Effect of Oxygen Content on Particle Generation in TiN Reactive Sputtering
S. S. Kulkarni, C. Fisher, P. S. Gilman
Abstract: Particle generation during TiN reactive sputtering is one of the main causes leading to yield losses in VLSI manufacturing. Particle generation has become a more serious concern with the eminent move towards < 0.18m m – 0.12m m feature dimensions. Several factors affecting particle generation have been identified including target geometry, sputtering process variables, surface finish, target microstructure and have been well documented. The overall target purity and oxygen content have been observed in some cases to affect particle and defect generation. Unfortunately, definitive studies correlating oxygen content to particle generation have not been reported. Consequently, there has been a lack of scientific explanations that correlate the oxygen content and overall purity to the particle generation.
In order to fill this gap, particle generation in TiN films sputtered from targets with varying oxygen contents has been studied. The overall purity levels, surface finishes and other factors affecting particle generation were held constant in order to isolate the effect of oxygen content on particle generation. It is seen that the particle count decreases with decreasing oxygen levels. A theory is proposed to co-relate the particle count to oxygen levels in the titanium targets.
Targets: G-12 Endura™ targets were sputtered in an Applied Materials Endura 5500, 8 inch system. The targets were fabricated from 5N titanium with an oxygen content ranging from 55ppm to 250 ppm. All the targets were fabricated with an enhanced surface finish, in order to ensure that surface imperfections were not the source of particle generation.
Process Conditions: The following process conditions were used during the TiN sputtering: -
In addition, titanium pasting was done after every 25 wafers and the shields were changed after every 250 KWh. The testing was carried out only to 650 KWh as at higher KWh’s, the main source of particles/defects is nodule flaking/rupture. This is makes is difficult to isolate the effect of target oxygen content on particle generation. The particles, ranging from 0.3m m – 7.5m m were measure on a Tencor 6200 system, after every 50KWh.
As seen from the graph above it is evident that the number of particles generated by the low oxygen target is lower than those generated by the high oxygen target. The average number of particles/cm2 for the low oxygen target is 0.03273 versus 0.05166 for the high oxygen target.
It has been demonstrated (1,2) that the main cause of nodule formation on the target surface is the differential between the sputtering rate and the redeposition rates. Factors like grain size i.e. grain boundaries, surface defects, strength of magnetic field influence this differential and hence affect particle generation during TiN reactive sputtering.
It is proposed that the oxygen levels in the target also influence nodule formation by affecting the sputtering versus re-deposition rates. During sputtering, oxygen from the target is ‘liberated’ at the target surface either in atomic form or in form of an oxide. Either way, the free oxygen atoms have a strong affinity to react with the titanium surface, which is free of any oxide layers. This results in a reaction at surface of the target, which effectively diminishes the sputtering rate, initially at a microscopic scale. At locations such as those near the racetracks, where the magnetic field is strong, the sputtering rate is high enough to overcome this decrease in sputtering rate. But in other locations such as those near the edge of the target, this decrease in sputtering rate can cause the nucleation of TixNy particles. These particles in-turn further affect the sputtering rate and lead to the growth of TixNy particles.
Thus decreasing the oxygen levels in TiN reactive sputtering targets can help in decreasing the defect density in the deposited thin film. Further tests are planned in order to get a quantitative link between these two factors.
On a side note, it has been observed that lowering the oxygen levels can also influence the target microstructure, i.e. grain size and texture. These factors also affect particle generation, and hence the microstructure of the target has to be correspondingly ‘tuned’ in order to derive maximum benefits of decreasing the oxygen levels.