You mean because sometimes it's cloudy? Couldn't you solve that problem by turning on LEDs just when it's cloudy?
Plants are pretty inefficient, it's true, but I don't think they're very efficient even with optimal wavelengths. But, yes, solar cells and LEDs can indeed convert green and infrared sunlight to red and blue, with about 3% efficiency under normal circumstances.
If your plan is "light pipes, but fall back to LEDs" that means you have to build two complete systems. You get double the installation costs, incur the combined complexity of both systems (plus overhead), and box yourself into the limitations of both systems.
That might be true, but I'm not sure it is. It seems to me that you escape the limitations of both systems. Also, I think lightpipe daylighting generally costs orders of magnitude less than sunlight-comparable LED illumination, although I don't have firm numbers on lightpipe costs; so I don't think the installation costs double, just increase slightly. If we're talking about a greenhouse, where a leaky roof is a minor annoyance comparable to a leaky faucet, rather than a life-destroying disaster, lightpipe costs should go down quite a bit compared to their use in office and residential daylighting.
The big question for me is, if you're going to build a hothouse, why would you build it with an opaque roof? That just seems perverse, like writing an HTTP server in bash or assembly language or something: http://canonical.org/~kragen/sw/dev3/server.s
Efficient light pipes require solar concentration and thus tracking. That will drive costs up. Current residential/office light pipes are using only small fraction of sunlight, so they aren't relevant to this problem at all.
Also you haven't said anything about cooling, which the translucent roof inevitably requires in summer.
These are good points. Lightpipes can use non-imaging optics in many cases and thus avoid the need for tracking, but maybe not in this case, and I hadn't thought about that.
Cooling, though, is an issue whether your light comes from LEDs or the sun—either way it comes with a lot of extra heat. With the LEDs I'm familiar with, this is a worse problem than with sunlight, but in another thread https://news.ycombinator.com/item?id=20908476 I see a report that there are now illumination LEDs that reach an astounding 40–50% efficiency level. If that's true, it would make the LED cooling problem a little smaller than the sunlight cooling problem, instead of much bigger; and of course spectral matching helps.
> Couldn't you solve that problem by turning on LEDs just when it's cloudy?
Alternatively, every time when it’s not sunny... meaning all night every night. Of course, that would require experimenting if it’s actually more efficient, but AFAIK that’s what a MIT experiment found regarding growing basil (though they were optimising for flavour, not energy efficiency).
Plants are pretty inefficient, it's true, but I don't think they're very efficient even with optimal wavelengths. But, yes, solar cells and LEDs can indeed convert green and infrared sunlight to red and blue, with about 3% efficiency under normal circumstances.