Sub-100nm technologies drive single-wafer wet cleaning
02/01/2002
Paul W. Mertens, Els Parton, IMEC, Leuven, Belgium
overview
The stringent requirements of sub-100nm technologies are pushing the need for single-wafer cleaning methods. As device dimensions shrink, the size of killer defects also decreases, requiring economical and environmentally friendly cleaning processes with increased particle removal efficiency. This article proposes a single-wafer, single-chemistry cleaning system with optimized megasonics and a drying method that leaves the wafer watermark-free.
 |
Cleaning processes are a fundamental step in the production of micro-electronic products. Because of inherently small geometries, metallic contamination on the wafers can seriously degrade device performance. Calcium and iron have been identified as the most detrimental metals, associated with gate oxide integrity (GOI) degradation. Now with shrinking of gate dimensions, the size of "killer defects" will reach the nanometer scale — 50nm for the 100nm technology node. In order to meet the stringent gate-oxide defect density requirements of the future, cleaning strategies with a higher performance have to be developed. To make them production-worthy, cost and environmental issues must also be addressed.
A combined solution
Following the design of device fabrication processes, wafer cleaning is going toward a clustered single-wafer mode. Single-wafer cleaning provides a number of advantages: shorter cleaning times to meet critical time constraints; the use of megasonic agitation, which can improve uniformity; a reduction of scrap wafers; shorter wafer-travel distances; and the option of single-sided cleaning. The biggest challenge is cleaning, rinsing, and drying in 1-2 min.
Critical on the cleaning side has been the development of a megasonic cleaning system that removes small particles without damaging the fine structures contained on the wafer. On the drying side, a fast, high-performance, watermark-free drying technique is necessary.
The cleaning roadmap is moving toward more environmentally friendly and cost-effective cleaning, implying a fast single-chemistry clean to reduce consumption of ultrapure water (UPW) and chemicals. An ideal cleaning solution combines megasonics, a single chemistry, and single-wafer cleaning.
Megasonics for nano-particles
Megasonics is a well-known cleaning method that uses acoustic waves in the MHz range to remove contaminants from wafers throughout the manufacturing process. Although the technique has been widely used, the detailed mechanism of how it removes particles is not yet clear. Factors such as acoustic streaming (induced flow in the cleaning solution), cavitation (the level of dissolved gases), and oscillatory effects are all believed to affect particle adhesion. Various sources have indicated particle size limitations of the megasonic cleaning technology. Some have stated that using state-of-the-art megasonics would not be adequate to remove particles with diameters <100nm [1]. Researchers at IMEC have concluded, however, that it is not a technology limitation, but an optimization issue. They compared the efficiency of two megasonic systems to remove SiO2 particles ranging from 30 to 140nm from silicon substrates [2, 3]. A KLA-Tencor SP1-TBI light scattering inspection tool was used to measure particles larger than 80nm [4], while light-scattering haze signals [5] were needed to measure smaller particles. The particle removal efficiency of the first megasonic system dropped as particles decreased in size. Specifically, 70% of 80nm particles were removed from the wafer, while only 20% of 30nm particles were. The second megasonic system, Verteq's Goldfinger, demonstrated particle removal efficiencies between 95 and 100% for particle sizes ranging from 30 to 140nm.
One concern about megasonics centers around the potential for physical damage that can cause disrupture of narrow lines on the wafer (Fig. 1). It is believed that this damage is related to the energy release of collapsing cavitation bubbles. The proprietary optimized megasonic system overcomes this issue with controlled and uniform megasonic energy distribution to the wafer surface. This allows for nanometer-particle removal — as well as particle removal from sensitive structured wafers — without damage.
 Figure 1. During wafer cleaning, megasonic agitation can potentially create physical damage to fine structures on the wafer surface. |
Fine aluminum lines were used as a vehicle to evaluate the removal of particles on patterned wafers. Most of the defects observed with a KLA-Tencor-AIT defect inspection tool resulted from the physical vapor deposition (PVD) process. These wafers were then subjected to a megasonic cleaning treatment in ultrapure water. High particle removal efficiency was observed. The wafer map of the "added" defects was checked with an optical inspection method and no wafer damage was observed. The effectiveness of this cleaning method is indicated in Fig. 2, where patterned wafer inspection shows a significant reduction in defects on patterned aluminum meander structures.
A single-chemistry clean
In traditional RCA cleaning, an ammonia peroxide mixture (APM) is used to remove organic contaminants and particles by continuously forming and dissolving a hydrous silicon oxide film (SC1 step) [6]. Because many metals are insoluble in this solution and re-deposit on the wafer surface, a second step (SC2 step) is needed, using a hydrochloric acid peroxide mixture (HPM). The challenge is to optimize the first step in such a way that metal deposition is prevented, thus allowing omission of the second step.
In order to improve the AMP clean with respect to metallic contamination, complexing agents were added to the ammonia peroxide mixture (APM+). Studies were performed to verify the effectiveness of APM+ in removing metal contaminants and preventing re-deposition. Because the effectiveness of APM for organic contaminants has already been proven [6], organic contaminants were not tested with APM+. It is very likely that complexing agents in APM+ will not negatively influence this capability.
 Figure 2. Where cleaned with this optimized megasonic method, a significant reduction in defects on patterned aluminum meander structures was indicated. |
APM concentrations and temperatures were varied and combined with the complexing agents, which were held at a fixed concentration. Parasitic metal deposition from the solution onto a clean wafer surface was first investigated. This was tested in several different experiments. Figure 3a shows the result of one of these tests in which APM and APM+ cleaning mixtures (0.25/1/5) were intentionally spiked with Fe, Cu, Zn, and Al at a weight concentration of 1ppb. It was observed that after a 10-min immersion at 50°C, followed by a 10-min overflow rinse and dry, the deposition of metal contaminants from the cleaning mixture to the wafer was significantly lower with the APM+ formula, relative to the standard APM (Fig. 3a).
While the prevention of metal re-deposition is important, the ultimate goal of the cleaning step is metal removal. To test this, the wafers were intentionally contaminated by exposing them to metal spiked solutions (Ca, Fe, Cu and Zn), resulting in typical contamination levels on the order of 1012 atoms/cm2 on the wafer surface. Figure 3b shows that most metals could not be removed with the standard APM solution, except for copper, which was removed to some extent by formation of soluble amine complexes. With the optimized APM+ mixture, metal cleaning was enhanced. The improvement of metal removal was independent of the APM concentration, though there is a general trend toward increased removal with higher temperatures [7]. Chelating agent concentrations were held at a fixed level in this test, but concentrations were varied in other tests and, as expected, metal removal increased with higher concentrations of complexing agents. Since modification of the APM solution did not adversely affect the other properties of APM, particle neutrality and removal efficiency were studied for APM+ and nonmodified APM. Both characteristics appeared unchanged [8].
Finally, and most important, the impact on GOI was studied using capacitor structures with an area of ~10mm2. The intent was to learn how the addition of complexing agents to standard APM affects GOI. Comparisons were made between APM and APM+ formulas, with and without the addition of metal contaminants. Wafers were cleaned using each cleaning formula at 50°C. As indicated in Fig. 4 (left), the addition of complexing agents to an ultrapure APM had no adverse effect on GOI. With the addition of trace amounts of metal contaminants (weight concentration of 1ppb), it was shown that the presence of complexing agents could actually improve the GOI yield from essentially zero to close to 100% (Fig. 4, right). These findings indicate that the APM+ single-chemistry clean may be a viable alternative for standard front-end-of-line (FEOL) cleans [7, 8].
Fast drying
Sub-100nm design rules require fast and efficient rinsing and drying techniques for single-wafer processing. Among the current techniques, spin-drying can deliver high throughput, but suffers from inadequate performance. This is due to the high-speed rotation that is used, which can result in static charging and particle addition from the ambient. Moreover, spin-drying involves considerable amounts of evaporation, resulting in a higher risk for residue and watermark formation. Furthermore, with reduced rinsing times, contaminants from the last cleaning step may remain close to the wafer surface and re-deposit when the film evaporates during the subsequent drying step. This contamination can be observed optically as "drying marks" or "stains" on the wafer.
Isopropyl alcohol (IPA)-vapor dryers, on the other hand, have better particle performance, but cannot meet the required drying time, which should be well below 1 min for a 300mm wafer. Furthermore, IPA vapor dryers are characterized by large IPA consumption and vapor emission. The use of high-temperature (near boiling point) IPA in this method also holds severe safety risks. This translates into high cost of ownership and environmental, safety and health issues. One way to overcome these problems is with a drying technique called Rotagoni. It is based on the Marangoni effect, a surface tension gradient effect that creates a force on parts of the liquid, resulting in very effective local scale liquid removal, while minimizing evaporation.
Using this drying technique, a horizontally placed wafer rotates at low speed (between 300 and 1000rpm) while ultrapure water and a tensioactive vapor are dispensed onto the wafer surface. (A tensioactive vapor is a vapor that, when dissolved into the liquid, reduces the surface tension of the liquid; for example, IPA vapor in nitrogen.) Low rotation speeds effectively reduce the amount of splash back and turbulence above the wafer, which can reduce static charging and entrapment of airborne particles.
Due to the Marangoni effect, the amount of evaporation taking place on the wafer surface during drying is minimized because the liquid is physically removed rather than evaporated. This was shown in an experiment comparing the evaporated film thickness for three spin techniques and Rotagoni drying. Film thickness was determined with a concentrated tracer solution (1000ppm KCl) dispensed on a wafer and dried using each technique. The Rotagoni process was performed in a worst-case mode because the ultrapure water normally used in this drying method was replaced by the same KCl solution. The amount of KCl remaining on the wafer was expected to be even lower, making evaporation and staining more likely.
The concentration of K and Cl after drying was analyzed by total-reflection X-ray fluorescence (TXRF), which calculates the amount of liquid evaporated. The estimated average evaporated film thickness determined in this way was a few microns for spin-drying, but below 50nm for Rotagoni drying (Fig. 5). Since the value obtained with Rotagoni drying was so low, it may still contain a significant component from the adsorption taking place during the KCL rinse prior to drying. This implies that the value obtained for Rotagoni may still be an upper-limit for the actual amount of evaporation taking place.
In a further comparison, spin drying in an IPA ambient, a variant of spin-drying, was evaluated using the Rotagoni set-up with the liquid supply turned off. The thickness of the evaporated liquid film was somewhat smaller than with conventional spin-drying, but still much higher than with Rotagoni drying [9]. Therefore, less staining is expected with the Rotagoni drying.
Conclusion
Combining a megasonic system with a Rotagoni dry potentially meets the critical cleaning applications of FEOL and BEOL, cleaning sensitive structures without damage and eliminating watermarks and residues commonly associated with a spin dry.
Acknowledgments
Rotagoni is a trademark of IMEC. Goldfinger is a trademark of Verteq.
References
- M. Olim, "A Theoretical Evaluation of Megasonic Cleaning for Submicron Particles, J. Electrochem. Soc., 144: 3657-3659, 1997.
- J. Lauerhaas et al., "Megasonic, Non-contact Cleaning Followed by Rotagoni Drying of CMP Wafers," UCPSS S Ost Belg sep 76-77: 251-254, 2001.
- J. Lauerhaas et al., "Single Wafer Cleaning and Drying: Particle Removal Via a Non-contact, Non-damaging Megasonic Clean Followed by a High Performance Rotagoni Dry," ISSM S Tok Sept, 2000.
- F. Holsteyns et al., "Optimization of Cleaning Processes for 100nm and Smaller Design Rules," YMS, Taiwan-Singapore, 2001.
- K. Xu et al., "Optimization of a Brush Scrubber for Nano-Particles," ECS S SanFran USA abstract nr. 1421, 2001.
- W. Kern et al., "Cleaning Solutions Based on Hydrogen Peroxide for Use in Silicon Semiconductor Technology," RCA Rev., 187-205, 1970.
- R. Vos et al., "Single Chemistry Cleaning Solution for Advanced Wafer Cleaning," UCPSS S Ost Belg Sept. 76-77: 119-122, 2001.
- B. Onsia et al., "Advanced Single Chemistry Alkaline Cleaning in a Singletank Tool," ECS S SanFran USA abstract nr. 1441, 2001.
- W. Fyen et al., "Reduction of Surface Metallic Contamination through Optimized Rinsing and Single-wafer Drying," ECS S SanFran USA abstr. 1437, 2001.
Paul W. Mertens obtained his masters degree in electrical-mechanical engineering and his PhD in applied sciences from the KULeuven. Since joining IMEC in 1984, he has investigated silicon wafer surface quality, thin gate dielectrics, cleaning processes, and contamination metrology. He is leader of the Ultraclean Processing Group.
Els Parton received her engineering degree and her PhD in applied biological sciences from the KULeuven. She joined IMEC in 2001 as scientific editor in the Business Development Div., IMEC, Kapeldreef 75, 3001 Leuven, Belgium; ph 32/16-281-467, [email protected].