n be the n-dimensional complex hyperbolic space, repreFor any integer n > 1, let HC
sented by the unit ball in Cn equipped with the Bergman metric of constant holomorphic
sectional curvature −4/(n + 1), on which the real Lie group U(n, 1) acts in a natural way.
Given a lattice Γ in U(n, 1), we consider the quotient YΓ := Γ\HC
Let M be a CM quadratic extension of a totally real number field F of degree d and
ring of integers OF , and let G be a reductive group over F defined by a hermitian form
on M 3 of signature (n, 1) at one infinite place ι and (n + 1, 0) at the others. A subgroup
Γ ⊂ G(F ) is arithmetic, if it is commensurable with G(OF ), and we will use the same
notation for its image ι(Γ) ⊂ G(Fι ) = U(n, 1), which is by definition an arithmetic lattice.
By the Baily-Borel theorem the arithmetic quotient YΓ has a structure of a normal, quasiprojective variety, in fact defined over a number field. It is projective if, and only if, G is
We will prove below three results for arithmetic YΓ , the first one concerning the compact
case, the second dealing with the general non-compact case, and the third giving precise
and explicit conclusions in the case of (non-compact) surfaces.
A quasi-projective variety X defined over a number field k is said to be Mordellic if, and
only if, for any finite extension k ′ of k, the set X(k ′ ) of k ′ -rational points of X is finite.
Lang conjectured in [L, Conjecture VIII.1.2] (see also [T, p.xviii]) that any hyperbolic,
projective X is Mordellic; this is consistent with the general philosophy of Vojta.
Theorem 0.1. If Γ is an arithmetic cocompact subgroup of U(2, 1) such that all its torsion
elements are scalar, then the projective surface YΓ is Mordellic.
Note that even though this theorem only concerns arithmetic subgroups, because F and
M can vary, it can be applied to infinitely many pairwise non-commensurable cocompact
discrete subgroups in U(2, 1).
Our results do not apply, however, to the analogous case of (a cocompact discrete subgroup of) a unitary group defined by a division algebra of dimension 9 over an imaginary
quadratic field M with an involution of the second kind. In that case it is known that the
Albanese of YΓ is zero for any congruence subgroup Γ; this was proved by Rapoport and
Zink [RZ] under a ramification hypothesis, and later by Rogawski [R1] using a different
method, without the hypothesis.
After an earlier version of this paper was written, we came to know of Ullmo’s work [U]
on Shimura varieties of abelian type which contains Theorem 0.1 for suitable covers since
Picard modular surfaces of congruence type are in this class. We hope that our (quite
different) approach is nevertheless of interest since we do not make use of the Shafarevich
conjecture. Instead we use certain key theorems of Rogawski [R1, R2] and of Faltings
[F1, F2] (see Propositions 2.4 and 3.2). We have also come to know recently of some
related work of Yeung [Ye] on rational points on ball quotients.
Consider now the more complicated case when G is isotropic, which necessarily implies
that F = Q and M is imaginary quadratic. Its toroidal compactification XΓ is not hyperbolic even if Γ is torsion free (for example, if n = 2 then XΓ is a union of YΓ with a
finite number of elliptic curves indexed by the cusps). However, by a result of Mumford
[M, Proposition 4.2], for Γ neat, XΓ is always of log general type. This implies that for
Γ sufficiently small, XΓ is of general type. The Bombieri-Lang conjecture predicts that
the points of XΓ over a given number field are not Zariski dense. This we establish in
Proposition 3.2 using a deep Theorem due to Faltings, when XΓ is smooth and does not
admit a dominant map to its Albanese variety, in particular when its irregularity is > n.
This allows us to solve an alternative of Ullmo and Yafaev [UY] regarding the Lang locus.
Theorem 0.2. For all Γ ⊂ U(n, 1) arithmetic and sufficiently small, YΓ is Mordellic.
Keeping the assumption that M is imaginary quadratic, say of fundamental discriminant
−D, and supposing in addition that n = 2, we restrict our attention to the corresponding
locally symmetric spaces YΓ , called Picard modular surfaces. For every ideal N ⊂ OM we
consider the congruence subgroups Γ(N) and Γ1 (N) (see Definition 1.5). We state now our
main theorem.
Theorem 0.3. Let D | D be the conductor of a simplest Hecke character over M (see (12)).
Γ1 (D) , if D ∈
/ {3, 4, 7, 8, 11, 15, 19, 20, 23, 24, 31, 39, 43, 47, 67, 71, 163},
(i) Let Γ = Γ(D) , if D ∈ {8, 15, 20, 23, 24, 31, 39, 47, 71},
Γ(D2 ) , if D ∈ {3, 4, 7, 11, 19, 43, 67, 163}.
Then YΓ is Mordellic, while XΓ is a minimal surface of general type.
(ii) Let N > 2 be a prime inert in M and not equal to 3 when D = 4. Then YΓ(N )∩Γ1 (D)
is Mordellic, while XΓ(N )∩Γ1 (D) is a minimal surface of general type.
At the heart of our proof are some arithmetic computations yielding, for each imaginary
quadratic field M , explicit congruence subgroups Γ such that XΓ does not admit a dominant
map to its Albanese variety. A geometric ingredient of the proof is a result of Holzapfel et
al that XΓ is of general type, implying by a theorem of Nadel [N] that YΓ does not contain
curves of genus ≤ 1.
In a related paper [DR], we will investigate further the structure of the Albanese variety
of Picard modular surfaces XΓ for specific families of congruence subgroups Γ, and exhibit
the abelian varieties B(p) of Gross as Albanese quotients (up to isogeny) and also explain
their relationship to the elliptic curves over the cusps. In addition, we will exploit the
Albanese quotients coming from residual automorphic forms, and present an alternate
method to deal with the possible curves on genus ≤ 1 by means of a formal immersion at
special points on the curves at infinity by analyzing the Fourier-Jacobi coefficients. This
leads to another approach for proving finiteness of rational points on the open surface. In
a third paper we will analyze the situation over Spec(OM ) with a view towards showing
the paucity, even lack, of rational points on the open surface for a suitable infinite class of
congruence subgroups. The ultimate aim of our program is to establish a weaker analogue
of Mazur’s theorem on modular curves, with a consequence for the boundedness of torsion
for principally polarized abelian 3-folds with multiplication by OM .
Acknowledgments. We would like to thank Don Blasius, Laurent Clozel, Jean-Fran¸cois Dat,
Najmuddin Fakhruddin, Dick Gross, Haruzo Hida, Barry Mazur, David Rohrlich, Matthew Stover
and Shing-Tung Yau for helpful conversations. In fact it was Fakhruddin who suggested our use of
Lang’s conjecture for abelian varieties. Needless to say, this Note owes much to the deep results
of Faltings. Thanks are also due to Serge Lang (posthumously), and to John Tate, for getting one
of us interested in the conjectural Mordellic property of hyperbolic varieties. Finally, we are also
happy to acknowledge partial support the following sources: the Agence Nationale de la Recherche
grants ANR-10-BLAN-0114 and ANR-11-LABX-0007-01 for the first author (M.D.), and from the
NSF grant DMS-1001916 for the second author (D.R.).
1. Basics: Lattices, general type and neatness
¯ = Γ/Γ ∩ U(1) denote its
Definition 1.1. Given a discrete subgroup Γ ⊂ U(n, 1) we let Γ
image in the adjoint group PU(n, 1) = U(n, 1)/U(1), where U(1) is centrally embedded in
U(n, 1). We put YΓ¯ := YΓ = Γ\HC
¯ ⊂ PU(n, 1) = PSU(n, 1) is the image of a discrete
Conversely any discrete subgroup Γ
¯ ∩ SU(n, 1).
subgroup of U(n, 1), namely U(1)Γ
Lemma 1.2. Let Γ be a lattice in U(n, 1).
(i) The analytic variety YΓ is an orbifold and one has the following implications:
¯ torsion free ⇒ YΓ is a hyperbolic manifold.
Γ neat ⇒ Γ torsion free ⇒ Γ
¯ is torsion free. Then the natural projection Hn → YΓ is an ´etale
(ii) Assume that Γ
¯ Moreover for every finite index normal
covering with deck transformation group Γ.
¯ Γ
¯ ′.
subgroup Γ′ of Γ the natural morphism YΓ′ → YΓ is an ´etale covering of group Γ/
n is a compact group, hence its intersection
Proof. The stabilizer in U(n, 1) of any point of HC
with the discrete subgroup Γ is finite, showing that YΓ is an orbifold. If Γ is neat, then no
element of it has a non trivial root of unity as an eigenvalue, in particular Γ is torsion free.
¯ is torsion free too. Under the latter assumption,
Since Γ ∩ U(1) is finite, this implies that Γ
n , and Γ
¯ acts freely and properly discontinuously on it, hence
Γ ∩ U(1) acts trivially on HC
n is simply connected, it is a universal covering space of Y with
YΓ is a manifold. Since HC
group Γ. In particular, YΓ is hyperbolic. The last claim follows from the exact sequence:
¯′ → Γ
¯ → Γ/
¯ Γ
¯ ′ → 1.
¯ is cocompact and torsion free. Then the projective
Proposition 1.3. Assume that Γ
variety YΓ is of general type and can be defined over a number field.
n implies by the Kodaira embedProof. The existence of the positive Bergman metric on HC
ding theorem that YΓ has ample canonical bundle, which results in YΓ being of general type;
it even implies that any subvariety is of general type. For surfaces one may alternately
use the hyperbolicity of YΓ to rule out all the cases in the Enriques-Kodaira classification
where the Kodaira dimension is less that 2, thus showing that YΓ is of general type.
Calabi and Vesentini [CV] have proved that YΓ is locally rigid, hence by Shimura [Sh1]
it can be defined over a number field.
For the convenient of the reader, we will provide a second, more direct proof when n = 2
and Γ is arithmetic, based on Yau’s algebra-geometric characterization of compact K¨ahler
2 . Since Y has an ample canonical bundle it can be embedded in
surfaces covered by HC
some projective space, hence is algebraic over C by Chow. Since YΓ¯ is uniformized by HC
the Chern numbers c1 , c2 of its complex tangent bundle satisfy the relation c21 = 3c2 . Since
everything can be defined algebraically, for any automorphism σ of C, the variety YΓ¯σ also
has ample canonical bundle and cσ2
1 = 3c2 . By a famous result of Yau [Y, Theorem 4], this
¯ σ \H2 for some cocompact discrete
is equivalent to the fact that YΓ¯σ may be realized as Γ
irreducible torsion free subgroup Γ
¯ is arithmetic, it has infinite index in its commensurator in PU(2, 1), denoted
Since Γ
¯ For every element g ∈ Comm(Γ)
¯ there is a Hecke correspondence
YΓ¯ ← YΓ∩g
¯ −1 Γg
¯ −→ Yg Γg
¯ −1 ∩Γ
¯ → YΓ
and the correspondences for g and g ′ differ by an isomorphism YgΓg
¯ −1 ∩Γ
¯ −→ Yg ′ Γg
¯ ′−1 ∩Γ
over YΓ¯ if, and only if, g ∈ Γg. By Chow (1) is defined algebraically, hence yields a
correspondence on YΓ¯σ = YΓ¯ σ :
YΓ¯ σ ← YΓ¯ 1 −→ YΓ¯ 2 → YΓ¯ σ ,
¯ 1 and Γ
¯ 2 of Γ
¯ σ . By the universal property of the covering
for some finite index subgroups Γ
2 , the middle isomorphism is given by an element of g ∈ PU(2, 1) ≃ Aut(H2 ).
space HC
¯ 1 gσ−1 , and by applying
Since Aut(HC /YΓ¯ i ) = Γi (i = 1, 2), it easily follows that Γ2 = gσ Γ
¯ σ gσ . It follows that gσ ∈ CommG (Γσ ) and one can check
¯1 = Γ
¯ σ ∩ g −1 Γ
σ −1 one sees that Γ
¯ σ gσ if, and only if, g ′ ∈ Γg.
¯ Therefore Comm(Γ
¯ σ )/Γ
¯ σ ≃ Comm(Γ)/
¯ Γ
¯ is infinite
that gσ′ ∈ Γ
too, which by a major theorem of Margulis implies that Γσ is arithmetic, providing an
alternative proof of a result of Kazhdan.
Consider now the action of Aut(C) on the set of equivalence classes of cocompact arithmetic subgroups Γ (modulo their center and up to conjugation by an element of U(2, 1)).
The group U(2, 1) has only countably many Q-forms, classified by central simple algebras
of dimension 9 over a CM field, endowed with an involution of a second kind and verifying
some conditions at infinity (see [PR, pp. 87-88]). Finally, there are only countably many
arithmetic subgroups for a given Q-form, since those are all finitely generated and contained in their common commensurator, which is countable. It follows that Γ is fixed by
an open subgroup of Aut(C), allowing one to conclude that YΓ is defined over a number
It is a well known fact that any compact orbifold admits a finite cover which is a manifold.
We will now provides such a cover explicitly for arithmetic quotients. Recall that an
arithmetic subgroup Γ ⊂ G(F ) is a congruence subgroup if there exists an integer N such
that Γ contains the principal congruence subgroup of level N , defined as:
Γ(N ) = ker (G(OF ) → G(OF /N OF )) .
The following lemma is well-known (see [H1, Lemma 4.3]).
Lemma 1.4. For any integer N > 2 the group Γ(N ) is neat.
Since G is defined by a hermitian form on M n+1 , we have an embedding G(OF ) ,→
GL(n + 1, OM ), through which we may view elements of G(F ) as (n + 1) × (n + 1) matrices.
This allows us to define the following finer congruence subgroups.
Definition 1.5. For every ideal N ⊂ OM we define the congruence subgroup Γ(N) (resp.
Γ1 (N)) as the kernel (resp. the inverse image of upper unipotent matrices) of the composed
G(OF ) ,→ GL(n + 1, OM ) → GL(n + 1, OM /N).
Lemma 1.6. Suppose that n = 2 and that M is an imaginary quadratic field of fundamental
discriminant −D ∈
/ {−3, −4, −7, −8, −24}. Then Γ1 ( −D) is neat.
Proof. Suppose that the subgroup of C× generated by the eigenvalues of some γ ∈ Γ1 ( −D)
contains a non-trivial root of unity ζ. If γ is elliptic then it is necessarily of finite order.
2 ⊂ P2 (C) hence is conjugated in GL(3, C) to a
Otherwise γ fixes a boundary point of HC
( α¯ ∗ ∗ )
, where β is necessarily a root of unity. Hence, it both cases,
matrix of the form 0 β −1
0 0 α
one may assume that ζ is an eigenvalue of γ.
By the Cayley-Hamilton theorem we have [M (ζ) : M ] ≤ 3 and since D ̸= 7 we may
assume (after possibly lifting γ to some power) that ζ has order 2 or 3. By the congruence
condition, each prime p dividing D has to divide also the norm of ζ − 1, hence D can be
only divisible by the primes 2 or 3. Hence D ∈ {3, 4, 8, 24} leading to a contradiction. 2. Irregularity of arithmetic surfaces
Non-vanishing (and unboundedness) of irregularity for sufficiently small arithmetic subgroups has been known since the works of Kazhdan [K] and Shimura [Sh2, Theorem 8.1].
The starting point for the arithmetic application of this paper was our knowledge that
Rogawski’s classification [R1, R2] of cohomological automorphic forms on G contributing
to H1 (YΓ , C) allowed one to prove such result for explicit congruence subgroups Γ which
can be chosen in various ways.
Throughout this section we assume that n = 2.
2.1. Automorphic forms contributing to the irregularity. Given a neat congruence subgroup Γ, we denote by q(YΓ ) the irregularity of YΓ , given by the dimension of
H0 (YΓ , Ω1YΓ ).
Fixing a maximal compact subgroup K∞ ≃ (U (2) × U (1)) × U (3)d−1 of the reductive
Lie group G∞ = G(F ⊗Q R) ≃ U(2, 1) × U(3)d−1 , we obtain a decomposition:
H1 (YΓ , C) ≃
H1 (Lie(G∞ ), K∞ ; π∞ )⊕m(π∞ ,Γ) ,
where π∞ runs over irreducible unitary representations of G∞ occurring with multiplicity
m(π∞ , Γ) in the discrete spectrum of L2 (Γ\G∞ ). When Γ is cocompact, the entire L2 spectrum is discrete and this decomposition follows from [BW, Chap.XIII]. When Γ is a
non-cocompact lattice, one gets such a decomposition a priori only for the L2 -cohomology
of YΓ . However, it is known for Picard modular surfaces that H1 (YΓ , C) is isomorphic to
the middle intersection cohomology (in degree 1) of the Baily-Borel compactification of YΓ ,
which is in turn isomorphic to the L2 -cohomology H1(2) (YΓ , C) (see [MR, §1]). In addition,
H1(2) (YΓ , C) is isomorphic to H1 (XΓ , C), where XΓ is a smooth toroidal compactification of
YΓ (see loc. cit.). In particular, H1 (YΓ , C) admits a pure Hodge structure and its dimension
is given by 2q(YΓ ).
At the distinguished Archimedean place ι where G(Fι ) = U(2, 1) there are exactly two
irreducible non-tempered unitary representations of G(Fι ), denoted πn+ and πn− , with nonzero relative Lie algebra cohomology in degree 1, while at the remaining infinite places,
the only irreducible unitary representation with non-zero relative Lie algebra cohomology
in degree 0 is the trivial one.
The restrictions to C× of the Langlands parameters of πn+ and πn− are given by:
z¯ 0
z 7→ 0 z/¯
z 0  ∈ L G0 = GL3 (C),
z −1
and its complex conjugate (see [La, p.62]).
We will now introduce the adelic setting which is better suited for computing the irregularity. For K a neat open compact subgroup of G(AF,f ), where AF,f denotes the ring of
finite adeles of F , we consider the adelic quotient
YK := G(F )\G(AF )/K∞ K.
Since G1 := ker(det : G → M 1 ) is simply connected and G1∞ is non-compact, by strong
approximation (see [PR, Theorem 7.12]) G1 (F ) is dense in G1 (AF,f ). It follows that YK is
a finite disjoint union, indexed by the class group
π0 (YK ) ≃ A1M /M 1 det(K)M∞
of surfaces YΓ for some neat congruence subgroups Γ ⊂ G(F ) and (2) can be rewritten as:
H1 (YK , C) ≃
H1 (Lie(G∞ ), K∞ ; π∞ )⊕m(πf ,K) ,
where π = π∞ ⊗ πf runs over all automorphic representation of G(AF ). By the above
description of π∞ , and Rogawski’s multiplicity one theorem [R2] one deduces the formula:
q(YK ) =
dim(πfK ) =
dim(πfK ),
π automorphic
π∞ =πι ⊗1⊗d−1 ,πι ≃πn
π automorphic
π∞ =πι ⊗1⊗d−1 ,πι ≃πn
where 1⊗d−1 denotes the trivial representation of U (3)d−1 .
2.2. Rogawski’s theory. Rogawski [R1, R2] gives an explicit description, in terms of
global Arthur packets, of the automorphic representations π of G(A) such that πι ≃ πn±
and πv = 1 at all Archimedean places v ̸= ι, which we will now present.
Let G′ denote the quasi-split unitary group associated to M/F , so that G is an inner
form of G′ (note that G ≃ G′ only for d = 1).
Let λ be a unitary Hecke character of M whose restriction to F is the quadratic character
associated to M/F , and let ν be a unitary character of A1M /M 1 .
At a place v of F which does not split in M , the local Arthur packet Π′ (λv , νv ) consists of
a square-integrable representation πs (λv , νv ) and a non-tempered representation πn (λv , νv )
of G′ (Fv ). Those can be described (see [R1, §12.2]) as the unique sub-representation and
the corresponding (Langlands) quotient of the induction of the character on the standard
upper-triangular Borel subgroup B(Fv ) which is trivial on the unipotent subgroup and
given on the diagonal torus T (Fv ) by:
α, β, α−1 ) 7→ λv (¯
α)|α|Mv νv (β), where α ∈ Mv× , β ∈ Mv1 .
If one considers unitary induction, then one has to divide the above character by the square
root of the modular character of B(Fv ), that is to say by (¯
α, β, α−1 ) 7→ |α|Mv .
At a place v of F which splits in M , G′ (Fv ) (resp. (M ⊗F Fv )1 ) can be identified with
GL(3, Mw ) (resp. Mw× ) where w is a place of M dividing v. The local Arthur packet
Π′ (λv , νv ) has a unique element πn (λv , νv ) which is induced from the character:
 h2
∗  7→ λw (det(h2 ))| det(h2 )|3/2
w νw (h1 )
0 0
of a maximal parabolic of G′ (Fv ) (see [R2, §1]).
For almost all v, πn (λv , νv ) is necessarily unramified. We set
Π′ (λ, ν) = ⊗v πv |πv ∈ Π′ (λv , νv ) for all v , and πv = πn (λv , νv ) for almost all v .
Recall that a CM type of M is the choice, for each Archimedean places v of F , of an
isomorphism M ⊗F,v R ≃ C. Suppose that λ (resp. ν) is algebraic of weight 1 (resp. −1)
relatively to a CM type Φ of M , in the sense that:
∏ z¯v
λ∞ (z) =
, for all z ∈ M∞ resp. ν∞ (z) =
zv , for all z ∈ M∞
|zv |
Denote by Ξ the set of such pairs (λ, ν).
Theorem 2.1 (Rogaswski [R1, R2]).
(i) For every (λ, ν) ∈ Ξ, Π′ (λ, ν) is a global
Arthur packet for G′ such that for all infinite v, πn (λv , νv ) = πn+ or πn− .
(ii) Π′ (λ, ν) can be transferred to an Arthur packet Π(λ, ν) on G such that Π(λv , νv ) =
{1} at all Archimedean places v ̸= ι, and Π(λv , νv ) = Π′ (λv , νv ) at the remaining
(iii) Denote by W (λνM ) ∈ {±1} the root number of the weight 3 algebraic Hecke character λνM , where νM (z) = ν(¯
z /z) is the base change character, and by s(π) the
number of finite places v such that πv ≃ πs (λv , νv ). Then
π ∈ Π(λ, ν) is automorphic if, and only if, W (λνM ) = (−1)d−1+s(π) .
(iv) Any automorphic representation π of G(A) such that πι ≃ πn± and πv = 1 at all
Archimedean places v ̸= ι, belongs to Π(λ, ν) for some (λ, ν) ∈ Ξ.
Proof. Let H = U(2) × U(1) be the unique elliptic endoscopic group, shared by G′ and all
its inner forms over F . The embedding of L-groups L H ,→ L G = L G′ depends on the choice
of a Hecke character µ of M , whose restriction to F is the quadratic character associated to
M/F , and allows one to transfer discrete L-packets on H to automorphic L-packets on G
(see [R2, §13.3]). The character µ being fixed, any couple of characters (λ, ν) ∈ Ξ uniquely
determine a (one-dimensional) character of H, whose endoscopic transfer is Π′ (λ, ν) (see
[R2, §1]).
Denote by WF (resp. WM ) the global Weil group of F (resp. M ). By loc.cit., the
restriction to WM of the global Arthur parameter
WF × SL(2, C) →
G = GL(3, C) o Gal(M/F )
of Π(λ, ν) is given by the 3-dimensional representation (λ ⊗ St) ⊕ (νM ⊗ 1), where St (resp.
1) is the standard 2-dimensional (resp. trivial) representation of SL(2, C). Comparing this
with the local parameter at infinity (3) yields πn (λv , νv ) = πn± at each infinite place v.
It follows that for every archimedean place v, Π′ (λv , νv ) is a packet containing a discrete
series representation of G′v , and hence by [R1, §14.4], there will be a corresponding Arthur
packet Π(λ, ν) of representations of G(AF ) such that at any archimedean place v ̸= ι,
Π(λv , νv ) is a singleton consisting of a finite-dimensional representation of the compact
real group G(Fv ) = U(3). In the notation of [R2, p.397] the representations πn+ and πn−
have parameters (r, s) = (1, −1) and (r, s) = (0, 1), respectively, and hence, by the recipe
on the same page, the highest weight of the associated finite-dimensional representation
equals (1, 0, −1). Therefore at every Archimedean v ̸= ι we have Π(λv , νv ) = {1}.
So far we have established (i) and (ii), while (iii) is the content of [R2, Theorem 1.1].
Conversely, any π as in (iv) is discrete, hence belongs to a Arthur packet Π on G, which
can be transferred to an Arthur packet Π′ on G′ (see [R1, Proposition 14.6.2] and [R1,
§14.4]). By definition Πv = Π′v at v = ι and at all the finite places v. In particular
πn± ∈ Πι = Π′ι , hence Π′ arises by endoscopy from H, that is to say equals Π′ (λ, ν) for
some unitary Hecke character λ of M whose restriction to F is the quadratic character
associated to M/F , and some unitary character ν of A1M /M 1 (see [R1, Theorem 13.3.6]).
Since Πv = {1} for all Archimedean places v ̸= ι, by the above mentioned recipe Π(λv , νv )
contains either πn+ or πn− , implying that λ (resp. ν) is algebraic of weight 1 (resp. −1)
relative to a unique choice of a CM type Φ of M .
2.3. Levels of induced representations. Let p be a prime of F divisible by a unique
prime P of M and let Fq be the residue field OF /p. In this section we exhibit open compact
subgroups K of G(Fp ) for which πn (λp , νp ) (resp. πs (λp , νp )) admit a non-zero K-invariant
subspace, and compute in some cases the exact dimension of this space.
For every integer n ≥ 1, we define the open compact subgroup K(Pn ) (resp. K1 (Pn )) of
G(Fp ) as the kernel (resp. the inverse image of upper unipotent matrices) of the composed
G(OF,p ) ,→ GL(n + 1, OM,P ) → GL(n + 1, OM /Pn ).
Lemma 2.2. Let n ≥ 1 be an integer such that the character (7) is trivial on K1 (Pn ) ∩
T (Fp ). Then both πn (λp , νp ) and πs (λp , νp ) have non zero fixed vectors under K1 (Pn ).
Proof. Let J denote the Jacquet functor sending admissible G(Fp )-representations to admissible T (Fp )-representations. The Jacquet functor is exact and its basic properties imply:
J(πs (λp , νp )) : (α
¯ , β, α−1 ) 7→ λp (¯
α)νp (β)|α|Mp = λp (¯
α)νp (β)|α|Mp · |α|Mp ,
J(πn (λp , νp )) : (α
¯ , β, α−1 ) 7→ λp (¯
α)νp (β)|α|Mp = λp (α−1 )νp (β)|α|Mp · |α|Mp .
One knows that the pro-p Iwahori subgroup K1 (Pn ) admits an Iwahori decomposition:
¯ (Fp )),
K1 (Pn ) = (K1 (Pn ) ∩ N (Fp )) · (K1 (Pn ) ∩ T (Fp )) · (K1 (Pn ) ∩ N
¯ (Fp )) denotes the unipotent of the standard (resp. opposite) Borel
where N (Fp ) (resp. N
containing T (Fp ). This is proved for the principal congruence subgroup K(Pn ) in [Cs,
Proposition 1.4.4] and the extension to K1 (Pn ) is an easy exercise. Now by [Cs, Proposition
3.3.6], given any admissible G(Fp )-representation V , one has a canonical surjection:
V K1 (P
J(V )K1 (P
n )∩T (F
Since both characters in (9) are trivial on K1 (Pn ) ∩ T (Fp )), the claim follows.
Lemma 2.3. Suppose that p is inert in M and that (λp , νp ) is unramified. Then the
dimension of the K(p)-fixed subspace of πs (λp , νp ) (resp. πn (λp , νp )) equals q 3 (resp. 1).
Proof. Since (λp , νp ) is unramified, restriction to the standard hyperspecial maximal compact subgroup Kp of G(Fp ) yields, by Iwasawa decomposition, the following exact sequence:
0 → πs (λp , νp )|Kp → IndB(F
(1) → πn (λp , νp )|Kp → 0.
p )∩Kp
The subspace of K(p)-invariant vectors in IndB(F
(1) identifies naturally with the
p )∩Kp
space of C-valued functions on the set:
(B(Fp ) ∩ Kp ) \Kp /K(p) ≃ B(Fq )\G(Fq ),
on which Kp /K(p) = G(Fq ) acts by right translation. By Iwahori decomposition, since
G(F )
G(Fq ) has rank one, the representation IndB(Fqq ) (1) has exactly two irreducible constituents
which are the trivial representation and the Steinberg representation. Since both πn (λp , νp )K(p)
and πs (λp , νp )K(p) are non-zero by Lemma 2.2, and since πs (λp , νp )Kp = 0, it follows that
πn (λp , νp )K(p) (resp. πs (λp , νp )K(p) ) is isomorphic to the trivial (resp. Steinberg) representation of G(Fq ), hence its dimension equals 1 (resp. q 3 ).
2.4. Irregularity growth. The positivity of q(YΓ ) is an essential ingredient in the proof
of our Diophantine results. Each step of the proof of Proposition 2.4 can be carried out
explicitly providing a precise level Γ, depending on M , at which q(YΓ ) > r. After the
completion of the work on this paper, we learned of Marshall’s interesting work [Ma] giving
sharp asymptotic bounds for q(YΓ ) when Γ shrinks, also by using Rogawski’s theory.
Proposition 2.4. For any arithmetic subgroup Γ of G(F ) and for any r > 0, there is a
finite index torsion free subgroup Γ′ of Γ such that q(YΓ′ ) > r.
Proof. Note that it suffices to find a neat congruence subgroup Γ′ such that q(YΓ′ ) > r,
since then, for any arithmetic Γ, the natural morphism YΓ∩Γ′ → YΓ′ is finite and surjective,
hence q(YΓ∩Γ′ ) ≥ q(YΓ′ ) > r.
Lemma 2.5. For any CM extension M/F and any CM type Φ, there exists an algebraic
Hecke character λ of weight 1 and CM type Φ.
∏ z¯v
× given by λ (z) =
Proof. Consider the character on M∞
. Since M/F is a CM
|zv |
× m
extension, the index m of (OF× )2 in OM
is finite, and λ∞ is trivial on (OM
) . By [C,
× m
Th´eor`eme 1] there exists an open compact subgroup U of AM,f such that U ∩OM ⊂ (OM
) ,
hence λ∞ can be extended (trivially) to M U M∞ . Finally, since AM /M U M∞ is a finite
abelian (class) group, there exists a character λ of A×
M /M extending λ∞ .
Let Π(λ, λ−1
) be the global Arthur packet on G associated to a character λ as in the
|M 1
lemma, and let v0 be a finite place of F which does not split in M . Choose a open compact
subgroup K(λ) = v K(λ)v of G(AF,f ) such that πs,v0 0 ̸= 0 and πn,v v ̸= 0 for all finite
places v, with K(λ)v being the standard hyperspecial maximal compact for all v ̸= v0
relatively prime to the conductor of λ.
Choose a finite place p of F inert in M relatively prime to v0 and to the conductor of
λ, such that q = |OF /p| > r. Let K(λ, p) be the subgroup of K(λ) with the maximal
compact K(λ)p replaced by K(p).
Consider an element π = ⊗v πv ∈ Π(λ, λ−1
) such that πι = πn± , πv = 1 for every infinite
|M 1
place v ̸= ι, πp = πs (λp , νp ), πv = πn,v for every finite v ̸= p, v0 , and finally:
, if W (λ3 ) = (−1)d , and
πv0 =
, if W (λ3 ) = (−1)d−1 .
By Theorem 2.1(iii), π is automorphic, and by Lemma 2.3 we have dim(πf
) ≥
q . By (5), for every character χ of the finite abelian group π0 (YK(λ,p) ), one still has
dim(πf ⊗ χ)K(λ,p) ≥ q 3 . By (6) we have q(YK(λ,p) ) ≥ q 3 |π0 (YK(λ,p) )|, hence there must exist
a connected component YΓ′ of YK(λ,p) such that q(YΓ′ ) ≥ q 3 > r.
2.5. Irregularity at low level. For the rest of this section M is imaginary quadratic.
˜ be an open compact subgroup of G(A
˜ Q,f ), where G
˜ ⊃ G is the group of unitary
Let K
˜ ∩ G(AQ,f ). Then Shimura variety
similitudes, and let K = K
SK˜ (C) = G(Q)\H
C × G(AQ,f )/K
has a canonical model SK˜ over its reflex field M and the set of geometrically connected com¯ ) ≃ π0 (S ˜ (C)) is a principal homogeneous space under Gal(M ′ /M )
ponents π0 (SK˜ ×M M
for some abelian extension M ′ of M (see [Go, §4]). Hence they are all Galois conjugates,
and in particular share the same irregularity.
By [Go, Lemma 2.4] the identity component of SK˜ (C) can be identified with YΓ , where
˜ · G(R)).
Γ = G(Q)
∩ (K
Since for any g ∈ G(R)
= GU(2, 1) we have ν(g)3 = | det(g)|2 > 0, implying ν(g) ∈ R×
b R = {1}. Hence Γ = G(Q) ∩ (K · G(R)) and the identity
we deduce that ν(Γ) ⊂ Q ∩ Z
component of YK can also be identified with YΓ . Therefore the connected components of
YK are a subset of those of SK˜ (C) and thus share the same irregularity.
Suppose that Γ is neat. Using Theorem 2.1(iii) one can easily transform (6) into:
2q(YΓ ) =
dim(πfK )(1 + W (λνM )(−1)s(π) ),
π0 (YK ) π∈Π(λ,ν)
π∞ ≃πn
where a (finite order) character χ of π0 (YK ) sends (λ, ν) ∈ Ξ to (λχ−1
M , νχ) ∈ Ξ. Note that
this action preserves the root number W (λνM ).
Proposition 2.6. Any Γ as in Theorem 0.3 is neat and q(YΓ ) > 2.
˜ of G(A
˜ Q,f ) such that Γ equals G(Q)
Proof. There exists a compact open subgroup K
(G(R) · K) as well as G(Q) ∩ (G(R) · K), where K = K ∩ G(AQ,f ).
Let λ be a unitary simplest Hecke character as in [Ya, p.88] of conductor:
, if D ̸= 3 is odd,
2 −D , if 8 divides D,
, if D = 3,
 −2D , otherwise.
In the first two cases those are the canonical characters studied in Rohrlich [Roh].
By definition, (λ, λ−1
) ∈ Ξ and is trivial on K1 (D) ∩ T (AQ,f ). Lemma 2.2 implies that:
|M 1
K1 (D)
̸= 0, for all π ∈ Π(λ, λ−1
|M 1
In case (ii) we fix a prime p dividing D and π = ⊗v πv ∈ Π(λ, λ−1
) such that πv =
|M 1
πn (λv , λ−1
) for all v ̸= p, N , πN = πs (λN , λ−1
1 ) and
πn (λp , λ−1 1 ) , if W (λ3 ) = −1,
πp =
πs (λp , λ−1 ) , if W (λ3 ) = 1.
|M 1
Since Γ(N ) is neat by Lemma 1.4, we can apply (11) which combined with Lemma 2.3
K(N )
q(YΓ(N )∩Γ1 (D) ) ≥ dim(πN
) ≥ N 3 ≥ 3.
We now turn to case (i) and suppose first that M has class number h ≥ 3. For any class
character ξ one has (λξ, λ−1
) ∈ Ξ giving h pairwise distinct elements in Ξ/b
π0 (YK1 (D) ).
|M 1
Fix a prime p | D and consider π = ⊗v πv ∈ Π(λξ, λ−1
) such that πv = πn (λv ξv , λ−1
) for
|M 1
all v ̸= p and
πn (λp ξp , λ−1 1 ) , if W (λ3 ) = 1,
πp =
πs (λp ξp , λ ) , if W (λ3 ) = −1.
|M 1
Since Γ1 (D) is neat by Lemma 1.6, one can apply (11) which combined with (13) yields
q(YΓ1 (D) ) ≥ h ≥ 3.
If M is one of the 18 imaginary quadratic fields of class number 2, then its fundamental
discriminant D has (exactly) two distinct prime divisors p < q. For each simplest character
λ on M , consider π ∈ Π(λ, λ−1
) such that πv = πn (λv , λ−1
) for all v ̸= p, q and
|M 1
(πn (λp , λ−1 1 ), πn (λq , λ−1 1 )) or (πs (λp , λ−1 1 ), πs (λq , λ−1 1 )) , if W (λ3 ) = 1,
(πp , πq ) =
(πn (λp , λ−1 ), πs (λq , λ−1 )) or (πs (λp , λ−1 ), πn (λq , λ−1 )) , if W (λ3 ) = −1.
|M 1
|M 1
|M 1
|M 1
If D ̸= 24 then Γ1 (D) is neat by Lemma 1.6 and (11) implies that q(YΓ1 (D) ) ≥ 2 · 2 = 4. If
D = 24 then Γ(D) is neat by Lemma 1.4, since 4 divides D, and again q(YΓ(D) ) ≥ 4.
Finally, we consider the nine imaginary quadratic fields of class number 1.
For D ∈ {7, 11, 19, 43, 67, 163} there is a unique simplest character λ (the canonical one).
Any character of (1 + −DOM /1 + DOM ) ≃ (Z/D) lifts to a finite order Hecke character
ξ of M with trivial restriction to Q, hence (λξ, λ−1
) ∈ Ξ. Let π = ⊗v πv ∈ Π(λξ, λ−1
) be
|M 1
|M 1
such that πv = πn (λv ξv , λ−1
) for all v ̸= D and
πn (λD ξD , λ−1 1 ) , if W (λ3 ) = 1,
πD =
πs (λ ξ , λ−1 ) , if W (λ3 ) = −1.
D D |M 1
Since Γ(D) is neat by Lemma 1.4, by (11) we get q(YΓ(D) ) ≥ D · dim(πD ) ≥ D.
For D = 3 the same argument with D2 instead of D, implies that q(YΓ(9) ) ≥ 3.
For D = 4 (resp. D = 8) the group Γ(8) (resp. Γ(2 −8)) is neat by Lemma 1.4 and it
is an easy exercise in class field theory to show that there are at least 3 weight one Hecke
characters on M whose restriction to Q is the quadratic character attached to M , and
whose conductor divides 8 (resp. 2 −8). It follows then from (11) and (13) that for D = 4
(resp. D = 8) one has q(YΓ(8) ) ≥ 3 (resp. q(YΓ(2√−8) ) ≥ 3).
Remark 2.7.
(i) The computation of the smallest level K for which there exists an
automorphic representation π ∈ Π(λ, ν) such that πfK ̸= 0 is analyzed in detail in
[DR]. In particular, if λ is a canonical character, we check that the level subgroup
at any p | D is precisely the one conjectured by B. Gross, namely the index 2
subgroup of the maximal parahoric subgroup with reductive quotient PGL(2).
(ii) One consequence of Rogawski’s theory is that the Albanese variety is of CM type
for any congruence subgroup (see [MR]). If M is imaginary quadratic and D
is prime, we will show in [DR] that the factor of the Albanese corresponding to
the canonical character λ turns out (at an appropriate prime to D level) to be
isogenous to the CM abelian variety B(D) defined by B. Gross in [G].
(iii) When Γ is not a congruence subgroup, there are examples of C. Schoen where the
Albanese is not of CM type (see [Sc]).
3. Mordellicity
We will deduce our main theorems from a more general proposition which is a consequence of the following powerful result of Faltings on the rational points of subvarieties of
abelian varieties.
Theorem 3.1 (Faltings [F2], [V]). Suppose A is an abelian variety over a number field k,
Z ⊂ A a closed subvariety. Then there are finitely many translates Zi of k-rational abelian
subvarieties of A, such that Zi ⊂ Z, and such that each k-rational point of Z lies on one
of the Zi .
Proposition 3.2. Let X be a smooth projective variety over a number field k which is
geometrically irreducible and does not admit a dominant map to its Albanese variety. Then
X(k) is not Zariski dense in X.
Proof of Proposition 3.2. If X(k) is empty, there is nothing to prove. Otherwise, use a
k-rational point of X to define the Albanese map over k:
j : X → Alb(X).
Then Z = j(X) is a closed, irreducible subvariety of Alb(X). Applying Theorem 3.1 with
A = Alb(X), we get a finite number, say m ≥ 1, of k-rational translates Zi of abelian
subvarieties of Alb(X) such that
Z(k) ⊂
Zi (k) and Zi ⊂ Z.
Since the Albanese map is defined over k, all the k-rational points of X are contained in
those j −1 (Zi ). Finally each j −1 (Zi ) is a proper closed sub-scheme of X (possibly singular
and reducible), since otherwise, the irreducibility of X would imply that Z = Zi = Alb(X),
contradicting the assumption that X does not admit a dominant map to its Albanese
Remark 3.3. When Lang originally made his conjecture on Mordellicity, his definition of
a variety X over k ⊂ C being hyperbolic required the Kobayashi semi-distance on X(C)
to be in fact a metric. Later it was established by R. Brody [B] that in the compact case
this was equivalent to requiring that there is no non-constant holomorphic map from C to
X(C). It is expected that every smooth projective irreducible variety X of general type
over C containing no curve of genus ≤ 1 is hyperbolic, and this is known if X is a surface
not admitting a dominant map to its Albanese variety.
Proof of Theorem 0.1. By Proposition 2.4 there exists a finite index subgroup Γ′ of Γ such
that q(YΓ′ ) > 2, which can be assumed to be normal. It follows that YΓ′ cannot admit a
dominant map to its Albanese variety. Moreover YΓ′ is a geometrically irreducible projective
surface, hence by Proposition 3.2 YΓ′ (k) is not Zariski dense in YΓ′ . If YΓ′ (k) is infinite, then
YΓ′ contains an irreducible curve C defined over k and such that C(k) infinite. Since C(k)
is Zariski dense in C, the curve C is geometrically irreducible and its geometric genus is at
most one by Faltings’ celebrated proof of Mordell’s conjecture [F1]. Taking a uniformization
of C yields a non-constant holomorphic map from C to YΓ′ , which is impossible since by
Lemma 1.2(i), YΓ′ (C) is a smooth compact hyperbolic manifold. Therefore YΓ′ is Mordellic.
By Lemma 1.2(ii), the natural morphism f : YΓ′ → YΓ is finite, etale and defined over
a number field k. Denote S the finite set of places of k where f ramifies. Then, for any
given number field k ′ ⊃ k,
f −1 (YΓ (k ′ )) ⊂
YΓ′ (k ′′ )
k ′′
where runs over all extensions of of degree at most the degree of f which are unramified
outside S. Since YΓ′ is Mordellic and there are only finitely many such extensions k ′′
(Hermite-Minkowski), it follows that YΓ (k ′ ) is finite, hence YΓ is Mordellic.
Proof of Theorem 0.2. The Lang locus of a quasi-projective irreducible variety Z is defined
as the Zariski closure of the union, over all number fields k, of irreducible components of
positive dimension of the Zariski closure of Z(k). It is clear that Z is Mordellic if, and
only if, its Lang locus is empty. The main theorem in [UY] states that, for Γ sufficiently
small, the Lang locus of a Baily-Borel compactification of YΓ is either empty or full, which
implies immediately that the same statement holds for YΓ itself.
For Γ neat, YΓ admits a smooth toroidal compactification XΓ defined over a number
field and by [Sh2, Theorem 8.1] one can assume by further shrinking Γ that q(XΓ ) > n. By
Proposition 3.2 the Lang locus of XΓ is not all, hence the Lang locus of YΓ is empty. Proof of Theorem 0.3. Let us first show that XΓ is of general type, hence its canonical
divisor KX is big (see [N, Definition 1.1]). Note that just like irregularity, the Kodaira dimension cannot decrease when going to a finite covering. By Holzapfel [H2,
Theorem 5.4.15] and Feustel [Feu] the surface XΓ1 (D) is of general type for all D ∈
{3, 4, 7, 8, 11, 15, 19, 20, 23, 24, 31, 39, 47, 71}. By the main theorem of Dˇzambi´c [Dˇz], the
surface XΓ(D) is of general type for all D ∈ {11, 15, 19, 20, 23, 31, 39, 47, 71} and a careful
inspection of his proof (using the prime above 3) shows that this is also true when D = 24.
The remaining varieties (D ∈ {3, 4, 7, 8}) are of general type by [H1, Proposition 4.13].
2 viewed as the unit ball
If g =
zj denotes the Bergman metric of HC
j dzi d¯
i,j=1 gi¯
{z = (z1 , z2 ) ∈ C2 , |z| < 1}, normalized by requiring that
Ric(g) =
∂ 2 log(g1¯1 g2¯2 − g2¯1 g1¯2 )
dzi d¯
zj = −g,
∂zi ∂ z¯j
then the holomorphic sectional curvature h is constant and equals −4/3 (see [GKK, §3.3],
3((1−|z|2 )δij +¯
zi zj )
where gi¯j =
(1−|z|2 )2
Since by Proposition 2.6 we have that Γ is neat and q(XΓ ) = q(YΓ ) > 2, Proposition
3.2 implies that XΓ (k) is not Zariski dense in XΓ . Then XΓ (k) is contained, up to a finite
set, in a union of geometrically irreducible curves C in XΓ which, by Faltings’ proof of
Mordell’s conjecture, can be assumed to be of geometric genus at most one Now applying
a result of Nadel [N, Theorem 2.1] with γ = 1 (so that −γ ≥ h = −4/3), we see that the
bigness of KX implies that each C is contained in the compactifying divisor, which is a
finite union of elliptic curves indexed by the cusps. It follows that YΓ (k) is finite and that
XΓ does not contain any rational curves, hence it is a minimal surface of general type. References
[BW] A. Borel and N. Wallach, Continuous cohomology, discrete subgroups, and representations of
reductive groups, vol. 67 of Mathematical Surveys and Monographs, American Mathematical Society,
Providence, RI, 2000.
[B] R. Brody, Compact manifolds and hyperbolicity, Trans. Amer. Math. Soc., 235 (1978), pp. 213–219.
[CV] E. Calabi and E. Vesentini, On compact, locally symmetric K¨
ahler manifolds, Ann. of Math., 71
(1960), pp. 472–507.
[C] C. Chevalley, Deux th´eor`emes d’arithm´etique, J. Math. Soc. Japan, 3 (1951), pp. 36–44.
[Cs] W. Casselman, Introduction to the theory of admissible representations of p-adic reductive groups,
preprint (1995).
[DR] M. Dimitrov and D. Ramakrishnan, Residual quotients of the Albanese of Picard modular surfaces,
and rational points, preprint (2014).
´, Invariants of some compactified Picard modular surfaces and applications, unpublished
[Dˇz] A. Dˇ
[F1] G. Faltings, Endlichkeitss¨
atze f¨
ur abelsche Variet¨
aten u
¨ber Zahlk¨
orpern, Invent. Math., 73 (1983),
pp. 349–366.
, The general case of S. Lang’s conjecture, in Barsotti Symposium in Algebraic Geometry (Abano
Terme, 1991), vol. 15 of Perspect. Math., Academic Press, San Diego, CA, 1994, pp. 175–182.
[Feu] J-M. Feustel, Zur groben Klassifikation der Picardschen Modulfl¨
achen, Math. Nachr., 118 (1984),
pp. 215–251.
[Go] B. Gordon, Canonical models of picard modular surfaces, in The zeta functions of Picard modular
surfaces, R. Langlands and D. Ramakrishnan, eds., Les Publications CRM, Montreal, 1992, pp. 1–30.
[GKK] R. Greene, K-T. Kim and S. Krantz, The geometry of complex domains, Progress in Mathematics, 291, Birkhuser Boston, Inc., Boston, MA (2011), xiv+303 pp.
[G] B. H. Gross, Arithmetic on elliptic curves with complex multiplication, vol. 776 of Lecture Notes in
Mathematics, Springer, Berlin, 1980. With an appendix by B. Mazur.
[H1] R.-P. Holzapfel A class of minimal surfaces in the unknown region of surface geography, Math.
Nachr., 98 (1980), pp. 211–232.
[H2] R.-P. Holzapfel Ball and surface arithmetics, Aspects of Mathematics, E29. Friedr. Vieweg & Sohn,
Braunschweig (1998), xiv+414 pp.
[K] D. Kazhdan, Some applications of the Weil representation, J. Analyse Mat., 32 (1977), pp. 235–248.
[L] S. Lang, Number theory III: Diophantine geometry, vol. 60 of Encyclopaedia of Mathematical Sciences,
Springer-Verlag, Berlin, 1991.
[La] R. P. Langlands, Les d´ebuts d’une formule des traces stable, vol. 13 of Publications Math´ematiques
de l’Universit´e Paris 7, Paris, 1983.
[Ma] S. Marshall, Endoscopy and cohomology growth on U (3), Compositio Math., 150 (2014), pp. 903–
[M] D. Mumford, Hirzebruch’s proportionality theorem in the noncompact case, Invent. Math., 42 (1977),
pp. 239–272.
[MR] V. K. Murty and D. Ramakrishnan, The Albanese of unitary shimura varieties, in The zeta
functions of Picard modular surfaces, R. Langlands and D. Ramakrishnan, eds., Les Publications
CRM, Montreal, 1992, pp. 445–464.
[N] A. Nadel, The nonexistence of certain level structures on abelian varieties over complex function
fields, Ann. of Math. 129 (1989), pp. 161–178.
[PR] V. Platonov and A. Rapinchuk, Algebraic groups and number theory, vol. 139 of Pure and Applied
Mathematics, Academic Press, Boston, MA, 1994.
die lokale Zetafunktion von Shimuravariet¨
aten. Monodromiefiltra[RZ] M. Rapoport and T. Zink, Uber
tion und verschwindende Zyklen in ungleicher Charakteristik, Invent. Math., 68 (1982), pp. 21–101.
[R1] J. D. Rogawski, Automorphic representations of unitary groups in three variables, vol. 123 of Annals
of Mathematics Studies, Princeton University Press, Princeton, NJ, 1990.
, The multiplicity formula for A-packets, in The zeta functions of Picard modular surfaces,
R. Langlands and D. Ramakrishnan, eds., Les Publications CRM, Montreal, 1992, pp. 395–419.
[Roh] D. Rohrlich, On the L-functions of canonical Hecke characters of imaginary quadratic fields, Duke
Math. J., 47 (1980), pp. 547--557.
[Sc] C. Schoen, An arithmetic ball quotient surface whose Albanese variety is not of CM type, preprint
[Sh1] G. Shimura, Algebraic varieties without deformation and the Chow variety, J. Math. Soc. Japan, 20
(1968), pp. 336–341.
, Automorphic forms and the periods of abelian varieties, J. Math. Soc. Japan, 31 (1979),
pp. 561–592.
[T] J. Tate, Introduction [Brief biography of Serge Lang], in Number theory, analysis and geometry,
Springer, New York, 2012, pp. xv–xx.
[U] E. Ullmo, Points rationnels des varits de Shimura, Int. Math. Res. Not., 76 (2004), pp. 4109–4125.
[UY] E. Ullmo and A. Yafaev, Points rationnels des varits de Shimura: un principe du “tout ou rien”,
Math. Ann. 348 (2010), pp. 689–705.
[V] P. Vojta, Arithmetic of subvarieties of abelian and semiabelian varieties, in Advances in number
theory (Kingston, ON, 1991), Oxford Sci. Publ., Oxford Univ. Press, New York, 1993, pp. 233–238.
[Ya] T. Yang, On CM abelian varieties over imaginary quadratic fields, Math. Ann., 329 (2004), pp. 87–
[Y] S.-T. Yau, Calabi’s conjecture and some new results in algebraic geometry, Proc. Nat. Acad. Sci.
U.S.A., 74 (1977), pp. 1798–1799.
[Ye] S-K. Yeung, Holomorphic one-forms, integral and rational points on complex hyperbolic surfaces, to
appear in J. Reine Angew. Math.
´ Lille 1, UMR 8524, UFR Mathe
´matiques, 59655 Villeneuve d’Ascq Cedex, France
E-mail address: [email protected]
Mathematics 253-37, Caltech, Pasadena, CA 91125, USA
E-mail address: [email protected]