A doubly splitting scheme for the Caginalp system with singular potentials and dynamic boundary conditions (Q1995437)

From MaRDI portal
scientific article
Language Label Description Also known as
English
A doubly splitting scheme for the Caginalp system with singular potentials and dynamic boundary conditions
scientific article

    Statements

    A doubly splitting scheme for the Caginalp system with singular potentials and dynamic boundary conditions (English)
    0 references
    0 references
    0 references
    23 February 2021
    0 references
    In recent years a great deal of attention has been devoted to the Caginalp phase-field model (CPFM) and its variants, keeping in view of their applicability in wide ranging fields. This article is devoted to the study of the Caginalp phase field system with dynamic boundary conditions and singular potentials. For this typical PFM, authors consider a nonlinear parabolic coupled system which governs the evolution of the (relative) temperature and of an order parameter, predicting long time behaviour. Now, singular potentials are also important from the physical point of view; in particular, authors have in mind the following thermodynamically relevant logarithmic potential. To justify mathematically, contrary to regular potentials, such singular potentials have the separation property which ensures that the order parameter remains strictly between \(-1\) and 1, as is expected from the physical point of view. Here authors are interested in the Caginalp system endowed with dynamic boundary conditions and with a singular potential and, in particular, with the logarithmic potential. They prove the existence and uniqueness of solutions, as well as their regularity. The main ingredient in this study consists in proving that the order parameter u is strictly separated from the singular values of the potential. In most of the literature related to phase field methods (PFM) the existence and uniqueness of solutions to various types of models differencing in assumptions tailored to achieve specific goals, has already been proved for regular potentials. Now, singular potentials are also important from the physical point of view; the authors are therefore concerned here with singular potentials. Literature survey conducted by the authors reveals that, \begin{itemize} \item The well-posedness and long-time behaviour of the Caginalp system with singular potentials and dynamic boundary conditions has been thoroughly analysed. \item From a numerical point of view, calculations for PFM with a logarithmic function were performed for regular solutions and for singular solutions, but only up to the singular time. \item To the best of the author's knowledge, up to the time of preparation of manuscript, the computation of singular solutions for Cahn-Hilliard type problems involving both a logarithmic potential and dynamic boundary conditions did not seem to have been addressed. \end{itemize} The author's line of research in this paper is therefore, to propose and to analyse a scheme which allows us to compute singular solutions to the problem even after the singularity occurs. A fundamental idea in their approach is to use the energy associated to the problem. A doubly splitting scheme for the Caginalp system, which forms part of the title of the paper, is based on two ideas: a splitting in time, which decouples the resolution of the equations in the problem at each time step and a convex splitting scheme (CSS) of the energy, forms a basis for the proof of the unconditional and unique solvability of the proposed scheme. A discrete version (computational model) of it is then obtained if the time step is vanishingly small. The main result of the paper is stated in the form of a proposition: By letting the time step tend to 0 and using a monotonicity argument, the time semi-discrete solution converges to an energy solution of the problem. A section of the paper is devoted to assumptions, notations and functional setting essential for rigorous proofs and introduction to the notion of energy solution. Under suitable assumptions, two Main Theorems, concerned with the existence and convergence of energy-dissipative solution and the scheme's is unconditional unique solvability, are proved theoretically and demonstrated numerically as the main results of the paper. Separate sections are devoted exclusively to focus on the analysis and numerical computation of 1d stationary singular solutions to CPFM and numerical computations of regular and singular solutions to the Caginalp system in two space dimension. Ample numerical experiments are performed to validate the accuracy and efficiency of the proposed scheme. The present semi-discrete method can solve one-, two- and three-dimensional problems with variable coefficients conveniently at low implementation cost. The new approach is simple and effective. The results of a numerical experiment are given, and the accuracy are discussed and compared. The simulations are carried with the FreeFem++ software. For the optimization problem, the truncated Newton algorithm from the ff-NLopt package was used. Observations and comments \begin{itemize} \item This is an important field of research with direct applications in many areas of material science. The purpose of the paper is to study the long-time behaviour of the Caginalp phase-field model with a logarithmic potential and dynamic boundary conditions for both the order parameter and the temperature. \item Number of numerical analysists have developed the method of time-splitting to divide complicated time-dependent partial differential equations into sets of simpler equations which could then be solved separately by numerical means over fractions of a time-step. In this paper splitting in time decouples the resolution of the equations in the problem at each step. \item Only development of faster computers is not enough, numerical analysis (fast and efficient algorithms) is needed. Double splitting scheme may lead a promising direction to the design of accurate and efficient numerical schemes for phase field modelling. \item Splitting methods of one form or another are frequently used in computing numerical solutions to partial differential equations. The Caginalp phase-field model being a system of time-dependent (evolutionary) partial differential equations, double splitting has an added advantage as explained by the authors. \item Three main aspects of the method are considered. The first is the accuracy and efficiency of the time-split method relative to un-split methods. The second is stability for split methods. Finally robustness of the method. The proposed scheme meets all the criteria. \item Authors have performed numerical simulations in two space dimension for the Caginalp system. The simulations were performed with the FreeFem++ software. \item Errors may trigger numerical instability which destroys convergence. Already existence and uniqueness of solution and regularity of solutions are handled in theory and using numerical experiments. Table 1 shows the L2 -- error between the solution from the doubly splitting scheme and the solution from the linearly implicit scheme at the final time T = 0:5. \item It is pointed out that for a regular solution, the linearly implicit scheme is much faster than the doubly splitting scheme. This is illustrated in Table 2. However, in the author's opinion, the computational time could be reduced by using a second order method involving the Hessian of the objective function and not only the gradient \item Visualisations of evolutionary behaviour of solutions for long-time-range to make predictions of the model are often desired. The Matlab software was used for the visualization. \item Numerical validation of the scheme is done by comparison of the solution obtained by the proposed scheme and the solution of CPFM computed with the linearly implicit scheme. \end{itemize}
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    Caginalp phase-field system
    0 references
    logarithmic potential
    0 references
    dynamic boundary conditions
    0 references
    maximal monotone operator
    0 references
    multivalued operator
    0 references
    convex splitting
    0 references
    0 references