Advertiser and

Buyer's Guide
Buyers Guide

Chip Shots blog

Greatest Hits of 2005
Greatest Hits of 2005

Featured Series
Featured Series

Web Sightings

Media Kit

Comments? Suggestions? Send us your feedback.



Interim edition includes no surprises, sets stage for 2005

As expected, no surprises lurk in the 2004 update of the semiconductor industry's technological guidebook. The 14 technology working groups reviewing the document determined that the 2003 version of the International Technology Roadmap for Semiconductors (ITRS) correctly assessed the industry's challenges and technological requirements. The incremental changes made in the latest update set the stage for the complete revision of the 2005 roadmap.

TECH EMERGENT: An overview of the emerging research devices chapter of the ITRS's 2004 update "illustrates the relationship of particular new concepts to the four functional categories that they each address."


The update notes that critical dimension (CD) control tolerances, the timing of the introduction of the 450-mm wafer size, and the necessity of improved yield models that include new materials are among the challenges facing the industry. The document also cited the need for new high-aspect-ratio inspection systems and improved data management for yield learning.

One significant change is the inclusion of all 15 years on the ITRS timeline. The addition of 2011, 2014, and 2017 means that the roadmap now features technology requirement parameters for each year, explains Peter Zeitzoff, a Sematech senior fellow active in the process integration, devices, and structures (PIDS) technology working group (TWG). The 2003 and previous versions of the roadmap contained technological specifications for only the short-term years—defined as the first seven years from 2003 through 2009. The segments of the map covering the long-term years specified challenges for six years up to 2018, but skipped 2011, 2014, and 2017. The updated document no longer has these gaps.

Here are other highlights in the 2004 edition:

The participants revised 128 of the 219 tables in the ITRS, added four tables, and updated 12 figures.

• Several chapters include new topics. In particular, a section on emerging research materials has been added to the chapter on emerging research devices.

• The low-k section of the roadmap remains unchanged for the first time in 10 years.

• Copper resistivity numbers are now included in the interconnect chapter to take into account surface- and sidewall-scattering effects.

• The yield enhancement chapter notes that edge exclusion is spread among that chapter as well as the chapters for front-end processes, lithography, and factory integration because of edge exclusion's importance to defect detection and characterization.

Comments obtained from participants in several key wafer-processing TWGs reveal the extent of the work that continues, both literally and figuratively, in the trenches. In the metrology chapter, the working group singled out CD measurement technology as an important new message.

Alain Diebold, a Sematech senior fellow and TWG participant, says the determination is supported by an SPIE conference presentation by Bryan Rice of Intel and by a report on scatterometry to the working group by Dan Wack of KLA-Tencor. Wack's "careful study" showed the impact of extending scatterometry to shorter wavelengths, while Rice's report showed CD-SEM and scatterometry of isolated lines measuring 16-nm wide with a 1:10 pitch as well as 36-nm nested lines with 1:3 and 1:1 pitches, Diebold points out.

"CD-SEM could resolve the line edges, and scatterometry and CD-SEM correlate well with each other," explains Diebold. "The need is for improved instrument performance so that a higher precision is achieved."

The metrology working group limns five difficult challenges at the 45-nm technology node or greater through 2009 and five difficult challenges beyond 2009 at nodes smaller than 45 nm. One key short-term obstacle for all metrology methods is the control of high-aspect-ratio technologies such as damascene processing. Group members note that the industry has yet to establish new process control requirements. In particular, they point out that 3-D measurements will be needed for trench structures in processes using low-k dielectric materials.

Diebold believes the industry is taking steps to address the issue. "The latest CD-SEMs have tilt capability," which permits the measurement of line shape and the ability to look down contact holes. He says the big challenge "is determining contact-hole dimensions in dual damascene."

The summary of challenges emphasizes that standards institutions such as the National Institute of Standards and Technology (NIST) must have rapid access to advanced development and manufacturing capability in order to develop the appropriate reference materials. Diebold says electron-beam lithography "is being pushed to its limits, and most places do not have access to a state-of-the-art system."

He stresses that National Science Foundation/National Nanotechnology Infrastructure Network (NSF-NNIN) facilities in the United States do have "this type of capability. For example, Cornell [University] has a new E-beam system." ATDF, a subsidiary of Sematech, offers fabrication of 300-mm wafers, Diebold adds.

The PIDS chapter features only minor changes from 2003. Among the most difficult challenges through 2010 at 45 nm and larger are the cost-effectiveness, process control, and reliability of very thin oxy-nitride gate dielectrics. Specifically, the summary says that the gate dielectric will be unable to meet gate-leakage current limits by 2006. Sematech's Zeitzoff points out that this inability applies only to such applications as low-standby-power and low-operating-power logic devices, "where the leakage current specifications are low, and it is absolutely critical to meet them. For high-performance logic, silicon oxy-nitride gate dielectric is projected to be able to meet the gate-leakage current limits until 2007.

"The industry's preferred approach to dealing with this issue is to implement high-k gate dielectric stacks, which have a greater thickness than oxy-nitride gate dielectric for the same equivalent oxide thickness," Zeitzoff continues. "As a result, if the high-k dielectric has a sufficient energy barrier relative to silicon, the gate-leakage current will be lower for the high-k dielectric."

How well will process integration teams cope with materials, process, and structural changes coming at such a rapid clip? Zeitzoff says it's unclear whether the industry will meet the projected deadlines for the potential solutions in the PIDS chapter. The PIDS group believes the introductions of the most critical innovations will follow this schedule: strained silicon for enhanced channel mobility, 2004; high-k gate dielectric for low-power logic, 2006; metal-gate electrode and high-k gate dielectric for high-performance logic, 2007; ultrathin-body, fully depleted (UTB-FD), single-gate silicon-on-insulator (SOI) MOSFETs, 2008; and UTB-FD multiple-gate MOSFETs such as FinFETs, 2010.

Transistor requirements will control the timing of these innovations, Zeitzoff points out. Current "solutions" considered inadequate to meet technological requirements are shown in the roadmap with projected timelines for the appropriate innovation needed to overcome the problem. He emphasizes that the timing of the innovations "is meant to represent the leading edge of the IC industry as a whole."

Because transistor scaling with its varied parameters is complicated, the industry will need to use "workarounds" for innovations that prove unworkable in their projected introductory year, the Sematech senior fellow says. "For example, if high-k gate dielectric is not available for deployment in 2006 as projected, then one approach would be to slow the scaling of EOT [equivalent oxide thickness] to meet the gate-leakage current limits even with oxy-nitride. To compensate, some of the other parameters, such as source/drain extension junction depth and the mobility enhancement from strained-silicon channels could be scaled more aggressively."

Zeitzoff insists that the industry has had good success at meeting these challenges, although the timing of the solutions is difficult to predict accurately. Notably, chipmakers are already using strained silicon in their processes according to projections. Nevertheless, "each of the other innovations has significant technical challenges remaining, including ensuring reliability and yield."

Those two factors, along with cost, are also major components in the factory integration focus area. "Factory integration plays a critical role in ensuring that our factories and enterprise systems are designed and integrated to enable process development and high-volume manufacturing so that the right products are made in the right volume on schedule with high yield, while meeting cost targets," says Mani Janakiram, cochairman of the factory integration TWG and a manager for analytics and visualization at Intel.

Chipmakers may have to turn to improvements in fab productivity, having wrung as much cost savings as possible from advancements such as reductions in feature size and near-perfect yields. Reaching the next level of cost efficiency requires breakthroughs in factory design and manufacturing. Janakiram attributes significant decreases in cost per function to these advancements, but he notes that much more cost reduction "is expected through improved factory design, system integration, and overall factory/enterprise optimization."

Janakiram and his TWG say the industry must focus on three priorities: airborne molecular contamination (AMC), conversion to the 450-mm wafer size, and proactive visualization. AMC's impact will be most severe at the 45-nm node, because "it could be a potential technology barrier and yield limiter," he points out.

The Intel manager notes that it took the industry more than nine years to "fully migrate from 200 mm to 300 mm, and it is now reaping the benefit of approximately [a] 30% improvement in cost. The same is expected from migrating to 450 mm." Because the roadmap projects the conversion to become effective in 2012, the industry must begin its assessment and analysis now, he insists.

Proactive visualization takes into account "the interdependence of all facets of semiconductor manufacturing," he says. The concept gives IC manufacturers "a seamless and transparent view" of a fab's information flow so that they can make production decisions based on good data.

Over the past decade, the importance of the environmental, safety, and health (ESH) area has grown significantly. Among the key ESH challenges cited in the 2004 update are the need for quality rapid assessment methodologies for new chemicals and the need to design energy-efficient process tools.

Jim Jewitt, chairman of the ESH working group, joined the team when it first formed in the early 1990s. He defines new chemicals as any "not being used today." The roadmap also defines advanced photoresists and high- and low-k materials as examples. Researchers need "some preliminary indications of ESH properties on potential candidate chemicals that do not already have sufficient ESH information."

Walter Worth, a Sematech fellow and ESH working group member, says examples of new materials include salts to raise the refractive index of fluids for 193-nm immersion lithography and chemicals such as hafnium and hafnium compounds for high-k processes. Methodologies could include supplementing data on material data safety sheets with a search of Internet databases and other resources in order to compile a matrix with approximately 75 ESH data types, he notes. Characterizing emissions from tools using the new materials is another potential method for rapid assessment.

Two long-term ESH challenges stand out in the updated document: the need to understand regulatory requirements that affect chemical restrictions and the need for broad material life-cycle analysis. Jewitt says that the industry has had great success mitigating ESH risk in chemical usage. However, simple mitigation efforts may no longer be sufficient.

"More-recent regulatory trends are for use restriction or prohibitions rather than mitigation," he points out. "The need is to comprehend potential for future restrictions during the early, or long-term, development of new materials, rather than learning that a material is not usable after 10 to 15 years of research has been completed."—JC

MicroHome | Search | Current Issue | MicroArchives
Buyers Guide | Media Kit

Questions/comments about MICRO Magazine? E-mail us at

© 2007 Tom Cheyney
All rights reserved.