SHEAR SENSORS AND USES THEREOF

Abstract

The present invention relates to shear sensors. In particular, the present invention relates to deformable sensors for measurement of shear forces during chemical mechanical polishing.

Description

FIELD OF THE INVENTION

The present invention relates to shear sensors. In particular, the present invention relates to deformable sensors for measurement of shear forces during chemical mechanical polishing.

BACKGROUND OF THE INVENTION

The semiconductor industry has worked its way into almost every facet of contemporary life. In the United States, the semiconductor industry leads all other industries in the total value of exports. In 2005, the industry sent almost $43 billion in product overseas. It's a key source for innovative technology as well, submitting more United States patents than any other industry and providing more than $100 million annually to support research at domestic universities. Furthermore, Moore's Law stating that the number of transistors possible on an integrated circuit (IC) doubles about every 2 years is constantly bolstered by the semiconductor industry's progress and computers are only becoming more efficient, ubiquitous, and important in modern society.

In this all important industry, chemical mechanical polishing (CMP) is a globally used tool critical to the processing of the IC and micro electromechanical system (MEMS). CMP involves the planarization of semiconductor wafer surfaces using a polishing pad and a chemically active and abrasive slurry. The polishing pad and wafer are pressed together and rotated around separate axes to precisely polish irregular wafer surfaces and achieve both global and local planarization. These irregularities are created by deposition of the multiple layers of conductors and dielectric insulators necessary in IC and MEMS fabrication.

One example of the need for CMP is a damascene process. This process is utilized to obtain a desired pattern of copper wires in an insulating layer of SiO₂. This vital process now has an annual economic impact over $1 billion and as IC feature sizes continue to shrink, the importance of planarizing substrates both globally and locally will only grow (Higgs et al., Journal of the Electrochemical Society, vol. 152, 2005, herein incorporated by reference in its entirety). To decrease feature sizes in transistors, the aperture of the lens used during lithography must be increased. Since depth of field is inversely proportional to the square of the numerical aperture, as the aperture is increased to gain the higher resolutions necessary for smaller feature sizes, depth of field decreases (Schellenberg, IEEE Spectrum, vol. 40, pp. 34-39, 2003; Dendukuri et al., Nature Materials, vol. 5, pp. 365-369, 2006, each of which is herein incorporated by reference in its entirety). Depth of field (also dependent on the lithography tool, process, pattern size and pattern geometry) can quickly move into the low nanometer scale, showing the necessity of CMP as a global planarization tool as feature sizes reduce further.

Individual CMP processes are well established within certain enterprises. However, the ability to change CMP parameters and predict the effects this will have on material removal rates is not yet realized (Evans et al., CIRP Annals—Manufacturing Technology, vol. 52, pp. 611-633, 2003, herein incorporated by reference in its entirety). The CMP process is widely used and removal rates for certain systems have been characterized experimentally in terms of many of the polishing parameters, but a comprehensive model involving all variables and their effects on material removal rates and non-uniformity remains elusive (Paul, Journal of the Electrochemical Society, vol. 148, pp. G355-G358, 2001, herein incorporated by reference in its entirety). The development and validation of a total CMP model is hindered by lack of knowledge of in situ shear forces present at the micro-scale (Cook, Journal of Non Crystalline Solids, vol. 120, pp. 152-171, 1990, herein incorporated by reference in its entirety). What is needed are improved methods for measuring forces during CMP.

SUMMARY OF THE INVENTION

The present invention relates to shear sensors. In particular, the present invention relates to deformable sensors for measurement of shear forces during chemical mechanical polishing.

In some embodiments, the present invention provides devices, systems and methods for measuring shear forces during chemical mechanical polishing. For example, in some embodiments, the present invention provides a system, comprising: at least one deformable sensor (e.g., an array of sensors), wherein the sensor is deformed under shear stress arising during chemical mechanical polishing; an apparatus for chemical mechanical processing (CMP), wherein the apparatus is in active communication with the sensor; and a detection device for detection of deformation of the sensor. In some embodiments, the detection device comprises a light source (e.g., a fiber optic light source), a microscope, and a CCD camera (e.g. a high speed camera), although a variety of different detection systems can be used.

In some embodiments, the sensor is composed of poly-dimethyl-siloxane (PDMS). In some embodiments, the sensor comprises a plurality of posts. In some embodiments, the posts are coated with a metal (e.g., chromium). In some embodiments, the post has diameter of at least 1 μm (e.g., approximately 30 μm to 100 μm, although other sizes may be used). In some embodiments, the post has a height:diameter ratio of less than approximately 4:1 (e.g., less than 3:1 or less than 2:1), although other ratios may be used. In some embodiments, the sensor further comprises molded markings. In some embodiments, the sensors are affixed to a polishing wafer. In some embodiments, the polishing wafer is affixed to the CMP apparatus using an acrylic or aluminum mating plate. In some embodiments, the acrylic mating plate comprises a plurality of viewing windows.

In other embodiments, the sensor is composed of silicon or a thick metal film (e.g., copper or nickel). In some embodiments, the sensor is circular. In some embodiments, the sensor is attached to a chemical mechanical processing apparatus using a flexible attachment configured to deform under shear stress. In some embodiments, the attachments move under shear stress.

The present invention further comprises methods of using the described devices and systems. In some embodiments, the method comprises the steps of contacting at least one deformable sensor, wherein the sensor is deformed under shear stress arising during chemical mechanical polishing, with an apparatus for chemical mechanical processing (CMP) under conditions such that the sensor is deformed. In some embodiments, the method further comprises the step of calculating shear forces on the sensor based on the deformation of the sensor (e.g., using a computer processor and computer software).

DESCRIPTION OF THE FIGURES

FIG. 1 shows a diagram of PDMS posts deflecting due to shear forces. The basic unit of the shear stress sensor is the recessed micro post (right image). An array of posts with as-designed height 100 microns, and varying diameter, d, are shown.

FIG. 2 shows a C-RICM image from showing asperity contact spacing for static contact between an IC 1000 polishing pad and a sapphire cover slip. The image of a 100 μm sensing post is superimposed to show the ability of the sensor to resolve individual asperity contacts. This image is taken from C. L. Elmufdi and G. P. Muldowney, “A novel optical technique to measure pad-wafer contact area in chemical mechanical planarization,” presented at Materials Research Society Symposium Proceedings, 2006.

FIG. 3 shows an exemplary design of a single sensor array of some embodiments of the present invention. The size of the array is approximately 2 mm×1 mm. Spacing between post arrays and individual posts within an array is 10 mm and 100 μm, respectively.

FIG. 4 shows stylus profilometer scans of a 120 μm long, 100 μm diameter post using downforces of 10, 30, and 50 μN (left) and five profilometer scans of a PDMS post that show repeatable deflection results and no plastic deformation (right).

FIG. 5 shows a diagram of the sensor wafer location relative to the slurry introduction point. The sensor wafer replaces the standard silicon wafer

FIG. 6 shows a diagram of the optical setup for determining shear stresses in situ.

FIG. 7 shows a diagram of the ⅛″ thick aluminum mounting plate and acrylic linkage which connect the sensor to the CMP axle. The linkage is removable to allow the optical setup views of all 4 windows in the mounting plate.

FIG. 8 shows microscope images of 2 layer SU-8-100 and SiO2 micromold. Left: top-down light microscope image. The inset shows the tick marks fabricated on the post and around the well edges. Right: An angled perspective light microscope image shows the three dimensional structure of the thick SU-8 layer.

FIG. 9 shows microscope images of the PDMS sensor array. Left: top-down light microscope image. The inset shows the tick marks and indicator numbers. Right: An angled perspective light microscope image shows the three dimensional structure of posts in wells.

FIG. 10 shows force-deflection curves for the smallest and largest diameter posts calibrated.

FIG. 11 shows a micrograph (using the optical setup detailed) of the final sensor in the polishing rig, obtained using the optical setup. The image is taken under static conditions with no applied load and no slurry present.

FIG. 12 shows in situ CMP micrographs of sensor posts. Left: 40 μm and 50 μm diameter posts being polished with a downforce of 10 lb and a pad rotation rate of 30 rpm. Right: 50 μm diameter post being polished with a downforce of 15 lb and a pad rotation rate of 30 rpm.

FIG. 13 shows post deflections charted for each post analyzed relative to their post wells.

FIG. 14 shows post deflections observed during polishing of 50 μm through 100 μm diameter posts at a pad rotation speed of 30 rpm and downforces of 5 lb, 10 lb, or 15 lb.

FIG. 15 shows post deflections observed during polishing of 50 μm through 100 μm diameter posts at a pad rotation speed of 60 rpm and downforces of 5 lb or 15 lb.

FIG. 16 shows modified force-deflection curves for 40 μm and 100 μm diameter posts.

FIG. 17 shows shear forces observed by 50 μm through 100 μm diameter posts during polishing at a pad rotation speed of 30 rpm and down forces of 5 lb, 10 lb, or 15 lb.

FIG. 18 shows shear forces observed by 50 μm through 100 μm diameter posts during polishing at a pad rotation speed of 30 rpm and down forces of 5 lb, 10 lb, or 15 lb.

FIG. 19 shows a schematic of a floating element sensor.

FIG. 20 shows a schematic of a floating element shear force sensor fabrication.

FIG. 21 shows a FEA model mesh for a 100 μm tall, 100 μm diameter post.

FIG. 22 shows a diagram of an alternative fabrication process for floating element sensors.

FIG. 23 shows calibration of PDMS posts. (a) Compared to finite element predictions. (b) Showing the change in stiffness of the post before treatment with O₂ plasma and metallization with Cr to after treatment and polishing.

FIG. 24 shows microscope pictures taken in situ during polishing with the high-speed microscopy setup. The two figures show a deflected (left) and undeflected (right) post reacting to pad-wafer interaction forces.

FIG. 25 shows a polar plot showing that the forces measured using the PDMS post shear sensor align with the direction of pad travel.

FIG. 26 shows an example of a time trace of the magnitude of the measured shear force as a function of time as measured with the PDMS post-like shear sensor.

FIG. 27 shows the results of 5 different conditions measured using the PDMS post-like shear sensors. The RMS force is plotted for each condition of variable down force and variable pad-wafer velocity.

FIG. 28 shows microscope pictures of the floating element sensors at intermediate steps in their fabrication. (Left) SEM picture showing a successful silicon etch early in process development. (Right) A light microscope picture showing a fabricated floating element at the second to last step in the process.

DEFINITIONS

As used herein, the term “wafer” generally refers to substrates formed of a semiconductor or non-semiconductor material. Examples of such a semiconductor or non-semiconductor material include, but are not limited to, monocrystalline silicon, gallium arsenide, and indium phosphide. Such substrates may be commonly found and/or processed in semiconductor fabrication facilities. A wafer may include one or more layers formed upon a substrate. For example, such layers may include, but are not limited to, a resist, a dielectric material, and a conductive material. Many different types of such layers are known in the art, and the term wafer as used herein is intended to encompass a wafer including all types of such layers.

DETAILED DESCRIPTION

The present invention relates to shear sensors for micromachining. In particular, the present invention relates to deformable sensors for measurement of shear forces during chemical mechanical polishing.

I. CMP Modeling

CMP modeling provides insight into the basic interactions that control polishing. However, the current state of the CMP model does not completely link input variables (pad and slurry properties and polishing parameters) to process variables (frictional forces, operating temperatures) and output variables (material removal rate and polishing defects) (Cook, supra). The most basic model used to describe CMP material removal rate comes from glass polish research—Preston's equation (Paul, Journal of the Electrochemical Society, vol. 148, pp. G355-G358, 2001; Cook, supra; Srinivasa-Murthy et al., Thin Solid Films, vol. 308-309, pp. 533-537, 1997, each of which is herein incorporated by reference in its entirety):

r=K_p·P·v  (1)

where r is the removal rate of the polished surface, P is the pressure applied to the wafer, v is the relative velocity between the polishing pad and the substrate, and K_p is a proportionality constant called the Preston coefficient. This coefficient has units of area over force. The removal rate in glass CMP is roughly linear with pressure and pad rotational speed. This equation adequately models the general chemical and mechanical processes present during CMP. However, it does not describe non-uniformities (NU) inherent in the polishing process such as dishing and inconsistent polishing rates across wafers. The Preston coefficient takes into account the chemical properties of the slurry and substrate surface as well as mechanical interactions at the substrate surface (Cook, Journal of Non-Crystalline Solids, vol. 120, pp. 152-171, 1990; Srinivasa-Murthy et al., Thin Solid Films, vol. 308-309, pp. 533-537, 1997, each of which is herein incorporated by reference in its entirety) but does not specifically detail relationships between polishing parameters. Furthermore, the linear relationship between the material removal rate and the pressure and velocity is not present when polishing all materials, specifically metals (Paul, Journal of the Electrochemical Society, vol. 148, pp. G355-G358, 2001 herein incorporated by reference in its entirety). As a result, many researchers have developed more complete models of the polishing process based on Preston's equation as described below.

Most of the models created attempt to describe interactions between the four elements of CMP: the slurry abrasive particles, slurry chemicals, polishing pad and the substrate (wafer). In general, the sequence of events involving these parameters begins with slurry chemicals promoting the creation of a thin reaction film at the surface of the substrate. In the absence of a mechanical abrasive, this reaction film can dissolve into the surrounding slurry. With a polishing pad and abrasives present, the reaction film may be mechanically removed by passing abrasive slurry particles and polishing pad asperities. Chemical dissolution is very slow compared to mechanical removal, yet the timescale of this chemical softening is on the same order as the mechanical abrasion to obtain effective polishing (Paul, supra). The polishing pad is the medium through which applied pressure is transferred to the abrasive particles and the substrate surface.

Although many CMP models agree with the above basic process, they differ on the details of material removal mechanisms. Specifically, the interactions between slurry particles, polishing pad, and wafer are often disputed. Some models attribute the majority of material removal to mechanical abrasion by slurry particles and determine that the material removal rate is indeed proportional to applied pressure and relative velocity between the substrate and pad as Preston's equation predicts (Cook, supra; Lui et al., Journal of the Electrochemical Society, vol. 143, pp. 716-721, 1996, herein incorporated by reference in its entirety). Others say that material removal is largely dependent on erosion enhanced by mechanical indentation of slurry particles rather than mechanical abrasion of slurry particles (Runnels, Journal of the Electrochemical Society, vol. 141, pp. 1900-1904, 1994; Tseng and Wang, Journal of the Electrochemical Society, vol. 144, pp. L15-L17, 1997; Zhang and Busnaina, Electrochemical and Solid-State Letters, vol. 1, pp. 184-187, 1998, each of which is herein incorporated by reference in its entirety). In these models, the material removal rate is non-linearly dependent on pressure and velocity, and fluid stress tensors are sought to detail the material removal at each point across a substrate. Experimental parameters such as pressures, speeds, substrates and slurries are not held constant from model to model which may explain their differences.

The active slurry particles in most CMP models are thought to be embedded in the polishing pad and pressed into the wafer. Slurry particles in the slurry that are not pressed against the wafer are not part of the removal process. Since the polishing pad transmits the force onto the slurry particles, pad topography is also critical to the CMP model. Some researchers assume that the polishing pad consists of asperities with heights following a Gaussian distribution. The contact area between the pad and the substrate, and thus the number of abrasive particles removing material from the substrate, is proportional to pressure. This model agrees with Preston's equation in that material removal rate is linearly related to pressure and velocity. Some modeling of pad topography relates the contact area between pad asperities and substrate using Hertz elastic contact theory so that contact area is nonlinearly dependant on pressure. This model also differentiates between polishing regimes when slurry particles are rolling along the substrate as opposed to dragging across the surface. When the particles are rolling their contribution to material removal is considered to be minimal due to the significant decrease in friction (Zhao and Shi, Electrochemical and Solid-State Letters, vol. 2, pp. 145-147, 1999, herein incorporated by reference in its entirety).

All of the models described above attempt to add depth to the initial Preston's equation and provide insight into the CMP mechanisms. Many of the differences between the models involve determining the correct sources of mechanical forces (pressure and shear force) during polishing. As shown in the models, material removal rate is strongly linked to these forces, imparted on the substrate by slurry particles. Characterizing slurry particle shear and normal forces during CMP sheds light on differences between theories and create a unified polishing model.

Furthermore, investigating local (micro scale) and global (coefficient of friction) shear forces in situ captures stick-slip phenomena and other vibration modes that time averaging techniques or post polishing analysis cannot, further adding depth to CMP models. Other differences between theories arise from attempting to characterize pad topography. Material removal is also dependant on the amount of contact between pad and substrate. Characterizing the number, size, distribution, and frequency of asperities contacting the substrate also aids in the creation of a complete CMP model.

The polishing model which connects all process variables is used to optimize the CMP procedure for the various semiconductor applications as current industrial needs require. Shear forces present during polishing directly influence material removal rate, processing temperature variations, and failure modes. Characterizing these surface forces in situ aids in CMP consumable development and CMP process optimization.

The importance of shear force characterization goes beyond CMP theory as well. The higher operating speeds and smaller feature size of emerging ICs require insulating layers with lower dielectric constants. Integrating these ‘low-k″ dielectrics into the standard semiconductor processes allow for lower capacitance and smaller special requirements in the final IC. Unfortunately, this is accomplished at the expense of mechanical strength as low-k and ultra low-k dielectrics are much weaker than conventional oxide dielectrics (Braun, Semiconductor International, vol. 24, pp. 54-56, 2001; Hosali and Busch, Solid State Technology, vol. 48, pp. 33-36, 2005, each of which is herein incorporated by reference in its entirety). The Young's modulus of many low-k dielectrics range from about 3 GPa up to about 12 GPa (Lee and Kumar, Solid State Technology, vol. 47, pp. 69-74, 2004, herein incorporated by reference in its entirety) whereas the modulus of standard SiO2 is around ten times greater at 70 GPa. The low-k property of these materials is obtained by introducing porosity between 20% and 50% of the dielectric (Hosali and Busch, supra; Braun, Semiconductor International, vol. 26, pp. 52-56, 2003, herein incorporated by reference in its entirety). CMP mechanical forces can physically move copper connection lines back and forth during polishing which damages the interconnects and can delaminate low-k dielectric insulating thin films (Braum 2001, supra). Characterizing shear forces present during CMP is exceptionally valuable in the semiconductor industry's pursuit of a 2.5 low-k value, the required value for the 45 nm node as predicted by the International Technology Roadmap for Semiconductors (ITRS) in 2004. Lowering shear forces while retaining material removal rates during CMP may be the key to integrating ultra low-k dielectrics into ICs.

Knowledge of shear forces present during CMP is also useful for polishing pad designs. In situ shear force knowledge also indicates the constancy of polishing pads. Pads are designed to deliver stable dynamic mechanical properties throughout their use. The ability to obtain long term in situ shear force measurements from pads aids in pad design optimization and process quality control.

Thus, in some embodiments, the present invention provides systems and methods for improved sensing of shear forces during CMP.

II. Shear Sensors

In some embodiments, the present invention provides shear sensors for the measurement of shear forces during CMP. These forces are caused by interactions between the substrate that is being polished and small asperities on the polishing pad. As described above, the sensor finds use in measuring shear forces in a variety of applications that utilize CMP.

For the most accurate determination of shear forces present during CMP, a shear sensor is preferably located at the interface between the substrate and the polishing pad to allow intimate contact with pad asperities and the reactive, abrasive slurry particles. A sensor for CMP shear force characterization is preferably chemically and physically robust, planar with the substrate surface, and easily fabricated with inexpensive materials and processes in order to support a disposable sensor design. Thus, in some embodiments, the present invention provides simple, inexpensive compositions and methods for measuring shear forces.

A. Force Estimates

In some embodiments, the present invention provides methods of estimating shear forces based on sensor movement. The below description is exemplified using certain experimental parameters utilized during development of embodiments of the present invention. However, the present invention is not limited to the parameters described below. One of skill in that art recognizes that experimental parameters may be altered depending on the desired application.

In some embodiments, post design is based on estimates of shear forces present while polishing a 4″ wafer using an IC 1000 pad with an applied pressure of 1.7 psi. The substrate size is chosen based on fabrication constraints while the applied pressure provides similar normal loads to previous studies. The total shear force on the wafer is thus approximately 30 N. At an applied pressure of 1.7 psi in a static situation, an IC1000 pad will contact only approximately 0.7% of the wafer through pad asperities as shown through previous work by others. For a 4″ wafer, this means a total asperity contact area of approximately 0.09 m². The mean radius of contact between a single asperity and the wafer at this pressure is approximately 5 μm. This is estimated by analyzing Confocal Reflectance Interference Contrast Microscopy (C-RICM) images of asperity-wafer contact and a single asperity contact area of roughly 8×10⁻¹¹ m² can be determined (C. L. Elmufdi and G. P. Muldowney, “A novel optical technique to measure pad-wafer contact area in chemical mechanical planarization,” presented at Materials Research Society Symposium Proceedings, 2006). Based on these estimates, a 4″ diameter sensor under a load of 1.7 psi will be in contact with approximately 700,000 asperities, calculated by dividing the total asperity contact area by the area of a single asperity contact. With a total shear force of approximately 30 N, the mean force delivered by a single asperity is expected to be on the order of 40 μN.

In some exemplary embodiments, the shear sensors are designed to handle a range extending one order of magnitude above and below this mean, from 4-400 μN. Fluid shear forces during CMP due to the slurry-wafer interactions are also considered in the design of the sensor. Fluid velocities in the gap between the wafer and the pad are assumed to be on the same order as the relative velocity between the wafer and pad. For the conditions used in experiments described herein, relative velocities are on the order of 1 m/s. Three dimensional numerical modeling of the fluid pass around the post was conducted using incompressible, viscous, nonlinear Navier-Stokes with 0.5 m/s fluid velocity and rigid sensor structures. Modeling was conducted using the FEA software COMSOL Multiphysics and results indicated that the total fluid force on a 100 μm diameter, 100 μm tall post under polishing conditions is less than 1 μN. This agrees well with other CMP models which sometimes omit fluid shear forces as negligible in force computations (Higgs et al., Journal of the Electrochemical Society, vol. 152, 2005; Levert et al., Tribology Transactions, vol. 41, pp. 593-599, 1998). It is therefore expected that fluid shear forces will not contribute significantly to the deflection of the posts.

B. Deformable Sensors

Embodiments of the present invention provide deformable sensors. Any deformable sensor configuration that permits calculation of shear forces may be used. The invention is exemplified below with post sensors and floating sensors.

1. Post Sensors

In some exemplary embodiments, the microfabricated posts in the shear sensor are made of PDMS, a mechanically and chemically robust silicone elastomer. The use of PDMS for sensor applications has many advantages over other materials. PDMS is very inexpensive, flexible, and easily fabricated. It is stable at high and low temperatures, impermeable to water, and permeable to gases. PDMS is also transparent. A transparent sensor allows deflections of posts at the sensor-pad interface to be obtained optically through the back of the wafer. No interconnects are necessary. FIG. 1 shows a cartoon of how a PDMS post reacts to asperities ‘polishing’ the sensor. The image on the right shows the basic unit of the shear stress sensor presented here: the recessed micro post with bulk PDMS surrounding it. Nearly 98% of the wafer surface is planar PDMS, it is only occasionally broken by the annular well region around sensor posts. This allows the majority of the normal force applied by the polishing pad to be carried by the bulk PDMS, thus not compressing or buckling the sensor posts. Measured forces remain primarily lateral shear forces. In the shown embodiment, the post and well are both circular, enabling shear force detection in both in-plane directions.

In some embodiments, PDMS sensor posts have a diameter of at least 1 micron (e.g. approximately 30 μm to 100 μm), although other sized posts may be utilized. In some embodiments, posts have a height:diameter aspect ratio of less than approximately 4:1, although other ratios may be utilized (e.g., less than 3:1 or less than 2:1). In some embodiments, the post is high enough to measure deflection but low enough not to break under pressure. The optimum dimensions vary by material used and manufacturing method utilized. In some embodiments, the aspect ratio of post sensors is optimized using the detection methods described herein.

In some embodiments, transparent PDMS sensors are coated with a metal film after molding to aid in visualization of the sensor. The present invention is not limited to a particular metal or other coating material. Any material that allows for visualization of sensor deflection during use may be utilized. Examples include, but are not limited to metal thin films of Chromium, Aluminum, Titanium, and Gold.

In some embodiments, PDMS posts are provided in arrays. Each PDMS structure comprises a plurality of posts. In some embodiments, the PDMS structures comprise molded marks on the top side of the sensor posts to aid in visualization of post deflection. In some embodiments, the surfaces of the sensor structures are dyed to aid in visualization of post deflection.

In some embodiments, the PDMS sensor uses a 4″ wafer platform, and is designed for use with a stiff polishing pad, for example, the Rodel IC 1000. Although the below description describes the sensor design and testing explicitly using this pad, a corresponding sensor for use with a soft pad is also designed using similar methods.

In some embodiments, the PDMS sensor structure is bonded to a polishing wafer. The wafer is then integrated into an existing CMP apparatus. The PDMS sensors of the present invention are suitable for use with any number of existing CMP devices. In some embodiments, the sensor wafer replaces the substrate being polished (e.g., forces are detected without polishing of a substrate). In other embodiments, sensors are integrated into a platform that allows for sensing and polishing of a substrate.

In some embodiments, the posts are oriented downward, with the tips in contact with the polishing pad. In some embodiments, to mate the sensor with the CMP axel and secure it during polishing, a mating plate is glued to the back of the wafer which carries the PDMS structures. In some embodiments, the plate has viewing windows cut through it to allow the posts to be observed optically through the back side of the wafer and PDMS.

In some embodiments, the shear sensor detection system further comprises a light source, microscope, and CCD camera or other camera for measuring post deflection. In some embodiments, a high speed camera is utilized. In some embodiments, the system further comprises a computer and computer software for calculating shear forces and displaying or otherwise reporting the results.

2. Floating Sensors

In some embodiments, the present invention provides floating sensors for measuring mechanical forces during chemical-mechanical polishing (CMP). In some embodiments, this sensor consists of a flexible sensor element that is designed to move when subjected to in-plane shear forces. When placed in contact with an operating CMP apparatus, the wafer/asperity forces cause the sensor elements to move in small increments. By measuring these deflections it is possible to calculate the shear forces.

In some embodiments, the sensor is silicon, although other suitable materials may be utilized. For example, in some embodiments, floating sensors are made of copper or other metals. In some embodiments, the sensor is made up of multiple arrays of individual sensor elements. In some embodiments, each array is composed of multiple sensor elements of the same size. The sensor can be used as either a full wafer using all of the arrays simultaneously, or as an individual array. An example of a representative silicon sensor is shown in FIG. 19.

In some embodiments, the individual sensors comprise a circular element supported by flexible attachments known as flexures. During polishing, shear forces act on the sensor elements causing them to pull at the flexures. This results in small lateral movements of the elements themselves. Sensor deflection is dependent upon the size of the sensor element as well as the stiffness of the flexures. In some embodiments, the circular elements vary in size, ranging in the hundreds of microns in diameter, although other sizes are suitable. The flexure dimensions vary in both thickness and width. Both of these dimensions range in the dozens of micron. In some embodiments, in order to prevent the sticking of the floating element to the supporting glass substrate, small raised bumps known as dimples are incorporated into the underside of the circular element.

In some embodiments, before measurements are taken using these sensors, the stiffness of the sensors is verified. Calibration elements are included on each sensor wafer. In some embodiments, the calibration structures are circular elements anchored with flexures identical to the ones attaching the sensor elements. These calibration elements are driven using electrostatic comb drives. Before polishing, each sensor class is calibrated using these elements. Calibration is performed by applying a known voltage across the comb drives. It is possible to calculate the force generated by the comb drives as a function of applied voltage. After applying this voltage, the sensors deflect. By measuring this deflection and comparing it to the applied electrostatic force one can determine the deflection of each sensor class when responding to an applied force.

In some embodiments, calibration is performed using an off-chip capacitively sensed MEMS force sensor and nanopositioning stage to calibrate the floating elements.

3. Additional Sensors

The present invention is not limited to posts and floating sensors. Sensors may be fabricated from a variety of different materials including, but not limited to, a variety of polymers including PDMS, and photosensitive epoxides such as SU-8, single crystal and polycrystalline silicon, and metals such as Nickel. In some embodiments, materials that are not amenable to casting or molding are utilized and other fabrication or machining methods are utilized.

III. Applications

As described above, shear sensors of the embodiments of the present invention find use in measuring shear forces during chemical mechanical processing (CMP). The present invention is not limited to the measurement of shear forces during CMP. Additional applications include, but are not limited to, use in aerospace applications (e.g., characterizing surfaces flows in wind tunnels, aircraft testing, etc) and robotics (e.g., surface sensing testing for gripping, locomotion and other robotic performances).

EXPERIMENTAL

The following examples are provided in order to demonstrate and further illustrate certain preferred embodiments and aspects of the present invention and are not to be construed as limiting the scope thereof.

Example 1

Post Sensor

A. Materials and Methods

Post Design

The size of the post structures was selected to produce measurable deflections over the estimated range of asperity forces. In PDMS casting using an SU-8 mold, it was determined that a preferred aspect ratio of the posts is below approximately 4:1 (height:diameter). The height of the posts was set at 100 μm and Euler-Bernoulli beam theory was used to obtain a range of PDMS post diameters that deflected between 5 μm and 50 μm when reacting to shear forces from 4 μN to 400 μN. The material is modeled as elastic and the modulus used is 750 kPa, the appropriate value for a 10:1 mixture of PDMS base to curing agent. Based on this analysis, the range of diameters for the post array was selected as 30 μm to 100 μm, meeting the sensitivity and dynamic range goals. This size also allows reasonable inplane spatial resolution for the sensor. The spacing between asperity contacts was estimated to be on the order of 80 μm equally spacing the estimated 700,000 asperity contacts over the substrate. FIG. 2 shows a CRIC-M image illustrating asperity spacing in a static situation. The image of a sensing post 100 μm in diameter is superimposed over the image to demonstrate the spatial resolution of the sensor. Based on asperities equally spaced by 80 μm and assuming a relative velocity between pad and substrate of 0.5 m/s, single asperities would pass a sensing post at a frequency of approximately 6.3 kHz, or every 160 μs. The resonant frequency of the sensor posts are approximately 13 kHz-20 kHz for the 40 μm to 100 μm posts, respectively, as calculated using COMSOL Multiphysics computed in vacuo. This gives the sensors a calculated temporal resolution between 50 μs and 75 μs, and thus will respond quickly enough to “see” the passing of individual asperities. As seen in FIG. 2, at times there may be more than a single asperity passing over a sensing post at one time and the true frequency of asperity contacts may span an order of magnitude on either side of this prediction. Time lengths between asperity contacts are expected between 10 μs and 1 ms.

Table 1 shows the posts used in the sensor array, their expected stiffnesses, and the resulting deflections from estimated asperity forces. As the table shows, an asperity force on the low end of the expected range, 4 μN, deflects the smallest post by approximately 50 μm. Similarly, an asperity force on the high end of the expected range, 400 μN, deflects the largest post approximately 50 μm as well.

TABLE 1
Post Diameter(μm)	30	40	50	60	70	80	90	100
Estimated Compliance (μm/μN)	14.25	4.51	1.85	0.89	0.48	0.28	0.18	0.12
Minimum Estimated Deflection	57.0	18.0	7.4	3.6	1.19	1.1	0.7	0.5
for a force of 4 μN (μm)
Minimum Estimated Deflection	5700.8	1803.6	738.6	356.3	192.3	112.7	70.4	46.2
for a force of 400 μN (μm)
**Asperity Forces: 4-400 μN**

FIG. 3 shows the basic design of a single sensor array using the photolithography mask layout editor software L-Edit. Spacing between individual sensing posts is 100 μm. Each array measures approximately 2 mm by 1 mm on a side. The arrays are repeated over the wafer with a spacing of approximately 10 mm.

Sensor Fabrication

Sensors were fabricated through a modified two-layer PDMS micromolding process similar to that described in Sia and Whitesides (Electrophoresis, vol. 24, pp. 3563-3576, 2003). In this process a master consisting of two layers, one silicon dioxide and one SU-8 photoresist, is microfabricated and used as a mold for the PDMS sensor structures. A single master mold can be used to create multiple copies of the full wafer PDMS sensor array.

The starting substrates were 100 mm diameter, <100> oriented, p-type 1-10 Ω·cm (boron) silicon wafers with 1 μm of silicon dioxide, grown by wet thermal oxidation. 1 μm of SPR220-3.0 positive photoresist (Rohm & Haas) was spun on and used as a mask for a buffered hydrofluoric acid etch (Ultratech NP 13:2) of the oxide. This etch creates the indicator numbers and deflection notches in the 1 μm oxide layer. The notches aid in analysis of deflection images while the indicator numbers identify where individual sensing posts are within the post array. Lithography was repeated with a second mask using the negative tone, high aspect ratio epoxy photoresist SU-8-100 (MicroChem Corp.). The SU-8-100 was spun at 500 rpm for 30 seconds and 3000 rpm for 70 seconds, prebaked at 65° C. for 15 minutes and 95° C. for 32 minutes. The exposure dose was 300 mJ/cm2 at I-line. A post exposure bake of 65° C. for 1 minute and 95° C. for 10 minutes was used followed by a 10 minute develop in PM Acetate developer. This results in an approximately 100 μm thick SU-8 structure, which, along with the thin patterned oxide, is used as the master mold for the PDMS.

The master mold was silanized in a rough vacuum desiccator with 2 to 3 drops of the silanizing agent tridecafluoro tetrahydroctyl trichlorosilane for 3 hours. A 10:1 PDMS ratio of base to curing agent was mixed (Dow Corning Sylgard 184), degassed in a rough vacuum chamber, and poured over the silanized master mold. The PDMS was cured in place on the master mold on a hot plate at 60° C. for at least 4 hours. The excess PDMS around the mold was then cut away and the resulting PDMS sensor was peeled from the SU-8. The PDMS structure is bonded to a Pyrex wafer (Corning type 7740 Pyrex glass) by exposing both surfaces to a 200 mT, 25 W oxygen plasma for 30 s, placing the PDMS and glass in contact, and heating on a hotplate at 60° C. for 15 minutes. Processing parameters are optimized for the materials and applications utilized.

Sensor Calibration

Sensor calibration was carried out using a microscale mechanical testing technique, called MAT-Test, developed by Hoperoft et al (Hoperoft et al., Fatigue and Fracture of Engineering Materials and Structures, vol. 28, pp. 735-742, 2005). The technique utilizes a contact surface profilometer to obtain force deflection curves for small structures. To determine the stiffness of the PDMS sensing posts, calibration posts (not recessed in wells) were fabricated. The posts are 80 μm in diameter, although posts from 30 μm to 100 μm in diameter have been fabricated. The force deflection curves obtained indicate the stiffness of the PDMS post tested. A Veeco Dektak 6M Stylus Profilometer was used to supply downforces between 10-150 μN. The resulting deflections were measured by the tool. The range of possible downforces fell within the estimated asperity force range (4-400 μN).

FIG. 4 shows the results of a typical profilometry scan. Letters have been placed in both the photograph and the scan data to show the sequential events that occur during the scan. Initially, the tip travels along the supporting PDMS slab (which had been sectioned using a razor blade). It reaches the end of the PDMS slab (location A) and drops down onto the post (location B). The tip then travels along the post, deflecting the post to a degree dependent on the downforce. The stylus tip eventually reaches the end of the post (location C) and the stylus begins to drop off of the post. The stylus tip has a radius of 12.5 μm, so 12.5 μm after reaching the end of the post the stylus starts to drop very rapidly (location D). The post deflection sensitivity in response to the stylus force were determined by plotting the measured deflection of the post against the applied force. An example of results from five identical profilometer scans along a 90 μm diameter calibration post at a downforce of 10 μN is shown in FIG. 4. Consecutive scans agree well with each other; all post deflection is purely elastic.

Post deflection is determined by plotting downforce vs. stylus vertical position and then applying a vertical position offset to the curve to force it to pass through the 0 μN, 0 μm point. This vertical position offset is used since the vertical distance measured by the stylus profilometer is referenced to an arbitrary zero point. Downforce is limited to 50 μN to ensure that the horizontal distance measured by the profilometer approximately reflects true distance along the curved posts.

Sensor Integration

The calibrated sensor is integrated into a current polishing setup used for CMP. The polisher sits on top of an AMTI force platform to allow real time measurements of the global friction and normal forces present during polishing. The polisher's plate accepts 12 inch diameter polishing pads. Applied pressure is transmitted through an aluminum axle pressing the wafer into the pad. A ½ hp Dayton motor was used to control the speed of wafer rotation, while the applied pressure was controlled by adding or removing weights. A custom 80/20 extruded aluminum frame was used as the mount for the motor assembly and additional research tools. The entire polishing setup rests on top of a 135 kg steel isolation table.

The sensor wafer replaces the standard silicon wafer being polished as shown in FIG. 5. The posts are oriented downward, with the tips in contact with the polishing pad. A Phantom v7.0 high speed camera with a 12 bit SR-CMOS sensor was coupled to a 15× relay lens and a 10× microscope objective to determine post deflection during CMP. The camera is able to capture images at rates up to 150,000 frames per second. For all experiments conducted, camera resolution and speed were set at 512×384 pixels and 10,000 frames per second, respectively. At 10,000 frames per second, under the assumption that the polishing pad's speed relative to the sensor is 0.5 m/s, asperities can be captured at 50 μm intervals. Pixel size is approximately 2 μm on a side. Light is provided to the sensor through the microscope objective using a fiber optic light guide and 90 degree soda lime plate beam splitter (50% reflection/50% transmission). FIG. 6 shows a diagram of the final optics design. Micropositioning stages are included to allow for image focus control (z direction) as well as one dimensional positioning of the image field parallel to the sensor plane (x direction). A 90° bend in the optics was used due to the space constraints of the CMP rig.

To mate the sensor with the CMP axle and secure it during polishing, a ⅛″ thick aluminum mating plate is glued to the back of the PDMS sensor wafer and connected to the axle through an acrylic linkage. The mounting plate has four main viewing windows to allow the posts to be observed optically through the back side of the PDMS. The sensor was designed to rotate as a traditional substrate would and the ⅛″ thick plate allows the microscope objective to maintain the approximately 10 mm working distance to the sensor posts. The mounting plate also contains smaller viewing windows separated by support beams to allow more sensor views as well as views of calibration posts (located at the edges of the wafer) while distributing applied load as much as possible. Any load concentrations can crack the Pyrex backing of the sensor wafer. The acrylic linkage connects the mounting plate to the rotating CMP axle by way of the axle pin. The design is shown in FIG. 7. This same axle pin connection method was used on the polishing rig. This allows quick replacement of the standard polished wafer with the sensor wafer.

During initial experiments, the sensor remained stationary in order to analyze the maximum number of sensor arrays for preliminary force observations as described below. In a static situation, the sensor can be mounted slightly off-axis from the CMP axle in the direction of the microscope objective. This allows more of the sensor arrays to be viewed by the objective which, due to space conflicts between the beam splitter and axle, observes a small number of arrays when the sensor wafer is mounted in line with the CMP axle.

Due to the off axis location of the sensor, the acrylic linkage is designed to be removable from the mounting plate allowing off-axis observations of different viewing windows by mounting the sensor in one of four orientations. Mating pins connect the linkage to the sensor mount and stabilizing pins are used to reduce rotation of the bottom acrylic plate about the axle pin. Such rotation loosens the mating pins and allows the sensor to be removed by pad shear forces.

The slurry used was a 9:1 water dilution of Cabot Microelectronics' Cab-O-Sperce SC-1 3.1 wt % abrasive. A magnetic stir bar was used during all experiments to ensure that the slurry did not aggregate and a peristaltic pump was used to deliver slurry to the pad surface at a rate of approximately 75 cc/min. The rotational velocity of the pad was either 30 rpm or 60 rpm and the loads applied were 5 lb, 10 lb, and 15 lb. At 15 lb, when accounting for the weight of the CMP axle, the pressure applied to the sensor was roughly 1.7 psi. Additional applied load resulted in the polishing pad sticking during experimentation as the friction between the sensor and the pad increased. The pad was not conditioned during experimentation due to space requirements of both the conditioner and the optical setup. Additionally, conditioning is not conducted prior to experimentation in order to minimize sticking between the sensor wafer and the polishing pad. Due to lack of conditioning, it is expected that the pad is glazed during experimentation. This may alter the shear forces present. In fact, a significant difference was observed when experimenting with a conditioned and a glazed polishing pad. The conditioned pad experienced static sticking between the PDMS sensor wafer and the pad. This effect was reduced greatly when the pad was allowed to glaze.

B. Results and Discussion

Fabrication

Sensors were successfully fabricated using the 2 layer micromold shown in FIG. 8.

Microscope images of a PDMS sensor are shown in FIG. 9. All indicator numbers and deflection notches (1 μm thick) resolved well and are clearly visible in both the micromold and the final sensor. The 100 μm through 40 μm diameter posts resolved well. The 30 μm diameter posts did not fully resolve in all cases. This can be seen in FIG. 9 where the rounded tops of these posts are not visible. For the fabrication process described above, an aspect ratio limitation of approximately 2:1 (post height:post diameter) was established.

The SU-8-100 master mold exhibited some variation in thickness across the wafer, varying from approximately 75 μm to 85 μm. The thickness of the mold determines the length of sensing posts and it is therefore preferred that the location of any post of interest is known so that its length, and hence its sensitivity to applied force, is also known.

One of the benefits of the micromolding process is that a single master mold can produce multiple sensors at minimal cost. Eight sensors were fabricated using the initial micromold. Sensor features were not degraded after 8 uses and the silanizing film did not visibly affected sensor production.

Modifications

After preliminary experimentation with the sensor wafer, it was determined that increased contrast was needed between the recessed sensing posts and their wells in order to obtain reliable deflection measurements. To obtain greater contrast, sensors were uniformly sputtered with a 15 nm thin layer of chromium. This thickness is not so great as to block light from passing through the sensor, but is thick enough to reflect incident light and provide the necessary contrast during in situ CMP image analysis. Gold was also used in this application, but it exhibited very poor adhesion relative to the chromium.

Calibration

FIG. 10 shows deflection vs. applied force obtained from calibration of the 40 μm and 100 μm diameter posts. All calibration and model results are reported at a location 100 μm from the bases of the posts. This location is selected because it is the designed post length and ensures that the stylus tip remains on the posts (ranging in length from approximately 110 μm to 130 μm) during scans. The displacement sensitivity (compliance) was 1.3 μm/μN (±0.08 μm/μN) for the 40 μm diameter post and 0.23 μm/μN (±0.02 μm/μN) for the 100 μm diameter post. Sensor deflection vs. applied force was linear over the range of applied forces. The dashed red lines in the graphs indicate the expected error in the FEA model. The expected error is computed by assign or subtracting 5% from the nominal post length. When accounting for this error, experimental data fall within the bounds of the expected error.

The stylus profilometry calibration data agrees well with a three dimensional finite element model. Note from the results shown that the more complex FEA model is preferred to predict the observed levels of deflection. A simple Euler-Bernoulli beam model, which was used to design the structures, is not sufficient as explained below. The FEA results shown in FIG. 10 are produced using Abaqus™ finite element analysis software. The C3D8R element, an eight node, linear, reduced integration continuum stress/displacement brick element with hourglass control, was used. This element uses a linear isotropic material model. An elastic modulus of 750 kPa and a Poisson ratio of 0.5 was selected based on the measurements given in previous work for a 10:1 base to curing agent ratio of the Dow Corning Sylgard 184 PDMS formulation. Nonlinear large deflection geometric effects are included in the computation, as the deflections of the structure are a significant fraction of its dimensions. Compliance of the PDMS supporting the base of the post is included by meshing and modeling a base region composed of the same material. The base region is 6 times the diameter of the post and 3 times as thick as the post is high. This size for the modeled base region was selected by increasing the size of the modeled base region under a constant load until the deflection of the tip no longer increased. Essential boundary conditions were used on the bottom of the base region, fixing both displacement and rotation. A distributed shear force was used on the top face of the post with a total integrated force equal to the applied load. The number of elements was increased until the deflection converged.

Measured post compliance was approximately twice the designed value due to the limitations of the Euler-Bernoulli beam model. There are three main phenomena included in the FEA simulation which are not in the beam model. First, the post is short compared to its width, so the assumptions of Euler-Bernoulli theory are violated. This is addressed by moving to a fully three dimensional elastic solid model. Second, the base of the post is attached to a large slab of PDMS which itself retains significant compliance, contributing to post deflection. The clamped base condition used for the beam theory model was overly stiff in this regard. This is addressed by including a compliant base region in the FEA model. Third, the deflections of the post at high forces are large compared to the diameter of the post. This violates linear beam theory. This is addressed by including nonlinear geometric effects in the FEA model. As can be seen in the results of FIG. 10, the FEA model is well validated against experimental data and can be applied for predicting post stiffness.

Sensor Integration

FIG. 11 shows an example of a micrograph obtained using the optical setup detailed above. The image shows 50 μm and 60 μm diameter posts on the final sensor, sputtered with chromium, in the polishing rig under static conditions and no applied load. The chromium produces excellent contrast between polishing posts and edges of the wells, and deflection notches and indicator numbers are clearly visible.

Using the mounting scheme described above in addition to chromium metallization, in situ CMP images of the sensor's micro post arrays were also successfully obtained. FIG. 12 shows an example of micrographs taken during polishing of 50 μm diameter posts at 30 rpm and applied loads of 10 lb or 15 lb. In the image on the left, the deflection notches and indicator numbers are still clearly visible and the contrast between the sensing post and the edge of the well is similar to that of the static situation. In the image on the right, all of the chromium on the surface of the sensor has been polished away. Deflection notches and indicator numbers are no longer visible. The posts in the left image are not in intimate contact with the polishing pad as indicated by the lack of chromium removal. In addition to adding necessary contrast to micrographs, the presence or absence of chromium is a useful indicator for whether or not certain regions of the wafer are being polished. The 50 μm post in FIG. 12 can be seen deflecting due to the shear stress as evident by the shadow elongated towards the upper left of the image (the direction of pad asperity movement). The chromium well is brighter in the image on the right due to intensity gain changes between the two images.

Sensor post deflections reported below are determined through image analysis of high speed camera images using the software ImageJ. A static image of the sensing post being analyzed is first taken and used as the base image to compare further dynamic images to. Exact circles are fit to the outside and the inside of this base image. These circles are concentric if the post is not deflected initially. The centers of these circles are obtained and any offset between the two is recorded.

The next step involves analyzing dynamic polishing high speed video obtained using the optical setup detailed above. Images were selected that demonstrate visible deflection of posts. The orientation of the camera setup was held constant for all experiments analyzed and the movement of polishing pad asperities in all images is towards the upper left, approximately 25° from the positive y axis. This is calculated by following asperity motion across images and calculating the angle of motion using the original location as the zero point.

The circle fit to the static sensing post in the initial step above is fit to the center of all posts analyzed. The centers of these circles are recorded. The final image analysis step includes determining the centers of the post wells. The circle fit to the outside of the static post is fit to the outsides of each post analyzed and their centers are recorded.

The x and y components of the centers of deflected post tops are subtracted from the corresponding x and y components of the centers of each post well to determine the deflection of each post relative to its well. This value is then referenced to the deflection of the static post relative to its well to determine the net deflection of the dynamic posts due to CMP shear stresses. In this approach, post tip deflection is recorded relative to the static base case and permanent deflections of posts are not mistaken for deflections due to shear forces present. FIG. 13 shows an example plot of dynamic post deflection charted against the static case. As in the example shown, many of the posts analyzed deflected in the direction of asperity movement as expected.

Observed Shear Forces

Deflections of 50 μm to 100 μm diameter posts were observed using the method described above. Only large deflections were analyzed in order to find the maximum shear forces on the PDMS posts. FIG. 14 shows the average deflections (during large deflection events) of sensor posts due to shear forces present at a pad speed of 30 rpm and downforces of 5 lb, 10 lb, or 15 lb. FIG. 15 shows the average deflections (during large deflection events) of sensor posts due to shear forces present at a pad speed of 60 rpm and downforces of 5 lb or 15 lb. Data points are made up of the average of 10 data points. The red bars indicate error due to uncertainty in fitting circles to the micro post tops and wells during image analysis. Table 2 shows the deflection range observed for each post under downforces and pad rotational speeds. The average angle of deflection throughout the data is approximately 26.6° from the y axis. This agrees very well with the calculated angle of pad asperity movement which is 25° from the y axis.

	TABLE 2
	Observed Post Deflections (μm)
	30 rpm	60 rpm
Post Diameter (μm)	5 lb	10 lb	15 lb	5 lb	15 lb
50			2.2-11.5		0.4-10.6
60	3.7-9		2.1-16.3	5.6-11.6
70	2.5-14.1	4.1-19.3		3.1-15.6
80	2.1-22.3	7.6-24.6		1.6-13.4	5.8-7.0
90	3.8-16.3	4.2-20.7	4.5-22.6	5.6-21.1	2.5-22.7
100	3.8-17.1	6.0-22.2	2.0-27.4	1.4-16.1	3.5-10.4

This deflection data can be converted into actual shear forces observed during sensor polishing by using the FEA model developed above (FIG. 21) along with the calibration results from the sensor wafer used in the experiments. The FEA model predicted sensor post deflection using the established PDMS 10:1 ratio elastic modulus of 750 kPa. The stiffness of the sensor posts have been measured at approximately two times the stiffness of previously calibrated posts. Knowing this modified post modulus, the FEA results are adjusted to determine the in situ CMP shear forces observed by the sensor. The shear forces presented here are the average maximum shear forces present while polishing the PDMS surface of the shear sensors.

FIG. 16 shows the force deflection curves for the recalibrated 40 μm and 100 μm diameter posts on the shear sensor used in experimentation. Local shear forces present during polishing with a pad rotation of 30 rpm are shown in FIG. 17. Local shear forces present during polishing with a pad rotation of 60 rpm are shown in FIG. 18.

The expected trend of decreasing shear forces with increasing pad rotation is visible between the two figures. The 100 μm diameter posts polished at 60 rpm show increasing shear force with increasing downforce. This trend does not show up in the other posts. There is a strong trend of increasing shear forces observed by increasing sensor post diameters at a pad rotation of 30 rpm. The present invention is not limited to a particular mechanism. Indeed, an understanding of the mechanism is not necessary to practice the present invention. Nonetheless, it is contemplated that this trend, visible under 60 rpm polishing conditions but not as strong, is due to the increased number of asperities that can contact the larger posts at a single time. At a constant downforce, shear forces observed by the 100 μm diameter posts are about 10 times larger than those observed by the 50 μm posts. Maximum shear forces of approximately 270 μN are obtained by the 100 μm diameter sensing posts under an applied load of 15 lb and a pad rotation speed of 30 rpm. The minimum resolvable shear force is approximately 5 μN.

The shear force delivered by a single asperity while polishing PDMS with an applied load of 15 lb was estimated at approximately 20 μN. This agrees well with the design assumption that asperity forces are on the order of 40 μN under an applied pressure of 1.7 psi. At a maximum applied load of 16.3 lb (the 1.3 lb weight of the CMP axle is added), the pressure on the sensor wafer is roughly 1.7 psi. This pressure is calculated by dividing the applied load by 80% of the total wafer area in order to account for the recessed calibration and deflecting post wells that are not in contact with the pad.

Example 2

Silicon Shear Sensor

This Example describes the fabrication of a floating silicon sensor. Sensor fabrication is performed using the bonded lost wafer process. This process entails the bonding of two semiconductor grade wafers, and then etching nearly all of one of those wafers away leaving a thin layer of devices attached to the remaining wafer.

The first wafer used in this process is made from PYREX glass. The second wafer can be either a silicon on insulator (SOI) wafer, or a silicon wafer which has been highly doped. In either case, the sensors are etched into the surface of the second wafer before anodically bonding its silicon surface to the PYREX wafer. After bonding, the bulk of the silicon is removed using either wet chemical or reactive ion etching. In the case of the SOI wafer, there is a thin layer of silicon dioxide left after the handle layer is etched away. This layer is removed using reactive ion etching. A diagram of the fabrication process used to make this sensor is shown in FIG. 19.

Patterns of both wafers are defined using photolithography, wet chemical etching and reactive ion etching. The first step in sensor fabrication defines both patterns in the device layer. There are two patterns etched into the device layer. The first pattern defines bonding pads and the dimples. The second etch extends through the device layer, using either the oxide layer or the doped silicon as an etch stop. This etch defines the sensor elements and the flexures. Next, wells are etched into the glass wafer. The fourth processing step utilizes anodic bonding to bond the glass wafer to the SOI wafer such that the sensor elements lie over the wells etched into the glass. Finally, the bulk and oxide layers are etched away to reveal the sensor elements. This final etch combines wet etching to remove most of the bulk layer with a final reactive ion etch to remove the oxide layer.

Example 3

Alternative Fabrication Process

This Example describes an alternative fabrication process for the floating element sensors (FIG. 22) below. An image of the fabrication process described in Example 1 is shown in FIG. 28. The alternative process described herein starts with the deposition of 5 microns of Parylene C onto a glass wafer by vapor phase Parylene deposition. Photoresist is patterned using photolithography to create the desired sacrificial layer geometry. A thin Chromium hardmask (100 nm) is sputtered on and lifted off. The Chromium hard mask is used to mask a medium power, medium pressure oxygen plasma etch of the Parylene C, resulting in the desired sacrificial layer geometry. The Chromium hard mask is stripped in a wet etch. Following this, photoresist is photolithographically patterned to produced the floating element shape. A thin Ti/Cu seed layer is sputtered on and patterned via liftoff. Thick copper (5-10 microns) is plated onto the seed layer using copper electroplating. After drying, the Parylene C sacrificial layer is etched out using a long, medium power, high pressure oxygen plasma etch. This results in the released floating element structure.

If desired, dimples are added to the bottom of the element by adding a second oxygen plasma etch of the Parylene C layer before Ti/Cu sputtering. This etch is masked using a lithographically patterned Chromium thin film. The etch is timed to extend only 1 or 2 microns into the Parylene, rather than all the way through. The dimples help to reduce stiction between the copper floating element and the glass wafer.

FIG. 4

Measurement of Shear Stress

Data showing the ability of the PDMS post-like shear stress sensors to measure wafer-pad interaction forces during CMP is presented in FIGS. 23-27 below. The figures show that the posts can be well calibrated and are linear, and can be applied to the measurement of dynamic forces at high speeds in the polishing environment. The forces are highly variable in time with time constants on the order of milliseconds and force magnitudes in the hundreds of micronewton range (for these conditions). These measurements represent the first in situ dynamic measurements of force in the polishing environment.

All publications and patents mentioned in the above specification are herein incorporated by reference as if expressly set forth herein. Various modifications and variations of the described method and system of the invention will be apparent to those skilled in the art without departing from the scope and spirit of the invention. Although the invention has been described in connection with specific preferred embodiments, it should be understood that the invention as claimed should not be unduly limited to such specific embodiments. Indeed, various modifications of the described modes for carrying out the invention that are obvious to those skilled in relevant fields are intended to be within the scope of the following claims.

Claims

1. A system, comprising:

a) at least one deformable sensor, wherein said sensor is deformed under shear stress arising during chemical mechanical polishing;

b) an apparatus for chemical mechanical processing (CMP), wherein said apparatus is in active communication with said sensor; and

c) a detection device for detection of deformation of said sensor.

2. The system of claim 1, wherein said sensor is composed of poly-dimethyl-siloxane (PDMS).

3. The system of claim 2, wherein said sensor comprises a plurality of posts.

4. The system of claim 3, wherein said post is coated with a metal.

5. The system of claim 4, wherein said metal is chromium.

6. The system of claim 3, wherein said post has diameter of at least 1 μm.

7. The system of claim 3, wherein said post has a diameter of approximately 30 μm to 100 μm.

8. The system of claim 3, wherein said post has a height:diameter ratio of less than approximately 4:1.

9. The system of claim 3, wherein said sensor further comprises molded markings.

10. The system of claim 1, wherein said system comprises an array of said sensors.

11. The system of claim 1, wherein said sensors are affixed to a polishing wafer.

12. The system of claim 11, wherein said polishing wafer is affixed to said CMP apparatus using an aluminum mating plate.

13. The system of claim 12, wherein said aluminum mating plate comprises a plurality of viewing windows.

14. The system of claim 1, wherein said sensor comprises a floating element supported by a flexure.

15. The system of claim 14, wherein said sensor is composed of a material selected from the group consisting of silicon and a thick metal film.

16. The system of claim 15, wherein said thick metal film is selected from the group consisting of copper and nickel.

17. The system of claim 15, wherein said floating element is circular.

18. The system of claim 1, wherein the detection device comprises a light source, a microscope, and a CCD camera.

19. The system of claim 18, wherein said camera is a high speed camera.

20. The system of claim 18, wherein said light source is a fiber optic light source.

21. A method, comprising contacting at least one deformable sensor, wherein said sensor is deformed under shear stress arising during chemical mechanical polishing, with an apparatus for chemical mechanical processing (CMP) under conditions such that said sensor is deformed.

22. The method of claim 21, further comprising the step of calculating shear forces on said sensor based on said deformation of said sensor.

23. The method of claim 22, wherein said shear forces are calculated using a computer processor and computer software.