Characterization and Modeling of Dynamic Thermal Behavior in CMP

Existing models of chemical mechanical polishing ~CMP! focus on the impact of pressure, velocity, slurry, and pad parameters on material removal rate. To complement these models, we present a model and experimental data for thermal effects in CMP. The energy source and thermal loss mechanisms are calculated and an energy balance formulation is used to predict the energy exchanged between the pad and slurry flowing over it. Experiments are shown to be consistent with this prediction, suggesting that the most significant elements are accounted for. We also describe the dynamic aspect of the process whereby heat is accumulated in the pad during polish and lost to the slurry flowing over the pad while it rotates around before entering under the head again. Our studies indicate that this heating and cooling cycle may be responsible for the transient behavior observed at the beginning of each polish. To analyze the transient thermal behavior observed in the pad, we propose a lumped parameter dynamic model. Simulation results are seen to match well with experimental data for copper polishes. © 2003 The Electrochemical Society. @DOI: 10.1149/1.1560642# All rights reserved. Characterization of chemical mechanical polishing ~CMP! removal rate and rate uniformity have traditionally focused on the use of Preston’s equation to model the mechanics of the polishing process, in which removal rate R is proportional to the product of pressure Pr and relative velocity v with proportionality given by Preston’s coefficient k p ,o rR 5 k pPrv. Depending on the hardness of the polishing film ~e.g., oxide, copper, tungsten! and barrier film ~e.g., nitride, TaN, TiN!, and the chemistry of the slurry, the temperature may vary significantly during polish. A factor not normally considered in Prestonian CMP models is the impact of changes in temperature on material removal. An experiment performed by Chiou et al. examined the chemical removal rates of copper at different temperatures and for different slurries and found that changes of 20°C can result in an order of magnitude increase in removal rates. 1 More recent studies of copper polishes have also observed similarities between the transient increase in temperature due to polish and the increase in removal rate. 2 Other work has observed common patterns between the spatial variation in removal rates for oxide and copper polishes and variation in temperature across the polishing pad. 3 Models for the evolution of topography at the chip and feature scale, including dishing and erosion, all require an equivalent blanket material removal rate parameter. 4,5 Thus an understanding of thermal behavior during polish has become increasingly important for better characterization of CMP at the wafer and chip scale.