As the semiconductor industry enters its fifth decade, one of its greatest advantages over other technology-driven fields is infrastructure. The integrated circuit (IC) industry collectively has an enormous amount of experience in making silicon-based CMOS manufacturing work. Alternative device technologies, from germanium channels to quantum dots, must operate within this structure and are unlikely to duplicate it. In fact, the sheer size of the CMOS industry is a nearly insurmountable barrier to alternative approaches.
As the IC industry matures, companies are seeking to apply these robust process technologies in new areas. One of the most successful of these is flat panel displays (FPDs). Once confined to wristwatches and portable computers, liquid crystal displays have now pushed CRTs out of offices and are starting to displace them from living rooms as well.
Many companies hope that photovoltaic devices represent the next great market for semiconductor manufacturers. After all, solar cells consume about 8 million sq. m. of silicon per year, while integrated circuits use only about 5 million sq. m. Shouldn’t the world’s leading experts in silicon processing be able to help photovoltaics manufacturers cut costs and improve efficiency?
Driven by high oil prices and government incentives, the photovoltaics industry is enjoying unprecendented growth, in excess of 30% per year. At the same time, the central challenge for photovoltaics manufacturers is cost reduction. Materials costs, manufacturing costs, and assembly costs all must come down if solar energy is to compete with other sources of electricity.
As Alain Harrus explained in the April issue of Solid State Technology (“Semiconductor processing technologies find a second life in photovoltaics,” p. 35), a variety of companies from the IC industry have found that silicon photovoltaics-currently the dominant technology-uses familiar materials and technologies. These firms have had some success in leveraging their hard-earned expertise in silicon processes. In March, Applied Materials announced that it has been selected to build and install a 200MW capacity thin film silicon PV production line for Moser Baer India Ltd.
Yet many of the most significant developments in photovoltaics lie far from the silicon mainstream. Tight supplies of silicon have driven a surge of interest in alternative materials. Aggressive cost reduction targets and increasing demand for portable electronics are motivating development of low-cost photovoltaic materials that can be printed on flexible substrates using roll-to-roll processes. Though these designs cannot yet match the efficiency of bulk silicon solar cells, they are finding their way into a variety of applications, from roofing materials to tents, that are inaccessible to bulkier designs.
How photovoltaic devices work
Silicon photovoltaic devices, like ICs, depend on the behavior of carriers in the vicinity of p-n junctions. In solar cells, the carriers are excited by photon absorption, rather than being injected from elsewhere in the circuit or thermally generated.
Figure 1. Solar spectrum and bulk silicon absorption spectrum. (Courtesy: Energy Research Centre of the Netherlands)
Photonic excitation creates carriers randomly throughout the material. Carriers that fall within, or diffuse to, the depletion region surrounding a junction are swept to the n-type side (holes) or the p-type side (electrons). Those that lie outside the junction region will eventually relax to the ground state, dissipating their energy as heat.
The photocurrent consists roughly of those carriers that reach the junction region before returning to the ground state. Carrier mobility and carrier lifetime are thus key contributors to internal quantum efficiency, the fraction of absorbed photons that actually excite free carriers. However, other losses reduce the net cell efficiency substantially.
For instance, most materials absorb only a portion of the solar spectrum, either reflecting the rest or allowing it to pass through. To excite free carriers, an absorbed photon’s energy must exceed the band gap, adjusted by the ionization energy of any dopants. Figure 1 illustrates the absorption spectrum of bulk silicon relative to the solar spectrum. Like most photovoltaic materials, bulk silicon has very little absorption in the infrared range.
When incident light falls within the material’s absorption spectrum, the absorption coefficient measures the thickness over which it will be completely absorbed. In thick layers, like in bulk silicon photovoltaics, almost all photons with appropriate energy will be absorbed. In thin layers with low absorption coefficients, reflectors and light trapping schemes may be used to increase absorption.
Not all of the free carriers generated by absorbed photons will actually reach the terminals of the solar cell. Some will recombine at defects or impurity sites, while others will dissipate their energy when they encounter resistance within the cell or at the metallic contacts to it. Indium tin oxide (ITO), a transparent oxide commonly used as a top electrode in thin film PV devices and FPDs, is a relatively poor conductor that tends to degrade net cell efficiency. Bulk silicon photovoltaics often use a grid of metallic contacts instead. Being opaque, these contacts reduce light absorption, but improve contact resistance.
The metal/semiconductor junctions at device contact points are an important source of resistive losses, especially for photovoltaic materials other than silicon. If the work function of the contact metal is close to the energy of the semiconductor’s conduction band, then electrons can easily flow across the junction. If the work function is close to the energy of the valence band, then holes can flow easily. If there is a substantial misalignment between the energy of the majority carriers in the semiconductor and the work function of the metal, then the junction will act as a diode preventing the free flow of carriers.
When all of these factors are considered, even the best bulk silicon cells achieve net efficiency of only around 25%. Shockley and Queisser  argued that the theoretical maximum efficiency for single-junction bulk silicon cells is in the neighborhood of 31%. Additional losses in the module assembly connecting the cell to the intended electrical load can cut this value by 10% or more. Designs that use several junctions to overcome thermodynamic limits have achieved cell efficiencies as high as 36%. Still, bulk silicon photovoltaics are unlikely to achieve significant efficiency improvements. Instead, photovoltaics manufacturers are concentrating on cost reduction through such means as more efficient use of raw materials and economies of scale due to increasing volumes.
Thin film PV designs are attractive because they do not depend on silicon wafers, the most expensive component of conventional solar cells. Instead, manufacturers hope to deposit thin film photovoltaics on metal foil, plastic film, and other inexpensive, flexible substrates.
CIGS leads the pack of thin films
Thin film PV designs cannot yet achieve cell efficiencies equal to that of bulk silicon, but the best are starting to come close. Copper indium gallium diselenide (CIGS), which has demonstrated 19% efficiency under laboratory conditions, holds the record for thin film cell efficiency. As with bulk silicon, integration into a complete module reduces the net efficiency substantially. So far, the best CIGS modules achieve efficiencies of around 13% in testing at the National Renewable Energy Laboratory. CIGS achieves such high efficiency in part because of its high absorption coefficient. More than 99% of the light striking the film will be absorbed within the first micron or so . CIGS cells also offer a tunable band gap as varying the indium:gallium ratio can match the bandgap to the illumination expected in a particular application. For example, indoor applications such as retail shelving can expect a different illumination spectrum from outdoor applications.
Figure 2. CIGS solar cell structure. (Courtesy: International Solar Electric Technology Inc.)
Typical CIGS cells (Fig. 2) depend on a heterojunction between an n-type CdS layer and a light-absorbing p-type CIGS layer. The high CIGS absorption coefficient helps here as well: a thin absorption layer reduces the risk that carriers will be lost to recombination before they can reach the p-n junction. Unfortunately for the commercial prospects of CIGS, CdS is toxic and a carcinogen. Though proper cell design minimizes the risk of environmental contamination, use of this material imposes additional costs and regulatory constraints on the entire CIGS supply chain. Use of toxic materials is especially problematic in a product that claims environmental friendliness as one of its key advantages.
Laboratory-scale CIGS cells are usually fabricated by co-evaporation of the component elements. This approach wastes a great deal of material, coating the walls of the process chamber as well as the target substrate. The need for a vacuum chamber also substantially increases equipment costs.
Commercialization of CIGS technology is still in its early stages. Statistics for CIGS sales are hard to find, but all thin film technologies combined command only a 7% share of the global photovoltaics market. Thin film silicon accounts for more than 80% of that fraction. Still, companies are using a variety of novel manufacturing methods for CIGS cells. GlobalSolar offers CIGS cells on stainless steel foil or polyimide film, deposited in thousand-foot lengths using a roll-to-roll process. Nanosolar claims to have developed printable CIGS inks, based on nanoparticles with the desired ratio of components. Nanosolar is currently building what it claims will be the world’s largest solar cell plant in San Jose, CA, and plans to begin production this year.
Though the manufacturing processes used in CIGS commercialization are unique, the cells themselves behave in familiar ways and depend on much the same underlying physics as conventional silicon cells. On the far horizon, other potential designs propose something entirely new. The DSSC, invented in 1991, depends on a photoelectrochemical process that is more closely related to photosynthesis than to conventional p-n junctions .
In conventional photovoltaic devices, the same material performs both light absorption and carrier transport. In a silicon cell, the silicon both supplies the photonically excited carriers and transports them to the external circuit. CIGS cells use two transport materials, but the CIGS layer is still responsible for light absorption.
DSSC: new approach to solar electricity
In DSSCs, in contrast, these two roles are performed by two different materials. First, a dye layer-usually based on ruthenium polypyridine compounds-absorbs the incident light. It then ejects electrons from the dye into TiO2 particles (see Fig. 3).
Figure 3. Dye-sensitized solar cell structure. (Courtesy: Swiss Federal Institute of Technology)
Electrons in the TiO2 then percolate to the anode, leaving the dye molecules in an oxidized, electron-poor state. Meanwhile, electrons entering the cell from the cathode, encounter a triiodide/iodide electrolyte. A platinum coating on the cathode catalyzes the conversion of triiodide (I3-) into iodide (I-), liberating an electron. These electrons reduce the dye to its original state, regenerating it for further use.
While charge separation in conventional PV cells depends on the contact potential at the p-n junction, in DSSC cells the driving force for separation is the difference between the TiO2 conduction band and the electrolyte’s chemical potential. This charge separation mechanism greatly reduces the chance that an electron-hole pair will recombine before it can contribute to the photocurrent. Once electrons are injected into TiO2, very few free holes are available for them to recombine with. For this reason, defects and impurities in TiO2 do not act as carrier recombination sites. Thus, DSSCs can tolerate more relaxed materials specifications than silicon cells.
The best DSSCs have demonstrated efficiencies around 11%, but this value is still very low for most commercial applications. The problem of efficiency in DSSCs is largely a problem of the preparation and structure of the TiO2 layer, according to Janne Halme at the Helsinki U. of Technology .
The lifetime of free carriers in the dye is limited, so the dye molecules in which the electrons originate must be adjacent to TiO2 particles. At the same time, a thicker dye layer maximizes light absorption. These competing goals are achieved by means of a nanoporous TiO2 sponge immersed in dye. The manufacturing process begins by coating the substrate with a colloidal slurry of TiO2. Sintering drives off solvents and other components of the slurry, leaving a nanoporous network. Immersion allows the dye to penetrate into the pores of the “sponge.” The surface area of the sponge determines the success or failure of dye uptake. At the same time, the pore size distribution within the sponge controls diffusion of the electrolyte, while the particle size distribution determines the cell’s optical properties. Finally, the interconnections between adjacent particles determine the rate at which electrons percolate through the structure.
One alternative, proposed by Matt Law and others at the U. of California (Berkeley), replaces the TiO2 sponge with a dense array of ZnO nanowires. The dye could penetrate such an array by capillary action. At the same time, conduction within the ordered nanowire structures is much more efficient than carrier diffusion through the particle sponge .
Neither TiO2 nor ZnO is a particularly good conductor. Oxide conductors are necessary because the aggressive electrolyte chemistry would corrode a metallic material. The volatile chemistries used pose other obstacles to commercialization as well. Especially at elevated temperatures-rooftop solar installations must tolerate temperatures in excess of 80°C-the electrolyte and dye solvents can attack the cell sealants and escape into the atmosphere. They are toxic, and their loss also degrades cell performance. A gelatinous, rather than liquid, electrolyte would simplify the manufacturing challenges associated with DSSC. A number of companies have licensed the original DSSC patents, but prospects for commercialization of the technology remain unclear.
In many ways, conventional solar cells and thin film cells target different design niches with different requirements. In large industrial installations, efficiency and lifetime are paramount, and silicon cells are likely to retain their advantage. Yet many emerging applications emphasize cost and portability. A tent made with solar materials might reduce the user’s need for generators or batteries. Solar-powered shelving might give a retailer more design flexibility, liberating shelf sensors from the building wiring. CIGS designs are already making gains in these markets, while DSSCs offer great promise for future applications.
- W. Shockley, H.J. Queisser, J. Appl. Phys., Vol. 32, No. 3, p. 510, 1961.
- J.E. Jaffe and Alex Zunger, “Theory of the Band-gap Anomaly in ABC2 Chalcopyrite Semiconductors,” Phys. Rev. B, Vol. 29, No. 4, pp. 1882-1906, 1984.
- Brian O’Regan, Michael Graetzel, “A Low-cost, High-efficiency Solar Cell-based on Dye-sensitized Colloidal TiO2 Films,” Nature, Vol. 353, No. 24, pp. 737-740, Oct. 24, 1991.
- Janne Halme, “Dye-sensitized Nanostructured and Organic Photovoltaic Cells: Technical Review and Preliminary Tests,” MS Thesis, Helsinki University of Technology, Ch. 4, 2002.
- Matt Law, et al., “Nanowire Dye-sensitized Solar Cells,” Nature Materials, Vol. 4, pp. 455-459, 2005.
Katherine Derbyshire is a contributing editor at Solid State Technology. She received her engineering degrees from the Massachusetts Institute of Technology and the U. of California, Santa Barbara. She is the founder of consulting firm Thin Film Manufacturing, email@example.com, http://www.thinfilmmfg.com.