Are no nodes good news?
By Hank Hogan
When Intel (Santa Clara, CA) recently announced volume manufacturing in the second half of 2007 of a 45 nm process, something was missing: mention of a technology node. Intel isn’t alone in abandoning nodes. The latest International Technology Roadmap for Semiconductors (ITRS) refers to product generation cycles instead.
For those in contamination control, this change could be of more than academic interest. Killer defects have traditionally been defined as a fraction of a node, which is the half pitch of a cell in a memory process or the minimum transistor width in other processes. Thus, the node served as shorthand for a contamination control requirement.
In some ways, that hasn’t changed, says John Goodman, senior vice president and chief technology officer for materials handling company Entegris (Chaska, MN). “The basic rule of thumb of a half linewidth for killer particles is still applicable.”
However, in other ways things have changed because the guideline isn’t always strictly applied anymore. For example, it may be too difficult to develop a filtration membrane to capture particles while maintaining flow. What’s more, the composition of the contaminant may be more important than its size. For those reasons, Goodman says Entegris works with its customers to come up with the right contamination control targets.
As for spotting contaminants, it isn’t always possible to follow a simple node-based formula. The technology to detect particles at low cost and in high volume sometimes doesn’t exist, notes Particle Measuring Systems (Boulder, CO) vice president of marketing John Mitchell.
Thus, those running the 45 nm process might not screen directly for 22 nm contaminants. “There’s no practical way to measure 20 nm particles, particularly in liquids,” says Mitchell.
Instead, he explains, semiconductor manufacturers monitor larger particles. If the distribution of particle sizes doesn’t change and sufficient care is taken in monitoring, this strategy works.
Another consequence of the demise of nodes is that comparisons between processes and contamination control requirements are more convoluted. When evaluating offerings from different manufacturers, Dan McGowan, a spokesperson for SEMATECH (Austin, TX), suggests asking them how their figures of merit, whatever they are, compare to a memory cell’s half pitch. “The answer should help determine how a certain technology advance can be viewed in context with ITRS terminology,” he says.
However, that comparison could be easier to describe than do. Kari Aakre, a spokesperson for Intel, notes that where the company’s technology stands in relation to the technology roadmap depends on which metric is used. She lists such possible measures as transistor gate length, contacted metal pitch, cell size of static random access memory, and others.
While these questions present problems, those who pushed for the dispatch of node nomenclature say the change solves problems. Bob Doering, Senior Fellow and technology strategy manager at Texas Instruments (Dallas, TX), notes company announcements have historically tended to tout process cycles rather than true technology nodes that are 0.7 times the previous generation, a difference that led to trouble in the ITRS. “We were getting almost a whole node out of phase between what we were calling a node and what the common parlance was. So now we don’t use that terminology,” he says.