2404 19439v1

INVARIANT DIVISORS AND EQUIVARIANT LINE BUNDLES
BORIS KRUGLIKOV† AND EIVIND SCHNEIDER†
Abstract. Scalar relative invariants play an important role in the theory of group actions on a
manifold as their zero sets are invariant hypersurfaces. Relative invariants are central in many
arXiv:2404.19439v1 [math.DG] 30 Apr 2024
applications, where they often are treated locally since an invariant hypersurface may not be a
locus of a single function. Our aim is to establish a global theory of relative invariants.
For a Lie algebra g of holomorphic vector fields on a complex manifold M , any holomorphic
g-invariant hypersurface is given in terms of a g-invariant divisor. This generalizes the classical
notion of scalar relative g-invariant. Any g-invariant divisor gives rise to a g-equivariant line
bundle, and a large part of this paper is therefore devoted to the investigation of the group
Picg (M ) of g-equivariant line bundles. We give a cohomological description of Picg (M ) in terms
of a double complex interpolating the Chevalley-Eilenberg complex for g with the Čech complex
of the sheaf of holomorphic functions on M .
We also obtain results about polynomial divisors on affine bundles and jet bundles. This
has applications to the theory of differential invariants. Those were actively studied in relation
to invariant differential equations, but the description of multipliers (or weights) of relative
differential invariants was an open problem. We derive a characterization of them with our
general theory. Examples, including projective geometry of curves and second-order ODEs, not
only illustrate the developed machinery, but also give another approach and rigorously justify
some classical computations. At the end, we briefly discuss generalizations of this theory.
Contents
1. Introduction 2
1.1. Background on relative invariants 2
1.2. A setup for global invariants 2
1.3. Overview of the novel results 3
2. Analytic invariant divisors and equivariant line bundles 5
2.1. Picard group and multipliers 6
2.2. A double complex 8
2.3. The equivariant Picard group 10
2.4. Line bundles admitting a transversal lift 16
2.5. Lie group vs Lie algebra approach 19
3. Invariant polynomial divisors on algebraic bundles 22
3.1. Lie algebra action on affine bundles 22
3.2. Lie algebra action on jet bundles 23
3.3. Example A: Three-dimensional Heisenberg algebra on the plane 25
3.4. Example B: Invariant divisors of curves in the projective plane 26
3.5. Example C: Second-order ODEs modulo point transformations revisited 30
4. Outlook 32
References 33
1
2 BORIS KRUGLIKOV AND EIVIND SCHNEIDER
1. Introduction
1.1. Background on relative invariants. Consider a manifold M together with a Lie group G
acting on M . Let F(M ) the algebra of functions on M and F(M )× the multiplicative subgroup
of nonvanishing functions. The action of g ∈ G on M induces the pullback action g∗ on F(M ).
A (scalar) relative invariant is a function R ∈ F(M ) satisfying
g∗ R = Λ(g)R ∀g ∈ G,
for some map Λ : G → F(M )× , called the multiplier, or weight, of R. If g ⊂ D(M ) denotes the
Lie algebra of vector fields on M corresponding to the Lie group action, then R also satisfies
X(R) = λ(X)R ∀X ∈ g,
for some (infinitesimal) multiplier λ ∈ g∗ ⊗ F(M ), or weight, of R. It follows from the definition
that the locus {R = 0} ⊂ M is G-invariant (resp. g-invariant).
In the case Λ = 1 (resp. λ = 0), the function R is called an absolute invariant, and each
level set {R = const} ⊂ M is invariant, so that we get an invariant foliation of M . Absolute
invariants are well understood in several different settings, see [30, 22, 24, 28] for the classical
invariant theory and [23, 18] for its differential counter-part.
For example, in the case of a regular smooth Lie group action on a smooth manifold, locally
by the Frobenius theorem, the number of functionally independent absolute invariants is equal
to the codimension of an orbit, and orbits are locally separated by that many invariants (see,
for example, Chapter 2 of [23]). In the case of an algebraic group action on an algebraic variety,
globally by the Rosenlicht theorem, orbits in general position are separated by rational absolute
invariants, and the number of algebraically independent rational absolute invariants is equal to
the codimension of a generic orbit (see, for example, Chapter 13 of [28]).
Relative invariants with nontrivial weight are less understood, although they appear in many
important applications (we refer to the introduction to [7] and also to the more recent [25]). In
particular, they are often used to describe g-invariant hypersurfaces containing singular orbits.
An infinitesimal multiplier λ is a 1-cocycle of the Chevalley-Eilenberg complex of g with coeffi-
cients in F(M ). Relationships between the weights of relative (differential) invariants and the
Chevalley-Eilenberg cohomology was discussed in [4, 23]. The question of realizability of a given
cocycle as the weight of some relative invariant was answered locally in the case of a regular
smooth G-action and F(M ) = C ∞ (M ) by M. Fels and P. Olver ([7] and [23, Th. 3.36]), also in
the context of vector-valued relative invariants. In the general case the answer is not known.
Note that rescaling of R by a non-zero function ef , f ∈ F(M ), changes λ by a coboundary
df , which naturally associates the Chevalley-Eilenberg cohomology class [λ] ∈ H1 (g, F(M )) to
the (equivalence class of the) relative invariant R. A proper version of this cohomology will be
central in our work.
1.2. A setup for global invariants. In general, the description of invariant hypersurfaces
(analytic subvarieties of codimension 1) by relative invariants works only locally: there exist
invariant hypersurfaces that cannot be described globally as the zero locus of a relative invariant.
In this paper we restrict to holomorphic actions on complex manifolds, where this problem can
be solved using the language of divisors. Some results extend to real analytic and algebraic
situations, but smooth versions of our global results in general are not available. Thus we
specialize our algebra of functions F(M ) to consist of holomorphic functions, and we will work
with the sheaf O = OM of such functions on a complex manifold M .
In most of the paper we will concentrate on the infinitesimal (Lie algebra) picture as it is
conceptually simpler and lends itself well to computations. Moreover, it is more general, as a
Lie group action always gives rise to a Lie algebra of (complete) vector fields, but not every Lie
3
algebra action can be integrated (the manifold M is not assumed compact; the Lie algebra may
be infinite-dimensional). It should be noted that for algebraic groups G (as well as for compact
Lie groups) the equivariant line bundles have been well studied, see [22, Ch. 1.3] and [3, §4.2]
for the definition and properties of the G-equivariant Picard group PicG (M ) in the context of
algebraic geometry. Our setup is more general, and we present the corresponding theory for Lie
groups in Section 2.5. The main object of study, however, will be the Picard group Picg (M ) of
g-equivariant line bundles defined for any Lie algebra g of holomorphic vector fields on M .
A divisor D on M is given by a collection of meromorphic functions fα defined on each chart
in an open cover {Uα } of M (if the functions fα are holomorphic, then D is called effective).
The functions fα are required to be consistent, in the sense that the zeros and poles of fα and
fβ agree on Uα ∩ Uβ , which is equivalent to fα /fβ being a nonvanishing holomorphic function on
Uα ∩ Uβ . (Our D correspond to Cartier divisors, which are equivalent to Weyl divisors for the
nonsingular analytic varieties we consider.) Analytic hypersurfaces of a complex manifold M
are given locally by the vanishing of a holomorphic function and globally by an effective divisor.
If g is a Lie algebra of vector fields on M and N ⊂ M is a g-invariant hypersurface defined
by the divisor D = {fα }, then each vector field of g is tangent to N , implying that for each α
X(fα ) = λα (X)fα ∀X ∈ g,
for some weight λα ∈ g∗ ⊗ O(Uα ), which is a 1-cocycle in the Chevalley-Eilenberg complex of g

with coefficients in the g-module O(Uα ) of holomorphic functions on Uα ⊂ M . Such a divisor is
called g-invariant. Multiplying each fα by nonvanishing holomorphic functions gives a different
representative of the same divisor, and the weight λα is in this case changed by a coboundary, so
the weights can be identified with elements in the Chevalley-Eilenberg cohomology H1 (g, O(Uα ))
or, more precisely, a slightly modified version thereof. A collection of such weights, or multipliers,
for each element of the cover {Uα }, that are compatible on overlaps, yields a multiplier group
that we will denote Mg (M ). Below we will define it in terms of a certain double complex.
As is well known, any divisor D on M gives rise to a line bundle [D] → M , with transition
functions gαβ = fα /fβ , on which fα are local defining functions of a particular section (and,
geometrically, D is the locus of this section). When D is g-invariant, then there exists a lift of
the Lie algebra g ⊂ D(M ) to a Lie algebra gλ ⊂ D([D]) defined locally in terms of the weights
λ = {λα } of D, meaning that ([D], gλ ) is a g-equivariant line bundle. Properly localized, the
obstruction for such a lift, and thus for the existence of invariant divisors, belongs in general to
the equivariant Picard group Picg (M ).
1.3. Overview of the novel results. Due to a close relationship between g-invariant divi-
sors and g-equivariant line bundles, Section 2.1 starts with an investigation of prerequisites for
the latter. The Picard group Pic(M ), consisting of holomorphic line bundles over M up to
equivalence, is isomorphic to the Čech cohomology group Ȟ1 (M, O× ).
In order to describe the group Picg (M ) of g-equivariant line bundles, we unite the Čech
complex with the Chevalley-Eilenberg complex into a double complex C •,• . The direct limit of
the first total cohomology of this complex (also called hypercohomology, cf. [12]) is exactly the
desired group: Picg (M ) := lim H1 (Tot• (C)).
−→
There exist natural homomorphisms Φ1 : Picg (M ) → Pic(M ) and Φ2 : Picg (M ) → Mg (M ).
The image of ̟ := Φ1 × Φ2 in Pic(M ) × Mg (M ) defines the reduced Picard group
̟
Picg (M ) −→ Picred
g (M ) → 0,
whence a double homomorphism (Ψ1 , Ψ2 ) such that Ψi ◦ ̟ = Φi and ker Ψ1 ∩ ker Ψ2 = 0:
Picred
g (M )
Ψ1 Ψ2
Pic(M ) Mg (M )
Theorem 1. The group Tg (M ) := ker(̟) of equivariant line bundles with trivial reduction is
defined by (7) and consists of the global lifts of g to the trivial line bundle over M that are locally
trivial, modulo globally trivial lifts.
When Tg (M ) = 0, Φ1 × Φ2 embeds Picg (M ) in Pic(M ) × Mg (M ) (Corollary 3 gives two

sufficient conditions for this); generally the same is true for Picg (M )/Tg (M ).
The homomorphisms Ψ1 , Ψ2 (and likewise Φ1 , Φ2 ) are neither injective nor surjective, in
general. We will describe ker(Ψi ) and im(Ψi ) in terms of the iterated cohomology of the double
complex C •,• . In particular, we will show that under certain topological conditions, if the
isotropy algebra gp of a generic point p ∈ M is perfect, then ker(Ψ1 ) = 0 and Picg (M ) ⊂ Pic(M ).
This is an infinitesimal version of Proposition 1.4 from [22], which gives sufficient conditions for
an algebraic group G to admit at most one linearization on any line bundle. The following
statements elaborate on the cases considered in [22] and [7] respectively.
Corollary 1. (i) If Mg (M ) = 0 and Tg (M ) = 0 then Φ1 : Picg (M ) → Pic(M ) is injective.

(ii) Likewise, if Pic(M ) = 0 and Tg (M ) = 0 then Φ2 : Picg (M ) → Mg (M ) is injective.
In Section 2.4 we consider the homomorphism
jg : Divg (M ) → Picg (M )
mapping a g-invariant divisor D with weight λ to the g-equivariant line bundle ([D], gλ ). The
non-equivariant map j : Div(M ) → Pic(M ), which takes D to [D], is well-understood: its kernel
and cokernel are given by exact sequence (10); for smooth projective varieties j is epimorphic
and Pic(M ) corresponds to the class group Cl(M ) of equivalent divisors, cf. [11]. In contrast,
even in the smooth projective case, the map jg is generally neither injective nor surjective.
We will give a necessary criterion for a g-equivariant line bundle (L → M, ĝ) to be the image
of a g-invariant divisor, namely that generic ĝ-orbits on L project bijectively (in our setup:
biholomorphic) to g-orbits on M (projection may be non-injective on singular orbits). We call
such Lie algebras transversal, borrowing the terminology from [1], although their notion of
transversality was a slightly stronger requirement.
Theorem 2. If D = {fα } is a g-invariant divisor and λ = {λα } is the corresponding weight,

then the lift gλ ⊂ D([D]) defined by λ is transversal.
Thus if ĝ ⊂ D(L) is not transversal, then the g-equivariant line bundle (L → M, ĝ) is not in
im(jg ). The condition (L, ĝ) ∈ im(jg ) restricts not only ĝ, but also L via im(Ψ1 ◦ jg ) ⊂ im(j).
The proof of Theorem 2 is based on a local argument and is similar to that of [23, Th. 3.36]
and [7, Th. 5.4], where lifts of g to the trivial bundle are considered. It is important to note
that in our general setting, contrary to the local regular settings of [7, 23], this criterion is only
necessary but not sufficient, which will be illustrated in examples. Yet, in an algebraic context
the converse statement holds true, up to an integer factor for the degree (see Theorem 12).
In Section 2.5 we show that the group of G-equivariant line bundles can be described by
a certain Lie group cohomology with coefficients in the sheaf O× , which combines the Čech
cohomology of O× and the continuous Lie group cohomology with coefficients in the G-module
5
O× (M ). This in turn is related to the equivariant Picard group PicG (M ), studied before in
particular situations when G is algebraic or compact. We also discuss its relation to Picg (M ).
Several examples of computation are spread throughout Section 2, demonstrating global con-
straints in the theory of g-invariant divisors and g-equivariant line bundles. For instance, when
M = CP 1 with the standard coordinate charts U0 , U∞ ⊂ CP 1 , and g = aff(1, C) is the 2-
dimensional Lie subalgebra of sl(2, C) ⊂ D(M ), then
Picg (U0 ) ≃ H1 (g, O(U0 )) = C, Picg (U∞ ) ≃ H1 (g, O(U∞ )) = C2 .
The isomorphism between the group of g-equivariant line bundles and the Chevalley-Eilenberg
cohomology group follows from the fact that all line bundles over C are trivial. On CP 1 , on the
other hand, there are only countably many line bundles, namely OCP 1 (k) for k ∈ Z. In this case
Picg (CP 1 ) = C × Z, where Z = Pic(CP 1 ). However, not all g-equivariant line bundles are of the
form [D] for some g-invariant divisor D. Instead, as a consequence of the necessary criterion of
Theorem 2, we have Divg (CP 1 ) = Z. For more details, see Example 5.
In Section 3 we focus on the important cases of projectable Lie algebras of vector fields on
affine bundles and on jet bundles. In these situations one can consider divisors whose restriction
to fibers are polynomial. Let ĝ be a projectable Lie algebra of vector fields on the total space of
an affine bundle π : E → M that preserves the affine structure on E, and let g = dπ(ĝ) ⊂ D(M ).
Theorem 3. If D is a ĝ-invariant polynomial divisor on the affine bundle E, then [D] = π ∗ L

for some g-equivariant line bundle L ∈ Φ1 (Picg (M )).
In other words, the ĝ-equivariant line bundle over E corresponding to a ĝ-invariant polynomial
divisor is the pullback of a g-equivariant line bundle over M . The same idea works for jet bundles
because the bundle πk+1,k : Jk+1 → Jk for k ≥ 1 has a natural affine structure in fibers. (For
jet spaces of sections of line bundles with the contact transformation algebra, the natural affine
structure in fibers starts at k = 2, with the corresponding modification of the claim.)
Theorem 4. Let g(k) ⊂ D(Jk ) be the prolongation of a Lie algebra g of point transformations
on J0 , 0 < k ≤ ∞. If D is a g(k) -invariant divisor that is polynomial in fibers of πk,1 : Jk → J1 ,
∗ L for some g(1) -equivariant line bundle L ∈ Φ (Pic
then [D] = πk,1 1
1 g(1) (J )).
This result provides our main application for classification of global relative invariants of
the prolonged g-action on J∞ , which is an essential step in the classification of all invariant
differential equations (see [20] for a series of examples of this technique). We note that while
the Gelfand-Fuks type cohomology H 1 (g(∞) , F(J∞ )) may be large and hard to compute, the
theorem reduces the problem to finite dimensions. To illustrate this, we will show how this
allows to effectively treat relative differential invariants of curves in CP 2 under the action of
the Möbius algebra of projective transformations as well as relative differential invariants of
second-order ODEs under the infinite-dimensional Lie algebra of point transformations.
In this paper we concentrate on the complex analytic and complex algebraic situation, using
notation CP n instead of Pn to stress that a part of our results extend to the real analytic and
real algebraic case, with examples like real projective spaces RP n , real jet spaces J∞ , etc. In
particular, examples A-C may be treated in the real context.
2. Analytic invariant divisors and equivariant line bundles

Let g ⊂ D(M ) denote a Lie algebra of holomorphic vector fields on the complex manifold
M . For M = Cn it is well-known that lifts of a Lie algebra g to the trivial line bundle M × C
are parametrized by the Chevalley-Eilenberg cohomology H1 (g, O(M )). This is also the space,
where weights of relative g-invariants take values. We refer to [6, 9] for the general Lie algebra
cohomology theory, to [4] for its relation to relative (differential) invariants, and to [7, 27] for a
relation to lifts.
The goal of this section is to generalize these results to arbitrary holomorphic line bundles
over complex manifolds, and replace the notion of relative g-invariant functions with g-invariant
divisors on M .
2.1. Picard group and multipliers. Let us start with a quick overview of holomorphic line
bundles, sufficient for our purpose (see [11, 15]). For an open subset U ⊂ M denote by O(U )
the space of holomorphic functions on U , and by O× (U ) the subspace of nonvanishing functions.
The corresponding sheaves on M are denoted by O and O× , respectively. Let π : L → M be a
line bundle and consider an open cover U = {Uα } of M that trivializes π, i.e., π −1 (Uα ) ≃ Uα ×C.
The line bundle is uniquely determined by its transition functions gαβ ∈ O× (Uα ∩ Uβ ), which
satisfy gαβ gβγ = gαγ . Two collections of transition functions {gαβ }, {g̃αβ } define the same
bundle if and only if g̃αβ = ffαβ gαβ for some functions fα ∈ O× (Uα ).
This leads to a description of of line bundles in terms of Čech cohomology. Define the complex
δ0 δ1
Y Y Y
0−
→ O× (Uα ) −→ O× (Uα ∩ Uβ ) −→ O× (Uα ∩ Uβ ∩ Uγ ) −
→ ··· ,
α α6=β α6=β6=γ6=α
with differentials given by

q+1
(−1)i+1
Y Y
q
(δ µ)α0 ···αq+1 = µα0 ···α̂i ···αq+1 , µ = {µα0 ···αq } ∈ O× (Uα0 ∩ · · · ∩ Uαq ).
Uα0 ∩···∩Uαq+1
i=0 α0 ,··· ,αq
In particular, δ0 and δ1 are defined in the following way:

Y
(δ0 µ)αβ = µα /µβ , µ = {µα } ∈ O× (Uα ),
α
1 ναγ Y
(δ ν)αβγ = , ν = {ναβ } ∈ O× (Uα ∩ Uβ ).
ναβ νβγ
α6=β
The first Čech-cohomology with respect to the fixed open cover U , defined by Ȟ1 (U , O× ) =
ker(δ1 )/im(δ0 ), is the group of transition functions on U modulo the above equivalence relation.
The Picard group Pic(M ) of equivalence classes of holomorphic line bundles over M can be
described in terms of this cohomology group as follows:
• If all line bundles are trivializable on the open charts in U (for instance, each Uα is
biholomorphic to a polydisc with a possible factor C× ) then Pic(M ) ≃ Ȟ1 (U , O× ).
• In general Pic(M ) ≃ Ȟ1 (M, O× ) := lim Ȟ1 (U , O× ) is the direct limit as U becomes finer.
−→
In both cases, the identification is a group isomorphism. In particular, if the conditions of
Leray’s theorem hold, the first description is applicable (see [11, p.40] or the simpler Theorem
12.8 of [8], which will usually be sufficient for us).
Definition 1. A lift of g ⊂ D(M ) to the line bundle π : L → M is a Lie algebra ĝ ⊂ Dproj (L) of
projectable vector fields, such that dπ : ĝ → g is a Lie algebra isomorphism and ĝ commutes with
the natural vertical vector field u∂u (u is a linear fiber coordinate). The pair (π, ĝ) is called a
g-equivariant line bundle. (We also refer to π or L as a g-equivariant bundle when a lift exists.)
For instance, the canonical line bundle KM = Λdim M T ∗ M (see [15, Ch. 2.2]) always admits
a canonical lift of g ⊂ D(M ). Thus it is an (often nontrivial) g-equivariant line bundle.
In general, the lift of a vector field X ∈ g can be defined on π −1 (Uα ) ≃ Uα × C by
X̂|Uα = X|Uα + λα (X)u∂u , λα ∈ g∗ ⊗ O(Uα ),
7
similar to formula (4.1) in [7]. To simplify notation, we will write X instead of X|Uα when there
is no room for confusion. The condition [X̂, Ŷ ] = [X, \ Y ] for each X, Y ∈ g implies that λα
satisfies
X(λα (Y )) − Y (λα (X)) = λα ([X, Y ]), ∀X ∈ g. (1)
Changing the coordinate function on the fiber, v = eµα u for some function µα ∈ O(Uα ), gives
X + λα (X) u∂u = X + (λα (X) + X(µα )) v∂v .
In this sense, two lifts λα , λ̃α on Uα are equivalent if and only if there exists a µα satisfying
λ̃α (X) = λα (X) + X(µα ), ∀X ∈ g. (2)
The conditions (1) and (2) can be interpreted in terms of Lie algebra cohomology of g with
coefficients in the g-module O(Uα ). Consider the Chevalley-Eilenberg complex
d0 d1
→ O(Uα ) −→ g∗ ⊗ O(Uα ) −→ Λ2 g∗ ⊗ O(Uα ) −
0− → ···
where the maps d0 and d1 are given by
(d0 µα )(X) = X(µα ), µα ∈ O(Uα ),
(d1 λα )(X, Y ) = X(λα (Y )) − Y (λα (X)) − λα ([X, Y ]), λα ∈ g∗ ⊗ O(Uα ),
for X, Y ∈ g (see [6]). Notice that Hom(g, F ) = g∗ ⊗ F when one of the factors is finite-
dimensional. If both factors are infinte-dimensional, a completion of the tensor product is
required. We omit this from the notation, understanding by default that Λi g∗ ⊗ F may stand
for Hom(Λi g∗ , F ) here and below.
Define the cohomology groups
H0 (g, O(Uα )) = ker(d0 ), Hi (g, O(Uα )) = ker(di )/im(di−1 ), i > 0.
It is clear that λα ∈ g∗ ⊗ O(Uα ) defines a lift of g to Uα × C if and only if d1 λα = 0. Furthermore,
two cocycles λα , λ̃α define equivalent lifts if and only if λ̃α = λα + d0 µα for some µα ∈ O(Uα ).
Thus, equivalence classes of lifts of g|Uα to Uα ×C are in one-to-one correspondence with elements
in H1 (g, O(Uα )). (Note also that H0 (g, O(Uα )) = O(Uα )g consists of g-invariants.)
Remark 1. If Uα is a polydisc for each α, then any function in O× (Uα ) is of the form eµ ,
and the argument above works. If Uα is a general open set, one replaces eµα fα with µα fα ,
where µα ∈ O× (Uα ). Then the local lifts are in one-to-one correspondence with elements in the
cohomology group of the complex
d0 log d1
→ O× (Uα ) −−−→ g∗ ⊗ O(Uα ) −→ Λ2 g∗ ⊗ O(Uα ) −
0− → ··· . (3)
We will use this slightly modified complex below with the notation
ker(d1 )
H̃1 (g, O(Uα )) = .
im(d0 log)
Elements in H̃1 (g, O(Uα )) yield local lifts of g to π −1 (Uα ) that may not glue together to a global
lift on L. On Uα ∩ Uβ a lift is given by both X + λα (X) uα ∂uα and X + λβ (X) uβ ∂uβ . The fiber
coordinates relate on overlaps by uα = gαβ uβ , where the transition functions {gαβ } represent an
element of Ȟ1 (U , O× (M )). Thus X + λα (X) uα ∂uα becomes X + (λα (X) − X(gαβ )/gαβ ) uβ ∂uβ ,
resulting in the following compatibility condition on Uα ∩ Uβ :
X(gαβ )
λα (X) − λβ (X) = = X(log gαβ ), ∀X ∈ g. (4)
gαβ
2.2. A double complex. To better understand the compatibility condition, consider the double
complex
d0,0 d1,0 d2,0
C 0,0 C 1,0 C 2,0
δ0,0 δ1,0 δ2,0
d0,1 d1,1 d2,1
C 0,1 C 1,1 C 2,1
δ0,1 δ1,1 δ2,1
d0,2 d1,2 d2,2
C 0,2 C 1,2 C 2,2
δ0,2 δ1,2 δ2,2
where C p,q are given by

Y
C 0,q = O× (Uα0 ∩ · · · ∩ Uαq ),
α0 ,··· ,αq
Y
C p,q = Λp g∗ ⊗ O(Uα0 ∩ · · · ∩ Uαq ), p ≥ 1,
α0 ,··· ,αq
and the differentials δp,q : C p,q → C p,q+1 and dp,q : C p,q → C p+1,q are defined for p = 0 by
q+1
(−1)i+1
Y
0,q
(δ µ)α0 ···αq+1 = µα0 ···α̂i ···αq+1 ,
Uα0 ∩···∩Uαq+1
i=0
X(µα0 ···αq )
(d0,q µα0 ···αq )(X) = X(log µα0 ···αq ) = ,
µα0 ···αq
and for p > 0 by

q+1
X
(δp,q µ)α0 ···αq+1 = (−1)i+1 µα0 ···α̂i ···αq+1 ,
Uα0 ∩···∩Uαq+1
i=0
p
X
p,q
(d µα0 ···αq )(X0 , . . . , Xp ) = (−1)i Xi (µα0 ···αq (X0 , . . . , Xi−1 , Xi+1 , . . . Xp ))
i=0
X
i+j
+ (−1) (µα0 ···αq ([Xi , Xj ], X0 , . . . , Xi−1 , Xi+1 , . . . , Xj−1 , Xj+1 , . . . , Xp )).
i<j
We will sometimes write C p,q (g, U ) for precision when there would otherwise be ambiguity.
The horizontal lines (q fixed) are nearly Chevalley-Eilenberg complexes of g with coefficients
in the g-modules O(Uα ), O(Uα ∩ Uβ ), etc; however (C 0,q , d0,q ) are adjusted in accordance with
Remark 1. The vertical lines (p fixed) are Čech complexes with respect to the open cover U .
Remark 2. For C 0,q it is natural to use multiplicative notation (with identity element 1) while
for C p,q for p ≥ 1 it is better to use additive notation (with identity element 0). Using these
notations consistently becomes difficult when we are dealing with this double complex, and even
more so when we work with the total complex defined below. We will therefore use 0 to denote
the identity element in these groups, and in the corresponding cohomology groups.
The total complex corresponding to the double complex C •,• is defined as follows:
Y X
Totr (C) = C p,q , ∂r = (dp,q + (−1)p δp,q ) : Totr (C) → Totr+1 (C).
p+q=r p+q=r
9
The identity ∂ i+1 ◦ ∂ i = 0 expresses the fact that the double complex is a commutative diagram.
The cohomology groups of the total complex are defined in the usual way:
ker(∂ i )
H0 (Tot• (C)) = ker(∂ 0 ), Hi (Tot• (C)) = .
im(∂ i−1 )
The double complex also gives us several complexes of cohomology groups. The cohomology
groups with respect to di,j (with j fixed) make up the following complexes:
H0d (C •,0 ) H1d (C •,0 ) H2d (C •,0 )

δ∗0,0 δ∗1,0 δ∗2,0
H0d (C •,1 ) H1d (C •,1 ) H2d (C •,1 )

δ∗0,1 δ∗1,1 δ∗2,1
H0d (C •,2 ) H1d (C •,2 ) H2d (C •,2 )

δ∗0,2 δ∗1,2 δ∗2,2
Simultaneously, the cohomology groups with respect to δi,j (with i fixed) also give complexes:
d0,0 d1,0 d2,0

H0δ (C 0,• ) ∗
H0δ (C 1,• ) ∗
H0δ (C 2,• ) ∗
d0,1 d1,1 d2,1

H1δ (C 0,• ) ∗
H1δ (C 1,• ) ∗
H1δ (C 2,• ) ∗
d0,2 d1,2 d2,2

H2δ (C 0,• ) ∗
H2δ (C 1,• ) ∗
H2δ (C 2,• ) ∗
Thus we get the induced cohomology groups

ker(δ∗j,i ) ker(di,j
∗ )
Hiδ (Hjd (C •,• )) = , Hid (Hjδ (C •,• )) = ,
im(δ∗j,i−1 ) im(d∗i−1,j )
for i ≥ 1 and
H0δ (Hjd (C •,• )) = ker(δ∗j,0 ), H0d (Hjδ (C •,• )) = ker(d0,j
∗ ).
In the general setting of the total complex, we have the two projections (homomorphisms)
H1 (Tot• (C))
Φ1 Φ2
H1δ (C 0,• ) H1d (C •,0 )
defined by
Φ1 ([(g, λ)]) = [g], Φ2 ([(g, λ)]) = [λ],
where [(g, λ)] denotes the equivalence class of (g, λ) ∈ ker(∂ 1 ). We use the notation g = {gαβ }
for an element in C 0,1 and λ = {λα } for an element in C 1,0 , and note that
Y
H1δ (C 0,• ) = Ȟ1 (U , O× ), H1d (C •,0 ) = H̃1 (g, O(Uα )).
α
Lemma 1. The maps Φ1 , Φ2 have the following kernels:

ker(Φ1 ) ≃ H1d (H0δ (C •,• )), ker(Φ2 ) ≃ H1δ (H0d (C •,• )).
Proof. The arguments for the two isomorphisms are similar to each other, so we prove the
statement only for ker(Φ2 ). If [(g, λ)] ∈ ker(Φ2 ), then there exists an element g̃ ∈ ker(δ0,1 ) such
that [(g̃, 0)] = [(g, λ)]. We have d0,1 g̃ = δ1,0 0 = 0, so that g̃ ∈ ker(d0,1 ). Furthermore, since
(g̃, 0)+∂ 0 (µ) = (g̃·δ0,0 µ, d0,0 µ), the freedom in choice of representative g̃ is exactly δ0,0 (ker(d0,0 )).
Thus ker(Φ2 ) = H1δ (H0d (C •,• )).
Lemma 1 is equivalent to exactness of the two sequences

0 −→ H1d (H0δ (C •,• )) −→ H1 (Tot• (C)) −→ im(Φ1 ) −→ 0, (5)
0 −→ H1δ (H0d (C •,• )) 1 •
−→ H (Tot (C)) −→ im(Φ2 ) −→ 0, (6)
and, furthermore, we have
lim H1δ (H0d (C •,• )) = Ȟ1 (M, (O× )g ), lim H1d (H0δ (C •,• )) = H̃1 (g, O(M )),
−→ −→
× g ×
where (O ) ⊂ O is the subsheaf of g-invariants.
Corollary 2. (i) If H1δ (C 0,• ) = 0, then H1 (Tot• (C)) ≃ H1d (H0δ (C •,• )).
(ii) Likewise, if H1d (C •,0 ) = 0, then H1 (Tot• (C)) ≃ H1δ (H0d (C •,• )).
Lemma 2. The images of Φ1 , Φ2 are given by the following exact sequences:

0 −→ im(Φ1 ) −→ H0d (H1δ (C •,• )) −→ H2d (H0δ (C •,• )),
0 −→ im(Φ2 ) −→ H0δ (H1d (C •,• )) −→ H2δ (H0d (C •,• )).
Proof. We give the proof for the first exact sequence. The proof for the second one is similar.
Consider an element [g] ∈ im(Φ1 ) ⊂ H1δ (C 0,• ). Since it lies in the image of Φ1 , there exists an
element λ ∈ C 1,0 satisfying δ1,0 λ = d0,1 g. This implies that d∗0,1 [g] = [d0,1 g] = [δ1,0 λ] = 0, and
thus [g] ∈ H0d (H1δ (C •,• )). The map H1δ (C 0,• ) ⊃ im(Φ1 ) → H0d (H1δ (C •,• )) is obviously injective.
Now, consider an element [g] ∈ H0d (H1δ (C •,• )). Since d∗0,1 [g] = 0, there exists an element
λ ∈ C 1,0 satisfying δ1,0 λ = d0,1 g. If λ̃ ∈ C 1,0 is another such element, then λ̃ − λ = λ0 ∈
H0δ (C 1,• ). The element d1,0 (λ + λ0 ) ∈ C 2,0 is δ2,0 -closed since δ2,0 ◦ d1,0 = d1,1 ◦ δ1,0 and
δ1,0 λ = d0,1 g. Thus d1,0 (λ + λ0 ) ∈ H0δ (C 2,• ). Since the freedom in representative λ + λ0 is
exactly H0δ (C 1,• ), we obtain a unique element [d1,0 λ] ∈ H2d (H0δ (C •,• )). We have [d1,0 λ] = 0, or
equivalently d1,0 λ ∈ d1,0 0
∗ (Hδ (C
1,• )), if and only if [g] ∈ im(Φ ).
1
While Pic(M ) = lim H1δ (C 0,• ) plays an important role, the group H1d (C •,0 ) will in general
−→
grow without bound as the cover U becomes finer. Therefore, as a counterpart to Ȟ1 (U , O× ) =
H1δ (C 0,• ), we define Mg (U ) := H0δ (H1d (C •,• )) which can be interpreted as the collection of local
(infinitesimal) multipliers of g with respect to the cover U that are equivalent on overlaps. Note
that this is also a reasonable definition in this context due to Lemma 2.
Definition 2. The group of multipliers of g on M is the direct limit Mg (M ) := lim Mg (U ).

−→
The group Mg (M ) should not be confused with the group of global multipliers H̃1 (g, O(M )),
which is often trivial as the algebra of global functions O(M ) may be small (for instance C
for compact M ). To see the difference, consider an element λ = {λα } ∈ ker(d1,0 ). We have
[λ] ∈ H1d (H0δ (C •,• )) if and only if λα = λβ on Uα ∩ Uβ , and [λ] ∈ Mg ({Uα }) = H0δ (H1d (C •,• )) if
and only if λα = λβ + d log µαβ for some µ = {µαβ } ∈ O× (Uα × Uβ ).
2.3. The equivariant Picard group. From the description of lifts at the end of Section 2.1
we see that the pair (g, λ) ∈ C 0,1 × C 1,0 defines a g-equivariant line bundle if and only if
δ0,1 g = 0, d1,0 λ = 0, d0,1 g = δ1,0 λ ⇔ (g, λ) ∈ ker(∂ 1 ).
11
The three conditions correspond to the cocycle condition for transition functions, the cocycle
condition for the local lift (1) and the compatibility condition (4), respectively. Rescaling the
fiber coordinates uα in the line bundle corresponds exactly to changing the cocycle (g, λ) by a
coboundary in im(∂ 0 ).
Definition 3. The group of equivalence classes of g-equivariant line bundles is called the g-
equivariant Picard group and denoted by Picg (M ) := lim H1 (Tot• (C)), where we exploit the
−→
direct limit by refinements (or use a fine cover U ) as before.
Denoting by Cg the modified Chevalley-Eilenberg sheaf complex (3), Picg (M ) may be iden-
tified with the first hypercohomology H1 (M, Cg ), cf. [11, Ch. 3.5] for a discussion of hypercoho-
mology Hq .
The maps Φ1 : H1 (Tot• (C)) → H1δ (C 0,• ) and Φ2 : H1 (Tot• (C)) → H1d (C •,0 ) induce maps
Φ1 : Picg (M ) → Pic(M ), Φ2 : Picg (M ) → Mg (M ),
denoted by the same letters (Lemma 2 justifies the choice of codomain for the second map). We
define Picred
g (M ) := im(Φ1 × Φ2 ) ⊂ Pic(M ) × Mg (M ), which we call the reduced g-equivariant
Picard group, and denote by Ψ1 , Ψ2 the projections
Ψ1 : Picred
g (M ) → Pic(M ), Ψ2 : Picg (M ) → Mg (M ).
Then ̟ := Φ1 × Φ2 epimorphically maps Picg (M ) to Picred
g (M ).
Proposition 5. Picred 1
g (M ) ≃ ker(∂ )/ ∼, where the equivalence relation is defined by (g, λ) ∼
(g̃, λ̃) if g̃ = g · δ0,0 ν and λ̃ = λ + d0,0 µ, where µ, ν ∈ C 0,0 satisfy
µ/ν ∈ ker(d0,1 ◦ δ0,0 ) = ker(δ1,0 ◦ d0,0 ).
Proof. The reduced equivalence relation is weaker, as the coboundaries for g and λ can be
chosen independently. Thus if (g, λ) ∈ ker(∂ 1 ) and (g · δ0,0 ν, λ + d0,0 µ) ∈ C 0,1 × C 1,0 is an
equivalent cocycle then it automatically satisfies the first two conditions: δ0,1 (g · δ0,0 ν) = 0 and
d1,0 (λ + d0,0 µ) = 0. However the third condition applied to the new pair is d0,1 (g · δ0,0 ν) =
δ1,0 (λ + d0,0 µ), which is equivalent to µ/ν ∈ ker(d0,1 ◦ δ0,0 ) = ker(δ1,0 ◦ d0,0 ).
Let us investigate the relationship between Picg (M ) and Picred g (M ). By Proposition 5 the
admissible pair (ν, µ) ∈ C 0,0 × C 0,0 characterizing the freedom in choice of representatives
(g, λ) ∈ ker(∂ 1 ) for the reduced group can be rewritten as (ν, µ) = (ν/µ, 1) · (µ, µ) ≃ (ν/µ, µ) ∈
ker(d0,1 ◦ δ0,0 ) × C 0,0 . It follows that (for a good cover) Picred
g (M ) is equal to
ker(∂ 1 ) ker(∂ 1 ) H1 (Tot• (C))

= = ,
˜ · im(∂ 0 )
im(∂) ˜
im(∂) ˜
im(∂)
˜
im(∂)∩im(∂ 0 )
· im(∂ 0 ) ˜
im(∂)∩im(∂ 0 )
where the map ∂˜ : ker(d0,1 ◦ δ0,0 ) → C 0,1 × C 1,0 is defined as δ0,0 × 0. We have:
im(∂ 0 ) = {(δ0,0 µ, d0,0 µ) ∈ C 0,1 × C 1,0 | µ ∈ C 0,0 },
˜ = {(δ0,0 κ, 0) ∈ C 0,1 × C 1,0 | κ ∈ ker(d0,1 ◦ δ0,0 )} ≃ ker(d0,1 ◦ δ0,0 )/ ker(δ0,0 ).
im(∂)
If d0,0 µ = 0, then µ ∈ ker(δ1,0 ◦ d0,0 ) = ker(d0,1 ◦ δ0,0 ), and therefore
˜ ∩ im(∂ 0 ) = {(δ0,0 µ, 0) ∈ C 0,1 × C 1,0 | µ ∈ ker(d0,0 )}
im(∂)
= δ0,0 (ker(d0,0 )) ≃ ker(d0,0 )/(ker(δ0,0 ) ∩ ker(d0,0 )).
It follows that
˜
im(∂) ker(d0,1 ◦ δ0,0 )
= .
˜ ∩ im(∂ 0 )
im(∂) ker(δ0,0 ) · ker(d0,0 )
Defining
ker(d0,1 ◦ δ0,0 )
Tg (U ) := , Tg (M ) := lim Tg (U ), (7)
ker(δ0,0 ) · ker(d0,0 ) −→
gives us the relation between Picg (M ) and Picred
g (M ):
Proposition 6. The following sequence is exact:

̟
→ Picred
0 → Tg (M ) −→ Picg (M ) − g (M ) → 0. (8)
The commutative diagram in Figure 1 gives relations between the groups we have considered,
allowing to derive the vanishing conditions for Picg (M ) and various isomorphisms. The diagram
0 0 0
0 Tg (M ) 0
δ∗0,0 d0,0
∗
Ȟ1 (M, (O× )g ) H̃1 (g, O(M ))
Picg (M )
Φ1 Φ2
Pic(M ) ̟ Mg (M )
Ψ1 Ψ2
Picred
g (M )
Figure 1. Commutative diagram: The dotted and dashed long sequences as

well as three straight line sequences are exact.
contains the short exact sequence (8), the (direct limit of) short exact sequences (5)-(6), and
also two longer exact sequences. For instance, exactness at δ∗0,0 and d∗0,0 can be seen as follows.
If µ ∈ ker(d0,1 ◦ δ0,0 ) then δ0,0 µ ∈ ker(d0,1 ) ∩ ker(δ0,1 ) and d0,0 µ ∈ ker(δ1,0 ) ∩ ker(d1,0 ), whence
[δ0,0 µ] ∈ H1δ (H0d (C •,• )) and [d0,0 µ] ∈ H1d (H0δ (C •,• )). We have [δ0,0 µ] = 0 if and only if there exists
an element µ0 ∈ ker(d0,0 ) such that δ0,0 µ = δ0,0 µ0 . This happens if and only if µ = µµ0 µ0 ∈
ker(δ0,0 ) · ker(d0,0 ). By a similar argument [d0,0 µ] = 0 if and only if µ ∈ ker(δ0,0 ) · ker(d0,0 ).
Note that the maps Ȟ1 (M, (O× )g ) → Picg (M ) and H̃1 (g, O(M )) → Picg (M ) in the commu-
tative diagram are defined by [g] 7→ [(g, 0)] and [λ] 7→ [(1, −λ)], respectively.
Corollary 3. If Ȟ1 (M, (O× )g ) = 0 or H̃1 (g, O(M )) = 0, then Picred

g (M ) ≃ Picg (M ).
Corollary 4. (i) We have H̃1 (g, O(M )) = 0 if and only if Φ1 : Picg (M ) → Pic(M ) is injective.
(ii) Likewise, Ȟ1 (M, (O× )g ) = 0 if and only if Φ2 : Picg (M ) → Mg (M ) is injective.
Notice im(Ψ1 ) = im(Φ1 ) and im(Ψ2 ) = im(Φ2 ), which are described by Lemmata 1 and 2.
Proposition 7. The group Tg (M ) of equivariant line bundles with trivial reduction corresponds
to global locally trivial lifts of g to the trivial line bundle over M modulo globally trivial lifts.
13
Proof. First note that an element [(g, λ)] ∈ Picg (M ) in the kernel of ̟ also belongs to the kernel
of Φ1 = Ψ1 ◦ ̟, so g determines a trivial line bundle. Similarly using Φ2 = Ψ2 ◦ ̟ we conclude
that λ yields a locally trivial lift (the multiplier is cohomologous to zero on open sets Uα ).
Next, applying d0,0 to both the numerator and the denominator of the right hand side of (2.3)
we get Tg (M ) ≃ ker δ |im d0,0 /d (ker δ0,0 ), whence the required interpretation.
1,0
0,0
Finally, by applying δ0,0 to (2.3) we conclude Tg (M ) ≃ ker d0,1 |im δ0,0 /δ0,0 (ker d0,0 ), which

corresponds to line bundles with g-invariant transition functions modulo global g-invariants.
Propositions 6 and 7 combine into Theorem 1.
Example 1 (Rational curve). Consider the projective space CP 1 with charts U0 ≃ C1 (x) and
U∞ ≃ C1 (y), with coordinates related by y = 1/x on U0 ∩ U∞ . The Lie algebra sl(2, C) acts
naturally on this space with the basis X, Y, Z given in local coordinates:
X|U0 = ∂x , Y |U0 = x∂x , Z|U0 = x2 ∂x , X|U∞ = −y 2 ∂y , Y |U∞ = −y∂y , Z|U∞ = −∂y .
Let λi be a representative
R of an element in H 1 (sl(2, C), O(Ui )) for i = 0, ∞.
Taking µ0 = e 0 0λ (X )dx (the integral sign denotes the anti-derivative on C) gives
(λ0 − d0,0 µ0 )(X0 ) = λ0 (X0 ) − ∂x (log µ0 ) = 0,
so we can without loss of generality assume that λ0 (X) = 0. The values λ0 (Y ) and λ0 (Z) are
now determined by the cocycle conditions
X(λ0 (Y )) − Y (λ0 (X)) = λ0 ([X, Y ]) = λ0 (X) = 0,
X(λ0 (Z)) − Z(λ0 (X)) = λ0 ([X, Z]) = 2λ0 (Y ),
Y (λ0 (Z)) − Z(λ0 (Y )) = λ0 ([Y, Z]) = λ0 (Z).
This leads to
λ0 (X) = 0, λ0 (Y ) = 12 A, λ0 (Z) = Ax.
Analogous computations on U∞ gives
λ∞ (X) = By, λ∞ (Y ) = 21 B, λ∞ (Z) = 0.
Next we require that d0,1 g0∞ = (δ1,0 λ)0∞ for some δ0,1 -cocycle g0∞ ∈ O× (U0 ∩ U∞ ). Evalu-
ating this on X, Y and Z leads to the following overdetermined system of ODEs:
∂x (g0∞ ) B x∂x (g0∞ ) A−B x2 ∂x (g0∞ )
=− , = , = Ax.
g0∞ x g0∞ 2 g0∞
The system has a solution if and only if A = −B in which case g0∞ = CxA . This solution
is holomorphic on U0 ∩ U∞ if and only if A ∈ Z. The constant C can be set equal to 1 by
multiplying g0∞ with (δ0,0 µ)0∞ for µ = {µ0 = 1, µ∞ = C} ∈ ker(d0,0 ). Thus the global lifts are
given by (λ0 , λ∞ ) with A = −B ∈ Z, and the corresponding sl(2, C)-equivariant line bundle has
transition function g0∞ = xA . To sum up, the cover U = {U0 , U∞ } is nice and we get
Picsl(2,C) (CP 1 ) ≃ Picred 1
sl(2,C) (CP ) ≃ Z.
The first isomorphism is a consequence of Ȟ1 (M, (O× )g ) ≃ H1δ (H0d (C •,• )) = 0 and Corollary
3. Since CP 1 is covered by two open charts, we have C 0,2 = 0, implying H0d (C •,2 ) = 0 and
H2δ (H0d (C •,• )) = 0. By Lemma 2, im(Ψ2 ) ≃ H0δ (H1d (C •,• )) = Mg ({U0 , U∞ }), which for this cover
is isomorphic to Z.
A straightforward generalization of this computation gives Tg (CP n ) = 0, Mg (CP n ) = C (for
n = 1 this is parametrized by the above A = −B, but with a finer cover U it is unconstrained:
A ∈ C) and Picg (CP n ) = Z for g = sl(n + 1, C).
Remark 3. Recall that Pic(CP n ) = {OCP n (k)}k∈Z ≃ Z, where OCP n (0) is the trivial line
bundle, OCP n (−1) is the tautological line bundle and for k > 0:
∗
OCP n (−k) = OCP n (−1)⊗k , OCP n (k) = OCP n (−1)⊗k .
The canonical line bundle is KCP n = Λn T ∗ CP n = OCP n (−n − 1), cf. [15, Ch. 2.2].
We will see later that Ψ1 and Φ1 may be non-injective. The following shows it for Ψ2 and Φ2 .
Example 2 (Elliptic curve). Consider C2 with coordinates (x, u) and two commuting maps
h1 (x, u) = (x + 1, u), h2 (x, u) = (x + ω1 , ω2 u),
where ω1 ∈ C\R, ω2 6= 0. Both of these maps respect the projection C2 → C given by (x, u) 7→ x,
and the vector field ∂x . Thus in the quotient by the Z2 action generated by h1 , h2 we get that
the vector field ∂x on the elliptic curve Γ = C/Z2 lifts to the vector field ∂x on the line bundle
C2 /Z2 over the elliptic curve. This line bundle Lω is topologically trivial but holomorphically
nontrivial for ω2 6= e2πiω1 , and all line bundles of this form lie in ker(Ψ2 ); see section 27 of [2]
for details. The general lift of g is given by ∂x + c u∂u , c ∈ C.
For holomorphic curves we have a short exact sequence (where c1 is the first Chern class)
c
0 → Pic0 (Γ) −→ Pic(Γ) −→
1
H 2 (Γ, Z) → 0,
and for elliptic curves Pic0 (Γ) = Div0 (Γ) ≃ Γ, x 7→ x − x0 , whence Pic(Γ) ≃ Γ ⊕ Z.
The summand Z in Pic(Γ) corresponds to divisors m · x0 , m ∈ Z, x0 ∈ Γ. However for
topologically nontrivial line bundles, m = c1 (L) 6= 0, the algebra g does not possess a lift to
L. Indeed,
h such ailift would define a flat connection, at which point we can use the formula
−1
c1 (L) = 2πi tr R∇ . Alternatively, denoting by π : C → Γ the quotient-projection by the lattice
h1, ω1 i, the pullback π ∗ L is trivial and can be identified with C2 (x, u), on which Z2 acts through
the above h1 , h2 . Invariance of the lift ∂x + f (x) u∂u gives periodicity f (x + 1) = f (x) and the
constraint f (x + ω1 ) − f (x) = 2πim, which are incompatible unless m = 0.
It is easy to see that Mg (Γ) = 0. Moreover Tg (Γ) = C as it corresponds to 0-cochains
cα esx ∈ O(Uα ) modulo local constants {cα } (so the quotient coordinate is s). This can be also
identified with H̃1 (g, O(M )) = C generated by global 1-form dx on Γ.
We conclude:
Picred
g (Γ) = Γ, Picg (Γ) = C2 /Z2 .
Note that the equivariant Picard group can be identified with Ȟ1 (M, (O× )g ) = (C× )2 but
simultaneously it corresponds to trivial one-dimensional bundle, with fibers C(c), over Γ. This
fits well the commutative diagram of Figure 1.
Corollary 4 gives a sufficient condition for Φ1 to be injective. For a connected algebraic group
G, Mumford’s Proposition 1.4 in [22] gives a sufficient condition for the map PicG (M ) → Pic(M)
to be injective, in terms of non-existence of a homomorphism G → GL(1, C). (The group of
G-equivariant line bundles will be discussed in Section 2.5.) This does not straight-forwardly
adapt to the infinitesimal analytic setting, yet below we obtain a result inspired by that of
Mumford.
Let gp ⊂ g denote the isotropy algebra of the point p ∈ M :
gp = {X ∈ g | Xp = 0}.
Let HkdR (M ) denote the holomorphic de Rham cohomology of M . It is known that in the affine
case (for Stein manifolds) as well as for the compact Kähler case this coincides with the singular
cohomology Hk (M, C), see [12, 11]. In general, the holomorphic de Rham cohomology HkdR (M )
is equal to the hypercohomology Hk (M, Ω•M ) of the sheaf of holomorphic forms on M .
15
Lemma 3. Let g be a transitive Lie algebra of vector fields on a manifold M . Define Z =

{[λ] ∈ H̃1 (g, O(M )) | λ(Y )p = 0 ∀Y ∈ gp , ∀p ∈ M } (the defining property is representative-
independent). Then we have a natural embedding Z ֒→ H1dR (M ).
In particular, if H1dR (M ) = 0 then every [λ] ∈ Z is exact: λ = d0,0 µ for some µ ∈ ker(δ0,0 ).
Proof. Consider [λ] ∈ Z. Since its representative λ is defined globally on M , the value λ(X)p ∈ C
is well-defined for any p ∈ M and any X ∈ g. Transitivity of g implies that for every v ∈ Tp M
there exists X ∈ g satisfying Xp = v, which means that the following linear function on Tp M is
well-defined:
αp : Xp 7→ λ(X)p .
This gives a well-defined 1-form α on M . Choosing Xi ∈ g for vi ∈ Tp M such that vi = (Xi )p
we get for every p ∈ M :
(dα)p (v1 , v2 ) = dα(X1 , X2 )p = (X1 (α(X2 )) − X2 (α(X1 )) − α([X1 , X2 ]))p = d1,0 λ(X1 , X2 )p = 0.
Thus dα = 0, so the closed 1-forms in two cohomologies correspond; the same clearly concerns
exact 1-forms. This yields the embedding [λ] 7→ [α].
If H1dR (M ) = 0 then α = d log µ for some µ ∈ O× (M ), and therefore λ = d0,0 µ.
Proposition 8. Let g be a transitive Lie algebra of vector fields on a manifold M such that
dim H̃1 (g, O(M )) > dim H1dR (M ). This holds, for instance, when H̃1 (g, O(M )) 6= 0 and H1dR (M ) =
0. Then for all p ∈ M there exists a surjective Lie algebra homomorphism gp → gl(1, C).
Proof. Consider an element [λ] ∈ H̃1 (g, O(M )) \ Z (by our assumption and Lemma 3 this set is
nonempty). Since δ1,0 λ = 0, λ defines a global lift of g to the trivial bundle M × C:
gλ = {X + λ(X)u∂u | X ∈ g}.
For any point p ∈ M , all vector fields of the Lie subalgebra
{Y + λ(Y )u∂u | Y ∈ gp } ⊂ gλ
are tangent to the fiber π −1 (p). Therefore the restriction to the fiber results in a Lie algebra
homomorphism:
gp ∋ Y 7→ λ(Y )p u∂u ∈ gl(1, C). (9)
By definition of Z, λ(Y )p 6≡ 0 for a generic point p. Thus, at this point (9) is surjective. Since,
for a transitive g, the isotropies gp at different points p are conjugate, the claim follows.
The existence of a surjective Lie algebra homomorphism to gl(1, C) implies that gp has an
ideal of codimension 1, the kernel of (9). Corollary 4 then leads to the following statement.
Corollary 5. Let g be a transitive Lie algebra of vector fields on a manifold M with H1dR (M ) = 0.
If gp does not have an ideal of codimension 1, then Picg (M ) → Pic(M ) is injective.
As a consequence, Φ1 is injective if gp is perfect [gp , gp ] = gp (this also applies to infinite-
dimensional Lie algebras g) and, in particular, if gp is semisimple (for finite-dimensional g).
Example 3 (Special affine algebra on the plane). Consider the Lie algebra
g = h∂x , ∂y , y∂x , x∂y , x∂x − y∂y i ⊂ D(C2 ).
The Lie algebra is transitive with simple isotropy gp . By Proposition 8 we have H̃1 (g, O(M )) = 0,
implying that Picg (C2 ) → Pic(C2 ) is injective by Corollary 4. Since Pic(C2 ) = 0 it follows that
Picg (C2 ) = 0.
Next, for the Lie algebra
h = hy∂x , −xy∂x − y 2 ∂y , ∂x , −x2 ∂x − xy∂y , 2x∂x + y∂y i ⊂ D(C2 )
the isotropy of the point p = 0
h0 = hy∂x , −xy∂x − y 2 ∂y , −x2 ∂x − xy∂y , 2x∂x + y∂y i
is 4-dimensional and solvable, and it has a 3-dimensional ideal. In this case Pich (C2 ) = C.
Note that g and h can be viewed as the same Lie subalgebra sl(2, C)⋉C2 ⊂ sl(3, C) ⊂ D(CP 2 )
restricted to two different open charts of CP 2 .
2.4. Line bundles admitting a transversal lift. We start by recalling some basic information
about divisors, cf. [11, 15]. Let O× denote the multiplicative sheaf of nonvanishing holomorphic
functions on a complex manifold M , and M× the sheaf of meromorphic functions that are not
identically zero on M . A divisor D is a global section of M× /O× . It is defined by a collection
of functions fα ∈ M× (Uα ) for an open cover {Uα } of M , such that fα /fβ ∈ O× (Uα ∩ Uβ ). Any
divisor D gives rise to a line bundle, denoted by [D], whose transition functions are given by
gαβ = fα /fβ ∈ O× (Uα ∩ Uβ ), and the long exact sequence (see [11]) relates the group of divisors
Div(M ) := Ȟ0 (M, M× /O× ) to the Picard group on M :
· · · → Ȟ0 (M, M× ) → Div(M ) → Pic(M ) → Ȟ1 (M, M× ) → · · · (10)
Here Ȟ0 (M, M× ) is the group of global meromorphic functions on M , and Div(M )/Ȟ0 (M, M× )
is the group of equivalence classes of divisors (equivalent divisors give equivalent line bundles).
Definition 4. Let g ⊂ D(M ) be a Lie algebra of vector fields on M . The divisor D = {fα }
defined on the open cover {Uα } of M is a g-invariant divisor if for each α
X(fα ) = λα (X)fα , ∀X ∈ g,
where λα ∈ g∗ ⊗ O(Uα ). The group of g-invariant divisors is denoted by Picg (M ). The collection
λ = {λα } is called the weight of D.
It follows that g is tangent to the set of zeros of D, and also to the set of poles. In this way
D defines a (possibly reducible) invariant hypersurface in M .
Proposition 9. Let g ⊂ D(M ) be a Lie algebra of vector fields on M , and let D = {fα } be a
g-invariant divisor with weight λ = {λα }. Set gαβ = fα /fβ and define g = {gαβ }. Then the
pair (g, λ) defines a g-equivariant line bundle L = [D], which is independent of the choice of
representative functions fα .
Proof. To show that the pair defines a g-equivariant line bundle, we must verify that ∂ 1 (g, λ) = 0.
It is clear that δ0,1 g = 0 since gαβ are transition functions of [D]. Next, the condition d1,0 λα = 0
holds for each α since fα 6≡ 0 and for arbitrary vector fields X, Y ∈ g we have
λα ([X, Y ])fα = [X, Y ](fα ) = X(Y (fα )) − Y (X(fα )) = (X(λα (Y )) − Y (λα (X)))fα .
What remains is to verify that the weights λα are compatible with the transition functions
gαβ = fα /fβ . On Uα ∩ Uβ we have

X(gαβ )
λα (X)fα = X(fα ) = X(gαβ fβ ) = X(gαβ )fβ + gαβ X(fβ ) = + λβ (X) fα ,
gαβ
which is equivalent to δ1,0 λ = d0,1 g.
Next, to show that the g-equivariant bundle is independent of representative functions fα of
D, take another representative f˜α = µα fα with µα ∈ O× (Uα ). This results in an equivalent
g-equivariant line bundle ({g̃αβ }, {λ̃α }): g̃αβ = gαβ µα /µβ and λ̃α (X) = λα (X) + X(µα )/µα for
all X ∈ g.
17
The lifted Lie algebra gλ ⊂ D([D]) is given locally on π −1 (Uα ) ≃ Uα × C by

gλ |π−1 (Uα ) = {X̂|π−1 (Uα ) = X|Uα + λα (X)u∂u | X ∈ g}.
The lift gλ has exactly the same form as the lifts in Section 2.1, the only difference being that
λ is now specifically determined by a g-invariant divisor. Thus we have a map
jg : Divg (M ) → Picg (M ), D 7→ ([D], gλ ).
In general this map is neither injective nor surjective. For instance, the kernel of this map
contains all global g-invariant functions in M× (M ). If g is transitive, then Divg (M ) = 0, but
Picg (M ) may be nontrivial as in Example 1. The next example exhibits non-trivial Divg (M ).
Example 4 (sl(2, C) ⊂ D(CP 2 )). The manifold CP 2 is covered by the three charts
U3 = {[x : y : z] ∈ CP 2 | z 6= 0},
U2 = {[x : y : z] ∈ CP 2 | y 6= 0},
U1 = {[x : y : z] ∈ CP 2 | x 6= 0},
on which coordinates are given respectively by
(x1 , x2 ) = (x/z, y/z), (y1 , y3 ) = (x/y, z/y), (z2 , z3 ) = (y/x, z/x).
Consider the Lie algebra sl(2, C) ⊂ sl(3, C) given in the respective charts by
hx2 ∂x1 , x1 ∂x2 , x1 ∂x1 − x2 ∂x2 i,
h∂y1 , −y12 ∂y1 − y1 y3 ∂y3 , 2y1 ∂y1 + y3 ∂y3 i,
h−z22 ∂z2 − z2 z3 ∂z3 , ∂z2 , −2z2 ∂z2 − z3 ∂z3 i.
A computation shows that the Chevalley-Eilenberg cohomology groups are
H1 (sl(2, C), O(U3 )) = 0, H1 (sl(2, C), O(U2 )) = C2 , H1 (sl(2, C), O(U1 )) = C2
with representative cocycles
λ3 = (0, 0, 0), λ2 = (0, B1 y1 + B2 y32 , −B1 ), λ1 = (C1 z2 + C2 z32 , 0, C1 ),
The holomorphic transition functions, compatible via overdetermined system (4), exist only for
B2 = C2 = 0, C1 = B1 = b, and are given by formulae
g32 = A1 y3b = A1 x−b
2 , g31 = A2 z3b = A2 x−b
1 , g21 = A3 z2b = A3 y1−b .
Requiring gαβ to be holomorphic gives the further restriction b ∈ Z. The constants A1 , A2 , A3
can be set equal to 1 by multiplying with an sl(2, C)-invariant δ0,0 -coboundary. We conclude:
Picsl(2,C) (CP 2 ) ≃ H1 (Tot• (C)) = Z.
In this case Divsl(2,C) (CP 2 ) ≃ Picsl(2,C) (CP 2 ). The unique divisor D = {f1 , f2 , f3 } corresponding
to b ∈ Z is given by
f1 = z3−b , f2 = y3−b , f3 = 1.
We will now describe an obstruction for the existence of invariant divisors, elaborating upon
[7]. The following definition is adapted from [1] where it was used for group actions.
Definition 5. For a Lie algebra g ⊂ D(M ) and a holomorphic line bundle π : L → M , a lift
ĝ ⊂ D(L) is called transversal if generic ĝ-orbits on L π-project biholomorphically.
Note that singular orbits of ĝ may project non-injectively (but indeed surjectively) to g-orbits
on M (see e.g. Example 5 below). The following is a reformulation of Theorem 2.
Proposition 10. Let (L, ĝ) be a g-equivariant line bundle over M . Suppose L ∈ im(Ψ1 ◦ jg ),
i.e., L = [D] for some g-invariant divisor D with weight λ and ĝ = gλ . Then ĝ is transversal.
Proof. Let D = {fα } be a g-invariant divisor with weight λ = {λα }, and [D] the corresponding
line bundle defined by transition functions gαβ = fα /fβ . Any element X̂ ∈ gλ takes on Uα the
form X̂|Uα = X|Uα + λα (X)u∂u for some X ∈ g. A straight-forward computation shows that
X̂(u)fα − uX̂(fα ) uλα (X)fα − uX(fα )
X̂(u/fα ) = 2
= = 0.
fα fα2
Thus, the function u/fα on Uα × C is a meromorphic absolute invariant (constant on gλ -orbits).
It follows that the dimension of generic g-orbits on Uα is equal to the dimension of generic
gλ -orbits on Uα × C. This holds simultaneously on each Uα , and therefore globally on M .
Remark 4. Local absolute invariants u/fα define a collection of local sections tangent to gλ ,
which are given by u = Cfα with C being an absolute g-invariant. Choosing C to be a global
invariant on M gives a global gλ -invariant section of [D].
Returning to Example 1 on sl(2, C) ⊂ D(CP 1 ), we observe that generic orbits of any nontrivial
lift are 2-dimensional. Thus there are no nontrivial invariant divisors, which also follows from
the fact that sl(2, C) is transitive on CP 1 . Here is another demonstration of Proposition 10.
Example 5 (aff(1, C) ⊂ D(CP 1 )). Consider again coordinate charts U0 ≃ C1 (x) and U∞ ≃
C1 (y) of CP 1 , with the Lie subalgebra g = aff(1, C) = hX, Y i ⊂ sl(2, C) given by
X|U0 = ∂x , Y |U0 = x∂x , X|U∞ = −y 2 ∂y , Y |U∞ = −y∂y .
General representatives λs of elements in H1 (aff(1, C), O(Us )), for s = 0, ∞, in basis (X, Y ) are
given by
λ0 = (0, A), λ∞ = (B2 y, B1 ), A, B1 , B2 ∈ C.
A general compatible transition function exists only when A = B1 − B2 , in which case it is
cohomologous to g0∞ = y B2 = x−B2 . Requiring g0∞ to be holomorphic results in B2 ∈ Z. The
local lifts corresponding to λ0 and λ∞ are given by
gλ0 = h∂x , x∂x + (B1 − B2 )u∂u i, gλ∞ = h−y 2 ∂y + B2 yu∂u , −y∂y + B1 u∂u i.
It is clear that the generic orbit dimension is 1 if and only if B2 = B1 . In this case we get the
invariant divisor D given by f0 = 1 and f∞ = y −B1 . Thus
Divg (CP 1 ) = Z $ C × Z = Picg (CP 1 ).
Note that the map Ψ1 : Picg (CP 1 ) → Pic(CP 1 ) is not injective: ker(Ψ1 ) ≃ H1d (H0δ (C •,• )) = C.
Similar to Example 1, we have Mg (M ) = C even though Mg ({U0 , U∞ }) = Z.
Proposition 10 can be viewed as a global version of [7, Th. 5.4]. According to it, locally, in
smooth regular case the statement allows a converse, giving a criterion for the (local) existence
of relative invariants. Globally, in general analytic context, there is no converse to Proposition
10, due to other reasons for non-existence of meromorphic invariant divisors/relative invariants.
This is shown in the following simple example, and also in a more complicated example of Section
3.4. Yet, in the following section, we will give a converse statement in the algebraic context.
Example 6. Consider the Lie algebra g = hx2 ∂x i ⊂ D(C). All line bundles over C are trivial,
Picg (C) = 0, while we have Picg (C) = C2 . A general representative cocycle of H̃1 (g, O(C)) has
the form λ(x2 ∂x ) = A + Bx with A, B ∈ C, and the corresponding lifted Lie algebra is
gλ = hx2 ∂x + (A + Bx)u∂u i ⊂ D(C × C).
19
Generic orbits of both g and gλ are 1-dimensional for any choice of A and B, thus gλ is transver-
sal. However, the general solution of the system x2 ∂x (f (x)) = (A + Bx)f (x) is
f (x) = xB e−A/x .
This is a (meromorphic) g-invariant divisor on C only when A = 0 and B ∈ Z, i.e., Divg (CP 1 ) =
Z and not all equivariant line bundles come from invariant divisors.
2.5. Lie group vs Lie algebra approach. Let G be a Lie group acting on M . We consider
the group PicG (M ) of G-equivariant line bundles over M . In the setting of algebraic schemes,
it was studied in [22, Ch. 1.3]. Here we give a different description of PicG (M ) emphasizing its
relation to Picg (M ) when g is the Lie algebra of vector fields corresponding to the Lie group
action, but demonstrate that, in general, PicG (M ) is not isomorphic to Picg (M ).
Definition 6. A lift ρ̂ of a group action ρ : G × M → M to a line bundle π : L → M is a map
ρ̂ : G × L → L such that ρg : L → L is a vector bundle automorphism for each g ∈ G and the
following diagram commutes:
ρ̂
G×L L
id×π π
ρ
G×M M
The pair (π : L → M, ρ̂) is called a G-equivariant line bundle. The space of such bundles,
modulo the natural equivalences, has the group structure with the operation of tensor product.
The group of G-equivariant line bundles PicG (M ) is called the G-equivariant Picard group.
We assume G acts by biholomorphisms on M . The general description of G-equivariant line
bundles over M can be done in terms of a cohomology theory that generalizes both the Čech
cohomology and the Lie group cohomology with coefficients in the G-module O× (M ).
Let π : L → M be a line bundle. Assume there exists a lift of the group action to L, i.e.
for each ϕ ∈ G there exists a (holomorphic) vector bundle automorphism ϕ̂ on L, satisfying
π(ϕ̂(p)) = ϕ(π(p)) for each p ∈ L (to simplify formulas, we use the notation ϕ = ρg and
ϕ̂ = ρ̂g ). Let U = {Uα } be a trivializing chart for L, and uα be a (linear) fiber coordinate on
π −1 (Uα ) ≃ Uα × C. Then ϕ̂ acts on uα in the following way:
ϕ̂∗ (uα ) = Λαβ (ϕ)uβ , Λαβ (ϕ) ∈ O× Uβ ∩ ϕ−1 (Uα ) .

(11)
Composing with a second element in the Lie group gives
ψ̂ ∗ (ϕ̂∗ (uα )) = ψ ∗ (Λαβ (ϕ))Λβγ (ψ)uγ
on Uγ ∩ ψ −1 (Uβ ∩ ϕ−1 (Uα )). Simultaneously, on Uγ ∩ ψ −1 (ϕ−1 (Uα )), we have
(ψ̂ ∗ ◦ ϕ̂∗ )(uα ) = Λαγ (ϕ ◦ ψ)uγ .
Thus on Uγ ∩ ψ −1 (Uβ ) ∩ ψ −1 (ϕ−1 (Uα )) we get:
ψ ∗ (Λαβ (ϕ))Λβγ (ψ) = Λαγ (ϕ ◦ ψ). (12)
When ϕ is equal to the identity transformation on M , equation (11) gives Λαβ (id) = gαβ , where
gαβ is the transition function of π on Uα ∩ Uβ . Setting ϕ = id in (12) gives
Λαγ (ψ) = ψ ∗ (gαβ )Λβγ (ψ)
while setting ψ = id leads to
Λαγ (ϕ) = Λαβ (ϕ)gβγ .
The last equality shows that if the transition functions are given, then Λαβ (ϕ) on Uβ ∩ Uα ∩
ϕ−1 (Uα ) is uniquely determined by Λαα (ϕ).
Next, changing the fiber coordinates vα = µα uα for µα ∈ O× (Uα ),
ϕ̂∗ (µα )Λαβ (ϕ)uβ = ϕ̂∗ (µα uα ) = ϕ̂∗ (vα ) = Λ̃αβ (ϕ)vβ = Λ̃αβ (ϕ)µβ uβ ,
results in the equivalence relation

ϕ∗ (µα )
Λαβ (ϕ) ∼ Λαβ (ϕ).
µβ
Introducing the differentials
ϕ∗ (µα )
(D 0 µ(ϕ))αβ = ,
µβ
Λαγ (ϕ ◦ ψ)
(D 1 Λ(ϕ, ψ))αβγ = ∗ ,
ψ (Λαβ (ϕ))Λβγ (ψ)
we see that Λ = {Λαβ } defines a G-equivariant line bundle over M if and only if D 1 Λ = 0.
Moreover, the G-equivariant line bundles defined by Λ and Λ̃ are equivalent if and only if
Λ̃ = Λ · D 0 µ for some µ. We define the action group cohomology for a given cover U
ker(D 1 )
H1U (G, O× ) =
im(D 0 )
and in general we use the direct limit of this cohomology, H1 (G, O× ) := lim H1U (G, O× ).
−→
Proposition 11. The group PicG (M ) of G-equivariant line bundles is isomorphic to H 1 (G, O× ).
Let us note that we consider not abstract, but rather continuous (van Est) group cohomology,
cf. [9]. In fact, the above specifies cochains to be holomorphic.
Remark 5. For a trivial line bundle we get the Lie group cohomology H1 (G, O× (M )) of the
Lie group G with the values in the module O× (M ). On the other hand, with ϕ and ψ being
idM , the above definition gives the Čech cohomology Ȟ1 (M, O× ) of M with the values in the
sheaf O× . Thus PicG (M ) interpolates between the two cohomologies.
Any Lie group action gives rise to a Lie algebra g of vector fields. Consider a one-parameter
group ϕt ⊂ G, and the corresponding vector field X. Denote the vector field on L corresponding
to ϕ̂t by X̂. For small t the set Uα ∩ ϕ−1
t (Uα ) is nonempty, and on this set we have
ϕ̂∗t (uα ) = Λαα (ϕt )uα .
When t approaches 0, then Uα ∩ ϕ−1

t (Uα ) approaches Uα , and the Lie derivative of uα with
respect to X̂ on Uα is given by
d
LX̂ (uα ) = Λαα (ϕt )uα .
dt t=0
Comparing this to the lifts X̂ = X + λα (X)uα ∂uα discussed in Section 2.1 results in the relation
d
λα (X) = Λαα (ϕt ).
dt t=0
Thus a G-equivariant line bundle on M yields a g-equivariant line bundle for g = Lie(G).
However, the map
PicG (M ) → Picg (M )
in general is neither injective nor surjective. Non-injectivity is illustrated by an action of a
discrete group, like Zm : z 7→ z m on CP 1 . Non-surjectivity is demonstrated as follows.
21
Example 7 (Projective action revisited). The Lie groups SL(2, C) and P GL(2, C) act on CP 1
by Möbius transformations. In the open cover given by charts U0 ≃ C(x) and U∞ ≃ C(y) the
action is
a b ax + b dy + c
ϕ= : ϕ∗ (x) = , ϕ∗ (y) = .
c d cx + d by + a
For SL(2, C) the lifts are given by
u0 u∞
ϕ̂∗ (u0 ) = A
, ϕ̂∗ (u∞ ) = ,
(cx + d) (by + a)A
where A ∈ Z, as in Example 1. On the other hand, for P GL(2, C) the lifts are given by
(ad − bc)A/2 u0 (ad − bc)A/2 u∞
ϕ̂∗ (u0 ) = , ϕ̂∗ (u∞ ) = ,
(cx + d)A (by + a)A
which is well-defined if and only if A = 2m ∈ 2Z. In other words, the line bundle OCP 1 (1) is
not P GL(2, C)-equivariant, but OCP 1 (2) is.
This example, borrowed from [22, Ch. 1.3], works in any dimension n: the line bundle OCP n (k)
is P GL(n + 1, C)-equivariant iff k ∈ (n + 1)Z, i.e., this group lifts only to the powers of the
canonical bundle KCP n . On the other hand, all bundles OCP n (k) are SL(n + 1, C)-equivariant.
(Note that the center of SL(n + 1, C) is Zn+1 and P GL(n + 1, C) = SL(n + 1, C)/Zn+1 .) This
difference can not be seen at the Lie algebra level, since the two Lie group actions give rise to
the same Lie algebra of vector fields. Summarizing we have:
PicSL(n+1,C) (CP n ) = Picsl(n+1,C) (CP n ) = Z = Pic(CP n ) ⊃ (n + 1)Z = PicP GL(n+1,C) (CP n ).
Note that in this example both groups P GL(n + 1, C) and SL(n + 1, C) are algebraic, so this
example illustrates a general result in [22, Cor. 1.6] on G-linearization of high powers Lm of an
algebraic line bundle L. Next we discuss a similar effect for invariant divisors.
Recall that an algebraic Lie algebra is g = Lie(G) for an algebraic Lie group G. If M is an
algebraic variety, we call a Lie algebra g ⊂ D(M ) algebraic if it is the Lie algebra of an algebraic
action by an algebraic Lie group on M . The following is a converse to Proposition 10 in the
algebraic context (there is a version of this statement for PicG (M )).
Theorem 12. Let (L, ĝ) ∈ Picg (M ) be a g-equivariant line bundle over an algebraic variety M
for an algebraic Lie algebra g of vector fields. Assume that the lift ĝ is algebraic and transversal.
Then there exists an integer m ∈ Z+ such that Lm ∈ im(Φ1 ◦ jg ), i.e., Lm = [D] for some
invariant divisor D with weight λ, and ĝ = gλ/m .
Proof. Since g is transversal, it admits on L an absolute invariant I = I(x, u), with x coordinate
on M and u a fiber coordinate on L, such that ∂u (I) 6≡ 0. This complements absolute invariants
J = J(x) obtained by pullback from M . By Rosenlicht’s theorem [26] the algebraicity of the
action implies that the invariant I can be chosen rational in proper (local) variables x, u (on Uα
with algebraic overlaps). Decompose I into its Laurent series by the fiber variable u
∞
X
I= hk (x)uk . (13)
k=−N
Since [u∂u , ĝ] = 0 we get that (u∂u )r (I) is an absolute invariant for every r. The spectrum
of the operator u∂u on generators uk is simple, and due to rationality the coefficients of I are
determined by a finite number of base functions h(x). Thus every term in the series (13) is an
absolute invariant. Choose such invariant of the lowest (in absolute value) degree by u. This
degree m does not depend on local coordinate chart Uα , α ∈ A, we are using, and we get:
um Iα um /fα (x) m fβ
Iα = α =⇒ 1= = αm = gαβ on Uα ∩ Uβ .
fα (x) Iβ uβ /fβ (x) fα
The collection of functions {fα ∈ O(Uα ) : α ∈ A} defines a g-invariant divisor D with weight
λα (X) = X(log fα ), X ∈ g, and the corresponding line bundle [D] has transition functions
fα m
g̃αβ = = gαβ .
fβ
Thus [D] = Lm and the claim follows.
Example 8. Consider the Lie algebra g = hx∂x i on C and its lift ĝ = hx∂x + Cu∂u i on the
m
trivial line bundle C × C. It is algebraic if C = pq ∈ Q with absolute invariant I = xuCm being
q
algebraic for minimal m = q, i.e., I = xup . Such a situation occurs for differential invariants
of curves in Euclidean plane with respect to the motion group, namely for the “square of the
curvature”, see the end of Introduction in [18]. The g-equivariant line bundle (C × C, ĝ) is in
im(jg ) if and only if C ∈ Z.
3. Invariant polynomial divisors on algebraic bundles

In this section we will consider Lie algebras of vector fields on bundles that have additional
structure on the fibers, and where it makes sense to consider divisors that are polynomial in
the fiber coordinates. More precisely, we will focus on affine bundles in Section 3.1 and on jet
bundles in Section 3.2. In the remaining three subsections, we will apply the obtained results
to examples involving Lie algebras of vector fields on jet spaces.
3.1. Lie algebra action on affine bundles. Let π : E → M be an affine bundle (of rank
r ≥ 1), and let ĝ ⊂ Dproj (E) be a Lie algebra of projectable vector fields on E that preserves
the affine structure in the fibers. In this section we will focus on ĝ-invariant divisors whose
restriction to fibers are polynomials. We define for U ⊂ M
P(U ) = {f ∈ O(π −1 (U )) | f |π−1 (p) is a polynomial for every p ∈ U },
where polynomiality is checked in affine coordinates on E. This space of functions is preserved
by automorphisms: If ϕ : E → E is an automorphism of affine bundles and f ∈ P(U ), then
ϕ∗ f ∈ P(ϕ−1 (U )).
Assume that U = {Uα } is an open cover for M such that π −1 (Uα ) ≃ Uα × Cr for any affine
bundle π. Then {π −1 (Uα )} is open cover for the total space E.
Definition 7. A divisor D = {fα } on an affine bundle π : E → M is called polynomial if its
defining functions fα ∈ O(π −1 (Uα )) are contained in P(Uα ).
In local coordinates xi , uj on π −1 (Uα ) ≃ Uα × Cr , fα takes the form
X
fα = Fσ (x)uσ , Fσ ∈ O(Uα ).
|σ|≤s
uσ (ui )mi
Q
Here = for the multi-index σ = (m1 . . . mr ). The defining functions of a polynomial
divisor D satisfy gαβ = fα /fβ ∈ O× (π −1 (Uα ) ∩ π −1 (Uβ )), where both the numerator and
denominator are polynomials in u1 , . . . , ur . It follows that the polynomials must cancel each
other out, which implies that the transition functions gαβ = fα /fβ are the pullback of functions
g̃αβ ∈ O× (Uα ∩ Uβ ). Thus we obtain the following proposition.
Proposition 13. Let D = {fα } be a polynomial divisor on the affine bundle π : E → M . Then
[D] = π ∗ L for some line bundle L → M .
In the above argument, it is clear that the degree s can be taken to be the same for each α.
We call the smallest such s the degree of the polynomial divisor D.
Next we let ĝ ⊂ Dproj (E) be a projectable Lie algebra of vector fields on E preserving the
affine structure on fibers, and consider ĝ-invariant polynomial divisors.
23
Proposition 14. Let π : E → M be a an affine bundle and let ĝ ⊂ Dproj (E) be a projectable
Lie algebra of vector fields on E preserving the affine structure on fibers; g = dπ(ĝ). If D is a
ĝ-invariant polynomial divisor on E, then [D] = π ∗ L for some g-equivariant line bundle L → M .
Proof. What remains to be proven is that the bundle π : L → M with transition functions g̃αβ
admits a g-lift. In local coordinates xi , uj on π −1 (Uα ) ≃ Uα ×Cr , each X ∈ ĝ takes the form X =
ai (x)∂xi + (bj0 (x) + bjl (x)ul )∂uj . Consider an invariant divisor D given by fα = |σ|≤s Fσ (x)uσ ,
P
a polynomial in u1 , . . . , ur of degree s. We have

X(fα ) = λα (X)fα .
Looking at the coordinate form of X, it is clear that X(fα ) is a polynomial in u1 , . . . , ur and,
furthermore, that its degree is ≤ s. Thus λα (X) = X(fα )/fα is a rational function in u1 , . . . , ur ,
and it is defined everywhere on Uα × Cr only if λα (X) is the pullback of a function λ̃α (X) ∈
O(Uα ). Thus the ĝ-equivariant line bundle over E defined by ({gαβ }, {λα }) is the pullback of
the g-equivariant line bundle over M defined by ({g̃αβ }, {λ̃α }).
Proposition 14 (which was reformulated in Theorem 3) is relevant, for instance, for investi-
gation of relative invariants of tensor fields (and other geometric objects like affine connections)
on a manifold M under the action of some Lie algebra g ⊂ D(M ).
Example 9. Consider the bundle E = S 2 T ∗ M → M whose sections are symmetric 2-forms

on M , and let g be the Lie algebra of holomorphic vector fields on M . There is a canonical
lift ĝ ⊂ D(E) of g. Let x1 , . . . , xn be coordinates on M and u11 , u12 , . . . , unn be the additional
induced coordinates on E. The function det([uij ]), with uji = uij for i < j, is the local expression
for a g-invariant divisor, and its (local) weight is −2divdx1 ∧···∧dxn . Globally, the line bundle given
by this g-invariant divisor is the pullback of the line bundle (Λn T ∗ M )⊗2 = (KM )⊗2 over M .
Let us make a brief remark about invariant rational divisors. Each such is a ratio of two
invariant polynomial divisors. The weights of invariant rational divisors form a lattice generated
by weights of invariant polynomial divisors. In other words, we have the following relation:

SpanZ Divpol
g (M ) = Divrat
g (M ). (14)
3.2. Lie algebra action on jet bundles. Now we consider polynomial divisors on jet bundles.
Most of the arguments here closely resemble those in Section 3.1, but some additional care must
be taken. Our introduction to jets will be very brief, and we refer to [19, 21, 23] for a more
comprehensive treatment.
Let J k (E, m) denote the space of k-jets of codimension-m submanifolds of E, and J k π the
space of k-jets of sections of the fiber bundle π. In statements that are true for both J k (E, m)
and J k π, we will use the notation Jk which can always be replaced with either of the two (an
exception to this convention occurs only in Section 3.5). There are natural bundle structures
πk,l : Jk → Jl for 0 ≤ l ≤ k, and πk : J k π → M .
Coordinates on J k π and J k (E, m) are induced from coordinates on the total space of π or E,
respectively. Given a bundle π : E → M , and an open cover {Uα } of coordinate charts of E, the
−1
collection {πk,0 (Uα )} is an open cover of J k π. The split coordinates x1 , . . . xn , u1 , . . . , um on Uα
induce additional canonical coordinates ujσ , |σ| ≤ k where σ is a multi-index, on πk,0 −1
(Uα ).
To get an open cover of J k (E, m), we let {Uα } be an open cover of coordinate charts of E
that trivializes the bundle J 1 (E, m) → E. On each Uα , for a given set of m + n coordinates,
we choose a splitting x1 , . . . , xn , u1 , . . . , um , and we denote the corresponding coordinate chart
on J 1 (E, m) by Uαi1 ···im with 1 ≤ i1 < · · · < im ≤ dim E. For each way of splitting there is
one chart. The split coordinates on Uα induce additional canonical coordinates uiσ (|σ| ≤ k) on
−1 −1
πk,1 (Uαi1 ···im ) for k ≥ 1. The collection {πk,1 (Uαi1 ···im )} is an open cover of J k (E, m). We define
for U ⊂ Ji
−1
Pi (U ) = {f ∈ O(π∞,i (U )) | f |π−1 (p) is a polynomial for every p ∈ U }.
∞,i
Polynomiality is defined with respect to the canonical coordinates described above. For example,
in the case of J k (E, m), then f ∈ Pj (U ) if and only if we have, for each α,
X
f −1 −1

i1 ···im
= Fτσ uτσ
π∞,j (U )∩π∞,1 Uα
j≤|σ|≤k,|τ |≤r
for some k, r ∈ Z≥0 and a collection {Fτσ } of functions on J j (E, m).

If ϕ : E → E is a point transformation and f ∈ Pi (U ), then (ϕ(∞) )∗ f ∈ Pi (ϕ−1 (U )) for i ≥ 1.
Similarly, if π : E → M is a fiber bundle, ϕ : E → E is a fiber-preserving biholomorphism, and
f ∈ Pi (U ), then (ϕ(∞) )∗ f ∈ Pi (ϕ−1 (U )) for i ≥ 0. In particular, point transformations preserve
P1 while fiber-preserving transformations do the same for P0 . Based on this, we introduce the
following notion.
Definition 8. A divisor on J k (E, m) is called polynomial if its defining functions fαi1 ···im ∈
−1
O(πk,1 (Uαi1 ···im )) are contained in P1 (Uαi1 ···im ). If π : E → M is a fiber bundle, a divisor on J k π
−1
is called polynomial if its defining functions fα ∈ O(πk,0 (Uα )) are contained in P0 (Uα ).
Proposition 15. (1) Let π : E → M be a fiber bundle, and let D = {fα } be a polynomial divisor
on J k π. Then [D] = πk,0
∗ L for some line bundle L → E = J 0 π.
(2) Let E be a manifold and D = {fαi1 ···im } be a polynomial divisor on J k (E, m). Then [D] =
∗ L for some line bundle L → J 1 (E, m).
πk,1
Proof. The proofs of (1) and (2) are very similar, so we prove only (2). Since fαi1 ···im /fβj1···jm are
−1
elements in O× (πk,1 (Uαi1 ···im ∩ Uβj1 ···jm )), the polynomial parts are required to cancel. Thus the
transition functions fαi1 ···im /fβj1 ···jm are the pullback of elements in O× (Uαi1 ···im ∩ Uβj1 ···jm ), which
are the transition functions of a line bundle over J 1 (E, m).
From the proof it follows that the order and degree of the polynomials fαi1 ···im and fβj1···jm agree.
Therefore, the order and degree are also well-defined notions for D = {fαi1 ···im }, and this is true
also for divisors on J k π. We define the weighted degree of the monomial c(x, y, yi )yσj11 · · · yσjss ∈
P1 (Uαi1 ···im ) (with |σl | ≥ 2 for each l) to be sl=1 |σl |, and the weighted degree of a sum of such
P
to be the maximal weighted degree of its monomial parts. The weighted degree can be defined
for a divisor in the same way that order and degree were defined above.
Next, consider a Lie algebra of vector fields g ⊂ D(J0 ); in the case J0 = J 0 π assume also
that g is π-projectable. The Lie algebra prolongs to a unique Lie algebra g(k) ⊂ D(Jk ), see, for
instance, [19, Sec. 1.5]. We are interested in polynomial g(k) -invariant divisors on Jk .
Proposition 16. (1) Let g be a Lie algebra of projectable vector fields on a fiber bundle π :
E → M and let D be a g(k) -invariant polynomial divisor on J k π. Then [D] = πk,0 ∗ L for some
g-equivariant line bundle L → E.

(2) Let g be a Lie algebra of point vector fields on J 0 (E, m) and D a g(k) -invariant polynomial
∗ L for some g(1) -equivariant line bundle L → J 1 (E, m).
divisor on J k (E, m). Then [D] = πk,1
−1
Proof. On J k (E, m) we will use the split coordinate charts πk,1 (Uαi1 ···im ). If π : E → M is a
−1
bundle, then the splitting is canonical and we use the charts πk,0 (Uα ) on J k π.
The prolongation of a vector field X = ai ∂xi + bj ∂yj ∈ g ⊂ D(E) is given by
X
X (k) = ai ∂xi + bjσ ∂yσj
0≤|σ|≤k
25
where bjσ are given recursively by (see [21, Th. 3.4])

j
bjσi = Dxi (bjσ ) − yσl Dxi (al ). (15)
When |σ| = d it is clear that bjσ is a sum of monomials of the form c(x, y)yσj11 · · · yσjss with |σl | ≤ d
for each l and |σ1 | + · · · + |σs | ≤ d + 1. Thus the weighted degree of bjσ is ≤ d + 1.
If D = {fαi1 ···im } is a polynomial invariant divisor on J k (E, m) of weighted degree d, then for
any X ∈ g the function X (k) (fαi1 ···im ) has weighted degree ≤ d + 1. The equality X (k) (fαi1 ···im ) =
−1
λiα1 ···im (X)fαi1 ···im implies that λiα1 ···im (X) = X (k) (fαi1 ···im )/fαi1 ···im is holomorphic on πk,1 (Uαi1 ···im )
if and only if the polynomial parts of the denominator is canceled out by the numerator. In
this case, λiα1 ···im (X) is polynomial of weighted degree ≤ 1, meaning that it is the pullback
of a function λ̃iα1 ···im (X) ∈ O(Uαi1 ···im ). Thus we see that the g(k) -equivariant line bundle over
J k (E, m) defined by the pair ({fαi1 ···im /fβj1···jm }, {λiα1 ···im }) is the pullback of a g(1) -equivariant
line bundle over J 1 (E, m).
If E → M is a fiber bundle, then we get a similar argument, but now λα (X) is the pullback
of a function in O(Uα ).
This proposition, whose second part was reformulated in Theorem 4, tells us that invariant
polynomial divisors on Jk are sections of pullbacks of equivariant line bundles over J1 or J0 .
In particular, they are controlled by H1 (Tot• (C)), where C p,q = C p,q (g(1) , {Uαi1 ···im }) or C p,q =
C p,q (g, {Uα }) or, more precisely, by Picg(r) (Jr ) for r = 1, 0 respectively. It is remarkable that
this fact is independent of the order k (one should compare to the statement of the Lie-Bäcklund
theorem [21, 23], although the proofs are different).
If the bundle π has, in addition, an affine structure, then we can consider divisors with local
defining functions fα in
−1
P−1 (Uα ) = {f ∈ O(π∞ (Uα )) | f |π∞
−1
(p) is a polynomial for every p ∈ Uα },
i.e., divisors that are polynomial on fibers of πk : J k π → M . These are preserved under (kth -
prolongation of) morphisms of affine bundles, and we will refer to them as “polynomial divisors”
in this context. For such a divisor D, we have [D] = πk∗ L for some line bundle L → M . We can
apply the same ideas as above to obtain the following result, which we leave without proof.
Proposition 17. Let g be a Lie algebra of projectable vector fields on an affine bundle π : E → M
that preserves the affine structure, and D be a g(k) -invariant polynomial divisor on J k π. Then
[D] = πk∗ L for some dπ(g)-equivariant line bundle L → M .
Example 10 (Riemannian geometry). Let π : S+ 2 T ∗ M → M denote the bundle of nondegenerate
symmetric 2-forms on M and g the Lie algebra of holomorphic vector fields on M , which induces
a Lie algebra g(k) of vector fields on J k π for k = 0, 1, . . . . If D a polynomial g(k) -invariant
divisor on J k π, then [D] is the pullback of a line bundle L → M . For example, if D is the
divisor on J 2 π that is given locally by the numerator of the scalar curvature of the metric, then
[D] = π2∗ (Λn T ∗ M )⊗4 .
Computations of invariants in jets often result in rational relative differential invariants, which
are related to polynomial differential invariants via a jet analogue of formula (14). This will be
demonstrated in the following examples.
3.3. Example A: Three-dimensional Heisenberg algebra on the plane. Consider the

following Lie algebra of vector fields on the plane:
g = h∂x , ∂y , y∂x i ⊂ D(C2 ).

It has the structure relations of the Heisenberg algebra and it prolongs naturally to the Lie
algebra g(1) of vector fields on J 1 (C2 , 1). Choosing y as the dependent variable gives
g(1) |U1 = h∂x , ∂y , y∂x − y12 ∂y1 i,
where U1 ⊂ J 1 (C2 , 1) denotes the open chart determined by our choice of dependent variable
on C2 . Taking instead x as the dependent variable results in a different chart U2 ⊂ J 1 (C2 , 1)
where the prolongation of g takes the form
g(1) |U2 = h∂x , ∂y , y∂x + ∂x1 i.
These two charts cover J 1 (C2 , 1) = U1 ∪ U2 . On overlap U1 ∩ U2 we get (x, y, y1 ) ≡ (x, y, 1/x1 ).
In each of the two charts we compute the Chevalley-Eilenberg cohomology:
H1 (g(1) , O(U1 )) = C2 , H1 (g(1) , O(U2 )) = 0.
A representative λ1 of a general element in H1 (g(1) , O(U1 )) takes the form
λ1 (∂x ) = 0, λ1 (∂y ) = 0, λ1 (y∂x − y12 ∂y1 ) = A + By1 , A, B ∈ C.
The compatibility condition λ1 (X)− λ2 (X) = X(g12 )/g12 , ∀X ∈ g(1) gives the general transition
function g12 = Cy1−B eA/y1 . This function is holomorphic on U1 ∩ U2 if and only if B ∈ Z.
Changing the representative (g, λ) ∈ C 0,1 × C 1,0 by the coboundary ∂ 0 µ where µ1 = 1, µ2 =
CeAx1 , we get g12 = y1−B and
λ1 = (0, 0, A + By1 ), λ2 = (0, 0, A).
Thus Picg(1) (J1 ) = C × Z → Pic(J1 ) = Z is epimorphic.
We identify a generating set (I, ∇) of absolute differential invariants in charts as follows:
y 1
2
− 3 , Dx on U1 ←→ x2 , Dy on U2 .
y1 y1
The invariant divisors on J 1 (C2 , 1) are generated by f1 = y1 , f2 = 1 of weight (A, B) = (0, −1).
Note that the invariant ODE y1 = 0 is not visible from the local computations on U2 . Indeed,
its solutions are y = const for the independent variable y, which are not graphs x = h(y).
General g(2) -invariant divisors on J 2 (C2 , 1) are generated by f = {f1 , f2 } and the absolute
invariant I. In particular, the irreducible invariant submanifolds of codimension 1 in J2 are
given by the divisors f˜ = {f˜1 , f˜2 } = {y2 − Cy13 , x2 + C} of weight (A, B) = (0, −3), parametrized
by C ∈ C.
Note that the non-zero parameter A above is not realizable by an invariant divisor (on J1 such
are y1−B ). Higher prolongations give no new weights of polynomial divisors and we conclude,
with the help of Proposition 16,
Z = jg(∞) Divrat
g(∞)
(J∞ ) ⊂ Picg(1) (J1 ) = C × Z.
3.4. Example B: Invariant divisors of curves in the projective plane. Consider the
Lie algebra sl(3, C) ⊂ D(CP 2 ) of projective vector fields. Differential invariants of curves in
the projective plane were studied already in 1878 by Halphen in his PhD thesis [15] (see also
the recent treatment [16] in the real case). In this section we demonstrate how the framework
developed in this paper sheds new light on those classical invariants.
The manifold CP 2 is covered by the three charts Ui = CP 2 \ {zi = 0}, i = 1, 2, 3, where
[z1 : z2 : z3 ] are homogeneous coordinates. Let us start by focusing on U3 with coordinates
x = z1 /z3 , y = z2 /z3 . In these local coordinates we have
sl(3, C)|U3 = h∂x , ∂y , y∂x , x∂y , x∂x − y∂y , x∂x + y∂y , x2 ∂x + xy∂y , xy∂x + y 2 ∂y i.
27
3.4.1. Equivariant line bundles. The cohomology group H1 (sl(3, C), O(U3 )) = C was computed
in [10, Table 3], and also in [27]. Our global approach shows that
Picsl(3,C) (CP 2 ) ≃ Pic(CP 2 ) = {OCP 2 (k) | k ∈ Z} ≃ Z.
Skipping the details of this computation, we instead focus on the corresponding computation
in J 1 (CP 2 , 1). Choosing y as the “dependent” variable we get an open coordinate chart U3y ⊂
J 1 (U3 , 1) in which the prolonged vector fields take the form
X1 = ∂x , X2 = ∂y , X3 = y∂x − y12 ∂y1 , X4 = x∂y + ∂y1 , X5 = x∂x − y∂y − 2y1 ∂y1 ,

(16)
X6 = x∂x + y∂y , X7 = x2 ∂x + xy∂y + (y − xy1 )∂y1 , X8 = xy∂x + y 2 ∂y + (y − xy1 )y1 ∂y1 .
Let us start by computing H1 (sl(3, C)(1) , O(U3y )). For a general cocycle λy3 , we define
ai (x, y, y1 ) := λy3 (Xi ) ∈ O(U3y ).
By subtracting a coboundary we can set a1 = 0. The cocycle condition involving X1 and X2

implies that ∂x (a2 ) = 0, and by subtracting a coboundary (now x-independent) we set a2 = 0.
The eight cocycle conditions
Xi (aj ) − Xj (ai ) − λ([Xi , Xj ]) = 0, 1 ≤ i ≤ 2, 3 ≤ j ≤ 6
reduce to Xi (aj ) = 0 and imply that a3 , a4 , a5 , a6 are independent of x and y.

By subtracting a coboundary (independent of x and y) we set a4 (y1 ) = 0. Then for the PDE
system defined by the remaining cocycle conditions, we get the general holomorphic solution:
a1 = 0, a2 = 0, a3 = A2 y1 , a4 = 0, a5 = A2 , a6 = A1 ,
3A1 + A2 3A1 − A2
a7 = x, a8 = A2 xy1 + y,
2 2
(1)
from which we see that H1 (sl3 , O(U3y )) = C2 .
A similar computation can be done in the open coordinate chart U3x ⊂ J 1 (U3 , 1), where x is
the dependent variable. In these coordinates, related to the previous by x1 = 1/y1 on overlap
U3y ∩ U3x , the prolonged vector fields take the form
X1 = ∂x , X2 = ∂y , X3 = y∂x + ∂x1 , X4 = x∂y − x22 ∂x1 , X5 = x∂x − y∂y + 2x1 ∂x1 ,

X6 = x∂x + y∂y , X̃7 = x2 ∂x + xy∂y + (x − yx1 )x1 ∂x1 , X8 = xy∂x + y 2 ∂y + (x − yx1 )∂x1 .
Defining bi (y, x, x1 ) := λx3 (Xi ) ∈ O(U3x ), and repeating the computations above, a general
(1)
representative of an element in H1 (sl3 , O(U3x )) is given by
b1 = 0, b2 = 0, b3 = 0, b4 = Ã2 x1 , b5 = −Ã2 , b6 = Ã1 ,

3Ã1 − Ã2 3Ã1 + Ã2
b7 = Ã2 yx1 + x, b8 = y,
2 2
implying H1 (sl(3, C)(1) , O(U3x )) = C2 .
The compatibility condition λy3 (X)−λx3 (X) = X(g33 yx yx
)/g33 implies that Ã1 = A1 and Ã2 = A2 .
yx
In this case the transition function has the form g33 = Cy1−A2 , and it is holomorphic if and only if
A2 ∈ Z. The constant C can be set equal to 1 via a suitable sl(3, C)(1) -invariant δ0,0 -coboundary.
Thus we conclude Picsl(3,C)(1) (J 1 (U3 , 1)) = C × Z.
Next, we perform similar computations on the remaining charts U1x , U1y , U2x , U2y of J 1 (CP 2 , 1).
In U2 ⊂ CP 2 we have coordinates (x̃, ỹ) = (z1 /z2 , z3 /z2 ). Choosing ỹ as dependent variable
results in coordinates (x̃, ỹ, ỹ1 ) on U2y ⊂ J 1 (CP 2 , 1). On U3y ∩ U2y we have x = x̃/ỹ, y = 1/ỹ and
y1 = ỹ1 /(x̃ỹ1 − ỹ). In these coordinates, the generators of sl(3, C)(1) are:
X1 = ỹ∂x̃ − ỹ12 ∂ỹ1 , X2 = −x̃ỹ∂x̃ − ỹ 2 ∂ỹ + (x̃ỹ1 − ỹ)ỹ1 ∂ỹ1 , X3 = ∂x̃ ,

X4 = −x̃2 ∂x̃ − x̃ỹ∂ỹ + (x̃ỹ1 − ỹ)∂ỹ1 , X5 = 2x̃∂x̃ + ỹ∂ỹ − ỹ1 ∂ỹ1 ,
X6 = −ỹ∂ỹ − ỹ1 ∂ỹ1 , X7 = −x̃∂ỹ − ∂ỹ1 , X8 = −∂ỹ .
Defining ci (x̃, ỹ, ỹ1 ) := λy2 (Xi ) ∈ O(U2y ) yields a general element in H1 (sl(3, C)(1) , O(U2y )):
3B1 + B2 3B1 + B2 3B1 − B2
c1 = ỹ1 , c2 = − x̃ỹ1 + ỹ, c3 = 0,
2 2 2
c4 = −B2 x̃, c5 = B2 , c6 = B1 , c7 = 0, c8 = 0.
The compatibility condition λy3 (X) − λy2 (X) = X(g32

yy yy
)/g32 implies that B1 = (A2 − A1 )/2 and
y y yy
B2 = (3A1 + A2 )/2. The transition function on U3 ∩ U2 is given by g32 = C̃ ỹ −B2 (x̃ỹ1 − ỹ)A2 . It
is holomorphic if and only if A2 , B2 ∈ Z. To sum up, we have
A2 ∈ Z and (3A1 + A2 )/2 ∈ Z. (17)
By doing a similar analysis on the intersection of the remaining charts, one gets
Picsl(3,C)(1) (J 1 (CP 2 , 1)) = Z2 . (18)
Furthermore, the map Picsl(3,C)(1) (J 1 (CP 2 , 1)) → Pic(J 1 (CP 2 , 1)) is injective, since we have
H̃1 (sl(3, C)(1) , J 1 (CP 2 , 1)) = 0.
Let us compare this to known bundles over J 1 (CP 2 , 1), starting with canonical bundles.
The line bundle Λ3 T ∗ J 1 (CP 2 , 1) → J 1 (CP 2 , 1) corresponds to (A1 , A2 ) = (−2, 2), while the
pullback of the line bundle Λ2 T ∗ CP 2 → CP 2 via π1,0 : J 1 (CP 2 , 1) → CP 2 corresponds to
(A1 , A2 ) = (−2, 0). This is easy to check by computing divergences of X1 , . . . , X8 with respect
to the volume forms Ω0 = dx ∧ dy and Ω1 = dx ∧ dy ∧ dy1 on U3 ⊂ CP 2 and U3y ⊂ J 1 (CP 2 , 1),
respectively: divΩ0 corresponds to (A1 , A2 ) = (2, 0) and divΩ1 corresponds to (A1 , A2 ) = (2, −2).
(Note that divergences with respect to different volume forms differ (locally) by a coboundary
in the modified Chevalley-Eilenberg complex.)
Furthermore, the pullback of the line bundle OCP 2 (1) → CP 2 corresponds to (A1 , A2 ) =
(2/3, 0) because of the relation between the canonical and tautological bundles over CP 2 (see
Remark 3). The vertical bundle V J 1 (CP 2 , 1) ⊂ T J 1 (CP 2 , 1) corresponds to (A1 , A2 ) = (0, −2),
while the subbundle hωi ⊂ T ∗ J 1 (CP 2 , 1) defined by the contact form ω ∈ Γ(T ∗ J 1 (CP 2 , 1))
corresponds to (A1 , A2 ) = (−1, 1). The subset (A1 , A2 ) ⊂ C2 satisfying (17) is generated by the
elements (2/3, 0) and (−1, 1). This leads to the following concrete description:
Proposition 18. Consider the standard realization of sl(3, C) ⊂ D(CP 2 ), and its prolongation
sl(3, C)(1) ⊂ D(J 1 (CP 2 , 1)). The equivariant Picard group (18) is
n o
Picsl(3,C)(1) (J 1 (CP 2 , 1)) = hωi⊗k1 ⊗ π1,0
∗
OCP 2 (k0 ) | k0 , k1 ∈ Z ≃ Z2 .
The integer parameters are related to the above weights like this: A1 = −k1 + 32 k0 , A2 = k1 .
3.4.2. Invariant divisors and absolute differential invariants. Generators for the absolute differ-
ential invariants are well-known, see e.g. [23, Table 5]. The field of rational absolute differential
invariants is generated by
R3 R2 R7
I7 = 78 , ∇ = Dx
R5 R53
29
−1
where R2 , R5 , R7 are expressed in the following way on π7,1 (U3y ):
R2 = y 2 ,
R5 = 9y22 y5 − 45y2 y3 y4 + 40y33 ,
R7 = 18y24 (9y22 y5 − 45y2 y3 y4 + 40y33 )y7 − 189y26 y62 + 126y24 (9y2 y3 y5 + 15y2 y42 − 25y32 y4 )y6
− 189y24 (15y2 y4 + 4y32 )y52 + 210y22 y3 (63y22 y42 − 60y2 y32 y4 + 32y34 )y5 − 4725y24 y44
− 7875y23 y32 y43 + 31500y22 y34 y42 − 33600y2 y36 y4 + 11200y38 .
We use a different set of generators than [23] in order to obtain rational invariants, which
by [18] are sufficient to separate orbits in general position. Table 5 in [23] also contains the
Lie determinant R2 R52 on the locus of which the orbit dimension drops. The Lie algebra
sl(3, C)(6) acts simply transitively on the complement of {R2 R5 = 0} ⊂ π6,1 −1
(U3y ); note that
dim J 6 (CP 2 , 1) = dim sl(3, C). A complete description of the orbit structure (over R) can be
found in [16].
Remark 6. Proposition 10 gains the following insight. Computing orbit dimensions of g =
sl(3, C) in J 4 (CP 2 , 1) shows that an invariant divisor exist only if A2 = 3A1 , in which case it
is y2−A1 , but this is meromorphic only if A1 ∈ Z. For k ≥ 5, the generic orbit dimension of
g(k) on J k (CP 2 , 1) is the same as that of (g(k) )λ , independently of λ. The general invariant
(A −3A1 )/6
divisor is given by R22A1 −A2 R5 2 , however this function is meromorphic if and only if
(2A1 − A2 ), (A2 − 3A1 )/6 ∈ Z. Together with (17) this implies that weights (A1 , A2 ) belong to
the lattice generated by (3, −3) and (2, 0).
−1
The polynomials R2 , R5 , R7 are local expressions, defined on π7,1 (U3y ), for invariant polynomial
divisors. But they extend uniquely to polynomial divisors on J 7 (CP 2 , 1). For R2 , R5 and R7 ,
the weight λy3 is given by (A1 , A2 ) = (−1, −3), (A1 , A2 ) = (−6, −12) and (A1 , A2 ) = (−16, −32),
respectively. In particular, R2 and R5 do not combine to a rational absolute differential invariant
(weight 0), which is consistent with the fact that g(6) has an open orbit on J 6 (CP 2 , 1). It is also
clear that R2 and R5 are local generators for polynomial invariant divisors on J 6 (CP 2 , 1) since
they generate a 2-dimensional space of weights.
Combining weights of the invariant divisors, we obtain the above absolute invariant I7 together
with the following invariant meromorphic tensor fields:
R5 ∗ R7
α5 = 4 dx ∧ dy ∈ Γ(π5,0 Λ2 T ∗ CP 2 ), α7 = 3 2 (dy − y1 dx) ∈ Γ(π7,1 ∗
hωi).
R2 R2 R5
R42
The inverse bivector α−1
5 = R5 Dx ∧∂y contracted with α7 gives the invariant derivation ∇ above.
Remark 7. These tensor fields can be compared to those of Theorem 5.1 of [16]. Their
2/3
R2−3 R5 (dy − y1 dx) is multi-valued over C, but its cube is the rational invariant tensor I7−1 α37 .
Note that in general polynomial divisors Divpol
g (M ) determine a weight sub-monoid in Picg (M ),
rat
while rational divisors Divg (M ), obtained as ratios of the former, determine a lattice.
Theorem 19. The lattice generated by polynomial divisors for g = sl(3, C) acting on J ∞ (CP 2 , 1)
is a sublattice of order 3 in the equivariant Picard group on 1-jets:

Z2 ≃ jg(∞) Divrat ∞ 2
( Picg(1) J 1 (CP 2 , 1) ≃ Z2 .

g(∞) J (CP , 1)
This is basically a summary of the computations. Indeed, from the tensor fields α5 , α7
we see that (pullbacks of) line bundles in Picsl(3,C)(1) (J 1 (CP 2 , 1)) are realized as [D] for some
rational sl(3, C)(7) -invariant divisor D on J 7 (CP 2 , 1) when k0 /3, k1 ∈ Z, where k0 and k1 are the
parameters used in Proposition 18. To understand why OCP 2 (1) is not realized in this way one
must consider which Lie group is acting here. The Lie algebra g = sl(3, C) on CP 2 integrates
to the Lie group G = P GL(3, C), and then results from Example 7 apply.
Remark 8. Non-degenerate curves in CP n up to projective transformations g = sl(n + 1, C)
were studied by Wilczynski [31]. He computed fundamental differential invariants via the corre-
spondence with linear ordinary differential equations of order n+1. Our results generalize to give
two-dimensional lattice Picg(1) J 1 (CP n , 1) , which constrains the weights of relative differential

invariants.
3.5. Example C: Second-order ODEs modulo point transformations revisited. Finally
for h ∈ O(U ), U ⊂ J 1 (CP 2 , 1), we consider scalar second-order ODEs
{y2 = h(x, y, y1 )} ⊂ J 2 (CP 2 , 1) (19)
together with the Lie algebra sheaf g = D(J0 ) of germs of holomorphic vector fields on J0 = CP 2 .
Here and throughout this section we use the notation Js = J s (CP 2 , 1), while J k (J1 ) consists of
k-jets of functions h on J1 . Our goal is to find generators for the invariant divisors on J 4 (J1 ).
Relative invariants were first found by A. Tresse in [29] via Lie theory and then by E. Car-
tan via his theory of moving frames [5]. We apply our global framework to justify the (two-
dimensional) weight lattice introduced in [17] and generate relative invariants for this classical
problem in a novel and conceptually transparent manner.
Any vector field on J0 prolongs uniquely to a vector field on J2 . This action induces an
(infinitesimal) transformation on the space of second-order ODEs. Choose local coordinates x, y
on CP 2 , denote p = y1 , u = y2 the induced coordinates on J2 , then an ODE is a hypersurface
u = h(x, y, p) in J 0 (J1 ) = J2 . Redefining g to be the image (prolongation) of D(J0 ) in J2 , its
further prolongation, the Lie algebra g(k) ⊂ D(J k (J1 )), is spanned by the vector fields of the
form X
aDx + bDy + cDp + Dσ(k) (ψ)∂uσ (20)
|σ|≤k
where a, b are functions of x, y, c = (∂x + p∂y )ϕ for ϕ = b − pa, Dx is the operator of total
derivative by x and similar for Dy , Dp , while Dσ is their composition for multi-indices of variables
(see [17]), and the function ψ is given by
ψ = (∂x + p∂y )2 ϕ + u(∂y ϕ − 2(∂x a + p∂y a) − aux − buy − cup .
The Lie algebra g(0) = g preserves the fibers of the affine bundle J 0 (J1 ) → J1 (and their
affine structure). Thus, in order to compute invariant divisors that are polynomial on fibers
of J k (J1 ) → J1 , we exploit Proposition 17 and start with classification of g-equivariant line
bundles on J1 .
3.5.1. g-equivariant line bundles. In Example B we saw that the sl(3, C)(1) -equivariant line bun-
dles on J1 were generated by the line bundles π1,0
∗ O ∗ 1
CP 2 (1) and hωi ⊂ T J . Since sl(3, C)
(1) ⊂ g,
we have a natural homomorphism

Picg (J 1 (CP 2 , 1)) → Picsl(3,C)(1) (J 1 (CP 2 , 1)). (21)
Proposition 20. Homomorphism (21) is an isomorphism.
Proof. We first prove that (21) is surjective. Clearly, the bundle OCP 2 (−3) ≃ π1,0 ∗ Λ2 T ∗ CP 2
admits a g-lift, due to naturality of the cotangent bundle. The bundle hωi ⊂ T J1 admits a
∗
g-lift since the prolongation preserves the Cartan distribution Ann(ω) ⊂ T J1 . What remains to
be seen is that OCP 2 (1) admits a g-lift. On OCP 2 (1), the local weight λ3 of a general vector field
X = a(x, y)∂x + b(x, y)∂y on U3 (for example) is λ3 (X) = (ax + by )/3, and it is not difficult to
check that this extends to a compatible weight for each X ∈ g.
31
Now we prove injectivity. Let [(g, λ)] ∈ Picg (J1 ) be in the kernel of (21). Then [g] = 0 ∈
Pic(J1 ) and there exists a representative for [λ] such that λ|sl(3,C)(1) = 0.
Take an arbitrary point in J1 and choose a chart with coordinates centered at this point
(origin). Due to transitivity of sl(3, C)(1) on J1 we can assume, without loss of generality, that
the coordinate chart is U3y from Section 3.4. We will compute λ|U3y . It is clear that if λ(X) 6= 0
for some X ∈ g, then λ(X)|U3y 6= 0 since U3y ⊂ J1 is a dense subset.
We continue with the notation from Section 3.4, so that sl(3, C)(1) |U3y = hX1 , · · · , X8 i with
Xi given by (16). We have λ(X1 ) = · · · = λ(X8 ) = 0. Next, consider the vector fields
Y1 = x2 ∂y + 2x∂y1 , Y2 = x2 ∂x − 2xy∂y − (4xy1 + 2y)∂y1 ,
Y3 = y 2 ∂y − 2xy∂x + (2xy12 + 4yy1 )∂y1 , Y4 = y 2 ∂x − 2yy12 ∂y1 .
The commutation relations
[X1 , Y1 ] = 2X4 , [X2 , Y1 ] = 0, [X4 , Y1 ] = 0,
[X1 , Y2 ] = 2X5 , [X2 , Y2 ] = −2X4 , [X4 , Y2 ] = −3Y1 ,
[X1 , Y3 ] = −2X3 , [X2 , Y3 ] = −2X5 , [X4 , Y3 ] = −2Y2 ,
[X1 , Y4 ] = 0, [X2 , Y4 ] = 2X3 , [X4 , Y4 ] = −Y3 ,
[X6 , Yi ] = Yi , i = 1, 2, 3, 4,
give four differential equations on each function λ(Yi ), implying λ(Y1 ) = · · · = λ(Y4 ) = 0.
Furthermore, all polynomial vector fields are generated by X1 , . . . , X8 and Y1 , . . . , Y4 . Indeed,
for j ≥ 3 we have
1
xi y j−i ∂x = [x2 ∂x , xi−1 y j−i ∂x ], i 6= 0, 3,
i−3
1
y j ∂x = [y 2 ∂y , y j−1 ∂x ],
j−1
1
x3 y j−3 ∂x = [x2 ∂y , xy j−2 ∂x ] + 2x2 y j−2 ∂y ,

j−2
and by swapping x and y we also generate xi y j−i ∂y for i = 0, . . . , j. Thus all vector fields with
polynomial coefficients of degree ≥ 3 are of the form [Z, Y ], where the coefficients of Y have
degree 2 and the coefficients of Z have degree strictly lower than those of [Z, Y ]. Then the
general cocycle condition
λ([X, Y ]) = X(λ(Y )) − Y (λ(X))
implies that λ(X) = 0 for any polynomial vector field X on U3y .
On any compact subset K ⊂ U3y , the subspace of vector fields in D(K) with polynomial
coefficients is dense in D(K). It follows that λ(X)|K = 0 for every X ∈ g for any K, and hence
that λ(X)|U3y = 0 for every X ∈ g. Thus λ = 0.
3.5.2. Invariant divisors. Now we compute the g(4) -invariant divisors on J 4 (J1 ). Let us work
in the coordinate chart τ4−1 (U3y ), where τ4 denotes the projection τ4 : J 4 (J1 ) → J1 . From
Proposition 20, we know that [λ] ∈ H 1 (g, O(U3y )) has a representative of the form
λ = C0 divdx∧dy + C1 divdx∧dy∧dy1 ,
where (C0 , C1 ) is related to (A1 , A2 ) by A1 = 2(C0 + C1 ) and A2 = −2C1 . Condition (17) is
equivalent to 3C0 , 2C1 ∈ Z. If f is a general polynomial of some fixed degree, then the system
X (k) f = λ(X)f, X ∈ g(0)
reduces to a linear system on the coefficients of f for each choice of (C0 , C1 ). By sequentially
setting C0 = 0, ±1/3, ±2/3, . . . and C1 = 0, ±1/2, ±1, . . . and letting f be a general polynomial
of degree 3 with undetermined coefficients, we get a series of linear systems determining the
coefficients of the polynomial. In this way, we obtain the solutions
f1 = upppp,
f2 = uxxpp + 2puxypp + 2uuxppp + p2 uyypp + 2puuyppp + u2 upppp + (uy uppp − up uypp − 4uyyp )p
− 3uuypp + (−uxpp + 4uyp )up + ux uppp − 3uy upp + 6uyy − 4uxyp ,
which have weights (C0 , C1 ) = (2, −5/2) and (−2, 1/2), respectively. Computing the rank of
prolonged vector fields at generic point we conclude that the action of g(4) has an open orbit
in J 4 (J1 ). Thus there are no (nonconstant) absolute invariants on J4 . Now, if f3 was another
invariant divisor of general weight (C0 , C1 ) = (2A − 2B, (B − 5A)/2) with rational A, B, then
for some integer m the ratio
f3m
f1Am f2Bm
is a rational function with weight (0, 0) and hence is an absolute differential invariant, and
therefore constant. Hence f3m is proportional to f1Am f2Bm .
Taking into account Proposition 17 we conclude the following.
Theorem 21. The lattice generated by polynomial divisors for the Lie algebra g = D(J0 ) acting
on J ∞ (J1 ) is a sublattice in the equivariant Picard group on 1-jets:

Z2 ≃ jg(∞) Divgrat ∞ 1
(J ) ( Picg J1 ≃ Z2 .

(∞) J
Let us note that cohomology of line bundles was explored in [14] to compute Cartan invariants
of projective connections, which correspond to a particular class of ODEs of the form (19) with
h cubic in y1 ; our methods though are quite distinct.
4. Outlook
In this work we proposed a theory of global scalar relative differential invariants, based on
familiar notions of divisors and line bundles. While G-equivariant line bundles were known for
algebraic and compact groups, the more general notions of equivariant Picard group Picg (M )
and invariant divisor group Divg (M ) for a Lie algebra g appear to be new and have certain
subtleties. (These notions even extend to Lie algebra sheaves, as seen in Example C.)
The basic setup is analytic, but we also consider polynomial divisors in affine bundles. Such
bundles arise in successive jet-prolongation, and polynomial relative differential invariants are
natural and sufficient in the equivalence problem of invariant hypersurfaces. We thus explore
polynomial divisors in jet spaces. While j Div(J∞ ) = j Div(J1 ) in Pic(J∞ ) = Pic(J1 ) (in the

case of fiber/affine bundle π this can be pushed down to J0 , resp. M ), the g-equivariant coun-
terpart is more complicated. In general, Picg(∞) (J∞ ) 6= Picg(1) (J1 ), and similarly for invariant
divisors. However weights of invariant polynomial divisors are 1-jet determined, as Propositions
16 and 17 state. This gives an effective bound on multipliers for relative invariants and, in many
cases, an algorithmic approach to compute them.
Invariant submanifolds of higher codimensions are related, in the same manner, to higher rank
equivariant vector bundles. While there are no general tools that classify analytic/algebraic
vector bundles of higher rank, some part of the theory generalizes. Weights of vector-valued
relative invariants are matrix-valued cocycles, leading to a more general cohomology theory.
Lastly, there is a differential algebra aspect to the theory of invariant divisors on jet bundles.
The structure theory of these global relative differential invariants will be discussed elsewhere.
33
Acknowledgments. The research leading to our results has received funding from the
Norwegian Financial Mechanism 2014-2021 (GRIEG project SCREAM, registration number
2019/34/H/ST1/00636) and the Tromsø Research Foundation (project “Pure Mathematics in
Norway”), as well as UiT Aurora project MASCOT.
References
[1] I. Anderson, M. Fels, Transverse group actions on bundles, Topology and Appl. 123, 443-459 (2002).
[2] V. I. Arnold, Geometrical Methods in the Theory of Ordinary Differential Equations, Springer (1988).
[3] M. Brion, Linearization of algebraic group actions. In: Handbook of group actions (Vol. IV), ALM 41,
291-340, Int. Press, Somerville, MA (2018).
[4] J. F. Cariñena, M. A. del Olmo, P. Winternitz, On the Relation Between Weak and Strong Invariance of
Differential Equations, Letters in Mathematical Physics 29, 151-163 (1993).
[5] E. Cartan, Sur les variétés à connexion projective, Bull. Soc. Math. France 52, 205-241 (1924).
[6] C. Chevalley, S. Eilenberg, Cohomology theory of Lie groups and Lie algebras, Trans. Amer. Math. Soc. 63,
85-124 (1948).
[7] M. Fels, P. Olver, On Relative Invariants, Math. Ann. 308, 701-732 (1997).
[8] O. Forster, Lectures on Riemann Surfaces, Springer, New York (1981).
[9] D. B. Fuks, Cohomology of infinite-dimensional Lie algebras, Consultants Bureau, New York (1986).
[10] A. González-López, N. Kamran, P. Olver, Lie Algebras of Differential Operators in Two Complex Variables,
American Journal of Mathematics 114, 1163-1185 (1992).
[11] P. Griffiths, J. Harris, Principles of Algebraic Geometry, John Wiley & Sons, Inc. (1978).
[12] A. Grothendieck, On the de Rham cohomology of algebraic varieties, Publications mathématiques de l’I.H.É.S.
29, 95–103 (1966).
[13] G.-H. Halphen, Sur les invariants différentiels, Gauthier-Villars (1878).
[14] J. C. Hurtubise, N. Kamran, Projective connections, double fibrations, and formal neighborhoods of lines,
Math. Ann. 292, 383-409 (1992).
[15] D. Huybrechts, Complex Geometry: An Introduction, Springer Berlin, Heidelberg (2005).
[16] N. Konovenko, V. Lychagin, On projective classification of plane curves, Global and Stochastic Analysis 1,
no. 2, 241-264 (2011).
[17] B. Kruglikov, Point classification of 2nd order ODEs: Tresse classification revisited and beyond. In: Differ-
ential Equations - Geometry, Symmetires and Integrability. Abel Symposia 5, 199-221, Springer (2009).
[18] B. Kruglikov, V. Lychagin, The global Lie-Tresse theorem, Selecta Math. 22, 1357-411, (2016).
[19] B. Kruglikov, V. Lychagin, Geometry of differential equations, in D. Krupka, D. Saunders (eds.), Handbook
of Global Analysis, 725-771, Elsevier (2016).
[20] B. Kruglikov, E. Schneider, ODEs whose Symmetry Groups are not Fiber-Preserving, Journal of Lie Theory
33, 1045-1086 (2023).
[21] A. M. Vinogradov, I. S. Krasil’shchik (editors), Symmetries and Conservation Laws for Differential Equations
of Mathematical Physics, American Mathematical Society, Providence (1999).
[22] D. Mumford, Geometric Invariant Theory, Springer-Verlag (1965).
[23] P. Olver, Equivalence, Invariants, and Symmetry, Cambridge University Press (1995).
[24] P. Olver, Classical invariant theory, Cambridge University Press (1999).
[25] P. Olver, Projective invariants of images, European Journal of Applied Mathematics 34, 936-946 (2023).
[26] M. Rosenlicht, Toroidal algebraic groups, Proc. Amer. Math. Soc. 12, 984-988 (1961).
[27] E. Schneider, Projectable Lie algebras of vector fields in 3D, Journ. Geom. and Physics 132, 222-229 (2018).
[28] W. Ferrer Santos, A. Rittatore, Actions and Invariants of Algebraic Groups, Chapman and Hall/CRC (2005).
[29] M. A. Tresse, Détermination des invariants ponctuels de l’équation différentielle ordinaire du second ordre
y ′′ = ω(x, y, y ′ ), S. Hirzel, Leipzig (1869).
[30] H. Weyl, The Classical Groups. Their Invariants and Representations, Princeton University Press (1939).
[31] E. J. Wilczynski, Projective differential geometry of curves and ruled surfaces, Leipzig, Teubner (1905).
†
Department of Mathematics and Statistics, UiT the Arctic University of Norway, Tromsø
9037, Norway.
Email addresses: [email protected] and [email protected] .

2404 19439v1

Uploaded by

Copyright:

Available Formats

2404 19439v1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2404 19439v1

Uploaded by

Copyright:

Available Formats

INVARIANT DIVISORS AND EQUIVARIANT LINE BUNDLES

BORIS KRUGLIKOV† AND EIVIND SCHNEIDER†

for some weight λα ∈ g∗ ⊗ O(Uα ), which is a 1-cocycle in the Chevalley-Eilenberg complex of g

whence a double homomorphism (Ψ1 , Ψ2 ) such that Ψi ◦ ̟ = Φi and ker Ψ1 ∩ ker Ψ2 = 0:

When Tg (M ) = 0, Φ1 × Φ2 embeds Picg (M ) in Pic(M ) × Mg (M ) (Corollary 3 gives two

Corollary 1. (i) If Mg (M ) = 0 and Tg (M ) = 0 then Φ1 : Picg (M ) → Pic(M ) is injective.

In Section 2.4 we consider the homomorphism

Theorem 2. If D = {fα } is a g-invariant divisor and λ = {λα } is the corresponding weight,

Theorem 3. If D is a ĝ-invariant polynomial divisor on the affine bundle E, then [D] = π ∗ L

2. Analytic invariant divisors and equivariant line bundles

with differentials given by

In particular, δ0 and δ1 are defined in the following way:

where C p,q are given by

and for p > 0 by

H0d (C •,0 ) H1d (C •,0 ) H2d (C •,0 )

H0d (C •,1 ) H1d (C •,1 ) H2d (C •,1 )

H0d (C •,2 ) H1d (C •,2 ) H2d (C •,2 )

d0,0 d1,0 d2,0

d0,1 d1,1 d2,1

d0,2 d1,2 d2,2

Thus we get the induced cohomology groups

Lemma 1. The maps Φ1 , Φ2 have the following kernels:

Lemma 1 is equivalent to exactness of the two sequences

Lemma 2. The images of Φ1 , Φ2 are given by the following exact sequences:

Definition 2. The group of multipliers of g on M is the direct limit Mg (M ) := lim Mg (U ).

ker(∂ 1 ) ker(∂ 1 ) H1 (Tot• (C))

Proposition 6. The following sequence is exact:

Ȟ1 (M, (O× )g ) H̃1 (g, O(M ))

Figure 1. Commutative diagram: The dotted and dashed long sequences as

Corollary 3. If Ȟ1 (M, (O× )g ) = 0 or H̃1 (g, O(M )) = 0, then Picred

Propositions 6 and 7 combine into Theorem 1.

Lemma 3. Let g be a transitive Lie algebra of vector fields on a manifold M . Define Z =

the isotropy of the point p = 0

h0 = hy∂x , −xy∂x − y 2 ∂y , −x2 ∂x − xy∂y , 2x∂x + y∂y i

· · · → Ȟ0 (M, M× ) → Div(M ) → Pic(M ) → Ȟ1 (M, M× ) → · · · (10)

The lifted Lie algebra gλ ⊂ D([D]) is given locally on π −1 (Uα ) ≃ Uα × C by

Next, changing the fiber coordinates vα = µα uα for µα ∈ O× (Uα ),

results in the equivalence relation

ϕ̂∗t (uα ) = Λαα (ϕt )uα .

When t approaches 0, then Uα ∩ ϕ−1

3. Invariant polynomial divisors on algebraic bundles

a polynomial in u1 , . . . , ur of degree s. We have

Example 9. Consider the bundle E = S 2 T ∗ M → M whose sections are symmetric 2-forms

for some k, r ∈ Z≥0 and a collection {Fτσ } of functions on J j (E, m).

g-equivariant line bundle L → E.

where bjσ are given recursively by (see [21, Th. 3.4])

Example 10 (Riemannian geometry). Let π : S+ 2 T ∗ M → M denote the bundle of nondegenerate

3.3. Example A: Three-dimensional Heisenberg algebra on the plane. Consider the

g = h∂x , ∂y , y∂x i ⊂ D(C2 ).

Picsl(3,C) (CP 2 ) ≃ Pic(CP 2 ) = {OCP 2 (k) | k ∈ Z} ≃ Z.

X1 = ∂x , X2 = ∂y , X3 = y∂x − y12 ∂y1 , X4 = x∂y + ∂y1 , X5 = x∂x − y∂y − 2y1 ∂y1 ,

ai (x, y, y1 ) := λy3 (Xi ) ∈ O(U3y ).

By subtracting a coboundary we can set a1 = 0. The cocycle condition involving X1 and X2

Xi (aj ) − Xj (ai ) − λ([Xi , Xj ]) = 0, 1 ≤ i ≤ 2, 3 ≤ j ≤ 6

reduce to Xi (aj ) = 0 and imply that a3 , a4 , a5 , a6 are independent of x and y.

X1 = ∂x , X2 = ∂y , X3 = y∂x + ∂x1 , X4 = x∂y − x22 ∂x1 , X5 = x∂x − y∂y + 2x1 ∂x1 ,

b1 = 0, b2 = 0, b3 = 0, b4 = Ã2 x1 , b5 = −Ã2 , b6 = Ã1 ,

X1 = ỹ∂x̃ − ỹ12 ∂ỹ1 , X2 = −x̃ỹ∂x̃ − ỹ 2 ∂ỹ + (x̃ỹ1 − ỹ)ỹ1 ∂ỹ1 , X3 = ∂x̃ ,

The compatibility condition λy3 (X) − λy2 (X) = X(g32