Doug at Nanoscale views gives a nice introduction (which I am going to quote in full) to this paper in Nature nanotechnology titled Nanomechanical detection of itinerant electron spin flop:
Many particles possess an internal degree of freedom called “spin” that is an intrinsic amount of angular momentum associated with that particle. The name is meant to evoke a spinning top, which has some rotational angular momentum about its axis when, well, spinning. Electrons have “spin 1/2”, meaning that if you pick a convenient axis of reference (“quantization axis”) that we’ll call z, the z-component of the electron’s spin angular momentum is either +1/2 hbar or -1/2 hbar. All too often we treat spin in a rather cavalier way. When people talk about “spintronics”, they are interested in using the spin degree of freedom of electrons to store and move information, rather than using the charge as in conventional electronics. One complication is that while charge is strictly conserved, spin is not. If you start off with a population of spin-aligned electrons and inject them into a typical solid, over time the spin orientation of those electrons will become randomized. Now, angular momentum is strictly conserved, so this relaxation of the electron spins must coincide with a transfer of angular momentum to the rest of the solid. Feynman pointed this out (somewhere in vol. III of his lectures on physics) – if you fire a stream of spin-polarized electrons into a brick hanging on the end of a thread, you are really applying a torque to the brick since you are supplying a flow of angular momentum into it, and the thread will twist to apply a balancing torque. Well, Zolfagharkhani et al. have actually gone and done this experiment. They use a ferromagnetic wire to supply a polarized spin current and an extremely sensitive nanomechanical torsional oscillator to measure the resulting torque. Very nice stuff.