Particle Removal and Feature Damage Reduction Using Carbonated Ammonia Solutions for Enhanced Megasonic Cleaning Processes

Abstract

In integrated circuit (IC) manufacturing, particulate contamination from hundreds of processe steps is a major cause of yield loss. The removal of particles is typically achieved through liquid chemical formulations aided by a sound field in the MHz frequency range. When liquid is irradiated with megasonic waves, dissolved gases play an important role in particle removal and feature damage. To take the advantage of the beneficial effect of CO₂ (aq.), this thesis describes the development and optimization of a megasonic cleaning process using a chemical system containing NH₄OH and NH₄HCO₃ at an alkaline pH in which a specific amount of aqueous CO₂ can be maintained to minimize feature damage. In addition, certain etching effects at a slightly alkaline pH were supported for achieving high particle removal. Sonoluminescence (SL) data were collected from these cleaning solutions and correlated with the cleaning performance. The intensity of SL is believed to be a sensitive indicator of transient cavitation during megasonic irradiation, which is thought to be responsible for fragile feature damage. To further analyze the SL signal with respect to the emission from hydroxyl radicals, single-band filters were used to collect the SL signal in different wavelength ranges. The study of particle removal and feature damage was performed using a single-wafer cleaning tool, MegPie® (ProSys, Inc.), which provided acoustic irradiation at a frequency of 0.925 MHz. Commercially available SiO₂ slurry with 200 ± 20 nm particles was used for particle contamination. Particle removal was investigated on both blanket SiO₂ samples and patterned samples. Feature damage studies were conducted on patterned samples by examining the number of line breakages per unit area. By adjusting the pH in NH₄OH/NH₄HCO₃ solutions from 7.8 to 8.5, the amount of CO₂ (aq.) was varied. At a pH of 8.2 with ~ 320 ppm CO₂ (aq.) in the cleaning solution, a high particle removal efficiency was achieved (> 90%) at an acoustic power intensity of 1 W/cm² for an exposure time of 60 s, and the feature damage was reduced by > 50%. For SL signal analysis, band filters in the wavelength range of (i) 280 – 305.5 nm, (ii) 300 – 340 nm, (iii) 335 – 375 nm, and (iv) 374.5 – 397.5 nm were used to resolve the SL spectrum in these wavelength ranges. The filters were sandwiched, one at a time, between the optical window and the photomultiplier tube (PMT) in the Cavitation Threshold (CT) cell. Air-, Ar-, and CO₂-containing DI water (at pH 4.53 with ~ 90 ppm aqueous CO₂) was pumped through the cell at a flow rate of 130 ml/min. The acoustic power was ramped from 0.1 to 4 W/cm² at an acoustic frequency of 0.925 MHz. The SL signal intensity showed the highest value in the ranges of 300 – 340 and 335 – 375 nm in air- and Ar-saturated DI water, which is due to the emission from excited hydroxyl radicals. These results are consistent with an SL spectrum analysis performed using expensive optical set-ups. In CO₂-containing DI water, the SL signal intensity was suppressed by a factor of 100. The methodology reported in this work is simple, inexpensive, and capable of capturing SL spectral features due to hydroxyl radicals.