W.C. Sinke

Photovoltaic materials: Present efficiencies and future challenges

Surveying the solar cell landscape The rate of development and deployment of large-scale photovoltaic systems over recent years has been unprecedented. Because the cost of photovoltaic systems is only partly determined by the cost of the solar cells, efficiency is a key driver to reduce the cost of solar energy. There are several materials systems being explored to achieve high efficiency at low cost. Polman et al. comprehensively and systematically review the leading candidate materials, present the limitations of each system, and analyze how these limitations can be overcome and overall cell performance improved. Science, this issue p. 10.1126/science.aad4424 BACKGROUND Photovoltaics, which directly convert solar energy into electricity, offer a practical and sustainable solution to the challenge of meeting the increasing global energy demand. According to the Shockley-Queisser (S-Q) detailed-balance model, the limiting photovoltaic energy conversion efficiency for a single-junction solar cell is 33.7%, for an optimum semiconductor band gap of 1.34 eV. Parallel to the development of wafer-based Si solar cells, for which the record efficiency has continually increased during recent decades, a large range of thin-film materials have been developed with the aim to approach the S-Q limit. These materials can potentially be deposited at low cost, in flexible geometries, and using relatively small material quantities. ADVANCES We review the electrical characteristics of record-efficiency cells made from 16 widely studied photovoltaic material geometries and illuminated under the standard AM1.5 solar spectrum, and compare these to the fundamental limits based on the S-Q model. Cells that show a short-circuit current (Jsc) lower than the S-Q limit suffer from incomplete light absorption or incomplete collection of generated carriers, whereas a reduced open-circuit voltage (Voc) or fill factor (FF) reflects unwanted bulk or interfacial carrier recombination, parasitic resistance, or other electrical nonidealities. The figure shows the experimental values for Jsc and the Voc × FF product relative to the S-Q limiting values for the different materials. This graph enables a direct identification of each material in terms of unoptimized light management and carrier collection (Jsc/JSQ < 1) or carrier management (Voc × FF/VSQ × FFSQ < 1). Monocrystalline Si cells (record efficiency 25.6%) have reached near-complete light trapping and carrier collection and are mostly limited by remaining carrier recombination losses. In contrast, thin-film single-crystalline GaAs cells (28.8%) show only minimal recombination losses but can be improved by better light management. Polycrystalline CdTe thin-film cells (21.5%) offer excellent light absorption but have relatively high recombination losses; perovskite cells (21.0%) and Cu(In,Ga)(Se,S)2 (CIGS) cells (21.7%) have poorer light management, although CIGS displays higher electrical quality. Aside from these five materials (Si, GaAs, CdTe, CIGS, perovskite) with efficiencies of >20%, a broad range of other thin-film materials have been developed with efficiencies of 10 to 12%: micro/nanocrystalline and amorphous Si, Cu(Zn,Sn)(Se,S)2 (CZTS), dye-sensitized TiO2, organic polymer materials, and quantum dot solids. So far, cell designs based on these materials all suffer from both light management and carrier management problems. Organic and quantum dot solar cells have shown substantial efficiency improvements in recent years. OUTLOOK The record-efficiency single-crystalline materials (Si, GaAs) have room for efficiency improvements by a few absolute percent. The future will tell whether the high-efficiency polycrystalline thin films (CdTe, CIGS, perovskite) can rival the efficiencies of Si and GaAs. Because the cost of photovoltaic systems is only partly determined by the cost of the solar cells, efficiency is a key driver to reduce the cost of solar energy, and therefore large-area photovoltaic systems require high-efficiency (>20%), low-cost solar cells. The lower-efficiency (flexible) materials can find applications in building-integrated PV systems, flexible electronics, flexible power generation systems, and many other (sometimes niche) markets. High-efficiency (>20%) materials find applications in large-area photovoltaic power generation for the utility grid as well as in small and medium-sized systems for the built environment. They will enable very large-scale penetration into our energy system, starting now and growing as the cost per kilowatt-hour is reduced further by a factor of 2 to 3. This can be achieved by nanophotonic cell designs, in which optically resonant and nonresonant structures are integrated with the solar cell architecture to enhance light coupling and trapping, in combination with continued materials engineering to further optimize cell voltage. Making big steps forward in these areas will require a coordinated international materials science and engineering effort. Limiting processes in photovoltaic materials. An efficient solar cell captures and traps all incident light (“light management”) and converts it to electrical carriers that are efficiently collected (“carrier management”). The plot shows the short-circuit current and product of open-circuit voltage and fill factor relative to the maximum achievable values, based on the Shockley-Queisser detailed-balance limit, for the most efficient solar cell made with each photovoltaic material. The data indicate whether a particular material requires better light management, carrier management, or both. Colors correspond to cells achieving <50% of their S-Q efficiency limit ηSQ (red), 50 to 75% (green), or >75% (blue). Recent developments in photovoltaic materials have led to continual improvements in their efficiency. We review the electrical characteristics of 16 widely studied geometries of photovoltaic materials with efficiencies of 10 to 29%. Comparison of these characteristics to the fundamental limits based on the Shockley-Queisser detailed-balance model provides a basis for identifying the key limiting factors, related to efficient light management and charge carrier collection, for these materials. Prospects for practical application and large-area fabrication are discussed for each material.

Density of amorphous Si

The density of amorphous Si has been measured. Multiple Si implants, at energies up to 8.0 MeV, were made through a contact mask to produce alternating amorphous/crystalline Si stripes with amorphous thicknesses up to ∼5.0 μm. For layers up to 3.4 μm (5 MeV), the amorphous Si is constrained laterally and deforms plastically. Above 5 MeV, plastic deformation of the surrounding crystal matrix is observed. Height differences between the amorphous and crystalline regions were measured for as‐implanted, thermally relaxed, and partially recrystallized samples using a surface profilometer. Combined with ion channeling measurements of the layer thickness, amorphous Si was determined to be 1.8±0.1% less dense than crystalline Si (4.90×1022 atom/cm3 at 300 K). Both relaxed and unrelaxed amorphous Si show identical densities within experimental error (<0.1% density difference).

Alkaline Etching for Reflectance Reduction in Multicrystalline Silicon Solar Cells

The reflection reducing properties of alkaline-etched multicrystalline wafers are investigated experimentally for high concentration saw-damage etching and low concentration texture etching. Saw-damage etch textures are too flat for multiple bounce reflectance in air, with only 1.6% of the multicrystalline wafer surface calculated to have facet tilt angles above 45° whereby double-bounce reflectance is guaranteed. Texture etching yields 3% lower reflectance in air, due to high angled ~up to 54.7°! pyramidal structures on near ~100! orientations, whereby 13% of the multicrystalline etch surface has tilt angles above 45°. However, under encapsulation, light is coupled more effectively into the silicon; reflectances for the saw-damage and texture-etched wafers compare only 7 and 5.5% higher, respectively, than upright pyramid textures on monocrystalline silicon~100!, compared to 18 and 15% higher in air. This is because a far larger proportion of the multicrystalline wafer ~around 40% for the two etches! has tilt angles above 20.9° whereby escaping light is totally internally reflected at the glass-air interface. For texture etching, not only

Defect states of amorphous Si probed by the diffusion and solubility of Cu

111% planes are stable to etching but the whole range of

SIMPLIFIED EVALUATION METHOD FOR LIGHT-BIASED EFFECTIVE LIFETIME MEASUREMENTS

XXY% crystallographic planes between these and

Improved treatment of the strongly varying slope in fitting solar cell I-V curves

110% orientations, contrary to the accepted texture etching theory.

Raman study of de‐relaxation and defects in amorphous silicon induced by MeV ion beams

The diffusivity and solubility of Cu impurities have been measured in different structural states of amorphous Si (a‐Si) formed by MeV Si implantation. The 2.2‐μm‐thick a‐Si layers were first annealed (structurally relaxed) at 500 °C and then implanted with 200 keV Cu ions, returning a 300‐nm‐thick surface layer to the as‐implanted state. After diffusion at temperatures in the range 150–270 °C, we observe solute partitioning at a sharp phase boundary between the annealed and Cu‐implanted layers, the partition coefficient being as large as 8.2±1.3. The diffusion coefficient in annealed a‐Si is 2–5 times larger than in as‐implanted a‐Si, with activation energies of 1.39±0.15 and 1.25±0.04 eV, respectively. The data show quite strikingly the role which defects can play in the a‐Si structure.

Metallisation patterns for interconnection through holes

In this letter, we present a simplified evaluation method for light-biased photoconductance decay measurements. The measured effective lifetime is shown to be a differential quantity τeff,d, which may differ significantly from the actual effective lifetime τeff. However, the actual effective lifetime can be approximated by integrating τeff,d directly over the incident power density of the bias light. The quality of the approximation depends mainly on the surface recombination velocity and the wavelength of the used bias light. However, the inaccuracy remains well below 10% for most practical cases.

Combined impurity gettering and defect passivation in polycrystalline silicon solar cells

Straightforward least squares fitting of I-V curves leads to nonoptimal fits: residuals around and above the open-circuit voltage dominate the fit, leading to a bad fit at the maximum power point and lower voltage values. To deal with this problem the authors have resorted to using weighting functions or to minimizing the area between data and fit instead of the least squares procedure. Both approaches lack a sound statistical basis. Voltage noise has a big influence on fitting due to the steep slope of an I-V curve for higher voltage values. For this reason we have used orthogonal distance regression (ODR), which is a mathematical method for fitting measurements with errors in both voltage- and current measurements. It allows for computing both the I-V curve parameters and their uncertainties.

Gettering in polycrystalline silicon solar cells

Raman spectroscopy is used as a probe of the state of amorphous Si (a‐Si) and damaged crystalline Si. MeV ion beams have been used to irradiate structurally relaxed a‐Si. When the density of Si atoms displaced by nuclear collisions exceeds 5%, the a‐Si is ‘‘de‐relaxed’’, and thus returns to its as‐implanted state. This behavior is an indication that point defect complexes exist in a‐Si and play an important role in the process of structural relaxation.