Semiconductor processing technologies find a second life in photovoltaics
04/01/2007
More energy from sunlight hits the Earth in just one hour than the vast amount of energy consumed by all humans in an entire year. Solar energy, as a resource, dwarfs all other sources of energy. With increasing attention toward carbon-neutral energy production, the interest and deployment of solar electricity or photovoltaic (PV) technology-the capture and conversion of the sun’s photons into electrons-is growing rapidly.
In 2005, 1.7 gigawatts (GWs) of PV modules were installed worldwide, generating more than US$12 billion in revenue. This number is expected to rise to 10.4GW and US$72 billion by 2010. The end goal is simple: generate electricity from the sun at a cost that is equal to, if not less than, the conventional methods used to produce grid electricity. Such methods-including coal, natural gas, and nuclear energy-are formidable competitors for electric power generation.
The rising demand for PV is driven by multiple trends. Developing nations like China and India are aggressively pursuing alternative energy sources to support their fast-growing economies. To augment their energy-independence initiatives, first-world countries like the US, Germany, and Japan are leveraging their economic and political muscle to offer rebates and other incentives to aspiring PV innovators. And the innovators are emerging en masse, ready to make a meaningful difference and harvest the potential financial windfall represented by what may well be the “next killer app.”
With the world’s most robust economies throwing their weight behind this global effort, it is reasonable to assume that the goal of shifting the world toward low-cost electricity generation is well within reach. Such momentum is largely enabled by the technology foundation established by a different industry more than 30 years ago: global semiconductor manufacturing. The PV industry is being jump-started by the maturation of semiconductor materials and manufacturing processes, as well as by the migration of seasoned semiconductor brains to staff the new space.
PV technology breakthroughs have been occurring at academic and government laboratories such as the National Renewable Energy Laboratory (NREL), which began operating in 1977 as the Solar Energy Research Institute. For the past three decades, these groups have worked on improving the PV cell efficiency-the energy output divided by the incident energy from the sun-and also the manufacturing technologies that would optimize the production of PV cells.
 Figure 1. Best reported efficiencies (% of photons converted into electrons) for various photovoltaic technologies in research. (Source: NREL) |
Figure 1 illustrates the best research cell efficiencies as a function of time for essentially four classes of material: silicon-based, thin-film based, multijunctions (including quantum dots at the research level), and emerging organic materials. For single bandgap materials under “one sun” conditions, the theoretical efficiency of 31% is called the Shockley-Queisser limit for quantum conversion. The only cells above this are the multijunction cells, which are upper-bound by the thermodynamic limit of 68%.
PV manufacturing models
Today’s PV materials, technologies, and manufacturing requirements bear a striking resemblance to the technologies that fueled the embryonic semiconductor manufacturing lines more than 30 years ago. That nascent production environment may well provide a roadmap for how the PV manufacturing environment will evolve over the coming decades.
The IC industry initially managed it all-from design to silicon and on to final testing-under the same roof. As the production process became more complex, device makers focused their energy on the process, outsourcing their equipment, raw materials, testing, and packaging requirements to an emerging supply chain. Further evolution brought us to today’s foundry model-a streamlined advanced manufacturing approach that builds on more than three decades of processing know-how. None of that exists in today’s PV manufacturing world.
While the casual observer of the current PV production environment may find a less sophisticated version of a semiconductor manufacturing fab, there are many similarities:
- Vertical integration. In today’s PV manufacturing lines, for either photocells or modules, the degree of vertical integration is very high. All elements of the manufacturing process are present in the same building, from raw material processing to finished devices and packaging.
- Raw materials. The raw materials are similar, either silicon wafers or thin-film materials for cells. Single crystal silicon manufacturing, polysilicon molding, and slicing thin silicon wafers are in common. Also included are materials such as cadmium telluride and copper-indium-gallium-selenide.
- Processing equipment. Most PV manufacturers build their own equipment or customize existing equipment for their process. The technologies include hot-wall furnaces, physical vapor deposition (PVD), chemical vapor deposition (CVD), evaporation, and nanopowders.
- Interdependence of equipment and processes. There is a strong interdependence between the equipment and the processes. As such, there are generally no “standard” processes. This is especially so for thin-film cells, which require radically different methods to manufacture the photo absorbers.
- IP and expertise. PV factories are run with a significant amount of proprietary know-how. The intellectual property is mostly focused on the methods of manufacturing the cells and modules.
- In-house packaging. The production line’s end products (PV cells) are typically packaged in-house (PV modules) and shipped directly to customers. This is reminiscent of the silicon devices sold decades ago in packaged devices to customers.
PV cell technologies
Figure 2 compares the cross sections of the two distinct architectures for PV cells: single crystal silicon and thin-film cells. In brief, the cells are manufactured on a substrate with the photoabsorber (silicon or thin film, a p-n junction formed therein) sandwiched between two conductive electrodes. The top electrode is transparent to admit the sun’s photons. The photons generate electron-holes pairs that are collected by the two electrodes to produce electrical current.
For silicon cells in today’s production, the thickness of the crystalline silicon is typically 190-210μm. With a strong silicon shortage expected to last for another few years, great effort is being exerted to reduce this thickness to possibly as low as 120μm.
In contrast, the thin-film cells are approximately 2μm in thickness, which lends itself to completely different approaches for manufacturing. The processes used to form the photo absorbers and the p-n junctions are familiar to anybody who has spent time in a semiconductor fab.
 Figure 2. Schematic cross-sections of the two major types of photovoltaic cells a) crystalline silicon, and b) thin-film. |
The majority of PV cells being shipped today are based on silicon technologies, single crystal, polycrystalline, or amorphous. Single crystal cells, with efficiencies as high as 22%, are gaining market share as the cost of manufacturing decreases. The process for the p-n junction formation is quite straightforward, using classic diffusion processes, wet etching, and CVD technologies. The metallization requirements are noncritical at >5μm geometries.
The challenge for manufacturing lies in the handling issue. A simple calculation yields the following: a 25MW annual production facility running 5-in. wafers (3W output at 20% efficiency for single crystal cells), 52 weeks at 80% uptime needs to process approximately 1200 wafers/hr. Today, this is accomplished in one of two ways: either with a large amount of proprietary or modified equipment to obtain the overall throughput, or an entirely new class of processing equipment built for throughput and reliability.
For thin-film based cells (Fig. 2b), the picture is different. The photo absorber such as CIGS (Cu (In, Ga) Se2) or CdTe can be formed by a variety of methods, including physical vapor deposition, nanopowder coating, electroplating, or dielectric bonding. Often, it involves a combination of technologies. No matter which absorber is used, the substrate tends to be stainless steel foil, with molybdenum back contacts typically formed using PVD.
Thin-film cell efficiencies today are in the 9-12% range. Control of the composition and tailoring of the myriad materials concentrations are critical to improving beyond the current best demonstrated ~19% efficiency. Because of the nature of the substrate and the low film thickness, the manufacturing methodology lends itself well to a roll-to-roll approach.
Again, this is accomplished with equipment that is customized to work with proprietary processes. Regardless of the manufacturing approach, no cleanrooms are required, and the need for metrology is unknown. Moreover, the industry is only beginning to investigate the importance of defectivity.
The PV industry is beginning its transition to high-volume manufacturing. As the costs decrease and the efficiencies rise, the volumes will increase. This is a highly vertically integrated industry that leverages many technologies more commonly used in the semiconductor industry.
As the PV industry matures over the next 20 years, it could become larger than today’s semiconductor industry. As such, it has the potential to provide a “second life” for conventional semiconductor manufacturing technologies.
Alain S. Harrus received his BS in mathematics and his MS in physics from the U. of Paris XI at Orsay, France, and his PhD in experimental solid-state physics from Temple U. in Philadelphia. He has spent more than 25 years in the semiconductor industry as a technologist and an investor in high-growth early-stage companies. He is a partner at San Francisco venture capital firm, Crosslink Capital, Two Embarcadero Center, Suite 2200, San Francisco, CA 94111; ph 415/617-1800, e-mail [email protected].