We characterize a TOPCon module by I-V curve measurements at temperatures between 15 degrees C and 45 degrees C. To derive the module I-V curve parameterization used in PV system simulations, typically the slope of the measured I-V curves near ISC towards larger voltages is interpreted as a shunt conductance. This method, applicable for single cells, leads for modules to the observation of an apparent intensity-dependent shunt RSh, parameterized in the model of De Soto and (differently parameterized) in PVsyst. However, this slope is rather the effect of cell current mismatches. We show, based on careful, yet standard state-of-the-art characterization, that there is no indication that this slope of the I-V curve corresponds to an actual shunt conductance. More importantly, we show that by not using such intensity dependent shunt term, we simplify the model, do not require any temperature dependent ideality ("fudge") factors and achieve better reproduction of the module performance. We show this by using a simple 1-diode model fit with constant series resistance. However, our detailed analysis shows two clearly identifiable exponential components ("2 diodes") in the I-V curves. The saturation currents follow expectable temperature dependencies. The series resistance RS exhibits a slight intensity dependence that merges towards low intensities into the separately-determined RS of the dark I-V curves. The temperature coefficient of RS is close to that of metals like copper or aluminum.

It is common practice in PV system simulation to use the De Soto model, which describes how to use the 1-diode equivalent circuit model for modules. De Soto's model scales the shunt with irradiance, making it disappear toward zero W/m2. Also, the commercial software PVsyst uses a parameterization that reduces the shunt effect when the irradiance goes down. However, the solar cells that make up a module typically do not have an illumination-dependent shunt. We therefore investigate the origin of the intensity-dependent apparent shunt in modules. We show that this apparent shunt (derived from the slope of the quasi-linear region from ISC onwards) is a misinterpretation for module I-V curves and has little to do with a shunt conductance, although this slope method serves well for determining the shunt conductance of individual cells. Instead, the module I-V curve slope of the quasi-linear region from ISC onwards is strongly influenced by even small ISC mismatches between the cells. Such mismatch can occur from small illumination inhomogeneity even for A+ solar simulators in the laboratory, or from cell production variation. Abandoning the practice of using the I-V curve slope to determine the shunt value for equivalent circuit models of modules (and the corresponding shunt scaling in the De Soto model or PVsyst) contributes to physically more meaningful I-V curve parameterizations and bears the opportunity for further improved accuracy of PV system energy yield prediction.

We explore how positioning high-albedo material on the ground impacts the performance gain obtainable by these selective ground albedo enhancements. The ground albedo can be improved, for example, by placing geosynthetic materials, white stones, or paint under or between PV panels. Specifically, we determine how the most effective positioning of albedo enhancement material (AEM) is influenced by geographic latitude, by the diffuse-light content in the total irradiance, by the choice between tracking versus fixed-tilt mounting, and by module mounting height. We find for fixed-tilt systems that albedo improvement strips with a width similar to the module table or less, should be under the tables, but shifted towards the sun-facing leading (lower) edge of the modules. This preference for placement towards the lower edge is slightly more pronounced the closer the location is to the equator. For typical/high-mounted (1.5m torque-tube height) tracked systems, for all albedo strip widths, the optimum placement is centered at the center of the module table. However, for low-mounted tracked systems and a ground area coverage by the AEM of about equal or less the module table width, we find that it is favorable to split the material into two off-centric strips rather than one strip centered to the trackers. Interestingly, we find for both fixed-tilt and tracking systems that the optimum placement of such albedo enhancement strips is not notably influenced by the diffuse content of the irradiation. However, the magnitude of the gain is influenced by the diffuse content and we find higher relative gain from albedo improvement in case of a high diffuse light content. Particularly for tracked systems the relative production gain by ground albedo enhancement is larger for higher diffuse content. Of course, the overall performance of tracked systems prefers low diffuse light content and dominance of DNI.

Photoluminescence is a powerful technique to analyze solar cell recombination across the manufacturing steps. However, this analysis requires accurate modelling of the phenomena caused by injection-dependent lifetimes. We present a transient photoluminescence technique that we apply to our current new-generation production cells. We extract the saturation current Jo and the bulk lifetime, in a refined version of the method developed by Kane and Swanson. We include band-gap narrowing effects occurring at the large operating voltages typical of our back-contact cells. Our approach results in excellent fittings of the carrier density decay (R 2 =99.9%). We find an area averaged J o of 6.2 fA/cm 2 and an inverse bulk Shockley-Read-Hall lifetime of 14 s −1 in a production cell measured after metal etching.

A model, adapted from the Shockley–Queisser detailed balance model to tandem solar cells with a monolithically grown GaAsxP1−x top junction on a Si bottom junction, has been developed. Updated data have been used for the absorption spectrums. Two surface geometries, flat and ideally textured, have been investigated. As an important improvement over existing models, the effects of threading-dislocations-related Shockley–Read–Hall recombinations in the GaAsxP1−x cell, due to the lattice mismatch between the GaAsxP1−x epilayers and the Si substrate, have been taken into consideration. Auger recombinations in the Si bottom cell and luminescent coupling between the cells have also been considered. For a dislocation-free 2-μm-thick top cell, maximal theoretical efficiencies of 41.6% and 39.1% have been calculated for a textured and a flat surface, respectively. For threading dislocation (TD) densities below 104cm−2, the impact of TDs in the GaAsxP1−x layers on the solar cell performances is very limited. With TD densities over 105cm−2, the top cell open-circuit voltage is reduced, hence the overall efficiency. For TD densities over 4×106cm−2, as the diffusion length of minority carriers in the base gets smaller than the base thickness, the short-circuit current in the top GaAsxP1−x cell is also reduced, resulting in a decrease in the optimal top cell bandgap. Using non-ideal EQEs and surface recombination rates from published experimental data, the long-term efficiency potential of the investigated technology has been estimated to be ~35.1% for an ideally textured GaAsxP1−x/Si tandem cell with a TD density of 105cm−2 (~33.0% with a flat surface).

Lattice-mismatched 1.7eV Al0.2Ga0.8As photovoltaic solar cells have been monolithically grown on Si substrates using Solid Source Molecular Beam Epitaxy (SSMBE). As a consequence of the 4%-lattice-mismatch, threading dislocations (TDs) nucleate at the interface between the Si substrate and III-V epilayers and propagate to the active regions of the cell. There they act as recombination centers and degrade the performances of the cell. In our case, direct AlAs/GaAs superlattice growth coupled with InAlAs/AlAs strained layer superlattice (SLS) dislocation filter layers (DFLSs) have been used to reduce the TD density from 1x10(9) cm(-2) to 1(+/- 0.2) x10(7) cm(-2). Lattice-matched Al0.2Ga0.8As cells have also been grown on GaAs as a reference.The best cell grown on silicon exhibits a V-oc of 964mV, compared with a V-oc of 1128mV on GaAs. Fill factors of respectively 77.6% and 80.2% have been calculated. Due to the lack of an anti-reflection coating and the non-optimized architecture of the devices, relatively low J(sc) have been measured: 7.30mA.cm(-2) on Si and 6.74mA.cm(-2) on GaAs. The difference in short-circuit currents is believed to be caused by a difference of thickness between the samples due to discrepancies in the calibration of the MBE prior to each growth. The bandgap-voltage offset of the cells, defined as E-g/q-V-oc, is relatively high on both substrates with 736mV measured on Si versus 572mV on GaAs. The non-negligible TD density partly explains this result on Si. On GaAs, non-ideal growth conditions are possibly responsible for these suboptimal performances.

Al0.2Ga0.8As photovoltaic solar cells have been monolithically grown on silicon substrates by Molecular Beam Epitaxy. Due to the 4% lattice mismatch between AlGaAs and Si, Threading Dislocations (TDs) nucleate at the III-V/Si interface and propagate to the active region of the cells where they act as recombination centers, reducing the performances of the devices. In order to reduce the Threading Dislocation Density (TDD) in the active layers of the cells, InAlAs Strained Layer Superlattice (SLS) Dislocation Filter Layers (DFLs) have been used. For one of the samples, in-situ Thermal Cycle Annealing (TCA) steps have additionally been performed during growth. For comparison purposes, reference Al0.2Ga0.8As solar cells have been grown lattice-matched on GaAs. For the sample grown on Si without TCA, the TDD has been reduced from over 7×109cm-2 at the III-V/Si interface to 3×107cm-2 in the base of the cells. With TCA, the TDD has been reduced throughout the sample from over 3×109cm-2 in the initial epilayers to 8(±2)×106cm-2 in the base of the cells. For the best devices, the Voc improves from 833mV on Si without TCA to 895mV using TCA, compared with 1070mV for the reference sample grown lattice-matched on GaAs. Similarly the fill factor improves from 73.7% on Si without TCA to 74.8% using TCA, compared with 78.4% on GaAs. The high bandgap-voltage offset obtained both on Si and GaAs indicates a non-optimal bulk AlGaAs material quality due to non-ideal growth conditions.

A model, derived from the detailed balance model from Shockley and Queisser, has been adapted to monolithically grown GaAsP/Si tandem dual junction solar cells. In this architecture, due to the difference of lattice parameters between the silicon bottom cell – acting as the substrate – and the GaAsP top cell, threading dislocations (TDs) arise at the IIIV/ Si interface and propagate in the top cell. These TDs act as non-radiative recombination centers, degrading the performances of the tandem cell. Our model takes into account the impact of TDs by integrating the NTT model developed by Yamaguchi et. al.. Two surface geometries have been investigated: flat and ideally textured. Finally the model considers the luminescent coupling (LC) between the cells due to reemitted photons from the top cell cascading to the bottom cell. Without dislocations, LC allows a greater flexibility in the cell design by rebalancing the currents between the two cells when the top cell presents a higher short-circuit current. However we show that, as the TD density (TDD) increases, nonradiative recombinations take over radiative recombinations in the top cell and the LC is quenched. As a result, nonoptimized tandem cells with higher short-circuit current in the top cell experience a very fast degradation of efficiency for TDDs over 104cm-2. On the other hand optimized cells with matching currents only experience a small efficiency drop for TDDs up to 105cm-2. High TDD cells therefore need to be current-matched for optimal performances as the flexibility due to LC is lost.

We present an analytical model for the current transport in polycrystalline (poly) Si/interfacial oxide/monocrystalline (c)-Si base junctions, which consistently describes the symmetrical behavior of an n(+) poly-Si emitter/p c-Si base and p(+) poly-Si emitter/n c-Si base configuration. Our model is focused on a regime within which the current transport is possibly dominated by a flow through oxide pinholes rather than by tunneling. For an emitter region assumed to form underneath the interfacial oxide by diffusion of dopants from the poly-Si into the c-Si, we calculate the minority charge carrier distribution and the resistance implied for majority charge carriers. With reasonable parameters, our model simultaneously reproduces the experimentally observed low emitter saturation current densities and low junction resistances values. Our model provides a plausible explanation for the high current gain observed in p-n-p and n-p-n bipolar transistors featuring a poly-Si emitter. In principle, the obtained correlation between recombination current and series resistance is analogous to the situation in a base region of a solar cell with local rear contacts. Thus, a poly-Si/c-Si junction can be explained within the framework of a classical p-n junction picture for a passivated, locally contacted emitter.

We present an experimental method to quantify the series resistance Ra-Si/ITO through the a-Si:H layers and the a-Si:H/ITO interface on test structures. In order to optimize Ra-Si/ITO, we apply different a-Si:H and ITO deposition parameters. We find the best value for R(p)-a-Si/ITO of 0.42 Ω·cm2 for an ITO double layer with a 10-nm-thin starting layer that provides good contact resistance and an additional 90-nm top layer that provides good conductivity. For R(n)-a-Si/ITO, we reach values below 0.1 Ω·cm2. We present an analysis of the series resistance and shading losses of our 100-cm2 bifacial screen-printed a-Si:H/cSi heterojunction solar cells, which show an open-circuit voltage of Voc = 733 mV, demonstrating the excellent level of interface passivation. The efficiency of 20.2% is limited by a low short-circuit current density of 37.1 mA/cm2 and fill factor of 76%.

We present an analytical model for the current transport in polycrystalline (poly)Si/interfacial oxide/monocrystalline ( c)-Si base junctions, which consistently describes the symmetrical behavior of an n+ poly-Si emitter/ p c-Si base and p+ poly-Si emitter/ n c-Si base configuration. Our model is focused on a regime within which the current transport is possibly dominated by a flow through oxide pinholes rather than by tunneling. For an emitter region assumed to form underneath the interfacial oxide by diffusion of dopants from the poly-Si into the c-Si, we calculate the minority charge carrier distribution and the resistance implied for majority charge carriers. With reasonable parameters, our model simultaneously reproduces the experimentally observed low emitter saturation current densities and low junction resistances values. Our model provides a plausible explanation for the high current gain observed in p-n-p and n-p-n bipolar transistors featuring a poly-Si emitter. In principle, the obtained correlation between recombination current and series resistance is analogous to the situation in a base region of a solar cell with local rear contacts. Thus, a poly-Si/ c-Si junction can be explained within the framework of a classical p-n junction picture for a passivated, locally contacted emitter.

Aluminum oxide provides an excellent surface passivation on p- and n-type crystalline silicon. On n-type silicon, however, the effective excess carrier lifetime τeff is often found to be injection dependent. Our experimental results show that, in our case, this effect depends mainly on the size of the lifetime samples. The fixed negative charges present at the Al2O3/c-Si interface induce an inversion layer at the surface, which results in a p–n-junction close to the surface of the sample and the inversion layer couples the sensing area with the poorly passivated or damaged edge of the sample. For smaller samples stronger injection-dependent lifetimes are measured, whereas large samples show a smaller injection dependence. In addition, photoconductance-calibrated photoluminescence lifetime images show that for low injection levels the lifetime decreases towards the sample edge. Device simulations for different sample sizes including the edge recombination are in agreement with the measured injection-dependent lifetimes. Therefore, it is necessary to use sufficiently large samples or decouple the sensing area form the edge when evaluating the injection dependence of the lifetime. For the samples used in this contribution, the injection dependence of the lifetime did not even fully vanish for an edge length of 12.5cm.

In this work, we investigate the applicability of counterdoping by ion implantation for the formation of pn-junctions for high efficiency interdigitated back contacted silicon solar cells. Counterdoping offers the possibility of creating the emitter with a blanket implantation and the back surface field with a masked implantation, leading to an elegant process without the need of precise alignment between the two implantation steps. We analyze I-V curves of diodes after implantation and high temperature annealing and compare the results with numerical simulations. Despite the presence of highly doped boron and phosphorous regions in contact to each other, neither trap assisted tunneling nor dominant recombination in the space-charge region is observed in forward direction. This result reflects the excellent removal of implant damage during the co-annealing step. In reverse direction, a sharp breakdown due to band-to-band-tunneling is observed at -8 V. Since it occurs very homogenously across the whole wafer and no local hot spots are observed, no implications for module reliability are implied.

We show the degradation of the front surface passivation by rear-side laser processing of thin silicon solar cells when using a laser with a pulse length of 8 ps. 45-μm-thick back-contact back-junction monocrystalline silicon solar cells show an energy conversion efficiency of 18.8% without rear-side laser processing, whereas they show only 7.5% with an additional rear-side laser process step for contact separation. This low efficiency is due to the degradation of the front surface passivation, which is confirmed by quantum efficiency measurements. The internal quantum efficiency at short wavelength is 0.88 without laser processing, whereas it is only 0.33 with the rear-side laser process step.