Atomic layer deposition (ALD), one of the semiconductor industry’s workhorse technologies, is responsible for creating ultra-thin films of exotic materials that allow so many of our appliances to work their electronic magic. Despite this indispensable role, however, ALD remains an expensive and highly inefficient process. It depends on precursors containing highly purified, rare metals such as hafnium, iridium or zirconium, very little of which actually makes their way into the finished product.
“You’re lucky to use one percent and typically it’s much less — 0.1 percent or 0.01 percent,” says Ken Cadien, a professor in the University of Alberta’s Department of Chemical and Materials Engineering. The problem, Cadien notes, is that deposition occurs within a flow reactor characterized by an A-B cycle, the first step being millisecond-scale pulses of precursor and the second a reactant pulse of oxygen (for growing oxides). In principle the precursor molecules should settle on a prepared surface during the first step. In reality, very few do so and many are just as likely to desorb from the surface.
Triratna Muneshwar, a post-doctoral fellow working with Cadien, took a closer look at the mechanics of this cycle and tweaked it in what seemed like a modest way. The result turned out to be far from modest. “It’s a fairly simple idea but he came up with a model that breaks up step A into much smaller steps,” Cadien says. “By breaking up the pulse into shorter pulses, the molecules actually wind up having more time to be accommodated on the surface. We predicted it with a mathematical model and the experiment nailed it. For some precursors, we got a 50 percent reduction in the required amount.”
That kind of reduction could mean big savings for semiconductor manufacturers like Intel, whose representatives displayed some surprise and a great deal of interest when Muneshwar presented the findings at an ALD conference in Portland, Ore. last year. He and Cadien subsequently published these results in the Journal of Applied Physics, where it was made a highlight article by the editor. “It’s one of these things where everybody says, ‘that’s obvious’ but nobody thought of it,” Cadien says. He concedes that firms in the field are tight-lipped about their work, so it is unclear if any of them had independently tried this approach. “It’s really an innovative way of looking at the process. It doesn’t fundamentally increase the process time, there’s no extra cost, no new equipment. It’s nice when theory comes together with application and saves people money while conserving critical materials.”