Abstract

A series of ultrathin InSb films grown on GaAs by low-pressure metalorganic chemical vapor deposition with different V/III ratios were investigated thoroughly using spectroscopic ellipsometry (SE), X-ray diffraction, and synchrotron radiation X-ray absorption spectroscopy. The results predicted that InSb films on GaAs grown under too high or too low V/III ratios are with poor quality, while those grown with proper V/III ratios of 4.20–4.78 possess the high crystalline quality. The temperature-dependent SE (20–300°C) and simulation showed smooth variations of SE spectra, optical constants (n, k, e1, and ε2), and critical energy points (, , , , and ) for InSb film when temperature increased from 20°C to 250°C, while at 300°C, large changes appeared. Our study revealed the oxidation of about two atomic layers and the formation of an indium-oxide (InO) layer of ∼5.4 nm. This indicates the high temperature limitation for the use of InSb/GaAs materials, up to 250°C.

1. Introduction

As a member of the III-V compound semiconductors family, the growth of ultrathin films of indium antimonide has attracted a great deal of attention for its use in midwavelength infrared detectors (viz., thermal imaging cameras and forward looking infrared systems), magnetic sensors, magnetoresistors, field-effect transistors, photoconductors, and high-speed electronic devices [133]. Among other semiconductors, the intrinsic binary InSb possesses the highest electron mobility (∼7.8 × 104 cm2/V·s), higher breakdown field (∼103 V/cm), high saturated electron drift velocity (∼5 × 107 cm/s), small effective mass ( ∼0.013 mo), lowest energy band gap E ∼0.18 eV (at 300 K), and ballistic length up to ∼0.7 μm (at 300 K). The InSb material can be used as a lattice-matched substrate for epitaxial growth of CdTe [34] and other relevant heterostructures and superlattices [35]. The InSb-based ternary as well as quaternary alloys are equally valuable for realizing midwavelength infrared detectors [3639]. More recently, there has been a growing challenge to explore the possibility of developing InSb-based quantum dots and nanostructured laser diodes [4042] for optoelectronic as well as photonic device applications in the 3–5 μm wavelength range for room temperature operations.

In optoelectronic devices, one must grow ultrathin layers of InSb on semi-insulating infrared transparent substrates to prevent current leakage. While CdTe is the only semi-insulating lattice-matched substrate available for InSb growth, it is difficult, however, to avoid the formation of the In2Te3 precipitates at the InSb/CdTe interface [43, 44]. Hence, many alternative materials (viz., Si, GaAs, InP, sapphire, and mica) have been chosen as substrates for preparing InSb epifilms [2, 1517, 1932] by molecular beam epitaxy, liquid-phase epitaxy [18, 30], metalorganic chemical vapor deposition (MOCVD) [2022, 31], metalorganic magnetron sputtering [32, 45], and two-step growth process [46] methods. Despite a large (14.6%) lattice mismatch between InSb and GaAs, the semi-insulating GaAs is considered as an attractive substrate due to high chemical stability and resistivity. Efforts have been made to grow InSb thin films on four-inch GaAs substrates [31, 32, 47]. Also many attempts made by MOCVD have shown improvements in InSb film growth on large area semi-insulating GaAs substrates [4749]. These efforts have certainly helped provided the increasing mass production needs of isolated InSb/GaAs (001) thin films for integration with III-V semiconductor platform to achieve better performance of both the electronic and photonic devices.

However, the incompatibility of the lattice constants between the two materials strongly limits the quality of InSb epilayers grown on GaAs [23]. In general, the epifilms suffer from high density of dislocations or defects near the film/substrate interface. These defects propagating throughout the entire material create antiphase domains and cause autodoping effects. The layers with intrinsic defects may also affect the carrier mobility and leakage current in both the electronic and photonic devices. To circumvent these difficulties and improve the crystallinity of InSb films, several strategies are pursued—especially in optimizing the growth conditions by controlling the V/III ratios, pressure, growth temperature, growth rate, and film thickness [2022, 31, 47, 48].

Accurate estimation of InSb epifilm thickness is of paramount importance using them for device engineering. Traditionally, the thin semiconducting films are characterized by using Hall measurements, XRD, ultraviolet-visible spectrophotometry, cross-sectional scanning electron microscopy [20, 21, 23, 49, 50], and so on. Being destructive, most of these methods are not convenient for assessing material quality required in electronic industry for device productions. Due to the narrow band gap of InSb, no reflectance interference fringes appear in the ultraviolet-visible wavelength range from the InSb/GaAs heterostructure. Therefore, a variety of nondestructive and penetrative tools have been exploited, providing nanoscale resolution for evaluating film thickness with greater degree of accuracy. Some of these techniques used in characterizing large area wafers include the atomic force microscopy, energy-dispersive X-ray spectroscopy, secondary-ion-mass spectrometry, and SE. RSS is another valuable and nondestructive tool to offer useful information on the crystalline quality and other parameters necessary for optimizing the InSb epifilm growth [31].

In this study, we present the results of our comprehensive investigations using SE, XRD, and SR-XAS, respectively, to assess the InSb film thickness and report the effects of V/III source ratios on the films crystalline quality for optimizing the MOCVD growth parameters, which were not exactly achieved from previous reports [31]. The SR-XAS was used to acquire the structural properties of the material at the atomic scale. Furthermore, temperature-dependent optical constants of InSb thin films between 25°C and 300°C (n, k, and ε) were probed by exploiting SE. This work provides a helpful guide to thin film characterization procedures required to monitor the growth processes, understanding the chemical and physical properties of materials and guiding the designs of high-performance InSb thin film devices.

2. Experimental

2.1. Material Growth

The growth of ultrathin uniform InSb thin epilayers was carried out on 4-inch semi-insulating (SI) GaAs (100) substrates using a low-pressure MOCVD method in vertical configuration by exploiting a high-speed rotating disk (180 mm diameter) reactor [31, 47]. The substrate orientation of 4-inch SI GaAs (100) was 2–4° off towards <110>. The trimethyl indium and trimethyl antimony sources were used as vapor-phase constituents of In and Sb into a reaction chamber at approximately ∼20°C and with the bubbler pressures set at 400 Torr for trimethyl indium and 323 Torr for trimethyl antimony, respectively. Hydrogen acted as carrier gas, and the pyrometer growth temperature was 395°C. In this paper, we have studied a series of five InSb/GaAs samples grown with different V/III source flow ratios and thickness. Table 1 summarizes the growth relevant parameters of the samples used from the MOCVD growth [31] and their thicknesses determined in this study. For InSb growth using trimethyl indium and trimethyl antimony, the surface morphology was found to be very sensitive to the V/III ratio and closely reflected the crystallinity of as-grown films. All InSb films with V/III ratio varied between 3.91 and 5.38 exhibit mirror-like surface morphology having thickness ranging from 34 to 64 nm. To examine the effects of V/III ratio on the InSb film characteristics, we named the five samples S1 to S5 in the order of V/III ratio corresponding to their growth run numbers (IA376, 380, 373, 370, and 371, resp. [31]).

2.2. Optical Measurements
2.2.1. Spectroscopic Ellipsometry

SE is a valuable nondestructive technique commonly used for determining thickness, optical constants of materials; the method has also been employed for a complete depth profiling in semiconducting epitaxially grown ultrathin or thin films. While the conventional SE approach suffers from the drawback of slow data acquisition process and covers a limited spectral range, the phase-modulated method that we have employed here offers the fast and precise data acquisition over a large wavelength range. A dual rotating-compensator Mueller matrix ellipsometer (ME-L ellipsometer, Wuhan Eoptics Technology Co. Ltd., China), equipped with Linkam Scientific heating and cooling stage device (THMSG600E), was used for the acquisition of ellipsometric spectra on ultrathin InSb epifilms. At room temperature, we acquired (Figure 1) two parameters Ψ and ∆, represented the ratio of amplitude decay between and p and s polarization of the reflected light and their phase difference, respectively, as a function of wavelength (λ) from 200 nm to 1670 nm at 1 nm step (or energy from 0.74 eV to 6.2 eV) at three angles of incidence 55°, 60°, and 65°.

2.2.2. Raman Scattering Spectroscopy

RSS is a powerful and nondestructive technique for providing valuable information on materials’ characteristics—especially for assessing epilayer thickness, strain, disorder, and site selectivity of defects. The RSS method is particularly suited for probing the local atomic and/or nanoscale structural changes in the InSb ultrathin films grown on SI GaAs while making the careful analysis of its subtle spectral variations. At the earlier time, we had quickly measured RSS on our InSb samples using a Raman spectrometer with the excitation from a HeNe laser 633 nm line. Later on, after more years, we performed further Raman measurements using 514-nm laser excitation on these samples, with similar results obtained. To avoid duplication, these data were not presented here.

2.2.3. X-Ray Diffraction Spectroscopy

As a common technique for material characterization, the XRD is widely used to evaluate the quality of crystal structure. It is sensitive for estimating stress/strains in epitaxially grown thin films. Furthermore, the peak position and the full width at half maximum (FWHM) of X-ray diffraction spectra give the information of crystal orientation and crystal quality. The diffractometer we used in experiments is Rigaku MiniFlex 600, Japan. The five InSb samples were measured by this X-ray diffractometer using a Cu Kα radiation (λ = 1.5406 Å).

2.3. Synchrotron Radiation X-Ray Absorption Spectroscopy

XAFS spectra were collected for the InSb thin film samples in X-ray fluorescence yield mode at the beam line 01C1 of the National Synchrotron Radiation Research Center in Hsinchu, Taiwan. The photon energy for the XAFS covered the range from 27,640 to 28,840 eV. The intensity of X-ray was monitored by a liquid N2–filled ionization chamber, and fluorescence emitted from the sample was measured by an argon-filled Stern-Held-Lytle detector. A Si (111) double-crystal monochromater with a 0.5-mm entrance slit was used. A filter was inserted between the sample and the detector window to reduce the noise from scattering and to improve the spectrum quality. The incident photon direction was 45 degree to surface of the sample, and the flux I0 of incident photon was monitored simultaneously by an ion chamber located just before the sample chamber. All measurements were made at RT, and all X-ray absorption spectra were normalized to I0.

3. Results and Discussion

3.1. Spectroscopic Ellipsometry at Room Temperature

The measured SE spectra are fitted empirically by minimizing the squared differences between the observed and calculated Ψ and Δ values, which generated from the fitting model at the corresponding wavelengths. The quality of the fit can be judged by the mean square error (MSE) defined as follows:where stands for the total number of experimental measurements of and for each chosen wavelength, represents the number of fitting parameters used, the superscripts “mod” and “exp” represent the appropriate values of model and experimental data, and signifies the MSE between the calculated and experiment data at each wavelength.

In this study, the SE data were fitted using Tauc–Lorentz multiple oscillator modes [48]:where E is photon transition energy, A and C are fitting constants, and estimates are obtained by fitting Ψ and . Equation (2) is useful for evaluating the dielectric function (or n, k). The real part of the dielectric function is obtained by exploiting the Kramers–Kronig integrations, that is,

In (3), the term P is the main part of the Cauchy integral, where is added as a fitting parameter.

Figure 1 shows experimental SE Ψ and ∆ spectra of five InSb/GaAs samples, measured with three incident angels at room temperature (RT). The spectral oscillations of the five samples are similar, indicating that they have similar optical properties. All of the Ψ and ∆ data with variable angles were considered in the calculation of the optical parameters like n and k. The thicknesses of InSb films were extracted through analyzing the SE data. The three-layer model, substrate/film/oxidation, or surface roughness was first tested for modeling. The simulation results showed the surface oxidation layer with the thickness of 0.01 nm or less for all sample. This is physically meaningless, that is, the thickness of the surface oxidation layer should be zero. Therefore, the two-layer model, substrate/film, was used for modeling on five InSb/GaAs samples here. Fitting results are listed in Table 1.

3.2. The X-Ray Diffraction Data and Fits

Figure 2 shows the wide-scan XRD spectrum of a typical InSb/GaAs sample (S2) in the angle range (2θ) from 55° to 69° at 0.1° step. The intensity of the GaAs (400) peak was observed to be weaker than that of the InSb (400) peak. This phenomenon was due to that the semi-insulating GaAs wafer cut 2–4° off (100) towards <110> was used as the substrate. The surface of GaAs wafer was deviated from the real (100) orientation. The GaAs wafer was not the real GaAs (100) substrate or not with the real (100) orientation surface. This type of semi-insulating GaAs (100) 2° off towards <110> substrate was successfully used for the growth of InSb thin films [47, 49]. During the growth process, it could lead the heteromismatch-induced dislocations to spread along the InSb/GaAs interface plane but not along the film growth normal direction, that is, not to the InSb film surface, so as to greatly decrease the dislocation density at the surface area of InSb film.

From this rough scan, the InSb (400) peak appeared as a single peak with the FWHM of 0.18°. Inset of Figure 2 shows fine-scan XRD spectrum in the angle (2θ) of 56°–58° at 0.005° step. The observation of the peak doublet splitting is caused due to the X-ray source Cu Kα1 and Cu Kα2 radiation. The well-separation of the InSb (400) peak doublet and the narrow FWHM of 0.07° from the InSb (400) Kα1 peak indicated good crystallinity of the film. A sharp diffraction peak was observed at 2θ = 56.88°, which corresponds to the (400) crystalline plane of InSb (S2). The other diffraction peak located at 66.20° can indexed to the (400) crystal plane of GaAs substrate. Table 2 lists the values of XRD peak and FWHM obtained by Lorentz fitting, for five InSb samples with different V/III ratios.

Farag et al. [51] calculated the crystallite size (D) of the InSb film grown on GaAs from the broadening of few XRD peaks using the Debye-Scherrer equation (4):where is the X-ray wavelength of Cu Kα (0.15418 nm), is the width of the peak at half maximum intensity for the thin film, is the Scherrer constant of the order of unity (0.95 for powder and 0.89 for film), and is the corresponding Bragg’s angle.

Usually, finite size gives broader peaks. The contribution of our InSb film thickness, , to the XRD broad peak width, , can be estimated by modified Debye-Scherrer formula (5),

The so-calculated values using the InSb film thickness obtained from SE are also listed in Table 2. It is seen that these values are much larger than the XRD FWHMs measured in our experiments. This predicts the high crystalline quality of our MOCVD grown InSb films on GaAs.

3.3. Simulations of SR-XAS Data

Indium K-edge extended X-ray absorption fine structure (EXAFS) was employed to study the local structure of InSb thin films on GaAs. Figure 3 plots the four In K-edges EXAFS of InSb, with all of In absorption edges near 27,940 eV. Replicate EXAFS scans were coadded to improve the signal-to-noise ratio. Due to numerous diffraction peaks in the In-EXAFS spectra, the samples were spun to average out the diffraction peaks in the EXAFS spectra. All EXAFS spectra were collected to 1,200 eV beyond the In K-edges. All EXAFS data were analyzed for wave vectors (k) from 3.0 Å−1 to 11.5 Å−1. The chi data were k2χ (k) weighted as shown in Figure 4 and Fourier transformed with a window width of k = 0.5 Å−1 to yield the R-space data. As shown in Figure 5, the simulated EXAFS spectra were generated, based on the documented crystallographic properties for In and Sb using ab initio-based theory.

We used ATHENA program to remove the background and extract the EXAFS oscillations from the k-space signals. To extract the bond length of RIn-Sb, a structural model was built using the package IFEFFIT. Fitting results of the Fourier transforms show a good agreement with the measurement shown in Figure 5. With the first neighbor peak being well defined in the In absorption, we extracted the from the In data listed in Table 3.

Fitting results show that the bond length and coordination number of the first In-Sb shell are about 2.80 Å and 4, respectively. The bond length and the coordination number of unstrained InSb are 2.80 Å and 4, respectively. The EXAFS fitting for five InSb/GaAs gave bond length values around 2.80 Å, and the sample S5 with the highest V/III ratio of 5.38 has its slightly less than 2.80 Å. The coordination number (CN) for all samples is close to but slightly below 4 (the regular value) with samples S4 and S5 nearest to 3.96–3.98, samples S2 and S3 at 3.92, and only sample S1 with the lowest V/III ratio of 3.91 significantly lower at 3.80. These variations could be caused by residual strain and too low/too high might cause higher lattice mismatch in the InSb/GaAs heterointerface, leading to higher residual strain within the epitaxial InSb film. This finding suggests that V/III ratios of 4.20 (S3) and 4.78 (S4) are the better growth parameter values, which is in consistent with the results from Raman scattering.

3.4. Further Analyses and Temperature-Dependent SE

The real part and imaginary part of the dielectric function of the bulk InSb and bulk GaAs as function of photon energy is shown in Figure 6. The peaks in these spectra reflect the high-energy-state transitions of bulk InSb and can be used to obtain several related critical energy points (, , , , and ) [52]. The high-energy-state transitions of bulk GaAs are , , , and [52]. Therefore, Figure 6 shows the peaks those are the high-energy-state transitions of the InSb film and none of GaAs substrate. D’Costa et al. [15] reported the SE study at room temperature for an InSb/GaAs grown by molecular beam epitaxy, showing similar critical energy points from InSb.

From the simulation of the SE Ψ and Δ data in Figure 1, the reflective index (n) and extinction coefficient (k) versus wavelength or energy can be extracted. Figures 7(a) and 7(b) exhibit the comparative n and k variation with energy (eV), for five InSb/GaAs, respectively. The critical energy points (, , , , and ) are marked as discussed in last paragraph.

Temperature-dependent (TD) SE measurements were performed in the temperature range from 20°C to 300°C. Figure 8 presents a set of TD-SE experimental data for sample S2, which were measured at three different angles, 55°, 60°, and 65°. Simulation fits were performed for all SE Ψ and Δ data. As mentioned in Section 3.1 for RT-SE, the three-layer model—substrate/film/surface oxide—was first used for modeling. The best fitting simulation results showed the surface oxide layer with the thickness of 0.01 nm or less, that is, zero, for all SE data measured from 20°C to 250°C. However, for the SE spectra (Ψ and Δ) data measured at 300°C, the three-layer model—substrate/film/surface oxide—has to be used, resulting in a surface oxide thickness of 5.4 nm for sample S2. Other samples had similar results, that is, SE data between RT and 250°C showed no surface oxide layer but SE at 300°C revealed a near 5-nm surface oxide layer, corresponding to about two atomic layer of indium oxides.

Figure 9 shows the variation of refractive index n and extinction coefficient k of InSb films in a temperature range from 20°C to 300°C. As seen, the InSb film contains several absorption peaks in the wavelength range from 200 to 1500 nm. Some absorption peaks shift to the longer wavelength (redshift) with increasing temperature. Kim et al. [53] performed TD-SE on bulk InSb over a temperature of 31 K to 675 K. A similar redshift was observed in our experimental InSb films on GaAs over the temperature range of 20°C to 300°C, that is, 293 K to 573 K, comparable with that observed in other study [53].

Figure 10 exhibits the complex dielectric function, real and imaging parts, and , calculated by and , of InSb film S2 with different temperatures. The high-energy critical points of , , , , and are indicated in the figure. The gradual redshifts of these critical energy peaks with temperature increasing from 20°C to 250°C are seen. However, these redshifts of critical energy peaks in Figure 10 and n and k spectra in Figure 9 become much more pronounced as temperature increases to 300°C. These large variations in n and k and and are due to the formation of an InO surface layer from the severe oxidation of the surface of InSb film at 300°C. This oxidation could be ignored for InSb film measured in 20°C–250°C because the SE fits with three-layer model leaded to an InO surface thickness of <0.01 nm, much thinner than one monoatomic layer. But the oxide layer cannot be ignored in sample heated at 300°C, that is, higher than 250°C, the severe oxidation appeared on InSb, leading to the formation of a thin indium-oxide layer. This result may predicate the high temperature limitation for the use of InSb/GaAs materials.

4. Conclusions

In summary, a series of InSb thin films grown on 4-inch GaAs substrates by MOCVD technique, with different V/III ratios, were investigated by SE, XRD, and SR-XAS, respectively. InSb thin film thickness was extracted by fitting the experimental SE data. The crystallinity of the films was high quality extracted from the well separation of the InSb (400) peak doublet and the narrow FWHM of the InSb (400) Kα1 peak in the XRD spectrum. Through advanced synchrotron radiation technique of XAS and data simulation, the atomic scale bonding length and coordination number were obtained.

Through combined multiple technological analyses, the results showed that InSb films on GaAs grown under too high or too low V/III ratios are with poor quality, while those grown with proper V/III ratios of 4.20 and 4.78 possess the high crystalline quality. These results are useful to the material growers for improving the growth processing.

The temperature-dependent (TD)-SE measurements (20–300°C) and simulation revealed a significant phenomenon: the SE spectra, optical constants (n, k, ε1, ε2), and critical energy points (, , , , and ) of the InSb thin films varied with temperature (T) gradually and smoothly as T increased from 20°C to 250°C, and all SE spectra can be fitted well by the two-layer model (substrate/film). However, up to 300°C, the SE spectra appeared to show large changes, and simulation revealed the existence of an indium-oxide (InO) layer of ∼5.4 nm, that is, about two atomic layers. The optical constants (n, k, ε1, ε2) and critical energy points (, , , , and ) had showed transilient changes from 250°C to 300°C, which are due to the top about two atomic layers oxidized. This indicates the high temperature limitation for the use of InSb/GaAs materials, up to 250°C, which provides a hinder to the device designers using InSb materials.

Data Availability

All types of data used to support the findings of this study are included within the article.

Conflicts of Interest

The authors declare that there are no conflicts of interest regarding the publication of this paper.

Acknowledgments

This work was supported by National Natural Science Foundation of China (nos. 61367004 and 61504030) and special funding (nos. T3120097921 and T3120099202) for Guangxi distinguished professors (Bagui Rencai and Bagui Xuezhe). The authors acknowledge the support and help from Prof. Enwei Liang, Prof. Gu Xu, Prof. Jingli Chen, Ms. Qingxuan Li, Ms. Fangze Wang, Prof. Shiyuan Liu, and Dr. Chuanwei Zhang.