betaFeedback

WorksMathematics

Rigid Lie Algebras from Semisimple Lie Algebras and Their Symplectic Modules

Abstract

We consider two extensions associated with a nonzero finite-dimensional symplectic module VV over a complex semisimple Lie algebra s\mathfrak{s}, and study their rigidity. The toral extension adjoins a derivation acting as the identity on VV and trivially on s\mathfrak{s}; the Heisenberg extension uses the invariant symplectic form on VV to define a central bracket. For the toral extension, we compute the first and second adjoint cohomology groups in terms of End⁡s(V)\operatorname{End}_{\mathfrak{s}}(V), and obtain rigidity precisely when VV is irreducible. These results hold for arbitrary nonzero finite-dimensional modules. For the Heisenberg extension, we express the second adjoint cohomology as the kernel of a map combining the symplectic centralizer with an equivariant Jacobi map. A Killing-form argument shows that every multiplicity-free sum of irreducible symplectic modules gives an algebraically rigid Heisenberg extension. We also treat constructions with independent toral generators or Heisenberg centers. Repeated summands lead to explicit nonrigid examples, including a deformation to sl3\mathfrak{sl}_3. The proofs use invariant cochains and recover several previously studied families.

Introduction

All Lie algebras and modules in our results are finite-dimensional over C\mathbb{C}. A Lie algebra is geometrically rigid if its change-of-basis orbit is Zariski open in the variety of Lie algebra laws, and algebraically rigid if its second adjoint cohomology vanishes. Algebraic rigidity implies geometric rigidity [17]; the converse fails in general [18].

Let s\mathfrak{s} be semisimple and let (V,ω)(V,\omega) be a nonzero symplectic s\mathfrak{s}-module. Regard VV as abelian. The direct product

g0=(s⋉V)×Ca(1)\mathfrak{g}_0=(\mathfrak{s}\ltimes V)\times\mathbb{C}a \tag*{(1)}

admits two linear deformations, defined by

τ(a,v)=v,η(u,v)=ω(u,v)a,\tau(a,v)=v,\qquad\eta(u,v)=\omega(u,v)a,

with all unspecified values zero. Their nonzero members have respective isomorphism types

gτ=(s⊕Ca)⋉V,gη=s⋉h(V,ω).\mathfrak{g}_{\tau}=(\mathfrak{s}\oplus\mathbb{C}a)\ltimes V,\qquad\mathfrak{g}_{\eta}=\mathfrak{s}\ltimes\mathfrak{h}(V,\omega).

In the first algebra, aa acts as the identity on VV. In the second, h(V,ω)=V⊕Ca\mathfrak{h}(V,\omega)=V\oplus\mathbb{C}a has bracket [u,v]=ω(u,v)a[u,v]=\omega(u,v)a. We compute the low-degree adjoint cohomology of these targets and show how their rigidity depends on the irreducible decomposition of VV.

2020 Mathematics Subject Classification. Primary 17B05; Secondary 17B56.

The toral calculation applies to any nonzero s\mathfrak{s}-module, without a symplectic assumption. With A=End⁡s(V)A = \operatorname{End}_{\mathfrak{s}}(V), it gives

H0(gτ,gτ)=0,H1(gτ,gτ)≃H2(gτ,gτ)≃A/CId⁡VH^{0}(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau}) = 0,\qquad H^{1}(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau}) \simeq H^{2}(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau}) \simeq A/\mathbb{C}\operatorname{Id}_{V}

(Proposition 3.1). Thus gτ\mathfrak{g}_{\tau} is complete and algebraically rigid when VV is irreducible. When VV is reducible, equivariant projections give nontrivial linear deformations, proving geometric nonrigidity. Adjoining an independent toral generator for each irreducible component instead produces an extension whose adjoint cohomology vanishes in every degree (Proposition 3.4).

Our main Heisenberg result applies to every nonzero multiplicity-free sum of irreducible symplectic s\mathfrak{s}-modules:

H0(gη,gη)≃C,H1(gη,gη)≃C,H2(gη,gη)=0.(2)H^{0}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta}) \simeq\mathbb{C},\qquad H^{1}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta}) \simeq\mathbb{C},\qquad H^{2}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta}) = 0. \tag*{(2)}

No exclusion of adjoint components from Λ2V\Lambda^{2}V is required (Theorem 4.2). The proof combines an invariant cochain calculation with a Killing-form argument: an equivariant map Λ2V→s\Lambda^{2}V \to\mathfrak{s} satisfying the corresponding Jacobi identity must vanish. The same argument proves rigidity when each irreducible symplectic component has its own Heisenberg center (Theorem 4.5). These algebras have a single diagonally acting Levi factor.

The general calculation underlying (2) expresses H2(gη,gη)H^{2}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta}) as the kernel of an explicit map involving the symplectic centralizer of the s\mathfrak{s}-action and the adjoint-valued two-cochains (Proposition 4.1). This formula retains the cancellations that occur when irreducible components are repeated. We exhibit two resulting failures of rigidity: three or more isomorphic Heisenberg blocks admit explicit central-valued deformations, while two copies of the defining sl2\mathfrak{sl}_{2}-module with one shared center admit a deformation to sl3\mathfrak{sl}_{3}.

These results belong to the cohomological approach to rigidity developed from the Hochschild–Serre factorization theorem and equivariant cochains. Richardson’s examples distinguish geometric from algebraic rigidity for semidirect products with abelian radical [18]. Related semidirect products involving so(3)\mathfrak{so}(3) were studied in [5]. Burde and Wagemann give explicit adjoint-cohomology calculations for semidirect products and low-dimensional perfect Lie algebras, including examples with Heisenberg radical [2, 3].

The toral extensions fit the external-torus framework of Carles [6, 7]; his cohomological study of generalizations of Borel subalgebras is also relevant to the full-torus calculation [8]. We give direct proofs for the extensions considered here and separate them from these broader structural results. The deformation defined by τ\tau is the split codimension-one construction of Grunewald and O’Halloran [9]; see also [10]. The deformation defined by η\eta instead changes the bracket on Λ2V\Lambda^{2}V.

Several irreducible Heisenberg families are already known. Campoamor-Stursberg proved cohomological rigidity for the Schrödinger algebra and its mass central extension in the nonexceptional spatial dimensions [4]. Khudoyberdiyev and Jumaniyozov obtained the corresponding vanishing results for mass central extensions of conformal Galilei algebras [13 Theorems 5.1 and 5.2]. Their d=1d=1 case includes all Heisenberg extensions by odd irreducible sl2\mathfrak{sl}_{2}-modules. We recover these results as applications of the general theorem. The symplectic oscillator algebras sp2n⋉h2n+1\mathfrak{sp}_{2n}\ltimes\mathfrak{h}_{2n+1} provide another family [15]. Complete extensions with Heisenberg nilradical, which include additional derivations, were studied in [16, 12].

The emphasis of this paper is the general multiplicity-free rigidity criterion, the kernel formula that extends beyond it, and the comparison between a shared center and separate Heisenberg centers. The explicit deformations describe some of the limitations of these criteria. Section 2 collects the cohomological and representation-theoretic tools and constructs the common source. Section 3 treats toral extensions. Section 4 proves the Heisenberg results, and Section 5 gives applications and nonrigid examples.

Acknowledgments

Some of the results of this paper were obtained as part of the Ph.D. thesis of the first author. The authors thank Edison Fernández-Culma and Nadina Rojas for valuable discussions concerning that thesis.

Cohomology, symplectic modules, and the common source

We use g\mathfrak{g} for a general Lie algebra, s\mathfrak{s} for a complex semisimple Lie algebra, and VV for an s\mathfrak{s}-module. The symbol g0\mathfrak{g}_0 denotes the undeformed direct product in (1); its targets are gτ\mathfrak{g}_\tau and gη\mathfrak{g}_\eta. A superscript (r)(r) indicates a construction with rr added generators. The corresponding radicals are denoted by r0\mathfrak{r}_0, rτ\mathfrak{r}_\tau, and rη\mathfrak{r}_\eta, with the same superscript convention. We write Ca\mathbb{C}a for the line spanned by aa, and C\mathbb{C} for the scalar field or the trivial module.

Geometric and algebraic rigidity

Let g\mathfrak{g} be a finite-dimensional complex Lie algebra with bracket μ\mu, and let L(g)\mathcal{L}(\mathfrak{g}) denote the variety of Lie algebra laws on its underlying vector space. The group GL⁡(g)\operatorname{GL}(\mathfrak{g}) acts on L(g)\mathcal{L}(\mathfrak{g}) by change of basis.

Definition 2.1. The Lie algebra g\mathfrak{g} is geometrically rigid if the orbit of μ\mu under GL⁡(g)\operatorname{GL}(\mathfrak{g}) is Zariski open in L(g)\mathcal{L}(\mathfrak{g}). It is algebraically rigid (also called cohomologically rigid) if

H2(g,g)=0,H^2(\mathfrak{g},\mathfrak{g})=0,

where the coefficients are taken in the adjoint module.

Algebraic rigidity implies geometric rigidity [17]. The converse does not hold in general [18].

Definition 2.2. A Lie algebra g\mathfrak{g} is complete if its center is zero and every derivation is inner; equivalently,

H0(g,g)=H1(g,g)=0.H^0(\mathfrak{g},\mathfrak{g})=H^1(\mathfrak{g},\mathfrak{g})=0.

Cohomology of semidirect products

For a Lie algebra g\mathfrak{g} and a g\mathfrak{g}-module MM, write Cq(g,M)=Hom⁡(Λqg,M)C^q(\mathfrak{g},M)=\operatorname{Hom}(\Lambda^q\mathfrak{g},M). Our differential is

(df)(x0,…,xq)=∑i=0q(−1)ixi⋅f(x0,…,xi^,…,xq)+∑0≤i<j≤q(−1)i+jf([xi,xj],x0,…,xi^,…,xj^,…,xq).(df)(x_0,\ldots,x_q)=\sum_{i=0}^{q}(-1)^i x_i\cdot f(x_0,\ldots,\widehat{x_i},\ldots,x_q)+\sum_{0\leq i<j\leq q}(-1)^{i+j}f([x_i,x_j],x_0,\ldots,\widehat{x_i},\ldots,\widehat{x_j},\ldots,x_q).

In particular, for adjoint coefficients, (dz)(x)=[x,z](dz)(x)=[x,z] and (df)(x,y)=[x,f(y)]−[y,f(x)]−f([x,y])(df)(x,y)=[x,f(y)]-[y,f(x)]-f([x,y]). All displayed cohomology isomorphisms are vector-space isomorphisms unless explicitly stated otherwise.

Let II be an ideal of g\mathfrak{g}. The actions of g\mathfrak{g} on II and MM induce an action on C∙(I,M)\mathrm{C}^{\bullet}(I,M), given by

(x⋅f)(u1,…,uq)=x⋅f(u1,…,uq)−∑j=1qf(u1,…,[x,uj],…,uq).(x\cdot f)(u_1,\ldots,u_q)=x\cdot f(u_1,\ldots,u_q)-\sum_{j=1}^{q}f(u_1,\ldots,[x,u_j],\ldots,u_q).

This action commutes with the differential and therefore induces a g\mathfrak{g}-module structure on Hq(I,M)H^q(I,M).

We use the following form of the Hochschild–Serre factorization theorem [11 Theorem 13, p. 603].

Theorem 2.3 (Hochschild–Serre factorization theorem). Let g\mathfrak{g} be a finite-dimensional complex Lie algebra, let MM be a finite-dimensional g\mathfrak{g}-module, and let II be an ideal such that g/I\mathfrak{g}/I is semisimple. Then, for every q≥0q\geq0,

Hq(g,M)≃⨁i+j=qHi(g/I,C)⊗Hj(I,M)g.(3)H^q(\mathfrak{g},M)\simeq\bigoplus_{i+j=q}H^i(\mathfrak{g}/I,\mathbb{C})\otimes H^j(I,M)^{\mathfrak{g}}. \tag*{(3)}

Here the summation denotes a direct sum of vector spaces.

Since g/I\mathfrak{g}/I is semisimple, we have H0(g/I,C)=CH^0(\mathfrak{g}/I,\mathbb{C})=\mathbb{C} and, by Whitehead’s lemmas,

H1(g/I,C)=H2(g/I,C)=0.H^1(\mathfrak{g}/I,\mathbb{C})=H^2(\mathfrak{g}/I,\mathbb{C})=0.

Consequently, restriction induces isomorphisms

Hq(g,M)≃Hq(I,M)g,q=0,1,2.(4)H^q(\mathfrak{g},M)\simeq H^q(I,M)^{\mathfrak{g}},\qquad q=0,1,2. \tag*{(4)}

Now write g=s⋉I\mathfrak{g}=\mathfrak{s}\ltimes I, where s\mathfrak{s} is a semisimple subalgebra complementary to II. By [14 (1.2), p. 1011],

Hq(I,M)g=Hq(I,M)s(q≥0).(5)H^q(I,M)^{\mathfrak{g}}=H^q(I,M)^{\mathfrak{s}}\qquad(q\geq0). \tag*{(5)}

Indeed, the action of II on its cohomology is trivial. Combining (4) and (5) gives the reduction used below.

Corollary 2.4. Let g=s⋉I\mathfrak{g}=\mathfrak{s}\ltimes I, with s\mathfrak{s} semisimple, and let MM be a finite-dimensional g\mathfrak{g}-module. Then, for q=0,1,2q=0,1,2,

Hq(g,M)≃Hq(I,M)s.(6)H^q(\mathfrak{g},M)\simeq H^q(I,M)^{\mathfrak{s}}. \tag*{(6)}

Moreover, every s\mathfrak{s}-invariant cohomology class has an s\mathfrak{s}-invariant cocycle representative, and every s\mathfrak{s}-invariant coboundary has an s\mathfrak{s}-invariant primitive [14 (1.3)–(1.4), p. 1011]. Equivalently, taking s\mathfrak{s}-invariants is exact on finite-dimensional s\mathfrak{s}-modules, and

(Bq(I,M))s=d((Cq−1(I,M))s).\left(B^q(I,M)\right)^{\mathfrak{s}}=d\left(\left(C^{q-1}(I,M)\right)^{\mathfrak{s}}\right).

Thus

Hq(I,M)s≃(Zq(I,M))sd((Cq−1(I,M))s)=Hq((C∙(I,M))s),(7)H^q(I,M)^{\mathfrak{s}}\simeq\frac{\left(Z^q(I,M)\right)^{\mathfrak{s}}}{d\left(\left(C^{q-1}(I,M)\right)^{\mathfrak{s}}\right)}=H^q\left(\left(C^{\bullet}(I,M)\right)^{\mathfrak{s}}\right), \tag*{(7)}

where C−1(I,M)=0C^{-1}(I,M)=0. The invariant complex in the last expression has degree-qq term Hom⁡s(ΛqI,M)\operatorname{Hom}_{\mathfrak{s}}(\Lambda^q I,M), with the restriction of the Chevalley–Eilenberg differential. Invariance is imposed on cochains, including both their arguments and their values.

These reductions apply to both abelian and nonabelian radicals. In each case the differential uses the bracket and coefficient action of the algebra being considered.

Zero-weight cochains

We also use the following consequence of the Cartan homotopy formula. Let g\mathfrak{g} be a Lie algebra, let MM be a g\mathfrak{g}-module, and let x∈gx \in\mathfrak{g}. Denote by Lx\mathcal{L}_x the endomorphism induced by xx on the cochain complex C∙(g,M)C^\bullet(\mathfrak{g},M) and put (ιxf)(y1,…,yq−1)=f(x,y1,…,yq−1)(\iota_xf)(y_1,\ldots,y_{q-1})=f(x,y_1,\ldots,y_{q-1}). Then

Lx=dιx+ιxd.(8)\mathcal{L}_x=d\iota_x+\iota_xd. \tag*{(8)}

Here

(Lxf)(y1,…,yq)=x⋅f(y1,…,yq)−∑i=1qf(y1,…,[x,yi],…,yq).(\mathcal{L}_xf)(y_1,\ldots,y_q)=x\cdot f(y_1,\ldots,y_q)-\sum_{i=1}^{q}f(y_1,\ldots,[x,y_i],\ldots,y_q).

Since dLx=Lxdd\mathcal{L}_x=\mathcal{L}_xd, the differential preserves every eigenspace of Lx\mathcal{L}_x. When Lx\mathcal{L}_x is diagonalizable, these eigenspaces therefore give a direct-sum decomposition of the cochain complex into subcomplexes. If ff is a cocycle of eigenvalue λ≠0\lambda\ne0, Cartan’s identity gives

f=d(λ−1ιxf).f=d(\lambda^{-1}\iota_xf).

Thus every nonzero-eigenvalue subcomplex has zero cohomology, and the inclusion of the zero-eigenvalue (zero-weight) subcomplex induces an isomorphism in cohomology. The same argument applies to (C∙(I,M))s(C^\bullet(I,M))^{\mathfrak{s}} when x∈Ix\in I and [x,s]=0[x,\mathfrak{s}]=0, since insertion of xx then preserves s\mathfrak{s}-invariance.

Symplectic modules

Let s\mathfrak{s} be a semisimple Lie algebra and let VV be an s\mathfrak{s}-module.

Definition 2.5. The module VV is called symplectic if it admits a nondegenerate s\mathfrak{s}-invariant alternating bilinear form

ω:V×V⟶C.\omega:V\times V\longrightarrow\mathbb{C}.

Thus

ω(x⋅v,v′)+ω(v,x⋅v′)=0\omega(x\cdot v,v')+\omega(v,x\cdot v')=0

for every x∈sx\in\mathfrak{s} and v,v′∈Vv,v'\in V.

If VV is irreducible, such a form, when it exists, is unique up to a nonzero scalar. Indeed, an invariant bilinear form is the same as an s\mathfrak{s}-homomorphism V→V∗V\to V^*. Hence an irreducible module admits a nonzero invariant bilinear form if and only if it is self-dual, and any such form is automatically nondegenerate. By Schur’s lemma, it is then either symmetric or alternating.

This distinction is measured by the Frobenius–Schur indicator. If cc denotes the transposition

c(v⊗v′)=v′⊗v,c(v\otimes v')=v'\otimes v,

we denote the indicator of an irreducible s\mathfrak{s}-module VV by

ν(V)=tr⁡(c∣(V⊗V)s).(9)\nu(V)=\operatorname{tr}\left(c\mid(V\otimes V)^{\mathfrak{s}}\right). \tag*{(9)}

The Frobenius–Schur theorem gives

ν(V)={0,V≄V∗,1,V admits an invariant symmetric form,−1,V admits an invariant alternating form.(10)\nu(V)= \begin{cases} 0, & V\not\simeq V^*,\\ 1, & V\text{ admits an invariant symmetric form},\\ -1, & V\text{ admits an invariant alternating form}. \end{cases} \tag*{(10)}

See [1] for the Frobenius–Schur indicators of finite-dimensional modules over complex semisimple Lie algebras. In particular, for an irreducible module,

V is symplectic⟺ν(V)=−1⟺Hom⁡s(Λ2V,C)≠0.(11)V\text{ is symplectic}\quad\Longleftrightarrow\quad\nu(V)=-1\quad\Longleftrightarrow\quad\operatorname{Hom}_{\mathfrak{s}}(\Lambda^{2}V,\mathbb{C})\ne0. \tag*{(11)}

In this case the last space is one-dimensional and dim⁡V\dim V is even.

The indicator is also computable directly from the highest weight. Let V(λ)V(\lambda) be the irreducible module with highest weight λ\lambda, let w0w_{0} be the longest element of the Weyl group, and let ρ∨\rho^{\vee} be the half-sum of the positive coroots. Tits’s formula [1 Theorem 3.6] states that

ν(V(λ))={0,λ≠−w0λ,(−1)λ(2ρ∨),λ=−w0λ.(12)\nu(V(\lambda))= \begin{cases} 0, & \lambda\ne-w_{0}\lambda,\\ (-1)^{\lambda(2\rho^{\vee})}, & \lambda=-w_{0}\lambda. \end{cases} \tag*{(12)}

Thus the symplectic condition is an explicit parity condition on a self-dual highest weight.

For a general module, write

V=⨁i=1kUi⊗Mi,V=\bigoplus_{i=1}^{k}U_{i}\otimes M_{i},

where the UiU_{i} are pairwise nonisomorphic irreducible modules and MiM_{i} are their multiplicity spaces. Schur’s lemma gives the following criterion.

Proposition 2.6. Let VV be an s\mathfrak{s}-module. For each irreducible s\mathfrak{s}-module UU, let mUm_{U} denote its multiplicity in VV, with mU=0m_{U}=0 when it does not occur. Then VV is symplectic if and only if the following conditions hold.

(a) Every irreducible module UU with ν(U)=1\nu(U)=1 occurs in VV with even multiplicity.

(b) For every non-self-dual irreducible UU, one has mU=mU∗m_{U}=m_{U^{*}}.

There is no parity restriction on the multiplicity of an irreducible module with indicator −1-1.

Proof. Write an isotypic component corresponding to a self-dual irreducible module UU as U⊗MU\otimes M, where MM is the multiplicity space. By Schur’s lemma, every invariant bilinear form on this component is the tensor product of the invariant form on UU with a bilinear form on MM. The tensor product is alternating precisely when one factor is alternating and the other is symmetric. Thus a symplectic UU requires a nondegenerate symmetric form on MM, which exists in every dimension, whereas an orthogonal UU requires a nondegenerate alternating form on MM, which exists precisely when dim⁡M\dim M is even.

If UU is not self-dual, an invariant form pairs its isotypic component only with the component corresponding to U∗U^{*}. It is nondegenerate precisely when the two multiplicity spaces have the same dimension. In that case a perfect pairing between them gives an invariant alternating form on the sum of the two components. Different duality orbits are mutually orthogonal, and the result follows. □

Thus a symplectic module need not be a sum of irreducible symplectic modules: orthogonal components can occur with even multiplicity, and a non-self-dual component can be paired with its dual. For example, W⊕W∗W\oplus W^{*} has the canonical symplectic form

ω((u,φ),(v,ψ))=ψ(u)−φ(v).(13)\omega((u,\varphi),(v,\psi))=\psi(u)-\varphi(v). \tag*{(13)}

By contrast, a multiplicity-free sum of irreducible symplectic modules means a sum of pairwise nonisomorphic irreducible modules, each of which is itself symplectic. This is the hypothesis of our Heisenberg rigidity theorems.

Lemma 2.7. If VV is an irreducible symplectic s\mathfrak{s}-module, then

Hom⁡s(V,s)=0.\operatorname{Hom}_{\mathfrak{s}}(V,\mathfrak{s})=0.

Proof. Suppose that there exists a nonzero s\mathfrak{s}-homomorphism from VV to s\mathfrak{s}. Since VV is irreducible, it identifies VV with an irreducible submodule of the adjoint module s\mathfrak{s}. Such a submodule is a simple ideal of s\mathfrak{s} and therefore carries a nondegenerate invariant symmetric bilinear form, given by the restriction of its Killing form. On the other hand, VV carries a nondegenerate invariant alternating form. This is impossible: by Schur’s lemma, the space Hom⁡s(V,V∗)\operatorname{Hom}_{\mathfrak{s}}(V,V^*) is one-dimensional, so two nonzero invariant bilinear forms on VV must be proportional. □

Lemma 2.8. If VV is a direct sum of irreducible symplectic s\mathfrak{s}-modules, then

Hom⁡s(Λ2V,V)=0.\operatorname{Hom}_{\mathfrak{s}}\left(\Lambda^2 V,V\right)=0.

Proof. Put h0=2ρ∨h_0=2\rho^\vee. By Tits’s formula, the highest weight of each summand takes an odd integer value on h0h_0. Every root takes an even integer value on h0h_0, since αi(ρ∨)=1\alpha_i(\rho^\vee)=1 for each simple root. All weights of VV therefore take odd values on h0h_0, whereas all weights of Λ2V\Lambda^2 V take even values. An equivariant map preserves these eigenspaces, so it must be zero. □

We record a few families of irreducible symplectic modules.

  • (a) For s=sl2\mathfrak{s}=\mathfrak{sl}_2, the irreducible module V(n)V(n) of highest weight nn satisfies

ν(V(n))=(−1)n.\nu(V(n))=(-1)^n.

Consequently, V(n)V(n) is symplectic precisely when nn is odd, or equivalently, when dim⁡V(n)=n+1\dim V(n)=n+1 is even.

  • (b) The defining module C2n\mathbb{C}^{2n} of sp2n\mathfrak{sp}_{2n} is irreducible and symplectic for every $n\ge 1.

  • (c) For s=sl4k+2\mathfrak{s}=\mathfrak{sl}_{4k+2}, with k≥0k\ge0, the irreducible module

Λ2k+1C4k+2\Lambda^{2k+1}\mathbb{C}^{4k+2}

is symplectic. Its invariant form is induced by the exterior product

Λ2k+1C4k+2×Λ2k+1C4k+2⟶Λ4k+2C4k+2≃C,\Lambda^{2k+1}\mathbb{C}^{4k+2}\times\Lambda^{2k+1}\mathbb{C}^{4k+2}\longrightarrow\Lambda^{4k+2}\mathbb{C}^{4k+2}\simeq\mathbb{C},

which is alternating because 2k+12k+1 is odd.

An equivariant Jacobi map

For an s\mathfrak{s}-module VV, define

J:Hom⁡s(Λ2V,s)⟶Hom⁡s(Λ3V,V),J(φ)(u,v,w)=φ(u,v)⋅w+φ(v,w)⋅u+φ(w,u)⋅v.(14)\begin{aligned} J:\operatorname{Hom}_{\mathfrak{s}}\left(\Lambda^2V,\mathfrak{s}\right)&\longrightarrow\operatorname{Hom}_{\mathfrak{s}}\left(\Lambda^3V,V\right),\\ J(\varphi)(u,v,w)&=\varphi(u,v)\cdot w+\varphi(v,w)\cdot u+\varphi(w,u)\cdot v. \tag*{(14)} \end{aligned}

The equation J(φ)=0J(\varphi)=0 is the Jacobi identity on VV for the bracket on s⊕V\mathfrak{s}\oplus V obtained by setting [u,v]=φ(u,v)[u,v]=\varphi(u,v).

Lemma 2.9 (Killing-form criterion). If VV is a multiplicity-free sum of irreducible symplectic s\mathfrak{s}-modules, then the map JJ in (14) is injective.

Proof. Let φ∈Hom⁡s(Λ2V,s)\varphi\in\operatorname{Hom}_{\mathfrak{s}}(\Lambda^{2}V,\mathfrak{s}) satisfy J(φ)=0J(\varphi)=0. On l=s⊕V\mathfrak{l}=\mathfrak{s}\oplus V, retain the bracket of s\mathfrak{s} and its action on VV, and put [u,v]=φ(u,v)[u,v]=\varphi(u,v). Equivariance of φ\varphi gives Jacobi on one input in s\mathfrak{s} and two in VV, while J(φ)=0J(\varphi)=0 gives Jacobi on three inputs in VV. Hence l\mathfrak{l} is a Lie algebra.

Let BB be its Killing form and write V=⨁iViV=\bigoplus_i V_i with the ViV_i pairwise nonisomorphic irreducible symplectic modules. There is no nonzero invariant symmetric bilinear form on VV: its restriction to each ViV_i must vanish by Schur’s lemma, and its cross terms vanish because Vi≄VjV_i\not\simeq V_j for i≠ji\ne j. Thus B(V,V)=0B(V,V)=0. Also B(V,s)=0B(V,\mathfrak{s})=0, by Lemma 2.7 and the self-duality of the adjoint module. Therefore VV is contained in the ideal k=ker⁡B\mathfrak{k}=\ker B.

The ideal k\mathfrak{k} is solvable. Indeed, ad⁡(x)\operatorname{ad}(x) acts trivially on l/k\mathfrak{l}/\mathfrak{k} for x∈kx\in\mathfrak{k}, so the Killing form of k\mathfrak{k} is the restriction of BB and is zero. Cartan’s solvability criterion applies. Consequently k∩s\mathfrak{k}\cap\mathfrak{s} is a solvable ideal of s\mathfrak{s} and must be zero. It follows that

im⁡φ=[V,V]⊆k∩s=0.\operatorname{im}\varphi=[V,V]\subseteq\mathfrak{k}\cap\mathfrak{s}=0.

The common deformation source

Let (V,ω)(V,\omega) be a nonzero symplectic s\mathfrak{s}-module. Put

r0=V⊕Ca,g0=(s⋉V)×Ca=s⋉r0,\mathfrak{r}_0=V\oplus\mathbb{C}a,\qquad \mathfrak{g}_0=(\mathfrak{s}\ltimes V)\times\mathbb{C}a=\mathfrak{s}\ltimes\mathfrak{r}_0,

where r0\mathfrak{r}_0 is abelian. Define alternating maps on g0\mathfrak{g}_0 by

τ(a,v)=v,η(u,v)=ω(u,v)a,\tau(a,v)=v,\qquad\eta(u,v)=\omega(u,v)a,

and set them equal to zero on all unspecified pairs.

Lemma 2.10. Let μ0\mu_0 be the Lie bracket of g0\mathfrak{g}_0. For every t∈Ct\in\mathbb{C}, the bilinear maps

μ0+tτandμ0+tη\mu_0+t\tau\qquad\text{and}\qquad\mu_0+t\eta

are Lie brackets.

Proof. For the first family, the only new bracket is

[a,v]t=tv.[a,v]_t=tv.

The endomorphism tId⁡Vt\operatorname{Id}_V commutes with the action of s\mathfrak{s} on VV, and hence aa acts as a derivation of the abelian Lie algebra VV in a way compatible with the action of s\mathfrak{s}. Thus the Jacobi identity holds.

For the second family, the only new bracket is

[v,v′]t=tω(v,v′)a.[v,v']_t=t\omega(v,v')a.

For this bracket, [a,s⊕V⊕Ca]t=0[a,\mathfrak{s}\oplus V\oplus\mathbb{C}a]_t=0. Hence the Jacobi identity for three elements of V⊕CaV\oplus\mathbb{C}a is automatic. The Jacobi identity involving an element x∈sx\in\mathfrak{s} and two elements v,v′∈Vv,v'\in V is precisely the identity

ω(x⋅v,v′)+ω(v,x⋅v′)=0.\omega(x\mathbin{\cdot}v,v')+\omega(v,x\mathbin{\cdot}v')=0.

This follows from the s\mathfrak{s}-invariance of ω\omega. □

For t≠0t\ne0, an isomorphism from the toral member at parameter tt to that at parameter one fixes s⊕V\mathfrak{s}\oplus V and sends aa to tata. For the Heisenberg family it sends aa to t−1at^{-1}a instead. Thus the nonzero targets are respectively gτ\mathfrak{g}_\tau and gη\mathfrak{g}_\eta.

Remark 2.11 (The Grunewald–O’Halloran construction). If II is a codimension-one ideal of (g,μ)(\mathfrak{g},\mu), write g=Cx⊕I\mathfrak{g}=\mathbb{C}x\oplus I and take D∈Der⁡(I)D\in\operatorname{Der}(I). The alternating map

φD(x,y)=D(y),φD(I,I)=0\varphi_D(x,y)=D(y),\qquad\varphi_D(I,I)=0

defines a linear deformation μ+tφD\mu+t\varphi_D: it replaces the derivation ad⁡μ(x)∣I\operatorname{ad}_{\mu}(x)|_I by ad⁡μ(x)∣I+tD\operatorname{ad}_{\mu}(x)|_I+tD [9, 10]. For the source above, take I=s⋉VI=\mathfrak{s}\ltimes V, x=ax=a, and D∣s=0D|_{\mathfrak{s}}=0, D∣V=Id⁡VD|_V=\operatorname{Id}_V. This gives τ\tau and the original action of aa on II is zero. The map η\eta instead changes the bracket on Λ2V\Lambda^2V and takes values in the complementary line.

Remark 2.12. The two directions do not give a two-parameter linear family. The Jacobi expression for μ0+λτ+tη\mu_0+\lambda\tau+t\eta on (a,u,v)(a,u,v) is −2λtω(u,v)a-2\lambda t\omega(u,v)a, so this is a Lie bracket precisely when λt=0\lambda t=0.

Toral extensions

A single toral generator

Let VV be any nonzero s\mathfrak{s}-module and put A=End⁡s(V)A=\operatorname{End}_{\mathfrak{s}}(V). No symplectic assumption is needed in this section.

Proposition 3.1. Let VV be a nonzero s\mathfrak{s}-module and let

gτ=(s⊕Ca)⋉V,[a,v]=v.\mathfrak{g}_{\tau}=(\mathfrak{s}\oplus\mathbb{C}a)\ltimes V,\qquad[a,v]=v.

Then

H0(gτ,gτ)=0,H1(gτ,gτ)≃A/CId⁡V,H2(gτ,gτ)≃A/CId⁡V.(15)H^0(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau})=0,\qquad H^1(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau})\simeq A/\mathbb{C}\operatorname{Id}_V,\qquad H^2(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau})\simeq A/\mathbb{C}\operatorname{Id}_V. \tag*{(15)}

If

V=⨁i=1kUi⊕mi,(16)V=\bigoplus_{i=1}^{k}U_i^{\oplus m_i}, \tag*{(16)}

where the UiU_i are pairwise nonisomorphic irreducible modules, then

dim⁡H1(gτ,gτ)=dim⁡H2(gτ,gτ)=∑i=1kmi2−1.(17)\dim H^1(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau})=\dim H^2(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau})=\sum_{i=1}^{k}m_i^2-1. \tag*{(17)}

Proof. Put rτ=Ca⋉V\mathfrak{r}_{\tau}=\mathbb{C}a\ltimes V. Corollary 2.4 gives

Hq(gτ,gτ)≃Hq(rτ,gτ)s,q=0,1,2.H^q(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau})\simeq H^q(\mathfrak{r}_{\tau},\mathfrak{g}_{\tau})^{\mathfrak{s}},\qquad q=0,1,2.

Cartan’s homotopy formula (8) reduces the calculation to

C0q={f∈Hom⁡s(Λqrτ,gτ):Laf=0}.C^q_0=\{f\in\operatorname{Hom}_{\mathfrak{s}}(\Lambda^q\mathfrak{r}_{\tau},\mathfrak{g}_{\tau}):\mathcal{L}_a f=0\}.

In degrees zero through three its terms are

C00=Ca,C^0_0=\mathbb{C}a,
C01=Cε⊕A,C^1_0=\mathbb{C}\varepsilon\oplus A,
C02={δT:T∈A}≃A,C^2_0=\{\delta_T:T\in A\}\simeq A,
C03=0.C^3_0=0.

Here all unspecified values are zero, and

ε(a)=a,T(v)=Tv,δT(a,v)=Tv.\varepsilon(a)=a,\qquad T(v)=Tv,\qquad\delta_T(a,v)=Tv.

Indeed, a zero-weight one-cochain can only send aa to Ca\mathbb{C}a and VV to VV, while a zero-weight two-cochain can only send a∧Va\wedge V to VV. Equivariance identifies the latter two types of maps with AA. A threefold exterior product is either in Λ3V\Lambda^3 V (weight three) or in a∧Λ2Va \wedge\Lambda^2 V (weight two); a product with two copies of aa is zero. Since the target has only weights zero and one, C03=0C^3_0=0.

For a one-cochain ff, we use

(df)(x,y)=[x,f(y)]−[y,f(x)]−f([x,y]).(df)(x,y)=[x,f(y)]-[y,f(x)]-f([x,y]).

A direct evaluation on v∈Vv\in V and on (a,v)(a,v) gives

da=−Id⁡V,d(λε+T)=λδId⁡V,dδT=0.(18)da=-\operatorname{Id}_V,\qquad d(\lambda\varepsilon+T)=\lambda\delta_{\operatorname{Id}_V},\qquad d\delta_T=0. \tag*{(18)}

Consequently the map C00→C01C^0_0\to C^1_0 is injective, the kernel in degree one is AA, and the space of boundaries is CId⁡V\mathbb{C}\operatorname{Id}_V. In degree two every δT\delta_T is a cocycle, while the boundaries are exactly the scalar multiples of δId⁡V\delta_{\operatorname{Id}_V}. This proves (15). Finally, A≃⨁iMmi(C)A\simeq\bigoplus_i M_{m_i}(\mathbb{C}) for the decomposition (16), which gives (17). The subtraction of one comes from the scalar identity: it is a one-coboundary, and its corresponding two-cochain is dεd\varepsilon. This remains true when VV contains trivial one-dimensional summands. □

Corollary 3.2. The algebra gτ\mathfrak{g}_\tau is algebraically rigid if and only if VV is irreducible. In that case gτ\mathfrak{g}_\tau is also complete.

Proof. By complete reducibility, dim⁡A=∑imi2\dim A=\sum_i m_i^2 equals one precisely when VV is irreducible. Apply Proposition 3.1. □

Every class represented by δT\delta_T integrates by replacing the action of aa with Id⁡V+tT\operatorname{Id}_V+tT.

Corollary 3.3. If VV is reducible, then gτ\mathfrak{g}_\tau is not geometrically rigid.

Proof. Choose a proper nonzero submodule and an s\mathfrak{s}-equivariant projection PP onto it. For t∈Ct\in\mathbb{C}, replace the action of aa by

[a,v]t=(Id⁡V+tP)v.(19)[a,v]_t=(\operatorname{Id}_V+tP)v. \tag*{(19)}

Because PP commutes with s\mathfrak{s}, this defines a linear deformation whose infinitesimal part is δP\delta_P. For t≠−1t\ne-1, this action is invertible. The solvable radical is Ca⋉V\mathbb{C}a\ltimes V, and any nilpotent ideal is contained in it. If such an ideal contained ca+vca+v with c≠0c\ne0, its brackets with VV would contain VV, on which ad⁡(ca+v)\operatorname{ad}(ca+v) is invertible. This contradicts nilpotence. Hence the nilradical is VV and is characteristic. The center of the quotient by VV is the line spanned by the class of aa. Hence an isomorphism with gτ\mathfrak{g}_\tau, the member at t=0t=0 of this new family, would have to conjugate a nonzero scalar multiple of Id⁡V+tP\operatorname{Id}_V+tP to Id⁡V\operatorname{Id}_V. For t≠0,−1t\ne0,-1, this is impossible: the action has two distinct eigenvalues, 11 and 1+t1+t, on nonzero subspaces. Thus arbitrarily close members of the family are not isomorphic to gτ\mathfrak{g}_\tau, and gτ\mathfrak{g}_\tau is not geometrically rigid. □

Independent toral generators

Choose a decomposition V=⨁i=1rViV=\bigoplus_{i=1}^r V_i into irreducible components, allowing repetitions, and let Zr=⨁i=1rCaiZ_r=\bigoplus_{i=1}^r\mathbb{C}a_i be a trivial s\mathfrak{s}-module. The projections onto the ViV_i define commuting semisimple derivations of s⋉V\mathfrak{s}\ltimes V. Hence

gτ(r)=(s⊕Zr)⋉V,[ai,vj]=δijvj(20)\mathfrak{g}^{(r)}_\tau=(\mathfrak{s}\oplus Z_r)\ltimes V,\qquad[a_i,v_j]=\delta_{ij}v_j \tag*{(20)}

is a Lie algebra. It is the nonzero target of the linear deformation of (s⋉V)×Zr(\mathfrak{s}\ltimes V)\times Z_r defined by τ(r)(ai,vj)=δijvj\tau^{(r)}(a_i,v_j)=\delta_{ij}v_j.

An external torus is an abelian subalgebra of semisimple derivations with zero intersection with the inner derivations. The derivations just described span a maximal torus in A≃⨁iEnd⁡(Mi)A \simeq\bigoplus_i \operatorname{End}(M_i) and form an external torus of s⋉V\mathfrak{s} \ltimes V. Indeed, an inner derivation vanishing on s\mathfrak{s} is defined by a vector in VsV^{\mathfrak{s}} and acts trivially on VV. The scalar derivation defining gτ\mathfrak{g}_{\tau} is the sum of these independent generators; its span is maximal precisely when VV is irreducible. This identifies the constructions within the external-torus framework of [6, 7]. The following direct calculation gives the cohomology of the full-torus extension; compare also the broader cohomological setting of [8].

Proposition 3.4 (The full-torus extension). Let V=⨁i=1rViV=\bigoplus_{i=1}^{r}V_i be any decomposition into nonzero irreducible s\mathfrak{s}-modules, allowing isomorphic or trivial summands. Let ZrZ_r act by [ai,vj]=δijvj[a_i,v_j]=\delta_{ij}v_j, and put gτ(r)=(s⊕Zr)⋉V\mathfrak{g}_{\tau}^{(r)}=(\mathfrak{s}\oplus Z_r)\ltimes V. Then

Hq(gτ(r),gτ(r))=0(q≥0).H^q(\mathfrak{g}_{\tau}^{(r)},\mathfrak{g}_{\tau}^{(r)})=0 \qquad(q\geq0).

In particular, this extension is complete and algebraically rigid.

Proof. Apply Theorem 2.3 with semisimple factor s\mathfrak{s} and radical rτ(r)=Zr⋉V\mathfrak{r}_{\tau}^{(r)}=Z_r\ltimes V. Cartan’s homotopy formula for the commuting operators aia_i reduces C∙(rτ(r),gτ(r))sC^\bullet(\mathfrak{r}_{\tau}^{(r)},\mathfrak{g}_{\tau}^{(r)})^{\mathfrak{s}} to its common zero-weight part. Let α1,…,αr∈Zr∗\alpha_1,\ldots,\alpha_r\in Z_r^* be the dual basis, so αi(aj)=δij\alpha_i(a_j)=\delta_{ij}. The output weights are zero and the αi\alpha_i; hence a cochain of weight zero has either only toral inputs and toral output, or exactly one input in ViV_i and output in that same ViV_i. Even if two modules are isomorphic, their torus weights are different. Schur’s lemma gives

C0q≃(ΛqZr∗⊗Zr)⊕(Λq−1Zr∗⊗Cr),C_0^q \simeq\left(\Lambda^q Z_r^*\otimes Z_r\right)\oplus\left(\Lambda^{q-1}Z_r^*\otimes\mathbb{C}^r\right),

where Λ−1Zr∗=0\Lambda^{-1}Z_r^*=0. If F:ΛqZr→ZrF:\Lambda^q Z_r\to Z_r is in the first summand, then

(dF)(z1,…,zq,vi)=(−1)q+1αi(F(z1,…,zq))vi.(dF)(z_1,\ldots,z_q,v_i)=(-1)^{q+1}\alpha_i(F(z_1,\ldots,z_q))v_i.

The differential is zero on the second summand: the action on its output cancels the action on its one ViV_i input, and VV is abelian. Thus the differential identifies the first summand in degree qq isomorphically with the second summand in degree q+1q+1. The whole zero-weight complex is acyclic. The factorization theorem then gives the assertion in every degree. □

Heisenberg extensions

The general kernel formula

Let (V,ω)(V,\omega) be a nonzero symplectic s\mathfrak{s}-module and put gη=s⋉h(V,ω)\mathfrak{g}_{\eta}=\mathfrak{s}\ltimes\mathfrak{h}(V,\omega), with [u,v]=ω(u,v)a[u,v]=\omega(u,v)a. We use the spaces

A=End⁡s(V),Ωs(V)=Hom⁡s(Λ2V,C),S=Hom⁡s(Λ2V,s).(21)A=\operatorname{End}_{\mathfrak{s}}(V),\qquad\Omega_{\mathfrak{s}}(V)=\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,\mathbb{C}),\qquad S=\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,\mathfrak{s}). \tag*{(21)}

The symplectic centralizer of the s\mathfrak{s}-action is

cω={T∈A:ω(Tu,v)+ω(u,Tv)=0 for all u,v∈V}.(22)\mathfrak{c}_{\omega}=\{T\in A:\omega(Tu,v)+\omega(u,Tv)=0\text{ for all }u,v\in V\}. \tag*{(22)}

Define B:A→Ωs(V)B:A\to\Omega_{\mathfrak{s}}(V) by

B(T)(u,v)=ω(Tu,v)+ω(u,Tv).(23)B(T)(u,v)=\omega(Tu,v)+\omega(u,Tv). \tag*{(23)}

This map is surjective. Indeed, for β∈Ωs(V)\beta\in\Omega_{\mathfrak{s}}(V) write β(u,v)=ω(Ru,v)\beta(u,v)=\omega(Ru,v). Invariance gives R∈AR\in A, and alternation gives ω(Ru,v)=ω(u,Rv)\omega(Ru,v)=\omega(u,Rv), so B(R)=2βB(R)=2\beta. The dilation derivation is given by

D∣s=0,D(v)=v,D(a)=2a.(24)D|_{\mathfrak{s}}=0,\qquad D(v)=v,\qquad D(a)=2a. \tag*{(24)}

Proposition 4.1. Suppose that Vs=0V^{\mathfrak{s}}=0 and Hom⁡s(V,s)=0\operatorname{Hom}_{\mathfrak{s}}(V,\mathfrak{s})=0. For gη=s⋉h(V,ω)\mathfrak{g}_{\eta}=\mathfrak{s}\ltimes\mathfrak{h}(V,\omega) one has

H0(gη,gη)≃C,H1(gη,gη)≃C[D]⊕cω.(25)H^{0}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta})\simeq\mathbb{C}, \qquad H^{1}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta})\simeq\mathbb{C}[D]\oplus\mathfrak{c}_{\omega}. \tag*{(25)}

Here DD is given by (24), and T∈cωT\in\mathfrak{c}_{\omega} represents the derivation

DT∣s=0,DT(v)=Tv,DT(a)=0.(26)D_{T}|_{\mathfrak{s}}=0,\qquad D_{T}(v)=Tv,\qquad D_{T}(a)=0. \tag*{(26)}

If, in addition, Hom⁡s(Λ2V,V)=0\operatorname{Hom}_{\mathfrak{s}}(\Lambda^{2}V,V)=0, define

K(T)(u,v,w)=ω(u,v)Tw+ω(v,w)Tu+ω(w,u)Tv.(27)K(T)(u,v,w)=\omega(u,v)Tw+\omega(v,w)Tu+\omega(w,u)Tv. \tag*{(27)}

Let JJ be the map in (14). Then

H2(gη,gη)≃ker⁡(K+J:cω⊕S⟶Hom⁡s(Λ3V,V)).(28)H^{2}(\mathfrak{g}_{\eta},\mathfrak{g}_{\eta})\simeq\ker(K+J:\mathfrak{c}_{\omega}\oplus S\longrightarrow\operatorname{Hom}_{\mathfrak{s}}(\Lambda^{3}V,V)). \tag*{(28)}

Proof. Put rη=h(V,ω)=V⊕Ca\mathfrak{r}_{\eta}=\mathfrak{h}(V,\omega)=V\oplus\mathbb{C}a, with [u,v]=ω(u,v)a[u,v]=\omega(u,v)a, so that gη=s⋉rη\mathfrak{g}_{\eta}=\mathfrak{s}\ltimes\mathfrak{r}_{\eta}. Corollary 2.4 reduces the calculation to the invariant complex (C∙(rη,gη))s(C^{\bullet}(\mathfrak{r}_{\eta},\mathfrak{g}_{\eta}))^{\mathfrak{s}}. In degrees zero and one it has terms

(C0(rη,gη))s=Ca,(C1(rη,gη))s=A⊕Cqa,(C^{0}(\mathfrak{r}_{\eta},\mathfrak{g}_{\eta}))^{\mathfrak{s}}=\mathbb{C}a,\qquad(C^{1}(\mathfrak{r}_{\eta},\mathfrak{g}_{\eta}))^{\mathfrak{s}}=A\oplus\mathbb{C}q_{a},

where T∈AT\in A is extended by T(a)=0T(a)=0, and qa(a)=aq_{a}(a)=a, qa(V)=0q_{a}(V)=0. With ηβ(u,v)=β(u,v)a\eta_{\beta}(u,v)=\beta(u,v)a and all other values zero,

dT=ηB(T),dqa=−ηω.(29)dT=\eta_{B(T)},\qquad dq_{a}=-\eta_{\omega}. \tag*{(29)}

Thus a one-cochain (T,c)(T,c) is a cocycle precisely when B(T)=cωB(T)=c\omega. Since B(Id⁡V)=2ωB(\operatorname{Id}_{V})=2\omega, it decomposes uniquely as

(T,c)=c2(Id⁡V,2)+(T−c2Id⁡V,0),(T,c)=\frac{c}{2}(\operatorname{Id}_{V},2)+(T-\frac{c}{2}\operatorname{Id}_{V},0),

with the second term in cω\mathfrak{c}_{\omega}. The invariant zero-cochain aa has zero differential because [a,gη]=0[a,\mathfrak{g}_{\eta}]=0. This proves the assertions in degrees zero and one.

Under the additional hypothesis, every invariant two-cochain is uniquely

ηβ+δT+φ,β∈Ωs(V), T∈A, φ∈S,\eta_{\beta}+\delta_{T}+\varphi,\qquad\beta\in\Omega_{\mathfrak{s}}(V),\ T\in A,\ \varphi\in S,

where δT(a,v)=Tv\delta_{T}(a,v)=Tv and φ\varphi is zero on a∧Va\wedge V. Every ηβ\eta_{\beta} is a cocycle and, by surjectivity of BB, a coboundary. The remaining differential is

d(δT+φ)(a,u,v)=−B(T)(u,v)a,(30)d(\delta_{T}+\varphi)(a,u,v)=-B(T)(u,v)a, \tag*{(30)}
d(δT+φ)(u,v,v′)=−K(T)(u,v,v′)−J(φ)(u,v,v′).(31)d(\delta_{T}+\varphi)(u,v,v')=-K(T)(u,v,v')-J(\varphi)(u,v,v'). \tag*{(31)}

These are all possible types of input triples. Therefore the cocycle conditions are T∈cωT\in\mathfrak{c}_{\omega} and K(T)+J(φ)=0K(T)+J(\varphi)=0. The invariant two-coboundaries have only the Ωs(V)\Omega_{\mathfrak{s}}(V) component, which proves (28). □\square

For dim⁡V≥4\dim V\ge4, KK is injective on End⁡(V)\operatorname{End}(V). For v′≠0v'\ne0, the quotient v′⊥/Cv′v'^{\perp}/\mathbb{C}v' is symplectic of dimension dim⁡V−2\dim V-2, so there are u,v∈v′⊥u,v\in v'^{\perp} with ω(u,v)=1\omega(u,v)=1. Then K(T)(u,v,v′)=Tv′K(T)(u,v,v')=Tv'. Nevertheless, a nonzero K(T)K(T) may cancel J(φ)J(\varphi) in (28); injectivity of KK alone is not a vanishing argument when S≠0S\ne0.

Multiplicity-free rigidity

Theorem 4.2. Let V=V1⊕⋯⊕VrV=V_1\oplus\cdots\oplus V_r be a nonzero multiplicity-free sum of irreducible symplectic s\mathfrak{s}-modules. For every invariant symplectic form ω\omega on VV, the algebra gη=s⋉h(V,ω)\mathfrak{g}_\eta=\mathfrak{s}\ltimes\mathfrak{h}(V,\omega) satisfies

H0(gη,gη)=Ca,H1(gη,gη)=C[D],H2(gη,gη)=0.(32)H^0(\mathfrak{g}_\eta,\mathfrak{g}_\eta)=\mathbb{C}a,\qquad H^1(\mathfrak{g}_\eta,\mathfrak{g}_\eta)=\mathbb{C}[D],\qquad H^2(\mathfrak{g}_\eta,\mathfrak{g}_\eta)=0. \tag*{(32)}

In particular, gη\mathfrak{g}_\eta is algebraically and geometrically rigid.

Proof. Each ViV_i is nontrivial, and Lemma 2.7 gives Hom⁡s(V,s)=0\operatorname{Hom}_{\mathfrak{s}}(V,\mathfrak{s})=0. Lemma 2.8 gives Hom⁡s(Λ2V,V)=0\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,V)=0. By Schur’s lemma,

A=⨁i=1rCId⁡Vi.A=\bigoplus_{i=1}^{r}\mathbb{C}\operatorname{Id}_{V_i}.

There are no invariant cross pairings between different ViV_i, so ω\omega restricts nondegenerately to every ViV_i. If T=⨁itiId⁡ViT=\bigoplus_i t_i\operatorname{Id}_{V_i} belongs to cω\mathfrak{c}_\omega, then 2tiω∣Vi=02t_i\omega|_{V_i}=0 for every ii. Thus cω=0\mathfrak{c}_\omega=0. Proposition 4.1 now gives the first two assertions and identifies H2(gη,gη)H^2(\mathfrak{g}_\eta,\mathfrak{g}_\eta) with ker⁡J\ker J. Lemma 2.9 proves the remaining vanishing. □

For an irreducible symplectic module, both nonzero targets of Lemma 2.10 are therefore algebraically rigid. For a multiplicity-free sum of two or more irreducible symplectic modules, the Heisenberg target is algebraically rigid whereas the single-generator toral target is geometrically nonrigid.

Remark 4.3. Whenever Vs=0V^{\mathfrak{s}}=0, the algebra gη\mathfrak{g}_\eta is perfect, and its radical and nilradical are both h(V,ω)\mathfrak{h}(V,\omega). Indeed, [s,V]=V[\mathfrak{s},V]=V by complete reducibility, and [V,V]=Ca[V,V]=\mathbb{C}a by nondegeneracy. Every nilpotent ideal maps to zero in the semisimple quotient, so it lies in the Heisenberg radical. Thus these targets belong to the perfect case in Carles’s structural alternatives for geometrically rigid algebras [6].

Remark 4.4 (The stronger adjoint exclusion). The condition Vs=0V^{\mathfrak{s}}=0 and S=0S=0 implies the hypotheses of Theorem 4.2, but is not needed there. To see this implication, let κ\kappa be the Killing form of s\mathfrak{s}. For T∈cωT\in\mathfrak{c}_\omega, the form bT(u,v)=ω(Tu,v)b_T(u,v)=\omega(Tu,v) is symmetric and invariant. Define ΦT:Λ2V→s\Phi_T:\Lambda^2V\to\mathfrak{s} by

κ(ΦT(u,v),x)=bT(x⋅u,v).\kappa(\Phi_T(u,v),x)=b_T(x\cdot u,v).

Symmetry and invariance make ΦT\Phi_T alternating and equivariant. If ΦT=0\Phi_T=0, then T(x⋅u)=0T(x\cdot u)=0 for all x,ux,u; since [s,V]=V[\mathfrak{s},V]=V, one has T=0T=0. Hence cω\mathfrak{c}_\omega injects into SS, and S=0S=0 forces cω=0\mathfrak{c}_\omega=0.

The multiplicity-space description in Proposition 2.6 identifies the factors of cω\mathfrak{c}_\omega with so(M,b)\mathfrak{so}(M,b) for a symplectic irreducible component, sp(M,b)\mathfrak{sp}(M,b) for an orthogonal one, and gl(M)\mathfrak{gl}(M) for a non-self-dual pair. Here bb is the corresponding nondegenerate symmetric or alternating form on the multiplicity space. These factors all vanish precisely when every component is symplectic and has multiplicity one. In particular, the stronger exclusion S=0S=0 implies this condition. Examples with S≠0S\ne0 are given in § 5.

Separate Heisenberg centers

Suppose V=⨁i=1rViV=\bigoplus_{i=1}^{r}V_i, with each ViV_i a nonzero symplectic s\mathfrak{s}-module carrying a form ωi\omega_i. Let Zr=⨁i=1rCaiZ_r=\bigoplus_{i=1}^{r}\mathbb{C}a_i be a trivial module and put

gη(r)=s⋉⨁i=1rh(Vi,ωi),[ui,vi]=ωi(ui,vi)ai.(33)\mathfrak{g}_{\eta}^{(r)}=\mathfrak{s}\ltimes\bigoplus_{i=1}^{r}\mathfrak{h}(V_i,\omega_i),\qquad[u_i,v_i]=\omega_i(u_i,v_i)a_i. \tag*{(33)}

Different radical blocks commute. The same Levi factor acts on all blocks; this is not a direct product of rr semidirect products. Its radical is rη(r)=V⊕Zr\mathfrak{r}_{\eta}^{(r)}=V\oplus Z_r.

The algebra (33) is a nonzero target of the linear deformation of (s⋉V)×Zr(\mathfrak{s}\ltimes V)\times Z_r given by

η(r)(ui,vi)=ωi(ui,vi)ai,η(r)(Vi,Vj)=0(i≠j).\eta^{(r)}(u_i,v_i)=\omega_i(u_i,v_i)a_i,\qquad\eta^{(r)}(V_i,V_j)=0\quad(i\ne j).

Jacobi follows from the centrality of ZrZ_r and the invariance of each ωi\omega_i. Rescaling the aia_i identifies all nonzero members. When the ViV_i are irreducible, this source also gives the full-torus target of (20).

Theorem 4.5. If V1,…,VrV_1,\ldots,V_r are pairwise nonisomorphic irreducible symplectic s\mathfrak{s}-modules, then

H0(gη(r),gη(r))=Zr,H^0(\mathfrak{g}_{\eta}^{(r)},\mathfrak{g}_{\eta}^{(r)})=Z_r,
H1(gη(r),gη(r))≅Cr,H^1(\mathfrak{g}_{\eta}^{(r)},\mathfrak{g}_{\eta}^{(r)})\cong\mathbb{C}^r,
H2(gη(r),gη(r))=0.(34)H^2(\mathfrak{g}_{\eta}^{(r)},\mathfrak{g}_{\eta}^{(r)})=0. \tag*{(34)}

The first cohomology is generated by the classes of the derivations DiD_i defined by Di∣Vi=Id⁡ViD_i|_{V_i}=\operatorname{Id}_{V_i}, Di(ai)=2aiD_i(a_i)=2a_i, and zero on the remaining components. In particular, gη(r)\mathfrak{g}_{\eta}^{(r)} is algebraically rigid.

Proof. Corollary 2.4 reduces the calculation to the s\mathfrak{s}-invariant cochain complex of rη(r)\mathfrak{r}_{\eta}^{(r)}. Schur’s lemma gives

End⁡s(V)=⨁iCId⁡Vi,Ωs(V)=⨁iCωi.\operatorname{End}_{\mathfrak{s}}(V)=\bigoplus_i\mathbb{C}\operatorname{Id}_{V_i},\qquad\Omega_{\mathfrak{s}}(V)=\bigoplus_i\mathbb{C}\omega_i.

Also Vs=0V^{\mathfrak{s}}=0, Hom⁡s(V,s)=0\operatorname{Hom}_{\mathfrak{s}}(V,\mathfrak{s})=0, and Hom⁡s(Λ2V,V)=0\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,V)=0, by the preceding lemmas.

The invariant zero-cochains are ZrZ_r and have zero differential. An invariant one-cochain has the form f(vi)=pivif(v_i)=p_iv_i, f(z)=Qzf(z)=Qz, with pi∈Cp_i\in\mathbb{C} and Q∈End⁡(Zr)Q\in\operatorname{End}(Z_r). Its differential is

(df)(ui,vi)=ωi(ui,vi)(2piai−Q(ai)),(35)(df)(u_i,v_i)=\omega_i(u_i,v_i)(2p_ia_i-Q(a_i)), \tag*{(35)}

and is zero on the other pairs. Thus it is a cocycle precisely when Q(ai)=2piaiQ(a_i)=2p_ia_i for all ii. This proves the assertions in degrees zero and one.

Every invariant two-cochain is uniquely a sum

β+δλ+γ+φ,\beta+\delta\lambda+\gamma+\varphi,

where β(ui,vi)=ωi(ui,vi)bi\beta(u_i,v_i)=\omega_i(u_i,v_i)b_i for arbitrary bi∈Zrb_i\in Z_r, δλ(z,vi)=λi(z)vi\delta\lambda(z,v_i)=\lambda_i(z)v_i for λi∈Zr∗\lambda_i\in Z_r^*, γ∈Hom⁡(Λ2Zr,Zr)\gamma\in\operatorname{Hom}(\Lambda^2Z_r,Z_r), and φ∈Hom⁡s(Λ2V,s)\varphi\in\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,\mathfrak{s}). Each map is extended by zero on the other types of pairs. The stated vanishing results exclude all further cochain types.

The differential on (z,ui,vi)(z,u_i,v_i) is

−ωi(ui,vi)(2λi(z)ai+γ(ai,z)).-\omega_i(u_i,v_i)(2\lambda_i(z)a_i+\gamma(a_i,z)).

Thus a cocycle satisfies γ(ai,z)=−2λi(z)ai\gamma(a_i,z)=-2\lambda_i(z)a_i. Taking z=aiz=a_i gives λi(ai)=0\lambda_i(a_i)=0. Taking z=ajz=a_j and exchanging i,ji,j gives

−2λi(aj)ai=γ(ai,aj)=2λj(ai)aj.-2\lambda_i(a_j)a_i=\gamma(a_i,a_j)=2\lambda_j(a_i)a_j.

For i≠ji\ne j, linear independence of ai,aja_i,a_j forces both coefficients to vanish. Hence λi=0\lambda_i=0 for all ii and γ=0\gamma=0. Every β\beta is a coboundary: in (35), take pi=0p_i=0 and Q(ai)=−biQ(a_i)=-b_i. The remaining cocycle equation on three inputs in VV is J(φ)=0J(\varphi)=0. Lemma 2.9 gives φ=0\varphi=0, proving the vanishing in degree two. □

Applications and nonrigid examples

Irreducible families

Corollary 5.1. Let V=C2nV=\mathbb{C}^{2n} be the defining module of sp2n\mathfrak{sp}_{2n}, where n≥1n\ge1. Both nonzero targets associated with (sp2n⋉V)×Ca(\mathfrak{sp}_{2n}\ltimes V)\times\mathbb{C}a are algebraically rigid. The Heisenberg target is the symplectic oscillator algebra sp2n⋉h2n+1\mathfrak{sp}_{2n}\ltimes\mathfrak{h}_{2n+1}.

Proof. The defining module is irreducible and symplectic. Apply Corollary 3.2 and Theorem 4.2. For the terminology, see [15]. □

Corollary 5.2. Let mm be a positive odd integer and let d=1d=1 or d≥3d\ge3. Put

sd={sl2,d=1,sl2⊕sod,d≥3,Vd={V(m),d=1,V(m)⊗Cd,d≥3.\mathfrak{s}_d= \begin{cases} \mathfrak{sl}_2, & d=1,\\ \mathfrak{sl}_2\oplus\mathfrak{so}_d, & d\ge3, \end{cases} \qquad V_d= \begin{cases} V(m), & d=1,\\ V(m)\otimes\mathbb{C}^{d}, & d\ge3. \end{cases}

where V(m)V(m) is the irreducible sl2\mathfrak{sl}_2-module of highest weight mm. Then both associated nonzero targets are algebraically rigid. The Heisenberg target is the complexified mass central extension of the conformal Galilei algebra of spin m/2m/2; for m=1m=1 this is the Schrödinger family.

Proof. For d≥3d\ge3, the module VdV_d is irreducible, and the tensor product of the alternating form on V(m)V(m) with the symmetric form on Cd\mathbb{C}^{d} is symplectic. For d=1d=1, V(m)V(m) itself is irreducible symplectic. Apply Corollary 3.2 and Theorem 4.2. □

The Heisenberg rigidity assertions in this corollary recover [13 Theorems 5.1 and 5.2], and for m=1m=1 the corresponding results of [4]. In particular, the case of a single odd irreducible sl2\mathfrak{sl}_2-module is already included in that literature. The factor so2\mathfrak{so}_2 is abelian, so the spatial dimension d=2d=2 is outside the semisimple-Levi formulation used here.

Multiplicity-free modules and adjoint cochains

For sl2\mathfrak{sl}_2, let V(n)V(n) denote the irreducible module of highest weight nn. The standard decompositions are

V(m)⊗V(n)=⨁j=0min⁡{m,n}V(m+n−2j).(36)V(m)\otimes V(n)=\bigoplus_{j=0}^{\min\{m,n\}}V(m+n-2j). \tag*{(36)}

and

Λ2V(n)=⨁j=0⌊(n−1)/2⌋V(2n−2−4j).(37)\Lambda^2V(n)=\bigoplus_{j=0}^{\lfloor(n-1)/2\rfloor}V(2n-2-4j). \tag*{(37)}

The second follows from the first because the flip acts on the component V(2n−2j)V(2n-2j) by (−1)j(-1)^j. For n=0n=0 the sum is empty. Also ν(V(n))=(−1)n\nu(V(n)) = (-1)^n, so ⨁nV(n)⊕mn\bigoplus_n V(n)^{\oplus m_n} is symplectic precisely when every even nn has even multiplicity mnm_n.

Corollary 5.3. Let V=⨁i=1rV(ni)V = \bigoplus_{i=1}^{r} V(n_i), where the nin_i are pairwise distinct positive odd integers. Then

H0(gτ,gτ)=0,H1(gτ,gτ)≃H2(gτ,gτ)≃Cr−1,H^0(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau}) = 0,\qquad H^1(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau}) \simeq H^2(\mathfrak{g}_{\tau},\mathfrak{g}_{\tau}) \simeq\mathbb{C}^{r-1},

whereas gη\mathfrak{g}_{\eta} satisfies (32) for every r≥1r \ge1. The extension with a separate Heisenberg center for each V(ni)V(n_i) satisfies (34).

Proof. Apply Proposition 3.1, Theorem 4.2, and Theorem 4.5. □

The adjoint module V(2)V(2) occurs in V(ni)⊗V(nj)V(n_i) \otimes V(n_j) for distinct odd ni,njn_i,n_j precisely when ∣ni−nj∣=2|n_i-n_j|=2. Thus Hom⁡s(Λ2V,s)\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,\mathfrak{s}) can be nonzero in this corollary. Such cochains are eliminated by the Jacobi equation in Lemma 2.9, rather than by a representation-theoretic exclusion. The same phenomenon occurs for other semisimple Levi factors.

Example 5.4. Take s=sl2⊕sl2\mathfrak{s}=\mathfrak{sl}_2\oplus\mathfrak{sl}_2 and

V=(V(1)⊗V(2))⊕(V(3)⊗V(2)).V = (V(1)\otimes V(2))\oplus(V(3)\otimes V(2)).

The two components are irreducible symplectic and nonisomorphic. Their cross tensor product contains V(2)⊗V(0)V(2)\otimes V(0), an adjoint component of s\mathfrak{s}, by (36). Hence Hom⁡s(Λ2V,s)≠0\operatorname{Hom}_{\mathfrak{s}}(\Lambda^2V,\mathfrak{s})\ne0, but both the single-center and separate-center Heisenberg targets are algebraically rigid.

Repeated blocks with separate centers

Repeated irreducible components need not preserve rigidity. We first consider the construction with one center for each copy.

Proposition 5.5. Let UU be a nonzero irreducible symplectic s\mathfrak{s}-module and let

gη(m)=s⋉⨁i=1mh(Ui,ωU),Ui≃U.\mathfrak{g}_{\eta}^{(m)}=\mathfrak{s}\ltimes\bigoplus_{i=1}^{m}\mathfrak{h}(U_i,\omega_U),\qquad U_i\simeq U.

For m≥3m\ge3, this algebra is geometrically nonrigid. Moreover,

dim⁡H2(gη(m),gη(m))≥m(m−1)(m−2)2.\dim H^2(\mathfrak{g}_{\eta}^{(m)},\mathfrak{g}_{\eta}^{(m)})\ge\frac{m(m-1)(m-2)}{2}.

Proof. Write uiu_i for the copy of u∈Uu\in U in UiU_i, and let Zm=⨁iCaiZ_m=\bigoplus_i\mathbb{C}a_i. Choose three distinct indices, say 1,2,31,2,3, and define an alternating map by

β(u1,v2)=ωU(u,v)a3,(38)\beta(u_1,v_2)=\omega_U(u,v)a_3, \tag*{(38)}

with all other types of pairs zero. If μ\mu is the original bracket, then μ+tβ\mu+t\beta is a Lie bracket for every tt: all radical brackets still take values in ZmZ_m, and the new bracket is s\mathfrak{s}-equivariant.

The cocycle β\beta is not a coboundary. By Corollary 2.4 and exactness of taking invariants, it is enough to check invariant one-cochains on the radical. Every such cochain has components in End⁡s(U⊕m)\operatorname{End}_{\mathfrak{s}}(U^{\oplus m}) and End⁡(Zm)\operatorname{End}(Z_m), since UU is nontrivial and Hom⁡s(U,s)=0\operatorname{Hom}_{\mathfrak{s}}(U,\mathfrak{s})=0. On U1∧U2U_1\wedge U_2, its differential takes values only in Ca1⊕Ca2\mathbb{C}a_1\oplus\mathbb{C}a_2: the original bracket of these two blocks is zero, and each remaining term brackets against either U1U_1 or U2U_2. It cannot equal the nonzero a3a_3 component in (38).

If the orbit of μ\mu were open, the curve μ+tβ\mu+t\beta would lie in that orbit for tt in a neighborhood of zero. Its tangent β\beta would then belong to the orbit tangent space, which is the space of two-coboundaries. This proves geometric nonrigidity.

For the dimension estimate, put M=CmM=\mathbb{C}^m. The central-valued invariant two-cochains on U⊗MU\otimes M, extended by zero on inputs in ZmZ_m, are all cocycles and have dimension

dim⁡(S2M∗⊗Zm)=m2(m+1)2.\dim(S^2M^*\otimes Z_m)=\frac{m^2(m+1)}{2}.

The invariant one-cochains have dimension 2m22m^2. Their kernel contains the mm independent block dilations, so the boundary space has dimension at most 2m2−m2m^2-m. Subtracting gives the stated bound. □

The proposition does not assert nonrigidity for two repeated blocks. The number of centers also matters, as the next example illustrates.

Two repeated modules with a shared center

Remark 5.6. For V=⨁nV(n)⊗MnV=\bigoplus_n V^{(n)}\otimes M_n, the Lie algebra of equivariant symplectic endomorphisms is

cω≃⨁n oddso(Mn)⊕⨁n evensp(Mn).\mathfrak{c}_\omega\simeq\bigoplus_{n\ \mathrm{odd}}\mathfrak{so}(M_n)\oplus\bigoplus_{n\ \mathrm{even}}\mathfrak{sp}(M_n).

with the symmetric or alternating multiplicity-space forms from Proposition 2.6. For a sum of odd components, Vsl2=0V^{\mathfrak{sl}_2}=0 and Hom⁡sl2(V,sl2)=Hom⁡sl2(Λ2V,V)=0\operatorname{Hom}_{\mathfrak{sl}_2}(V,\mathfrak{sl}_2)=\operatorname{Hom}_{\mathfrak{sl}_2}(\Lambda^2V,V)=0, so Proposition 4.1 computes H1(gη,gη)H^1(\mathfrak{g}_\eta,\mathfrak{g}_\eta) and gives the kernel formula for H2(gη,gη)H^2(\mathfrak{g}_\eta,\mathfrak{g}_\eta) with arbitrary multiplicities. Repeated odd components introduce adjoint-valued cochains through

Λ2(V(n)⊗Mn)≃(Λ2V(n)⊗S2Mn)⊕(S2V(n)⊗Λ2Mn).\Lambda^2(V^{(n)}\otimes M_n)\simeq(\Lambda^2V^{(n)}\otimes S^2M_n)\oplus(S^2V^{(n)}\otimes\Lambda^2M_n).

since V(2)V^{(2)} occurs in S2V(n)S^2V^{(n)} for every positive odd nn. The kernel formula retains both the centralizer and these adjoint-valued cochains. Even components require separate checks of all the hypotheses of Proposition 4.1.

Example 5.7 (Two copies of the defining module). Let U=C2U=\mathbb{C}^2 with its standard symplectic form ϵ\epsilon, and let V1,V2V_1,V_2 be two copies of UU. Write V=V1⊕V2V=V_1\oplus V_2, with form ω=ϵ⊕ϵ\omega=\epsilon\oplus\epsilon, and denote by ui∈Viu_i\in V_i the copy of u∈Uu\in U. Then

Λ2V≃V(2)⊕V(0)⊕3.\Lambda^2V\simeq V^{(2)}\oplus V^{(0)\oplus3}.

Use a single shared generator aa for both copies, so that [ui,vi]=ϵ(u,v)a[u_i,v_i]=\epsilon(u,v)a and [V1,V2]=0[V_1,V_2]=0. For gη=sl2⋉h(V,ω)\mathfrak{g}_\eta=\mathfrak{sl}_2\ltimes\mathfrak{h}(V,\omega),

dim⁡H0(gη,gη)=1,dim⁡H1(gη,gη)=2,dim⁡H2(gη,gη)=1.\dim H^0(\mathfrak{g}_\eta,\mathfrak{g}_\eta)=1,\qquad\dim H^1(\mathfrak{g}_\eta,\mathfrak{g}_\eta)=2,\qquad\dim H^2(\mathfrak{g}_\eta,\mathfrak{g}_\eta)=1.

Moreover, gη\mathfrak{g}_\eta admits a linear deformation whose nonzero members are isomorphic to sl3(C)\mathfrak{sl}_3(\mathbb{C}). In particular, gη\mathfrak{g}_\eta is not geometrically rigid, whereas every nonzero member is algebraically and geometrically rigid.

To see this explicitly, identify sl2=sp(U)\mathfrak{sl}_2=\mathfrak{sp}(U) and set

σ(u,v)x=ϵ(u,x)v+ϵ(v,x)u.\sigma(u,v)x=\epsilon(u,x)v+\epsilon(v,x)u.

Let T(u1)=u2T(u_1)=u_2 and T(u2)=−u1T(u_2)=-u_1, and define an alternating map ψ\psi by ψ(a,v)=Tv\psi(a,v)=Tv, \quad ψ(u1,v2)=−13σ(u,v),ψ(Vi,Vi)=ψ(sl2,gη)=0\psi(u_1,v_2)=-\frac{1}{3}\sigma(u,v), \quad\psi(V_i,V_i)=\psi(\mathfrak{sl}_2,\mathfrak{g}_\eta)=0.

Here cω=CTc_\omega=CT and Hom⁡sl2(Λ2V,sl2)\operatorname{Hom}_{\mathfrak{sl}_2}(\Lambda^2V,\mathfrak{sl}_2) is spanned by the cross-pair cochain φ(u1,v2)=σ(u,v)\varphi(u_1,v_2)=\sigma(u,v). The two-dimensional symplectic identity

ϵ(u,v)v′+ϵ(v,v′)u+ϵ(v′,u)v=0\epsilon(u,v)v' + \epsilon(v,v')u + \epsilon(v',u)v=0

gives, on a triple with two inputs in the first copy,

J(φ)(u1,v1,v2′)=−3ϵ(u,v)v1′=3K(T)(u1,v1,v2′).J(\varphi)(u_1,v_1,v'_2)=-3\epsilon(u,v)v'_1=3K(T)(u_1,v_1,v'_2).

The same calculation with the copies exchanged gives J(φ)=3K(T)J(\varphi)=3K(T) on all triples. Since K(T)≠0K(T)\ne0, the kernel in (28) is precisely the line represented by ψ=δT−13φ\psi=\delta_T-\frac{1}{3}\varphi. This proves the cohomology dimensions.

The map ψ\psi is itself a Lie bracket. Indeed, ψ(sl2,gη)=0\psi(\mathfrak{sl}_2,\mathfrak{g}_\eta)=0 and ψ(Λ2V)⊂sl2\psi(\Lambda^2V)\subset\mathfrak{sl}_2, so the Jacobi identity on three vectors of VV is automatic. The only remaining condition is ψ(Tu,v)+ψ(u,Tv)=0\psi(Tu,v)+\psi(u,Tv)=0, which follows from the displayed formulas. Here the vector subspace sl2\mathfrak{sl}_2 lies in the center for the bracket ψ\psi alone, not for the original bracket μ\mu. If μ\mu denotes the bracket of gη\mathfrak{g}_\eta, then dψ=0d\psi=0 shows that μ+tψ\mu+t\psi is a linear deformation of gη\mathfrak{g}_\eta. For t≠0t\ne0, put u±=u1∓iu2u_\pm=u_1\mp iu_2 and view ϵ(v,−)\epsilon(v,-) as a row covector. An isomorphism onto sl3(C)\mathfrak{sl}_3(\mathbb{C}) is given, for x∈sl2x\in\mathfrak{sl}_2, by

x+λa+u++v−⟼(x+itλ3Id⁡Uu−4it3ϵ(v,−)−2itλ3).x+\lambda a+u_++v_-\longmapsto \begin{pmatrix} x+\frac{it\lambda}{3}\operatorname{Id}_U & u\\ -\frac{4it}{3}\epsilon(v,-) & -\frac{2it\lambda}{3} \end{pmatrix}.

Indeed, [a,u±]t=±it u±[a,u_\pm]_t=\pm it\,u_\pm, the brackets within either copy {u±:u∈U}\{u_\pm:u\in U\} vanish, and

[u+,v−]t=2ϵ(u,v)a−2it3σ(u,v).[u_+,v_-]_t=2\epsilon(u,v)a-\frac{2it}{3}\sigma(u,v).

The identity u ϵ(v,−)=12(σ(u,v)−ϵ(u,v)Id⁡U)u\,\epsilon(v,-)=\frac{1}{2}\bigl(\sigma(u,v)-\epsilon(u,v)\operatorname{Id}_U\bigr) identifies this mixed bracket with the matrix commutator; the other brackets follow from the sl2\mathfrak{sl}_2-invariance of ϵ\epsilon. Whitehead’s second lemma shows that every nonzero member is algebraically, and hence geometrically, rigid. Its center is zero, whereas Z(gη)=CaZ(\mathfrak{g}_\eta)=\mathbb{C}a, so these members are nonisomorphic to gη\mathfrak{g}_\eta. This proves the geometric nonrigidity of gη\mathfrak{g}_\eta. The cohomology dimensions also agree with the corresponding eight-dimensional entry in [3 Proposition 5.3].

References

FaCET, Universidad Nacional de Tucumán, Argentina

Email address: [email protected]

CIEM – FaMAF, CONICET – Universidad Nacional de Córdoba, Argentina

Guangdong Technion Israel Institute of Technology, China

Email address: [email protected]

References

  1. [1]M. Abu-Hamed and S. Gelaki, Frobenius–Schur indicators for semisimple Lie algebras, J. Algebra 315 (2007), no. 1, 178–191.
  2. [2]D. Burde and F. Wagemann, Sympathetic Lie algebras and adjoint cohomology for Lie algebras, J. Algebra 629 (2023), 381–398.arxiv.org/abs/1908.05963
  3. [3]D. Burde and F. Wagemann, Cohomology of perfect Lie algebras, arXiv:2411.14952 [math.RT] (2024).arxiv.org/abs/2411.14952
  4. [4]R. Campoamor-Stursberg, Cohomological rigidity of the Schrödinger algebra S(N) and its central extension \widetilde{S}(N), J. Lie Theory 27 (2017), no. 2, 315–328.
  5. [5]R. Campoamor-Stursberg, Some structural properties of semidirect sums of so(3) and abelian Lie algebras, Geometry, Integrability and Quantization, Proceedings Series 22 (2021), 88–106.
  6. [6]R. Carles, Sur la structure des algèbres de Lie rigides, Ann. Inst. Fourier (Grenoble) 34 (1984), no. 3, 65–82.DOI
  7. [7]R. Carles, Sur certaines classes d’algèbres de Lie rigides, Math. Ann. 272 (1985), 477–488, doi:10.1007/BF01455861.DOI
  8. [8]R. Carles, Sur la cohomologie d’une nouvelle classe d’algèbres de Lie qui généralisent les sous-algèbres de Borel, J. Algebra 154 (1993), no. 2, 310–334.
  9. [9]F. Grunewald and J. O’Halloran, Deformations of Lie algebras, J. Algebra 162 (1993), no. 1, 210–224.DOI
  10. [10]J. F. Herrera-Granada and P. Tirao, The Grunewald–O’Halloran conjecture for nilpotent Lie algebras of rank ≥ 1, Comm. Algebra 44 (2016), no. 5, 2180–2192.
  11. [11]G. Hochschild and J.-P. Serre, Cohomology of Lie algebras, Ann. of Math. (2) 57 (1953), no. 3, 591–603.DOI
  12. [12]C. Jiang and D. Meng, The realization and structure of complete Lie algebras whose nilpotent radicals are Heisenberg algebras, Chinese Sci. Bull. 43 (1998), 383–385.
  13. [13]A. Khudy... and D. Juman...iyozov, Cohomological rigidity of conformal Galilei algebras and their central extensions, arXiv:2508.17752 [math.RA] (2025).
  14. [14]G. F. Leger, Jr., On cohomology of Lie algebras, Proc. Amer. Math. Soc. 8 (1957), no. 6, 1010–1020.DOI
  15. [15]G. Liu and K. Zhao, The category of weight modules for symplectic oscillator Lie algebras, Transform. Groups 27 (2022), 1025–1044.arxiv.org/abs/1908.04534
  16. [16]D. Meng, Complete Lie algebras and Heisenberg algebras, Comm. Algebra 22 (1994), 5509–5524.DOI
  17. [17]A. Nijenhuis and R. W. Richardson, Jr., Deformations of Lie algebra structures, J. Math. Mech. 17 (1967), no. 1, 89–105.DOI
  18. [18]R. W. Richardson, Jr., On the rigidity of semi-direct products of Lie algebras, Pacific J. Math. 22 (1967), no. 2, 339–344.DOI

Paper details

vlnc:W_m2uvwnrf
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 1

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…
Contents