Post-chemical mechanical polishing cleaning of silicon wafers with laser-induced plasma

Transcription

1 J. Adhesion Sci. Technol., Vol. 18, No. 7, pp (2004) VSP Also available online - Post-chemical mechanical polishing cleaning of silicon wafers with laser-induced plasma VAMSI KRISHNA DEVARAPALLI 1,YINGLI 2 and CETIN CETINKAYA 3, 1 Department of Mechanical and Aeronautical Engineering, Wallace H. Coulter School of Engineering, Clarkson University, Potsdam, NY , USA 2 IBM Microelectronics Division, Hopewell Junction, NY 12533, USA 3 Department of Mechanical and Aeronautical Engineering, Center for Advanced Materials Processing, Wallace H. Coulter School of Engineering, Clarkson University, Potsdam, NY , USA Received in final form 15 March 2004 Abstract During chemical-mechanical polishing (CMP), wafers are subjected to various particle sources such as slurry, polishing pads and polishing machines. Consequently, wafer post-cmp cleaning is crucial in micro- and nano-manufacturing to improve yield. In conventional cleaning process, the wafer is subjected to forces whose magnitudes are large enough to potentially cause substrate damage. This damage concern is becoming more severe as the characteristic feature size shrinks to sub-100-nm region. The development of dry, rapid, non-contact and non-destructive particle removal methods has been emerging as a critical requirement for post-cmp cleaning. In recent years a novel technique for particle removal using the pressure field generated by pulsed laser-induced plasma (LIP) has been introduced for cleaning surfaces and trenches. In the current study, it is demonstrated that the LIP removal technique is capable of removing ceria CMP particles with size of 100 nm and above without any substrate damage. The results reported in this study prove that LIP can also be applied over extended areas for post-cmp cleaning. It is possible to remove smaller particles at lower gap distances but the minimum particle size that can be removed is limited by the risk of substrate damage. Results reported in this study indicate that the LIP particle removal technique has significant potential for post-cmp cleaning in the near future. Keywords: Nanoparticles; ceria; particle removal; laser cleaning; chemical-mechanical polishing (CMP); micro-contamination; nano-contamination. 1. INTRODUCTION Chemical-mechanical polishing (CMP) is an established technology for the planarization of dielectric and metal layers in micro- and nano-manufacturing. During To whom correspondence should be addressed. Tel.: (1-315) Fax: (1-315) cetin@clarkson.edu

2 780 V. K. Devarapalli et al. CMP, wafers are subjected to various particle sources such as slurry, polishing pads and polishing machines. Wafer cleaning after CMP, therefore, is a critical step in micro- and nano-manufacturing in order to minimize yield loss. As the complexity of nano-fabrication increases, surface cleaning for post-cmp and during many other processes, such as reactive ion etching and thin-film deposition, becomes extremely crucial in many industries, including semiconductors, optics, photonics and microelectromechanical systems (MEMS). A dry, rapid, non-contact and non-destructive particle removal method is desirable to reduce the risk of substrate damage and chemical usage while removing smaller particles. Minimizing the use of chemical agents in LIP cleaning is also a major advantage for workplace safety and environmental conservation. A brief review of the laser removal techniques based on substrate acceleration is presented. Pulsed-laser particle removal methods based on substrate acceleration and shock-wave generation from superheated thin liquid films have been employed in the past decade. It is known that a direct-pulsed laser irradiation generates high surface accelerations and strong shear waves in the near field [1] due to rapid thermal expansion in a thin top layer. Particle removal using pulsed lasers has been reported in numerous contexts since it was first reported in 1991 [2, 3]. However, possible substrate damage in direct-pulsed irradiation has been a concern for submicrometer particles [4]. Laser steam cleaning, in which particle removal is due to superheating of a liquid film deposited on the substrate to nucleate cavitation bubbles under pulsed laser heating, appears more effective [5] than the directirradiation method. However, it is inadequate when dry cleaning of a surface is required and/or particle-liquid/surface-liquid interactions at high temperatures are not desirable, for instance in the removal of dry pharmaceutical powders. Direct laser cleaning has been applied to a large number of cleaning applications such as wafers, disk drive heads, stencils and sputtering targets [6, 7]. In direct surface irradiation methods, recent interest has been in thermo-elastic stresses generated on the substrate [4]. Also, the interaction between the laser beam and the particle (near-field diffraction), and consequent substrate damage has been a concern [8, 9]. It is feared that this interaction presents a serious challenge for the direct method in nanoparticle removal applications since lasers with short wavelengths are preferred in direct laser cleaning due to their higher absorption coefficients. In recent years a novel method for particle removal using the pressure field generated by pulsed laser-induced plasma (LIP) has been introduced for cleaning surfaces [10 12] and trenches [13]. The schematic in Fig. 1 depicts a simple set-up for the technique. In the experiments reported in the current study, an incident laser beam is focused through a convex lens. Near the focal point of the lens, the energy density is large enough to cause the breakdown of air into plasma [14]. The pressure field generated by the expanding plasma core acts over the surface of the particle, creating an equivalent moment about the contact zone [15]. If the moment due to the pressure field is strong enough, an interfacial crack is initiated between the particle and the substrate. The crack propagates at the leading edge and closes at the trailing

3 Post-CMP cleaning of silicon wafers with LIP 781 Figure 1. Schematic of the laser-induced plasma removal set-up. d is the distance from substrate to the center of the beam, f l the focal length of the lens and r 0 the incident beam radius. edge as the particle rolls. The removal of the particle begins when the particle loses contact with the surface due to various reasons, such as interactions with other particles and surface features. It is observed that the removal effectiveness depends heavily on the distance between the center of the plasma and the substrate, since it dictates the magnitude of the applied pressure field on the surface of the particle [11]. Moving the substrate closer to the blast center increases the magnitude of the moment experienced by the particle, which increases the effectiveness of the removal process. However, the risk of substrate damage increases as the substrate is moved closer to the plasma core. These two issues must be successfully mitigated for damage-free particle removal. The LIP experiments performed to date indicate that the main reason for surface damage is the physical contact between the plasma and the substrate. The required pressure levels predicted in Ref. [15] and measured in Ref. [16] are in the range of kpa, which are low compared to the yield strengths of typical substrate materials (e.g. silicon). Therefore, it is reasonable to conclude that surface damage risk due to the magnitude of shockwave pressure is low and the main cause of damage is chemical reactions between the plasma and the surface. 2. EXPERIMENTAL PROCEDURE FOR LASER-INDUCED PLASMA REMOVAL The substrates used in the experiments were reclaimed 6-inch, [111] n-type doped silicon wafers with approximately 1-µm-thick thermal oxide layer. The wafer was cut into small square pieces of 1.5 cm 1.5 cm area using a diamond scriber to avoid the geometrical constraints, while the sample was imaged and analyzed under a scanning electron microscope (SEM). These samples were individually washed with de-ionized water and ethyl alcohol to get rid of the initial contamination. In order to locate the same area before and after pulsed-laser irradiation, a diamond-shaped reference pattern was engraved on the square wafer piece with the same diamond

4 782 V. K. Devarapalli et al. Figure 2. Particle size distribution bar chart (V p ) from the ultrafine particle size analyzer (MICROTRAC-UPA150) indicates that the ceria slurry used in experiments has particles with diameters (D) in the range of nm. The left y-axis corresponds to the cumulative volume percentage (C v ) of the slurry passing through the analyzer channels versus the particle diameter (D). scriber. The shape of the pattern was observed to change with the application of LIP, which hindered the process of locating the exact cleaned area. To eliminate this problem, prior to slurry deposition the bare sample piece was cleaned by laser shots at 1.4 mm firing distance with 5 pulses over the diamond-shaped area. Commercially available ceria slurry supplied by NYACOL Nano Technologies (Ashland, MA, USA), with a particle size in the range of nm was used (Fig. 2). The slurry was diluted with methanol in order to reach a concentration where particles could be deposited with minimal agglomeration. The diluted CMP slurry was deposited onto the silicon wafer using the drop-agitation technique. The objective of deposition was to achieve a uniform distribution of slurry particles with minimal agglomeration and proper density to facilitate analysis in the target areas of the wafer. The wafer pieces were attached to a rigid post positioned in a jewelrycleaning machine. The surface vibrations in the jewelry-cleaning machine were used in order to reduce particle aggregation while the suspension was drying on the surface. The diluted slurry was applied to the central area of the wafer. The wafer was allowed to completely dry following deposition. To demonstrate that LIP could be applied over extended areas the complete 6-inch wafer was taken, cleaned and slurry was deposited according to the procedure explained above and then the pulsed-laser irradiation was carried out. Both initial cleaning and deposition were conducted in a class-10 cleanroom. The wafers were then taken to the laser setup outside of the cleanroom for application of the LIP. The pulsed laser utilized in the experiments was a Q-switched Nd:YAG operating at a fundamental wavelength of 1064 nm (Quantel Brilliant series Q44)

5 Post-CMP cleaning of silicon wafers with LIP 783 with pulse energy of 370 mj. It had a pulse length of 5 ns, a repetition rate of 10 Hz, and a beam diameter of 5 mm. A 25 mm diameter, 100 mm focal length lens with a 1064-nm-thick specific antireflective coating was used to converge the beam. The wafer was placed onto a translation stage and held in place with a light adhesive tape on an aluminum stud. Horizontal translation was achieved in two dimensions using sliding posts with millimeter markings. More exact translation was required in the vertical dimension to control the critical parameter d (Fig. 1). A linear translation stage with a 20 ± 10 µm resolution was used for this purpose. In this particular experimental setup, d was set at 1.4 mm. A He-Ne laser was employed for positioning of the lens and vertical alignment of the sample with the Nd:YAG beam. A diode laser was used to mark the horizontal position of the plasma and to align the sample. Two sets of experiments were carried out, one on a 6-inch wafer and the other on a small wafer piece; this is because the 6-inch wafer could not be analyzed under the SEM, because of the geometrical limitations in the SEM chamber. The pulsed-laser irradiation procedures for the two samples were as described below Irradiation procedure for the wafer piece sample The wafer was positioned on the translation stage and the required test area where LIP was to be applied was identified with the diode laser. Previous experimental results showed that a single pulse affected a circular area with a diameter of about 2 3 mm [11]. As the test area was approximately 2 mm 2, a single attempt was sufficient to cover the whole test area. In this particular experiment 10 pulses were imparted Irradiation procedure for the 6-inch wafer The wafer was placed on the translation stage. LIP was performed at the center of the wafer in an 8 mm 8 mm grid (Fig. 3a). The size of the grid compared to the wafer is depicted in Fig. 3b by a square with white lines. The wafer was positioned so that the lower left grid corner was the first to be cleaned; the laser irradiation then proceeded sequentially over the entire grid for cleaning the entire zone with an area of 8 mm 8 mm. After each alignment, a single pulse was applied to each point on the grid. Optical microscope (Olympus model BH2) images were captured on the entire 6-inch wafer. Particle counting was achieved using a surface analysis system (SAS) (PMS SAS 3600 XP by Particle Measuring Systems, Boulder, CO, USA). SEM analysis was then conducted to characterize the particles and to obtain the before and after images of smaller square shaped wafer pieces. A JEOL model 6300 microscope was employed for capturing SEM images.

6 784 V. K. Devarapalli et al. (a) (b) Figure 3. SAS analyses of the silicon substrate before LIP exposure (a) with the 8 mm 8mm cleaning grid in the inset and after LIP exposure (b). The white square in (b) indicates the cleaning area grid.

7 3. PARTICLE REMOVAL RESULTS Post-CMP cleaning of silicon wafers with LIP 785 The ceria slurry was analyzed in an ultrafine particle analyzer (MICROTRAC- UPA150, Honeywell, Phoenix, AZ, USA). Figure 2 shows the particle size distribution (V p ) chart of the CMP slurry used in the experiments. It was observed that the slurry had particles with diameter D in the range of nm. The findings of the experiments for the 6-inch wafer and the wafer piece samples are summarized below Results from the 6-inch wafer Single pulses were imparted on the 6-inch wafer. In Fig. 4, the before and after optical microscope pictures of the affected areas of the silicon wafer at 100 magnification factors are presented. These images verify that larger particles and agglomerates in these areas have been removed. It is evident that the LIP technique at a gap distance of d = 1.4 mm has been quite successful in removing the smaller particles also, including the clumped particles. Since smaller particles are not clearly visible under the optical microscope and as the 6-inch wafer has a geometrical constraint to be analyzed under SEM, so experiments with small wafer pieces were carried out with a higher number of pulses to check the removal of particles Results from the wafer piece sample In Fig. 5, a set of SEM images of the deposited ceria slurry particles on the wafer is presented. These images were captured for the characterization and verification of the particle size distribution in the deposited ceria slurry. The characterization process is also important for a more complete understanding of removal from the entire area and nature of the deposited particles. The SEM images confirm a particle diameter (D) range of approximately nm. This estimation seems to agree with the range of particle diameters found using the ultrafine particle analyzer in Fig. 2. Figure 6 presents a set of SEM images of the substrate before and after LIP removal. It is evident that a vast majority of the particles deposited on the marked areas on the wafer has been removed. While no damage is detected by the SEM analysis in the cleaning areas, there is some substrate damage observed near the walls of the inscribed diamond reference, which can be attributed to the change in the surface roughness imparted by the diamond scriber while marking the reference. 4. CONCLUSIONS AND REMARKS It is demonstrated that the laser-induced plasma (LIP) removal technique is capable of removing particles as small as 100 nm present in the ceria CMP slurry from a silicon wafer. The objective of the study was to evaluate the potential of LIP as a post-cmp cleaning technique. The wafer used as a substrate was n-type

8 786 V. K. Devarapalli et al. (a) (b) Figure 4. Optical microscope images (100 ) of the silicon substrate (a) before and (b) after LIP exposure.

9 Post-CMP cleaning of silicon wafers with LIP 787 (a) (b) Figure 5. SEM images of deposited ceria slurry on the silicon wafer showing the particle size distribution with (a) and (c) at 4000, (b) at 5000 and (d) at magnifications. The approximate size of some of the individual particles measured by SEM is shown in panels (a), (b) and (c).

10 788 V. K. Devarapalli et al. (c) Figure 5. (Continued). (d)

11 Post-CMP cleaning of silicon wafers with LIP 789 (a) (b) Figure 6. SEM images of ceria CMP slurry particles deposited on a silicon wafer. (a d) at 1000, (e h) at 3000 and (i, j) at 6000 magnifications are the pairs of images before and after LIP exposure. Approximate size of some of the individual particles are identified on all the before images. The dot-dash lines indicate the boundaries of the markers and the cleaning area.

12 790 V. K. Devarapalli et al. (c) Figure 6. (Continued). (d)

13 Post-CMP cleaning of silicon wafers with LIP 791 (e) Figure 6. (Continued). (f)

14 792 V. K. Devarapalli et al. (g) Figure 6. (Continued). (h)

15 Post-CMP cleaning of silicon wafers with LIP 793 (i) Figure 6. (Continued). (j)

16 794 V. K. Devarapalli et al. doped [111] silicon with an approximately 1-µm-thick thermal oxide layer. The data from the ultrafine particle analyzer confirmed that the particles in the ceria slurry ranged from 30 nm to 400 nm, and the SEM analysis conducted for removal verification confirmed the particle sizes to range from roughly 35 nm to 450 nm. The particles were deposited on the substrate with the drop-agitation technique. In the after cleaning SEM images, no substrate damage was detected at the 1.4-mm gap distance. Successful experiments were performed to prove that LIP could also be applied over extended areas for post-cmp cleaning. It is possible to remove ceria particles smaller than 100 nm at lower gap distances, but the lower limit for a gap distance is constrained by the risk of substrate damage. Further trials are needed to verify the ability of the LIP technique to remove particles smaller than 100 nm. The current results provide a good indication that the LIP removal technique has significant potential in post-cmp cleaning as a near-future cleaning technique. Acknowledgements The authors acknowledge the National Science Foundation (Nanoscale Exploratory Research Program, Award ID ), the New York State Science and Technology Foundation, and the Center for Advanced Materials Processing (CAMP) for their partial financial supports for this research project. REFERENCES 1. C. Cetinkaya, C. Li and J. Wu, J. Sound Vibr. 231, (2000). 2. W. Zapka, W. Ziemlich and A. C. Tam, Appl. Phys. Lett. 58, 2217 (1991). 3. K.Imen,S.J.LeeandS.D.Allen,Appl. Phys. Lett. 58, 203 (1991). 4. J. Lin and C. Cetinkaya, J. Adhesion Sci. Technol. 17, (2003). 5. M. Mosbacher, V. Vobler, J. Boneberg and P. Leiderer, Appl. Phys. A. 70, (2000). 6. X. Wu, PhD Thesis, Universite de Montreal, Montreal, Québec (1999). 7. Y. F. Lu, W. D. Song, M. H. Hong, B. S. Teo, T. C. Chong and T. S. Low, J. Appl. Phys. 80, (1996). 8. B. S. Lukyanchuk, Y. W. Zheng and Y. F. Lu, Proc. SPIE 4065, (2000). 9. M. Mosbacher, H. J. Münzer, J. Zimmermann, J. Solis, J. Boneberg and P. Leiderer, Appl. Phys. A 72, (2001). 10. J. M. Lee and K. G. Watkins, J. Appl. Phys. 89, (2001). 11. R. Vanderwood, MS Thesis, Clarkson University, Potsdam, NY, USA (2002). 12. C. Cetinkaya, R. Vanderwood and M. Rowell, J. Adhesion Sci. Technol. 16, (2002). 13. R. Vanderwood and C. Cetinkaya, J. Adhesion Sci. Technol. 17, (2003). 14. P. D. Maker, R. W. Terhune and C. M. Savage, in: Proc. 3rd International Quantum Electronics Conference, P. Grivet and N. Bloembergen (Eds), Vol. 2. Columbia University Press, New York, NY (1963). 15. T. Hooper and C. Cetinkaya, J. Adhesion Sci. Technol. 17, (2003). 16. C. Cetinkaya and M. D. Murthy Peri, Nanotechnology 15, (2004).

17