I can't seem to export the sketch from the program on my desktop, so I have to describe it tonight. The idea is to use three different narrower bandgap panels; so one for say, red, one for blue and one for green. And instead of letting all incident light hit each pn junction, putting a dielectric reflector for green and blue on the red bandgap material and the blue and red dielectric reflector on the green bandgap and green and red dielectric reflector on the blue bandgap, so that each surface reflects the light that it cannot convert into a current.
You arrange each of these panels into a long triangular prism with one end have a long narrow opening going the length of a parabolic trough. So that you have a large surface area of reflected light focusing into a long thin line the length of the trough right at the opening of the solar panels. You're familiar with a prism:
So you make each side of the prism a solar panel with the dielectric surface facing inward the toward the other pn junctions, and the backing can be stainless steel, but I think a thin film solar cell with a silvered back contact would increase reflectance. However in this case, the medium in the prism is air, not glass. And two sides of the prism don't quite meet to close the 'triangle' of the prism, but are instead separated a bit to form a long aperture, which corresponds to the focus of the parabolic reflector. Mylar might work as a cheap reflector, it's easy to get , and really smooth for the price, not sure how it stands up to exposure. Anyway, the light that goes in has to keep being recycled around the inside of the prism striking from surface to surface before it can exit the prism. So the chances of any of the photons being absorbed as they recirculate goes up. I believe the collection efficiency should improve of the triple junction solar concetrators which use fresnel lenses, though you could use a fresnel concentrator and just put the prism aperture below the fresnel lens in the direction of the sun. It's just a difference of whether you use a mirror or a lens, though i think the mirror might be cheaper.
To cut heat, you could try a UV/IR cut filter, or just use a bandgap and dielectric coating for those too, and increase the number of panels. You really could instead of using a triangular prism, use a four sided, five sided, etc, even up to a circular tube with different band gap material at different parts along the arc of the cylinder, but then you'd have to mass produce them that way. Let other people work on that type of stuff. Also instead of cutting the UV/IR you might be able to use a phosphorescent wavelength conversion or fluorescent or quantum dots.
The heat sink should keep the solar cells sufficiently cool. The design is not merely to increase concentration of light on the surface, which increases temperature. The design is to recycle light, a trick similar in lighting applications using a light tunnel or the like. In this case There are potentially better ways to do this, like flexible solar material. If you wrap the solar material into a tube and just make the backing a silvered reflector, with a silvered cap at one end and an aperture at the other. And in addition to a heat sink,you could use a parabolic dish and cool the solar cell with liquid, or some other heat transfer fluid.
This particular design is one that should be usable with industry infrastructure as is. The narrower band gaps should be more efficient over the narrower band than the standard broad band solar cells over the same band gap, but I'm not sure on that. The industry however doesn't manufacture narrower band solar cells. Though the narrower bandgap stuff should have already been manufactured it'll be expensive for the higher efficiency yield it should produce because industry is focused on manufacturing the larger, less efficient solar panels.
