Can Lithography Go to the Extreme?
As semiconductor manufacturers press the limits of current optical lithography, many are scrambling to find a next-generation lithography (NGL) system. Devising an NGL is consuming many in the industry because the health and viability of tomorrow's electronics industry will hinge on it. Simply put, to continue on with Moore's Law--the doubling of power of electronic chips every 18 months, which has become the industry standard for advancement and success--one or more NGLs must be developed.
A leading NGL candidate is extreme ultraviolet lithography, or EUVL. But during its 10-plus years of R&D, EUVL has been both the darling and the black sheep of the industry.
EUVL is being developed in programs in a number of countries like Japan, France, and the US. It has the backing of industry trade groups, such as the Semiconductor Industry Association and International Sematech. It is identified in the International Technology Roadmap for Semiconductors as a potential solution to future lithography challenges. And in the US, it also enjoys the backing of major chip-making companies, including Intel, IBM, Motorola, Advanced Micro Devices, and several others.
A consortium of these companies called the EUV LLC, has an active research program to advance EUVL. The EUV LLC, with a budget of $250 million, is working with three national laboratories to essentially take defense technologies, like those to make highly precise mirrors and optics, and apply them to the circuit manufacturing.
This project is so big in scope and challenge that it has even been called the semiconductor industry's "moon shot" of technology. Recently, the EUV LLC demonstrated the capabilities of a completed "alpha tool" system. To many, this was vindication and proof that EUVL can work.
"EUVL will be the production technology of the future," promises Intel Corp.'s, Santa Clara, Calif., Chuck Gwyn, who is the program director of the EUV LLC. "We think this because of the reasonable cost of ownership, and the fact that we have basically demonstrated all aspects of the technology."
"There is no question in my mind that this is the correct solution to producing devices that have geometries of 0.05 microns [50 nm] and lower," adds Art Zafiropoulo, CEO of Ultratech Stepper Inc., San Jose, Calif. "This is the only technology that will be successful."
But to many others, the EUV "alpha tool" is a big, complex system. Proving it can lay down one layer of a circuit pattern in a laboratory test is far removed from the intricacies and tradeoffs that rule real-world manufacturing environments. The technology will demand almost total perfection in mask making, mirror shaping, and wafer handling.
"There are many people who are skeptical about EUV's viability in manufacturing," says Henry Smith, a professor of electrical engineering at Massachusetts Institute of Technology, Cambridge.
"Demonstrating one layer is pretty good proof," admits Dale Ibottson, director/VLSI process development at Agere Systems, Allentown, Pa., which was Lucent Microelectronics Group prior to a recent spin-off. "Demonstrating you can come up with a whole infrastructure to support both the commercialization and the commercial production of these tools, in my mind, is the big question. It is as much [a question of] the infrastructure as it is some of the technical hurdles that they have to overcome."
EUV was once thought of as too exotic to ever become practical. But the semiconductor industry continued to explore the technique as a possible replacement to optical lithography because they believed the technology could support multiple process generations and be cost effective for producing all types of semiconductor chips.
Optical lithography currently resides in the in 130-nm "node," industry jargon for the technology's capability plateau. In 2003, circuit lines of advanced devices will be down to the 100-nm node, and by 2005 it is expected to be in the 70-nm range.
"When we get into those ranges," says Juri Matisoo, VP/technology at the Semiconductor Industry Association, San Jose, Calif., "we are starting to really push the current optical lithography capability."
The reason for this is because with the current tricks of the lithography trade, as Matisoo calls them, semiconductor manufacturers can print line widths of about half of the smallest wavelength they employ. Practical optical lithography pretty much ends at 157 nm, making the 70-nm threshold a critical one. To go lower will require a completely different lithography technology.
EUVL is but one of several candidate NGLs that have been explored. Electron beam projection lithography (EPL) is also being developed both in the US by companies like Agere and Applied Materials and overseas, most notably by Nikon in Japan and ASM Lithography Holding NV in the Netherlands. It is still seen as a possible NGL candidate, but for many only after EUVL.
Other NGLs, such as proximity x-ray lithography, electron beam direct write, and ion beam projection lithography, "have fallen by the wayside," says Matisoo.
"This is a wavelength issue," he explains. "The EUV wavelength is in the 13-nm range. The most advanced optical is 157 nm. It's an order of magnitude shorter wavelength, so you should be able to print 13-nm line widths with it, and it lets you get out of this box of subline width resolution."
Smaller and smaller line widths would mean faster and faster circuits, denser memories, and more capable and cost-effective computing. For example, chips made by the EUVL process could have 10-GHz processors, compared to the fastest Pentium 4 processor speed of 1.5 GHz, according to the EUV LLC. It could lead to microprocessors that are 30 times faster than those today, and a 1,000-fold increase in computer memory. This capability could lead to telephone systems that translate your voice into the appropriate foreign language for cross-continent telephone calls, reduce the relative cost of computing by an order of magnitude, and provide super computer power for thousands of dollars instead of the millions it costs today.
Conventional optical lithography focuses a beam of light onto a transparent mask that contains images describing the circuitry for a chip. The circuit images then pass through a series of reduction lenses that are reproduced on a wafer coated with photosensitive resist material.
EUVL is similar to optical lithography but uses very short wavelength illumination (10 to 14 nm). Since EUV radiation is absorbed by all materials, the lithography process must be performed in a vacuum and use reflective optics.
EUV radiation is produced from a 45-eV plasma by heating a supersonic xenon gas jet with a high-powered laser (the alpha tool uses an Nd:YAG laser). The radiation is collected in a condenser and shaped into a narrow arc-like beam and focused onto a reflecting reticle or mask.
The reflected radiation passes through a four-time reduction camera and is imaged onto the resist-covered wafer. The entire patterned field is illuminated by scanning the reticle through the beam. Correspondingly, the wafer is scanned at one-fourth the reticle speed in the opposite direction to reproduce the mask image on the wafer. Conventional silicon processing can be used to define the mask patterns on the wafer.
The alpha tool demonstration earlier this year showed "the integration of the technology into one system," says Gwyn, who calls it a milestone for the project. The system printed full-field images in the 100-nm range. It also operated in the 80-nm range, but the quality of those images weren't up to snuff.
"We were able to demonstrate good print patterns over the full field of 24 X 32.5 mm, which is a good-sized integrated circuit," he adds.
"They still don't have the capability for multi-layer printing," says SIA's Matisoo. "There is no registration capability of the machine. But it's a big, big step forward and they pulled everything together to demonstrate printing capability at that wavelength."
But critics of the EUVL system say it is very complex; it requires system precision that has never even been attempted before, and key questions remain as to whether it can be cost effective, if infrastructure to support it can be built up, and if technical challenges, both seen and not yet foreseen, can be overcome.
"The problems with EUVL and EPL are only slowly being revealed," says MIT's Smith, who favors x-ray techniques for NGL. "And it's not at all clear that they can be solved in time to meet the roadmap timetable, if ever.
"That whole system depends on fabricating optics that are smooth and can follow a certain figure to within a fraction of a nanometer," explains Smith, who pioneered development of techniques for fabricating nanostructures. "It's possible, but it's a horrendous problem. Then they have to make masks that are just as good as the lenses. The masks are going to be incredibly expensive, and every layer of the chip has to have a different mask."
"They have made some tremendous advances in EUVL, but I think as we get closer to defining the commercial tool and its limitations, there will be further technical hurdles," adds Agere's Ibottson. "I'm not sure they can be overcome."
"EUV doesn't come free," admits Matisoo, a proponent of the system and of EUV LLC.
Chief among the technical challenges is the power source. The source in the alpha tool, which by Gwyn's own estimation, needs to be 10 times more powerful to achieve the targeted throughput rate of 80 300-mm wafers/hr--a rate that will define the overall system's cost effectiveness. Gwyn says that while a candidate for a commercial system has not yet been identified, groups around the world and at the EUV LLC are working on promising replacements. He points to discharge sources, especially a capillary discharge source, as potential solutions.
"I'm confident it will be solved," he says.
The masks, by directing such fine line widths themselves, have to be defect-free. The mirrors of the system likewise have to be perfect in order to operate as designed. Zafiropoulo explains that the mirrors, called aspheres, require considerable care in manufacturing. One set of aspheres currently takes one year to build. The mask blank--essentially a flat piece of very high purity glass--will have up to 80 coatings on it. Each coating will have to be controlled to the atomic scale to unparalleled precision.
"If one layer has a defect in it, then the whole thing is void. It will not work," says Zafiropoulo. He adds that the software system is "exotic--you are talking about stages that have a precision of 5 nm."
The EUV LLC is using defense technologies to aid in obtaining such performance. They are employing EUV source technology from Sandia National Laboratories, Albuquerque, N.M., and optical design and multilayer coating technology from Lawrence Livermore (Calif.) National Laboratory. But to turn all of this around and transform specialty technologies into commercial systems will be challenging. To do it on schedule will be daunting.
Currently, the EUV LLC's schedule includes transferring EUVL technology to stepper companies and getting them to scale up and produce beta tools (about 10) by late 2003 or early 2004. Quantities (in the range of 10/quarter) of production tools are scheduled for late 2005 or early 2006.
"In addition, masks (blanks and patterned), EUV source components (laser and jet or discharge), resists, and mask inspection equipment must be commercially available in that time frame," says Gwyn.
While Zafiropoulo is a proponent of the EUV LLC, he says the schedule might be too ambitious. Because of infrastructure concerns, "there won't be 10 machines running in the manufacturing world at least until 2007."
And to Ibottson, who assesses new technologies and tries to incorporate the best of them for his company, "timing is an extremely important factor. I have some skepticism that [EUVL] is going to arise and be cost effective relative to alternatives," he says.
Ibottson prefers EPL to EUVL based on cost and time-to-market issues. EPL, he says, can, with technical tradeoffs, achieve comparable resolution to what is believed to be the ultimate resolution of EUV.
"It is the net cost per wafer pass of the system when it is commercialized" that makes EPL look more attractive, he says. "I'll buy whatever is cost effective."
The largest question to be answered concerning EUV might well be how the semiconductor industry responds. Most everyone in the industry says infrastructure could determine the success or failure of the technology.
Can companies build up the capabilities to supply these perfect pieces, like masks and mirrors, and do so on a production basis? Can the system meet its targeted through-put rate of 80 300-mm wafers/hr? Who will be able to afford such an expensive piece of equipment thought to be in the $30 to 40 million range?
While these questions loom large, many feel the semiconductor industry will again rise to the occasion and deliver what was thought improbable.
"Years ago, we didn't think we could put somebody on the moon," says Zafiropoulo. "It will take those kind of resources and effort and brain power to do it. I would not rule out the ingenuity of American scientists or American companies. I think it's huge, but I love it."
Semiconductor Lithography Roadmap Desired Feature technology size Lithography options availability 180 nm Optical 1999 130 nm Optical 2002 100 nm Optical, x-ray 2005 70 nm Optical, x-ray, EUL 2008 50 nm EUV, charged beam 2011 35 nm EUV, charged beam 2014 Source: International Technology Roadmap for Semiconductors
Tiny Structures Impact Industry
In June, researchers at Intel Corp., Santa Clara, Calif., announced that they had built transistors for microprocessors and other logic chips with structures only 20 nm wide, roughly four times wider than a single atom.
Such tiny structures could have a huge impact on semiconductor manufacturing, resulting in microprocessors that contain up to a billion transistors, running at speeds approaching 20 GHz and operating at less than 1 V. Today's best Pentium 4 chip runs at 1.7 GHz and has 42 million transistors.
The result will be chips that are 30% smaller and run 25% faster than the industry's current fastest transistors. More importantly, Intel's researchers say the new advance, which they announced at the 2001 Silicon Nanoelectronics Workshop in Kyoto, Japan, means Moore's Law is safe for the next decade.
"We still have not found a fundamental limit for making silicon transistors smaller," says Robert Chau, Intel Fellow and direct of transistor research, Intel Logic Technology Development. "The pace of silicon development is accelerating, not decelerating."
"Our transistor research shows that we are able to extend Moore's Law scaling for at least another three generations beyond our current technologies," adds Gerald Marcyk, director of the Components Research Lab in Intel's technology and manufacturing group.
Intel researchers expect the new technology, which when it's in production will require the use of extreme ultraviolet lithography, to be implemented by 2007.
Reflectometer Measures EUV Mask Blanks
Researchers at EUV Technology, Martinez, Calif., in collaboration with researchers at Lawrence Livermore (Calif.) National Laboratory (LLNL), have developed a device that measures the reflectivity and uniformity of multilayer coatings for extreme ultraviolet (EUV) lithography mask blanks, without removing them from a clean environment, in the deposition chamber or shop floor. The blanks are made by coating 200-mm silicon wafers with molybdenum silicon (MbSi) multilayer Bragg reflectors by ion-beam sputtering in a vacuum coating chamber.
Their EUV Reflectometer model LPR 1016 is installed on the Low Defect Deposition system for depositing ultraclean multilayers at LLNL. Lithography masks created in this chamber are transferred from the deposition chamber to the measurement chamber with robotics, eliminating the risk of particulate contamination.
The reflectometer operates by focusing a Nd:YAG Q-switched pulsed laser onto a gold target. The source spectrum produced is essentially continuous in the EUV region between 10 and 15 nm. A narrow band around the region of interest is selected by the monochromator. The output wavelength is tuned by rotating the grating. After leaving the exit slit, the light is reflected from the sample mounted at a fixed angle to the detector. A small portion of the measurement beam is split off to an auxiliary detector after the slit to monitor the beam intensity and enable shot-to-shot variations to be normalized.
The time per measurement (about 2 min) for this reflectometer is comparable to that of the synchrotron radiation reflectometer (SRR) used at the Lawrence Berkeley (Calif.) National Laboratory's Advanced Light Source and the National Institutes of Standards and Technology's SURF facility. However, due to its in situ measurements, the setup time is considerably faster (10 sec vs 1 hr) and the turnaround time is even faster (5 min vs 8 hr).
While the main application of the reflectometer is to measure EUV lithography photo mask blanks, it also can be used to measure the reflectivity of multilayer coatings on EUV optics. In these applications, it offers the same advantages over synchrontron radiation measurement techniques.
The EUV consortium EUV LLC purchased the first tool for SEMATECH and installed it at the Dept. of Energy's Lawrence Livermore (Calif.) National Laboratory in September 2000.
LPR 1016 Specs:
Measurement area: 200 mm dia Spot size: Less than 2 mm dia Wavelength range: 10 to 16 nm Wavelength accuracy: 0.0025 nm (0.02%) Reflectivity accuracy: Less than 1%
More Info: EUV Technology Martinez, CA 94553 925-212-7394 www.euvl.com
Lawrence Livermore National Laboratory Livermore, CA 94551 925-423-5398 www.llnl.gov
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||commercial usefulness of extreme ultraviolet lithography|
|Comment:||Can Lithography Go to the Extreme?(commercial usefulness of extreme ultraviolet lithography)|
|Publication:||R & D|
|Article Type:||Brief Article|
|Date:||Jul 1, 2001|
|Previous Article:||Don't ignore dot-com mania.|
|Next Article:||CVD Integrates Inorganics.|