We analyse a novel subdomain scheme for time-spectral solution of initial-value partial differential equations. In numerical modelling spectral methods are commonplace for spatially dependent systems, whereas finite difference schemes are typically applied for the temporal domain. The Generalized Weighted Residual Method (GWRM) is a fully spectral method that spectrally decomposes all specified domains, including the temporal domain, using multivariate Chebyshev polynomials. The Common Boundary-Condition method (CBC) here proposed is a spatial subdomain scheme for the GWRM. It solves the physical equations independently from the global connection of subdomains in order to reduce the total number of modes. Thus, it is a condensation procedure in the spatial domain that allows for a simultaneous global temporal solution. It is here evaluated against the finite difference methods of Crank-Nicolson and Lax-Wendroff for two example linear PDEs. The CBC-GWRM is also applied to the linearised ideal magnetohydrodynamic (MHD) equations for a screw pinch equilibrium. The growth rate of the most unstable mode was efficiently computed with an error <0.1%.

Numerical weather prediction (NWP) is a difficult task in chaotic dynamical regimes because of the strong sensitivity to initial conditions and physical parameters. As a result, high numerical accuracy is usually necessary. In this chapter, an accurate and efficient alternative to the traditional time stepping solution methods is presented; the time-spectral method. The generalized weighted residual method (GWRM) solves systems of nonlinear ODEs and PDEs using a spectral representation of time. Not being subject to CFL-like criteria, the GWRM typically employs time intervals two orders of magnitude larger than those of time-stepping methods. As an example, efficient solution of the chaotic Lorenz 1984 equations is demonstrated. The results indicate that the method has strong potential for NWP. Furthermore, employing spectral representations of physical parameters and initial values, families of solutions are obtained in a single computation. Thus, the GWRM is conveniently used for studies of system parameter dependency and initial condition error growth in NWP.

Finite difference methods are traditionally used for modelling the time domain in numerical weather prediction (NWP). Time-spectral solution is an attractive alternative for reasons of accuracy and efficiency and because time step limitations associated with causal CFL-like criteria, typical for explicit finite difference methods, are avoided. In this work, the Lorenz 1984 chaotic equations are solved using the time-spectral algorithm GWRM (Generalized Weighted Residual Method). Comparisons of accuracy and efficiency are carried out for both explicit and implicit time-stepping algorithms. It is found that the efficiency of the GWRM compares well with these methods, in particular at high accuracy. For perturbative scenarios, the GWRM was found to be as much as four times faster than the finite difference methods. A primary reason is that the GWRM time intervals typically are two orders of magnitude larger than those of the finite difference methods. The GWRM has the additional advantage to produce analytical solutions in the form of Chebyshev series expansions. The results are encouraging for pursuing further studies, including spatial dependence, of the relevance of time-spectral methods for NWP modelling. Program summary: Program Title: Time-adaptive GWRM Lorenz 1984 Program Files doi: http://dx.doi.org/10.17632/4nxfyjj7nv.1 Licensing provisions: MIT Programming language: Maple Nature of problem: Ordinary differential equations with varying degrees of complexity are routinely solved with numerical methods. The set of ODEs pertaining to chaotic systems, for instance those related to numerical weather prediction (NWP) models, are highly sensitive to initial conditions and unwanted errors. To accurately solve ODEs such as the Lorenz equations (E. N. Lorenz, Tellus A 36 (1984) 98–110), small time steps are required by traditional time-stepping methods, which can be a limiting factor regarding the efficiency, accuracy, and stability of the computations. Solution method: The Generalized Weighted Residual Method, being a time-spectral algorithm, seeks to increase the time intervals in the computation without degrading the efficiency, accuracy, and stability. It does this by postulating a solution ansatz as a sum of weighted Chebyshev polynomials, in combination with the Galerkin method, to create a set of linear/non-linear algebraic equations. These algebraic equations are then solved iteratively using a Semi Implicit Root solver (SIR), which has been chosen due to its enhanced global convergence properties. Furthermore, to achieve a desired accuracy across the entire domain, a time-adaptive algorithm has been developed. By evaluating the magnitudes of the Chebyshev coefficients in the time dimension of the solution ansatz, the time interval can either be decreased or increased.

The Semi-Implicit Root solver (SIR) is an iterative method for globally convergent solution of systems of nonlinear equations. We here present MATLAB and MAPLE codes for SIR, that can be easily implemented in any application where linear or nonlinear systems of equations need be solved efficiently. The codes employ recently developed efficient sparse matrix algorithms and improved numerical differentiation. SIR convergence is quasi-monotonous and approaches second order in the proximity of the real roots. Global convergence is usually superior to that of Newton's method, being a special case of the method. Furthermore the algorithm cannot land on local minima, as may be the case for Newton's method with line search. 

Time-spectral solution of ordinary and partial differential equations is often regarded as an inefficient approach. The associated extension of the time domain, as compared to finite difference methods, is believed to result in uncomfortably many numerical operations and high memory requirements. It is shown in this work that performance is substantially enhanced by the introduction of algorithms for temporal and spatial subdomains in combination with sparse matrix methods. The accuracy and efficiency of the recently developed time spectral, generalized weighted residual method (GWRM) are compared to that of the explicit Lax-Wendroff and implicit Crank-Nicolson methods. Three initial-value PDEs are employed as model problems; the 1D Burger equation, a forced 1D wave equation and a coupled system of 14 linearized ideal magnetohydrodynamic (MHD) equations. It is found that the GWRM is more efficient than the time-stepping methods at high accuracies. The advantageous scalings Nt**1.0*Ns**1.43 and Nt**0.0*Ns**1.08 were obtained for CPU time and memory requirements, respectively, with Nt and Ns denoting the number of temporal and spatial subdomains. For time-averaged solution of the two-time-scales forced wave equation, GWRM performance exceeds that of the finite difference methods by an order of magnitude both in terms of CPU time and memory requirement. Favorable subdomain scaling is demonstrated for the MHD equations, indicating a potential for efficient solution of advanced initial-value problems in, for example, fluid mechanics and MHD. 

A fully spectral multi-domain method has been developed and applied to three applications within ideal MHD, compressible Navier-Stokes, and a two-fluid plasma turbulence model named the Weiland model. The time-spectral method employed is the Generalized Weighted Residual Method (GWRM), where all domains such as space, time, and parameter space are spectrally decomposed with Chebyshev polynomials. The spectral decomposition of the temporal domain allows the GWRM to reach spectral accuracy in all dimensions. The GWRM linear/nonlinear algebraic equations are solved using an Anderson Acceleration (AA) method and a newly developed Quasi Semi-Implicit root solver (Q-SIR). Up to 85% improved convergence rate was obtained for Q-SIR as compared to AA and in certain cases only Q-SIR converged. In the most challenging simulations, featuring steep gradients, the GWRM converged for time intervals roughly two times larger than typical time steps for explicit time-marching schemes, being limited by the CFL condition. Time intervals up to 70 times larger than those of explicit time-marching schemes were used in smooth regions. Furthermore, the most computationally expensive algorithm, namely the product of two Chebyshev series, has been GPU accelerated with speedup gains of several thousands compared to a CPU.

Time-spectral methods may feature substantial advantages over time-stepping solvers for solution of initial-value ODEs and PDEs, but their effciency depends on the smoothness of the solution. We present two methods to overcome this problem. The first involves transforming the differential equations to an equation for a new variable, related to the time-integrated solution, before applying the solution algorithm. In the second method, a procedure for transformation to exact differential equations of a running average is outlined. Examples of solution of stiff problems and problems with multiple time scales are presented, employing the time-spectral Generalized Weighted Residual Method (GWRM). It is found that the smoothing algorithms have a significant positive effect on convergence.