Robins Magnus
Robins Magnus
Robins Magnus
Abstract: The experimental observation by Robins1, that a projectile spinning about its
axis of travel experiences a transverse force (lift), was refuted by Euler2 purely as a
contradiction to expected symmetry of fluid flow. This, undoubtedly had taken away the
precedence of finding this effect by Robins and subsequently the same was credited to
Magnus3, which is a testimony of belief overtaking physical observation. In the last
century Prandtl 4 looked at this problem once again and came up with an upper limit on
lift that a spinning cylinder will experience. This is now considered a fundamental tenet
in explaining aerodynamic phenomenon of lift generation. However, over the last two
decades evidences are accumulating via experimental 5,6 and numerical investigations 7 ,8
that a new temporal instability affects this flow at high rotation rates, negating the above
mentioned upper limit. In this note, we trace the origin of this particular effect to its
present day status with respect to flow past a rotating cylinder.
1. The Beginning:
It must be pointed out that both Euler and Robins had mutual admiration for each
others work. For example, Robins published in 1739 Remarks on M. Eulers Treatise of
Motion. So the misinterpretation of Robins work was based truly upon the personal
belief of Euler. We also note that Euler enunciated his famous equations of fluid motion
only in 1752, the first mathematical model of inviscid flow. Even today, the Robins-
Magnus effect cannot be explained by solution of Eulers equation. Hence, Eulers
observations were his intuitive feeling that a spinning symmetric body (with top-down
symmetry) cannot experience an asymmetric force in the symmetric direction. In trying
to explain the occurrence of transverse force experienced by a spinning body Prandtl 4
advanced an explanation that was based on steady irrotational flow model. In doing so, he
also advanced a maximum limit to this transverse force. This is the next development in
this subject area that is interesting and discussed next.
2. Maximum Principle Enunciated:
In putting forward his results, Prandtl4 came up with an upper limit on the value
of this transverse force that a rotating cylinder will experience, as its rotation rate is
increased. This can be readily explained with the sketches of the flow field in Figure 1. If
one defines a non-dimensional rotation rate by = * D / 2U , where the cylinder of
diameter D rotates at * while being placed in a uniform stream of velocity U , then
U D
one can define a Reynolds number by Re = for this flow field. In Figure 1, the top
frame (a) depicts the steady inviscid irrotational flow field when the cylinder does not
rotate and one can note a perfect top-down and fore-aft symmetry of the flow field. In
frame (b), a case is depicted for < 2 , where both the front and rear stagnation points
(half-saddle points) are deflected downwards, causing the flow to exert an upward force
on the cylinder. With increase of to 2, these stagnation points move towards each
other and merge at the lowermost point on the cylinder (as shown in Figure 1(c)). For this
location of stagnation point, it is easy to show that the corresponding non-dimensional lift
value is given by the coefficient C Lmax = 4 . The lift value attained is a maximum was
reasoned by Prandtl heuristically, because if the flow field continues to be steady, then
with further increase of rotation rate the single half-saddle point of Figure 1(c) would
move in the flow field as a full saddle-point located on the closed streamline that
demarcates the flow field into two parts (as shown in Figure 1(d)). The region located
inside the closed streamline is insulated from the region outside and would be permanent
if the flow is steady. This suggests that the circulation cannot increase beyond the
rotation rate for the case shown in Figure 1(c). Thus, the Prandtls observation is based
on a steady flow using an inviscid irrotational flow model. In an actual viscous flow, the
circulation will be created at the solid wall continually that is convected and diffused
according to the governing Navier- Stokes Equation. A steady flow model, as proposed
by Prandtl, presupposes that an equilibrium exist between the process of creation of
vorticity at the wall and its viscous diffusion. While in the proposed model of Prandtl
such an equilibrium is assumed for all rotation rates up to the critical value ( = 2) - it is
assumed that the equilibrium is maintained even when the rotation rate is increased
further. How realistic is the model proposed by Prandtl? It seemed very real as no
counter-examples were encountered for decades where it was violated and it became a
very standard argument in all the text books about the infallibility of Prandtls limit.
However, some recent experimental and numerical observations seems to suggest
otherwise and they are casting serious doubts about the correctness of Prandtls logic.
3. Maximum Principle Violated:
In7,8,16, a third order upwind scheme was used for capturing these temporal
instabilities by taking very small time steps ( t = 10 5 ). In accurate numerical
computations, apart from achieving high spectral resolution it has to be ensured that the
discretization should not alter the physical dissipation that is only a second derivative of
the variable. This is the cause for the success of third order upwind schemes that adds
fourth order dissipation only. There are other methods that have been also used17,18 for
this problem and they have produced results that are totally different from experimental
observations. For example, in17, the author used a Streamline- Upward/Petrov-Galerkin
(SUPG) method, along with Pressure- Stabilizing/ Petrov-Galerkin (PSPG) numerical
stabilization terms for a case where the cylinder rotated eccentrically. The author stated
that these computations are also important from the point of view that in a real
situation it is almost certain that the rotating cylinder will be associated with a certain
degree of wobble. This was prompted by the authors earlier attempt 18 in computing the
flow for Re = 3800 and = 5 that resulted in an error of 400% at t = 250 . The reason
for the failure by SUPG/ PSPG is due to the massive artificial stabilization of the
numerical scheme by added second derivative term that interferes with physical
dissipation. Additionally, the time steps used are about 10 4 times larger than that used
in7,8,13-16 that would be simply not able to capture the transient events recorded in7,8,13-16 .
A typical set of results are shown in Figure 2, where the time variation of the
force and moment coefficients have been computed for Re = 3800 and = 5 . Here,
more accurate compact different scheme has been used 19 for computing the same flow
field. The instability displayed in this case consists of sharp discontinuous jump in the
values of force and moment coefficients at discrete times. Such instabilities were also
seen in the results reported in 7 ,8,16 . The physical mechanism behind the instability is
already explained in 8 by an equation based on energy principle derived from full Navier-
Stokes equation without any approximation. It was noted that the basic equilibrium
solution is destabilized by an intense interaction between the velocity and vorticity fields
of the primary and disturbance field. In the context of computations, the disturbance field
arises via truncation error and accumulated round off error. In simulating physical
instabilities, the chosen numerical schemes must be neutrally stable. If there is numerical
dissipation present that interferes with the physical dissipation, then the physical
instabilities will be subdued. This is the case with the results of 17 ,18 , where excessive
second order dissipation suppresses physical instabilities. It is interesting to note that the
time at which instabilities appear would depend on the amplitude of accumulated error.
As the compact scheme 19 has higher accuracy than the third order upwind scheme 8 , the
first temporal instability occurs later for the compact scheme. However, the value of C l
at which the first instability appears remains the same.
Closing Remarks:
The evolution of this problem has run parallel with the webs and tides of developments in
fluid mechanics through last three hundred years. It began with experimental observation
of Robins that was negated by heuristic observation of Euler. However, when the
analytical fluid mechanics was in its prime, Prandtl, another leading aerodynamicist, not
only explained the phenomenon theoretically, but also proposed a new limit on the
phenomenon based on heuristic logic and this survived for many decades before
evidences started accumulating that this limit may be violated. It is interesting to note that
the two experimental observations on this supposed violation is visual in one case and
uses analytical model to arrive at the observation in the other case. In contrast, the
computational evidences are based on the full governing equations. However, the
detection by numerical calculation depends on the existence of ambient noise (numerical
error) and the accuracy of the numerical method used. It shows the need to study and
develop models for the actual background noise that is present in experiments. A realistic
noise model with very high accuracy computational algorithms- that preserves
fundamental physical principles- would provide conclusive evidence of this and many
other problems of instabilities.
References: