PV update – thin-film silicon keeps on going July 8, 2013Posted by Maury Markowitz in solar.
Tags: solar power, thin-film
A while back I posted about Twin Creeks, a startup who’s looking to dramatically reduce the cost of PV through the use of much thinner cells. Since then I’ve been reading up on a number of such approaches, so I’d like to post a little overview of what’s going on.
What we do now
Almost all modern solar panels are identical in basic construction. They start with a process that grows a large block of silicon, using chemistry and a little heat for poly-silicon (pSi) or less chemistry and more heat for mono-sillicon (mSi). pSi processes give you a large cube of material, mSi produces a long cylinder.
To make cells, you cut slices off those blocks. The best “knives” we have for this are similar to a band saw, but use a thin steel wire covered in silicon carbide dust. There’s a lower limit to how small the wire can be before it starts snapping all the time, and that size is about the same thickness as a solar cell. That means that about 1/2 of the original silicon material in the block ends up as sawdust, or “kerf loss” as they call it in the industry.
Once the cells are sawn from the block, they are cleaned up, etched so they are less reflective, “doped” to make them electrical active, and then covered on the back with a grid of aluminum and the front with many fine wires of silver (solder) to pick up the electricity they generate. Now you have a cell.
Then the cells are glued to a sheet of glass, wired together, and covered with a thin sheet of plastic on the back to protect them from the environment. This sealed-up plate is known as the “laminate”. To finish the panel they glue the laminate into an aluminum frame, and connect the wires to a “junction box” (or “j-box”) with plug-n-play connectors on it. Now you have a panel.
Silicon feedstock represents the majority of the cost of the panel, but precise and up-to-date numbers are jealousy guarded. We can get some sort of idea of the costs from this report, which shows the relative cost declines between the end of 2010 and 2012. The first lighter blue bar on the left shows that in those two years the cost of the silicon feedstock fell by 27 cents per watt, representing a decline of 45%.
So let’s do a little math… if the price of silicon fell by 45% and that was 27 cents, then the original price was something like 27 / 0.45 = 60 cents per watt. If the rest of the panel fell by 11 cents and that was 19%, then it’s maybe 11 / 0.19 = 58 cents per watt.
So in 2010 the relative cost of the silicon was 60/(60+58) ~= 50% of the cost of the materials. In 2012 those costs were (60-27) = 33 cents and (58-11) = 47 cents, so now the silicon is 33/(33+47) ~= 40% of the cost.
The relative cost of the silicon is falling, but it’s still huge!
From the start, research teams around the world have been trying to attack the cost of PV by reducing the cost of the active portion – the silicon. Some have tried to do this using different forms of silicon, amorphous silicon to be exact, which can be deposited in very thin layers. Considerably more effort has been invested in alternate materials, the two main contenders being CIGS and CdTe.
FirstSolar had some success with their CIGS panels. With that exception every other thin-film player has either failed or been pushed into a niche. The reason is simple – the price of silicon declined 45% in two years. What were once competitive alternate technologies aimed at lowering the costs are now more expensive than traditional silicon techniques.
So this brings us to the topic of this column: new ways to make conventional cells without the kerf losses.
What these processes are attempting to do is hook into the enormous industrial machine for making the silicon feedstock, but change the processing steps to reduce waste. If the current cost of silicon is 33 cents, then if your process uses all of it instead of half, that implies the cost will fall to about 16 cents. But these processes generally go further, and attempt to eliminate several processing steps as well, and go directly from cut to finished cell ready for the wiring. This can push out another cent or two.
We talked about Twin Creeks previously, and there’s another player in the same space, SiGen. These companies are trying to develop a new “saw” that uses a particle accelerator instead of a wire to cut the silicon cells off the block. This process wastes little silicon, maybe 10 to 20% instead of 50%. It also leaves the surface nicely textured, even better than traditional methods. The downside is that machines are slower than wire saws, and use more power to operate. Twin Creeks was bought out recently, and SiGen’s been around for a while with no commercial use. So it’s not clear where their technology is going, if anywhere.
1366 technologies takes a different approach, making cells directly from molten silicon. This skips several additional steps, especially the lengthy cooling periods when the original material cools into blocks (or cylinders). 1366 opened a new factory in early 2013, so we’ll get to see how this scales out over the next few years.
Finally there are a couple of new players using a gaseous deposition technique. In these systems, silicon is mixed with other chemicals that turn it into a gas, which is then sprayed on a substrate, often aluminum or copper. The other chemicals then evaporate, leaving the silicon. It’s basically the same process as painting.
This is the same basic concept used in the amorphous silicon thin-film approach, but this version creates extremely high-quality “epitaxial” cells. Amorphous silicon tends to have very low efficiencies, around 10%. Solexel is boasting that their technique produces cells with efficiencies around 21%. Crystal Solar and Sifiniti are taking the same approach, but are nowhere as far along the development process as Solexel.
AmberWave is another concept based on the same deposition technique, but with the added wrinkle that they also deposit small islands of very high-performance cells within the substrate. Their idea is to use the bulk silicon as a carrier. But they are at the very start of development, and their web pages seem to indicate this is a pure-cost play, one that would not be competitive in efficiency terms.
Here’s the problem
Now let’s say that one of these approaches works, and we can reduce the cost of the silicon feedstock and processing, bringing our costs down into the 15 cent range – hell, let’s call it 12 cents for fun. And we’ll say that they get this running in the next two years, in Q4 2015.
Now look at that chart up above. In the two years from Q4 2010 to 2012, the price of “all the rest of the stuff” declined by 11 cents. And there’s no reason to believe this rate will change in the next two years… it’s aluminum and glass. In fact, it might be more likely that they won’t decline at all. But let’s assume the best-case scenario, and say that it will drop by another 11 cents by Q4 2015.
Ok, so that means our panel’s raw materials are 12 cents for the cells, and (47-11) = 36 cents for the rest of the materials. So our total costs fell from 33+47 = 80 cents to 12+36 = 48 cents. So the total price fell only 40%. Now lets say a miracle occurs and we half the price of the silicon again. Then we go to 6+36 = 42 cents, another 12%.
The problem here is that we’ve squeezed the manufacturing process so hard already that we seem to be well into the area of diminishing returns. Radical reductions in cost won’t translate into equality radical drops in overall price. And the reduction is even less, almost invisible, when you consider all the rest of the parts that make up a solar power system – things like the racking that holds the panels on the roof, the inverter that converts DC to AC power, or the wiring to get the power from the roof to your power panel.
Now there is one wrinkle here… all of these thin-film products are semi-flexible. Conventional cells are extremely brittle so we have to glue them to a sheet of glass so they don’t get bent too much. But these cells don’t need nearly that much protection. This means that we might be able to reduce the packaging costs – maybe they could be placed on a sheet of coroplast and covered with saran wrap.
This has the potential to greatly reduce the price of the panel as a whole – large sheets of tempered glass are not cheap. How much of an effect this will have I can’t say, I don’t have the raw numbers I’d need to make a guess.
The bottom line
I suspect that if any of these pans out, and it looks like 1366 and Solexel will at least give us some idea over the next year or two, we should expect panel prices to fall into the 40 cent/watt range well before 2020. At that price we can expect total system costs just under $1 a watt, and installed costs in the $1.50-$1.75 range.
At that price, the NREL LCoE calculator turns up a price of 9.2 cents per kilowatt hour, which is the equivalent of today’s grid power in the 6.5 cent range (add 20 years of inflation to 6.5 cents and you get 9.2 cents). That’s the same cost as a nuclear plant (at least here in Ontario) and very competitive with wind and natural gas. It’s certainly well below grid parity if you consider time-of-use.
So the path to widespread grid parity is there, now we need to see if any of these technologies pans out. We only need to wait a year or two to find out.