Asymptotic integration of differential equations with oscillatory terms of large amplitudes. I (Q2567824)

The system of \(m+n\) nonlinear equations \[ \frac{dx} {dt}=f(x,y,t, \omega t), \quad \frac{dy}{dt}=\varphi(x,y,t,\omega t)+ \omega\chi(x,t,\omega t),\tag{1} \] where \(\omega\) is a large parameter, is considered. Here, \(f(x,y,t,\tau)\), \(\varphi(x,y,t,\tau)\) are defined on the set \(\Omega_1=\{(x,y,t,\tau):(x,t,\tau) \in\Omega_0,y\in G\}\), where \(\Omega_0=\{(x,t,\tau):x=(x_1,x_2,\dots,x_m)\in D\), \(t\in[0,T]\), \(T> 0\); \(\tau\in[0,\infty)\}\); \(D,G\) are bounded domains in the spaces \(\mathbb{R}^m,\mathbb{R}^n\), respectively. \(\chi(x,t,\omega t)\) is defined on \(\Omega_0\). These vector functions are \(\ell\)-periodic, \(\ell>0\), with respect to \(\tau\) and together with the derivatives \(\frac{\partial\chi}{\partial x}(x,t, \tau)\) and \(\frac{\partial x}{\partial t}(x,t,\tau)\) they are continuous and satisfy a Lipschitz condition with respect to \(x\) with a common constant for all values of their arguments; the average of \(\chi(x,t,\tau)\) with respect to \(\tau\) is zero; the corresponding averaged problem to (1) has a solution \(u^0(t),p^0(t)\); \(p^0(t)+q[u^0(t),t,\tau]\in G\), \(t\in[0,T]\), \(\tau\in[0,\ell]\). Under these conditions, the author proves, that for each \(\varepsilon>0\), there exists an \(\omega_0=\omega_0(\varepsilon)>0\) such that for each \(\omega>\omega_0\) there exists a unique solution \((x_\omega(t),y_\omega(t))\) of the Cauchy problem for (1) and the inequality \(|x_\omega(t)-u^0(t)|+|y_\omega(t)-p^0(t)-q[u^0(t), t,\omega t]|<\varepsilon\) is valid for all \(t\in[0,T]\). Moreover, the author proves that if the data of problem (1) satisfy the above-mentioned conditions and some additional smoothness conditions, then for each positive integer \(n\), there exist positive numbers \(\omega_n\) and \(c_n\) such that if \(\omega>\omega_n\), then one can effectively construct on approximation \((x^n_\omega(t),y^n_\omega(t))\) satisfying the estimate \(|x_\omega(t)-x^n_\omega(t)|+|y_\omega(t)-y^n_\omega(t)| <c_n \omega^{-(n+1)}\) for \(t\in[0,T]\); moreover, the estimate remains valid for \(n=0\) as well.