Shear-Induced Structures and Thickening in ... - ACS Publications

Sep 24, 2013 - Cabot Microelectronics Corporation, Aurora, Illinois 60504, United States ... ABSTRACT: Chemical mechanical polishing (CMP) is an essen...
0 downloads 0 Views 4MB Size
Article pubs.acs.org/Langmuir

Shear-Induced Structures and Thickening in Fumed Silica Slurries Nathan C. Crawford,† S. Kim R. Williams,‡ David Boldridge,§ and Matthew W. Liberatore*,† †

Department of Chemical and Biological Engineering and ‡Department of Chemistry and Geochemistry, Colorado School of Mines, Golden, Colorado 80401, United States § Cabot Microelectronics Corporation, Aurora, Illinois 60504, United States S Supporting Information *

ABSTRACT: Chemical mechanical polishing (CMP) is an essential technology used in the semiconductor industry to polish and planarize a variety of materials for the fabrication of microelectronic devices (e.g., computer chips). During the high shear (∼1,000,000 s−1) CMP process, it is hypothesized that individual slurry particles are driven together to form large agglomerates (≥0.5 μm), triggering a shear thickening effect. These shear-induced agglomerates are believed to cause defects during polishing. In this study, we examined the shear thickening of a 25 wt % fumed silica slurry with 0.17 M added KCl using in situ small-angle light scattering during rheological characterization (rheo-SALS). The salt-adjusted slurry displays a ∼3-fold increase in viscosity at a critical shear rate of 20,000 s−1 during a stepped shear rate ramp from 100 to 25,000 s−1. As the shear rate is reduced back to 100 s−1, the slurry displays irreversible thickening behavior with a final viscosity that is 100-times greater than the initial viscosity. Corresponding rheo-SALS images indicate the formation of micrometer scale structures (2−3 μm) that directly correlate with the discontinuous and irreversible shear thickening behavior of the fumed silica slurry; these micrometer scale structures are 10-times the nominal particle diameter (∼0.2 μm). The scattering patterns from the 25 wt % slurry were corroborated through rheo-SALS examination of 27 and 29 wt % slurries (CKCl = 0.1 M). All slurries, regardless of ionic strength and solids loading, display scattering patterns that are directly associated with the observed thickening behavior. Scattering was only observable during and after thickening (i.e., no scattering was detected in the absence of thickening). This work serves as the first in situ observation of micrometer scale structures within the fumed silica CMP slurry while under shear.



INTRODUCTION In order to meet the growing demand for faster and more powerful microprocessors, the semiconductor industry must continually work to reduce the feature size of today’s microelectronic devices.1 The most advanced silicon chips contain over one billion transistors in a square centimeter of surface area.2,3 As a result, this places great demands on the chemical mechanical polishing (CMP) community. The formation of surface defects, including gouges, pits, scratches, etc., are a key limiting factor to the success of the CMP process. It is well understood that CMP-induced defects can degrade microelectronic device performance4 and can even lead to catastrophic device failure.3 Thus, the reduction of all types of surface defects has become essential to the fabrication of microelectronic devices. During the CMP process, slurry is introduced onto a relatively “rough” polymer polishing pad, commonly blown polyurethane, that consists of an open macroscopic porous structure and has a patterned surface to aid slurry transport.5 The pad can be up to 3 feet in diameter and rotates at speeds near 100 rpm (∼4 m/s at the outer radius of the polishing pad). A semiconductor wafer is mounted in a carrier and then pressed face down into near contact with the pad at a given pressure (∼2 to 6 psi6). The colloidal slurry is then sandwiched into a thin film (1−100 μm) between the two disks that rotate in © 2013 American Chemical Society

unison. The combination of the orthogonal force from the wafer, the abrasive nature of the slurry, and the grinding action of the polishing pad leads to synergistic mechanical removal of material. The active chemicals in the slurry enhance removal rates and improve selectivity between different surface materials.5 The high speed polishing environment exposes the slurry to shear rates in excess of 1,000,000 s−1.7 It is believed that during this intense shearing, individual slurry particles are driven together to form large agglomerates (≥0.5 μm),8 which cause a spike in the slurry’s viscosity known as shear thickening.9,10 Shear thickening (or dilantacy) is defined as an increase in a material’s viscosity with increasing deformation (i.e., shear rate or stress). The growth in viscosity is usually believed to be the result of transient stress bearing particles (hydroclusters) that form when close range hydrodynamic lubrication forces dominate over interparticle repulsive forces under flow.10,11 Hydrocluster formation has been extensively examined using simulations12−14 and a myriad of experimental techniques.15−20 Currently, hydroclusters are considered the defining feature of the shear thickened state.10 Received: July 16, 2013 Revised: September 20, 2013 Published: September 24, 2013 12915

dx.doi.org/10.1021/la402631p | Langmuir 2013, 29, 12915−12923

Langmuir



Recent rheological studies have shown that fumed silica CMP slurries will shear thicken under high shear rates (>10,000 s−1).21−23 The slurries’ thickening response can be tuned by adjusting both ionic strength21,24 and particle surface hydration.22 The shear thickening behavior of these modestly concentrated (∼25 wt %) fumed silica slurries is discontinuous and, unlike many shear thickening systems, irreversible. The unique thickening of these slurries continues to puzzle researchers, displaying unexplained cation dependencies in the presence of added salts24 and additional close range stabilizing forces22,24 not included in Derjaguin−Landau− Verwey−Overbeek (DLVO) theory.25,26 Although multiple studies have examined the high shear rheological behavior of these fumed silica slurries, little work has been done to elucidate the underlying mechanism(s) of the observed irreversible thickening behavior. In order to gain a greater understanding of the driving force(s) behind the slurries’ irreversible thickening behavior, a more robust method than standard bulk rheology must be employed. Currently, a substantial body of literature exists for reversible shear thickening systems and their study using rheooptical techniques.11,15−17,19,27 Rheo-optics are based on the notion that a material’s microstructure controls its rheology and that material deformation has a distinct optical fingerprint.28 The optical signature of a material under flow can then be related to microstructural changes (i.e., orientation, size parameters, phase behavior, etc.). The choice of optical technique (i.e., birefringence, light scattering, Raman spectroscopy, optical microscopy, etc.) is governed by the material’s optical properties, the level of microstructure to be quantified, and the size scale of interest. For this study, we are concerned with correlating viscosity changes (i.e., shear thickening) with changes in particle size on the micrometer scale while under flow. Therefore, rheological characterization coupled with in situ small-angle light scattering (rheo-SALS) will be employed. Rheo-SALS probes for shearinduced structure formation on the micrometer length scale (∼1−5 μm), a size range 5× to 10× larger than the nominal particle diameter (0.1 to 0.2 μm) of the fumed silica aggregates. Thus, the large agglomerates (≥0.5 μm) believed to be responsible for generating CMP-induced defects1,29−31 and the reported shear thickening behavior should be detectable by rheo-SALS. Ideally, the utilization of SALS with rheological characterization will allow for real time analysis of shearinduced structure formation and their correlation with changes in viscosity. High shear rheo-optical measurements are challenging and, as a result, are scarce. For a rotating rheometer, shear rates up to 8,000 s−1 were used to study thixotropic solutions of high molecular weight polymers, but not shear thickening.32 Thus, to our knowledge, the work presented here is the first observation of a shear thickening system using rheo-SALS under high shear (≥10,000 s−1). For this study, the formation of a detectable scattering pattern directly correlates with the thickening response of the slurry. As the scattering pattern evolves throughout the slurries’ flow curve, size and orientation of the shear-induced agglomerates are determined. This study provides insight into the mechanism behind the irreversible thickening behavior of these fumed silica slurries and links the observed thickening to the formation of micrometer sized structures, which are believed to be the source of CMP-induced defects.

Article

EXPERIMENTAL METHODS

For this study, 25, 27, and 29 wt % fumed silica slurries (d = 200 ± 25 nm, determined via dynamic light scattering) with and without the addition of salt (0.1 and 0.17 M KCl) were employed. The fumed silica aggregates (Figure 1) were electrostatically stabilized at pH 11

Figure 1. Transmission electron microscopy image of fumed silica (provided by Cabot Microelectronics Corporation). (well above silica’s isoelectric point of pH 2) through the addition of KOH. All slurry material was provided by Cabot Microelectronics Corporation (Aurora, IL) and was a simplified version of the commercial product; consisting of fumed silica, water, and KOH. Concentrated slurries (34.7 wt %) were diluted to a target solids fraction of 25, 27, and 29 wt % using ultrapure DI water (18 Ω) and KCl solutions (Mallinckrodt Chemicals, Phillipsburg, NJ) of 0.52 M (for the 25 and 29 wt % slurries) and 0.39 M (for the 27 wt % slurry). The final KCl concentration (CKCl) for the 25 wt % slurry was 0.17 and 0.1 M KCl for the 27 and 29 wt % slurries. Salt and silica concentrations were altered to ensure that thickening occurred below the maximum achievable shear rate in our high shear rheo-SALS setup (30,000 s−1). After dilution, slurries were stored under ambient conditions for 24 h before commencing rheological tests. High solids (>20 wt %) and high salinity (>0.1 M) slurries are inadequately stable and typically form large agglomerates and/or complete gel-like networks in 1−2 weeks; please see our previous publications for further details.8,23,24 Rheology and SALS data were collected simultaneously using an AR-G2 rheometer (TA Instruments, New Castle, DE) with the commercially available SALS attachment. A transparent, quartz parallel-plate configuration (50 mm diameter) with a 100 μm gap was used for all tests. The commercial laser housing includes a lens that is embedded into the Peltier plate surface to allow the light source to traverse through the sample thickness (i.e., the gradient direction). The transition from the Peltier surface to the lens is “rough” to the touch and can disrupt the flow pattern, leading to erroneous results at small gap heights (≤300 μm). Therefore, an additional transparent plate was affixed to the Peltier plate using vacuum grease to establish a “smooth” bottom surface. The added plate is housed in a metal surround for additional stability during experimentation (Figure 2). All measurements were conducted at 20 °C, with temperature control of ±0.1 °C provided by the Peltier plate. A stepped shear rate ramp (and reduction) was employed to monitor the slurries’ flow behavior at shear rates from 100 to 30,000 s−1. Each shear rate was held constant for 35 s, while rheological data were collected every 3 s and SALS images were captured every 10 s (instrument limit). Rheo-SALS tests were repeated in triplicate (at minimum) to confirm reproducibility. Rheo-SALS information was averaged over the peak hold time (35 s) and a single viscosity and scattering pattern are reported per measured shear rate. Longer peak hold experiments were conducted (∼1−5 min), but the scattering patterns showed little time dependent behavior. The size scale of 12916

dx.doi.org/10.1021/la402631p | Langmuir 2013, 29, 12915−12923

Langmuir

Article

Figure 2. Photograph (left) and schematic (right) of the modified TA Instruments’ SALS attachment used for shear rates >10,000 s−1. Schematic adapted from TA Instruments’ Accessory Manual.33 scattering objects captured by this SALS setup is 1.0−4.6 μm (q = 1.4−6.1 μm−1). SALS images were analyzed using ImageJ with standard protocols for subtracting the background and removing the beam stop from the raw images.33,34 The characteristic length (aC) and aspect ratio (aR) of the shearinduced structures were determined using Debye-Bueche35 plots (I−0.5 vs q2).11,36,37 A linear fit to the radially averaged data plotted in the Debye-Bueche format was calculated for each scattering image using the method of least-squares (example plot included in the Supporting Information). The slope and intercept of the linear fit were used to calculate aC (aC = [slope/intercept]0.5) for the scatter-inducing structures. A calibration of the Debye-Bueche characteristic length was performed using a Microbead NIST Traceable Particle Size Standard (Polysciences, Inc., Warrington, PA) of polystyrene microspheres (3 μm diameter). A correction factor was determined using the ratio of the calculated aC (3.2 μm) with the reported diameter of the standard (3 μm). The correction factor (∼1.07) was then applied to our calculations of aC in order to report “true” versus “apparent” characteristic lengths for the fumed silica slurries (this is standard procedure outlined in the instrument manual33). A correlation between the characteristic length and volume fraction of solids within the slurry was avoided because the “true” solids volume fraction is unknown. Entrained liquid within the porous structure of the fumed silica particles can cause the nominal, weight-based volume fraction to differ from the “true” volume fraction by as much as 30-fold.38 In addition, the orientation of the shear-induced structures was analyzed using the aspect ratio of the scattering patterns, determined by taking the ratio of the q values in the vorticity and flow directions for a given value of the measured intensity. Particle sizing, before and after a rheo-SALS experiment, was achieved via dynamic light scattering (DLS). Before commencing DLS analysis, fully concentrated slurries (collected directly from the rheoSALS tooling) were diluted 1:104 (by volume) using a stock solution of KOH (pH = 11). Dilution with DI water was avoided to prevent particle agglomeration through alteration of the particles’ surface charge via pH adjustment. Brookhaven Instruments (Holtsville, NY) ZetaPALS at a scattering angle of 90° was employed for DLS sizing.

to be the result of both particle layering, where layers of particles slide past one another more readily than if they were randomly distributed,40 and when entropic viscosity contributions become negligible due to relatively large hydrodynamic stresses.20 The fumed silica slurries’ viscosity is compared to the viscosity of DI water (Figure 3). As expected, DI water behaves

Figure 3. Stepped shear rate ramp (filled symbols) and reduction (open symbols) for DI water (triangles) and a 25 wt % silica slurry with a concentration of added KCl, CKCl, of 0.17 M (squares). Letters (a) to (h) and (w) to (z) correspond to the locations of the reported SALS images in Figure 4 for the slurry and water samples, respectively.

as a Newtonian fluid displaying a viscosity that is virtually independent of shear rate. The average measured viscosity (across the full shear rate ramp and reduction experiment) of the DI water control was (7.5 ± 0.4) × 10−4 Pa·s, which is within 25% of the reported value for the viscosity of water at 20 °C (1.0 × 10−3 Pa·s41). Large systematic measurement errors (>40%), due to “non-parallelism,” are expected when using a parallel-plate geometry at small gap heights (≤100 μm).23,42,43 Once the shear rate is increased to 20,000 s−1 the slurry makes an abrupt transition from shear thinning to shear thickening (indicated by the dashed line in Figure 3). The 3fold increase in viscosity displayed by the fumed silica slurry at 20,000 s−1 is deemed discontinuous shear thickening and has



RHEO-SALS DURING SHEAR THICKENING The rheology of the 25 wt % fumed silica slurries undergoes several changes as a function of shear rate. Slight shear thinning behavior is exhibited with increasing shear rate from 100 to 15,000 s−1, where a 30% decrease in viscosity is observed. Shear thinning is a common response displayed by colloidal suspensions.9,10,39 Reduction in suspension viscosity is believed 12917

dx.doi.org/10.1021/la402631p | Langmuir 2013, 29, 12915−12923

Langmuir

Article

10,000 s−1) for the 25 wt % silica slurry (Figure 4a and b). No scattering indicates that only subdetectable particles (