INSTITUTE OF MATHEMATICS POLISH ACADEMY OF SCIENCES
WARSZAWA 1995
THE SYMMETRIC PLURICOMPLEX GREEN FUNCTION
U R B A N C E G R E L L
Department of Mathematics, University of Ume˚ a S-90187 Ume˚ a, Sweden
1. Introduction. Let Ω be an open and connected set in R
n; x
0∈ Ω. Then the classical Green function G
Ω(x, x
0) is the solution to the Dirichlet problem
G
Ω(x, x
0) = 0, ∀x ∈ ∂Ω
∆
xG(x, x
0) = δ
x0.
In [7], Klimek introduced the pluricomplex Green function g
Ω, that can be defined as solution to
g
Ω(z, z
0) ∈ PSH (Ω) g
Ω(z, z
0) = 0, ∀z ∈ ∂Ω (dd
czg
Ω(z, z
0))
n= (2π)
nδ
z0where Ω is a domain (open, bounded, and connected set) in C
n; z
0∈ Ω.
An alternative definition of g
Ωfor any domain Ω in C
n, z
0∈ Ω is g
Ω(z, z
0) = sup{ϕ(z); ϕ ∈ PSH (Ω), ϕ ≤ 0, ϕ(z) − log |z − z
0|
bounded above near z = z
0}.
It is well known that the classical Green function is symmetric: G
Ω(x, x
0)=
G
Ω(x
0, x). However, the pluricomplex Green function need not be symmetric.
It was shown by Bedford and Demailly [2] that there exists a strictly pseudo- convex smooth Ω such that g
Ω(z, z
0) 6= g
Ω(z
0, z).
2. The symmetric pluricomplex Green function. In [3], we introduced the symmetric pluricomplex Green function W
Ω(z, ω),
W
Ω(z, ω) = sup{ϕ(z, ω) ∈ 2 − PSH (Ω × Ω), ϕ ≤ 0, ϕ(z, ω) ≤ log |z − ω| − log max[d(z, {Ω), d(ω, {Ω)]}.
1991 Mathematics Subject Classification: Primary 32F05.
The paper is in final form and no version of it will be published elsewhere.
[135]
Here, 2 − PSH (Ω × Ω) denotes the subharmonic functions that are also sep- arately plurisubharmonic. The purpose of this note is to consider some basic properties of W
Ω.
Definition. A domain Ω is said to be strongly hyperconvex if W
Ω(z, ω) is an exhaustion function for Ω for each fixed ω ∈ Ω.
R e m a r k. Every strictly pseudoconvex set is strongly hyperconvex.
Lemma 1. Suppose ϕ is plurisubharmonic near zero and that |ϕ(z) − log |z|| <
K near zero for some constant K. Then µ(0) ≥ (2π)
nwhere µ is the weak
∗-limit of (dd
cmax[ϕ, t])
n, t → −∞.
P r o o f. Let 1 > r > 0 so that ϕ is plurisubharmonic and so that
|ϕ(z) − log(z)| ≤ K on B(0, r).
Given 0 < < 1, then
|z| > e
−1⇒ log |z| ≥ −1
⇒ − log |z| < 1 ⇒ (ϕ − log |z| > −K) ⇒
ϕ − log |z| > −(K + 1) − log |z| ⇒ ϕ(z) > (1 − ) log |z| − (K + 1).
Thus
Ω
= {z ∈ B(0, r); ϕ(z) < (1 − ) log |z| − (K + 1)}
is a neighborhood of zero and relatively compact in B(0, r) if e
−1< r.
Let t < inf
z∈∂Ω(1 − ) log |z| − (K + 1) = δ < 0 and define ϕ
2t= max[ϕ, 2t].
Then Ω
t= {z ∈ B(0, r); ϕ
2t< max[(1 − ) log |z| − (K + 1), t]} is a neighbor- hood of zero and relatively compact in B(0, r). Thus
R
Ωt
(dd
cϕ
t) ≥ R
Ωt
(dd
cmax[(1 − ) log |z| − K + 1, t])
n= (2π)
n(1 − )
nand since Ω
t⊂ Ω
⊂ B(0, e
−1),
R
B(0,e− 1)
(dd
cϕ
2t)
n≥ (2π)
n(1 − )
n.
So if µ is the weak
∗-limit of (dd
cϕ
2t)
n, t → −∞, then µ(0) ≥ (2π)
nwhich proves the lemma.
Theorem 1. Suppose Ω is strongly hyperconvex. Then g
Ω≥ W
Ωwith equality if and only if
τ (z) = R
Ω
(dd
cξmax[W
Ω(z, ξ), −1])
n= (2π)
n, ∀z ∈ Ω.
P r o o f (cf. [3, Prop. VII:2). Note first that R
Ω
(dd
czmax[g
Ω(z, ξ), t])
n=
(2π)
n,
∀t < 0, ∀z ∈ Ω and that
τ (z) = R
Ω
(dd
cξmax[W
Ω(z, ξ), t])
nis independent of t for all negative t. Also, it follows from definitions that g
Ω≥ W
Ω. It follows from Lemma 1 that τ (z) ≥ (2π)
nwith equality if W
Ω≡ g
Ω.
On the other hand, assume τ (z) ≡ (2π)
n. Again, by Lemma 1, (dd
cW
Ω(z, ξ))
n= 0 on z 6= ξ. Let ξ ∈ Ω be given and consider for 0 < < 1, (1 − )W
Ω(z, ξ).
Then (1 − )W
Ω(z, ξ) = 0 on ∂Ω, (1 − )W
Ω(z, ξ) ≥ g
Ω(z, ξ) for z near ξ. Since (dd
cz(1 − )W
Ω(z, ξ))
n= 0 outside ξ, (1 − )W
Ω(z, ξ) ≥ g
Ω(z, ξ) on Ω. Letting
& 0, we find that W
Ω= g
Ω.
Lemma 2. Let Ω
1⊂ C
n, Ω
2⊂ C
nbe two open and connected sets. Then max(W
Ω1, W
Ω2) ≤ W
Ω1×Ω2.
P r o o f.
0 ≥ max[W
Ω1(z
1, ω
1), W
Ω2(z
2, ω
2)]
≤ max(log |z
1− ω
1| − log max[d(z
1, {Ω
1), d(ω
1, {Ω
1)],
log |z
2− ω
2| − log max[d(z
2, {Ω
2), d(ω
2, {Ω
2)])
≤ log |(z
1, z
2)−(ω
1, ω
2)|− log min[d(z
1, {Ω
1), d(ω
1, {Ω
1), d(z
2, {Ω
1), d(ω
2, {Ω
2)]
so the inequality now follows from the definition of W
Ω1×Ω2via [3, Cor. VII:1].
Example. Denote by C
Ωthe Carath´ eodory pseudodistance on Ω. We give an example of a bounded pseudoconvex set Ω, such that
log tanh C
Ω6= W
Ω6= g
Ω.
Let Ω
1= {z ∈ C;
12< |z| < 1} and let Ω
2be any strictly pseudoconvex domain where W
Ω2(z
20, ω
20) < g
Ω2(z
20, ω
20) for a point (z
20, ω
20) ∈ Ω
2× Ω
2(by [2], such a set exists). Note first that W
Ω1= g
Ω1and that
W
Ω1(z
1, ω
1) > log tanh C
Ω(z
1, w
1), ∀z
16= ω
1∈ Ω
1(cf. Klimek [7], p. 234–235]).
Then log tanh C
Ω1×Ω2((z
1, z
2), (ω
1, z
2)) < W
Ω1(z
1, ω
1) ≤ max[W
Ω1(z
1, w
1), W
Ω2(z
2, z
2)] ≤ W
Ω1×Ω2((z
1, z
2), (ω
1, z
2)) by Lemma 2. Thus log tanh C
Ω1×Ω26=
W
Ω1×Ω2; it remains to prove that W
Ω1×Ω26= g
Ω1×Ω2. Suppose W
Ω1×Ω2≡ g
Ω1×Ω2. Since Ω
1and Ω
2are pseudoconvex, it follows from Theorem 9. 6 in [6] that g
Ω1×Ω2= max[g
Ω1, g
Ω2] so W
Ω1×Ω2((z
1, z
2), (z
1, ω
2)) = g
Ω2(z
2, ω
2) is plurisub- harmonic in ω
2which is a contradiction to the assumption
W
Ω2(z
20, ω
02) < g
Ω2(z
20, ω
02)
by Proposition VII:2 in [3].
3. Some estimates. If Ω is a domain in R
n, regular for the classical Dirich- let problem, then for every function ϕ, subharmonic near Ω we have the Riesz representation formula:
ϕ(ω) = R
Ω
G(ξ, ω)∆ϕ(ξ) + R
∂Ω
ϕ(ξ)dσ
ω(ξ), ω ∈ Ω
where G is the Green function for Ω and dσ
ωis the harmonic measure relatively Ω and ω.
Stokes theorem gives a similar formula for plurisubharmonic functions (cf.
Demailly [4], [5] and Ko lodziej [10]). Suppose V, ϕ and ψ ∈ PSH (Ω) ∩ L
∞(Ω) and define
s(r) = {z ∈ Ω; ϕ(z) = r}; B(r) = {z; ϕ(z) < r}.
We assume that B(r) ⊂⊂ Ω ∀r < 0. Consider
R
S(r)
V d
cϕ ∧ (dd
cψ)
k−1= (Stokes)
= R
B(r)
dV ∧ d
cϕ ∧ (dd
cψ)
n−1+ R
V dd
cϕ ∧ (dd
cψ)
n−1= R
B(r)
d(ϕ − r) ∧ d
cV ∧ (dd
cψ)
n−1+ R
B(r)
V (dd
cϕ) ∧ (dd
cψ
n−1= (Stokes)
= − R
B(r)
(ϕ − r)dd
cV ∧ (dd
cψ)
n−1+ R
B(r)
V dd
cϕ ∧ (dd
cψ)
k−1. Hence
R
B(r)
V (dd
cϕ) ∧ (dd
cψ)
n−1= R
B(r)
(ϕ − r)dd
cV ∧ (dd
cψ)
n−1(1)
+ R
S(r)
V d
cϕ ∧ (dd
cψ)
n−1.
We now claim that d
cϕ ∧ (dd
cψ)
n−1is a positive measure on S(r).
For let 0 ≤ h ∈ C
∞be given. Let > 0 and define ϕ
= max{ϕ, r − }. Then
R
S(r)
hd
cϕ ∧ (dd
cψ)
n−1= R
S(r)
h(d
cϕ
) ∧ (dd
cψ)
n−1= R
S(r)
dϕ
∧ d
ch ∧ (dd
cψ)
n−1+ R
B(r)
h(dd
cϕ
) ∧ (dd
cψ)
n−1= R
r−≤ϕ<r
dϕ
∧ d
ch ∧ (dd
cψ)
n−1+ R
r−≤ϕ<r
hdd
cϕ
∧ (dd
cψ)
n−1= R
r−≤ϕ<r
d(ϕ
− (r − )) ∧ d
ch ∧ (dd
cψ)
n−1+ R
r−≤ϕ<r
hdd
cϕ
∧ (dd
cψ)
n−1= R
S(r)
(ϕ
− (r − ))d
ch ∧ (dd
cψ)
n−1− R
r−≤ϕ<r
(ϕ
− (r − ))dd
ch ∧ (dd
cψ)
n−1+ R
r−≤ϕ<r
hdd
cϕ
∧ (dd
cψ)
n−1. Here, the last term is nonnegative so
R
S(r)
hd
cϕ ∧ (dd
cψ)
n−1≥ −
R
B(r)
dd
ch ∧ (dd
cψ)
n−1− R
|dd
ch ∧ (dd
cψ)
n−1| → 0, → 0 which proves the claim.
Example. Let 0 ≤ V ∈ PSH (Ω), ψ ∈ PSH ∩L
∞(Ω) and ϕ
t= max[W (z, ξ), −t].
Then (1) gives
R
B(r)
V (dd
cϕ
t) ∧ (dd
cψ)
n−1= R
B(r)
(ϕ
t− r)dd
cV ∧ (dd
cψ)
n−1+ R
s(r)
V d
cϕ
t∧ (dd
cψ)
n−1. Letting r → 0 we get
R
Ω
− ϕ
tdd
cV ∧ (dd
cψ)
n−1≤ R
Ω
V dd
cϕ
t∧ (dd
cψ)
n−1≤ sup
Ω
V R
Ω
dd
cϕ
t∧ (dd
cψ)
n−1so if we choose ψ = P
nj=1
|z
i|
2then
(i) R
Ω
− W (z, ξ)∆V ≤ kV k
L∞R
Ω
∆
zW (z, ξ), if we choose ψ = V , then
(ii) R
Ω
− W (z, ξ)(dd
cV (z))
n≤ kV k
L∞R
Ω
dd
czW (z, ξ)(dd
cV )
n−1, and finally if ψ = ϕ
t,
R − ϕtdd
cV ∧ (dd
cϕ
t)
n−1≤ R
V (dd
cϕ
t)
nand so
(iii) R
− W (z, ξ)dd
cV ∧ (dd
cW )
n−1≤ R
V (dd
cξW (z, ξ))
n.
4. Integrability of plurisubharmonic functions. Suppose µ is a positive measure on Ω. How do we know there is a ϕ ∈ PSH ∩L
∞loc (Ω) with (dd
cϕ)
n= µ?
Here is a necessary condition.
Proposition 1. Let R > 1 fixed , B the unit ball. Then there exists a constant c such that
R
B¯
− ϕ(dd
cu)
n≤ c R
B
− ϕdV for all 0 ≥ ϕ ∈ PSH (RB ) and −1 ≤ u ≤ 0, u ∈ PSH (RB ).
P r o o f. See [3, Prop. VI:2].
Let now Ω be hyperconvex with exhaustion function ψ. Let µ be a positive measure and assume 0 ≥ V ∈ PSH ∩L
∞(Ω), (dd
cV )
n= µ. For m > 0, define V
m= max(V, mψ). Then, by (1),
R Vm(dd
cψ)
n = R
ψdd
cV
m∧ (dd
cψ)
n−1≤ 1 m
R Vmdd
cV
m∧ (dd
cψ)
n−1≤ . . . ≤ 1 m
n−1
R ψ(ddcV
m)
n. so
0 ≤ R
− ψ(dd
cV
m)
n≤ m
n−1R
− V
m(dd
cψ)
n≤ m
n−1(sup
z∈Ω
−V (z)) R
(dd
cψ)
n. If τ (z) ≡ (2π)
n, we take ψ(ξ) = W (z, ξ) and get
(2π)
nm
n−1V
m(z) ≤ R
W (z, ξ)(dd
cV
mξ)
n≤ 0.
If supp µ is compact, then V
m= V near the support of µ for m large enough and therefore
0 ≤ R
−W (z, ξ)dµ(ξ) ≤ m
n−1R
−V (ξ)(dd
cW (z, ξ))
n≤ m
n−1(sup
ξ∈Ω
−W )τ (z)).
We are thus led to consider the pluricomplex potential Ω3z7→ R W (z, ξ)dµ(ξ) for positive measures µ. We have just proved
Theorem 2. Suppose −1 ≤ u ≤ 0, u ∈ PSH (Ω) and that Ω is strongly hyper- convex. Then
0≤ − R
W
Ω(z, ξ)(dd
cmax[u(ξ), mW
Ω(η, ξ)])
n≤−m
n−1R
max[u(ξ), mW
Ω(η, ξ)](dd
cξW
Ω(z, ξ))
n≤ m
n−1τ (z), z ∈ Ω, η ∈ Ω.
5. A metric defined by W . It is known that g
Ωgives rise to an infinitesimal
Finsler pseudometric, cf. [1], [8] and [9]. We show here that W
Ωalso defines an
infinitesimal Finsler pseudometric.
Definition. Let w ∈ Ω, ξ ∈ C
n. We define T (ω, ξ) = lim
|l|→0 l∈C
W
Ω(w + lξ, ω) − log |l|.
Proposition 2. T (ω, ξ) is upper semicontinuous on Ω × C
n.
P r o o f. Note that (ω, ξ, l) 7→ W
Ω(ω +lξ, ω)−log |l|, l 6= 0 is upper semicontin- uous and subharmonic in l, for ω, ω + lξ ∈ Ω. Also, for ω, ξ fixed W
Ω(ω + lξ, ω) ≤ c+log |l|. Therefore W
Ω(ω +lξ, ω)−log |l| has a uniquely determined subharmonic extension over l = 0. Also
T (ω, ξ) = lim
|l|→0
W
Ω(ω + lξ, ω) − log |l| = lim
r&0
W
Ω(ω + rξ, ω) − log r.
By the mean value property for subharmonic functions, 1
2π
2π
R
0