High-resolution X-ray diffraction and micro-Raman scattering studies of Ge(:Ga) thin films grown on GaAs (001) substrates by MOCVD

Abstract

We have employed high-resolution X-ray diffraction (HRXRD) and micro-Raman scattering to study the structural and lattice vibrational dynamic properties of heavy Ga-doped Ge thin films (i.e., Ge:Ga) epitaxially grown by metalorganic chemical vapor deposition on GaAs (001) substrates. Reciprocal space mapping revealed that the ∼1.0 μm thick Ge:Ga films are coherently stressed on the GaAs (001) substrates and the in-plane compressive strain increases with Ga incorporations. In contrast, ∼90% strain has been relaxed in a 10 μm thick unintentionally doped Ge thin film. The compressive strain caused by such Ga incorporations in the Ga-doped Ge thin films, having a maximum of 866 PPM, plays a minor role in the Raman shift of the Ge–Ge longitudinal optical (LO) phonon that has been observed up to −11.63 cm⁻¹. The large phonon softening has been discussed on the bases of hole concentrations and ‘alloy-disorder’ of the ‘self-annealing’ induced atomic interdiffusions, specifically with the help of Raman scattering and HRXRD from the cross-section of the Ge(:Ga)/GaAs (001) heterostructures.

---

I. Introduction

Germanium (Ge), one of the first generation semiconductors, has been extensively investigated and matured to industrial applications. Recent developments in tandem solar cells and broadband photodetectors have caused an increasing research interest in materials integration between group-IV and III–V semiconductors.^1,2 Monolithic growth of Ge-on-GaAs and/or GaAs-on-Ge has thus been a natural focus for band broadening because of the small lattice mismatch (∼0.083%) between Ge and GaAs. Unfortunately, atomic interdiffusions across the grown interface during the overgrowth caused by the ‘self-annealing’ effect pose a great challenge,^3,4 beside the polar/nonpolar mismatch problem when growing GaAs-on-Ge,⁵ for materials integration of Ge and GaAs by any epitaxial growth method. Because of atomic interdiffusions, doping control of the epilayer is quite difficult; violent interdiffusions may even give rise to solid-state reactions and recrystallizations at the heterostructural interfaces.⁶ In Ge crystal, Ga and As atoms are p- and n-type dopants, respectively. However, beyond the doping levels, recrystallization of Ga_xGe_1−x, Ge_1−xAs_x, and/or (GaAs)_1−x(Ge₂)_x alloys may also occur.^7,8

Lattice vibration dynamics has long been studied in understanding the doping behavior of group-III atoms in Ge, which is also complicated in monolithic Ge(:Ga)/GaAs heterostructures due partly to the self-annealing induced atomic interdiffusions. Sanson et al.⁹ have studied Ge:Al (i.e., Al-doped Ge) by employing micro-Raman scattering coupled with a small-angle beveling technique. The large difference in the atomic weights of Al and Ge makes the localized vibration mode of Al in Ge (LVM_Al) matrix, i.e., substitutions, detectable at heavy doping levels. By monitoring the Raman shift of the Ge–Ge longitudinal optical (LO) phonon and the intensity evolutions of LVM_Al as a function of doping concentrations (depending on depth in the implanted and annealed samples, see ref. 9), they observed a strong correlation between the phonon-softening of LO_Ge–Ge and the intensity increases of LVM_Al and thus the hole concentrations of Ge. However, the doping activation efficiency was unclear in their studied samples. In other words, the effects of hole concentrations and inactivated Al atoms (e.g., interstitials and/or their splitting complex associated with Ge vacancies) on the observed phonon-softening are still convoluted; moreover the LVM_Ga mode is indistinguishable in Ge:Ga due to the small difference in the atomic weights of Ga and Ge. Recently, Kabuyanagi et al.¹⁰ have provided direct evidence for hole-induced Raman shifts in Ge thin films via back-gate bias induced hole-accumulation in p⁻ or p⁺-type Ge channels of a field-effect transistor. In their experiments, the variation of dopants and thus the effect of ‘alloy-disorder’ on the Raman shift of LO_Ge–Ge were completely excluded. Unfortunately, the hole concentrations accumulated by the back-gate tuning technique reported in ref. 10 were limited to n_h < 1.0 × 10¹⁹ cm⁻³.

In this work, we have studied epitaxial growth of Ge thin films, with and without intentionally incorporated Ga atoms, on GaAs (001) substrates by metalorganic chemical vapor deposition (MOCVD). For these studies we have employed high-resolution X-ray diffraction (HRXRD), micro-Raman scattering spectroscopy, energy-dispersive X-ray spectroscopy (EDX), and Hall-effect measurement. The results obtained indicate a phonon-softening of LO_Ge–Ge, up to 11.63 cm⁻¹ at a hole concentration of 2.87 × 10²⁰ cm⁻³, with the increase in Ga incorporations. By comparing the evolutions of cross-sectional Raman spectra and the HRXRD curves across the interfaces of the Ge:Ge/GaAs and Ge/GaAs heterostructures, we shed some light on determining the effects of hole concentrations and alloy-disorder on the phonon-softening of LO_Ge–Ge.

II. Experiment

The Ge(:Ga) films were grown at 550–630 °C by MOCVD on unintentional miscut GaAs (001) substrates employing germane (GeH₄), trimethylgallium (TMGa), and tertiarybutylarsine (TBAs) as the Ge, Ga, and As reaction precursors, respectively. The precursors, when feeding into the reactor, were carried by highly purified H₂ and the chamber pressure was set at 10 kPa during growth. Before loading into the growth chamber, the GaAs (100) substrates were cleaned in a hydrogen fluoride solution (10%), followed by sonication in acetone, isopropenyl acetylene, and deionized water in sequence. The residual oxide on the GaAs surface was thermally decomposed at 600 °C under TBAs for a few minutes. To improve the surface smoothness of the substrate, a GaAs buffer layer (∼100 nm thick) is usually grown after the surface treatment. For growing Ge thin films, the flow rate of GeH₄ was set at 920 μmol min⁻¹. To incorporate and vary the Ga concentrations, the flow rate of TMGa was adjusted in the range of 5.6–8.4 μmol min⁻¹. More detailed growth parameters and procedures can be found in our earlier publications.^11,12 For the current study, three kinds of samples were selected. Sample A was unintentionally doped, ∼10 μm in thickness, and grown at 575 °C; sample B was Ga-doped, ∼1.0 μm thick, grown at 550 °C; and sample C was heavier Ga-doped than sample B, ∼1.0 μm thick, grown at 600 °C. They are summarized in Table 1.

Table 1 Summary of the studied Ge(:Ga) epitaxial layers on GaAs (001) substrates

Samples	Thickness (μm)	TGrowth (°C)	Ga-Doped	Lattice constant^a (Å)			ε^∥ (PPM)
				a^∥	a^⊥	a
a The lattice constant of GaAs (5.6533 Å) was taken as a reference.
A	10	575	No	5.6575	5.6584	5.6581	−106
B	1.0	550	Yes	5.6533	5.6616	5.6580	−830
C	1.0	600	Yes	5.6533	5.6619	5.6582	−866

---

PANalytical HRXRD (CuKα₁) system designed for large wafers (up to 8-inch diameter) was used in this study. For reciprocal space mapping (RSM), a PIXel 3D detector was used to speed up the high quality data collections; however, a triple-crystal setup with a position sensitive detector and a crystal analyzer was used for 2θ–θ scans to improve the resolutions. Micro-Raman scattering was carried out at room temperature using a confocal micro-Raman system (WITec alpha 300). An argon ion laser with the beam wavelength of 532 nm was used as the excitation source, the beam size, when focused on the sample surface, was about 0.9 μm. The step of the sample-stage movement for collecting the backscattering Raman spectra from the cross-section of the Ge(:Ga)/GaAs heterostructures can be decreased to ∼10 nm. The EDX spectra were collected in the chamber of a field-emission scanning-electron microscope (FESEM); point focus and linear-scanning modes were used for analyzing the localized elements and the elemental distributions, respectively.

III. Results and discussion

A. Morphology comparisons

Fig. 1(a) and (c) show the typical photographs taken from samples A and C, respectively. Likewise, Fig. 1(b) and (d) present their surface morphologies reordered by atomic-force microscopy (AFM) using a tapping mode, showing a significant difference between the thick Ge and thin Ge:Ga films. A comparison between Fig. 1(a) and (b) indicates that the dark networks in Fig. 1(a) and the corresponding bright ones in Fig. 1(b) are walls of surface pits. A careful look at Fig. 1(a) and (b) reveals that the surface pits are formed in the shape of inverted square pyramids (ISP) as schematically shown in the inset of Fig. 1(a). The cycles in Fig. 1(b) highlight the typical ISP-like pits; many of the others are adjacently truncated along various in-plane directions, forming the observed networks. These truncated ISP-like pits are more clearly seen in Fig. 1(a) where the regular and collectively oriented white lines correspond to the lateral edges of ISP (a typical structure is indicated by the arrow). Their orientations with respect to the in-plane orientation of the GaAs (001) substrate indicate that the side faces of the ISP pits are (11n) atomic planes of Ge. A height profile collected along one of the base edges of the ISP pits is shown in Fig. 1(e). One sees that both the upward- and the downward-slopes exhibit paralleling characters. This observation suggests that the side faces of the ISP pits are of a same group of Ge (11n) atomic planes. However, the value of n derived from Fig. 1(e) is not reliable since the sectional scan across the apex of the individual pits is not guaranteed due to the lateral truncations of adjacent pits. Nevertheless, it is quite clear that the dark networks in Fig. 1(a) and the corresponding bright ones in Fig. 1(b) originate from the surface ISP pits of the Ge epilayer rather than islands coalescence induced domains.

Fig. 1 Surface morphologies recorded by optical microscope (a) and AFM (b) from sample A, i.e., a 10 μm thick unintentionally doped Ge on GaAs (001) substrate. Likewise, (c) optical micrograph and (d) AFM image recorded from sample C, i.e., a 1.0 μm thick Ga-doped Ge on GaAs (001) substrate. The scale bars in (a)–(c) are 10 μm and that in (d) is 2 μm. The cycles in (b) indicate typical inverted-square-pyramid (ISP) pits; an ISP pit is schematically shown in the inset of (a). The arrow in (a) indicates a typical top-view of a surface pit. (e) Height profile derived from the AFM image in (b) along a direction parallel to the base edge of the ISP pits.

In general, the presence of inverted-pyramid-like pits on epitaxial films is related to threading dislocations that formed to relax the lattice strain during the epitaxial growth of heterostructures. This phenomenon is frequently observed, and even unavoidable, in epitaxy with large lattice mismatches such as GaN on sapphire or Si. The end of threading dislocations connects the apex at the bottom of the surface pits.¹³ Although the mismatch between Ge and GaAs is small (∼0.083%), the lattice stress accumulates and builds-up with the increase in film thickness, which is eventually relaxed via generating dislocations when the film thickness exceeds a critical value [i.e., critical thickness, ∼2.0 μm for Ge-on GaAs (001)]. The strain relaxation of sample A has been further confirmed by the HRXRD measurements, typically from its cross-section, which reveals an apparently peak separation of Ge (220) and GaAs (220) (see next section).

In comparison, when Ga is introduced with a flow rate of TMGa in the range of 5.6–8.4 μmol min⁻¹ into the growth of Ge, Ga-rich droplets are formed on the surface [see Fig. 1(c)–(d) of sample C and detailed dot-formation mechanism in ref. 12]. Localized EDX analysis on the film surface, at the flat areas away from the dot structures, reveals that the Ga concentration increases from 5.2% to 5.6% and 5.8% when the growth temperature is increased from 550 °C (sample B) to 600 °C (sample C) and 630 °C, respectively. It has to be noted that due to the segregation of Ga atoms towards the film surface during the growth (the solid solubility of Ga in Ge is ≤1.1 at%), the Ga concentrations might be overestimated by EDX. However, their relative variations as a function of the growth temperatures are consistent with the evolution of the surface dot structures (see ref. 12) and thus can be a relative indicator of Ga incorporations.

B. Strain and atomic diffusions induced disorder

Fig. 2(a) and (b) present RSMs, carried out in an asymmetric configuration with a high incident angle aiming at the Ge(:Ga) (115) and GaAs (115) atomic planes, of samples A and C, respectively. It is seen that the diffraction peak of the Ge(:Ga) epilayer and that of the GaAs substrate are separated in Q_x in Fig. 2(a) rather than in Fig. 2(b). This result indicates that onset of train relaxations occurred in sample A but not in sample C. Because of the large layer thickness HRXRD from the cross-section of sample A is possible. In this light, sample cleavage along the (110) atomic planes of the substrate was processed by mechanical force and a typical cross-section micrograph is shown in the inset of Fig. 2(c). The HRXRD (220) 2θ–θ scan, collected with the incident beam along the interface direction, is shown in Fig. 2(c). One sees that the Ge (220) and GaAs (220) diffractions are apparently separated, indicating the difference in their in-plane lattice constants and thus the strain relaxations. A diffraction shoulder at the lower-angle side of GaAs (220), indicated by the arrow, is also observed. This observation indicates an existence of interfacial layer between the Ge epilayer and the GaAs substrate. This interlayer is not distinguishable in the HRXRD scans from the sample surface due to the large film thickness of sample A. Using GaAs (220) as a reference the in-plane lattice constant of the Ge epilayer can be determined, which is about a^∥_Ge = 5.6575 Å. This value is slightly smaller than that of a^⊥_Ge = 5.6584 Å measured from the (004) diffractions (see latter context). The alignment of the diffraction peaks in Fig. 2(b) indicates an absence of any strain relaxation in sample C and thus a^∥_Ge:Ga = a_GaAs. Fig. 2(d) presents the element distributions of Ge, Ga, and As across the Ge:Ga/GaAs interface of sample C collected by carrying out linear-scanning EDX measurement on a cleaved cross-section. The background of Fig. 2(d) is a typical SEM image (2.0 μm in its horizontal dimension) taken from the cleaved cross-section of the Ge:Ga/GaAs heterostructure. Although the interface is indistinct in the SEM image, the elemental analysis provides clear evidence for the presence of atomic interdiffusions. The interdiffusions even smeared out the intentional Ga dopants in terms of the comparisons between Ga and As profiles toward the film surface. This observation and the comparisons between Fig. 1(a) and (c) support the segregation and clustering of Ga on the growing surface that in turn form the dot structures.

Fig. 2 (a) and (b) are reciprocal space mappings aiming at the (115) atomic planes of the Ge/GaAs (001) (sample A) and Ge:Ga/GaAs (001) (sample C) heterostructures, respectively. (c) and (d) are the HRXRD 2θ–θ (220) curve and elemental distribution profiles measured from the cross-sections of samples A and C, respectively. The background of (c) is an optical micrograph recorded from the cross-section of sample A and that of (d) is a SEM image recorded from the cross-section of sample C during the EDX measurement. The arrow in (c) indicates a diffraction shoulder, suggesting an existence of interfacial layer. The straight line in (d) indicates the location that the elemental analyses were processed.

Fig. 3(a) and (b) show the HRXRD 2θ–θ scans of sample A from the film surface aiming at Ge/GaAs (004) and (002) atomic planes, respectively. Likewise, those of sample C are shown in Fig. 3(c) and (d). Taking the GaAs (004) peak as a reference and assuming that the lattice constant of the GaAs substrate is not changed by the MOCVD growth, the lattice constant of the epilayer in the growth direction can be derived from a^⊥_Ge(:Ga) = a_GaAs sin(θ_GaAs)/sin(θ_GaAs − Δθ). Here, θ_GaAs is the Bragg angle of GaAs (004) and Δθ is the angle difference between the diffraction peaks of GaAs (004) and Ge(:Ga) (004). Together with the in-plane lattice constants obtained above, the lattice constant of fully relaxed Ge(:Ga), a_Ge(:Ga), can be obtained via ε^⊥ = (−2C₁₂/C₁₁)ε^∥ with ε^⊥ = (a^⊥_Ge(:Ga) − a_Ge(:Ga))/a_Ge(:Ga) and ε^∥ = (a^∥_Ge(:Ga) − a_Ge(:Ga))/a_Ge(:Ga). It should be noted that the stiffness constants C₁₁ and C₁₂ of Ge:Ga are unavailable and we took those of Ge (i.e., C₁₁ = 12.85 × 10¹² N m⁻² and C₁₂ = 4.83 × 10¹² N m⁻²), instead, for the calculations.¹⁴ The calculation results, together with the in-plane strains, are summarized in Table 1. It is seen that there is not any apparent difference in the full relaxed lattice constants at all of the studied samples A, B and C. In terms of the linear increase in lattice constant of GaGe alloys with the increase in Ga compositions reported by Greiner et al.,^15–17 i.e., every percent of Ga incorporation tends to increase the lattice constant of relaxed GaGe by 0.0004 Å, the incorporation of Ga in our Ge:Ga thin films is smaller than 1%. This value is much smaller than that probed by EDX from the sample surfaces (see above discussion), which also provides indirect evidence from the surface segregation of Ga atoms during the MOCVD growth.

Fig. 3 (a) and (b) are HRXRD 2θ–θ scans of the thick Ge/GaAs (sample A) aiming at the (004) and (002) atomic planes, respectively. Likewise, (c) and (d) are those of the thin Ge:Ga/GaAs (sample C) aiming at (004) and (002), respectively. Please note the vertical axis breaks in (a), (c), and (d). Integrated peak intensity ratios are also presented along with the individual scans for comparisons.

It has been well known that the HRXRD (002) diffraction peak exists in zinc-blende crystal structures (e.g., GaAs) but not in diamond ones (e.g., Ge) due to the average and canceling of atomic scattering factors.¹⁸ However, the comparison in Fig. 3(b) and (d) shows that the (002) diffractions from the epilayer are really missing in the thick Ge film [see Fig. 3(b)] but existing in the thin Ga-doped Ge film [see Fig. 3(d)]. This result indicates a presence of long-range disorder that forced the lattice matrix of Ge:Ga away from the diamond structure. We have shown above that the Ga concentration in the Ge lattice matrix should be smaller than 1%, which is negligible when compared with the atomic diffusions as shown in Fig. 2(d). In this regards, we believe that the long-range disorder of the Ga:Ge thin films is most likely caused by the self-annealing induced atomic diffusions rather than Ga-doping although its hole concentration is as high as 2.87 × 10²⁰ cm⁻³. On the other hand, the film growth is faster than the atom diffusions, leading to a decrease in the alloy-disorder towards the film surface. As a result, it is undetectable in the thick Ge film of sample A [see Fig. 3(b)] although the presence of an interfacial layer is seen in Fig. 2(c) as probed by HRXRD (220) from the cross-section.

C. Raman shift of LO_Ge–Ge and hole concentrations

Fig. 4(a) presents the micro-Raman spectra collected from the surface of samples A, B, C. Because of the large optical absorption coefficient of Ge at the excitation wavelength of 532 nm (the absorption length is <20 nm),¹⁹ Raman scattering of the GaAs substrate is thus impossible from the surface in the backscattering setup. LO_Ge–Ge mode at 288–300 cm⁻¹, together with the second-order optical modes at 550–600 cm⁻¹ as indicated by the arrow in Fig. 4(a), is clearly seen.²⁰ The broadband at 150–200 cm⁻¹ could be attributed to second-order acoustic overtones.²⁰ The vibration mode of Ge–Ga alloy is not detectable in the spectrum, which is different from other Ge alloys, e.g., GeSn and SiGe,^21,22 and most likely due to the similar atomic weight between Ga (69.74) and Ge (72.59). The inset in Fig. 3(a) shows the expanded spectrum of the LO_Ge–Ge mode, which exhibits apparent phonon-softening, broadening, and asymmetry with the increase in Ga incorporations.

Fig. 4 (a) Micro-Raman spectra collected at room temperature, using a 532 nm wavelength laser as the excitation source, from the surface of the Ge(:Ga)/GaAs samples A, B, and C; the inset highlights the Raman peak shifts with the increase in Ga incorporations. (b) Photon-softening of the LO_Ge–Ge as a function of hole concentrations in this work together with those reproduced with permissions from AIP Advances 6, 015114 (2016) and Physical Review B 23, 6592 (1981).

Theoretically, the Raman peak shift of epitaxial semiconductor thin film alloys can be caused by carrier concentrations, alloy-disorder, and lattice strain.^23–25 For Ge thin films, the peak shift of LO_Ge–Ge induced by biaxial strains reads ΔΩ ≅ −kε^∥ and the proportionality factor k is ∼410.53 cm⁻¹.²⁶ This factor k is a bit smaller than that of uniaxial-strain-induced peak shift of bulk Ge (438 cm⁻¹) while the latter is nearly the same as that of Ge nanowires (434 cm⁻¹).^27,28 These comparisons indicate that k = 410.53 cm⁻¹ is a reasonable choice for estimating the strain-induced Raman peak shift. In this light, taking the maximum compressive in-plane strain of the Ge:Ge thin films (i.e., 866 PPM, see Table 1) into account, a phonon-hardening of the LO_Ge–Ge mode of smaller than 0.35 cm⁻¹ is resulted in the Ge:Ge thin films. This amount of phonon-hardening, as compared with the observed phonon-softening in Fig. 4(a), is negligible.

In the case of semiconductors with high density of free carriers, phonon-electron coupling may shift the Raman scattering from optical phonons at the first Brillouin-zone-center.²⁹ For heavy Ga-doped p-type Ge, the hole concentrations induced phonon-softening complicates the analysis of the alloy-disorder effect on the Raman shift of LO_Ge–Ge. This is different from those of SiGe and GeSn alloys, where the alloying does not introduce any changes in the concentration of free carriers. It has been mentioned above that Kabuyanagi et al.¹⁰ have measured the effect of hole accumulations on the phonon-softening of LO_Ge–Ge via bake-gate tuning of p⁻ or p⁺-type Ge channels of field-effect transistors. The effects of dopants, and thus the effects of the alloy-disorder, on the Raman shift of LO_Ge–Ge are completely excluded. Unfortunately, the hole concentrations accumulated by the back-gate tuning were limited to n_h ≤ 10¹⁹ cm⁻³; nevertheless, the phonon-softening of LO_Ge–Ge exhibits a linear relationship with respect to the hole concentrations for both the p⁻ and p⁺-Ge.¹⁰ In this regard, we have fitted the experimental data reported by Kabuyanagi et al.¹⁰ [reproduced with permissions from AIP Advance 6, 015114 (2016)] and extended the linear relationships up to n_h = 3.2 × 10²⁰ cm⁻³, which are shown in Fig. 4(b). Those experimental data, reproduced with permissions from Physical Review B 23, 6592 (1981) reported by Olego et al.,³⁰ and their linear fittings are also presented in Fig. 4(b) for comparisons. In the experiment by Olego et al.,³⁰ bulk single crystals with hole concentrations introduced by Ga-doping were used. One sees in Fig. 4(b) that both the experiments using the 488 nm wavelength excitation show similar phonon-softening behavior, i.e., 4.79 ± 0.44 cm⁻¹ by every increase of 10²⁰ cm⁻³ in hole concentrations regardless of the dopants (i.e., Ga atoms). An increase in the excitation wavelength tends to reduce the phonon-softening due to the dispersion of electron-phonon coupling with the heavy-hole band of Ge.³⁰ To this end, the comparisons in Fig. 4(b) suggest that the high hole concentration, rather than the alloy-disorder (see discussions in the previous section), dominates the phonon-softening detected from the sample surface (with a penetration depth < 20 nm) although the Ga concentrations might be higher than the hole concentrations wherein.

D. Evolutions of Raman scattering and HRXRD across the interface

Fig. 5(a) and (b) present the micro-Raman spectra collected from the cross-sections, created by mechanical force cleavage along the GaAs (110) planes, of samples A and C, respectively. The spectra were artificially shifted along the vertical axis and individually normalized to the highest peak intensity. Because of the limited spatial resolution of the micro-Raman system (with the beam size of ∼900 nm on the sample) and the atomic interdiffusions across the interface [see Fig. 2(d)], it is difficult to accurately locate the interface due to spatial convolutions. Nevertheless, both the GaAs-like transverse optical phonon (TO_GaAs) and the Ge-like LO_Ge–Ge mode are observed in Fig. 5 and their scattering intensities are monotonically and oppositely varying across the interface areas (highlighted by the pattern-bars). The GaAs-like LO mode is forbidden and thus undetectable from the (110) cross-section in the backscattering configuration.

Fig. 5 Micro-Raman spectra collected from the cleaved (110) surface of the (a) thick Ge and (b) thin Ge:Ga films on GaAs (001) substrates. The pattern-bars indicate the interfacial areas of the heterostructures; the dashed lines indicate the Raman shifts of transverse optical (TO) mode and the longitudinal optical (LO) mode of bulk GaAs and Ge, respectively.

The most striking observation in Fig. 5 is the evolution of the Raman peak shift across the interface areas. They are following a same trace for both samples A and C except for the larger peak shifts, with respect to those of pure GaAs and Ge (indicated by the dashed lines), of sample C than those of sample A. When moving from the substrates toward the Ge(:Ga) films (i) the GaAs-like TO mode shifts to lower frequencies (softening) accompanied by a peak broadening; (ii) emerging of the Ge-like LO mode at a frequency lower than that of pure Ge, especially in the Ge:Ge/GaAs sample, and the Ge-like LO mode also shifts to lower frequencies as approaching the interface areas; and (iii) the Ge-like LO mode shifts back to higher frequencies when moving away from the interface areas towards the Ge(:Ga) thin films. These frequency evolutions, as well as the asymmetric broadening to the lower frequencies, of the Raman peaks are very much like those of metastable (A^IIIB^V)_1−x(C^IV₂)_x thin film alloys with x increasing from 0 to 1, for example, see ref. 8 for those of (GaSb)_1−x(Ge₂)_x. The frequency evolution similarly even includes phonon-hardening of the A^IIIB^V-like TO in (A^IIIB^V)_1−x(C^IV₂)_x for x ≥ 0.5,⁸ which corresponds to those observed in Fig. 5 (see the comparisons between the topmost two spectra crossed by the patterned-bars) as moving away from the interface areas towards the Ge(:Ga) films. These observations, together with the elemental distributions across the Ge:Ge/GaAs interfaces detected by EDX [see Fig. 2(d)] as well as the shoulder of HRXRD GaAs (220) diffractions [see Fig. 2(c)], suggest that an interfacial layer, most likely (GaAs)_1−x(Ge₂)_x alloys, is recrystallized by the self-annealing induced atomic diffusions during the MOCVD growth. As a consequence, it is quite clear that the phonon-softening of the GaAs-like TO and the Ge-like LO with respect to those of TO_GaAs and LO_Ge at the interfacial areas is dominated by alloy-disorder while the excess shifts of the Ge-like LO mode of sample C (Ge:Ga/GaAs), as compared with those of sample A (Ge/GaAs), are dominated by the hole concentrations. The latter case is indeed what we have observed in the Raman scattering from the film surfaces as shown in Fig. 4, where the probed region is far away from the disordered interface areas due to the limited laser penetration depth (<20 nm).

HRXRD (220) 2θ–θ scans from the cross-section of sample A [i.e., the thick Ge/GaAs (001) heterostructure], together with their fittings are presented in Fig. 6(a). The measurements were carried out by setting the diffraction beam-planes (formed by the incident and reflected beams) parallel to the interface plane of the heterostructure while moving the sample step-by-step, fronted by the Ge epilayer, into the diffraction beam-planes along their perpendicular direction. The diffraction curves were individually normalized to their highest peak and artificially bottom-up shifted in Fig. 5(a) for comparisons. Due to the much larger beam size than the layer thickness and the limited resolution in the sample-stage movement, a quantitative determination of the layer thickness is not possible. Nevertheless, one sees that when the Ge layer is moved into the diffraction beam-planes only the Ge (220) peak is detected [see the bottom curve in Fig. 5(a)]. Along the with the sample-stage movement, the Ge (220) peak increases in intensity and the peaks from the interfacial layer and the GaAs substrate emerge in sequence. Plotted in Fig. 5(b) are the peak intensity ratios of I_Interlayer/I_Ge(220) and I_GaAs(220)/I_Ge(220) as a function of the sample movement. The intensity evolutions show that the curves in Fig. 5(a) were collected at the initial steps of sample movements; neither the interlayer nor the substrate completely entered the diffraction beam-planes. In fact, when the sample is fully covered by the beam-planes the increases in the intensity ratios of I_Interlayer/I_Ge(220) and I_GaAs(220)/I_Ge(220) are saturated at 5.1 and 31.0, respectively [see the 2θ–θ (220) curve and its fittings shown in the inset of Fig. 5(b)]. The saturation of the increase in I_Interlayer/I_Ge(220) occurs much earlier than that of I_GaAs(220)/I_Ge(220) (curves not shown for the sake of brevity). These observations provide clear cut evidence for the existence of an interfacial layer in the Ge/GaAs heterostructure of sample A although its growth temperature is a bit lower than that of the Ge:Ga/GaAs sample C.

Fig. 6 (a) HRXRD 2θ–θ (220) scans, and their fittings, collected from the cleaved (110) surface of the Ge/GaAs heterostructure with the interface plane parallel to the diffraction beam-planes (formed by the incident and reflected beams). The scans, artificially shifted in the vertical direction, from bottom up were collected by moving the sample, fronted by the Ge epilayer, step-by-step into the diffraction beam-planes. (b) Integrated intensity ratios of I_Interlayer/I_Ge(220) and I_GaAs(220)/I_Ge(220) derived from the fittings in (a), they are saturated at 5.1 and 31.0, respectively. The scan, and its fittings, in the inset of (b) is measured after a certain sample movement, in which both the increases in I_Interlayer/I_Ge(220) and I_GaAs(220)/I_Ge(220) were saturated.

IV. Conclusion

In conclusion, micro-Raman scattering and HRXRD have been employed to compare the lattice vibrational and structural properties of a thick Ge film and thin Ga-doped Ge films grown by MOCVD on GaAs (001) substrates. The measurements have been carried out in both surface and cross-section configurations for the heterostructures. Reciprocal space mapping of the (115) atomic planes in an asymmetric configuration revealed that the strain relaxations occurred in the thick Ge film but not in the thin Ge:Ga films. Threading dislocations, generated due to the strain relaxations, gave rise to inverted-square-pyramid-like pits on the film surface with a density of ∼2.5 × 10⁶ cm⁻². Although there is no strain relaxation in the Ga-doped Ge thin films, the surface segregation of Ga atoms resulted in the Ga-rich droplets. The surface segregation of Ga atoms also caused an overestimation of Ga incorporation probed by EDX from the surface of the samples. The maximum compressive strain built in the studied Ge:Ga thin films is about 866 PPM, which tends to cause a phonon-hardening of about 0.35 cm⁻¹ to the LO mode of the Ge–Ge lattice vibrations. Such amount of Raman peak shift is negligible as compared with the phonon-softening of up to 11.63 cm⁻¹ induced by the Ga-doping (probed from the film surface). By comparing with the linear relationships derived and extended from the Raman peak shifts as function of hole concentrations available in the literature, we believe that the phonon-softening probed from the film surface is dominated by the heavy hole concentrations, up to 2.87 × 10²⁰ cm⁻³, regardless of the surface segregation of Ga atoms. On the other hand, GaAs-like TO and Ge-like LO phonons have been observed from the [110]-oriented cross-sections of the Ge(:Ge)/GaAs (001) heterostructures. Their intensity and peak-shift evolutions, as well as the elemental distributions, across the interfaces suggest an existence of an interlayer in between the Ge(:Ga) epilayer and the GaAs substrate. This interlayer is mainly caused by self-annealing induced atomic diffusions and solid-state reactions, and it consists of, most likely, (GaAs)_1−x(Ge₂)_x alloys. This interlayer, undetectable in the Ge/GaAs structure from the film surface direction due to the thick Ge epilayer, induced an apparent long-range lattice disordering in the thin Ge:Ga film, which manifested itself as the emergence of the HRXRD (002) diffraction peak. Finally, HRXRD from the cross-section of the Ge/GaAs (001) heterostructure, together with the evolutions of diffraction peak intensities as a function of sample movements in the direction perpendicular to the diffraction beam-planes, provides clear cut evidence for the presence of the interlayer.

References

1. J. Yang, Z. Peng, D. Cheong and R. Kleiman, Fabrication of high-efficiency III-V on silicon multijunction solar cells by direct metal interconnect, IEEE Journal of Photovoltaics, 2014, 4, 1149.
2. E. H. Steenbergen, M. J. DiNezza, W. H. G. Dettlaff, S. H. Lim and Y.-H. Zhang, Optically-addressed two-terminal multicolor photodetector, Appl. Phys. Lett., 2010, 97, 161111.
3. J. A. Carlin, S. A. Ringel, E. A. Fitzgerald, M. Bulsara and B. M. Keyes, Impact of GaAs buffer thickness on electronic quality of GaAs grown on graded Ge/GeSi/Si substrates, Appl. Phys. Lett., 2000, 76, 1884.
4. C. K. Chia, J. R. Dong, D. Z. Chi, A. Sridhara, A. S. W. Wong, M. Suryana, G. K. Dalapati, S. J. Chua and S. J. Lee, Effects of AlAs interfacial layer on material and optical properties of GaA/Ge(100) epitaxy, Appl. Phys. Lett., 2008, 92, 141905.
5. M. K. Hudait and S. B. Krupanidhi, Self-annihilation of antiphase boundaries in GaAs epilayers on Ge substrates grown by metal-organic vapor-phase epitaxy, J. Appl. Phys., 2001, 89, 5972.
6. H. F. Liu, A. S. W. Wong, G. X. Hu and H. Gong, Observation of interfacial reactions and recrystallization of extrinsic phases in epitaxial grown ZnO/GaAs heterostructures, J. Cryst. Growth, 2008, 310, 4305.
7. F. A. Trumbore, Solid solubilities of impurity elements in germanium and silicon, Bell Syst. Tech. J., 1960, 39, 205.
8. T. C. McGlinn, M. V. Klein, L. T. Romano and J. E. Greene, Raman-scattering and electron-microscopy study of composition-dependent ordering in metastable (A^IIIB^V)_1−x(C^IV₂)_x alloys, Phys. Rev. B: Condens. Matter Mater. Phys., 1988, 38, 3362.
9. A. Sanson, E. Napolitani, M. Giarola, G. Impellizzeri, V. Privitera, G. Mariotto and A. Carnera, Non-conventional characterization of electrically active dopant profiles in Al-implanted Ge by depth-resolved micro-Raman spectroscopy, Appl. Phys. Express, 2013, 6, 042404.
10. S. Kabuyanagi, T. Nishimura, T. Yajima and A. Toriumi, Experimental evidence of zone-center optical phonon softening by accumulating holes in thin Ge, AIP Adv., 2016, 6, 015114.
11. Y. J. Jin, C. K. Chia, H. F. Liu, L. M. Wong, J. W. Chai, D. Z. Chi and S. J. Wang, P-type Ge epitaxy on GaAs (100) substrate grown by MOCVD, Appl. Surf. Sci., 2016, 376, 236.
12. H. F Liu, Y. J. Jin and C. Y. Yang, Droplet-induced dot, dot-in-hole, and hole structures in GaGe thin films grown by MOCVD on GaAs substrates, CrystEngComm, 2016 10.1039/c6ce00778c.
13. H. F. Liu, W. Liu, A. M. Yong, X. H. Zhang, S. J. Chua and D. Z. Chi, Effects of annealing on structural and optical properties of InGaN/GaN multiple quantum wells at emission wavelength of 490 nm, J. Appl. Phys., 2011, 110, 063505.
14. C. Kittel, Introduction to Solid State Physics, Wiley, 7th edn, 1996.
15. E. S. Greiner, Constitution of germanium-rich germanium-gallium alloys, J. Met., 1952, 4, 1044.
16. E. S. Greiner and P. Breidt Jr, Melting point of germanium and the continuation of some Ge-Ga alloys, Trans. Metall. Soc. AIME, 1955, 203, 187.
17. R. W. Olesinski and G. J. Abbaschian, The Ga-Ge (gallium-germanium) system, Bull. Alloy Phase Diagrams, 1985, 6, 258.
18. S. I. Shah, B. Kramer, S. A. Barnett and J. E. Greene, Direct evidence for an order/disorder phase transition at x ≃ 0.3 in single-crystal metastable (GaSb)_(1−x)(Ge₂)_x alloys: High-resolution X-ray diffraction measurements, J. Appl. Phys., 1986, 59, 1482.
19. D. E. Aspnes and A. A. Studna, Dielectric functions and optical parameters of Si, Ge, GaP, GaAs, GaSb, InP, InAs, and InSb from 1.5 to 6.0 eV, Phys. Rev. B: Condens. Matter Mater. Phys., 1983, 27, 985.
20. H. D. Fuchs, C. H. Grein, C. Thomsen, M. Cardona, W. L. Hansen, E. E. Haller and K. Itoh, Comparison of the phonon spectra of ⁷⁰Ge and natural Ge crystals: Effects of isotopic disorder, Phys. Rev. B: Condens. Matter Mater. Phys., 1991, 43, 4835.
21. A. Mosleh, M. A. Alher, L. C. Cousar, W. Du, S. A. Ghetmiri, T. Pham, J. M. Grant, G. Sun, R. A. Soref, B. Li, H. A. Naseem and S. Q. Yu, Direct growth of Ge_1−xSn_x films on Si using a cold-wall ultra-high vacuum chemical-vapor-deposition system, Front. Mater. Sci., 2015, 2, 30.
22. H. Brugger and G. Abstreiter, Si-Ge strained layer superlattices, J. Phys., Colloq., 1987, 48, C5–C321.
23. R. Cheng, W. Wang, X. Gong, L. Sun, P. Guo, H. Hu, Z. Shen, G. Han and Y.-C. Yeo, Relaxed and strained patterned germanium-tin structures: a Raman scattering study, ECS J. Solid State Sci. Technol., 2013, 2, P138.
24. H. F. Liu, N. Xiang, S. Tripathy and S. J. Chua, Raman scattering probe of anharmonic effects due to temperature and compositional disorder in GaN_xAs_1−x, J. Appl. Phys., 2006, 99, 103503.
25. H. Chen, Y. K. Li, C. S. Peng, H. F. Liu, Y. L. Liu, Q. Huang and J. M. Zhou, Crosshatching on a SiGe film grown on a Si (001) substrate studied by Raman mapping and atomic force microscopy, Phys. Rev. B: Condens. Matter Mater. Phys., 2002, 65, 233303.
26. D. Nam, D. Sukhdeo, A. Roy, K. Balram, S. Cheng, K. C. H. Z. Yuan, M. Brongersma, Y. Nishi, D. Miller and K. Saraswat, Strained germanium thin film membrane on silicon substrate for optoelectronics, Opt. Express, 2011, 19, 25866.
27. F. Cerdeira, C. J. Buchenauer, F. H. Pollak and M. Cardona, Stress-induced shifts of first-order Raman frequencies of diamond- and zinc-blend-type semiconductors, Phys. Rev. B: Solid State, 1972, 5, 580.
28. J. Greil, A. Lugstein, C. Zeiner, G. Strasser and E. Bertagnolli, Tuning the electro-optical properties of germanium nanowires by tensile strain, Nano Lett., 2012, 12, 6230.
29. H. F. Liu, W. Liu, S. B. Dolmanan, S. Tripathy, S. J. Chua and D. Z. Chi, Effects of doping and indium inclusions on the structural and optical properties of InN thin films grown by MOCVD, J. Phys. D: Appl. Phys., 2012, 45, 195102.
30. D. Olego and M. Cardona, Self-energy effects of the optical phonons of heavily doped p-GaAs and p-Ge, Phys. Rev. B: Condens. Matter Mater. Phys., 1981, 23, 6592.

---