of

surface tension of the interface. For a two-dimensional surface deflection with η = η(x)
only, Eq. (1-86) becomes

(1-87)

We see from this relation that, when the interface smiles (concave upward, positive
curvature), p > pa, while a frowning interface (concave downward) results in p > pa.
Equations (1-85) and (1-86) represent complex nonlinear conditions, but they can be
evaluated numerically.

In the range 0 to 100°C, a clean air–water interface has a nearly linear variation of
surface tension with temperature:

(1-88)
with accuracy of ± 1 percent. Measured values of air–water surface tension are given in Table
1-4 for temperatures up to the critical point. These are ideal for a very clean interface. Under
field conditions, ℑ can vary greatly due to the presence of surface contaminants or slicks.

TABLE 1-4
Surface-tension coefficient for an air–water interface

†Critical point.

In large-scale problems, such as open-channel or river flow, the free surface deforms only
slightly and surface-tension effects become negligible. Equations (1-85) and (1-86) then
simplify into

(1-89)

These are obviously very attractive linearized conditions.

Note: The present discussion concerns the deformation of an interface by uniform surface
tension. If the surface tension varies along the interface due, for example, to a temperature
gradient, a flow called Marangoni convection will be induced from the hot surface toward
the cold surface. This mechanism is beyond the scope of the present text. For an example of
Marangoni flow analysis, see the paper by Sasmal and Hochstein (1994) or the reviews by
Ostrach (1982) and Davis (1987).

1-4.5 Conditions at a Deformable Fluid–Solid Interface

In most of the problems of this text, a solid boundary will be taken to be a rigid interface that
merely imposes no-slip and no-temperature-jump conditions on the fluid. The dynamics and
thermodynamics of the solid are neglected. However, there is a growing field of research in
which the solid is coupled to the fluid through deformable and dynamic interactions. In this
case, we must match the velocity, stress, temperature, and heat transfer across the interface,
and the equations of motion for both the fluid and the solid must be solved simultaneously.
Material–behavior Page 31relations can be used when the fluid is Newtonian and the solid
satisfies Hooke’s law. For details, see the texts by Blevins (2001) and Au-Yang (2001).

SUMMARY
This chapter overviews the introductory concepts of fluid motion with which the reader
should already be conversant. A brief history and some sample viscous-flow problems are
outlined, followed by an extensive discussion of the different quantities that distinguish a
fluid: (1) the kinematic properties, (2) the transport properties, and (3) the thermodynamic
properties. The chapter closes with a cursory look at various boundary conditions relevant to
viscous-fluid flow.

PROBLEMS
1-1. A sphere 1.4 cm in diameter is placed in a freestream of 18 m/s at 20°C and 1 atm.
Compute the diameter Reynolds ​number of the sphere if the fluid is (a) air, (b) water,
and (c) hydrogen.

1-2. A telephone wire 8 mm in diameter is subjected to a crossflow wind and begins to

shed vortices. From Fig. 1-7, what wind velocity (in m/s) will cause the wire to “sing”
at middle C (256 Hz)?

1-3. If the wire in Prob. 1-2 is subjected to a crossflow wind of 12 m/s, use Fig. 1-8 to
estimate its drag force (in N/m).

1-4. For oil flow in a pipe far downstream of the entrance (Figs. 1-9 and 1-10), the axial
velocity profile is a function of r only and is given by u = (C/μ) (R2 − r2), where C is
a constant and R is the pipe radius. Suppose the pipe is 1 cm in diameter and umax =
 30 m/s. Compute the wall shear stress (in Pa) if μ = 0.3 kg/(m ⋅ s).

1-5. Glycerin at 20°C is confined between two large parallel plates. One plate is fixed
and the other moves parallel at 17 mm/s. The distance between plates is 3 mm.
Assuming no-slip, estimate the shear stress in the glycerin, in Pa.

1-6. A plane unsteady viscous flow is given in polar coordinates by

where C is a constant and ν is the kinematic viscosity. Compute the vorticity ωz (r, t)
and sketch an array of representative velocity and vorticity profiles for various times.

1-7. A two-dimensional unsteady flow has the velocity components:

Find the equation of the streamlines of this flow which pass through the point (x0, y0)
at time t = 0.

1-8. Using Eq. (1-3) for inviscid flow past a cylinder, consider the flow along the
streamline approaching the forward stagnation point (r, θ) = (R, π). Compute (a) the
distribution of strain rates εrr and εrθ along this streamline and (b) the time required
for a particle to move from the point (2R, π) to the stagnation point.

1-9. A commonly used equation of state for water is approximately independent of

temperature:

where A ≈ 3000, n ≈ 7, p0 = 1 atm, and ρ0 = 998 kg/m3. From this formula, compute
(a) the pressure (in atm) required to double the density of water, (b) the bulk modulus
of water at 1 atm, and (c) the speed of sound in water at 1 atm.

1-10. As shown in Fig. P1-10, a 3 × 3-ft plate slides down a long 30° incline on which
there is a film of oil 0.005 in. thick with viscosity μ = 0.0005 slug/(ft ⋅ s). Assuming
that the plate does not deform the oil film, estimate (a) the terminal sliding velocity
(in ft/s) and (b) the time required for the plate to accelerate from rest to 99 percent of
the terminal velocity.
 FIGURE P1.10

Page 32

1-11. Estimate the viscosity of nitrogen at 36 MPa and 49°C, and then compare with the
measured value of 45 μPa ⋅ s.

1-12. Estimate the thermal conductivity of helium at 420° C, 1 atm, and compare with
the measured value of 0.28 W(m ⋅ K).

1-13. According to Table C-5 and Fig. 1-15, at what pressure is the viscosity of CO2
equal to approximately 30 × 10−5 Pa ⋅ s when the temperature is 800°R?

1-14. Some measured values for the viscosity of ammonia gas are as follows:

Fit these data measurements, in the least-square-error sense, to the power law, Eq. (1-
33), and the Sutherland law, Eq. (1-34).

1-15. Experimental data for the viscosity of helium at low pressure are as follows:

Fit these values to a suitable formula.

1-16. Repeat the analysis of the velocity profile between two plates (Fig. 1-13) for a
Newtonian fluid but allow for a slip velocity δu ≈ ℓ(du/dy) at both walls. Compute the
shear stress at both walls. [τw = μV/(h + 2ℓ ) at both walls.]

1-17. By considering the equilibrium of forces on the element shown in Fig. P1-17,
 derive Eq. (1-86), which expresses the pressure jump across a curved surface due to
surface tension.

FIGURE P1.17

1-18. Two spherical bubbles of radii R1 and R2, which contain air, coalesce into a single
bubble of radius R3. If the ambient air pressure is p0 and the merging process is
isothermal, derive a formula that relates R3 to ( p0, R1, R2, ℑ ).

1-19. In Prob. 1-1, if the temperature, sphere size, and velocity remain the same for
airflow, at what air pressure will the Reynolds number ReD be equal to 10,000?

1-20. A solid cylinder of mass m, radius R, and length L falls concentrically through a
vertical tube of radius R + ΔR, where ΔR ≪ R. The tube is filled with gas of viscosity
μ and mean free path ℓ. Neglect fluid forces on the front and back faces of the
cylinder and consider only shear stress in the annular region, assuming a linear
velocity profile. Find an analytic expression for the terminal velocity of fall, V, of the
cylinder (a) for no slip and (b) with slip [Eq. (1-80)].

1-21. Solve P1-20 for the terminal fall velocity for no slip if the cylinder is aluminum,
with diameter 4 cm and length 10 cm. The tube has a diameter of 4.02 cm and is filled
with argon gas at 20°C.

1-22. In Fig. P1-22 a disk rotates steadily inside a disk-shaped container filled with oil
of viscosity μ. Assume linear velocity profiles with no slip and neglect stress on the
outer edges of the disk. Find a formula for the torque M required to drive the disk.
 FIGURE P1.22

Page 33

1-23. Based on Eq. (1-76), it can be shown that the coefficient of thermal expansion of a
perfect gas is given by β = 1/T. Use this approximation to estimate β of ammonia gas
(NH3) at 20°C and 1 atm and then compare with the accepted value from a data
reference.

1-24. The rotating-cylinder viscometer in Fig. P1-24 shears the fluid in a narrow
clearance Δr, as shown. Assuming a linear velocity distribution in the gaps, if the
driving torque M is measured, find an expression for μ by (a) neglecting and (b)
including the bottom friction.

FIGURE P1.24

1-25. Consider 1 m3 of a fluid at 20°C and 1 atm. For an isothermal process, calculate
the final density and the energy, in joules, required to compress the fluid until the
pressure is 10 atm for (a) air and (b) water. Discuss the difference in results.
1-26. Equal layers of two immiscible fluids are being sheared between a moving and a
fixed plate, as in Fig. P1-26. Assuming linear velocity profiles, find an expression for
the interface velocity U as a function of V, μ1, and μ2.

FIGURE P1.26

1-27. Use the inviscid-flow solution of flow past a cylinder, Eq. (1-3), to (a) find the
location and value of the maximum fluid acceleration along the cylinder surface. Is
your result valid for gases and liquids? (b) Apply your formula for amax to airflow at
10 m/s past a cylinder of diameter 1 cm and express your result as a ratio compared to
the acceleration of gravity. ​Discuss what your result implies about the ability of fluids
to withstand acceleration.

1-28. The coefficient of thermal expansion is defined as

Determine β for an ideal gas with p = ρRT. Show your work in detail.

1-29. Starting with Maxwell’s low-density approximation of the viscosity, namely,

and Newton’s expression of the wall shear stress as a function of the
velocity gradient, τw = μ(∂u/∂y)w, express Maxwell’s slip velocity, uw = (∂u/∂y)w,
(a) as a function of the shear stress, density, and speed of sound a;
(b) as a function of the Mach number, the mean-flow velocity U, and
the skin friction coefficient, Cf = 2 τw/(ρU2). Page 34

1-30. Consider a hydraulic lift with a 50 cm diameter shaft sliding inside a housing with
an inside diameter of 50.02 cm. If the shaft travels at 0.25 ms, calculate the shaft
resistance to motion per unit length. You may use water as the working fluid.

1-31. Consider a thin air gap of 1 mm that is formed between two parallel surfaces that
are maintained at 20°C and 40°C, respectively. In the case of a quiescent medium (say
 still air), calculate the heat transfer rate across the gap per unit area.

1-32. In the presence of viscosity, the pressure drop associated with a fully developed
laminar motion in a horizontal tube of length L and diameter D may be evaluated
analytically. One finds:

where μ stands for the dynamic viscosity and denotes the volumetric flow
rate. Show that the corresponding head loss may be written as

What value of flam do you obtain?

1-33. A time-dependent, two-dimensional motion has three velocity components that are
given by

where a and b are pure constants. The objective of this problem is to compare and
contrast the streamlines in this flow with the pathlines of the fluid particles.
(a) Find the equations governing the streamline that passes through the
point (1, 1) at time t.
(b) Calculate the path of a particle that starts at r0 = (x0, y0) = (1, 1) at t
= 0. Determine the location of a particle at t = 1, denoted as r1.
(c) Use the results of part (a) to determine the condition under which
the streamlines and pathlines coincide.
1-34. A tornado may be simulated as a two-part circulating flow in cylindrical
coordinates, with υr = υz = 0,

(a) Calculate the divergence of the velocity. Is the flow compressible or

incompressible?
(b) Determine the vorticity. Is the flow rotational or irrotational?
(c) Determine the strain rates in each segment of the flow. What is the
sum of the three normal strain rates?
 In modeling the motion of an 8-meter diameter tornado rotating at an angular speed of
1-35. ω at the point of maximum swirl, it is possible to use the Maicke–Majdalani
profile (Maicke and Majdalani 2009) as a piecewise approximation for which υr = υz
= 0 and the tangential velocity is given by

(a) State whether the flow is 1D, 2D, or 3D, steady or unsteady, and
specify υθ (r) as r → ∞.
(b) Calculate the divergence of the velocity. Is the flow compressible or
incompressible?
(c) Determine the vorticity. Is the flow rotational or irrotational?
(d) Determine the strain rates and the shear stresses in the inner and
outer flow segments.
(e) What is the limit of υθ (r) as r → 0 ? Hint: In taking the limit, it is
helpful to remember that and that, in the inner segment, the
tangential velocity can be rewritten as

1-36. The Taylor profile, which has been used to describe the bulk gaseous motion in
planar, slab rocket chambers (Maicke and Majdalani 2008), corresponds to a self-
similar profile in porous channels that bears symmetry with respect to the chamber’s
midsection plane. Using normalized Cartesian coordinates, the streamfunction may be
written as where l represents the aspect ratio of the
chamber (i.e., the length of the porous surface normalized by the chamber half
height). In this problem, the velocity vector, normalized by the wall injection speed,
may be expressed as V(x, y) = u i + υ j.
(a) Determine the axial and normal velocity profiles using:

Page 35

(b) Evaluate the velocity divergence and determine whether the motion
is compressible or incompressible.
(c) Evaluate the vorticity and determine whether the motion is
rotational or irrotational.
 FIGURE P1.36

1-37. In the Taylor flow problem described above, determine the following:
(a) The strain rates.
(b) The Lagrangian time-dependent coordinates xj (t) and yj (t) for a
particle j that enters the porous chamber at the sidewall where yj = 1
and xj = Xj at t = 0. Recall that

(c) The pathlines of a particle j entering the chamber at Xj = 1, 2, 3, 4,

5. Display your results in the (x, y) plane.Page 36
1-38. The Taylor–Culick profile, which describes the bulk gaseous motion in solid
rocket motors (Culick 1966), corresponds to an axisymmetric, self-similar profile in
porous tubes. Using normalized cylindrical coordinates, the streamfunction may be
written as where l represents the aspect ratio of the
motor (i.e., the length of the porous surface normalized by the chamber radius). In
this problem, the velocity vector, normalized by the wall injection speed, may be
expressed as V(r, z) = υr er + υz ez.
(a) Determine the axial and radial velocity profiles using Stokes’
definition:

(b) Evaluate the velocity divergence and determine whether the motion
is compressible or incompressible.
(c) Evaluate the vorticity and determine whether the motion is
 rotational or irrotational.

FIGURE P1.38

1-39. In the Taylor–Culick flow problem described above, determine the following:
(a) The strain rates.
(b) The Lagrangian time-dependent coordinates rj (t) and zj (t) for a
particle j that enters the cylindrical rocket chamber at the sidewall
where rj = 1 and zj = Zj at t = 0. Recall that

(c) The pathlines of a particle j entering the chamber at Zj = 1, 2, 3, 4,

5. Display your results in the (r, z) plane.
1-40. The Vyas–Majdalani profile, which describes the helical motion of a cyclonic
chamber (Vyas and Majdalani 2006), consists of a three-component velocity profile,
V(r, z) = υr er + υθ eθ + υz ez, where

Here a denotes the radius of the cyclonic chamber, U stands for the average tangential
velocity at r = a, and κ represents a dimensionless offset swirl parameter that gauges
the relative importance of axial and tangential speeds.
(a) Is the flow one-dimensional, two-dimensional, or three-
dimensional?
(b) Is the flow steady or unsteady?
(c) Calculate the divergence of the velocity. Is the flow compressible or
 incompressible?
(d) Determine the vorticity. Is the flow rotational or irrotational?
(e) Determine the strain rates. What is the sum of the three normal
strain rates?
(f) Assuming a circular opening of radius calculate the volumetric
flow rate by integrating the axial velocity from

FIGURE P1.40

1-41. The Maicke–Majdalani profile, which may be used to describe the motion of an
unbounded tornado (Maicke and Majdalani 2009), can be expressed as a simple
piecewise approximation for which υr = υz = 0 and

where R denotes the radius at which the wind’s angular speed ω may be measured and
X represents the fraction of the radius R where the free vortex behavior ceases.
(a) Is this flow one-dimensional, two-dimensional, or three-
dimensional?
(b) Is this flow steady or unsteady?
(c) Calculate the divergence of the velocity. Is the flow compressible or
incompressible?
(d ) Determine the vorticity. Is the flow rotational or irrotational?
(e) Determine the strain rates and the shear stresses in each segment of
 the flow.
(f) What is the limiting value of υθ (r) as r → 0 ?

1-42. Assuming ϕ to be a scalar and V a vector, evaluate the following quantities:

(a) ∇ × ∇ ϕ
(b) (∇ ϕ) × (∇ ϕ)
(c) (∇ ϕ) ⋅ (∇ ϕ)
(d) ∇ ⋅ (∇ ϕ)
(e) ∇ ⋅ (∇ × V)
Page 37

1-43. Assuming U, V, and W to be Cartesian vectors, prove the following identities:

†Note, however, that flows involving chemical or nuclear reactions require an extended concept of equilibrium. Such flows
typically involve knowledge of reaction rates and are not treated here.

‡A solid has two sound speeds: (K/ρ)1/2 is the dilatation or longitudinal-wave speed, and (G/ρ)1/2 is the rotational or shear-
wave speed. Fluids have only one sound speed.
Page 38

CHAPTER
2

FUNDAMENTAL EQUATIONS OF COMPRESSIBLE

VISCOUS FLOW

The more we learn of science, the more we see that its wonderful mysteries are all explained
by a few simple laws so connected together and so dependent upon each other, that we see
the same mind animating them all.

Olympia Brown (1835–1900)

2-1 INTRODUCTION

THE EQUATIONS of viscous flow have been known for more than 100 years. In their
complete form, these equations are difficult to solve, even on modern digital computers.
In fact, at high Reynolds numbers (under turbulent-flow conditions), the equations are, in
effect, impossible to solve with present mathematical techniques because the boundary
conditions become randomly time dependent. Nevertheless, it is very instructive to derive
and discuss these fundamental equations because they provide many insights, yield several
particular solutions, and can be reduced according to several modeling laws. Moreover, these
exact equations can be simplified, using Prandtl’s boundary-layer approximations, as it will
be later shown. The resulting simpler system is very practical and yields many fruitful
engineering solutions.

2-2 CLASSIFICATION OF THE FUNDAMENTAL

EQUATIONS
The basic equations considered here represent the three conservation principles for physical
systems:
1. Conservation of mass (continuity).
2. Conservation of momentum (Newton’s second law).
3. Conservation of energy (first law of thermodynamics).

The three unknowns that must be obtained simultaneously from these three basic equations
are the velocity V, the thermodynamic pressure p, and the absolute temperature T. We
consider p and T to be the two required independent thermodynamic variables. However, the
final forms of the conservation equations also contain four other thermodynamic variables:
the density ρ, the enthalpy h (or the internal energy e), and the two transport properties μ and
k. Using our tacit assumption of local thermodynamic equilibrium, the latter four properties
are uniquely determined by the values of p and T. Thus, the system is completed by
specifying four state relations

(2-1)
which can be provided in the form of tables, charts, or semi-theoretical formulas based on
kinetic theory. Many useful analyses simply assume that ρ, μ, and k are constants and that h is
proportional to T through h = cpT.

Finally, to specify a particular problem completely, we must impose conditions on V, p,

and T at every point of the boundary of the flow regime.
The preceding considerations apply to a fluid of assumed uniform, homogeneous
composition, i.e., where ​diffusion and chemical reactions are not considered.
Multicomponent reacting fluids must consider at least two more relations:

4. Conservation of species.
5. Chemical reaction laws.Page 39

Additional relations that must be considered consist of the diffusion coefficient D = D(p, T),
chemical-equilibrium constants, reaction rates, and heats of formation. This text does not
consider reacting boundary-layer flows [see Kee et al. (2003)].

Finally, even more relations are necessary if one considers the flow to be influenced by
electromagnetic effects. This is the subject of the field of magnetohydrodynamics. Such
effects are not considered in the present text either.
Instead, let us proceed to derive the three basic equations of a single-component fluid,
bearing in mind that the results will also apply to uniform nonreacting mixtures, such as air
or liquid solutions.†

2-3 CONSERVATION OF MASS: THE EQUATION OF

CONTINUITY
As mentioned in the discussion of Eq. (1-5), all three conservation laws are Lagrangian in
nature; i.e., they apply to a fixed system of particles. Thus, in the Eulerian system appropriate
to fluid flow, our three laws rely on the particle derivative

(1-8)

which is a formidable expression. In Lagrangian terms, the law of conservation of mass is

surpassingly simple:

(2-2)
where ϑ is the volume of a particle. In Eulerian terms, this is equivalent to

(2-3)

We can relate Dϑ/Dt to the fluid velocity by noting that the total dilatation or normal-strain
rate is equal to the ​relative time rate of change of the volume of a particle:

(2-4)

Further, we can substitute for the strain rates from our kinematic relations in Eqs. (1-21) to
retrieve:

(2-5)

Combining Eqs. (2-3) through (2-5) to eliminate ϑ, we obtain the equation for continuity in
its most general form,

(2-6)

If the fluid is incompressible, Eq. (2-6) reduces to the simpler condition for a solenoidal
velocity, namely,

(2-7)
which is equivalent to requiring that particles retain a constant volume.‡ It may be instructive
to note that although all constant density fluids are incompressible, the converse is not true. It
is possible to have incompressible fluids with varying densities. A clear example is salt water
which maintains a constant volume although its density varies with depth.

2-3.1 A Useful Strategy: The Stream Function

If the continuity Eq. (2-6) reduces to only two nonzero terms, it can be satisfied identically,
and therefore replaced, by the so-called stream function ψ. The idea dates back to the French
mathematician Joseph-Louis Lagrange in 1755. The Cartesian incompressible-flow stream
function is treated in detail in Sec. 2-11. Here, we illustrate its properties using a steady, two-
dimensional compressible flow in the xy plane, for which the continuity equation reduces to

(2-8)

Page 40

If we now define the stream function ψ such that

(2-9)

we see by direct substitution that Eq. (2-8) is satisfied identically, assuming of course that ψ
is continuous to the second-order derivative. Thus, the continuity equation can be discarded
and the number of dependent variables reduced by one. The penalty, however, is that the
remaining velocity derivatives are increased by one order.

The stream function is not only useful but has physical significance:

(2-10)

which means that a change in the stream function represents a change in the mass flow rate.
Moreover, lines of constant ψ, along which dψ = 0, represent lines across which there is no
mass flow (dṁ = 0); that is, they are streamlines of the flow. Also, the difference between the
values of ψ of any two streamlines is numerically equal to the mass flow between those
streamlines.

2- 4 CONSERVATION OF MOMENTUM: THE NAVIER–

STOKES EQUATIONS
This relation, commonly known as Newton’s second law, expresses a proportionality between
an applied force and the resulting acceleration of a particle of mass m:

(2-11)
where a = F/m reminds us that the acceleration represents a force per unit mass. If the system
is a fluid particle, it is convenient to divide Eq. (2-11) by the volume of the particle, so that
the analysis is carried out in terms of the fluid density instead of the particle mass. It is also
traditional to reverse the terms and place the acceleration on the left-hand side such that

(2-12)

where f is the applied force per unit volume on the fluid particle. Note that in our chosen
Eulerian system, the acceleration is the rather complicated particle derivative from Eq. (1-9)
and f is divided into two types: surface forces and body forces.

The so-called body forces are those that apply from a distance to the entire mass of the
fluid element. Such forces are usually due to external fields such as gravity or an applied
electromagnetic potential. We ignore magnetohydrodynamic effects here and consider only
the gravitational body force per unit volume,

(2-13)
where g is the gravitational acceleration vector.

The surface forces are those applied by external stresses on the sides of the element. The
quantity stress σij is a tensor, just as the strain rate ϵij in Sec. 1-3. The sign convention for
stress components on a Cartesian element is shown in Fig. 2-1, where all stresses are
positive. The stress tensor, which comprises both viscous and surface pressure effects, can be
written as

FIGURE 2-1
Notation for stresses.

(2-14)

where the viscous stress tensor may be represented by the nine components
and δij stands for the Kronecker delta function (δij = 1 if i = j and δij = 0 if i ≠ j as described in
App. D). Note that, unlike the viscous stresses, the pressure can be exerted only normally to
the surface of the fluid element. Following the strain rate ϵij, σij, and τij form symmetric
tensors such that σij = σji and τij = τji; i ≠ j. This symmetry is required to maintain the
equilibrium of moments about the three axes of the element.§Page 41

The positions of the stresses in the array of Eq. (2-14) are not arbitrary; the rows
correspond to applied forces in each coordinate direction. Considering the front faces of the
element in Fig. 2-1, the total force in each direction due to stress is

(2-15)

For an element in equilibrium, these forces would be balanced by equal and opposite forces
on the back faces of the element. If the element is accelerating, however, the front- and back-
face stresses will be different by differential amounts. For example,

(2-16)

Hence, the net force on the element in the x direction, for example, will be due to three
derivative terms:

or, on a unit volume basis, dividing by dx dy dz, since σij = σji, we get

(2-17)

which we note is equivalent to taking the divergence of the vector (σxx, σxy, σxz). Similarly, fy
and fz are the divergences of the second and third rows of σij. Recalling that

it may be seen that Eq. (2-17) is simply f Along similar lines, the
total vector surface force may be expressed as
 (2-18)

Page 42where the divergence of σij is to be interpreted in the tensor sense, so that the result is
a vector.

Newton’s law, Eq. (2-12), now becomes:

(2-19)

and it remains only to express σij in terms of the velocity V. This is done by relating τij to ϵij
through the assumption of some viscous deformation-rate law, e.g, for a Newtonian fluid.**

2-4.1 The Fluid at Rest: Hydrostatics

From the definition of a fluid (Sec. 1-3), viscous stresses vanish if the fluid is at rest. The
velocity and shear stresses are zero, and the normal stresses become limited to the hydrostatic
pressure contribution. Equation (2-19) reduces to the hydrostatic relation if V = 0:

(2-20)

If we take the z coordinate as up and assume ρ and g are constants, the pressure will vary
linearly with z, according to δp = −ρgδz. As such, pressure will increase downwardly and
proportionately to the specific weight of the fluid, as covered in the hydrostatics
subdiscipline, e.g., White (2016). Here, we must ensure that our dynamic momentum
equation reduces to Eq. (2-20) when V = 0.

2-4.2 Deformation Law for a Newtonian Fluid

By analogy with Hookean elasticity, the simplest assumption for the variation of viscous
stress with strain rate is a linear law. These considerations were first made by Stokes (1845)
and, as far as we know, the resulting deformation law is satisfied by all gases and most
common fluids. Stokes’ three postulates are

1. The fluid is continuous, and its stress tensor σij is at most a linear function of the strain
rates ϵij.
2. The fluid is isotropic, i.e., its properties are independent of direction, and therefore the
deformation law is independent of the coordinate axes in which it is expressed.
3. When the strain rates are zero, the deformation law must reduce to the hydrostatic
 pressure condition, σij = − pδij.

Note that the isotropic condition 2 requires that the principal stress axes be identical to the
principal strain-rate axes. This makes the principal axes a convenient place to begin the
deformation-law derivation. Let x1, y1, and z1 be the principal axes, where the shear stresses
and shear strain rates vanish [see Eq. (1-24)]. With these axes, the deformation law could
involve at most three linear coefficients, C1, C2, and C3. For example,

(2-21)
The term −p is added to satisfy the hydrostatic condition 3 above. But the isotropic condition
2 requires that the crossflow effect of ϵ22 and ϵ33 be identical, i.e., that C2 = C3. Therefore,
there are really only two independent linear coefficients in an anisotropic Newtonian fluid.
We can rewrite Eq. (2-21) in the simpler form

(2-22)
where K = C1 − C2, for convenience. Note also that ϵ11 + ϵ22 + ϵ33 = ∇ ⋅ V from Eq. (2-5).

Now let us transform Eq. (2-22) to some arbitrary axes x, y, z, where shear stresses are
not zero, and thereby find an expression for the general deformation law. With respect to the
principal axes x1, y1, z1, let the x axis have Page 43direction cosines l1, m1, n1, let the y axis
have l2, m2, n2, and let the z axis have l3, m3, n3. Recall that for any set of
direction cosines. Then, the transformation rule between a normal stress or strain rate in the
new system and the principal stresses or strain rates is given by, for example,

(2-23)

Similarly, the shear stresses (strain rates) may be related to the principal stresses (strain rates)
by the following transformation law:

(2-24)

We can now eliminate σ11, ϵ11, σ22, etc., from Eq. (2-23) by using the principal-axis
deformation law, Eq. (2-22), and the fact that l2 + m2 + n2 = 1. The result is

(2-25)
with exactly similar expressions for σyy and σzz. Similarly, we can eliminate σ11, ϵ11, etc.,
from Eq. (2-24) to retrieve

(2-26)
and precisely analogous expressions for xz and σyz. Note that the direction cosines have all
vanished systematically. Equations (2-25) and (2-26) are, in effect, the desired general
deformation law. By comparing Eq. (2-26) and Eq. (1-30) for shear flow between parallel
plates, we see that the linear coefficient K is equal to twice the ordinary coefficient of
viscosity, K = 2μ. The coefficient C2 is new and independent of μ and may be called the
second coefficient of viscosity. In linear elasticity, C2 is called Lamé’s constant and is given
the symbol λ, which is also adopted here. Since λ is associated only with volume expansion, it
is customary to call it the coefficient of bulk viscosity, λ.

Equations (2-25) and (2-26) can be combined and rewritten as a single general
deformation law for a viscous Newtonian fluid. We get

(2-27)
where we have written ϵij in terms of the velocity gradients from Eq. (1-27). This
deformation law was first given by Stokes (1845) and assumes knowledge of the geometric
rules for coordinate transformation of stresses and strain rates. The interested reader may
refer to the textbooks on continuum mechanics by Lai et al. (1995), Malvern (1997), or
Talpaert (2003). The scalar forms of τij are provided in App. B for incompressible flow using
both Cartesian and cylindrical coordinates.

2-4.3 Thermodynamic Pressure versus Mechanical Pressure

Stokes (1845) pointed out an interesting consequence of Eq. (2-27). By analogy with the
strain relation [Eqs. (1-25)], the sum of the three normal stresses σxx + σyy + σzz leads to an
invariant tensor. In this context, the mechanical pressure is defined as the negative one-
third of this sum, i.e., the average compression stress on the element. Then, by ​summing Eq.
(2-27), we obtain

(2-28)

where p represents the thermodynamic pressure. Thus, the mean pressure in a deforming
viscous fluid is not equal to the thermodynamic property called pressure. This distinction is
rarely important, since div V is usually very small in typical flow problems, but the precise
meaning of Eq. (2-28) has been a controversial subject for more than a century. Stokes
himself simply resolves the issue using an assumption that is as good as any other, namely,
by setting

Taking the second coefficient of viscosity to be simply eliminates the discrepancy

between the mechanical and thermodynamic pressures. However, the available experimental
evidence from the measurement of sound-wave attenuation, as reviewed by Karim and
Rosenhead (1952), suggests that λ for most liquids is actually positive, rather Page 44than
−2μ/3, and is often much larger than μ. However, the experiments themselves have been a
matter of some ​controversy [Truesdell (1954)].

A second type of flow condition will also make equal to p regardless of the value of λ,
specifically

Incompressible flow:

As such, the bulk viscosity cannot affect a truly incompressible fluid, but in fact, it does
affect certain phenomena occurring in nearly incompressible fluids, e.g., sound absorption in
liquids. Meanwhile, if ∇ ⋅ V ≠ 0 (e.g., compressible flow), we may still avoid the problem if
viscous normal stresses are negligible. This is the case in boundary-layer flows of
compressible fluids, for which only the first coefficient of viscosity μ proves to be important.
However, the normal shock wave problem represents a case where the coefficient λ cannot be
neglected. The second such case is the previously mentioned problem of sound-wave
absorption and attenuation.

It appears, then, that the second viscosity coefficient is still a controversial quantity. In
fact, λ may not even be a thermodynamic property since it is apparently frequency dependent.
Fortunately, the disputed term, λ ∇ ⋅ V, is ​typically so very small that the effect of λ may be
ignored altogether. An interesting discussion of the second ​coefficient of viscosity may be
found in Landau and Lifshitz (1959, Sec. 78) and Panton (1996).

2-4.4 The Navier–Stokes Equations

The desired momentum equation for a general linear (Newtonian) viscous fluid is obtained
by substituting the stress relations, Eq. (2-27), into Newton’s law [Eq. (2-19)]. The result is
the famous equation of motion that bears the names of Navier and Stokes. In fact, this
equation brackets the works of four scientists: Navier (1827), Poisson (1831), Saint Venant
(1843), and Stokes (1845). In scalar form, we obtain

(2-28a)

These are the Navier–Stokes equations, fundamental to the subject of viscous fluid flow.
Considerable economy is achieved by rewriting them as a single-vector equation, with and
without the indicial notation (App. D):

(2-29b)

As usual, the indicial form may be further compacted into

2-4.5 Incompressible Flow: Thermal Decoupling

If the fluid is assumed to be of constant density, div V vanishes due to continuity [Eq. (2-7)],
and the vexing bulk coefficient λ disappears from Newton’s law. Equations (2-29a) and (2-
29b) are not greatly simplified, though, if the first viscosity μ is still permitted to vary with
temperature and pressure (and hence with position). If, however, we assume that μ is
spatially uniform, many terms vanish, leaving us with a much simpler Navier–Stokes
equation for an incompressible fluid with constant viscosity:

(2-30)

Page 45Most of the problems in this text correspond to incompressible flows that can be
solved using Eqs. (2-7) and (2-30). Note that it is sufficient for ρ and μ to be constant for
these equations to become independent of the temperature. One may then solve the continuity
and momentum equations for the velocity and pressure, and later solve for the temperature
from the thermally decoupled energy equation described in Sec. 2-5. This approximation
often divides textbooks into those devoted to the discipline of “fluid mechanics” and those of
“heat transfer.” Despite the main focus of the present text on fluid dynamical concepts, it will
maintain a limited heat-transfer coverage throughout.

2-4.6 Steady Inviscid Flow: The Euler and Bernoulli Equations

If we further assume that viscous terms are negligible, Eq. (2-30) reduces to

(2.30a)

This is called Euler’s equation (derived by Leonhard Euler in 1755) for inviscid flow. It is
first order in V and p and thereby simpler than the second-order Navier–Stokes Eq. (2-30). In
the absence of viscosity, the no-slip condition must be relaxed at a fixed wall, and the
tangential velocity can be allowed to slip. Much research has been reported for Euler’s
equation: analytical (Currie 1993), numerical (Tannehill et al. 1997), and mathematical
theorems (Kreiss and Lorenz 1989). As shown in several undergraduate texts (e.g., White
2016), Euler’s equation for steady, incompressible, frictionless motion may be integrated
along a streamline between any two points 1 and 2 to yield

(2.30b)

where z is vertical upward. This is Bernoulli’s equation for steady frictionless flow. Though
approximate, since all fluids are viscous, it has many applications in aeronautics and
hydrodynamics and serves as an outer boundary matching condition in boundary-layer theory
(Chap. 4). The unsteady form of Bernoulli’s equation will be given in Sec. 2-10.

Bernoulli’s equation stands at the basis of streamline analysis and can be rewritten in
units of length, which is a more convenient form to use when analyzing problems with
irreversible losses. The latter are often lumped into an energy head loss term, hL, which
reduces the total amount of energy left in the stream at the downstream location. Thus, going
from point 1 to 2, one can write

(2.30c)

where the head loss may be determined using the procedures described in White (2016) and
summarized in App. E on head loss calculations.

2-4.7 Steady Inviscid Axisymmetric Flow: The Bragg–Hawthorne Equation

In the treatment of steady, inviscid, and incompressible flow where the motion is
axisymmetric, Euler’s equation reduces to a compact, scalar, second-order partial differential
equation in the stream function. Known as the Bragg–Hawthorne equation (Bragg and
Hawthorne 1950), this relation represents the projection of Euler’s steady-state equation
along the axis of rotation.

(2.30d)

Page 46At this stage, the incompressible Stokes stream function may be invoked to
replace the radial and axial velocities, namely, by letting As for the
tangential velocity υθ, it may be expressed in terms of the angular momentum by recognizing
that the tangential momentum equation for this axisymmetric case reduces to

The requirement for a vanishing material derivative of the angular momentum enables us to
write rυθ = B(ψ), which must be conserved along a streamline. In fact, 2πB, which represents
the flow circulation, as well as the total head H, cannot change along a streamline. Forthwith,
backward substitution into Eq. (2-30d) leads to

which may be readily rearranged and simplified into

This second-order partial differential equation in ψ (r, z), also referred to as the Hicks
equation (Hicks 1898) or the Long–Squire equation (Long 1953; Squire 1956), proves to be
extraordinarily important in the treatment of axisymmetric and swirl-dominated flows,
particularly, those arising in cyclone separators and wall-bounded vortex rocket engines
(Bloor and Ingham 1987; Vyas and Majdalani 2006; Majdalani 2012; Maicke, Cecil, and
Majdalani 2017). Although an equivalent form does not conveniently exist in Cartesian
coordinates, a spherical analog with several important applications is available, as shown in
App. F (e.g., Majdalani and Rienstra 2007; Barber and Majdalani 2017).

2-5 THE ENERGY EQUATION (FIRST LAW OF

THERMODYNAMICS)
The first law of thermodynamics for a system states that the change in system energy is
simply the sum of the work and heat added to the system:

(2-31)
The quantity Et denotes the total energy of the system; hence, in a moving system, such as a
flowing fluid particle, Et will include not only the internal energy but also the kinetic and
potential energies. Thus, for a fluid particle, the energy per unit volume is

(2-32)

where e = internal energy per unit mass

r = particle displacement or position vector
g = gravitational acceleration vector
Just as in conservation of mass and momentum, the energy equation for a fluid may be
conveniently written as a time rate of change, following the particle. Thus, Eq. (2-31)
becomes

(2-33)
From Eq. (2-32), we then have

(2-34)

It remains to express Q and W in terms of fluid properties.

It is assumed that the heat transfer Q to the element is given by Fourier’s law. From Eq.
(1-35), the heat flow per unit area is given by the vector

(1-35)
Referring to Fig. 2-2, the heat flow into the left face of the element is qx dy dz, while the heat
flow out of the right face is

FIGURE 2-2
Heat and work exchange on the left and right sides of an element.

Page 47A similar situation holds for the upper and lower faces involving q y, and the front
and rear faces involving q z. In each case, the net heat flow is out of the element. Hence, the
net heat transferred to the element is

Dividing by the element volume dx dy dz, we have the desired expression for the heat-
transfer term neglecting internal heat generation:
 (2-35)

Referring again to Fig. 2-2, the rate of work done to the element per unit area on the left face
is and the rate of work done by the right-face stresses is

Again, the other faces are similar. In just the same fashion as the heat transfer, then, the net
rate of work done on the element may be determined from

(2-36)

Using the indicial notation (App. E), this becomes, very simply,

(2-37)

The ensuing expression can be further decomposed in a very convenient fashion into

(2-38)

where the “:” symbol is termed double contraction or double-dot product and has the
property At this point, the divergence of the stress tensor may be directly related
to the momentum equation via

(2-19)

Hence, after some vector algebra,†† we get

(2-39)

Page 48which are exactly the kinetic- and potential-energy terms in Eq. (2-34). Thus, the
kinetic and potential energies vanish when we substitute for Et, Q, and W in Eq. (2-33) from
Eqs. (2-34), (2-35), and (2-38):

(2-40)

This is a widely used form of the first law of thermodynamics for fluid motion. If we split the
stress tensor into pressure and viscous terms, using Eq. (2-27), we have
 (2-41)

From the equation of continuity [Eq. (2-6)], we have

(2-42)

Combining Eqs. (2-40) to (2-42) gives

(2-43)

The final expression consolidates the internal and pressure energies into the fluid enthalpy,

(2-44)
Generally, enthalpy will prove to be a more useful function than internal energy, particularly
in boundary-layer flows. Furthermore, the pressure term Dp/Dt may often be neglected in Eq.
(2-43), while the related term p∇ ⋅ V in Eq. (2-40) will be nonnegligible for the same flow.

The last term in Eq. (2-43), involving viscous stresses, is customarily called the
dissipation function Φ:

(2-45)

The term Φ is always positive definite, in accordance with the second law of
thermodynamics, since viscosity cannot add energy to the system. For a Newtonian fluid,
using τ from Eq. (2-27), we obtain

(2-46)

which is always positive since all the terms are quadratic. On the other hand, λ may actually
be negative, e.g., Stokes’ hypothesis with It is an interesting exercise, using Eq. (2-
46), to prove that the correct conditions in which Φ may not be negative are

(2-47)
Using this short notation, the energy relation (2-43) takes the final form

(2-48)

where Φ is given by Eq. (2-46).

2-5.1 The Incompressible-Flow Approximation
From the thermodynamic identity (1-77), we can rewrite Eq. (2-48) as

(2-49)

Now, if the flow velocity scale U becomes smaller, while heat transfer remains important, the
fluid kinetic energy U2 will eventually become much smaller than the enthalpy change cp ΔT.
Since both Dp/Dt and Φ are of order U2, the limit of low-velocity or incompressible flow will
be

(2-50)

Page 49If we further assume uniform thermal conductivity, we obtain the more familiar
incompressible heat-convection ​equation:

(2-51)

Note that the correct specific heat is cp, not cυ, even in the incompressible-flow limit of near-
zero Mach number. A very good discussion of this point is given in Panton (1996, Sec. 10.9).

2-5.2 Summary of the Basic Equations

To summarize, the three basic laws of conservation of mass, momentum, and energy have
been adapted for use in fluid motion. They are, respectively,

(2-6)

(2-19)

(2-48)

where and, for a linear (Newtonian) fluid, the viscous stresses are

(2-27)

As mentioned in the beginning of this chapter, Eqs. (2-6), (2-19), and (2-48) involve seven
variables, of which three are assumed to be primary: p, V, and T (say). The remaining four
variables are assumed known from auxiliary relations and data of the form
 (2-1)
Finally, we note that these relations are fairly general and involve only a few restrictive
assumptions: (1) the fluid forms a (mathematical) continuum, (2) the particles are essentially
in thermodynamic equilibrium, (3) the only effective body forces are due to gravity, (4) the
heat conduction follows Fourier’s law, and (5) there are no internal heat sources.

2-6 BOUNDARY CONDITIONS FOR VISCOUS HEAT-

CONDUCTING FLOW
The various types of boundary conditions have been discussed in detail in Sec. 1-4 for the
three different basic boundaries: (1) a fluid–solid interface, (2) a fluid–fluid interface, and (3)
an inlet or exit section. These various types of conditions are sketched in Fig. 2-3. The
general results are that, for the majority of viscous-flow analyses, the following conditions
must be imposed:

FIGURE 2-3
Various boundary conditions in fluid flow.

1. At a fluid–solid interface, there must be no slip

(2-52)
and either no temperature jump (when the wall temperature is known)

(2-53)
or equality of heat flux in the normal direction n to the surface (when the solid heat flux
 is known)

(2-54)

2. At the interface between a liquid and a gaseous substance, such as the standard
atmosphere, there must be kinematic equivalence in the normal direction to the interface

(2-55)
Page 50

where η is the surface coordinate, and there must be equality of normal momentum flux

(2-56)
We must also have equality of tangential shear stress and normal heat flux at the
interface, which translate into

(2-57)

where Vt is the tangential velocity component parallel to the direction of shear. These
simplifications occur when the gaseous substance has negligibly small transport
coefficients relative to those of the liquid. Also, in some problems, such as the analysis
of confused, stormy seas, we need to know the moisture evaporation rate ṁ at the
interface.

3. At any inlet section of the flow, we need the three quantities V, p, and T at every point on
the boundary. At the exit section, we generally need to know V and T, but not the
pressure p. Exit conditions are difficult because of the formation of wakes and other a
priori unknown outflow behavior. One approximation is to let the streamwise flow
gradients vanish far downstream of the flow field of interest.
Note: If the pressure can be eliminated, as, for example, with the vorticity/stream-
function approach to be discussed in Sec. 2-11, then there is no need to impose a pressure
boundary condition. The computer-oriented vortex methods, explained in the monograph by
Cottet and Koumoutsakos (1999), do not compute the pressure at all.

2-7 ORTHOGONAL COORDINATE SYSTEMS

Most of the previous discussion has been illustrated by Cartesian coordinates, and only lip
service has been paid to other important orthogonal systems. The basic equations of motion,
Eqs. (2-6), (2-19), and (2-48) are, of course, valid for any coordinate system when written in
tensor form; the problem for non-Cartesian systems is to derive the correct formula for the
gradient vector ∇ plus the related expressions for divergence and curl. A straightforward
procedure is to use the metric stretching factors h i to relate the new curvilinear coordinates
to a Cartesian system, as discussed in standard mathematical references, e.g., Pipes (1958).
Let the general curvilinear system (x1, x 2, x 3) be related to a Cartesian system (x, y, z) so
that the element of arc length ds is given by

(2-58)
which defines the factors hi. Note that the hi stretching factors in general will be functions of
the new coordinates xi. Then, in the new system, the components of the gradient of a scalar ϕ
are

(2-59)

The divergence of any vector A = (A1, A2, A3) may be reconstructed from

(2-60)

Page 51

and the spatial components of the vector B = ∇ × A may be obtained from

(2-61)

with exactly similar expressions for B2 and B3. Equations (2-59) to (2-61) are sufficient to
derive the equations of motion in the new coordinate system xi. We illustrate with the two
classic (and probably most important) systems, cylindrical and spherical.

2-7.1 Cartesian Coordinates

Just to review our previous results, let xi = (x, y, z), from which hi = (1, 1, 1). If V = (u, υ, w),
the vector operations are
 (2-62)

2-7.2 Cylindrical Polar Coordinates

These radial, azimuthal, and axial (or vertical) coordinates (r, θ, z) are related to the Cartesian
system (x, y, z) by

(2-63)
Page 52

from which we find that h1 = 1, h2 = r, and h3 = 1. Let the cylindrical polar velocity
components be (υr, υθ, υz). Then, the new vector relations are

(2-64)
2-7.3 Spherical Polar Coordinates
As clarified in Fig. 2-4, the spherical polar coordinates (r, φ, θ), which stand for radial, polar,
and azimuthal variables, respectively, are defined by the relations
 FIGURE 2-4
Spherical coordinate system with the same azimuthal angle as in cylindrical
coordinates

(2-65)
from which we find that h1 = 1, h2 = r, and h3 = r sin φ. Let the velocity components be V =
(υr, υφ, υθ). Then, the desired new vector relations are

(2-66)
It is seen that the terms in non-Cartesian systems are somewhat more complicated in
form but relatively straightforward. The complete equations of motion in cylindrical and
spherical coordinates [i.e., the special-case forms of Eqs. (2-6), (2-19), and (2-48)] are given
in App. G.
2-8 MATHEMATICAL CHARACTER OF THE
EQUATIONS OF MOTION
The character of the basic relations, Eqs. (2-6), (2-19), and (2-48), is extremely complex.
There are at least two factors that hinder our analysis: (1) the equations are coupled in the
three variables V, p, and T, and (2) each equation contains one or more nonlinearities. Are
these equations of the boundary-value type, initial-value type, or are they wavelike in nature?
The answer is that they contain mixtures of boundary-value, initial-value, and wavelike
characteristics.

The general theory concerning the character of partial differential equations has
developed from the study of the interesting but rather specialized quasi-linear second-order
equation, which remains linear in its highest derivatives:

(2-67)

Page 53where the coefficients A, B, and C may be constants or functions of (x1, x2), and
where D may be a nonlinear function of x1, x2, ϕ, ∂ ϕ/∂ x1, and ∂ϕ/∂x2. By developing an
analytic continuation of the second derivatives, it is found that the character of Eq. (2-67)
changes radically, depending on the sign of the discriminant B2 − 4AC. In particular,

(2-68)

The names elliptic, parabolic, and hyperbolic are related to the conic sections of analytic
geometry. Like the conic sections, these names denote a vastly different character for Eq. (2-
67):

1. If the equation is elliptic, it can be solved by specifying the boundary conditions on a

complete contour enclosing the region; it is a boundary-value problem.
2. If the equation is parabolic, boundary conditions must be closed in one direction but
remain open at one end of the other direction; it becomes a mixed initial- and boundary-
value problem.
3. If the equation is hyperbolic, it can be solved in a given region by specifying conditions
at only one portion of the boundary, the other boundaries remaining open; it proves to be
an initial-value problem.
Many of the partial differential equations of mathematical physics may be classified with
reference to Eq. (2-67), albeit expressed in terms of two variables only. For example, the
most popular forms may be expanded in Cartesian coordinates and written as
 (2.69a)

(2.69b)

(2.69c)

In applying the two-variable canonical form of Eq. (2-67) to Eq. (2-69), one may use x1 =
x and x2 = y or else x1 = t and x 2 = x (when both spatial and temporal variables are present),
while ignoring the remaining spatial variables, to ​classify the corresponding two-variable
problem. In general, multidimensional equations written in two- or three-dimensional space
exhibit the same character as their reduced one- or two-dimensional spatial forms. Some
equations may even exhibit a mixed character that depends on the value of their discriminant.
The Euler–Tricomi equation, represents a linear partial differential
equation with a variable coefficient, which proves to be elliptic for x > 0, parabolic for x = 0,
and hyperbolic for x < 0.
Page 54From a physical standpoint, it may be worth mentioning that in parabolic and
elliptic equations, any perturbation of the initial or boundary-distributed values is felt at once
everywhere in the domain. Conversely, solutions of hyperbolic equations exhibit a wavelike
character. Hyperbolic equations tend to be the most challenging to solve because when their
initial values are disturbed, the disturbances cannot be immediately sensed at all spatial
locations. In a hyperbolic problem, the disturbances travel at a finite propagation speed along
the so-called characteristics of the equation. For further examples and discussion of these
partial-differential-equation classifications, see Chap. 21 of the text by Kreyszig (2011).
Most importantly, the viscous-flow equations, (2-6), (2-19), and (2-48), display a mixture
of elliptic, parabolic, and hyperbolic behavior depending on whether they are steady or
unsteady, viscous or inviscid, and compressible or incompressible. While parabolic behavior
may be associated with time dependence and strongly directional motions, elliptic and
hyperbolic characters prove to be dependent on the relative dominance of viscous diffusion
and spatial convection terms. For example, the Navier–Stokes equations, which are parabolic
in the presence of viscosity and unsteadiness, become elliptic under steady flow conditions.
Conversely, Euler’s inviscid equations, which prove to be hyperbolic for unsteady
(compressible) supersonic conditions, become elliptic for steady subsonic conditions (both
compressible and incompressible), and parabolic for unsteady incompressible-flow
conditions. As for the energy equation, it follows the Navier–Stokes equation in its display of
a parabolic or elliptic character depending on its time dependence. Finally, in the treatment of
thin shear layers and boundary-layer flows, velocity gradients in the streamwise direction
become significantly smaller than velocity gradients in the cross-streamwise direction. The
corresponding equations are left with only one second-order spatial derivative of the velocity
and hence become parabolic under both steady and unsteady flow conditions. The
monograph by Kreiss and Lorenz (1989) develops a variety of mathematical properties and
theorems for the Navier–Stokes equations.
It is instructive now to consider the special case of diffusion.

2-8.1 Low-Speed Diffusion: The Prandtl Number

Consider an incompressible Newtonian fluid at low speed with constant ρ, μ, and k. Let the
streamwise changes be small, that is, neglect the convective derivatives compared to local
changes. Then, the momentum and energy ​equations reduce to

(2-70)

(2-71)

We recognize Eq. (2-71) as a multidimensional form of Fourier’s heat-conduction equation

(2-69b). The temperature has elliptic boundary-value behavior in space (x, y, z) and parabolic
“marching” behavior in time. We only need an initial condition T (x, y, z, 0) to get started and
can “march” forward in time indefinitely if we know boundary values on T at all times.

The behavior of Eq. (2-70) is not so obvious. But we can eliminate V by taking the
divergence of the entire equation, noting that ∇ ⋅ V = 0 for incompressible motion. The result
is

(2-72)
which is elliptic based on Eq. (2-69a). Recalling that ω = ∇ × V represents the fluid vorticity,
the pressure can be entirely eliminated by taking the curl of Eq. (2-70); one gets

(2-73)
This relation, like Eq. (2-71), is also a heat-conduction equation in the vorticity, with
diffusivity coefficients that are purely fluid properties:

(2-74)

Note that the kinematic viscosity ν and thermal diffusivity α have no mass units. The units of
both coefficients are m2/s, identical to the mass diffusivity D from Eq. (1-42). As we saw in
Chap. 1, their ratios constitute fundamental dimensionless fluid-property groups that provide
a measure of relative rates of diffusion:

(2-75)

Page 55Note that Pr = Sc Le. Since mass diffusion is not emphasized here, we limit our
attention to viscous and thermal effects. Table 2-1 catalogs values of the Prandtl number for
various fluids at 20°C. It is clear that the Prandtl number is small for liquid metals, slightly
less than unity for gases, somewhat higher than unity for light liquids, and very large for oils
because of their relatively large viscosities. In short, there are significant differences among
fluids in their abilities to spread viscous and thermal effects.

TABLE 2-1
Prandtl number of various fluids at 20°C

Figure 2-5 shows low-speed viscous flow past a hot wall as an example. The sketches
illustrate low Reynolds number or laminar-flow effects. High Reynolds number or turbulent
flows agree qualitatively with Fig. 2-5, but the profile differences are not so broadly obvious.
 FIGURE 2-5
Prandtl number effects on viscous and thermal diffusion. (Hot wall shown for ​-
convenience.)

A quantitative formula for the relative spreading of viscous and thermal effects may be
deduced from the two diffusion equations

(2-76)

(2-77)

This may be accomplished by defining new spatial variables: for the heat-conduction
equation and for the vorticity transport equation. The transformed equations will have
identical unit diffusion constants and hence spreading rates. For a given unit time, the
thermal spreading distance will then be identical to the viscous spreading length
:

(2-78)

Equation (2-78) proves to be a good approximation for all laminar boundary-layer flows,
even at high speeds.

Page 56

2-9 DIMENSIONLESS PARAMETERS IN VISCOUS FLOW

Since our basic equations of motion are challenging to say the least, it is essential to recast
them in their most efficient form, thereby increasing the usefulness, portability, and
parametric independence of whatever solutions we find. This is accomplished by
nondimensionalizing the governing equations and their boundary conditions.
 For simplicity, we may assume constant c p and cυ, approximately true for all gases, and
neglect the second coefficient of viscosity, λ, which is seldom needed, unless compressibility
effects are significant. Then, our four variables, p, ρ, V, and T, will depend upon space, time,
and eight parameters that occur in the basic equations and boundary conditions:

(2-79)
The eight parameters, μ through ℑ, are assumed to be known from data or thermodynamic
state relations. Note that ℑ represents the coefficient of surface tension defined in Chap. 1.
Given a constant g = −gk, several reference properties appropriate to the flow may be used,
specifically:

(2-80)

Steady viscous flows have no characteristic time scale of their own, so a particle residence
time of L/U may be chosen as a reference time scale. In seeking more generality, however, a
reference time scale of ω−1 may be associated with an unsteady flow with a characteristic
frequency ω. The corresponding dimensionless variables may be denoted by an asterisk and
written as

(2-81)

2-9.1 Nondimensionalizing the Basic Equations

After substituting the above variables into our basic equations (2-6), (2-19), and (2-48), we
may collect terms and divide out dimensional constants so that all remaining terms are made
dimensionless. We thus arrive at our nondimensional equations of motion:

(2-82)
It can be immediately seen that the Navier–Stokes equation contains three
dimensionless groupings:

(2-83)
The Reynolds number is the most important dimensionless group in fluid mechanics. Almost
all viscous-flow phenomena depend on it. The Froude number is important only if there is a
free surface in the flow. Moreover, the degree of unsteadiness is concisely captured by the
dimensionless frequency, or the Strouhal number, St = ωL/U, which represents the ratio of
unsteady and steady inertial forces. Since St appears as the coefficient of the unsteady terms
in Eq. (2-82), all time-dependent terms are dropped altogether in steady or quasi-steady flows
where the Strouhal number is negligible. Conversely, in the field of linear acoustics, all terms
are ignored except for the unsteady inertial and pressure gradient terms in the momentum
equation.

As we move to the dimensionless energy equation, we recover two additional parameters:

(2-84)

Page 57For a perfect gas, the more familiar Mach number, Ma = U/a0, can replace the Eckert
number through the transformation:

(2-85)

The Prandtl number is always important in convective heat-transfer problems. In high-speed

flows, Re, Pr, γ, and Ma are all important (Chap. 7), but at low speeds (Ma < 0.3), the
Eckert/Mach number terms are often negligible and the energy equation reduces to
which is the non-dimensional form of Eq.
(2.51). The product RePr in the conduction term is called the Peclet number, Pe.

An important point, sometimes neglected in elementary approaches, is that the variables

ρ, μ, and k must be specified as thermodynamic functions for the particular fluid studied:
Thus, if different fluids are being compared, Re, Pr, and Ec might not be
sufficient if there are large temperature variations. As we have seen in Chap. 1, gases and
liquids have markedly dissimilar temperature-dependent ρ, μ, and k to the extent that
temperature corrections might be needed. We recall that when a set of parameters is known to
affect a problem, the Buckingham Pi theorem may be used to identify the corresponding
dimensionless groupings, as summarized in App. H1.

2-9.2 Momentum with Free Convection

In free or “natural” convection, there is no freestream U. The flow is driven by gravity acting
on slight density changes. Pressure is nearly constant, and the density can be modeled as
with the coefficient of thermal expansion,

(2-86)
The parameter U is replaced by the velocity grouping μ/( ρ0L). The reader is invited (as a
problem exercise) to nondimensionalize the Navier–Stokes equation again and find a new
parameter characteristic of free convection:

(2-87)

As we shall see in Sec. 4-13, Gr and Pr correlate heat-transfer results in free convection.

2-9.3 Nondimensionalizing the Boundary Conditions

Just analyzing the basic equations is not enough. The boundary conditions also contain
important dimensionless parameters. In the freestream, where V = U and T = T0, the
conditions simply become V* = 1 and T* = 0. Subsequently, if we substitute the same
variables from Eqs. (2-81) into the boundary conditions of Sec. 2-6 at a fixed wall, we obtain

(2-88)

So the temperature conditions specify either a unit temperature ratio or a dimensionless wall
heat transfer parameter that is called the Nusselt number, Nu. The computed results at the
wall will then be the opposite: either a Nusselt number or a (nonunit) temperature ratio.

If there is slip or temperature jump at the wall, Eqs. (1-80) and (1-83) become

(2-89)

Slip conditions introduce the Knudsen number, Kn = ℓ/ L, and the specific-heat ratio, γ =
cp/cυ.

Finally, at an interface where surface tension is important, Eq. (1-86) for interfacial
pressure becomes

(2-90)

This equation introduces two additional dimensionless groups:

(2-91)

Page 58As seen earlier in Eq. (2-83), the Froude number is fundamental to all free-surface
flows and can never be neglected. The Weber number is important only if it is small, of
Ο(10) or less, which is usually due to a small length scale L. The cavitation number is
appropriate if the pressure p0 is interpreted as the vapor pressure of the liquid, pυ. If Ca ≪ 1,
the liquid might vaporize (cavitate) when the local pressure drops below pυ. For further
details, see the text by ​Brennen (1995).

To summarize, the following parameters are important for any particular flow:

1. All viscous flows, especially those with surface boundaries (to guide the motion):
Reynolds number.
2. Variable-temperature problems: Prandtl and Eckert (or Mach) numbers.
3. Flow with free convection: Grashof and Prandtl numbers.
4. Wall heat transfer: temperature ratio or Nusselt number.
5. Slip flow: Knudsen number and specific-heat ratio.
6. Free-surface conditions: Froude number (always); Weber number (sometimes), and
cavitation number (sometimes).

It is interesting to note that, in spite of our care, we have missed two parameters which
become important at higher Reynolds numbers, where the flow becomes turbulent (a mean
flow plus a superimposed random unsteadiness). These two parameters, which occur in
Chaps. 5–7, are the degree of surface roughness (a departure from geometric similarity) and
the amount of turbulence (percentage of random fluctuation) in the reference velocity U
(called freestream turbulence). Reference to these two effects was made in the discussion of
the drag of a cylinder in Fig. 1-8. To the novice, these parameters express unexpected effects
—timely reminders that fluid mechanics is a daunting subject, containing many hidden
surprises to confound or delight the would-be analyst.

As the subject of fluid dynamics continues to evolve, new dimensionless parameters are
still being found, thereby adding to our repertoire of scaling tools. In the interest of clarity, an
extensive list of dimensionless parameters is provided in App. H2. For example, let us
consider y p to be the penetration depth of a rotational wave in a right-cylindrical porous
tube; y p defines the fraction of the radius (measured from the porous wall) where
rotationality prevails. Pursuant to an asymptotic study of the penetration depth in a rocket
chamber, which can be modeled as a porous tube with uniformly distributed sidewall
injection, a similarity parameter, Mj, coined penetration number, was first reported by
Majdalani to W. K. Van Moorhem in 1993:

where Re = UL / ν and St = ωL/U denote the wall injection (crossflow) Reynolds number and
the Strouhal number associated with the dimensionless oscillation frequency in a chamber of
radius L. Majdalani’s penetration number gauges the relative importance of unsteady inertial
and viscous forces and has been shown to control the thickness of the oscillatory Stokes layer
over a porous surface with hard blowing (Majdalani and Van Moorhem 1995; Majdalani
1999; Majdalani and Roh 2000). As illustrated in Fig. 2-6, by fixing Mj at 0.1, the frequency
(or Strouhal number) Page 59can be varied from one end of the spectrum to the other without
affecting the penetration depth. Such a parameter does not appear in the governing equations
of motion and is, therefore, difficult to guess because it depends on a nonlinear combination
of the Reynolds and Strouhal numbers. Nonetheless, its usage enables us to collapse large
families of curves characterizing the depth of penetration, and taken at different spatial
locations, Re and St, into single curves taken at specific axial locations within a porous
chamber (see Fig. 2-7).

FIGURE 2-6
Analytical and numerical solutions for the axial wave in a simulated porous tube
driven by sidewall mass injection (Majdalani 1995; Majdalani and Roh 2002).
Oscillations are shown for the first fundamental mode and three specific values
of the Strouhal (St) and Reynolds numbers (Re) based on the sidewall injection
speed. By fixing the penetration number Mj, the depth of penetration remains
constant despite the increase in the Strouhal number.
 FIGURE 2-7
Depth of penetration of the oscillatory rotational velocity in a porous tube driven
by sidewall mass injection (Majdalani 1995): (a) in the form of multiple curves
computed at one axial station and different Strouhal and kinetic Reynolds
numbers, St and Rek = ReSt; (b) in the form of curves that converge into single
lines at multiple axial stations when plotted versus Mj. Note that the entire
family of curves in (a) collapses into a single curve in (b) corresponding to x*/L0
= 0.5.

Another important parameter that arises in the context of wall-bounded cyclonic motions
is the vortex Reynolds number, namely,

Vortex Reynolds number:

where R and L represent the radius and length of the cyclonic chamber, while Re = UR / ν and
σ = R2/Ai refer to the Reynolds and modified swirl number, respectively. This parameter
combines the effects of the Reynolds number, modified swirl number, and chamber aspect
ratio, thereby taking into account variations in the finite body length in a wall-bounded
cyclonic chamber, which can have a significant bearing on the corresponding helical
structure (Majdalani 2014). Moreover, the vortex Reynolds number is shown to control the
value of the maximum tangential (swirl) velocity as well as the thickness of the forced vortex
core, which develops around the axis of rotation in a cyclonic flow.

2-10 VORTICITY CONSIDERATIONS IN

INCOMPRESSIBLE VISCOUS FLOW
As discussed in Chap. 1, the vorticity vector ω = curl V is a measure of rotational effects,
being equal to twice the local angular velocity of a fluid element. While vorticity is not a
primary variable in flow analysis, the velocity itself being far more useful and significant, it
is still instructive to examine the impact of the vorticity vector on the ​Navier–Stokes
equation. To keep the lesson from being lost in a maze of algebra, we assume constant ρ and
μ, so that Eq. (2-30) applies. We also take g = − gk to the extent of reducing the Navier–
Stokes equation into

(2-92)

As per App. A, vorticity can be introduced into both the acceleration and the viscous terms
using the two vector identities,

(1-10)
 (2-93)
Page 60Then remembering that ∇ z ≡ k and ∇ (∇⋅ V) = 0 when ρ is constant, Eq. (2-92)
becomes

(2-94)

Equation (2-94) is most illuminating. The left-hand side is the sum of the classic Euler terms
for inviscid flow. The right-hand side vanishes if the vorticity is zero, regardless of the value
of the viscosity. Thus, if ω = 0 identically, which is the classic assumption of irrotational
flow, the velocity vector must by definition be a potential function, V = ∇ ϕ, where ϕ is the
velocity potential, and Eq. (2-94) becomes the celebrated Bernoulli equation for unsteady
incompressible flow:

(2-95)

It follows that Bernoulli’s equation is valid even for viscous fluids if the flow is irrotational.
Put another way, Eq. (2-94) shows that every potential-flow solution of classical
hydrodynamics is in fact an exact solution of the full Navier–Stokes equations. The difficulty
is, of course, that potential flows do not and cannot satisfy the no-slip condition at a solid
wall, which requires both the normal and tangential velocities to vanish. This is because the
assumption of irrotationality eliminates the second-order velocity derivatives from Eq. (2-
92), leaving only first-order derivatives, so that only one velocity condition can be satisfied at
a solid wall. In potential flow, then, we require only the normal velocity to vanish at a hard
wall and place no restrictions on the tangential velocity, which is allowed to slip. It would
appear at this point that potential solutions are no aid in viscous analysis, but in fact at high
Reynolds numbers, viscous flow past a solid body is primarily potential flow everywhere
except close to the body, where the velocity drops off sharply through a thin viscous
boundary layer to satisfy the no-slip condition. In many cases, the boundary layer is so thin
that it does not really disturb the outer potential flow, which can be determined by the
methods of classical hydrodynamics described, for example, in the texts by Milne-Thomson
(1968) or Robertson (1965). This was the case in Fig. 1-1a. It benefits the reader, therefore,
to master potential-flow analysis as a natural introduction to the study of viscous flow.

We can ascertain when irrotational flow occurs by deriving an expression for the rate of
change of vorticity. One form is obtained by taking the curl of Eq. (2-94), with the result

(2-96)

which is the vorticity transport equation for an incompressible fluid with a uniform and
constant viscosity. By dropping the second term involving the kinematic viscosity, one
recovers the famous Helmholtz equation of hydrodynamics. The first term on the right arises
from the convective derivative and is called the vortex-stretching term; this term was zero in
the special case of Eq. (2-77). The second term is obviously a viscous-diffusion term; if it is
neglected (for an ideal fluid), Eq. (2-96) leads to Helmholtz’ theorem that the strength of a
vortex remains constant and also to Lagrange’s theorem that ω = 0 for all time if it vanishes
everywhere at t = 0. Still a third result of neglecting the viscous term is Kelvin’s theorem that
the circulation about any closed path moving with the fluid is a constant. All three theorems
are valid and useful in the inviscid approximation, but all three fail when the viscosity is
finite because the second term in Eq. (2-96) is constrained by the no-slip condition not to
vanish near solid walls. Thus, in all viscous flows, vorticity is generally present, being
generated by relative motion near solid walls. However, vorticity can also be present in
inviscid flows. If the Reynolds number is large, the vorticity is swept downstream and
remains close to the wall. A boundary layer is formed, outside of which the vorticity may be
taken as zero. If the flow is between two walls, as in duct flow, the two boundary layers will
meet and fill the duct with vorticity, as in Fig. 1-10, so that the potential-flow model is rarely
applied to internal duct flow.

2-11 TWO-DIMENSIONAL CONSIDERATIONS: THE

STREAM FUNCTION
In Eq. (2-9) we noted that the stream function is an exact solution of the continuity equation
when only two independent spatial variables appear in the flow. For maximum benefit, let us
examine the simple case of constant μ and ρ and two-dimensional flow in the xy plane, i.e., a
flow with only the two velocity components u (x, y, t) and υ (x, y, t). The equations of motion
here are the continuity and the two components of the Navier–Stokes equations:

(2-97)

(2-98)

(2-99)

Page 61These are to be solved for the three dependent variables u, υ, and p. Note that the
momentum equation has been uncoupled from the energy equation by the assumption of
constant μ and ρ, so that T (x, y, t) can be solved later from Eq. (2-50) if needed. By analogy
with Eq. (2-9), Eq. (2-97) can be immediately satisfied by a stream function ψ (x, y, t) of the
type

(2-100)

where ψ takes on the meaning of volume flow rather than mass flow, since the density does
not appear in its definition. Meanwhile, the pressure and gravity can be eliminated from Eqs.
(2-98) and (2-99) by cross-differentiation, i.e., by taking the curl of the two-dimensional
vectorial momentum equation. The result is
 (2-101)

Since V = V(x, y, t) only, the vorticity reduces to a single nonvanishing component ω z. At

the same time, the gradient operator becomes independent on z. By comparison with the
more general vorticity transport relation, Eq. (2-96), we see that the vortex-stretching term (ω
⋅ ∇)V vanishes identically because, for this special case of flow in the xy plane, the vorticity
vector ω is perpendicular to the gradient of V. Combining Eqs. (2-100) and the definition of
ω z, we have the intriguing relation

(2-102)
so that Eq. (2-101) can be rewritten as a fourth-order partial differential equation with the
stream function as the only variable:

(2-103)

At this stage, the boundary conditions can be expressed in terms of the first derivatives of ψ,
from Eq. (2-100). For example, in the flow of a uniform stream in the x direction past a solid
body, the auxiliary conditions can be written as

(2-104)

This stream-function vorticity approach is an alternative to the direct or “primitive” variable

approach of solving for (u, υ, p) in Eqs. (2-97) to (2-99). Equations (2-103) and (2-104) are
convenient for numerical analysis as exemplified in the first (hand-calculated) CFD solution
by Thom (1933).

2-11.1 Creeping Flow: Re ≪ 1

If the viscosity is very large, the Reynolds number UL/ν is very small, Re ≪ 1, and the right-
hand side of Eq. (2-103) dominates. We are left with a linear fourth-order equation,

(2-105)
which is the two-dimensional biharmonic equation for “creeping flow.” Many solutions are
known from elasticity [ Timoshenko and Goodier (1970)], which can be adapted to creeping
flow. A wide variety of low Reynolds number flows are treated in the monographs by
Langlois (1964) and by Happel and Brenner (1983). We shall briefly treat this subject in Sec.
3-9.

2-11.2 High Reynolds Number, Quasi-Viscous Flow: Re ≫ 1

At the opposite side of the spectrum, if the Reynolds number is sufficiently large, the right-
hand side of Eq. (2-103) becomes so small that it can be either dropped or used at the basis of
a matched-asymptotic expansion. In such series expansions, the far field is generally
represented by the outer inviscid solution, and a small viscous correction is added, if needed,
using a stretching transformation of the inner region. Moreover, if the motion is steady or
quasi steady, the far field can be captured by solving a lower-order equation, namely,

(2-106)

Page 62Due to its symmetry, it can be easily shown that the resulting vorticity transport
equation is fully satisfied by any rotational flow so long as its vorticity can be expressed as
ωz = f (ψ ), including the linear relation ωz = C 2 ψ. The latter can be used to derive several
interesting solutions, such as the Taylor profile for a porous channel with uniform sidewall
injection [e.g., Taylor (1956), Griffond and Casalis (2000, 2001), and Féraille and Casalis
(2003)]. Many such profiles are termed “quasi-viscous” as they can be made to satisfy the
no-slip requirement at some of the walls despite their inviscid character. A modest collection
of these is reviewed by Majdalani and Saad (2012).

2-11.3 Axisymmetric Flow: The Stokes Operator

Besides its convenient two-dimensional planar form, the vorticity transport equation can be
equally simplified for axisymmetric flows. As we have seen in Sec. 2.4-7, the radial and axial
velocities can be related to the incompressible Stokes stream function using

(2-107)

Since the motion is confined to the rz plane, the vorticity reduces to a single component in
the θ direction, namely,

(2-108)

Instead of recovering the Laplacian of the stream function, as we have seen in Eq. (2-102),
we now stumble upon the Stokes operator D2, which is a minor variant of the Laplacian in
cylindrical coordinates. Specifically, we get,

(2-109)

As for the vorticity transport equation, since (ω · ∇) = 0 for all axisymmetric flows bearing
only one vorticity component, Eq. (2-96) can be rearranged into

(2-110)
Note that the Laplacian and Stokes operators only differ by the sign of the curvature term.
Nonetheless, compared to Eq. (2-101), the expression we obtain is less compact, especially
when expanded. In this vein, the axisymmetric counterpart of Eq. (2-103) can be written as

(2-111)
Although the resulting right-hand side is slightly longer than the biharmonic creeping
flow relation for Re ≪ 1, the large Reynolds number case leads to an appreciable
simplification. Thus, by limiting our attention to a steady flow with Re ≫ 1, we are left with

(2-112)

Here too, any rotational motion with ωθ = = rf(ψ ) can be shown to satisfy the vorticity
transport equation identically. In fact, by using ωθ = C 2 r ψ, several solutions have been
advanced and shown to remain viable for Reynolds numbers as low as 200. Two simple
examples that will be later discussed correspond to the Taylor-Culick and Vyas-Majdalani
profiles for which C = π and 2 π, respectively [Culick (1966); Vyas and Majdalani (2006)].
These have been used extensively as bulk flow idealizations in both cylindrically-shaped
rocket motors and cyclonically-driven rocket engines. As for the case of a swirling rocket
motor, it is considered separately in Sec. 3-6.4, namely, in the treatment of ​rotating porous
tubes.Page 63

2-12 NON-INERTIAL COORDINATE SYSTEMS

By far the majority of flow problems are treated in a fixed, or inertial, coordinate system.
Cases do arise, though, such as the flow over rotating turbine blades or the geophysical
boundary layer on a rotating earth, where we may wish to use non-inertial coordinates that
move with the accelerating system. Then we must modify Newton’s law, Eq. (2-11), which is
valid only if a is the absolute acceleration of the particle relative to inertial coordinates.

Suppose that (X, Y, Z) belong to an inertial frame and that our chosen coordinates (x, y, z)
are translating and rotating relative to that frame. Let R and Ω be the displacement and
angular velocity vector of the (x, y, z) system relative to (X, Y, Z). Then, by straightforward
vector calculus [see Eq. (2-106) in Greenwood (1987, pp. 49–51.)], we can relate the
absolute acceleration a of a particle to its displacement r and velocity V relative to the
moving system:

(2-113)

Thus, if V is a non-inertial velocity vector, the entire formidable right-hand side of Eq. (2-
113) must replace the derivative DV/Dt in Eq. (2-12) or (2-19). However, we would be
expected to know the functions R(t) and Ω(t), which relate the two systems, so that in a sense
we are merely adding known inhomogeneities to the problem.

Geophysical Flows: The Rossby Number

A classic example of a non-inertial system is the use of earth-fixed coordinates to calculate

large-scale (geophysical) motions, where the earth’s rotation cannot be neglected. In this
case, the first three terms of Eq. (2-113) can be neglected, for three different reasons: (1) the
first term is the earth’s acceleration relative to the fixed stars and is surely almost zero; (2)
the second term vanishes because dΩ / dt is nearly zero for the earth; and (3) the third term is
nothing more than the earth’s centripetal acceleration, which essentially accounts for the
variation of the gravity vector g with latitude. Thus, for geophysical flows,

(2-114)

the second term of which is called the Coriolis acceleration, named after the French
mathematician G. Coriolis, who was the first, in 1835, to study this phenomenon. If, as
before, we use U and L to nondimensionalize this expression, we obtain

(2-115)

where Ro = U/(ΩL) is the dimensionless Rossby number (App. H2). It follows that the
Coriolis term can be neglected if Ro is large, which will be true if the motion scale L is small
compared to the earth’s radius. We shall study a geophysical viscous flow (Ekman flow) in
Chap. 3.

2-13 CONTROL-VOLUME FORMULATIONS

The expressions that we have seen so far have all been differential relations: fluid flow in the
small, so to speak. We also need to use integral relations to calculate the gross fluxes of mass,
momentum, and energy passing through a finite region of the flow. This is the so-called
integral, control-volume approach now widely used in fluid mechanics [see White (2016),
Chap. 3]. We limit ourselves here to a fixed control volume; the text by Hansen (1967) treats
moving and deformable control volumes.

Let the closed solid curve in Fig. 2-4 represent a finite region (control volume) through
which a fluid flow passes. At any instant t, the region is filled by an aggregate of fluid
particles that define the system, i.e., a quantity of known identity. Our goal here is to calculate
the rate of change dB / dt of any extensive gross property B (mass, kinetic energy, enthalpy,
etc.) of the system that depends on its mass at that instant t. We do so by considering a
limiting process: At time t + Δt, the system has passed on slightly to the right, as indicated by
the dotted lines in Fig. 2-8 and suggested by the streamline arrows. Page 64
 FIGURE 2-8
Sketch of a control volume.

The slight motion outlines three regions, I, II, and III, as marked on the figure. The limit
we are looking for is

(2-116)

As shown in undergraduate texts, such as White (2016), Sec. 3.2, the region II terms in the
brackets become the rate of change of B within the control volume. The region III term is the
outflow of B, and the region I term is the inflow of B into the control volume. By defining the
area vector dA = n dA as having the direction of the outward normal unit vector to the
control surface (see Fig. 2-8), we can account for outflow and inflow with a single relation
known as the Reynolds transport theorem, specifically

(2-117)

where β = dB / dm represents the mass-independent intensive property associated with the

quantity B that happens to be enclosed within the control volume. Accordingly, the time rate
of change of an extensive property associated with a fixed system crossing a control volume
must equal the rate of accumulation of that property within the control volume plus the net
flux of that property out of the control surface. In what follows, we apply this formulation to
three extensive properties, namely, the mass, momentum, and energy.

2-13.1 Conservation of Mass

The relevant property is the mass B = m for which β = dm/dm = 1. Equation (2-117) becomes

(2-118)
This relation is true for any fixed control volume and will be very useful in boundary-layer
theory, Chap. 4. As an exercise, one can show, using Gauss’ theorem, that Eq. (2-118) is
identical to the differential continuity equation given by Eq. (2-6).

2-13.2 Conservation of Linear Momentum

The intensive property corresponding to the linear momentum B = mV is the velocity itself, β
= d(mV)/dm = V. Equation (2-117) becomes equal to the total force ΣF acting on the system
passing through the control volume:

(2-119)

It must be emphasized that this relation holds only for an inertial control volume and that the
force ΣF incorporates both surface forces acting on the control volume and body forces
acting from a distance on the mass within. Equation (2-119) was first applied to viscous
boundary layers by Kármán (1921).Page 65

2-13.3 Conservation of Energy

When the extensive property is taken to be the total, mass-dependent energy B = E, the
corresponding intensive quantity becomes the specific energy β = et = dE/dm. Recalling the
first law of thermodynamics dE = dQ + dW, Eq. (2-117) becomes

(2-120)

where

The work term dW includes work on the boundaries of normal and shear stresses plus any
shaft work added to the system:

(2-121)
Equation (2-120) may be applied to boundary layers to determine the heat transfer.

2-13.4 The Steady-Flow Energy Equation

An important special case of Eq. (2-120) occurs in steady flow when the control volume
consists of fixed solid walls plus a simple one-dimensional inlet and outlet, as shown in Fig.
2-9. The work of boundary forces is zero at the fixed walls because of the no-slip condition.
Further, if the inlet and outlet flows are approximately uniform and parallel, we may
reasonably neglect viscous normal stresses locally, so that the only important boundary work
is due to pressure forces at the inlet and outlet. The term d/dt vanishes for steady flow, and
Eq. (2-120) reduces to
 FIGURE 2-9
Sketch for the one-dimensional steady-flow energy equation.

(2-122)

But the only contributions to V ⋅ dA are − V1 A1 and + V2A2 at the inlet and outlet,
respectively. Thus, we can carry out the integration to yield two terms

(2-123)

where the subscript s denotes shaft work. For steady flow, equals the mass
flow rate dm/dt through the control volume. Also, the quantity p/ρ + e is the fluid enthalpy h.
We may now divide through by dm/dt to achieve the familiar form

(2-124)

which we recognize as the so-called steady-flow energy equation, stating that the change in
total enthalpy equals the sum of the heat and shaft-work addition per unit mass.
The formula does not hold in boundary-layer theory unless we are careful to define our
control volume to match Fig. 2-9 conditions.

SUMMARY
The role of this chapter is to derive and discuss the basic equations of viscous fluid flow:
conservations of mass, momentum, and energy plus the auxiliary state relations. The various
boundary conditions and coordinate systems are discussed, and dimensionless parameters are
derived and related to particular classes of flow. Stream-function and vorticity approaches are
discussed, and the basic relations are also derived in integral form for finite control volumes.
It is the purpose of the remainder of this text to tackle these equations as best as we can for
particular classes of viscous flow.Page 66

PROBLEMS
2-1. By consideration of the cylindrical elemental control volume shown in Fig. P2-1,
 use the conservation of mass to derive the continuity equation in cylindrical
coordinates [App. G, Eq. (G-3)].

Figure P2.1

2-2. Simplify the equation of continuity in cylindrical coordinates (r, θ, z) to the case of
steady compressible flow in polar coordinates ∂/∂ z = 0 and derive a stream function
for this case.

2-3. Simplify the equation of continuity in cylindrical coordinates to the case of steady
compressible flow in axisymmetric coordinates ∂/∂ θ = 0 and derive a stream function
for this case.

2-4. For steady incompressible flow with negligible viscosity, show that the Navier–
Stokes relation [Eq. (2-30)] reduces to the condition that along a
streamline of the flow, where h denotes the height of the fluid particle above a
horizontal datum. This is the weaker form of the so-called Bernoulli’s relation.

2-5. Show that for incompressible steady flow with negligible viscosity and thermal
conductivity, the energy Eq. (2-40) reduces to the condition that
along a streamline of the flow. This is the stronger form of
Bernoulli’s relation.

2-6. Consider the proposed incompressible axisymmetric flow field υz = C(R2 − r2) in
the region 0 ≤ z ≤ L, 0 ≤ r ≤ R, where C is a constant. (a) Determine if this is an exact
solution of the Navier–Stokes equation. (b) What might it represent? (c) If an
axisymmetric stream function ψ (r, z) = 0 exists for this flow, find its form.

2-7. Investigate the two-dimensional stream function ψ = Cxy, with C a constant, to

determine whether it can represent (a) a realistic incompressible frictionless flow and
(b) a realistic incompressible viscous flow. Sketch some streamlines of the flow.
 Investigate a proposed plane stream function for isothermal incompressible flow
2-8. where C is a constant. Determine whether this flow is an exact solution
for constant μ and, if so, find the pressure distribution p(x, y) and plot a few
representative streamlines.

2-9. Consider the incompressible plane unsteady flow by Lamb-Oseen (1932):

where C and ν are constants, and gravity is neglected. Is this an exact solution of the
continuity and Navier–Stokes equations? If so, plot some velocity profiles for
representative times. Is there any vorticity in the flow? If so, plot some vorticity
profiles.

2-10. Using the expression for dissipation Φ from Eq. (2-46), prove the inequalities
about μ and λ given by Eq. (2-47).

2-11. The differential equation for irrotational plane compressible gas flow [Shapiro
(1954), Chap. 9] is

where ϕ is the velocity potential and a the (variable) speed of sound in the gas. In the
spirit of Sec. 2-9.1, nondimensionalize this equation and define any parameters which
appear.

2-12. . In flow at moderate to high Reynolds numbers, pressure changes scale with
as in Eq. (2-83). In very viscous (low Reynolds number) flow, pressure scales μU0/L.
Make this change in Eqs. (2-83) and repeat the nondimensionalization of the Navier–
Stokes equation. Define any parameters which arise and show what happens if the
Reynolds number is very small.

2-13. The equations of motion for free convection near a hot vertical plate for
incompressible flow with constant properties are

Page 67

Introduce the dimensionless variables

 where L is the length of the plate. Use these variables to nondimensionalize the free
convection equations and define any parameters that arise.

2-14. For laminar flow in the entrance to a pipe, as shown in Fig. P2-14, the entrance
flow is uniform, Using the integral relations of Sec. 2-13, show that the viscous
drag exerted on the pipe walls between 0 and x is given by

Figure P2.14

2-15. To illustrate “boundary-layer” behavior, i.e., the effect of the no-slip condition for
large Reynolds numbers, Prandtl in a 1932 lecture proposed the following model
(linear) differential equation:

where ε mimics the effects of small fluid viscosity. The boundary conditions are (1)
u(0) = 2, and (2) u remains bounded as y becomes large. Solve this equation for these
conditions and plot the profile u(y) in the range 0 < y < 2 for ε = 0, 0.01, and 0.1.
Comment on the results. Is the plot for ε = 0 the same as that obtained by setting ε = 0
in the original differential equation and then solving?

2-16. Consider the plane, incompressible, Cartesian stream function in the region 0 ≤ y ≤
∞:

where (a, b, c) are positive constants. (a) Determine if this is an exact solution to the
continuity and Navier–Stokes equations if gravity and the pressure gradient are
neglected. (b) What are the dimensions of (a, b, c)? (c) If y = 0 represents a wall, does
the no-slip condition hold there? (d) Is there any vorticity in the flow field? If so,
what is its form?

2-17. Use Eq. (2-86) for density in the gravity term of the Navier–Stokes equation (2-
29b), let ρ ≈ ρ0 in the acceleration term, and let μ = constant. For free convection,
with U replaced by μ / ( ρ0 L), nondimensionalize the Navier–Stokes equations and
 show that the Grashof number appears.

2-18. Flow through a well-designed contraction or nozzle is nearly frictionless, as

shown, for example, in White (2016), Sec. 6-12. Suppose that water at 20°C flows
through a horizontal nozzle at a weight flow of 50 N/s. If entrance and exit diameters
are 8 cm and 3 cm, respectively, and the exit pressure is 1 atm, estimate the entrance
pressure from Bernoulli’s equation.

2-19. Show, using Gauss’ theorem [Kreyszig (2011), Sec. 9.8] that the control-volume
mass relation, Eq. (2-118), leads directly to the partial differential equation of
continuity, Eq. (2-6).

2-20. In discussing incompressible flow with constant μ, Eq. (2-30), we cavalierly said,
“many terms vanish” from Eq. (2-29a). Be less cavalier and show that the many
viscous terms in Eqs. (2-29a) do indeed reduce to the single vector term μ∇2V in three
dimensions.

2-21. In deriving the basic equations of motion in this chapter, we skipped over the
partial differential equation of angular momentum. Did we forget? Do some reading,
perhaps in Lai et al. (1995) or Malvern (1997), and explain the significance of the
angular momentum equation.

2-22. Normalize the Navier–Stokes equation with constant properties, specifically,

(a) by setting:

(b) Define the dimensionless parameters that appear in the following

nondimensional form:

Page 68

2-23. In order to better analyze the boundary-layer structure over a flat plate with a
characteristic length L, the traditional variables are rescaled using
 where ε ≈ δ / L ≪ 1 and δ denotes the boundary layer thickness. Show that ∂2u/∂ x2
≪ ∂2u/∂ y2 in the axial momentum equation written for the boundary-layer region.
For steady-state motion with constant fluid properties, you may start with the axial
momentum equation given by:

2-24. Given the dimensionally scaled expressions of the basic fluid dynamical forces:

where the velocity gradient is approximated using dV/dy ≈ U/L, verify the following
characteristic ratios:

2-25. In 1851, Stokes introduced an expression for the drag force, FD = 3πμVD, in the
context of laminar flow over a smooth sphere of diameter D. Express the drag
coefficient as a function of the Reynolds number. Recall that

A should be the projected area of the sphere. Answer:

Page 69

2-26. In normalizing the basic equations of motion in Sec. 2-9.1, we have assumed a
characteristic time constant that scales with the unsteady period, i.e., the reciprocal of
 the frequency, τ = 1/ω, when we set t * = ωt. Show that by taking a different t* =
tU/L, for which τ = L/U represents the time for a fluid particle to travel a distance L at
the characteristic mean-flow velocity, (a) the continuity and momentum equations
reduce to

(b) the energy equation reduces to

2-27. Renormalize the basic equations of motion in Sec. 2-9.1 assuming an

incompressible fluid with negligible dissipation function Φ as well as constant
viscosity and thermal conductivity, μ = μ0 and k = k0. The normalization will
correspond to

(a) Show that the continuity and momentum equations can be

simplified into

(b) Show that the energy equation reduces to

(c) Under what circumstances does the Euler number become

important?
(d) What happens to this set of equations if the density ρ = ρ0 is
assumed to be constant?
2-28. In deep water, the average speed V of a free-surface wave can be expressed as a
function of the density, ρ, gravity, g, depth, D, and spatial wavelength, λ. Obtain a
dimensionless representation of the free-surface wave speed as a function of these
controlling parameters.
2-29. A model sphere of diameter D is to be immersed in a water tunnel running at a
speed of V in order to obtain data that can be used to predict the drag force on a
weather balloon that is traveling in air at 1.5 m/s. For a model diameter of 50 mm, the
measurement of the drag force on the small sphere yields 3.78 N. Assuming similarity
conditions have been established between the spherical lab-scale model and the air
balloon, what should be the water tunnel speed if the diameter of the balloon is 3 m?
Estimate the drag force on the full-scale balloon.

2-30. Surface tension is responsible for causing capillary waves to develop on a liquid-
free surface with an average speed, V, that is dependent on the fluid density, ρ, surface
tension, ℑ, and spatial wavelength, λ. Obtain the nondimensional relations linking the
speed of the capillary waves to these three independent quantities.

2-31. The lifting force F on an air-launched rocket depends on its length L, velocity V,
diameter D, angle of attack α, density ρ, dynamic viscosity μ, and speed of sound c.
Using the Buckingham Pi theorem, determine the dimensionless groups or Pi
parameters needed to characterize the functional relation between the lifting force and
the quantities that affect it.

2-32. A solid propellant rocket chamber is often modeled as a porous tube with sidewall
injection (Majdalani 1999). The oscillatory Stokes boundary layer that develops
inside a porous tube with sidewall injection (due to pressure oscillations) is affected
by variations in several parameters. These include the injection velocity U, the
frequency of oscillations ω, the kinematic viscosity ν, the hydraulic radius of the
chamber R, the chamber length L, and the axial location within the chamber X. The
dependence of the Stokes boundary-layer thickness Δ on these parameters may be
expressed as Δ = f (U, ω, ν, R, L, X). Using scaling principles, show that the
penetration depth of the unsteady rotational wave, yp = Δ/R, may be related to the
following dimensionless parameters

Note that the square, square-root, or reciprocal of a dimensionless parameter as well

as the product of two dimensionless parameters, such as X / R and R/L, can also
produce a nondimensional group.

2-33. The fluid hammer or hydraulic shock can occur when a valve is closed rapidly in a
fluid supply line. Given a circular tube, the pressure impulse Δp = p1 − p2, also known
as the strength of the fluid hammer, depends on the fluid density, ρ, the speed of
sound in the fluid, c, the diameter of the tube, D, the dynamic viscosity, μ, and the
mean-flow velocity, V. Using the Buckingham Pi Theorem, identify the dimensionless
groups that will minimize the number of test cases needed to fully characterize the
pressure impulse sensitivity to these five independent parameters. Note that the
 dimensionless pressure impulse is traditionally referenced to the speed of sound and
that the Reynolds number based on the speed of sound is known as the acoustic
Reynolds number.

2-34. The drag force acting on a submarine can be taken to be a function of the water
density, ρ, speed, V, viscosity, μ, and displaced volume, ϑ. Using a 1:50 scaled-down
model of the submarine in a laboratory-size water tunnel, start by specifying the
dimensionless parameters needed to guide your drag force measurements. Assuming
a leveling of the drag force Page 70coefficient at increasing speeds, provide a best
guess of the drag force on a full-scale prototype traveling at 27 knots if a value of 13
N is obtained in a water tunnel running at 10 knots.

2-35. Experimental measurements of the pressure drop across a sudden contraction in a

circular tube suggest that Δp = p1 − p2 can be a function of the density, ρ, dynamic
viscosity, μ, mean-flow velocity, V, and both large and small diameters, D1 and D2.
Identify the dimensionless groups that will minimize the number of test cases needed
to fully characterize the pressure dependence on these five independent parameters.

Figure P2.35

2-36. Water at 15°C flows through a 0.1 m internal diameter galvanized iron pipe (ε =
0.00015 m) at a mass flow rate, of 15 kg/s. We are told that:

The total pipe length, which is L = 150 m, runs in a horizontal plane.

The pipe contains one check valve (Le/D = 55), and four 90° elbows (Le/D = 30 each).

The properties of water are:

(a) Find the Reynolds number. Is the flow turbulent?
(b) Find the Darcy friction factor f.
(c) Find the major head loss, hM.
(d) Find the minor head loss, hm.
 (e) What percentage of the total head loss is due to minor losses?
(f) Find the pressure drop across this pipe.

2-37. Consider a water pipe of length 100 m, diameter of 0.1541 m (6 inches), roughness
of 4.6 × 10−5 m (DN 150 Schedule 40 Steel), viscosity of 1.08 × 10−5 m2/s, specific
weight of 8630 N/m3, and an inlet pressure of 120 kPa at Point 1. In the presence of
minor losses, determine:
(a) The maximum permissible volumetric flow rate (in m3/s) and
average velocity at Point 2 that will ensure that the pressure drop in
the pipe does not exceed 60 kPa.
(b) The total head loss in meters.
(c) The percentage of the total head loss that corresponds to minor
losses.
We are told that the pipe runs in a horizontal plane and that it contains an open
butterfly valve (Le/D = 45), and two standard 90° elbows (Le/D = 30 each).

2-38. A 200 m long cast iron pipe has a roughness of ε = 0.00025 m and a diameter of
0.05 m. It carries water with a density of 1000 kg/m3 and a kinematic viscosity of
10−6 m2/s. The gage pressure in the water main is 1500 kPa. The supply line will
require the installation of eight elbows (L e/D = 30 each), two standard tees (Le/D =
60 each), and one square-edged entrance (K = 0.5), in the total length of 200 m. The
gage pressure required at the discharge point is 500 kPa. Find the discharge flow rate
Q.

2-39. Water at 15°C flows through a galvanized iron pipe with a roughness of ε =
0.00015 m, at a mass flow rate, of 20 kg/s. We are told that:

The total pipe length, which is L = 200 m, runs in a horizontal plane.

The pipe contains one check valve (Le/D = 55), and four 90° elbows (Le/D = 30 each).

The properties of water are: ρ = 1000 kg/m3, μ = 0.00114 N ⋅ s/m2, and g = 9.81 m/s2.

The total pressure drop across the pipe is 169 kPa. What is the diameter of the pipe?

2-40. Water with a density of ρ = 1000 kg/m3 negotiates a 180° turn through an elbow as
shown below. As the water enters the elbow, its gage pressure is 200 kPa. The
pressure at the outlet section is atmospheric. Assuming uniform properties
everywhere, we have A1 = 0.002 m2, A2 = 0.0006 m2, and U1 = 2 m/s. Determine the
 horizontal force that is needed to prevent the elbow from separating. Are the bolts in
tension or compression?Page 71

Figure P2.40

2-41. Consider a two-dimensional planar channel with one inlet and two outlets that
include a reducing bend. The velocity distribution at the inlet varies linearly from
zero to U1. The velocity at the two outlets is uniform and the motion may be taken to
be incompressible and steady. Determine (a) the maximum velocity at Section 1 using
the integral, control-volume approach and (b) the reaction forces needed to support
the channel.

Figure P2.41

2-42. Consider the steady incompressible motion of a fluid across an axisymmetric tube
of length L and radius R = 3 cm.
(a) Determine the uniform velocity at the inlet, U, if the velocity profile
at the outlet can be approximated using

(b) Evaluate your result for Umax = 10 m/s and water with a density of
ρ = 1000 kgm3.
 (c) Determine the difference in the fluid momentum flux between the inlet and outlet
sections of the tube.

2-43. Consider a contraction between two pipes in a hydraulic system, which is known
as a reducer. With the pressure and ​velocity measurements available across the
reducer, and given the empty volume and mass of the reducer (ϑr, mr), determine the
reaction forces that must be supplied by the surrounding system to hold the reducer.
You may assume that gravity is ​downward.

Figure P2.43

Page 72

2-44. A two-dimensional velocity field is given by

where A is a constant. Does the field satisfy continuity?

2-45. For a certain incompressible, three-dimensional flow field, the velocity

components in the x and y directions are given by

(a) Determine the simplest velocity component w in the z direction so

that the volume dilatation rate is zero.
(b) Determine the components of the vorticity vector. Is this an
 irrotational flow field?
2-46. For a two-dimensional flow field, the velocity is defined by
Determine whether this motion is compressible or not.

2-47. For an incompressible, two-dimensional flow field, the velocity component in the
y direction is given by

Determine the velocity component u in the x direction so that the volume dilatation
rate is zero.

2-48. The three components of the velocity in a flow field are given by

Determine the divergence (volumetric dilatation rate) and interpret the results.
Determine an expression for the vorticity vector. Is this an irrotational flow field?

2-49. The velocity vector in a certain flow field is prescribed by

where (x, y) have units of feet and t has units of seconds. Determine expressions for
the local (unsteady) and convective (spatial) acceleration terms. Evaluate the
magnitude and direction of the velocity and acceleration at the point x = y = 2 ft at
time t = 0 s.

2-50. For steady flow, show that the curl of the inviscid momentum equation reduces to
∇ × (V × ω) = 0, where ω = ∇ × V. Start with the steady expression of Euler’s
equation, namely,

2-51. For planar flow in the xy plane, show that the vorticity has only one nonvanishing
component in the z direction, specifically, ω = ωz k. Show that the vorticity, when
written in terms of the incompressible stream function, becomes

2-52. For steady, inviscid, planar flow in the xy plane, (a) show that ∇ × (V × ω) = 0
 reduces to

(b) Show that any solution of the form ωz = f (ψ) will satisfy this
condition, including the linear relation ωz = C2ψ, where C is a
constant.
2-53. For planar flow in the xy plane, any solution of the form ωz = C2ψ can be shown to
satisfy the steady, inviscid vorticity transport equation. When this relation is inserted
into the vorticity expression, one gets

(a) Apply the classification test given by Eq. (2-68) to determine the
character of the equation.
(b) How many boundary conditions will be needed in order to solve
this equation?
2-54. The Taylor profile, which has been used to describe the bulk gaseous motion in
two-dimensional rocket chambers, corresponds to a planar profile in porous channels
that bears symmetry with respect to the chamber’s midsection plane at y = 0. Page
73Assuming unit wall injection and unit chamber half height, the domain may be
defined over 0 ≤ x ≤ l and 0 ≤ y ≤ 1, as shown in Fig. P2-54. Using the stream-
function vorticity approach, one may take ωz = C2ψ to secure the vorticity transport
equation and then eliminate Page 74the velocities in favor of the stream function
everywhere. The resulting vorticity equation in terms of the stream function becomes

(a) Using the method of separation of variables with ψ(x, y) = X(x)

Y(y), find the general form of the solution.
(b) Express the problem’s four fundamental boundary conditions in
terms of the stream function knowing that:
 (c) Find the solution that satisfies these boundary conditions.
(d ) Determine the velocities associated with this solution and verify that they satisfy
the four boundary conditions stated in Part (b).

Figure P2.54

2-55. For axisymmetric flow in the rz plane, show that the vorticity has only one
nonvanishing component in the θ direction, specifically, ω = ωθ eθ. Show that the
vorticity, when written in terms of the incompressible stream function, becomes

2-56. For steady, inviscid, axisymmetric flow in the rz plane, (a) show that the vorticity
transport equation ∇ × ( V × ω) = 0 reduces to

(b) Show that any solution of the form ωθ = rg(ψ) satisfies this
condition, including the linear relation ωθ = C 2rψ, where C is a
constant.
2-57. For axisymmetric flow in the rz plane, any solution of the form ωθ = C2rψ can be
shown to satisfy the steady, inviscid vorticity transport equation. When this relation is
inserted into the vorticity expression, one gets

(a) Apply the classification test given by Eq. (2-68) to determine the
 character of the equation.
(b) How many boundary conditions will be needed to solve this
equation?
(c) How should you define B(ψ) and H(ψ) in the Bragg–Hawthorne Eq.
(2-30e) to reproduce the same partial differential equation considered
here?
2-58. The Taylor–Culick profile, which has been extensively used to describe the bulk
gaseous motion in axisymmetric rocket chambers (Culick 1966), corresponds to a
self-similar axisymmetric profile in porous tubes. Assuming a unit wall injection and
a unit chamber radius, the domain may be defined over 0 ≤ r ≤ 1 and 0 ≤ z ≤ l, as
shown in Fig. P2-58. Using the stream-function vorticity approach, one may take ωθ
= r C2ψ to secure the vorticity transport equation and then eliminate the velocities in
favor of the Stokes stream function everywhere. The resulting vorticity equation in
terms of the Stokes stream function becomes

(a) Using the method of separation of variables with ψ (r, z) = R(r)Z

(z), find the general form of the solution.
(b) Express the problem’s four fundamental boundary conditions in
terms of the stream function knowing that:

(c) Find the solution that satisfies these boundary conditions.

(d) Determine the velocities associated with this solution and verify
that they satisfy the four boundary conditions stated in Part (b).
 Figure P2.58

2-59. In order to derive an exact solution for the velocity field in a wall-bounded
cyclonic flow evolving in a cylindrical chamber (Fig. P2-59), Vyas and Majdalani
(2006) assumed that the flow is steady, incompressible, inviscid, and axisymmetric
with respect to the centerline. Because of the presence of swirl, a three-component
velocity profile is used, namely, In this problem, a and
denote the outer and outlet radii of the cyclonic chamber, U stands for the
circumferential tangential velocity at r = a, Ai = Qi/U denotes the injection area for a
volume flow rate of Qi, and κ = Ai/(2πaL) represents a dimensionless offset swirl
parameter that gauges the relative importance of the axial-to-tangential speed ratio.
The boundary conditions for this problem can be chosen to be

(a) Guided by experimental data and numerical simulations, we may

assume υθ = υθ (r) to be an axially invariant function. Under this
assumption, the azimuthal momentum equation in cylindrical
coordinates, namely,

decouples from its axial and radial counterparts. Recalling that the motion is
axisymmetric (i.e., there are no variations in θ), show that the solution to the
azimuthal momentum equation given above reduces to

(b) Realizing that the solution to Part (a) constitutes a statement of

conservation of angular momentum, rυθ = B = const, solve for B and
υθ (r) using the first boundary condition, υθ (a) = U.
(c) Express the remaining four boundary conditions as functions of the
Stokes stream function using
Page 75

(d ) Because the vorticity transport equation ∇ × ( V × ω) = 0 reduces to

use partial differentiation to prove that any solution of the form ωθ = rg(ψ) satisfies
this condition, including the linear relation ωθ = C2rψ, where C is a constant.
(e) Substitute the Stokes stream function into the tangential vorticity to
show that

( f ) Substitute ωθ = C2rψ, which satisfies the vorticity transport equation, into the
vorticity equation to obtain

(g) Using the method of separation of variables with ψ(r, z) = R(r) Z(z),
find the general form of the solution to the resulting partial differential
equation.
(h) Find the particular solution that satisfies the four boundary
conditions formulated in Part (c).
(i) Show that the velocities associated with this profile are:

( j) Verify that the velocities satisfy the problem’s boundary conditions.

(k) Show that the same governing equation may be restored from the
Bragg–Hawthorne Eq. (2-30e), where B = const and
may be readily substituted. The form of the latter may be easily
deduced from a pressure analysis along a streamline.
 Figure P2.59

†Note, however, that air flowing at very high temperatures will undergo spontaneous diffusion and chemical reactions, as
will many other mixtures. Even single-component fluids, such as oxygen, will dissociate into atomic oxygen at high
temperatures.

‡But not necessarily a constant shape, i.e., the element may undergo normal strains of opposite sign plus shear strains of any
sign.

§Here, we assume the absence of concentrated coupled stresses.

**Realizing that ∇j (δij p) = ∇i p = ∂ p/ ∂ xi, an equivalent indicial form of Eq. (2-19) is