BACUS reveals latest trends in RET, etch, and e-beam
11/01/2000
More than 600 attendees of the 20th annual BACUS Symposium on Photomask Technology discussed the latest efforts to optimize application of reticle enhancement technology (RET). Particularly evident at the conference, held in September in Monterey, CA, was the industry's current "roadmapped" progression of phase shift mask (PSM) technology from attenuated PSM to more difficult, costly alternating PSM technology (see "Chairman Morgan: Applied wisdom at BACUS"). Attendees also seemed to be looking for significant advances in dry etch technology for maskmakers and longing for evidence of the sorely needed next generation of mask writing tools.
Reticle enhancement
Clearly, the components of RET are increasingly being developed into advanced production lithography.
Among the many examples discussed at BACUS, Byeongsoo Kim from Samsung Electronics' Semiconductor R&D Center, Yongin, Korea, detailed his company's application of KrF lithography with off-axis illumination (OAI), optical proximity correction (OPC) assist features, and attenuated PSM for 110nm DRAM device structures. "We have used half-tone PSM and strong annular OAI to improve process margin for cell pattern. We've applied assist features to isolated line, reducing iso-dense bias in the peripheral area. These methods are being applied for full chip level gate-poly patterning of DRAM devices with 110nm design rules," said Kim.
At the leading edge of production lithography, a fair share of the 49 papers presented at BACUS (the conference also included 59 poster papers) addressed the difficulty of applying PSM with emerging ArF lithography for 100nm circuit patterns. For example, Haruo Iwasaki from NEC, Kanagawa, Japan, reported the evaluation of attenuated-PSM with ArF to print 100nm-line, 150nm-space patterns for logic gates. Using annular illumination with an 0.60NA ArF scanner and chemically amplified positive resist, NEC engineers have suppressed a 23nm iso-dense bias using OPC and achieved a 0.4mm common depth of focus (DOF).
As reported repeatedly in BACUS presentations (and discussed at breaks), the emerging, apparently winning, advanced lithography application for 100nm and beyond combines ArF (or even KrF) with a double exposure alternating PSM scheme — an initial exposure for critical layers using dark field PSM followed by a trim mask exposure that eliminates unwanted patterns from the first mask (see "A new strong-PSM paradigm: Phase Phirst!" on p. 47).
Sang-Sool Koo from Hyundai Electronics Industries, Ichon, Korea, said, "We have achieved 60nm transistor patterns with good process margin and good resist pattern fidelity (less resist top loss and line-edge roughness). Our CD variation was controlled within 5nm. This work shows the possibility of sub-100nm patterning by using 4x masks and double exposure alternating PSM technology with ArF lithography (the ASML PAS 5500/900)." Hyundai lithography engineers used an in-house developed ArF resist in this work. In addition, phase error for lithography with the 193nm wavelength was estimated by using experimentation and simulation, the latter from Hyundai's own OPC simulation tool base on diffused aerial image modeling.
In related work reported by other engineers at Hyundai, simulation and experiments have shown that 0.63NA ArF lithography using attenuated PSM achieved a 15% resolution improvement compared to 0.70NA KrF.
Speaking specifically about the viability of double exposure PSM, Tomohiko Yamamoto, an engineer at Fujitsu, Mie, Japan, said, "Investigating the impact of fabrication error in alternating PSM to determine the quality required for the dual exposure and focusing on phase accuracy, undercutting, and quartz and chrome defects, our experimental results showed that they do not need to be as strictly controlled as previously thought."
Dry etch
The challenges in today's maskmaking could benefit from a more robust dry etch technology, but success has been difficult. Here, "loading effect" — a mask's pattern density effect on plasma — is one of the great issues. One advance reported at BACUS came from Tatsuya Fujisawa, an engineer at Selete, Yokohama, Japan. Using a new etcher capable of achieving high plasma density at low pressure (1mTorr), Selete engineers have reduced loading effect in experimental work that obtained 10.8nm (3s) etch uniformity for 0.5mm lines and space patterns on a 123mm x 123mm area. "We obtained vertical chrome profile on both clear field and dark field 0.5mm patterns," said Fujisawa.
A dry etch system for photomask processing has been developed by PKL, Choongnam, Korea. Its new plasma source achieves a plasma density of ~1011-1012 cm-3 and <5% plasma uniformity, both over a 6 in. x 6 in. surface. The system incorporates chromium and MoSi end point detection via monitoring Cr and Mo radical emission intensity. PKL engineer Hyuk-Joo Kwon said, "The three sigma of CD uniformity of Cr pattern in a 132mm x 132mm area was <10nm, where the target CD was 0.8mm. Further, Cr and MoSi slopes are 88-90°, which shows that we are achieving a highly anisotropic etch." In this work the resist to Cr selectivity was >1.3 at a clear area ratio of <60%. "Selectivity was mainly affected by oxygen partial pressure and clear area ratio," said Kwon. In work with PSM, the phase uniformity achieved was 180±1° and transmittance uniformity within ±0.2%.
Writing tools
With maskmakers looking for next generation writing tools, particularly for the 100nm technology node, BACUS attendees were looking for signs that the contenders will soon be available. The possible solutions discussed at the conference included two systems still in R&D or beta site testing.
One senses that Etec Systems' (Hayward, CA) new tool will soon be available. This is a raster based e-beam solution with improved resolution and CD uniformity, positional accuracy and throughput. The e-beam system includes a column with 50kV accelerating voltage and a new gray-beam writing technique. The latter incorporates a pulse-width modulated blanking system, per pixel deflection, retrograde scanning, and multiphase and multipass writing, using conventional high contrast resists. Details presented at BACUS revealed that this system meets the challenges of the 130nm device generation with extendibility to at least 100nm, but it is not yet commercially available.
Micronic Laser Systems, Taby, Sweden, is developing a new breed of DUV spatial light modulator (SLM) technology that provides a means for high resolution and massive parallel exposure for maskmaking. The parallel exposure uses an array of micro mirrors individually and electrostatically addressed to a reflecting or nonreflecting state. Company engineer Tor Björn Sandström said, "Analog modulation of each pixel provides an address grid satisfying the requirements of the industry roadmap without speed penalty.
The process repeats as in a stepper." While the new SLM technology is still very much in the laboratory, Sandström notes, "it has the resolution to rival that of e-beam pattern generators yet provides the productivity of laser patterning." —P.B.
Chairman Morgan: Applied wisdom at BACUS
Answering questions after his keynote address at the 20th annual BACUS Symposium on Photomask Technology held recently in Monterey, CA, James Morgan, chairman and CEO of Applied Materials, was asked for advice about how maskmakers can price their products for the value that they provide to semiconductor manufacturers.
"That is a tough one," said Morgan. "Customers don't necessarily appreciate how much value you provide. We had the same problem with 300mm. Customers didn't take ownership of 300mm as they did [with other wafer size changes] in the past; they didn't put any commitment into the 300mm transition. Our [equipment supplier] industry carried the whole risk.
"Of course we have had a really strong upturn, so it doesn't appear that we have been hurt, but our industry spent several billions of dollars at total risk with no commitment by our customers to go forward. I have never done that in my business career; it is a stupid thing to do. But your customers can be pretty persuasive, even make you do stupid things. It will work out, but if you think about it, we had to bring out 22 tools and about 80 processes in order to meet that commitment."
"All of us [equipment vendors] are under a lot of pressure to enable our customers to have this technology and to provide the capital investment to make it happen. Some of them may not appreciate how much investment you have to make in order to be successful over time. If they don't appreciate that, then only the ones with capital are going to survive. [In the education process] you have to go higher up because the purchasing manager has a diode mind." —P.B.
A new strong-PSM paradigm: Phase Phirst!
At one of the better attended BACUS paper presentations, Marc D. Levenson of M.D. Levenson Consulting, Saratoga, CA, revealed a creative new approach to strong PSMs (as well as giving a lesson in effective use of a speaker's allotted 20 minutes). Levenson calls it a new paradigm and dubs it "Phase Phirst! — a method to reduce PSM costs and even delay the need for next generation lithography techniques."
Setting the challenge, Levenson said, "The mask contribution to the cost of production is becoming very large and even prohibitive for manufacturers of application specific ICs (ASICs) and system-on-chip (SOC) ICs. Remember that half of all reticles are used to make less than 570 wafers." Levenson and his collaborators have examined the question: What would it take to make alternating PSM affordable for ASIC and SOCIC manufacturers? "The answer to that question is found in sidewall chrome alternating aperture (SCAA) masks," said Levenson, who published the technique in 1992. "Basically, you make a defect-free chromeless mask with 'the right phase pattern,' re-coat it with chrome and resist, and write the needed exposure pattern in the chrome layer. You end up with a PSM structure where all quartz walls are covered by opaque chrome and all chrome is supported by quartz. With all silica sidewalls covered and all chrome supported, the SCAA mask is largely immune to the phase and amplitude anomalies that cause space-width alternations as well as the design, fabrication, and cleaning difficulties that plague other PSM structures," he said.
Briefly described, Phase Phirst! starts with a prepatterned substrate, with the pattern chosen from a small select list. "The mask house does one write, one develop, one etch, one inspect, and maybe one repair, and ships a finished SCAA mask with a turnaround time similar to today's chrome on glass (COG) mask. You end up with a strong PSM with the same effort for what you do for COG, using a substrate with the right phase pattern on it," said Levenson (see illustration).
Behind the development of Phase Phirst!, Levenson worked with co-authors John Petersen and David Gerold at Petersen Advanced Lithography, Austin, TX, and Chris Mack at the Finle Division of KLA-Tencor, Austin, TX, as well as collaborators from DNP, Canon, JSR, Sigma-C, and KLA-Tencor.
"My assertion is that there is a limited number of useful phase patterns that will address a majority of ASIC and SOC designs. With the right trim mask and the right pattern in the chrome over these phase steps, you get the desired circuit pattern," said Levenson.
Referring to many of the BACUS papers that reported on the difficulty in getting PSM just right, Levenson said, "PSM structures matter; I think you have heard that today. But the SCAA mask has the advantage that the environment of each space is identical independent of the phase. That means that the electric field at the plane of the aperture and the brightness of the image in focus is the same, which is not the case with dual trench [strong] PSM designs where the amount of light going through the aperture is more controlled by the bottom of the etched trench."
Early wafer results with the Phase Phirst! paradigm have shown good results down to 125nm half pitch. "When we tried 110nm, the resist fell over, so while what we did shows no pitch walking, we were limited by resist collapsing," said Levenson.
Wrapping up his presentation in record time for a BACUS presentation, Levenson outlined his paradigm. "Maskmakers produce identical substrates with useful phase patterns using wafer fab techniques and stepper (or even projection or proximity) lithography. A good substrate here doesn't necessarily mean perfect. It is going to be used for dark field exposure (most of the substrate is going to be covered with chrome) so the defects can be mapped and accounted for in the design process. Then, you design circuits to place the fine features at predefined phase shift locations; interconnects can go anywhere because they are on the other masks. The phase substrate becomes the dark field PSM and is used with conventional COG trim masks."
What is this going to cost? "Guaranteed defect-free SCAA mask substrates will be manufactured in large quantity and at low cost if the design grids become standardized." Levenson believes that, depending on the mask pattern, as volumes approach 1000 to 10,000 masks, the cost plateaus at something not much more than last generation COG. "Clearly, SCAA masks can be made, and they work better," he said. The main question is how are we going to get the designers to do it this way? The design tool companies say that this is easy to do. But you have to have the designers using these tools. There is a lot of money at stake here. If the designers that you have now will not use these tools, fire them and hire new ones who have more hunger," Levenson quipped.
As an answer to the stated question for the special BACUS "k1-limbo" session — "How low can you go?" — Levenson replied "$9999.95, with Phase Phirst! in lots of 1000!"