While it's by no means clear which of the competing technologies will actually deliver practical throughput, chipmakers in Europe and Japan are clearly hedging their bets on next-generation lithography by looking seriously at maskless e-beam systems for writing at least the critical layers of low-volume products.
Progress recently reported by little Mapper Lithography, of Delft, Netherlands, working with Philips, joins programs by Leica Microsystems with STMicroelectronics, Canon with Hitachi, and TEL/Ebara/Dai Nippon Screen with Toshiba in targeting throughput of better than 10 wafers/hour by 2005 or so, with yet another approach.
Mapper aims to increase throughput with more beams; in fact, it has made 13,000 beams in parallel, to write a 26mm x 32mm field at one go, says CEO Boudewijn Baud. Contrary to previous press reports, and indeed contrary to what it still says on the company's own Web site, forget the old business plan with the special converter plate to change light to e-beams for high-volume production. The company is now using a split beam, aiming at a maskless tool for low volume applications at the 45nm node. "The converter plate is out," Baud told WaferNews, though apparently a micro lens array is still in, to split the beam into multiple parts. Currently the process is at the demonstration phase, starting in at really writing with the machine, at around one wafer/hour or so. "Our aim is to come out first at the end of 2005 with a tool for process development that can be used at customers," says Baud. Philips will produce the tool for Mapper, which has only 20 employees.The low-energy Gaussian beam system targets throughput of 10 wafers/hour.
Mapper was founded in 2000 by Peter Kruit, professor of physics at Delft University of Technology, and others, to use his Multi-Aperature Pixel-by-Pixel Enhancement of Resolution (MAPPER) technology. CEO Baud comes from the Dutch space industry's Fokker Space, and several other executives are recent physics graduate students from the university. The company has funding from the university; from Residex Venture Capital Network, part of a Dutch banking and insurance group; and from KLA-Tencor's KT Venture Group. Baud says they've also received some European government support.
Its revised split-beam approach moves upstart Mapper closer to competing directly with lithography giant Canon's similar maskless e-beam tool. Though Canon is very tight-lipped about this technology, Pennwell partner Nikkei Microdevices reports from Japan that the company currently has throughput up to several wafers/hour and aims to complete the basic tool development this year, ship a model for pilot production to users to evaluate in 2005, and ship a production tool by 2007. Target application is devices with surface area of 10mm2 or less, made in volumes of 60,000 units or fewer. Canon reported simulations at Sematech meetings last year suggesting throughput of up to 34 wafers/hour could be possible at 70nm resolution.
Canon's tool runs its e-beam through a collimator lens and then a correcting lens array, to split it into a 64 x 64 grid of 4,096 individual beams. A pair of lenses and deflectors within the correcting lens array for each beam adjusts its focal point to correct the aberration from the system. The idea is for each beam to simultaneously write its own 4 micron square micro field, creating a 256 micron square subfield at one shot, then electrically repeat these shots along a 4mm path, and mechanically move the stage to repeat these 4mm x 256 micron mainfields across the chip. The pilot tool will start with 2 micron square fields per beam for 64 micron square subfields and a 1mm x 64 micron main field. The delicate mechanical movements are made with an air stage that works in a vacuum, which Canon will also use in its other advanced lithography tools.
Leica Microsystems, meanwhile, aims to increase throughput while using one beam by varying the beam's shape to print the entire part of a pattern at one time in development work with STMicroelectronics in Crolles, France. Rather than a Gaussian beam tool with a round spot, Leica uses shapes such as rectangles, slanted-side rectangles, and triangles, with 45nm resolution, according to Jurgen Gramss of Leica, and the shapes can be varied in steps of 1nm. This greatly improves throughput, he added. Leica is also working with Gerhard Gross, formerly of Sematech, on a multi-beam e-beam system called ML2, for Mask-Less Lithography.
"STM in Crolles is very interested in direct write e-beam," says VSLI Research Europe director Bob Mariner, in Bedford, England. "The vast majority of product volume is done in low volume lots."
Tokyo Electron, Ebara, and Dai Nippon Screen, working with Toshiba, similarly use only one column for their maskless lithography tool, also targeting 10 wafers/hour throughput by 2005. But these companies say they get improved results from using a low energy (5KeV) and asymmetrical column, and speed up the process by breaking the pattern down not just into a few simple shapes, but into a library of 100 or so commonly repeated cells that can each be exposed in one shot. (See WaferNews, Volume 9, Number 37, Sept 16, 2002).
Advantest uses both multiple beams and a cell-pattern library, in yet another maskless e-beam alternative technology under development. Its multiple beams, however, are each generated by its own electron gun (See WaferNews, Volume 9, Number 42, October 21, 2002).