Rolled sheeting for next generation of processors

[MadRat]

Materials can be cast into almost any shape or form these days. I think that it should be possible to coat some sort of sheeting with purified silicon or some other insulation in order to make processors. This would eliminate the need to cast purified silicon cyclinders for the creation of microprocessors. Cutting the material into a thin wafer is eliminated. Trying to come up with ever increasingly complicated ways of cooling wafers become simplified, as thin sheeting can literally be pasted onto a perfectly smooth thermally-conductive surface. I wouldn't see why existing techniques to attach interconnects on wafers to the cpu package couldn't also be used to attach the interconnects from sheeting down to the packaging.

I personally think that the sheeting would be the best way to pursue this idea. It would have to be some sort of thin but sturdy, inert material that has a high tensile strength so that it cannot easily be stretched. (Mylar plastic is but one material that quickly comes to mind.) The idea behind using sheeting is that it can be created very thin, the need to be only as thick as necessary for manipulating the material. Etching all sorts of odd shapes of design, hardening the material from EMP effects, and using stacks of the sheeting then become just a few of the possibilities. I wouldn't doubt if current machinery used in etching silicon couldn't be used with materials made from this yet to be discovered sheeting.

There is always the possibility that the sheeting could be attached to the packaging prior to the etching process, too. It looks like AMD and Intel have pretty well moved to using PCB in the packaging process. Large sheets of PCB, complete with the pins, could be made first. (This is probably how they do it now anyhow.) The cores, in sheet form, could be used to solder down to cores to the PCB array. The PCB/unfinished core combination would then be run through the etching process. Once the process is complete then the individual processors would be cut from the PCB. Rolls of PCB and core materials could be sent through one long integrated process without having to stop and manipulate individual wafer slices.

How would one go about researching the electro-chemical properties necessary for good processor manufacturing in order to find a material suitable for use as sheeting in processors?

 

[Eskimo]

Interesting idea, a couple problems with it though.

To have the necessary mobility and doping characteristics of a good semiconductor it is necessary for the silicon to be crystalline in nature. There is no known method of deposition of silicon in a purely crystalline form without a seed layer present like that in epitaxial growth. If you just try to deposit silicon you will end up with polycrystalline or amorphous silicon depending on pressure and temperature. Even if you could achieve a similar lattice structure to thermally formed silicon you would have a large number of crystal point defects due to film stress during deposition.

Assuming you got around the first problem you would have to take care to ensure your sheeting material closely matched silicon in terms of thermal coefficient of expansion. There are numerous high temperature steps involved in IC production with many involving very high ramp rates. If your TEC doesn't match you are going to have delamination of your silicon film. Also can your film handle high temperatures up to 1100C without curling? It is critical for cross wafer/'sheet' uniformity to have near perfect flatness. Some tools chuck a wafer on the edge with no backside support, wafers are as thick as they are to combat graviatational sag and provide rigidity in such cases, would your material be able to do this? During high temperature low pressure diffusion steps what is the phase diagram of your material? Will it outgass or will atoms diffuse through into the silicon?

Is your material going to be a conductor? insulator? semiconductor? You'll have to be aware of issues involving charging in plasma systems where charge could build up on your 'sheet' and lead to damage of the gate oxide. If we are talking about plasma etch systems what will your sheet material's selectivity be to the chemistry used to etch silicon, oxide or nitride (CF4, C2F6, HBr, O2, SF6 and so on...). What about chemical resistance to wet cleans involving but not limited to HF, Hot Phosphoric, Pirhana, HCl, NH3, TMAH, Acetone, and any of the other numerous chemicals used.

Those are just some of the challenges that come to me off the top of my head. Good luck with your research.

 

[MadRat]

Okay, just thinking out of the box here.

If the material is inert then it would have to be similar to silicon, but doesn't need to be silicon. Assume that the material will not detach from the PCB if pre-attached. (i.e. curling or delamination) Unfortunately I'd almost have to guess then that the PCB would be damaged by 1100 degree temperatures.

If the material is pre-attached to the PCB will this make it impossible to etch? I'm also going to have to assume the PCB under the core would be free of metal that would absorb electromagnetic fields. (I assume you hinted at the possibility that electromagnetic fields in the etching process may cause a buildup in the layer that would damage the gates.) Obviously the etching would have to be performed on the top of the sheeting layer versus underneath the core like on flipchips.

Will it outgass or will atoms diffuse through into the silicon?

I assume the material is either going to need to resist absorbtion or its probably not going to work.

I'm not familiar with HF, Hot Phosphoric, Pirhana. I'll have to look this stuff up.

 

[dejitaru]

Some companies are working on that now.

 

[Eskimo]

Originally posted by: MadRat
Okay, just thinking out of the box here.

Just trying to help you out with some issues you might face, not trying to slam your idea.

I'm not familiar with HF, Hot Phosphoric, Pirhana. I'll have to look this stuff up.

HF is hydrofluoric acid, in addition to about being the nastiest stuff you'll ever want to encounter even at the 10-50% concentration level. It's used in relatively large volumes in wet clean environments since it etches silicon dioxide very very fast compared to silicon (1000:1 or thereabouts depending on concentration). Hot Phosporic Acid is comprised H3PO4 at approximately 150-200C and is used for the removal of silicon nitride. Piranha (mispelled it earlier) is a mixture of sulfuric acide and hydrogen peroxide used for the removal of organic contaminants.

 

[MadRat]

Awww, HF is something I do understand. I'm was sheet metal worker once.

You guys use alot of chemicals that work very good as polishers.

 

[f95toli]

HF is really nasty. If you get some on your skin it will just "diffuse" through the skin and flesh, you will not notice anything until it starts eating away your skeleton (I am not kidding, this is the way it was exaplained to me in a safety-course).

Back to the topic: There are a few ways to deposit crystaline sillicon on sheets, but they are all very expensive. Basically what you do is that you use a "template", another material with a structure very similar to silicon (it also need to expand in the same way when heated). Some of these "templates" can be grown on sheets.There are a few oxides that work (strontium-titanate with a buffer layer of cerium-oxide fo example) but the problem is that they are very expensive.
A silicon-wafer costs next to nothing compared to the other materials. Silicon is really cheap, it is the processing that is expensive; not the raw material.