Ellipsoidal Particles Immersed Viscous Fluid.

161

ratio is 8 per cent, a t ordinary tem peratures which rem ains constant till
about 120°, and then falls off, rapidly a t first and then slowly to about
1-2 per cent, a t the critical point, There is no ehauge of imperfection of
polarisation on passing through the critical point. The correction due to this
in the expression for the intensity of the scattered light is given.

In conclusion, I have great pleasure in recording my indebtedness to.

Prof. C. V. Raman who suggested the research, and who continued to take an
inspiring interest in its progress. The experim ental work was carried out in
the Physical Laboratory of the Indian Association for the Cultivation of
Science, Calcutta.

The Motion o f Ellipsoidal Particles Immersed a Viscous Fluid.

By G. B. J e f f e r y , M.A., D.Sc., Fellow of U niversity College, London.

(Communicated by Prof. L. N. G. Filon, F.R.S. Received January 30, 1922.)

1 1. Introduction.
In both physical and biological science, we are often concerned w ith the
properties of a fluid, or plasma, in which small particles or corpuscles are
suspended and carried about by the motion of the fluid. The presence of the
particles will influence the properties of the suspension in bulk, and, in
particular, its viscosity will be increased. The most complete m athem atical
treatm ent of the problem, from this point of view, has been th a t given by
Einstein,* who considered the case of spherical particles and gave a simple
formula for the increase in the viscosity. W e have extended this work to
the case of particles of ellipsoidal shape.
The second section of the paper is occupied w ith the requisite solution of
the equations of motion of the fluid. The problem of the motion of a viscous
fluid, due to an ellipsoid moving through it w ith a small velocity of transla­
tion in a direction parallel to one of its axes, has been solved by Oberbeck,f
and the corresponding problem for an ellipsoid rotating about one of its axes
by Edwards.^ In both cases the equations of motion are approxim ated by
neglecting the terms involving the squares of the velocities. I t may be seen,

* “ Eine neue Bestimmung der Molekiildimensionen,” ‘ Ann. d. Physik,’ vol. 19, p. 289
(1896) ; with a correction in vol. 34, p. 591 (1911).
t 4 Crelle,5 vol. 81 (1876).
+ ‘ Quart. Jour. Math.,’ vol. 26 (1892).
 1G2 Dr. G. B. Jeffery. TheMotion o
a posteriori, that the condition for the validity of this approximation is th at
the product of the velocity of the ellipsoid by its linear dimensions shall be
small compared with the “ kinem atic coefficient, of viscosity ” of the fluid.
Xn relation to our present problem, it will therefore be satisfied either for
sufficiently slow motions, or for sufficiently small particles.
The third section is devoted to the consideration of the forces acting upon
the particle as a whole. I t is found th a t these reduce to two couples, one
tending to m ake the particle adopt the same rotation as the surrounding fluid,
and the other tending to set the particle with its axes parallel to the
principal axes of distortion of the surrounding fluid.
In the fourth section we investigate the motion of a particle which is
subject to no forces except those arising from the pressure of the fluid on its
surface. I t is found th a t there is a variety of possible motions corresponding
to different initial conditions. They are periodic and of a general character
somewhat like the motion of a top. There is thus a certain degree of inde­
term inacy in the problem as treated by the approxim ation which neglects the
squares of the velocities.
Sections five and six are concerned w ith the dissipation of energy and with
the viscosity of suspensions. I t is found th at the increase of viscosity may
still be represented by a formula of E instein’s type with a modified num erical
factor. Owing to the indeterm inacy revealed in section four, it is not possible
to give a definite value for this num erical factor. I t is, however, possible to
specify an upper and a lower lim it for this factor, and the difference between
these lim its diminishes to zero as the particles approach a spherical shape.
Section seven is devoted to a tentative suggestion for the removal of this
indeterminacy. I t is, th at when a variety of motions is possible for given
boundary conditions under the approxim ated equations of motion, the actual
motion will be th at which corresponds to minimum dissipation of energy. If
this hypothesis be granted, it is possible to state many of the results of the
paper with greater precision.

§ 2. The Solution of theEquations of Motion.

Consider, in the first instance, the modification produced in the motion of
a viscous fluid by the presence of a single ellipsoidal particle. W e shall
assume that, apart from the disturbance produced in the immediate neighbour­
hood of the particle, the motion is steady, and varies in space on a scale which
is large compared with the dimensions of the particle. I t appears from the
work of Oberbeck that, if the fluid at a distance from the particle has a
velocity of translation relative to the particle, the latter is acted upon by a
resultant force. I t will appear in the course of the investigation th at any
 Particles Immersed in a Viscous Fluid. 163

other motion of the fluid can only give rise to a resu ltan t couple. The
particle will therefore ultim ately assume the velocity of translation appropriate
to th a t p a rt of the fluid which it displaces. On the above assum ptions this
velocity of translation will be sensibly uniform, and we may therefore take
the centre of the particle as a t rest.
Let x, y, zbe rectangular co-ordinates referred to axes fixed in the particle
and moving w ith it, and let the surface of the particle be the ellipsoid
a2+ y2 jb2+ / e2 = 1. (1)
The undisturbed motion of the fluid in the neighbourhood of the particle is
then given by
Mo = a:r + h y + gz + y
vo = h r + b y + X
Wo = g.r + f y + GZ+ %y—r}X j
where a, b, C, f, g, h , £, y,£ are the com ponents of distortion and
the fluid. These are taken to be constant in space through a volume which
is large compared w ith the dimensions of the particle. They will, however,
vary w ith the time, since the particle will ro tate under th e influence of the
fluid, and the axes of x, y, z will be rotating axes.
Let the spins of the ellipsoid about its axes be &>i, &>2, Take a set of
axes, x ', ,y's', in fixed directions w ith their origin a t the centre of the particle,
and let the components of distortion and rotation of the fluid in its
undisturbed motion referred to these axes be a ', b ', c ' , .... These are
constant, but the corresponding unaccented quantities are linear functions of
these, w ith coefficients which depend upon the direction cosines of the two
sets of axes, and which therefore vary w ith the tim e in consequence of
(Ol, 0)2, 0)3.
Neglecting squares and products of the velocities, the equations of motion
of an incompressible viscous fluid referred to moving axes are of the type
du
•7^ —W3V+ (o^w

where p, p,, p are respectively the density, coefficient of viscosity, and mean
pressure of the fluid.
The spins on, 0)2, 0)3 are produced by the motion of the fluid, and, as will
appear explicitly later, they are of the same order as the velocities. Hence to
our order of approximation all products like co^v may be neglected. I t is not
quite so obvious th at the remaining term s on the right-hand side of (3) are of
the second order. I t will appear later th a t tt, v, w are homogeneous linear
functions of a , b , c, ..., with constant coefficients. They are, therefore,
homogeneous linear functions of the constants a ', b ', c ', ..., with coefficients
164 Dr. G. B. Jeffery. The Motion o f Ellipsoidal
which are functions of the direction cosines of the moving axes. Hence the
only term s which can arise in ?/idtare of the type ana
second order of small (quantities, and is therefore to be neglected.
The equations of motion therefore reduce to
r fu = w
= I-

with the equation of continuity

du j d o , dw _ ~
G)

W e have to find a solution of (4) and (5) which agrees w ith (2) a t great
distances from the origin, and which gives on the surface of the ellipsoid
— (OzZ—G&zy, V = <03*— V) — (6)
As usual, let X. denote the positive root of
x2 z2
■+ . - + • 1, G)
a 2+ X b2+ X+ X
and A = {(a2+ X)(&2+ X)(c2+ X )p , (3)
and let

Jx ( 2+ X) A (b2+ \ ) A’ U (e2+ X) A' (9)

i:
We will also write
Too
dX XdX
( 10)
L (b2+ X) (c2a.+(b2
X)+AX)
’ (c2 + X) A
with symmetrical integrals for so th at
a' - 7 ~ f t >>_ t? ft-(? y
( 11)
V2— b2—c2
We will denote the corresponding integrals, in which the lower lim it of
integration has been replaced by zero, by fto ', __
W e have the following well-known solutions of Laplace’s eq u atio n :—
a;2 22 dX
•+ ( 12)
a2-f- X d2 + X A
and
X l - a'yz,X2 = ft'zx, Xz = y'xy. (13)
We shall seek a solution in which u,w are exp
of the first and second differential coefficients of XI and the first differential
coefficients of %i, %2>%3- These latter can, of course, be expressed in term s of
the second differential coefficients of XI, but as the corresponding terms enter
into the solution in a somewhat different way, it will be convenient to keep
them separate.
 Particles Immersed a Viscous Fluid. 165

From (7) we have

_ 2#P2 3\ 2;yP2 3\ 2zP2
(14)
3r a2+ X.’ By b2+ X' 3T c2+ A.
where
1 z2
(15)
P2 (a2.+ X f + (P + X )2'1'
W e then have a num ber of formulas of which the following are ty p ic a l:
0n _ 3^2
32I2 =
_ ^
o. 4a;2P2
4 r2P 2 32H = _ 4yzP2
3a?
= 2a&’,
tta2 X (a2+ X)2A' 3 y3 z (52+ X) (A
. (16)
and
a%i _ <h& _ a^.3 _ 2xyzP2
3a; 3y 3z
2 fz P 2 dXi 2yz2P2
(17)
ii

= *V
(b2+ X)2(c2+ X)A ’ 3: (b2+ X)(c2+ X)2&'
All these differential coefficients tend to zero a t infinity.
W e assume
v 3X2
Uo + 3- (% 1 + s^2 + rl %3) + W 3 & _
ay 3?

+ A ( a ; ^ - ^ ) + H (a;
32X2 3n\ 32n 3n
—) 4- Gr' (£C
dx* t y ) 3®3:
Sx) + t i {X 3
, / „ , ? n i , u oKl , ,, 32!! \
+ !/(H <fc» + B dx?,ij+ F 3i;3:j

+ s (g ^ + F ' | A + C I - 2-), (18)

\ dar dte dy d&azj

v„ + t (% 1 + S*s + TXS) + U ^ -■W ^G

+ x{ A -H ' Q i n , 32P \
1 dxdy * /2 ' 3y

+ H' y
' 32n 3q lT) / 32q 3 fi\ , _ / 3212 3n
, dxd y 3k + B ( ^ - W ) + F l y a ^ '
1 | f ^ + F'-3^q , p 32q \ (19)
3?/2 ^ " BycS/’
and
w = Wo+ ^ ( r %1+s %2+ t %3) + v
3y
166 Dr. G. B. Jeffery. TheMotion of Ellipsoida
I t may be seen by inspection th a t these values of u, v, satisfy (5), since
a + b + C = 0 for an incompressible fluid, and Q and %i, %2, %3 satisfy
Laplace’s equation.
Substituting these expressions into (4), it is easily seen th at these are
satisfied provided that

, - * + * { A g + B « + 0 g + (F + D g “

( 21)
+ <g + g ' > H + (h + h ' > H } '

where p 0 is the constant mean pressure at a distance from the ellipsoid. I t

therefore remains to satisfy the conditions (6) on the surface of the ellipsoid
X = 0. Substituting from (16) and (17) into (18), (19), (20), we have

u = £ { a + 7 'W - £ ,V - 2 ( a + /3 + Y)A}
+ y { h - r + 7 /T - 2 /3 H + 2«H '}

+ z { g + ri+/3' S - 2 7G' + 2«G}

2%P2
[R +2 (b2+ \ ) F + 2 (c2+ X) F'} yz/(b2+ X) (c2 + X)
(a2-j- X) A

+ (S + 2 (c2 + X) G + 2 (a2+ X) G '} (a2+ X)

+ {T + 2(«2+ X )H + 2(52+ X ) H '} ^ / ( a 2+ X) (62+ X)

+ {W - 2 (a2 + X) A + 2 (b2+ X) B } + X)2

- {V - 2 (c2+ X) C + 2 (a2+ X) A} + X)2 , (22)

v = # {h -b £+ 7 'T + 2/3H —2 « H '}

-\-y{b + a’U —<y'W—2 (a + /3 + 7)B}
+ 2 {f - f + a 'R - 27F + 2/3F'}
2yP2
{R + 2 (J2+ X) F + 2 (e2+ X) F '} yzj(b2+ X) (c2+ X)
G2+X ) A

+ {S + 2 (c2 + X) G + 2 (a2+ X) G ’} X) (a2+ X)

+ {T + 2 (a 2+ X )H + 2 G 2+ X ) H '} ^ / ( a 2+ X)(&2+ X)
+ {U —2 (52+ X) B + 2 (c2+ X) C} z2/(c2+ X)2

- {W - 2 (a2+ X) A + 2 (62+ X) B}x2/(a2+ X)2 , (23)

 Particles Immersed in a Viscous Fluid. 167

and
w = £{g— rj + fi 'S —2«G + 27G'}
+ y {f -\-1 + « 'R + 27F - 2/3F'}
+ z { c + / 3 'V - x 'U - 2 ( a + /3 + y) C}
2zB2 {B + 2(b2+ A)B + 2(c2 + A )F } ijz/(b 2+ A)(c2+ A)
(c2+ A)A
+ {S -f 2 (c2+ ) r + 2 (a2+ A ) G '} / ( c -f
A
G (a2+
+ {T + 2 (a2+ A) H + 2 (b2+ X) H '} xy\(a 2+ A) ( + A)
+ { \ - 2 ( c 2+ A)G + 2(a2+ A )A } x 2/(a 2+ A)2
- { U - 2 ( b 2+ A )B + 2(c2+ A )C } if/(b 2+ A)2 . (24)

P utting A —0, identifying w ith (6), and equating coefficients, we obtain

equations which determ ine the constants. On solution we find their values
to be as follows :—
A = i{ 2 a o //a - A //b - 7 o ,,c } / ( A /V / + 7o,'«o'/ + «o/W ,)-l
B = i{ 2 /3 o ,/b - 7o,/C - « o ,/a } / ( / 3 o 'V , + 7 o /,«o,' + «o/W /) l , (25)
C = £ {27o,/C—ao,/a — @o"b}/(/3o"j
p — /^of—c2«o(£—Q>l) 7 o f+ 5 2« o (| —&)i) >
F'
2«0' (62A> + e27o) ’ 2«0' (&2/5o+ c27o)
q _ 7og —u.2j80 (77—02) »og + c2/30 (7; —co2)
G' s (26)
T 2/3oV 7o + «2«oV 2/30' ( c27 o+ a2
H = a»4l —&27o(£—&>3) H' /3oh + a~7<» (£—^3)
2y0, («2«o + & W ’ 270' (o2«0 + b2/3()) J

E = S = - g / / S 0', T = - h / 7o'. (27)

The values of U, Y, W will not be needed for our purpose. They are
given by
U = 252B —2e2C ^
Y = 2fi2C —2a2A j>. (28)
W = 2a2A —2 J
Thus the constants in (18), (19), and (20) are all uniquely determ ined and
the velocity of the fluid is known at all points.
For later use we note the following relatio n s:—-
A + B + C = 0, (29)
U + Y + W = 0, (30)
and, using the relation a + b + c = 0,
A a + B b + Cc = i («o//a 2+ ^o,,b 2+ 7 o " c 2)/(/8o/V , + 7o//«o" + «o///eo//)- (31)
168 Dr. G. B. Jeffery. TheMotion of Ellipsoid

§ 3. The Resultant Couple on the Particle.

For an incompressible fluid the stresses are given, with the usual nota­
tion, by
(CX = -p + 2^ |j , yz = y, ( | ? ■+1 ) . (32)

The calculation of the values of these stresses on the surface of the

ellipsoid is greatly simplified by noting th a t most of the brackets in (22),
(23), and (24) vanish for X = 0, and th at, therefore, the only term s surviving
in the surface values of the differential coefficients of the velocity compo­
nents are those arising from the differentiation of the brackets.
If we denote the components of force per unit area acting on the surface of
the ellipsoid by X, Y, Z, we have
—N ^ "X .>—N
X {xxx/ct2+ xy y1

Y {yx x /a 2 + yy -yf- yz zjc2} (33)

N ^ X ✓ N
Z { zx x j a 2+ zy y/b2+ zz

The reduction is straightforw ard and, after simplifying as far as possible by

means of the relations between the constants, we obtain
8/xP
x = abc { > 3 + h £ + G ' 1 ] - —4/^P (ao A + /3o B + 70C)
pH

PM

> —4/^P («0A + /3o B 4* 70 C)

II
1
0

abc
"“fed **
CSJ

1
II

> —4/zP ( acqA + /3oB + 7 oC)

+
o

abc l cf * + r h + G P J c
(34)
Since, apart from P, which is even in x, y, only first powers of the
co-ordinates appear in these expressions, they will separately vanish on
integration over the surface of the ellipsoid. Hence there is no resultant
force acting on the particle. Denoting the components of the resultant
couple acting on the particle by L, M, X, we have on integration over the
surface of the ellipsoid
L = jj (yZ -

which on substitution from (34), om itting terms which obviously vanish on

integration, gives
 Particles Immersed in a Viscous 169

This integral is readily evaluated in term s of the volume of the ellipsoid,

and we have from this and two sim ilar expressions
L = ^ 7 r At(F / - F ) , M = fi(G ' —G),
T N = ^ -7 ry a (
S ubstituting from (26), these give
167Tfl
L {(&3- c 2)f+(& a+ c3)(£-<>n)}>
S(b2/30+ c2y0)

M _ ^ _ {(c2_ a 3 ) g + ( t , + a 2 ) ( , _ m3)}> j- (36)

N
3 ( a w V ) {(tt3- y ) h + (<,3+ia)(?~ M3)}-
The resultant couple acting upon the ellipsoid m ay thus be regarded as
consisting of two p a r ts : one which vanishes when the ellipsoid has the same
resultant spin as the fluid, and one which vanishes when the axes of the
ellipsoid coincide w ith the principal axes of distortion of the fluid.

§ 4. TheMotion of the Particle— Laminar Motion.

If the particle is subject to no forces except those exerted by the fluid
upon its surface, then in the slow motions to which our assum ptions have
already restricted us, the resu ltan t couple on the particle m ust vanish at
every instant. Hence, from (36),
(b2 c2) wi =
(c2-}- a2) o)2 = c2(tj + g) g) j>,

or, from (35),

(a2+ b2)g>3 = a3( £ + h ) + 52( £ - h ) J
F ' = F, G ' = G, H ' = H. (38)
These relations will simplify many of our formulae and, in particular, we
may note th a t (26) are replaced by
F = ^ (i — &h= k
2«o '(b2+ c2)’2/3o, (c3+ «3)’ 27 o V

I t has not been found possible to obtain a solution of equations (37) which
would give the motion of the particle for the most general possible undis­
turbed motion of the fluid. Some light may, however, be throw n on the
problem by the consideration of a particular, but im portant, case.
Take axes x ', y ’, ?J,fixed in directio
velocity in the undisturbed motion be , w f parallel to these axes.
Consider the “ laminar motion ” given by u f = = 0, u f = rcy\ where k is
a constant. Let the direction cosines of the axes of the ellipsoid ( ., the
axes ,xy, )z referred to the axes of x', ?/, be (h, ?nh Wi), ( n2)
170 Dr. G. B. Jeffery. The Motion of Ellipsoidal

(Is, ms, nz) respectively. Then employing the usual formulae for the tra n s­
formation of axes, we obtain
cl — , b = Km27l2,C =
f = ±k (m2nz + m zii2), g — \ tc (mZ7h + mi%3), h = k (m\n24- m2nx),

£= \ k (m2nz—mzn2) , y = |/c (w 3?ii —wii

and equations (37) become

= K ( b2m2nz —
c2m zn2)
(e2+ <w2 = /C (c2mzn\ —a2vi\ns) (41)
(a2+ &2)6>3 = /C (a2m\n2 —b2m2n\)

These may be regarded as the equations of motion of the ellipsoid and give
its spins for any position of its axes. I t may be seen from these th at a
possible type of motion is th a t in which any one of the three axes of the
ellipsoid lies perm anently along the axis of i.e., perpendicular both to the
direction of the velocity and to th a t of the velocity gradient in the undis­
turbed motion, while the ellipsoid rotates about this axis with a variable spin.
A typical motion of this class is th a t for which

l\,mi, nx —1, 0, 0 ;
k,mz, nz = 0, —sin %, cos y ; &>i = %; eo2 = &)3 = 0. (42)>

Equations (41) then give

( b4-
2 c2) y = kb2
( cos2 %
or, om itting a4constant of integration which is obviously immaterial,
, b, bcKt /10,
tan x = - ten (43)

The equations readily adm it of integration in the case of an ellipsoid of

revolution. Let b= c,and introduce the three E ule
between the axes of x and x ', <£ the angle between the planes
f t the angle between the planes of x'x and xy. Then

6)1 = 0 cos 8+ ft, co2= 8in

s ft— ij)sin
and
mi — sin 8cos (f), 7>i = sin sin <£ "
m2 = —sin (f)sin
n2= cos (f>sin ft + cos 8 sin <f>cos f t
mz = —sin (f) cos f t —cos 8 cos <f>sin f t
nz = cos <f>cos
 Particles Immersed a Fluid. 171

S ubstituting in (41), we have after some reduction,

</>cos 6= \ k cos 0, (45)
ip} Q— K(
a2— sin 6 cos 6 sin <£ cos <£, (46)
(a2 + b2) <jb = ka3
( cos3 (f) + b2 sin3 <£). (47)
From (47) we have, if t = 0 when <£ = 0,
. ft , Kabt
t o n * = j t a n o> + J, (48)

and dividing (46) by (47) and integrating,

f nn2 a%2 g —a u (49)
k2 (a2cos2 + sin2 7
where k is a constant of integration.
I t appears th a t the motion of the ellipsoid is periodic. Its axis of revolu­
tion describes a cone about the perpendicular to the plane of the undisturbed
motion, to which its inclination varies between the lim its tan and
ta n -1 ( b/k.)
This statem ent requires modification in the degenerate cases of a disc and a
th in rod respectively. P u ttin g a= 0 in equations (4
find th a t a disc can take any position in which its faces are entirely composed
of stream-lines of the undisturbed motion of the fluid. I t thus moves through
th e fluid edge-on with its axis at any inclination to the fluid shear. If we
put 1)= c.—0 in the same equations we find w hat is otherwise obvious, th at a
thin rod can set itself in any direction in a plane which is a plane of constant
velocity for the undisturbed motion of the fluid. In w hat follows, however,
we shall regard rods and discs as spheroids of small bu t finite thickness.
This investigation reveals no tendency on the p art of the ellipsoid to set its
axis in any particular direction with regard to the undisturbed motion of the
fluid. The motions corresponding to all values of are consistent with the
boundary conditions and with the approxim ated equations of motion.

§ 5. The Dissipation of Energy.

As the velocity of the fluid is finite a t great distances from the ellipsoid,
the total rate of dissipation of energy will be infinite. W e will accordingly
calculate the excess of the rate of dissipation of energy over th at which obtains
in the same motion of the fluid in the absence of the ellipsoidal particle.
Difficulties which arise in connection w ith the boundary conditions, make it
imperative that we should define precisely the two fluid motions which we
are thus comparing.
Describe a sphere S, of large but finite radius R, whose centre is at the
centre of the particle. Let the velocity of the fluid at points on S be given
172 Dr. G. B. Jeffery. TheMotion of Ellipsoida
by equations (2), and let the ellipsoid move freely under the fluid pressures on
its surface so th at relations (38) are satisfied. We will calculate the excess
of the rate of dissipation of energy over th a t which would occur if the ellipsoid
were removed, the sphere S completely filled w ith fluid, and the velocity a t
all points on its surface the same as in the first case.
The direct calculation of the rate of dissipation of energy presents very
great difficulties. For our purpose, it will be sufficient to consider the rate
at which work m ust be done over the boundary of the sphere S in order to
m aintain the motion. This, of course, differs from the rate of dissipation of
energy by the rate of increase of the kinetic energy of the fluid. For the
periodic motions investigated in the last section, however, the average rate of
work of m aintaining the motion m ust be the same as the average rate of
dissipation of energy.
As we shall ultim ately suppose the radius of the sphere S to tend to
infinity, we may adopt such a degree of approxim ation th a t negative powers
of R are neglected in the final result. This implies th a t the stresses m ust be
correct to order R -3 and the velocities to order R -2.
A t great distances from the origin we have, neglecting term s of order r~'A,
n —1 /r,%] = %2 = %s
Substituting in (18), (19), (20), and rem em bering th a t (38) are satisfied, and
writing
= A x2+ By2+
we have, neglecting term s of order r~ 4,
u — Uo—4x$>/rb,v = v()—Ay<&/ r 5, w = w()—4z$>/rb,(52)
while (21) gives
p = po—8yu,<3>/r5. (53)
Superpose a motion, finite at all points w ithin the sphere S, and such as
to cancel the second term s in the right-hand members of equations (52) a t
points on its surface. Such am otion is easily found by the ordinary spherical
harmonic methods, and when combined with the motion represented by
(52), we have,
u _ M R 5—r 5) . 5 (R2—r 2) 0 ^ q
RV5 R5 ete
4y (R5— r5), J 5 (R2— 3<J)
R'V5 ~l R5 (54)
R'V5 dy r >
4z (R5—r 5) ^ 5 (R2- ■r2) 0<E>
w0-
RV + R5
which give on the surface of S
p = p0—50yU,fl>/R5. (55)
 Particles Immersed a Viscous Fluid. 173

These velocities will not accurately satisfy the boundary conditions on the
surface of the ellipsoid, but proceeding by successive approxim ations it is
easily seen th a t the additional term s will be of order II-7 on the surface of S,
and may therefore be neglected. The stresses a t r = K are calculated from
(54) and (55), by the aid of (32). Typical results are
0 , 1A [ 5 A . dx2 ^ 2.x c<3>\
XX - -2?o+ ~ /* a + 1 0 / 4 ^ <1 R6 ^ J

S---N Z0<I> y
yz = 2yu.f-f 10/4 (56)
K1 B* dz S '

From these we have

xr — (x ocx + y xy + z xz / r)R

= - § * > » + ! (® * + h y + g * )+ id /t (o7)

and two sim ilar equations.

The rate a t which work is done in m aintaining the motion is given by
x—S X— s
j’J (u{) xt + Vo yr + tvo dS
where the integration extends over the surface of S. On substitution from
(57), this integral is readily evaluated and it gives
| ttE3 . 2/4(a2+ b 2+ c2+ 2 f2+ 2g 2+ 2 h 2)
+ - J- TTya(a A + bB + cC + 2f F + 2gG + 21iH). (58>
The first term is the work necessary to m aintain the motion in the absence
of the p a rtic le ; the second represents the addition due to its presence. On
making the radius of the sphere S tend to infinity, this result becomes exact.
On substitution from (31) and (39). this gives for the additional rate of doing
work owing to the presence of the particle
d W _ 16 _ |f f «o"a2+ /3 0" b 2+ y o " c 2
dt s ^ l/3o,,7o,'+ 7 o ,/«o/' + « o"A /'
2f2 2g2 2h2
«o' (b2+ c2)+ £ 0#V 2+ «2) + 70 + b2)
I t is interesting to note th at if due regard is not paid to the conditions on
the bounding sphere, and if the calculation is made directly from equations
(52), the result is only one-fifth of th a t given above.
In the case of an ellipsoid of revolution ( = c) we have 70', 70" respectively
equal to fid, /So", and it is easy to show th at + 2 « 0,/ = 2 Equation
(59) then takes the simplified form
dW «0" a 2 b 2+ c2+ 2f2 2 (g 2+ h 2)1
TT Try (60)
dt 2b2ad fid' ‘2b2ad
 174 Dr. G. B. Jeffery. The Motion of Ellipsoidal
If the fluid is moving in lam inar motion, this gives on substitution from
(40) and (44)
dW 1 7r/AK2 : + sin4 6 sin2 2 </>
dt { v 2 # W ft' 26=W f t > 2 + &2)
2
4- j — 7 cos2 6 4- sin2 0 (61)
b/5o
The average values of cos2 9 and sin4 9 sin2 2 are calculated from (48) and
(49) as follows:
hcab [•277 rU Z»2
27r(a2+ 62)Jo °°S* 0 d<f>d<f> ~
and
/cab r2 2a?b2 P 4- +2
S r(« * + F )J . sm 4f,sm 32* ^ # - (a * -iy
If D is the average rate of dissipation of energy, (61) then gives
2f 2 2a2b2f a2+ b2+
rfXK L ( ^ - 2)2
(a2- b W { ( a 2+ k2)
< — -2------- 1- — - __________£ _____ 1
L2&2«o,/3o// 2&2V f t ' G 24-&2) J
f 1
1. (62)
+ v/ {(a2
4- F ) (b2+
W e will show th at this has no maxima or minima for finite, non-zero
values of k. Differentiating w ith regard to k2, we obtain
D 27T/./C2 [£2f t " { 2V «0' - (a?+ )ft ' } + q ^ f t / ( * , / '- f t
V
(63)
d{¥) 3&2«0'ft'ft"
From their definition in equations (10), we see that a0', «</', f t ', f t " are
all positive, while <*o"—f t " is of the same sign as a2—b2. F urther

= ( (B3 + X)5/3(J2 + X)3.

after an integration by parts. The latter integral is clearly positive, and

hence the left-hand side is of the same sign as a2—b2. I t follows th at the
right-hand side of (63) never vanishes for finite values of and is of opposite
sign to a2—b2. Accordingly, the motion which gives m inimum average
dissipation of energy will correspond to k = oo for a prolate spheroid, and to
k = 0 for an oblate spheroid. The average rate of dissipation of energy will,
in all cases, lie between the values corresponding to = 0 and
We will write
D = VVfMK2 (64)
 Particles Immersed a Viscous Fluid. 175

where vsi a factor and v = 47 rah?/3 ,a nd is there

particle. E quation (62) then gives, — for oo ,
v —1 fa (65)
and for = 0
1 f a 25 ^
( 66)
* “ aP (a + b)2 l /30" 2V
In the case of spheroids the integrals in (9) and (10) can be evaluated in
finite term s. W e obtain,
(1) for prolate spheroids, = = cos #
abUQ = £ cosec4 # {(2 —5 cos3 9)sec4 #4- 3©},
a5/3o' = cosec4 # { 2 + sec3 0 —3©},
a3/3o"= cosec4# {(2 + cos2# ) © —3}, (67)

where © = cosec # log tan

(2) for oblate spheroids, b — c — a sec 6,

bbu l — | cosec4 6 {3© - cos 0 ( 5 —2 cos3 #)},
Z>5/3o' = cosec4# {(2 + cos3# ) s e c # —3©},
bs(3o" = cosec4 # {(1 + 2 cos3 #) © —3 cos #}, (68)
where © = # /sin # .
The maximum and minimum values of v for spheroids of different shapes
are given in Tables I and I I. They are given in term s of the “ ellipticity ”
of the meridian section of* the particle, as this is generally more convenient
for m easurem ent than the eccentricity. The ellipticity e is defined to be the
difference of the greatest and least diam eters divided by the greatest, so th at
e= ( a —b)/afor a prolate spheroid and e = ( for an oblate spheroid.

Table I.—Prolate Spheroids. Table I I .— Oblate Spheroids.

j
Ellipticity Ellipticity
£ — (i —Jj Minimum Maximum Minimum Maximum
v. V.
€ = b— a V. v .
a b

0 2 ‘5 2-5 0 2-5 2-5

o -l 2*431 2-540 0 1 2-464 2-582
0-2 2*361 2-586 0-2 2-426 2 -683
0-3 2 -295 2-645 0-3 2-388 2-818
0-4 2-232 2-719 0-4 2*348 3-003
0-5 2-174 2 819 0-5 2-306 3-267
0 6 2*120 2-958 0-6 2*262 3 -670
0-7 2*073 3 -170 0-7 2*216 4-354
0-8 2 035 3-548 0-8 2 -168 5 -744
0 9 2 -010 4-485 0-9 2-116 9 -960
1 0 2-000 QO l'O 2 061 00
I
176 Dr. G. B. Jeffery. The Motion o f Ellipsoidal

I t will be noted th at the m aximum value of v tends to infinity in each case

as the ellipticity tends to unity. This does not, of course, mean th at the
dissipation of energy becomes infinite, since the volume of the particle at the
same time tends to zero.

| 6. The Viscosity Suspensions.

The results of the last section may be applied, as Einstein has applied the
corresponding results for a sphere, to the determ ination of the viscosity of a
fluid containing solid spheroidal particles in suspension. If the suspension is
so dilute th at the distances between neighbouring particles are great compared
with their dimensions, we may regard each particle as the cause of a certain
increase in the dissipation of energy.
In the case of spherical particles the motion is steady and D is given
accurately, without averaging, by the second term in (58). I t is then easy to
show from (10), (25) and (39), that, when a = b = c,
A /a = B /b = c / C = E /f = G /g = H / h = 5a3/8. (69)
If v is the volume of the particle, so th a t v — 47ra3/ 3, we have,
D = vp(5a2+ b 2+ c 2+ 2i 2+ 2g2+ 2h2).
Hence, if v is the total volume of the particles in u n it volume of th e
suspension, the total rate of dissipation of energy is
2ya (1 + 2-5 Y) (a2+ b 2+ c2+ 2f2 + 2g2+ 2h2).
Erom the point of view of the dissipation of energy, we may, therefore,
regard the suspension as a homogeneous fluid whose coefficient of viscosity
is /x,* where
/x* = /a (l + 2-5Y). (70)
This agrees with E instein’s well-known result. I t does not require the
particles to be equal in size, so long as they are all spherical in shape. I t
applies only to suspensions which are sufficiently dilute and in which there is
no tendency to form aggregates. W ith these restrictions it is true for any
steady motion of the suspension, subject only to the condition that, if a, b , ...
vary in space, they shall do so only on a scale which is large compared with
the dimensions of the particles, and th a t this variation shall not promote a
variation in the concentration of the particles. For example, in applying this
formula to measurements of viscosity by capillary tube methods, we should
have to be assured: (1) th at the diameter of the tube is large compared with
th at of the particles ; (2) th a t there is no tendency for the particles to select
particular paths, say, along the axis of the tube.
Some further light may be thrown upon this last point. I t is easily seen
that, within the lim itations of the present theory, there is no resultant force
 Particles Immersed in a Viscous Fluid. 177

oil th e particle in a direction perpendicular to the axis of the tube. 11 the

axis of the tube is taken as axis of zt the undisturbed m otion is given by
u - v — 0, w = A {a2— x2— y
and A is a known constant. Referred to parallel axes through the centre of
a particle (Ji, k), th is gives u — v —0.
This may be regarded as the superposition of two m otions : (1) = v 0,
w = a2 - h2 — k2— x2 — y2 ; (2) u = v = 0, w = —
of these m otions can give rise to any force on the particle perpendicular to
the axis of s — the motion (1), for reasons of sym m etry, and the m otion
(2), by the investigations of the present paper. I t follows th a t the combined
motion can give rise to no such force.
In the case of lam inar motion, we can extend E in stein ’s result to ellipsoidal
particles. W e have then from (64) for the average total dissipation of energy
per unit volume of the suspension
fXK2(1 + vV )

(71)

The values of vfor prolate and oblate spheroids of different ellipt

meridian section are given in Tables I and I I respectively.
I t will be noted th a t the m inim um value of v is always less th an the value
(2-5) for spherical particles. In this connection it m ay be rem arked that,
although in some cases E instein’s form ula agrees well w ith experim ent, in
others it agrees only if v has a value greater th an Smoluchowski’f
proposes a formula of the type (71) for non-spherical particles, and, with
regard to the factor v, he rem arks : “ For the sphere this value {v — 2‘5) is
evidently a minimum, since by virtue of its rolling motion it influences the
shearing motion of the least volume of the fluid.” This argum ent is not
valid, for we have shown th a t for any spheroid there are possible motions for
which the increase in the dissipation of energy is less than it is for a sphere
of the same volume.

§ 7. A Hypothesis on the Dissipation o f Energy.

I t appears from the foregoing paragraphs th a t we are able only to specify
lim its for the numerical factor in E instein’s formula. I t is obviously unde­
sirable to leave a problem, which is physically quite determ inate, in this
indeterm inate form.
* For an account of the many experimental investigations of this problem, see
Arrhenius, ‘ Biochem. Jour.,’ vol. 11, p. 112 (1917).
t “ Theoretische Bemerkungen iiber die Viskositat der Kolloide,” ‘ Kolloid Zeitsch.,’
vol. 18, p. 191 (1916).
178 Dr. G. B. Jeffery. TheMotion of
We m ight assume th at at any instant the axes of the particles are in
random directions. On this hypothesis it is easy to calculate the rate of
dissipation of energy. I t is found th at this gives a value for E instein’s
factor which is greater than the maximum values shown in Tables I and II.
The explanation is th at random directions for the axes of the particles do
not correspond to a perm anent steady state.
Again, we can determ ine v if we know the num ber of particles for which
the k of (49) lies between k and k + dk, say This, however, is ju st
what our present theory fails to reveal: according to it, all values of k are
possible, and there is no indication of their relative probability. I t would
appear th a t this failure is due to the lim itations of the theory of the slow
motion of a viscous fluid.
I t will be remembered th a t the only stable direction of perm anent tra n sla ­
tion of an ellipsoid in a perfect fluid is th a t of its least axis. If an attem pt
is made to solve the corresponding problem for a viscous fluid by Overbeck’s
method, it is found th at no couple is exerted upon the ellipsoid by the fluid
if its motion is one of pure translation in any direction whatever. A heavy
ellipsoid falling through a viscous fluid would fall w ith different velocities for
different directions of its axes with regard to the vertical, but apparently it
would not tend to set its axes in any particular directions. This is due to
the fact that in the case of the perfect fluid the resultant couples arise from
those terms in the pressure which depend upon the square of the velocity,
term s which are neglected in the ordinary viscous theory.
I t seems not improbable that, in our present problem, a more complete
investigation would reveal the fact th at the particles do tend to adopt special
orientations with respect to the motion of the surrounding fluid. Our
suggestion is that the particles will tend to adopt that motion , of all the
motions possible under the approximated equations, corresponds to the least
dissipation of energy. If this be true, it reconciles the perfect fluid theory,
the viscous theory, and experim ent in the case of an ellipsoid falling under
gravity through a fluid.
Our assumption is reminiscent of the theorems of Helmholtz and Korteweg
on the rate of dissipation of energy in the motion of a viscous fluid. A
little reflection, however, will show th at these theorems are not really
applicable to our present problem. Eor one thing, except in the case of
spheroidal particles, no steady motion is possible. Further, these theorems
are themselves deduced from the approximate equations of motion obtained
by neglecting the squares and products of the velocities, and our investiga­
tions have been based upon exact solutions of these same approximate
equations. I t is impossible that a general theorem, based upon a set of
 Particles Immersed in a Viscous Fluid. 179

differential equations, should give, w ith regard to a particular case, inform a­

tion which is not contained in the corresponding exact solution of the
equations.
If this hypothesis be granted, w7e are able to assert the following-
propositions :—
1. Prolate spheroidal particles imm ersed in a fluid in lam inar motion will
tend to set themselves w ith th eir axes perpendicular to the plane of the
undisturbed motion of the fluid. They will rotate about their axes w ith
constant angular velocity, and the whole motion will be steady.
2. Oblate spheroidal particles imm ersed in a fluid in lam inar motion will
tend to set themselves with an equatorial diam eter perpendicular to the plane
of the undisturbed motion of the fluid. They will rotate about this diam eter
with a variable angular velocity, and the whole motion will be periodic but
not steady.
3. For m otion in a capillary tube, the particles will tend to concentrate
along the axis of the tube.
4. The effective value of the num erical factor in E in stein ’s formula will be
the m inimum v,as shown in Tables I and II. This is always less than th
2-5 appropriate to spherical particles. Thus, for particles of given volume,
the increase of the viscosity is greatest when the particles are spherical.
An experimental investigation of some of these provisional conclusions
would be valuable.