Can't Get Enough Activation

In an ideal world, dopant atoms such as phosphorus, arsenic or antimony are added to silicon in such a way that they nestle nicely into the silicon lattice, substituting for one silicon atom and bonding to the four surrounding silicon atoms. This way, they are electrically active and able to donate a free electron to the flow of electrons that make up electrical current. It's desirable to have a large current flow from the source to drain in a transistor when the transistor is "on," yet extremely little current flowing when the device is "off."

The problem, according to the International Technology Roadmap for Semiconductors (ITRS), is that, to get adequate drain current, transistor source and drain regions need higher concentrations of current carriers, and it doesn't look like that will be possible with existing technology. Future generations of silicon technology will require free-electron densities in excess of 10²¹ electrons/cm³, but today's best approaches provide no more than 5 × 10²⁰ electrons/cm³.

What researchers have found is that, even when they pack the silicon with so many dopant atoms that the material becomes supersaturated, they still can't achieve high enough levels of charge carriers. They have determined that, as dopant atom concentrations approach the high levels required for smaller devices — ~10²¹ dopants/cm³ — the electron concentration in the conduction band does not exceed 6-7 × 10²⁰ electrons/cm³.

What this means is that not enough of the dopants are getting activated. To find out why, researchers at Bell Labs , the R&D arm of Lucent Technologies (Murray Hill, N.J.), recently undertook a study using an advanced scanning transmission electron microscope (STEM), combined with rigorous sample preparation techniques, to analyze highly antimony-doped silicon. They believe it is the first time anyone has ever been able to examine individual dopant atoms as they exist under the silicon surface.

Top to bottom, a ball-and-stick model, a simulated micrograph and an experimental micrograph for a two-atom antimony cluster. (Source: Bell Labs)

What they found, as reported in the April 25 issue of Nature, was that two-atom clusters were the predominant deactivating mechanism. The paired defects outnumbered the three- and four-atom clusters by roughly 50:1. Although the concentration of dopants in a material is only a fraction of a percent — even in supersaturated materials — these two-atom clusters (Figure) apparently form just by chance. "You can get two atoms that are very close to each other, just from random statistics. When two atoms get close to each other, they can deactivate, if that's energetically favorable," Muller said.

"When you get to this concentration, there's a certain probability that another dopant will have a nearest neighbor," said Paul Peercy, a professor at the University of Wisconsin , Madison, and formerly president of International SEMATECH. "As long as the dopants are distributed randomly, I think it's going to be a challenge." The problem: Nobody knows how to add dopants non-randomly.

Today, most doping is done by ion implantation, which is followed by an anneal. The anneal helps remove some of the damage caused by ions plowing into the silicon, and also helps activate the dopants. "Ion implantation adds an extra wrinkle because it creates all this extra damage," Muller noted. "But if you can imagine some way of doing it without damage, by laser annealing for instance, then you've still got this problem of when these atoms get to very high concentrations — not all of them go substitutional and stay put. They can actually form pairs and clusters."

It should be noted here that phosphorus, arsenic and antimony are n-type dopants. P-type dopants, such as boron, work differently in that they contribute positive charge carriers (called "holes," indicating the lack of an electron) to current flow. CMOS devices use both p- and n-type doping in PMOS and NMOS transistors, respectively. But the need to get high concentrations of charge carriers and a high current density is the same for both types of devices.

For additional information on wafer processing, go to www.semiconductor.net/wafer