主要演算子が非線形でも使える抽象的な平均化法

We develop an abstract averaging method for equations $Mx=\varepsilon N(x,\varepsilon)$ in Banach spaces. The method applies when the equation can be separated into an averaged compatibility condition and a nonlinear complementary problem, and it allows the principal operator $M$ to be genuinely nonlinear. We prove that every nondegenerate zero of the corresponding averaged equation generates, for all sufficiently small nonzero values of $\varepsilon$, a nearby solution. Under a suitable small Lipschitz condition this solution is locally unique, while in the differentiable case it belongs to a unique local $C^1$ branch. The framework contains the abstract linear Fredholm averaging theory as a special case, but also covers quasilinear problems for which the principal operator has no invertible linearization at the reference state. We also determine the leading asymptotic profile of the solutions. Applications are given to a vector $p$-Laplacian Duffing system and to a nonsmooth problem in a Hilbert space.