Hodge theory tells us that the cohomology of a compact Kähler manifold is a pure Hodge structure. What if we consider an algebraic variety (over
) in general? The question whether its cohomology should have a pure Hodge structure has negative answer. Consider a nodal elliptic curve
.
It turns out that
, hence it does not admit a complex structure and, in particular, is not a pure Hodge structure of weight 1. If we resolve the singularity, we get
with exceptional divisor
such that
and we have a long exact sequence of relative homology groups

and the relative exact sequence in cohomology

Since

, we can see

as the extension of

and

, which are, respectively, pure Hodge structures of weight 0 and 1. More precisely, we have an increasing filtration

such that the graded piece

is a pure Hodge structure of weight 1.
Definition 11.4
A mixed Hodge structure
consists of a free abelian group
together with an increasing filtration of 

and a dicreasing filtration of


such that

defines a pure Hodge structure of weight

on the graded piece

A morphism of mixed Hodge structures is a
-linear map which is compatible with the two filtrations of filtered vector spaces.
Proposition 11.1
Any morphism
of mixed Hodge structures is strict, i.e. every element of
which is in the image of
comes from
and similarly the same holds for the weight filtration.
In view of the (singular) case above, we would like to have a (canonical and functorial) result independent of resolution and compactification, which allows us to associate a mixed Hodge structure to any algebraic variety over
. Indeed, we have the following.
Theorem 11.1
Quasi-projective varieties have canonical mixed Hodge structure on their cohomology.
Moreover, we get that the filtrations are independent of the choice of resolution, and so, by Hironaka's theorem, the result is extended to any algebraic variety over
.