Alexander E. Braun, Senior Editor -- Semiconductor International, 2/20/2008

With photovoltaic (PV) technology rapidly approaching grid parity — some predict in two years — this old/new technology finally seems prepared to take center stage as one of the more important renewable energy sources. Grid parity is achieved when it is possible to buy and install a PV system and, over its life, the cost of the electricity generated is no greater than if it had been bought from conventional sources, such as a utility.

At the recently held SEMI (San Jose) Silicon Valley Lunch Forum, the subject of “Solar: Will It Light Up the Semiconductor Market? What are the Real Opportunities?” was explored.

Paul Basore, vice president, technology, REC Group (Moses Lake, Wash.), one of the two presenters, indicated that there are currently three dominant PV module markets in the world. “Japan was the first to develop, next came Europe — primarily Germany — and now expanding to other EU nations, with California considered as the next great market,” he said.

Grid parity is defined not only by a PV system’s cost, installation and service lifetime, but also by some variable factors, such as location. A system in a sunny region, such as Phoenix, would attain grid parity easier that one in Seattle. “In 2003, we came up with a cost of electricity to the end user — if installed in a favorable area — of between $0.28 and $0.35 per kWh,” Basore said. “This allowed us to project forward in time and predict where cost reductions could be had over the entire value chain. We saw that costs would quickly drop over the next few years, but that this would level out due to fundamental limitations in the technology. Three years later, we revisited this projection and found that costs had dropped, although not as much as projected. Then, last year, we found we were back on the original curve because, between 2005 and 2007, costs fell faster than expected. Some of the fundamental limitations envisioned in 2003 weren’t as fundamental. Now, the cost for the end user seems very favorable, as it continues to drop linearly. If current R&D is successful, by 2010 further significant reductions might occur.”

Globally, over $100B of PV electricity is purchased yearly at prices in excess of $0.20 per kWh. Thus, even today costs on installed systems could be met (in reasonably sunny locations). However, not all of this $100B in electricity originates from favorable locations. The goal is to reduce costs to drive up profit margins and cause faster growth. A way of determining the PV market size is to figure out how fast the demand for electricity is growing in different parts of the world, and how the amount of manufactured PV compares with that. “If all the PV systems produced in the world in 2009 were taken to California, all its electrical needs would be met without new power plants. By 2012, all of Japan’s power needs would be met, and by 2014, world PV production could meet America’s electricity needs,” Basore said.

PV’s energy payback has dogged the industry from the beginning. Originally, it was estimated that a PV system had to be in the sun seven years before it produced the initial energy required to manufacture and install it. “PV systems last anywhere from 20 to 30 years, but if one-third of that represents restoring the energy used to produce them, this isn’t a particularly good option,” Basore said, adding that by the late 1990s, that number was halved to about three-and-a-half years, with a current industry average now being two years. “Taking into account the technology’s expected developments, by 2010 top systems should have an energy payback time of under a year. This is only for wafer-based modules; thin film will probably do even better.” For significant cost reductions, it is not enough to address one component — all steps and values must be considered; for example, reducing the cost of burning silicon into the wafers. This requires thinner wafers. “We’re now making wafers approximately 180 µm thick by cutting them with wires that are approximately 140 µm thick, so much is wasted in silicon sawdust,” Basore said. Experiments are underway to produce 160-µm-thick wafers cut with 120-µm-thick wire. To meet 2010 requirements, only wafers 120-160 µm thick should be used. At those dimensions, traditional cell fabrication processes encounter problems; specifically, issues with the screen printing of the metal paste and subsequent firing — wafers that thin tend to warp when heated. Cell process must be redesigned and efficiency increased because, currently, cells are only about 15-16% efficient in yielding modules that are only 14% efficient.

The industry’s argument in favor of thin films instead of wafers is less material use. “That’s misinformed — silicon makes up a quarter of the Earth’s crust,” Basore said. “We have already enough refining capacity to supply the needs of any thin-film silicon technology. The key point to focus on thin-film opportunities is reducing the cost depositing the material.” To be competitive, thin films must be low in cost and offer a reasonably high efficiency. Because they are less efficient, for any given system with a certain power rating, a larger area is needed, requiring more land or roof space, affecting price. Also, many thin-film technologies come in small modules, requiring more pieces, increasing installation costs.

Finally, thin films lack a durability track record in the field, while wafer modules have been fielded in significant numbers for over 30 years. Eventually, thin films will catch up. Meanwhile, the market perceives the difference, resulting in another barrier to wider use. Lastly, there are other, non-silicon technologies based on elements such as indium and tellurium, which have availability questions.

Although government support continues being an important factor in mainstreaming PV, huge investments are taking place, both in private funding and organic growth in companies already involved. Over the next couple of year, political incentives will begin losing importance and the industry will become less dependent on them.

According to Mark Pinto, CTO and vice president, general manager of Applied Materials Energy and Environmental Solutions (Santa Clara, Calif.), everything orbits around cost. “We’re asked why go into solar since it is all about making cheap equipment. In our industry, semiconductors, cost has been a predominant factor — look at the cost per transistor, per bit. In 2006, an iPod that cost about $200 would have cost $1B in 1976 and would have been very large. When costs fall, markets open. This has been fundamental both to the IC and the display business. In the 1990s, a purchase order for a flat panel monitor was difficult to get through; today you no longer see CRTs and a 40-inch HDTV is $1000.”

From a manufacturing perspective, cost is composed of functions. This can be a bit, display area or watts per solar. For ICs, it is how many transistors can be put in without radically increasing process costs. Most markets are at least equally driven by process cost per unit area. Displays is almost 100% process cost per unit area; pixels per unit area do not drive cost, it is how much it costs to process a square area of display. This gets driven by using large substrates.

“Big substrates require big equipment,” Pinto said. “Cluster tools inline, web tools, everything needed to lower unit area cost. However, material uniformity and integrity is still on a nanometer scale; even for architectural glass, you can see the difference if you don’t have uniformities on the order of percentages of tens of nanometer films, which require good quality control and throughput. The same is with solar. Over the past 20 years, module cost per watt has changed. In the 1980s, it was over $20 a watt. We were 20 times further from grid parity than now. As this nears $2.00 a watt, grid parity is obtained in some areas and markets begin taking off.”

Solar looks at production cost per watt as the cost per unit area, as well as efficiency. While there is no 100% efficiency, cost per unit area can be driven down. Scale is important. It took some 20 years to go from a 1 MW fab, half a megawatt line peak per year, to 100 MW fabs today, and to gigawatt fabs in the near future. Just as with ICs, scale drives down costs, all driven by large equipment platforms.

“There are multiple technologies — crystalline and thin film — that have different applications today,” Pinto said, “crystalline being best for residential rooftops, thin film for ground-mounted commercial installations. We’re working on both because they will have significant impact in different segments of the market.”

Applied is also focused on thinner wafers and less waste. “In a 50 MW fab, there are multiple lines, mostly process equipment and automation between steps,” Pinto said. “A key step is laying the antireflective coating and passivation, which is done with a PVD nitride tool. High throughput with a deposition system is 3000 to 5000 wafers per hour; not done by any single PVD or PECVD system. In terms of scaling in the materials use area, a GW fab uses about 200 times a 300 mm fab’s silicon area.”

When considering efficiency, watts/m² is crucial. This is attained through more complicated cell technologies, which means more processes and more materials. Another way is through yield and better controlled processes to get more watts out of a fab. This is extendable to new materials and higher efficiency applications.

“With thin film, you start with glass, put an interlayer transparent conducting material, the absorber is silicon p-i-n or more complicated structures, and a back metallization,” Pinto said. “We’re installing factories to do this. The panel our customers will provide is 5.7 m², slightly bigger than a Gen 8.5 display substrate. When cut into quarters, you get the standard size thin film module — all this means lower cost. Big modules are cheaper to install, they have fewer junction boxes, less wires. In a ground-mounted system, they require one-quarter of the wires, less labor. This savings is the equivalent of two to three percentage points more in conversion rate. With silicon panels that size, it’s the equivalent of three percentage points more efficiency. We’re used to Moore’s Law, big changes in transistor costs. But PV is within factors of grid parity in all of Europe, so small tens of percentage changes can be significant.”

Large panels also enable building-integrated PV, where the material is part of the construction, instead of a later add-on to an existing structure — whether as glass on a building or a facade.

It would appear that at least a part of the semiconductor industry is ready to transition over to PV. Pinto sees no major difficulties. “In Applied, we have some 800 people working full-time on solar. Over 400 came from different parts of the company, not just acquisitions. We’ve converted 400 people into PV experts and for us, it has not been a big change — it’s still about productivity.”

Paradoxically, PV’s future over the next few years will be determined not so much by its adoption by a continent — Europe, in this case — but by that of a state. As Basore put it, “Although PV is bigger overseas than in the U.S., everyone’s radar there is on California, which is widely viewed as the next great market area. It may be difficult to make inroads into the European PV business, but I want to be at the frontline for the industry’s growth in California.”