Ultralow-energy amorphization of contaminated silicon samples investigated by molecular dynamics

Ion beam processes related to focused ion beam milling, surface patterning, and secondary ion mass spectrometry require precision and control. Quality and cleanliness of the sample are also crucial factors. Furthermore, several domains of nanotechnology and industry use nanoscaled samples that need to be controlled to an extreme level of precision. To reduce the irradiation-induced damage and to limit the interactions of the ions with the sample, low-energy ion beams are used because of their low implantation depths. Yet, low-energy ion beams come with a variety of challenges. When such low energies are used, the residual gas molecules in the instrument chamber can adsorb on the sample surface and impact the ion beam processes. In this paper we pursue an investigation on the effects of the most common contaminant, water, sputtered by ultralow-energy ion beams, ranging from 50 to 500 eV and covering the full range of incidence angles, using molecular dynamics simulations with the ReaxFF potential. We show that the expected sputtering yield trends are maintained down to the lowest sputtering yields. A region of interest with low damage is obtained for incidence angles around 60° to 75°. We also demonstrate that higher energies induce a larger removal of the water contaminant and, at the same time, induce an increased amorphization, which leads to a trade-off between sample cleanliness and damage.

: Evolution of the amorphization coefficient for a specific case comparing 50 and 500 eV for each angle.     Figure S7: Probability to fragmentation a water molecule per impact with respect to the fluence for each energy at 0°, 30°, 45°, 60°, 75°, 83°. Figure S8: Probability to fragment a water molecule per impact with respect to the fluence for each angle for a) 50 eV impacts and b) 500 eV impacts. Table S1: Force field parameters, taken from [1].  60°, e) 75° and f) 83°. Each energy is displayed along the y axis, while the implantation depth is along the x and the number of counts is displayed along the z axis.

References
S6 Figure S4: Implantation depths of hydrogen for incidence angles of a) 0°, b) 30°, c) 45°, d) 60°, e) 75° and f) 83°. Each energy is displayed along the y axis, while the implantation depth is along the x and the number of counts is displayed along the z axis.

Sputtering of clusters
For higher impact energies, material can be sputtered in different ways: -Direct sputtering by elastic collisions [46] from the incident argon. This means the argon collides directly with either a contaminant, a silicon or an argon atom and sputters it.
-Indirect sputtering by elastic chocs from the collision cascade. In this case, the initial collision between the incident argon and a target atom transmits the energy into the lattice, which leads possibly to other displaced atoms in the sample. These displacements can sputter particles when occurring at the sample surface.
-Backscattering of Argon particles. This mechanism usually concerns only argon that is a noble gas and therefore does not form bonds in the sample. This causes argon to be very volatile.
These mechanisms play a role in the final sputtering yields. An interesting outcome of the simulations are the mechanisms leading to the sputtering of clusters. In Figure S5 we plotted the results of the analysis of the sputtered clusters with the highest sputter yields.
We included several other partial sputtering yields in the supporting information, such as oxygen -oxygen. A note on the method, clusters were detected using a threshold on interatomic distance. For silicon clusters, the interatomic threshold we set is inferior to the Si -Si bond length for clustering. Consequently, a direct observation for silicon -silicon clusters in general is that in most cases, silicon atoms are sputtered as Si2 clusters. The partial Si -Si yield is also a non-negligible fraction of the pure silicon yield (up to 40% for the angles between 60 and 75°). This mechanism is enhanced for increasing impact energies and for larger incidence angles. The second most abundant sputtered clusters S8 are entire water molecules. Since they are deposited on the sample surface, there is at the beginning of the bombardment a high probability to sputter intact water molecules, or to fraction them. Following bombardments can also sputter silicon -oxygen clusters. Due to the strong bond between silicon and oxygen, the silicon -oxygen pair has the same sputtering yield than silicon -hydrogen, despite the fact of hydrogen being twice as abundant in the sample. Finally, very few oxygen -hydrogen, oxygen -oxygen and hydrogen -hydrogen clusters are sputtered (cf. Figure S6).
The previously described trends are increased by higher impact energies and larger incidence angles: the sputtering yield of silicon -silicon clusters is enhanced in these conditions similarly to the silicon sputtering yield. We can also observe a similar trend for water related clusters. The trend for water removal is biased by the non-renewal of the water layer, yet we can observe that higher impact energies and incidence angles favor the removal of water molecules, with a maximum at 500 eV and 75°. We can also observe good conditions of water removal in combination with a lower sputtering yield for silicon at 83°. Since the water layer is not renewed in between bombardments, another interesting observation is to measure the probability per collision to fraction a water molecule in order to determine if some water molecules remain intact on the surface after the 500 bombardments.
From these observations we can conclude that silicon sputtering is mainly done via clusters. When comparing the silicon yield to the cluster yield, we measure that in approximatively 70% of the cases, silicon is sputtered as a Si2 cluster. It is interesting to observe that a non-negligible fraction of the sputtered clusters are including a contaminating particle, approximatively 20% when adding the yield of Si -O and Si -H.

S11
Considering the fixed amount of water molecules initial, and a substantially lower count of single hydrogen and oxygen particles in the sample, this indicates that the contaminating particles will strongly interact with the silicon particles. Furthermore, even though the atomic ratio of oxygen should be half the atomic ratio of hydrogen, the measured clusters is similar between Si -O and Si -H. This underlines the strong interaction happening between silicon and oxygen particles. Once the contaminating particles are split, they will stick to silicon and consequently will modify the local structure, allowing for these clusters to be sputtered in the later part of the bombardments.

Evolution of the fraction of intact water molecules
The implantation of water species and the amorphization of the sample is strongly correlated to the fragmentation of water molecules during the argon bombardment.
Indeed, when water molecules are dissociated, free oxygen atoms or O -H fragments are liberated near the surface, increasing the probability of the formation of a siliconoxygen bond. Once formed, a silicon -oxygen bond will be notably hard to break and will tend to be sputtered as a whole, as shown in previous section. Thus, water fragmentation plays a key role in the amorphization of the sample, since the bond of Si -O clusters will be i) much stronger than Si -Si or Si -H, and ii) shorter than Si -Si bonds. Therefore, it is easy to understand how the formation of Si -O bonds will modify the local structure, and thus, as shown in previous sections, have an impact on the amorphization coefficient.
One of the objectives, i.e. minimizing the impact of the water contamination on the sputtering process, is to remove as much water molecule as possible by sputtering, while minimizing its fragmentation and implantation. In Figure S7 the probability (in percentage) S12 to fragment one water molecule per impact is plotted with respect to the fluence. Since the water layer is not renewed throughout the simulations, at the end of the bombardments the number of intact water molecules tends towards zero. The evolution of the water dissociation probability can be fitted by the ( / ) function. A steep curve indicates fast water fragmentation, while a slowly decreasing curve indicates the presence of intact water particles until high fluences. With increasing bombardment energy, we can observe an increased water fragmentation probability: a higher energy induces bigger collisions cascades, which in return have a higher probability to interact with water molecules and thus have a higher fragmentation probability. Despite this increased probability of fragmentation, we observed that even at a 50 eV impact energy, there are almost no intact water molecules left at the end of the simulations. We also observe that an incidence angle of 60° maximizes the water fragmentation for low impact energies. For high impact energies, the water dissociation probability is highest for large incidence angles. When considering the overall behavior of the water molecules, there is a competition between the water molecule fragmentation and its sputtering. We observed in the sputtering yield section that the intact water molecule yield is maximum at higher incidence angles and impact energies. The probability of fragmentation also increases with respect to the angle and energy. In Figure S8 we plotted the water dissociation probability for 50 and 500 eV collisions, for each angle we selected. From this plot we can deduce a lower chance of water fragmentation at lower energies. At an incidence angle of 83° we observe the lowest probability to fragment water molecules for both 50 and 500 eV. On the other hand, at 500 eV, we still have a significant intact water molecules S13 sputtering yield which could indicate an interesting set of parameters for water removal while also retaining a shallow amorphous layer on the sample surface. When combining the information of the water fraction with the sputtered cluster, it becomes clearer how much the water contamination will impact the sputtering processes.
Due to the high probability of fractioning a water molecule in the earlier stages of the irradiation process, the remaining oxygen and hydrogen particles will have a strong probability to directly interact with silicon particles. Due to this interaction, it will create scenario in which the presence of oxygen or hydrogen will enhance the sputtering of silicon. This behaviour in our simulation remains limited (up to 20% as shown in the previous paragraph), yet, in a scenario where the water layer would be renewed, this could lead to a substantial increase in the sputtering yield of silicon due to the presence of water.