74.7% Efficient GaAs-Based Laser Power Converters at 808 nm at 150 K

Abstract

High-efficiency multijunction laser power converters are demonstrated for low temperature applications with an optical input at 808 nm. The photovoltaic power converting III-V semiconductor devices are designed with GaAs absorbing layers, here with 5 thin subcells (PT5), connected by transparent tunnel junctions. Unprecedented conversion efficiencies of up to 74.7% are measured at temperatures around 150 K. At temperatures around 77 K, a remarkably low bandgap offset value of W_oc = 71 mV is obtained at an optical input intensity of ~7 W/cm². At 77 K, the PT5 retains an efficiency of 65% with up to 0.3 W of converted output power.

1. Introduction

Impressive laser Optical Power Converters (OPCs) results have been obtained for various wavelength ranges and output power capabilities [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47]. Our recent Power Converter Performance Chart [41,47] clearly highlights that multijunction OPCs are most advantageous to obtain high device performance. The research related to photovoltaic devices also suggests other potential future device improvements [48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67], as well as new optical wireless power transmission (OWPT) applications and design strategies for future systems [68,69,70,71,72,73,74,75,76,77,78].

In typical cases, the OPC devices are expected to operate near room temperature, or more generally to function efficiently in a range between −40 °C and 85 °C. However, space applications or scientific experiments requiring cryogenic temperatures [79] could benefit from OPC devices capable of high-performance at temperatures as low as 77 K.

From the laser diode perspective, many of the OPC developments have historically been achieved at wavelengths around 808 nm due to the ubiquity and the maturity of GaAs-based lasers at this wavelength. An input wavelength at 808 nm will actually be close to optimal for OPCs designed with GaAs absorbing layers as the band-edge shifts to a wavelength of 822 nm at 77 K. Indeed, having the optical input wavelength only at ~14 nm (~26 meV) above the edge of the semiconductor bandgap is advantageous for minimizing the photocarrier thermalization losses and for optimizing the OPC’s efficiency.

In this study, we therefore measure the characteristics of vertical multijunction OPCs with 5 thin GaAs subcells (PT5) at cryogenic temperatures, with an input wavelength at 808 nm. The temperature dependence of the key OPC parameters is measured for a PT5 based on the Vertical Epitaxial Heterostructure Architecture (VEHSA) design [57]. This study focuses on a PT5 design which was originally intended for operating near room temperature at a input wavelength of near 850 nm. This design is also expected to be near ideal at low temperatures due to the GaAs bandgap shift combined with the swap of the optical input from 850 nm to 808 nm.

2. Materials and Methods

The schematic of the PT5 heterostructure is depicted in Figure 1a. It is based on the previously described VEHSA design [57,58]. The Beer-Lambert law was used to calculate the individual subcell’s absorber thicknesses, here with each subcell absorbing ~ 1/5 of the incident light. Specifically, the thicknesses for the GaAs absorber subcells used here were 192, 246, 346, 581, and 2636 nm, from top to bottom, respectively. It should be noted that methods have now also been put forward based on machine learning and genetic algorithm for improving the optimization of the individual thicknesses [80]. Such a model suggests for example that the evaluation using the Beer-Lambert model might be underestimating the required thickness of the first subcell.

The photovoltaic vertical multijunction structure was built for operation at T ~ 20 °C with the optical input from a laser source emitting in a spectral range peaking around 850 nm. The predominant change in the GaAs absorption coefficient between 20 °C and 77 K is a shift towards shorter wavelengths. Therefore, we expect the same design will be near-optimal for an optical input wavelength of 808 nm at cryogenic temperatures. The PT5 is designed with 5 optically transparent photovoltaic semiconductor subcells interconnected with tunnel junctions, labelled TJi in Figure 1a. Each individual subcell comprises an n-type emitter and a p-type base. The TJs are made to be transparent to the input beam, utilizing AlGaAs-AlInGaP alloys lattice-matched to GaAs.

The underlying TJ’s current-voltage (I-V) characteristics can be deduced from PT5 I-V characteristics taken under optical input intensities that are exceeding the TJ’s peak current capability. For example, Figure 1b shows the PT5 I-V curve rescaled to distinguish the TJ characteristic more clearly. The horizontal axis plots Voc—V, whereas the vertical axis is inverted compared to the plot of a typical p/n junction I-V curve. This approach reveals the negative differential resistance (NRD) region of one of the TJs (NDR observed for 0.3 V < V < 0.9 V), thus confirming the tunnel current characteristics at 77 K. As previously observed when the NDR arises, the illuminated I-V is discontinuous for that region. This particular PT5 structure had a peak tunneling current capabilities of the order of A/cm², but we are confident that designs with significantly higher peak current capability are readily achievable [57].

The epitaxial layers are grown using commercial production Aixtron Metal Organic Chemical Vapor Deposition (MOCVD) reactors. The total thickness of all the emitter and base layers from the different subcells is such that the impinging optical beam is almost completely absorbed for the condition of 850 nm at 20 °C, or similarly for 808 nm at 77 K. As described previously [41,58], to realize the required photocurrent matching condition, the structure usually has increasing subcell thicknesses from the top subcell (thinnest) toward the bottom subcell (thickest). Furthermore, the vertical multijunction devices can also benefit from strong photon coupling and recycling within, and between, the constituent subcells [81,82,83,84]. Potentially the photon coupling and recycling effects could be even more significant at low temperatures, in which case the radiative recombination is typically prominent.

The PT5 wafers were fabricated into chips with an area of 0.03 cm². The device fabrication included standard blanket back-metallization, front ohmic contacts, and antireflection coatings (ARC) constructed from layers of Al₂O₃ and TiO₂. The ARC typically reduces the reflectivity (R) of the incident beam to R < 4% for the spectral range of interest.

A 808 nm fiber-coupled laser diode manufactured by BWT was used [85]. It had a numerical aperture of NA ~0.22, using a multi-mode fiber core diameter of 400 μm and cladding of 440 μm. The PT5 devices were packaged in Broadcom’s regular power housing equipped with an FC optical connector [37]. The I-V characteristics were acquired using a Keithley 2601B source-meter. For most of the I-V measurements, the fiber-coupled laser was connected to the packaged PT5 using an FC connector. The FC connection was further sealed using a Kapton tape and the device was immersed in liquid nitrogen for the 77 K measurements. For variable temperature measurements, either the device was let to warm up after the liquid nitrogen was all evaporated, or a liquid nitrogen cryostat was used equipped with a standard 1 kOhms resistive temperature device (RTD) to directly measure the device temperature. Quick I-V scans were used to avoid significant chip heating or temperature drifts between the measurements.

3. Results

The PT5 characteristics at 77 K are shown in Figure 2. The measured I-V curves are shown in Figure 2a for various optical input powers between P_in = 96 mW and P_in = 372 mW. The dashed (pink) curve of Figure 2a is an ideal diode model fitted to the 96 mW data. A good fit is obtained, here using 5 diodes all with the same ideality factor of n = 1.6 and a quantum efficiency of EQE = 79%. The fitted photocurrent ratios for the 5 subcells (from bottom to top) are respectively 100%, 99%, 98%, 97%, and 96%. The fit reproduces well the data when the overall series resistance is set to be smaller than ~0.1 Ohm.

Figure 2b shows that the output power P_mpp has a measured slope efficiency of Eff ∼65% at 77 K, with negligible deviations from a linear regression for optical input powers up to ~0.5 W. Here, for this particular PT5, the input power was limited by the tunnel junction peak current capability. Based on our other manufacturing data and from alternative TJ designs tested at low temperatures, we expect future PT5 runs will have input power capabilities about an order of magnitude higher.

Remarkably, for example for the 372 mW curve of Figure 2a, the open-circuit voltage (V_oc) reaches a value of V_oc = 7.184 V, while the maximum power point voltage (V_mpp) is then 6.875 V. It corresponds to an average voltage of 1.437 V per subcell, yielding a bandgap voltage offset value of W_oc = 0.071 V, where W_oc = (Eg/q) − V_oc with Eg being the bandgap energy (here, 1.508 eV for GaAs at 77 K) and q is the electronic charge. The W_oc values obtained with the PT5 at low temperatures are therefore significantly better than the best values obtained at room temperature for GaAs with W_oc (20 °C) = 0.181 V [41], and also for the long wavelength PT10-InGaAs/InP OPCs with W_oc (20 °C) = 0.187 V [47].

The ideal diode model of Figure 2a can also be used to further explore better optimized, but realistic, conditions: a conversion efficiency of Eff ~ 75% would require improving the EQE to ~87%, while keeping W_oc and the diode ideality factors the same. An EQE of 93% would yield an Eff ~80% at 77 K. Such EQE improvements will most likely be achieved by further reducing the mismatch in the subcell’s photocurrents. Increasing the optical input intensities in the tens of W/cm² could also help to increase the efficiency if the W_oc value can be further reduced under higher optical intensities.

The temperature dependence of the PT5 properties are analyzed in more details in Figure 3. The output voltage is shown in Figure 3a for a 1 cm² chip mounted into a cryostat equipped with a 1 kOhm RTD and used to directly assess the OPC’s temperature while the device is warming up from liquid nitrogen temperature to room temperature. The open-circuit voltage (V_oc) is decreased only by few millivolts between liquid nitrogen temperature and about 145 K. The V_oc of this larger chip is lower than the V_oc obtained for the smaller chip of Figure 2, predominantly because of the relatively low optical intensity used for Figure 3a. More importantly, for the range between 175 K and up to above room temperature, the V_oc varies linearly with temperature with a slope of −7 mV/K, as shown from the linear regression in Figure 3a. This temperature coefficient can then be used to calibrate the device temperatures from the measured V_oc. For example, Figure 3b shows the measured efficiency as a function of the measured V_oc for the PT5 of Figure 2 with an optical input of 100 mW at 808 nm (blue curve). The PT5 parameters are extracted from the full I-V curves taken while the PT5 is warming up from 77 K to room temperature. The PT5 efficiency clearly increases as the device warms up from 77 K and reaches a maximum value of Eff ~72.4% when the V_oc reaches 6.7 V. Similar temperature dependence data, as in Figure 3b, were measured for other optical input powers. Input-power-adjusted V_oc temperature coefficients (e.g., Figure 3a) are then used to plot the temperature dependence of the conversion efficiency, as shown in Figure 4 for an optical input power of 100 mW and 193 mW (black curve). A maximum conversion efficiency of Eff = 74.7% is here measured for this PT5 at a temperature of about 150 K at P_in = 193 mW.

Applying again the ideal diode model to the 150 K I-V curves, we evaluate that for an optical input intensity of ~62 W/cm², such PT5 would be expected to have a W_oc value of 57 mV and an efficiency of Eff ~ 77.7% (from the diode model, not shown). The latter evaluation is therefore based on the experimental results at lower optical input combined with the diode model projected at higher optical intensities. We expect future runs will achieve these conditions.

4. Discussion

The highest output voltage values are measured at the lowest temperatures. This is expected from the temperature dependence of the bandgap. However, as shown in in Figure 4, the conversion efficiency of the PT5 peaks at intermediate temperatures. Record efficiencies of Eff > 70% are obtained for temperatures between about 130 K and 180 K. This optimum in performance is obtained because the best current matching conditions, for this specific layer design, are realized for that temperature range. For example, Figure 5 shows the temperature dependence of the output voltage in Figure 5a, of the external quantum efficiency measured at the maximum power point (EQE at mpp) in Figure 5b, and of the measured bandgap offset (W_oc) in Figure 5c.

As can be observed in Figure 5b, at an input wavelength of 808 nm, a good current matching condition is indeed obtained in the range of 150 K < T < 180 K, with EQE values of about 90% at the maximum power point. Further incremental improvements could be obtained by insuring a better current matching in all the subcells simultaneously, minimizing the device reflectivity, and minimizing the gridline shadowing. At lower temperatures the EQE decreases to ~79%, as was also deduced from the ideal diode model of Figure 2a. The EQE also decreases at higher temperatures, reaching about 70% at room temperature, because this particular PT5 was design for an optical input around 850 nm at 20 °C. It can therefore be deduced that at 808 nm at the warmer temperatures, the upper subcells are somewhat overdriven and the bottom subcells are current-starved. Conversely, near 77 K, the upper subcells would be generating less photocurrent relative to the lower subcells.

The V_oc data of Figure 5a is used, with the GaAs energy gap calculated using the Varshni equation [86], to extract W_oc in Figure 5c. W_oc increases linearly above 130K with a slope of 1.2 mV/K for each individual subcells. This can be explained by the shift of the Fermi levels in the n-type and p-type side of each p/n junction with temperature. An estimate suggests this change should be ~0.3 mV/K for each GaAs junction. For temperatures below 130 K, record W_oc values in the range of 75 mV are obtained with negligible temperature dependence. The flatter temperature dependence of W_oc below 130 K may be caused by changes in the density of states of GaAs affecting the Fermi levels and also from current mismatch in the individual cells affecting slightly the measured V_oc. In previous temperature studies of GaAs solar cells [87,88,89,90,91], Philipps et al. have attributed the decrease of the open-circuit voltage with increasing temperature to the temperature dependence of the effective density of states, the charge carrier densities and the band parameter of GaAs [87].

5. Conclusions

In conclusion, record efficiencies have been demonstrated at cryogenic temperatures with the vertical multijunction VEHSA optical power converter. Conversion efficiencies of Eff ~75% have been measured for an input wavelength of 808 nm with the PT5 OPCs. For the specific PT5 design studied, the maximum power point EQE values are reaching ~90% for an optimal temperature range of 150 K < T < 180 K. Record bandgap offset values (W_oc) have been obtained, with W_oc as low as 71 mV for temperatures below 130 K. The measured data and the corresponding ideal diode multijunction model suggest that multijunction GaAs OPCs for cryogenic operation with conversion efficiencies in the range of 75% to 80% are expected to be realistic with further optimization. The study confirms that OPC devices designed for high performance at cryogenic temperature with an optical input at 808nm should be expected to also operate at high conversion efficiencies at, or near, room temperature using an optical input at 850 nm. These results are expected to be of benefit for a number of experiments requiring electrical isolation under cryogenic conditions, for example conducted in liquid argon [79,92].