PHYSICA L R EVIEW LET T ERS week ending

VOLUME 93, N UMBER 14 1 OCTOBER 2004

Falling Paper: Navier-Stokes Solutions, Model of Fluid Forces, and Center of Mass Elevation
Umberto Pesavento1,* and Z. Jane Wang2,†
1
Department of Physics, Cornell University, Ithaca, New York 14853, USA
2
Theoretical and Applied Mechanics, Cornell University, Ithaca, New York 14853, USA
(Received 28 January 2004; published 27 September 2004)
We investigate the problem of falling paper by solving the two dimensional Navier-Stokes equations
subject to the motion of a free-falling body at Reynolds numbers around 103 . The aerodynamic lift on a
tumbling plate is found to be dominated by the product of linear and angular velocities rather than
velocity squared, as appropriate for an airfoil. This coupling between translation and rotation provides a
mechanism for a brief elevation of center of mass near the cusplike turning points. The Navier-Stokes
solutions further provide the missing quantity in the classical theory of lift, the instantaneous
circulation, and suggest a revised model for the fluid forces.

DOI: 10.1103/PhysRevLett.93.144501 PACS numbers: 47.32.Ff, 47.85.Gj, 47.11.+j

A piece of paper or a leaf flutters and tumbles down in a problem in the most general case, i.e., free fall of a three
seemingly unpredictable manner (see Fig. 1). This rich dimensional flexible sheet, is daunting and unrealistic at
dynamical behavior has inspired many experimental and Reynolds numbers of the order of 103 . To simplify the
modeling works on falling plates [1– 4] and disks [5,6], as problem, we note that bodies of relatively large span-to-
well as observations of dispersing seeds [7]. Even before chord ratios fall essentially along a two dimensional
the establishment of aerodynamic theory, Maxwell of- plane [4]. In the case of a business card, the span-to-chord
fered a qualitative explanation of the correlation between ratio is about 1.8, and the motion in the spanwise direction
the sense of rotation and the drift direction of a tumbling is negligible. Taking advantage of these observations we
card [8]. More recently, Willmarth et al. classified dy- focus on a falling rigid plate in a two dimensional fluid
namics at different Reynolds numbers and dimensionless governed by the incompressible Navier-Stokes equations.
moment of inertia [5]. Belmonte et al. further identified a The choice of a rigid plate is convenient for numerics and
dimensionless number governing the transition from flut- for comparison against existing theories. Obtaining ac-
tering to tumbling [2]. Mahadevan et al. observed a curate numerical solutions in this regime turns out to be
relation between tumbling speed and the thickness of a nontrivial due to the thin tip of the geometry, the moving
falling card [3]. Models based on inviscid theory [9–14] boundary coupled to the fluid forces, and the small fluid
also exhibit dynamics qualitatively similar to those seen torque. Our numerical method is based on a vorticity-
in experiments. stream function formulation in a conformal grid fitted to
What appears to be lacking is a model of the fluid force the plate [16,17]. The conformal grid concentrates points
and torque that is constructed and tested against experi- at the edges of the plate and can be mapped onto a
ments or computations. The force predicted by inviscid Cartesian grid in which we discretize and solve the
theory includes added mass and a lift proportional to the Navier-Stokes equations efficiently via fast Fourier trans-
product of velocity and circulation [10]. An unresolved form [18]. In the mapped Cartesian grid the Navier-
issue is the choice of the circulation around the falling Stokes equations take the following form:
object, which cannot be determined from inviscid theory.
Previous models either assumed the circulation to be a
constant [14] or to be linearly proportional to the trans-
lational velocity [2,9,11]. The latter is appropriate for a
steady translating airfoil at a small angle of attack, as
given by the celebrated Kutta-Joukowski condition which
requires flow velocity to be finite at the singular trailing
edge [15]. The resulting lift is quadratic in velocity. While
the Kutta-Joukowski condition works remarkably well in
this special case, there is no direct evidence that it holds
for an object fluttering or tumbling in a fluid.
Theoretical progress is in part hindered by the lack of
simultaneous measurements of instantaneous forces and
flows around a falling object. Here we solve the Navier-
Stokes equations for the fluid subject to the dynamics of a FIG. 1. Rise of a falling journal cover under windless con-
falling body in two dimensions. Solving the falling paper ditions, selected frames from a footage filmed at 300 frames=s.

144501-1 0031-9007=04=93(14)=144501(4)$22.50  2004 The American Physical Society 144501-1

 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 93, N UMBER 14 1 OCTOBER 2004

@
S! p


1 2 (b)

(a) y x

Su  r!  r !; (1) v
b
@t Re a
u
(c) Ω
p


0

r 
Su  0; (2) θ
1 g

y' [chords]
2

where ! and u are, respectively, the vorticity and velocity 3

fields, and S is the scaling factor associated with the 4
conformal mapping from the exterior of the plate to a
5 1 mg
semi-infinite strip. The use of body-fixed coordinates 2
3
6 4
eliminates spatial interpolation as the plate moves with -1 0 1 2 3 4 5 6 7 8

respect to the fluid. This turns out to be crucial for x' [chords]
obtaining the almost periodic trajectories seen in Fig. 2. (d)
1 2
The forces on the plate are calculated by integrating the
-1.8 -0.67 0 0.67 1.8
stress tensor along the body [19]. The updated velocity is ω [u t /a]
then fed back to the Navier-Stokes solver through the
boundary conditions. The method is applicable to arbi-
trary shapes of the cross section, and for simplicity, we 3 4
choose an ellipse. To gauge the grid dependence of the
results, we repeat simulations using grid systems of 512 
1024 and 256  512 and find that the results presented
below hold for both resolutions.
A falling ellipse is characterized by six parameters: the
major and minor semiaxes, a and b, the densities of the FIG. 2 (color online). Navier-Stokes solution of a tumbling
ellipse, b , and of the fluid, f , the kinematic viscosity of ellipse at Re  1100, I
 0:17, and e  0:125. (a) Body-fixed
coordinate system and kinematic variables. (b) Trajectory and
the fluid , and the gravitational acceleration g [see
orientation of the chord (major axis of the ellipse) over five
Fig. 2(a)]. From these parameters, three dimensionless periods of motion. (c) The history of the chord and the force
quantities can be defined: the Reynolds number, Re  vector, equally spaced in time for the first period of the
2ut a=, using the terminal velocity estimated by balanc- trajectory in (b). The chords numbered from 1 to 4 correspond
ing gravity against the fluid force on a plate of size 2a and to the frames shown in (d) and to the times marked with dots on
q


































drag coefficient 1, ut  bg
b = f 1; the dimension- the force history of Fig. 3. (d) Vorticity field at four instants
2 2 during a full rotation. The frames display an area of 5  2:5
less moment of inertia, I
 b
a2ab
3
 b
; and the aspect ratio chords and they are 4a=ut time units apart. The vorticity field is
f
of the ellipse e  b=a. Note that the dimensionless iner- color coded on a logarithmic scale.
tia is related to the Froude number Fr defined in previous
work [2], I
/ Fr2 . center of mass elevation for a rigid plate without ambient
In Fig. 2 we show the Navier-Stokes solution of a wind. The swinging up motion may be familiar to those
tumbling ellipse for Re1100, I
 0:17, and e  who have seen sailplane stunts performing dead loops. At
0:125, released from rest with an initial angle of 0:2rad sufficiently high Reynolds numbers, Joukowski’s theory
with respect to the horizontal. After fluttering for a short predicts phugoid motion which swings up periodically in
transient, the ellipse tumbles with an almost periodic the special case where the angle of attack is constant and
motion [five periods are shown in Fig. 2(b)], which con- drag is negligible [9]. The situation here is different. The
sists of gliding and quick 180 rotations. During the Reynolds number is relatively low, about 103 , and drag is
glide, the ellipse pitches up due to the fluid torque. The non-negligible. A straightforward modification of
increased angle of attack results in an increased drag; Joukowski’s model to incorporate the lift-drag polar at
thus the ellipse slows down while pitching up. The small Reynolds number about 103 predicts no center of mass
translational velocity results in a cusplike shape near the elevation [20].
turning points. As the plate initiates a turn, the wake To investigate what might be missing in the Joukowski-
becomes unstable and breaks up into vortices due to like model of a falling plate, we turn to the instantaneous
Kelvin-Helmholtz instability. The vortices have a char- forces (Fig. 3). It is instructive to decompose the pressure
acteristic size of half of a chord. force into the contribution of the added mass and the lift
A casual observer of falling leaves or paper might proportional to circulation, as described in inviscid the-
notice that while falling downward on average, they can ory [10,14]. The viscous force and torque Fx , Fy , and 
rise momentarily as if picked up by wind. In Fig. 1 we are relatively small (see Fig. 3) and can be treated as
verify this observation by filming a falling journal cover perturbations. The force components, Fx and Fy , and
at high speed. The case shown in Fig. 2 is an example of the torque  are defined with respect to body-fixed axes

144501-2 144501-2
 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 93, N UMBER 14 1 OCTOBER 2004

[see Fig. 2(a)] and can be modeled as and the other given by the Kutta-Joukowski condition (see
Fig. 3).
Fx  f v  m12 v m11 u_ 
f b Agsin  Fx ;
(3)   cR a2   cL ajvj sin
2; (6)
where jvj is the speed of the ellipse and  is its angle of
Fy  f u m21 u m22 v_ 
f b Ag cos  Fy ;
attack, defined as the angle between the major axis of the
(4) ellipse and its velocity vector,  2 0; 2.
For the falling ellipse shown in Fig. 2, the rotational
 
m11 m22 uv Ia _   ; (5) contribution is about 10 times larger than the translational
where u
t and v
t are the components of the ellipse one. The lift corresponding to the circulation  is about
velocities along its major and minor axes, 
t is its 75% of the total lift, the remaining lift being generated by
angular velocity, 
t is the angle between the x axis the added mass terms with coefficients m12 and m21 . It is
and the direction of gravity [see Fig. 2(a)],  is the worth noting that values mij from the force fit differ from
circulation around the body, A is the area of the ellipse, those given by inviscid theory (see the caption of Fig. 3).
mij is the components of the added mass tensor, and Ia is Skin friction gives a contribution of about 25% of the
the added moment of inertia [15]. In the expression for total force and can be approximated with an expansion in
the forces Fx and Fy , the first term is the lift due to the the kinematic variables the ellipse u, v, and  (see the
circulation around the ellipse, the second and the third caption of Fig. 3). The pressure torque is 2 orders of
terms correspond to the added mass, the fourth term is the magnitude smaller than the torque on an ellipse steadily
buoyancy corrected gravity, and the F s are the viscous translating with speed ut . However, it can be modeled by
forces. a term proportional to uv as predicted by inviscid theory
The circulation  is unknown and needs to be modeled (see Fig. 3).
to complete the equations for the pressure force. Here, we The circulation model of Eq. (6) can be validated by
fit the pressure force from the Navier-Stokes solution by integrating the velocity field outside the ellipse. Figure 4
using Eqs. (3) and (4), where the added mass coefficients shows the circulation obtained both by fitting the pressure
are left as free parameters and  has two contributions, force with Eqs. (3) and (4) and by integrating the velocity
one proportional to the angular velocity of the ellipse  field. The circulation displays a strong dependence on the
motion of the ellipse and cannot be modeled by a constant
0.5
4 value as in [14] or by the classical expression for a trans-
3
lating airfoil as in [2,11]. Instead, the circulation is better
Fx [m g]

0
2 approximated by Eq. (6). The negative peaks in the cir-
-0.5
1
culation correspond to vortices shed at the turning points
t [a/ut]
5 10
1
15
2
20 25 30
of the ellipse trajectory.
1
In terms of the traditional decomposition of forces into
Fy [m g]

0
3
lift and drag, the proportionality between the circulation
-1
4
around the ellipse and its rotational velocity corresponds
5 10 15 20 25 30
t [a/ut]
τ [m g chord]

1.5
0.02
2
1

0 4 1
3
-0.02
5 10 15 20 25 30 t [a/u ] 0.5
t
Γ

0
FIG. 3. The fluid force and torque on the tumbling ellipse
shown in Fig. 2. Solid lines are computed forces, with the total -0.5

force shown as thin lines and the pressure force as thick lines.
The corresponding dashed lines are the best fits of the data 30 40 50 60 70 80 90
t [a/u ]
t
based on the quasisteady model of Eqs. (3) –(5). The added
mass tensor m11  0:53m and m12  3:1m, m22  1:5m, m21  FIG. 4. Circulation as a function of time for the falling
0:56m, and the circulation around the ellipse   2:6a2   ellipse of Fig. 2. The thick line corresponds to the value of
0:49ajvj sin
2 are obtained from the pressure force. These mij the circulation found by integrating the vorticity field up to a
differ from those predicted by inviscid theory (minv inv
11  m21  radius of 3=2 chords from the center of the ellipse. The thin line
0:0491m and minv 22  m inv
12  3:14m). The viscous corrections corresponds to the circulation obtained from fitting the pres-
are modeled with the expansions Fx  11 u 12 u2 , Fy  sure forces, the dashed line to the contribution of the rotational
21 v 22 v2 , and   31 uv 32  33 jj, with term of Eq. (6) alone. The peaks of negative circulation not
11  0:18, 12  0:0075, 21  0:070, 22  0:054, 31  captured by the fit correspond to the vortices shed by the ellipse
0:31, 32  0:05, and 33  0:16. at the turning points of its trajectory.

144501-3 144501-3
 PHYSICA L R EVIEW LET T ERS week ending
VOLUME 93, N UMBER 14 1 OCTOBER 2004

1 address the nature of the transition between fluttering and

(b)
0
(a) tumbling [23]. Understanding free-falling bodies might
also have interesting applications to insect flight, an area
−1
of research that partly motivated this study. Although
−2 insects might take advantage of both active and passive
−3 mechanisms to control their flapping wings, only pre-
y’ [chords]

scribed motions have been considered so far [24,25].

−4
Falling paper is a beautiful example of a passive flight.
−5 We hope that the model presented here will be also
−6
relevant to descriptions of forces in general flapping
motion.
−7
We wish to thank M. Salganick for discussion at the
mg
−8 early stage of this work and A. Andersen for discussion of
−9
the ODE model. This work is supported by NSF, by ONR,
−4 −2 0 2 4 6 8 by AFSOR, and by Packard Foundation.
x’ [chords]

FIG. 5 (color online). Trajectories obtained from ordinary

differential Eqs. (3) –(5) for different values of the translational
*Electronic address: [email protected]
and rotational lift coefficients cL and cR : (a) cR  2:6, cL  †
Electronic address: [email protected]
0:49, from the fit of Fig. 3; (b) cR  0, cL  1:5, as in classical
[1] P. Dupleich, NACA Technical Report No. 1201, 1941.
translational lift. Center of mass elevation occurs in (a) but is
[2] A. Belmonte, H. Eisenberg, and E. Moses, Phys. Rev.
absent in (b).
Lett. 81, 345 (1998).
[3] L. Mahadevan, W. S. Ryu, and A. D. Samuel, Phys. Fluids
to a lift proportional to jvj instead of jvj2 as in the case 11, 1 (1999).
of a translating plate. In the classical context, jvj is the [4] H. J. Lugt, Annu. Rev. Fluid Mech. 15, 123 (1983).
predicted lift for an airfoil translating and pitching at [5] W.W. Willmarth, N. E. Hawk, and R. L. Harvey, Phys.
Fluids 7, 197 (1964).
small amplitude [21]. This is particularly important at the
[6] S. Field, M. Klaus, M. Moore, and F. Nori, Nature
turning points of the trajectory, where the translational (London) 387, 252 (1997).
velocity is small. There, the increased angular velocity  [7] C. K. Augspurger, Am. J. Bot. 73, 353 (1986).
compensates for the decreasing velocity jvj and the lift [8] J. C. Maxwell, in The Scientific Papers of James Clerk
generated is sufficient for the center of mass of the ellipse Maxwell (Dover, New York, 1890), pp. 115–118.
to elevate. This mechanism for lift augmentation has been [9] A. A. Andronov, A. A. Vitt, and S. E. Khaikin, Theory of
recently emphasized in insect hovering [22]. Oscillators (Dover, New York, 1966).
To further verify the connection between center of [10] H. Lamb, Hydrodynamics (Dover, New York, 1945).
mass elevation and rotational lift, we arbitrarily vary [11] Y. Tanabe and K. Kaneko, Phys. Rev. Lett. 73, 1372
the coefficients of the rotational and translational contri- (1994).
butions in Eq. (6). The tumbling trajectories obtained [12] L. Mahadevan, H. Aref, and S.W. Jones, Phys. Rev. Lett.
75, 1420 (1995).
with such a procedure are shown in Fig. 5. Models with-
[13] Y. Tanabe and K. Kaneko, Phys. Rev. Lett. 75, 1421
out rotational lift display center of mass elevation only for (1995).
unphysical values of the lift coefficient (cL > 7). On the [14] L. Mahadevan, C. R. Acad. Sci. Paris 323, 729 (1996).
other hand, models including rotational lift show center [15] G. K. Batchelor, An Introduction to Fluid Dynamics
of mass elevation for the coefficients obtained from (Cambridge University Press, Cambridge, 1967).
Navier-Stokes solutions (cL  0:49 and cR  2:6 in the [16] Z. J. Wang, J. Fluid Mech. 410, 323 (2000).
case shown in Fig. 5). [17] Z. J. Wang, J. M. Birch, and M. H. Dickinson, J. Exp. Biol.
Finally the flow-induced coupling between translation 207, 449 (2004).
and rotation can decrease the speed of descent. The tum- [18] W. E and J. G. Liu, J. Comput. Phys. 126, 122 (1996).
bling ellipse shown in Fig. 2 has an average descent speed [19] H. J. Lugt and H. J. Haussling, J. Fluid Mech. 65, 711
of 0:4ut . In contrast, an identical ellipse parachuting (1974).
[20] M. Salganik and Z. J. Wang (unpublished).
down with its major axis perpendicular to the direction
[21] M. M. Munk, NACA Technical Report No. 217, 1925.
of motion would reach a terminal velocity of 0:77ut . It [22] M. H. Dickinson, F. O. Lehmann, and S. P. Sane, Science
would be interesting to find out whether the slow descent 284, 1954 (1999).
and the stable direction of tumbling motion are exploited [23] A. Andersen, U. Pesavento, and Z. J. Wang (to be pub-
by nature, for example, in seed dispersion. lished).
In our current work we are further validating the model [24] S. P. Sane, J. Exp. Biol. 206, 4191 (2003).
presented in this Letter experimentally and using it to [25] Z. J. Wang, Annu. Rev. Fluid Mech. (to be published).

144501-4 144501-4