Chemical Functionalization of Silica and Alumina Particles for

4816

Langmuir 2005, 21, 4816-4823

Articles Chemical Functionalization of Silica and Alumina Particles for Dispersion in Carbon Dioxide Pamela M. Visintin,† Ruben G. Carbonell,‡ Cynthia K. Schauer,*,† and Joseph M. DeSimone*,†,‡ Department of Chemistry, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina 27599-3290, Department of Chemical Engineering, North Carolina State University, Raleigh, North Carolina 27695-7565 Received August 31, 2004. In Final Form: March 16, 2005 The steric stabilization and flocculation of modified silica and alumina particle suspensions in condensed CO2 were studied. Silica particles (average diameters of 7 and 12 nm) were functionalized using chlorosilanes of the form CnF2n+1CH2CH2Si(CH3)2Cl (n ) 8, 4, or 1) to give CnF2n+1-silica. Alumina particles (diameter of 8-14 nm) were grafted with C8F17CH2CH2Si(OEt)3 and chemically modified with perfluorononanoic acid to yield C8F17-alumina and C8F17COOH-alumina, respectively. Elemental analysis and thermogravimetric analysis on the derivatized particles were carried out, and surface coverage was calculated. The stabilization of these modified particles in condensed CO2 was quantified using turbidimetry. Particle stability was found to increase with increasing fluorinated tail length, temperature, and CO2 density. Unmodified particles and those modified with only -CF3 tails were unstable in condensed CO2. Stabilization in supercritical CO2 is continuous up to 24 h for the CnF2n+1-silica (n g 4) particles and 96 h for the C8F17-alumina particles. The C8F17COOH-alumina particles gave a significantly higher graft density than the C8F17-alumina particles but are not as stable in CO2. The C8F17-alumina particles were stable at lower CO2 densities than the modified silica particles. This stability difference may be attributed to the precursor organosilanes being monofunctional (modified silica) versus trifunctional (modified alumina), producing different structures on the surface.

Introduction The compatibility of liquid and supercritical (sc) CO2 with nanometer-length feature sizes and copper/low-k technology make it an advantageous solvent for semiconductor device fabrication. Several applications using CO2 as the solvent are being developed because CO2 has favorable wetting properties, a low dielectric constant, and it makes possible a “dry in-dry out” process.1 Other advantages of CO2 (Tc ) 31 °C, Pc ) 73.8 bar) are that it is nontoxic, nonflammable, environmentally benign, and readily available. Chemical mechanical polishing (CMP) has emerged as the primary technology for implementing copper into semiconductor devices. CMP is a process which uses a slurry composed of etchant reagents and nanometer-sized abrasive particles to remove excess copper from uneven topography on a wafer until a flat surface is achieved. We recently demonstrated the “chemical” feasibility of a CO2based CMP process by oxidizing and chelating copper metal in CO2 media.2,3 The “mechanical” aspect of a CO2based CMP process will require the stabilization of the inorganic abrasive particles for precision polishing. It is * Authors to whom correspondence should be addressed. Email: [email protected] (C.K.S.); [email protected] (J.M.D.). † University of North Carolina at Chapel Hill. ‡ North Carolina State University. (1) Weibel, G. L.; Ober, C. K. Microelectron. Eng. 2003, 65, 145. (2) Bessel, C. A.; Denison, G. M.; DeSimone, J. M.; DeYoung, J.; Gross, S.; Schauer, C. K.; Visintin, P. M. J. Am. Chem. Soc. 2003, 125, 4980. (3) Visintin, P. M.; Bessel, C. A.; White, P. S.; Schauer, C. K.; DeSimone, J. M. Inorg. Chem. 2005, 44, 316.

important that the particles in the slurry be stable to surface deposition (sedimentation), as sedimentation onto the wafer surface results in a higher density of scratches and adhesion of the particles to the surface, which may make particles difficult to remove during post-CMP cleaning.4 The most commonly used abrasive particles in aqueous-based CMP polishing slurries are silica and alumina.5,6 Previous studies by Johnston and co-workers have shown that hydrophilic silica particles (average diameter g 175 nm) in condensed CO2 are not stable to aggregation without (1) physisorbed high-molecular-weight fluorinated surfactants or (2) surface-grafted poly(1H,1H-dihydroperfluorooctyl methacrylate) (PFOMA) or poly(dimethylsiloxane) (PDMS) polymers.7-9 In the case of the surfacegrafted PDMS tails, 15 wt% hexane needs to be added as a co-solvent to CO2 for the particles to be stabilized.8 Carbon dioxide has far weaker van der Waals forces than hydrocarbon solvents due to its low polarizability/volume and refractive index, and consequently, “CO2-philic” segments, such as those containing fluorine and silicon, (4) Singh, R. K.; Lee, S.-M.; Choi, K.-S.; Basim, G. B.; Choi, W.; Chen, Z.; Moudgil, B. M. MRS Bull. 2002, 27, 752. (5) MacDonald, M. J.; Murray, M. P.; Chamberlin, T. S. Slurry and Use thereof for Polishing. U.S. Patent 2,001,013,506, 2001. (6) Hellring, S. D.; Y., L.; McCann, C. P.; Babu, S. V.; S., N. Silica and a Silica-Based Slurry. U.S. Patent 2,003,094,593, 2003. (7) Calvo, L.; Holmes, J. D.; Yates, M. Z.; Johnston, K. P. J. Supercrit. Fluids 2000, 16, 247. (8) Yates, M. Z.; Shah, P. S.; Johnston, K. P.; Lim, K. T.; Webber, S. J. Colloid Interface Sci. 2000, 227, 176. (9) Sirard, S. M.; Castellanos, H.; Hwang, H. S.; Lim, K. T.; Johnston, K. P. Ind. Eng. Chem. Res. 2004, 43, 525.

10.1021/la047823c CCC: $30.25 © 2005 American Chemical Society Published on Web 04/23/2005

Functionalization of Silica and Alumina Particles

Figure 1. Structures of the fluorinated ligands and the corresponding modified silica and alumina particles used in the study of sterically stabilized abrasive particles in condensed CO2.

are needed to prevent aggregation of the particles.8,10-12 These “CO2-philic” tails have low cohesive energy densities, which help overcome the attractive van der Waals forces between the particles and thereby provide steric stabilization. The objective of this study is to examine the steric stabilization and flocculation in condensed CO2 of modified silica and alumina particles using low-molecular-weight “CO2-philic” tails as the stabilizing agent. Hydrophilic silica particles (average diameters of 7 and 12 nm) were functionalized using monochlorosilanes, C8F17CH2CH2Si(CH3)2Cl (1), C4F9CH2CH2Si(CH3)2Cl (2), and CF3CH2CH2Si(CH3)2Cl (3). The syntheses of fluorosilyl-derivatized alumina particles (diameter of 8-14 nm) and alumina particles with chemisorbed perfluorononanoic acids are also reported, which allows a comparison between covalently attached tails and hydrogen-bonded tails. The structures of the fluorinated ligands and corresponding modified silica and alumina particles are shown in Figure 1. Finally, turbidity measurements were used to assess modified particle stability as functions of fluorinated tail length, temperature, and CO2 density. To the best of our knowledge, this is the first report of the dispersion and stabilization of alumina particles in CO2 media. Experimental Section Materials and General Procedures. All reactions involving air- or moisture-sensitive materials were conducted in ovendried glassware under an atmosphere of argon using standard Schlenk techniques. Argon was purified by passage through columns of BASF R3-11 catalyst and 4-Å molecular sieves. Chlorodimethyl-3,3,3-trifluoropropylsilane (Lancaster), chlorodimethyl-3,3,4,4,5,5,6,6,6-nonafluorohexylsilane (Aldrich), and 1H,1H,2H,2H-perfluorodecyldimethylchlorosilane (Lancaster) were stored and handled in an argon-filled Vacuum Atmospheres (10) DeSimone, J. M.; Maury, E. E.; Menceloglu, Y. Z.; McClain, J. B.; Romack, T. J.; Combes, J. R. Science 1994, 265, 356. (11) Harrison, K.; Goveas, J.; Johnston, K. P.; O’Rear, E. A., III. Langmuir 1994, 10, 3536. (12) O’Neill, M. L.; Cao, Q.; Fang, M.; Johnston, K. P.; Wilkinson, S. P.; Smith, C. D.; Kerschner, J. L.; Jureller, S. H. Ind. Eng. Chem. Res. 1998, 37, 3067.

Langmuir, Vol. 21, No. 11, 2005 4817 glovebox. The nonporous, fumed silica particles Aerosil A200 and A300 were supplied by the Degussa Corporation and dried overnight at 125 °C. The average particle diameters were 12 (Aerosil A200) and 7 nm (Aerosil A300), and the specific surface areas were 200 ( 25 and 300 ( 30 m2/g, respectively.13 γ-Alumina (99.99%, Accumet Materials) with a particle diameter of 8-14 nm and a specific surface area of 250 ( 50 m2/g was dried at 125 °C for several days. For the silica and alumina particles, the surface area/volume ratio (3/radius) divided by the particle density gave values that correspond to the reported specific surface area. This calculation confirmed that the silica and alumina particles are relatively nonporous, and therefore, the tails bind primarily to the surfaces. Triethoxysilane (TES), 1H,1H,2H-perfluoro-1-decene, and perfluorononanoic acid were from Aldrich, and 1H,1H,2H,2H-perfluorodecyltriethoxysilane (Rf17TES) was from Lancaster. A 50 mM dicyclopentadienylplatinum(II) chloride, [Cp2]PtCl2, solution in dry chloroform (Aldrich) was used to catalyze the hydrosilylation. The [Cp2]PtCl2 catalyst was synthesized according to a procedure developed by Apfel et al.14 Toluene and THF were distilled from sodium benzophenone ketyl, and triethylamine was distilled from CaH2 immediately prior to use. Unless otherwise noted, all solvents and reagents were purchased in reagent grade from commercial suppliers and used without further purification. Carbon dioxide (SFC/SFE grade) was purchased from Air Products. Characterization Methods. Infrared spectra were recorded using KBr pellets on a Thermo Nicolet Magna-IR 750 spectrometer. NMR spectra were recorded at room temperature on a Bruker Avance 400 MHz spectrometer. Elemental analyses were performed by Atlantic Microlab, Inc. (Norcross, GA). Thermogravimetric analyses (TGA) experiments were carried out on a Seiko Model RTG220 instrument at a heating rate of 2 °C/min in air. The differential scanning calorimetry (DSC) trace (in air) was recorded on a Seiko Model DSC220C instrument. An uncovered aluminum pan was loaded with approximately 10 mg of sample, heated to 110 °C at a rate of 10 °C/min and held at 110 °C for 1 h. The temperature was then raised to 590 °C at a rate of 10 °C/min. Silylation of Silica Particles (4-5). Silanes were covalently attached to the two silica samples, 12 and 7 nm, following a modified procedure.15 Silica particles (0.50 g, 8.32 mmol), 10 times molar excess (based on 2-3 SisOH/nm2)13 of the respective chlorosilane 1-3, and dry toluene (40 mL) were stirred in a 100mL three-necked round bottom flask equipped with a reflux condenser under argon at room temperature. Dry triethylamine (0.4 mL) was added as a catalyst. The reaction mixture was then stirred at 100 °C for 3 d. After this time, the derivatized silica particles were centrifuged, washed with THF (30 mL) to remove excess chlorosilane, washed with absolute ethanol (3 × 30 mL) to remove the triethylammonium chloride byproduct, washed with pentane (30 mL), and dried under vacuum for 48 h. 1H NMR spectroscopy (CD2Cl2) showed no residual triethylammonium chloride. Graft densities, δ (mmol/g), of the covalently attached fluorosilyl groups were determined by fluorine elemental analysis (eq 1) and TGA (eq 2), where gE is the F content, ME is the molecular weight of F, Z is the number of F atoms per

gE MEZ

(1)

∆w MGroup

(2)

δ) δ)

fluorosilyl group, ∆w is the weight loss, and MGroup is the molecular weight of the fluorosilyl group. Fluorosilyl groups/nm2 (χ) (eq 3) and fluorosilyl groups/particle (ψ) (eq 4) were calculated from Avogadro’s number (AN), the specific surface area (S), and the average particle radius (r). Sonicating the particles prior to and/ or during the reaction did not enhance surface coverage. The (13) Michael, G.; Ferch, H. Technical Bulletin Pigments, Basic Characteristics of AEROSIL; Degussa Corporation: Akron, OH, 1997. (14) Apfel, M. A.; Finkelmann, H.; Janini, G. M.; Laub, R. J.; Lu¨hmann, B.-H.; Price, A.; Roberts, W. L.; Shaw, T. J.; Smith, C. A. Anal. Chem. 1985, 57, 651.

4818

Langmuir, Vol. 21, No. 11, 2005

Visintin et al.

Table 1. Characterization of the Various Fluorinated Groups on Silica and Alumina as Determined by Fluorine Elemental Analysis modified particle abbrev.

F elemental analysis (%)

graft density δ (mmol/g)

no. of groups χ (per nm2)

no. of groups ψ (per particle)

7-nm C8F17-silica (4a) 7-nm C4F9-silica (4b) 7-nm CF3-silica (4c) 12-nm C8F17-silica (5a) 12-nm C4F9-silica (5b) 12-nm CF3-silica (5c) 8-14-nm C8F17-alumina (6a) 8-14-nm C8F17-alumina (6b) 8-14-nm C8F17COOH-alumina (6c)

17.97 10.19 3.70 14.25 7.58 3.25 11.74 19.00 41.57

0.56 0.60 0.65 0.44 0.44 0.57 0.36 0.59 1.3

1.1 1.2 1.3 1.3 1.3 1.7 0.88 1.4 3.1

1.7 × 102 1.8 × 102 2.0 × 102 6.0 × 102 6.0 × 102 7.8 × 102 3.3 × 102 5.4 × 102 1.2 × 103

Table 2. Characterization of the Various Fluorinated Groups on Silica and Alumina as Determined by TGA modified particle abbrev.

weight loss (%)

graft density δ (mmol/g)

no. of groups χ (per nm2)

no. of groups ψ (per particle)

7-nm C8F17-silica (4a) 7-nm C4F9-silica (4b) 7-nm CF3-silica (4c) 12-nm C8F17-silica (5a) 12-nm C4F9-silica (5b) 12-nm CF3-silica (5c) 8-14-nm C8F17-alumina (6a) 8-14-nm C8F17-alumina (6b) 8-14-nm C8F17COOH-alumina (6c)

25.81 18.51 10.86 21.10 13.12 9.30 19.21 31.40 59.38

0.49 0.58 0.63 0.40 0.41 0.54 0.36 0.60 1.3

1.0 1.2 1.3 1.2 1.2 1.6 0.88 1.4 3.1

1.5 × 102 1.8 × 102 2.0 × 102 5.5 × 102 5.6 × 102 7.4 × 102 3.3 × 102 5.5 × 102 1.2 × 103

results of the modified silica particles 4-5 are shown in Tables 1 and 2. The yields, which were based on 2.5 SisOH/nm2, were approximately 50%.

χ)

δAN S

ψ ) χ(4πr2)

(3) (4)

Preparation of C8F17-Alumina via Silanization/Hydrosilylation (6a). The silicon hydride-modified alumina intermediate was prepared according to a procedure developed by Pesek et al.16 with minor modifications. Dried alumina particles (1.0 g, 9.80 mmol) and 30 mL of p-dioxane were stirred at 80 °C in a 100-mL three-neck round-bottom flask equipped with a reflux condenser. To this suspension, 1 mL of 3.1 M HCl and then 4.60 mL (25.1 mmol) of TES (dropwise) were added. This mixture was stirred at 80 °C for 5 h. After this time, the silicon hydridemodified alumina particles were centrifuged, washed with THF/ water (80:20) (2 × 30 mL), THF (2 × 30 mL), and diethyl ether (2 × 30 mL), and then dried under vacuum for 24 h at 110 °C. These particles were stored in an oven (125 °C) until needed: IR (KBr): 3451(br) ν(OH), 2259(s) ν(Si-H), 1635(m), 1143(vs), 1073(vs), 835(vs) ν(AlO), 748(w), 634(w) cm-1. As a control study to ensure that the TES did not selfpolymerize, 4.6 mL of TES, 30 mL of p-dioxane, and 1 mL of 3.1 M HCl were stirred and heated at 80 °C for 5 h. A homogeneous colorless solution (no polymer formation) was observed throughout the 5 h duration. The silicon hydride-modified alumina intermediate was derivatized via hydrosilylation using a terminal olefin in the presence of a catalyst on the basis of the procedure by Pesek et al.16 1H,1H,2H-perfluoro-1-decene (15 mL) and 100 µL of 50 mM [Cp2]PtCl2 in dry chloroform were placed in a 100-mL threenecked round-bottom flask equipped with a reflux condenser. This mixture was heated to 70 °C and stirred until a clear solution was obtained (approximately 1.5 h). The hydride-modified alumina particles (0.5 g) were slowly added to the olefin/catalyst solution over a 1 h period, and then the temperature was raised to 85 °C. The reaction mixture was stirred at 85 °C for 96 h. After this time, the derivatized alumina particles were centrifuged, washed with toluene (4 × 30 mL), CH2Cl2 (2 × 30 mL), and diethyl ether (2 × 30 mL), and then dried under vacuum at 110 (15) Fadeev, A. Y.; Eroshenko, V. A. J. Colloid Interface Sci. 1997, 187, 275. (16) Pesek, J. J.; Sandoval, J. E.; Minggong, S. J. Chromatogr. 1993, 630, 95.

°C overnight. Surface coverage data (eqs 1-4) as determined by elemental analysis and TGA results are shown in Tables 1 and 2. IR (KBr): 3448(br) ν(OH), 2963(w) ν(CH), 2925(w) ν(CH), 2958(w) ν(CH), 2259(s) ν(Si-H), 1628(m), 1445(w), 1372(w), 1321(w), 1245(w) ν(CF), 1208(w) ν(CF), 1149(s) ν(CF), 1115(w) ν(CF), 1070(s), 835(vs) ν(AlO), 748(w), 659(w) cm-1. Preparation of C8F17-Alumina via Silanization (6b). Fluorinated silanes were covalently attached to alumina surfaces on the basis of the procedure by Rong et al.17 Alumina particles (1.0 g, 9.80 mmol) and 95% EtOH (50 mL) were placed in a 100mL three-necked round-bottom flask equipped with a reflux condenser and stir bar. Rf17TES (1.75 mL, 4.04 mmol) was then added dropwise over a 15 min period, and the mixture was refluxed at 80 °C for 4 h. After this time, the derivatized alumina particles were centrifuged, extracted with absolute EtOH for at least 12 h to remove the excess fluorinated silane adsorbed onto the particles, and then dried under vacuum at 80 °C overnight. Surface coverage data (eqs 1-4) as determined by elemental analysis and TGA results are shown in Tables 1 and 2. IR (KBr): 3448(br) ν(OH), 2963(w) ν(CH), 2925(w) ν(CH), 2958(w) ν(CH), 1628(m), 1445(w), 1372(w), 1321(w), 1242(s) ν(CF), 1207(vs) ν(CF), 1153(s) ν(CF), 1115(m) ν(CF), 1069(m), 1022(w), 816(vs) ν(AlO), 706(w), 659(vs) cm-1. As a control study to ensure that the Rf17TES did not selfpolymerize, 1.75 mL of Rf17TES and 50 mL of 95% EtOH were stirred and heated at 80 °C for 4 h. A homogeneous, colorless solution (no polymer formation) was observed throughout the 4 h duration. Preparation of 7-nm C8F17-Silica via Silanization. The procedure for the modification of the 7-nm silica particles (0.59 g, 9.82 mmol) with Rf17TES (1.75 mL, 4.04 mmol) is identical to that used for the modification of alumina particles 6b: Anal. Calcd: 1.78% F, δ ) 0.055 mmol/g, χ ) 0.11 tails/nm2, ψ ) 17 tails/particle (eqs 1-4). Preparation of Alumina with Adsorbed Perfluorononanoic Acids (6c). Perfluorononanoic acid (1.14 g, 2.45 mmol) and HPLC water (35 mL) were placed in a 100-mL three-necked round-bottom flask. This mixture was stirred and slightly heated until a thick, homogeneous solution formed. Hydrated alumina (1.00 g, 4.90 mmol) was added, and the flask was then connected to a condenser. The mixture was vigorously stirred at 55 °C for 96 h. After this time, the alumina particles were centrifuged, washed with absolute EtOH (2 × 30 mL), THF (2 × 30 mL), and diethyl ether (30 mL), and then dried under vacuum at 80 °C overnight. Surface coverage data (eqs 1-4) as determined by (17) Rong, M. Z.; Ji, Q. L.; Zhang, M. Q.; Friedrich, K. Eur. Polym. J. 2002, 38, 1573.

Functionalization of Silica and Alumina Particles

Langmuir, Vol. 21, No. 11, 2005 4819

elemental analysis and TGA results are shown in Tables 1 and 2. IR (KBr): 3444(br) ν(OH), 1670(vs) ν(CdO), 1480(m), 1416(w), 1371(m), 1328(w), 1241(sh) ν(CF), 1207(vs) ν(CF), 1151(s) ν(CF), 1119(m) ν(CF), 1039(m), 990(w), 978(w), 818(vs) ν(AlO), 656(vs), 565(vs) cm-1. Silica Turbidimetry Studies. On the basis of work by Johnston and co-workers, we chose turbidimetry as the method to assess the stability of our modified silica and alumina slurries in CO2.7-9,18 Turbidity (τ), defined as the attenuation of light caused by scattering, was calculated from the path length (L), the initial intensity (I0), and the measured intensity (I), according to eq 5. The turbidity measurements were collected in a 2.5-mL

()

1 I0 τ ) ln L I

(5)

0.85, and 0.91 g/mL), yielding the same results within experimental error.

Results and Discussion Syntheses of Derivatized Silica and Alumina Particles. Silica and alumina particles are the most commonly used abrasives in current CMP processes, and therefore, it seemed logical to start with those particles to develop a CO2-based CMP process. To form stable dispersions in condensed CO2, silica and alumina particles must be sterically stabilized by groups bound to the surface since electrostatic stabilization processes are ineffective in such low-dielectric media. Unmodified silica and alumina particles were first studied in condensed CO2. At CO2 densities from 0.64 to 0.91 g/mL, the unmodified particle suspensions did not form stable colloidal dispersions in CO2, as determined visually (Figure S1a and b) and by turbidity measurements. These particles completely settled within a few seconds once the stirring was stopped due to the attractive van der Waals interactions between particles. Since it is well known that fluorinated compounds are highly soluble in CO2, covalently attaching fluorinated ligands to the particles should enhance the particlesolvent interaction if the solvent quality is sufficient, and thereby promote dispersion of the particles in CO2. In working toward an industrially viable CO2-based CMP process, it is preferable that the stabilizing ligands are readily available and inexpensive and that the particle modification procedure is relatively facile, reproducible, and gives high yields of derivatized particles. Additionally, it is essential that the stabilizing ligands be short enough to ensure that the abrasive properties of the particles are not lost. The functionalization of silica surfaces with chlorosilanes has been well-established.15,20-24 Monochlorosilanes 1-3 were covalently attached to the surface silanol groups (Sis denotes a surface silicon atom) of the 12- and 7-nm silica samples through a base-catalyzed condensation reaction (eq 6).15 This procedure yielded silica derivatives 4a, 4b, 4c, 5a, 5b, and 5c (Figure 1), which allowed a

316 stainless steel high-pressure cell equipped with two 1-in.diameter × 5/8-in.-thick sapphire windows with a path length of 1.0 cm. The incident beam of a He-Ne laser (Uniphase, λ ) 633 nm) was passed through the high-pressure view cell and detected with a LaserPAD semiconductor sensor (Coherent). To investigate the steric stabilization of the particles, the turbidity was measured as a function of time at the fixed wavelength. The turbidity-time profiles were collected at various CO2 densities and temperatures. The CO2 densities investigated were 0.64, 0.71, 0.79, 0.85, and 0.91 g/mL. Turbidity measurements were recorded at 25, 35, 40, and 50 °C beginning with the lowest density. The physical properties of the particle/CO2 suspensions were assumed to be equivalent to that of pure CO2. The CO2 density was obtained from the database provided by the National Institute of Standards and Technology.19 The particles (0.005 g, 0.2% w/v) were placed in the cell containing a Teflon-coated stir bar, and the cell was charged with the desired pressure of CO2. To determine how much time was needed to reach equilibrium conditions for the particles stabilized in CO2, turbidity measurements were conducted at 2 and 16 h. No difference in the turbidity-time profiles was observed between the 2 and 16 h data; therefore, the particles were stirred for 2 h prior to taking the initial measurement. All subsequent measurements were stirred for 5-10 min prior to taking measurements since the system was already at equilibrium, and data taken at 5, 10, and 20 min gave the same result. After stirring for the specified time, the stirring was stopped and the turbidity measurements were immediately started. At each temperature/density, the turbidity was measured every 5 s for 20 min. The 20-min measurement period was an adequate amount of time to obtain the initial slope of the turbidity-time profile. There is not a significant difference in the measured turbidimetry at 20 min and 1 h, indicative of the fact that the sterically stabilized particles only slowly sediment over time. The dispersion stability (defined as dτ/dt) was then determined from the initial slope of the turbidity-time profile.7-9,18 The closer dτ/dt is to 0, the more stable the suspension. Each temperature/ density measurement was performed in duplicate, and the slopes were averaged. To assess the reproducibility of the measurements, measurements for particles 4a were repeated several weeks later at 35 °C (d ) 0.64, 0.71, 0.79, 0.85, and 0.91 g/mL), yielding the same results within experimental error. Alumina Turbidimetry Studies. The same general procedure and amount of particles were used as described above. The only difference is that the particles were stirred in CO2 for 16 h prior to taking the initial measurement, as it was determined that at least 12 h was needed to reach equilibrium conditions of the particles stabilized in CO2. As the system was already at equilibrium, all subsequent measurements only had to be stirred for 5-10 min prior to taking measurements, as the data at 5, 10, and 20 min gave the same result. To assess the reproducibility of the measurements, measurements for particles 6b and 6c were repeated several weeks later at 35 °C (d ) 0.64, 0.71, 0.79,

homologous series of fluorinated silyl chain lengths to be studied. Fluorine elemental analysis and TGA on the modified particles were used to calculate surface coverage (Tables 1 and 2). The results obtained from TGA are very similar to those from elemental analysis. At a given mass, the 7-nm silica particles have a higher graft density (tails/g particle) than the 12-nm silica particles due to the higher surface area per gram for the smaller particles (Figure 2). On a per-particle basis, the 7-nm silica particles have fewer tails per particle than the 12-nm silica particles (Figure 2); with a coverage of 2.5 SisOH/nm2, the 7- and 12-nm particles would theoretically have 385 and 1130 SisOH/ particle, respectively. Since the modified silica studies indicate that the -C8F17 tails are most stable to flocculation in CO2 at moderate solvent densities, this fluorinated tail length was chosen as the common stabilizer to compare modified alumina to modified silica. The analogous reaction of alumina par-

(18) O’Neill, M. L.; Yates, M. Z.; Harrison, K. L.; Johnston, K. P.; Canelas, D. A.; Betts, D. E.; DeSimone, J. M.; Wilkinson, S. P. Macromolecules 1997, 30, 5050. (19) Lemmon, E. W.; McLinden, M. O.; Friend, D. G. Thermophysical Properties of Fluid Systems. In NIST Chemistry WebBook, NIST Standard Reference Database Number 69. http://webbook.nist.gov.

(20) Berendsen, G. E.; Pikaart, K. A.; de Galan, L. Anal. Chem. 1980, 52, 1990. (21) Ulman, A. Chem. Rev. 1996, 96, 1533. (22) Fadeev, A. Y.; McCarthy, T. J. Langmuir 1999, 15, 3759. (23) Fadeev, A. Y.; McCarthy, T. J. Langmuir 2000, 16, 7268. (24) Cao, C.; Fadeev, A. Y.; McCarthy, T. J. Langmuir 2001, 17, 757.

SisOH + CnF2n+1CH2CH2Si(CH3)2Cl f CnF2n+1CH2CH2Si(CH3)2-OSis + HCl (6)

4820

Langmuir, Vol. 21, No. 11, 2005

Visintin et al.

Figure 2. Graft density (solid line) and total number of groups per particle (dashed line) as a function of fluorosilyl chain length for the 7- (0) and 12-nm (b) modified silica particles. The graph was plotted according to the elemental analyses results.

Figure 3. Comparison of the graft density (solid line) and total number of groups per particle (dashed line) among 4a, 5a, 6b, and 6c. The graph was plotted according to the elemental analyses results.

ticles with a chlorosilane is unsuccessful because AlsO-Si bonds are unstable under acidic conditions (Als denotes a surface aluminum atom).25 It has been shown that alumina may be adequately modified through a twostep silanization/hydrosilylation reaction scheme.16,26 Following this method, silicon hydride-modified alumina particles (diameter of 8-14 nm) (eq 7) were synthesized. IR spectroscopy and TGA data confirm that the silanization process was successful. The IR spectrum shows a strong Si-H stretch at ∼2260 cm-1 (Figure S2). The TGA curve (in air) exhibits an initial weight gain at approximately 440 °C and the DSC trace (in air) shows a single exothermic peak at 545 °C (Figure S3), indicative of the thermooxidative degradation of SiH groups into SiOH groups.27,28 The lack of a second exothermic peak in the DSC trace shows that the TES did not self-polymerize (polyhydrosiloxane gives a sharp peak around 375 °C).16,26 The silicon hydride alumina particles were further modified by the catalytic hydrosilylation of C8F17CHdCH2 (eq 8) to give 8-14-nm C8F17-alumina (6a). IR spectroscopy, TGA, and fluorine elemental analysis confirm that the hydrosilylation reaction was successful. The IR spectrum of 6a has the expected C-H stretching bands between 2858 and 2963 cm-1, C-F stretches at 1115, 1208, and 1245 cm-1, and a Si-H band of reduced intensity (Figure S2). The graft density of 6a as determined by elemental analysis was 0.365 mmol/g (Table 1). This value is significantly lower than that of the silica particles 4a

determined by elemental analysis is 0.588 mmol/g (Table 1), which is similar to 4a (Figure 3).

(0.556 mmol/g) and 5a (0.441 mmol/g). It has been reported that silicon hydride-modified alumina particles catalytically bonded with C6H13CHdCH2 and C16H33-CHdCH2 give graft densities of 0.269 and 0.216 mmol/g, respectively.16 This implies that the graft density of 6a is within the typical range. Particles 6a were found to be unstable at CO2 densities from 0.66 to 0.91 g/mL as determined visually. For this reason, a direct silane coupling methodology was explored to increase the graft density. Organosilanization of the alumina particles was carried out by reacting Rf17TES with the alumina surfaces (eq 9) to give 8-14-nm C8F17alumina (6b). Strong stretching bands at 1242, 1207, 1153, and 1115 cm-1 in the IR spectrum readily identify the C-F groups (Figure S4). The graft density of 6b as (25) Pesek, J. J.; Matyska, M. T. J. Chromatogr. A 2002, 952, 1.

A second series of particles was synthesized to enable the difference in stability of alumina particles with tails covalently bonded to the surface and tails hydrogenbonded to the surface to be investigated. It is well-known that carboxylic acids strongly adsorb onto alumina surfaces through hydrogen-bonding with AlsOH.29-32 Chemisorption of perfluorononanoic acids onto the alumina particles surfaces gave 8-14-nm C8F17COOH-alumina (6c) (Figure 1). The IR spectrum showed a sharp band at 1670 cm-1, which was assigned to the CdO stretch for the perfluorononanoic acid groups (Figure S7). This band was shifted 88 cm-1 to lower energy from the ν(CdO) stretch in uncoordinated perfluorononanoic acid (1758 cm-1), indicative of chemisorption. The graft density of 6c as determined by elemental analysis was 1.29 mmol/g (Table 1), which indicates a significantly greater ligand surface coverage than for 4a, 5a, and 6b (Figure 3). Turbidimetry Measurements. Turbidimetry was used to quantitatively determine the steric stabilization of the silica and alumina particle suspensions in condensed CO2. The solvent conditions, choice of stabilizer (tail), and synthetic mode of attaching the tails onto the surfaces influence this stabilization, and are discussed below. Derivatized Silica Particles. Unmodified silica particles in condensed CO2 provide a baseline for the stabilities of 4-5. Particles 4a and 5a are both well dispersed in scCO2 (40 °C, 214 bar, 0.4% w/v) but gave different appearances (Figure S1c and d). Suspensions 4a (smaller particles) and 5a (larger particles) in CO2 media were translucent and opaque, respectively. Visual observations through a view cell indicated that stabilization in CO2 (T g 40 °C, CO2 densities g 0.85 g/mL) was continuous up to 24 h for 4a despite the low-molecular-weight surface groups. Turbidity measurements were conducted on the particle suspensions. Figure 4 depicts the suspension turbidity over time at 55 °C and CO2 densities from 0.64 to 0.91 g/mL for 4a. Turbidity-time profiles for 5a at 55 °C are (26) Pesek, J. J.; Tang, V. H. Chromatographia 1994, 39, 649. (27) Sandoval, J. E.; Pesek, J. J. Anal. Chem. 1989, 61, 2067. (28) Chu, C.-H.; Jonsson, E.; Auvinen, M.; Pesek, J. J.; Sandoval, J. E. Anal. Chem. 1993, 65, 808. (29) Laibinis, P. E.; Hickman, J. J.; Wrighton, M. S.; Whitesides, G. M. Science 1989, 245, 845. (30) Haky, J. E.; Blauvelt, T. M.; Wieserman, L. F. J. Liq. Chromatogr. 1996, 19, 307. (31) Mao, Y.; Fung, B. M. J. Colloid Interface Sci. 1997, 191, 216. (32) Mao, Y.; Fung, B. M. Chem. Mater. 1998, 10, 509.

Functionalization of Silica and Alumina Particles

Figure 4. Turbidity-time profiles for 4a in CO2 at 55 °C. The unmodified silica suspension (dashed line) in CO2 (55 °C, 0.91 g/mL) and an ideal stability curve (O) are shown for comparison.

Langmuir, Vol. 21, No. 11, 2005 4821

Figure 6. Turbidity-time profiles for 4a (solid line) and 5a (dashed line) in CO2 at 35 °C. An ideal stability curve (O) is shown for comparison.

Figure 5. Steric stabilization (dτ/dt) versus CO2 density for 4a in CO2 at 55, 40, 35, and 25 °C.

nearly identical. As the solvating power was increased by increasing the CO2 density, the modified silica suspensions became more stable. A similar trend was observed at 40, 35, and 25 °C for both the 7-nm (4a, Figure 5) and 12-nm (5a, Figure S5) modified silica particles. Figure 5 also demonstrates that the modified particles were more stable at low CO2 densities when the temperature (T) was increased. At T g 35 °C, 4a was stable (dτ/dt ≈ -0.038 cm-1 min-1) at CO2 densities g 0.79 g/mL, and at 25 °C, they were stable at CO2 densities g 0.85 g/mL. In comparison, 5a was only stable at a CO2 density of 0.79 g/mL at high temperatures (g55 °C), and at 25 °C, they were only stable at high CO2 densities (g0.91 g/mL). Particles 4-5 were not stable (dτ/dt g -1.4 cm-1 min-1) at CO2 densities e 0.71 g/mL. These stability trends correlate well with those reported by Johnston and coworkers for the stabilization of silica particles with highmolecular-weight fluorinated surfactants adsorbed to the surface or fluoropolymers covalently attached to the surface; however, stabilization of these systems in the liquid CO2 regime (T < 31 °C) was not reported.7,9 Notable differences were observed between the 7- and 12-nm modified silica particles. The smaller particles modified with fluorinated ligands were more stable in condensed CO2 than the corresponding larger particles. First, 5a was not as stable as 4a at temperatures e 40 °C. This is clearly demonstrated in Figure 6 at 35 °C. Second, 5b was not stable in CO2 up to 0.91 g/mL and 55 °C (Figure S6), whereas 4b was stable at 55 °C and CO2 densities g 0.85 g/mL, as well as at 40 °C and CO2 densities g 0.91 g/mL (Figure 7). At 40 °C and 0.85 g/mL, turbidity measurements show that 4b slowly settled over time after the stirring was stopped. Stabilization in CO2 (T ) 55 °C, CO2 density ) 0.91 g/mL) was found to be continuous up to 24 h for 4b on the basis of visual observation. The lack of steric stabilization in CO2 of 5b compared to the moderate stability of 4b may be explained on the

Figure 7. Comparison of stabilities of 7-nm CnF2n+1-silica (n ) 8, 4, or 1) suspensions in CO2 at (a) 55 °C and 0.91 g/mL, (b) 55 °C and 0.85 g/mL, (c) 40 °C and 0.91 g/mL, and (d) 40 °C and 0.85 g/mL.

basis of theoretical predictions.33,34 As described by Johnston and co-workers,8,9 the propensity to flocculate is proportional to the radius of the particle. As the radius of a particle increases, the particle-particle separation distance required to overcome the van der Waals attraction increases because the contact surface area increases. Therefore, flocculation in CO2 for the 12-nm particles is expected at larger particle-particle separation distances than the 7-nm particles. As a result, the 12-nm particles must be modified with longer fluorinated tails (i.e., -C8F17) to be stable in CO2. The longer fluorinated tails better separate the particles in CO2, thereby strengthening the particle-solvent interaction and eliminating flocculation. Both the 7- and 12-nm silica particles modified with -CF3 tails (4c and 5c, respectively) were found to be unstable in CO2 under all the conditions studied. The -CF3 chains apparently are simply not long enough to provide a steric barrier to aggregation, and consequently, the particles settled within seconds in CO2. We conclude that silica particle stability in CO2 is highly dependent on the covalently attached fluorinated tail length. Stability increases in the order CF3-silica , C4F9-silica < C8F17silica. Derivatized Alumina Particles. Unmodified alumina particles in condensed CO2 provided a baseline for measuring the stabilities of 6b and 6c. After 6b was stirred in scCO2 (40 °C, 214 bar, 0.4% w/v) for 12 h, the suspension was virtually transparent (Figure S1e). Turbidity measurements were conducted on 6b and 6c. As observed (33) Israelachvili, J. Intermolecular Forces & Surface Forces. 2nd ed.; Academic Press: San Diego, 1992. (34) Hiemenz, P. C.; Rajagopalan, R. Principles of Colloid and Surface Chemistry, 3rd ed.; Marcel Dekker: New York, 1997.

4822

Langmuir, Vol. 21, No. 11, 2005

Figure 8. Turbidity-time profiles for 6b in CO2 at 35 °C. The unmodified alumina suspension (dashed line) in CO2 (55 °C, 0.91 g/mL) and an ideal stability curve (O) are shown for comparison.

Figure 9. Steric stabilization (dτ/dt) versus CO2 density for 6b and 6c compared to 4a in CO2 at 35 °C. The modified alumina particles are significantly more stable than the modified silica particles.

Figure 10. Turbidity-time profiles for 6c in CO2 at 35 °C. The unmodified alumina suspension (dashed line) in CO2 (55 °C, 0.91 g/mL) and an ideal stability curve (O) are shown for comparison.

with the modified silica particles, these particles were also stable to lower CO2 densities at higher temperatures. Figure 8 shows that 6b is stable (dτ/dt ≈ -0.0027 cm-1 min-1) in scCO2 at T g 35 °C and CO2 densities g 0.71 g/mL. In the liquid CO2 regime (25 °C), these particles were stable at CO2 densities g 0.79 g/mL. Visual observations through a view cell indicated that stabilization in CO2 (T g 40 °C, CO2 densities g 0.85 g/mL) was continuous up to 96 h for 6b. The stabilization of 6b in CO2 was significantly better at lower CO2 densities than any of the modified silica particles reported elsewhere,7-9 as well as in our studies (Figure 9). Turbidity measurements demonstrated that 6c was not as stable in CO2 as 6b (Figures 9 and 10). Particles 6c were stable (dτ/dt ≈ -0.023 cm-1 min-1) in scCO2 at T > 35 °C and CO2 densities g 0.85 g/mL (versus 0.71 g/mL as seen with 6b). In the liquid CO2 regime (25 °C), 6c was stable at CO2 densities g 0.85 g/mL (versus 0.79 g/mL as seen with 6b). It may be concluded that tails covalently linked to the surface better stabilize the particles in CO2 than hydrogen-bonded tails.

Visintin et al.

Factors Influencing Particle Steric Stabilization in Condensed CO2. Turbidimetry studies establish that the alumina particle suspensions in condensed CO2 are significantly more stable than the silica particle suspensions. In a given CO2 medium with identical fluorinated tail lengths, differences in the stability of dispersions between the silica (4a and 5a) and alumina (6b) particles may arise from (i) the size of the particles, (ii) a difference in the refractive index of the particles, (iii) the graft density (total tails/g particle), and/or (iv) the mode of attaching tails to the surfaces. Since the sizes of the silica (average diameters of 7 and 12 nm) and alumina (diameter of 8-14 nm) particles used in our studies were approximately the same, the difference in stability of the silica and alumna particles is not likely to originate from (i). A large difference in refractive indexes of the particles could obscure the comparison between particles of a different type. The specific refractive index increment (dn/ dc), typically 0.1 mL/g, was determined using eq 10, where n is the refractive index and c is the inorganic particle density.35 The refractive indexes for silica, alumina, and CO2 (40 °C, 215 bar, 0.85 g/mL) are 1.45, 1.7, and 1.2004 g/mL, respectively, and the densities of silica and alumina are 2.19 and 3.7 g/mL, respectively.13,36,37 The dn/dc values for silica and alumina were calculated to be 0.11 and 0.13 mL/g, respectively. Given the similarities of the dn/dc values, it is unlikely that (ii) is the origin of the difference in the measured turbidimetry behavior.

dn nparticle - nsolvent ) dc dparticle

(10)

Higher graft densities on the particles play a role in dictating the stability of particle dispersions in CO2. As shown in Tables 1 and 2, 6b has a graft density that is approximately 11% and 29% higher than 4a and 5a, respectively. Therefore possibility (iii) likely contributes to the greater steric stabilization of 6b compared to 4a and 5a in condensed CO2. A major difference between the derivatized alumina and silica surfaces is the nature of the bonding of the stabilizing tails to the surface. Monofunctional organosilanes (one attachment site per tail) used for the derivatization of silica particles form only a single bond to the surface per tail. The trifunctional organosilane (three attachment sites per tail) used for the derivatization of alumina particles may give rise to several surface structures arising from the possibility of one or more interactions of a tail with the particle surface or with other tails.23 The difference in the structure of the tails bound to the surface between the modified silica particles (4a and 5a) and the modified alumina particles (6b) is a likely factor influencing particle steric stabilization in condensed CO2. To investigate if the difference in stability of the modified silica particles (4a and 5a) and the modified alumina particles (6b) in CO2 was due to the different structures on the surface (iv), an attempt was made to derivatize 7-nm silica particles with Rf17TES using a procedure identical to that for 6b. Unfortunately, the graft density for these derivatized particles as determined by elemental analysis was only 0.055 mmol/g and these particles were not stable in condensed CO2. The low surface coverage of the silica particles in comparison to the alumina particles (35) Huglin, M. B. Light Scattering from Polymer Solutions; Huglin, M. B., Ed.; Academic Press: New York, 1972. (36) CRC Handbook of Chemistry and Physics, 74th ed.; Lide, D. R., Ed.; CRC Press: Boca Raton, FL, 1993. (37) Burns, R. C.; Graham, C.; Weller, A. R. M. Mol. Phys. 1986, 59, 41.

Functionalization of Silica and Alumina Particles

did not allow us to draw any conclusions regarding the role of surface structure on particle stability. Conclusions The steric stabilization of silica and alumina particles modified with low-molecular-weight fluorinated groups has been determined by turbidimetry. Fluorinated tail length, temperature, and CO2 density play a major role in stabilizing the particles. Silica particles modified with -C8F17 segments (4a and 5a) are stable to flocculation at moderate solvent densities. Higher temperatures improve the CO2 solvent strength, leading to stable particle suspensions at lower solvent densities. Modified silica, 4a, and alumina, 6b, particles give the most stable dispersions in CO2 and, therefore, are likely candidates in a CO2-based CMP slurry. The windows of operability for 4a in the supercritical and liquid CO2 regimes are T > 35 °C, CO2 densities g 0.79 g/mL and 25 °C, CO2 densities g 0.85 g/mL, respectively. The windows of operability for 6b in the supercritical and liquid CO2

Langmuir, Vol. 21, No. 11, 2005 4823

regimes are T g 35 °C, CO2 density g 0.71 g/mL and 25 °C, CO2 densities g 0.79 g/mL, respectively. These studies on abrasive particle stabilization and flocculation in condensed CO2 are important to identify candidates for abrasive particles in a CO2-based CMP process. Acknowledgment. We acknowledge the Kenan Center for the Utilization of CO2 in Manufacturing for financial support. This work made use of STC shared experimental facilities supported by the National Science Foundation under Agreement No. CHE-9876674. We are grateful to the Degussa Corporation for their generous contribution of the silica particles. We also thank Professor Richard J. Spontak at North Carolina State University for lending us the He-Ne laser. Supporting Information Available: Figures S1-S7 (PDF). This material is available free of charge via the Internet at http://pubs.acs.org. LA047823C