IS this the end of the use of photolithography in the manufacture of next gen chips?

[klaviernista]

just went to a conference where this nano-imprinting technique was presented. Other people in this forum have posted articles relating to the focusing of light to a point smaller than its wavelength, which allows the formation og 60nm lines. The tools for doing this are extremely expensive (~$ 15million a tool or more). This nanoimprinting process can make lines as small as 20nm for ~$1 million a tool. The technique still needs work, but if perfected it might signify the end of photolithography as the tool that will make next generation circuits/cpu's. Check out the technology at:

http://www.molecularimprints.com

 

[f95toli]

As far as I know the technique is somewhat limited because of difficultites with the "masks" (or rather templates). So right now it can not replace photolithography and I don't think it ever will.

We actually have a nanoimprinter in the university cleanroom where I work sometimes, it is rarely used because of the limitations (it is a couple of years old), but it is nice for things like nanoparticles.

You don't happen to remember what kind of material they were using to make the template?

I wonder what the price of a template with 20 nk details would be (if it can be done, I doub't it). it would probably take a week to draw a full waffer using an e-beam...

 

[klaviernista]

I do not recall the actual photopolymer they are using to form the imprinted surface, but you are correct in stating that the main issue is with the template or stamper. The stampers are made via e-beam lithography, and so all of the error associated with the e-beam process are transferred to the stamper. e-beam lithography has difficulty forming sub 40nm features. This is a big problem with the stamping process because even slight errors in the stamper will generate errors from field to field and will result in large increases in signal loss in CMOS devices. At the conference I attended Dr. Sreenivasan (who apparently co-invented this particular process) was touting that the process could consistently reproduce features that were 40-60nm. He attributed this to the fact that the e-beam process could create a much more accurate stamper at this feature size. In addition, when the feature size gets this small, depending on the type of feature being made, the stamper is easily damaged. This is particularly true with stampers that are used to make nanometer sized pits (which could be used as is for next gen optical media or filled with magnetic material for next gen magnetic media), as the stamper has small projections that are only nm wide and nm high projecting from its surface.

Regardless, most commerical processes now are ~.13 microns (130nm), and the industry is driving towards 60nm with the light filtering photolithography technique I mentioned in my previous post. Even if this process could only make consistent 60nm features it would still be big advance.

 

[f95toli]

I remember the name of the company that made our nano-imprinter
Obducat

It is this model http://www.obducat.com/sida_78.asp

It has been up and running for a couple of years.

 

[klaviernista]

Yeah, the obducat systems operate non a substantially different premise then the moleculr imprints machines. The obducat machines heat high tg polymers to just above their melting point and then compress the stamper into the melt. The melt is then cooled below its Tg, where it solidifies, and the stamper is removed, reuslting in the stamped article. This process has been shown to have major problems with stamper damage.

The molecular imprints design however utilizes no heat and almost zero pressure. The surface of the substrate is first sprayed with picoliter droplets of photopolymer.Almost immediately after the surface is sprayed, the stamper is very lightly pressed into the droplets. By capillary action, the photopolymer conforms to the surface of the stamper. The photopolyer is then cured via incident irradiation through the stamper (which is transparent). The stamper is then removed. The stamper is also coated with a monolayer of fluorpolymer to prevent the photopolymenr from sticking to the stamper during the removal step.

 

[Smilin]

Yeah guys I'm just not getting how they are doing the Stamp/template.

It's like the underwear gnomes from southpark. Step one: collect underwear, Step three: profit!

I think Step two (somehow make a Stamp/template from underwear) is the difficult part.

 

[pm]

These things always take longer to get to market than one might think that they would. CMP was adopted fairly quickly as these things go and that took quite a while for full industry acceptance. Immersion litho is next and will probably extend 193nm litho to 65nm and maybe down to 40nm or so depending on adjustments to the NA using various (not yet determined) fluids. That's 2009 or later before something else could step in.

 

[klaviernista]

Originally posted by: Smilin
Yeah guys I'm just not getting how they are doing the Stamp/template.

Basically, the stamper is made by conventional electron beam lithography. You can learn about that tech here: http://dot.che.gatech.edu/henderson/introduction_to_electron_beam_lithography.htm.

The other thing about the molecular imprints process is that the master is much smaller then the size of a standard wafer. I think the speaker said they could get 4 useable masters from a single standard e-beam processed wafer. This is an advantage in the sense that it reduces the cost of each master. ITs a disadvantage in that the ratio of the size of the feature on the master and those imprinted on the substrate is ~1:1. Compare this to the current photolithogray techniques, where the features on the photomask are roughly 4 times the size of the features formed on the substrate (light passes through the photomask and is then focused down to the correct feature size. Thus, the masters for this imprinting process are much more difficult to manufacture than currently used photomasks, as the features have to be extemely precise but 1/4 the size.

 

[CanOWorms]

Don't people always say photolithography is going to end soon, but it ends up being extended? I thought liquid immersion was the next thing that's supposed to extend the lifetime.

 

[0]

Way back when they said e-beam litho was going to supplant regular litho. That never happened.

Biosensors use self-assembling techniques, perhaps that is the wave of the future?

 

[klaviernista]

Yeah, but in many ways e-beam litho has supplanted regualr litho, particularly when features that are smaller then those that are currently attainable by conventional litho are required.

Also, to answer your question canowaorms, yes, liquid immersion is going to be the next gen. of photolithography. HOwever, the extension of photolith likely will not go on forever because the tools almost exponentially increase in cost as the feature size is reduced. Currently, a single photolith tool for IC chips runs about 15million, with next generation tools estimated at ~60million. Considering a fab would have at least 10-20 of these tools, thats 600 million-1.2 billion just for the tools (and doesn't include the infrastrucutre of the rest fo the fab). Those tools are striving to achieve 60nm features.

Camparatively, the molecular imprints tool costs ~1-2 million per tool, and can make accurate features field to field as small as 40nm. Thats smaller features for a fraction of the price. BUsiness sense indicates that this is a vaibel alternative. The only problem is is that industry is entrenched in photolith technology right now.

 

[f95toli]

E-beam lithograohy is used a lot in research and development.
You can use it in direct-write mode to make very small lines (about 20 nm depending on the material). But mainly the e-beam is uised to make masks for photolithography,

However, an e-beam is way too slow for producttion, even very simple circuits can take many hours to write,

 

[Eskimo]

Originally posted by: klaviernista
Currently, a single photolith tool for IC chips runs about 15million, with next generation tools estimated at ~60million. Considering a fab would have at least 10-20 of these tools, thats 600 million-1.2 billion just for the tools (and doesn't include the infrastrucutre of the rest fo the fab). Those tools are striving to achieve 60nm features.

Your statement about the rising cost of photo tools is correct and increasingly forcing companies to maximize the throughput and uptime of existing tools. Your prices however are not entirely correct, while those may be the 'sticker' prices the litho vendors quote we don't pay nearly that much, one advantage of buying in volume i suppose. The reason that immersion lithography is so attractive to the industry is the relatively low cost of change. Except for the fluid delivery and immersion/shower component there is not much new in the way of new materials or methods being introduced. Since the fluid currently under consideration is DI water that further decreases the amount of learning necessary for adoption. Immersion also allows for existing platforms to be used by the litho vendors which reduces their R&D costs and ultimately our system costs. It's hard to bring about radical change in this industry for any process. We spend enough time and resources sustaining established processes that new processes aren't desired unless absolutely necessary to continue device scaling/performance.

I know that Motorola is a big supporter of the nano-imprint since they invested a considerable sum in the start up of Molecular Imprints. The rest of the industry will likely take a wait and see approach until someone like Moto can show the integration of nano imprint technology into an actual production line. Personally I haven't done a lot of research or had a lot of experience with the technology but I have heard reservations from people regarding the defects introduced from the imprint process. I would imagine as you remove that stamp from the photopolymer you are liberating particles during the seperation and some residue would remain on the stamper to be redeposited on another part of the wafer. It is a promising technology that might find applications, should be interesting to watch.

*Not speaking for any company

 

[rimshaker]

Not quite just yet. Photolithography fortunately just got a simple major modification where using a fluid medium between the wafer and light source can easily handle smaller resolutions for the next few nodes.

 

[klaviernista]

good post eskimo.

your correct in that motorol invested several million dollars into Molecular Imprints, and your logic regarding residue on the stamper is well founded. This same question was asked at the conference I attended and apparently has been dealt with in an interesting manner. According to the speaker, a monolayer of a fluoropolymer similar to teflon is vapor deposited on the surface of the master to prevent this problem from occurring. Apparently they are having much less trouble with residue on the stamper then you would think. The main problem they've been having in terms of errors have been noticed at sub 40nm feature sizes. These errors are attributed to the error in the e-beam process used to make the masters, not the imprinting process itself, though I'm sure that at that feature size its probably a little of both.

As for the price of tools, I'm sure you are correct. Nonetheless the tools are much more expensive than this imprinting tool. HOwever, you are again right in that radical processing changes will be strongly resisted by any major manufacturer as a result of the large R&D cost required. Heck, radical change in any established process takes forever.

 

[f95toli]

I saw a paper today in Applied Physics Letters (the June 28th issue), it was written by a group from Princeton.
5 nm linewidth and 14 nm pitch in the resist...Not bad..

 

[klaviernista]

Yep, saw that article as well. Preety cool stuff. The princeton tech is based on the same premise as that marketed by molecular imprints. The third reference cited was authored by Sreevinasen (and others), and he is one of the founders of molecular imprints).

Chances are the molecular imprints guys and the princeton guys know each other.

Oh, those interested in the article can access it here (you might need a subscription though):

http://scitation.aip.org/getpdf/servlet/GetPDFServlet?filetype=pdf&id=APPLAB000084000026005299000001&idtype=cvips

 

[feitianren]

To f95toli,
we are going to install the same nanoimprinter as your's from obducat. Is it easy to operate and maintain? Do you have a electronic copy of the system mannul? I want to read it before the installation, will you please share it with me? Thank you