MICRO: Vereecke

Investigating the role of gas cavitation in megasonic nanoparticle removal

Guy Vereecke, Els Parton, Frank Holsteyns, Kaidong Xu, Rita Vos,
and Paul W. Mertens, IMEC; Michael O. Schmidt, Infineon; and
Theresia Bauer, Wacker Siltronic

One challenge facing the IC industry in maintaining historical trends in productivity is to find increasingly efficient ways to clean wafers between the different process steps. As devices continue to shrink, new cleaning technologies must be developed to remove particles smaller than 50 nm. Many people involved in semiconductor manufacturing doubt that megasonic cleaning will remain the success story it has been until now. Its physical action might not be sufficient to remove ultrasmall particles attached to the wafer surface with ultralarge adhesion forces, while its harmful side effects may outshine its cleaning capability in future technology nodes.

To push the megasonic cleaning technique to its ultimate limits, insight into its underlying properties is necessary. Research in this area, based on a new particle-measuring approach known as the haze method, has revealed that the use of different system parameters during megasonic cleaning results in varying removal efficiencies at the wafer level. The conclusions from that research all point in the same direction: gas cavitation is probably the key to efficient cleaning in megasonic tanks.

Megasonics: A Sound Technique for Tackling Nanoparticles

Traditional cleaning solutions typically use highly concentrated chemicals to perform underetching—the removal of a thin substrate layer to detach particles from the wafer surface. However, that method has become unthinkable in recent technology nodes, in which layers are scaled to the smallest dimensions possible and substrate consumption must be reduced to a minimum, creating the need for dilute solutions. Furthermore, less-concentrated chemicals are preferable for cost and environmental reasons.

Seeking a cost-efficient and ecologically sound cleaning method, scientists at RCA in the 1970s thought that by adding the physical action of ultrasound waves to a cleaning solution, they could reduce chemical consumption.¹ Ultimately, that technique began to use frequencies in the megahertz range (hence the term megasonics) to eliminate the harmful effect of cavitation—the nucleation, growth, and implosion of bubbles in the cleaning solution. Experts believed that cavitation did not occur at higher frequencies. Although the use of megasonic energy has become widely accepted in wet-cleaning processes and its effectiveness has been demonstrated in numerous studies, it is not clear how it removes particles.

Phenomena such as acoustic streaming, Schlichting streaming, cavitation microstreaming, and cavitation bubble implosion (bubble collapse) are all candidates for revealing the mystery of the megasonic cleaning principle.^2–4 Acoustic streaming refers to large-scale fluid motion in the megasonic tank during megasonic irradiation. On a smaller scale, Schlichting streaming is fluid motion with vortices in the thin acoustic boundary layer at the wafer surface. At sufficiently high power levels, many micron-sized cavitation bubbles are generated in the fluid and the boundary layer near contaminating particles. These bubbles exhibit a pulsating motion, causing cavitation microstreaming to occur around them. The bubbles can also undergo cavitation bubble implosion, generating even larger velocities and pressures. All of these phenomena are believed to produce high shear stresses on the wafer surface, leading to the physical removal of contaminants.

Ongoing research is attempting to gain deeper insight into the mechanisms behind particle removal in a megasonic energy field. Indeed, acquiring knowledge of the ultrasonic cleaning mechanism is essential for improving and optimizing the cleaning process, especially in the case of ultrasmall particles.

Inspection Tool Limitations and the Haze Method

In order to evaluate and develop future cleaning techniques and chemistries, suitable detection methods are necessary. Traditionally, inspection tools have used scattered light to reveal information about the sizes and locations of contaminants known as light-point defects (LPDs). However, state-of-the-art inspection tools cannot detect particles smaller than about 60 nm with an LPD signal. To overcome this detection roadblock, researchers at IMEC (Leuven, Belgium) extended the capability of the inspection tool and further developed the haze method.^5,6 This innovative method is based on the low-intensity background signal (haze) collected from traditional light-scattering tools, which provides information on the surface roughness of full wafers. The relationship between LPDs and haze is presented in Figure 1.

Figure 1: The relationship between LPDs and haze in a light-scattering inspection tool.

Researchers observed that in a particular range of surface density, the haze signal can detect particles of known size as surface roughness. They saw that the haze signal is proportional to the surface density of deposited particles, which is proportional to the volume particle concentration in the bath. Figure 2 shows experimental data and a fit model for added haze versus particle density for SiO₂ particles of different diameters added to an Si₃N₄ substrate.⁵ Based on this proportionality, calibration curves were generated, enabling the researchers to obtain haze maps showing regions with different densities of nanosized particles. This breakthrough in detection capability paved the way for a new understanding of megasonic cleaning.

Figure 2: Experimental data and fit model for added haze versus particle density for SiO₂ particles of different diameters added to an Si₃N₄ substrate.

Megasonic Cleaning Experiments

To study the influence of hardware and process parameters on the uniformity and efficiency of megasonic cleaning to remove nanoparticles from the wafer surface, experiments were conducted using 200-mm wafers. First, the wafers underwent an ozone and DI water/hydrogen fluoride cleaning step, HCl/rinse and oxide regrowth/rinse steps, and a Marangoni drying step.⁷ Then they were subjected to controlled contamination by being immersed in a particle-contaminated bath containing SiO₂ particles with 126-, 78-, and 34-nm diameters. Finally, the wafers were rinsed in an overflow bath and Marangoni dried again.

The contaminated wafers were cleaned in a megasonic tank equipped with two arrays of four transducers each (2 X 300 W at about 720 kHz) that were mounted at the bottom of the tank. The transducers were operated in multiplexed, or pulsed, mode, with two transducers (one from each array) being fired at the same time for one second. Particle counts were determined using an SP1^TBI light-scattering tool from KLA-Tencor (San Jose), which detected 126- and 78-nm particles using the LPD method and 78- and 34-nm particles using the haze method. Particle removal efficiency (PRE) was calculated by comparing particle counts before and after controlled contamination and cleaning.

Influence of Bath Aeration. To test the influence of bath aeration on megasonic removal efficiency, a degasification system was installed in the megasonic system to remove dissolved gases from the ultrapure water. Downstream of the degasification unit, a gasification unit to add different gases (O₂ in this case) and a spiking unit to add chemicals (APM, NH₄OH/H₂O₂/H₂O) were installed.

Figure 3: Influence of dissolved oxygen on the removal efficiency of 34-nm SiO₂ particles as a function of megasonic power in DI water at room temperature.

In all experiments involving degassed water and degassed APM solutions at 1:1:250, 1:1:200, and 1:1:100 dilutions, the PRE for 34-nm particles was 0%, unless O₂ was spiked into the cleaning bath. This implies that acoustic streaming and Schlichting streaming, on the one hand, and vapor cavitation (bubbles filled with water vapor), on the other, do not have cleaning capacity, since these phenomena can occur in degassed solutions that have a PRE of 0%. However, because gas cavitation occurred only in the case of the oxygenated solutions, it was a suitable candidate for explaining the success of the megasonic cleaning process in these experiments. Figure 3 demonstrates the influence of dissolved oxygen on the removal efficiency of 34-nm SiO₂ particles as a function of the megasonic power level in DI water at room temperature, while Figure 4 illustrates the influence of dissolved oxygen on the removal of 34-nm SiO₂ particles as a function of APM dilution level at room temperature. The process time for both tests was 5 minutes.

Figure 4: Influence of dissolved oxygen on the removal efficiency of 34-nm SiO₂ particles as a function of APM solution level at room temperature.

Influence of Carrier Type and Wafer Spacing. Different types of carriers exist to support the wafers in the cleaning bath. Since megasonics produce focused beams traveling through the bath solution, the waves can be deflected or blocked by a carrier's structural elements, an effect known as shadowing. In IMEC's work, the influence of wafer carrier type on the PRE of 34-nm particles was tested by comparing standard low-profile carriers with low-mass carriers having a reduced support area. Tests using these carriers were performed with single wafers or full batches. In addition, single wafers were immersed in the bath using vacuum tweezers. All tests were conducted in aerated DI water for 5 minutes.

The haze maps from the single-wafer cleans clearly indicate shadowing: the maps show regions of reduced PRE near the wafer carriers' structural elements. The PRE for the low-mass carriers was higher than that of the standard carriers. The best overall PRE and the highest cleaning uniformity were accomplished when vacuum tweezers were used. Of course, the use of tweezers under commercial conditions is impractical, but modern 300-mm tools achieve minimal shadowing by using edge-grip carriers, which contact the wafer at only two points.

Cross sections of the two carrier types and of the vacuum tweezer setup used in these experiments are illustrated in Figure 5. Also pictured are haze maps of cleaned wafers with 34-nm particles. The figure indicates that it was more effective to perform megasonic cleaning on single wafers than on full batches. In the standard low-profile carriers, full-batch wafers had an average PRE of 48% while single wafers had a PRE of 54%. In the low-mass carriers, full-batch wafers had an average PRE of 59% while single wafers had a PRE of 79%. And single wafers held with a vacuum tweezer had an average PRE of 86%. The higher average PRE and cleaning uniformity for the single-wafer cleans indicated that PRE decreases in direct proportion to tighter wafer spacing.

Figure 5: Haze maps of cleaned wafers with 34-nm particles: (a) low-profile carriers, (b) low-mass carriers, and (c) vacuum tweezers were used to support the wafers in single-wafer or batch cleans performed in aerated DI water. (Wafers were placed 6.3 cm from the tank floor.)

The haze maps from the batch-cleaned wafers revealed a consistent pattern of vertical stripes. Contrary to the researchers' expectations, PRE was lower in the line of the beam, above the center of the transducers, than above the edges of the transducers, although acoustic pressure is greatest where megasonic waves pass.

Interestingly, relatively large bubbles were observed on the sides of the beams (the position of the beams was visible as an accumulation of solution). Originally formed above the center of the transducers, small cavitation bubbles moved toward the side surface, or boundaries, of the beam region, which exhibits lower pressure than the center of the beam. At the same time, the bubbles agglomerated and were visible to the naked eye. A haze map showing the vertical lines and the difference in PRE levels at different points above the transducers is shown in Figure 6a. The gas bubbles that were observed at the edges of the megasonic beam are illustrated in Figure 6b.

Figure 6: (a) Haze map showing vertical lines with lower PREs above the transducers' center and higher PREs above the edges of the transducers, and (b) diagram of gas bubbles in the cleaning solution at the edges of the megasonic beam.

As demonstrated in the aeration experiment, more gas results in higher PRE. Hence it was concluded that solution degassing in the center of the pulsed megasonic beams caused the striped pattern on the cleaned wafers. This effect was less pronounced in the single-wafer tests than in the batch-wafer tests, probably resulting from differences in fluid flow. The replacement of old solution with fresh, aerated solution occurred in three dimensions for the single wafers but only in two dimensions for the full-batch wafers.

Influence of Wafer Position in the Bath and Chemical Flow Rate. In the next experiment, wafer position in the tank and the effect of flow rate were investigated. To eliminate the influence of carrier shadowing, vacuum tweezers were used in this experiment to hold the wafers in position.

Wafers were placed 0.6 and 6.3 cm from the tank floor. The chemical flow rate was 8 L/min. Haze maps of wafers contaminated with 34-nm SiO₂ particles after megasonic cleaning in APM showed a low-PRE region (<15%) at the bottom of the wafers that were placed 0.6 cm from the tank floor. Higher PRE levels were obtained from the wafers that were placed 6.3 cm from the tank floor. The results of this test are presented in Figure 7.

Figure 7: Haze maps of wafers contaminated with 34-nm SiO₂ particles after megasonic cleaning in APM with wafers (a) 0.6 cm and (b) 6.3 cm from the tank floor.

The effect of chemical recirculation on the PRE levels of wafers contaminated with 34-nm SiO₂ particles that were placed 6.3 cm from the tank floor was tested by using either no flow rate or a flow rate of 8 L/min. When no flow rate was used, the region with a low PRE level (<15%) was very large. This test is summarized in the haze maps shown in Figure 8.

Figure 8: Haze maps of wafers contaminated with 34-nm SiO₂ particles after megasonic cleaning in APM with a flow rate of (a) 0 L/min and (b) 8 L/min.

To explain these findings, pressure measurements were performed with a hydrophone at different bath depths. The values obtained were not related to flow rate. Moreover, while near-field variations in pressure were measured, they could not be correlated with variations in PRE. Consequently, differences in PRE levels based on wafer position in the bath were not caused by pressure differences.

Again, gas cavitation may explain why the lowest PRE was observed when no flow was used and why the placement of the wafer farther from the floor of the bath resulted in a better PRE. Archimedes' principle holds that if the mass of an immerged object is less than the mass of the water displaced by the object, the object will rise. Hence acoustic pressure and buoyancy caused gas bubbles to rise to the bath surface. When local gas depletion at the bottom of the tank was not compensated for by chemical recirculation, it was believed that much of the tank was filled with degassed solution.

Influence of Chemistry. After investigating the influence of carrier type, wafer spacing, wafer position in the bath, and the chemical flow rate, the researchers studied the composition of the cleaning solution. Based on the findings from the previous experiments, a low-mass carrier, a high wafer position in the tank, and optimal recirculation conditions were employed. Wafers were cleaned with or without megasonic action. Megasonic cleans were performed in aerated DI water or 1:1:500, 1:1:50, and 1:1:5 concentrations of APM solution. Figure 9 shows the PRE for 78- and 34-nm SiO₂ particles with and without megasonic irradiation at fixed bath temperatures of 30° and 50°C. Process time was 5 minutes.

Interestingly, neither the 34- nor the 78-nm particles could be removed without megasonic cleaning. In all test cases, the use of megasonic energy resulted in the removal of nearly all 78-nm contaminants. The removal of the 34-nm particles improved greatly with the use of increasingly concentrated APM, especially at 50°C.

Figure 9: Particle removal efficiency for 78- and 34-nm SiO₂ particles in aerated DI water and APM solutions (a) with megasonic irradiation and (b) without megasonic irradiation.

The positive effect of APM on megasonic cleaning efficiency can be explained in different ways. First, it is known that in the presence of chemicals, the pH of the solution causes electrostatic repulsion between the particles removed and the wafer, preventing the particles from redepositing and readhering to the wafer surface. However, tests with blank wafers indicated that in DI water, just as in APM, particles did not redeposit on the wafer surface, ruling out this mechanism.

Second, the etching activity of the chemicals may account for the improved PRE. Although the slight underetch of the particles by APM chemicals is not sufficient to completely release the particles, the underetch effect may help cavitation to loosen the particles from the wafer surface.

Third and most plausibly, APM solutions contain more gases than DI water. While the presence of chemicals alone will not result in the enhanced formation of gas bubbles (since no PRE was seen in degassed solutions, as shown in Figure 4), it is possible that chemicals influence the cavitation properties of existing gas bubbles. Because of its higher vapor pressure, ammonia can vaporize into bubbles, while the decomposition of H₂O₂ by cavitation can provide an additional source of O₂. As in all the other experiments discussed here, gas cavitation seemed to be the prime cause of higher PREs.

Conclusion

Megasonic cleaning techniques must be improved to remove ultrasmall (e.g., 30-nm) particles, requiring an understanding of the physical and chemical processes that take place in the megasonic tank. With the aid of a novel detection technique known as the haze method, the influence of wafer carrier type, wafer position in the tank, chemical flow rate, aeration, and chemical concentration were studied. It was concluded that gas cavitation is the basic cause of particle removal in megasonic tanks.

While cavitation can be used effectively on blanket wafers to improve cleaning efficiency, its use in front- and back-end-of-line processes, where damage to device structures can occur, is more challenging. Therefore, a compromise must be reached between particle removal efficiency and device damage, requiring further insights into the phenomenon of cavitation. For example, if the parameters influencing the intensity of cavitation bubble implosion can be determined and controlled (i.e., if smaller bubbles and/or less energetic implosion phenomena can be achieved), more-benign megasonic cleaning techniques can be developed. Perhaps a cleaning technique can be created that minimizes the role of cavitation. In any case, megasonics has not yet revealed its ultimate capabilities.

References

1. A Mayer and S Schwartzman, "Megasonic Cleaning: A New Cleaning and Drying System for Use in Semiconductor Processing," Journal of Electronic Materials 8, no. 6 (1979): 855–864.

2. SL Cohen et al., "Studies of the Relationship between Megasonics Surface Etching and Particle Removal in SC-1 Solutions," in Proceedings of the Materials Research Society Symposium, vol. 386 (Warrendale, PA: Materials Research Society, 1995), 13–19.

3. AS Busnaina and F Dai, "Megasonic Cleaning," Semiconductor International 20, no. 9 (1997): 85–90.

4. M Olim, "A Theoretical Evaluation of Megasonic Cleaning for Submicron Particles," Journal of the Electrochemical Society 144, no. 10 (1997): 3657–3659.

5. K Xu et al., "Relation between Particle Density and Haze on a Wafer: A New Approach to Measuring Nano-Sized Particles," Solid State Phenomena 92 (2003): 161–164.

6. SH Yoo et al., "Particle Removal Efficiency Evaluation at 40 nm Using Haze Particle Standard," Solid State Phenomena 76–77 (2001): 259–262.

7. M Meuris et al., "The IMEC Clean: A New Concept for Particle and Metal Removal on Si Surfaces," Solid State Technology 38, no. 7 (1995): 109–114.

Guy Vereecke, PhD, is responsible for research on advanced cleaning techniques at IMEC (Leuven, Belgium). He joined the institute's ultraclean processing group in 1993 to work on gaseous contaminants and dry cleaning methods. He received MS and PhD degrees in materials science from the Catholic University of Louvain-La-Neuve in Belgium. (Vereecke can be reached at +32 16 281527 or guy.vereecke@imec.be.)

Els Parton, PhD, is a scientific editor at IMEC, where she is responsible for authoring and editing the research organization's numerous technical documents and publications. She received a PhD in engineering from the University of Leuven. (Parton can be reached at +32 16 281467 or els.parton@imec.be.)

Frank Holsteyns, is a member of the ultraclean processing group at IMEC, where he is involved in research into megasonic cleaning and wafer inspection. He received an MS in engineering from the Catholic University of Leuven. (Holsteyns can be reached at +32 16 281325 or frank.holsteyns @imec.be.)

Kaidong Xu has been a PhD student at IMEC for almost four years, where he has developed the haze method for counting nanosized particles on wafers. He is also working on brush scrubber cleaning. He received a BS in chemistry from Peking University in China. (Xu can be reached at +32 16 281276 kaidong. xu@imec.be.)

Rita Vos, PhD, joined IMEC in 1995 as a researcher in the field of metallic contamination removal and is responsible for the development of new chemistries for ultraclean processing. She received a PhD in chemistry from the Catholic University of Leuven. (Vos can be reached at +32 16 281534 or rita. vos@imec.be.)

Paul W. Mertens, PhD, heads the ultraclean processing group at IMEC. He joined the institute in 1984 to work on silicon-on-insulator technologies. Since 1990, his main field of research has been silicon wafer surface quality, particularly for ULSI applications. His work involves thin gate dielectrics, defect control, the effects of contamination, cleaning processes, and related metrology. He received MS and PhD degrees in applied sciences from the Catholic University of Leuven. (Mertens can be reached at +32 16 281352 or paul. mertens@imec.be.)

Michael O. Schmidt, PhD, is a senior process engineer in the area of wafer cleaning technology and BEOL processing at Infineon Technologies (Dresden, Germany). He is serving as an industrial resident at IMEC. He received a PhD in inorganic chemistry from the University of Cologne in Germany. (Schmidt can be reached at +49 351 8867520 or mo.schmidt@infineon.com.)

Theresia Bauer, PhD, is the senior manager in the cleaning and etch R&D area at Wacker Siltronic (Burghausen, Germany). She has 18 years of experience in those process areas. She received a PhD in physical chemistry from Friedrich Alexander University in Nuremberg-Erlangen, Germany. (Bauer can be reached at +49 8677 835114 or theresia. bauer@wacker.com.)

Wet Surface Technologies

Investigating the role of gas cavitation in megasonic nanoparticle removal