The continuous classical Heisenberg ferromagnet equation with in-plane asymptotic conditions. I: Direct and inverse scattering theory (Q2425207)

The authors investigate the continuous classical Heisenberg spin chain which serves as a simple and integrable model of ferromagnetism. The object of interest is the three-dimensional magnetization vector \(\mathbf{m}(x,t)\in\mathbb{S}^2\) at position \(x\in\mathbb{R}\) and time \(t\) which satisfies the Heisenberg ferromagnet equation with in-plane asymptotic boundary conditions \begin{gather*} \mathbf{m}_t = \mathbf{m} \wedge \mathbf{m}_{xx} \tag{1a} \\ \mathbf{m}(x) \to \cos(\gamma)\mathbf{e}_1 - \sin(\gamma)\mathbf{e}_2 \quad\text{as}\ x\to\pm\infty \tag{1b} \end{gather*} where \(\gamma\in[0,2\pi)\) is a given angle. It is known that (1a) is integrable and admits a Lax pair representation [\textit{L. A. Takhtajan}, Phys. Lett. 64A, 235--237 (1977)] \[ \begin{cases} V_x & = \mathbf{A} V = [i\lambda (\mathbf{m}\cdot\boldsymbol{\sigma})]V\\ V_t & = \mathbf{B} V = [-2i\lambda^2(\mathbf{m}\cdot\boldsymbol{\sigma}) - i\lambda(\mathbf{m}\wedge\mathbf{m}_x\cdot\boldsymbol{\sigma})]V \end{cases} \tag{2} \] where \(\boldsymbol{\sigma}=(\sigma_1,\sigma_2,\sigma_3)\) is the vector of Pauli matrices, \((\mathbf{A},\mathbf{B})\) is the Lax pair, and \(\lambda\) the spectral parameter. Thus, the inverse scattering transform can be applied to the initial-value problem (1a) when \(\mathbf{m}(x,0)\) is known. The goal of this and the subsequent paper [\textit{F. Demontis} et al., Ric. Mat. 68, No. 1, 163--178 (2019; Zbl 1423.35324)] is to develop the direct and inverse scattering theory, as well as to generate closed-form soliton solutions for (1a) with the asymptotic boundary condition (1b). In particular, the above tools will be used to obtain explicit expressions for reflectionless solutions of (1). Their approach to develop the scattering theory for (1) is to recast problem (1a) with the boundary condition (1b) into the problem with perpendicular (easy-axis) asymptotic boundary conditions \[ \mathbf{m}(x) \to \mathbf{e}_3 \quad\text{as}\ x\to\pm\infty\,. \tag{3} \] In fact, the inverse scattering theory for equation (1a) with these boundary conditions is well studied in the literature, see, e.g., [\textit{L. Landau} and \textit{E. Lifshitz}, Phys. Z. Sowjetunion 8, 153--169 (1935; Zbl 0012.28501)] and [\textit{F. Demontis}, \textit{S. Lombardo}, \textit{M. Sommacal}, \textit{C. van der Mee} and \textit{F. Vargiu}, Commun. Nonlinear Sci. Numer. Simul. 64, 35--65 (2018)]. The development of direct scattering theory begins with the study of analyticity properties (Proposition 1) and the asymptotic behavior for large spectral parameters \(\lambda\) (p. 151f and Section 2.5) of the Jost matrix solutions \(\Psi=(\psi,\overline{\psi})\) and \(\Phi=(\overline{\phi},\phi)\) of the first equation in system (2) and the scattering data. In particular, the associated Faddeev functions \(\mathrm{e}^{-i\lambda x}\psi(x,\lambda)\) and \(\mathrm{e}^{i\lambda x}\phi(x,\lambda)\) are analytic in \(\lambda\in\mathbb{C}^+\), continuous in \(\lambda\in\overline{\mathbb{C}^+}\), and have a finite limit as \(\lambda\to\infty\). The actual construction of the scattering matrix \(\mathbf{S}(\lambda)\) (which contains part of the scattering data) associated to the first equation in system (2) relies on the transition matrix \(\mathbf{T}(\lambda)\) which relates the two Jost solutions with each other. From the asymptotic behavior of the Jost solutions for large \(\lambda\), it is inferred that there is an \(\alpha\in\mathbb{R}\) such that \(\mathbf{S}(\lambda)\to\mathrm{e}^{-i\alpha} I_2\) as \(\lambda\to\pm\infty\). Moreover, the reflection coefficients \(R_{\ell,r}(\lambda)\) can be expressed as Fourier transforms of \(L^1(\mathbb{R})\)-functions \(\ell\) and \(\rho\). Subsequently, the authors study the time evolution of the scattering data and find that the transmission coefficient is time-independent whereas the reflection coefficients from the left and the right satisfy \(R_{\ell,r}(\lambda,t) = \mathrm{e}^{\pm 4i\lambda^2 t}R_{\ell,r}(\lambda,0)\). In particular, the time evolution is equivalent to that of the focusing cubic NLS, but see also Propositions 1 and 2 and Corollary 2 in [\textit{V. E. Zakharov} and \textit{L. A. Takhtajan}, Theor. Math. Phys. 38, 17--23 (1979)] in the case of the boundary condition (3). Eventually, the authors formulate and solve the inverse scattering problem, i.e., the reconstruction of the (unique) \(\mathbf{m}(x)\) once the scattering data (in particular, \(\rho\) and \(\ell\)) is known, see their formulae (47) and (48) as well as (39), (40), and (41). This is done by applying the so-called Marchenko method (which was already used in [\textit{T. Aktosun} et al., Inverse Probl. 23, No. 5, 2171--2195 (2007; Zbl 1126.35058)] in the context of the focusing NLS). The above analysis relies on the relaxed regularity assumptions that \((\mathbf{m}\cdot\boldsymbol{\sigma})_x\) exists almost everywhere and belongs to \(L^1(\mathbb{R})\) and that \(1+\cos(\gamma) m_1(x)-\sin(\gamma) m_2(x)>0\) holds. In the development of the direct scattering theory, it is assumed that no spectral singularities exist, i.e., the transition matrix \(\mathbf{T}(\lambda)\) has a non-vanishing diagonal for all \(\lambda\in\mathbb{R}\) which implies in particular that the transmission coefficient is finite for all \(\lambda\in\mathbb{R}\).