Atomically Controlled Processing for Group IV Semiconductors. Junichi Murota* and Masao Sakuraba

Transcription

1 / The Electrochemical Society Atomically Controlled Processing for Group IV Semiconductors Junichi Murota* and Masao Sakuraba Laboratory for Nanoelectronics and Spintronics, Research Institute of Electrical Communication, Tohoku University, Katahira, Aoba-ku, Sendai , Japan. *Tel: , Fax: One of the main requirements for Si-based ultrasmall device is atomic-order control of process technology. Here, we show the concept of atomically controlled processing for group IV semiconductors based on atomic-order surface reaction control in Si-based CVD epitaxial growth. Self-limiting formation of 1-3 atomic layers of group IV or related atoms in the thermal adsorption and reaction of hydride gases on Si(100) and Ge(100) are generalized based on the Langmuir-type model. Moreover, Si or Si 1-x Ge x epitaxial growth over N, P or C layer already-formed on Si(100) or Si 1-x Ge x (100) surface is achieved and the capability of atomically controlled processing for advanced devices is demonstrated. Additionally, the strain control of the Si 1-x Ge x /Si heterostructure due to stripe patterning is discussed. Introduction Atomically controlled processing for group IV semiconductors has become indispensable for the fabrication of ultrasmall MOS devices and Si-based heterodevices for ultra-large-scale integrations, because high performance devices require atomic order abrupt heterointerfaces and doping profiles as well as strain engineering due to introduction of Ge into Si. Especially for processing involving surface reaction processes like CVD, the advancement of the process technology requires atomic-order surface reaction control. Improvements in the quality of gases and equipment have enabled ultraclean low-temperature CVD processing for atomic-order control (1-3). Our concept of atomically controlled processing for group IV semiconductors is based on atomic-order surface reaction control by CVD (3-5). The final goal is generalization of atomic-order surface reaction processes and creation of new properties in Si-based ultimate small structures which will lead to nanometer scale Si devices as well as Sibased quantum devices. Figure 1 summarizes the capabilities of atomically controlled processing for the fabrication of ultrasmall and nanodevices. Based on the investigation of surface reaction processes, the concept has been demonstrated for high performance Si 0.65 Ge 0.35 channel pmosfets with a 0.12 µm gate length by utilizing in-situ impuritydoped Si 1-x Ge x selective epitaxy on the source/drain regions at 550 o C (6). Additionally, in in-situ doped Si 1-x Ge x epitaxial growth on the (100) surface in a SiH 4 -GeH 4 -dopant (PH 3, or B 2 H 6 or SiH 3 CH 3 )-H 2 gas mixture, the deposition rate, the Ge fraction and the dopant concentration have been expressed quantitatively by modified Langmuir-type rate equations (3,5). In this paper, the surface reactions of hydride gases (SiH 4, GeH 4, NH 3, PH 3, B 2 H 6, CH 4 and SiH 3 CH 3 ) on Si(100) and Ge(100) for atomic-order growth are reviewed based on the Langmuir-type adsorption and reaction scheme. Furthermore, typical atomic layer doping by the epitaxial growth of Si or SiGe over the material already-formed on (100) surface Downloaded on to IP address. Redistribution subject to ECS 111 terms of use (see ecsdl.org/site/terms_use)

2 Fig. 1: Atomically controlled processing for group IV semiconductors for ultrasmall and nanodevices. and their capability of atomically controlled processing for Si-based group IV heterodevices are demonstrated. Additionally, strain control and electrical properties of stripe patterned Si/Si 1-x Ge x /Si(100) heterostructures are discussed for deep submicron stripe width. Self-Limited Surface Reaction of Hydride Gas Self-limiting conditions for some hydride gases on the Si(100) and Ge(100) surfaces are summarized in Table 1. Surface reactions of hydride gases (SiH 4, GeH 4, NH 3, PH 3, B 2 H 6, CH 4 and SiH 3 CH 3 ) on Si(100) and Ge(100) for atomic-order growth are explained based on kinetics of Langmuir-type adsorption models (Fig. 2) (3,5,7). Table 1: Typical self-limiting conditions for some hydride gases on Si(100) and Ge(100)(5, 7) Downloaded on to IP address. Redistribution subject to ECS 112 terms of use (see ecsdl.org/site/terms_use)

3 Fig. 2: Schematic images of (a) self-limited adsorption and (b) self-limited reaction of hydride for atomic-order growth based on Langmuir-type model. In the case of SiH 4 and GeH 4 adsorption on monohydride Si(100) and Ge(100) surfaces respectively, the adsorption proceeds according to Langmuir-type kinetics (Fig. 2(a)). The adsorbed SiH 4 and GeH 4 do not react and desorb by removing SiH 4 and GeH 4 gases in the reactor. In order to perform atomic-layer growth, the reaction must be induced during the short time, in which the next adsorption scarcely proceeds using flash heating. In the other cases, hydride molecules are adsorbed and react simultaneously on the surface based on Langmuir-type model shown in Fig. 2(b). Most representative one is low-temperature atomic-order NH 3 nitridation of Si surface and SiH 4 reaction on Ge(100). Especially, in the case of the SiH 4 reaction, it is found that single atomic layer growth of Si occurs for the hydrogen free surface formed by preheating at 350 o C in Ar, and for the hydrogen-terminated surface with the dimer structure formed by preheating at 350 o C in H 2, as shown in Fig. 3. The density of the SiH 4 reaction sites on the hydrogen-terminated Fig. 3: θ SiH4 time product dependence of Si atomic amount on the Ge(100). The samples were exposed to SiH 4 at 260 o C after H 2 - and Ar-preheating at 350 o C. SiH 4 partial pressures are 100 Pa ( ), 300 Pa ( ) and 500 Pa (, ). The solid and dotted curves are calculated from the equations in Fig. 2(b) with n T =4.5x10 15 cm -2 for H 2 -preheating, n T =6.2x10 15 cm -2 for Ar-preheating, k r =3.0x10-3 s -1 and K=7.5x10-3 Pa -1. Downloaded on to IP address. Redistribution subject to ECS 113 terms of use (see ecsdl.org/site/terms_use)

4 Ge surface with the dimer structure is lower than that on the hydrogen-terminated unreconstructed surface and hydrogen-free surface. In the case of SiH 3 CH 3 reaction, it appears that SiH 3 CH 3 is adsorbed without breaking the Si-C bond at o C. In the case of PH 3 reaction on Si(100) and Ge(100), the PH 3 reaction is suppressed on the hydrogen-terminated Si and Ge surfaces, but PH 3 is adsorbed dissociatively on the hydrogen-free Si and Ge surfaces at 300 and 200 o C, respectively. As a result, the P atomic amount on the surface tends to saturate below one atomic layer. On the Si surface at o C, the P atomic amount tends to saturate to about two or three atomic layers. It was also reported that the P coverage becomes lower under high pressure of H 2 because of the adsorption of hydrogen even at 350 o C. At 450 o C, the P atomic amount is independent of PH 3 partial pressure ( Pa). On the Ge surface at o C, the P atomic amount tends to saturate to about one atomic layer. In the case of B 2 H 6, the B atom amount tends to saturate self-limitedly at around 1.4x10 15 cm -2 (2 AL) at 180 o C. At 500 o C, the B atom amount increases with B 2 H 6 exposure time and exceeds 2 AL. It is clear that continuous B 2 H 6 reaction at 500 o C proceeds with H desorption on B atoms (7). Atomic-Layer Doping in Si and SiGe Epitaxial Growth Atomic-layer doping is performed by Si or Si 1-x Ge x epitaxial growth over the material already-formed on Si(100) or Si 1-x Ge x (100) surfaces. At growth temperatures below 500 o C, atomic-layer doping of a half atomic layer of N, C, P, B is achieved. In their structures, N, C and B atoms are confined within of about 1 nm thick layers, which is in the order of the measurement accuracy. Nitrogen Atomic Layer Doping in Si and Si 1-x Ge x Epitaxial Growth In the case of Si/half atomic layer of N/Si(100) heterostructure growth, the Si film is epitaxially grown by SiH 4 exposure at 500 o C on the nitrided Si(100) surface formed by NH 3 reaction at 400 o C. On the atomic-layer order nitrided Si(100), an incubation period for Si deposition was observed (8). The incubation period is caused by lowering of SiH 4 adsorption and/or reaction rates at the nitrided sites. In Si epitaxial growth on the surface with N atomic amount below 3x10 14 cm -2 after the incubation period, the deposition rate is the same as that on the Si surface without nitridation. After capping with Si, most of the N atoms are buried in the initially nitrided region within the thickness of about 1 nm. That is within the measurement accuracy. It should be noted that N atoms tend to segregate at the grown surface with increasing Si epitaxial growth temperature. On the other hand, if the N atomic amount is 6x10 14 cm -2, amorphous Si is grown. This is caused by the generation of Si 3 N 4 (8). High quality epitaxial growth of multi-layer N-doped Si film composed of N layers of 3x10 14 cm -2 and 3.0 nm thick Si spacer was achieved (9). In such N-doped Si film, N atoms act as a donor. The total sheet carrier concentration increases with increasing the N amount up to 1x10 14 cm -2 /layer, and decreases. In other words, donor activation ratio tends to decrease with increasing the N amount, and the typical ratio is about 0.4 % at the N amount of 5x10 13 cm -2 /layer (10). Such a low donor activation ratio has been also reported for samples doped by ion implantation and annealed (11). Ionization energy of the donor level is estimated to be about mev based on the assumption that there is no acceptor compensation. The Hall mobility is much larger than that (about 100 cm 2 /(V s) at 300 K) of the uniformly P doped Si with P concentration of cm -3, and is as high as that of cm -3 (12) (Fig. 4) (10, 5). Because the local concentration of Downloaded on to IP address. Redistribution subject to ECS 114 terms of use (see ecsdl.org/site/terms_use)

5 Fig. 5: Measured temperature dependence of the Hall mobility for N AL doped Si films. Solid lines are the values for the uniformly doped n-type Si (12). Hall mobility of uniformly P doped Si epitaxial film grown by low-pressure CVD on Si(100) with P concentration of cm -3 is also shown ( ). the ionized donor is estimated to be about cm -3 in 1 nm-thick N doped region, it is considered that scattering in electron transport is reduced by atomic layer doping structure. Moreover, there is a possibility that the carrier mobility is enhanced by highly condensed strain near the N atomic layer doped region. Since N atoms tend to diffuse and a part of them segregate at the surface at 750 o C, the application into device fabrication requires very low-temperature processing. N atoms in Si 1-x Ge x are preferably combined with Si atoms (13). In the Si 0.5 Ge 0.5 epitaxial layer shown in Fig. 5, a N doping dose of 6x10 14 cm -2 is confined in an about 1.5 nm-thick region even after 650 o C heat treatment in contrast to the result for Si cap layer growth on the thermally nitrided Si(100) with a N doping dose of 6x10 14 cm -2 which was found to be amorphous (14). Fig. 5. Depth profiles of Ge fraction and N concentration in the as-deposited and heat-treated N atomic layer doped Si/Si 0.5 Ge 0.5 /Si(100). Downloaded on to IP address. Redistribution subject to ECS 115 terms of use (see ecsdl.org/site/terms_use)

6 Phosphorus Atomic-Layer Doping in Si Epitaxial Growth on Si 1-x Ge x /Si(100) In the case of Si growth on the (100) surface with P atomic amount of cm -2 at a rather low temperature of 450 o C and a rather high SiH 4 partial pressure of 220 Pa, P incorporation into the Si film and the epitaxial growth of heavily P doped Si film have been achieved, although the tailing towards surface is also observed and a part of the P atoms segregate or desorb during Si growth. In such case, heavily P-doped epitaxial Si film on Si(100) with average P concentration of cm -3 are formed with 7 nm-thick spacers. Average carrier concentration reaches as high as 3.7x10 20 cm -3 and the resistivity as low as 2.7x10-4 Ω-cm (Fig. 6) (15, 5). After heat treatment at tempera-tures higher than 550 o C, a reduction of the carrier concentration is observed (16). Because the electrically inactive P atoms such as P clustering are formed by the heat treatment, the higher activity of P for the as-deposited films is expected to be out of thermal equilibrium. Using such atomiclayer doping technique, a very low contact resistivity of about 5x10-8 Ω-cm 2 between Ti and the Si film has been obtained (17). On the other hand, it is found that the Si film is also epitaxially grown on the P layer already-formed on Si 1-x Ge x /Si(100) by Si 2 H 6 exposure at 450 o C (18), and that P atoms of 4x10 14 cm -2 are incorporated in about 1 nmthick region for higher Si 2 H 6 partial pressure for the Si capping layer deposition (19). This means that, in order to suppress the segregation, Si growth at lower temperature and higher Si growth rate becomes effective. Carbon Atomic-Layer Doping in Si/Si 1-x Ge x /Si(100) structure Fig. 6: Heat treatment temperature dependence of the carrier concentration and the resistivity of the P doped epitaxial Si films. Heat treatment time is 60 min. Ge Fraction without C 1.0 Cap. Si Si0.55Ge0.45 Si(100) As Depo. 650 O C H.T. Si Si 0.55 Ge 0.45 Ge Si(100) Depth from Surface (nm) Ge Fraction with C 1.0 Cap. Si C Si0.55Ge0.45 Si(100) Si(100) (a) Si/Si 0.55 Ge 0.45 /Si(100) (b) Si/C/Si 0.55 Ge 0.45 /Si(100) Fig. 7. Depth profiles of Ge fraction and C atom amount in (a) Si/Si 0.55 Ge 0.45 / Si(100) and (b) Si/C/Si 0.55 Ge 0.45 /Si(100) heterostructures before and after heat treatment at 650 o C As Depo. 650 O C H.T. Ge C Si Si 0.55 Ge 0.45 Ge C Depth from Surface (nm) Carbon Atom Amount (x10 14 cm -2 ) Downloaded on to IP address. Redistribution subject to ECS 116 terms of use (see ecsdl.org/site/terms_use)

7 In order to obtain atomic-order abrupt heterointerface, it is essential to suppress the intermixing between Si and Ge at Si/Si 1-x Ge x /Si heterointerface as well as dopant diffusion. It was reported that C introduction into the heterostructures is effective to control lattice strain and B diffusion (20). In the case of the heterostructure without C, an intermixing between Si and Ge is observed after heat treatment at 650 o C (Fig. 7(a)). The intermixing at the heterointerface is effectively suppressed by carbon atomic-layer doping at the heterointerface (21) (Fig. 7(b) (22)). In the case of 40nm thick strained Si 0.55 Ge 0.45 /Si, the strain amount is reduced by the carbon atomic-layer doping of (1-3)x10 14 cm -2 C at the heterointerface and the reduction becomes larger with the heat treatment. While, in the case of 4nm thick strained Si 0.55 Ge 0.45 /Si, strain relaxation as well as intermixing between Si and Ge are suppressed, although Ge fraction for the undoped heterostructure is reduced and the strain comes to be relaxed after the 650 o C heat treatment (22). These results indicate that, by low-temperature atomic layer doping, very high carrier concentration and higher carrier mobility of Si-based group IV semiconductors could be achieved compared with doping under equilibrium conditions, and that suppression of intermixing and strain relaxation at the heterointerface are obtained. Strain Control and Electrical Properties of Stripe-Patterned Si/Si 1-x Ge x /Si(100) Heterostructures Strain engineering has become indispensable for electron and hole mobility improvement of ULSIs as scaling down of device dimension (23). In order to realize the high performance strained Si layer, it is very important to create a defect-free relaxed Si 1-x Ge x layer. By stripe patterning of Si/strained Si 0.6 Ge 0.4 /Si(100) heterostructures (24), both of the visible-raman scattering peaks from the Si 0.6 Ge 0.4 layer in Fig. 8(a) and the UV-Raman scattering peaks from the Si capping layer in Fig. 8(b) are shifted lower with decreasing Fig. 8: (a) Visible and (b)uv Raman scattering spectra obtained from (1)blanket, (2)165nm and (3)120 nm-wide stripe patterned Si (5nm)/Si 0.6 Ge 0.4 (25nm)/Si(100) heterostructures. Peaks at 520 cm -1, cm -1 and cm -1 are derived from the unstrained Si, strained Si and strained Si 0.6 Ge 0.4 layers, respectively. Downloaded on to IP address. Redistribution subject to ECS 117 terms of use (see ecsdl.org/site/terms_use)

8 Fig. 9:Relationship between Raman shift changes for the Si-Si vibration modes from the Si capping and from the Si 0.6 Ge 0.4 layers in the Si/ Si 0.6 Ge 0.4 (25 nm)/si(100) heterostructure with various Si capping thickness and stripe width. The dashed line is drawn with assumption of perfect in-plane lattice matching between the Si capping and Si 0.6 Ge 0.4 layers. the stripe width (25). It should be noted that the peak of the Si-Si vibration mode in the tensile strained Si capping layer is expected to appear at cm -1 (26) on the assumption that the Si capping layer is strained due to relaxation of Si 0.6 Ge 0.4 layer. From the result, it is considered that, by stripe patterning, the strained Si 0.6 Ge 0.4 and Si capping layers tend to become the strained Si 0.6 Ge 0.4 and Si capping layers tend to become relaxed and tensile-strained, respectively. In the case of the 5 nm-thick Si capping layer, values of the Raman shift changes are in excellent agreement with those shown by a dashed line (27) as shown in Fig. 9. On the other hand, in the case of the 10 nm-thick Si capping layer, the Raman shift change in the Si capping layer (i.e. degree of tensile strain) is smaller than the value shown by the dashed line. Since the UV-Raman results clarify the strain amount in the 4-5 nm-thick surface region of the Si capping layer as discussed above, it is considered that strain near the surface in the Si capping layer is smaller than that near the heterointerface. Using the four terminal structure and pn junction isolation, resistivity for Si capping and Si 0.6 Ge 0.4 layer of the patterned Si/Si 0.6 Ge 0.4 /Si(100) heterostructure was evaluated (28). For the strained Si layer on the Si 0.6 Ge 0.4 /Si(100), it is found in Fig. 10 that both n- type and p-type resistivity becomes smaller at the narrower stripe width below 250 nm, although change of the resistivity for the unstrained Si is negligibly small. In the width range of nm, typical resistivity reduction of the strained Si for the stripe is about (a) (b) Fig. 10: Stripe width dependence of the resistivity for 10 nm-thick tensile strained Si ( ), 10 nm-thick unstrained Si ( ), 25 nm-thick Si 0.6 Ge 0.4 ( ) with (a)p-doping concentration of cm -3 and (b)b-doping concentration of cm -3 in the stripe patterned structures. Especially to evaluate the resistivity of the n- and p-si 0.6 Ge 0.4, the capping p- and n-si layer in the contact area was removed by wet etching before epitaxial growth of P-doped Si 0.6 Ge 0.4 for contact, respectively. Downloaded on to IP address. Redistribution subject to ECS 118 terms of use (see ecsdl.org/site/terms_use)

9 10-20 % and 5-10 % for electron and hole, respectively. It is clear that the resistivity reduction corresponds to the increase of strain amount in the Si layer. Additionally, it is found that resistivity of the p-type Si 0.6 Ge 0.4 layer becomes larger at the narrower stripe width below 250 nm, although the change of resistivity for the n-type Si 0.6 Ge 0.4 is negligibly small. The resistivity enlargement corresponds to the decrease of strain amount in the Si 0.6 Ge 0.4 layer. Conclusions Self-limiting formation of 1 3 atomic layers of group IV or related atoms in the thermal adsorption and reaction of hydride gases on Si(100) and Ge(100) are generalized based on the Langmuir-type model. By the epitaxial growth of Si and Si 1-x Ge x over the material already-formed on (100) surfaces, atomic layer doping of N, P and C is achieved. Atomic layer doping results indicate that very high carrier concentration and higher carrier mobility of Si-based group IV semiconductors could be achieved compared with doping under equilibrium conditions, and that suppression of intermixing and strain relaxation at the heterointerface are obtained. Additionally, it is confirmed that the band engineering for group IV semiconductors becomes possible by the strain control of the Si 1-x Ge x /Si heterostructure due to stripe patterning. These results open the way to atomically controlled technology for channel engineering, heavy impurity doping and so on in ULSIs. Acknowledgements This study was partially supported by a Grant-in-Aid for Scientific Research and a Grant-in-Aid for Priority Area Research (No ) from the Ministry of Education, Culture, Sports, Science and Technology of Japan. References 1. B. S. Meyerson, Appl. Phys. Lett. 48, 797 (1986). 2. J. Murota, N. Nakamura, M. Kato, N. Mikoshiba and T. Ohmi, Appl. Phys. Lett. 54, 1007 (1989), p J. Murota and S. Ono, Jpn. J. Appl. Phys. 33, 2290 (1994). 4. B. Tillack, B. Heinemann and D. Knoll, Thin Solid Films 369, 189 (2000). 5. J. Murota, M. Sakuraba and B. Tillack, Jpn. J. Appl. Phys. 45, 6767 (2006). 6. D. Lee, S. Takehiro, M. Sakuraba, J. Murota and T. Tsuchiya, Appl. Surf. Sci. 224, 254 (2004). 7. H. Tanno, M. Sakuraba, B. Tillack and J. Murota, 4th Int. SiGe Technol. and Device Meeting (ISTDM2008), May-11-14, 2008, Abstr.No.Wed-S8-04, p.100 (2008) : Solid-State Electron. (2009) in press. 8. Y. Jeong, M. Sakuraba and J. Murota, Appl. Phys. Lett. 82, 3472 (2003). 9. Y. Jeong, M. Sakuraba and J. Murota, Appl. Surf. Sci. 224, 197 (2004). 10. Y. Jeong, M. Sakuraba and J. Murota, Mat. Sci. Semiconductor Processing 8, 121 (2005). 11. K.L. Brower, Phys. Rev. Lett. 44, 1627(1980); Phys. Rev. B 26, 6040 (1980). 12. S. S. Li, W. R. Thurber, Solid-State Electron. 20, 609 (1977). Downloaded on to IP address. Redistribution subject to ECS 119 terms of use (see ecsdl.org/site/terms_use)

10 13. N. Akiyama, M. Sakuraba, B. Tillack and J. Murota, Appl. Surf. Sci. 254, 6021 (2008). 14. T. Kawashima, M. Sakuraba, B. Tillack and J. Murota, 6th Int. Conf. Si Epitaxy and Heterostructures (ICSI-6), Los Angeles, May-17-22, Y. Shimamune, M. Sakuraba, T. Matsuura and J. Murota, J. Phys. IV (France) 11, Pr3-255 (2001). 16. R.B. Fair and J.C.C. Tsai, J. Electrochem. Soc. 124, 1107(1977). 17. J. Noh, M. Sakuraba, J. Murota, S. Zaima and Y. Yasuda, Appl. Surf. Sci , 679 (2003). 18. Y. Chiba, M. Sakuraba and J. Murota, Semicond. Sci. Technol. 22, S118 (2007). 19. Y. Chiba, M. Sakuraba, B. Tillack and J. Murota, 6th Int. Conf. Si Epitaxy and Heterostructures (ICSI-6), Los Angeles, May-17-22, H. Rücker and B. Heinemann, Solid-State Electron. 44, 783 (2000). 21. K. Takahashi, M. Sakuraba and J. Murota, Appl. Surf. Sci , 193 (2003). 22. T. Hirano, M. Sakuraba, B. Tillack and J. Murota, 6th Int. Conf. Si Epitaxy and Heterostructures (ICSI-6), Los Angeles, May-17-22, T. Ghani, M. Armstrong, C. Auth, M. Bost, P. Charvat, G. Glass, T. Hoffmann, K. Johnson, C. Kenyon, J. Klaus, B. McIntyre, A. Murthy, J. Sandford, M. Silberstein, S. Sivakumar, P. Smith, K. Zawadzki, S. Thompson and M. Bohr, IEDM Tech Dig., p.978 (2003). 24. J. Uhm, M. Sakuraba and J. Murota, Thin Solid Films 508, 239 (2006). 25. J. Uhm, M. Sakuraba and J. Murota, Thin Solid Films 517, 300 (2008). 26. K. Arimoto, D. Furukawa, J. Yamanaka, K. Nakagawa, K. Sawano, S. Koh, Y. Shiraki, N. Usami, Mat. Sci. Semicond. Processing 8, 181 (2005). 27. J. Lockwood and J.-M. Baribeau, Phys. Rev. B 45, 8565 (1992). 28. J. Uhm, M. Sakuraba and J. Murota, Semicond. Sci. Technol. 22, S33 (2007). Downloaded on to IP address. Redistribution subject to ECS 120 terms of use (see ecsdl.org/site/terms_use)