Fabrication of silicon-on-aln novel structure and its residual strain characterization

Transcription

1 Journal of Crystal Growth 244 (2002) Fabrication of silicon-on-aln novel structure and its residual strain characterization Zhenghua An a,c, *, Chuanling Men b, Xinyun Xie a, Miao Zhang a, Paul K. Chu c, Chenglu Lin a a State Key Laboratory of Functional Materials for Informatics, Shanghai Institute of Microsystem and Information Technology, Chinese Academy of Sciences, Room 8-310, Changning Road 865, Shanghai, China b Institute of Microelectronic Materials, School of Material Science and Engineering, Tong-ji University, Shanghai , China c Department of Physics and Materials Science, City University of Hong Kong, Tat Chee Avenue, Kowloon, Hong Kong Received 27 March 2002; accepted 26 June 2002 Communicated by C.R. Abernathy Abstract Self-heating effects in silicon-on-insulator (SOI) devices limit the applicability of SOI materials in electronics in cases where high power consumption is expected. AlN becomes a promising alternative to SiO 2 layer in traditional SOI materials. For the first time, a novel silicon-on-aluminum-nitride (SOAN) structure has been fabricated by the smartcut process to alleviate the self-heating effects. The AlN films were synthesized on 4 00 Si(1 0 0) substrate by ion-beamenhanced deposition technique, followed by the smart-cut process. Cross-sectional transmission electron microscopy micrograph confirms the formation of the SOAN structure. High-resolution transmission electron microscopy image and spreading resistance profile results serve evidence that the top silicon has good crystalline quality and electrical quality similar to the Si substrate. High-resolution X-ray diffraction was employed to study the residual strain in the formed SOAN structure and indicates that the residual lattice strain in the top silicon layer varied from tensile to little compressive after as-received SOI samples annealed at 11001C for an hour. r 2002 Elsevier Science B.V. All rights reserved. PACS: 81.15; 77.55; Keywords: A1. High resolution X-ray diffraction; A3. Ion-beam-enhanced deposition; B1. Nitrides 1. Introduction A buried oxide layer (BOX) has been largely used in silicon-on-insulator (SOI) materials offering several advantages as compared to bulk silicon *Corresponding author. Tel.: ; fax: address: anzhenghua@mail.sim.ac.cn (Z. An). due to easier electrical isolation and a significant reduction of parasitic capacitances. In spite of the best electrical performances and operating frequency of SOI power devices, the low thermal conductivity of the BOX leads to an undesirable thermal isolation, becoming critical in high power devices due to the large delivered current level. Hence, severe self-heating effects can eventually degrade the device electrical characteristics under /02/$ - see front matter r 2002 Elsevier Science B.V. All rights reserved. PII: S (02)

2 28 Z. An et al. / Journal of Crystal Growth 244 (2002) prolonged high temperature operation. Several attempts have been performed to reduce the above effects [1,2]. Recently, AlN has received increasing interest due to its unique properties: high electrical resistivity, high breakdown dielectric strength, high chemical and thermal stability and a coefficient of thermal expansion close to that of silicon. In particular, the thermal conductivity of AlN is almost 200 times higher than that of SiO 2 (3.2 W cm 1 K 1 against W cm 1 K 1 ) [3]. All these outstanding properties make AlN thin film a potential candidate for buried insulator material in SOI structure, namely forming the silicon-on-aluminum-nitride (SOAN) structure [4]. Various techniques have been reported for synthesis of AlN films, including molecular beam epitaxy (MBE), reactive sputtering, chemical vapor deposition (CVD) and pulsed laser deposition (PLD). Ion-beam-enhanced deposition (IBED) [5 8] technique is another competing technique for the synthesis of AlN films. Compared with other deposition methods, IBED can offer an accurate and direct control over the fundamental deposition parameters. Furthermore, smart-cut process [9] is a versatile technique for the buildup of various novel structures, such as silicon-onglass, SiGe-on-insulator, etc. However, up to now, no one has succeeded in fabricating SOAN structure with AlN buried layer grown by IBED technique by means of the smart-cut process. In this paper, AlN films on Si(1 0 0) substrate prepared by IBED were bonded directly at room temperature to a hydrogen-implanted wafer through the smart-cut process. Then, the obtained SOAN structure was characterized by crosssectional transmission electron microscopy (XTEM) and spreading resistance profile (SRP). In particular, high-resolution X-ray diffraction (HRXRD), a nondestructive tool for characterization of the SOI film [10,11], was applied to measure residual strain of the SOAN film in bonded wafers. the wafer was implanted with 20 kev N 2 + and N + (60%N 2 +, 40%N + ) with the ion flux density of 25 macm 2, and meanwhile Al was evaporated at the rate of 0.5 ( As 1. A detailed procedure for the AlN film preparation is given in Ref. [12]. The main steps of fabricating the SOAN structure by the smart-cut process are illustrated in Fig. 1, including hydrogen implantation (dose H + cm 2 and energy 150 kev), synthesis of AlN thin film and hydrophilic bonding at room temperature of the two wafers; the implanted wafer split into two parts at 4501C giving SOAN structure, then annealing was performed at 11001C for an hour in a N 2 gas atmosphere to increase the bonding strength. The SOAN structure was investigated by XTEM, HRXRD and SRP. XTEM images were carried out on a JEM-4000EX. Samples for XTEM were prepared in the conventional way by mechanical thinning and Ar + ion. HRXRD was performed on a Philips X pert equipped with a two-crystal four-reflection Ge[2 2 0] diffractometer. Cu Ka 1 radiation with a wavelength of ( A and fourfold Ge(2 2 0) reflection were used in the experiment. Rocking curves were 2. Experimental procedure The AlN films used in this work were grown on 4 00 Si(1 0 0) substrate by IBED. During the growth, Fig. 1. Process of fabricating SOAN structure with AlN as buried insulating layer by the smart-cut technique.

3 Z. An et al. / Journal of Crystal Growth 244 (2002) measured on the working plane (0 0 4) with an o=2y scan. 3. Results and discussion To fabricate SOAN structure by the smart-cut process, the IBED-derived AlN thin films should have a smooth surface to meet the need of bonding. Typically bonding requires that the mean microroughness of the wafer surface should be in the range of 1 2 ( A. AFM observation was performed to examine surface topography of the AlN film (shown in Fig. 2). It can be found that the surface of the film is smooth and uniform, and the surface roughness RMS values are 1.3 ( A over a 1 1 mm 2 area, enough for direct wafer bonding without the intermediate polishing step. The AlN film surface was rinsed in de-ionized water and spun dry before bonding and the hydrogenimplanted wafer passed through an RCA cleaning procedure, thus made it hydrophilic and also spun dry. After that, direct bonding was performed at room temperature immediately. The yield of the bonding procedure is extremely dependent on the surface roughness as well as on the stress in the film. Our successful bonding further confirms the smooth surface of the AlN films. Furthermore, we think that both van der Waals interactions and hydrogen bridge bonds contribute to our successful bonding, although they may be somewhat different from Si SiO 2 bonding. The subsequent heat treatment at 4501C split the bonded pair from the hydrogen-implantation-induced bubble layer. As a result, SOAN structure was obtained and the bonded interface was strengthened with the high temperature annealing (11001C). Fig. 3 shows the SRP of the SOAN samples. Three layers similar to traditional SOI, including the top silicon layer, buried AlN and substrate, can be clearly distinguished. For the as-received samples, the top Si resistance is about O and unexpectedly not so uniform. This may be due to the nonuniform distribution of implantation damage or displacement atoms. For the annealed sample, the top Si resistance becomes more uniform (about O) due to the recovery of implantation damages. What is more, it exhibits a more steeper slope from top silicon to buried AlN, implying a good interface quality between top silicon and AlN. The thickness of top silicon is about 1.25 mm, consistent with the simulation result of hydrogen implantation by TRIM-96. Also, the resistance lowering of top Si is originated from activation of impurity during high temperature annealing. The spreading resistance of AlN buried layer is higher than 10 8 O (exceeding the measurement range of the SRP apparatus) Fig. 2. Surface topography obtained by AFM of the AlN film formed at 0.5 ( As 1 evaporation rate of Al. Fig. 3. Spreading resistance profile of the SOAN samples.

4 30 Z. An et al. / Journal of Crystal Growth 244 (2002) both for the as-received and annealed samples, suggesting excellent dielectric properties of the buried AlN layer and thus the SOAN material becoming a promising alternative to traditional SOI material. Cross-sectional TEM was adopted to investigate the structure of SOAN sample annealed at 11001C for 1 h (shown in Fig. 4). It gives a direct evidence of the formation of SOAN structure. The thickness of the top Si layer is about 1.2 mm, well consistent with the simulation result by TRIM-96 and SRP result. From the image, the complete closure of the interface appears clearly. The AlN buried layer exhibits the thickness uniformity across the whole wafer. Both the upper and lower interfaces are very flat and steep. The highresolution TEM image of the upper interface (bonded interface) is also shown in Fig. 4. It can be seen clearly that there is a straight and smooth boundary between the top silicon and AlN buried layer. The top silicon has a good single crystal quality with no resolvable defects. At the same time, it could be derived from the figure that the buried AlN has a thickness of about 60 nm. The structure of the AlN layer is amorphous and neither voids nor bubbles are observable. The amorphous structure of AlN is very suitable to act as insulator in SOAN structure ensuring low leakage currents due to the absence of the columnar growth, grains, and threading dislocations, whereas grain boundaries in crystalline AlN may serve as high leakage paths [13,14]. In addition, the silicon /110S spot diffraction Fig. 4. Cross-sectional TEM micrograph of the SOAN sample annealed at 11001C for 1 h (left) and high-resolution XTEM image of the top Si/buried AlN interface (right). pattern (not shown) of the select-area electron diffraction (SAD) also shows good crystallization of the top Si with perfect crystalline lattice similar to the substrate silicon, which can meet the requirement of manufacturing devices ensuring stability and reliability. To measure well-defined diffraction peaks from both the top layer and the substrate, and then evaluate the variation in the residual lattice strain in the top layer, HRXRD was employed. An o=2y scan was performed on the working plane (0 0 4), where o is the angle between the incident beam and the wafer surface and 2y is the angle between the incident beam and the detector (reflected beam). This technique is very sensitive to strain in the direction perpendicular to the working diffraction plane of the crystal. Following Bragg s law, the HRXRD spectra build up as a result of summing over many periodic atomic planes. Thus, contributions to the HRXRD spectrum are expected both from the crystalline top Si layer and the substrate, while AlN amorphous layer does not contribute. Rocking curves of the SOAN samples are shown in Fig. 5. No Pendellosung fringes (thickness fringes) are clearly resolved because of a relatively shorter integration time (about 10 min for each scan) of measurement with our present equipment. But it does not matter since we only want to determine the strain in top Si, but not the thickness. Each curve is computer simulated and divided into two peaks, relating to contribution from top Si and from substrate, respectively. The angular separation corresponds to the tilt between the top layer and substrate silicon planes. When two wafers are bonded, some tilt and rotational misalignment may exist, leading to an asymmetric set of planes in the substrate not parallel to a similar set of planes belonging to the top silicon layer [11]. The angular separation, a unique feature of bonded wafers, is also observed in our SOAN structure. The residual strain in the top silicon layer may be determined by taking two measurements of the same reflection at two j values which differ by 1801 (j denotes a rotation in the plane of the wafer) [15]. Assuming that the difference in the diffraction angles at j between the substrate and the top silicon layer is DR 0 ij ; and at j þ 1801; this

5 Z. An et al. / Journal of Crystal Growth 244 (2002) Fig. 5. The HRXRD patterns of the SOAN samples (a) j ¼ 01; as-received sample, (b) j ¼ 01; annealed at 11001C for 1 h, (c) j ¼ 1801; as-received, (d) j ¼ 1801; annealed. angle is DR 180 ij : Then: Dy ¼ðDR 0 ij þ DR180 ij Þ=2; ð1þ where Dy is the difference in the Bragg angle due to the different lattice spacings in the substrate and the top layer. Since Dy can be also expressed as Dy ¼ tan yðd top d sub Þ=d sub ; ð2þ where d top and d sub are the diffraction plane spacings in the top layer and the substrate, and y is the substrate Bragg angle, the residual lattice strain is given by e ¼ Dd=d sub ¼ðd top d sub Þ=d sub ¼ ðdr 0 ij þ DR180 ij Þ=2tan y: ð3þ The positive and negative signs of e refer to compressive strain and tensile strain, respectively, because the XRD measurements refer to the change in the inter-planar spacing normal to the substrate surface.for the as-received SOAN sample, the residual lattice strain in the top silicon layer is about , indicating that a tensile lattice strain, i.e., lateral tensile stress applied parallel to the surface was measured. The initial strain in the top Si layer is mainly due to the difference in the thermal expansion coefficients of Si ( K 1 ) and AlN ( K 1 ) and is produced during the wafer bonding process. Whereas, after being annealed at 11001C, the strain is cut down deeply to about , which is one order of magnitude lower than that of the as-received sample. At the same time, it is very clear that the residual lattice strain in the top silicon layer varied from tensile (negative) to compressive (positive) as the annealing temperature was increased. The significant change of the strain in the top silicon layer implies that an almost stress-free condition is established owing to

6 32 Z. An et al. / Journal of Crystal Growth 244 (2002) the viscosity of AlN. It is possible that the residual lattice strain is removed from the top Si layer by the annealing process as long as the annealing temperature is sufficiently higher than the temperature required for the onset of viscous flow in the AlN. In addition, we note that the diffraction spectra also give some indication as to the top silicon layer quality. After annealing (see Figs. 5c and d), the diffraction peak width becomes much narrower than that of as-received SOAN samples, indicating the improved crystalline quality of the top Si layer. This is due to the removal of implantation damages in the top Si layer at high temperature, which ensures the applicability of SOAN materials. that our method for forming SOAN structure is feasible. Thus SOAN material could be used in micro-electronic applications where thermal effects are to be taken into account, particularly, when thermal dissipation is needed. Acknowledgements This work is supported by the special Funds for major state Basic Research projects No. G , the National Nature Science Foundation of China under grant Nos and , and Hong Kong RGC CERG Cityll 1052/02 E or Conclusions In summary, a novel SOAN structure addressed to decrease the self-heating effect inherent to SOI devices is achieved by using AlN thin film rather than SiO 2 as buried insulating layer in SOI structure. The AlN films were synthesized on 4 00 Si(1 0 0) substrate using the IBED technique. The subsequent smart-cut process proved that the AlN film could be bonded to hydrogen-implanted silicon wafer at room temperature. The small microroughness contributes to our successful bonding. XTEM micrograph confirms the formation of SOAN structure. After high temperature annealing, the top Si has a crystal and electrical quality almost as good as the substrate. Besides, it can also be pointed out that, for the as-received SOAN sample, there exists a higher tensile strain in the top silicon probably owing to the difference in the thermal expansion coefficients between Si and AlN, the implantation damages as well as the tilt and the rotational misalignment induced during the wafer bonding process. While, in the case of the annealed sample, the residual strain is reduced greatly to compressive strain as a result of viscous flow of the AlN. These results demonstrate References [1] J. Roig, D. Flores, M. Vellvehi, J. Rebollo, J. Millan, Microelectron. Reliab. 42 (2002) 61. [2] F. Udrea, W. Milne, A. Popescu, Electron Lett. 33 (1997) 907. [3] Z.Y. Fan, G. Rong, J. Browning, N. Newman, Mater. Sci. Eng. B 67 (1999) 90. [4] S. Bengtsson, M. Bergh, M. Choumas, C. Olesen, K.O. Jeppson, Jpn. J. Appl. Phys. 35 (1996) [5] Y. Watanabe, Surf. Eng. 14 (1998) 427. [6] X. Wang, A. Kolitsch, F. Prokert, W. Moller, Surf. Coat. Tech. 103/104 (1998) 334. [7] J.H. Edgar, C.A. Carosella, C.R. Eddy, D.T. Smith, J. Mater. Sci. Mater. Electron. 7 (1996) 247. [8] X.J. He, S.Z. Yang, K. Tao, Y.D. Fan, Mater. Chem. Phys. 51 (1997) 199. [9] M. Bruel, Electron. Lett. 31 (1995) [10] S. Kimura, A. Atsushi, Jpn. J. Appl. Phys. Part 1 37 (1998) [11] G.M. Cohen, P.M. Mooney, E.C. Jones, K.K. Chan, P.M. Solomon, Appl. Phys. Lett. 75 (1999) 787. [12] C.L. Men, Z. Xu, Z.H. Zheng, X.Z. Duo, M. Zhang, C.L. Lin, Chin. Phys. Lett. 18 (2001) [13] K.J. Hubbard, D.G. Schlom, J. Mater. Res. 11 (1996) [14] T. Adam, J. Kolodzey, C.P. Swann, M.W. Tsao, J.F. Rabolt, Appl. Surf. Sci (2001) 428. [15] T. Lida, T. Itoh, D. Noguchi, Y. Takano, J. Appl. Phys. 87 (2000) 675.