Polysilicon forecast is cloudy
for solar cell, IC industries
PRETTY POLY: The polysilicon market’s double-digit growth in both volume and revenues, fueled by increased demand by the solar cell industry, could lead to worsening shortages of the material. (click image to enlarge)
Here comes the sun. There goes the polysilicon. A dark cloud hangs over the solar cell industry’s aggressive growth plans because of a shortage of much-needed silicon wafers. The tight supply also has semiconductor manufacturers casting a wary eye at their bottom lines as prices have risen significantly.
Users can expect the skies to clear in the next few years when polysilicon producers unveil production ramp-ups to meet the strong demand for substrates. For the moment, though, rapid growth in the solar, or photovoltaic (PV), cell industry will remain constrained, while chipmakers contend with higher prices and a drizzle of uncertainty as they grapple with the 300-mm transition.
“There is definitely a shortage worldwide,” acknowledges Tak Takahashi, senior marketing representative for Hemlock Semiconductor, a manufacturer of polycrystalline silicon in Hemlock, MI. “Depending on who you talk to, the shortage could be very severe. Some [PV] companies may have to close their doors for lack of feedstock.”
Hemlock makes and sells semiconductor- and solar-grade polycrystalline silicon for the two major applications, Takahashi says. He emphasizes that chipmakers are in a healthier position than their PV cell counterparts because of their long-standing relationships with their suppliers.
“IC manufacturers would be pretty much—I wouldn’t say ‘protected’—in far better shape than people in the PV industry, because the semiconductor industry has been well established. Their substrate suppliers have long-term contracts signed with the polysilicon suppliers. On the other hand, the PV industry is still in its infancy stage, and there are a lot of newcomers coming in the door. They [the newcomers] don’t even have a conception of what a long-term contract is all about.”
In a presentation on the outlook for substrates and lithography materials, Lara Chamness addressed these issues at SEMI’s annual Strategic Materials Conference (SMC), held in Half Moon Bay, CA, in mid-January. The senior market analyst agrees with Takahashi that the shortage most likely will have a major impact on the PV cell industry. “I haven’t heard any of my silicon manufacturers complain about lack of poly, and they’re supplying record volumes,” she said in a post-SMC interview. “It hasn’t been an issue at the Silicon Manufacturers Group meetings.”
Chamness points out that the last two calendar quarters have seen record area shipments for silicon. In the fourth quarter of 2005 shipments totaled 1826 million square inches—an all-time high for a quarter. Silicon shipments had hit a new all-time of 1748 million square inches in just the previous quarter. “That’s way above anything done in 2000 for sure, and it was higher than 2004, which itself was a record year for materials,” she adds.
At the moment, prices are the big concern. In 2004, polysilicon sold for approximately $30 per kilogram, Chamness says. By early February 2006 she was hearing reports of “over $70 per kilogram or higher.” Spot prices in 2004 hovered around $32 per kilogram. By the last week of December 2005 or the first week of January 2006, it was selling “in the $80 range, with reports of more than $100.”
The conversion to 300-mm wafers could put more pressure on polysilicon supplies because of a “surprisingly strong demand” for the larger substrates, Takahashi says. The wafers require more material because the substrate itself is thicker and “you end up with higher kerf loss while you’re doing the processing, and the yield is not up to the 200-mm standard yet. Those things combined mean you need more silicon going into every square inch of chips.”
At a Sematech- and SEMI-sponsored meeting after Semicon West in 2005, an industry group called the Critical Materials Council (CMC) discussed the expected shortfalls of semiconductor-grade polysilicon through 2008. A press release from Sematech noted that the IC industry consumes approximately two-thirds of manufactured silicon, with the remainder taken by the PV industry, which was growing at a 30–40% annual clip.
At the time, an executive from Advanced Silicon Materials (ASiMi) said that not enough polysilicon was available “to support the growth of the photovoltaic industry after 2008.” Dave Keck, vice president of ASiMi, added that the shortage certainly would not meet the growing needs of chipmakers during the conversion to 300-mm wafers. He and others present predicted a shortage of 4000 metric tons in 2005 that would increase to 6000, 12,000, and 20,000 metric tons in 2006, 2007, and 2008, respectively.
Dan Tracy, SEMI’s senior director of industry research and statistics, said that chipmakers alone were straining supplies, since some 40 300-mm fabs were expected to be in operation by 2006.
But since that post-Semicon meeting, both Takahashi and Chamness say the picture has become a little sunnier as Hemlock and other suppliers have announced major expansion plans. “Since July 2005, many polysilicon producers have made public announcements about their capacity expansion programs, like Wacker in Germany,” notes Takahashi. “They announced a 5000 metric tons capacity expansion. MEMC recently announced a 4000 metric tons capacity expansion, and Hemlock’s capacity expansion was made public for 6800 metric tons coming online in 2008. There may still be some shortage, but it is not as big a magnitude as mentioned in July.” He also says that Hemlock is expanding its capacity incrementally, adding 30% over its 2005 capacity.
A consultant who attended the SMC believes the semiconductor industry should have access to the wafers it needs—provided growth proceeds at a modest clip. “Half of the big semiconductor wafermakers will be able to get poly for the next couple of years, as long as the industry grows at the consensus forecast of 6–7% per year,” says Richard Winegarner, president of Sage Concepts, a Healdsburg, CA–based research firm that tracks the silicon market.
Concerning growth rates, he questions “all the big discussions about whether the semiconductor industry has matured” and will never see the violent market swings that have defined it. “I don’t know that that’s true. Perhaps it just hasn’t seen large new applications.” The question remains whether the industry is maturing in the way of traditional industries such as steel or tires, “where the shape of the curves were flat.”
Wafer supply could be very tight, Chamness notes. While the issue was “on their radar” at a recent meeting between poly suppliers and device manufacturers, “they remained very quiet about potential concerns.
“Something that came through again at ISS and SMC this year is there are so many consumer applications out there,” Chamness notes. “A lot of people are saying there isn’t going to be one killer app any more, but you’ve got this killer market called consumer electronics.”
Any critical shortage of polysilicon would have a twofold impact on the device manufacturers and their suppliers, she points out. “It’s with the wafers themselves and also with the components that go into some of the fab tools,” such as electrodes and focus rings. “Those two components are made of polysilicon, and it’s a lower grade. But if there’s a question of whether there’s enough poly, and if you cannot make those components, it’s an Achilles heel.”
In the mid- to late-1990s those flat growth curves mentioned earlier by Winegarner prompted several steel companies to enter the silicon market. “Somebody in Japan wrote a white paper [saying] silicon would be a good thing for Japanese steel companies to get into because their current business had stagnated,” he recalls.
“So, four or five of them in one form or another got into the semiconductor industry silicon chain. All of them lost money and have withdrawn; not only Japanese companies, but actually Korean and Taiwanese too.”
Winegarner says the most significant effect on the polysilicon market of the mid- to late 1990s was an expansion that proceeded with the expectation that the global semiconductor industry would reach the $300-billion mark by 2000. “They put in the capacity to supply a $300-billion market, and it went down to $135 billion, so there was gross overcapacity.” In 1998 a decrease in silicon wafer demand caused by lower IC sales, smaller die sizes, a decrease in the use of test wafers, and an increase in reclaimed wafers combined to reduce wafer industry revenues by 30% from $6.6 billion to $4.6 billion, according to Winegarner.
“This is just the semiconductor industry alone,” Winegarner continues. “Until 2000, the solar guys hardly used any of what you would call virgin polysilicon; they were all using scrap. All the poly plants have been operating as underperforming assets. They were all losing money, so they’ve all pared down their staff, and their customers were holding what are called ‘reverse auctions,’ in which you’d get all your suppliers on the line and you’d say, ‘this is what I want to buy. Which one of you is going to sell it to me for the lowest price?’
“So,” he adds, “not only were [the polysilicon producers] losing money, there was a lot of animosity. Gone were the long personal friendships between salesmen and purchasing agents. Those went down the drain.”
One major trend arising from this scenario is that upcoming polysilicon capacity expansions “are being funded by the solar guys, because polysilicon learned from the mid-1990s mistakes,” Chamness says.
What mistakes exactly? “If you build it, they will come,” she replies. “They’re locking them into long-term contracts. The solar guys, they’ve made the commitment, they’ve gone down the aisle, they’re at the altar and exchanging vows.”
What Winegarner calls the “transition year” for solar cells occurred in 2004, “when demand was actually bigger than supply, but there was some inventory lying around that helped meet that need. Last year was actually the point where solar energy, instead of being able to grow at 40%, probably got more near to 7 or 8% because it couldn’t get material.”
There is a sunny future on the horizon for the PV industry, which shares several of the process steps of its semiconductor counterpart. “The solar industry will actually use as much silicon [in metric tons] as the semiconductor industry within the next 12 months,” asserts Tom Werner, CEO of SunPower, a Sunnyvale, CA–based manufacturer of PV cells.
Cypress Semiconductor purchased a majority stake in SunPower in 2002, helping it to develop its high-efficiency A-300 solar cell. (The company went public in 2005, filing an IPO in November. Cypress owns 85% of the shares.) The key to the technology is that the metal contacts for collecting and conducting electricity are on the back of the cell, where they are away from the sunlight for improved cell performance, according to the company.
“Yes, there’s a shortage, so the solar market is actually constrained in its growth because of the shortage of polysilicon,” Werner says. Despite this state of affairs, prices are increasing “and that means [profit] margins are increasing, which means that there will be new entrants in the polysilicon markets because markets are God, which means the shortage will last between one and three years. Capitalism will work.”
The polysilicon industry “has gone through cycles where they’ve added capacity and the promise of solar didn’t come to fruition, and the semiconductor industry went through a downturn and there was nobody around to use [the polysilicon].” Werner adds that “significant capacity is being engineered as we speak.”
Dick Swanson, the former Stanford electrical engineering professor who founded SunPower and is its president and CTO, addressed the synergies with the semiconductor industry in mid-January at SEMI’s Industry Strategy Symposium (ISS), held the same week in Half Moon Bay as SMC.
Swanson says his presentation “hit a note. Everyone was excited about it. I think it’s just the emergence of a market that’s growing this fast—40% per year historically—in recent times that is also similar to the industry that they’re familiar with. Everyone’s brain is going 100 miles an hour about the opportunities they may explore in that space.”
Swanson says that solar cell manufacturing is similar to making integrated circuits in that it uses diffusion, film deposition, and metal deposition. On the bright side financially, one process that is not used is high-resolution patterning, Swanson emphasizes.
“When we first started working with Cypress, they would ask us what the design rules were, and when we’d say, ‘100 µm,’ they’d kind of smirk, and after that they got working on it.
“The big difference is the volume,” Swanson continues. “They take the basic semiconductor steps they’re all used to. [Then it’s] how do you do one wafer per second? We started 50,000 wafers per day.” That’s equal to the “monthly start rate” of some semiconductor fabs. Cost is “brought down by amortizing the manufacturing overhead. It’s more akin to the DVD industry. The actual process steps—the basic process know-how that these companies have—is essentially applicable. They just have to think ‘one wafer a second.’ ”
Werner says the purity level of silicon for solar cells varies, “but not significantly. It’s not a big economic driver.” Swanson adds that the growth method for turning silicon into a single-crystal ingot is more important than the purity level, because when silicon crystallizes, it rejects impurities by nature of that process.
“What solar does do is use a lot more scrap,” Swanson says. “There’s material that’s scrap from the IC industry; even the so-called pot scrap left over in the aftergrowth is sometimes recycled. That material also tends to be pretty pure.”
Takahashi says the purity level for polysilicon in the semiconductor industry is 11 9s to 12 9s, measured by parts per billion or parts per trillion. “The solar industry can live with purity levels two orders of magnitude higher than that. Nine 9s would be pretty much acceptable for solar applications.”
Once the extra polysilicon capacity becomes available around 2008, the sky would seem to be the limit for the PV industry. Werner, Swanson, Takahashi, and Winegarner all take note of some of the larger economic and technology trends, singling out solar-friendly government incentive programs in Germany, Portugal, Spain, South Korea, and China as positive signs in particular.
For its part, the semiconductor industry may find itself in a bind if it starts to grow, Winegarner says, because the polysilicon suppliers will be “contractually obligated to sell to the solar guys. So even if it looks like it’s there, the polysilicon is not there for them. Of course, semiconductor users of polysilicon are going to have a tough transition to go from a position of power—they were holding reverse auctions—to a position of weakness where they’re going to have to bid. In the interim what’s happened is that most of the semiconductor guys had long-term contracts—something on the order of three years—and as those are expiring they’re getting their supplies, but the prices are doubling.”
If there’s a winner in this scenario, it may be the polysilicon houses. “Poly was looked down upon three or four years ago,” the consultant muses. “Now it’s a hot commodity.” — JC
Search | Current Issue | MicroArchives
Buyers Guide | Media Kit
Questions/comments about MICRO Magazine? E-mail us at firstname.lastname@example.org.
© 2007 Tom Cheyney
All rights reserved.