<<A Feynman sprinkler, also referred to as a Feynman inverse sprinkler or as a reverse sprinkler, is a sprinkler-like device which is submerged in a tank and made to suck in the surrounding fluid. A regular sprinkler has nozzles arranged at angles on a freely rotating wheel such that when water is pumped out of them, the resulting jets rotate the wheel; a Catherine wheel works on the same principle. The question of which way an inverse sprinker would turn (so, with the sprinker sucking the water in rather than pumping it out) was the subject of an intense and remarkably long-lived debate. The problem is now commonly associated with theoretical physicist Richard Feynman, who mentions it in his popular autobiographical book Surely You're Joking, Mr. Feynman!. The problem did not originate with Feynman, however, nor did he publish a solution to it.
The first documented treatment of the problem is in Ernst Mach's textbook The Science of Mechanics, first published in 1883. There, Mach argued that the device shows "no distinct rotation." As a graduate student in theoretical physics at Princeton University in the early 1940s, Richard Feynman became intrigued by the problem and built a make-shift experiment within the facilities of the university's cyclotron laboratory. The experiment ended with the explosion of the glass container he was using as part of his setup.
The sprinkler problem attracted a great deal of attention after the incident was mentioned in Surely You're Joking, Mr. Feynman!, a book of reminiscences published in 1985. Feynman neither explained his understanding of the relevant physics, nor did he describe the results of the experiment. In an article written shortly after Feynman's death in 1988, John Wheeler, who had been his doctoral advisor at Princeton, revealed that the experiment at the cyclotron had shown “a little tremor as the pressure was first applied [...] but as the flow continued there was no reaction.” The sprinkler incident is also covered in James Gleick's biography of Feynman, Genius, published in 1992, where Gleick claims that a sprinkler will not turn at all if made to suck in fluid.
The behavior of the reverse sprinkler is qualitatively quite distinct from that of the ordinary sprinkler, and one does not behave like the other "played backwards." Most of the published theoretical and experimental treatments of this problem have claimed (as did Mach and Gleick) that a sprinkler will not turn when made to suck in the surrounding fluid. It is now understood, however, that an ideal reverse sprinkler (i.e., one which can turn without friction and is surrounded by an ideal fluid) will accelerate towards the incoming fluid as the suction is being switched on, and come to a stop as the suction is switched off. The ideal reverse sprinkler will not experience any torque in its steady state. This behavior may be understood in terms of conservation of angular momentum: in its steady state, the amount of angular momentum carried by the incoming fluid is constant, which implies that there is no torque on the sprinkler itself.
Most experimental setups fail to detect any turning of the reverse sprinkler because the transient torque is not large enough to overcome the friction of the sprinkler's bearing. On the other hand, experiments with very low-friction bearings find a small torque on the reverse sprinkler even in its steady state.
This is now understood to be a consequence of the viscosity of the fluid being sucked into the sprinkler, which leads to the dissipation of some of the energy of the incoming fluid and imparts some angular momentum to the surrounding tank. This torque induced by viscosity causes the reverse sprinkler to turn weakly towards the incoming fluid (i.e., in the direction contrary to the motion of a regular sprinkler expelling water).>>