Solidified F C C metals usually have a {100} fibre texture due to the (100 ) preferential growth direction of dendrites [1-3]. However, under planar interface growth conditions such as in the zonerefining process, the {100} texture is inhibited and {111} fibre texture develops instead [4]. The mechanism of the {111} texture formation is unclear. It is known that planar growth occurs only under directional solidification conditions and, for alloys, only under low growth rate or high-temperature gradient conditions [5]. Velocity of planar growth is much lower than the growth in casting. Under planar growth conditions, the solid/liquid interface is smooth and stable [5]. In this case, grains having low solid/liquid interfacial energies should grow preferentially and overlap those having high solid/liquid interfacial energies: this is preferable, since the overall solid/liquid interracial energy can be minimized. In the case of FCC metals, {111} and {100} planes are the two most closely packed planes having the lowest solid/liquid interracial energies [6]. However, using a molecular-dynamics simulation technique, Broughton and Abraham have demonstrated that the { 100} plane has a lower solid/liquid interfacial energy than the {111} plane [7]. Therefore, grains having their {10 0} planes parallel to the solid/liquid interface should grow preferentially, leading to a {100} texture rather than the {i 11} texture. Such a deduction, of course, disagrees with the experimental observation [4]. So there must be some other factors also affecting the texture formation. In order to understand the preferential growth of the {111} texture, Chadwick used a hard-sphere model to simulate the advance of {111} and {100} planes in solidification [8]. The simulation was made in the following way. Some plastic balls were glued together in the form of square rafts to represent the {111} and {100} planes of a FCC crystal structure, and these rafts were placed at the bottom of a box to act as nucleating planes for crystal growth. Then, the box was filled with several thousand identical balls arranged randomly to simulate the liquid phase. Chadwick observed that the {100} plane grew preferentially when the liquid was loose-packed (the box was filled incautiously by simply dumping the balls), whereas the {111} plane grew preferentially when the liquid was close-packed (by vibrating the apparatus to make the liquid packed down tighter). Chadwick analysed the solid/liquid interface structure in terms of the surface-packing density in the