Contents lists available at
SciVerse ScienceDirectC. R. Acad. Sci. Paris, Ser. I
www.sciencedirect.com
Complex Analysis
Squares of positive ( p , p ) -forms
Carrés de ( p , p ) -formes positives
Zbigniew Błocki a , Szymon Pli´s b , 1
aInstytut Matematyki, Uniwersytet Jagiello´nski, Łojasiewicza 6, 30-348 Kraków, Poland bInstytut Matematyki, Politechnika Krakowska, Warszawska 24, 31-155 Kraków, Poland
a r t i c l e i n f o a b s t r a c t
Article history:
Received 30 November 2012
Accepted after revision 21 January 2013 Available online 7 February 2013 Presented by Jean-Pierre Demailly
We show that if
α
is a positive(
2,
2)
-form, then so isα
2. We also prove that this is no longer true for forms of higher degree.©
2013 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved.r é s u m é
Nous montrons que si
α
est une(
2,
2)
-forme positive alorsα
2l’est aussi. Nous prouvons également que ceci n’est plus vrai pour les formes de degré supérieur.©
2013 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved.1. Introduction
Recall that a ( p , p ) -form α in C
nis called positive (we write α 0) if for ( 1 , 0 ) -forms γ
1, . . . , γ
n−pone has:
α ∧ i γ
1∧ ¯ γ
1∧ · · · ∧ i γ
n−p∧ ¯ γ
n−p0 .
This is a natural geometric condition, positive ( p , p ) -forms are for example characterized by the following property: for every p-dimensional subspace V and a test function ϕ 0, one has:
V
ϕα 0 .
It is well known that positive forms are real (that is α ¯ = α ) and if β is a ( 1 , 1 ) -form then
α 0 , β 0 ⇒ α ∧ β 0 . (1)
It was shown by Harvey and Knapp
[5](and independently by Bedford and Taylor
[1]) that (1) does not hold for all( p , p ) and ( q , q ) -forms α and β , respectively. We refer to Demailly’s book
[2], pp. 129–132, for a nice and simple introduction topositive forms.
Dinew
[3]gave an explicit example of ( 2 , 2 ) -forms α , β in C
4such that α 0, β 0 but α ∧ β < 0. We will recall it in the next section. The aim of this note is to show the following, somewhat surprising result:
Theorem 1. Assume that α is a positive ( 2 , 2 ) -form. Then α
2is also positive.
E-mail addresses:Zbigniew.Blocki@im.uj.edu.pl,umblocki@cyf-kr.edu.pl(Z. Błocki),splis@pk.edu.pl(S. Pli´s).
1 The second named author was partially supported by the NCN grant No. 2011/01/D/ST1/04192.
1631-073X/$ – see front matter
©
2013 Académie des sciences. Published by Elsevier Masson SAS. All rights reserved.http://dx.doi.org/10.1016/j.crma.2013.01.009
It turns out that this phenomenon holds only for ( 2 , 2 ) -forms:
Theorem 2. For every p 3, there exists a ( p , p ) -form α in C
2psuch that α 0 but α
2< 0.
We do not know if similar results hold for higher powers of positive forms.
The paper is organized as follows: in Section
2we present Dinew’s criterion for positivity of ( 2 , 2 ) -forms in C
4, which reduces the problem to a certain property of 6 × 6 matrices. Further simplification reduces the problem to 4 × 4 matrices.
We then solve it in Section
3. This is the most technical part of the paper. Higher degree forms are analyzed in Section4,where a counterpart of Dinew’s criterion is showed and Theorem
2is proved.
2. Dinew’s criterion
Without loss of generality we may assume that n = 4. Let ω
1, . . . , ω
4be a basis of (C
4)
∗such that:
dV := i ω
1∧ ¯ ω
1∧ · · · ∧ i ω
4∧ ¯ ω
4= ω
1∧ · · · ∧ ω
4∧ ¯ ω
1∧ · · · ∧ ¯ ω
4> 0 .
Set
Ω
1:= ω
1∧ ω
2, Ω
2:= ω
1∧ ω
3, Ω
3:= ω
1∧ ω
4, Ω
4:= ω
2∧ ω
3, Ω
5:= − ω
2∧ ω
4, Ω
6:= ω
3∧ ω
4.
Then
Ω
j∧ Ω
k=
ω
1∧ · · · ∧ ω
4, if k = 7 − j ,
0 , otherwise .
With every ( 2 , 2 ) -form α we can associate a 6 × 6-matrix A = ( a
jk) by
α =
j,k
a
jkΩ
j∧ ¯ Ω
k.
For
β =
j,k
b
jkΩ
j∧ ¯ Ω
kwe have:
α ∧ β =
j,k
a
jkb
7−j,7−kdV . (2)
The key will be the following criterion from
[3]:Theorem 3. α 0 if ¯ z Az
T0 for all z ∈ C
6with z
1z
6+ z
2z
5+ z
3z
4= 0.
Sketch of proof. For γ
1= b
1ω
1+ · · · + b
4ω
4, γ
2= c
1ω
1+ · · · + c
4ω
4, we have i γ
1∧ ¯ γ
1∧ i γ
2∧ ¯ γ
2=
4 j,k,l,m=1b
jb ¯
kc
lc ¯
mω
j∧ ω
l∧ ¯ ω
k∧ ¯ ω
m=
j<l k<m
( b
jc
l− b
lc
j)( b
kc
m− b
mc
k) ω
j∧ ω
l∧ ¯ ω
k∧ ¯ ω
m.
It is now enough to show that the image of the mapping:
C
8( b
1, . . . , b
4, c
1, . . . , c
4)
−→ ( b
1c
2− b
2c
1, b
1c
3− b
3c
1, b
1c
4− b
4c
1, b
2c
3− b
3c
2, − b
2c
4+ b
4c
2, b
3c
4− b
4c
3) ∈ C
6is precisely { z ∈ C
6: z
1z
6+ z
2z
5+ z
3z
4= 0 } . Indeed, it is a well-known fact that the image of the Plücker embedding of the 4-dimensional Grassmannian G ( 2 , 4 ) in P (Λ
2C
4) P
5is the quadric defined by the above equation. 2
Using Theorem
3and an idea from
[3], we can show:Proposition 4. The form
α
a=
6 j=1Ω
j∧ ¯ Ω
j+ a Ω
1∧ ¯ Ω
6+ ¯ a Ω
6∧ ¯ Ω
1is positive if and only if | a | 2.
Proof. We have:
¯
z Az
T= | z |
2+ 2 Re ( a ¯ z
1z
6) 2 | z
1z
6| + 2 | z
2z
5+ z
3z
4| + 2 Re ( a ¯ z
1z
6).
If z
1z
6+ z
2z
5+ z
3z
4= 0 and | a | 2 we clearly get ¯ z Az
T0. If we take z
1, z
6with ¯ z
1z
6= −¯ a, | z
1| = | z
6| and z
2, . . . , z
5with z
2z
5+ z
3z
4= − z
1z
6then ¯ z Az
T= 2 | a |( 2 − | a |) . 2
By (2):
α
a∧ α
b= 2
3 + Re ( a b ¯ ) dV .
Therefore, α
2, α
−2are positive, but α
2∧ α
−2< 0.
In view of Theorem
3, we see that Theorem1is equivalent to the following:
Theorem 5. Let A = ( a
jk) ∈ C
6×6be hermitian and such that z Az ¯
T0 for z ∈ C
6with z
1z
6+ z
2z
5+ z
3z
4= 0. Then
6 j,k=1a
jka
7−j,7−k0 .
We will need the following technical reduction:
Lemma 6. For every ( 2 , 2 ) -form α in C
4, we can find a basis ω
1, . . . , ω
4of (C
4)
∗such that:
α ∧ ω
1∧ ω
2∧ ¯ ω
1∧ ¯ ω
j= α ∧ ω
1∧ ω
2∧ ¯ ω
2∧ ¯ ω
j= 0 (3) for j = 3 , 4.
Proof. We may assume that α = 0, then we can find ω
1, ω
2∈ (C
4)
∗such that
α ∧ ω
1∧ ω
2∧ ¯ ω
1∧ ¯ ω
2= α ∧ i ω
1∧ ¯ ω
1∧ i ω
2∧ ¯ ω
2= 0 . (4) By V
1denote the subspace spanned by ω
1, ω
2and by V
2the subspace of all ω ∈ (C
4)
∗satisfying (3) with ω
jreplaced by ω . Then dim V
1= 2, dim V
22, and by (4) we infer V
1∩ V
2= { 0 } , hence ( C
4)
∗= V
1⊕ V
2. 2
3. Proof of Theorem
5By Lemma
6we may assume that the matrix from Theorem
5satisfies a
26= a
36= a
46= a
56= 0
and
a
62= a
63= a
64= a
65= 0 .
Then
6 j,k=1a
jka
7−j,7−k=
5 j,k=2a
jka
7−j,7−k+ 2
a
11a
66+ | a
16|
2.
Therefore Theorem
5is in fact equivalent to the following result:
Theorem 7. Let A = ( a
jk) ∈ C
4×4be hermitian and such that z Az ¯
T0 for z ∈ C
4with z
1z
4+ z
2z
3= 0. Then
4 j,k=1a
jka
5−j,5−k0 . (5)
Proof. Write
A =
⎛
⎜ ⎝
λ
1a b α
¯
a λ
2β − c b ¯ ¯β λ
3− d
¯
α −¯ c −¯ d λ
4⎞
⎟ ⎠ .
It satisfies the assumption of the theorem if and only if for every z ∈ C
4of the form z = ( 1 , ζ, w , −ζ w ) one has ¯ z Az
T0.
We can then compute
¯
z Az
T= λ
1+ 2 Re ( a ζ ) + λ
2|ζ |
2+ 2 Re
b − α ζ + β ¯ζ + c |ζ |
2w +
λ
3+ 2 Re ( d ζ ) + λ
4|ζ |
2| w |
2.
Therefore A satisfies the assumption if λ
j0,
| a |
λ
1λ
2, | b |
λ
1λ
3, | c |
λ
2λ
4, | d |
λ
3λ
4, (6)
and for every ζ ∈ C
b − α ζ + β ¯ζ + c |ζ |
22λ
1+ 2 Re ( a ζ ) + λ
2|ζ |
2λ
3+ 2 Re ( d ζ ) + λ
4|ζ |
2. (7)
In our case (5) is equivalent to
4 Re ( a d ¯ + b c ¯ ) 2 (λ
1λ
4+ λ
2λ
3) + 2
| α |
2+ |β|
2.
We will in fact prove something more:
4 Re ( a d ¯ + b c ¯ ) (
λ
1λ
4+
λ
2λ
3)
2+
| α | + |β|
2. (8)
Without loss of generality, we may assume that:
Re ( a d ¯ ) > 0 , Re ( b c ¯ ) > 0 ,
for if for example Re ( a d ¯ ) 0 then by (6) 4 Re ( a d ¯ + b c ¯ ) 4 Re ( b c ¯ ) 4
λ
1λ
2λ
3λ
4(
λ
1λ
4+ λ
2λ
3)
2.
Set u := Re ( a d ¯ ) and ζ := − r d ¯ / | d | , where r > 0 will be determined later. Then we can write the right-hand side of (7) as follows:
λ
1− 2ur
| d | + λ
2r
2λ
3− 2r | d | + λ
4r
2=
λ
1+ λ
2r
2λ
3+ λ
4r
2+ 4ur
2− 2r
λ
1| d | + λ
3u
| d | + r
2λ
2| d | + λ
4u
| d |
λ
1+ λ
2r
2λ
3+ λ
4r
2+ 4ur
2− 4r
2(
λ
1λ
4+ λ
2λ
3) √
u
= (
λ
1λ
4+
λ
2λ
3− 2 √
u )
2r
2+
λ
1λ
3−
λ
2λ
4r
22.
For r = (
λλ12λλ34)
1/4from (7) we thus obtain:
b r + d ¯
| d | α − d
| d | β + cr
λ
1λ
4+
λ
2λ
3− 2 √ u .
We also have:
b r + d ¯
| d | ( α − ¯β) + cr
b r + cr
− | α | − |β| 2
Re ( b c ¯ ) − | α | − |β|
and therefore:
2
Re ( a d ¯ ) + 2
Re ( b ¯ c )
λ
1λ
4+
λ
2λ
3+ | α | + |β|.
To get (8), we can now use the following fact: if 0 a
1x, 0 a
2x and a
1+ a
2x + y then a
21+ a
22x
2+ y
2. This can be easily verified: if a
1+ a
2x then a
21+ a
22x
2and if a
1+ a
2x then
x
2+ y
2x
2+ ( a
1+ a
2− x )
2= a
21+ a
22+ 2x ( x − a
1)( x − a
2). 2
4. (
p,
p) -Forms in C C C
2pIn C
2pwe choose the positive volume form:
dV := i dz
1∧ d ¯ z
1∧ · · · ∧ i dz
2p∧ d ¯ z
2p= dz
1∧ · · · ∧ dz
2p∧ d z ¯
1∧ · · · ∧ d ¯ z
2p.
By I we will denote the set of subscripts J = ( j
1, . . . , j
p) such that 1 j
1< · · · < j
p2p. For every J ∈ I there exists unique J
∈ I such that J ∪ J
= { 1 , . . . , 2p } . We also denote dz
J= dz
j1∧ · · · ∧ dz
jpand ε
J= ± 1 is defined in such a way that:
dz
J∧ dz
J= ε
Jdz
1∧ · · · ∧ dz
2p.
Note that:
ε
Jε
J= (− 1 )
p. (9)
With every ( p , p ) -form α in C
2pwe can associate an N × N-matrix ( a
J K) , where N = I = ( 2p ) !
( p !)
2,
by
α =
J,K
a
J Ki dz
j1∧ d ¯ z
k1∧ · · · ∧ i dz
jp∧ d ¯ z
kp= i
p2J,K
a
J Kdz
J∧ d ¯ z
K(10)
(note that ( − 1 )
p(p−1)/2i
p= i
p2). Then α
2=
J,K
ε
Jε
Ka
J Ka
JKdV (11)
and for γ
1, . . . , γ
p∈ (C
2p)
∗α ∧ i γ
1∧ ¯ γ
1∧ · · · ∧ i γ
p∧ ¯ γ
p=
J,K
a
J Kγ
1∧ · · · ∧ γ
p∧ dz
J∧ ( γ
1∧ · · · ∧ γ
p∧ dz
K).
Therefore ( a
J K) has to be positive semi-definite on the image of the Plücker embedding
C
2p∗p( γ
1, . . . , γ
p) −→
γ
1∧ · · · ∧ γ
p∧ dz
Jdz
1∧ · · · ∧ dz
2pJ∈I
∈ C
N(12)
which is well known to be a variety in C
N(see, e.g.,
[4], p. 64).We are now ready to prove Theorem
2:Proof of Theorem
2. First note that it is enough to show it for p= 3. For if α is a ( 3 , 3 ) -form in C
6such that α 0 and α
2< 0 then for p > 3 we set:
β := i dz
7∧ d z ¯
7+ · · · + idz
2p∧ d z ¯
2p.
We now have α ∧ β
p−30 but ( α ∧ β
p−3)
2= α
2∧ β
2p−6< 0.
Set p = 3, so that N = 20, and order I = { J
1, . . . , J
20} lexicographically. Then the image of the Plücker embedding (12) is in particular contained in the quadric:
z
1z
20− z
10z
11+ z
5z
16− z
2z
19= 0 . (13)
For positive a , λ, μ to be determined later define:
α := i
λ( dz
J1∧ d z ¯
J1+ dz
J20∧ d ¯ z
J20) + μ
k∈{2,5,10 11,16,19}
dz
Jk∧ d ¯ z
Jk+ a ( dz
J1∧ d ¯ z
J20+ dz
J20∧ d ¯ z
J1)
.
Then, similarly as in the proof of Proposition
4,¯
z Az
T= λ
| z
1|
2+ | z
20|
2+ μ | z
2|
2+ | z
5|
2+ | z
10|
2+ | z
11|
2+ | z
16|
2+ | z
19|
2+ 2a Re (¯ z
1z
20)
2 (λ − a ) | z
1z
20| + 2 μ | − z
10z
11+ z
5z
16− z
2z
19|
= 2 (λ + μ − a ) | z
1z
20|
if z satisfies (13). Therefore α 0 if a λ + μ .
On the other hand, by (11) and (9):
α
2= 2
λ
2+ 3 μ
2− a
2dV .
We see that if we take a = λ + μ and λ > μ > 0, then α 0 but α
2< 0. 2 References
[1] E. Bedford, B.A. Taylor, Simple and positive vectors in the exterior algebra ofCn, preprint, 1974.
[2] J.-P. Demailly, Complex Analytic and Differential Geometry, monograph, 1997, available athttp://www-fourier.ujf-grenoble.fr/~demailly.
[3] S. Dinew, On positiveC(2,2)(C4)forms, preprint, 2006.
[4] J. Harris, Algebraic Geometry. A First Course, Grad. Texts in Math., vol. 133, Springer, 1995.
[5] R. Harvey, A.W. Knapp, Positive(p,p)forms, Wirtinger’s inequality, and currents, in: R.O. Kujala, A.L. Vitter III (Eds.), Value Distribution Theory, Part A, Dekker, 1974, pp. 43–62.