APPLICATION TO CONCENTRATION IN DEGREE FOR MICROFUNCTIONS AT THE BOUNDARY
By Andrea D’Agnolo and Giuseppe Zampieri (Received July 15, 1992)
§0. Introduction.
Let X be a complex analytic manifold, let M ⊂ X be a real analytic hypersurface, and denote by σ the symplectic form on the cotangent bundle π : T ∗ X → X.
Assume that the conormal bundle T M ∗ X is symplectic with respect to the imaginary part of σ.
In this paper, we discuss the link between the (real) symplectic form σ| T∗
M
X , and the classical Levi form L M on the complex tangent plane T C M = T M ∩ iT M . In particular, we prove that if M is the boundary of a strictly pseudoconvex (or pseudoconcave) domain, and if S is a submanifold of M such that Σ = S × M T M ∗ X is regular involutive in T M ∗ X, then S is generic in X.
Define e Σ to be the union of the complexifications of the bicharacteristic leaves of Σ. For p ∈ Σ, set λ S (p) = T p T S ∗ X, λ 0 (p) = T p π −1 π(p). In §2 we show the relationship between λ S ∩ iλ S and the null-space of L M | TCS . We then show that:
dim R (T e Σ ∩ λ 0 ) = 1 + dim C (λ S ∩ iλ S ).
Denote by s ± (M ) (resp. s ± (S)) the numbers of positive and negative eigenvalues of L M (resp. L M | TCS ), and set γ(S) = dim C (λ S ∩ iλ S ∩ λ 0 ). Denote by µ W (O X ) the Sato microlocalization of the sheaf O X of holomorphic functions along a real submanifold W ⊂ X. In §3 we assume that the dimension of λ S ∩ iλ S is constant respect to p. Hence e Σ = T W ∗ X for a suitable submanifold W of X, and we show that µ W (O X ) is concentrated in degree d = 1 + dim C (λ S ∩ iλ S ) + s − (S) − γ(S).
When s − (M ) − s − (S) = dim C (λ S ∩ iλ S ) − γ(S), we also show that there is a natu- ral morphism µ W (O X ) → µ M (O X ) which induces an injective morphism between the cohomology groups in degree d. This is, in a generalized sense, a theorem of concentration in degree for microfunctions at the boundary.
§1. Symplectic forms and Levi forms.
Let X be a complex manifold of dimension n, π : T ∗ X → X the cotangent bundle, ∂ the holomorphic differential on X. Let X R denote the real underlying manifold to X and d = ∂+ ¯ ∂ its real differential. We shall always identify T ∗ (X R ) → ∼ (T ∗ X) R . (If φ is a C 1 function on X R , the identification is given by dφ(x) 7→
∂φ(x).) Let ω be the (complex) canonical 1-form on T ∗ X, σ = ∂ω the canonical
Appeared in: Japan. J. Math. (N.S.) 19 (1993), no. 2, 299–310.
1
2-form, H: T ∗ T ∗ X → T T ∗ X the corresponding Hamiltonian isomorphism. We shall consider the induced one-forms ω R = 2 Re ω, ω I = 2 Im ω and two-forms σ R = 2 Re σ, σ I = 2 Im σ on T ∗ X R .
Let M be a real analytic hypersurface of X defined by the equation φ(z) = 0 at z 0 ∈ M . Let p = (z 0 , ∂φ(z 0 )), and set:
T z0M = {v ∈ T z0X; Rehv, ∂φi = 0}
X; Rehv, ∂φi = 0}
T z C0M = T z0M ∩ iT z0M = {v ∈ T z0X; hv, ∂φi = 0}
M ∩ iT z0M = {v ∈ T z0X; hv, ∂φi = 0}
X; hv, ∂φi = 0}
λ M (p) = T p T M ∗ X λ 0 (p) = T p π −1 π(p)
ν(p) = C H(ω(p)).
The morphism:
M → T M ∗ X z 7→ (z, ∂φ(z)).
induces the morphism
ψ : T z0M → λ M (p)
v 7→ (v, ∂h∂φ, vi + ∂h ¯ ∂φ, ¯ vi)
by which T z C0M is identified to {(v, ∂h ¯ ∂φ, ¯ vi); h∂φ, vi = 0}. Let L M be the Levi form of M at p. Recall that, if (z 1 , . . . , z n ) is a local system of coordinates at z 0 , L M is the Hermitian form on T z C0M represented by the matrix (∂ i ∂ ¯ j φ(z 0 )) i,j . One immediately checks that:
M represented by the matrix (∂ i ∂ ¯ j φ(z 0 )) i,j . One immediately checks that:
ψ ∗ (σ| ψ(TC
z0
M ) ) = L M .
Let S be a real analytic submanifold of M with codim X M = 1, codim M S = r.
We shall set:
γ(S, p) = dim C (λ S (p) ∩ iλ S (p) ∩ λ 0 (p)).
One says that S × M T M ∗ X is regular when ω| S×MT
∗
M
X (= 2 i ω I | S×
MT
∗M
X ) 6= 0.
Proposition 1.1. The following assertions are equivalent:
(1.1) S × M T M ∗ X is regular at p;
(1.2) dim R (λ S (p) ∩ ν(p)) = 1.
Proof. Let S be locally defined by the equations φ i = 0 (i = 1, ..., r + 1), where φ 1 ∼ = φ is a local equation for M , and recall that p = (z 0 , ∂φ(z 0 )). We have:
ψ ∗ (ω I ) = − Re(i ∂φ).
Thus:
S × M T M ∗ X is regular ⇔ i ∂φ
∂z (z 0 ) / ∈ (T S ∗ X) z0
⇔ dim R (λ S (p) ∩ ν(p)) = 1.
Remark 1.2. Let codim X S = 2, textcodim X M = 1. Then S × M T M ∗ X is regular iff γ(S, p) = 0 (i.e. S is “generic”).
We shall assume from now on that
(1.3) T M ∗ X is I-symplectic
i.e. that σ I | λM is non-degenerate. This is equivalent to λ M (p) ∩ iλ M (p) = {0} or else to
L M is non-degenerate,
(cf [S1]). In fact via ψ one may easily identify λ M (p) ∩ iλ M (p) with the null-space of L M (cf [D’A-Z1]).
Corresponding to the non-degenerate forms σ I and L M we may thus consider two Hamiltonian isomorphisms H I and e H I :
T ∗ T M ∗ X −−−−→ T T HI M ∗ X
ψ
y ψ
x
(T C M ) ∗ H e
I
−−−−→ T C M.
Consider the morphism π ∗ M : T z ∗0M → T p ∗ T M ∗ X induced by the projection π M : T M ∗ X → M .
Proposition 1.3. For θ ∈ T z ∗
0
M , we have
(1.4) H I (π M ∗ (θ))| ψ(TCM ) = ψ(v)
(in the identification λ M (p) f →λ ∗ M (p) given by w 7→ σ(w, ·)), where v ∈ T z C
0
M is the unique solution of
(1.5) ∂h ¯ ∂φ, ¯ vi = i
2 θ| TCM (in the identification T C M f →T C M ∗ given by v 7→ hv, ·i).
Proof. The vectors v which satisfy (1.4) must verify for every u ∈ T z C0M : 1
2 [hθ, ui + h¯ θ, ¯ ui] = σ(ψ(v), ψ(u))
= −[ih∂h ¯ ∂φ, ¯ vi, ui − ih∂h ¯ ∂φ, ¯ ui, vi]
= −[ih∂h ¯ ∂φ, ¯ v, i, ui + ih∂h ¯ ∂φ, ¯ vi, ui], and hence:
Reh i
2 θ, ·i| TCM = Reh∂h ¯ ∂φ, ¯ vi, ·i| T
CM Reasoning in the same way for iu we get the conclusion.
Let us remark now that (when (1.3) holds):
dim R (λ S (p) ∩ ν(p)) = 1 ⇔ i H I (π ∗ (p)) / ∈ λ S (p),
where π ∗ : T z ∗0X → T p ∗ T ∗ X denotes the map associated to π : T ∗ X → X. We may
then correspondingly rephrase the equivalent conditions of Proposition 1.1. We also
have
Proposition 1.4. Let S ⊂ M ⊂ X with codim X M = 1 and with L M being positive or negative definite at p. Then
(1.6) S × M T M ∗ X is regular involutive ⇒ γ(S, p) = 0.
Proof. Let S be locally defined by the equations φ i = 0 (i = 1, ..., r + 1), where φ 1 ∼ = φ is a local equation for M . We have already seen that:
S × M T M ∗ X is regular ⇔ i ∂φ / ∈ R∂φ 1 + · · · + R∂φ r+1
⇔ ∂φ 2 | TCM , . . . , ∂φ r+1 | T
CM are R-independent.
(1.7)
Let v i solve (1.4) for θ i = ∂φ i . Let χ and ψ be functions on M which vanish on S and let u, v ∈ ⊕
i Rv i be such that: ψ(u) = H I (π M ∗ (dχ)), ψ(v) = H I (π ∗ M (dψ)).
Owing to Proposition 1.3, we have:
{χ, ψ} I = −i L M (¯ u ∧ v), ({·, ·} I denoting the Poisson bracket). Hence:
(1.8) {χ, ψ} I = 0 ∀χ, ψ (i.e. S × M T M ∗ X is involutive) ⇒ u 6= i v ∀u, v.
(In fact if v = i u, then σ(H u, H v) = i σ I (H u, i H u) = −2i L M (¯ u ∧ u) 6= 0 since L M definite (> 0 or < 0).) We also remark that
v 7→ h∂h ¯ ∂φ, ¯ vi, ·i| TCM is injective,
due to (1.4). By (1.7), (1.8), this implies that ∂φ i | TCM , i ≥ 2 are C-independent.
This is in turn equivalent to the fact that ∂φ i , i ≥ 1 are C-independent.
§2. Degenerate R-Lagrangian submanifolds.
Let X be a complex manifold of dimension n, and let ¯ X be the complex conjugate manifold to X. We shall identify X R to the diagonal X × X X. The manifold X × ¯ ¯ X is then a natural complexification of X R .
Let M be a C ω -hypersurface of X R defined at z 0 ∈ M by the equation φ(z) = 0, and set p = (z 0 , ∂φ(z 0 )). We shall identify C ⊗ R T z0M with {(u, ¯ v) ∈ T z0X × T z0X; h∂φ, ui + h ¯ ¯ ∂φ, ¯ vi = 0} and consider:
X × T z0X; h∂φ, ui + h ¯ ¯ ∂φ, ¯ vi = 0} and consider:
ψ C : C ⊗ R T z0M → λ M (p) + i λ M (p)
(u, ¯ v) 7→ (u; ∂h∂φ, ui + ∂h ¯ ∂φ, ¯ vi).
(2.1)
Let S ⊂ M be a submanifold with codim M S = r. Consider the null space:
NS(L M | TC
z0
S ) = {v ∈ T z C
0S; ∂h ¯ ∂φ, ¯ vi ∈ π ∗ ((T S ∗ X) z0)}.
Using T X ,→ T X × T X T ¯ X and ψ, we get
Proposition 2.1. We have an identification (2.2) λ S (p) ∩ iλ S (p) ∼ = NS(L M | TC
z0
S ) ⊕ (λ S (p) ∩ iλ S (p) ∩ λ 0 (p)).
Proof. We shall often omit the indices z 0 and p in the following. The projection R : Hπ ∗ (T S ∗ X + iT S ∗ X) −→ Hπ ∗ (T S ∗ X) is well defined modulo λ S ∩ iλ S ∩ λ 0 . It is easy to check that (cf also [D’A-Z1]):
λ S ∩ iλ S λ S ∩ iλ S ∩ λ 0
= {ψ C (v) − 2R(∂h ¯ ∂φ, ¯ vi); v ∈ T C S, ∂h ¯ ∂φ, ¯ vi ∈ Hπ ∗ (T S ∗ X + iT S ∗ X)}.
Then (2.2) follows.
We suppose all through this section that T M ∗ X is I-symplectic and that S × M T M ∗ X is a regular involutive submanifold of T M ∗ X. We also set Σ = S × M T M ∗ X, and define e Σ to be the union of the complexifications of the bicharacteristic leaves of Σ. This is a germ of R-Lagrangian submanifold of T ∗ X R .
Proposition 2.2. We have an identification
(2.3) T p Σ ∩ iT e p Σ ∼ e = ψ({v ∈ T z0S + iT z0S; ∂h ¯ ∂φ, ¯ vi ∈ Hπ ∗ (T S ∗ X z0 + iT S ∗ X z0)}).
S; ∂h ¯ ∂φ, ¯ vi ∈ Hπ ∗ (T S ∗ X z0 + iT S ∗ X z0)}).
)}).
Proof. We have by definition
(2.4) T e Σ = T Σ + iT Σ ⊥ ,
(where · ⊥ denotes the symplectic orthogonal in (T M ∗ X, σ I ).) But
(2.5) T Σ ⊥ = ψ({u ∈ T C M ∩ T S; ∂h ¯ ∂φ, ¯ ui| TCM ∈ iHπ ∗ (T S ∗ X) | T
CM }).
Then (2.3) follows immediately.
Proposition 2.3. Under the above assumptions, we have (2.6) dim C (T p Σ ∩ iT e p Σ) = r. e
Proof. We have T Σ ∩ iT Σ = 0 (due to the fact that T M ∗ X is I-symplectic). It follows:
T e Σ ∩ iT e Σ = (T Σ + iT Σ ⊥ ) ∩ (T Σ ⊥ + iT Σ)
= T Σ ⊥ + iT Σ ⊥ = C ⊗ R T Σ ⊥ .
For v ∈ T e Σ + iT e Σ (resp v ∈ T Σ + iT Σ, resp v ∈ λ S + iλ S ) let us denote by v cTΣe (resp v c
T Σ, resp v cλS) the “conjugate” with respect to T e Σ (resp T Σ, resp λ S ).
) the “conjugate” with respect to T e Σ (resp T Σ, resp λ S ).
This is well defined modulo T e Σ ∩ iT e Σ = (T e Σ + iT e Σ) ⊥ (resp T Σ ∩ iT Σ = {0}, resp λ S ∩ iλ S = (λ S + iλ S ) ⊥ ). We remark now that T e Σ + iT e Σ = T Σ + iT Σ and thus (2.7) σ(v, w cTe
Σ)| v,w∈(T e Σ+iT e Σ)∩λ
0
∼ σ(v, w cT Σ)| v,w∈(T Σ+iT Σ)∩λ0
(where “∼” means equivalent in signature and rank). On the other hand:
( ψ C (T C S) + (λ S ∩ λ 0 ) C = (λ S + iλ S ) ∩ λ 0 ψ C (T C S) ∩ (λ S ∩ λ 0 ) C = λ S ∩ iλ S ∩ λ 0 and:
(T Σ + iT Σ) ∩ λ 0 = ψ C (T C S) ⊕ (λ S ∩ λ 0 ) C
(where · C = · + i· and where we make the identification T C S ,→ {0} × T ¯ X). It follows:
(2.8) σ(v, w cT Σ)| v,w∈(T Σ+iT Σ)∩λ0 ∼ σ(v, w cλS)| v,w∈(λS+iλ
S)∩λ
0.
∼ σ(v, w cλS)| v,w∈(λS+iλ
S)∩λ
0.
+iλ
S)∩λ
0.
Let τ denote the inertia index for a triple of Lagrangian planes in the sense of [K-S1] and [K-S2]. We recall that 1 2 τ (T e Σ, iT e Σ, λ 0 ) and n − dim R (T e Σ ∩ λ 0 ) + 2 dim C (T e Σ ∩ iT e Σ ∩ λ 0 ) − dim C (T e Σ ∩ iT e Σ) are respectively the signature and the rank of σ(v, w cTe
Σ)| v,w∈(T e Σ+iT e Σ)∩λ
0