi_{V'}], Hom ( V ⊗ V ′ , V ′ ) : ϱ ′ = ϱ + [ τ , π V ′ ] ,
Hom ( ∧ 2 V ′ , V ′ ) : π V ′ ′ = π V ′ . \operatorname{Hom}(\wedge^{2}V',V'):\quad\pi'_{V'}=\pi_{V'}. Hom ( ∧ 2 V ′ , V ′ ) : π V ′ ′ = π V ′ .
The above three equations give us the equalities (9) , (8) , and (7) respectively. Hence, we get
( π V ′ ′ , ϱ ′ , ω ′ ) = τ ⋅ ( π V ′ , ϱ , ω ) . (\pi'_{V'},\varrho',\omega')=\tau\cdot(\pi_{V'},\varrho,\omega). ( π V ′ ′ , ϱ ′ , ω ′ ) = τ ⋅ ( π V ′ , ϱ , ω ) .
Conversely, let ( π V ′ ′ , ϱ ′ , ω ′ ) = τ ⋅ ( π V ′ , ϱ , ω ) (\pi'_{V'},\varrho',\omega')=\tau\cdot(\pi_{V'},\varrho,\omega) ( π V ′ ′ , ϱ ′ , ω ′ ) = τ ⋅ ( π V ′ , ϱ , ω ) in GZ 2 ( g , V ′ ) \operatorname{GZ}^{2}(\mathfrak{g},V') GZ 2 ( g , V ′ ) , i.e., the linear map τ : V → V ′ \tau:V\to V' τ : V → V ′ satisfies the equations (7) , (8) , and (9) . Then by direct calculation, one can see that ζ : = ω + ϱ + π V ′ \zeta:=\omega+\varrho+\pi_{V'} ζ := ω + ϱ + π V ′ and ζ ′ : = ω ′ + ϱ ′ + π V ′ ′ \zeta':=\omega'+\varrho'+\pi'_{V'} ζ ′ := ω ′ + ϱ ′ + π V ′ ′ are gauge equivalent through e ad τ e^{\operatorname{ad}\tau} e ad τ , which means that τ ∈ Hom Del ( L V , V ′ , V ′ ) ( ζ , ζ ′ ) \tau\in\operatorname{Hom}_{\operatorname{Del}(L_{V,V'},V')}(\zeta,\zeta') τ ∈ Hom Del ( L V , V ′ , V ′ ) ( ζ , ζ ′ ) .
By the fact that L V , V ′ L_{V,V'} L V , V ′ is abelian, the Baker-Campbell-Hausdorff multiplication reduces to the addition of linear maps. We get
exp ( L V , V ′ 0 ) ≅ ( Hom ( V , V ′ ) , + ) \exp(L_{V,V'}^{0})\cong(\operatorname{Hom}(V,V'),+) exp ( L V , V ′ 0 ) ≅ ( Hom ( V , V ′ ) , + )
as algebraic groups. The equivariance follows from summarizing the above computations into the following equality:
e τ ⋆ ( ω + ϱ + π V ′ ) = τ ⋅ ( π V ′ , ϱ , ω ) . e^{\tau}\star(\omega+\varrho+\pi_{V'})=\tau\cdot(\pi_{V'},\varrho,\omega). e τ ⋆ ( ω + ϱ + π V ′ ) = τ ⋅ ( π V ′ , ϱ , ω ) .
Therefore, we conclude that GZ 2 ( g , V ′ ) / / Hom ( V , V ′ ) ≅ Del ( L V , V ′ ) \operatorname{GZ}^{2}(\mathfrak{g},V')//\operatorname{Hom}(V,V')\cong\operatorname{Del}(L_{V,V'}) GZ 2 ( g , V ′ ) // Hom ( V , V ′ ) ≅ Del ( L V , V ′ ) as algebraic groupoids. The second statement follows from taking π 0 \pi_{0} π 0 of the above groupoids. ∎
Formal deformations and independence of the section
Introduce the quotient stacks
Z g , V ′ : = [ G Z 2 ( g , V ′ ) / Hom ( V , V ′ ) ] , \mathcal{Z}_{\mathfrak{g},V'} := [GZ^{2}(\mathfrak{g},V')/\operatorname{Hom}(V,V')], Z g , V ′ := [ G Z 2 ( g , V ′ ) / Hom ( V , V ′ )] ,
E p : = [ G ext ( g , V ′ ) / G p ] , \mathcal{E}_{p} := [G\operatorname{ext}(\mathfrak{g},V')/G_{p}], E p := [ G ext ( g , V ′ ) / G p ] ,
E p u n : = [ G ext ( g , V ′ ) / G p u n ] . \mathcal{E}_{p}^{\mathrm{un}} := [G\operatorname{ext}(\mathfrak{g},V')/G_{p}^{\mathrm{un}}]. E p un := [ G ext ( g , V ′ ) / G p un ] .
Write
M C ( L V , V ′ ) : = [ MC ( L V , V ′ ) / exp ( L V , V ′ 0 ) ] . \mathfrak{MC}(L_{V,V'}) := [\operatorname{MC}(L_{V,V'})/\exp(L_{V,V'}^{0})]. MC ( L V , V ′ ) := [ MC ( L V , V ′ ) / exp ( L V , V ′ 0 )] .
For a stack X \mathcal{X} X and an object x ∈ X ( k ) x \in\mathcal{X}(k) x ∈ X ( k ) , we use the formal deformation functor in Definition Definition 2.12
Def X , x ( A ) = X ( A ) × X ( k ) { x } , A ∈ A r t k . \operatorname{Def}_{\mathcal{X},x}(A) = \mathcal{X}(A) \mathbin{\times}_{\mathcal{X}(k)} \{x\}, \qquad A \in\mathbf{Art}_{k}. Def X , x ( A ) = X ( A ) × X ( k ) { x } , A ∈ Art k .
Theorem 4.5. There is an equivalence of stacks
Z g , V ′ ≃ M C ( L V , V ′ ) . \mathcal{Z}_{\mathfrak{g},V'} \simeq\mathfrak{MC}(L_{V,V'}). Z g , V ′ ≃ MC ( L V , V ′ ) . For a global cocycle z = ( π , ϱ , ω ) z=(\pi,\varrho,\omega) z = ( π , ϱ , ω ) , let ζ = π + ϱ + ω \zeta=\pi+\varrho+\omega ζ = π + ϱ + ω . Then
Def Z g , V ′ , [ z ] ≃ Def L V , V ′ , ζ , \operatorname{Def}_{\mathcal{Z}_{\mathfrak{g},V'},[z]} \simeq\operatorname{Def}_{L_{V,V'},\zeta}, Def Z g , V ′ , [ z ] ≃ Def L V , V ′ , ζ , naturally as groupoid-valued functors on A r t k \mathbf{Art}_{k} Art k . Therefore, the formal deformation problem of global 2 2 2 -cocycles is controlled by the dgLa L V , V ′ L_{V,V'} L V , V ′ .
Proof. By Theorem 4.4 , there exists an equivalence of quotient stacks, which carries [ z ] [z] [ z ] to [ ζ ] [\zeta] [ ζ ] . Hence, Theorem 2.14 implies
Def Z g , V ′ , [ z ] ≃ Def M C ( L V , V ′ ) , [ ζ ] ≃ Def L V , V ′ , ζ . \operatorname{Def}_{\mathcal{Z}_{\mathfrak{g},V'},[z]} \simeq\operatorname{Def}_{\mathfrak{MC}(L_{V,V'}),[\zeta]} \simeq\operatorname{Def}_{L_{V,V'},\zeta}. Def Z g , V ′ , [ z ] ≃ Def MC ( L V , V ′ ) , [ ζ ] ≃ Def L V , V ′ , ζ .
Corollary 4.6. A linear section s : V → V ′ ′ s:V\to V'' s : V → V ′′ induces an equivalence
E p ≃ Z g , V ′ . \mathcal{E}_{p} \simeq\mathcal{Z}_{\mathfrak{g},V'}. E p ≃ Z g , V ′ . For μ ′ ′ ∈ G ext ( g , V ′ ′ ) ( k ) \mu''\in G\operatorname{ext}(\mathfrak{g},V'')(k) μ ′′ ∈ G ext ( g , V ′′ ) ( k ) and z s = Ψ s ( μ ′ ′ ) z_{s}=\Psi_{s}(\mu'') z s = Ψ s ( μ ′′ ) , where Ψ s \Psi_{s} Ψ s is given by (10) , it induces a natural equivalence
Def E p , [ μ ′ ′ ] ≃ Def Z g , V ′ , [ z s ] . \operatorname{Def}_{\mathcal{E}_{p},[\mu'']} \simeq\operatorname{Def}_{\mathcal{Z}_{\mathfrak{g},V'},[z_{s}]}. Def E p , [ μ ′′ ] ≃ Def Z g , V ′ , [ z s ] .
Proof. By Proposition 3.8 , there is an isomorphism of algebraic action groupoids carrying μ ′ ′ \mu'' μ ′′ to z s z_{s} z s . Passing to quotient stacks and taking homotopy fibers at the corresponding objects, we prove the result.
Note that we need to choose a section s s s in Corollary 4.6. However, the following result shows that it is independent of the choice when we study Def E p , [ μ ′ ′ ] \operatorname{Def}_{\mathcal{E}_{p},[\mu'']} Def E p , [ μ ′′ ] via dgLa.
For a global extension structure μ ′ ′ ∈ G ext ( g , V ′ ′ ) \mu''\in G\operatorname{ext}(\mathfrak{g},V'') μ ′′ ∈ G ext ( g , V ′′ ) , take a section s : V → V ′ ′ s:V\to V'' s : V → V ′′ of p p p . Denote by
L V , V ′ s : = L V , V ′ ( π V ′ s , ϱ s , ω s ) , L V , V ′ u n , s : = ( L V , V ′ u n ) ( π V ′ s , ϱ s , ω s ) L_{V,V'}^{s}:=L_{V,V'}^{(\pi_{V'}^{s},\varrho^{s},\omega^{s})}, \qquad L_{V,V'}^{\mathrm{un},s}:=(L_{V,V'}^{\mathrm{un}})^{(\pi_{V'}^{s},\varrho^{s},\omega^{s})} L V , V ′ s := L V , V ′ ( π V ′ s , ϱ s , ω s ) , L V , V ′ un , s := ( L V , V ′ un ) ( π V ′ s , ϱ s , ω s )
the twisted dgLas of L V , V ′ L_{V,V'} L V , V ′ and L V , V ′ u n L_{V,V'}^{\mathrm{un}} L V , V ′ un by the corresponding Maurer–Cartan element ω s + ϱ s + π V ′ s \omega^{s}+\varrho^{s}+\pi_{V'}^{s} ω s + ϱ s + π V ′ s of the global 2 2 2 -cocycle ( π V ′ s , ϱ s , ω s ) (\pi_{V'}^{s},\varrho^{s},\omega^{s}) ( π V ′ s , ϱ s , ω s ) respectively, where ( π V ′ s , ϱ s , ω s ) (\pi_{V'}^{s},\varrho^{s},\omega^{s}) ( π V ′ s , ϱ s , ω s ) is given by (11) . Let s ′ : V → V ′ ′ s':V\to V'' s ′ : V → V ′′ be another linear section of p p p . Define τ : V → V ′ \tau:V\to V' τ : V → V ′ by
τ ( x ) = s ( x ) − s ′ ( x ) , ∀ x ∈ V . (19) \tau(x)=s(x)-s'(x), \qquad\forall x\in V. \tag*{(19)} τ ( x ) = s ( x ) − s ′ ( x ) , ∀ x ∈ V . (19)
Proposition 4.7. The dgLas L V , V ′ s L^{s}_{V,V'} L V , V ′ s and L V , V ′ s ′ L^{s'}_{V,V'} L V , V ′ s ′ are isomorphic.
Proof. Since s , s ′ s,s' s , s ′ are sections of p p p , it is shown in [5 Lemma 2.6] that the triples ( π V ′ s , ϱ s , ω s ) (\pi^{s}_{V'},\varrho^{s},\omega^{s}) ( π V ′ s , ϱ s , ω s ) and ( π V ′ s ′ , ϱ s ′ , ω s ′ ) (\pi^{s'}_{V'},\varrho^{s'},\omega^{s'}) ( π V ′ s ′ , ϱ s ′ , ω s ′ ) are equivalent global 2-cocycles. As in Theorem 4.4 , ω s + ϱ s + π V ′ s \omega^{s}+\varrho^{s}+\pi^{s}_{V'} ω s + ϱ s + π V ′ s and ω s ′ + ϱ s ′ + π V ′ s ′ \omega^{s'}+\varrho^{s'}+\pi^{s'}_{V'} ω s ′ + ϱ s ′ + π V ′ s ′ are gauge equivalent Maurer–Cartan elements of L V , V ′ L_{V,V'} L V , V ′ via e τ e^{\tau} e τ . By [29 Corollary 5.10] , the two dgLas are isomorphic. □
Theorem 4.8. For every global extension structure μ ′ ′ \mu'' μ ′′ and every linear section s s s of p p p , there is a natural equivalence
Def E p , [ μ ′ ′ ] ≃ Def L V , V ′ s . \operatorname{Def}_{\mathcal{E}_p,[\mu'']} \simeq\operatorname{Def}_{L^{s}_{V,V'}}. Def E p , [ μ ′′ ] ≃ Def L V , V ′ s . Explicitly, for A ∈ A r t k A \in\mathbf{Art}_k A ∈ Art k ,
Def E p , [ μ ′ ′ ] ( A ) ≃ Del ( L V , V ′ s ⊗ m A ) . \operatorname{Def}_{\mathcal{E}_p,[\mu'']}(A) \simeq\operatorname{Del}(L^{s}_{V,V'} \otimes\mathfrak{m}_A). Def E p , [ μ ′′ ] ( A ) ≃ Del ( L V , V ′ s ⊗ m A ) . The controlling dgLa is independent of s s s up to the isomorphisms in Proposition 4.7 .
Proof. By Theorem 4.5 and Corollary 4.6, there are natural equivalences
Def E p , [ μ ′ ′ ] ≃ Def Z g , V ′ , [ z s ] ≃ Def MC ( L V , V ′ ) , [ ζ s ] ≃ Def L V , V ′ s . \operatorname{Def}_{\mathcal{E}_p,[\mu'']} \simeq\operatorname{Def}_{\mathcal{Z}_{g,V'},[z_s]} \simeq\operatorname{Def}_{\operatorname{MC}(L_{V,V'}),[\zeta_s]} \simeq\operatorname{Def}_{L^{s}_{V,V'}}. Def E p , [ μ ′′ ] ≃ Def Z g , V ′ , [ z s ] ≃ Def MC ( L V , V ′ ) , [ ζ s ] ≃ Def L V , V ′ s .
Similarly, for the unframed equivalence, we also have the following characterization.
Theorem 4.9. The stack of global extension structures with unframed isomorphisms has the two isomorphic presentations
E p u n = [ Gext ( g , V ′ ) / G p u n ] ≅ [ MC ( L V , V ′ u n ) / H V , V ′ ] . \mathcal{E}^{\mathrm{un}}_p = [\operatorname{Gext}(\mathfrak{g},V')/G^{\mathrm{un}}_p] \cong[\operatorname{MC}(L^{\mathrm{un}}_{V,V'})/H_{V,V'}]. E p un = [ Gext ( g , V ′ ) / G p un ] ≅ [ MC ( L V , V ′ un ) / H V , V ′ ] . Every G p u n G^{\mathrm{un}}_p G p un -torsor over Spec k \operatorname{Spec} k Spec k is trivial. Consequently, after choosing the section of p p p , there are isomorphisms of groupoids
E p u n ( k ) ≃ Gext ( g , V ′ ) ( k ) / / G p u n ( k ) ≅ MC ( L V , V ′ u n ) ( k ) / / H V , V ′ ( k ) , \mathcal{E}^{\mathrm{un}}_p(k) \simeq\operatorname{Gext}(\mathfrak{g},V')(k)//G^{\mathrm{un}}_p(k) \cong\operatorname{MC}(L^{\mathrm{un}}_{V,V'})(k)//H_{V,V'}(k), E p un ( k ) ≃ Gext ( g , V ′ ) ( k ) // G p un ( k ) ≅ MC ( L V , V ′ un ) ( k ) // H V , V ′ ( k ) , and the set of connected components is GExt u n ( g , V ′ ) \operatorname{GExt}^{\mathrm{un}}(\mathfrak{g},V') GExt un ( g , V ′ ) . The natural map sending a global extension to the Lie algebra structure on the kernel is equivariant and induces a morphism
q ~ : E p u n ⟶ [ Lie ( V ′ ) / GL ( V ′ ) ] . (20) \widetilde{q}:\mathcal{E}^{\mathrm{un}}_p \longrightarrow[\operatorname{Lie}(V')/\operatorname{GL}(V')]. \tag*{(20)} q : E p un ⟶ [ Lie ( V ′ ) / GL ( V ′ )] . (20) Proof. The intrinsic action of G p u n G^{\mathrm{un}}_p G p un becomes the action of H V , V ′ H_{V,V'} H V , V ′ under the chosen section, which proves the equivalence of quotient stacks. Consider the split exact sequence
1 ⟶ Hom ( V , V ′ ) ⟶ H V , V ′ ⟶ GL ( V ′ ) ⟶ 1. 1 \longrightarrow\operatorname{Hom}(V,V') \longrightarrow H_{V,V'} \longrightarrow\operatorname{GL}(V') \longrightarrow1. 1 ⟶ Hom ( V , V ′ ) ⟶ H V , V ′ ⟶ GL ( V ′ ) ⟶ 1. By the associated long exact sequence of pointed sets, we have
1 ⟶ H 1 ( k , Hom ( V , V ′ ) ) ⟶ H 1 ( k , H V , V ′ ) ⟶ H 1 ( k , GL ( V ′ ) ) = 1. 1 \longrightarrow H^{1}(k,\operatorname{Hom}(V,V')) \longrightarrow H^{1}(k,H_{V,V'}) \longrightarrow H^{1}(k,\operatorname{GL}(V')) = 1. 1 ⟶ H 1 ( k , Hom ( V , V ′ )) ⟶ H 1 ( k , H V , V ′ ) ⟶ H 1 ( k , GL ( V ′ )) = 1. which implies that H 1 ( k , G p u n ) H^{1}(k,G^{\mathrm{un}}_p) H 1 ( k , G p un ) is trivial, since Hom ( V , V ′ ) \operatorname{Hom}(V,V') Hom ( V , V ′ ) is unipotent and has trivial H 1 H^{1} H 1 . Hence, every G p u n G^{\mathrm{un}}_p G p un -torsor over Spec k \operatorname{Spec} k Spec k is trivial, which further implies that
E p u n ( k ) ≃ Gext ( g , V ′ ) ( k ) / / G p u n ( k ) . \mathcal{E}^{\mathrm{un}}_p(k) \simeq\operatorname{Gext}(\mathfrak{g},V')(k)//G^{\mathrm{un}}_p(k). E p un ( k ) ≃ Gext ( g , V ′ ) ( k ) // G p un ( k ) . For ( τ , B ) ∈ H V , V ′ (\tau,B) \in H_{V,V'} ( τ , B ) ∈ H V , V ′ and ζ ∈ MC ( L V , V ′ u n ) \zeta\in\operatorname{MC}(L^{\mathrm{un}}_{V,V'}) ζ ∈ MC ( L V , V ′ un ) , restricting the transported bracket to V ′ V' V ′ , we have
q ( ( τ , B ) ⋅ ζ ) = B ⋅ q ( ζ ) , ( B ⋅ π ) ( u , v ) = B π ( B − 1 u , B − 1 v ) . q((\tau,B)\cdot\zeta)=B\cdot q(\zeta), \qquad(B\cdot\pi)(u,v)=B\pi(B^{-1}u,B^{-1}v). q (( τ , B ) ⋅ ζ ) = B ⋅ q ( ζ ) , ( B ⋅ π ) ( u , v ) = B π ( B − 1 u , B − 1 v ) . Thus, the map q : MC ( L V , V ′ u n ) ⟶ Lie ( V ′ ) q:\operatorname{MC}(L^{\mathrm{un}}_{V,V'})\longrightarrow\operatorname{Lie}(V') q : MC ( L V , V ′ un ) ⟶ Lie ( V ′ ) is equivariant for the homomorphism H V , V ′ ⟶ GL ( V ′ ) H_{V,V'}\longrightarrow\operatorname{GL}(V') H V , V ′ ⟶ GL ( V ′ ) , and it descends to (20) . □
Theorem 4.10. L V , V ′ u n L^{\mathrm{un}}_{V,V'} L V , V ′ un controls the formal deformation of global extension structures under the unframed equivalence, i.e.,
Def p , [ μ ′ ′ ] u n ≃ Def L V , V ′ u n , s . \operatorname{Def}^{\mathrm{un}}_{\mathfrak{p},[\mu'']} \simeq\operatorname{Def}^{\mathrm{un,s}}_{L_{V,V'}}. Def p , [ μ ′′ ] un ≃ Def L V , V ′ un , s .
Proof. According to Theorem 4.9 and Proposition 2.13 , we obtain
Def p , [ μ ′ ′ ] ( A ) ≃ MC ( L V , V ′ ) μ ′ ′ ( A ) / / H 1 ( A ) , \operatorname{Def}_{\mathfrak{p},[\mu'']}(A) \simeq\operatorname{MC}(L_{V,V'})_{\mu''}(A)//H_1(A), Def p , [ μ ′′ ] ( A ) ≃ MC ( L V , V ′ ) μ ′′ ( A ) // H 1 ( A ) ,
where H 1 ( A ) = ker ( H V , V ′ ( A ) ⟶ H V , V ′ ( k ) ) H_1(A)=\ker\left(H_{V,V'}(A)\longrightarrow H_{V,V'}(k)\right) H 1 ( A ) = ker ( H V , V ′ ( A ) ⟶ H V , V ′ ( k ) ) . Since MC ( L V , V ′ u n ) = MC ( L V , V ′ ) \operatorname{MC}(L^{\mathrm{un}}_{V,V'})=\operatorname{MC}(L_{V,V'}) MC ( L V , V ′ un ) = MC ( L V , V ′ ) , by Theorem 4.8 , we have
MC ( L V , V ′ u n ) μ ′ ′ ( A ) = MC ( L V , V ′ ) μ ′ ′ ( A ) = MC ( L V , V ′ s ⊗ m A ) = MC ( L V , V ′ u n , s ⊗ m A ) . \operatorname{MC}(L^{\mathrm{un}}_{V,V'})_{\mu''}(A)=\operatorname{MC}(L_{V,V'})_{\mu''}(A)=\operatorname{MC}(L^{\mathrm{s}}_{V,V'}\otimes\mathfrak{m}_A)=\operatorname{MC}(L^{\mathrm{un,s}}_{V,V'}\otimes\mathfrak{m}_A). MC ( L V , V ′ un ) μ ′′ ( A ) = MC ( L V , V ′ ) μ ′′ ( A ) = MC ( L V , V ′ s ⊗ m A ) = MC ( L V , V ′ un , s ⊗ m A ) .
Since m A \mathfrak{m}_A m A is nilpotent and k k k has characteristic zero, the finite exponential and logarithm series identify H 1 ( A ) H_1(A) H 1 ( A ) with exp ( L V , V ′ u n , 0 ⊗ m A ) \exp(L^{\mathrm{un},0}_{V,V'}\otimes\mathfrak{m}_A) exp ( L V , V ′ un , 0 ⊗ m A ) . Thus, we obtain that
Def p , [ μ ′ ′ ] ( A ) ≃ Del ( L V , V ′ u n , s ⊗ m A ) , \operatorname{Def}_{\mathfrak{p},[\mu'']}(A)\simeq\operatorname{Del}(L^{\mathrm{un,s}}_{V,V'}\otimes\mathfrak{m}_A), Def p , [ μ ′′ ] ( A ) ≃ Del ( L V , V ′ un , s ⊗ m A ) ,
which is natural in A A A . □
Examples
At the end of this section, we classify global extensions up to framed equivalence for two simple examples based on Theorem 4.3 and Theorem 4.4 , which shows the effectiveness of our machinery.
Example 4.11. Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) be a finite-dimensional abelian Lie algebra over k k k . Denote by k ⊕ V → p V k\oplus V\xrightarrow{p}V k ⊕ V p V the projection, i.e.,
p ( r , x ) = x , ∀ x ∈ V , r ∈ k . p(r,x)=x,\qquad\forall x\in V,r\in k. p ( r , x ) = x , ∀ x ∈ V , r ∈ k . In the following, we describe the set GExt ( g , k ) \operatorname{GExt}(\mathfrak{g},k) GExt ( g , k ) of global extensions on k ⊕ V k\oplus V k ⊕ V with respect to p p p and V V V . For ω + ϱ + π k ∈ L V , k \omega+\varrho+\pi_k\in L_{V,k} ω + ϱ + π k ∈ L V , k , where
π k ∈ Hom ( ∧ 2 k , k ) = { 0 } , ϱ ∈ Hom ( V ⊗ k , k ) ≅ V ∗ , ω ∈ Hom ( ∧ 2 V , k ) ≅ ∧ 2 V ∗ , \pi_k\in\operatorname{Hom}(\wedge^2 k,k)=\{0\},\qquad\varrho\in\operatorname{Hom}(V\otimes k,k)\cong V^*,\qquad\omega\in\operatorname{Hom}(\wedge^2V,k)\cong\wedge^2V^*, π k ∈ Hom ( ∧ 2 k , k ) = { 0 } , ϱ ∈ Hom ( V ⊗ k , k ) ≅ V ∗ , ω ∈ Hom ( ∧ 2 V , k ) ≅ ∧ 2 V ∗ , we have that ω + ϱ + π k ∈ MC ( L V , k ) \omega+\varrho+\pi_k\in\operatorname{MC}(L_{V,k}) ω + ϱ + π k ∈ MC ( L V , k ) if and only if
ϱ ( x ) ω ( y , z ) + ϱ ( y ) ω ( z , x ) + ϱ ( z ) ω ( x , y ) = 0 , ∀ x , y , z ∈ V . \varrho(x)\omega(y,z)+\varrho(y)\omega(z,x)+\varrho(z)\omega(x,y)=0,\qquad\forall x,y,z\in V. ϱ ( x ) ω ( y , z ) + ϱ ( y ) ω ( z , x ) + ϱ ( z ) ω ( x , y ) = 0 , ∀ x , y , z ∈ V . That is, ϱ ∧ ω = 0 \varrho\wedge\omega=0 ϱ ∧ ω = 0 .
If ϱ ≠ 0 \varrho\ne0 ϱ = 0 , there exists ς ∈ V ∗ \varsigma\in V^* ς ∈ V ∗ such that ω = ϱ ∧ ς \omega=\varrho\wedge\varsigma ω = ϱ ∧ ς .
If ϱ = 0 \varrho=0 ϱ = 0 , every ω ∈ Hom ( ∧ 2 V , k ) \omega\in\operatorname{Hom}(\wedge^2V,k) ω ∈ Hom ( ∧ 2 V , k ) satisfies ϱ ∧ ω = 0 \varrho\wedge\omega=0 ϱ ∧ ω = 0 .
Therefore,
GZ 2 ( V , k ) ≅ MC ( L V , k ) ≅ ( ⨆ ϱ ∈ { V ∗ ∖ { 0 } } { ϱ } × ( V ∗ / k ϱ ) ) ⨆ ( { 0 } × ∧ 2 V ∗ ) \operatorname{GZ}^2(V,k)\cong\operatorname{MC}(L_{V,k})\cong\left(\bigsqcup_{\varrho\in\{V^*\setminus\{0\}\}}\{\varrho\}\times(V^*/k\varrho)\right)\bigsqcup\left(\{0\}\times\wedge^2V^*\right) GZ 2 ( V , k ) ≅ MC ( L V , k ) ≅ ϱ ∈ { V ∗ ∖ { 0 }} ⨆ { ϱ } × ( V ∗ / k ϱ ) ⨆ ( { 0 } × ∧ 2 V ∗ ) ≅ ( ⨆ ϱ ∈ { V ∗ ∖ { 0 } } { ϱ } × k dim V − 1 ) ⨆ ( { 0 } × k ( dim V − 1 ) dim V 2 ) . \cong\left(\bigsqcup_{\varrho\in\{V^*\setminus\{0\}\}}\{\varrho\}\times k^{\dim V-1}\right)\bigsqcup\left(\{0\}\times k^{\frac{(\dim V-1)\dim V}{2}}\right). ≅ ϱ ∈ { V ∗ ∖ { 0 }} ⨆ { ϱ } × k d i m V − 1 ⨆ ( { 0 } × k 2 ( d i m V − 1 ) d i m V ) . Moreover, by the definition of gauge equivalence of Maurer–Cartan elements, ω + ϱ \omega+\varrho ω + ϱ and ω ′ + ϱ ′ \omega'+\varrho' ω ′ + ϱ ′ are equivalent if and only if
ϱ = ϱ ′ = 0 , ω = ω ′ , \varrho=\varrho'=0,\qquad\omega=\omega', ϱ = ϱ ′ = 0 , ω = ω ′ , or there exists a linear map τ ∈ V ∗ \tau\in V^* τ ∈ V ∗ such that
ϱ = ϱ ′ ≠ 0 , ω − ω ′ = ϱ ′ ∧ τ . \varrho=\varrho'\ne0,\qquad\omega-\omega'=\varrho'\wedge\tau. ϱ = ϱ ′ = 0 , ω − ω ′ = ϱ ′ ∧ τ .
Thus, by Theorem 3.9 (1) and Theorem 4.4 , there are bijections
GExt ( g , k ) ≅ π 0 ( Del ( L V , k ) ) ≅ GH 2 ( g , k ) ≅ ( V ∗ ∖ 0 ) □ k ( dim V − 1 ) dim V 2 \operatorname{GExt}(\mathfrak{g},\mathfrak{k}) \cong\pi_{0}(\operatorname{Del}(L_{V,\mathfrak{k}})) \cong\operatorname{GH}^{2}(\mathfrak{g},\mathfrak{k}) \cong(V^{*}\setminus0)\mathbin{\square} \mathfrak{k}^{\frac{(\dim V-1)\dim V}{2}} GExt ( g , k ) ≅ π 0 ( Del ( L V , k )) ≅ GH 2 ( g , k ) ≅ ( V ∗ ∖ 0 ) □ k 2 ( d i m V − 1 ) d i m V
Moreover, the stabilizer of ( ϱ , ω ) (\varrho,\omega) ( ϱ , ω ) is k ϱ \mathfrak{k}_{\varrho} k ϱ if ϱ ≠ 0 \varrho\ne0 ϱ = 0 , and is V ∗ V^{*} V ∗ if ϱ = 0 \varrho=0 ϱ = 0 .
Example 4.12. Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) be a one-dimensional Lie algebra over k \mathfrak{k} k , and let V ′ V' V ′ be a finite-dimensional vector space over k \mathfrak{k} k . Denote by V ′ ⊕ V → p V V'\oplus V \xrightarrow{p} V V ′ ⊕ V p V the projection, i.e.
p ( u , x ) = x , ∀ x ∈ V , u ∈ V ′ . p(u,x)=x,\qquad\forall x\in V,u\in V'. p ( u , x ) = x , ∀ x ∈ V , u ∈ V ′ .
For the triple ω + ϱ + π V ′ ∈ L V , V ′ 1 \omega+\varrho+\pi_{V'}\in L^{1}_{V,V'} ω + ϱ + π V ′ ∈ L V , V ′ 1 , where
π V ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , ϱ ∈ Hom ( V ⊗ V ′ , V ′ ) ≅ Hom ( V ′ , V ′ ) , ω ∈ Hom ( ∧ 2 V , V ′ ) = { 0 } , \pi_{V'}\in\operatorname{Hom}(\wedge^{2}V',V'),\qquad\varrho\in\operatorname{Hom}(V\otimes V',V')\cong\operatorname{Hom}(V',V'),\qquad\omega\in\operatorname{Hom}(\wedge^{2}V,V')=\{0\}, π V ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , ϱ ∈ Hom ( V ⊗ V ′ , V ′ ) ≅ Hom ( V ′ , V ′ ) , ω ∈ Hom ( ∧ 2 V , V ′ ) = { 0 } ,
we have that ϱ + π V ′ ∈ MC ( L V , V ′ ) \varrho+\pi_{V'}\in\operatorname{MC}(L_{V,V'}) ϱ + π V ′ ∈ MC ( L V , V ′ ) if and only if
0 = π V ′ ( π V ′ ( u , v ) , w ) + π V ′ ( π V ′ ( v , w ) , u ) + π V ′ ( π V ′ ( w , u ) , v ) 0=\pi_{V'}(\pi_{V'}(u,v),w)+\pi_{V'}(\pi_{V'}(v,w),u)+\pi_{V'}(\pi_{V'}(w,u),v) 0 = π V ′ ( π V ′ ( u , v ) , w ) + π V ′ ( π V ′ ( v , w ) , u ) + π V ′ ( π V ′ ( w , u ) , v )
ϱ ( π V ′ ( u , v ) ) = π V ′ ( ϱ ( u ) , v ) + π V ′ ( u , ϱ ( v ) ) , \varrho(\pi_{V'}(u,v))=\pi_{V'}(\varrho(u),v)+\pi_{V'}(u,\varrho(v)), ϱ ( π V ′ ( u , v )) = π V ′ ( ϱ ( u ) , v ) + π V ′ ( u , ϱ ( v )) ,
which means that π V ′ \pi_{V'} π V ′ is a Lie bracket on V ′ V' V ′ and ϱ \varrho ϱ is a derivation of π V ′ \pi_{V'} π V ′ . Thus,
GZ 2 ( g , V ′ ) ≅ MC ( L V , V ′ ) ≅ ∐ π V ′ ∈ MC ( L V ′ ) Der ( V ′ , π V ′ ) , \operatorname{GZ}^{2}(\mathfrak{g},V')\cong\operatorname{MC}(L_{V,V'})\cong\coprod_{\pi_{V'}\in\operatorname{MC}(L_{V'})}\operatorname{Der}(V',\pi_{V'}), GZ 2 ( g , V ′ ) ≅ MC ( L V , V ′ ) ≅ π V ′ ∈ MC ( L V ′ ) ∐ Der ( V ′ , π V ′ ) ,
where Der ( V ′ , π V ′ ) \operatorname{Der}(V',\pi_{V'}) Der ( V ′ , π V ′ ) is the set of derivations of the Lie algebra ( V ′ , π V ′ ) (V',\pi_{V'}) ( V ′ , π V ′ ) . If ϱ + π V ′ \varrho+\pi_{V'} ϱ + π V ′ and ϱ ′ + π V ′ ′ \varrho'+\pi'_{V'} ϱ ′ + π V ′ ′ are equivalent in MC ( L V , V ′ ) \operatorname{MC}(L_{V,V'}) MC ( L V , V ′ ) , then there exists τ ∈ Hom ( V , V ′ ) ≅ V ′ \tau\in\operatorname{Hom}(V,V')\cong V' τ ∈ Hom ( V , V ′ ) ≅ V ′ such that
π V ′ = π V ′ ′ , ϱ = ϱ ′ + ad ( τ ) . \pi_{V'}=\pi'_{V'},\qquad\varrho=\varrho'+\operatorname{ad}(\tau). π V ′ = π V ′ ′ , ϱ = ϱ ′ + ad ( τ ) .
Hence, by Theorem 3.9 (1), there is
GExt ( g , V ′ ) ≅ GH 2 ( g , V ′ ) = ∐ π V ′ ∈ MC ( L V ′ ) Out ( V ′ , π V ′ ) , \operatorname{GExt}(\mathfrak{g},V')\cong\operatorname{GH}^{2}(\mathfrak{g},V')=\coprod_{\pi_{V'}\in\operatorname{MC}(L_{V'})}\operatorname{Out}(V',\pi_{V'}), GExt ( g , V ′ ) ≅ GH 2 ( g , V ′ ) = π V ′ ∈ MC ( L V ′ ) ∐ Out ( V ′ , π V ′ ) ,
where Out ( V ′ , π V ′ ) \operatorname{Out}(V',\pi_{V'}) Out ( V ′ , π V ′ ) is the vector space of outer derivations of the Lie algebra ( V ′ , π V ′ ) (V',\pi_{V'}) ( V ′ , π V ′ ) .
Non-abelian extensions of dgLas and fibrations of groupoids
In this section, we investigate the relation between the Deligne groupoid corresponding to the global cohomology of Lie algebras and the one corresponding to the non-abelian cohomology of Lie algebras.
We first recall the relation between the Deligne groupoid of a specific dgLa and the second non-abelian cohomology of Lie algebras.
Proposition 5.1. ([16 ] , Lemma 2.1, Corollary 2.1, Proposition 2.2)
(1) Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) and g ′ = ( V ′ , μ ′ ) \mathfrak{g}'=(V',\mu') g ′ = ( V ′ , μ ′ ) be Lie algebras. Define a graded vector space L ~ V , V ′ = ⨁ n ∈ Z L ~ V , V ′ n \widetilde{L}_{V,V'}=\bigoplus_{n\in\mathbb{Z}}\widetilde{L}^{n}_{V,V'} L V , V ′ = ⨁ n ∈ Z L V , V ′ n by
L ~ V , V ′ n = { 0 , n ≤ − 1 ; Hom ( V , V ′ ) , n = 0 ; ⨁ i + j = n + 1 i ≥ 1 Hom ( ∧ i V ⊗ ∧ j V ′ , V ′ ) , n ≥ 1. \widetilde{L}^{n}_{V,V'}=
\begin{cases}
0, & n\le-1;\\
\operatorname{Hom}(V,V'), & n=0;\\
\bigoplus_{\substack{i+j=n+1\\ i\ge1}}\operatorname{Hom}(\wedge^{i}V\otimes\wedge^{j}V',V'), & n\ge1.
\end{cases} L V , V ′ n = ⎩ ⎨ ⎧ 0 , Hom ( V , V ′ ) , ⨁ i + j = n + 1 i ≥ 1 Hom ( ∧ i V ⊗ ∧ j V ′ , V ′ ) , n ≤ − 1 ; n = 0 ; n ≥ 1.
Give the bracket and differential on L ~ V , V ′ \widetilde{L}_{V,V'} L V , V ′ as follows
[ ⋅ , ⋅ ] L ~ V , V ′ : L ~ V , V ′ × L ~ V , V ′ ⟶ L ~ V , V ′ , d ~ : L ~ V , V ′ ⟶ L ~ V , V ′ , [\cdot,\cdot]_{\widetilde{L}_{V,V'}}:\widetilde{L}_{V,V'}\times\widetilde{L}_{V,V'}\longrightarrow\widetilde{L}_{V,V'},\qquad
\widetilde{d}:\widetilde{L}_{V,V'}\longrightarrow\widetilde{L}_{V,V'}, [ ⋅ , ⋅ ] L V , V ′ : L V , V ′ × L V , V ′ ⟶ L V , V ′ , d : L V , V ′ ⟶ L V , V ′ ,
( f , g ) ⟼ [ f , g ] L ~ V , V ′ = [ f , g ] , f ⟼ d ~ ( f ) = [ μ + μ ′ , f ] , (f,g)\longmapsto[f,g]_{\widetilde{L}_{V,V'}}=[f,g],\qquad
f\longmapsto\widetilde{d}(f)=[\mu+\mu',f], ( f , g ) ⟼ [ f , g ] L V , V ′ = [ f , g ] , f ⟼ d ( f ) = [ μ + μ ′ , f ] ,
for any f ∈ L ~ V , V ′ k f\in\widetilde{L}^{k}_{V,V'} f ∈ L V , V ′ k , g ∈ L ~ V , V ′ l g\in\widetilde{L}^{l}_{V,V'} g ∈ L V , V ′ l , where [ ⋅ , ⋅ ] [\cdot,\cdot] [ ⋅ , ⋅ ] is given in Lemma Lemma 4.1 . Then, the dgLa ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) is nilpotent.
(2) The set of connected components of Deligne groupoid of ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) is isomorphic to the second non-abelian cohomology of g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) with values in g ′ = ( V ′ , μ ′ ) \mathfrak{g}'=(V',\mu') g ′ = ( V ′ , μ ′ ) , i.e.,
π 0 ( Del ( L ~ V , V ′ ) ) ≅ H n a b 2 ( V , μ ; V ′ , μ ′ ) . \pi_{0}\left(\operatorname{Del}(\widetilde{L}_{V,V'})\right)\cong H^{2}_{\mathrm{nab}}(V,\mu;V',\mu'). π 0 ( Del ( L V , V ′ ) ) ≅ H nab 2 ( V , μ ; V ′ , μ ′ ) .
Indeed, the three dgLas ( L V ′ , [ ⋅ , ⋅ ] N R ) (L_{V'},[\cdot,\cdot]_{\mathrm{NR}}) ( L V ′ , [ ⋅ , ⋅ ] NR ) , ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) and ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) , introduced respectively in Example Example 2.6 , Proposition Proposition 5.1 and Proposition Proposition 4.2 , fit together into a short exact sequence of dgLas, as shown below. Here, in the construction of ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) , V ′ V' V ′ is regarded as a trivial Lie algebra.
Lemma 5.2. With the above notations, ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) is a non-abelian extension of ( L V ′ , [ ⋅ , ⋅ ] N R ) (L_{V'},[\cdot,\cdot]_{\mathrm{NR}}) ( L V ′ , [ ⋅ , ⋅ ] NR ) by ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) , i.e., the following sequence is exact
0 ⟶ L ~ V , V ′ → ι L V , V ′ → p r L V ′ ⟶ 0. 0\longrightarrow\widetilde{L}_{V,V'}\xrightarrow{\iota}L_{V,V'}\xrightarrow{\mathrm{pr}}L_{V'}\longrightarrow0. 0 ⟶ L V , V ′ ι L V , V ′ pr L V ′ ⟶ 0. where
ι n : L ~ V , V ′ n ⟶ L V , V ′ n , f ⟼ ( f , 0 ) , ∀ f ∈ ⨁ i + j = n + 1 i ≥ 1 Hom ( ∧ i V ⊗ ∧ j V ′ , V ′ ) , \iota_n:\widetilde{L}^{n}_{V,V'}\longrightarrow L^{n}_{V,V'},\qquad
f\longmapsto(f,0),\qquad
\forall f\in\bigoplus_{\substack{i+j=n+1\\i\geq1}}\operatorname{Hom}(\wedge^{i}V\otimes\wedge^{j}V',V'), ι n : L V , V ′ n ⟶ L V , V ′ n , f ⟼ ( f , 0 ) , ∀ f ∈ i + j = n + 1 i ≥ 1 ⨁ Hom ( ∧ i V ⊗ ∧ j V ′ , V ′ ) , and
p r n : L V , V ′ n ⟶ L V ′ n , ( g 0 , ⋯ , g n + 1 ) ⟼ g n + 1 , ∀ g i ∈ Hom ( ∧ n + 1 − i V ⊗ ∧ i V ′ , V ′ ) , \mathrm{pr}_n:L^{n}_{V,V'}\longrightarrow L^{n}_{V'},\qquad
(g_0,\cdots,g_{n+1})\longmapsto g_{n+1},\qquad
\forall g_i\in\operatorname{Hom}(\wedge^{n+1-i}V\otimes\wedge^{i}V',V'), pr n : L V , V ′ n ⟶ L V ′ n , ( g 0 , ⋯ , g n + 1 ) ⟼ g n + 1 , ∀ g i ∈ Hom ( ∧ n + 1 − i V ⊗ ∧ i V ′ , V ′ ) , with n ≥ 1 n\geq1 n ≥ 1 . In the degree 0 case, the inclusion is identity and the projection is 0.
By the fact that the three dgLas are nilpotent and Proposition Proposition 2.5 , the maps
ι : ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) ⟶ ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) , p r : ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) ⟶ ( L V ′ , [ ⋅ , ⋅ ] N R ) , \iota:(\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d})\longrightarrow(L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d),\qquad
\mathrm{pr}:(L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d)\longrightarrow(L_{V'},[\cdot,\cdot]_{\mathrm{NR}}), ι : ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) ⟶ ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) , pr : ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) ⟶ ( L V ′ , [ ⋅ , ⋅ ] NR ) ,
induce morphisms of Deligne groupoids
ι ∗ : Del ( L ~ V , V ′ ) ⟶ Del ( L V , V ′ ) , p r ∗ : Del ( L V , V ′ ) ⟶ Del ( L V ′ ) . \iota_*:\operatorname{Del}(\widetilde{L}_{V,V'})\longrightarrow\operatorname{Del}(L_{V,V'}),\qquad
\mathrm{pr}_*:\operatorname{Del}(L_{V,V'})\longrightarrow\operatorname{Del}(L_{V'}). ι ∗ : Del ( L V , V ′ ) ⟶ Del ( L V , V ′ ) , pr ∗ : Del ( L V , V ′ ) ⟶ Del ( L V ′ ) .
At the level of groupoids, we have the following results.
Proposition 5.3. (1) ι ∗ : Del ( L ~ V , V ′ ) ⟶ Del ( L V , V ′ ) \iota_*:\operatorname{Del}(\widetilde{L}_{V,V'})\longrightarrow\operatorname{Del}(L_{V,V'}) ι ∗ : Del ( L V , V ′ ) ⟶ Del ( L V , V ′ ) is an inclusion of groupoids,
(2) p r ∗ : Del ( L V , V ′ ) ⟶ Del ( L V ′ ) \mathrm{pr}_*:\operatorname{Del}(L_{V,V'})\longrightarrow\operatorname{Del}(L_{V'}) pr ∗ : Del ( L V , V ′ ) ⟶ Del ( L V ′ ) is a fibration of groupoids.
Proof. By the fact that ι \iota ι is injective on L ~ V , V ′ 0 \widetilde{L}^{0}_{V,V'} L V , V ′ 0 and L ~ V , V ′ 1 \widetilde{L}^{1}_{V,V'} L V , V ′ 1 , the first statement is obvious. Notice that Im p r ∗ ( Obj ( Del ( L V , V ′ ) ) ) = Obj ( Del ( L V ′ ) ) = MC ( L V ′ ) \operatorname{Im}\mathrm{pr}_*(\operatorname{Obj}(\operatorname{Del}(L_{V,V'})))=\operatorname{Obj}(\operatorname{Del}(L_{V'}))=\operatorname{MC}(L_{V'}) Im pr ∗ ( Obj ( Del ( L V , V ′ ))) = Obj ( Del ( L V ′ )) = MC ( L V ′ ) . Suppose π ∈ MC ( L V ′ ) \pi\in\operatorname{MC}(L_{V'}) π ∈ MC ( L V ′ ) , then there exists an object κ ∈ Obj ( Del ( L V , V ′ ) ) \kappa\in\operatorname{Obj}(\operatorname{Del}(L_{V,V'})) κ ∈ Obj ( Del ( L V , V ′ )) such that
p r ∗ ( κ ) = p r ( κ ) = π . \mathrm{pr}_*(\kappa)=\mathrm{pr}(\kappa)=\pi. pr ∗ ( κ ) = pr ( κ ) = π .
Let ψ ∈ St π ( Del ( L V ′ ) ) \psi\in\operatorname{St}_{\pi}(\operatorname{Del}(L_{V'})) ψ ∈ St π ( Del ( L V ′ )) be a morphism with source π \pi π and target π ′ \pi' π ′ . Then there exists a ∈ L V ′ 0 a \in L^0_{V'} a ∈ L V ′ 0 such that ψ = e a \psi=e^a ψ = e a . By the surjectivity of pr \operatorname{pr} pr on L V , V ′ 0 L^0_{V,V'} L V , V ′ 0 , there exists an element a ~ ∈ L V , V ′ 0 \tilde{a}\in L^0_{V,V'} a ~ ∈ L V , V ′ 0 such that pr ( a ~ ) = a \operatorname{pr}(\tilde{a})=a pr ( a ~ ) = a . Define ψ ~ = e a ~ \tilde{\psi}=e^{\tilde{a}} ψ ~ = e a ~ , we have
pr ∗ ( e a ~ ) = e pr ( a ~ ) = e a . \operatorname{pr}_*(e^{\tilde{a}})=e^{\operatorname{pr}(\tilde{a})}=e^a. pr ∗ ( e a ~ ) = e pr ( a ~ ) = e a .
Define κ ′ = e a ~ ⋆ κ \kappa'=e^{\tilde{a}}\star\kappa κ ′ = e a ~ ⋆ κ , we obtain that
pr ∗ ( κ ′ ) = pr ( e a ~ ⋆ κ ) = e pr ( a ~ ) ⋆ pr ( κ ) = e a ⋆ π = π ′ . \operatorname{pr}_*(\kappa')=\operatorname{pr}(e^{\tilde{a}}\star\kappa)=e^{\operatorname{pr}(\tilde{a})}\star\operatorname{pr}(\kappa)=e^a\star\pi=\pi'. pr ∗ ( κ ′ ) = pr ( e a ~ ⋆ κ ) = e pr ( a ~ ) ⋆ pr ( κ ) = e a ⋆ π = π ′ .
In other words, pr ∗ ( ψ ~ ) = ψ \operatorname{pr}_*(\tilde{\psi})=\psi pr ∗ ( ψ ~ ) = ψ . Hence, the morphism pr ∗ \operatorname{pr}_* pr ∗ is a fibration of groupoids. □ \square □
Lemma 5.4. If α ∈ MC ( L V , V ′ ) \alpha\in\operatorname{MC}(L_{V,V'}) α ∈ MC ( L V , V ′ ) , then ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ α = d ~ + [ α , ⋅ ] L V , V ′ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}^{\alpha}=\widetilde{d}+[\alpha,\cdot]_{L_{V,V'}}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d α = d + [ α , ⋅ ] L V , V ′ ) is a dgLa, denoted by L ~ V , V ′ α \widetilde{L}^{\alpha}_{V,V'} L V , V ′ α .
Proof. By Lemma 5.2 , ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) is an ideal of ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) (L_{V,V'},[\cdot,\cdot]_{L_{V,V'}},d) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d ) as dgLas. Hence, [ α , ξ ] L V , V ′ ∈ L ~ V , V ′ [\alpha,\xi]_{L_{V,V'}}\in\widetilde{L}_{V,V'} [ α , ξ ] L V , V ′ ∈ L V , V ′ for any α ∈ L V , V ′ \alpha\in L_{V,V'} α ∈ L V , V ′ , ξ ∈ L ~ V , V ′ \xi\in\widetilde{L}_{V,V'} ξ ∈ L V , V ′ . Thus, by Proposition 2.3 , d ~ α ( L ~ V , V ′ ) ⊂ L ~ V , V ′ \widetilde{d}^{\alpha}(\widetilde{L}_{V,V'})\subset\widetilde{L}_{V,V'} d α ( L V , V ′ ) ⊂ L V , V ′ , which means that ( L ~ V , V ′ , [ ⋅ , ⋅ ] L ~ V , V ′ , d ~ α = d ~ + [ α , ⋅ ] L V , V ′ ) (\widetilde{L}_{V,V'},[\cdot,\cdot]_{\widetilde{L}_{V,V'}},\widetilde{d}^{\alpha}=\widetilde{d}+[\alpha,\cdot]_{L_{V,V'}}) ( L V , V ′ , [ ⋅ , ⋅ ] L V , V ′ , d α = d + [ α , ⋅ ] L V , V ′ ) is a dgLa. □ \square □
Using the Deligne groupoid of the twisted dgLa introduced above, we characterize the preimage of a Lie algebra structure on V ′ V' V ′ under the map pr \operatorname{pr} pr .
Theorem 5.5. Let π ∈ MC ( L V ′ ) \pi\in\operatorname{MC}(L_{V'}) π ∈ MC ( L V ′ ) and α 0 = ( 0 , 0 , π ) ∈ Obj ( pr ∗ − 1 ( π ) ) ⊂ L V , V ′ 1 \alpha_0=(0,0,\pi)\in\operatorname{Obj}(\operatorname{pr}_*^{-1}(\pi))\subset L^1_{V,V'} α 0 = ( 0 , 0 , π ) ∈ Obj ( pr ∗ − 1 ( π )) ⊂ L V , V ′ 1 . Then the twisted dgLa L ~ V , V ′ α 0 \widetilde{L}^{\alpha_0}_{V,V'} L V , V ′ α 0 governs non-abelian extensions of g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) by ( V ′ , π ) (V',\pi) ( V ′ , π ) . Moreover, we have
pr ∗ − 1 ( π ) ≅ Del ( L ~ V , V ′ α 0 ) . \operatorname{pr}_*^{-1}(\pi)\cong\operatorname{Del}(\widetilde{L}^{\alpha_0}_{V,V'}). pr ∗ − 1 ( π ) ≅ Del ( L V , V ′ α 0 ) . In particular, when π = 0 ∈ MC ( L V ′ ) \pi=0\in\operatorname{MC}(L_{V'}) π = 0 ∈ MC ( L V ′ ) is the trivial Lie algebra structure on V ′ V' V ′ , then we have pr ∗ − 1 ( 0 ) = Del ( L ~ V , V ′ ) \operatorname{pr}_*^{-1}(0)=\operatorname{Del}(\widetilde{L}_{V,V'}) pr ∗ − 1 ( 0 ) = Del ( L V , V ′ ) .
Proof. Let π ∈ MC ( L V ′ ) \pi\in\operatorname{MC}(L_{V'}) π ∈ MC ( L V ′ ) and let α 0 = ( 0 , 0 , π ) ∈ MC ( L V , V ′ ) \alpha_0=(0,0,\pi)\in\operatorname{MC}(L_{V,V'}) α 0 = ( 0 , 0 , π ) ∈ MC ( L V , V ′ ) so that pr ( α 0 ) = π \operatorname{pr}(\alpha_0)=\pi pr ( α 0 ) = π . In other words, the set of objects
Obj ( pr ∗ − 1 ( π ) ) = { κ ∈ MC ( L V , V ′ ) : pr ( κ ) = π } \operatorname{Obj}(\operatorname{pr}_*^{-1}(\pi))=\{\kappa\in\operatorname{MC}(L_{V,V'}):\operatorname{pr}(\kappa)=\pi\} Obj ( pr ∗ − 1 ( π )) = { κ ∈ MC ( L V , V ′ ) : pr ( κ ) = π } is non-empty. By Lemma 5.4 , L ~ V , V ′ α 0 \widetilde{L}^{\alpha_0}_{V,V'} L V , V ′ α 0 is a dgLa. Since d ~ α 0 ( γ ) = [ μ + π , γ ] \widetilde{d}^{\alpha_0}(\gamma)=[\mu+\pi,\gamma] d α 0 ( γ ) = [ μ + π , γ ] for all γ ∈ L ~ V , V ′ \gamma\in\widetilde{L}_{V,V'} γ ∈ L V , V ′ , by Proposition 5.1 , the twisted dgLa L ~ V , V ′ α 0 \widetilde{L}^{\alpha_0}_{V,V'} L V , V ′ α 0 governs non-abelian extensions of g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) by ( V ′ , π ) (V',\pi) ( V ′ , π ) .
For any κ ∈ Obj ( pr ∗ − 1 ( π ) ) \kappa\in\operatorname{Obj}(\operatorname{pr}_*^{-1}(\pi)) κ ∈ Obj ( pr ∗ − 1 ( π )) , we have that pr ( κ − α 0 ) = 0 \operatorname{pr}(\kappa-\alpha_0)=0 pr ( κ − α 0 ) = 0 , so κ − α 0 ∈ L ~ V , V ′ \kappa-\alpha_0\in\widetilde{L}_{V,V'} κ − α 0 ∈ L V , V ′ . By the fact that κ \kappa κ and α 0 \alpha_0 α 0 are Maurer–Cartan elements, we have
d ~ α 0 ( κ − α 0 ) + 1 2 [ κ − α 0 , κ − α 0 ] L ~ V , V ′ = d ~ ( κ ) + 1 2 [ κ , κ ] − d ( α 0 ) − 1 2 [ α 0 , α 0 ] = 0. \widetilde{d}^{\alpha_0}(\kappa-\alpha_0)+\frac{1}{2}[\kappa-\alpha_0,\kappa-\alpha_0]_{\widetilde{L}_{V,V'}}=\widetilde{d}(\kappa)+\frac{1}{2}[\kappa,\kappa]-d(\alpha_0)-\frac{1}{2}[\alpha_0,\alpha_0]=0. d α 0 ( κ − α 0 ) + 2 1 [ κ − α 0 , κ − α 0 ] L V , V ′ = d ( κ ) + 2 1 [ κ , κ ] − d ( α 0 ) − 2 1 [ α 0 , α 0 ] = 0. Hence, we obtain that κ − α 0 ∈ MC ( L ~ V , V ′ α 0 ) \kappa-\alpha_0\in\operatorname{MC}(\widetilde{L}^{\alpha_0}_{V,V'}) κ − α 0 ∈ MC ( L V , V ′ α 0 ) , which means Obj ( pr ∗ − 1 ( π ) ) − α 0 ⊂ MC ( L ~ V , V ′ α 0 ) \operatorname{Obj}(\operatorname{pr}_*^{-1}(\pi))-\alpha_0\subset\operatorname{MC}(\widetilde{L}^{\alpha_0}_{V,V'}) Obj ( pr ∗ − 1 ( π )) − α 0 ⊂ MC ( L V , V ′ α 0 ) .
Let u ∈ MC ( L ~ V , V ′ α 0 ) u\in\operatorname{MC}(\widetilde{L}^{\alpha_0}_{V,V'}) u ∈ MC ( L V , V ′ α 0 ) , so pr ( u + α 0 ) = pr ( α 0 ) = π \operatorname{pr}(u+\alpha_0)=\operatorname{pr}(\alpha_0)=\pi pr ( u + α 0 ) = pr ( α 0 ) = π . Since
d ( u + α 0 ) + 1 2 [ u + α 0 , u + α 0 ] L V , V ′ = ( d ( u ) + 1 2 [ u , u ] + [ u , α 0 ] ) + ( d ( α 0 ) + 1 2 [ α 0 , α 0 ] ) = 0 , d(u+\alpha_0)+\frac{1}{2}[u+\alpha_0,u+\alpha_0]_{L_{V,V'}}=\left(d(u)+\frac{1}{2}[u,u]+[u,\alpha_0]\right)+\left(d(\alpha_0)+\frac{1}{2}[\alpha_0,\alpha_0]\right)=0, d ( u + α 0 ) + 2 1 [ u + α 0 , u + α 0 ] L V , V ′ = ( d ( u ) + 2 1 [ u , u ] + [ u , α 0 ] ) + ( d ( α 0 ) + 2 1 [ α 0 , α 0 ] ) = 0 , we obtain MC ( L ~ V , V ′ α 0 ) ⊂ Obj ( pr ∗ − 1 ( π ) ) − α 0 \operatorname{MC}(\widetilde{L}^{\alpha_0}_{V,V'})\subset\operatorname{Obj}(\operatorname{pr}_*^{-1}(\pi))-\alpha_0 MC ( L V , V ′ α 0 ) ⊂ Obj ( pr ∗ − 1 ( π )) − α 0 . It implies that on the object level, we have
Obj ( pr ∗ − 1 ( π ) ) ≅ Obj ( Del ( L ~ V , V ′ α 0 ) ) . \operatorname{Obj}(\operatorname{pr}_*^{-1}(\pi))\cong\operatorname{Obj}(\operatorname{Del}(\widetilde{L}^{\alpha_0}_{V,V'})). Obj ( pr ∗ − 1 ( π )) ≅ Obj ( Del ( L V , V ′ α 0 )) .
On the morphism level, let κ , κ ′ ∈ Obj ( pr − 1 ( π ) ) \kappa,\kappa' \in\operatorname{Obj}(\operatorname{pr}^{-1}(\pi)) κ , κ ′ ∈ Obj ( pr − 1 ( π )) and η ∈ L V , V ′ 0 = L ~ V , V ′ 0 \eta\in L^0_{V,V'}=\widetilde{L}^0_{V,V'} η ∈ L V , V ′ 0 = L V , V ′ 0 such that κ ′ = e η ⋆ κ \kappa'=e^\eta\star\kappa κ ′ = e η ⋆ κ . Denote u = κ − α 0 u=\kappa-\alpha_0 u = κ − α 0 , u ′ = κ ′ − α 0 u'=\kappa'-\alpha_0 u ′ = κ ′ − α 0 , then we have
u ′ = u + ∑ n = 0 ∞ ad n ( η ) ( n + 1 ) ! ( [ η , u ] − d ~ α 0 ( η ) ) = e η ⋆ u . u'=u+\sum_{n=0}^{\infty}\frac{\operatorname{ad}^{n}(\eta)}{(n+1)!}\left([\eta,u]-\widetilde{d}^{\alpha_0}(\eta)\right)=e^\eta\star u. u ′ = u + n = 0 ∑ ∞ ( n + 1 )! ad n ( η ) ( [ η , u ] − d α 0 ( η ) ) = e η ⋆ u .
Hence, if η ∈ Hom pr − 1 ( π ) ( κ , κ ′ ) \eta\in\operatorname{Hom}_{\operatorname{pr}^{-1}(\pi)}(\kappa,\kappa') η ∈ Hom pr − 1 ( π ) ( κ , κ ′ ) , then η ∈ Hom Del ( L ~ V , V ′ α 0 ) ( κ − α 0 , κ ′ − α 0 ) \eta\in\operatorname{Hom}_{\operatorname{Del}(\widetilde{L}^{\alpha_0}_{V,V'})}(\kappa-\alpha_0,\kappa'-\alpha_0) η ∈ Hom Del ( L V , V ′ α 0 ) ( κ − α 0 , κ ′ − α 0 ) .
Suppose that η ∈ Hom Del ( L ~ V , V ′ α 0 ) ( κ − α 0 , κ ′ − α 0 ) \eta\in\operatorname{Hom}_{\operatorname{Del}(\widetilde{L}^{\alpha_0}_{V,V'})}(\kappa-\alpha_0,\kappa'-\alpha_0) η ∈ Hom Del ( L V , V ′ α 0 ) ( κ − α 0 , κ ′ − α 0 ) , then we obtain
κ ′ = κ + ∑ n = 0 ∞ ad n ( η ) ( n + 1 ) ! ( [ η , κ ] − d ~ ( η ) ) = e η ⋆ κ , \kappa'=\kappa+\sum_{n=0}^{\infty}\frac{\operatorname{ad}^{n}(\eta)}{(n+1)!}\left([\eta,\kappa]-\widetilde{d}(\eta)\right)=e^\eta\star\kappa, κ ′ = κ + n = 0 ∑ ∞ ( n + 1 )! ad n ( η ) ( [ η , κ ] − d ( η ) ) = e η ⋆ κ ,
which implies that η ∈ Hom pr − 1 ( π ) ( κ , κ ′ ) \eta\in\operatorname{Hom}_{\operatorname{pr}^{-1}(\pi)}(\kappa,\kappa') η ∈ Hom pr − 1 ( π ) ( κ , κ ′ ) . Thus, pr − 1 ( π ) ≅ Del ( L ~ V , V ′ α 0 ) \operatorname{pr}^{-1}(\pi)\cong\operatorname{Del}(\widetilde{L}^{\alpha_0}_{V,V'}) pr − 1 ( π ) ≅ Del ( L V , V ′ α 0 ) .
In particular, when taking π = 0 ∈ MC ( L V ′ ) \pi=0\in\operatorname{MC}(L_{V'}) π = 0 ∈ MC ( L V ′ ) , the trivial Lie algebra structure on V ′ V' V ′ , the isomorphism of the groupoids pr − 1 ( 0 ) \operatorname{pr}^{-1}(0) pr − 1 ( 0 ) and Del ( L ~ V , V ′ ) \operatorname{Del}(\widetilde{L}_{V,V'}) Del ( L V , V ′ ) in the above discussion reduces to the identity.
Corollary 5.6. The Deligne groupoid of L V , V ′ L_{V,V'} L V , V ′ is the disjoint union of the Deligne groupoids of the twisted dgLas L ~ V , V ′ ( 0 , 0 , π ) \widetilde{L}^{(0,0,\pi)}_{V,V'} L V , V ′ ( 0 , 0 , π ) indexed by the Lie algebra structures π \pi π on V ′ V' V ′ , i.e.
Del ( L V , V ′ ) ≅ ⨆ π ∈ MC ( L V ′ ) ( k ) Del ( L ~ V , V ′ ( 0 , 0 , π ) ) . (21) \operatorname{Del}(L_{V,V'})\cong\bigsqcup_{\pi\in\operatorname{MC}(L_{V'})(k)}\operatorname{Del}\left(\widetilde{L}^{(0,0,\pi)}_{V,V'}\right). \tag*{(21)} Del ( L V , V ′ ) ≅ π ∈ MC ( L V ′ ) ( k ) ⨆ Del ( L V , V ′ ( 0 , 0 , π ) ) . (21)
Proof. Since Del ( L V ′ ) \operatorname{Del}(L_{V'}) Del ( L V ′ ) is totally disconnected [12 p. 218] , pr \operatorname{pr} pr is surjective on MC ( L V , V ′ ) \operatorname{MC}(L_{V,V'}) MC ( L V , V ′ ) . By Theorem 5.5 , we obtain the desired disjoint union decomposition.
Remark 5.7. By the fact that π 0 ( Del ( L ~ V , V ′ ( 0 , 0 , π ) ) ) ≅ H n a b 2 ( V , μ ; V ′ , π ) \pi_0(\operatorname{Del}(\widetilde{L}^{(0,0,\pi)}_{V,V'}))\cong H^2_{\mathrm{nab}}(V,\mu;V',\pi) π 0 ( Del ( L V , V ′ ( 0 , 0 , π ) )) ≅ H nab 2 ( V , μ ; V ′ , π ) for all π ∈ MC ( L V ′ ) \pi\in\operatorname{MC}(L_{V'}) π ∈ MC ( L V ′ ) [16 Proposition 2.2] , taking the set of connected components of the Deligne groupoids in equation (21) , we recover Theorem 3.12 .
Obstructions to lifting and extending deformations
In this section, we study infinitesimal deformations and obstructions for global extensions. We construct an obstruction class, the vanishing of which captures the existence of a lift of an infinitesimal deformation of the Lie algebra structure on V ′ V' V ′ to an infinitesimal deformation of the global extension structure on V ′ ′ V'' V ′′ .
Extending a finite-order deformation
Let ( V ′ ′ , p , g ) (V'',p,\mathfrak{g}) ( V ′′ , p , g ) be a global extension datum, where g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) , and fix a global extension structure μ ′ ′ \mu'' μ ′′ . Let V ′ = ker p V'=\ker p V ′ = ker p .
A deformation of μ ′ ′ \mu'' μ ′′ over k [ t ] / ( t n + 1 ) k[t]/(t^{n+1}) k [ t ] / ( t n + 1 ) is called an n n n -th order deformation. First-order deformations are also called infinitesimal deformations. We write D = k [ t ] / ( t 2 ) D=k[t]/(t^2) D = k [ t ] / ( t 2 ) for the k k k -algebra of dual numbers.
Choose a linear section s : V → V ′ ′ s:V\to V'' s : V → V ′′ of p p p , and let ( π V ′ s , ϱ s , ω s ) (\pi^s_{V'},\varrho^s,\omega^s) ( π V ′ s , ϱ s , ω s ) be the corresponding global 2 2 2 -cocycle defined in (11) . Recall the graded spaces L V ′ L_{V'} L V ′ , L V , V ′ L_{V,V'} L V , V ′ , L ~ V , V ′ \widetilde{L}_{V,V'} L V , V ′ and L ~ V , V ′ \widetilde{L}_{V,V'} L V , V ′ from Example 2.6 , Proposition 4.2 and Proposition 5.1 . Since π V ′ s \pi^s_{V'} π V ′ s is a Maurer–Cartan element of L V ′ L_{V'} L V ′ , π V ′ s + ϱ s + ω s \pi^s_{V'}+\varrho^s+\omega^s π V ′ s + ϱ s + ω s is a Maurer–Cartan element of L V , V ′ L_{V,V'} L V , V ′ , and L ~ V , V ′ \widetilde{L}_{V,V'} L V , V ′ is an ideal of L V , V ′ L_{V,V'} L V , V ′ . These graded vector spaces give rise to four cochain complexes ( L V , V ′ , δ V , V ′ ) (L_{V,V'},\delta^{V,V'}) ( L V , V ′ , δ V , V ′ ) , ( L V , V ′ u n , δ V , V ′ ) (L^{\mathrm{un}}_{V,V'},\delta^{V,V'}) ( L V , V ′ un , δ V , V ′ ) , ( L ~ V , V ′ , δ ~ V , V ′ ) (\widetilde{L}_{V,V'},\widetilde{\delta}^{V,V'}) ( L V , V ′ , δ V , V ′ ) and ( L V ′ , δ V ′ ) (L_{V'},\delta^{V'}) ( L V ′ , δ V ′ ) , where
δ V , V ′ = [ μ + π V ′ s + ϱ s + ω s , ⋅ ] , δ ~ V , V ′ = [ μ + π V ′ s + ϱ s + ω s , ⋅ ] = δ V , V ′ ∣ L ~ V , V ′ , δ V ′ = [ π V ′ s , ⋅ ] . \delta^{V,V'}=[\mu+\pi^{s}_{V'}+\varrho^{s}+\omega^{s},\mathord{\cdot}],\qquad\widetilde{\delta}^{V,V'}=[\mu+\pi^{s}_{V'}+\varrho^{s}+\omega^{s},\mathord{\cdot}]=\left.\delta^{V,V'}\right|_{\widetilde{L}_{V,V'}},\qquad\delta^{V'}=[\pi^{s}_{V'},\mathord{\cdot}]. δ V , V ′ = [ μ + π V ′ s + ϱ s + ω s , ⋅ ] , δ V , V ′ = [ μ + π V ′ s + ϱ s + ω s , ⋅ ] = δ V , V ′ L V , V ′ , δ V ′ = [ π V ′ s , ⋅ ] .
Denote by H i ( L V , V ′ ) δ H^{i}(L_{V,V'})^{\delta} H i ( L V , V ′ ) δ , H i ( L V , V ′ u n ) δ H^{i}(L^{\mathrm{un}}_{V,V'})^{\delta} H i ( L V , V ′ un ) δ , H i ( L ~ V , V ′ ) δ ~ H^{i}(\widetilde{L}_{V,V'})^{\widetilde{\delta}} H i ( L V , V ′ ) δ and H i ( L V ′ ) δ H^{i}(L_{V'})^{\delta} H i ( L V ′ ) δ the i i i -th cohomology of the corresponding complexes.
Proposition 6.2. There is a short exact sequence of the cochain complexes:
0 ⟶ ( L ~ V , V ′ , δ ~ V , V ′ ) → ι ( L V , V ′ , δ V , V ′ ) → p r ( L V ′ , δ V ′ ) ⟶ 0 , 0\longrightarrow(\widetilde{L}_{V,V'},\widetilde{\delta}^{V,V'})\xrightarrow{\iota}(L_{V,V'},\delta^{V,V'})\xrightarrow{\mathrm{pr}}(L_{V'},\delta^{V'})\longrightarrow0, 0 ⟶ ( L V , V ′ , δ V , V ′ ) ι ( L V , V ′ , δ V , V ′ ) pr ( L V ′ , δ V ′ ) ⟶ 0 , where ι ( ξ ) = ( ξ , 0 ) \iota(\xi)=(\xi,0) ι ( ξ ) = ( ξ , 0 ) and p ( ξ , f ) = f \mathrm{p}(\xi,f)=f p ( ξ , f ) = f for all f ∈ L V ′ f\in L_{V'} f ∈ L V ′ and ξ ∈ L ~ V , V ′ \xi\in\widetilde{L}_{V,V'} ξ ∈ L V , V ′ . Consequently, there is a long exact sequence of the cohomology groups:
0 ⟶ H 0 ( L ~ V , V ′ ) δ ~ ⟶ H 0 ( L V , V ′ ) δ ⟶ 0 ⟶ H 1 ( L ~ V , V ′ ) δ ~ 0\longrightarrow H^{0}(\widetilde{L}_{V,V'})^{\widetilde{\delta}}\longrightarrow H^{0}(L_{V,V'})^{\delta}\longrightarrow0\longrightarrow H^{1}(\widetilde{L}_{V,V'})^{\widetilde{\delta}} 0 ⟶ H 0 ( L V , V ′ ) δ ⟶ H 0 ( L V , V ′ ) δ ⟶ 0 ⟶ H 1 ( L V , V ′ ) δ ⟶ H 1 ( L V , V ′ ) δ ⟶ H 1 ( L V ′ ) δ ⟶ H 2 ( L ~ V , V ′ ) δ ~ ⟶ ⋯ . \longrightarrow H^{1}(L_{V,V'})^{\delta}\longrightarrow H^{1}(L_{V'})^{\delta}\longrightarrow H^{2}(\widetilde{L}_{V,V'})^{\widetilde{\delta}}\longrightarrow\cdots. ⟶ H 1 ( L V , V ′ ) δ ⟶ H 1 ( L V ′ ) δ ⟶ H 2 ( L V , V ′ ) δ ⟶ ⋯ . Proof. The short exact sequence follows from a straightforward computation. □
In fact, various deformations information about global extensions can be read from the cohomology of the dgLa, as in the classical setting. A typical example is that the first cohomology can be used to characterize infinitesimal of μ ′ ′ \mu'' μ ′′ .
Proposition 6.3. Equivalence (resp. unframed equivalence) classes of infinitesimal deformations of the global extension structure μ ′ ′ \mu'' μ ′′ are classified by the first cohomology of ( L V , V ′ , δ V , V ′ ) (L_{V,V'},\delta^{V,V'}) ( L V , V ′ , δ V , V ′ ) (resp. ( L V , V ′ u n , δ V , V ′ ) (L^{\mathrm{un}}_{V,V'},\delta^{V,V'}) ( L V , V ′ un , δ V , V ′ ) ). In other words, there are bijections
D μ ′ ′ ( k [ t ] / ( t 2 ) ) ≅ H 1 ( L V , V ′ ) δ , D μ ′ ′ u n ( k [ t ] / ( t 2 ) ) ≅ H 1 ( L V , V ′ u n ) δ , D_{\mu''}\bigl(k[t]/(t^{2})\bigr)\cong H^{1}(L_{V,V'})^{\delta},\qquad D^{\mathrm{un}}_{\mu''}\bigl(k[t]/(t^{2})\bigr)\cong H^{1}(L^{\mathrm{un}}_{V,V'})^{\delta}, D μ ′′ ( k [ t ] / ( t 2 ) ) ≅ H 1 ( L V , V ′ ) δ , D μ ′′ un ( k [ t ] / ( t 2 ) ) ≅ H 1 ( L V , V ′ un ) δ , where D μ ′ ′ = π 0 D e f E p , [ μ ′ ′ ] D_{\mu''}=\pi_{0}\mathop{\mathrm{Def}}\nolimits_{E_{p,[\mu'']}} D μ ′′ = π 0 Def E p , [ μ ′′ ] and D μ ′ ′ u n = π 0 D e f E p , [ μ ′ ′ ] u n D^{\mathrm{un}}_{\mu''}=\pi_{0}\mathop{\mathrm{Def}}\nolimits_{E^{\mathrm{un}}_{p,[\mu'']}} D μ ′′ un = π 0 Def E p , [ μ ′′ ] un are set-valued deformation functors induced by D e f E p , [ μ ′ ′ ] \mathop{\mathrm{Def}}\nolimits_{E_{p,[\mu'']}} Def E p , [ μ ′′ ] and D e f E p , [ μ ′ ′ ] u n \mathop{\mathrm{Def}}\nolimits_{E^{\mathrm{un}}_{p,[\mu'']}} Def E p , [ μ ′′ ] un in § 4 respectively.
Proof. This follows from Theorem 4.8 , Theorem 4.10 and [32 Lemma 6.4.1] . □
For a fixed algebraic structure, the question of whether and under what conditions an n n n -th order deformation can be extended to an ( n + 1 ) (n+1) ( n + 1 ) -th order deformation has also been extensively studied in the literature. These problems can be transformed into a cohomological problem, which is now known as the obstruction theory.
Proposition 6.4. Let μ ′ ′ \mu'' μ ′′ be a fixed global extension structure on V ′ ′ V'' V ′′ , and μ n ′ ′ \mu''_{n} μ n ′′ an n n n -th order deformation of μ ′ ′ \mu'' μ ′′ . Then μ n ′ ′ \mu''_{n} μ n ′′ is extensible to an ( n + 1 ) (n+1) ( n + 1 ) -th order deformation if and only if the obstruction class
Ob n ( μ n ′ ′ ) : = 1 2 [ ∑ i + j = n + 1 [ η i , η j ] ] ∈ H 2 ( L V , V ′ ) δ \operatorname{Ob}_{n}(\mu''_{n}):=\frac{1}{2}\left[\sum_{i+j=n+1}[\eta_{i},\eta_{j}]\right]\in H^{2}(L_{V,V'})^{\delta} Ob n ( μ n ′′ ) := 2 1 [ i + j = n + 1 ∑ [ η i , η j ] ] ∈ H 2 ( L V , V ′ ) δ vanishes. Here, μ ′ ′ = μ + ∑ i = 1 n t i η i \mu''=\mu+\sum_{i=1}^{n}t^{i}\eta_{i} μ ′′ = μ + ∑ i = 1 n t i η i , and η i = π V ′ , i + ϱ i + ω i \eta_{i}=\pi_{V',i}+\varrho_{i}+\omega_{i} η i = π V ′ , i + ϱ i + ω i are the coefficients of t i t^{i} t i in the expansion of the corresponding global 2 2 2 -cocycle of μ n ′ ′ − μ ′ ′ \mu''_{n}-\mu'' μ n ′′ − μ ′′ as a polynomial in t t t .
Proof. This follows from Theorem 4.8 and [31 ] [83]. □
Lifting a deformation of the Lie bracket on the kernel
Fix a linear section s s s of p p p . For convenience, we will omit the superscript s s s if there is no ambiguity. By Corollary 4.6, an infinitesimal deformation μ D ′ ′ \mu_D^{\prime\prime} μ D ′′ of μ ′ ′ \mu^{\prime\prime} μ ′′ corresponds to a global 2 2 2 -cocycle
( π V ′ D , ϱ D , ω D ) = ( π V ′ + t π V ′ ′ , ϱ + t ϱ ′ , ω + t ω ′ ) , (\pi_{V'}^D,\varrho^D,\omega^D)=(\pi_{V'}+t\pi_{V'}',\varrho+t\varrho',\omega+t\omega'), ( π V ′ D , ϱ D , ω D ) = ( π V ′ + t π V ′ ′ , ϱ + t ϱ ′ , ω + t ω ′ ) ,
where
π V ′ ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , ϱ ′ ∈ Hom ( V ⊗ V ′ , V ′ ) , ω ′ ∈ Hom ( ∧ 2 V , V ′ ) . \pi_{V'}'\in\operatorname{Hom}(\wedge^2V',V'),\qquad\varrho'\in\operatorname{Hom}(V\otimes V',V'),\qquad\omega'\in\operatorname{Hom}(\wedge^2V,V'). π V ′ ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , ϱ ′ ∈ Hom ( V ⊗ V ′ , V ′ ) , ω ′ ∈ Hom ( ∧ 2 V , V ′ ) .
In particular, π V ′ D \pi_{V'}^D π V ′ D is an infinitesimal deformation of the Lie bracket π V ′ \pi_{V'} π V ′ on the kernel.
Conversely, given such a deformation π V ′ D = π V ′ + t π V ′ ′ \pi_{V'}^D=\pi_{V'}+t\pi_{V'}' π V ′ D = π V ′ + t π V ′ ′ , we ask whether it extends to an infinitesimal deformation of the global cocycle ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) , and determine the obstruction to finding suitable ϱ ′ \varrho' ϱ ′ and ω ′ \omega' ω ′ such that
( π V ′ D , ϱ D , ω D ) : = ( π V ′ + t π V ′ ′ , ϱ + t ϱ ′ , ω + t ω ′ ) (\pi_{V'}^D,\varrho^D,\omega^D):=(\pi_{V'}+t\pi_{V'}',\varrho+t\varrho',\omega+t\omega') ( π V ′ D , ϱ D , ω D ) := ( π V ′ + t π V ′ ′ , ϱ + t ϱ ′ , ω + t ω ′ )
forms an infinitesimal deformation of the given global 2 2 2 -cocycle ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) .
Recall that if π V ′ + t π V ′ ′ \pi_{V'}+t\pi_{V'}' π V ′ + t π V ′ ′ is an infinitesimal deformation of π V ′ \pi_{V'} π V ′ , according to [35 §2] , π V ′ ′ \pi_{V'}' π V ′ ′ is a cocycle in the corresponding Chevalley–Eilenberg complex. Thus, δ V ′ ( π V ′ ′ ) = 0 \delta^{V'}(\pi_{V'}')=0 δ V ′ ( π V ′ ′ ) = 0 .
Proposition 6.5. There exists a pair ( ϱ ′ , ω ′ ) ∈ Hom ( V ⊗ V ′ , V ′ ) × Hom ( ∧ 2 V , V ′ ) (\varrho',\omega')\in\operatorname{Hom}(V\otimes V',V')\times\operatorname{Hom}(\wedge^2V,V') ( ϱ ′ , ω ′ ) ∈ Hom ( V ⊗ V ′ , V ′ ) × Hom ( ∧ 2 V , V ′ ) such that
( π V ′ D , ϱ D , ω D ) : = ( π V ′ + t π V ′ ′ , ϱ + t ϱ ′ , ω + t ω ′ ) (\pi_{V'}^D,\varrho^D,\omega^D):=(\pi_{V'}+t\pi_{V'}',\varrho+t\varrho',\omega+t\omega') ( π V ′ D , ϱ D , ω D ) := ( π V ′ + t π V ′ ′ , ϱ + t ϱ ′ , ω + t ω ′ ) is an infinitesimal deformation of the given global 2 2 2 -cocycle ( π V ′ , ϱ , ω ) (\pi_{V'},\varrho,\omega) ( π V ′ , ϱ , ω ) if and only if the cohomology class Ob π V ′ ( π V ′ , ϱ , ω ) ∈ H 2 ( L ~ V , V ′ ) δ \operatorname{Ob}_{\pi_{V'}}(\pi_{V'},\varrho,\omega)\in H^2(\widetilde{L}_{V,V'})^\delta Ob π V ′ ( π V ′ , ϱ , ω ) ∈ H 2 ( L V , V ′ ) δ , defined by
Ob π V ′ ( π V ′ , ϱ , ω ) = [ [ ϱ + ω , π V ′ ′ ] ] , \operatorname{Ob}_{\pi_{V'}}(\pi_{V'},\varrho,\omega)=\left[[\varrho+\omega,\pi_{V'}']\right], Ob π V ′ ( π V ′ , ϱ , ω ) = [ [ ϱ + ω , π V ′ ′ ] ] , vanishes.
Proof. By (13) , we know that [ μ , π V ′ ′ ] = 0 [\mu,\pi_{V'}']=0 [ μ , π V ′ ′ ] = 0 . Since δ V ′ ( π V ′ ′ ) = [ π V ′ , π V ′ ′ ] = 0 \delta^{V'}(\pi_{V'}')=[\pi_{V'},\pi_{V'}']=0 δ V ′ ( π V ′ ′ ) = [ π V ′ , π V ′ ′ ] = 0 , we have [ ϱ + ω , π V ′ ′ ] = δ ~ V , V ′ ( π V ′ ′ ) [\varrho+\omega,\pi_{V'}']=\widetilde{\delta}^{V,V'}(\pi_{V'}') [ ϱ + ω , π V ′ ′ ] = δ V , V ′ ( π V ′ ′ ) . Thus, there is a cohomology class Ob π V ′ ( π V ′ , ϱ , ω ) ∈ H 2 ( L ~ V , V ′ ) δ \operatorname{Ob}_{\pi_{V'}}(\pi_{V'},\varrho,\omega)\in H^2(\widetilde{L}_{V,V'})^\delta Ob π V ′ ( π V ′ , ϱ , ω ) ∈ H 2 ( L V , V ′ ) δ .
If Ob π V ′ ( π V ′ , ϱ , ω ) = 0 \operatorname{Ob}_{\pi_{V'}}(\pi_{V'},\varrho,\omega)=0 Ob π V ′ ( π V ′ , ϱ , ω ) = 0 , then there exist ϱ ′ ∈ Hom ( V ⊗ V ′ , V ′ ) \varrho'\in\operatorname{Hom}(V\otimes V',V') ϱ ′ ∈ Hom ( V ⊗ V ′ , V ′ ) and ω ′ ∈ Hom ( ∧ 2 V , V ′ ) \omega'\in\operatorname{Hom}(\wedge^2V,V') ω ′ ∈ Hom ( ∧ 2 V , V ′ ) such that
[ ϱ + ω , π V ′ ′ ] = δ ~ V , V ′ ( − ϱ ′ − ω ′ ) , [\varrho+\omega,\pi_{V'}']=\widetilde{\delta}^{V,V'}(-\varrho'-\omega'), [ ϱ + ω , π V ′ ′ ] = δ V , V ′ ( − ϱ ′ − ω ′ ) ,
which is equivalent to the following equations
[ ϱ , π V ′ ′ ] = − [ π V ′ , ϱ ′ ] , [ ω , π V ′ ′ ] = − [ μ , ϱ ′ ] − [ ϱ , ϱ ′ ] − [ π V ′ , ω ′ ] , [ μ , ω ′ ] = − [ ϱ , ω ′ ] − [ ϱ ′ , ω ] . [\varrho,\pi_{V'}']=-[\pi_{V'},\varrho'],\qquad[\omega,\pi_{V'}']=-[\mu,\varrho']-[\varrho,\varrho']-[\pi_{V'},\omega'],\qquad[\mu,\omega']=-[\varrho,\omega']-[\varrho',\omega]. [ ϱ , π V ′ ′ ] = − [ π V ′ , ϱ ′ ] , [ ω , π V ′ ′ ] = − [ μ , ϱ ′ ] − [ ϱ , ϱ ′ ] − [ π V ′ , ω ′ ] , [ μ , ω ′ ] = − [ ϱ , ω ′ ] − [ ϱ ′ , ω ] .
Therefore, ( ω + t ω ′ , ϱ + t ϱ ′ , π V ′ + t π V ′ ′ ) (\omega+t\omega',\varrho+t\varrho',\pi_{V'}+t\pi_{V'}') ( ω + t ω ′ , ϱ + t ϱ ′ , π V ′ + t π V ′ ′ ) satisfies the equations (4) , (5) , and (6) . Define ϱ D = ϱ + t ϱ ′ \varrho^D=\varrho+t\varrho' ϱ D = ϱ + t ϱ ′ and ω D = ω + t ω ′ \omega^D=\omega+t\omega' ω D = ω + t ω ′ . Since
ϱ D ( x ⊗ 1 ) ( u ⊗ 1 ) = ϱ ( x ) u ⊗ 1 + ϱ ′ ( x ) u ⊗ t , ω D ( x ⊗ 1 , y ⊗ 1 ) = ω ( x , y ) ⊗ 1 + ω ′ ( x , y ) ⊗ t , \varrho^D(x\otimes1)(u\otimes1)=\varrho(x)u\otimes1+\varrho'(x)u\otimes t,\qquad\omega^D(x\otimes1,y\otimes1)=\omega(x,y)\otimes1+\omega'(x,y)\otimes t, ϱ D ( x ⊗ 1 ) ( u ⊗ 1 ) = ϱ ( x ) u ⊗ 1 + ϱ ′ ( x ) u ⊗ t , ω D ( x ⊗ 1 , y ⊗ 1 ) = ω ( x , y ) ⊗ 1 + ω ′ ( x , y ) ⊗ t ,
it follows that ( π V ′ D , ϱ D , ω D ) (\pi_{V'}^D,\varrho^D,\omega^D) ( π V ′ D , ϱ D , ω D ) is a global 2 2 2 -cocycle, which induces an infinitesimal deformation μ D ′ \mu_D' μ D ′ of μ ′ ′ \mu'' μ ′′ .
If ( π V ′ D , ϱ D , ω D ) (\pi_{V'}^D,\varrho^D,\omega^D) ( π V ′ D , ϱ D , ω D ) is a global 2 2 2 -cocycle, then there exists
( ϱ ′ , ω ′ ) ∈ Hom ( V ⊗ V ′ , V ′ ) × Hom ( ∧ 2 V , V ′ ) (\varrho',\omega')\in\operatorname{Hom}(V\otimes V',V')\times\operatorname{Hom}(\wedge^2V,V') ( ϱ ′ , ω ′ ) ∈ Hom ( V ⊗ V ′ , V ′ ) × Hom ( ∧ 2 V , V ′ )
such that
ϱ D = ϱ + t ϱ ′ , ω D = ω + t ω ′ . \varrho^D=\varrho+t\varrho',\qquad\omega^D=\omega+t\omega'. ϱ D = ϱ + t ϱ ′ , ω D = ω + t ω ′ .
The global 2 2 2 -cocycle conditions imply the following identities:
[ ϱ , π V ′ ′ ] = − [ π V ′ , ϱ ′ ] , [\varrho,\pi_{V'}']=-[\pi_{V'},\varrho'], [ ϱ , π V ′ ′ ] = − [ π V ′ , ϱ ′ ] ,
[ ϱ , π V ′ ′ ] = − [ μ , ϱ ′ ] − [ ϱ , ϱ ′ ] − [ π V ′ , ω ′ ] [\varrho,\pi'_{V'}]=-[\mu,\varrho']-[\varrho,\varrho']-[\pi_{V'},\omega'] [ ϱ , π V ′ ′ ] = − [ μ , ϱ ′ ] − [ ϱ , ϱ ′ ] − [ π V ′ , ω ′ ]
[ μ , ω ′ ] = − [ ϱ , ω ′ ] − [ ϱ ′ , ω ] . [\mu,\omega']=-[\varrho,\omega']-[\varrho',\omega]. [ μ , ω ′ ] = − [ ϱ , ω ′ ] − [ ϱ ′ , ω ] .
Consequently, these relations can be unified as
[ ϱ + ω , π V ′ ′ ] = δ ~ V , V ′ ( − ϱ ′ − ω ′ ) , [\varrho+\omega,\pi'_{V'}]=\widetilde{\delta}_{V,V'}(-\varrho'-\omega'), [ ϱ + ω , π V ′ ′ ] = δ V , V ′ ( − ϱ ′ − ω ′ ) ,
which implies that Ob π V ′ ′ ( π V ′ , ϱ , ω ) = 0 \operatorname{Ob}_{\pi'_{V'}}(\pi_{V'},\varrho,\omega)=0 Ob π V ′ ′ ( π V ′ , ϱ , ω ) = 0 .
Example 6.6. We explicate the d g L a dgLa d g L a description for Example 4.11. We assume that g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) has even-dimension 2 n 2n 2 n . Let { y 1 , ⋯ , y n , z 1 , ⋯ , z n } \{y_1,\cdots,y_n,z_1,\cdots,z_n\} { y 1 , ⋯ , y n , z 1 , ⋯ , z n } be a basis of V V V . Define ϱ \varrho ϱ and ω \omega ω by
ϱ = 0 , ω ( y i , z j ) = δ i j , ω ( y k , y l ) = ω ( z p , z q ) = 0. \varrho=0,\qquad\omega(y_i,z_j)=\delta_{ij},\qquad\omega(y_k,y_l)=\omega(z_p,z_q)=0. ϱ = 0 , ω ( y i , z j ) = δ ij , ω ( y k , y l ) = ω ( z p , z q ) = 0.
Then the global extension ( k ⊕ V , [ ⋅ , ⋅ ] ω ) (\mathbb{k}\oplus V,[\cdot,\cdot]_\omega) ( k ⊕ V , [ ⋅ , ⋅ ] ω ) is the ( 2 n + 1 ) (2n+1) ( 2 n + 1 ) -dimensional Heisenberg Lie algebra, where
[ ( r 1 , x 1 ) , ( r 2 , x 2 ) ] ω = ( ω ( x 1 , x 2 ) , 0 ) , ∀ ( r i , x i ) ∈ k ⊕ V , i = 1 , 2. [(r_1,x_1),(r_2,x_2)]_\omega=(\omega(x_1,x_2),0),\qquad\forall(r_i,x_i)\in\mathbb{k}\oplus V,\quad i=1,2. [( r 1 , x 1 ) , ( r 2 , x 2 ) ] ω = ( ω ( x 1 , x 2 ) , 0 ) , ∀ ( r i , x i ) ∈ k ⊕ V , i = 1 , 2.
By Theorem 4.8 , the d g L a dgLa d g L a
L V , k ϱ = ( L V , k , [ ⋅ , ⋅ ] L V , k , d = [ ω , ⋅ ] L V , k ) L^{{\varrho}}_{V,\mathbb{k}}=\left(L_{V,\mathbb{k}},[\cdot,\cdot]_{L_{V,\mathbb{k}}},d=[\omega,\cdot]_{L_{V,\mathbb{k}}}\right) L V , k ϱ = ( L V , k , [ ⋅ , ⋅ ] L V , k , d = [ ω , ⋅ ] L V , k )
controls the deformation of ( k ⊕ V , [ ⋅ , ⋅ ] ω , V , p ) (\mathbb{k}\oplus V,[\cdot,\cdot]_\omega,V,p) ( k ⊕ V , [ ⋅ , ⋅ ] ω , V , p ) , where
L V , k i = { 0 , i ≤ − 1 ; Hom ( V , k ) , i = 0 ; Hom ( ⋀ i + 1 V , k ) ⊕ Hom ( ⋀ i V ⊗ k , k ) , i ≥ 1. L^i_{V,\mathbb{k}}=
\begin{cases}
0, & i\leq-1;\\
\operatorname{Hom}(V,\mathbb{k}), & i=0;\\
\operatorname{Hom}(\bigwedge^{i+1}V,\mathbb{k})\oplus\operatorname{Hom}(\bigwedge^iV\otimes\mathbb{k},\mathbb{k}), & i\geq1.
\end{cases} L V , k i = ⎩ ⎨ ⎧ 0 , Hom ( V , k ) , Hom ( ⋀ i + 1 V , k ) ⊕ Hom ( ⋀ i V ⊗ k , k ) , i ≤ − 1 ; i = 0 ; i ≥ 1.
By Proposition 6.3 , the set of equivalence classes of infinitesimal deformations of ( k ⊕ V , [ ⋅ , ⋅ ] ω , V , p ) (\mathbb{k}\oplus V,[\cdot,\cdot]_\omega,V,p) ( k ⊕ V , [ ⋅ , ⋅ ] ω , V , p ) is
D [ ⋅ , ⋅ ] ω ( k [ t ] / ( t 2 ) ) ≅ H 1 ( L V , k ϱ ) , \mathcal{D}_{[\cdot,\cdot]_\omega}(\mathbb{k}[t]/(t^2))\cong H^1(L^{{\varrho}}_{V,\mathbb{k}}), D [ ⋅ , ⋅ ] ω ( k [ t ] / ( t 2 )) ≅ H 1 ( L V , k ϱ ) ,
where D [ ⋅ , ⋅ ] ω \mathcal{D}_{[\cdot,\cdot]_\omega} D [ ⋅ , ⋅ ] ω is the set-valued deformation functor induced by D e f p , [ [ ⋅ , ⋅ ] ω ] \mathcal{D}\mathrm{ef}_{\mathfrak{p},[[\cdot,\cdot]_\omega]} D ef p , [[ ⋅ , ⋅ ] ω ] . By (13), we have
d f = [ ω , f ] L V , k = 0 , ∀ f ∈ Hom ( V , k ) , and [ ω , Hom ( ⋀ 2 V , k ) ] L V , k = 0. df=[\omega,f]_{L_{V,\mathbb{k}}}=0,\qquad\forall f\in\operatorname{Hom}(V,\mathbb{k}),\qquad\text{and}\qquad[\omega,\operatorname{Hom}(\bigwedge^2V,\mathbb{k})]_{L_{V,\mathbb{k}}}=0. df = [ ω , f ] L V , k = 0 , ∀ f ∈ Hom ( V , k ) , and [ ω , Hom ( ⋀ 2 V , k ) ] L V , k = 0.
For ϖ ∈ Hom ( V ⊗ k , k ) ≅ V ∗ \varpi\in\operatorname{Hom}(V\otimes\mathbb{k},\mathbb{k})\cong V^* ϖ ∈ Hom ( V ⊗ k , k ) ≅ V ∗ , since the differential is 0 0 0 on Hom ( ⋀ 2 V , k ) \operatorname{Hom}(\bigwedge^2V,\mathbb{k}) Hom ( ⋀ 2 V , k ) , a direct computation shows that
H 1 ( L V , k ϱ ) ≅ ⋀ 2 V ∗ ⊕ Ker ( V ∗ → ϖ ↦ ϖ ∧ ω ⋀ 3 V ∗ ) . H^1(L^{{\varrho}}_{V,\mathbb{k}})\cong\bigwedge^2V^*\oplus\operatorname{Ker}\left(V^*\xrightarrow{\varpi\mapsto\varpi\wedge\omega}\bigwedge^3V^*\right). H 1 ( L V , k ϱ ) ≅ ⋀ 2 V ∗ ⊕ Ker ( V ∗ ϖ ↦ ϖ ∧ ω ⋀ 3 V ∗ ) .
Notice that ⋀ 3 V ∗ = 0 \bigwedge^3V^*=0 ⋀ 3 V ∗ = 0 for n = 1 n=1 n = 1 , and the map ϖ ↦ ϖ ∧ ω \varpi\mapsto\varpi\wedge\omega ϖ ↦ ϖ ∧ ω is injective for n = 2 n=2 n = 2 , which implies that
H 1 ( L V , k ϱ ) ≅ { ⋀ 2 V ∗ ⊕ V ∗ , n = 1 , ⋀ 2 V ∗ , n ≥ 2. H^1(L^{{\varrho}}_{V,\mathbb{k}})\cong
\begin{cases}
\bigwedge^2V^*\oplus V^*, & n=1,\\
\bigwedge^2V^*, & n\geq2.
\end{cases} H 1 ( L V , k ϱ ) ≅ { ⋀ 2 V ∗ ⊕ V ∗ , ⋀ 2 V ∗ , n = 1 , n ≥ 2.
The following example shows that the obstruction in Proposition 6.5 can be nonzero.
Example 6.7. Let V = span k { x } V=\operatorname{span}_{\mathbb{k}}\{x\} V = span k { x } be a one-dimensional abelian Lie algebra with μ = 0 \mu=0 μ = 0 , and let V ′ = span k { u , v } V'=\operatorname{span}_{\mathbb{k}}\{u,v\} V ′ = span k { u , v } . Denote by V ′ ′ = V ′ ⊕ V → p V V''=V'\oplus V\xrightarrow{p}V V ′′ = V ′ ⊕ V p V the projection to the second summand, s : V → V ′ ′ s:V\to V'' s : V → V ′′ the natural embedding. The induced global 2 2 2 -cocycle is
π V ′ = 0 , ϱ ( x ) = id V ′ , ω = 0. \pi_{V'}=0,\qquad\varrho(x)=\operatorname{id}_{V'},\qquad\omega=0. π V ′ = 0 , ϱ ( x ) = id V ′ , ω = 0.
It determines a global extension on V ′ ′ = V ′ ⊕ V V^{\prime\prime}=V^{\prime}\oplus V V ′′ = V ′ ⊕ V , whose Lie bracket is given by
μ ′ ′ ( x , u ) = u , μ ′ ′ ( x , v ) = v . \mu^{\prime\prime}(x,u)=u,\qquad\mu^{\prime\prime}(x,v)=v. μ ′′ ( x , u ) = u , μ ′′ ( x , v ) = v .
Let D = k [ t ] / ( t 2 ) D=k[t]/(t^2) D = k [ t ] / ( t 2 ) . Define
π V ′ ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , π V ′ ′ ( u , v ) = v . \pi_{V^{\prime}}^{\prime}\in\operatorname{Hom}(\wedge^2V^{\prime},V^{\prime}),\qquad\pi_{V^{\prime}}^{\prime}(u,v)=v. π V ′ ′ ∈ Hom ( ∧ 2 V ′ , V ′ ) , π V ′ ′ ( u , v ) = v .
Since π V ′ = 0 \pi_{V^{\prime}}=0 π V ′ = 0 , we have [ π V ′ , π V ′ ′ ] = 0 [\pi_{V^{\prime}},\pi_{V^{\prime}}^{\prime}]=0 [ π V ′ , π V ′ ′ ] = 0 . Thus
π V ′ D = π V ′ + t π V ′ ′ = t π V ′ ′ \pi_{V^{\prime}}^{D}=\pi_{V^{\prime}}+t\pi_{V^{\prime}}^{\prime}=t\pi_{V^{\prime}}^{\prime} π V ′ D = π V ′ + t π V ′ ′ = t π V ′ ′
is an infinitesimal deformation of the Lie bracket on the kernel. We compute the obstruction in Proposition 6.5. Set
O = [ ϱ + ω , π V ′ ′ ] = [ ϱ , π V ′ ′ ] . O=[\varrho+\omega,\pi_{V^{\prime}}^{\prime}]=[\varrho,\pi_{V^{\prime}}^{\prime}]. O = [ ϱ + ω , π V ′ ′ ] = [ ϱ , π V ′ ′ ] .
By direct calculation, we obtain
O ( x , u , v ) = π V ′ ′ ( ϱ ( x ) u , v ) + π V ′ ′ ( u , ϱ ( x ) v ) − ϱ ( x ) π V ′ ′ ( u , v ) = v + v − v = v . O(x,u,v)=\pi_{V^{\prime}}^{\prime}(\varrho(x)u,v)+\pi_{V^{\prime}}^{\prime}(u,\varrho(x)v)-\varrho(x)\pi_{V^{\prime}}^{\prime}(u,v)=v+v-v=v. O ( x , u , v ) = π V ′ ′ ( ϱ ( x ) u , v ) + π V ′ ′ ( u , ϱ ( x ) v ) − ϱ ( x ) π V ′ ′ ( u , v ) = v + v − v = v .
In particular, O ≠ 0 O\ne0 O = 0 .
Moreover, since dim V = 1 \dim V=1 dim V = 1 , we have
L ~ V , V ′ 1 = Hom ( V ⊗ V ′ , V ′ ) , L ~ V , V ′ 2 = Hom ( V ⊗ ∧ 2 V ′ , V ′ ) , L ~ V , V ′ 3 = 0. \widetilde{L}_{V,V^{\prime}}^{1}=\operatorname{Hom}(V\otimes V^{\prime},V^{\prime}),\qquad\widetilde{L}_{V,V^{\prime}}^{2}=\operatorname{Hom}(V\otimes\wedge^2V^{\prime},V^{\prime}),\qquad\widetilde{L}_{V,V^{\prime}}^{3}=0. L V , V ′ 1 = Hom ( V ⊗ V ′ , V ′ ) , L V , V ′ 2 = Hom ( V ⊗ ∧ 2 V ′ , V ′ ) , L V , V ′ 3 = 0.
The differential on L ~ V , V ′ \widetilde{L}_{V,V^{\prime}} L V , V ′ is
δ ~ V , V ′ = [ μ + π V ′ + ϱ + ω , − ] = [ ϱ , − ] . \widetilde{\delta}^{V,V^{\prime}}=[\mu+\pi_{V^{\prime}}+\varrho+\omega,-]=[\varrho,-]. δ V , V ′ = [ μ + π V ′ + ϱ + ω , − ] = [ ϱ , − ] .
For every ϱ ′ ∈ Hom ( V ⊗ V ′ , V ′ ) \varrho^{\prime}\in\operatorname{Hom}(V\otimes V^{\prime},V^{\prime}) ϱ ′ ∈ Hom ( V ⊗ V ′ , V ′ ) , the bracket [ ϱ , ϱ ′ ] [\varrho,\varrho^{\prime}] [ ϱ , ϱ ′ ] belongs to Hom ( ∧ 2 V ⊗ V ′ , V ′ ) = 0 \operatorname{Hom}(\wedge^2V\otimes V^{\prime},V^{\prime})=0 Hom ( ∧ 2 V ⊗ V ′ , V ′ ) = 0 . Consequently,
δ ~ V , V ′ ( L ~ V , V ′ 1 ) = 0. \widetilde{\delta}^{V,V^{\prime}}\left(\widetilde{L}_{V,V^{\prime}}^{1}\right)=0. δ V , V ′ ( L V , V ′ 1 ) = 0.
Hence O O O is a cocycle which is not a coboundary, and
Ob π V ′ ( π V ′ , ϱ , ω ) = [ O ] ≠ 0 in H 2 ( L ~ V , V ′ ) . \operatorname{Ob}_{\pi_{V^{\prime}}}(\pi_{V^{\prime}},\varrho,\omega)=[O]\ne0\quad\text{in }H^2(\widetilde{L}_{V,V^{\prime}}). Ob π V ′ ( π V ′ , ϱ , ω ) = [ O ] = 0 in H 2 ( L V , V ′ ) .
By Proposition 6.5, the deformation π V ′ D \pi_{V^{\prime}}^{D} π V ′ D of the kernel does not lift to an infinitesimal deformation of the given global extension.
This can also be seen directly. Any such lift would satisfy
ϱ D ( x ) = id V ′ + t ϱ ′ ( x ) , ω D = 0. \varrho^{D}(x)=\operatorname{id}_{V^{\prime}}+t\varrho^{\prime}(x),\qquad\omega^{D}=0. ϱ D ( x ) = id V ′ + t ϱ ′ ( x ) , ω D = 0.
The requirement that ϱ D ( x ) \varrho^{D}(x) ϱ D ( x ) is a derivation of π V ′ D \pi_{V^{\prime}}^{D} π V ′ D , evaluated on u , v u,v u , v , would give
ϱ D ( x ) π V ′ D ( u , v ) = π V ′ D ( ϱ D ( x ) u , v ) + π V ′ D ( u , ϱ D ( x ) v ) , \varrho^{D}(x)\pi_{V^{\prime}}^{D}(u,v)=\pi_{V^{\prime}}^{D}(\varrho^{D}(x)u,v)+\pi_{V^{\prime}}^{D}(u,\varrho^{D}(x)v), ϱ D ( x ) π V ′ D ( u , v ) = π V ′ D ( ϱ D ( x ) u , v ) + π V ′ D ( u , ϱ D ( x ) v ) ,
that is, t v = 2 t v tv=2tv t v = 2 t v , a contradiction in V ′ ⊗ D V^{\prime}\otimes D V ′ ⊗ D .
An application: contraction from g ⋊ ad g \mathfrak{g}\rtimes_{\operatorname{ad}}\mathfrak{g} g ⋊ ad g to V ab ⋊ ad g V_{\operatorname{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ⋊ ad g
Definition 7.1. [15 §1.1] Let A A A be an augmented commutative k k k -algebra. An algebraic deformation of a Lie algebra g = ( V , ν ) \mathfrak{g}=(V,\nu) g = ( V , ν ) over A A A is an A A A -bilinear Lie algebra structure on the free module V ⊗ A V\otimes A V ⊗ A whose reduction along the augmentation is ν \nu ν .
Remark 7.2. This definition does not require A A A to be local or Artinian. A closely related notion is the algebraic contraction. An algebraic contraction of ν 1 \nu_1 ν 1 to ν 0 \nu_0 ν 0 is a family over k [ t ] k[t] k [ t ] , augmented by t ↦ 0 t\mapsto0 t ↦ 0 , whose special fiber is ν 0 \nu_0 ν 0 and whose restriction to G m = A 1 ∖ { 0 } G_m=\mathbb{A}^1\setminus\{0\} G m = A 1 ∖ { 0 } is isomorphic to the constant family with fiber ν 1 \nu_1 ν 1 .
We will also deal with the case when ( V ′ ′ , μ ′ ′ ) (V^{\prime\prime},\mu^{\prime\prime}) ( V ′′ , μ ′′ ) is a global extension structure. In this case, the notions of deformation and contraction are defined similarly, keeping track of the extension structure.
Definition 7.3. We call the contraction kernel-framed if the isomorphism over G m \mathbb{G}_{m} G m fixes V ′ V^{\prime} V ′ pointwise, and kernel-unframed if V ′ V^{\prime} V ′ is only stable under the isomorphism over G m \mathbb{G}_{m} G m .
Let g = ( V , μ ) \mathfrak{g}=(V,\mu) g = ( V , μ ) , and let V ′ V^{\prime} V ′ be a second copy of its underlying vector space. Denote the transported bracket on V ′ V^{\prime} V ′ by μ V ′ ( u , v ) = [ u , v ] \mu_{V^{\prime}}(u,v)=[u,v] μ V ′ ( u , v ) = [ u , v ] and the adjoint action by ad x ( u ) = [ x , u ] \operatorname{ad}_{x}(u)=[x,u] ad x ( u ) = [ x , u ] . Set A = k [ t ] A=k[t] A = k [ t ] , augmented by t ↦ 0 t\mapsto0 t ↦ 0 , and write V A ′ = V ′ ⊗ A V^{\prime}_{A}=V^{\prime}\otimes A V A ′ = V ′ ⊗ A and g A = ( V ⊗ A , μ ⊗ 1 A ) \mathfrak{g}_{A}=(V\otimes A,\mu\otimes1_{A}) g A = ( V ⊗ A , μ ⊗ 1 A ) .
Proposition 7.4. The formula
[ ( u , x ) , ( v , y ) ] t = ( [ x , v ] − [ y , u ] + t [ u , v ] , [ x , y ] ) (22) [(u,x),(v,y)]_{t}=([x,v]-[y,u]+t[u,v],[x,y]) \tag*{(22)} [( u , x ) , ( v , y ) ] t = ([ x , v ] − [ y , u ] + t [ u , v ] , [ x , y ]) (22) defines an A A A -Lie algebra structure on ( V ′ ⊕ V ) ⊗ A (V^{\prime}\oplus V)\otimes A ( V ′ ⊕ V ) ⊗ A . The projection to g A \mathfrak{g}_{A} g A is a Lie algebra homomorphism. With respect to the augmentation t ↦ 0 t\mapsto0 t ↦ 0 , it is an algebraic deformation of V a b ′ ⋊ ad g V^{\prime}_{\mathrm{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ′ ⋊ ad g .
Proof. Endow V A ′ V^{\prime}_{A} V A ′ with the bracket [ u , v ] t = t [ u , v ] [u,v]_{t}=t[u,v] [ u , v ] t = t [ u , v ] . Its Jacobiator is t 2 t^{2} t 2 times the Jacobiator of g \mathfrak{g} g , so it vanishes. The adjoint action satisfies
[ x , [ u , v ] t ] = [ [ x , u ] , v ] t + [ u , [ x , v ] ] t [x,[u,v]_{t}]=[[x,u],v]_{t}+[u,[x,v]]_{t} [ x , [ u , v ] t ] = [[ x , u ] , v ] t + [ u , [ x , v ] ] t
and [ ad x , ad y ] = ad [ x , y ] [\operatorname{ad}_{x},\operatorname{ad}_{y}]=\operatorname{ad}_{[x,y]} [ ad x , ad y ] = ad [ x , y ] . Therefore, we get an action of g A \mathfrak{g}_{A} g A on ( V A ′ , [ ⋅ , ⋅ ] t ) (V^{\prime}_{A},[\cdot,\cdot]_{t}) ( V A ′ , [ ⋅ , ⋅ ] t ) by derivations. Moreover, the corresponding Lie algebra structure is given by (22) , which specializes to V a b ′ ⋊ ad g V^{\prime}_{\mathrm{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ′ ⋊ ad g when t = 0 t=0 t = 0 .
Proposition 7.5. The family in Proposition 7.4 is flat. Its fiber at t = 0 t=0 t = 0 is V a b ′ ⋊ ad g V^{\prime}_{\mathrm{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ′ ⋊ ad g , where V a b ′ V^{\prime}_{\mathrm{ab}} V ab ′ is abelian. Its fiber at t = 1 t=1 t = 1 is g ⋊ ad g \mathfrak{g}\rtimes_{\operatorname{ad}}\mathfrak{g} g ⋊ ad g , where the second factor carries its original bracket. For every a ∈ k × a\in k^{\times} a ∈ k × , the map
θ a : ( V ′ ′ , [ ⋅ , ⋅ ] a ) ⟶ ( V ′ ′ , [ ⋅ , ⋅ ] 1 ) , θ a ( u , x ) = ( a u , x ) , \theta_{a}:(V^{\prime\prime},[\cdot,\cdot]_{a})\longrightarrow(V^{\prime\prime},[\cdot,\cdot]_{1}),\qquad\theta_{a}(u,x)=(au,x), θ a : ( V ′′ , [ ⋅ , ⋅ ] a ) ⟶ ( V ′′ , [ ⋅ , ⋅ ] 1 ) , θ a ( u , x ) = ( a u , x ) , is a kernel-unframed global extension isomorphism. Consequently, g ⋊ ad g \mathfrak{g}\rtimes_{\operatorname{ad}}\mathfrak{g} g ⋊ ad g contracts to V a b ′ ⋊ ad g V^{\prime}_{\mathrm{ab}}\rtimes_{\operatorname{ad}}\mathfrak{g} V ab ′ ⋊ ad g at t = 0 t=0 t = 0 in the kernel-unframed sense of Definition 7.3 .
The punctured line A 1 ∖ { 0 } = G m \mathbb{A}^{1}\setminus\{0\}=\mathbb{G}_{m} A 1 ∖ { 0 } = G m is a G m \mathbb{G}_{m} G m -torsor under multiplication. This action lifts to the punctured family and extends to the A 1 \mathbb{A}^{1} A 1 -family by
λ ⋅ ( t ; u , x ) = ( λ t ; λ − 1 u , x ) , ∀ λ ∈ G m , t ∈ A 1 , u ∈ V ′ , x ∈ V . (23) \lambda\cdot(t;u,x)=(\lambda t;\lambda^{-1}u,x),\qquad\forall\lambda\in\mathbb{G}_{m},\ t\in\mathbb{A}^{1},\ u\in V^{\prime},\ x\in V. \tag*{(23)} λ ⋅ ( t ; u , x ) = ( λ t ; λ − 1 u , x ) , ∀ λ ∈ G m , t ∈ A 1 , u ∈ V ′ , x ∈ V . (23)
Proof. Since the underlying A A A -module is free, the family is flat. By substituting t = 0 t=0 t = 0 and t = 1 t=1 t = 1 in (22) , we get the descriptions of the corresponding fibers. If a ∈ k × a\in k^{\times} a ∈ k × , then
θ a ( [ ( u , x ) , ( v , y ) ] a ) = ( a [ x , v ] − a [ y , u ] + a 2 [ u , v ] , [ x , y ] ) = [ θ a ( u , x ) , θ a ( v , y ) ] 1 . \begin{aligned}
\theta_{a}([(u,x),(v,y)]_{a})&=(a[x,v]-a[y,u]+a^{2}[u,v],[x,y])\\
&=[\theta_{a}(u,x),\theta_{a}(v,y)]_{1}.
\end{aligned} θ a ([( u , x ) , ( v , y ) ] a ) = ( a [ x , v ] − a [ y , u ] + a 2 [ u , v ] , [ x , y ]) = [ θ a ( u , x ) , θ a ( v , y ) ] 1 .
In particular, the restriction a Id V ′ a\operatorname{Id}_{V^{\prime}} a Id V ′ is a Lie algebra isomorphism from ( V ′ , a μ V ′ ) (V^{\prime},a\mu_{V^{\prime}}) ( V ′ , a μ V ′ ) to ( V ′ , μ V ′ ) (V^{\prime},\mu_{V^{\prime}}) ( V ′ , μ V ′ ) , since
a Id V ′ ( a [ u , v ] ) = a 2 [ u , v ] = [ a u , a v ] . a\operatorname{Id}_{V^{\prime}}(a[u,v])=a^{2}[u,v]=[au,av]. a Id V ′ ( a [ u , v ]) = a 2 [ u , v ] = [ a u , a v ] .
We remark that the map a Id V ′ a\operatorname{Id}_{V^{\prime}} a Id V ′ does not belong to Aut ( V ′ , μ V ′ ) \operatorname{Aut}(V^{\prime},\mu_{V^{\prime}}) Aut ( V ′ , μ V ′ ) , and it is a Lie algebra isomorphism only after we rescale the Lie bracket on V ′ V^{\prime} V ′ . Hence, the restriction of the family to G m \mathbb{G}_{m} G m is trivialized by
( t ; u , x ) ⟼ ( t ; t u , x ) . (t;u,x)\longmapsto(t;tu,x). ( t ; u , x ) ⟼ ( t ; t u , x ) .
Moreover, the kernel at t = 0 t = 0 t = 0 is abelian. This produces a particular example of the usual Inönü–Wigner contraction construction as in [39 ] .
It remains to check that the action actually lifts. Let Θ λ ( u , x ) = ( λ − 1 u , x ) \Theta_\lambda(u,x)=(\lambda^{-1}u,x) Θ λ ( u , x ) = ( λ − 1 u , x ) . Then
Θ λ ( [ ( u , x ) , ( v , y ) ] t ) = ( λ − 1 [ x , v ] − λ − 1 [ y , u ] + λ − 1 t [ u , v ] , [ x , y ] ) = [ Θ λ ( u , x ) , Θ λ ( v , y ) ] λ t . \Theta_\lambda([(u,x),(v,y)]_t)=(\lambda^{-1}[x,v]-\lambda^{-1}[y,u]+\lambda^{-1}t[u,v],[x,y])=[\Theta_\lambda(u,x),\Theta_\lambda(v,y)]_{\lambda t}. Θ λ ([( u , x ) , ( v , y ) ] t ) = ( λ − 1 [ x , v ] − λ − 1 [ y , u ] + λ − 1 t [ u , v ] , [ x , y ]) = [ Θ λ ( u , x ) , Θ λ ( v , y ) ] λ t .
Hence, the action in (23) carries the fiber at t t t isomorphically to the fiber at λ t \lambda t λ t . Moreover, it is direct to check
Θ λ η = Θ λ ∘ Θ η , ( λ η ) t = λ ( η t ) , \Theta_{\lambda\eta}=\Theta_\lambda\circ\Theta_\eta,\qquad(\lambda\eta)t=\lambda(\eta t), Θ λ η = Θ λ ∘ Θ η , ( λ η ) t = λ ( η t ) ,
so that these fiberwise Lie algebra isomorphisms define a G m \mathbb{G}_m G m -action on the punctured family. □
Proposition 7.6. Let s a b = ad ∈ MC ( L V , V ′ ) s_{\mathrm{ab}}=\operatorname{ad}\in\operatorname{MC}(L_{V,V^\prime}) s ab = ad ∈ MC ( L V , V ′ ) be the Maurer–Cartan element for the special fiber at t = 0 t=0 t = 0 . Relative to the twist L V , V ′ s a b L^{s_{\mathrm{ab}}}_{V,V^\prime} L V , V ′ s ab , the family is
β ( t ) = t μ V ′ . \beta(t)=t\mu_{V^\prime}. β ( t ) = t μ V ′ . For every n ≥ 1 n\geq1 n ≥ 1 ,
t μ V ′ ∈ MC ( L V , V ′ s a b ⊗ ( t ) K [ t ] / ( t n + 1 ) ) . t\mu_{V^\prime}\in\operatorname{MC}\left(L^{s_{\mathrm{ab}}}_{V,V^\prime}\otimes(t)\mathbb{K}[t]/(t^{n+1})\right). t μ V ′ ∈ MC ( L V , V ′ s ab ⊗ ( t ) K [ t ] / ( t n + 1 ) ) . If g \mathfrak{g} g is non-abelian, the first-order classes
[ μ V ′ ] ∈ H 1 ( L V , V ′ s a b ) , [ μ V ′ ] ∈ H 1 ( L V , V ′ u n , s a b ) [\mu_{V^\prime}]\in H^1(L^{s_{\mathrm{ab}}}_{V,V^\prime}),\qquad[\mu_{V^\prime}]\in H^1(L^{\mathrm{un},s_{\mathrm{ab}}}_{V,V^\prime}) [ μ V ′ ] ∈ H 1 ( L V , V ′ s ab ) , [ μ V ′ ] ∈ H 1 ( L V , V ′ un , s ab ) are nonzero. Thus, the formal germ at the degenerate fiber is nontrivial.
Proof. It is straightforward to see that the absolute Maurer–Cartan element of (22) is
s ( t ) = ad + t μ V ′ = s a b + t μ V ′ . s(t)=\operatorname{ad}+t\mu_{V^\prime}=s_{\mathrm{ab}}+t\mu_{V^\prime}. s ( t ) = ad + t μ V ′ = s ab + t μ V ′ .
Since the adjoint action is by derivations, together with the Jacobi identity, we have
[ ad , μ V ′ ] N R = 0 , [ μ V ′ , μ V ′ ] N R = 0. [\operatorname{ad},\mu_{V^\prime}]_{\mathrm{NR}}=0,\qquad[\mu_{V^\prime},\mu_{V^\prime}]_{\mathrm{NR}}=0. [ ad , μ V ′ ] NR = 0 , [ μ V ′ , μ V ′ ] NR = 0.
The projection pr : L V , V ′ → L V ′ \operatorname{pr}:L_{V,V^\prime}\to L_{V^\prime} pr : L V , V ′ → L V ′ in Lemma 5.2 sends d s a b τ d^{s_{\mathrm{ab}}}\tau d s ab τ to zero for every τ ∈ L V , V ′ 0 \tau\in L^0_{V,V^\prime} τ ∈ L V , V ′ 0 . It sends the cocycle μ V ′ \mu_{V^\prime} μ V ′ to the nonzero bracket μ V ′ \mu_{V^\prime} μ V ′ . Hence μ V ′ \mu_{V^\prime} μ V ′ cannot be a boundary when g \mathfrak{g} g is non-abelian.
The same argument applies to the enlarged degree-zero term. If ξ ∈ L V , V ′ u n , 0 = Hom ( V ′ ′ , V ′ ) \xi\in L^{\mathrm{un},0}_{V,V^\prime}=\operatorname{Hom}(V^{\prime\prime},V^\prime) ξ ∈ L V , V ′ un , 0 = Hom ( V ′′ , V ′ ) , then the restriction of d s a b ξ = [ μ + s a b , ξ ] N R d^{s_{\mathrm{ab}}}\xi=[\mu+s_{\mathrm{ab}},\xi]_{\mathrm{NR}} d s ab ξ = [ μ + s ab , ξ ] NR to V ′ ∧ V ′ V^\prime\wedge V^\prime V ′ ∧ V ′ is zero, because the Lie bracket on the kernel of the special fiber is zero. The restriction of μ V ′ \mu_{V^\prime} μ V ′ to V ′ ∧ V ′ V^\prime\wedge V^\prime V ′ ∧ V ′ is the nonzero original bracket. Therefore, μ V ′ \mu_{V^\prime} μ V ′ is not a boundary in L V , V ′ u n , s a b L^{\mathrm{un},s_{\mathrm{ab}}}_{V,V^\prime} L V , V ′ un , s ab either.
Notice that the isomorphism θ a \theta_a θ a changes the framing on the kernel, so that it is an arrow only in the kernel-unframed groupoid. After inverting t t t , the formal family becomes isomorphic to the constant family with fiber g ⋊ ad g \mathfrak{g}\rtimes_{\operatorname{ad}}\mathfrak{g} g ⋊ ad g via the same change. □
The following result is a direct consequence of the kernel-unframed global extension formulism in Theorem 4.9 .
Proposition 7.7. Assume that g \mathfrak{g} g is non-abelian. Let
c : A 1 ⟶ MC ( L V , V ′ ) , c ( t ) = s ( t ) = s a b + t μ V ′ . c:\mathbb{A}^1\longrightarrow\operatorname{MC}(L_{V,V^\prime}),\qquad c(t)=s(t)=s_{\mathrm{ab}}+t\mu_{V^\prime}. c : A 1 ⟶ MC ( L V , V ′ ) , c ( t ) = s ( t ) = s ab + t μ V ′ . Then the following statements hold.
(i) For λ ∈ G m \lambda\in\mathbb{G}_m λ ∈ G m , let
Θ λ = λ − 1 Id V ′ ⊕ Id V ∈ H V , V ′ . \Theta_\lambda=\lambda^{-1}\operatorname{Id}_{V'}\oplus\operatorname{Id}_V\in H_{V,V'}. Θ λ = λ − 1 Id V ′ ⊕ Id V ∈ H V , V ′ .
This is a kernel-unframed global extension isomorphism from the fiber at t t t to the fiber at λ t \lambda t λ t , and transport of brackets satisfies
Θ λ ⋅ c ( t ) = c ( λ t ) . \Theta_\lambda\cdot c(t)=c(\lambda t). Θ λ ⋅ c ( t ) = c ( λ t ) .
Consequently, the restriction of the family to G m \mathbb{G}_m G m is constant in the kernel-unframed stack M C ( L V , V ′ u n ) \mathop{\mathcal{MC}}(L^{\mathrm{un}}_{V,V'}) MC ( L V , V ′ un ) , while its value at 0 0 0 is the contraction limit. Thus, c c c is an algebraic deformation of global extensions and a kernel-unframed algebraic contraction.
(ii) Put ν a b = μ + s a b \nu_{ab}=\mu+s_{ab} ν ab = μ + s ab . In the kernel-framed and kernel-unframed based formal deformation problems, the germ of c c c is the Maurer–Cartan family t μ V ′ t\mu_{V'} t μ V ′ in, respectively, L V , V ′ s a b L^{s_{ab}}_{V,V'} L V , V ′ s ab and L V , V ′ u n , s a b L^{\mathrm{un},s_{ab}}_{V,V'} L V , V ′ un , s ab . Its tangent vectors are the nonzero classes
[ μ V ′ ] ∈ H 1 ( L V , V ′ s a b ) , [ μ V ′ ] ∈ H 1 ( L V , V ′ u n , s a b ) . [\mu_{V'}]\in H^1(L^{s_{ab}}_{V,V'}),\qquad[\mu_{V'}]\in H^1(L^{\mathrm{un},s_{ab}}_{V,V'}). [ μ V ′ ] ∈ H 1 ( L V , V ′ s ab ) , [ μ V ′ ] ∈ H 1 ( L V , V ′ un , s ab ) .
Moreover, the rescaling trivializes the punctured algebraic family.
Proof. On the kernel, the rescaling has the identity
λ − 1 Id V ′ ( t [ u , v ] ) = λ − 1 t [ u , v ] = [ λ − 1 u , λ − 1 v ] λ t . \lambda^{-1}\operatorname{Id}_{V'}(t[u,v])=\lambda^{-1}t[u,v]=[\lambda^{-1}u,\lambda^{-1}v]_{\lambda t}. λ − 1 Id V ′ ( t [ u , v ]) = λ − 1 t [ u , v ] = [ λ − 1 u , λ − 1 v ] λ t .
Together with the compatibility of the adjoint action, this gives the kernel-unframed isomorphism Θ λ : ( V ′ ′ , [ ⋅ , ⋅ ] t ) → ( V ′ ′ , [ ⋅ , ⋅ ] λ t ) \Theta_\lambda:(V'',[\cdot,\cdot]_t)\to(V'',[\cdot,\cdot]_{\lambda t}) Θ λ : ( V ′′ , [ ⋅ , ⋅ ] t ) → ( V ′′ , [ ⋅ , ⋅ ] λ t ) and therefore Θ λ ⋅ c ( t ) = c ( λ t ) \Theta_\lambda\cdot c(t)=c(\lambda t) Θ λ ⋅ c ( t ) = c ( λ t ) . Notice that λ − 1 Id V ′ \lambda^{-1}\operatorname{Id}_{V'} λ − 1 Id V ′ is an isomorphism from ( V ′ , t μ V ′ ) (V',t\mu_{V'}) ( V ′ , t μ V ′ ) to ( V ′ , λ t μ V ′ ) (V',\lambda t\mu_{V'}) ( V ′ , λ t μ V ′ ) . It follows that every point c ( a ) c(a) c ( a ) with a ≠ 0 a\ne0 a = 0 lies in the torus orbit of c ( 1 ) c(1) c ( 1 ) . The trivialization over G m \mathbb{G}_m G m is given by θ t = t Id V ′ ⊕ Id V \theta_t=t\operatorname{Id}_{V'}\oplus\operatorname{Id}_V θ t = t Id V ′ ⊕ Id V .
The last statement follows from Proposition 7.6 together with the two equivalences
D e f E p , [ ν a b ] ( A ) ≃ Del ( L V , V ′ s a b ⊗ m A ) , D e f E p u n , [ ν a b ] ( A ) ≃ Del ( L V , V ′ u n , s a b ⊗ m A ) \mathop{\mathrm{Def}}_{E_p,[\nu_{ab}]}(A)\simeq\operatorname{Del}(L^{s_{ab}}_{V,V'}\otimes\mathfrak{m}_A),\qquad
\mathop{\mathrm{Def}}_{E_p^{\mathrm{un}},[\nu_{ab}]}(A)\simeq\operatorname{Del}(L^{\mathrm{un},s_{ab}}_{V,V'}\otimes\mathfrak{m}_A) Def E p , [ ν ab ] ( A ) ≃ Del ( L V , V ′ s ab ⊗ m A ) , Def E p un , [ ν ab ] ( A ) ≃ Del ( L V , V ′ un , s ab ⊗ m A )
in Theorem 4.5 , Corollary 4.6 and Theorem 4.10 . □
References
Department of Mathematics, Jilin University, Changchun 130012, Jilin, China
Email address: junjiang@jlu.edu.cn
Department of Mathematics, The University of Hong Kong, Pokfulam, Hong Kong, China
Email address: wangchuijia@u.nus.edu
Department of Mathematics, Jilin University, Changchun 130012, Jilin, China
Email address: xukh24@mails.jlu.edu.cn