J. Fluid Mech. (2014), vol. 754, R2, doi:10.1017/jfm.2014.

428

Inverse Magnus effect on a rotating sphere:

when and why

Jooha Kim1 , Haecheon Choi1,2, †, Hyungmin Park1 and Jung Yul Yoo1
1 Department of Mechanical & Aerospace Engineering, Seoul National University,
Seoul 151-744, Korea
2 Institute of Advanced Machines and Design, Seoul National University, Korea

(Received 26 April 2014; revised 26 April 2014; accepted 22 July 2014;

first published online 6 August 2014)

In some specific conditions, a flying spinning ball deflects in a direction opposite to

that predicted by the Magnus effect, which is known as the inverse Magnus effect.
To elucidate when and why this effect occurs, we measure the variations of the drag
and lift forces on a rotating sphere and the corresponding flow field with the spin
ratio (the ratio of the rotational velocity to the translational one). This counterintuitive
phenomenon occurs because the boundary layer flow moving against the surface of
a rotating sphere undergoes a transition to turbulence, whereas that moving with
the rotating surface remains laminar. The turbulence energizes the flow and thus the
main separation occurs farther downstream, inducing faster flow velocity there and
generating negative lift force. Empirical formulae are derived to predict the location
where the flow separates as a function of the Reynolds number and the spin ratio.
Using the formulae derived, the condition for the onset of the inverse Magnus effect
is suggested based on the negative lift generation mechanism.

Key words: separated flows, wakes/jets

1. Introduction
The Magnus effect is a well-known phenomenon by which a spinning ball moving
through the air deflects in the direction of rotation (Magnus 1853). One can easily find
the Magnus effect in many sports, such the baseball, golf, tennis, cricket and soccer
(Mehta 1985). For example, in soccer, a ball spinning in the clockwise direction
curves away from its principal flight path to the right. Applications of this effect are
not limited to sports but include external ballistics (Swanson 1961), flying machines
(Seifert 2012), ship stabilization (Morisseau 1985) and the saltation of particles
(White & Schulz 1977). Moreover, the Magnus effect is exploited in a number of
natural designs (Wauthy et al. 1998; Dickinson, Lehmann & Sane 1999; Vogel 2013).

† Email address for correspondence: choi@snu.ac.kr

c Cambridge University Press 2014 754 R2-1
 J. Kim, H. Choi, H. Park and J. Y. Yoo
The inverse Magnus effect refers to the deflection of a flying spinning ball in the
direction opposite to that predicted by the Magnus effect. This counterintuitive
phenomenon occurs only at some specific Reynolds numbers (Re = Ud/ν; U is the
flying speed, d is the ball diameter and ν is the kinematic viscosity) and spin ratios
(α = ωd/2U; ω is the spinning angular velocity). Many earlier studies (Maccoll 1928;
Davies 1949; Taneda 1957; Briggs 1959; Tanaka, Yamagata & Tsuji 1990; Aoki et al.
2003a,b; Barlow & Domanski 2008; Kray, Franke & Frank 2012; Muto, Tsubokura &
Oshima 2012a,b) have experimentally and numerically measured the inverse Magnus
force, but no relationship between Re and α generating the inverse Magnus effect has
been provided so far.
The classical explanation of the Magnus effect relies on the Bernoulli principle,
which states that for an inviscid flow, an increase in the speed of the fluid occurs
simultaneously with a decrease in the pressure and vice versa. However, since the
concept of the boundary layer was introduced by Prandtl in 1904, the conventional
Magnus effect has been explained in terms of a delayed separation on the retreating
side where the sphere surface moves with the flow. Interestingly, when the inverse
Magnus effect occurs, the flow separates farther downstream on the advancing side
(the sphere surface moves against the flow) than on the retreating side (Taneda 1957;
Kray et al. 2012; Muto et al. 2012a,b). This counterintuitive shift of the separation
delay has been explained as being caused by the turbulence being generated only
on the advancing side (Davies 1949; Krahn 1956; Taneda 1957; Briggs 1959; Aoki
et al. 2003a,b; Barlow & Domanski 2008; Kray et al. 2012; Muto et al. 2012a,b),
but the detailed process of the delayed separation has not been fully investigated due
to the difficulty of quantitatively measuring the flow near the surface of a rotating
sphere. To explain the asymmetric flow separation for the inverse Magnus effect, some
authors (Davies 1949; Krahn 1956) have introduced the concept of the effective (or
relative) Reynolds numbers on the advancing and retreating sides, i.e. Readv = (U +
ωd/2)d/ν = Re(1 + α) and Reret = (U − ωd/2)d/ν = Re(1 − α), respectively. However,
at a given Reynolds number, the return to a positive lift force after the inverse Magnus
effect with increasing spin ratio could not be explained with these effective Reynolds
numbers. Moreover, the conditions for the onset of the inverse Magnus effect, derived
from the effective Reynolds numbers (Krahn 1956), did not match those from the
experimental study (Swanson 1961).
In the present study, we investigate the mechanism of the inverse Magnus effect
and suggest a predictive model for the onset of the effect by measuring the drag and
lift forces directly from a rotating sphere by varying the Reynolds number and spin
ratio, and the velocity fields near a rotating sphere surface using DPIV (digital particle
image velocimetry).

2. Experimental set-up
Figure 1(a) shows the schematic diagram of the present experimental set-up for
the force measurement. The experiment was conducted in a closed-type wind tunnel
(Göttingen type), whose test section is 0.9 m wide, 0.9 m high and 4 m long. The
maximum wind speed in the test section is 60 m s−1 and the uniformities of the
mean streamwise velocity and the turbulence intensity are both within 0.3 % at a
free-stream velocity of 18 m s−1 . The free-stream turbulence within 0.5 % is low and
has little effect on the flow over a sphere in the Reynolds number range investigated
in the present study (Son et al. 2010). The sphere with the diameter of 150 mm was
made of acrylonitrile butadiene styrene (ABS) resin. A small DC motor was installed
754 R2-2
 Inverse Magnus effect on a rotating sphere: when and why

(a) Closed-type wind tunnel (b) Closed-type wind tunnel

CCD camera y
Sphere
Flow Flow
d x
DC motor
0.08d

0.027d Supporter Laser sheet

Eppler strut aerofoil Three force sensors Laser optics

Power supply

Amplifier Nd:YAG laser Timing hub PC

F IGURE 1. Schematic diagram of the experimental set-up for (a) the force measurement
and (b) the DPIV measurement.

inside the sphere and rotated it about the vertical axis in the clockwise direction.
The vibration of the sphere due to the rotation was minimized (less than 2 % of the
sphere diameter) by preventing resonance in the system. The rotational speed of the
sphere was varied by adjusting the voltage applied to the motor and was measured
using a tachometer. The sphere with the motor was fixed to a supporter whose cross-
section was designed as an Eppler strut aerofoil to minimize interference caused by
the support. The drag and lift coefficients on a non-rotating sphere that are measured
with the present experimental set-up are in good agreement with previous results (not
shown here), ensuring negligible support interference on the forces measured.
The ranges of the Reynolds number and the spin ratio for the experiment were
Re = 0.6 × 105 –1.8 × 105 and α = 0 (no spin)–1.7, respectively. The three-dimensional
forces on the sphere were measured simultaneously by three force sensors (one CAS
BCL-3L and two AND LCB-03s) attached to the supporter. The calibration curve was
linear and the uncertainty of the force measurement was ±2.5 %. The output from
the force sensor was amplified and sampled for 60 s at a rate of 32 kHz to obtain
the fully converged mean force. The data from the measurement were transferred to
a computer through an A/D converter (NI PCI-6251), after which they were post-
processed. The measured forces were corrected by subtracting those of the isolated
supporter which were measured separately.
The velocity measurement was carried out in the same wind tunnel as was used
for the force measurement. This facility was equipped with a glass wall and a
transparent acrylic ceiling to provide the necessary optical access. As shown in
figure 1(b), the DPIV system consisted of a fog generator (SAFEX), an Nd:YAG
laser (New Wave) operating at 120 mJ, a CCD camera (Vieworks VH-4M) with a
2048 pixel × 2048 pixel resolution, and a timing hub (Integrated Design Tools). The
fog generator produced liquid droplets of approximately 1 µm in diameter, which
were introduced into the wind tunnel. A streamwise (x–y) centre plane parallel to the
mean flow and normal to the axis of rotation was illuminated from one side of the
test section with a thin laser light of 3 mm thickness using laser optics. The flow
field in the shaded area behind the sphere was obtained by measuring the velocity
fields while rotating the sphere in the counterclockwise direction. The camera used
had a 60 mm lens to provide a 160 mm × 160 mm field-of-view (FOV). The camera
754 R2-3
 J. Kim, H. Choi, H. Park and J. Y. Yoo
was mounted on a three-dimensional traversing unit which allowed it to be moved
along the x and y directions. Eight FOVs were used to cover the velocity field of
−0.6 < x/d < 2.5 and −1 < y/d < 1. An iterative cross-correlation analysis was
conducted with an initial window size of 64 pixel × 64 pixel and with 32 × 32 final
interrogation windows. The interrogation window was overlapped by 50 %, leading
to a spatial resolution of 0.0086d. Spurious vectors were removed using a local
median filter that rejected vectors greater than three times the standard deviation of
a 3 × 3 window. Removed vectors were replaced by vectors resulting from a linear
interpolation in each direction from the surrounding 3 × 3 set of vectors.

3. Results and discussion

3.1. Lift and drag variations
The force coefficients on the rotating sphere are shown in figure 2. At a given Re,
the lift coefficient, CL = L/(0.5ρU 2 A), increases almost linearly with increasing spin
ratio, where L is the lift, ρ is the density of air and A (=πd2 /4) is the cross-sectional
area of the sphere. However, with a further increase in α, the lift coefficient abruptly
decreases at a critical spin ratio and then even becomes negative (i.e. the inverse
Magnus effect). The critical spin ratio becomes smaller as the Reynolds number
increases, and the magnitude of the negative lift force is greater for higher Reynolds
number within the range of our experimental parameters. After it reaches its minimum,
the lift coefficient increases and then becomes positive again. The spin ratio for the
return to a positive lift force also decreases with increase in the Reynolds number.
The behaviour of the drag coefficient, CD = D/(0.5ρU 2 A), is quite similar to that of
the lift coefficient in that it increases with the spin ratio, falls off and then increases
again, where D is the drag. The drag coefficient starts to decrease at the same spin
ratio as that at which the lift coefficient falls off, but it increases at a higher α than
the lift coefficient. We will discuss this point in detail later in this paper.

3.2. Mean and instantaneous velocity fields

The variations of the forces on the sphere are directly associated with the separation
points on both the retreating and the advancing surfaces. To observe the separation
points, we measured the velocity field around the sphere by means of DPIV and
estimated the separation angles (θsep ) based on the MRS criterion (Rott 1956; Sears
1956; Moore 1958): uθ = 0 and ∂uθ /∂r = 0, where (r, θ) are the polar coordinates
with the origin at the centre of a sphere. As shown in figure 3(a) (Re = 1.4 × 105 ),
the separation angle on the retreating side (θret ) gradually increases with the spin
ratio. On the other hand, on the advancing side, the separation angle (θadv ) is smaller
than that on the retreating side at low spin ratios, but rapidly increases to θadv = 140◦
at the critical spin ratio (α = 0.28), much larger than θret at the same spin ratio.
At this critical spin ratio, the shear layer separated earlier on the advancing side
reattaches to the rear surface of the sphere through turbulence generation from
the shear-layer instability (figure 3c,i; see below). Main separation delay by this
separation–reattachment process has also been found in other bluff-body flows and
their controls (Jeon et al. 2004; Choi, Jeon & Choi 2006; Choi, Jeon & Kim 2008;
Son et al. 2011). With this main separation delay, the radius of curvature, R, of
the streamlines becomes smaller on the advancing side than that on the retreating
side. The smaller R on the advancing side means a larger normal pressure gradient
according to the so-called Euler-n equation, ∂p/∂n = ρV 2 /R, where n is the local
754 R2-4
 Inverse Magnus effect on a rotating sphere: when and why

0.7 Magnus effect

0.6 Retreating CL > 0
side
0.5 Flow
CD
0.4 Advancing
side CL < 0
0.3
CL and CD

Inverse Magnus effect

0.2
CL CD Re
0.1 0.6
0 0.8
1.0
–0.1 1.2
1.4
–0.2
1.6
–0.3 1.8
0 0.4 0.8 1.2 1.6 2.0

F IGURE 2. Variations of the lift and drag coefficients with the spin ratio at Re = 0.6 ×
105 –1.8 × 105 . The lift coefficient is defined to be positive when the lift force is exerted
from the advancing to the retreating side (i.e. the Magnus effect occurs) and vice versa. It
should be noted that the eight circles on the left of the figure denote the cases investigated
in figure 4(a–d).

direction normal to the streamline and V is the local flow speed. Then, the pressure
is smaller on the advancing side than on the retreating side, resulting in negative
lift (figure 3a). With a further increase in the spin ratio, the separation angle on the
retreating side continuously increases, whereas that on the advancing side decreases,
resulting in an increase in CL (figure 3a). To understand these separation-point
movements according to the Reynolds number and the spin ratio, the instantaneous
velocity vector fields near the separation point are measured and shown together
with the normalized r.m.s. (root-mean-square) azimuthal velocity fluctuations in
figure 3(e–j). On the retreating side (figure 3e–g), the sphere surface rotates in the
same direction as the free stream and thus the momentum near the surface increases
with the spin ratio. With the increased near-wall momentum, the separation point on
the retreating side gradually moves downstream with the spin ratio (figure 3a). On
the other hand, on the advancing side, at α = 0.21 (figure 3h), the sphere surface
rotates against the free stream and thus the flow loses its momentum near the surface,
separating earlier than that on the retreating side. Therefore, the conventional Magnus
effect with positive lift coefficient occurs at this spin ratio (figure 3a). However, at
α = 0.28 (figure 3i), the flow on the advancing side initially separates but reattaches
due to the turbulence generation by the shear-layer instability. The reattached flow
with high momentum near the wall overcomes the strong adverse pressure gradient
formed on the rear sphere surface, resulting in a delay of the main separation.
This flow phenomenon is critical in the sense that the drag and lift forces rapidly
decrease. The formation of a secondary separation bubble, i.e. a closed-loop streamline
consisting of separation and reattachment, is a typical phenomenon for the critical
flow regime (Achenbach 1972; Choi et al. 2006, 2008; Son et al. 2011). With a
further increase in the spin ratio, i.e. at α = 0.60 (figure 3j), no secondary separation
bubble is observed on the advancing side because a turbulent boundary-layer flow is
formed without the formation of a separation bubble.
754 R2-5
 J. Kim, H. Choi, H. Park and J. Y. Yoo
(a) (b) (c) (d)
0.2
(b)
CL 0.1
0 (e) (f) (g)
(d)
–0.1
–0.2 (c)
140
130 (h) (i) ( j)
120
110
Inverse Lift Separation Normalized vorticity
100
Magnus
90 Flow 40 20 0 –20 –40
effect
80
–0.2 0 0.2 0.4 0.6 0.8 Normalized azimuthal r.m.s. velocity
Separation 0 0.18 0.36

(e) (f) (g)

(h) (i) Reattached flow with ( j)

high momentum near
the wall

Separation
bubble

Turbulence generation by
the shear-layer instability

F IGURE 3. (a) Variations of the lift coefficient and separation angles on the retreating
(θret ) and advancing (θadv ) sides with the spin ratio at Re = 1.4 × 105 . Here, the separation
angle is defined as the angle between the stagnation point and the separation point.
(b–d) Time-averaged vorticity contours and streamlines at (b) α = 0.21, (c) 0.28 and
(d) 0.60. (e–j) Instantaneous velocity vectors and contours of the r.m.s. azimuthal velocity
fluctuations on the retreating (e–g) and advancing (h–j) sides for (e,h) α = 0.21, (f,i) 0.28
and (g,j) 0.60.

As mentioned previously, the drag coefficient starts to decrease at the same spin
ratio as that at which the lift coefficient falls off, but starts to increase at a higher
α than the lift coefficient (figure 2). To investigate why this occurs, we show the
variation of the time-averaged streamlines with the spin ratio at Re = 1.0 × 105 in
figure 4(a–d). As the spin ratio increases from 0 (no spin) to 0.34, the separation
delay occurs on the retreating side only, and the upper recirculation bubble inclines to
the lower sphere surface with its centre closer to the surface (figure 4b), decreasing
the pressure behind the sphere. Therefore, the drag coefficient increases at α = 0.34
with a positive lift force (figure 2). At α = 0.53, the separation on the advancing
side is delayed more in the downstream than on the retreating side, and the
754 R2-6
 Inverse Magnus effect on a rotating sphere: when and why

(a) 1.0 (b) 1.0

0.5 0.5

0 0

–0.5 –0.5

–1.0 –1.0
–0.5 0 0.5 1.0 1.5 2.0 –0.5 0 0.5 1.0 1.5 2.0

(c) 1.0 (d) 1.0

0.5 0.5

0 0

–0.5 –0.5

–1.0 –1.0
–0.5 0 0.5 1.0 1.5 2.0 –0.5 0 0.5 1.0 1.5 2.0

F IGURE 4. Time-averaged streamlines with the spin ratio at Re = 1.0 × 105 : (a) α = 0
(no spin); (b) 0.34; (c) 0.53; (d) 0.80.

lower recirculation bubble moves closer to the surface (figure 4c). The shrinkage
of recirculation bubbles is noticeable owing to the separation delay on both the
retreating and the advancing sides, and thus the drag coefficient rapidly decreases
with a negative lift force (figure 2). With a further increase in the spin ratio to
α = 0.80, the flow on the retreating side fully attaches to the surface, whereas the
separation point on the advancing side moves upstream (figure 4d). Therefore, the lift
coefficient becomes positive again and the drag coefficient decreases more (figure 2)
in the absence of the recirculation bubble. At α > 0.80, the separation point on the
advancing side would move further upstream and thus both the drag and the lift
coefficients would increase. It should be noted that the wake is deflected upwards
with a negative lift force (figure 4c), whereas it is deflected downwards with a
positive lift force (figure 4b,d).

3.3. Predictive model of the inverse Magnus effect

Given that the inverse Magnus effect occurs due to the change in the boundary-layer
development on the advancing side, an important question is when this happens. To
find the relationship between Re and α that generates the inverse Magnus effect, we
derive an empirical relationship focusing on the variation of the separation angles on
both sides of a rotating sphere. Figures 5(a,b) show the variations of the separation
angles on the retreating and advancing sides with the spin ratio, respectively. On the
retreating side, the separation angles at different Reynolds numbers collapse well into
one line and increase with the spin ratio. This can therefore be expressed as a function
of the spin ratio only, as follows:
θret = 1.14α + 1.62 (rad). (3.1)
754 R2-7
 J. Kim, H. Choi, H. Park and J. Y. Yoo

(a) 2.8 (b) 2.8 (c) 2.8

2.6 2.6 2.6
2.4 2.4 2.4
2.2 2.2 2.2
Re
2.0 2.0 2.0
0.8
1.8 1.0 1.8 1.8
1.6 1.2 1.6 1.6
1.4 1.4 1.4 1.4
1.2 1.6 1.2 1.2
1.8
1.0 1.0 1.0
0 0.2 0.4 0.6 0.8 1.0 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 105 106 107
Re
(d ) 0.5 (e) 6 ( f ) 2.8
0.4 2.6
5
0.3 2.4
Re 4 2.2
0.2 0.8
2.0
CL

0.1 1.0 3
1.2 1.8
0
1.4 2 1.6
–0.1 1.6 1.4
–0.2 1.8
1
1.2
–0.3 0 1.0
0 0.2 0.4 0.6 0.8 2.0 0 0.2 0.4 0.6 0.8 1.0 105 106 107

(g) (h) 0.3

Maccoll (1928)
2.0 Tanaka et al. (1990) 0.2
Barlow & Domanski (2008)
1.5 Present 0.1
CL
Re

0
1.0
–0.1
0.5
–0.2

0 –0.3
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 –0.8 –0.4 0 0.4 0.8

F IGURE 5. (a) Separation angle on the retreating side (θret ) with the spin ratio, θret =
1.14α + 1.62 (solid line from best linear fit). (b) Separation angle on the advancing
side (θadv ) with the spin ratio. (c) Separation angle with the Reynolds number for a
non-rotating sphere (Achenbach 1972). Here, Reci and Recf denote the initial and final
Reynolds numbers for the critical flow regime, respectively. (d) Lift coefficient with the
spin ratio. Here, the triangles and rectangles denote the initial and final spin ratios for the
critical region on the advancing side. (e) Normalized effective Reynolds number (Re∗ =
Re∗adv /Re) with the spin ratio, Re∗ = exp(2.45α) (solid line from best exponential fit).
( f ) Separation angle on the advancing side (θadv ) with the effective Reynolds number,
θadv = −0.016/log(Re∗adv /Reci ) + 0.34 sin(2π log(Re∗adv )/1.89 + 2.66) + 2.25 (solid line from
best fit). (g) Region of the inverse Magnus effect (green) on the Re–α plane derived from
the condition θadv > θret . Here, symbols denote the Re–α values where the measured lift
coefficients were negative in the previous and present studies. (h) Lift coefficient as a
function of the difference between the separation angles θret and θadv , CL = 0.42(θret − θadv )
(solid line from best linear fit).

754 R2-8
 Inverse Magnus effect on a rotating sphere: when and why
This implies that the spin ratio is the most important parameter to determine the
boundary-layer characteristics on the retreating side, irrespective of the Reynolds
number. On the other hand, the separation angle on the advancing side is a function
of both Re and α. The critical region for a non-rotating sphere is bounded by the
initial and final critical Reynolds numbers, Reci and Recf , respectively, as shown in
figure 5(c). The triangles in figure 5(d) denote the spin ratios at which sudden delay
of separation occurs; that is, the effective Reynolds numbers at these α values equal
Reci . The rectangles denote the spin ratios at which the separation bubble on the
advancing side disappears; that is, the effective Reynolds numbers at these α 0 s equal
Recf . Thus, we newly model the effective Reynolds number on the advancing side as
follows:
Re∗adv = Re f (α). (3.2)
To find f (α), we plot the normalized effective Reynolds number Re∗ = Re∗adv /Re
with the spin ratio (figure 5e). This normalized effective Reynolds number can be
fitted using the exponential function, Re∗ = Re∗adv /Re = exp(2.45α). Consequently, the
effective Reynolds number on the advancing side is expressed as a function of Re
and α, as follows:
Re∗adv = Re exp(2.45α). (3.3)
Using the results obtained, we plot θadv versus Re∗adv in figure 5(f ). The separation
angles from different Reynolds numbers collapse well into one curve (for Re∗adv > Reci ),

0.016 2π log(Re∗adv )
 
θadv = − + 0.34 sin + 2.66 + 2.25 (rad), (3.4)
log(Re∗adv /Reci ) 1.89
and θadv rapidly increases when the effective Reynolds number is within the critical
flow regime. The inverse Magnus effect occurs when the separation is delayed further
more on the advancing side than on the retreating side. Therefore, we plot the region
of the inverse Magnus effect on the Re–α plane from the modelled separation angles
on the retreating and advancing sides using the following condition:

θadv > θret . (3.5)

Figure 5(g) shows the region of the inverse Magnus effect (green) on the Re–α
plane, derived from (3.1), (3.4) and (3.5), together with the previous and present
experimental results. As shown, the present model predicts the inverse Magnus
effect very well. It should be noted that the modelled region is valid when the
boundary-layer separation remains laminar on the retreating side (i.e. Re < Reci ).
Figure 5(h) shows the variation of the lift coefficient with the difference between
the separation angles on the retreating and advancing sides. The lift coefficients at
different Reynolds numbers and spin ratios collapse well into one line and increase
linearly with θret − θadv ,
CL = 0.42(θret − θadv ). (3.6)

4. Summary and concluding remarks

In this paper, we have shown that the inverse Magnus effect is caused by the
difference in the boundary-layer growth and separation along the advancing and
retreating sphere surfaces. That is, at the critical spin ratio, the shear layer separated
on the advancing side reattaches to the rear surface of the sphere through the transition
to turbulence from the shear-layer instability, whereas the boundary-layer flow on the
754 R2-9
 J. Kim, H. Choi, H. Park and J. Y. Yoo
retreating side remains laminar and separates earlier than that on the advancing side.
By introducing an effective Reynolds number newly defined in the present study, we
have provided the separation point as a function of the effective Reynolds number
and suggested a predictive model of the inverse Magnus effect based on the negative
lift generation mechanism.
Recently, lifting devices using the Magnus effect have received attention from the
Society of Unmanned Aerial Vehicle (UAV) Design and Testing due to their high
lift forces (or high wing loading) and stall resistance, despite the disadvantage of
requiring additional driving mechanisms and complexity compared with conventional
wings (Seifert 2012). The inverse Magnus effect occurs not only on a rotating sphere
but also on rotating objects such as spheroids (Tanaka et al. 1990) and cylinders
(Swanson 1961). A UAV designer thus should consider and avoid the inverse Magnus
effect because small UAVs operate in the Reynolds number range of 104 –105 based
on the diameter of the cylinder-shaped rotor (Seifert 2012), at which breakdown of
the lift force may occur. The present condition for the onset of the inverse Magnus
effect should be helpful in the design of more effective and safer Magnus-force-based
vehicles.

Acknowledgements
This work was supported by the research programs 2011-0028032,
2014M3C1B1033980 of NRF, MSIP, Korea.

References
ACHENBACH , E. 1972 Experiments on the flow past spheres at very high Reynolds numbers.
J. Fluid Mech. 54, 565–575.
AOKI , K., K INOSHITA , Y., N AGASE , J. & N AKAYAMA , Y. 2003a Dependence of aerodynamic
characteristics and flow pattern on surface structure of a baseball. J. Vis. 6, 185–193.
AOKI , K., O HIKE , A., YAMAGUCHI , K. & N AKAYAMA , Y. 2003b Flying characteristics and flow
pattern of a sphere with dimples. J. Vis. 6, 67–76.
B ARLOW, J. B. & D OMANSKI , M. J. 2008 Lift on stationary and rotating spheres under varying
flow and surface conditions. AIAA J. 46, 1932–1936.
B RIGGS , L. J. 1959 Effect of spin and speed on the lateral deflection (curve) of a baseball; and
the Magnus effect for smooth spheres. Am. J. Phys. 27, 589–596.
C HOI , J., J EON , W.-P. & C HOI , H. 2006 Mechanism of drag reduction by dimples on a sphere.
Phys. Fluids 18, 041702.
C HOI , H., J EON , W.-P. & K IM , J. 2008 Control of flow over a bluff body. Annu. Rev. Fluid Mech.
40, 113–139.
DAVIES , J. M. 1949 The aerodynamics of golf balls. J. Appl. Phys. 20, 821–828.
D ICKINSON , M. H., L EHMANN , F.-O. & S ANE , S. P. 1999 Wing rotation and the aerodynamic
basis of insect flight. Science 284, 1954–1960.
J EON , S., C HOI , J., J EON , W.-P., C HOI , H. & PARK , J. 2004 Active control of flow over a sphere
for drag reduction at a subcritical Reynolds number. J. Fluid Mech. 517, 113–129.
K RAHN , E. 1956 Negative Magnus force. J. Aeronaut. Sci. 23, 377–378.
K RAY, T., F RANKE , J. & F RANK , W. 2012 Magnus effect on a rotating sphere at high Reynolds
numbers. J. Wind Engng Ind. Aerodyn. 110, 1–9.
M ACCOLL , J. W. 1928 Aerodynamics of a spinning sphere. J. R. Aero. Soc. 28, 777–798.
M AGNUS , G. 1853 Ueber die Abweichung der Geschosse, und: Ueber eine auffallende Erscheinung
bei rotirenden Körpern. Ann. Phys. 164, 1–29.
M EHTA , R. D. 1985 Aerodynamics of sports balls. Annu. Rev. Fluid Mech. 17, 151–189.
M OORE , F. K. 1958 On the separation of the unsteady laminar boundary layer. In Boundary Layer
Research (ed. H. G. Görtler), pp. 296–311. Springer.

754 R2-10
 Inverse Magnus effect on a rotating sphere: when and why
M ORISSEAU , K. C. 1985 Marine application of Magnus effect devices. Nav. Engrs J. 97, 51–57.
M UTO , M., T SUBOKURA , M. & O SHIMA , N. 2012a Negative Magnus lift on a rotating sphere at
around the critical Reynolds number. Phys. Fluids 24, 014102.
M UTO , M., T SUBOKURA , M. & O SHIMA , N. 2012b Numerical visualization of boundary layer
transition when negative Magnus effect occurs. J. Vis. 15, 261–268.
ROTT, N. 1956 Unsteady viscous flow in the vicinity of a stagnation point. Q. Appl. Maths 13,
444–451.
S EARS , W. R. 1956 Some recent developments in airfoil theory. J. Aeronaut. Sci. 23, 490–499.
S EIFERT, J. 2012 A review of the Magnus effect in aeronautics. Prog. Aerosp. Sci. 55, 17–45.
S ON , K., C HOI , J., J EON , W.-P. & C HOI , H. 2010 Effect of free-stream turbulence on the flow over
a sphere. Phys. Fluids 22, 045101.
S ON , K., C HOI , J., J EON , W.-P. & C HOI , H. 2011 Mechanism of drag reduction by a surface trip
wire on a sphere. J. Fluid Mech. 672, 411–427.
S WANSON , W. M. 1961 The Magnus effect: a summary of investigations to date. Trans. ASME J.
Basic Engng 83, 461–470.
TANAKA , T., YAMAGATA , K. & T SUJI , Y. 1990 Experiment of fluid forces on a rotating sphere
and spheroid. In Proceedings of the 2nd KSME–JSME Fluids Engineering Conference, Seoul,
Korea, October, pp. 10–13.
TANEDA , S. 1957 Negative Magnus effect. Rep. Res. Inst. Appl. Mech. 5, 123–128.
VOGEL , S. 2013 Comparative Biomechanics: Life’s Physical World. Princeton University Press.
WAUTHY, G., L EPONCE , M., BANAI , N., S YLIN , G. & L IONS , J.-C. 1998 The backward jump of a
box moss mite. Proc. R. Soc. Lond. B 265, 2235–2242.
W HITE , B. R. & S CHULZ , J. C. 1977 Magnus effect in saltation. J. Fluid Mech. 81, 497–512.

754 R2-11