Molecular-dynamics simulations of electronic sputtering E. M. Bringa and R. E. Johnson Engineering Physics, University of Virginia, Charlottesville, Virginia 22903 M. Jakas Engineering Physics, University of Virginia, Charlottesville, Virginia 22903 and Departamento de Fı ´sica Fundamental y Experimental, Universidad de La Laguna, 38201 La Laguna Tenerife, Spain Received 11 May 1999 Following electronic or collisional excitation of a solid by a fast ion, an energized cylindrical region is produced which can lead to sputtering. Here ejection from such a region is studied via molecular-dynamics simulations using Lennard-Jones and Morse potentials. Over the full range of excitations studied the yield vs the energy release per unit path length in the solid, which we call dE / dx , is shown to scale with the binding energy and with the density of the material for all materials studied and at all dE / dx . This allows the simulation results to be applied to low-temperature, condensed-gas solids and to more refractory solids over a broad range of dE / dx . The effect of a distribution of energies for the initial energizing events, and the effect of a spatial distribution of such events for a given dE / dx are examined. Three regimes have been identified. When the energy release per excitation event is greater than the escape energy, sputtering is linear in dE / dx at low dE / dx . With increasing dE / dx a spikelike regime occurs in which the yield again becomes nonlinear with dE / dx . For fixed cylindrical radius ejection then saturates so that at very high dE / dx the yield again becomes nearly linear with dE / dx . In this regime the size of the yield increases with the initial radial extent of the track and is determined by the removal of energy radially by the pressure pulse and by the transport of energy from depth to the surface. Therefore, the clear nonlinearities observed in the knock-on sputtering yields by heavy ions require consideration of the radial extent of the cascades. For electronic sputtering yields of condensed-gas solids, the observed nonlinearity in the sputtering yield suggests that the radial extent of the excited region varies in a manner different from that predicted or that the energy release to the lattice is nonlinear in the stopping power. S0163-18299902146-3 I. INTRODUCTION In two recent papers 1,2 hereafter I and IImolecular- dynamics MDsimulations of a cylindrically energized re- gion were carried out in order to describe the transport of energy from a track of excitations produced by a fast ion and to test ‘‘thermal’’ spike models for sputtering. Although spike models have been used for over 50 years to describe the nonlinear aspects of sputtering, 3,4 ion-beam induced mixing, 5,6 and track formation 7,6 by fast ions, there have been relatively few tests using atomistic simulations. One of our goals is to understand the applicability of spike models to laboratory results on the electronic sputtering of low- temperature condensed-gas solids. 8–10 Such experiments pro- vide one of the few means of determining the nonradiative electronic relaxation pathways in molecular insulators. 8,9,11 The electronic sputtering of low-temperature ice is also of interest as it produces atmospheres on the moons of Jupiter. 12 Laboratory studies of fast ions incident on low- temperature condensed-gas solids show that the yields for the molecular solids all exhibit a roughly quadratic depen- dence on dE / dx at high dE / dx . Here dE / dx is the energy deposited per unit path length by a fast ion and is also called the stopping power. 13 Spike models, in which the energy transport is diffusive, also give yields quadratic in dE / dx for the cylindrical geometry appropriate to fast ions. In such models, of course, dE / dx is the kinetic energy per unit path length of the moving atoms or molecules in the spike. As- suming this value of dE / dx is proportional to the stopping power, the laboratory observations have been parametrized using spike models. 14 This parametrization allows one to beautifully scale data for different targets over a broad range of dE / dx . 14 However, we found to our surprise that the yield calculated using MD simulations for initial conditions asso- ciated with a spike is not quadratic in dE / dx for those values of dE / dx appropriate to the nonlinear electronic sputtering regime. 15,2 Quite remarkably, the sputtering yields calculated using MD for fixed track radius are nearly linear in dE / dx at high dE / dx even though the transport processes are clearly nonlinear. That is, above some ‘‘threshold’’ value of dE / dx saturation sets in unlike in the analytic spike models. We showed that the difference in the dependence of the yield on dE / dx obtained in the MD simulations from that dependence obtained using a spike model is not due to a lack of thermal equilibration locally in the rapidly evolving spike, and we showed earlier that local equilibration was not required in order to obtain the quadratic dependence of the yield. 14 In addition, Jakas 16 showed that the quadratic dependence in spike models in which the energy transport is diffusive is robust since it persists even when realistic thermal properties which allow melting are used rather than the analytic model properties typically used. The difference between our MD simulations and the spike models is due to two factors. First, the energy trans- ported away from the excited cylindrical region at high dE / dx cannot be described diffusively but is more closely PHYSICAL REVIEW B 1 DECEMBER 1999-II VOLUME 60, NUMBER 22 PRB 60 0163-1829/99/6022/1510710/$15.00 15 107 ©1999 The American Physical Society