When water is squirted out of a sprinkler, the sprinkler head rotates in the direction opposite to the water flow (called the normal direction) due to the reaction force on the head. This is like a Hero's engine, the rotational equivalent of a rocket. Q: If water is sucked into the sprinkler head (actually pushed by a larger ambient pressure in the water vessel) will the sprinkler head (a) rotate in the normal direction, (b) rotate in the inverse direction, or (c) remain motionless? A: When water is sucked into the sprinkler head, the sprinkler head will rotate in the direction opposite the water flow, or inverse, direction.
Filling the tank, tube, and bucket with water, the system then functions as a siphon. Raising the bucket above the tank, the water flows out of the sprinkler, creating the normal sprinkler mode. In the normal sprinkler mode, the sprinkler head experiences a reaction force like a rocket and rotates in the "normal" direction. Its angular speed increases until it is limited by the viscosity of the water bath, and the rotation continues for a while after the water flow ceases. Lowering the bucket below the tank, the water flows into the sprinkler, creating the inverse sprinkler mode. In the inverse sprinkler mode, the sprinkler head starts into rotation in the inverse direction, opposite to the direction of the water flow, during the transient period while the water flow is starting. The angular speed of the head remains constant while the water flows at a constant rate, then ceases immediately when the water flow ceases.
A problem with analyzing this in terms of conservation of angular momentum involves the following observation: While water is flowing into the nozzle, reach in and stop the nozzle with your hand. Then release it. If conservation of angular momentum is the important concept, then the nozzle should remain at rest after it is released. However, it starts to rotate at the same angular speed it had attained before being stopped. Go figure it out and let us know what is happening.
The Edgerton Laboratory at MIT has an interesting hallway version of an idea related to the inverse sprinkler. Lamentably, it uses a complicated model that has too much friction and is probably too rigid, and therefore draws an incorrect conclusion.
The original paper on this demonstration is: Richard E. Berg and Michael R. Collier, The Feynman inverse sprinkler problem: A demonstration and quantitative analysis, AJP 57, 654-657, (1989). A video, Inverse Sprinkler Models, by Berg and Collier, produced at the University of Maryland to accompany the paper, can be viewed using the link below. The video demonstrates the experimental features of three different models of inverse sprinkler systems.
The article Alejandro Jenkins: An elementary treatment of the reverse sprinkler, presents more recent data and is probably more complete in its analysis than the earlier article by Berg and Collier using the apparatus in this demonstration and listed in the references linked below (Richard E. Berg and Michael R. Collier, The Feynman inverse sprinkler problem: A demonstration and quantitative analysis, AJP 57, 654-657, (1989)).