Design and evaluation of pad grooves for copper CMP

Variations in the chemical mechanisms of copper chemical-mechanical planarization (CMP) can appear due to the effect of pad grooving on (i) net flow under the wafer, (ii) pad, wafer, and slurry temperature, and (iii) reactants and polish debris concentration. Furthermore, changes in the mechanical abrasion of the passive film might appear due to the effect of pad grooving on (i) slurry film thickness under the wafer, (ii) friction force at the pad-wafer interface, (iii) pad compressibility, and (iv) pad-wafer contact area. The effective transport of slurry in and out of the pad-wafer interface becomes critical particularly for processes in which by-products are detrimental to polishing rates. By combining logarithmic and spiral grooves, paths are created to introduce fresh slurry into, and spent slurry and debris out of, the pad-wafer interface. The experimentally grooved pads were tested and statistically compared to a commercial pad in terms of removal rate (RR), average coefficient of friction, and average pad leading edge temperature. Also a flat (i.e., not grooved) pad was included in this study to evaluate in general the effect of pad grooves in copper CMP. The results indicate that the pad achieving the highest relative values for RR, coefficient of friction ((COF) over bar), and (T) over bar (p) is the one that combines a negatively directed logarithmic groove with a positively directed spiral groove. This pad results in a 24% increase in RR and a 28% increase in (COF) over bar compared to the concentrically grooved pad. To establish the mechanical and chemical contributions to the process, experimental data were then theoretically evaluated. A three-step model in combination with a previously developed flash heating (FH) temperature model was proposed for copper CMP. In all cases, the model root-mean-square (rms) error fell in the range of 322-674 angstrom/min, while the experimental repeatability error was in the range of 118-1100 angstrom/min. This model presented an expression to characterize the rate of oxide growth (k(1)) and the addition of a third step to characterize the dissolution rate of copper oxide (k(3)). The relative values of k(1) and k(2) (mechanical rate constant) as a function of pV showed that the process was more limited by film removal through mechanical abrasion, especially at low values of pV. However, as pV increased this limitation was reduced and there was a transition to a more balanced process.