Intel does High-K and metal gates

INTEL HAS THIS way of putting out breathtaking scientific achievements in dull as dishwater ways. Today, it announced high-K + metal gate transistors at 45 nanometres, and it was as lively as watching a wafer go through a fab. Fear not, we shall endeavour to inform and entertain you about this subject.

There are two parts of this announcement, the high-K part and the metal gate part, and both are very different. K is the dielectric constant of the material used, high-K means a higher dielectric constant than the 3.9 of silicon dioxide, low-K means a lower number. You want high-K materials in some parts of your transistor, low in others. You can have a transistor with some parts high and others low, that is quite desirable actually.

Traditionally, the gate dielectric material is silicon dioxide, and for years, there has been a search for a material with a higher K to reduce leakage and have other beneficial properties. Basically, it would be a huge step in the otherwise shatteringly dull field of advanced semiconductor chemistry.

In a standard transistor, you have several major parts, the source, drain, gate dielectric and the gate electrode. It looks like this, S is source, D drain, and E is the metal gate electrode. The rest you can probably figure out for yourself, but do not go after the dragons without a raiding party of level 60 or higher, trust me.

It works like this, the electricity moves from the source to the drain, and in a perfect world, when the transistor is on, you get 100% of the electricity, 0% when off. Current flow when off is called leakage, and that is wasted power, therefor bad. When on, it never hits 100%, and how much voltage it takes and how long it takes to go from 0 to full on determines performance. The lower the voltage and the faster the ramp, the better.

The electrode on top basically turns it the transistor on or off and the dielectric modifies the on and off switch. The higher the K value, the better it is at controlling the modulation of the parts underneath it, basically on is closer to 100% off closer to 0%. If you get a K greater than 3.9, you decrease leakage and increase drive current. Life is good.

Intel's announcement today is that the silicon dioxide gate dielectric was replaced with one based on Hafnium, a material with a much higher K value. This makes things work better, faster and cooler. Sadly, Hafnium is extremely non-toxic unless dropped on one's head in large quantities, thus depriving Inq writers of juicy toxicity headlines for years to come. It also does not taste like raspberries.

The other half of the announcement is that they are using metal gates, which is the part right above the dialectic, basically the electrode. The old way is to use polysilicon for this material, Intel is now using metal for it. This removes a problem where the silicon dioxide and polysilicon reacted and lessened the effectiveness of the dielectric. Intel would not say what the metal is, with any luck, this one will be toxic, especially to bunnies.

In the end, you have the first really new way of doing things in years for a mass production semiconductor. It should be better in every way than the chemicals it replaces. Intel is claiming the usual 2x increase in transistor density, a given on the shrink between 65nm and 45nm, and better performance on top of it. 30% reduction in switching power, 20% faster switching speed or 5x lower source-drain leakage, and a 10x reduction in leakage through the dielectric layer. Not bad.

What you end up with is a chip that runs faster, cooler and takes less voltage. All good things. It will allow Intel to continue Moore's law from 65nm to 45, and possibly below. Nothing toxic, nothing exploding, just chips that work better and are cheaper to make. Exciting but dull. µ