In Situ Investigation of Slurry Flow Fields during CMP N. Mueller, a C. Rogers, a V. P. Manno, a R. White, a,z and M. Moinpour b a Tufts University, Medford, Massachusetts 02155, USA b Intel Corporation, Santa Clara, California 95052, USA The objective of this work is to obtain in situ slurry fluid flow data during the chemical mechanical planarization CMP process. Slurry flow affects the material removal processes, the creation of defects, and consumable use during CMP, and therefore impacts the cost and quality of polishing. Wafer-scale flow visualization using seeded slurry was accomplished for a variable applied load 0.3–2.5 psi downforce, wafer rotation speed 0 and 33 rpm, slurry injection locations, and various pad types flat, XY grooved, and AC grooved. In situ pad conditioning was employed in all experiments. The data indicated complex slurry flow fields on the pad surface in the wafer vicinity, which are influenced by slurry injection point, pad grooving, downforce, and wafer/conditioner rotation. Injection location and pad type were shown to have the strongest impact on the variation in the fluid flow fields obtained. © 2009 The Electrochemical Society. DOI: 10.1149/1.3223562 All rights reserved. Manuscript submitted December 29, 2008; revised manuscript received July 31, 2009. Published October 7, 2009. The semiconductor industry relies heavily on chemical mechani- cal planarization CMP to create planar surfaces on silicon sub- strates during the integrated circuit IC manufacturing process. As the IC feature size continues to shrink, the need to create surfaces with micrometer-level global planarity is of utmost importance. 1-4 CMP is a complex, multimechanism process 1 in which the material removal mechanisms are not well characterized, 5 leading to a high reliance on indirect empirical data. 6 In situ data acquisition during CMP is difficult due to concurrent processes, complex geometries, and the combined opaque nature of the pads, wafers, and slurry during polish. In most cases, CMP data are gathered ex situ or ob- tained from oversimplified systems that may not fully represent in- dustrially relevant polishing conditions. Mechanistic models have been developed to explain the phenomena, 3,6-9 but there are only limited empirical data available to test these models. The polishing parameters and their effect on pol- ish quality, which is influenced by fluid transport and polishing com- ponents, must be better understood to advance the state of the art in CMP. 1,3,9 The slurry velocity flow fields at the wafer–slurry–pad interface have not yet been fully characterized and a repeatable and reliable method to measure these polishing indicators is lacking. 10 We present data from two flow characterization approaches in this paper: qualitative flow visualization and slurry velocity measure- ment using particle image velocimetry PIV. Experimental A photograph of the polisher utilized in this work is presented in Fig. 1. The primary component is a modified half-scale Stuers RotoPol-31 CMP industrial polisher that is surrounded by an 80 /20 aluminum frame. The polishing system sits atop a 136 kg solid steel isolation table. Optically clear BK-7 glass wafers 75 mm in diam- eter  3 in. and 12.5 mm thick  0.5 in. were rotated and loaded with downforce through the centrally located aluminum shaft driven by a 0.2 hp Dayton motor. A transparent glass wafer was used as a surrogate for a silicon wafer because it has similar polishing char- acteristics as the semiconductor wafer, but allows optical data ac- quisition through the wafer to the underwafer region. The wafer was attached to the shaft via a pin connection that allowed the wafer to gimbal. 11 A selection of polyurethane polishing pads 300 mm in diameter  12 in. was used along with Cabot Microelectronic’s Cabo-sil SC-1 oxide polishing slurry with 100 nm fumed silica abra- sive particles. 4 Slurry flow visualization was achieved using tracer particles that were added to the slurry to allow optical tracking of the fluid mo- tion. Pepper particles were used as tracers because they are inexpen- sive, easy to work with, environmentally benign, and can be imaged under ambient light. To accurately follow the flow, tracer particles need to be small and light enough to travel with the fluid and not perturb the flow. This is usually quantified by the Stokes number of the particle, which should be less than 0.1. 12 However, this is only relevant for neutrally buoyant particles immersed in the flow. The pepper particles used in this study were small enough and had a large enough surface area to be trapped by surface tension at the surface of the flow. We could not, therefore, completely rely on a Stokes number evaluation. However, qualitatively the particles travel with the flow; bubbles and other particulates of various sizes close to the surface of the flow traveled at the same rate as the pepper. Therefore, the pepper behavior was thought to be a good estimate for the fluid behavior. Video recordings taken at 30 frames /s using a Canon Vixia HG10 high definition video camera were analyzed manually to yield a qualitative description of the slurry flow. Individual flow visual- ization data acquisition runs were limited to less than 1 min to avoid tracer particle clogging in the pad surface, especially in the grooves. Clogging the pad grooves was avoided because it distorts data in- terpretation and modifies the flow field. Figure 2 is a sample image in which clogged grooves are apparent. Several different experimental conditions, including wafer speed, downforce, injection location, and pad type, were examined and compared. The pads used included Freudenberg FX9 flat pads, Freudenberg FX9 XY grooved pads, and Cabot’s DC100 ac grooved pads. These pads contained no pattern, a cross-hatched pattern, and a concentric circle pattern, respectively. The specific experimental z E-mail: r.white@tufts.edu Force Table (under RotoPol-31) Wafer Steel Platen with 12” Polishing Pad RotoPol-31 Motor Aluminum Axle Figure 1. Color online The CMP experimental rig used to gather the in situ data. In this photograph the conditioner is not shown so as not to obscure the view, but it is always present during experiments. Journal of The Electrochemical Society, 156 12 H908-H912 2009 0013-4651/2009/15612/H908/5/$25.00 © The Electrochemical Society H908 Downloaded 08 Oct 2009 to 130.64.80.153. Redistribution subject to ECS license or copyright; see http://www.ecsdl.org/terms_use.jsp