Burgos de Jong

arXiv:1604.00686v2 [math.AG] 9 Jul 201 Forum of Mathematics, Sigma (2018), vol.
, e1, 54 pages
doi:
1
SINGULARITIES OF THE BIEXTENSION METRIC FOR

FAMILIES OF ABELIAN VARIETIES
JOSÉ IGNACIO BURGOS GIL1 , DAVID HOLMES2 and

ROBIN DE JONG3
1
Instituto de Ciencias Matemáticas (CSIC-UAM-UCM-UCM3), Calle Nicolás
Cabrera 15, Campus UAM, Cantoblanco, 28049 Madrid, Spain; burgos@icmat.es
2
Mathematical Institute, Leiden University, PO Box 9512, 2300 RA Leiden, The
Netherlands; holmesdst@math.leidenuniv.nl
3
Mathematical Institute, Leiden University, PO Box 9512, 2300 RA Leiden, The
Netherlands; rdejong@math.leidenuniv.nl
Abstract
In this paper we study the singularities of the invariant metric of the Poincaré bundle over a
family of abelian varieties and their duals over a base of arbitrary dimension. As an application
of this study we prove the effectiveness of the height jump divisors for families of pointed abelian
varieties. The effectiveness of the height jump divisor was conjectured by Hain in the more general
case of variations of polarized Hodge structures of weight −1.
2010 Mathematics Subject Classification: 14H10 (primary); 11G50; 14D07
1. Introduction
1.1. Families of curves. By way of motivation of the general results in this

paper, consider the following situation. Let X be a smooth complex algebraic
variety of dimension n, and let π : Y → X be a family of smooth projective curves
parametrized by X. Let A, B be two relative degree zero divisors on Y → X, with
disjoint support. To these divisors we can associate a function h : X → R, given
by the archimedean component of the Néron height pairing
h(x) = hA x , B x i∞ ,
c The Author(s) 2018. This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence
(<http://creativecommons.org/licenses/by/3.0/>), which permits unrestricted re-use, distribution, and reproduction in any medium,
provided the original work is properly cited.
J. I. Burgos Gil, D. Holmes and R. de Jong 2
where x ∈ X. Let X ֒→ X be a smooth compactification of X with D = X \ X

a normal crossings divisor. We are interested in the behavior of the function
h close to the boundary divisor D. As is customary to do, we assume that
the monodromy operators on the homology of the fibers of Y → X about all
∼
irreducible components of D are unipotent. Let x0 be a point of X, and U − → ∆n
a small enough coordinate neighborhood of x0 such that D ∩ U is given by
q1 · · · qk = 0. Thanks to a result of Brosnan and Pearlstein [6] (see also [16]
[19] for the case where X has dimension 1), there exist a continuous function
h0 : U \ Dsing → R and rational numbers f1 , . . . , fk such that on U \ D the equality
X
k
h(q1 , . . . , qn) = h0 (q1 , . . . , qn ) − fi log |qi | (1.1)
i=1
holds. Since h0 is continuous on U \ Dsing , this determines the behavior of h

close to the smooth points of D. The question remains what happens when we
approach a point of Dsing . In other words, what kind of singularities may h0 have
on Dsing ?
From work by G. Pearlstein [21] we find a more precise statement. Let
x0 ∈ X be as above. Then there exists a homogeneous weight one function
f ∈ Q(x1 , . . . , xk ) such that the following holds. Consider a holomorphic test
curve φ : C → X that has image not contained in D, a point 0 ∈ C such that
φ(0) = x0 , and a local analytic coordinate t for C close to 0. Assume that φ is
given locally by

t 7→ tm1 u1 (t), . . . , tmk uk (t), qk+1(t), . . . , qn (t) ,
where m1 , . . . , mk are non-negative integers, u1 , . . . , uk are invertible functions
and qk+1 , . . . , qn are arbitrary holomorphic functions. Then the asymptotic
estimate
h(φ(t)) = b′ (t) − f (m1 , . . . , mk ) log |t| (1.2)
holds in a neighborhood of 0 ∈ C. Here b′ is a continuous function that extends
continuously over 0.
Naively one might expect that the function f is linear and f (m1 , . . . , mk )
is just a linear combination of the numbers fi with coefficients given by the
multiplicities mi of the curve C. In general, however this turns out not to be
the case. Examples of non-linear f can be found in [3] and [8]. In [1], [3]
and [16] one finds a combinatorial interpretation of the function f in terms of
potential theory on the dual graphs of stable curves.
As a special case of one of the main results of this paper we will have a
stronger asymptotic estimate. Namely
h(q1, . . . , qn) = b(q1 , . . . , qn ) + f (− log |q1 |, . . . , − log |qk |)
Singularities of the biextension metric for abelian varieties 3
on U \ D, where b : U \ D → R is a bounded continuous function that extends

in a continuous manner over U \ Dsing . The boundedness of b can be seen as
a uniformity property on the asymptotic estimates for different test curves. In
general, as shown by Example 3.3 below, the function b can not be extended
continuously to Dsing , thus the boundedness of b is the strongest estimate that
can be hoped for.
As a concrete example of the shape of the function f , consider the stable
curve Y0 obtained by glueing two projective lines at zero and infinity, and
marking the point (1 : 1) in both components. Let Y → X be a versal deformation
of Y0 . The locus in X where the morphism Y → X is not smooth is a normal
crossings divisor, locally defined by q1 q2 = 0, say. The examples in [3] and
[8] show that the function f (x1 , x2 ) is given, up to linear forms in x1 and x2 , by
x1 x2 /(x1 + x2 ).
One may ask for further properties of h. For example, a result of T. Hayama
and G. Pearlstein [15, Theorem 1.18] implies that h is locally integrable. Another
question is whether the same can be said about the forms ∂h and ∂∂h ¯ and their
powers. As seen in [8] in a case where X is two-dimensional this may lead to
interesting intersection numbers between infinite towers of divisors. We plan
to address this question in full generality in a subsequent work. In this paper
we will focus on the one-dimensional case because it is the only case needed to
treat Conjecture 1.2 below. Thus assume that the dimension of X is one. Let
h0 be the function appearing in equation (1.1). Then we prove that the 1-form
∂h0 is locally integrable on U with zero residue. Also the 2-form ∂∂h ¯ 0 is locally
integrable on U.
1.2. Admissible variations of mixed Hodge structures. The correct general
setting for approaching these issues is to consider a variation of polarized pure
Hodge structures H of weight −1 over X, see for instance [13] and [14]. Let
H∨ be the dual variation. Let J(H) → X and J(H∨ ) → X be the corresponding
families of intermediate jacobians. Then on J(H) × J(H ∨ ) one has a Poincaré
X
(biextension) bundle P = P(H) with its canonical (biextension) metric. The
polarization induces an isogeny of complex tori λ : J(H) → J(H ∨ ). Let
ν, µ : X → J(H) be two sections (with good behavior near D, more precisely
admissible normal functions). Then we define
L = Pν,µ = (ν, λµ)∗ P ,

def
as a metrized analytic line bundle on X. We put Pν = Pν,ν . This “diagonal”

case will be of special interest to us. One important example, discussed at length
V
in [14] and [6], is given by the normal function in J( 3 H1 (Yx )) = H3 (J(Yx ))
associated to the Ceresa cycle [Yx ]−[−Yx ] in J(Yx ), for a family of curves Y → X.
A second example is provided by the sections determined by two relative

degree zero divisors A, B on a family of smooth projective curves, as above. Let
H be the variation of Hodge structure given by the homology of the fibers of the
family of curves Y → X. Then J(H) is the usual jacobian fibration associated
to Y → X. It is principally polarized in a canonical way. The divisors A, B
give rise to sections ν, µ of J(H) → X. The Deligne pairing associates to the
line bundles OY (A) and OY (B) a line bundle hA, Bi on X, in a functorial and
bi-multiplicative way, see [10]. The line bundle hA, Bi comes with a canonical
rational section sA,B , as well as a canonical hermitian metric ||·||A,B . The metric
on hA, Bi is determined by the archimedean height pairing. More precisely, we
have the identity

h(x) = hA x , B x i∞ = − log ||sA,B ||A,B (x)
for all x ∈ X. There is a canonical isometry
∼
hA, Bi⊗(−1) −
→ Pν,µ .
Thus the singularity near x0 of the biextension metric of the local rational section
sA,B precisely gives the singularity of the function h near x0 as discussed above.
Returning to the general set-up, the result of Brosnan and Pearlstein [6,
Theorems 24 and 79] is that some power L⊗N extends as a continuously metrized
line bundle over X \ Dsing . Here we need to impose the condition that the
monodromy operators on the fibers of H about all irreducible components of
D are unipotent. Moreover, Theorem 233 and Remark 234 of [6] provide a
canonical extension of L⊗N on X \ Dsing to an analytic line bundle over the whole
of X (though the metric will in general not extend continuously over Dsing ).
Note that if the line bundle L⊗N on X \ Dsing is algebraic, then it has a unique
extensionh to an algebraic
i line bundle on X. We denote the resulting line bundle
⊗N
on X by L , ||−|| . This extension is commonly known as the Lear extension
X
of L⊗N , though the first general proof of its existence is due to Brosnan and
Pearlstein in [6]. In order to remove the dependence on the choice of N we
will
h adopt i of Q-line bundles, and consider the Lear extension
i theh formalism
1 ⊗N
L, ||−|| = N L , ||−|| as a Q-line bundle on X.
X X
We are interested in the behavior of the biextension metric on L when we
approach a point x0 in the singular locus Dsing . Let s be a section of L = Pν,µ
on U ∩ X that corresponds to an admissible biextension variation of mixed
Hodge structures. Pearlstein [21, Theorem 5.19] has proved that there exists
a homogeneous weight one function f s ∈ Q(x1 , . . . , xk ) such that for each
holomorphic test curve φ : C → X as above the asymptotic estimate
− log ks(φ(t))k = b′ (t) − f s (m1 , . . . , mk ) log |t| (1.3)
holds in a neighborhood V of 0 ∈ C, with b′ (t) continuous on V.

Now assume that the polarized variation H is torsion-free and of type (−1, 0),
(0, −1) over X, so that the family J(H) → X is a family of polarized abelian
varieties over X. Under this assumption we are able to strengthen the result of
Pearlstein’s.
1.3. Statement of the main results. Recall that we work with a smooth
complex algebraic variety X, provided with a partial compactification X with
D = X \ X a normal crossings divisor, and a polarized pure variation of Hodge
structures H of weight −1 over X.
∼
→ ∆n be a coordinate chart on X such that D ∩ U =
Let (q1 , . . . , qn ) : U −
{q1 · · · qk = 0}. Denote by Di the local component of D with equation given by
qi = 0. For any 0 < ǫ < 1 write
Uǫ = {(q1 , . . . , qn ) ∈ U : |qi | < ǫ for all i = 1, . . . , n} .
Note that Uǫ ∩ X is identified via the coordinate chart with (∆∗ǫ )k × ∆n−k
ǫ .
Theorem 1.1. Assume that H is a variation of torsion-free polarized pure Hodge

structures of type (−1, 0), (0, −1) on X. Assume that the monodromy operators
on the fibers of H about the irreducible components of D are unipotent. Let
ν, µ : X → J(H) be two admissible normal functions of J(H) over X. Then there
exist an integer d, a homogeneous polynomial Q ∈ Z[x1 , . . . , xk ] of degree d with
no zeroes on Rk>0 and, for each section s of Pν,µ corresponding to an admissible
biextension variation of mixed Hodge structures over U ∩ X, a homogeneous
polynomial P s ∈ Z[x1 , . . . , xk ] of degree d + 1 such that the homogeneous weight
one rational function f s = P s /Q satisfies the following properties.
1. For all ǫ ∈ R>0 small enough, the function
b(q1 , . . . , qn) = − log||s|| − f s (− log |q1 |, . . . , − log |qk |)
is bounded on Uǫ ∩ X and extends continuously over Uǫ \ Dsing .
2. The function f s is uniquely determined by the previous property. Moreover,
if s′ is another section of Pν,µ over U ∩ X, such that
X
k
′
div(s /s) = a i Di , (1.4)
i=1
then the difference

X
k
f s′ − f s = ai (− log |qi |)
i=1
is linear in the functions − log |qi |.
3. The function f s : Rk>0 → R extends to a continuous function f s : Rk≥0 → R.
4. In the case that µ = ν, the function f s is convex as a function on Rk>0 and

the function f s is convex as a function on Rk≥0 .
We make a few remarks about Theorem 1.1. First of all, by [6, Theorem 81], if
U is small enough, admissible sections s as in Theorem 1.1 exist.
Next, by [6, Corollary 177], if U is small enough the set of admissible biex-
tension variations on U ∩ X is a non-empty torsor over the group of meromorphic
functions with poles only on D. Hence the admissibility of s and condition (1.4)
imply the admissibility of s′ .
Clearly, the function f s from Theorem 1.1 coincides with the f s from Pearl-
stein’s asymptotic estimate (1.3). However, we do not assume [21, Theorem 5.19]
in our proof, hence our arguments give an independent proof of (1.3) for the case
of polarized, torsion-free variations of type (−1, 0), (0, −1).
If the family J = J(H) of jacobians is algebraic, that is, J is an abelian scheme
over X, then any two algebraic sections µ and ν of J over X are admissible, and
for such µ, ν the Lear extension of Pν,µ over X is an algebraic Q-line bundle.
Let the rank of H be 2g. Our proof of Theorem 1.1 in section 4 will show
that the function f s in the theorem has the shape
Xk Xk Xk
f s (x1 , . . . , xk ) = ( xi Ai ci )t ( xi Ai )−1 ( xi Ai ci ), (1.5)
i=1 i=1 i=1
where the Ai (i = 1, . . . , k) are positive semidefinite g × g matrices such

P
that ki=1 Ai is positive definite, the ci are in Qg , and are determined by the
monodromy of µ and ν about the branches of the divisor D. Thus the singularity
of − log||s|| has the shape
f s (− log|q1 |, . . . , − log|qk |)
 k t  k −1  k 
X  X  X 
=  − log|qi |Ai ci   − log|qi |Ai   − log|qi |Ai ci  .
i=1 i=1 i=1
Finally, Example 3.3 below will show that, in general, the locus of indeterminacy
Dsing of b can not be reduced to a smaller set.
We next turn to the issue of local integrability, in dimension one. R. Hain
has made the following conjecture (see [14, Conjecture 6.4]). Assume we work
with an arbitrary polarized variation of Hodge structures (H, λ) of weight −1,
whose underlying local system of abelian groups is torsion-free, and let P be
its Poincaré bundle. Let ν be an admissible normal function of the family of
intermediate jacobians J(H) over X.
Conjecture 1.2 (Hain). Write P̂ = (id, λ)∗ P and let ω = c1 (P̂) be the first Chern
form of the pullback of the Poincaré bundle with its canonical metric. Assume
that X is a curve. Let L = Pν = ν∗ P̂ with induced metric ||−|| and let N ∈ Z>0
be hsuch that i L⊗N extends as a continuous metrized line bundle over
h X. Let
i
c1 L⊗N , ||−|| be the first Chern class of the extended line bundle L⊗N , ||−|| .
X X
Then the 2-form ν∗ ω is integrable on X, and the equality
Z Z
1
ν∗ ω = c1 L⊗N , ||−||
X N X X
holds.
∗
h ν ω =i c1 (Pν ), and that the integral on the right hand side equals
Note that
1 ⊗N
N degX L , ||−|| X . We prove the following result, which implies Hain’s con-
jecture in the case of a variation of torsion-free polarized Hodge structure of type
(−1, 0), (0, −1).
Theorem 1.3. Assume that the polarized variation H over X is torsion-free and
pure of type (−1, 0), (0, −1), and that the monodromy operators on the fibers of
H about all irreducible components of D are unipotent. Let s be a section of Pν,µ
corresponding to an admissible biextension variation of mixed Hodge structures
over U ∩ X and assume that dim X = 1. Write
− log ksk = b(z) − r log |t|
on U ∩ X with r ∈ Q and with b bounded continuous on U, as can be done by
the existence of the Lear extension of Pν,µ over X. Then the 1-form ∂b is locally
integrable on U with zero residue. Moreover the 2-form ∂∂b¯ is locally integrable
on U.
As also ∂∂¯ log |t| is locally integrable, we find that ∂∂¯ log||s|| is locally integrable.
Since moreover the 1-form ∂b has no residue on U, so that d[∂b] ¯ = [∂∂b],
¯ upon
globalizing using bump functions and applying Stokes’ theorem we find
Z Z h i
1
c1 (Pν,µ ) = c1 P⊗N ν,µ , ||−|| = deg Pν,µ , ||−|| .
X N X X X
In the diagonal case, we mention that by [14, Theorem 13.1] or [20, Theorem 8.2]
the metric on Pν is non-negative. Thus Theorem 1.3 implies that actually the
inequality h i
deg Pν , ||−|| ≥ 0 (1.6)
X
holds. We mention that in a letter to P. Griffiths, G. Pearlstein sketches a proof
of Conjecture 1.2, and hence of the inequality (1.6), without the assumption that
the type be (−1, 0), (0, −1).
We return again to the setting where the parameter space X is of any

dimension. However, we specialize to the “diagonal” case where µ = ν.
Consider, as before a test curve φ : C → X that has image not contained in
D, and a point 0 ∈ C such that φ(0) = x0 . Let φ denote the restriction of φ to
−1
C \ φ D. The Q-line bundle
h i⊗−1 ∗h i
φ∗ (Pν , ||−||) ⊗ φ Pν , ||−||
C X
−1
has a canonical non-zero rational section, as it is canonically trivial over C\φ D.
We call its divisor the height jump divisor J = Jφ,ν on C. R. Hain has made the
following conjecture (see [14, end of §14]).
Conjecture 1.4. For all holomorphic test curves φ : C → X with image not
contained in D, the height jump divisor J = Jφ,ν on C is effective.
Choose coordinates in a neighbourhood U of x0 as before so that x0 has co-

ordinates (0, . . . , 0) and let f s ∈ Q(x1 , . . . , xk ) be as in Pearlstein’s asymptotic
estimate (1.3), based on the choice of some admissible section s of Pν on U ∩ X.
It can be shown that the function f s : Rk>0 → R extends to a continuous function
f s : Rk≥0 → R. Locally around 0 the map φ can be written as
φ(t) = (tm1 u1 (t), . . . , tmk uk (t), qk+1 (t), . . . , qn(t)),

def
where, for i ∈ [1, k], mi > 0 and ui (0) , 0. Write f s,i = f s (0, . . . , 0, 1, 0, . . . , 0)
(the 1 placed in the i-th spot), then
X
k
ord0 J = − f s (m1 , . . . , mk ) + mi f s,i . (1.7)
i=1
Note that indeed ord0 J is independent of the choice of s. The rational number
ord0 J is called the “height jump” associated to the test curve φ, the admissible
normal function ν and the point 0 ∈ C.
The terminology is due to R. Hain [14], who also observed a first instance
where the height jump is non-zero. We refer the reader to the monograph [6]
by P. Brosnan and G. Pearlstein, where an extensive study of the height jump in
complete generality is given. Note that the height jumps precisely when f s is not
linear. We mention that Conjecture 1.4 about the height jump was stated in [14]
only for the normal function on Mg associated to the Ceresa cycle, but it seems
reasonable to make this broader conjecture.
In this paper we prove Conjecture 1.4 in the case of admissible normal
functions of families of polarized abelian varieties.
Theorem 1.5. Assume that the polarized variation H over the smooth complex
variety X is torsion-free and pure of type (−1, 0), (0, −1), and that the monodromy
operators on the fibers of H about all irreducible components of D are unipotent.
Let ν be an admissible normal function of the family of intermediate jacobians
J(H) over X. Then for all holomorphic test curves φ : C → X with image not
contained in D, the associated height jump divisor J = Jφ,ν on C is effective.
Combining with inequality (1.6) we obtain
Corollary 1.6. Assume that C is smooth and projective. i Then under the as-
∗h
sumptions of Theorem 1.5, the Q-line bundle φ Pν , ||−|| has non-negative de-
X
gree on C.
The key to our proof of Theorem 1.5 is the convexity of the homogeneous
function f s , as asserted in Theorem 1.1(4). We have the following explicit
expression for ord0 J. In equation (1.5) we already gave an expression for f s
and hence f s in terms of matrices Ai and vectors ci for i = 1, . . . , k. We will
see in subsection 3.4 that f s,i = cti Ai ci for i = 1, . . . , k. Following the general
expression (1.7) this gives
Xk X
k X
k X
k
ord0 J = −( mi Ai ci )t ( mi Ai )−1 ( mi Ai ci ) + mi cti Ai ci
i=1 i=1 i=1 i=1
for the height jump in our setting.

Turning again to the case of the Ceresa cycle, note that since the intermediate
Jacobian of the primitive part of H3 (J(Yx )) is a compact complex torus but not
an abelian variety, we can not apply directly our results for families of abelian
varieties to this case.
In the special case of families of jacobians of curves Conjecture 1.4 has
been proved in [3]. The proof in this special case makes heavy use of the
combinatorics of dual graphs of nodal curves, and so cannot readily be extended
to families of abelian varieties, nor does it seem practical to reduce the general
case to that of jacobians of curves.
Remark 1.7. After the initial submission of the present paper to arxiv, two proofs
of Conjecture 1.4 have appeared, see [6] and [7].
1.4. Overview of the paper. We review the content of the different sections
of this paper. In the preliminary section 2 we start by recalling the notions of
Q-line bundle and of Lear extension, and the Poincaré bundle on the product of a
complex torus and its dual, together with its associated metric. We also recall the
explicit description of the Poincaré bundle and its metric on a family of polarized
abelian varieties. Also we study the period map associated to a family of pointed
polarized abelian varieties. Moreover we give a local expansion for the metric
on the pullback of the Poincaré bundle under this period map. The functions that
appear as the logarithm of the norm of a section of the pullback of the Poincaré
bundle will be called norm-like functions.
In section 3 we study norm-like functions and give several estimates on their
growth and that of their derivatives. Finally in section 4 we prove the main results
on local integrability and positivity of the height jump.
We fix some notation that we will use throughout. Let r be a positive
integer. For any commutative ring R we will denote by Colr (R) (respectively
Rowr (R), Mr (R) and S r (R)) the set of column vectors of size r with entries in R
(respectively row vectors, matrices and symmetric matrices of size r-by-r).
We denote by S r++ (R) ⊂ S r (R) (respectively S r+ (R) ⊂ S r (R)) the cone of
positive definite (respectively positive semidefinite) symmetric real matrices. We
denote by Hr Siegel’s upper half space of rank r, and by Pr its compact dual.
By a variety we mean an integral separated scheme of finite type over C.
2. Preliminary results
2.1. Lear extensions. We start by recalling the formalism of Q-line bundles.
Details can be found in [3, Definition 2.10].
Definition 2.1. Let X be a complex variety. An (algebraic resp. analytic) Q-

line bundle over X is a pair (L, r) where L is an (algebraic resp. analytic) line
bundle on X and r > 0 is a positive integer (informally, we think of it as
L⊗1/r ). A metrized Q-line bundle is a triple (L, ||−||, r), where (L, r) is a Q-line
bundle and ||−|| is a continuous metric on L. An isomorphism of Q-line bundles
(L1 , r1) → (L2 , r2 ) is an equivalence class of pairs (a, f ) where a is a positive
integer and f : L1⊗ar2 → L2⊗ar1 is an isomorphism, where the equivalence relation
is generated by setting (a, f ) ∼ (an, f ⊗n). An isomorphism of metrized line
bundles is an isometry if one (equivalently all) of the corresponding morphisms
of line bundles is an isometry. Every line bundle L gives rise to a Q-line
bundle (L, 1). Note that, if L is a line bundle and r > 1 is an integer, then
there is a canonical isomorphism (L⊗r , r) ≃ (L, 1). Moreover, if L is a torsion
line bundle so that L⊗r ≃ OX , then there is an isomorphism of Q-line bundles
(L, 1) → (OX , r). If we do not need to specify the multiplicity r, a Q-line bundle
will be denoted by a single letter. We note that the group of isomorphism classes
of Q-line bundles on X is equal to Pic(X) ⊗Z Q.
We denote
RatQ (X) = (O(X) \ {0}, ×) ⊗ Q.
If (L, r) is a Q-line bundle, a Q-rational section of (L, r) (or rational section

1
for short) is an equivalence class of symbols s rd , where s is a non-zero rational
section of L⊗d . Two symbols s11/rd1 and s21/rd2 are equivalent if
(s1 )⊗d2 = (s2 )⊗d1
as a section of L⊗d1 d2 . The space of rational sections of (L, r) is a torsor over

RatQ (X). Moreover, if s and s′ are rational sections of (L, r) and (L′ , r′ ) then
′
s ⊗ s′ is a rational section of (L⊗r ⊗ (L′ )⊗r , rr′ ), but there is no additive structure
of rational sections.
1
The divisor of the section s rd is
1 1
div(s rd ) = div(s).
rd
Definition 2.2 (Lear extension). Let X ⊆ X be an open immersion of smooth

def
complex varieties, such that the boundary divisor D = X\X has normal crossings,
and L a line bundle on X with continuous metric ||−||. A Lear extension of L is
a Q-line bundle (L, r) on X together with an isomorphism α : (L, 1) → (L, r)|X
and a continuous metric on L|X\Dsing such that the isomorphism α is an isometry.
Since Dsing has codimension at least 2 in X, if a Lear extension exists then it is
unique
h upito a unique isomorphism. If a Lear extension of L exists we denote it
by L, ||−|| . Note that the isomorphism class of the Lear extension of L depends
X
not only on L but also on the metric on L.
1
If s is a rational hsectioni of L, writing s = (s⊗r ) r , it can also be seen as
a rational section of L, ||−|| . We will denote by divX (s) the divisor of s as
X
i section of L and by divX (s) the divisor of s as a rational section of
ha rational
L, ||−|| .
X
2.2. Poincaré bundle and its metric. In this section we recall the definition
of the Poincaré bundle and its biextension metric. Moreover we make the
biextension metric explicit in the case of families of polarized abelian varieties.
In the literature one can find small discrepancies in the description of the
Poincaré bundle, see Remark 2.5. These discrepancies can be traced back to two
different choices of the identification of a complex torus with its bidual. More-
over, there are also different conventions regarding the sign of the polarization
of the abelian variety. Since one of our main results is a positivity result it is
worthwhile to fix all the signs to avoid these ambiguities.
Complex tori and their duals. Let g ≥ 0 be a non-negative integer, V a g-
dimensional complex vector space and Λ ⊂ V a rank 2g lattice. The quotient
T = V/Λ is a compact complex torus. It is a Kähler complex manifold, but in

general it is not an algebraic variety.
We recall the construction of the dual torus of T . We denote by V ∗ =
HomC (V, C) the space of antilinear forms w : V → C. This is not the dual V ∨ of
V. In fact, let V denote the abelian group V with the complex structure · given
by
α · v = α · v.
∗ ∨
Then V = V .
The bilinear form
h·, ·i : V ∗ × V → R, hw, zi = Im(w(z))
def
is non-degenerate. Thus
Λ∨ = {λ ∈ V ∗ | hλ, Λi ⊂ Z}
def
is a lattice of V ∗ . The lattice Λ∨ is canonically isomorphic to the dual of the

lattice Λ. The quotient T ∨ = V ∗ /Λ∨ is again a compact complex torus, called
the dual torus of T .
We can identify V with HomC (V ∗ , C) by the rule
z(w) = w(z) (2.1)
so that the bilinear pairing
(V ∗ ⊕ V) ⊗ (V ∗ ⊕ V) → R, (w, z) ⊗ (w′ , z′ ) 7→ Im(w(z′ )) + Im(z(w′ ))
is antisymmetric. With this identification the double dual (T ∨ )∨ gets identified
with T .
The points of T ∨ define homologically trivial line bundles on T giving an
isomorphism of T ∨ with Pic0 (T ). We recall this construction. Let C1 denote the
subgroup of C× of elements of norm one. Let w ∈ V ∗ . Denote by [w] its class in
T ∨ and by χ[w] ∈ Hom(Λ, C1 ) the character
χ[w] (µ) = exp(2πihw, µi). (2.2)
The line bundle associated to [w] is the line bundle L[w] with automorphy factor
χ[w] . In other words, consider the action of Λ on V × C given by
µ(z, t) = (z + µ, t exp(2πihw, µi)).
Write L[w] = (V × C)/Λ. The projection V × C → V induces a map L[w] → T .
It is easy to check that L[w] is a holomorphic line bundle on T that only depends
on the class [w]. Note that the identification between T ∨ and Pic0 (T ) is not
completely canonical because it depends on a choice of sign. We could equally
well have used the character χ−1 [w] .
The Poincaré bundle. Note that, although the cocycle (2.2) is not holomorphic
in w, the line bundle L[w] varies holomorphically with w, defining a holomorphic
line bundle on T × T ∨ called the Poincaré bundle. See [4, § 2.5] for details.
Definition 2.3. A Poincaré (line) bundle P is a holomorphic line bundle on

T × T ∨ that satisfies
1. the restriction P|T ×{[w]} is isomorphic to L[w] ;
2. the restriction P|{0}×T ∨ is trivial.
A rigidified Poincaré bundle is a Poincaré bundle together with an isomorphism

∼
P|{0}×T ∨ −
→ O{0}×T ∨ .
To prove the existence of a Poincaré bundle, consider the map
aP : (Λ × Λ∨ ) × (V × V ∗ ) → C×
given by

aP ((µ, λ), (z, w)) = exp π (w + λ)(µ) + λ(z) . (2.3)
This map is holomorphic in z and w. Moreover, since for (µ, λ) ∈ Λ × Λ∨ ,

1
hλ, µi = (λ(µ) − λ(µ)) ∈ Z,
2i
the map aP is a cocycle for the additive action of Λ × Λ∨ on V × V ∗ . Hence, it
is an automorphy factor that defines a holomorphic line bundle P on T × T ∨ =
V × V ∗ /Λ × Λ∨ .
For a fixed w ∈ V ∗ ,
aP ((µ, 0), (z, w)) = exp(πw(µ)).
This last cocycle is equivalent to the cocycle (2.2). Indeed,
exp(πw(µ)) exp(πw(z + µ))−1 exp(πw(z)) = exp(2πihw, µi),
and the function z 7→ exp(πw(z)) is holomorphic in z. Thus the restriction

P|T ×{[w]} is isomorphic to L[w] . Moreover
aP ((0, λ), (0, w)) = 1,
which implies that the restriction P|{0}×T ∨ is trivial. The uniqueness of the
Poincaré bundle follows from the seesaw principle (see [4, Appendix A]).
We conclude
Proposition 2.4. A Poincaré bundle exists and is unique up to isomorphism. A

rigidified Poincaré bundle exists and is unique up to a unique isomorphism.
Remark 2.5. Using the above identification of T with the dual torus of T ∨ we
have that, for a fixed z ∈ V, the restriction P|{[z]}×T ∨ agrees with L[z] . In fact
aP ((0, λ), (z, w)) = exp(πλ(z)),

and, arguing as in the proof of Proposition 2.4, this cocycle is equivalent to the
cocycle
exp(2πi Im(λ(z))) = exp(2πihz, λi).
Note that the definition of the Poincaré bundle in [13, §3.2] states that P|{[z]}×T ∨ =
L[−z] . The discrepancy between [13] and the current paper is due to a different
choice of identification between T and (T ∨ )∨ .
Remark 2.6. As we will see later, in equation (2.12), the cocycle (2.3) is not
optimal because it does not vary holomorphically in holomorphic families of
tori.
Group theoretical interpretation of the Poincaré bundle. We next give a group

theoretic description of the Poincaré bundle. We start with the additive real Lie
group W given by
W = V × V ∗.
e the semidirect product W
Denote by W e = W ⋉ C× , where the product in W e is
given by

(z, w), t · (z′ , w′ ), t′ = (z + z′ , w + w′ ), tt′ exp(2πihw, z′ i) . (2.4)
Clearly the group
W Z = Λ × Λ∨ (2.5)
e
is a subgroup of W.
Consider the space
P = V × V ∗ × C×
def
(2.6)
and the action of W e on P by biholomorphisms given by

(µ, λ), t · ((z, w), s) = z + µ, w + λ, ts exp(π(w + λ)(µ) + πλ(z)) . (2.7)
The projection P → V × V ∗ induces a map WZ \P → T × T ∨ . The action of C×
on P by acting on the third factor provides WZ \P with a structure of C× -bundle
over T × T ∨ . Denote by PT = (WZ \P) ×× C the associated holomorphic line
C
bundle. The structure of P as a product space induces a canonical rigidification
PT |{0}×T ∨ = O{0}×T ∨ .
Proposition 2.7. The line bundle PT is a rigidified Poincaré line bundle.

Proof. From the explicit description of the cocycle (2.3) and of the action (2.7)
we deduce that PT is a Poincaré bundle.
The metric of the Poincaré bundle. The Poincaré bundle has a metric that is
determined up to constant by the condition that its curvature form is invariant
under translation. On a rigidified Poincaré bundle, with given rigidification
∼
PT |{0}×T ∨ −
→ O{0}×T ∨ , the constant is fixed by imposing the condition k1k = 1.
We now describe explicitly this metric.
Let W e 1 = W ⋉ C1 with the product described before. Denote by P× the
T
Poincaré bundle with the zero section deleted. Since P×T = WZ \P, the invariant
metric of PT is described by the unique function k · k : P → R>0 satisfying the
conditions
1. (Norm condition) For (z, w, s) ∈ P, we have
k(z, w, s)k = |s|k(z, w, 1)k.
e 1 ) For g ∈ W
2. (Invariance under W e 1 and x ∈ P, we have
kg · xk = kxk
3. (Normalization) k(0, 0, 1)k = 1.

Using the explicit description of the action given in (2.7), we have that
(z, w, s) = (z, w, 1) · (0, 0, s exp(−πw(z))),
from which one easily derives that the previous conditions imply

k(z, w, s)k2 = |s|2 exp − π w(z) + w(z) . (2.8)
Holomorphic families of complex tori. Let X be a complex manifold and T → X

a holomorphic family of dimension g complex tori. This means that T is
defined by a holomorphic vector bundle V of rank g on X and an integral
local system Λ ⊂ V of rank 2g such that, for each s ∈ X, the fiber Λ s is a
lattice in V s and the flat sections of Λ are holomorphic sections of V. Indeed
Λ is the local system s 7→ H1 (T s , Z) and V the holomorphic vector bundle
s 7→ H1 (T s , C)/F 0 H1 (T s , C).
We now want to give to the dual family of compact tori a holomorphic
structure. That is, we want to construct a holomorphic family of compact tori
T ∨ with a canonical identification (T ∨ ) s = (T s )∨ . This construction is not
completely obvious because the vector spaces (V s )∗ vary anti-holomorphically
with s. We will use the lattice Λ to define a holomorphic structure on this family
of vector spaces.
Write HC = Λ ⊗ OX . It is a holomorphic vector bundle, with a holomorphic
surjection HC → V and an integral structure that determines a complex
conjugation in HC . The kernel F 0 = Ker(HC → V) is a holomorphic vector
bundle. For every s ∈ X, the surjection HC → V allows us to identify F 0 s with
V s , hence F s0 with V s . Let Λ∨ be the dual local system to Λ. On the dual vector
bundle H ∨ = Λ∨ ⊗ OX consider the orthogonal complement (F 0 )⊥ to F 0 . Then
(F 0 )⊥ is isomorphic with the dual vector bundle V∨ . The quotient H ∨ /(F 0 )⊥ is
a holomorphic vector bundle that we denote by V∗ . The identification F s0 = V s
gives us the equality
(V∗ ) s = (H ∨ /(F 0 )⊥ ) s = (F s0 )∨ = (V s )∨ = (V s )∗ ,
that explains the notation.

Then the dual family of tori is defined as
T ∨ = V∗ /Λ∨ .
Let U ⊂ X be a small enough open subset such that the restriction of T to U

is topologically trivial. Choose s0 ∈ U and an integral basis
(a, b) = (a1 , . . . , ag , b1, . . . , bg )
of Λ s0 such that (a1 , . . . , ag) is a complex basis of V s0 . By abuse of notation,

we denote by ai , bi , i = 1, . . . , g the corresponding flat sections of Λ. We can
see them as holomorphic sections of HC and we will also denote by ai , bi their
images in V. After shrinking U if necessary, we can assume that the sections ai
form a frame of V, thus we can write
(b1 , . . . , bg ) = (a1 , . . . , ag )Ω (2.9)
for a holomorphic map Ω : U → Mg (C). We call Ω the period matrix of the

variation on the basis (a, b). Note that condition (2.9) is equivalent to saying that
F 0 ⊂ HC is generated by the columns of the matrix
!
−Ω
.
Id
Writing HR for the real vector subbundle of HC formed by sections that are
invariant under complex conjugation, we have that F 0 ∩ HR = 0. This
implies that Im Ω is non-degenerate. The complex basis (a1 , . . . , ag ) gives us
an identification of V|U with the trivial vector bundle Colg (C) and the basis
(a, b) identifies Λ with the trivial local system Colg (Z) ⊕ Colg (Z). With these
identifications, the inclusion Λ → V is given by
(µ1 , µ2) 7→ µ = µ1 + Ωµ2 .
Let now (a , b ) = (a∗1 , . . . , a∗g , b∗1, . . . , b∗g) be the basis of Λ∨s0 dual to (a, b).
∗ ∗
As before we extend the elements a∗i , b∗i , i = 1, . . . , g to flat sections of Λ over

U. Then b∗1 , . . . , b∗g is a frame of V∗ . One can check that, on V∗ , the equality
(a∗1 , . . . , a∗g ) = −(b∗1 , . . . , b∗g)Ωt
holds. Thus if we identify V∗ with the trivial vector bundle Rowg (C) using the
basis (b∗ ) and Λ∨ with the trivial local system Rowg (Z) ⊕ Rowg (Z) using the
∗ ∗ ∨ ∗
basis (a , b ) we obtain that the inclusion Λ → V is given by
(λ1 , λ2 ) 7→ λ = −λ1 Ω + λ2 . (2.10)
∗
In the fixed bases, one can check that the pairing between V and V is given by
w(z) = −w(Im Ω)−1 z̄, (2.11)
where w ∈ Rowg (C) and z ∈ Colg (C), while the pairing between the lattice Λ
and its dual Λ∨ is given by
h(λ1 , λ2 ), (µ1 , µ2 )i = λ1 µ1 + λ2 µ2 ,
where λ1 , λ2 ∈ Rowg (Z) and µ1 , µ2 ∈ Colg (Z). Clearly the pairing between Λ
and Λ∨ has integer values.
The cocycle aP from equation (2.3) can now be written down explicitly as
aP ((µ1 , µ2), (λ1 , λ2 ), (z, w))
= exp(−π((w − λ1 Ω + λ2 )(Im Ω)−1 (µ1 + Ω̄µ2 ) + (−λ1 Ω̄ + λ2 )(Im Ω)−1 z)),
(2.12)
which is not holomorphic with respect to Ω. Thus it does not give us on the nose
a holomorphic Poincaré bundle in families. Nevertheless the construction of the
Poincaré bundle can be given a holomorphic structure.
Proposition 2.8. Let X be a complex manifold and T → X a holomorphic family
of dimension g complex tori. Let ν0 : X → T × T ∨ be the zero section. Then
X
1. the fiberwise dual tori form a holomorphic family of complex tori T ∨ → X;

2. on T × T ∨ there is a holomorphic line bundle P, together with an
X
∼
isomorphism ν∗0 P − → OX , called the rigidified Poincaré bundle, which is
unique up to a unique isomorphism, and is characterized by the property
that for every point p ∈ X, the restriction P|T p ×T p∨ is the rigidified Poincaré
bundle of T p ;
3. there is a unique metric on P that induces the trivial metric on ν∗0 P = OX

and whose curvature is fiberwise translation invariant.
Proof. Fix an open subset U ⊂ X as before. The dual family of tori T ∨ is

holomorphic by definition.
In order to prove that the Poincaré bundle defines a holomorphic line bundle
on the family we need to exhibit a new cocycle that is holomorphic in z, w
and Ω and that, for fixed Ω, is equivalent to aP holomorphically in z and w.
Write λ = −λ1 Ω + λ2 and µ = µ1 + Ωµ2 as before with λ1 , λ2 ∈ Rowg (Z) and
µ1 , µ2 ∈ Colg (Z). Consider the cocycle
bP ((λ, µ), (z, w)) = exp(2πi((w − λ1 Ω + λ2 )µ2 − λ1 z)) (2.13)
for w ∈ Rowg (C) and z ∈ Colg (C). Then bP is holomorphic in z, w, and Ω.
Consider also the function
ψ(z, w) = exp(−πw(Im Ω)−1 z), (2.14)
which is holomorphic in z and w. Since
bP ((µ, λ), (z, w)) = aP ((µ, λ), (z, w))ψ(z, w)ψ(z + µ, w + λ)−1
we deduce that the cocycle bP determines a line bundle that satisfies the prop-
erties stated in item (2) from the proposition over the open U. The uniqueness
follows again from the seesaw principle. By the uniqueness, we can glue together
the rigidified Poincaré bundles obtained in different open subsets U to obtain a
rigidified Poincaré bundle over X.
The fact that the invariant metric has invariant curvature fixes it up to a
function on X that is determined by the normalization condition. Thus if it
exists, it is unique. Since the expression for the metric in (2.8) is smooth in
Ω and the change of cocycle function in (2.14) is also smooth in Ω we obtain an
invariant metric locally. Again the uniqueness implies that we can patch together
the different local expressions.
Remark 2.9. Since the cocycle aP does not vary holomorphically in families,
the frame for the Poincaré bundle used in equation (2.8) is not holomorphic
in families. The cocycle bP and the rigidification do determine a holomorphic
frame of the Poincaré bundle over X × V × V ∗ . In this holomorphic frame the
metric is given by
k(z, w, s)k2 = |s|2 exp(−π(w(z) + w(z)))|ψ(z, w)|2
= |s|2 exp(4π Im(w)(Im Ω)−1 Im(z)), (2.15)
where ψ is the function given in (2.14).
Abelian varieties. We now specialize to the case of polarized abelian varieties.

A polarization on the torus T = V/Λ is the datum of an antisymmetric non-
degenerate bilinear form E : Λ × Λ → Z such that for all v , w ∈ V,
E(iv , iw) = E(v , w), −E(iv , v ) > 0, for v , 0.
Here we have extended E by R-bilinearly to V = Λ ⊗ R. Note that the
standard convention in the literature on abelian varieties is to ask E(iv , v ) to be
positive. But this convention is not compatible with the usual convention in the
literature on Hodge Theory. We have changed the sign here to have compatible
conventions for abelian varieties and for Hodge structures.
Since E is antisymmetric and non-degenerate we can choose an integral basis
(a, b) such that the matrix of E on (a, b) is given by
!
0 ∆
, (2.16)
−∆ 0
where ∆ is an integral diagonal matrix. We will call such basis a Q-symplectic
integral basis. From a Q-symplectic integral basis (a, b) we can construct a
symplectic rational basis (a∆−1 , b).
With the choice of a Q-symplectic integral basis, the condition E(iv , iw) =
E(v , w) is equivalent to the product matrix ∆Ω being symmetric. Thus Ωt ∆ =
∆Ω. The condition −E(iv , v ) > 0 is equivalent to ∆ Im Ω being positive definite.
This last condition is equivalent to that any of the symmetric matrices (Im Ω)t ∆,
((Im Ω)−1 )t ∆ or ∆(Im Ω)−1 is positive definite.
Recall from (2.9) that Ω ∈ Mg (C) is determined by the relation b = aΩ. The
polarization E defines a positive definite hermitian form H on V given by
H(v , w) = −E(iv , w) − iE(v , w),
so that we recover the polarization E as the restriction of − Im(H) to Λ × Λ.
In the basis (a1 , . . . , ag ) of V, the hermitian form H is given by ∆(Im Ω)−1 =
((Im Ω)−1 )t ∆. That is, under the identification V = Colg (C), we have
H(v , w) = v t ∆(Im Ω)−1 w. (2.17)
∨
The polarization defines an isogeny λE : T → T that is given by the map
V → V ∗ , v 7→ H(v , −). Under the identification V ∗ = Rowg (C) given by the
basis (b∗ ), by equations (2.11) and (2.17), we deduce that λE is given by
λE (v ) = −v t ∆. (2.18)
The fact that ∆Ω is symmetric and ∆ is integral implies that this map sends Λ to
Λ∨ defining an isogeny. The dual polarization E ∨ on V ∗ is given by the hermitian
t
form H ∨ (e, f ) = e(Im Ω)−1 ∆−1 f so that the map V → V ∗ is an isometry.
Consider now the composition of the diagonal map with the polarization map
on the second factor (id, λE ) : T → T × T ∨ and let P be the Poincaré bundle on
T × T ∨ . Then (id, λE )∗ P is an ample line bundle on T whose first Chern class
agrees with the given polarization of T .
Theorem 2.10. The metric induced on the bundle (id, λE )∗ P is given by the
function k · k : V × C× → R>0 ,
k(z, s)k2 = |s|2 exp(−4π Im(z)t ∆(Im Ω)−1 Im(z)). (2.19)
Proof. This follows from equations (2.15) and (2.18).
Hodge structures of type (−1, 0), (0, −1). Recall that a pure Hodge structure of
type (−1, 0), (0, −1) is given by
1. A finite rank Z-module, HZ .
def
2. A decreasing filtration F • on HC = HZ ⊗ C such that
F −1 HC = HC , F 1 HC = 0, HC = F 0 HC ⊕ F 0 HC .
A polarization of a Hodge structure of type (−1, 0), (0, −1) is a non-degenerate

antisymmetric bilinear form Q : HZ ⊗ HZ → Z which, when extended to HC by
linearity, satisfies the “Riemann bilinear relations”
1. The subspace F 0 HC is isotropic.
2. If x ∈ F 0 HC , then iQ(x, x) > 0.
We will be interested only in torsion-free Hodge structures. We recall that the
category of torsion-free Hodge structures of type (−1, 0), (0, −1) and the category
of compact complex tori are equivalent cf. [4, Exercise 1.5.10]. If H = (HZ , F • )
is such a Hodge structure, we write V = HC /F 0 and π : HC → HC /F 0 for the
def
projection. Then Λ = π(HZ ) is a lattice in V, that defines a torus T = V/Λ. This
torus is denoted J(H) and called the Jacobian of H.
Conversely, if T is a complex torus, then H1 (T, Z) is torsion-free and has a
Hodge structure of type (−1, 0), (0, −1).
If (HZ , F • ) has a polarization Q then, identifying Λ with HZ and writing
E = Q, we obtain a polarization of T . We finish by verifying that, indeed E is a
polarization in the sense of complex tori. That E is non-degenerate follows from
the non-degeneracy of Q. Let v , w ∈ V, choose x̄, ȳ ∈ F 0 HC such that π( x̄) = v
and π(ȳ) = w. Write x, y for the complex conjugates of x̄ and ȳ respectively. Then
x + x̄ ∈ HZ ⊗ R and π(x + x̄) = v , while ix − i x̄ ∈ HZ ⊗ R and π(ix − i x̄) = −iv .
Thus by the first Riemann bilinear relation
E(iv , iw) = Q(−ix + i x̄, −iy + iȳ) = Q(x, ȳ) + Q( x̄, y)
E(v , w) = Q(x + x̄, y + ȳ) = Q(x, ȳ) + Q( x̄, y),
Thus E(iv , iw) = E(v , w). Moreover, by the second bilinear relation
H(v , v ) = −E(iv , v ) = −Q(−ix + i x̄, x + x̄) = 2iQ(x, x̄) > 0.
2.3. Nilpotent orbit theorem. The aim of this section is to formulate a version
of the Nilpotent orbit theorem that allows us to deal with variations of mixed
Hodge structures, in a setting with several variables. Such a Nilpotent orbit
theorem is stated and proved in [22]. In order to formulate this theorem, we
need quite a bit of background material and in particular define the notion of
“admissibility” for variations of mixed Hodge structures. Also we need to take a
detailed look at the behaviour of monodromy on the fibers of the underlying local
systems. Most of the introductory material below is taken from [25, Section 14.4]
and [22].
Variations of polarized mixed Hodge structures. Let X be a complex manifold.

A graded-polarized variation of mixed Hodge structures on X is a local system
H → X of finitely generated abelian groups equipped with:
1. A finite increasing filtration
W• : 0 ⊆ . . . ⊆ W k ⊆ W k+1 ⊆ . . . ⊆ HQ
of HQ = H ⊗ Q by local subsystems, called the weight filtration,
2. A finite decreasing filtration
F•: HC ⊗ OX ⊇ . . . ⊇ F p−1 ⊇ F p ⊇ . . . ⊇ 0
of the vector bundle H = HC ⊗ OX by holomorphic subbundles, called the

Hodge filtration,
3. For each k ∈ Z a non-degenerate bilinear form
Qk : GrW W
k (H Q ) ⊗ Grk (H Q ) → Q X
of parity (−1)k ,
such that:
1. For each p ∈ Z the Gauss-Manin connection ∇ on H satisfies the “Griffiths

transversality condition” ∇F p ⊆ Ω1X ⊗ F p−1 ,
2. For each k ∈ Z the triple (GrW • W

k (H Q ), F Grk (H), Qk ) is a variation of pure
polarized rational Hodge structures of weight k. Here for each p ∈ Z we
write F p GrW p W
k (H) for the image of F H ∩ W k H in Grk (H C ) under the
projection map W k H → GrW k (H C ).
A variation of polarized mixed Hodge structures will be called torsion-free if

H is a local system of torsion-free abelian groups. A Q-variation of polarized
mixed Hodge structures is defined analogously with the difference that H is a
local system of finite dimensional Q-vector spaces.
Period domains. If (H, W• , F • ) is a mixed Hodge structure, then HC has a unique
bigrading I •,• such that
s,r
F p HC = ⊕r≥p,s I r,s , Wk HC = ⊕r+s≤k I r,s , I r,s = I mod ⊕ p<r,q<s I p,q .
The integers hr,s = dim I r,s are called the Hodge numbers of (H, W• , F • ).
Given a quadruple (H, W• , Qk , h) with H a rational vector space, W• an
increasing filtration of H, Qk a collection of non-degenerate bilinear forms of
parity (−1)k on GrW k (H), and a partition of dim(H) into a sum of non-negative
integers h = {hr,s } satisfying the symmetry condition hr,s = h s,r , there exists
a natural classifying space (also known as a period domain) M = M(h) =
M(H, W• , Qk , h) of mixed Hodge structures (W• , F • ) on H which are graded-
polarized by Qk .
We recall the construction of M from [22, § 3]. Write
X X
p
fp = hr,s and fk = hr,k−r
r≥p, s r≥p
and let M̌ be the set of all decreasing filtrations F • of HC satisfying

p
dim(F p ) = f p , dim(F p GrW p W
k ) = fk , and Qk (F Grk , F
k−p+1
GrW
k )= 0
The group n o
GC = g ∈ GL(HC )W Grk (g) ∈ AutC (Qk )
is a complex algebraic group that acts transitively on M̌ giving to it a structure
of complex manifold. The manifold M̌ is usually called the “compact dual” of
M by analogy with the pure case, although in general it is not compact.
The period domain M is the subset of M̌ formed by the filtrations F • such
that (H, W• , F • , Q) is a polarized Q-mixed Hodge structure. By [22, Lemma
3.9] M is an open subset of M̌, hence it has an induced structure of complex
manifold. By the same lemma, the group
n o
G = g ∈ GL(H )W Gr (g) ∈ Aut (Q )
P C k R k
acts transitively on M. We also consider the group

n o
GR = g ∈ GL(HR )W Grk (g) ∈ AutR (Qk ) . (2.20)
Note that we have inclusions
GR ⊂ G P ⊂ GC .
Remark 2.11. The group GR acts transitively on the subset MR of filtrations

defining a mixed Hodge structure that is split over R. If the filtration W has only
two non-trivial weights that are adjacent, that is, if there is a k such that
0 = Wk−2 ⊂ Wk−1 ⊂ Wk = H,
then any mixed Hodge structure on M(H, W, Q, h) is split over R. Therefore
MR = M and GR acts transitively on M. This will hold for the case of interest
to us in Section 2.4.
Relative filtrations. Let H be a rational vector space, equipped with a finite
increasing filtration W• . We let N denote a nilpotent endomorphism of H,
compatible with W• . We call an increasing filtration M• of H a weight filtration
for N relative to W• if the two following conditions are satisfied:
1. for each i ∈ Z we have N Mi ⊆ Mi−2 ,
2. for each k ∈ Z and each i ∈ N we have that N i induces an isomorphism
∼
N i : Grk+i
M
GrW
k H −
M
→ Grk−i GrW
k H
of vector spaces.
It can be verified that if H has a weight filtration for N relative to W• , then it is
unique. We call N strict if N(H) ∩ Wk = N(Wk ) for all k ∈ Z. By [27, Proposition
2.16], if the filtration W• has length two (in the sense that H = Wk and Wk−2 = 0
for some k), and if H has a weight filtration for N relative to W• , then N is strict.
Admissible variations of mixed Hodge structures. Now let (H, W • , F • , Qk ) be
a variation of graded-polarized mixed Hodge structures over the punctured unit
disc ∆∗ . Let s0 ∈ ∆∗ and (H, W• , F • , Qk ) the fiber of (H, W • , F • , Qk ) over s0 .
Let γ be a generator of the fundamental group π1 (∆∗ , s0 ) and T the monodromy
operator defined by γ acting on H. Since W • is a filtration by local subsystems,
the monodromy operator preserves W • . The operator T can be written as
T = T s T u , where T s is semisimple and T u is unipotent. The monodromy is
said to be quasi-unipotent if T sr = Id for certain integer r ≥ 1. We denote by
N = log T u the logarithm of the unipotent part of the monodromy. Clearly N is
nilpotent.
By [11, II Remarque 5.5] the vector bundle H = H ⊗C O∆∗ can be “canoni-
cally” extended to a vector bundle H̃ on the unit disk. Moreover, the subbundles
Wk = W k ⊗ O∆∗ also extend canonically to subbundles W̃k of H̃. These exten-
sions are really canonical when the monodromy is unipotent. In case it is not, the
extensions depend on the choice of a logarithm, but, as explained in [11, § II 5]
this choice can be made once and for all.
Following [17], [25] and [27] we call the variation (H, W • , F • , Qk ) pre-
admissible if
1. the monodromy is quasi-unipotent,
2. the logarithm N of the unipotent part of the monodromy has a weight

filtration M• (H, W• , N) relative to W• on H,
3. the subbundles F • of H extend to subbundles F˜ • of H̃ in such a way that

p
the coherent sheaves GrF˜ GrW̃
k are locally free.
Assume now that X is an open submanifold of a manifold X, where D =

X \ X is a normal crossings divisor. Let (H, W • , F • , Qk ) be a graded-polarized
variation of mixed Hodge structures over the complex manifold X. We then
call the variation (H, W • , F • , Qk ) admissible if for every holomorphic map
f¯ : ∆ → X with f¯(∆∗ ) ⊂ X, the variation f ∗ H on ∆∗ is pre-admissible. Here
we denote by f the restriction of f¯ to ∆∗ .
In algebraic geometry, admissible variations of mixed Hodge structures come
about as follows. Let π : Y → X be a morphism of complex algebraic varieties.
Then there is a non-empty open subset ι : U → X such that the constructible
sheaf H = ι∗ Ri π∗ ZY is a local system that has a canonical structure of admissible
graded-polarized variation of mixed Hodge structures (H, W • , F • , Qk ). More-
over, there is a finite étale map g : U e → U such that g∗ H has unipotent mon-
odromy.
In general, the usual cohomological operations like direct images or relative
cohomology will produce mixed Hodge modules [23, 24] which are a generaliza-
tion of the notion of admissible variations of polarizable mixed Hodge structures,
where the local system H is replaced by a perverse sheaf of Q-vector spaces.
There is a criterion for when a mixed Hodge module is indeed an admissible
variation of mixed Hodge structures: given a mixed Hodge module H (with a
given polarization), if the underlying perverse sheaf is a local system, then H is
an admissible Q-variation of polarized mixed Hodge structures. See for instance
[2] for a survey on mixed Hodge modules.
For admissible variations of mixed Hodge structures we have the follow-
ing compatibility between the graded polarization and the monodromy. Let
(H, W• , F • , Qk ) be a reference fiber of the variation near the boundary divisor
D = X \ X of the smooth algebraic variety X. We denote the local monodromy
operators around the branches of D by T 1 , . . . , T m , and the corresponding log-
arithms of the unipotent part by N1 , . . . , Nm . We denote by gC the Lie algebra
Lie GC of the group CC defined above. Then, in this generality, the T i belong
to GC , and the Ni belong to gC , for each i = 1, . . . , m. The R>0 -span C of the
local monodromy logarithms Ni inside gC is called the open monodromy cone of
the reference fiber (H, W• , F • , Qk ). Each element of C is nilpotent, and it can be
proved that the relative weight filtration M• of (H, W• ) is constant on C.
The period map. To an admissible graded-polarized variation of mixed Hodge

structures (H, W • , F • , Qk ) over X = (∆∗ )k × ∆n−k we can associate a period map,
as follows. Let M = M(h) be the period domain associated to (H, W• , Qk ) and
set GC as above. Let Γ ⊂ GC be the image of the monodromy representation
ρ : π1 (X, x0 ) → GC . The period map φ : X → Γ\M is the map that associates to
x ∈ X the Hodge filtration of H x . The period map is holomorphic.
Let H ⊂ C be the upper half plane. Let e : Hk → (∆∗ )k be the uniformization
map given by (z1 , . . . , zk ) 7→ (exp(2πiz1 ), . . . , exp(2πizk )). Then along e the
period map φ lifts to a map φ̃ : Hk × ∆n−k → M. In other words, we have
the following commutative diagram
φ̃
Hk × ∆n−k // M
(e,id)
φ
(∆∗ )k × ∆n−k // Γ \ M
P the right hand arrow is the canonical projection. As Ni ∈ Lie GC we find

where
exp( ki=1 zi Ni ) ∈ GC for all z1 , . . . , zk ∈ H. Let ψ̃ : Hk × ∆n−k → M̌ be the map
given by
X
k
ψ̃(z1 , . . . , zk , qk+1 , . . . , qn) = exp(− zi Ni ).φ̃(z1 , . . . , zk , qk+1 , . . . , qn ) .
i=1
Then ψ̃ descends to an “untwisted” period map ψ : (∆∗ )k × ∆n−k → M̌, fitting in

a commutative diagram
ψ̃
Hk × ∆n−k // M̌
tt::
ψ ttt
ttt
tt
(∆∗ )k × ∆n−k
Note that, importantly, the map ψ takes values in the compact dual M̌, and not
in a quotient of it.
Nilpotent orbit theorem. The following result is the starting point of G. Pearl-
stein’s Nilpotent orbit theorem (cf. [22, Section 6]) for admissible graded-polarized
variations of mixed Hodge structures and is enough for showing the estimates we
need.
Theorem 2.12. (G. Pearlstein) Let (H, W • , F • , Qk ) be an admissible graded-
polarized variation of mixed Hodge structures over X = (∆∗ )k × ∆n−k . Then
the untwisted period map ψ extends to a holomorphic map ψ : ∆n → M̌.
2.4. Families of pointed polarized abelian varieties.

Period vectors. Assume that (H, F • , Q) is a polarized pure Hodge structure
of weight −1, torsion-free, of type (−1, 0), (0, −1), and of rank 2g. Recall
that given a Q-symplectic integral basis (a1 , . . . , ag , b1, . . . , bg ) of (H, Q), there
exists a unique basis (w1 , . . . , wg ) of F 0 HC determined by demanding that
Pg
wi = − j=1 Ωi j a j + bi for some (period) matrix Ω ∈ Mg (C) (cf. equation (2.9)).
We call this new basis the associated normalized basis. As we have seen, the
Riemann bilinear relations imply that Ωt ∆ = ∆Ω and ∆ Im Ω > 0.
Assume an extension
0 → H → H ′ → Z(0) → 0 (2.21)
in the category of mixed Hodge structures is given. Then H ′ has weight filtration
W• : 0 ⊂ W−1 = HQ ⊂ W0 = HQ′ .
Taking F 0 (−)C in (2.21) yields the extension

0 → F 0 HC → F 0 HC′ → C → 0
of C-vector spaces. As can be readily checked, for each a0 ∈ H ′ that lifts
the canonical generator of Z(0) in (2.21) there exists a unique w0 ∈ F 0 HC′
such that w0 ∈ a0 + C-span(a1 , . . . , ag ). Given such a lift a0 , we let δH ′ =
(δ1 , . . . , δg)t ∈ Colg (C) be the coordinate vector determined by the identity
P
w0 = a0 + gj=1 δ j a j . We call δH ′ the period vector of the mixed Hodge structure
(H ′ , F • , W• ) on the basis (a0 , a1, . . . , ag, b1 , . . . , bg ) of the Z-module H ′ . It can be
verified that replacing a0 by some element from a0 + H changes δ by an element
of Zg + ΩZg . The resulting map Ext1MHS (Z(0), H) → Cg /(Zg + ΩZg ) is finite,
and gives Ext1MHS (Z(0), H) a canonical structure of complex torus.
Let A = J(H) be the Jacobian of H. Thus A is a polarized complex abelian
variety of dimension g with H = H1 (A). Let ν ∈ A, and write H(ν) for the relative
homology group H1 (A, {0, ν}). There is an extension of mixed Hodge structures
0 → H → H(ν) → Z(0) → 0
canonically associated to (A, ν). Here Z(0) is to be identified with the reduced
homology group H̃0 ({0, ν}). The map A → Ext1MHS (Z(0), H) given by sending
ν to the extension H(ν) is a bijection, compatible with the structure of complex
torus on left and right hand side.
The period map of a family of pointed polarized abelian varieties. Let X be a
smooth complex variety with an open immersion X ⊂ X into a smooth complex
algebraic variety, with D = X \ X a normal crossings divisor. Let (H, F • , Qk ) be
a variation of polarized torsion-free pure Hodge structures of weight −1 and type
(−1, 0), (0, −1) over X. We note that such a pure polarized variation is necessarily
admissible. Write Y = J(H) and let π : Y → X be the associated analytic family
of polarized abelian varieties over X, with polarization λ : Y → Y ∨ .
We will now work locally complex analytically. Thus we will suppose that
X is the polydisk ∆n , and D is the divisor given by the equation q1 · · · qk = 0, so
that X = (∆∗ )k × ∆n−k . We further assume that all local monodromy operators
T 1 , . . . , T k about the various branches determined by q1 , . . . , qk are unipotent (for
instance, this is the case if the family extends as a semiabelian scheme Y → X).
Let g be the relative dimension of Y → X. Also, we will henceforth usually
suppress the polarization from our notation.
Let (H, F • ) be a reference fiber of H near the origin. Let N be any element
of the open monodromy cone of H. Then we have N 2 = 0 and the filtration
associated to N simply reads
0 ⊂ M−2 ⊂ M−1 ⊂ M0 = HQ
with M−2 = Im N and M−1 = Ker N. Since, in this case the group GR acts
transitively on M, the operator N belongs to the Lie algebra of GR , thus there
exist a Q-symplectic integral basis (a1 , . . . , ag, b1 , . . . , bg ) of (H, Q) and a non-
negative integer r ≤ g such that:
1. M−2 = span (a1 , . . . , ar ),
2. M−1 = span (a1 , . . . , ag , br+1, . . . , bg ).
In particular, (ār+1 , . . . , āg, b̄r+1 , . . . , b̄g ) is a Q-symplectic integral basis of the

M
pure polarized Hodge structure Gr−1 H of type (−1, 0), (0, −1). Clearly, with
respect to this basis, each local monodromy operator N j has the form
!
0 A′j
Nj = .
0 0
Let ∆ be the matrix associated to the polarization as in equation (2.16). Each
def
A′j is integral and the g-by-g matrices A j = ∆A′j are symmetric and positive
semidefinite. Moreover, the left upper r-by-r block of A j is positive definite. See
[9] for more details in the above construction.
To avoid the appearance of the polarization matrix ∆ and thus to simplify the
notation we will sometimes replace the Q-symplectic integral basis (a, b) by the
symplectic Q-basis (a∆−1 , b). In this new basis each local monodromy operator
N j has the form
!
0 Aj
Nj = .
0 0
On this new basis we can realize the period domain associated to H as the
usual Siegel’s upper half space Hg of rank g. We have GR = Sp(2g, R), and the
action on Hg is given by the usual prescription
! !
A B −1 A B
· M = (AM + B)(CM + D) , ∈ Sp(2g, R) , M ∈ Hg .
C D C D
In this representation the period map Ω : X → Γ \ Hg is made explicit by
associating to each x ∈ X the matrix Ω(x) = ∆ΩYx , where ΩYx is the period
matrix of the fiber Yx on the chosen Q-symplectic integral basis of H. Here Γ is
the image of the monodromy representation into Sp(2g, R). In the new basis, the
monodromy representation sends the local monodromy operator T j to the matrix
!
1 Aj
∈ Sp(2g, R) .
0 1
We will now extend this picture to include an admissible normal function ν.
Definition 2.13. Let U ⊂ X be a non-empty open subset. An admissible normal

function of Y on U is a holomorphic section ν : U → Y|U such that the associated
graded-polarized mixed Hodge structures H x (ν), x ∈ U form an admissible
variation over U.
Let ν be an admissible normal function of Y on X. Varying x ∈ X we thus find

an extension
0 → H → H(ν) → Z(0) → 0
of admissible variations of graded-polarized mixed Hodge structure. The weight
filtration of this variation looks like
W• : 0 ⊂ W −1 = HQ ⊂ W 0 = H(ν)Q ,
so that GrW W
−1 H(ν)Q = H Q and Gr0 H(ν) = Q(0). We denote the Hodge filtration
•
of H(ν) by F . We start by taking a reference fiber H(ν) of H(ν) and augmenting
our chosen Q-symplectic integral basis of H by an a0 ∈ H(ν) lifting the canonical
generator of Z(0).
Since by assumption, H(ν) is an admissible variation of polarized mixed
Hodge structures, the relative weight filtration M•′ on our reference fiber H(ν)
exists. Let N ′ be an element of the open monodromy cone of H(ν) such that
N = N ′ |H . We will now proceed to determine the matrix shape of N ′ on the basis
(a0 , a∆−1, b) of H(ν). As N ′2 = 0, the filtration associated to N ′ on H(ν) is
L• : 0 ⊂ L−1 ⊂ L0 ⊂ L1 = H(ν)Q ,
with L−1 = Im(N ′ ), L0 = Ker(N ′ ). As the monodromy action on GrW 0 = Q(0)

is trivial, we have that Im(N ′ ) ⊂ HQ , so that N ′−1 HQ = H(ν)Q (here by
N ′−1 HQ we denote the inverse image of HQ under N ′ ). As W• has length
two, and as by admissibility the weight filtration of N ′ relative to W• exists,
as we noted above it follows that N ′ is strict. Explicitly, we have that H(ν)Q =
N ′−1 HQ = HQ + Ker(N ′ ). The equality H(ν)Q = HQ + Ker(N ′ ) implies that
Ker(N ′ ) % Ker(N) and hence that Im(N ′ ) = Im(N).
The period domain associated to (H(ν), W• ) can be realized as Cg × Hg . The
group GR in this case is
 
! 


  1 0 0  


 
 g A B 
GR = 
 

m A B 
 : m, n ∈ R , ∈ Sp(2g, R)
 .

 n C D C D 

The action of GR on Cg × Hg is given by

 
 1 0 0 
 
m A B  (v , M) = (v + m + Mn, (AM + B)(CM + D) ) , v ∈ C , M ∈ Hg .
−1 g
n C D
Varying x ∈ X = (∆∗ )k × ∆n−k and then taking F 0 we obtain a period map
associated to the variation H(ν)
(δ, Ω) : X → Γ \ (Cg × Hg )
that is given by
(δ(x), Ω(x)) = (∆δH(ν(x)) , ∆ΩYx ).
We denote by
(δ̃, Ω̃) : Hk × ∆n−k → Cg × Hg
the lift of the period map along the map (e, id) : Hk × ∆n−k → X, where we recall
that we denote by e : Hk → (∆∗ )k the map
e(z1 , . . . , zk ) = (exp(2πiz1 ), . . . , exp(2πizk )).
Theorem 2.14. There exist a holomorphic map ψ : ∆n → S g (C), a holomorphic

map α : ∆n → Cg , and vectors c1 , . . . , ck ∈ Qg with ∆−1 A j c j ∈ Zg for j = 1, . . . , k
such that for (z, t) ∈ Hk × ∆n−k with e(z) sufficiently close to zero the equalities
X
k X
k
Ω̃(z, t) = z j A j + ψ(e(z), t) , δ̃(z, t) = z j A j c j + α(e(z), t)
j=1 j=1
hold in S g (C) resp. Cg .

Proof. Let N j denote the local monodromy operator of H around the branch of
D determined by q j = 0. We have
!
z 1 z jA j
exp(z j N j ) = T j j =
0 1
and hence exp(z j N j ).M = z j A j + M for each M ∈ Hg , z j ∈ U, and j = 1, . . . , k

(here U is an open subset of H consisting of points with sufficiently large
imaginary part). Denote by Pg the compact dual of Hg . The untwisted period
P
map ψ : ∆n → Pg obtained by Theorem 2.12 extending exp(− kj=1 z j N j ).Ω̃(z, t)
factors through S g (C) ⊂ Pg . We obtain the equalities
X
k X
k
Ω̃(z, t) = exp( z j N j ).ψ(e(z), t) = z j A j + ψ(e(z), t)
j=1 j=1
in S g (C).
Let N ′j denote the local monodromy operator of H(ν) around the branch of D
determined by q j = 0. The equality Im(N ′j ) = Im(N ′j |HQ ) on H(ν)Q that follows
from our above considerations shows that N ′j has a matrix
 
 0 0 0 
 −1 −1 
∆ A j c j 0 ∆ A j 
0 0 0
on the integral basis (a0 , a1 , . . . , ag , b1, . . . , bg), for some c j ∈ Qg . Since the
monodromy is integral in such basis, we deduce that ∆−1 A j c j has to be integral.
In the Q-basis (a0 , a∆−1 , b), the matrix of N ′j is
 
 0 0 0 
 
A j c j 0 A j 
0 0 0
Then for (v , M) ∈ Cg × Hg and z j ∈ U we have exp(z j N ′j ).(v , M) =

(v + z j A j c j , M + z j A j ). Let (α, ψ) : ∆n → Cg × Pg denote the untwisted period
map. We find the equalities
X
k X
k
′
δ̃(z, t) = exp( z j N j ).α(e(z), t) = z j A j c j + α(e(z), t)
j=1 j=1
in Cg .
The norm of a section. Denote by P the Poincaré bundle on Y ×X Y ∨ with its

canonical C ∞ hermitian metric as described in Section 2.2. Given two admissible
normal functions ν, µ : X → Y we will denote
Pν,µ = (ν, λµ)∗ P, Pν = Pν,ν ,
where λ : Y → Y ∨ is the isogeny provided by the polarization. We are interested
in studying the singularities of the metric of Pν,µ when we approach the boundary
of X.
Consider the five maps
m, p1,3, p1,4 , p2,3, p2,4 : Y ×X Y ×X Y ∨ ×X Y ∨ −→ Y ×X Y ∨ ,
where m(x, y, z, t) = (x + y, z + t) and pi, j is the projection over the factors i, j.
Then we have a canonical isomorphism
∼
m∗ P −
→ p∗1,3 P ⊗ p∗1,4P ⊗ p∗2,3 P ⊗ p∗2,4P (2.22)
of holomorphic line bundles over Y ×X Y ×X Y ∨ ×X Y ∨ , in other words, the Poincaré
bundle is a biextension on Y ×X Y ∨ in the sense of [12, Exposé VII]. The explicit
description of the cocycle bP in equation (2.13) and of the metric of the Poincaré
bundle in Remark 2.9 shows that the canonical isomorphism (2.22) is in fact an
isometry for the canonical induced metrics on left and right hand side. We obtain
in particular
Lemma 2.15. Let ν1 , ν2 , µ1 , µ2 be holomorphic sections of the family Y → X.
Then we have a canonical isometry
∼
(ν1 + ν2 , λ(µ1 + µ2 ))∗ P −
→ (ν1 , λµ1 )∗ P ⊗ (ν1 , λµ2)∗ P ⊗ (ν2 , λµ1 )∗ P ⊗ (ν2 , λµ2 )∗ P
of hermitian line bundles on X.
Let f : Y ×X Y ∨ → Y ×X Y ∨ be the map given by (x, ℓ) 7→ (λ∨ (ℓ), λ(x)). Then
∼
we have a canonical isometry f ∗ P −→ P⊗d , where d = det ∆ is the degree of the
polarization λ. This leads to canonical isometries
∼ ∼
P⊗d → P⊗d
ν,µ − µ,ν and P⊗2d → P⊗d
ν,µ −
⊗(−d)
ν+µ ⊗ Pν ⊗ P⊗(−d)
µ .
Hence, in order to study the singularities of the metric on Pν,µ it suffices to study
the singularities of the metric on Pν , Pµ and Pν+µ . In particular, for the purpose
of proving our main results, it suffices to focus on the diagonal cases Pν .
Let ν be an admissible normal function of the family Y → X. Let s be a
section on X of Pν . Interpreting the Poincaré bundle as parametrizing biextension
variations of mixed Hodge structures (see [13]) we can canonically associate to
the section s a biextension variation of mixed Hodge structures over X. We say

that the section s is admissible if this variation is admissible.
Let x0 be a point of X. The purpose of the present section is to give an
asymptotic expansion of the logarithm of the norm of an admissible section of
Pν near x0 . From equation (2.19) it follows that it suffices to give asymptotic
expansions of the period matrix of the family Y → X, and of the period vector
(see below) associated to ν.
We use now Theorem 2.14 to obtain an expression of the norm of the section
s. Let ||−|| denote the canonical metric on Pν = ν∗ P. Continuing the notation
from Theorem 2.14, let a = 2π Im α and B = 2π Im ψ. For j = 1, . . . , k let
x j = − log |t j |.
Corollary 2.16. For every admissible section s of Pν on (∆∗ )k ×∆n−k there exists
a meromorphic function h on ∆n which is holomorphic on (∆∗ )k × ∆n−k , such that
the identity
 k t  k −1  k 
X  X  X 
− log||s|| = − log|h| +  x j A j c j + a  x j A j + B  x j A j c j + a (2.23)
j=1 j=1 j=1
holds on (∆∗ )k × ∆n−k .
Proof. The vector z and the matrix Ω in Theorem 2.10 are expressed in the
integral basis (a, b), while δ(x) and Ω(x) are expressed in the Q-basis (a∆−1 , b).
Writing z = ∆−1 δ(x) and Ω = ∆−1 Ω(x), we obtain that
− log||s(x)|| = − log|h(x)| + 2π(Im δ(x))t (Im Ω(x))−1 (Im δ(x))
for a suitable meromorphic function h on ∆n which is holomorphic on (∆∗ )k ×
∆n−k . Note that, even though Ω(x), δ(x) are multivalued, their imaginary parts
1
are single valued. From Theorem 2.14 we obtain, noting that Im z j = − 2π log |t j |,
1 X 1 X
k k
Im Ω(x) = − A j log |t j | + Im ψ , Im δ(x) = − A j c j log |t j | + Im α .
2π j=1 2π j=1
Combining we find equation (2.23).
3. Normlike functions
The purpose of this section is to carry out a systematic study of the functions
 k t  k −1  k 
X  X  X 
ϕ =   
x j A j c j + a   
x j A j + B   x j A j c j + a
j=1 j=1 j=1
that appear on the right hand side of the equality in Corollary 2.16. We call such
functions normlike functions.
Let f : Rn → R be a function. The recession function of f is the function
1
rec( f )(x) = lim f (λx).
λ
λ→+∞
If the recession function exists, it is homogeneous of weight one, in the sense

that
rec( f )(µx) = µ rec( f )(x), for µ ∈ R>0 .
We show that normlike functions ϕ have a well-defined recession function
rec(ϕ) with respect to the variables x j , and we are able to calculate rec(ϕ)
explicitly. In our main technical lemma Theorem 3.2 we give bounds for the
difference ϕ − rec(ϕ) and, in the case where k = 1, for the first and second order
derivatives of ϕ − rec(ϕ). The bound on the difference will be key to the proof of
our first main result Theorem 1.1, the bounds on the derivatives will be used in
our proof of Theorem 1.3. In section 3.4 we prove, among other things, that the
recession functions rec(ϕ) are convex. This will lead to the effectivity statement
in Theorem 1.5.
3.1. Some definitions. Recall that we have denoted by Mr (R) the space of r-
by-r matrices with real coefficients, by S r+ (R) ⊂ Mr (R) the cone of symmetric
positive semidefinite real matrices inside Mr (R), and by S r++ (R) ⊂ S r+ (R) the
cone of symmetric positive definite real matrices.
Lemma 3.1. Let N1 , . . . , Nk be a finite set of positive semidefinite symmetric real

g-by-g matrices such that N1 +· · ·+Nk has rank r. Then there exists an orthogonal
matrix u ∈ Og (R) such that, upon writing Mi = ut Ni u for i = 1, . . . , k, we have
!
Mi′ 0
Mi = ,
0 0
P
with all Mi′ ∈ S r+ (R) and Mi′ ∈ S r++ (R).
Proof. It will be convenient to use the language of bilinear forms. If Q is a

symmetric positive semidefinite bilinear form on Rg and f1 , . . . , fg is a basis of
Rg such that Q( fα , fα ) = 0 for α = r+1, . . . , g, then Q( fα , fβ ) = 0 for β = 1, . . . , g
and α = r + 1, . . . , g. Indeed, for all λ ∈ R we have Q(λ fα − fβ , λ fα − fβ ) ≥ 0,
that is
−2λQ( fα , fβ ) + Q( fβ , fβ ) ≥ 0 .
Since this inequality is satisfied for all λ we deduce that Q( fα , fβ ) = 0.
Let N = N1 + · · · + Nk , and denote by Q the symmetric positive semidefinite
bilinear form that N defines on the standard basis (e1 , . . . , eg) of Rg . Note that Q
has rank r. By the spectral theorem, upon replacing the basis (e1 , . . . , eg ) of Rg
by ( f1 , . . . , fg ) = (e1 , . . . , eg )u for some orthogonal matrix u we can assume that
the expression of Q in the basis ( f1 , . . . , fg ) is
!
A 0
M= ,
0 0
with A ∈ S r+ (R) invertible and diagonal. In particular, Q( fα , fα ) = 0 for
α = r + 1, . . . , g. For i = 1, . . . , k let Qi denote the symmetric positive semi-
definite bilinear form that Ni defines on the standard basis (e1 , . . . , eg ) of Rg .
Note that Q = Q1 +· · ·+ Qk . Since all the Qi are positive semidefinite, we deduce
that Qi ( fα , fα ) = 0 for i = 1, . . . , k. Note that Mi = ut Ni u is the expression of Qi
in the basis ( f1 , . . . , fg ). By the previous discussion we have
!
Mi′ 0
Mi = ,
0 0
P
with Mi′ ∈ S r+ (R) and Mi′ = A ∈ S r++ (R), proving the lemma.
Suppose we are given the following data:

- three integers k ≥ 0, m ≥ 0, g ≥ 0;
- a real number κ ≥ 0;
- a compact subset K ⊆ Rm ;
- matrices A1 , . . . , Ak ∈ S g+ (R) all of rank ≥ 1;
- vectors c1 , . . . , ck ∈ Rg ;
- functions a : K → Rg and B : K → S g (R) which are restrictions of smooth
functions on some open neighbourhood of K;
such that for all (x1 , . . . , xk , λ) ∈ Rk>κ × K, we have that
def
X
k
P(x1 , . . . , xk , λ) = xi Ai + B(λ) > 0 . (3.1)
i=1
Note that if g = 0, then necessarily k = 0.

To these data we associate a smooth function ϕ : Rk>κ × K → R by
ϕ(x1 , . . . , xk , λ) =
 k t  k −1  k 
X  X  X 
 xi Ai ci + a(λ)  xi Ai + B(λ)  xi Ai ci + a(λ) . (3.2)

i=1 i=1 i=1
By condition (3.1), the function ϕ is well-defined and its values are non-negative.
We call ϕ the normlike function associated to the 4-tuple ((Ai ), (ci ), a, B). We call
P
the natural number k the dimension of ϕ. Write r = rk ki=1 xi Ai for some (hence
all) (x1 , . . . , xk ) ∈ Rk>κ . Note that r ≥ 1 if k > 0.
Let u ∈ Og (R). Replacing the vector ci by u−1 ci , a by u−1 a, the matrix B
by ut Bu and Ai by ut Ai u one checks that the function ϕ remains unchanged. By
Lemma 3.1 we can thus restrict to considering normlike functions where the Ai
have the shape !
A′i 0r,g−r
Ai = ,
0g−r,r 0g−r,g−r
P
with each A′i ∈ S r+ (R) and such that xi A′i ∈ S r++ (R) for all (x1 , . . . , xk ) ∈ Rk>κ
(hence for all (x1 , . . . , xk ) ∈ Rk>0 ).
From now on we assume that the matrices Ai indeed have this shape. We
write ! ! !
c′i a1 B11 B12
ci = , a= , and B =
⋆g−r a2 B21 B22
where c′i and a1 have size r, and B11 is an r-by-r matrix. The second block of the
vector ci is marked with an asterisk because the function ϕ is independent of its
value. Condition 3.1 implies that B22 (λ) is positive definite for all λ ∈ K, and the
symmetry of B implies that B21 = Bt12 .
We define another smooth function f : Rk>κ × K → R by
 k t  k −1  k 
X  X  X 
f (x1 , . . . , xk , λ) =  xi A′i c′i   xi A′i   xi A′i c′i  . (3.3)
i=1 i=1 i=1
Pk
This function f is well defined as i=1 xi A′i is positive definite on Rk>0 . The
function f depends trivially on λ and is clearly homogeneous of degree 1 in the
xi , and so defines a smooth function Rk>0 → R, which we also call f . Again, the
values of f are non-negative. By convention, if k = 0, the function f is zero.
Finally, the “recession” of ϕ is defined as the pointwise limit
rec(ϕ) : Rk>κ × K → R
(x1 , . . . , xk , λ) 7→ limµ→∞ µ1 ϕ(µx1 , . . . , µxk , λ),
if it exists. Again, if k = 0, then rec(ϕ) = 0.

3.2. Statement of the technical lemma. We can now state the “main technical
lemma”:
Theorem 3.2. In the notation of the previous section, write ϕ0 = ϕ − f . Note that
ϕ0 is a smooth function on R>κ × K. Then
1. the function |ϕ0 | is bounded on Rk>κ′ × K for some κ′ ≥ κ. The recession

of ϕ exists and is equal to f . In particular, rec(ϕ) is independent of the
parameter λ;
2. the function f is bounded on the open simplex ∆0 = {(x1 , . . . , xk ) ∈ Rk>0 :
Pk
i=1 xi = 1};
3. when k = 1,
(a) the function ϕ0 : R>κ ×K → R extends continuously to a function from
R>κ × K to R, where by R>κ we denote R>κ ⊔ {∞} with the natural
topology;
(b) the derivatives of ϕ0 satisfy the estimates
∂ϕ0 ∂2 ϕ0
= O(x−2
1 ) and = O(x−3
1 ),
∂x1 ∂x21
as x1 → ∞, where the implicit constant is uniform in K.
Example 3.3. When k > 1, in general we can not extend ϕ0 to a continuous

k
function on R>κ × K as the following example shows. Put g = 1, k = 2, m = 0,
A1 = A2 = 1, c1 = 1, c2 = 2, B = 0, κ = 1 and a = 1. Then
2(x1 + 2x2 ) + 1
ϕ0 = ϕ − f = .
x1 + x2
The sequences {(n, n)}n≥1 and {(n, 2n)}n≥1 converge, when n → ∞, to the point
2
(∞, ∞) ∈ R>1 . Nevertheless
10
lim ϕ0 (n, n) = 3, lim ϕ0 (n, 2n) = ,
n→∞ n→∞ 3
2
showing that ϕ0 can not be continuously extended to R>1 .
Before starting the proof of Theorem 3.2 we recall a few easy statements
related to Schur complements and inverting a symmetric block matrix. For a
symmetric block matrix !
A B
M= t
B C
with C invertible we call A − BC −1 Bt the Schur complement of the block C in M.
We have a product decomposition
! ! ! !
A B 1 BC −1 A − BC −1 Bt 0 1 0
M= t = .
B C 0 1 0 C C −1 Bt 1
In particular, M is invertible if and only if A − BC −1 Bt is invertible, and if these

conditions are satisfied we have
!
−1 (A − BC −1 Bt )−1 −(A − BC −1 Bt )−1 BC −1
M = .
−C −1 Bt (A − BC −1 Bt )−1 C −1 + C −1 Bt (A − BC −1 Bt )−1 BC −1
Also, if M is positive semidefinite, then so is the Schur complement A − BC −1 Bt .
3.3. Proof of the technical lemma. First we observe that, if k = 0, then ϕ is a
continuous function on a compact set, hence is bounded. Moreover, the function
f is zero. Thus the statements are trivially true and we are reduced to the case
k > 0 and hence g > 0.
Assume that we have already shown that |ϕ− f | is bounded on Rk>κ′ ×K. Then,
for each (x1 , . . . , xk , λ) ∈ Rk>κ′ × K we have
1 1
lim ϕ(µx1 , . . . , µxk , λ) = lim f (µx1 , . . . , µxk ) .
µ→∞ µ µ→∞ µ
The latter limit exists and is equal to f (x1 , . . . , xk ) by weight-one-homogeneity

of f . Thus the recession function of ϕ exists and agrees with f . In consequence,
in order to prove Theorem 3.2.1 and 3.2.2 we only need to show the boundedness
of |ϕ − f | and of f on the required subsets.
We next show that we can assume a simplifying hypothesis.
Definition 3.4. We say that the set of symmetric positive semidefinite matrices
A1 , . . . , Ak satisfies the flag condition if Ker(Ai ) ⊆ Ker(Ai+1 ), for i = 1, . . . , k − 1.
Consider the subset

U = {0 < x1 ≤ x2 ≤ · · · ≤ xk } ⊂ Rk>0 .
Since [
Rk>κ = σ−1 U ∩ Rk>κ
σ∈Sk
and [
∆0 = σ−1 U ∩ ∆0 ,
σ∈Sk
by symmetry it is enough to prove the boundedness of |ϕ − f | in U ∩ Rk>κ and P of f

in U ∩ ∆0 . Writing y1 = x1 , yi = xi − xi−1 for i = 2, . . . , k we find that xi = ij=1 y j
and that U ∩ Rk>κ is parametrized by the set y1 > κ, y2 , . . . , yk ≥ 0.
Note that
X
k X
k X
k X
k X
k X
k
xi Ai = yi Aj and xi Ai ci = yi A jc j.
i=1 i=1 j=i i=1 i=1 j=i
P
Lemma 3.5. Writing Ãi = kj=i A j we have that Ker Ãi ⊆ Ker Ãi+1 . Moreover we
Pk
have Im(Ãi ) = j=i Im(A j ).
Proof. We first observe that, if A is a symmetric positive semidefinite real

matrix, then Ax = 0 if and only if xt Ax = 0. Indeed, clearly Ax = 0 implies
xt Ax = 0. Conversely, assume that xt Ax = 0 and let y be any vector. Then, for
all λ ∈ R,
0 ≤ (y + λx)t A(y + λx) = yt Ay + 2λyt Ax
which implies that yt Ax = 0. Therefore Ax = 0.
T
We show that this observation implies that Ker Ãi = kj=i Ker A j . We have
x ∈ Ker Ãi if and only if
X
k
0 = xt Ãi x = xt A j x.
j=i
Since the matrices A j are positive semidefinite this implies that xt A j x = 0,

T
j = i, . . . , k. Therefore x ∈ kj=i Ker A j . The converse is clear. As a result
\
k \
k
Ker Ãi = Ker A j ⊆ Ker A j = Ker Ãi+1 .
j=i j=i+1
Since, for a symmetric positive semidefinite matrix A, the image Im(A) is the
orthogonal complement of Ker(A) we deduce
\
k ⊥ X
k X
k
Im(Ãi ) = Ker(Ãi )⊥ = Ker(A j ) = Ker(A j )⊥ = Im(A j ) .
j=i j=i j=i
This proves the lemma.
It follows from the Lemma that there exist vectors c̃i ∈ Rg such that
X
k
A j c j = Ãi c̃i .
j=i
Replacing Ai by Ãi , xi by yi and ci by c̃i we are reduced to proving the

boundedness of |ϕ − f | on R>κ × Rk−1
≥0 × K and of f on the set
X
k
{(x1 , . . . , xk ) ∈ Rk≥0 : x1 > 0 , xi ≥ 0 for all i > 1, (k − i + 1)xi = 1}
i=1
under the extra hypothesis that the matrices A1 , . . . , Ak satisfy the flag condition
from Definition 3.4. Clearly, by the homogeneity of f it is enough to prove the
boundedness of f on the set
X
k
H = {(x1 , . . . , xk ) ∈ Rk≥0 : x1 > 0 , xi ≥ 0 for all i > 1, xi = 1}.
i=1
From now on we assume the flag condition and we write ri = rk(Ai ). Then
r = r1 ≥ · · · ≥ rk ≥ 1. Thanks to the flag condition, we can assume furthermore
that the basis of Rg has been chosen in such a way that
!
′ A′′i 0
Ai = , (3.4)
0 0
with A′′i ∈ S r++
i
(R).
The following is our main estimate.
Lemma 3.6. There exists a constant c such that for all 1 ≤ α, β ≤ r and all
(x1 , . . . , xk ) ∈ R>0 × Rk−1 ≥0 we have the following bound on the α, β-entry in the
P
inverse of the r-by-r matrix ki=1 xi A′i :

X k −1 c c
xi A′i ≤ X ≤ .
i=1 αβ xj x1
j : r j ≥min(α,β)
Proof. This follows immediately from two intermediate results:

Claim 3.6.1. There exists a constant c1 > 0 such that for all (x1 , . . . , xk ) ∈
R>0 × Rk−1
≥0 we have the bound
X
k Y
r X
det xi A′i ≥ c1 xi > 0.
i=1 j=1 i:ri ≥ j
To prove this claim, define the r-by-r matrix

!
Idri 0
Ji = . (3.5)
0 0
Since A′′i is positive definite, there exists ǫ > 0 such that for all i, we have that
A′i − ǫ Ji is positive semidefinite. Then
X X X
xi A′i = xi (A′i − ǫ Ji ) + xi ǫ Ji . (3.6)
i i i
The proof of the following lemma is left to the reader.

Lemma 3.7. Let A, B be positive semidefinite symmetric r × r matrices. Then

det(A + B) ≥ det(A) + det(B).
Combining Lemma 3.7 with equation (3.6), we deduce

X X Y
r X
det xi A′i ≥ det xi ǫ Ji = ǫ r xi > 0
i i j=1 i:ri ≥ j
as required. The second intermediate result is as follows:
Claim 3.7.1. There exists a constant c2 > 0 such that for all 1 ≤ α, β ≤ r and all
(x1 , . . . , xk ) ∈ R>0 × Rk−1 ≥0 we have the following bound on the cofactors of the
P
matrix ki=1 xi A′i :

Xk Y r X
cof α,β xi A′i ≤ c2 xi .
i=1
α′ =1 i:ri ≥α′
α′ ,min(α,β)
P
To prove this second claim, write A = i xi A′i . Then there is a constant c3 such
that for 1 ≤ α′ , β′ ≤ r one has
X X
Aα′ ,β′ ≤ c3 xi ≤ c3 xi .
i:ri ≥max(α′ ,β′ ) i:ri ≥α′
∼
Let σ : {1, . . . , α̂, . . . , r} −
→ {1, . . . , β̂, . . . , r} be a bijection (the ˆ means “omit”).
Then Y Y X
A ′ ′ ≤ cr−1 xi
α ,σ(α ) 3
α′ ,α α′ ,α i:ri ≥α′
and since Aα′ ,σ(α′ ) = Aσ(α′ ),α′ we also have

Y Y X
A ′ ′ ≤ cr−1 xi .
α ,σ(α ) 3
α′ ,α α′ ,β i:ri ≥α′
Choosing the smaller upper bound of the two we find

Y Y X
A ′ ′ ≤ cr−1 xi
α ,σ(α ) 3
α′ ,α α′ ,min(α,β) i:ri ≥α′
and hence Y X
cof α,β (A) ≤ (r − 1)!cr−1
3 xi .
α′ ,min(α,β) i:ri ≥α′
This proves the second claim and, consequently, Lemma 3.6.

Proof (Proof of Theorem 3.2 (2)). From Lemma 3.6 we deduce the existence of
a constant c4 > 0 such that, for all 1 ≤ α, β ≤ r,
P P
X X X
′ ′ t ′ −1 ′ ′ j : r j ≥α x j i : ri ≥β xi
xi Ai ci xi Ai xi Ai ci ≤ c4 · P
j : r ≥min(α,β) x j
α α,β β
X j
= c4 · xj
j : r j ≥max(α,β)
and hence
X X t X −1 X X X
0≤ f = xi A′i c′i xi A′i xi A′i c′i ≤ c4 xj .
α α,β β
α,β α,β j : r j ≥max(α,β)
This is clearly bounded on H. This proves Theorem 3.2 (2).

Proof (Proof of Theorem 3.2 (1)). We start by noting that
P !
xi A′i + B11 B12
P= ,
B21 B22
with
P B′ 22 invertible. −1Moreover, as P is invertible, so is the Schur complement
xi Ai + B11 − B12 B22 B21 of B22 in P. If we put
X −1
Q= xi A′i + B11 − B12 B−1
22 B21
then we get !
−1 Q −QB12 B−122
P = −1 . (3.7)
−B−1 −1 −1
22 B21 Q B22 + B22 B21 QB12 B22
P
Write A = xi A′i and M = B11 − B12 B−1 −1
22 B21 so that Q = (A + M) . Recall that
−1 −1 −1
A is invertible, so that Q = (Idr + A M) A .
Claim 3.8. There exists a κ′ > κ such that the series
A−1 − A−1 MA−1 + A−1 MA−1 MA−1 + · · · + (−1)m (A−1 M)m A−1 + · · ·
converges to Q on R>κ′ × Rk−1
≥0 × K.
Proof. The entries of the matrix M are continuous functions on the compact set
K, hence bounded. Let c be the constant of Lemma 3.6, choose
κ′ > max(cr2 max(Mαβ ), c, κ)
and put ǫ = cr2 max(Mαβ )/κ′ . Note that 0 < ǫ < 1. Moreover, by Lemma 3.6
and the condition x1 ≥ κ′ ,
c (c max(Mαβ )r2 )m
|(A−1 M)m A−1 | ≤ < ǫm.
αβ x1 κ′m
It follows that the series converges absolutely. By construction, the limit of the
series is (A + M)−1 = Q finishing the proof of the claim.
Write M1 = (Idr + MA−1 )−1 and M2 = (Idr + A−1 M)−1 . Then Q = A−1 M1 =
M2 A−1 . An argument similar to that of Claim 3.8 shows that the entries of M1
and M2 are bounded on the set R>κ′ × Rk−1
≥0 × K.
We deduce from Lemma 3.6 that there is a constant c2 such that
c2
|Qαβ | ≤ P
j : r j ≥min(α,β) xj
on the same set. It follows that

X X
Q( ′ ′
xi Ai ci ) and ( xi A′i c′i )t Q
β α
−1
are bounded on R>κ′ × Rk−1
≥0 × K. Moreover, since Q − A = A−1 M3 A−1 with
again M3 having bounded entries, we deduce that there is another constant c3
such that
Q − A−1 ≤ c3
P ,
αβ P
j : r j ≥α x j i : ri ≥β xi
and consequently
X t X
xi A′i c′i Q − A−1 xi A′i c′i
is bounded. Finally, to prove that |ϕ − f | is bounded we compute

X t X X
ϕ− f = xi A′i c′i Q − A−1 xi A′i c′i + 2at1 Q xi A′i c′i
X
+ at1 Qa1 − 2at2 B−1
22 B 21 Q xi A′i c′i
− 2at2 B−1 t −1 −1 −1
22 B21 Qa1 + a2 (B22 + B22 B21 QB12 B22 )a2
and we use the previously obtained bounds. This proves Theorem 3.2 (1).
Proof (Proof of Theorem 3.2 (3)). From now on we assume that k = 1 so we

have ϕ : R>κ × K → R and f : R>0 → R. Explicitly,
ϕ(x1 , λ) = (A1 x1 c1 + a)t P−1 (A1 x1 c1 + a)
with P = A1 x1 + B, and f = ct1 A1 c1 x1 . Recall that we write ϕ0 = ϕ − f . Put

w0 = a − Bc1 .
Lemma 3.9. We have
ϕ0 (x1 , λ) = 2at c1 − ct1 Bc1 + wt0 P−1 w0 .

Proof. We compute
ϕ0 (x1 , λ) = (A1 x1 c1 + a)t P−1 (A1 x1 c1 + a) − ct1 A1 c1 x1
= (w0 + Pc1 )t P−1 (w0 + Pc1 ) − ct1 A1 c1 x1
= wt0 P−1 w0 + 2ct1 w0 + ct1 Pc1 − ct1 A1 c1 x1
= wt0 P−1 w0 + 2ct1 a − ct1 Bc1 .
We continue to assume that k = 1. It follows that A′1 is invertible.
Lemma 3.10. In the above notation and with k = 1, we have
! !
−1 0 0 1 A′−1
1 −A′−1
1 B12 B−1
22 −2
P = + −1 + O(x1 )
0 B−1 22 x1 −B−1 ′−1
22 B21 A1 B−1 ′−1
22 B21 A1 B12 B22
and !
1 1 0
P−1 A1 = + O(x−2
1 )
x1 −B−1
22 B21 0
as x1 → ∞, where the implicit constants are uniform in K.
Proof. From equation (3.7) we obtain
! !
−1 0 0 Q −QB12 B−122
P = + −1 .
0 B−122 −B−1 −1
22 B21 Q B22 B21 QB12 B22
Also recall that Q = (Idr + A−1 M)−1 A−1 with A = A′1 x1 and M bounded. This
yields Q = x−1 ′−1 −2
1 A1 +O(x1 ) as x1 → ∞. The first estimate readily follows. Upon
recalling that !
A′1 0
A1 =
0 0
the second estimate also follows.
To finish the proof of Theorem 3.2.3, note that by combining Lemma 3.9 and
Lemma 3.10 that
!
t 0 0
t t
ϕ0 (x1 , λ) = 2a c1 − c1 Bc1 + w0 w0 + O(x−1
1 )
0 B−1
22
as x1 → ∞. From this it is immediate that ϕ0 extends continuously to a function
from R>κ × K to R. Next, from Lemma 3.9 we have
∂ϕ0 ∂2 ϕ0
= −wt0 P−1 A1 P−1 w0 , = 2wt0 P−1 A1 P−1 A1 P−1 w0 .
∂x1 ∂x21
Combining this with Lemma 3.10 we find the estimates
∂ϕ0 ∂2 ϕ0
= O(x−2
1 ), = O(x−3
1 ),
∂x1 ∂x21
completing the proof of Theorem 3.2.3.
3.4. On the recession function of a normlike function. Let f : Rk>0 → R be

the recession function of the normlike function ϕ associated to ((Ai ), (ci ), a, B) as
above. The purpose of this section is to list a number of useful properties of f .
Proposition 3.11. The function f is convex, that is, for all x, y ∈ Rk>0 and all
λ ∈ [0, 1] we have f (λx + (1 − λ)y) ≤ λ f (x) + (1 − λ) f (y).
Proof. Example 3.4 on p. 90 of [5] states that the function hg : Rg ×S g++ (R) → R
given by hg (x, Y) = xt Y −1 x is convex. The function f : Rk>0 → R is the
composition of hg with the linear map
 k 
X X
k 
k g ++
R>0 → R × S g (R), (x1 , . . . , xk ) 7→   ′ ′
xi Ai ci , xi Ai  .
′
i=1 i=1
Since the composition of a linear map followed by a convex function is again

convex, we deduce that f is convex.
Proposition 3.12. The function f extends to a continuous function f : Rk≥0 →

R≥0 . The function f is homogeneous of weight one and convex.
Proof. By Theorem 3.2 (2) we know that the function f is bounded on the open
standard simplex ∆0 . Define
f : ∆ → R≥0
by the formula
f (x1 , . . . , xk ) = inf (pl )l →(x1 ,...,xk ) liminf l→∞ f (pl );

here the infimum is over sequences in ∆0 tending to the point (x1 , . . . , xk ). This
function f is well-defined because f is bounded on ∆0 . It follows easily from the
definition of f that f is convex and lower semi-continuous. Since ∆ is a convex
polytope, it follows from [26, Theorem 10.2] that f is continuous. Now extend f
to Rk≥0 \{0} by homogeneity. By sending in addition 0 to 0 we obtain the required
continuous and convex function f : Rk≥0 → R≥0 .
We can make the function f explicit as follows. Let I ⊆ {1, . . . , k} be any subset,
I
and set J = {1, . . . , k}\I. We consider the restriction of f to the subset R>0 ⊆ Rk≥0
given by setting x j equal to zero for all j ∈ J. Let
 
X 
rI = rk   xi Ai  : xi > 0,
i∈I
and for i ∈ I set !

A′′ 0
Ai = i (3.8)
0 0
where A′′i has size rI , and similarly
!
c′′i
ci = ,
⋆
where c′′i has length rI .
J
Note that, if I , ∅, then rI ≥ 1. Let K ⊂ R>0 be an arbitrary compact subset.
Write xI = (xi )i∈I and x J = (x j ) j∈J . We define the function
I
fI : R>0 × K → R, (xI ; x J ) 7→ f (x1 , . . . , xk ) .
Write X X
a(x J ) = x j A′j c′j , B(x J ) = A′j x j ,
j∈J j∈J
then we see that

 t  −1  
X  X  X 
fI (xI ; x J ) =  xi Ai ci + a(x J ) 
′ ′
xi Ai + B(x J ) 
′
xi Ai ci + a(x J )
′ ′
i∈I i∈I i∈I

I
and hence by Theorem 3.2 fI has a recession function rec( fI ) : R>0 → R which
can be written
 t  −1  
X  X  X 
rec( fI (xI )) =  xi Ai ci  
′′ ′′
xi Ai  
′′
xi Ai ci  ,
′′ ′′
i∈I i∈I i∈I
when I , ∅, and rec( f∅ ) = 0. Note that rec( fI ) is independent of the choice of K.

I
Also note that, by Theorem 3.2, | fI − rec( fI )| is bounded on R>0 × K.
Proposition 3.13. Let I ⊆ {1, . . . , k} be any subset. We have
f |R>0
I = rec( fI ).
Proof. When I = ∅ the equality is trivially true. We assume that I , ∅. Choose

J I
c ∈ R>0 and xI ∈ R>0 arbitrarily. By Theorem 3.2(1) with K = {c}, there exists a
constant δ > 0 depending on c and xI such that for all λ ∈ R>0 we have
|(rec( fI ))(λxI ) − fI (λxI ; c)| ≤ δ .
We deduce that for all λ ∈ R>0 we have
c δ
|(rec( fI ))(xI ) − f (xI , )| ≤ . (3.9)
λ λ
As f extends f continuously we have

c
lim f (xI , ) = f |R>0
I (x I )
λ→∞ λ
independently of the choice of c. Combining with the bound (3.9), we find that
(rec( fI ))(xI ) = f |R>0

I (x I ),
as required.
A special case of interest is when |I| = 1. For each 1 ≤ i ≤ k, set

!
Ae 0
Ai = i (3.10)
0 0
where Aei has size ri = rk Ai and hence is positive definite; here e is short for
“essential”. Similarly set
!
cei
ci = ,
⋆
where cei has length ri . Define
µi = cti Ai ci = (cei )t Aei cei .
Then µi ≥ 0 and we have for all xi > 0
f (0, . . . , 0, xi, 0, . . . , 0) = (rec( f{i} ))(xi ) = xi (cei )t (Aei )t x−1 e −1 e e

i (Ai ) xi Ai ci = xi µi .
In particular, the function f (0, . . . , 0, xi, 0, . . . , 0) is homogeneous linear in xi ,

and
µi = f (0, . . . , 0, 1, 0, . . . , 0) .
We call µ1 , . . . , µk ≥ 0 the coefficients associated to ϕ.
4. Proofs of the main results

In this section we prove our main results. We will continue to work with the
“diagonal case” where we consider the pullback Poincaré bundle Pν associated
to a single admissible normal function ν of our family π : Y → X. As was
explained at the beginning of Section 2.3, by the biextension property of the
Poincaré bundle this is sufficient for the purpose of proving the main results as
stated in the introduction.
4.1. Singularities of the biextension metric. In this section we will prove

Theorem 1.1.
Proof (Proof of Theorem 1.1). Following Theorem 2.14, take
- a small enough ǫ > 0,
- matrices
A1 , . . . , Ak ∈ S g (R)
−1
of positive rank such that ∆ Ai ∈ Mg (Z) for i = 1, . . . , k,
- vectors
c1 , . . . , ck ∈ Qg
such that ∆−1 Ai ci ∈ Zg for i = 1, . . . , k,
- bounded holomorphic maps α : ∆nǫ → Cg and ψ : ∆nǫ → Pg ,
such that the multi-valued period mapping
(Ω, δ) : Uǫ ∩ X → M = Hg × Cg (4.1)
of the variation of mixed Hodge structures H(ν) on Uǫ is given by the formula
 k 
X log q j Xk
log q j 
(q) = (q1 , . . . , qn ) 7→   Aj + ψ(q), A jc j + α(q) (4.2)
j=1
2πi j=1
2πi
(recall that Uǫ was defined in Section 1.3). Put a = 2π Im α, B = 2π Im ψ, and

define κ ∈ R via κ = − log ǫ. As above define the function ϕ : Rk>κ × ∆nǫ → R≥0
via
 k t  k −1  k 
X  X  X 
ϕ(x1 , . . . , xk ; q) =   
xi Ai ci + a   
xi Ai + B   xi Ai ci + a .
i=1 i=1 i=1
n
Choose any 0 < ǫ ′ < ǫ. The restriction of ϕ to Rk>κ × ∆ǫ ′ is then a normlike
function of dimension k. Let f : Rk>0 → R≥0 be the associated recession function
f = rec(ϕ). Recalling the explicit expression (3.3) for f , the conditions
∆−1 Ai ∈ S g (R) ∩ Mg (Z) , ∆−1 Ai ci ∈ Zg ,
for each i = 1, . . . , k imply that the Ai and Ai ci are themselves integral and that f
is the quotient of two homogeneous polynomials in Z[x1 , . . . , xk ]. In particular
f ∈ Q(x1 , . . . , xk ). It is clear that f is homogeneous of weight one, and by
Proposition 3.11 the function f is convex when seen as a real-valued function on
Rk>0 .
Let s be an admissible section of Pν over Uǫ ∩ X. Following Corollary 2.16

we have
− log||s|| = − log|h| + ϕ(− log |q1 |, . . . , − log |qk |; q)
on Uǫ ∩ X with h a meromorphic function on Uǫ , holomorphic on Uǫ ∩ X. As
s is locally generating over Uǫ ∩ X we have that h has no zeroes or poles on
Uǫ ∩ X. Hence there is a linear form l ∈ Z[x1 , . . . , xk ] and a holomorphic map
u : Uǫ → C∗ such that
− log |h| = l(− log |q1 |, . . . , − log |qk |) + log |u|
on Uǫ ∩ X. The image of U ǫ ′ under the map u is compact.
Put f s = f + l in Q(x1 , . . . , xk ). Then f s is again homogeneous of weight
one and convex as a function on Rk>0 . Our claim is that f s satisfies all the
requirements of Theorem 1.1. We need to show first of all that − log||s|| −
f s (− log |q1 |, . . . , − log |qk |) is bounded on U ǫ ′ ∩ X and extends continuously over
U ǫ ′ \ Dsing .
In order to see this, put ϕ0 = ϕ − f on Rk>κ × ∆nǫ . Then
− log||s||(q) = f s (− log |q1 |, . . . , − log |qk |)+log |u|+ϕ0 (− log |q1 |, . . . , − log |qk |; q)
on Uǫ ∩ X. Note that log |u| extends in a continuous and bounded manner over
the whole of U ǫ ′ . We are reduced to showing that ϕ0 (− log |q1 |, . . . , − log |qk |; q)
is bounded on U ǫ ′ ∩ X and extends continuously over U ǫ ′ \ Dsing .
For this we invoke Theorem 3.2.1. This readily gives the boundedness of ϕ0
via the map
(− log | · |, id) : (∆∗ǫ )k × ∆nǫ → Rk>κ × ∆nǫ .
Let p ∈ (D \ Dsing ) ∩ U ǫ ′ . Up to a change in the order of the variables, we can
assume that the coordinates of p satisfy q1 = 0, qi , 0 for i = 2, . . . , N. We take
a small polydisk Vǫ ′′ ⊂ U ǫ ′ of small radius ǫ ′′ with center at p such that Vǫ ′′ ∩ X
can be identified with ∆∗ǫ ′′ × ∆n−1
ǫ ′′ and hence Vǫ ∩ D can be identified with the
′′
n
divisor q1 = 0 on ∆ǫ ′′ . Write
r = (r2 , . . . , rk ) = (− log |q2 |, . . . , − log |qk |)
for q ∈ Vǫ ′′ ; then r can be assumed to move through a compact subset K ′ ⊂ Rk−1 .
n
Put K ′′ = K ′ ×∆ǫ ′ . We define functions ϕ′ : R>κ ×K ′′ → R≥0 and f ′ : R>κ ×K ′′ →
R≥0 via
ϕ′ (x1 ; r, q) = ϕ(x1 , r; q) , f ′ (x1 ; r) = f (x1 , r) .
Then both ϕ′ , f ′ are normlike of dimension one. Write
! !
A′ 0 A′′ 0
A1 = 1 , A′1 = 1 ,
0 0 0 0
with A′1 positive semidefinite of size r, and A′′1 = Ae1 positive definite of size and
rank r1 . Then it is readily verified that both rec( f ′ ) and rec(ϕ′ ) are equal to the
linear function x1 µ1 = x1 ct1 A1 c1 = f (x1 , 0, . . . , 0). Note that
ϕ0 (− log |q1 |, . . . , − log |qk |; q) = ϕ′ (− log |q1 |; r, q) − f ′ (− log |q1 |; r)
on Vǫ ′ ∩ X. We are done once we show that ϕ′ − f ′ extends continuously over

R>κ ×K ′′ . Following Theorem 3.2.3 we have that both ϕ′ −rec(ϕ′ ) and f ′ −rec( f ′ )
extend continuously over R>κ × K ′′ . As rec(ϕ′ ) = rec( f ′ ) we find the required
extension result.
The second item of Theorem 1.1 is clear. As f s is up to a linear form the
recession function of a normlike function we have that f s is convex, and by
Proposition 3.12 that f s extends as a convex, continuous homogeneous weight
one function f s : Rk≥0 → R. This finally proves items (3) and (4) of Theorem
1.1.
4.2. The Lear extension made explicit. Let s be an admissible section of

P = Pν . We recall that this means that s corresponds to an admissible
biextension variation of mixed Hodge structures h on
i X. Then s can also be seen
as a rational section of the Lear extension P, ||−|| of P over X. We can now
Xh i
compute the global Q-divisor divX (s) that represents P, ||−|| . We do this after
X
a little digression.
Write U = Uǫ ′ , V = Vǫ ′′ to reduce notation. Let µ1 , . . . , µk ∈ Q be the
coefficients of ϕ (see end of Section 3.4 for the definition), and νi = ordDi h for
i = 1, . . . , k, and ai = µi + νi . Here Di is the divisor on X given locally on U by
qi = 0. We obtain from the above proof that
− log||s||(q) = −a1 log |q1 | + ψ1 (q) (4.3)
on V ∩ X where ψ1 (q) extends continuously over V.

We say that p ∈ X is of depth k if p is on precisely k of the irreducible
divisors Di . The set Σk of points of depth k on X is a locally closed subset of
X and for k ≥ 1 they yield a stratification of D = X \ X. For p ∈ Σk take a
coordinate neighborhood U ⊂ X such that p = (0, . . . , 0) and D ∩ X is given
by the equation q1 · · · qk = 0. Theorem 1.1 yields an associated homogeneous
weight-one function f p,s ∈ Q(x1 , . . . , xk ).
Lemma 4.1. The map Σk → Q(x1 , . . . , xk ) given by p 7→ f p,s is locally constant.

Proof. Take p, U as above and let y = (0, . . . , 0, yk+1, . . . , yn) ∈ U be another

point of depth k. Let q′i = qi for i = 1, . . . , k, q′i = qi − yi for i = k + 1, . . . , n.
Then q′ are coordinates centered around y and we have
− log||s|| = f p,s (− log |q1 |, . . . , − log |qk |) + ψ p (q)

= fy,s (− log |q′1 |, . . . , − log |q′k |) + ψy (q′ )
= fy,s (− log |q1 |, . . . , − log |qk |) + ψy (qi − yi )
on U ∩ X with ψ p , ψy bounded on U ∩ X. We find that f p,s − fy,s is bounded on
Rk>κ and, being homogeneous of weight one, it vanishes identically.
In order to compute the divisor divX (s) that represents the Lear extension
of Pν over X we are interested in the behavior of the function f s : Σ1 → Q(x)
obtained from Lemma 4.1 by restricting to k = 1. Note that Σ1 = D \ Dsing .
S
Let D = dα=1 Dα be the decomposition of D into irreducible components. Take
any irreducible component Dα . Since Dα \ Dsing is connected, we deduce from
Lemma 4.1 that the function
f s,α : Dα \ Dsing → Q(x)
is constant. Its value is a homogeneous linear function which we write as
f s,α (x) = aα x, with aα ∈ Q. In this notation we find:
Corollary 4.2. Let s be a section
h ofiP corresponding to an admissible biexten-
sion variation on X. Let L = P, ||−|| be the Lear extension of P over X. Then
X
L is represented by the Q-divisor
X
d
divX (s) = a α Dα
α=1
on X.
4.3. Local integrability. Our next task is to investigate ∂∂¯ log||s|| over curves.
Proof (Proof of Theorem 1.3). We use the estimates from Theorem 3.2.3. We
assume k = n = 1, but otherwise keep the notation and assumptions from
Section 4.1. In particular we have the normlike function ϕ(x1 , q1) on R>κ × ∆ǫ
and the associated recession function f = rec(ϕ) on R>0 . Put ϕ0 = ϕ − f . Put
ϕ1 (q1 ) = ϕ0 (− log |q1 |, q1 ). By Corollary 2.16 on Uǫ ∩ X, noting that f is linear,
we have
− log||s||(q1 ) = − log |h|(q1 ) + ϕ1 (q1 )
for some meromorphic function h. Note that

1 ∂ϕ0 dq1 ∂ϕ0
∂ϕ1 = − + dq1 .
2 ∂x1 q1 ∂q1
Here ∂ϕ0 /∂q1 is smooth and bounded on Uǫ ′ , and by Theorem 3.2.3 we have a
constant c1 such that

∂ϕ0 ≤ c · x−2 .
∂x1 1 1
Hence for a smooth vector field T with bounded coefficients we find a constant
c2 such that
1
|∂ϕ1 (T )| ≤ c2 ·
(− log |q1 |)2 |q1 |
on Uǫ ∩ X. A similar argument yields
¯ 1 (T )| ≤ c2 · 1
|∂ϕ
(− log |q1 |)2 |q1 |
on Uǫ ∩ X. In particular, there is a constant c3 such that
Z
ǫ
∂ϕ1 ≤ c3 .
∂U ǫ (log ǫ)2 ǫ
Thus the residue res0 (∂ϕ1) of ∂ϕ1 at zero is zero.

Next, there exists a smooth (1, 1)-form ζ on Uǫ such that
¯ 1= 1 ∂2 ϕ0 1
∂∂ϕ dq1 dq1 + ζ .
4 ∂x21 |q1 |2
By Theorem 3.2.3 we have a constant c4 such that

2
∂ ϕ0 −3
2 ≤ c4 · x1 .
∂x1
Hence for smooth vector fields T, U with bounded coefficients we find a constant
c5 and an estimate
¯ 1 (T, U)| ≤ c5 · 1
|∂∂ϕ
(− log |q1 |)3 |q1 |2
¯ 1 is locally integrable on Uǫ .
on Uǫ ∩ X. This shows that ∂∂ϕ
4.4. Effectivity of the height jump divisor. In this section we prove Theorem
1.5. We continue again with the notation as in Section 4.1. In particular we have
U = Uǫ , we have s an admissible section of Pν on U ∩ X, and f s : Rk>0 → R the
associated homogeneous weight one function such that
− log ksk − f s (− log |q1 |, . . . , − log |qk |)
is bounded on U∩X and extends continuously over X\Dsing . Moreover f s extends
as a convex homogeneous weight one function f s : Rk≥0 → R (cf. Theorem 1.1).
It is clear that a convex homogeneous weight one function is subadditive, hence
we have the estimate
X
k
f s (x1 , . . . , xk ) ≤ f s (0, . . . , 0, xi, 0, . . . , 0) (4.4)
i=1
on Rk≥0 .
Now let φ : C → X be a map from a smooth curve, sending a point 0 in C
to p = (0, . . . , 0), and such that there exists an open neighbourhood V of 0 in C
such that φ maps V into U. We also assume that φ does not map V into D. Then
φ is given locally at 0 ∈ C by
φ(t) = (tm1 u1 , . . . , tmi ui , . . .) ,
where t is a local coordinate on C at 0, the mi are non-negative integers, and ui
are units. Write φ for the restriction of φ to V \ {0}.
Proposition 4.3. We have an equality of Q-divisors on V:

div (φ∗ s) |V = f s (m1 , . . . , mk ) · [0] ,
h i
where φ∗ s is viewed as a rational section of the Lear extension φ∗ (Pν , ||−||) .
V
Proof. It suffices to show that

− log||φ∗ s|| ∼ − f s (m1 , . . . , mk ) log |t|
on V \ {0}, where ∼ denotes that the difference is bounded and extends continu-
ously over V. As by Theorem 1.1
− log||s|| − f s (− log |q1 |, . . . , − log |qk |)
is bounded on U ∩ X we obtain the boundedness by pullback along φ. The
continuous extendability over V then follows from the boundedness combined
with the existence of a Lear extension for φ∗ (Pν , ||−||).
Proposition 4.4. We have an equality of divisors on V:

X
k
φ∗ (divX (s)) = f s (0, . . . , 0, mi, 0, . . . , 0) · [0] ,
i=1
h i
where s is viewed as a rational section of the Lear extension Pν , ||−|| .
U
Proof. This follows immediately from Corollary 4.2.
Proof (Proof of Theorem 1.5). Combining Propositions 4.3 and 4.4 one sees
that the line bundle
h i⊗−1 ∗h i
φ∗ (Pν , ||−||) ⊗ φ Pν , ||−||
C X
has a canonical non-zero rational section, whose divisor is
 
 X
k 
− f (m1 , . . . , mk ) + f s (0, . . . , 0, mi, 0, . . . , 0) · [0]
 s
i=1
on V, which is indeed independent of the choice of rational section s. This divisor

is effective by the subadditivity of f s expressed by inequality (4.4). In particular
the section is global.
Acknowledgements
We would like to thank R. Hain and G. Pearlstein for several discussions and
useful hints. We also are very grateful to the referees for a number of important
corrections and clarifications. Finally we would like to thank the hospitality of
the Mathematical Institute of Leiden University and the Instituto de Ciencias
Matemáticas where the authors could meet to work on this paper.
The first author has been partially supported by the MINECO research projects
MTM2013-42135-P and MTM2016-79400-P and ICMAT Severo Ochoa project
SEV-2015-0554 and the DFG project SFB 1085 “Higher Invariants”.
References
[1] O. Amini, S. Bloch, J. Burgos Gil, J. Fresán, Feynman amplitudes and limits of heights.
Izv. Math. 80 (2016), no. 5, 813–848.
[2] M. Asakura, Motives and algebraic de Rham cohomology. In: The arithmetic and geometry
of algebraic cycles (Banff, AB, 1998), 133–154, CRM Proc. Lecture Notes 24, 2000.
[3] O. Biesel, D. Holmes, R. de Jong, Néron models and the height jump divisor. Trans. AMS
369 (2017), 8685–8723.
[4] C. Birkenhake, H. Lange, Complex Abelian Varieties. Second edition. Grundlehren der
Mathematischen Wissenschaften 302. Springer Verlag, Berlin.
[5] S. Boyd, L. Vandenberghe, Convex Optimization. Cambridge University Press, 2009.
[6] P. Brosnan, G. Pearlstein, Jumps in the archimedean height. Preprint, arxiv:1701.05527.

[7] J.I. Burgos Gil, D. Holmes, R. de Jong, Positivity of the height jump divisor. IMRN,
https://doi.org/10.1093/imrn/rnx169
[8] J. I. Burgos Gil, J. Kramer, U. Kühn, The singularities of the invariant metric on the
line bundle of Jacobi forms. Recent Advances in Hodge Theory 45–77, (M. Kerr and
G. Pearlstein eds.) London Math. Soc. Lecture Note Ser. 427, Cambridge Univ. Press.
Cambridge, 2016.
[9] E. Cattani, Mixed Hodge structures, compactifications and monodromy weight filtrations.
In: Topics in Transcendental Algebraic Geometry 75–100, (P. Griffiths ed.), Ann. of Math.
Stud. 106, Princeton Univ. Press, Princeton, NJ, 1984.
[10] P. Deligne, Le déterminant de la cohomologie. In: Current trends in arithmetical algebraic
geometry (Arcata, Calif., 1985), Contemp. Math. 67 (1987), 93–177.
[11] P. Deligne, Equations différentielles à points singuliers réguliers, Lecture Notes in Math.,
163, Springer Verlag Berlin, 1970.
[12] A. Grothendieck. Groupes de monodromie en géométrie algébrique, Lecture Notes in
Math., 288, Springer Verlag Berlin, 1972.
[13] R. Hain, Biextensions and heights associated to curves of odd genus. Duke Math. J. 61
(1990), 859–898.
[14] R. Hain, Normal functions and the geometry of moduli spaces of curves. In: G. Farkas and
I. Morrison (eds.), Handbook of Moduli, Volume I. Advanced Lectures in Mathematics,
Volume XXIV, International Press, Boston, 2013.
[15] T. Hayama, G. Pearlstein, Asymptotics of degenerations of mixed Hodge structures. Adv.
Math. 273 (2015), 380–420.
[16] D. Holmes, R. de Jong, Asymptotics of the Néron height pairing. Math. Res. Lett. 22 no. 5
(2015), 1337–1371.
[17] M. Kashiwara, A study of variation of mixed Hodge structure. Publ. RIMS, Kyoto Univ.
22 (1986), 991–1024.
[18] K. Kato, C. Nakayama, S. Usui, SL(2)-orbit theorem for degeneration of mixed Hodge
structure. J. Algebraic Geom. 17 (2008), no. 3, 401–479.
[19] D. Lear, Extensions of normal functions and asymptotics of the height pairing. PhD thesis,
University of Washington, 1990.
[20] G. Pearlstein, C. Peters, Differential geometry of the mixed Hodge metric. Preprint
arxiv:1407.4082. To appear in Comm. Analysis and Geom.
[21] G. Pearlstein, SL2 -orbits and degenerations of mixed Hodge structure. J. Diff. Geometry
74 (2006), 1–67.
[22] G. Pearlstein, Variations of mixed Hodge structure, Higgs fields, and quantum cohomology.
manuscripta math. 102 (2000), no. 3, 269–310.
[23] M. Saito, Modules de Hodge polarisables. Publ. Res. Inst. Math. Sci. 24 (1988), 849–995.
[24] M. Saito, Mixed Hodge modules. Publ. Res. Inst. Math. Sci. 26 (1990), 221-333.
[25] C. Peters, J. Steenbrink, Mixed Hodge structures. Ergebnisse der Mathematik und ihrer
Grenzgebiete. 3. Folge. A Series of Modern Surveys in Mathematics, 52. Springer Verlag,
Berlin, 2008.
[26] R. T. Rockafellar, Convex analysis, Princeton Math. Series, vol. 28, Princeton Univ. Press,
1970.
[27] J. Steenbrink, S. Zucker, Variation of mixed Hodge structure. I. Invent. Math. 80 (1985),
489–542.

Burgos de Jong

Uploaded by

Copyright:

Available Formats

Burgos de Jong

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Burgos de Jong

Uploaded by

Copyright:

Available Formats

arXiv:1604.00686v2 [math.AG] 9 Jul 201 Forum of Mathematics, Sigma (2018), vol.

SINGULARITIES OF THE BIEXTENSION METRIC FOR

JOSÉ IGNACIO BURGOS GIL1 , DAVID HOLMES2 and

1.1. Families of curves. By way of motivation of the general results in this

where x ∈ X. Let X ֒→ X be a smooth compactification of X with D = X \ X

holds. Since h0 is continuous on U \ Dsing , this determines the behavior of h

on U \ D, where b : U \ D → R is a bounded continuous function that extends

L = Pν,µ = (ν, λµ)∗ P ,

as a metrized analytic line bundle on X. We put Pν = Pν,ν . This “diagonal”

A second example is provided by the sections determined by two relative

holds in a neighborhood V of 0 ∈ C, with b′ (t) continuous on V.

Theorem 1.1. Assume that H is a variation of torsion-free polarized pure Hodge

then the difference

3. The function f s : Rk>0 → R extends to a continuous function f s : Rk≥0 → R.

4. In the case that µ = ν, the function f s is convex as a function on Rk>0 and

where the Ai (i = 1, . . . , k) are positive semidefinite g × g matrices such

We return again to the setting where the parameter space X is of any

Choose coordinates in a neighbourhood U of x0 as before so that x0 has co-

φ(t) = (tm1 u1 (t), . . . , tmk uk (t), qk+1 (t), . . . , qn(t)),

for the height jump in our setting.

Definition 2.1. Let X be a complex variety. An (algebraic resp. analytic) Q-

If (L, r) is a Q-line bundle, a Q-rational section of (L, r) (or rational section

(s1 )⊗d2 = (s2 )⊗d1

as a section of L⊗d1 d2 . The space of rational sections of (L, r) is a torsor over

Definition 2.2 (Lear extension). Let X ⊆ X be an open immersion of smooth

T = V/Λ is a compact complex torus. It is a Kähler complex manifold, but in

is a lattice of V ∗ . The lattice Λ∨ is canonically isomorphic to the dual of the

Definition 2.3. A Poincaré (line) bundle P is a holomorphic line bundle on

1. the restriction P|T ×{[w]} is isomorphic to L[w] ;

2. the restriction P|{0}×T ∨ is trivial.

A rigidified Poincaré bundle is a Poincaré bundle together with an isomorphism

To prove the existence of a Poincaré bundle, consider the map

This map is holomorphic in z and w. Moreover, since for (µ, λ) ∈ Λ × Λ∨ ,

aP ((µ, 0), (z, w)) = exp(πw(µ)).

This last cocycle is equivalent to the cocycle (2.2). Indeed,

exp(πw(µ)) exp(πw(z + µ))−1 exp(πw(z)) = exp(2πihw, µi),

and the function z 7→ exp(πw(z)) is holomorphic in z. Thus the restriction

aP ((0, λ), (0, w)) = 1,

Proposition 2.4. A Poincaré bundle exists and is unique up to isomorphism. A

aP ((0, λ), (z, w)) = exp(πλ(z)),

Group theoretical interpretation of the Poincaré bundle. We next give a group

Proposition 2.7. The line bundle PT is a rigidified Poincaré line bundle.

3. (Normalization) k(0, 0, 1)k = 1.

Holomorphic families of complex tori. Let X be a complex manifold and T → X

that explains the notation.

Let U ⊂ X be a small enough open subset such that the restriction of T to U

(a, b) = (a1 , . . . , ag , b1, . . . , bg )

of Λ s0 such that (a1 , . . . , ag) is a complex basis of V s0 . By abuse of notation,

(b1 , . . . , bg ) = (a1 , . . . , ag )Ω (2.9)

for a holomorphic map Ω : U → Mg (C). We call Ω the period matrix of the

As before we extend the elements a∗i , b∗i , i = 1, . . . , g to flat sections of Λ over

1. the fiberwise dual tori form a holomorphic family of complex tori T ∨ → X;

3. there is a unique metric on P that induces the trivial metric on ν∗0 P = OX

Proof. Fix an open subset U ⊂ X as before. The dual family of tori T ∨ is

Abelian varieties. We now specialize to the case of polarized abelian varieties.

A polarization of a Hodge structure of type (−1, 0), (0, −1) is a non-degenerate

Variations of polarized mixed Hodge structures. Let X be a complex manifold.

1. A finite increasing filtration

of HQ = H ⊗ Q by local subsystems, called the weight filtration,