In this paper, we study systems of functional equations (FEs) and firs
t-order partial differential equations (PDEs) suggested in [SIAM J. Sc
i. Statist. Comput., 12 (1991), pp. 607-647 and SIAM J. Numer. Anal.,
29 (1992), pp. 1741-1768] as approximations for the computation of inv
ariant tori. The main new ideas of this paper are, first, to investiga
te these systems in the setting of Hilbert spaces rather than in the s
etting of Banach spaces and, then, to employ Fourier methods instead o
f difference methods for a numerical solution. Based on the setting of
Sobolev spaces H-s(T-p), proper conditions for the PDE and FE systems
to be dissipative are described, and some regularity results for the
FE system are proved. We studied two kinds of Fourier methods, the spe
ctral method and the pseudospectral method, in detail under dissipativ
ity conditions. Convergence and optimal error estimates are shown theo
retically for these Fourier methods in the case of general linear syst
ems. Numerical results for three examples provided in the last section
indicate that the Fourier method behaves very well not only for smoot
h solutions but also for nonsmooth solutions.