Chapter 4 Growth Rate of SiGe and Ge on Si(001)
4.1 Growth model by Chemical Vapor Deposition
Silane undertakes the following reactions for CVD growth in ultra-high vacuum (UHV) conditions (clean silicon surface) [1]:
A. SiH4(g) => Si(s) + 2H2(g) [total mechanism]
B. SiH4(g) + 2 (*) => H* + SiH3(ad) [silane adsorption]
C. SiH3(ad) + (*) => H* + SiH2 (ad) [H decomposition]
D. 2SiH2(ad) => H2(g) + 2SiH (ad) [H decomposition, H desorption]
E. 2SiH (ad) => H2 (g) +2 Si (ad) [H decomposition, H desorption]
F. 2H (ad) = > H2(g) + 2 (*) [H desorption]
G. Si (ad) => Si (s) + (*) [surface migration]
59
where (*) denotes an open site (a dangling bond) for adsorption, (g) represents gas phase, H* is an adsorbed surface hydrogen atom, and (ad) represents an adsorbed species. Equation (A) is the overall surface reaction for the growth process with silane.
Equations (B) through (G) are the individual reaction steps. If we balance Equations (B)-(G) (divide Equations (D) and (E) by 2) we end up with the original reaction (Equation (A)). Equation (B) is the silane adsorption step. Equations (B)-(E) are the hydrogen decomposition steps. Equation (F) is the open site generation step via mono-hydride desorption. Equation (G) is the surface diffusion step for silicon.
According to Gates, [1], [2], and [3], Equation (b) is the rate-limiting step when there are open sites available (i.e. the growth rate is limited by silane (SiH4(g)) adsorption, controlled by the number of open sites (*)). From the Fig. 4-1 shown below, illustrates Equation (B) describing how silane adsorbs onto the hydrogenated 2x1 reconstructed silicon surface. In order for silane to adsorb on the surface, it also requires two adjacent open sites to be generated on the growing surface. In Fig. 4-1, only the adsorption on two adjacent open sites from separate dimer rows is shown. The silane adsorption can occur on any two adjacent open sites. Silane dissociatively adsorbs, splitting a Si-H bond and forming SiH3 on one site and an H on another open site. At the same silane partial pressure; the open site fraction is significantly less in a hydrogen environment.
Fig.4-1 The side view of silane adsorption reaction on a (2 x 1) reconstructed silicon surface. The dashed triangle underneath the silicon surface represents the reconstruction of the surface silicon atoms in/out of the plane of the figure
Fig. 4-2 below illustrates the adsorption steps for silane on a hydrogenated 2x1 reconstructed silicon surface which is established by S.M. Gates. Equations (B) to (D) and Equation (F) are illustrated in the Fig. 4-2. Note that in the Fig. 4-2, the adsorption of silane is shown along the dimer row as opposed to adsorbing on two separate dimer rows in Fig. 4-1. (a) illustrates the adsorption of silane as a SiH3 and a hydrogen atom (Equation (B)). (b) Depicts hydrogen desorption (Equation (F)) to form two open sites for (c) the SiH3 to split into SiH2 (Equation (C)). Another two hydrogen desorption step occurs in (d) to create two open sites and for SiH2
to be reduced to SiH (Equation (d)). Step (e) shows the SiH bonding with its two nearest neighboring silicon atoms (i.e. incorporated into the solid) and the new reconstruction (f) on the surface. Therefore, not only the adsorption of the precursor, but also the open sites created by hydrogen desorption on the surface are the crucial rules of the dissociated precursor molecules adsorption for Si growth. To realize the effects of carrier gas, precursor on hydrogen desorption and precursor adsorption, the SiGe and Ge grown on Si(100) by ASM Epsilon RTCVD are discuss below.
Fig.4-2 Top view of the adsorption process of silane onto a Si:H (100) (2x1) reconstructed surface based on references [1],[3]. Red circles indicate the change, and a red line indicates an open surface site.
4.2 SiGe Growth and Characteristics on Si(001)
The applications of SiGe film as high mobility channel and stressor are discussed in chapter 1. Having higher Ge% can have more benefits for the channel and stressor. The SiGe films using liquid dichlorosilane (SiCl2H2, DCS) for Si source and 10% germane (GeH4) in H2 as Ge source are studied here. The growth temperature and pressure is 650oC and 80 torr, respectively. To minimize the contamination, wafer is directly loaded into the load-lock chamber after the 10% HF dips with the subsequent 1100oC prebaking in H2 for 2 minutes. H2 with the flow rate at 20 standard liter per minute is fixed in the whole growth process. Having better epitaxial surface than wafer, the ~100 nm Si buffer layer was grown firstly on Si wafers. The SiGe layer using gas flows of DCS fixed at 130sccm and GeH4 are 25~250sccm, respectively, to grow SiGe with increasing Ge content. The morphology and thickness are observed by atomic force microscope (AFM) and the cross-sectional TEM images. Besides, the quality and composition are determined by x-ray diffraction (XRD) and photoluminescence (PL) at
~60K taken in liquid nitrogen. The excitation source is 671nm laser with power of 2W/cm2.
The -2 scans around the (004) XRD which uses Cu K1 radiation with wave length at 0.15406 nm and is selected by a Ge (220) four reflection channel-cut monochromator is used to study the Ge content of SiGe films. The incident beam provides a beam having a divergence in the scattering plane (00L) of about 12 arc sec.
With the X'pert epitaxy by Philips simulator, the fully strained Si1-xGex with x~0.27 is demonstrated in the Fig. 4-3. The lattice constant of Ge and Si for simulation is 0.5657nm and 0.5431nm, respectively. And the Possion ration for simulation is 0.270 and 0.279 for Ge and Si, respectively.
The PL spectrum using 671nm laser in Fig.4-4 (a) shows the strongest SiGe peak with no-phonon (NP) peak and transverse optical (TO) peak at lower energy. The NP peak is attributed to lattice disorder (mainly alloy fluctuations) and relaxes the momentum conservation requirement. The TO phonon replicas relates to Si-Si, Si-Ge, and Ge-Ge vibrations. Using an electron-hole plasma (EHP) model [4], bandgap of extracted SiGe from the cut-off NP peak of out sample is 901meV.
The formation of the EHP phase is discussed below. At low carrier injection, when an electron in the valence band is excited into the conduction band, the free electron and hole are bound together through coulomb attractive forces and exist in the excitonic phase [11] [12]. As the carrier concentration increases above (3±1)×1016 cm-3 according to the Mott transition [13], due to the screening effects among the electrons and holes, the bound exciton dissociates into the electron hole plasma. The spectral distribution of the infrared radiation is considered as the result of the electron-hole recombination in the degenerate non-equilibrium plasma. We can thus fit the luminescence spectra by a convolution of the electron and hole distribution functions.
The spectra from the EHP can be modeled with the equation shown below:
(4.1)
where I0 is relative intensity, De and Dh are the density of states of electron and hole, Efn
and Efp are the quasi-Fermi energies, h is the energy of the emitted photon, T is the temperature, Eg is the low-energy edge of the spectrum, fe (E, Efn, T) is the Fermi-Dirac distribution of electrons and fh(hv -Eg-E, Efp, T) is the Fermi-Dirac distribution of holes.
The quasi-fermi energy of electrons and holes is defined to be zero at the conduction 𝐼(ℎ𝑣) = 𝐼0 𝑑𝐸 ∙ 𝐷𝑒 𝐸 ∙
ℎ𝑣−𝐸𝑔
0 𝐷ℎ(ℎ𝑣 − 𝐸𝑔 − 𝐸) ∙ 𝑓𝑒(𝐸, 𝐸𝑓𝑛, 𝑇) ∙ 𝑓𝑒(ℎ𝑣 − 𝐸𝑔 − 𝐸, 𝐸𝑓𝑝, 𝑇)
band minimum and valence band maximum, respectively, and is positive when deeper into the bands. The Fermi-Dirac distribution for electron and holes are shown below:
(4.2)
The density of states has a three dimensional distribution for both holes and electrons for PL measurements, is due to the localization in the accumulation layer. Taking the 3D density of states for the electrons and holes, respectively, are given below:
(3.3)
Substituting the above equations into the EHP equation would lead to the EHP model for PL measurements. The theoretical spectrum from the EHP model is a convolution between the electron and the hole population. Based on the Bean’s calculation about the function of Ge content on strained SiGe energy bandgap [5], the extracted Ge content is also ~28%. Higher Ge content estimated from PL than from XRD could come from the fully strain estimation of SiGe. Partial relaxation could make higher Ge content
64
Fig.4-3 The -2 XRD of SiGe on Si(001). Note the Ge content is 27% under the fully strain fitting.
Fig.4-4 The PL spectrum of Si0.73Ge0.27 on Si with EHP fitting. The bandgap extracted from the cut-off of NP peak is 907meV.
33 34 35 36
The cross-section high resolution TEM and scanning transmission electron microscope (STEM) images of 7.4nm Si cap/24.8nm Si0.73Ge0.27 on Si shows no dislocation to relax strain (Fig.4-5(a)) and the interface is abrupt (Fig. 4-5(b)). Using STEM, which has the higher-angle detector, the atom with higher atomic number can be brighter than the smaller ones From the AFM (Fig.4-6), no cross-hatch is found and RMS roughness is only 0.1nm, which is smooth as the Si substrate surface (~0.1nm).
No strain relaxation through the island or dislocation growth is found. The abrupt interface between SiGe/Si reflects no interdiffusion during the growth. Base on the SiGe thickness from TEM, the SiGe growth rate using DCS/GeH4= 130/25 is ~0.41nm/sec.
Fig.4-5 (a) The high resolution TEM of Si(7.4nm)/Si0.73Ge0.27 (24.8nm)/Si buffer on Si(001). No dislocation is found. (b) From the STEM images, abrupt interface shows no interdiffusion between Si0.73Ge0.27 and Si.
5 nm
(a) (b) Si cap 7.4 nm
With increasing GeH4 from 25sccm to 250sccm, the Si0.5Ge0.5 can be obtained, in which the Ge content is also determined by XRD and PL (Fig. 4-7). High quality Si cap /Si0.5Ge0.5 on Si is realized with surface roughness at ~0.21nm (Fig. 4-8). No cross-hatch on the surface also indicates the SiGe film is dislocations free (Fig.4-8).
From SiGe thickness by the TEM, the 5nm Si cap/17nm Si0.5Ge0.5 on Si reflects the growth rate of Si0.5Ge0.5 is 2.9nm/sec. To study the Ge content effects on the SiGe growth rate, growth rate with different Ge content (x) of Si1-xGex film is studied in Fig.
4-10. It is found that, with increasing GeH4, not only Ge content can increase but also the SiGe growth rate can be enhanced (Fig.10). From the previous discussion about growth rate, the H desorption is the crucial role on growth rate. For one atom adsorbed, two open sites are needed. The catalytic effect of germane on the growth rate comes from the lower H desorption energy from Ge (1.5eV) than from Si (2.1eV) [6].
RMS roughness=0.1nm
20
(nm) 0 10
0um 5um
5um
Fig.4-6 The AFM image of Si/ Si0.73Ge0.27/Si. The RMS roughness is only 0.1nm, which is close to Ge substrate surface. Note that no cross-hatch is found on surface.
67
Increased hydrogen desorption from Ge sites on the growing surface leads to more open sites on surface. Since the decomposed DCS converts to HCl by H2, the non-linear growth rate enhancement of SiGe using DCS and GeH4 could come from the etching effect of HCl.
Fig.4-7 (a) The -2 XRD and (b) the PL spectrum of Si0.5Ge0.5 on Si. Note the Ge%
extraction from both XRD and PL is close to 50%.
Fig.4-8 The AFM image of Si/Si0.5Ge0.5/Si. The RMS roughness is only 0.21nm.
Note that no cross-hatch is found on surface.
32 34 36
600 700 800 900 1000 1100 1200 1300
0.0
Fig.4-9 The high resolution TEM of Si(5nm)/Si0.5Ge0.5 (17nm)/Si buffer on Si(001). No dislocation is found and interface is abrupt.
Fig.4-10 The SiGe growth rate of Si1-xGex with 0.27<x<0.5 Note the SiGe growth rate increases with increasing Ge content.
25 30 35 40 45 50
1 2 3
SiGe g ro wth rate (nm /sec)
Ge%
SiGe by DCS/GeH
4
T=650oC P=80torr
4.3 Carrier Gas Effects on SiGe Growth Rate on Si(001)
Using pure SiH4 and the GeH4 with the flow rate at 20sccm and 19sccm, higher growth rate of 0.56 nm/sec can have than its using DCS (130sccm)/GeH4(25sccm).
However, the SiGe using DCS /GeH4 can have relatively stronger PL intensity of SiGe peak (Fig.4-11) than using SiH4/GeH4. The etching effect of HCl, which can clean and etching the growing surface, could be the reason for stronger PL and lower growth rate.
Using the same methods discussed before, the Ge content using SiH4/ content and stronger SiGe PL can have by DCS/GeH4
To observe the carrier gas effect on SiGe growth, the N2 is used instead of H2 for comparison under other conditions are fixed. It is found that when using SiH4/GeH4 the growth rate can be enhanced three times (Fig. 4-12(a)). However, when using DCS/GeH4, the growth rate is 35% reduced Fig. 4-12(b).
In conventional CVD there are two growth regimes. One is mass-transport-limited, which is limited by the transport of the precursor to the surface of the wafer, via diffusion through a boundary layer. The other is the reaction-rate-limited which is limited by surface reaction between the gas and the surface. The former occurs at high growth temperatures and the latter occurs at low growth temperatures. For the SiGe growth at 650oC, it is in the reaction-rate-limited regime, the growth rate is dominated not by the gas flow but by the reaction-limited region, where both the H desorption to create the surface open site and the adsorption of adatom are the crucial roles. Note that two open sites needed on the growing surface for silane adsorption. When using SiH4/GeH4 in the N2 ambient, open sites created only at Ge sites. On the other hand, in the case using DCS/GeH4, low H2 partial pressures disable the Si adsorption by SiCl2
removing the Cl with H2 and makes lower growth rate and higher Ge%.
SiCl2 + H2 Si+HCl (g)
However, when using SiH4/GeH4 in the H2 ambient, even the partial pressure of SiH4 is larger than GeH4, lower SiH4 adsorption rate than GeH4 makes large Ge%. In the N2
ambient, where most of open sits are available and adatom adsorption is irrelevant, larger SiH4 partial pressure than GeH4 leads to low Ge%
4.4 Ge Growth on Si(001)
Due to the ~4.1% lattice mismatch between Si (5.431 Ǻ) and Ge (5.658 Ǻ), hetero-epitaxial growth of Ge on Si beyond critical thickness generally results in the formation of misfits and associated threading dislocations even islanding .There\are a number of ways to grow high quality single crystalline Ge on Si including using a graded GexSi1-x buffer [8, 9] or overgrowing Ge in a Si/SiO2 template [10]. Here we use a two-step growth approach [11, 12] with cycling growth and anneal to grow high quality Ge on Si substrate.
The low-temperature buffer layer is to prevent the Ge from Stranski-Krastanov (S-K) Fig.4-12 The SiGe growth rate and Ge content using (a) SiH4/GeH4 with H2 and
growth. Base on S-K growth theory, only Ge layer less than its critical thickness (~101 nm) can be free from dislocations and induce strain (compressive strain in this case) in Ge. Above critical thickness, the energy of the strain stored in the Ge can be relaxed by the formation of misfit dislocation (a plastic process) or three-dimensional island. To prevent such island growth, the temperature must be low enough to kinetically prohibit adatoms moving to form islands with adequate kinetic energy. Besides, hydrogen on the surface can act as surfactant at low temperature and can hinder the nucleation of the 3-dimensional islands [13]. In practice, the low temperature can also prevent the complete equilibrium condition where dislocation may start to form even before the critical thickness. Once the strain energy is fully relaxed plastically by misfit dislocation, raised Ge growth temperature for growth rate at higher temperature can have on the low-temperature Ge buffer layer, which is in the homo-epitaxial case. The low temperature 320oC with 375oC is used for low and high growth temperature for our Ge, respectively.
Cycling thermal annealing at 825oC after growth is performed to reduce decrease threading dislocations which propagate to the edge of the substrate at annealing. In the ref. 14, 1-2 order of reduction of threading dislocation density can have using thermal annealing at 900oC. These threading dislocations act as both defect centers (non-radiative recombination centers) which degrade the electrical and optical properties of material. Fig. 4-13 is the 4.1mm epitaxial Ge on Si(001) with the 2.7x106 cm-2 dislocation density which is observed by etching pits using Schimmel etching. Due to the larger thermal expansion coefficient of Ge than Si, Ge tends to shrink more by
post annealing and induces ~0.16% in-plane tensile strain in Ge film upon cooling (Fig.
4-14).
The room temperature PL is taken for comparison with bulk Ge. The PL peaks at 695 meV and 780 meV are attributed to the indirect band transition and direct band transition, respectively (Fig.4-15). For the radiative recombination in an indirect bandgap material such as Ge, momentum conservation is achieved through the longitudinal acoustic phonon (28 meV) assisted transition [15]. In a direct transition, the momentum conserved is achieved without phonon. The energy difference of conduction band minimum between valley and L valley is 140 meV at room temperature [16].
The spectra of the indirect and direct band gap emission in both bulk Ge and epi-Ge on Si sample were fitted by using the electron-hole plasma recombination model [17] [15]
and the direct band gap recombination model [18], respectively. The band tail of absorption edge model is also taken into consideration in the direct band gap recombination model [19]. The integrated PL intensity ratios of direct to indirect band gap transition for the bulk n-Ge and epi-Ge on Si sample are 0.05 and 2.5, respectively.
For the indirect radiative recombination in bulk Ge, a dominant LA phonon (~28 meV) is involved to satisfy the momentum conservation between the L valleys and the zone center in the valance band.
For the epi-Ge on Si sample, the large defect density in the Ge film may lead to a spread of trap levels in momentum space. These trap levels enhance the non-radiative recombination rate because the more phonons with different momentum can be involved in the momentum conservation. Thus, the relative intensity of the indirect band gap transition is lower in the epi-Ge on Si sample than in the bulk Ge [20].
Threading dislocation density: 2.7 × 106cm-2
4.1um
Fig.4-13 The high resolution TEM image of 4.1m Ge on Si(001). The inset is the Nomarski micrograph of Ge with Shimmel etching for etching pits observation.
Fig.4-14 The -2 XRD of 4.1 m Ge on Si(001). Note the 0.16% tensile strain in Ge film.
64 65 66 67 68 69 70 71 72
Inten sity (a.u.)
2
(degree)
4.1um Ge/Si X'pert Fitting relaxed Ge
e~0.16%
To observe the carrier gas effect on Ge growth, 375oC, 40torr with nitrogenand hydrogen are used for comparison. Both samples have post annealing in hydrogen at 825oC to reduce the threading dislocations. It is found that the Ge growth rate enhanced three times in N2 over in H2 (Fig.4-16 (a)). At 375oC for Ge growth, it is in the reaction-rate-limited regime, where H desorption more in nitrogen ambient can be the main reason for higher growth rate than in hydrogen. More open sites created in nitrogen ambient also lead to smoother surface than in hydrogen(Fig.4-16 (b)). Using N2 as carrier gas strongly reduces the H2 partial pressure compared to a deposition in hydrogen as carrier gas. The balance between the germane and hydrogen adsorption and the hydrogen desorption is modified depending on the nature of the carrier gas [21].
Fig.4-15 The PL spectra of the bulk Ge and epitaxial Ge on Si at room temperature.
The integrated PL intensity ratios of the direct to indirect band gap transition of the bulk n-Ge and epitaxial Ge on Si sample are 0.05 and 2.5, respectively. The bulk Ge is ~15 times the integrated intensity of PL of the epitaxial Ge sample.
Lower hydrogen partial pressure favors germane adsorption and enhance hydrogen desorption to enhance Ge growth rate.
The surface roughness reduction can be explained in terms of the Ge diffusion barrier as compared to the Ge–H cluster [22]. The effect of attaching H to Ge reduces the with surface RMS roughness of 2nm
(a)
4.5 Summary
To reduce the cost of tool investment for epitaxial growth by CVD, the high growth rate of epitaxial layer is crucial for device development. With the growth model by CVD, the growth rate can be changed by Ge content of SiGe film, precursors, or carrier gas. Due to lower H desorption energy from Ge than from Si, more open sites created leads to higher growth rate when Ge% increasing. When using SiH4/ GeH4 for SiGe films and GeH4 for Ge films growth with N2 instead of H2 as carrier gas. The growth condition having more surface open site in N2 than in H2 is the main reason for growth
To reduce the cost of tool investment for epitaxial growth by CVD, the high growth rate of epitaxial layer is crucial for device development. With the growth model by CVD, the growth rate can be changed by Ge content of SiGe film, precursors, or carrier gas. Due to lower H desorption energy from Ge than from Si, more open sites created leads to higher growth rate when Ge% increasing. When using SiH4/ GeH4 for SiGe films and GeH4 for Ge films growth with N2 instead of H2 as carrier gas. The growth condition having more surface open site in N2 than in H2 is the main reason for growth