Studies of copper electropolishing in phosphoric acid-based electrolytes and its application to planarization of ULSI copper wafers.

摘要

As semiconductor device feature size continues to shrink, integration of copper and porous low-k dielectrics becomes inevitable choice for further improvement in device performance. This will provide substantial challenges for chemical mechanical planarization (CMP), which is currently used for copper planarization. These challenges arise primarily from the weak mechanical strength of such dielectrics, which may not withstand the force applied during CMP. Augmenting or replacing CMP with alternatives such as electropolishing becomes necessary to address these challenges and the tight control of copper planarization in the coming years.; Cu electropolishing in phosphoric acid electrolytes has been a prototype for studies of electropolishing. Despite intensive investigation, the mechanism and the identity of rate-limiting species remain controversial. In this work, electropolishing is observed visually as a brightening of the Cu surface in a well-defined mass-transport-controlled limiting-current plateau. The limiting-current density depends on factors such as electrolyte kinematic viscosity, electrode rotation rate, effective diffusion coefficient, and bulk concentration of the rate-limiting species, as predicted by the Levich equation for a rotating disc electrode (RDE). A resistive boundary layer, which is ruled out to be salt film by electrochemical impedance studies, formed at anode surface during electropolishing. The rate-limiting step during Cu electropolishing was found to be diffusion of water, which acts as acceptor species for the dissolving Cu2+ ion. Water diffusion coefficients during electropolishing were approximated from transient and steady-state analyses of current transient curves and in the 10-9 cm2/s and 10 -8--10-7 cm2/s range when the number of water as acceptors for each dissolved Cu2+ ion (sH) is assumed to be 1 and 6 (for two extreme cases), respectively.; In this work, electropolishing of blanket and patterned Cu wafers was investigated. The Cu removal rate of a rotating Cu-wafer electrode at steady state depends on the wafer rotation rate, bulk concentration and diffusion coefficient of water, and electrolyte kinematic viscosity. Uniform Cu polishing rate was obtained across wafer surface, and a better leveling effect was observed in electrolytes with higher acid concentrations or viscosities. Electropolishing was suppressed in the presence of organic diluents. Endpoint of Cu electropolishing could be judged from current transient curve. Planarization efficiency was affected by the geometrical parameters such as profile step height, wavelength, and the diffusion-layer thickness. Over features with large feature widths, step heights and space widths, leveling efficiency was very low, although narrow features were well planarized. To address the topography and pattern dependence of electropolishing on patterned wafers, three different potential methods, "superpolishing", pulse-reverse, and "two-step" methods, have been investigated and evaluated. The experiment results show that planarization efficiency was not enhanced and step heights were not reduced or eliminated by the "superpolishing" and pulse-reverse methods. The "two-step" method in which CMP is performed firstly and then electropolishing is perhaps of greater promise due to smaller Cu dishing produced in this method, compared to the other two methods and the conventional CMP-only or electropolishing-only process.