We consider here the problem of constructing an efficient algorithm fo
r a classical Monte Carlo simulation of crystalline polyethylene with
unconstrained bond lengths and angles. This macromolecular crystal, pr
esents a particular example of a system with many different energy sca
les, ranging from soft ones represented by nonbonded van der Waals int
eractions, to stiff ones, represented in particular by bond stretching
. A proper sampling of all the energy scales poses a problem and it is
shown that a standard Metropolis algorithm employing just local moves
is not very efficient at low temperatures. As a solution it is propos
ed to employ also global moves consisting of displacements of the cent
er of mass of the whole chains in all three spatial directions as well
as rotations of the chains around an axis parallel to the It-axis, wh
ich act on the degrees of freedom associated with the chain packing. I
t is shown that by properly alternating such global moves with standar
d local moves, the statistical inefficiency of the algorithm is consid
erably reduced.