A wire-on-well (WoW) structure was fabricated using InGaAs/GaAs/GaAsP superlattice device growth technology. This structure modifies the local concentration of carriers in the quantum well and lengthens the carrier lifetime to increase carrier transport efficiency. However, the reason for this remains unclear. Therefore, we investigated the detailed carrier transition properties using photoluminescence (PL) and photoreflectance measurements. Regarding the PL spectra at 4 K, two characteristic peaks at 1.39 and 1.34 eV were observed. Both transitions are attributed to the recombination between the first quantum level of the electron (e1) and that of the heavy hole (hh1). We also discussed the carrier distribution in the WoW structure and found that the maximum carrier existing probabilities for e1 and hh1 are located at different positions. The less overlapping of the wavefunctions causes low transition probability and results in the observed long carrier lifetime. An additional prominent result in the WoW structure is the blue shift attributed to the 1.34 eV PL peak induced by increasing laser excitation power. We found that the blue shift occurred by the screening of the electric field caused by the compressive strain, as in the case of the quantum-confined Stark effect.

The maximum conversion efficiency of solar cells consisting of a single junction, such as monocrystalline silicon, is theoretically calculated to be approximately 30%, and no further improvement could be expected (Shockley–Queisser limit).1 Therefore, multijunction solar cells using multiple semiconductor materials are attracting attention owing to their higher conversion efficiencies. In particular, lattice-matched, InGaP/GaAs/Ge triple-junction solar cells are extensively used on the ground and in space development because of their superior radiation resistance and device weight.2–5 However, the light absorption wavelength region of the middle GaAs cell is narrower than that of the top and bottom cells, thus resulting in a lower generated current. Each cell was connected in series; therefore, the current flowing through the entire cell was limited to the lowest value of the constituent cell. An alternative method is to replace the middle GaAs cells with semiconductor materials that provide the same current as the other cells. A semiconductor material that can be lattice-matched with Ge and has a bandgap energy (Eg) of ∼1.2 eV is required. InGaNAs is a candidate material that meets this requirement, and its Eg can be adjusted to 1.2 eV by altering its composition. However, it is difficult to achieve high-quality crystal growth. This is because the introduction of the N atoms during the growth becomes challenging to control Eg because of the negative and large bowing parameters.6–8 

Multiple quantum well (MQW) solar cells, including semiconductor materials containing a multiple quantum well (QW) structure in the light-absorbing layers of the solar cell, have been proposed.9–11 MQW needs to be multilayered to increase the amount of light absorption. However, defects at the interface of InGaAs/GaAsP and recombination of the photoexcited carriers in the quantum wells cause a decrease in carrier collection efficiency (CCE). A superlattice (SL) structure, wherein the barrier layer of MQW is thinned down to a few nanometers, has been studied.12 In this structure, wavefunctions overlap in adjacent quantum wells to form minibands. Since carriers are transported by the tunneling effect through the minibands, recombination within the SL structure is reduced, and the CCE is improved.13–15 However, if the barrier layer is extremely thin, the strain compensation condition cannot be met, and the crystal quality deteriorates. Therefore, it was not easy to stack more than 100 layers, which is necessary for optical absorption.

Therefore, the strain-relaxation-layer-inserted SL structure is being considered.16 A layer with an intermediate lattice constant is inserted to satisfy the strain compensation condition between the well and barrier layers. The strain accumulation can be suppressed, and the stacking of >100 layers becomes possible.17,18 In this case, carriers are transferred by tunneling through the minibands, thus suppressing nonradiative recombination and enabling efficient carrier transport.19 The improvement of CCE is then confirmed. The effect of induced strains in this structure has been studied from the viewpoint of radiative and nonradiative transitions.19–22 The rate equation for the photoexcited carriers has been investigated. Four types of carrier-dissipating processes were considered for both electrons and holes. These include radiative recombination, nonradiative recombination, thermal escape from the QW, and tunneling followed by thermal excitation. We found that the activation energy of the nonradiative transition had a maximum that corresponded to the minimum nonradiative transition probability at critical phosphorus compositions. A trade-off between the strain balance condition over the entire range of the SL structure and the local residual strain at the interfaces in the SL structures explains this result.21 

Conventional SL, referred to as planar superlattice (PSL) in this paper, face difficulties when extensive thinning of the barrier layer is attempted by enhancing the phosphorus composition owing to strain compensation concerns. Moreover, optimizing the carrier collection efficiency is formidable without substantially reducing the thickness of the barrier layer. Thus, we cannot achieve a high-carrier collection efficiency using our PSL. Recognizing these challenges, our research study focuses on the wire-on-well (WoW) structure.

An undulated superlattice (WoW) solar cell obtained by growing a strain-relaxation-layer-inserted PSL on a tilted substrate has been proposed.23 With detailed temperature control and the use of triethylgallium (TEGa) as a source of Ga in this solar cell, step bunching occurs.23 Periodically arranged InGaAs nanowires (NWs) were formed on the QWs owing to step bunching. Experiments confirmed that a WoW with this shape changes the local concentration of electrons and holes and improves the carrier lifetime and I–V characteristics for solar cell applications.24 Concerning the improvement of the carrier lifetime, Ref. 25 also showed that WoW has longer carrier lifetime than PSL. References 24 and 25 performed time-resolved photoluminescence for WoW and PSL with different compositions and thicknesses, respectively, grown under the same conditions. The results showed that WoW has a longer carrier lifetime than the PSL. An important implication of these results is that a PSL structure and long carrier lifetime of WoW can be reliably obtained via controlled growth conditions.

WoW also has a thin barrier region, where the carrier tunneling transport efficiency is high. This structure enables us to realize a thin barrier and low-phosphorous composition at the same time. Solar cells using WoW are expected to achieve high-conversion efficiency from these device characteristics. However, the detailed carrier transition process has not yet been clarified owing to its complex structure. Therefore, in this study, photoluminescence (PL) and photoreflectance (PR) measurements were performed to discuss the band structure and photoexcited carrier transition processes.

We recently developed a WoW structure based on SL fabrication technology.26,27 An intrinsic GaAs (i-GaAs) was stacked on an n-type GaAs substrate using metal-organic vapor phase epitaxy (MOVPE), and the SL structure was stacked within the i-layer for 20 cycles, as shown in Fig. 1(a). The SL consisted of a 1.93 nm In0.3Ga0.7As well layer and a 2.9 nm GaAs0.6P0.4 barrier layer, with a 2.6 nm GaAs strain relaxation layer inserted between them. SL growth was performed at 530 °C, and TEGa was used as the Ga source.23 InGaAs/GaAsP WoW was fabricated by growing it on a GaAs vicinal substrate. When the substrate was tilted by 6° with respect to the (001) plane toward [111] B, the thickness of the InGaAs layer was modulated periodically and continuously. The layer formed in a geometry like a triangle. We refer to this InGaAs layer as a triangularly shaped QW. At the same time, this InGaAs QW layer is structured like nanowires (NWs), extending toward the [ 1 ¯ 10 ] direction. Cross-sectional scanning transmission microscopy (STEM) images of the samples used in this study were not obtained. However, to better understand the sample structure, STEM images of samples were cited from Ref. 24 as displayed in Fig. 1(b). The upper part of Fig. 1(b) shows a STEM image of the WoW structure in the ( 1 ¯ 10 ) plane perpendicular to the NWs.24 The cross-sectional image of the triangular-shaped QW is drawn in the form of a white triangle in the figure. The local carrier densities within the InGaAs QW spatially change along the [ 1 ¯ 1 ¯ 0 ] direction. The carrier lifetime increased by a factor of four, and the fill factor of the I–V characteristics was improved, as stated in the introduction.24 The PSL sample was also grown on the GaAs (001) plane for reference, as shown in the lower STEM image of Fig. 1(b).24 Both samples were placed in the reactor of the MOVPE system during crystal growth at the same time so that the growth conditions other than the tilted angle of the substrate used for growth were identical.

FIG. 1.

(a) Schematic of planar superlattice (PSL) sample stacked with 20 layers of InGaAs/GaAs/GaAsP superlattice grown on the GaAs (001) substrate. The wire-on-well (WoW) sample was grown at an inclination of 6° with respect to the substrate. (b) The cross-sectional images of the scanning transmission microscopy for the WoW (upper figure) and PSL (lower figure).24 

FIG. 1.

(a) Schematic of planar superlattice (PSL) sample stacked with 20 layers of InGaAs/GaAs/GaAsP superlattice grown on the GaAs (001) substrate. The wire-on-well (WoW) sample was grown at an inclination of 6° with respect to the substrate. (b) The cross-sectional images of the scanning transmission microscopy for the WoW (upper figure) and PSL (lower figure).24 

Close modal

A halogen lamp was used as the probing light source in the PR measurements. A semiconductor laser (532 nm, 500 W/m2, Ventus 532, Laser Quantum) was used as the modulating light source to generate an electric field. It was passed through a mechanical chopper set at 270 Hz to irradiate the n-type GaAs surface. The probing light was incident on the sample surface at an angle of 45°. The alternating current and direct current components of the optical reflection signal were detected using a Si photodiode (S8746, Hamahoto), and their ratio (ΔR/R) was calculated. The sample was placed on a cold finger in a cryostat. A closed-cycle refrigerator was used to control the sample temperature from 4 to 300 K. In this study, the critical energy (ECR) was estimated by fitting the third-derivative Aspnes function to the acquired ΔR/R spectrum.28 However, if the oscillators are closely spaced with overlapping extrema, the Kramers–Kronig method proposed by Hosea becomes useful.29,30 This is also available when the observed derivative-like signal line intensity is too small to fit by the Aspnes function. We then used the PR modulus spectrum obtained after the Kramers–Kronig transformation of the ΔR/R spectrum. The ECR was estimated as the peak energy by fitting the Lorentzian function. The PL measurement was performed with the same semiconductor laser as an excitation light source, and the luminescence signal from the sample surface dispersed by a monochromator was detected by a Si photodiode. The PL measurements were performed at 4 K by changing the excitation light intensity from 5 to 2.5 × 104 W/m2. The sample temperature was also changed from 4 to 250 K, with the excitation light intensity set at 1.5 × 104 W/m2.

Figure 2 shows the PR modulus spectra of the WoW and PSL samples at room temperature. The peaks with intensities higher than 1.4 eV were attributed to the Eg of the GaAs substrate, buffer, and cap layers.31 Three peaks [A (1.31 eV), B (1.33 eV), and C (1.38 eV)] were observed in the WoW samples. The observed peaks below the bandgap region of GaAs were attributed to quantized levels. Because no peak was observed below peak A, we considered that this peak was attributed to the excitation of the carriers from the highest quantized level of the heavy hole band (hh1) to the lowest electron band (e1) in the triangular QW structure of the WoW sample. Peaks B and C are attributed to higher-order quantized-level transitions. However, these transition energies are difficult to calculate theoretically. Thus, we discussed only Peak A to understand the carrier transition. On the other hand, only one peak of B′ (at 1.33 eV) was observed for the PSL samples. In this structure, the transition energy from hh1 to e1 was theoretically calculated to be equal to 1.329 eV using Nextnano software for semiconductor nanodevices.32 We then attributed the observed lowest energy peak B′ to the hh1-e1 transition in the PSL samples. Note that the energy of peak B in the WoW samples is the same as that of B′ in the PSL samples.

FIG. 2.

Photoreflectance (PR) modulus spectra of WoW (a) and PSL (b) samples at room temperature. Three other signals for WoW (A, B, and C) and one for PSL (B′) were observed in addition to those obtained from the GaAs substrate.

FIG. 2.

Photoreflectance (PR) modulus spectra of WoW (a) and PSL (b) samples at room temperature. Three other signals for WoW (A, B, and C) and one for PSL (B′) were observed in addition to those obtained from the GaAs substrate.

Close modal

The low-temperature PL spectra of the WoW and PSL samples are shown in Figs. 3(a) and 3(b), respectively. The excitation intensity varied from 500 to 2.5 × 104 W/m2. Below the bandgap region of GaAs, peaks A (1.34 eV), B (1.39 eV), and B′ (1.40 eV) were observed at an excitation intensity of 2.5 × 104 W/m2 in the WoW and PSL sample cases, respectively. We consider that the peaks observed below 1.45 eV were due to the transitions between the quantum levels and were assigned to A, B, and B′ based on the similarities of the PR spectra, as shown in the figures. The differences in the peak energies due to the measured temperature change between the PR and PL spectra are discussed in Sec. IV.

FIG. 3.

PL spectra of WoW (a) and PSL (b) samples at 4 K as a function of excitation light intensity. Peak A blueshifts at decreasing excitation intensity; peaks B and B′ do not exhibit blue shifts.

FIG. 3.

PL spectra of WoW (a) and PSL (b) samples at 4 K as a function of excitation light intensity. Peak A blueshifts at decreasing excitation intensity; peaks B and B′ do not exhibit blue shifts.

Close modal

The peak energies are plotted in Fig. 4 as a function of excitation intensities to determine the excitation power dependence of the PL signals. The observed PL peaks A and B in the WoW samples were decomposed using a Gaussian distribution function to estimate the PL peak energies. The observed signal intensities were proportional to the excitation intensities. The characteristic feature of peak A in the WoW sample is that the peak energy shifts to higher energy regions at increasing excitation intensities. The amount of the blue shift was 0.007 eV per digit. This is not the case for peak B in the cases of WoW samples, as indicated by the red solid triangles. No excitation power dependence was observed. At the same time, no shift was observed for peak B′ in the PSL sample cases (not shown in the figure). PL signals from the bandgap region of the GaAs substrate were also observed, as in the case of the PR spectra. The PL peak energies in this bandgap region did not change at increasing excitation intensities in the PSL samples, except for one peak at ∼1.49 eV, as shown in Fig. 3(b). The observed shift may be explained by considering that this peak was due to donor–acceptor recombinations.33 

FIG. 4.

Observed photoluminescence (PL) peak energies of WoW shown as a function of excitation light intensity. Peak A (WoW10, 20, and 40) exhibits a blueshift when the excitation light intensity increases. No blue shift was observed for peak B.

FIG. 4.

Observed photoluminescence (PL) peak energies of WoW shown as a function of excitation light intensity. Peak A (WoW10, 20, and 40) exhibits a blueshift when the excitation light intensity increases. No blue shift was observed for peak B.

Close modal

Figure 5(a) shows the temperature dependence of the PL spectra of the WoW samples between 4 and 250 K. The excitation power intensity was 1.5 × 104 W/m2. The observed peaks shifted to a lower-energy region when the temperature increased. PL spectra acquired at room temperature could not be observed due to their small signal-to-noise ratios. The PL peak energies are plotted in Fig. 5(b) as a function of temperature. In this figure, the temperature dependencies of the signal peak expected from the energy gap of the GaAs buffer and substrate, the so-called Varshni's relation, are depicted using the black- and red-dashed curves for PSL and WoW samples, respectively. The observed data for peaks B and B′ were well reproduced using Varshni's relation. The observed PR peak energy at room temperature (from Fig. 2) is also marked using black arrows in Fig. 5(b). This shows that the temperature behavior is well understood.

FIG. 5.

(a) PL spectra of WoW from 4 to 250 K and (b) temperature dependence of the PL. PL peak shifts to the lower energy region when the temperature increases. The peaks A and B merge into one broad peak at temperatures above 100 K.

FIG. 5.

(a) PL spectra of WoW from 4 to 250 K and (b) temperature dependence of the PL. PL peak shifts to the lower energy region when the temperature increases. The peaks A and B merge into one broad peak at temperatures above 100 K.

Close modal

Figure 6 shows the observed PL spectra at 4 K for three samples with different WoW stacking numbers. Only the sample with 20 stacks has been discussed up to the present. However, to determine the origin of the observed PL peaks, it would be useful to compare the spectra of the samples with stacking numbers 10 and 40. In this figure, the vertical axis is drawn on a linear scale to understand the change in the intensities of the samples with different stacking numbers. When the stacking number increased, the intensity of peak A increased, and the peak energies shifted to lower energies. In contrast, the situation is different for peak B. The intensity decreased with increasing stacking numbers. The peak energy did not change. Figure 4 also shows the excitation power dependencies of the A peak for the samples with 10 (WoW10) and 40 stackings (WoW40) (dashed line blue curves).

FIG. 6.

The PL spectra of WoW samples with the SL stacked numbers 10, 20, and 40 at 4 K. Peak A increases and peak B decreases as the number of layers increases. A redshift of the peak energy is observed for peak A when the stacked number increases.

FIG. 6.

The PL spectra of WoW samples with the SL stacked numbers 10, 20, and 40 at 4 K. Peak A increases and peak B decreases as the number of layers increases. A redshift of the peak energy is observed for peak A when the stacked number increases.

Close modal

Band structure calculations were performed on the WoW and PSL structures. Regarding the PSL structure, band structures were easily calculated by Nextnano.32 The widths of the InGaAs QW, GaAs strain relaxation, and GaAsP barrier layers were set at 1.9, 2.6, and 2.9 nm, respectively. The number of SL stacks was 10 in the calculation. This is because no difference was observed in the number of stackings between 40 and 10. We used material parameters, such as the bandgap and effective mass, in the simulator databases. The calculated energy of the e1-hh1 transition of the PSL samples is 1.33 eV, which is well reproduced by the PR spectra at room temperature, as shown in Figs. 2 and 5(b). However, the model calculation is not easy for WoW samples. The QW width changes along the [ 1 ¯ 1 ¯ 0] direction, and the stacking sequence is complicated. Therefore, we used the simple model structure shown in Fig. 7(a) for the WoW sample. Only one layer of the WoW structure was considered, and the shape is triangular, as shown by the red dotted lines in the figure. The sum of the thicknesses of the InGaAs QW, GaAs strain-balanced, and GaAsP barrier layers was kept constant at 10 nm. The calculated carrier distributions, the existence probability that is the square of the wavefunction, in the e1, hh1, and lh1 quantized states are shown in Fig. 7(b). The bright region indicates higher carrier concentrations.

FIG. 7.

(a) Schematic of the structure of WoW sample projected in the ( 1 ¯ 10 ) plane for the theoretical band calculations using Nextnano. (b) Calculated strain within WoW and carrier existence probability distributions, the square of the wavefunction, of the first quantum level of electrons, heavy holes, and light holes. The proposed carrier recombination path is denoted using yellow arrow.

FIG. 7.

(a) Schematic of the structure of WoW sample projected in the ( 1 ¯ 10 ) plane for the theoretical band calculations using Nextnano. (b) Calculated strain within WoW and carrier existence probability distributions, the square of the wavefunction, of the first quantum level of electrons, heavy holes, and light holes. The proposed carrier recombination path is denoted using yellow arrow.

Close modal

Carriers accumulated in the inner part of the triangular wells. The electron energy at the right-hand side in the triangular-shaped QW is smaller because the quantized energy in the wider well-width region is low. The carriers may be concentrated in the wider-well region near the triangular vertex. The position of the maximum carrier distribution in the hh1 state is slightly displaced toward the narrower well-width region on the left-hand side, as shown in the figure. This may be attributed to the difference in the carrier effective masses of the electrons and heavy holes. The energy level changed more sensitively to the slight changes in the well width when the effective mass was smaller. It is, thus, more accessible to electrons in the broader well-width region. The hole distribution of the lh1 state, which is close to that of e1, supports this explanation.

Another possible reason is the strain induced in the QW. The calculated result of the elastic energy stored in the relevant position in the InGaAs of WoW using a Nextnano software calculator is shown in Fig. 7(b). The blue color region (wider well-width region) draws the more strained region. Since the lattice constant of the InGaAs is larger than that of GaAs, the compressive strain was induced at the interfaces. The larger compressive strain induced the larger bandgap energy (Eg).34 Therefore, the wider well-width region (right-hand side) becomes more strained, and Eg in this region increases compared with the narrower well-width region. The carriers redistribute to the left-hand side from the triangular edge. The effect of the strain and the well-width mentioned before determines the actual carrier distributions.

Next, we considered the reason of the formation of the peak A of the WoW samples, as shown in Fig. 3(a). The energies of this peak are 1.33 and 1.35 eV for the excitation power intensities of 500 and 2.5 × 104 W/m2, respectively. Accordingly, a blue shift is observed as the excitation power increases. As this was the lowest peak observed in the PL spectra, it is attributed to the electron transition from e1 to hh1 in the WoW structures. Theoretical prediction based on the Nextnano calculation suggests that if the structure is PSL, the 1.33 and 1.35 eV transitions correspond to the e1–hh1 transition of the uniform QW width of 5.7 and 5.1 nm, respectively. However, Nextnano's calculations show that the e1–hh1 transition in the present WoW structure occurs at 1.29 eV. In this case, the calculation of carrier distribution was conducted using a single set of triangular-shaped WoW structures, as shown in Fig. 7(a). No effect of the multi-stacked WoW structures was considered. The actual transition energy of WoW is affected by the carrier distribution and strain-induced distortion, which differ from those of the PSL structure owing to its two-dimensional structure, as depicted in Fig. 7(a). Therefore, we can conclude that the observed peak A is due to the e1–hh1 transition in the WoW samples.

The carrier transition process is discussed based on the probability of the existence of carriers. The highest carrier distribution probability in the triangular QW is not the same for e1 and hh1, as shown in Fig. 7(b). The most probable positions for the electrons and holes were assumed to be at the horizontal axes around 50 and 45 nm, respectively. Since the overlapping of the wavefunctions of electrons and holes determines the transition probability, less overlapping induces a low transition rate and a longer lifetime. This could explain the reported longer carrier lifetime of the WoW structure.

An interesting result in the present study is that peak A showed a blue shift with increasing excitation power, as shown in Figs. 3(a) and 4. The shift was 7 meV per digit. Then, we discuss the reasons for these blue shifts next. It is known that the PL emission peak by the donor-to-acceptor pair transition shifts to higher energy at increasing excitation power levels. This is due to the change in the distance between the paired donor and acceptor regarding the Coulombic interaction. However, we could not apply this idea to explain our results. The amount of the blue shift is reported to be on the order of 1.3 meV per digit for direct bandgap semiconductors, such as GaAs.35 The observed blue shift was too large. Of course, we cannot completely exclude the possibility that the donor and acceptor are formed in the QW structure. However, there may be no evidence that such donor–acceptor pairs were introduced in the present case.

As already mentioned, compressive strain causes the larger Eg on the right side of the triangle. This band tilt inclination is one of the reasons for the difference in the position of the maximum existing probability of electrons and holes. This spatial bandgap change might generate the electric field. This causes the electrons to be forced to the left and have their maximum probability at the position indicated in the calculation. Now consider the case where the carrier concentration increases due to increased excitation light intensity. It is possible that the increase in the number of carriers causes a screening effect on the electric field. When an electric field exists in the conventional QW, electrons and holes are separated in opposite directions.36 The transition energy under such an electric field is smaller than that without the field as known by the quantum confined Stark effect (QCSE).37 When the excitation light intensity increased, the electric field was screened by photogenerated carriers. Therefore, the QCSE weakens, which also weakens the narrowing of the bandgap, resulting in a blue shift of the PL peak.

We also discuss the origin of the observed PL peaks other than the peak A in the WoW samples in Fig. 3(a). For peak B at 1.39 eV, the peak energy did not change as a function of the excitation intensity. This is also the case for peak B′ (at 1.40 eV) in the PSL samples. We have already reported that residual PSL structures appear near the boundary between the WoW region and the buffer/substrate. Even when the substrate was tilted at approximately 6°, three or four layers formed in the quasi-PSL structure. This is also justified by the STEM image depicting 200 periods of In0.3Ga0.7As (2 nm)/GaAs (2.7 nm)/GaAs0.6P0.4 (3 nm)/GaAs (2.7 nm) layers grown on a GaAs (001) substrate with a tilt of 6° toward the [111] direction.26 Therefore, if we consider peak B to be due to these PSL components in the WoW sample near the substrate, the observed experimental results are understood. The difference in the peak energy of ∼0.015 eV between the observed peaks B and B′ may be due to the difference in the well widths of the two different sample structures. No peak shift with increasing excitation power intensity is also explained.

Another evidence that justifies the peak B is due to the quasi-PSL structure in the WoW sample is shown in Fig. 6. When the stacking number increased, the intensity of peak A increased, whereas that of peak B decreased. A few layers of quasi-PSL were considered to be formed for all samples near the boundary of the WoW and GaAs buffer layers. When the number of triangular WoW layers increased, the intensity of peak A increased. In contrast, the PL emission from the few quasi-PSL layers should remain constant whenever the stacking number increases. Because the energy of peak B is larger than that of peak A, reabsorption of the PL emission from the quasi-PSL layer occurs, and the intensity decreases. This was the case in Fig. 6. Accordingly, our explanation for peaks A and B is considered to be reasonable. The energies of peak A exhibited a red shift when the stacking number increased. The peak energies were 1.360, 1.353, and 1.349 eV for stacking numbers 10, 20, and 40, respectively. We considered that this difference was caused by the change in the quantum well dimension modified by the accumulation of the stacking of the WoW layers.

Carrier transition properties of InGaAs/GaAs/GaAsP SL solar cells of wire on well-type structures with high carrier collection efficiency were studied. A WoW structure was fabricated using the InGaAs/GaAs/GaAsP SL device fabrication technology. When the substrate was tilted by 6°, the thickness of the InGaAs layer was modulated periodically and continuously. The layer was formed in a geometry like a triangle. This structure modified the local concentration of carriers in the QW and increased the carrier lifetime. We investigated the carrier recombination properties using photoreflectance (PR) and photoluminescence (PL) measurements to elucidate the detailed carrier transition process. Non-tilted (PSL) samples were grown for comparison with WoW samples. Two typical electron transitions with energies equal to 1.34 (peak A) and 1.39 eV (peak B) were observed at 4 K in the PL spectra of the WoW samples. Temperature dependences of these peak energies and intensities were discussed, and we concluded that these are the transitions from the e1 to hh1 levels. We also discussed the carrier distribution in the WoW structure and found that the maximum carrier existing probabilities for e1 and hh1 are located at different positions. The less overlapping of the wavefunctions causes low transition probability and results in an observed long carrier lifetime of WoW.

When the laser excitation power increased, a blue shift was observed for peak A. No such shift was observed for peak B. To explain the reason for this blue shift, we considered the effect of induced strain at the interfaces of GaAs and InGaAs. Compressive strain causes the larger bandgap energy on the right side of the triangle. This band tilt inclination results in the difference in the position of the maximum existing probability of electrons and holes. The number of carriers screens the electric field when the carrier concentration increases due to increased excitation light intensity. Therefore, the quantum confined Stark effect (QCSE) weakens. This fact also hinders the bandgap narrowing, resulting in a blue shift of the PL peak energy.

This work was supported by JSPS KAKENHI under Grant Nos. 16H04648, 18K04876, and 21H01365.

The authors have no conflicts to disclose.

Shintaro Komaba: Conceptualization (lead); Data curation (lead); Formal analysis (lead); Investigation (supporting); Project administration (supporting); Validation (supporting); Visualization (lead); Writing – original draft (equal). Nana Taketa: Conceptualization (supporting); Data curation (supporting); Formal analysis (lead); Investigation (lead); Project administration (lead); Validation (lead); Visualization (supporting); Writing – review & editing (supporting). Meita Asami: Resources (lead); Supervision (supporting); Writing – review & editing (lead). Masakazu Sugiyama: Resources (supporting); Writing – review & editing (supporting). Tetsuo Ikari: Conceptualization (lead); Formal analysis (supporting); Funding acquisition (equal); Methodology (equal); Resources (supporting); Supervision (lead); Visualization (supporting); Writing – original draft (equal); Writing – review & editing (lead). Atsuhiko Fukuyama: Conceptualization (lead); Funding acquisition (equal); Methodology (equal); Resources (lead); Supervision (lead); Writing – review & editing (supporting).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1.
W.
Shockley
and
H. J.
Queisser
, “
Detailed balance limit of efficiency of p-n junction solar cells
,”
J. Appl. Phys.
32
(
3
),
510
519
(
1961
).
2.
T.
Takamoto
,
M.
Kaneiwa
,
M.
Imaizumi
, and
M.
Yamaguchi
, “
Ingap/GaAs-based multijunction solar cells
,”
Prog. Photovolt. Res. Appl.
13
(
6
),
495
511
(
2005
).
3.
R. R.
King
,
D. C.
Law
,
K. M.
Edmondson
,
C. M.
Fetzer
,
G. S.
Kinsey
,
H.
Yoon
,
R. A.
Sherif
, and
N. H.
Karam
, “
40% efficient metamorphic GaInP∕GaInAs∕Ge multijunction solar cells
,”
Appl. Phys. Lett.
90
(
18
),
183516
(
2007
).
4.
M.
Yamaguchi
,
T.
Takamoto
,
K.
Araki
, and
N.
Ekins-Daukes
, “
Multi-junction III–V solar cells: Current status and future potential
,”
Sol. Energy
79
(
1
),
78
85
(
2005
).
5.
M.
Yamaguchi
, “
Radiation-resistant solar cells for space use
,”
Sol. Energy Mater. Sol. Cells
68
(
1
),
31
53
(
2001
).
6.
T.
Ikari
,
S.
Fukushima
,
Y.
Ohta
,
A.
Fukuyama
,
S. D.
Wu
,
F.
Ishikawa
, and
M.
Kondow
, “
Experimental study of hydrostatic and shear deformation potential in Ga 1-y In y N x As 1-x alloys using a piezoelectric photothermal spectroscopy
,”
Phys. Rev. B
77
(
12
),
125311
(
2008
).
7.
I. A.
Buyanova
and
W. M.
Chen
,
Physics and Applications of Dilute Nitrides
, 1st ed. (
Taylor & Francis
,
New York
,
2004
).
8.
A.
Tukiainen
,
A.
Aho
,
G.
Gori
,
V.
Polojärvi
,
M.
Casale
,
E.
Greco
,
R.
Isoaho
,
T.
Aho
,
M.
Raappana
,
R.
Campesato
, and
M.
Guina
, “
High-efficiency GaInP/GaAs/GaInNAs solar cells grown by combined MBE-MOCVD technique
,”
Prog. Photovolt. Res. Appl.
24
(
7
),
914
919
(
2016
).
9.
K. W. J.
Barnham
and
G.
Duggan
, “
A new approach to high-efficiency multi-band-gap solar cells
,”
J. Appl. Phys.
67
(
7
),
3490
3493
(
1990
).
10.
Y.
Wang
,
R.
Onitsuka
,
M.
Deura
,
W.
Yu
,
M.
Sugiyama
, and
Y.
Nakano
, “
In situ reflectance monitoring for the MOVPE of strain-balanced InGaAs/GaAsP quantum-wells
,”
J. Cryst. Growth
312
(
8
),
1364
1369
(
2010
).
11.
N. J.
Ekins-Daukes
,
J. M.
Barnes
,
K. W. J.
Barnham
,
J. P.
Connolly
,
M.
Mazzer
,
J. C.
Clark
,
R.
Grey
,
G.
Hill
,
M. A.
Pate
, and
J. S.
Roberts
, “
Strained and strain-balanced quantum well devices for high-efficiency tandem solar cells
,”
Sol. Energy Mater. Sol. Cells
68
(
1
),
71
87
(
2001
).
12.
M.
Courel
,
J. C.
Rimada
, and
L.
Hernández
, “
An approach to high efficiencies using GaAs/GaInNAs multiple quantum well and superlattice solar cell
,”
J. Appl. Phys.
112
(
5
),
054511
(
2012
).
13.
T.
Aihara
,
A.
Fukuyama
,
Y.
Yokoyama
,
M.
Kojima
,
H.
Suzuki
,
M.
Sugiyama
,
Y.
Nakano
, and
T.
Ikari
, “
Detection of miniband formation in strain-balanced InGaAs/GaAsP quantum well solar cells by using a piezoelectric photothermal spectroscopy
,”
J. Appl. Phys.
116
(
4
),
044509
(
2014
).
14.
K.
Toprasertpong
,
H.
Fujii
,
Y.
Wang
,
K.
Watanabe
,
M.
Sugiyama
, and
Y.
Nakano
, “
Carrier escape time and temperature-dependent carrier collection efficiency of tunneling-enhanced multiple quantum well solar cells
,”
IEEE J. Photovolt.
4
(
2
),
607
613
(
2014
).
15.
H.
Shen
,
S. H.
Pan
,
F. H.
Pollak
,
M.
Dutta
, and
T. R.
AuCoin
, “
Conclusive evidence for miniband dispersion in the photoreflectance of a GaAs/Ga 0.74 Al 0.26 As coupled multiple-quantum-well structure
,”
Phys. Rev. B
36
(
17
),
9384
9387
(
1987
).
16.
Y. P.
Wang
,
S. J.
Ma
,
K.
Watanabe
,
M.
Sugiyama
, and
Y.
Nakano
, “
Management of highly-strained heterointerface in InGaAs/GaAsP strain-balanced superlattice for photovoltaic application
,”
J. Cryst. Growth
352
(
1
),
194
198
(
2012
).
17.
H.
Fujii
,
Y.
Wang
,
K.
Watanabe
,
M.
Sugiyama
, and
Y.
Nakano
, “
Suppressed lattice relaxation during InGaAs/GaAsP MQW growth with InGaAs and GaAs ultra-thin interlayers
,”
J. Cryst. Growth
352
(
1
),
239
244
(
2012
).
18.
H.
Fujii
,
K.
Toprasertpong
,
Y.
Wang
,
K.
Watanabe
,
M.
Sugiyama
, and
Y.
Nakano
, “
100-period, 1.23-eV bandgap InGaAs/GaAsP quantum wells for high-efficiency GaAs solar cells: Toward current-matched Ge-based tandem cells
,”
Prog. Photovolt. Res. Appl.
22
(
7
),
784
795
(
2014
).
19.
A.
Watanabe
,
T.
Ikari
,
R.
Furukawa
,
M.
Sugiyama
, and
A.
Fukuyama
, “
Reduction of non-radiative recombination by inserting a GaAs strain-relaxation interlayer in InGaAs/GaAsP superlattice solar cells investigated by photo-thermal spectroscopy
,”
J. Appl. Phys.
128
(
19
),
195702
(
2020
).
20.
T.
Nakamura
,
A.
Fukuyama
,
M.
Sugiyama
, and
T.
Ikari
, “
Effect of strain relaxation layer insertion on carrier recombination and escaping processes in superlattice solar cell structures using photoluminescence spectroscopy
,”
J. Phys. D: Appl. Phys.
52
(
4
),
045104
(
2019
).
21.
A.
Fukuyama
,
N.
Yamamoto
,
R.
Furukawa
,
M.
Sugiyama
, and
T.
Ikari
, “
Photothermal investigation for optimizing a lattice strain relaxation condition of InGaAs/GaAsP superlattice photovoltaic structures from a nonradiative transition point of view
,”
J. Phys. D: Appl. Phys.
56
(
4
),
045101
(
2023
).
22.
T.
Nakamura
,
A.
Fukuyama
,
M.
Sugiyama
, and
T.
Ikari
, “
Effects of barrier thickness on carrier non-radiative relaxation process in InGaAs/GaAsP superlattice solar cells by piezoelectric photothermal and surface photovoltage spectroscopies
,”
Jpn. J. Appl. Phys.
58
(
11
),
112001
(
2019
).
23.
H.
Fujii
,
T.
Katoh
,
K.
Toprasertpong
,
H.
Sodabanlu
,
K.
Watanabe
,
M.
Sugiyama
, and
Y.
Nakano
, “
Thickness-modulated InGaAs/GaAsP superlattice solar cells on vicinal substrates,
J. Appl. Phys.
117
(
15
),
154501
(
2015
).
24.
M.
Sugiyama
,
H.
Fujii
,
T.
Katoh
,
K.
Toprasertpong
,
H.
Sodabanlu
,
K.
Watanabe
,
D.
Alonso-Álvarez
,
N. J.
Ekins-Daukes
, and
Y.
Nakano
, “
Quantum wire-on-well (WoW) cell with long carrier lifetime for efficient carrier transport
,”
Prog. Photovolt. Res. Appl.
24
(
12
),
1606
1614
(
2016
).
25.
D.
Alonso-Álvarez
,
T.
Thomas
,
M.
Führer
,
N. P.
Hylton
,
N. J.
Ekins-Daukes
,
D.
Lackner
,
S. P.
Philipps
,
A. W.
Bett
,
H.
Sodabanlu
,
H.
Fujii
,
K.
Watanabe
,
M.
Sugiyama
,
L.
Nasi
, and
M.
Campanini
, “
Ingaas/GaAsP strain balanced multi-quantum wires grown on misoriented GaAs substrates for high efficiency solar cells
,”
Appl. Phys. Lett.
105
(
8
),
083124
(
2014
).
26.
H.
Cho
,
K.
Toprasertpong
,
H.
Sodabanlu
,
K.
Watanabe
,
M.
Sugiyama
, and
Y.
Nakano
, “
Stability and controllability of InGaAs/GaAsP wire-on-well (WoW) structure for multi-junction solar cells
,”
J. Cryst. Growth
464
,
86
93
(
2017
).
27.
H.
Fujii
,
H.
Sodabanlu
,
M.
Sugiyama
, and
Y.
Nakano
, “
Low-temperature MOVPE using TEGa for suppressed layer undulation in In x Ga 1−x As/GaAs 1−y P y superlattice on vicinal substrates
,”
J. Cryst. Growth
414
,
3
9
(
2015
).
28.
D. E.
Aspnes
, “
Third-derivative modulation spectroscopy with low-field electroreflectance
,”
Surf. Sci.
37
,
418
442
(
1973
).
29.
T. J. C.
Hosea
, “
Estimating critical-point parameters from Kramers-Kronig transformations of modulated reflectance spectra
,”
Phys. Status Solidi B
182
(
1
),
K43
K47
(
1994
).
30.
T. J. C.
Hosea
, “
Estimating critical-point parameters of modulated reflectance spectra
,”
Phys. Status Solidi B
189
(
2
),
531
542
(
1995
).
31.
I.
Vurgaftman
,
J. R.
Meyer
, and
L. R.
Ram-Mohan
, “
Band parameters for III–V compound semiconductors and their alloys
,”
J. Appl. Phys.
89
(
11
),
5815
5875
(
2001
).
32.
See http://www.nextnano.de for ver. 4.3.1.14 (Munich, Germany).
33.
H. B.
Bebb
and
E. W.
Williams
, “
Photoluminescence, chapters 4 and 5
,” in
Semicondocturs and Semimetals
, edited by
R. K.
Willardson
and
A. C.
Beer
(
Academic Press
,
New York
,
1972
), p.
8
.
34.
S. L.
Chuang
,
Physics of Optoelectronic Devices
(
John Willey and Sons Inc.
,
1995
).
35.
R. C. C.
Leite
and
A. E.
DiGiovanni
, “
Frequency shift with temperature as evidence for donor-acceptor pair recombination in relatively pure n-type GaAs
,”
Phys. Rev.
153/3
,
841
843
(
1967
).
36.
J. H.
Davis
,
The Physics of Low Dimensional Semiconductors
(
Cambridge University Press
,
1998
).
37.
D. A. B.
Miller
,
D. S.
Chemla
,
T. C.
Damen
,
A. C.
Gossard
,
W.
Wiegmann
,
T. H.
Wood
, and
C. A.
Burrus
, “
Band-Edge electroabsorption in quantum well structures: The quantum-confined stark effect
,”
Phys. Rev. Lett.
53
,
2173
(
1984
).