Asymptotic behavior of the Maxwell-Klein-Gordon system (Q1741664)

The authors consider the Maxwell-Klein-Gordon equation for a scalar field $\phi: \mathbb{R}^{1+3} \rightarrow \mathbb{C}$ and a potential $A_\alpha: \mathbb{R}^{1+3} \rightarrow \mathbb{R}$: \[ D^\alpha D_\alpha \phi =0, \qquad \partial^\beta F_{\alpha \beta}= J_\alpha \tag{1} \] where $D_\alpha = \partial_\alpha + i A_\alpha$ is the covariant derivative, $F_{\alpha \beta} = \partial_\alpha A_\beta - \partial_\beta A_\alpha$ is the curvature, and $J_\alpha = \Im( \phi \overline{D_\alpha \phi})$ is the current. It is taken $x^0 = t$, the indices are raised and lowered with respect to the Minkowski metric $m = \text{diag }(-1, 1, 1, 1)$. The summation over Greek indices is over $\alpha = 0, \dots, 3$, and the Latin indices are summed over the spatial variables $j=1, 2, 3$; $\partial^\alpha = m^{\alpha \beta} \partial_{\beta}$ and $\partial^0 = - \partial_t$. The authors adopt the Lorenz gauge condition \[ \partial^\alpha A_\alpha = 0 \tag{2} \] and impose the data \[ \big( \phi(0), D_0 \phi(0) \big) = (\phi_0, \dot{\phi}_0), \qquad \big(A_j(0), \partial_t A_j(0) \big) = \big( a_j, \dot{a}_j\big) \text{ for } j=1, 2, 3.\tag{3} \] They also introduce a \textit{null frame}. The first two components are the null generators of forward and backward light cones: $L=\partial_t + \partial_r$, $\underline{L}=\partial_t - \partial_r$, $r=|x|$. Let $\{e^0_B\}_{B=1,2}$ be a local orthonormal frame for the unit sphere in $\mathbb{R}^3$, and for each value of the radial variable define: $S_B = \frac{1}{r}\, e^0_B$. Then $\{S_B\}_{B=1,2}$ forms an orthonormal basis on each sphere $\{r=const\}$, for each fixed time slice. The potential $A_\mu$ can be expressed in the frame $\{L, \underline{L}, S_1, S_2\}$ and by writing $L = L^\mu \partial_\mu$: \[ A_L = L^\mu A_\mu = A_0 + \omega^j A_j, \qquad A_{\underline{L}} = \underline{L}^\mu A_\mu = A_0 - \omega^j A_j. \] \par Define the weighted Sobolev spaces: \[ \|T\|_{H^{k,s_0}(\mathbb{R}^3)}^2 = \sum_{|I|\leqslant k} \int_{\mathbb{R}} (1 + r^2)^{s_0 + |I|} |\nabla_x^I\, T|^2 \ dx \] and assume that for some $1/2<s_0<3/2$ and sufficiently small $\epsilon>0$ the initial data satisfy \[ \| (a_1, a_2, a_3) \|_{H^{k+1, s_0 -1}} + \| (\dot{a}_1, \dot{a}_2, \dot{a}_3) \|_{H^{k, s_0}} + \| \phi_0 \|_{H^{k+1, s_0-1}} + \| \dot{\phi}_0 \|_{H^{k, s_0}} \leqslant \epsilon.\tag{4} \] Given $s_0$, let $s$ and $\gamma$ be any numbers obeying \[ \frac{1}{2}<s<1,\qquad 0<\gamma < \frac{3}{2}-s\qquad\text{and}\qquad {s_0^\prime} = s+\gamma<s_0.\tag{5} \] \par The main result are the following theorems. \par Theorem 1. Let $k\geqslant 7$ with $k\in\mathbb{N}$ and assume that (4) and (5) hold. Then provided that $\epsilon$ is sufficiently small, if $(A_\mu, \phi)$ is the unique global solution to (1) with constraint (2) and data (3), for any $q\in \mathbb{R}$, $\omega \in \mathbb{S}^2$ the limits \begin{align*} & \Phi_0(q, \omega)= \lim_{t \rightarrow \infty} \big( r e^{i \frac{1}{4\pi} {\mathbf{q}} \ln (1+r)}\phi\big)\big( t, (t + q) \omega\big), \\ & \mathcal{A}_{S_B}(q,\omega)=\lim_{t\to\infty} (r {A}_{S_B})\big(t,(q+t)\omega\big), \qquad \qquad \mathcal{A}_{{\underline{L}}}^0(q, \omega)= \lim_{t \to \infty} ( r A^{\mathrm{mod}}_{{\underline{L}}})\big( t, (t + q) \omega \big) \end{align*} exist, where we define \begin{align*} &A^{mod}_{{\underline{L}}}(t,r\omega) = A_{{\underline{L}}}\big( t, r\omega\big) -\frac{1}{2r} \int_{r-t}^\infty \mathcal{J}_{{\underline{L}}}(\eta, \omega) \ln\Big( \frac{\eta + t+r}{\eta + t-r}\Big) d\eta, \\ &\mathcal{J}_{{\underline{L}}}(q,\omega) = - 2\Im\Big( \Phi_0(q, \omega) \overline{\partial_q \Phi_0(q, \omega)} \Big). \end{align*} Moreover, for $r=|x|>\frac{t}{2}$ and $\omega = \frac{x}{|x|}$ we have \begin{align*} & \big| ( e^{ i {\frac{1}{4\pi} \mathbf{q}} \ln{(1+r)}} r \phi) (t,x) - \Phi_0(r-t, \omega)\Big| \lesssim \epsilon \langle{t+r}\rangle^{\frac{1}{2} - (s+\gamma)} + \epsilon\langle{t+r}\rangle^{-1} \langle{t-r}\rangle^{\frac{3}{2}-s} S^0(t,r) \mathbf{1}_{\{t>r\}}, \\ &\big| (r A_L)(t,x) - {\frac{1}{4\pi} \mathbf{q}} \big| + \big| (r A_{S_B})(t,x) - \mathcal{A}_{S_B}(r-t, \omega)\big|+ \big| ( r A^{mod}_{{\underline{L}}})(t,r\omega) - \mathcal{A}_{{\underline{L}}}^0(r-t, \omega) \big| \\ &\lesssim \epsilon \langle{t+r}\rangle^{\frac{1}{2}-(s+\gamma)} + \epsilon^2 \frac{\langle{t-r}\rangle}{\langle{t+r}\rangle} S^0\, \mathbf{1}_{\{t>r\}}. \end{align*} Here $\mathbf{1}_{\{t>r\}}$ is the characteristic function for the set where $t>r$ and \[ S^0(t,r) = \frac{t+r}{r} \ln\Big( \frac{\langle{t+r}\rangle}{\langle{t-r}\rangle}\Big). \] \par Theorem 2. Let $k\geqslant 7$ with $k\in\mathbb{N}$ and assume that (4) and (5) hold. Given $y \in \mathbb{B}^3=\{y\in \mathbb{R}^3; \,|y|<1\}$, we have the limit \[ \lim_{t \to \infty}tA_\mu(t, ty ) = \mathcal{K}_\mu(y),\quad\text{where}\quad \mathcal{K}_\mu(y)=\frac{1}{4\pi}\int_{-\infty}^\infty\int_{\mathbb{S}^2}\frac{\mathcal{J}_\mu(q, \omega)}{1-\langle{y, \omega}\rangle}\, dS(\omega) \, dq \,, \] where $\mathcal{J}_\mu$ is the asymptotic source term \[ \mathcal{J}_\mu(q,\omega) = L_\mu(\omega)\Im\left(\Phi_0(q, \omega)\overline{\partial_q\Phi_0(q, \omega)}\right). \] Here $L_\mu(\omega)=m_{\mu\nu} L^\nu(\omega)$, where $L=\partial_t+\partial_r=L^\nu(\omega)\partial_\nu$. \par Additionally, we have the following bound on the difference when $|x|<t$ \[ \left|\,tA_\mu(t, x) - \mathcal{K}_\mu(x/t)\right| \lesssim \epsilon\big|\,t-|x|\big|^{\,1-2s} S^0+ \epsilon\big|\,t-|x|\big|^{\,1/2-s-\gamma}. \]