GaAs_1−yBi_y Raman signatures: illuminating relationships between the electrical and optical properties of GaAs_1−yBi_y and Bi incorporation

The dilute III-V ternary compound semiconductor, GaAs_1−yBi_y, is a promising materials system for a range of device applications. The incorporation of bismuth into GaAs induces a strong bandgap reduction and considerably enhances the spin-orbit splitting energy.^1–3 These features make GaAs_1−yBi_y a good candidate for long wavelength optoelectronic devices and spintronic devices. Research in the synthesis and properties of GaAs_1−yBi_y has been primarily focusing on determining relationships between epitaxial growth conditions and Bi incorporation.^3–5 In general, incorporating Bi is very challenging due to the large differences in the Group V atomic radii (Bi is 31% larger than As) and electronegativities (Bi is 9% smaller than As).^3,6 Recent studies indicate that Bi forms acceptor levels⁷ introducing holes in nominally undoped GaAs_1−yBi_y.⁸ 

Herein we report on two GaAs_1−yBi_y Raman signatures which can be used to determine the Bi composition (y) and predict 300K photoluminescence intensity. These signatures are based only on GaAs vibrational modes, as opposed to the relatively low intensity GaBi modes.⁹ Recently, J. A. Steele et al.,¹⁰ showed that the relative intensity of the Bi-induced longitudinal-optical-plasmon-coupled (LOPC) mode to the GaAs longitudinal-optical (LO) phonon (Γ) is positively related to hole concentration, as shown earlier in numerous earlier studies of doped GaAs.^11–13 The LOPC/LO ratio can therefore be used to study the acceptor nature of Bi. Herein, we report on a new observed relationship between 300K photoluminescence (PL) intensity, or emission efficiency, and the hole concentration as determined using the LOPC/LO ratio. Furthermore, we show that the Raman signature characterizing Fröhlich scattering, a process dominated by impurity-induced scattering, can be used to determine the Bi incorporation efficiency.^14,15

In this study, a Riber 2300 molecular beam epitaxy (MBE) system is used to synthesize GaAs_1−yBi_y thin films on GaAs (001) substrates. The thicknesses of the GaAs_1−yBi_y films are all approximately 250 nm. A detailed description of growth conditions and procedure can be found elsewhere.¹⁶ Two sets of samples were prepared and are described below.

Set A is comprised of data collected across a single sample grown on 1/4 of a two-inch substrate at 320°C and growth rate of ∼0.5μm/h. Normally, GaAs_1−yBi_y samples are rotated during MBE growth in order to produce compositionally-uniform films since the beam fluxes vary spatially across the substrate.³ However, one drawback resulting from rotation is the difficulty in finely controlling the V/III ratio, to which the Bi incorporation is very sensitive.⁴ In order to study the impact of the V/III ratio on Bi incorporation, we chose not to rotate the sample during growth in order to create flux gradients as shown in Fig. 1. The dotted line across the wafer indicates the area of interest for all of the measured properties. The parameter x (mm) describes the position on the line. Near-stoichiometric flux conditions are roughly achieved near the center of the dotted line with an error of ∼2 mm. Therefore, the left end of the line delineates the portion of the sample growth under As-rich conditions, while the right end was grown with Ga-rich conditions.

The x=0 point is defined as the position at which near-stoichiometric , i.e. As/Ga absolute flux ratio close to 1:1, flux conditions occur. This position is determined after growth and assumed to occur at the boundary between regions with and without Ga droplets. Therefore, positive x represents the distance from a point in the Ga-rich region to the x=0 point, and negative x represents the distance from a point in the As-rich region to the x=0 point. The flux distribution is given by:¹⁷ 

where C is a constant and Ω is the solid angle. Of most importance to this study is how the relative fluxes change from the x=0 point into both the As-rich and Ga-rich regions of the sample. We can write the fluxes as a function of position (x) as:

where A(x) is the relative flux ratio at point x to the flux at x=0 and is determined using equation (1) and the specific geometry of the MBE system. Fig 2 shows A(x) for Ga, As, and Bi fluxes, as well as the Bi concentration, determined using high-resolution x-ray diffraction (HRXRD) (004) rocking curves, as a function of position. Bi incorporation is close to 0 when x < 0. In this study, we only use data from points x≥0. This set of data points comprises the A set which represents GaAs_1−yBi_y grown under the same temperature and growth rate with varying Bi concentration resulting solely from variations in flux ratios.

Set B is comprised of different samples that were prepared using substrate rotation and grown with various growth temperatures (290°C to 350°C) and rates (0.16μm/h to 1.35μm/h). The resultant Bi composition varies from 0.58% to 3.07%. More information on these samples can be found in our previous work.¹⁶ 

Typical GaAs_1−yBi_y Raman spectra can be found in two prior studies.^9,10 Four polarized quasi-backscattering geometries can be used for the GaAs zinc blende structure:

where $x= 100 , y= 010 , z= 001 , X= 1 1 ̄ 0 , Y= ( 110 )$⁠.

If we focus on first-order GaAs-related modes, three peaks can be observed: the GaAs LO mode (∼292 cm⁻¹), the GaAs TO mode (∼268 cm⁻¹), and the LOPC mode (∼265 cm⁻¹). Steele et al.¹⁰ chose to use the −z(Y,  Y) z backscattering geometry, in which two components, the allowed deformation potential (DP) scattering and the impurity-induced Fröhlich scattering (F) of the LO mode are summed together.¹⁸ Our Raman spectra are shown in Figure 3(a). In our spectra, the GaAs TO mode can be neglected. Therefore, we focus only on the GaAs LO mode at Γ and the LOPC mode. We use two geometries, −z(x,  y) z  and the −z(x,  x) z, in order to differentiate DP and F scattering, respectively. The symmetry-allowed geometry −z(x,  y) z is used to analyze the relationship between the relative intensity ratio I_LOPC/I_LO, which represents the hole concentration,¹⁵ and the 300K photoluminescence (PL) intensity, or GaAs_1−yBi_y emission efficiency. In addition, the Fröhlich scattering, which is forbidden by the Raman selection rules, is analyzed in relation to Bi incorporation.

In general, the I_LOPC/I_LO ratio is positively correlated to carrier concentration in GaAs.¹⁰ M. Seon et al. studied the relationship between the GaAs I_LOPC/I_LO and doping density.¹⁵ Steele’s work extended this analysis to GaAs_1−yBi_y wherein the relative intensity ratio is an indication of the hole concentration and, therefore, the acceptor nature of Bi. In this study, the relationship found in Seon et al.’s work is utilized to convert the Raman signature I_LOPC/I_LO to hole concentration. The incorporated Bi is treated the same way as other dopants in GaAs. A linear relationship between log₁₀ (I_LOPC/I_LO) and log₁₀ (carrier concentration) is used for conversion.

300K PL was measured with 785nm diode laser with 25 mW incident power. A 100x objective was used to focus the incident light to a ∼4μm $2$ area. An InGaAs detector was used and the spectra were adjusted according to the response curve of the detector. A typical PL spectrum is shown in Figure 3(b). The GaAs peak is from the substrate. The GaAs_1−yBi_y peak intensities are recorded for comparison.

Figure 4 shows the relationships between the Bi concentration, the hole concentration as determined from the Raman data and the 300K PL intensity for both sets. For Set A, wherein the data is collected for varying Bi composition synthesized at the same temperature and growth rate, the Bi composition is nearly linearly related to the hole concentration, as shown in Fig 4(a). In addition, the PL Intensity is also nearly linearly related to the Bi content, as shown in Fig 4(b). The correlation coefficients for both relationships are larger than 0.98. However, if the samples are grown with different substrate temperatures and/or growth rates, as is the case for Set B, the Bi content and hole concentration appear, to some extent, linearly uncorrelated, as shown in Fig 4(c). We infer from this that Bi-related defects, such as antisites or clusters, which are dependent on growth temperature and rate, modify the behavior of Bi as an acceptor. In addition for the Set B samples, the 300K PL intensity is no longer linearly related to the Bi composition, as shown in Figure 4(d), showing that the optical emission is also clearly dependent upon the concentration of Bi-related defects determined presumably by how Bi is incorporated in relation to growth temperature and growth rate.

Interestingly, in Fig 5(a) and 5(b), we observe that, in general, the PL intensity is positively related to the hole concentration in both sets of samples. Therefore, we conclude that growth conditions, specifically the temperature and growth rate, affect Bi-related defects and acceptor formation. The resultant hole concentration is found to correlate with PL emission efficiency. In our previous study,¹⁶ we found that higher growth temperature and lower growth rate can enhance PL emission efficiency, and herein, we show that these growth conditions also enhance the hole concentration.

The hole concentrations, shown in Figures 4 and 5, are found to be in the range of 1x10¹⁸ to 1x10¹⁹ cm⁻³, which is consistent with Steele et al.’s results.¹⁰ However, as indicated by Steele, these concentrations are higher than those measured by Pettinari et al. Hall measurement was performed on these samples. However, the resistance was higher than the system limit and no signal could be captured. As a result, Eddy current non-contact sheet resistance measurements were performed on our samples. The average sheet resistance was found to be ∼1x10⁶ Ohms/sq. Considering that the thicknesses of our samples is ∼250nm, the mobility of the holes must be in the range of ∼0.02 cm²/(V ⋅s). Two possible reasons may lead to such low hole mobility. First, the Bi concentration is much higher than that of typical GaAs dopants and Bi interacts strongly with the valence band. Similarly, the incorporation of nitrogen in GaAs perturbs the conduction band and significantly reduces the electron mobility, for example, with only 0.2% N incorporation, the electron mobility decreases by almost 2 orders of magnitude.¹⁹ Second, our mobility may be lower than Pettinari et al.’s due to strong scattering from ionized centers. As shown above, the hole concentration is not only related to the Bi concentration, but also to the specific growth conditions, and, therefore, Bi-related defect concentrations. In general, samples grown at higher temperatures and lower growth rates show a higher hole concentration. However, the electrical characterization described in Pettinari et al.’s work was for a GaAs_1−yBi_y sample with a Bi concentration of ∼10%. In order to achieve this high Bi concentration, more ‘extreme’ growth conditions must be used, such as an extremely low temperature (in Pettinari’s work the growth temperature was 270°C). Therefore, in their case, in spite of a higher Bi content, we conclude that the Bi acceptor concentration should be much lower than that for our samples.

We also observe a Raman signature directly related to the concentration of Bi in GaAs_1−yBi_y. The LO mode intensity in the −z(x,  x) z backscattering geometry results from Fröhlich scattering. Generally, Fröhlich scattering is induced by impurities, including alloy elements and doping.^14,15 Little existing literature explores Fröhlich scattering in GaAs, perhaps due to its low scattering efficiency even in heavily-doped samples, as well as the presence of Raman modes directly from alloy elements. For example, InAs modes in InGaAs have been extensively studied. However, Fröhlich scattering should be of interest in dilute compounds such as GaAs_1−yBi_y since the Bi concentration is high enough to induce observable Fröhlich scattering, but the GaBi mode intensity is still relatively low making it difficult to observe if at all.⁹ As shown in Fig 6(a) and 6(b), the intensity of the LO mode caused by Fröhlich scattering shows a good linear relationship to the Bi concentration in both sets of samples, which indicates that this scattering mechanism is only related to Bi content, and is independent of the concentration of Bi-related defects.

In this work, we study the relationship between hole concentration as determined by the Raman signature, I_LOPC/I_LO, to 300K PL intensity. We show that the electrical and optical behavior of Bi is dependent on Bi content, but also growth conditions and therefore, by inference, to the creation of Bi-related defects. Also, another Raman signature LO (F) is discovered to be proportional to the Bi content independent of growth conditions. These signatures can be very useful for Raman studies of dilute bismides.