[ajuc]

But there was no gradient of temperatures before, and there is now - it's possible to use this gradient to produce energy. So this energy had to come from somewhere. It didn't come from electricity, because LED already has >100% efficiency (so there's more energy in the photons, than in electricity we put into the system).

So it has to come from temperature differnce between the place we cooled, and the place we heat up by photons, I think. So this place had to be cooler than the LED before experiment, so gradient of temperatures was there from the start, and we are only decreasing it.