A computational study of the double hysteresis phenomenon during reactive sputtering

Abstract

When process parameters such as the reactive gas partial pressure or the discharge voltage are studied as a function of the reactive gas flow during reactive magnetron sputtering, the obtained curve has an S-shape. A direct consequence of this behaviour is that process control based on the reactive gas flow exhibits hysteresis. Under specific conditions, it is possible to observe two S-shaped curves: one when the reactive gas pressure is increased, the other during the return to the initial state by decreasing the reactive gas pressure. This behaviour has been described as double hysteresis behaviour. The origin of the double hysteresis behaviour is computationally studied by high-throughput calculations using a previously developed model. The influence of different process and material parameters were evaluated based on newly developed measures to characterize the calculated process curves. This high-throughput analysis reveals that the double hysteresis behaviour is linked to the difference in the removal rate of non-reacted implanted ions during the increase and decrease of the reactive gas pressure. Within the parameter space a region can be defined for which the double hysteresis behaviour is strong. The latter can not only assist further experiments to study this behaviour but also defines conditions to limit its impact. For Al, a discharge current density of approximately 0.025~A/cm2 was found to maximize double hysteresis.

Influence of chemisorption on the double hysteresis phenomenon during reactive sputtering

“…The model is implemented in a freely available software package [16]. Details on the model can be found in the references [2,15,4,14]. To ensure a good readability, the same symbols as in Van Bever et al [14] have been used in this paper.…”

“…For our analysis, abstraction is made of the exact shape of the process curves. Instead, hysteresis quantification measures were introduced [14].…”

“…The high-throughput analysis of Van Bever et al [14] was continued. The process and material parameters in the RSD-model are first scanned over a wide range.…”

“…Within the approximations of the RSD-model, the target state is determined by the reactive gas fraction [15] and the discharge voltage can be determined from the target state by using the IV-characteristics of the magnetron sputtering discharge [15,14]. Hence, the lack of a substrate does not affect the V (P R 2 ) process curve and B hyst [V (P R 2 )] does not disappear.…”

“…In our previous paper [14] the double hysteresis was computationally studied, and its origin was quantitatively explained. The implanted reactive gas ions have a limited time to react with the target material before they are removed by sputtering.…”