A typical radiometer (partial vacuum) does not work by light pressure, but by heating up the molecules of near the black-colored vane. These heated molecules are therefore moving faster and "bang" into the black vane more than they do the white vane. Now, if you lower the pressure in the radiometer more (closer to a true vacuum), at some point the vanes will stop and then reverse direction because there aren't enough molecules of heated air to override the increased pressure of light on the white vane. Elastic impacts cause more force than inelastic impacts, so the black vane has about half the pressure of the white vane (assuming pure black, pure white, a perfect vacuum, and both vanes facing directly into the light source).

The real answer is thermal transpiration... Light is absorbed by the black side of the vanes more than the white side, which reflects the light. If light was a stream of water this would push the vanes round with the black side leading - but the radiometer goes the other way! (In fact light is, of course, a stream of particles and a very very sensitive radiometer with a vacuum inside would rotate the expected way). What is happening in the real radiometer is that the black side heats up more than the dark side and the gas on the dark side is warmer than that on the light side. It has been suggested that molecules of the hot gas hitting the black surface would push it round - but the surface and gas are at the same temperature, so no net momentum is transferred. But right at the edge of the vane there is an asymmetry which causes a small force called "thermal creep" (I don't quite understand this yet - perhaps it is due to a hot gas molecule hitting a slow one and locally changing the pressures) You need a low pressure in the radiometer to reduce friction to the extent that the vanes can move. You can't have a hard vacuum because it is the gas that causes it to move. EDIT: This paper talks about Einstein's equation for thermal creep, but doesn't say how Einstein got the equation! http://arxiv.org/PS_cache/physics/pdf/0402/0402011v1.pdf If I figure it out I'll let you know! EDIT TWO: I think I got it roughly : hotter gas is less dense - so there are more gas molecules next to the edge of the cold white side than there are next to the edge of the hot black side - so the excess molecules on the white side creep round the edge to attempt to equalize the density! By doing that they increase the pressure on the edge of the black side compared to the edge of the white side and push the vanes round.