Chapter 1 Introduction
1.3 Channel Material Engineering
The channel material engineering is one of most applications for the SiGe or Ge epitaxial growth. The SiGe channel pMOSFET was investigated in the previous work, and the enhancement of the hole mobility and the driven current are at least 3 times to the bulk Si device, as shown in Fig. 1-3 [9]. Due to the highest hole mobility of Ge among group III-V and IV materials [10]. The Ge pMOSFETs with high hole mobility and high drive current have already been reported [11, 12]. The Ge (111) nMOSFETs with record high electron mobility (2200 cm2/V s) exceeding Si universal mobility proved in 2012 [13], as shown in Fig. 1-4. However, the high cost of Ge wafers limits the application of Ge MOSFET in CMOS technology. The high quality, defect free of SiGe or Ge epitaxial growth directly on Si substrate has drawn lots of attention in the last decade.
However, the heteroepitaxy of Ge and SiGe on Si have the fundamental limit from the high lattice mismatch (up to 4.2%) between Si and Ge. Therefore, the misfit dislocation network at interface to relax strain is needed for SiGe or Ge growth on Si.
These dislocation can degraded the crystalline quality and consequently compensate the electron and hole mobility advantage expected from the SiGe or Ge channel. One of conventional solution is to grown a gradually but eventually fully relaxed Si1-xGex (x ramps from 0 to 1) buffer layer on the Si substrate, serving as the so-called virtual substrate. In the Fig. 1-5, the defect free Si0.5Ge0.5 can be grown on the virtual substrate, which is graded buffer layer with Si1-xGex (0<x<0.5). In order to reduce the density of threading dislocations, these technologies typically require the thick growth (typically >
1 m) to gradually ramps up Ge fraction to relax strain, is not appreciated in practical industry applications. Therefore, it is highly interested and demanded if one can have
some innovative ideas to produce thinner and dislocation-free virtual substrate layers, or even direct growth of high quality Ge structures on Si. Several approaches proposed in the represent project proposal are thus attempts to find a practical solution to make defect-free Ge layer growth for heterogeneous integration of high mobility channel.
Fig.1-3 (a) Drain current characteristics of the SiGe QW device and bulk Si device. (b) The hole mobility of SiGe QW device and bulk Si device from the split C-V measurement [9].
Fig.1-4 Substrate orientation effects on electron mobility in our planar devices compared to Si universal mobility. The peak mobility on Ge(111) is about 2200 cm2/V.s at RT, which is 2 times enhancement compared to that of Ge(100) [13].
Fig.1-5 The TEM image of the 500nm Si0.5Ge0.5 on Si1-xGex relaxed buffer layer. Note that the strain is gradually relaxed by the relaxed buffer layer to make Si0.5Ge0.5 be defect free.
1.4 Source/Drain Material Engineering
Due to the good gate control of FinFETs, the strain technology is more important than the EOT scaling. Si1-xGex source/drain stressor [14]-[16] is one of the most successful strain techniques for pMOS performance. Besides, the Si1-xGex S/D reduces the resistance of S/D substantially [17]. The SiGe S/D has been used in the Intel 22 nm node FinFETs [18]. The shape stressor along the <110> FinFETs channel direction changes from rectangular to diamond (Fig. 1-6), but is conformal along the <100>
channel direction. The reason for the diamond shape could be the fin orientation and the orientation dependent growth rate difference (Fig. 1-7) [19].
Fig.1-6 The TEM of the PMOS channel in the S/D region showing the SiGe epitaxy in the S/D region [18].
Fig.1-7 TEM images of Si60Ge40 layer grown on FinFET structures defined on (100) Si:
(a) fin oriented along <110> direction, (b) fin oriented along <100> direction [19].
1.5 Dissertation Organization
The motivation of the SiGe and Ge epitaxial growth for advanced device applications are given in the previous sections of this chapter. The importance of the SiGe or Ge epitaxial growth for high mobility channel and S/D stressor are emphasized.
In chapter 2, using Unaxis UHV/CVD, where the growth temperature and unexpected contamination is low, the Si and Ge atoms surface diffusion and selective growth for the SiGe nanoring formed by Ge out-diffusion mechanism in different ambient are discussed in this chapter. When the central area of quantum dot is passivated by H, the epitaxial Si capping layer cannot cover the whole quantum dots and formed uncapped quantum dot. SiGe nanoring can be formed by 500oC in situ annealing in 1hr. When the epitaxial Si capping layer covers the whole quantum dots grown in He ambient, more than 4hrs in situ annealing is needed to form SiGe nanorings. Si capping layer grown in less H ambient can retard the SiGe nanoring formation is demonstrated.
In chapter 3, a transition from three- to two-dimensional growth for Si grown on Ge(100) is first observed in this chapter. With the increasing Si deposition, the Si growth changes from three- to two-dimensional growth mode and the dots disappear gradually. Finally, the surface is smooth with the roughness of 0.26nm, similar to the original Ge substrate, when 15 nm Si is deposited. More Ge segregation on the wetting layer leads to more open sites to increase the subsequent Si growth rate on the wetting layer than on the Si dots. The in-plane x-ray diffraction by synchrotron radiation is used to observe the evolution of tensile strain in the Si layer grown on Ge (100) substrate.
Growth rate enhancement with more Ge content leads to the morphology change for Si growth on Ge.
In chapter 4, basic theory and mechanism factors of epitaxial SiGe growth mechanism is discussed previously. Open sites creation and precursor adsorption are the main causes affect the SiGe growth rate. Using ASM Epsilon RTCVD, the SiGe growth rate is studied in different conditions, including growth ambient, Ge content, and gas precursors. Base on the previous observation of SiGe growth, the SiGe growth rate can be enhanced by having more open sites on surface. Due to easier H desorption from Ge than from Si, more Ge content surface can enhance SiGe growth rate by having more open sites than its with less Ge content. Besides, when using silane (SiH4) in N2 ambient instead of H2 can enhance both the SiGe and Ge growth rate by 2x~3x. But, harder gas phase SiCl2 desorption in N2 ambient when using dichlorosilane (SiH2Cl2) as Si source can retard the Si and Ge adsorption and reduce the SiGe growth rate. X-ray diffraction and photoluminescence are used to determine the Ge content. And higher growth rate results to lower Ge content of SiGe is found.
From the chapter 5 to chapter 6, the two doping technologies for Ge using CVD are studied. Conventional doping technology, like ion implantation, can create defects even after implant anneal. For n-type dopant, the point defects such as Ge vacancies are the main root cause of fast diffusion of dopants and low activation fraction. For p-type dopant, stable defects by implantation-induced crystal disorder remain after long time annealing. All of these defects degrade the Ge diode performance to have non-ideal forward current, low on/off ratio, and high reverse saturation current. In chapter 5, the
solid phase phosphorous layer and boron layer deposited by CVD are used as n-type and p-type doing, respectively. Doping Ge by the diffusion of dopant layers on Ge has much lower defect density as compared to ion implantation. From the n+p and p+n Ge junction formed by phosphorous and boron layers, respectively, the low defect density leads to abrupt dopant profiles, and good diode characteristics.
In chapter 6, in situ doping technology which is also free from implantation damage is studied. To reduce the dopant desorption during in situ doping, low temperature growth with post activation annealing is used to have high activated carrier density. For phosphorous doped Ge, limited dopant diffusion due to the nanosecond duration of the laser annealing is used. Higher and steeper dopant distribution can have that it with rapid thermal annealing and in situ H2 annealing. 2x1020 cm-3 n-type Ge can have using in situ phosphorous doped Ge activated by laser. For boron doped Ge, laser anneal
could introduce interstitial defects. These defects can result to dopant diffusion or even inactivation by boron-interstitial-cluster. 3x1020 cm-3 p-type Ge can have using in situ boron doped Ge activated by in situ annealing.
In the chapter 7, the important contribution and conclusion of SiGe and Ge epitaxy growth and doping are summarized. Besides, the future works on CVD epitaxy which can used in the improved device application is discussed in the last part of this thesis.
Reference:
[1] http://en.wikipedia.org/wiki/Transistor_count.
[2] Y. Taur and T. H. Ning, Fundamentals of Modern VLSI Devices, Cambridge, (2002).
[3] The website of Intel Corp: http://newsroom.intel.com/docs/DOC-2032.
[4] M. Lundstrom, International Electron Devices Meeting (IEDM), Technical Digest, 789 (2003).
[5] C. Jungemann and B. Meinerzhagen, International Electron Devices Meeting (IEDM), Technical Digest, 191 (2003).
[6] A. G. O’neill and A. D. Antoniadis, IEEE Trans. Electron Devices, 43, 911 (1996).
[7] B. S. Meyerson, IBM J. Res. Develop., 44, 132 (2000)
[8] F. W. Smith and G. Ghiddini, Journal of the Electrochemical Society, 133, 1300 (1984)
[9] C. -Y. Peng, F. Yuan, C. -Y. Yu, P. -S. Kuo, M. H. Lee, S. Maikap, C. -H. Hsu, and C. W. Liu, Appl. Phys Lett, 90, 012114 (2007)
[10] R.Pillarisetty, “Academic and Industry research progress in Germanium Nanodevices”, Nature, vol. 479,17 (2011).
[11] S. -H. Lee, J. Huang, P. Majhi, P. D. Kirsch, B. -G. Min, C. -S. Park, J. Oh, W. -Y.
Loh, C. -Y. Kang, B. Sassman, P. Y. Hung, S. McCoy, J. Chen, B. Wu, G. Moon, D. Heh, C. Young, G. Bersuker, H. -H. Tseng, S. K. Banerjee, and R. Jammy,
Symposium on VLSI Technology Digest of Technical Papers, 74 (2009).
[12] N. Taoka, W. Mizubayashi, Y. Morita, S. Migita, H. Ota, and S. Takagi, Symposium on VLSI Technology Digest of Technical Papers, 80 (2009).
[13] Shu-Han Hsu, Hung-Chih Chang, Chun-Lin Chu, Yen-Ting Chen,Wen-Hsien Tu, Fu Ju Hou, Chih, Hung Lo, Po-Jung Sung, Bo-Yuan Chen,Guo-Wei Huang, Guang-Li Luo, C.W.Liu, Chenming Hu,and Fu-Liang Yang, International Electron Devices Meeting (IEDM), Technical Digest, 23.6 (2012).
[14] M.D.Giles, M.Armstrong, C.Auth, S.M.Cea, T.Ghani, T.Hoffmann, R.Kotlyar,
P.Matagne, K.Mistry, R.Nagisetty, B.Obradovic, R.Shaheed, L.Shifren, M.Stettler,
S.Tyagi, X.Wang, C.Weher, K.Zawadzki, Symposium on VLSl Technology Digest of Technical Papers, pp. 118-119, (2004).
[15] Z. Luo, Y. Chong, J. Kim, N. Rovedo, B. Greene, S. Panda, T. Sato, J. Holt, D.
Chidambarrao, J. Li, R. Davis, A. Madan, A. Turansky, O. Gluschenkov, R. Lindsay,
Ajmera, J. Lee, S. Mishra, R. Amos, D. Schepis, H. Ng, and K. Rim, IEDM Tech.
Dig., pp. 495–498, (2005).
[16] F. Nouri, P. Verheyen, L. Washington, V. Moroz, I. De Wolf, M. Kawaguchi, S.
Biesemans, R. Schreutelkamp, Y. Kim, M. Shen, X. Xu, R. Rooyackers, M. Jurczak, G. Eneman, K. De Meyer, L. Smith, D. Pramanik, H. Forstner, S. Thirupapuliyur, and
G. Higashi, IEDM Tech. Dig., pp. 1055–1058, (2004).
[17] V. Moroz, G. Eneman, P. Verheyen, F. Nouri, L. Washington, L. Smith, M.
Jurczak, and D. Pramanik, Proc. SISPAD, pp. 143–146, (2005).
[18] C. Auth, C. Allen, A. Blattner, D. Bergstrom, M. Brazier, M. Bost, M. Buehler, V.Chikarmane, T. Ghani, T. Glassman, R. Grover, W. Han,D. Hanken, M.
Hattendorf, P. Hentges, R. Heussner, J. Hicks*, D. Ingerly, P. Jain, S. Jaloviar, R.
James, D. Jones, J. Jopling, S. Joshi, C. Kenyon, H. Liu, R. McFadden, B.
McIntyre, J. Neirynck, C. Parker, L. Pipes, I. Post, S. Pradhan, M. Prince, S.
Ramey*, T. Reynolds, J. Roesler, J. Sandford, J. Seiple, P. Smith, C. Thomas, D.
Towner, T. Troeger, C. Weber, P. Yashar, K. Zawadzki, K. Mistry Symposium on VLSl Technology Digest of Technical Papers, pp. 131-132, (2012).
[19] Hikavyy , W. Vanherle, B. Vincent, J. Dekoster, H. Bender, A. Moussa, L. Witters, T. Hoffman, R. Loo, Thin Solid Film, 520, 3179–3184, (2012).
Chapter 2
Growth and Control of SiGe Nanoring Formation
2.1 Introduction
Self-assembled quantum dots (QDs) have attracted much interest for the potential applications of nanoelectronics [1–3]. Nanoring structures have also been widely observed in III-V system [4,5] and applied to optoelectronics [6]. SiGe nanorings on Si(100) grown both by molecular beam epitaxy [7] and ultrahigh vacuum chemical vapor deposition (UHV/CVD) [8] have been reported previously. The UHV/CVD growth of nanoring at 600 oC by Si surface diffusion can only be observed in very limit process window due to specific growth mechanism and this kind of SiGe nanorings can hardly meet the device requirements.
Both the increasingly Si cap growth and the high thermal budget can destroy this kind of nanoring structures [9]. The nanoring formed at 500oC by Ge out-diffusion under Si cap region was preferred to control the nanoring formation. For the blank growth of SiGe, carrier gas (H2 or He) was well-known to change the growth kinetics [10]. The hydrogen passivation, which can influence the Ge concentration and the strain in the SiGe nanostructures, plays a crucial role in the SiGe quantum wells (QWs) and SiGe QDs growth. For QWs growth, The H2 carrier gas increases Si segregation during the
QW growth and leads to the rougher SiGe/Si bottom interface [11]. The He carrier gas can improve the uniformity distribution and volume of QDs as compared to H2 [12]. For the nanoring formed by Ge out-diffusion at 500oC, the Si capping grown on QDs and Ge out-diffusion during subsequent annealing can be also controlled by the carrier gas.
The dependence of Ge out-diffusion on Si caps as well as the ambient environments (H2, or He) is studied in this work. Nanorings were characterized by atomic force microscopy (AFM), high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM), and energy dispersive x-ray spectroscopy (EDS).
2.2 Experiments
SiGe nanorings on Si(100) were grown by UHV/CVD at 500 oC. Pure silane (SiH4) and germane (GeH4) were used as reactant gases. Due to the hot-wall system, the growth temperature and temperature are fixed during the growth. The hot-wall system can prevent the process environment from the out-gasing of the chamber wall. The background impurity can be reduced, and the film quality would thus be improved. The base pressure of UHV/CVD system is ~10-9 torr and 10-3 torr during process. HF-dip last is used to have a hydrogen passivated wafer surface. After clean, the wafer is loaded into the load lock chamber directly. The process wafers will be put into the loader chamber first, and pump to ~10-6 torr. When the pressure inside the loader chamber matches the set point, the chamber door will open, and the wafer is transformed into the process chamber. After the transformation, the wafers would be baked in the process chamber with 300 sccm H2 flow at the process temperature first to keep the Si surface hydrogen passivated. The ~3 nm Si buffer layer was grown first using SiH4 after the HF
dip of Si wafers. The gas flows of GeH4 and He carrier gas were fixed at 5 and 35 sccm, respectively, to grow uniformly distributed QDs. After Stranski-Krastanov mode growth of Si0.2Ge0.8 QDs, the epi-Si layer using SiH4 at 50 sccm with H2 (50 sccm) or He (50 sccm) was deposited for comparison. The samples were then in situ annealed in vacuum, in He ambient and in H2 for comparison. Carrier gas effects on Si capping on QDs and annealing for out-diffusion are studied. Note that Ge QDs transform into Si0.2Ge0.8
alloys due to Si/Ge inter-diffusion [13], as measured by the EDS.
2.3 Carrier Gas Effects on Si cap for Nanoring Formation
To observe the carrier gas effects on Si capping on QDs, the high resolution cross-section TEM images at Fig. 2-1(a) and HAADFSTEM images of epi-Si layers using SiH4/H2 grown on QDs for contrast between Si and Si0.2Ge0.8 at Fig. 2-1(b). Since the H can passivate the open sites on the QD surface to reduce the silane adsorption at UHV/CVD growth pressure of 10-3 torr, the epi-Si layer deposited using SiH4 and H2
cannot form completely on the QDs, [Fig. 2-1(a) and (b)]. The tops of QDs are more relaxed [14] than the bottoms, and have a larger in-plane lattice constant than Si. If some Si atoms are accidentally deposited on the central area of the QDs, the tensile strain given by the tops of QDs can drive the diffusion of Si to the wetting layers [8].
For comparison, the TEM images at HAADF-STEM images of epi-Si layers using SiH4/He can grow not only on the wetting layer but also on the QDs (Fig. 2-2(a) and (b)). Since the annular dark field image formed only by very high angle, incoherently scattered electrons — as opposed to Bragg scattered electrons for conventional TEM —
it is highly sensitive to variations in the atomic number of atoms in the sample by HAADF-STEM.
It is found that the QDs can be totally covered using SiH4/He growth. Besides, due to the hydrogen passivation, the Si layer (~5 nm, Fig. 2-1) on the wetting layer for SiH4/H2
growth is also thinner than that for SiH4/He growth (~7 nm, Fig. 2-2). Note the SiGe wetting layer is thinner than 1 nm for original QDs.
For 1 h annealing at 500oC in vacuum (10-9 torr) after SiH4/H2 growth for partially covered QDs, nanorings were observed [Fig. 2-3(a)]. These Ge atoms can out diffuse from the central area of uncapped QDs to form nanorings [Fig. 2-3(b)]. On the contrary, for the epi-Si layer deposited using SiH4 and He, the H on the QDs surface was taken away by the He gas flow and left more surface open sites on QDs. The SiH4 can be adsorbed and deposited on the whole QDs to form capped QDs [Fig. 2-4(b)]. Due to the Si cap, the Ge surface diffusion is relatively slow [15] No nanoring formation was observed after the same annealing condition (1 h at 500 oC) in vacuum[Fig. 2-4(a)].
(b)
Fig.2-1 (a) High resolution cross-sectional TEM images and (b) HAADF-TEM images of the uncapped QD using SiH4 and H2. Note that the QDs is only partially capped and the Si thickness on wetting layer is only ~5nm.
a)
Fig.2-2 (a) High resolution cross-sectional TEM images and (b) HAADF-TEM images of the capped QD using SiH4 and He. Note that the QDs is totally capped and the Si thickness on wetting layer is ~7nm which is thicker than its using SiH4 and H2.
/H2.
Fig.2-3 (a) The AFM images (5m x 5m) of nanorings transformed from uncapped QDs after 1h in-situ vacuum annealing and (b) The schematics of corresponding growth model of nanoring formation transformed from uncapped QDs. Note H passivation can prevent Si growth on top of the QD and enhance the Ge out-diffusion for nanoring formation
Si surface diffusion to strain free region
Ge out-diffusion (b)
: Ge : Si
(b)
Fig.2-4 (a) The AFM images (5m x 5m) of capped QDs after 1 h in-situ annealing in vacuum at 500oC and (b) The schematics of corresponding growth model for no nanoring formed from capped QDs. Note that Si can be grown on top of QD with no H passivation and retard Ge out-diffusion to form nanorings
: Ge : Si
(a)
Fig. 2-5 is the three dimension AFM topography images of uncapped QDs and nanoring formed from uncapped QDs. The average width and height of uncapped QDs is 75nm and 25nm, respectively (Fig. 2-5(a)). And the average width and depth of nanorings is 130nm and 9nm, respectively (Fig. 2-5(b)). It is found that since Ge at the center of the uncapped QD diffuses outward to form nanoring, the width of the nanoring is larger than the uncapped QD. The Ge out-diffusion could be also observed from the high resolution TEM cross-section with dispersive x-ray spectroscopy (EDS) measurement of the nanoring (Fig.2-6). From the EDS, the Ge content at the ring edge is 27% higher than that at the center (11%). It is also an evidence of Ge out-diffusion to form nanorings. The result of higher Ge content at edge is different from molecular beam epitaxy (MBE) samples [16] because our growth temperature of 500oC is lower than the MBE temperature of 680 oC, and Si diffusion is not important in our work.
Note that our early work [17] at higher growth temperature (600 oC) has similar results of MBE samples [16] grown at 680 oC.
(a)
Fig.2-5 (a) The 3D AFM images of uncapped QDs
Fig.2-5 (b) The 3D AFM images of uncapped QDs after 1 h annealing in vacuum at 500oC to form nanoring. Note that the width of nanoring is larger than QDs due to Ge out-diffusion.
A
B
130 nm
~ 9 nm 30 nm
(width)
A B
(Width)
(b)
Atomic%
Ge 28%
Atomic%
Ge 11%
Atomic%
Ge 27%
50 nm
Fig.2-6 The high resolution TEM images and corresponding EDS measurement of the nanoring.
To extensively study the evolution of QDs to nanorings by in situ annealing at 500oC, different gases during in situ annealing are used for comparison. The base pressure during annealing is about ~10-9 torr in vacuum. Fig. 2-7 plots the nanoring density versus in situ annealing time in vacuum for the uncapped QDs. The density of uncapped QDs using SiH4/H2 for Si capping layer growth can increase from 3.5x108cm-3 to
To extensively study the evolution of QDs to nanorings by in situ annealing at 500oC, different gases during in situ annealing are used for comparison. The base pressure during annealing is about ~10-9 torr in vacuum. Fig. 2-7 plots the nanoring density versus in situ annealing time in vacuum for the uncapped QDs. The density of uncapped QDs using SiH4/H2 for Si capping layer growth can increase from 3.5x108cm-3 to