Chapter 6

Equation of motion and vorticity transport

6.1 Newton's second law for the motion of a parcel

6.2 Integral momentum balance
6.3 Cauchy's equation of motion
6.4 Euler's and Bernoulli's equations
6.5 The Navier-Stokes equation
6.6 Vorticity transport
6.7 Dynamic similitude, the Reynolds number
and dimensionless numbers in fluid dynamics

Fluid flow is established in response to an external action mediated

by boundary motion, by the application of a surface force, or by the pres-
ence of a body force. The evolution of a transient flow and the structure
of a steady flow established after an initial start-up period are governed
by two fundamental principles of thermodynamics and classical mechan-
ics: mass conservation, and Newton's second law for the motion of fluid
parcels. The implementation of Newton's law in fluid mechanics leads
to Cauchy's equation of motion, which provides us with an expression
for the point particle acceleration in terms of the stresses, and to the
vorticity transport equation governing the point particle rate of rota-
tion. The derivation and intepretation of these equations in general and
specific forms, and their solution for simple flow configurations will be
the theme of our discussion.

6.1 Newton's second law

for the motion of a parcel
Consider a fluid parcel in motion, as illustrated in figure 5.1.1 (a).
Newton's second law of motion requires that the rate of change of the
linear momentum of the parcel, denoted by Mp, be equal to the sum of
the forces exerted on the parcel, including the body force due to gravity
given in equation (5.1.1), and the surface force given in equation (5.1.2).
Expressing these forces in terms of surface and volume integrals, we
obtain
(6.1.1)
where f is the hydrodynamic traction exerted on the parcel surface. Ex-
pressing further the traction in terms of the stress tensor, as shown in
equation (4.2.9), we find

(6.1.2)
where the unit normal vector n points into the parcel exterior. Our next
task is to relate the rate of change of the parcel momentum to the fluid
density and velocity.

6.1.1 Rate of change of linear momentum

An expression for the linear momentum arises by subdividing the
parcel into elementary subparcels of volume dVp and mass dmp = pdV,
and then summing the contributions by integration, obtaining

(6.1.3)
where u is the fluid velocity. The rate of change of linear momentum is
given by

(6.1.4)
Now, because the integration is over the volume of the parcel which is
not stationary but changes in time, interchanging the time differentiation
and the volume integration on the right-hand side of (6.1.4) is permissible
only if the time derivative is replaced by the material derivative D/Dt
under the integral sign. Thus,

(6.1.5)
Conservation of mass requires that the material derivative of the elemen-
tary mass drrip vanish, yielding the simplified expression
 (6.1.6)

6.1.2 Equation of parcel motion

Substituting the right-hand side of (6.1.6) into the left-hand side of
(6.1.2), we obtain the desired equation of fluid parcel motion,

(6.1.7)
involving the point particle acceleration, the stress tensor, and the body
force. Explicitly, the x, y, and z components of (6.1.7) are given by

(6.1.8)
Equations (6.1.8) are valid irrespective of whether the fluid is compress-
ible or incompressible.

6.1.3 Two-dimensional flow

The counterpart of equation (6.1.7) for two-dimensional flow in the
xy plane is

(6.1.9)
Figure 6.1.1 Illustration of a fluid parcel with an instantaneous rect-
angular shape, drawn with the solid line, in two-dimensional flow.
The fact that the parcel generally deforms to obtain a warped
shape, drawn with the dashed line, does not prevent us from ap-
plying Newton's law of motion in its integral form over the instan-
taneous parcel shape.

where dA is the differential of the surface area, and dl is the differential

of the arc length along the contour of a parcel in the xy plane. Explicitly,
the x and y components of (6.1.9) are given by

(6.1.10)
We note again that these equations are valid irrespective of whether the
fluid is compressible or incompressible.

Motion of a rectangular pared

As an application, we consider the motion of a fluid parcel with an

instantaneous rectangular shape whose sides are parallel to the x or y
axes, as depicted in figure 6.1.1. It is important to note that the parcel
will remain rectangular only if the fluid exhibits rigid-body motion; under
more general conditions, the parcel will deform to obtain a warped shape
drawn with the dashed line in figure 6.1.1. The fact that the parcel
will generally deform, however, does not prevent us from evaluating the
integrals in (6.1.10) over the instantaneous rectangular shape.
For simplicity, we assume that the density of the fluid is uniform, and
the acceleration of gravity is constant. Taking into consideration that
the unit normal vector is parallel to the x or y axis over each side, we
find that equations (6.1.10) assume the simpler forms

(6.1.11)
and

(6.1.12)
The first integral on the right-hand side of (6.1.11) involves normal
stresses exerted on the vertical sides, and the second integral involves
shear stresses exerted on the horizontal sides; the converse is true for
(6.1.12).

Steady unidirectional flow

In the case of steady unidirectional flow along the x axis, the point
particles move along the x axis with constant velocity and vanishing
acceleration, Du/Dt = O. The left-hand side of the equation of motion
(6.1.9) vanishes, yielding a balance between the hydrodynamic and body
force,

(6.1.13)
Restricting our attention to Newtonian fluids and using the consti-
tutive equation (4.5.4), we find the following:
• In the absence of axial and transverse stretching, dux/dx = O and
duy/dy = O, the normal stresses axx and ayy are equal to the
negative of the pressure p, axx = <jyy = —p.
 • The shear stresses axy — cryx are independent of streamwise posi-
tion x, but may depend on the lateral position y.
Subject to these simplifications, the balance equations (6.1.11) and
(6.1.12) reduce to

(6.1.14)
where Ax = x% - x\ and Ay = j/2 — J/i- The second of equations (6.1.14)
is clearly satisfied when

(6.1.15)
reflecting the hydrostatic pressure variation. The first of equations (6.1.14)
is satisfied when

(6.1.16)
where G is a free parameter called the modified pressure gradient. Phys-
ically, G is determined by the physical mechanism driving the flow:
1. When G = O, the first of equations (6.1.16) shows that the pres-
sure variation in the x direction is hydrostatic, and the second
of equations (6.1.16) shows that the shear stress ayx is constant,
independent of y. This is the case of shear-driven flow.

2. When the streamwise pressure drop vanishes, px=x2,y = PX=XI,y? the

first of equations (6.1.16) requires that G = pgx, and the second
of equations (6.1.16) shows that the difference in the shear stress
(&yx)x,y=y2 ~ (cryx)x,y=y1 is equal to -p gx Ay. This is the case of
gravity-driven flow.
3. When the flow is horizontal, gx = O, the first of equations (6.1.16)
shows that G is the negative of the streamwise pressure gradient,
and the second of equation (6.1.16) shows that the difference in
the shear stress (&yx)x,y=y2 ~ (&yx)x,y=yi is equal to -G Ay. This
is the case of pressure-driven flow.
 Problem

Problem 6.1.1 Body force in terms of a surface integral.

Show that the body force expressed by the second integral on the
right-hand side of (6.1.7) may be expressed as a surface integral in the
form
(6.1.17)
Hint: Use the Gauss divergence theorem (2.6.25).

6.2 Integral momentum balance

Consider the integrand of the rate of the change of momentum on
the left-hand side of equation (6.1.7). Using the rules of product differ-
entiation and the continuity equation (2.8.5), we write

(6.2.1)
where V-u = dux/dx+duy/dy+duz/dz is the divergence of the velocity.
If the fluid is incompressible, the second term on the right-hand side of
(6.2.1) is absent.
The x component of the vectorial expression (6.2.1) may be manip-
ulated to give

(6.2.2)

where the time derivative d/dt is taken keeping the spatial position fixed.
Combining the last four terms on the right-hand side of (6.2.2), we find
 (6.2.3)
Working in a similar fashion with the y and z components of (6.2.1),
we derive the corresponding expressions

(6.2.4)
and

(6.2.5)

6.2.1 Momentum tensor

To collect equations (6.2.3) - (6.2.5) into a unified form, we introduce

the momentum tensor M^, defined as

MiJ = PUiUj, (6.2.6)

where the indices i and j range over x, y, and z or, correspondingly, 1,

2, and 3. It is evident from the definition (6.2.6) that the tensor M is
symmetric, Mij = M^.
Moreover, we introduce the divergence of the momentum tensor de-
fined as a vector whose ith component is given by

(6.2.7)
where summation of the repeated index j is implied. For example, the
x component of the divergence of M is given by

(6.2.8)
Subject to these definitions, equations (6.2.3) - (6.2.5) are expressed
by the collective form

(6.2.9)
where i = £,y, or z. The corresponding vector form is

(6.2.10)

The right-hand sides of equations (6.2.9) and (6.2.10) involve Eulerian

derivatives, that is, derivatives with respect to time and spatial coordi-
nates.

6.2.2 Integral momentum balance

Substituting now (6.2.10) into the left-hand side of the equation of

parcel motion (6.1.7), we derive the alternative form

(6.2.11)

To this end, we use the Gauss diverence theorem stated in equation

(2.6.25) to convert the volume integral of the divergence of the momen-
tum tensor into a surface integral over the parcel volume, obtaining

(6.2.12)

where the unit normal vector n points outward from the parcel. In index
notation,

(6.2.13)
where summation is implied over the repeated index j.

6.2.3 Control volumes

It is important to remember that equation (6.2.13) has originated
from Newton's second law for the motion of a fluid parcel. In the process
of expressing the material derivative in terms of Eulerian derivatives
taken with respect to time and position in space, however, the parcel
has lost its significance as a material body of fluid, and has become
relevant only insofar as to define the volume that it occupies in space.
To signify the new interpretation, we rewrite (6.2.13) in identical for-
m, except that the domain of integration is now called a control volume,
denoted by CV. Using the definition of the momentum tensor shown in
(6.2.6), we express the integral momentum balance in the form

(6.2.14)
The four integrals on the left- and right-hand side of (6.2.14) have the
following interpretation:
1. The first integral is the rate of change of the iih component of mo-
mentum of the fluid residing inside the control volume. At steady
state, this term vanishes.
2. The scalar UjUj of the second integrand on the left-hand side of
(6.2.14) is the component of the fluid velocity normal to the bound-
ary of the control volume; accordingly, the corresponding integral
expresses the rate of convective transport of the ith component of
the fluid momentum across the boundary of the control volume.
3. The first integral on the right-hand side of (6.2.14) is the ith com-
ponent of the surface force exerted on the boundary of the control
volume.
4. The second integral on the right-hand side of (6.2.14) is the ith
component of the body force exerted on the control volume.
We proceed now to discuss a particular application illustrating the
usefulness of the integral momentum balance in engineering analysis.

1
6.2.4 Flow through a sudden enlargement
Consider steady flow in a duct through the sudden enlargement, as
illustrated in figure 6.2.1, and identify the control volume with the sec-
tion of the duct confined between the vertical planes labelled 1 and 2.
1
ThIs example, and the one discussed in problem 6.2.1 was borrowed from the
pioneering text by Bird, Stewart, and Lightfoot cited in the bibliography, where a
collection of engineering problems are solved by use of integral mass, momentum, and
energy balances.
 PIane l
Plane 2

Figure 6.2.1 Simplified model of flow in a duct through a sudden en-

largement. An integral momentum balance allows us to compute
the pressure rise P^ — PI as a function of the inlet and outlet cross-
sectional areas 5i and 52.

Assuming that the density of the fluid is uniform and the velocity profile
is flat at the inlet and outlet, neglecting the shear stress at the walls,
approximating the normal stress at the inlet and outlet with the negative
of the pressure, assuming that the pressure at the washer-shaped area is
equal to the inlet pressure, and neglecting the effects of gravity, we find
that the integral momentum balance (6.2.14) at steady state simplifies
to
(6.2.15)
where Si and $2 are the cross-sectional areas of the inlet and outlet.
The three terms on the right-hand side of (6.2.15) are approximations to
the first integral on the right-hand side of (6.2.14) for the outlet, washer-
shaped area, and inlet. Mass conservation requires U\S\ = U^S^ solving
this equation for C/i, and substituting the result into (6.2.15), we find an
expression for the pressure drop

(6.2.16)
which predicts a rise in pressure in agreement with laboratory observa-
tion.
 Plane 1 Plane 2

Figure 6.2.2 Schematic illustration of an ejector pump. The pressure

rise P<2~Pi may be estimated by performing an integral momentum
balance.

Problem

Problem 6.2.1 Operation of an ejector pump.

Figure 6.2.2 shows a schematic illustration of an ejector pump. At
plane 1, two fluid streams merge; one with uniform velocity Ui over a
cross-sectional area Si, and the second with uniform velocity UQ over a
cross sectional area SQ. At plane 2, the velocity profile is uniform over
the cross-sectional area S? = SQ + Si. The pressure is assumed to be
uniform over the cross-section of the inlet and outlet, respectively, equal
to PI and P2, and the fluid density is assumed to be uniform throughout
the flow.
Derive an expression for the rise in pressure P^ — PI in terms of
p, UQ, Ui, SQ, and Si, similar to that shown in equation (6.2.16).

6.3 Cauchy's equation of motion

Equation (6.1.7) contains two volume integrals and one surface inte-
gral over the boundary of a fluid parcel. If we managed to convert the
surface integral into a volume integral, we would be able to collect all
integrands into one common expression. Since the volume of the parcel
is arbitrary, the unified integrand would have to vanish, providing us
with a differential equation.
6.3.1 Hydro dynamic volume force
Transforming the surface integral of the traction into a volume inte-
gral can be done using, once again, the Gauss divergence theorem stated
in equation (2.6.25). Identifying the vector h with each one of the three
columns of the stress tensor, we obtain

(6.3.1)
In index notation,

(6.3.2)

where summation of the repeated index j is implied; the index i is free

to vary over x,y, or z.
The divergence of the stress tensor V • a under the integral sign on
the right-hand side of (6.3.1) or (6.3.2) is a vector denoted by

S
- V ' a' (6.3.3)
with Cartesian components

(6.3.4)

Physically, S is the hydrodynamic force per unit volume of fluid; in

contrast, the traction f is the hydrodynamic force per unit surface area.

6.3.2 Force on an infinitesimal parcel

To confirm identity (6.3.2), we consider a small fluid parcel with the
shape of a rectangular parallelepiped centered at the origin, as illustrated
in figure 5.1.1(b). The six flat sides of the parcel are perpendicular
to the x, y, or z axis, the lengths of the three edges are, respectively,
equal to Ax, Ay, and Az, and the volume of the parcel is equal to
AF = Ax Ay Az.
Consider the surface integral on the left-hand side of equation (6.3.2).
Over the sides that are perpendicular to the x axis, located at x = ^r,
and x = -^f, designated as the first and second side, the unit normal
vector is parallel to the x axis; over the first side nx — 1, and over the
second side nx = — 1. Because the size of the parcel is small, the stresses
over each side may be approximated with the corresponding values at
the side center. Subject to this approximation, the surface integral on
the left-hand side over the first and second side take the forms

(6.3.5)
where the parentheses enclose the coordinates of the evaluation point.
Adding the two contributions, and factoring out the common product
AyAz, we obtain

(6.3.6)
Next, we observe that, in the limit as Ax tends to zero, the ratio of
the differences

(6.3.7)
tends to the partial derivative daxi/dx evaluated at the origin. Corre-
spondingly, the difference (6.3.6) reduces to

(6.3.8)
where the derivatives are evaluated at the origin.
Working in a similar fashion with pairs of sides that are perpendicular
to the y or z axis, and summing the contributions, we find that the left-
hand side of (6.3.2) takes the approximate form

(6.3.9)
where the quantity enclosed by the parentheses is evaluated at the origin.
Expression (6.3.9) is an approximation to the volume integral on the
right-hand side of (6.3.2), confirming the identity.

6.3.3 The equation of motion

Substituting now (6.3.1) into (6.1.7), consolidating the integrals, and

arguing that, because the volume of integration is arbitrary, the com-
posite integrand must vanish, we obtain Cauchy's differential equation

(6.3.10)

governing the motion of a compressible or incompressible fluid. In index

notation,

(6.3.11)

where summation of the repeated index j is implied, and the index i is

free to vary over x, y, or z.
Using equations (2.8.7) and (6.2.10), we derive two equivalent forms
of (6.3.10) involving Eulerian derivatives,

(6.3.12)

and

(6.3.13)

Explicitly, the three scalar components of (6.3.12) are given by

(6.3.14)
The terms enclosed by the parentheses on the left-hand sides are the
Cartesian components of the point particle acceleration; the right-hand
sides include the Cartesian components of the volume force due to the
hydrodynamic stresses and the components of the body force.
We emphasize again that equations (6.3.10) - (6.3.13) are valid for
both compressible and incompressible fluids.

6.3.4 Evolution equations

Given the instantaneous velocity and stress fields, we may evaluate

the the right-hand sides of (6.3.10) and (6.3.12), as well as the second
term on the left-hand side of (6.3.12), and thereby compute the rates of
change Du/Dt and du/dt. This observation suggests that the equation
of motion (6.3.10) is, in fact, an evolution equation for the point particle
velocity, and equation (6.3.12) is an evolution equation for the velocity
at a fixed point in the flow.
A similar evolution equation for the density was derived in Chapter
2 on the basis of the continuity equation, as shown in (2.7.15). The evo-
lution equations for the density and velocity originate from fundamental
physical laws: mass conservation is required by thermodynamics, and
Newton's second law of motion expresses a fundamental law of classical
mechanics.

6.3.5 Cylindrical polar coordinates

In the cylindrical polar coordinates illustrated in figure 1.3.2, the
hydrodynamic volume force defined in equation (6.3.3) is expressed by
S - Ex ex + S^ eff + S^ e^. (6.3.15)
Using the rules of coordinate transformation and the chain rule of dif-
ferentiation, we find

(6.3.16)
The cylindrical polar components of the equation of motion are
 pax = Y>x +pgx, pap = Za +pga, p Q1^ = S^ + p #<p, ( 5 3 ^ 7 )

where ax,aa, and a^ are the cylindrical polar components of the point
particle acceleration given in equations (2.8.10).
Using the alternative expressions (2.8.11), we find

(6.3.18)

Centrifugal force

The first term on the right-hand side of the second of equations

(6.3.18), PU^/a, is the centrifugal force familiar from classical mechanics.
This term expresses an effective volume force in the a direction due to
fluid motion in the meridional (p direction. The centrifugal force arises,
for example, in the flow generated by the rotation of a circular cylinder
about its axis in a viscous liquid, as will be discussed in Section 7.5.

Coriolis force

The negative of the first term on the right-hand side of the third of
equations (6.3.18), pua u^/'a, is the Coriolis force. This term expresses
an effective force in the (p direction when flow occurs in both the a and
(p directions. The Coriolis force arises, for example, in the flow due to a
spinning circular disk immersed in a tank of fluid.

6.3.6 Spherical polar coordinates

In the spherical polar coordinates depicted in figure 1.3.3, the hydro-
dynamic volume force defined in equation (6.3.3) is expressed by

E = Er er + £0 e# + S^ e<p. ,g g ig.
Using the rules of coordinate transformation and the chain rule of dif-
ferentiation, we find

(6.3.20)

The spherical polar components of the equation of motion are

(6.3.21)

where a r ,a#, and a^ are the spherical polar components of the point
particle acceleration given in equations (2.8.13).

6.3.7 Plane polar coordinates

In the plane polar coordinates depicted in figure 1.3.4, the hydrody-

namic volume force defined in equation (6.3.3) is expressed by

S = S7. er + £0 e0. (6.3.22)

Using the coordinate transformation rules and the chain rule of differen-
tiation, we find

(6.3.23)

The plane polar components of the equation of motion are

(6.3.24)
where ar and CIQ are the plane polar components of the point particle
acceleration given in equations (2.8.15).
Alternative expressions are

(6.3.25)

involving, respectively, the centrifugal force and the negative of the Cori-
olis force on the right-hand side.

6.3.8 Vortex force

Returning to equation (6.3.12), we replace the second term in the

parentheses on the left-hand side with the right-hand side of (2.8.23),
and find
(6.3.26)
where
U2 = U2+U2+U2 ( 6 - 3 - 27)

is the square of the magnitude of the velocity. The third term on the
left-hand side of (6.3.26) is called the vortex force. This force appears
when the vorticity vector is non-parallel to the velocity vector, otherwise
their cross product will vanish. A flow wherein the vorticity vector is
parallel to the velocity vector at every point, is called a Beltrami flow.

Problem

Problem 6.3.1 Beltrami flow.

Explain why a two-dimensional or an axisymmetric flow cannot be a
Beltrami flow.

6.4 Euler's and Bernoulli's equations

Euler's equation derives from the equation of motion (6.3.10) by sub-
stituting in it the simplest possible constitutive equation for the stress
tensor corresponding to an ideal fluid, expressed by (4.5.15). Considera-
tion of the individual components of the volume force S, given in (6.3.4),
shows that

(6.4.1)
The equation of motion (6.3.10) then reduces to the Euler equation

(6.4.2)
The associated Eulerian form is

(6.4.3)
Explicitly, the three Cartesian components of (6.4.3) are given by

(6.4.4)
The cylindrical, spherical, and plane polar components of Euler's
equation follow readily from equations (6.3.17), (6.3.21), and (6.3.24),
by using the constitutive equations (4.7.7), (4.7.14), and (4.7.20) with
vanishing fluid viscosity.
Using the expression for the point particle acceleration shown on the
left-hand side of equation (6.3.26), we find an alternative form of Euler's
equation in terms of the vortex force,

(6.4.5)
where u2 = u^ + Uy + u^ is the square of the magnitude of the velocity.

6.4.1 Boundary conditions

Euler's equation is a first-order differential equation for the velocity

and the pressure in the spatial variables. To compute a solution, we
require one scalar boundary condition or two scalar jump conditions
involving the velocity or pressure along each boundary:
 • Over an impermeable surface, we require the no-penetration con-
dition.

• Over a free surface, we require that the pressure be equal to the

ambient pressure increased or decreased by an amount that is equal
to the product of the surface tension and twice the local mean
curvature.

• Over a fluid interface, we require a kinematic and a dynamic conti-

nuity or jump condition. The kinematic condition requires that the
normal component of the fluid velocity remain continuous across
the interface. The dynamic condition requires that the pressure
undergo a discontinuity by an amount that is equal to the product
of the surface tension and twice the local mean curvature.

6.4.2 Bernoulli's equation for irrotational flow

If the flow is irrotational, the third term on the left-hand side of

(6.4.5) vanishes. Expressing the velocity in terms of the gradient of the
velocity potential 0, as shown in equation (3.2.3) and more explicitly
in equations (3.2.10), we find that Euler's equation (6.4.5) assumes the
simplified form
(6.4.6)
The order of time and space differentiation in the gradient of the poten-
tial may be freely switched in the first term on the left-hand side.
Assuming now that the density of the fluid is uniform, we express
the acceleration of gravity as the gradient of the scalar s = g • x, writing

g = Vs = V(g • x) = V(gx x + gyy + gz z). (g.4.7)

Substituting this form into (6.4.6), and collecting all terms under the
gradient, we find

(6.4.8)
Since the spatial derivatives of the scalar quanitity enclosed by the
parentheses on the left-hand side vanish, this quantity must be spatially
uniform but possibly time-dependent. On the basis of this argument,
Euler's equation for irrotational flow reduces to Bernoulli's equation for
the irrotational flow of a uniform-density fluid,
 (6.4.9)
where c(t) is an unspecified function of time.

Evolution of the potential

Bernoulli's equation (6.4.9) may be regarded as an evolution equa-

tion for the harmonic potential: given the instantaneous velocity and
pressure, we may evaluate the second, third, and fourth terms on the
left-hand side, compute the time derivative d(/)/dt, and advance the po-
tential over a small period of elapsed time. The last term c(i) causes the
potential to increase or decrease uniformly at the same rate throughout
the fluid. But since the velocity is computed by taking derivatives of the
potential with respect to the spatial coordinates, this uniform change
does not affect the distribution of the velocity in the fluid.
In problems involving a free surface with a prescribed pressure on one
side, it is appropriate to convert the Eulerian time derivative d(f)/dt on
the left-hand side of (6.4.9) to the material derivative D(J)/Dt. Invoking
the definition of the material derivative, and expressing the velocity as
the gradient of the potential, as shown in equations (3.2.10), we find

(6.4.10)
Combining equations (6.4.9) and (6.4.10), we find

(6.4.11)
which provides us with the rate of change of the potential following a
point particle.

Fluid sloshing in a tank

Consider, for example, the sloshing of a fluid in a container, as

illustrated in figure 6.4.1. The pressure at the free surface on the side of
the liquid, denoted by PFS, is related to the ambient pressure PAITH by
the dynamic boundary condition

PFS = PAim + 7 2 Km, /g 4 12)

Figure 6.4.1 Irrotational motion due to the sloshing of a fluid in a con-
tainer. Bernoulli's equation provides us with an evolution equation
for the potential following the motion of point particles distributed
over the free surface.

where 7 is the surface tension, and Km is the mean curvature of the free
surface. Applying equation (6.4.11) at a point on the free surface, and
using (6.4.12), we find that the rate of change of the potential following
a point particle is given by

(6.4.13)
Integrating this equation in time, we obtain a boundary condition for
the potential over the free surface.

6.4.3 Bernoulli's equation for steady irrotational flow

At steady state, the time derivative of the potential on the left-hand

side of (6.4.9) vanishes, yielding the best-known version of Bernoulli's
equation
(6.4.14)
The time-dependence of the constant c(t) on the right-hand side accounts
for a possible uniform change in the level of the pressure throughout the
fluid.
The three terms on the right-hand side of the (6.4.14) express, re-
spectively, the kinetic energy, the potential energy due to the pressure,
Figure 6.4.2 Illustration of gravity-driven drainage of a fluid from a
tank. The exit velocity may be computed using Bernoulli's equa-
tion for irrotational flow, and is given by Toricceli's law.

and the potential energy due to the body force, all three per unit mass
of fluid. Bernoulli's equation requires that the sum of the three energies
be uniform thoughout the domain of flow.

Toricelli's law

Bernoulli's equation allows us to carry out approximate engineering

analyses of several classes of internal and external high-speed flows. Con-
sider, for example, the gravity-driven drainage of a fluid from a tank, as
illustrated in figure 6.4.2. If the rate of drainage is sufficiently slow, the
flow may be assumed to be in a quasi-steady state; that is, the magni-
tude of the time derivative is small compared to the rest of the terms
in the unsteady Bernoulli equation (6.4.9), and the steady version of
Bernoulli's equation (6.4.14) can be employed.
To compute the velocity at the point of drainage, denoted by C/, we
evaluate the left-hand side of (6.4.14) first at the free surface and second
at the point of drainage, and set the two expressions equal. Consider-
ing the velocity at the free surface negligible, and setting the pressure
at the free surface and at the point of drainage equal to the ambient
 Test section
Plane 2

Plane 1

Figure 6.4.3 Flow in a wind or water tunnel with a contraction that

dampens small perturbations.

atmospheric pressure, we derive Toricelli's law

U = ^Flgh, (6.4.15)
also describing the velocity of a rigid body in free gravitational fall.

Decay of perturbations in a wind or water tunnel

Wind and water tunnels are used widely in studies of high-speed

flows. To obtain a desirable uniform velocity profile, the tunnel is de-
signed with a smooth contraction upstream from a test section where
measurement or observation takes places, as illustrated in figure 6.4.3.
Consider a small perturbation of the otherwise flat upstream velocity
profile at plane 1, as illustrated in figure 6.4.3. The pressure is assumed
to be uniform over any cross-section along the contraction. Applying
Bernoulli's equation (6.4.14) for the fluid outside or inside the perturbed
region, and neglecting the effects of gravity, we find

(6.4.16)
Combining these equations to eliminate the pressure, and rearranging,
we find
(6.4.17)
Figure 6.4.4 Irrotational free-surface flow of a horizontal stream over
a hump.

Now, because the perturbation has been assumed small, the actual
velocities u\ and u^ may be replaced with the unperturbed velocities
Ui and C/2 in the numerator and denominator of the fraction on the
right-hand side of (6.4.17), yielding

(6.4.18)

Combining the approximate mass balance Ui Si — C/2 82 with equation

(6.4.18), and rearranging, we obtain

(6.4.19)

which shows that the relative amplitude of the perturbation decays like
the square of the contraction ratio 3%/Si, confirming that the contraction
aids in the establishment of a uniform velocity profile.

Flow of a horizontal stream over a hump

As a third example, we consider steady two-dimensional irrotational

flow of a horizontal stream over a gently sloped hump, called the Venturi
flume, as illustrated in figure 6.4.4. The free surface is located at y =
h(x] + d(x), where h(x) is the height of the hump and d(x) is the depth
of the stream. As x tends to plus or minus infinity, h(x) tends to zero.
The profile of the streamwise velocity is assumed to be uniform across
the stream; thus, Ux = u(x).
 Applying Bernoulli's equation (6.4.14) first at a point at the free
surface located far upstream, and second at an arbitrary point at the
free surface, neglecting the y component of the free-surface velocity and
the pressure drop across the free surface due to surface tension, and
noting that the gravitational acceleration vector is given by g = (O, -#),
we find

(6.4.20)
where UQ is the upstream velocity, and do = d(x = -oo) is the upstream
depth. Combining now the mass conservation equation UQC[Q = u(x)d(x)
with equation (6.4.20) to eliminate u(x), and rearranging the resulting
expression, we derive a cubic equation for the reduced layer depth d(x) =
d(z)/d 0 ,

(6-4.21)
where h(x) = /i(x)/do is the reduced height of the hump. We have
introduced the dimensionless ratio

(6.4.22)
expressing the relative magnitude of inertial and graviational forces,
called the Proude number.
In practice, the Venturi flume is used to deduce the flow rate from
measurements of the deflection of the free surface from the horizontal
position. As the Froude number tends to zero, gravitational forces dom-
inate, and equation (6.4.21) has the obvious solution d(x) = 1 — h(x),
which shows that the free surface tends to become flat. On the other
hand, as the Proude number tends to infinity, inertial forces dominate,
and equation (6.4.21) has the obvious solution d(x) = 1, which shows
that the depth of the stream remains constant, and the free surface fol-
lows the topography of the hump. Intermediate values of the Froude
number yield free-surface profiles with a downward deflection (problem
c.6.4.1).

6.4.4 Bernoulli's equation for steady rotational flow

Let us return to Euler's equation (6.4.5), and consider a rotational

flow at steady state. The time derivative on the left-hand side vanishes,
yielding
Figure 6.4.5 A system of Cartesian coordinates with the x axis tangen-
tial to a streamline, used to derive Bernoulli's equation for steady
rotational flow, equation (6.4.26).

(6.4.23)

where u2 = u% + Uy + u2z is the square of the magnitude of the velocity.

The x component of equation (6.4.23) reads

(6.4.24)

Next, we place the origin of the Cartesian axes at a point in the

fluid, identify the streamline that passes through that point, and orient
the x axis tangentially to the streamline and thus parallel to the local
velocity, as illustrated in figure 6.4.5. By construction then, the y and
z components of the velocity vanish at the origin, and the second and
third terms on the left-hand side of equation (6.4.24) are equal to zero.
Taking advantage of these simplifications, we derive the reduced form

(6.4.25)
where the left-hand side is evaluated at the origin. Equation (6.4.25)
states that the rate of change of the quantity enclosed by the parentheses
on the left-hand side with respect to distance along the streamline should
vanish; thus, the quantity enclosed by the parentheses must remain con-
stant along the streamline. An equivalent mathematical statement is

(6.4.26)

where the function /(x,y, z) remains constant along a streamline. In

two-dimensional or axisymmetric flow, the function /(:r,y,z) may be
considered as a function of the stream function if) which, by definition,
is constant along a streamline.

Problems

Problem 6.4.1 Flow through a sudden enlargement.

Consider the flow through a sudden enlargement depicted in figure
6.2.1. Use Bernoulli's equation to compute the rise in pressure P^ — PI,
and compare your result to that shown in equation (6.2.16) obtained by
an approximate integral momentum balance.
Problem 6.4.2 Bernoulli's equation for two-dimensional flow with
uniform vorticity.
The velocity field u of a two-dimensional flow in the xy plane with
uniform vorticity LJZ = Jl may be decomposed into the velocity field
of a simpler two-dimensional flow with uniform vorcitity Q, denoted by
v, and a potential flow expressed by the harmonic potential </>, so that
u — v + V</>. Two examples of simpler flows are the simple shear flow
with velocity v — (—fi j/, O), and a flow expressing rigid-body rotation
with velocity v = § (—y, x). Using this decomposition, derive Bernoulli's
equation
(6.4.27)

where ^ is the stream function and u = |u| is the magnitude of the

velocity of the decomposed flow.
Problem 6.4.3 Flow due to an unsteady point source or point vortex.
(a) Discuss whether the flow due to a two- or three-dimensional point
source with time-dependent strength satisfies the Euler equation for in-
viscid flow.
(b) Repeat (a) for a two-dimensional point vortex.
Problem 6.4.4 Force on a sphere in accelerating potential flow .
Consider unsteady irrotational flow past a sphere that is held station-
ary in an accelerating stream with velocity Vx(t). The velocity potential
and Cartesian components of the velocity are given by equations (3.6.10)
and (3.6.11). Use Bernouli's equation (6.4.9) to evaluate the pressure,
and then compute the force exerted on the sphere by evaluating the
surface integral
(6.4.28)
where n is the unit vector normal to the sphere pointing into the fluid.
Based on this result, evaluate the force exerted on a sphere held station-
ary in a non-accelerating steady flow, and discuss the physical relevance
of the assumption of irrotational motion.

Computer problems

Problem c.6.4.1 Flow over a hump.

Consider the flow of a horizontal stream over a hump illustrated in
figure 6.4.4. The height of the hump is described by the parabolic shape
function h(x) — h$ [I - (x/a) 2 ], for -a < x < a, where a is the half-
length of the hump and /IQ is the maximum height. Substituting this
profile into (6.4.21), we find

(6.4.29)
for -1 < x < 1, where x = x/a is the reduced distance from the mid-
point. Compute and plot the reduced layer depth d against x for /IQ/do —
0.01, 0.05, 0.10 and Fr = 0.01, 0.1, 10, 100, and discuss the free-surface
shapes.

6.5 The Navier-Stokes equation

The Navier-Stokes equation derives from the equation of motion
(6.3.10) using the constitutive equation for the stress tensor for an in-
compressible Newtonian fluid, equation (4.5.3). If the viscosity of the
fluid is uniform, the hydrodynamic volume force is given by

S = V - a = V - ( - p I + / i 2 E ) = -Vp + /^ 2 V • E, (6.5.1)
where I is the identity matrix.
Working in index notation, we find that the ith component of twice
the divergence of the rate of deformation tensor E on the right-hand side
of (6.5.1) is given by

(6.5.2)

where summation of the repeated index j is implied. The term enclosed

by the parentheses on the right-hand side of (6.5.2) is equal to the diver-
gence of the velocity; because the fluid has been assumed incompressible,
this term is equal to zero. The first term on the right-hand side of (6.5.2)
is the Laplacian of the ith component of the velocity,

(6.5.3)

Using these results to simplify expression (6.5.1), we find that the

hydro dynamic volume force is given by

S EE V • a = -Vp + n V 2 U.
(6.5.4)

Correspondingly, the equation of motion (6.3.10) reduces to the Navier-

Stokes equation

(6.5.5)

which is distinguished from Euler's equation (6.4.2) by the presence of

the viscous force represented by the product of the viscosity and the
Laplacian of the velocity on the right-hand side.
The associated Eulerian form is

(6.5.6)
whose three Cartesian components are:

(6.5.7)

6.5.1 Pressure and viscous forces

The first term on the right-hand side of (6.5.6), equal to the negative
of the pressure gradient, represents the pressure force. The second term,
equal to the Laplacian of the velocity multiplied by the viscosity, is the
viscous force.
Working in index notation under the assumption that the fluid is
incompressible and thus the velocity field is solenoidal, V • u = O, we
find that the Laplacian of the velocity is equal to the negative of the curl
of the vorticity,
V 2 U = -V x a; (6-5.8)
(problem 6.5.1). An important consequence of this identity is that, if
the flow is irrotational, or if the vorticity is constant, or if the vorticity
field is irrotational, then the viscous force vanishes even though the fluid
may not be inviscid. In this case, the Navier-Stokes equation reduces to
Euler's equation which may be integrated to yield Bernoulli's equation
(6.4.9) for irrotational flow, or equation (6.4.26) for steady rotational
flow.

A radially expanding or contracting bubble

An example of a viscous flow with non-zero viscous stresses but van-

ishing viscous forces is provided by the irrotational flow generated by
the radial expansion or contraction of a spherical bubble with time-
dependent radius a(t). The velocity may be represented in terms of a
three-dimensional point source with time dependent strength m(i) placed
at the bubble center. In spherical polar coordinates with the origin at
the bubble center, the velocity potential and radial component of the
velocity are given, respectively, by

(6.5.9)
where m(t) is the strength of the point source. The no-penetration con-
dition at the surface of the bubble requires da/dt = ur(r — a), which
may be rearranged to give an expression for the strength of the point
source in terms of the bubble radius,

(6.5.10)
Substituting (6.5.10) into equations (6.5.9), we find

(6.5.11)
Referring now to Bernoulli's equation (6.4.9) for unsteady irrotational
flow, we compute the first and second terms on the left-hand side,

(6.5.12)
Substituting these expressions into Bernoulli's equation (6.4.9), and solv-
ing for the pressure, we find

(6.5.13)
Far from the bubble, the first and second terms on the right-hand side
of (6.5.13) vanish, and the pressure is given by the linear and possibly
time-dependent distribution Poo(x5 *) — P [c(t) + g • x],
The normal stress arr undergoes a jump across the surface of the
bubble, as determined by the bubble radius and surface tension 7. Using
the simplified version of the interfacial condition (4.3.13) for constant
surface tension, we find
 (6.5.14)
where ps(t) is the pressure in the interior of the bubble, and Acm = I/a is
the mean curvature of the interface. Substituting the second of (6.5.11)
into the Newtonian constitutive equation orr — -p + 2 p, dur/dr, and
the result into (6.5.14), we find

(6.5.15)
Finally, we apply (6.5.13) at the surface of the bubble, evaluate the pres-
sure from (6.5.15), neglect hydrostatic variations over the diameter of the
bubble, and rearrange the resulting expression to obtain the generalized
Rayleigh equation

(6.5.16)
where x# is the location of the bubble center.
Equation (6.5.16) is a second-order nonlinear ordinary differential
equation governing the evolution of the bubble radius. To compute the
solution, we require the initial bubble radius, the initial rate of expansion
da/dt, and also the bubble pressure and pressure at infinity. The bubble
pressure may be further related to the bubble volume by means of an
appropriate equation of state provided by thermodynamics.

6.5.2 Boundary conditions

The Navier-Stokes equation is a second-order differential equation

for the velocity with respect to the spatial coordinates. To compute a
solution, we require one scalar boundary condition for each component
of the velocity or traction over each boundary:
• Over an impermeable solid surface, we require the no-penetration
and no-slip boundary conditions.
• Over a free surface, we require that the normal component of the
traction be equal to the ambient pressure increased or decreased
by an amount that is equal to the product of the surface tension
and twice the local mean curvature, and the tangential component
vanish.
 • Over a fluid interface, we require kinematic and dynamic conti-
nuity or jump conditions. The kinematic condition requires that
all components of the velocity be continuous across an interface.
The dynamic condition requires that the normal component of the
traction undergo a discontinuity by an amount that is equal to the
product of the surface tension and twice the local mean curvature,
and the tangential component of the traction undergo a discontinu-
ity that is determined by the Marangoni tractions due to variations
in surface tension.

6.5.3 Cylindrical polar coordinates

The cylindrical polar components of the hydrodynamic volume force

for a Newtonian fluid arise by substituting the constitutive relations
(4.7.7) into expressions (6.3.16). After a fair amount of algebra, we find

(6.5.17)
The cylindrical polar components of the Navier-Stokes equation arise
by substituting these expressions into the right-hand sides of (6.3.17) or
(6.3.18).

6.5.4 Spherical polar coordinates

The spherical polar components of the hydrodynamic volume force

arise by substituting the constitutive relations (4.7.14) into expressions
(6.3.20). After a fair amount of algebra, we find
 (6.5.18)
The Laplacian operator V2 in spherical polar coordinates was defined in
equation (3.2.17). The spherical polar components of the Navier-Stokes
equation arise by substituting these expressions into the right-hand sides
of (6.3.21).

6.5.5 Plane polar coordinates

The plane polar components of the hydrodynamic volume force arise
by substituting the constitutive relations (4.7.20) into expressions (6.3.23).
The result is

(6.5.19)
The plane polar components of the Navier-Stokes equation arise by sub-
stituting these expressions into the right-hand sides of (6.3.25).

Problems

Problem 6.5.1 Viscous force.

Prove identity (6.5.8). Hint: Set the vorticity equal to the curl of
the velocity; express the curl of the vorticity in index notation; and then
use identity (2.3.9).
Problem 6.5.2 Steady flow.
Consider a flow at steady state. Explain why it is not generally
permissible to specify an arbitrary solenoidal velocity field that satisfies
the boundary conditions, and then compute the pressure by solving the
Navier-Stokes equation (6.5.6). Hint: Consider the conditions subject to
which the equation Vp = F, where F is a vector function, has a solution,
and note that the curl of the gradient of a function vanishes, as shown
in equation (6.6.10).
Problem 6.5.3 Expansion of a bubble.
Show that, when the right-hand side of (6.5.16) vanishes and vis-
cous stresses and surface tension are insignificant, an exact solution to
equation (6.5.16) is given by

(6.5.20)

6.6 Vorticity transport

In Section 6.3, we interpreted the equation of motion as an evolution
equation determining the rate of change of the velocity of a point par-
ticle, or the rate of change of the fluid velocity at a fixed point in the
flow. Descendant evolution equations governing the rate of change of the
spatial derivatives of the velocity comprising the velocity gradient tensor
and its symmetric and skew-symmetric components may be derived by
straightforward differentiation.
Of particular interest is the evolution of the skew-symmetric part of
the velocity gradient tensor related to the vorticity as shown in equation
(2.3.13). The availability of an evolution equation for the vorticity allows
us to study the rate of change of the angular velocity of a small fluid
parcel as it translates and deforms while it is convected in a flow.

6.6.1 Two-dimensional flow

To begin, we consider the evolution of the strength of the vorticity
uz in a two-dimensional flow, defined in terms of the velocity in equation
(2.3.15). To derive an evolution equation for u;2, we divide both sides
of the equation of motion (6.3.12) by the density to remove it from
the left-hand side, and then take the y derivative of the x component
of the resulting equation and subtract it from the x derivative of the
corresponding y component. The result is

(6.6.1)
Expanding out the derivatives on the left-hand side, and using the conti-
nuity equation dux/dx + duy/dy = O, we obtain the remarkably simpler
form

(6.6.2)
The left-hand side of (6.6.2) expresses the material derivative of the
vorticity which, according to equation (2.3.7), is equal to twice the rate
of change of the angular velocity of a small fluid parcel.
Next, we expand out the derivatives on the right-hand side of (6.6.2),
and expess the hydrodynamic volume force in terms of the stresses using
the two-dimensional counterparts of equations (6.3.4), and thus obtain
the general form of the vorticity transport equation for an incompressible
fluid,

(6.6.3)

Baroclinic production of vorticity

The term enclosed by the first set of parentheses on the right-hand

side of (6.6.3) expresses generation of vorticity due to density inhomo-
geneities, known as baroclinic production of vorticity. To illustrate the
physical mechanism that is responsible for this term, we consider a verti-
cal column of fluid whose density increases upward in the direction of the
y axis, as depicted in figure 6.6.1; thus, dp/dy > O. The x component
of the hydrodynamic volume force E^ causes the column to accelerate in
the positive direction of the x axis; because the density and hence the
inertia of the fluid increases with height, the top will accelerate less than
the bottom. As a result, the column will buckle backwards exhibiting
counterclockwise rotation expressed by the second terms within the first
set of parentheses on the right-hand side of (6.6.3).
Figure 6.6.1 Vorticity is generated when a column of fluid that is heav-
ier at the top buckles in acceleration under the influence of a volume
force.

Inviscid fluids

If viscous forces are insignificant, the shear stresses vanish, the normal
stresses are equal to the negative of the pressure, and the hydrodynamic
volume force is equal to the negative of the pressure gradient. Conse-
quently, the term enclosed by the square brackets on the right-hand side
of (6.6.3) makes no contribution, and the term expressing baroclinic pro-
duction obtains a simple form, yielding the vorticity transport equation

(6.6.4)

where ez is the unit vector along the z axis which is perpendicular to

the xy plane of the flow.
For a fluid with uniform density, equation (6.6.4) predicts

(6.6.5)

showing that a small fluid parcel rotates at a constant angular velocity

as it is convected by the flow. The physical origin of this remarkably
simple result can be traced back to conservation of angular momentum
in the absence of shear stresses imparting a torque.
 Incompressible Newtonian fluids
Considering next the evolution of the vorticity in an incompressible
Newtonian fluid with uniform density and viscosity, we substitute the
constitutive equation for the stress tensor shown in equations (4.5.4) into
the right-hand side of (6.6.3), and simplify the resulting expression by
use of the continuity equation to derive the vorticity transport equation

(6.6.6)
where v = IJL/p is a physical constant called the kinematic viscosity.
The following table displays the kinematic viscosities of water and air
at three temperatures. Note that the kinematic viscosity of air is higher
than that of water by two or three orders of magnitude. In contrast,
the viscosity of water is higher than that of air by one or two orders of
magnitude.
The right-hand side of (6.6.6) expresses diffusion of vorticity in the xy
plane. Like temperature or concentration of a species, vorticity spreads
out from regions of highly rotational flow, that is, regions where small
spherical parcels exhibit intense rotation, to regions of weakly rotational
or irrotational flow. The actual mechanism by which this occurs will
be exemplified in subsequent chapters with reference to unsteady and
boundary-layer flows.

Kinematic viscosity, v

Temperature Water Air

0
C IQ-2Cm2 sec l(T cra2 sec
2

20 1.004 15.05
40 0.658 18.86
80 0.365 20.88

6.6.2 Axisymmetric flow

Next, we consider an axisymmetric flow without swirling motion, and

refer to the cylindrical polar coordinates (#, a, (p] depicted in figure 6.6.2.
Figure 6.6.2 The vorticity of a point particle in axisymmetric flow in-
creases as the particle moves farther away from the axis of symme-
try, due to vortex stretching.

Working as previously for two-dimensional flow, we derive the counter-

part of the vorticity transport equation (6.6.6) for an incompressible
Newtonian fluid with uniform density and viscosity,

(6.6.7)

The second-order linear differential operator E2 on the right-hand side

of (6.6.7), defined in equations (2.9.12) and (2.9.15), is the counterpart
of the Laplacian operator for two-dimensional flow shown in (6.6.6).

Vortex stretching

If viscous forces are negligible, the right-hand side of (6.6.7) vanishes,

and the resulting vorticity transport equation for inviscid flow takes the
form

(6.6.8)

This equation requires that the strength of the vorticity of a point par-
ticle, Up, be proportional to the distance of the point particle from the
axis of symmetry, a, so that the ratio between them remains constant
in time and equal to the initial value, as illustrated schematically in fig-
ure 6.6.2. This fundamental evolution law expresses a physical process
known as vortex stretching. The significance of vortex stretching will be
discussed in Chapter 11 in the framework of vortex flows.
6.6.3 Three-dimensional flow

Generalizing the preceding discussion, we set out to derive an evo-

lution equation for the vorticity vector of a three-dimensional flow. For
simplicity, we restict our attention to incompresible Newtonian fluids
and assume that the density and viscosity are uniform throughout the
domain of flow. Our point of departure is the Navier-Stokes equation
(6.4.2).
Using the expression for the point particle acceleration shown on the
left-hand side of equation (6.3.26), we find the following alternative form
of the Navier-Stokes equation in terms of the vortex force,

(6.6.9)
where u2 = u*. + vfc + ILJ, is the square of the magnitude of the velocity.
To derive an evolution equation for the vorticity, we take the curl of
both sides of equation (6.6.9). A vector identity states that the curl of
the gradient of a smooth scalar function of position / vanishes,

V x V/ = O. (6.6.10)
The proof follows readily working in index notation: the ith component
of the left-hand side of (6.6.10) is given by

(6.6.11)
The symmetry of the second derivative on the right-hand side of (6.6.11),
combined with the inherent antisymmetry of the alternating tensor, re-
quires that the right-hand side vanish.
Using identity (6.6.10), we find that the curl of the second term on
the left-hand side of (6.6.9) and the curl of the first term on the right-
hand side of (6.6.9) both vanish. Invoking the definition u = V x u, we
obtain the vorticity transport equation for three-dimensional flow,

(6.6.12)
where v ~ IJL/p is the kinematic viscosity.
The second term on the left-hand side of (6.6.12) can be manipu-
lated to acquire a physical interpretation. In index notation, the ith
component of the vector expressed by this term is given by
 (6.6.13)
Using property (2.3.9), we recast the right-hand side of (6.6.13) into the
form

(6.6.14)
An identity states that the divergence of the curl of a vector field vanish-
es; a consequence of this identity is that the vorticity field is solenoidal
(problem 2.3.4). Because the fluid has been assumed incompressible, the
velocity field is also solenoidal. Consequently, the second and third terms
on the right-hand side of (6.6.14) vanish. Substituting the result back
into equation (6.6.12), we find the desired vorticity transport equation

(6.6.15)
or in vector notation,

(6.6.16)
The left-hand side of (6.6.15) or (6.6.16) is the material derivative of the
vorticity, that is, the rate of change of the vorticity following the motion
of point particles.

Vorticity rotation and stretching

To understand the nature of the first term on the right-hand side

of (6.6.16), we consider a small material vector rfl, and label the first
point as A, and the last point as B. Using a Taylor series expansion, we
find that the difference in the velocity across the end-points is given by
u5 — UA = d\ - Vu. Comparing this expression with the first term on
the right-hand side of (6.6.16), we find that the vorticity vector behaves
like a material vector convected by the fluid: it rotates and stretches or
compresses under the influence of the local flow.
 In the case of two-dimensional flow, because the vorticity vector is
normal to the plane of the flow, neither rotation nor stretching or com-
pression can take place. In the case of axisymmetric flow, because the
vorticity vector points in the meridional direction, rotation is prohibit-
ed, but stretching or compression can take place, as discusssed in Section
6.6.2.
Persistence of irrotational motion in inviscid flow
One important consequence of (6.6.16) for inviscid fluids is that, if the
vorticity of a point particle vanishes at the initial instant, it will vanish
at all times. Thus, volumes of rotational fluid remain rotational, volumes
of irrotational fluid remain irrotational, and the interface between them
remains sharp and well-defined at all times.
Source of vorticity in viscous flow
In practice, a fluid flow is established from the state of rest; conse-
quently, the initial vorticity distribution is equal to zero. Since the right-
hand side of the vorticity transport equation (6.6.16) vanishes through-
out the fluid, the initial rate of production of vorticity is equal to zero,
and this seemingly suggests that the flow will remain irrotational at all
times, which is known not to be true. The paradox is resolved by ob-
serving that vorticity, like heat, enters the fluid by diffusion across the
boundaries. The precise mechanism by which this occurs will be dis-
cussed in Chapter 7.

Problems

Problem 6.6.1 Reduction to two-dimensional flow.

Show that the vorticity transport equation (6.6.16) reproduces the
transport equation (6.6.6) for the strength of the vorticity uz of a two-
dimensional flow.
Problem 6.6.2 Convection of vorticity.
Prove the identity
u dui j duj
* Wi = " a£> ^6-6-17)
which allows us to express the first term on the left-hand side of (6.6.16)
in the alternative form (Vu) • u. Hint: Begin with the identity u x uj =
U x V x u = O, and then work in index notation using identity (2.3.9).
6.7 Dynamic similitude, the Reynolds number,
and dimensionless numbers
in fluid dynamics
Consider streaming flow along the x axis with velocity U\ past a
stationary body with desingated size L\ as illustrated in figure 6.7.1(a),
and another streaming flow along the x axis with velocity 172 past a
second body that arises by shrinking or expanding the first body by a
certain factor, as illustrated in figure 6.7.1(b). If the surface of the first
body is described by the equation

f(x,y,z) = 0, (W.I)

then, the surface of the second body is described by the equation

(6.7.2)

where
(6.7.3)

is a scaling factor.
For example, if the first body is a sphere with diameter L\ centered
at the the point (xx^yc^zc)^ then

(6.7.4)

and

(6.7.5)

Setting the right-hand side of (6.7.5) equal to zero, we obtain the equa-
tion of a sphere with diameter L^ centered at the point (a xc, a yc, a ZQ).
Both fluids are assumed to be incompressible and Newtonian. Let pi
and //i be the density and viscosity of the first fluid, and p% and /^2 be
the density and viscosity of the second fluid.
Figure 6.7.1 Flows in two similar domains. If the Reynolds numbers
are equal, as shown in equation (6.7.6), the velocity and pressure
field of the second flow may be deduced from those of the first flow
by rescaling, and vice versa.

6.7.1 Similitude and the Reynolds number

We will show that, when the values of the four control and physical
parameters characterizing the first flow, Li,C/i,pi, and p\, and the cor-
responding values of the control and physical parameters characterizing
the second flow, Z/2, ^2>P2> &nd /^5 are related by the equations

(6.7.6)

then the structure of the second flow may be inferred from the structure
of the first flow, and vice versa, by carrying out a simple computation
described as rescaling, as will be explained in the following subsection.
 The ratio on the left-hand side of (6.7.6) is defined as the Reynolds
number of the first flow, and the the ratio on the right-hand side of
(6.7.6) is defined as the Reynolds number of the second flow.

6.7.2 Rescaling

To deduce the structure of the second flow from the structure of the
first flow, and vice versa, we introduce the dynamic pressure established
due to the flow, defined as the pressure deviation from the hydrostatic
distribution,

P?= Pi-Pi S'^ P2 =P2~ P 2 g - x , (6.7.7)

where the superscript D stands for "Dynamic" or, more accurately, "Hy-
drodynamic". In the absence of an imposed flow, the pressure assumes
its hydrostatic distribution, and the dynamic pressure vanishes though-
out the fluid.
Consider an arbitrary point in the first flow, denoted by XI, and a
corresponding point in the second flow whose coordinates are given by

X2 = Q f X i . (6.7.8)
Equations (6.7.1) and (6.7.2) ensure that, if XI lies at the surface of the
body in the first flow, then X2 will lie at the surface of the body in the
second flow. In Section 6.7.3, we will show that, when relation (6.7.6)
is fulfilled, the velocity and dynamic pressure at the second point in the
second flow are related to those at the first point in the first flow by the
equations

U 2 (X 2 ) = (J ui(X 2 ), P?(X2) -/35 2 ^f(X 2 ), (6.7.9)

where we have defined the ratio of the velocities of the incident flow, £,
the viscosity ratio A, and the density ratio /3,
U<2 2 p2
oA -
= —, \ -= ^—,
A R-
p = —. /fi 7 IQ\
Ui IJLI pi ^D./.luj
The equality of the Reynolds numbers expressed by (6.7.6) requires
/3 6 a = A.
Relations (6.7.9) are also valid for unsteady flow, provided that the
velocity field of the first flow at the designated origin of time is related
to the velocity of the second flow by the first of equations (6.7.9), and
the comparison is made at times ti and £2 related by
 (6.7.11)

6.7.3 Dimensional analysis

To prove relations (6.7.9), we consider the Navier-Stokes equation

(6.5.6) and the continuity equation V - U = O governing the structure
and dynamics of the two flows with appropriate physical constants corre-
sponding to the two fluids, subject to appropriate far-field and boundary
conditions, and work as follows.

First flow

Considering the first flow, we introduce the dimensionless indepen-

dent variables

(6.7.12)
and the dimensionless dependent variables

(6.7.13)
Solving for the dimensional variables in terms of their dimensionless
counterparts denoted by the hats, and substituting the result into the
Navier-Stokes equation and into the continuity equation, we find

(6.7.14)
and
Vi-ui-0, (6.7.15)
where Re\ = pi Ui LI/HI is the Reynolds number of the first flow shown
on the left-hand side of (6.7.6); we have introduced the dimensionless
gradient Vi = (d/dxi,d/dyi,d/dzi) and associated Laplacian operator

The far-field condition requires that, far from the body, the dimen-
sionless velocity uxi tends to unity, whereas uyi and uzi tend to van-
ish. The no-slip and no-penetration boundary conditions require that
the velocity vanishes at points (o;,j/,z) that satisfy equation (6.7.1) or,
equivalently, points (#1,3/1,21) that satisfy
 /(Li £i, LI yi, LI 21) = O. (6.7.16)

Second flow

Considering next the second flow, we introduce the dimensionless

independent variables

(6.7.17)
and the dimensionless dependent variables

(6.7.18)
Solving for the dimensional variables in terms of their dimensionless
counterparts, and substituting the result into the Navier-Stokes equa-
tion and into the continuity equation, we find

(6.7.19)
and
V2 • U2 = O, (6.7.20)
where Re2 = p2 U2 L2/'p>2 is the Reynolds number shown on the right-
hand side of (6.7.6); we have introdued the dimensionless gradient V2 =
(d/dx2,d/dy2,d/dz2) and associated Laplacian operator

V22 = (d2/dxld*/dyid2/dzl).
The far-field condition requires that, far from the body, the dimen-
sionless velocity uX2 tends, to unity, whereas uy2 and uZ2 tend to van-
ish. The no-slip and no-penetration boundary conditions require that
the velocity vanishes at points ( x , y , z ) that satisfy equation (6.7.2) or,
equivalently, points (£2, $2,^2) that satisfy

/(Li X 2 , LI 3/2, LI Z2) = O. (6.7.21)

 Comparison

To this end, we compare one by one the equations and boundary con-
ditions governing the two flows in the dimensionless variables indicated
by a hat, and deduce the following:

1. The Navier-Stokes equation (6.7.14) is identical to the Navier-

Stokes equation (6.7.19) provided that the two Reynolds numbers
are equal, as stated in (6.7.6).

2. The continuity equation (6.7.15) is identical to the continuity

equation (6.7.20).

3. The far-field conditions are identical; both dimensionless velocities

tend to (1,0,0).

4. The boundary conditions on the first body described by (6.7.16) are

identical to the boundary conditions on the second body described
by (6.7.21).

These results suggest that, when the Reynolds numbers are equal,
the values of the dimensionless dependent variables in the two flows at
corresponding dimensionless times and corresponding dimensionless po-
sitions are equal. Setting, for example, P^(XI) — p® (£2), and using the
definitions (6.7.13) and (6.7.18), we find the second of relations (6.7.9)
subject to (6.7.10).

6.7.4 Structure of a flow as a function of the Reynolds number

Generalizing the preceding discussion, we consider a flow in a domain

with characteristic length scale L, identify an appropriate characteristic
velocity C/, and compute the Reynolds number

= pLU - LE.
M v ' (6.7.22)

where v = fji/p is the kimematic viscosity. We then introduce the dimen-

sionless independent variables

-_x
X =
- -. y
y= Z=
-_z -_ tu
t=
L> L> L> ^T> (6.7.23)
and the dimensionless dependent variables

(6.7.24)

Solving for the dimensional variables in terms of their dimensionless

counterparts, and substituting the result into the Navier-Stokes equa-
tion and into the continuity equation, we find

(6.7.25)

and
V-u-0. (6.7.26)

Our earlier analysis suggests that the structure of the flow depends on
L, C/, p and IJL collectively through the dimensionless Reynolds number,
in the sense of dynamic similitude expressed by equations (6.7.9) and
(6.7.10).

Characteristic scales

The choice of characteristic velocity and length scale may be obvi-

ous in some cases but subtle in others. If all terms in the dimensionless
Navier-Stokes equation (6.7.25) are of order unity, then the Reynolds
number clearly expresses the relative importance of inertial forces, as-
sumed to scale with p U2/L, and viscous forces, assumed to scale with
IJL U/L2, so that their ratio is the Reynolds number defined in (6.7.22).
If an alternative scaling for these forces is available on physical ground,
then they should be used in place of a generic scaling that lacks physical
insight.

Stokes flow

Inspecting the dimensionless Navier-Stokes equation (6.7.25), we note

that, when the Reynolds number is small, viscous forces dominate and
the left-hand side makes a negligible contribution. The dimensionless
pressure gradient also appears to make a negligible contribution in this
limit, but this is only a mathematical illusion: the dimensionless pressure
arose from the arbitrary scaling shown in the last of equations (6.7.24),
which may be contrasted with the physical scaling of the position vector
and velocity in terms of the unambiguous length and velocity scales
L and U. As a consequence, the dimensionless pressure gradient may
become singular as the Reynolds number tends to vanish, suggesting
that an alternatice scaling is required. To prevent this occurrence, we
retain the dimensionless pressure gradient in the dimensionless form of
the Navier-Stokes equation.
Reverting to dimensional variables, we find that the Navier-Stokes
equation reduces to the Stokes equation

O - -Vp + IJL V 2 U + p g, (6.7.27)

describing steady or unsteady creeping flows with negligible inertial forces.

The analysis and computation of these flows will be the exclusive topic
of our discussion in Chapter 9.

Flows at high Reynolds numbers

Inspecting (6.7.25), we find that when the Reynolds number is large,

viscous forces are negligible and may be neglected throughout the do-
main of flow. This, however, is permissible only when the velocity does
not change rapidly over small distances across fluid layers that are thin
compared to the global size of the boundaries, otherwise the preceding
scaling with respect to U and L may not be valid. Such thin layers
typically occur along flow boundaries or interfaces between two adjacent
streams of the same or different fluids. In Chapter 10, we shall demon-
strate that viscous forces may be substantial or even dominant within
these layers, even though the bulk of the flow may occur at high Reynolds
numbers.

Laminar and turbulent flows

When the Reynolds number exceeds a certain threshold, an unsteady

small-scale motion characterized by rapid fluctuations in the velocity
and vorticity field is spontaneously established. This turbulent motion
is superposed on a steady or unsteady slower-evolving macroscopic or
large-scale flow. Flows below the critical Reynolds number are called
laminar to indicate that the streamlines are smooth, and flows above
thie critical Reynolds number are called turbulent to indicate that the
instantaneous streamlines are highly convoluted.
The transition from laminar to turbulent motion may occur by several
mechanisms including the amplificiation of internal waves. The critical
Reynolds number where transition occurs may be estimated theoretically
by carrrying out a stability analysis, as will be discussed in Chapter 10.
The dynamics of turbulent motion may be studied by several methods
including statistical analysis, nonlinear dynamical systems theory, and
vortex dynamics, as will be discussed in Chapter 10.

6.7.5 Dimensionless numbers in fluid mechanics

We have demonstrated that two geometrically related flows occurring
at the same Reynolds numbers are similar, in the sense that one may
be deduced from the other by rescaling. Arguments leading us to this
conclusion have been made with reference to a flow that is bounded by a
solid surface over which the no-slip and no-penetration boundary condi-
tions are required. Moreover, a time-independent velocity was imposed
in the far-field boundary condition in lieu of a driving mechanism.
If the driving mechanism is time-dependent, or if the flow is bounded
by fluid interfaces and free surfaces, additional conditions for dynamic
similitude requiring the equality of further dimensionless numbers are
required. These dimensionless numbers enter the problem formulation
either through the governing equations or through boundary and inter-
facial conditions.

Frequency parameter for a time-dependent flow

An externally imposed time-dependent flow with a time scale T is

characterized by the dimensionless frequency parameter

R_ L*
^=TV (6.7.28)
where v is the kinematic viscosity. In the case of periodic flow with
angular frequency u due, for example, to an oscillating pressure gradient,
T may be identified with the period T = 2?r/a;.

Froude number

The relative importance of inertial and gravitational forces in a flow

bounded by a free surface, such as the flow due to the propagation of
water waves in the ocean, is determined by the Froude number

Fr = -?-,
V^' (6.7.29)
where g is the acceleration of gravity. In the case of flow over a hump
discussed in Section 6.4, the Froude number takes the specific form shown
in equation (6.4.22).

Bond number

The relative importance of gravitational forces and surface tension in

a fluid bounded by a free surface or interface is determined by the Bond
number
P9L
Bo-
Bo
=~^~i
" (6.7.30)

where 7 is the surface tension (problem 6.7.3).

Weber number

The relative importance of inertial forces and surface tension in a

fluid bounded by a free surface or interface is determined by the Weber
number
H7
We ?U2L
= —7-- (6.7.31)
For example, the Weber number determines the deformation and nature
of the flow around a bubble rising or convected at high speed through
an ambient liquid.

Problems

Problem 6.7.1 Characteristic scales.

Identify the characteristic velocity scale /7, length scale L, and the
Reynolds number of (a) simple shear flow past a stationary particle, (b)
flow due to the settling of a particle in the atmosphere, and (c) flow due
to a breaking wave in the ocean.
Problem 6.7.2 Reynolds number.
Compute the Reynolds number of (a) an ant crawling, (b) a person
running, (c) a car moving at 100 km per hour, and (d) an elephant
running across a plain at maximum speed.
Problem 6.7.3 Bond number in hydrostatics.
Explain how the Bond number emerges from the scaling of the Laplace-
Young equation (5.4.6) in hydrostatics.