That would require the solar panel to be more efficient at capturing solar energy than a plant

Aaand this is true! https://en.wikipedia.org/wiki/Photosynthetic_efficiency

1-2% of best plants vs 22% of best commercial panels! :)

This is not exactly correct. The link explains why by breaking down the stages of efficiency loss. The solar panels only mitigate the first step, so you are look at 22% vs 47% for plants. Then, the 22% is not including losses from converting the electricity to lighting (or heck, the energy to create those panels in the first place).

LEDs can select the most favorable wavelengths however so maybe it can approach parity again.

You could generate chemically costly preproducts for the plant synthetically too to help its efficiency. Things like increasing CO2 and providing at least nitrogen.

But while you are at it, you could directly to synthetic food... Hydrogen, methane, formaldehyde, sugars.

Sounds like the best solution mid term, before we'll be uploaded to computers. :)

I wonder if it's really necessary to convert the light to electricity as an intermediate step. I'm imagining a filter that's mostly transparent but has a fluorescent coating to convert UV to visible light that the plants can use, effectively brightening the sun in the useful part of the spectrum. Unfortunately Google is giving me the wrong kind of results for "fluorescent filter".

The material you are looking for is called a scintillator. It converts higher energy photons to visible. Of course it has its own issues / cost. Plus, you have to get the light from the rooftop to plants.

But you are not getting rid of the low plant efficiency. any other losses will simply come on top. nothing beats direct sunlight.

I guess that it may be possible. The plants probably have hard time optimizing for this and just eclipsing or poisoning other plants is better strategy for them.

You would however be also limited with efficiency of the lamps.