Article pubs.acs.org/IECR
Use of Multifunctional Carboxylic Acids and Hydrogen Peroxide To Improve Surface Quality and Minimize Phosphine Evolution During Chemical Mechanical Polishing of Indium Phosphide Surfaces John B. Matovu,†,‡ Patrick Ong,‡ Leonardus H. A. Leunissen,‡,§ Sitaraman Krishnan,† and S. V. Babu*,† †
Department of Chemical and Biomolecular Engineering and the Center for Advanced Materials Processing, Clarkson University, Potsdam, New York 13699, United States ‡ Interuniversity Microelectronics Center (IMEC), Kapeldreef 75, 3001 Heverlee, Belgium S Supporting Information *
ABSTRACT: During the chemical mechanical polishing (CMP) of an indium phosphide buffer layer for the fabrication of InGaAs n-channels in CMOS devices, there is interest in controlled removal of In and P atoms from the surface and avoidance of the generation of toxic phosphine (PH3) gas. We report InP removal rates and phosphine generation during InP CMP in hydrogen peroxide based silica particle dispersions in the presence and absence of three different multifunctional chelating carboxylic acids, namely oxalic acid, tartaric acid, and citric acid. The presence of these acids in the polishing slurry resulted in highly smooth surfaces (about 0.1 nm RMS surface roughness) with good InP removal rates (about 400 nm min−1) and very low phosphine generation (6 M), to our knowledge, there are no reports on the effect of multifunctional carboxylic acids on phosphine release. Peddeti et al.3,4 measured phosphine evolution during CMP of InP and provided a preliminary analysis of the reactions that lead to its generation. They measured InP removal rates using silica particle dispersions. They found that the removal rates were negligible with only pH-adjusted deionized water, that is, in the absence of the silica particles. High removal rates and high phosphine evolution were observed with silica slurries at acidic pH. The removal rates and phosphine generation were both low under alkaline conditions, despite high silica particle concentration in the slurry (10 wt %). These experiments clearly demonstrated a combined effect of chemical dissolution and mechanical abrasion on InP removal. On the basis of these results, it would appear that the polishing of InP in alkaline pH would be a more attractive choice because no phosphine is formed. However, recently, Waldron et al.19 observed that an alkaline slurry that is normally used to polish polysilicon layers resulted in poor surface quality of InP in the STI structures. On the other hand, an acidic slurry (that was previously used to polish W) resulted in better surface quality. Therefore, there is still a need to develop slurries that not only minimize phosphine release but also provide good surface quality. Peddeti et al.3,4 also found that the addition of hydrogen peroxide to the silica-based slurries lowered phosphine generation, but at the expense of lower InP removal rates. It is expected that the addition of multifunctional carboxylic acids to these slurries would increase InP removal rates because of chelation of indium by these acids, as proposed by Bandaru and Yablonovitch,13 but not increase phosphine generation. In a study on InP CMP using a combination of an α-hydroxy acid (citric acid) and an oxidant (sodium hypochlorite, NaOCl), Morisawa et al.20 found that, under acidic pH conditions, InP removal rates increased with an increase in the concentration of NaOCl. However, the dissolution was accompanied by an evolution of potentially toxic chlorine gas. To our knowledge, there are few reports on slurry formulations for InP chemical mechanical polishing that yield good InP removal rates and smooth surfaces, without evolution of toxic gases. Hence, we focus here on minimizing or even completely suppressing phosphine release while maintaining good InP removal rates and surface quality, at a relatively low silica particle concentration in the polishing slurry (3 wt %), the latter to minimize defects and ease post-CMP cleaning. Therefore, we investigated the effect of adding multifunctional carboxylic acids, namely oxalic, tartaric, and citric acids, to the hydrogen peroxide/silica based slurries on InP removal rates and phosphine evolution. Of these acids, citric and tartaric acids
2. EXPERIMENTAL METHODS 2.1. InP Chemical Mechanical Polishing. InP blanket coupons (1.8 × 1.8 cm2), diced from an 8-in. wafer consisting of a 1-μm thick InP film grown on a 1-μm thick Ge seed layer on a Si substrate, were polished on an Alpsitec E460 Mecapol Polisher. The process for growing this InP film stack has been discussed by Wang et al.,1 and more details can be found there. Dispersions of colloidal silica abrasives, Nexsil 35A (30 wt % in water), with average particle diameter of 35 nm, and of fumed silica, Semi-Sperse 12E (SS12, 12 wt % in water), were obtained from Nyacol Nanotechnologies and Cabot Microelectronics, respectively. Hydrogen peroxide solution (30 wt % in water), and the carboxylic acids, oxalic acid (anhydrous, CAS no. 144-62-6, MW 90.03 g mol−1, ≥97.0%), tartaric acid (CAS no. 89-69-4, MW 150.09 g mol−1, ≥99.0%), and citric acid (CAS no. 77-92-9, MW 192.12 g mol−1, ≥99.5%), purchased from Sigma-Aldrich, were used as additives in the polishing slurry, without further purification. The structures of the carboxylic acids are shown in Figure 1. Deionized (DI) water was used to prepare the slurries.
Figure 1. Chemical structures of the carboxylic acids that were used in this study.
As a representative example, the polishing slurry was prepared by mixing 100 g of Nexsil 35A colloidal silica dispersion, 33.33 g of hydrogen peroxide solution, 866.67 g of DI water, and 7.20 g of oxalic acid using a magnetic stirrer for about 5 min. The pH was adjusted using nitric acid or potassium hydroxide as necessary, and the slurry was used immediately for CMP experiments thereafter. The concentrations of Nexsil 35A, hydrogen peroxide, and oxalic acid in the slurry were 3 wt %, 0.3 mol dm−3 (∼1 wt % of the slurry), and 0.08 mol dm−3 (∼0.72 wt % of the slurry), respectively. The molar concentrations of hydrogen peroxide and oxalic are based on the volume of the aqueous phase. The polishing was performed at 24.1 kPa (3.5 psi) operating pressure, 80/72 rpm carrier/platen rotation speed, and 200 mL min−1 slurry flow rate. The coupons were polished for 30 s using an IC1000 K-groove pad. The pad was conditioned ex situ using a diamond grit conditioner for 30 s before each run for all the experiments. InP removal rates were calculated from the difference in mass, determined using a Metryx Mentor balance with a precision of 1.0 μg, before and after polishing. The polished coupons were dried in a nitrogen stream before weighing. A density of 4.81 g cm−3 was used to convert InP gravimetric removal rates to the rate of reduction in thickness.21 The reported rates were obtained after averaging over 3 experiments for each process condition. 10665
dx.doi.org/10.1021/ie400689q | Ind. Eng. Chem. Res. 2013, 52, 10664−10672
Industrial & Engineering Chemistry Research
Article
2.2. Phosphine Measurements. Phosphine concentrations in the air above the InP substrates were measured using a Zellweger MDA Scientific CM4 four point continuous gas monitor (Honeywell) by connecting a gas collection tube to one of its four sampling ports. The headspace gas was sampled at a location 3 cm above the pad and 5 cm from the center of the wafer. The flow system of the instrument included a pump that sampled gas at a constant rate of 720 cm3 min−1 (at standard ambient temperature and pressure). PH3(g) concentration was determined using the Chemcassette detection system, which optically measured the density of the stain developed by the gas on a reactive Chemcassette tape. The concentration of PH3 in the gas mixture (v/v), which is proportional to the stain density, is reported in parts-per-billion (ppb). 2.3. InP Dissolution Rates. Dissolution rates were measured at 40 °C, to approximately match the surface temperature during CMP,22 using InP blanket coupons (1.8 × 1.8 cm2) in a 500 mL beaker containing 300 mL of the test solution. The InP coupon was washed with DI water, dried in a nitrogen stream, and weighed using a Sartorius analytical balance of 0.1 mg precision. It was thereafter immersed in the test solution that was stirred using a magnetic stirrer for 1 min. The coupon was removed from the solution, rinsed with DI water, dried in a nitrogen stream, and reweighed. Each reported value is an average from three different measurements. 2.4. Surface Roughness. Surface roughness values of blanket coupons, before and after CMP, were determined using the tapping mode of a Nanoscope IVa Dimension 3100 atomic force microscope (AFM) system (Digital Instruments). Scans over 2 × 2 μm2 regions of the surface were used to determine the root-mean-square surface roughness. 2.5. Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy. Clean InP blanket coupons (1.8 × 1.8 cm2) were dipped separately in solutions containing 0.3 M hydrogen peroxide and 0.08 M oxalic, tartaric, or citric acid at pH 6 for 1 h (long time exposure was used to ensure high adsorption density of the carboxylic acids on the InP surfaces) and then in DI water for 1 min to remove any weakly bound acid species. The coupons were then dried in a nitrogen stream before FTIR measurements. A Bruker Vector 22 FTIR spectrophotometer was used to obtain the FTIR spectra using the attenuated total reflectance (ATR) mode at ambient temperature. A ZnSe crystal was used for total internal reflection of the IR beam at an angle of 45°. 2.6. X-ray Photoelectron Spectroscopy (XPS). InP blanket coupons (1.8 × 1.8 cm2) were dipped in solutions containing 0.3 M hydrogen peroxide and 0.08 M oxalic acid at pH 6 for 1 min, and then in DI water for 1 min to remove any weakly bound acid species. The surfaces were subsequently dried in a stream of nitrogen gas. XPS data were acquired using a Theta 300 system (Thermo Instruments) in the angle resolved mode, with a monochromatized Al Kα X-ray source (1486.6 eV) and a spot size of 400 μm at normal incidence. The survey and high resolution scans were performed at emission angles of 78° and 21°, measured relative to the surface normal. Curve fitting was performed using the Tougaard background and Gaussian−Lorentzian peak shape functions. Peak areas were converted to atomic concentrations using tabulated atomic sensitivity factors. XPS spectra were also acquired for the as-received InP surface. 2.7. Zeta Potential Measurements. Zeta potentials of silica particles, before and after the addition of oxalic, tartaric,
and citric acids, were measured using a Brookhaven ZetaPlus zeta potential analyzer and a palladium-coated Uzgiris electrode. Hydrogen peroxide was not added to these silica dispersions due to formation of gas bubbles on the electrode surface because of Pd-catalyzed decomposition of hydrogen peroxide.23 The pH was adjusted over the range 2−12 using nitric acid and potassium hydroxide solutions. Each reported value of zeta potential is the average of four measurements. 2.8. Surface Wettability. Water contact angles were measured using a Dataphysics OCAH230L contact angle goniometer. The contour of a water droplet on the surface was recorded by means of a CCD camera, and the contact angle was calculated using drop shape analysis.
3. RESULTS AND DISCUSSION 3.1. Dissolution Studies (Chemical Etching without Mechanical Polishing). Dissolution rates of InP were measured as a function of pH (in the range 2−12), for surfaces immersed in a 0.3 M aq solution of hydrogen peroxide, and for surfaces immersed in 0.3 M aq solutions of hydrogen peroxide containing 0.08 M of the carboxylic acid (oxalic, citric, or tartaric). No measurable dissolution rates were observed in any of these solutions, which is consistent with the findings of DeSalvo et al.9 that InP is virtually unaffected by an aq solution of citric acid and hydrogen peroxide etchant. Increasing the dissolution time did not result in a measurable increase in mass loss. The immersion of an InP blanket coupon in a stirred (1000 rpm) solution of pH-adjusted DI water (pH = 2, 4, 5, or 7) at 40 °C for 2 h still did not result in any significant dissolution. Vos et al.15 made similar observations during the etching of InP in acidified water and found that the dissolution rate was below 2 nm min−1, except in the case of concentrated HCl (37 wt %, that is, 12 M), in which the dissolution rate was more than 10 000 nm min−1, evidently due to the formation of highly water-soluble InCl3. Likewise, Cuypers et al.24 recently reported a low dissolution rate (