In the Penryn processors that we recently began fabricating, most of their transistors' features measure around 45 nm, though one is as small as 35 nm. It's the first commercial microprocessor to have features this small; all other top-of-the-line microprocessors in production as this article is being written have 65-nm features. In other words, Penryn is the first of the 45-nm generation of microprocessors. Many more will soon follow.

The thickness of the SiO₂ insulation on the transistor's gate has scaled from about 100 nm down to 1.2 nm on state-of-the-art microprocessors. The rate at which the thickness decreased was steady for years but started to slow at the 90-nm generation, which went into production in 2003. It was then that the oxide hit its five-atom limit. The insulator thickness shrank no further from the 90-nm to the 65-nm generation still common today.

The reason the gate oxide was shrunk no further is that it began to leak current [see illustration, “Running Out of Atoms”]. This leakage arises from quantum effects. At 1.2 nm, the quantum nature of particles starts to play a big role. We're used to thinking of electrons in terms of classical physics, and we like to imagine an electron as a ball and the insulation as a tall and narrow hill. The height of the hill represents how much energy you'd need to provide the electron to get it to the other side. Give it a sufficient push and—sure enough—you could get it over the hill, busting through the insulation in the process.

But when the hill (the oxide layer) is so narrow that you are counting individual atoms of thickness, the electron looks less like a ball and more like a wave. Specifically, it's a wave that defines the probability of finding the electron in a particular location. The trouble is that the wave is actually broader than the hill, extending all the way to the other side and beyond. That means there is a distinct probability that an electron that should be on the gate side of the oxide can simply appear on the channel side, having “tunneled” through the energy barrier posed by the insulation rather than going over it.

In the mid-1990s, we at Intel and other major chip makers recognized that we were fast approaching the day when we would no longer be able to keep squeezing atoms out of the SiO₂ gate insulator. So we all launched research programs to come up with a better solution. The goal was to identify a gate dielectric material as a replacement for SiO₂ and also to demonstrate transistor prototypes that leaked less while at the same time driving plenty of current across the transistor channel. We needed a gate insulator that was thick enough to keep electrons from tunneling through it and yet permeable enough to let the gate's electric field into the channel so that it could turn on the transistor. In other words, the material had to be physically thick but electrically thin.

The technical term for such a material is a “high-k” dielectric; k, the dielectric constant, is a term that refers to a material's ability to concentrate an electric field. Having a higher dielectric constant means the insulator can provide increased capacitance between two conducting plates—storing more charge—for the same thickness of insulator. Or if you prefer, it can provide the same capacitance with a thicker insulator [see illustration, “The High-k Way”]. SiO₂ typically has a k of around 4, while air and a vacuum have values of about 1. The k-value is related to how much a material can be polarized. When placed in an electric field, the charges in a dielectric's atoms or molecules will reorient themselves in the direction of the field. These internal charges are more responsive in high-k dielectrics than in low-k ones.

Incidentally, back in 2000, leading semiconductor firms began to change the material used to insulate the metal wires that connect transistors to each other from SiO₂ to low-k dielectrics. In the case of interconnects, you do not want the electric field from one wire to be felt in other nearby wires, because it creates a capacitor between the wires and can interfere with or slow down the signals on them. A low-k dielectric prevents the problem.