IBM’s thin-film PV offers hybrid nonvacuum approach

PV World talks with IBM’s David Mitzi about the company’s new hybrid approach to thin-film photovoltaics, which solves some key problems in materials performance, and discusses the near-term goal for conversion efficiency at both the cell and module level.

by James Montgomery, news editor

February 22, 2010 – IBM has a long history (going back to the 1970s) of work in solar photovoltaics, mainly with silicon-based solar cells; in recent years it has expanded to explore thin-film PV, with one particular focus on low-cost solution-based approaches for depositing thin-film chalcogenide material. It’s a perfect fit for David Mitzi, manager of PV science and technology at IBM Research in Yorktown Heights, NY — he has a background in solution processing for metal chalcogenides in other technologies, notably phase-change memory.

Bringing a focus to thin-film PV hinges on several benefits vs. silicon-based photovoltaics, Mitzi explained. Silicon material has an indirect bandgap, which complicates its ability to absorb solar radiation; tricks and enhancements such as light-trapping can help, but these also add complexity and costs. On the other hand, thin-film materials such as CIGS/CIS and CdTe possess a direct bandgap, meaning thinner layers of them can absorb solar radiation. “The physics points to using thinner layers of material,” he said. Thin-film materials also are less susceptible to “grain boundaries,” meaning there’s less efficiency-dragging recombination vs. silicon, Mitzi noted.

Using a thin-film and thinner material layer opens up a wider range of form-factors, Mitzi said. 50-100μm of Si “is a pretty brittle material” — but a few microns of thin-film material may offer more flexibility, e.g., perhaps utilizing lower-cost manufacturing methods such as putting them on flexible substrates, and a wider range of applications such as flexible cells, building materials, and even clothing.

Moreover, IBM’s use of a particular collection of earth-abundant kesterite materials (copper, tin, zinc, sulfur, and selenium — aka, Cu2ZnSn(Se,S)4)  — solves another issue. Materials such as cadmium and indium are not only scarce, they’re also highly sought after for other electronics applications. Limited supply and high demand makes them exceedingly expensive and adds to overall PV manufacturing costs.

Click to Enlarge
Magnified view of the kesterite cross section. (Source: IBM)

Work in Japan with a different group of materials — a copper-tin-zinc-sulfide combination, using a vacuum-based deposition approach — has resulted in 6.7% efficiency. Mitzi and his team, though, are also focusing on a better way to work with the materials that reduces complexity and cost. Certain technologies, such as CdTe, lend themselves well to vacuum-based deposition, Mitzi noted, where the material is evaporated and lands on a substrate in a nice film. But the more complex a material is (such as CIGS, or Cu2ZnSn(Se,S)4 containing four or five elements), makes it more complicated to get everything to work properly — one would have to evaporate everything together (requiring simultaneous precise control), or do sputtering, put layers down, cosputter, and selenize in a hydrogen selenide atmosphere (and here the problem is dealing with the hydrogen selenide, Mitzi noted).

Another cost reducer is to go to a very large-area deposition, for which it’s arguably best to simplify processes and costs — and this, too, means finding an alternative to the vacuum environment’s complexity of chambers, pumps, components, etc.

There are nonvacuum “ink-based” approaches (liquid solutions and suspensions) for chalcogenide-based absorber layer deposition that can replace vacuum-based techniques,; these involve mixing at a molecular level to form smooth homogenous films. (This is demonstrated with spin-coated CIGS absorber layers, Mitzi noted.) A problem with solution processing, though, is the solubility of some materials (e.g., zinc) and the fact that drying can stress and crack the film — it may take several passes to lay down a desired thicker film (say, micron-sized) due to cracking/delamination, he said. On the other hand, in a suspension, it’s hard to achieve single-phase crystallization. And particle-based approaches require organic agents for wetting and dispersion and to avoid cracks and delamination — but these also introduce contaminants to the processes.

The challenge, then, is this: How to deposit thin films using nonvacuum processes (solutions and/or suspensions)? And with those, how to get an insoluble material into a solution to do solution-based processing, using an array of different solution-based processing approaches (e.g., dipcasting, doctor blading)?

The IBM team’s answer, explained Mitzi, is a hybrid solution-particle approach that offers the benefits and solves drawbacks of both approaches. The slurry/ink is a copper-tin chalcogenide solution, with in situ formation of dispersible particle-based zinc chalcogenide precursors. This deposition “combines the best of both worlds,” Mitzi said: solutions have solid particles, metal and chalcogenide elements that integrate into the final film. Solubility limitations are resolved, as the dissolved components can be engineered as binding media for particles (no need for organic binders). The solid particles add stress-relief and crack-deflection centers, so thicker layers can be deposited in a single deposition step. And contact between the two phases is a rapid reaction, homogeneous phase formation. Moreover, the process should be adaptable to various ultrahigh-throughput printing methods (flexographic, screen, inkjet, roll-to-roll) and coatings (spin, spray, curtains, slits).

Click to Enlarge
Solar cell in a working device. (Source: IBM)

With an NREL-confirmed 9.6% cell efficiency in hand, the short-term goal is improving that to 12% “over the next 12-18 months,” Mitzi said — that would translate at the module level to about 8%-9% efficiency, which is roughly “the low end of commercializable.” Ideally that could be pushed further to about 15% cell efficiency/10%-11% module efficiency — and he says there is “no reason” the technology cannot be “every bit as high as CIGS” (i.e., approaching 20% in small cells). Once higher efficiency is demonstrated, the next step is to tailor the rheology to match whatever is the chosen large-area deposition tooling. The team hasn’t really begun to optimize the devices, so there are “many knobs to tweak,” he said — starting with the absorber and related layers, which in CIGS depends on moly diselenide deposition. “That bottom layer, and the glass beneath, is very important,” he explained. “How permeable is it to ions from the glass substrate? How does the interfacial layer look?”

Ultimately, despite its background in developing and making related technologies (it has a large fab complex in New York), IBM doesn’t want to get into the solar manufacturing game, Mitzi said — they’ll look to license the technology to a partner to pursue commercialization.