Search

This article presents for the first time the application of wet-chemical interfacial oxide from an industrial inline processing system for poly-Si-based passivating contacts. An excellent passivation quality is achieved by creating an interfacial oxide with a very short exposure time of 90 s in ozonized water and by adjusting the annealing temperature in a tube furnace, resulting in surface recombination current densities of 4 fA/cm2 and 1.2 fA/cm2 before and after a hydrogenation step, respectively. Detailed electrical characterization reveals the interplay of in-diffusion of P into the wafer and hydrogenation step. Our investigation shows that the optimum annealing temperature can differ before and after the hydrogenation step. The developed wet-chemical interfacial oxide is successfully implemented in back junction solar cells on large-area gallium-doped p-type silicon wafers (156.75 × 156.75 mm2) featuring a phosphorus-doped poly-Si-based passivating contact at the rear side. The best cell has an efficiency of 23.6% and an open-circuit voltage of 719 mV, independently confirmed by ISFH CalTeC in Germany. Our cost calculation shows a saving of up to 17.2% in capital expenditure, 5.2% p.a. in operating expense, and 9.0% in the footprint if the interfacial oxide is formed by an inline wet-chemical processing system instead of a plasma chamber.

This article investigates the impact of the back-surface-field (BSF) thickness variation within a local aluminum contact on the performance of passivated emitter and rear contact solar cells. A significant difference of BSF thickness between contact endings and the center of dash-shaped contacts is verified experimentally by a comprehensive statistical analysis using scanning electron microscopy. The impact of local BSF thickness differences on the cell performance is studied with 3-D technology computer-aided design (TCAD) device simulations. Several device parameters such as BSF thicknesses, the doping concentration in the BSF profile at rear contacts, or the metallized area fraction at the cell rear side are varied. Our simulation study shows that the open-circuit voltage is mainly affected by locally reduced BSF thicknesses, resulting in an efficiency loss up to 0.14%abs or 0.84%abs, respectively, if an area fraction of 1% or 20% within a local contact has reduced BSF thicknesses. This effect can be minimized either by reducing the metallized area fraction at the cell rear side or by increasing the doping concentration in the BSF profile at aluminum rear contacts. In addition, we demonstrate that the 3-D simulations can be approximated with 2-D simulations by applying a single doping profile with an average BSF thickness, calculated with the harmonic mean.

Sufficiently deep local back-surface-fields (LBSF) are crucial for achieving low contact recombination and thus high solar cell efficiencies in passivated emitter and rear contact (PERC) solar cells. In this article, we investigate spatial variations of LBSF thickness 1) across dashed contacts, 2) between lateral positions within a cell, and 3) from cell to cell. The objective of this experimental study is to proof the impact of LBSF thickness on open-circuit voltage in PERC solar cells. Our measurements and analysis of variance (ANOVA) reveal that variations in LBSF thickness across dashed contacts exceed observed variations in LBSF thickness between lateral positions within a cell as well as cell-to-cell variations. Thus, we statistically treat all LBSF measurements of a single experimental group of PERC solar cells as one population of measurements. The two investigated experimental sets of PERC cells fired using different fast-firing profiles show a difference in LBSF thickness of about 1 μm. The resulting difference in open-circuit voltages arises to 1.3 mV. ANOVA-based analysis shows for our experimental samples that at least 19 random measurements are necessary in order to resolve LBSF thickness difference of 0.5 μm with sufficient statistical significance.

We investigate the transport mechanism of poly-Sibased carrier-selective junctions using the two-dimensional numerical semiconductor device simulations. The detailed transport model considers the charge carrier transport through the pinholes as well as tunneling through a very thin silicon oxide simultaneously. For the verification of the simulation model, the complete temperature dependent transfer length method is modeled and its results are verified with measurements of two different samples. By means of rigorous simulations, the influence of different pinhole geometrical and material parameters on junction resistivity are investigated and explained in detail. From the presented results, the fundamental understanding needed for optimizing the polySi-based carrier selective junction in respect to the main design parameters such as doping level in poly-Si, annealing time, silicon oxide thickness, and pinhole density is given. The detailed analysis shows the pinhole channel plays the most crucial role in the design of poly-Si-based carrier-selective junctions if the silicon oxide layer thickness is larger than 2 nm.

We present a novel cell concept that combines the tandem cell approach with the passivated emitter and rear cells (PERC) mainstream technology. As an interface between Si bottom and top cell, we utilize passivating n+-type polysilicon on oxide (POLO) contacts and a p+ poly-Si/n+ poly-Si tunneling junction. Our full area PERC+ Si bottom cells are fabricated within a typical industrial process sequence where the POCl3 diffusion and SiNx deposition are replaced by the POLO junction formation processes. The implied open-circuit voltage iVoc that is measured on these devices reaches up to 708 mV (684 mV) under 1 sun (under filtered spectrum to simulated top cell absorption). On sister cells with planar front side, the respective iVoc values are 718 mV (696 mV). In order to understand the device physics of our ultra-abrupt p+ poly-Si/n+ poly-Si tunneling junction, we determined the carrier lifetime in the poly-Si by time-resolved photoluminescence. The extracted lifetimes of 42–54 ps enter as input parameter for numerical Sentaurus Device simulations. These simulations reveal the importance of band-to-band and trap-assisted tunneling for a low tunneling junction resistivity of 2.95 mΩ·cm2. Experimentally, an upper limit for the combined junction resistance of the p+ poly-Si/n+ poly-Si/SiOx stack of 100 mΩ·cm2 is determined.

We investigate the enhancement in transparency and conductivity of aluminum doped zinc oxide (ZnO:Al) layers upon high-temperature annealing and its impact on contact resistance, as well as, on passivation properties of carrier selective junctions based on doped polycrystalline Si on a passivating silicon oxide (POLO). The temperature stability of these junctions allows annealing of the ZnO:Al/POLO combination up to 600 °C. We prepare ZnO:Al films by dc magnetron sputtering at room temperature. We determine the complex refractive index of ZnO:Al in dependence of post-deposition annealing (PDA) temperature by spectroscopic ellipsometry. High-temperature annealing improves the conductivity and reduces the absorption within ZnO:Al. The optical losses in a ZnO:Al/POLO stack are rather limited by the poly-Si layer than by the ZnO:Al. The sheet resistance improves from roughly 20 000 Ω/sq for 80 nm thick as-deposited ZnO:Al films to 72 Ω/sq after fast firing at 600 °C. At the same time, PDA cures the damage induced in the POLO junctions during ZnO:Al deposition. After PDA with AlxOy capping layers, the passivation quality even surpasses the initial level. A transmission electron microscopy analysis of the interface between the ZnO:Al and the underlying poly-Si reveals the formation of a silicon oxide like interfacial layer after PDA at 400 °C. This interfacial layer causes a high contact resistivity of the metal/ZnO:Al/POLO-junction and could limit the thermal budget for cell processing. Our results indicate that after successful process adjustment, ZnO:Al could substitute In-based transparent conductive oxides on POLO cells for cost reasons, as well as, enable a high efficiency potential.

We report on an industrial large area, screen-printed, double-side contacted cell with polysilicon on oxide (POLO) junctions on both sides and an energy conversion efficiency of 22.3% (A = 244.15 cm 2, Voc = 714 mV, FF = 81.1%, Jsc = 38.5 mA/cm2, measured in-house). This cell shows an extraordinarily low series resistance below 0.05 Ω cm2. This confirms the low specific junction resistance observed recently for POLO junctions. The present cell suffers from 1) low short-circuit current due to parasitic absorption in the rather thick poly-Si (30 nm), as well as in the indium tin oxide, 2) deterioration of the recombination behavior upon sputter deposition of a transparent conductive oxide (TCO), and 3) shunts near the edge due to nonadapted TCO edge exclusion. We address all of these limitations experimentally. In particular, we developed a plasma-enhanced chemical vapor deposition process for ZnO:Al, which does not compromise the passivation of the POLO junctions underneath. An estimation of the efficiency potential (based on the two-diode model and the assumption that all these building blocks can be successfully combined on a cell level) shows that 25.3% can be achieved with this cell concept. We also look into potential cost advantages of the POLO junction scheme for this cell structure, such as the usage of p-type Cz-Si material and the omission of Ag fingers.

The determination of the spectral responsivity is an essential part of solar cell characterization. Since solar simulators only approximate the reference spectrum, a spectral mismatch correction has to be performed. This correction procedure requires spectral responsivity data. Apart from the complete differential spectral responsivity procedure, the IEC 60904-8 standard defines four simplifications. In this paper, we provide information on the variations in the spectral responsivity curves for these simplifications. We show that for nonlinear front junction cells, deviations predominantly occur at wavelengths above 700 nm and become largest around 1000 nm. While we found a maximum deviation of 30% for the simplification with lowest requirements in bias irradiance, all other simplifications yield deviations below 10%. For a nonlinear cell measured relative to a world photovoltaic scale reference cell using a class A solar simulator, this transfers to a deviation below 0.01% in the spectral mismatch factor. If one depends on the use of a simplification, we recommend using the multicolor approach. Even though the singlecolor approach might yield lower deviations, this approach requires knowledge about the maximum in the spectral responsivity, which is not generally known in advance of the measurement. Accepting a slightly higher deviation, the white bias approach is a recommendable alternative.

Many manufacturers choose the passivated emitter and rear cell (PERC) approach in order to surpass the 20% cell efficiency level in mass production. In this paper, we study the efficiency potential of the PERC approach under realistic assumptions for incremental improvements of existing technologies by device simulations. Based on the most recent published experimental results, we find that the PERC structure is able to reach about 24% cell efficiency in mass production by an ongoing sequence of incremental improvements. As a guideline for future developments, we provide a method to improve cell efficiency most effectively by monitoring the current losses at the maximum power point. By means of numerical device modeling, we identify some key technologies toward 24% efficient PERC cells and provide its technology-related target requirements.

Within the silicon photovoltaics (PV) community, there are many approaches, tools, and input parameters for simulating solar cells, making it difficult for newcomers to establish a complete and representative starting point and imposing high requirements on experts to tediously state all assumptions and inputs for replication. In this review, we address these problems by providing complete and representative input parameter sets to simulate six major types of crystalline silicon solar cells. Where possible, the inputs are justified and up-to-date for the respective cell types, and they produce representative measurable cell characteristics. Details of the modeling approaches that can replicate the simulations are presented as well. The input parameters listed here provide a sensible and consistent reference point for researchers on which to base their refinements and extensions.