The Log-effect for \(p\)-evolution type models (Q945891)

The authors consider the Cauchy problem for the one evolution operators, that is, \[ D^2_t- \sum^{2p}_{|\alpha|=0} a_\alpha(t) D^\alpha_x u- \sum^{p-1}_{|\alpha|= 0} b_\alpha(t) D^\alpha_x D_t u= 0,\quad t> 0,\;x\in\mathbb{R}^n, \] \[ u(0, x)= u_0(x),\qquad u_t(0, x)= u_1(x). \] They suppe that the principal part satisfies \[ C_0|\xi|^{2p}\leq \sum_{|\alpha|= 2p} a_\alpha(t)\xi^\alpha\leq C^{-1}_0 |\xi|^{2p},\quad \xi\in\mathbb{R}^n \] and \[ |D^l_t a_\alpha(t)|\leq C\Biggl({1\over t}\biggl(\log{1\over t}\biggr)^\gamma\Biggr)^l,\quad l= 1,2,\;|\alpha|= 2p, \] where \(\gamma\in [0,1]\). Moreover, the remaining coefficients of the principal part satisfy \[ |D^l_t a_\alpha(t)|\leq C\Biggl({1\over t}\biggl(\log{1\over t}\biggr)^\gamma\Biggr)^l,\quad l= 0,1,2,\;|\alpha|= p+1,\dots, 2p-1, \] where \[ \begin{aligned} 0\leq \sigma_{0\alpha}\leq {2p-|\alpha|\over p},\quad \sigma_{1\alpha}&= {(5p- 2|\alpha|)(1- \sigma_{0\alpha})+ |\alpha|- p\over 2(|\alpha|- p)},\\ \sigma_{2\alpha}&= {(3p- |\alpha|)(1- \sigma_{0\alpha})+ |\alpha|- p\over |\alpha|- p}. \end{aligned} \] Moreover, \(b_\alpha\) satisfies \[ |D^l_t b_\alpha(t)|\leq C\Biggl({1\over t}\biggl(\log{1\over t}\biggr)^\gamma\Biggr)^{\theta_{l\alpha}},\quad l= 0,1,2,\;|\alpha|= 1,\dots, p, \] where \[ \begin{aligned} \theta_{0\alpha}&\leq {p-|\alpha|\over p},\;\theta_{1\alpha}= {(p- |\alpha|)(1- \theta_{0\alpha})+ |\alpha|\over |\alpha|},\\ \theta_{2\alpha}&= {(2p- |\alpha|)(1- \theta_{0\alpha})+ |\alpha|\over |\alpha|}, \end{aligned} \] and \(a_\alpha\), \(b_0\in L^1(0,T)\) for \(|\alpha|= 0,\dots, p\). Then the Cauchy problem is \(H^\infty\) well-posed with loss regularity, that is, there are \(C_1,C_2> 0\) such that \[ \|(\langle D\rangle^p u,D_tu)(t,\cdot)\|_{H^s}\leq C_1\| e^{C_2(\log\langle D\rangle)^\gamma}\,(\langle D\rangle^p u_0,u_1)\|_{H^s}. \] Furthermore, the authors showed that the above conditions are optimal by constructing the counter-examples.