3.

5 SPECIAL FUNCTIONS 87

Now when p is zero, we can express the gamma function Thus, Jk(x) has the same representation as given in
as a factorial, Gðn þ 1Þ ¼ n!; so the zero-order Bessel func- Eq. 3.149, except we replace Gðn þ k þ 1Þ ¼ ðn þ kÞ!:
tion of the first kind is represented by
X
1
ð
1Þn ðð1=2ÞxÞ2n 3.5.2 Modified Bessel’s Equation
J 0 ðxÞ ¼
n¼0
n!  n! As we saw in Example 3.3, another form of Bessel’s equa-
ðð1=2ÞxÞ2 ðð1=2ÞxÞ4 tion arises when a negative coefficient occurs in the last
¼1
þ
 ð3:150Þ
ð1!Þ2 ð2!Þ2 term, that is,

The reader can see, had we applied the method of

d 2y dy
Frobenius, the second solution would take the form x2 þx
ðx2 þ p2 Þy ¼ 0 ð3:156Þ
" dx2 dx
X1
ð
1Þmþ1 ðð1=2ÞxÞ2mþ2
y2 ¼ b0 J 0 ðxÞ  ln x

½ðm þ 1Þ!2 This can be obtained directly from Eq. 3.143 by replacing x
m¼0
 # with ix (since i2 ¼
1). The solution when p is not integer
1 1 or zero yields
 1 þ þ  þ ð3:151Þ
2 mþ1
y ¼ AJ p ðixÞ þ BJ
p ðixÞ ð3:157Þ
The bracketed function is called the Neumann form of the
second solution. However, the most widely used and and if p is integer or zero, write
tabulated function is the Weber form, obtained by adding
ðg
ln 2ÞJ 0 ðxÞ to the above, and multiplying the whole y ¼ AJ k ðixÞ þ BY k ðixÞ ð3:158Þ
by 2/p. This is the standard form tabulated, and is given
the notation However, because of the complex arguments, we introduce

   the modified Bessel function, which contains real argu-
2 1 ments, so if p is not integer (or zero), write
Y 0 ðxÞ ¼ ln x þ g J 0 ðxÞ
p 2
y ¼ AI p ðxÞ þ BI
p ðxÞ ð3:159Þ
2X 1
ð
1Þmþ1 ðð1=2ÞxÞ2mþ2

wðm þ 1Þ ð3:152Þ
p m¼0 ½ðm þ 1Þ!2 or, if p is integer k (or zero), write

where as before y ¼ AI k ðxÞ þ BK k ðxÞ ð3:160Þ

1 1
wðm þ 1Þ ¼ 1 þ þ    þ The modified Bessel functions can be computed from the
2 mþ1
general result for any p
The Euler constant g is defined as
  X1
ðð1=2ÞxÞ2nþp
1 1 1 I p ðxÞ ¼ ð3:161Þ
g ¼ lim 1 þ þ þ    þ
ln ðmÞ ¼ 0:5772 n¼0
n!Gðn þ p þ 1Þ
m!1 2 3 m
ð3:153Þ If p is integer k, then replace Gðn þ k þ 1Þ ¼ ðn þ kÞ! The
second solution for integer k is after Weber
Thus, the general solution when p ¼ 0 is represented using
symbols as
  
1
K k ðxÞ ¼ ð
1Þkþ1 ln x þ g I k ðxÞ
2
yðxÞ ¼ AJ 0 ðxÞ þ BY 0 ðxÞ ð3:154Þ  2m
k
1 X ð
1Þ ðk
m
1Þ! 1
k
1 m
þ x
It is obvious that limx!0 Y 0 ðxÞ ¼
1; so that for conserva- 2 m¼0 m! 2
tive systems, which include x ¼ 0 in the domain of the
physical system, obviously one must require B ¼ 0 to attain 1X 1
ð
1Þk ðð1=2ÞxÞ2mþk
þ ½wðm þ kÞ þ wðmÞ
finiteness at the centerline (symmetry). 2 m¼0 m!ðm þ kÞ!
When p is integer we again use the Weber form for the ð3:162Þ
second solution and write (after replacing p with integer k)
Plots of J0(x), J1(x), Y0(x), I0(x), and K0(x) are shown
yðxÞ ¼ AJ k ðxÞ þ BY k ðxÞ ð3:155Þ in Fig. 3.1.
88 SERIES SOLUTION METHODS AND SPECIAL FUNCTIONS

where
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
s  ffi
1 1
a 2
p¼
c ð3:165Þ
s 2

The types of Bessel functions that arise depend on the char-

acter of (d)1/2=s and the values of p.
pffiffiffi
1. If d =s is real and p is not integer (or zero), then Zp
denotes Jp and Z
p denotes J
p.
pffiffiffi
2. If d =s is real and p is zero or integer k, then Zp
denotes Jk and Z
p denotes Yk.
pffiffiffi
3. If d =s is imaginary and p is not zero or integer, then
Zp denotes Ip and Z
p denotes I
p.
pffiffiffi
4. If d =s is imaginary and p is zero or integer k, then
Zp denotes Ik and Z
p denotes Kk.

FIGURE 3.1 Plots of J0(x), J1(x), Y0(x), I0(x), and K0(x).

EXAMPLE 3.5
3.5.3 Generalized Bessel’s Equation
Very often, Bessel’s equation can be obtained by an ele- Pin promoters of the type shown in Fig. 3.2a and 3.2b are used in
mentary change of variables (dependent, independent, or heat exchangers to enhance heat transfer by promoting local wall
both variables). For the general case, we can write turbulence and by extending heat transfer area. Find an expression
to compute the temperature profile, assuming temperature varies
d 2y dy mainly in the x-direction. The plate temperature Tb, fluid tempera-
x2 2
þ xða þ 2bxr Þ
dx dx ture Ta, and heat transfer coefficient h are constant.
  We first denote the coordinate x as starting at the pin tip, for
þ c þ dx
bð1
a
rÞxr þ b2 x2r y ¼ 0
2s
ð3:163Þ
geometric simplicity (i.e., similar triangles). Heat is conducted
Representing Z p ðxÞ as one of the Bessel functions, a along the pin axis and is lost through the perimeter of incremental
general solution can be written area As ¼ 2pyDs; so that writing the steady-state conservation law
gives, where at any plane through the pin, the cross-sectional area
" pffiffiffiffiffiffi ! pffiffiffiffiffiffi !#
jd j s jd j s is A ¼ py2
ð1
a=2Þ
ðbxr =rÞ
y¼x e AZ p x þ BZ
p x
s s
ðqx AÞjx
ðqx AÞjxþDx
h2pyDsðT
T a Þ ¼ 0 ð3:166Þ
ð3:164Þ

FIGURE 3.2 (a) Pin promoters attached to heat exchange surface. (b) Geometry of a single pin.
 3.5 SPECIAL FUNCTIONS 89

Now, similar triangles show y=x ¼ b=H and the incremental length so we see that this solution is admissible at z ¼ 0. It is easy to show
Ds ¼ Dx=cosb; so that replacing y and dividing by Dx yields for small arguments
pffiffiffi K nffiffiffiðxÞ ’ 2n
1 ðn
1Þ!x
n ; hence approx-
that p
imately K 1 ð2 zÞ ’ 0:5= z; so that in the limit
dðqx AÞ 2pxbh pffiffiffi


ðT
T a Þ ¼ 0 ð3:167Þ lim z
1=2 K 1 ð2 zÞ ! 1
dx Hcosb z!0

Replacing Thus, this function is inadmissible in the domain of z ¼ 0, so take

 2 B0 ¼ 0. This same conclusion would have been reached following
dT b a Frobenius analysis, since the series solution would obviously
qx ¼
k and A ¼ px2
dx H contain ln (x) as taught in the previous sections.
Whereas boundary conditions were not explicitly stated, it is
gives clear in Fig. 3.2a that T(H) ¼ Tb or u(lH) ¼ Tb
Ta; hence, the
arbitrary constant A0 can be evaluated
 
 pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
d dT 2h A0 ¼ lH ðT b
T a Þ=I 1 ð2 lH Þ ð3:174Þ
x2
x ðT
T a Þ ¼ 0 ð3:168Þ
dx dx kðb=H Þcosb
In terms of x, the final solution is
The equation can be made homogeneous by replacing (T
Ta) ¼ rffiffiffiffi pffiffiffiffiffiffi
TðxÞ
T a H I 1 ð2 lxÞ
u, and we shall denote the group of terms as ¼ pffiffiffiffiffiffiffi ð3:175Þ
Tb
Ta x I 1 ð2 lH Þ
2h
l¼ The dimensionless temperature (T(x)
Ta)=(Tb
Ta) arises natu-
½kðb=H Þcos b
rally and can be computed directly either from expansions for I1
so differentiating yields or from tables (Abramowitz and Stegun 1965).

d 2u du We need the differential or integral properties for Bessel

x2 þ 2x
xlu ¼ 0 ð3:169Þ functions to compute the net rate of heat transfer. We dis-
dx2 dx
cuss these properties in the next section, and then use them
to complete the above example.
We now introduce a dimensionless length scale, z ¼ lx to get
finally
3.5.4 Properties of Bessel Functions
d 2u du
z 2
þ 2z
zu ¼ 0 ð3:170Þ It is easily verified that all power series presented thus far
dz2 dz
as definitions of Bessel functions are convergent for finite
values of x. However, because of the appearance of ln(x)
Comparing this with the generalized Bessel relation, Eq. 3.163,
in the second solutions, only Jp(x) and Ip(x) are finite at
indicates we should let
x ¼ 0ðp  0Þ: Thus, near the origin, we have the important
results:
b ¼ 0; a ¼ 2; c ¼ 0; d ¼
1; s ¼ 12
k ¼ 0; J 0 ð0Þ ¼ I 0 ð0Þ ¼ 1 ð3:176Þ
and moreover, Eq. 3.165 suggests
k > 0ðintegerÞ; J k ð0Þ ¼ I k ð0Þ ¼ 0 ð3:177Þ
1 pffiffiffi
p¼1 and d ¼ 2iðimaginaryÞ ð3:171Þ
s
p > 0; J
p ð0Þ ¼ I
p ð0Þ ! 1 ð3:178Þ
Comparing with Eq. 3.164, and item (4) in the list, we arrive at
The sign in the last expression depends on the sign of
1  pffiffiffi pffiffiffi  Gðm þ p þ 1Þ; as noted in Eqs. 3.149 and 3.161. However,
u ¼ pffiffiffi A0 I 1 ð2 zÞ þ B0 K 1 ð2 zÞ ð3:172Þ it is sufficient to know that a discontinuity exists at x ¼ 0 in
z
order to evaluate the constant of integration. We also
observed earlier that ln(x) appeared in the second solutions,
We first check for finiteness at z ¼ 0 (i.e., x ¼ 0), using the expan-
so it is useful to know (e.g., Example 3.5)
sion in Eq. 3.161
" pffiffiffi pffiffiffi #
Y k ð0Þ ¼ K k ð0Þ ! 1 ð3:179Þ
1 pffiffiffi 1 z 1 ð zÞ3
lim pffiffiffi I 1 ð2 zÞ ¼ lim pffiffiffi þ pffiffiffi þ    ¼ 1 ð3:173Þ
z!0 z z!0 z 1 z 1!2! hence only Jk(x) and Ik(x) are admissible solutions.
90 SERIES SOLUTION METHODS AND SPECIAL FUNCTIONS

Asymptotic expressions are also useful in taking limits TABLE 3.1 Selected Values for Bessel Functionsy
or in finding approximate solutions; for small values of x,
x J0(x) J1(x) I0(x) I1(x)
the approximations are
0 1.0000 0.0000 1.0000 0.0000
1 xp 2p x
p 1 0.7652 0.4401 1.266 0.5652
J p ðxÞ ’ p ; J
p ðxÞ ’ ðx  1Þ
2 Gðp þ 1Þ Gð1
pÞ 2 0.2239 0.5767 2.280 1.591
3
0.2601 0.3391 4.881 3.953
ð3:180Þ 4
0.3971
0.0660 11.30 9.759
5
0.1776
0.3276 27.24 24.34
and for integer or zero orders, we have 6 0.1506
0.2767 67.23 61.34
7 0.3001
0.0047 168.6 156.0
2n ðn
1Þ!
n
Yn ’
x ðx  1; n 6¼ 0Þ ð3:181Þ 8 0.1717 0.2346 427.6 399.9
p 9
0.0903 0.2453 1094 1031
y
2 Taken from Abramowitz and Stegun (1965).
Y 0 ðxÞ ’ lnðxÞ ðx  1Þ ð3:182Þ
p
TABLE 3.2 Zeros for Jn(x); Values of x to Produce Jn(x) ¼ 0y
The modified functions for small x are
n¼0 n¼1 n¼2
xp 2p x
p
I p ðxÞ ’ p ; I
p ðxÞ ’ ðx  1Þ 2.4048 3.8371 5.1356
2 Gðp þ 1Þ Gð1
pÞ 5.5201 7.0156 8.4172
ð3:183Þ 8.6537 10.1735 11.6198
11.7915 13.3237 14.7960
14.9309 16.4706 17.9598
K n ðxÞ ’ 2n
1 ðn
1Þ!x
n ðx  1; n 6¼ 0Þ ð3:184Þ 18.0711 19.6159 21.1170
y
Taken from Abramowitz and Stegun (1965).
K 0 ðxÞ ’
lnðxÞ ð3:185Þ

For large arguments, the modified functions sustain of Jp(x) plays an important part in finding eigenvalues for
exponential-type behavior and become independent of partial differential equations expressed in cylindrical coor-
order (p may be integer or zero): dinates, as we show in Chapter 10.
It is easy to show by variables’ transformation that
pffiffiffiffiffiffiffiffi Bessel functions of 12 order are expressible in terms of ele-
I p ðxÞ ’ expðxÞ= 2px ð3:186Þ
mentary functions:
rffiffiffiffiffiffi rffiffiffiffiffiffi
p 2
K p ðxÞ ’ expð
xÞ  ð3:187Þ J 1=2 ðxÞ ¼ sinðxÞ ð3:190Þ
2x px
rffiffiffiffiffiffi
However, for large arguments, Jp(x) and Yp(x) behave in a 2
J
1=2 ðxÞ ¼ cosðxÞ ð3:191Þ
transcendental manner: px
rffiffiffiffiffiffi rffiffiffiffiffiffi
2  p p 2
J p ðxÞ ’ cos x

p ð3:188Þ I 1=2 ðxÞ ¼ sinhðxÞ ð3:192Þ
px 4 2 px
rffiffiffiffiffiffi
rffiffiffiffiffiffi  I
1=2 ðxÞ ¼
2
ð3:193Þ
2 p p px
coshðxÞ
Y p ðxÞ ’ sin x

p ð3:189Þ
px 4 2
TABLE 3.3 Values of x to Satisfy x  J1 ðxÞ ¼ N  J0 ðxÞy
where p can be any real value including integer or zero. It
is also clear in the limit x ! 1 that Jp(x) and Yp(x) tend # N/x ! x1 x2 x3 x4 x5
to zero. The oscillatory behavior causes Jp(x) and Yp(x) to
pass through zero (called zeros of Jp(x)) and these are sepa- 0.1 0.4417 3.8577 7.0298 10.1833 13.3312
1.0 1.2558 4.0795 7.1558 10.2710 13.3984
rated by p for large x. Values of several Bessel functions are
2.0 1.5994 4.2910 7.2884 10.3658 13.4719
listed in Table 3.1 and some zeros of Jp(x) are shown in
10.0 2.1795 5.0332 7.9569 10.9363 13.9580
Table 3.2. Table 3.3 illustrates the zeros for a type III homo-
y
geneous boundary condition. The transcendental behavior Taken from Abramowitz and Stegun (1965).
 3.5 SPECIAL FUNCTIONS 91

3.5.5 Differential, Integral, and Recurrence Relations Eqs. 3.194 and 3.195 are exact differentials and yield the
key integral properties directly, for example,
The following differential properties may be proved with
reference to the defining equations and are of great Z
utility in problem solving (Mickley et al. 1957; Jenson and lxp J p
1 ðlxÞdx ¼ xp J p ðlxÞ ð3:204Þ
Jeffreys 1977):
( )
d  p  lxp Z p
1 ðlxÞ; Z ¼ J; Y; I and
x Z p ðlxÞ ¼
dx
lxp Z p
1 ðlxÞ; Z ¼ K Z
ð3:194Þ lxp I p
1 ðlxÞdx ¼ xp I p ðlxÞ ð3:205Þ

( )
d 
p 
lx
p Z pþ1 ðlxÞ; Z ¼ J; Y; K
x Z p ðlxÞ ¼ Later, in Chapter 10, we introduce the orthogonality pro-
dx lx
p Z pþ1 ðlxÞ; Z ¼ I perty, which requires the integrals
ð3:195Þ
Z x
x
8 p 9 J k ðljÞJ k ðbjÞj dj ¼ ½lJ k ðbxÞ  J kþ1 ðlxÞ
>
< lZ ðlxÞ
Z ðlxÞ; Z ¼ J; Y; I>
= 0 l
b2
2
d   p
1
x
p
Z p ðlxÞ ¼
bJ k ðlxÞJ kþ1 ðbxÞ
dx >
:
lZ p
1 ðlxÞ
p Z p ðlxÞ; Z ¼ K >
;
x ð3:206Þ
ð3:196Þ
and if l ¼ b, this gives the useful result
By applying the recurrence relations, these can also be writ-
ten in the more useful form Z x  
8 p 9 ½J k ðljÞ2 jdj ¼ 12 x2 J 2k ðlxÞ
J k
1 ðlxÞ J kþ1 ðlxÞ
>
<
lZ ðlxÞ þ Z ðlxÞ; Z ¼ J; Y; K>
= 0
d   pþ1
x
p
ð3:207Þ
Z p ðlxÞ ¼
dx >
: lZ pþ1 ðlxÞ þ p Z p ðlxÞ; Z ¼ I >
;
x
ð3:197Þ For k integer or zero, we saw in Eq. 3.201J
k ðlxÞ ¼
ð
1Þk J k ðlxÞ; so that if k ¼ 0, the right-hand side of
Most tables of Bessel functions present only positive order Eq. 3.207 may be written as
values, so the recurrence relations are needed to find nega-  2 
tive order values 1 2
2x J 0 ðlxÞ þ J 21 ðlxÞ
lx  
Z p ðlxÞ ¼ Z pþ1 ðlxÞ þ Z p
1 ðlxÞ ; Z ¼ J; Y
2p
ð3:198Þ EXAMPLE 3.6

lx  
I p ðlxÞ ¼ I pþ1 ðlxÞ
I p
1 ðlxÞ ð3:199Þ We were unable to complete the solution to Example 3.3 in order
2p to find the effectiveness factor for cylindrical catalyst pellets.
Thus, we found the expression for composition profile to be
lx  
K p ðlxÞ ¼ K pþ1 ðlxÞ
K p
1 ðlxÞ ð3:200Þ  pffiffiffiffiffiffiffiffiffiffiffi
2p
CA I 0 r k=DA
¼  pffiffiffiffiffiffiffiffiffiffiffi ð3:208Þ
Also, for n integer or zero, the following inversion propert- C As I 0 R k=DA
ies are helpful:

J
n ðlxÞ ¼ ð
1Þn J n ðlxÞ ð3:201Þ The effectiveness factor hA is defined as the ratio of actual molar
uptake rate to the rate obtainable if all the interior pellet area is
exposed to the reactant, without diffusion taking place. Thus, this
I
n ðlxÞ ¼ I n ðlxÞ ð3:202Þ maximum uptake rate for species A is computed from

K
n ðlxÞ ¼ K n ðlxÞ ð3:203Þ W max ¼ pR2 LkC As ð3:209Þ

92 SERIES SOLUTION METHODS AND SPECIAL FUNCTIONS

and the actual net positive uptake rate is simply exterior area times from the cone surface)
flux   
dT 
  Q ¼ ðpb2 Þ þ k  ð3:216Þ
dx x¼H
dC A
W A ¼ 2pRL þDA ð3:210Þ
dr r¼R pffiffiffiffiffiffiffiffiffi
This will require differentiation of I 1 ð2 ðlxÞÞ: To do this, define
pffiffiffiffiffiffiffiffiffi
We thus need to differentiate CA in Eq. 3.208, and we can use uðxÞ ¼ 2 ðlxÞ; so that from Eq. 3.196
Eq. 3.197 to do this; we find rffiffiffi

dI 1 ðuÞ dI 1 ðuÞ du l 1
 pffiffiffiffiffiffiffiffiffiffiffi ¼ ¼ I 0 ðuÞ
I 1 ðuÞ ð3:217Þ
dI 0 r k=DA dx du dx x u
dC A CAs
¼  pffiffiffiffiffiffiffiffiffiffiffi ð3:211Þ
dr I 0 R k=DA dr Thus, the net rate is
"  pffiffiffiffiffiffiffi #
kpb2 ðT b
T a Þ pffiffiffiffiffiffiffi I 0 2 lH
and from Eq. 3.197 we see Q¼
1 þ lH  pffiffiffiffiffiffiffi ð3:218Þ
H I 1 2 lH
sffiffiffiffiffiffi! sffiffiffiffiffiffi! sffiffiffiffiffiffi!
d k k k
I0 r ¼ I1 r ð3:212Þ It is clear from Table 3.1 that xI 0 ðxÞ > I 1 ðxÞ; so that the net rate is
dr DA DA DA positive as required. To find the effectiveness factor for Pin pro-
moters, take the ratio of actual heat rate to the rate obtainable if
so we finally obtain WA as the entire pin existed at temperature Tb (base temperature), which
is the maximum possible rate (corresponding to 1 conductivity)
 pffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffi I 1 R k=DA pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
W A ¼ 2pRLC As kDA  pffiffiffiffiffiffiffiffiffiffiffi ð3:213Þ Qmax ¼ þh  Acone  ðT b
T a Þ ¼ hpb b2 þ H 2  ðT b
T a Þ
I 0 R k=DA ð3:219Þ

Defining the Thiele modulus as

hence, we find
pffiffiffiffiffiffiffiffiffiffiffi   "pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi #
L¼R k=DA Q k 1 lH I 0 ð2 lH Þ
h¼ ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
1
Qmax hH
1 þ ðH=bÞ2 I 1 ð2 lH Þ
the effectiveness factor is
ð3:220Þ
2 I 1 ðLÞ WA For large arguments (lH), we can obtain the asymptotic result,
hA ¼ ¼ ð3:214Þ
L I 0 ðLÞ W A;max since I 0 I 1 (Eq. 3.186)

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Curves for hA versus L are presented in Bird et al. (1960). u k
u 2
h’t  ð3:221Þ
hb 1 þ ðH=bÞ2 cos b
EXAMPLE 3.7
Thus, the Biot number, defined as hb=k, and the geometric ratio
We were unable to express net heat flux for the Pin promoters in (H=b) control effectiveness. Thus, p for
ffiffiffi ambient air conditions,
Example 3.5, in the absence of differential properties of Bessel Bi 2, and if H ¼ b, then cos b ¼ 1= 2; hence an effectiveness is
functions. Thus, the temperature profile obtained in Eq. 3.175 was estimatedpffiffito
ffi be 0.84 (84% effectiveness). Under these conditions,
found to be expressible as a first-order, modified Bessel function lH ¼ 4 2; which can be considered large enough (see Table 3.1)
to use the approximation I 0 I 1 .
rffiffiffiffi pffiffiffiffiffiffi For small values of (lH), we can use the series expansion in
H I 1 ð2 lxÞ
TðxÞ ¼ T a þ ðT b
T a Þ pffiffiffiffiffiffiffi ð3:215Þ Eq. 3.161 to see that
x I 1 ð2 lH Þ
pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi pffiffiffiffiffiffiffi
I 0 ð2 lH Þ ’ 1 þ lH and I 1 ð2 lH Þ ’ lH þ ð lH Þ3 =2!
where

2h so approximately
l¼
½kðb=H Þcos b
k ð1=2ÞlH 1
h qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3:222Þ
To find the net rate of transfer to a single pin, we need to compute hH 1 þ ð1=2ÞlH
the heat leaving the base (which must be the same as heat loss 1 þ ðH=bÞ2
 PROBLEMS 93

Moreover, since lH  1; then we have finally that h ! 1 for (b) Define a dimensionless radial coordinate as
small lH p(to see this, replace l ¼ 2hH=(kb cos b) and
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
cos b ¼ H= H 2 þ b2 Þ. rffiffiffiffiffiffi
2h
This problem illustrates the usefulness of the asymptotic x¼r
approximations given in Eqs. 3.186–3.189. bk

and introduce y ¼ T
Ta, and thus show the ele-

PROBLEMS mentary equation

3.12. The Taylor series expansion of f (x) around a point d 2y dy

x2 þx
x2 y ¼ 0
x ¼ x0, can be expressed, provided all derivatives of dx 2 dx
f(x) exist at x0, by the series
describes the physical situation.
@f ðx0 Þ 1 @ f ðx0 Þ
2
(c) Apply the method of Frobenius and find the roots
y ¼ f ðxÞ ¼ f ðx0 Þ þ ðx
x0 Þ þ ðx
x0 Þ2
@x 2! @x2 of the indicial equation to show that c1 ¼ c2 ¼ 0.
1 @ n f ðx0 Þ (d) Complete the solution and show that the first few
þ  þ ðx
x0 Þn
n! @xn terms of the solution are
" #
x 2 x 4  1 2
Functions that can be expressed this way are said to y ¼ a0 1 þ þ þ 
be “regular.” 2 2 2!
( " #
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi x 2 x 4 1
(a) Expand the function ð1 þ xÞ around the point þ b0 lnðxÞ 1 þ þ þ 
x0 ¼ 0 by means of Taylor’s series. 2 2 ð2!Þ2
(b) Use part " #)
(a) to deduce the useful approximate
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi x 2 3 x 4 1
result ð1 þ xÞ
1 þ x=2: What error results
þ þ 
when x ¼ 0.1? 2 2 2 ð2!Þ2
(c) Complete Example 2.24 for the case n ¼ 1/2 by
Taylor expansion of exp(
E/RT) around the point 3.43. The Graetz equation arises in the analysis of heat
T0, retaining the first two terms. transfer to fluids in laminar flow
3.22. Determine the roots of the indicial relationship d 2y dy
for the Frobenius method applied to the following x2 2 þ x þ l2 x2 ð1
x2 Þy ¼ 0
dx dx
equation:
Apply the method of Frobenius and show that the only
d 2 y 1 dy solution that is finite at x ¼ 0 has the first few terms as
(a) x 2 þ þy¼0
dx 2 dx
 2   
l l2 l2 4
d2y dy y ¼ a0 1
x2 þ 1þ x

(b) x 2 þ 2 þ xy ¼ 0 4 16 4
dx dx
d 2y dy 3.53. Use the method of Frobenius to find solutions for the
(c) x2 2 þ x þ ðx2
1Þy ¼ 0 following equations:
dx dx
3.32. Thin, metallic circular fins of thickness b can be d 2y dy
(a) xð1
xÞ
2 þ 2y ¼ 0
attached to cylindrical pipes as heat transfer promot- dx2 dx
ers. The fins are exposed to an ambient temperature d 2y dy
Ta, and the root of each fin contacts the pipe at posi- (b) x 2 þ ð1
2xÞ
y¼0
dx dx
tion r ¼ R1, where the temperature is constant, Tw.
The fin loses heat to ambient air through a transfer 3.61. Villadsen and Michelsen (1978) define the Jacobi
coefficient h. The metallic fin transmits heat by con- polynomial as solutions of the equation
duction in the radial direction. d2y dy
xð1
xÞ þ ½b þ 1
ðb þ a þ 2Þx
(a) Show that the steady-state heat balance on an ele- dx2 dx
mentary annular element of fin yields the equation þ nðn þ a þ b þ 1Þy ¼ 0
   
1d dT 2h Show how this can be obtained from the conventional
r
ðT
T a Þ ¼ 0
r dr dr bk definition of Jacobi’s ODE, Eq. 3.145.
94 SERIES SOLUTION METHODS AND SPECIAL FUNCTIONS

3.71. In the method of orthogonal collocation to be L ¼ height of fin

described later, Villadsen and Michelsen (1978) W ¼ thickness of fin at base
define Legendre’s equation as u ¼ half wedge angle of fin
d 2y dy (a) Find the general solution of the above equation
xð1
xÞ 2
þ ð1
2xÞ þ nðn þ 1Þy ¼ 0
dx dx with the provision that temperature is at least
Use the change of variables u(x) ¼ 2x
1 and show finite at the tip of the fin.
this leads to the usual form of Legendre equation (b) Complete the solution using the condition T ¼ TH
given in Eq. 3.146. at the base of the fin.
3.82. (a) Show that the solution of Problem 3.3 can be con- (c) If a fin is 5 mm wide at the base, 5 mm high, and
veniently represented by the modified Bessel has a total length 71 cm (28 in.), how much heat is
functions transferred to ambient (desert) air at 20 C? Take
the heat transfer coefficient to be 10 Btu/h ft2 F
y ¼ AI 0 ðxÞ þ BK 0 ðxÞ (0.24  10
3 cal/cm2  s  C) and the barrel temper-
ature (as a design basis) to be 400 C. The conduc-
(b) Evaluate the arbitrary constants, using the bound- tivity of steel under these conditions is 0.10 cal/
ary conditions (cm  s   C). From this, given the rate of heat
rffiffiffiffiffiffi
2h generation, the number of fins can be specified.
y ¼ T w
T a @ x ¼ R1 Today, circular fins are used in order to reduce
bk
dy bending stresses caused by longitudinal fins.

k ¼ hy @ r ¼ R2 ðouter fin radiusÞ
dr 3.112. Porous, cylindrical pellets are used in packed beds
for catalytic reactions of the type A
k! s
Products.
3.92. Rodriques’ formula is useful to generate Legendre’s
Intraparticle diffusion controls the reaction rate
polynomials for positive integers n
owing to the tortuous passage of reactant through
the uniform porous structure of the pellet.
1 d n ðx2
1Þn
Pn ðxÞ ¼ n
2 n! dxn (a) Ignoring transport at the end caps, perform a
steady material balance for the diffusion
Show that the first three Legendre polynomials are
reaction of species A to obtain for linear kinetics

P0 ðxÞ ¼ 1; P1 ðxÞ ¼ x; P2 ðxÞ ¼ 12 ð3x2
1Þ  

1d dCA
DA r
k s as C A ¼ 0
r dr dr
and then prove these satisfy Legendre’s equation.
where DA is pore diffusivity of species A, CA is
3.103. A wedge-shaped fin is used to cool machine-gun bar-
molar composition, ks is surface rate constant,
rels. The fin has a triangular cross section and is L
and as represents internal area per unit volume
meters high (from tip to base) and W meters wide at
of pellet.
the base. It loses heat through a constant heat trans-
fer coefficient h to ambient air at temperature TA. (b) If the bulk gas composition is CA0, and the gas
The flat base of the fin sustains a temperature TH. velocity is so slow that a finite film resistance
Show that the temperature variation obeys exists at the particle boundary (where r ¼ R),
then introduce the variables transformation
  rffiffiffiffiffiffiffiffiffi
d 2 y dy 2hL sec u CA k s as
x þ
y¼0 y¼ ; x¼r
dx2 dx kW CA0 DA

where and show the equation in part (a) becomes

d 2y dy
x2 2 þ x
x2 y ¼ 0
x ¼ distance from tip of fin dx dx
y ¼ T
TA
(c) Apply the boundary condition
T ¼ local fin temperature
TA ¼ ambient temperature 
h ¼ heat transfer coefficient dCA 

DA ¼ kc ðC A jr¼R
C A0 Þ
k ¼ thermal conductivity of fin material dr r¼R
 REFERENCES 95

and show 3.13 . Darcy’s law can be used to represent flow pressure
drop through uniform packed beds
 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
CA I 0 r ks as =DA k dp
¼  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi V 0z ¼
ðk is permeabilityÞ
CA0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
DA ks as =kc I 1 R ks as =DA þ I 0 R ks as =DA m dz

where V0z is the superficial velocity along the axial

Note: As k c ! 1, the solution is identical to (z) direction. The Brinkman correction has been
Example 3.3. suggested (Bird et al. 1960), so that pipe wall effects
can be accounted for by the modification
3.123. Thin, circular metal fins are used to augment heat
 
transfer from circular pipes, for example, in home dp m 1d dV 0z
heating units. For such thin fins, heat is conducted 0 ¼

V 0z þ m r
dz k r dr dr
mainly in the radial direction by the metal, losing
heat to the atmosphere by way of a constant heat (a) For a constant applied pressure gradient such
transfer coefficient h. In Problem 3.3, we introduced that
dp=dz ¼ DP=L; show that a modified
the follow variables: Bessel’s equation results if we define

y ¼ T
TA y ¼ V 0z
V 0z ;
rffiffiffiffiffiffi DP k r
x ¼r
2h V 0z ¼ ðDarcy velocityÞ; x ¼ pffiffiffi
bk L m k

hence, obtain
to obtain for circular fins
d2y dy
2 x2 2
þx
x2 y ¼ 0
d y dy dx dx
x2 2
þx
x2 y ¼ 0
dx dx
(b) Use the symmetry condition to show

(a) If the pipe of radius Rp takes a temperature Tp,

DP k  pffiffiffi
and the outer rim of the fin at position R exists at V 0z ðrÞ ¼ þ AI 0 r= k
ambient temperature TA, show that the tempera- L m
ture profile is What is the remaining boundary condition to
find the arbitrary constant A?
T
TA I 0 ðxÞK 0 ðxR Þ
K 0 ðxÞI 0 ðxR Þ
¼
T p
T A I 0 ðxp ÞK 0 ðxR Þ
I 0 ðxR ÞK 0 ðxp Þ
REFERENCES
where
rffiffiffiffiffiffi Abramowitz, M. and I.A. Stegun. Handbook of Mathematical
2h Functions. Dover Publications, Inc., New York (1965).
xR ¼ R
bk Bird, R. B., W. E. Stewart, and E. N. Lightfoot. Transport
rffiffiffiffiffiffi Phenomena. John Wiley & Sons, Inc., New York (1960).
2h Boas, M. L. Mathematical Methods in Physical Science. 2nd ed.,
xp ¼ Rp
bk John Wiley & Sons, Inc., New York (1983), p. 12.
Jenson, V. G. and G. V. Jeffreys. Mathematical Methods in Chemi-
(b) For small arguments, such that xp < x < xR < 1, cal Engineering. 2nd ed., Academic Press, New York (1977).
show that the temperature profile in (a) reduces Mickley, H. S., T. K. Sherwood, and C. E. Reid. Applied Mathe-
to matics in Chemical Engineering. McGraw Hill, New York
(1957).
Villadsen, J. and M. L. Michelsen. Solution of Differential Equa-
T
TA lnðr=RÞ

 tion Models by Polynomial Approximation. Prentice Hall,
T p
T A ln Rp =R Englewood Cliffs, N.J. (1978).