On Lie algebroids and Poisson algebras

Dennise Garca-Beltran

, Jose A. Vallejo

and Yuri Vorobjev

Facultad de Ciencias, Universidad Autonoma de San Luis Potos, Mexico,

Departamento de Matematicas, Universidad de Sonora, Mexico.

Email: dennise.gb@alumnos.uaslp.edu.mx,jvallejo@fc.uaslp.mx,yurimv@guaymas.uson.mx
June 8, 2011
Abstract
We introduce and study a class of Lie algebroids associated to faithful modules which is
motivated by the notion of cotangent Lie algebroids of Poisson manifolds. We also give a
classication of transitive Lie algebroids and describe Poisson algebras by using the notions of
algebroid and Lie connections.
1 Introduction
Let (M, ., .) be a Poisson manifold equipped with the bracket ., ., which is determined by a
Poisson bivector P
2
TM. It is well known that the cotangent bundle T

M carries a natural Lie

algebroid structure (see Section 2 for denitions), that is, on dierential 1forms (sections of T

M)
the following bracket can be dened (see [2], Proposition 14.19),
[[, ]] = i
P
d i
P
d + d((P)), (1)
where P : T

M TM is the vector bundle morphism canonically induced by P through (P) =

P(, ). Moreover, the following Leibniz-like rule is satised, for any smooth function f (

(M):
[[, f]] = f[[, ]] + (P)(f) .
That means that the corresponding anchor map is just P. Note that in the case when , in (1) are
exact ( = df and = dg for some f, g (

(M)), we have P(, ) = f, g and formula (1) reads:

[[df, dg]] = df, g.
Moreover, the bracket (1) of two closed forms is again closed.
In fact, the properties above characterize the Lie algebroid structure on the tangent bundle that
comes from a Poisson bracket on the base manifold (see [6]). More precisely, given a Lie algebroid
(T

M, [[, ]], ) on the tangent bundle T

M, there exists a Poisson bracket ., . on M such that

= P if and only if the following conditions are satised:
(a) is skew-symmetric, i.e. (()) = (()), for all ,
1
(M).
(b) If ,
1
(M) are closed, then [[, ]] is also closed.
The aim of the present paper is to put the study of cotangent Lie algebroids of Poisson manifolds
in an algebraic framework. In the rst part (Section 3), we study and characterize a class of Lie
algebroids with properties similar to (a), (b), which we call Lie algebroids of Poisson type. To this

Corresponding author.
1
end, we will work in a slightly more general context than that of vector bundles over manifolds, using
the notion of Lie-Rinehart algebra (see [9, 10]), also called a Lie pseudoalgebra (see [14, 17, 18, 19]
and, particularly, [15] for many interesting remarks on the evolution of these notions and as a general
reference), although we will continue using the denomination Lie algebroid for them (as they are
the algebraic version of the geometric Lie algebroids). We give several examples illustrating the
dierent situations that can appear.
In the second part (Sections 4 and 5), we deepen in the relationship between transitive algebroids
and Poisson structures for a certain class of spaces, those of the form Der(/) V , where / is
a commutative algebra and V an /module. We describe parametrizations of the transitive Lie
algebroids on Der(/) V following the techniques exposed in [20, 21], which are based on the use
of a connection on a Lie algebroid. For completeness, we include a subsection in the preliminaries
devoted to the topic of connections in an algebraic setting. Once the parametrization is given, we
apply it to prove that a transitive algebroid endowed with a connection is isomorphic to one of the
form Der(/) V . Finally, we obtain new classes of Poisson algebras on /V starting from Poisson
algebras on /.
2 Preliminaries
Throughout the paper, unless otherwise explicitly stated, / denotes an associative, commutative
algebra with identity element 1
A
, over a commutative ring 1 with identity element 1
R
.
2.1 Derivations and connections in commutative algebras
In subsequent sections, we will need to introduce connections on an algebroid. In our algebraic
setting, the most appropriate notion of connection is Koszuls one, which is given in terms of deriva-
tions.
Denition 2.1. A derivation of the algebra / over 1 is a map X Hom
R
(/, /) satisfying the
Leibniz rule
X(f g) = X(f) g + f X(g).
The set of derivations of / over 1 is denoted Der
R
(/) or simply Der(/) when there is no risk of
confusion about the ring 1.
Remark 2.2. The denition just given can be extended to the case of derivations of / over 1 into
an /module M. These are abelian groups morphisms X : / M satisfying the Leibniz rule
above, and form an /module denoted Der
R
(/, M) or simply Der(/, M).
The set Der(/) has an 1Lie algebra structure when endowed with the commutator of endomor-
phisms, given by [X, Y ] = X Y Y X.
Note also that, if / as an 1module is faithful, then for every X Der(/) we have X(1
A
) = 0 and
indeed X(r) = 0 for every r 1 viewed as a subalgebra of /.
Denition 2.3. Let M be a unitary /module. A derivation law, or Koszul connection, on M is
an /linear mapping : Der(/) Hom
R
(M, M) (the image of X Der(/) denoted
X
) such
that

X
(f m) = X(f) m + f
X
(m).
Not every /module M admits a connection in this sense, but it is easy to see that any free
/module does. Of course, arbitrary /modules do not need to be free. So, in order to obtain a
big enough class of modules for which we can guarantee the existence of a Koszul connection, we
will make a brief digression on modules of dierentials and Connes connections (see [3]).
Let
1
(/) be the /module dened by the kernel of the multiplication /
R
/ /. Dene the
map d : /
1
(/) by da = 1 aa1, which is a derivation of / over 1 with values into
1
(/).
2
It is clear from the denition that
1
(/) = Span
A
df : f /: Since the elements of
1
(/) lie in
the kernel of the multiplication map, if

a
j
b
j

1
(/), then

a
j
b
j
= 0 and therefore

a
j
b
j
=

(a
j
b
j
a
j
b
j
1) =

a
j
db
j
.
In fact,
1
(/) is the submodule of C
1
(Der(/), /) (the 1component of the dierential algebra
C(Der(/), /) of Chevalley-Eilernberg cochains of the Lie algebra Der(/) with values in the Der(/)module
/) generated by the elements df, f / (see [5]). Note, in particular, that this implies
1
(/)
Der

(/).
Denition 2.4. Let M be an /module. A Connes connection on M is an /linear map : M

1
(/)
A
M such that, for all f /, m M,
(fm) = f(m) + da
A
m.
Remark 2.5. Connes denition of a connection (see [3]) actually does not require that / be a
commutative algebra. The denition goes back to a work by N. Katz [12].
Starting from a Connes connection, we can obtain a Koszul one. If X Der(/), then we dene
a right /linear pairing : Der(/)
R

1
(/) / by
(X,

a
j
db
j
)

a
j
X(b
j
).
The Koszul connection associated to can be constructed as follows: for X Der(/),
X

Hom
R
(M, M) is the map given by applying the connection and then contracting the
1
(/) com-
ponent with . Thus, if m M is such that (m) =

a
j
db
j
m
j
, for certain a
j
, b
j
/ and
m
j
M, we have for each f / that (fm) = f

a
j
db
j
m
j
+ df m, and

X
(fm) =

(X, fa
j
db
j
)m
j
+ (X, df)m
= f

a
j
X(b
j
)m
j
+ X(f)m
= f
X
(m) + X(f)m.
A basic result obtained by J. Cuntz and D. Quillen (see [4]) is that Connes connections on an
/module M are in bijective correspondence with /linear splittings of the natural action /
R
M M. As a consequence, M admits a Connes connection if and only if it is projective.
As said earlier, we will need later on to work with (Koszul) connections, so we need conditions on
/ to assure their existence. From what we have seen, these connections exist on any /module
M which is free or projective. Indeed, note that a free module is always projective, but there are
projective modules which are not free. In the literature, there are several well-known conditions on
/ guaranteeing the projective character of M (for example, that / be semi-simple as a ring). When
we talk of a connection on M, unless otherwise explicitly stated, we will mean that any one of these
conditions is satised and that the connection is Koszul.
2.2 Lie algebroids
Denition 2.6. Let T be a faithful /-module. A Lie algebroid is a triple (T, [[, ]], ), where [[, ]]
is a Lie bracket on T and : T Der(/) is a morphism of /-modules, called the anchor map, such
that:
[[X, fY ]] = f[[X, Y ]] + (X)(f)Y,
for all f / and for all X, Y T.
Remark 2.7. Sometimes, the condition that the anchor map be a morphism of Lie algebras is
included in the denition of Lie algebroid. However, this fact is a consequence of the conditions
in denition 2.6, as have been noted by J. C. Herz , Y. Kosmann-Schwarzbach, F. Magri and J.
Grabowski among others (see [8, 14, 7]).
3
Denition 2.8. Let (T, [[, ]], ) and (T

, [[, ]]

) be Lie algebroids (over the same algebra / and

the same ring 1). A morphism of Lie algebroids is a morphism of /-modules : T T

such that

= and ([[X, Y ]]) = [[(X), (Y )]]

,
for all X, Y T.
Let us consider some examples. The rst is the classical one.
Example 2.9. Let M be a manifold. Let E

M be a vector bundle over M and T = (E) the
(

(M)-module of sections of E (i.e. / = (

(M) and 1 = R). Then, the Lie algebroid structure

on (E) is dened by a Lie bracket [[, ]] on (E) with an anchor map
q : (E) Der
R
(

(M)

= (TM),
such that for all f (

(M) and for all X, Y (E):

1. [[X, fY ]] = f[[X, Y ]] + q(X)(f)Y,
2. q(fX + Y ) = fq(X) + q(Y ).
In particular, if E = T

M, then the Lie algebroid structure is given by the bracket (1) where the
anchor is the Poisson mapping P. For E = TM we have the trivial Lie algebroid, where q = Id
TM
.
Example 2.10. Consider the 1-algebra of dual numbers over /,
/

= /[] = x + y : x, y /,
2
= 0,
with the obvious operations. Clearly, /

is an /

-module and we can endow it with the Lie algebra

structure given by the bracket:
[[x
1
+ y
1
, x
2
+ y
2
]] = (x
1
y
2
y
1
x
2
)
for x
1
+ y
1
, x
2
+ y
2
/

. Thus (/

, [[, ]], ) is a Lie algebroid with anchor map

: /

Der(/

),
x + y ad
x
for x + y /

. Here ad
x
(x
1
+ y
1
) = [[x, x
1
+ y
1
]] is the adjoint map of x = pr
1
(x + y).
The next example will be relevant in Section 4.
Example 2.11. Consider the /-module Der(/) /. Denote by pr
1
the projection onto the rst
factor, pr
1
: Der(/) / Der(/), and dene the following bracket:
[[(D
1
, a
1
), (D
2
, a
2
)]] = ([D
1
, D
2
], D
1
(a
2
) D
2
(a
1
))
for (D
1
, a
1
), (D
2
, a
2
) Der(/) /, where [D
1
, D
2
] is the commutator of endomorphisms. Then,
(Der(/) /, [[, ]], pr
1
) is a Lie algebroid.
3 Lie algebroids of Poisson type
Denition 3.1. A Poisson algebra (/, , ) is an associative algebra / together with a Lie bracket
which is also a derivation for the product in /, that is, there is an 1bilinear operation , :
// / such that
1. f, g = g, f (skew-symmetry),
2. f, g, h +g, h, f +h, f, g = 0 (Jacobi identity),
4
3. f, gh = f, gh + gf, h (Leibniz identity),
for all f, g, h /.
If (/, , ) is a Poisson algebra, then we can dene the adjoint map ad : / Der(/) by
ad
f
(g) = f, g,
for all g /. Then, extending the mapping df ad
f
by linearity, we get a morphism :
1
(/)
Der(/) uniquely dened by
(df) = ad
f
, f /. (2)
Sometimes (by analogy with Poisson manifolds), ad
f
is referred to as the Hamiltonian vector eld
corresponding to f /, and denoted by X
f
(we will use this notation and terminology later in
Section 5).
Also, given an
1
(/), we can dene d through the usual formula
d(X, Y ) = X((Y )) Y ((X)) ([X, Y ]),
for all X, Y Der(/).
Theorem 3.2. If (/, , ) is a Poisson algebra, then (
1
(/), [ , ], ) is a Lie algebroid with anchor
map dened by (2) and the Lie bracket
[[, ]] =
()
d
()
d + d((())). (3)
Proof. Let f / and ,
1
(/). Then,
[[, f]] =
()
d(f)
(f)
d + d(f(()))
=
()
(df ) + f
()
d f
()
d + df(()) + fd((()))
= f[[, ]] +
()
(df ) + df(())
= f[[, ]] + df(()) (())df + df(())
= f[[, ]] + ()(f).
The skew-symmetry and the 1-bilinearity of [[, ]] are obvious. To verify the Jacobi identity, let us
consider a system df
i

iI
of generators of
1
(/), so that for arbitrary elements , ,
1
(/),
we have = g
i
df
i
, = h
j
df
j
, = m
k
df
k
for some g
i
, h
j
, m
k
/. Note that for any f, g /, we
have
[[df, dg]] =
(df)
d(dg)
(dg)
d(df) + d(dg((df))) = df, g.
Then,
[[, ]] = [[g
i
df
i
, h
j
df
j
]]
= h
j
g
i
[[df
i
, df
j
]] h
j
(df
j
)(g
i
)df
i
+ g
i
(df
i
)(h
j
)df
j
= g
i
h
j
df
i
, f
j
h
j
f
j
, g
i
df
i
+ g
i
f
i
, h
j
df
j
.
A straightforward computation gives
[[[[, ]], ]] = m
k
h
j
g
i
df
i
, f
j
, f
k
m
k
f
k
, h
j
g
i
df
i
, f
j
+ h
j
g
i
f
i
, f
j
, m
k
df
k
m
k
h
j
f
j
, g
i
df
i
, f
k
+ m
k
f
k
, h
j
f
j
, g
i
df
i
h
j
f
j
, g
i
f
i
, m
k
df
k
+m
k
g
i
f
i
, h
j
df
j
, f
k
m
k
f
k
, g
i
f
i
, h
j
df
j
+ g
i
f
i
, h
j
f
j
, m
k
df
k
,
It follows from here that the cyclic sum [[[[, ]], ]] is zero,
[[[[, ]], ]] = m
k
h
j
g
i
d

(f
i
, f
j
, f
k
)
+ ((m
k
f
k
, h
j
g
i
+ g
i
m
k
f
k
, h
j
+ h
j
m
k
f
k
, g
i
)df
i
, f
j
)
+ ((h
j
g
i
f
i
, f
j
, m
k
+ h
j
f
j
, g
i
f
i
, m
k
h
j
f
j
, g
i
f
i
, m
k

g
i
f
i
, h
j
f
j
, m
k
+ g
i
f
i
, h
j
f
j
, m
k
)df
k
)
= g
i
h
j
((f
i
, f
j
, m
k
+f
j
, f
i
, m
k
f
i
f
j
, m
k
)df
k
) = 0.
5
Example 3.3. Let (M, , ) be a Poisson manifold, i.e., ((

(M), , ) is a Poisson algebra. Using

Theorem 2.9 we have that
1
((

(M)) =
1
(M) is a Lie algebroid with anchor map = P (dened
by (2)) and the bracket (1):
[[, ]] =
()
d
()
d + d((())),
for all ,
1
(M). The proof is based on the fact that there exist a nite subset g
1
, ..., g
k

C

(M) (with k 2dimM +1) such that the C

(M)module
1
(M) is spanned by dg
1
, ..., dg
k
.
This, in turn, is a consequence of Whitneys embedding theorem and the fact that the sheaf of germs
of smooth functions is soft (see [1]).
Theorem 3.2 tells us that given a Poisson algebra we have a Lie algebroid canonically associated
to it. We are interested now in the reciprocal: When does a Lie algebroid (
1
(/), [[, ]], ) determine
a Poisson structure on /?.
Given a Lie algebroid (
1
(/), [[, ]], ), for any f, g / we can dene
f, g = (dg)((df)) = ((df))(g). (4)
The bracket , dened in this way is clearly 1-bilinear and satises the Leibniz rule
fg, h = ((d(fg)))(h) = ((fdg + gdf))(h) = f((dg))(h) + g((df))(h)
= fg, h + gf, h,
for all f, g, h /.
Thus, in order to get a Poisson structure on / we only need to take care of the skew-symmetry and
the Jacobi identity.
Denition 3.4. The anchor map of a Lie algebroid (
1
(/), [[, ]], ) is said to be skew-symmetric
if
(()) = (())
for all ,
1
(/).
If we assume that the anchor map is skew-symmetric, then the new operation dened by (4) is
also skew-symmetric. Now, let us turn our attention to the Jacobi identity.
Theorem 3.5. Let (
1
(/), [[, ]], ) be a Lie algebroid, and dene Q
2
(Der(/)) by Q(df, dg) =
dg((df)) for f, g /. The following conditions are equivalent:
(i) The bracket , dened by f, g = (df)(g) satises the Jacobi identity.
(ii) [Q, Q]
SN
= 0
(iii) [(df), (dg)] = (df, g)
Here, [, ]
SN
denotes the Schouten-Nijenhuis bracket on multiderivations [13].
Proof. It is based on the computation of
1
2
[L
Q
,
Q
] =
1
2

[Q,Q]
SN
, where on the left-hand side we
have the graded commutator of derivations on the graded algebra (/). Due to the /-linearity we
only need to apply this operator to basis elements. The direct application of the denition of the
operators gives
1
2
[L
Q
,
q
](df dg dh) =
1
2
(L
Q

Q
(1)
2

Q
L
Q
)(df dg dh)
=
1
2
(
Q
d
Q
+
Q
d
Q
)(df dg dh)
=
Q
d(
Q
(df dg)dh +
Q
(dg dh)df +
Q
(dh df)dg)
=
Q
(dQ(df, dg) dh + dQ(dg, dh) df + dQ(dh, df) dg)
= Q(dQ(df, dg), dh) + Q(dQ(dg, dh), df) + Q(dQ(dh, df), dg).
6
The equivalence of the conditions (i) and (ii) follows from the following identities
Q(dQ(df, dg), dh) + Q(dQ(dg, dh), df) + Q(dQ(dh, df), dg)
= Q(df, dg), h +Q(dg, dh), f +Q(dh, df), g
= f, g, h +g, h, f +h, f, g.
On the other hand, this term can be computed as follows
Q(dh, dQ(df, dg)) + Q(df, dQ(dh, dg)) Q(dQ(df, dh), dg)
= (dh)(Q(df, dg)) + (df)(Q(dh, dg)) (dQ(df, dh)(g)
= (dh)((df)(g)) + (df)((dh)(g)) (df, h)(g)
= [(df), (dh)](g) (df, h)(g)
= ([(df), (dh)] (df, h))(g).
But, from
1
2

[Q,Q]
SN
( , df, dh)(dg) =
1
2
[Q, Q]
SN
(df, dg, dh), we have the equality

1
2

[Q,Q]
SN
( , df, dh)

(g) = ([(df), (dh)] (df, h))(g),

which proves the equivalence between items (ii) and (iii).
This result motivate the following denition.
Denition 3.6. A Lie algebroid (
1
(/), [[, ]], ) is of Poisson type if:
(1) The anchor is skew-symmetric,
(2) One of the equivalent conditions (ii)-(iii) in Theorem 3.5 holds.
In other words, a Lie algebroid (
1
(/), [[, ]], ) is of Poisson type if it determines a Poisson structure
on the algebra /.
Another important issue for us is to determine the form of the bracket of a Lie algebroid of Poisson
type. We know that the classical example of Poisson manifolds leads to brackets of the type (1).
The following result characterizes the class of such algebroids.
Proposition 3.7. If (
1
(/), [[, ]], ) is a Lie algebroid for which is skew-symmetric and [[df, dg]] =
df, g, then the bracket of the Lie algebroid is of the form
[[, ]] =
()
d
()
d + d((())).
Proof. Note rst that, if = df and = dg for some f, g, /, then

()
d
()
d + d((())) = d(dg((df))) = df, g = [[df, dg]] = [[, ]].
Since every element of
1
(/) is a linear combination of elements of the form df
i
(f
i
/), it is
enough to prove the statement for = f
1
dg
1
and = f
2
dg
2
:

()
d
()
d + d((()))
=
(f1dg1)
df
2
dg
2

(f2dg2)
df
1
dg
1
+ f
2
d(dg
2
(f
1
(dg
1
))) + df
2
dg
2
(f
1
(dg
1
))
=df
2
((f
1
dg
1
))dg
2
dg
2
((f
1
dg
1
))df
2
df
1
((f
2
dg
2
))dg
1
+ dg
1
((f
2
dg
2
))df
1
+ f
2
d(f
1
dg
2
((dg
1
))) + f
1
df
2
dg
2
((dg
1
))
=f
1
df
2
((dg
1
))dg
2
f
1
dg
2
((dg
1
))df
2
f
2
df
1
((dg
2
))dg
1
+ f
2
dg
1
((dg
2
))df
1
+ f
1
f
2
d(dg
2
((dg
1
))) + f
v
dg
2
((dg
1
))df
1
+ f
1
dg
2
((dg
1
))df
2
=f
1
df
2
((dg
1
))dg
2
f
2
df
1
((dg
2
))dg
1
+ f
2
dg
1
((dg
2
))df
1
+ f
1
f
2
dg
1
, g
2
f
2
dg
1
((dg
2
))df
1
=f
1
df
2
((dg
1
))dg
2
f
2
df
1
((dg
2
))dg
1
+ f
1
f
2
[[dg
1
, dg
2
]]
=[[f
1
dg
1
, f
2
dg
2
]]
=[[, ]]
7
The hypothesis of this Proposition can be reformulated in an alternative way.
Proposition 3.8. If (
1
(/), [[, ]], ) is a Lie algebroid with anchor skew-symmetric, then the
following assertions are equivalent:
(i) [[df, dg]] = d(dg((df))) = df, g for all f, g /,
(ii) d = 0 = d implies d[[, ]] = 0.
Proof. First, let us assume that item (i) holds. By Proposition 3.7 we know that the bracket is of
the form
[[, ]] =
()
d
()
d + d((())),
and hence condition (ii) holds,
d[[, ]] = d(
()
d
()
d + d((()))) = d
2
((())) = 0,
whenever d = 0 = d.
Now let us assume that condition (ii) holds. Dene C(, ) = [[, ]] d((())) for ,
1
(/).
Notice that C is skew-symmetric and
dC(, ) = d([[, ]]) d
2
((()) = 0,
for any closed and .
Let us evaluate the following expression
C(df, hdh) = [[df, hdh]] d(hdh((df)))
= h[[df, dh]] + (df)(dh)dh dh(dh((df))) hd(hdh((df)))
= h[[df, dh]] hd(hdh((df)))
= hC(df, dh).
Now, applying the operator d and taking into account that hdh is closed, we get
0 = dC(df, hdh) = d(hC(df, dh)) = dh C(df, dh) + hdC(df, dh) = dh C(df, dh).
Interchanging the roles of f and h, we have
0 = dC(dh, fdf) = df C(dh, df) = df C(df, dh).
These relations imply that df, dh and C(df, dh) are linearly dependent for arbitrary f and h. In
particular, if df and dh are linearly independent, then C(df, dh) = 0, and hence
[[df, dh]] = d(dh((df))) = df, h.
Let us summarize these results.
Theorem 3.9. Let (
1
(/), [[, ]], ) be a Lie algebroid. Then, there is a Poisson algebra structure
, : // / on / such that
(df) = ad
f
, f /,
if and only if:
(a) is skew-symmetric
(b) One of the following conditions holds:
(1) d = 0 = d implies d[[, ]] = 0 ;
8
(2) [[df, dg]] = d(dg((df))) for all f, g /.
Under these conditions, the Lie bracket [[, ]] is reconstructed from by the formula
[[, ]] =
()
d
()
d + d((())).
Of course, the basic example of this situation is the cotangent Lie algebroid of a symplectic manifold.
A Lie algebroid structure (
1
(M), [[, ]], ) induces a Poisson bracket on (

(M) if and only if is

skew-symmetric and, whenever d = 0 = d, then d[[, ]] = 0. For this kind of examples, the
property [[df, dg]] = df, g follows directly from the injectivity of the anchor map and the fact that
it is a Lie algebra morphism: [[df, dg]] = [[(df), (dg)]] = (df, g). Let us consider the following
example, where the anchor map is not injective but the property [[df, dg]] = df, g still holds.
Example 3.10. Let 1 = R and / = R[x
1
, x
2
, x
3
]. Then, Der(/) = Span
1
,
2
,
2
and
1
(/) =
Spandx
1
, dx
2
, dx
3
. Dene the following bracket
[[p
i
dx
i
, q
j
dx
j
]] =((p
1
(
2
+
3
) + p
2
(
1
+
3
) + p
3
(
1

2
)) (q
i
)
(q
1
(
2
+
3
) + q
2
(
1
+
3
) + q
3
(
1

2
)) (p
i
)) dx
i
,
and the anchor map as
:
1
(/) Der(/)
p
i
dx
i
(p
2
+ p
3
)
1
+ (p
1
p
3
)
2
+ (p
1
+ p
2
)
3
.
Note that the matrix representation of relative to the given basis in Der(/) and
1
(/) is
=

0 1 1
-1 0 1
-1 -1 0

.
Therefore, is skew-symmetric and of rank 2 ( is not injective). A long but straightforward
computation shows that (
1
(/), ][, ]], ) is a Lie algebroid.
Let us show that this Lie algebroid is of Poisson type by checking the property [[df, dg]] = df, g.
We have for p, q /:
p, q = d((dp)(q)) = d

i
pdx
i

(q)

= d((
2
p +
3
p)
1
q + (
1
p
3
p)
2
q + (
1
p +
2
p)
3
q) .
The dx
1
factor in the expansion of this expression (the other cases are similar) is:
(
2
p +
3
p)
2
11
q + (
1
p
3
p)
2
12
q + (
1
p +
2
p)
2
13
q
(
2
12
p +
2
13
p)
1
q + (
2
11
p
2
13
p)
2
q + (
2
11
p +
2
12
p)
3
q.
On the other hand, the Lie algebroid bracket is
[[dp, dq]] = [[
i
pdx
i
,
j
qdx
j
]]
= (dp)(
k
q) (dq)(
k
p))dx
k
.
For k = 1, we compute the coecient of dx
1
:
(
2
p +
3
p)
2
11
q + (
1
p
3
p)
2
12
q + (
1
p +
2
p)
2
13
q
+ (
2
q +
3
q)
2
11
p (
1
q
3
q)
2
12
p (
1
q +
2
q)
2
13
p,
which is the same as above. Thus, (
1
(/), [[, ]], ) is a Lie algebroid of Poisson type and its bracket
is just given by formula (3).
9
Remark 3.11. The anchor map of the cotangent Lie algebroid is not injective in general. It is only
true in the symplectic case.
Finally, let us consider an example of Lie algebroid of Poisson type whose bracket does not have
the form (3).
Example 3.12. Let 1 = R and / = R[x
1
, x
2
, x
3
]. Dene the structure of an Abelian Lie algebra
with generators dx
1
, dx
2
, dx
3
,
[[dx
i
, dx
j
]] = 0, i, j 1, 2, 3.
Extending by 1-bilinearity and the Leibniz identity gives
[[p
i
dx
i
, q
j
dx
j
]] = p
i
(dx
i
)(p
j
)dx
j
q
j
(dx
j
)(p
i
)dx
i
.
Next, let us think of the following skew-symmetric morphism of /-modules as the anchor map:
:
1
(/) Der(/),
dx
1
x
3

2
,
dx
2
x
3

1
,
dx
3
0.
With these denitions the bracket in
1
(/) can also be expressed in a form suitable for explicit
computations, as
[[p
i
dx
i
, p
j
dx
j
]] = x
3
((p )
3
q
i
(q )
3
p
i
)dx
i
,
where an element p / is viewed as a vector p = (p
1
, p
2
, p
3
), = (
1
,
2
,
3
) and the subindex 3
denotes the third component of the cross product p .
It can be checked that (
1
(/), [[, ]], ) is a Lie algebroid of Poisson type. However, if p = 2x
1
+ x
2
and q = x
1
+ x
2
, then
[[dp, dq]] = [[2dx
1
+ dx
2
, dx
1
+ dx
2
]] = 0,
whereas
dp, q = d(dq((dp))) = d((dx
1
+ dx
2
)(2dx
1
+ dx
2
))
= d((dx
1
+ dx
2
)(2x
3

2
x
3

1
)) = dx
3
.
4 Transitive Lie algebroids
To motivate the denition of transitive Lie algebroids that we will give, let us consider for a moment
the geometric example of a Lie algebroid (E, [[, ]], q), where E M is a vector bundle over a
manifold M (recall Example 2.9). If the anchor map q : (E) (TM) is an epimorphism, the
algebroid (E, [[, ]], q) is said to be transitive. In this case, it is possible to construct the so-called
Atiyah sequence of the algebroid, which is the short exact sequence
g
E TM


//
j
// //
q
where g = Kerq. Thus, the existence of a section for the anchor q (equivalently, a linear connection
on E) implies that, locally, E = TM g. Note also that the bre of the bundle g over the point
x M, g
x
is a Lie algebra (called the isotopy Lie algebra of the algebroid E at x M) with the
bracket given, for , g
x
, by
[, ] = [[X, Y ]],
where X, Y (E) are any sections such that X(x) = and Y (x) = .
Denition 4.1. A Lie algebroid (T, [[, ]], ) is transitive if there exist a short exact sequence of
/-modules:
V T Der(/)


//
j
// //

Remark 4.2. There is a more general notion, the extension of a Lie-Rinehart algebra, that gener-
alizes the transitivity condition for a geometric Lie algebroid (see [11]).
10
4.1 Transitive algebroids induced by connections
Let V be a unitary A-module and a connection on V .
Denition 4.3. The curvature of is the mapping C

: Der(/) Der(/) Hom(V, V ) dened

by
C

(X, Y ) = [
X
,
Y
]
[X,Y ]
.
Denition 4.4. If V is endowed with a Lie algebra structure [, ]
V
, then a connection is said to
be a Lie connection if

X
[v
1
, v
2
]
V
= [
X
v
1
, v
2
]
V
+ [v
1
,
X
v
2
]
V
,
for all v
1
, v
2
V and for all X Der(/).
This subsection is devoted to the proof of the following result (see, also [17], [18]).
Theorem 4.5. Let V be an /-module endowed with a Lie algebra structure [ , ]
V
which is /-linear
(i.e., [fX, Y ]
V
= f[X, Y ]
V
f /, X, Y V ). Let be a Lie connection on V . If there exists
a 2form B
2
(/; V ) with values in V , such that, for any X
1
, X
2
, X
3
Der(/) and v V the
following conditions hold:
(a) [B(X
1
, X
2
), v]
V
= C

(X
1
, X
2
)(v),
(b) The cyclic sum (
X1
(B(X
2
, X
3
)) B([X
1
, X
2
], X
3
)) = 0,
then (Der(/) V, [[, ]], ) is a transitive Lie algebroid with anchor map = pr
1
and bracket
[[(X
1
, v
1
), (X
2
, v
2
)]] = ([X
1
, X
2
], [v
1
, v
2
]
V
+
X1
v
2

X2
v
1
B(X
1
, X
2
)). (5)
Moreover,

2
(B(X, Y )) =
1
([X, Y ]) [[
1
(X),
1
(Y )]], (6)
for X, Y Der(/). Here
1
: Der(/) Der(/) V and
2
: V Der(/) V are the inclusion
maps.
Proof. It is clear that Der(/) V is an /-module. We must show that the bracket dened by (5) is
Lie. The skew-symmetry and the 1-bilinearity are immediate. To check the Jacobi identity, let us
pick X
1
, X
2
, X
3
Der(/) and v
1
, v
2
, v
3
V . Then, we have:
11
([[[[(X
1
, v
1
), (X
2
, v
2
)]], (X
3
, v
3
)]])
= ([[X
1
, X
2
], X
3
], [[v
1
, v
2
]
V
, v
3
]
V
+ [
X1
v
2
, v
3
]
V
[
X2
v
1
, v
3
]
V
[B(X
1
, X
2
), v
3
]
V
+
[X1,X2]
v
3

X3
([v
1
, v
2
]
V
+
X1
v
2

X2
v
1
B(X
1
, X
2
)) B([X
1
, X
2
], X
3
))
+([[X
2
, X
3
], X
1
], [[v
2
, v
3
]
V
, v
1
]
V
+ [
X2
v
3
, v
1
]
V
[
X3
v
2
, v
1
]
V
[B(X
2
, X
3
), v
1
]
V
+
[X2,X3]
v
1

X1
([v
2
, v
3
]
V
+
X2
v
3

X3
v
2
B(X
2
, X
3
)) B([X
2
, X
3
], X
1
))
+([[X
3
, X
1
], X
2
], [[v
3
, v
1
]
V
, v
2
]
V
+ [
X3
v
1
, v
2
]
V
[
X1
v
3
, v
2
]
V
[B(X
3
, X
1
), v
2
]
V
+
[X3,X1]
v
2

X2
([v
3
, v
1
]
V
+
X3
v
1

X1
v
3
B(X
3
, X
1
)) B([X
3
, X
1
], X
2
))
= ([[X
1
, X
2
], X
3
] + [[X
2
, X
3
], X
1
] + [[X
3
, X
1
], X
2
],
[[v
1
, v
2
]
V
, v
3
]
V
+ [[v
2
, v
3
]
V
, v
1
]
V
+ [[v
3
, v
1
]
V
, v
2
]
V
+[
X1
v
2
, v
3
]
V
+ [v
2
,
X1
v
3
]
V

X1
([v
2
, v
3
]
V
)
+[
X2
v
3
, v
1
]
V
+ [v
3
,
X1
v
2
]
V

X2
([v
3
, v
1
]
V
)
+[
X3
v
1
, v
2
]
V
+ [v
1
,
X2
v
3
]
V

X3
([v
1
, v
2
]
V
)
[B(X
1
, X
2
), v
3
]
V
+ C

(X
1
, X
2
)(v
3
)
[B(X
2
, X
3
), v
1
]
V
+ C

(X
2
, X
3
)(v
1
)
[B(X
3
, X
1
), v
2
]
V
+ C

(X
3
, X
1
)(v
2
)
+
X1
B(X
2
, X
3
) B([X
1
, X
2
], X
3
) +
X2
B(X
3
, X
1
) B([X
2
, X
3
], X
1
)
+
X3
B(X
1
, X
2
) B([X
3
, X
1
], X
2
))
= (0, 0)
Now let us show that satises Leibniz. We have
[[(X
1
, v
1
), f(X
2
, v
2
)]]
= ([X
1
, fX
2
], [v
1
, fv
2
]
V
+
X1
(fv
2
)
fX2
(v
1
) B(X
1
, fX
2
))
= (X
1
(f)X
2
+ fX
1
X
2
fX
2
X
1
, f[v
1
, v
2
]
V
+
X1
(fv
2
)
fX2
(v
1
) B(X
1
, fX
2
))
= (f[X
1
, X
2
] + X
1
(f)X
2
, f[v
1
, v
2
]
V
+
X1
(fv
2
) f
X2
(v
1
) fB(X
1
, X
2
)).
On the other hand,
f[[(X
1
, v
1
), (X
2
, v
2
)]] + (X
1
, v
1
)(f)(X
2
, v
2
)
= f[(X
1
, v
1
), (X
2
, v
2
)] + X
1
(f).(X
2
, v
2
)
= (f[X
1
, X
2
] + X
1
(f)X
2
, f[v
1
, v
2
]
V
+ f
X1
v
2
f
X2
v
1
fB(X
1
, X
2
) + X
1
(f)v
2
)
= (f[X
1
, X
2
] + X
1
(f)X
2
, f[v
1
, v
2
]
V
f
X2
v
1
fB(X
1
, X
2
) +
X1
(fv
2
)).
Thus, [[(X
1
, v
1
), f(X
2
, v
2
)]] = f[[(X
1
, v
1
), (X
2
, v
2
)]] + (X
1
, v
1
)(f)(X
2
, v
2
).
Since = pr
1
is clearly /-linear, we have that Der(/) V is a Lie algebroid. The transitivity is
obvious in view of the sequence
V Der(/) V Der(/).


//
2
// //
1
Moreover, a direct computation shows that

1
[X, Y ] [[
1
(X),
1
(Y )]] = ([X, Y ], 0) [[(X, 0), (Y, 0)]]
= ([X, Y ], 0) ([X, Y ], B(X, Y ))
= (0, B(X, Y )) =
2
(B(X, Y )).
12
Remark 4.6. The conditions (a) and (b) in this theorem have the following interpretation. Condi-
tion (a) states that the curvature of the connection is given by the composition C

= ad
R
B,
where ad
R
: V V is the right adjoint with respect to the bracket on V . On the other hand, (b)
expresses a modied Bianchi identity (see also [17], [18]).
4.2 Parametrization of transitive algebroids on Der/V
The following result says that the converse of Theorem 4.5 is also true.
Theorem 4.7. Let (Der(/) V, [[, ]], ) be a transitive Lie algebroid with = pr
1
. Then:
(i) The bracket on V , dened for v
1
, v
2
V by
[v
1
, v
2
]
V
= pr
2
([[
2
(v
1
),
2
(v
2
)]]), (7)
is an /-linear Lie bracket.
(ii) The mapping : Der(/) Hom(V, V ) given by

X
(v) = pr
2
([[
1
(X),
2
(v)]]),
for v V and X Der(/), is a Lie connection on V .
(iii) The mapping B : Der(/) Der(/) V dened by
B(X, Y ) = pr
2
([[
1
(X),
1
(Y )]]
1
([X, Y ])),
is /-bilinear, skew-symmetric and it satises conditions (a) and (b) of Theorem 4.5.
Proof. First let us show that [ , ]
V
dened by (7) is a Lie bracket and /-linear. The skew-symmetry
and 1bilinearity are inherited from [[, ]] and
2
. Let v
1
, v
2
, v
3
V , and f /. For the Jacobi
identity we have:
[[v
1
, v
2
]
V
, v
3
]
V
= (pr
2
[[
2
(pr
2
[[
2
(v
1
),
2
(v
2
)]]),
2
(v
3
)]])
= (pr
2
[[[[
2
(v
1
),
2
(v
2
)]],
2
(v
3
)]]) = 0.
And for the /-linearity:
[fv
1
, v
2
]
V
= pr
2
[[
2
(fv
1
),
2
(v
2
)]]
= pr
2
[[f
2
(v
1
),
2
(v
2
)]]
= pr
2
(f[[
2
(v
1
),
2
(v
2
)]] (
2
(v
2
))(f)
2
(v
1
))
= fpr
2
([[
2
(v
1
),
2
(v
2
)]]
= f[v
1
, v
2
]
V
.
Now, let us show that is a Lie connection on V . Let X, Y, Z Der(/), f /, v, v
1
, v
2
V and
s R. First, we observe that is /-linear:

fX+Y
(v) = pr
2
([[
1
(fX + Y ),
2
(v)]])
= pr
2
([f
1
(X),
2
(v)]] + [[
1
(Y ),
2
(v)]])
= pr
2
(f[[
1
(X),
2
(v)]] (
2
(v))(f)(
1
(X)) + [[
1
(Y ),
1
(v)]])
= fpr
2
([[
1
(X),
2
(v)]] + [[
1
(Y ),
2
(v)]])
= f
X
(v) +
Y
(v).
Also,
X
is 1linear,

X
(av
1
+ v
2
) = pr
2
([[
1
(X),
2
(av
1
+ v
2
)]])
= pr
2
([[
1
(X), a
2
(v
1
) +
2
(v
2
)]]) = a
X
(v
1
) +
X
(v
2
).
13
The Leibniz rule can be veried as follows

X
(fv) = pr
2
([[
1
(X),
2
(fv)]])
= pr
2
([[
1
(X), f
2
(v)]])
= pr
2
(f[[
1
(X),
2
(v)]] + (
1
(X))(f)(
2
(v))
= pr
2
(f[[
1
(X),
2
(v)]] + X(f)(
2
(v))
= fpr
2
([[
1
(X),
2
(v)]]) + X(f)pr
2
(
2
(v))
= f
X
(v) + X(f)v.
Also, has the Lie property,

2
([
X
v
1
, v
2
]
V
+ [v
1
,
X
v
2
]
V
) =
2
(pr
2
([[
2
(
X
v
1
),
2
(v
2
)]])
+pr
2
([[
2
(v
1
),
2
(
X
v
2
)]]))
=
2
(pr
2
([[
2
(pr
2
([[
1
(X),
2
(v)]])),
2
(v
2
)]])
+pr
2
([[
2
(v
1
),
2
(pr
2
([[
1
(X),
2
(v
2
)]]))]]))
= [[[
1
(X),
2
(v
1
)]],
2
(v
2
)]]
+[[
2
(v
1
), [[
1
(X),
2
(v
2
)]]]]
= [[
1
(X), [[
2
(v
1
),
2
(v
2
)]]]]
= [[
1
(X),
2
([v
1
, v
2
]
V
)]]
=
2
(pr
2
([[
1
(X),
2
([v
1
, v
2
]
V
)]]))
=
2
(
X
([v
1
, v
2
]
V
)),
which implies that
X
[v
1
, v
2
]
V
= [
X
v
1
, v
2
]
V
+ [v
1
,
X
v
2
]
V
.
Let us now check the properties of B: Let X
1
, X
2
, X
3
Der(/) and f /. Then B is skew-
symmetric,
B(X
1
, X
2
) = pr
2
([[
1
(X
1
),
1
(X
2
)]]
1
([X
1
, X
2
]))
= (pr
2
([[
1
(X
2
),
1
(X
1
)]]
1
([X
2
, X
1
])))
= R(X
2
, X
1
),
and /-linear,
B(fX
1
+ X
2
, X
3
) = pr
2
([[
1
(fX
1
+ X
2
),
1
(X
3
)]]
1
([fX
1
+ X
2
, X
3
]))
= pr
2
([f
1
(X
1
),
1
(X
3
)]] + [[
1
(X
2
),
1
(X
3
)]]

1
([fX
1
, X
3
]) + [X
2
, X
3
]))
= pr
2
(f[[
1
(X
1
),
1
(X
3
)]] (
1
(X
3
))(f)
1
(X
1
)
+[[
1
(X
2
),
1
(X
3
)]]
1
(f[X
1
, X
3
]
X
3
(f)X
1
+ [X
2
, X
3
]))
= pr
2
(f[[
1
(X
1
),
1
(X
3
)]] + [[
1
(X
2
),
1
(X
3
)]]

1
(f[X
1
, X
3
] X
3
(f)X
1
+ [X
2
, X
3
]))
= fB(X
1
, X
3
) +B(X
2
, X
3
).
Finally, it remains to see that conditions (a) and (b) of Theorem 4.5 are satised. For (a), we
have:

2
(C

(X, Y )(v)) =
2
([
x
,
Y
](v)
[X,Y ]
(v))
=
2
(
X
(
Y
)(v))
2
(
Y
(
X
)(v))
[[
1
([X, Y ]),
2
(v)]]
= [[
1
(X),
2
(
Y
(v))]] [[
1
(Y ),
2
(
X
(v))]]
[[
1
([X, Y ]),
2
(v)]]
= [[
1
(X), [[
1
(Y ),
2
(v)]]]]
+[[
1
(Y ), [
2
(v),
1
(X)]]]]
[[
1
([X, Y ]),
2
(v)]]
= [
2
(v), [[
1
(X),
1
(Y )]]]]
[[
1
([X, Y ]),
2
(v)]]
= [[([[
1
(X),
1
(Y )]]
1
([X, Y ]),
2
(v)]]
= [[
2
(pr
2
([[
1
(X),
1
(Y )]]
1
([X, Y ])),
2
(v)]]
=
2
([pr
2
([[
1
(X),
1
(Y )]]
1
([X, Y ]), v]
V
)
=
2
([(B(X, Y )), v]
V
).
14
Hence, [(B(X, Y )), v]
V
= C

(X, Y )(v). Next,

2
(
X
(B(Y, Z)) B([X, Y ], Z)) = [[
1
(X),
2
(B(Y, Z))]]
2
(B([X, Y ], Z))
= [[
1
(X),
1
[Y, Z]]]
[[
1
(X), [[
1
(Y ),
1
(Z)]]]]

1
[[X, Y ], Z] + [[
1
[X, Y ],
1
(Z)]]
= [[
1
(X),
1
[Y, Z]]] + [[
1
[X, Y ],
1
(Z)]]
= 0.
Thus,
X
(B(Y, Z)) B([X, Y ], Z) = 0.
As a consequence of Theorems 4.5 and 4.7, we have the following.
Corollary 4.8. Let V be an /-module. There is a one to one correspondence between transitive Lie
algebroids (Der(/) V, [ , ]], ) with anchor = pr
1
and triples ([, ]
V
, , B) consisting of:
(i) A /-linear Lie bracket [, ]
V
in V ;
(ii) A Lie connection on V ;
(iii) A 2form B
2
(/; V ) satisfying conditions (a) and (b) of Theorem 4.5.
4.3 Algebroid connections
In this subsection, we rst generalize Theorem 4.7 to the case of a transitive Lie algebroid which
is not necessarily of the form Der(/) V , but we require that it be endowed with a Lie algebroid
connection (see [16, 18]), considered as a section of the anchor map. Then, in Theorem 4.11 we
construct a Lie algebroid structure on Der(/) V which is isomorphic to the given algebroid.
Denition 4.9. Let (T, [[, ]], ) be a Lie algebroid. A morphism of /-modules : Der(/) T is
called a Lie algebroid connection (on T) if it is a section of , that is, = Id
Der(A)
.
Theorem 4.10. Let
V T Der(/)


//

// //

(8)
be a transitive Lie algebroid and : Der(/) T a Lie algebroid connection for T. Then:
(i) The bracket [, ]
V
dened for v
1
, v
2
V by
[v
1
, v
2
]
V
= [[(v
1
), (v
2
)]], (9)
is an /-linear Lie bracket.
(ii) The mapping : Der(/) Hom(V, V ) given by

X
(v) = [[(X), (v)]],
is a Lie connection on V (called the adjoint connection).
(iii) The mapping B : Der(/) Der(/) V , dened by
B(X, Y ) = [[(X), (Y )]] ([X, Y ]),
is /-bilinear, skew-symmetric and satises the conditions (a) and (b) of Theorem 4.5.
15
Proof. First, note that the exactness of (8) allows us to identify V with its image under in T, and
that Im = ker in T. Then, since is a morphism of Lie algebras, we have
[[(v
1
), (v
2
)]] = [(v
1
), (v
2
)] = 0.
Thus, the bracket [, ]
V
is well-dened. By a similar argument it can be proved that and B are well-
dened. Now, let us show that [ , ]
V
dened by (9) is a Lie bracket and /-linear. The 1bilinearity
and skew-symmetry are again inherited from [[, ]]. For the rest of properties, let v
1
, v
2
, v
3
V and
f /. The Jacobi identity results from
[[v
1
, v
2
]
V
, v
3
]
V
= [[([[(v
1
), (v
2
)]]), (v
3
)]]
= [[[[(v
1
), (v
2
)]], (v
3
)]] = 0.
The /-linearity can be checked as follows:
[fv
1
, v
2
]
V
= [[(fv
1
), (v
2
)]]
= [[f(v
1
), (v
2
)]]
= f[[(v
1
), (v
2
)]] ((v
2
))(f)(v
1
))
= f[[(v
1
), (v
2
)]]
= f[v
1
, v
2
]
V
.
Next, we show that is Lie connection in V . The 1linearity is inherited from and [[, ]]. If
X, Y, Z Der(/), f /, and v, v
1
, v
2
V, s R, the /-linearity of follows from the computation

fX+Y
(v) = [[(fX + Y ), (v)]]
= [[f(X), (v)]] + [(Y ), (v)]]
= f[[(X), (v)]] ((v))(f)((X)) + [(Y ), (v)]]
= f[[(X), (v)]] + [(Y ), (v)]]
= f
X
(v) +
Y
(v).
For a xed X Der(/),
X
satises the Leibniz identity,

X
(fv) = [[(X), (fv)]]
= [[(X), f(v)]]
= f[[(X), (v)]] + ((X))(f)((v))
= f[[(X), (v)]] + X(f)((v))
= f[[(X), (v)]] + X(f)k((v))
= f
X
(v) + X(f)v,
and is a Lie connection,
([
X
v
1
, v
2
]
V
+ [v
1
,
X
v
2
]
V
) = ([[[(X), (v
1
)]], v
2
]
V
+ [v
1
, [[(X), (v
2
)]]]
V
)
= [[[[(X), (v
1
)]], (v
2
)]] + [[(v
1
), [[(X), (v
2
)]]]]
= [[(X), [[(v
1
), (v
2
))]]]]
= [[(X), ([v
1
, v
2
)]
V
)]]
= (
X
[v
1
, v
2
]).
This implies that
X
[v
1
, v
2
]
V
= [
X
v
1
, v
2
]
V
+ [v
1
,
X
v
2
]
V
.
By similar computations it can be shown that B
2
(/; V ). The proof of the property which relates
B to the curvature of also follows the same guidelines,
(C

(X, Y )(v)) = ([
x
,
Y
](v)
[X,Y ]
(v))
= (
X
(
Y
)(v)) (
Y
(
X
)(v)) [[([X, Y ]), (v)]]
= [[(X), (
Y
(v))]] [[(Y ), (
X
(v))]] [[([X, Y ]), (v)]]
= [[(X), [[(Y ), (v)]]]] + [[(Y ), [[(v), (X)]]]] [[([X, Y ]), (v)]]
= [[(v), [[(X), (Y )]]]] [[([X, Y ]), (v)]]
= [[([[(X), (Y )]] ([X, Y ]), (v)]]
= [[([[(X), (Y )]] ([X, Y ])), (v)]]
= ([[[(X), (Y )]] ([X, Y ]), v]
V
)
= ([B(X, Y ), v]
V
).
16
Therefore, [B(X, Y ), v]
V
= C

(X, Y )(v). Finally, we check the modied Bianchi identity,

(
X
(B(Y, Z)) B([X, Y ], Z)) = [[(X), (B(Y, Z))]] (B([X, Y ], Z))
= [[(X), ([Y, Z])]]
[[(X), [[(Y ), (Z)]]]]
([[X, Y ], Z]) + [[([X, Y ]), (Z)]]
= [[(X), ([Y, Z])]] + [[([X, Y ]), (Z)]] = 0.
Thus,
X
(B(Y, Z)) B([X, Y ], Z) = 0.
Theorem 4.11. Let
V T Der(/)


//

// //

be a transitive Lie algebroid and : Der(/) T a Lie algebroid connection on T. Then (T, [[, ]], )
is isomorphic to (Der(/) V, , ), pr
1
), where the bracket is dened through
[(X
1
, v
1
), (X
2
, v
2
)] = ([X
1
, X
2
], [v
1
, v
2
]
V
+
X1
(v
2
)
X2
(v
1
) B(X
1
, X
2
)),
and [, ]
V
,
X
, B are given in Theorem 4.10.
Proof. We know that (Der(/) V, , ), pr
1
) is a Lie algebroid by Theorem 4.5. Moreover,
: Der(/) V T
(X, v) (X) + (v)
is an /-module isomorphism such that
( )(X, v) = ((X) + (v)) = X = pr
1
(X, v)
and
((X
1
, v
1
), (X
2
, v
2
))) = ([X
1
, X
2
], [v
1
, v
2
]
V
+
X1
(v
2
)
X2
(v
1
) B(X
1
, X
2
))
= ([X
1
, X
2
], [[(v
1
), (v
2
)]] + [[(X
1
), (v
2
)]]
[[(X
2
), (v
1
)]] [[(X
1
), (X
2
)]] ([X
1
, X
2
]))
= ([X
1
, X
2
]) + ([[(v
1
), (v
2
)]]) + ([[(X
1
), (v
2
)]])
([[(X
2
), (v
1
)]]) ([[(X
1
), (X
2
)]]) ([X
1
, X
2
])
= [[(v
1
), (v
2
)]] + [[(X
1
), (v
2
)]] [[(X
2
), (v
1
)]] + [[(X
1
), (X
2
)]]
= [[(X
1
) + (v
1
), (X
1
) + (v
1
)]]
= [[(X
1
, v
1
), (X
2
, v
2
)]].
5 Poisson algebras on /V
Recall that, if (/, , ) is a Poisson algebra, then we dene for every f / the Hamiltonian
derivation X
f
Der(/) as the adjoint map X
f
= f, . The following standard properties will be
used below:
(i) X
f1f2
= f
1
X
f2
+ f
2
X
f1
,
(ii) The mapping f X
f
is a Lie algebra morphism, i.e, X
{f1,f2}
= [X
f1
, X
f2
].
Now, on /V we can dene a product given by
(f
1
, v
1
) (f
2
, v
2
) = (f
1
f
2
, f
1
v
2
+ f
2
v
1
), (10)
for f
1
, f
2
/ and v
1
, v
2
V . This makes /V a commutative ring. Under certain conditions, this
ring is also a Poisson algebra, as shown by the following result.
17
Theorem 5.1. Let (/, ,
A
) be a Poisson algebra. Suppose we have a transitive Lie algebroid
V Der(/) V Der(/),


//
2
// //

with anchor = pr
1
. Then, /V is also a Poisson algebra with product dened by (10) and Poisson
bracket
(f
1
, v
1
), (f
2
, v
2
) = (f
1
, f
2

A
,
X
f
1
(v
2
)
X
f
2
(v
1
) + [v
1
, v
2
]
V
+B(X
f1
, X
f2
)), (11)
where [, ]
V
, and B are given in Theorem 4.7.
Proof. Let f
1
, f
2
, f
3
/ and v
1
, v
2
, v
3
V . The skew-symmetry of the bracket is immediate
(f
1
, v
1
), (f
2
, v
2
) = (f
1
, f
2

A
,
X
f
1
(v
2
)
X
f
2
(v
1
) + [v
1
, v
2
]
V
+B(X
f1
, X
f2
))
= (f
2
, f
1

A
,
X
f
1
(v
2
)
X
f
2
(v
1
) [v
2
, v
1
]
V
B(X
f2
, X
f1
))
= (f
2
, v
2
), (f
1
, v
1
).
Next we verify the Leibniz identity, (where [[, ]] denotes the Lie bracket on Der(/) V )
(f
1
, v
1
), (f
2
, v
2
) (f
3
, v
3
)
= (f
1
, v
1
), (f
2
f
3
, f
2
v
3
+ f
3
v
2
)
= (f
1
, f
2
f
3

A
,
X
f
1
(f
3
v
2
+ f
2
v
3
)
X
f
2
f
3
(v
1
) + [v
1
, f
3
v
2
+ f
2
v
3
]
V
+B(X
f1
, X
f2f3
))
= (f
2
f
1
, f
3

A
+ f
1
f
2
, f
3

A
, X
f1
(f
3
)v
2
+f
3

X
f
1
(v
2
) + X
f1
(f
2
)v
3
+ f
2

X
f
1
(v
3
)

X
f
2
f
3
(v
1
) + f
2
[v
1
, v
3
]
V
+ f
3
[v
,
v
2
]
V
+pr
2
([[
1
(X
f1
),
1
(X
f2f3
)]]
1
([X
f1
, X
f2f3
]))
= (f
2
f
1
, f
3

A
+ f
1
f
2
, f
3

A
, f
2
pr
2
([[
1
(X
f1
),
2
(v
3
)]])
+f
2
pr
2
([[
2
(v
1
),
1
(X
f3
)]]) + f
2
[v
1
, v
3
]
V
+ f
2
pr
2
([[
1
(X
f1
),
1
(X
f3
)]])
f
2

1
([X
f1
, X
f3
]) +f
1
, f
3
v
2
+f
3
pr
2
([[
1
(X
f1
),
2
(v
2
)]]) + f
3
pr
2
([[
2
(v
1
),
1
(X
f2
)]]) + f
3
[v
1
, v
2
]
V
+f
3
pr
2
([[
1
(X
f1
),
1
(X
f2
)]]) f
3
pr
2
(
1
([X
f1
, X
f2
]) +f
1
, f
2
v
3
)
= (f
2
, v
2
)(f
1
, v
1
), (f
3
, v
3
) + (f
3
, v
3
)(f
1
, v
1
), (f
2
, v
2
),
Here we have used that the bracket [ , ]
V
is /-linear. Finally, for the Jacobi identity we have
(f
1
, v
1
), (f
2
, v
2
), (f
3
, v
3
)
= (f
1
, f
2

A
,
X
f
1
(v
2
)
X
f
2
(v
1
) + [v
1
, v
2
]
V
+B(X
f1
, X
f2
)), (f
3
, v
3
)
= (f
,
f
2

A
, f
3

A
,
pr
2
([[
1
([X
f1
, X
f2
]),
2
(v
3
)]]
[[
1
(X
f3
), [[
1
(X
f1
),
2
(v
2
)]]]]
+[[
1
(X
f3
), [[
1
(X
f2
),
2
(v
1
)]]]]
[[
1
(X
f3
), [[
2
(v
1
),
2
(v
2
)]]]]
[[
1
(X
f3
), [[
1
(X
f1
),
1
(X
f2
)]]]]
+[[
1
(X
f3
),
1
([X
f1
, X
f2
])]]
+[[[
1
(X
f1
),
2
(v
2
)]],
2
(v
3
)]]
[[[
1
(X
f2
),
2
(v
1
)]],
2
(v
3
)]]
+[[[
1
(X
f1
),
1
(X
f2
)]],
2
(v
3
)]]
[[
1
([X
f1
, X
f2
]),
2
(v
3
)]]
+[[
1
([X
f1
, X
f2
]),
1
(X
f3
)]]

1
([[X
f1
, X
f2
], X
f3
]))
+[[v
1
, v
2
]
V
, v
3
]
V
).
18
It follows that
(f
1
, v
1
), (f
2
, v
2
), (f
3
, v
3
)
= (f
1
, f
2

A
, f
3

A
+f
2
, f
3

A
, f
1

A
+f
3
, f
1

A
, f
2

A
,
+ pr
2
([[
1
(X
f3
), [[
1
(X
f1
),
2
(v
2
)]]]]
[[
1
(X
f1
), [[
1
(X
f2
),
2
(v
3
)]]]]
[[
1
(X
f2
), [[
1
(X
f3
),
2
(v
1
)]]]])
+ pr
2
([[
1
(X
f3
), [[
2
(v
1
),
2
(v
2
)]]]]
[[
1
(X
f1
), [[
2
(v
2
),
2
(v
3
)]]]]
[[
1
(X
f2
), [[
2
(v
3
),
2
(v
1
)]]]])
+ pr
2
([[
1
(X
f3
), [[
1
(X
f1
),
1
(X
f2
)]]]])
+ pr
2
(
1
([[X
f1
, X
f2
], X
f3
]))
+ ([[v
1
, v
2
]
V
, v
3
]
V
)) = (0, 0).
Corollary 5.2. V is an ideal of /V with respect to the Poisson bracket dened by (11).
The proof of this result follows from the fact 0, f
A
= 0.
Corollary 5.3. If (/, ,
A
) is a Poisson algebra and we have a transitive Lie algebroid
V Der(/) V Der(/),


//
2
// //

endowed with an algebroid connection , then, /V is a Poisson algebra.

Proof. By Theorem 4.11, we know that (T, [[, ]], ) is isomorphic to (Der(/) V, , ), pr
1
). The
statement follows from Theorem 5.1.
Acknowledgements
JAV and DGB express their gratitude to the members of the Department of Mathematics of the
University of Sonora (where part of this work was done), particularly to R. Flores-Espinoza and G.
Davila-Rascon, for their warm hospitality. JAV also thanks the National Council of Science and
Technology in Mexico (CONACyT), for the research grant JB2-CB-78791.
References
[1] C. J. Atkin, J. Grabowski: Homomorphisms of the Lie algebras associated with a symplectic
manifold. Comp. Math. 76 (1990) 315-349.
[2] R. Abraham, J. Marsden: Foundations of Mechanics. W. A. Benjamin (1967), Menlo Park, CA.
[3] A. Connes: Non-commutative dierential geometry. Inst. Hautes Etudes Sci. Publ. Math. 62
(1985), 257-360.
[4] J. Cuntz, D. Quillen: Algebra extensions and nonsingularity. J.Amer. Math. Soc. 8 (1995),
251-289.
[5] M. DuBois-Violette, P. W. Michor: Derivations et calcul dierentiel non commutatif II. Compt.
Rend. Acad. Sci. Paris 319 Serie I (1994) 927-931.
[6] J. P. Dufour, N. Tien-Zung: Poisson Structures and Their Normal Forms. Birkhauser (2005),
Basel.
[7] J. Grabowski: Quasi-derivations and QD-algebroids. Rep. Math. Phys. 52 3 (2003), 445-451.
19
[8] J. C. Herz: Pseudo-algebres de Lie. C. R. Acad. Sci. Paris. 236 (1953), 1935-1937 and 2289-2291.
[9] J. Huebschmann: Poisson cohomology and quantization. J. reine angew. Math. 408 (1990),
57-113.
[10] J. Huebschmann: Lie-Rinehart algebras, Gerstenhaber algebras, and Batalin- Vilkovisky alge-
bras. Ann. de lInstitut Fourier 48 (1998) 425-440.
[11] J. Huebschmann: Extensions of Lie-Rinehart algebras and the Chern-Weil construction. In:
Festschrift in honour of Jim Stashes 60th anniversary. Contemp. Math. 227 (1999), 145-
176.
[12] N. M. Katz: Nilpotent connections and the monodromy theorem; applications of a result of
Turritin. Publ. Math. Inst. Hautes Etudes Sci. 39 (1970) 175-232.
[13] I. Kolar, J. Slovak, and P. W. Michor: Natural operations in dierential geometry. Springer-
Verlag (1993) Berlin.
[14] Y. Kosmann-Schwarzbach, F. Magri: Poisson-Nijenhuis structures. Ann. Inst. Henri Poincare,
Phys. Theor. 53 (1990), 35-81.
[15] Y. Kosmann-Schwarzbach, K. Mackenzie: Dierential operators and actions of Lie algebroids.
In: Quantization, Poisson Brackets and Beyond. Contemp. Math. 315 (2002) 213-233.
[16] J. Kubarski: The Chern-Weil homomorphism of regular Lie algebroids. Publications du
Departement de Mathematiques, Universite Claude Bernard Lyon 1, nouvelle serie, (1991),
1-70
[17] K. Mackenzie: Lie algebroids and Lie pseudoalgebras. Bull. London Math. Soc. 27 (1995), 97-
147.
[18] K. Mackenzie: General theory of Lie groupoids and Lie algebroids. Cambridge UP. London Math.
Soc. Lect. Note Ser. 212 (2005) Cambridge.
[19] G. Rinehart: Dierential forms for general commutative algebras. Trans. Amer. Math. Soc. 108
(1963), 195-222.
[20] Yu. Vorobjev: Coupling tensors and Poisson geometry near a single symplectic leaf. In Lie
Algebroids, Banach Center Publ. 54 (2001) 249-274.
[21] Yu. Vorobjev: On Poisson realizations of transitive Lie algebroids. J. of Nonlinear Math. Phys.
11, Suppl. (2004) 43-48.
20