Averaging of time-periodic systems without a small parameter (Q819493)

This paper presents a new averaging method for determining a periodic solution of a non-Hamiltonian ODE with periodic forcing. With respect to the classical analytical methods using the averaging approach, the results do not depend on the existence of a small parameter in the equation data. In this framework, the method consists in obtaining a suitable nonlinear equivalence problem, which can be solved formally using a Lie transform to linearize it. This text does not consider the solution's stability. More precisely, from the solution of a nonlinear functional equation, called ``equivalence problem'', the determination of the underlying time dependent family of diffeomorphisms is possible. This functional equation plays an essential role, permitting a generalization of the Lie transform formalism. The difference with other approaches is that this equation involves a notion of ``conjugacy'' (defined in Section 2) between a nonautonomous field in the original problem and an autonomous one in the transformed problem. Formal coordinate transformations, associated with both first-order and higher-order averaging, are derived. Then, it is possible to correct the solution of an averaged system by recovering the oscillatory components of the original equation. The paper shows that this approach permits gaining CPU time with respect to the standard numerical methods (the averaged system is integrated with a larger step size), which is confirmed by an example (atmospheric chemistry). This paper would have been more convincing by taking into account the critical analysis of the methods for determination of periodic solutions, and their limitations, given in \textit{I. Gumowski}'s book [Oscillatory evolution processes. Quantitative analysis arising from applied science. Nonlinear Science: Theory and Applications. Manchester etc.: Manchester University Press. (1989; Zbl 0695.34030)] (cf. in particular Chapter 2). The example given in the paper does not permit to see the parameter dependence of the solution. Any of the simpler examples given in Chapter 2 of the book above gives a possibility of comparison with the classical methods, and also with the Gumowski's one, which is different from the ordinary perturbation methods, i.e., it does not necessitate the presence of a small parameter in the original ODE, as for the author's method. Another point about the example: it is not clear that the solution, having the period of the forcing term, is unique. Indeed, other solutions with the same period, or its subharmonics, or fractional harmonics, depending on the initial conditions, might coexist with it. This fact is well known in particular in the case of electronic oscillators.