chemistry of additives in damascene copper plating, The

IBM Journal of Research and Development, Jan 2005 by Vereecken, P M, Binstead, R A, Deligianni, H, Andricacos, P C

When the solutions were not purged with nitrogen, the ring currents for Cu were typically much lower because of the fast competitive oxidation of Cu by dissolved oxygen (not shown). Accordingly, the excess currents for accelerated copper deposition at low overpotentials were also lower in the ambient solutions. This observation supports the notion that the acceleration is controlled by the near-surface concentration of Cu and its complexes. The SPS accelerator merely enhances the formation of stable Cu^sup ^ complexes, which enhances the current density as described in previous sections. Dissolved oxygen, on the other hand, decreases the Cu^sup ^ concentration near the surface, thus slowing down the reaction. The role of oxygen in damascene plating may have been underestimated. Depletion of oxygen in trench and via cavities enhances the differential in cuprous concentration and hence the plating kinetics.

In summary, the rotating ring-disk experiments did not show any evidence for the direct electrochemical reduction of SPS to MPS. This observation excludes a catalytic mechanism whereby the adsorbed cuprous thiolate intermediary is formed through the interaction of Cu^sup 2 ^ with MPS. Evidence of considerable SPS consumption was found in the potential range of cuprous ion formation, which was correlated with the formation of Cu(I)(thiolate)^sub ad^. In addition, the presence of SPS in the copper sulfate solutions was found to increase the amount of soluble cuprous complexes near the surface.

Differential plating kinetics

From our results, it is concluded that the near-surface concentration of cuprous ions, intermediates, or complexes is linked to the overpotential and thus the kinetics of copper electrodeposition. The depolarizing effect of rate-accelerating additives (Cl^sup -^ and SPS accelerator) and the polarizing effect of rate-inhibiting additives (suppressor and levelers) follow directly from this cuprous-overpotential linkage. Time-dependent differences in cuprous species inside trench and via cavities and at the planar field may then account for the differential plating kinetics responsible for supcrconformal plating or superfilling. Because of the cuprous-regulating properties of the plating additives, this concept does not contradict existing models based on physical adsorption and the time-dependent differential additive coverage in cavities and the field such as the curvature-enhanced accelerator coverage (CEAC) mechanism. In these models the copper deposition kinetics is determined by the surface coverage O of accelerator or catalyst (and indirectly by suppressor coverage, 1 - θ), and the rapid change in surface coverage inside cavities accounts for the fast bottom-up plating kinetics [6-16]. The actual catalytic acceleration mechanism is of no concern in these models, since the kinetic parameters are obtained from electrochemical measurements. From our results, it follows that acceleration indeed occurs where Cu(I)(thiolate)^sub ad^ is adsorbed at the surface. The increased copper deposition rate is the result of enhanced cuprous-complex formation at these adsorption sites, accounting for the so-called catalytic effect. In addition, a large differential in copper deposition rate is formed between Cu(I)(thiolate)^sub ad^ sites and adsorbed suppressor-chloride sites, where cuprous formation is inhibited. A major advantage of this physical-chemical approach is that concentration changes due to in- and out-diffusion of reagents and products, together with other less obvious solution components, such as Cl^sup -^, O2, and oxidation byproducts, can also be accounted for. A disadvantage is the complexity of the mechanism and the need for sophisticated modeling [42]. However, the main contribution of the physical-chemical mechanism is that it provides an understanding of the underlying mechanisms.