Analysis of In-Situ Shear and Normal Forces During Metal Cmp, Electrical Interface States of Sulfur-Passivated Sige Moscaps, and Reaction Kinetics of Pyrite Nanocrystal Purity

Abstract

The first study presented in this dissertation concerns the synthesis of pyrite (FeS2) nanocrystals (NC). Since 1980s, pyrite has been studied as a potential replacement of silicon as the absorber layer in photovoltaics. The main objective in our study has been to synthesize pyrite NCs in one stage by reacting ferrous chloride with elemental sulfur in oleylamine (OLA) without needing an additional ligand for NC growth. Results show that the synthesized pyrite NCs are cubic and 88 ± 14 nm on each side. NCs are synthesized at 200 °C, with a sulfur to iron precursor ratio of 6, and 1 hour of reaction time in OLA. We establish that the phase purity of the pyrite NCs depends on the sulfur to iron ratio. For near stoichiometric ratios, at least 24 hours of reaction time is required to achieve phase purity. This is remarkable, because, up until this study, literature indicated that there was a threshold to achieving phase purity, specifically at a ratio of 6. The rate-determining step of the synthesis reaction is shown to involve sulfur, as addition of sulfur precursor increases the rate of pyrite formation. We propose a mechanism for pyrite NC formation. FeS, H2S and polysulfides of the form Sn2- form in OLA. The slow step is the reaction between FeS and these molecules. Then, Fe2+S2- is attacked nucleophilically by H2S and Sn2, with S2- converting to S- and the remaining sulfur transferred to form pyrite Fe2+S2-. The second study presented is an investigation into the next generation transistor material, silicon germanium (SiGe). SiGe has better p-type carrier mobility than silicon, and should, theoretically, produce a better performing transistor. One challenge in implementing this material is the unstable native germanium oxides that form on its surface during processing. These oxides are electrical defects. Our goal in this study is to remove all native silicon and germanium oxides, and then passivate the germanium atoms with sulfur to prevent re-oxidation. Capacitors (MOSCAPs) from the passivated starting surface are fabricated to mimic the gate the stack in the transistor. X-ray photoelectron spectroscopy (XPS) is used extensively to monitor surface oxides on the passivated SiGe surface. Capacitance-voltage profiling is employed to monitor electrical defects from the oxides or otherwise on the SiGe surface, and bulk electrical defects within the dielectric layer grown on this surface in the process of making the full capacitor. An alloy of 25% germanium is used for all experiments in the study. Wet passivation of the germanium oxides is accomplished through aqueous solutions (non-acidic and more acidic) of ammonium sulfide. This study finds that, in non-acidic ammonium sulfide solutions, sulfur does not bond with germanium and oxides of both silicon and germanium regrow during the passivation treatments, after successfully being removed through a hydrofluoric acid/hydrochloric acid solution treatment. Acidic aqueous ammonium sulfide solutions result in sulfur bonding with germanium, but oxide regrowth is even more severe. However, any passivation treatment results in a decrease in density of electrical defects, compared to the non-passivated MOSCAPs. Bulk fixed charge oxide defects in the dielectric later are present in all MOSCAPs. The third, fourth and fifth studies presented analyze spectral signatures and variances of in-situ forces during chemical mechanical planarization (CMP) of metal thin films for the purposes of improving endpoint detection, in particular, and helping us gain a better fundamental understanding of CMP processes, in general. A new parameter, directivity () is defined and used throughout all three studies. In the third study, CMP directivity is introduced as a ratio of shear and normal forces generated during polishing. This parameter is inspired by a parameter of the same name used to assess violins. Sound quality of individual violins is often evaluated through directivity by violin makers and physicists, which is the ratio of the variance of pressure exerted on the top plate of the violin to the variance of pressures exerted on its bottom plate. Older violins are known to have higher values of directivity than modern violins. A directivity greater than unity is indicative of sound traveling from the violin being more unidirectional. The physical components of the CMP polisher have quite a few parallels to the violin. For example, the strings of the violin are like the pad in the polisher. Typical coefficients of friction (COF) at the bow-rosin-string and wafer-slurry-pad interfaces are even similar, being in the range of 0.3 to 0.8 for both violins and CMP processes. This study observes the natural resonances of the polisher through spectral shear force signatures, and finds that the signature changes with added vibrations from the environment. The classic hammer strike test, which is common in testing the spectral signature response in violins, is applied to a polisher to show this effect. In the fourth and fifth studies, CMP of copper and tungsten is carried out with different pad surface micro-textures, generated from two different, commercially available conditioning discs. In-situ shear and normal force measurements are captured for all polishing experiments. The experiments are conducted with a variety of polishing pressures and sliding velocities. In a separate study, the removal rate (RR) from these studies is determined, as well as the coefficient of friction (COF). The surface of the pad after in-situ conditioning and polishing is imaged with confocal microscopy (CM) and various parameters are derived from the optical image data such as contact area between pad and wafer (CA), contact density between pad and wafer () and the microhydrodynamic lubrication layer (MHDLL) thickness. From the raw in-situ force data, variances of forces and spectral signatures are analyzed. For spectral analysis, the time domain is converted into the frequency domain with a fast Fourier transform, to distinguish a fingerprint of the experiment, or common force peaks between experiments. We then focus on the shear force because it represents the force that is most closely associated with friction at the pad-slurry-wafer interface. Variance of shear force is interpreted as stick-slip events at the interface. Results show that higher directivity values correlate in nearly all experiments (regardless of wafer material) with higher RR, larger CA and thinner MHDLL thickness. This clear correlation highlights the potential for directivity being used for endpoint detection in CMP. Several of the FFT shear force peaks from 0 to 150 Hz in every polishing experiment are shown to be harmonic peaks of force vibrations corresponding to the sliding velocity of the platen.