Computer Simulation of Sputtering II

Abstract

The conclusions of the first paper in this series have been confirmed by simulations in which the copper target is represented by a composite potential function consisting of a Born-Mayer repulsive potential segment, a cubic potential matching segment, and a Morse potential attractive segment. Surface layer relaxation has been included, and surface layer atom binding energies for the most primitive planes of copper have been determined to be: Eb(100) = 2.4 f 0.1 ev, Eb(100)= 2.1 f 0.1 eV and Eb(ll1) = 2.4 f 0.1 ev. For argon sputtering copper there is no detectable change in the spot patterns between the two models, and the sputtering yields agree within the uncertainty of the simulation. Sputtering yield vs. energy curves now agree quite closely with the experimental data. For sputtering at 5 ke-V the energy distribution of the supttered atoms appears to have the form dN/d - E. The argon-copper pusttering efficiency matches smoothly into the polycrystaline experiemental data reported by H. H. Andersen. Oblique ion incidence simulationson the (100)copper surface are compared with the5keV data of Southern, Willis, and Robinson, and with the 20 keV data from the Amsterdam group. The simulations indicate that there may be fine structure in the yield vs. angle of rotation curves. Well-defined peaks with widths less than one degree have been identified at points on the curve where they might reasonably be expected to appear. The simulations exhibit a finestructure anomoly which may explain the “cloudy region” experimental anomoly in the neon-copper system. Simulations at 20 keV have succeeded in producing focuson sputtering from a sixth atomic layer primary collision. These focuson events occur with very low probability, High energy (> 10 keV) oblique incidence sputtering has been simulated as a two stage process. In the first stage energy is deposited in primary knock-on atoms at various depth below the crystal surface. The second stage is a yield per primary knock-on atom as a function of depth, energy, and direction of recoil. Except for (110) direction recoils, the recoil atom collision mechanism is characteristically a cascade rather than a collection of focusons. This two stage simulation produces yields which agree surprisingly well with the 20 keV experimental data.