Vincent P. Manno

Measurements of Slurry Film Thickness and Wafer Drag during CMP

dIntel Corporation, Santa Clara, California 95052-8119, USA Chemical mechanical planarization ~CMP! 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.

Liquid crystal imaging for temperature measurement of electronic devices

The liquid crystal imaging (LCI) technique has been shown to be a viable method for part level temperature measurement. The steps necessary to develop this capability and its sensitivity to various parameters are discussed and highlighted. Comparison of LCI and point measurement showed that high accuracy can be achieved with this method. The spatial resolution which plays a critical role in this measurement was shown to be attainable for part level measurements. Even higher spatial resolution can be attained with modifications to the optics and image processing software. Calibration procedures are also described.<<ETX>>

Characterization of laminar jet impingement cooling in portable computer applications

A thermal characterization study of laminar air jet impingement cooling of electronic components within a geometry representative of the CPU compartment of a typical portable computer is reported. A finite control volume technique was used to solve for the velocity and temperature fields. Convection, conduction and radiation effects were included in the simulations. The range of jet Reynolds numbers considered was 63 to 1500; the applied compartment heat load ranged from 5-15 W. Radiation effects were significant over the range of Reynolds numbers and heat loads considered, while the effect of natural convection was only noticeable for configurations when the ratio Gr/Re/sup 2/ exceeded 5. The predicted importance of Re rather than jet size was confirmed with test data. Proof of concept was demonstrated with a numerical model representative of a full laptop computer. Both simulations and lab tests showed that low flow rate JI cooling schemes can provide cooling comparable to a high volume flow rate configuration, while using only a fraction of the air flow. Further, under the conservative assumption of steady state, fully powered components, a hybrid cooling scheme utilizing a heat pipe and laminar JI was capable of cooling the processor chip within 11 C of the vendor specified maximum temperature for a system with a total power dissipation of over 21 W.

A method for reduction of numerical diffusion in the donor cell treatment of convection

Abstract This article is concerned with the donor cell treatment of convection in numerical simulations of convective-diffusive flow. The two sources of numerical diffusion in this treatment, truncation error and crossflow diffusion, are explained and quantified. Truncation error occurs due to the use of approximate profile assumptions while crossflow diffusion arises due to cell-wise homogenization of the convected quantity in multidimensional problems. Crossflow diffusion is the dominant source of error in many cases. A corrective scheme is introduced in this article which compensates for the effect of crossflow diffusion by reducing the effective anisotropic diffusion coefficient used in the diffusion portion of the simulation. Relationships are developed quantifying the crossflow diffusional error and requirements for explicit numerical stability when the error correction technique is employed. The magnitudes of the diffusional error and the improvements realized using the corrective scheme are demonstrated through computational examples.

In Situ Investigation of Slurry Flow Fields during CMP

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.

In Situ Measurement of Pressure and Friction during CMP of Contoured Wafers

In situ fluid film pressure and interfacial friction measurements during chemical mechanical planarization (CMP) are reported over a range of applied loads (27.6-41.4 kPA or 4-6 psi) and relative pad/wafer velocities (0.35-0.58 m/s). The slurry film pressure beneath contoured test wafers was measured using a novel experimental setup that enables dynamic data collection. The friction data have a repeatability of ∼10%. The uncertainty of the pressure measurements and the computed down forces were ′2.1 kPa(′0.3 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 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.

Energy savings achievable through liquid cooling: A rack level case study

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.

Determining Pad-Wafer Contact using Dual Emission Laser Induced Fluorescence

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 2.5 µm/pixel. At this magnification, some regions imaged contain contact, whereas others do not. For the contact regions discussed in this paper, contact percentage varies from 0.07% to 0.27% using a Cabot Microelectronics D100 polishing pad. The asperity contact area increases with applied load, which was varied from 0.28 to 3.1 psi.

The Effect of Wafer Shape on Slurry Film Thickness and Friction Coefficients in Chemical Mechanical Planarization

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.

A NOTE: A MODEL OF STEAM INJECTOR PERFORMANCE

A simple, mathematical model of the steady state performance of a condensing steam injector is developed based on one-dimensional conservation equations. The model involves two distinct flow regions–a stratified annular flow and a downstream dispersed flow. The former regime is modeled using a two fluid, non-equilibrium formulation while the latter is described by a homogeneous non-equilibrium model. The flow regime transition point is not calculated from first principles but must be described empirically. The model is applied to the simulation of experimental data. Good accuracy is achieved especially in cases where a significant condensation potential exists. The major modeling deficiencies are the lack of a predictive flow transition and the need for empirical closure relations for interphasial exchanges and shock calculations. More fundamental experiments are required before qualitative analytical model improvements are possible.