Scale-Dependent Nanomechanical Properties of CMP Polishing Pads

Chemical Mechanical polishing, or CMP, has four components in the polishing interface; namely the polishing pad, the polishing fluid, the abrasives in the fluid, and the surface being polished. Understanding the way that load is transferred between these components is key to understanding and predicting wafer polishing. Numerous models of CMP and the different load sharing arrangements which occur between these four components have been proposed. These include models by Luo and Dornfield (1) and Fu, Chandra and colleagues (2). Other models focus on just one or two loading arrangements, such as `hydroplaning' when all load is transmitted by the liquid in the case of Runnels and Eyman (3). In this work, we consider four possible combinations of load transfer from pad to wafer in CMP. These combinations are listed below in Table 1 and are illustrated schematically in Figure 1. It is important to note the role of scale in this process; the illustrations in Figure 1 are drawn to scale from actual profilometer scans obtained over a length of 5 with particles of 150 nm in diameter. Table 1 - Load Transfer Combinations Load Regime Illustration Wafer - Fluid - Pad Figure la Wafer - Particles/Fluid - Pad Figure lb Wafer - Particles - Pad Figure lc Wafer-Pad Figure ld The interest in load-sharing between the system components is because some configurations are anticipated to achieve higher polishing rates than others. For instance, in hydroplaning, or hydrodynamic lubrication, the lateral force exerted on the polished surface by the particles can be no higher than the fluid shear force developed in the gap between the pad and wafer. This is not expected to generate high rates of material removal from the surface (4). For contact that is largely between the pad and the wafer, such as the regime depicted in Figure ld, the lateral load is transmitted to the wafer by the pad and is hence limited by the pad's low rigidity. This regime is not expected to generate high material removal rates either. The highest levels of material removal are expected from regimes where the mechanical load exerted on the system is transmitted by the particles, in combination with either the fluid or the pad. Under what circumstances does the polishing regime change from one to another? What must occur in order to initiate or terminate contact between the system components? This is an ongoing question in the field of tribology, and has traditionally been viewed in terms of the pressure and velocity applied to the system, the thickness of the fluid film between the two surfaces, and the roughness of those surfaces (4). In this paper, we propose a new mechanism ? that the load on the particles may be sufficient to totally embed them in the polishing pad, at which point they are no longer able to transmit load and polish the wafer surface. This is illustrated schematically in Figure 1(d).