For the cleaning of silicon substrates, a mixture of an oxidizing and an etching agent is often used. One of the first successful mixtures (1970) is the well known APM or SC-1 cleaning (NH4OH/H2O2/H2O) (1). Since then, other mixtures (oxidizing agent/etching agent) such as HNO3/HF (2), H2O2/HF (3) were developed. For the cleaning of silicon substrates, a mixture of an oxidizing agent and an etching agent always results in good particle performance, since the Si under the particles is continuously oxidized and etched simultaneously (2,3).

Wafer cleaning studies have been performed so as to understand the influence of acoustic (megasonic) energy on particle removal in dilute SC-1 solutions. Surface etching alone (up to 60Å) has been found to be insufficient to completely remove silicon nitride surface particles from native oxide surfaces in the absence of megasonics. For megasonic cleaning processes the minimum surface etching required for complete nitride particle removal is significantly lower (between 3–12Å) than for a non-megasonic process. The exact 'threshold' for surface etching will depend on the chemical nature of the particle/surface and the megasonics power. Megasonics energy does not appear to enhance chemical etching of the substrate, at least for silicon oxide substrates, however, it significantly improves particle removal. This data suggests that the particle removal process can benefit from both a thermally activated component (etching) as well as an acoustic component (cavitation/ acoustic streaming).

A statistical design of experiments (DOE) approach has been employed to evaluate the effects of megasonic input power, solution chemistry, bath temperature, and immersion time on particle removal in SC-1 chemistries. Megasonic input power was the dominant factor in the response surface model. Substantially diluted chemistries, performed with high megasonic input power and moderate-to-elevated temperatures. were shown to be very effective for small particle removal. Follow-on studies to the original DOE have led to an investigation of ultradilute SC-1 chemistries with megasonic power. These chemistries ranged from 0 to 1000 ppm of NH4OH and H2O2. Post processing light point defect (LPD) counts differ substantially between bare n and p-type Si<100>, and ambient lighting conditions are shown to influence LPD counts on p-type Si<100>. Solution properties such as pH and oxidation potential have been studied, and an investigation of post processing silicon surface properties is underway.

Abstract Wet chemical removal of a silicon nitride/silicon Poly Buffer LOCUS (PBL) film stack with a phosphoric/nitric acid solution was characterized. Though silicon nitride etch rates remain constant, silicon etch rates decrease as a function of loading which severely limits the usable lifetime of the bath. This is caused by buildup of etching products which limits the amount of silicon that can be dissolved in the solution. Addition of HF to the phosphoric/nitric acid solution enhances the dissolution chemistry presumably by decomposing the silica reaction products and shifting the equilibrium. Characterization of this process was done to determine whether it could be applied towards PBL removal. Depending on the relative amounts of HF and HNO3 added, silicon etch rates could be enhanced by two orders of magnitude (200–400A/min) and silicon nitride etch rates by a factor of two. Selectivities for etching silicon to oxide and silicon nitride to oxide were typically between 8–12:1 and 8–20:1 respectively.

With the goal of optimizing point-to-point etch uniformity in a continuous flow wet processing system, a comprehensive fluid dynamics study was undertaken. Aspects of this work included developing a computational, finite-element, fluid dynamic model; performing a series of empirical studies based on dye-injection with photo detection and video analysis; and observing platinumwire bubble generation with transparent wafers.

Various flow modification designs were tested including the base case of no inserted device, a stationary “spinner” insert, and a stationary “showerhead” insert. Unsteady, non-uniform, flow distribution and zones of substantial flow recirculation were observed in the case of using no flow modification device. These observations were confirmed with the computational fluid dynamic model.

Used in conjunction with the two different types of flow inserts, a new supplementary Teflon screen was designed and tested. The combined optimum configuration using this screen with an insert was then identified. Significantly enhanced point-to-point etch uniformity, with minimal recirculation and prompt fluid displacement, resulted from this design.

We report on the use of the Surface PhotoVoltage (SPV) technique to monitor the Si surface bonding arrangement, and the impurity metallic contamination level prior to critical diffusion processes via the indirect measurement of surface charge and diffusion length, respectively. We show that the effectiveness of the pre-diffusion wet chemical cleaning and rinsing sequences can be accurately monitored via the real-time, nondestructive SPV measurement. In particular the nature of the surface passivation/chemical oxide formed during the cleaning and rinsing operations can be monitored by quantitative surface charge measurements. The importance of the prior wafer history is highlighted, as is the role of the Si starting material and measurement parameters.

A wafer gap conductivity cell determined which rinsing parameters control the removal of wafer cleaning and etching solutions. The first part of this study focused on set up and calibration of the conductivity cell. Sodium chloride solutions with known ionic conductivities are used as model fluids. Wafer rinsing experiments showing the concentration as a function of time are analyzed. The sensitivity of the wafer gap conductivity cell is compared to a wall mounted probe, the typical method used to measure rinsing efficiency. The conductivity results are explained using a fluid dynamics model of the wafer gap.

The second part of this study focused on using the wafer gap conductivity cell to study the removal of chemicals typically used in wafer cleaning processes. Experimental results show the effect of tank geometry and flow parameters on the time required for rinsing. Analysis shows the rinse process can be modeled by assuming convective fluid flow and ideal mixing. These results provide critical insight into the most important wafer rinsing parameters.

Electrochemical polarization experiments were performed on Si wafers in ammoniacal solutions maintained at a pH in the range of 9.5 to 11.5. Anodic polarization of silicon yielded curves which are typical for materials that undergo passivation. The values of open circuit potential and passivation potential for p-type Si wafers were more anodic than for the n-type Si wafers. Corrosion current density of p-type Si wafers of low resistivity was lower than that of wafers of high resistivity. Corrosion current densities correlated well with surface roughness induced in alkaline solutions. Addition of surfactant or H2O2 to alkaline solutions reduced critical current density for passivation and corrosion current density.

Cleaning effects of pure water containing dissolved oxygen of very low concentration (LDO water) to metallic contaminants on silicon wafer surface were confirmed. To maintain the concentration of the dissolved oxygen in water, experiments were performed in a glove box in which ambience was controlled so as to satisfy Henry's law between the water and the ambient gas. In the experiment using intensionally contaminated wafers, residual metal contaminants except copper on Si-surface decreased from 1014 atoms/cm2 to 1011 atoms/cm2 after the 1ppb hot LDO treatment at boiling point. This effect depended on the concentration of dissolve oxygen, treatment temperature, and rinsing time. Contact angle of the wafer surface increased gradually from about 10 [deg] with decrease in the residual metals and jumped up to about 90 [deg]. when the amount of residual metals reached to minimum. Then absorption peak of Si-O bonds in FT-IR-RAS spectra also disappeared. These results therefore show that hot LDO water removed metal contaminants from the wafer surface with etching of the native oxide.

Removing the native oxide from the poly-Si surface prior to WSix deposition is essential for achieving high quality silicides as well as sufficient film adhesion, particularly after high temperature anneal or oxidation. Contact angle studies have been used to determine initial and time-dependent surface characteristics of several types of silicon surfaces following immersions in HF-based etchants for varying amounts of time. The morphological characteristics of the surfaces before and after exposure to etchants, as well as the relative etch rates and wetting capabilities of the etchants have been used to explain the following results: With respect to initial contact angle studies, the implanted & annealed polycrystalline silicon surface has the lowest contact angle followed by polycrystalline and monocrystalline surfaces. Longer immersion times yield lower initial contact angles. The 0.1% lightly-buffered HF solution results in the highest contact angle followed by the 1% buffered HF solution with surfactant, and the 1% HF solution. With respect to contact angle changes during ambient air exposure time, the asdeposited polycrystalline silicon surface is most stable followed by monocrystalline, and implanted & annealed polycrystalline silicon surfaces. Longer immersion times improve surface stability while the 0.1% lightly-buffered HF solution results in the most stable surface followed by the 1% buffered HF solution with surfactant, and the 1% HF solution.

Time zero breakdown electrical measurements, SIMS, Bias Temperature Stress (BTS), and Triangular Voltage Sweep (TVS) techniques performed on MOS capacitors have been used to characterize two types of defects produced in TEOS thin films subjected to CMP process planarization. Microcracking of the insulator, resulting in degradation in breakdown characteristics, and uptake of K+ ions (from the polishing slurry) are found to occur. These defects are thought to be caused by the chemical and mechanical stress to which the oxide surface is subjected during the CMP process. The effects of polish pad type, non K+ ion containing slurry and post-CMP wet etch steps were all found to influence the extent of damage observed in polished films.

The polishing technology used for manufacturing ultraflat and smooth Si surfaces on a large scale is the chemomechanical polishing (CMP) technique. This technique combines the chemical corrosive removal of silicon atoms and the mechanical transport of the agents. The removal rates strongly depend on the interaction of mechanical parameters and the chemistry involved in the polishing process like the pH of the alkaline polishing slurry used. Removal of Si during CMP is explained by a nucleophilic attack of OH− to silicon atoms catalyzing the corrosive reaction of H2O resulting in cleavage of silicon backbonds. Characterization of the surface chemistry of the silicon wafer after polishing by X-Ray Photoelectron Spectroscopy and High-Resolution Electron Energy Loss Spectroscopy reveals an oxide free, predominantly hydride covered silicon surface displaying hydrophobic properties. Morphological features like microroughness as well as localized surface irregularities on the silicon surface, also referred to as Light Point Defects, depend on different strongly interacting process parameters. Microroughness is reduced by CMP by several orders of magnitude as characterized by lightscattering techniques and Atomic Force Microscopy.

Chemical-Mechanical Planarization (CMP) of SiO2 is performed using alkaline silica slurries while CMP of tungsten (W) utilizes acidic slurries with alumina as the abrasive. Proposed mechanisms for the two CMP processes, with more emphasis on SiO2-CMP, have been discussed in literature. However, much less is known about the removal mechanism of residual slurry particles from the planarized surfaces - a crucial step for subsequent device processing. We discuss the chemical and physical basis of post-CMP cleaning by double-side scrubbing using polyvinyl alcohol (PVA) brushes and show how the interactions between the wafer surface, slurry, and the brush material affect the overall cleaning efficiency. Using the zeta potential concept the common features for cleaning surfaces after SiO2-CMP and W-CMP are established and the differences between these two systems are highlighted. We present surface particle levels for two model systems as a function of cleaning chemistries and discuss their influence on post-CMP surface metal levels.

The effects described in this paper are specific to cleaning with mechanical brush scrubbing. Oxides, both thermally grown and deposited are common cleaning applications for mechanical brush scrubbing. Thermally grown oxides present higher final defect counts after scrubbing with deionized water compared to deposited oxides. In this paper we present our results on the cleaning of unpolished and polished oxide surfaces and show differences in these results to be dependent on the chemistry used for cleaning and the degree of hydrophilicity of the wafer surface.

The surface cleaning technology with the use of ice scrubber cleaning has been developed to remove the particles after Chemical Mechanical Polishing (CMP) process. The ice particles with a high speed nearly equal to the sound velocity bombarded the Si wafer surface, as a result, the residue from the slurry solution was reduced from ∼5/cm2 to ∼0.05/cm2 and the metal impurities are completely eliminated below the defect limitation for ICP mass spectroscopy. The charge build-up damage due to the high speed particles is not introduced into the the MOS capacitors. This technology is quite effective, compared with the conventional brush scrubber method and is applicable for the cleaning process below the quarter micron devices.

The effect of bath temperature, megasonic power, and NH4-OH:H2O ratio are studied for particle removal efficiency, surface roughness, and surface Fe concentration in SC-1 cleaning solutions. Experimental results are presented which show removal efficiencies better than 97% on bare silicon wafers for optimized process conditions. These results are related to the etch rate of thermal oxides and a model is developed for reducing surface roughness and minimizing Fe contamination levels while maximizing particle removal efficiency.

The slight etch and accompanying roughening of the silicon by the APM solution are of great concern as device geometrys and gate oxide thicknesses decrease. This report covers the variations in the SiO2 etch rate and surface roughening of an APM solution as a function of NH4OH and H2O2 concentration and temperature. In general, the etch rate and roughening increased with increasing temperature and NH4OH concentration but was unaffected by H2O2 concentration. Other portions of the larger APM study covering the addition and removal of metals and the removal of particles by APM are published elsewhere. Together, these studies provide a complete set of process response surfaces for the SC-1 chemistry. These surfaces can be used to select the most promising regime of parameter space for any particular process.

Direct surface analysis of Si wafers and environmental materials such as polymers for wafer carriers and for high purity water systems is important to identify contaminants and their sources. Organic contaminants on the surfaces are difficult to analyze; however, their adverse effects have been cited in recent years. This paper demonstrates the detection and identification of surface contaminants using Time-of-Flight Secondary Ion Mass Spectrometry (TOF-SIMS) on semiconductor related materials. The analytical information provided by TOF-SIMS can be useful for maintaining material cleanliness or for failure analysis.

The solutions DIH20: NH40H: H202 and DIH20: HCL: H202 have been used extensively in the semiconductor industry for the removal of transition and alkali metals from wafer surfaces. The former, when used in a ratio of approximately 5:1: 1 effects the wet oxidation removal of organic surface films and trace metals (Cu, Au, Ag, Ni, etc.) while the latter, in an approximate 6:1:1 ratio dissolves alkali ions and hydroxides of Al+3, Fe+3, and Mg+2 desorbing by complexing residual metals. Of concern is that these mixtures are known to plate out charged ionic species, particularly Fe, which remains on the Si surface after exposure to these solution. This effect adds an enhanced surface potential barrier which can be negative or positive depending on the concentrations of species deposited.

While the effects of Fe as deposited by the NH40H solution is well-known, we are concerned in this study with the deposition of Ca, K, and the lighter ions, Na and Al. Conventional analytical methods such as ICP-MS or AAS or TXRF have been used to detect these ions but are problematic since each has a limited range of detectable species or is destructive to the surface when used with VPD to increase resolution. In this study, we have verified the deposition of these species by a ToF-SIMS (Time-of-Flight Secondary Ion Mass Spectrometry) analysis since it has the ability to detect the complete range of the desired ions. In addition, we have applied high-injection surface photovoltage (SPV) to obtain an average surface charge on each sample. To illustrate the use of these methods, we have applied them to the analysis of various closed acid processing systems. Our interest here is the amount of these selected ions being deposited from these solutions and the influence of using a standard canister or bulk acid supply system. Comparisons were also made to standard and megasonic RCA cleans and hot phosphoric nitride strips. We will discuss ToF-SIMS and how it was applied to analyze the addition of these charged trace ions. This work suggests that the problem of trace ion detection on Si must include the entire range of transition and alkali metals, and that ToF-SIMS is a promising analytical tool for this purpose.

Secondary ion mass spectroscopy (SIMS) coupled with oxygen flooding of the silicon surface during analysis provides an analytical technique capable of detecting ≤1010 atoms/cm2 of many surface elemental contaminants. Of particular importance to meet the future needs of the semiconductor industry is the current ability to detect Al and Fe contamination at a level of 2×109 atoms/cm2.