The work of Butler & Gribben (1968) on a general formulation ofthe problem of non-linear waves in non-uniform plasmas is extended. The particular case treated in Butler & Gribben, § 6, is discussed again (the distribution functions fs and electrostatic potential depend only on one space co-ordinate, that in the direction of propagation of the wave, the wave is slowly varying only with respect to this co-ordinate and time, the magnetic field vanishes and relativistic effects are negligible). Conditions necessary to avoid secular terms in the solutions of the Vlasov and Poisson equations are rederived, but without resorting to assumed series expansions in e, the perturbation parameter describing the non-uniformity, for the dependent variables, and using a different notation which simplifies and retains the symmetry of the equations. The corresponding general boundary condition, to be satisfied by fs innergy space at the boundary of the trapped particle region, is also derived.

Particular attention is devoted to the Vlasov equation, and the derived general result is used to obtain the appropriate necessary condition from this equation, correct to O(ε). This shows that, unlike the leading-order theory, at the next stage an extra length scale appears in the equations, which would be needed if e.g. the theory were to be used to discuss shock waves. It is argued that this result, taken with the mixing process described in Butler & Gribben (which supplies a mechanism for the formation of shocks), appears to place the theory at least on as satisfactory a level as the Navier-Stokes theory for the discussion of shock wave structures.

Another aspect of the analysis is the comparison with the Whitham (1965a) averaging method for treating the propagation of non-linear waves in other fields. Similar features are pointed out, and it seems likely that the averaging method is equivalent to the leading-order theory. It is thought that the present approach might prove to be useful in other wave problems.