07-10-2016, 11:17 AM
1458151252-073428761.docx (Size: 458.96 KB / Downloads: 6)
Abstract—The highest solar cell conversion efficiencies are achieved with four-junction devices under concentrated sunlight illumination. Different cell architectures are under development, all targeting an ideal bandgap combination close to 1.9, 1.4, 1.0, and 0.7 eV. Wafer bonding is used in this work to combine materi-als with a significant lattice mismatch. Three cell architectures are presented using the same two top junctions of GaInP/GaAs but dif-ferent infrared absorbers based on Germanium, GaSb, or GaInAs on InP. The modeled efficiency potential at 500 suns is in the range of 49–54% for all three devices, but the highest efficiency is expected for the InP-based cell. An efficiency of 46% at 508 suns was already measured by AIST in Japan for a GaInP/GaAs//GaInAsP/GaInAs solar cell and represents the highest independently confirmed effi-ciency today. Solar cells on Ge and GaSb are in the development phase at Fraunhofer ISE, and the first demonstration of functional devices is presented in this paper.
A. Theoretical Modeling
Theoretical calculation were performed using a self-built code for calculating the absorption in each layer of a four-junction cell, using the transfer matrix method [17]. It was assumed that light is coherent throughout the whole device structure. To achieve a proper optical termination, the substrate was defined to have an infinite thickness. Absorption coefficients for the materials were taken from [18]–[22], as well as from in-house measurements at Fraunhofer ISE. For unknown compositions, the data were interpolated from the nearest available neigh-bors. A morphing algorithm was used for the interpolation of n(λ) and k(λ) data, which takes into account critical energy points where the slope of the dielectric functions changes signifi-cantly. Linear interpolation with composition was performed be-tween these critical energy endpoints. The interpolation method was tested for Alx Ga1-x As compounds and found to lead to excellent agreement with experimental results. The dielectric function of ternary and quaternary III–V compounds was then determined by using this morphing algorithm, and the transfer matrix method was used to calculate the expected absorption in each solar cell layer. The current of a subcell is then calcu-lated for the AM1.5d spectrum by assuming that each absorbed photon creates one electron–hole pair, which contributes to the photocurrent.
The algorithm for the optimization of the four-junction solar cell followed a sequence of
1) finding the optimum bandgap combination by maximizing a fitness function defined as the sum of all bandgap ener-gies multiplied by the photocurrent of the current limiting subcell; this represents an approximation to the power of the device;
2) applying a single-diode model to each junction to deter-mine the overall dark-current characteristics of the four-junction cell;
3) optimizing PM P P by assuming J(V) = smallest photocur-rent of all subcells minus the sum of the dark currents of
the subcells.
Typical values have been assumed for the series resistance (15 mΩ•cm−2) and grid shading (4%). These values are realistic and may be even improved in the future. The parallel resistance was taken as infinite. The reverse saturation current Jrs is a