Sei sulla pagina 1di 140

Notes for the 2nd revised edition of

TRANSPORT PHENOMENA
by
R. B. Bird, W. E. Stewart, and E. N. Lightfoot

by R. B. Bird
7 December 2009

These "Notes," prepared in 2009, are intended for use by


students and instructors to fill in the missing steps in some of the
derivations in the text. Comments and corrections will be greatly
appreciated. Also, if there are places in the text where you feel
additional explanation is needed, it would be appreciated if you
would let us know.
Many of these notes involve the Leibniz formula for
differentiating integrals, the hyperbolic functions, the error function,
Taylor series, and the gamma functionsall topics for which many
undergraduates have received inadequate instruction. These topics
are all reviewed in Appendix C.

Pages in the text for which "Notes" have been prepared:

p. 5 Conservation laws in binary collisions


p. 18 Evaluation of stress-tensor components
p. 26 Dimensional consistency in fluid dynamics
p. 35 Convective momentum flux
p. 51 Flow of kinetic energy in tubes
p. 52 Conduits with circular and triangular cross sections
p. 54 Velocity distribution in annular flow
p. 55 Limiting cases of annular flow
p. 58 Flow of immiscible fluids
p. 59 Flow around a sphere
p. 78 Normal stresses at solid surfaces
p. 81 Equation of change for mechanical energy
p. 82(i) Proof that ( :v ) is positive for Newtonian fluids
p. 82(ii) Conservation equation for angular momentum
p. 86 Vector identity needed for the Bernoulli theorem
p. 90 Torque balance
p. 117 Slope of the complementary error function

1
p. 121 Fluid motion near an oscillating plate
p. 123 Equation for the stream function
p. 125 Alternative method of getting Stokes' law
p. 139 The Falkner-Skan equation
p. 154 Average velocity in turbulent flow in tubes
p. 170 Turbulent flow in a circular jet
p. 198 Derivation of macroscopic balances
p. 199 Efflux from a spherical tank
p. 242 Power-law flow in circular tubes
p. 248 Viscoelastic flow near an oscillating plate
p. 255 Polymer flow analyzed with a FENE-P model
p. 275 Dimensional consistency in heat transfer
p. 286 Enthalpy of an ideal monatomic gas
p. 299 Temperature profile in flow with viscous heating
p.309 Checking the cooling fin solution
p. 315 Temperature profile in tube flow
p. 337 Alternative equation of change for temperature
p. 341 The equation of change for entropy
p. 343 Tangential annular flow with viscous heating
p. 346 Temperature profile for transpiration cooling
p. 375 One-dimensional time-dependent heat conduction
p. 377 Two solutions for the slab heating problem
p. 379 Unsteady heat conduction with sinusoidal heating
p. 386 Steady-state potential flow of heat in solids
p. 388 Boundary layer flow with heat transfer
p. 413 Turbulent flow in tubes with heat transfer
p. 415 Turbulent flow in circular jets with heat transfer
p. 454 Derivation of macroscopic energy balance
p. 494 Planck's radiation law and Wien's displacement law
p. 529 Dimensional consistency in diffusion
p. 534 Binary formulas from multicomponent formulas
p. 535 Two formulations of Fick's law of diffusion
p. 547(i) Diffusion through a stagnant gas film
p. 547(ii) Taylor expansion of diffusion problem result
p. 555 Diffusion with chemical reaction
p. 563(i) Diffusion with solid dissolution
p. 563(ii) Evaluation of mass flux from concentration profiles
p. 565 Verification of solution of diffusion problem
p. 585 Diffusion with convection and chemical reaction
p. 589 Energy equation for multicomponent mixtures
p. 591 Euler's theorem for homogeneous functions
p. 615 Time-dependent evaporation of a liquid

2
p. 622 Diffusion with time-dependent interfacial area
p. 626 Diffusion with chemical reaction
p. 692 Interaction of phase resistances
p. 766 Driving force in multicomponent diffusion
p. 767 Simplification of multicomponent diffusion result
p. 768 Relating Maxwell-Stefan and Fick diffusivities
p. 769 Illustrating interrelations between diffusivities

3
Note to p. 5

Section 0.3 is important for emphasizing some of the basic


concepts and definitions. Here we work through some of the missing
steps in Section 0.3, going from Eq. 0.3-3 to Eq. 0.3-4, and from Eq.
0.3-5 to Eq. 0.3-6.

(a) In Eq. 0.3-3, replace rA1 by rA + R A1 and make analogous


replacements for rA2 , rB1 , and rB2 . We also let mA1 = mA2 = 12 mA . With
these substitutions, we then get:
1
2 (
mA rA + R A1 + 2 mA r
1
)
A + R A2 + 2 mB r
1
(
B + R )
B1 + 2 mB r
1
(
B + R B2 = ) ( )
1
2
mA ( rA + R
) + 1 m ( r + R
A1 2 A A
) + 1 m ( r + R
A2 2 B B
) + 1 m ( r + R
B1 2 B B
) (1)
B2

But, according to the drawing in Fig. 0.3-2, R A1 = R A2 and,


analogously, R B1 = RB2 , so that

mA rA + mB rB = mA rA + mB rB (2)

This is the law of conservation of momentum in terms of the


molecular masses and velocities.

(b) We start with Eq. 0.3-5, and replace rA1 by rA + R A1 as above. We


also let mA1 = mA2 = 12 mA . Then Eq. 0.3-5 becomes:

1 1
2 2 ( (
mA ( rA rA ) + 2 rA R
) (

A1 + R A1 R A1 ))

(
+ 12 12 mA ( rA rA ) + 2 rA R A 2 + R
R
A2) (

A2 + A )
(
+ 12 12 mB ( rB rB ) + 2 rB R
( ) (

B1 + R B1 R B1 ))
+
(
+ 12 12 mB ( rB rB ) + 2 rB R B2 + R
R
B2) (
B2 B )
= 1 1
2 2 ( (
mA ( rA rA ) + 2 (rA R A1 ) + R
R
A1

A1 ))

4
+
( ) (
+ 12 12 mA ( rA rA ) + 2 rA R A 2 + R
R
A2 A2 A )
( (
+ 12 12 mB ( rB rB ) + 2 rB R ) (
+ R
B1
R
B1

B1 ))
+
( ) (
+ 12 12 mB ( rB rB ) + 2 rB R B2 R
+ R B2 B2 B) (3)

The single-underlined terms just exactly cancel the doubly


underlined terms in the following line, because R A1 = R A2 and also
R B1 = RB2 . Hence we get

1
2
mA ( rA rA ) + 12 mA1 ( R
R ) + 1 m ( R
A1 A1 2 A2
R ) +
A2 A2 A

+ m ( r r ) + m ( R
1
2 B B B 2
1 R
B1
) + m (R R ) +
B1
1
B1 2 B2 B2 B2 B

= 1
2
mA (rA rA ) + mA1 (R A1 R A1 ) + mA 2 (R A2 R A 2 ) + A
1
2
1
2

+ 1
2
mB(rB rB ) + 12 mB1 (R B1 R B1 ) + 12 mB2 (R B2 R B2
) + B (4)

In the first line of the equation above, the terms have the following
significance: Term 1 is the kinetic energy of molecule A in a fixed
coordinate system; Term 2 is the kinetic energy of atom A 1 in a
coordinate system fixed at the center of mass of molecule A; Term 3 is
the kinetic theory of atom A2 in a coordinate system fixed at the
center of mass of molecule A; Term 4 is the potential energy of
molecule A as a function of rA2 rA1 , the separation of the two atoms
in molecule A. The sum of terms 2 to 4 we call the "internal energy"
uA of molecules A, and Eq. 4 may be rewritten in the form of Eq. 0.3-
6.
This discussion of the collision between two diatomic
molecules is interesting, for several reasons. It shows how the idea of
"internal energy" arises in a very simple system. We encounter this
concept later in 11.1 where the terms "kinetic energy" and "internal
energy" are used in connection with a fluid regarded as a continuum.
When the fluid is regarded as a continuum, it may be difficult to
understand how one goes about splitting the energy of a fluid into
kinetic and internal energy, and how to define the latter. In
considering the collision between two diatomic molecules, however,
the splitting is quite straightforward.

5
Another point is that, having seen the need for splitting the
energy into two parts, one might be led to ask: why don't we need to
split the momentum into two parts in a similar way? Here again, for
the collision of diatomic molecules, the need for dividing the
momentum into two parts is not necessary.
The subject of transport phenomena is built up on the laws of
conservation of mass, momentum, angular momentum, and energy.
The application of these laws to the system of two colliding diatomic
molecules is relatively straightforward. However, when applying
them to a moving fluid, some notational problems arise associated
with the necessity of dealing with fluid bodies in three dimensions.

6
Note to p. 18

In Fig. 3B.2 there is shown a duct with cross-section of an


equilateral triangle. The height of the triangular cross-section is H,
and the side length is 2H 3 . We want to evaluate the viscous stress
tensor components for the incompressible flow in the z-direction, for
which the velocity in the z-direction is given as a function of x and y
in Eq. 3B.2-1:

(P 0 P L )
vz ( x, y ) =
4 LH
( y H )( 3x 2 y 2 ) (1)

and vx = 0 and v y = 0 . Here L is the length of the duct (which goes


from z = 0 to z = L ) and P 0 P L is the difference in modified
pressure between the ends of the duct. What are the stresses at the
surface y = H according to Eq. 1.2-6?
For the velocity distribution given above, the nonzero
components are yz = zy and xz = zx . From Eq. 1.2-6, we get:

vz (P P L ) 3x 2 y y3 3Hx 2 + Hy 2
yz =
y
= 0
4 LH y
( )
(P 0 P L )
=
4LH
( 3x 2
3y 2 + 2Hy ) (2)

vz (P P L ) 3x 2 y y3 3Hx2 + Hy2
xz =
x
= 0
4 LH x
( )
(P P L ) 6x y H
= 0
4LH
( ( ) ) (3)

At the surface y = H:

(P 0 P L )
yz
y= H
=
4LH
( 3x 2
)
H 2 ; xz y=H
=0 (4,5)

7
Note to p. 26

It is very important to make a habit of checking equations for


dimensional consistency. Show that the following equations in the
text are dimensionally consistent: Eq. 1.4-14, Eq. 1.5-11, and Eq. 1.7-2.
Do this by replacing the symbols in the formulas by the dimensions
corresponding to the symbols in the table beginning on p. 872. Omit
any numerical factors that appear.

ML2
( M ) 2 (T )
M tT
(a) Lt = (1)
( )
L2 ()

1 ML2
moles t
M
(b) Lt = (2)
L3
moles

M M M L L M M M L L
(c) 2 = 2 + 3 t t = 2 + 2 + 3 t t
Lt Lt L Lt Lt L
(3)

In each case, dimensional consistency is found. In (c) the unit tensor


is a dimensionless quantity.

8
Note to p. 35

In Fig. 3B.7 we show the flow into a slot of width 2B. Far away
from the slot, the velocity components are given by Eqs.3B.7-2 to 4.
The convective momentum flux tensor is given by vv , where vv is the
dyadic product of v with v. The components of vv in Cartesian
coordinates are shown in the last three columns of Table 1.7-1.
a. What are these components for the flow in Problem 3B.7?
b. What is the convective momentum flux through a plane
perpendicular to the x axis?

a. When we make use of the Cartesian components of the


velocity given in Eqs. 3B.7-2 to 4, we get

2
1 2w x6
( vv )xx = vx vx = +
W
(1)
(x )
4
2
+y 2

2
1 2w x5 y
( vv )xy = vx v y = +
W
(2)
(x )
4
2
+y 2

( vv )xz = vx vz = 0 (3)
2
1 2w x5 y
( vv )yx = v y vx = +
W
(4)
(x )
4
2
+y 2

2
1 2w x4 y2
( vv )yy = vy vy = +
W
(5)
(x )
4
2
+y 2

( vv )yz = v y vz = 0 (6)
( vv )zx = vz vx = 0 (7)
( vv )zy = vz v y = 0 (8)
( vv )zz = vz vz = 0 (9)

b. For a plane perpendicular to the x axis, the vector n is x .


The components of the momentum flux are then

9
2
1 2w x6
x vv x = vx v x = + (10)
W
(x )
4
2
+y 2

2
1 2w x5 y
x vv y = v x v y = + (11)
W
(x )
4
2
+y 2

x vv z = vx vz = 0 (12)

Then x vv x is the amount of x momentum flowing per unit time


through a unit area of surface perpendicular to the x axis, x vv y
is the amount of y momentum flowing per unit time through a unit of
surface perpendicular to the x axis, and x vv z is the amount of z
momentum flowing per unit time through a unit of surface
perpendicular to the x axis.
Verify that the units of the expressions on the right sides of Eqs.
10 and 11 do indeed have the units of momentum per area per time.

10
Note to p. 51

On this page several quantities are obtained from the velocity


distribution of Eq. 2.3-18. We could also ask ourselves: how much
kinetic energy is flowing per unit time in the axial laminar flow of a
fluid in a circular tube?
The volume rate of flow through an element of cross section
rdrd is vz rdrd . The kinetic energy per unit volume of the fluid is
1
2
vz2 , since the only nonzero velocity component is vz . Therefore the
total amount of kinetic energy per unit volume flowing through the
tube is:

0 0 ( 12 vz ) vz rdrd = 2 0 ( 12 vz ) vz rdr
2 R 2 R 2

( )
1 1 3
= 2 12 R 2 0 vz3 d = R 2 vmax
3
0 1 d
2
(1)

In the second line, we have introduced the dimensionless coordinate


= r R and the maximum velocity vmax = (P 0 P L ) R 2 4 L . Now all
that remains is to evaluate the integral:

R 2 vmax
3 1 2
( 4 6
) 2 3 1 3 3 1
0 1 3 + 3 d = R vmax 2 4 + 6 8
4 6 + 4 1 1
= R 2 vmax
3
= R 2 vmax
3
8 8
1
(
= R 2 vmax vmax
4 2
)
1 2
(2)

The last form suggests a volume rate of flow multiplied by a kinetic


energy per unit volume, both quantities evaluated for the maximum
velocity.

11
Note to p. 52

The laminar flow in a circular tube with radius R is discussed in


2.3, and the laminar flow in tubes with equilateral triangular cross-
section of height H is described in Problem 3B.2. Both tubes have the
same length, L. We want to compare these two flow problems.
a. Compare the mass rates of flow for the two tubes when their
cross-sectional areas are the same.
b. Compare the mass rates of flow for the two tubes when the
perimeters of their cross sections are the same.

a. For flow in circular and triangular tubes we have for the


mass flow rates (see Eq. 2.3-21 and Eq. 3B.2(b)):

(P 0 P L ) R 4 3 (P 0 P L ) H 4
w = w = (1,2)
8L 320 L

In Eq. 1, R is the tube radius; in Eq. 2, H is the height of the triangular


cross section, and 2H 3 is the length of a side of the triangle. To
make the comparison, we need to express the flow rates in terms of
the cross-sectional areas. Since for circular tubes A = R 2 , and for
1 2
equilateral triangular tubes, A = H ,
3

w =
( )
(P 0 P L ) A 2 2
; w =
3 3 (P 0 P L ) A 2
(3,4)
8L 180 L

Therefore

w 3 3 (16 )
= 8 = 0.726 (5)
w 180

b. The perimeters of the two tubes are P = 2 R for circular


tubes, and P = 2 3H for triangular tubes. Therefore

12
(
3 (P 0 P L ) P 2 3 )
4
(P 0 P L ) ( P 2 )
4

w = w = (6,7)
8 L 180 L

Taking the ratio, as before

w
=
3

( 8)( 2 ) = 0.265
4

(8)
( )
2 3 (180 )
4
w

For the square cross section (see Problem 3B.3), the ratios
corresponding to Eqs. 5 and 8 may be found to be

w
= 0.884 (same cross-sectional areas) (9)
w

w
= 0.545 (same perimeters) (10)
w

What, if any, conclusions can you draw from this problem?

[The triangular duct problem is discussed on p. 58 of Landau and


Lifshitz, Fluid Mechanics, Addison Wesley (1959); our H is their a
multiplied by 12 3 .]

13
Note to p. 54

Let's check a few things about Fig. 2.4-1.


a. There the graph shows the transport of z-momentum in the
positive r-direction. This quantity is negative when r < R , positive
when r > R , and 0 when r = R , where is defined by Eq. 2.4-12.
We need to verify that this graph is consistent with Eq. 2.4-13.
b. The figure also shows the velocity distribution for flow in an
annulus, as given in Eq. 2.4-14. What is the location of the maximum
velocity? Show that the position of the maximum is nearer the inner
cylinder of the annulus.
c. What is the velocity at the maximum in the curve?

a. Eq. 2.4-13 may be rewritten as

(P P L ) R r 2 R (P0 PL )R 2 R
2
R
rz = R r = 1
0
(1)
2L 2L r r

If r R < , then the bracket in Eq. 2b.6-1 is negative, whereas if


r R > , then the bracket in Eq. 2b.6-1 is positive. This is in
agreement with the graph of rz vs. r in Fig. 2.4-1.

b. To find the maximum of the expression in brackets in Eq. 2.4-14, as


a function of r R = s , we have to differentiate [ ] with respect to s as
follows 9 (alternatively, one may set rz equal to zero):

d 1 2 1 1 2 1
1 s ln = 0 2s +
2
(2)
ds ln (1 ) s ln (1 ) s

Setting the right side equal to zero, and solving for s gives

1 2
smax = (3)
2 ln (1 )

Hence, choosing the plus sign (why?), we get for the location of the
maximum in the velocity curve

14
12
rmax =R (4)
2 ln (1 )

The half-way point between the inner and outer cylinders is given by
s = 12 (1 + ) . Therefore we now have to prove that

1 2 1 2 (1 )
< (1 + ) or <1 (5a,b)
2 ln (1 ) 2 (1 + ) ln (1 )
It is easy to calculate the left side of Eq. 5b as a function of :

2 (1 ) 2 (1 )

(1 + ) ln (1 ) (1 + ) ln (1 )
0.1 0.711 0.6 0.980
0.2 0.829 0.7 0.991
0.3 0.895 0.8 0.999+
0.4 0.935 0.9 0.999+
0.5 0.962

It is evident that for > 0.8, the maximum is very close to being half
way between the two cylinders, and that is to be expected, inasmuch
as the annular-slit flow approaches a flat-slit flow. Problem 2B.5 gives
a discussion of the interrelation of the flow in a plane slit and the
flow in a narrow annulus.

c. The maximum velocity is then

vz,max =
(P 0 P L ) R 2 rmax
1
2


1 2 R
ln
4 L ln (1 ) rmax
R

=
(P 0 P L ) R2
1
1 2

1 2
ln
2 ln (1 )
(6)
4 L 2 ln (1 ) ln (1 ) 1 2

This result may also be written in terms of , as in Eq. 2.4-15.

15
Note to p. 55

It is good practice to check limiting cases whenever possible.


For example, one should show that Eq. 2.4-17 becomes the Hagen-
Poiseuille formula for tube flow (Eq. 2.3-21) in the limit that 0 .
(See the comment in the paragraph that begins four lines after Eq. 2.4-
14.)

Solution: We have to show that the bracket quantity in Eq. 2.3-21


becomes equal to unity when 0 . In this limit, the various terms
inside the bracket become:

0
( )
lim 1 4 = 1 (1)

lim (1 )
2
2
=1 (2)
0
l im (1 ) = (3)
0

Thus the bracket quantity becomes:

) ( )
1 2 lim
2
1

(
1
4
ln (1 ) 0 1

=1 (4)

and the Hagen-Poiseuille formula is recovered.

16
Note to p. 58

Let us verify that the average velocities in the two regions are
given by Eqs. 2.5-20 and 21.
In region I, the average velocity is given by

(p pL ) b2 1 0 2 I I II x x
2

b b I + II I + II b b
v I
= 0
+ dx
2 I L
z

(p pL ) b 2
2 I I II
0
= 1 I + II I + II
+ d where = x b
0 2

2 I L

=
( p0 pL ) b 2 2 I 1 I II 1
I
I II II
2 I L + 2 + 3

=
(
I
(
p0 pL ) b 2 6 2 3 2 +
I II
) (I II


)
2 I L I + II

=
( p0 pL ) b 2 7 I + II
(1)
I II
2 I L +

Similarly for Region II we have

(p
pL ) b 2 1 b 2 II I II x x
2

2 II L b 0 I + II I + II b b
v II
z = 0
+ dx

( p0 pL ) b 2 1 2 II I II
2 II L 0 I + II I + II
= + 2
d

=
( p0 pL ) b 2 II 1 I II 1
2

+

2 II L I + II 2 I + II 3

=
(
I
(
p0 pL ) b 2 6 2 + 3 2 +
II II I
) ( II


)
2 II L I + II

=
( p0 pL ) b 2 I + 7 II
(2)

2 II L I + II

17
Note to p. 59

The verification of some equations requires a lot of algebraic


detail that falls in the category of "straightforward but tedious."
Nonetheless, such derivations should be done. Here are two
examples:

a. Verify the expressions for the stress components rr and r in Eqs.


2.6-5 and 2.6-6. From Eqs. B.1-8 and 2.6-1, we get

1 3 R 3 R
2 4
vr
rr = 2 = 2 v + cos
r R 2 r 2 r

3 v R R
2 4

= + cos (1)
R r r

And from Eq. B.1-11 and Eqs. 2.6-1 and 2, we find

1 R 3 R R
2 4
v 1 vr
r = r + = v sin
r r r R r 2 r r

1 R 3 R 1 R
2 4

v ( sin )
R r 2 r 2 r

4 4
3 v R 3 v R
=+ r sin = + r sin (2)
2 R 2 R

This is the "form drag" result given in Eq. 2.6-8.

b. Show how Eqs. 2.6-9 and 2.6-12 are obtained by doing the
necessary integrations.
To get the normal force acting on the sphere (from Eq. 2.6-7),
we start by noting that rr on the surface of the sphere is zero. (This is
a special case of the general result given in Example 3.1-1.) The
pressure p on the surface of the sphere is given in Eq. 2.6-8. Therefore
the z-component of the force acting normal to the surface of the
sphere is:

18
3 v
F( ) = cos (cos ) R 2 sin d d
2
0 0 p0 + gRcos + 2
n

R
3 v
= 2 0 p0 + gRcos + cos ( cos ) R 2 sin d
2 R
3 v
= 2 p0 R 2 0 cos sin d + 2 R 2 gR + cos2 sin d (3)
2 R 0

The first integral is zero, and the second is 2/3, so that

3 v 2 4
F ( ) = 2 R 2 gR +
n
= R3 g + 2 Rv (4)
2 R 3 3

To get the z-component of the tangential force acting on the


sphere, we substitute Eq. 2.6-11 into Eq. 2.6-10 and integrate:

3 v
F (t) = 2 0 sin ( sin ) R 2 sin d
2 R
4
= 3 Rv 0 sin 3 d = 3 Rv = 4 Rv (5)
3

This is the "friction drag" result displayed in Eq. 2.6-12.

19
Note to p. 78

The general result for normal stresses at solid surfaces given in


Eq. 3.1-6 is very important. We have already seen in Eq. 2.6-5 that the
normal stresses for creeping flow around a sphere are exactly zero at
the sphere surface, r = R. Verify that Eq. 2.6-5 is correct for rr , ,
and by using Eqs. B.1-15, 16, and 17 and Eqs. 2.6-1, 2, and 3.

Solution:
a. First, we get rr from Eq. B.1-15:

1 3 R 1 R
2 4
vr
rr = 2 = 2 v + 3 cos
r R 2 r 2 r

3 v R R
2 4

= + cos (1)
R r r

b. Then we get from Eq. B.1-16:

1 R 3 R 1 R
2 4
1 v vr
= 2 + = 2 v + + cos
r r R r 4 r 4 r

1 R 3 R 1 R
2 4

2 v + cos
R r 2 r 2 r

2 v 3 R 3 R
2 4

= + cos
R 4 r 4 r

3 v R R
2 4

= + cos (2)
2R r r

c. And, finally, we get from Eq. B.1-17:

20
1 v vr + v cot
= 2 +
r sin r
1 R 3 R 1 R R 3 R 1 R
3 3

= 2 v 1 + cos + 1 + + cos
R r 2 r 2 r r 4 r 4 r

3 v R2 R4
= + cos (3)
2R r r

When r = R, all of the normal stresses are zero, in agreement with


Example 3.1-1.

21
Note to p. 81

Here we give the details of the derivation of Eq. 3.3-1 from Eq.
3.2-9. Although Eq. 3.3-1 itself is not much used, the integral of Eq.
3.3-1 over large flow systems is widely used. We call this the
"macroscopic mechanical energy balance"; the term "engineering
Bernoulli equation" is also used.
The derivation we work through here is an excellent exercise in
using some of the "del" relations given in Appendix A (see
particularly A.4).
We start by forming the dot product of the local velocity v with
Eq. 3.2-9. The last term presents no problems:

( v g ) = (v g ) (1)

The term involving [ ] may be rearranged using Eq. A.4-29 in


Example A.4-1:

( ) ( )
v [ ] = [ v ] + ( : v ) (2)

The term containing p may be similarly rearranged by using Eq.


A.4-19:

( v p ) = ( pv ) + p ( v ) (3)

We now tackle the remaining two terms by first putting both of


them on the left side of the equation:


(
v t v + v [ vv ] )

t t
( )
= v v + ( v v ) + v [ v v ] + ( v v ) ( v ) (4)

------------- ----------------

In the first term, we differentiate the product with respect to t, and in


the second term, we use Eq. A.4-24. In the second line, the dashed
underlined terms then sum to zero by using the equation of

22
continuity; furthermore, the first term is split up into two terms, and
the third term is rearranged, thus:

1 1
=
t 2 2 t
(
( v v ) ( v v ) + v [ v v ] ) (5)

We again use the equation of continuity to rewrite the second term in


Eq. 5:

1 1
=
t 2
( v v ) ( )
+ 2 ( v v ) ( v ) + v [ v v ] (6)

Now the second and third term may be combined by using Eq. A.4-
19 with s replaced by 12 ( v v ) and v replaced by v :

1 1
= ( v v ) + ( v v ) v
t 2 2
1 2 1 2
= v + v v (7)
t 2 2

Clearly knowing that the final result is Eq. 3.3-1 is very helpful in
doing the last several steps.

23
Note to p. 82 (i)

We want to verify that ( :v ) , when written for a Newtonian


fluid, may be written as a sum of squares as shown in Eq. 3.3-3, and is
hence positive.
First define a tensor = v + ( v ) , and then ( :v ) may be

written for Newtonian fluids (for which is symmetric) as:

( :v ) = 12 ( : ) = 12 ( 23 ( v ) : ) + 12 ( v )(: ) (1)

where Eq. 1.2-7 has been used. Next, since ( : ) = 2 ( v ) , we get

( :v ) = 12 ( : ) 43 ( v )2 + ( v )2 (2)

This is equivalent to:

( :v ) = 12 ( 23 ( v ) : 23 ( v ) ) + ( v )2 (3)

which is another way of writing Eq. 3.3-3. The last step may be seen
to be true by expanding the double-dot product in Eq. 3 thus:

( 2
3( v ) : 23 ( v ) )
= ( : ) 43 ( v ) 43 ( v ) + 49 3 ( v )
2 2 2

= ( : ) 43 ( v )
2
(4)

which is the coefficient of 1


2
in Eq. 2.

24
Note to p. 82 (ii)

Eq. 3.4.1 is obtained by taking the cross product of the position


vector with the equation of motion in Eq. 3.2-9. To do this, we form
the cross product, term by term:

a. The time-derivative term is straightforward; for the ith component:


r t v = t [ r v ]i (1)
i

because the position vector r is independent of the time t.

b. For the next term in the equation, we consider only the i


component and expand the expression in terms of its components:


r [ vv ] = ijk x j vl v k
l xl
i
j k

= ijk x j vl vk ijk vl vk jl
j k l xl j k l


= v x v v v (2)
j k l xl l ijk j k j k ijk j k

In the second line, we have moved the x j inside the differentiation


and subtracted off a compensating term. In the third line, we have
rearranged the first term and performed the sum on l in the second
term. It can be seen that the second term is zero (inasmuch as it
involves a double sum on a pair of indices that appear symmetrically
in one factor ( v j vk ) and antisymmetrically in another ( ijk ); see
Exercise 5 on p. 815). Now we convert Eq. 2 back to bold-face
notation:

r [ vv ] = v [ r v ] (3)
i i

c. Next we examine the term containing the pressure:

25

[r p]i = ijk x j x p = ijk x p ijk jk p
xk j
(4)
j k k j k j k

The last term is zero, since it involves a double sum on a pair of


indices that appear symmetrically in one factor and antisymmetric in
another. The other term can be rearranged as follows:


x
j k l

l
{ }
ijk x j p kl =
l

xl
ijk x j p kl
j k
(5)

On the right side, the quantity within the braces can be recognized as
the cross product of a vector with a tensor (see text just after Eq. A.3-
19). Therefore the above result in Eq. 5 can be written as:


x {r p }il = x {r p }li = {r p }

(6)
l l l l
i

In order to write Eq. 6 as the divergence of a tensor, the indices must


be as in Eq. A.4-13. This requires introducing the transpose of the
cross product as indicated above.

d. The term containing the tensor can be treated in somewhat the


same manner as in part (c) above:


r [ ] = ijk x j [ ]k = ijk x j (7)
i
j k x lk j k l l

Next we write this intermediate result as the sum of two terms:


ijk x x j lk ijk lk x
xl j
j k l l j k l



= ijk j kl ijk lk jl
x
l xl
j k j k l

{ }
{ }

= r ijk jk = r + ikj jk
l xl l xl
il li
j k j k

26
{
= r } + [ : ]

(8)
i
i

If the stress tensor is symmetric, the [ : ]i term vanishes, and may


be replaced by in the first term.

e. The ith component of the external force term, [r g ]i , is


straightforward.

When all the terms are collected, Eq. 3.4-1 results.

27
Note to p. 86

In Example 3.5-1 it was pointed out that, to derive the Bernoulli


equation, we need the vector identity:

( v v [ v ]) = 0 (1)

To show that this relation is true, we use A.2 and Eq. A.4-10.
The proof requires replacing the dot and cross operations by
their expressions in terms of vector components. This is most
efficiently done by making use of summations. First we write the dot
product in terms of the components:

( v v [ v ]) = v v [ v ]
i
i i
(2)

Next we write the cross product operations using the ijk symbol:


= vi ijk v j [ v ]k = v i ijk v j klm v (3)
i j k i j k l m xl m

Then we rearrange the expression and make use of the cyclic


property of the permutation symbol, i.e., klm = mkl = lmk :


= ijk klm vi v j vm = ijk lmk vi v j v (4)
i j k l m xl i j k l m xl m

We can now use Eq. A.2-7 to replace the sum on k of products of two
permutation symbols:

i j l m
( )
= il jm im jl vi v j

v
xl m
(5)

After doing the sums on l and m we get two terms:

28

= vi v j v j vi v j v (6)
i j xi i j x j i

In the second summation, we replace i by j and j by i to get:


= vi v j v j v j vi v =0 (7)
i j xi i j xi j

It may now be seen that both terms are the same. Consequently their
difference is zero. Therefore, we have proven the identity in Eq. 1.

29
Note to p. 90

An alternative method of solving the problem in Example 3.6-3


is given here. Eq.3.6-21 may be set up by making a shell torque
balance on a shell of thickness r and height L. Then the torque at
radius r is equal to 2 rL r , whereas the torque at radius r + r
r r
is 2 ( r + r ) L r r +r ( r + r ) . The torque is (force per unit area) x
(area) x lever arm. When these torques are equated, we get after
dividing by r and letting r go to zero:

d 2 d v
d 2
dr
( )
r r = 0 or r r =0
dr dr r
(1a,b)

Eq. B.1-11 has also been used. Here we show that this is equivalent to
Eq. 3.6-21, which is

d 1 d d 2 v 1 dv v
( rv
dr r dr
) =0 or
dr 2
+
r dr r 2
=0 (2)

We start by performing the differentiations in the large


parentheses in Eq. 1b:

d v 1 dv v dv
r 2 r = r 2 r 2 = r 2 rv (3)
dr r r dr r dr

Next do the differentiation with respect to r (in Eq. 1b) and set the
result equal to zero:

2
d 2 dv dv 2 d v dv
r rv = 2r + r r v
dr dr dr dr 2 dr
2
2 d v 1 dv v
=r 2 + 2 =0 (4)
dr r dr r

30
Therefore either r 2 is zero, or the quantity in parentheses is zero. But
r 2 cannot be zero, and hence Eq. 2which came from Example 3.6-
3must be the same as Eq. 1b (from the torque balance).

31
Note to p. 117

When you look at Fig. 4.1-2, you might wonder whether the
slope at y = 0 is 1 . You can answer that question by differentiating
Eq. 4.1-15 with respect to :

d d 2 2
d
erfc =
d
1

0 e d
=0
=0

2 2 2
= e = = 1.1287 (1)
=0

Here we have used the Leibniz formula for differentiating integrals in


C.3 and the definition of the complementary error function in C.6.
As may be seen from Fig. 4.1-2, the slope is somewhat steeper than
minus 1.
Notice that in Eq. 4.1-14, as well as in Eq. 1 above, we used a
bar over the to make a distinction between the variable of
integration and the upper limit on the integral. When applying the
Leibniz theorem, it is vital to make this distinction.

32
Note to p. 121

Example 4.1-3 should be straightforward, except possibly for


(a) the line immediately after Eq. 4.1-49, and (b) the line following Eq.
4.1-53. Here we show how to obtain the expressions given in these
two locations.

(a) Let the real and imaginary parts of the arguments be:
for z1 : z1r and z1i
for z2 : z2r and z2i
for w: wr and wi
Then the expression { z1 w} = {z2 w} becomes

{ } { }
( z1r + iz1i ) ( wr + iw i ) = ( z2r + iz2i ) ( wr + iwi ) (1)

Then, taking the real parts of both sides, we get

z1r wr z1i wi = z2r wr z2i wi (2)

Rearranging gives:

wr ( z1r z2r ) = wi ( z1i z2i ) (3)

Since w, and hence wr and wi , are arbitrary, the only way that this
equation can be satisfied is if z1r = z2r and z1i = z2i (i.e., z1 = z2 ).

(b) The square root of i will be some number in the complex


plane, say a + bi , so that

i = a + bi (4)

where the real quantities a and b must be determined. When the left
and right sides of Eq. 4 are squared, we get

( )
i = a 2 b 2 + 2abi (5)

33
To find a and b, we equate the real and imaginary parts of the left and
right sides:

a2 b2 = 0 and 2ab = 1 (6)

Eliminating b between these two equations gives

1
a4 = (7)
4

Hence

1 1 1 1
a2 = , and a= , i (8)
2 2 2 2

whence

1 1 1
b= , (9)
2 2 i

and the two square roots of i are

1
i = a + bi = (1 + i) (10)
2

Alternatively, one can write i in polar form and use the fact that
i has a unit length:

i = re i = 1e i 2
(11)

Then the square root of i is

1 1
i = e i 4
= cos + i sin = +i (12)
4 4 2 2

34
Note to p. 123

Here we show how Eq. (A) of Table 4.2-1 is obtained.

The velocity is of the form

v ( x, y ) = x vx ( x, y ) + y v y ( x, y ) (1)

The velocity components are related to the stream function according


to the relations in the 3rd column.
The first term in Eq. (A) comes from the first term in Eq. 4.2-1,
which is

v y vx
t
[ v ] z t [
= v ]z z t x y
=

2 2
= z 2 + 2 = z 2 (2)
t x y t

The right side of Eq. (A) comes from the right side of Eq. 4.2-1:

2 2
[ v ] = 2 + 2 [ v ]
2

x y
2 2 2 2
= z 2 + 2 2 + 2 = z 4 (3)
x y x y

The second term on the left side of Eq. (A) can be taken care of
similarly:

v [ v ] = v z 2

( )
= x vx + y v y z 2

= y vx 2 + x v y 2

35
x y z x y z

=
x y z x y z
0 v x 2 0 v y 2 0 0

= + z

x
(
vx 2 + z )

y
v y 2 ( )
v v y 2
= z x + + z vx 2 + v y 2 (4)
x y x y

The first term is zero, because of the assumption of incompressibility


(see statement just above Eq. 4.2-1). Then we get finally

2 2
= z +
y x x y

=
x y
=
(
, 2 ) (5)
2 2 ( x, y )

x y

When the results in Eqs. 2, 3, and 5 are combined, we get same


expression that is given in Eq. (A) of the table.

36
Note to p. 125

Here we want to fill in the details of getting Eq. 4.2-20 from Eqs.
4.2-18 and 4.2-19. This is a "straightforward but tedious exercise."
We begin by evaluating each of the four squared terms in Eq.
4.2-19 using the velocity components in Eqs. 4.2-13 and 14; it is
convenient to introduce the dimensionless variable = r R .

2
vr
( )
2
9 v 2
2 = 1 3 cos (1)
2
r 2r

2
1 v vr 2v2
( )
2
2 + = 2 1 + 34 1 + 14 3 cos + 1 32 1 + 12 3 cos
r r r
9 v2
( )
2
= 1 + 3 cos (2)
2
8r

2
v cot vr 2v2
( )
2
2 + = 2 1 + 34 1 + 14 3 cos + 1 32 1 + 12 3 cos
r r r
9 v2
( )
2
= 1 + 3 cos (3)
2
8r

2 2
v 1 vr v v 1 vr
r +
r r r = r r + r
2
3 1 + 3 3 sin + 1 3 1 1 3 sin
v2 4 4 4 4
=
r 1 3 1 + 1 3 sin
2

2 2
v2 2
= 3 3 sin (4)
2 2
r

When the above results are combined, we get

( :v ) r 2 = v2 274 2 27 4
2
+ 27 6
4
cos2 + v2 94 6 sin 2
(5)

37
Then the kinetic contribution to the force on the sphere is given by

R ( :v ) r

Fk v = 2 0 2
drsin d

= 2 23 v2 R 274 2 27 4
2
+ 27 6
4
dr + 2 43 v2 R 94 6 dr
(6)

0 cos sin d = 0 sin d = 43 . Then, finally
2 2 3
since 3
and


Fk = 9 v R 1 2 2 4 + 6 d + 6 v R 1 6 d

= 9 v R 1 + 32 3 15 5 + 6 v R 15 5
1 1

= 9 v R 1 + 2
3
1
5
+ 6 v R 15
= 9 v R 52 + 6 v R 15 = ( 24
5
+ 6
5 ) v R

= 6 v R (7)

which is Stokes' law.

38
Note to p. 139

We show here how to get the Falkner-Skan equation in Eq. 4.4-35 for
the flow near a corner. For this system, the external flow ve was
found earlier (see Eqs. 4.3-42 and 43) to be

2c ( 2 )
ve ( x ) = c' x ( )
2
x (1)
2

We then have to solve Eq. 4.4-11 to get the velocity distribution in the
neighborhood of the wedge shown in Eq. 4.3-4:

vx vx dve 2 vx
vx + vy = ve + (2)
x y dx y 2

This equation can be rewritten in terms of the stream function


( x, y ) , by using the expressions for the velocity components in the
first row of Table 4.2-1:

2 2 dve 2
y xy + x y 2 = ve dx + y 2 y (3)

Insertion of Eq. 1 into this equation then gives Eq. 4.4-32:

2 2 c' 2 1 3
= (4)
y xy x y 2 2 x( 23 ) ( 2 ) y 3

Next we want to rewrite this equation in terms of f and :

( x, y ) = c' ( 2 )x ( ) f ( ) Ax ( ) f ( )
1 2 1 2
(5)

c' y y
( x, y ) = B (6)
( 2 ) x( ) ( )
1 2
x( ) ( )
1 2

39
We start by converting the various derivatives from ( x, y ) to
derivatives of f ( ) :

1 2 df 1
= Ax ( ) = ABx ( ) f ' (1 ) ( 2 ) = ABx ( ) f ' (7)
1 2 2
y d y x
2 ( )
2
( 2 ) 2 x 1
= ABx f " = AB (1 ) ( 2 ) f " = AB2 (1 2 ) ( 2 ) f " (8)
y 2
y x x
3 1 1
= AB 2 (1 2 ) ( 2 ) f = AB3 ( 23 ) ( 2 ) f (9)
y 3
x y x
d 1 2 1 2 df
= A x ( ) f Ax ( )
x dx d x
x ( )1 1 2 By (1 ) (1 ) ( 2 )1
1 2
= A f Ax ( ) x f'
(2 ) x ( 2 )

= A
1 x ( )
1 2
f +A
(1 ) x1 ( 2 )
f' (10)
(2 ) x (2 ) x
2 1 x ( )
1 2
( 1 ) x1 ( 2 ) d
= A f '+ A ( f ')
yx ( 2 ) x y ( 2 ) x y d
1 x ( )
1 2
( 1 ) x1 ( 2 )
= AB f '+ AB ( f "+ f ') (11)
( 2 ) x(1 ) ( 2 ) x ( 2 ) x(1 ) (2 ) x
The terms on the right side of Eq. 4.4-32 are then:

c'2 1 1 c'2 1
2 ( 23 ) ( 2 ) + AB ( 23 ) ( 2 ) f = 2 ( 23 ) ( 2 ) ( + f )
3

x x x
(12)

Next we write down the terms on the left side of Eq. 4.4-32:

1 x ( )x ( ) 2
2 1 2
2 2
=A B
2 2
f
y xy x y 2 ( 2 ) x(1 ) ( 2 ) x

40
(1 ) x (2 ) x1 ( 2 ) f f "+ f 2
( 2 ) x(1 ) ( 2 ) x ( )
A B2 2

x ( )
1 2
1
A B2 2
ff
( 2 ) x(12 ) (2 ) x
+A B2 2 (1 ) x1 ( 2 ) f f "
( 2 ) x(1 2 ) ( 2 ) x
1
x (
23 ) ( 2 )
x (
23 ) ( 2 )
= c'2 f 2 c' 2 ff
2 (2 )
= c' 2
1
(2 )
x (
23 ) ( 2 )
( f 2 ff ) (13)

Combining the results in Eqs. 12 and 13 gives

c' 2
1
(2 )
x (
23 ) ( 2 )
( )
f 2 ff = c' 2
1
(2 )
x (
23 ) ( 2 )
( + f )
(14)
or

f 2 ff = + f (15)

which is the same as Eq. 4.4-35, the Falkner-Skan equation.

[Note: In earlier printings of the textbook, the prime was omitted from
c', and the quantity c' was not defined.]

41
Note to p. 154

To get the average flow velocity from Eq. 5.1-4, we integrate the
velocity distribution over the circular tube cross-section:

1 ( r R )
2 R 2 R 17

vz =

0 0
v z rdrd
= vz ,max

0 0
rdrd
2 R 2 R
0 0
rdrd
0 0
rdrd
2
1 ( r R ) rdr = 2vz ,max [1 ] d
R 17 1

17
= vz ,max (1)
R2 0 0

where = r R . To evaluate the integral, we make a change of variable


1 = . Then

v z = 2vz ,max 1 7 (1 )d
1
(2)
0

This can then be written as the sum of two integrals, which can be
evaluated:

( )
1
1 1 8 7 15 7
v z = 2vz ,max 0 d 0 d = 2vz ,max 8 7 15 7 = 0.82vz ,max
17 87

0
(3)

Since 0.82 is approximately 4/5, relation in Eq. 5.1-5 is verified.

42
Note to p. 170

Here we fill in some of the missing steps following Eq. 5.6-18.


Setting C1 = 0 in Eq. 5.6-18 and rewriting the equation, we get:

FF = ( F + F ) 2F (1)

where the primes indicate differentiation with respect to . Next we


note that Eq. 1 may be put into the form

1
2 ( F ) = (F ) 2F
2
(2)

Each term may now be integrated with respect to to give

1
2
F 2 = F 2F C2 (3)

which is the same as Eq. 5.6-19. The constant C2 is zero according to


Eq. 5.6-16, with a, b, and d set equal to zero. [Note: The comment
about setting = ln does not seem to be helpful.]
Eq. 3 is a separable, first-order differential equation

dF dF d
= 2F + 12 F 2 or = (4)
d (
2F 1 + 14 F )

Then, according to a table of integrals (e.g., Formula 101.1 of


Dwight's Table of Integrals), Eq. 4 gives, on integration

1 + 14 F 1 + 14 F
ln1
2
= ln + ln C3 or ln = ln + ln C3 (5a,b)
F F

Next take the antilog of the equation and then square the result to
obtain

43
F
1 + 14 F
= ( C3 ) 2
(6)

Now either a "plus" sign or a "minus" sign may be inserted inside the
absolute value bars. Since we have no reason to prefer one over the
other, we consider the two cases separately:

Case I (plus sign):


( C3 )
2

( )
F = + 1 + 14 F ( C3 )
2
or F= (7)
1 14 ( C3 )
2

Case II (negative sign):


( C3 )
2

( )
F = 1 + 14 F ( C3 )
2
or F= (8)
1 + 14 ( C3 )
2

When F in Eq. 7 is plotted against ( C3 ) , it tends toward


2

infinity as ( C3 ) = 4 is approached from below, and approaches


2

minus infinity when approached from above. Hence Case I is


physically unreasonable behavior for the stream function.
When F in Eq. 8 is plotted versus ( C3 ) , is it monotone
2

decreasing over the entire range of ( C3 ) . Since this is physically


2

reasonable, we choose the solution in Case II (which agrees with Eq.


5.6-20 in the textbook).
When Eq. 8 (or Eq. 5.6-20) is inserted into Eq. 5.6-12 and 13, Eqs.
5.6-21 and 22 follow immediately.
Then substitution of Eq. 5.6-21 into the expression for J in Eq.
5.6-2 gives:

( 2C )
2
2

J = 2 0 v z2 rdr = 2 ( )
3
0
t 2
4
d (9)
1 +

1
4 (C3 ) 2

We now let 1
4 (C3 )2 = x 2 so that d = ( 4 C32 ) xdx ; then

44
xdx
J = 32 ( ) C32 0
t 2

( )
4
1 + x2


( t)2 2 1 = 16 ( t )2 C 2
= 32 C3 (10)

( )
3 3
3
6 1 + x
2
0

whence Eq. 5.6-23 follows at once.


Similarly the mass rate of flow is

w = 2 0
( )
t
( 2C ) 2
3
vz rdr = 2 0 2
d z 2
z 1 +

1
4 (C3 )2

4 xdx 1
= 4 ( ) zC32 (
t)
= 16 z
t

C2 0
3 (1 + x ) 2
2
2 1 + x2
( )
0
= 8 ( ) z
t
(11)

45
Note to p. 198

Here we show how to obtain the macroscopic mass and momentum


balances from the corresponding equations of change.

Macroscopic Mass Balance

The equation of change for conservation of mass is given in Eq.


3.1-4. We want to integrate this equation over the system pictured in
Fig. 7.0-1 on p. 197:


V (t) t dV = V (t) ( v ) dV (1)

We now apply the 3-dimensional Leibniz formula (Eq. A.5-5) to the


left side and the Gauss divergence theorem (Eq. A.5-2) to the right
side:

d
dV S(t) ( n vS ) dS = S(t) (n v ) dS
dt V (t )
(2)

in which n is the outwardly directed unit vector on the surface S (t ) .


The surface is a function of time t, because there may be moving parts
in the system. Eq. 2 may be rewritten as

d
(
dV = S(t) n ( v vS ) dS
dt V (t )
) (3)

We note that the mathematical surface S (t ) defining the system


consists of several parts that we identify as follows:
the "inlet" surface S1 (on which v S = 0)
the "outlet" surface S2 (on which v S = 0)
the "fixed" surface S f (on which v = vS = 0 )
the "moving" surface Sm (on which v = vS 0 )
with v being the fluid velocity and v S the surface velocity. The
surface integrals are then split into four parts corresponding to the
four types of surfaces.

46
The integral on the left side is the total mass mtot in the system.
The surface integrals over S f and Sm are zero, because v = v S .
Therefore we are left with

d
m = S ( n v ) dS S ( n v ) dS (4)
dt tot 1 2

We now introduce the vectors u1 and u 2 , which are unit vectors in


the direction of flow at planes "1" and "2", respectively. Thus the n in
the S1 integral will be u1 , whereas the n in the S2 integral will be
+ u 2 . Now we make the assumptions that (i) the density is a constant
over the cross section, and (ii) that the velocity is always parallel to
the walls of the entry and exit tubes, so that v = u at plane "1" and
v = u at plane "2". Then Eq. 4 becomes

d
m = + 1 S vdS 1 S vdS (5)
dt tot 1 2

Here v is the velocity in the direction of flow, which varies across the
cross section. Therefore integrations over the surfaces S1 and S2 give

d
m = 1 v1 S1 2 v 2 S2 = w1 w2 (6)
dt tot

where is the average value over the cross section; w1 and w2 are
the mass rates of flow at the inlet and outlet, respectively. Eq. 6 is just
the same as Eq. 7.1-1 in the textbook, which was written down by
using elementary arguments (i.e., common sense).

Macroscopic Momentum Balance

When the equation of motion of Eq. 3.3-9 is integrated over the


volume of the flow system in Fig. 7.0-1, we get


V (t) t v dV = V (t) [ vv ]dV V (t) pdV V (t) [ ]dV + V (t) gdV
(7)

47
Next apply the Leibniz formula to the left side and the Gauss
divergence theorem (or Eq. A.5-2) to the surface integrals on the right
side to get

d
vdV S(t) v ( n v S ) dS = S(t) [ n vv ]dS S(t) npdS S(t) [ n ]dV
dt V (t )
+ V t gdV (8)
()

The integral on the left side is just the total momentum Ptot in the
flow system. If g does not change over the volume of the flow system,
then it may be removed from the integral, which gives mtot . Then

d
P = S t n ( v v S ) v dS S t npdS S t [n ]dS + mtot g (9)
dt tot () () ()

We now consider the three surface integrals seriatim:


The first integral is zero on fixed and moving surfaces, and v S
is zero at the entry and exit planes, so that

S(t ) n ( v v S ) v dS = + S u1 u1u1 v 2 dS S u 2 u 2 u 2 v 2 dS
1 2

= + 1 v12 S1u1 2 v22 S2 u 2 (10)


Here it has been assumed that the flow at the inlet and outlet planes is
parallel to the container wall.
The second integral will contribute both at the inlet and outlet
planes and also to the force on the various solid surfaces:

S(t ) npdS = + S u1 pdS S u 2 pdS S +Sm


npdS
1 2 f

= p1S1u1 p2S2 u 2 F(f s


) p
(11)

the last contribution being the force exerted by the fluid on the solid
surfaces by the pressure.
Finally the third integral will be

S(t ) [ n ] dS = + S u1 dS S u 2 pdS S +S [ n ] dS
1 2 f m

48
F(f s
)

(12)

Note that we have omitted the contributions at the entry and exit
planes because they would normally be quite small compared to the
pressure terms. Therefore we are left with just the force exerted by
the fluid on the solid surfaces because of the viscous forces.
When all the forces are combined we get for the macroscopic
momentum balance (with F (f s
p)
+ F (f s
)
= F f s = Fs f ):

d
Ptot = 1 v12 S1 u1 2 v22 S2 u 2 + p1S1u1 p2S2 u 2 + Fs f + mtot g (13)
dt

This is the same as Eq. 7.2-11 (or 7.2-12) in the textbook, obtained by
elementary reasoning.

[Note: The derivation of the macroscopic mechanical energy balance is


given in 7.8, the derivation of the macroscopic energy balance is given
in the Note to p. 454.]

49
Note to p. 199

Here we work through all the details of Example 7.1-1, (a)


explaining how to get Eq. 7.1-4, (b) giving the details of how the
difference in the modified pressures is obtained, and (c) working
through the algebraic details of the remainder of the problem.

a. (This development was given by Professor L. E. Wedgewood,


University of Illinois at Chicago)
We have to find the volume of liquid in the sphere below the
liquid level. We imagine that the sphere is generated by a circle in the
xz-plane, with its center at z = R and x = 0. The tank is draining in the
negative z-direction, and the exit from the sphere into the attached
tube is located at z = 0. The sphere is created by rotating the
generating circle around the z-axis.
The generating circle has the equation:

x2 + ( z R ) = R2
2
or x 2 = 2Rz z 2 (1a,b)

Then we visualize the liquid volume as being made up of a stack of


thin circular disks of thickness dz, each with a volume of

( )
dV = x 2 dz = 2Rz z 2 dz (2)

Then the total volume of the liquid is:

( ) ( ) ( )
h h
V= 0 2Rz z dz = Rz 3 z = Rh2 13 h3
2 2 1 3
(3)
0

Thus the liquid volume is:

1 h
V = Rh 2 1 (4)
3 R

We may check this result at three points where we know the result:
Tank full: h = 2R V = 43 R 3
Tank half full: h=R V = 23 R3

50
Tank empty: h=0 V =0

b. Next we want to apply Eq. 7.1-2 to the system. No liquid is


entering at plane "1" so that w1 = 0 , and, if the diameter of the exit
tube is sufficiently small that the flow in it is laminar, then w2 will be
given by the Hagen-Poiseuille formula (Eq. 2.3-21). Hence

d 2 1 h (P 2 P 3 ) D4
Rh 1 = (5)
dt 3 R 128 L

To get the modified pressures we must specify a datum plane; we


choose this to be at the tube outlet (i.e., plane "3"), so that:

P 2 = p2 + gh2 = ( gh + patm ) + gL (6)

That is, p2 is the pressure due to the liquid in the sphere above plane
"2" plus the atmospheric pressure, and h2 is the height of plane "2"
above the datum plane (i.e., plane "3"). Furthermore,

P 3 = p3 + gh3 = patm + 0 (7)

since p3 is the pressure atmospheric pressure at the tube outlet, and


h3 is the distance upward from the datum plane (i.e., plane "3"),
which is zero. Hence, Eq. 5 becomes

1 h ( gh + gL ) D
4
d
Rh 2 1 = (8)
dt 3 R 128 L

c. Eq. 8 may be rewritten as

R d 2 1 h 3 gD4
h = A (9)
h + L dt 3 R 128 L

(which defines the quantity A) or

51
R h 2 dh
2h =A (10)
h + L R dt

Now we introduce a new variable H = h + L in order to facilitate the


solution of the differential equation

2R ( H L ) ( H L ) dH
2

=A (11)
H dt

or

1 dH
H 2 ( R + L ) + ( 2R + L ) L
H dt
=A (12)

Now we integrate this equation and make use of the initial and final
conditions:

L
1 2
2 H 2 ( R + L ) H + ( 2R + L ) L ln H = At
tefflux
0
(13)
2R+L

This gives

1 2 1 L
L ( 2R + L ) 2 ( R + L ) L + 2 ( R + L ) ( 2R + L ) + ( 2R + L ) L ln
2

2 2 2R + L
= Atefflux (14)
or after some cancellations

1 2 2R + L
tefflux =
A
2R + 2RL ( 2R + L ) L ln
L
(15)

When L2 is factored out of the bracket expression, Eq. 7.1-8 of the


textbook is obtained.

52
Note to p. 242

By working through all the intermediate steps in Example 8.3-1


we get a better idea how to solve problems involving the power-law
model. First we have to obtain the expression for the scalar that
appears in Eq. 8.3-5:

3 3
= 1
2 ( : ) = 1
2 ij ji and = v + ( v )

(1,2)
i=1 j=1

where i and j take on the values 1 = r, 2 = , and 3 = z, since we are


dealing with cylindrical coordinates. The components of in
cylindrical coordinates may be obtained from Eqs. (S) to (AA) in
Table A.7-2. Since the only component of v that is nonzero is the z-
component and since that is a function of r only, the only components
of that we need are the rz- and zr-components,

rz = ( v )rz + (v )zr = vz r + vr z = vz r + 0 (3)


zr = ( v )zr + ( v )rz = vr z + vz r = 0 + vz r (4)

Therefore

= 1
2 ( : ) = ( dvz dr ) = dvz dr
2
(5)

where the minus sign must be chosen, since dv z dr is negative. Then


the shear stress rz will be

n1 n
dv dvz dvz
rz = m z = (6)
dr dr dr

By going through the above procedure, it is guaranteed that, when


we take the fractional powers of the quantities in parentheses (see
Table 8.3-2 for sample values of n), we will not get any imaginary
quantities. When Eq. 6 is substituted into Eq. 2.3-13, we then get:

53
n 1n
dv P PL dvz P 0 P L
m z = 0 r or = r1 n (7,8)
dr 2L dr 2mL

Integration of Eq. 8 gives

r( )
1n 1 n +1
P PL
vz = 0 +C (9)
2mL (1 n ) + 1
Application of the no-slip condition requires that vz = 0 at r = R :

R( )
1n 1 n +1
P PL
0 = 0 +C (10)
2mL (1 n ) + 1
Subtraction of Eq. 10 from Eq. 9 eliminates the integration constant C
and leads to

r ( ) R( )
1n 1 n +1 1 n +1
P PL
vz = 0 (11)
2mL (1 n) + 1
Rearrangement then gives

(P P L ) R r (1 n )+1
1n
R
vz = 0 1 (12)
2mL
(1 n) + 1 R

The mass rate of flow w is then obtained by integrating the velocity


distribution over the tube cross-section:

2 R
w = 0 0 vz rdrd
(P P L ) R
1n

= 2 R 2 0
2mL

R
(1 n ) + 1 (
0
1
)
1 ( ) d
1 n +1
(13)

where = r R . Performing the integration gives

54
(P P L ) R (1 n ) + 1
1n
R
w = 2 R 0
2

2mL
(1 n) + 1 2 (1 n) + 3
(P P L ) R
1n
1
= R 0
3
(14)
2mL (1 n ) + 3
This is the power-law analog of the Hagen-Poiseuille equation for
Newtonian fluids.

55
Note to p. 248

Here we work through the missing steps in Example 8.4-2. In


Eq. 8.4-20, we make the change of variable s = t t after the first line
in order to get a factor e i t to appear explicitly on the right side of the
equation (to match a similar factor on the left side):

d v i t 0 s 1 i s
{ }
2 0
0 i t
i v e = 2 e 0 e e ds (1)
dy 1

Next, we perform the integration over s:

2 0

s i s
{
i v e 0 i t
} d v
= 2 e i t 0
1
1 (1 1 ) i e 1e


dy 0
d 2 v 0 i t 0
= 2 e (2)
dy 1 + i1

We may now remove the real-operator sign from both sides, as well
as the common multiplier e i t , to get:

d 2 v 0 0 d 2 v 0 i (1 + i1 ) 0
i v =0
or = v (3a,b)
dy 2 1 + i1 dy 2 0

Eq. 3b is a differential equation for the complex function v0 ( y ) . The


equation is of the form of Eq. C.1-4, where [] is a2 . Since this is a
complex quantity, we write it as ( + i ) , where and are real, so
2

that we can write the solution as

v0 = Ae (
+ + i ) y
+ Be (
+ i ) y
(4)

Then according to Eq. 8.4-19,

56
{ }
vx ( y,t ) = v0 ( y ) e i t = Ae ({(
+ +i ) y
+ Be (
+ i ) y
) }
e i t (5)

Now we know that the oscillatory disturbance will not propagate


with increasing amplitude as the distance from the wall increases, so
that A will have to be set equal to zero and will have to be positive.
Also we know that the amplitude of the disturbance right at the wall
will be v0 . Therefore B will have to be set equal to v0 . Therefore Eq. 5
can be rewritten as

{
vx ( y,t ) = v0 e ( }
+ i ) y i t
e {
= v0 e y e ( }
i t y )
= v0 e y cos ( t y )
(6)

Now we must find out what and are. To do this, we have to go


back to Eq. 3:

i (1 + i1 )
= ( + i )
2
(7)
0

We next equate the real and imaginary parts of both sides:

1 2
=
2 2
and 2 = (8, 9)
0 0

This gives us two equations for the two unknowns and . We can
eliminate between the two equations and get an equation for :

2
1 1 2

2
= (10)
20 2 0

or, after multiplying through by 2

2
1 2 2
+
4
=0 (11)
0 20

57
This can be regarded as a quadratic equation in 2 , which has the
solution:

2 2
1 2 1 1 2
=
2
+ 4
20 2 0 2
0

1 2 1 2 2
= 1 1 + 2 2 = 1 + (12)
20 1 20 1 1

We now extract the square root of both sides to get:

2 2
12
= 1 + (13)
20 1 1

Because must be real and positive, the sign must be chosen to be


+, and the sign must also be chosen to be "plus." Thus we finally
get:

+ 1 + 2 2
12
=+ 1 (14)
20 1

+ 1 + 2 2
1 2
=+ 1 (15)
20 1

However, we could just as well have chosen both of the signs to be


"minus." Then we would have


12
= 1 +
2 2
(16)
20 1 1


1 2
= 1 + 1 1
2 2
(17)
20

The quantity given in Eq. 16 is still real and positive. So, how do
we make a choice between Eqs. 14 and 15 on the one hand, and Eqs.
16 and 17 on the other? It is easy to show that the former do not
satisfy Eq. 8, whereas the latter do. Therefore, we conclude that Eqs.

58
16 and 17 are the proper quantities to insert in Eq. 6. Thus Eqs. 8.4-24
and 25 of the textbook are correct.

59
Note to p. 255

Since the FENE-P dumbbell model (a molecular model)


describes many of the observed rheological phenomena of polymer
solutions, it is important to understand how we go from the
constitutive equation in Eqs. 8.6-2 to 4 to the expressions for the
rheological properties. Here we show how to get the non-Newtonian
viscosity and the first normal-stress coefficient in steady shear flow.
For the steady shear flow, vx = y , the rate-of-strain tensor
may be displayed as a matrix (see A.9) thus:

xx xy xz 0 0

= yx yy yz = 0 0 0 (1)

zy
zz 0 0 0
zx

Then Eq. 8.6-4 may be written in matrix form as follows:

p,xx p,xy p,xz 2 p,xy p, yy p, yz



Z p, yx p, yy p, yz H p, yy 0 0
p,zy p,zz 0 0
p,zx p, yz
0 1 0
= nKT H 1 0 0 (2)

0 0 0

From this we get at once that the only nonzero components of p are
p,xx and p,xy = p, yx . Then from the matrix equation, Eq. 2, we can
write down the two nonvanishing component equations as:

Z p,xx = 2 p, yx H and Z p, yx = nKT H (3,4)

in which

( ) (
Z = 1 + ( 3 b ) 1 tr p 3nKT = 1 + ( 3 b ) 1 p,xx 3nKT
) (5)

60
In order to make the algebraic manipulations somewhat easier, we
introduce the following dimensionless quantities:

Dimensionless shear stress: S = p, yx 3nKT (6)

Dimensionless normal stress: N = p,xx 3nKT (7)

Dimensionless shear rate: = H (8)

Then Eqs. 3, 4, and 5 become:

ZN = 2S (9)
ZS = 13 (10)
Z = 1 + ( 3 b ) (1 N ) (11)

Use the second of these equations to eliminate Z and rewrite Eqs. 9


and 11 thus:

N 3
= 2S and = 1 + (1 N ) (12,13)
3S 3S b
Then N may be eliminated between these two equations to get a cubic
equation for the dimensionless shear stress S (for method of solving
cubic equations, see a mathematics handbook):

b+3 b
S3 + S+ =0 or S 3 + 3pS + 2q = 0 (14,15)
18 54

Eqs. 14 and 15 serve to define p and q. Eq. 15 has three solutions, two
imaginary solutions (of no interest here) and one real solution:

S = 2p1 2 sinh ( 1
3
arcsinh p3 2 q ) (16)

or, in terms of the original variables:

61
p, yx b + 3
+1 2 b + 3 3 2 b
= 2 sinh 3 arcsinh
1
H (17)
54

3nKT 54 108

From this result one can plot the non-Newtonian viscosity as a


function of the shear rate.
If one wants only the limiting values of the non-Newtonian
viscosity at zero and infinite shear rates, this information can be
obtained from Eq. 17 or 16. Very small shear rate corresponds to very
small q, so that

S = 2p1 2 sinh ( 1
3 )
p3 2 q = 2p1 2 { 1
3 }
p3 2 q + = 23 p1 q
(18)

if we keep only the terms linear in q in the Taylor series expansion of


the hyperbolic sine and the arc hyperbolic sine functions:

sinh x = x + 16 x 3 + arcsinh x = x 16 x 3 + (19,20)

In terms of the original variables, Eq. 18 is

1
b + 3 b b
p, yx = 3nKT
2
H = nkT (21)
3
54 108 b + 3 H

This corresponds to

b
s = nkT H (22)
b + 3

in the limit of zero velocity gradient.


In the limit of infinite velocity gradient, we make use of the fact
that for large values of the argument, sinh x 12 exp x (see Appendix
C.5). Therefore

S = 2p1 2 sinh ( ln ( 2p q)) = 2p


1
3
3 2 12 1
2
exp 13 ( ln ( 2p q)) = ( 2q)
3 2 13

(23)

62
In the original variables, this becomes

13 13
b b
p, yx = 3nKT 2 H = nKT 54 H (24)
108 108

This corresponds to a "power-law function":

13
b 1
s = nKT H (25)
2 H

in the infinite velocity gradient limit.


Now the results in Eqs. 22 and 25 can be obtained directly from
Eq. 15 in another way. If the velocity gradient (and hence, q) is quite
small, the cubic term in S may be omitted, and one gets then

S = 32 p1 q (26)

directly (see Eq. 18). Similarly, if the velocity gradient is quite large,
then the linear term in S may be omitted, and one gets

S = ( 2q )
13
(27)

immediately (see Eq. 23).


From Eq. 17, one can also find the molecular stretching as a
function of the shear rate, as described at the top of p. 255. In
addition, from Eq. 3 the first normal stress coefficient can be obtained

2 ( s )
2

1 = (28)
nKT

this formula being predicted for the entire shear-rate range.

[Note: For more information on the FENE-P dumbbell model, see


Teaching with FENE Dumbbells, by R. B. Bird, Rheology Bulletin,
January 2007.]

63
Note to p. 275

As an exercise in identifying the dimensions of thermal


quantities, we verify that the following equations are dimensionally
consistent: Eq. 9.3-12, Eq. 9.8-6, and Eq. 9.8-8. We do this by
replacing the symbols in these equation by the corresponding
dimensions, making use of the "Notation" table on pp. 872 et seq.

ML2 ML2
( ) 2 (T )
M t 2T
ML tT
(a) 3 = (1)
tT L2 M

2
M M L L M L2 L M L M
(b) 3 = 3 t t + 3 2 t + 2 t + 3 (2)
t L L t Lt t

L3
L2 L2 L3 M M
M
(c) 2 = 2 (T ) + M 2 + (T ) T 2 (3)
t t T Lt ( ) Lt
In each of these cases, each term has the same dimensions.

64
Note to p. 286

We illustrate the use of Eq. 9.8-8 for an ideal monatomic gas.


For the ideal monatomic gas, the heat capacity at constant pressure is
given by

5 R
C p = (1)
2M

as given two lines after Eq. 9.3-15.


The ideal gas law equation of state is

RT
pV = (2)
M

Therefore, the bracket in Eq. 9.8-8 is

V
V T = T RT
V = T R = 0
V (3)
T pM pM
T p p

Hence, the only term that survives is the heat-capacity term, which is:

H
H
2M
(
o = 5 R T To ) (4)

65
Note to p. 299

In connection with Fig. 10.4-2, we want to find the location of


the maximum in the temperature vs. distance curve. We will express
the result in terms of the Brinkman number, Br. We also need to
know what happens when the Brinkman number goes to zero (i.e.,
negligible viscous heating).
To simplify the discussion, we introduce the following
dimensionless variables:

T T0 x
= (temperature) = (distance) (1)
Tb T0 b

so that Eq. 10.4-9 becomes:

1
= Br (1 ) + (2)
2

To get the location of the maximum of the temperature, we


differentiate with respect to and set the derivative d d equal to
zero:

d 1
= Br (1 2 ) + 1 = 0 (3)
d 2

From this we get the location of the maximum in the temperature


curve:

1 1
max = + (4)
2 Br

When there is negligible viscous heating, Br 0 , and, according to


Eq. 4, max . This result is nonsense, since the system extends only
to = 1 . Therefore, Eq. 4 has to be restricted to max 1 , and when
max = 1 , Br = 2. Alternatvely, Eq. 4 has to be limited to 2 Br< .

66
Note to p. 309

It's always a good idea to check the solutions to problems. Here


we verify that the expression in Eq. 10.7-13 for the dimensionless
temperature in the cooling fin satisfies the differential equation in Eq.
10.7-9, and the boundary conditions in Eqs. 10.7-10 and 11.
First calculate the derivatives from Eq. 10.7-13 and Eqs. C.5-10
and 11:

d sinh N (1 )
d
=
cosh N
( N ) (1)

cosh N (1 )
d 2
d 2
=
cosh N
( +N ) 2
(2)

When the expression for (in Eq. 10.7-13) and its second derivative
(in Eq. 2) are substituted into Eq. 10.7-9, an identity results. Therefore,
Eq. 10.7-13 satisfies the differential equation.
The boundary condition at = 0 is satisfied, since

cosh N (1 )
= =1 (3)
cosh N
=0

and the boundary condition at = 1 is also satisfied, since

d sinh N (1 )
d =1
=
cosh N
( N ) =0 (4)
=1

The hyperbolic sine is shown in Figure C.5-2.

67
Note to p. 315

Here we verify the determination of the constants of integra-


tion, C0 , C1 , and C2 in Eq. 10.8-27.

C1 : Boundary condition 1 requires that ( 0, ) = finite :

( 0, ) = C0 + C1 ln 0 + C2 (1)

This can be satisfied only if C1 = 0 .

C0 : Boundary condition 2 requires that = 1 at = 1 :

2 4 3 1
= C0 = C (2)
=1 4 16 4 0
=1

from which it follows that C0 = 4 .

C2 : The dimensionless form of condition 4 is given in Eq. 10.8-25.


Substitution of Eq. 10.8-27 into that equation gives:

1 2 4
4 16
(
= 0 4 + 4 + C2 1 2 d ) (3)

Next we evaluate the integrals:

1
0 4 1( 2
d)
= 4
1
0 ( 3
d)
= 4
1 1
2 4 = (4)

2 4 1 5 5 7
( )
3
1
0 4 4 16 1 d = 4 0 4 16 + 16 d =
2


1 5 1 7
4 + = (5)
16 6 16 8 16 24 4

68
1
( ) 1
C2 0 1 2 d = C2
4
(6)

Combining the results of Eq. 3 to 6, we find that C2 = 7 24 .

69
Note to p. 337

In the textbook, we transformed Eq. 11.2-2 into an equation for


the enthalpy (Eq. 11.2-3) and then used an equilibrium thermo-
dynamic formula to get the equation of energy in terms of C p and T,
given in Eq. 11.2-5 (and also Eq. (J) in Table 11.4-1). Specifically, we
used the equation for H (T, p ) :

H
H
V
dT + p dp = Cp dT + V T T dp

dH = (1)
T p T p

appropriately rewritten for a "particle" of fluid moving with the local


velocity v.
Alternatively, we can begin with Eq. 11.2-2 and use another
equilibrium thermodynamic formula to get the equation of energy in
terms of C V and T (see Eq. (I) in Table 11.4-1). Specifically, we use
( )
T ,V :
the equation for U


= U dT + U dV = C dT + p + T p dV
dU (2)
T V T
V V T V

Then we can use this equation to obtain Eq. (I) in Table 11.4-1.
First we rewrite Eq. 2 for a fluid particle moving with the fluid,
and then we multiply by the density of the fluid. This gives:

DU DT + p + T p DV
= C (3)
Dt V
Dt T Dt
V

Since V = 1 , we may rewrite DV Dt as follows:

DV D 1 1 D
= = 2 Dt (4)
Dt Dt

70
Then if we use Eq. (A) of Table 3.5-1, we may rewrite this last relation
as:

DV 1

Dt
( )
= 2 ( v ) = ( v ) (5)

Now substitute this expression into Eq. 3, and then make use of Eq.
11.2-2 to get:

DT p
C V + p + T ( v ) = ( q ) p ( v ) ( :v ) (6)
Dt T V

When the terms involving ( v ) are moved to the right side, there is
some cancellation and Eq. (I) of Table 11.4-1 is obtained:

DT p
C V = ( q ) T ( v ) ( :v ) (7)
Dt T V

[Note: See footnote 1 at the bottom of p.338 for a discussion of


"incompressible fluids."]

71
Note to p. 341

Verify that Eq. (T) of Table 11.4-1 is the equation of change for
entropy.
We can start with the thermodynamic expression

= TdS pdV
dU (1)

which is written for a small isolated mass of fluid that is stationary.


For a small mass of fluid that is moving with the fluid, we can then
write:


DU DS DV
=T p (2)
Dt Dt Dt

where we are now assuming that this expression can be applied


locally. When this equation is multiplied by and then combined
with Eq. (G) of the table we get:

DS DV
( q ) p ( v ) ( :v ) = T p (3)
Dt Dt

Next, divide by T and replace V by 1 / to get

DS 1 1 1 p D 1
= ( q ) p ( v ) ( :v ) + (4)
Dt T T T T Dt

The second and fourth terms on the right side may be seen to cancel
one another is one makes use of the equation of continuity in the
form of Eq. (A) of Table 3.5-1, so that we get

DS 1 1
= ( q ) ( :v ) (5)
Dt T T

When we make use of Eq. 3.5-4, Eq. 5 becomes

72

t
(
S + Sv )
= 1 ( q ) 1 ( :v )
T T
(6)

This is equivalent to Eq. (T) in Table 11.4-1. Eq. (T) is written in a


form that emphasizes the entropy flux (q / T ) and the entropy
production (the terms in brackets:


t
( )
q + 1 (q T ) 1 ( :v )
S = Sv T 2
T T

(7)

See Problem 11D.1 and 24.1 for more on this subject.

73
Note to p. 343

Verify that Eq. 11.4-14 gives the location of the maximum


temperature. We'd also like to know whether the maximum is nearer
the inner cylinder or the outer.
Differentiate in Eq. 11.4-13 with respect to and set the
derivative equal to zero:

d 1 2 1 1
= + N 3 1 2 =0 (1)
d ln ln

Solving for 2 2 gives:

2 1 1 1
= + 1 2 (2)
2
N ln ln

whence, the location of the maximum temperature rise is

2 ln (1 )
max = (3)
( )
1 2 1 (1 N )

Now is max greater than or less than 12 ( + 1) ? Clearly this cannot be


answered until N is known; but N depends on the geometry, the
temperature difference, the thermal conductivity, and the viscosity. If
we take N to be infinity, then we get:

1
2
1
3
1
4

1
2 ( + 1) 0.75 0.67 0.625

max 0.67 0.51 0.42

This suggests that the maximum occurs nearer the inner wall, where
the velocity gradient is larger.

74
Note to p. 346

Here we want to derive Eq. 11. 4-27 from Eq. 11.4-26 by follow-
ing the instructions in the text. First we rewrite Eq. 11.4-26 in terms of
dimensionless variables:

T T1
= = dimensionless temperature (1)
T T1
r
= = dimensionless radial coordinate (2)
R

Then Eq. 11.4-26 becomes:

d R d 2 d
wr C p
= where R0 = (3a,b)
d R0 d d 4 k

Now introduce the change of variable u = 2 ( d d ) , and rewrite the


differential equation as:

u R du
= (4)
2 R0 d

This equation may be solved to get:

1 R R R 1
= ln u + ln C1 or C1 u = exp 0 (5a,b)
R0 R0 R

where the constant of integration has been written as ( R R0 ) ln C1 .


Reverting to the original variable gives:

d R 1
C1 2 = exp 0 (6)
d R

Further integration then gives:

75
1 2 R0 1 v R
C1 = exp R d + C2 = 1 exp 0
R
v dv + C2

e ( 0 ) e ( 0 )
R R v R R
= + C2
( R0 R )
e ( 0 ) e ( 0 )
R R R R
= + C2 (7)
( R0 R )

The constants of integration may be determined from the boundary


conditions that: at = , = 1 , and at = 1 , = 0 . From the second
of these boundary conditions it is evident that C2 = 0 . From the first
boundary condition, C1 may be found. Then the final expression for
the dimensionless temperature is:

e ( 0
R )
e ( 0 )
R R R
= R (8)
e ( 0
R )
e ( 0 )
R R

which agrees with Eq. 11.4-27.

76
Note to p. 375

a. First we'll verify the formula for differentiation of the error


function in Eq. C.6-2.
b. Then we'll show that Eq. 12.1-8 is a solution to Eq. 12.1-3 and
the associated boundary and initial conditions.

a. The differentiation of the error function is given by Eq. C.6-2:

d 2 u 2 du
erf u = e (1)
dx dx

where it is understood that u is a function of x. Now differentiate the


error function in Eq. C.6-1 with respect to x using the Leibniz formula
of C.3:

d d 2 u u 2 2 u u 2 u 2 u 0 2 0
erf u =
dx
0 e du =

0 x
e du + e
x
e
x (2)
dx

Since u is a dummy variable of integration, it is not a function of x,


and therefore the first term is zero. The last term is also zero. Hence
the second term is the only term in the parenthesis that contributes to
the derivative of erf u, and that leads directly to Eq. 1. [Note: It is
important to designate the dummy variable of integration, u , and the
upper limit in the integral, u, by two different symbols. This example
emphasizes the importance of this statement.]

b. Now turn to Eq. 12.1-8, where the left side is the dimension-
less temperature difference . Form the derivatives that appear in
Eq. 12.1-3:

2 y2 4 t y 1 3 2
= e t (3)
t 4 2
2 y 2 4 t 1
= e (4)
y 4 t
2 2 y 2 4 t 1 2y
= e (5)
y 2 4 t 4 t

77
When the results in Eqs. 3 and 5 are inserted into Eq. 12.1-3 it is found
that the equation yields an identity. Therefore, Eq. 12.1-8 does satisfy
the differential equation.
Eq. 12.1-8 also satisfies the boundary conditions at y = 0 and
y = , as may be seen from Fig. 4.1-2. At t = 0 , Fig. 4.1-2 tells us that
= 0.

[Note: Here, and elsewhere we have made use of the Leibniz formula
for differentiating an integral. For additional information and anec-
dotes regarding the Leibniz formula see R. P. Feynman, Surely You're
Joking Mr. Feynman, Bantam Books, New York (1986), p. 72 and p. 93.
Professor Feynman was a Nobel Prize winner in physics.]

78
Note to p. 377

As pointed out in the footnote, there are two solutions to this


problem, one that converges rapidly for long times (see Eq. 12.1-31):

( 1)n exp n +( ) (
2 cos n + 12 )
2
= 2 1
(1)
n=0 (n + ) 1
2
2

and one that converges rapidly for small times:


= 1 ( 1) erfc
n + 12 12
n (
+ erfc
)
n + 12 + 12

( ) (2)
n=0

We want to verify that both of these solutions satisfy the partial


differential equation (Eq. 12.1-14), the initial condition (Eq. 12.1-15),
and the boundary conditions (Eq. 12.1-16).

a. Solution in Eq. 1:
When Eq. 1 is substituted into Eq. 12.1-14, we get by
differentiating the exponential once with respect to and the cosine
twice with respect to

( 1)n
(
exp n + 12 ) 2 cos n + ( ) (n + ) 2 =
2 2
2 1 1

n=0 (n + ) 1
2
2 2

( 1)n
(
exp n + 12 ) 2 cos n + ( ) (n + ) (n + )
2
2 1 1 1

n=0 (n + ) 1
2
2 2 2

(3)
and this is clearly an identity, so that the partial differential equation
is satisfied.
When is set equal to zero and = 1 in Eq. 1, we get

( 1)n
1 = 2
n=0 (n + )
1
(
cos n + 12 ) (4)
2

79
We have to prove that this is an identity. To do this, multiply both
( )
sides by cos m + 12 and integrate over from 1 to +1:

( 1)n
+1
1 cos m + ( 1
2 ) d = 2 (n + )
n=0
1
+1
1 ( ) (
cos m + 12 cos n + 12 d )
2
(5)
Then performing the integrations we get
( 1) +1 cos2 n + n
1
( ) ( ) d
+1
sin m + 1
= 2 mn 1
1

( m + 12 ) 2 1
n=0 n + 2
1
( ) 2

or
( 1) m

(
2
m + 12
sin
)
m + 1
2
= 2 (
m + 12
) ( )
(7)

This is an identity, since sin m + 12 = ( 1) . ( ) m

When is set equal to +1 or 1, we get = 0 , since


cos m + ( 1
2 ) = 0 .
b. Solution in Eq. 2
When Eq. 1 is substituted into Eq. 12.1-14, we get by
differentiating the complementary error functions once with respect
to and then twice with respect to

(
n + 12 12 ) n + 1 1 2
( )
erfc =
2
exp

2 2
n + 1 1 1 3 2
2 ( ) ( )

2 2


(8)
( ) n + 1 1 1
( )
2
n + 12 12
erfc =
2
exp
2 2
2 (9)


2 (
n + 12 12 2 ) n + 1 12
exp
( )

2 2
erfc =
2

80
( )
n + 1 1 1 1
2
2 2

2 2
(10)

Thus, the nth term satisfies the partial differential equation, and
hence the entire series does.
At = 0 , all of the complementary error functions are zero, so
that = 1 , and the initial condition is satisfied.
At = 1 , = 0 as may be seen as follows. First consider
= +1 :

0 1 1 2 2 3
= 1 erfc + erfc
erfc + erfc +
erfc + erfc

(11)
Hence there is cancellation of between the nth and (n + 1)th terms,
and, since erfc 0 = 1, the dimensionless temperature is zero. A similar
argument can be made for = 1 .

81
Note to p. 379

In Example 12.1-3 going from Eqs 12.1-38 and 39 to Eq. 12.1-40


presents a few problems. Therefore we go through the details.
Substitution of Eq. 12.1-38 into Eq. 12.1-39 gives:


k (T0 T ) = q0 y e
2 y
cos t y dy (1)
2

Bars have been added to the variable of integration to distinguish it


from the lower limit on the integral. It is easier to perform the
integration if the cosine is converted into an exponential of a complex
quantity, thus:

T T0 =
q0
k
2 y
y e {
e
i ti 2 y
}dy
=
q0
k { 2 y i 2 y
e i t y e dy }

q0 i t e (1+i ) 2 y
= e
k (1 + i ) 2
y

q0 i t e (1+i ) 2 y
= e (2)
k
(1 + i ) 2

Next we remove from the braces that portion of the exponential that
is real; we also multiply numerator and denominator by (1 i ) . Then
we have

q e i t e i 2 y (1 i )
2 y
T T0 = 0 e
k 2 2
q 2 2 y
= 0
2k
e {
e i t e
i 2 y
}
(1 i ) (3)

82
In order to proceed, we need to rewrite (1 i ) in the form re i . Then
we find r and as follows:

1 i = re i = r ( cos + i sin ) (4)

so that equating real and imaginary parts gives

1 = r cos and 1 = r sin (5,6)

Taking the ratio of these two equations we get

r sin 1
= or tan = 1 or = 34 , 14 (7,8,9)
r cos 1

Since (1 i ) is in the 4th quadrant, the appropriate choice is = 14 .


Next we square both Eq. 5 and Eq. 6

1 = r 2 cos2 and 1 = r 2 sin 2 (10)

Adding the two equations then gives:

r2 = 2 or r= 2 (11)

The plus sign must be chosen, since r must be non-negative.


Therefore we have shown that

1 i = 2e i 4
(12)

Returning now to Eq. 3, we get:

T T0 =
q0
2k
2 2 y

e {
2e i t
i 2 yi 4
}
q 2 y
= 0
k
e (
cos t 2 y 14 ) (13)

This agrees with Eq. 12.1-40 in the textbook.

83
Note to p. 386

The derivation of Eq. 12.3-6 from Eq. 12.3-5 is given here. First
we note that, if z = x + iy = re i , then

ln z = ln r + i = ln x 2 + y 2 + i arctan ( y x ) (1)

Now we have to resolve Eq. 12.3-5 into its real and imaginary parts.
We introduce the abbreviated notation Z = z b , X = x b , and
Y = y b . Then

1 sin Z 1 1 sin ( X + iY ) 1
w= ln = ln + i
sin Z + 1 sin ( X + iY ) + 1

1 sin X cos iY + cos X sin iY 1


= ln
sin X cos iY + cos X sin iY + 1

1 sin X cosh Y + icos X sinh Y 1


= ln
sin X cosh Y + icos X sinh Y + 1

( sin X cosh Y 1) ( sin X cosh Y + 1) + cos 2 X sinh 2 Y



1 +i cos X sinh Y (sin X cosh Y + 1) cos X sinh Y ( sin X cosh Y 1)
= ln
( sin X cosh Y + 1 ) 2
+ ( cos X sinh Y ) 2

= ln
( 2 2
)
1 sin X cosh Y 1 + cos X sinh Y + 2i cos X sinh Y
2 2

(2)
(sin X cosh Y + 1)2 + (cos X sinh Y )2

In going from the third to the fourth line, we have multiplied the
numerator and denominator by the complex conjugate of the
denominator.
The imaginary part of the expression in Eq. 2 is then

84
1 2cos X sinh Y
= arctan
sin X cosh Y 1 + cos X sinh Y
2 2 2 2

1 2cos X sinh Y 1 2 ( cos X sinh Y )


= arctan = arctan
sinh Y cos X
2 2
1 ( cos X sinh Y )2

1 2 tan A
arctan (3)
1 tan A
2

The last expression serves to define A . But the quantity in


parentheses is just tan 2A (see, e.g., formula 406.02 of Dwight's Tables
of Inegrals and Other Mathematical Data, 4th edition), and the "angle
whose tangent is tan 2A " is just 2A (i.e., arctan(tan2A ) = 2A.
However, A = arctan ( cos X sinh Y ) so that, finally

2 cos X 2 cos x b
= arctan = arctan (4)
sinh Y sinh y b

which is the result in Eq. 12.3-6.

85
Note to p. 388

Here we work through the missing steps to get the result in Eq.
12.4-16. We begin by evaluating the integral in the first term on the
right side of the equals sign in Eq. 12.-4-4 (the second term is zero,
because ve = v = a constant in this problem):

vx vx
v x ( v vx ) dy = v2 ( x ) 0
1
0 v 1
v
d (1)

Here ( x ) is the velocity boundary-layer thickness, and = y ( x ) is


the dimensionless coordinate in the y-direction. In the second integral
we have changed the upper limit to "1" because 1 ( vx v ) in Eq.
12.4-6 and 7 is zero beyond = 1 . Then substituting the assumed
velocity profile into Eq. 1 gives:

v x ( v vx ) dy

0
( )(
= v2 ( x ) 0 2 2 3 + 4 1 2 + 2 3 4 d
1
)
= v ( x ) ( 2 4 )
1
2
0
2
2 3 + 9 4 4 5 4 6 + 47 8 d
(
= v2 ( x ) 1 43 12 + 95 32 74 + 12 19 )
= v ( x )(
2

315+56718035
315 )= 37
315
v2 ( x ) (2)

Similarly, the integral appearing in Eq. 12.4-5 may be


evaluated:

C p vx (T T ) dy

0

v (T T ) ( x ) 1 v x T T d
= C p 0 T 0 v T T T
0

T T0 T0 T
v (T T ) ( x ) 1 v x
= C p 0 T 0 v T T T T dT
0 0

86

v (T T ) ( x ) 1 v x 1 T0 T d
= C (3)
p 0 T 0 v T0 T T

in which T = y T ( x ) = y ( x ) , and is assumed to be inde-


pendent of x. Then, inserting the postulated profiles for velocity and
temperature into Eq. 3, we get:

C p vx (T T ) dy

0
0 T 0 T T T (
v (T T ) ( x ) 1 2 2 3 3 + 4 4
= C p ) (1 2T )
+ 2T3 T4 dT
v (T T ) ( x )
= C p 0 T ( 2
15
140
3
3 + 180
1
4 ) (4)

When the expressions in Eqs. 2 and 4 as well as Eqs. 12.4-6 to 9 are


substituted into Eqs. 12.4 and 5, we get differential equations for the
boundary-layer thicknesses as a function of the distance along the
plate:

2 v 37 d
= v2 (5)
315 dx

2k (T T0 )
T
= ( 2
15
140
3
3 + 180
1
4 C p )
v (T T )
0
d

dx T
(6)

Eq. 5 for ( x ) may be solved as follows:

d 315 2 v 630 x 1260 x

dx
=
37 v2
; 0 d = 37 v
0 dx ; =
37 v
(7,8,9)
and Eq. 6 for T ( x ) may also be solved:

2k (T T0 )
T
= ( 2
15
140
3
3 + 180
1
4 C p )
v (T T )
0
d

dx T
(10)

d 1 2k
T T = (11)
dx ( 2
15
140
3
3 + 180
1
4 ) C vp

87
T 2 x
0 T d T = 0 dx (12)
( 2
15
140
3
3 + 180
1
4 ) v

4 x
T = (13)
( 2
15
140
3
3 + 180
1
4 ) v

It remains to find = T ( x ) ( x ) as a function of the physical


properties. Forming the ratio we get:

T 4 ( 37 1260 )
= = (14)
( 2
15
140
3
3 + 180
1
4 )

Squaring both sides and collecting all the terms on the left side, we
find:
35 1 k 1
2
140 + 180 =
3 3 5 1 6
Pr= = (15)
15
315 Pr C p C p k

For large Prandtl numbers, we can drop all but the lead term on the
left side and get

= 3 35 15 Pr 1 3 = 3 0.880Pr 1 3 = 0.958Pr 1 3 (16)


315 2

As pointed out in the textbook, 0.958 may be replaced by 1 to fit the


exact curve in Eq. 12b.2-15 within 5%. This replacement leads to Eq.
12.4-16 in the textbook.

[Note: In earlier printings of the textbook, the second integrals in Eqs.


12.4-10 and 11 had an upper limit of infinity rather than 1.]

88
Note to p. 413

The object here is to fill in the missing steps between Eq. 13.4-11
and Eq. 13.4-16. First we multiply Eq. 13.4-11 by

d ( ) d
t

1 + = C0 (1)
d d

Then first integration with respect to gives:

( ) d
t
= C0 0 d + C1 C0 I ( ) + C1

1 + (2)
d

where ( ) is the dimensionless velocity defined just after Eq. 13.4-6,


and the abbreviation I ( ) is introduced. Eq. 2 may now be rewritten:

d C0 I ( ) C1
= + (3)
( )
d 1 + (t ) 1 + (t )
( )
A second integration with respect to gives:

= C0 0
()
I
d + C1 0
1
d + C2 (4)
(
1 + ( )
t
) (
1 + ( )
t
)
in which we must remember that ( ) is now a function of . If next
t

we consider the limit of the above expression as 0 , it may be seen


that the first term goes to zero and the second term goes to infinity
(and therefore violates B. C. 1); therefore C1 must be taken to be zero.
Hence the expression for the dimensionless temperature becomes:

89
( , ) = C0 + C0 0
()
I
d + C2 (5)

t
(
1 + ( )
)
Next, apply the boundary condition at = 1 :

1 = C0
()
I
= C0 I (1) or C0 = I (1)
1
(6)

t
(
1 + ( )
=1 )
inasmuch as ( ) is zero at the wall. Next we want to get the driving
t

force 0 b , which is the dimensionless wall temperature minus


the dimensionless bulk temperature (defined in Eq. 10.8-33):

1 ()
I
0 0 d d

0 b = C0 0
1 I ( )
d C0
(
1 + ( )
t
)
1 + ( )

t
( ) 1
0 d

I ( ) ()
( ) 1 + ( ( ) ) d
1 C0 1 1 I
= C0 0 d d (7)
(
1 + ( )
t
) I (1) 0

t

In the second expression we have interchanged the order of integra-


tion. The second term in Eq. 7 may be rewritten:

()

C0 1

I ( 1) 0 ( 1
0 d 0 d

) ( )
1 +
I
(t )
d

I ()
=
C0 1
I (1) I
() d
I (1 ) 0 1 + ( )

t
( )

90
I ( ) I ( )
2
1 C 1
= C0 0 d + 0 0 d (8)

(
(
1+

t)
) I (1 )
1+

(
(
t)
)
The first term in Eq. 8 above just cancels the first term in Eq. 7, and
hence we are left with:

I ( ) I ( ) I (1)
2 2
C 1 1
0 b = 0 0 d = 0 d (9)
(
I (1) 1 + ( t )
) 1+
( ( t)
)

This is in agreement with Eq. 13.4-16 in the textbook.

91
Note to p. 415

Some additional material is given here on 13.5.


Perform the indicated substitutions into Eq. 13.5-1 to get for the
partial differential equation for :

( ) F (t ) F (t ) 1
t
F + = r (1)
z r z z Pr(t ) r r r

where the primes indicate differentiations with respect to . We next

(
make the change of variables = r z and = ( ) w z , and further

t
)
let ( r, z ) = ( , ) = f ( ) , so that

1 f 1
= + = + 0 = (2)
r r r z z

r ( ) f r
t
f
= + = 2 + = 2 2
z z z z w z z
(3)
Substitution of these expressions into Eq. 1 gives

( ) F f 1 (t ) F f r f
t
F
z z z z 2 2 z
( ) 1 f 1
t
= (t ) 2 (4)
Pr z

Multiplication of the entire equation by z 2 ( ) gives


t

F f 1 1 f
F [ f ] + ( F ) f + =
Pr(t )
(5)

92
On the left side, the terms involving F f cancel, and the equation
may be rewritten as follows:

1 d 1 1 f
d
( Ff ) = (t )
(6)
Pr

This equation can be multiplied by Pr( ) and integrated once to give


t

f
Pr( ) Ff =
t
+C (7)

According to Eq. 5.6-20, F = 0 at = 0 , which means that C = 0. A


further integration from 0 to then yields

f ( ) C32
(C )
2
F
= Pr ( ) d = Pr( ) 0 d = Pr( ) ln 1 +
t 2
0
t t 1
ln
f (0)
(C )
2 4 3
1+ 1
4 3
(8)
or, taking the antilogarithm of both sides

f ( ) 2Pr( )
t

= 1 + (C )
2
1
(9)
f ( 0 ) 4 3

When this result is compared with Eq. 5.6-21 for the velocity profile
in a circular jet, and use is made of the definition in Eq. 13.5-8, we
obtain finally

Pr( )
t

vz
= (10)
max vz ,max

which is a rather simple, and apparently fairly satisfactory, result.

93
Note to p. 454

Verify that Eq. 15.1-1 can be obtained from Eq. 11.1-9 by the method
described on p. 454.
We start by integrating Eq. 11.1-9 over the volume of the flow
system shown in Fig. 7.0-1:

V (t) t ( 2 v
1 2
+ U )
dV =
+
V (t)
( ( 1
2
v 2 + U +
v dV))
(
V (t ) ( q )dV V ( t ) ( pv )dV V (t ) [ v ] dV ) (1)

We next apply the Leibnitz formula to the left side of the equation
and the Gauss divergence theorem to the right side:

d

dt V (t )
( 1
2
v 2 + U + )
dV n
S(t ) ( ( 1
2
+
v 2 + U v dS
S ) )
( (
= S(t ) n 1
2
v 2 + U ))
v dS ( n q )dS
+
S(t )

( )
S( t ) ( n pv )dS S(t ) n [ v ] dS (2)

The integral in the first term on the left side is the total energy
(kinetic + internal + potential energy). The second term on the left
side can be combined with the first term on the right side. Thus we
get

d
dt
( ( (
Ktot + Utot + tot ) = S t n 12 v 2 + U
( )
+ ( v v ) dS
S ) )
S t ( n q )dS S t ( n pv )dS S t n [ v ] dS
() () () ( ) (3)

We now analyze the terms on the right side seriatim:


The first term can be seen to contribute nothing on the fixed
surface S f and the moving surface Sm . At the inlet cross section S1
and the outlet section S2 , the surface velocity v S is zero and collinear
with the outwardly directed unit vector n. The fluid velocity vector v
is assumed to point in the direction opposite to the n vector at the
entry plane, and in the same direction as as the n vector at the exit

94
plane; therefore, at the entry ( n v ) = v , and at the exit ( n v ) = +v .
We make the further assumption that the internal energy and the
potential energy are constant over the cross section. Then when the
integration over the cross sectional area is performed we get:

S(t ) n ( ( 1
2
+
v 2 + U ) )
( v v S ) dS

v S +
= 12 1 v13 S1 + 1U
1 1 1 1 1 v1 S1

v S +
12 2 v23 S2 + 2U
2 2 2 2 2 v2 S2 (4)

The second term on the right side (the q-term) is the integral
over all surfaces of the normal component of the heat flux vector and
is thus the rate of total heat addition to the system, Q:

S
f +Sm +S1 +S2
(n q)dS = Q (5)

It is assumed that the heat addition at surfaces S1 and S2 is usually be


small compared to the heat added at the solid surfaces.
The third term on the right (the p-term) has to be evaluated at
all the surfaces. At the inlet and outlet planes, we will get

S
1 +S2
(n pv )dS = p1 v1 S1 p2 v2 S2 (6)

by the same arguments leading to the internal energy terms in Eq. 4.


These terms represent the rate of doing work on the system at the
entry and exit planes. On the solid surfaces we get

S
f +Sm
(n pv )dS = Wmp (7)

This term is the rate that pressure does work on the system at the
moving surfaces Sm ; there is no work done at the fixed surfaces S f ,
inasmuch as the rate of doing work is a force times a velocity, and at
S f the surface velocity is zero.
The fourth term on the right (the -term) is evaluated similarly
to the p-term. First the contributions at surfaces S1 and S2 are
considered, but it is assumed that these will be small compared to the

95
pressure terms in Eq. 6. On the solid surfaces there will be a
contribution similar to that for the pressure forces on the moving
surfaces:

S
f +Sm (n [ v ])dS = W
m (8)

and, here again, the contribution at S f will be zero. The contributions


from Eqs. 7 and 8 will be added to give Wmp + Wm = Wm , the total
work done on the system through the moving surfaces.
When all the contributions in Eqs. 4 through 8 are added up we
get Eq. 15.1-1 of the textbook:

d
dt
( (
Ktot + Utot + tot ) = 12 1 v13 + 1U
v +
1 1
)
1 1 v1 S1

(
12 2 v 23 + 2U v +
2 2
)
2 2 v2 S2 + Q + Wm

(
+ p1 v1 S1 p2 v2 S2 ) (8)

or, introducing the enthalpy

d
dt
( (
Ktot + Utot + tot ) = 12 1 v13 + 1 H
v +
1 1
)
1 1 v1 S1

(
12 2 v 23 + 2 H v +
2 2
)
2 2 v2 S2 + Q + Wm (9)

Either Eq. 8 or Eq. 9 is referred to as the unsteady-state macroscopic


energy balance.

96
Note to p. 494

The derivations of the Stefan-Boltzmann law and Wien's law


from the Planck black-body distribution law are quite important and
therefore it is a good idea to understand all the intermediate steps in
the development.

a. The Stefan-Boltzmann law


We integrate Planck's distribution law over all wavelengths:

(e) 2 c 2 h 1
q(b ) =

0 0
e
qb d = d (1)
5 e ch KT 1

Next we make a change of variable x = ch KT . Then

5
1 KT ch 1
2 c h 5 d = 2 c h x 5 2 dx
2 2

ch KT x
2 ( KT )
4

= 2 3
c h
( x ) dx
3
(2)

Hence the integral becomes

2 ( KT )
4
x3
q(b ) =

0
e
dx (3)
2 3
c h ex 1

Then expand the denominator of the integrand as a Taylor series


about x = 0, to get:

(
e x 1 = 1 + e x + e 2x + e 3x + 1 ) (4)

Then a term by term integration of Eq. 3 gives:

2 ( KT ) 2 ( KT ) 1 2 ( KT ) 4
4 4 4


0 3 nx
x e dx = 6 4 = (5)
c 2 h3 n=1 c 2 h3 n=1 n c 2 h3 15

97
On p. 171 of Planck's book, The Theory of Radiation, Dover, New York
(1959), which is a translation of Vorlesungen ber die Theorie der
Wrmestrahlung, 5th edition, Barth, Leipzig (1923), the Stefan-
Boltzmann equation is obtained as shown above. However, Planck
did not evaluate the summation in Eq. 5 exactly. Instead, he simply
evaluated the sum numerically as:

1 1 1
=1+ + + + = 1.0823 (5)
24 34 44

Nowadays, even in a small integral table (such as H. B. Dwight,


Tables of Integrals and Other Mathematical Data, Macmillan, New York,
Fourth Edition (1961)), the integral over x in Eq. 3 may be found; see
Formula 860.33 on p. 231.
When Eq. 5 is compared with the Stefan-Boltzmann law for a
black body q( ) = T 4 , then we get the Stefan-Boltzmann constant:
e

2 5K4
= (6)
15 c 2 h3

which interrelates key constants from several different fields of


physics.

b. Wien's displacement law


First rewrite Eq. 16.3-7 in terms of x thus:

5
( e) 2 c 2 h 1 KT x 5
qb = = 2 c h x
2
(7)
5
e ch KT 1 ch e 1

We can then differentiate this with respect to x to get

5
dq(b )
e
2 KT 5x
4
x5 e x
= 2 c h (8)
ch e x 1 e x 1
( )
2
dx

Then setting the derivative equal to zero gives the value of x at which
the maximum in the q(b ) curve occurs:
e

98
xmax e xmax
5 =0 (9)
e xmax 1

whence

(
xmax = 5 1 e xmax ) (10)

from which one can find, by trial and error, xmax = 4.9651... , or
maxT = 0.2884 cm K . This is Wien's displacement law.

99
Note to p. 529

We want to verify that Eq. 17.4-4, Eq. (I) of Table 17.8-1, and Eq.
(E) of Table 17.8-2 are dimensionally consistent.
If we put the dimensions of the quantities into the equation
instead of the mathematical symbols (and omit the numerical factors)
we get:

L2 M
t Lt
1
(a) = (1)
ML2 L
t 2 T (T )

moles moles L
(b) 2 = (2)
Lt L3 t

M L M L2 1
(c) 3 t = 3 t L (3)
L L

In each case, the dimensions on the left side are the same as the
dimensions on the right side.

100
Note to p. 534

It is important to know how to simplify multicomponent


relations to their corresponding binary equations. We illustrate this
procedure by showing how to get the binary formula in Eq. (Q') from
the multicomponent formula in Eq. (Q) in Table 17.7-1.
In the sum in Eq. (Q), the indices and can take on only the
values of A and B in a binary system. Then in the sum, must be B.
Therefore, Eq. (Q') becomes for a binary system:

MA
A = M + xA ( MB MA ) xB (1)
M2
Next use Eqs. (M) and (J) of the table, written for a binary system:

MA
A = ( x A MA + xB MB ) + x A ( MB MA ) ( x A )
( xA MA + xB MB ) 2

(2)
Within the bracket, the terms xA MA cancel each other, and the
remaining terms may be combined, since xA + xB = 1 . We are then left
with:

MA MB
A = + xA (3)
( xA MA + xB MB ) 2

which is just Eq. (Q').

101
Note to p. 535

Here we want to verify that Eq. 17.7-4 can be derived from Eq.
17.7-3 using the relations in Tables 17.7-1 and 2.
First we transform the last term in Eq. 17.7-3 into the analogous
term in Eq. 17.7-4, with a multiplying factor:

(
D AB A = ( cM ) D AB MA MB M 2 xA )
= cD AB xA ( MA MB M ) (1)

Here, Eq. (F) of Table 17.7-1 was used, as well as Eq. ( Q ).


Next we transform the mass concentration times the diffusion
velocity as follows:

A ( v A v ) = ( c A MA ) v A ( A v A + B v B )
= ( c A MA ) B ( v A v B ) = (c A MA ) ( xB MB M ) ( v A v B )
(
= ( c A MA ) ( MB M ) v A ( xA v A + xB v B ) )
= cA ( v A v * ) ( MA MB M )
(2)

In this development, we end up with the same factor ( MA MB M )


appearing. In the first step above, we used Eq. (B) of Table 17.7-2. In
the second step we used Eq. (K) of Table 17.7-1. In the third step Eq.
(O) of Table 17.7-1 was used. In the fourth step we used Eq. (J) of
Table 17.7-1, and in the last step Eq. (C) of Table 17.7-2.
The rest of the proof makes use of the definitions

jA = A ( v A v ) JA = c A ( v A v * ) (3a,b)

which are given in Eqs. (E) and (I) of Table 17.8-1. It remains, then, to
verify that

jA = JA ( MA MB M ) (4)

This may be done by rewriting Eq. 4 as

102
jA = JA ( M ) ( MA M ) ( MB M ) = JA ( c ) ( A x A ) ( B xB ) (5)

Here Eqs. (G) and (O) of Table 17.7-1 have been used. The result in
Eq. 5 may also be written thus:

jA JA
= (6)
A B cx A xB

This important equation is also given in Eq. 17B.3-1.

103
Note to p. 547 (i)

Starting with the concentration profile for species A in Eq. 18.2-


11, derive the subsequent results up through Eq. 18.2-16.
First obtain the expression for xB,avg :

0 ( xB2 xB1 ) d
1
( xB2 xB1 )
1
xB,avg
= = (1)
xB1 1
ln ( xB2 xB1 )
0 d 0

where we have used the integral a x dx = a x ln a + C . Therefore( )


xB,avg
=
( xB2 xB1 ) 1 or xB,avg =
xB2 xB1
( xB )ln (2a,b)
xB1 ln ( xB2 xB1 ) ln ( xB2 xB1 )

Hence the rate of evaporation of A at the gas-liquid interface is:

cD AB dxB cD AB dxB d cD AB d ( xB xB1 )


NA = = = (3)
z= z1
xB1 dz z= z1
xB1 d =0
dz z2 z1 d
=0

Then, using the derivative da x dx = a x ln a , we get:

cD AB x x cD AB xB2
NA = B2 ln B2 = ln (4)
z= z1
z2 z1 xB1 xB1 z2 z1 xB1
=0

Then, multiplying the numerator and denominator by ( xB2 xB1 ) :

cD AB ( xB2 xB1 ) cD AB ( xA1 xA2 )


NA = = (5)
z= z1
z2 z1 ( xB )ln z2 z1 (1 xA )ln

Next we obtain the solution for very small values of xA1 and xA2 :

104
cD AB ( xA1 x A2 ) ln (1 xA2 ) (1 xA1 )
NA =
z= z1
z2 z1 (1 xA2 ) (1 xA1 )
cD AB
= ln (1 xA2 ) ln (1 x A1 )
z2 z1

=
cD AB
z2 z1 (
xA2 1 xA2
2
2
1 3
3
x A2 + x ) (
A1 + 1 2
2
x A1 + x
1 3
3 A1 )

= 1+ 2
2
(
cD AB ( xA1 xA2 ) 1 xA1 xA 2
2

+3
1
3
x A1 ) (
xA2
3
+
)
z2 z1

( A1 A 2 ) ( A1 A2 )
x x x x

cD AB ( xA1 xA2 )
=
z2 z1 2 3 (
1 + 1 ( xA1 + xA2 ) + 1 x A1
2
+ x A1 x A2 + x )
2
A2 +

(6)

The Taylor expansion in Eq. C.2-3 has been used for expanding the
logarithms in line 2.

105
Note to p. 547 (ii)

a. The result in Eq. 18.2-11 may be written as follows:


xB xB2 z z1
= in which = (1,2)
xB1 xB1 z2 z1

We want to expand this result in a Taylor series in to get a result of


the form

xB
= 1 ( ) + (3)
xB1

b. Next we rework the problem in 18.2 by omitting the xA


term in the denominator of Eq. 18.2-1 (that is, assume that xA is so
small that it can be neglected with respect to unity). We can then
show that this gives the first two terms of the series in Eq. 3.

a. From Eq. C2.1 we get by expanding about =0:


xB xB2 d xB2
=
xB1 xB1
+
d xB1
( 0) +
=0 =0

x
= 1 + 1 ln B2 1 + (4)
xB1

Now from Eq. C.2-3, with (1 + x) replaced by x, and x replaced by


(1 x), we have, for 0 < x 2

ln x = ( x 1) 12 ( x 1) + 13 ( x 1) ( x 1) 4 +
2 3 1
4
(5)

Therefore,

106
xB x 1 x
2
1 x
3

= 1 + B2
1 B2
1 + B2
1 + + (6)
xB1 xB1 2 xB1 3 xB1

If xB2 is only slightly greater than xB1 (which would be the case if
species A is present only in a small amount), then we need retain
only the first term inside the bracket. The final result is then:

xB x
= 1 + B2 1 (7)
xB1 xB1

b. When xA can be neglected with respect to unity in Eq. 18.2-1,


the differential equation for xA as a function of z is:

d 2 xA
=0 (8)
dz 2

Since xA + xB = 1 , the same differential equation (with A replaced by


B) is valid for species B. Furthermore, since z and are related by the
linear expression given in Eq. (2) we may write:

d 2 xB
=0 (9)
d 2

This equation may be integrated to give:

xB x
= 1 + B2 1 (10)
xB1 xB1

in agreement with Eq. 7.

107
Note to p. 555

We wish to verify that Eq. 18.4-9 does satisfy Eq. 18.4-7 by


substituting into the differential equation.
Differentiate Eq. 18.4-9 with respect to , using Eq. C.5-10:

d sinh (1 ) ( )
= (1)
d cosh

A second differentiation then gives, using Eq. C.5-11:

d2
=
( )
cosh (1 ) + 2
(2)
d 2 cosh

By substituting the second derivative from Eq. 2 and the expression


in Eq. 18.4-9 into Eq. 18.4-7, it is seen that an equality is obtained.

108
Note to p. 563 (i)

Here we verify that Eq. 18.6-8 is a solution to the differential


equation given in Eq. 18.6-2.
Because of the definition of f just above Eq. 18.6-6, f = cA c A0
must satisfy Eq. 18.6-2. Therefore, we calculate the derivatives using
the Leibniz formula given in C.3:

f
z
1
= 4 exp 3
3 () z
( )
=
1
( )
( )( )
4
exp 3 13 yz 4 3 ( a 9D AB )
13
(1)
3

f f
z
1
= 4 exp 3
3 () z y
( )1
= 4 exp 3
3 y () ( )
=
1
( )
( )
4
exp 3 ( a 9D AB z )
13
(2)
3

2 f
y 2
=
1
( )
4
( )
exp 3 ( a 9D AB z )
13
( 3 ) ( a 9D z )
2
AB
13
(3)
3

Then, substituting these expressions into Eq. 18.6-2, we get:

( )
ay + 13 yz 4 3 ( a 9D AB )
13
( )
= D AB +3 2 ( a 9D AB z )
23
(4)

If both sides are multiplied by ( 9D AB z a ) , and replace


13
2 by
y 2 ( a 9D AB z )
23
, we get

ay 2
= 3D AB y 2 ( a 9D AB z ) (5)
3z

which is an identity. This concludes the proof.

109
Note to p. 563 (ii)

The concentration profiles are given by Eq. 18.6-8.

c A0 3

( )
cA =
e d (1)
4
3

where = y ( a 9D AB z ) . We show how to get the next two equations


13

from this result.


Notice in Eq. 1 the bar over , which is used to make a
distinction between the dummy variable of integration and the
dimensionless distance. This distinction is vital when we apply the
Leibniz rule (Eq. C.3-2) for differentiation of an integral to get the
local molar flux at the wall is then:

c A D AB c A0 d 3 d
N Ay = D AB = e (2)
y=0 y y=0 ()
43 d dy
y=0

Note that only the third term in Eq. C.3-2 contributes to the
derivative (i.e., the term involving the lower limit of the integral):

D AB c A0 3 a
( )
13 13
D AB c A0 a
N Ay = e = (3)
y=0 43
() 9D AB z
=0
43 9D AB z ()
Then the total molar flow across the surface of width W and
length L is given by the integral over that surface:

13
W L D c a L
WA = 0 0 N Ay dzdx = AB4 A0 W 0 z 1 3 dz
y=0 9D ()
3 AB

L
D c a
13
z2 3 2D AB c A0 WL a
13

= AB4 A0 W 0
= (4)
9D ()
3 AB
2
3
4
3
43 () 9D L
AB

110
In the last step, Eq. C.4-4 can be used to replace 4
3
( ) by ( ) to
4
3
7
3
agree with Eq. 18.6-10.

111
Note to p. 565

Let us verify that Eq. 18.7-9 satisfies the differential equation in


Eq. 18.7-6. To simplify the problem, it is a good idea to introduce
dimensionless variables:

r c k1aR 2
= ; = A ; = (1,2,3)
R c AR DA

Then the differential equation in Eq. 18.7-6 and the solution in Eq.
18.7-9 may be rewritten as:

1 d 2 d 1 sinh
= 2 ; = (4,5)
d d
2
sinh

First we use Eq. 5 to evaluate the derivative (see also C.5):

d 1 sinh cosh
= 2 + (6)
d sinh sinh

Multiplication by 2 then gives:

d sinh cosh
2 = + (7)
d sinh sinh

Further differentiation gives:

d 2 d cosh cosh 2 sinh


= + + (8)
d d sinh sinh sinh

Division by 2 gives:

1 d 2 d 1 sinh
= 2 (9)
d d
2
sinh

112
But the right side of the equation is just , according to Eq. 5. We
have therefore shown that the differential equation is satisfied.

113
Note to p. 585

The equation to be solved is in Example 19.1-1 is

dc A d 2 cA
v0 = D AB k1c A (1)
dz dz 2

which is of the form of Eq. C.1-7a. We know that the concentration of


A will decrease with increasing distance from the porous plug, so we
assume that it will have the form c A = exp ( az ) , where a is a positive
constant. When this is substituted into the Eq. 1, we get (after
canceling exp ( az ) from each term):

v0 a = D AB a2 k1 or D AB a2 + v0 a k1 = 0 (2)

This quadratic equation can be solved for a to get:

v0 v02 + 4D AB k1
a= (3)
2D AB

To insure that a is positive, we must choose the plus sign. Hence a is


given by:


( )
v
a = 1 + 1 + 4D AB k1 v02 0
2D AB
(4)

and the concentration profile is:


( )
v z
c A = exp +1 1 + 4D AB k1 v02 0
2D AB
(5)

which is the result in Eq. 19.1-20.

114
Note to p. 589

It is desired to show how to obtain Eq. (H) of Table 19.2-4 from Eq.
(E).
We move the ( q) term to the left side of the equation and
then perform mathematical operations on the new left side. First we
use Eq. 3.5-4 to rewrite the new left side as:

DH
Dt

+ ( q) = H
t
( )
+ ( q )
+ vH (1)

For a multicomponent mixture, the heat flux vector q may be written


as described in Fn. (a) of Table 19.2-4, and then neglect the
contribution q( ) . Then Eq. 1 may be rewritten as
x


N
( )
N H
DH
+ ( q ) = c H + vH
( kT ) + j
(2)
Dt t =1 =1 M

Here we have also used the relation Eq. 19.3-9 to rewrite the first
term on the right side. Next we combine the second and fourth terms
on the right side to get


DH N N
+ ( q ) = c H + vcH + J H ( kT )
Dt t =1 =1
N N N

= c H + v c H + J H ( kT ) (3)
t =1 =1 =1

Then, since c v + J = N from Table 17.8-1, Eqs. (G) and (H),


DH N N

+ ( q ) = c H + N H ( kT ) (4)
Dt t =1 =1

Combining this with Eq. (E) then gives Eq. (H).

115
Note to p. 591

The theorem of Euler (pronounced "Oiler") for homogeneous


functions is used in order to get Eq. 19.3-9. We here give a proof of
Euler's theorem.
First we give a definition of a homogeneous function. A function
of n variables is said to be "homogeneous of degree k" if

f ( x1 , x2 , x3 , xn ) = k f ( x1 , x 2 , x3 ,xn ) (1)

That is, if in the function f ( x1 , x 2 , x3 ,xn ) , we replace x1 by x1 , x2


by x2 , etc., then this will give a result that is the same as multiplying
the original function by k .
If we differentiate both sides of Eq. 1 with respect to , we get

f ( x1 ) f ( x2 ) f ( xn )
+ + = k k1 f ( x1 , x2 , x3 ,xn )
( x1 ) ( x2 ) ( xn )
(2)
where, on the left side, it is understood that by f we mean the
function f ( x1 , x2 , x3 , xn ) . It is understood that the function f
has continuous first partial derivatives. We now perform the
differentiations on the left side to get

f f f
x1 + x2 + xn = k k1 f ( x1 , x2 , x3 ,xn ) (3)
( x1 ) ( x2 ) ( xn )

Next, we set = 1, to get

f f f
x1 + x2 + xn = kf ( x1 , x2 , x3 ,xn ) (4)
x1 x2 xn

which is Euler's theorem. Here, the functionality of f is exactly the


same on both sides of the equation. Since the enthalpy is a
homogeneous function of degree "1" Eq. 19.3-9 follows directly.

116
H
n n =H (5)
n ( ) ,T ,p

or

n H =H (6)

This result is frequently used in discussions of mixtures. An example


is the relation immediately after Eq. 17C.1-3, where, for a binary
mixture

nAVA + nBVB = V

or, when the entire equation is divided by V,

c AVA + cBVB = 1

Another example of Euler's equation is in going from Eq. 1 to Eq. 2 in


the Note to p. 589.

117
Note to p. 615

We want to verify that Eqs. 20.1-16 and 17 are a solution to 20.1-


9 and 10. We make use of C.6 on the error function.
First we find the first and second derivatives of X with respect
to Z:

1 erf ( Z )
X= (1)
1 + erf

dX 1 2 ( Z )2
= e (2)
dZ 1 + erf

d2X 1 2 ( Z )2
= e 2 ( Z ) (3)
dZ2 1 + erf

When the derivatives in Eqs. 1, 2, and 3 are substituted into Eq. 20.1-
9, it is seen that the latter equation is satisfied.
Next we substitute the first derivative from Eq. 2 into Eq. 20.1-
10 to get:

1 x 1 2 ( Z )2
= A0 e
2 1 xA0 1 + erf Z=0
2
x e
= A0 (4)
1 xA0 (1 + erf )

which agrees with Eq. 20.1-17.

118
Note to p. 622

Here we give a more detailed discussion of the development


between Eqs. 20.1-67 and Eq. 20.1-74.
When the bracket in Eq. 20.1-67 is set equal to unity, we get:

d2 g dg
+ 2 =0 (1)
d 2 d

where g = c A c A0 and = z ( t ) , which has the solution (by analogy


with Example 4.1-1)

cA z
= 1 erf (2)
cA0

with given by Eq. 20.1-70. To get Eq. 20.1-72, we differentiate the


concentration profile:


c A z
N Az0 = D AB = D AB c A0 erf
z z
4D AB 0 S ( t ) S (t ) dt
t 2
z=0
z=0
(3)
Use Eq. C.6-2 to differentiate the error function, and get

1 2
2
4D AB 0 S ( t ) S (t ) dt
t 2
N Az0 = +D AB c A0

1 2
D AB 1
0 S ( t ) S ( t ) dt
t 2
= c A0 (4)
t t

The total number of moles of A that have crossed the mass-transfer


surface S(t) at time t is then given by:

MA = S ( t ) c A0 0 1 erf ( z / ) dz = S ( t ) c A0 0 erfc ( z / )dz



(5)

119
The complementary error function "erfc (...)" is defined in C.6. We
now insert this function into Eq. 5:

2 2
MA = S ( t ) c A0 z / e d dz = S (t ) c A0 ( z ) d
2 2
0 e d
0 0

(6)

In the second form, the order of integrations over and z has been
interchanged. Now the integral over z can be performed to give

2 2 1
MA = S ( t ) c A0 d = S ( t ) cA0
2
0 e (7)
2

Inserting the expression for from Eq. 20.1-70, we then get

1 4D AB
MA = S ( t ) c A0 4D AB 0 S ( t ) S ( t ) dt = c A0 0 S ( t )
t 2 t 2
dt

(8)
which is Eq. 20.1-73.
Another expression can be obtained from integrating Eq. 20.1-
72:

0 S ( t ) N Az0 ( t )dt
t
MA =
4D AB t S( t )
= c A0
0 dt (9)
0 S ( t ) S ( t )
t 2
dt

It is relatively easy to show that Eqs. 8 and 9 are the same. We first
note that the S ( t ) in the denominator can be removed from the
integral, so that Eq. 9 can be rewritten as

S ( t )
2
4D AB t
MA = c A0
0 dt (10)
0 S ( t )
t 2
dt

120
Then we make use of the fact that the square root in the denominator
contains no t and hence can be removed from the integral over t to
give
0 S ( t )
t 2
4D AB dt
MA = c A0

()
2
0 S t dt
t

4D AB
0 S ( t ) dt
t 2
= c A0

4D AB
0 S ( t )
t 2
= c A0 dt (11)

This is the same as Eq. 8. Proving that the two expressions for MA are
the same is a stronger statement than that the two expressions give
the same results for dMA dt .
This example is also a good illustration of the importance of
using a symbol for the dummy variable of integration that is different
from that used as one of the limits in the integral.

121
Note to p. 626

We want to verify that the limiting solutions for slow and fast
reactions in Eqs. 20.2-16 to 18 and 20.2-19 are correct.

(a) Slow reactions


We start by rewriting Eq. 20.2-13 in terms of dimensionless
variables. We do this by using the definition of in Eq. 20.2-17:

1 4 d 3
= + 3 + 2 (1)
Sc 3 d

Then, using the expansion in Eq. 20.2-16, we have

(
= Sc 1 3 1 + a1 + a2 2 + ) (2)

( ( )
2 = Sc 2 3 1 + 2a1 + a12 + 2a2 2 + ) (3)

3 = Sc 1 (1 + 3a + ( 3a + 3a )
1
2
1 2
2
+ ) (4)

Substitution of these three expressions into Eq. 1 gives

1=
4
3
( ( ) ) (
3a1 + 2 3a12 + 3a2 2 + + 1 + 3a1 + 3a12 + 3a2 2 + ( ) )
(
+ Sc1 3 1 + 2a1 + a12 + 2a2 2 + ( ) ) (5)

We now equate terms in the same powers of :

Zeroth power of : 1=1 (6)

First power of : 0 = 4a1 + 3a1 + Sc1 3


or a1 = 17 Sc1 3 (7)

Second power of : 0= 8
3 (a 3
1 ) ( )
+ 4a1 a2 2 + a13 + 4a1 a2 2 + 2Sc1 3 a1 2
or ( )
0 = 11 a12 + a2 + 2Sc1 3 a1

122
or 0 = 11 ( 1
49 ) (
Sc 2 3 + a2 + 2Sc1 3 17 Sc1 3 )
or a2 = + 539
3
Sc 2 3 (8)

Thus we have obtained the first two coefficients in the expansion of


Eq. 20.2-16, and hence the first few terms in the expression for as a
function of .

(b) Fast reactions


When the trial function for is taken to be of the form

= K m m<0 (9)

When this is substituted into Eq. 1, we get

1 4
= ( 3m ) K 3 3m + K 3 3m + K 2 2m+1 (10)
Sc 3

The ratio of the last term to either of the first two terms is
proportional to m+1 , which is a positive quantity. For large , the
last term is the dominant term on the right side. Therefore Eq. 1
becomes

1
= 2 (11)
Sc

and therefore the solution to Eq. 1 for this case is

= ( Sc )
1 2
(12)

in agreement with Eq. 20.2-19.

123
Note to p. 692

We begin by formulating the problem as in Problem 12.1-4:

Solid Liquid
Cs 1 2 Cs Cs
= 2 (1) = NCl (5)
=1

Cs = finite at = 0 (2)
Cs = Cl at = 1 (3)
Cs = 1 at = 0 (4)

Taking the Laplace transform of the problem we get

Solid Liquid
1 d 2 dCs dCs
pCs 1 = 2 (6) = NCl (9)
d d d =1

Cs = finite at = 0 (7)
Cs = Cl at = 1 (8)

The solution to the homogeneous equation corresponding to Eq. 6 is


the complementary function

K1 K
Cs,cf = cosh p + 2 sinh p (10)

The particular integral of Eq. 6 is, by inspection

1
Cs,pi = (11)
p

The complete solution to Eq. 6 is then

K1 K 1
Cs = cosh p + 2 sinh p + (12)
p

124
The boundary condition in Eq. 7 requires that K1 = 0 . The boundary
condition in Eq. 8 requires that

1
Cs =1
= K 2 sinh p + = Cl (13)
p

and combination of Eqs. 9 and 12 gives

(
NCl = K 2 sinh p + p cosh p ) (14)

Elimination of Cl between Eqs. 13 and 14 leads to an expression for


the integration constant K 2 , and hence also to Cs :

1 N sinh p
Cs = (15)
p p p cosh p + ( N 1) sinh p

Next, the Laplace transform of the total amount of A within the


sphere is

MA 1 1 N p x sinh x
= 0 Cs d = 2 0
2
dx
4 R 3 c 0 3p p p cosh p + ( N 1) sinh p
1 N p x sinh x
= 2 0 dx
3p p p cosh p + ( N 1) p
p
1 N x cosh x sinh x
= 2
3p p p cosh p + ( N 1) sinh p
0

1 N p cosh p + ( N 1) sinh p Nsinh p


= 2
3p p p cosh p + ( N 1) p
1 N Nsinh p
= 2 1
3p p p cosh p + ( N 1) p

125
1 N p cosh p + ( N 1) sinh p Nsinh p
=
3p p 2 p cosh p + ( N 1) sinh p
1 N Nsinh p
= 2 1
3p p p cosh p + ( N 1) sinh p
1 N N2
= + (16)
3p p 2 p 2 p coth p + ( N 1)

Now we take the inverse Laplace transform of the above; the


transforms of the first two terms may be found in an elementary table
of transforms:


MA 1 2 1 1
= N + N L (17)
4 R 3 c 0 3 p
2
( p coth p + ( N 1)
)
To get the inverse transform of the last term, we can use the
Heaviside partial fractions expansion theorem for repeated roots,
which is:

f ( p ) = N ( p ) D ( p ) with D ( p ) = ( p a1 ) ( p a2 )m ( p an )m
m1 2 n
If ,
N ( p ) is a polynomial of degree less than ( m ) 1 , and
j ai ak for
i k , then

n mk kl ( ak )
f (t ) = t mk l e akt
k=1 l=1 ( mk l ) ! ( l 1) !

d l1 N ( p ) D ( p)
with kl ( p ) = and D ( p ) = .
dp l1 Dk ( p ) ( p ak )
k mk

The contribution from the factor p 2 is then (with a1 = 0 , m1 = 2 , k=1)

11 ( 0 ) 12 ( 0 )
t+ = 11 ( 0 ) + 12 ( 0 ) (18)
( 2 1) !(1 1) ! ( 2 2 ) ! ( 2 1) !

126
where

N2 N2
11 ( 0 ) = = = N (19)
p coth p + ( N 1) 1 + ( N 1)
p=0

d N2
12 (0 ) =
dp p coth p + ( N 1 ) p=0


= N 2
(
( d dp ) p coth p + ( N 1) ) =
1
(20)

( ( ) )
2
+ 3
p coth p N 1
p=0

The contributions in Eqs. 19 and 20 just exactly cancel the first two
terms in Eq. 17.
The contribution from the remaining factor in L1 { } is

MA ( ) 1 N ( p )
N ( ak ) + a
= L = e k (21)
4 R c0 D ( p ) k= 2 D' ( ak )
3


e + ak
=N 2

k= 2 p2
( 1
2
p1 2 coth p 12 csch 2 p + 2p) ( p coth p + ( N 1) ) p= ak
------------------------------

where the prime on D indicates differentiation with respect to p, and


the ak are the zeros of the denominator in the braces in Eq. 17 (except
for the double zero from p 2 , which we have already taken into
account). This means that the dashed underlined term in Eq. 21 may
be omitted.
We know on physical grounds that the quantity M(t) must be a
decreasing function of time. Therefore, the ak must be negative. This
can be guaranteed by setting ak equal to k2 where the k are real
numbers. Then we have

127
N ( p ) sinh 2 p 2
== 2N e k
1 2
L
D ( p ) (
k= 2 p 2 p 1 2 sinh p cosh p 1
p= k2 )
i 2 sin 2 k i k 2
= 2N 2
e k
k= 2 ( )( i sin
4
k k cos k ik )
sin 2 k 2
= 2N 2
e k (22)
k= 2 ( )
k3 ( k sin k cos k )

Therefore, the total amount of A within the sphere is at any time t is:

MA ( t )
( )

= 6N 2 Bn exp n2D AB t / R 2 (23)
4
3
R 3 c0 n=1

where the n are determined from

n cot n + ( N 1) = 0 (24)

and the Bn are

N 2 sin 2 n
Bn = (25 )
n3 ( n sin n cos n )

For infinite N these last two expressions may be simplified. If, in Eq.
24, N is infinite, sin k must be zero, and therefore k must be n . If,
in Eq. 25, N

(
Bn 3
n cot n ) sin 2 n
2
n2
= 4 =
1
(26)
n ( n sin n cos n ) n (n )2

Thus we obtain the results in Eqs. 22.4-34 and 35.

128
Two references should have been cited here:
E. N. Lightfoot, in Lectures in Transport Phenomena, AIChE, New
York (1969), pp. 59-60.
H. Grber, S. Erk, and U. Grigull, Die Grundgesetze der Wrme-
bertragung, Springer-Verlag, Berlin, 3rd edition (1961), pp. 55-62.

129
Note to p. 766

Here we work through the details of the development on p. 766,


leading up to the expression for the generalized diffusional driving
force at the bottom of the page.
When Eq. 24.1-2 is applied to a moving element of fluid, we
write


DU DS DV N G D
=T p + (1)
Dt Dt Dt =1 M Dt

in which


DU N
= ( q ) ( :v ) + ( j g ) Table 19.2-4, Eq. D (2)
Dt =1
[Note footnote b to Eq. D!!]
DV D 1 1 D 1
=
Dt Dt
= = +

( v ) Table 19.2-3, Eq. A (3)
2 Dt

D
= ( j ) + r Table 19.2-3, Eq. B (4)
Dt

Substitution of these expressions into Eq. 1 gives

DS 1 N
p
= ( q ) ( :v ) + ( j g ) + ( v )
Dt T =1 T


1 N G

T =1 M
(
( j ) + r )
1 N G 1 1 N G
= ( q)
T

( j ) T (:v ) T M r
=1 M =1

1 N
+ ( j g )
T =1

130
1 N G 1
= q j q 2 T +
T =1 M T
N 1 G 1 1 1 N G
j g ( :v ) r (5)
=1 T M T T T =1 M

Comparison of Eq. 5 with Eq. 24.1-1 gives the entropy flux vector and
the entropy production rate (Eqs. 24.1-3 and 4):

1 N G
s = q j (6)
T =1 M

1 N 1 G 1 1 1 N G
gS = q 2 T j g ( :v ) r
T =1 T M T T T =1 M
(7)

( H )
N
Next we rewrite Eq. 6 by replacing q by q( ) +
h
M j --that is,
=1
we subtract off the heat flux associated with the diffusion of the
chemical species. This gives us then for the entropy flux (Eq. 24.1-5):

1 ( h) N S
s = q j (8)
T =1 M

The entropy flux is now written as the sum of two terms: the first
term is the entropy flux associated with heat flow, and the second
term is that connected with diffusion of the chemical species. (For

( H )
N
more on the reasons for replacing q by q( ) +
h
M j , see S. R.
=1
de Groot and P. Mazur, Non-Equilibrium Thermodynamics, North-
Holland, Amsterdam (1962), pp. 24-25.)
In order to rewrite the entropy production term, we make use
of the Gibbs-Duhem relation in the entropy representation (see H. B.
Callen, Thermodynamics and an Introduction to Thermostatics, Wiley,
New York (1985), pp. 60-62):

131
1 p N G
Ud + Vd n d = 0 (9)
T T =1 T

where n is the number of moles of species . Division by V and


doing some elementary manipulations gives us the form of the
Gibbs-Duhem equation that we need:

N H 1 1 N 1 G
T + p
=0 (10)
=1 M T 2 T =1 T M

Next, in Eq. 7, we add four terms inside the bracket in the second
term on the right side in such a way that the term is not changed:

1 G 1 H 1
p + 2 T
N j
1 T M T M T


gS = q 2 T
T =1
H 1 1 1 N
2 T g + g
M T T T =1
1 1 N G

T
( ) T M r
:v (11)
=1

The terms containing H clearly cancel one another, and the terms
containing p and g do not contribute to the sum, inasmuch as
j = 0 . Thus, Eq. 11 may now be rewritten as:

1 G 1 H 1
p + 2 T
N j T M T
h 1 M T
gS = q( ) 2 T
T =1 1 1 N
g + g
T =1

T

1 1 N G
( :v ) r (12)
T T =1 M
or

132
( ) j 1 1 N G
N
TgS = q( ) lnT cRTd T ( :v ) T M r
h

=1 =1
(13)

The d are the generalized driving forces for diffusion . Note that the
above development guarantees that d = 0 , as is required, since
j = 0 . We can see that d = 0 , since the first three terms in the
bracket in the first line of Eq. 12 sum to zero according to the Gibbs-
Duhem equation, and the fourth and fifth terms clearly combine to
sum to zero.
The above development has led us to the expression for the
generalized driving forces for diffusion:

G N
cRTd = c T + c H lnT p g + g (14)
T =1

To get the alternative form for d given in the second line of Eq. 24.1-
8, we follow the discussion of Ref. 5 on p. 766.
The quantities G , H , and p are functions of the state of a fluid
element, which may be described by the temperature T, the pressure
p, and the set of (N 1) mole fractions x where = 1, 2,3, ... ( N 1).
Then the first term in Eq. 14 may be written by expanding G using
the chain rule of partial differentiation to give:

G
c T = c G c G lnT
T
N 1 G G G
= c x x + c T + c p c G lnT
T

=1 p
N 1 G
= c x x c TS lnT + c V p c G lnT (15)
=1

133
When Eq. 15 is combined with Eq. 14, it is seen that the lnT terms
exactly cancel, and the p terms may be combined. Furthermore,
when the relation dG = RTdlna is used, we get finally

N 1
lna N
cRTd = c RT x + ( ) p g + g
=1 lnx =1
T ,p,x
(16)

in which = c V is the volume fraction of species , and the


subscript x stands for all the x except x and xN . Thus, we see that
the driving forces d include the contributions from the mole fraction
gradients, the pressure gradient, and the external forces.
Equation 16 is in agreement with Eq. 7.8 of Ref. 5 at the bottom
of p. 766, and also with Eq. 11.1-27 of Molecular Theory of Gases and
Liquids, by Hirschfelder, Curtiss, and Bird, Wiley, New York (1964).
Therefore, the first term of the second line of Eq. 24.1-8 of BSL may be
misleading. Here and elsewhere (Eqs. 24.2-8, 24.2-9, 24.2-10, 24.4-1,
and 24.5-4) the abbreviated notation

N 1 lna
x T ,p lna (17)
=1 lnx
T ,p,x

is used (this is just the chain rule applied to lna with the p and
T terms omitted). This notation has also been used by E. N.
Lightfoot, Transport Phenomena and Living Systems, Wiley, New York
(1974), Eqs. 1.2.7 and 12, on p. 163, and W. M. Deen, Analysis of
Transport Phenomena, Oxford University Press (1998), Eq. 11.6-2.
Deen, however, states that his treatment is for dilute mixtures. Both
of these authors, however, appear to have their summations going
from 1 to N (instead of 1 to (N 1).
The discussion in Multicomponent Mass Transfer, by R. Taylor
and R. Krishna, Wiley, New York (1993), also uses essentially the
notation of Eq. 17 above, with the summation going from 1 to (N 1)
as may be seen on pp. 23, 24, and 29 (their Eq. 2.3.10 is, aside from
some notational differences, the same as Eq. 16 above).

134
In Advanced Transport Phenomena, by J. C. Slattery, Cambridge
University Press (1999), p. 450, Eq. 8.4.3-4, gives a result that is
consistent with Eq. 16 above, although the notation is considerably
different from ours.
There is also a discussion of multicomponent systems in
Transport Phenomena Fundamentals, by J. L. Plawsky, pp. 66-69 (heat
flux) and pp. 69-73 (mass fluxes); however, no derivations of the
expressions are given.
Thermodynamics of Irreversible Processes, by G. D. C. Kuiken,
Wiley, New York (1994), has a discussion of the driving force d on
p. 183.

135
Note to p. 767

Just as a check, we show that Fick's law for a binary system


may be obatained by simplifying Eq. 24.2-3. This equation is, for a
binary system with the two components labeled "A" and "B":

jA = + A ( D AAd A + D ABd B ) (1)

Use Eqs. (A) and (C) of Table 24.2-1 and the equation d A + d B = 0 (see
line 2 on p. 767), to eliminate the generalized Fick diffusivities in
favor of the Maxwell-Stefan diffusivities:

B2
jA = + A DABd A A B DABd A (2)
x A xB x A xB

This may be rewritten, using the fact that A + B = 1 , to give

A B
jA = D d (3)
xA xB AB A

Next use the second line of Eq. 24.1-8, omitting the terms for thermal
diffusion, pressure diffusion, and forced diffusion. This yields:

A B ln aA
jA = DAB ( xA ln aA ) = A B DAB x A (4)
xA xB xA xB ln xA

Then, to obtain the molar flux of species "A" we use Eq. 17B.3-1 to get
the relation J*A = ( cxA xB A B ) jA . Finally, combine this equation
with Eq. 4 and introduce the binary diffusivity D AB from Eqs. 24.3-2
and 4, to get:

J*A = cD AB xA (5)

This is the form of Fick's (first) law given in Eq. B of Table 17.8-2 for a
binary system.

136
Note to p. 768

Here we show how to derive the generalized Maxwell-Stefan


relations (Eq. 24.2-4) from the generalized Fick equations (Eq. 24.2-3).
This proof was first given by H. J. Merk, Appl. Sci. Res., 73-99 (1959),
5, Eqs. 89 and 90.
We start by introducing the abbreviation j# = j + DT lnT , in
order to avoid having to carry along the thermal diffusion term
throughout the derivation. Then the generalized Fick equation, Eq.
24.2-3, for species may be written as

j# N
= D d (1)
=1

where j# = 0, d = 0 , D = D , and D = 0 (see above Eq.



24.2-3). Next, write a similar equation for species and form the
difference of the two equations; then multiply both sides of the
equation by the quantity x x D to get

x x j# j# x x
(D )
N
= D d (2)
D D =1

Now we sum both sides over the index

x x j# j# N
( )d
x x
D = D D D (3)
=1

If the bracket quantity on the right side is set equal to , we


see that we get

x x j# j# N
( )
N
D d = d (0) = d
= d = d
=1 =1

137
x x j# j#
or = d (4)
D

This is exactly the generalized Maxwell-Stefan equation for species


(see Eq. 24.2-4). At the same time we get

(D )
x x
D D = (5)

From this we will get the relations between the D and the D .
One might wonder why we replace the bracket expression in
Eq. 3 by rather than simply , which would, after all, also
lead to the generalized Maxwell-Stefan equation. If Eq. 5 is multiplied
by and summed over , one gets the identity 0 = 0, because of
D = 0 . But if the is not included on the right side of Eq. 5,

multiplication of the equation by and summing over will give 0


= . Similar arguments may be made for not setting the bracket
expression equal to A , where A is an arbitrary constant.
Let us now return to 5 and get the relation between the Fick
and the Maxwell-Stefan multicomponent diffusivities. We start by
defining a matrix B with matrix elements ( B ) = D + D ; the
matrix B is an ( N 1) ( N 1) matrix, with the ( = )-row and the
( = )-column missing. We now rewrite Eq. 5, omitting the equation
for = :

x x
D
(B ) = ( ) (6)

Next we perform some operations on the relation D = 0:


(D )
D + D = 0 or ( B ) +D = 0 (7)

138
Then we multiply Eq. 6 by B1 ( )

and sum on to get:

x x
D
= B1

( )
(8)

Multiplying the second form of Eq. 7 by B1 ( )


and summing on
gives


( )
= D B1

(9)

Then multiplying the reciprocal of Eq. 8 by Eq. 9, and replacing the


index by the index , gives Eq. 24.2-7 of the textbook:

x x
D ( adjB )

D = (10)
( adjB )

In the last step we have also made use of the fact that

(B ) = (adjB ) (det B )
1
(11)

where adjB is the matrix adjoint to B and det B is the determinant


of the matrix B . The matrix adjB is the transpose of the matrix of
cofactors (or "signed minors") of B .

139
Note to p. 769

As a check on Eq. 24.2-7, we use it to get Eq. (B) of Table 24.2-2.


We first write Eq. 24.2-7 for the specific case of the 1-2 pair of a
ternary system:

x1 x2 D12 ( adjB1 )22 + D13 ( adjB1 )32


D12 = (1)
1 2 (adjB1 )22 + (adjB1 )32
The 2 x 2 matrix ( B1 ) may be displayed thus:

(B1 )22 (B1 )23 D22 + D12 D 23 + D13


(B1 ) = = (2)
( B1 )32 (B1 )33 D32 + D12 D + D
33 13

The adjoint matrix is the transpose of the matrix of cofactors (a


cofactor is a "signed minor"):

( adj B1 )22 (adj B1 )23 (B1 )33 (B1 )23


(adj B1 ) = =
( adj B1 )32 (adj B1 )33 (B1 )32 (B1 )22
D + D13 D 23 D13
= 33 (3)
D32 D12 D 22 + D12

We are now ready to substitute into Eq. 24.2-7:

x1 x2 D12 ( D33 + D13 ) + D13 (D32 D12 )


D12 =
1 2 ( D33 + D13 ) + (D32 D12 )
x1 x2 D12D33 + D13D32
= D D + D + D
1 2 12 33 13 32

x1 x2 +D12D33 D13D 23
= (4)
1 2 D12 + D 33 D13 D 23

In the last step, we have made use of the symmetry of the D .

140

Potrebbero piacerti anche