Chemical mechanical planarization ͑CMP͒ is a process widely used for the manufacture of silicon integrated circuits. In this work, we measured the thickness of the slurry film between the wafer and the pad during polish while simultaneously measuring the frictional drag. All experiments are performed on a 1:2 scale laboratory tabletop rotary polisher with variable pad speed and wafer downforce control. Dual emission laser-induced fluorescence techniques optically measured the slurry film thickness through a dual-camera imaging system. The resulting data are discussed for wafers polished with a 3.1 wt % abrasive concentration slurry solution on Freudenberg's FX-9 polishing pads. It was found that the degree of surface curvature of the wafer substrate significantly influences the slurry film thickness and wafer drag, and therefore, the polish. The convex wafer shows the expected behavior of increased downforce reduces the slurry film thickness and increases the coefficient of friction. Further, as the pad speeds up, the slurry thickness increases and the friction decreases. The concave wafer shows no change in slurry film thickness and a decrease in the frictional coefficient with increasing downforce. Both the film thickness and frictional coefficient appear to decrease slightly with increasing pad speed. This difference between the two wafer shapes reflects the different fluid mechanics in each case.The chemical mechanical planarization ͑CMP͒ process is generally accepted as the technique used for the manufacture of integrated circuits ͑ICs͒, yet the understanding of the fundamental mechanisms involved in this process is limited. In recent years fundamental research has been done to both experimentally understand polishing characteristics and analytically model the processes involved. 1-3 To model the removal rate, one must first understand how the thin ͑ϳ20 m͒ film of slurry between the wafer and polishing pad behaves. Thin-film thicknesses also imply that the asperities in the pad reach through the slurry and rub on the pad. Wafer nonuniformities in turn lead to defects during production. Slurry film thickness during polishing is largely determined by how the hydrodynamic pressure of the fluid and the pad asperities support the wafer downforce. Levert, Mess, and Tichy have found that during polishing there is a vacuum created underneath a static wafer, pulling the wafer into the pad. 4-6 Sundararajan has found positive fluid pressure developing in the gap between the wafer and polishing pad. 7 In our experiments we witness both cases: positive pressure with a convex wafer ͑Fig. 1a͒ and negative pressure ͑vacuum͒ with a concave wafer ͑Fig. 1b͒. Frictional measurements during polishing help characterize the interaction between the wafer, abrasive ͑slurry͒, and pad. 8 This in turn helps identify the removal rate and its dependence on the lubrication regime. 3,[9][10][11][12] The separation distance between the polishing pad and the wafer substrate during CMP can be characterized by the thickness of the slurry laye...

The fluid film thickness and drag during chemical-mechanical polishing are largely dependent on the shape of the wafer polished. In this study we use dual emission laser induced fluorescence to measure the film thickness and a strain gage, mounted on the polishing table, to measure the friction force between the wafer and the pad. All measurements are taken during real polishing processes. The trends indicate that with a convex wafer in contact with the polishing pad, the slurry layer increases with increasing platen speed and decreases with increasing downforce. The drag force decreases with increasing platen speed and increases with increasing downforce. These similarities are observed for both in-situ and ex-situ conditioning. However, these trends are significantly different for the case of a concave wafer in contact with the polishing pad. During ex-situ conditioning the trends are similar as with a convex wafer. However, in-situ conditioning decreases the slurry film layer with increasing platen speed, and increases it with increasing downforce in the case of the concave wafer. The drag force increases with increasing platen speed as well as increasing downforce. Since we are continually polishing, the wafer shape does change over the course of each experiment causing a larger error in repeatability than the measurement error itself. Different wafers are used throughout the experiment and the results are consistent with the variance of the wafer shape. Local pressure measurements on the rotating wafer help explain the variances in fluid film thickness and friction during polishing.

No abstract

Largely due to cooling power consumption needed for their thermal management, data centers consume over 2% of the electricity produced in the United States. A fundamental understanding of data center energy management is both environmentally and economically important. This paper develops a model of data center energy utilization for thermal management based on thermo-fluid first principles. The model is applied to a single rack system that can be used as an experimental platform for its validation.Models are developed for conventional air-cooling and hybrid cooling, where microprocessors are liquid cooled and other components are air cooled. The air-cooling model considers a single chiller with a variable speed compressor and an intermediate air-to-water heat exchanger. The hybrid model considers cold plates rather than air-cooled heat sinks on the microprocessors and includes an additional chiller. Compiled results quantify the power consumption of the two cooling schemes over ranges of key variables, such as ambient temperature. For example, the air-cooling model shows a 10% reduction in energy consumption for thermal management when the chiller setpoint is raised by 5°C, and the water flow rate pumped through the air-to-liquid heat exchanger is increased to accommodate this change. Hybrid model results show that energy consumption decreases as much as 40% over the parametric ranges investigated relative to air-cooling. The models can be utilized for future system design and trade-off studies. These models do not include variations in air flow rates or power consumption by the server as part of the sensitivity study.

It is becoming increasingly clear that understanding the small scale polishing mechanisms operating during CMP requires knowledge of the nature of the pad-wafer contact. Dual Emission Laser Induced Fluorescence (DELIF) can be used to study the fluid layer profile between the polishing pad and the wafer during CMP. Interactions between the polishing pad surface and the wafer can then be deduced from the fluid layer profile. We present a technique and some preliminary data for instantaneous measurement of in-situ pad-wafer contact, defined as the point at which the fluid film thickness goes to zero, using DELIF. The imaging area is 1.30mmx1.74 mm with a resolution of 2.

No abstract

The objective of this work is to obtain in situ slurry fluid flow data during the chemical mechanical planarization ͑CMP͒ 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 downforce͒, wafer rotation speed ͑0 and 33 rpm͒, slurry injection locations, and various pad types ͑flat, XY grooved, and AC grooved͒. 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.The semiconductor industry relies heavily on chemical mechanical planarization ͑CMP͒ to create planar surfaces on silicon substrates during the integrated circuit ͑IC͒ 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 obtained from oversimplified systems that may not fully represent industrially 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 polish quality, which is influenced by fluid transport and polishing components, 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 measurement using particle image velocimetry ͑PIV͒. ExperimentalA 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 diameter ͑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 chara...

In this paper we document in-situ fluid film pressure and interfacial friction measurements during Chemical Mechanical Planarization (CMP) over a range of applied loads and relative pad/wafer velocities. The slurry film pressure beneath contoured test wafers is measured using a novel experimental setup that enables dynamic data collection. The friction data have a repeatability of approximately 10%.The uncertainly of the pressure measurements and the computed down forces are ±20.7 kPa (±3 psi) and 20%, respectively. The data indicate that wafer shape, specifically global curvature, is a significant factor in determining the lubrication regime during CMP. Full hydrodynamic lubrication, in which the slurry fluid film supports the entire applied load, was not realized for CMP of either concave (center high) or convex (center low) wafers. The data for concave wafers show that -6% to 37% of the applied load is supported by the slurry film, where the negative sign indicates suction conditions that were obtained at the lowest applied load condition. CMP of convex wafers is found to operate closer to full hydrodynamic lubrication, with the fluid layer supporting 36% to 64% of the applied downforce. In all cases, the measured friction coefficient decreased as the support of the fluid layer increased (higher positive pressures). CMP of concave wafers is more sensitive to changes in applied downforce, while the convex wafer type was most affected by changes in the wafer/pad rotation speed, which in turn determines effective slurry film velocity beneath the wafer. Overall, the CMP conditions seen in these scaled experiments operate primarily in the partial lubrication regime shifting closer to hydrodynamic lubrication for convex wafers at the high load, high speed conditions.