You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Since we have applied a conformal rescaling, the energy of vh is uniformly bounded, so (vh) converges<br />
up to a subsequence to some J-holomorphic map v in the C0 loc ([0, +∞[×[0, +∞[, T ∗M4 )<br />
topology (more precisely, we have C∞ loc convergence once the domain [0, +∞[×[0, +∞[ is transformed<br />
by a conformal mapping turning the portion near the boundary point (1, 0) into a neighborhood<br />
of (0, 0) in the upper-right quarterÀ+ =]0, +∞[×]0, +∞[). The J-holomorphic map has<br />
finite energy, so by removal singularities it has a continuous extension at ∞ (again, by Proposition<br />
6.4 together with a suitable conformal change of variables). By (203), (204), and (205) it satisfies<br />
the boundary conditions<br />
π ◦ v(0, t) = (c(0), c(0)) for t ≥ 0, (208)<br />
v(s, 0) ∈ N ∗ ∆M×M for 0 ≤ s ≤ 1, (209)<br />
v(s, 0) ∈ N ∗ ∆ Θ M<br />
for s ≥ 1. (210)<br />
Since the boundary conditions are of conormal type and the Liouville one-form η vanishes on<br />
conormals, we have<br />
|∇v| 2 ds dt = v ∗ (ω) = v ∗ (dη) = dv ∗ �<br />
(η) = v ∗ (η) = 0,<br />
�À+<br />
�À+<br />
�À+<br />
�À+<br />
so v is constant. By (208), π ◦ v = (c(0), c(0)). In particular,<br />
Since<br />
the thesis follows.<br />
∂À+<br />
lim<br />
h→+∞ π ◦ vh(s, 0) = (c(0), c(0)) uniformly in s ∈ [0, 1].<br />
π ◦ vh(s, 0) = π ◦ ũh(αhs, 0) = (π ◦ uh(αhs, 0), π ◦ uh(s, 1)) = (dh(s), dh(s)),<br />
Localization. We fix a positive number A, playing the role of the upper bound for the action.<br />
Then the union of all spaces of solution M K 0 (γ1, γ2; x), where γ1 ∈ P Λ (L1), γ2 ∈ P Λ (L2), and<br />
x ∈ P Θ (H1 ⊕ H2) satisfy the index identity (193) and the action estimates<br />
ËL1(γ1) +ËL2(γ2) ≤ A,�H1⊕H2(x) ≤ A, (211)<br />
is a finite set. Let us denote by qi, i ∈ {1, . . ., m}, the points in M such that π ◦ ui(0, 0) =<br />
π ◦ ui(0, 1) = (qi, qi) for some u in the above finite set. We choose the indexing in such a way that<br />
the points qi are pair-wise distinct, and we fix a positive number δ such that<br />
Bδ(qi) ∩ Bδ(qj) = ∅, ∀ i �= j.<br />
We may assume that the positive constant δ chosen above is so small that Bδ(qi) ⊂ M is diffeomorphic<br />
toÊn . Lemma 6.7 implies the following localization result:<br />
6.8. Lemma. There exists a positive number α(A) such that for every α ∈]0, α(A)], every γ1 ∈<br />
P Λ (L1), γ2 ∈ P Λ (L2), x ∈ P Θ (H1 ⊕ H2) satisfying the index identity (193) and the action<br />
bounds (195), each solution u ∈ M K α (γ1, γ2; x) satisfies<br />
for some i ∈ {1, . . ., m}.<br />
π ◦ u([0, α] × {0}) = π ◦ u([0, α] × {1}) ⊂ B δ/2(qi) × B δ/2(qi),<br />
The chain homotopy. Since the Grassmannian of subspaces of some given dimension inÊn is<br />
connected and since the δ-ball around each qi is diffeomorphic toÊn , there exist smooth isotopies<br />
ϕij : [0, 1] ×Ê3n → Bδ(qi) 4 × Bδ(qj) 4 ⊂ M 8 , ∀i, j ∈ {1, . . .,m},<br />
102