You are on page 1of 11

TAYLOR SPECTRUM FOR MODULES OVER LIE ALGEBRAS

BORIS BILICH

Abstract. In this paper we generalize the notion of the Taylor spectrum to modules
over an arbitrary Lie algebra and study it for finite-dimensional modules. We show that
the spectrum can be described as the set of simple submodules in case of nilpotent and
semisimple Lie algebras. We also show that this result does not hold for solvable Lie
algebras and obtain a precise description of the spectrum in case of Borel subalgebras of
arXiv:2108.12415v1 [math.RT] 27 Aug 2021

semisimple Lie algebras.

1. Introduction
In 1970, Taylor introduced the notion of the joint spectrum for a finite tuple of com-
muting operators T = (T1 , ..., Tn ) on a Banach space V in [16]. It was defined as the set
of points λ = (λ1 , ..., λn ) ∈ Cn such that a certain Koszul complex associated to λ is not
exact. The Taylor spectrum of a single operator coincides with the classical notion of
spectrum. The same year, Taylor estabilished the existence of the holomorphic functional
calculus in a neighborhood of the spectrum in [17]. In 1972, he proposed a framework for a
noncommutative functional calculus in [18] but the notion of spectrum for non-commuting
tuples of operators was not yet developed.
The first step in this direction was made by Fainshtein in [11]. He generalized the Taylor
spectrum to tuples of operators generating a finite dimensional nilpotent Lie algebra. In [3],
Boasso and Larotonda independently introduced the Taylor spectrum for representations
of solvable Lie algebras. It is given by
σgBL (V ) = {λ ∈ (g/[g, g])∗ : TorUk g (V, Cλ ) 6= 0 for some k ∈ Z≥0 } (1)
where g is a solvable Lie algebra and V is a right Banach g-module.
The spectrum σ BL is equivalent to the definition of Fainshtein in the nilpotent case.
Moreover, we can think about commuting tuples T = (T1 , ..., Tn ) as representations of
the n-dimensional abelian Lie algebra and it turns out that the spectrum for abelian Lie
algebras is equivalent to the original Taylor spectrum.
The Taylor spectrum for Lie algebras was actively studied in works of several authors. A
variety of classical theorems were generalized, such as the projection property and different
kinds of the spectral mapping theorem. The details can be found in the monograph [1]
and in a series of works by Dosi (see for example [7, 8, 9]). In 2010, Dosi estabilished a
noncommutative holomorphic functional calculus in a neighborhood of the spectrum for
supernilpotent Lie algebras of operators [10].
In this paper we introduce the notion of Taylor spectrum for modules over an arbitrary
finite-dimensional Lie algebra and study it for finite-dimensional modules. Let ĝ be the

2020 Mathematics Subject Classification. Primary 17B56, Secondary 17B30, 47A13.


Key words and phrases. Taylor spectrum, Lie algebra cohomology.
1
2 BORIS BILICH

set of isomorphism classes of simple g-modules. The spectrum of a g-module is given as


a subset of ĝ by the non-vanishing property of some homology groups (Definition 3.1).
The article is organized as follows. In Section 2 we recall neccesary facts about Lie
algebra cohomology. In Section 3 we give the definition of the spectrum and study its
properties. We obtain a complete description for the spectrum of finite-dimensional mod-
ules over a semisimple Lie algebra in Theorem 3.5. In this case, the spectrum coincides
with the set of isomorphism classes of simple submodules. This result holds even for
Banach modules as shown in Corollary 3.7. We finish the section with Proposition 3.8,
which shows the behaviour of the spectrum for Lie algebra extensions.
Section 4 is devoted to the spectrum for solvable Lie algebras. Although this case was
studied before (see [1, 3]), the authors considered mostly infinite-dimensional modules.
We explore the spectrum for finite-dimensional modules. It turns out that the spectrum
is hard to compute even for the trivial module. An approximation is given by Theorem
4.1, which happens to be a refolmutaion in terms of the Taylor spectrum of an earlier
result by Millionschikov [15]. We describe completely the spectrum in case of nilpotent
Lie algebras (Theorem 4.4) and Borel subalgebras of semisimple Lie algebras (Theorem
4.7).
Acknowledgments. The author is grateful to his scientific advisor Alexei Pirkovskii for his
careful guidance during writing this paper. Thanks should also go to Egor Alimpiev and
Marat Rovinsky for language corrections to an early draft and to Simon Wadsley and
Yves de Cornulier for useful discussions on the Mathoverflow website (see answers to [2]).

2. Preliminaries
2.1. Notation. In this article all algebras, including Lie algebras, are defined over the
field of complex numbers. We use the notation Ug for the enveloping algebra of a Lie
algebra g. We denote by g-mod and mod-g the categories of left and right g-modules,
respectively. We write ĝ for the set of isomorphism classes of simple finite-dimensional
g-modules and C for the trivial module. If S is a simple g-module we also denote by S ∈ ĝ
its isomorphism class. This will not lead to a confusion.

2.2. Functors on categories of modules. For the rest of this section we denote by g
an arbitrary finite-dimensional Lie algebra. Let V be a g-module. We define vector spaces
V g = {v ∈ V : g · v = 0 ∀g ∈ g},
called invariants and
Vg = V /gV,
called coinvariants. The assignments V 7→ V g and V 7→ Vg define functors from g-mod (or
mod-g) to the category of vector spaces. These functors are isomorphic to HomU g (C, V )
and C ⊗U g V respectively.
We define two functors ∗ : g-modop → g-mod and ◦ : g-mod → mod-g as follows.
The duality functor ∗ maps a g-module V to its dual vector space with the left g-action
given by
(g · f )(v) = −f (g · v), for all f ∈ V ∗ , v ∈ V, g ∈ g.
TAYLOR SPECTRUM FOR MODULES OVER LIE ALGEBRAS 3

The antipode functor ◦ maps V to itself as a vector space with the right g-action given
by
v · g = −g · v, for all v ∈ V, g ∈ g.
The antipode functor defines an equivalence of the categories g-mod and mod-g. We
also denote by ∗ and ◦ the functors on the category of right g-modules, defined in the
same way. It is easy to see that (◦ )◦ is naturally isomorphic to the identity functor.
However, (∗ )∗ is naturally isomorphic to the identity functor only on the subcategory of
finite-dimensional modules.
Another important functor is  ⊗C  : mod-g × g-mod → g-mod. If V ∈ mod-g and
W ∈ g-mod, then V ⊗C W is the tensor product of V and W as vector space with the
action of g, which is determined by the formula
g · (v ⊗ w) = v ⊗ (g · w) − (v · g) ⊗ w, for all w ∈ W, v ∈ V, g ∈ g.
For V, W ∈ g-mod (resp. mod-g), we denote by V ⊗ W the left (resp. right) g-module
V ◦ ⊗C W (resp. V ⊗C W ◦ ).
For V ∈ g-mod and and a simple g-module S ∈ ĝ, we write VS for the g-module S ⊗C V .
If S is one-dimensional, then it can be described by a character λ ∈ (g/[g, g])∗ . In this
case we write Vλ for VS . For example, let C be the trivial g-module. With use of the above
notation Cλ stands for the one-dimensional module with the action given by g · s = λ(g)s
for all s ∈ Cλ and g ∈ g. We also use the notation V−S for the module S ∗ ⊗C V . If S is
one-dimensional of character λ, then V−S is isomorphic to V−λ .
2.3. Lie algebra (co)homology. In this subsection we briefly recall necessary definitions
and properties of Lie algebra cohomology. For details, we refer the reader to any related
textbook (see for example [19, Chapter 7]).
Definition 2.1. Let V be a left g-module. For all k ∈ Z≥0 , the k-th homology of g with
coefficients in V is defined as
Hk (g, V ) = TorUk g (C, V ).
Dually, the k-th cohomology is defined as
H k (g, V ) = ExtkU g (C, V ).
(Co)homologies are functors from the category g-mod to the category of vector spaces.
They can be computed with the Chevalley-EilenbergVfree resolution of the trivial g module
C (see [19, Definition 7.7.1]). It has Fk = Ug ⊗C k g in degree k with the differential
given by
X
p
d(u ⊗ g1 ∧ · · · ∧ gp ) = (−1)i+1 ugi ⊗ g1 ∧ · · · ∧ ĝi ∧ · · · ∧ gp +
i=1
X
+ (−1)i+j u ⊗ [gi , gj ] ∧ · · · ∧ ĝi · · · ∧ ĝj · · · ∧ gp , where u ∈ Ug, gi ∈ g.
i<j
We need the following fact about homology.
Lemma 2.2. Let V ∈ mod-g and W ∈ g-mod. Then
TorUk g (V, W ) ∼ Ug
= Tork (C, V ⊗C W ) = Hk (g, V ⊗C W ),
for all k ∈ Z≥0 .
4 BORIS BILICH

Proof. Since the functors C ⊗U g ( ⊗C W ) and  ⊗U g W are naturally isomorphic, it


suffices to show that if P• → V is a flat resolution of V , then P• ⊗C W is a flat resolution
of V ⊗C W .
By definition, the flatness of Pk means the exactness of the functor Pk ⊗U g . We
use the associativity of the tensor product to obtain a natural isomorphism of functors
(Pk ⊗C W ) ⊗U g  and Pk ⊗U g (W ⊗C ). The last functor is the composition of exact
functors, so it is also exact. If follows that Pk ⊗C W is flat and we obtain the natural
isomorphism TorUk g (V, W ) ∼ Ug
= Tork (C, V ⊗C W ). 
A useful variation of the Poincaré duality holds for finite dimensional LieValgebras. Let
n be the dimension of g. We endow the one-dimensional vector space n g with the
structure of a left g-module, which extends the adjoint action by the Leibnitz rule. That
is,
X
n
g · g1 ∧ · · · ∧ gn = g1 ∧ · · · ∧ [g, gi ] ∧ · · · ∧ gn
i=1
for g, gi ∈ g.
Proposition 2.3 (Poincaré duality). For 0 ≤ k ≤ n and a left g-module V , there are
vector space isomorphisms
^
n
= Hk (g, V )∗ and H k (g, V ) ∼
H k (g, V ∗ ) ∼ = Hn−k (g, ( g)∗ ⊗C V ),
natural in V . Consequently
^
n
H k (g, V ∗ ) ∼
= H n−k (g, g ⊗C V )∗ .
Proof. [14, Theorem 6.10]. 
3. Taylor spectrum of g-modules
3.1. Definition and first properties. Let g be an arbitrary Lie algebra of dimension
n and V be a left g-module. Recall that ĝ is the set of isomorphism classes of simple
finite-dimensional g-modules.
Definition 3.1. The Taylor spectrum of V is the subset of ĝ defined as
σg (V ) = {S ∈ ĝ : TorUk g (S ∗ , V ) 6= 0 for some k ≥ 0}.
We will occasionaly use the notation σ(V ) instead of σg (V ) when it is clear what g is
meant. We will also write the spectrum for the Taylor spectrum.
Suppose that g is a solvable Lie algebra. By Lie’s theorem, every simple g-module is
one-dimensional. Thus, ĝ can be identified with the space (g/[g, g])∗ of characters. In
this case, the Taylor spectrum can be expressed in terms of the spectrum (1), defined by
Boasso and Larotonda in [3]:
σg (V ) = −σgBL (V ◦ ).
We may think about σg as a generilization of σgBL .
The spectrum has other useful characterizations.
Proposition 3.2. For an arbitrary S ∈ g, the following are equivalent:
(1) S ∈ σ(V );
TAYLOR SPECTRUM FOR MODULES OVER LIE ALGEBRAS 5

(2) Hk (g, VV−S ) 6= 0 for some k;


(3) H k (g, Vn g ⊗C V−S ) 6= 0 for some k;
(4) S ∗ ⊗C n g ∈ σ(V ∗ ).
Proof. 1 ⇔ 2 This follows immediatley from Lemma 2.2.
2 ⇔ 3 This is just the Poincaré duality (Proposition 2.3).
3 ⇔ 4 Again, by the Poincaré duality we have
^
n
H (g, g ⊗C V−S )∗ = H n−k (g, (V−S )∗ ).
k

∗ ∼
Vn Vn ∗
Observe
Vn that (V −S ) ∗
is isomorphic to S ⊗C V = g ⊗ C ( g) ⊗C S ⊗ V ∗ ∼ =
∗ V
g ⊗C (V )−S ∗ ⊗C g . Using the equivalence of (1) and (3) we obtain the desired
n

result.

Lemma 3.3. Let V , V ′ , and V ′′ be g-modules such that there is a short exact sequence
0 → V ′ → V → V ′′ → 0.
Then
(1) σg (V ) ⊂ σg (V ′ ) ∪ σg (V ′′ ).
(2) If the short exact sequence splits, i.e., V ∼
= V ′ ⊕ V ′′ then σg (V ) = σg (V ′ ) ∪ σg (V ′′ ).
Proof. The tensor product over C is exact. So for any simple g-module S the sequnce
′ ′′
0 → V−S → V−S → V−S →0
is exact. It induces the homology long exact sequence
∂ ′ ′′ ∂ ′
→ Hk (V−S
− ) → Hk (V−S ) → Hk (V−S )−
→ Hk−1 (V−S )→.
′ ′′
If Hk (V−S ) = Hk (V−S ) = 0 for all k, then Hk (V−S ) are also zero. This proves (1).
For (2) observe that the connecting homomorphisms ∂ are all zero if the exact sequence
′ ′′
splits. So, Hk (V−S ) = Hk (V−S ) ⊕ Hk (V−S ) and σg (V ) = σg (V ′ ) ∪ σg (V ′′ ). 
3.2. Semisimple
Vn Lie algebras. Let g be a semisimple Lie algebra of dimension n. Note
that g is isomorphic to the trivial g-module. It follows from Proposition 3.2 that the
Taylor spectrum of a g-module V can be described as the set
σg (V ) = {S ∈ ĝ : H k (g, V−S ) 6= 0 for some k}.
The next lemma helps us to compute the spectrum for finite-dimensional g-modules.
Lemma 3.4. Let S be a simple g-module. Assume that S ∼ 6= C. Then
H k (g, S) = 0 for all k.
Proof. [19, Theorem 7.8.9] 
If the module is trivial, then, obviously, H 0 (g, C) ∼
= C. The following theorem gives a
complete description of the spectrum for semisimple Lie algebras.
Theorem 3.5. Let V be a finite-dimensional g-module. The Taylor spectrum of V coin-
cides with the set of simple components of V :
σg (V ) = {S ∈ ĝ : S is isomorphic to a submodule of V }. (2)
6 BORIS BILICH

Proof. By Lemma 3.3(2), the spectrum of V is the union of spectra of its simple submod-
ules, so we may assume that V is simple. Let S be a simple g-module. Let V−S = ⊕m i=1 Si
be a simple decomposition of V−S . Then we have H k (g, V−S ) ∼ ⊕m
= i=1 H k
(g, S i ). By Lemma
3.4, H k (g, V−S ) is nonzero for some k if and only if Si ∼
= C for some i. On the other hand,
the module V−S is isomorphic to HomC (S, V ) and (V−S )g ∼ = Homg (S, V ), so it is one-

dimensional if S = V and zero otherwise. It follows that S lies in the spectrum of V if
and only if it is isomorphic to a submodule of V . 
In fact, formula (2) holds even for Banach g-modules. We need the following fact to
prove this.
Proposition 3.6. Any Banach module V over a semisimple Lie algebra g is the union
of its finite-dimensional submodules.
Proof. [1, Corollary 5, §30] 
In other words, V is the colimit of the filtered diagram V of its finite-dimensional
submodules. Recall that the homology and the tensor product commute with filtered
colimits. Thus, we can strengthen Theorem 3.5.
Corollary 3.7. The Taylor spectrum of a Banach g-module V coincides with the set of
isomorphism classes of simple submodules of V .
Proof. The filtered diagram V of modules induces the filtered diagram H∗ (g, V−S ) of
vector spaces. Moreover, H∗ (g, V−S ) is the colimit of H∗ (g, V−S ). Since all the maps in
V are inclusions, so are the maps in H∗ (g, V−S ). Thus, H∗ (g, V−S ) is nonzero if and only
if S ∈ σ(W ) for some W ⊂ V . This means that S ∈ σ(V ) if and only if S is isomorphic
to a submodule of V . 
3.3. Extensions of Lie algebras. Let h be an arbitrary Lie algebra and λ be a character
of h. By a one-dimensional extension of h we mean an exact sequence of Lie algebras
π
0 → Cλ → g −
→ h → 0.
Here Cλ is an ideal of g and the commutator in g satisfies [g, c] = λ(π(g))c for all g ∈ g and
c ∈ Cλ . The fundamental result of the Lie algebra cohomology theory states that isomor-
phism classes of such extensions are in one-to-one correspondence with the set H 2 (h, Cλ ) =
V
Ext2U h (C, Cλ ). We recall how to construct the bijection. Let ξ ∈ HomC ( 2 h, Cλ ) be a
cocycle. We define a Lie bracket on the vector space g = h ⊕ Cλ by
[(h1 , c1 ), (h2 , c2 )]ξ = ([h1 , h2 ], λ(h1 )c2 − λ(h2 )c1 + ξ(h1 ∧ h2 )),
for all hi ∈ h, ci ∈ Cλ . It can be shown that this formula satisfies the axioms of a Lie
bracket and that cohomologous cocycles induce isomorphic extensions (see [19, Theorem
7.6.3]).
We adapt the techniques for computing cohomology of Lie algebra extensions from
[12, Chapter II.§6]. From now on and till the end of the sectionVwe write g for a one-
dimensional extension of h, represented by a cocycle ξ ∈ HomC ( 2 h, Cλ ). We also use
the notation c for the ideal Cλ ⊂ g. Let V be a h-module. We denote by V π the g-module
which is obtained from V via the homomorphism π. Note that if S ∈ ĥ is an ireducible
h-module, then S π is also irreducible as a g-module. We use this to identify ĥ with the
subset of ĝ.
TAYLOR SPECTRUM FOR MODULES OVER LIE ALGEBRAS 7

Assume that we have an h-module V and we want to compute the spectrum σg (V π ).


The following proposition gives us some clues.
Proposition 3.8. Let g and V be as above. Then
(1) σg (V π ) ⊂ ĥ ⊂ ĝ;
(2) If S ∈ σg (V π ), then either S ∈ σh (V ) or S−λ ∈ σh (V );
(3) If the extension is central, i.e., λ = 0, then σg (V π ) = σh (V ).

Proof. Let S ∈ ĝ \ ĥ. It is easy to verify that cS = {c · s : s ∈ S, c ∈ c} is a submodule of


S. Since S is irreducible, it is either a zero submodule or the whole S. If it is zero, then
S is actually an h-module and it contradicts our assumption. Thus, cS = S and Sc = 0.
The same argument can be used to show that S c = 0.
2
The Hochschild-Serre spectral sequence Ep,q = Hp (h, Hq (c, V−S )) converges to Hk (g, V−S )
(see [19, 7.5]). It suffices to prove that Hq (c, V−S ) = 0 for all q ≥ 0. The only possible
nonzero homology groups are H0 (c, V−S ) = (V−S )c ∼ = (S ∗ )c . ⊗C V = 0 and H1 (c, V−S ) =
c ∼
(V−S ) = (S ) ⊗C V = 0 since V is c-invariant. Therefore S is not in σg (V π ) by Proposition
∗ c

3.2. We have proved (1).


Suppose now that S ∈ ĥ. Then the h-modules H0 (c, V−S ) and H1 (c, V−S ) are isomorphic
to V−S and V ⊗C Cλ respectively. Since Hk (c, V−S ) = 0 for k > 1, the Hochschild-Serre
spectral sequence stabilizes on the third page. The differntials in E 2 are dk : Hk (h, V−S ) →
Hk−2 (h, V−S ⊗C Cλ ). The spectral sequence collapses if and only if all the differentials are
isomorphisms of vector spaces.
If dk is not an isomorphism for some k, then either Hk−2(h, V−S ⊗C Cλ ) or Hk (h, V−S )
are nonzero. This means that S or S−λ are in σh (V ). This proves (2).
Suppose that λ = 0. Since S ∼ = S−λ , we have σg (V π ) ⊂ σh (V ) by (2). Let S be in
σh (V ). There is an integer k ≥ 0 such that H k (h, V−S ) 6= 0 and H i (h, V−S ) = 0 for i < k.
The differential dk : Hk (h, V−S ) → Hk−2 (h, V−S ) is not an isomorphism, so Hk (g, V−S ) 6= 0.
This proves (3). 

4. Case of solvable Lie algebras


4.1. Trivial module. In this section g denotes an arbitrary solvable Lie algebra of di-
mension n. By Lie’s theorem, every simple g-module is one-dimensional, so we identify ĝ
with the space (g/[g, g])∗ of characters.
Let V be a finite-dimensional g-module. By weights ω(V ) ⊂ ĝ we mean the set of
diagonal matrix entries in a triangular basis for V . It is independent on the choice of such
basis and can be also described as the set of one-dimensional subquotients of V (Jordan-
Hölder theorem). Consider the adjoint representation ad g ∈ g-mod. The weights ω(ad g)
are called Jordan-Hölder values of g. We denote by 2ρ the sum of all Jordan-Hölder values
with multiplicites. If (g1 , ..., gn ) is a triangular basis of ad g, then we have
g · g1 ∧ · · · ∧ gn = 2ρ(g)g1 ∧ · · · ∧ gn
Vn
for g ∈ g. So, g is a one-dimensional g-module with character 2ρ.
Theorem 4.1. Let g be a solvable Lie algebra of dimension n. If λ ∈ ĝ is in the spectrum
σg (C), then it is the sum of at most n Jordan-Hölder values of g. Moreover, if λ ∈ σg (C),
then 2ρ − λ is also in σg (C).
8 BORIS BILICH

Proof. The proof is by induction on n. The theorem is trivial for n = 1. Assume that
the statement holds for all solvable Lie algebras of dimension n − 1. Choose any one-
dimensional ideal c in g and denote the corresponding character by µ. By Proposition
3.8.(2) we know that the spectrum σg (C) is the subset of {0, µ} + σg/c (C). By induction,
any ν ∈ σg/c (C) is the sum of at most n − 1 Jordan-Hölder values of g/c, which are also
Jordan-Hölder values of g. Since µ is a Jordan-Hölder value of g, we conclude that any
element of the spectrum σg (C) is the sum of at most n Jordan-Hölder values of g.
For the second assertion we use Proposition 3.2.V The trivial module is isomorphic to its
dual, so if Cλ is in the spectrum of C, then C∗λ ⊗C n g ∼
= C2ρ−λ is also in the spectrum. 
Remark. An equivalent statement was first obtained by Millionschikov in [15, Theorem
3.1]. Theorem 4.1 is a slight reformulation of that result in terms of the Taylor spectrum.
Obviously, 0 is always in the spectrum and, hence, so is 2ρ. So, generally, there are
often more than one element in the spectrum of C.
Example 4.2. Let g be the 3-dimensional solvable Lie algebra with basis e1 , e2 , e3 and
the commutator given by [e1 , e2 ] = e2 and [e1 , e3 ] = λe3 for some λ ∈ C. The space of
characters is one-dimensional, so we identify it with C by the evaluation at e1 . Then
σg (C) = {0, 1, λ, 1 + λ}. Indeed, 0 and 2ρ = 1 + λ are always in the spectrum and for 1
and λ the first homology groups are non-vanishing.
4.2. Nilpotent Lie algebras. The following well-known fact in representation theory
for nilpotent Lie algebras is crucial for the computation of the Taylor spectrum.
Lemma 4.3. Let V be a finite-dimensional indecomposable module over nilpotent Lie
algebra g. Then the set of weights ω(V ) consists of one element.
Proof. [4, Chapter VII, Proposition 9] 
We call modules with only one weight monoweighted. Lemma 4.3 shows us that any
finite-dimensional module over a nilpotent Lie algebra can be decomposed in the sum of
monoweighted submodules. We are ready to formulate the main result.
Theorem 4.4. Let g be a nilpotent Lie algebra and V be a finite-dimensional g-module.
Then the spectrum σg (V ) coincides with the set of isomorphism classes of one-dimensional
submodules of V .
Proof. The spectrum of V is the union of spectra of its indecomposable submodules by
Lemma 3.3(2), so we may assume that V is indecomposable. Then by Lemma 4.3, V is
monoweighted. If µ is the only weight of V , then ω(V−µ ) = {0} and σg (V−µ ) = σg (V ) − µ,
so we may additionally assume that ω(V ) = {0}. The proof is by induction on the
dimension of V .
By Engel’s theorem, all Jordan-Hölder values of g are zero. Using Theorem 4.1 we
conclude that the assertion holds for V ∼ = C. If λ is a character of g, then σg (Cλ ) =
σg (C) + λ. This means that the assertion holds for one-dimensional modules.
Suppose now that dim V = m and that the theorem holds for modules of dimension
less than m. Choose a one-dimensional trivial submodule of V . We have a short exact
sequence of modules
0 → C → V → V /C → 0.
TAYLOR SPECTRUM FOR MODULES OVER LIE ALGEBRAS 9

By Lemma 3.3(1), σ(V ) ⊂ σ(C) ∪ σ(V /C). But σ(C) = σ(V /C) = {0} by induction. On
the other hand, 0 is in the spectrum of V since H0 (g, V ) = Vg 6= 0. This completes the
proof. 

4.3. Borel subalgebras of semisimple Lie algebras. Here we use the terminology
from the theory of semisimple Lie algebras. We refer the reader to [13, Chapter 2] for
details.
For the rest of the section s is a semisimple Lie algebra and g is a Borel subalgebra
of s. It is known that g is isomorphic to a semidirect product h ⋉ n, where h is a Cartan
subalgebra of s and n = [g, g]. We denote the root system of s relative to h by ∆ ⊂ h∗ = ĝ.
We write ∆+ ⊂ ∆ for the subset of positive roots. Elements of ∆+ are simply nonzero
Jordan-Hölder values of g. We use the notation W (∆) for the Weyl group. For any
element w ∈ W its length is denoted by l(w).
The aim of the subsection is to describe the spectrum of irreducible s-modules consid-
ered as g-modules. First we describe the cohomology of such modules over n by Kostant’s
theorem. Then we use the formula for the cohomology of a semidirect product to compute
the cohomology over g.
Let V be a s-module of highest weight λ. The cohomology groups H k (n, V ) are naturally
h-modules after a canonical identification h = g/n. Kostant’s theorem gives an explicit
formula for these modules.
Lemma 4.5 (Kostant’s theorem). As an h-module, H k (n, V ) is the sum of one-dimensional
modules with weights
w(λ + ρ) − ρ, w ∈ W (∆), l(w) = k,
where ρ is the half-sum of positive roots (or, equivalently, Jordan-Hölder values).
Proof. [14, Theorem 6.12] 
We say that an abelian Lie algebra h acts torally on an h-module W if W is a direct sum
of one-dimensional submodules. If h is a Cartan subalgebra of a semisiple Lie algebra s,
then h acts torally on any finite dimensional s-module. If g = h⋉n is the Borel subalgebra
of s, then h also acts torally on n.
Lemma 4.6. For a finite-dimensional s-module V we have
M ^
p
k
H (g, V ) = h∗ ⊗C H q (n, V )h
p+q=k

Proof. It is a special case of [6, Theorem 4]. 


The following theorem gives a complete description of the Taylor spectrum for Borel
subalgebras in the special case of modules restricted from a semisimple algebra.
Theorem 4.7. Let s be a semisimple Lie algebra and V be a simple s-module of highest
weight λ. Let g be a Borel subalgebra of s. Then the Taylor spectrum of V considered as
a g-module is given by
σg (V ) = {ρ + w(λ + ρ) : w ∈ W (∆)}, (3)
where ρ is the half-sum of positive roots and W (∆) is the Weyl group.
10 BORIS BILICH

k
Vn
Proof. By Proposition
Vn 3.2, ν ∈ σ g (V ) if and only
Vnif H (g, g⊗C V−ν ) is nonzero for some
k. The module g is isomorphic to C2ρ , so ∼
g ⊗C V−ν = V2ρ−ν . By Lemma 4.6, we
have
M ^ p
M ^ p
k q
H (g, V2ρ−ν ) = ∗
h ⊗C H (n, V2ρ−ν ) = h
h∗ ⊗C (H q (n, V )2ρ−ν )h .
p+q=k p+q=k

It follows that ν ∈ σg (V ) if and only if (H q (n, V )2ρ−ν )h is nonzero for some q. By Lemma
4.5, H q (n, V )2ρ−ν is the sum of one-dimensional modules with weights w(λ+ρ)+ρ−ν with
w ∈ W of length l(w) = q. We conclude that ν ∈ σg (V ) if and only if ν = w(λ + ρ) + ρ
for some w ∈ W (∆). 
Remark. A formula equivalent to (3) was obtained for the trivial module in [5, Section 4.4].
Theorem 4.7 generalizes this to arbitrary s-modules and gives the description in terms of
the Taylor spectrum.

References
[1] Daniel Beltiţă and Mihai Şabac, Lie algebras of bounded operators, Operator Theory: Advances and
Applications, vol. 120, Birkhäuser Verlag, Basel, 2001.
[2] Boris Bilich, Homology of solvable (nilpotent) Lie algebras, URL: https://mathoverflow.net/q/330854
[3] Enrico Boasso and Angel Larotonda, A spectral theory for solvable Lie algebras of operators, Pacific
J. Math. 158 (1993), no. 1, 15–22.
[4] Nicolas Bourbaki, Lie groups and Lie algebras. Chapters 7–9, Elements of Mathematics, Springer-
Verlag, Berlin, 2005.
[5] Leandro Cagliero and Paulo Tirao, Total cohomology of solvable Lie algebras and linear deformations,
Trans. Amer. Math. Soc. 368 (2016), no. 5, 3341–3358.
[6] Vincent E. Coll, Jr. and Murray Gerstenhaber, Cohomology of Lie semidirect products and poset
algebras, J. Lie Theory 26 (2016), no. 1, 79–95.
[7] Anar Dosiev, Ultraspectra of a representation of a Banach Lie algebra, Funct. Anal. Appl. 35 (2001),
no. 3, 226–229.
[8] Anar Dosiev, Algebras of power series of elements of a Lie algebra and the Slodkowski spectra, J.
Math. Sci., New York 124, no. 2, 4886–4908.
[9] Anar Dosiev, Projection property of spectra for solvable Lie algebra representations, Trans. Acad. Sci.
Azerb., Ser. Phys.-Tech. Math. Sci. 23 (2003), no. 1, Math. Mech., 27–32.
[10] Anar Dosi, Taylor functional calculus for supernilpotent Lie algebra of operators, J. Operator Theory
63 (2010), no. 1, 191—216.
[11] Alexander S. Fainshtein, Taylor joint spectrum for families of operators generating nilpotent Lie
algebras, J. Operator Theory 29 (1993), no. 1, 3—27.
[12] Alain Guichardet, Cohomologie des groupes topologiques et des algèbres de Lie, Textes
Mathématiques, 2. CEDIC, Paris, 1980.
[13] James E. Humphreys, Introduction to Lie algebras and representation theory, Second printing, revised.
Graduate Texts in Mathematics, 9. Springer-Verlag, New York-Berlin, 1978.
[14] Anthony W. Knapp, Lie groups, Lie algebras, and cohomology, Mathematical Notes, 34. Princeton
University Press, Princeton, NJ, 1988.
[15] Dmitri V. Millionschikov, Cohomology of solvable Lie algebras, and solvmanifolds. (Russian), Mat.
Zametki 77 (2005), no. 1, 67–79; translation in Math. Notes 77 (2005), no. 1-2, 61–71
[16] Joseph L. Taylor, A joint spectrum for several commuting operators, J. Functional Analysis 6 (1970),
172–191.
[17] Joseph L. Taylor, The analytic-functional calculus for several commuting operators, Acta Math. 125
(1970), 1—38.
[18] Joseph L. Taylor, A general framework for a multi-operator functional calculus, Advances in Math.
9 (1972), 183-–252.
TAYLOR SPECTRUM FOR MODULES OVER LIE ALGEBRAS 11

[19] Charles A. Weibel, An introduction to homological algebra, Cambridge Studies in Advanced Mathe-
matics, 38. Cambridge University Press, Cambridge, 1994.

Department of Mathematics, HSE University, Usacheva str. 6, 119048, Moscow, Rus-


sian Federation
Email address: bilichboris1999@gmail.com

You might also like