IBM Recycles Waste CPU Wafers Into Solar Panels

Luyseyal writes "IBM has developed a process for scrubbing waste silicon wafers clean, allowing the otherwise highly secret waste to be sold. The silicon quality usually necessary for solar production is very high and the cost of solar panels reflects it. Recycling this waste should help bring down the cost in the long run and add a new profit vector for chip manufacturers. The article notes that IBM has such a high profile in the chip business that this recycling tech should spread rapidly."

I'll wait for AMD to do this

[by Anonymous Coward]

Their scrap wafers turned into solar power should generate more power at a fraction of the cost.

Consider the power consumption in another way

[FoxconnGuy (997669)]

I read on a book about a Germany based solar plate vendor produces their solar plates using the power generated from the solar plates they produced and installed outside their building. This is not a perfect solution. But better then totally rely on power generated from other not-green sources.

Re:

[jx100 (453615)]

Yeah, but then you can't go to Europa.

Silicon wafers are not the answer for longterm PV

[EmbeddedJanitor (597831)]

While most PV is currently constructed from wafer silicon, this is not a viable long-term strategy because it takes so much energy to make a wafer. To make real progress, PV needs to move to alternative technologies.

How Much?

[Doc Ruby (173196)]

How much energy does it take to make a PV wafer, per usable area? Not including what it takes to deliver and install (and later recycle) it?

Re:How Much?

[EmbeddedJanitor (597831)]

To give you an idea... First you have to melt sand. Not cheese on a pizza, but sand.... then keep in melted while building up the silicon boule which takes a good long time. Then you cut it into wafers and a lot of the material gets lost in the kerf. Then there is doping where the wafers need to be kept at very high temperatures for many hours while the dopants get absorbed into the wafers...

That's a lot of heating that needs to be done very cleanly so uses electrical power which is far more wasteful than trying to get the same heat from a primary source (gas/oil etc).

No wonder PV has such long energy payback times and costs so much.

To get energy input (and thus $/watt too) to practical levels requires a change from wafer-based technology.

Re:

[Doc Ruby (173196)]

I understand the process. But what is the actual quantity of energy required to make a square meter of PV? How many joules to produce a 15%, or for a 20%, or a 23%, or even the new (not just silicon) 42% PVs? That makes discussions of the PV energy budget actually tractable, instead of just speculations about whether the breakeven is met.

And how much is saved by using these IBM "scrubbed chips" instead of starting from scratch, for what %efficiency?

Re:

[EmbeddedJanitor (597831)]

I don't think there are any easy answers due to different processes required to make different grade silicon and for other differences too (strenght, temperature etc). The crap they put into solar calculators etc could be made on a barbeque at home (well almost). Different doping levels require different amounts of baking and highly uniform doping requires doing it slowly.

Waste wafers get you past the boule stage. You'd need to redope them though.

Re:

[dbIII (701233)]

instead of just speculations about whether the breakeven is met.

The breakeven thing is a bit of a distraction since it is no longer 1960 - the time to break even would vary depending on the process and lattitude where you use the things. I don't have any numbers on this, it needs a bit more than what is in the textbooks but in short the wide use of semiconductors resulted in it being worth making even small improvements to save a lot of energy and money.

Prior Art

[mdsolar (1045926)]

Hope IBM is not planning on patenting their method. This kind of thing has been studied already to understand the energy savings from recycling solar cells. Recycling solar cells requires about one third the energy of making new cells: http://www.solarworld.de/solarmaterial/english/press/8AV.3.14.pdf [solarworld.de]. And, basically, you scrape off what was on the waffer before and then start again. Note that in the link, they assume about 2.7 peak equivilent sun hours per day. A typical value for the US is 5 so that the energy payback time would be about 2 years for a new panel and 8 months for a recycled panel. For 40 years of use you get EROEIs of 20 and 60 for new and recycled respectively. But, you have to wait 40 years to start getting the cheaper deal ;-)
--
Rent solar and save: http://mdsolar.blogspot.com/2007/01/slashdot-users-selling-solar.html [blogspot.com]

Re:How Much?

[sholden (12227)]

First you have to melt sand. Not cheese on a pizza, but sand

God damn it! No wonder my attempts have never worked. You have no idea how many different types of cheese I have tried...

Re:

[rgravina (520410)]

Sort of like bootstrapping solar power!

Re:How Much?

[Ogemaniac (841129)]

It actually takes an enormous amount of energy to make solar (or IC) grade silicon. The estimates I have seen calculate that about 20% of the total energy produced by a typical crystalline silicon solar panel is necessary to construct and install the cell. Roughly half that energy is embedded in the silicon itself.

I disagree with the parent's parent post. There is no reason that silicon cells are not viable renewable energy sources. They produce five units of energy over the long haul for every one put in (excluding sunlight, of course!) - and that one could be renewable itself.

Silicon for IC and solar is so expensive and energy intensive because it must be so pure. To produce it, SiO2 (quartz, sand, etc) first reduced with carbon (similar to how iron oxide is made into iron). This requires lots of energy. This product, however, is crude. To purify it, it must be gassified to various chlorosilane molecules and then distilled (lots of energy in both steps). The highly pure gas species are again reduced to silicon metal and then recrystallized carefully to eliminate even more impurities...again, energy intensive. In most cases, these steps are undertaken at different facilities or companies, requiring shipment at each step as well.

Re:

[Average (648)]

One factor that may make a difference is that power at point A is not equivalent to power at point B. The places where power is ultra-cheap tend to export it in the form of manufactured products (i.e., the Columbia River historically being a cheap place to refine Aluminium lead Boeing in Seattle). If energy were somewhat more expensive, something that is net Energy-Return-on-Energy-Invested positive (making solar cells) will be even more net-positive in a cheap energy climate.

Transport cost for silicon

[mdsolar (1045926)]

You presume to much. The transportation cost for moving coal to a power plant is 200 times higher for the amount of energy produced than the transportation cost of taking a solar panel from a factory to your roof. Solar beat nuclear in this measure as well if you consider that uranium is mined in Austraia and enriched in France. Of course it makes sense to make panels where there is hydro power. But, there is hydro power in a lot of places so why the globalization jab? Shipping solar panels a long way

Re:

[Doc Ruby (173196)]

I understand the process. But what is the actual quantity of energy required to make a square meter of PV? How many joules to produce a 15%, or for a 20%, or a 23%, or even the new (not just silicon) 42% PVs?

And how much is saved by using these IBM "scrubbed chips" instead of starting from scratch, for what %efficiency?

You say about 20% of the energy the PV will produce is consumed in construction and installation - 10% in manufacturing the silicon. A square meter of PV will last maybe 30 years, getting may

Re:

[Ogemaniac (841129)]

25Gj for a panel comes out to a little under 2000 Kj/mol of silicon. That is a totally plausible number given the number of steps and high temperatures required (above 1000C in some phases). The other 10% is not just installation, but also includes wires, back-panels, sealents, etc.

Putting things into space is enormously energy intensive. You would never come out ahead unless you built a space elevator first. Unfortunately, no material known to man is strong enough to build one, not even in theory (c

Re:How Much?

[Doc Ruby (173196)]

Another way to calculate it is about 777.6Kj:Kg [slashdot.org], which is 18.624Gj for the 8" wafer, in the range of what we ran down.

I left out the only 20% efficiency solar -> DC conversion factor, so the cells I described produce only about 50Gj in their lifetime, or 37% total energy inefficiency from manufacturing. Seems like a lot.

I'm not sure we'd have to put the silicon into space. I saw reports of a NASA demo a few years ago of a lunar robot making solar cells from lunar dust. There's about 20 trillion square meters of Moon facing the Sun at any time, getting about 1.3KW:m^2, or 26 petawatts. Even at 1% conversion/transmission/conversion efficiency, that's 260TW, or 17x total human energy consumption. Which means well under 6%, perhaps even 0.6%, of the Moon's surface would replace all Earth power generation. Of course, orbiting solar platforms could offer even larger energy return. And consider the amount of energy wasted on war and fuel distribution that could be saved. If the space "factories" are productive enough, the energy budget balances well in favor of doing it.

Re:

[dbIII (701233)]

First I'll say that it is very true that it is a lot of energy but the part below I have to dispute and ask where it came from:

The estimates I have seen calculate that about 20% of the total energy produced by a typical crystalline silicon solar panel is necessary to construct and install the cell

What year was this written? What estimate did they have for the lifetime of the cell? Is it possible that it was written longer ago than their estimate for the lifetime of a cell and that improvements in the dec

Re:

[Ogemaniac (841129)]

I heard it from someone in my own company and have heard similar numbers before. Since we (actually our subsidiary) make the silicon, I doubt we are skewing the data in favor of the nuclear industry. Instead, the point was that we have to get our energy costs down and figure out a better way to make this stuff. Indeed we have, but we are still the biggest consumer of electricity in the state now due to our efforts to make silicon for the solar industry. How ironic.

Re:How Much?

[dbIII (701233)]

It really depends on how big the ingot is that the wafers are cut from and how well the factory is set up. It used to be a very slow energy intensive operation - you are growing very big single crystals of pure silicon after all in a process known as "zone refining". There were a lot of improvements in the 1970s to change the size of the zone melted, some changes to reduce the amount of this that has to be done and doing things in large volumes (big ingots with big wafers) brings the energy use per square millimetre of silicon components right down. There are still people that look at the numbers from the first solar cells produced for the early space program by experimental methods and assume (or pretend as an arguement ploy) the costs are the same today despite the very wide use of semiconductors and the improvements and economies of scale.

As for recyling - it would be a matter of grinding the top off by whatever method is easiest (eg. Silicon carbide grinding and finer particles of the same to polish) to give you a single silicon crystal to turn into whatever you want it to be. In a lab progressively finer grades of normal sandpaper and a retail brass cleaner gets enough of a polish to see a mirror finish under a microscope at 400x.

To add an answer to somebody else's question here there are other methods like "sol-gel", the name actually somes from solution and not solar. This method for multi-crystalline coatings including some solar cell materials is effectively mixing up some goo in a bucket, painting it on and then heating it up in an oven. The solar materials made this way are not as effective but really cheap due to not needing very high temperatures to fabricate - you don't have to melt silicon.

Re:

[Doc Ruby (173196)]

Can I get a number of joules per square millimeter to produce 20% efficient PV wafers about 8" wide, like the one in the article's picture looks?

Re:

[mikael (484)]

This web page [processpecialties.com] provides an explanation of the process. You are going to be melting sand into liquid, which is going to require industrial furnaceswhich precision controlled temperatures. High temperature furnaces can use up to 14 kilowatts of energy for a good few hours per ingot of silicon.

Then the other machines are required to grind and slice the ingot into wavers. These will use a standard industrial supply since they are just doing mechanical motion and maybe some water/gas cooling.

But that wouldn't tak

Re:

[Doc Ruby (173196)]

OK, so that's 684Mj:Kg * 2 = 1.296Gj:Kg *.6 = 777.6Mj:Kg.

An 8" wafer (like the one in the IBM story's pic) at 0.5mm thick is about 16246.5736 cubic millimeters, 16.2465736 cubic centimeters, (at 2.57g:cm^3), about 41.75g, or 18.624Gj [google.com] to produce. That's about 145 gallons of gasoline [google.com] to make a wafer that can produce something like 50Gj (using those numbers * .2 efficiency I left out) in its 30 year lifetime [slashdot.org], about 37% inefficiency from manufacturing. Long way to go.

Re:

[HonkyLips (654494)]

Yes, you're right- on a comparable scale of cost per megawatt produced our current PV technology costs more than 10x that of all other common energy sources. And although PV panels don't have any emmissions, the manufacturing process isn't exactly environmentally friendly either. However it is great for specific circumstances, mostly isolated (off-grid) applications, or for peak-demand in sunny areas (solar-powered air conditioners!). So developments like this one are welcome but it would be wrong to thi

Dear IBM,

[tjstork (137384)]

If you just give me a few hundred thousand dollars, I'll buy a little boat and just dump all your trash in the ocean.

signed,

Nigeria

ha ha bad pun

[rleamon (895852)]

"the cost of solar panels reflects it" Slow day at the news desk.

This is great

[Fengpost (907072)]

However, it would more impressive if someone can recycle the waste of LCD substrate. The LCD generates huge amount of waste as well.

What do they do with SOI wafers?

[cyfer2000 (548592)]

So what do they do with SOI wafers? Remove the whole buried oxide layer by CMP?

Oh man!

[Derek Loev (1050412)]

"Are those solar panels real??!! They're huge!"
"No way man, that's got to be silicon. There's no way it's natural."

hmmm

[thatskinnyguy (1129515)]

I watched a video once showing how processors are made. Hard to believe the highly polished and uniform wafers start out as a giant glass turd [siliconvalleyvisitor.com]. All kidding aside, the video also showed all the waste produced. And with silicon being worth a billion dollars an acre [news.com], a little bit of payback would be appreciated by chip manufacturers. I'm sure.

Re:

[rah1420 (234198)]

I've done better than visit a museum; I used to work at what was once Lucent Microelectronics in Allentown PA where, before they tore it down to make way for a ball stadium, they had wafer fabs and even a crystal growing installation onsite.

The best part of that job was signing up for being a chaperone for "Take your Daughter to Work Day" (it was still daughters only then) and herding the kids around to the different areas. We watched the ingots growing and being cut into wafers, polished, kerfed, and then

Question about solar power

[Entropius (188861)]

Aren't there ways to get solar power without futzing with photovoltaics?

What sort of efficiency can we get out of focusing sunlight on water (using cheap Fresnel lenses), making steam, and using it to turn a turbine? Is this cheaper per watt of generating capacity to build?

Seems like if you did this on seawater (on a big barge or similar), you could extract the water once the steam recondensed and getting desalination for free. If desalination becomes necessary to supply freshwater this might be worth it.

Re:

[dbIII (701233)]

Yes - but you have to build something big, which means a big capital cost. There have even been pilot projects that store enough steam to drive the turbines at night and another that splits ammonia during the day and reforms it at night to give base load. Since oil and coal are cheap there is little incentive to build very large solar thermal plants and it doesn't work well at the low end so we see small photovoltaic generating projects instead. If you double the size of the photovoltaics you get twice t

Re:

[mspohr (589790)]

This has been done. Look at this link for some projects in California. http://www.energy.ca.gov/siting/solar/index.html [ca.gov]

Look at the Stirling engine projects.

Re:

[Doctor Faustus (127273)]

What sort of efficiency can we get out of focusing sunlight on water (using cheap Fresnel lenses), making steam, and using it to turn a turbine?
You're talking power plants, now. Photovoltaics are good for rooftops, and when somebody has an acre of land they're not using somewhere. These are usually a lot closer to where the electricity is going to be used, so you save in transmission losses.

If you're just using the sunlight for heat, most of the newer projects use something other than water to collect hea

Re:

[khallow (566160)]

As long as none of those other technologies continue to be competitive with silicon-based solar cells, people will continue to care about silicon-based solar cells.

Re:

[Black Copter Control (464012)]

... Because now you can have a Beowulf cluster of solar panels?

Re:As the 8th Most Common Element (by Mass)...

[dwywit (1109409)]

I think it's got more to do with the "energy debt" that silicon wafers acquire during their manufacture. Anything is better than starting from scratch.

Re:As the 8th Most Common Element (by Mass)...

[Mark_MF-WN (678030)]

I remember reading about how the entire concept of "recycling" started with aluminum -- because the difference between refining new aluminum from bauxite and reprocessing existing aluminum is so incredibly great. Even iron is recycled to an extraordinary degree. And they say that 99% of all the gold that has ever been mined is still in use. There are even a few companies that believe that they can profitably recover platinum from the dust on America's highways left behind by catalytic converters!

Is it any surprise that silicon, being so expensive to purify, would ultimately start to see at least some measure of waste recovery?

Re:

[kestasjk (933987)]

I imagine that when it gets doped it's very difficult to un-dope. I guess that's why the silicon is only being used in photovoltaic cells and isn't suitable to be made into new chips, but I'm a layman.

Re:

[petermgreen (876956)]

I don't think doping generally goes all that deep, you can probablly get rid of most of it by stripping back the surface with a suitable chemical.

Re:

[weirdcrashingnoises (1151951)]

"The silicon quality usually necessary for solar production"

apparently u couldn't even bother to RTFS

Re:As the 8th Most Common Element (by Mass)...

[dbIII (701233)]

It's the same as with Aluminium, common as dirt but if we want to use it as something other than dirt that takes a lot of effort. Once it's in a usable form it's an incedible amount easier to use the metal than it is to turn the mined material into metal all over again.

With Silicon you have the added problem that you want really big crystals since you do not want a grain boundary halfway across your electronic component. The wafers are cut from a single large crystal and it takes a lot of effort to grow this crystal. Silicon is very hard so cutting it into wafers is not that easy either.

Forgive me, this is a bit offtopic

[causality (777677)]

The AC said "As the 8th most common element (by mass) in the Universe. Do we really need to worry about recycling?"

And for this he/she was modded Troll. That the AC missed the point that recycling the CPU wafers is about not wasting the effort and energy that went into creating them and is not about the abundance of unrefined silicon is most likely a simple careless mistake and there is no evidence to the contrary. Assuming that it's a deliberate troll attempt and wasting mod points that could have be

Re:

[lena_10326 (1100441)]

Assuming that it's a deliberate troll attempt and wasting mod points that could have been used to promote the responses that corrected it, in my mind, says more about the moderator who did this than about the AC who was factually wrong (for whatever reason)...

...

Because I have noticed a decline in the quality of judgment calls made by some moderators (certainly not all and not most of them) and it tends to express itself in this way...

There are more trolls moderating... than there are posting. Just

Re:

[pipatron (966506)]

How would you know, if you don't have an account?

Or maybe you're just being grumpy because people think you're an asshole and mod you down, thus posting as AC gives you a higher starting score?

Re:Not real news

[networkBoy (774728)]

But what am I going to use for pellet gun targets now?

Seriously, a 1x2, slotted to hold wafers, A-framed, and backstopped by a heavy tarp fed to a 55 gal drum is the most awesome way to dispose of scrap wafers ever. We generate about 100 a year at my site and they pile up in 5 gallon buckets waiting to be sent to scrap, I just like helping along the process...
-nB

Re:Not real news

[Animats (122034)]

Right. I heard the same thing from an Applied Materials VP.

Besides, the serious players in the solar business are now making solar cells five square meters at a time, using gear based on LCD panel fab technology. Solar panel production has gone way beyond using recycled IC wafers.

There's been commercial wafer recycling [processpecialties.com] for years.

Re:

[Lord Bitman (95493)]

Anyone care to clue me in on why the parent was modded "troll"?
IBM is bragging about having developed a cleaner way to do a wholely unneccessary process. Is that not fact?