J . Fluid Mech. (1987), vol. 183, p p .

199-218 199
Printed in Great Britain

The lift force on a spherical body in a

rotational flow
By T.R. AUTONt
Department of Applied Mathematics and Theoretical Physics, University of Cambridge,
Silver Street, Cambridge CB3 9EW, UK

(Received 21 December 1981 and in revised form 31 December 1986)

This paper concerns the flow about a sphere placed in a weak shear flow of an inviscid
’fluid.The secondary velocity resulting from advection of vorticity by the irrotational
component of the flow is computed on the sphere surface, and on the upstream axis.
The resulting lift force on the sphere is evaluated, and the result is confirmed by an
analytical far-field calculation. The displacement of the stagnation streamline, far
upstream of the sphere, is calculated more accurately than in previous papers.

1. Introduction
The magnitude of forces acting on small bubbles, droplets and particles in
non-uniform and unsteady flows is still a subject of some controversy, even when the
flow is inviscid and the shapes are spherical. The force on a stationary or moving
sphere in a uniform unsteady flow is well known (e.g. Batchelor 1967). The force on
a sphere in a non-uniform irrotational flow was first derived by Taylor (1928) and
Tollmien (1938), and subsequently re-derived by a number of authors (see Auton,
Hunt & Prud’homme 1987); but there has been no published calculation of the force
on a sphere in an inviscid rotational flow.
As well as being of fundamental interest for many problems in fluid mechanics,
there are practical applications of knowledge of the forces on a sphere in inviscid flow.
The chief one is in calculating the motion of bubbles in pure water when the Reynolds
number is large, the wake is thin and the outer flow is close to that calculated in
inviscid flow. Calculating the force on bubbles in rotational flow could lead to a better
understanding of the distribution of bubbles in pipe flow and in vortices, situations
of importance but whose analysis remains quite uncertain.
Whereas the analysis of irrotational flow around a sphere is linear, that of
rotational flow is, in general, nonlinear. However the latter analysis is approximately
linear if the strength of the vorticity of the rotational flow is weak enough that the
change in incident velocity across the sphere (Iwla) is much less than the relative
differencebetween the velocity of the sphek Uand the incidentvelocity u, on the centre
of the sphere, i.e.
I4 a 4 IU-u,l. (1.1)

Given the approximation (1.l),Lighthill (1956a, b) showed how the velocity field
can be calculated to a Grst approximation by calculating how the vorticity of the
weak rotational velocity field is distributed by the primary, irrotational velocity field
proportional to I U- u,l. Once the distorted vorticity is calculated, the secondary flow
t Present address : Imperial Chemical IndustriesPLC, Central Toxicology Laboratory, Alderley
Park, Macolesfield, Cheshire SKlO 4TJ, UK.
200 T.R.Auton
associated with these changes can be computed by the Biot-Savart law, and using
the method of images to satisfy the appropriate boundary conditions on the sphere.
I n this paper Lighthill's technique is followed in detail to compute the velocity
field, the surface pressure and thence the force acting on the sphere. It is shown that
certain arithmetic (or typographic) errors were made by Lighthill (1957)and Cousins
(1969,1970)in their calculations of the velocity and pressure fields. Neither author
calculated the complete pressure field or the net force acting on the sphere. Analysis
shows that this must be a lift force equal to
FL = CL$tU3p*(U - U o ) A 0, (1.2)
where p* is the fluid density and C , the lift coefficient.
Computations of CL gives a value of 0.500,and a subsequent control volume
analysis given in 56 confirms that CL = t.
This result is used in a more general analysis of spherical particles in non-uniform
flows by Auton et al. (1987).

2. Statement of the problem

A slightly sheared plane parallel flow approaches a fixed sphere of radius a. The
problem isto calculate the velocity fieldu(x)around the sphere, the pressuredistribution
on the sphere p, and the lift force FL acting on the sphere, subject to the assumptions
that the viscosity is zero, the density is uniform and the shear is weak but uniform.
Formally a solution is to be found to the governing equations

(U'V)W = ( O ' V ) U , (2.1)

where 0 = v A U (2.2)
and v - u = 0, (2.3)
subject to the boundary conditions

u = (uo+Ay, 0, O)T as z+- co (2.4)

and u*n=O onr=a; (2.5)
uoand A are constants. We seek an approximate solution when
aA
B=-41
UO

(which is the first term in an asymptotic expansion as A j - 0 ) .

The solution is based upon that of Lighthill (1956u,b, 1957).u is the sum of the
a
irrotational (or primary) flow V past a sphere (i.e. = 0) plus a small perturbation
(or secondary) velocity field u, proportional to a,,,
i.e.

u= v+u, (2.7)
where vAv=o
and V+ (uo,0, O)T as r+ co

V*n=O onr=a.
 Li$ force on a spherical body in a rotational flow 201

D
m
L
I m
X

m
m

Therefore, in terms of spherical polar coordinates ( r ,8, A) as shown on figure 1, the

solution to Vis

v, = 0. I
The primary-flow streamlines are defined by

( 3
r2 sin28 1 -- = pi, (2.10)

where po is the distance of the streamline from the axis far upstream of the sphere.
From (2.1),(2.6)and (2.8) it follows that to first order u satisfies
(V.V)W= (O'V)v (2.11)

where O=VAU
(2.12)
and O+ ( O , O , -A)T as x+- a,
) and ( o * V )u of O(A2)can be neglected when
where the second-order terms ( V O W o
A 6 1. (This approximation is not uniformly valid; it is not valid in particular near
to the stagnation points of the primary flow.) The solution to (2.11)for the vorticity
of the secondary flow, subject to (2.12) is defined by the primary velocity field.
Lighthill (1956b) showed how this solution for w can be expressed in terms of the
drift function t defined by
& = -da:
= - - dy
- - dz (2.13)
v v, K 7
X
t--+0 asx-t-a. (2.14)
UO

Notice that the surfaces of constant t describe the motion of planes initially
perpendicular to the flow far upstream when convected by the primary flow. Cousins
(1969) gives the following expression for t , which is convenient for numerical
integration as it removes the singularity in the integrand when 0 +.R :

uot(p07)' = loz+
(( 1
r'
-A)
' ~8') ) sin26 dB' +po cot 6,
+ ( ~ ~ / rsin (2.15)
202 T.R. Auton

e p / a = 0.1Ooo 0.2500 0.50oO 1.oooo 2.oooo

0.00 - 13.0203 - 4.7568 - 1.8393 -0.4006 -0.0240
10.00 -3.3288 -3.5797 - 1.7331 -0.3916 -0.0235
20.00 -0.6889 - 1.5231 - 1.2732 -0.3493 -0.0208
30.00 0.0209 -0.5039 -0.6988 -0.2578 -0.0129
40.00 0.3650 0.0239 -0.2362 -0.1257 0.0025
50.00 0.5828 0.3437 0.0992 0.0203 0.0249
60.00 0.7345 0.5555 0.3380 0.1556 0.0511
70.00 0.8393 0.6980 0.5040 0.2648 0.0759
80.00 0.9045 0.7880 0.6114 0.3402 0.0941
90.00 0.9333 0.8334 0.6687 0.3787 0.1018
100.00 0.9277 0.8390 0.6812 0.3805 0.0978
110.00 0.8893 0.8079 0.6527 0.3481 0.0834
120.00 0.8205 0.7431 0.5865 0.2869 0.0623
130.00 0.7238 0.6474 0.4860 0.2066 0.0396
140.00 0.6025 0.5235 0.3564 0.1220 0.0203
150.00 0.4602 0.3740 0.2089 0.0525 0.0076
160.00 0.2999 0.2022 0.0744 0.0129 0.0017
170.00 0.1216 0.0390 0.0068 o.oO09 o.Ooo1
180.00 0 0 0 0 0
TABLE1. q , / A sin h

where r’ is the positive root of

r ’ z ( 1 - g ) = p i cosecze. (2.16)

The vorticity is then given in terms oft by (Lighthill 19563, equation (58))

(2.17)

WA
r sin8
=-A COSA-.
Po I
Notice the very simple form of wA. The ring vorticity is independent o f t , as the
axisymmetric primary flow causes stretching, but not rotation, of ring vorticity.
It is convenient to subtract from o the uniform oncoming vorticity field. So we
define
o1= w - o 0 , (2.18)
where woY= -A sinh sine,
woe = - A sin 8 cos 8,
wOA= -A COS~. I
Numerically calculated values of o1and r/a as functions of p o / a and 8 are given
in tables 1 4 . The values of q,agree with those given by Lighthill (19563) to within
(2.19)

0.1. Lighthill’s results are not as accurate as those given here, as he used an
asymptotic approximation to calculate t, rather than a numerical integration of
(2.15).
 Lift force on a spherical body in a rotational Jlow 203

e p/a = 0.1000 0.2500 0.5000 1.oooo 2.oooo

0 0 0 0 0 0
10.00 2.1611 0.7837 0.3057 0.0672 0.0039
20.00 3.9730 1.5256 0.5716 0.1213 0.0059
30.00 5.5280 2.1547 0.8122 0.1601 0.0054
40.00 6.8714 2.6622 1.0150 0.1889 0.0031
50.00 7.9870 3.0562 1.1693 0.2121 0.0008
60.00 8.8498 3.3377 1.2724 0.2305 0.0003
70.00 9.4376 3.5052 1.3243 0.2427 0.0024
80.00 9.7345 3.5578 1.3266 0.2474 0.0069
90.00 9.7329 3.4963 1.2815 0.2434 0.0127
100.00 9.4336 3.3244 1.1924 0.2302 0.0180
110.00 8.8463 3.0487 1.0639 0.2077 0.0212
120.00 7.9896 2.6791 0.9018 0.1762 0.0213
130.00 6.8902 2.2288 0.7137 0.1372 0.0181
140.00 5.5829 1.7146 0.5099 0.0934 0.0126
150.00 4.1105 1.1595 0.3053 0.0508 0.0067
160.00 2.5269 0.6009 0.1268 0.0183 0.0023
170.00 0.9261 0.1363 0.0200 0.0026 0.0003
180.00 0 0 0 0 0
TABLE2. wl0/A sinA

e p / a = 0.10oO 0.2500 0.5000 1.m 2 . m

0 0 0 0 0 0
10.00 -0.9277 -0.1378 -0.0203 -0.0026 -0.0003
20.00 -2.5176 -0.6096 -0.1325 -0.0194 -0.0025
30.00 -4.0667 -1.1663 -0.3247 -0.0575 -0.0077
40.00 -5.4797 - 1.7007 -0.5420 -0.1129 -0.0162
50.00 -6.7040 -2.1729 -0.7484 -0.1757 -0.0270
60.00 -7.6987 -2.5603 -0.9238 -0.2352 -0.0384
70.00 -8.4324 -2.8474 - 1.0563 -0.2832 -0.0483
80.00 -8.8819 - 3.0238 - 1.1385 -0.3141 -0.0551
90.00 -9.0333 - 3.0833 - 1.1663 -0.3247 -0.0575
100.00 -8.8819 -3.0238 -1.1385 -0.3141 -0.0551
110.00 -8.4324 -2.8474 - 1.0563 -0.2832 -0.0483
120.00 -7.6987 -2.5603 -0.9238 -0.2352 -0.0384
130.00 -6.7040 -2.1729 -0.7484 -0.1757 -0.0270
140.00 -5.4797 - 1.7007 -0.5420 -0.1129 -0.0162
150.00 -4.0667 -1.1663 -0.3247 - 0.0575 -0.0077
160.00 -2.5176 -0.6096 -0.1325 -0.0194 -0.0025
170.00 -0.9277 -0.1378 -0.0203 -0.0026 -0.0003
180.00 0 0 0 0 0
TABLE3. w1JA cosh

3. The secondary velocity field

The secondary velocity field u consists of four parts:
u& the uniform shear-flow perturbation;
+
4, an irrotational flow field such that u, n = (ug 4) n = 0 on the surface of the
sphere ;
u:y the Biot-Savart velocity field due to the vorticity m1;
204 T. R. Auton

e p / a = 0.1Ooo 0.2500 0.5000 1 .oOOo 2.oooo

10.00 1.1101 1.6380 2.9379 5.7738 11.5213
20.00 1.0285 1.1765 1.6556 2.9806 5.8622
30.00 1.0133 1.0832 1.3247 2.1149 4.0309
40.00 1.0081 1 .Om4 1.1995 1.7314 3.1619
50.00 1.0057 1.0355 1.1412 1.5348 2.6813
60.00 1.0044 1.0278 1.1107 1.4263 2.3980
70.00 1.0038 1.0236 1.0941 1.3656 2.2312
80.00 1.0034 1.0215 1.0857 1.3344 2.1427
90.00 1.0033 1.0208 1 .0832 1.3247 2.1149
100.00 1.0034 1.0215 1.0857 1.3344 2.1427
110.00 1.0038 1.0236 1.0941 1.3656 2.2312
120.00 1.0044 1.0278 1.1107 1.4263 2.3980
130.00 1.0057 1.0355 1.4112 1.5348 2.6813
140.00 1.0081 1.0504 1.1995 1.7314 3.1619
150.00 1.0133 1.0832 1.3247 2.1149 4.0309
160.00 1.0285 1.1765 1.6556 2.9806 5.8622
170.00 1.1101 1.6380 2.9379 5.7738 1 1.5213
TABLE4.r / a

ui, an irrotational flow field such that u, +

n = (uf u:) n = 0 on the surface of the
sphere.
ug and t$ are easily determined. u: has velocity potential
Aa5
4; = -sin8 cos8 cash (3.1)
3r9
and uo is given by
uOr= ArsinOcosecosA

Aa5
Vo1 = --
3r4 cos 0 sin A.

) shown that Ui, is the BiotiSavart field of a system of vorticity

Lighthill ( 1 9 5 6 ~has
within the sphere. Corresponding to an element of vorticity with strength
(mir, w ~ B , @ ~dV
A)
at the point ( r ,8, A) external to the sphere, there is an image system of vorticity with
strength U U
(? u l r , -,mle, -;wlA)d~
, A) together with a uniform line vortex of strength
at the image point ( a 2 / r8,

per unit length between the centre of the sphere and the image point. Rewriting this
) terms of vectors, the BioeSavart integral
expression given in Lighthill ( 1 9 5 6 ~in
for u, becomes
‘s
ul(r) = - dr’d8’dh’ r f 2sin 8’F(r,r’),
4.11
(3.3)
 L(ftforce on a spherical body in a rotationalpow 205

and
- 2(0, r’)r‘
0 , = 0,- ?.’2 *

4. Evaluation of u,
The main computational difficulty arises from the evaluation of the three-
dimensional Biot-Savart integral (3.3) for ul(r).The in&grand has singularities at
r’ = r and also as r’+u, or @‘-to,or 8’+x, where (at/apo)e,and consequently o1is
singular.
We shall require the values of ul(r)on the surface of the sphere and also on the axis
where 8 = 0 or x . Consequently, it is convenient to perform the integration with
respect to pk, rather than r‘, for when this is done, all the singularities are on the
surface pi = 0. We rewrite the integral, then, in the form

& Jam dr’ Jon de’ J:x dA’ r f 2sin 8’ F(r,r’)

8 dA’rt2
= d J o w d p ~ J ~ d 2x ’ ~ o sin8

where r’ = r’(& el)

is the positive solution of (2.16),and

Thus, to determine q ( r ) at any point, one must perform a three-dimensional

integration of the vector function F. Some simplification is possible, however, using
the axisymmetry of the system. When r = a or when 8 = 0 or x
v,, = 0;
when A = 0 or x , we also have
VIA =0
and when A = 01
Vl/j = 0.
F(r,r’) is a linear function of m,(r’), so we can write it as
F(r,r’) = wl(r’) G(r,r’) (4.3)
If G is written in terms of the local polar coordinate directions at the point r’, then
G depends upon A and A’ only through the combination A-A’. As ml(t’) can be
written in the form
wli(p6, e’,A’) = wla(p;, e’,0) cos A’ pi,B’, in) sin A’ (4.4)
where i stands for r‘, 8’ or A‘, it follows that
vlr(r,8, A ) = vlt(r,8 , O ) cash + vlt(r,8,in)sin A. (4.5)
206 T . R. Auton
This is a very useful result. Instead of integrating the vector F(r,r’) over r‘ for each
value of 8 and h where ul(a,8,A) is required, one only need integrate the scalar
quantities Fe(r,r’) for each value of 8 with A = 0, and FA(r,r’) for each value of 8
with h = $ and use the following equations:

This greatly reduces the effort required to calculate u, over the surface of the sphere,
The numerical integrations cannot be performed satisfactorily by a standard
numerical integration package because of the singularities of the integrand. Special
techniques were developed for the integration and these will be described in the rest
of this section. For definiteness we shall describe the calculation of v&, 8,O).The
calculation of wlA(a,8,in) and the velocity on the axis is very similar.

4.1. The A’ integration

The integration with respect to A‘ is performed first. We denote the result of this
integration by
jo
1 2n
(3
4 ( r , p i ,8’)= 4R U’ r’z sin 8’ 7 ~ ~ r’,
( el,
r ,A’). (4.7)

The integrand has a sharp peak near A’ = h when po is small. To handle this
integration efficiently and accurately we use a change of variable from A‘ to 6, where
tan (+A’) = c+ tan (M) (4.8)

with
+
r2 r’2 +
-2rr’ cos (8 el)
C=
r2+r‘2-2rr’ cos (8-0’)’ (4.9)

and integrate using an adaptive routine from the NAG library (DOlAJF),which is
designed to integrate functions with singularities, or near-singular behaviour. The
calculations were performed with a requested relative error tolerance of
Generally the accuracy attained was better than this, but for a few values of pi and
el this accuracy could not be attained due to rounding error; however, in these cases
the absolute value of the integral is small and makes a negligible contribution to the
secondary velocity.
4.2. The el integration
When pi is small it is found that q ( r , p i ,el) has a sharp spike around the value of
el a t which
s2 = r2+r‘2-2rr’ cos(0-8’) (4.10)
is minimized with r’ = r’(pi, el). The spikiness of the integrand may be substantially
reduced by the following change of variable:
fI = fI0-a arcsinh (B‘gd),
- (4.11)
where b and S are chosen such that

(4.12)
 Lift force on a spherical body in a rotational flow 207

tz A

B', 5
FIGCRE2. The smoothing of the 8' integrand &(r,p;, 6'') caused by change of variable from 8' to
6, (4.13), T = a, 6'= 67t/6, A = 0, pi = 0.113.

and u and 5, are chosen such that

5 =0 when 8' = 0,
and =R when 8' = A .
With this change of variable the integral is transformed to

(4.13)

Xotice that cosh ([-E0/a)has a minimum a t the place where q ( r ,p i , 6') has a peak.
The smoothing effect of this change is shown in figure 2.
The integration was performed using the same software as the A' integration, and
the same error tolerance, (To be exact, a copy of the routine with different
subroutine and variable names was used, as Fortran is not a recursive language.)
We shall denote the result of the 8' integration by 4 ( r , p i ) .
4.3. The pi integration
The most obvious difficulty in the pi integration is that the range of integration is
infinite. A further complication arises at the stagnation points, where r = a , 8 = 0
or A, as F'(r,pi) becomes infinitely large as pi tends to zero. These difficulties are
handled using a Gauss-Rational quadrature scheme, using the NAG routine DOlBCF
to evaluate the weights and abscissae. The scheme is exact for integrals of the form

(4.14)

where a>0, d-l>c>l

and Pnis a polynomial of order 2n - 1 or less, n being the number of evaluation points.
The values of a , c and d are chosen by the user on the basis of his knowledge of
208 T.R.Auton

1
G
0.01

0.01 0.1 1.o 10.0

P;
FIGURE3. Logarithmic graph of integrand of the pi integrationq ( r ,p i ) when r is, (I),the front and,
(11), the rear stagnation point on the sphere. Notice the singular behaviour as p;+O. The dashed
curves are the asymptotic approximations (4.17) and (4.18) and the dot-dashed curve is (4.19).

the integrand. Except at the stagnation points the asymptotic behaviour of F2(r,pi)
for small and large pi is t
Pf+ O ,
O(1) a s A
a
(4.15)
A Pi
as----foo,
a
and the chosen values of a, c and d are 1, 0 and 2 respectively. A t the stagnation
points the behaviour of F, is singular as pi+O and the numerical values plotted in
figure 3 suggest that
a
(4.16)

a
and corresponding values of c and d are -t and $ respectively. The asymptotic
behaviour of 4 ( r , p i )for small pi can also be obtained from the integral expressions
given by Lighthill (1957, equation (10.13)). The dominant contribution to F2comes
from the vorticity at points very close to T and detailed analysis, not given here, shows
that
(4.17)
at the front stagnation point r = a, 8 = x , and

(4.18)
 Lijt force on a spherical body in a rotational flow 209

e ve(a, 8,O) vA(a,8, t)

Aa Aa
0 -1.1803 1.1803
10 -0.2284 0.5343
20 0.0255 0.3118
30 0.0706 0.1917
40 -0.0152 0.1393
50 -0.1734 0.1376
60 -0.3525 0.1747
70 -0.5079 0.2416
80 -0.6050 0.3302
90 -0.6207 0.4335
100 -0.5451 0.5448
110 -0.3817 0.6583
120 -0.1461 0.7685
130 0.1361 0.8705
140 0.4331 0.9601
150 0.7104 1.0337
160 0.9357 1 .Of383
170 1.0824 1.1220
180 1.1327 1.1327
ve(a,8, A ) = v,(a, 8,O) cosA
vA(a,8, A) = vA(a,8,$) sin h
5. Secondary velocity on sphere surface
TABLE

1 .oooo 1.1327
1.0196 0.9469
1.0824 0.7383
1 .2027 0.5185
1.4142 0.3221
1.8Ooo 0.1704
2.6131 0.07032
5.1258 0.01650
vr(r,x , 0) = 0
vn(r,x , 0) = 0
TABLE
6.Secondary velocity on upstream axis

at the rear stagnation point, r = a, 8 = 0. The constant C is given by

-
C = 33B(g,%) 0.2764,
where B is a beta function (Abramowitz & Stegun 1965, p. 258). The behaviour as
pi+ 00 can also be determined, and

(4.19)

for both the stagnation points. A contribution 7c/64p? comes from the 8-vorticity
as stated by Lighthill (1957, equation 21) and the remaining 4x/64pZ from the
210 T.R.Auton
1.6

1.o

-1.0

-1.6 I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
0 90 10
0 (degrees)
FIQURE 4. Secondary 8-velocities on the surface of the sphere in the plane of symmetry:
v&, 8,O)lAu:. ,
the numerical results of this paper, 0,
Cousins (1969)numerical results, and the
curve + is Hall’s (1956)approximate analytical results.
1.5

1
I

1.o

0.5

.,
1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

0 (degrees)
FIQURE
5. Secondary A-velocities on the surface of the sphere: v A ( a , 8 , ~ n ) / A a : the numerical
results of this paper; 0, those of Cousins (1969).
 21 1

0 I I I I

1 2 3 4 5 6
rla

FIGURE6. Downwaah on upstream axis times T a l a s : (v0(r, x , O)/Aa)(P/a*): D,numerical results of

this paper ; + , Hall’s (1966) approximation,and the daahed lines are the aaymptotic limits of the
two graphs.

FIQURE
7. Diagram showing directions and magnitudes of secondary velocities in the oncoming
shear flow and on the sphere surface.

A-vorticity. These asymptotic formulae are also plotted on figure 3 and the close
agreement between the asymptotic and numerical results provides an independent
check upon the calculation.
The pi integrations were performed with 32, or in some cases 64, points of
evaluation.
The calculated values of V e and v, (including v,,) on the surface of the sphere, and
the downwash, ve on the upstream axis are given in tables 5 and 6.As stated above,
the A’ and 0’ integrations are accurate to at least four significant figures. Halving
the number of integration points used for the pk integration gives results that differ
from those of tables 5 and 6 by less than The values given in the tables are
believed to be accurate to at least three significant figures.
212 T.R.Auton
These results are plotted on figures 4, 5 and 6. To give a clearer picture, the
magnitude and direction of the secondary velocity at certain points on the sphere
surface and on the upstream axis is shown on figure 7.

5. The lift force on the sphere

By Bernoulli’s theorem, neglecting gravity,
h = p+lglu12
is constant along the streamlines. Therefore, on the surface of the sphere the pressure
is
P = Po+lgl~o12-~l~12~ (5.1)
where p, and uo are the pressure and velocity far upstream on the stagnation
streamline.
To first order in the oncoming vorticity, A,

and the lift force on sphere is therefore

=p 6 s,””
d8 dh a2 sin O#uo sin 8w&, 8,O) cos An, (5.3)

where S is the surface of sphere and (4.6)has been used for ve(a,8,A).
In Cartesian coordinates

sin 8 sin h
By inspection Fyis the only non-zero component of FL and

y 2 0
s‘
F = -3K p * a2 uo d8 sin38 w,(a, 8,O)
(5.4)
A,
x 2.09454p*asuo
and in vector notation
F’ = CLp*$u3u0 A w,,

CL x 0.500035. I
This result is obtained by numerical integration using Simpson’s rule and the values
(5.5)

of given in table 5.
This result is used in developing a general expression for the lift force and the total
force acting on a small sphere in an arbitrary rotational straining flow, by Auton
et al. (1987).
 Lifl force on a spherical body in a rotational flow 213

6. An analytical evaluation of the lift coefficient

The reader will have noticed that the computed value of the lift coefficient is
remarkably close to +.I n this section we present an analytical argument to prove that
the value is exactly t.
Applying the momentum-integral theorem to a volume surrounding the sphere we
find that L

4
-F - @*(u*n)u+pn)dS,

where S is the external surface of the volume of integration. It is convenient to choose

for S a deformed cylinder, whose curved surface is parallel to the streamlines of the
primary flow, and bounded axially by circular disks of radius Z centred on the x-axis
at positions z = & X (see figure 8). Z and X are chosen such that
U
a<Z<-,
A
X%Z. (6.3)
Using Bernoulli’s theorem we can expand p in terms of the velocity
p = po+14p*(U-m)e-1 *
4p (u)z, (6.4)
where u-O0 is the velocity far upstream on the same streamline and po is the uniform
upstream pressure. The total velocity is made up of the irrotational primary flow V,
proportional to u, and a secondary velocity uyproportional to A. The terms involving
only Vcontribute nothing to the total integral as there is no net force due to a uniform
streaming motion, the interaction of Vand t) leads to a force in the y-direction, which
is proportional to A. I n the following analysis we shall consider only those terms that
are first order in A and only the y-component of the force integral.
Consider now the far field of u. I n the wake, far downstream of the sphere, the
vorticity is aligned in the z-direction, with
dX
o, = A sinh-
do
(Lighthill 19563) where
X = Lim uot(z,p).
z-*m

This vorticity induces a flow in the (y,2)-plane, given by

w, =
dp’X(p’) -t( 1 -cos 2h) X@)
I
independent of x. Outside the wake region, the secondary flow far from the sphere
is given by
-
u Ap coshe,+-(-p
AaS
2rS (6.7)

(Lighthill 19563). We can now evaluate the leading-order terms in the force integral.
214 T.R. Auton

FIQURE
8. Sketch showing volume of integration used in the analysis of $6.

On the upstream disk all terms of the integral vanish asymptotically, as z/u+ co .
On the curved surface S , the leading-order term is
r
") v-" ds,
p* Js, - lzy (u - 1)-

where V-" = uoex (6.9)

and v- O0 = Ap- ex cos A (6.10)
are the asymptotic limits of V and 0, and p-" of p, as z+.- 00 along a streamline.
To leading order, from (6.7) and (6.10)
(v- u - y x
Apa3 cos h
- 6r3
(6.11)

as P-P-" - u3
@P' (6.12)

To leading order we can approximate p by Z,and the integral becomes

p* -X
&J,' c dA( -cos A)
Au3Z COB hu,
+
6(Z2 z2)f
On the downstream disk S, the leading-order term is

(6.14)

and substituting for v,, its asymptotic form in the far downstream wake region (6.6),
this becomes

P* JozP d~ JO2' auo (- Acos2h pa

P
Jo p' dp'X@') -#A( 1-cos 2 A ) X@))

--&,Ap*JmpdpX(p) =-&,Ap*Vn =-$a3p*uoA, (6.15)

01

where V, = #V is the drift volume associated with the primary flow (Darwin 1953).
Putting the separate contributions together, we conclude that
F, = -ina3p*uo Ae, (6.16)
and c, = a. (6.17)

7. Discussion
I n this paper Lighthill's secondary-flow method is used to expose features of the
flow around a stationary sphere placed in a simple shear flow. The secondary flow
on the sphere surface has been calculated numerically, and used to estimate the
resulting lift force on the sphere.
 Lijl force on a spherical body in a rotational $flow 215
Lighthill’s work is mainly concerned with calculating the secondary velocity on
the upstream axis, which he calls the downwash, D(r),and the displacement of the
stagnation streamline 6; 6 is used to correct the results of Pitot-tube velocity
measurements from sheared velocity fields. In our notation
D(r) = V e ( r ,X , 0). (7.1)
Wherever possible, the results of this paper have been compared with Lighthill’s and
this has provided a check upon the analysis. This comparison has shown a couple
of minor errors in Lighthill (1957, hereinafter denoted L).
In the final, pi, integration for D(r),Lighthill claims that the integrand De(r,p;)
tends to zero as pi+O (L, p. 500). (The suffix 8 here denotes the BioeSavart
contribution to D(r)from u18 - he obtains the contribution from ulhanalytically and
D,(r,p;)+O as claimed.) This is only true, however, when T > a ; at r = a careful
asymptotic analysis of the expression given by Lighthill (L, equation 13) shows that
D(a,p i ) - ~ p i - t as pi +0. (7.2)
This result is confirmed by the numerical results shown in figure 3. Consequently
Lighthill’s value for D(a) of 0.97Aa is too small by 15% compared with our result
of 1.133Aa. A t larger values of r Lighthill’s results are confirmed: he gives

D ( d 2 a )= 0.33Aa, (7.3)
compared with 0.322Aa found in this paper, and Lighthill’s asymptotic result (L,
equation (22))

is also confirmed by our numerical results.

The displacement of the stagnation streamline is given by (L, equations (31),(32))
=
u0
J-m
a
D(r)dr = iJ* D(a coseca)
(1--3/rs)i uo (1-sinaa)f
cosec a cot a da, (7.5)

where a = arcsin a/r. Lighthill estimates 6 using Simpson’srule, with interval i x and
the second integral above, and finds
Aas x
6 X --{(isi+7.0360(1/2a)+#D(a)).
uo 12
However, the coefficient of D(a) should be $ here and Lighthill’s estimate of D(a)
is itself too small. The effect upon 6 is quite small, though, and Lighthill’s value of
0.89Aaa/uois quite close to our estimate, based upon the figures in table 6 , of
0.938Aa3/uo;his value of 6 is accurate to one decimal place, despite these errors,
which is all the accuracy he claims.
A similar calculation to that given here has also been made by Cousins (1969,
1970). He used Lighthill’s theory to calculate numerically the secondary velocities
over a portion of the forward surface of the sphere, where 8 > 120O. He claims an
accuracy of one decimal place in his values of tr but makes no mention of any special
procedures being used because of singularities of the Biot-Savart integrand. The
accuracy of his results must be questioned as he obtains the same, erroneous, value
as Lighthill, 0.97Aa, for the downwash at the front stagnation point. He does not
appear to have noticed the trigonometric relations (4.6), which would have
216 T.R.Auton
substantially reduced his computational labours. In terms of Cartesian coordinates,
as used by Cousins, the relations (4.6) imply

1
v J r , 6, A) = vJr, 6,O)cos A,
v J r , 6, A ) = v J r , 6,O)cos2A +v,(r, 6,in)sin2A, (7.7)
v,(r, 6 , A ) = v,(r, O , $ ) sin2A.
Alternatively, these relations can be derived quite easily from the expressions given
in Appendix I of Cousins’ (1969)paper, but they are not satisfied by Cousins’ results,
even allowing an error of kO.1 on each figure. For comparison some of his results
are plotted on figures 4 and 5.
Hall (1956)has found an approximation to the secondary velocity in the plane of
symmetry. He assumes that the two-dimensional divergence of v in this plane is zero
so that
-+-
av, avu = 0.
ax ay
This is only an approximation, as av,/az =+ 0 in general. The curl of the velocity in
this plane is

with w, the same as that used in this paper,

-A
w, = (7.10)
(1-a3/r3)
With this approximation an analytic solution in the plane of symmetry can be found,
viz.
-(---)
2= 1 r a3 sin28
Aa 2 a r3
(7.11)

For comparison with our results these velocities are also plotted on figures 4 and 6.
Hall’s method generally overestimates the downwash and, hence the displacement
6.Using Hall’s theory one finds
@ x 1.24 (7.12)
Aa2
which is 30 yo larger than our result. On the surface of the sphere, Hall’s values of
ve are surprisingly close to ours for 6 > in,but on the downstream face of the sphere,
the other vorticity components, w,. and wo, neglected in Hall’s theory, become more
significant and the results diverge. The lift coefficient, estimated using Hall’s values
for ve,is 0.375 and is about 30 % smaller than the result of this paper.
The numerical value for the lift coefficient is confirmed by a more elegant analytical
argument, using the momentum theorem. It is instructive to reiterate the assump-
tions underlying this analysis. The momentum theorem is applied to a control volume
surrounding the sphere, which is a slightly deformed cylinder, the generators being
streamlines of the primary flow. An approximation to the momentum integral is then
found, using far-field results for the secondary flow, published by Lighthill 1956b,
in the case that the cylinder diameter is much larger than the sphere diameter, and
 Lift force on a spherical body in a rotational flow 217
its length much greater than its diameter. However, as Lighthill (1957) has pointed
out the approximations to the far-field secondary velocity are only valid when
a 4r 4a/A.
The expression for the lift force is thus found as an approximation to a momentum
integral over a finite volume, and should be viewed as the first term in an asymptotic
expansion in A.
To obtain this result, it is necessary to assume that the cylinder is much longer
than its diameter. This is analogous to Darwin’s (1953) observation that, to calculate
the drift volume of the sphere moving through an infinite fluid at rest, it is necessary
to take limits in the %-directionand in the p-direction, in that order. This same point
has recently been reiterated by Benjamin (1986). It is interesting to note, however
that the integral for FLhas leading-order contributions from the curved surface S, and
from the downstream disk S,, whilst the drift volume has contributions only from
the upstream and downstream disks.
In a later paper (Auton et al. 1987) it will be shown how the result found here for
FLmay be generalized to more complex, straining and time-dependent inviscid flows.
In a real fluid, viscosity will act to cause a drag force opposing the relative motion
of the fluid, and may cause separation of the flow from the rear of the sphere. The
effect of flow separation is difficult to determine, but is likely to be less significant
for bodies with mobile surfaces, such as bubbles and drops, than for rigid particles.
The effects of finite drag on the motion of bubbles is discussed in Thomas et al.
(1983), and some of the experimental data on the forces on spherical bodies is reviewed
in Auton (1984).

This work was supported by the Science and Engineering Research Council and
the Central Electricity Research Laboratory under a CASE studentship. I wish to
acknowledge my debt of gratitude to my Ph.D. supervisors, Dr J. C. R. Hunt and
Dr N. H. Thomas, for much helpful guidance and encouragement and to the referees
of earlier drafts of this paper for their comments.

REFERENCES
ABRAMOWITZ, M. & STEQUN, I. A. 1965 Handbook of Mathematid F u n c t i m . National Bureau of
Standards.
AUTON, T. R. 1984 The dynamics of bubbles drops and particles in motion in liquids. Ph.D. thesis,
University of Cambridge.
AUTON,T. R., HUNT,J. C. R. & PRUD’HOMME, M. 1987 On the motion of a circular cylinder and
a sphere in inviscid rotational flow. J . Fluid Mech. (submitted).
BATCHELOR, G. K. 1967 An Introduction to Fluid Dynamics. Cambridge University Press.
BENJAMIN, T. B. 1986 Note on added mass and drift. J . Fluid Mech. 169,251.
COUSINS, R. R. 1969 Shear flow paat a sphere. N P L rep. MA80.
COUSINS, R. R. 1970 A note on the shear flow paat a sphere. J . Fluid Mech. 40,543.
DARWIN, C. 1953 Note on hydrodynamics. Proc. Camb. Phil. Soc. 49, 342.
HALL, I. M. 1956 The displacement effect of a sphere in a two-dimensional flow. J . Fluid Mech.
1, 142.
LIQHTHILL, M. J. 1956a The image system of a vortex element in a rigid sphere. Proc. Camb. Phil.
Soc. 32, 31.
LIQHTHILL, M. J. 1956b Drift. J . Fluid Mech. 1, 31 (and Corrigendum 2, 311).
LIQHTHILL, M. J. 1957 Contributions to the theory of the Pitot tube displacement effect. J . Fluid
Mech. 2 , 493.
218 T. R. Auton
TAYLOR, G. I. 1928 The forces on a body placed in a curved or converging stream of fluid. Proc.
R. Soc. Lond. A70,260.
THOMAS, N. H., ATJTON, T. R., SENE,K. J. & HUNT,J. C. R. 1983 Entrapment and transport of
bubbles by transient large eddies in multiphme turbulent shear flows. BHRA Intl Conf. on the
Phy.si& Modelling of Mdtiphasejkws, Coventry, April 1983 , p. 169. (Also 1984 Entrapment
and transport of bubbles by plunging water. In aaSTransfer at Water Surfaces [ed.W. Brutsaert
& G. H. Jirka], p. 255. Reidel.)
TOUMIEN,W. 1938 Uber Krafte und Momente in schwaoh gekrummten oder konvergenten
Stromungen. Ing.-Arch. 9, 308.