Ron Reis, REC Silicon’s VP of technology, laid out the reasons why FBR was chosen as the successor to the ol’ Siemens polysilicon warhorse. While he credited the incumbent as “a proven technology for producing ultrapure polysilicon” that “meets the market’s needs,” he went on to detail what he called its “several disadvantages.” His summarized explanation shines a technical and economic light on why the company chose to go with the fluidized bed reactor (FBR) approach.
Siemens has “high-energy consumption in the deposition stage itself, which is the most energy-intensive stage of making polysilicon. It’s a batch process, which means that there’s a fair amount of nonproductive capacity,” he said. “At the end of a cycle you have to harvest the material out, you have to reseed the filaments and restore it to where it’s ready to receive the reactive gas--and that’s a lot of nonproductive time.”
“Then the finished product needs additional finishing,” he continued. “The rods are harvested, they either need to be chunked for the Czochralski process or they need to be processed for the float zone, to be received into the float-zone growers. So there’s a fair amount of additional work.”
“This process is basically a cold-wall process,” Reis said of Siemens. “Through electrically resistive heating, driving voltage and current through the starting polysilicon filaments, you create heat, and that heat is what is needed for the silicon-bearing gases to decompose, whether it’s TCS [trichlorosilane] or silane, it’s the same. But you only want the deposition to occur there, you don’t want it to happen anywhere else where there’s a hot surface.”
“So these [reactors, like the ones shown above] are cold wall because water or some other cooling medium is used to keep the whole rest of the reactor cool, so that the heat is focused right at the point you want to maximize your deposition. Because of that, a lot of the energy that is raining off these rods gets carried away in the form of warm water, which is basically taking the energy away. The gases come in and they’re removed at [the reactor outlet].”
Enough about Siemens. Why FBR? In a word (well, two)—energy consumption.
“That’s why we’ve invested in FBR,” the good professor explained. “It’s all about energy efficiency.” He showed a graph that represented the range of energy consumed by those using the Siemens process (kilowatt-hours per kilogram produced)—incumbent company and newcomer, new and aging equipment—and compared it with the FBR performance. The Siemens numbers ranged from 60-120 (call it 100 for comparative purposes), while the FBR’s granular method comes in at about 10KWh/kg—an order of magnitude difference.
This is not insignificant, since for many years one of the immutable objects on the crystalline-silicon PV manufacturing bill of materials/cost roadmap has been poly, and with its new fluidized scheme, REC Si thinks it has a way to whittle that cost down quite a bit.
Reis then detailed more FBR in situ (and apparently de facto) advantages: from a certain thermal perspective, it’s another case of “location, location, location.”
“Using fluid bed technology, which has been in existence around the chemical processing industry for a long time, it’s all about mixing the reactants and making sure the temperature gets to the right location. Because of this process, understanding how to control it, you’re able to put the reactive gases in, have them decompose into silicon when they soak down on the seed particles, but you don’t have to cool the wall. So all that hot water generated in a Siemens process is not needed.
“The energy that goes in is utilized just to keep the silicon hot for good decomposition then is removed out. That’s why it’s essentially 10% of the Siemens energy usage.”
Another advantage? It’s a continuous process. “Basically, we’re continually harvesting the deposited silicon,” the tech veep pointed out. “There’s a seed stage where very fine silicon particles are put in, a starting place for the silicon to be decomposed. That continuous nature helps lower the cost of the polysilicon coming out.”
Then there’s the handling, something that’s simplified when the end-product is basically pelletized material. “Granular polysilicon is a very useful form factor. It’s easy to handle, it flows well, we’re able to take it out and essentially just classify it, by its visible size, and then load it into bulk containers. So the labor usage and further processing of poly has gone down tremendously.
“At the customer receiving end, as they learn to use this form factor, they’re now able to get their costs down, because in a lot of cases it’s manual loading of crucibles which isn’t needed when you have a flowable product.”
Getting back to talking about to the intricacies of how they make the stuff, Reis showed a diagram with a complicated tangle of causes and effects, underscoring the “complexities associated with the FBR process, some of the interrelationships that go into taking silane and making granular poly.”
(Three or four of the ovals on the chart were empty, something the he told me in a subsequent email he couldn’t fill in because they were part of “our proprietary knowledge.” So we don’t know, for example, what an apparently key performance factor might be that receives causal input from silane concentration, residence time, fluidization, and temperature, and then effects byproducts and granular silicon yield. If any industry experts can offer any help filling in that blank, please submit a comment!)
He did elucidate one process pathway—fluidization. “Being able to fluidize includes the kinetics of the deposition of the gas, in our case silane, which is important for yield, because unconverted silane lowers your granular poly yield. We have essentially 100% conversion of what goes on. I can tell you that in the pilot phase, we did not have that in the early days.”
“Also associated with fluidization is stability, [since] by maintaining stability you’re able to maximize your run lengths,” he continued. “The stability of the operation is highly important to the quality also. You can see some of the influences that come into the front end of this,” such as reactor diameter, bottom head and nozzle designs, and the number of nozzles, as well as the operating pressure, bed height, and hydrogen feed—all of which have a direct causal link to fluidization.
Of course, this heavy lifting on the production side would be meaningless if the quality of the granular polysilicon were not up to snuff. “This program was chartered and built to make the lowest cost, best value poly for the PV industry,” stipulated Reis. “So as we expected based on the design, this is not electronic-grade poly, but it’s not that far off. The quality model of the pilot phase reflects, over the run length of a reactor run, what we anticipated.
“Process stability is the key, providing for longer run lengths. Early in 2009, we weren’t achieving these kinds of durations that we’re seeing now”; the “steep process operator learning curve has benefited us tremendously in terms of the stability.”
Apparently, the customers are pleased with the level of granular poly quality they are seeing. Kurt Levens, REC Si’s VP of commercial development, said that the “Silicon 3.0 FBR products have been qualified and accepted in six different applications,” with “18 different customers…on three continents that have used or are using the products.”
Back to the technical track, Reis broke down the list of usual quality consideration suspects when it comes to polysilicon. “If you start with donors and acceptors, because we start with silane, the purest form of silicon, donors and acceptors are not a concern—this has been verified. Similarly, using silane in the way the FBR process was designed, carbon and oxygen are not introduced.
“In the Siemens process, you have carbon chucks that hold the filaments and those quite often can migrate into some forms of the product. Because of the continuous process, we don’t have the filaments chuck issue to deal with.”
One of the more interesting revelations of the day—and a point of repeated questioning by the field-trip invitees—had to do with how the customers are integrating the handling of the poly granules into their forests of crystal pulling tools at the wafering plants. As with many aspects of poly mixology—both in production and consumption--there is no one preferred cocktail recipe.
“The experience right now is that blend tests have been done in ratios of 50-100% of FBR granules [to regular poly], with results coming back that the customers are experiencing normal cell efficiencies,” explained Reis.
“What’s really the focus of 2010 is getting more familiarity with our product. It’s a different form factor, so it melts and it crystallizes in a different manner than chunk poly. It’s a little less dense than chunk,” he said. There are lots of ongoing efforts to optimize “the blend ratio to get maximum parameters for the melting and crystallization, and [optimize] productivity.”
In the next installment of the Moses Lake blog series, more details will be revealed about the FBR production flow and the facility itself, as well as a discussion of root causes and other topics with REC Si’s VP of operations. To read part one of the series, click here.
PHOTOS COURTESY OF REC SILICON AND REC WAFER