The roots of the company's hyperpure-silicon business in Burghausen go back to 1953, when research on raw materials for electronics began there. Initial production of polysilicon for semiconductor applications in float zone (FZ) silicon started in 1958, with 1961 marking the first manufacturing of crucible-pulled (or CZ, short for Czochralski) crystals, as well as the first epitaxial wafers. The company founded a wholly owned subsidiary in 1968, Wacker-Chemitronic, and commissioned a new plant for polysilicon deposition the following year. Other milestones include the launch of 200-mm-wafer pilot production (1984), the first 300-mm wafer experiments (1990) and creation of the large-wafer business unit (1995), and the reorganization of all sites into international business units under the name Wacker Siltronic (1994).

Wacker's Burghausen site has 130 crystal pullers, half of which are dedicated to 200- and 300-mm production.

The Burghausen plant produces all the necessary raw materials for wafer making, starting from production of ultrapure polysilicon via trichlorosilane. The polysilicon chunks are then melted in the quartz crucibles of the rows of crystal pullers in the factory. Then crystal growth and the pulling of the ingots takes place, followed by the sawing of the silicon rods, edge rounding of the newly sliced wafers, grinding and lapping, polishing, and cleaning. The process concludes with final inspection and packaging of the prime wafers. The factory floor varies from high-end cleanrooms in the front and back ends of the process to machine-shop-like conditions for some of the dirtier work, from state-of-the-art metrology to technicians using rulers to measure slug diameters.

Siltronic's Burghausen plant produces the widest range of silicon among the company's six worldwide manufacturing facilities, growing crystal and making wafers for everything from 4-inch (100-mm) applications to the latest 12-inch (300-mm) substrates. Although the pedigree and scope of the facility make it one of the world's most compelling wafer-making sites, its role as a leading 300-mm facility was really what inspired my desire to visit the plant after Semicon Europa concluded in mid-April.

My hosts were Hermann Fusstetter, vp and head of the large-wafer business unit, and Volker Braetsch, the company's vp of marketing and business development. Two tour guides showed me around nearly every part of the manufacturing facility: Peter Gerlach, senior manager of CZ crystal growing for advanced wafers; and Hans-Joachim Stadter, senior production manager for the large-wafer business unit. All are Wacker veterans, though with 22 years at the company, Fusstetter's the senior man.

Most people who know something about the silicon business understand the difficulties the companies face, both in terms of profitability and customer perceptions. Those on the wafer side sometimes feel like the Rodney Dangerfield of the semiconductor world. "It's not a commodity like gases," said Fusstetter. "[The industry] likes to consider it a commodity, but a wafer is a highly engineered, high-volume product, with many pages of specifications. We need to create respect for it again.... Keep in mind that the valley around the Bay Area is not called Equipment Valley, it's called Silicon Valley."

Braetsch piped in with similar sentiments. "It's a little bit ironic how low the value and respect is for silicon in this whole industry. It's anywhere between a $5, $7, $8 or $9 billion business, depending on the year, but if you shut that down, you shut down 95% or 98% of a $150–$200 billion IC industry, and you shut down close to a $1 trillion final-product industry."

"There is a pretty obvious correlation between the availability of silicon and the value of it," he continued. "I would say that quite a lot of customers tend to call silicon a commodity in downturns. When there is excess capacity in silicon and you can easily get it, and you can squeeze the suppliers, it's a commodity. Whenever the picture changes, then it reverses course, and you hear comments like, 'No, silicon isn't a commodity, silicon is strategic.' There are even some customers that have procurement or materials committees, and they use the term 'strategic-material suppliers.'"

Although the pendulum is hopefully swinging back in an upturn direction, there can be little doubt of the current strategic value of 300-mm wafers. Recent studies by Gartner Dataquest and SEMI show growth in both revenue and unit area for the large substrates. Companies such as Wacker that have increased 300-mm sales have been able to "control revenue losses...below the market average," according to the Dataquest study.

Fusstetter was tight-lipped about the company's plans, both in terms of timing and location, for adding large-wafer capacity with the so-called Fab 300-2 project. "It depends very much on the speed of the recovery, on the speed of getting the first high-volume fab, Fab 300-1, to capacity.

"Remember that the plant you're visiting was originally intended as a small pilot line (the first stage was commissioned in 1998 for 20,000 wafers per month), but the last acceleration before the downturn created the need to simply not just build a green-field fab, but to expand the existing pilot line way beyond a pilot line into our first high-volume fab. With final buildout, when we really have all the equipment ready, we will have a capacity of 75,000 [300-mm wafers out per month], and that in turn is equivalent to about 180,000 200-mm wafers." The company also has about 10,000 large wafers per month coming out of its Wacker Nippon Steel joint-venture site in Japan, he added.

  Once the raw materials are in place, the beginning of any wafer-making process is crystal growth and pulling, done in the tall machines known as pullers. The Burghausen site has 130 pullers, with half of them dedicated to producing 100- , 125- , and 150-mm ingots while the others work on the 200- and 300-mm side. My tour guide for the crystal-growing hall, Peter Gerlach, showed me row after row of pullers in several different rooms, from older, small-substrate-capable tools to the beta version of a newer 300-mm design. Along the way, he offered an on-the-fly tutorial in the art and science of producing single-crystal ingots.

All 300-mm wafers are double-side and edge-notch polished, followed by cleaning steps to remove the slurry.

"The form of the crystal is the same, whether you grow a 4 inch, 8 inch, or 12 inch," he said, using English measurement units for the metrically challenged American editor. "One of the challenges is that the bigger the diameter [of the ingot] grows, the shorter the crystal gets with the same [polysilicon] charge size. Therefore, you are forced to use bigger and bigger charge sizes, which means bigger and bigger quartz crucibles, which means bigger and bigger [pulling] machines.... With longer crystals, you have increased efficiencies. The main cost factor is the yield, so the challenge is to obtain a good yield point, whether you use a big or small charge size."

Charge sizes for 300-mm crystals range from 120 to 200 kg, to the 350 kg that can be handled by the new pulling tool, a far cry from the 300-g charges shown in a 1963 photo on a wall outside the halls. "To grow a crystal takes approximately 2 1/2 days for 300 mm," explained Gerlach. "Half of the process is crystal growing, the rest is heating up and cooling down the furnace.... You have to take some time to heat up the graphite parts, and the other part is...you determine the defects inside the crystal by the cooling-down time, and you can control the defect growth by controlling the temperature during cooling down of the ingot."

Defects can be troublesome yield limiters, both in the crystal stage and closer to the end of the line. "In crystal growing, you do specific tests to produce specific defects," the manager notes. "For example, stacking fault tests enlarge the grown-in defects to a stacking fault, then by etching you can see a characteristic form...which points in certain crystallographic directions, so it's easy to detect. Stacking faults correlate to either metal contamination or crystal-grown defects."

While stacking faults have been around for nearly as long as modern crystal growing has and are well understood, Gerlach described a more recent anomalous phenomenon, which he called "the D defect." "This correlates to the gate oxide integrity (GOI) yield, so a high number of D defects means a bad GOI yield. At a very early stage you can get an impression how the GOI yield in your process will be. These defects have a certain etch mechanism that is different from the oxygen in the stacking faults... and are similar to the COPs (crystal originated particles), which are detected during the polishing process.... You can control these defects, as well as the stacking faults and the COPs, by changing your crystal growth parameters."

Gerlach pointed out two reasons why he believes Wacker may have a competitive manufacturing advantage: a design feature of their crystal-pulling machines and the facility layout of their most recently built halls. "We have metal shafts on our pullers, compared with Japan, where they mostly have cable pullers on which the crystal hangs. A cable puller has certain advantages in case of earthquakes, because it allows the crystal to swing, but we don't have earthquakes here. We think we might have an advantage because we can use a bigger charge size with the shafts."

On the facility side, the newer hall built six years ago incorporated a radical design change. Rather than the high ceilings and large cleanroom areas found in the older halls, cleanroom space in the 200-/300-mm hall has been cut in half by putting in a lower ceiling. "This means all your air-exchange numbers and so forth are much lower, making it less expensive to maintain," says Gerlach. "The other point is that the upper part of the machine sticks into a separate [mezzanine] floor, so you can do your maintenance work on the motors and shaft in a separate room that doesn't have to be a cleanroom. So it's cleaner, less noisy, and you only have on this floor what you need to operate the machine."

My escort for the rest of the plant tour, Hans-Joachim Stadter, took me through everything from crystal sawing to final cleaning, inspection, and packaging. The silicon rod portions, up to 400 mm long and weighing well over 30 kg, are loaded into wire saws equipped with an abrasive cutting slurry. "The wire spools have up to 100 kilometers of wire in one tool," he noted, citing one of the more amazing equipment stats in the semiconductor world. "It takes about 10–12 hours per ingot for cutting [them into wafers]. We try and put smaller ingots together or one larger ingot to fill one tool."

Next come the edge rounding, grinding, and etching steps. The edges are rounded with a disk grinder, then the new wafers are machined to a smooth, flat surface with a combination of surface grinding and chemical etching. Stadter said the grinding wheels have to be changed every 1000 wafers or so, and then the wheels are sent away to be refurbished and regrooved. He showed me the wet benches where an etch step takes place before the wafers go on to the polishing tool set.

"All 300-mm wafers are double-side and edge-notch polished," he explained. Here, the factory returns to a cleanroom configuration, and many wafer cassettes in various states of fullness rest on carts and racks, and near machines. We observed a technician carefully loading three wafers at a time into a polishing tool, smoothing off excess fluids by hand and aligning the substrates with a laser marker. "If you don't do it in this manner, the wafers will swim in the machines, and you will have some breakage," Stadter pointed out. "The breakage is terrible because the whole load will be wasted. You would have to clean all the systems and everything."

After the polishing, there are cleaning steps to remove the slurry from the wafers, and eventually visual inspection takes place within a dark room. "There's an operator inside, and he takes the wafer and looks under a special lamp for scratches and other visual defects. Then he decides if it's okay or not okay. Then we have another measurement where we generate flatness data for the double-side polishing."

The final polishing step is performed on the front sides of the wafers with a CMP tool. This process provides a final defect- and haze-free surface. To prepare for future tightening of requirements, the Wacker team is developing processes on CMP equipment from three of the major suppliers.

The final building on the tour housed the epitaxy area, cleaning, metrology, and packaging. While talking about the epi process, Stadter said they were trying to eliminate the LTO [low-temperature oxidation] thermal process for 300-mm/pp+ epitaxial wafers, because it's "too costly.... We don't want too much process variation."

A view inside the metrology area revealed several ADE, KLA-Tencor, and other inspection and measurement tools. Stadter pointed out a tool being used to test the quality of carriers and shipping boxes for particles, resistivity, and other metrics. "The shipper must be clean, and sometimes we have troubles with that," he admitted. He also complained about the attitude of a few "monopolists" in the market. "Their tools are not so good, because they take too much time to install, adjust, and get ready for production."

"Our customers insist that we have to provide more and more statistical data, and more and more process capability data," explains Braetsch. "We also try to get our processes down to a three, four, five sigma capability so that we can reduce sampling frequency, because many of the [testing] processes are destructive. The same holds for our suppliers, that's part of the cooperation with our suppliers. We would also like to get rid of incoming quality control and rely more and more on supplier certification as well as on their process capabilities, rather than an 'I have to look at every single piece' kind of control."

When I asked Fusstetter whether he felt that some suppliers treated him differently than their IC supplier customers, he nodded affirmatively. "What we see is they [the chip companies] constitute a much bigger equipment market and have much higher buying power. Also, and especially in 300 mm, wafer makers are the test beds for their new equipment, [since] they cannot afford consortia such as Sematech and Selete. Plus, equipment makers cannot use large volumes of wafers to mature their equipment for the market, because the equipment is already needed to generate those wafers. It's a Catch-22!"

It seems that whether it's the perceived commodification of their fundamental product or the lack of consistent customer service and attention from their suppliers, the wafer guys have to work hard to get respect. Hopefully, the 300-mm production ramp will finally cause the semiconductor manufacturing community to hold them in higher esteem and help bring the silicon companies back to profitability, through what Fusstetter sees as a combination of "volume, standardization, and cost-reducing cooperation efforts up and down the food chain."