High on a hill was a lonely sunflower. Not a normal sunflower, mind you; that would hardly be very notable. This sunflower is a solar sunflower that combines both photovoltaic solar power and concentrated solar thermal power in one neat, aesthetic package that has a massive total efficiency of around 80 percent.
The Solar Sunflower, a Swiss invention developed by Airlight Energy, Dsolar (a subsidiary of Airlight), and IBM Research in Zurich, uses something called HCPVT to generate electricity and hot water from solar power. HCPVT is a clumsy acronym that stands for "highly efficient concentrated photovoltaic/thermal." In short, it has reflectors that concentrate the sun—"to about 5,000 suns," Gianluca Ambrosetti, Airlight's head of research told me—and then some highly efficient photovoltaic cells that are capable of converting that concentrated solar energy into electricity, without melting in the process. Airlight/Dsolar are behind the Sunflower's reflectors and superstructure, and the photovoltaics are provided by IBM.
The two constituent technologies of the Solar Sunflower—concentrated solar thermal power and photovoltaic solar power—are both very well known and understood at this point, and not at all exciting. What's special about the Sunflower, however, is that it combines both of the technologies together in a novel fashion to attain much higher total efficiency. Bear with me, as this will take a little bit of explaining.
The reflectors are simply slightly curved, mirrored panels. Airlight has tried a variety of different reflector materials, from glass to mylar, but it looks like they have finally settled on aluminium foil, which isn't prohibitively expensive and has very high reflectance. Aluminium foil does need additional material to protect it from the elements, though, as it's very flimsy. The Sunflower has six "petals," each consisting of six reflectors. At the focal point of the 36 reflectors there are six collectors, one for each block of six reflectors.
The collectors are where most of the magic occurs. To begin with, each collector has an array of gallium-arsenide (GaAs) photovoltaic cells. GaAs is much more efficient at converting sunlight into electricity (38 percent in this case, versus about 20 percent for silicon), but it's much, much more expensive. With the Sunflower, though, space is at a premium: the sunlight is only focused on a very small region, so you need to use the absolute best cells available. The GaAs array in each collector only measures a few square centimetres, and yet it can produce about 2 kilowatts of electricity (so, one Sunflower generates about 12kW of electricity in total).
Photovoltaic cells, like most semiconductors, become less efficient as they get hotter. The GaAs cells used by the Sunflower have a max operating temperature of around 105°C. The problem is, when you focus the power of 5,000 suns on a single point, things get a lot hotter than 105°C. During one test, Airlight told me that they used the reflectors to melt a hole in a lump of iron (which has a melting point of 1538°C); during another test, the reflectors were misaligned and "we had molten aluminium dripping everywhere."
How then do you stop your collectors turning into very expensive puddles of molten metal?
Cooling with hot water? Cool.
The answer is a very clever cooling system, borrowed from one of IBM's specialities: supercomputers. In high-performance computing (HPC) installations, heat is one of the limiting factors on how much processing power you can squeeze into a given volume. Computer chips are tiny; you could physically cram hundreds of them into a 1U server chassis, if you so wished. Cooling them, though, is another matter entirely. You can use fancy heat pipes, or perhaps immerse the whole thing in liquid, but it only gets you so far.
Over the last few years, IBM has been working on advanced methods of liquid cooling, primarily to boost compute density, but also to reduce the amount of waste heat (getting rid of it increases efficiency and reduces costs). In a conventional liquid cooling setup, there's a water block (a lump of metal with some channels for liquid to flow through), a pump, and a radiator. The liquid is usually water, but it could be something else, like Fluorinert or mineral oil. Heat is transferred from the computer chip to the liquid, and then carried to the radiator and released into the atmosphere. This is inefficient for two reasons: there's a limit to how much heat can be "picked up" as the fluid passes through the block; and the heat being radiated into the atmosphere is wasted.
IBM solves both of these problems with its hot-water cooling technology. First, instead of the hot water passing through a radiator and venting the thermal energy out into the atmosphere, the hot water is simply used as hot water: to heat homes, or to drive industrial processes, such as desalination, pasteurisation, drying, cooking, etc. IBM already has an example of this in Aquasar, a supercomputer at ETH in Zurich, where the hot water is used to heat university buildings.
Second, to increase heat transfer from the chip to the water, IBM has replaced the dumb ol' water block with a piece of silicon with microfluidic channels. This piece of silicon, which is then stuck to the back of the computer chip like a tiny water block, has thousands of tiny channels that bring the water to within just a few microns of those pesky heat-generating transistors. This massively increases the amount of heat that can be dissipated, plus all of those discrete channels do a lot better job of dealing with chip hot spots (small regions that are more active than others) than the handful of giant channels in a conventional water block.