The paper describes the development of a semi-implicit solver designed for
the need to follow slow evolution flows encountered in nonlinear resistive
computational plasma dynamics (CPD). Spatial discretization uses a second o
rder finite difference approximation while temporal advance is achieved by
a second order semi-implicit predictor-corrector scheme that reduces the se
vere time step constraints imposed by the fast magnetosonic and Alfven wave
s on standard explicit schemes. An efficient preconditioner adapted to magn
etohydrodynamics (MHD) problems and associated with conjugate gradient-like
linear solvers considerably increases the CPU saving when compared with an
explicit advance. This scheme enables the use of a large time step as well
as the necessary high spatial resolution. An application of this scheme to
a class of astrophysical nonlinear MHD problems allows one to perform nume
rical experiments relevant to the slow MHD evolution of the magnetic field,
dominating the outer atmosphere of the sun, that leads to small scales for
mation.