It's basically converting the heat difference into light.
So you can't use it just to cool, it doesn't work that way.
No, that's only if you are in outer space. On earth it's the temperature of the air, or the walls.
It doesn't matter if the light eventually gets to outer space - what matters is that there is something hotter near you (namely the air).
i.e. it's not where the photons end up that matters, it's the total of all the photons that are being sent in your (the lamp's) direction that counts.
> it's not where the photons end up that matters, it's the total of all the photons that are being sent in your (the lamp's) direction that counts.
This is what I was referencing when I siad "the local background radiation." The common minimum was supposed convey high frequency when choosing a random spot in the known universe.
> On earth it's the temperature of the air, or the walls.
The temperature of the air only comes into play if it is emiting radiation at the frequency of light produced by the led and at greater intensity. I do not see why this would be required.
Blackbody radiation. Air (everything) always emits light.
> I do not see why this would be required.
It's a fundamental law of thermodynamics that requires a heat source and a cold source in order to convert heat into some other form of energy.
The air around you is the cold source, and the LED lamp is the hot source. If the LED is colder than the air it will not work - the LED must be hotter (in this particular case they put it in an oven).
The thing I think you're not taking into account is that this device has the effect of slightly increasing your black-body radiation. So you're at 1300K, radiating at 1300K, but we apply a little electricity and you start radiating a little more-- maybe you now look like you're at 1301K. Your heat hasn't actually increased, just apparent radiation, and as a consequence you cool down a little faster than your black-body profile would suggest.
So say you're at room temperature, radiating at room temperature, being warmed by the air, and not losing any heat. We apply current again, and you start radiating at a little over ambient temperature. The air is still warming you, but you're cooling a little faster than before, with the extra heat being deposited wherever the radiation is absorbed. To the air, you now feel cool.
Why should the absolute temperature of the ambient air have an effect on the process?
From the description given, the diode uses current to exchange ambient heat for emitted light. It cools the device, and heats whatever the light falls on, essentially acting as a remote heat pump. We can convert waste heat into a somewhat larger amount of "waste light", effectively increasing black-body radiation, which is great if light is more useful to us than heat.
Of course you don't get a free ride, and you can't make something cool without making something else hot (eventually), but conventional air conditioning is bound by the same limitation and is still plenty good enough for our purposes.
In principle, what prevents me from covering my roof with these things, radiating my waste heat (less atmospheric absorption) into deep space, and thus cooling my house?
Again, this supposing that the device could be scaled up to produce usable amounts of visible light while retaining the local cooling effect at room temperature, none of which have yet been demonstrated. Apparently their technique is to run the diode at as low of a voltage as possible, which presents some obvious problems.
[0] http://physics.aps.org/synopsis-for/10.1103/PhysRevLett.108....
Actually the abstract states it acts as a "thermodynamic heat engine", which would support what you're saying-- but given that it requires an input of electrical energy, I don't see how that could be right. They also liken it to "thermoelectric coolers". Possibly that was just mistyping or sloppy terminology, or possibly I'm missing something.
Fridge lights? House lighting in warm regions? Hmmm. On second thought it could have interesting applications depending how much it cools.
I doubt that they'll be able to scale this up too much though, since it probably relies on an effect that only occurs below a certain threshold.
I'd love a computer that glows brighter to release heat when it gets too hot, though :-)