1. Introduction

Lasers operating in the near-infrared spectral region (at 1.3 and 1.55 $\mu$m) are critical for telecommunication applications. Conventionally, sources in this wavelength range are InP-based quantum wells (QWs) using GaInPAs active regions. However, the low heat conductivity of GaInPAs layers and the small refractive index contrast between alloys lattice-matched to InP motivated the investigation of GaAs-based sources. The high index of refraction contrast and small lattice mismatch between Al(Ga)As and GaAs makes GaAs substrates an ideal choice for vertical-cavity surface emitting lasers (VCSELs) [1,2]. However, few choices exist for near-infrared active regions nearly lattice-matched to GaAs emitting at 1.3 $\mu$m or 1.55 $\mu$m. Highly-strained InGaAs/GaAs QW lasers have been demonstrated with low-threshold currents; however, the increasing strain in the QW with increasing In concentration degrades the optical quality thus limiting the emission wavelengths to $\lambda <$ 1.3 $\mu$m [3–5]. Another approach involves the introduction of nitrogen or bismuth to InGaAs, forming a highly mismatched alloy. The addition of nitrogen to InGaAs resulted in a reduction in not only the compressive strain in the QW, but also a favorable reduction in the bandgap due to band anticrossing interaction [6] Similarly, the introduction of dilutes amount of bismuth rapidly reduces the bandgap allowing for emission at 1.3 $\mu$m [7]. Dilute-nitride lasers demonstrated extended emission wavelengths, beyond 1.55 $\mu$m, on GaAs; however, the challenging growth of these alloys has limited their adoption [8].

A relatively unexplored alloys for this application are the dilute borides. The small size of the boron atom reduces the lattice mismatch of InGaAs grown on GaAs, enabling longer emission wavelengths. However, B-III-V alloys are highly-mismatched alloys due to the large difference in size between the boron atom and remaining atoms [9]. The highly-mismatched nature of the alloy previously complicated its growth and limited boron incorporation in B-III-V alloys. Until recently the amount of boron incorporated in BGaInAs was limited to $\leq$ 4% [10]. As the In concentration in BGaInAs was increased, the amount of boron incorporated decreased to $\leq$ 2% [11,12]. The challenging synthesis of these alloys limited the understanding of the effects on the band structure arising from addition of boron. However, recent advancements in the molecular beam epitaxy growth of BGa(In)As alloys has enabled boron concentrations up to 18% in coherently-strained BGaAs epitaxial layers on GaP [13], increased simultaneous B and In concentrations in BGaInAs, [14] and the demonstration of BGaInAs active region light-emitting diodes [15]. These recent improvements in growth of the alloys has enabled further understanding of the effects of alloying with boron on band structure [16].

In this work, BGaInAs/GaAs QWs are explored as an alternative to InGaAs/GaAs QWs. For this purpose, we measured photoreflectance (PR), power-dependent photoluminescence (PDPL), and temperature-dependent photoluminescence (TDPL) spectra for a set of samples with $\sim$40% of indium and up to 4.75% of boron incorporated. We examined effects of boron addition into the thin film of InGaAs/GaAs QWs on relative valence band offset ratio and optical quality.

2. Materials and methods

The samples were grown in a solid-source molecular beam epitaxy (MBE) system equipped with an arsenic cracker to supply As$_{2}$, gallium, indium, and bismuth SUMO cells, and an electron beam evaporator to source boron. BGaInAs/GaAs QWs with a thickness of approximately 10 nm were grown with a bismuth surfactant as described in [14,15]. The alloy composition and thickness of each sample were determined from high-resolution X-ray diffraction $\omega$-2$\theta$ measurements, see details in [14]. The measurements showed that the QW in the reference sample (0% B) is 11 nm thick, whereas QWs in other samples (with boron) are 9.5 nm thick.

A "bright configuration" experimental setup was used to measure the PR spectra [17]. A single grating 0.55 m focal-length monochromator and a thermoelectrically-cooled InGaAs photo-diode were used to disperse and detect the light reflected from the samples. A 150 W tungsten-halogen bulb with the long pass filter cutting at 0.8 $\mu$m was used as a probe source and a continuous wave laser (405 nm line) as a pumping source. Both beams were focused onto the samples to a diameter of $\sim$2 mm, and the power of the laser beam was set to $\sim$100 mW. The pump beam was modulated by a mechanical chopper at a frequency of 280 Hz. A phase sensitive detection of the PR signal was accomplished using a lock-in amplifier. Photoluminescence measurements were taken employing a 532 nm diode-pumped solid-state laser for excitation and a Hamamatsu TG-cooled NIR-I mini-spectrometer C9913GC for detection. The laser beam was focused on the sample to the diameter of around 0.2 mm with its intensity reduced to values below 120 mW by a gray filter.

Energy levels at the center of the first Brillouin zone (CFBZ) and band structures are computed based on the 8-band $\boldsymbol {k} \cdot \boldsymbol {p}$ model with an envelope function approximation. Oscillator strengths are calculated for transverse electric (TE) and transverse magnetic (TM) polarization of light within the electric dipole approximation. For the details on the $\boldsymbol {k}\cdot \boldsymbol {p}$ Hamiltonian see our previous papers [18–20]. Band structure parameters and bowing parameters, used during $\boldsymbol {k} \cdot \boldsymbol {p}$-based calculations, are listed in the Tables 1 and 2, respectively. The parameters are: valence band offset ($VBO$), spin-orbit splitting ($\Delta _{so}$), direct energy gap at 0 K ($E_g$), Varshni’s parameters ($\alpha$ and $\beta$), effective mass of electrons in conduction band ($m_e$), Luttinger parameters ($\gamma _1$, $\gamma _2$, and $\gamma _3$), deformation potentials ($a_c$, $a_v$, and $b$), lattice constant ($a_{lc}$), elastic constants ($C_{11}$ and $C_{12}$), and Kane’s energy ($E_p$). GaAs and InAs are known semiconductor binary compounds, for which almost all band parameters are taken from a review work of Vurgaftman et al. [21]. With exception to valence band offsets (VBOs) taken after Hart & Zunger [9], who calculated them including d-orbitals that allowed to obtain more reliable values. BAs remains under investigated compared to other III-V binary compounds; therefore, its band parameters have not been fully established yet. In this work, a set of parameters proposed and discussed in our previous papers is used for this material [22,23]. For convenience, all VBO parameters gathered in this work are decreased by the value of the VBO for GaAs since GaAs forms the barriers of the QWs. Parameters that are not available (N.A.) for BaAs ($\alpha$, $\beta$ and $E_p$) are approximated by proper parameters of In$_{0.4}$Ga$_{0.6}$As. This approximation is proposed since only small mole fractions of boron are considered in the alloys containing around 40 % of the indium. Since hydrostatic deformation potentials for BaAs are not available separately for conduction and valence bands another approximation is applied. The hydrostatic deformation potential for valence bands ($a_v$) is set to zero, and for conduction bands ($a_c$) is set to the value derived theoretically by Chimot et al. [24] for the direct bandgap. Three bowing parameters ($\Delta _{so}$, $m_e$, and $a_c$) in InGaAs are taken after the review publication of Vurgaftman et al [21]. Alloys BGaAs were recently investigated for bowings describing $\Delta _{so}$ and $E_g^{0K}$ [16]. The resultant parameters are used in this work. Remaining bowing parameters listed in the Table 2 are results of our research. They are marked with asterisks and described further in this paper.

Table 1. Parameters of BAs, GaAs, and InAs binary compounds.

View Table | View all tables in this article

Table 2. Bowing parameters of InGaAs, BGaAs, and BInAs ternary alloys.

View Table | View all tables in this article

3. Results and discussion

Figures 1(a) and (b) show PR spectra measured at room temperature (RT), around 300 K, and at low temperature (LT), 20 K, for BGaInAs/GaAs single QWs with In mole fractions of $\sim$ 40% and various mole fractions of boron. The strongest resonances around 1.42 eV and 1.52 eV at RT and LT respectively are associated with photon absorption in GaAs barriers. The shape of this signal changes from sample to sample due to a different band bending in GaAs layers resulting in individual contributions to the entire signal [27]. The PR oscillations observed at lower energies are associated with optical transitions in BGaInAs/GaAs single QWs. The resonances are analyzed using the low-field electromodulation Lorentzian line-shape function form given by Eq. (1), known also as the Aspnes formula [28], where $C_n$, $\phi _n$, $\Gamma _n$ and $E_n$ are the resonance amplitude, phase, broadening, and energy of the $n$-th transition respectively.

 

Fig. 1. PR spectra of BGaInAs/GaAs QWs with various mole fractions of boron. Experimental results are denoted by solid gray lines. Solid blue lines represent theoretical fits based on Eq. (1). Moduli of excitonic transitions are plotted by dotted red lines. Moduli of 11H, 12H, and 22H transitions are plotted by solid black lines. Moduli of other observed transitions are plotted by solid colored lines. The signal above 1.4 eV on each spectrum is related to the GaAs substrate. Since the signal from the substrate is relatively strong, it is presented after down-scaling. Results are obtained at (a) RT and (b) LT.

Download Full Size | PDF

Since observed resonances are related to states in QWs of typical thicknesses, it is reasonable to analyze them on the basis of a series of calculations within the 8-band $\boldsymbol {k}\cdot \boldsymbol {p}$ model for QWs. Figure 2(a) shows an exemplary energy profile and electronic band structure of the QW with 2.37% of boron in the thin film. This profile describes one of the measured samples, with identical boron fraction. Three electron states (1E, 2E, and 3E) and five hole states at the $\Gamma$ point (heavy holes: 1H, 2H, and 3H; light holes: 1L and 2L) are found in the majority of examined QWs. Up to six resonances in the Fig. 1 are identified with satisfying accuracy basing on transition energies between the states calculated for each sample. Pairs of the states related to these transitions are symbolically presented in the Fig. 2(b). The notation used in this paper, $nm$H(L), denotes transitions between states in an $n$-th conduction subband and states in an $m$-th heavy-hole (light-hole) subband.

 

Fig. 2. Interpretation of determined transitions in terms of localized states in 9.5 nm QWs with 40% of indium. (a) Energy profile of chosen QW with marked and labeled energy levels, electronic band structure. (b) Scheme presenting sets of wave functions involved in experimentally identified transitions. (c) Oscillator strengths, obtained in the PR measurements and calculated energies of transitions in QWs at RT. Oscillator strengths associated with transverse electric (TE) and transverse magnetic (TM) polarization are denoted by yellow and blue circles respectively. Experimental results obtained at the RT are depicted by black squares. Solid black lines show energies of observed transitions involving heavy holes. Dashed black lines are related to measured transitions involving light holes. Thin dotted lines show nominally forbidden transitions not verified by the PR experiment. A linear function fitted to the 11H transition is shown my dotted red line.

Download Full Size | PDF

Up to five allowed transitions (11H, 22H, 11L, 22L, and 33H) and one nominally forbidden transition (12H) are identified for each sample with boron. The majority of detected resonances are each related to a unique optical transition. The remaining few resonances, that correspond to multiple transitions, are also included in the fitted function. Such resonances are observed especially for a reference QW (the one without boron), since it is the thickest one. In this sample, many transitions are of similar energy making it impossible to distinguish them on the basis of the PR spectra. Also, contrasting to the samples with boron, there are no light-hole-related transitions in the reference sample, which is an effect of strain and type-II band alignment for light-holes.

In the experiments performed at LT, an additional excitonic transition is observed in the samples with 0%, 0.6%, and 1.2% of boron. They are located up to 5 meV below an optical transition 11H, therefore with bounding energy typical for QW regime [29]. A line fitted for the sample with 2.37% of boron does not match the experimental data as well as lines of the other samples, because the excitonic transition (still existing in the sample) has larger broadening and therefore contributes to a background-like signal without very well defined energy. Omitting such broad excitonic transition does not influence determination of the energies related to other transitions; however, it is visible in the photoluminescence (PL) spectra discussed further. No excitonic transitions is resolved for the sample with 4.75% of boron due to larger broadeing of PR resonances, which is attributed to larger alloy inhomogeneities in this sample.

It is worth noting, that a deep confinement is present for electrons in all samples, since three conduction subbands always contribute to the spectra. Such confinement is expected for this system because InGaAs alloys with high contents of indium (here around 40%) are already known to form an attractive type-I band alignment in InGaAs/GaAs heterostructures, where carriers of both kinds are strongly confined in the InGaAs layer.

Strong resonances observed in PR spectra are related to calculations held for all the samples at RT, since the material parameters used for the calculations within the model provide better agreement in these conditions. The experimentally derived and the calculated transition energies together with oscillator strengths are shown in the Fig. 2(c). The oscillator strengths are non-zero only for allowed transitions in the BGaInAs/GaAs QWs, even though forbidden transitions can be clearly observed within the experiment, such as H12, due to an electric field build in the structure. Based on this comparison, it is visible, that first four resonances with boron can be clearly assigned to one of single optical transitions: 11H, 12H, 22H, or 11L in almost all cases. Two strong resonances around 1.27 eV and around 1.31 eV are distinguished for the samples with 0.6%, 1.2%, and 2.37% of BAs. The one of the lower energy is most probably related to the allowed transitions 22L, but also may have its source in a transition 32H for the lower boron fractions. The higher-energy resonance matches the calculated transition 33H very well. It should be noted, that these QWs experience the lateral strain of about 3.5%, which is a relatively high value for such structures. This strain is an additional source of difficulty in an accurate modeling of the QWs since it amplifies errors arising from the limited certainty of deformation potentials used in computations. Therefore, some divergence between calculated values and experimental ones are visible in the presented results, especially for the light-hole-related transitions. An other issue appears in the case of the sample with the highest boron fraction (4.74%). The lower optical quality of this sample results in a higher broadening of all observed resonances, additionally decreasing the accuracy of the energy determination. However, in all the cases, the lowest-energy transitions are always found with at least 5 meV accuracy. The energies of the first confined states are also computed with the highest accuracy, which allows conducting a more conclusive analysis based on the lowest-energy transitions.

The main part of the analysis is preceded by agreeing the 11H transition (the fundamental transition energy) calculated with the measured one for the reference sample (without boron). Creating a set of parameters based only on the values reported up to date for InGaAs alloys did not allow to reproduce the energy of 11H for this sample obtained by the fitted line, and therefore for QWs with low boron mole fraction. However, a slight modification of the bowing parameter of the direct energy gap, encouraged by the large range of reported values in the literature, enabled us to reach agreement. It is obtained with the bowing parameter of the energy gap equal 0.62 eV, which is only 0.04 eV higher than the established one for high indium contents [25]. The data related to 11H for the rest of the samples (with boron) are fitted with a linear function shown in the Fig. 2(c). Reproduction of the fitted dependence in calculations is obtained by treating the bowing parameter of the direct gap in BInAs alloys as another fitting parameter. The experimental determination of this bowing parameter has not been previously reported. Here, we experimentally determine the BInAs direct bandgap bowing parameter to be 6 eV. Since the QWs exhibit deep confinements for both electrons and holes, the transition 11H is not sensitive to the change of the offset (if the confinements remain deep after the changes), which allows to obtain the latter bowing independently from the following analysis of the offsets.

One of practical parameters commonly applied to explore band alignments with the least number of additional variables is a relative valence band offset ratio ($Q_v$), defined as presented in the Fig. 3(a). It can be used further to derive values of VBOs in the samples. A very sensitive approach to determine $Q_v$ (or other parameters of the QWs) based on the experimentally obtained energies of optical transitions is to analyze differences between the energies as function of the $Q_v$. An additional advantage of such method is an elimination of residual inconsistencies related to the energy gap. Accurate determination of the band alignment in this work is based on the transitions energy differences (TEDs), 22H-11H and 22H-12H, since they are satisfactory sensitive to changes of the $Q_v$.

 

Fig. 3. Analysis of the transitions energy differences, 22H-11H and 22H-12H, obtained from PR measurements. (a) A schematic introducing the valence band offset ratio $Q_v$ (unstrained structure) and the thickness of QWs, which are treated as variables in the analysis. (b,c) Transitions energy differences, 22H-11H and 22H-12H as a function of (b) $Q_v$ and (c) thickness. (b,c) Theoretical calculations based on $\boldsymbol {k}\cdot \boldsymbol {p}$ model in RT are depicted by solid black lines. Results obtained in PR measurements are marked by horizontal lines. (b) Vertical arrows denote values of $Q_v$ and thicknesses determined by collation. (c) Gray bands indicate estimated accuracy of nominal thicknesses.

Download Full Size | PDF

Figure 3(b) shows calculated TEDs as a function of $Q_v$ for QWs with fractions such as in the measured samples and with thicknesses equal to the nominal ones. The calculations are held only for the cases at RT, since the used material parameters are the most accurate at the RT. However, it is expected that TEDs are independent from the temperature, since the biggest effect of the change of the temperature is attributed to the change of the energy gaps that is already removed from this analysis. In addition, $Q_v$ usually is not expected to be dependent on the temperature. Therefore, the experimental energies obtained both at RT and LT are included in this analysis and appear to be in good agreement with each other. Intersections of the calculated curves and experimental horizontal lines provide the finest values of the $Q_v$ for the given TED at RT or LT. Average values of the resulting $Q_v$ for each samples are calculated and marked by vertical arrows. These averages are presented in the Fig. 4(a) and fitted with a linear function given by the Eq. (3).

 

Fig. 4. Valence band alignments of heterostructures based on BAs, GaAs, and InAs binary compounds. (a) Values of $Q_v$ extracted from the analysis of 22H and 11H transitions are depicted by black squares. A linear function fitted to the data is denoted by a dashed red line. Values of $Q_v$ resulting from chosen VBOs of constituent binary compounds and bowing parameters are depicted by a solid black line.(b) Conduction band minimums CB($\Gamma _6$) and valence band maximums VB($\Gamma _8$) as functions of the lattice constant for unstained binary compounds (BAs, GaAs, and InAs) with their alloys (BInAs, BGaAs, GaInAs). Bowing parameters of direct energy gaps and VBOs used in this work are added to the plot.

Download Full Size | PDF

The consistency of the derived fitted line of the $Q_v$ is further verified. Values of $Q_v$, for each sample, are taken from the fitted line and used to compute dependencies of TEDs on the thickness of the QWs. In the Fig. 3(c) these dependencies are plotted together with the experimental result in a similar manner as in case of the $Q_v$. Thicknesses determined in such a way confirm the nominal values almost perfectly in most of the cases. All calculated transition energies shown in the Fig. 2(c) also agree with the experiment accurately, despite the TEDs-based analysis involves only the transitions: 11H, 12H, and 22H.

The fitted line of experimental $Q_v$ shown in the Fig. 4(a) is used to conclude on bowing parameters of VBOs for boron-containing alloys: BInAs and BGaAs. The $Q_v$ strongly depends on the VBOs of constituent binary compounds and their bowing parameters. The VBOs taken from Hart & Zunger [9] appear to be the best choice for agreeing the experimental with the calculated values of $Q_v$. With this set of VBOs, the agreement for the reference sample is achieved with a reasonable bowing parameter of VBO for GaInAs equal −0.18 eV. Agreement for higher mole fractions of boron is obtained by fitting bowing parameters of VBOs for boron-containing alloys: BInAs and BGaAs. The slope of $Q_v$ line is used to determine the VBO bowing parameters with an assumption that they equal to some fraction of bowing parameters of respective energy gaps in the BInAs and BGaAs alloys. With this approach it is found that the bowing parameters of VBOs are equal around 65% of the bowing parameters of energy gaps in respective alloys. Therefore, they take value of around $-2.3$ eV for BGaAs and $-3.9$ eV for BInAs. The $Q_v$ resulting from the choice of these parameters is shown in the Fig. 4(a). These two bowing parameters have not been reported yet in the literature. It is worth noticing that while the obtained values are only approximations of the real bowing parameters of VBOs, non-zero bowing parameters of the VBO for the alloys with boron exist without any doubts. Resulting band alignments for unstrained materials are presented in the Fig. 4(b).

TDPL and PDPL measurements are performed for each of the samples as experiments complementary to the PR, which allow to evaluate the optical quality of the investigated samples. Representative spectra and analysis relating to all the samples are shown in the Fig. 5. Panels 5(a)-(d) present sets of TDPL with the excitation power equal 40 mW and PDPL at LT equal 20 K. The spectra are presented for the samples with nominal fractions of 2.37% and 4.75%. The spectra obtained for the samples with BAs mole fractions: 0%, 0.6%, 1.2%, and 2.37%, show no relevant qualitative difference, hence the sample with the fraction 2.37% can be treated as the representative one for all of them. The sample with the highest mole fraction of boron (4.75%), in contrast, exhibited spectra of the shapes suggesting an existence of a significant alloy inhomogeneity on the basis of observed Stokes shift whose magnitude decreases with the excitation power that is attributed to saturation of localized states. Moreover, the energy shift of the PL spectra for this sample can be also affected by screening of the build-in electric field.

 

Fig. 5. Results of the photoluminescence measurements. (a,b) Exemplary set of TDPL spectra measured in temperatures from 20 K up to 300 K with excitation power of 40 mW. (c,d) Exemplary set of PDPL spectra measured at temperature of 20 K with excitation power up to 120 mW. (e) PL peak position as a function of the temperature. (f) Integrated PDPL intensity at temperature of 20 K as a function of excitation power. (g) Broadening of the fundamental transition extracted from the TDPL measurements as a function of temperature. (h) The broadening at temperatures equal 20 K and 300 K as a function of boron mole fraction. (i) External quantum efficiency as a function of boron mole fraction, estimated basing on TDPL measurements according to Eq. (5). (j) Integrated PL intensity at excitation power equal 40 mW as a function of reciprocal temperature. (e, f, g, j) Results for each sample are presented with different open symbol according to legend above all the charts.

Download Full Size | PDF

Based on the spectra such as in the Fig. 5(a) and (b), peak energies of the fundamental transitions for all samples are determined and plotted in the Fig. 5(e) as functions of temperature. Energies of 11H transitions obtained from the PR measurements at LT and RT are additionally depicted by solid marks. The energies agree with the TDPL measurements at RT within the measurement accuracy, but there are significant discrepancies at LT. The PL at LT is found up to 5 meV under the 11H transition in all samples, but for one with 4.75% of boron is larger and equals 10 meV. The observed Stokes shift between PL peak and PR resonance means that the localized emission dominates at LT. However, none of presented TDPL spectra exhibit an S-shaped behavior reported by Hidouri et al. within their work related to BInGaAs/GaAs QW grown by metal organic chemical vapor deposition [30]. This means that our QWs are more uniform and therefore the effects of carrier location are weaker.

Measurements of PDPL are performed to additionally determine and confirm the nature of the fundamental transitions. Integrated intensities of the PDPL as functions of excitation power for all the samples are plotted in the Fig. 5(f). No spectral shift is observed for the QW with 2.37 % and a significant shift is visible for the QW with larger B (4.75 %), see Fig. 5(c) vs (d). This observation is very consistent with the magnitude of the Stokes shift observed for these samples. As the Stokes shift is twice larger in the QWs with 4.75 % B the observed blueshift of PL peak with the increase in the excitation power is attributed to an easier saturation of localized excitons with larger localization energy due to their lower density of states.

A power law function given by Eq. (4) is fitted to each data set. In Eq. (4), $I$ is integrated intensity around the peak, $P$ is an excitation power, and $\alpha$ is an exponent, which equals 1 for an excitonic transition and 2 for a band-to-band transition [31].

In our case $\alpha$ above 1 is obtained for all QW samples. The obtained exponents mean that the luminescence peak is mostly related either to a free- or a bound-exciton emission and possess some contribution of band-to-band emission. This contribution is the highest for the sample with 4.75% B, which could be related to lower exciton binding energy.

Broadening of the fundamental peaks of TDPL, shown in the Fig. 5(g), are estimated for another insight for the optical quality of the samples. For all of the samples, except the one with 4.75% of B, it varies comparably with the temperature in range from 10 meV at 20 K up to 20 meV at 300 K. Observed broadening of the sample with the highest mole fraction of B (4.57 %) is equals to around 25 meV without any clear dependence on temperature. The higher broadening of this sample indicates its larger inhomogeneity. The broadening dependence on the boron admixture at LT and at RT is plotted in the Fig. 5(h). The decrease of it with addition of boron up to 2.37% implies some improvement of the structural quality that could be attributed to a strain homogenization after the incorporation of boron. Generally, the broadening of the optical transition is expected to increase as an additional component is added and its concentration increased. In our samples this behavior is also observed, but for a higher concentration of B (QW sample with 4.57 % B). This shows that the dilute concentration of B is very beneficial for BGaInAs/GaAs QWs.

Such conclusion suggests that internal quantum efficiency should also get improved by incorporating the B atoms into the QWs. The internal quantum efficiency can be roughly estimated basing on TDPL with assumption, that nonradiative recombination at LT is absent [32]. It is calculated based on Eq. (5), where $I\left (T\right )$ denotes integrated intensity at the temperature T and a given excitation power, here 40 mW. $I\left (LT\right )$ denotes integrated intensity at the lowest temperature, here 20 K.

The results of analysis of IQE at tree temperatures are shown in the Fig. 5(i). The internal quantum an intensity clearly increases with the addition of B atoms in a regime of small mole fractions at all considered temperatures. At higher temperatures, however, the advantage of incorporation B into the material drops down for lower mole fractions and is unusually favorite for the sample with 4/57 % B. This can be attributed to an overestimated quantum yield at LT for this sample, which can be far from 100 %. For the remaining samples, the quantum yield at LT can also be overestimated, but from this analysis we can certainly conclude that adding B to GaInAs/GaAs QWs does not necessarily deteriorate their optical quality.

Integrated intensities of TDPL for each of the sample excited with the power of 40 mW are shown in the Fig. 5(j). The activation energies are determined based on the thermal quenching according to phenomenological expression given by an Eq. (6), where $I(T)$ is an integrated intensity around the peak of a given sample at the temperature $T$, $I_0$ is the integrated intensity at $T=0$, $\gamma _i$ are ratios between the radiative and non-raditaive lifetimes of energy levels participating in the thermal quenching processes, $E_{Ai}$ are $i$-th activation energies, and $k_B$ is the Boltzmann’s constant.

The first activation energy $E_{A1}$ equal to around 20 meV is found for all samples. The second activation energy $E_{A2}$ equal to around 100 meV is found for all samples except that with 4.75% of B. For the sample with the mole fraction 4.75% B, the energy $E_{A2}$ equals approximately 50 meV.

The activation energies can’t contribute to the transitions from the localized states in the QWs to their barriers since the energies are independent from the mole fraction of the boron, notably affecting the band alignment in the structures. Also, the found values are significantly smaller than the energy of freeing electron-hole pairs formed by 1E and 1H states or 2E and 2H states. Therefore, the activation energies in all cases are related to some nonradiative processes involving defects in the thin films or their vicinity in the barriers. The sample with the highest mole fraction of boron, however, shows qualitatively different routs of escaping carriers from the QW, since the $E_{A2}$ found is different than in the rest of the samples.

PR and PL measurements show that adding boron to GaInAs/GaAs QWs does not broaden optical transitions and does not deteriorate optical properties as is the case with nitrogen or bismuth diluted QWs [33,34]. The effect of strong enhanced broadening of optical transitions in nitrogen (or bismuth) diluted QWs is very strong and results from a strong band gap change per % of N (Bi) [35,36]. For boron diluted BGa(In)As, this change is very small in the regime of diluted concentration of boron [16] and hence the optical transitions almost does not broaden after the incorporation of small amount of boron.

Another issue is the nonradiative recombination, which is related to point defects. In the studied samples the efficiency of PL for boron-containing QWs is very comparable to the efficiency of the reference boron-free QW. Therefore, we claim that the optical quality of BGaInAs/GaAs QWs is also very satisfactory and can still be improved by further optimizing the growth process or the post-grown annealing.

4. Conclusion

Alloying conventional III-V alloys with B provides the possibility for strain engineering long-wavelength sources on GaAs, and even Si. Here, we experimentally determined for the first time the effects of adding B on the band alignments of BGaInAs QWs. An addition of B into the BGaInAs/GaAs QWs with around 40% of In results in an increase of valence band offset $Q_v$ by around 4% per 1% of B incorporated and in an increase of the fundamental transition energy by around 8 meV per 1% of B incorporated. Bowing parameters of the direct energy gap are chosen to find an agreement between the experiment and calculations: i) for the InGaAs is adjusted to 0.62 eV, ii) for the BInAs is set to the value of 6 eV. Bowing parameter of the VBO for InGaAs is set at value −0.18 eV such that it reproduces fundamental transition energy in the reference sample. In performed analysis of optical transitions in the set of samples, it was found that around 65% of the band-gap bowing parameter resides in the valence bands. This which suggests that type-II band alignment in BAs/GaAs and BAs/InAs heterostructures with small mole fractions of boron. Therefore bowing parameters of VBOs in BGaAs and BInAs are chosen at −2.3 eV and at −3.9 eV, respectively, for the calculations. Observed tendencies undoubtedly signify, that there is a non-zero bowing in the valence bands. Additionally, an optical quality of the measured samples is checked within TDPL and PDPL. The results suggest no degradation of optical quality with an addition of boron up to 2.75% and its deterioration for higher mole fractions of B. Alloying small amounts of B with GaInAs can clearly be used for successful strain reduction in QWs. This approach opens a possibility to reduce the band gap by an increase amount of indium incorporated, which is a very promising way to extend an emission spectral range of GaInAs-based semiconductor structures towards longer wavelengths.