Semiconductor Technologies Guide

An anonymous reader writes "X-bit labs have posted an interesting article on manufacturing technologies used in the semiconductor industry. Good reading if you want to get a really indepth idea of technologies used for semiconductor manufacturing by IBM, Intel, AMD, and others."

Comments

Nano-tube technology and it's application:

[antispamist]

Can carbon nanotube be used in CPU production? From what I understand some are conductive with low radiation and others are excellent insulators.

The article talks about "spontaneous electron movement from the negatively charged silicon substrate of the channel to the positively charged gate."

I guess I am just curious as I recently wrote a paper on their applications and I would like to hear from someone a little more technically knowledgable than me. Anyone have any real knowledge or some *easy* links they could share?

Re:Nano-tube technology and it's application:

[Wakkow]

I'm not sure I can answer your question, but I want to clarify something. I found this on google: diagram of a mosfet transistor.

Here's a simplified explanation. Think of a switch with source at one terminal and drain on the other. When sitting without a voltage on the gate, the source and drain are not connected. When the switch is turned on (ie. gate high), electrons are allowed through the pathway created.

Anyways, the yellow in the diagram is an insulator. The switching is all done without touching the doped silicon connecting the source and the drain. My point is that the silicon needs to be there. It's integral to how the switching works. I don't know anything about nano-tubes but it cannot replace the silicon unless it can act like a semiconductor (both as a conductor and insulator depending on temp, etc). Perhaps it could replace the SiO2 currently used as the insulating layer but no matter what, the smaller the channels get, the more the electrons are going to want to jump..

Anyone that knows more about SOI want to comment?

Re:Nano-tube technology and it's application:

[Anonymous Coward]

They aren't currently used in CPU production, but IBM has developed a "Carbon NanoTube Field Effect Transistor" (CNTFET) which they seem to think will replace silicon. Look up "carbon nanotube FET" on google for links.

Re:Nano-tube technology and it's application:

[Anonymous Coward]

Carbon nanotubes are still very much in the experimental phase. While there are a lot of uses in which they may be useful, like as interconnects on ICs, the technology for mass manufacturing of nanotubes still does not exist.

I believe they can only be grown at the moment on AFM tips, or at least that is the only way they can really test them at the moment.

Re:Nano-tube technology and it's application:

[brarrr]

Well, the idea of using carbon nanotubes as transistors uses a completely different transistor model than a mosfet. The conductance of CNTs come about due to their orientation (imagine a plane of graphite rolled up - there are many ways to make the ends meet, akin to the steepness of a circular staircase). A few orientations are conducting, the rest are insulating.

Semiconducting CNTs occur when you put a twist on the CNT - i don't know the mechanism for that nor the theory behind it, but it is unrelated to the traditional transistor.

SOI is silicon on insulator, which is just an advanced technique used in transistor design - not necessary for the fundamentals of mosfets.

Re:Nano-tube technology and it's application:

[Oo.et.oO]

there are thousands of different types of transistors, most of which do not use silicon as the semiconductor, most of which do not resemble a MOSFET, or even a FET at all.

silicon, and even electrons, are not critical to the operation of all transistors.

Re:Nano-tube technology and it's application:

[Cougar1]

Carbon nanotubes can be mass produced. The problem is that modern production techniques are not selective, so the result is a jumble of nanotubes, some of which possess the desired electrical properties and many of which do not. The challenge then becomes removing the undesirable nanotubes and then organizing the remaining nanotubes into electrical circuits. Researchers are making great strides in removing the undesired nanotubes, but organizing those that remain is much more difficult.

On a small scale this can be done using AFM tips and I believe this is how IBM and others have demonstrated working nanotube-based devices. However, such a technique is impractical for large-scale production, where you need to interconnect millions of devices and significant work still remains to reach this goal.

woah

[ergonal]

Wow, I'd like to say sorry to whoever's member is shown on the last page! That's TINY man!

Re:woah

[CanadaDave]

Yeah I noticed that too. It's crazy. Just wait until it gets turned on... and the current will start to flow from source to drain. Just don't turn it on too high, or it will blow.

Well...

[Anonymous Coward]

If anyone comments on this within 10 minutes, they didn't read the article ;-)

Re:Well...

[Anonymous Coward]

Speak for yourself...slow reader that you are.

check out the comparison b/c transistor and virus

[lingqi]

between the transistor and an influenza (closely related to SARS) virus, no less.

I don't think it shows the smallness of the transistor as much as I suddenly realized how much further we have to go before hitting biological complexity.

the surface of the virus has crazy number of protein receptors that allows it to latch onto only the proper cells, and inside a strand of genetic material that contains thousands, if not millions of ACGT pairs - which puts information density of our most hardcore RAM at a great shame. Actually there are probably other stuff inside, but IANAVirologist.

Looooong road ahead...

side note: I don't think the gearheads are so obsessed about the manufacturing process for cars, nor the martha-stuart followers the manufacturing process for triple flower-pattern guest-only bath towels, why are geeks sooooo into the photolithography process?

Anybody wants to offer an explanation?

is this guy a moron?

[Anonymous Coward]

whoever this guy is, he does not know what he's talking about. if you want real information, look at

http://britneyspears.ac/lasers.html

The XBIT labs starts off wrong, and just gets worse. for example, his descriptions of leakage current are all wrong. but, what do we care, we're slashdotters and we don't know our ass from our brain.

Re:check out the comparison b/c transistor and vir

[Anonymous Coward]

and inside a strand of genetic material that contains thousands, if not millions of ACGT pairs

Actually it's 1501 base pairs (so 3002 bits of information) in one strain of the Influenza B virus.

Re:check out the comparison b/c transistor and vir

[Anonymous Coward]

Sorry, that was for just one segment, have a browse through NCBI for the other segments to find out the total size (go on, it's fun :)

Re:On the subject

[Timesprout]

Gene stop trying to plug your book

Re:check out the comparison b/c transistor and vir

[fishbert42]

"... ACGT pairs - which puts information density of our most hardcore RAM at a great shame."

Perhaps; but we can write and rewrite to our "most hardcore RAM" much faster than genetic information in ACGT pairs can mutate.

Re:check out the comparison b/c transistor and vir

[alannon]

My guess is that no other industry in the world has the pace of tangible progress as microprocessors (except perhaps magnetic storage). Look back 30 years. Microprocessors were still in their infancy. The product of microprocessors have gotten 10's of thousands of times more powerful. Automobile manufacturing was, however, much as it still is today. The vast majority of textiles manufacturing is similarly unchanged in the last 30 years. History has demonstrated, though, that any fundamentally new technology goes through a very rapid period of development in a relatively short period of time. Eventually, all technologies level off in their pace of development.

etching

[romit_icarus]

I used to be part of the surface physics research community that looked into how to etch more finer than we currently do so. Saw this so I thought I'd add the little I learnt there.

Conventional etching uses plasma. Plasma is a soup of charged particles of energy of the order of a kev (1000 electron volts). It's use is mostly dictated by the fact that we understand how to create it quite well, rather than in our understanding of what is/are the consituents of plasma that actually effect the etching action.

What is found that the etching process sensitive to the charge element, to the quantum state of the incident ions, to its energy, to the angle of incidence. And of course this is purely from the beam side, from the surface point of view there are a lot more variables...

Did you know that the molecular process that does etching is very similar to that that creates a radio blackout in the space shuttle while reentry? Somewhat cool

Re:etching

[brarrr]

what you're talking about is actually called sputtering, in that high energy particles are used to remove surface attoms. it is not the commonly used definition, but nevertheless, it is the correct definition for sputtering. you're more likely to hear of sputtering as a surface deposition process.

etching for silicon is more likely to done with a chemical etch (pirhanna sol'n, HF, etc.) than a physical one.

Re:etching

[Bender_]

etching for silicon is more likely to done with a chemical etch (pirhanna sol'n, HF, etc.) than a physical one.

Boy, thats 70ies stuff. Today, wet processing is avoided as much as it is possible. But you are right in one point, purely physical etching is not used frequently. However there are combined physical/chemical methods. Do a websearch on Reactive Ion Etching, Plasma Etching etc.

Just 50 comments in 2 hours?

[flokemon]

Don't tell me /.ers are busy reading the article...

General semiconductor physics information

[pellik]

Seriously, this is one of the best semiconductor physics sites I've ever seen. Informative too.

http://www.britneyspears.ac/lasers.htm

In case it gets slashdotted..

[Anonymous Coward]

In this article we take a close look at the contemporary manufacturing technologies used in the semiconductor industry. We tried to raise the curtain of mystery over the transistors and other chips manufacturing process, and to evaluate the prospects of the current and upcoming technological processes.

- Why can't a man fly like a bird?
- The manufacturing process doesn't allow.

Why do car engines consume liters of gasoline in a hundred miles, but not just grams? And why gasoline, but not water? Why is there no portable cool fusion device in every household? And, at last, why do processors consist of just millions of transistors working at a few gigahertz frequency? The answer is very simple: the today's technological process, or the technological capabilities of the manufacturer, don't allow anything beyond that yet.

However, we shouldn't confuse the actual technical capabilities with potential ones. If Intel wanted to produce 65nm processors, they could make them even today, but the production volume would be measured in pieces, or dozens of pieces. And these processors would also cost much more. As we are all living in a big marketplace, we can put it down in the following way: the technological process is a compromise between the theoretically possible level of some technology and the price acceptable for a mass consumer of products made with this technology.

Of course, we see this compromise in the PC CPU market too, although there is some misbalance towards performance prevailing over the price, which is mostly caused by vast and successful marketing campaigns CPU manufacturers have been carrying out. Thus, Intel has been putting about $300 million or even more into promotion of each of its latest processors (Pentium 4, Pentium-M).

But there are sales plans and marketing cycles, there is Moore's law, at last (see our article about Moore's Law for more details). Various economic and marketing reasons, and Moore's law in particular, are a kind of locomotive power for the whole processor industry. By the way, this law, which was originally just a simple observation, but later was puffed up by Intel a little, works now mostly in Intel's favor, because it is most profitable for those who have the highest technological and manufacturing potential.

Besides, the fact that this law was fulfilled so easily didn't do much good to the industry. Engineers didn't have to worry about effective functioning of various processor units: why bother if you could allot as mush transistors as you wanted for every unit. Or at least close to that. Now it has become clear that Moore's law is going to die out one day and such effects as leakage current threaten further development, so, they have finally started talking about a more thorough approach to processor architectures development. This is a very pleasing fact.

Of course, if we would like to talk about the manufacturing technology process used for PC CPU production, we should start with Intel. The company has the biggest production facilities today (25 fabs, at least ten of them make CPUs) and one of the most advanced production processes.

Re:fp!

[Anonymous Coward]

is gluthethimide available over the counter? where do you get it?

Re:This article officially owned by trolls

[CausticWindow]

What's the official troll ownage ratio? 2/3?

Re:Dude, this is slashdot

[Anonymous Coward]

Rikki don't lose that number!

Easy question

[Control Group]

Because geeks are attracted to that which is complex, and have a fundamental desire to understand the "how" of things. As a geek who also works on cars, I am interested in the manufacturing process for cars - but as a poster below noted, that doesn't change anywhere near as fast. Once I learned it, I stopped being as interested in it.

Besides, photolithography is hella cool.

Re:check out the comparison b/c transistor and vir

[brarrr]

There is an industry with much higher rate of advancement. Since the introduction of disposable diapers, their ability to absorb effectively has outpaced the shrinkage rate for effective processors. Thats one of the little 'facts' we bring up in our intro to engineering classes in materials engineering. Weird, but true.

M of N gates

[bestguruever]

My knowledge in this field could probably fit in one of these new transitor channels, but I was wondering:

Could the Tri-Gate transitor be a major boon to producing M of N gates with hysterisis? The asynchonous logic proponents would love this. I'm thinking you would might need more layers of interconnects.

Anybody ideas?

holy poorly written batman!

[Oo.et.oO]

this is THE most poorly written article i've actually tried to read on the web in years.

"Besides, the characteristics of the channel become more predictable, while the transistor itself - more robust to various errors, like those provoked by space particles that may get into the channel and ionize it."

i'm not even going to get into the english, which quite frankly is horrid. "robust to errors"? "space particles" (yes i know what he's trying to say)? "ionize it"? ionize the channel? (yes again, i understand he means the carriers, but that's not what happens)

most of his information is unsubstantiated at best. There are no references nor citations. Most of it seems somewhat accurate but only because i understand what they were TRYING to say. If i didn't know about fab tech to begin with i would have been very misled.

If you want to know about semiconductor fabrication technology, do yourself a favor and borrow a textbook from your EE budies, then read up in journals.

Re:holy poorly written batman!

[Anonymous Coward]

THis is troll territory, beat it, mr. insightful!!

Re:holy poorly written batman!

[Cougar1]

this is THE most poorly written article i've actually tried to read on the web in years.

I take it you don't read many articles on the web... :D

Anyway I agree the article is not really well written and the English is horrible, but it does give a nice, though very simplified, overview of some of the key problems of semiconductor processing and what various companies are doing to overcome them.

Being completely technically accurate for such an article is quite difficult. For example, when discussing high-k dielectrics, the article states that, "This material should be 10,000 more effective in preventing electron leakage from the channel to the gate than SiO2. If you have been reading attentively, you should realize the importance of this: the thickness of the nonconductive layer may now be reduced to tenths of nanometer keeping an acceptable gate leakage value."

The above contains many errors. For example the quantum mechanical tunneling through a layer (any material) a few tenths of a nanometer thick is enormous, so it can't be true that the high-k layer is thinner than SiO2. The reality is that high-k means the material has a higher dielectric constant than SiO2. This means that electric fields are transmitted more readily through the high-k material and so a thick high-k film exhibits the same electrical behavior as a much thinner SiO2 film.

Thus, the reality is that replacing SiO2 with a high-k material allows the use of a much thicker film, which behaves electrically as if it were thinner than the corresponding SiO2 film. Quantum mechanical tunneling of electrons through the gate dielectric drops exponentially with increasing thickness, so the leakage current is reduced dramatically through the use of the thicker film.

uckling

[Anonymous Coward]

FUCing shit CUnt face fucker someoth gbithc fucker shit!

haha

[Anonymous Coward]

poopy head anonymous coward person!! xbit labs is just ONE company. so it'd be "has posted" not "have posted"