In lines 8-11 of Lu (2009) [18, p. 2977] we wrote: "For integer m 3, if M is C m -smooth and C m−1 -smooth L : R × T M → R satisfies the assumptions (L1)-(L3), then the functional L τ is C 2 -smooth, bounded below, satisfies the Palais-Smale condition, and all critical points of it have finite Morse indexes and nullities (see [1, Prop. 4.1, 4.2] and [4])". However, as proved in Abbondandolo and Schwarz (2009) [2] the claim that L τ is C 2 -smooth is true if and only if for every (t, q) the function v → L(t, q, v) is a polynomial of degree at most 2. So the arguments in Lu (2009) [18] are only valid for the physical Hamiltonian in (1.2) and corresponding Lagrangian therein. In this note we shall correct our arguments in Lu (2009) [18] with a new splitting lemma obtained in Lu (2011) [20].
Proof. (i) For the neighborhoods V (i) in Claim 2.5 let us take open neighborhoodsThen we have the commutative diagrams: