No abstract

Pattern-dependent effects are a key concern in chemical-mechanical polishing (CMP) processes. In oxide CMP, variation in the interlevel dielectric (ILD) thickness across each die and across the wafer can impact circuit performance and reduce yield. In this work, we present new test mask designs and associated measurement and analysis methods to efficiently characterize and model polishing behavior as a function of layout pattern factors-specifically area, pattern density, pitch, and perimeter/area effects. An important goal of this approach is rapid learning which requires rapid data collection. While the masks are applicable to a variety of CMP applications including back-end, shallow-trench, or damascene processes, in this study we focus on a typical interconnect oxide planarization process, and compare the pattern-dependent variation models for two different polishing pads. For the process and pads considered, we find that pattern density is a strongly dominant factor, while structure area, pitch, and perimeter/area (aspect ratio) play only a minor role.

The electrochemical behavior of tungsten during chemical mechanical polishing (CMP) was observed in order to investigate a proposed blanket passivation and abrasion mechanism for tungsten removal. The experiments were performed in a cell that allowed electrochemical measurements to be made during polish. Polish rates were determined from the same samples used in the cell. Alumina-based polish slurries containing potassium iodate, ferric nitrate, or ammonium persulfate were used. DC polarization experiments show no evidence of passive film formation on the tungsten duiing polish. Tungsten oxidation rates measured during polish account for removal rates that are ito 2 orders of magnitude below the measured polish rate. Values of the charge-transfer resistance (measured by ac impedance spectroscopy) during polish are ito 2 orders of magnitude higher than expected from the polish rate, thus corroborating the dc-based data. Polish rates under potentiostatic conditions were also measured. The current required to maintain the metal anodic of the open-circuit potential is well below the current expected from measured polish rates, assuming complete oxidation of the tungsten. The polish rate during cathodic potentiostatic conditions (-0.5 V with regard to the open-circuit potential) was similar to the polish rate at open circuit. We conclude that the formation of a blanket passive layer does not significantly contribute to tungsten removal during CMP. InfroductionChemical mechanical polishing (CMP) is the most effective and now the predominant method for the removal of excess tungsten (W) deposited by nonselective chemical vapor deposition (CVD) for the formation of contacts and vias used in integrated circuit (IC) multilevel interconnects. Figure la depicts the CVD tungsten film and patterned oxide prior to polish. Figure ib depicts the same surface after CMP. The majority of W CMP research to date has focused on empirical cause and effect relationships in which process variables, such as slurry composition, pad type, applied pressures, and platen and carrier speeds, are empirically modeled. These empirical models allow for adequate manufacturing process control; however, they provide little information on the fundamental * Electrochemical Society Active Member.Oxide Fig. 1. The result of a blanket tungsten deposition is shown in the top sketch. The tungsten has been deposited in the vias opened in the inter-level dielectric, but is also present as a blanket film on the surface. The excess tungsten has been polished back to the interlevel dielectric in the bottom sketch.tungsten removal mechanisms that occur during polish. Clearer understanding of the removal mechanism(s) will benefit next-generation designs of slurries and pads and will improve W CMP manufacturing processes.In this work we investigate the role of tungsten oxidation and passive film formation in the mechanism of tungsten removal during CMP, by comparing measurements of the electrochemical behavior of the CVD tungsten film with tungsten removal rates obta...

No abstract

To date, chemical mechanical planarization ͑CMP͒ models have relied heavily on parameters such as pressure, velocity, slurry, and pad properties to describe material removal rates. One key parameter, temperature, which can impact both the mechanical and chemical facets of the CMP process, is often neglected. Using a modified definition of the generalized Preston's equation with the inclusion of an Arrhenius relationship, thermally controlled polishing experiments are shown to quantify the contribution of temperature to the relative magnitude of the thermally dependent and thermally independent aspects of copper and interlayer dielectric ͑ILD͒ CMP. The newly defined Preston's equation includes a modified definition of the activation energy parameter contained in the Arrhenius portion, the combined activation energy, which describes all events ͑chemical or mechanical͒ that are impacted by temperature during CMP. Studies indicate that for every consumable set combination ͑i.e., slurry and polishing pad͒ a characteristic combined Arrhenius activation energy can be calculated for each substrate material being polished.