Imagine you have a heavy flywheel composed of lead or bismuth, wrapped in fiberglass or Kevlar so the flywheel won't fly apart under the apparent centrifugal force when being spun at high RPM.
Now imagine you have an identical flywheel right next to (within 1/100th inch of) the first, aligned so both have the same axis of rotation. The two flywheels are not touching or connected in any way.
Now imagine you attach a motor capable of high RPM to the first flywheel, place a vacuum bell over the whole thing and evacuate the bell to a high vacuum.
Now you spin up the motor to ~25,000 RPM. Mysteriously, the second flywheel begins rotating (although at a lower RPM), too!
"Weird,", you may think, "there must still be some air in the vacuum bell and it's dragging between the two flywheels."
But the second flywheel is rotating in the opposite direction of the first flywheel, so it can't be air drag causing the rotation.
Why does this happen?
It's a weird quantum effect called the Barnett Effect, where an uncharged body (in this case, the first flywheel) undergoing angular acceleration experiences spin polarization of its electrons, thereby generating a magnetic field.
"But why does the second flywheel rotate opposite to the first?", you ask.
Well, that's another weird quantum effect known as the Einstein-deHaas Effect (conservation of angular magnetic momentum). The Barnett Effect and the Einstein-deHaas Effect are two sides of the same coin.
How can we exploit this phenomenon?
Well, if the two flywheels were mechanically connected such that they could still rotate in opposite directions, but the second flywheel was geared to spin faster than the first, it would create an imbalance between the two spin-polarized fields.
This imbalance results in a spin differential between the two flywheels any time they're rotating, which should result in a vector force attempting to push the flywheels to spin faster in attempting to conserve angular magnetic momentum. If the vector force is sufficient to overcome system friction, the flywheels should accelerate, and the faster they spin, the faster they try to spin.
We've essentially created a mechanical analog of a permanent magnet. The same effect occurs in a permanent magnet by dint of electron spin, which is why domain flipping occurs (as explicated in the thread, 'Understanding And Exploiting Physical Phenomena') via the exchange interaction to minimize the internal energy of the magnet.
This has implications for permanent magnets... it implies that one pole (one predominant magnetic moment direction) must be stronger than the other (as Howard Johnson wrote about in his book 'The Secret World of Magnets', to wit: "The north element (vortex) is dominant, and has proven to be the stronger vortex with higher gauss ratings."), since the vector potential A field is induced by the external magnetizing field, causing a rotoreflected B field with one predominant magnetic moment direction, then domain-flipping occurs (a reflection of the rotoreflection) after the external field is removed, creating a second predominant magnetic moment direction to minimize the magnet's internal energy:
http://www.aboveunity.com/thread/understanding-and-exploiting-physical-phenomena/?order=all#comment-10076212-cc32-42dd-8afc-a970005258be
So we have a vector (the induced vector potential A field (Ainduced)), a pseudovector (the rotoreflected B field... Brotoreflected=curl(Ainduced)), a pseudovector of that pseudovector, the reflected B field (Breflected ≅ Brotoreflected, which isn't a vector because it's a similarity transform (a flipping of domains) rather than a rotoreflection) and a vector (the rotoreflected vector potential A field... Arotoreflected=curl(Breflected)).
Which pole is dominant depends upon the conditions during magnetization. For modern magnets, it'll likely be a very small difference (since the crystallographic lattice of permanent magnets are easy-axes aligned prior to sintering, thus there are only two directions in which magnetic moment can be aligned and result in the lowest internal energy; and there's nothing stopping half the domains flipping due to the exchange interaction because the cubic crystallographic lattice of the magnetic material isn't unidirectionally magnetostricted, it's bidirectionally magnetostricted).