Impact of gallium concentration in the gas phase on composition of InGaAsN alloys grown by AP-MOVPE correlated with their structural and optical properties

The dependencies of the composition (In and N contents in a solid phase) and growth rate (GR) of InGaAsN epilayers on the Ga/(Ga + In) molar ratio determined on the basis of simulation of the measured diffraction curves are depicted in Fig. 1.

According to our expectations, the incorporation of indium drastically decreases, while nitrogen concentration increases with increasing gallium concentration in the gas phase. For the highest value of Ga/(Ga + In) = 0.970, indium content in InGaAsN alloy is negligible, while nitrogen content increases more than double. For the lower values of Ga/(Ga + In) = 0.821, 0.908, 0.935 (samples NI129, NI135, NI136), InGaAsN epilayers reveal a composition inhomogeneity [19‐25] typical for dilute nitrides, with a linear increase of indium content toward the sample surface. In the case of the growth rate, the obtained dependence is more complex (Fig. 1b). Initially, the value of GR decreases with increasing Ga concentration in the gas phase, while above the value of Ga/(Ga + In) = 0.93 ÷ 0.94 slightly increases. It shows that the epitaxial growth of InGaAsN alloys is mainly promoted by In incorporation (more probability of InGaAs crystallization than GaAsN), which confirms a very low N incorporation efficiency in GaAs under the applied growth conditions. The GR and composition for all the investigated structures are summarized in Table 1. In the case of the samples NI129, NI135 and NI136, the values of In/N ratio were calculated for the highest values of In content located at the surface, which has the greatest effect on the stresses generated in the InGaAsN layer.

The obtained diffraction curves of the samples grown at the gas Ga/(Ga + In) molar ratio changed from 0.821 to 0.970 are shown in Fig. 2a. For the lowest Ga/(Ga + In) values of 0.821 and 0.908 (samples NI129, NI135; blue curves), which favours higher In incorporation, the broad and blurred signal corresponding to the InGaAsN layer is placed to the left (for the smaller Bragg angles) of the main high intensity peak coming from the GaAs substrate. It indicates that the lattice parameter of these InGaAsN epilayers is greater than the GaAs which corresponds to In/N ratio higher than 3 (In/N = 3 guarantees a lattice match to GaAs [3]). For Ga/(Ga + In) > 0.908 (samples NI130-132 and NI136, red and black curves in Fig. 2a) the indium incorporation decreases, and the InGaAsN-related signal shifts to the bigger Bragg angles towards the GaAs substrate. For sample NI136, it is nearly overlapped with GaAs, while for Ga/(Ga + In) > 0.935 (samples NI130, NI131, NI132), it appears distinctly to the right of the GaAs peak. This means the last three mentioned samples have the lattice parameter smaller than the GaAs which corresponds to In/N ratio below 3. The simulations of the measured rocking curves of the compressive strained InGaAsN alloys with the highest In content (samples NI 129, NI135, NI136) reveal a composition gradation [20, 24, 25] typical for dilute nitrides. The indium content increases linearly toward the sample surface, while the N content remains constant (Table 1, subsection 3.1). This tendency is not observed for the tensile strained InGaAsN with In content below 1% (samples NI130, NI131, NI132), which exhibit a homogeneous composition. Figure 2b presents the dependence of the lattice mismatch ∆a/a (where ∆a is a difference between InGaAsN layer lattice parameter and GaAs substrate lattice parameter a), determined from the rocking curves for all investigated samples on the Ga/(Ga + In) molar ratio. The best lattice matching of InGaAsN to GaAs was obtained for the Ga/(Ga + In) molar ratio of 0.935 (sample NI136). This layer is compressively strained with the lattice mismatch ∆a/a = 605.61 ppm. A linear approximation of the obtained results (the equation is shown in Fig. 2b) allowed estimation of the value of Ga/(Ga + In) = 0.94, which guarantees the lattice matching of InGaAsN to GaAs (the zero place of the determined linear function, ∆a = 0).

Figure 3 presents the reciprocal space maps around the asymmetrical reflection (2 2 4) of heterostructures containing InGaAsN with the highest In content (sample NI129), InGaAsN with the best lattice-matched to GaAs (sample NI136) and InGaAsN with the largest N content (sample NI132). They show that all investigated structures are fully strained. In the case of the most compressively strained InGaAsN layer (sample NI129, Fig. 3a), a diffused scattering along the horizontal axis representing the in-plane scattering vector component Q_x is distinctly visible. It may be caused by the presence of structural defects [26] or morphological peculiarities. This scattering was described by full width at half maximum (FWHM) of main peak, determined from (0 0 4) symmetrical space maps reflection (not included here). For sample NI129, the highest value of 12 arc sec was obtained, while only 7 arc sec revealed the samples NI136 and NI132. It confirms the results coming from AFM measurements, which showed the worst surface roughness of sample NI129, which is probably connected with 2-D and/or 3-D island growth modes.

The main aim of the AFM measurements was to evaluate the surface roughness of the investigated structures depending on the gas Ga/(Ga + In) molar ratio, which directly corresponds to In and N contents in the InGaAsN solid phase, the composition inhomogeneity and the induced local strains. Typical for dilute nitrides growth conditions: low growth temperatures, a small amount of AsH₃ together (in our case) with the use of the atmospheric pressure process, preventing the step-flow mode growth and favouring quasi-two-dimensional (2-D) or three-dimensional (3-D) growth modes [27‐30].

Figure 4 shows the AFM images (10 × 10, 5 × 5 and 2 × 2 μm² scans) and the evaluated root mean square roughness (R_RMS) for: the reference GaAs layer (Fig. 4a) grown under the same conditions as GaAs buffer in the test structures, InGaAsN with the highest In content (sample NI129, the greatest compressive strains—Fig. 4b), InGaAsN lattice-matched to GaAs (sample NI136, the smallest value of the introduced stress—Fig. 4c) and InGaAsN with the largest N content and negligible In content (sample NI132, the greatest tensile strains—Fig. 4d). The surface morphology of the investigated InGaAsN epilayers distinctly differs in comparison to the reference GaAs layer with a very smooth (R_RMS = 1.69 ÷ 2.04 Å) step/terrace structure, characteristic of the step-flow mode growth. In the case of InGaAsN, the visible surface features indicate 2-D/3-D island formation on terraces and a different way of their subsequent coalescence. As previously mentioned, this island structure is the most likely for our InGaAsN growth parameters as: a low growth temperature (T_g = 585 °C) resulting in a small surface adatom diffusion length, a low AsH₃ concentration in the gas phase (As/(As + N) molar ratio of 0.27) and the use of the atmospheric pressure MOVPE process, causing a higher amount of incorporated impurities during InGaAsN growth (mainly organic precursor debris). Additionally, a small misorientation of GaAs substrates (± 0.05 ÷ 0.08°) causes a large step separation, which also favours the 2-D nucleation on the middle area of the terraces [27]. All InGaAsN surfaces show a high density of nanometer-size islands initiating the nucleation process. Their height (4.0 ± 0.5 nm) and diameter (35 ± 2 nm) were determined based on detailed AFM imaging performed with a supersharp AFM tip (radius < 2 nm) in order to minimize the effect of a tip shape on the measured surface features. The exemplary 150 × 150 nm² surface image of sample NI136 and the measured profiles used for calculating the islands’ dimension are presented in Fig. 5a and b, respectively.

The formation of these islands can be connected with the precursor debris [12] and/or the disrupted step edges [27, 28]. In the case of the compressively strained InGaAsN layer (sample NI129, Fig. 4b), elongated bulges (about 0.5 µm in length) beside these islands are also visible. The surface roughness R_RMS changes from 11.7 Å (2 × 2 μm² scan) to 17.5 Å (10 × 10 μm² scan), which indicates the 2-D and/or 3-D island growth mode. The elongated shape of the formed islands can be connected with a higher mobility and a higher surface-diffusion length of the Ga and In adatoms (the lowest Ga/(Ga + In) ratio of 0.821) [27, 29]. The surface morphology of lattice-matched InGaAsN (sample NI136, Fig. 4c), grown at a higher Ga/(Ga + In) molar ratio of 0.935, shows a mesh-like structure described in [30]. Firstly, they are locally formed and randomly distributed islands, which in the second stage join together to create the observed mesh morphology. The R_RMS roughness is varied from 10.05 Å (2 × 2 μm² scan) to 11.83 Å (10 × 10 μm² scan), which is smaller in comparison to the most compressively strained NI129 sample. The observable effect of increasing the root mean square roughness parameter with enlarging the AFM image size (which is most pronounced for NI129 sample) is connected with the roughness area dependence mechanism [31, 32]. It shows the scaling behaviour of surface topography, when, by extending the sampling area, the larger-scale surface topography components are included in the R_RMS parameter calculation, resulting in its increase. In the case of the highest Ga/(Ga + In) molar ratio of 0.970, when the most tensile-strained InGaAsN was obtained with a negligible In content (probably the GaAsN epilayer with N content of 0.91% was grown based on HRXRD results), the surface morphology shows a slightly elongated network structure similar to tangled “spaghetti” (sample NI132, Fig. 4d). The evaluated R_RMS roughness is comparable to the lattice-matched InGaAsN (samples NI136). The presented AFM studies suggest that the network morphology of InGaAsN epilayers created for the highest values of Ga/(Ga + In) is not directly connected with introduced strains, but is probably related to the decreased group III adatom surface mobility (diffusion-length) resulting from an increase of Ga concentration in the gas phase. It is also confirmed by the changes of the R_RMS parameter for the samples NI129, NI136 and NI132 (2 × 2 μm² and 5 × 5 μm² scans). With the increasing Ga/(Ga + In) molar ratio, the R_RMS parameter firstly decreases, and then increases again. It can be explained by the competitive influence of the strains and group III adatoms surface mobility on InGaAsN layers roughness. The smaller roughness of sample NI136 in comparison to sample NI129 means that a distinct reduction of compressive strains for Ga/(Ga + In)=0.935 has a greater impact on improvement of InGaAsN layer flatness than the lower adatoms mobility, in this case, worsens its surface quality. For sample NI132, grown at highest value of Ga/(Ga + In)=0.970, the R_RMS parameter slightly increases in comparison to sample NI136, but remains smaller than for sample NI129. It means the tensile strains introduced into this layer (much lower than for sample NI129) have a negligible influence on the InGaAsN surface quality, and a formation of the network morphology with a higher roughness is related to exceedingly low group III adatoms mobility.

The obtained PL spectra reveal composition inhomogeneity (except of the samples NI135 and NI136, best lattice-matched to GaAs), visible as broad or splitting spectra shown in Fig. 6a. This inhomogeneity is typical for metastable alloys, such as dilute nitrides, and is spontaneously created during their growth, especially at higher process temperatures that enhance phase segregation [33]. This fact can be explained by a dominance of surface kinetics, the nonlinear dependence of N incorporation on In content [34], and of the strain gradient [25]. Both alloy content fluctuations and an inhomogeneous distribution of the N-related defects result in a band gap variation [35]. It is distinctly visible in the PL spectra recorded for InGaAsN with the highest In (sample NI129) and N (sample NI132) contents—Fig. 6b. They exhibit overlapping emissions: a main PL peak at about 1.21 ÷ 1.22 eV (band gap-related emission), and some response at about 1.16 ÷ 1.17 eV (labeled as E₂ in Fig. 6b). The broadening of PL spectra FWHM_PL is the largest (75 ÷ 79 meV) for the structures with the highest N content (samples NI130 ÷ NI132), which results from a broad emission (E₂), probably connected with a high density of N-related states localized close to the conduction band minimum [35]. In the case of sample NI129 with the highest In content, the value of FWHM_PL is smaller (66 meV), and a distinct split between the main emission signals is clearly visible, which suggests an influence of a phase separation effect.

The PL spectrum of the InGaAsN layer, the best lattice-matched to GaAs presented in Fig. 6b (sample NI136), shows a well-resolved band gap-related emission at 1.2 eV, with a small broadening (FWHM_PL = 55 meV). The dependence of the parameters evaluated from the PL spectra (E_PLmax—the energy of PL signal maximum and FWHM_PL—the PL signal broadening) on the applied Ga/(Ga + In) molar ratio is presented in Fig. 6c. All determined parameters are listed in Table 2.

The measured CER spectra of the test InGaAsN/GaAs heterostructures recorded at room temperature (RT) are presented in Fig. 6a. They exhibit, besides distinct optical transitions related to GaAs at energy of 1.43 eV, the weak resonances detected at about 1.2 eV associated with the InGaAsN layer. Additionally, CER spectra related to InGaAsN with three specific composition ratios: In/N = 28 (the highest In content, sample NI129), In/N ratio = 4 (the best lattice-matched to GaAs, sample NI136) and In/N = 0 (indium absence, sample NI132), are shown in Fig. 7b.

The band gap energy (E_g) of the InGaAsN epilayers and CER transition broadening (Γ parameter) were estimated by fitting the corresponding resonances using a low-field electro-modulation Lorentzian line shape functional form, described in [36, 37]. The values of the obtained parameters, as well as the calculated ratio of InGaAsN to GaAs CER signal amplitudes A_InGaAsN/A_GaAs, are listed in Table 3.

The InGaAsN epilayers with the highest and the lowest In/N composition ratios (the most lattice-mismatched to GaAs) have a very week CER signal, whereas the narrowest resonance reveals sample NI129 with the highest In content. A similar relationship was observed for the InGaAs structures with decreasing In content of 12.8, 8.9 and 3.4% for which the values of Γ parameter were 12.6, 13.9 and 17.2 meV, respectively [38]. Samples NI135 and NI136 (the best lattice-matched to GaAs) have the best optical properties, which exhibit the strongest InGaAsN-related transition with its broadening (Γ parameter) below 50 meV. The determined value of the InGaAsN band gap is about 1.2 eV for all investigated heterostructures. Therefore, an increase of the gas phase molar ratio Ga/(Ga + In) allows to increase the N content and a growth of InGaAsN alloys lattice-matched to GaAs, but further investigations are needed for lowering their band gap to the applications value of 1.0 eV, required for SCs.