Contamination control challenges extend beyond fabs to toolmakers
By Hank Hogan
It was after production units started shipping that the contamination problems cropped up, says Kimberly Subrahmanyan, engineering director for the semiconductor link processing group at Electro Scientific Industries, Inc. (ESI; Portland, OR). ESI’s first ultraviolet, laser-based link-processing offering was intended to allow its customers to manufacture advanced memory products, those with links spaced more closely together than possible with the company’s existing infrared laser-based systems.
The engineering prototypes had worked well. That was also the case with the production units-at least initially. Then the contamination-related issues started to appear.
“We were finding organics on the lenses. We were losing transmission through the optical path of our system,” recalls Subrahmanyan. “We were seeing some of the contaminants, like particles and such, burned onto the lenses.”
In switching from a light source with a wavelength longer than red to one with a wavelength shorter than blue, ESI’s engineering team thought they’d accounted for the contamination control challenge. The problem, though, had proven to be more difficult than they’d anticipated.
So they revamped their entire manufacturing process, cleaning it up in multiple ways. As a result, they succeeded in fixing the problem and now ship UV systems routinely. That effort carried with it some lessons, tips that could be of use as ultraviolet lasers and light sources work their way into more and more cleanrooms and a wider number of manufacturing steps or processes.
Contamination brought to light
In ESI’s case, the manufacturing changeover began because the company wanted to satisfy its customers’ needs. ESI’s systems “blow” links, which are found in semiconductor memory chips. These links electrically switch in spare good-memory locations for those with defects, allowing memory manufacturers to repair less-than-perfect product and improve yield. The technique is widely used in the industry and, in ESI’s products, involves firing a laser at the link.
The company had been doing this with infrared lasers operating at wavelengths of 1,300 and 1,064 nanometers, well above the 700 nm of the visible color red. Due to the fundamentals of optics, the spot size that a laser can be focused down to is a function of the wavelength. The shorter the wavelength, the smaller the spot created. As features on chips became smaller, memory manufacturers wanted to have the links take up less space and that meant ESI had to make some changes. “In support of shrinking structures, we needed to go down to smaller spot sizes in order to blow the links,” says Subrahmanyan.
The company chose to go to a laser operating at 355 nm, in the ultraviolet below the 400 nm visible to the eye. In doing so, the spot size decreased by the same ratio as the wavelength.
However, the energy of those photons rose similarly. Thus, while a 1,064 nm photon had energy of 1.2 electron volts (eV), one at 355 nm packed 3.5 eV. Because of the increased energy of UV photons, chemical reactions happen upon exposure to UV that don’t take place when visible or infrared light is involved. The engineering staff at ESI knew this and they therefore changed handling procedures during their manufacturing process to reduce contamination.
Although this worked with the prototype, it wasn’t successful when production ramped up. Bryan Bolt, manager of strategic technology development at ESI, had some previous experience with UV at 257 nm, where the energy climbs to 4.8 eV. He knew that optics failing in three weeks would, with 90 percent contamination reduction, not see their service life lengthen similarly and hit 11 months out of a 12 month target. “It may only increase service life to eight weeks or so,” he says. “That last 10 percent makes a big difference.”
As for what is considered a problem, Bolt points to anything in a vapor phase, such as heavy hydrocarbon chains and acid gases. These condense on the lens and photopolymerize into a film when the UV hits, thereby degrading the optics.
To eliminate the problem, ESI first ensured that they weren’t introducing contaminants into the process by their choice of materials. They then instituted a number of handling and cleaning procedures, such as using an ultrasonic clean followed by a vacuum bake of parts prior to assembly to remove any contaminants. Finally, they upgraded their research and development cleanroom as part of an overall remodeling process (see Figs. 1 and 2). They used this to prove out the revamped process. For production, they built a new cleanroom.
Cleanroom entrance at ESI headquarters’ R&D center. Photo courtesy of Electro Scientific Industries, Inc.
According to Brian Green, semiconductor link processing group product manager at ESI, the new cleanroom will go live this month. It will include a 4,200-square-foot area that’s ISO Class 5, equivalent to the old Class 100 standard, and another 6,400 square feet that’s ISO Class 6 (Class 1,000). Those area figures include production, gowning, pass-through, and materials-handling space. The more sensitive components, such as those that experience the full intensity of the laser, will be assembled in the cleaner section. The Class 6 area uses HEPA filters, while the Class 5 area uses ULPA filters.
Green noted that in designing and building the new cleanroom, the team had to consider more than just the particulate load. “We had to take into account the molecular risk,” he says.
Testing room at ESI headquarters’ R&D center. Photo courtesyof Electro Scientific Industries, Inc.
Heavier hydrocarbons, those with more than six carbons in them, were among the chemical constituents that concerned them. Green reports that so far everything appears to be in good shape, with the design and construction helping the cleanroom to hit its contamination control goal.
The combination of changes in materials, cleaning procedures, and manufacturing process paid off. ESI now consistently delivers its UV product to customers, with good field results.
Twenty or forty to one
The ESI story is one that may be repeated, in different forms, over and over as semiconductor technology advances. The ITRS roadmap, which the industry jointly produces to highlight future manufacturing needs, calls for widespread use of UV lasers for defect inspection, feature size and other measurement, via drilling and chip dicing. The wavelengths mentioned range from 355 to 266 to 193 nm (see Table). At the latter two wavelengths, light hydrocarbons and a host of other airborne contaminants become a problem.
An example of what’s to come can be found in Applied Materials (Santa Clara, CA) UVision system, which uses a 266 UV laser for illumination. The company claims the device can spot critical defects as small as 30 nm, which will be needed for manufacturing of 65 nm and smaller processes. It’s able to do so, in part, because it uses a UV light source.
The contamination control challenge presented by such systems isn’t just in the construction of the laser-driven products. It’s also in their use, which can be in less-than-pristine areas. The link blowers that ESI makes, for instance, are typically found in test and assembly areas and not ultraclean semiconductor front ends.
So far, most of the deep UV lasers, those below 266 nm, have been in lithography steppers, but that won’t be the case in the future. Joe LaChapelle, CEO and chairman of Deep Photonics (Corvallis, OR), notes that there were 20 to 40 inspection stations for every lithography machine. Deep Photonics, a start-up, hopes to use some advanced technology it’s developed to provide relatively inexpensive solid state deep UV lasers that operate at 266 nm and below. The company claims its products will have longer lifetimes and will offer higher power than what’s currently available.
If Deep Photonics succeeds in bringing these products to market, the use of UV may become more prevalent. In particular, LaChapelle says a 266 nm source would make various materials-processing tasks easier, not only in semiconductor but also in electronics manufacturing. Printed circuit boards via holes, for example, can be drilled more easily with a 266 nm source than one at 355 nm.
However, he acknowledged that the new technology won’t get around the basic contamination control problem. “There’s no question that cleanliness, both during the assembly and manufacturing of our laser heads and then also in their implementation to the OEM capital equipment, is absolutely critical for maintaining lifetime and stable power,” he said.
Deep Photonics assembles its laser heads in a cleanroom. Then they’re nitrogen-purged to remove trace contaminants, and hermetically sealed. Because of the wavelength of the light, that discipline might have to be extended to the work area, with nitrogen-purging and other measures to reduce contamination. That’s particularly true if the intended use is not in a typical semiconductor front-end cleanroom.