Triple and Quadruple Junctions Thermophotovoltaic Devices Lattice Matched to InP
Thermophotovoltaic (TPV) conversion of IR radiation emanating from a radioisotope heat source is under consideration for deep space exploration. Ideally, for radiator temperatures of interest, the TPV cell must convert efficiently photons in the 0.4-0.7 eV spectral range. Best experimental data for single junction cells are obtained for lattice-mismatched 0.55 eV InGaAs based devices. It was suggested, that a tandem InGaAs based TPV cell made by monolithically combining two or more lattice mismatched InGaAs subcells on InP would result in a sizeable efficiency improvement. However, from a practical standpoint the implementation of more than two subcells with lattice mismatch systems will require extremely thick graded layers (defect filtering systems) to accommodate the lattice mismatch between the sub-cells and could detrimentally affect the recycling of the unused IR energy to the emitter. A buffer structure, consisting of various InPAs layers, is incorporated to accommodate the lattice mismatch between the high and low bandgap subcells. There are evidences that the presence of the buffer structure may generate defects, which could extend down to the underlying InGaAs layer. The unusual large band gap lowering observed in GaAs(1-x)N(x) with low nitrogen fraction [1] has sparked a new interest in the development of dilute nitrogen containing III-V semiconductors for long-wavelength optoelectronic devices (e.g. IR lasers, detector, solar cells) [2-7]. Lattice matched Ga1-yInyNxAs1-x on InP has recently been investigated for the potential use in the mid-infrared device applications [8], and it could be a strong candidate for the applications in TPV devices. This novel quaternary alloy allows the tuning of the band gap from 1.42 eV to below 1 eV on GaAs and band gap as low as 0.6eV when strained to InP, but it has its own limitations. To achieve such a low band gap using the quaternary Ga1-yInyNxAs1-x, either it needs to be strained on InP, which creates further complications due to the creation of defects and short life of the device or to introduce high content of indium, which again is found problematic due to the difficulties in diluting nitrogen in the presence of high indium [9]. An availability of material of proper band gap and lattice matching on InP are important issues for the development of TPV devices to perform better. To address those issues, recently we have shown that by adjusting the thickness of individual sublayers and the nitrogen composition, strain balanced GaAs(1-x)N(x)/InAs(1-y)N(y) superlattice can be designed to be both lattice matched to InP and have an effective bandgap in the desirable 0.4- 0.7eV range [10,11]. Theoretically the already reduced band gap of GaAs(1-x)N(x), due to the nitrogen effects, can be further reduced by subjecting it to a biaxial tensile strain, for example, by fabricating pseudomorphically strained layers on commonly available InP substrates. While such an approach in principle could allow access to smaller band gap (longer wavelength), only a few atomic monolayers of the material can be grown due to the large lattice mismatch between GaAs(1-x)N(x) and InP (approx.3.8-4.8 % for x<0.05, 300K). This limitation can be avoided using the principle of strain balancing [12], by introducing the alternating layers of InAs(1-y)N(y) with opposite strain (approx.2.4-3.1% for x<0.05, 300K) in combination with GaAs(1-x)N(x). Therefore, even an infinite pseudomorphically strained superlattice thickness can be realized from a sequence of GaAs(1-x)N(x) and InAs(1-y)N(y) layers if the thickness of each layer is kept below the threshold for its lattice relaxation