This project deals with energy dissipation processes at solid surfaces that are induced by the impact of atoms or molecules with velocities ranging from about 10 m/s to 10 km/s. The imposed kinetic energy dissipates by means of elastic collisions (“nuclear stopping”) and electronic excitation processes (“electronic stopping”). For impact energies in the keV range, it is well known that nuclear stopping largely dominates the energy loss experienced by the projectile, thus generating a cascade of atomic subsurface collisions which can lead to the emission of surface atoms into the gas phase (sputtering). However, part of the kinetic energy is converted into electronic excitation, which manifests, for instance, as the presence of hot electrons in excited states above the Fermi level.

Both processes are investigated by i) mass spectrometric characterization of the sputtered particle fluxes emitted from the surface and ii) spectrometric detection of hot electrons with excitation energies below the work function of the sample. Particular emphasis is put on the comparison between processes induced by atomic and polyatomic projectile species, since in the latter case the collision cascades induced by the constituent atoms overlap in space and time, generating a nonlinear superposition effects which are to be studied here.

Mass Spectrometry

An example for the type of experiments performed to characterize the flux of material emitted from the surface is shown in Fig. 1, which depicts the measured velocity distribution of neutral atoms and clusters emitted from a polycrystalline indium surface under bombardment with Au- and Au2- projectiles of the same impact velocity. These distributions carry a clear signature of the nature of the sub-surface collision processes induced by the projectile impact. For the monoatomic projectile, the data coincides with what is expected theoretically for a so-called linear collision cascade (solid line). Note that the distributions measured for atoms and dimers are largely different.

Switching to diatomic projectiles leads to a drastic change of the distributions towards low emission velocities. Moreover, atoms and clusters are now ejected with practically identical velocity distributions. Both observations indicate a drastic change the nature of the emission process, which is now dominated by a phase explosion from a dense collisional spike.

Electron Spectroscopy

To detect hot electrons in low-energy excited states located between the Fermi level and the vacuum niveau, we utilize a Metal-Insulator-Metal (MIM) tunnel junction. The working principle of such a device is illustrated in . The top electrode represents the actual surface bombarded by the projectile beam. The central idea is that weakly excited electrons (~ 1 eV above the Fermi level) generated in this layer can travel to the insulator interface, tunnel through the thin oxide barrier and are detected as an ion bombardment induced tunneling current in the underlying aluminum base electrode. In order to allow ballistic transport of the hot electrons to the interface and – at the same time - prevent damage to the oxide interface due to penetration of projectiles, the top electrode thickness is chosen as about 20 nm.

As is seen in Fig. 2 a tunneling current into the base electrode is only observed if the impact zone of the pulsed projectile beam is located inside the junction area (defined by the overlap of top and base electrodes). Kinetic excitation mechanisms like electronic friction or electron promotion in close collisions may lead to direct transfer of kinetic energy to the electronic system. These processes must strongly depend on the impact energy of the projectile. Furthermore, kinetic excitation can be superimposed by effects induced by the potential energy of the projectile charge. In order to differentiate between both excitation mechanisms, Fig. 3 shows the measured tunneling yield (i.e., the tunneling current normalized to the projectile flux) as a function of the projectile energy for singly charged Ar+ ions as well as for neutral Ar atoms impinging onto a polycrystalline silver surface.

The results show a clear dependence on the kinetic energy, thus indicating that the excitation is due to kinetic processes. It can be seen that the signals for Ar+- and Ar0-bombardment are practically the same for low projectile energies, whereas for higher impact energies electronic excitation induced by the ions exceeds that induced by the neutrals. This finding is understandable in terms of the interaction time of the projectile with the surface before the actual impact. When the ion approaches the surface slowly, electron transfer processes like Auger neutralization are fast enough to fully neutralize the ion before it penetrates into the solid.

In this case, there is obviously no difference between the signals measured for ion and neutral projectile bombardment. At higher velocities, the neutralization process is not fully completed, leading to a growing probability to reach the surface in the charged state. Since the electronic stopping power for the charged particle is larger than that for the corresponding neutral species, a larger tunneling current is observed for the initially charged projectile.

The data displayed in Fig. 3 have been acquired with zero bias voltage across the tunnel junction. By applying a negative bias voltage to the aluminum base electrode, it is possible to generate an energy dispersive element permitting energy spectroscopy of the excited electrons. As an example, Fig. 4 shows the measured tunneling yield as a function of the bias voltage. At first sight, the observed signal decrease with increasing negative bias voltage appears to be qualitatively expected from the increasing energy discrimination. The same is true for the signal increase at positive bias voltages. A very interesting finding is the change in the sign of the tunneling yield for voltages below -1V. Analysis shows that this observation is due to the detailed shape of the tunneling barrier, leading to contributions of both excited electrons and holes to the measured current, the relative weight of both contributions being controlled by the bias voltage. A detailed interpretation of this experiment is currently under way, simulating the effect of ion induced excitation by means of a temporally increased electron temperature in the silver layer during the lifetime of the collisional spike induced by the projectile impact. Preliminary results reveal electron temperatures of the order of 1000 K to be reached within the collision cascade initiated by the projectile impact.

Theory

The experimental results are compared to a new computer simulation model which allows to incorporate low-energy electronic excitation into the molecular dynamics (MD) simulation of atomic collision cascades in solids. The transfer of kinetic into electronic excitation energy is described in the frame of a quasi-free electron gas model. It is assumed that the friction-like electronic energy loss experienced by all moving atoms leads to a space and time dependent electronic excitation that spreads around the original point of generation with a diffusivity D. The resulting excitation energy density E(r,t) is parametrized in terms of a space and time dependent electron temperature Te. Although the underlying assumption of quasi-instantaneous energy equipartition among the electron system is not generally justified, corresponding ab-initio DFT-calculations carried out in Project B5 reveal that the original energy spectrum generated by electronic friction is dominated by low-energy excitations and therefore closely resembles a Fermi distribution at all times.

First pioneering calculations of electronic excitations in an atomic collision cascade initiated by the 5-keV impact of an atomic silver particle onto a monocrystalline Ag(111) surface have not only demonstrated that electron temperatures of several thousand Kelvin may be reached, but also clearly indicated the necessity to account for the influence of space- and time-dependent reduction of crystallographic order caused by the evolution of the collision cascade on the diffusivity D. In our most recent work, we have therefore extended the model towards a full three dimensional treatment based on a finite differences approach which allows for a spatial and temporal variation of D, which can then be coupled both with local temperature and lattice disorder. The latter is quantified in terms of a local order parameter calculated from the time dependent atom positions delivered by the MD calculation.

As an example of such a calculation, Fig. 1 depicts temporal snapshots of the resulting lateral distribution of (a) the electron temperature and (b) the diffusivity D evaluated immediately at the surface, i.e., in the uppermost cell layer of the model crystallite. In order to emphasize the role of lattice dynamics, Fig.1 (c) shows a three-dimensional animation of the corresponding MD trajectory.

A more complete set of animations can be viewed electronically at our web address http://www.exp.physik.uni-due.de/wucher/.

Figure 5: Snapshots of (a) the electron temperature Te and (b) the diffusivity D at the surface; (c) shows the corresponding trajectory pictures in a perspective view.

At 150 fs after the projectile impact, the temporal and spatial evolution of the collision cascade has already led to sputter emission of two surface atoms as well as to a significant damage of the crystallographic order within a circular near-surface volume of approximately 1 nm diameter around the original impact point. By inspecting the diffusivity distribution at that time, the correlation between lattice disorder and the magnitude of D becomes clearly visible. The surface electron temperature exhibits a much broader distribution in connection with much smaller absolute values, mainly caused by the rapid diffusion of the original excitation. This distribution is superimposed by local excitation sources originating from fast recoil atoms moving within the surface layer.

The next snapshots captured at t=350 fs reveal (i) the onset of massive sputtering and (ii) the spatial spread of the collision cascade predominantely propagating in direction towards the front-right crystal edge. Looking at the diffusivity plot, one finds a nearly homogenous distribution with rather low values of D?0.5-5 cm2/s . The corresponding electron temperature distribution peaks at values around 1000 K and still shows prominent local structure reflecting the ongoing kinetic heating of the electronic sub-system by fast recoils. Finally, at t=750 fs an electron temperature distribution is obtained which exhibits a gaussian-like shape with a maximum of approximately 2000 K. The location of the maximum excitation correlates with the core of enhanced collision dynamics and crystallographic disorder. To allow a better visualization of the time dependence of surface excitation, the temporal evolution of Te has been calculated for different radial distances r from the impact point.

To allow a better visualization of the time dependence of surface excitation, the temporal evolution of Te has been calculated for different radial distances r from the impact point.

There are several interesting observations. Shortly after the impact, the predicted excitation exhibits a sharp peak of about 10 fs duration. The maximum electron temperature reached in this time interval is calculated as 6500 K, decreasing with increasing distance from the impact point. After the projectile has passed the surface layer, the electron temperature is found to rapidly decay due to the onset of diffusion without any notable electronic energy source term, until at times of about 50 fs values close to room temperature are reached.

Probably the most striking observation in Figure 6 is the fact that, after this initial decay, the surface excitation is found to rise again, leading to a second maximum of Te at times around 500 fs. This finding is of utmost importance, since it reflects the trapping of electronic excitation in a collision cascade by means of local heating and, more importantly, atomic disorder. The electron temperature reached in the later stage of the cascade is clearly sufficient to influence the ionization and excitation processes of sputtered particles leaving the surface. Moreover, the time scale at which these temperatures are reached almost perfectly coincides with that of maximum particle emission . A shortcoming of the present model is the fact that electronic excitations via electron promotion in violent binary collisions have not yet been taken into account. Our ongoing present work focuses on the proper inclusion of those processes into our nonlinear diffusion model in order to study the relative role of both (electronic friction and electron promotion) excitation.