Chapter 2 Growth and Control of SiGe Nanoring Formation
2.4 Carrier Gas Effects on Annealing for Nanoring Formation
Except the Si capping layer effect on nanoring formation, the carrier gases during in-situ annealing for the possible nanoring formation are also studied. In the Fig
2-12(a)~(c), for capped QDs, the Ge cannot diffuse out to form nanorings by in-situ annealing in vacuum, in H2 and in He. Si cap is the crucial role to retard the Ge out-diffusion for nanoring formation. The Si capping layer has also been a good candidate to passivate the Si0.2Ge0.8 p-type channel metal oxide semiconductor field effect transistors [20].
1 μm
1hr anneal In vacuum
No Nanoring
1 μm
1 hr anneal In He
No Nanoring
1 μm
1 hr annealing In H
2No Nanoring
Fig.2-12 The AFM images (5m x 5m) of capped QDs with in-situ anneal at 500oC in vacuum, in He ambient and in H2 ambient.
For uncapped QDs, Ge can diffuse out to form nanoring when in-situ annealing at 500oC in vacuum for only one hour (Fig.2-13), where the base pressure in the chamber is about 10-9 torr. The nanoring density can increase with the in-situ annealing in vacuum time and reach 7.5x108 cm-2. Even for uncapped QDs, no nanorings can be formed by in-situ annealing in H2 or He at 500oC for 1hr. No nanoring can be formed from uncapped QDs even after annealing in H2 up to 5 h. The nanoring can be formed only with annealing both in vacuum and in He ambient (Fig. 2-13).
1 μm
Nanoring
1hr anneal In vacuum
1 μm
Nanoring
1 hr anneal In He
1 μm
No Nanoring
1 hr annealing In H
2Fig.2-13 The AFM images (5m x 5m) of uncapped QDs with in-situ anneal at 500oC in vacuum, in He ambient and in H2 ambient. Note no nanoring can be formed with in- situ anneal in H2even from uncapped QDs.
The Fig.2-14 is the model for uncapped QDs with different annealing. With the same hydrogen passivation effect, the QDs surface which covers with hydrogen can retard Ge out-diffusion. In the case with in situ annealing in He, H can be taken away by He flow and create unpassivated QDs, nanorings can be formed after 1 h at 500oC (Fig. 2-14).
Therefore the hydrogen can not only retard the Si capping layer growth for uncapped QDs formation, but also retard the Ge out-diffusion during anneaing. For the application, post annealing with hydrogen instead of vacuum can be used for SiGe channel devices without capping layer. The time evolution of Raman spectra for H2 annealing, in which no nanoring can be formed, shows a negative shift of Ge-Ge peak (Fig. 2-15), probably due to the diffusion of underneath Si into QDs.
HHHHH
Fig.2-14 The schematics for annealing effects on nanoring formation. Even for the uncapped QDs, the nanoring cannot be formed with H2 annealing by H passivation.
: Ge : Si
HHHHHH HHHHHH
X X
Fig.2-15 Raman spectra as a function of in-situ annealing time in H2 for uncapped QDs
2.5 Summar
yThe Si layer has been used to passivate the Si0.2Ge0.8 p-channel metal oxide semiconductor field effect transistors [20]. The Ge out-diffusion from QDs is the key factor to form nanorings at 500oC, and the nanoring formation is also a good indicator of Ge out-diffusion. Both the Si cap and the hydrogen passivation can reduce the Ge out-diffusion while the annealing in vacuum and in He can enhance the Ge out-diffusion to form nanorings. The conditions to form nanoring is summarized in Table 2-1 Using the carrier gas effect, Ge out-diffusion which is good for nanoring formation but not desired for integral circuit process can be well controlled.
Table 2-1 The summary conditions for nanoring formation at 500oC.
Annealing in vacuum 0 hr 1 hr 2 hr 4 hr 6 hr
Reference:
[1] G. Jin, J. L. Liu, and K. L. Wang, Appl. Phys. Lett. 83, 2847 (2003).
[2] Y. Q. Wu, F. H. Li, J. Cui, J. H. Lin, R. Wu, J. Qin, C. Y. Zhu, Y. L. Fan, X. J. Yang, and Z. M. Jiang, Appl. Phys. Lett. 87, 223116 (2005).
[3] H. J. Kim and Y. H. Xie, Appl. Phys. Lett. 79, 263 (2001).
[4] D. Granados and J. M. Garcia, Appl. Phys Lett. 82, 2401 (2003).
[5] G. Huang, W. Guo, P. Bhattacharya, G. Ariyawansa, and A. G. U. Perera, Appl.
Phys. Lett. 94, 101115 (2009).
[6] M. W. Kim and P.-C. Ku, Appl. Phys. Lett. 98, 201105 (2011).
[7] J. Cui, Q. He, X. M. Jiang, Y. L. Fan, X. J. Yang, F. Xue, and Z. M. Jiang,Appl.
Phys. Lett. 83, 2907 (2003).
[8] S. W. Lee, L. J. Chen, P. S. Chen, M.-J. Tsai, C. W. Liu, T. Y. Chien, and C. T. Chia, Appl. Phys. Lett. 83, 5283 (2003).
[9] C.-H. Lee, Y.-Y. Shen, C. W. Liu, S. W. Lee, B.-H. Lin, and C.-H. Hsu, Appl. Phys.
Lett. 94, 141909 (2009).
[10] W. A. P. Claassen and J. Bloem, J. Crytal Growth 51, 443 (1981).
[11] E. Rudkevich, F. Liu, D. E. Savage, T. F. Kuech, L. McCaughan, and M. G.
Lagally, Phys. Rev. Lett., 81, 3467, (1998).
[12] C.-H. Lee, C.-Y. Yu, C.M. Lin, C.W. Liu, H. Lin, and W.-H. Chang, Appl. Surf. Sci.
254, 6257 (2008).
[13] M. H. Liao, C.-H. Lee, T.-A. Hung, and C. W. Liu, J. Appl. Phys. 102, 053520 (2007).
[14] A. Marzegalli, V. A. Zinovyev, F. Montalenti, A. Rastelli, M. Stoffel, T.
Merdzhanova, O. G. Schmidt, and L. Miglio, Phys. Rev. Lett. 99, 235505 (2007).
[15] H. Liu and R. Huang, J. Appl. Phys. 97, 113537 (2005).
[16] M. Stoffel, A. Malachias, A. Rastelli, T. H. Metzger, and O. G. Schmidt, Appl.
Phys. Lett. 94, 253114 (2009).
[17] S. W. Lee, Y. L. Chueh, H. C. Chen, L. J. Chen, P. S. Chen, L. J. Chou, and C. W.
Liu, Thin Solid Films 508, 218 (2006).
[18] P. H. Tan, K. Brunner, D. Bougeard, and G. Abstreiter, Phys. Rev. B 68, 125302 (2003).
[19] C.-H. Lee, C. W. Liu, H.-T. Chang, and S. W. Lee, J. Appl. Phys. 107, 056103 (2010).
[20] 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, 12114 (2007).
Chapter 3 A Transition from Three to
Two Dimensional Si growth on Ge (100) substrate
3.1 Introduction
SiGe quantum wells (QWs) [1, 2] and quantum dots (QDs) [3, 4] have drawn much attention in the applications of nanoelectronics and optoelectronics. The three dimension growth mode, Stranski-Krastanov (SK) growth mechanism, has been well-known to dominant the SiGe epitaxial growth on Si substrate [5, 6]. A few monolayers of Si directly grown on Ge has been investigated to reduce the surface roughness of the Ge p-channel metal-insulator-semiconductor field effect transistor [7,8]. Moreover, a 10 nm doped epi-Si-passivation layer, which can eliminate Fermi level pinning, has also been demonstrated above the Ge substrate to reduce the contact resistivity [9]. However, only limit literatures have discussed about the growth mechanism of Si on Ge growth. The dot growth of 4~20 monolayer of Si on Ge(001) by molecular beam epitaxy (MBE) system [10] was studied over a wide range of growth
temperatures, and the growth mechanism of Si dots on Ge seems similar to that of Ge dots on Si. By the ultra-high vacuum chemical vapor deposition (UHV/CVD) system the growth can happen even at the temperature lower than 300oC due to the hydrogen desorption from the surface [11, 12]. However, the surface open site for SiH4 adsorb can change during growth. Due to lower Ge-H bond energy than Si-H, open sites can be created by the Ge segregation on Si surface during growth and will be responsible for the growth mode transition from the traditional three dimension (3D) to two dimension (2D) mode.
In this chapter, the transition from 3D to 2D growth for Si on Ge, which is different from the Ge on Si case, was observed for the first time by the atomic force microscopy (AFM) and the cross-sectional transmission electron microscopy (TEM). The strain in Si film on Ge was analyzed through the Raman and x-ray diffraction (XRD) using synchrotron radiation source.
3.2 Experiments
All the samples were grown by the UHV/CVD system at 550oC. The base pressure was ~10-9 torr. Pure silane (SiH4) at a fixed 100 sccm flow was used for Si growth. The Si is directly grown on Ge without the buffer layer. The Si dots grown at 550oC are shown in Fig. 3-1 with the dot density of ~ 7 ×108 cm-2 and the surface root-mean-square (RMS) roughness of ~1.21 nm. The much lower density as compared to the Ge dots on Si (~1010 cm-2) probably due to the impedance of three-dimensional dot growth by the tensile strain [13, 14].
There are two types of Si dots on Ge (Fig. 3-2). The average dot height (h) and dot base (b) for the type-I Si QDs are 15 nm and 144 nm, respectively; while those are 2.5 nm and 60 nm, for type-II QDs. The three-dimensional (3D) AFM images and the line profile of two types Si QDs are shown in Fig. 1(b). For the type-I Si QDs, the {1 3 11}, {1 0 5}, and {1 1 3} facets can be observe. However, for the type-II Si QDs, the low-index facets, such as {1 0 7} and {1 0 9} appear on the surface. All these facets which can be observed above the Si QDs have been defined from the Si QDs grown in oxide/nitride in the previous investigation. Lower index of type-II QDs facets could come from the lower surface energy. Note that the h/b ratio of the type-I Si QDs (0.1) is larger than that of the type-II Si QDs (0.04). The larger aspect ratio of the islands leads to larger strain relaxation. The results between the type-I and type-II Si QDs indicate a larger strain relaxation in the type-I Si QDs as compared to the type-II Si QDs. Besides, the aspect ratios of Si dots on Ge are from ~ 0.05 to ~ 0.1, which are smaller than the Ge dots on Si (0.13-0.17) [14].
The cross sectional TEM image of the type-II Si dot having the aspect ratio of ~ 0.05 is shown in Fig, 3-3. The wetting layer of the Si dot is ~5 nm, which is thicker than that of the Ge dot on Si (~1 nm). The thicker wetting layer of Si dot on Ge was also report in the ref.10. The observed thicker wetting is probably due to the tensile strain [13]. It is evident that the SK mode growth is still valid for the tensile strained Si growth on Ge (001) [10].
41
Fig.3-1 The AFM images (5μm × 5μm) of Si QDs grown on Ge substrate. Two types of Si QDs (indicated by dash circles) can be observed.
5.0nm 10nm
3.3 Transition from 3D to 2D Growth on Ge(001)
The transition from 2D to 3D growth of Ge on Si(001) is well-known as SK mode growth. The Ge islands continue to grow with increasing growth time and coalescences with other islands [21, 22]. The SK mode growth is still valid in the Si on Ge(001) growth [10].
Two types Si QDs, which have been discussed before, were observed when the Si film was deposited for 7.5 min (Fig. 3-2) with the wetting layer thickness of ~5 nm (Fig. 3-3).
However, while the Si thickness increases, the surface morphology changes. No dots were observed on the surface with the Si thickness of ~11 nm (Fig.3-4). Instead, the smooth surfaces (RMS roughness ~0.91 nm) with scattered ring-like structures were observed on the surface (Fig, 3-5 (a)). The density of the ring-like structure is ~ 2 × 107 Fig.3-3 The cross-sectional TEM image of a type II Si dot with the wetting layer of ~5 nm and the dot height of ~3 nm.
10nm
cm-2.
The Fig.3-6 shows the 3D AFM image and the line profile of the ring-like structure. The ring average width and depth is ~129.4nm and ~1nm, respectively. For Ge growth on Si, the ring could be due to the SiO2 particles from other’s work [15] or surface diffusion of adatoms to relatively strain-free regions from our early work [16].
For Si growth on Ge, since the strain is almost relaxed at Si dot facet on Ge substrate [17], the Si adatoms can diffuse to relatively strain-free facet regions and aggregate on the facet of the Si dot to form the ring-like structure in this work.
Moreover, no oxide at Si/Ge interface was detected by the EDS. The three-dimensional to two-dimensional transition of Si growth would not be expected if there was some oxide on the initial Ge surface. In the TEM image (Fig. 3-4), it is found that the dots disappeared after the growth of ~ 11 nm Si film. Due to the small aspect ratio (~0.007) of the ring-like structure, it is difficult to observe the ring-like structure on the cross sectional TEM image. Note that some dislocations were observed in the Si film which can relax tensile strain. The average width of the ring structures is smaller than the base width of the type-I Si QDs. Moreover, the ring density of 2 × 107 cm-2 is similar to the dot density of the type-I Si QDs. Based on these results, the ring-like structures in Fig. 3-6 may probably come from the type-I Si QDs. The specific growth mechanism for the Si directly grown on Ge should be responsible for this.
11 nm
dislocation 5 nm
Fig.3-4 The cross-sectional TEM image of the ~11 nm Si on Ge. Note the dislocation appears to relax strain.
Fig.3-5 The AFM image of the ~11 nm Si film with ring-like structures on the surface.
5.0nm 10nm 0.0nm
RMS roughness: 0.91nm
2.5m 5m
2.5m 5m
0m
3.4 Growth Model of Si growth on Ge(001)
The Ge and Si atoms have naturally 4.2% lattice mismatch, and the Ge film would under compressive strain when grown on the Si substrate. The SK growth mechanism should dominate the Ge on Si growth. When the Ge film becomes thicker than the critical thickness, the strain in the Ge film tends to relax, and the 3D islands will form on the surface. The Ge layers could transform into SiGe alloys due to Si/Ge.
Fig.3-6 The 3D AFM images and line profile of the ring-like structure. The average width and depth is ~129.4 nm and ~1nm, respectively.
3nm
0nm
~131 nm
A B
A B
3nm
0nm
~129.4 nm
A B
interdiffusion. With the increasing Ge growth, the SiGe QDs which have already formed above the surface will become larger dome due to more Ge atoms coalescence.
Moreover, the small islands can also form in the space without the SiGe QDs. That is why both the dot densities of the larger dots and small dots increase with the increasing Ge growth time at the beginning. In conventional, there are three different types of islands found from the Ge growth on Si: large multifaceted domes,[17] square-based pyramids,[18] and elongated {105} faceted hut clusters.[19]. Pyramids and domes usually form at high temperatures, whereas the much smaller hut clusters nucleate at lower temperatures.
When the SiGe QDs become bigger, there would be less space without SiGe QDs.
That is why the dot density would become lower after even longer SiGe QDs growth.
The dot density of the large dots may increase, but that of the small dots would drop with the increasing Ge growth time. The dot density ratio between the large dots and small dots may thus increases. From previous works [18]. the SiGe dot density is ~8 × 108 cm-2, while the average dot height and dot base width for the SiGe QDs are 16 nm and 124 nm, respectively. The dot density ratio between the large dot (h/b > 0.133) and small dot (h/b < 0.1) is ~0.6. When the growth time of Ge on Si increases to 20 min, the dot density slightly increase to ~1.1 × 109cm-2, and the average dot height and dot base width for the SiGe QDs increase to 22 nm and 154 nm, respectively. Both the dot density and the dot size increase with the increasing Ge growth time. Moreover, the dot density ratio between the large dots and small dots becomes ~0.9. More large dots appear on the surface as compared to the small dots after the further Ge growth.
From previous work about SiGe dots grown on different orientations Si substrate, the square-based pyramids by Ge growth on Si(001) at 550oC (Fig.3-7(a)). It is found
that the shape of the pyramids depends on the surface orientation. The shape is hexagonal for pyramids on Si(110) (Fig.3-7(b))and triangle for pyramids on Si(111) (Fig.3-7(c)), respectively[20]. The shapes of epitaxially grown islands usually follow the symmetry of underlying substrate [21].
For the growth of Si growth on Ge, the growth rate enhancement on the wetting layer plays a crucial role in the growth mode transition from 3D to 2D. The Ge can segregate on Si surface with the activation energy ~ 1.4eV [22]. The Ge segregation in our samples is observed by the energy dispersive x-ray spectroscopy (EDS) measurement. For the Si dot grown on Ge, the Ge content at the wetting layer surface is
~ 37%, which is much higher than that at the dot (~10%) ( Fig. 3-8 ).
Since the desorption energy of hydrogen from Ge (100) surface (~1.51eV) is lower than that from Si (100) surface (~2.05eV), the hydrogen desorption increases due to the increasing Ge coverage on the surface [23]. From the EDS, there is more Ge segregation
Fig.3-7 The AFM SiGe pyramid images on (a) Si(001) (b) Si(110), and (c) Si(111).
(a) (b) (c)
on the wetting layer (~5 nm) than on peak of Si dots for Si growth on Ge surface after the three-dimensional dot growth, more open sites can be created on dot peak than on wetting layer surface. It is known that not only Si atom adsorption, but also the hydrogen desorption is the crucial role for Si growth, which will be discussed with detail in Chapter 4. Two open sites are needed for one Si atom adsorption. Therefore more Ge segregation can yield a higher Si growth rate on the Si wetting layer than on the Si dots [24]. Higher subsequent Si growth rate at the wetting layer than the Si dots leads to the transition of three-dimensional Si dot growth to the two-dimensional Si film growth (Fig. 3-9).
[Ge] 37%
10 nm
[Ge] 10%
*
*
Fig.3-8 The EDS measurement for a Si dot on Ge. Note the Ge content is more on the wetting layer surface than on the QDs peak.
After ~15 nm growth of Si on Ge (Fig.3-10), neither Si dots nor ring-like structures were observed on the surface (Fig. 3-11). The surface RMS roughness is only about
~0.26 nm, which is similar to the bulk Ge substrate (~0.25 nm). From the EDS measurement (Fig. 3-11), the Ge content is ~47% near the bottom of “Si film” and gradually decreases to ~2% near the top due to Ge diffusion into Si. With the assistance of the enhanced growth rate at the initial Si wetting layer, the transition from 3D to 2D growth mode was observed for Si growth on Ge. For even thicker Si, less Ge content near the top is expected. Note that there have some dislocations found to relax tensile strain (Fig.3-10).
Fig.3-9 The growth model of Ge segregation effects of Si grown on Ge. Higher growth rate due to the more open sites on the wetting layer than on the dot leads to the smooth surface.
: Si
H
: Ge
H
H H H H
H H
H H H
H H
10 nm
[Ge] 2%
*
*
*
dislocation
[Ge] 47%
[Ge] 11%
Fig.3-10 The cross-sectional TEM images of ~15 nm Si grown on Ge and Ge content by EDS measurement.
Fig.3-11 The AFM images of the ~15 nm Si grown on Ge. Note the surface roughness (~0.26 nm) is similar to bulk Ge.
The evolution of strain in Si growth directly on Ge, which is determined by the lateral lattice constant, was analyzed by the in-plane XRD using synchrotron radiation source and Raman. Raman spectroscopies with 488 nm laser excitation were measured with resolution of 0.2 cm-1 to analyze the strain in the Si on Ge samples, as shown in Fig. 3-12. The Si-Si peak of bulk Si is located at ~520.8 cm-1. The increase of relaxation inside Si film can lead to a positive shift of Si-Si peak. Note that the Ge-Ge peaks of all these samples does not compare here. No peak shifts of the Ge-Ge peaks can be found in this case. The Ge-Ge peaks in the Si on Ge samples should come from the Ge substrate, and the volume of the Ge substrate is much larger than that of the Si film.
Only the lattice constants for the Ge atoms near the Si/Ge interface become smaller, while those in the Ge substrate are still the same as the initial stage. The strain response of the Ge atomic layers near the Si/Ge interface is too weak, and can hardly be observed.
For the Si-Si peak, when the Si QDs were initially grown, the Si-Si peak shift to a lower wave number (517.5 cm-1) as compared to the bulk Si case (520.8 cm-1), which means the Si film may suffer ~0.4% tensile strain in this stage. This can be easily understood since the formation of the Si dots can relax the strain inside the Si film (Fig.3-1). However, the formation of the Si dots can only partially relax the tensile strain and remain ~0.4% tensile strain in Si. With the increasing Si deposition (11 nm and 15 nm), the strain relaxations become even larger due to the appearance of threading dislocations in the Si layer (Fig. 3-4 and Fig. 3-10). The Si-Si peaks of these two samples show negative wave number shifts and moves much closer to the bulk Si peak.
The XRD in-plane radial scans using synchrotron radiation source across Si (400) and Ge (400) peaks for the samples with Si dots and 15 nm Si on Ge are shown in
The XRD in-plane radial scans using synchrotron radiation source across Si (400) and Ge (400) peaks for the samples with Si dots and 15 nm Si on Ge are shown in