Given a geometry, electronic structure may be calculated for even very complex materials within the local density approximation. However, geometrical sturctures of very complex materials or disordered systems are often not known. A major step toward the achievement of first principle predictions of the structural and dynamical properties of very complex materials has been made possible by the Car-Parrinello method of ab intio molecular dynmics stimulation. While this approach when implemented on supercomputers allows the simulation of many interesting problems of even more importance are still to large. On the otherhand, the appearance of the massively parallel processor architecture of the machine is taken into account. Preliminary testing of Car-Parrinello simulation on the Connection Machine indicates that it maps well onto parallel architectures. Provided that the code is fully optimized for the Connection Machine it will be possible to predict strutural and dynamical properties of complex materials form first principles. Examples would include simulation of surface oxidation, molecular/cluster interaction with surfaces and ab initio studies of scanning tunnel microscope and atom force microscope images.
