The place is loud but not cacophonous, bright but not migraine-inducing. Under the high ceiling, overhead cranes swing around, slick and shiny custom-made process machinery churns away (or not) within clear modular separation walls, and workers shout to each other or share a laugh.
The fab is highly automated, with robots and conveyor systems of various sizes and shapes--from AGVs scurrying past on the floor to large, strong-armed palletizing robots, their limbs moving around with that mindless yet almost anthropomorphic precision of the industrial robot class. Scores of 25-high stacks of tube-module trays, each holding 48 CIGS PV cylinders, are grouped on the floor in various spots.
The company took the keys to the Fremont facility in February 2007, and after some construction and equipment installation, had its first PV tube modules coming off the line by late summer of that year, according to my tour guide, company VP of sales/marketing/biz dev Kelly Truman. Over the next year, the team focused on improving tube performance and enhancing yields. His boss, CEO Chris Gronet, told me earlier that the first volume shipments of Solyndra's panels began in July 2008 and that 10 beta sites are running in the U.S. and Europe to validate the systems' performance.
The company puts a nameplate capacity of 110 MW on what it calls its "front-end" fab. It has borrowed this terminology from the chipmaking realm to differentiate its core CIGS tube production from the "back-end" processing down the road in Milpitas, where an outer tube sheaths the inner one, the units' endcaps are plugged in and sealed, the optical coupling liquid is injected between the inner and outer tubes, and the finished cylinder modules are inserted in the panel arrays, 40 at a time. (I didn't tour the back-end plant, but I have seen the video.)
Because Solyndra's front-end fab building once housed a succession of HDD fabs and thus has not been purpose built for the company's process-- or as my Truman put it, "the building was not built from scratch to fit the tools." The floorplan isn't bad, especially with the level of automation helping out, but the flow is less than optimal. He told me that in the second factory, they "will do the layout optimally for the logistics."
He leads me around to the large, multi-dunk-tank cleaning equipment where the incoming soda-lime glass tubes are cleaned and prepped for process. Truman pointed out that there's quite a bit of worldwide capacity with this kind of glass, since the medtech and pharma crowd have been switching over to plastic for their test tubes and vials.
"Unlike conventional panels, where there seems to be a bit of a fuss about the availability of flat panels of glass, we actually have a segment of the glass industry where there's an excess of supply," he related.
The robots transport the tubes to the molybdenum deposition tool, where, like with CIGS on flat glass or flexible foil, the metallic back contacts are put down on the glass substrates. The individually RFIDed tubes are rotated, with careful attention to film uniformity. The Solyndra fab boasts a proprietary manufacturing control system with an arsenal of sensors, creating a closed-loop metrology system reminiscent of a flat-panel display or chipmaking plant.
"We have alot of custom metrology in these things," Truman explained proudly. "At every step of the way every tube is tested, whether's it's in deposition or scribing for the monolithic integration. Even during our monolithic integration steps, there are metrology devices on the head of the scribing machines measuring a variety of properties as we go along. Every tube is tracked through the automation system and through every aspect of the deposition--each tube has its own identifier."
With six 70-MW lines slated for that new facility, soon to start construction just several hundred meters up Kato Road from Fab 1 on what is now a 30-acre vacant lot, the new place will be able to take full advantage of that touted full automation/advanced process control approach.
After the moly layers are deposited and a quick patterning-step stop, the PV cylinders are robotically schlepped to the most important equipment set on the floor--the CIGS absorber tool. Solyndra designed and built (with the help of subcontractors) the 45-MW system, which boasts an impressive footprint, stretching about 100 feet. The company uses the coevaporation approach to laying down its copper, indium, gallium, and (di)selenide, because, as CEO Chris Gronet reminded me, the highest conversion efficiencies for CIGS have been achieved using that type of deposition process.
Once the CIGS film stack has been put down (which Solyndra claims is a factor of two thinner than anyone else's layers), the tubes move to the junction partner/buffer layer tool. The company uses a proprietary wet solution process for this, in which the tubes are spray-coated with nanometer-scale layers of cadmium sulfide (although there's apparently work going on to make the buffer cadmium-free). Finally, the transparent conductive oxide (TCO) topcoat, reportedly an optimized i-ZnO/Al:ZnO cocktail, is sputtered as the last film in the process.
Then it's off to the laser-scribing tools (again based on Solyndra's own IP) for the monolithic integration process. Truman told me that the scribes, done six tubes at a time, are mostly helical, with "one linear scribe done at the very end to define each cell." He shows me how the individual cells, rather than being cylindrical per se, are slightly curved. Searching for a term descriptive of the shape, we hit on "croissant."
Before the cylinders are sent down to the back-end facility for packaging and paneling, each and every one of them is tested for its performance, electrical output, and the like on a tool familar to any PV manufacturer--the bright pulsing lights of the solar simulator.
Before we exited the factory floor, I asked Truman about the company's intellectual property and whether it might consider licensing its technology at some point. "We have a significant IP portfolio," he told me. "Four patents have been issued, multiple ones have been published, and there a large number of submitted applications. Among the ones that have been issued, there's one in particular that's very broad and fundamental, which basically says that if you put cells in series on an elongated object, the patent covers that.
"At this point, everything we've done is organic, we've not licensed anything internally nor have we approached anyone about licensing anything. We see the capacity expansion being done through own growth, not through licensing in the near term." Gronet said that once Fab 2 is built and running at volume, the company will eventually start looking for other manufacturing sites.
Now that Solyndra has emerged from stealth mode and I've seen its once-mysterious Fab 1 with my own eyes, the question growled in a Tom Waits' song a few years back, "what the hell are they building in there?" has been answered. While the facility may be impressive and the company's gameplan not without serious merit, there are unanswered questions and clarifications that Chris Gronet and this team will need to address at some point.
In the third and final part of my blog series on Solyndra, I discuss some of those questions.
To read the first Solyndra blog, click here.