APPENDIX

Couette Flow

A great variety of possible viscous flow patterns, including complex and shifting
character patterns, occurs between two rotating cylinders, as described by the University of
Berkeley in 2003. In the experiment, there are two stationary concentric cylinders with a
constant amount of oil in a clearance between them. Both the outer transparent cylinder and
the inner cylinder, have the possibility to rotate at various speeds around their stationary,
collinear axes. With a fine aluminum powder as a suspension in the oil, the flow patterns are
easy to see. When either outer or inner cylinder rotates slowly, nothing unexpected happens.
However, when the inner cylinder rotates at a higher rotational speed, while the outer
cylinder is at rest, the fluid breaks into horizontal bands. When the outer cylinder rotates at a
similar higher rate, with the inner one at rest, no such effect occurs. How can it be that there
is a difference between rotating the inner as opposed to the outer cylinder? The answer is
found by looking at the clearance cross-sections (Figure A7.1). When the inner layers of the
fluid are moving more rapidly than the outer ones, they tend to move outward because
the centrifugal force is larger than the pressure holding them in place. A whole layer cannot
move out uniformly because the outer layers are in the way. It must break into cells and
circulate like the convection currents in a room, which has hot air at the bottom. When the
inner cylinder is at rest and the outer cylinder has a high velocity, the centrifugal forces build
up a pressure gradient, which keeps everything in equilibrium, as in a room with hot air at
the top.

Figure A7.1 Couette flow: fluid flow patterns in the clearance between inner and outer stationary cylinder, when
the inner cylinder rotates around a fixed axis, collinear with the inner cylinder.
1017

2005 by Taylor & Francis Group, LLC

1018

ROTORDYNAMICS

Figure A7.2 Centrifugal forces break Couette flow into bands (a), waves (b), and spirals (c) (case of slow
rotation of the outside cylinder in the opposite direction).

As the inner cylinder is speeded up, at first the number of bands increases. Then suddenly
the bands become wavy, and the waves travel around the cylinder. The speed of these waves
is easily measured. For high rotational speeds, they approach the fraction 1/3 of the inner
cylinder speed.
So far, there is no explanation, of why this speed equals 1/3 of the rotational speed. The
whole mechanism of the wave formation is not quite understood, and yet the flow is steady
and laminar.
If at certain speed of the inner cylinder, the outer cylinder starts also rotating, but in the
opposite direction, then the Couette flow pattern begins to break up. Wavy regions get
formed, alternating with apparently quiet regions (Figure A7.2), making a spiral pattern. In
these quiet regions, however, the flow is quite irregular; it is, in fact completely turbulent.
With increasing speed difference, the wavy regions also begin to show irregular turbulent
flow. If the cylinders are rotated still more rapidly, the whole flow becomes chaotically
turbulent.
Note that these two cylinders do not represent a model of a bearing, as the journal in the
fluid-lubricated bearing has a freedom to move laterally, while the Couette flow occurs only
in the clearance of two cylinders with stationary axes, restricted from lateral motion. In
addition, the fluid flow in bearings is forced in and out, in order to remove heat and wear
debris.
In the experiment described above, there are many interesting regimes of flow, which are
quite different, and yet are all contained as solutions of simple equations for various values of
one parameter the Reynolds number. With rotating cylinders, one can see many of the
effects which occur in the flow past a cylinder: first, there is a steady flow; second, a flow sets
in a pattern, which varies in time but in a regular, smooth way; finally, the flow becomes
completely irregular.
A similar effect can be seen in the column of smoke rising in still air. There is a smooth
steady column, followed by a series of twistings, as the stream of smoke begins to break up,
ending finally in an irregular, churning cloud of smoke.
The main lesson to be learned from this experiment is that a tremendous variety of
behaviors can be hidden in a relatively simple set of equations, which model these
phenomena. All the solutions are for the same equations, only with different values of
parameters. The difficulty is that there is still no mathematical power today to analyze them,
except for very small Reynolds numbers, i.e. in the completely viscous case.

2005 by Taylor & Francis Group, LLC