We have developed a machine learning empowered computational framework to facilitate design space exploration for optoelectronic devices. In this work, we apply dimensionality reduction and clustering machine learning algorithms to identify optimal ten-junction C-band photonic power converter (PPC) designs. We outline our framework, design optimization procedure, calibrated optoelectronic model, and experimental calibration devices. We report on top performing device designs for on-substrate and flat back-reflector architectures. We comment on the design sensitivity for these PPCs and on the applicability of dimensionality reduction and clustering algorithms to assist in optoelectronic device design.
Using a model developed in Crosslight PICS3D, we have compared simulated gain and device performance for InAs/InP quantum dash ridge waveguide lasers with experimental data from fabricated devices. We investigated the change in device behaviour as the energy spectrum of the dashes is varied and inhomogeneous broadening is changed to represent a distribution of dash sizes and composition. We observed a distinct asymmetry of the dash layer occupation due to inefficient thermionic emission hindering hole transport across the quantum dash layer stack. We have quantified how this effect can be utilized to achieve higher threshold current temperature stability.
Photonic power converters designed to operate in the telecommunications O-band were measured under non-uniform 1319 nm laser illumination. Two device architectures were studied, based on lattice-matched InGaAsP on an InP substrate and lattice-mismatched InGaAs grown on GaAs using a metamorphic buffer. The maximum measured efficiencies were 52.9% and 48.8% for the lattice-matched and -mismatched designs respectively. Both 5.4-mm2 devices were insensitive to the incident laser spot size for input powers of < 250 mW and exhibited better performance for larger spot sizes with more uniform illumination profiles at higher powers.
The effects of spectral albedo on bifacial silicon heterojunction photovoltaic cell performance is explored in six locations in North America using an optoelectronic drift-diffusion model. We model seven spectral albedos using the scaled rear irradiance method and compare to broadband values. Cell performance varies geographically, with the maximum efficiency of 22.6% calculated for Cambridge Bay (69°N) with snow ground cover, and maximum output power of 216 W/m2 for Mexico City (20°N) with white sand. Neglecting spectral effects of albedo can under or over-estimate power by >2%, which can significantly impact system-level energy yield.
Non-uniform irradiance on the rear side of bifacial photovoltaic (PV) modules causes electrical mismatch between cells and energy loss across the module. Racking structures increase this non-uniformity through shadows and reflections that vary throughout the day. However, commercial software typically use constant values to estimate mismatch losses in annual simulations. We investigate the impact of torque tube shading and reflection on rear side irradiance mismatch in bifacial PV modules in one-in-portrait (1P) and two-in-portrait (2P) horizontal single-axis trackers with a range of ground albedos over a typical meteorological year in Livermore, California, USA. Irradiance simulations use a version of bifacial_radiance, the National Renewable Energy Laboratory’s python wrapper for the RADIANCE ray tracing software, which we modified for arbitrary 2D irradiance sampling of the module(s) under investigation. For a torque tube reflectivity of 0.745, torque tube reflection accounts for 3.0% and 5.5% of the annual rear insolation in 1P and 2P configurations, respectively, for a 0.2 albedo; or 2.9% and 3.1% for a 0.6 albedo. Torque tube reflection decreases annual rear insolation mismatch from 11.8% to 10.7% in 1P configurations, and from 11.5% to 9.8% in 2P configurations with 0.2 albedo. Similarly, with 0.6 albedo, annual rear insolation mismatch decreases from 12.6% to 11.6% in 1P configurations, and from 11.9% to 10.4% in 2P configurations. However, we demonstrate that annual figures are insufficient for capturing the impact of torque tube reflection; seasonal and diurnal variations must also be considered.
The use of III-V and group IV compounds in the same heterostructure is of great interest for high performances solar cells under concentration. In fact, the combination of these III-V and group IV compounds can lead to interesting strategic bandgap choices and engineering to better match the absorption of the solar spectrum, and therefore better solar cell performance. The series-connected quad-junctions (4J) solar cell strategies have the potential to improve solar cells performance and therefore enable low-cost concentrator photovoltaic (CPV) systems, allowing lower levelized cost of electricity (LCOE) from a CPV system. This work presents the investigation of the performance of dual junction (2J) GaInP/GaAs that might be implemented as upper cells with group IV (SiGeSn) cells as bottom cells. The aim of this study is to validate the epitaxial structure and the fabrication process for future 4J cells development. Pitch is varied from 125 μm to 400 μm for two different size of cells, in order to optimize solar cells performance under concentration (X) In the range of 100X to 1000X. Solar cells demonstrated high fill factor (FF) values and ideality factors (n) approaching unity per subcell have been obtained in the range of 100X to 500X. A FF of 85% and 88% are obtained at a concentration of 1000X for the bigger and smaller cells respectively, for the narrowest pitch. These results close to the state-of-the-art are encouraging for the implementation of this 2J with IV bottom subcell for the purpose of high performance 4J.
Bifacial photovoltaics present a clean and cheaper alternative to diesel generators for high-latitude remote communities; however, solar cells are tested at air mass 1.5, while average air mass increases with increasing latitude. For example, Cambridge Bay (69ºN) has an irradiance-weighted average air mass of 3.1. We demonstrate improved efficiency of bifacial silicon heterojunction modules under high air mass spectra due to reduced incident UV light. We implement air mass correction in our bifacial PV modelling software, and we quantify the impact of air mass on energy yield for fixed-tilt and tracked systems in high latitude locations.
Photonic power converters (PPCs) are one of the main components of optical power transmission systems, generating electrical power via the photovoltaic effect. We simulate ultrathin PPCs designed for operating at the telecommunication wavelength of 1310 nm with 9 and 12 times thinner absorbing layers using cubic and pyramidal nanostructured back reflectors (BRs), respectively. While increasing efficiency by 13% (rel.) over conventional PPCs, results also show superior light trapping for pyramidal BR with twice the absorption of a simple double pass absorber layer of the same thickness and higher short-circuit current for pillar BR reaching 94% of an ideal Lambertian surface.
Bifacial photovoltaics present a clean and cheaper alternative to diesel generators for high-latitude remote communities; however, solar cells are typically tested at 0° angle of incidence, 25°C, and AM1.5, from which high-latitude conditions vary greatly. A bifacial silicon photovoltaic cell optimized for high-latitude conditions will improve energy yield for these systems. We integrate experimentally-derived cell parameters with a systems-level model capable of fixed-tilt and tracked energy yield predictions. We optimize to find the most efficient cell design for high-latitude environments in Sentaurus and SunSolve and determine the resulting improvement in energy yield for an entire panel in MATLAB.
KEYWORDS: Silicon, Germanium, Interfaces, Multijunction solar cells, Photovoltaics, Electrochemical etching, Transmission electron microscopy, Chemical mechanical planarization, Solar cells
III-V solar cell cost reduction and direct III-V/Si integration can both be realized by depositing a thin layer of high-quality Ge on relatively low-cost Si substrates. However, direct epitaxial growth of Ge on Si substrates is difficult due to the 4% lattice mismatch between the film and the substrate. Threading dislocations (TDs) introduced within the Ge layer have a detrimental effect on device performances. The goal of this research is to address the perennial need to minimize the defect density of Ge epilayers grown on a Si substrate. We seek to accommodate the effects of the lattice mismatch by introducing a porous Si interface layer to intercept dislocations and prevent them from reaching the active layers of the device. The porous Si layer is formed through dislocation-selective electrochemical deep etching and thermal annealing. The porous layer created beneath the top Ge layer can both act as dislocation traps and as a soft compliant substrate, which displays high flexibility. Transmission electron microscopy (TEM) analysis of the Ge/porous Si interface shows that the lattice mismatch strain of the Ge films was almost relaxed. The surface roughness of this modified Ge/Si substrate has been reduced using chemical mechanical polishing (CMP) process to fulfil the requirements for epitaxy of III-V alloys. Finally, we present simulation results exploring the effect of threading dislocations on device performance.
Distributed circuit models (DCM) divide photovoltaic devices into discrete elementary units. Each unit is assigned an equivalent circuit based on geometry and location, with circuit parameters being fit to or extrapolated from experimental results. Interconnection of these elementary units with ohmic resistors representing lateral and vertical resistances within the layers of the device forms the complete circuit model. DCMs allow grid design optimization, simulation of chromatic aberration, luminescent coupling and analysis of power losses due to regionally specific resistances, which are not possible with simple lumped models. Previous DCMs have been limited to 1-3 junction devices, using a 2D surface model, or use of a one-diode circuit model for the cell junctions. Furthermore, a DCM can be used to simulate complex multi-junction devices with non-uniform illumination, whereas in comprehensive physics-based simulators like Synopsys TCAD Sentaurus this would require vastly greater computational resources.
In this work, a parameterized 3D distributed circuit model was developed to calculate the performance of III-V solar cells and photonic power converters (PPC) with a variable number of epitaxially stacked pn junctions. We validated these calculations against published results using a similar 3D model for a 1-junction solar cell. Furthermore, experimental results from Azastra Opto’s 20-junction PPC illuminated by an 845 nm diode laser are compared. These devices are designed with many pn junctions to achieve higher voltages and to operate under non-uniform illumination profiles from a laser or LED. The effect on device performance of varying both these parameters will be discussed.
Four-junction solar cells for space and terrestrial applications require a junction with a band gap of ∼1 eV for optimal performance. InGaAsN or InGaAsN(Sb) dilute nitride junctions have been demonstrated for this purpose, but in achieving the 14 mA/cm2 short-circuit current needed to match typical GaInP and GaAs junctions, the open-circuit voltage (VOC) and fill factor of these junctions are compromised. In multijunction devices incorporating materials with short diffusion lengths, we study the use of thin junctions to minimize sensitivity to varying material quality and ensure adequate transmission into lower junctions. An n-i-p device with 0.65-μm absorber thickness has sufficient short-circuit current, however, it relies less heavily on field-aided collection than a device with a 1-μm absorber. Our standard cell fabrication process, which includes a rapid thermal anneal of the contacts, yields a significant improvement in diffusion length and device performance. By optimizing a four-junction cell around a smaller 1-sun short-circuit current of 12.5 mA/cm2, we produced an InGaAsN(Sb) junction with open-circuit voltage of 0.44 V at 1000 suns (1 sun=100 mW/cm2), diode ideality factor of 1.4, and sufficient light transmission to allow >12.5 mA/cm2 in all four subcells.
A monolithic compound semiconductor phototransducer optimized for narrow-band light sources was designed for and has achieved conversion efficiencies exceeding 50%. The III-V heterostructure was grown by MOCVD, based on the vertical stacking of a number of partially absorbing GaAs n/p junctions connected in series with tunnel junctions. The thicknesses of the p-type base layers of the diodes were engineered for optimal absorption and current matching for an optical input with wavelengths centered in the 830 nm to 850 nm range. The device architecture allows for improved open-circuit voltage in the individual base segments due to efficient carrier extraction while simultaneously maintaining a complete absorption of the input photons with no need for complicated fabrication processes or reflecting layers. Progress for device outputs achieving in excess of 12 V is reviewed in this study.
A twelve-junction monolithically-integrated GaAs phototransducer device with >60% power conversion efficiency and >14 V open-circuit voltage under monochromatic illumination is presented. Drift-diffusion based simulations including a luminescent coupled generation term are used to study photon recycling and luminescent coupling between each junction. We find that luminescent coupling effectively redistributes any excess generated photocurrent between all junctions leading to reduced wavelength sensitivity. This broadened response is consistent with experimental measurements of devices with high-quality materials exhibiting long carrier lifetimes. Photon recycling is also found to significantly improve the voltage of all junctions, in contrast to multi-junction solar cells which utilize junctions of differing bandgaps and where high-bandgap junctions benefit less from photon recycling.
The high-efficiency conversion of photonic power into electrical power is of broad-range applicability to many industries due to its electrical isolation from the surrounding environment and immunity to electromagnetic interference which affects the performance and reliability of sensitive electronics. A photonic power converter, or phototransducer, can absorb several watts of infrared laser power transmitted through a multimode fiber and convert this to electrical power for remote use. To convert this power into a useful voltage, we have designed, simulated, and fabricated a photovoltaic phototransducer that generates >5 V using a monolithic, lattice-matched, vertically-stacked, single-cell device that eliminates complex fabrication and assembly steps. Experimental measurements have demonstrated a conversion efficiency of up to 60.1% under illumination of ~11 W/cm2 at a wavelength of 835 nm, while simulations indicate that efficiencies reaching 70% should be realistically achievable using this novel design.
Triple-junction AlGaInP/InGaAs/Ge solar cells with embedded InAs quantum dots are presented, where typical
samples obtain efficiencies of > 40% under AM1.5D illumination, over a range of concentrations of 2- to 800-suns
(2 kW/m2 to 800 kW/m2). Quantum efficiency measurements show that the embedded quantum dots improve the
absorption of the middle subcell in the wavelength range of 900-940 nm, which in turn increases the overall
operating current of the solar cell. These results are obtained with 1 cm2 solar cells, and they demonstrate the
solar cells' low series resistance, which and makes them ideal for the current generation in commercial
concentrator systems. The thermal management and reliability of the solar cell and carrier is demonstrated by
testing the experimental samples under flash (up to 1000-suns) solar simulator and continuous (up to 800-suns)
solar simulator. Under continuous solar illumination, the solar cell temperature varies between ~Δ3°C at 260-suns
linearly to ~Δ33°C at 784-suns when the solar cell is mounted with thermal paste, and ~Δ27°C at 264-suns linearly
to ~Δ91°C at 785-suns when no thermal paste is used. The solar cells experience the expected shift in open circuit
voltage and efficiency due to temperature, but otherwise operate normally for extended periods of time.
KEYWORDS: Resistance, Solar cells, Oscilloscopes, Multijunction solar cells, Time metrology, Doping, Remote sensing, Power supplies, Calibration, Germanium
The current density-voltage characteristic of an AlGaAs/AlGaAs tunnel junction is determined by taking a time-averaged
measurement across the device. A tunnelling peak of ~950A/cm2 is recorded by this method. Measurements of the
tunnelling peak and valley currents by the time averaging method are obscured due to the unstable nature of the negative
differential resistance region of the current density-voltage characteristic. This AlGaAs/AlGaAs tunnel junction is then
biased inside the negative differential resistance region of the current density-voltage characteristic, causing the current
and the voltage to oscillate between the peak and the valley. The current and voltage oscillations are measured over time
and then currents and voltages corresponding to the same time stamps are plotted against each other to form a timedependent
curve from which a tunnelling peak of a value larger than 1100A/cm2 is determined. The peak determined by
this method is 11-20% larger than previously determined using the time averaged measurement. An AlGaAs/InGaP
tunnel junction having no negative differential resistance region is also presented.
Simulations of AlxGa1-xAs/GaAs (x = 0.3) and AlxGa1-xAs/AlxGa1-xAs (x < 0.2) tunnel junction J-V characteristics are
studied for integration into a 2D metamorphic multi-junction solar cell model composed of GaInP/GaAs/InGaAs. A
comparison of the simulated solar cell J-V characteristics under AM1.5D spectrum is discussed in terms of short circuit
current density (Jsc), open circuit voltage (VOC), fill factor (FF) and efficiency (η) for both tunnel junction designs. Using
AlxGa1-xAs/GaAs top and bottom tunnel junctions, the metamorphic solar cell obtained values of Jsc = 12.3 mA/cm2, VOC
= 2.56 V, FF = 0.81 and η = 25.5%, whereas the solar cell with the AlxGa1-xAs/AlxGa1-xAs top and bottom tunnel
junctions reported values of Jsc = 12.3 mA/cm2, VOC = 2.22 V, FF = 0.81 and η = 22.1%. At open circuit voltage, energy
band diagrams show minimal curvature in the electron and hole quasi Fermi levels; furthermore, the difference between
the top sub-cell electron quasi Fermi level and the bottom sub-cell hole quasi Fermi level is shown to be equal to qVOC
for both designs. The energy band diagram of the complete structure is compared for both tunnel junction designs,
showing the difference in energy levels that correspond to the difference in measured open circuit voltage. The observed
decrease in open circuit voltage was ΔVOC = 0.34 V, which was attributed to the difference in tunnel junction material
band parameters such as bandgap, valence and conduction band offsets at heterojunctions and Fermi level degeneracies
due to doping concentration differences.
InAs quantum dots in a GaAs matrix are studied. Those quantum dots are used in applications to enhance the overall
efficiency of multi-junction solar cells beyond 40%. Photoluminescence measurements at 77 K using a 532 nm laser
have been performed on an epitaxially grown structure of self-assembled InAs quantum dots in a GaAs matrix upon a Ge
substrate, where three energy levels are determined at En=0=1.01 eV, En=1=1.07 eV and En=2=1.13 eV. Theoretical
calculations of the energy levels determine the quantum dots to be 7 nm high and have a 37 nm base diameter, which is
close to atomic force microscopy measurements performed on the samples. Intensity dependant photoluminescence
measurements reveal the saturation of the first excited energy level at 5×106 W/m2. A general model for the saturation of
the first quantum dot excited energy level is then developed. This saturation model is applied to the AM1.5D solar
spectrum at 297 K to determine the concentration of solar energy needed to saturate the first excited energy level within
a multi-junction solar cell. Saturation was determined to be at ~1.56×104 suns (where 1 sun = 1000 W/m2). Since current
solar concentrations are between 500-1000 suns concentration, the saturation of such quantum dots will not occur.
KEYWORDS: Solar cells, Receivers, Sun, Solar concentrators, Multijunction solar cells, Calibration, Temperature metrology, Computer simulations, Copper, Device simulation
The thermal performances of multi-junction solar cells, mounted on receivers, are studied to determine the change in
device efficiency with respect to sunlight concentration under continuous illumination. Experimental characterization of
the device was performed by measuring the solar cell current-voltage curve using both flash and continuous-illumination
solar simulators. We are able to extract the change in efficiency and open circuit voltage with respect to the change in
concentration from experiments with respect to the application of thermal paste between the receiver and the heat
exchange. We show the efficiency linearly decrease at a rate of -0.0094%/°C (no paste) and -0.0043%/°C (paste). We
used the calibrated numerical model to determine the solar cell temperature and incorporate the corresponding efficiency
when scaled up to 2000 sun concentrations under continuous illumination.
Photovoltaic solar cells are a route towards local, environmentally benign, sustainable and affordable energy solutions.
Antireflection coatings are necessary to input a high percentage of available light for photovoltaic conversion, and
therefore have been widely exploited for silicon solar cells. Multi-junction III-V semiconductor solar cells have achieved
the highest efficiencies of any photovoltaic technology, yielding up to 40% in the laboratory and 37% in commercial
devices under varying levels of concentrated light. These devices benefit from a wide absorption spectrum (300-
1800 nm), but this also introduces significant challenges for antireflection coating design. Each sub-cell junction is
electrically connected in series, limiting the overall device photocurrent by the lowest current-producing junction.
Therefore, antireflection coating optimization must maximize the current from the limiting sub-cells at the expense of
the others. Solar concentration, necessary for economical terrestrial deployment of multi-junction solar cells, introduces
an angular-dependent irradiance spectrum. Antireflection coatings are optimized for both direct normal incidence in air
and angular incidence in an Opel Mk-I concentrator, resulting in as little as 1-2% loss in photocurrent as compared to an
ideal zero-reflectance solar cell, showing a similar performance to antireflection coatings on silicon solar cells. A transparent conductive oxide layer has also been considered to replace the metallic-grid front electrode and for inclusion as part of a multi-layer antireflection coating. Optimization of the solar cell, antireflection coating, and concentrator system should be considered simultaneously to enable overall optimal device performance.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.