Optical Lithography at 45nm

[optoman]

I went to a talk last night at the local OSA (optical society of america) chapter and they found a way to produce 45 nm semiconductors. It was relatively easy to do and Intel, with all the other boys should be getting there faster because of their recent breakthroughs. They best part about it was that they could use the current wavelengths at 193 nm instead of going to a shorter wavelength bypassing all the problems associated with a shorted wavelength.

We were the first group of people outside the research group to have seen their work. They showed us pics of photo resist with 45 nm waveguides in them. Forget 90 nm chips, I want the 45 nm ones.

 

[pm]

What kind of breakthroughs were discussed? OPC and phase-shifting improvements? Or something beyond these two?

 

[optoman]

I really don't know all the technical terms but basically they were using immersion optical lithography with water. Said they could write as small 45 nm and had pics to prove it. All the components they used were readily available except for a couple of optical pieces which had to be custom made. They were using NAs small enough not to give them problems.

 

[Eskimo]

Yes immersion lithography is definately an exciting possiblity to extend current ArF systems to process at much smaller dimensions. I was lucky enough to have some exposure with one of the first university research groups to work on immersion. We were able to print 65nm features using a relatively rudimentary 193nm optical column.

For those of you who haven't heard of it basically by replacing the air (actually nitrogen in today's systems) between the objective lens and the wafer with a liquid medium you can realize a significant gain in how small of a feature you can resolve. Water is the most promising canidate since we know the most about it and it's interactions with the wafers and resists in use. Theory, in the form of Rayleigh's equation, tells us that the minumum resolution one can print is expressed as R=(k1*lamda)/NA. Where k1 is a process determined constant, lamda is your expsoure wavelength and NA is the numerical aperature of your lens system. Traditionally we would express NA as equal to the sine of the half angle subtended by the lens. In otherwords NA=sin(alpha). But this expression is actually a simplification as we are ignoring the medium in which the lens is present since the index of refraction of air is esentially 1.0. The correct expression would be NA=n*sin(alpha). Water at 193nm has an index of ~1.47, so in essence your effective NA is multiplied by that much allowing you to resolve much smaller features.

For a nice little summary of the topic see
this short paper

 

[optoman]

Eskimo hit it right on the head. These guys did exactly that and are ahead of the curve. He speculated that we should see it being used around 2005 instead of 2007. After they finish fine tuning the 45 nm process they are going after the 22nm. This is theoretically the limit for optical lithography because the wavelength of the source will go to 153nm and water is no longer transparent. If someone can create a medium that acts like water at this wavelength then they might have a chance of going smaller then 22nm. They are thinking something like teflon might be a good candidate and are working with Dupont to find something off the shelf.

 

[f95toli]

Below 22 nm is almost impossible. Even with e-beam lithography it is difficult to go below 20 nm and there you are free to choose tha wavelength (by changing the acceleration voltage in the column). There are issues with the sensitivity of the resist, scattering and last but not least quantum mechanical effects that will eventually start to affect the process as the dimensions gets smaller.

And right now I do not see the point, you would need a whole new set of design concepts and materials in order to be able to make ICs with a 20 nm linewidth since quantum mechanical effects will be dominating. There are more important problems like insulators and gate materials that needs to be solved first.

 

[tart666]

The correct expression would be NA=n*sin(alpha).

actually, an even more correct expression is lambda_water=lambda_vacuum / n

NA is defined in terms of angle, and geometry of the situation, which does not change. The wavelength actually physically changes here, the benefits of which we see in improved resolution. Btw, the DOF will scale linearly with n this way, instead of going down with the square of NA, to our benefit.

To optoman: there will always be pictures of small resist lines. Heck, i printed some 30 nm lines at 90 nm pitch last week. This does not translate into working devices at those densities for a long time.

(edit: DOF goes down, not up. that would make it too easy...)