An article about a recent Szilard Engine experiment has left me confused about the logic underpinning such an engine.
For those who aren't familiar with this engine, just consider a closed cylinder with a single particle inside. Now suppose you find out in which (longitudinal) half of the cylinder the particle is located. Slide a piston into the *other* half of the cylinder until it is at the half-way mark. As the particle bounces around, the piston expands. The significance of this simple arrangement is that we have used information (or a decrease in entropy resulting from our knowledge of where the particle is) to transfer energy (some of the particle's energy is transferred to the piston).
Feel free to rephrase that more carefully if you like, but I don't think it will change the thrust of my question, which is this. How is our knowledge of the particle's location essential here? To see what I mean, replace the asymmetric piston assembly by a movable disk that we can slide into the center of the cylinder at will, thereby creating a partition. Now surely it doesn't matter in which half the particle is located, the disk will move away from the region containing the particle thereby transferring energy from the particle to the movable disk. Well, I guess that's a question ;-)
I've got a feeling I'm missing the whole point of the Szilard engine, so I'm looking forward to hearing what others think.