It's the custom here at DKos to announce "breakthroughs" in renewable energy. These typically range from violations of thermodynamics to nice ideas that are unlikely to be practical in the foreseeable future. This one, however, is for real.
John Rogers' group here at the University of Illinois has figured out a practical new way to stack cells with different wavelength sensitivities to use the full solar spectrum rather than the narrow slice possible with a single junction. (news story,article)The idea of multiple junctions is old but the technique of making them is altogether new. It's combined with another idea already in use- inexpensive focussing using molded lenses. That allows the relatively expensive multi-junction material to cover only a fraction of the surface area.
The actual module efficiency efficiency exceeds 35%. At more than twice the efficiency of currently installed solar panels, and with anticipated reasonable fabrication costs, this should really open up the path to solar use. The authors emphasize the usefulness of ultra-high-efficiency panels for utility-scale projects, but they should also be very good for distributed small set-ups, in which available light can be patchy and the cost of installing each panel can be larger than the cost of the panel.
update: Thanks to Visceral for the reminder that I should have pointed out that these focussed systems only work with direct sunlight, not the diffuse light on a cloudy day. They do require a tracking set-up. That still leaves many applications in the broad regions where most days are clear. If the costs gradually come down on the multi-junction material, following typical electronics cost-vs-time curves, it should be possible at some point to skip the focussing and use them in diffuse light too.