Two-step photon absorption in InP/InGaP quantum dot solar cells

First conceptualized by Martí et al.,¹ the concept of intermediate band solar cells (IBSCs) promises a drastic increase in single-junction efficiency (theoretically over 60% under concentrated sunlight²) exceeding theoretical predictions made by the Shockley-Queisser limit.³ In this concept, an intermediate band (IB) is formed within the forbidden gap of the host material, which acts as an additional energy level where photons with energy less than the host bandgap energy can be absorbed. Consequently, a second photon can excite the accumulated carriers from the IB to the conduction band (CB). This greatly reduces the transmission loss of single junction solar cells. Several methods exist to form a semi-discrete energy level within the band-gap, such as utilizing quantum dots (QDs), growing highly mismatched alloys, or introducing deep level defects in bulk materials (impurity bands).^4–12 So far, utilizing InAs and GaSb QDs has been the most studied for IB formation within single junction GaAs and/or AlGaAs solar cells. Unambiguous observation of two-step photon absorption (TSPA) has been observed in these systems at temperatures exceeding 200 K.¹³ InP QDs in an In_xGa_1-xP host lattice matched to GaAs, however, is another promising material system. Several advantages of this system warrant further study, such as In_0.49Ga_0.51P (hereafter just InGaP) having close to the ideal host bandgap of ∼1.9 eV for IBSC as well as being one of the most radiation resistant materials for space applications.^14,15 With QDs adding to the radiation hardness,^16,17 InP QD/InGaP IBSC could potentially be a viable replacement for expensive triple-junction solar cells used in space. As proposed by Tayagaki and Sugaya, InP QDs with type-II band alignment create a more ideal energy profile for TSPA than InAs or GaSb QDs, forming a 300 meV well in the CB and a negligibly small 5 meV barrier in the valence band (VB).¹⁸ This band structure leads to an increased lifetime of the carriers within the InP QDs, giving them sufficient time to absorb low energy photons and be collected as photocurrent rather than being lost to thermal emission or recombination. In this study, InP QD/InGaP IBSC grown via metal-organic chemical vapor deposition (MOCVD) with TSPA response up to 250 K is demonstrated, one of the highest observations reported to date using QD systems.

Figure 1(a) shows the IBSC concept in schematic form. There are three significant energy levels in this concept: E_G, E_H, and E_L, where E_G is the bandgap energy of the host material, E_H is the IB energy level with respect to the VB, and E_L is the energy level difference between the CB and IB. Thus, three energy levels exist (VB to CB, VB to IB, and IB to CB) for photon absorption. The VB to CB transition is the typical single-junction absorption path, whereas the VB to IB and IB to CB transitions are formed through the insertion of the IB energy state. Figure 1(b) shows a simulated (6-band k·p method) type-II InP QD/InGaP band energy profile (cut through the growth direction) of a single InP QD (diameter and height of 40 nm and 15 nm, respectively) buried in InGaP at 300 K and is in good agreement with simulated results reported previously by Pryor et al.¹⁹ on the same material system. The E_G, E_H, and E_L energy levels in this system are predicted to be 1.9, 1.6, and 0.3 eV, respectively, with confinement for electrons only. Although E_L of 0.3 eV is not the most ideal (ideal is E_L of 0.7 eV) for IBSC, it is deep enough to prevent significant thermal escape at higher temperatures as evidenced by the experimental results presented below.

Epitaxial growth was carried out in an Aixtron close-coupled showerhead metal-organic chemical vapor deposition system (CCS-MOCVD) using trimethylgallium (TMGa), trimethylindium (TMIn), arsine (AsH₃), and phosphine (PH₃) precursors. Disilane (Si₂H₆) and diethylzinc (DEZn) were used for n-type and p-type doping, respectively. Zn-doped GaAs with a 2° offcut towards the (110) was used as the substrate. To optimize the QD growth conditions, multiple InP QD layers were formed on top of a 50 nm InGaP layer grown on a 100 nm undoped GaAs buffer layer. Growth was done at a temperature of 650 °C and growth rates of 2 μm/h for the GaAs buffer and 1 μm/h for the InGaP layer, respectively. The temperature was ramped down to 530 °C during QD growth at a growth rate of approximately 0.1 ML/s for a total nominal thickness of 5.0 MLs. The substrate temperature and growth rates were extracted from an in-situ LayTec EpiTT monitor. The initial QD growth conditions were based on the values reported in the literature,^20–23 and a detailed table of growth conditions is given in supplementary material Table S1. The geometrical and optical characteristics of the surface InP dots are shown in Fig. 2. Figure 2(a) shows a tilt-view 3-dimensional atomic force microscopy (AFM) image, with a scan area of 2 μm × 2 μm. Dimensional analysis using commercial scanning probe image processing software (SPIP) indicates the formation of dimensionally homogeneous QDs, with the dot density, diameter, and height of 0.7 × 10¹⁰ cm⁻², 56 ± 10 nm, and 18 ± 2.8 nm, respectively, as shown in supplementary material Fig. S1. According to Pryor et al., the height of the QD is important in forming type-II band alignment on InGaP, which is driven by strain and must be taller than ∼15 nm.¹⁹ The composition of our InGaP has been verified by high resolution x-ray diffraction (HRXRD). As can be seen, the InGaP is lattice matched to the GaAs substrate, with the In composition to be around 49% with a lattice constant of 5.63 Å, leading to a lattice mismatch between InP and InGaP to be around 3.7%. The relatively smaller lattice mismatch results in larger QDs compared to InAs QDs on GaAs, which has a lattice mismatch of nearly 7%. The symmetrical (004) reciprocal space map (RSM) of the five-layer InP QD samples (supplementary material Fig. S2) shows that the InP QD superlattice (SL) is indeed compressively strained by 1496 ppm. The strain was calculated by fitting the superlattice peaks SL(0), SL(-1), and SL(1) of the 2θ-ω scan obtained from integrating the RSM scan (supplementary material Fig. S3). Insertion of optimized strain relief layers, such as InGaAsP at the end of each SL period, may help reduce the strain by applying a tensile stress to balance the compressive stress caused by the QD layer.²⁴ 

To verify the type-II band alignment of the grown dots, optical characterization of the test structures was performed. First, the photoluminescence (PL) wavelength of the type-II QDs with those reported in the literature was compared. The QD sample was mounted in a closed cycle He cryostat, excited by a 532 nm (2.33 eV) diode laser and detected by a thermoelectrically cooled Si diode detector. Several reports agree upon a PL wavelength of approximately 770–800 nm at room temperature,^18,20,21,25 matching the 300 K PL spectra shown in Fig. 2(b). Clear PL peaks of the InP QD (∼800 nm), GaAs buffer (∼875 nm), and InGaP spacer (∼685 nm) can be seen. This energy also agrees with the simulated QD transition (InP QD ground state to the valence band) shown in Fig. 1(b). Temperature dependent PL spectra (supplementary material Fig. S4) show a blue-shift of the QD peak from 800 nm at 300 K to 750 nm at 50 K, showing good agreement with previous reports in this temperature range.^18,26 Even at low temperatures, the PL signal from type-I InP QDs was not detected, which emits at a higher energy of ∼1.7 eV (730 nm) at 100 K,¹⁸ attesting to the good homogeneity of type-II QDs. The full width at half maximum (FWHM) of the QD peak is approximately 40 meV at 40 K, indicating narrow distribution in size and, therefore, reduced inhomogeneous broadening of the PL emission. Additionally, the excitation power dependent PL peak shift was measured at low temperatures. A blue-shift in the PL peak wavelength has been observed by several groups in type-II confinement structures with increasing laser excitation at low temperatures, owing to the band bending near the QD/host matrix interface with an increase in excited carriers.²⁷ The shift in the wavelength is typically known to be linearly dependent on the third root of the excitation laser power,²⁸ which is observed in the QD PL peak, as shown in Fig. 2(c). Additionally, the time-resolved PL (TRPL) response of the InP QD sample was measured at 15 K using a 405 nm GaN laser for excitation and a streak camera to measure the PL decay profile. The decay profile [Fig. 2(d)] shows an initial fast decay (∼6–17 ns) component followed by a very long (∼230–250 ns) decay component in the energy range of 1.57–1.63 eV for one and five layers of InP QDs. The slow decay component, quite long compared to other type-II QD systems, clearly indicates the existence of type-II confinement due to the spatial separation of electrons and holes. The initial fast decay component remained even at low excitation, ruling out the possibility of it being due to Auger recombination. It could be caused by degraded interfacial characteristics due to non-optimized InGaP spacer growth conditions, as a rough emitter surface was observed after growth of several QD layers. No PL from the wetting layer was measured. Due to the thick 50 nm InGaP spacer layer between each QD layer, no quantum mechanical coupling between adjacent QD layers or SL effects are expected in these structures. Not only is the thickness of the spacer InGaP too thick for any meaningful quantum mechanical coupling but also a large red-shift in the wavelength and a decrease in the spectral linewidth have been observed for vertically coupled systems but were not observed for the structure grown in this work (supplementary material Fig. S5).^29,30

Several 1 × 1 cm² solar cells were grown and fabricated using standard III-V wet etch, lithography, and metallization techniques, consisting of (a) reference InGaP solar cell, (b) 3-layer InP QD solar cell, and (c) 5-layer InP QD solar cell. A detailed device structure with doping and growth conditions is given in supplementary material Table S1. A double layer ZnS/MgF anti-reflection coating (ARC) was deposited on all samples before the measurement. External quantum efficiency (EQE) was measured for each cell using a Newport IQE 200 system, as shown in Fig. 3(a). A general decrease in EQE for both QD samples is observed over all relevant wavelength ranges at energies greater than the bandgap, with an increase in sub-bandgap EQE due to the InP QDs. The decreased EQE likely indicates a reduction in the material quality of the emitter region or is caused by carriers being trapped by the InP QDs during the transit of the intrinsic region. A fitting of the EQE using a modified form of the drift-diffusion equation³¹ for the reference and five-layer InP QD sample is given in supplementary material Fig. S6. As expected, the InP QD device shows a reduced minority carrier diffusion length in the emitter region compared to the reference cell by approximately a factor of 3 (0.85 μm vs. 0.25 μm), which corresponds to the increased strain observed from the HRXRD analysis.

To better understand the carrier transport within each region of the solar cell, the carrier collection efficiency (CCE) for the InP QD cell was measured. The CCE is an effective parameter to probe the efficiency of carrier transport at various depths of the solar cell by varying the photon wavelength.³² Shorter wavelengths are entirely absorbed in the emitter region, while longer wavelengths penetrate deeper into the solar cell following the Beer-Lambert law. In general, CCE can be enhanced by applying a reverse bias voltage to the solar cell diode. As the reverse bias voltage increases, a strong electric field facilitates carrier transport through the cell even for carriers created in the base. The ratio between the saturated EQE and EQE at specific bias voltages can be defined as the carrier collection efficiency, CCE = EQE(V, λ)/EQE_sat(λ), where V and λ are the bias voltage and wavelength, respectively. At high enough reverse bias voltages, the collection efficiency saturates. As shown in the inset of Fig. 3(a), there is a larger drop in CCE for the shorter wavelength at zero bias voltage with respect to the saturated CCE bias of −4 V. As the wavelength increases, the photons have a higher probability of creating carriers deeper into the cell (i.e., in the base region). As expected, the CCE near the base, where the growth quality is unaffected by the QD SL, is higher than the CCE near the emitter region.

Current-voltage curves at AM0 illumination were measured on all three cell types using a solar simulator equipped with tungsten halogen and metal halide arc lamps (2-zone TS Space Systems). The open-circuit voltage (V_OC), short-circuit current density (J_SC), efficiency, and fill factor for each cell are listed in Table I. As can be seen in Fig. 3(b), a decrease in both J_SC and V_OC is observed for the InP QD cells compared to the reference cell. Consistent with the EQE fitting (supplementary material Fig. S5) and CCE analysis discussed previously, the reduction in J_SC and V_OC compared to the reference cell may be attributed to degradation in the emitter growth quality after QD formation and reduced chance of electrons generated in the base region to be collected due to recombination in the QD region. Optimization in the emitter region growth condition, the use of strain balancing techniques,³³ and doping the QDs may improve the performance.

Finally, TSPA characteristics were measured using two light sources, a tungsten lamp through a monochromator and a 1.3 μm (0.95 eV) laser with a power density of 100 mW/cm², to excite carriers from the VB to IB and then from the IB to CB, respectively, as shown in Fig. 4(a). Note that the 1.3 μm laser cannot produce any photocurrent unless there are carriers present in the IB state. In this setup, photons with energy larger than the bandgap of InGaP (i.e., E_G ∼1.9 eV) excite electrons from the VB to the CB, which is efficiently absorbed and collected by the n-type emitter region contact. Once the photon energy drops below E_G, the monochromatic light populates the IB energy state, which is subsequently excited by the 1.3 μm laser into the CB and collected as photocurrent. Figure 4(b) shows the temperature dependence of the photocurrent generated by the 1.3 μm laser, chopped at a frequency of 166 Hz and measured via a lock-in amplifier. TSPA photocurrent is generated at a photon energy of ∼1.9 eV and persists until ∼1.6 eV. A slight tail is observed below E_H, which is likely due to the non-uniformity of the QD dimensions. TSPA photocurrent is clearly measured up to 250 K, and the onset wavelength at which TSPA is measured roughly matches the E_G dependence of InGaP as a function of temperature (estimated by the temperature dependent photocurrent measurement shown in supplementary material Fig. S7). TSPA photocurrent as a function of the 1.3 μm laser is shown in Fig. 4(c). A clear decrease in TSPA current is observed with decreasing laser power, with no current measured at a power of 0 mW. This verifies that the measured current is solely due to carriers pumped by the 1.3 μm laser from the IB to the CB. No TSPA signal was detected on the reference sample without the InP QDs, verifying that the QDs indeed form a sub-band energy level within the forbidden gap, allowing absorption of photons with below the bandgap energy of the InGaP host.

In conclusion, high temperature TSPA up to 250 K is demonstrated on InP QD/InGaP solar cells grown by MOCVD. PL measurements indicate homogeneous growth of type-II InP QDs with long carrier lifetimes. These results clearly indicate that InP QD in the InGaP host is suitable for high temperature IBSC and could be a promising candidate for space photovoltaic applications.

See supplementary material for a detailed structure of the InP QD solar cell, additional InP QD characterizations, EQE analysis, and temperature dependent photocurrent results.

The authors would like to acknowledge the U.S. Air Force Research Laboratory (Grant No. STTR FA9453-15-C-0404) for their support. The work in AIST was supported by the New Energy and Industrial Technology Development Organization (NEDO) under the Ministry of Economy, Trade and Industry (METI).