In future semiconductor technology generations cleaning processes will face the challenge of removing nano-particles with no damage to fragile structures and virtually no etching of the substrate. In this study we have evaluated the capabilities of a representative set of present megasonic cleaning tools to meet this challenge in process conditions where the etching of thermal silicon dioxide was lower than 0.5 Å. The tests vehicles for particle removal and damaging consisted in 34 nm SiO2 particles on hydrophilic Si wafers and in poly-on-gate lines of line-width ranging from 150 to 70 nm, respectively. No tool reached the target of high particle removal efficiency (PRE) and low damage to the 70 nm lines in the presen series of tests. Lower damage could only be obtained at the cost of lower PRE, by decreasing the megasonic power. Wafer maps for PRE and damage showed patterns that were tool-specific. Only two systems out of five seemed to show a simple direct correlation between PRE and damage at wafer level, indicating that more fundamental research is needed to understand the cleaning and damaging mechanisms in megasonic systems.
INTRODUCTION
In semiconductor manufacturing, as features sizes are scaling down below 100 nm, particles with a diameter of a few tens of nanometers need to be considered as killer defects. For example with the 90nm technology node particles of a size larger than 45nm are believed to be potential killer defects for devices in chips.1 For several reasons related to substrate consumption budget, cost, and environmental impact, present cleans make use of diluted chemistries with low etching capability and need additional physical mechanisms, such as megasonic agitation, to remove contaminant particles.2 As particle size decreases, the ratio of adhesion force over cleaning force increases, thereby potentially compromising the particle removal efficiency (PRE).3 On the other hand, wafer surfaces may present fine structures with fairly high aspect ratios, such as gate electrodes or low k isolation patterns, which become vulnerable to sideward impact by physical forces.4,5 The combination of all these trends results in a collapse of the process window to the extent that cleaning of nano-particles is becoming a major challenge in production and the future use of traditional cleaning methods is questioned. Previous studies have demonstrated that, even though cleaning of nano-particles is becoming increasingly more difficult as particle size decreases, it is actually possible toremove particles with a diameter as small as about 30 nm by megasonic cleaning.6,7 PRE was shown to depend strongly on the presence of dissolved gas in the cleaning solution, indicating that cavitation was playing an important role in the cleaning mechanism.8,9 The decrease in PRE at smaller particle size was associated with a decrease in cleaning uniformity.
In this work various megasonic cleaning systems were evaluated for removal efficiency of nano-particles, cleaning uniformity, and damaging of fragile structures, providing a snapshot of the capabilities of present tools to respond to the new challenges. Additionally comparison of wafer maps for PRE and damaging was used to better characterize the relationships between PRE and damaging.
MATERIALS AND METHODS
The cleaning tests were performed in various systems (A, C to F) differing by their configuration (batch or single-wafer, position of transducers, carrier), solution flow (recirculation or single-pass), transducer operation (continuous or multiplexed), and wafer size (200 or 300 mm) (Table 1). Transducer frequencies ranged from 0.7 to 1 MHz. The effect of dissolved gases was studied by comparing results obtained with degassed and aerated DIW. PRE was determined with hydrophilic 200 or 300 mm Si wafers contaminated with SiO2 particles of 34nm diameter purchased as slurry (Clariant Elexsol). Wafers were contaminated using an immersion based controlled contamination (CC) procedure and used within a few hours after preparation.10 Particles numbers were determined using light scattering on a KLA Tencor SP1TBI using the haze channel.10,11 Particle removal efficiencies (PRE) were calculated from measured particle counts after CC and after clean, taking the initial count pre-CC into account.
Table 1. Overview of tested megasonic cleaning tools.
The test vehicle for damaging consisted in 200 mm wafers with poly-on-gate lines that were inspected by laser-light scattering (KLA-Tencor AIT) and SEM (Philips XL810). The lines were 8 mm long and about 170 nm tall, with widths and aspect ratioranging from about 150 to 70 nm and about 1.1 to 2.5, respectively (Fig. 1). The lines were printed in groups of nine, differing by line-width and spacing. External lines in groups of narrow spacing were printing thinner as a result of so-called proximity effects during photolithographic exposure (see Fig. 1, left). Finally the pressure distribution at the surface of wafers during cleaning was determined with wafers covered with a pressure indicative sensor film (Pressurex Micro, from Sensor Products Inc.).
RESULTS AND DISCUSSION
With all megasonic-cleaning tools damaging of poly lines showed up in SEM inspection as the removal or bending of small pieces of line with a length of about 1 to 2 µm (Fig. 2). The localized character of damage suggests that cavitation was probably the cause. Full wafer inspection with a KLA-Tencor AIT could only detect the removed pieces of line but allowed to determine defects statistics and the spatial distribution of defects on wafers.
