Ann. Sc. Norm. Super. Pisa Cl. Sci.

(5)
Vol. XI (2012), 903-919

Convergence in capacity on compact Kähler manifolds

S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

Abstract. The aim of this note is to study the convergence in capacity for func-
tions in the class E(X, ω). We study the problem under several restrictions on the
Monge-Ampère measures of the functions considered, such as common domina-
tion by a fixed measure or control on the variation.

Mathematics Subject Classification (2010): 32W20 (primary); 32Q15 (sec-

ondary).

1. Introduction

In [2,3] Bedford and Taylor laid down the foundations of the theory of the complex
Monge-Ampère operator which is nowadays a central part of pluripotential theory.
In [3] the notion of relative capacity Cn was introduced (see Section 2 for the defi-
nitions of all the notions appearing in this note). Initially Bedford and Taylor used
this capacity to solve deep problems concerning small sets in pluripotential the-
ory. It was soon realized, however, that capacities are very useful technical tools
in solving Monge-Ampère equations with singular data. Especially the discovery
of Xing [21], who proved that the complex Monge-Ampère operator is continuous
with respect to convergence in capacity, attracted much interest. This is in contrast
to convergence in L p , 1 < p < ∞, since it is known [8] that the Monge-Ampère
operator is discontinuous with respect to such topology. Recently convergence in
capacity in the setting of domains in Cn (which for brewity will be referred to as the
local setting in this note) was studied by many authors. We refer to [4–6, 9, 15, 17],
which is by far an incomplete list of recent contributions, where the reader may
obtain a complete picture of the developments in the field.
Quite recently alternative pluripotential theory on compact Kähler manifolds
was developed by Guedj and Zeriahi [12,13] and Ko!lodziej [18]. It should be noted

The first named author was partially supported by Polish ministerial grant N N 201 271135.
The final version of the note was finished while the second named author was a post doctoral
fellow in CIRM, Trento. He wishes to thank this institution for its hospitality and perfect working
conditions.
Received March 24, 2010; accepted May 20, 2011.
904 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

that both theories differ significantly especially when it comes to global results -
some of the differences were discussed in [13] and [11]. In [18] Ko!lodziej intro-
duced the capacity C X,ω on a compact Kähler manifold (X, ω), which is modelled
on the relative capacity Cn of Bedford and Taylor from the local setting. Despite its
non nocal character this notion plays similar fundamental role as Cn does in solving
complex Monge-Ampère equations.
In [13] Guedj and Zeriahi (building on local desults due to Cegrell [5, 6]) in-
troduced the new Cegrell class E (X, ω) of ω-psh functions for which the complex
Monge-Ampère operator is well-defined. Roughly speaking, it is the largest class
of ω-psh functions where the Monge-Ampère operator behaves like it does in the
bounded functions setting. The authors obtained various basic properties of this
function class. Later on, building on their work Xing [22] studied the convergence
in capacity C X,ω (one should also mention that various results in the bounded func-
tions setting were also known (compare [14, 19])). As Xing noticed, however, vari-
ous technical difficulties appear in the analysis in the whole E (X, ω) and therefore
some of his results are proven only in the smaller class E 1 (X, ω). This leads to the
question of what remains true in E (X, ω).
The aim of the present note is to answer this question. The main goal will be to
study the convergence in capacity C X,ω in the class E (X, ω) – Section 3. We obtain
fairly complete description of that convergence under various additional assump-
tions such as common domination of the Monge-Ampère measures of the functions
by a fixed measure or assumptions on the measures’ variation. Analogous results
in the local case (except Theorem 3.6 and Lemma 3.7 which are completely new)
were obtained by Ko!lodziej and Cegrell in [9]. We wish to point out that while
some of the results look as a straightforward generalizations of the corresponding
theorems due to Xing [22], the applied techniques are by necessity of different
nature, which allows us to deal with the formerly intractable case of functions in
E (X, ω) \ E 1 (X, ω). These essential new technical tools, namely the partial com-
parison principle and the uniqueness of solutions to the Monge-Ampère equation in
the class E (X, ω), are borrowed from [11].
The results obtained, in the authors’ opinion, yield better understanding of
the class E (X, ω) and the non-linear nature of the Monge-Ampère operator in this
singular setting.

ACKNOWLEDGEMENTS . The authors would like to thank an anonymous referee

of the previous draft of this manuscript whose remarks improved significantly the
exposition.

2. Preliminaries

First we recall some elements of pluripotential theory that will be used in the paper.
We refer the reader to [20] for a recent general overview of pluripotential theory
both in the local and in the Kähler setting.
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 905

Given a domain " in Cn one associates to any compact K ⊂ " the Bedford-
Taylor relative capacity Cn introduced in [3]:
Definition 2.1. Let K be a compact subset of a relatively compact domain " in Cn .
Then the relative capacity Cn (K , ") is defined by
!" #
Cn (K , ") = sup (dd c u)n : u ∈ PSH("), −1 ≤ u ≤ 0 ,
K

where PSH(") denotes the space of plurisubharmonic functions in ".

Throughout the note X will denote a (fixed) compact Kähler manifold of com-
$ ndimension n equipped with a fundamental form ω = ω X normalized by
plex
X ω = 1.
We shall deal with the ω-plurisubharmonic functions on X, which are defined
as follows:
Definition 2.2. An upper semicontinuous function ϕ : X → [−∞, +∞) is called
ω-plurisubharmonic (ω-psh for short) if ϕ ∈ L 1 (X) and ωϕ := ω + dd c ϕ ≥ 0,
where the inequality is understood in the sense of currents. By PSH(X, ω) (re-
spectively PSH− (X, ω)) we denote the set of ω-psh (respectively negative ω-psh)
functions on X.
Modelling on the definition of Cn above Ko!lodziej in [18] introduced the ca-
pacity C X,ω :
Definition 2.3. Given any Borel set E ⊂ X we define its capacity C X,ω on X as
!" #
C X (E) = C X,ω (E) = sup ωϕn : ϕ ∈ PSH(X, ω), −1 ≤ ϕ ≤ 0 ,
E

where ωϕn = (ω + dd c ϕ)n and n = dim X.

We refer to [12, 18] for more information about this capacity.
It follows from the locality of the Monge-Ampère operator that it can be de-
fined for any bounded ω-psh function [3]. As noticed by Guedj and Zeriahi this
is not so for unbounded ones and therefore one has to restrict to a subclass of
PSH(X, ω). The following class of ω-psh fuctions was introduced by Guedj and
Zeriahi in [13]:
Definition 2.4.
! " " #
n
E (X, ω) = ϕ ∈ PSH(X, ω) : lim ωmax(ϕ,− j) = ωn = 1 .
j→∞ {ϕ>− j} X

Intuitively the space E (X, ω) consists of those functions which have mild singular-
ities (or do not have −∞ poles at all), so that no Monge-Ampère mass concentrates
near those poles. In particular it can be proven (see [13]) that any such function
must have zero Lelong numbers everywhere.
906 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

Let us also define

E − (X, ω) = E (X, ω) ∩ P S H − (X, ω).
We refer to [13], where the reader can find the basic properties of these function
classes.
The convergence in capacity is defined as follows:
Definition 2.5. Let u j , u ∈ PSH(X, ω). We say that {u j } converges to u in C X if
C X ({|u j − u| > δ}) → 0
as j → ∞, for every (fixed) δ > 0.
For our later reference we shall need the notion of uniform absolute continuity
with respect to capacity. It is modelled on the uniform continuity with respect to a
measure.
Definition 2.6. A family of positive measures {µα } on X is said to be uniformly
absolutely continuous with respect to C X -capacity if for every & > 0 there exists
δ > 0 such that for each Borel subset E ⊂ X satisfying C X (E) < δ the inequality
µα (E) < & holds for all α. We denote this by µα ) C X uniformly for α.
Note that by [12, Theorem 6.2] we know that the sets satisfying C X (E) = 0
are precisely the ω-pluripolar sets. In particular all measures that are continuous
with respect to capacity must vanish on pluripolar sets.
The next two propositions are well-known:
Proposition 2.7 (cf. [13, Theorem 1.5]). Let u ∈ E (X, ω), v ∈ PSH(X, ω). Then
" "
k
ωmax(u,v) ∧T = ωuk ∧ T,
{u≤v} {u≤v}

for all 1 ≤ k ≤ n and T = ωϕ1 ∧ . . . ∧ ωϕn−k with ϕ1 , . . . , ϕn−k ∈ E (X, ω).

Proposition 2.8 (cf. Proposition 3.6 in [12]). Let u ∈ PSH− (X, ω). Then for t ≥ 0
% %
% %
%sup u % + c
% %
X
C X ({u < −t}) ≤ ,
t
where the positive constant c does not depend on u.
Next we state a simple criterion for uniform continuity in terms of capacity.
Proposition 2.9. Let u j ∈ E − (X, ω). Then the following two statements are equiv-
alent:
i) inf j≥1 sup X u j > −∞ and ωun j ) C X uniformly for j ≥ 1;
$
ii) limt→+∞ lim j→∞ {u j ≤−t} ωun j = 0.
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 907

Proof. i) ⇒ ii) This is a direct application of Proposition 2.8.

ii) ⇒ i) We claim that inf j≥1 sup X u j > −∞.
Suppose on contrary that inf j≥1 sup X u j = −∞. Hence
" "
lim ωun j = lim ωun j = 1,
j→∞ {u j ≤−t} j→∞ X

but this contradicts the assumption that

"
lim lim ωun j = 0.
t→+∞ j→∞ {u ≤−t}
j

Let us fix & > 0. We choose t0 and then j0 = j0 (t0 ) > 1 such that
"
ωun j < &,
{u j ≤−t0 }

for all j ≥ j0 . For each Borel set E ⊂ X can estimate

" " "
n n
ωu j = ωu j + ωun j
E E∩{u j ≤−t0 } E∩{u j >−t0 }
" "
≤ ωun j + n
ωmax(u j ,−t0 )
{u j ≤−t0 } E∩{u j >−t0 }
"
≤ ωun j + t0n C X (E) ≤ & + t0n C X (E)
{u j ≤−t0 }

& j0
for all j ≥ j0 . On the other hand, since k=1 ωunk ) C X we can choose δ1 > 0
& j0
such that k=1 ωunk (E) < & for all Borel sets E ⊂ X with C X (E) < δ1 . Hence
ωun j (E) < 2& for all j ≥ 1 and all Borel sets E ⊂ X, such that C X (E) < δ =
min(δ1 , t&n ).
0

The next proposition is a modified version of [22, Lemma 2]:

Proposition 2.10. Let u j , v j ∈ E − (X, ω) be such that u j ≥ v j for j ≥ 1. Assume
that ωvn j ) C X uniformly for j ≥ 1. Let also inf j≥1 sup X v j > −∞. Then
ωun j ) C X uniformly for j ≥ 1.

Proof. By [13, Theorem 1.5] we have

" " " "
ωun j ≤ 2n u
ω n
u j ≤ 2n
u
ωvn j ≤ 2n ωvn j ,
j
{u j <−2t} {v j < 2 −t} 2 {v j < 2j −t} {v j <−t}

for every t > 0. By Proposition 2.9 we obtain ωun j ) C X uniformly for j ≥ 1.

908 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

We remark that the proof shows that the result still holds for any family of
functions u α and vα , α ∈ ' instead of just sequences.
Proposition 2.11. Let u j ∈ E (X, ω), u ∈ PSH(X, ω) be such that u j → u in C X .
Then the following two statements are equivalent:
i) u ∈ E (X, ω);
ii) ωun j ) C X uniformly for j ≥ 1.
Proof. i) ⇒ ii) By Proposition 2.7 we have
" " " "
ωun j = n
ωmax(u j ,−t)
≤ n
ωmax(u j ,−t)
+ n
ωmax(u j ,−t)
{u j ≤−t} {u j ≤−t} {u≤−t+1} {|u j −u|>1}
"
n
≤ ωmax(u j ,−t)
+ t n C X ({|u j − u| > 1}).
{u≤−t+1}

By coupling the quasi-continuity of u (in [3, Theorem 3.5]) and the facts that
n
ωmax(u n
) C X uniformly for j ≥ 1 and ωmax(u,−t) ) C X uniformly for t > 0
j ,−t)
(see Proposition 2.10), we get
" "
lim lim ωun j ≤ lim n
ωmax(u,−t) = 0.
t→+∞ j→∞ {u ≤−t} t→+∞ {u≤−t+1}
j

By Proposition 2.9 we obtain ωun j ) C X uniformly for j ≥ 1.

ii) ⇒ i) This is a part in [22, Theorem 3].
Finally we would like to mention that recently Benelkourchi, Guedj and Zeri-
ahi [1] have developed analogous theory for compact complex $ manifolds equipped
with big forms (i.e. smooth semi-positive forms ω satisfying X ωn > 0). We leave
to the reader as an exercise to check that our results still hold in this more general
setting.

3. Convergence in capacity

Below we prove various results regarding convergence in capacity in the class

E (X, ω). We begin with a technical theorem which will be used later on. This
result is a generalization of similar ideas in [14, Lemma 2.3] (where the case of
bounded functions is considered) and in [22, Theorem 2], where the author works
in the smaller function class E 1 (X, ω).
Theorem 3.1. Let u j , v j ∈ E (X, ω) be such that
i) ωun j ) C X uniformly for j ≥ 1;
$
ii) lim j→∞ {u j <v j −δ} ωun j = 0 for all δ > 0.

Then lim j→∞ C X ({u j < v j − δ}) = 0 for all δ > 0.

 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 909

Proof. We can assume, by adding constants to both u j and v j if necessary, that

sup X u j = 0 for all j ≥ 1. Set u jt := max(u j , −t). For each k = 0, . . . , n we will
prove inductively that for every δ > 0
'" (
lim sup ωun−k
j
∧ ωϕk : ϕ ∈ PSH(X, ω), −1 ≤ ϕ ≤ 0 = 0. (3.1)
j→∞ {u j <v j −δ}

Note that if k = 0 then (3.1) holds by assumption.

Assume that (3.1) holds for k − 1. We will prove that
'" (
lim sup ωun−k
j
∧ ωϕk : ϕ ∈ PSH(X, ω), −1 ≤ ϕ ≤ 0 = 0
j→∞ {u j <v j −3δ}

for any δ > 0. We fix t ≥ 1 and ϕ ∈ PSH(X, ω) satisfying −1 ≤ ϕ ≤ 0. For

notational ease we denote the set {u j + δt u jt < v j + δt ϕ − 2δ} by U and let also
v j + δt ϕ−2δ
η j , β j be the ω-psh functions defined respectively by η j := , β j :=
1+ δt
u j + δt u jt
. Then we have
1+ δt
" "
ωun−k
j
∧ ωϕk ≤ ωun−k
j
∧ ωϕk ,
{u j <v j −3δ} U

δϕ−δu
since t
jt
> −δ.
The measure on the right-hand side can be enlarged to

t +δ
ωη j ∧ ωun−k
j
∧ ωϕk−1 .
δ
Thus, by consecutive application of [11, Theorem 2.3] and yet another set inclusion
obtained as the one above we get
" "
t +δ n−k k−1 t +δ
ωη j ∧ ωu j ∧ ωϕ ≤ ωβ j ∧ ωun−k
j
∧ ωϕk−1
δ U δ U
" "
t n−k+1 k−1
≤ ω ∧ ωϕ + ωu jt ∧ ωun−k ∧ ωϕk−1 .
δ {u j <v j −δ} u j {u j <v j −δ}
j

Using form [13, Corollary 1.7 ] this last sum can be in turn estimated by
" "
t +δ n−k+1 k−1
ωu j ∧ ωϕ + ωu jt ∧ ωun−k
j
∧ ωϕk−1
δ {u j <v j −δ} {u j ≤−t}
" "
t +δ
≤ ωun−k+1
j
∧ ωϕ
k−1
+ ωun−k+1
j
∧ ωϕk−1 .
δ {u j <v j −δ} {u j ≤−t}
910 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

By the induction hypothesis we get

'" (
lim sup ωun−k
j
∧ ωϕk : ϕ ∈ PSH(X, ω), −1 ≤ ϕ ≤ 0
j→∞ {u j <v j −3δ}
'" (
≤ sup ωun−k+1
j
∧ ωϕk−1 : ϕ ∈ PSH(X, ω), −1 ≤ ϕ ≤ 0, j ≥ 1 ,
{u j ≤−t}

for every t ≥ 1.
u +ϕ
Note that j 2 ≥ u j − 1, thus by Proposition 2.10 we get that ωnu j +ϕ ) C X
2
uniformly for j ≥ 1. Expanding the left-hand side we obtain

ωun−k+1
j
∧ ωϕk−1 ) C X uniformly for j ≥ 1.

Now by letting t → +∞ (and using ωun−k+1

j
∧ ωϕk−1 ) C X uniformly for j ≥ 1),
from Proposition 2.8 we obtain
'" (
lim sup ωun−k
j
∧ ωϕk : ϕ ∈ PSH(X, ω), −1 ≤ ϕ ≤ 0 = 0.
j→∞ {u j <v j −3δ}

Theorem 3.1 has various consequences regarding continuity of the Monge-Ampère

operator. Below we list some of them.
As a direct corollary we obtain the following theorem:

Theorem 3.2. Let u j , v j ∈ E (X, ω) be such that

i) ωun j + ωvn j ) C X uniformly for j ≥ 1;

$ $
ii) lim j→∞ [ {u j <v j −δ} ωun j + {v j <u j −δ} ωvn j ] = 0 for all δ > 0.

Then u j − v j → 0 in C X .

Recently the second named author obtained a characterization of convergence

in capacity for bounded ω-psh functions (in [14, Theorem 2.1]). Our next result
provides such a characterization in the class E (X, ω):

Theorem 3.3. Let u j ∈ E (X, ω) and u ∈ P S H (X, ω). Then the following three
statements are equivalent:

i) u ∈ E (X, ω) and u j → u in C X ;
ii) ωun j ) C X uniformly for j ≥ 1 and u j → u in C X ;
$
iii) ωun j ) C X uniformly for j ≥ 1, lim j→∞ u j ≤ u and lim j→∞ n
{u j <u−δ} ωu j =
0 for all δ > 0.
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 911

Proof. i) ⇔ ii) This equivalence is the content of Proposition 2.11.

ii) ⇒ iii) This implication is trivial.
iii)⇒i) This is a direct application Theorem 3.1, coupled with Proposition 2.11.

Recall that the weak convergence of Monge-Ampère measures does not imply
the weak convergence of the corresponding functions (and vice versa). Thus these
convergence problems are considered with additional restrictions. Usually results
in this direction are called stability theorems in the literature, since they show that
suitable small modification of the Monge-Ampère measures do not lead to large
deviation of solutions (i.e. the solutions are stable). In [9] good convergence prop-
erties were obtained (in the setting of domains in Cn ) under the assumption that
all the Monge-Ampère measures are dominated by a fixed measure vanishing on
pluripolar sets. In the Kähler setting the corresponding problem was studied by
Xing in [22]. He worked in the subclass E 1 (X, ω) ⊂ E (X, ω), however all his ar-
guments can be applied in the whole E (X, ω) provided one has uniqueness (modulo
an additive constant) for the solutions of the Monge-Ampère equation

ωϕn = µ, ϕ ∈ E (X, ω),

$ $
where µ is any positive Borel measure satisfying X dµ = X ωn and vanishing on
pluripolar sets. This uniqueness statement was recently obtained in [11]. Thus cou-
pling Xing’s arguments with Theorem 3.2 one gets the following stability theorem:

Theorem 3.4. Let u j ∈ E (X, ω) and u ∈ PSH(X, ω). Assume that ωun j ≤ dµ for
some measure dµ ) C X . Let also sup X u j = sup X u. Then the following three
statements are equivalent:

i) u j → u in C X .
ii) u j → u in L 1 .
iii) ωun j → ωun weakly.

Proof. The implications ii) ⇔ iii) are essentially due to Xing. In particular in [22,
Corollary 1] (and Proposition 2.11) yields ii) ⇒ iii). For the other implication one
can proceed exactly as in [22, Theorem 8]: below we sketch the details from [22]
for the reader’s convenience.
Due to the normalization assumption one can extract a subsequence from u j
convergent in L 1 to some function v ∈ PSH(X, ω). Note that one also has sup X v =
sup X u. The result follows if one can show that v actually coincides with u, since
the chosen convergent subsequence is arbitrary. But the Monge-Ampère measures
of the functions in the subsequence are again dominated by µ, hence from [22,
Theorem 4] one concludes that v belongs to E (X, ω) and moreover the Monge-
Ampère measures tend weakly to ωvn . Thus one gets ωun = ωvn , so by uniqueness
(now in E (X, ω)) and normalization u = v. Thus, since any subsequence has
subsequence convergent in L 1 to u one concludes that u j → u in L 1 .
912 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

Since implication i) ⇒ ii) is trivial, we only have to show that ii) implies i).
But we know that ωun j ) C X (since all the measures are dominated by µ). By [7]
(see also in [9, Lemma 1.4]) we know that
"
∀M ∈ R lim max(u j , M) − max(u, M)dµ = 0.
j→∞ X

(Indeed, the results in [7] and [9] are proven in the local case, but this convergence
statement is of purely local nature and hence can be easily transplanted to the Kähler
setting).
Thus we obtain
" " " "
n n 1
ωu j + ωu ≤ dµ+ |max(u j ,M)−max(u,M)|dµ.
{u j <u−δ} {u<u j −δ} {u j <M}∪{u<M} δ X

If M is sufficiently negative by the assumption µ ) C X and Proposition 2.8 we

obtain that the first term on the right-hand side can be made arbitrarily small (inde-
pendently of j). Finally we obtain
)" " *
lim ωun j + ωun = 0.
j→∞ {u j <u−δ} {u<u j −δ}

Thus by Theorem 3.2 u j → u in C X .

Before we proceed further we recall the notion of variation of a measure. Given

any Borel measure µ (non-necessarily positive) one defines its variation ||µ|| on a
Borel set A by ' (
+∞
||µ||(A) := sup |µ(Rk )| ,
k=1

where Rk , k ∈ N is any disjoint at most countable partition of the set A by Borel

sets. The classical Hahn decomposition theorem states that in fact that it is enough
to consider only the special partition given by the two Hahn sets – intutively – the
pieces where µ is positive and negative.
Remark 3.5. It would be interesting to know whether the above result still holds
if instead of domination by a fixed measure one assumes merely that ωun j ) C X .
Indeed large parts of Xing’s proof can be dealt verbatim, yet some details remain
unclear to us.
Our next result concerns a new type of stability. Instead of controlling the
deviation of u j from v j we impose an assumption on the variation of their Monge-
Ampère measures. Since in concrete situations the Monge-Ampère measures of
functions in concern are given (while the information on the functions themselves
is very limited) this result seems more suitable for applications than Theorem 3.2.
Such a theorem is new even for bounded ω-psh functions.
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 913

Theorem 3.6. Let u j , v j ∈ E (X, ω) be such that

i) lim j→∞ | sup X u j − sup X v j | = 0;
ii) ωun j + ωvn j ) C X uniformly for j ≥ 1;
$
iii) lim j→∞ X ||ωun j − ωvn j || = 0.

Then u j − v j → 0 in C X .
Note that the first assumption is an obvious normalization which can always
be assumed after adding suitable constants to the functions in concern. Before we
start the proof we state an auxiliary lemma which might be of independent interest.
Lemma 3.7. Let u, v ∈ E (X, ω). Then for any k ∈ {0, · · · , n} we have the estimate
" ," - n−k
n
||ωuk ∧ ωvn−k − ωun || ≤2 ||ωun − ωvn || .
X X

Proof. Set
1
µ = [ωun + ωvn ].
2
We choose f, g ∈ L 1 (dµ), f, g ≥ 0 such that

ωun = f dµ, ωvn = gdµ.

By the inequality for mixed Monge-Ampère measures ( [10], see also in [11, Theo-
rem 2.1]) and the Hölder inequality we get
" " " "
k n−k
k n−k n
ωu ∧ ωv − ωu ≥ f n g n dµ − f dµ
E" E E " E
k
. n−k n−k
/ %
k % n−k
%
n−k %
= f n g n − f n dµ ≥ − f n %g n − f n % dµ
E E
" ," - k ," - n−k
k n−k n n
≥− f |g − f |
n n dµ ≥ − f dµ |g − f |dµ
E E E
," - n−k ," - n−k
n n
≥− |g − f | dµ ≥− ||ωun − ωvn ||
E X

for any Borel set E ⊂ X. Similarly we get

" " ," - n−k
n
ωuk ∧ ωvn−k − ωun ≥− ||ωun − ωvn || .
X\E X\E X

Hence
" " ," - n−k
n
ωuk ∧ ωvn−k − ωun ≤ ||ωun − ωvn ||
E E X
914 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

for all Borel sets E ⊂ X. A combination of these inequalities yields

%" " % ," - n−k
% % n
% ωk ∧ ωn−k − n%
ωu % ≤ n n
||ωu − ωv || ,
% u v
E E X
for all Borel E ⊂ X. This implies that
" ," - n−k
n
k n−k n n n
||ωu ∧ ωv − ωu || ≤ 2 ||ωu − ωv || .
X X

Proof of Theorem 3.6. Our strategy will be to start with the bounded functions case.
It will turn out that the general case follows in a rather simple manner from this one.

Case I: u j , v j ∈ PSH(X, ω) ∩ L∞ (X). Set

, !" #- 1
4
k n−k m n−m
a j := sup ||ωu j ∧ ωv j − ωu j ∧ ωv j || : 1 ≤ k, m ≤ n .
X
Note that our assumptions coupled with Lemma 3.7 yield lim j→∞ a j = 0. From
now on we assume that a j ’s are small.
The crucial point in the proof will be the following equality
'" (
lim sup ωun j = 1. (3.2)
j→∞ t∈R {|u j −v j −t|≤a j }

Indeed, suppose (3.2) is already proved. Then we can choose k j ∈ R such that
"
lim ωun j = 0.
j→∞ {|u j −v j −k j |>a j }

Moreover from iii) we get

"
lim ωvn j = 0.
j→∞ {|u j −v j −k j |>a j }

By Theorem 3.2 we obtain u j − v j − k j → 0 in C X . Also from i) we get

lim j→∞ k j = 0. Hence u j − v j → 0 in C X .
Thus we have to prove (3.2). Suppose on contrary that there exists &0 > 0 such
that "
ωun j ≤ 1 − 2&0
{|u j −v j −t|≤a j }
for all t ∈ R, j ≥ 1. Set
' " (
t j := sup t ∈ R : ωun j ≤ 1 − &0 .
{u j <v j +t+a j }
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 915

$
Replacing v j + t j by v j we can assume that t j = 0. Then {u j <v j +a j } ωun j ≤ 1 − &0
$
and {u j ≤v j +a j } ωun j ≥ 1 − &0 . Hence
" " "
ωun j = 1 − ωun j = 1 − ωun j
{v j <u j +a j } {u j +a j ≤v j } {u j ≤v j +a j }
"
+ ωun j ≤ 1 − &0 .
{v j −a j <u j ≤v j +a j }
0 1
Consider now the interval [−a j + a 2j , a j − a 2j ]. It contains 1/a j − 2 disjoint
0 1
closed intervals each with length 2a 2j , ( x denotes the integral part of x). Observe
that the sets {|u j − v j − p| < a 2j } are pairwise disjoint when p varies among
$
the centers of those intervals. Since {|u j −v j |≤a j } ωun j ≤ 1 we can thus choose
s j ∈ [−a j + a 2j , a j − a 2j ] satisfying
"
ωun j ≤ 2a j .
{|u j −v j −s j |<a 2j }

(For otherwise we would get 1 ≥ 2a j ([1/a j ] − 2) ≥ 2a j (1/a j − 3) > 1 for all a j

small enough).
Replacing v j + s j by v j we can assume that s j = 0. One easily obtains the
following inequalities
" " "
n n
ωu j ≤ 1−&0 , ωu j ≤ 1−&0 , ωun j ≤ 2a j . (3.3)
{u j <v j +a 2j } {v j <u j +a 2j } {|u j −v j |<a 2j }

In [13, Theorem 4.6] we can find ρ j ∈ E (X, ω), such that sup X ρ j = 0 and ωρn j =
1 n
+c j 1{u j ≥v j } ωun j (the constant c j ≥ 0 is chosen so that the measure
1−&0 1{u j <v j } ωu j
has total mass 1, while 1 A denotes the characteristic function of the set A). Set
U j = {(1 − a 3j )u j < (1 − a 3j )v j + a 3j ρ j } ⊂ {u j < v j }.
By the inequality for mixed Monge-Ampère measures [10] we get
a 3j
ωun−1
j
∧ ω(1−a 3 )v j +a 3 ρ j ≥ (1 − a 3j )ωun−1
j
∧ ωv j + 1
ωun j
j j
(1 − &0 ) n

on U j . In [11, Theorem 2.3], our assumptions and the definition of a j we obtain

" a 3j " "
3 n−1 n
(1 − a j ) ωu j ∧ ωv j + 1
ωu j ≤ ω(1−a 3 )v j +a 3 ρ j ∧ ωun−1
j
Uj (1 − &0 ) n U j U j
j j
" " "
≤ ω(1−a 3 )u j ∧ ωun−1
j
= (1 − a 3j ) ωun j + a 3j ω ∧ ωun−1 j
j
Uj Uj Uj
2" 3 "
≤ (1 − a 3j ) ωun−1
j
∧ ωv j + a 4j + a 3j ω ∧ ωun−1
j
.
Uj Uj
916 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

Hence
 
" " "
1 1
1
 ωun j − 6 n 
1−a 3j 7 ωu j ≤ 1
ωun j
(1 − &0 ) n {u j ≤v j −a 2j } ρ j ≤− aj (1 − &0 ) n Uj
(3.4)
" "
≤ aj + ω ∧ ωun−1
j
≤ aj + ω ∧ ωun−1
j
.
Uj {u j <v j }

Similarly to ρ j we define ϑ j ∈ E (X, ω), such that sup X ϑ j = 0 and ωϑn j =

1 n
1−&0 1{v j <u j } ωu j + d j 1{v j ≥u j } ωun j (d j plays the same role as c j above). Set

V j = {(1 − a 3j )v j < (1 − a 3j )u j + a 3j ϑ j } ⊂ {v j < u j }.

Similarly we get
 
" " "
1 1

1
ωun j − 6 1−a 3
n
7 ωu j  ≤ 1
ωun j
(1 − &0 ) n {v j ≤u j −a 2j } ϑ j ≤− a j (1 − &0 ) n Vj
" "
j (3.5)
≤ aj + ω ∧ ωun−1
j
≤ aj + ω ∧ ωun−1
j
.
Vj {v j <u j }

Note that coupling (3.3), (3.4) and (3.5) one obtains

 
"
1 1 − 2a j − 2 6 n 
1 1−a 3j 7 ωu j
(1 − &0 ) n ρ j ≤− aj
 
" "
1
≤ 1
 ωun j − 2 6 n 
1−a 3j 7 ωu j
(1 − &0 ) n 2
{|u j −v j |≥a j } ρ j ≤− a
j

≤ 2a j + 1.

Now, by Proposition 2.8 and assumption ii), if we let j → ∞ we would obtain

1
1
≤ 1,
(1 − &0 ) n
a contradiction.
Thus (3.2) and the whole proof in the bounded functions case is finished.
Below we consider the general case.
Case II: u j , v j ∈ E (X, ω). We can assume, adding constants if$necessary, that
sup u j = sup X v j = 0. Choose t j → +∞ such that lim j→∞ {u j ≤−t j } ωun j +
$ X n
{v j ≤−t j } ωv j = 0. Using Case I for max(u j , −t j ) and max(v j , −t j ) we obtain
max(u j , −t j ) − max(v j , −t j ) → 0 in C X . Therefore u j − v j → 0 in C X .
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 917

Our last result is a generalization from [22, Theorem 5] and from [19, Theorem
3.4]. Ko!lodziej proved the statement for bounded functions, while Xing needed the
additional assumption that ∀ j ≥ 1, v j ≥ v0 for some fixed function v0 ∈ E (X, ω).
Note that such an assumption is quite hard to be checked especially if we know
a priori only the Monge-Ampère measures of each u j . In fact we show that this
condition is superflous.
Theorem 3.8. Let u j , v j , v ∈ E (X, ω), u ∈ PSH(X, ω) and A > 1 be such that
u j → u in L 1 (X) and ωun j ≤ Aωvn j for j ≥ 1. Assume that v j → v in C X . Then
u ∈ E (X, ω) and u j → u in C X .
Proof. For each t > 0 we set
u jt := max(u j , −t), u t := max(u, −t), v jt := max(v j , −t), vt := max(v, −t),
n−1
+
T jt := ωvk jt ∧ ωvn−1−k
t
.
k=0

We have " "

ωun j ≤A ωvn j .
{u j <u−δ} {u j <u−δ}

The latter integral can be estimated by

" " "
n
ωv j ≤ ωvn j + ωvn jt
{u j <u−δ} {u j ≤−t}∪{u≤−t}∪{v j ≤−t} {u jt <u t −δ}
" "
1
≤ ωvn j + |u t − u jt |ωvn jt
{u j ≤−t}∪{u≤−t}∪{v j ≤−t} δ X
" "
1 &
≤ ωvn j + |u t + & − u jt |ωvn jt + ,
{u j ≤−t}∪{u≤−t}∪{v j ≤−t} δ X δ

where we have used from [13, Corollary 1.7] in the first inequality and Chebychev’s
inequality in the second one.
Note that the middle term can be estimated by
" "
1 n t +& &
(u t − u jt )ωv jt + ωvn jt +
δ X δ {u jt >u t +&} δ
" "
1 1
≤ (u t − u jt )(ωvn jt − ωvnt ) + (u t − u jt )ωvnt (3.6)
δ X δ X
"
t +& &
+ ωvn jt + .
δ {u jt >u t +&} δ

Integration by parts in the first term yields (recall the definition of T jt )

" "
(u t − u jt )(ωvn jt − ωvnt ) = (vt − v jt )(ωu jt − ωu t ) ∧ T jt . (3.7)
X X
918 S!L AWOMIR D INEW AND P HA. M H O ÀNG H I Ê. P

$ n
Now coupling (3.6) and (3.7) one can estimate {u j <u−δ} ωv j by
" "
1
ωvn j
+ |vt − v jt |(ωu jt + ωu t ) ∧ T jt
{u j ≤−t}∪{u≤−t}∪{v j ≤−t} δ X
" "
1 t +& 2&
+ (u t − u jt )ωvnt + ωvn jt +
δ X δ {u jt >u t +&} δ

for all t > 0 and all & > 0.

Observe that the second term goes to 0 as j → ∞, since the integrated function
converges to 0 in capacity and all the measures are uniformly absolutely continuous
with respect to C X for j ≥ 1 (and fixed t). By [7] (recall the argument used in the
proof of Theorem 3.4) we also get
"
lim (u t − u jt )ωvnt = 0.
j→∞ X

Note that Hartogs’ lemma (see [16, Theorem 3.2.13 ]) yields C X ({u jt > u t +&}) →
0 as j → ∞. Moreover, since ωvn jt ) C X uniformly for j ≥ 1 we obtain
"
lim ωvn jt = 0.
j→∞ {u jt >u t +&}

Coupling all these and letting j → ∞ we get

" "
n 2&
lim ωu j ≤ sup ωvn j + ,
j→∞ {u j <u−δ} j≥1 {u j ≤−t}∪{u≤−t}∪{v j ≤−t} δ

for all t > 0, & > 0. Letting & → 0 and t → +∞ we obtain

"
lim ωun j = 0.
j→∞ {u j <u−δ}

By Theorem 3.3 we obtain that u j → u in C X .

References
[1] S. B ENELKOURCHI , V. G UEDJ and A. Z ERIAHI, A priori estimates for weak solutions of
complex Monge-Ampère equations, Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) 7 (2008), 81–96.
[2] E. B EDFORD and B. A. TAYLOR, The Dirichlet problem for a complex Monge-Ampère
operator, Invent. Math. 37 (1976), 1–44.
[3] E. B EDFORD and B. A. TAYLOR, A new capacity for plurisubharmonic functions, Acta
Math. 149 (1982), 1–40.
[4] T. B LOOM and N. L EVENBERG, Capacity convergence results and applications to a
Bernstein-Markov inequality, Trans. Amer. Math. Soc. 351 (1999), 4753–4767.
 C ONVERGENCE IN CAPACITY ON COMPACT K ÄHLER MANIFOLDS 919

[5] U. C EGRELL, Pluricomplex energy, Acta Math. 180 (1998), 187–217.

[6] U. C EGRELL, The general definition of the complex Monge-Ampère operator, Ann. Inst.
Fourier Grenoble 54 (2004), 159–179.
[7] U. C EGRELL, Convergence in capacity, Technical report, Isaac Newton Institute for Math-
ematical Sciences, 2001 (arXiv: math/0505218).
[8] U. C EGRELL and S. KO!L ODZIEJ, The Dirichlet problem for the complex Monge-Ampère
operator: Perron classes and rotation invariant measures, Michigan Math. J. 41 (1994),
563–569.
[9] U. C EGRELL and S. KO!L ODZIEJ, Equation of complex Monge-Ampère type and stability
of solutions, Math. Ann. 334 (2006), 713–729.
[10] S. D INEW, An inequality for mixed Monge-Ampère measures, Math. Z. 262 (2009), 1–15.
[11] S. D INEW, Uniqueness in E(X, ω), J. Funct. Anal. 256 (2009), 2113–2122.
[12] V. G UEDJ and A. Z ERIAHI, Intrinsic capacities on compact Kähler manifolds, J. Geom.
Anal. 15 (2005), 607–639.
[13] V. G UEDJ and A. Z ERIAHI, The weighted Monge-Ampère energy of quasiplurisubharmonic
functions, J. Funct. Anal. 250 (2007), 442–482.
[14] P. H. H I Ê. P, On the convergence in capacity on compact Kähler manifolds and its applica-
tions, Proc. Amer. Math. Soc. 136 (2008), 2007–2018.
[15] P. H. H I Ê. P, Convergence in capacity, Ann. Polon. Math. 93 (2008), 91–99.
[16] L. H ÖRMANDER, “Notions of Convexity”, Birkhäuser, Boston, 1994.
[17] N. V. K HUE and P. H. H I Ê. P, A comparison principle for the complex Monge-Ampère oper-
ator in Cegrell’s classes and applications, Trans. Amer. Math. Soc. 361 (2009), 5539–5554.
[18] S. KO!L ODZIEJ, The Monge-Ampère equation on compact Kähler manifolds, Indiana Univ.
Math. J. 52 (2003), 667–686.
[19] S. KO!L ODZIEJ, The set of measures given by bounded solutions of the complex Monge-
Ampère equation on compact Kähler manifolds, J. London Math. Soc. 124 (2005), 225–238.
[20] S. KO!L ODZIEJ, The complex Monge-Ampère equation and pluripotential theory, Memoirs
Amer. Math. Soc. 178/840 (2005), 1–64.
[21] Y. X ING, Continuity of the complex Monge-Ampère operator, Proc. Amer. Math. Soc. 124
(1996), 457–467.
[22] Y. X ING, Continuity of the complex Monge-Ampère operator on compact Kähler manifolds,
Math. Z. 263 (2009), 331–344.

Institute of Mathematics
Jagiellonian University
! ojasiewicza 6
ul. L
30-348 Kraków, Poland
slawomir.dinew@im.uj.edu.pl

Department of Mathematics
University of Education
(Dai hoc Su Pham Ha Noi)
CauGiay, Hanoi, Vietnam
phhiep vn@yahoo.com