SYMPLECTIC HODGE THEORY ON LIE ALGEBROIDS

SHANE RANKIN

Abstract. We explore the natural analogues of the Brylinksi condition, Strong Lefschetz condition, and
dδ-lemma in Symplectic Geometry originally explored by Brylinksi, Mathieu, Yan, and Guillemin to the
Symplectic Lie Algebroid case. The equivalence of the three conditions is re-established as a purely algebraic
statement along with a primitive notion of the dδ-lemma shown establsihed by Tseng, Yau, and Ho. We then
arXiv:2411.06012v1 [math.SG] 8 Nov 2024

show that the natural analogues of these in the Lie Algebroid setting holds as well with examples given

Contents
1. Introduction 1
1.1. Background 1
1.2. Results 2
2. Lefschetz Modules 3
2.1. The Lie Superalgebra 3
2.2. Equivalence of Surjectivity 6
2.3. Equivalence of Triples 7
2.4. Weakly Lefschetz Modules 10
2.5. Primitive Equivalence 11
3. Lefschetz Algebroids 14
3.1. Symplectic Lie Algebroids 14
3.2. Kahler-Weil Identities 17
3.3. Cohomology with Coefficients 19
4. Examples 21
4.1. Kodaria-Thurston Manifold 21
4.2. Symplectic Lie Algebras of Dimension Four 21
4.3. Six-Dimensional Nilmanifolds 22
4.4. E-manifolds 23
4.5. Kähler Lie Algebroids 24
References 25

1. Introduction
1.1. Background. Hodge theory for Symplectic Manifolds was introduced by Ehresmann and Liberman and
rediscovered by Brylinski [2]; Brylinksi conjectured that on a Symplectic Manifold (M, ω) of dimension 2m
every deRham cohomology class admits a symplectic harmonic representative. Mathieu [14] and Yan [25]
independently showed that the Brylinksi conjecture holds true if and only if (M 2m , ω) satisfies the Strong
Lefschetz Property, that is for all 0 ≤ k ≤ m, the map
[L]k : H m−k (M ) → H m+k (M )
[α] 7→ [ω k ∧ α]
is surjective. This result was then improved by Merkulov in [18], and Guillemin in [7] by showing that these
two conditions are equivalent to the symplectic dδ-lemma in the case that M was a compact manifold using
1
Poincaré duality. In a key departure from traditional Riemannian Hodge theory the Symplectic Laplacian
is no longer an elliptic operator as it vanishes identically, and as such the results that heavily makeup the
theory no longer work. Instead we primarily work with the observation that in the presence of a symplectic
form, the space of differential forms on a smooth manifold admits the structure of an sl2 (R)-module. This
“Symplectic Hodge Theory” has been extended to new settings in work by Lin [11] and others, which require
reestablishing many conditions in specific geometric settings. In this paper we work towards streamlining
this by proving analogous statements about modules in a relevant module category, and applying them
to Symplectic Lie Algebroids, a minimal smooth setting for the notions involved to make sense. We also
generalize the operators ∂+ , ∂− of Tseng and Yau [24] to this strictly algebraic setting (which we rename d
and d respectively), though we do not explore the related Laplacians, or Appelli and Bott-Chern cohomologies
corresponding to certain combinations of these operators. We then relate the original trio of equivalences
back to ∂+ ∂− -lemma of [8] by reestablishing the equivalence of this to the original triple in our algebraic
framework rather than the original topological framework considered. Through this generalization we are
also able to lift the compactness condition, though it was free to assume in the classical case of the tangent
bundle by Poincaré Duality and standard Algebraic Topological facts relating to the relevant conditions.
1.2. Results. We define a category of modules Cn over a suitable Lie Superalgebra with properties that
axiomatize the algebra of differential forms on a Symplectic Manifold much in the spirit of [15].
Definition 1. Let g be the 5 dimensional Lie Superalgebra with basis e, f, h, d, δ with e, f, h degrees 2, −2, 0
respectively and d, δ in degrees 1, −1 respectively all subject to the following relations
[e, f ] = h [e, h] = −2e [f, h] = 2f
[e, d] = 0 [f, d] = δ [h, d] = d
[e, δ] = d [f, δ] = 0 [h, δ] = −δ
[d, d] = 0 [d, δ] = 0 [δ, δ] = 0
where all commutators are graded commutators. As a Lie Superalgebra this is isomorphic to the 5-dimensional
Orthosymplectic Lie Superalgebra
It is well known that the space of forms of any symplectic manifold admits the structure of a g-module
where the even part of the algebra corresponds to the standard sl2 structure, d corresponds to the deRham
differential, and δ is defined in [2]. Since the space of forms admits finitely many degree, 0 through 2n, we
want to restrict our attention to modules for which h acts with finitely many eigenvalues. Moreover, since
h acts on the form integrally and via a diagonalizable action, we want to restrict to modules for which this
occurs as well.
Definition 2. For n ≥ 0, let Cn denote the full subcategory of g-mod such that h acts integrally via
diagonalizable action with finitely many eigenvalues between −n and n
Objects of this category are also referred to as “Modules of Finite h-type” or “Modules of Finite h-spectrum”
in the literature. In the case that the underlying geometric structure is Kähler, or more generally has the
Hard Lefschetz Property, the module enjoys similar properties and is called a “Lefschetz Module”.
Theorem. Let V ∈ Ob(Cn ). Then the following are equivalent
(1) The map [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all i
(2) The inclusion (ker(δ), d) ֒→ (V, d) is a quasi-isomorphism
(3) The dδ-lemma holds for V
(4) The d d-lemma holds for V
After introducing the notion of the dδ-lemma holding weakly in section 2.4, we also weaken this to the
following theorem
Theorem. Let V ∈ Ob(Cn ). Then the following are equivalent
(1) The dδ-lemma holds for V up to degree s
2
 (2) The map [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all |i| ≥ s
(3) The inclusion (ker(δ), d) ֒→ (V, d) is a quasi-isomorphism in degree |i| ≥ s
Applying the first of these two theorems to a Symplectic Lie Algebroid with representation we have the
following theorem:
Theorem. Let (A → M, ρ, ω) be a Symplectic Lie Algebroid of fiber dimension 2m, with representation
(E, ∇). Then the following are equivalent:
m−k m+k
(1) The map [L]k : HA (M, E) → HA (M, E) given by [ω]k ∧ − is an isomorphism for all k
(2) The inclusion of complexes (ker( ∇ ), ∇) ֒→ (Ω•A (M, E), A ∇) is a quasi-isomorphism
A ∗ A

(3) The (A, E)-dδ-lemma holds for A

(4) The (A, E)-d d-lemma holds for A
Choosing E to be the trivial line bundle on M , A = T M , and with the standard deRham differential d as
the connection we recover the main results of [25], [14], and [8]. In section 2 we introduce the relevant Lie
Superalgebra and associated module category and go on to establish the four results equivalence as a purely
algebraic fact as well as reestablish the notion of the “weak” Lefschetz property as seen in [4]. In section 3
we introduce Symplectic Lie Algebroids, and define operators on the space of A-forms analogous to those
on a Symplectic Manifold. Section 4 contains examples of the theorems proven. First, the classical example
KT 4 - The Kodaira-Thurston manifold, fits well into the framework as it should. Next, using Ovando’s
[22] classification of symplectomorphism classes of 4-dimensional Lie Algebras, we discuss which admit the
structure of a Lefschetz Module. Moving on we consider six-dimensional Nilmanifolds, and discuss which of
the 26 isomorphism classes admitting symplectic forms admit Lefschetz maps that are isomorphisms. We
then discuss E-manifolds as studied by Miranda and Scott [19], and finally Kähler Lie Algebroids.

2. Lefschetz Modules
2.1. The Lie Superalgebra.
Proposition 2.1. If V ∈ Ob(Cn ), then V admits a decomposition
Mn
V = Vi
i=−n
where h acts on Vi as scaling by i, i.e. i · idVi
Lemma 2.1. Every cyclic V ∈ Ob(Cn ) is finite dimensional
P
Proof. Let v generate V as a U (sl2 (R))-module, and decompose v into it’s components v = vi for vi ∈ Vi .
It suffices to show that the submodule each vi generate is finite dimensional. By the Poincaré-Birkhoff-Witt
theorem, this module is spanned by the vectors
hvi i = spanm,r,k {er f m hk vi }
Since h acts by scaling on vi , it suffices to check then on
hvi i = spanm,r {er f m vi }
Now, since there are only finitely many eigenvalues, we have that r, m ≤ n. From this we can note that there
are finitely many combinations of m and r such that
−n ≤ −2m + 2r + i ≤ n
Thus there are finitely many generators for hvi i 
Now, we’d like to be able to talk about primitive elements as normal in sl2 representation theory.
L
Definition 3. Suppose V ∈ Ob(Cn ) with decomposition V = i Vi . We call a vector vi ∈ Vi primitive if
f vi = 0 or equivalently ei+1 vi = 0
Using this definition, we reprove a common structure theorem for sl2 -modules
3
Theorem 2.2. Let V ∈ Ob(Cn ). Then every v ∈ V can be written as a finite sum
X
v= ei vi
where the vi are primitive
Proof. Start by replacing V with hvi. This is again an object of Cn , and as it’s cyclic must be finite
dimensional. We can then decompose it into a finite sum of irreducibles, and write v as it’s components
in these, so it suffices to show elements in these irreducibles is of the form desired. Suppose v ∈ V−n
and u ∈ V−n+1 , both of which are primitive by degree considerations. Now, since f ek v = λ1 ek−1 v and
f ek u = λ2 ek−1 u for nonzero constants λi , we know that they span a nonzero submodule together. Since
we’re in the irreducible case, this must be the entire module. Moreover, they all correspond to distinct
eigenvalues of h, so they must be linearly independent, i.e. they form a basis of V . 
Another useful fact is the following classical result
L
Proposition 2.2. Suppose V ∈ Ob(Cn ) with decomposition V = i Vi . Then the map ei : V−i → Vi is a
bijection
Proof. This follows from classical sl2 (R) representation theory 
As well as the following bracket relations
Lemma 2.3. [ei , δ] = iei−1 d
Proof. We proceed by induction, the base case is given as definition. Suppose this holds for the i − 1 case,
then we have that
ei δ − δei = e(δei−1 + (i − 1)ei−2 d) − ei δ
= eδei−1 + (i − 1)ei−1 d − δei
= (δe + d)ei−1 + (i − 1)ei−1 d − δei
= iei−1 d

Lemma 2.4. [h, ei ] = 2iei
Proof. This holds from definition for i = 1, so suppose it holds for i − 1. Then we have that
hei − ei h = (ei−1 h + 2(i − 1)ei−1 )e − ei h
= ei−1 he + 2(i − 1)ei − ei h
= ei−1 (eh + 2e) + 2(i − 1)ei − ei h
= 2iei

We also want to bring the notion of the ⋆ involution present in classical Symplectic Hodge Theory:
Definition 4. On any V ∈ Ob(Cn ), define the “star operator” ⋆ : Vi → V−i to be the involution such that
δ|Vi = (−1)i+1 ⋆ d⋆
Now, in the spirit of classical deRham cohomology we arrive at the following definitions:
Definition 5. Let V ∈ Cn , and v ∈ V . Then we define the following
(1) v is “closed” if v ∈ ker(d)
(2) v is “exact” if v ∈ im(d)
(3) v is “coclosed” if v ∈ ker(δ)
(4) v is “coexact” if v ∈ im(δ)
4
 More generally we can adopt the language of harmonicity as follows
Definition 6. An element v ∈ V , where V ∈ Cn , we say that v is “harmonic” if v ∈ ker(d) ∩ ker(δ). The
submodule of harmonic elements of V will be denoted Vb
Proposition 2.3. For V ∈ Ob(Cn ), Vb is a g-submodule of V .
Proof. This is clearly a vector subspace and closed under the action of d and δ. If v ∈ Vb , then dev = edv = 0
and δev = eδv − dv = 0. Likewise df v = f dv + δv = 0 and δf v = f δv = 0. The closedness under h follows
from e and f 
One can regard V as a bi-differential Z-graded complex after writing V out as
d d d d
... Vi−1 Vi Vi+1 ...
δ δ δ δ
L
Definition 7. Let V ∈ Ob(Cn ) with decomposition V = i Vi . The cohomology of V regarded as the above
complex in d at the i-th position is denoted H i (V, d)
L
Definition 8. Let V ∈ Ob(Cn ) with decomposition V = i Vi . The homology of V regarded as the above
complex in in δ at the i-th position is denoted Hi (V, δ)
Lemma 2.5. Let V ∈ Ob(Cn ), and v ∈ V . If v is closed, then ⋆v is coclosed. Similarly, if v is coclosed,
then ⋆v is is closed.
Proof. Suppose that dv = 0. Then we have that
δ(⋆v) = ± ⋆ d ⋆ ⋆v = ± ⋆ dv = 0
Now, suppose δv = 0. Then we have that
δv = 0
±⋆d⋆v =0
±d ⋆ v = ⋆0 = 0
=⇒ d(⋆v) = 0

Lemma 2.6. Let V ∈ Ob(Cn ), and v ∈ V . If v is exact, then ⋆v is coexact. If v is coexact, then ⋆v is exact
Proof. Suppose that v is exact, that is v = dw. Then we have that
⋆v = ⋆dw = ⋆d ⋆ ⋆w = ±δ(⋆w)
Similarly, suppose v = δw. Then we have
⋆v = ⋆δw = ± ⋆ ⋆d ⋆ w = d(± ⋆ w)

Theorem 2.7. For V ∈ Ob(Cn ), there is an isomorphism
H i (ker(δ), d) ∼
= H −i (ker(d), δ)
Given by the ⋆ map
Proof. Define a map
ϕ : Vbi /d(Vi−1 ∩ ker(δ)) → Vb−i /δ(Vi+1 ∩ ker(d))
Given by sending [v] 7→ [⋆v]. This is well-defined by the preceding lemmas, and is surjective as for any class
[v], it has preimage [⋆v]. To see injectivity, suppose that ϕ[v] vanishes, then by definition we have
⋆v = δw = ± ⋆ d ⋆ w
5
For some w ∈ ker(d) ∩ Vi+1 , and the sign depending on the grading. Applying ⋆ to the above relation we
have
v = ±d ⋆ w = d(± ⋆ w)
Which means that [v] was the zero class. The only possible point of confusion is why ± ⋆ w is in the kernel
of δ, but this again follows from the preceding lemmas 
2.2. Equivalence of Surjectivity. The following theorem was established independently by Yan and Math-
ieu.
Theorem 2.8 (Yan [25], Mathieu [14]). Let (M 2n , ω) be a symplectic manifold. Then the following are
equivalent:
n−k n+k
(1) The strong Lefschetz map [ω k ] : HdR (M ) → HdR (M ) is surjective
(2) Every deRham Cohomology class has has a (symplectic) harmonic representative
We can rephrase this purely as an algebraic equivalence: Let V ∈ Ob(Cn ), then the following are equivalent:
(1) For all 0 ≤ i ≤ n, the map [ei ] : H −i (V, d) → H i (V, d) is surjection
(2) The inclusion (ker(δ), d) ֒→ (V, d) induces a surjection in cohomology H i (ker(δ), d) → H i (V, d)
Lemma 2.9. [ei ] carries harmonic harmonic representatives to harmonic representatives
Proof. Suppose [v−i ] ∈ H −i (V, d) has a harmonic representative [v−i0
]. We claim that [ei v−i
0
] is a harmonic
i 0
representative of [e v−i ]. First, note that they’re cohomologous as if v−i = v−i + dτ−i−1 , then we have that
ei v−i = ei v−i
0
+ ei dτ−i−1 = ei v−i
0
+ dei τ−i−1
Moreover, we have that ei v−i
0
is harmonic since
δei v−i = ei δv−i − iei−1 dv−i = 0

The following is essentially a rephrasing of Yan’s original proof [25]:
Theorem 2.10 (D. Yan). Suppose V ∈ Ob(Cn ), then the following statements are equivalent
(1) The inclusion (ker(δ), d) ֒→ (V, d) induces a surjection in cohomology H i (ker(δ), d) → H i (V, d)
(2) For all 0 ≤ i ≤ n, the map [ei ] : H −i (V, d) → H i (V, d) is a surjection
Proof. Suppose condition 1 holds, that is the maps H i (ker(δ), d) → H i (V, d) are surjections, or equivalently
we can say we have surjections Vbi → H i (V, d). We also know by classical sl2 -representation theory that
ei : V−i → Vi is a bijection, and this restricts to a bijection on Vb−i → Vbi since the harmonic elements make
up a g-submodule of V . Together these give us the diagram
ei
Vb−i Vbi

H −i (V, d) H i (V, d)
[ei ]

Where the top and vertical arrows are surjections, implying the bottom map is surjective. This is the exact
map in condition 2. Now, suppose that [ei ] is surjective for all i. We’d like to show that the inclusion
map (ker(δ), d) ֒→ (V, d) is a surjection in cohomology, i.e. given any cohomology class [v] ∈ H • (V, d), we
can replace it by a class [w] where w ∈ ker(δ). First, we claim that it suffices to check this on cohomolo-
gies in nonpositive degree. To see this, fix [vi ] ∈ H i (V, d) for i > 0. Then as [ei ] as surjective, we have
that [vi ] = [ei ][w−i ] for some [w−i ] ∈ H −i (V, d), and we know by Propostion 2.9 that [ei ] will carry a har-
monic representative of [w−i ] to a harmonic representative of [vi ]. Now, we claim that for i > 0, H −i (V, d)
decomposes
H −i (V, d) = im([e]) + ker([ei+1 ])
6
Fix a class [v−i ] ∈ H −i (V, d) and consider [ei+1 v−i ] ∈ H i+2 (V, d). Since the [ei ] are surjective, we have that
there is class [β−i−2 ] ∈ H −i−2 (V, d) such that [ei+2 β−i−2 ] = [ei+1 v−i ]. Now, we can (trivially) write
[v−i ] = ([v−i − eβ−i−2 ]) + [e]([β−i−2 ])
It’s clear that the second summand is in the image of [e], we need only to check that the first is in the kernel
of [ei+1 ]. This follows from the definition as
[ei+1 ]([v−i − eβ−i−2 ]) = [ei+1 v−i − ei+2 β−i−2 ] = [ei+1 v−i − ei+1 v−i ] = [0]
We now proceed to prove the claim by induction. First, note that every class in H −n (V, d) and H −n+1 (V, d)
is harmonic. If [v−n ] ∈ H −n (V, d), then δv−n ≡ 0 trivially, and so v−n is harmonic. If [v−n+1 ] ∈ H −n+1 (V, d),
then we have that
δv−n+1 = [d, f ]v−n+1 = df v−n+1 − f dv−n+1 = 0
Since f v−n+1 ∈ V−n−1 ≡ 0, and v−n+1 is closed. So we only need to check on −n + 2 ≤ i ≤ 0, and the −n + 1
degree serves as the base case. Now, suppose the statement holds for H −i−1 (V, d) and we want to show that
it holds on H −i (V, d). Fix [w−i ] ∈ H −i (V, d). By the above decomposition we have that
[β−i ] = [α−i ] + [e]([β−i−2 ])
For [β−i−2 ] ∈ H −i−2 (V, d) and [α−i ] ∈ H −i (V, d) ∩ ker([ei+1 ]). By inductive hypothesis [β−i−2 ] possesses a
harmonic representative, and [e] carries harmonic representatives into harmonic representatives, so we can
freely assume [e]([β−i−2 ] is harmonic. We only need to show that [α−i ] admits a harmonic representative. By
definition of [α−i ] we have that dα−i = 0 and ei+1 α−i = dβi+1 for some βi+1 ∈ Vi+1 . Now, by the bijectivity
of ei+1 : V−i−1 → Vi+1 , there exists γ−i−1 ∈ V−i−1 such that βi+1 = ei+1 γ−i−1 . We claim that α−i − dγ−i−1
is a harmonic representative of [α−i ]. It’s clear that [α−i ] = [α−i − dγ−i−1 ], and that α−i − dγ−i−1 ∈ ker(d),
we only need to show that it’s in ker(δ). Since δ = [d, f ], this reduces to showing that α−i − dγ−i−1 ∈ ker(f ),
i.e. is primitive. Primitivity here is equivalent to showing that α−i − dγ−i−1 ∈ ker(ei+1 ). Computing we
have that
ei+1 (α−i − dγ−i−1 ) = ei+1 α−i − d(ei+1 γ−i−1 ) = ei+1 α−i − dβi+1 = ei+1 α−i − ei+1 α−i = 0


2.3. Equivalence of Triples. In the previous subsection we’ve shown that the following two statements are
equivalent for V ∈ Ob(Cn ):
(1) The map [ei ] : H −i (V, d) → H i (V, d) is a surjection
(2) The inclusion (ker(δ), d) ֒→ (V, d) induces a surjection in cohomology
Which in the original paper(s), Mathieu claims is equivalent to a third condition: The dδ-lemma.
Definition 9. The dδ-lemma holds for V ∈ Ob(Cn ) if the following equalities hold
im(d) ∩ ker(δ) = im(δ) ∩ ker(d) = im(dδ)
This is equivalent to the above two if we replace surjection with isomorphism in both claims; in the original
setting of the tangent bundle of a compact manifold these are equivalent by Poincare duality. The first of the
equivalences - that dδ-lemma is equivalent to the inclusion (ker(δ), d) ֒→ (V, d) follows from the following:
Theorem 2.11. Let V ∈ Ob(Cn ). Then the dδ-lemma is equivalent to the inclusion (ker(δ), d) → (V, d)
inducing an isomorphism in cohomology
Proof. See lemma 5.4.1 in [12] 

Proposition 2.4. Let V ∈ Ob(Cn ), and suppose that the dδ-lemma is satisfied. Then [ek ] : H −k (V, d) →
H k (V, d) is an isomorphism for all k
7
Proof. Since dδ-lemma implies that there are harmonic representatives, we have that that [ek ] is surjective
by the prior work. This also allows us to check the validity of this isomorphism by restricting to [ek ] :
H −k (ker(δ), d) → H k (ker(δ), d). So, suppose that [α−k ] ∈ ker([ek ]). Then we have that ek α−k = dτk−1 for
some τk−1 ∈ Vk−1 ∩ ker(δ). Then, note that
δ(ek (α−k )) = δdτk−1 = −dδτk−1 = 0
So, the form ek α−k ∈ im(d) ∩ ker(δ), so by the dδ-lemma we have that there is a ρk such that
ek α−k = dδρk
Now, since ek is a bijection for all k, let φk denote it’s inverse. Then we have that
α−k = φk (dδρk ) = −φk δdρk
Now, since [ek , δ] = kek−1 d, we have that
φk δ = δφk − kφk ek−1 dφk
Meaning that
a−k = −(δφk − kφk ek−1 dφk )(dρk )
= −δφk dρk + kφk ek−1 φk d2 ρk
= −δφk dρk
= dδφk ρk
So α−k ∈ im(d) for δφk ρk ∈ ker(δ) 
Proposition 2.5. Let V ∈ Ob(Cn ). If [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all i, then
ker(d) ∩ im(δ) = im(d) ∩ im(δ).
Proof. What we want to show is that for a general module element v ∈ V , we have that dδv = 0 implies
that δv = dφ for some φ. First, we claim that it suffices to check this on primitive elements. Towards this,
suppose that the statement P holds on primitive elements, and fix arbitrary v ∈ V . We can decompose v into
primitive elements, i.e. v = ei vi where the vi are primitive. Now, since dδv = 0, we have that
X X X
0 = dδv = dδei vi = d(ei δ − ie−i1 d)vi = ei dδvi
Since this sum is direct, this forces ei dδvi = 0 for all i. Since ei is acting on the negative weights in this sum,
it acts injectively, so dδvi = 0 for all i. By hypothesis, we have then that there exist φi such that δvi = dφi ,
and thus
X
δv = δei vi
X
= ei δvi − iei−1 dvi
X
= ei dφi − iei−1 dvi
X 
=d ei φi − iei−1 vi
Now, to see that this holds on primitive form we proceed as follows. Suppose v−i ∈ V−i is primitive and
dδv−i = 0. Then we have that
ei+1 δv−i = δei+1 v−i − (i + 1)ei dv−i
However, ei+1 v−i = 0 since v−i is primitive, and we’re left with
ei+1 δv−i = d(−(i + 1)ei v−i )
Rephrased this says that [δv−i ] ∈ ker[ei+1 ], which by the Strong Lefschetz Property is an isomorphism,
meaning that [δv−i ] is the zero class in cohomology, so δv−i = dφ for some φ ∈ V−i−2 . The only degree
where this argument does not hold is V−n , but for any v ∈ V−n , δv ≡ 0, so it’s trivially in the image of d, i.e.
δv = 0 = d(0). 
8
Proposition 2.6. Let V ∈ Ob(Cn ), and suppose that [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all
i. Then ker(d) ∩ im(δ) = ker(δ) ∩ im(d)
Proof. From the preceding proposition, we have that ker(d) ∩ im(δ) = im(d) ∩ im(δ) ⊂ ker(δ) ∩ im(d). So we
need to show that im(d) ∩ ker(δ) ⊂ im(δ) ∩ ker(d). So suppose α ∈ im(d) ∩ ker(δ), that is α = dβ and δα = 0.
Then we have
0 = δα = ± ⋆ d ⋆ α
Where sign depends on degree, but is not of importance for this proof. Since ⋆ is an isomorphism, we have
that d ⋆ α = 0,i.e. ⋆α ∈ ker(d). Now using exactness we have
⋆α = ⋆dβ = ± ⋆ d ⋆ ⋆β = ±δ ⋆ β
Thus ⋆α ∈ ker(d) ∩ im(δ). By the preceding proposition yet again we have then that ⋆α ∈ im(d) ∩ im(δ), and
so ⋆α = dρ. Using this we then have that
α = ⋆ ⋆ α = ⋆dρ = ⋆d ⋆ ⋆ρ = δ(± ⋆ ρ)
Giving us α ∈ im(δ) 
Proposition 2.7. Let V ∈ Ob(Cn ), and suppose that [e]i : H −i (V, d) → H i (V, d) is surjective for all i. Then
im(d) ∩ im(δ) = im(dδ)
Proof. Since the two operators anticommute, the ⊃ direction is clear, and we need only to prove the forward
inclusion which we’ll prove by induction. First note that the statement trivially holds elements in Vn or
V−n since if v ∈ im(δ) ∩ Vn , then v = δ0 = 0, and likewise for V−n . Our base case is on Vn−1 , so choose
vn−1 ∈ Vn−1 , and let
vn−1 = dγn−2 = δwn
Now, since dwn is trivially zero it represents a cohomology class, and so we can choose a harmonic represen-
tative of said class giving us wn = wn0 − dτn−1 where wn0 is harmonic. We then have that
vn−1 = δwn
= δ(wn0 − dτn−1 )
= −δdτn−1
= dδτn−1
Now, suppose this holds for Vk for some −n < k < n, and fix vk−1 ∈ Vk−1 ∩ im(d) ∩ im(δ). By definition we
have that
vk−1 = dγk−2 = δβk
From this, define ξk+1 = dβk , and observe that
δξk+1 = δdβk = −d(δβk ) = −d2 γk−2 = 0
So ξk+1 ∈ im(d)∩ker(δ). By the remark preceding this proposition we have that im(d)∩ker(δ) = im(d)∩im(δ),
which by inductive hypothesis extends to im(d) ∩ ker(δ) = im(d) ∩ im(δ) = im(dδ), so ξk+1 = dδφk+1 . Then
note that d(βk − δφk+1 ) = 0, so it admits a harmonic representative, i.e. βk − δφk+1 = βk0 − dτk for some
τk ∈ Vk . Then applying the definition of vk−1 we have
vk−1 = δβk
= δ(βk0 − dτk + δφk+1 )
= −δdτk
= dδτk

At this point, we have enough to establish the equivalence of the triple: dδ-lemma, Strong Lefschetz
property, and Brylinksi property. We’ve already proven all the of the relevant pieces, but for clarity state
them here in one place
9
Theorem 2.12. Let V ∈ Ob(Cn ). The the following conditions are equivalent:
(1) The dδ-lemma holds for V
(2) The map [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all i
(3) The inclusion (ker(δ), d) ֒→ (V, d) is a quasi-isomorphism
Proof. We have that the first and third conditions are equivalent by Theorem 2.11. Then we have that the
first implies the second by Proposition 2.4, and the reverse direction follows from the Proposition 2.5, 2.6,
and 2.7. 
In light of this theorem, we make the following definition
Definition 10. If V ∈ Ob(Cn ) satisfies any of the above equivalent conditions in Theorem 2.12 we call V a
“Lefschetz Module”
Corollary 2.12.1. Let V ∈ Ob(Cn ) be a Lefschetz module. Then H ∗ (V, d) is an sl2 (R)-module
Proof. We can see from the g relations that
[e, d] = 0 [f, d] = δ [h, d] = d
∗
So on H (V, d), we see that e and h pass to cohomology, i.e. it’s always a B-module. The only possible
problem lies in well-definedness of f . Suppose [v] = [w] ∈ H ∗ (V, d), that is v = w + dφ for some φ ∈ V . Then
we have that
f [v] = [f v] = [f (w + dφ)] = f [w] + [f dφ] = f [w] + [(df + δ)φ] = f [w] + [δφ]
Now δφ ∈ im(δ) ∩ ker(d), so since V is Lefschetz, there exists some ϕ such that δφ = dδϕ so
f [v] = f [w] + [dδϕ] = f [w]

2.4. Weakly Lefschetz Modules. There are examples on manifolds where the Lefschetz condition is only
partially met, i.e. the maps [L]k are isomorphisms for some k, but not all 0 ≤ k ≤ n
Definition 11. Given V ∈ Ob(Cn ), we say that V is “s-Lefschetz” if ek is a isomorphism for 0 ≤ s ≤ k ≤ n
The same theorems from before hold in this setting as well, but only on the portions on the modules for
which the maps ek are isomorphisms.
Theorem 2.13. Suppose V ∈ Ob(Cn ) , then the following statements are equivalent
(1) The inclusion (ker(δ), d) ֒→ (V, d) induces a surjection in cohomology H i (ker(δ), d) → H i (V, d) in
degrees |i| ≥ s
(2) For all 0 ≤ s ≤ i ≤ n, the map [ei ] : H −i (V, d) → H i (V, d) is a surjection
Proof. The forward direction follows in the same exact way it does in the proof of Theorem 2.10. The reverse
direction followed by induction that crucially used the fact that we could decompose H −i (V, d) for i > 0.
This decomposition was the only part of the proof that relied on the reverse direction, and the rest of the
proof holds if we only assume that the [ei ] are surjective from V−n up to V−s . 
The upgraded equivalence also holds under the added assumption that ei is in fact an isomorphism
Definition 12. Given a module V ∈ Ob(Cn ), we say that V has the “dδ-lemma up to degree s” if
im(d) ∩ ker(δ) = im(dδ) = im(δ) ∩ ker(δ) on Vi for − n ≤ i ≤ s
im(d) ∩ ker(δ) = im(dδ) on Vs+1
for −n ≤ s ≤ −1
Note that by ⋆ duality if this equality holds on Vi for −n ≤ i ≤ −s it then also holds in Vi for s ≤ i ≤ n.
From this observation we’ll only prove statements about it in non-positive degree, as the duality will extend
the results to positive degrees.
10
Theorem 2.14. Let V ∈ Ob(Cn ). Then the following are equivalent
(1) The dδ-lemma holds for V up to degree s
(2) The map [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all |i| ≥ s
(3) The inclusion (ker(δ), d) ֒→ (V, d) is a quasi-isomorphism in degree |i| ≥ s
Proof. The proof will follow similarly to the original. The first and third condition are equivalent by [12],
whose argument works in each degree. The first implies the second using Proposition 2.4 and acknowledging
the proof is symmetric in it’s degree arguments about the zero weight. Finally, to see that the second
condition implies the first, note that proposition 2.5 and proposition 2.6 are degree agnostic, and proposition
2.7 is done by induction, which we can once again stop early when the hypothesis fail in degree −s and
achieve the result 
2.5. Primitive Equivalence. Much of this section is rephrasing of work in Chung-I Ho’s thesis [8] and
Tseng’s paper [24]
L
Definition 13. Suppose V ∈ Ob(Cn ) with decomposition k Vk . Then the space of primitive elements of
degree k will be denoted Pk = ker(f ) ∩ Vk
Throughout the rest of this section, fix V ∈ Ob(Cn ), so that we may speak of elements vi ∈ P i freely.
Lemma 2.15. Suppose that wi ∈ Pi . Then the primitive decomposition of dwi only has two terms
Proof. First, note that
f 2 dwi = f (df + δ)wi = f δwi = δf wi = 0
As wi is primitive. Taking a primitive decomposition of dwi we have
X
dwi = ek vi+1−2k
k≥0
2
So applying f we have X
0= f 2 ek vi+1−2k
k≥0
Where vi+1−2k ∈ Pi+1−2k Which means for all k ≥ 2, we had to have had that ek vi+1−2k = 0, but as ei acts
injectively we have that vi+1−2k = 0 for these k, leaving us with
dwi = vi+1 + evi−1

Definition 14. Given vi ∈ Pi with dvi = vi+1 + evi−1 , define
d vi = vi+1
d vi = vi−1
so on P i we have dvi = vi+1 + evi−1 = d vi + e d vi
These are the same operators as ∂+ and ∂− as introduced in [24].
P These extend to the entire
P module by
taking primitive decomposition of any element v ∈ V , i.e. if v = vi for vi ∈ Vi , and vi = k≥0 ek vi−2k
then we have that
XX
dv = ek d vi−2k
i k≥0
XX
dv = ek d vi−2k
i k≥0

Lemma 2.16. We have that [d, e] = [d, e] = [d, d] = [d, d] = 0 where all are graded commutators, as well as
e(d d) = −e(d d)
On Vn−1 , Vn and [d, d] = 0 on Vk for k < n − 1
11
Proof. This follows the same as it did in [24], Lemma 2.5 
Proposition 2.8. For vi ∈ P i , we have that f d vi = 0
Proof. Since vi is primitive, we have that [f, d]vi = f d vi . By the Jacobi identity we have that
1
[f, d] =
[h, [d, f ]] + [f, d]
2
Thus we have to have that h and [d, f ] commute. Checking this on vi gives us that
(i − 1)[d, f ]vi = i[d, f ]vi
Forcing [f, d]vi = 0 
Proposition 2.9. For vi ∈ P i δvi = (1 − i) d vi
Proof. Using the definition of δ we have
δvi = f dvi − df vi = f dvi = f d vi + f e d vi
We know the first summand vanishes by the preceding lemma, and on the second we have that
f e d vi = (ef − h) d vi = −h d vi = −(i − 1) d vi

Lemma 2.17. On P i , we have that dδ = (1 − i) d d
Proof. Using the preceding lemmas we have
dδvi = (i − 1)d d vi = (i − 1)(d +e d) d vi = (1 − i) d d vi

Definition 15. We say that v ∈ V is “symphonic” if d v = d v = 0
This was previously called “primitive harmonic” in the literature, see [8], however to avoid the phrase
“harmonic primitive harmonic” we introduce this terminology.
Lemma 2.18. v ∈ V is harmonic if and only if each term of it’s primitive decomposition is symphonic
P
Proof. Let v = vi with vi = k≥0 ek wi−2k for wi−2k ∈ P i−2k . Then if each wi−2k , since d = d +e d it’s clear
that v is closed. Using the commutation relation given in Lemma 2.3, we have
XX
δv = δek wi−2k
i k≥0
XX
= (ek δ − kek−1 d)wi−2k
i k≥0
XX
= ek (1 − i) d wi−2k
i k≥0
=0
P
Now, suppose that v = i vi for vi ∈ Vi is harmonic. this occurs if and only if each vi is harmonic, which
occurs if and only if the term in the primitive decomposition of each vi is harmonic, so it suffices to chow
ci . Then
this for a primitive harmonic element. Towards this, suppose v ∈ P
0 = δv = (i − 1) d v
So v ∈ ker(d). Moreover we have
0 = dv = d v + e d v = d v
Thus v ∈ ker(d) 
12
Definition 16. Let V ∈ Ob(Cn ). We say that V satisfies the d d-lemma if the following equalities hold
im(d) ∩ ker(d) ∩ P k = im(d d) ∩ P k −n + 1 ≤ k ≤ −1
im(d) ∩ ker(d) ∩ P −n = im(d d) ∩ P −n
im(d) ∩ ker(d) ∩ P 0 = im(d d) ∩ P 0
Lemma 2.19. The dδ-lemma holds for V ∈ Ob(Cn ) if the d d-lemma does
Proof. It suffices to check this on each weight, so fix vi ∈ Vbi and first suppose that vi ∈ im d so that
vi = dwi−1 . Then taking primitive decompositions of both we have
X X
vi = ek vi−2k wi−1 = ek wi−1−2k
k≥0 k≥0

Combining this with the fact that vi = dwi−1 we have

X
ek (vi−2k − dwi−1−2k ) = 0
k≥0

Forcing vi−2k = dwi−1−2k = d wi−1−2k + e d wi−1−2k for all k. Since vi is harmonic, each vi−2k is symphonic,
so we must have that d wi−1−2k ∈ im d ∩ ker d and d wi−1−2k ∈ im d ∩ ker d. By the d d-lemma then we must
have that there are γi−2k , ρi−2−2k ∈ im(d d) (possibly zero) such that
d wi−1−2k = d d γi−2k d wi−1−2k = d d ρi−2−2k
From here, we have that
X
vi = ek (d wi−1−2k + e d wi−1−2k )
k≥0
X
= ek (d d γi−2k + e d d ρi−2(1+k) )
k≥0
X  
γi−2k ρi−2(1+k)
= ek dδ + edδ
1 − (i − 2k) 1 − i + 2(1 + k)
k≥0
X  
γi−2k ρi−2(1+k)
= dδ ek +e
1 − (i − 2k) 1 − i + 2(1 + k)
k≥0

Where we repeatedly use the g relations, and make note that the constants that appear in the denominator
are non-vanishing. The only possible point of confusion is why the γi−2k and ρi−2(1+k) are primitive. If
they’re not, then taking a primitive decomposition of them, it’s easy to see that we can adjust by constants
to force this to be true. A similar argument holds for the case when v ∈ im(δ). 
Lemma 2.20. The d d-lemma holds for V ∈ Ob(Cn ) if the dδ-lemma does
Proof. Suppose that vi ∈ ker(d) ∩ ker(d) ∩ P i and recall that this forces vi to be harmonic and each of it’s
primitive decomposition terms to be symphonic. First, suppose that vi ∈ im(d), i.e. vi = d βi+1 . Then taking
primitive decompositions we have
X X
vi = ek vi−2k βi+1 ek βi+1−2k
k≥0

Using the coexactness we have must have that

 
βi+1−2k
vi−2k = d βi+1−2k = δ
−i + 2k
So vi−2k ∈ im(δ) ∩ ker(d) = im(dδ). This then gives us that there exist ρi−2k such that
vi−2k = dδρi−2k = d d(1 − i + 2k)ρi−2k
13
So we have X X
vi = ek vi−2k = d d ek ((1 − i + 2k)ρi−2k )
k≥0 k≥0
Again it may be unclear as to why the ρi−2k are primitive, but this can be chosen to happen as before. Now,
suppose that vi ∈ im(d) so vi = d βi−1 for some βi−1 . Then equating primitive decompositions we have that
vi−2k = d βi−1−2k where X X
vi = ek vi−2k βi−1 = ek βi−1−2k
k≥0 k≥0

Since vi is symphonic, we have that d d βi−1−2k = 0 for all k. We then have that
X
dδβi−1 = dδek βi−1−2k
X
= ek dδβi−1−2k
k≥0
X 1
= ek d d βi−1−2k
2(1 − k) − i
k≥0
X −1
= ek d d βi−1−2k
2(1 − k) − i
k≥0
=0
So δdβi−1 = 0, meaning that dβi−1 ∈ ker(δ) ∩ im(d). By the dδ-lemma then we have that there exist γi such
that dβi−1 = dδγi . Taking a primitive decomposition of γi we have that
X
dδγi = ek dδγi−2k
k≥0
X
= (1 − i + 2k)ek d d γi−2k
k≥0
X
= dd (i − 1 − 2k)ek γi−2k
k≥0

Which shows us that

 
X
vi = d βi−1 − e d βi−1 = d d (i − 1 − 2k)ek γi−2k − eβi−1 
k≥0

That is, vi ∈ im(d) from which the result follows from the previous case. 
Theorem 2.21. Let V ∈ Ob(Cn ). Then the following are equivalent
(1) The dδ-lemma holds for V
(2) The map [ei ] : H −i (V, d) → H i (V, d) is an isomorphism for all i
(3) The inclusion (ker(δ), d) ֒→ (V, d) is a quasi-isomorphism
(4) The d d-lemma holds for V
Proof. The first three are equivalent by 2.12. The first and last are equivalent by Lemmas 2.20 and 2.19 

3. Lefschetz Algebroids
3.1. Symplectic Lie Algebroids. Lie Algebroids were originally introduced by Pradines [23] as a general-
ization of the tangent bundle of a manifold to the Differentiable Groupoid setting.
Definition 17. A “Lie Algebroid” (A, ρ, M ) over a C ∞ -manifold M is a vector bundle A → M , equipped
with an “anchor” ρ : A → T M , and a bracket [·, ·]A on sections of A that is bilinear, alternating, and satisfies
the Jacobi identity, such that
14
 (1) [X, f Y ]A = f [X, Y ]A + (ρ(X) · f )Y for all X, Y ∈ Γ(A) and f ∈ C ∞ (M )
(2) ρ([X, Y ]A ) = [ρ(X), ρ(Y )]T M for all X, Y ∈ Γ(A)
Definition 18. Given a Lie Algebroid (A, ρ, M ), we can define the “deRham-Chevalley-Eilenberg complex”
of A, where our cochain complex is made of
ΩkA (M ) = Γ(Λk A∗ )
And we define M
Ω∗A (M ) = ΩkA (M )
k
k+1
This comes with an A-differential, dA : ΩkA (M ) → ΩA (M ), defined as
k+1
X
dA η(X1 , ..., Xk+1 ) = ci , ..., Xk+1 )
(−1)i+1 ρ(Xi ) · η(X1 , ..., X
i=1
X
+ ci , ..., X
(−1)i+j η([Xi , Xj ], X1 , ..., X cj , ..., Xk+1 )
i<j

For Xi ∈ Γ(A) and η ∈ ΩkA (M ).

A standard computation shows that d2A = 0, and as such we can consider the cohomology of this complex
k
Definition 19. Let (A, ρ, M ) be a Lie Algebroid. The “Lie Algebroid Cohomology” of A, denoted HA (M ) is
the cohomology of the complex
dA k−1 dA dA dA
... ΩA (M ) ΩkA (M ) Ωk+1
A (M ) ...
that is  
dA
ker ΩkA (M ) −−→ Ωk+1
A (M )
k  
HA (M ) =
k−1 dA k
im ΩA (M ) −−→ ΩA (M )
which we can rephrase busing A-closed and exact terminology
k dA -closed forms
HA (M ) =
dA -exact forms
The standard Cartan Calculus from the Tangent Bundle also extends to Lie Algebroids, however we’ll only
use the notion of the exterior and interior products
Definition 20. Given X ∈ Γ(A), we define “A-interior multiplication” with X, or “contraction by X”,
k−1
denoted ιX , as the operator ιX : ΩkA (M ) → ΩA (X) such that
ιX η(X1 , ..., Xk−1 ) = η(X, X1 , ..., Xk−1 )
Where Xi ∈ Γ(A) and η ∈ ΩkA (M )
Symplectic Lie Algebroids were originally introduced by Nest and Tsygan [20] and much work has been done
on them in recent years by may authors. As in classical Symplectic Geometry, Symplectic Lie Algebroids
arise naturally as the “phase space” of any Lie algebroid as originally uncovered by Martínez [13]. Most
recently, Lin et al. [10] proved an analogue of the Darboux-Marsden-Weinstien Theorem and described the
processes of symplectic reduction in this context.
Definition 21. Let (A, ρ, M ) be a Lie Algebroid. An “A-Symplectic form” ω is a dA -closed, nondegenerate
element of Ω2A (M )
This allows us to define Symplectic Lie Algebroids
Definition 22. A “Symplectic Lie Algebroid” is a Lie Algebroid A equipped with a A-symplectic form.
15
Corollary 3.0.1. Let (A → M, ρ, ω) be a Symplectic Lie Algebroid. Then the fiber dimension of A is even
and A is orientable
Proposition 3.1 (Martinez). Let B → N be an arbitrary Lie Algebroid, and π : B ∗ → N be the projection.
Define the bundle
A = π ! B = T B ∗ ×T N B
Which is a Lie Algebroid over B ∗ . Then ω = −dA α is a canonical symplectic structure on B ∗ , where
α ∈ Ω1A (B ∗ ) is defined as follows after identifying elements of A as tuples in the product bundle
α(x, p, v, b) = p(b) x ∈ N, p ∈ Bx∗ , v ∈ Tp B ∗ , b ∈ Bx
Satisfying Tp π(v) = ρB (b).
Proof. See originally Martinez [13] or Lin [10] 

Proposition 3.2. Given Symplectic Lie Algebroids (A1 → M1 , ω1 ) and (A2 → M2 , ω2 ), the product (A1 ×
A2 → M1 × M2 , ω) where ω = π1∗ ω1 + π2∗ ω2 is a Symplectic Lie Algebroid
Proof. The product of two Lie Algebroids has a unique Lie Algebroid structure [17], so we need to check that
ω is a symplectic form. It’s clear that it’s nondegenerate as it’s made from the pullback of two nondegenerate
forms. To see that it’s closed note that πi are Lie Algebroid morphisms, and a map f : A → B is a Lie
Algebroid morphism if and only if the induced map f ∗ : Ω∗B → Ω∗A is a chain map. Using this we have that
πi∗ are chain maps and so
dA1 ×A2 ω = dA1 ×A2 π1∗ ω1 + dA1 ×A2 π2∗ ω2 = π1∗ (dA1 ω1 ) + π2∗ (dA2 ω2 ) = 0


Lemma 3.1. Let (A → M, ρ, ω) be a Symplectic Lie Algebroid of fiber dimension 2m. Then for all p ∈ M ,
there exists a smooth symplectic local frame of A, i.e. there exists a frame {ui , vi }m i=1 ⊂ Γ(A) such that
{ui (q), vi (q)}m
i=1 is a symplectic basis of Aq for all q ∈ U , an open neighborhood of p.

Proof. Fix p ∈ M , and let B0 = {si }2m

i=1 be a smooth local frame of A over a neighborhood V of p. Now,
define u1 = s1 , and choose ǫ1 ∈ Γ(A|U ) such that
 
s2 + ǫ 1
ω u1 , =1
C1
Which (after possibly restricting to a smaller open set p ∈ V1 ⊂ V ) exists by non-degeneracy of ω, and we
rescale by C1 to have the desired constant 1, so that for v1 := s2C+ǫ
1
1
we have ω(u1 , v1 ) = 1. Now, let

u2 = s3 − ω(s3 , v1 )u1 + ω(s3 , u1 )v1

Then ω(u1 , u2 ) = ω(v1 , u2 ) = 0 by design. For v2 , start by choosing ǫ2 ∈ Γ(A|U ) such that
 
s4 + ǫ 2
ω u2 , =1
C2
s4 +ǫ2
Which again is possible by non-degeneracy of ω after possibly shrinking to p ∈ V2 ⊂ V1 . Call v2′ = C2 ,
and now define
v2 = v2′ − ω(v2′ , v1 )u1 + ω(v2′ , u1 )v1
Then we have that
ω(u1 , v2 ) = ω(v1 , v2 ) = 0 ω(u2 , v2 ) = 1
We can continue this process m times giving us a frame of the desired form over Vm := U . 
16
 Note here that we’re not insisting that the algebroid 1-forms that make up our local coframe of A∗ are
differentials of coordinate functions, or that they’re exact. In fact the coframe we define is given by
ξi = u♭i µi = −vi♭
Where ♭ : Γ(A) → Γ(A∗ ) is given by X 7→ ω(X, −). The local frame {µi , ξi }m ∗ m
i=1 of A is dual to {ui , vi }i=1 ,
but a geometrically inclined choice of normal coordinates that correspond to this are harder to pin down.
There are cohomological obstructions to making such a choice as shown by Miranda and Scott [19] (Theorem
29). For this reason we will not make an appeal to a full “Darboux Theorem” though work has been done
on this see [10]. In many settings such as b-manifolds there are results on local normal forms, and in these
settings the following sections results can be more easily understood geometrically, see [16], [19].
3.2. Kahler-Weil Identities. Throughout this section, let (A → M, ρ, ω) be a Symplectic Lie Algebroid of
fiber dimension 2m. We introduce the main two operators involved in this result. The “Lefschetz” map
L : ΩkA (M ) → Ωk+2
A (M )
α 7→ ω ∧ α
And if we let λ denote the dual bivector to ω, we define
k−2
Λ : ΩkA (M ) → ΩA (M )
α 7→ ιλ α
By our weaker analogue of the Darboux Theorem, Lemma 3.1 , we have local symplectic frames of A and A∗
on an open set U given by
A|U = spanC ∞ (M) {ui , vi }m
i=1 A∗ |U = spanC ∞ (M) {µi , ξi }m
i=1
In such a frame we can write local expressions for L and Λ
X X
L= eµi eξi Λ= ιvi ιui
i i
2m−k
In order to understand these operators better define ⋆ : ΩkA (M ) → ΩA (M ) as the operator
m
ω
α ∧ ⋆β = λk (α, β)
m!
For all α, β ∈ ΩkA (M ). A standard result shows the following
Proposition 3.3. ⋆⋆ = Id
k−1
Definition 23. The “Brylinski” differential, d⋆A : ΩkA (M ) → ΩA (M ) defined as follows for α ∈ ΩkA (M )
d⋆A α = (−1)k+1 ⋆ dA ⋆ α
This operator also creates a chain complex on the Ω•A (M ) in the opposite direction of dA since d⋆A ◦d⋆A = 0.
The homology of this complex will not be explored here. As in the classical setting, we have the following
definition
Definition 24. We say that a form α ∈ Ω∗A (M ) is “A-symplectic harmonic” or just “harmonic” if dA α =
b k (M ) and the set of all harmonic forms as Ω
d⋆A α = 0, and denote the set of harmonic k-forms as Ω b ∗ (M ).
A A

Proposition 3.4. Λ = ⋆L⋆

Proof. This is a strictly Linear Algebraic fact, see [6] for details 
Lemma 3.2. If α ∈ ΩkA (M ), then [L, Λ]α = (k − m)α
Proof. These are local, C ∞ (M )-linear operators, so fix a point p ∈ M and an open neighborhood U ∋ p
such that we have the symplectic frames {ui , vi }m m ∗
i=1 of A and {µi , ξi }i=1 of A . From here we can proceed
as follows in [6] on page 8. 
We then extend this result to the entire space of forms using the following definition
17
Definition 25. We define the “counting operator” H : Ω∗A (M ) → Ω∗A (M ) as the map
X
H= (k − m)πk
k

Where πk : Ω∗A (M ) → ΩkA (M ) is the canonical projection

Proposition 3.5. Given the operators L, Λ, and H, we have the following sl2 -commutation relations
[L, Λ] = H [H, L] = 2L [H, Λ] = −2Λ
Proof. The first identity is the result of the lemma 3.2, so let’s check the second one on α ∈ ΩkA (M )
[H, L]α = H(Lα) − L((k − m)α)
= (k + 2 − m)Lα − (k − m)Lα
= 2Lα
The third identity follows by a similar argument
[H, Λ]α = H(Λα) − Λ(Hα)
= (k − 2 − m)Λα − (k − m)Λα
= −2Λα


Lemma 3.3. We have the commutation relations

(1) [L, dA ] = 0
(2) [H, dA ] = dA
(3) [H, d∗A ] = −d∗A
(4) [Λ, d∗A ] = 0
(5) [L, d∗A ] = dA
(6) [Λ, dA ] = d∗A
Proof. We prove these in order for α ∈ ΩkA (M )
(1) Using that ω is closed we have
[L, dA ]α = ω ∧ dA α − dA (ω ∧ α)
= ω ∧ dA α − (dA ω ∧ α + ω ∧ dA α)
=0
(2) Using only degree considerations we have
[H, dA ]α = HdA α − dA Hα = [(k + 1) − n]dA α − (k − n)dA α = dA α
(3) Again using only degree considerations we have that
[H, d⋆A ] = Hd⋆A α − d⋆A Hα = [(k − 1) − n]d⋆A α − (k − n)d⋆A α = −d⋆A α
(4) Using the star form of the operators we have
[Λ, d⋆A ]α = (−1)k+1 ⋆ [L, d] ⋆ α = 0
(5) Using the Jacobi identity we have
[L, d⋆A ] + [dA , [L, Λ]] + [Λ, [dA , L]] = 0
Now we know the last term on the right hand side vanishes as dA and L commute, and using the
Kähler-Weil identity we have
[L, d⋆A ] = [H, dA ] = dA
18
 (6) Using the Jacobi identity along with the previous work we have that
[Λ, [L, d⋆A ]] + [d⋆A , [Λ, L]] + [L, [d⋆A , Λ]] = 0
[Λ, d] + [d⋆A , −H] + 0 = 0
Giving us that
[Λ, d] = d⋆A

Proposition 3.6. The space of A-forms of a Symplectic Lie Algebroid of fiber dimension 2m with the
operators L, Λ, H, dA , d⋆A is an object of Cm under the assignment
e 7→ L f 7→ Λ h 7→ H d 7→ dA δ 7→ d⋆A
Proof. The commutation relations follow proposition 3.5 and lemma 3.3 
Theorem 3.4. Suppose (A → M, ρ, ω) is a Symplectic Lie Algebroid of fiber dimension 2m. Then the
following statements are equivalent:
(1) The inclusion (ker(d⋆A ), dA ) ֒→ (Ω∗A (M ), d) is a quasi-isomorphism
m−k m+k
(2) For all 0 ≤ k ≤ m, the map [L]k : HA (M ) → HA (M ) is an isomorphism
⋆
(3) (A → M, ρ, ω) satisfies the A-dd -lemma
(4) (A → M, ρ, ω) satisfies the A-d d-lemma
where the first three statements hold equivalence weakly, i.e. in degree as in the statement of Theorem 2.14
Proof. This is the statement of Theorem 2.21 under the representation 
Definition 26. We say that a Symplectic Lie Algebroid (A, ρ, ω) is a “Lefschetz Algebroid” if any of the four
conditions in the preceding theorem hold
∗
Corollary 3.4.1. For any Lefschetz Algebroid (A → M, ρ, ω), the space HA (M ) is an sl2 (R)-module
Proof. This follows from Corollary 2.12.1 
Corollary 3.4.2. The deRham-Chevalley-Eilenberg complex of a Lefschetz Lie Algebroid is formal
Proof. Condition 1 in Theorem 3.4 shows that a Lefschetz Lie Algebroid satisfies the criteria for the in Manin
[12], from which the result follows. 
We also have the weak version of this theorem
Theorem 3.5. Suppose (A → M, ρ, ω) is a Symplectic Lie Algebroid of fiber dimension 2m, and 0 ≤ s ≤ m.
Then the following statements are equivalent:
(1) The inclusion (ker(d⋆A ), dA ) ֒→ (Ω•A (M ), d) is a quasi-isomorphism in degrees ≤ s or ≥ s + m
m−k m+k
(2) The map [L]k : HA (M ) → HA (M ) is an isomorphism for k ≥ s
⋆
(3) The A-dA dA -lemma holds up to degree s
Proof. This is the statement of Theorem 2.14 under the representation. 
3.3. Cohomology with Coefficients. This whole story works as well for Lie Algebroid cohomology with
coefficients in a representation of the Lie Algebroid. This extension was originally pointed out in [25] for
traditional vector bundles with connections.
Definition 27. Let A → M be a Lie Algebroid over a smooth manifold M . An “A-connection” on a vector
bundle E → M is an R-bilinear map
A
∇ : ΓA × ΓE → ΓE
(X, s) 7→ ∇X s
19
Which is C ∞ (M ) linear in the first entry and satisfies
A
∇X (f s) = f A ∇X s + Lρ(X) (f )s
A
The “curvature” of the connection ∇ is the End(E)-valued algebroid 2-form
R∇ (X, Y ) = A ∇X A ∇Y − A ∇Y A ∇X − A ∇[X,Y ]
A
We say that ∇ is “flat” if R∇ ≡ 0
Definition 28. Let A → M be a Lie Algebroid over a smooth manifold. A “representation” of A is a vector
bundle E equipped with a flat A-connection.
Using the connection we can create the space of forms with coefficients in a representation which inherits
a grading from ΩkA (M ) and comes with a differential given by the covariant derivative
Definition 29. Let A → M be a Lie Algebroid and E be a representation of A. Then the space of “Algebroid
k-forms with coefficients in E” is defined as
k
^
ΩkA (M, E) = Γ( A∗ ⊗ E)
We can assemble these into the differential graded module of all forms
M
Ω∗A (M, E) = ΩkA (M, E)
k≥0

With differential A ∇ : ΩkA (M, E) → Ωk+1

A (M, E) given by the induced covariant derivative, i.e. we extend
A
∇ : Ω0A (M, E) → Ω1A (M, E) via enforcing
A
∇(α ⊗ s) := (dA α) ⊗ s + (−1)|α| α ⊗ (A ∇s)
For α ⊗ s ∈ ΩkA (M, E). Since the connection is flat, this is a cochain complex.
Proposition 3.7. Let (A, ρ, ω) be a Symplectic Lie Algebroid of fiber dimension 2m and E be a representation
of A. Then we have operators
L∇ : ΩkA (M, E) → Ωk+2
A (M, E)
α ⊗ e 7→ (ω ∧ α) ⊗ e = (Lα) ⊗ e
k−2
Λ∇ : ΩkA (M, E) → ΩA (M, E)
α ⊗ e 7→ (ιω−1 α) ⊗ e = (Λα) ⊗ e
H∇ : ΩkA (M, E) → ΩkA (M, E)
α ⊗ e 7→ ((k − m)α) ⊗ e = (Hα) ⊗ e
A ∗ k−1
∇ : ΩkA (M, E) → ΩA (M, E)
α ⊗ e 7→ [Λ∇ , A ∇](α ⊗ e)
A
These together with ∇ satisfy the relations of g, and as such the space of forms with coefficients are an
object of Cn .
Proof. The proof follows similarly to that of Lemma 3.3 
Definition 30. Let (A, ρ, M ) be a Symplectic Lie Algebroid of fiber dimension 2m, E → M be a repre-
sentation of A with connection A ∇, and α ∈ Ω∗A (M, E). We say that α is “ A ∇-Symplectic Harmonic” if
A
∇α = A ∇∗ α = 0.
Corollary 3.5.1. Let (A → M, ρ, ω) be a Symplectic Lie Algebroid of fiber dimension 2m, with representation
(E, ∇). Then the following are equivalent:
m−k m+k
(1) The map [L]k : HA (M, E) → HA (M, E) given by [ω]k ∧ − is an isomorphism for all k
(2) The inclusion of complexes (ker( ∇ ), ∇) ֒→ (Ω•A (M, E), A ∇) is a quasi-isomorphism
A ∗ A

20
 (3) The (A, E)-dδ-lemma holds for A
(4) The (A, E)-d d-lemma holds for A
and the first three statements hold equivalence weakly, i.e. in degree as in the statement of Theorem 2.14.
The above corollary is the statement of Theorems 2.14 and 2.21. Taking E to be the trivial line bundle
recovers Theorem 3.4, however it would be interesting to look at what this theorem means for the choice
top
^ top
^
E = QA = A∗ ⊗ T ∗M
As defined in [3] as this always admits the structure of a representation of A.

4. Examples
4.1. Kodaria-Thurston Manifold. One of the first examples of a symplectic manifold that does not admit
a Kähler structure is given by the “Kodaira-Thurston” manifold KT 4 which is constructed as follows. Take
R4 with coordinates (x1 , x2 , x3 , x4 ) and choose a, b, c, d ∈ Z. Then consider the quotient of R4 by identifying
(x1 , x2 , x3 , x4 ) ∼ (x1 + a, x2 + b, x3 + c, x4 + d − bx3 )
The resulting quotient is the connected smooth manifold KT 4. We can see that we have a basis of one forms
{ei }4i=1 given by
e1 = dx1 e2 = dx2 e3 = dx3 e4 = dx4 + x2 dx3
We have a symplectic structure on KT 4 given by ω = e1 ∧ e2 + e3 ∧ e4 , and the deRham cohomology of KT 4
has given by generators
0
HdR (KT 4 ) = h[1]i 1
HdR (KT 4 ) = h[e1 ], [e2 ], [e3 ]i
2
HdR (KT 4 ) = hω, [e12 − e34 ], [e13 ], [e24 ]i 3
HdR (KT 4 ) = h[ω ∧ e1 ], [ω ∧ e2 ], [ω ∧ e4 ]i
4 1
HdR (KT 4 ) = h[ ω 2 ]i
2
Now we can consider the Lefschetz maps
1
[L] : HdR (KT 4 ) → HdR
3
(KT 4 ) 0
[L]2 : HdR (KT 4 ) → HdR
4
(KT 4 )
The map [L]2 is clearly an isomorphism, but [L] is not as it has nontrivial kernel, and doesn’t surject either.
This satisfies the criteria of Theorem 3.4 of degree s = 1 where we take A = T M , the identity as the anchor,
and the differential as the classical deRham differential.

4.2. Symplectic Lie Algebras of Dimension Four. If we consider Symplectic Lie Algebras as Symplectic
Lie Algebroids over the one-point space, the Lie Algebroid Cohomology is exactly the Chevalley-Eilenberg
cohomology, and our theorem can be checked by examining the surjectivity of the maps [L]k : H m−k (g) →
H m+k (g). For clarification sake by a Symplectic Lie Algebra we mean a (real) Lie Algebra g equipped with
V2 ∗
a nondegenerate ω ∈ g that is closed under the Chevalley-Elienberg differential. The 2-dimensional case
has only two isomorphism classes; the abelian one which satisfies Theorem 3.5 and the other one which does
not. Things get more interesting in the 4-dimensional case, where the number of symplectomorphism classes
grows significantly. In these case we need to check the following two maps in cohomology in the diagram:

[L]2

[L]1

0 H 0 (g) H 1 (g) H 2 (g) H 3 (g) H 4 (g) 0

A result by Ovando [22] gives a classification of all 4-dimensional Real Solvable Symplectic Lie Algebras up
to symplectomorphism. Recomputing the cohomology of these classes we can check by hand which maps
[L]i are isomorphisms, surjections, or neither. Using the notation of Ovando and numbering for different
21
symplectic forms, we can check the maps in cohomology which results in the following table where s denotes
the s as in Theorem 3.5, i.e. s = 2 is the Lefschetz condition.

Lie Algebra s [L] [L]2

rh3 1 Neither Isomorphism
rr3,0 0 Surjection Surjection
rr3,−1 2 Isomorphism Isomorphism
rr′3,0 2 Isomorphism Isomorphism
r2 r2 0 Surjection Surjection
′
r2 1 0 Isomorphism
n4 1 0 Isomorphism
r4,0 0 Surjection Surjection
r4,−1 0 0 Surjection
r4,−1,β , β ∈ (−1, 0) 0 Isomorphism Surjection
r4,−1,−1 0 Neither Surjection
r4,α,−α , α ∈ (−1, 0) 0 Surjection Surjection
r′4,0,δ , δ > 0 0 Isomorphism Surjection
d4,1 0 Surjection Surjection
d4,2 , ω1 0 0 Surjection
d4,2 , ω2 0 Isomorphism Surjection
d4,2 , ω3 1 Isomorphism Surjection
d4,λ , λ ≥ 1/2 0 Surjection Surjection
d′4,δ , δ > 0 0 Surjection Surjection
h4 0 Surjection Surjection
Table 1. Lie Algebra Lefschetz Maps

4.3. Six-Dimensional Nilmanifolds. Nilmanifolds are homogeneous spaces N/Γ where N is a simply
connected nilpotent real Lie Group, and Γ is a lattice in N of maximal rank. Such spaces have been classified
up to isomorphism in dimension 6, and there are 34 such isomorphism classes. It is known [5] that exactly 26
of the 34 isomorphism classes admit symplectic structures. The data of such a space is typically presented
by two pieces of information: Its class, given as a 6-tuple, and its symplectic form. For example, the tuple
(0, 0, 0, 0, 12, 14 + 25) with symplectic form 13 + 26 + 45 conveys the data of the Lie Algebra with dual
generators {ei }6i=1 such that
de1 = de2 = de3 = de4 = 0
de5 = e1 ∧ e2
de6 = e1 ∧ e4 + e2 ∧ e5
with symplectic form ω = e1 ∧ e3 + e2 ∧ e6 + e4 ∧ e5 . Only one of these classes admit a Kähler structure,
and the other classes have provided interesting (counter)-examples in Symplectic Geometry. The fact that
the Lie Algebra data is sufficient to classify these is evident in the following theorem:
•
Theorem (Nomizu
V• ∗ [21]). The deRham complex Ω (M•) of a nilmanifold M = N/Γ is quasi-isomorphic to
the complex n of left-invariant forms on N , thus HdR (M ) ∼ •
= HCE (n).
Examining the maps [L]k : H 3−k (n) → H 3+k (n) where [α] 7→ [ω k ∧ α] for k = 1, 2, 3 allows to check
when Theorem 3.5 holds. The first of these maps [L]3 is assured to be an isomorphism by compactness
of the nilmanifold, and the remaining two are explicitly calculable given the standard data of a symplectic
nilmanifold described above. By Poincareé Duality, the maps [L]k are between vector spaces of the same
finite dimension, and so being surjective is equivalent to being an isomorphism. Only one class will admit all
22
 Signature Symplectic Form [L] [L]2 [L]3
(0,0,12,13,14,15) 16+34-25 No No Yes
(0,0,12,13,14,23+15) 16+24+34-26 No No Yes
(0,0,12,13,23,14) 15+24+34-26 Yes No Yes
(0,0,12,13,23,14-25) 15+24-35+16 Yes No Yes
(0,0,12,13,23,14+25) 15+24+35+16 Yes No Yes
(0,0,12,13,14+23,24+15) 16 + 2 × 34 − 25 No No Yes
(0,0,0,12,13,14+23) 16 − 2 × 34 − 25 No No Yes
(0,0,0,12,13,24) 26+14+35 No No Yes
(0,0,0,12,13,14) 16+24+35 No No Yes
(0,0,0,12,13,23) 15+24+36 No No Yes
(0,0,0,12,14,15+23) 13+26-45 No No Yes
(0,0,0,12,14+15+23+24) 13+26-45 No No Yes
(0,0,0,12,14,15+24) 13+26-45 No No Yes
(0,0,0,12,14,15) 13+26-45 No No Yes
(0,0,0,12,14,13+42) 15+26+34 No No Yes
(0,0,0,12,14,23+24) 16-34+25 No No Yes
(0,0,0,12,14,15+34) 16+35+24 No No Yes
(0,0,0,12,14+23,13+42) 15 + 2 × 26 + 34 No No Yes
(0,0,0,0,12,15) 16+25+34 No No Yes
(0,0,0,0,12,14+25) 13+26+45 No No Yes
(0,0,0,0,12,14+23) 3+26+45 No No Yes
(0,0,0,0,12,34) 15+36+24 No No Yes
(0,0,0,0,12,13) 16+25+34 No No Yes
(0,0,0,0,13+42, 14+23) 16+25+34 No No Yes
(0,0,0,0,0,12) 16+23+45 No No Yes
(0,0,0,0,0,0) 12+34+56 Yes Yes Yes
Table 2. Six-Dimensional Nilmanifold Lefschetz Maps

three isomorphisms, as in this setting all three maps being isomorphisms is equivalent to admitting Kähler
structure [1]. Table 2 lists whether or not a given nilmanifold class has maps [L]k that are isomorphisms; Yes
indicating it’s an isomorphism, and No indicating it’s not an isomorphism.

4.4. E-manifolds. E-manifolds are a generalization of a collection of different geometric settings: b-manifolds,
bm -manifolds, c-manifolds, and Regular Foliations. Miranda and Scott have thoroughly explored the coho-
mology and symplectic geometry of E-manifolds in many settings, e.g. [19]. The definition given in the above
paper of Miranda and Scott of an E-manifold is restated here for convenience.

Definition 31. Let E be a locally free submodule of the C ∞ module X(M ) of vector fields on M . By the
Serre-Swan theorem, there is an E-tangent
∗ bundle E T M whose sections (locally) are
sections of E, and an
E-cotangent bundle T M := T M . We will call the global sections of Λp E T ∗ M E-forms of degree p,
E ∗ E

and denote the space of all such sections by E Ωp (M ). If E satisfies the involutivity condition [E, E] ⊆ E,
there is a differential d : E Ωp (M ) → E Ωp+1 (M ) given by
  
E

X
dη V0 , . . . , Vp = (−1)i Vi η V0 , . . . , V̂i , . . . , Vp
i
X   
+ (−1)i+j η Vi , Vj , V0 , . . . , V̂i , . . . , V̂j , . . . , Vp .
i<j
23
 The cohomology of this complex is known as E-cohomology and is the same as Lie Algebroid Cohomology
if we view this as a Lie Algebroid with anchor given by inclusion. If we have an E-closed nondegenerate
E-2-form ω, we call this an E-symplectic form and the triple (E, M, ω) an E-symplectic manifold. These
are Symplectic Lie Algebroids. The cohomology of such an object in general is rather difficult to compute,
however Miranda and Scott showed the following
Theorem 4.1 (Miranda and Scott [19]). Let M = R2 , and let E be the involutive subbundle of X(R2 )
generated by
v1 = x∂x + y∂y
v2 = −y∂x + x∂y
With dual bundle generated by
xdx + ydy
v1∗ =
x2 + y 2
−ydx + xdy
v2∗ =
x2 + y 2
Then the E-Cohomology or Lie Algebroid Cohomology is given by


 i = 0, 2
R
E k 2
H (M ) = R i=1


0 i≥3

Theorem 4.2. The Symplectic Lie Algebroid defined by Miranda and Scott (E → R2 , ρ = id, ω = v1∗ ∧ v2∗ )
satisfies theorem 3.4
Proof. We only need to check that the map [L] : E H 0 (R2 ) → E H 2 (R2 ) is an isomorphism, which amounts
to showing that ω = v1∗ ∧ v2∗ is not an exact form. Suppose towards a contradiction that it is, i.e. there are
f, g ∈ C ∞ (R2 ) such that ω = E d(f v1∗ + gv2∗ ). Unfolding the definition this implies that
ω = (v1 (g) − v2 (f ))ω
2
Which means for all (x, y) ∈ R we have that
v1 (g) − v2 (f ) = 1
Or equivalently    
∂g ∂f ∂g ∂f
− x+ + y=1
∂x ∂y ∂x ∂x
However at (0, 0) this fails, so ω is not exact 
4.5. Kähler Lie Algebroids. There has been recent attention towards Kähler Lie Algebroids, see [9]. By
the Hard Lefschetz Theorem, Kälher manifolds are always Lefschetz manifolds. This relationship still holds
true for a suitable class of Kähler Lie Algebroids
Definition 32. A “Kähler Lie Algebroid” (A → M, ρ, g) is a Lie Algebroid with a Hermitian metric g whose
associated canonical (1, 1)-A-form ω ∈ Ω1,1
A (M ) is closed

Corollary 4.2.1. Kähler Lie Algebroids are Symplectic Lie Algebroids

Throughout this section, let J be the associated complex structure for A. Reconsidering the construction
of d⋆A with the presence of a complex structure reveals that
Jd⋆A J −1 = d†A
Where d†A is the hermitian adjoint to dA , for details see [6].
24
Theorem 4.3. Kähler Lie Algebroids which admit harmonic representatives for every A-cohomology class
in the Riemannian sense are Lefschetz Lie Algebroids
Proof. It suffices to show that every cohomology class admits a symplectic harmonic representative. Fix any
p,q
class [α] ∈ HA (M, C), where we’ve chosen α to be a harmonic representative. Then dA α = d†A α = 0. By
our above observations we have
0 = d†A α = Jd⋆A J −1 α
p−q ⋆
Implying that (−1) dA α = 0, so α is symplectic harmonic 

References
[1] Chal Benson and Carolyn S. Gordon, Kähler structures on compact solvmanifolds, Proc. Amer. Math. Soc. 108 (1990),
no. 4, 971–980. MR993739
[2] Jean-Luc Brylinski, A Differential Complex for Poisson Manifolds, Journal of Differential Geometry 28 (1988), no. 1, 93
–114.
[3] Sam Evens, Jiang-Hua Lu, and Alan Weinstein, Transverse measures, the modular class and a cohomology pairing for Lie
algebroids, Quart. J. Math. Oxford Ser. (2) 50 (1999), no. 200, 417–436. MR1726784
[4] Marisa Fernández, Vicente Muñoz, and Luis Ugarte, The dδ–lemma for weakly lefschetz symplectic manifolds, 2005.
[5] Michel Goze and Yusupdjan Khakimdjanov, Nilpotent lie algebras, Springer, London, 1996.
[6] Victor Guillemin, Hodge theory, 1997. Available at https://math.mit.edu/~vwg/shlomo-notes.pdf.
[7] , Symplectic hodge theory and the dδ-lemma, 2001.
[8] Chung-I Ho, Topological Methods in Symplectic Geometry, ProQuest LLC, Ann Arbor, MI, 2011. Thesis (Ph.D.)–University
of Minnesota. MR2912247
[9] Tengzhou Hu, A Generalized Kodaira Vanishing Theorem for Lie Algebroids and a Riemann-Roch Theorem for Singular
Foliation, ProQuest LLC, Ann Arbor, MI, 2023. Thesis (Ph.D.)–Washington University in St. Louis. MR4722602
[10] Yi Lin, Yiannis Loizides, Reyer Sjamaar, and Yanli Song, Symplectic reduction and a Darboux-Moser-Weinstein theorem
for Lie algebroids, Pure Appl. Math. Q. 19 (2023), no. 4, 2067–2131. MR4671391
[11] Yi Lin and Reyer Sjamaar, Equivariant symplectic Hodge theory and the dG δ-lemma, J. Symplectic Geom. 2 (2004), no. 2,
267–278. MR2108377
[12] Yu.Ĩ. Manin, Three constructions of Frobenius manifolds: a comparative study, Asian J. Math. 3 (1999), no. 1, 179–220.
Sir Michael Atiyah: a great mathematician of the twentieth century. MR1701927
[13] Eduardo Martínez, Lagrangian mechanics on Lie algebroids, Acta Appl. Math. 67 (2001), no. 3, 295–320. MR1861135
[14] Olivier Mathieu, Harmonic cohomology classes of symplectic manifolds, Commentarii Mathematici Helvetici 70
(1995/12/01), no. 1, 1–9.
[15] , Homologies associated with Poisson structures, Deformation theory and symplectic geometry (Ascona, 1996), 1997,
pp. 177–199. MR1480723
[16] Anastasia Matveeva and Eva Miranda, Reduction theory for singular symplectic manifolds and singular forms on moduli
spaces, Adv. Math. 428 (2023), Paper No. 109161, 42. MR4603782
[17] Eckhard Meinrenkin, Lie groupoids and lie algebroids lecture notes, fall 2017, University of Toronto, 2017.
[18] S. A. Merkulov, Formality of canonical symplectic complexes and frobenius manifolds, Internat. Math. Res. Notices 14
(1998), 727–733. MR1637093
[19] Eva Miranda and Geoffrey Scott, The geometry of E-manifolds, Rev. Mat. Iberoam. 37 (2021), no. 3, 1207–1224.
MR4236806
[20] Ryszard Nest and Boris Tsygan, Deformations of symplectic lie algebroids, deformations of holomorphic symplectic struc-
tures, and index theorems, Asian J. Math. 5 (2001), no. 4, 599–635. MR1913813
[21] Katsumi Nomizu, On the cohomology of compact homogeneous spaces of nilpotent Lie groups, Ann. of Math. (2) 59 (1954),
531–538. MR64057
[22] Gabriela Ovando, Four dimensional symplectic Lie algebras, Beiträge Algebra Geom. 47 (2006), no. 2, 419–434. MR2307912
[23] Jean Pradines, Théorie de Lie pour les groupoïdes différentiables. Calcul différenetiel dans la catégorie des groupoïdes
infinitésimaux, C. R. Acad. Sci. Paris Sér. A-B 264 (1967), A245–A248. MR216409
[24] Li-Sheng Tseng and Shing-Tung Yau, Cohomology and Hodge theory on symplectic manifolds: II, J. Differential Geom. 91
(2012), no. 3, 417–443. MR2981844
[25] Dong Yan, Hodge structure on symplectic manifolds, Adv. Math. 120 (1996), no. 1, 143–154. MR1392276

25