Академический Документы
Профессиональный Документы
Культура Документы
APPLIED MECHANICS
VOLUME 8
ADVANCES IN
APPLIED MECHANICS
Ed itors
TH.
H. L. DRYDEN
VON
KLRMLN
Managing Editor
G. KUERTI
Case Institute of Technology, Cleveland, Ohio
Associate Editors
F. H.
VAN DEN
DUNGEN
L. HOWARTH
VOLUME 8
1964
ACADEMIC PRESS
COPYRIGHTQ1964, BY ACADEMICPRESSINC.
ALL RIGHTSRESERVED
N O PART OF T H I S BOOK
FROM T H E PUBLISHERS.
LIBRARY
OF
NUMBER:48-8503
P R I N T E D I N T H E U N I T E D STATES OF AMERICA
CONTRIBUTORS
TO VOLUME
8
BERNARD
D. COLEMAN,
Mellon Institute, Pittsburgh, Pennsylvania
HERSHELMARKOVITZ,
Mellon Institute, Pittsburgh, Pennsylvania
N. N. MOISEEV,Computing Centre of the U.S.S.R. Academy of Sciences,
Moscow, U.S.S.R.
E. L. RESLER,
JR., Cornell University Ithaca, N e w York
H. S. RIBNER,Institute of Aerophysics, University of Toronto, Canada
V. V. RUMYANTSEV,
Institute of Mechanics of the U.S.S.R. Academy of
Sciences, Moscow, U.S.S.R.
W. R. SEARS,
Cornell Universit31, Ithaca, N e w Yurk
...
Vlll
in this period there was work being done on structures, aerofoil theory
(steady and unsteady), supersonic flow, turbulence, and meteorology, and
in all these KBrmPn was entirely at home; wherever possible experiment
and theory proceeded hand in hand.
With the onset of World War I1 it was inevitable that KBrmrins services
should be in great demand. This, together with his interest in post-war
years in international scientific cooperation, meant that Pasadena saw less
of him in person but continued to benefit from his guidance. He felt very
keenly the death of his mother and later of his sister, and thereafter
S. Marengo always held sad memories for him.
G . I. Taylor has written of KBrmBns work as essentially that of a
broadminded and deepminded engineer whose thought extended over the
whole range of engineering science, of his work in founding the International
Congresses, and his wider and later work on international cooperation in
scientific matters as, for instance, in founding AGARD. I t would be supererogatory for me to attempt to add to what he has written.
KrirmBn had a great gift of exposition whether it was of his own numerous
original works or in presenting that of others. One could always look to
him to go straight to the nub of any sound idea or expose a fallacy in the
kindest possible way. His work will forever remain embedded in the literature
but to us who knew him it is the loss of KBrmin - the friend, counselor,
and source of inspiration - that we mourn above all else.
L. HOWARTH
Contents
CONTRIBUTORS
TO VOLUME
VIII . .
Theodore von Kbrmbn: A Tribute .
....................
. . . . . . . . . . . . . . . . . . . .
vii
1
4
6
23
60
64
BY HERSHELMARKOVITZ
A N D BERNARD
D COLEMAN.
Mellon Institute. Pittsburgh
Pennsylvania
. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Relation to General Simple Fluids . . . . . . . . . . . . . . . . . .
. Steady Simple Shearing Flow . . . . . . . . . . . . . . . . . . . .
. Viscometric Flows . . . . . . . . . . . . . . . . . . . . . . . . .
. Steady Extension of a Cylinder . . . . . . . . . . . . . . . . . . .
. Relation to Classical Viscoelasticity . . . . . . . . . . . . . . . . .
I
I1 .
111
IV
V
VI
VII .
.
69
71
77
79
87
89
96
100
100
.
.
.
........................
ix
104
105
106
109
109
115
119
142
142
CONTENTS
V I I.
VIII
I X.
X.
XI .
. Refraction
. .
. .
. .
. .
..
.
.
.
.
.
.
151
In!)
163
169
115
118
178
Notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
I . Simplest Cases of Motion: the Cavity is Completely Filled . . . . . . .
I1. Stability of Motion of a Solid-Liquid Body with Respect to a Part of the
Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . .
I11. Stability of Steady Motion of a Solid Body with Liquid-Filled Cavity .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
183
184
186
203
214
230
.
.
AUTHORINDEX
SUBJECTINDEX
.
.
..
.
.
..............................
.............................
233
235
. 268
287
288
291
295
AND
E . L . RESLER,
JR
a
A
.4
Speed of sound. v y x p
.4lfv6n.wave velocity
Magnitude of the vector A
- 1)W
(Atm*
Mass of particle
Mass of an electron
Mass of an ion
Density of neutral particles
Density of electroris
Wave-normal vector
Pressure
Pressure perturbatlion
Pressure in related non-magnetic flow
pH4/8n
Total pressure, p
Magnetic Prandtl Number 4npoa
Electron partial pressure
Fluid-velocity vector
Magnitude of the vector q
Velocity vector in related non-magnetic flow
1
T'
U
U'
U
U
-,
U
UP
V
V
U'
VP
Y
ii
W
W,
x , Ys z
YD
a
Pa
Y
c'
F
8
1
Id
V
5
P
P'
n
7
71
7e
7m
Temperature
Perturbation of temperature, T - T,
x component of q
x component of v
Free-stream velocity vector; i.e., value of q a t large distances from
an obstacle
Magnitude of the vector U
Value of U in related non-magnetic flow
x component of particle velocity V
Average velocity component of particle along x
Velocity vector of particle
y component of q
y component of v
y component of V
Perturbation velocity vector q - U
Average velocity component of particle along y
z component of q
z component of Y
Cartesian coordinates; sometimes x , y are cylindrical coordinates
Distance t o damp t o I/e in y direction
p H a / 4np U 2
(1 - Mm')l/2
Specific-heat ratio
= 0 for plane flow, = 1 for axisymmetric flow
z component of
Complex constant in the expression $ a exp ( i l x - 8 y )
Heal constant in the expression # a exp ( i l x - @y)
Magnetic permittivity of fluid, e.m.u.
Kinematic viscosity of fluid
curl H, therefore 4nJ
Mass density of fluid
Perturbation of density, p - pm
Electrical conductivity of fluid, e.m.u.
Mean time between collisions of electrons with ions
Mean time between collisions of ions with neutrals
Mean time between collisions of electrons with neutrals
Mean collision time for electrons in collisions with ions and neutrals,
(Tc-l
+ T-1)-1
Undisturbed condition
Conditions a t outside edge of boundary layer in Chapter I .
INTRODUCTION
The class of magneto-fluid dynamic problems that involves streaming
flow about solid obstacles has attracted attention by virtue of the interesting
phenomena that are predicted by the theory, rather than its practical utility.
Some of these phenomena ar,e analogous to the familiar features of conventional gasdynamics - standing waves, elliptic and hyperbolic fields, boundary
layers, and wakes - and others present surprising contrasts.
Such flows should be capable of realization in the laboratory, even if
the great difficulties involved in any experiments with conducting gases
are admitted. Furthermore, as will be seen in the examples that follow,
some of the phenomena predicted for these flows are prominent and should
be easily discernible and measurable. Thus, some of these configurations
should be suitable for diagnosis of flows, determination of fluid properties,
and experimental verification of theoretical predictions. This makes this
category of flows important, for there is a dearth of such experimental results
throughout the whole field of magneto-fluid dynamics.
In this report we shall consider almost exclusively steady flows that
consist of a uniform stream at great distances from the solid bodies, and
upon which a magnetic field, also uniform at large distances, is imposed by
external means at some angle to the stream. Whenever this angle is different
from zero it is assumed that an electric field is also applied, so that the
region of uniform flow and uniform magnetic field far from the body is
current-free.
Throughout the report, th'e approximations of continuum flow are made.
This implies, of course, that the density is so large - or, more precisely,
that the Knudsen Number is :SO small - that it is legitimate to assign to the
fluid a t any point a pressure, density, temperature, and velocity. I t will
also be assumed that thermodynamic equilibrium is maintained at every
point, and the perfect-gas law will be used to describe the thermodynamic
state. The assumption of thermodynamic equilibrium also implies that the
mechanical stresses acting are adequately described under Stokes' Hypothesis ;
i.e., that the mechanical equations of the problem are the Navier-Stokes
equations.
In the same spirit and with substantially the same validity and limitations, the electrical properties of the fluid are described, throughout the
greater part of this paper, by assigning to it an electrical conductivity and
magnetic and electrical permittivities (both usually equal to those of empty
space). As is customary in magneto-fluid dynamics, that portion of the
electric current that is due to the transport of charge by the fluid motion,
the so-called convection current, is neglected. This does not mean that the
fluid is everywhere neutral in charge, which it cannot be in general, but only
that the charge density is always so small that the convection current is
much smaller than the conduction current. Also, relativistic effects are
neglected throughout .
Even with these simplifications, the equations describing the flow are
found to be intractable in most of the interesting cases, and further approximations must be introduced to permit solutions to be constructed. In
particular, the electro-magnetic properties of the fluid, such as conductivity
and permittivity, are taken to be constants, rather than functions of the
thermodynamic state a t every point. This implies that the temperature,
in particular, does not vary widely - an assumption that may be of doubtful
validity in some experimental situations.
Rather than writing out the full equations appropriate to the fluid
model just described - i.e., a viscous, compressible, conductive gas - we
shall undertake here to expose and explain the principal phenomena of flow
about bodies by beginning with the simplest model, an incompressible perfect
conductor, and proceeding subsequently toward more complicated cases.
To justify this approach, we shall argue that the main features are deducible
from rather simple equations and that further elaboration of the fluid model
results in modifications of these features rather than entirely new results.
Thus we shall begin with a study of flows of incompressible fluids of
infinite electrical conductivity and then proceed to investigate the main
effects of electrical resistivity. These studies will permit us to investigate
subsequently the effects of compressibility in perfectly conducting, inviscid
gases with some knowledge of the modifications due to viscosity and resistivity. Some details of these modifications, at least those due to resistivity,
will, however, be worked out as well. This results in a relatively clear understanding of the main features of flows in real, conducting gases in the
absence of Hall Effect - i.e., a t high densities and moderate magnetic-field
strengths. Finally, it will be pointed out that even perfect conductors may
exhibit appreciable Hall Effect, and some effects of the Hall phenomenon
- the drift of charged particles across the electric field due to the presence
of the magnetic field - will be worked out.
I . Boundary Conditions
To solve problems in magneto-fluid dynamics consistent with the above
discussion, solutions for the electromagnetic variables must be obtained
throughout all space. Thus conditions on both the magnetic- and electricfield vectors at surfaces across which the material properties, including
conductivity and permittivities, change, as at a solid-fluid interface, must
be formulated consistent with the electromagnetic equations. The general
statements of these interface conditions follow.
In all cases the component of the magnetic-induction vector B normal
to the surface and the component of the electric-field vector E along the
surface must be continuous, since the former is divergence-free and the latter,
in steady flow, curl-free.
Similarly, the component of magnetic-field strength H along the interface
must be continuous, because the curl of this vector, according to Ampbres
Law, is proportional to the electric-current density j. To permit a discontinuity in the tangential component of H would mean that the current
density becomes locally infinite, which is not realistic in actual fluids of
finite electrical conductivity. However, in some cases the approximation
of infinite conductivity will be introduced as a limiting case of fluids of large
conductivity. When this is done it may be necessary to admit infinite
current density in surface-current layers, i.e., to allow the tangential
component of II to be discontinuous. When such an approximation is made,
an effort will also be made to provide an explanation of the detailed nature
of the current layer in a real substance.
The boundary conditions on the velocity vector q at a fluid-solid interface
are the usual ones; namely both normal and tangential components of q
must be continuous, i.e., must vanish. If the approximation of vanishing
viscosity is made, the condition on the tangential component must be
relaxed; i.e., a vortex sheet must be permitted at the interface. The analogy
between the current layer of infinite current density and the vortex sheet
of infinite vorticity will be emphasized later in this report.
2. Units
INCOMPRESSIBLE
FLOW
Continuity :
div q = 0,
(1.1)
Momentum:
(q -C7 )q +-g
1 rad p =v V2 q +-j P
P
x H.
div B = 0,
Amplres Law:
4nj = curl H,
(1.4)
Faradays Law:
curl E = 0.
(1.5)
= o(E
+ q x B).
u-l curl j
= curl
(q x B) = ,u curl (q x H)
The equation of momentum, Eq. (1.2) can also be combined with AmpQes
Law to eliminate the current density j ; the result is
B = pH.
(q*V)rFz- ( S 2 . V ) q = v V 2 S 2 + ~ c u r l [ ( H V)H].
.
4ZP
W. R. SEARS A N D E . L. R E S L E R , J R .
Rm
= 4npaUL
where L and U are characteristic length and flow speed, respectively. This
number then measures the ratio of magnetic-field diffusion-time 4npaL
to the characteristic flow-time LIU.
I n view of the correspondence between the equations governing H in
magneto-fluid dynamics and G? in conventional fluid mechanics, the various
phenomena typically associated with ordinary aerodynamics a t different
Reynolds Numbers Re can be expected to have their counterparts in various
regimes of the Rm spectrum.
Using the kinetic-theory approximation to v for gases, namely v = 0.499 d
the magnetic Prandtl Number Pr, becomes 0.499 (4np)oEl; i.e., it is the
magnetic Reynolds humber based on the mean free path and the mean
thermal speed E. For ionized gases under typical terrestrial conditions, this
material property is much smaller than 1, as substitution of typical values
into this formula will verify. The same is true of Pr, for other fluid conductors, such as liquid metals, electrolytic solutions, and solutions of alkali
metals in ammonia. Consequently, we may not expect that the Re and R m
spectrums will coincide ordinarily, but rather that the large-Re approximation will frequently be appropriate for a wide range of values of Rm.
We shall return to this point later in the report.
(1.12)
q=U+v
and
H=H,+h
and
J h (<< (Hrnl.
where
(1.13)
Iv(
<(
IUI
Both equations (1.7)and (1.8)can then be linearized and assume the forms
(1.14)
1
V2h = (Ha V)V - (U * V)h
4npa
- __
and
(1.15)
or if 51 denotes curl q, the vorticity, and
times the current density,
(1.16)
and
(1.17)
Eqs. (1.16)and (1.17)are clearly the linearized counterparts of Eqs. (1.7)
and (l.lO), respectively.
(1.18)
The solution of this equation is simple: both 51 and 6 consist of parts
that are constant along lines having the directions sketched in Fig. 1.1,
having the directions U
A, and U - A,, where A, denotes the AlfvCnwave velocity Vpu/4npH,.
10
The meaning of Eqs. (1.18) is now clear: vorticity and current are
propagated at the AlfvCn speed in the direction of the magnetic field, relative
to the moving fluid.
The solutions for velocity and magnetic-field vectors are found by
integration of
and p. It is clear that they consist of rotational parts
that are constant along the directions shown in Fig. 1.1 plus curl-free parts.
By returning to the equation of momentum and Ohms Law, Eqs. (1.15)
and (1.14) with Y = 0-1
0, one can relate the rotational parts of v and h
and also their irrotational parts.
11
By cross-differentiation of Eqs. (1.16 and 1.17), the fourth-order differential equation satisfied by both S and 5 in the case of incompressible,
steady, viscous flow with finite conductivitv is obtained :
(1.19)
12
where E, and B, are the electric and magnetic fields appropriate to the given
boundary conditions with no flow. In the special case we are discussing
here, where E and B are uniform at large distances, 8, and B, are constants,
but the theory can be used to investigate many other geometries [15], [16],
[171, WI, [191, [201.
In this small-Rim approximation, therefore, the equation of momentum,
Eq. (1.2), becomes independent of the induction equation, Eq. (1.7). The
damped-AlfvCn-wave character of the flow is lost at this limit; the flow
is everywhere rotational. The conductivity is so small that the magneticfield disturbances and accompanying vorticity are diffused to great distances
within the typical flow-time. This type of approximation has been used
by Rossow [21], [22] in boundary-layer studies, in addition to the authors
mentioned above.
Attention should also be called to a more general study by Stewartson [23]
in which Rm is considered arbitrary in magnitude and it is shown how the
various limiting cases such as R m - 0 , 00 (and also cases of very strong
magnetic field, which are mentioned below) are obtained from a general
formulation, within the small-perturbation approximation.
2. A ligned-Fields Flow
A special case that has attracted some attention is the category of flows
with U parallel (or anti-parallel) to B, ; these are called aligned-fields /lows
and are in some respects singular in the family of flows with varying angle
between the vectors. They are of some interest because they represent a
particularly simple experimental arrangement, namely flow in a pipe or
wind tunnel which is itself enclosed in a solenoid.
For aligned-fields flow that is uniform a t large distances, E must vanish
at large distances. Two cases of particular interest are plane and axisymmetric
flows; for these, Eq. (1.5) requires that E be constant: hence it must be
zero throughout the flow. The set of equations (1.7), (1.8) can therefore
be simplified by replacing the former by its integral (i.e., by Eq. (1.6a)), viz.,
(1.21)
--
4npa
curl H = q x H.
Ideal Conductor
Suppose now that u -+ co; Eq. (1.21) then states that q is parallel to H.
The magnetic lines of force are said to be frozen into the fluid and therefore
congruent with the streamlines. Moreover, in view of Eq. (1.21) and the
divergence conditions it follows that q o( H, and evaluating the constant
at large distances from the body we have
(1.22)
qlU
= H/H,.
13
Although we have derived Eq. (1.22) only for plane and axisymmetric
flow, this frozen-fields configuration describes a possible flow pattern for
any aligned-fields flow of a perfect conductor, regardless of symmetry, as
we shall see. Let us therefore carry out the analysis without any special
assumption regarding symmetry.
The momentum equation, Eq. (1.8) becomes, in view of Eq. (1.22),
(12 3 )
where am2 denotes the ratio pHm2/4npU2,i.e., the ratio of the magnetic
pressure to the dynamic pressure in the undisturbed stream. am is also the
reciprocal of the Al/vLn Number of the stream; if a, > 1 the stream speed
is less than the speed of Alfvkn waves in the undisturbed flow and may be
called sub-AlfvLnnic; if a, < 1, it may be called super-Alfvkic.
Moreover, if the total pressure p
p H 2 / 8 nis replaced by the symbol I,
this momentum equation (1.23), takes on the form of the familiar momentum
equation of ordinary incompressible, viscous flow. The additional relation
needed for solution of Eq. (1.23) is the continuity equation, div q = 0.
Hasimoto [24] has shown how solutions of ordinary viscous-flow problems
can be transformed to solutions of Eq. (1.23). He introduces the fictitious
velocity 4 and pressure $, defined by
(1 - aa2)q= ij
(1.24)
and
(1 - a,2)P = j .
(1.25)
(q
V)q
1
+ -grad$
P
= vP2q
and div q = 0.
Thus any viscous, incompressible, steady flow involving an undisturbed
uniform stream can be interpreted to yield a family of aligned-fields flows
for various values of a,. If the undisturbed stream speed of the viscous
flow is 0,the stream speed of the related aligned-fields flow depends on
u,2, viz.,
(12 7 )
u = 0 / ( 1 - a,2).
This requires that the flow direction be reversed if a m 2 > 1; i.e., if the
stream is sub-AlfvCnic. If the stream speed is equal to the AlfvCn speed the
transformation fails.
14
H,
(C
FIG. 1.2. Related non-magnetic (a) and aligned-fields, viscous, incompressible flows
according to Hasimotos theory. In (b) the stream is super-AlfvCnic (ama< 1 ) ; in
(c) it is sub-AlfvCnic (a,* > 1).
15
(1.28)
where the subscript 1 denotes conditions just outside the vortex-current
layer. But
can be calculated from Bernoullis equation, so that
(1.29)
Thus, the pressure at the body surface and the resulting force and moment
on a body are profoundly affected; their variations along the body contour
* The reader who undertakes to calculate the pressure should not be concerned
by the fact that the pressures in the related non-magnetic flow, according to Eq. (1.25),
are negative for sub-Alfvhic flow. The related flow is. of course, unrealistic to this
extent, but it should be recalled that incompressible flow patterns (velocity and pressuregradient distributions) are unaffected by changes of pressure level, i.e., by adding or
subtracting a constant pressure.
16
W. R . SEARS A N D E. L. RESLER, J R .
17
FIG. 1.3. Sketch of plane flow with verv strong magnetic field.
Hz
= 0(1),
v = 0 (Rm- l/2),
H,
= O(Rm-1/2),
j = O(Rm1/2),
Q = O(Rm1/2).
boundary layer are both zero. The tangential velocity component u remains
OP).
(d) In the viscous sublayer the orders of magnitude, by contrast, are
u = 0(1),
2'
= O(Ra-'/2),
H , = O ( R M - ' / ~ ) O(PY,'/~),
H,= O(Rm-1'2),
= O(Rm1/2),
Q = O(Re1/').
18
19
(wU(l
V
-am2)
where
(1.33)
11,2= 2L {(Re
___
v
a
V./a,
The nature of the flow for small values of Pr, can be determined approximately from Eq. (1.33). Let
6 = Rml Re = Pr,
(1.34)
<< 1.
Then
(1.35)
k 2
1 Re
2 r (1
+ 6 f (1 - 6 + 2am26)}
or
(1.36)
1, w Re/L
A2 w (1 - aa2)Rm/L.
and
Thus, current and vorticity are confined to wakes whose lateral dimensions are, respectively, of orders
(F)
1/2
(1.37)
and
),
4npc(1 X- am2)U
20
W. R. SEARS A N D E. L. RESLER, J R .
and it is clear that the latter, i.e., the thicker wake, expands in the upstream
direction when a m 2 > 1.
In other words, for arbitrary Rm and small P r , the flow appears to be
of the character already deduced above, involving a narrow viscous wake
and a much thicker inviscid wake or precursor. To be sure, studies of this
nature do not cast much light on the question of how the wakes or precursors
join the flow near the solid body, which for sub-AlfvCnic flow is the unsolved
problem alluded to above.
Lary [32] considered the limiting casePr, = 0, i.e., inviscid, smallperturbation, aligned-fields flow at arbitrary Rm. His results agree
with what has been said above, and he was able to solve boundary-value
problems of slender two-dimensional and axisymmetric bodies by distribution
of singular solutions. Thus he did answer the question of the joining of the
far-field to the flow near the solid body, but only within the accuracy of
the linearized small-perturbation approximation. Larys results lend some
support to the general picture we have deduced for aligned-fields flow but
cannot be considered definitive.
We shall not reproduce Larys detailed results here, but only draw upon
them to assess the strength of the upstream wake in typical cases. For a
body of revolution whose cross-sectional area is S ( x ) , in sub-AlfvCnic flow,
Larys formula (62) of [32] gives the following expression for the streamwise
velocity perturbation u at a distance - x ahead of the body nose ( x = 0):
L
(1 3 8 )
0
(1.39)
0
21
where the integral is recognized as the body volume. It is clear that the r 3
term in Eq. (1.39) must be ignored when magnetohydrodynamic effects
are present because comparable terms have been neglected in the magnetohydrodynamic term; we have retained it to show explicitly the leading
non-magnetic upstream influence.
Thus the upstream influence in the sub-AlfvCnic precursor, measured
by the velocity deficiency, is an order of magnitude larger than the conventional effect. For example, if Rm = n and a, = 2, the wake velocity
deficiency is x/2L times as great as the deficiency without magnetohydrodynamic effects, i.e., 5 times as great at a distance of 10 body lengths.
The precursor should therefore be easily discernible in the laboratory.
3. Flow with Very Strong Magnetic Field
FIG. 1.4. Sketch of aligned-fields flow with very strong magnetic field [38], [39].
general case where B, and U are not parallel, the conclusion is not
unexpected in view of the results obtained earlier in this report: for either
very large or very small Rm, the flow is undisturbed except by standing
waves lying in the B, direction, and conditions are invariant along these
waves (Fig. 1.3).
This type of flow has been studied in more detail by Ludford and Singh
[361, [371.
For aligned-fields flow, in the same limit a, -+ bo, Stewartson [38], [39]
finds an analogous type of flow (Fig. 1.4). This result leads Stewartson
to the interesting conjecture that all steady, sub-AlfvCnic, aligned-fields
flows are analogous to that sketched in Fig. 1.4; they involve wake-like
disturbances both upstream and downstream of the body, whose widths
22
W.
R. S E A R S A N D E. L. RESLER, J R .
are not small, even for streamline bodies, but are of the same order as the
body thickness.
In view of our difficulties in finding acceptable solutions to the equations
of the inviscid boundary layer, mentioned earlier, it seems possible that this
conjecture is correct. But still other flow patterns have been suggested;
e.g., by Tamada [a],
and the correct solution seems to await more elaborate
calculations or experimental observations.
4. Summary
In brief summary of what has been found thus far concerning the steady
flow of incompressible conducting fluids past solid obstacles, we emphasize
that for large Rm the picture is dominated by the phenomenon of AlfvCn
waves. In real conducting liquids such waves are diffused by both
viscosity and electrical resistivity; the latter is typically a relatively
larger effect.
With increasing resistivity (smaller Rm), this diffusion of the standing
waves becomes so great that the waves actually become broad regions that
expand parabolically away from the obstacle in the AlfvCn-wave directions.
These are sometimes called wakes, even when they do not lie in the stream
direction, because of their shapes. Finally, for small Rm, the diffusewave nature of the flow is lost, for magnetic diffusion is essentially
instantaneous.
The category of aligned-fields flows is singular, however, for here the
AlfvCn-propagation direction coincides with the flow direction. For large Rm
the flow is irrotational and current-free outside of thin boundary layers and
the wakes to which they connect. Detailed study of these phenomena
reveals that, for small Pi,, both boundary layers and wakes are essentially
two-layered, with a thin viscous layer imbedded within a much thicker,
nearly inviscid layer. The inviscid boundary layer appears t o grow in
thickness from rear to front and to connect with an upstream wake or
precursor, whenever the flow is sub-AlfvCnic. The detailed nature of this
sub-AIfvCnic flow is not clear, however, and the conjecture has been
advanced that an entirely different kind of flow pattern actually occurs
a t sub-AlfvCnic speeds, viz., slug flow, in which there are dead-water
wakes both upstream and downstream of a body.
In the next chapter the effects of compressibility of the fluid are considered. It will be shown that only a few changes must be made in the
descriptions given here in order to describe qualitatively the flow of a
conducting gas past an obstacle.
23
11. COMPRESSIBLEFLOW
1. Aligned-Fields Flow of Inviscid Gases
To begin our discussion of gas flows past solid obstacles, we shall reverse
the order of the preceding chapter and treat first the special case of alignedfields flows. We shall consider first inviscid flows, with some confidence
that the principal effects of viscosity, viz., diffusion of waves, boundary
layers, and wakes, are not qualitatively different from what occurs in nonmagnetic flows.
The equations of the problem are Eqs. (1.2) to (1.6), inclusive, where
now p is variable and the viscosity term in Eq. (1.2) is omitted, but Eq. (1.1)
must be replaced by
div (pq) = 0
(2.1)
1
-V2H = (H V)q
4nP7
- (q * C7)H - H div q.
(2.3)
=9 p T .
cv(q P)T
+p(q - P)p-l=
j2./!a,
av
ah
p,U-----HH,-+grad
ax
4n
ax
ax
+ p,
div v = 0,
p'
is
p - pa, and
24
4nj = curl h,
(2.7)
1
av
ah
-__
V2h = H, -- U - - Hmdivv,
ax
4nv
ax
and the energy equation. But according to Eq. (2.7) the current density
is a first-order quantity, and therefore the Joule-heating rate is second-order
and can be neglected. The energy equation becomes a statement that the
flow is homentropic, and the temperature can be eliminated from the
equation of state:
(2.10)
P =py
or
P'1P.X
= yp'lp,.
p,Uu'
+ P' = function of y
and z.
= Pa
i- p ' =
pco
- pmuu'.
p,UQ -
fi
8 = function of y and z
4n
where again 5 denotes curl H, or 4xj. Once again the function of y and z is
seen to vanish by consideration of conditions a t large n. Thus,
(2.14)
25
av
ah
aui
1
- U -ax
- M,2H, -.
2&i V2h= H , ax
ax
Ideal Conditctor
At this point let us make the approximation of infinite conductivity in
the equations just derived. Only Eq. (2.16) is affected; its left-hand side
becomes zero and it can be integrated with respect to x :
H,v - M,Hmu
(2.17)
Uh
where the parentheses enclose vectors and Pm2is the Prandtl-Glauert factor
1 - MWa. This is an interesting generalization of Eq. (1.22), to which it
reduces a t small Mach number.
In principle, Eqs. (2.14) and (2.17) are to be solved to determine the
flow pattern. For simplicity let us illustrate this by assuming that the flow
is either plane or axisymmetrical, so that a small-perturbation stream function $ can be used ( $ C , 11 of [41]), i.e.,
(2.19)
p,w
= y--Ia+lay,
VI =
- y-Ea+iax
(2.20)
(2.21)
26
W. R . S E A R S AND E. L. RESLER, J R .
CONVENTIONAL
I3
J5
P
ELUPTIC
FIG.2.1. Diagram illustrating the regimes of hyperbolic and elliptic flow problems for
the aligned-fields case. Conventional aerodynamics occupies the extreme upper region.
where
(2.24)
A w 2 =
8w2-1
aW2pm2
-1
M , 2a,
-2
+ aW-2 - 1
MW2
c,a
= am2
+ A,% - a , ~ A w 2 i U 2
27
of a,-l against M , , as in Fig. 2.1.* The most striking features are the
existence of subsonic-hyperbolic and supersonic-elliptic regimes. As indicated,
conventional (non-magnetic) flow occurs in this diagram at large values of
a,-l and consists simply of elliptic-subsonic and hyperbolic-supersonic
regimes. The parabolic loci are the straight lines M , = 1 and a,-l = 1.
The transition at the quarter-circle M m 2 a,-2 = 1 does not have any
counterpart in non-magnetic fluid mechanics, for the flow in the hyperbolic
area adjacent to it is hypersonic: the ratio A, is very large, not because
the flow speed is large but because the critical speed c , is small. The name
hypercritical has therefore been suggested for this transition [a].
$(x,Y) = f ( x
+ m m ~ +)
S(X
- mmy)
where f and g are arbitrary and rn, denotes l V d r n 2- 11. A typical, simple
flow is sketched in Fig. 2.2 for a subsonic sub-AlfvCnic stream. It should
be of interest that the streamline spacing is narrower than free-stream above
the inclined plate, and wider below. In view of Eqs. (2.10) and (2.12), the
relationships among speed, pressure, and density are conventional ; thus,
Diagrams of this type were used in References [42] and [43];the latter is a study
of the characteristics of the nonlinear equations of inviscid, perfectly-conducting. gas
flow. This emphasizes the fact that the small-perturbation results derived here are
locally correct in nonlinear problems.
28
since the flow is subsonic, the pressure is greater in the flow below the plate
than above. For the case sketched, in fact, it is
(2.27)
=pa
- pcoU2mmd/Pm2
=p
+ pmU%J/Pm2
FIG.2.2. Sketc of plane, aligned-fields sub-AlfvCnic, subsonic flow over a flat pla .e
at angle of incidence 6.
qxH=O
As indicated, Eqs. (2.17) and (2.18) are linearized statements of this result.
A consequence is that, just as in the corresponding incompressible case,
the magnetic field must vanish within any solid obstacle of finite dimensions.
Once again there appears a current- and vortex-layer a t the fluid-solid
interface, and the surface pressure differs from the fluid pressures calculated
29
P = p1= Pl +
HI2
P Z
(2.32)
to first order. Now It,, can be evaluated from Eq. (2.18)for small-perturbation
flow; hence,
(2.33)
FIG. 2.3. Sketch of plane, aligned-fields sub-Alfvhic, supersonic flow over an airfoil.
30
W. R. SEARS A N D E. L. RESLER, J R .
But in sketching Fig. 2.3 we have implicitly assumed that the pattern
can be carried over from incompressible flow without altering the circulation.
In a kind of flow that involves both viscous and inviscid boundary layers,
and in which some boundary layers appear to grow from rear to front on a
body, it surely must be necessary to reexamine the principles - inherently
viscosity-related - that fix the value of the circulation in actual flows.
FIG.2.4. Diagram relating wave-propagation speed c to the wave angle in steady flow.
31
speed A , , and the angle e, between the wave normal and the magnetic-field
vector.
But the waves of steady flow patterns are just these magnetosonic waves
viewed in a moving system. Since they are stationary in this system, their
propagation speed c must be - U cos e,, i.e., the component of U in the
normal direction must be the propagation speed c (Fig. 2.4). Thus the
propagation speed c can be identified with c, of Eq. (2.25), which clearly
has the same meaning. Equation (2.25) can then be written as
+ A m 2- am2Am2
cos2 (p
C2
(2.36)
c2 = a m 2
or
(2.37)
c4
- (a,2
+ Am2)c2+ am2Am2cos2q~ = 0.
Further study of the modes of the respective waves discloses that the flow
produced by the waves whose speed is A , is not plane but involves velocity
components normal to the plane of H , and the wave normal. I t is for this
reason that this root was eliminated from our formulas for plane flow and
(if azimuthal velocities are assumed not to occur, for reasons of symmetry)
from axisymmetric flow as well.
The roots of Eq. (2.38)are usually plotted in a polar diagram of c versus e,,
as in Fig. 2.5.
There is also a companion diagram to Fig. 2.5, due to Friedrichs and
Kranzer [47], giving the shape of a self-similar disturbance that expands
from a point according to the speed-inclination relationship of Fig. 2.5.
It can be constructed from Fig. 2.5 by a graphical process [49] that is
equivalent to Huygens' construction. The result is shown in Fig. 2.6. In
geometrical terminology, Fig. 2.5 gives the pedal curves of Fig. 2.6. This
figure has the property that is is shape-preserving if every element propagates
according to Fig. 2.5; it is therefore the shape of the disturbance that
propagates from every point along the path of a moving point-disturbance
and is the magnetogasdynamic analog of the spherical disturbance of radius
a, in ordinary gasdynamics.
Figure 2.6 is a figure of revolution, symmetrical about the B, direction.
The disturbance front caused by the fast waves of Fig. 2.5 is an oblate
spheroid. The slow waves produce two cusped figures of revolution that
32
sLowKd
WAVES
0-
FIG.2.6.
a, = V Z A , .
(The dimen-
33
lie of course, within the spheroid; it can be shown that the volume inside
of these figures is undisturbed [50]. The intermediate waves produce only
two point-disturbances propagating at the Alfvkn-wave speed along the B,
direction. The innermost parts of the slow-wave curves of Fig. 2.5, being
approximately circular, also produce point-disturbances ; these are the
points propagating with speed (urnd2 Am-2)-1/2 at the inner ends of the
cusped figures. A three-dimensional sketch of the disturbance diagram is
given in Fig. 2.7.
Now the standing-wave results derived above can be understood in
relation to this diagram (Fig. 2.6) and the general nature of more general,
aligned-fields or non-aligned steady flows can be predicted as well.
FIG. 2.7. Three-dimensional sketch of the disturbance pattern of Fig. 2.6 (not to scale).
First, all of the elliptic and hyperbolic regions of Fig. 2.1, which pertains
only to plane and axisymmetric flow, can be understood. Stationary waves
are produced whenever the sequence of disturbances produced by a moving
body forms an envelope. This can be ascertained by drawing the body's
velocity vector in Fig. 2.6 and asking whether a tangent can be drawn
from the head of this vector to any point of the disturbance figure. If the
flow is plane, the envelope is a plane; if the flow is axisymmetric, the
envelope is a cone. Thus, body-velocity vectors that terminate in OA
(Fig. 2.6) produce no tangents and the flow is elliptic. Those that terminate
in A B produce tangelits to the cusped, slow-wave figures; these represent
the forward-facing waves of subsonic sub-Alfvknic flow. When the flowspeed vector terminates in BC, no tangent can be drawn; this is the elliptic
regime of subsonic super-Alfvknic flow. When the speed U is greater than
OC, tangents can be drawn to the fast-wave diagram, giving correctly the
hyperbolic, supersonic, super-Alfvknic regime.
34
W . R. SEARS
The preceding discussion has been based on the assumption that a , > A ,
or M , < a,-l. Study of Eq. (2.38) will disclose that when a, < A , the
fast- and slow-wave curves in Fig. 2.5 remain unaltered except that the
dimensions a, and A , in the figure must be interchanged. The intermediatewave circles, of course, must still have diameter A , . Therefore, when
a, < A , , Fig. 2.6 is altered only by interchanging the dimensions a, and A ,
and moving the intermediate-wave point to point C. Then the regime
<U<
is the regime of elliptic, supersonic, sub-AlfvCnic flow.
O
x
BC
Let the stream vector U have any arbitrary direction. By viewing the
flow in a coordinate system moving along the cylindrical axis with a speed
equal to the axial component of ti, one can always convert the flow to one
35
in which both H, and the stream vector lie in a plane perpendicular to the
axis. We shall therefore confine our discussion to this case.
Now, the disturbances set up by the relative motion are cylindrical;
i.e., they are the envelopes of the three-dimensional disturbances of Fig. 2.6
and consist of cylindrical wave fronts having the cross section of Fig. 2.6.
By symmetry the flow pattern can be expected to be plane.
The nature of the resulting flow patterns can again be ascertained by
drawing the body-velocity vector in Fig. 2.6 and looking for tangents, which
represent plane envelopes. Figure 2.8 illustrates one such case: the bodyvelocity vector - U is so great that its arrowhead lies outside the fast-wave
front ; hence, four plane, stationary, wave families are produced.
The flow pattern about an obstacle, when the stream speed is U and
its magnitude is greater than that of the fast-wave front, is therefore doubly
hyperbolic, made up of four families of waves whose directions are
given by the construction of Fig. 2.8. The perturbation stream function $,
for example, is given by an expression of the form
(2.39)
36
W. R. SEARS A N D E. L. RESLER, J R .
a layer acts in a direction perpendicular to the magnetic field. In alignedfields flows this is in equilibrium with the pressure jump across the layer,
which we discussed and evaluated a t an earlier point in this report. Hut in
crossed-fields flows the body force has a tangential component, which cannot
be balanced.
It is interesting to notice that, according to this discussion, there is one
less boundary condition at each surface in aligned-fields flow than in more
general geometries. But the aligned-fields case is just the singular one in
which one family of waves is lost (cf. Figs. 2.7 and 2.8), so that the flow
pattern remains determinate.
(2.40)
+(x,
Y) == C / i ( x
+ WY) +
+e
r-1
Three-Dimensional Flows
Without going into any more detail concerning two-dimensional flows,
let us proceed to the more complicated, general case. I t has already been
remarked that the Friedrichs disturbance diagram, Fig. 2.6, is a figure
37
38
-.
39
where the normal velocity component lies in K i before the shock and K,
behind. (Note that, since the processes being considered involve finite
increments, all of the quantities defining the diagram, such as a, A , and B
are different on the two sides of the shock, in general.) The requirement
> i results in increasing entropy across all of these shocks; they are
therefore thermodynamically acceptable,. and all of them are compression
waves.
FIG.2.10. Same as Fig. 2.5 but with notation added to define the six possible kinds
of magneto-gasdynamic shock waves. The left half (a) is drawn for A , > a,, the
right half for a,,., > A , .
40
T o make the situation more difficult, there are grave doubts about the
stability of some of the six families, i.e., whether they will actually be
produced in real-gas flows [59], [60], [61]. There seems to be general
agreement among these investigators, that four of the families, namely
all those that cross the dashed line in Fig. 2.10, are unstable.
It is difficult to predict how and whether the stable strong shocks will
occur in flows past bodies. Cabannes [58] has shown that attached strong
shocks, namely those of the families R, -+K, and R, .-+ R4,can appear in
the calculated flow patterns for a very simple type of flow, namely steady
sub-Alfvknic flow past an infinite wedge.
One more investigation of steady, plane, inviscid flows about obstacles,
involving shock waves, has been undertaken by Geffen [@L]. The particular
case studied is sub-AlfvCnic, supersonic, plane flow. As has already been
pointed out, this flow is paradoxical in that its streamline patterns are
elliptical but its speed-area relationships are supersonic. As the streamlines
widen to pass around the body, the local speed should increase. How, then,
does such a flow negotiate the leading edge of an obstacle ? This question
was posed in References [63j and [64], and the flow of Fig. 2.11, involving
a strong shock of the family R,
R,, was proposed as a conjecture.
-.
FIG. 2.11. Conjectured flow pattern for a case of sub-Alfvhic supersonic flow. References [63], !64].
41
FIG.2.12. Sketch of calculated flow pattern for a case of sub-AlfvCnicsonic flow. After
N. Geffen [62].
42
(2.4'2)
Let us rewrite this equation in terms of the previously defined symbol Vry,
(Eq. (2.24)); the equation is then
43
- = {am2Rm(Mm2+
(2.47)
- l)}-*
Pro#erties
01
Sinusoidal Solutions
s/A,
(2.48)
r4
+ r2 IM m 2- 2 - _iRm
___
M_m 2+
}
2 n 4,=
1
pm2
{+
(2.49) r2 - 1 = 1 Mm2 i R m 2 - 1)
2
2nAm
~
kvq
R
;-(1
2
.
-am2)
1----- iRm
)+
iR;nm2
2n#,2
This equation in its present form is unwieldy, and more insight can be
gained by making various approximations first. The general case is treated
in the literature j13].
Using the definition of Am,Eq. (2.24), this formula can be rewritten
as follows:
44
W . R. SEARS A N D E. L. RESLER, J K
(2.50)
We can then treat equation (2.50) for both large and small K .
Making an expansion for small K , i.e., Rm << 2 x M m 2 ,in equation (2.50)
we find
(2.52)
r2 = 1
+ iKM,2,
1 - M,2
+ iKM,2(Mm4 - l ) a m 2 .
The first root corresponds to the positive sign in equation (2.50) and the
second to the negative sign. (We will consistently adhere to this convention
in what follows, referring to the roots as the first and the second, respectively.)
Substituting for K we have, after solving for r ,
41 a exp i A
(E)
x =F
y exp
Ay.
(The subscript 1 indicates that this is only the contribution to I/J from the
first root.)
This part of the flow therefore consists of wavy disturbances set up a t
the boundary and damped rapidly with distance from the boundary. The
signs in Eq. (2.54) must be chosen to result in positive damping; thus the
negative signs are appropriate for positive y , and the lines of constant phase
are inclined downstream from vertical a t an angle proportional to Km.
This root represents the effect of currents generated a t the boundary, which
diffuse into the stream and are carried downstream by the fluid.
Note that the damping with y depends on the wave-length; longerwave disturbances extend further out into the stream. When Rm = 0,
45
this root would indicate a harmonic disturbance pattern; this is the solution
corresponding to putting Vz# = 0 in Eri. (2.44). As mentioned, it actually
disappears in the limit Rm =- 0, for a compressible flow.
The second root is the one that, in the limit Rm = 0, reduces to the
ordinary aerodynamic solution, namely either Prandtl-Glauert or Ackeret
flow. l h e part of (I, appropriate to this root is
Consider first the supersonic case, i.e., M , greater than one. Note that
for Rm == 0 we have Ackeret flow: the streamline pattern is propagated
-~
unchanged along lines of slope d y / d x = f 1/l/Mm2
- 1, the Mach lines.
Since the slope is independent of wave-length, this result is true of the whole
disturbance pattern after Fourier superposition.
For small Rm the slope of the lines along which the disturbances stand
is not changed, but they are damped. As y increases the upper sign must
be chosen in order for the disturbance to disappear at large y , while if y
decreases the lower sign must be chosen. Thus the waves are inclined
downstream. The amplitude of the disturbance carried by these waves is
damped to l/e a t a distance y equal to
46
w.
(2.57)
Of course, as Rm -0 we have the Prandtl-Glauert solution, which can
be related to flow fields with M , = 0 by the usual transformations. The
effect of increasing Rm in this case is not to damp the waves but to incline
the lines of constant phase. Since the lower sign must be chosen for large y ,
it is seen that these lines are inclined upstream. In this case the angle of
inclination, measured from the cross-flow direction y, is proportional to the
product aW2Km,which is the so-called interaction parameter.
Qualitatively, this result tells us that the Prandtl-Glauert disturbance
pattern of subsonic flow past a body, which extends symmetrically fore and
aft in the case of a symmetrical shape, will be deflected upstream by
magnetogasdynamic effects, for small Rm.
The reader is cautioned that we cannot discuss the case M , = 0 in the
above context. That is, the phenomena of incompressible flows have been
excluded in this approximation. I t is permissible to consider small Mm2,
but only if Rm is still smaller; for M , = 0, Rm has also disappeared.
Making now an expansion for large K , i.e., Rm >> 2 n M m 2 , in equation
(2.50), we obtain
(2.58)
.u,4
iK M m 2 ( 1-A:?
Mm2-U,2
Mm2
(2.59)
I n discussing this expression we must not let Rm approach zero for a
finite M,, but we can now discuss incompressible flows, M , = 0, and in
this limit we can make Rm as small as we like. I t is important to realize
that incompressible flow a t arbitrary Rm is essentially different from flow
47
at moderate M, and small Rm. This is brought out by the fact that
Eqs. (2.59) are completely different from their counterparts in the preceding
discussion, Eqs. (2.53).
The first root for the present case can be written
(2.60)
yl =
in(1
A , 2
-M,2)l,p]
RmMm2
with
n(l
Ql oc e
x p i l { x F 2 E ( l -
(2.61)
+ Am2
- Mm2)A?a2
RmMm2
n(1
+ Am2
-M m 2 ) d ,
") I y .
RmMm2
for am2< 1 .
Thus, for am2< 1 (pmU2/2> ,uH,2/8n, i.e., dynamic pressure greater
than magnetic pressure) we must choose the upper sign to insure that the
disturbance vanishes at large y. The slope of the lines of constant phase
in the region above the body is therefore
(2.63)
The larger Rm, the smaller the slope for given am, and for given Rm,
the stronger the magnetic field the less the slope. Thus, again, the effect
of increasing conductivity is to inhibit diffusion, while the effect of increasing
magnetic-field strength is to channel the disturbance and confine it.
However, if the flow is sub-Alfvknic; i.e., am2> 1 (pmLi2/2< ,~~CcH,~/87c),
this part of Q becomes
for a m 2 > 1 .
48
Now for positive damping with increasing y the lower signs must be
chosen; thus the lines of constant phase lie along the slope
(2.65)
$2 a
exp i A ( x
_VAma
- 1 y ) exp f YAW2
-1
7d
Aw4
-___
Rm 1 - a,-2
Iy.
-1
n
tan a, R m sin4
1
- ___._
AnYt,
= 1.
l(1 - a,*-2)
49
(2.68)
This formula can be used to estimate the Rm necessary to see forwardfacing waves in the laboratory. Although the other solution,
is also
superimposed on the field, it is damped out much closer to the body for a
given Rm.
In the case that A,< 1 the form of $2 becomes
(2.69)
$,
a exp zl x
----,11
__ n
"uw4
f 1/1- Am2
y)exp f
Rm 1 - u r n ~
-___
-d m 2 2 y
for A,< 1.
To damp with positive y, we need the lower signs, and then
(2.70)
50
damped even a t small Rm,(b) the appearance of Mach (sonic) waves, diffused
by resistivity, in supersonic flow, and (c) a tendency for Prandtl-Glauert
flow patterns to be skewed upstream as Rm increases from zero in subsonic
flow.
For flows with arbitrary angle between U and B,, we may anticipate
that these effects will have their counterparts, except, of course, that the
boundary-layer phenomenon will be replaced by another family of diffuse
standing waves. To understand in more detail the effects of varying Rm
on these patterns in compressible conductors will require further investigation.
111. HALLEFFECTS
1. Particle Motions
(3.2)
dV
m -itt
= e(E
+ V x B)
Taking the scalar components of this vector equation one finds that
the motion of the particle in the z or magnetic-field direction is independent
of the presence of B. One also finds
t:)
u p = k,cos - t
(3.3)
fp = - k ,
sin t:t)
t:)
+&sin - t
+ k, cos t:t) + C
(3.4)
m Va
-=
eVB
51
eB
or
eE,
up = -sin
mw
ot,
(3.5)
eE,
mw
vp = ---cosot
- _eE,
_.
mw
F(t) = e-:I7.
(3.6)
(3.7)
up(t)neF(t)dt = t up(t)e- dt
0
and similarly for 5. Using the expressions given in Eq. (3.5) for u&) and
vp(t), we find
52
j x = axxEx= npzi,
(3.10)
j y = ayxEx= np6,
(3.12)
t = mean
ti = mean
t,= mean
(3.13)
53
Here*
K
= l/W,t,
Kc
= IlWeTe,
Ki
we = eB/m,,
m, = mass of electron,
w, = eB/mi,
m; = mass of ion,
+ n),
9, = electron pressure.
f = n/(n,
= K,/(K;
= 1f2WiTi
KJ,
--
Note that in all cases K , / K ~ = Vme/m,;this is a very small number and will
be neglected in what follows.
It is not often that the equation above, Eq. (3.13), must be used in
its entirety. The following forms are often used:
Slightly Ionized Gas with Ion Slip: Here the electron pressure gradient Vp8
may be neglected; also n,/n (< 1.
(3.14)
where a again denotes the electrical conductivity of the gas; for this case,
where there are three kinds of particles, it is given by n,e2t,,,/m,where 5,
is the mean collision time for electrons in collisions with neutrals and
ions, viz.,
(3.14a)
T.=(;+
);
-1
The last term in Eq. (3.14) represents the phenomenon called ion slip,
which arises from the fact that the ions and neutrals are not perfectly coupled.
Thus, if the mean time between collisions of these species is appreciable,
there is an imperfect transfer of electromagnetic effects from the charged
particles t o the neutrals that make up the bulk of the gas.
Slightly Ionized Gas without Ion Slip: This is the case in which the number
of charged particles is so small compared to the number of neutrals that
collisions between electrons and ions are very rare (t oo),whereas collisions
with neutrals are frequent. The ion-slip term is then negligible compared
to the j x B term in Eq. (3.14). The result is
-.
(3.15)
OeT,
j = a ( E + q x B ) - - j xBB ,
* We retain the factor 1/2 in the definition of K; to conform with [65]. Some authors
do not retain it, and care must be used in comparing results.
54
tc=ti=00.
(3.16)
where u denotes n,e%/m,. This is the same result as Eqs. (3.10) and (3.11),
provided that the electric field due to electron pressure gradient is included
in the definition of E.
Note the similarity between Ohms Laws for fully ionized and partially
ionized gases : they differ by the presence of the electron-pressure gradient
term and the definition of u.
Although the term tensor conductivity is used quite freely, these
equations (3.14)-(3.16) are vector equations and should be used as such.
If this is not done the tensor conductivity changes with the boundary
conditions.
The tendency for the electric current to flow across the direction of the
electric field, under the influence of the magnetic field, is called the Hall
Effect.
The electrical conductivity of an ionized gas is a measure of one of the
transport properties. Other transport properties are similarly affected by
the magnetic field. The flux of the transported property is reduced if it
is carried by charged particles in the presence of a magnetic field. As in
charge transport, the transport in a direction across a magnetic field is
reduced by a factor (1 ~ , ~ int a~fully
) -ionized
~
gas, and an induced
~
perpendicular to both the
flux proportional to o,t(l4-w , T ~ ) -appears,
gradient of the property driving the flux and the magnetic field. For example,
if R is the usual heat conductivity of electrons in a fully ionized gas with no
magnetic field applied, the heat flux in the presence of a magnetic field is
given by [67]
(3.17)
3. Aligned-Fields Flow of a Fully Ionized Gas or a Slightly Ionized Gas Without
Ion Slip
In Chapters I and I1 we treated only flows of fluids having scalar electrical
conductivity u. We shall now develop the corresponding theory for ionized
gases, in particular for fully ionized gases and for slightly ionized gases
without ion slip. It has already been noted that their Ohms Laws, Eqs. (3.16)
and (3.16) are nearly the same. Under conditions such that the tensor
55
character of the conductivity disappears, i.e., when the Hall Effect terms
in Eqs. (3.15) and (3.16) are negligible, the results will reduce to those
previously obtained. Conditions under which the Hall Effect dominates the
flow will also be defined.
Once again we shall consider small-perturbation flow with uniform aligned
fields I! and B, a t large distances from a slender body.
As before (equation (2.14)) the proportionality between the vorticity
and the current density in small-perturbation flow still holds true as well as
the linearized Bernoulli's equation. These were both derived from the
momentum equation. Taking the curl of the appropriate Ohm's Law and
using the isentropic energy relations, which are still appropriate, one obtains
(3.18)
av
aui
aB
a B , - - uB,M,2-- all - = curl j
ax
ax
ax
aj
+ W,T ax
(3.18)
V 2 - 4np0U(1 - U,
56
One often hears the term collisionless plasma used to denote highly
conducting plasmas. Rut a collisionless plasma is one in which a is large
because the time between collisions t is large; hence it is not the same as
the ideal conductors considered in Chapters I and 11. Suppose we now
consider this case, t -+ 00. From Eq. (3.22) we have, noting that Km is
proportional to t, a = ne2t/m., and p = nm,,
(3.23)
where again w , denotes the ion cyclotron frequency eB,/m,.
57
The quantity li so defined has been called the characteristic ion Larmor
radius; it is seen to be actually a material property. The quantity up
is the plasma frequency. The quantity 1, depends on fundamental constants
as well as mi and n ; substituting for these constants,
(3.25)
Zi2=
Z
-5.15 x
IOl4
(cm2)
Z in this equation is the atomic weight of the ion. Thus this distance is small
for high-density plasmas.
we have for
(3.27)
kl2r4
y2,
W. R. SEARS A N D E. L. RESLER, J R
(3.30)
I .
59
(3.32)
Again we can eliminate div v by using the mass-conservation equation and
the isentropic relation between p and p to eliminate the density:
(3.33)
laking the curl of the linearized equation of momentum, including the
time-dependent term, we have
(3.34)
Substituting these results into Eq. (3.32), after differentiation with respect
to time, one has
(3.36)
60
To eliminate 9, take the divergence of the linearized equation of momentum and find
(3.36)
Finally, eliminating fi between (3.35) and (3.36),we have as an equation for j,
angle with respect to this axis and therefore with the plane-wave normal
n,,, (Fig. 3.1).
Suppose that j is proportional to exp {ill - z f y / c } ; since the x - and
z-derivatives are zero, we find
AY2= pHY2/4nprn,
A x 2 = pHX2/4np,,
miy
= eHyp/m,.
If the scalar components of this equation are written out, the y component
is uncoupled :
(3.39)
61
For any given situation, i.e., given am2, A2, pa, and oiy, the wavespeed diagram can be used to construct a disturbance-shape diagram
analogous to Fig. 2.6, but, in general, such a diagram now pertains only
to a given frequency f. In other words, it tells us the shape of the propagating constant-phase loci due to a steady-state, oscillatory, point disturbance of frequency f at the origin. The motion of a solid body is equivalent,
through Fourier superposition, to combined oscillatory disturbances of many
frequencies, and each of these has a different diagram. Obviously the result
of superposition requires a good deal of calculation, but one expects
qualitative results to be similar to the effects at some dominant frequency,
probably of the order of magnitude UIL. Thus it can be expected that the
Mach cones of three-dimensional large-% flow (Chapter 11) will be
broadened as well as rendered somewhat indistinct by diffusion. The two
narrow, cusped cones of Cumberbatch will be connected by a thin, flat
sheet of diffusion lying in the plane of lJ and B,.
In the first two chapters of this presentation the treatment was for
frequencies / small compared with the ion cyclotron frequency wiy, so that
the right-hand side of equation (3.40) can be put equal to zero; thus we obtain
(unless
(3.41)
and
am2 Ay2
if )
+ + 4npa
Ax2
c2
+ am2Ay2+ ifam2
-= 0.
and
(3.44)
where A 2 = Ax2
+ Ay2.
iy = 0)
4npua
62
Of course, for uam2 -+ bo we get only the van de Hulst roots mentioned
before. For other limiting cases, e.g. A = 0,one gets ordinary sound waves
00 or incompressible
and current diffusion (skin-depth phenomena) ; for a,
flow, A l f v h waves and again current diffusion. Notice that again it is the
quantity ua,2 - proportional to Rm/Ma2 of Chapter I1 - that appears,
and care must be taken in going to various limits.
In any but the limiting cases named, c as given by Eqs. (3.43) and (3.44)
is complex. The phase velocity to be plotted on a Friedrichs-type diagram
is then the reciprocal of the real part of c-l. If the phase velocity is plotted
as a function of the angle pl for different values of fluam2,one finds that,
as the quantity f/uam2goes to infinity, the outer curve, of om1 shape, always
FIG.3.2. Wave-speed diagrams (fast, slow, and intermediate phase speeds vs. wave
inclination) for plasma without Hall Effect for steady oscillation at frequency /. The
numbers on the curves are values of //4npa&. When this parameter is infinite the
slow- and intermediate-wave diagrams are circles of infinite radius. urn = 1/2A
.,,,
approaches the sonic circle, while the inner curves, having the shape of the
cu sign, approach a circle of infinite radius. The circle of infinite radius
occurs because we have not retained relativistic terms in our equations ;
if these terms were retained, the figures M would approach a circle with
radius of the speed of light.
In Fig. 3.2 are plotted the phase-velocity curves for a gas without
Hall Effect but with several values of the resistivity parameter / / 4 y ~ u a , ~ .
These will give the reader some idea of how the Friedrichs Diagram is affected
by resistivity. (See also [70].)
As a second example, let us consider a so-called cold coIlisionless plasma.
This is the case represented by uoo= 0 and u = OQ in Eqs. (3.39) and (3.40).
In this case the equations become
(3.45)
c 2 - A Y 2 -- 0
63
and
c2
- A2 cosz f$ = 0
and
(3.48)
c4 - c2A2(1
+ C O S ~p) + A 4
C O S ~p
f 2 A 4 C O S ~p = 0.
-w'lY
In this case, even though u = do, the fast and slow phase velocities are
frequency-dependent. Comparison with other references can be made by
noting that
(3.49)
FIG. 3.3. Wave-speed diagrams (fast, slow, and intermediate phase speeds vs. wave
inclination) for cold collisionless plasma, for steady oscillation at frequency /. The
numbers on the curves are values of ( f / ~ i ~ )The
* . intermediate-wave diagram is the
same for all values of this parameter.
(3.50)
64
This equation can be compared with the results given in Reference 1711.
In Fig. 3.3 are plotted phase-velocity curves for the cold collisionless plasma
for several values of the frequency parameter ( f / w i J 2 .
6. Summary
In this chapter we have shown that the Hall Effect may be important
even in gases with very large electrical conductivity. If the high value of u
occurs because the plasma is rarefied, i.e., the collision time t is large, the
result is that the Hall phenomena remain appreciable. They are manifested,
for example, in alterations of the flow pattern near the solid-fluid interface
and deflection of magneto-sonic waves towards ordinary Mach angles,
as well as the appearance of cross-flow.
On the other hand, if large conductivity is achieved by making the
electron density large in a gas of high density, the Hall term becomes
negligible, and the model used in the earlier chapters is appropriate.
The case of an ionized gas with arbitrary conductivity and Hall Effect
is relatively intractable. Some qualitative information can be obtained by
considering the propagation of sinusoidal plane waves in such a fluid, leading
to a generalization of the wai.e-propagation and disturbance-shape figures.
The relationship of these to the disturbances set up by the steady motion
of bodies is not so close as before, however, since the pertinent effects of
resistivity and Hall phenomenon are frequency-dependent.
References
1. LAMB,H., Hydrodynamics. Cambridge, 1932. 32R. Eq. ( 8 ) .
2. ELSASSER,
W. hI.. Hydromagnetic dynamo theory, Rev. Mod. P k y s . 28, 135-163
(1956).
3. A L F V ~ N H.,
,
On the existence of electromagnetic-hydrodynamic waves, A rkiv /or
Mafenzatik Asfronomi och F y s i k , 2BH, No. 2 (1943) ; see also Nalure 160, 405-406
(1942).
4. SEARS,W. R . , and RESLER,E. L., J R . , Theory of thin airfoils in fluids of high
electrical conductivity, J . Fluid Mecli. 6 , 257-273 (1959).
5. LUDFORD,
G. S. S., Note on airfoil theory in hydromagnetics, J . Aero. Sci. 2H.
511-512 (1961).
8. LUDFORD,
G. S. S., Further note on airfoil theory in hydromagnetics. J . .4ero. Sci.
28, 741-742 (1961).
7. SUNG.K. S., Linearized Magneto-gas Dynamics for Arbitrary A l f v h Number,
Mach Number, and Magnetic Reynolds Number. Ph. D. Thesis, Harvard
University, Cambridge, Mass. (1903).
8. CUMBERBATCH,
E., SARASON,
L., and WEITZMER,
H., Magnehohydrodynamic flow
past a thin airfoil, A . Z . A . A . Journal 1, 079-690 (1963).
9. STEWARTSOS,
K., Magneto-fluid dynamics of thin bodies in oblique fields I, Zetfs.
1. angew. M a t h . und Phys. 12, 261-271 (1961).
65
11). HASIMOTO.
H., Magnetohydrodynamic wakes in a viscous conducting fluid, Rev.
Mod. Phys. 82, 860-866 (1960) : also Magneto-Fluid Dynamics (F. N. Frenkiel
and W. R . Sears, eds.), Natl. Acad. of Sci.-Natl. Kes. Council, Publ. No. 829, 1960.
11. CLAUSER,F. H . , Concept of field modes and the behavior of the magnetohydrodynamic field, The Physics of Fluids, 6 , 231-253 (1963).
12. RESLER,E. L., J R . and MCCUNE. J. E., Electromagnetic interaction with aerodynamic flows, i n The Magnetodynamics of Conducting Fluids (D. Rershader,
ed.), pp. 120-135, Stanford, 1959.
13. RESLER,E. L., J R . and MCCUNB,J . E., Some exact solutions in linearized magnetoaerodynamics for arbitrary magnetic Reynolds numbers, Rev. Mod. P h y s . 82,
848-854 (1960); also Magneto-Fluid Dynamics (F. N. Frenkiel and W. R . Sears,
eds.), Natl. Acad. of Sci.-Natl. Res. Council, Publ. No. 829, 1960.
14. MCCWNE,J . E., On the motion of thin airfoils in fluids of finite electrical concluctivity, J . Fluid Mech. 7 , 449-468 (1960).
1.5. LUDFORD,
G. S. S., Inviscid flow past a body at low magnetic Reynolds number,
Re?,. Mod. P h y s . 82, 1000-1003 (1960); also Magneto-Fluid Dynamics (F. N .
Frenkiel and W. R. Sears, eds.), Natl. Acad. of Sci.-Katl. Res. Council, Publ.
No. 829, 1960.
16. LUDFORD,
G. S. S. and M U R R A Y ,J . D.. On the flow of a conducting fluid past a
magnetized sphere, J . Fluid Mech. 7 , 516-528 (1960).
17. CHESTER,
W., The effect of a magnetic field on Stokes flow in a conducting fluid,
J . Fluid Mech. 3, 304-308 (1957).
18. CHESTER,W., On Oseens approximation, J . Fluid Mech. 13, 557-569 (1962).
19. LUDFORD,
G. S. S., The effect of an aligned magnetic field on Oseen flow of a conducting fluid, Arch. Rational Mech. 4, 405-41 1 (1960).
20. T A M A D AKO.,
,
Flow of a slightly conducting fluid past a circular cylinder with
strong, aligned magnetic field, Phys. of Fluids, 6 , 817-823 (1962).
21. Rossow, V. J . , On the flow of electrically condocting fluids over a flat plate in
the presence of a transverse magnetic field, N A C A T N No. 3971 (1957).
22. Rossow. V. J., On magneto-aerodynamic boundary layers, Zeits. /. angew. Math.
und P h y s . Bb, 519-527 (1958).
K., Magneto-fluid dynamics of thin bodies in oblique fields 11.
23. STEWARTSDN,
Zeals. f . angew. Math. und P h y s . 18, 242;255 (1962).
24. HASIMOTO,
H., Viscous flow of a perfectly conducting fluid with a frozen magnetic
field, P h y s . of Fluids 2, 337-338 (1959).
25. SEARS,W. R., On a boundary-layer phenomenon in magneto-fluid dynamics,
Astronautica Acta 7 , 223-236 (1961).
26. LYKOUDIS,
P. S., On a class of compressible laminar boundary layers with pressure
gradient for an electrically conducting fluid in the presence of a magnetic field,
Proc. I X t h Internatl. .4stronautical C o n t r . , Amsterdam (l958), 168-180, Springer,
Vienna (1959).
27. SEARS,M:. R. and MORI, Y,,Studies of the inviscid boundary layer of magnetohydrodynamics, i n Progress in Applied Mechanics (The William Prager
Anniversary Volume). (D. C. Drucker, ed.), Macmillan, New York, 1963.
28. LEWELLEN,W. S., An Inviscid Boundary Layer of Magnetohydrodynamics,
M. Aero. E. Thesis, Cornell University, Ithaca, New York (1959): also A F O S R TN-59-927.
H. P. and CARRIER,
G. F., The. magnetohydrodynamic flow past a
29. GREENSPAN,
flat plate, J . Fluid Mech. 6 , 77-96 (1959).
30. GREENSPAN,
H. P., Flat plate drag in magnetohydrodynamic flow, Phys. of Fluids
8, 581-588 (1960).
31. GOURDINE,
M. C., On Magnetohydrodynamic Flow over Solids, Ph. D. Thesis,
v
Calif. lnst. of Tech., Pasadena, Calif. (1960). See also On the role of ? J ~ S C O S l t and
66
32.
33.
34.
38.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49,
67
50. WEITZNER.H., On the Greens Function for Two-Dimensional Magnetoliytlrodynamic Waves. I I , Inst. of Math. Sci.. New York University, NYO-948!). I!J60.
51. CHU, C. K . and L Y N N ,Y. M . , Steady magnetohydrodynamic flow past a nonconducting wedge A . I . A . A . Journal 1, 1062-1067 (1963).
.id. CUMBERBATCH.
E., Magnetohydrodynamic Mach cones, .I
.4ero.
. Sci. 89, 1476 1471)
(1 962).
.53. HELFER.H . L., Magnetohytlrodynamic sliock waves, Asfrophysical J . 117, l77-1!1!)
( 1 953).
54. LUST, H . , Magneto-hydrodynaniische StoUwellen in einem Ilasma rinentllicher
Leitfahigkeit, Zeits. /. Nafurforsch. 8 A . 377-284 ( 1953).
5 5 . BAZER,J. and ERICSON,W. B.. Hydromagnetic shocks. .4strophysacal ,/. 1811.
758-785 (1959).
56. ERICSON,W. B. and BAZER,J., On certain properties of hydromagnetic shocks,
Phys. of Fluids 3, 631-640 (1960).
J . A., One-dimensional magnetogasdynamics in oblique fields, ,/. F l u i d
57. SHERCLIFF.
Mech. 9, 481-505 (1960).
5 8 . CABANNES,
H., Attached stationary shock waves in ionizetl gases, R e v . M o d . Ihys.
32, 973-976 (1960) ; also Magneto-Fluid Dynamics (F. N. Frenkiel ancl
W. R. Sears, eds.), Natl. Acad. of Sci.-Natl. Res. Council, Publ. No. 82!), 1960.
59. GERMAIN,
P., Shock waves and shock-wave structure in magneto-fluid dynamics.
Rev. Mod. P h y s . 32, 951-958 (1960) ; also blagneto-Fluid Dynamics
(F. N. Frenkiel and W. R. Sears, eds.), Natl. Acad. of Sci.-Natl. l<es. Council,
Publ. No. 829, 1960.
G. I . . and POLOVIN,
I<. V., The stability of shock
60. AKHIEZER, A. I . , LIURARSKII,
waves in magnetohydrodynamics, J . E x p t l . 7heoref Ihys. 36, 731-737 ( I 9 R H ) ;
Translated in Soviet P h y s . J E T P . H. 507-511 (1959).
61. TANIUTI,
T., A note on the evolutionary condition on hydromagnetic shocks,
Progv. Theor. P h y s . 28, 756-757 (1962).
62. GEFFEN,N . , On Aligned-fields Magnetogasdynamic Flows With Shocks, Ih. I).
Thesis, Cornell University, Ithaca, New York, l!)63; also available from University Microfilms, Ann Arbor, Michigan.
63. SEARS,W. R.. High-field effects in magnetohydrotlynamics, i n High Magnetic
Fields (H. Kolm, B. Lax, R Bitter and I<. Mills, eds.). 652-658, MIT Press
and J . Wiley, New York, 1962.
64. SEARS,W. R., Some paradoxes of sub-Xlfvenic flow of a compressible conducting
fluid, i n Proc. Symposium Electyomagnetics and Ffuid Dynumics of Gaseous Plusma
363-372, Polytechnic Press, Brooklyn, 1962.
T . G., Magnetohydrodynamics, Interscience, New York, 1957.
65. COWLING,
A , , Electrical conductivity of highly
66. LIN, S. C., RESLER,E. L. and KANTROWITZ,
ionized argon produced by shock waves. ,journal of A p p f i r d Physics 28, 95-10!)
(1955).
67. CHAPMAN,
S. and COWLING,T. G., The Mathematical Theory of Non-Uniform
Gases, Cambridge University Press, p. 337, 1952.
68. RESLER,E. L., J R . , Aerodynamic Aspects of Magnetohydrodynamics i n Developments in Mechanics, Vol. 1, pp. 503-509, Plenum Press, New York, 1961.
Proceedings of 7th Midwestern Mechanics Conference, East Lansing, Mich .,
Sept. 1961. (Lay, J. E. and Malvern. L. E., eds.).
69. SONNERUP,
B., Some effects of tensor conductivity in magnetoliydrodynamics,
J . Aevo. Sci. 28, 612-621, 643-644 (1961).
70. KESLER.
E. L., JR., Some remarks on hydromagnetic waves for finite conductivity,
Rev. M o d . P h y s . 82, 866-867 (1960); also Magneto-Fluid Dynamics (F. N.
Frenkiel and W. R. Sears, eds.). Natl. Acad. of Sci.-Natl. Res. Council, Pnhl.
No. 829, 1960.
68
71. FISHMAS,
F. J., KANTROWITZ,
.I. H. antl PETSCHEK,
H . E., Magnetohydrodynamic
72. XAPOLITANO,
L. G. and COXTcRSl, C;., "Magneto-Fluid Dynamics: Current Papers
and .Abstracts", AC.4 RI) Bibliography I, Enlarged Edition. I'ergarnon, Oxford,
London, S e w York. antl Paris. 1962.
73. Par, S H I H - I ."Magnetogasdynamics
,
and Plasma I)!~namics". Springer, Vienna. 1962.
AND
BERNARD D. COLEMAN
I. Introduction . . . . . . . . . . . . .
11. Relation to General Simple Fluids . . .
111. Steady Simple Shearing Flow . . . . .
11'. Viscometric Flows . . . . . . . . . .
1. Steady Couette Flow . . . . . . . .
2. Axial Flow between Concentric Pipes
3. Approximate Results . . . . . . . .
Steady torsional flow . . . . .
Steady cone and plate flow . .
4. Appendix . . . . . . . . . . . . .
V. Steady Extension of a Cylinder . . . .
VI. Relation t o Classical Viscoelasticity . .
V I I . Nonsteady Shearing Flows . . . . . .
1. Sinusoidal Oscillations . . . . . . .
References . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Page
69
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
71
77
79
79
82
84
84
85
86
87
89
96
98
,100
I, INTRODUCTION
An iiicompressible simple material is a substance whose mass density
never changes and for which the stress is determined, to within an arbitrary
hydrostatic pressure, by the history of the gradient of the deformation. An
incomfiressible simple fluid is an incompressible simple material with the
property that all local configurations are equivalent in response, with all
observable differences in response being due to definite differences in
history. *
In recent years much has been learned about the mechanical behavior
of general incompressible simple fluids. It has been found that there are
several boundary-value problems which can be solved for incompressible
simple fluids in general, using no special constitutive assumptions beyond the
definition of fluidity. Yet there are many important hydrodynamical
problems which can be discussed in great detail for incompressible Newtonian
fluids, sometimes even to within explicit calculation of the velocity field,
____-
70
but about which little can he said for general simple fluids. Indeed, when
one considers that whereas a Newtonian fluid is defined by a linear isotropic
function of a tensor, a general simple fluid is defined by an isotropic functional (which need not be linear) of a tensor-valued function, it is rather
surprising that any problems can he solved for general simple fluids.
If one adds to the definition of a-simple fluid a smoothness postulate
which makes mathematical the physical notion of gradually fading memory,
i.e. the notion that deformations which occurred in the distant past should
have less effect on the present value of the stress than deformations which
occurred in the recent past, then one can prove theorems justifying the
intuitive idea that for slow motions the local behavior of a simple fluid
should be approximated by the behavior of a Newtonian fluid [2]. I n other
words, on assuming fading memory one can prove that the theory of
Kewtonian fluids gives to the theory of perfect fluids a correction for viscoelastic effects which is complete to within terms of order greater than
m e in the time scale. The theorems which yield this result also enable u s
to find a constitutive equation which gives corrections for viscoelastic effects
that are complete to within terms of order greater than two in the time
scale :* in Cartesian components that constitutive equation reads as follows:t
(1.2a)
-4 (1)1, = U f , ,
(l.2b)
A(NJr7 = A ( N - 1 J r k U k . j
tj.1
+ A(N
-1)lkvk.i
A(N-l)l7;
71
11. RELATION
TO GENERAL
SIMPLEFLUIDS
Consider a material point P in a deformable body and let x i , j = 1, 2, 3,
be the jth Cartesian coordinate of P at time t. Let ~ ( ~ ) t~ ( xs )~ give
,
the
ith Cartesian coordinate of P at time t - s, 0 s < 00; ~ ( ~ ) i) (is called
the relative deformation function for P . The tensor FM)ij(t- s ) , defined by
<
72
H E R S H E L MARKOVITZ A N D B E R N A R D D. COLEMAN
det
(F(t)ij(t- S ) ) =
1.
(2.3)
where p ( t ) is the indeterminate hydrostatic pressure and Gjj is a symmetrictensor-valued functional of the function F(t)al(t- .). In (2.3) Ck&) is some
tensor-valued measure of strain from the configuration a t time t to a special
time-independent reference configuration. The functional Gii is said to
characterize the material a t P. If Gii is independent of the strain
parameter Cii(t), then we say that the material described by (2.3) is a simple
fluid. Hence, for a simple fluid
where G,, is a functional which depends on only the point P under consideration. In words, (2.4) states that if the deformation gradient, relative to the
present configuration as reference, is known for all past configurations,
then the extra-stress, S,j psi,, is determined; the operator G,,, which so
determines the stress involves no preferred configurations and is the same
operator whatever the present configuration may be.
and aij is an isotropic functional; i.e., for every orthogonal tensor Qii, g,j
obeys the identity
73
det [ (GkI(s))
+ &I]
= 1,
<s <
do.
We assume that .9 is the set of all histories Gkl( ) with finite norm llGkl( ) / I ;
then we can say that 9,together with the norm (2.9),is a Hilbert space, H.
We regard the norm of a history in H as a measure of the total intensity
of all the past strains (relative to present configuration) at a material point P.
I t will be observed that jlGkl( )I1 is defined in such a way that it places
greater emphasis on the values Gkl(s)for small s (recent past) than on values
for very large s (distant past).
If a neighborhood about P has never been distorted, i.e. has been in its
present configuration (to within a rigid rotation) a t all times in the past,
then the history Gkl( ) of P reduces to
(2.10)
GhI(S)
= 0,
<s <
~ i ) .
74
The function (2.10) is called the rest history. We note the norm of the rest
history is zero and the norm of an arbitrary history G k l ( ) may be interpreted
as the distance of Ghl( ) from the rest history.
A functional $j,,is said to be a memory junctional (of order n ) if $,i has
n) a t the point in X corFrCchet functional derivatives (of all orders
responding to the rest history (2.10) and if these functional derivatives obey
certain technical requirements of continuity. For our present purposes it
suffices to put n = 3, and we say that a simple fluid has gradually f a d i q
memory if $j,i in (2.5) is a memory functional of order three.
Not all simple fluids have gradually fading memory. In fact there is an
important class of simple fluids, called Rivlin-Ericksen fluids, which do not.
To discuss these fluids we need to consider the Rivlin-Ericksen tensors
which were defined in (1.2) but which can be shown to be also given by*
<
(2.11)
In 1955 Rivlin and Ericksen 131 considered the theory of isotropic materials
for which the stress depends on only the spatial gradients of the velocity,
acceleration, second acceleration, . . . , (M - 1)th acceleration. They showed
that for an incompressible material of this type, the extra-stress Si,f pS,,
must be given by an isotropic function of the tensors
N = 1, 2,. . . , M ;
i.e.
(4.12a)
sij
+ psi,
( l j k d (2)kl.
. . ., A ( M ) k l )
(2.12b)
, A(M)kI)
= hii(QrmQlnA
(ljmni Q k m Q d (2)mn.s
,QkmQlnA (Mjmn).
(2.13)
Sbj
+ PSij = qoA(l)ij
aA(l)tkA(l)ki
* Equation (2.11) Seems to be due to Sol1 [I]; Hivlin and Ericksen start with (1.2)
as their definition. Other kinematical formulae for the tensors A(,)ij, a summary of
their history, and a proof of the equivalence of (2.11) and (1.2) are given in the
encyclopedia article of Truesdell and Toupin [7].
75
A theory based on the general equations (2.5) and (2.7) of a simple fluid
is compatible with stress relaxation. In fact, it appears to us that (2.5)
and (2.7) express, for materials which can be considered incompressible,
a concept of fluidity sufficiently general to cover almost all purely mechanical
effects in real fluids.+
The Rivlin-Ericksen fluid corresponds to that special case of (2.5) and
(2.7) in which $jijis such that its value is determined by a finite number of
derivatives of G&) a t s = 0. If we restrict our attention to only those
histories Gki( ) which are polynomials in s, then it is true that (2.5) and (2.7)
imply (2.12) for all functionals $ i j . We cannot, however, make such a
restriction, because there are many important physical situations in which
Gkl( ) is not a polynomial (for example, stress-relaxation experiments).
Yet this observation does not give the whole story of the logical relation
of Rivlin-Ericksen fluids to general simple fluids. For, in the limit of very
slow motions, general simple fluids with fading memory are approximated,
albeit in a rather drastic sense, by certain particular Rivlin-Ericksen fluids,
even when Gki( ) is not a polynomial. We now make this remark more
precise.
Let Gij( ) be any history; from it we can construct for each a, 0 < a
a new history G$( ) as follows
(2.15)
G$(s) = Gjj(as),
0 \< s <
< 1,
oc,;
76
+ ps,,
(a) +
= 7?04(h,
p ~ ; d
~
1)rk
+ YAI$r, + O(a3).
( 2 1
(IJkl
Equation (2.17) tells us that, for a simple fluid with fading memory, the
equation (2.13) of a Newtonian fluid furnishes a complete first-order correction, for viscoelastic effects, to the theory of perfect fluids in the limit of
slow motions. Equation (2.18) exhibits all second-order corrections for
viscoelastic effects.* In (2.17) and (2.18), q0, ,8, and y are material constaizts,
i.e. numbers depending on only the functional sj,, defining the simple fluid
under consideration.
The constitutive equation obtained by striking out the error term O(a3)
in (2.18) is just our equation ( l . l ) ,i.e. the constitutive equation of a secondorder fluid.
I t is, of course, a special case of (2.12) and hence does not
describe (exactly) a fluid with gradually fading memory, yet it approximates
such fluids in sufficiently slow flows.
It is now clear that the theorem leading to (2.18) justifies our assertion
that ( 1 . 1 ) gives corrections for viscoelastic effects that are complete to
within terms of order greater than two in the time scale. Kote that the
theory of Reiner-Rivlin fluids (2.14) would not yield the term in (2.18)
with coefficient y ; hence that theory cannot give a complete second-order
correction to the Xewtonian fluid.
(1963).
77
111. STEADYSIMPLE
SHEAKINC
FLOW
Steady simple shearing flow is a flow for which there exists a Cartesian
coordinate system xl,xz,x3 in which the components of the velocity field zli
have the form
(3.1)
v1
= 0,
v2 = v ( x , ) ,
v3 = 0.
- A(1)kIS
G(t)kZ(S) =
4A[2)kfs2,
A(.L)k[,
relative
(3.4)
(Sij) ==
s33
where
(3.5a)
(3.5b)
(3.5c)
Let us put
(3.6)
q(K)
==
T(K)/K;
q(K) =
?(-
K),
K:
Ui(K)
=a;(-
K).
78
Moreover, the functions q ,oz,q (and hence t) depend only on the material
under consideration, i.e. are determined by the functional B,,of Eq. (2.5).
For this reason (I,,
u2, and q are called material junctions. Of course, q is
just the shear-dependent viscosity of the rheological literature, while
u1 and gZ govern the normal stress effects.
We turn now to the behavior of a second-order fluid in steady simple
shearing flow. On substituting (3.3) into (1.1) we find that for a secondorder fluid
(3.8a)
(3.8b)
V(K)
vo,
(3.9b)
Ul(K)
(p -k 2 Y ) K 2 ,
(3.9c)
Dz(K)
= 8..
i.e.
T(K)
=V
~ K ,
- .s,
It follows from Eqs. (2.18) and (3.!1) that whenever a general simple
fluid with fading memory is approximated in a slow flow by a second-order
fluid, the material constants qo, P , y, of the second-order fluid must be
related as follows to the material functions q , u,, of the simple fluid [lo; :
(I,
(3.10a)
. --=
d4
qo = Iim
r+O
~(0)
79
Equations (3.10) tell us that a knowledge of 17, ul, and u, for a simple fluid
with fading memory suffices to determine the constants yo, p, and y and
hence suffices to completely determine the second-order fluid which approximates the simple fluid in slow flows.
Because of Eq. (3.10a), ljro is called the zero-shear-rate viscosity. Since
there is a large literature on the experimental determination of V ( K ) for
small K, and hence of q(O),*having now identified q,, with q(0) we can leave
consideration of this material constant and turn to a detailed discussion
of p and y .
IV. VISCOMETRICFLOWS
A motion is called a (global) viscometric flow j15] if, for each material
point P and each time t, G,,,( ) has the form (3.2) and there exists an
orthonormal basis e i ( P , t ) , i = 1 , 2 , 3, relative to which the components
have the forms (3.3) with K independent of t and depending
A(l)kt,
only on P; here e , ( P , t ) may, but need not, be the physical basis of an
orthogonal coordinate system.
Examples of viscometric flows are steady simple shearing flow, steady
Couette flow, steady Poiseuille flow, and steady helical flow in an annulus.
Viscometric flows are of interest to us here because they constitute the
class of flows for which the stress Sii in a simple fluid is determined by the
same three material functions t, (I,, and 0, which determine Sii in a steady
simple shearing flow. In fact, for any simple fluid in a viscometric flow the
matrix of the components of the stress tensor relative to the basis ei(P,t)
has the form (3.4) with elements obeying (3.5).+ Because of this the theory
of viscometric flows is a manageable theory in which one can actually solve
hydrodynamical problems for simple fluids without specializing the functional $jij [II, 171. We now consider some results in this theory which
permit the determination of the material functions u, and u2 and hence of
the material constants and y.
1. Steady Couette Flow
80
axis of the cylinders. The physical components v i of the velocity field relative
to r,O,z are assumed to have the form
(4.1)
v' = 0,
YO
= YW(Y),
v' = 0 ,
M is the torque per unit height required to maintain the motion. Assuming
that the fluid adheres to the cylindrical walls (which are taken to be very
long), we have the boundary conditions
w(R,)= ill,
(4.3)
w(R2)= Q2.
(4.4a)
(4.4b)
(4.44
where S,, S,,, So,, and SIB are physical components: all other physical
components of the stress tensor are zero.
It follows from (3.9a) that for a second-order fluid
(4.5)
(4.6)
RZ2- R 1
We note that in deriving (4.6) we use only the function t which is the same
for Newtonian fluids and second-order fluids. Hence, we are not surprised
81
that (4.6) agrees with results known to follow from the Navier-Stokes
equations* and yields the familiar formula
(4.7)
Q2
- 62, =
-i'&)'
M
47%
R,
(4.8)
where p is the mass density. It follows from (3.9) and (4.5) that for a secondorder fluid
(4.9a)
and
(4.9h)
On substituting (4.6) and (4.9) into (4.8) we find that the integrations in
(4.8) can be performed explicitly. The result is
where 9 is the contribution of the inertia of the fluid to the stress difference.
For given values of R,, Sl,, R,, R2 and p , 9 is the same for h'ewtonian fluids
and second-order fluids and is independent of ?j:'
(4.10b)
**
82
For a general simple fluid with fading memory, Eqs. (4.10) give an expression
for S,(R,) - S,,(R,) which is valid for small angular velocity difference
& - R, and which involves an error of order O((R, Hence if we
measure S,,(R,) - S,(R,) as a function of R, - R, for a simple fluid, in an
apparatus of fixed geometry, we can evaluate y :
We remark that when the gap between the cylinders is small, or, more
precisely, if we keep R, and L2, fixed and let K ,
K,,
R,, in such a
way that R,(R, - LZ,)/(R, -- K , ) remains fixed (which means, approximately, that the rate of shear is kept fixed), then Eq. (4.10) is approximated by
-+
-+
Sw(R2) (4.12)
where the first term on the right arises from the inertia and vanishes i f
either of the cylinders is stationary.' The second term is written in its
present form to emphasize that, for a second-order fluid, S,,(R,) - S,,(R,)
goes to zero linearly in ( R , -- K , ) / K , if (R,-- H , ) / K , is made to approach
zero a t fixed R,(R, - iZl)/(R2 - Rl).
z" = 0,
1''
= V(Y),
LIB = 0 ,
/I(
H,) = U( R,) = 0.
Like steady Couette flow, this viscometric flow is a special case of steady
helical flow in an annulus [16] which has been shown, in the theory of simple
fluids, to be compatible with the equations of motion 1171.
This term was not included in the otherwise more general Eq. (4.28) of reference [lo].
83
For a simple fluid in axial flow between concentric pipes, the difference
in the radial thrusts per unit area on the inner and outer cvlinders is [lo]
(4.15)
((Y)
= by-'
1
- -UY.
(4.17a)
and a is the driving force per unit volume in the axial direction:
(4.18)
a = [n(R,2 - R12)]-1/;
/ is the applied force per unit length in the z direction exerted on the annulus
of fluid.
For a second-order fluid, it follows from (3.Ya), (4.16), and (4.17a) that
(4.17b)
on doing this integration and using (4.18) one finds that
(4.19)
b=
i
4n In (R2m
'
Substitution of (3.9a), (3.9b), and (4.19) into (4.15) enables explicit evaluation of the integral in (4.15), and we find
For a general simple fluid with fading memory, Eq. (4.20) gives an expression
for S,,(R,) - S,,(Rl) which is valid for small f and which involves an error
There are misprints in Ey. (4.14) of reference [lo] and Eq. (3.7) of reference [17].
a4
/? + 2 y = - 8n2q02(R22- K 1*)
u' = 0,
= rw(z),
u' = 0.
I t is assumed that the flow takes place in a right cylindrical region with
axis z, height 1 and radius R . We imagine two rigid discs, perpendicular to
the z axis and rotating about it. The disc a t z = 0 has angular velocity R,
and that at z = 1, R,. The boundary conditions expressing adherence to the
d'iscs are
(4.24)
w ( 0 ) = 4,
w(1) = Q,.
(4.25)
w(z) =
a, - Qo
1
85
+ Q,.
Here we are interested in the normal thrust S,, on the upper disc. This is
determined by the equation [IS, 20, 211
(4.26a)
where
(4.26b)
and
(4.28)
B =~ I I I kr,
y = (kr - kIr)/2.
(4.30)
Steady Cone and Plate Flow. This is a flow whose velocity field has the
physical components
(4.31)
v e = 0,
v1
= r(sin 8 ) o ( 8 ) ,
vr = 0
86
.,(x/2) = no,
For a simple fluid a flow field of the form (4.31) cannot satisfy both the
boundary conditions (4.33) and the equations of motion unless certain
approsimations are made. One such approximation, believed to be satisfactory in many cases, is to neglect the inertia and to assume that I) is small,
i.e. t o put PY[w(e)] = 0 , and also cos 8 = 0 , sin 0 = 1 forn/2 - 4 < 0 < x/X,
in the dynamical equations.
We are here interested in the radial dependence of the normal thrust on
the cone. Under the assumptions made above, it can be shown that [SO, 21 j
(4.33a)
where
(4.33b)
(0,- flo)/I).
1he normal pressure should be a linear function of In Y (at least for r small
enough so that the experimental complications due to uncertain edge effects
are ignorable).
For a general simple fluid with fading memory, one should be able to
obtain (p + y ) from measurements of a.Seeia In Y as a function of n, - Ro
in a cone and plate apparatus of fixed angle $:
(4.35)
4. Appendix
87
V. STEADYEXTENSION
OF
CYLINDER
A steady extension is a flow for which the velocity field has the form
(54
v1 = alx,,
v 2 = a2x2,
213
= a3x3
in some fixed Cartesian coordinate system x,,x2,x3. Here the a' are constants
obeying the relation
a'
(5.2)
+ a2 + a3 = 0,
(5.3)
+ pS+ = [ a i r + (ai)2A)S,,
(no summation)
_-
(5.4)
1-1
1-1
(5.5)
- 1IS',
(no summation)
A ( I ) ,=
~ 2a'&,,
S,
+ pS,,
= S,,[a'27j0
+ ( u ' ) ~ ~+( Pr ) ] ,
(no summation)
and on comparing this with (5.3)we see that for a second-order fluid
and A are given by
(5.8a)
r = 27j0 = const.
(5.8b)
A = 4(p
+ y ) = const.
88
It follows from Eqs. (2.18) and (5.8) that whenever a general simple
fluid with fading memory is approximated by a second-order fluid, the
material constants qo, ,!?,and y must be related as follows to the material
functions of the simple fluid:
(5.9a)
70
(5.9b)
= tT(O,O),
+ y = iA(0,O).
a' = a,
a2 = a3 = - & a ;
d
d
a=-lnL=:-2--1n
dt
dt
(5.11)
L ( t ) are, re-
R.
S,,(O,r)
(5.12b)
SlI(L.,y) = UT
=a T
+ a2A + pa2r2/8
+ a2A + p(aZL2 + a2r2/4)/2.
For a general simple fluid (assuming fading memory and neglecting inertia)
Eq. (5.13) gives an expression for S,, which involves an error of order O(a3).
Hence for such a fluid l;lo and ,!? y may be determined as follows from
measurements of S,, versus a :
89
1 . dS,,
qo = - lim __
2 a-0 da
(5.14a)
(5.14b)
P+y=-lim-.
8
a-+O
dZS,,
da2
+ PGij =
m(s)Gij(s)ds
(6.1)
[a(S~,SZ)Gik(si)Gkj(sZ)b(s,,s,)Gkk(si)Gii(s,)1dsidsz
00
+ o(llGkI(
)It3).
)/I2)
(6.2)
90
= SUP
[G,~(S)C,,(S)]*
m >s>O
is small: in other words, to consider motions for which the strain relative
to some fixed configuration is always small. Such steady flows as those
considered in Sections 111, IV, and 1-do not have this property, but oscillatory motions of small amplitude do. So do histories occurring in stressrelaxation experiments involving small imposed deformations.
Let us consider now the linear theory (6.2). When IIG(t,u( )I1 is made
very small by making e of (6.3) small, (6.2) is approximated by Boltzmanns
theory [24] of infinitesimal linear viscoelasticity ; i.e., for a general simple
fluid with fading memory [ 5 ]
m
(6.4)
Si,
+ pS;i = 2G(O)E;i(t)+ 2
1
~
E,,(t - s)ds
+0 ( c 2 )
In a shearing stress relaxation experiment, i.e. when E,, has the special
time dependence
91
(6.7)
cc
(6.8d)
mm
s,, = - p .
Comparing this result with Eqs. ( 3 4 , which by Eq. (2.18) must hold for
the same fluid also to within an error of order O ( K ~ we
) , see thatt
(6.9a)
qo = - sm(s)ds,
0
(6.9b)
y = - s2m(s)ds,
2
* Cf. H. Leaderman [25]. Note that the symbol c: used in that committee report
represents, in our present notation, 2E,,.
The Eqs. (6.9) can be obtained directly by considering Eqs. (5.14) and (7.4) of
reference [2] and Eqs. (6.1) and (6.4) of reference [5]. without recourse to a special
kinematical situation. Such a derivation was carried out by Coleman and Markovitz [26] :
it shows that our Eqs. (6.9a) and (6.9b)arise from the fact that the terms in (2.18)
which are linear in Rivlin-Ericksen tensors involve, a t bottom, only the first Fr6chet
derivative of 8;j a t the rest history (2.10). The derivation we give here has, however,
the advantage that it does not require familiarity with the theory of functional
derivatives, if one grants Eq. (2.18).
9s
(6.!k)
m2(s)k-Z(s)ds
must be finite. Since we here assume that as s - - bo, h(s) ---c 0 faster than
it follows that m(s) also goes to zero faster than s - ~ ,and hence, by
(6.5) that
s-3,
(6.10)
lim sZG(s)
=:
0.
(6.1la)
(6.11b)
y = - sG(s)ds.
On the basis of a similar argument, Coleman and No11 LJ71 have shown t h a t in
an oscillatory simple shearing motion of small amplitude, to within an error of order
o(e*). the normal stress difference S,, - S,, is determined by G and so is the difference
in radial thrusts on the inner and outer cylinders in a periodic Couette flow. The
material function p of references [ 5 ] and [27] is equal to our m while the function #
of those references equals l j 2 our G. There is a misprint in Eq. ( 8 ) of reference i27];
the lower limit on the integration should be s, not 0.
93
(6.12)
and
< 0.
El&)
(6.13)
Here Ei2 is a real constant. From (6.4) we see that for such a strain, according
to infinitesimal linear viscoelasticity, the shear stress is given by
Through integration by parts and use of (6.10), Eq. (6.14) can be recast
into the form
m
(6.15)
S,,/2E:.L =
0
(6.16)
94
is known as the shear storage modulus [25]; it determines the part of the
stress S,, that is in phase .with the strain E12. The material function G,
defined by the equation [28]
(6.17)
is called the shear loss modulus; it determines the part of the stress S,, that
is in phase with the rate of strain dE,,/dt. The dynamic shear viscosity, ~ ( w ) ,
defined as
(6.18)
y ( ~=
) G(w)/w = G(s) cos wsds
n
lim f ( w )
(6.19)
u+o
G(s)ds
lim $ ( w ) = ?lo;
w+o
i.e. that the dynamic shear viscosity a t zero frequency is equal to the steady
shear viscosity measured at low shearing. It follows from (6.16) that
(6.21a)
lim G(w) = 0
w-+o
(6.21b)
and
(6.21c)
lim
s+o
dG(w)
-0
dw
~
95
It thus becomes possible to calculate, from the results of sinusoidal experiments a t low frequencies, the normal stress difference across the annulus
in Couette shearing at low speeds of rotation, using (4.10a) and (6.22).
I t is frequently assumed that G(t) can be written as the Laplace transform
of some function* N ( s ) , i.e.
G(t)= e-N(s)ds.
(6.23a)
G(t) = e- I7F(z)dt,
(6.23b)
or
G ( t )=
(6.23~)
e-17H(z)d(lnt).
-m
1+ 1
m
(6.24)
G(w) =
s2
w2
ds =
H (t)
w 2t2
F (z)w 2t2
2dt=
l+wt
--a
and
(6.25) m j ( w ) = G(w) =
[----+
N(s)ws
w2
F(t)wt
l f w t
s2
,dt=
-a
(6.26)
s-lN(s)ds = t F ( t ) d t =
i
- x
tH(t)d(lnt)= qo,
96
V I I . NONSTEADY
SHEARISC
FLOWS
The viscometric flows discussed in Sections I11 and I V and the steady
extension treated in Section V are all special cases of a class of motions
called substantially stagnant motions [15, 301 or motions with constant stretch
history [311. All those boundary value problems involving substantially
stagnant motions for which exact solutions are known (under conservative
body forces) for second-order fluids, or even for Newtonian fluids, can also
be solved for simple fluids in general. Furthermore, all of the hydrodynamical
problems for which exact solutions are known for genelal simple fluids yield
subst ant ially stagnant mot ions.
Yet there are some hydrodynamical problems which can be solved
exactly for Newtonian fluids but which, by their very nature, involve motions
which are not substantially stagnant. Although solutions to such problems
for general simple fluids seem beyond our present grasp, it is of interest to
see whether solutions can be found for second-order fluids.
Consider, for example, an isochoric rectilinear flow which in Cartesian
coordinates x1,xz,x3 has a velocity field with components
(7.1)
v l = 0,
212 = v ( x , t ) ,
v3
=:
0,
(XI=
x).
When v(x,t) is constant in t such a flow reduces to the simple shearing flow
discussed in Section 111. Here we assume that v is not constant in t , and we
call a motion obeying (7.1) a nonsteady simple shearing flow. Such a flow is
not substantially stagnant.
In the case of a Kewtonian fluid, substitution of (7.1) into the NavierStokes equations yields the observation [32: that v(x,t) is determined by a
We note that the integral appearing on the right of Eq. (6.11b)has been related
by Fujita (as quoted on page 56 of [29:) to the steady-state shear compliance J ( c )
m
tlOp](c).
97
%,= - p
+ (P+
dy)
(:y
--
(7.db)
(7.84
(7.8d)
s,
= -p.
a2v
81
at1
t l ax
o ~ + y - -w=ap t- -
(7.4b)
aP _
(7.4c)
--
ax
ap
az
at
ap
a~
ax
= 0.
98
1 . Sinusoicial Oscillations
We consider a second-order fluid, free from body forces, undergoing
a nonsteady shearing flow between two parallel plates. We assume that the
fluid adheres to the plates and that a t all times t , - 00 < t < m, one plate
lies in the plane x = 0 and is at rest, while the other, located at x = I ,
undergoes sinusoidal oscillation in the x 2 direction with velocity
cos rut.
We further assume that there is no pressure gradient in the direction of flow.
It follows then that d ( t ) zz 0 in (7.5) and that (7.6) becomes
(7.7)
We now seek a solution of (7.7) with the form
~ ( x , t=
) Kt1 {I*(x)eiylfj
(7.8)
where Re stands for real part of, and C(x) is a complex-valued function
of x obeying the boundary conditions
l.(O) = 0,
(7.!))
l(1) = C,
/I
(7. I la)
(7.1 Ib)
Equation (7.8) is equivalent to the assertion that u ( x , f ) may be written in
the form:
(7.12)
v ( x , t ) = b,(x) cos xt
- F 2 ( x )sin wt.
Here Fl and L, are real valued functions giving, respectively, the real and
imaginary parts of C ( x ) . It follows from (7.10) that
(7.13a)
I, = I,q[sinh
QX
(7.13b)
99
where
(7.13~)
= sinh2 a1 cos2 hl
q-1
+ sin2bl cosh2al.
From (7.2), (7.5) and (7.12), it is seen that p - l and the normal-stress
differences [(S,, - S,,), etc.] oscillate with twice the circular frequency m
of the velocity gradient & / a x . It is also true that for a general simple fluid
in certain periodic simple shearing flows, including sinusoidal oscillations,
the normal-stress differences oscillate with twice the frequency of the velocity
field [ 101. A special characteristic of a second-order fluid is that a sinusoidal
velocity gradient gives rise to sinusoidal normal-stress differences which
are in phase with
If the spacing between the planes is so small that at<< 1, expansion
of the hyperbolic functions in (7.10a) yields the result
(7.14)
V ( x ) = V,x/l
+ O((al)2),
(7.15)
+ O((al)2).
(7.16)
( p + 2y)(1/,/1)2cos2wt+ [ ( t ) .
We may use Eqs. (7.2) to calculate the values of the stress components in
this limiting case. They are also independent of x and are given by:
(7.17a)
(7.17b)
(7.17~)
S,, w - p
(7.17d)
Sa3% - p .
+ tan-l[ym/q,])
(Vo/l)(qo2 y2ze~2)12cos
(wt
+ /?( VO/J!)~
C O S ~~t
It will be noticed that the shearing stress S,, = S,, oscillates with the
same circular frequency w as the velocity but is out of phase with the velocity
by an angle equal to tan- (yw/qo). This result for S,, is the same as that
which would be obtained by solving the present boundary-value problem
using the constitutive Eq. (6.4) of infinitesimal linear viscoelasticity,
taking the limits al-. 0 and zl-. 0, and then using Eqs. (6.20) and (6.22).
However, on working within the framework of infinitesimal linear viscoelasticity, one does not obtain information about normal-stress differences,
100
while our present calculation yields the interesting result that, for a secondorder fluid and small separation I, the normal-stress differences oscillate
with a circular frequency twice w and are in phase with the square of the
velocity of the moving plate. We also note that the particular normal-stress
difference
(7.18)
References
1. SOLL,
W.,
101
14. MARKOVITZ,
H., ELYASH,L. J., PADDEN,
F. J., J R . , DEWITT,T. W.. .4 cone-andplate viscometer, J . Colloid Sci. 10, 165-173 (1955).
16. COLEMAN,B. D., Kinematical concepts with applications in the mechanics and
thermodynamics of incompressible viscoelastic fluids, Arch. Rational Mech. Anal.
9, 273-300 (1962).
16. RIVLIN,R . S., Solution of some problems in the exact theory of visco-elasticity,
J . Rational Mech. Anal. 6 , 179-188 (1956).
17. COLEMAN,
B. D., and NOLL, W.. Helical flow of general fluids, ,I. AppZ. Phys.
30, 1508-1512 (1959).
18. PAI.Shih-I, Viscous Flow Theory. I. Laminar Flow. Van Nostrand, Princeton,
1956, p. 54.
19. PADDEN,F. J., and DeWITT, T. W., Some rheological properties of concentrated
polyisobutylene solutions, J . APPI. Phys. 26, 1086-1091 (1954).
30. NOLL, W., and TRUESDELL,
C., The non-linear field theories of mechanics i n Handbuch der Physik (S. Fliigge, ed.) v. VIII/2. Springer, Berlin. To be published.
21. MARKOVITZ,
H., Normal stress effect in polyisobutylene solutions. 11. Classification and application of rheological theories, Trans. SOC.Rheol. 1, 37-52 (1957).
H., and BROWN,D. R., Normal stress measurements on a polyiso22. MARKOVITZ.
butylene-cetane solution in parallel plate and cone-plate instruments, Proc.
f n t l . Symposium on Second-Order Effects i n Elasticity, PEasticity and Fluid
Dynamics, Haifa, 1962 (in press).
23. COLEMAN,
B. D., and NOLL, W., Steady extension of incompressible simple fluids.
Phys. Fluids 6 , 840-843 (1962).
24. BOLTZMANN,
L., Zur Theorie der elastischen Nachairkung, Sztaber. Kaiserlich.
Akad. Wiss. ( W i e n ) , Math.-Naturwiss. KI. 70, Sect. 11, 275-306 (1874).
65. LEADERMAN,
H., Proposed nomenclature for linear viscoelastic behavior, Trans.
SOL. Rheol. 1, 213-222 (1957).
26. COLEMAN,
B. D., and MARKOVITZ,H.. Normal stress effects in second-order fluids,
J . A p p l . Phys. 86, 1-9 (1964).
27. COLEMAN,
B. D., and NOLL, W., Normal stresses in second-order viscoelasticitv.
Trans. SOC. Rheol. 6 , 41-46 (1961).
28. GROSS,B., Mathematical Structure of the Theories of Viscoelasticity, Hermann
and Cie, Paris, 1953.
29. FERRY,
J . D., Viscoelastic Properties of Polymers. Wiley. New York, 1961.
B. D., Substantially stagnant motions, Trans. SOL. Rheol. 6, 293-300
30. COLEMAN,
(1962).
31. XOLL, W . , Motions with constant stretch history, Arch. Rational Mech. .4nal.
11, 97-105 (1962).
37. STOKES,G. G., Mathematical and Physical Papers. Cambridge University Press,
Cambridge, 1901, vol. 111, p. 19, Eq. (8).
33. LAMB,H., Hydrodynamics, Cambridge University Press, Cambridge, 1895.
$ 5 298-300.
1. Introduction . . . . . . .
TI . Resume of Major Features
Iage
105
i06
. . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . .
A . S I M W F I E I I PHYSICAL xcouxr
111. Physics of Sound Generation . . . . . . . . . . . . .
1 . Simple Source or Monopole . . . . . . . . . . . . .
2 . Force Source or Dipole . . . . . . . . . . . . . .
3 . Stress Source or Quadrupole . . . . . . . . . . .
1. Equivalent Patterns of Simple Sources, Dipoles and
I c. Equivalent Aerodynamic Generators of Sound . . . .
I . Reynolds Stresses (Quadrupole Sources) . . . . . .
2 . Vortex Stretching (Dipole Sources) . . . . . . . .
3 . Fluid Dilatations (Simple Sources) . . . . . . . . .
1. Sound Radiated from a Jet . . . . . . . . . . . . . .
I . Structure of a Round Turbulent Jet . . . . . . .
2 . Sound Radiated from a Volume Element . . . . .
3 . The Ujs, yo. y-. f z and f - 2 Laws . . . . . . . . .
4 . Self-noise and Shear-noise . . . . . . . . . . . . .
.
.iEffects of Convection and Refraction . . . . . . .
6 . Shifts of the Spectrum Peak . . . . . . . . . . .
7. Power and Efficiency vs . Mach Number . . . . . .
X . Effects of Density and Temperature . . . . . . . .
!I . Reduction of Turbojet Noise . . . . . . . . . . .
.
.
.
.
. . . . . .
109
. . . . . .
109
. . . . . .
111
. . . . . . . 112
Quadriipoles
. .
113
. . . . . . . . 115
. . . . . . . . 115
. . . . . . . . 116
. . . . . . . . 117
. . . . . . .
II!)
. . . . . . . . II!)
. . . . . . . . 122
. . . . . . . . 123
. . . . . . .
125
. . . . . . . . 126
. . . . . . . . 131
. . . . . . . . 133
. . . . . . . . 136
. . . . . . . .
138
R . MATHEMATICAL DEVELOPMEN1
\I . Governing Equations . . . . . . . . . . . . . . . . . . . . . . . .
142
I . Wave Equations for a n Inviscid Fluid . . . . . . . . . . . . . . . 142
2 . Solutions for Radiated Sound . . . . . . . . . . . . . . . . . . . 143
3 . Spectrum of Radiated Sound . . . . . . . . . . . . . . . . . . . 146
4 . Developments of Lighthill and Proudman . . . . . . . . . . . . . 147
\.I1 . Convection Effects for a Simplified Model of Turbulence . . . . . . . 151
I . Moving vs . Stationary Axes . . . . . . . . . . . . . . . . . . . 151
2 . Directivity of Sound from Unit Volume . . . . . . . . . . . . . . 153
3 . Spectrum of Sound from Unit Volume . . . . . . . . . . . . . . 155
1. Moving Jets . . . . . . . . . . . . . . . . . . . . . . . . . . .
157
V11I . Refraction Effects Due to the Mean Flow . . . . . . . . . . . . . . 159
I . Convected Wave E(Iuation . . . . . . . . . . . . . . . . . . . .
159
2 . Form of Solution . . . . . . . . . . . . . . . . . . . . . . . .
161
3 . Qualitative Effects on Directivity . . . . . . . . . . . . . . . . . 162
IS. Improved Model: Isotropic Turbulence Superposed on Mean Flow . . . . 163
1. Remarks on Lilleys Approach
. . . . . . . . . . . . . . . . . . 163
n. s.
104
RIBNEH
SOTAT 10s
The more important symbols are collected here for convenience. Other symbols of
limited use are explained in the text.
Sozzle cross-section area
Convection factor (C = [ ( I - Mc cos 818 + a*kfCaj14
Speed of sound
Nozzle diameter (D = 2 H )
Frequency
.4 scale of turbulence
u/c
Eddy convection speed/co (often taken as U,/Zc,)
Flight speed/c,
m
z.
P(y)dsy =
P(yl)dy, =
z-ye)zx sin W O ,
P(w)dw =
0
etc.)
105
p(')
5;
t
t'
w
wt
WP
Ouevbars, eic.
In Part A, effective time-space average over jet: in Part B, local time average
Double partial time derivative
Associated with single-frequency source pattern
Used with different meanings: defined where used
~
..
SubscriPts
0
i
i
Other
[ I
(
)AV
I. INTRODUCTION
The purpose of this article is the explanation of major features of jet
noise. The framework has evolved from the fundamental work of Lighthill
[I, 21 and has been contributed to by a substantial number of workers
(see list of References). A further extension has been made in the course
of preparing the present survey; it takes the form of an advance in the
handling of spectra and directivity based on a new treatment of 'self-noise'
and 'shear-noise'. After some preliminaries, a single coherent analysis is
developed in detail, drawing in the main from [l-131. Alternative approaches
are touched on but not treated exhaustively.
A number of excellent review articles and a bibliography have been
written on this and the more inclusive subject of aerodynamic sound,
[3, 14-24, 901. Of these [lF] and [18] are comprehensive surveys of experimental results. The present article should be regarded as complementary
in nature. I t provides a state-of-the-art account of the theory as it has now
been extended.
Ideas from these earlier reviews have been drawn on where appropriate,
and the writer would like to record here his debt to the authors concerned.
Of particular help have been the notions of Powell [2l] and \t-estervelt [25]
for Section 111 and of Lighthill [3] for Section V.9.
As in Lilley's work [4] no mention is made of the noise emission from
choked jets or supersonic jets that are under- or overexpanded. This noise
is dominated by certain resonance phenomena [26, 271 excited by the
interaction of turbulence with shock waves [28-30]; the generation process
106
H.
S. RIBNER
is quite different from that for the normal shock-free jet. A simple brief
account is given, e.g., in [8!.
The following section gives a short summary of the major features of
jet noise with a qualitative explanation. The casual reader may stop here.
The remainder is divided into two parts. Part A is a greatly amplified
account of the material given in the earlier resume. The emphasis is on
the physical explanation and the mathematics is kept to a minimum.
Moreover, analogies are used from time to time in place of genuine derivations. Part 3 is the serious mathematical development documenting the
conclusions of Part A, with extensions. There is some overlapping in the
expository material of the two parts, but an effort has been made to keep
this within bounds.
Thus the material is presented on three successive levels of elaboration.
I t is hoped in this way to serve the interests of a wider range of readers.
Perhaps, also, the mathematical development will be clearer if there is a
qualitative account to fall back on for perspective.
11. RESUMEOF MAJOR FEATIJHES
A turbulent jet flow involves inertial forces associated with the fluctuations of momentum flux. Equivalently, the forces arise from the movement
of vorticity. These forces are effective in generating sound (Figs. L(b) and 2).
Since the resultant force field must be zero, the fluctuating forces may be
thought of as occurring as opposed pairs as in stresses. A fluid element subject
to such a force pair suffers a fluctuating quadrupole deformation: this is
a second mechanism of sound generation (Pigs. l(c) and 3). Associated with
the inertial forces will be pressure gradients in the flow. A region of high
pressure will be slightly compressed and conversel).. The transient compressions and expansions (dilatations) constitute a third mechanism of
sound generation (Figs. l(a) and 4). Mathematically, all three are equivalent
in the sound they produce when summed over the region disturbed by the
flow.
The sound power generated from a volume element d l of turbulence
of effective scale L , mean square velocity
and dominant frequency f
in low speed flow comes out to be
(2.1)
where po is the ambient density and co is the ambient speed of sound. In an
idealized model of jet flow we assume 3 4 U 2 , the square of the local
U / L with the result
mean velocity, and f
(2.2 )
dP
dVpoU8/Lcos.
107
(2.3)
U 3 ,L
r/.
D,U
Ui to obtain the
poUj8D2/~,5
dp
dy
Iyo
)yP7
in mixing region
in developed jet
for the distribution of sound power emission with distance y measured along
the jet axis (Fig. 10). Furthermore, if the 'slice' of jet dy a t y emits as a simplifying assumption -. only a single frequency f(y) it follows under
the assumed similarity that
(2.5)
-4
dp
f2
df
/-2
gives the slope of the frequency spectrum on either side of the peak (Fig. 10).
A more detailed analysis indicates that the sell-izoise due to the
turbulence is accompanied by shear-noise due to cross-coupling with the
mean flow. The shear-noise spectrum is peaked an octave below the selfnoise spectrum and has a relative factor
c0s4 0, where 0 is the angle of
emission relative to the flow direction (y-axis). The low frequency shearnoise spectrum dominates at small values of 8 (e.g. 30") and the high frequency self-noise spectrum dominates near 90' where C O S ~8
0 (Fig. 17).
Convection of the eddies by the mean flow (when subsonic) crowds the
sound waves in the downstream direction (Fig. 11). This causes an effective
Doppler shift of frequency in the ratio C-l and an associated amplification
C'-* (Figs. 16 and 17) where
--+
(2.6)
108
H . S. H I R N E K
109
speeds (model jets, turbojets) very accurately follow a U; law, and there
is a transition at supersonic speeds (afterburning jets, rockets) to something
approximating a U j 3 law for the limited region of the data (Fig. 19).
Division of the two-slope sound power law U? - U j 3 by the kinetic
power of the jet
Ui3 gives a two-slope efficiency
U: and constant.
Thus the steep rise for subsonic jets levels off to a constant limiting efficiency
(of 0.3 to 0.80;,) for rockets.
The experimental rocket data do not extend beyond UiM Sc, (the
effective eddy speeds are half this) and the use of the convected quadrupole
(or dilatation) deductions for higher speeds is purely speculative, as has
been indicated. The approach of 0. M. Phillips - an asymptotic theory for
high values of Ui/co- predicts that the efficiency must widely used diminish
like U , - 3 / 2 .
Substantial reduction in jet noise can be accomplished, for a given
thrust, by the reduction in velocity associated with a larger jet diameter
according to the U; law. This is exploited in the increasingly popular
bypass or turbofan engines. In the pioneer Rolls Royce Conway the noise
reduction is 7 dh in one case cited.
Corrugated and multi-tube nozzles are the most widely used means for
quieting existing turbojets. Their development has been motivated by
conflicting interpretations of the theory, and the explanation of their
behavior remains a matter of controversy and speculation. It is generally
agreed that reduction in shear plays an important role, presumably reducing
overall turbulence levels : these jets entrain external air into a restricted
region, giving it some forward velocity. Another notion is that the sound
from one small jet (or corrugation) is reflected and refracted (shielded)
by the temperature and velocity field of nearby jets so that the aggregate
sound has a less peaky directional distribution. A group headed by R. Lee
of General Electric has had some success with a semi-empirical computerized
method for predicting muffler behavior (and flow development of interfering
jets) in some detail. Their results suggest that the shielding principle cited
above may play only a minor role.
4
110
H. S. RIRNEH
with the balloon rempved will have the same effect. Thus the
signs have
a dual interpretation as radial acceleration (of the balloon) and as a time
rate of injection of sources of matter.
The sound pressure radiated from the balloon is proportional to the product
of surface area times radial acceleration and density; this is a measure
of the total strength of +Is. The value at a given distance is proportional
to p u s / , where p is the fluid density, u the rms velocity, S the surface area
and f the frequency. The corresponding radiated sound power is
P,
(3.1)
(puS)2/2/pr
Oscillating
rigid sphere
Oscillating
force
Dipole
Dipole pair
Deforming sphere
where c is the speed of sound. Since p u s is also the rate Q a t which mass
is supplied, an alternative form is
(3.2)
P,
Q2pyPc
(aQ/at)2/pc.
(3.3)
P,
rn21d6/2/pc
(am/a1)21a6/pc
111
- -
(3.4)
The model for a dipole source of sound is a rigid sphere that oscillates
back and forth along a line ( e g , a pendulum bob). The outward or normal
acceleration of the forward and rearward hemispheres is now opposite.
A t a given instant we may put +Is on the face with positive normal acceleration and -'s on the face with negative normal acceleration. The spherical
sound waves that are produced are likewise 180" out of phase in front and
rear. The intensity pattern is a figure of eight.
and
are equal. This would seem to imply a
The sum total of
mutual cancellation : the outflow from the sources tends to circulate around
the sphere to be ingested by the sinks. However, occurrences at one face
will be transmitted at the speed of sound to the other, i.e., at a time
L/c
later, where L is the diameter. The transmitted pressure will fall short of
cancelling the local pressure disturbance by a fraction given by the phase
shift
f L / c ; the transmitted velocity will behave similarly. Hence the
acoustic power acquires a factor
(fL/c)* compared with that for the
sources or sinks separately,
--Is
+Is
P,
(3.5)
(puS)2L2f4/pc-
= 1.5p
(volume) (auli3t)
pSLfu.
112
H. S. RIBNER
P2
(34
pUSD2/c34 pU3D2M3.
This result may be applied to Aeolian tones: these are the sounds generated
by a cylinder in an airstream when there is an oscillating lift associated
with an eddying wake. The cylinder (unlike our assumed sphere) may be
perfectly stationary, the force being a reaction. Curle's analysis [32] shows
that (3.7) and (3.8) are still applicable to the radiated sound when supplemented by a term corresponding to quadrupole generators. The
quadrupole power has, however, a relative strength's Mach number
squared; it tends to be much smaller a t the subsonic speeds of the reported
measurements. These measurements lie very close to the U S law of (3.8)
(e.g., [96-991).
3. Stress Source or Quadrupole
P,
F L ~ / L2(a2F/at2)2/pCti.
~ C ~
I13
(3.10)
~1
P3
pUSD2/c5 pU9D2M5.
This resuit contains the famous U8D2 law that has been widely confirmed
by experiment for the noise of subsonic jets.
The efficiency of sound generation may be defined as the ratio of the
acoustic power to the mechanical power or flux of kinetic energy
pU3D2
in the flow. These conversion efficiencies for systems of sources, dipoles,
or quadrupoles are proportional respectively to M ,M3, and M 5 (see (3.4),
(3.8), (3.10))for systems with the assumed similarity. It is evident that for
M < 1 the sequence of decreasing efficiency is source : dipole : quadrupole
for equal efficiencies at M = 1.
The decreasing efficiency results from the Stokes effect : the partial
cancellation of the output of the two sources of the dipole, and, carrying
the process a step further, of the two dipoles of the quadrupole [ 1, 31. In
more physical terms, the fluid pushed out in front of the oscillating sphere
dipole is sucked in again at the rear, thus greatlv reducing the tendency
to create compression waves or sound. The effect is further compounded
in the deforming sphere quadrupole, with a further reduction in efficiency.
114
H. S. RIBNEH
earlier for the individual sources and sinks making up a dipole or quadrupole
will likewise apply to some degree in a spatial source distribution whose
total strength is zero. Lighthill has therefore made the point [l-31 that
the emission from such a source distribution has the character of quadrupole radiation. However, this characterization is oversimplified and some
further discussion is in order.
We can imagine a continuous random distribution of sources and sinks
in two dimensions (for convenience) as a random system of hills and hollows:
the hills represent sources and the hollows, sinks, and the height or depth
the local strength. As the pattern fluctuates in time, the hills and hollows
change in an erratic way; however, they have an average size, or radius.
,.
1his source radius is a measuret of the average effective size of coherently
radiating source or sink patches.
Only regions of sources whose extent is small compared with a typical
wave length of sound can be regarded as emitting simultaneously: for larger
regions sources more distant from an observer must be counted earlier
because of the travel time. If the source radius is very small compared with
a sound wavelengthlh, then such regions each contain a number of sourcesink patches that may be treated as emitting simultaneously. If the
and --Is essentially cancel in these regions, the emission figured in this way
will be essentially zero. Then if one wants to retain the simultaneous emission
approximation, a better approach is to replace the source-sink pattern by
the equivalent dipole pattern. If this, too, virtually vanishes the replacement is by the equivalent quadrupole pattern.
+ I s
115
IV. EQUIVALENT
AERODYNAMIC
GENERATORS
OF SOUND
1. Reynolds Stmsseg ((juadrupole Sources)
116
H . S. RIBNER
seems to ensure the vanishing of the resultant force. The magnitude of the local stress is the rate of momentum flow pnu from which the inertial forces are
derived. This is closely related to the Reynolds stress in the theory of the
turbulent boundary layer. K e shall retain the name 'Reynolds stress' for
puu but observe that here it is an instantaneous stress and moreover it now
includes the contribution of the mean flow.
The Reynolds stress may also be written pupj (i = 1, 2, or 3: i = 1, 2,
or 3) to show that it has nine components: e.g., the component pula2
represents the force in the u1 direction acting on unit area normal to the ug
direction. This is a shear stress and corresponds to a lateral quadruple.
A term like puI2 is a longitudinal stress and corresponds to a longtudinal
quadrupole. These are illustrated in Fig. 3 [l-31.
is epuivolent to
-+
FIG.
+-
is equivalent to
+-
3. Longitudinal and lateral quadrupole sound sources associated with the put'
and pu,ul Reynolds stresses, respectively (after Lighthi11 r3:).
117
118
H. S. RIBNEK
5, dilatation rate =
ap/at
effective
source strength,
am
at
azp(o)
at2
I azp(o)
c2
at2
p(0)
p(I)
flow.
field dominates within and near the turbulence a t subsonic speeds (Fig. 5 ) ,
constituting what is known as the acoustic near field. Further out it is
overriden by the acoustic radiation field $ ( I ) , which decays more slowly
with distance. The p(") field has virtually the characteristics of the pressure
field in an incompressible flow, being dominated by inertial rather than
compressional effects; in particular, it exhibits no wave propagation : hence
the name 'pseudosound'.
Relatively small at subsonic flow speeds
119
-.
V. SOUNDRADIATEDFROM
A JET
Lilley [4] has organized the available experimental data on the round
turbulent jet (especially that of Laurence [39]) into a form suited to the
estimation of jet noise. More recently the data have been extended by
Davies, Rarratt and Fisher [40]. \Ve shall summarize briefly the relevant
material, following Lilleys plan in the main.
Self -similar
profiles
I
I
I
I
I
I Mixing
I
I
I
I
1
Nozzle
Y
region
Transition
region
Mean
FIG. 6. Idealized model of a round turbulent jet. In a real jet similarity of turbulence
profiles is not attained in H until a t least 50 diameters from nozzle.
120
H. S . RIBNER
The radial profiles of some major mean and turbulent quantities are
approximately self-similar in the mixing region (A), changing only in scale.
For example (Fig. 7(a)), the profile of turbulent intensity, normalized to
unity at the peak, fits a universal function of = ( I - R ) / y a t all axial
stations ( R is the diameter).* The profile of the mean velocity (7 (not shown)
is likewise a single function of 5 [43],
(b)
(0)
-!
L,
(5.2)
(5.3) WID
-= 0.71,0.46,0.37,0.33
ui
0<y
M 0.13 .y
for
< 611
121
[=r/y
characterizing the fully developed jet. The transition region extends from
about 4.6 to 8 diameters. Here 5 has been redefined as r l y , r is the local
!:;$&Dovies
0
et all
M=0.3 (Laurence)
0.05 I
YdD
radius and the origin of y is taken 1.0 diameter upstream of the jet exit.
Figure 7(b) shows that the turbulent intensity profiles are not fully selfsimilar in terms of r/y as early as 8 diameters; in fact, Townsend [43]
-0.1
0
Radial distance
0.I
0.2
Y2-R
Yl
FIG.9. Profiles of mean-flow speed (dotted) and convection speed (solid) across
mixing region (reproduced from Davies, Barratt, and Fisher [40]).
122
H. S. RIBNER
cl
Volume Eiemeiit
dP
dV(am/at)PL3/poco
where N equals the kolume ratio dV/L3 and we are being a little more precise
now, designating p and c as the ambient values po and c,,. In this equation
L3 may be approsimated as (average
the effective volume of the source
scale)3 of the turbulence - in crude terms the volume of an eddy. A more
precise measure is the correlation volume defined in Part B.
In the dilatation theory the effective source strength (in terms of a
fictitious rate of mass injection m ) is given by
4
(4.2)
amiat =
co-Wp(o)iat2
dP
CT
d V po(U2)zf4L3/~06.
123
(When the jet is hot or consists of a foreign gas, po in p(O)+ poGmust be replaced
by the local (time average) value p ; this leads to p2/po in place of po in (5.6).
The speed of sound, here c,, must also be reexamined. Cf. Sec. V.8.)
This same equation is readily derived from (3.9) based on the Reynolds
stress or quadrupole model of flow noise. It can also be obtained from the
vortex model, a qualitative argument being that leading from (3.7) to (3.9).
3. The U;, yo, y-, fa and
Laws
In the mixing region (A) of a jet (cf. Sec. V . l ) the profiles of mean and
turbulent quantities are self-similar or invariant with axial distance y when
expressed nondimensionally : along certain rays the turbulent velocity
and scales and the mean velocity maintain a fixed proportionality. A second
region (B) of self-similar mean velocity profiles starts about 8 diameters
downstream of the nozzIe defining the fully developed jet (Fig. 6). Full
self-similarity of the profiles of turbulent velocity and scale is not attained
until some 50 or more diameters downstream [43](Fig. 7). In our analysis
we shall deal instead with an idealized model of jet flow in which full selfsimilarity of the mean and turbulence profiles throughout region B is
postulated, starting at 8 diameters. The differences are to be borne in mind
in applying the results to a real jet.
According to the assumed behavior the ratio G/U2is the same for similar
elements d V . We assume a further similarity in the scaling of frequency,*
f cc U I L ; this is roughly approximated by the limited experimental data
covering region (A) (Sec. V . l ) . With these relations (5.6) becomes
dP
(5.71
dVpOU8/Lco5
yo Law:
dP
dY
poUjaD - constant,
CC--
co5
mixing
region (A)
~1
y2dy (disc),
124
(5.9)
H. S. RIBNER
y-Law:
(;)-
dp
po;;?
- - dY
developed
jet (B)
5
Y
The area under the curve is the total sound power emitted by the jet.
By (5.8) and (5.9) this must be of the form
(5.10)
Ui8Law:
P -poU?Da/cos
( U , = nozzle velocity)
in agreement with (3.10) obtained by less detailed considerations. The righthand side of (5.10) with De replaced by the nozzle cross-section area -4 (to
allow for noncircular nozzles) has been termed the Lighthill parameter by
a number of writers.
For simplicity we shall imagine that a given slice of jet emits just a
single characteristic frequency : the actual rather peaked spectrum of the
slice (cf. Part B) is considered squeezed into a single line. Then the spectrum
emitted by the jet as a whole can be approximated as a continuous array
of these lines side by side by writing [6, 211
(5.11)
This is not a curve of quadrupole strength distribution along the jet as might
be inferred from remarks in [3]. The (quadrupole strength) differs by a factor c (frequency)- which varies strongly with y.
125
f-2
(5.12)
f2Law:
(5.13)
d P *PoUi0D f-2,
Law:
dP
df
Cob
--
poUjbD6
Cob
f2*
mixing
region (A);
developed
jet (R)
after some simplification [el. These results and their derivation are illustrated
in Fig. 10.
Equations (5.12) and (5.13) may be combined into
(5.14)
126
H. S. RIBNER
(5.16)
+ COS~O
where the factor of 2 equalizes the integral over frequency of the two
H-functions.
This result does not include the very important effects of convection
and refraction which are introduced in the next Section.
5. Effects o/ Convection and Refraction
We shall now consider how the foregoing results are modified, first by
convection of the eddies, and next by refraction due to the wind - the
mean jet flow.
Hotwire studies employing correlation techniques [40] show that, roughly
speaking, an eddy in a jet is convected about three times its length before
it has decayed, i.e., before it has lost its identity due to fluctuation. Thus
we have the picture of the moving eddy emitting sound pulses during a
short lifetime. This is illustrated qualitatively in Fig. 11. For subsonic
convection speeds the sound waves are crowded in the downstream direction
and more widely spaced in the upstream direction. This powerfully enhances
the intensity in directions making an acute angle with the flow. The polar
plot above the sketch shows the variation in intensity (at a fixed radial
distance) with emission angle 0 , for a continuous succession of such wave
trains.
For supersonic convection speeds (Mc> 1) the effects are more extreme
in that the sound waves can coalesce t o form a concentrated envelope or
annular segment of a Mach cone. The sound intensity shows a strong peak
normal to this cone; the peak direction is defined by M , cos 8 = 1 from
aerodynamic considerations.
The wave pattern grows even after the eddy or sound source has died,
enveloping both upstream and downstream points. Thus there is no zone
of silence as in steady supersonic flow. (The difference lies in our use of a
127
stationary frame of reference that does not follow the moving source.)
Further, the finite lifetime of the source accounts for the truncation of the
Mach cone
In a jet the successive passage of eddies will provide a succession of
wave trains leading to a distribution of such truncated Mach cones, one
within the other, expanding outward with time. A given point will receive
a continuous fluctuating sound pressure instead of a short burst of pulses.
Further, the randomness in time - but not the randomness in space - will
tend to smear out or otherwise impair the sharpness of the envelope at
M , cos 6 = 1. This will moderate an otherwise infinite intensity peak which
would occur for moving eddies emitting a single frequency (or even zero
frequency (d.c.)) [ll].
--
Subsonic
Peak sound
pressure
Supersonic
FIG. 1 1 . An eddy moving from left to right emits a series of sound pulses during its
short lifetime. Upper curves show rms pressure variation provided by crowding of
sound waves, given by factor C 5 I 2 = [ ( l - M c cos O)*
aaM,a]-5/4 where speed is
Mcco and eddy length/decay distance is a.
(5.17)
(1 - M , cos 6 )-5
128
H . S. RIBNEH
analysis in effect postulated very large decay times for the eddies (or frequencies approaching zero). An allowance for finite decay time via a characteristic radian frequency w1 and scale L was made in the work of Ribner
[9-111 and Williams [13], giving finally*
c-6 = [ (1 - M , cos e ) 2 + o
~ ~ L ~ / x=
c ~[ ( I~ ] -M,
~ cos
/ ~q2 O L ~ M , ~ ] - ~ / ~
(5.18)
in place of (5.17) (equating the two forms serves to define a, with M , = Uc/co).
The derivation according to the method of [ l l ] is given in Part B herein.
This is supplemented by a more detailed physical argument than that given
for Fig. 11.
The early work of [9] and [lo] dealt with simple sources and yielded
C-1 in effect as the convection factor for intensity (cf. [ l l ] ) . The phase-shift
FIG. IS. Refraction and reflection of sound rays by a semi-infinite plane flow. Plane
waves are assumed, the rays being normal to the wave planes.
cancellation effects discussed in Sec. IV.3 for the constituent sources and
sinks of stationary quadruples show up again in the convection process;
these give an additional factor C-2 in sound pressure or C-* in (pressure)2*intensity, so that the overall factor is C-s. The double time
derivative in the equivalent dilatation source behaves similarly, likewise
providing the additional C-4 factor.
It is well known [44]from theory and experiment that sound rays are
refracted by velocity gradients - that is, by wind. For the case of plane
waves in a semi-infinite plane flow the angle of refraction for sound leaving
the jet (Fig. 12) and the refracted intensity can be accurately predicted
[45, 461. There exists a minimum refraction angle given by
(5.19)
-.
M, cos 0
129
outward rotation of the peak sound intensity (peak of the polar of Fig. 13)
with increasing speed of sound: the sequence was 15", 30", 42" for air
(ci = 1030 fps), turbojet (cj M 1800 fps), helium (ci = 2910 fps).
We can now summarize the overall directivity of the jet noise approximately as follows. The directivity is governed by the combined factor
(5.15) and (5.18), namely,
(5.20)
(1
from near 180" to near 40". Below roughly 40" (turbojets), 20' (model jets)
refraction provides a deep dimple (reduction) compared with this factor.
In the above the two terms of the factor (1 cos40) respectively describe
the basic directivity of the self-noise and shear-noise, and the factor C-5
describes the directive amplification provided by convection of the eddies.
In Fig. 14 the directivity factor (5.20), with GC chosen as 0.55 for
best fit, is compared with measurements for turbojets and with the factor
(1 - M , cos 13-5. (The experimental a from space-time correlations of
turbulence in the mixing region 1401 is nearer 0.33 for unheated air jets;
there are no measurements on turbojets known to the writer.) It is seen
that (5.20) with the chosen a is quite good above the refraction cutoff. The
130
H . S. RIBNER
\.
- .\.
---- h4pO.W(JSe-P-Sengine)
30-
4.0.82 ( I - M,
,,.'.
COS
e)*
t
~
L"
FIG. 14. Relative intensity of turbojet noise versus angle 8 from flow direction (referred
to zero db at 90'). Experimental data from Pietrasanta [91j is compared with the
theoretical factors. (Allowance is not made for the cutoff below 40a, presumably due
to refraction.)
8
0.I
0.3
0.5
07
09
10
Mc
131
omax =
17 loglo [(fo/uj)(rj/TJ2.11
where T,/Tois the ratio of jet temperature (at nozzle) to ambient in "R, and
emaxis the angle of the peak of the noise polar in degrees. The points show
considerable scatter from the line (5.21), but it does exhibit the general
influence of the principal variables. The equation applies to both round
and corrugated-nozzle jets, D being the diameter of a round nozzle of
equivalent cross-sectional area.
Jet velocity makes itself felt in two competing ways: via the convective
factor C-s and via refraction. The former, which involves velocity to the
fifth power, is presumably more sensitive to speed. In the writer's view,
with increasing jet velocity the convection strongly enhances the downstream amplification to encroach even more on the refractive valley of Fig. 13.
This rotates the peak intensity inward (smaller emax)in qualitative conformity
with the equation.
6 . Shifts of the Spectrum Peak
132
H. S. RIBNER
frequency pushes the peak slightly to the left of the true Doppler shift
(1 - M , cos O ) - l giving the rather smaller effective shift
C-1= [ (1 - M , cos 0)2
+ a2Mcz]-1/2
Radion frequency
(cf. [ill)).
FIG. IS. Sound spectrum radiated by unit volume of a simplified model of turbulence.
Convective peak amplification and 'Doppler' shift are given by factors C-4 and C-l,
respectively, to give power amplification C6.
02
0.5
I
2
5
Relative frequency/flow speed
10
FIG. 17. Variation of theoretical idealized jet noise spectrum with emission angle 0.
Relative heights of "bass" and "treble" without allowance for convection are in approximate ratio 2 c0s4 0 t o 1. Convective amplification coupled with Doppler shift yields
final (upper) 30" curve. (This figure is oversimplified compared with (5.22) or (10.22).
The separate convection factors (C-* in the equations) for the bass and treble spectra
are lumped together.)
133
Number
The low speed analysis (Sec. V.3) predicted a variation of sound power
with jet nozzle velocity Ui proportional to U T . We shall now reexamine
this, taking account of the modifications due to convection and other effects.
* This effect alone would yield U,'.
11s.Mach
134
H. S. RIBNER
(1
+ Mc2)/(1- Mc2)4
for
M,
< 1,
00
for M,> 1,
which results from integration over a unit sphere. With the effective convection Mach number M , taken as Uj/2c,, (mixing region of the jet) (5.23)
predicts infinite augmentation at jet speeds Uj of twice the speed of sound
-10
Convection speed
Speed of sound outside jet
FIG. 18. Directional average of convection factor, ( C d S ) ~ v= (1/2) sC-S sin Ode,
0
in db relative to value at M c = 1. This factor multiplies low-speed power law (e.g. Uj*).
(Weighting factor ( 1
~ 0 ~ 4 has
8 ) not been included in average.)
135
T H E G E N E R A T I O N OF S O U N D BY T U R B U L E N T J E T S
Rockets
t engines
terburners)
500
loo0
2000
5000
uj (ft/sec)
I(
100
- -
FIG. 19. Noise power vs. jet speed, showing transition from
U,a to
U , S , approximately, when convection speed Lj/Z exceeds the speed of sound (Reproduced from
Powell [31]).
X/D
Lossiter-Hubbord 3
0.02 0.04
0.1
0.2 0.4
r/D
0.5
1.0~10~
Reynolds NO.(pUjD//d
FIG. 20. Dependence of turbulence intensity (in yo of U,) on jet nozzle velocity Uj
(after Lassiter and Hubbard [87], with data from Laurence [39] replotted on a Reynolds
number basis).
The deduction of
for subsonic nozzle speeds (before multiplying
by the convection factor) is actually faulty: a reconsideration of the turbulence data indicates instead more nearly Ui7. (This tends to cancel out
the subsonic rise in the convection factor, Fig. 18.) The Ui7 results because
the rms turbulent velocity increases about like U?I4 rather than linearly
136
H. S. RIBNER
with U , as first assumed (Fig. 20). I t is not clear whether the relative intensity
levels off again at higher speeds to restore a basic U? law as seems to be
implied by the noise data.
A marked deviation from an eighth-power speed dependence is to be
expected eventually from considerations of efficiency. Division of noise
power (- Ui8) by mechanical power (- U,) of the jet gives an efficiency
Ui5 for subsonic values of Ui/2. The corresponding division of noise
power (- via)by jet power (m U,) for the supersonic range of Uj/2 gives
a constant efficiency. More precisely (according to experiment for air jets,
turbojets, and rockets)
?I = K1Mj5
(5.25)
= K2
where K , M x~
po/pj [5l], K , = 0.3 to 0.8yu [52]. The point is that
the efficiency cannot continue t o increase like Mi5 without exceeding 100%:
it must level off somewhere (or even decrease), corresponding to a marked
reduction of the speed exponerit.
The experimental rocket data do not extend beyond Mi(= Ui/co)M 8,
which corresponds to an effective convection Mach number M, M 4. The
use of the convected quadrupole (or dilatation) inference of constant efficiency
a t higher speeds is purely speculative because the Uis law on which it depends
is just an assumption there. The approach of Phillips [5] - an asymptotic
theory for high values of Ui/co - predicts that the efficiency must ultimately
diminish as Miw9l2for constant assumed turbulent intensity.
8. Effects of Density and Temperature
When the jet consists of one gas exhausting into another the jet density
may differ substantially from the ambient value po. We can generalize
from remarks following (5.6) that the sound power may be expected to vary
as p2/po2, where i; is an effective average density in the jet. If the density
of the jet gas is pi at the nozzle, then the ratio p2/po2 will be nearer unity
than pi2/po2 because of the mixing.
Experimental measurements [47] show the sound power depends more
strongly on density than the theoretical j2/po2, in fact, the more extreme
pj2/po2fits the data very well over a range from 16.7 (Freon 12/air) to 0.027
(helium/air). The reason for the discrepancy is not clear. I t is tacitly
assumed in the theory that the mixing dynamics are unaltered by change
in density. This is false for a hot jet [53] and unproved for an unheated jet
of a foreign gas.*
(Added in proof.) More recent information discussed in 1941 shows a dependency
of jet spreading on density ratio (and on Mach number).
137
The pj2/po2 variation of sound power breaks down when the density
change pi arises from heating. Indeed, measurements on air jets heated
to as much as 1000"F [47, 541 and turbojets at much higher temperatures
[55] indicate no sensible effect of temperature : the proportionality between
sound power and the Lighthill parameter p o A U t / c 2 (5.10) "holds for a
wide range of jets from very small jets, both hot and cold, up through several
sizes of jet engines" [54].
A 1000" F air jet has about 1.3 times the rate of spread of a 90' F jet [53],
and this appears to be associated with a 1/1.3 reduction in typical sound
frequencies [54], presumably due to the reduced shear. This would not
account for the failure to scale with p2/po2, as the effect would be in the
wrong direction (a further decrease in emission). Furthermore, the frequency
effect is probably largely offset by increases in turbulence scale and effective volume of the noise-emitting region.
The writer suggests that turbulent heat transport generates additional
sound from entropy fluctuation offsetting the reduction associated with
reduced density. The appropriate entropy term can be found e.g. in [lo, l l j
(Appendix A). The entropy spottiness arises from the turbulent nature of
the transport of hot fluid from the interior to the cool exterior. Furthermore,
the entropy fluctuation in time may be expected to be rather similar to
that of the turbulent velocity.
The foregoing has left out of account differences between the mean
square speed of sound inside (2) and outside (co2) the jet. Lighthill [2]
argues that a ratio c o 2 / s large compared with unity, e.g. 5.3 for a Freon
jet in air, may substantially augment the sound emission (especially at the
higher frequencies), aside from the direct density effect. Experimentally,
only the density effect - mentioned above - appears to be in evidence [47].
We may reexamine this in terms of the source strength on the dilatation
theory. In terms of the local mean square speed of sound 3 the source
strength is - ($)-1a2fi(o)/at2 (4.2). But this source is radiating into a medium
characterized by .? nearby with a transition to co2 outside the jet. Moreover,
a t low speeds the jet dimensions << typical radiated sound wave length.
A crude analogy is an oscillatory point source Q, centered in a sphere of
imbedded in ambient gas
stationary gas of density ii, speed of sound (S)ll2,
of density po and speed of sound co. We ask, what is the source strength Qo
that will generate the same far-field sound when p = Po, 3 = co2. A simple
analysis shows that, for a sphere with circumference <( wave length,
Qo = (po/p)Q,. If the inner and outer gases are the same, but at different
temperature, then p o / p = $/co2. Thus the effective source strength in the
equivalent unheated jet is
138
H . S. RIBNER
in which the speed of sound co outside the jet appears. This means that jet
temperature should have no direct effect on the sound power radiation for
wave lengths >> the jet dimensions. (The associated density effect from
f~@l
remains, however, in theory.)
I t will be noted that (5.26) still holds approximately when the gas sphere
of the analogy is helium or Freon
12 imbedded in air, with no heating.
That is, for these gases po/p M c2/co2within f 20%. Thus it is not surprising
that no effect of speed of sound on noise power was found experimentally [47].
(Sound speed differences have, however, a powerful refractive effect ; this
was discussed in Section V.5.)
Fs
The basic IJ8D2law for the noise power from similar jets may be
compared with the U 2 D 2 relation governing the thrust. Thus an increase
in jet diameter a t constant thrust will provide a net reduction in noise: the
increase in D will be offset four-fold by the corresponding reduction in U .
(We overlook here the modest effect of jet density.) This powerful effect
of reducing the jet velocity by going to a larger diameter engine has been
exploited in the bypass or turbofan engines. In this approach the enlarged
compressor constitutes in part a fan to provide an annular flow of air that
bypasses the combustion chamber. In the pioneer Rolls Royce Conway
the bypass ratio is 40% and in one case cited [56] the noise is reduced 7 db.
Coupled with the reduction in jet noise, the bypass principle has other
virtues that make it attractive. The thermodynamic efficiency is good, and
the increased diameter increases the propulsive efficiency. There is a net
gain in economy. The favorable features have led to a trend to engines of
much larger bypass ratio, but with the penalty of fan noise emerging as an
increasing offender.
The first generation of jet engines did not employ bypass and even the
bypass engines require further quieting as the power goes up. Thus since
the early 1950s engineers have been faced with the urgent requirement for
quieting the engines. The mathematical theory has not, unfortunately,
provided a clearcut guide. The techniques have been motivated by conflicting
interpretations of the theory, but have nevertheless met with a fair degree
of success. Furthermore, the explanation of successful muffler behavior
is still a matter of controversy and speculation. In view of the very limited
understanding of the subject (as it appears to the writer) no very detailed
discussion will be attempted here.
The inference has been drawn that reduction in the mean-flow shear
should be beneficial in two ways. First, it should reduce the expected direct
noise enhancement due to mean shear [2]. It is now thought, however, that
the amplifying effect of shear is approximately offset by an associated
reduction in scale [3, 111. Second, reduced shear should reduce the
T H E G E N E R A T I O N OF S O U N D BY T U R B U L E N T JETS
139
140
H. S. RIBNER
quencies can be scattered by the larger eddies in a jet, the criterion being
wave length comparable with or less than eddy size [62].
In the direction of peak sound a corrugated nozzle produces a bellshaped
curve of sound reduction vs. frequency to be subtracted from the spectrum
obtained with a standard nozzle. The peak reduction shifts toward higher
frequencies as the width of the convolutions is decreased (their number
increased) [59]. Multiple nozzles behave similarly [61]. It is argued [63]
that the convolutions tend to eliminate turbulent eddies of dimensions
comparable with the convolution width; thus sound of the associated band
of frequencies is minimized. An alternative speculation of the present writer
is that the shielding effect may be optimized with a certain ratio of corrugation width to sound wave length.
The spectral curve of sound reduction provided by these nozzles goes
somewhat negative at the upper end; i.e., the high frequencies are increased,
rather than attenuated. This is to be expected from the greatly increased
perimeter, compared with that of a round jet, of the initial mixing region
responsible for the higher frequencies (cf. the derivation of the f - 2 spectrum
law). Unhappily, the higher frequencies are substantially more irritating
to people, it is now realized [64]. Thus although the corrugated and multinozzles substantially reduce the noise intensity, there is little reduction in
degree of annoyance. That is to say, there is little cHange in the PN db
(perceived noise level, db), a rating scale that weights the frequencies according to annoyance [64-681.
Despite the pessimistic tone of the early remarks of this Section,
R. Lee et al. [48] have had an impressive degree of success with an ambitious
semi-empirical formalism for predicting nozzle-suppressor behavior. It is
unfortunate that there has not been time in preparing this review to give
their voluminous report the study it deserves. Several alternative formalisms
are proposed and the results compared with relevant parts of a comprehensive
set of scale model tests which included, besides the conical-nozzle jet, a
rectangular jet, a pair of interfering rectangular jets, and the jet from an
%lobe corrugated nozzle.
One simple notion therein is, in effect, that the sound power radiated
from a slice of jet is related solely to the local mean velocity according to
(5.27)
141
with constants a and @ obtained from source tracing (cf. e.g. [67]). (Also
included is an alternative to (5.27) that replaces Us by U s x (turbulent
shear
this is somewhat closer to the accepted theory.) Experimental
patterns of U for e.g. the interfering rectangular jets and the 8-lobe nozzle,
are used to evaluate (5.27), and this with (5.28) yields the predicted spectrum
for each case.
The agreement with the measured acoustical spectrum data varies
from fair to very good, with the spectral distortions being well predicted.
Thus although (5.27) would appear to be a gross oversimplification, its
success in conjunction with the easier-to-rationalize (5.28) is significant.
I t may well be that the interaction between the turbulent quantities and the
mean flow provides a measure of validity to (5.27) as an integral.
Another important development in the report of Lee et al. [48] is a
digital computer procedure for computing the development - the mean
velocity U (and shear stress) as function of position - of one or more parallel
jets of arbitrary initial cross section. Good agreement with experiment is
obtained in the example of two interfering rectangular jets. This computation
provides the data for (5.27); it appears, however, to have been a parallel
development and was applied only in the alternative form cited under (5.28).*
The method is a development of a principle of mixing due to Reichardt.
Here again the assumptions are oversimplified with respect to accepted
theory, but substantial success may be claimed.
The results for overall power are extended to directivity by this hypothesis based on the experimental data: the directivity characteristics of
jet noise are functions only of frequency and are essentially independent
of nozzle (or suppressor) configuration. Satisfactory agreement of overall
directivity predicted on this basis with experimental values was found for
a 6-tooth nozzles, an %lobe segmented nozzle, the same with a shroud,
and a 21-tube nozzle. These results suggest that the shielding principle put
forward above may play only a minor role.
The apparent dependence of directivity primarily on frequency despite
changes in nozzle configuration can perhaps be explained. Nozzle configuration will probably not alter convection speed in the dominant noise generator,
the mixing region, so the directivity due to convection will be unaltered.
The refractive effect may depend primarily on frequency via the ratio
wave-length/mean diameter. Finally, the proportions of bass and treble
emitted from unit volume on the two spectrum theory vary in a definite
way with direction (Secs. V.4 and V.6).
* The relation for frequency was also somewhat modified. The agreement between
predicted and measured spectra was good for a conical nozzle and fairly good for an
8-lobe nozzle.
142
H. S . RIBNER
R. MATHEMATICAL DEVELOPMENT
VI. GOVERNINGEQUATIONS
1. Wave Equations
where c2ap/ax, has been written in place of the pressure gradient aplax,
by virtue of the isentropy. Elimination of pui by cross-differentiation and
subtraction yields
* The case of the hot jet requires consideration of entropy fluctuation (cf. Sec. V.8) :
see .Appendix A of [ l l ] for appropriate equations.
143
The acoustic waves in the flow may be neglected in the forcing term if
they are weak. That is, the aerodynamic flow may be approximated as
incompressible at low speeds. Stated in another way, the back reaction of
the sound field on the flow generating the sound is negligible in the quasiincompressible range of flow speeds. This should cover jet turbulence in the
schockfree range.
It is equally valid, if the fluid is unbounded, to treat the sound field as
generated by dipoles of strength a(puiui)/aci or by quadrupoles of strength
puiui (cf. Sec. 111.3). This follows from the fact that the source strength
has the form of a double divergence and puiui approaches zero outside the
region of flow. The implications of the equivalence of such patterns of
monopoles, dipoles, and quadrupoles are discussed in Sec. 111.4.
We may split the pressure perturbation into the pseudosound pressure
p ( O ) and the acoustic pressure # ( I ) :
(6.3)
- #lo = p ( 0 )
+ $(I.
azp
7 at2
v2p =
a2(puiuj)
axiaxi
Lighthill
Equation
Upon subtraction of (6.4)
Since the dilatation equation (6.6) has the simpler source term - the
complexities are buried in the pseudosound pC0) - we shall deal with it first ;
later we shall exhibit the corresponding solution of the Lighthill equation.
Let a volume element in the source field be d3y at point y. Then the solution
144
H. S. RIBNEK
pl)
(6.7)
I:I[
#(l)(x,t)= - -4nco2x
d3y
___
(Dilatation).
Component sound waves too long to meet the criterion must be excluded.
The corresponding solution of the Lighthill equation is discussed in
Section 11.4; a convenient form due to Proudman [7] is
(6.9)
Far
Field
p(l(x,t)= ___
4nco2x
1[
____
d3y
(LighthillProudman)
where u, is the component velocity, including the mean flow it any, in the
direction of x. The equivalence is proved in [ I l l .
The two equations are of parallel form. Further developments will
refer at first to (6.8): they can be converted to refer to (6.9)by replacement
of the pseudosound pressure p ( O ) by the x-momentum flux density puX2.
The acoustic intensity at x is given by pc,2/poco. The power radiated
in the direction of x (per unit solid angle) is therefore
(6.10)
p(e)= x 2 p ( x , e ) / p o c o
where x makes the polar angle 8 with the jet flow direction. (This and
what follows applies to a round jet: for the more general case where there
is an azimuth dependence P(8) becomes P(O,$)). The square of (6.8) may
be written in a special way for insertion into (6.10) to yield
(6.11)
11-
P(8) = (16n2poco5)-l
p(0)(y,i)p(o)(y,i)d3yd3y,
the average being over time. The integrand is the correlation of P ( O ) a t the
two points y and y a t the different retarded times i and 2. This space-
145
time correlation will be a function solely of the space separation y' - y"
and of the time separation 1' - t". Thus it will be convenient to make the
definitions
(6.12)
g = y' - y",
y = (1/2)(y'
+ y"),
t=
k - P,
and call the integrand R(y,g,t). It will be noticed that y lies halfway between
the two correlated points. The integration limits in y and 5 are again infinite;
however, in practice the limit on y may be reduced to the effective flow
volume I/ while retaining for convenience in calculation the infinite limits
on g. Then the power in direction 8 is
(6.13)
u
(6.14)
where Ro(y,g,t) = p(o)(y',?)p(o)(y'',~')is the space-time correlation of
pseudosound pressure.
The difference (6.12)of the sound travel times from y' and y" to the
observer a t x is given by t in
cot = (x - y"l -
Ix - y'I
Cot
=5*
x/x
that is, the projection of the point separation = y' - y" on the vector x.
This relation is to be inserted into (6.13)or (6.14); in the latter case the
?Plat4 operation must be carried out first.
The integral over 5 in (6.13)may be interpreted as the value of (p(o))2
at point y multiplied by a certain volume, the 'correlation volume'. This
volume is an effective eddy size, or region within which values of p ( O ) are
well correlated. Because of the time delay (6.15)in the integral, the correlation volume may depend on the direction 6 of the emission vector x. Thus
the integral states in effect that
146
H. S. RIBNER
(6.17)
where P(8) is the acoustic power per unit solid angle in direction 8 and
P(8,w) is the spectral density of this power. An expression for P ( 0 , o )can
be obtained from the theory of stochastic processes as
m
that is, as x2/poco times the Fourier cosine transform of the sound pressure
autocorrelation p(')(~,t)p(~)(x,t
z') at point x. (The factor x2/pocuarises in
the conversion (6.10) from mean square pressure to power.) The autocorrelation is the correlation of the value at a given time with the \ralue a
certain time increment t' later.
The far-field sound pressure autocorrelation appearing in (6.18) mav
be written as a generalization of (6.14) in the form (cf. (6.10))
where t satisfies (6.15) and t' is constant for the integration. l'he Fourier
cosine transform (6.18) of (6.19) may be carried under the y and integrals
to be performed on Ro first, if convenient. This would give
(6.20)
The cos w t ' may be replaced by exp (- z w t ' ) in (6.18) if the limit zero is
replaced by - 03 and (2/n)by (l/n). Correspondingly (6.20) may be
replaced by
147
- +
- +
?he Proudman form of the Lighthill equation for radiated sound, already
introduced as (6.9), will be central to many of the developments herein.
This equation and underlying developments of Lighthill on which it rests
will be derived in the present section. Lighthills treatment of the amplifying effect of shear will also be sketched, although it will not figure
in the approach followed in the present article.
This will be a convenient place to generalize the conservation equations
(6.1) to allow for viscosity, as Lighthill did. The exact equations for mass
and momentum may be written
148
I
S.
. RIBNER
(6.22)
(6.23)
where co2ap/axi has been added to both sides of the momentum equation
such that
(6.24)
T81. .- puiuj
tij
+ ( p - co2p)dij
The right-hand side is an effective source strength and the solution for
p(x,t) is of the form
(6.26)
where [ ] again denotes evaluation at the retarded time (cf. after (6.7)).
Two applications of the divergence theorem yield (see e.g. [23],pp. 62-64)
(6.27)
m
If further r is large compared with the dimensions of the flow this reduces
to the far-field relation
(6.29)
149
P ( x / x ) -- x ~ $ ~ ~ ( x ) / p o c o
The square of (6.29) may be written in a special way for insertion into (6.10)
to yield
(6.30)
The integrand here, which is a summation of correlations as i, j , fi, 1, take
on different values, is developed further in [I].
The term Tii can be identified as a generalization of the quadrupole
strength discussed earlier. In a jet flow, if we esclude cases with density
greatlv different from ambient, Tii will be dominated by the momentum
flux or Reynolds stress pu,uj. Equation (6.29) (with the left side replaced by
p(*)/co2)thus reduces to
(6.31)
W
* 38p0f(U,2)5/?co-2
as the total acoustic power generated by unit volume in terms of the mean
rate of dissipation E per unit mass. This was, of course, radiated uniformly
in all directions. Lilley [a] has applied this in the calculation of the socalled 'self-noise' generated by the turbulence in a jet - the part not
150
H. S. RIBNER
according to the argument. Equation (6.34) states that the time fluctuations
of momentum flux (or Reynolds stress) are dominated by the product of
pressure and rate-of-strain. (The latter defines the distortion of a fluid
element .)
In [2] - and this is further developed in [4] - a strong presumption
is made that the mean shear d,, overrides the fluctuating shear within the
mixing region of a jet. Thus the dominant quadrupole is T,, w pu1u2,and
(6.31) reduces to
p ( x , t )w
(6.35)
xlxz
__I
4nc02x3
5 [z]
E,,
d3y.
The interesting feature of this is the directivity factor, which for the mean
square sound pressure for a round jet takes the form [2]
(6.36)
[1
151
VII. CONVECTIONEFFECTS
FOR
SIMPLIFIED
MODEL OF TURBULENCE
semi--froz&
pattern
(a)
Ur
semi-frozen
pattern
(b)
__
may be expected to look like R,, Fig. 2l(a). If, now, the turbulence is
convected with velocity Uc the correlation pattern will appear like Fig. 21(b)
to the stationary observer. The transformation is
* Actually, Lilleys analysis [ 4 ] appears to suggest 4 rather than unity. The constant
must be 4 in order that the sound power associated with sine 20 be about four times
that associated with the omnidirectional part, the ratio stated by Lilley.
162
H. S. RIBNER
153
The acoustic power radiated in direction 8 (per unit solid angle) from
unit volume of turbulence at y in a round jet is designated P(8,y). From
(6.14) this is
(7.4)
P(e,Y)= ( 1 6 7 ~ 2 p ~ c ~ 5 ) -at4
1a4 R0 (Y*6*t.)d3E
a3
Ro = Rp = pC0)(y e/2, t
+ t ) p ( ~- 6/28
-___-
t)
(74
Ro = Rx = pux2(y
wherein p has been replaced by its local mean p. This equation exhibits
the general character of a convected fluctuating pattern shown in Fig. 21(b).
A typical radian fluctuation frequency is of. An effective eddy volume
or low-speed correlation volume is the value L3 of the volume integral
of (Rs/p2T2)d36
when t is set equal to zero. L may be interpreted as a
154
H . S. RIBNER
directional-average scale of turbulence. (The scale anisotropy of jet turbulence may be simulated by the use of longitudinal and transverse scales
L,, L,, L, [13, 111 but the refinement seems unwarranted in view of the
basically oversimplified form represented by (7.7).)
The integration of (7.4) with Ro given by (7.7) and with the time delay
therein specified from (6.15) as (Fig. dl(b))
cot = t1cos 8
(7.8)
+ & sin 8
yields the sound power radiated in direction 8 (per unit solid angle) from unit
volume at y as [ l l ]
(7.9)
c = [(I - M,cos
0,2~2/nC02~1/2.
An alternative form is
(7.11)
C EZ [ ( l
- M,cos 0 ) 2+ a2Mc2]1/2
+ & sin 0 )
(1 - M,cos 8)6C-5.
By virtue of the (1 - M , cos 8)-5 factor outside, the final convection factor
is jtist the C-5 of (7.9). Thus the moving-frame integral with time delay
possesses a zero that exactly cancels the (1 - M , cos /3)-s singularity. (This
zero follows from (7.3) which implies that the path of integration in the
E, - U c t plane lies along the Uct-axis for 1 - M,cos 8 = 0. Since the
integrand is a t-derivative that vanishes at infinity, the integral vanishes.)
155
+ + t32]}
cos coot.
R,,
[ ( w t v3-1exp - ~ U , ~ / ~ U J / ~ I R A U J ~
(7.15)
where the weight factor in square brackets is the Fourier cosine transform of
esp - o ~ , ~ t 2 .
Suppose now the spectral density P(B,y,cn) associated with the singlefrequency correlation (7.14) has been determined. Then the spectral density
associated with the random pattern described by (7.15) is
(7.16)
J,~/~UJ~~~P~B,~,UJ~~UJ~
-+
156
H. S. RIBNER
of equation (E 14). The results may be expressed in terms of P(8,y) (equation (7.9)) as
0 z 1 - McCOS8,
(7.10)
P(e,Y,w) = P ( e , Y ) w P - w o i o p ) .
wp = 2
V5o,,/c.
157
4. Moving Jets
where again a uniform speed of sound ($)*/' = co has been assumed. When
the stream speed Uo is limited to subsonic values this has the solution (e.g.,
[37], Eqs. (3.5.0) and (3.5.2))
(7.25)
where
,.
t^ = t + [M&l
(7.26)
- Yl)
-~1/C0PO2~
with
Po2 E 1 - U0'/c 0 2 -= 1 - M 0 2.
Evaluation of the mean square pressure involves the two-point spacethat figures in u. We shall
time correlation of pseudosound pressure
assume the convected Gaussian form
R, = p e x p {-
n ~ - 2 [ t,
(v,+ ~ , ) i ] 2- ~
-w
~ - 2 6 ~n ~2 - 2 ~ , 2
, ~ )
(7.27)
(7.28)
p y x , e )=
3014p72L3 1 M , cos 0,
4n2c04x2 (1 - Mo2sin20)C6
158
H. S. RIBNER
cos 8, =
Stationary 'jet'
FIG. 22. Comparison of directional sound patterns of simulated stationary and moving
supersonic jets. M , = Mach number of external uniform stream ( - motion of jet
nozzle). M + M,, = Mach number of convection of eddies through effective jet volume V
(based on external sound speed c").
Local normal
to Mach cone
origin
stream) for a constant relative eddy speed M , = 2.0 (.- excess of jet speed
over stream speed). The results are shown in Fig. 22 as curves of rms sound
pressure versus 8 for a fixed radial distance x .
159
The sweepback of the noise peak results from convection of the wave
pattern. The peak at 8, is normal to the Mach cone and satisfies M , cos 8, = 1.
The angle 0 of the M , = 0.8 peak is related to the angle 8, of the M , = 0
peak by the construction shown in Fig. 23.
1Jpon deletion of the cross-hatched Mach cone and its normal Fig. 23
applies to directions other than the noise peak. The figure then shows the
relation between the angle 8 measured from the actual origin (in a frame
attached to the moving aircraft) and the angle 0, measured from a certain
effective origin, at the timr of signal reception. A t the earlier time of signal
emission the jet nozzle was located back at the effective origin.
By the geometry of Fig. 23 it can be shown that
(7.30)
x2(1 - MO2sin2
0) = xC2(1 M,cos 1 9 , ) ~ .
If the left-hand factor is replaced by the right hand factor in (7.28) the latter
takes a form given (for the directional part) by Williams 1131. Williams
formula thus applies to the geometry a t the time of signal emission and our
formula (7.28) to the geometry a t the time of signal reception. The former
gives a snapshot of a transient sound pattern whereas the latter remains
statistically steady in time (for fixed .Y) relative to an origin in the moving jet.
A feature associated with the moving origin is the sweepback of the
sound polar exhibited in Fig. 22. A corresponding polar reckoned relative
to the stationary effective origin of \Villiams (for fixed x,) is unswept.
VIII. REFRACTION
EFFECTSDUE TO
THE
MEAN FLOW
Earlier qual-
1. Convected W a v e Equatioii
a 2 ( p ~ ~a>fiaYl
t~~)/
in the inviscid approximation, where the space coordinate has been changed
from x to y. This same term also governs the pseudosound pressure pi0)
that figures in the dilatation theory (cf. ((5.4)). In the low-speed approximation the density p of (8.1) is treated as a constant po, an incompressibleflow assumption. Actually, however, the density derivatives can be important
at the higher flow speeds, and even at low speeds they account for refraction
effects.
Thus (8.1) may be expanded as [lo, 11, 23, 411
160
H. S. RIBNER
which shows how the flow velocity weights the density derivatives. If this
is inserted
- into the wave equation (6.2) together with the isentropic relation
d p M c2dp there results
(8.3)
where
of the dilatation equation. The source term on the right-hand side is the time
derivative of p:) following the mean motion.
Here p:) is a modified pseudosound pressure defined by (8.5). It can
be argued [ll] that p:) approximates the local pressure field within the
flow up to much higher Mach numbers than does the original pC0) of (6.4).
Moreover, the replacement of the right-hand side of (8.5) by incompressibleflow values is here a much better approximation than for pcO).
The remaining pressure )#!
in (8.8) may be called the acoustic pressure.
Whereas the pseudosound p:) is dominant in the near field, the less rapidly
decaying
is dominant in the far-field. Equation (8.6) is thus an equation
for the far-field pressure.
pi)
161
2. Form of Solution
1-
&)(x,t) = - (4n)- [ ( ~ ~ ) - * D ~ p / D t ~ ] ~ , t f G ( ~ l y ) d ~ y
V
![
i
1 ap
ar p
1 ar ap dS
+(x,t) = - - -- +--+--4n r an
an r2 cor an at
S
that is, as the emission into air at rest from a certain virtual distribution
of sources and dipoles on S. This equation may be applied to the emission
from a point source a t y within S to obtain G ( x / y )for an external point x.
(Kirchoffs formula is applicable as well to the entire aerodynamic sound
field generated by the jet [77] ; this is a point of some practical significance.)
162
H. S. RIBNER
In the far field, defined by x >> dimensions of S' and 2nx >> wave length,
the distance I = IX -- yl in (8.11) approaches x . The pressure p radiated
by the surface distribution approaches asymptotically an x - l dependence
and depends also on the direction of x. Thus we may write
Far
(8.12)
Field
where 0 is the polar angle between x and the jet axis and # is the azimuth
angle.
For a circular. jet x may be restricted to the zl,x2 plane with no loss
in generality because of the axisymmetry. TheJ! , I dependence then disappears,
3. Qualitative Effects on Directivity
The factor K(O,i,h,y) of the Green's function describes the directional
distribution of sound pressure in the far-field radiated from a point acoustic
source at point y in a specified jet flow: the refractive and diffractive effects
of the flow are embodied in this directivity function. The analytical determination of K(e,#,y) is a very formidable task (see [78] and [79] for similar
problems) and only very idealized cases have been treated [80-83].
M.1.2
o0
-i
-I
\Source
FIG.24. Idealized example of refraction of sound by a flow field: transmitted directivity pattern for oscillating line source near a velocity discontinuity.
P. Gottlieb from ms of [80].
(Supplied by
163
IX. IMPROVED
MODEL: ISOTROPIC
TURBULENCE
SUPERPOSED
ON MEAN FLOW
I . Remarks
011
Lilleys Approach
164
H. S. RIBNER
We return to the formula for the directional power from unit volume at y,
(7.4)
where R, is taken as the Proudman form
(9.1)
R,(O,y,g,t) = F2ziX2zixt2
(9.2)
zi, = u,
+ ux
__
165
Three of the other terms are time-independent and differentiate out in (7.4).
The others are zero or virtually zero, depending on the assumed statistics.
Hence in what follows the other terms in (9.3) may be neglected.
To make progress we shall replace the real nonuniform turbulence by
turbulence postulated as homogeneous and isotropic within a correlation
volume. The statistical values may, however, vary from one correlation
volume to another; that is, they may vary with the position y, the centerpoint
of the correlation volume
over which Q and Q are allowed to range. The
isotropy ensures that u x u i is independent of the direction of X, when
expressed relative to-axes
one of which lies along x. If this axis be chosen
_ _
as the 1-axis, then uxu; = ulul.
We require further that the joint probability density of u, and u, be
Gaussian. Then it can be shown that
(9.4)
-2
uZ2ui=
2 u12
+ 2R12,
where
R,
~
G
u X u l G ~ulul.
Also U , = U cos 0, where 8 is the angle between the observer vector x and
the local mean velocity U. Thus with the assumed flow model and the
other terms neglected
(9.5)
jF2Rx(8,%,t)=
q2+ 2R12(S,t)
+ 4R,(P,t)UU(Q
cos28
self-noise
shear-noise
166
H. S. RIBNER
Accordingly we shall examine (7.4) in the moving frame with zero time
delay. Moreover, in this frame it would seem reasonable to assume that
R,(g,t) is factorable into a space function and a time function; more
specifically we shall assume that
exp - wltl
RlK4 =
(9.6)
shear-noise
self-noise
R I ( ~=
) ?[f 4- (5z2
h2)f/261
t 2= El2
/ = exp (- nE2/L2);
+ 622+ t32
R,(Q = ~ ~ - (nE2/L2)
~ [ 1 (nE12/L2)]
exp (- nE2/L2).
self-noise term
-2
= 2- 3/2014u12L3.
167
a jet. To this end we rotate the axes of (9.10) so that the 1-axis lies along U
with which x makes the angle 8:
(9.12)
A crude approximation to UU'(8) for points near the middle of the mixing
region may be made in the form
(9.13)
+ 0 ) - 3 / 2 w , 4 U 2 ~ L 3 13
COS~
I t follows that the two kinds of noise from a volume element in the
mixing region are in the ratio
(9.15)
1/% V 2 cos48.
+ 7
shear-noise - a
self-noise
4(1
u)3l2
Laurence's data [39] for shear in relation to the lateral scale suggest an
effective
value for u of the order of 0.45. If this is applied together with
u12/U2= (0.28)2there results
(9.16)
shear-noise
= 1.16 cos4e
self-noise
on the basis that the zone of peak turbulence in the mixing region dominates
in sound generation. The numerical factor is not sensitive to the value
of a.
The writer regards this result, aside from the
variation, as indicating
no more than an order-of-magnitude agreement between the shear-noise
and the self-noise. The uncertainty arises from the joint uncertainties in
the forms taken for R,(Q (e.g., the deviation from the assumed isotropy)
and U U ' ( 8 ) : the fact that the volume integral of R,(Q by itself is zero
makes the nonzero integral of the product UU'R, quite sensitive to the
assumed forms. The experimental noise directivity curves (Fig. 14) - the
details were discussed in Part -4 - appear to be compatible with a co&
content of the order of (9.16).
168
H . S. RIBNER
where B ( y )U2/uT is estimated as about unity for a point in the center of the
mixing region. Because of the difficulty and uncertainty in estimating the
it is thought that the assumption
value of B(y) 3 a1/2/4(1
(0.18)
has been carried over without change from (7.9); it provides a generalization
of the low-speed results of the last section (wherein C is replaced by unity)
to allow for convection at nonnegligible Mach number M,. This C was
calculated on the simplified model of turbulence of Section VII, and a new
calculation with the improved model of the present section would be expected
to show some change. I t is thought, however, that the change would be
minor, as (7.10) seems already to exhibit the major features expected on
physical grounds and has other evidence (Part A) to recommend it.
The factor (1 cosW) accounts for the basic directivity in (9.18) and
the factor C-s for the downstream beaming due to convection. The effects
of refraction, discussed qualitatively in Secs. V.5 and VIII.3, are not included.
We may approximate the spectrum of the shear-noise term of (9.18)
by the bell-shaped curve (7.21) peaking at
(7.23)
wp
=2
vw,/c.
The corresponding spectrum of the self-noise term will peak an octave higher
at 2wp. This frequency doubling arises from the presence of the square
of the turbulent velocity component in the self-noise (cf. (9.6) and (9.7)).
The spectral density may be written out as
1ti9
Like C,this spectrum was derived for the simplified model of turbulence,
not the present model with correlation (9.6). The overall spectrum of the
jet, however, is not sensitive to the shape of the spectrum from unit volume,
so long as it is sufficiently peaked. It suffices if the peak frequency is
properly located and the area is normalized to unity. In fact, it will appear
later that the bell-curve (7.21) and the &function spectrum of Part A yield
the same f 2 and f- asymptotic laws for the complete jet spectrum.
X. SOUNDEMISSION
FROM
C.OMPLETE JET
FIG. 23. System of coordinates for mixing region. For the developed jet the y,-axis
is shifted to the jet axis.
d3y = n(D
170
H. S. RIBNER
where P(O,y,)dy, is the power emitted from the slice dy, in direction 8 from
the jet axis. The variables on the right-hand side are functions of y1 alone,
being taken a t that value of y 2 for which g(y2,yl)peaks in the radial direction.
0
For the idealized model of a jet discussed in Part A the function g(y2,y,)
is of the form g(y2/y1) (with suitably chosen origin for yl) in regions A
and B. (Strictly speaking, d / c cannot scale with y1 in B, but this will
be ignored in the present development.) With dS taken as n D d y , (an
approximation) in A and 2ny2dy2 in B the integral of g(y2/y1)dSover a slice
of jet gives
(10.5)
y,D in A
y12 in B
(mixing region),
(developed jet).
171
('Oe6)
p(8,y,)
~1
+ cos48
p2ui8D { Y o in A
poco6Cs Y-7 in B
(mixing region),
(developed jet),
where Y has been written for the distance along the jet axis in nozzle
diameters, y J D .
When P ( 6 , y l ) is integrated over a unit sphere the factor (1 cos48)
yields a weighted average convection factor (C-')Av. Thus the total noise
power emitted by a slice of jet at y , satisfies the proportionality (10.6) with
(C-S)Av instead of Cs.
This may be written in the form
(10.7)
< <
172
H. S. RIBNER
f2
and f - 2 Laws
The spectral density of the sound power emitted in direction 6 (per unit
solid angle) from a slice of jet at y1 will be obtained as the first step.
the time scale V. &,-l of the turbulence correlation
Experimentally [a]
function is nearly constant through a slice in the mixing region. Thus when
the integration of (9.19)over a slice of jet is carried out, as was done for (10.4),
the spectral shape for the slice approximates the spectral shape for the
element. The slice spectrum is therefore a generalization of (10.4)
(10.9)
+(we,Yltw/~)
P(e,Y,,w) = W , Y , , ~ ) coS4e
where wpLC/U has been written in place of w,LIU with a change in the
proportionality factor to (215%2)-?
The spectrum P(8,w) from the entire jet is the integral over the slices
along y1; the corresponding @-integral or basic spectrum may be designated
@(O,w). With the similarity assumptionsp
U 2and (in effect) w p U / L C
that have been made for bath regions A and B of the jet this is
V.
V.
(10.11)
The variable of integration may be changed from y1 to wP-l and the slice
spectrum F(w/wp)may be inserted from (7.21) and (7.21)
V.
(10.13)
@(e,w)A
2$)d(E).
@(O,w), VI
P0Co6C6
+ 2a2 + 1) exp (-
w 2 [ 1 - (2a4
2a2)]/8.
173
< <
(10.15)
(10.16)
P
The bracketed expression is nearly constant (< 8yo variation) in the range
0 w
0.2 wb(b) ( p 0.2). Thus we may write
< <
<
(10.18)
< <
~1
(10.19)
( w c / w * ) 2=
qe,w)
(/c/f*)2,
(wC/w*)-2 = (/C/f*)-2,
w
2wp(u)
00.
H(Y)
= Y2/(1
@)2;
Y = w c / o * = fC//*
174
H . S. RIBNEH
The implications of this equation as to the shift of the spectrum peak with 0
and jet speed have been discussed in Part A for subsonic jets and illustrated in Fig. 17.
In conclusion we note that the total radiated sound power is related to
p(e,u))by
n m
(10.23)
(Direct substitution of (10.24) into (10.23) yields an identity and serves, with
use of (10.20), to define the weighted average convection factor ( C - 5 ) A , . . )
175
XI. ASYMPTOTIC
BEHAVIOR
AT HIGHMACH NUMBER
1. Inferences from Convected Quadrupoles or Dilatations
where 4, is the local mean velocity and the speed of sound is ultimately
required to vary but slowly with distance (gradients are neglected) to yield
(11.3)
ax, ax,
176
H. S. RIBNER
(11.4)
177
Fig. 26 for several Mach numbers M . For large M the bulk of the radiation
is nearly normal to the flow, the peak a t a,,,lying just a little below the
cutoff at cos a = (2M)-l. This property suggested to Phillips the possibility
"of a simple estimate of the acoustic energy flux N per unit area from the
shear zone using expressions derived by Ribner [88]". He wrote
(11.5)
10-
8-
f(a)
64-
FIG. 26. Relative mean square sound pressure / ( a ) in Mach waves radiated in direction a from supersonic turbulent shear layer. (Replotted from Phillips [ 5 ] . )
If the turbulence relative intensity and scale are essentially constant with M
(as assumed in deriving the low-speed U8 law), then this equation implies an
efficiency decreasing as M-3/2. This is an even greater change from the
subsonic M 5 law than the constant supersonic efficiency estimated on the
convected quadrupole or dilatation models.
The experimental data for afterburning jets and rockets (Fig. 19) suggests
a constant supersonic efficiency ( U3 law) ; however, the range is too short for
extrapolation to the very high Mach numbers for which (11.6) is most
applicable.
Other comparisons cited by Phillips deal with the directivity. Lassiter
and Heitkotter [89] measured the directional distribution of the sound
from a rocket whose exit Mach number (based on external sound speed)
was 3.16. They found a directional peak a t about 50", compared with the
57" predicted by the theory for 2M = 3.16. Moreover, as a increased above
60" the experimental intensity dropped some 60 db, approximating the
cutoff shown in the theoretical curves of Fig. 26.
178
H. S. RIBNER
ACKNOWLEDGMENT
This review was supported in part by the Air Force Office of Scientific Research
under Grants AF-AFOSR-62-267 and AF-AFOSR-223-63. Unpublished experimental
work of W. T. Chu (briefly touched upon) was supported also by the Defence Research
Board of Canada. The author wishes to record his appreciation to both organizations
and to Dr. G. N. Patterson, Director of the Institute of Aerophysics, for their support
over many years of the UTIA program of research on aerodynamic noise which provided
the background and motivation for the present study.
References
1. LIGHTHILL,
M. J., On sound generated aerodynamically I. General theory, Proc.
ROY. SOC.A211, 564-587 (1952).
2. LIGHTHILL,
M. J., On sound generated aerodynamically 11. Turbulence as a source
of sound, Proc. Roy, SOC.A%?% 1-32 (1954).
3. LIGHTHILL,
M. J ., The Bakerian Lecture, 1961. Sound generated aerodynamically,
PVOC.ROY. SOC.A267, 147-182 (1962).
4. LILLEY,G. M., On the Noise from Air Jets, A R C 20, 376-N40-FM 2724 (1958).
5. PHILLIPS,0. M., On the generation of sound by supersonic turbulent shear layers,
J. Fluid Mech. 9, 1-28 (1960).
6. POWELL,A., Similarity considerations of noise production from turbulent jets,
both static and moving, Douglas Aircraft Co.. Rep. SM-23246 (1958); abridged
in J . Acoust. SOC.Amer. 81, 812-813 (1959).
I., The generation of noise by isotropic turbulence, Proc. Roy. SOC.
7. PROUDMAN,
A214, 119-132 (1952).
8 . RIBNER,
H. S., On the strength distribution of noise sources along a jet, Univ. of
Toronto, Inst. of Aerophysics, U T I A Rep. 51 (AFOSR TN 58-359. A D 154 264)
(1958); abridged in J. Acoust. SOC.Amev. 80, 876 (1958).
9. RIBNER,H. S., New theory of jet-noise generation, directionality and spectra,
J . Acousf. SOC.Amer. 81, 245-246 (1959).
10. RIBNER,H. S., A Theory of the Sound from Jets and Other Flows in Terms of
Simple Sources, Univ. of Toronto, Inst. of Aerophysics, U T I A Rep. 67
(AFOSR TN 60-950) (1960).
11. RIBNER.H. S., Aerodynamic sound from fluid dilatations - a theory of the sound
from jets and other flows, U. of Toronto, Inst. of Aerophysics, U T I A Rep, 86
(AFOSR TN 3430) (1962).
12. WILLIAMS,
J. E. Ff., Some Thoughts on the Effects of Aircraft Motion and Eddy
Convection on the Noise from Air Jets, Univ. of Southampton. Dept. of
Aero./Astro., U S A A Rep. 155 (1960).
13. WILLIAMS,
J. E. Ff., The noise from turbulence convected at high speed, A R C
23, 323-N. 18PFM 3138 (1961); superseded by Phil. Trans. Roy SOC.Lond.,
Series A, 266, 469-503 (1963).
E. J., Research on aerodynamic noise from jets and associated problems,
14. RICHARDS,
J. Royal. Aero. SOC.67, 319-342 (1953).
15. POWELL,A., Survey of Experiments on Jet-Noise, Aircraft Engineering 26, 2-9
(1954).
L. R. and KORBACHER,
G. K., Review of aerodynamic noise, Univ. of
16. FOWELL.
Toronto, Inst. of Aerophysics, U T I A Review No. 8 (1955).
17. RICHARDS,
E. J., Noise Research in the United Kingdom, Canadian Aero. J . 8,
341-357 (1957).
18. VON GIERKE, H. E., Aircraft noise sources, in "Handbook of Noise Control"
(C. M. Harris, ed.), pp. 88-30 to 88-65. McGraw-Hill, New York. 1957.
179
180
H. S. RIBNER
42. MOLLO-CHRISTENSEN,
E., Jet flow and Jet noise, Mass. Inst. of Tech., Aeroelastic
and Struct. Res. Lab., Rep. ASRL 1006 (1963).
43. TOWNSEND,
A. A., The Structure of Turbulent Shear Flow. Cambridge Univ.
Press, Cambridge, 1956.
44. RAYLEIGH,
LORD,Theory of Sound (2 Vols). Dover Reprint 1945, Vol. 11,
pp. 132-137.
45. RIENER,H. S., Reflection, transmission, and amplification of sound by a moving
medium, J . Acoust. SOC.Amcr. 29, 435-441 (1957).
46. MILES,J. W., On the reflection of sound at an interface of relative motion, J . Acousf.
Soc. Amer. 29, 226-228 (1957).
L. W. and HUBBARD,
H. H., Experimental studies of noise from subsonic
47. LASSITER,
jets in still air, NACA T N 2757 (1952).
48. LEE, ROBERTet al., Research investigation of the generation and suppression of
jet noise (report for Navy, Bu. Weapons), General Electric Co., Flt. Propuls.
Lab. Dept. Cincinnati, 1961.
J. H., An investigation of the noise produced by a subsonic air jet,
49. GERRARD,
J. Aero. Sci. 28, 855-866 (1956).
50. LEE, R., Free field measurements of sound radiated by subsonic air jets, David
Taylor Model Basin, Washington, D.C., DTMB Rep. 868 (1953).
51. HOWES,W. L., Similarity of far noise fields of jets, NASA TR R-52 (1960).
52. MAYES,W. H., Some near-and far-field noise measurements for rocket engines
operating at different nozzle pressure ratios, J. Acousf. SOC.Amcr. 81, 1013-1015
(1959).
S., and UBEROI,hi. S., Further experiments on the flow and heat transfer
53. CORRSIN,
in a heated turbulent air jet, NACA Rep. 998 (1950) (Supersedes NACA T N
1865).
54. ROLLIN,V. G., Effect of jet temperature on jet-noise generation, NACA T N 4217
(1958).
9.5. COLES,W. D. and CALLAGHAN,
E. E., Investigation of far noise field of jets.
11-comparison of air jets and jet engines, NACA T N 3591 (1956).
.?6. HARDCASTLE.
D.. Noise measurements on a Conway engine. Unpublished Rolls
Royce report.
57. LAURENCE,
J. C. and BENNINGHOFF,
J. M., Turbulence measurements in multiple
interfering air jets, NACA T N 4029 (1957).
58. WESTLEY,R. and LILLEY,G. M., An Investigation of the Noise Field from a Small
Jet and Methods for Its Reduction, College of Aeronaut. (Cranfield) Rep. 53 (1952).
59. GREATREX,
F. B., Jet noise, Inst. of Aeronaut. Sci., Z A S Preprint 559 (1955).
60. GREATREX,
F. B., Noise suppressors for avon and conway engines, Amer. SOC.
Mech. Engrs., Paper No. 59-AV-49 (1959).
H. W., Silencing the jet aircraft, Paper delivered at 24th Ann. Mtg.,
61. WITHINGTON,
Znst. Aeronaut. Sci., Jan. 1956.
E., A. and MATSCHAT,
K. R., The Scattering of Sound by a Single Vortex
62. M ~ L L E R
and by Turbulence, U.S. Air Force Office of Sci. Research, AFOSR-TN-69-337
(1959).
63. RICHARDS,E. J., Noise research in the united kingdom, Canadian Aero. J . 8,
341-357 (1957).
64. KRYTER,K. D., Scaling human reactions to the sound from aircraft, J . Acousl.
SOC.Amer. 81, 1415-1129 (1959).
65. KRYTER,K. D., The meaning and measurement of perceived noise level, Noise
Control 6 (1960).
66. BROCH,J. T., Loudness evaluation; a review of current methods, Bruel & Kjaer
(Copenhagen), Tech. Rev. 2 (1962).
181
67. DYER,I., Distribution of sound sources in a jet stream, J. Acoust. Soc. Amer. 81,
1016-1021 (1959).
68. KRAICHNAN,
R. H., The Scattering of Sound in a Turbulent Medium, J. Acoust.
Soc., Amer. 26, 1096-1104 (1953).
69. MAWARDI,
0. K., On the spectrum of noise from turbulence, J. Acoust. Soc. Amer.
27, 442-445 (1955).
70. FRANZ,G. J., The near-sound field of turbulence, David Taylor Model Basin
(Washington, D.C.), D T M B Rep. 982 (1959).
71. BATCHELOR,
G. K., The Theory of Homogeneous Turbulence. Cambridge Univ.
Press, Cambridge, 1953.
J.. and DUMAS,R., Some measurements of time and space
72. FAVRE,A., GAVIGLIO,
in wind tunnel. In French. La Recherche Aeronaut 82 (1953). Translated
N A C A T M 1370 (1955).
J., Round jets in a general stream, A R C Rep. and
73. SQUIRE,H. B. and TROUNCER,
Memo. 1904 (1944).
74. RIBNER,H. S., Field of flow about a jet and effect of jets on stability of jet
propelled airplanes, N A C A Wartime Rep. L-213 (1946).
J. F. J.. A round jet in an ambient co-axial stream, J. Fluid Mech.
75. MACZYNSKI,
13, 597-608 (1962).
76. POWELL,A,, On the reduction of choked jet noise, Proc. Phys. SOC.67, 313-327
(1954).
77. RIBNER,H. S., Review of AFOSR-TN-1308 (1959) by S. I. Cheng On the
aerodynamic noise of a jet, AppI. Mech. Rev. 18,A M R 5494, 777-778 (1960).
78. PRIDMORE-BROWN,
D. C. and INGARD,U., Tentative method for calculation of
the sound field about a source over ground considering diffraction and scattering
into shadow zones, N A C A T N 3779 (1956).
79. PRIDMORE-BROWN, D. C., N A C A RM 57B25 (1957).
80. GOTTLIEB.P., Acoustics in moving media, Ph. D. Thesis, Physics Dept., Mass.
Inst. of Tech. (1959); also, Sound source near a velocity discontinuity, J. Acoust.
Soc. Amer. 82, 1117-1122 (1960).
81. MORETTI,G., and SLUTSKY,
S., The noise field of a subsonic jet, General Appl.
Sci. Labs., GASL Tech. Rep. No. 150 (AFOSR TN-59-1310) (1959).
S., and TAMAGNO,
J., Sound field distribution about a jet, Gen. Appl.
82. SLUTSKY,
Sci. Lab., Tech. Rep. 259 (AFOSR TN 1935) (1961).
83. SLUTSKY,S., Acoustic field of a cylindrical jet due to a distribution of random
sources or quadrupoles, Gen. Appl. Sci. Lab., Tech. Rept. 281 (1962).
84. MORSE,P. M. and FESHBACH,H., Methods of Theoretical Physics Vol. 11.
McGraw-Hill, New York, 1953, pp. 1361-1362.
W. R., The interaction of plane and cylindrical sound waves with a
85. JOHNSON,
stationary shock wave, Univ. of Michigan, Tech. Rep. 2539-8-T (1957), pp. 45-50.
86. SANDERS,
N. D. and LAURENCE,
J. C., Fundamental investigation of noise generation by turbulent jets, SOC.Automotive Engineers, S A E Trans. 66,244-249 (1957).
H. H., Some results of experiments relating to the
87. LASSITER,
L. W. and HUBBARD,
generation of noise in jets, J . Acoust. SOC.Amev. e7, 431-437 (1955).
88. RIBNER,H. S., Note on acoustic energy flow in a moving medium, U. of Toronto,
Inst. of Aerophysics, UTZA T N 21 (AFOSR TN 68-360, AD 154265) (1958).
89. LASSITER,
L. W. and HEITKOTTER,
R. H., Some measurements of noise from three
solid-fuel rocket engines, N A C A T N 3316 (1954).
R. L., Structural aspects of acoustic loads, AGARDograPh 65 (1960).
90. CLARKSON,
A. C., Noise measurements around some jet aircraft, J . Acoust. SOC.
91. PIETRASANTA,
Amer. %3, 434-442 (1956).
92. LYAMSHEV,
L. M., Analysis of acoustic radiation from a turbulent aerodynamic
182
H. S. RIBNER
flow (in Russian), Akusticheskii Zh. 6, 472-477 (1960) : translated in Sou. Phys.ACOUS~.
6, 472-476 (1961).
93. LYAMSHEV,
L. M..
BY V . V . RUMYANTSEV
Institute of Mechanics of the U.S.S. R . Academy of Sciences. Moscow
Puge
Introduction . . . . . . . . . . . . . . . . . . . .
184
I. Simplest Cases of Motion; the Cavity is Completely Filled . . . . . . . 186
1. Irrotational Motion . . . . . . . . . . . . . . . . . . . . . . . .
186
2. Two Examples of Irrotational Motion . . . . . . . . . . . . . . . . 191
3. Uniform Vortex Motion of a Liquid in an Ellipsoidal Cavity . . . . . 196
11. Stability of Motion of a Solid-Liquid Body with Respect to a Part of the
203
Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1. Equations of Motion . . . . . . . . . . . . . . . . . . . . . . .
203
2. Stability with Respect to a Part of the Variables . . . . . . . . . . 206
3. Two Examples (Viscous Liquid, Partially Filled Cavity) . . . . . . . 209
111. Stability of Steady Motion of a Solid Body with Liquid-Filled Cavity . . 214
1. Some Relationships . . . . . . . . . . . . . . . . . . . . . . . .
214
2. A Stability Theorem. . . . . . . . . . . . . . . . . . . . . . . .
217
3. The Problem of Minimum . . . . . . . . . . . . . . . . . . . . .
224
References. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
230
NOTATIOK
Space-fixed system of coordinates
Body-fixed system of coordinates
Cosines of angles formed by axis C with axes x,y,z
Velocity vector of point with coordinates x.y,z
Velocity vector of point 0
Angular velocity vector of solid body
Density of liquid
Viscosity coefficient
Kinematic viscosity coefficient
Unit vector along the external normal to the surface of liquid u
consisting of wet walls q of the cavity and free surface u
Region bounded by surface (I
Moments of inertia of solid (s = 1) and liquid (s = 2) about axes x,y.z
Coordinates of the centre of mass of solid (s = 1 ) and of liquid (s = 2)
Mass of solid (s = 1). and of liquid (s = 2)
Mass of system
Acceleration of gravity
Kinetic energy of solid (s = l), and of liquid (s = 2)
Kinetic energy of system
183
184
V.
V. RUMYANTSEV
INTRODUCTION
The problem of the motion of a solid body with cavities completely filled
with a liquid began to attract the attention of scientists over a century ago.
Stokes [ l , 21 appears to have been the first to draw attention to this
interesting problem of mechanics; it was also taken up by Helmholtz [3],
G. Lubeck [4], and Lamb [5] who investigated a number of special cases:
F. Neumann [6] became interested in it when studying the motion of solid
bodies in a liquid.
N . E. Zhukovskii 171 for the first time made a thorough general study
of the motion of a solid body whose cavities are completely filled with a
homogeneous incompressible liquid. He proved that the irrotational motion
of the liquid in the cavity is determined by the rotation of the body and
does not depend on its forward motion, while the motion of the body proceeds
as if the liquid had been replaced by an equivalent solid body.
Simultaneously with these studies interest arose in the question of
stability of that motion. Kelvin who made experiments with a thinwalled
spheroidal gyroscope containing a liquid [8] found that the gyroscope is
stable if the cavity is sufficiently oblate and if the gyroscope revolves rapidly.
If, however, its shape is slightly elongated, it becomes very unstable
irrespective of its angular speed. The mathematical explanation of this
phenomenon was the object of studies by A. G. Greenhill 191, F. Slutskii [lo],
S. Hough [ l l ] , Poincare [12] and A . B. Basset [13], who analyzed the
homogeneous vortex motion of a liquid in an ellipsoidal cavity. Hough [I13
studied small oscillations about the state of uniform rotation of solid and
liquid as a whole, the axis of rotation being the principal axis of inertia.
He obtained necessary conditions of stability and applied them to a shell of
negligibly small mass. PoincarC [ 121 likewise investigated this problem,
considering also the elasticity of the shell and the non-uniformity of the
liquid.
Considerable interest in this question arose again in our time, especially
in the past 20 years, when the new problem of a solid with cavities tzot
completely filled with liquid came up and was extensivly explored. Numerous
papers of many researchers were devoted to this topic: G. E. Pavlenko [14],
S. L. Sobolev [15], N. G. Chetaev [16], L. N. Sretenskii [17], D. E. Okhotsimskii [18], S. G. Krein [19], N. N. Moiseev [20], G. S. Narimanov [21],
S. V. Malashenko [22], A. Y. Ishlinskii, and M. E. Temchenko [23],
K. Stewartson [24] and many others (see reviews [25, 261).
Two theoretical approaches to the problem of stability are known :
1) the study of the linearized equations of the perturbed motion of the
system, with application of the theory of small oscillations and of the spectral
theory of operators; 2 ) the study of the full non-linear equations of the
perturbed motion with application of Lyapunovs methods [27-291.
185
The linear theory for this problem has been substantially developed
and achieved considerable success. I t provides a quite adequate description
of the motion of the system and its conclusions have been confirmed
experimentally to a certain extent. Lyapunovs criticism of certain investigations by PoincarC [30] concerning the stability of the equilibrium figure of
a rotating liquid may, however, be fully applied to the linear theory also
in the present case. Lyapunov 1291 emphasized that the conclusions about
stability, reached by studying linear equations that have been obtained by
deleting some terms considered as small in the differential equations of the
problem, cannot be regarded as rigorously established. In linearizing
the equations of motion, one problem is in point of fact replaced by another
with which the former may in some cases have nothing in common. There
are no criteria at present for the correlation of these two problems in the
theory of small oscillations of solid bodies containing a liquid; the establishment of such criteria would be of considerable importance for substantiating the linear theory. The only exception is provided by the
conditions obtained according to the linear theory [20] from the Lagrangian
theorem: since the latter is valid by virtue of the full equations [47], these
conditions may be regarded in first approximation as sufficient conditions
of stability. But the complexity and clumsiness of the characteristic equation, for which the signs of the real parts of the roots must be investigated,
are also typical for the linear theory of the stability problem in our case [24].
In this respect Lyapunovs non-linear methods compare favourably with
the linear theory. These methods are mathematically rigorous and permit
in a number of cases to obtain effective sufficient conditions of stability.
When we pose the problem of stability of motion of a solid body containing
a liquid (a system with an infinite number of degrees of freedom) the question
arises: In what sense and in relation to which variables does stability interest
us in this problem? In his famous book [27], Lyapunov, defining stability
for a system with a finite number of degrees of freedom, emphasized the
necessity of specifying those quantities which are supposed to remain small,
since the term stability in itself contains nothing absolute [29].
The present review of non-linear methods in the stability of stationary
motions of a body with enclosed liquid has three chapters. The first deals
with the simplest case, in which the motion of the liquid is fully defined by
a finite number of variables. Such a case is evidently possible only when the
cavity is completely filled with an inviscid liquid. Cases of irrotational motion
and of uniform rotation of the liquid are investigated. Here the problem
would naturally be posed as a stability problem according to Lyapunov for
a system with a finite number of degrees of freedom [27, 311.
The second chapter is devoted to the study of the stability of motion
of a body with a cavity partly or completely filled with an inviscid or viscous
liquid, no assumptions being made regarding the nature of the motion of
186
V. V. RUMYANTSEV
the liquid. The state of the system is now described by an infinite number
of variables, and the problem of stability becomes considerably more complex.
However, in this case, too, it is possible to pose the problem of stability with
respect to a finite number of variables by introducing certain quantities
which integrally define the motion of the liquid.
The third chapter outlines another approach to the problem of stability.
It develops Lyapunovs ideas [28, 291 proceeding from the theory of stability
of the equilibrium figure of a rotating liquid. Proof is supplied that our
problem reduces to a minimum problem for a certain expression, and a
solution of the latter is given. The Lagrangian theorem is also established
for a body with a liquid.
All three chapters supply examples illustrating the application of the
theory.
It will be noted that the solid body with a cavity completely or partly
filled with a liquid is treated as a single mechanical system. The equations
of motion of such a system can be derived [32, 331 from the HamiltonOstrogradskii principle. 7 hese equations are mathematical expressions of
the general theorems of mechanics about momentum and moment of momentum of a system and of Eulers or Navier-Stokes equations of the motion
of a liquid. Under certain conditions, the equations of the motion of the
system have integrals of energy and areas as well as some other first integrals
whose existence is important in applying the methods under consideration.
In most cases we confine ourselves, for the sake of brevity, to motions of
the solid body with a single fixed point.
I. SIMPLEST
CASESOF MOTION;THE CAVITY IS COMPLETELYFILLED
We consider cases in which the motion of the liquid in the cavity can
be fully described by a finite number of variables. Evidently, this is possible
only when the cavity is completely filled (no free surface). It is well known [ri]
that if in this case the liquid is in irrotational motion or uniformly rotating
(quasi-solid), the motion of the system can be described by ordinary differential equations. Such cases are analyzed below.
1. Irrotational Motion
187
is harmonic in x,y,z; on the wall of the cavity, 0 , its normal derivative equals
the normal component of the wall velocity.*
Hence the determination of the velocity potential is reduced to solving
Neumanns problem; and the irrotational motion of liquid in an n-fold
connected cavity is completely determinate, if one gives the normal component of the velocity for every point of the boundary of the cavity and the
value of the circulation for each of the n - 1 independent irreducible closed
curves (circuits) which can be drawn in the cavity [5].
Following Zhukovskii [7], let us resolve the motion of the solid body
into a translational motion proceeding with the velocity of some body-fixed
point 0 and a rotary one about that point. The progressive motion has a
single-valued potential wlx
w,y
w3z; it is directly passed on to the
liquid and does not affect its internal motion. The latter is changed only
by the rotary motion about point 0 and will be the same if we consider
this point as space-fixed and rotate the body about it in the same way as
it rotates about the point 0 moving in space.
The velocity potential d(x,y,z,t) is expressed as
d=x+v
(1.1)
(1.4)
q(x,y,z,t) =
2 wi$i(x,Y,z)
i= 1
188
V . V. RUMYANTSEV
I t follows from the formulae that I$ depends on the time t only by means
of the mi. If, after some period of time the body comes to rest, then 4
becomes equal to x, i.e. the liquid inside the cavity makes the same motion
it made initially [7], when the body was a t rest.
We now turn to the kinetic energy and the moment of momentum of
the liquid which makes an irrotational motion with potential I$. According
to Kelvin's generalization of Green's theorem [5] for a multiply-connected
region
where ui is the area of one of the jth partitions which can be put in the
cavity without destroying its connectivity.
Taking into account the boundary conditions (1.2), we have
r.j= I
A* = p $ ~ ~ (-y zzm)du,
U
B* = p
189
Ry (1.7) the kinetic energy of the liquid, T,, is the sum of the kinetic
energy T,(')(q,w2,u3)of the motion with potential p and of the constant
kinetic energy T2(*)(kl,.
. .,kn- ,) associated with potential x.
The moment of momentum of the liquid relative to point 0, given by
pJ,r x grad +It, is likewise made up of the moments of momentum associated
with p and
The projections of the moment of momentum belonging to
are P,Q,R; they are represented by linear functions of the principal circulations ki and do not change in time.
The projections of the resultant moment of momentum of the solid-body
motion and the motion of the liquid with potential 9 are equal to aT/aw,
(i = 1, 2 , 3 ) .
Equation T,(')(x,y,z)= 1 represents some ellipsoid in the body-fixed
system. Zhukovskii established that the latter always includes the inertial
ellipsoid of the liquid.
Let us now consider some solid body with an ellipsoid of inertia for some
fixed point
x.
T , ( X , Y , ~ ) TZ'')(x,y,z) = 1.
(1.9)
The same inertial ellipsoid can be obtained by joining to the given solid
body, instead of the liquid in the cavity, another solid body with an ellipsoid
of inertia
T,("(x,Y,z)= 1
about the same fixed point 0. Its mass equals that of the liquid to be replaced,
and its centre of gravity coincides with that of the liquid. Stokes called
such a body the equivalznt solid body; it does not depend on the location
of the fixed point 0 and has smaller moments of inertia in relation to any
axis than the corresponding liquid mass, i.e. A* < A , , B* < B,, C* < C2 [7].
If axes x,y,z are directed along the principal axes of the ellipsoid (1.9),
its equation is
AX2
(1.10)
+ By2 + c z 2 = 1
where A,B,C denote the sums of the moments of inertia of the solid body
and of the equivalent solid body, i.e.
(1.11)
A =A,
+ A*,
B = B,
+ B*,
C = C,
+ C*.
190
V . V. RUMYANTSEV
We obtain the equations of motion of the system from the theorem about
the moment of momentum:
dw
A2
dt
+ (C - B ) 0 2 0 ,+ w2R - w ~ Q= L,
~ d w +, ( B - A)w1w2f w @ dt
wzP= N .
Here the moments of the external forces in relation to the mobile axes
x,y,z are L , M , N .
These equations are identical with those of the motion of a solid body
to which a rotating gyroscope is joined. The direction cosines of the axis
of rotation of the gyroscope are in the ratio P : Q : R , while the product of
the initial angular velocity (with the solid body at rest) and the moment
of inertia about the axis of rotation equals (P2 Q2 R2)'/'.
Consequently, the mechanical effect of a liquid in the cavity withozd
initial velocity is identical with that of some equivalent body. If the liquid
in the multiply-connected cavity has some initial velocities then its additional
action is similar to the action of a rotating gyroscope joined to the body.
This important theorem was given by N. E. Zhukovskii [7]. Hence,
for the irrotational motion of liquid in the cavity, the motion of the solid
body together with the liquid is fully defined by a finite number of variables,
with respect to which the stability problem should be posed according to
Lyapunov [27]. The solution of this problem for a number of cases was
given by N. G . Chetaev [16].
It should be noted that the moments of external forces, L , M , N , applied
to the system, may be functions not only of the time and of the angular
velocities mi, but also of some other variables. In such cases the differential
equations for these variables must be joined to the equations (1.12). For
example, in many practical cases we shall have [35]:
+ +
(1.13)
191
W 3 )=
- ay3
if axis ( is directed along the vector of velocity of the mass centre, and
a = const.
For the case of a central force field
if axis O,( is directed from the centre of attraction 0, along the straight
line OIO. Here f is the constant of gravitation, p ( x , y , z ) is the density at
point x,y,z of the solid or liquid, 11 is the volume of the solid and the liquid,
and R is the distance 0,O.
If R is much larger than the dimensions of the body, then one can take
[36] the approximate value of the force function instead of (1.17)
(1.18)
2. Two Examples of Irrotational Motion
Consider two problems dealing with the stability of motion of a solidliquid body with the liquid in irrotational motion.
+ M~~ +
2T = M(w12
+ A o I 2+ B
w ~ CO,~.
~
2T = const,
Aw,w,
7u:
+ w: + w:
= const.
w2 = w3 = 0,
w1= 0,
0 2
= u 3 = 0.
192
V. V. RUMYANTSEV
ru,=w+5
and retaining the previous designations for the rest of the variables, write
down the first integrals of the equations of the perturbed motion
w22
(1.19)
I = v1 - 2AV2
XV,
+ v,2
= (M
~ ~ )
where
1 = o/w,
+x =Ao2/w2.
If
A>B,
(1.20)
A>C,
then function 1. will be positive definite in all the six variables wj,wi and
its time derivative V will vanish as a result of the equations of the perturbed
motion. Proceeding from Lyapunovs theorem about stability [27], it is
possible to state that the inequalities (1.20) are sufficient conditions of
stability (with respect to w ~ , o J ~of) the permanent helicoidal motion of a
solid body with a liquid-filled cavity.
Example 2. Uniform rotation about a fixed axis.
First consider the solid-liquid body moving about a fixed point under
the action of forces possessing a force function U ( y 3 ) . Suppose the cavity
is multi-connected and the initial motion of the liquid is such that P=Q =0,
R # 0. For example, let the cavity h a r e the shape of a torus obtained by
rotating about z some circle with centre not on the z axis. Then [7]
R = M2k/2n, where k is the circulation.
I t can be easily seen that the equations of motion (1.12) and (1.14)
in the case under consideration have first integrals of energy
Awl2
Aqy,
and of areas
1!13
(1.21)
= B.
(1.22)
w1= wz= 0,
y1 = yz = 0,
w, = w ,
y3 = 1
which describes the uniform revolution of a solid body about the lixed axis
of symmetry z, coinciding with axis 5.
Let us assume that in the perturbed motion
y3=1+q1
w,=w+E,
Vl
Am?
V, =A q y ,
V , zz y12
zl)l (
q-
+ Bw,y2 + C ( r q + E + Eq) + Rq
__
q2
+ . ..
= const,
= const,
+ + q 2 + 2q = 0
y22
V,
if condition (1.21) is fulfilled.
(au/aY3)1
6 = const
= (au/aY3)y,=l.
194
V. V. RUMYANTSEV
This is possible, provided the polynomial to the left has two different real
roots:
(1.24)
M'ith this condition satisfied, the constant 1 can be so selected that c' is
positive definite with respect to all variables q , y i (i = 1,2),5',17; in addition,
I/' = 0. Consequently, according to Lyapunov's theorem [27], inequality
(1.24) is a sufficient condition of stability.
Let us now demonstrate that (1 2 4 ) is also necessary for stability. Consider
the function r38j
(1.25)
W = W l Y 2 - %Yl
and its time derivative, taken by means of the equations of the perturbed
motion,
A aY3 1
According to Sylvester's criterion, W' will be positive definite in the
variables q , y , (z' = 1,2), if
(1 26)
holds. Here it is assumed that the variable 17 preserves all the time the
order of smallness of the variables q , ~ otherwise
)~:
the motion would
evidently be unstable with respect to y3.
When condition (1.26) is fulfilled, the unperturbed motion (1.22) is
unstable since the function W meets in this case the conditions of Chetaev's
instability theorem 1311. Hence condition (1.24) is necessary and sufficient
for the stability of the unperturbed motion (1.22).
If there is no initial motion of the liquid in the multi-connected cavitv,
then R = 0, and we obtain from (1.24) the condition of stability
(1 2 7 )
195
= y o = 0,
U(Y3) = - MgzOy,,
(CU
+ R ) 2- 4 A M g z 0 > 0
[3H],
(1.30)
3B
+ (Cu2 + (g),
+ Rw) (y: + y; + q2)+
where the constant
(1.32)
+ RU +
- >O.
(
(C- A)w
+ RW - MgzO>
[39],
and with K = 0. as
(1.34)
(C - A ) d > MgzO
(381.
196
V. V. RUMYANTSEV
of the cavity and is proportional to the density p of the liquid. For instance,
for a cavity in the shape of a round cylinder with radius a and height 2h [16]
where [, are the roots of the equation dJ,(C)/dS = 0, zo and z, are the z
co-ordinates of the cylinder faces, and M , = 2npa2h.
For a cavity in the shape of an ellipsoid of revolution with centre on
the z axis at zo
M (a2- c2)2
A*=> 5 a2+c2
+ M2z02.
This is another simple case, in which the motion of the solid-liquid body
is fully determined by a finite number of variables. Let the cavity have the
shape of an ellipsoid
(1.35)
The motion of the liquid in the cavity can be described by the formulae [9-111
(1.36)
(1.37)
Q2
a3
+ [ ( a 2- b2)xy + 2b2xoy - 2a2yox] a2 + b2
0 3
satisfies the boundary condition for an ideal liquid on the walls of cavity
6.
197
(a2
(1.38)
dn,
= 2b2(*
dt
dt
,+c2
a 2 + b2
+ a2
c2
w3R2, - 2R3Q1
- ___
b2
+ b2)(a2+ c2)
b2(a2- c2)
(b2
+ c2)(a2+ b2)
c2(b2 - a21
(a2 c2)(b2 c2) .
+ c2
- ~ (-y y o ) ] ,
- zO)],
where p, q, and r are functions of time only. PoincarC [12] assumed that
the velocities of the liquid particles are linear functions of their co-ordinates.
He termed such a motion of a liquid a simple motion. Proceeding from
Helmholtz theorem, he pointed out that if the motion of the liquid a t the
initial instant was simple it would remain so all the time, provided the
cavity has the shape of an ellipsoid. In this case the projections on the
mobile axes of the relative velocities are
24 = - Ql(Z
C
- 20) -
r,(y
Yo),
z, = C r,(x
b
- xo) - a P,(Z
- zo),
G,
= A*wl
+ AR,,
G,
B*u, + BR,,
G3 = C*O,
+ CR3
198
V. V. RUMYANTSEV
where
M (b2 -. c2)
A*=-?
5 h2 + c2
(1.40)
(c2c2
B* = L
5
+ M2(y,2 + zO2),
a2),
+ a2 + M,(zZ +
(a8- c , ) ~
c*=M2
+ M&,2
5 a2 + b2
4 M,b2c2
5 b2 c2
-I__
R = 4 M 2 a 2 c 2
XZ),
+ Y$),
c2
+ a2
4 M,a2b2
C = - ___
A a2 f b2
=A,
+ M2(yo2+ zo2),
c c, +
=
M,(XO2
B1
+ M2(zo2 + xo2),
+ yo2).
[The equations of motion (1.38) and (1.41) can also be obtained from
PoincarCs equations [12] in group variables [33].]
Thus in the present case of a solid body with an ellipsoidal cavity filled
with an ideal liquid, the motion of the system is fully characterized by a
finite number of variables. Therefore the problem of stability can, in this
case too, be posed and solved as one of stability according to Lyapunov.
Exanifilc 3. Let us consider a solid body moving about a fixed point
0, = 0, and with a cavity of the kind (1.35), where x, = yo = 0. Suppose
the system is under the action of forces with a force function of the form
C(y,) and assume that the moments L , M , N are determined according to
formulae (1.13).
199
Awl2 + B w , ~ Cw32
AwIy,
in addition,
y12
+ +
y22
732 =
1;
the integral
b2c2RI2
(03
= (0,
(1.43)
R,
= R2 = 0,
y3 =
R,
= R,
y1 = yz = 0,
1,
describing the uniform rotation of the solid body about the z axis (coinciding
with the fixed C axis), and a relative elliptical rotation [7] around the same
axis with components
Let us assume that this is the unperturbed motion and analyze its stability
with respect to variables w,,R,,y,(i = 1,2,3).
It is worth noting that in the particular solution (1.43), magnitudes
w and R may have arbitrary values. Of practical interest are values R in
the interval 0 < E
R w , where E is infinitesimally small, for, as shown
by experiment, during the uniform rotation of the body about a fixed axis
passing through the centre of the cavity, the liquid, unless it has been put
into vortical motion earlier, first makes an irrotational motion (it is at rest),
and then is increasingly involved in the motion of the body by friction
until it moves along with it as a single solid body. In what follows we shall
confine ourselves precisely to this case E
Q
oi.
< <
< <
V.
200
V. RUMYANTSEV
Q3=R+lj7,
y3=1+5
and retaining the previous designations for the rest of the variables.
I t is evident that in the general case the equations of the perturbed
motion will possess the following first integrals:
I/,
E Awl2
1/
I/,
- 2wV2
= Ao,2
+ AV4 + pV3 $- ~ X L , ,
+ py,2
+ (x+,+J)P+
...
A=C-
o-R
Ra2b2
11
= (CW
+CQw +
201
(1.47)
+ C'WO+
(C - B)w2
a2 - c2
4a2c2
( @ 2 + ,2)2
C ' d > 0.
202
V. V. RUMYANTSEV
v = v, -!-2 1 v 2 -
Cwl
C(C
(1.50) = A(wI2
+ -C2
p+
A
02*)
-A)
+ 2 A l ( ~ , y +,
- CwR -
~ 2 ~ 2 )
( ) (Y12+
~
aY3 1
Y22)
2c1tg -
+ (A + pa2c2)(Q,Z+
Q22)
3- 2AI(Q,y,
Q2y2)
- CQA(y12
+ (C + pa4)172 + zcnq2- C m p + . . .
where 1 is some constant and ,u = - C(A + R)/Qa4.
+ Y22)
For simplicity, (1.50) may be regarded as the sum of six quadratic forms
of two variables each*. The sum of the first three quadratic forms in (1.50)
is similar to Lyapunovs function (1.23); it will be positive definite when
condition (1.27) is satisfied. The constant coefficient 1 may have any value
in the interval
I , < 1 < 1,
where &,A2 are the roots of AA2 + CwA - (aU/ay3)1= 0.
(1 51)
The fourth and fifth quadratic forms in the right part of equality (1.50)
will be positive definite when the following condition holds for negative 1:
(15 2 )
2A
n.
203
Some particular cases of this problem have been analyzed in [41]. For
inertial motion about the centre of the cavity (1.35), when a = 6 and the
shell is so thin that its moments of inertia may be neglected in comparison
with the liquid, the sufficiency of the conditions
(15 3 )
a>c
or
c>3a
Ell]
11. STABILITY
OF MOTION OF A SOLID-LIQUID
BODYWITH RESPECTTO
PARTOF THE VARIABLES
(2.1)
204
V. V. RUMYANTSEV
where the G, (i = 1,2,3) denote the projections on the mobile axes of the
moments of momentum of the liquid. As these quantities are not known
beforehand, equations (2.1) must be taken together with those of the motion
of the liquid
dv,
___
dt
1afi + v A v ~ ,
+ w3vl - w1v3 = Y - P aY
+ wlvz - w2v1 = 2
dv
3
at
avl
av,
afi
a2
- --
+ vdv,,
av3
-+-+--=0,
ax
ay
a2
where P(x,y,z,t) is the pressure, d the Laplacian operator, and X,Y,Z the
projections of the body force per unit mass acting on the liquid.
The projections of the absolute velocity of the liquid on the mobile axes
may be represented as
(2.3)
~ 1 =~ 2 2
+ U,
V%= W ~ X 012
V,
~3
=W
~ Y W ~ X W,
where u,v,w designate the projections of the velocity of the liquid relative
to the solid body.
The solutions of equations (2.2) should meet the following boundary
conditions on the boundary of region T , occupied by the liquid. The velocity
of the liquid on the wall a1 of the cavity with which the viscous liquid is in
contact at a given moment of time should be equal to that of the corresponding point of the wall, i.e.
u=v=w=O
(1.4)
on
= 0 ) only
al.
ul+vm+wn=O
al.
The kinematic and dynamic conditions must hold on the free surface a,
of the liquid. The first implies that the velocity of the displacement of any
point on surface ac and the velocity of the particles of the liquid adjoining
the surface a t this point should have the same projections on the normal
to the surface, i.e.
(2.6)
af
af
af
af
ax
ay
a2
at
-u+-v+-w+-=O
on
a;,
205
pn = - p0n
on
a,,
where Po is the constant air pressure on the free surface. In the case of an
ideal liquid, we have instead of (2.7) only the constancy of the hydrodynamic
pressure
(2.8)
=Po
on
uc
Hence, the problem of motion about a fixed point of a solid body with a
liquid in its cavity is formulated by equations (2.1) and (2.2) with boundary
conditions (2.4)-(2.8).
Let us now give some integral relationships and first integrals for the
equations of motion (2.1) and (2.2).
If the forces acting on the solid body and the liquid possess a force
function U , the following equation of energy dissipation is valid [ 5 ] :
(2.9)
It follows that motion of a viscous liquid (p # 0) without energy dissipation
is possible only when everywhere in the liquid
(2.10)
which implies that there is no lengthening or shortening of linear elements
anywhere [5]. This is possible only when the liquid and the solid body
in whose cavity it is located, move as one solid body.
In the case of an ideal liquid ( p = 0 ) we obtain from (2.9) the integral
of energy
(2.11)
T - U = const.
If the moment of the external forces about some space-fixed axis 5' equals
zero, equations (2.1) have the integral of areas
which expresses the constancy of the projection on the 5' axis of the moment
of momentum of the system. If, however, the moment of the external
V. V.
206
RUMYANTSEV
forces about the point 0 equals zero, then we have the integral of constancy
of the moment of momentum of the system
(2.13)
(Aim1
If the liquid is ideal, if the ellipsoid of inertia of the body for point 0
and the cavity have the shape of bodies of revolution about the same axis,
and if the moment about that axis of the external forces acting on the
system vanishes, then the projection of the instantaneous angular velocity
of the body on that axis remains constant throughout the motion [32].
Let for example A , = B, and let the above conditions be satisfied for the
axis z, then the equations of motion have the first integral
m3 = const.
(2.14)
<
qj' = /j'(t)
(i = 1,2,3),
( j = 1,.
. .,%),
= F.,(x,y,z,t),
qio = / i (to)1
qjo' = /j'(t0)
vi(x,y,z,to)= pi(x,y,z)
( j = 1,. . .,%),
(i = 123)*
* The
207
where the .ci,ei' are real constants and the E ; ( x , Y , z )are real functions satisfying
the continuity equation. We shall term the E'S perturbations. By giving
the perturbations, the perturbed motion of the system is fully determined
since the forces acting on the system are assumed to be invariable [31].
We shall consider the absolute value of the perturbations as sufficiently
small.
The equations of the perturbed motion are obtained from the original
equations by substituting for qi,qj',vi,p the quantities
208
V. V. RUMYANTSEV
y,
(r = 1, . . . ,K )
= Q, - F ,
<
<
IQ,
<
- F r J< L
hold for any t 2 to, then the unperturbed motion is stable with respect
to the quantities Q,, otherwise it is unstable [ Z ? ] .
We now prove a theorem [44] which may be useful in some cases as a
modification of Lyapunovs theorem about stability.
Let us assume that there exists a real, continuous, bounded, one-valued
function p(t,xi,xi,ui)which vanishes when the variables xi = xi = u, = 0,
and whose time derivative satisfies cp
0 owing to the equations of the
perturbed motion. Let the inequality
<
(2.19)
Y ( t , y , ,* .
< p(t,xj,xj>ui)
, ~ k )
hold for all values of the variables t,xi,x,,2ti under consideration and the
corresponding values of the variables y,, where !P(t,yl,. . . , y k ) is a real,
continuous, bounded, one-valued function of its variables, which disappears
when all y , = 0 ( I = 1,. , .,k).
<
809
F(t,y1,* * * 9Yk)
w ( Y 1 , . .Yk)
is valid.
Suppose A > 0 is an arbitrarily small number and 1 > 0 is the exact
lower boundary of W on the sphere Z f = 1 y s 2= A .
The function
Q ) ( ~ ~ , X ~ , X ~ ,being
U,)
clearly independent of time, admits an infinitely small
upper limit;* consequently, for given 1 > 0 a number 1 > 0 can be found
such that for all values of the variables xi,xj,u, satisfying
< 1,
(2.20)
lxjj
Jxjl
< 1,
<1
F(io$yi,*
#Yk)
,Yk)
<y(t,yi,
< y(t,xj,xj.ui)<
*Yk)
wg = w ,
Gl= G2 = 0,
G, = C2w,
y1=y2=0,
y3=
1,
(2.21)
u = II = ZE, = 0,
which describes the uniform rotation about axis z coinciding with axis 5
of the solid body and the liquid in its cavity.
Let us investigate the stability of this motion relative to the projections
of the instantaneous angular velocity of the body OJ,, the projections of the
moment of momentum of the liquid G,, and the direction cosines y,, assuming
that in the perturbed motion
< a determines
* Since 1~1(1,,0,0,0)
= 0, the set ( x j , x j , u i ) defined by q~(l,,xj,xj,ui)
for a n y positive a a non-empty neighborhood of 0.
210
V. V. RUMYANTSEV
and retaining the previous designations for the rest of the variables.
First we obtain an expression for the kinetic energy of the liquid in
terms of the G, [45]. During the motion of the system, the G, will be functions of time. Now introduce instead of G,new unknown functions of time x,,
defining them by
(2.22)
If the velocities of the liquid ui are known, this in itself determines
functions x,. Let us also introduce u(i)(x,y,z,t),defined by
(2.23)
~ ( 1=
) vI+ x3y
- x#,
~ ( 2=
) ~2
X ~ Z X ~ X , ~ 1 3=
) 23
X ~ X XI).
By making use of equalities (2.22) and (2.23), the kinetic energy of the
liquid can be represented as
+ Gz2 +
G32,
= max
(A2,B2,C2).
x1 = x, = 0,
x3 = 0
V(l)
= V l 2 ) = V ( 3 ) = 0.
-<o
dt
211
I t can also be easily seen that the equations of the perturbed motion have
first integrals
(C - A ) W , - MgzO> 0.
The problem of stability of a symmetrical gyroscope with an axisymmetrical cavity completely filled with an ideal liquid making a vortex
212
V. V. RUMYANTSEV
C>A>B.
Hence the permanent rotation of a solid body with a cavity filled completely
with a viscous or inviscid liquid about the small axis of the central ellipsoid
of inertia of the system is stable.
This result may be regarded as supplementary to Zhukovskiis theorem [7]
about the motion of a solid body with a cavity completely filled with a
viscous liquid.*
In a similar approach, paper [46] provides the solution of the stability
problem for a free solid body with a cavity completely filled with a viscous
liquid, if the body is attracted by a fixed centre according to Newtons law.
w1 = w2 = w3 = 0,
G, = G,
= G, = 0,
y1 = yz = 0,
y3 = 1
if the centre of gravity of the system is on the z axis. Solution (2.31) means
equilibrium of the body and liquid, whose free surface is the plane z = zl.
Let us analyze the stability of the equilibrium (2.31) in relation to
wi,G,,y, setting in the perturbed motion y3 = 1 7 and retaining the
previous notations for the rest of variables. The equations of the perturbed
motion have the integral relationship and the first integral
213
I/, G
(2.32)
+ T , + M,gzloq - U , - M,gz,O)
+ + q 2 + 2q = 0.
2(T,
v, = y12
= const,
y22
where K denotes the largest of the principal moments of inertia of the liquid
a t anyone instant [28]. It can also be easily seen [4Y] that the potential
energy of the liquid in the perturbed motion
where - Uz* denotes the potential energy of the liquid when it has the
plane xyl
yy,
zy3 = z1 as free surface. It is easily calculated that
where A,,B, are the principal central moments of inertia of the segment Q
of plane z = z1 limited by the walls of the cavity.
Let us now consider the function
Hi = 2 T 1 +
(GI2
(2.33)
It clearly follows from the above that
v, 2 H,,
(2.34)
!?'
=H,
(2.35)
- MgzooI/, = 2 T 1 + - (G,2
K
- (MZOO
+ G,2 + G,2)
y<v
where
q) = v1- Mgz0OV,.
'*
*3
214
V. V. RUMYANTSEV
If the inequality
(2.36)
MzoO+A*p<O
(A*>B*)
where 1; is the potential energy of the forces acting on the system and h
an integration constant.
Let us determine the position of the solid body relative to the inertial
frame O,&[ by Lagrangian co-ordinates q,.. . .,qn(n 6). The potential
energy of the system I. will, generally speaking, be a function of the coordinates q,,. . .,q, (r n) and of the shape of the liquid inside the cavity.
We also assume that the constraints permit rotation of the whole system
as a single solid body around some fixed straight line, and that the forces
acting on the system have no moment about that line. In this case the
<
<
215
Gc = const
(3.2)
v=n+oxr.
The kinetic energy and the projection on axis ( of the moment of momentum
of the system may be represented as
(3.3)
T = 2 + n)Gc'
+ -21 0 2 S ,
Gc = Gcl
+ oS.
Here
denote the kinetic energy and the projection on the axis of the kinetic
moment of the system in its motion relative to frame 0161q15;
s = 2 m 1 ( t u 2+ q u 2 )
I'
is the moment of inertia of the system about asis 5. (In these formulae the
summation refers to all mass points of the system.) The angular velocity LO
of the frame O l ~ l q l [ may be taken arbitrarily; let us choose it in such a
way that the projection on axis 5 of the moment of relative momentum of
the system, Gcl, vanishes at anv time, which by (3.2) and (3.3) is equivalent to
(3.4)
WS L'k = const
(3.5)
(IJ
1291.
k2
X+--+V=h.
2s
216
V. V. RUMYANTSEV
With the above assumptions about the constraints and the acting forces,
the system under consideration can indeed carry out a wheel rotation,
that is a uniform rotation as a single solid body, about the fixed axis 5.
In this case the system will be in equilibrium relative to the frame Ol(,q,~,
rotating about axis [ with the angular velocity of uniform rotation ooof
the system. Formally, this follows from
(3.6)
5v
=-
0025v,
q v f f = - w02qv,
C = 0,
- wo26
=0
or as
(3.7)
6U=O
if the notation
1 ko2
w=--+v
2 s
where KO is the value of the constant k in the law of areas for the present
case of uniform wheel rotation. The change in this function on a virtual
displacement of the system is
217
By comparing 6W with
1
2
6U = - w026S - 6V
and using woS, = k,, we see that equation (3.7) is equivalent to
6W = 0.
(3.10)
Thus, in this case of steady motion of the system, expression (3.9) is stationary.
According to definition (3.9), W depends on those co-ordinates of the
body ql,. .,qn- on which S and V actually depend, as well as on the
shape of the liquid and on the magnitude of the constant k,.
The quantity k, can be regarded as a variable parameter; we may then
apply to our system the results of the general theory of equilibrium of
material systems with potential energy depending on a parameter [31!.
Condition (3.10) leads, as can be easily seen, to equations
(3.11)
for the free surface, if the liquid does not fill the cavity completely. Here
U,(t,q,C) denotes the force function of the body force per unit of mass
acting on the liquid, and the potential energy of the liquid is given by
= - pJrU&t.
For a fixed value of the parameter k,, equations (3.11) and (3.12) determine
the co-ordinates of the solid body and the shape of the free surface of the
liquid in this steady motion.
v,
2. A Stability Theorem
218
V. V. RUMYANTSEV
I = max (PP,) .
Introduce also V as a measure of the deviation of f from f,:V is the
volume of that part of f which is outside ,/ or, which amounts to the same,
the volume of that part of fo which is outside f.
With a separation of given value I, the deviation V will have some
maximum which may be written in the form 1#(Z):
max V = Z#(Z).
1= const
219
> 0.
I g A
But the minimum of the deviation V for the given value of separation always
equals zero [29j. For continuous motion of body and liquid, 1 and I7 will
evidently be continuous functions of time.
We are now ready for the following definition of the stability of motion
of our system when the cavity is not completely filled: Consider the perturbed
motion of the system following an initial perturbation. If the initial separation and the initial relative velocities of the particles of the liquid as well
as the initial desiations and relative velocities of the body can be selected
so small that 1qiI,Iq,'1,1221,
and I remain smaller than some arbitrarily small
preassigned bounds, either throughout the motion or a t least until the
deviation has itself become smaller than some arbitrarily small preassigned
de\+ation, then the motion of the system under consideration is stable;
otherwise it is unstable.
According to this definition, the following is a necessary and sufficient
condition for the stability of the unperturbed steady motion: To any small
given numbers L , and L, a positive number 1 can be so selected that for
initial values of generalized co-ordinates q,(')) and velocities q i ' ( 0 ) of the
body, of separation
of deviation VO)and of relative velocities of the
liquid U ( ~ ) , V ( ~ ) , Wwhich
(~),
satisfy conditions
(3.13)
at t
= to,
lq,(O)l, l q i f ( 0 ) l lI(O)l,
,
Iu(O)I,1v(O)1,jw(O)I all
for any t
(3.14)
> to, or
a t least until
v >El,
the inequalities
(3.15)
1 4 ~ ' I 1 1 q< L?
are satisfied. Here E denotes a positive number smaller than the minimum
of the function $(1) under condition lIj
L,, while 1 may be regarded as a
possible deviation of the liquid. I t is worth noting that the condition related
t o (3.14) should be omitted, if the cavity is completely filled.
\Ve shall also need the concept minimuni o i expression W . If u' represents
function IY(q,,. . . , q n - I , k o ) , we mean by minimum of this function (at
fixed k,) an isolated minimum with respect to those variables q l , . . . ,qn- ,,
on which it actually depends. If the liquid does not fill the cavity completely,
then we accept with Lyapunov 1281 the following definition of the minimum
of W : If for the steady motion under consideration (when qi = 0, I = 0,
<
220
V. V. RUMYANTSEV
< E,
d E,
V > 1
<
formed for all slightly perturbed states near the state of steady motion,
has an isolated minimum W,, then the unperturbed motion is stable.
Proof [29]. Let us impart to the points of the system some sufficiently
small initial deviations and velocities. If left to itself, our system will move
on further in conformity with the integral of energy (3.5)which we rewrite as
where again (0) denotes the initial perturbed value and k is the magnitude
of the constant of areas in the perturbed motion. Suppose A is an arbitrarily
small positive number which does not exceed the number L , and which
we shall assume a t any rate smaller than number E above. Let us designate
by IY, the least possible value which W can assume if the separation I or
one of the co-ordinates q, are absolutely equal to A , while the remaining
quantities and the deviation satisfy the conditions
lQll
d A,
111
<A ,
v 2 El.
w,> w,.
221
But in choosing /I) and 1qil sufficiently small, and B > EE, expression W can
be made to differ from Woas little as desired. Let us choose A so small that
IW -
(3.17)
L,
E(O)
W(O)<W,.
(3.18)
<A,
1E(O)j
<A,
V(O)> d(0).
Whatever the initial position and the shape of the liquid, the initial velocities
of the points of the system can be so chosen that the constant magnitudes
IQil < A ,
Ill B A
are fulfilled. Now choose R, which appears in the stability condition (3.13)
and determines the region of initial perturbations, such that with conditions
(3.13) satisfied inequality (3.19) will be fulfilled for all values of S under
conditions (3.20). With such a choice of initial conditions, we have
throughout the subsequent time of motion as long as (3.20) holds
(3.21)
2+
w< w,
by (3.16) and (3.19). Hence, W < W,, and this will hold at least until lqil
or Ill exceeds A .
But qi(O) and E(O) are smaller than A , and V(O)> .do),
and since qi, 1 and
change continuously in course of time, lqil and 111 cannot exceed A without
first becoming equal to A . And inequalities
lqil = A ,
111 = A ,
222
V. V . RUMYANTSEV
I t follows from inequality (3.%1),account taken of (3.17), that 171 < L,.
On this ground we conclude that
and none will cease to hold as long as the last of them is valid. The theorem
has thus been proved.
I t is worth noting that for a completely filled cavity the conditions for 1
and V become superfluous; under the conditions of the theorem the following
inequalities hold for any 12 t o :
/qal < L,,
(221
< L,.
Note 1. As outlined above, the liquid was considered ideal. But the
theorem remains valid also for a viscous liquid as in this case the only thing
required is to replace equation (3.16) by the inequality
223
and if the minimum of the expression Wjnf occurs for the equilibrium shape
of the liquid, then this shape is stable [28, 291.
Corollary. If in the equilibrium position of a solid body with a liquidfilled cavity -(KO = 0 ) , the potential energy of the system V has an isolated
minimum V,,, then the equilibrium is stable [48].
Note that this result can also be applied in the case of relative equilibrium
of the same system. To have something definite in mind, we assume that
our system is acted upon, in addition to conservative forces, by nonconservative forces which are the cause of a moment N directed along
axis C whose magnitude is such that throughout the motion the angular
velocity w of the body about 5 remains constant. In this case we have
instead of the integrals of energy (3.1) and areas (3.2) equations
d(T
+ V ) = Nwdt,
dGc = N
-
at
151 1
+ V - OCC= const.
Introducing again the mobile frame O1(lq,C and bearing in mind formulae
(3.3), we can rewrite this integral as
(3.22)
1+ V--dS=const.
2
w*=v - - w212 s
has an isolated minimum, this position is stable.
As noted above, condition (3.7) is equivalent to condition (3.10) if (3.4)
holds. Owing to this, we are able to compare the state of relative equilibrium
of our system at constant angular velocity w with that of steady motion
224
V. V. RUMYANTSEV
possessing an integral of areas (3.2). If for some position of relative equilibrium W , has a minimum, W likewise has a minimum for the corresponding
steady motion [51], as can be easily seen. Indeed, let W , be a minimum of
expression W,, i.e.
I -
: (f,
-ko2
:)
--- -
It+
v,>o,
lqil ,< H
where S is the moment of inertia of the system about axis 5 for given q,,
and the unknown shape of the free surface of the liquid is to be determined.
By carrying out the variations under the integral sign and taking into
account the equation of continuity and the boundary conditions for the
225
liquid, we find that (3.24) is possible if and only if the free surface of the
liquid appears in the form
(3.25)
For steady motion with q, = 0 equation (3.25) takes the form (3.12).
Under rather general assumptions regarding the smoothness of the
cavity walls, the Ievel of the liquid and the nature of the body forces acting
on the liquid it can be shown [50] that expression W has a minimum if for
fixed q, the free surface of the liquid is determined by equation (3.25). We
shall not give the proof of this statement.
To any given set of values q, belonging to region (3.23), the solid body
with liquid in its cavity can be made to correspond to a certain solid body,
which we shall call transformed and which consists of the given solid body
and the liquid with free surface (3.25) solidified. According to what has been
stated in the foregoing paragraph, W for the transformed body will have a
minimum compared with W-values that belong to any free surface permissible
for the liquid and sufficiently close to surface (3.25).
Theorem 3.3. For the existance of a minimum of W in the steady motion
of a solid body with a liquid in its cavity the existence of a minimum of W
at q = 0 for the transformed solid bodies in region (3.23) is necessary and
sufficient.
Proof. Suppose expression W for all possible transformed solid bodies
in region (3.23) has a minimum W,, with q, = 0 ; then W - W,> 0. For
a solid body with a liquid, the difference W - W , will be still more positive,
by the definition of the transformed body. This proves the sufficiency.
To prove necessity, suppose expression W for the solid body with a liquid
has a minimum at q, = 0. This means that for every possible set of q,, I ,
and V , satisfying 1q,1
H , )I1 H , 2 EZ, all the values of W - W , will
remain positive, becoming zero only for q, = 1 = = 0. Consequently,
this difference will also be positive for values of separation 111 < H that
characterize the transition from shape f o of the liquid in the unperturbed
motion to shapes f determined by equation (3.25) with any q, from region
(3.23). This in turn means that expression W has the minimum W , for
transformed bodies with q, = 0. The theorem has thus been proved.
Hence, the problem of the minimum of expression W is reduced to a
minimum problem for a function of a finite number of variables W for
a solid body with a liquid bounded by walls o1 of the cavity and the free
surface (3.25).
Let us find the change in magnitude W for the transformed solid body
when passing from a position corresponding to steady motion of the system
at q, = 0 to a perturbed position in region (3.23). This passage can be
<
<
226
V. V.
RUMYANTSEV
d W = AIW
(3.26)
+d2W
Neglecting third order terms in q,,
+ 7[2d1S . A 2 S + (A2S)] + . . .
w2
2SO
r$1(x>Y,z>qi)= c1
t l ,or
227
(3.30)
Neglecting second order terms we have
n-
(3.31)
4(x,Y,z,,qi) - c
=AC
+ w2
(x' + y2)dS + . . .
SO
-
Thus we obtain
(3.34)
228
V. V. RUMYANTSEV
w.
f$(x,y,z,O) = 3 d ( x 2
+ y2) - gz = c.
The potential energy of the system and its moment of inertia about axis [
in the perturbed position equal respectively
+(x,y,z,Yi) = - . w 2 [ x 2
2
(3.36)
- 2XYYlY2 - W
Y 1 +
Y Y 2 ) v1
- Y12 - Yz2I
______
- d Y l + YY2 + v1 - Y12 - Y 2 2 ) ,
Let us assume that region Q represents a circular ring with radii R,
and R, ( R , > R2). In this case (3.33) holds, and dc = d2S = 0. Then
pJr,d(XrJ),ZjQi)dt = h(y12
yf2)a, where
+ ... .
229
(3.37)
+ V 2 = b2
and coincides with the latter in the limit, if 2g/w2+O. Let us consider
this extreme case which occurs when the liquid is weightless.
Instead of (3.36) we have now
and
(%)o
=d
b,
(
,I
= - w2bz(cos87,
+ sin By2).
Supposing cylinder (3.37) intersects the cavity surface u1 along circles with
centres on the axis z a t -7 = h & d . In first approximation the condition
of volume preservation becomes now
hence
AC = A,S = 0.
Finally, we find
p $dt=zpb2w2
71
3h2
+ d2
(YI2
+Y a 2 P
230
V. V. RUMYANTSEV
The minimum condition for W thus reduces to the single inequality [48]
References
(Titles of Russian publications are translated)
1. STOKES,G. G., On some cases of fluid motion, Trans. Cambridge Phil. Sot. 8 (1849).
-.> STOKES,G. G., On the critical values of sums of periodic series, Trans. Cambridge
Phtl. SOC.8 (1849).
H. v., Uber Reibung tropfbarer Flussigkeiten, Sitzungsbevichte der
3. HELMHOLTZ,
K. Akademie der Wissenschaften zu Wien, 40 (1860).
4. LOBECK,G., Uber den EinfluO, welchen auf die Bewegung eines Pendels mit einem
kugelformigen Hohlraume eine in ihm enthaltene reibende Flussigkeit ausubt,
Journal f . reine u . a n t . Math, 7 7 (1873).
5. LAMB,H., Hydrodynamics, Cambridge, 1932.
F., Hydrodynamische Untersuchungen, Leipzig, 1883.
8. NEUMANN,
N. E., On the motion of a rigid body having cavities filled with a
7. ZHUKOVSKII,
homogeneous liquid, Collected Works, vol. 3, Moscow 1938.
8. KELVIN,LORD,Mathematical and Physical Papers, vol. 4, Cambridge, 1882.
A. G., On the general motion of a liquid ellipsoid, Proc. Camb. Phtl.
9. GREENHILL.
SOC.4 (1880).
10. SLUTSKII,F., De la rotation de la terre suppose6 fluide a son interieur, Bull. de la
Socikttt des Naturalistes de Moscou 0 (1895).
11. HOUGH,S., The oscillations of a rotating ellipsoidal shell containing fluid, Phil.
Trans. (A) 186 (1895).
12. PO IN CAR^, H., Sur la precession des corps d6formables. Bull. Asfronontique 27 (1910).
13. BASSET,A. B., On the steady motion and stability of liquid in an ellipsoidal vessel,
Quart. J . Math. 46 (1914).
G. E., The Heaving of Ships. Moscow, 1935.
14. PAVLENKO,
S. L., On the motion of a symmetrical top with a liquid-filled cavity,
15. SOBOLEV,
in Zb. Prikl. Mekh. Tekh. Fiz., No. 3, 1960.
N. G., On stability of the rotating motions of a solid body, whose cavity
16. CHETAEV,
is filled with an ideal liquid, Prikl. Mat. Mech. 21 (1957).
L. N., Oscillation of a liquid in a moving vessel, Izu. Akad. Nauk
17. SRETENSKII,
S S S R , Old. Tekh. Nauk. No. 10, 1951.
D. E., On the theory of the motion of a body, having cavities partial
18. OKHOTSIMSKII,
filled with liquid, Prikl. Mar. Mech. 20 (1958).
19. KREIN,S. G., and MOISEEV,N. N., About oscillations of a solid body containing
liquid with a free surface, Prikl. Mat. Mech. 21 (1957).
20. MOISEEV,N. N.. The problem of the motion of a solid body containing liquid
masses having a free surface, Math. Zb. 82 (1953).
G. S., About the motion of a solid body having a cavity partially
21. NARIMANOV,
filled with liquid, Prikl. Mat. Mech. 20 (1958).
S. V.. and TEMCHENKO,
M. E., About a method of experimental
22. MALASHENKO,
investigation of the stability of motion of a top, inside of which is a cavity filled
with liquid, in Zb. Frikl. Mekh. Tekh. Fiz. No. 3, 1960.
A. Yw.,and TEMCHENKO,
M. E., Small vibrations of the vertical axis
23. ISHLINSKII,
of a top, having a cavity completely filled with an ideal incompressible fluid,
i n Zb. Prikl. Mekh. Tekh. Fiz. No. 3, 1980.
231
24. STEWARTSON,
K., On the stability of a spinning top containing liquid, J . of Fluid
Mech. 6 , 577 (1959).
25. COOPER,R. M., Dynamics of liquids in moving containers, A H S Journal 80 (1960).
26. RUMYANTSEV,
V. V., Stability of motion of a solid body with liquid-filled cavities,
in Proceedings of the All-Union Congress on Theoretical and Applied Mechanics,
Moscow, 1962.
27. LYAPUNOV,
A. M.. General Problem of the Stability of Motion. Moscow, 1950.
Russian reprint of the original that appeared in 1892. A French translation,
dating from 1907, is available as ProblBme GCn6ral de la Stabilit6 du Mouvement by M. A. Liapounoff, republished as Number 17 of Annals of Mathematics Studies, Princeton University Press, 1949.
28. LYAPUNOV,
A. M., On the stability of ellipsoidal forms of equilibrium of rotating
liquids, Collected works, vol. 3, Moscow, 1959.
29. LYAPUNOV,
A. M., The problem of minimum in the question on stability of figures
of equilibrium of rotating fluid, Collected works, vol. 3, Moscow, 1959.
H., Sur 16quilibre dune masse fluide animee dun mouvement de rota30. POINCAR&.
tion, Acta Math. 7 (1885).
31. CHETAEV,N. G., The Stability of Motion. Moscow, 1955, English Translation
(same title) published by Pergamon Press 1961.
32. RUMYANTSEV,
V. V., Equations of the motion of a solid body, having cavities
partially filled with liquid, Prikl. M a t . Mech. 18 (1954).
V. V.. On the equations of the motion of a solid body with a liquid33. RUMYANTSEV,
filled cavity, Prikl. Mat. Mech. 19 (1955).
34. KOCHIN,N. E., KIBEL, I. A , , and ROSE, N. V., Theoretical Hydromechanics.
Moscow, 1955.
35. GORYACHEV,
D. N., Some general integrals in the problem on the motion of a
solid body. Warsaw, 1910.
36. BELETSKII,V. V., Some questions of the motion of a solid body in a Newtonian
force field, Prikl. M a t . Mech. 21 (1957).
37. CHETAEV,
N. G., On a property of Poincar6s equations. Prikl. Mat. Mech. 19 (1955).
38. RUMYANTSEV,
V. V., On the stability of permanent rotations of a solid body about
a fixed point, Prikl. M a t . Mech. 21 (1957).
39. RUMYANTSEV,
V. V., About the stability of the motion of gyrostats, Prikl. Mat.
Mech. 26 (1961).
40. RUMYANTSEV,
V. V., Stability of the rotation of a solid body, having an ellipsoidal
cavity filled with liquid, PrikI. M a f . Mech. 21 (1957).
41. ZHAK,S. V., About the stability of some particular cases of motion of a symmetrical
gyroscope, containing liquid masses, Prikl. M a t . Mech. 20 (1958).
42. RUMYANTSEV,
V. V., About stability of motion with respect to a part of the variables,
Vestn. Mosk. U. Ser. Mat. Mekh. Nr. 4, 1957.
43. RUMYANTSEV,
V. V.. About the stability of rotational motions of a solid body with
liquid content, Prikl. Mat. Mech. 23 (1959).
44. RUMYANTSEV,
V. V., A theorem on the stability of motion, Prikl. Mat. Mech.
24 (1960).
45. RUMYANTSEV,
V. V., About the stability of the motion of a top having a cavity
filled with a viscous fluid, Prikl. Mat. Mech. 24 (1960).
46. KOLESNIKOV,
N. N., About the stability of a free solid body having a cavity filled
with a n incompressible viscous fluid, Prikl. Mat. Mech. 08 (1962).
47. RUMYANTSEV, V. V., About the stability of equilibrium of a solid body, having
cavities filled with a liquid, Dokl. A . N . S S S R 124 (1959).
48. RUMYANTSEV,
V. V., About the stability of steady motions of solid bodies with
liquid-filled cavities, Prikl. Mat. Mech. 26 (1962).
232
V. V. RUMYANTSEV
49. POZHARITSKII,
G. K., The problem of the minimum in the problem of stability of
equilibrium of a solid body having a cavity partialy filled with liquid, PrikZ.
M a t . Mech. 06 (1962).
50. POZHARITSKII,
G. K., and RUMYANTSEV,
V. V., The problem of the minimum in
the question of stability of the motion of a solid body with a liquid, PrikZ. Mat.
Mech. 27 (1963).
51. APPELL,P., Figures dbquilibre dune masse liquide homoghe en rotation, Trait6 de
Mecanique Rationnelle, vol. 4. Paris, 1932.
A brief introduction to Lyapunovs main idea can be found in LaSalle, J., and Lefschetz, S., Stability by Liapunovs Direct Method. Academic Press, 1961.
BY N. N. MOISEEV
Computing Centre of the U . S . S . R . Academy of Sciences, Moscow, U . S . S . R .
Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
I. Survey of Special Problems . . . . . . . . . . . . . . . . . . . . . .
1. Oscillations of a Heavy Liquid Enclosed in a Vessel a t Rest . . . .
2. The Stokes-Zhukovskii Problem . . . . . . . . . . . . . . . . . .
3. The Pendulum Problem . . . . . . . . . . . . . . . . . . . . . .
4. Oscillations of a Conservative System with a Liquid Member . . . .
5 . Torsional Oscillations of a Beam with a Liquid-Containing Cavity . .
6. Torsional-Flexural Oscillations of a Beam with a Liquid-Containing Cavity
11. General Properties of the Equations . . . . . . . . . . . . . . . . .
1. Oscillations of a Liquid in a Vessel . . . . . . . . . . . . . . . .
2. Oscillations of a Conservative System with a Liquid Member . . . .
3. Oscillations of a Beam with a Liquid-Containing Cavity . . . . . . .
4. Investigation of Static Stability . . . . . . . . . . . . . . . . . .
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Selected Bibliography . . . . . . . . . . . . . . . . . . . . . . . . .
Page
233
235
. 230
. 243
246
. 252
. 258
266
. 268
. 268
. 270
. 273
. 280
287
288
INTRODUCTION
The dynamics of a body containing a liquid I as a history that goes back
to the classical contributions of nineteenth-century scientists, but only in
the last two decades was the case taken up when the liquid enclosed within
the body has a free surface on which waves may appear. Almost simultaneously, and independently of one another, several scores of papers, devoted
to this problem, were published in various countries. This interest was
evoked by a great number of technical problems calling for a theory of
motion of a body with a cavity containing liquid under conditions when the
cavity is partly filled. These include, above all, the problem of seismic
oscillations of structures under water pressure and of fuel reservoirs, the
dynamics of rockets, the calculation of an airplane wing, taking into account
the fuel which it may contain, etc.
The considerable efforts made by scientists and engineers have produced
the result that a t present our knowledge about oscillations of a body with
a liquid represents a rather comprehensive theory branching out in several
directions.
It is not possible to expound within this article all the aspects of this
theory, and the author is faced with the difficult problem of selecting the
233
234
N . N . MOISEEV
235
The article does not deal with the two directions in which intensive
investigations are being currently conducted. These are non-linear oscillations and the problem of damping. A study of these questions involves
difficulties of a fundamental nature. A great many algorithms pertaining
to the theory of non-linear oscillations have been published, but all of them
are still very clumsy and, most important, no one has so far succeeded in
proving their convergence. Moreover, the very question of the existence of
periodic solutions of resulting non-linear systems still remains open. Still
more complicated is the problem of oscillations of a viscous liquid. The
very formulation of the problems comprises here a great number of difficulties.
The problems of the dynamics of a body with a liquid when the latter
is under conditions of weightlessness have become pertinent most recently,
but only the first steps have so far been made in this direction and it is still
premature to speak of results.
The article supplies references to those papers only which may clarify
the essence of the material in question and facilitate the reading of the
article. A selected bibliography is given in addition.
1. SURVEY
OF SPECIAL
PROBLEMS
236
N . N. MOISEEV
6L = 0
(1.1)
L = .(T-L!)dl.
0
FIG.1
v=
vcp;
2, condition of continuity
v v = 0;
(1.3)
if
v,=O
PEZ;
4. kinematic condition
(1.51
Ct = u,
if
P ES.
The second and the third conditions are determined by the physics of
the problem, and the fourth is a corollary of the second for small amplitudes.
The condition of irrotationality, as will be seen below, may be disregarded
in the linear theory.
237
,.,.
pt
+ gc
= 0,
p E s.
(1.7) is the known condition of constancy of pressure on the free surface [3].
By using (1.7), we can represent L in the following way:
t
is harmonic in
T;
9 satisfies conditions
238
N . N . MOISEEV
91 = H / .
H is the integral operator
gc
+ Herr
= 0.
(1.11)
3. We study here the question of natural oscillations of the liquid. Let
us assume, therefore, that
y ( P , t ) = cos at@(P),
(1.12)
239
or
H#.#dS -
(1.14)
#'dS,
s (A@)2dt
(1.15)
S@2dS
'
J" (VQ2dt
1, = min *
J@,dS
in the class of functions orthogonal to @, if
is a function solving the
variational problem (1.15),etc. Orthogonality refers here to the metric
defined in L,, the functions themselves being defined in S.
5 . To solve the variational problem (1.1)for the functional (1.13)it is
advantageous to apply Ritz' method. The standard procedure reduces
to the following: a system of coordinate functions { x , ( P ) } is introduced,
and the solution of the problem is sought as a sum
iv
Then, proceeding from (1.13) and (l.ll), we arrive at the following system
of algebraic equations
N
(1.16)
2 am(%,,,- ilp,,,,)
9U-l
= 0,
n = 1,4,.. ., N ,
240
N. N . MOISEEV
where
with
anm = am%,
Pnm
Pmn.
lanrn
-
=0
which we shall call the equation of frequencies, We shall designate its zeroes
by I n ( n = 1,2,. . . N ) , Owing to the symmetry of matrices awmand prim,
the eigenvalues In are real. Once they are known, the natural frequencies
are determined by
(1.18)
aw2=
In&
but slightly.
11. The boundary conditions for QI belong to the category of natural
conditions, and therefore it need not be required that the functions xn should
strictly satisfy all boundary conditions. According to the general theory,
the minimizing sequence constructed by Ritz method will nevertheless
converge to the exact solution.
may be selected rather
111. Hence, the system of coordinate functions
roughly. I t suffices to provide only for the completeness of the system.
It is therefore advisable to select the
as eigenfunctions of some volume
which contains the given volume but has a simpler shape. For instance, if
the liquid oscillates inside a conical tank, we may take as system of coordinate
functions the eigenfunctions of the liquid in a cylindrical vessel whose cross
section equals the largest cross section of the cone.
24 1
6. Let us now consider the variational problem (1.15) for two volumes
equal free surface S , but with surface Zlenveloping surface Z2
(see Fig. 2). Hence t l > t2,and for any function @
tland t2with
(1.19)
@(2)
FIG. 2
Since
A(')
A'2'
<
@)
by
J (V@(l')2dT
J- @ ( 1)2dS
S
+I
<
Thus, if we have two vessels with the same area of free surface but such
that Zlof the first vessel envelops Z2 of the second vessel, then the corresponding natural frequencies will be greater in the vessel whose volume
is larger.
242
N. N. MOISEEV
FIG. 3
(1.21)
(1.22)
v,= V O ,
Ap,=O,
VU=O,
PET;
(1.23)
(1.24)
by
j-
(v1,v2) = v1 v,dt.
243
(u,vo)=
pnr&
z+s
(u,vo) = 0.
We shall express this result as follows:
vo = n0v
where I7, is the operator of orthogonal projection from E on E,.
Vector v satisfies Eulers equation
(126)
Vf
nowe
obtain
P
!%
+no(v.v)v
=-v
at
P
- gzo.
ut
(1.27)
+ I7,(V,B)V
= 0.
For the infinitesimal motions under consideration, the linear terms of (1.26)
and (1.27) produce
(2
-+p++z
)= o
or
3
+ p + gz = const,
at
111
= 0.
The last equation has a simple physical sense: the vortical component
of the velocity vector remains constant at every point of volume t if second
order terms are neglected.
The pressure and the shape of the free surface is determined with the
same degree of accuracy by the irrotational component alone which can be
found in this approximation independently of the vortical component as a
solution of the problem under investigation in this section. Conversely, the
omission of condition (1.2), i.e. of irrotationality, will not affect the firstorder accuracy of the solution.
2. The Stokes-Zhukovskii Problem
1. Before proceeding to a liquid with free surface, we have to consider
the classical problem of the motion of a body with a cavity completely filled
by an ideal incompressible liquid.
244
N. N. MOISEEV
Let the liquid occupy volume t (but the plane S is no longer a free
boundary) and assume in addition that volume t is moving and that this
motion is known to us. This means that vector v,, of its instantaneous
translational motion and vector w of its instantaneous angular velocity
are known. The velocity potential of the absolute motion of the liquid should
satisfy condition
_
aP an - v,,
P E Zf
where
v, = von0
(12 8 )
+ (ox r)no,
+
no is the unit vector of the normal to the surface Z S and r is the radiusvector of the points of this surface. Let us denote by voi,wi,uz,xi(i = 1 , 2 , 3 )
the projections of the vectors vo,w,noand r on the axes of the system x1,xz,x3,
rigidly connected to the moving volume t. We seek the harmonic function p,
the potential of the absolute motion of the liquid, in its dependence on the
coordinates of the mobile system.
As has already been noted by Stokes, the solution of problem (1.28)
should be sought as a linear function of scalars zloi and w,,
3
p=
voipi*
i=1
+ 2 wipi*+s.
1
=1
Functions pi* are determined only by the geometry of region t and do not
depend on its motion. They are harmonic and satisfy the following boundary
conditions :
--
--
an - az z h2,
an
-a%* - u3xz - Q X ,
an
(1.29)
= h,,
_
a%*_
-azxl - ulxz
an
= h,.
an
an
=h
31
= h,,
- UIX3 - U3Xl
-a%*
--
_
-- h i ,
an
PEZ+S
t,twice
245
for
i=l,21...16,
*; +
u;.
-Au;=A#;=f,
PET;
(1.30)
#i:
where
Since C does not depend on the unknown function, the solution of the
variational problem for functional F is reduced to solving the variational
problem for functional
I;* = (Vp;*)'dt S
[p;*h;- $;h;]dS,
+z
S +z
246
but
N. N. MOISEEV
s t,bihidS
s +r
disregarded.
Hence, the problem of determining the Stokes-Zhukovskii potentials is
equivalent to finding the minimum of functional
r
J
S
(1.31)
7
FIG.4
by a rigid weightless rod. Let us place the fixed system of coordinates (OZY)
a t the point of suspension and connect the mobile system (ozy) to the free
surface in the state of rest, as shown in the sketch. To analyze the plane
oscillations of the pendulum requires, generally speaking, some additional
assumptions regarding the shape of the cavity.
We begin with constructing an expression for the kinetic energy of the
system body liquid
T = T , + T,;
(1.32)
T,is the kinetic energy of the shell, Jo is its moment of inertia about the
hinge 0 ; T , is the kinetic energy of the oscillating liquid, p is its density,
and v is the velocity vector of the liquid on the assumption that the free
surface has been replaced by a solid wall:
247
v = wvV*
where ~ 1 is
* the Stokes-Zhukovskii potential and is determined by the shape
of the vessel only, and tp is the potential of wave motion in the vessel. Thus
T=
82
( V ~ ) ~ d8p
t VVVq*dt
+ pe2
(Vp*)2dt
T
The integral in the last term depends on the geometry of the cavity only.
For this reason we shall call the quantity
an adjoint mass (in this case an adjoint moment of inertia) of the liquid.
On writing J = J 0 + m, (1.32) becomes
(1.33)
T
We consider this problem in the linear sense and therefore should retain
only the quadratic terms in the expression for the kinetic energy. Consequently, we should carry out the integration in (1.33) over the volume
which the liquid occupies in the equilibrium position.
Let us now construct in a similar way an expression for the potential
energv of the system
where
the pendulum,
Mo is the mass of the pendulum shell, and I, is the distance from its centre
of inertia to point 0. In this expression the integration extends over the
volume occupied by the liquid,
248
N. N. MOISEEV
and plane S. The first item is the potential energy possessed by the liquid
if the free surface is replaced by a solid lid. With pg f Z d r = pgZ*, where
7.
Z = (z-L)cosO+ ysin8
where 1 = 00. We have
[ ( z - 2 ) cos 8
s
7,
+ y sin 8]dz
C2dS
+ pg8
CydS
+ O ( P );
s CdS = 0.
n=-2 + Bpg
u2e2
CydS
+p
C2dS.
Here
u2 = loM&
+ pgt$*.
Note: Subsequently we shall need yet another expression for the kinetic
energy. Applying Green's formula to (1.33) and taking into account that
aP
-=
an
Sr,
P ES,
we find that
(1.35)
L=
(T-Wdt
249
while T and 17 are taken as in (1.33) and (1.34). By using Greens formula
and the kinematic relationship (1.5), we obtain
(1.36)
(1.37)
250
N. N. MOISEEV
0 = X sin at,
(1.39)
cp = @ cos at,
5 = $sin at,
X ( K~ ]a2)
+p
* ( g y - cp*a2)dS= 0,
(1.40)
+ pg*
X p ( g y - cp*u2)
- p@H* = 0.
$ = - XR,(gy
- q*a2).
g+ - 02H# = f
as
where
FIG. ?.I
In our case
251
252
N. N. MOISEEV
Equation (1.41) can most simply be studied graphically (see Fig. 5). The
unknown zeroes are the points of intersection of the straight line Fl = ic2 - Ja2
with the curve
v = vp,+v*
It has been shown in Section 2 that
3
(1.42)
v* =
2 v,vp,,* + 2 o,vq:+3
i=l
I=
253
and assume the expression for the kinetic energy of the system as follows:
Here
m$7. .- mij
0
+p
Vpi*Vqi*dt,
and {mt.} is the matrix of the coefficients of the quadratic form representing
the kinetic energy of the solid body. Using Zhukovskiis terminology, we
shall call mii the masses of the equivalent solid body.*
Let us now consider the small oscillations of a conservative system K
which has n degrees of freedom, and let 2, (z = 1,2,. . . , n) be the generalized
coordinates of this system. Supposing there is a solid body with a liquid
cavity among the members of this oscillatory system. Then, without loss of
generality, the kinetic energy of system K can be described by (1.43) in
which the summation now extends from 1 to n.
2. We now compute the potential energy 17 of system K. If the free
surface is covered by a lid (in this case we shall denote it by KO),then
(1.44)
n = no= 2 ,rbj7ZjZj.
1
i,j = 1
If we consider only the case when the equilibrium of system KOis stable,
154
N. N . MOISEEV
IZ, = 22,
s f,CdS
(1.45)
lI= 5
2 biiZiZj +
i=l
ij=1
no
T -0-2
2 Y,I2;
?I
IZ0
2 pi2Yi2.
i=l
i=l
,uKare the natural frequencies of the system when the free surface is covered
with a lid.
In the new variables we have
vi*
and f,.
By eliminating cp and
(1.47)
4. Applying Hamiltons principle and repeating the considerations used
in the construction of the pendulum equations, we obtain the following
system of equations for the oscillations of system K :
255
System (1.36),
which determines the oscillations of a pendulum, was a
particular case of the system of integro-differential equations (1.48).
References [7]and [ll] bring the analysis of a number of specific problems
whose motion is reduced to the study of system (1.48).
5 . Let us now consider the problem of free oscillations. Suppose
Y i= X,sin at,
(1.49)
= @ cos at,
5 = (G sin at.
Construct
= f ( T - Z7)dt,
and (1.47),
and integrate from 0 to 2n/a. Omitting a non-essential multiplier,
we obtain
(1.50)
s=1
Y (a2A,, - &)a,
(ae- p L 2 ) X +
, A
(1.51)
n
i = I$,.
. .,n;
2 ( a z ~-, ~B,,)x, + C
r=l
Here
= 0,
s=1
1-1
- As)a,= 0,
(a20(Is
s = 1,2. . .N .
256
N. N. MOISEEV
It can be easily seen that ais = asi by applying Green's formula to the
integral.
(1.51) is a system of homogeneous linear equations for Xl,. ... X,,
al,.... aN. T o have non-trivial solutions, a2 must be a root of
dN(a2)= 0
where A N is the determinant
...
...
...
...
...
...
...
...
...
...
...
...
...
...
The elements symmetrical to the main diagonal are equal, thus the roots
of A(a2) = 0 are real.
We have noted in the preceding section [after (1.40)] that the problem
is simplified if the normalized eigenfunctions of the free-oscillation problem
are taken as
since in this case
xu,
Hxn = 7
xn,
O n
(Xn.Xm)
n#m
=m'
+ 2 (a2Ai, - Bis)as= 0,
i = 1 , 2 . . .n ;
s=l
U
(1.52)
s = 1,2.. N ;
i =1
as = pglws2,
P, = pg.
257
(a2
- pi2)Xi + &
(a2Ais - ~ i s ) =
a ~
0,
s=1
(153)
n
= 0.
i =1
k =O
k =O
a2=
ake&.
k =O
For the zero-order quantities Xi,,, aso and no2 we have the system
XiO(UO2- /Liz) = 0,
(15 4 )
aso(ao2as- Ps) = 0.
System (1.54) has a solution if uo2= pi2 or ao2= Ps/as. Since we wish
to determine the frequencies of system K close to those of system KO,we
must assume that a, = pi. For definiteness suppose a, = ,ul ; then
X, = X, = . . X , = a, = 0. We may assume without loss of generality
that X , = 1.
For the first-order quantities Xi, and asl we have the equations
i # 1,
aI2= 0.
258
N. N. MOISEEV
X i , = O ( i > I),
XI, and
a,,
= 0.
Bl,
- AlSPl2 - -.
pl2aS - P s
(1.55)
X&02
To make this equation solvable, it is necessary and sufficient that the right
member of (1.56) be equal to zero. Hence,
(1 55)
259
and subject the function u ( y ) to boundary conditions that make the operators
in question self-adjoint. For instance, it may be assumed that u(0) = u'(Z)= 0
(the beam is rigidly clamped a t one end). We shall call the boundary value
i
FIG. 8
Iu,u,dy
= 0,
n f m;
(15 6 )
[KU,,'U,'
+ Bu,,u,]dV
#m
0,
)a
p,~,
n = m'
260
N. N. MOISEEV
1
where
This implies that the natural frequencies of the beam are equal to those of
the associated boundary value problem.
Note: Investigation of the eigenfunctions referred to above is particularly convenient if the parameters of the beam are independent of y .
If they change slowly, it is advisable to use the asymptotic solution of
the associated problem,
(1.57)
0
261
Here q~ is the potential of the absolute motion of the liquid, and t the volume
occupied by the liquid.
Let us assume
0 = cos wt6(y),
9 = - o sin odrP(x,y,z),
and construct the Lagrangian:
aP
an 5 21,
(0
= (thyo x r)no.
no = a x 0 + Pyo + yzo,
r = xx0
+ yyo + zzo,
we obtain
where
f(x,y,z) = za
- xy.
a@
-= 6f.
an
6=
2
i-1
262
N . N. MOISEEV
t, determined
as solutions of the
-a@u -u,f
an
or
on = HH,(P)f(P),
PE
2.
am [w2(anm
+ prim) -
rum1 = 0.
bum = p
V@,V@,dt.
For the same problem let us construct yet another equation of oscillations, using Hamilton's principle for this purpose
Carrying out
s~ = {ze,se, - Ke,se,
0 0
- seseyyat
jrvm,f. v ~ ~ e ~ f a ~ a t
OT
by partial integration of the first term (remembering that the variations are
isochronous and vanish at y = 0 and y = I) and by applying Green's formula
to the second term, we obtain
L I
263
a
-Hu
an
= 'u,
0 YI l g
Here the minimum and maximum values of y on Z are y1 and yz. We can
always integrate over y from 0 to 1, since I , = 0 if y < y1 or y > yz.
By collecting terms and using the arbitrariness of variations, we arrive
at the following equation for the torsional oscillations of a beam containing
a cavity filled completely with a liquid:
( 15 9 )
mt+ p l f w 6 t t d l . - (KO,), + Be = 0.
'Y
* Cf.
264
N. N. MOISEEV
for a beam without liquid, but if there is liquid inside the beam, the acceleration in this cross section also depends on the angular accelerations of all the
other cross sections. This effect arises from the term
W f e t P ,= p j fW,)
'Y
lv
H(py,p)f(p)e,(P,t)dPdP,
fi
FIG.7
where q$') is the velocity potential which the liquid would have had if
surface S had not been free
~ ( l=) Hv,(P),
PEL'+ S,
265
where 5' is the elevation of the free surface above the unperturbed level in
the absolute system of coordinates. Hence
q ' 2 ' = H*(5't - v,) = H*ut
T*=L
[VHV,
2p
+ PH*utl2dt.
where tois the volume occupied by the liquid in equilibrium, and tlis the
volume bounded by the free surface and the plane S. The second term can
be written as
are only determined by the elastic displacements and have the same structure
as the potential energy of the external forces. Therefore we can include
them in the potential energy of the external forces without loss of generality.
Thus we write the kinetic and potential energy of the system beam liquid
in the following way
266
N. N. MOISEEV
[ K e y 2 B02]dy
0
+
5
The integrals in (1.61) and (1.62) depending on 6, are written in the system
of coordinates oxlylzl.
Accordingly, the equations of motion will appear as
18;;
+p
JY
(1.63)
+p
fHfettdb
dY
fH*%tdd, - (KO,),
+ Be +p g
+ p ~ * u t+t p g f e + pgu
pHettf
f d d , = 0,
dY
where d , is the segment which lies at the intersection of the region S and
the plane of the normal cross section whose coordinate equals y . Thus the
oscillations of a beam with a liquid are determined by a system of two
integro-differential equations.
Note. Ritz' method may again be recommended for calculating the
principal oscillations and natural frequencies. If the frequencies of the free
oscillations of the liquid in a fixed vessel and of the torsional oscillations of
a beam are not close to each other, the spectrum of the combined problem
(beam + liquid) will be given approximately by the superposition of the
spectra of both problems. We have to select two systems of coordinate
functions. In this case it is advisable to choose the system of eigenfunctions
of the operator H as one of the systems, and the system of eigenfunctions
of the associated boundary value problem as the other system. To calculate
the spectrum of the combined problem one may advantageously employ the
methods of the perturbation theory outlined in the preceding section.
6. Torsional-Flexural Oscillations of a Beam with a Liquid-Containing Cavity
1. Let us denote the deflection in the plane xoy by X , and in the plane
yoz by letter 2. Then the kinetic and potential energy of the beam (without
a liquid) is given by the following formulae well-known from the elements
of strength of materials (see e.g. [ I S ] ) :
I
267
INTRODUCTION TO T H E T H E O R Y OF OSCILLATIONS
Here the C, are the flexural and torsional stiffnesses, the B,, characterize
the external conservative forces (B,?= B J , the A,, are coefficients characterizing the distribution of masses or moments of inertia (in the preceding
section I @ ) ) ,and the A,, characterize the asymmetry of the beam, i.e. the
non-coincidence of the shear centre and the centroid of the cross section.
If there is a liquid-containing cavity inside the beam and the liquid has
a free surface, the kinetic and potential energy of the system are
To
+ ?!2-
(C7HylXt
( I .65)
ZZ =
+2
u2dS
+ pg
u[Xy,
+ Zy2 + 8y3Jds.
Here
y1 = cos (nx),
y2 = cos (nz),
268
N . N . MOISEEV
+ B,,Z + ~,,e + p g
Y3udl = 0,
dv
+
+ pH*utt + pgy&
+ PgY38 + Pg* = 0
pHXtty1+ pHZtty2
+ PgY,z
pmty3
11. GENERALPROPERTIES
OF THE EQUATIONS
This chapter is devoted to a study of the general properties of the equations derived in the first chapter. The structure of the spectrum, the
existence of natural oscillations, and the possibility of employing Ritz
method are elucidated for the case of conservative external forces. In t h e
last section the basic theorem of this theory is proved: for the stability of
the equilibrium of the system body
liquid it is necessary and sufficient
that some other, considerably simpler system should be stable. The body
may be either rigid or elastic (but again subject to the hypothesis of
Bernoullis beam t heorey).
AH# - # = 0,
To find the solutions, Ritz method was recommended. But certain
questions remained open. They concern the solvability of this problem,
the structure of the spectrum, the completeness of the system of eigenfunctions, and the convergence of Ritz method with N
00.
---+
269
(2.2)
ss
Here H(P,Q)is Green's function of Neumann's problem for region t. Accordingly the inner integral
and satisfying
Consequently,
(Hf,!) = ( V u ) 2 d t .
1
This proves the assertion, since the right member is zero if and only if u G 0 ,
0.
or, which is the same, if f
The physical meaning of this statement is quite evident. The scalar
product (2.2) determines the kinetic energy of an oscillating liquid if the
velocity of the points of the free surface is determined by f(Q) sin at. But
it is well-known that the kinetic energy of liquid in irrotational motion in a
simply-connected vessel at rest can vanish if and only if f = 0.
270
N. N. MOISEEV
(c) The eigenfunctions $, of operator If, which describe the principal forms
of free oscillations are such that sequence 1&,. .. $,. .. is complete in E .
(d) The eigenvalues and eigenfunctions may be determined by Ritz' method.
This theorem covers the essence of all the theoretical questions about
a liquid in free oscillations of infinitely small amplitude.
2. Oscillations of a Conservative System with a Liquid Member
1. Here we study certain general properties of the system of integrodifferential equations (1.48) which are the subject of analysis in 1131. We
shall follow the outline given in that article. First of all let us put system
(1.48) in operator form. To do this we introduce the n-dimensional space
E, of vectors Y with components Y , (i = 1, 2 , . ... n) and the space E of
functions C ( P E S),square-integrable in S. Let us define the scalar products.
n
2 Y,Z,
(Y,z),=
in E,,
,=1
(C,d =
C(P)V(P)dP
in E.
+ (pp).
= (YCl),YCZ)),
L, =
... 0
0 1 ... 0
.........
.........
1 0
p1
...
pz
...
0 0
............
............
0 0 ... P*
...
M, =
27 1
(b) acting in E :
L,,
=H ,
M,,= p g ;
Ll*Y = P(qJ**,Yo),
M,, = ( Y ' Y ) ,
where we denote by q** and v the vectors with projections qi** and v,,
respectively (i = 1, 2,. . ., n) ;
(d) acting from E to E,:
M,,C
L,,C = y*,
= v*
(I
{
. .p
y* = p q,**CtdS,.
S
S I
q7n**tdS ,
v* = [v,CdS.
.. . jVnCdS,1
+ Lo,tt + M,Y +
L,OY'' + L,& + M,,Y + M,,t
L,Y"
M0,C = 0,
= 0.
This can be expressed still more briefly if the operators L and M acting
in d are introduced:
(2.3)
Lx"
+ M X= 0
(Lx,x)= (x,Lx),
( M x , x=
) (x,Mx).
272
N. N. MOISEEV
( M - u2L)q = 0 .
Thus the quantities A
M-lL:
cq = Aq.
(2.5)
(Mx19x2).
(MM-Lx1,x2)= ( L x , x ) = (x1,Lx2)
= (XI,MM-Lx,) =
(Mx,,Cx,)
= (x1,Cx,)F.
In addition, ( C X , , X =
~ ) (~L x , , ~ ,>
) 0, if x1 f 0. Hence, proceeding from
the general theorems of analysis, it follows (see [2]) that:
(a) for operator C in space E there exists a complete system of eigenelements
orthogonal in Friedrichs metric
(b) all un2> 0 and on2 do with ~t-+ 00
(c) the eigenelements q,, and quantities u, can be determined by Ritz method
which converges.
273
[ = Zeid ;
% -
u~,u;
E Ei,
i = 1,2, 3 ;
274
N. N. MOISEEV
(x,Y)=
(.iVi)i.
i=l
Lipi = A+,
+p
yiHyiujdZ,
i,j = 1 , 2 , 3 ;
I,
j = 1,2,3;
Lpju, = pHiyjuj,
Lj4u4= p ylH*u4dl,
j = 1,2,3;
I?.
L,u,
=pH*q;
M33u3 = - C3U3Y)Y -I
B33U3,
M . . u .- B..u.
,I I ,
i,j = 1 , 2 , 3 ,
M4pj = pgyiu,,
j = 1 , 2 , 3;
Mj4u4 = p g y,u4dl,
M44u4 = pgu4;
i#j;
j = 1,2,3.
d?.
275
In matrix notation
= llLijlll
llMiill
Lx*t
(2.7)
+ M x = 0.
( L X , Y ) = (X,LY).
(2.8)
i,i = 1
i,j = I
= (ulL14v4)1.
The rest of the relationships proving the validity of (2.9) are established in a
similar way.
Unlike operator L of the preceding section (see (2.3)),the present operator
L is not completely continuous. The difference is caused by the fact that,
previously, the space Eo was finite-dimensional; but now the part of E ,
is played by E ,
E,
E, which is infinite-dimensional.
+ +
276
N. N . MOISEEV
(2.10)
( M x , x ) = 2z7.
It appears that only those problems are of practical interest, in which the
equilibrium position is statically stable, i.e. I? in the position of equilibrium
is a minimum. We shall therefore consider operator M as positive definite.
Hence, the difference between equations (2.7) and (2.3) consists in the
fact that operator L in (2.7) is not completely continuous. This prevents
us from applying directly the general theorems about linear operators.
Note. If operator M - l had been completely continuous, the above
difficulty would have been of no significance, since operator M - l L owing
to the boundedness of L would have been completely continuous.
4. Instead of (1.66), let us construct a system of equations equivalent
to it. For this purpose we calculate the variation of the Lagrangian
I
(2.11)
lc
- y l x - y2z -
y3e.
If we now set up the equations for the variables X , Z, 8, and v , and omit the
clumsy calculations (for more details see [19]), we arrive a t the following
system of integro-differential equations :
+ ~ +~~~~e~~
z +
~ (clxyy)vv
~
+ B;X + B : ~ Z
+ B:,e +
y1[ H l x I I y l+ H
~ + HlettY3
z ~ + ~~ * v ~~ ~=1~0,d s
A,,X~~A
lv-dv
+~A x~ +~~~~~e~~
z~ +~ (czz,,),,
~
+ ~ 2 :+x ~2*zz
+
+ B,*,e + Y z [ H I x I t y +l HIZttYz+ HletIY3+ H*vttids
1 ,--J
3
= 0,
277
~ 3 1 ~ t t
Y3[H1Xttyl
1Y- d Y
p [HIXttyl $- H1ztty,
H1eltY3
H*vlt
+ gvl
= O*
In these equation H , = H - H*, Bii* = B,i - pli, and p 11. = p"7% are
functions of variable y , depending only on the configuration of volume z
(the expressions are given in greater detail in the next section).
The main difference between the transformed system (2.12) and (1.66)
is that its first three equations do not include terms containing (they include
only derivatives of v with respect to t ) .
If we introduce the space E,' = El E, E3 and let w E E,' where w
is the vector with projections X,Y,O; then (2.12) may be written as
+ +
(2.13)
4 1 '
(2.14)
43'
Mll'
(2.15)
42'
M'
L32'
M12'
'33'
Ml3'
M32'
Y1
r = Y2
Y3
M33'
278
and
N. N. MOISEEV
L'
A=l/PHII'.
= E,'
+ E,
by
N = l / Il l
M'
PH*ilJ
Pg
LIZ,
(2.16)
+ NZ= 0.
(z,Az,) = (x,Lx,),
and for this purpose it suffices, in turn, to spell out the corresponding scalar
products and to make in the resulting expressions the substitutions
+ H*,
= Hl
v=u
+ y , x + 722 + y3e.
M ' s =f.
(2.17)
e m = ] G ~ P . Q ) ~ ( w Q= c,f.
0
279
(2.19)
w=
w cos ot,
v=
v cos ot.
(2.20)
Now substitute
M- 1/2LM- 112
R=
1
MI- ll2D VPT
P H,rM-
__
VPg
1/2
P H*
VPg
1
~
VPg
is completely continuous.
The symmetry of operator R is established by a simple verification.
Hence the problem is reduced to the eigenvalue problem
280
N . N . MOISEEV
= AX.
( M - l L h , h )=
~ (Lh,h)
where the scalar product to the right is the kinetic energy. Consequently,
unless h z 0, this expression cannot equal zero.
4. Investigation of Static Stability
281
+ 2B12XZ + Bz2Z2
The substitution (2.11), which we already used in the preceding section, will
again be applied, but the calculations will now be made in greater detail.
In the new variables, I7 appears as
where
IT* = { C I X t , + CzZ;,
2
1
(2.22)
y2y3ZBdS -
- pg yly2XZdS - pg yly3XOdS
y12X2dS- - pg yz2Z2dS- pg y 3 V S .
2 5
S
This expression can be given a more telling form that clarifies the meaning
of a number of terms.
As y1 = cos (nx) and yz = cos (nz) are constant on S (the normal to S
is always vertical) and as y3 = zyl - xy2, we can write
282
Here 1
, and J , denote the static moments of the interval d , ( J , = Jxdl,. . . ),
dY
while Jz,, J,,,JlX are its moments of inertia. Thus the quantities p , , = pi,
depend only on the geometry of the cavitv.
If we introduce the designations
R?
- B..
- plj,
I1 I1
functional Z7* will appear as
1
(2.23)
Theorem 1. For the system beam liquid to be statically stable, i.e. for
functional 17 to be positive definite, it is necessary and sufficient that functional 17* should be positive definite.
This theorem has a simple mechanical significance. For static stability
of a beam with a liquid it is necessary and sufficient that some beam without
liquid, possessing the same elastic characteristics but under the action of
other external forces, should be stable.
It has been shown above that there exists no beam without liquid which
would be dynamically equivalent to a beam with liquid. For the stability,
283
t=v-
1
-2ViYi,
Pg
n=II1+n*
where
But
284
N. N. MOISEEV
where J s is the moment of inertia of the free surface. Thus, the condition
of static stability may be represented as
(2.24)
K'
> pgJ.s
I,
MI1
-pJs
where I , is the given length of the pendulum whose free surface is covered
by a lid, and M is the mass of the whole system. Hence, the presence of
the liquid can only weaken the stability of the pendulum.
5. Analysis of the regularity of functional
in the general case is not
trivial, and there exist no criteria which would not only be sufficient but also
necessary, We shall only point here to a few conditions that are sufficient
for static stability.
Functional (2.23) can be written as:
n*
If the coefficients C, are everywhere positive (a case that is typical for many
applications), then only U2*must be positive definte. Therefore Sylvester's
inequalities should hold :
> 0.
Conditions (2.25) impose limitations on the geometrical characteristics and
can be expressed more directly. In the case of torsional oscillations of a
beam, (2.25) reduces to p:> 0 or
(2.27)
= 0,
B33 - P g J x x
285
B3, > 0.
In other words, it is sufficient for the stability of the beam that the external
forces should be of the nature of a restoring force. In this case the liquid
does not affect a t all the stability of the beam.
6. Let us consider now the bending oscillations of the beam in two mutually
perpendicular directions. IT,* will appear as
1
d
Conditions (2.25) yield two inequalities:
(2.28)
286
N. N. MOISEEV
n*= 2
(2.30)
I
I
C3By2dy
1
2 BS+S02dy.
0
BG2 -4
Y E
[W],
FIG.I)
where
287
LI* that
This condition imposes a lower bound on C,. I t will be satisfied if, for instance,
function C, is larger than some constant q, which depends on the external
forces and the geometry of cavity.
References
(Titles of Russian publications are translated)
1. MOISEEV,N. N., About two liquid-filled pendulums, P M M 16 (1952).
2. MIKHLIN,
S. G., Variational methods in mathematical physics. GIFML, Moscow,
1957. A German translation is available : S. G. Michlin, Variationsmethoden
der mathematischen Physik, Akademie-Verlag, Berlin, 1962 ; an English
288
N . N . WOISEEV
14. MIKHLIN,S. G., The minimum problem for a quadratic functional. GIFML,
Moscow, 1951.
15. RIESZ, F., et SZ.-NAGY.B., Leqons danalyse fonctionelle, 3d ed., Acad. Sci.
de Hongrie, Budapest, 1955.
16. SRETENSKII.
L. N., Theory of wave motions of a liquid. GITTL, Moscow, 1936.
17. PONTRYAGIN,
L. S., Hermitian operators in a space with indefinite metric, Izv.
A N S S S R , ser. mat. 8 (1944).
18. TIMOSHENKO,
S., Strength of Materials. Van Nostrand, New York, 1930.
19. MOISEEV,N . N., Variational problems in the theory of oscillations of a liquid and
of a body with liquid, i n Variational methods in problems of oscillations of
a liquid and of a body with liquid. Vychisl. Tsentr A N S S S R , Moscow, 1962.
Selected Bibliography
1 . ABRAMSON,H. N., CHUWEN-HWA,RANSLEBEN,
G. E., J R . , Representation of fuel
sloshing in cylindrical tanks by a n equivalent mechanical model, A R S Jouvnal
11, 1697-1705 (1961).
2. BUBLIK,B. N., and MERKULOV,
V. 1.. Dynamic stability of thin elastic shells,
reinforced by rigid ribs and filled with liquid. i n 29. 119-179.
3. COOPER,R. M.. Dynamics of liquids in moving containers, A R S Journal 80, 725-729
( 1960).
4. KIRILLOV,V. V., investigation of the oscillations of a liquid in a n immovable
container with consideration of drainage. T r . Mosk. fiz.-tekhn. instituta, 1960,
Nr. 5, 62-72.
5 . LAWRENCE,
H. K.,WANG,C. J . and HEDDY,H . R., Variational solution of fuel
sloshing modes, Jet Propulsion 28. 729-736 (1958).
ti. MILES,J. W., On the sloshing of liquid in a flexible tank, J . A p p l . Mech. 26, 277-283
(1958).
7. MILES, J . W., Note on the damping of free-surface oscillations due t o drainage,
J . Flrtrd Mech. 12. 438-440 (1962).
8 . MOISEEV,N. N., The problem of small oscillations of a n open container with liquid
under the action of a n elastic force, Llhuazn. rnatern. zh. 4, 168-173 (1952).
9. MOISEEV. N. N., Dynamics of a ship with liquid loads, Izv. A N S S S R . Old. tekhn. n.
1954, Nr. 7.
10. RIoIsEEv, N. N., On a problem in the theory of waves on the surface of a limited
volume of liquid, P M M 19, 342-347 (1955).
11. MOISEEV,N. N., On the boundary-value problems for the linearized Navier-Stokes
equations in the case of small viscosity, Z h V M i M F 1 (1961).
17. MOISEEV, N . N., On mathematical methods for the investigation of nonlinear
oscillations of liquids, in Proceedings of the international symposium on nonlinear
oscillations, Kiev, 1961.
13. MOISEEV,N. N., Some mathematical questions concerning the motion of a satellite.
Lecture given in the international symposium on the dynamics of satellites,
Paris, 1962.
14. NARIMANOV,
G. S.. On the motion of a rigid body whose cavity is filled with liquid,
P M M 20, 21-38 (1956).
15. NARIMANOV,
G. S., On the oscillations of liquid in movable cavities, I z v . A N S S S H ,
Otd. tekhn. n. 1957, Nr. 10, 71-74.
16. NARIMANOV,
G. S., On the motion of a vessel partially filled with liquid, the motion
of the latter not being considered as small, P M M 21, 513-524 (1957).
17. PENNEY,
W. G. and PRICE,A. T., Some gravity-wave problems in the motion of
perfect liquid, Philos. Trans. Roy. SOC.London A244, Nr. 882. 254 (1952).
289
PM llil
,4 N
= Akad. nauk
Zh LM z ,VIF
GIFICfL
GI T T L
=
=
Author Index
Numben in parentheses are reference numbers and are included to assist in locating references in which authors'
names are not mentioned in the text. Numbers in italics refer to pages on which the references are listed.
B
Barratt, M. J., 119, 120, 121, 122(40),
123(40), 126(40), 129(40), 150(40),
151(40), 171(40), 172(40), 179
Basset, A. B., 184, 230
Batchelor, G. K., 149(71), 166(71), 187
Bazer, J., 38(55, 56), 67
Beletskii, V. V., 191(36), 231
Benninghoff, J. M., 139(57), 180
Blokhintsev, D. I., 118(37), 179
Boltzmann, L., 90, 101
Boynton, F. P., 136(94),'782
Broch, J. T., 140(66), 180
Brown, D. R., 85, 101
C
Cabannes, H., 38(58), 40, 67
Carrier, G. F., 18, 65
Chapman, S., 54(67), 67
Chester, W., 12(17, 18), 65
Chetaev, N. G., 184, 185(31), 190,
191(16, 37), 192(37), 194, 195(16, 31),
196(16), 207(31), 208(31), 217(31),
224(31), 230, 231
Chu, C. K., 35, 67
Clarkson, B. L., 105(90), 181
Clauser, F. H.,11, 65
Coleman, B. D., 70(2), 73(6), 76(10),
77(10, l l ) , 78(10), 79(11, 15, 17),
81(10, l l ) , 82(10. 17), 83(10), 87(23),
88(23), 89(5), W 2 , 51, 91(2, 5 ) , 92(6),
96(15, 30), 99(10), 100. 701
Contursi, G., 68(72), 68
Cooper, R. M., 184(25), 231
D
Davies, P. 0. A. L., 119, 120, 121, 122(40),
123(40), 126(40), 129(40), 150(40),
151(40), 171(40), 172(40), 179
De Witt, T. W., 79(14), 81(19), 93, 701
Dumas, R., 150(72), 181
Dyer, I., 141(67), 181
E
Eisenberg, H., 79(13), 100
Elsasser, W. M., 8(2), 64
Elyash, L. J., 79(14). 101
Ericksen, J. L., 70, 71, 74, 100
Ericson. W. B., 38(55, 56), 67
Etkin, B., 105(20), 179
F
Favre, A., 150(72), 181
Ferry, J. D.. 95, 96(29), 701
Feshbach, H., 162(84). 781
Fisher, M. J., 119, 120, 121, 122(40),
123(40), 126(40), 129(40). 150(40),
151(40), 171(40), 172(40), 179
Fishman, F. J., 64(71), 68
Ford, G. W., 118(35), 179
Fowell, L. R., 105(16), 178
Franken, P. A., 105(22), 159(22), 179
Franz, C. J., 148(70), 181
Frei, E. H., 79(13), 700
Friedrichs, K. O., 30, 31, 66
291
292
AUTHOR INDEX
H
Hardcastle, D., 138(50), 780
Hasimoto, H., 11, 13, 65
Heitkotter, R. H., 177, 187
Helfer, H. L., 38(53), 67
Helmholtz, H. v., 184, 230
Herlofsen, N., 30(48), 66
Hough, S., 184, 190(11), 203(11), 230
Howes. W. L., 133(51), 130(51), 780
Hubbard, H. H., 129, 135, 130(47),
137(47), 138(47), 171(87), 780, 787
I
Ishlinskii, A. Yu., 184, 230
J
Johnson, W. R., 162(85), 787
K
Kantrowitz, A., 54(66), 67
Kantrowitz, A. R., 04(71), 68
Kelvin, Lord, 184, 230
Kibel, I. A., 197(34), 205(34), 237, 237(3),
287
180, 187
M
McCune, J. E.. ll(12, 13), 29(45), 35,
36(45), 43(13). 65, 66
Maczynski, J. F. J., 157(75), 781
Malashenko, S. V., 184, 230
Markovitz, H., 79(14), 85(21), 86(2l), 91,
707
293
AUTHOR INDEX
N
Napolitano, I,. G., 68(72), 68
Narasimha, R., 119(41), 175(41), 179
Narimanov, G. S., 184, 230
Neumann, F., 184, 230
Noll, W., 69, 70(2), 73(6), 74, 76(10),
77(10, 1l ) ,78(10),79(11, 17),81 (10, 11).
82(10, 17), 83(10), 85(20), 86(20),
87(23), 88(23), 89(5), 90(2, 5), 91(2, 5).
92(5), 96(31), 99(10), 100, 107
0
Okhotsimskii, D. E., 184, 230, 249(9), 287
P
Padden, F. J . , 81(19), 101
Padden, F. J., Jr., 79(14), 101
Pai, Shih-I., 68(73), 68, 81(18), 93, 107
Pavlenko, G. E., 184, 230
Petschek, H. E., 64(71), 68
Phillips, 0. M., 105(5), 127(5), 136,
175(5), 176(5). 177, 778
Pietrasanta, A. C., 130, 181
Poincark, H., 184, 185, 197, 198, 230, 231
Polovin, R. V., 40(60). 67
Pontryagin, L. S., 288
Powell. A , , 105(6, 15, 19, 24, 26, 27).
106(2), 116, 124(6, 21), 125(6), 129(15),
135, 159(15, 76). 162(76), 163(15).
171(6), 175(6, 26). 178, 179, 781
Pozharitskii, G. K., 213(49), 214(49, 50),
225(50), 227(49), 232
R
Ram, G. S., 105(29), 179
Rayleigh, Lord, 128(44), 780
Reiner, M., 74, 100
Resler, E. L., 54(66), 67
Resler, E. L., Jr., lO(4). 11(4), 29(45), 35,
36(45),43(13), 56(68),62(70),64, 66.67
Ribner, H. S., 105(8, 9, 10, 11. 20, 28, 29),
117(9, 10, 11). 118(9, 10, l l ) , 119(11,
38). 124(8), 127(11), 128(45), 129(9,
10, 11). 130(11), 132(11), 137(10, 11).
138(1l ) , 142(1l ) , 144(1l ) , 145(11).
147(9, 10, l l ) , 148(11), 151(11),
153(10, 11). 154(11), 155(11), 156(10,
I l ) , 157(4, 74). 159(9, 10, 11). 160(11),
161(77), 162(45), 163(9, 10, l l ) ,
171(8), 175(10, 11). 176(9, 10, 11). 177,
178, 179, 181
Richards, E. J., 105(14, 17), 140(63), 178,
180
Riesz, F., 288
Rivlin, R. S., 70, 71(4), 74, 79(16), 82(16),
85(16), 100
Rollin, V. G., 137(54). 180
Rose, N. V., 197(34), 205(34), 237
Rossow, V. J., 12, 18(21, 22), 6 5
Roze, N . V., 237(3), 287
Rumyantsev, V. V., 184(26), 185(47),
186(32, 33), 194(38), 195(38, 39),
196(40), 197, 198(33), 203(42, 43, 44),
206(32), 208/44), 210(45), 212(46),
214(32, 50), 223(48), 225(50), 230(48),
231. 232
S
Sanders, N. D., 163(86), 181
Sarason, L., lO(8). 64
Savic, P., 116(34), 179
Sears, W. R., l0(4), 11(4, 12, 13), l6(25),
18(27), 25(41), 27(42), 31(49), 40(63,
64), 64, 65, 66, 67
Seebass, A. R., 27(44), 66
294
AUTHOR INDEX
T
Tamada, KO.,12(20), 22, 65, 66
Tamagno, J., 162(82), 181
Taniuti, T., 27(43), 40(61), 66, 67
Temchenko, M. E., 184122). 230
Timoschenko, S., 266(18), 288
Toupin, R. A., 74(7), 100
Townsend, A. A., 120(43), 121, 123(43),
180
U
Uberoi, M. S., 136(53), 137(53), 780
V
Van de Hulst, H. C., 30, 66
von Gierke, H. E., 105(18), 178
W
Weitzmer, H., 10(8), 33(50). 64, 67
Westervelt, P. J., 105, 106(25), 179
Westley, R., 139, 180
Williams, J . E., 105(12, 13), 127, 128,
135(13), 152(13), 154(13), 156(13).
159, 178
Withington, H. W., 139(61), 140(81), 180
z
Zhak, S. V., 203(41), 231
Zhukovskii, N. E., 184, 187, 18817).
189(7), 190. 192(7), 195(7), 197,
199(7), 212, 230, 244(10), 287
Subject Index
A
B
Beam with liquid-containing cavity, 273
Boltzmann's theory of linear viscoelasticity. 75, 90
Boundary conditions (MHD). 5
Boundary layer, inviscid, in MHD. 16 f.,
42 f.
F
Faraday's law, 7
Fast waves, 31
Fading memory, 70, 73, 74
Flexural-torsional oscillations, 258
Flow over a n airfoil (MHD), 29
over a plate, 21
295
296
S U R J E C T INDEX
H
Hall effect, 5. 50, 54, 55, .i8
Hamiltons principle, 234, 236, 254, 262,
267
Hasimotos theory, 14
flow, 16
Heavy liquid in a vessel, 235
Helical flow, 79, 82
Helicoidal motion, 191
High-density plasma, 58
History function, 73
Histories with small norm, 90
Hyperbolic regimes, 26
Hyperbolic-supersonic, 27
Hypercritical, 27
I
Ideal conductor, 12, 25
Inconipressible simple fluid, 69
Incompressible flow (MHD), 6
Infinitesimal strain tensor, 90
Influence function, 73
Intermediate waves, 33
Inviscid perfect conductor, 9, 15
Ion slip, 53
Irrotational motion (stab.), 186
in linearized problems, 242
Isochronism of the variations, 249
Isolated minimum, 219
Isotropic turbulence, 149 f., 163, 166
Isotropic functional, 72
L
Leading-edge stagnation-point flow
(MHD), 18
Lighthill parameter (turb.), 124
Linear theory (stab.), 185
h r e n t z force, 50
M
Mach number redefined (MHD), 26
Mach wave, 127, 176
emission, 134
Magnetic lines frozen, 57
Magnetic poles, absence of, 7
Magnetic Prandtl number (def.), 8
Magnetic Reynolds number (def.). 8
Magnetosonic Mach cone, 37
waves, 31
Maievskiis condition, 195
Material function, 78, 87, 89
Material objectivity, principle of, 72
Memory functional, 74, 76
Mixing region, 119 ff., 123, 125, 169 ff., 174
Monopole, 109
Moving-axes (turb.), 152 If., 166
Moving jets, 157 f.
N
Keumanns operator, 237
Newtonian fluid, 70, 74, 76, 78
Norm, 73
Normal stress, 81, 95, ion
effects, 71. 75, 78
measurements, 86
Normal stress coefficients, 92
Not completely continuous operator, 176
0
Ohms law for ioized gases, 51, 52, 5!b
Oscillations of a liquid in a vessel, 268
P
Particle motions (MHD), 50
Pedal curve, 31
Pendulum problem (osc.), 246. 283
Perfect fluid (MHD). 74
Plasma frequency, 57
Plasma, collisionless, 56
cold, 62 f.
Plasma, composition of, 52
Poiseuille flow, 79
297
SUBJECT INDEX
Q
Quadrupole, 112, 143, 149, 153, 17.7
sources, 115
R
Kate of shear, 77
Real history, 74
Refraction (turb.). 126, I28 f., 131, 159,
168
S
Scale of turbulence, 1.70, 154
Scale anisotropy, 154
Second-order fluid, 70, 76
Self-noise, 125 f., 133, 149 f., 164, 166 ff.
Separation (stab.), 218
Shear compliance, 96
Shear-dependent viscosity, 75
Shear loss modulus, 94
relaxation modulus, 71, 91 f.
storage modulus, 94
Shear-noise, 125 f., 133, 164, 168 ff.
Shearing flow, steady simple, 77
nonsteady simple, 96
Shock waves (MHD), 38
Simple fluid, 69, 72 f., 78. 81, 85 f., 87
problem, 243
Stokes effect, 113
Storage modulus, 7 1
Strain measure, 89
Stress source (turb.), 112
Stress relaxation, 75, 90
Sub-AlfvCnic, 13, 47
Sublayer, viscous (MHD), 1 6 f.
Substantially stagnant motions, 96
Super-Alfvbnic, 13
Surface-current layers, 6
Surface presslire (MHD), 18, 28 f.
T
Temperature effects (turb.), 136
Tensor conductivity, 52, 54
Torsional flow, 84
298
SUBJECT INDEX
U
Uniform rotation (stab.), 192
Uniform vortex motion (stab.), 196
Units in MHD, 6
Upstream wake (MHD), 14
V
Viscoelasticity, linear, 71, 89, 93
second order, 71, 89
Viscometric flow, 79
Viscosity effects (MHD), 11
Viscosity function, 78
Vortex layer (MHD), 6, 15, 28, 36
Vortex stretching, 116
W
Waves, damping of (MHD), 45
Waves, forward-facing, (MHD), 49
Wave-propagation (MHD), 30
Wave-speed diagram, 62 f.
generalized, 58
Wake, inviscid, (MHD), 20
z
Zhukovskii's theorem, 190