by Debra Vogler, Senior Technical Editor, Solid State Technology
The tradition of the ISMI Manufacturing Symposium as an “in the trenches” conference, by and for the people doing the actual day-to-day work in the fabs, continued last week (Oct. 24-25, Austin, TX). Particular attention was given to topics on yield improvement/productivity methodologies, ESH, and sustainability. Specific talks included how to fix litho “hot spots,” cleaning wafer chucks without DI water or solvents, guidelines for detecting/preventing electrostatic discharge (ESD) events, and why process stability may be the deciding factor in the battle over 450mm wafers.
TI hits up hot spot reduction for scanners
Eric Wagner of TI’s DMOS 6 equipment engineering group discussed that fab’s No.1 cause of photo rework — hot spots — and what the group has been doing to address it. Most of the hot spots are attributed to backside particles and result in excessive scanner downtime, high cycle times due to rework, and yield excursions, he noted. Data from DMOS 6 showed an ~5% overall impact on scanner availability due to hot spots and as much as a ~7% impact on i-line scanner availability. Hot spots were the cause of >30% of all photo rework.
The key challenge to finding the root causes of hot spots, Wagner pointed out, is that the backside contamination that causes them comes and goes. Furthermore, backside contamination is often seen as the problem of the lithography group, he added, even though it may have been caused prior to the wafer’s arrival at photo, so the litho group is left with the task of addressing it.
TI is using a three-pronged approach to detect hot spots at photo: a focus spot monitor (FSM) on-board the ASML scanners, a KLA Viper tool for macro post-pattern/pre-etch inspection, and YE inspection for pre-etch and post-etch. These efforts are augmented by ASML’s ultralow-contact wafer chuck, called Mini Burl. According to Wagner, the Mini Burl uses the same material as the standard Twinscan chuck and its contact points are in the same location as the standard chuck, but there is reduced contact area between the wafer the and wafer table. The Mini Burl is also less sensitive to wafer backside contamination.
By using the Mini Burl, TI has achieved up to a 65% reduction in hot spot rework and increased scanner availability by >3%, Wagner noted. The Mini Burl also requires less chuck cleans, hence less machine wear and other mechanical issues.
TI is also testing an automatic chuck cleaner option on Twinscan that came out in 2Q07, called SpotLess, which is a cleaning stone in a holder under the scanner frame that cleans a chuck as it moves underneath the stone. Noting that SpotLess is currently available only on XT platforms, Wagner appealed to ASML to consider bringing the option to the AT platform, as its evaluation at DMOS 6 on an ASML XT450 scanner resulted in being able to clean a single spot in ~1 minute and the whole chuck in ~20 minutes. In comparison, manually cleaning a hot spot from tool down to tool up takes 1-3 hours.
AMAT improves litho throughput with focus spot removal clean
Among its many business areas, Applied Materials is involved in improving the throughput of lithography tools. Brent Ames, global business manager for Applied’s Global Services FOS Silicon Technologies Group, provided details on a polymer film cleaning method developed by the services group to clean wafer chucks in scanner tools — in this case, ASML Twinscan models — in lieu of the industry standard DI water or solvent-based clean, or even stoning.
Using a polymer film (on a wafer) to trap (embed) particles, even particles in-between pads can be eliminated as the pads conform to the chuck, Ames explained. Data showed that focus and chuck spots were completely removed by the single pass of the polymer wafer on an ASML tool. The upshot of the method, he noted, is that it cannot outgas or leave residue behind, nor does it transfer contamination to other areas of the tool. The time for focus spot removal and tool re-qualification for production was 20 min vs. the typical 1 hr.
Making a pitch for SEMI E-78
Andrew Rudack of SEMATECH made a pitch for using ESD event detection as a predictive maintenance tool, as well as using the SEMI E78 standard (“Guide to Assess and Control Electrostatic Discharge (ESD) and Electrostatic Attraction (ESA) for Equipment”). The latest edition of the document was approved in May 2006. Rudack developed an ESD event detection technique for use in several different litho applications including reticle damage detection, as well as in a 300mm stepper and a 200mm scanner.
The last word on 450mm manufacturing
Between the discussions about going to 450mm wafers that took place at ISSM and the comments from the ISMI symposium (see stories: Keys to pushing 450mm: productivity and “agility,” and Will a transition to 450mm wafers leave equipment suppliers behind?), it seems pointless to continue dishing about this still hotly-contested issue. That said, many dedicated people are hard at work on developing requirements and the beginnings of standards, so there are sure to be a number of productivity improvements that come out of the endeavors even if they do not make it to 450mm manufacturing.
Michael O’Halloran of CH2M Hill, a panelist on the “Trends in Fab Design” session, summed up the dilemma in response to a question about 450mm. He noted that there was also much discussion and preliminary work done in the years preceding the transition to 300mm wafers, but nothing really happened until there was stability in the manufacturing process. “Until you can predict what kinds of tools you will have at 450mm (and that really means knowing what’s ahead 2-3 nodes in advance), it’s too risky to move forward.”
O’Halloran’s comments speak volumes for where the industry is today. Some major IC manufacturers clearly will be pushing for 450mm, attracted by gains in productivity and lower costs. But with the uncertainty and wide variety that lies ahead with respect to materials selection at 32nm and beyond, the status of next-generation lithography beyond 32nm, as well as the economics and source of R&D funding for the 450mm effort, many in the industry think that the bulk of the attention should be paid to 300mm Prime and the specific choices that will be made for 32nm and 22nm. Manufacturing at 450mm, at least for now, seems a distant dream. — D.V.
Next week’s WaferNEWS will take a closer look at a specific ISMI presentation that described how the statistical technique of multi-variate analysis can be applied to chamber matching.