Homoclinic orbits for the perturbed sine-Gordon equation (Q4487955)

The aim of the authors is to construct homoclinic solutions of the following perturbation of the sine-Gordon equation \[ u_{tt}- a^2 u_{xx}- \sin u= 2\varepsilon^\mu u_{txx}+\varepsilon f(t,u,u_t).\tag{1} \] The solution \(u\) is supposed \(2\pi\)-periodic and even in \(x\); one assume \(0<\varepsilon\leq 1\), \(1\leq \mu< 2\) and \(a\in ({1\over 2}, 1)\). \(f(t,u,u_t)\) is smooth, \(T\)-periodic in \(t\) and has the form \[ f(t,u,d)= \gamma u+ f_1(t,u,d),\quad (\partial_u f_1)(t,0,0)= (\partial_d f_1)(t,00)= 0\tag{2} \] with \(\gamma\) some constant. For \(\varepsilon= 0\) the inverse scattering method yields a well known homoclinic solution \[ b(x,t)= 4\tan^{-1}((\sigma\cos x)(a\cosh\sigma t)^{-1}),\tag{3} \] where \(\sigma= (1- a^2)^{{1\over 2}}\). One has \(b(x,t)= 0(e^{-\sigma| t|})\). From ODE-theory one finds a unique \(x\)-independent solution \(p(t,\varepsilon)\) of \[ u_{tt}- \sin u=\varepsilon f(t,u,u_t)\tag{4} \] which is \(T\)-periodic in \(t\); one has \(p(t,\varepsilon)= 0(\varepsilon)\). One now seeks a solution \(u(x,t,\varepsilon)\) of (1) of the following form \[ u(x,t,\varepsilon)= p(t,\varepsilon)+ b(x,t-\alpha)+ r(x,t\varepsilon),\tag{5} \] where \(r(x,t,\varepsilon)\) is the function one is looking for and which has to satisfy \(\lim\{r(t)\}_{H^1}= 0\) as \(| t|\to\infty\). Here \(H^1\) is the Sobolev space of functions \(\sum a_n\cos nx\), \(\sum| a_n|^2n< \infty\), while \(\alpha\in [0,T]\) is to be determined. The main result (Theorem 2.7) asserts that if a certain technical condition involving \(\alpha\) is satisfied, then for \(\varepsilon\) sufficiently small, a solution \(u(x,t,\varepsilon)\) of (1) of the form (5) exists. This solution is thus homoclinic to \(p(t,\varepsilon)\). The first step in the rather involved proof amounts to insert the right-hand side of (5) into (1) so as to obtain an equation for \(r(t,x,\varepsilon)\) which assumes the form \[ r_{tt}- a^2 r_{xx}- 2\varepsilon^\mu r_{txx}- (\cos p+\varepsilon f_u(t, p,p_t)) r- \varepsilon f_d(t,p,p_t) r_t+ mr+ m_1 r_t= \varepsilon F+ G,\tag{6} \] where \(m(x,t,\varepsilon, \alpha)\), \(m_1(x,t,\varepsilon, \alpha)\), \(F(x,t,\varepsilon,\alpha)\) and \(G(x,t,r,r_t, \varepsilon,\alpha)\) are expressed in terms of \(f\) and its constituents in (2). After a study of the linear part on the left-hand side of (6), a Lyapunov-Schmidt method is applied to (6). The orthogonality condition pertinent to the Lyapunov-Schmidt method then leads to the technical condition in Theorem 2.7 which amounts to require that a certain Melnikov function \(E(\alpha)\) has a simple zero \(\alpha\in [0,T]\).