On the convective nature of the instability of a front undergoing a supercritical Turing bifurcation (Q1038113): Difference between revisions

The author studies some stability questions related to the parabolic system \[ \begin{aligned} & \partial_tu_1=\partial^2_xu_1+\tfrac12(u_1-c)(1-u^2)+\gamma_1u^2_2\\ & \partial_tu_2=-(1+\partial^2_x)^2u_2+\alpha u_2-u^3_2-\gamma_2u_2(1+u_1)\end{aligned}\tag{1} \] already considered in [(*) \textit{A. Ghazaryan} and \textit{B. Sandstede}, SIAM J. Appl. Dyn. Sys. 6, No.~2, 319--347 (2007; Zbl 1210.37054)]. In (1) one has \(x\in \mathbb R\), \(t\geq 0\), while \(\gamma_1\in\mathbb R\), \(\gamma_2>0\) and \(c\in(0,1)\) are fixed, \(\alpha\), assumed to be small, is considered as a bifurcation parameter. For every \(\alpha\), system (1) admits a front solution \({\mathcal U}_h\) given by \[ {\mathcal U}_h(x-ct)=(h(x-ct),a)\text{ where }h(x)=\tanh(x/2).\tag{2} \] This front solution connects the equilibrium \((-1,0)\) at \(x=-\infty\) with the equilibrium \((1,0)\) at \(x=+\infty\). The author now discusses the stability \({\mathcal U}_h\) in (2) under the proviso that the parameters \(\gamma_1,\gamma_2,c\) are subject to certain restrictions, while \(\alpha\) is small. A first condition on \(\gamma_1,\gamma_2,c\) is given by \[ \gamma_1\gamma_2>3(1+c)(5+c)(11+3c)^{-1}\text{ and }\gamma_2>0.\tag{3} \] In (*) one also assumes \(\gamma_1>0\) while below one supposes \[ \gamma_2>0\text{ and }\gamma_2-\sqrt{2}<\gamma_1<0.\tag{4} \] The Banach spaces on which the stability analysis is based, are the locally uniform Sobolev spaces \(H^k_{u\ell}\), defined and discussed [cf. \textit{A. Mielke} and \textit{G. Schneider}, Nonlinearity 8, No.~5, 743--768 (1995; Zbl 0833.35016)]. With \(\beta>0\) one associates a weight function \(\rho_\beta(x)\) which has to satisfy \[ \rho_\beta(x)=e^{\beta x}\text{ for }x\leq -1,\quad \rho_\beta(x)=1\text{ for }x\geq 1\text{ and }\rho'_\beta(x)\geq 0\text{ for }x\in \mathbb R. \tag{5} \] The result below depends on \(\beta\) but not on the particular choice of \(\rho_\beta\). Now let (3), (4) be satisfied. By theorem 4 there are \(\alpha_*,\beta_*, K\) and \(\Lambda_*\) as follows. If \(0<\alpha<\alpha^*\) and if \(V_a=(v^1_a,v^2_a)\) satisfies \[ \|V_a\|_{H^1_{u\ell}}\leq \alpha,\quad \|\rho_{\beta_*}V_0\|_{H^{-1}_{u\ell}}\leq \alpha \] then (1) has a unique global solution \(\mathcal U\) with initial data \({\mathcal U}_h+V_a\), havig the representation \[ {\mathcal U}(x,t)={\mathcal U}_h(x-ct+q(t)]+V(x,t) \] for suitable \(q(t)\), and there is \(q^*\in\mathbb R\) such that for \(g\geq 0\), \[ \|V(\cdot,t)\|_{H^1_{al}} \leq K\alpha^{\frac14},\quad |q(t)|\leq K\alpha\text{ and }\rho_{\beta_*}(\cdot-at)V(\cdot,t)\|_{H^1_{ul}}+|q(t)-q_*|\leq Ke^{-\Lambda_*t}. \] The proof of the above statement requires lengthy computations and relies strongly on arguments put forward in (*).