Clean Energy Exchange

Four New Approaches to Ultrathin Silicon

Want to reduce solar PV costs from the start? Use less material and associated processes.

 The past few weeks have seen announcements and de-stealths from several companies all targeting the same part of the solar photovoltaic supply chain: reduce costs by eliminating most of the silicon material and related processes.

Finding ways to use less silicon made a lot more sense a few years ago when the material cost hundreds of dollars per kilogram — then prices nosedived to and below $20/kg, which made many of those ventures not as attractive, and many fell by the wayside. Nevertheless this remains the highest-cost part of the upstream half of the equation, so it's a clear target for further improvements.

Efforts in this area differ, from gas deposition to a molten formation (a la 1366 Technologies to an implant/cleave process at Twin Creeks and SiGen), but generally they all offer technology that promises to vastly reduce the use of silicon needed in the ballpark of 80-90 percent and eliminate several steps in the process flow, shooting for a sweetspot silicon thickness of a few tens of microns (30-50 or a bit more). Using less silicon means saving costs, simplifying processes, saving time, while still preserving efficiency. Thinner silicon has some issues with fragility and handling, though, so these players also have figured out ways to incorporate carrier substrates and/or other protective layers.

We've seen several recent announcements and updates from companies pursuing the model of using far less silicon. Here's four of the newest crop of "kerfless" wafer companies, who have a number of things in common, points out Fatima Toor, lead analyst for solar components at Lux Research. They all are based on high-temperature (1000-1100°C) chemical vapor deposition of epitaxial silicon, a type of thin crystalline silicon that conducts both laterally and longitudinally, unlike amorphous silicon that only conducts longitudinally. They all have some kind of lift-off process — except Scifiniti which says it needs none). Cell efficiencies range from 21 percent (Solexel, on a full-size 156×156mm area) to around 15 percent (Crystal Solar on a 125×125mm area, and Amberwave on a 1 cm2 area).

How is this race in ultrathin silicon shaping up? Solexel still has plans to get its Malaysian production facility up and running by 2014. Crystal Solar says that over the next 12 months it will demonstrate 19 percent cell efficiency, build several kWs of large-area modules to test for standards certifications, and build a 25-MW pilot factory with an industrial partner. Scifiniti and Amberwave are less far along the commercialization path, Toor says; they "are just getting their hands dirty."

Solexel Adds Funding, Boosts Cell Efficiency

Solexel, a graduate of the DoE's PV Incubator program, first unveiled at last year's Intersolar North America (July 2012): epitaxial deposition via trichlorosilane gas on a reusable mono-crystalline silicon template, resulting in a thin porous layer (around 35 microns) on a full-size square 156×156 mm back-contact cells. A flexible backplane attached to the back of the wafer — a resin material used in printed circuit board manufacturing — helps prevent breakage and enables electronics integration onto the backside of the cell, such as cell-level shade management capabilities. Lift-off of the thin cells and backplane is done by laser-cutting around the four sides and it all pops off. At the time cell efficiencies were 19 percent according to the company, with "a roadmap" heading toward 23.5 percent cell efficiency and 22 percent module efficiency, and ultimately a solar PV module manufacturing cost of $0.42/Watt.

Last week Solexel quietly submitted an SEC filing acknowledging another $14.8 million in new equity funding in its Series C round, spread among 12 total investors. The company raised $25 million in May 2012 in the first close of its Series C round, earmarked to help scale up the company's manufacturing process and eventually ramp manufacturing in a planned 200 MW/year plant in Malaysia's Senai Hi-Tech Park. Total funding is over $51 million, according to Mark Kerstens, Solexel's chief sales and marketing officer and acting CFO. This most recent closure did not have any industry investors, from the solar or any other sector, he noted; SunPower invested in the May 2012 round.

Also, the company's cells have now been NREL-certified at 21.2 percent, according to Kerstens. Because the cell is flexible and more sturdy than typical thin-silicon it doesn't need glass in the front or a typical frame, which opens up applications in flat roofs with weight issues, or for roofing shingles with cells built right in (Solexel was working with Owens-Corning on a building-integrated solar PV (BIPV) solar roofing shingle). Panels made from these back-contact cells are all black with no gridlines, a potential aesthetic benefit in certain residential and commercial applications, he added. Plans for doing cell manufacturing in Malaysia are still very much part of the plan, but he wouldn't disclose the timing except to say that construction has not begun. 

Crystal Solar Shares DoE Results, Scale-Up Strategy

Crystal Solar is now talking about its Department of Energy (DoE)-funded project to demonstrate viability of very thin monocrystalline silicon (~80 microns) using its in-house developed processes and equipment.

Results and takeaways from the 18-month SunShot program are posted here. By the end of 2011, six months into the first phase of its SunShot program, they had delivered 15.2-percent efficient wafers (125×125mm) bonded to glass; two months later they made five of the same cells fabricated and lifted off a silicon substrate, and made mini modules out of them (4×4 cells). At the end of nine months, concluding Phase I, the company's tool had throughput of 25 wafers/hour and 3 microns/minute deposition rate.

Crystal Silicon's technology, dubbed "Epi Thin-Silicon," is claimed to eliminate several steps in the current process flow, while retaining high efficiency of monocrystalline silicon. The approach uses high-temperature chemical vapor deposition (CVD) of trichlorosilane, and low-cost chemicals to reclaim the surface; it also involves a silver screen-printing process developed with Georgia Tech, and other steps in texture etching, diffusion, and nitride antireflection. The result is a 50-65 percent reduction in module costs to ~$0.40/W, or direct manufacturing costs "approaching $0.50/Wp." Capacity expansions can occur in 6-9 months instead of three years to ramp a polysilicon plant, the company notes.

The type of process that Crystal Solar uses has been talked about before in a research context (e.g. Fraunhofer, ISFH, imec), but the company has taken it further with refined production-grade equipment, multiple cell architectures, and commercial-scale demo modules, explained Ashish Asthana, Crystal Solar founder and EVP for business development and technology programs. Cell conversion efficiencies have exceeded 17.5 percent in "plain vanilla-type cells," he said, and he predicts 20 percent efficiencies on optimized cell architectures over the next few months. The company is piloting its epi thin silicon technology now and plans to ramp to high-volume production in the second half of 2014, utilizing multiple partners with their own cell and module technologies — "a sort of 'Intel Inside' model," he said, changed to "'epi thin silicon inside.'" Sales will come through equity and revenue sharing, he said.

Crystal Solar and Solexel both target thin silicon using epi, but Asthana differentiates that the latter uses a more complex proprietary process flow based on interdigitated back contacts, while Crystal Solar has "simpler, cheaper, conventional process flows" and a "production worthy toolset, with an intrinsically low cost-both capital and running costs."

The company showed a plan for a 100-MW factory using its gas-to-module approach with eight epi reactors, calculating a baseline production capital cost of $77 million and total cost model of $0.45/W. That compares, they say, to a typical c-Si line of $1.50/W, and best module costs of around $0.60/W.

Scifiniti Decloaks with Thin Silicon Wafers

Another startup, Scifiniti (née Integrated Photovoltaics), destealthed last week to join the ranks of companies seeking to lower the usage and cost of silicon PV, but with a catch.

Scifiniti says its technologycreates an ultrathin layer of multicrystalline silicon on top of a lower-cost silicon substrate. The ultrathin silicon layer cuts in half the energy required for processing silicon into wafers; reduces silicon use by more than 90 percent; eliminates defects associated throughout the typical silicon wafer process, from ingot growing ("no tail to top dopant variations," the company says) to wire sawing; and reduces overall wafer costs by half.

Along with the technology debut, Scifiniti has pulled in $10 million in Series B funding from its existing investors: Alloy Ventures, Firelake Capital, I2BF Global Ventures, and Peninsula Ventures.

Sharone Zehavi, chairman/president/CEO, and Curt Vass, EVP of bizdev, explained how it works. A low-cost metallurgical grade silicon powder is sintered and formed into a carrier substrate, in much the same way that ceramic kitchen tiles are made.  On top of that, a very high-throughput atmospheric chemical vapor deposition (CVD) tool deposits a barrier layer and reflective layer, followed by a layer of high-quality multicrystalline silicon to about 30-50 microns thickness. The resulting "SmartWafer" is the same size as a standard solar wafer (around 180 microns thick in total, 156 × 156mm) and can fit as a "drop-in replacement" for a standard wafer in a cell maker's existing lines and processes without changing any methods or purchasing new equipment (though processes may have to be gently tweaked).

Both the sintered carrier wafer and the thin multicrystalline layer on top eliminate many process steps, and costs, associated with conventional silicon formation and processing, from growing the ingot to forming/shaping and sawing wafers. Bottom line, says Vass: a typical silicon solar wafer has about 5.5-6.0 grams/Watt of silicon. Scifiniti uses just 0.5 grams/W. With so much of the silicon and associated steps and costs removed, the company's SmartWafer will cost less than half of a typical solar wafer's $0.25-$0.30/W.

While the basic value proposition for SmartWafers right now is purely cost-comparison with conventional silicon wafers, the company has other plans for the future. The metallurgical-grade silicon carrier wafer is conductive, which would reduce the need for some conductive paste currently applied to solar cells, Zehavi noted. And there are "all kinds of other next steps you can do," he said, including adding in materials into the sintered silicon to "play with the bandgap" (germanium is a popular choice) and improve the material's light-harvesting capabilities. Of course there are many other solar cell technologies that incorporate different light-harvesting materials to hike efficiency, but typically these are too expensive to be competitive with silicon-based PV, Zehavi noted. Additionally, the SmartWafers are mechanically stronger than typical silicon wafers, Vass pointed out; that should improve yield losses due to breakage which can add up over time, which also translates to cost savings.

Target customers are the usual roster of cell and module manufacturers, Vass and Zehavi said. Scifiniti wants to sell only the wafers — not handing the technology and IP behind it to customers so they can do it themselves — which will maintain control over the technology and its progress. That decision was made early on after Zehavi arrived in late 2009, he explained: it's better to sell the material instead of equipment (and supporting services), the IP can be better managed and extended to next generations, and the barrier to entry is made as low as possible. Introducing more capital equipment, and proving it out at scale, would mean convincing each customer to write million-dollar checks, a much harder sell than cheaper drop-in wafers.

Now decloaked, the company will shift its focus to improving the technology and finding both customers to evaluate it and partners to help scale it up and establish facilities to make SmartWafers, which could mean sites near cell/module firms, but also with chemical or materials suppliers, Zehavi said. It costs about $14 million to equip 100 MW worth of SmartWafer capacity; a 300-MW capacity facility would thus cost less than $45 million. Zehavi figures the company will need around $20-$25 million more for a manufacturing facility. For future funding, Vass acknowledged the "natural next step" is to find a "partner investor" from somewhere in the industry, be it a cell/module maker or a supplier of chemicals or materials. (We've heard other silicon-targeting startups entertaining possible tie-ups with downstream solar players that are connecting the dots: more efficiency per cell and panel means fewer panels per project and lower costs, and/or more energy in an equal area.) Cell/module customers could be easily-identified as investors, Zehavi said, but many chemical companies and conglomerates are looking for a solar sector play as well.

Atomic Solar PV on the Horizon, Maybe

How thin can you go? Most of the startups we've been talking with are in roughly the same ballpark with ultrathin silicon, in the few tens of microns (30-50), searching for the sweetspot of conversion efficiency and durability and associated processing costs. But in the research lab, things get very interesting all the way down to single-atom thickness — and not necessarily with mainstream silicon, either.

Last week MIT researchers declared that a stacked sheet of one-molecule-thick materials could create an effective solar cell, sharing their findings in the journal Nano Letters. Two such layers only have around 1-2 percent conversion efficiency, compared with conventional silicon PV's best marks in the 20 percent ranges — but the films are just 1nm thick, and stacking lots of them could boost efficiency significantly. Pound-for-pound these cells could produce up to 1,000× more power than conventional PV, they say. That thinness has a potential play in applications where weight and footprint mean more than costs, such as space applications. The materials being studied, graphene and molybdenum disulfide, also are relatively stable in open air, under UV light, and in moisture, and wouldn't need protection under heavy glass sheets.

On the flip side, this work is about as far from commercial as one can get. It's all based on computer modeling of those two candidate materials, neither of which are easy or cheap to obtain and work with — or even able to be made at large-scale at this point. "Manufacturability is an essential question," acknowledges Jeffrey Grossman, the Carl Richard Soderberg Associate Professor of Power Engineering at MIT, "but I think it's a solvable problem."