By Paula Doe
WaferNews Contributing Editor
Nikon reports progress on both the optics and the infrastructure for its e-beam projection lithography, and argues it may still be the best available option for printing critical contact layers at 70 to 35nm, even with low throughput.
The company’s strategy is to first introduce its e-beam tool as a contact level printer. “We see EPL as complementary to 157, because 157 will still have trouble with contacts,” says John Wiesner, VP engineering, Nikon Precision Inc.
He argues that optical lithography has trouble with the inherently three-dimensional contact structures, and will have to do two or even three passes, using a very expensive phase shift mask and then a trim mask or two to fix the problems from the first pass, reducing throughput to about a third. But 3D features are far easier for e-beam, with its greater depth of focus. And there are three to four of those critical contact layers on every chip.
“It’s still a huge market,” says Wiesner, “Well worth the investment.”
Nikon plans to install two alpha research tools in 2003, one in the US and one in the Far East, which apparently means one will go to Selete. It aims to have a beta tool out by 2004, which will probably tweak the PREVAIL optics developed by IBM by shortening the column and reducing the geometric aberrations to allow an open convergence angle to improve performance.
Then Nikon plans significant changes – which it will discuss at SPIE – in the optical parameters for the generation after the beta tool, using a new approach to controlling the electron interactions to improve throughput without sacrificing resolution.
Nikon researchers say they can control coulomb interactions by reducing the convergence angle and increasing the size of the exposure subfield, which also reduces the space-charge effect.
They claim to get 10 times better depth of focus than optical lithography, allowing 0.07-micron contact holes, 0.045-micron lines and spaces, and individual 0.035-micron lines – at 30 to 40 wafers/hour – according to a paper in the December issue of Nikkei Microdevices.
Throughput of the current R&D system is under 10 WPH, but researchers say the 70nm production tool should be able to do 15 WPH, and then adding the new optics and the high speed stage should get that up to 30 WPH for 50nm and 35nm.
Though it now looks like chip makers will be able to make their 100nm generation devices with 193 optical scanners using the improved high NA 0.78 lenses and enhanced resists, the 70nm node remains up for grabs.
“Don’t ignore these [e-beam] folks,” says Griff Resor, president of Resor Associates, Boxborough, MA. “It’s still not clear that 157 is going to be real. Something sure delayed it a year or two. They need resists, pellicles, CaF2.”
He notes that while it looks like software fixes will be able to adjust for the birefringence problem, the first round of lenses still won’t be very good.
And EUV?
“They have a 4W bulb, and they need to be at 150W,” says Resor. “This point source technology has proved to be very difficult.”
Resor also notes that printing contacts means the mask is mostly opaque, which simplifies dealing with those hard-to-control e-beams. And there are some infrastructure advantages to e-beam, with good resists available and lots of people around who know how to repair e-beam systems.
Nikon has drummed up good support from Japan’s suppliers of the necessary infrastructure. HOYA is making mask blanks and intends to provide patterned masks, planning to ship samples and possibly orders of small research quantities this year. Dai Nippon Printing and Toppan will supply the masks too, either on HOYA blanks or their own.
“The Japanese suppliers are totally committed,” says Wiesner, “Though Photonics and Dupont are taking a more wait and see approach.”
Researchers in Japan say they have 100mm diameter 2-micron silicon masks now, and expect 200mm versions later this year. Wiesner notes that now that rival Leepl is using a mask technology similar to Nikon’s, with support bars dividing the mask into subfields, its masks will be much more expensive too, and all the pattern placement and alignment is much easier with Nikon’s 4x reduction than with Leepl’s 1x approach.
Seiko Instruments is testing data conversion software that takes output from the e-beam tool and converts it directly to the mask writer, based on its software that converts CAD system data.
“So far it doesn’t flunk,” says Wiesner. “And the times appear acceptable.”
Nikon has also developed an air stage that works in the vacuum required for e-beam, using differential pumping to remove the air in stages. The servo control of this stage moves at more than 100Hz, and the vacuum within the chamber remains at 10-4 Pa or less. The company is now just finishing integrating the wafer alignment, mask alignment and auto focus systems from its optical scanners with the EB optics.
“I was personally blown away by the air-bearing-in-a-vacuum-with-magnetically-shielded-linear-motor technology that they are using for their stages,” says another lithography expert, who still remains skeptical about the PREVAIL optics. “Those guys are serious.”
WaferNews