Hydrogen-induced surface blistering of sample chuck materials in hydrogen plasma immersion ion implantation

Transcription

1 Hydrogen-induced surface blistering of sample chuck materials in hydrogen plasma immersion ion implantation Paul K. Chu a) and Xuchu Zeng Department of Physics and Materials Science, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon, Hong Kong Received 29 January 2001; accepted 14 May 2001 Hydrogen plasma immersion ion implantation PIII coupled with ion cut is an economical way to synthesize silicon-on-insulator wafers. In order to avoid premature surface blistering caused by the coalescence of hydrogen microcavities, the implantation temperature must be low ( 300 C), and sample cooling is usually required due to the high ion flux in hydrogen PIII. In addition, the entire sample chuck including the silicon wafer and all the exposed surfaces are bombarded by ions and sputtered impurities from the sample holder can be reimplanted or deposited onto the silicon wafer. Ideally, the problem can be solved if the sample chuck is made of silicon but engineering a silicon sample chuck with sufficient electrical conductivity and a cooling mechanism is very complicated. In addition, the hydrogen ions implanted into the exposed silicon chuck surface can cause surface blistering and exfoliation similar to the silicon wafer. The silicon particles released into the vacuum chamber will reduce the process yield. One practical approach is to engineer the sample chuck with stainless steel and then coat the surface with a material compatible with silicon. If the blistering resistance of the coating is better and the lifetime of the coating is sufficiently long, periodic cleaning can ensure particle and contamination free operation. In this work, we investigate the blistering behavior of three such materials, single-crystal silicon, polycrystalline/amorphous silicon, and silicon dioxide. Our results show that silicon dioxide is the best candidate, followed by polysilicon. However, the insulating nature of silicon dioxide must be considered. Our theoretical simulation results show that an oxide layer several micrometers thick will not affect the surface potential significantly even at a relatively low bias voltage American Vacuum Society. DOI: / I. INTRODUCTION As the dimensions of modern integrated circuits approach the deep submicrometer regime, silicon-on-insulator SOI is often preferred over bulk silicon as the substrate to reduce or eliminate deleterious effects such as alpha-particle soft errors, latchup, short-channel effects, source/drain punchthrough, and hot carriers. 1,2 It is the more desirable substrate material for low-power high-speed complementary metal oxide silicon CMOS chips. There are several competing techniques to produce SOI and hydrogen plasma immersion ion implantation PIII coupled with ion cut has been demonstrated to be an effective and economical means. 3 5 The biggest advantage of PIII compared to conventional beamline ion implantation is its high throughput due to the high ion flux ( cm 2 s 1 ) and that the implantation time is independent of the wafer size. 6,7 However, while the lack of an ion filter reduces the cost of a PIII instrument, all ion species, including contaminants that are sputtered or etched from all exposed surfaces including the inner wall of the vacuum chamber and sample stage, are coimplanted into the silicon wafers Fig. 1. Even though the sputtering yield of hydrogen is relatively small, the dose required in the ion-cut process is quite high between and cm 2 and there exist heavier atmospheric contaminants C, O, etc. a Author to whom all correspondence should be addressed; electronic mail: paul.chu@cityu.edu.hk in the residual vacuum because PIII equipment is typically not designed for ultrahigh vacuum UHV operation. 8 The interior wall of the vacuum chamber can be shielded by a liner made of aluminum or silicon, 9 but it is difficult to cover the sample stage entirely. Metallic impurities sputtered from the exposed areas will be coimplanted or deposited onto the silicon wafer. The obvious solution is to manufacture the sample stage out of silicon. However, because of the high ion flux in PIII, sample heating can be quite severe, especially in a commercial environment that demands a high sample throughput. Integration of a mechanical cooling system into a silicon sample chuck is very complicated. The alternative is to manufacture the sample chuck using stainless steel 10,11 but coat the exposed surface with a material compatible with silicon such as single-crystal silicon, polycrystalline or amorphous silicon, or silicon dioxide. However, similar to the silicon wafer, hydrogen implantation into the materials can form shallow buried microcavities, 12 and, if the accumulative implant dose is high enough, surface exfoliation or blistering can result. Silicon particles emitted from these surfaces into the vacuum chamber can reduce the yield of the implantation process. Hence, it is important to study the blistering characteristics of these three materials after hydrogen plasma immersion ion implantation. In addition, the insulating nature of silicon dioxide may affect the ion trajectories as well as the implant dose and energy uniformity. In this work, we 2301 J. Vac. Sci. Technol. A 19 5, SepÕOct Õ2001Õ19 5 Õ2301Õ6Õ$ American Vacuum Society 2301

2 2302 Paul K. Chu and Xuchu Zeng: Hydrogen-induced surface blistering 2302 TABLE I. Experimental parameters and visual classification of surface blistering using optical microscopy 500. Normal implantation conditions: atoms/cm 2, T 150 C Si single crystal silicon, SiO nm SiO 2 on silicon, poly-si 300 nm polysilicon on silicon. Sample no. Annealing conditions Dose Observation FIG. 1. Schematic of plasma immersion ion implantation showing that ions bombard all surfaces of the sample chuck. conduct experiments to systematically investigate the hydrogen-induced surface blistering phenomenon in singlecrystal silicon, polysilicon, and silicon dioxide as well as theoretical calculation to determine the effects of the oxide thickness on the surface potential under different bias voltages. II. EXPERIMENT Silicon wafers 150 mm in diameter were implanted with different hydrogen doses at below 150 C using our semiconductor PIII instrument. 13 They were 1 regular silicon 100 wafers, 2 silicon wafers with 180 nm of thermally grown silicon dioxide on the surface, and 3 silicon wafers with 300 nm of chemical vapor deposited CVD polysilicon on the surface 600 C CVD and 900 C dopant drive in to increase the electrical conductivity. After hydrogen PIII, the samples were annealed at different conditions and examined using an optical microscope to investigate the extent of surface blistering. The experimental parameters are listed in Table I. The visual inspection results that will be discussed in detail in the next section are shown in the last column. In order to evaluate the effects of the degree of crystallinity, a second set of polysilicon wafers 200 nm deposited by CVD at 550 C without the subsequent high temperature dopant diffusion treatment were implanted and investigated. The results are displayed in Table II. These samples should be more amorphous on account of the lower processing temperature. III. RESULTS AND DISCUSSION Figure 2 displays the H, N, and O depth profiles acquired by secondary ion mass spectrometry SIMS froma20kv hydrogen plasma implanted silicon wafer. The presence of coimplanted atmospheric contaminants such as N and O due to outgassing and leaks in the system is quite evident. This is because PIII equipment is usually not designed to operate under UHV conditions. Hence, residual gas contaminants are 1 Si 600 C for 10 min 1 bubbles 2 Si as implanted 0.75 smooth 3 Si 200 C for 13 h 1 smooth 4 Si 200 C for 13 h, and then 600 C 1 bubbles for 10 min 5 Si as implanted 4 bubbles 6 Si as implanted 1.5 starting to blister 7 Si implanted at 300 C 3 bubbles 8(SiO 2 as implanted 1 smooth 9(SiO 2 ) 600 C for 15 min 7 smooth 10 (SiO 2 ) 600 C for 10 h 1 smooth 11 (SiO 2 ) as implanted 5 smooth 12 poly-si 600 C for 35 min 1 bubbles 13 poly-si as implanted 4 bubbles 14 poly-si 250 C for 4 h 1 smooth 15 poly-si as implanted 2 bubbles 16 poly-si 250 C for 4 h, and then 600 C 1 bubbles for 35 min 17 poly-si as implanted 1 smooth inevitable even though the purity of the hydrogen gas used in our experiments is %. Based on the SIMS data, the calculated H, O, and N doses are , , and atoms/cm 2, respectively. In terms of percentage, the relative concentration of oxygen and nitrogen to hydrogen is only about 2%, but N and O are much heavier than hydrogen and cause substantially more sputtering of the sample stage than hydrogen. Table III compares the relative number of iron atoms sputtered by H, O, and N. The sputtering yields by H, O, and N are estimated using TRIM. Inour process, the dominant ion species are H 3,O 2, and N 2. The net implantation voltages are thus 6.7 kv for hydrogen, 10 kv for oxygen, and 10 kv for nitrogen. Our simulation results show that the sputtering yield of either oxygen or nitrogen is a hundred times higher than that of hydrogen. That is, while a hydrogen ion bombarding the stage sputters off the equivalent of iron atom, 0.02 atom of oxygen or nitrogen ion will sputter off and 1.8 TABLE II. Experimental parameters and visual classification of surface blistering on the low temperature low T polysilicon using optical microscopy 100. Implantation conditions: atoms/cm 2, T 150 C low temperature poly-si 200 nm polysilicon on silicon. Sample no. Annealing conditions Dose Observation 20 low T as implanted 1 smooth 21 low T 250 C for 10 h 1 smooth 22 low T 500 C for 15 min 1 bubbles 23 low T 250 C for 10 h, and then 500 C 1 bubbles for 15 min 24 low T as implanted 4 smooth 25 low T 500 C for 15 min 4 large bubbles 26 low T as implanted 10 smooth 27 low T 500 C for 15 min 10 large bubbles J. Vac. Sci. Technol. A, Vol. 19, No. 5, SepÕOct 2001

3 2303 Paul K. Chu and Xuchu Zeng: Hydrogen-induced surface blistering 2303 FIG. 2. SIMS depth profiles of H, O, and N in a 20 kv hydrogen plasma implanted silicon wafer. The projected ranges of oxygen and nitrogen are in line with 10 kev O and N implantation predicted to be about 28 nm for N and 25 nm for O by TRIM iron atom, respectively. Hence, it is obvious that metallic contaminants sputtering from the exposed area of the sample chuck by these adventitious gaseous ions are severe. In fact, our calculation shows that sputtering due to hydrogen only accounts for 10% of the total contamination, and the number of the sputtered iron atoms that can subsequently be ionized in the plasma and erode the sample stage to introduce more contamination is orders of magnitude lower than that by the atmospheric species. Constructing a sample stage entirely from silicon is technologically complicated, and even if it can be implemented the silicon surface can blister after prolonged hydrogen-implantation releasing particles into the vacuum chamber. The alternative is to coat the sample chuck with a protective coating that is compatible with silicon processing and shows higher blistering resistance than single-crystal silicon. The optical micrographs acquired from the four sets of samples: single-crystal silicon, silicon with 180 nm of surface SiO 2, silicon with 300 nm of high temperature polysilicon, and silicon with 200 nm of low temperature poly- or TABLE III. Number of iron atoms sputtered by 6.7 kev hydrogen, 10 kev oxygen, and 10 kev nitrogen calculated by TRIM. The total doses are determined experimentally by SIMS. H O N Total dose cm relative dose sputtering yield relative number of sputtered atoms FIG. 3. Optical micrographs 500 magnification acquired from the singlecrystal silicon samples ( atoms/cm 2 ). amorphous silicon are depicted in Figs The normal hydrogen dose (1 ) described here is a typical value that will result in layer transfer in the ion-cut process. We perform a low temperature anneal on the implanted singlecrystal silicon wafers first to find out whether the implanted hydrogen can be released slowly without giving rise to surface bubbles. The process is equivalent to heating the sample stage in situ. However, as shown in Fig. 3, even a 13 h anneal at 200 C cannot accomplish the purpose because a subsequent anneal at 600 C for 10 min blisters the surface severely samples No. 3 and No. 4. Therefore, a low temperature anneal of hydrogen implanted single-crystal silicon cannot solve the problem. In addition, it is observed that at the normal hydrogen PIII temperature, a 1.5 hydrogen dose will result in surface blistering. Our experiments thus conclude that single-crystal silicon is not the suitable coating material. Figure 4 exhibits the optical micrographs 500 magnification acquired from the SiO 2 samples. The results are much more encouraging because no bubbles can be observed, including the samples annealed at a high temperature samples No. 9 and No. 10 and that implanted with a high dose sample No. 11, 5 dose. The blistering resistance of polysilicon falls between single-crystal silicon and silicon JVST A - Vacuum, Surfaces, and Films

4 2304 Paul K. Chu and Xuchu Zeng: Hydrogen-induced surface blistering 2304 FIG. 4. Optical micrographs 500 magnification acquired from the 180 nm SiO 2 on single crystal silicon samples ( atoms/cm 2 ). dioxide. The optical micrographs of polysilicon samples deposited at 600 C and then undergoing a 900 C predep process are displayed in Fig. 5. The 1 samples do not exhibit blistering after annealing, but the 4 sample No. 13 does. The 100 magnification micrographs acquired from the low temperature polysilicon are depicted in Fig. 6. The 1 and 4 samples do not show surface blister as implanted or FIG. 6. Optical micrographs 100 magnification acquired from the 200 nm polysilicon on single-crystal silicon samples 550 C CVD, atoms/cm 2. annealed at 250 C. However, severe blistering results after 500 C annealing. Figure 7 shows the 50 magnification optical micrographs acquired from the low T polysilicon implanted with 10 dose. Again, severe blistering is observed after 500 C annealing. Our results thus suggest that a SiO 2 coating has the best blistering resistance and is the best coating material for the sample chuck. It is believed that the improvement in the blistering resistance stems from traps in the SiO 2 and polysilicon to tying up free hydrogen thereby impeding coalescence of the hydrogen microcavities. In PIII, the voltage is applied to the sample stage and so it is imperative that the materials making up the sample chuck are electrically conducting. Hence, engineering a sample chuck using entirely quartz is not feasible. The alternative is FIG. 5. Optical micrographs 500 magnification acquired from the 300 nm polysilicon on single-crystal silicon samples 600 C CVD plus 900 C predep, atoms/cm 2. FIG. 7. Optical micrographs 50 magnification acquired from the 200 nm polysilicon on single-crystal silicon samples 550 C CVD, atoms/cm 2. J. Vac. Sci. Technol. A, Vol. 19, No. 5, SepÕOct 2001

5 2305 Paul K. Chu and Xuchu Zeng: Hydrogen-induced surface blistering 2305 FIG. 10. Theoretical calculation of the voltage on the silicon dioxide which thickness varies from 100 nm to 1 cm. FIG. 8. Optical photograph of the as-implanted 150 mm diameter Si wafer. Areas a and c show more severe surface blistering due to higher hydrogen doses locally. to cover the exposed sample chuck with a quartz tube and shroud. Figure 8 depicts an optical micrograph of a hydrogen plasma implanted silicon sample placed on a stainless steel sample stage surrounded by a thick quartz shroud on the side and underneath. The blistered areas, a and c, have been implanted with higher hydrogen doses locally, and our simulation study reveals that the ion dose nonuniformity is caused by insulating nature of the quartz shroud. 14 Even though the use of a bulk insulator can be excluded, a thin SiO 2 coating which does not affect the ion trajectories significantly under high voltage working conditions should suffice. Figure 9 shows the calculated projected range of the various hydrogen ions in the plasma, namely H,H 2, and H 3 in addition to He that is used in some ion-cut experiments despite its FIG. 9. Projected ranges of H,H 2 labeled at H2, H 3 labeled as H3, and He in SiO 2 calculated from TRIM. higher induced damage. From the standpoint of ion penetration or stopping, several micrometers of SiO 2 will suffice in typical hydrogen PIII experiments. Figure 10 shows the calculated surface voltage on the surface of SiO 2 coatings of various thicknesses as a function of the applied voltage. In the calculation, 15 we use an implant dose of cm 2 and a pulse repetition rate of 100 Hz. We assume that the charges stay at the surface during the pulse and are annihilated during the off cycle by electrons from the plasma. The ion species used in the simulation is H which gives rise to the largest charging effects. Surface charging will be less severe for H 2 and H 3 due to the smaller charge to number of hydrogen atoms ratio 1 2 for H 2 and 1 3 for H 3. Our results show that the SiO 2 coating can be as thick as 100 m without appreciable voltage reduction on the surface even at an implantation voltage of 20 kv. Therefore, silicon dioxide appears to be the best coating material polysilicon being second best for the sample chuck. However, it should be noted that we have not conducted extremely high implant dose e.g., 100 experiments to simulate the situation in the field, but even if the SiO 2 coating does exhibit surface blistering at such high implant doses, in situ argon plasma cleaning will be sufficient to expose a new and clean surface. Considering that the thickness of the coating can be as large as 100 m, a well designed and constructed sample chuck will last a reasonably long time. IV. CONCLUSION In order to reduce metallic contamination sputtered from the exposed sample stage, a material compatible to silicon processing must be used to construct the sample chuck. Engineering the sample chuck out of silicon is impractical due to the complexity and surface blistering of silicon after prolonged hydrogen PIII can release particles into the vacuum. Depositing a SiO 2 coating on the sample chuck is a viable alternative. Our experiments show that silicon dioxide pos- JVST A - Vacuum, Surfaces, and Films

6 2306 Paul K. Chu and Xuchu Zeng: Hydrogen-induced surface blistering 2306 sesses the best blistering resistance. Our theoretical calculation suggests that if the coating is thin enough, the surface potential on the oxide coating is not altered even at a low bias voltage. Thus, the ion dose and energy uniformity on the silicon wafer is not affected. ACKNOWLEDGMENTS The authors thank Z. N. Fan for valuable contributions. This work was jointly supported by Hong Kong Research Grants Council Earmarked Grants CERG No or CityU 1003/99E and No or CityU 1032/00E, and City University of Hong Kong Strategic Research Grant No J. P. Colinge, Silicon-on-Insulator Technology: Materials to VLSI Kluver, Boston, MA, H. Vogt, G. Burbach, J. Belz, and G. Zimmer, Semicond. Sci. Technol. 34, P. K. Chu, X. Lu, S. S. K. Iyer, and N. W. Cheung, Solid State Technol. 40, S X. Lu, S. S. K. Iyer, C. M. Hu, N. W. Cheung, J. Min, Z. N. Fan, and P. K. Chu, Appl. Phys. Lett. 71, X. Lu, N. W. Cheung, M. D. Strathman, P. K. Chu, and B. Doyle, Appl. Phys. Lett. 71, P. K. Chu, S. Qin, C. Chan, N. W. Cheung, and L. A. Larson, Mater. Sci. Eng., R. R17, P. K. Chu, N. W. Cheung, C. Chan, B. Mizuno, and O. R. Monteiro, Semiconductor Applications, inhandbook of Plasma Immersion Ion Implantation and Deposition, edited by A. Anders Wiley, New York, 2000, Chap. 11, p Z. Fan, X. Zeng, P. K. Chu, C. Chan, and M. Watanabe, Nucl. Instrum. Methods Phys. Res. B 155, U.S. patent No P. K. Chu and C. Chan, Removable Liner Design for Plasma Immersion Ion Implantation. 10 P. K. Chu, B. Y. Tang, Y. C. Cheng, and P. K. Ko, Rev. Sci. Instrum. 68, J. Matossian, G. A. Collins, P. K. Chu, C. P. Munson, and J. V. Mantese, Design of a PIII&D Processing Chamber, inhandbook of Plasma Immersion Ion Implantation and Deposition, edited by A. Anders Wiley, New York, 2000, Chap. 6, p P. K. Chu and N. W. Cheung, Mater. Chem. Phys. 57, P. K. Chu, S. Qin, C. Chan, N. W. Cheung, and P. K. Ko, IEEE Trans. Plasma Sci. 26, Z. Fan, P. K. Chu, C. Chan, and N. W. Cheung, Appl. Phys. Lett. 73, G. A. Emmert, J. Vac. Sci. Technol. B 12, J. Vac. Sci. Technol. A, Vol. 19, No. 5, SepÕOct 2001