Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
this intersection is a finite set of points, each of which comes with an orientation sign ±1. Denoting<br />
by ne! (x, y) the algebraic sum of these signs, we conclude that the homomorphism<br />
Mk(f, g) → Mk−n(f0, g0), x ↦→ �<br />
y∈crit(f0)<br />
m(y;f0)=k−n<br />
ne! (x, y)y, ∀x ∈ critk(f),<br />
is a chain map of degree −n, and that it induces the Umkehr map e! in homology (by the identification<br />
of the homology of the Morse complex with singular homology described in section 2.1).<br />
Let us conclude this section by describing the Morse theoretical interpretation of the intersection<br />
product in homology. Let M be a finite-dimensional oriented manifold, and consider the<br />
diagonal embedding e : ∆M ֒→ M ×M, which is n-codimensional and co-oriented. The intersection<br />
product is defined by the composition<br />
and it is denoted by<br />
Hj(M) ⊗ Hk(M) ×<br />
−→ Hj+k(M × M) e!<br />
−→ Hj+k−n(∆M) ∼ = Hj+k−n(M),<br />
• : Hj(M) ⊗ Hk(M) −→ Hj+k−n(M).<br />
The above description of e! and the description of the exterior homology product × given in section<br />
2.3 immediately yield the following description of •. Let g1, g2, g3 be complete metrics on M, and<br />
let fi ∈ F(M, gi), i = 1, 2, 3, be such that −gradfi satisfies the Morse-Smale condition. The<br />
non-degeneracy conditions (14) and (15), necessary to represent e!, are now<br />
x ∈ crit(f1) ∩ crit(f2) =⇒ m(x; f1) + m(x; f2) ≥ n, (16)<br />
x ∈ crit(f1) ∩ crit(f2) ∩ crit(f3) =⇒ m(x; f1) + m(x; f2) ≥ m(x; f3) + n. (17)<br />
These conditions are implied for instance by the generic assumption that f1 and f2 do not have<br />
common critical points. We can now perturb the metrics g1, g2, and g3 on M in such a way that<br />
for every triplet xi ∈ crit(fi), i = 1, 2, 3, the intersection<br />
W u ((x1, x2); −grad g1×g2 f1 ⊕ f2) ∩ a(W s (x3; −grad g3 f3)),<br />
a : M → M × M being the map a(p) = (p, p), is transverse in M × M, hence it is an oriented<br />
submanifold of ∆M of dimension m(x1; f1) + m(x2; f2) − m(x3; f3) − n. By compactness and<br />
transversality, when m(x3; f3) = m(x1; f1) + m(x2; f2) − n, this intersection, which can also be<br />
written as<br />
{(p, p) ∈ W u (x1; −gradf1) × W u (x2; −gradf2) | p ∈ W s (x3; −gradf3)} ,<br />
is a finite set of points, each of which comes with an orientation sign ±1. Denoting by n•(x1, x2; x3)<br />
the algebraic sum of these signs, we conclude that the homomorphism<br />
Mj(f1, g1) ⊗ Mk(f2, g2) → Mj+k−n(f3, g3), x1 ⊗ x2 ↦→ �<br />
n•(x1, x2; x3)x3,<br />
x3∈crit(f3)<br />
m(x3;f3)=j+k−n<br />
where x1 ∈ critj(f1), x2 ∈ critk(f2), is a chain map of degree −n from the complex M(f1, g1) ⊗<br />
M(f2, g2) to M(f3, g3), and that it induces the intersection product • in homology (by the identification<br />
of the homology of the Morse complex with singular homology described in section 2.1).<br />
2.5 Lagrangian action functionals, and Morse theoretical interpretation<br />
of the homomorphisms c∗, ev∗, and i!<br />
We are now in a good position to describe the homomorphisms appearing in diagram (8) in a<br />
Morse theoretical way. The first thing to do is to replace the Banach manifolds Ω(M, q0), Λ(M),<br />
and Θ(M) by the Hilbert manifolds<br />
Ω 1 (M, q0) := � γ ∈ Λ 1 � 1 1,2<br />
(M) | γ(0) = q0 , Λ (M) := W (Ì, M),<br />
Θ 1 (M) := � (γ1, γ2) ∈ Λ 1 (M) × Λ 1 (M) | γ1(0) = γ2(0) � ,<br />
16