How to avoid order reduction when Lawson methods integrate nonlinear initial boundary value problems (Q2132426)

It has been observed that the application of Lawson methods to solve numerically initial boundary value problems (IBVPs) in certain scenarios leads to an order reduction. The main objective of this paper is to present a novel approach that avoids such order reduction. An usual approach used to integrate numerically IBVPs is the so-called MOL approach (method of lines approach): a spatial discretization that converts the IBVP in an initial value problem (IVP) is followed by a time integration of the semi-discrete IVP. In this paper, for the quasilinear IBVP (2.1) a different approach is followed: the time integration is performed using exponential Lawson methods (2.8)--(2.11), with some terms depending on the solutions of specified linear IBVP, is followed by the space integration. We remark that the IBVP (2.1) satisfies a set of smoothness assumptions ((A1)--(A7)) and the nonlinearity is defined by a reaction term with a convenient structure ((2.5)). The main problem is how to choose the boundary conditions for the auxiliary linear IBVPs that lead to a specific convergence order. Here, the Lawson methods are completed by the auxiliary IBVP (3.4), (4.1) or the one defined in the beginning of Section 5, depending on the convergence order of the fully discrete method and, for the space discretization presented in a general form, the authors assume the assumptions (H1)-a), b) and (H2)-a)-c). The final methods are then defined by (3.6), (4.2) and (5.1). Method (3.6) is analysed in Section 3. The main results of this section are Theorem 3.1 that establishes the conditions that lead to second and third orders for the local errors of the time discretization, Theorem 3.2 that states the orders of the local error of the fully discrete method, and Theorem 3.3 where the convergence order for (3.6) is deduced. Method (4.2) is studied in Section 4. Theorems 4.1, 4.2 establish the conditions that lead to orders 3, 4 for the local error of the time semidiscretization and convergence orders of the local errors of the fully discrete scheme, respectively. Theorem 4.3 states the orders of the global error assuming a compatibility conditions between the time and space discretizations (Condition (4.4)). Particular attention is giving to the error estimates for the Dirichlet and Robin/Neumann boundary conditions. Method (5.1) is investigated in Section 5 where the authors establish results analogous to the ones stated in Section 4 for method (4.2). In Section 6, some numerical experiments are included for a quasilinear one-dimensional heat equation with time dependent boundary conditions. It should be remarked that all the results proved require important smoothness conditions for the IBVP that may reduce the range of application of the obtained results.