The standard radiometer that you get at science supply or home decorations house is already pumped to its proper operating pressure (somewhere around 50 mTorr give or take). The result is well known, shine light on the vanes and they spin in a direction away from the dark sides of the 4 vanes. (Kind of like a politician - shine light on one and you get lots of spin.)
Many moons ago I described a simple adaptation to allow one to connect a standard $8 radiometer to a vacuum pump. Some more detail may be found in "The First Five Years" but the scheme is as shown here:
As one would expect, when the device is pumped to below something like 100 mTorr, the vanes will begin to spin when illuminated. They really wind up at a few 10s of mTorr below this.
The operation of the radiometer is widely misunderstood (as I was until a physicist struck me on the upside of the head). The original argument that the spinning action is due to the pressure exerted by light is, of course, wrong and easily disproved by the direction of rotation. The most widely held belief is that the motion is due to molecular recoil from the dark (warmer) sides of the vanes. This too is incorrect. The rotation is really due to thermal creep aka thermal transpiration. I won't go into that here as there are good explanations on the web e.g.
http://math.ucr.edu/home/baez/physics/General/LightMill/light-mill.html. In brief, gas molecules are "pumped" from the cooler side to the warmer side. The flow past the vane edges is what creates the action.
Now, if you are careful during the pumpdown cycle you will see some interesting effects. Assuming that the innards of the device have been exposed to room air, when you get to around 1 Torr the vanes will rotate when strongly illuminated (I use a tactical flashlight with a xenon bulb). This is almost certainly due to outgassing.
As the pressure goes a bit lower, the vanes will not move and this remains the case until you get to about 0.1 Torr. Once the pressure is below this, the vanes will spin, slowly at first. The next effect that can be observed is, as the light continues to shine, the vanes will stop spinning. If you look at the pressure curve, the pressure rises (more outgassing) and when it rises enough the spinning will stop. You can continue this for a period of time and eventually the heat from the flashlight will complete the outgassing process and the vanes will spin continuously.
This makes a rather nice demo of the interplay of outgassing and thermal transpiration.
Is there any practical purpose to this? There is a device called a Knudsen pump that consists of small volumes interconnected with narrow passageways (the passageways are equivalent to the vane edges). Alternating volumes are warm and cold. The net effect of this is a real pumping action. Micromachined Knudsen pumps are being explored as miniature vacuum pumps.
