Solutions To Rainville's Special Functions - Tom Cuchta, Leon Hall, Sylvester Pagano, Compiled 2 June 2015

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)
SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA
Contents
0.
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
23.
Preface
Chapter 1 Solutions
Chapter 2 Solutions
Chapter 3 Solutions
Chapter 4 Results Cited
Chapter 4 Solutions
Results from Chapter 5 used
Chapter 5 Solutions
Chapter 6 Solutions
Chapter 7 Solutions
Chapter 8 Solutions
Chapter 9 Solutions
Chapter 10 Solutions
2
4
9
17
23
24
39
40
48
64
67
76
77
98
102
110
122
140
146
153
156
170
171
184
0. Preface
Earl D. Rainville began giving lectures on Special Functions at the University of
Michigan in 1946. The course was well received, and his notes became the basis for
his book, Special Functions, published in 1960. Also in 1946, Sylvester J. Pagano
received his B.S. degree in Electrical Engineering at the Missouri School of Mines
and Metallurgy (MSM). That fall, Pagano was appointed Instructor in Mathematics
at MSM. In 1950, Pagano, now Assistant Professor, spent the summer at Michigan,
where he and Rainville presumably met. In the summers of 1962, 1963, and 1964,
Pagano was again at Michigan, this time as a National Science Foundation Science
Faculty Fellow. Rainville passed away in 1966, the same year Pagano was promoted
to Professor at the University of MissouriRolla (UMR, MSM under its new name).
In the spring of 1966, I was a sophomore at UMR and was taking Elementary Differential Equations; the textbook was the third edition or so of Rainvilles Elementary
Differential Equations. This would be a better story if Pagano had been the instructor in that class, but he wasnt. I did have a class from Pagano later, when
I was a beginning graduate student; it was Operational Calculus, and we used the
Operational Mathematics book by R.V. Churchill. Churchill was Rainvilles Ph.D.
advisor (Michigan 1939), but I knew nothing of any of these connections at the
time. Pagano was a member of both my M.S. and Ph.D. committees at UMR.
Figure 1. Photograph of Earl D. Rainville from the University of

Michigan.
Figure 2. These photographs are of Sylvester J. Pagano as a

college senior in 1946.
Pagano retired in 1986, a year after I had returned to Rolla as a faculty member.
Somehow, in the process of him cleaning out his office, I got his collections of worked
problems from Rainvilles Special Functions, Rainvilles Intermediate Differential
Equations, and Churchills Operational Mathematics. The way I remember it is
that I knew these problem solutions existed, and when Pagano retired, I asked him
if I could have them, a request to which he graciously agreed. As to how I knew he
even had this material, I dont remember for sure. These problem solutions (hand
written) almost certainly go back to the three summers Pagano spent at Michigan
in the 1960s, and some might even date back to his earlier 1950 visit; Pagano
continued to work on them as he taught the courses himself in the 1960s1980s.
Beginning in the mid1990s I began to teach both Operational Calculus and
Special Functions from time to time at UMR. For Operational Calculus, I used
Churchills book, and Paganos problem solutions were quite useful. I developed
my own notes for Special Functions because it was a summer course designed so our
graduate students who taught in the summer would have something to take, and
some of these students had not yet studied complex variables. The last time I offered
Special Functions was the summer of 2012, and I was able to talk Tom Cuchta,
one of the students in that class, into working on transcribing Paganos problem
solutions from Rainvilles Special Functions into electronic form using LATEX. To
my surprise and delight, Cuchta finished the job by the end of 2012! We learned,
however, that Pagano had not provided solutions to all the problems he wrote up
solutions to 196 out of 231 problems in the book. Thats 85%, so is pretty good,
but I decided I would work on finishing the job, and Cuchta agreed to keep adding
to the TEXfile as more problems were completed. As of now (April 2015), there are
less than 10 problems left to finish. The end is in sight.
Sylvester Pagano was a good example of a type of mathematics professor that
seems to be disappearing these days. He didnt publish any papers, but he was an
outstanding teacher and he kept building his knowledge of mathematics throughout

his career. When I talk with alumni from our department, and often with engineering graduates from other departments, the person they most frequently ask about
is Professor Pagano; they fondly remember him as one of the good ones from their
student days. These alums are righthe was one of the good ones, and he should
be remembered. These problem solutions are mostly his, and the rest were inspired
by him. There is a lot of interesting mathematics here. I hope you enjoy it.
Leon M. Hall
Professor Emeritus, Mathematics
Missouri S&T
1. Chapter 1 Solutions
Problem 1. Show that the following product converges and find its value:

Y
6
.
1+
(n + 1)(2n + 9)
n=1
Solution 1. (Solution by Leon Hall) By Theorem 3, page 3, this product converges
X
6
converges absolutely.
absolutely because
(n + 1)(2n + 9)
n=1
1+
6
(n + 1)(2n + 9)
(n + 1)(2n + 9) + 6
(n + 1)(2n + 9)
2n2 + 11n + 15
=
(n + 1)(2n + 9)
(2n + 5)(n + 3)
=
.
(n + 1)(2n + 9)
n
Y
1+
So, if
Pn
k=1
n
Y
6
(k + 1)(2k + 9)
(2k + 5)(k + 3)
(k + 1)(2k + 9)
k=1
[7 9 (2n + 5)][4 5 (n + 3)]
=
[2 3 (n + 1)][11 13 (2n + 9)]
[7 9][(n + 2) (n + 3)]
=
[2 3][(2n + 7) (2n + 9)]
21 (n + 2)(n + 3)
=
2 (2n + 7)(2n + 9)
=
then
21 n2 + 5n + 6
21
=
.
n
n 2 4n2 + 32n + 63
8
Note: The use of Theorem 3 is not needed because finding the value of the infinite
product is sufficient itself to show convergence.
lim Pn = lim
Problem 2. Show that

Y
n=2
1
1 2
n

=
1
.
2
Solution 2. (Solution by Leon Hall)

n2 1
(n + 1)(n 1)
1
=
=
2
n
n2
n2
1
Let
Pn
=
=

n
Y
1
1 2
k
k=2
n
Y
(k + 1)(k 1)
k2
[3 4 (n + 1)][1 2 3 (n 1)]
=
(2 3 4 n)2
(n + 1)
.
=
2n
k=2
n+1
= lim
n 2n
1
=
2

Y
1
=
1 2 .
n
n=2
lim Pn

Y
n=2
1
n

diverges to 0.

n
Y
k1
1 2 3 (k 1)
1
=
=
Pn =
k
234n
n
k=2
Since
1
=0
n
the product diverges to 0. [The product does not converge to 0 because none of the
terms in the product are 0.]
lim Pn = lim
Problem 4. Investigate the product
(1 + z 2 ) in |z| < 1.
n=0
Solution 4. (Solution by Leon Hall) Let Pn =
n
Y
(1 + z 2 ). Then
k=0
P0 = 1 + z,
P1 = (1 + z)(1 + z 2 )
P2 = (1 + z)(1 + z 2 )(1 + z 4 ) = (1 + z)(1 + z 2 + z 4 + z 6 ).

2(2n 1)
Assume Pn = (1 + z)
(z 2 )k =
2n+1
X2
(z 2 )k . Then
k=0
Pn+1
k=0
n+1
= Pn (1 + z 2 )
n+1
= Pn + z 2h Pn
n+1
= (1 + z) 1 + z 2 + + z 2
= (1 + z)
n+1
+ z2
n+1
+ z2
+2
n+2
+ + z2
2n+2
X2
(z 2 )k
k=0
So we have shown by induction that

Pn = (1 + z)
2n+1
X2
z 2k ,
k=0
1
n
, for |z| < 1. |1 + z 2 |
2
1z
Y
1
2n
.
and same process works. Thus,
(1 + z ) converges absolutely to
1
z
n=0
which is a geometric series converging to (z + 1)

1 + |z|2

1
exp
diverges.
n
n=1
n
Y
exp
k=1
Sn = log Pn =

1
and let
n
n
X
1
.
k
k=1
Sn is the nth partial sum of the harmonic series, which diverges. As in the proof
of Theorem 2, page 3, Pn = exp Sn and
lim Pn = lim exp Sn = exp lim Sn .
Thus, because {Sn } diverges, so does {Pn }.

Y
1
exp
diverges to 0.
n
n=1
Solution 6. (Solution by Leon Hall) Let
log Pn = Sn =

n
n
X
X
1
1
=
k
k
k=1
k=1
as in Problem 5. Then
n
X
1
Pn = exp Sn = exp
k
k=1
!
=
1
!.
n
X
1
exp
k
k=1
Because
X
1
diverges to we have
k
k=1
lim Pn = 0
and so
1
exp
n
n=1
diverges to 0.
Problem 7. Test

Y
n=1
z2
n2

.
Solution 7. (Solution by Leon Hall) The product diverges to 0 for any z such
z2
that z = m, m a positive integer. For all z such that 1 2 6= 0, we have by
n

Y
X
z2
z2
Theorem 3, page 3, that
1 2 is absolutely convergent because
2
n
n
n=1
n=1
is absolutely convergent. In fact,

X
z2
z2 2
.
2 =
n
6
n=1

Y
(1)n+1
converges to unity.
1+
n
n=1

n
Y
(1)k+1
1+
.
k
k=1
Case 1: n is even. Then

21
3+1
41
n
n1
1+1
Pn =
1
2
3
4
n1
n
n!
Rearranging, we get Pn =
= 1 for even n.
n!
Case 2: n is odd. Then

1+1
21
3+1
41
n1
n2
n+1
n+1
Pn =
=
.
1
2
3
4
n2
n1
n
n
In both cases lim Pn = 1.
n

Y
1
Problem 9. Test for convergence:
1 p for real p 6= 0.
n
n=2
Solution 9. (Solution by Leon Hall) For p > 1 the series of positive numbers

Y
X
1
1
is known to be convergent (e.g., by the Integral Test). Thus,
1 p
np
n
n=2
n=2
is absolutely convergent by Theorem 3.
1
> 0 and so convergence and absolute convergence are the
np
X
1
diverges for 0 < p 1, our product diverges by
same. Because the series
np
n=2
Theorem 3.
For p < 0, let p = q where q > 0. Then
1
1 p = 1 nq = 1 + (nq ).
n
But lim (nq ) 6= 0, so in this case our product diverges by Theorem 1.
n

Y
1
Summary:
1 p diverges when p 1 (an p 6= 0), and converges when
n
n=2
p > 1.
For 0 < p 1, 1
Y
sin( nz )
z
n
n=1
is absolutely convergent for all finite z with
the usual convention at z = 0.

Solution 10. (Solution by Leon Hall) Let
sin( nz )
z
n
Then
an (z)
= 1 + an (z).
sin( nz )
z
n
1 z2
1 z4
1 z6
=
+
+ ...
2
4
3!n 2 5! n 7!
n6
1
1
z
= 2 +O
.
n
6
n2
Thus, there exists a constant M such that
|an (z)| <
M
,
n2
X
M
and because
converges, the product
n2
n=1
(1 + an (z)) =
n=1
sin( nz )
z
n
converges absolutely and uniformly for all finite z by Theorems 3 and 4.

If z = 0 the product is, with the usual convention,
1 = 1.
n=1
Problem 11. Show that if c is not a negative integer,

z
Y
z
1
exp
c+n
n
n=1
is absolutely convergent for all finite z.
Solution 11. Let

z
z
1 + an (z) = 1
exp
c+n
n

z
1 z3
1
1 z4
1 z2
z2
1 z3
= 1+ +
+
+ ...
+
+ ...
z+
+
2
3
c+n
n 2! n2
3! n3
n 2! n 3! n

1
1
1
1
1
1
2
=1+
z+
z +
z3 + . . .
n c+n
2!n2
n(c + n)
3!n3
2!n2 (c + n)
X
c
cn
c (k 1)n k
=1+
z+ 2
z2 +
z
n(c + n)
2n (c + n)
k!nk (c + n)
k=3

c
1
1
=1+
.
z
z2 + O
n(c + n)
2n(c + n)
n2
Thus, for c not a negative integer and for any finite z, there is a constant M such
M
that |an (z)| 2 and so by Theorems 3 and 4 the product converges absolutely and
n
uniformly.
0 (z)
1 X
Problem 1. Start with ()
=
(z)
z n=1

1
1
, prove that
z+n n
20 (2z) 0 (z) 0 (z + 21 )
= 2 log 2,
(2z)
(z)
(z + 12 )
and thus derive Legendres duplication formula, page 24.
Solution 1. Applying () three times and simplifying yields
20 (2z) 0 (z) 0 (z + 12 )
(2z)
(z)
(z + 12 )
X
X

X
1
1
1
1
2
1
2
2
1
++ ++
+
= 2
2z
z
z+n n
z + 21 n=1 2z + n n
z + 21 + n n
n=1
n=1

2n
n
n
X
X
X
2
2
2
1
1
1
2
=
lim
+ lim
+ lim
n
n
2z + 1 n
2z + k k
z+k k
2z + 1 + 2k k
k=1
k=1
k=1
"
#
2n
n
n
X
X
X
2
2
2
2
=
+ lim
+ 2H2n +
Hn +
Hn
2z + 1 n
2z + k
2z + 2k
2z + 2k + 1
k=1
k=1
"k=1
#
2n
2n+1
X
X
2
2
2
+ lim
+
+ 2H2n 2Hn
=
2z + 1 n
2z + k
2z + k
k=1
k=2
2
2
2
=
+
+ lim
+ lim (2H2n 2Hn )
2z + 1 2z + 1 n 2z + 2n + 1 n
= 0 + 0 + 2 lim [(H2n log 2n) (Hn log n) + log 2n log n]
n
= 2[ + log 2]
= 2 log 2.
Problem 2. Show that 0 ( 21 ) = ( + 2 log 2) .
10 SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA
Solution 2. By Problem 1, we know that

20 (2z) 0 (z) 0 (z + 21 )
= 2 log 2.
(2z)
(z)
(z + 12 )
Now let z =
1
to get
2
0 (1) 0
2
(1)
1
2
1
2
0 (1)
= 2 log 2,
(1)
and so, algebra yields

0 ( 12 )
0 (1)
1 = (1) 2 log 2.
( 2 )
1
0
But (1) = 1, (1) = , ( 2 ) = , hence
0 ( 12 )
= 2 log 2,
1
and by rearrangement,
1
0 ( ) = ( + 2 log 2) .
2
Z
Problem 3. Use Eulers integral form (z) =
et tz1 dt to show that
0
(z + 1) = z(z).
Z
et tz1 dt, for R(z) > 0, integration by parts yields

Z
t
et tz dt
dv = e dt (z + 1) =
0
Z
z t
t
= [t e ]0 + z
v = e
et tz1 dt
Solution 3. From (z) =
u = tz
du = ztz1
= 0 + z(z),
where lim tz et converges for R(z) > 0
t
.
Problem 4. Show that (z) = lim nz B(z, n).

n
Solution 4. From page 28 (1), we know

(n 1)!nz
,
n
(z)n
(z) = lim
but
B(z, n) =
(z)(n)
(z)(n 1)!
(n 1)!
=
=
.
(z + n)
(z)n (z)
(z)n
Hence
(z) = lim nz B(z, n).
n
Problem 5. Derive the following properties of the beta function:

(a) pB(p, q + 1) = qB(p + 1, q);
(b) B(p, q) = B(p + 1, q) + B(p, q + 1);
(c) (p + q)B(p, q + 1) = qB(p, q);
(d) B(p, q)B(p + q, r) = B(q, r)B(q + r, p).
Solution 5. (a) We know B(p, q) =

pB(p, q + 1) =
(p)(q)
, so
(p + q)
(p + 1)q(q)
p(p)(q + 1)
=
= qB(p + 1, q).
(p + q + 1)
(p + 1 + q)
(note: p q and q p is this the symmetric property?)

(b)
B(p, q)
(p)(q)
(p + q)
(p)(q)
=
(p + q + 1)
p+q
(p + q)(p)(q)
=
(p + q + 1)
q(p)(q)
p(p)(q)
+
=
(p + q + 1) (p + q + 1)
(p + 1)(q)
=
(p)(q + 1)
(p + q + 1) +
(p + q + 1)
= B(p + 1, q) + B(p, q + 1).
=
(c)
(p + q)B(p, q + 1)
(p + q)(p))(q + 1)
(p + q + 1
(p + q)(p)(q + 1)
=
(p + q)(p + q)
(p)(q + 1)
=
(p + q)
(p)q(q)
=
(p + q)
= qB(p, q).
(d)
B(p, q)B(p + q, n)
(p)(q) (p + q)(n)
(p + q) (p + q + n)
(p)(q)(n)
=
(p + q + n)
(q)(n) (p)(q + n)
=
(q + n) (p + q + n)
= B(q, n)B(q + n, p).
=
11
Problem 6. Show that for positive integral n, B(p, n + 1) =
n!
.
(p)n+1
Solution 6. For integer n and using Theorem 9 (pg. 23),

B(p, n + 1)
=
=
=
=
=
(p)(n + 1)
(p + n + 1)
(p)(n + 1)
(p + 1)n (p + 1)
(p)n!
(p + 1)n p(p)
n!
p(p + 1)n
n!
.
(p)n+1
(1 + x)p1 (1 x)q1 dx.
Problem 7. Evaluate
1
(1 + x)p1 (1 x)q1 dx. Now let y =
Solution 7. Let A =
1
1+x
, x = 2y
2
1, 1 x = 2 2y = 2(1 y). Hence

Z 1
A =
2p1 y p1 2q1 (1 y)q1 2dy
0
Z 1
= 2p+q1
y p1 (1 y)q1 dy
0
= 2p+q1 B(p, q).
Problem 8. Show that for 0 k n,

(1)k ()n
.
(1 n)k
Note particularly the special case = 1.
()nk =
Solution 8. Consider ()nk for 0 k n. Then

()nk
= ( + 1) . . . ( + n k 1)
( + 1) . . . ( + n k 1)[( + n k)( + n k + 1) . . . ( + n 1)]
=
( + n 1)( + n 2) . . . ( + n k)
()n
=
( + n k)k
()n
=
(1)k (1 n)k
(1)k ()n
=
.
(1 n)k
Note for = 1, that (n k)! =
(1)k n!
.
(n)k
13
Problem 9. Show that if is not an integer,

(1 n)
(1)n
=
.
(1 )
()n
Solution 9. Consider for not equal to an integer
(1 n)
(1 )
as desired.
(1 n)
()
(1 n)
=
()( 1)( 1)
(1 n)
=
()( 1) . . . ( n + 1)(1 n)
1
=
,
(1)n ()n
1
In the following problems, the function P (x) := x bxc .
2
Z x
Problem 10. Evaluate
P (y)dy.
0
P (y)dy when P (y) = y byc
Solution 10. To evaluate

0
1
. Let m be an
2
integer so that m 0. If m x < m + 1, then bxc = m and

Z x
Z x
P (y)dy =
P (y)dy
0

Zmx
1
=
ym
dy
2 #
m
"
2 x
1
1
ym
=
2
2
"
2 m 2 #
1
1
1
xm
=
2
2
2

1
1
=
P 2 (x)
2
4
1 2
1
= P (x) .
2
8
Problem 11. Use integration by parts and the result of the above exercise to show
that
Z

P (x)dx
1

8(1 + n) .
1
+
x
n
Z
P (x)
dx and use integration by parts
Solution 11. Consider
1+x
n
1
1
2
2
Z
Z
1 P (x) 4
P (x)
1 P (x) 4
1
dv = P (x)dx
u = (1 + x)
dx =
dx.
+
1+x
2
1+x
2 n
(1 + x)2
n
v=
1 2
1
P (x)
2
8
du = (1 + x)2 dx

Now max P 2 (x)
Z

1
1
1
= and P 2 (n) = implies

4
4
4
P (x)
1
dx = 0 0 +
1+x
2
P 2 (x)
1
4 dx
(1 + x)2
and
Z

1

Z
P (x) 1
1
1
4
dx
=
dx
1+x
2 n (1 + x)2
8 1+x n
or

P (x) 1
1
dx
.
1+x
8 1+n

Z
Problem 12. With the aid of the above problem, prove the convergence of
0
Z
Solution 12.
0
cise 11,
P (x)
dx converges lim
n
1+x
Z

lim
n
Z
Hence
0
P (x)dx
.
1+x
P (x)
dx = 0 but from Exer1+x

1
P (x)
dx lim
= 0.
N 8(1 + n)
1+x
P (x)
dx < .
1+x

Z
Z
Z
P (x)dx X n+1 P (x)dx X 1 (y 21 )dy
=
=
.
1+x
1+x
1+n+y
0
n=0 n
n=0 0
Then prove that
lim n
and thus conclude that

0
(y 21 )dy
1
=
1+n+y
12
P (x)dx
is convergent.
1+x
Solution 13. (Solution by Leon Hall) Because P (x) is periodic with period 1, it is
clear that
Z
Z n+1
X
P (x)
P (x)
dx =
dx.
1+x
1+x
0
n=0 n
Let x = n + y. Then
Z
n+1
P (x)
dx
1+x
=
Z0 1
=
Z0 1
=
0
P (n + y)
dy
1+n+y
P (y)
dy
1+n+y
y 21
dy.
1+n+y
15
This establishes the first set of equalities.

Z 1
Z 1
y 12
3
1
1 n+
dy =
dy
2 y+n+1
0 y+n+1
0

1
3
= y n+
log(y + n + 1)
2
0

3
n+2
=1 n+
log
2
n+1

3
1
=1 n+
log 1 +
2
n+1

1
1
3
1
1
+
+ ...
=1 n+
2
n + 1 2(n + 1)2
3(n + 1)3
4(n + 1)4
Z 1
X
(y 21 )
To determine the convergence of
dy we compare with the known
y+n+1
n=0 0
X
1
convergent series
using the limit comparison test.
2
n
n=1
Z 1
(y 12 )

dy
1
3 X
0 y+n+1
2
2
=
n
n
n
+
1
2
k(n
+
1)k
n2

k=1

3
1
1
1
1
= n2 n2 n +
+
+
O
2
n + 1 2(n + 1)2
3(n + 1)3
n + 1
3
2
2
3
2
6n (n + 1) n (n + 2 )[6(n + 1) 3(n + 1) + 2]
1
=
+O
6(n + 1)3
n +1

15 2
3
6n5 + 18n4 + 18n3 + 6n2 [6n5 + 18n4 + 37
1
2 n + 2 n ]
+O
=
3
n+1

6(n + 1)
1 3
2n
1
=
+O
.
6(n + 1)3
n+1
Thus,
lim n
and
Z
X
n=0
Z
0
(y 21 )
1
dy = ,
y+n+1
12
(y 21 )
dy =
y+n+1
Z
0
P (x)
dx
1+x
converges.
Problem 14. Apply Theorem 11, page 27, to the function f (x) = (1 + x)1 ; let
n and thus conclude that
Z
1
=
y 2 P (y)dy.
2
1
Solution 14. (Solution by Leon Hall) Let f (x) =
1
. Theorem 11, page 27
1+x
gives with p = 0, q = n,

Z n
Z n
n
X
1
1
1 1
1
=
dx + +
+
f 0 (x)B1 (x)dx.
1+k
2 2 1+n
0 1+x
0
k=0
So,
n
X
k=0
1
log(1 + n)
1+k
1 1
= +
2 2
1
1+n
+
0
B1 (x)
dx
(1 + x)2

Z n+1
1 1
1
B1 (y + 1)
dy
= +
+
2 2 1 + n Z1
y2
n+1
1 1
1
B1 (y)
= +
dy
2 2 1+n
y2
1
y := x + 1
Problem 15. Use the relation (z)(1 z) =

sin x sin y =
and the elementary result

sin z
1
[cos(x y) cos(x + y)]
2
to prove that
1
(2 c)(c 1)(c a b)(a + b + 1 c)

(c)(1 c)(c a b)(a + b + 1 c)
=
.
(c a)(a + 1 c)(c b)(b + 1 c)
(a)(1 a)(b)(1 b)
Solution 15. Note that

1 (c a b) = a + b + 1 c,
1 (c a) = a + 1 c,
and
1 (c b) = b = 1 c,
so we can use the gamma function relation four times to get
1
(c)(1 c)(c a b)(a + b + 1 c)

(c a)(a + 1 c)(c b)(b + 1 c)
2 sin (c a) sin (c b)
2 sin(c) sin (c a b)
sin(c) sin (c a b) sin (c a) sin (c b)
=
.
sin(c) sin (c a b)
=1
Now using the given trig identity, we get, continuing the equality:
+ b) cos (2c a b)] 12 [cos (b a) cos (2c a b)]
1
2 [cos (a + b) cos (2c a b)]
1
2 [cos (b a) cos (a + b)]
= 1
2 [cos (a + b) cos (2c a b)]
sin(a) sin(b)
=
sin(c) sin (c a b)
0 sin(a) sin(b)
=
.
sin (c 1) sin (c a b)
=
1
2 [cos (a
Canceling minus signs and multiplying and dividing by 2 yields

=
as desired.
(c 1)(2 c)(c a b)(a + b + 1 c)

(a)(1 a)(b)(1 b)
17
Problem 1. With the assumptions of Watsons lemma, show, with the aid of the

X
k
convergence of the series F (t) =
an expt
1 in |t| a + , that for 0
r
k=1
t a, there exists a positive constant c1 such that

n

X
k
n+1

ak expt
1 < c1 expt
1 .
F (t)

r
r
k=1
Solution 1. We wish to show that there exists c1 such that for 0 t a (see
problem (2) for t > a),

n

X
n+1
k

1
ak t r < c1 t r 1
F (t)

k=1
under the condition of Watsons lemma. By the convergence of
an t r 1 = F (t)
n=1
in |t| a we write

n

X
k

1
ak t r
F (t)

k=1

X

k

1
ak t r
=

k=n+1

X
kn1
n+1

1
=t r
ak t r

k=n+1

X
kn1
n+1

1
t r
ak a r

k=n+1
< c1 t
n+1
r 1
X

kn1

where c1 >
ak a r . Remember from Watsons lemma

k=n+1
F (t) =
an t r 1
k=1
for |t| a + where > 0.
Problem 2. With the assumptions of Watsons lemma, page 41, show that for
t > a, there exist positive constants c2 and such that

n

X
k
n+1

ak expt
1 < c2 expt
1 et .
F (t)

r
r
k=1
Solution 2. Under the assumption of Watsons lemma we wish to show that for
t > a, there exist constants c2 , such that

n

X
n+1
k

ak t r 1 < c2 et t r 1 .
F (t)

k=1
Now for t > 0, we have given |F (t)| < kebt . Hence

n

n

X

X
kn1
k
n+1

ak t r 1 < kebt + t r 1
ak t r
F (t)

k=1
k=1
n
X

k

= kebt +
ak t r 1 ,

k=1
but k n 1 < 0 and t > a, so

n

n

X

X
kn1
n+1
k

bt
1
1
ak t r < ke + t r
ak a r
F (t)

k=1
k=1

X

k

but
ak a r converges. Hence, since t > a, there exist constants M1 , M2 such

k=1
that

n

a n+1
X
k
n+1
n+1
r 1

1
ak t r < M1 ebt t r 1
+ M2 t r 1
F (t)

t
k=1
n+1
< M1 ebt t r 1 + M2 t
n+1
< c1 et t r 1 .
n+1
r 1
ebt
Problem 3. Derive the asymptotic expansion (6) immediately preceding these exercises by applying Watsons lemma to the function
Z xt
te
0
f (x) =
dt
t + t2
0
and then integrating the resultant expansion term by term.
Solution 3. Consider Watsons lemma with F (t) =
F (t) =
Choose r =
Hence
t
. Hence
1 + t2
X
X
t
n+1 2n+1
=
(1)
t
=
(1)n t2n1 ; 0 t < 1.
1 + t2
n=0
n=1
1
1
t
1
1
, a = , = . Also for t 0, et 1 and
< .
2
2
2
3
1+t
2
|F (t)| < 1 et ; t
1
,
2
satisfying condition (2) of Watsons Lemma, and

F (t) =
(1)n t 1/2 1
n=1
for |t|
1 1
5
+ = .
2 3
6
Hence by Watsons lemma for f (x) =

0
f 0 (x)
n
(1)
X
n=1
x2n
n
1/2
text
dt we have
1 + t2

=
X
(1)n (2n 1)!
.
x2n
n=1
19
Then
Z
f 0 (s)ds
Z
X
n=1
(1)n (2n 1)!

ds.
s2n
After integrating, this becomes

Z
X
(1)n (2n 2)!
.
x2n1
n=1
f 0 (s)ds
()
x
Now
d
dx
ext
dt =
1 + t2
Z
0
text
dt f 0 (x).
1 + t2
Let A label the integral in the middle of the last formula. Hence
Z
f (x) =
0
ext
dt,
1 + t2
which we label as B. Also note that integrals A and B are uniformly convergent. Hence
Z
ext
dt = 0.
lim f (x) =
lim
x
x 1 + t2
0
Therefore
Z

f 0 (s)ds = f (s) = 0 f (x).
x
By (),
0 f (x)
X
(1)n (2n 2)!
; |x| , R(x) > 0.
x2n1
n=1
So
f (x)
ext
dt
1 + t2
0
X (1)n (2n 2)!
x2n1
n=1
X (1)n (2n)!
; |x| , R(x) > 0.

x2n+1
n=0
Z
Problem 4. Establish (6), page 43, directly, first showing that

f (x)
n
X
(1) (2k)!x
k=0
2k1
= (1)
n+1
Z
0
ext t2n+2
dt,
1 + t2
and thus obtain not only (6) but also a bound on the error made in computing with
the series involved.
X
1
=
(1)k t2k we have
1 + t2
Solution 4. (Solution by Leon Hall) Because
k=0
1
1 + t2
=
=
=
n
X
k=0
n
X
k=0
n
X
k=0
So
(1)k t2k +
(1)k t2k
k=n+1
k 2k
(1) t
+ (1)n+1 t2n+2
(1)k t2k
k=0
t2n+2
.
(1)k t2k + (1)n+1
1 + t2
ext
dt
1 + t2
Z0
Z
n
X
=
ext
(1)k t2k dt + (1)n+1
Z
f (x)
n
X
(1)k
k=0
Z
ext t2n+2
dt
1 + t2
ext t2n+2
dt.
1 + t2
ext t2k dt + (1)n+1
k=0
Z
0
Integration by parts give the reduction formula (for x as specified)

Z
Z
2k(2k 1) xt 2k2
xt 2k
e t
dt.
e t dt =
x2
0
0
Z
1
This, plus the fact that
ext dt = , yields
x
0
Z xt 2n+2
n
X
(2k)!
e t
f (x) =
(1)k 2k+1 + (1)n+1
dt.
x
1 + t2
0
k=0
Then

n

X

k (2k)!
(1) 2k+1
f (x)

x
k=0
Z xt 2n+2

e t
=
dt
2
1+t
Z 0
<
|ext |t2n+2 dt
Z0
<
eRe(x)t t2n+2 dt
0
=
In the region |arg x|
|x|
(2n + 2)!
.
[Re(x)]2n+3
, > 0, if Re(x) > N , then |x| >

and as
2
sin

(2n + 2)!
= O |x|2n2 ,
2n+3
[Re(x)]
and so
f (x)
(1)k
n=0
(2k)!
x2k+1
as |x| in the sector |arg x| < , > 0.

2
21
Problem 5. Use integration by parts to establish that for real x ,

Z
X
et t1 dt ex
(1)n n!xn1 .
x
n=0
et
dt as x . Using integration by parts,
t
x
Z

1
f (x) = et
t2 et dt
t
x

x
Z

1 t
1
1
e dt
= ex + 2 et + 2
3
x
t x
t
0
Z
1 t
1 t
1
x 1
=e
2 3e
23
e dt
4
x x2
t
t
x
x
Z
Solution 5. Consider f (x) =

u=
1
t
u=
u=
dv = et dt
1
t2
v = et
1
t2
du = 2
dv = et dt
1
t3
v = et
= ....
This pattern can clearly continue forever, so we can write

Z
n
X
(1)k k!
n+1
f (x) = ex
+
(1)
(n
+
1)!
t(n+2) et dt.
xk+1
x
k=0
Now since 0 < x < t,

Z t
n

X
(1)k k!
e
x
x
dt
= (n + 1)!e
e f (x)
n+2

xk+1
t
k=0
Zx
(n 1)! x
<
e
et dt
xn+2
x
(n + 1)!
.
<
xn+2
Hence

n
X
1
(1)k k!
1
ex f (x)
=
O
=
O
,
xk+1
xn+2
xn+1
k=0
so
ex f (x)
X
(1)n n!
k=0
xn+1
or
Z
t1 et dt ex
X
(1)n n!
as x .
xn+1
n=0
Problem 6. Let the Hermite polynomials Hn (x) be defined by

exp(2xt t2 ) =
X
Hn (x)tn
n!
n=0
for all x and t, as in Chapter 11. Also let the complementary error function erfc x
be defined by
Z
2
erfc x = 1 erf x =
exp( 2 )d.
x
Apply Watsons lemma to the function F (t) = exp(2xt t2 ); obtain

2 Z
X
1
exp x s
exp( 2 )d
Hn (x)sn1 , s ,
1
2
sx
2
n=0
and thus arrive at the result
"
2 #
X

1 1
1 1
1 1
exp
t
t x
t x
erfc
Hn (x)tn , t 0+ .
2
2
2
n=0
Solution 6. Because expt (n 1) = tn1 , and if
X
Hn (x) n
F (t) = exp(2xt t2 ) =
t ,
n!
n=0
we can write
F (t) =
X
Hn1 (x)
expt (n 1)
(n 1)!
n=1
so the first condition in Watsons Lemma is satisfied for any fixed x with r = 1,
a = 1, and any > 0. Also,
2
F (t) = exp(2xt t2 ) = et e2xt < 2e2xt

for t 0, so the second condition in Watsons Lemma is satisfied with k = 2 (or
anything > 1) and b = 2x. Thus, we get
Z
X
Hn1 (x)(n)
est F (t)dt
(n 1)!sn
0
n=1
as s , or
Z

X
1
2
Hn (x)sn1 ,
exp t + 2 x s t dt
2
n=0
as s . Completing the square in the exponent leads to

"
Z
2 #

X
1
1
exp t + s x
dt
Hn (x)sn1
exp x s
2
2
0
n=0
1
as s , and if we let = t + s x we get
2
"
2 # Z
X
2
1
exp x s
e d
Hn (x)sn1 ,
1
2
sx
2
n=0
as s .
1
Using the definition of erf c, and then making the substitution t =
(not the
s
inverse Laplace transform) we get
"
2 #

X
1
1
exp x s
erfc
sx
Hn (x)sn2 ,
2
2
2
n=0
as s or
exp
2t
as t 0+ as desired.
"
1
x
2t
2 #

erfc
1
x
2t
X
n=0
Hn (x)tn ,
23
Problem 7. Use integration by parts to show that if R() > 0, and if x is real,
Z
X
(1)n ()n
et t dt x1 ex
, x ,
xn+1
x
n=0
of which Problem 5 is the special case = 1.
Solution 7. Integration by parts with u = t and dv = et dt yields
Z
Z
t
x
e t dt = e x
et t(+1) dt.
x
The same integration by parts with replaced by + 1 applied to the last integral
gives
Z
Z
h
i
et t dt = ex x 1
+ ( + 1)
et t(+2) dt.
x
x
x
Continuing, after n + 1 integrations by parts, we have
Z
Z
n
X
(1)k ()k
t
x +1
n+1
e t dt = e x
+ (1)
()n+1
et t(+n+1) dt
xk+1
x
x
k=0
or
x 1
e x
e t
x
dt
n
X
(1)k ()k
k=0
xk+1
x 1
=e x
(1) ()n+1
et t(+n+1) dt.
Thus,
the desired asymptotic series if the right side of the last equation is
we have
1
+ 1 as x . This is true because
O
xn

Z
x 1

t (+n+1)
n
e x
(1)
()
e
t
dt
n+1

x
x 1

Z
e x
()n+1 t

e
dt

+n+1
x x
x
e ()n+1 x
=
e
xn+2
|()n+1 |
=
xn+2
1
=O
xn+1
as x .
4. Chapter 4 Results Cited
By definition,
F (a, b; c; 1)
X
(a)n (b)n
.
(c)n n!
n=0
Theorem 1. If R(c a b) > 0 and if c is neither zero nor a negative integer,

F (a, b; c; 1) =
(c)(c a b)
.
(c a)(c b)
pg.66, (2):
a a+1
1
,
;
a,
a
b
+
;
2
4z
x2
(1 + z)a F
=F
1
(1 + z)2
b
+
;
1 + a b;
2

z
< 1,
Theorem 2. If |z| < 1 and
1 z
a, c b;
a, b;
z
z = (1 z)a F
F
.
1z
c;
c;
Theorem 3. If |z| < 1,

F (a, b; c; z) + (1 z)cab F (c a, c b; c; z).
1
Theorem 4. If 2b is neither zero nor a negative integer and if |y| <
and
2

y

1 y < 1,
a a+1
,
;
2
2
a, b;
2
y
=F
2y .
(1 y)a F
2
(1
y)
2b;
1
b+ ;
2
1
Theorem 5. Kf a + b + is neither zero nor a negative integer, and if |x| < 1 and
2
|4x(1 x)| < 1,
a, b;
4x(1 x) .
F
1
a+b+
2
Theorem 6. If c is neither zero nor a
|4x(1 x)| < 1,
a, 1 a;
x = (1 z)c1 F
F
c;
negative integer, and if both |x| < 1 and

ca c+a1
,
;
2
2
4x(1 x) .
c;
a, b;
d
F
dx
c;
a + 1, b + 1;
ab
x = F
c
c + 1;
x .
25
Solution 1. From
F (a, b; c; 1)
X
(a)n (b)n
,
(c)n n!
n=0
we get
d
F (a, b; c; x)
dx
=
=
X
(a)n (b)n xn1
(c)n (c 1)!
n=1
X
(a)n+1 (b)n+1 xn
(c)n+1 n!
X
(a + 1)n (b + 1)n xn
ab
=
c n=0
(c + 1)n n!
ab
= F (a + 1, b + 1; c + 1; x).
c
n=0

2a, 2b;
1
a+b+ ;
2

1
1
a+b+
1
2
2

.
2 = 1
1
1
1
1
c+ a
c a+
2
2
2
2
2
2a, 2b;
Solution 2. Wish to evaluate F
1
a+b+
2
2 . From Theorem 5,
1
a+b+ ;
2
2a, 2b;
x =F
a, b;
1
a+b+
2
4x(1 x)
1
1
for |x| < 1, |4x(1x)| < 1. We need to use x = , but note that R(a+b+ ab) >
2
2
0. Hence, by Theorem 1,
2a, 2b;
a+b+
1
2
2 =F
a, b;
a+b+
1
2

1
1
a+b+
2
2
1 =

.
1
1
b+
a+
2
2
a, 1 a;
F
c;

1
2 (c)
2
1

=
1
1
ca+1
2
c+ a
2
2
2
1c
a, 1 a;
Solution 3. Consider F
c;
c1 c+a1
,
;
2
x = (1x)c1 F 2
4x(1 x) .
c;
By Theorem 6, for |x| < 1, |4x(1 x)| < 1,
ca c+a1
a, 1 a;
,
;
2
x = (1 x)c1 F 2
F
4x(1 x) .
c;
c;

c a c a 1
Since R c + +
> 0, we may use Theorem 1 to conclude that
2 2 2 2 2
c1 c+a1
a, 1 a;
c1
,
;
1
1
2
F
F 2
=
1
2
2
c;
c;

1
(c)
2

,
=
c
+
a
ca+1
2c1
2
2
as desired.
Problem 4. Obtain the result
n, b;
F
c;
(c b)n
1 =
.
(c)n
Solution 4. Consider F (n, b; c; 1). At once, if R(c b) > 0,

F (n, b; c; 1) =
(c)(c b + n)
(c b)n
=
.
(c + n)(c b)
(c)n
Actually the condition R(c b) > 0 is not necessary because of the termination of
the series involved.
n, a + n;
F
c;
(1)n (1 + a c)n
1 =
.
(c)n
Solution 5.
n, a + n;
1 =
F
c;
(c)(c a)
.
(c + n)(c a n)
By Exercise 9, Chapter 2, if (c a) is nonintegral,

(1 n)
(1)n
=
.
(1 )
()n
Hence,
n, a + n;
F
a;
(1)n (1 c + a)n
1 =
,
(c)n
27
as desired.
n1 b n;
F
a;
Solution 6.
n, 1 b n;
F
a;
1 =
1 =
(a + b 1)2n
.
(a)n (a + b 1)n
(a)(a 1 + b + 2n)
(a 1 + b)2n
=
.
(a + n)(a 1 + b + n)
(a)n (a 1 + b)n
Of course a 6= nonpositive integer, as usual.

Problem 7. Prove that if gn = F (n, ; 1 + n; 1) and is not an integer, then
gn = 0 for n 1, g0 = 1.
Solution 7. Let gn = F (n, ; 1 + n; 1). Then
gn =
n
X
k=0
X n!()k ()k
(n)k ()k
=
.
k!(1 + n)k
n!(n k)!()n
k=0
Hence, compute the series
X
()n gn tn
n!
n=0
n
X
X
()k ()nk tn
k!(n k)!
!
!
X ()n tn
()n tn
=
n!
n!
n=0
n=0
= (1 t) (1 t)
= 1.
n=0 k=0
Therefore, g0 = 1 and gn = 0 for n 1. (Note: easiest to choose 6= integer, can

actually do better than that probably.)

dn a1+n
x
F (a, b; c; x) = (a)x xa1 F (a + n, b; c; x).
n
dx
Solution 8. Consider Dn [xa1+n F (a, b; c; x)] (D
Dn [xa1+n F (a, b; c; x)]
= Dn
=
=
X
k=0
X
k=0
d
). We have
dx
X
(a)k (b)k xn+k+a1
k=0
(c)k k!
(a)k (n + k + a 1)(n + k + a 2) . . . (k + a)xk+1a (b)k

(c)k k!
(a)k (a)n+k xk+a1 (b)k
(c)k (a)k k!
(a + k)n (a)n xk+a1 (b)k

k!(c)k
k=0
a1
= (a)n x
F (a + n, b; c; x).
=
Problem 9. Use equation (2), page 66, with z = x, b = n, in which n is a

non-negative integer, to conclude that
1 1
1
a,
a
+
;
n, a;
2 2
2
4x
x = (1 x)a F
F
.
(1 x)2
1 + a + n;
1 + a + n;
Solution 9. From (2) on page (66) we get
a a+1
,
;
2
2
a
4z
(1 + z) F
(1 + z)2
1 + a b;
Use z = x, b = n to arrive at
a a+1
,
;
2
2
a
(1 x) F
1 + a + n;
a, b;
=F
1
+
a
b;

.
4x
(1 x)2
a, n;
=F
1 + a + n;
x ,
as desired.
1
Problem 10. In Theorem 23, page 65, put b = , a = + , 4x(1 + x)2 = z and
2
thus prove that
1
, + ;
2
F
21

1
2
.
z = (1 z) 1 + 1 z
Solution 10. Theorem 4 gives us
a, b;
4x
(1 + x)2a F
=F
(1 + x)2
ab;
Put b = , a = +
1
and
2
1
a, a b + ;
1
b+ ;
2
4x
= z.
(1 + x)2
Then
zx2 + 2(z 2)x + z = 0
p
zx = 2 3 z 2 4z + 4 32 = 2 z 2 1 z.
Now x = 0 when z = 0, so
zx = 2 z 2 1 z = 1 z + 1 2 1 z
or
x=
(1
1 z)2
(1 1 z[1 (1 z)]
=
.
z
z(1 + 1 z)
x2
.
Thus
29
1 1z
x=
1+ 1z
and
1+x=
.
1+ 1z
Then we obtain
4x
4(1 1 z) (1 + 1 z)2
= z,
(1 + x)2
4
1+ 1z
1
a check. Now with b = , a = + , Theorem 4 yields
2
1
21

+
,
;
2
F
=F
z
1+ 1z
2;
1
+ , 1;
x2
1
+ ;
2
1;
x2
=1F0
;
= (1 x2 )1 .
2
2 1z
and 1 + x =
,
Since 1 x =
1+ 1z
1+ 1z
4 1z
2
(1 x ) =
.
(1 + 1 z)2
Thus we have
1
, + ;
2
F
2+1
2
2
2
1
=
(1 z) 2
1+ 1z
1+ 1z

21
2
21
= (1 z)
,
1+ 1z

2;
as defined. Now we use Theorem 3 to see that
1
1
, ;
, + ;
1
2
2
2
F
2 = (1 z) F
2;
2;
so that we also get

1
, ;
2
F
2;
as desired.
z =
1+ 1z
21
,
Problem 11. Use Theorem 6 to show that

1
1 1
1
1
c a, c a + ;
2
2
2
2
2
x = (12x)ac F
c;
a, 1 a;
(1x)1c F
c;
Solution 11. By Theorem 6,
a, 1 a;
x
(1 x)1c F
a;
ca c+a1
,
;
2
2
=F
c;
ca c+a1
,
;
2
=F 2
4x(x 1
.
(1 2x)2
4x(1 x)
1 (1 2x)2
c;
ca
ca+1
;
a c 2 ,c
2
2
= (1 2x) 2 F
1 + (1 2x)2
(1 2x)2
c;
ca ca+1
,
;
2
2
ac
= (1 2x) F
c;
4x(x 1)
,
(1 2x)2
which we wished to obtain.

Problem 12. In the differential equation (3), page 54, for
w = F (a, b; c; z)
introduce a new dependent variable u by w = (1 z)a u, thus obtaining
z(1 z)2 u00 + (1 z)[c + (a b 1)z]u0 + a(c b)u = 0.
z
. Show that the
Next change the independent variable to x by putting x =
1z
equation for u in terms of x is
d2 u
du
+ [c (a + c b + 1)x]
a(c b)u = 0,
dx2
dx
and thus derive the solution
a, c b;
z
w = (1 z)a F
.
1z
c;
x(1 x)
Solution 12. We know that w = F (a, b; c; z) is a solution of the equation

(1)z(1 z)w00 + [c (a + b + 1)z]w0 abw = 0.
In (1) put w = (1 z)a u. Then
w0 = (1 z)a u0 + a(1 z)a1 u,
w00 = (1 z)a u00 + 2a(1 z)a1 u0 + a(a + 1)(1 z)a2 u.
31
Hence the new equation is

z(1z)u00 +2azu0 +a(a+1)z(1z)1 u+cu0 +ca(1z)1 u(a+b+1)zu0 a(a+b+1)z(1z)1 uabu = 0,
or
z(1z)u00 +[c(ba+1)z]u0 +(1z)1 [(a2 +a)z+ca(a2 +ab+a)zab(1z)]u = 0,
or
(2)z(1 z)2 u00 + (1 z)[c + (a b 1)z]u0 + a(c b)u = 0.
z
x
1
. Then z =
,1 z =
, and we use equation (12)
1z
1x
1x
dx
1
= (1 x)2 :
on page 12 of IDE for the change of variable. First,
=
dz
(1 z)2
d2 x
2
=
= 2(1 x)3 . The old equation (2) above may be written
dz 2
(1 z)3

d2 u
c
a b 1 du
a(c b)
+
+
+
u = 0,
dz 2
z(1 z)
1z
dt
z(1 z)2
Now put x =
which then leads to the new equation

c(1 x)2
du a(c b)(1 x)3
d2 u
+ (a b 1)(1 x)
u = 0,
(1x)4 2 + 2(1 x)3 (1 x)2
dx
x
dx
x
or
x(1 x)
d2 u
du
+ [2x {c(1 x) + (a b 1)x}]
a(c b)u = 0,
dx2
dx
or
d2 u
du
a(c b)u = 0.
+ [x (a b + c + 1)x]
2
dx
dx
Now (3) is a hypergeometric equation with parameters = c, + beta + 1 =
a b + c + 1, = a(c b). Hence = a, = c b, = c. One solution of (3) is
(3)x(1 x)
u = F (a, c b; c; x),
so one solution of equation (1) is
a, c b;
W = (1 z)a F
z
.
1z
c;
Problem 13. Use the result of Exercise 12 and the method of Section 40 to prove
Theorem 2.

z
< 1,
Solution 13. We know that in the region in common to |z| < 1 and
1 z
there is a relation
a, c b;
z
1c
(1z)a F
= AF (a, b; c; z)+Bz F (a+1c, b+1c; zc; z).
1z
c;
Since c is neither zero nor a negative integer, the last term is not analytic at z = 0.
Hence B = 0. Then use z = 0 to obtain 1 1 = A 1, so A = 1. Hence
a, c b;
z
F (a; b; c; z) = (1 z)a F
.
1z
c;
Problem 14. Prove Theorem 3 by the method suggested by Exercises 12 and 13.
Solution 14. (Solution by Leon Hall) Note that the first two parameters in F (a, b; c; z)
are interchangeable, so results involving one of them also apply to the other. By
Exercise 12,

z
F (a, b; c; z) = (1 z)a F a, c b; c;
.
1z
z
Let w =
so this becomes
1z
F (a, b; c; z) = (1 z)a F (a, c b; c; w).
Again, by Exercise 12, applied to the second parameter,

w
a
(cb)
F (a, b; c; z) = (1 z) (1 w)
F c a, c b; c;
.
1w
w
But 1 w = (1 z)1 , and
= z, so
1w
F (a, b; c; z)
= (1 z)a ((1 z)1 )(cb) F (c a, c b; c, z)

= (1 z)cab F (c a, c b; c; z)
as desired.
Problem 15. Use the method of Section 39 to prove that if both |z| < 1 and
|1 z| < 1, and if a, b, c are suitably restricted,
a, b;
a, b;
(c)(c
b)
1z
z =
F
F
(c a)(c b)
1 + b + 1
c;
c;
c a, c b;
cab
(c)(a + b c)(1 z)
1 z .
+
F
(a)(b)
c a b + 1;
Solution 15. (Solution by Leon Hall) We denote the hypergeometric differential
equation:
z(1 z)w00 (z) + [c (a + b + 1)z]w0 (z) abw(z) = 0
by HGDE. If we make the change of variable z = 1 y, then HGDE becomes
y(1 y)w00 (y) + [c (a + b + 1)y]w0 (y) abw(y) = 0
where c = a + b + 1 c. Thus, two linearly independent solutions are
F (a, b; c; y)
and
y 1c F (a + 1 c, b + 1 c; 2 c; y).
These solutions as function of z are valid in |1 z| < 1 and are
F (a, b; a + b + 1 c; 1 z)
33
and
(1 z)cab F (c a, c b; c a b + 1; 1 z).
Thus, in the region D where both |z| < 1 and |1 z| < 1,
F (a, b; c; z) = AF (a, b; a+b+1c; 1z)+B(1z)cab F (ca, cb; cab+1; 1z)
for some constants A and B.
Assume Re(c a b) > 0 and c 6= 0 or a negative integer and let z 1 inside
the region D to get
F (a, b; c; 1) = A 1 + B 0.
Thus, by Theorem 18, page 49, we get
A=
(c)(c a b)
.
(c a)(c b)
Now, let z 0 inside the region D and assume Re(1c) > 0 and neither a+b+1c
nor c a b + 1 is zero or a negative integer. Then
1 = AF (a, b; a + b + 1 c; 1) + BF (c a, c b; c a b + 1; 1)
and we get
1 AF (a, b; a + b + 1 c; 1)
.
F (c a, c b; c a b + 1; 1)
Again using Theorem 18, this becomes
B=
1
B=
(c)(cab)(a+b+1c)(1c)
(ca)(cb)(b+1c)(a+1c)
(cab+1)(1c)
(1b)(1a)
By Exercise 15, page 32 the numerator is equal to

(2 c)|Gamma(c 1)(c a b)(a + b + 1 c)
.
(a)(1 a)(b)(1 b)
Hence,
B
=
=
(2 c)(c 1)(c a b)(a + b + 1 c)

(a)(b)(c a b + 1)(1 c)
(1 c)(1 c) (c)
c1 (a + b c)(a + b c)
(a)(b)(c a b)(c a b)(1 c)

(c)(a + b c)
=
.
(a)(b)
This yields the desired formula for F (a, b; c; z) in terms of the given hypergeometric
functions of 1 z.
Problem 16. In a common notation for the Laplace transform
Z
L{F (t)} =
est F (t)dt = f (s); L1 {f (s)} = F (t).
0
Show that
L1
1
s
a, b;
F
s + 1;
a, b;
z =F
1;
z(1 et ) .
Solution 16. Let A = L 1
1
s
z . We wish to evaluate A. Now
a, b;
F
1 + s;

X
(a)n (b)n z n 1
1
A=
L
.
n!
s(1 + s)n
n=0
But
1
s(s + 1)n
(1 + s)
s(1 + s + n)
1 (1 + s)(1 + n)
=
sn! (1
+ s + n)(1)
n, s;
1
1
=
F
n!s
1 + s;
n
X
(n)
k (s)k
1
= n!s
.
k!(1 + s)k
=
k=0
Hence
n
1
1 X (n)k
.
=
s(s + 1)n
n!
k!(s + k)
k=0
Then
L 1
1
s(s + 1)n

=
1
1 X (n)k ekt
= (1 et )n .
n!
k!
n!
k=0
Therefore
a, b;
n
t n
X
(a)n (b)n z (1 e )
A=
=F
n!n!
n=0
1;
z(1 et ) .
There are many other ways of doing Exercise 16. Probably the easiest, but most
undesirable, is to work from right to left in the result to be proved. It is hard to see
any chance for discovering the relation that way.
Problem 17. With that notation of Exercise 16 show that
n 3+n
1+ ,
;
2
2
a(n
+
2)
a2
L{tn sin at} =
F
2
n+2
s
s
3
;
2
Solution 17. We wish to obtain the Laplace Transform of tn sin at. Now
n
t sin at =
X
(1)k a2k+1 tn+2k+1
k=0
(2k + 1)!
and
L {tm } =
(m + 1)
.
sm+1
Hence
L {tn sin at}
X
(1)k a2k+1 (n + 2k + 2)
(2k + 1)!sn+2k+2
a(n + 2) X (1)k (n + 2)2k a2k
k=0
sn+2
(2)2k s2k

n+2
n+3
k
(1)
a2k
2
2
a(n + 2) X
k
k

=
3
sn+2
2k
k=0
k!
s
2 k
n+2 n+3
,
;
2
2
a(n + 2)
a2
.
F
=
sn+2
s2
3
;
2
k=0
Problem 18. Obtain the results

log(1 + x) = xF (1, 1; 2; x),

1 1 3 2
arcsin x = xF
, ; ;x ,
2 2 2

1
3
arctan x = xF
, 1; ; x2 .
2
2
Solution 18. Using
(1)n
n!
=
we know that
(n + 1)!
(2)n
log(1 + x)
X
(1)n xn+1
n+1
n=0
X (1)n (1)n (1)n xn

=x
(2)n n!
n=0
= xF (1, 1; 2; x).
Next, start with

2 12
(1 y )
X
n=0
using
y 2n
n!
1
2
1
=
2n + 1
n+
1
2 n

3
2 n
1
2 n
1
2
=
x
to get
2 21
(1 y )
dy =
X
n=0
Thus we arrive at
arcsin x =
X
n=0
= xF
1
2 n
1
2 n
x2n+1
.
n!(2n + 1)
1
2n+1
2 nx

3
2 n n!
1 1 3 2
, ; ;x .
2 2 2
35
Finally form
(1 + y 2 )1 =
(1)n y 2n
n=0
to obtain
Z
x
2 1
(1 + y )
0
X
X
(1)n
(1)n x2n+1
dy =
=
2n + 1
n=0
n=0
or

arctan x = xF
1
2 n

3
2 n
x2n+1
(1)n
n!

1
3
2
, 1; ; x .
2
2
Problem 19. The complete elliptic integral of the first kind is

Z 2
d
p
.
K=
0
1 k 2 sin2

1 1
Show that K = F
, ; 1; k 2 .
2
2 2
Z 2
d
p
Solution 19. From K =
we obtain
0
1 k 2 sin2
2n 2n
Z 2 X
1
sin d
2 nk
.
K=
n!
0 n=0
But
Z
2n
sin
0
Hence

1
2 12 12 n
n + 21 12
1 1
2 n
1
=
=
=
.
d = B n + ,
2
2 2
2(n + 1)
2n!
2 n!
X
K=
2 n=0
1
2 n
1
2 n
k 2n
n!n!
= F
2

1 1
2
, ; 1; k .
2 2
Problem 20. The complete elliptic integral of the second kind is

Z 2 p
1 k 2 sin2 d.
E=
0

1 1
2
, ; 1; k .
2 2
Z 2 p
Solution 20. From E =
1 k 2 sin2 d, we get
Show that E = F
2
X
12
k 2n sin2n d
n!
0 n=0
1 2n
1
X
2 n 2 n k
=
.
2 n=0
n!n!
Z
Hence
E=
F
2

1 1
, ; 1; k 2 .
2 2
Problem 21. From the contiguous function relations 1-5 obtain the relations 6-10.
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)
(15)
37
(a b)F = aF (a+) bF (b+),

(a c + 1)F = aF (a+) (c 1)F (c),
[a + (b c)z]F = a(1 z)F (a+) c1 (c a)(c b)zF (c+),
(1 z)F = F (a) c1 (c b)zF (c+),
(1 z)F = F (b) c1 (c a)zF (c+),
[2a c + (b a)z]F = a(1 z)F (a+) (c a)F (a),
(a + b c)F = a(1 z)F (a+) (c b)F (b),
(c a b)F = (c a)F (a) b(1 z)F (b+),
(b a)(1 z)F = (c a)F (a) b(1 z)F (b+),
[1 z + (c b 1)z]F = (c a)F (a) (c 1)(1 z)F (c),
[2b c + (a b)z]F = b(1 z)F (b+) (c b)F (b),
[b + (a c)z]F = b(1 z)F (b+) c1 (c a)(c b)zF (c+),
(b c + 1)F = bF (b+) (c 1)F (c),
[1 b + (c a 1)z]F = (c b)F (b) (c 1)(1 z)F (c),
[c 1 + (a + b + 1 2c)z]F = (c 1)(1 z)F (c) c1 (c a)(c b)zF (c+).
Solution 21. From (3) and (4) we get

[a + (b c)z (c a)(1 z)]F = a(1 z)F (a+) (c a)F (a),
or
[2a c + (b a)z]F = a(1 z)F (a+) (c a)F (a).
(6)
From (3) and (5) we get

[a + (b c)z (c b)(1 z)]F = a(1 z)F (a+) (c b)F (b),
or
(7)
[a + b c]F = a(1 z)F (a+) (c b)F (b).

[(a b)(1 z) 2a + c (b a)z]F = (c a)F (a) b(1 z)F (b+),
or
(8)
[c a b]F = (c a)F (a) b(1 z)F (b+).

(9)
(b a)(1 z)F = (c a)F (a) (c b)F (b).
Use (2) and (6) to obtain

[(a c + 1)(1 z) 2a + c (b a)z]F = (c a)F (a) (c 1)(1 z)F (c),
or
[1 a + (c b 1)z]F = (c a)F (a) (c 1)(1 z)F (c).
(10)

[a + b c (a b)(1 z)]F = b(1 z)F (b+) (c b)F (b),
or
(11)
[2b c + (1 b)z]F = b(1 z)F (b+) (c b)F (b),
which checks with (6). Easier method: in (6) interchange a and b. From (1) and
(3) we get
[a + (b c)z (a b)(1 z)]F = b(1 z)F (b+) c1 (c a)(c b)zF (c+),
or
(12)
[b + (a c)z]F = b(1 z)F (b+) c1 (c a)(c b)zF (c+),
more easily found by changing b to a and a to b in (3). In (2) interchange a and b

to get
(13)
(b c + 1)F = bF (b+) (c 1)F (c).
In (10) interchange a and b to get
(14)
[1 b + (c a 1)z]F = (c b)F (b) (c 1)(1 z)F (c).

[a + (b c)z (a c + 1)(1 z)]F = (c 1)(1 z)F (c) c1 (c a)(c b)zF (c+),
or
(15)
[c1+(a+b2c+1)z]F = (c1)(1z)F (c)c1 (ca)(cb)zF (c+).
Problem 22. The notation used in Exercise 21 and in Section 33 is often extended
as in the examples
F (a, b+) = F (a 1, b + 1; c; z),
F (b+, c+) = F (a, b + 1; c + 1; z).
Use the relations (4) and (5) of Exercise 21 to obtain
F (a) F (b) + c1 (b a)zF (c+) = 0
and from it, by changing b to (b + 1) to obtain the relation
(c 1 b)F = (c a)F (a, b+) + (a 1 b)(1 z)F (b+),
or
(c 1 b)F (a, b; c; z) = (c a)F (a 1, b + 1; c; z) + (a 1 b)(1 z)F (a, b + 1; c; z),
another relation we wish to use in Chapter 16.
Solution 22. From Exercise 21 equation (4) and (5) we get
(1)(1 z)F = F (a) c1 (c b)zF (c+),
(2)(1 z)F = F (b) c1 (c a)zF (c+).
From the above we get
F (a) F (b) + c1 (b a)zF (c+) = 0.
Now replace b by b + 1 to write
F (a, b+) F + c1 (b + 1 a)zF (b+, c+) = 0,
or
F (a, b; c; z) = F (a 1, b + 1; c; z) + c1 (b + 1 a)zF (a, b + 1; c + 1; z).
Problem 23. In equation (9) of Exercise 21 shift b to b + 1 to obtain the relation
(c 1 b)F = (c a)F (c, b+) + (a 1 b)(1 z)F (b+),
or
(c 1 b)F (a, b; c; z) = (c a)F (a 1, b + 1; c; z) + (a 1 b)(1 z)F (a, b + 1; c; z),
another relation we wish to use in Chapter 16.
39
Solution 23. Equation (9) of Exercise 21 is

(b a)(1 z)F = (c a)F (a) (c b)F (b)
from which we may write
(b + 1 a)(1 z)F (b+) = (c a)F (a, b+) (c b 1)F,
or
(c b 1)F (a, b; c; z) = (c a)F (a, b + 1; c; z) + (a 1 b)(1 z)F (a, b + 1; c; z).
6. Results from Chapter 5 used
Theorem 7. If z is nonintegral,
.
sin z
Theorem 8. (Dixons Theorem) The follow is an identtiy of a, b, and c are so
restricted that each of the functions involved exists:
(z)(1 z) =
3 F2
a, b, c;
1 =
1 + a b, 1 + a c;
(1 + 12 a)(1 + a b)(1 + a c)(1 + 21 a b c)

.
(1 + a)(1 + 12 a b)(1 + 21 a c)(1 + a b c)
Theorem 9. If R() > 0, R() > 0, and if k and s are positive integers, then
insdie the region of convergence of the resultant series
Z t
a1 , . . . , ap ;
cxk (t x)s
x1 (t x)1 p Fq
0
b1 , . . . , b q ;
+1
+k1 +1
+s1
, ,
,...,
;
a1 , . . . , a p , k , k , . . . ,
k
s
s
s
= B(, )+1 p+k+s Fq+k+s
+ ++1
++k+s1
b1 , . . . , bq ,
,
,...,
;
k+s
k+s
k+s
Theorem 10. If z is nonintegral,
.
(z)(1 z) =
sin z
Theorem 11. (Dixons Theorem) The follow is an identtiy of a, b, and c are so
restricted that each of the functions involved exists:
a, b, c;
(1 + 21 a)(1 + a b)(1 + a c)(1 + 21 a b c)
1 =
.
3 F2
(1 + a)(1 + 12 a b)(1 + 21 a c)(1 + a b c)
1 + a b, 1 + a c;
Theorem 12. If R() > 0, R() > 0, and if k and s are positive integers, then
insdie the region of convergence of the resultant series
Z t
a1 , . . . , ap ;
cxk (t x)s
x1 (t x)1 p Fq
0
b1 , . . . , b q ;
k k ss ctk+s
(k + s)k+s
+1
+k1 +1
+s1
, ,
,...,
;
a1 , . . . , a p , k , k , . . . ,
k
s
s
s
= B(, )+1 p+k+s Fq+k+s
+ ++1
++k+s1
b1 , . . . , bq ,
,
,...,
;
k+s
k+s
k+s
;
;
x
F
F
0 1
0 1
a;
b;
1 1
1
1
1
a + b, a + b ;
2
2
2
2
2
x = 2 F3
4x .
a, b, a + b 1;
Solution 1. Consider the product

0 F1 (; a; x)0 F1 (; b; x)
=
=
xn+k
(a)k (b)n k!n!
n,k=0
n
X
X
xn
n=0 k=0
n
X
X
(a)k (b)nk k!(n k)!
(1 b n)k (n)k
xn
(a)k k!
(b)n n!
n=0 k=0
n,
1
n;
X
xn
1
=
F
.
(b)n n!
n=0
a;
=
We then use the result in Exercise 6, page 69, to get
X
(a + b 1)2n xn
0 F1 (; a; x)0 F1 (; b; x) =
(b) (a)n (a + b 1)n n!
n=0 n

a+b1
X
2
n

=
(a)n (b)n (a + b 1)n n!
a+b
2n
n=0
2 xn
n
2
a+b a+b1
,
;
2
= 2 F3 F 2
4x .
a, b, a + b 1;
1 3
, ;
Z t
4 4
1
1
1
1
2
2
x 2 (t x) 2 [1 x (t x) ] 2 dx = t2 F1
2
0
1;
Solution 2. We use theorem 12 on the integral

Z t
1
1
1
A=
x 2 (t x) 2 [1 x2 (t x)2 ] 2 dx.
0
t4
.
16
k k ss ctk+s
(k + s)k+s
Now
41

1
; ; x2 (t x)2 dx,
2
0
3
1
1
so in Theorem 12 we pput |alpha = , = , p = 1, q = 0, a1 = , c = 1, k =
2
2
2
2, s = 2. The result is
1 3 5 1 3
, , , , ;
2 4 4 4 4

3 1
22 22 t2+2
,
A=B
,
t 5 F4
4
2 2
4
2 3 4 5
, , , ;
4 4 4 4
or

1 3
1 3
3
1
4, 4;

4, 4;
2
2
4
t
t4
A=
t 2 F1
= t2 F1
,
(2)
2
16
16
1, 1;
1;
Z
A=
x 2 (t x) 2 1 F0
as desired.
Problem 3. With the aid of Theorem 10, show that
(1 + 21 a)
cos 21 a(1 a)
=
(1 + a)
(1 21 a
and that
sin (b 12 a)(b 21 a)
(1 + a b)
.
=
sin (b a)(b a)
(1 + 12 a b)
Thus put Dixons theorem (Theorem 11) in the form
a, b, c;
cos 21 a sin (b 12 a) (1 a)(b 21 a)(1 + a c)(1 + 21 a b c)
1 =
.
3 F2
sin (b a)
(1 21 a)(b a)(1 + 12 a c)(1 + a b c)
1 + a b, 1 + a c;
Solution 3. We first note that, since (z)(1 z) =

,
sin z

1
1
1
1
1
1+ a 1 a
a
a 1 a
2
2
2
2
2
=
(1 + a) (1 a)
a(a)(1 a)
sin a
=
a
2 sin
2
2 cos a sin a
2
2
=
2 sin a
a 2
= cos
.
2

a
1
1
1
cos
sin b a (1 a) b a (1 + a c) 1 + a b c
a, b, c;
2
2
2
2

1
F
=

,
3 2
1
1
sin (b a)
1 + a b, 1 + a c;
1 a (b a) 1 + a c (1 + a b c)
2
2
1
1
so long as a, a, b a, b a are not integers and the gamma functions involved
2
2
have no poles.
But now we have arrived at an identity (for non-integral values of certain parameters) which has the property that both members are well-behaved if a is a negative
integer or zero. It follows that the identity continues to be valid for a = n, n a
a
= 0.
nonnegative integer. If a = (2n + 1), an odd negativer integer, cos
2
Problem 4. Use the result in Exercise 3 to show that if n is a non-negative integer,
2n, , 1 2n;
(2n)!()n ( )n
1 =
.
3 F2
n!()2n ()n
1 2n, ;
Solution 4. If a = 2n in the identity of Exercise 3 above, and if we chose
b = , c = 1 2n , we obtain
2n, , 1 2n;
cos(n) sin ( + n) (1 + 2n)( + n)(1 2n 1 + + 2n)(1 n 1 + + 2n)
1 =
3 F2
sin ( + 2n)
(1 + n)( + 2n)(1 n 1 + + 2n)(1 2n 1 + + 2n)
1 2n, ;
(1)n (1)n sin (2n)!()n ()( + n)
=
sin
n!()2n ( + n)( )
(2n)!()n ( )n
=
,
n!()2n ()n
as desired.
Problem 5. With the aid of the formula in Exercise 4 prove Ramanujans theorem:
, ;
;
;
x2
x
x =2 F3
.
F1
1 F1
4
1 1
1
;
;
1
, , + ;
2 2
2
Solution 5. Consider the product
1 F1 (; ; x)1 F1 (; ; x)
X
n
X
(1)k ()k ()nk xn
()k ()nk k!(n k)!
n=0 k=0
X
n
X
(n)k ()k (1 n)k ()n xn
=
k!()k (1 n)k n!()n
n=0 k=0
n, , 1 n;
X
()n xn
1
=
.
3 F2
n!()n
n=0
, 1 n;
Since the product of the two 1 F10 s is an even function of x, we may conclude that
3 F2
2n 1, , 1 2n 1;
1 =0
, 1 2n 1;
43
and that
1 F1 (; ; x)1 F1 (; ; x)
3 F2
2n, , 1 2n;
, 1 2n;
(2n)!()n ( )n ()2n x2n
=
,
n!()2n ()n
(2n)!()2n
n=0
n=0
by Exercise 4. Hence we get Ramanujans theorem
()2n x2n
1
(2n)!()2n
, ;
1 F1 (; ; x)1 F1 (; ; x)
= 2 F3
1 1
1
, , + ;
2 2
2
x
4
Problem 6. Let n =3 F2 (n, 1an, 1bn; a, b; 1). Use the result in Exercise 3
to show that 2n+1 = 0 and
2n =
(1)n (2n)!(a + b 1)3n

.
n!(a)n (b)n (a + b 1)2n
Solution 6. From Exercise 3, we get
, , ;
cos a
2 sin ( 2 ) (1 )( 2 )(1 + )(1 + 2 )
1 =
.
3 F2
sin ( )
(1 2 )( )(1 + 2 )(1 + )
1 + , 1 + ;
Consider
n n = 3 F2 (n, 1 a n, 1 b n; a, b; 1). We wish to use = n, but
cos
= 0 for n odd. Hence 2n+1 = 0 and
2
2n = 3 F2 (2n, 1 a 2n, 1 b 2n; a, b; 1).
Therefore in the result from Exercise ?? we put = 2n, = 1 a 2n, =
1 b 2n, and thus obtain
n
cos(n) sin (1 a n) (1 + 2n)(1 a n)(b)(1 + a + b + 3n)

sin (1 a)
(1 + n)(1 a)(b + n)(1 + a + b + 2n)
cos2 (n) sin (1 a) (2n)!(1)n (a + b 1)3n
=
,
sin (1 a)
n!(a)n (b)n (a + b 1)2n
=
so that
n =
(1)n (2n)!(a + b 1)3n

.
n!(a)n (b)n (a + b 1)2n
Problem 7. With the aid of the result in Exercise 6 show that
1
1
1
(a + b 1), (a + b), (a + b + 1);
3
3
3
0 F2 (; a, b; t)0 F2 (; a, b; t) =3 F8
1
1 1 1
1 1
1
a, b, a + , b, b + , (a + b 1), (a + b);
2
2 2 2
2 2
2
27t2
64
Solution 7. Let us consider the product

(t)
= 0 F2 (; a, b; t)0 F2 (; a, b; t)
X
n
X
=
n=0 k=0
X
n
X
tn
(n)k (1 a n)k (1 b n)k
k!(a)k (b)k
n!(a)n (b)n
n=0 n=0
n
X
t
=
n
n!(a)
n (b)n
n=0
X
2n t2n
=
,
(2n)!(a)2n (b)2n
n=0
in terms of the n of Exercise 6 above. We already knew that 2n+1 = 0 which

checks with the fact that (t) is an even function of t. Since, by Exercise 6,
n =
(1)n (2n)!(a + b 1)3n

,
n!(a)n (b)n (a + b 1)2n
we have
(t)
(1)n (a + b 1)3n t2n

n!(a)n (b)n (a)2n (a + b 1)2n
n=0
a+b a+b+1 2n
X
(1)n 33n a+b1
t
3
n 3
n 3
n

=
a+1
b+1
2n b
2n a+b1
2n a
2
2
n!(a)
(b)
2
n
n
2 n
2
2 n
2
2
n=0
n
n
n
=
,
a+b
2
n
or
F
(1,
a,
b;
x)
F
(;
a,
b;
x)
=
F
0 2
0 2
3 8
1
3 (a
1
1
+ b 1), (a + b), (a + b + 1);
3
3
a a+1 b b+1 a+b1 a+b

, ,
,
,
;
a, b, ,
2
2
2
2
2
2
27t2
64
Problem 8. Prove that

n
X
(1)nk ( b c)nk ( b)k ( c)k xnk
k=0
k!(n k)!()k
3 F2
k, b, c;
( b)n ( c)n (1 x)n
3 F2
n!()n
1 + b k, 1 + c k;
1
1
1
n, n + , 1 n;
2
2
2
1 + b n, 1 + c n;
and note the special case = b + c, Whipples theorem.
4x
(1 x)2
45
b, c;
t(1 x + xt) . Then
Solution 8. Let = 2 F1
;
=
=
=
=
X
( b)n ( c)n tn [(1 x) + xt]n
()n n!
n=0
X
n
X
( b)n ( c)n (1 x)nk xk tn+k
k=0 k=0
[n
X
2]
X
n=0 k=0
[n
X
2]
X
k!(n k)!()n
( b)nk ( c)nk (1 x)n2k xk tn
k!(n 2k)!()nk
( b)n ( c)n (1 x)n tn

(n)2n (1 n)k (1)k xk
,
k!(1 + c n)k (1 + b n)k (1 x)2k
n!()n
n=0 n=0
or
3 F2
n=0
n n1
,
, 1 n;
2
2
( b) ( c) (1 x)n tn
n
n
4x
.
n!()
n
(1 x)2
1 + b n, 1 + c n;
But also, since 1 t(1 x + x) = (1 t)(1 + xt),
b, c;
= (1 t)b+c (1 + xt)b+c 2 F1
t(1 x + xt) .
;
Hence
= (1 t)b+c (1 + xt)b+c
= (1 t)b+c1 (1 + xt)b+c
X
(b)n (c)n tn [1 x(1 t)]n
n!()n
n=0
n
X
k
X
(1) (b)n (c)n xk (1 t)k tn
n=0 k=0
k!(n k)!()n
or
X
(1)k (b)n+k (c)n+k xk (1 t)x tn+k
k!n!()n+k
n,k=0
b + k, c + k;
X
1
= (1 + xt)b+c (1 t)b+c
2 F1
+ k;
k=0
= (1 t)b+c (1 + xt)b+c
Now
2 F1
b + k, c + k;
b, c;
t = (1 t)bck 2 F1
+ k;
t .
+ k;
Therefore
= (1 + xt)b+c
2 F1
b, c;
(1)k (b)k (c)k (xt)k

t
k!()k
+ k;
k
k n+k
(1)
(b)
(c)
k
k ( b)n ( c)n x t
= (1 + xt)b+c
k!n!()n+k
n,k=0
n
X X (1)s (b)s (c)s ( b)ns ( c)ns xs tn
= (1 + xt)b+c
s!(n s)!()n
n=0 s=0
n, b, c;
X
(1)n ( b c)n xn tn X
=
3 F2
n!
n=0
n=0
1 + b n, 1 + c n;
k=0
x ,
or
=
n
X
X
n=0 k=0
3 F2
k, b, c;
nk
(1)
x
1 + c b, 1 + c x;
( b c)nk ( b)k ( c)k xnk n

t .
k!()k (n k)!
By equating coefficients of tn in the two expansions we obtain the desired identity

Exercises 9-11 below use the notation of the Laplace transform as in Exercise 16,
page 71.
a1 , . . . , ap ;
L tcp Fq
b1 , . . . , bq ;
1 + c, a1 , . . . , ap ;
(1 + c)
zt =
Fq
s1+c p+1
b1 , . . . , b q ;
Solution 9. We know that L {tm } =
a1 , . . . , a p ;
L tc p Fq
b1 , . . . , bq ;
z
.
s
(m + 1)
. Then,
sm+1
X
(a1 )n . . . (ap )n z n
zt
=
L {tn+k }
(b
)
.
.
.
(b
)
n!
1 n
q n
n=0
(a1 )n . . . (ap )n z n
(1 + c)n (1 + c)
n+k+1
(b
)
.
.
.
(b
)
n!s
1
1 n
p n
n=0
1 + c, a1 , . . . , ap ;
z
F (1 + c)
=
.
p+1 Fq
5
s1+c
b1 , . . . , b q ;
a1 , . . . , ap ;
1
L1
Fq+1
sp
s + 1, b1 , . . . , bq ;
z =p Fq+1
a1 , . . . , ap ;
z(1 e1 ) .
1, b1 , . . . , bq ;
Solution 10. In Chapter 4, Exercise 16 we found that

1
(1 et )n
1
L
=
.
s(s + 1)n
n!
It follows that
L 1
F
s p q+1
z = p Fq+1
a1 , . . . , ap ;
1 + s, b1 , . . . , bq ;

L1
sk
(s z)k+1
47
a1 , . . . , ap ;
z(1 et ) .
1, b1 , . . . , bq ;
k + 1;
zt
= 1 F1
1;
1
1
1
s
=
= 1 F0
(s z)k+1
s (1 54 )k+1
s
By Exercise 9 with c = 0,
k + 1;
(1)
1
L
1 F0
s
;
= L 1
k + 1;
(1)
z
.
5
1, k + 1;
2 F1
k + 1;
zt
= t0 F
1;
k + 1;
zt .
=F
1;
1;
d
Fq
dz p
p
Y
a1 , . . . , ap ;
z =
b1 , . . . , bq ;
m=1
q
Y
am
p Fq
bj
a1 + 1, . . . , ap + 1;
z .
b1 + 1, . . . , bq + 1;
j=1
Solution 12.
a1 , . . . , ap ;
d
p Fq
dz
b1 , . . . , b q ;
X
(a1 )n . . . (ap )n z n1
(b1 )n . . . (bq )n (n 1)!
n=1
X
(a1 )n+1 . . . (a0 )n+1 z n
(b1 )n+1 . . . (bq )n+1 n!
n=0
a1 + 1, . . . , ap + 1;
a1 . . . ap
=
p Fq
b 1 . . . bq
b1 + 1, . . . , bq + 1;
z .
Problem 13. In Exercise 19 page 71, we found that the complete elliptic integral
of the first kind is given by

1
1 1
2
K(k) = 2 F1
, ; 1; k .
2
2 2
Show that
t
p
K( x(t x))dx = arcsin

1
t .
2
Solution 13. We are given that

1
K(k) = 2 F1
2

1 1
2
, ; 1; k .
2 2
Now consider
Z
A=
p
K( x(t x))dx.
By the integral of Section 56, (i.e. Theorem 57) with = 1, = 1, k = 1, s = 1,

etc.
t

1 1
, ; 1; x(t x) dx
2 2
0
1 1 1 1
, , , ;
2 2 1 1
= B(1, 1)t21 4 F3
2 3
1, , ;
2 2

(1)(1)
1 1 3 t2
=
t 2 F1
, ; ;
2 (2)
2 2 2 4!
2

1 1 3 t
t
, ; ;
=
2 F1
2
2 2 2 2

t
= arcsin
,
2
A =
2 F1
1 1t2
1
22
by Exercise 18, Chapter 4.
Problem 1. By collecting powers of x in the summation on the left, show that
1
J2n+1 (x) =
2
n=0
J0 (y)dy.
0
Solution 1.
J2k+1 (x)
k=0
X
(1)n ( x2 )2n+2k+1
n!(n + 2k + 1)!
k,n=0
X
n
X
(1)nk ( x )2n+1
2
n=0 k=0
X
n
X
(n k)!(n + k + 1)!
(n)k (1)n ( x2 )2n+1

(n + 1)k n!(n + 1)!
n=0 k=0
n, 1;
X
(1)n ( x2 )2n+1
1
=
2 F1
n!(n + 1)!
n=0
n + 2;
X (n + 2)(2n + 1)(1)n ( x )2n+1

2
=
(2n
+
2)(n
+
1)n!(n
+
1)!
n=0
n x 2n+1
X
(1) ( 2 )
=
(2n + 1)n!n!
n=0
Z xX
(1)n ( y2 )2n
1
dy
=
2 0 n=0
n!n!
Z x
1
J0 (y)dy.
=
2 0
=
Problem 2. Put the equation of Theorem 39, page 113, into the form

X
1
(A)
exp z(t t1 ) = J0 (z) +
Jn (z)[tn + (1)n tn ].
2
n=1
Use equation (A) with t = i to conclude that
cos z = J0 (z) + 2
(1)k J2k (z),
k=1
sin z = 2
(1)k J2k+1 (z).
k=0
Solution 2. We know that

X
X
X
1
1
n
n
Jn (z)t =
Jn (z)t + J0 (z) +
Jn (z)tn .
exp z(t t ) =
2
n=
n=
n=1
Now Jn (z) = (1)n Jn (z). Hence

X
X
z
1
exp (t
=
Jn (z)E n + J0 (z) +
Jn (z)T tn
2
t
n=1
n=1
X
n
= J0 (z) +
Jn (z)[t + (1)n tn ].
n=1
Now use t = i. Then (1)n in = i2n in = in and

1
exp
i
= exp(iz) = cos z + i sin z.
2
i
49
Therefore
cos z + i sin z = J0 (z) + 2
in Jn (z).
n=1
But Jn (z) = (1)n Jn (z), so we have

cos z i sin z = J0 (z) + 2
(1)n in Jn (z).
n=1
(note: above argument can be done more easily by equating even and odd function
of z) Thus we get
cos z = J0 (z) + 2
i2n d2n (z) = J0 (z) + 2
n=1
(1)n J2n (z)
n=1
and
i sin z = 2
i2n+1 J2n+1 (z),
n=0
or
sin z = 2
(1)n J2n+1 (z).
n=0
Problem 3. Use t = e
in equation (A) of Exercise 2 to obtain the results
cos(z sin()) = J0 (z) + 2
J2k (z) cos(2k),
k=1
sin(z sin()) = 2
J2k+1 (z) sin(2k + 1).
k=0
Solution 3. Put t = ei . Then tn + (1)n tn = eni + (1)N eni and

1
z
t
= exp(iz sin()) = cos(z sin()) + i sin(z sin()).
exp
2
t

z
1
ni
n ni
n
n
Also e +(1) e
= [1+(1) ] cos(n)+[1(1) ] sin(n). From exp
t
=
2
t
X
J0 (z) +
Jn (z)[tn + 9 1)n tn ] we thus obtain
n=1
cos(z sin )+i sin(z sin ) = J0 (z)+
Jn (z)[(1+(1)n ) cos(n)+(1(1)n ) sin(n)].
n=1
Now equate even function of z on the two sides, then odd functions of z on the two
sides to get
X
cos(z sin ) = J0 (z) + 2
J2k (z) cos(2k)
k=1
and
sin(z sin ) = 2
X
k=0
J2k+1 (z) sin(2k + 1).
51
Problem 4. Use Bessels integral, page 114, to obtain for integral n in the relations
Z
2
(B)
[1 + (1)n ]Jn (z) =
cos(n) cos(z sin())d,
0
Z
2
n
(C)
[1 (1) ]Jn (z) =
sin(n) sin(z sin())d.
0
With the aid of (B) and (C) show that for integral k,
Z
1
J2k (z) =
cos(2k) cos(z sin(thetaA))d,
0
Z
1
J2k+1 (z) =
sin(2k + 1) sin(z sin())d,
0
Z
cos(2k + 1) cos(z sin()d = 0,
0
Z
sin(2k) sin(z sin())d = 0.
0

Jn (z) =
cos(n z sin )d.

0
Then
Jn (z) =
cos(n) cos(z sin )d +

0
sin(n) sin(z sin )d.

0
Now change z to (z) to get

Z
Z
1
1
cos(n) cos(z sin )d
(1)n Jn (z) =
0
0
We then obtain
(B)
2
[1 + (1) ]Jn (z) =
cos(n) cos(z sin )d

0
Z
2
0
Use (B) with n = 2k and (C) with n = 2k + 1 to obtain
Z
1
J2k (z) =
cos(2k) cos(z sin )d,
0
Z
1
J2k+1 (z) =
sin(2k + 1) sin(z sin )d.
0
Use (B) with n = 2k + 1 and (C) with n = 2k to obtain
Z
cos((2k + 1)) cos(z sin )d = 0,
(C)
[1 (1)n ]Jn (z) =
sin(2k) sin(z sin )d = 0.

0
Problem 5. Expand cos(z sin()) and sin(z sin()) in Fourier series over the interval < < . Thus use Exercise 4 to obtain in another way the expansions
in Exercise 3.
Solution 5. In the interval < < , the Fourier approximation of f () is

f () =
X
1
a0 +
(an cos(n) + bn sin(n)),
2
n=1
in which
Z
1
f () cos(n)d,

Z
1
bn =
f () sin(n)d.

Consider first f () = cos(z sin ), an even function of . For this function
Z
2
an =
cos(n) cos(z sin )d, bn = 0.
0
an =
By the results in Exercise 4 we obtain a2k+1 = 0, a2k = 2J2k (z). Hence

cos(z sin ) = J0 (z) + 2
J2k (z) cos(2k), < < .
k=1
Next, let f () = sin(z sin ), an odd function of . For this function,

Z
2
an = 0, bn =
0
From Exercise 4 we get b2k = 0, b2k+1 = 2J2k+1 (z). Hence
sin(z sin ) = 2
J2k+1 (z) sin((2k + 1))), < < .
k=0

1
1
x(t t1 ) by exp x(t t1 ) , obtain
2
2
the coefficient of t0 and thus show that
X
J02 (x) + 2
Jn2 (x) = 1.

Problem 6. In the product of exp
n=1
1
For real x conclude that |J0 (x)| 1 and |Jn (x)| 2 2 for n 1.

Z
1
exp
t
=
Jn (x)tn
2
t
n=
and thus that

Z
1
exp
t
=
(1)k Jk (x)Jn (x)tn+k .
2
t
n=
The coefficient of t0 on the right is (k = n)
(1)n Jn (x)Jn (x) = 0,
n=
from which we obtain

J02 (x) + 2
(1)n Jn (x)Jn (x) = 1.
n=1
53
But J2n (x) = (1)n Jn (x). Hence

J02 (x) + 2
Jn2 (x) = 1.
n=1
It follows at once, for real x, that

|J0 (x)| 1,
1
|Fn (x)| , n 1.
2
Problem 7. Use Bessels integral to show that |Jn (x)| 1 for real x and integral
n.
Solution 7. Bessels integral is
1
Jn (x) =
cos(n x sin )d.

0
For real x (and n), |cos(n x sin )| 1. Hence

Z
1
|Jn (x)|
d = 1.
0
Problem 8. By iteration of equation (8), page 111, show that
2m
m
X
dm
J
(z)
=
(1)mk Cm,k Jn+m2k (z),
n
dz m
n=0
where Cm,k is the binomial coefficient.

Solution 8. We have, with D =
d
, from (8), page 100,
dz
2DJn (z) = Jn1 (z) Jn+1 (z).

Then
22 D 2 Jn (z)
= 2DJn1 (z) 2DJn+1 (z)

= Jn2 (z) Jn (z) Jn (z) + Jn+2 (z)
= Jn2 (z) 2Jn (z) + Jn+2 (z).
Let us use induction. Assume

2m D m Jn (z) =
m
X
(1)mk Cm,k Jn+m2k (z),
k=0
as we k now is true for m = 1, 2. Then

2m+1 D m+1 Jn (z)
=
=
=
m
X
(1)mk Cm,k [Jn+m2k1 (z) Jn+m2k+1 (z)]
k=0
m+1
X
m
X
k=1
k=0
(1)mk+1 Cm,k1 Jn+m+12k (z) +
m
X
(1)m+1k Cm,k Jn+m+12k (z)
(1)m+1k [Cm,k + Cm,k1 ]Jn+m+12k (z) + (1)m+1 Jn+m+1 (z) + Jnm1 (z).
k=1
Now Cm,k + Cm,k1 = Cm+1,k (Pascals triangle), so that (using the last two terms
also),
m+1
X
2m+1 D m+1 Jn (z) =
(1)m+1k Cm+1,k Jn+m+12k (z),
k=0
which completes the induction.

Problem 9. Use the result in Exercise 1, page 105, to obtain the probduct of two
Bessel functions of equal argument.
Solution 9. We know already that
a+b a+b1
,
;
2
2
0 F1 (; a; x)0 F1 (; b; x) = 2 F3
4x .
a, b, a + b 1;
Then
z n z m
Jn (z)Jm (z)
z2
z2
2
2
)0 F1 (; m + 1; )
0 F1 (; n + 1;
(n + 1)(m + 1)
4
4
z n+m
n+m+2 n+m+1
,
;
2
2
2
=
2 .
2 F3
z
(n + 1)(m + 1)
n + 1, m + 1, n + m + 1;
=
Problem 10. Start with the power series for Jn (z) and use the form (2), page 18,
of the Beta function to arrive at the equation
Z 2
2( 21 z)n
Jn (z) =
sin2n cos(z cos )d,
( 12 )(n + 21 ) 0
1
for R(n) > .
2

1
z 2k+n
(1)
k 2k+n
X
X
2
(1) z
k
Jn (z) =
=
.
22k+1 k!(k + n + 1)
2n (2k)!(k + n + 1)
k
k=0
k=0
Also

1
2 k
(k + n + 1)

1
1
n+
k+
2
2

=
1
1
(k + n + 1) n +
2
2

1
1
B k + ,n +
2
2

=
1
1
n+
2
2
Z 2
2

=
cos2k sin2n d.
1
1
0
n+
2
2

Therefore
z n
2
Jn (z) = 2
1
n+
2
z n
2
= 2
1
n+
2
Z
1
2
1
2
sin2n cos(z cos )d
sin2n cos(z cos )d.
Problem 11. Use the property

d
1 du
0 F1 (; a; u) =
0 F1 (; a + 1; u)
dx
a dx
to obtain the differential recurrence relation (6) of Section 60.
d
1 du
0 F1 (; a; u) =
0 F1 (; a + 1; u). Since
dx
a dx
z n
(1)
Jn (z) =

z2
2
0 F1 ; 1 + n;
(1 + n)
4
we obtain
d n
[z Jn (z)]
dz

1
1 z
z2
F
;
2
+
n;
0
1
2n (1 + n) 1 + n
2
4
z n+1

z2
= z n 2
0 F1 ; 2 + n;
(2 + n)
4
= z n Jn+1 (z),
which yields (6) of Section 60.

Problem 12. Expand
0 F1
1 + ;
2xt t2
in a series of powers of x and thus arrive at the result

t 2x
t
12
p
X
J+n (t)xn
J ( t2 2xt) =
.
n!
n=0
55

1
Solution 12. Consider 0 F1 ; 1 + ; (2xt t2 ) . We obtain
4
X
tn (2x t)n
2xt t2
F
=
=
0 1
2n
2 (1 + )n n!
4
n=0
1 + ;
X
n
X
(1)k (2k)n tn+2k
=
2n+2k
w
(1 + )n+k k!n!
n=0 k=0
X
X
tn (2x)n
(1)k t2k
=
22k k!(1 + + n)k 22n n!(1 + )n
n=0 k=0
n
t

xn (1 + )
X
t2
2
=
.
0 F1 ; 1 + + n;
4
n!(1 + + n)
n=0
Now
+n
t

t2
2
J+n (t) =
,
0 F1 ; 1 + + n;
( + n + 1)
4
so we obtain
;
0 F1
1 + ;
2xt t
4
= (1 + )
X
Jn+ (t)xn
t
2
n!
n=0
or
t2 2xt
2
!
(1 + )J
or

t 2x
t
2
p
X

Jn+ (t)xn
t
,
t2 2xt = (1 + )
2
n!
n=0
p
X
Jn+ (t)xn
.
J ( t2 2xt) =
n!
n=0
Problem 13. Use the realtions (3) and (6) of Section 60 to prove that: For real
x, between any two consecutive zeros of xn Fn (x), there lies one and only one zero
of xn Fn+1 (x).
Solution 13. We are given that
d n
[x Jn (x)] = xn Jn1 (x),
dx

d n
x Jn (x) = xn Jn+1 (x).
dx
We know that Jn (x) has exactly n zeros at x = 0. Let the others (we have proved
there are any) on the axis of reals be at 1,n , 2,n , . . . . The curve y = xn Jn (x) has
its real zeros only at the s. By Rolles theorem we see that the zeros 1,n1 , 2,n2 , . . .
of
y 0 = xn Jn+1 (x)
are such that an odd number of them lie between each two consecutive s. The
curve
y2 = xn+1 Jn+1 (x)
57
has its zeros at x = 0 and at the s. But

y20 = xn+1 Jn (x),
so the s lie between consecutive s.
Problem 14. For the function In (z) of Section 65 obtain the following properties
by using the methods, but not the results, of this chapter:
(1)zIn0 (z) = zIn1 (z) nIn (z),
(2)zIn0 (z) = zIn+1 (z) + nIn (z),
(3)2In0 (z) = In1 (z) + In+1 (z),
(4)2nIn (z) = z[In1 (z) In+1 (z)].
= in Jn (iz)
In (z)
z n
2

z2
=
0 F1 ; 1 + n;
(1 +
n)
4

2 k
n
z
X
1
z
2
=
(1 + n)
(1 + n)k k! 4
k=0
X
z 2k+n
.
=
22k+n k!(k + n + 1)

k=0
(1 + n)k = (1 + n)(2 + n) . . . (k + n),

so
(1 + n)(1 + n)k = (k + n + 1).
So, as in the method of Section 60,
d n
[z In (z)]
dz
X
k=0
= zn
=z
z 2k+n2n1
+ k)
22k+n1 k!(n
k=0
X
k=0
z 2k+n1
22k+n1 k!(n + k)
z 2k+n1
+ (n 1) + 1)
22k+n1 k!(k
= z n In1 (z),
or
z n In0 (z) + nz n1 In (z) = z n In1 (z),
which is equivalent to
zIn (z) = zIn1 (z) nIn (z),
which is (1).
Similarly,
d n
[z In (z)]
dz
=
=
d X
z 2k
dz
22k+n k!(k + n + 1)
k=0
X
z 2k1
k=1
22k+n1 (k 1)!(k + n + 1)
z 2k+1
22k+n+1 k!(k + n + 1 + 1)
k=0
n
= z In+1 (z),
=
and
z n In0 (z) nz n1 In (z) = z n In+1 (z),
or
zIn0 (z) = nIn (z) + zIn+1 (z),
which is (2).
Adding (1) and (2):
2zIn0 (z) = zIn1 (z) + zIn+1 (z)
or
2In0 (z) = In1 (z) + In+1 (z),
which is (3).
Equating the right sides of (1) and (2):
zIn1 (z) nIn (z) = zIn+1 (z) + nIn (z),
or
2nIn (z) = z[In1 (z) In+1 (z)],
which is (4).
Problem 15. Show that In (z) is one solution of the equation
z 2 w00 + zw0 (z 2 + n2 )w = 0.
Solution 15. (Solution by Leon Hall) Because In is a 0 F1 function times z n , we
know from Section 46 that u = 0 F1 (; b; y) is a solution of
y
d2 y
du
+b
u = 0,
dy 2
dy

z2
and so 0 F1 ; 1 + n;
is a solution of
4
d2 u
du
zu = 0.
+ (2n + 1)
dz 2
dz
Thus is w = z n u, In (z) will be a solution of
z
z n+1 w00 2nz n w0 + n(n + 1)z n1 w + (2n + 1)[z n w0 nz n1 w] z n+1 w = 0

or
z n [zw00 (2n 2n 1)w0 [n(n + 1)z 1 + n(2n + 1)z 1 + z]w] = 0
or
z 2 w00 + zw0 (n2 + z 2 )w = 0.
59
1
Problem 16. Show that, for Re(n) > ,
2
n
Z 2
1
2 2z

sin2n cosh(z cos())d.
In (z) =
12 n + 12 0
Solution 16. (Solution by Leon Hall) For n not a negative integer,
In (z) = in Jn (iz),
and using the result of Problem 10,
In (z) = in
2( 12 iz)n
1
( 2 )(n + 12 )
1
2
sin2n cosh(z cos )d.
The powers of i cancel, and because cos(iw) = cosh w we get

Z 12
2( 21 z)n
In (z) =
sin2n cosh(z cos )d
( 12 )(n + 21 ) 0
for Re(n) >
1
as desired.
2
Problem 17. For negative integral n define In (z) = (1)n In (z), thus completing
the definition in Section 65. Show that In (z) = (1)N In (z) and that
"
#
X
1
n
1
exp z(t + t ) =
In (z)t .
2
n=
Solution 17. (Solution by Leon Hall) We have
In (z)tn
n=
1
X
(1)n In (z)tn +
n=
In (z)tn
n=0
(1)
n+1
In+1 (z)t
n1
In (z)tn .
n=0
n=0
Now proceed exactly

difference beingthat
as in the 2proof
of Theorem 39, the only
z
z2
In (z) involves 0 F1 ; 1 + n;
whereas Jn (z) involves 0 F1 ; 1 + n;
, to
4
4
get

X
1
In (z)tn = exp z(t + t1 ) .
2
n=
Problem 18. Use the integral evaluated in Section 56 to show that

Z tp
p
n+ 1
t
n
n
2
1
[ x(t x)] Jn ( x(t x))dx = 2
t
Jn+ 2
.
2
0
Solution 18.
Z
0
p
p
[ x(t x)]n Jn ( x(t x))dx =
1
2n (1
+ n)
xn (tx)n 0 F1
1 + n;
x(z x)
dx.
4
Now use Theorem 37 with = n + 1, = n + 1, p = 0, q = 1, b1 = 1 + n, c =

1][4
k = 1, s = 1. The result is
,
n + 1, n + 1;
Z tp
t2
p
1
[ x(t x)]n Jn ( x(t x))dx = n

B(1 + n, 1 + n)t2n+1 F
44
2 (1 + n)
0
2n + 2 2n + 3
1 + n,
,
;
2 2
t2
(1 + n)(1 + n) 2n+1
= n
t
0 F1
16
2 (1 + n)(2 + 2n)
3
n+ ;
2
;

n+ 12
2
3
t
1
t
tn+ 2
(1 + n) n +
2n+1
2
r
2

F
=
0 1
3
4
(2 + 2n) n +
3
2
n+ ;
2

3

(1 + n)
+n
t
2
1
= 2n+1 tn+ 2
Jn+ 12
.
(2 + 2n)
2

1
2z1
2
(z) z +
2
, we get
By Legendres duplication formula, (2z) =

3
1+2n
2
(1 + n)
+n
2
(2 + 2n) =
.
Hence
Z
p
p
[ x(t x)]n Jn ( x(t x))dx

1
2n+1 tn+ 2
t
1
J
n+ 2
1+2n
2
2

n+ 1
t
n
=2
t 2 Jn+ 12
.
2
=
Problem 19. By the method of Exercise 18 show that

Z 1
1 x sin( x)dx = 1 J2 (),

0
and, in general, that

c
Z 1
2
c1 12 n
(1 x) x Jn ( x)dx = (c)
Jn+c ().
0
Z
1 x sin( x)dx.
We know that

3 z2
sin z = z 0 F1 ; ;
.
2
4
61
Hence
1
1 x sin( x)dx =

1
1
3 2 x
x 2 (1 x) 2 0 F1 ; ;
dx
2
4
We now use Theorem 3 with t = 1, =
2
, k = 1, s = 0. We get
4
3
3
3
, = , p = 0, q = 1, b1 = , c =
2
2
2
1 x sin( x)dx
Now let us turn to

Z 1
1
(1 x)c1 x 2 n Jn ( x)dx =
0
2
= B
1 F
4
3 3
, ;
2 1

3
3

2
2
2
=
0 F1 ; 3;
(3)
4

3
;
2
( 2 )n
(1 + n)
3 3
,
2 2
1
c1 n
(1 x)
x 0 F1
2 x
; 1 + n;
4

dx
and use Theorem 37 with = n + 1, = c, p = 0, q = 1, b1 = 1 + n, t = 1, c =

2
, k = 1, s = 0. The result is
4
n + 1;
Z 1
n
2
(2)
B(n + 1, c)F
(1 x)t1 xtn Jn ( x)dx =
1
(1
+
n)
4
0
1 + n,c + n + 1;

( 2 )n
(n + 1)(c)
2
F
=
;
c
+
n
+
1;
0 1
(1
+n) (n + c + 1)
4

c
= (c)
Jn+c (),
2
as desired.
Z t
Z t
exp[2x(t x)]I0 [2x(t x)]dx =
exp( 2 )d.
0
exp[2x(t x)]I0 [2x(t x)]dx. Now

0
exp[2x(t x)]I0 [2x(t x)] = exp[2x(t x)]0 F1 [; 1; x2 (1 x)2 ].

In Kummers second formula we have

1 z2
.
1 F1 (a; 2a; 2z) = e 0 F1 ; 1 + ;
2 4
z
Use a =
1
and z = 2x(t x) to get
2

exp[2x(t x)]I0 [2x(t x)] = 1 F1

1
; 1; 4x(t x) .
2
Then,
t
Z
exp[2x(t x)]I0 [2x(t x)]dx =

1 F1

1
; 1; 4x(t x) dx,
2
to which we may apply Theorem 37 with = 1, = 1, p = 1, q = 1, a1 =

1, c = 4, k = 1, s = 1. We thus get
1 1 1
, , ;
2 1 1
Z t
t2
exp[2x(t x)]I0 [2x(t x)]dx = B(1, 1)tF
4
4
0
2 3
1, , ;
2 2

1 3
2
= t 1 F1
; ; t
2 2
1
n
t2n+1
(1)
X
2 n

=
3
n=0
n!
2 n
X
(1)n tn+1
=
n!(2n + 1)
n=0
Z tX
(1)n 2n
=
d
n!
0 n=0
Z t
=
exp( 2 )d.
1
, b1 =
2

Z t
1
1 2
1 2
t I0
t .
[x(t s)] 2 exp[4x(t x)]dx = exp
2
2
0
Solution 21.
Z t
Z t
1
1
12
[x(t x)] exp[4x(t x)]dx =
x 2 (t x) 2 0 F0 (; ; 4x(t x))dx.
0
1
1
, = , p = q = 0, k = 1, s = 1, c = 4 :
2
2
1 1
, ;
2 2

1 1 0
4t2
exp[4x(t x)]dx = B
,
t F
2 2
4
1 2
, ;
2 2
( 12 )( 12 )
1
=
F
; 1; t2
1 1
(1)

2

1 2
t4
= exp
t 0 F1 ; t;
16
22 2
t
t
= exp
I0
.
2
2
We use Theorem 37 with =
Z
0
[x(t x)] 2
63
Problem 22. (Solution by Leon Hall) Obtain Neumanns expansion

z n
2
X
(n + 2k)(n + k 1)!Jn+2k (z)
k!
k=0
, n 1.
Solution 22. Let

Fn (z) =
X
(n + 2k)(n + k 1)! z n
k!
k=0
Then

d z n
Jn+2k (z)
dz
2
Jn+2k (z).
n z n
0
Jn+2k
(z)
Jn+2k (z)
2 2
z2 n h
i
n
0
Jn+2k
(z) Jn+2k (z)
=
2
z

( z2 )n
n(n + 2k)
0
0
nJn+2k (z) + 2kJn+2k (z)
Jn+2k (z).
=
n + 2k
z
=
z n
Using (8), Section 60, and (1), Section 61, and simplifying gives

( z )n
d z n
Jn+2k (z) = 2
[kJn+2k1 (z) (n + k)Jn+2k+1 (z)] .
dz
2
n + 2k
So
Fn0 (z)
#
"
z n X
X
(n + k)!
k(n + k 1)!
=
Jn+2k1 (z)
Jn+2k+1 (z).
2
k!
k!
k=0
k=0
Note that the k = 0 term in the first series is zero, and so shifting the index in the
first series yields the second series.
Thus, Fn0 (z) = 0, making Fn (z) = constant. From the structure of the Bessel
functions, we see that
z n n! z n
Fn (0) =
=1
2
n! 2
and so Fn (z) = 1, from which Neumanns expansion immediately follows.
Problem 23. (Solution by Leon Hall)
Prove Theorem 39, page
113, by
forming
1 1
1
zt and the series for exp zt
.
the product of the series for exp
2
2
Solution 23. Theorem 39 is: for t 6= 0 and for all finite z,

X
z
1
exp
t
=
Jn (z)tn .
2
t
n=
We know
exp
z X
zn n
t =
t
2
2n n!
n=0
and
0
z
X
X
(1)m z m m
(1)n z n n
exp t1 =
t
=
t .
2
2m m!
2n (n)!
n=
m=0
Now, in the product
!
an t
0
X
bn t
, the coefficient of tn , n 0, is
n=
n=0
given by
an+k bk
k=0
and the coefficient of tn , n > 0 is given by
ak bn+k .
k=0
z

hz
z
i
t exp t1 , or exp
t 1t , the coefficient of
Thus, in the product exp
2
2
2
tn for n 0 is:

X
k=0
z k+n
2k+n (k + n)!

(1)k z k
2k k!
(1)k z 2k+n
22k+n k!(k + n)!
k=0
= Jn (z),
=
and the coefficient of tn , n > 0 is:

k
X
X
z
(1)k+n z k+n
(1)k z 2k+n
n
=
(1)
2k k!
2k+n (k + n)!
22k+n k!(k + n)!
k=0
k=0
n
= (1) Jn (z)
= Jn (z).
Problem 1. The function

2
erf (x) =
exp(t2 )dt
was defined on page 36. Show that

2x
erf (x) = 1 F1

1 3
2
; ; x .
2 2
Solution 1. Let
2
erf x =
Then,
Z
0
exp(t2 )dt.
erf x
=
=
=
=
t2n dt
(1)
2 X
0
n!
n=0
2 X (1)n x2n+1
n=0 n!(2n + 1)
1
2x X (1)n 2 n x2n
n=0
n! 23 n

2x
1 3
1 F1
; ; x2 .
2 2
Problem 2. The incomplete Gamma function may be defined by the equation

x
et t1 dt, R() > 0.
(, x) =
0
Show that
(, x) = 1 x 1 F1 (; + 1; x).
Solution 2. Let
Z
et t1 dt; Re() > 0.
(, x) =
0
Then,
Z
(, x)
=
0
xX
(1)n tn+1
n!
n=0
X
(1)n xn+
.
=
n!( + n)
n=0
Now, ( + n) =
( + 1)n
. Hence
()n
(, x) = 1 x
X
(1)n ()n xn
= 1 x 1 F1 (; + 1; x).
n!(
+
1)
n
n=0
Problem 3. Prove that

dk z
e 1 F1 (a; b; z) = (1)k (b a)k ez 1 F1 (a; b + k; z).
dz k
You may find it helpful to use Kummers formula, Theorem 42.
(b)k
Solution 3. Let D =
d
. Consider D[ez 1 F1 (a; b; z)]. We know that
dx
1 F1 (a; b; z)
Hence
= ez 1 F! (b a; b; z).
65
D k [ez 1 F! (a; b; z)]
We used D 1 F! (a; b; z) =
= D k 1 F1 (b a; b; z)
(b a)k (1)k
=
1 F! (b 1 + k; b + k; z)
(b)k
(1)k (b a)k z
=
e 1 F! (a; b + k; z).
(b)k
a
1 F! (a + 1; b + 1; z) k times.
b

Z
1
et t1 0 F1 (; b; zt)dt.
1 F1 (a; b; z) =
(a) 0
Z
(x) =
et tx1 dt, Re(x) > 0.
0
Then,
1 F1 (a; b; z)
=
=
X
(a)n z n
n!(b)n
n=0
1 X (a + n)z n
(a) n=0 n!(b)n

Z
a+n1 n
X
t
z
1
dt
=
et
(a) 0
n!(b)
n
n=0
Z
1
=
et ta1 0 F1 (; b; tz)dt, Re(a) > 0.
(a) 0
Problem 5. Show, with the aid of the result in Exercise 4, that
exp(t2 )t2an1 Jn (zt)dt =

z2
(a)z n
F
.
a;
n
+
1;
1 1
2n+1 (n + 1)
4
Solution 5. We obtain
Z
exp(t2 )t2an1 Jn (zt)dt =
A=
0
Z
0

2
et t2an1 z n tn
z 2 t2
;
1
+
n;
F
dt.
0 1
2n (1 + n)
4
Put t = . Then

Z
zn
1 a1
z2
A = n
e
d
0 F1 ; 1 + n;
2 (1 + n) 2 0
4

n
2
z
(a)
z
= n+1
,
1 F! a; 1 + n;
2
(1 + n) 1
4
as desired.
Problem 6. If k and n are non-negative integers, show that

k, + n;
;k > n
0
(n)k
1 =
F
; 0 k n.
;
()k
Solution 6. Let V (k, n) = F (k, + n; ; 1). Then

V (k, n) =
k
X
(k)s ( + n)s
s!()s
s=0
k
X
s=0
(1)s k!()n+s
.
s!(k s)!()s ()n
Hence
=
=
X
V (k, n)tk
k=0
X
k
X
k=0 s=0
k!
(1)s ()n+s tk
s!(k s)!()s ()n
(1)s ()n+s tk+s

s!k!()s ()n
k,j=0
X (1)s ( + n)s ts
= et
s!()s
s=0
= et 1 F! ( + n; ; t)
= et et 1 F! (n; ; t)
n
X
(n)k tk
=
.
k!()k
=
k=0
Thus
(n)k
V (k, n) =
()k
;0 k n
; k > n.
note: this method requires material in chapter 7 (on line 5)

Can do by Chapter 4 if first show that et 1 F1 (n; ; t) = 1 F1 ( + n; ; t).
Problem 1. From et (xt) =
n (x)tn , show that
n=0
n (xy) =
n
X
y k (1 y)nk k (x)
(n k)!
k=0
and in particular that

2n n
X
n
1
k (x)
x =
.
2
(n k)!
k=0
Solution 1. Let
et (xt) =
X
n=0
Then
n (x)tn .
67
e (xyt) =
n (xy)tn
n=0
and
eyt (xyt) =
n (x)y n tn .
n=0
But
et (xyt) = e(1y)t eyt (xyt)
so that
n (xy)t
!
!
X
X
(1 y)n tn
n n
n (x)y t
=
n!
n=0
n=0
X
n
X
(1 y)nk y k k (x) n
t .
=
(n k)!
n=0
n=0
k=0
Hence
n (xy) =
n
X
y k (1 y)nk k (x)
(n k)!
k=0
For y =
1
, we obtain
2

2 n
X
n
k (x)
1
x =
.
2
(n k)!
k=0
Problem 2. Consider the set (called Appell polynomials) n (x) generated by

ext A(t) =
n (x)tn .
n=0
Show that 00 (x) = 0, and that for n 1, n0 (x) = n1 (x).

Solution 2. Consider n (x) defined by
ext A(t) =
(A)
n (x)tn .
n=0
From (A) we obtain

text A(t) =
n0 (x)tn .
n=0
Hence
X
n=0
or
n (x)tn+1 =
X
n=0
n0 (x)tn ,
n1 (x)tn =
n=1
69
n0 (x)tn .
n=0
Therefore, 00 (x) = 0, and, for n 1, n0 (x) = n1 (x).

Problem 3. Apply Theorem 50, page 141, to the polynomials n (x) of Section 73
and thus obtain Theorem 45.
Solution 3. We have
et (xt) =
n (x)tn
n=0
and wish to apply Theorem 50, page 239. In the notation of Theorem 50 we have
A(t)et , H(t) = t, (t) =
n tn , 0 6= 0.
n=0
n
X
1
t
,, so an = ; H(t) = t, so h0 = 1, hn = 0 for n 1.
Now A(t) = et =
n!
n
n=0
Then also
X
tet
tA0 (t)
= t =t=
n tn+1 ,
A(t)
e
n=0
so that 0 = 1, n = 0 for n 1.
Furthermore,
X
tH 0 (t)
t1
=
=1=1+
n tn+1
H(t)
t
n=0
so that n = 0 for n 0. Hence by Theorem 50,

xn0 (x) nn (x) = 1 n10 (x) = n1 (x),
which is Theorem 45, page 224.
Problem 4. The polynomials n (x) of Exercise 3 and Section 73 are defined by
(A)
et (xt) =
n (x)tn ,
n=0
but by equation (9), page 134, they also satisfy

(B)
(1 t)c F
xt
1t

=
(c)n n (x)tn ,
n=0
for a certain function F . By applying Theorem 50, page 141, to (B), conclude
that the n (x) of (A) satisfy the relation
(c)n [xn0 (x) nn (x)] =
n1
X
(c)k [ck (x) + xk0 (x)],
k=0
for arbitrary c.
Solution 4. For the n (x) of Exercise 3 above, we know that

c
(1 t)
(B)

F
xt
1t

=
(c)n n (x)tn ,
n=0
where
F (u) =
(c)n n un
n=0
in terms of (u) =
n un .
n=0
We now use Theorem 50 on the polynomials fn (x) = (c)n n (x) of (B).

Here
A(t) = (1 t)c , H(t) =
t
.
1t
Then
log A(t) = c log(1 t),
H(t) = 1 +
1
1
; H 0 (t) =
,
1t
(1 t)3
c
A0 (t)
=
,
A(t)
1t
X
tH 0 (t)
t
1t
1
tn .
=
=
=
1
+
H(t)
(1 t)2 t
1t
n=1
Hence
tA0 (t) X n+1

=
ct
, so n = c for n 0;
A(t)
n=0
X
tH 0 (t)
=1+
tn+1 , so n = 1 for n 0.
H(t)
n=0
Thus Theorem 50 yields
x(c)n n0 (x)n(c)n n (x) =
n1
X
c(c)n1k n1k (x)x
k=0
n1
X
0
1(c)n1k n1k
(x),
k=0
or, with reversal of order of summation,

(c)n [xn0 (x) nn (x)] =
n1
X
(c)k [ck (x) + xk0 (x)].
k=0
Problem 5. Apply Theorem 50, page 141, to the polynomials yn (x) defined by (1),
page 135. You do not, of course, get Theorem 47, since that theorem depended upon
the specific character of the exponential.
71
Solution 5. On page 228 we find

A(t) exp
xt
1t

=
yn (x)tn .
n=0
In the notation of Theorem 50 we have

A(t) = A(t), H(t) =
1
t
=1
1t
1t
from which
H 0 (t) =
X
1
tH 0 (t)
1t
t
1
,
=
=
1
+
tn .
(1 t)2 H(t)
(1 t)2 t
1t
n=1
so
tA0 (t) X
=
n tn+1
A(t)
n=0
and
X
tH 0 (t)
=1+
tn+1 ,
H(t)
n=0
so that n = 1 for n 0.
Therefore, we may conclude that there exist constants n such that
xyn0 (x) nyn (x) =
n1
X
k yn1k (x) x
k=0
n1
X
0
1 yn1k
(x).
k=0
Problem 6. Apply Theorem 50 to the Laguerre polynomials through the genrating

relation 914), page 135, to get
()
xDL()
n (x) nLn (x) =
n1
X
()
k=0
for the Laguerre polynomials.

Solution 6. From page 228 we have
(1 t)1 exp
()
[(1 + )Lk (x) + xDLk (x) = 0
xt
1t

=
Ln() (x)tn .
n=0
We may therefore use Theorem 50 with

A(t) = (1 t)1 ,
H(t) =
t
1
=1
,
1t
1t
log A(t) = (1 + ) log(1 t),

H 0 (t) =
1
,
(1 t)2
X
tA0 (t)
(1 + )t
=
= (1 + )
tn+1 ,
A(t)
1t
n=0
X
X
1t
t
1
tH 0 (t)
n
=
=
=
1
+
t
=
1
+
tn+1 .
H(t)
(1 t)2 t
1t
n=1
n=0
Hence
n = 1 + , n 0;
n = 1, n 0.
Therefore we obtain
xDLn() (x)
nLn() (x)
= (1 + )
n1
X
()
Ln1k (x)
n1
X
k=0
()
DLn1k (x)
k=0
or
xDLn() (x) nLn() (x) =
(A)
n1
X
()
()
[(1 + )Lk (x) + xDLk (x)].
k=0
On page 230 we had

DLn() (x)
(B)
n1
X
()
Lk (x).
k=0
Using (B) with (A) we obtain

xDLn() (x) nLn() (x) = (1 + )DLn() (x) + xD 2 Ln() (x),
or
[xD 2 + (1 + x)D + n]Ln() (x) = 0,
as desired.
Problem 7. The Humbert polynomials hn (x) are defined by
(1 3xt + t3 )p =
hn (x)tn .
n=0
Use Theorem 52, page 144, to conclude that

xh0n (x) nhn (x) = h0n2 (x).
Solution 7. For hn (x) we have
(1 3xt + t3 ) =
hn (x)tn = 1 F0
n=0
In the notation of Theorem 52 we have
3xt t3 .
A(t) = 1, H(t) = 3t, g(t) = t3 , (t) = (1 t) .

Then
tA0 (t)
tH 0 (t)
3t
tg 0 (t)
t(3t2 )
= 0,
=
= 1,
=
= t2 .
A(t)
H(t)
3t
H(t)
3t
Hence
n = 0, n = 0; n = 0, n 0; 1 = 1, n = 0 for n = 0, n 2.
Therefore Theorem 52 yields
xh0n (x) nhn (x) = h0n2 (x).
Problem 8. For the yn (x) of Section 74 show that

F = A(t) exp
xt
1t
satisfies the equation

x
F
F
(1 t)tA0 (t)
F
t
= t2
F
x
t
t
A(t)
and draw what conclusions you can about yn (x).

X
xt
=
yn (x)tn .
Solution 8. Let F = A(t) exp
1t
n=0
Then
F
t
=
A(t) exp
x
1t
F
x
=
A(t) exp
t
(1 t)2
xt
1t
xt
1t
+ A0 (t) exp
xt
1t
Thus we obtain
x
F
F
A0 (t)
t(1 t)
= t(1 t)
F,
x
t
A(t)
or
x
F
F
F
tA0 (t)
t
= t2
(1 t)
F.
x
t
t
A(t)
Now let
tA0 (t) X
=
n tn+1 .
A(t)
n=0
Then

.
73
[xyn0 (x) nyn (x)]tn
n=0
=
=
yn (x)ctn+1
n=1
!
n tn+1
n=0
nyn (x)t
n=0
(n
n=2
n+1
!
yn (x)tn
n=0
X
n
X
X
n
X
n+1
!
n tn+2
n=0
nk yk (x)t
+
nk yk (x)t
n=0 k=0
n=0 k=0
n1
n2
X
X
X
X
1)yn1 (x)tn
n1k yk (x)tn +
n2k yk (x)tn .
n=1 k=0
n=2 k=0
n2
X
xyn0 (x) nyn (x) = (n 1)yn2 (x)
(n1k n2k )yk (x) 0 yn1 (x).
k=0
We may also write

F
F
A0 (t)
F
t(1 t)
= xt
t(1 t)
F,
x
t
x
A(t)
or
F
F
xt F
tA0 (t)
t
=
F,
x
t
1 t x
A(t)
from which it follows that
x
[xyn0 (x)
n=0
nyn (x)]y
= x
!
n+1
n=0
n
X
X
= x
!
yn0 (x)tn
n=0
yk0 (x)tn+1
n
X
X
!
n+1
n t
n=0
nk yk (t)tn=1 .
n=0 k=0
n=0 k=0
Hence, for n 1,
xyn0 (x)
nyn (x) =
n1
X
[xyk0 (x) + n1k yk (x)].
k=0
Problem 9. For polynomials an (x) defined by

c
(1 t)

A
xt
1t

=
an (x)tn
n=0
obtain what results you can parallel to those of Theorem 48, page 137.
Solution 9. Let an (x) be defined by
(1)
(1 t)

A
xt
1t

=
X
n=0
Let
(2)
yn (x)tn
n=0
n+2
Hence we get y00 (x) = 0, xy10 (x) y1 (x) = 0 y0 (x), and for n 2,
x(1 t)
A(u) =
X
n=0
n un .
an (x)tn .
X
n=0
!
n
yn (x)t
75
Put
F = (1 t)c A
xt
1t

.
Then
F
= t(1 t)c1 A0
x
F
= c(1 t)c1 A x(1 t)c2 A0 .
t
Hence
x
F
F
t(1 t)
= ctF.
x
t
From which we may write

(3)
F
F
F
t
= ctF t2
.
x
t
t
From (3) we obtain
[xa0n (x) nan (x)]tn
= c
n=0
an (x)tn+1
n=0
ngn (x)tn+1
n=0
[c + n 1]gn1 (x)tn .
n=1
Therefore,
a00 (x)
= 0 and for n 2,
xa0n (x) nan (x) = (c + n 1)an1 (x).
We may rewrite equation (3) in the form

x(1 t)
F
F
F
t(1 t)
= ctF xt
.
x
t
x
This leads to the identity

x
F
F
ct
xt F
t
=
F
x
t
1t
1 t x
from which we get
[xa0n (x)
n=0
nan (x)]t
= c
= c
=
n=0
X
n
X
n=0 k=0
n1
X
X
!
n
an (x)t
n=0
n+1
ak (x)t
X
n
X
k=0 k=0
[cak (x) + xa0k (x)]tn .
n=0 k=0
Hence we obtain the relation
!
n+1
!
t
n+1
n=0
a0k (x)tn+1
X
n=0
!
a0n (x)tn
xa0n (x) nan (x) =
n1
X
[cak (x) + xa0k (x)].
k=0
v
1
t
. Then t =
and 1 t =
.
In (1), let us put v =
1t
1t
1v
Hence (1) becomes
(1 v)c A(xv) =
X
an (x)(v)n
,
(1 v)n
n=0
or
n xn v n
n=0
=
=
=
X
an (x)(1)n v n
(1 v)n+c
n=0
X (1)n an (x)(c)n+k v n+k

n,k=0
n
X
X
n=0 k=0
n
X
X
n=0 k=0
k!(c)n
(c)n (1)nk ank (x) n
v
k!(c)nk
(1)k (c)n ak (x) n
v ,
(c)k (n k)!
which can be put in various other forms.

Next, from (1) we get
an (x)t
n=0
X
n (x)n tn
=
(1 t)n+c
n=0
X
n (1)n xn (c)n+k tn+k
=
(c)n k!
n,k=0
X
n
nk nk
X
(1)
x
nk (c)n tn
=
k1(c)nk
n=0
k=0
or
an (x)tn =
n=0
X
n
X
(1)k xk k (c)n tn
n=0 k=0
(n k)!(c)k
Therefore,
an (x) =
n
X
(1)k xk k (c)n
k=0
(n k)!(c)k
No problems in Chapter 9.
77
Problem 1. Start with the defining relation for Pn (x) at the beginning of this
chapter. Use the fact that
1
(1 2xt + t2 ) 2 = [1 (x +
x2 1)t] 2 [1 (x
x2 1)t] 2
and thus derive the result
Pn (x) =
n
X
( 1 )k ( 1 )nk (x +
2
x2 1)nk (x
k!(n k)!
k=0
x2 1)k

1
(1 2xt + t2 ) 2 =
Pn (x)tn .
n=0
Now
[1 (x +
x2 1)t][1 (x
x2 1)t] = (1 xt)2 (x2 1)t2 = 1 2xt + t2 .
Hence
X
n=0
Pn (x)tn
1
1
x2 1)t] 2 [1 (x x2 1)t] 2
!
!
X
X ( 1 )n (x x2 1)n tn
( 21 )n (x + x2 1)n tn
2
=
n!
n!
n=0
n=0
n
X
X
( 12 )k ( 21 )nk (x + x2 1)nk (x x2 1)k tn
.
=
k!(n k)!
n=0
= [1 (x +
k=0
Hence we obtain
Pn (x) =
X
( 1 )k ( 1 )nk (x +
2
k=0
x2 1)nk (x
k!(n k)!
x2 1)k
Problem 2. Use the result in Exercise 1 to show that
1
n; ;
2
+ x2 1)
Pn (x) =
(x x2 1)2
2 F1
.
n!
1
n;
2
Solution 2. From Exercise 1 we get, since x + x2 1 = (x x2 1)1 ,

( 21 )n (x
x2 1)n (x x2 1)2k
k!n!(1 12 n)k
k=0
1
n, ;
1
n
2
2
( )n (x + x 1)
= 2
(x x2 1)2
2 F1
n!
1
n;
2
Problem 3. In Section 93, equation (4), page 166, is
1
1
1
n,
n
+
;
2
2
2
( 21 )n (2x)n
1
.
Pn (x) =
2 F1
n!
x2
1
n;
2
We know from Section 34 that the 2 F1 equation has two linearly independent
solution:
Pn (x)
n
X
( 1 )k (n)k ( 1 )n (x +
2
2 F1 (a, b; c; z)
and
z 1c F (a + 1 c, b + 1 c; 2 c; z).
Combine these facts to conclude that the differential equation
(1 t2 )y 00 2ty 0 + n(n + 1)y = 0
has the two linearly independent solutions y1 = Pn (t) and y2 = Qn (t), where
Qn (t) is as given in equation (4) of Section 102.
n n 1
,
;
2
2
( 1 )n (2x)n
1
,
Pn (x) = 2
2 F1
2
n!
x
1
n;
2
from which w = xn Pn (x) is a solution of the 2 F1 differential equation with
n
n1
1
a = ,b =
, c = n, z = x2 .
2
2
2
Then consider

d2 w
3
dw n(n 1)
1
z(1 z) 2 +
n
n z
w=0
dz
2
2
dz
4
dz
dx
1
and put z = x2 . Then,
= 2x3 ,
= x3 ,
dx
dz
2
dw
dx dw
1 dw
=
= x3
dz
dz dx
2 dx

2
2
d w
1 d w 3
1
dw
1
3
= x6 2 x2 x3
= x6 w00 + x5 w1 .
dz 2
4 dx
2
2
dx
4
4
79
The equation in w and x becomes

x
(1x

1 6 00 3 5 2
3
n(n 1)
2
0 1 3 1
2
) x w + x x (1x )w x
n
n x
w = 0,
w0
4
4
2
2
2
4
or
x2 (x2 1)w00 + 3x(x2 1)w0 x[(1 2n)x2 (3 2n)]w0 n(n 1)w = 0,
or
x2 (x2 1)w00 + x[(2 + 2n)x2 2n]w0 n(n 1)w = 0.
Now put
w = xn y
w0 = xn y 0 nxn1 y
w00 = xn y 00 2nxn1 y 0 + n(n + 1)xn2 y
or
xn w = y
xn w0 = y 0 nx1 y
xn w00 = y 00 2nx1 y 1 + n(n + 1)x2 y
and the differential equation becomes
x2 (x2 1)y 00 2nx(x2 1)y 0 +n(n+1)(x2 1)y+x[(2+2n)x2 2n]y 0 n[(2+2n)x2 2n]yn(n1)y = 0,
or
x2 (x2 1)y 00 + x[2x2 ]y 0 + [n(n + 1)x2 (1 2) n2 n + 2n2 n2 + n]y = 0,
or
(1)
(1 x2 )y 00 2xy 0 + n(n + 1)y = 0.
Since the differential equation for w has the two linearly independent solutions
w1 = xn Pn (x) and
n1 1
n 1
+ + n;
+ + n;
2
2
2
2
1+n 2+n
,
;
2
2
12n
1 21 +n
z =x
w2 = z
2 F1
2 F1
3
3
+ n;
+ n;
2
2
Then the differential equation (1) for y has the linearly independent solutions
y1 = Pn (x) and
1+n 2+n
,
;
2
2
1
1n
,
y2 = x
F
2 1
x2
3
+ n;
2
which is a constant multiple of Qn (x) as given on page 314.
x2
.
Note that the whole thing could have been done much more simply by using the
properties of the Riemann P-symbol.
[xPn0 (x) nPn (x)]tn = t2 (1 2xt + t2 ) 2
n=0
and
n
[2]
X
X
(2n 4k + 1)Pn2k (x)tn = (1 2xt + t2 ) 2 .
n=0 k=0
Thus conclude that

[ n2
2 ]
xPn0 (x) nPn (x) =
(2n 4k 3)Pn22k (x).
k=0

1
(1 2xt + t2 ) 2 =
Pn (x)tn .
n=0
Put
1
F = (1 2xt + t2 ) 2 .
Then
3
F
= t(1 2xt + t2 ) 2
x
and
3
F
= (x t)(1 2xt + t2 ) 2 .
t
Hence
3
F
F
t
= t2 (1 2xt + t2 ) 2 ,
x
t
from which it follows that
(1)
[xPn0 (x) nP (x)]tn = t2 (1 2xt + t2 ) 2 .
n=0
Next we form the series

n
[2]
X
X
(2n 4k + 1)Pn2k (x)tn
n=0 k=0
(2n + 1)Pn (x)tn+2k
n,k=0
2 1
= (1 t )
(2n + 1)Pn (x)tn .
n=0
By equation (5), page 271, we get
81
0
0
(2n + 1)Pn (x) = Pn+1
(x) Pn1
(x), n 1.
Hence
n
[2]
X
X
"
(2n 4k + 1)Pn2k (x)t
0
Pn+1
(x)tn
n=2
Now
= 1 and
P00 (x)
n=0
= 0. Hence we have
[2]
X
X
#
0
Pn1
(x)tn
= (1 t2 )1 1 +
n=1
n=1
"
#
X
X
0
n1
0
n+1
2 1
= (1 t )
1+
PN (x)t
Pn (x)t
.
n=0 k=0
P10 (x)
"
(2n 4k + 1)Pn2k (x)t
2 1
= (1 t )
n=0 k=0
2 1
= (1 t )
Pn0 (x)tn1
n=0
1 F
F
t
t x
x
#
Pn0 (x)tn+1
n=0
1 F
t x
3
= (1 2xt + t2 ) 2 .
=
Now we use (1) to conclude that
[xPn0 (x)
n=0
nPn (x)]t
[2]
X
X
(2n 4k + 1)Pn2k (x)tn+2
n=0 k=0
n2
[X
2 ]
X
(2n 4k 3)Pn22k (x)tn .
n=2 k=0
Hence, for n 2,
[ n2
2 ]
xPn0 (x)
nPn (x) =
(2n 4k 3)Pn22k (x).
k=0
There are many other methods of obtaining this result.

Problem 5. Use Batemans generating function (3), page 163, with x = 0, t = 2y
to conclude that

1 1 2
.
0 F1 (; t; y)0 F1 (; 1; y) = 0 F3 ; 1, 1, ; y
2 4
Solution 5. We have, from Bateman,

X
t(x 1)
t(x + 1)
Pn (x)tn
F
;
1;
F
;
1;
=
.
0 1
0 1
2
2
(n!)2
n=0
Use x = 0, t = 2y to obtain
0 F1 (; 1; y)0 F1 (; 1; y)
X
2n Pn (0)y n
.
(n!)2
n=0
(1)k ( 12 )k
. Therefore
k!
X
22k (1)k ( 21 )k y 2k
F
(;
1;
y)
F
(;
1;
y)
=
0 1
0 1
k![(2k)!]2
k=0
X (1)k 22k ( 1 )k y 2k
2
=
k!22k k!( 12 )k 22k k!( 21 )k
k=0
X
(1)k y 2k
=
22k (k!)3 ( 21 )k
k=0

1 1 2
= 0 F3 ; 1, 1, ; y ,
2 4
now P2k+1 (0) = 0 and P2k (0) =
as desired.
Problem 6. Use Brafmans generating function, page 168, to conclude that
2 F1
c, 1 c;
1t
1 + t2 2 F1
c, 1 c;
1+t
2
1;
1 1
1 1 1
1
c, c + , c, 1 c;
2 2
2 2 2
2
t2
= 4 F3
1
1, 1, ;
2
Solution 6. Brafmans generating relation, page 287, is
c, 1 c;
2 F1
1;
1tp
2 F1
2
c, 1 c;
1;
2 F1
1 + t2
1 + t p X (c)n (1 c)n pn (x)tn

=
,
(n!)2
2
n=0
with p = (1 2xt + t2 ) 2 . We use x = 0. Note that P2n+1 (0) = 0 and P2n (0) =
(1)n ( 12 )n
. Then
n!
c, 1 c;
c, 1 c;
2
(c)2n (1 c)2n (1)n ( 21 )n t2n
1+tp
1t 1+t F
=
2 1
n![(2n)!]2
2
2
n=0
1;
1;
1c
2c
n 2n
X
( 2c )n ( c+1
2 )n ( 2 )n ( 2 )n (1) t
=
n!n!( 12 )n n!
n=0
c c+1 1c 2c
,
,
,
;
2
2
2
2
t2
= 4 F3
,
1
1, 1, ;
2
as desired.
Problem 7. Use equation (5), page 168, to obtain the results
sinn Pn (sin ) =
n
X
(1)k Cn,k cosk ()Pk (cos ),
k=0
Pn (x) =
n
X
(1)k Cn,k (2x)nk Pk (x),
k=0
Pn (1 2x2 ) =
n
X
(2x)k Cn,k Pk (x).
k=0
Solution 7. We are given

(1)
Pn (cos ) =
First use =
Thus (1) leads to
sin
sin
n X
n

Cn,k
k=0
sin( )
sin
nk
Pk (cos ).
. Then cos = sin , sin = cos , sin( ) = 1.

2

n X
nk

n
cos
1
Pn (sin ) =
=
Cn,k
Pk (cos ),
sin
cos
k=0
or
(sin )n Pn (sin ) =
(2)
n
X
(1)k Cn,k cosk Pk (cos ).
k=0
Next let us use = in (1). We thus obtain

n X

nk
n
sin
2 sin cos
Cn,k
Pk (cos )
sin
sin
k=0
n
X
(1)nk Cn,k (2 cos )nk Pk (cos )
= (1)n

Pn (cos )
n
X
k=0
(1)k Cn,k (2 cos )nk Pk (cos ).
k=0
With cos = x we get

(3)
Pn (x) =
n
X
(1)k Cn,k (2x)nk Pk (x).
k=0
Finally put = 2 in (1). The result is

Pn (cos(2)) =
or
2 sin cos
sin
n X
n
k=0

Cn,k
sin
2 sin cos
nk
Pk (cos ),
83
Pn (cos 2)
= (2 cos )n
=
n
X
n
X
(1)nk Cn,k (2 cos )kn Pk (cos )
k=0
(1)nk Cn,k (2 cos )k Pk (cos ).
k=0
Put cos = x. Then cos(2) = 2 cos2 () 1 = 2x2 1 and

Pn (2x2 1) = (1)n Pn (1 2x2 ).
Hence we have
Pn (1 2x2 ) =
(4)
n
X
(1)k Cn,k (2x)k Pk (x).
k=0
Problem 8. Use the technique of Section 96 to derive other generating function

relations for Pn (x). For instance, obtain the results
1 F2 (n; 1, 1; y)Pn (x)t
n=0
yt(x t + )
22
yt(x t )
0 F1
2
2
= 1 0 F1
1;
1;
1
in which = (1 2xt + t2 ) 2 , and
2 F1 (n, c; 1; y)Pn (x)t
n=0
1 1
1
c, c + ;
2 2
2
2c1 2
2 c
=
( + xyt yt ) 2 F1
.
y 2 t2 (x2 1)
(2 + xyt yt2 )2
Also sum the series
3 F2 (n, c, 1
c; 1, 1; y)Pn (x)tn .
n=0
Solution 8. We start with
(1)
n1

Pn
xt
p

=
X
(n + k)!Pn+k (x)tk
k=0
k!n!
In the generating relation

0 F1
t(x 1)
; 1;
2
we replace x by

X
t(x + 1)
Pn (x)tn
=
,
0 F1 ; 1;
2
(n!)2
n=0
xt
yt
and t by
to get
p
p
, p = (1 2xt + t2 ) 2 .

0 F1
yt
; 1;
2p
xt
1
p

0 F1
yt
; 1;
2p
xt
+1
p

=
=
=
=
85
n n
X
(1)n pn1 Pn ( xt
p )y t
(n!)2
n=0
(1)n (n + k)!Pn+k (x)y n tn+k

k!n!(n!)2
n,k=0
X
n
X
n!(1)nk Pn (x)y nk tn
n=0 k=0
X
n
X
n=0 k=0
k![(n k)!]3
(1)k n!y k
Pn (x)tn ,
(k!)3 (n k)!
or

X

n;
yt(x
t
+
p)
yt(x
p)
F
F
;
1;
=
p1 0 F1 ; 1;
0
1
1
2
2p2
2p2
n=0
1, 1;
y Pn (x)tn .
Next let us apply the same process to the known generating relation
c c+1
,
;
2
2
(1 xt)c 2 F1
X
(c)n Pn (x)tn
t2 (x2 1)
.
=
n!
(1 xt)2 n=0
1;
xt
yt
We replace x by
, t by
and note that (1 xt) becomes
p
p
1+
yt(x t)
= p2 [p2 + xyt yt2 ]
p2
while t2 (x2 1) becomes

y 2 t2
p2
"
xt
p
2
#
1 =
y 2 t2 2
y 2 t2 (x2 1)
2
2
[x
2xt
+
t
t
+
2xt
t
]
=
.
p4
p4
We thus obtain
c c+1
,
;
2
p2c1 [p2 + xyt yt2 ]c 2 F1 2
2 2
y t (x 1)
+ xyt yt2 )2
(p2
n n
X
(1)n (c)n pn1 Pn ( xt
p )y t
=
n!
n=0
=
=
=
X
(1)n (c)n (n + k)!Pn+k (x)y n tn+k
k!n!n!
n,k=0
n
nk
X X (1)
(c)nk n!Pn (x)y nk tn
n=0 k=0
X
n
X
n=0 k=0
We have thus obtained
k![(n k)!]2
(1)k (c)k n!y k Pn (x)tn
.
(k!)2 (n k)!
c c+1
,
;
2
p2c1 [p2 +xytyt2 ]c 2 F1 2
n, c;
X
=
2 F1
n=0
1;
y 2 t2 (x2 1)
2
(p + xyt yt2 )2
y Pn (x)tn .
Finally let us sum the series
3 F2 (n, c, 1
c; 1, 1; y)Pn (x)t
n=0
=
=
X
n
X
(1)k n!(c)k (1 c)k Pn (x)y k tn
(k!)3 (n k)!
n=0 k=0
(1) (n + k)!(c)k (1 c)k Pn+k (x)y k tn+k

(k!)3 n!
n,k=0
X X (n + k)!Pn+k (x)tn (1)k (c)k (1 c)k (yt)k

(k!)2
k!n!
k=0 n=0
With the aid of (1) we get
3 F2 (n, c, 1
n=0
k=0
= p1
k
k
pk1 Pk ( xt
p )(1) (c)k (1 c)k (yt)
(k!)2
yt k
X
(c)k (1 c)k Pk ( xt
p )( p )
k=0
=p
c; 1, 1; y)Pn (x)tn
(k!)2
c, 1 c;
r
2 2
2y(xt)t
1+ yt
+ yp2t
1+
p
p2
2 F1
1
c, 1 c;
p+yt
= p1 2 F1
"
2 F1
12xt(1y)+(1y)2 t2
2p
c, 1 c;
1 yt
p p
pyt
1;
1
Problem 9. With = (1 2xt + t2 ) 2 , show that
Pn
1 xt

=
n
X
(1)k Cn,k tk Pk (x).
k=0
p2 +2yt(xt)+y 2 t2
2
c, 1 c;
2 F1
1;
12xt(1y)+t2 (1y)2
2
pn Pn
n=0
1 xt
p
yn
87
2 2 2
= [1 2 1xt
p yp + y p ]
1
= [1 2y(1 xt) + y 2 p2 ] 2
1
= [1 2y + 2xyt + y 2 2xy 2 t + y 2 t2 ] 2
1
= [(1 y)2 + 2xyt(1 y) + y 2 t2 ] 2
1
yt
yt
= (1 y)1 [1 2x( 1y ) + ( 1y )2 ] 2
X
(1)k Pk (x)y k tk
=
(1 y)k+1
k=0
X (1)k Pk (x)(n + k)!tk y n+k

=
n!k!
n,k=0
n
k
X X (1) Pk (x)n!tk y n
=
k!(n k)!
n=0 k=0
n
X
X
(1)k Cn,k Pk (x)tk y n .
=
n=0 k=0
It follows that
pn Pn
1 xt
p

=
n
X
(1)k Cn,k tk Pk (x)
k=0
as desired.
Problem 10. With the aid of the result in Example 7 pg. 31, show that
n, n + 1, ;
(1 + x)1 (1 x)1 Pn (x)dx = 2+1 B(, )3 F2
1 .
1, + ;
Investigate the three special cases = 1, = 1, + = n + 1.


1v
.
Pn (x) = 2 F1 n, n + 1; 1;
2
Then
1
(1+x)1 (1x)1 Pn (x)dx =
(1+x)1 (1x)1
n
X
(n)k (n + 1)k (1 x)k
k=0
2k (k!)2
But, by Ex.7, page ? we have

Z
(1 + x)1 (1 x)+k1 dx = 2++k1 B(, + k) =
Hence
2++k1 ()( + k)
.
( + + k)
dx.
(1 + x)1 (1 x)1 Pn (x)dx
n
X
(n)k (n + 1)k 2++k1 ()( + k)
k=0
+1
2k (k!)2 P ( + + k)
n
()() X (n)k (n + 1)k ()k
( + )
k=0
(k!)2 ( + )k
Therefore
Z
(A)
n, n + 1, ;
(1+x)1 (1x)1 Pn (x)dx = 2+1 B(, )3 F2
1 .
1, + ;
Let use choose = 1 in (A). We get
n, n + 1, ;
2
(1)()
F
(1 x)1 Pn (x)dx =
( + 1)
1
1, 1 + ;
1 .
We now turn to Theorem 30 with a = 1, b = 1 . To see that
3 F2
n, 1 + n, ;
1 + , 1;
(1 )n
1 =
.
(1 + )n
Thus we get
Z
(1 x)1 Pn (x)dx =
(1)
1
2 (1 )n
2 (1 )n
.
(1 + )n
()n+1
Next choose = 1 in equation (A). We thus get

Z
(1 + x)1 Pn (x)dx
n, n + 1, 1;
()(1)
F
= 2
( + 1)
1, +1;
n, n + 1;
1
=
2 F1
+ 1;
2
( + 1)()
=
( + 1 + n)( n)
2 (1)n (1 )n
=
(1 + )n
(1)n 2 (1 )n
=
.
()n+1
Finally, we choose + = n + 1 in (A) and obtain

Z
1
1
(1+x)
1
Hence
(1x)
n, n + 1, n + 1 ;
+ 1 )
F
(n + 1)
1, n + 1;
n ()(n
Pn (x)dx = 2
1 .
n, n + 1 ;
2n ()(1 )(1 )n
1
=
2 F1
n!
1;
2n ()(1 )(1 )n (1)n (1 + 1 1)n
=
,
n!
(1)n
(1 + x)1 (1 x)n Pn (x)dx
in which we used Ex.5, page 119. Therefore we have

Z
(1 + x)1 (1 x)n Pn (x)dx =
(1)n 2n (1 )n (1 )n
.
sin()(n!)2
Problem 11. Obtain from equation (5), page 168, the result
!
r
n
X
n
1+x
n
2
2
(1 + x) Pn
Cn,k Pk (x)
=2
2
k=0
and use it to evaluate the integral

Z
n
2
(1 + x) Pn
1
1+x
2
!
Pm (x)dx.
Solution 11. We put = 2 and cos = x in the known relation

(1)
Pn (cos ) =
sin
sin
n X
n

Cn,k
k=0
sin( )
sin
nk
Pk (cos )
to get

Pn
cos
2
1
2 cos( 2 )
!n
n
X
Cn, Pk (cos ),
k=0
or
r
n
2
(2)
[1 + x] Pn
1+x
2
!
n
= 2 2
n
X
Cn,k Pk (x).
k=0
Then
Z
B=
n
2
(1 + x) Pn
1
1+x
2
!
n
Pm (x)dx = 2 2
n
X
k=0
= 21 2
Cn,k
If n < m, then k < m and = 0. If n m,

B
89
n
X
(1)k (n)k ( 1 )k
2
k!( 32 )k
k=0
1
n, ;
n
1
= 21 2 2 F1
.
3
;
2
Pk (x)Pm (x)dx.
1
Problem 12. Evaluate

1
Z
1 1 n
x Pn2k (x)dx,
2 1
0
and check your result by means of Theorem 65, page 181. Thus show that
Z
xn Pn2k (x)dx =
xn Pn2k (x)dx =
n!
2n k!( 32 )nk
and, equivalently, that

1
xn+2k Pn (x)dx =
(n + 2k)!
2n+2k k!( 32 )n+k

1
1
x Pn2k (x)dx =
2
n
A=
0
xn Pn2k (x)dx.
We have
Z
A=
0
n + 2k, n 2k + 1;
1x
dx
2
x n 2 F1
1;
and we apply Theorem 37 with = n + 1, = 1, t = 1, p = 2, q = 1, a1 =
1
n + 2k, a2 = n 2k + 1, b1 = 1, c = , k = 0s = 1. The result is
2
n + 2k, n 2k + 1, 1;
Z 1
1
xn P n 2k(x)dx = B(n + 1, 1)F

.
2
0
1, n + 2;
By Example 3, page 39
a, 1 a;
2 F1
21c (c)( 12 )
1
=
ca+1 ,
( c+a
)
2
2 )(
2
c;
which we use with a = n + 2k, 1 a = 1 + n 2k, c = n + 2. Then
(n + 1)(1) 2n1 (n + 2)( 12 )
2n2k+3
(n + 2) ( 2k+2
)
0
2 )(
2
3
( 2 )
n!
= n
2 k!(n k + 32 )
n!
= n 3
.
2 k!( 2 )nk
Recall that we found in Theorem 65 that
Z
xn Pn2k (x)dx
[2]
(2n + 4s + 1)Pn2s (x)
n! X
.
x = n
2 s=0
s!( 32 )ns
n
91
We may then write

1
1
=
2
x Pn2k (x)dx
0
xn Pn2k (x)dx
n
[2]
(2n 4s + 1)
n! X
2n+1
R1
1
Pn2k (x)Pn2s (x)dx
s!( 32 )ns
s=0
.
.
Only the term with s = k remains. We get

Z
xn Pn2k (x)dx =
n! 2n 4k + 1
2
n!
.
= n 3
2n+1 k!( 32 )nk 2n 4k + 1
2 k!( 2 )nk
Thus a check on the earlier result.

A simple change from n to (n + 2k) yields
1
xn+2k Pn (x)dx =
(n + 2k)!
.
2n+2k k!( 32 )n+k
Problem 13. Use the formula (5), page 104, to obtain the result
n

t
1
xn (1 x)n dx
=
Q
,
n
2 )n+1
(1
x
2
t
0
where Qn (t) is the function given in (4), page 182.
Z

Z t
xn (t x)n
dx
=
xn (t x)n 1 F0 (n + 1; ; x2 )dx.
2 n+1
0 (1 x )
0
We may use Theorem 37 with = n + 1, = n + 1, p = 1, q = 0, a1 = n + 1, c =
1, k = 2, s = 0 to get
Z
B1 =
n + 1,
B1 = B(n + 1, n + 1)t2n+1 F
n+1 n+2
,
;
2
2
2n + 2 2n + 3
,
;
2
2
t2
,
or
Z
0
n+1 n+2
,
;
2
2
(n + 1)(n + 1) 2n+1
F
=
t
(2n + 2)
3
n+ ;
2

(n!)2 t2n+1 2n ( 1t )n+1 ( 32 )n
1
=
Qn
,
(2n + 1)!
n!
t
xn (t x)n
dx
(1 x2 )n+1
t2
in terms of the !n (t) of page 182. We thus obtain, since n!22n ( 32 )n = (2n + 1)!,
Z
0
xn (t x)n
dx =
(1 x2 )n+1
n

t
1
Qn
.
2
t

n, n;
Pn (x) =
2n ( 12 )n (x
1)
n!
2n;
.
1x

X
n
k
(1)k (n + k)!( 1x
1x
2 )
=
.
Pn (x) = 2 F1 n, n + 1; 1;
2
2
(k!) (n k)!
k=0
Reversing the order of the terms, we get

Pn (x)
=
=
n
X
(1)nk (2n k)!( 1x )nk
2
k![(n k)!]2
k=0
n X

n
2 k
(2n)!(n)k (n)k ( 1x
)
x1
(2n)k k!(n!)2
n, n;
2n ( 21 )n (x 1)n
2
=
.
2 F1
n!
1x
2n;
2
k=0

n, n;
Pn (x) =
2n ( 21 )n (x
n!
+ 1)
F
2n;
.
1+x
Solution 15. Since Pn (x) = (1)n Pn (x), it follows from the result in Exercise 14
that
n, n;
2n ( 21 )n (x + 1)n
2
Pn (x) =
.
2 F1
n!
1+x
2n;
Problem 16. Show that for |t| sufficiently small
(2n + 1)Pn (x)tn = (1 t2 )(1 2xt + t2 ) 2 .
n=0
Solution 16. From

1
(1 2xt + t2 ) 2 =
Pn (x)tn
n=0
we obtain
1
t(1 2xt2 + t4 ) 2 =
Pn (x)t2n+1 .
n=0
We then differentiate both members with respect to t to get
93
(2n + 1)Pn (x)t2n = (1 2xt2 + t4 ) 2 + t(2xt 2t3 )(1 2xt2 + t4 ) 2 .
n=0
Now replace t by t2 :
(2n + 1)Pn (x)tn
= (1 2xt + t2 ) 2 + t(2x 2t)(1 2xt + t2 ) 2
n=0
= (1 2xt + t2 ) 2 (1 2xt + t2 + 2xt 2t2 )

3
= (1 t2 )(1 2xt + t2 ) 2 ,
as desired.
Problem 17. Use Theorem 48, page 137, with c = 1, x replaced by
n =
( 21 )n
to arrive at
n!
(1 x)n = 2n (n!)2
n
X
(1)k (2k + 1)Pk (x)
k=0
(n k)!(n + k + 1)!
1
(1 x) and
2
Solution 17. We know the generating relation
Pn (x)tn
= (1 2xt + t2 ) 2
n=0
= [(1 t)2 2t(x 1)] 2

1

2t(x 1) 2
1
= (1 t)
1
2
(1 t)

1
2t(x
1)
1
= (1 t) 1 F0
; ;
.
2
(1 t)2
In the above, put x = 1 2v to get
1
;
2
1
(1 t) 1 F0
X
4vt
=
Pn (1 2v)tn .
2
(1 t)
n=0
We may now use Theorem 48 with n =
( 12 )n
and c = 1. This yields
n!
vn =
22n n!( 12 )n n! X (1)k (2k + 1)Pk (1 2v)

,
22n ( 12 )n k=0 (n k)!(n + k + 1)!
or
n
(1 x) = 2 (n!)
n
X
(1)k (2k + 1)Pk (x)
k=0
as desired.
(n k)!(n + k + 1)!
Problem 18. Use Theorem 48, page 137, to show that

(1 x)Pn0 (x) + nPn (x)
0
= nPn1 (x) (1 x)Pn1
(x)
n1
n1
X
X
=
Pk (x) 2(1 x)
Pk0 (x)
k=0
n1
X
k=0
(1)
nk+1
(1 + 2k)Pk (x).
k=0
Solution 18. Using the transformation in Exercise 17 above we obtain

v
d
Pn (1 2v) nPn (1 2v)
dv
d
Pn1 (1 2v)
dv
n1
X
d
=
[Pk (1 2v) + 2v Pk (1 2v)]
dv
k=0
n1
X
(1)nk (2k + 1)Pk (1 2v).
=
= nPn1 (1 2v) v
k=0
dy
dy
Since x = 1 2v, we know that v
= (1 x) .
dv
dx
Hence the above relations become
0
(1 x)Pn0 (x) + nPn (x) = nPn1 (x) (1 x)Pn1
(x),
(1 x)Pn0 (x) + nPn (x) =
n1
X
[Pk (x) 2(1 x)Pk0 (x)],
k=0
(1 x)Pn0 (x) + nPn (x) =
n1
X
(1)nk+1 (2k + 1)Pk (x).
k=0
Problem 19. Use Rodrigues formula, page 162, and successive integrations by
parts to derive the orthogonality property for Pn (x) and to show that
Z
Pn2 (x)dx =
2
.
2n + 1
1
Solution 19. We know Pn (x) = n D n (x2 1)n .
2 n!
Then
Z
Pn (x)Pm (x)dx
1
Z 1
1
= n
[D n (x2 1)n ]Pm (x)dx
2 n! 1
Z 1
1
1 n1 2
1
= n
{D
(x 1)n }Pm (x) 1 n
[D n1 (x2 1)n ][DPm (x)]dx
2 n!
2 n! 1
= ...
Z
(1)n 1 0 2
= n
[D (x 1)n ][D n Pm (x)]dx.
2 n! 1
If n > m, D n Pm (x) 0. Also the original integral is symmetric in n and m.

Hence
95
Pn (x)Pm (x)dx = 0, m 6= n.
1
But we also have

Z
Pn2 (x)dx =
Now Pn (x) =
2n ( 21 )n xn
n!
(1)n
2n n!
(x2 1)n [D n Pn (x)]dx.

1
.
2 n
+ n1 , so D n Pn (x) = 2n
Therefore
Z
Pn2 (x)dx =
( 21 )n
n!
(1 x2 )n dx =
( 12 )n
n!
(1 x)n (1 + x)n dx.
By Example 7, page 51, we have

Z
Pn2 (x)dx =
22n+1 ( 21 )n (n + 1)(n + 1)
( 12 )n n+1+n+11
2
B(n + 1, n + 1) =
.
n!
n!(2n + 2)
Hence
Z
Pn2 (x)dx
22n+1 ( 21 )n n!
(2n + 1)!
22n+1 ( 21 )n n!
= 2n 3
2 n!( 2 )n
2 21
=
n + 12
2
.
=
2n + 1
=
Problem 20. Show that the polynomial yn (x) = (n!)1 (1 x2 ) 2 Pn ((1 x2 ) 2 )

has the generating relation

X
1 2 2
e 0 F1 ; t; x t =
yn (x)tn
4
n=0
t
and that Theorem 45, page 133, is applicable to this yn (x). Translate the result
into a property of Pn (x), obtaining equation (7), page 159.
Solution 20. Consider the polynomial
n
(1 x2 ) 2
yn (x) =
Pn
n!
We find that
1 x2

.
yn (x)tn
n=0
P
X
n

(t 1 x2 )n
1
1x2
n!
n=0

1
t2 (1 x2 )
1
1 x2
= et 0 F1
1;
;
t 2 x2
.
4
= et 0 F1
1;
Problem 21. Let the polynomials wn (x) be defined by

ext [t2 (x2 1)] =
wn (x)tn ,
n=0
with
(u) =
n un .
n=0
Show that
(c)n wn (x)tn = (1 xt)c
n=0

(c)2kk
k=0
t2 (x2 1)
(1 xt)2
k
and thus obtain a result parallel to that in Theorem 46, page 134. Apply your
new theorem to Legendre polynomials to derive equation (2), page 164.
Solution 21. (Solution by Leon Hall) Multiplication of series gives
!
!
X
X
X
xn t n
n
2
n 2n
wn (x)t
=
n (x 1) t
n!
n=0
n=0
n=0
n
X
X
(2k + 1 + tx)
nk x2k (x2 1)nk t2n ,
=
(2k
+
1)!
n=0
k=0
and so
n
X
nk (x2 1)nk x2k
w2n (x) =
k=0
and
w2n+1 (x) =
(2k)!
n
X
nk (x2 1)nk x2k+1
k=0
(2k + 1)!
An index shift gives

w2n (x) =
n
X
k (x2 1)k x2n2k
k=0
and
w2n+1 (x) =
(2n 2k)!
X
k (x2 1)k x2n2k+1
k=0
(2n 2k + 1)!
97
Thus
(c)n wn (x)tn
n=0
X
n
X
(c)2n k (x2 1)k
n=0 k=0
(2n 2k)!
x2n2k t2n +
X
n
X
(c)2n+1 k (x2 1)k
n=0 k=0
(2n 2k + 1)!
x2n2k+1 t2n+1 .
If we expand the finite sum, the coefficient of k (x2 1)k for k = 0, 1, . . . , n is
X
(c)2n x2n2k t2n X (c)2n+1 x2n2k+1 t2n+1
+
.
(2n 2k)!
(2n 2k + 1)!
n=k
n=k
Noting that (c)2n+2k = (c + 2k)2n (c)2k and (c)2n+2k+1 = (c + 2k)2n+1 (c)2k we have
#
"
X
X
X (c + 2k)2n (xt)2n X
(c + 2k)2n+1 (xt)2n+1
n
2
k 2k
(c)n wn (x)t =
+
(c)2k k (x 1) t
(2n)!
(2n + 1)!
n=0
n=0
n=0
k=0
where the second factor is equivalent to
X
1
(c + 2k)n (xt)n
=
.
c+2k
n!
(1
xt)
n=0
1
) Now we have
(1 xt)c+2k
2 2
k
X
X
t (x 1)
n
c
(c)n wn (x)t = (1 xt)
(c)2k k
,
(1 xt)2
n=0
(do a Taylor series around 0 for
k=0
which is similar to the result of Theorem 46, pg. 134. For the second part, note
that the formula (4) on pg.165 is

X
Pn (x)tn
1 2 2
xt
= e 0 F1 ; 1, t (x 1) ,
n!
4
n=0
where Pn is the Legendre polynomial of degree n. If we now let
Pn (x)
n!
wn (x) =
and
(u) =
X
k=0
uk
,
22k (k!)2
1
we get k = 2k
, and our result gives
2 (k!)2
2 2
n
X
X
(c)n Pn (x)tn
(c)2n
t (x 1)
.
= (1 xt)c
n!
22n (n!)2 (1 xt)2
n=0
n=0
c c 1
(c)2n
Note that 2n =
+
and that
2
2 n 2 2 n
2 F1
(1 , 2 ; 1; z) = 1 +
X
(1 )n (2 )n z n
.
n!
n!
n=1
c
c 1
t2 (x2 1)
we have
, 2 = + , and z =
2
2 2
(1 xt)2

X
(c)n Pn (x)tn
c c 1
t2 (x2 1)
= (1 xt)c 2 F1
, + ; 1;
n!
2 2 2
(1 xt)2
n=0
Thus, with 1 =
which is equivalent to equation (2) on pg.164.

Problem 1. Use the fact that

exp(2xt t2 ) = exp(2xt x2 t2 ) exp[t2 (x2 + 1)]
to obtain the expansion
n
Hn (x) =
[2]
X
n!Hn2k (1)xn2k (x2 + 1)k
k!(n 2k)!
k=0
Solution 1. From
exp(2xt t2 ) = exp(2xt x2 t2 )exp[t2 (x2 + 1)]
we obtain
n=0
X
Hn (1)(xt)n
n!
n=0
X
(x2 + 1)n t2n
n!
n=0
[2]
X
X
Hn2k (1)(x)n2k (x2 + 1)k tn
k!(n 2k)!
n=0 k=0
It follows that
n
Hn (x) =
[2]
X
n!Hn2k (1)xn2k (x2 + 1)k
k!(n 2k)!
k=0
Problem 2. Use the expansion of xn in a series of Hermite polynomials to show

that
Z
2k
exp(x )x Hn2k (x)dx = 2
Note in particular the special case k = 0.

n
[2]
X
n!Hn2s (x)
x =
.
2n s!(n 2s)!
s=0
n
Then
n!
.
k!
exp(x )x Hn2k (x)dx =
[2]
X
s=0
n!
n s!(n 2k)!
99
ex Hn2k (x)Hn2s (x)dx.
The integrals involved on the right are zero except for s = k. Hence
Z
2
n!
2
ex Hn2k
(x)dx
n
2 k!(n 2k)!
n!2n2k (n 2k)!
=
n
2 k!(n 2k)!
n!
= 2k .
2 k!
For k = 0 the right member becomes n! .

Z
exp(x2 )xn Hn2k (x)dx
Problem 3. Use the integral evaluation in equation (4), page 192, to obtain the
result
Z
exp(x2 )H2k (x)H2n+1 (x)dx =
(1)k+s 22k+2s ( 12 )k ( 32 )s
.
2s + 1 2k

Z
ex Hn (x)Hm (x)dx = 0, m 6= n.
If m and n are both odd or both even, the integrand above is an even function of
x and we obtain
ex Hn (x)Hm (x)dx = 0
2
0
for m n mod 2.
Now consider
Z
ex H2n (x)H2s+1 (x)dx.
By equation (4), page 331, we get
ex H2n (x)H2s+1 (x)dx
2(2n 2s 1)
0
h
i
2
0
(x) 202n (x)H2s+1 (x)
= ex H2n (x)H2s+1
0
0
0
(0) + H2n
(0)H2s+1 (0)
= H2n (0)H2s+1
0
= H2n (0)H2s+1 (0).
0
(0) from page 323, we get
Using the values of H2n (0) and H2s+1
Z
0
ex H2n (x)H2s+1 (x)dx
(1)n 22n ( 21 )n (1)s 22s+1 ( 32 )s

2(2n 2s 1)
(1)n+s 22n+2s ( 12 )n ( 23 )s
=
.
2s + 1 2n
=
100SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA
Problem 4. By evaluating the integral on the right, using equation (2), page 187,
ans term-by-term integration, show that
(A)
Pn (x) =
n!
exp(t2 )tn Hn (xt)dt,
which is Curzons integral for Pn (x), equation (4), page 191.

Solution 4. Consider the integral
2
n!
2
=
et tn Hn (xt)dt
et tn
[2]
X
(1)k (2xt)n2k dt
k=0
[n
2]
X (1)k (2k)n2k
k!(n 2k)!
k=0
[n
2]
k!(n 2k)!
Z
1
1
e nk 2 d
0
X (1)k (2x)n2k (n k + 1 )
2
k!(n 2k)!
( 12 )
k=0
[n
2]
X (1)k ( 1 )nk (2x)n2k

2
k!(n 2k)!
k=0
Hence
2
n!
et tn Hn (xt)dt = Pn (x).
Problem 5. Let vn (x) denote the right member of equation (A) of Exercise 4.
Prove (A) by showing that
vn (x)y n = (1 2xy + y 2 ) 2 .
n=0
Solution 5. Put vn (x) =
n!
Then
vn (x)y
n=0
et tn Hn (xt)dt.
Z
n
n
2 X t Hn (xt)t
2
=
et
dt
n!
0
n=0
Z
2
2 2
2
=
et e2(xt)yty t dt
Z0
2
2
2
=
et [12xy+y ] dt.
0
p
Use t 1 2xy + y 2 = to get
2
vn (x)y n = p
1 2xy + y 2
n=0
Hence vn (x) = Pn (x).
Z
0
e d = (1 2xy + y 2 ) 2 =
X
n=0
Pn (x)y n .
101
Problem 6. Evaluate the integral on the right in

Z
x
n+1
2
exp(t2 )tn+1 Pn
(B)
Hn (x) = 2
exp(x )
dt
t
x
by using
n
(2t) Pn
x
t
[2]
X
n!(x2 t2 )k (2x)n2k
(k!)2 (n 2k)!
k=0
derived from equation (1), page 164, and term-by-term integration to prove the
validity of (B), which is equation (5), page 191.
Solution 6. We know
n
Pn (x) =
[2]
X
n!(x2 1)k xn2k
22k (k!)2 (n 2k)!
k=0
from (1), page 280. Then

n
(2t) Pn
x
t
[2]
X
n!(x2 t2 )k (2x)n2k
(k!)2 (n 2k)!
k=0
which leads to the result
n+1 x2
t2 n+1
Pn
x
t
dt =
[2]
X
k=0
n!ex
2
(k!) (n 2k)!
et (x2 t2 )k (2x)n2k (2tdt).
Put t x2 = . Then
n+1 x2
t2 n+1
Pn
[2]
Z
X
n!(2x)n2k
e (1)k k d
dt =
t
(k!)2 (n 2k)! 0
x
k=0
[n
2]
X n!(1)k (2x)n2k k!
k=0
[n
2]
(k!)2 (n 2k)!
X n!(1)k (2x)n2k
k=0
k!(n 2k)!
Hence
2
n+1 x2
et tn+1 Pn
x
t
dt = Hn (x).
Problem 7. Use the Rodrigues formula

exp(x2 )Hn (x) = (1)n Dn exp(x2 ); D =
an iteration integration by parts to show that

Z
0
exp(x2 )Hn (x)Hm (x)dx =
2n n!
d
dx
; m 6= n
;m = n

2
Solution 7. We know Hn (x) = (1)n ex D n ex .

Then
Z
[D n ex ]Hm (x)dx
Z
hn
o
i
2
= (1)n D n1 ex Hm (x)
+ (1)n+1
ex Hn (x)Hm (x)dx
= (1)n
= ...
= (1)2n
ex [D n Hm (x)]dx.
Hence
Z
ex Hn (x)Hm (x)dx = 0, m 6= n
and, since Hn (x) = 2n xn + n2 (x) (where n2 (x) denotes a polynomial of

degree (n 2)),
Z
Z
2
x2 2
e
Hn (x)dx =
ex [D n Hn (x)]dx
2
n
= 2 n!
ex dx
= 2n n! .

1
H2n (x) = (1)n 22n n!Ln 2 (x2 ),

1
H2n+1 (x) = (1)n 22n+1 n!xLn2 (x2 ).

n
Solution 1. From Hn (x) =

H2n (x)
[2]
X
(1)k n!(2x)n2k
k=0
n
X
=
=
k=0
n
X
k=0
n
X
k!(n 2k)!
we get
(1)k (2n)!(2x)2n2k
k!(2n 2k)!
(1)nk (2n)!(2x)2k
(2k)!(n k)!
(1)n (2n)!(n)k x2k

k!( 12 )k n!
k=0

1
1 2
= (1)n 22k
F
n;
;
x
1 1
2
2 n
1
1
n!
(
)
= (1)n 22n
Ln 2 (x2 ).
2 n ( 12 )n
Hence
[D n1 ex ][DHm (x)]dx
( 12 )
H2n (x) = (1)n 22n n!Ln
103
(x2 ).
In the same way
H2n+1 (x)
=
=
n
X
(1)k (2n + 1)!(2x)2n+12k
k=0
n
X
k=0
n
X
k!(2n + 1 2k)!
nk
(1)
(2n + 1)!(2x)2k+1
(n k)!(2k + 1)!
(1)n (n)k 22n n!( 23 )n 22k+1 x2k+1

n!22k k!( 32 )k
k=0

3
= (1)n 22n ( 32 )n (2x)1 F1 n; ; x2
2
( 12 ) 2
n 2n
= (1) 2 n!(2x)Ln (x ).
=
Hence
(1)
H2n+1 (x) = (1)n 22n+1 n!xLn 2 (x2 ).

Problem 2. Use Theorem 65, page 181, and the method of Section 118 above to
derive the result
1
1
(n k), (n k 1);
2
2
Pn() (x)
n
X
=
2 F3
3
k=0
1
1
+ k, (1 + + k), (2 + + k);
2
2
2
[n]
2
X
(2x)n
(2n 4k + 1)Pn2k (x)
=
.
n!
k!( 32 )nk
k=0
Then
1
4
(1)k (1 + )n (2k + 1)Pk (x)
.
2k (n k)!( 3 ) (1 + )
k
2 k
Ln() (x)tn
n=0
X
n
X
(1)s (1 + )n xs tn
n=0 s=0
s!(n s)!(1 + )s
(1)s (1 + )n+s xs tn+s

s!n!(1 + )s
n,s=0
s
[2]
X
X
(1)s (1 + )n+s (2s 4k + 1)Ps2k (x)tn+s
=
2s n!(1 + )s k!( 23 )sk
n,s=0 k=0
X
(1)s (1 + )n+s+2k (2s + 1)Ps (x)tn+s+2k
=
2s+2k k!n!(1 + )s+2k ( 23 )s+k
n,k,s=0
k
n+k+2s
change letters X (1) (1 + )n+k+2s (2k + 1)Pk (x)t
=
3
2k+2s s!n!(1 + )k+2s ( 2 )k+s
n,k,s=0
n
[2]
X
X
(1)k (1 + )n+k (2k + 1)Pk (x)tn+k
=
s!(n 2s)!(1 + )k+2s ( 32 )k+s 2k+2s
n,k=0 s=0
n
[2]
X
X
(n)2s 22s
(1)k (1 + )n+k (2k + 1)Pk (x)tn+k
s!(1 + + k)2s ( 23 + k)s
(1 + )k ( 32 )k 2k n!
n,k=0 s=0
n n1
,
;
2
2
k
n+k
X
1
(1) (1 + )n+k (2k + 1)Pk (x)t
=
.
2 F3
3
4
2k (1 + )k ( 2 )k n!
3
n,k=0
1++k 2++k
+ k,
,
;
2
2
2
=
We thus get
Ln() (x)tn =
n=0
X
n
X
nk nk1
,
;
2
2
F
2 3
3
n=0 k=0
1++k 2++k
+ k,
,
;
2
2
2
1
4
(1)k (1 + )n (2k + 1)Pk (x)tn
2k (1 + )k ( 32 )k (n k)!
Hence
nk nk1
,
;
2
2
1
1
(n k), (n k 1);
2
2
n
X
L()
2 F2
n (x) =
1
k=0
1
(1 + + k), (2 + + k);
2
2
n
X
(1) (1 + )n (2k + 1)Pk (x) .

=
2 F3
3
k
4
3
2 (1 + )k ( 2 )k (n k)!
k=0
1++k 2++k
+ k,
,
;
2
2
2
Problem 3. Use formula (4), page 194, and the method of Section 118 to derive
the result
Ln() (x)
1
4
(1)k (1 + )n Hk (x)
k!(n k)!2k (1 + )k .
105
Solution 3. From
[n]
2
X
(2x)n
Hn2s (x)
=
n!
s!(n 2s)!
s=0
we obtain
Ln() (x)tn
n=0
X
n
X
(1)k (1 + )n xk tn
n=0 k=0
X
n,k=0
k!(n k)!(1 + )k
(1)k (1 + )n+k xk tn+k

k!n!(1 + )k
k
[2]
X
X
(1)k (1 + )n+k Hk2s (x)tn+k
=
n!(1 + )k 2k s!(k 2s)!
n,k=0 s=0
X
(1)k+2s (1 + )n+k+2s Hk (x)tn+k+2s
=
n!(1 + )k+2s 2k+2s s!k!
n,k,s=0
n
[2]
X X (1)k (1 + )n+k Hk (x)tn+k

.
s!(n 2s)!(1 + )k+2s 2k+2s k!
s=0
n,k=0
Thus we obtain
X
n=0
Ln() (x)tn
[2]
X
X
(n)2s 22s (1)k (1 + )n+k Hk (x)tn+k

s!(1 + + k)2s
n!(1 + )k 2k k!
n,k=0 s=0
n n1
,
;
2
2
k
n+k
X
1
(1) (1 + )n+k Hk (x)t
=
F
2 2
k
k!n!2 (1 + )k
4
1++k 2++k
n,k=0
,
;
2
2
nk nk1
2
2
n
X
k
n
X
(1) (1 + )n Hk (x)t .
=
2 F2
k
4
1++k 2++k
2 k!(n k)!(1 + )k
n=0 k=0
,
;
2
2
=
Problem 4. Use the results in Section 56, page 102, to show that
Z
Ln [x(t x)]dx =
0
(1)n H2n+1 ( 2t )
.
22n ( 32 )n
Solution 4. We use Theorem 37 to evaluate

Z
0
Ln [x(t x)]dx =
t
1 F1 (n; 1; x(t
x))dx.
In Theorem 37 we use = 1, = 1, p = 1, q = 1, a1 = n, b1 = 1, k = 1, s =
1, c = 1 to get
Ln [x(t x)]dx
= B(1, 1)t3 F3
2 3
1, , ;
2 2

3 t2
= t1 F1 n; ;
2 4

1 2
n!
( 12 )
= 3 Ln
t .
4
( 2 )n
n, 1, 1;
t2
4
By Exercise 1 above we have

(1)
Ln 2
1 2
t
4

=
(1)n H2n+1 ( 12 t)
.
22n tn!
Hence
Z
Ln [x(t x)]dx =
(1)n
H2n+1
22n ( 32 )n

1
t .
2
Problem 5. Use the results in Section 56, page 102, to show that
Z
0
p

H2n ( x(t x))dx
1
1 2
n
2n
p
= (1) 2
Ln
t .
2 n
4
x(t x)
Solution 5. Since
H2n (x) = (1)n 22n

1 2
1
F
;
x
,
n;
1 1
2 n
2
by Exercise 1, we may write

t

Z t
p
p
1
n 2n 1
21
21
H2n ( x(t x)) x(t x)dx = (1) 2 ( )n
x (tx) 1 F1 n; ; x(t x) dx.
2
2
0
We then apply Theorem 37 with =

1
, c = 1, k = 1, s = 1, to get
2
1
1
, = , p = 1, q = 1, a1 = n, b1 =
2
2
1 1
n, , ;

2 2
1 1 0
n 2n 1
= (1) 2 ( 2 )n B
,
t 3 F3
2 2
1 1 2
, , ;
2 2 2
n;
( 12 )( 12 )
t2
n 2n 1
= (1) 2 ( 2 )n
1 F1
(1)
4
1;

1
= (1)n 22n ( 12 )n Ln
t2 .
4
Z
0
p
H2n ( x(t x))
p
dx
x(t x)
t2
4
107
Problem 6. Show that if m is a non-negative integer and is not a negative

integer,
()
(+2m)
n
X
(m)k Lk (x)Lnk (x)
(1 + )n
.
(1 + )k
(1 + 12 + 12 m)n k=0
L()
n (x) =
Solution 6. On page 366 we had

Ln() (x) =
(1)
()
(2c2)
n
(x)
(1 + )n X (1 + c)k Lk (x)Lnk
.
(c)n
(1 + )k
k=0
1
1
1
1
In (1) choose c = 1 + + m. Then 2c 2 = m and 1 + c = m.
2
2
2
2
Hence
Ln() (x) =
()
(m
n
X
( m
(1 + )n
2 )k Lk (x)Lnk (x)
.
(1 + )k
(1 + 12 + 21 m)n k=0
Problem 7. Show that if m is a non-negative integer and is not a negative

integer,
L()
n (x) =
()
(+2m)
(1 + )n (1 + )m X (m)k Lk (x)Lnk (x)

.
(1 + )m+n
(1 + )k
k=0
Solution 7. In equation (1) of Exercise 6 above choose c = 1 + + m. Thus we

get
()
Ln (x)
=
=
()
(+2m)
n
(x)
(1 + )n X (m)k Lk (x)Lnk
(1 + + m)n
(1 + )k
k=0
()
(+2m)
n
X
(m)k Lk (x)Lnk
(x)
(1 + )n (1 + )m
(1 + )m+n
(1 + )k
k=0
Problem 8. Use integration by parts and equation (2), page 202, to show that
Z
()
ey Ln() (y)dy = ex [Ln() (x) Ln1 (x)].
x
Solution 8. From (2), page 350, we get

()
Ln() (x) = DLn() (x) DLn+1 (x).

Now put
Z
An =
and integrate by parts with u = e
()
()
v = Ln (x) Ln+1 (x) to get
h
An = e
x
y
ey Ln() (y)dy
()
and dv = Ln (y)dy du = ey dy and
n
oi Z
()
()
Ln (y) Ln+1 (y)
+
x
h
i
()
ey Ln() (y) Ln+1 (y) dy.
h
i
()
An = ex Ln+1 (x) Ln() (x) + An An+1 .
Hence
h
i
()
An+1 = ex Ln+1 (x) Ln() (x) ,
so that, by a shift of index, we get
Z
An =
h
i
()
ey Ln() (y)dy = ex Ln() (x) Ln1 (x) .

Z
x (t x)1 Ln() (x)dx =
(1 + )() (1 + )n t+ (+)
L
(t).
(1 + + ) (1 + + )n n
Solution 9. Let us use Theorem 37 to evaluate

t
x (t x)
(1 + )n
=
n!
Ln() (x)dx
x (t x)1 1 F1 (n; 1 + ; x)dx.
In Theorem 37 use = + 1, = , p = 1, q = 1, a1 = n, b1 = 1 + , c =
1, k = 1, s = 0 to arrive at
x (t x)1 Ln() (x)dx
(1 + )n
B( + 1, )t+ 2 F2
n!
n,
+1
;
1
1 + , + + 1;
(1 + )n (1 + |alpha)() +
=
t
1 F1 (n; + + 1; t)
n!
( + + 1)
(1 + + n) (1 + )()t+
n!
(+)
(t)
=
Ln
n!(1 + )
(1 + + ) (1 + + )n
(1 + )n ((1 + )() + (+)
=
(t).
t
Ln
(1 + + )n (1 + + )
Problem 10. Show that the Laplace transform of Ln (t) is

Z
st
e
0
1
Ln (t)dt =
s

n
1
1
.
2
Solution 10.
Z
0
est Ln (t)dt
=
0
n
X
est
n
X
(1)k n!tk dt
(k!)2 (n k)!
k=0
(1)k n!
=
k!(n k)!sk+1
k=0

n
1
1
=
1
.
s
s
109
Problem 11. Show by the convolution theorem for Laplace transforms, or otherwise, that
t
Lm+n (x)dx = Lm+n (t) Lm+n+1 (t).
Ln (t x)Lm (x)dx =
0

n
1
1
Solution 11. We know that L
1
= Ln (t). (where L denotes
s
s
Laplace transform and Ln denotes the Laguerre polynomial).
Then by the convolution theorem,

n
m
Z t
1
1
1
1
Ln (t x)Lm (x)dx =
1
1
L
s"
s
s
s
0

n+m #
1
1 1
.
1
=
s s
s
1
Hence
Z
( "
n+m #)
1
1
1
1
.
Ln (t x)Lm (x)dx = L 1
s s
s
But
L
(
n+m )
1
1
1
= Ln+m (t).
s
s
Hence
Z
Ln (t x)Lm (x)dx =
Ln+m ()d.
But, by (2), page 350,

Ln (x) = DLn (x) DLn+1 (x),
so that
Z
Ln (t x)Lm (x)dx
= [Ln+m () Ln+m+1 ()]0

= Ln+m (t) Ln+m+1 (t).
Problem 12. Evaluate the integral

Z
2
x ex [L()
n (x)] dx
0
of (7), page 206, by the following method. From (4), Section113, show that
Z
X
n=0
x e
2
2n
[L()
n (x)] dxt

x(1 + t)
dx
= (1 t)
x exp
1t
0
= (1 t2 )1 (1 + )
X
(1 + + n)t2n
=
.
n!
n=0
22

(1 t)1 exp
xt
1t

=
Ln() (x)tn .
n=0
Then
(1 t)22 exp
2xt
1t

=
X
n
X
()
()
Lk (x)Lnk (x)tn .
n=0 k=0
()
Then, because of the orthogonality property of Ln (x) we get
x e
0
()
()
Lk (x)Lnk (x)tn dx
= (1t)
22
Z
0
n=0 k=0

2xt
dt,
x exp x
1t
or
Z
X
Z
h
i2
x ex Ln() (x) dxt2n = (1 t)22
n=0
x exp
In the last integral put x

Z

x(1 + t)
dx.
1t
1+t
= and thus obtain
1t

+1 Z

+1
x(1 + t)
1t
1t

x exp
dx =
e d = (1 + )
.
1t
1+t
1+t
0
Therefore we have
Z
X
n=0
h
i2
x ex Ln() (x) dxt2n
= (1 t)1 (1 + t)1 (1 + )
= (1 + )(1 t2 )1
X
(1 + )n (1 + )t2n
=
n!
n=0
X
(1 + + n)t2n
=
.
n!
n=0
Hence
Z
0
h
i2
(1 + + n)
x ex Ln() (x) dx =
.
n!
Problem 1. Prove Theorem A (Sheffer): If n (x) is of Sheffer A-type zero,

gn (m, x) = Dm n+m (x) is also of Sheffer A-type zero and belongs to the same
operator as does n (x).
111
Solution 1. Let n (x) be of Sheffer A-type zero. Put

gn (m, x) = D m n+m (x); D
d
.
dx
We know there exists an A, , and H such that

(1)
A(t)(xH(t))
=
=
n (x)tn
n=0
n+m (x)tn+m .
n=m
Now the n (x) form a simple set. Hence D m k (x) = 0 for k < m. Therefore,
using the operator D m on (1) we obtain
A(t)[H(t)]m (m) (xH(t)) =
D m n+m (x)tn+m .
n=0
Since H(0) = 0, we may write

m
(2)
H(t)
A(t)
t
(m)
(xH(t)) =
D m n+m (x)tn ,
n=0
from which it follows at once (Theorem 72) that D m n+m (x) are polynomials
of Sheffer A-type zero an that they belong to the same operator as do the n (x),
because the H(t) in (1) and (2) are the same function.
Problem 2. Prove Theorem B: If n (x) is of Sheffer A-type zero,
n (x) = n (x)
"m
Y
#1
(1 + i )n
i=1
is of Sheffer A-type m.
Solution 2. We are given that n (x) is of Sheffer A-type zero and wish to consider
n (x) =
n (x)
m
Y
(1 + i )n
i=1
Let n (x) belong to the operator J(D). By Theorem 74, there exists k , vk such
that
n1
X
(k + xvk )D k+1 n (x) = nn (x).
k=0
Hence there exists operator

B = B(x, D) =
(k + xvk )D k+1 = B1 (D) + xB2 (D)
k=0
such that
Bn (x) = nn (x).
Then
(B = i )n (x) = (n + i )N (x),
so that
(
J(D)
m
Y
)
[B(x, D) + i ] n (x) = J(D)
i=1
n (x)
m
Y
(1 + i )n1
i=1
n1 (x)
m
Y
= n1 (x).
(1 + i )n1
i=1
Since n (x) belongs to one operator whose coefficients have maximum degree m,
n (x) is of Sheffer A-type m and n (x) belongs to the operator
J1 (x, D) = J(D)
m
Y
[i + B(x, D)].
i=1
Two operators arise here. Above implies commutativity of the operators [i +

B(x, D)] and it implies that J1 is a proper operator, that is transforms every polynomial into one of degree one lower than the original.
Let us now prove the desired commutativity property. Let
i = i +
k D k+1 + x
vk D k+1 .
k=0
k=0
We shall show that 1 2 = 2 1 . Since

D m+1 [+k D k+1 +xvk D k+1 ] = D m+1 +k D m+k+2 +xvk D m+k+2 +(m+1)vk B m+k+1 ,
we obtain
"
1 2
= 1 +
m D m+1 + x
m=0
= 1 2 + 1
#"
vm D m+1
2 +
m=0
k D
k+1
+ 1 x
k=0
k=0
k D k+1 + x
k=0
vk D
k+1
+ 2
X
k=0
m D
m+1
m=0
+ x2
#
vk D k+1
"
vm D
m=0
which is symmetric in 1 and 2 . Hence 1 2 = 2 1 .

We shall now prove a theorem of value in discussing what operators transform
polynomials of degree n into polynomials of degree (n 1).
Theorem : If the operator J(x, D) is such that a simple set of polynomials n (x)
belongs to it in the Sheffer sense,
(1)
J(x, D)n (x) = n1 (x), J(x, D)0 (x) = 0,
then J(x, D) transforms every polynomial of degree precisely n into a polynomial
of degree precisely n 1.
Proof : Let gn (x) be of degree exactly n. Then we know there exists the expansion
gn (x) =
n
X
A(k, n)nk (x), A(0, n) 6= 0,
k=0
merely because the n (x) form a simple set. Then, because of (1),
m+1
X
m=0
(m + xvm )D m
J(x, D)gn (x) =
(2)
n1
X
113
A(k, n)n1k (x),
k=0
and the right member of (2) is of degree exactly (n 1).

Hn (x)
n!
is of Sheffer A-type zero, obtain the associated functions J(t), H(t), A(t), and draw
what conclusions you can from Theorems 73-76.
Solution 3. (Solution by Leon Hall) From the defining relation for the Hermite
polynomials (p.187) we have
X
Hn (x)tn
n!
n=0
= exp(2xt t2 )
= exp(t2 ) exp(2xt).
Thus by Theorem 72, n (x) is a Sheffer A-type zero with A(t) = exp(t2 ) and
t
H(t) = 2t. Because J(x) is the inverse of H(t) we get J(t) = .
2
Hn (x)
The operator J may be found directly as follows. Denote
by n (x). Then
n!
we need (see p.189 for the first few Hermite polynomials):
T0 (x)D1 (x) = 0 (x)
T0 (x)D(2x) = 1.
So,
T0 (x) =
1
.
2
Continuing,
T1 (x)D2 2 (x) = 1 (x) T0 (x)D2 (x)
1
4T1 (x) = 2x 4x = 0.
2
So, T1 (x) = 0. In general we need
Tn (x)D
n+1
n+1 (x) = n (x) T0 (x)Dn+1 (x)
n1
X
k=1
Because
Hn0 (x)
= 2nHn1 (x), we have

0 (x) =
2nHn1 (x)
= 2n1 (x)
n!
(also see Theorem 76) and so

T0 (x)Dn+1 (x) = n (x)
and
Dk+1 n+1 (x) = 2k+1 nk (x).
Tk (x)Dk+1 n+1 (x).
This makes, for n 2,

Tn (x)2n+1
= n (x) n (x)
n1
X
2k+1 Tk (x)nk (x)
k=1
n1
X
2k+1 Tk (x)nk (x).
k=1
Because we have T1 (x) = 0, this makes T2 (x) = 0, which means T3 (x) = 0, and so
on. Therefore, from Theorem 70,
X
1
J(x, D) =
Tk (x)Dk+1 = D.
2
k=0
From the formulas for A(t) and H(t) it is easily seen that, in Theorem 73,

0
; k 6= 1
ak =
2 ; k = 1
and
0 ; k = 1, 2, . . .
2 ; k = 0.
This makes equation (11) from Theorem 73, for n 2,
1
(xD D2 )n (x) = nn (x)
2
or
00n (x) 2x0n (x) + 2nn (x) = 0,
which is Hermites differential equation. Hermites differential equation is also a
1
result of Theorem 74 using k = 0 except 1 = and v0 = 1, vk = 0; k 1.
2
Equation (17) in Theorem 75, using the k and k values we found before, becomes
k =
2xn1 (x) 2n2 (x) = nn (x)

and this can be written as
2x
Hn (x)
Hn1 (x) 2Hn2 (x)
=
(n 1)!
(n 2)!
(n 1)!
or
1
[2xHn1 (x) 2(n 1)Hn2 (x) = Hn (x)] ,
(n 1)!
where the equation in brackets is the pure recurrence relation for Hermite polynomials.
()
Problem 4. Show that Ln (x) is of Sheffer A-type zero, and proceed as in Exercise 3.
Solution 4. (Solution by Leon Hall) A generating function for the Laguerre polynomials (see (4), p.202) is
X

1
xt
exp
=
Ln() (x)tn ,
(1 t)1+
1t
n=0
()
so Theorem 72 says Ln (x) is of Sheffer A-type zero with

A(t) = (1 t)1
115
and
t
.
1t
An easy inverse calculation for H(t) yields
H(t) =
J(t) =
X
t
=
tn+1 .
1t
n=0
From
1+ X
A0 (t)
=
=
k tk
A(t)
1t
k=0
we have k = 1 + for all k, and from

H 0 (t) =
(k + 1)tk
k=0
we have k = (k + 1). Thus, (11) in Theorem 73 is

n1
X
()
(1 + x(k + 1))J k+1 (D)L()
n (x) = nLn (x).
()
k=0
By definition
()
J(D)Ln (x)
n1
X
()
= Ln1 (x) so
J k+1 (D)L()
n (x) =
n1
X
()
Ln1k (x) =
()
Lk (x),
k=0
k=0
k=0
n1
X
and so by Section 114, equation (3),

n1
X
()
Lk (x) = DL()
n (x)
k=0
(see Theorem 76). The left side of () then becomes

n1
X
(1 + x kx)J k+1 (D)Ln() (x)
= (1 + x)
k=0
n1
X
J k+1 (D)L()
n (x) x
k=0
= (1 +
n1
X
k=0
()
x)DLn (x)
()
= (1 + x)DLn (x) x
n1
X
k=0
n1
X
kJ
k+1
()
Rewrite the last sum as

n1
n2
n3
1
X ()
X ()
X ()
X
()
()
kLn1k (x) =
Lk (x) +
Lk (x) + . . . +
Lk (x) + L0 (x)
k=0
k=0
()
()
k=0
()
()
= DLn1 (x) DLn2 (x) . . . DL2 (x) DL1 (x)

n1
X ()
= D
Lk (x)
k=0
()
= D2 Ln (x).
Thus, () becomes, finally
2 ()
()
(1 + x)DL()
n (x) xD Ln (x) = nLn (x)
(D)Ln() (x)
kLn1k (x).
k=1
k=1
kJ k+1 (D)L()
n (x)
or
()
()
xD2 L()
n (x) + (1 + x)DLn (x) + nLn (x) = 0,
which is Laguerres differential equation. Equation (17) in Theorem 75 becomes,

after an index shift,
n
X
()
(1 + kx)Lnk (x) = nL()

n (x).
k=1
In Theorem 74, with u = J(t) =
t
, we get
1t
uA0 (u)
= (1 + )t
A(u)
and
uH 0 (u) = t t2 ,
making 0 = (1 + ), v0 = 1, v1 = 1, and the rest of the k and vk zero.
Then, using these values in equation (14) of Theorem 74 we again get Laguerres
differential equation.
Problem 5. Show that the Newtonian polynomials
(1)n (x)n
n!
are of Sheffer A-type zero, and proceed as in Exercise 3.
Nn (x) =
Solution 5. (Solution by Leon Hall) A standard Taylor series expansion about

t = 0 shows that
X
X
(1)n (x)n n
Nn (x)tn =
t = (1 + t)x = ex log(1+t) ,
n!
n=0
n=0
so in Theorem 72 we have A(t) = 1 and
H(t) = log(1 + t) =
X
(1)k
k=0
k+1
tk+1 .
Because J(t) is the inverse of H(t),

J(t) = et 1 =
X
k=0
tn+1
.
(n + 1)!
A(t) constant means all the k = 0 in Theorems 73 and 75 and that all the k = 0
in Theorem 74. Because
H 0 (t) =
X
1
=
(1)k tk ,
1+t
k=0
we have k = (1) in Theorems 73 and 75 and, for u = J(t),

uH 0 (u) = 1 et =
X
(1)k k+1
t
(k + 1)!
k=0
we get vk =
117
(1)k
in Theorem 74. Note that also if Nn (x) Nn (x+1)Nn (x),
(k + 1)!
then
(1)n
(1)n
((x + 1))n
(x)n
n! n1
n!

x+1 x+1n
(1)
(x)n1
=
(n 1)!
n
n
= Nn1 (x).
Nn (x)
Theorem 73 now gives

n1
X
(1)k xJ k+1 (D)Nn (x) = nNn (x)
k=0
or using Theorem 75
x
n1
X
(1)k Nn1k (x) = nNn (x).
k=0
And because k+1 Nn (x) = Nn1k (x),

x
n1
X
(1)k k+1 Nn (x) = nNn (x).
k=0
Theorem 74 yields
x
n1
X
k=0
(1)k k+1
D
Nn (x) = nNn (x).
(k + 1)!
Finally, from Theorem 76, recalling from H(t) that hk =

n1
X
k=0
(1)k
, we have
k+1
(1)k
Nn1k (x) = DNn (x)
k+1
or
1
1
(1)n1
Nn0 (x) = Nn1 (x) Nn2 (x) + Nn3 (x) . . . +
N0 (x).
2
3
n
With some work, this last result could also be obtained by logarithmic differentiation.
()
Ln (x)
(1 + )n
is of Sheffer A-type unity but that with chosen to be
n (x) =
= D( + )
the polynomials n (x) are of -type zero.
()
n (x) =
()
Ln (x)
.
(1 + )n
Since Ln (x) is of Sheffer A-type zero (Exercise 4), it follows by Exercise 2

that n (x) is of Sheffer A-type one.
Now
()
X
Ln (x)tn
= et 0 F1 (; 1 + ; xt),
(1 + ) n
n=0
so that, by Theorem 79, n (x) is of -type zero with

= D( + 1 + 1 a) = D( + ).
Problem 7. Determine the Sheffer operator associated wit hthe set
Ln (x)
,
(n!)2
and thus show that n (x) is of Sheffer A-type 2.
n (x) =
Ln (x)
. Since Ln (x) is of Sheffer A-type zero (by
(n!)2
Exercise 4), we may use Exercise 2 to conclude that n (x) is of Sheffer A-type 2.
We now wish to find the operator to which n (x) belongs.
Let
Solution 7. Consdier n (x) =
J(x, D) =
Tk (x)D k+1 .
k=0
We first proceed by brute strength methods. We know that

L0 (x) = 1,
L1 (x) = 1 x,
1
L2 (x) = 1 2x + x2 ,
2
3
1
L3 (x) = 1 3x + x2 x3 ,
2
6
1
2
L4 (x) = 1 4x + 3x2 x3 + x4 ,
3
24
then J(x, D)n (x) = n1 (x) requires that
T0 (x)D
1x
1
=
; T0 (x)(1) = 1, T0 (x) = 1.
(1!)2
(0!)2
Then also
[T0 (x)D + T1 (x)D 2 ]
1
1
L2 (x) =
L1 (x),
(2!)2
(1!)2
or

1 2
[D + T1 (x)D ] 1 2x + x = 4(1 x)
2
2
(2 + x) + T1 (x)(1) = 4 4x; T1 (x) = 2 3x.

Next we have
119

3
1
1
[D + (2 3x)D 2 + T2 (x)D 3 ] 1 3x + x2 x3 = 9 1 2x + x2 .
2
6
2
or

1
3 + 3x x2 + (2 3x)(3 x) + T2 (x)(1) = 9 18x + 9x2 ,
2
or
9
1
3 3x + x2 + 6 11x + 3x2 T2 (x) = 9 18x + x2 ,
2
2
T2 (x) = 4x + x2 ; T2 (x) = 4x x2
Then

1
3
1
2
[D+(23x)D 2 +(4xx2 )D 3 +T4 (x)D 4 ] 1 4x + 3x2 x3 + x4 = 16 1 3x + x2 x3 ,
3
24
2
6
or

1
1
8
4 + 6x 2x2 + x3 +(23x) 6 4x + x2 +(4xx2 )(4+x)+T4 (x) = 1648x+24x2 x3 .
6
2
3
Thus we get
x3
(1 + 9 + 6 16) = x2 ; T4 (x) = x2 .
6
Let us consider the effect of the operator
T4 (x) = x2 +
J(x, D) = D + (2 3x)D 2 + (4x x2 )D 3 + x2 D 4

Ln (x)
upon n (x) =
.
(n!)2
Problem 8. Prove that if we know a generating function
y(x, t) =
n (x)tn
n=0
or the simple set of polynomials n (x) belonging to a Sheffer operator J(x, D),
no matter what the A-type of n (x), we can obtain a generating fuction (sum the
series)
n (x)tn
n=0
for any other polynomial n (x) belonging to the same operator J(x, D).
Solution 8. We are given that n (x) is a simple set of polynomials and
(1)
y(xt) =
X
n=0
n (x)tn .
Let n (x) belong to the operator J(x, D) in the Sheffer sense. Let n (x) belong
to the same operator J(x, D). Then, by Theorem 71, there exist constants bk such
that
(2)
n (x) =
n
X
bk nk (x).
k=0
It follows from (1) and (2) that
n (x)t
n=0
X
n
X
bk nk (x)tn
!
!
X
n
n
bn t
n (x)t
n=0 k=0
X
n=0
n=0
= B(t)y(x, t).
Note that the bk s can be computed identically.
Problem 9. Obtain a theorem analogous to that of Exercise 8 but with Sheffer
A-type replaced by -type.
Solution 9. Follow the proof in Exercise 8 except for two substitutions:
(1): Replace J(x, D) by J(x, );
(2): Replace Theorem 71 by Theorem 78.
Problem 10. Show that if Pn (x) is the Legendre polynomial,

1
x
(1 + x2 ) 2 n
Pn
n!
1 + x2
is a simple set of polynomials of Sheffer A-type zero.
n (x) =

n
(1 + x2 ) 2
n (x) =
Pn
n!
1 + x2

.
We know that

X
Pn (u)v n
v 2 (u2 1)
uv
= e 0 F1 ; 1;
.
n!
4
n=0
Therefore
X
n=0
n (x)t
P
X
n
x
1+x
(1 + x2 ) 2 tn
n!
n=0
=e
t 1+x2
x
1+x2
xt
= e 0 F1
1;
0 F1
;
t2 (1 + x2 )
1;
t2
.
4

4
x2
1+x2

1
121

t2
Hence n (x) is of Sheffer A-type zero with H(t) = t, A(t) = 0 F1 ; 1;
.
4
Problem 11. Let Pn (x) be the Legendre polynomial. Choose = D and show
that the polynomials
(x)n
Pn
n (x) =
(n!)2
are of -type zero for that .
x+1
x1

(x 1)n
Pn
n (x) =
(n!)2
where Pn (x) is the Legendre polynomial.
We know that

(1)
0 F1
; 1;
v(u 1)
2
In (1) put v = t(x 1), u =

0 F1
; 1;
x+1
x1

,
v(u + 1)
2

=
X
Pn (u)v n
.
(n!)2
n=0
x+1
1
u+1
x
u1
=
,
=
, and
. Then
x1
2
x1
2
x1
(1) yields
(2)
0 F1 (; 1; t)0 F1 (; 1; xt) =
n (x)tn .
n=0
Using Theorem 79 we may conclude from (2) that n (x) is of -type zero with
= D( + 1 1) = D and
A(t) = 0 F1 (; 1; t),
H(t) = t,
(t) = 0 F1 (; 1; t).
Problem 12. Prove that if the operator J(x, D) is such that a simple set of polynomials n (x) belongs to it in the Sheffer sense, then J(x, D) transforms every
polynomial of degree precisely n into a polynomial of degree precisely (n 1).
Solution 12. This result was proved as part of our solution to Problem 2 of this
chapter. See the theorem at the end of that solution.
Theorem : If the operator J(x, D) is such that a simple set of polynomials n (x)
belongs to it in the Sheffer sense,
(1)
J(x, D)n (x) = n1 (x), J(x, D)0 (x) = 0,
then J(x, D) transforms every polynomial of degree precisely n into a polynomial

of degree precisely n 1.
Proof : Let gn (x) be of degree exactly n. Then we know there exists the expansion
gn (x) =
n
X
A(k, n)nk (x), A(0, n) 6= 0,
k=0
merely because the n (x) form a simple set. Then, because of (1),
J(x, D)gn (x) =
(2)
n1
X
A(k, n)n1k (x),
k=0
and the right member of (2) is of degree exactly (n 1).

Problem 1. Show that the Batemans polynomial (see Section 146, page 285)
Zn (t) = 2 F2 (n, n + 1; 1, 1; t)
has the recurrence relation
n2 (2n3)Zn (t)(2n1)[3n2 6n+22(2n3)t]Zn1 (t)+(2n3)[3n2 6n+2+2(2n1)t]Zn2 (t)(2n1)(n2)2 Z

Solution 1. Let
Zn (t) = 2 F2 (n, N = 1; 1, 1; t) =
n
X
(1)k (n + k)!tk
(k!)3 (n k)!
k=0
or
Zn (t) =
(k, n).
k=0
Then
Zn1 (t) =
X
(1)k (n 1 + k)!tk
k=0
Zn2 (t) =
k=0
(k!)3 (n 1 k)!
X
nk
k=0
n+k
(k, n),
(n k)(n k 1)(n k 2)
(k, n).
(n + k)(n + k 1)(n + k 2)
Also
tZn1 (t) =
X
(1)k (n 1 + k)!xk+1
k=0
(k!)3 (n 1 k)!
X
(1)k1 (n 2 + k)!xk
k=1
[(k 1)!]3 (n k)!
or
tZn1 (t) =
X
k=0
tZn2 (t) =
k 3
(k, n)
(n + k)(n + k 1)
X
(1)k (n 2 + k)!xk+1
k=0
(k!)3 (n 2 k)!
X
(1)k1 (n 3 + k)!xk
k=1
[(k 1)!]3 (n 1 k)!
tZn2 (t) =
X
k=0
123
k 3 (n k)
(k, n).
(n + k)(n + k 1)(n + k 2)
We may now proceed to a recurrence relation of the form

Zn (t) + [A + Bt]Zn1 (t) + [C + Dt]Zn2 (t) + EZn3 (t) = 0,
in which A, B, C, D, E depend upon k. For the determination of those coefficients
we need to satisfy the identity in k:
(n + k)(n + k 1)(n + k 2) + A(n k)(n + k 1)(n + k 2) Bk 3 (n + k 2)
+C(n k)(n k 1)(n + k 2) Dk 3 (n k) + E(n k)(n k 1)(n k 2) = 0.

Use
k=n:
2n(2n 1)(2n 2) Bn3 (2n 2) = 0 :
coef f k 4 :
B + D = 0 :
k = 2 n : D(2 n)3 (2n 2) + E(2n 2)(2n 3)(2n 4) = 0
(n 2)2
D
2(2n 3)
(2n 1)(n 2)2
=
n2 (2n 3)
D(n 2)3 + 2E(n 2)(2n 3) = 0
E=
k = 1 n : B(1 n)3 (1) + C(2n 1)(2n 2)(1) D(1 n)3 (2n 1) + E(2n 1)(2n 2)(2n 3) = 0,
or
B(n 1)3 2C(n 1)(2n 1) + D(n 1)3 (2n 1) + E 2(n 1)(2n 1)(2n 3) = 0
Thus we have
2C(2n 1)
= 2E(2n 1)(2n 3) + D(n 1)2 (2n 1) B(n 1)2

2(2n 1)2 (n 2)2 (2n 3) 2(n 1)2 (2n 1)2
2(n 1)2 (2n 1)
=
+
.
2
2
n (2n 3)
n
n2
THen
n2 C
2(2n 1)
n2
2(2n 1)
D=
n2
B=
= (2n 1)(n 2)2 + (n 1)2 (2n 1) (n 1)2

= (2n 1)[n2 + 4n 4 + n2 2n + 1] (n 1)2
= (2n 1)(2n 3) (n 1)2
= 4n2 8n + 3 n2 + 2n 1
= 3n2 6n + 2
Hence
C=
3n2 6n + 2
.
n2
(2n 1)(2n 2)(2n 3) + A(1)(2n 2)(2n 3) B(n 1)3 (2n 3) D(n 1)3 (1) = 0
or
2(2n 1)(2n 3) + 2A(2n 3) B(n 1)2 (2n 3) D(n 1)2 = 0,
or
2(2n 1)(n 1)2 (2n 3) 2(2n 1)(n 1)2
+
.
2A(2n 3) = 2(2n 1)(2n 3) +
n2
n2
Hence we get
k =n1:
n2 (2n 3)A
= n2 (2n 1)(2n 3) + (2n 1)(n 1)2 (2n 3) + (2n 1)(n 1)2

= (2n 1)[(2n 3)(n2 + n2 2n + 1) + (n 1)2 ]
= (2n 1)[4n2 + 8n 3 + n2 2n + 1]
= (2n 1)(3n2 6n + 2).
We now have
(2n 1)(3n2 6n + 2)
2(2n 1)
3n2 6n + 2
2(2n 1
(2n 1)(n 2)2
,B =
, C+
,D =
,E =
.
2
2
2
2
n (2n 3)
n
n
n
n2 (2n 3)
The desired recurrence relation is therefore
A=

n2 (2n 3)Zn (t) (2n 1) 3n2 6n + 2 2(2n 3)t Zn1 (t)
+(2n 3)[3n2 6n + 2 + 2(2n 1)t]Zn2 (t) (n 2)2 (2n 1)Zn3 (t).
Note the one sign change from that in Sister Celines paper. The sign was correct
in her thesis.
Problem 2. Show that Sister Celines polynomial fn (a; ; x), or

1
Fn (x) = 3 F2 n, n + 1, a; 1, ; x
2
has the recurrence relation
nfn (x)[3n24(n1+a)x]fn1 (x)+[3n44(n1a)x]fn2 (x)(n2)fn3 (x) = 0.

1
fn (x) = 3 F2 n, n + 1, a; 1, ; x ,
2
for which we seek a pure recurrence relation.
Now

fn (x) =
n
X
(1)k (n + k)!(a)k xk
k=0
k!(n k)!k!(k)k
n
X
(1)k (n + k)!(a)k (4x)k
k=0
k!(2k)!(n k)!
Put
(k, n) =
Then
(1)k (n + k)!(a)k (4x)k

.
k!(2k)!(n k)!
fn (x) =
125
(k, n),
k=0
fn1 (x) =
X
(1)k (n 1 + k)!(a)k (4x)k
k!(2k)!(n 1 k)!
k=0
fn2 (x) =
X
nk
k=0
X
(n k)(n k 1)
k=0
fn3 (x) =
(n + k)(n + k 1)
n+k
(k, n),
X
(n k)(n k 1)(n k 2)
k=0
(k, n),
(n + k)(n + k 1)(n + k 2)
(k, n).
Furthermore,
(4x)fn1 (x)
=
=
=
X
(1)k (n 1 + k)!(a)k (4x)k+1
k=0
X
k=1
X
k=0
k!(2k)!(n 1 k)!
k1
(1)
(n 2 + k)!(a)k1 (4x)k
(k 1)!(2k 2)!(n k)!
k 2k(2k 1)
(k, n),
(n + k)(n + k 1)(a + k 1)
and
(4x)fn2 (x) =
X
(1)k (n 2 + k)!(a)k (4x)k+1
k=0
k!(2k)!(n 2 k)!
X
(1)k1 (n 3 + k)!(a)n1 (4x)k
k=1
(k 1)!(2k 2)!(n 1 k)!
or
(4x)fn2 (x) =
X
k=0
k(2k)(2k 1)(n k)
(k, n).
(n + k)(n + k 1)(n + k 2)(a + k 1)
We therefore seek an identity of the form

fn (x) + [A + B(4x)]fn1 (x) + [C + D(4x)]fn2 (x) + Efn3 (x) = 0,
in which A, B, C, D, E may depend upon n. We are thus led to the identity (in
k)
(1)
1+A
nk
2k 2 (2k 1)
(n k)(n k 1)
B
+C
n+k
(n + k)(n + k 1)(a + k 1)
(n + k)(n + k 1)
2k 2 (2k 1)(n k)
(n k)(n k 1)(n k 2)
+E
0,
(n + k)(n + k 1)(n + k 2)(a + k 1)
(n + k)(n + k 1)(n + k 2)
or
(2)
(n+k)(n+k1)(n+k2)(a+k1)+A(nk)(n+k1)(n+k2)(a+k1)
2Bk 2 (2k 1)(n + k 2) + C(n k)(n k 1)(n + k 2)(a + k 1)
2Dk 2 (2k 1)(n k) + E(n k)(n k 1)(n k 2)(a + k 1) 0.

We need five equations for the determination of A, B, C, D, E.
Use
k=n:
2n(2n 1)(2n 2)(a + n 1) 2Bn2 (2n 1)(2n 2) = 0,

or
a + n 1 Bn = 0
k =1a:
B=
a+n1
.
n
2B(1 a)2 (1 2a)(n 1 a) 2D(1 a)2 (1 2a)(n 1 + a) = 0,

or
(n 1 a)B
;D =
a+n1
n1a
=
n
B(n 1 a) + D(n 1 + a) = 0
k = 2 n : 2D(2 n)2 (3 2n)(2n 2) + E(2n 2)(2n 3)(2n 4)(a n + 1) = 0,

or
n2
E=
1+an
n2
=
n
D(n 2) + E(a n + 1) = 0;
k=0:
and
coeff k 4 :
k=0:
n(n 1)(n 2)(a 1) + An(n 1)(n 2)(a 1) + Cn(n 1)(n 2)(a 1)

+En(n 1)(n 2)(a 1) = 0
1 a 4B + C + 4D E = 0
1 + A + C + E = 0.
Hence
2 + 2C 4B + 4D = 0.
Therefore we have
C
= 1 + 2B 2D
2E + 2n 2 2n 2 2a
= 1 +
+
n
n
from which we get

C=
3n 4
.
n
Finall,
A = 1 C E = 1
so
A=
Thus we obtain
3n 4 n 2
+
n
n
3n 2
.
n
127
nfn (x)[(3n2)(4x)(a+n1)]fn1 (x)+[3n44x(n1a)]fn2 (x)(n2)fn3 (x) = 0.

Problem 3. Show that Rices polynomial (see Section 147, page 287)
Hn = Hn (, p, v) = 3 F2 (n, n + 1, ; 1, p; v)
satisfies the relation
n(2n3)(p+n1)Hn (2n1)[(n2)(pn+1)+2(n2)(2n3)2(2n3)(+n1)v]Hn1 +(2n3)[2(n1)2 n(p

Solution 3. Now
Hn (, p, v)
= 3 F2 (n, n + 1, ; 1, p; v)
n
X
(1)k (n + k)!()k v k
=
(k!)2 (n k)!(p)k
k=0
X
(k, n).
=
k=0
Then
Hn1 =
X
(1)k ()k (n 1 + k)!v k
k=0
(k!)2 (p)k (n 1 k)!
Hn2 =
X
nk
k=0
X
(n k)(n k 1)
k=0
Hn3 =
(n + k)(n + k 1)
n+k
(k, n),
(k, n),
X
(n k)(n k 1)(n k 2)
(k, n),
((n + k)(n + k 1)(n + k 2)
k=0
vHn1 =
X
k=0
X (1)k1 ()k1 (n 2 + k)!v k

(1)k ()k (n 1 + k)!v k+1
=
2
(k!) (p)k (n 1 k)!
[(k 1)!]2 (p)k1 (n k)!
k=1
vHn1 =
X
k=0
k 2 (p + k 1)
(k, n),
(n + k)(n + k 1)( + k 1)
and
vHn2 =
X
k=0
k 2 (p + k 1)(n k)
(k, n).
(n + k)(n + k 1)(n + k 2)( + k 1)
Then there exists a relation

Hn + (A + Bv)Hn1 + (C + Dv)Hn2 + EHn3 = 0,
in which A, B, C, D, E depend upon n alone. We are led to the identity in k:
(n + k)(n + k 1)(n + k 2)( + k 1) + A(n k)(n + k 1)(n + k 2)( + k 1)
Bk 2 (p + k 1(n + k 1) + C(n k)(n k 1)(nk 1)(n + k 2)( + k 1) Dk 2 (p + k 1)(n k)
+E(n k)(n k 1)(n k 2)( + k 1) = 0
We now solve for A, B, C, D, E :
k =1 :
B(1 )2 (p )(n 1) D(1 )2 (p )(n 1 + );
k=n:
(2n)(2n 1)(2n 2)( + n 1) Bn2 (p + n 1)(2n 2) = 0;
k = 2 n : D(2 n)2 (p + 1 n)(2n 2) + E(2n 2)(2n 3)(2n 4)( + 1 n) = 0;
B(n 1)
.
n+ 1
2(2n 1)(n + 1)
B=
n(n + p 1)
2(2n 1)(n 1)
D=
n(n + p 1)
(n 2)(p + 1 n)
E=
D
2(2n 3)( + 1 n)
(n 2)(2n 1)(p + 1 n)
=
n(2n 3)(n + p 1)
D=
k = 1 n : B(1 n)2 (p n)(1) + C(2n 1)(2n 2)(1)( n) D(1 n)2 (p n)(2n 1)

+E(2n 1)(2n 2)(2n 3)(p n) = 0;
we have
B(n1)(pn)2C(2n1)(n)D(n1)(pn)(2n1)+2E(2n1)(2n3)(pn) = 0,
or
2(2n 1)(n 1)(n 1)(p n)(2n 1)
(2n 1)(n + p 1)(n 1)(p n)
2C(2n1)(n)+
n(n + p 1)
n(n + p 1)
+
2(n 2)(2n 1)(p + 1 n)92n 1)(2n 3)( n)

= 0,
n(2n 3)(n p 1)
or
n( n)(n + p 1)C
= (n + 1)(n 1)(p n) + (n 1)(n 1)(p n)(2n 1)

+(n 2)(p + 1 n)(2n 1)(p n)
= (n 1)(p n)[n + 1 + (n 1)(2n 1)] + (n 2)(p + 1 n)(2n 1)( n)
= 2(n 1)2 (p n)( n) + (n 2)(p + 1 n)(2n 1)( n).
Hence
n(n + p 1)C
= 2(n 1)2 (p n) + (n 2)(2n 1)(p + 1 n)

= 2(n 1)2 (p n + 1) + 2(n 1)2 + (n 2)(2n 1)(p + 1 n)
= 2(n 1)2 + (p n + 1)[2n2 + 4n 2 + 2n2 5n + 2]
= 2(n 1)2 n(p n + 1),
so
C=
2(n 1)2 n(p n + 1)

.
n(p + n 1)
Finally,
coeff k 4 : 1 A B + C + D E = 0.
Then
A=1B+C +DE
so
n(2n 3)(p + n 1)A
129
= n(2n 3)(p + n 1) 2(2n 3)(2n 1)(n + 1)

+(2n 3)[2(n 1)2 n(p n + 1)] 2(2n 3)(2n 2)(n 1)
(n 2)(2n 1)(p + 1 n),

or
n(2n 3)(p + n 1)A
= (2n 3)[pn + n2 n + 2n2 4n + 2 pn + n2 n]

(2n 1)[2(2n 3)(n + 1) + 22(2n 3)(n 1) + (n 2)(p + 1 n)]
= 2(2n 3)(n 1)(2n 1) 2(2n 1)(2n 3)(n + 1)
(2n 1)[2(2n 3)(n p 1) + (1 2)(p + 1 n)]
= (2n 1)[2(2n 3)(n + 1 n + + 1) + 2(n 1)(2n 3) (n 2)(p + 1 n)
= (2n 1)[4(n 1)(2n 3) + 2(n 1)(2n 3) (n 2)(p + 1 n)]
= (2n 1)[(n 2)(p + 1 n) + 2(n 1)(2n 3)].
Hence we arrive at the recurrence relation

n(2n 3)(p + n 1)Hn
(2n 1)[(n 2)(p + 1 n) + 2(n 1)(2n 3) 2(2n 3)(n + p 1)v]Hn1
+(2n 3)[2(n 1)2 n(p + 1 n) 2(2n 1)(n p 1)v]Hn2
+(n 3)(2n 1)9p + 1 n)Hn3 = 0.
Problem 4. Show that the polynomial
fn (x) = 1 F2 (n; 1 + , 1 + ; x),
which is intimately related to Batemans Jnu,v of Section 147, page 287, satisfies
the relation
(+n)(+n)fn (x)[3n2 3n+1+(2n1)(+)+x]fn1 (x)+(n1)(3n3++)fn2 (x)(n1)(n2)fn3 (

Solution 4. Let fn (x) = 1 F2 (m; 1+, 1+; x) =
n
X
k=0
(1)k nxk
.
k!(1 + )k (1 + )k (n k)!
fn (x)
Put gn (x) =
. Then
n!
n
X
X
(1)k xk
gn (x) =
=
(k, n).
k!(1 + )k (1 + )k (n k)!
k=0
k=0
Now
gn1 (x) =
X
k=0
X
(1)k xk
=
(n k)(k, n)
k!(1 + )k (1 + )k (n 1 k)!
k=0
and
gn2 (x) =
(n k)(n k 1)(k, n)
k=0
gn3 (x) =
(n k)(n k 1)(n k 2)(k, n).
k=0
Also
xgn1 (x) =
X
k=0
X
(1)k xk+1
(1)k1 xk
=
k!(1 + ) k(1 + )k (n 1 k)!
(k 1)!(1 + )k1 (1 + )k1 (n k)!
k=1
or
xgn1 (x) =
k( + k)( + k)(k, n).
k=0
Then there exists the relation

gn (x) + (A + Bx)gn1 (x) + Cgn2 (x) + Dgn3 (x) = 0,
with A, B, C, D dependent only on n.
We are led to the identity in k
1+A(nk)Bk(+k)(+k)+C(nk)(nk1)+D(nk)(nk1)(nk2) = 0.
Then use
1
.
n( + n)( + n)
1
coeff k 3 :
B D = 0;
D=
n( + n)( + n)
(n 1)( + n 1)( + n 1) n( + n)( + n)
k = n 1 : 1 + A B(n 1)( + n 1)( + n 1) = 0; A =
n( + n)( + n)
3n2 3n + 1 + (2n 1)( + ) +
=
n( + n)( + n)
k=0:
1 + An + Cn(n 1) + Dn(n 1(n 2) = 0
1 Bn(a + n)( + n) = 0;
k=n:
B=
Then
n(n 1)( + n)( + n)C
= ( + n)( + n) + 3n 3n + 1 + ( + )(2n 1) + + (n 1)(n 2)

= n2 ( + )n + 3n2 3n + 1 + ( + )(2n 1) + + n2 3n + 2
= 3n2 6n + 3 + ( + )(n + 2n 1)
= 3(n 1)2 + (n 1)( + )
= (n 1)[3n 3 + + ]
So
C=
3n 3 + +
.
n( + n)( + n)
We thus find that

n( + n)( + n)gn (x) [3n2 3n + 1 + n 1)( + ) + x]gn1 (x)
+(3n 3 + + )gn2 (x) gn3 (x) = 0.
Now gn (x) =
131
fn (x)
. Hence we get
n!
( + n)( + n)
fn1 (x) 3n 3 + +
fn3 (x)
fn (x)[3n2 3n+1+n1)(+)+x]
+
fn2 (x)
= 0,
(n 1)!
(n 1)!
(n 2)!
(n 3)!
or
( + n)( + n)fn (x) [3n2 3n + 1 + (2n 1)( + ) + x]fn1 (x)
+(n 1)[3n 3 + + ]fn2 (x) (n 1)(n 2)fn3 (x) = 0.
Problem 5. Define the polynomial wn (x) by
wn (x) =
n
X
(1)k n!Lk (x)
k=0
(k!)2 (n k)!
in terms of the Laguerre polynomial Lk (x). Show that wn (x) possesses the pure
recurrence relation
n2 wn (x)[(n1)(4n3)+x]wn1 (x)+(6n2 19n+16+x)wn2 (x)(n2)(4n9)wn3 (x)+(n2)(n3)wn4 (x) = 0.
Solution 5. Let
wn (x) =
n
X
(1)k n!Lk (x)
k=0
(k!)2 (n k)!
in terms of the simple Laguerre polynomial. We know that

(1)
nLn (x) (2n 1 x)Ln1 (x) + (n 1)Ln2 (x) = 0
and we seek a recurrence relation for wn (x).
wn (x)
(1)k Lk (x)
Put
= n (x) and
= gk (x).
n!
(k!)2
Then
(1)k (k!)2 kgk (x)(1)k1 [(k1)!]2 (2k1x)gk1 (x)+(1)k2 [(k2)!]2 (k1)gk2 (x) = 0,
or
k 3 (k 1)gk (x) + (k 1)(2k 1 x)gk1 (x) + gk2 (x) = 0.
Also
n (x) =
X
X
gk (x)
gk (x)
=
.
(n k)!
(n k)!
k=0
k=0
Then
n1 (x) =
n3 (x) =
X
k=0
X
k=0
(n k)
gk (x)
,
(n k)!
gk (x)
(n k)(n k 1)(n k 2)
,
(n k)!
n2 (x) =
n4 (x) =
(n k(N k 1)
k=0
gk (x)
,
(n k)!
(n k)(n k 1)(n k 2)(n k 3)
k=0
gk (x)
.
(n k)!
We first wish to fins A, B, C, D, E so that

An (x) + Bn1 (x) + Cn2 (x) + Dn3 (x) + En4 (x) =
X
k 3 (k 1)gk (x)
k=0
(n k)!
The above requires that A to E satisfy the identity in k,

A+B(nk)+C(nk)(nk1)+D(nk)(nk1)(nk2)+E(nk)(nk1)(nk2)(nk3) = k 3 (k1).
We thus get
A = n3 (n 1)
E=1
D E(n + n 1 + n 2 + n 3) = 1
or
D + 4n 6 = 1; D = (4n 7)
k = n 1 : A + B(1) = (n 1)3 (n 2)
B = (n 1)3 (n 2) n3 (n 1) = (n 1)[(n 1)2 (n 3) n3 ]
B = (n 1)(4n2 5n + 2).
k = n 2 : A + 2B + 2 1 C = (n 2)3 (n 3)
2C = (n 2)3 (n 3) N 3 (n 1) + 2(n 1)(4n2 5n + 2)
= 12n2 30n + 20
k=n:
coeff k 4 :
coeff k 3 :
So C = 2n2 15n + 10. We now have

(1)
n3 (n 1)n (x) (n 1)(4n2 5n + 2)n1 (x) + (5n2 15n + 10)n2 (x)
X
k 3 (k 1)gk (x)
.
(4n 7)n3 (x) + n4 (x) =
(n k)!
k=0
From
n1 (x) =
X
k=0
gk (x)
,
(n 1 k)!
we get, with our usual convention that gs (x) 0 for s < 0,

n1 (x) =
X
gk1 (x)
(n k)!
k=0
and in the same way

n2 (x) =
X
(n k)gk1 (x)
k=0
n3 (x) =
(n k)!
(n k)(n k 1)
k=0
gk1 (x)
, etc
(n k)!
We now wish to find F, G, H, so that

F n1 (x) + Gn2 (x) + Hn3 (x) =
(k 1)(2k 1)
k=0
gk1 (x)
.
(n k)!
133
We thus arrive at the identity

F + G(n k) + H(n k)(n k 1) = (k 1)(2k 1).
Hence
k=n:
F = (n 1)(2n 1)
coeff k 2 : H = 2
coeff k : G H(n + n 1) = 3
G + 2(2n 1) = 3
G = (4n 5)
Therefore we get
(2)
(n1)(2n1)n1 (x)(4n5)n2 (x)+2n3 (x) =
X
(k 1)(2k 1)gk1 (x)
(n k)!
k=0
From A1 n1 (x) + B1 n2 (x) =
X
(k 1)gk1 (x)
(n k)!
k=0
we determine A1 and B1 by
A1 + B1 (n k) (k 1).
Then A1 = (n 1), B1 = 1. Hence
(n 1)n1 (x) + n2 (x) =
(3)
X
(k 1)gk1 (x)
(n k)!
k=0
Also
(4)
n2 (x) =
X
k=0
X gk2 (x)
gk (x)
=
.
(n 2 k)!
(n k)!
k=0
Since
X
k 3 (k 1)gk (x) + (k 1)(2k 1 x)gk1 (x) + gk2 (x)
k=0
(n k)!
= 0,
we may combine (1), (2), (3), (4) to get

n3 (n 1)n (x) (n 3)(4n2 5n + 2)n1 (x) + (6n2 15n + 10)n2 (x)
(4n 7)n3 (x) + n4 (x) + (n 1)(2n 1)n1 (x) (4n 5)n3 (x)
+2n3 (x) (n 1)xn1 (x) + xn2 (x) + n2 (x) = 0,
or
n3 (n 1)n (x) (n 1)[(4n2 7n + 3 + x]n1 (x) + (6n2 19n + 16 + x)n2 (x)
(4n 9)n3 (x) + n4 (x) = 0
wn (x)
We recall that n (x) =
and use it to obtain the final result:
n!
n2 wn (x)(4n2 7n+3+x)wn1 (x)+(6n2 19n+16+x)wn2 (x)(n2)(4n9)wn3 (x)+(n2)(n3)wn4 (x) = 0.
Problem 6. Show that the polynomial wn (x) of Exercise 5 may be written
n
n + k;
X
(n)k (x)k
1
wn (x) =
.
1 F1
(k!)3
1 + k;
k=0
wn (x) =
n
X
(1)k n!Lk (x)
(k!)2 (n k)!
k=0
Let us form
=
=
=
=
=
X
wn (x)tn
n!
n=0
X
n
X
(1)k Lk (x)tn
(k!)2 (n k)!
n=0 k=0
(1)k Lk (x)tn+k
(k!)2 n!
n,k=0
X
X
(1)k+s xs tn+k
n,k=0 s=0
X
k,s,n=0
k!n!(s!)2 (k s)!
(1)k xs tn+k
(s!)2 k!k!(k + s)!
Then
n
X
X
(1)k xs tn+s
(s!)2 k!(n k)!(k + s)!
n,s=0 k=0
n;
X
xs tn+s
1
=
1 F1
n!(s!)3
n,s=0
1 + s;
X
n
n + s;
X
xs t n
1
=
.
1 F1
3
(s!) (n s)!
n=0 s=0
1 + s;
=
Hence
wn (x)
n
X
1 F1
1 F1
1 + s;
n + s;
s=0
s=0
n
X
n + s;
1 + s;
n!xs
s)!
(s!)3 (n
(n)s (x)
1
,
(s!)3
as desired.
Problem 7. Define the polynomial vn (x) by (see Section 131, page 251)
vn (x) =
n
X
(1)k n!Pk (x)
k=0
(k!)2 (n k)!
135
in terms of the Legendre polynomial Pk (x). Show that vn (x) satisfies the recurrence relation
n2 vn (x)[4n2 5n+2(2n1)x]vn1 (x)+[6n2 15n+11(4n5)x]vn2 (x)(n2)(4n72x)vn3 (x)
+(n 2)(n 3)vn4 (x) = 0.
vn (x) =
n
X
(1)k n!Pk (x)
k=0
(k!)2 (n k)!
vn (x)
(1)k Pk (x)
= gk (x).
= n (x) and
n!
(k!)2
We know that Pk (x) satisfies the relation
Put
kPk (X) (2k 1)xPk1 (x) + (k 1)Pk2 (x) = 0.

Hence gk (x) satisfies
(1)k k(k!)2 gk (x)(1)k1 [(k1)!]2 (2k1)xgk1 (x)+(1)k2 (k1)[(k2)!]2 gk2 (x) = 0,
or
(1)
k 3 (k 1)gk (x) + (k 1)(2k 1)xgk1 (x) + gk2 (x) = 0.
Also
(2)
n (x) =
X
gk (x)
.
(n k)!
k=0
Because of the identity of form of equation (2) with the corresponding relations
of Exercise 5, we may write immediately
(3) n3 (n 1)n (x) (n 1)(4n2 5n + 2)n1 (x) + (6n2 15n + 10)n2 (x)
(4n 7)n3 (x) + n4 (x) =
X
k 3 (k 1)
k=0
(n k)!
gk (x).
In the same way we use the work in Exercise 5 to conclude that
(4)
(n1)(2n1)n1 (x)(4n5)n2 (x)+2n3 (x) =
X
(k 1)(2k 1)gk1 (x)
(n k)!
k=0
and
(5)
n2 (x) =
X
gk2 (x)
.
(n k)!
k=0
Since
X
k 3 (k 1)gk (x) + (k 1)(2k 1)xgk1 (x) + gk2 9x)
k=0
(n k)!
= 0,
we get from (3), (4), (5) that

n3 (n 1)n (x) (n 1)[4n2 5n + 2 (2n 1)x]n1 (x)
+[6n2 15n + 11 (4n 5)x]n2 (x) [4n 7 2x]n3 (x) + n4 (x) = 0.
vn (x)
But n (x) =
. Hence we arrive at the result
n!
n2 vn (x) [4n2 5n + 2 (2n 1)x]vn1 (x) + [6n2 15n + 11 (4n 5)x]vn2 (x)
(n 2)(4n 7 2x)vn3 (x) + (n 2)(n 3)vn4 (x) = 0.
Problem 8. Show that the vn (x) of Exercise 7 satisfies the relations
(1 x2 )vn00 (x) 2xvn0 (x) + n(n = 1)vn (x) = 2n2 vn1 9x) n(n 1)vn2 (x)
and
0
(1x2 )vn0 (x)+nxvn (x) = [(2n1)x1]vn1 (x)(n1)xvn2 (x)+(1x2 )vn1
(x).
Solution 8.
Problem 9. Let
(1)k ( 21 )nk (2x)n2k
,
k!(n 2k)!
so that the Legendre polynomial of Chapter 10 may be written
(k, n) =
Pn (x) =
(k, n).
k=0
Show that
xPn1 (x) =
X
(n 2k)(k, n)
k=0
xPn0 (x) =
0
Pn1
(x)
2n 2k 1
(n 2k)(k, n),
k=0
X
2k(k, n)
,
2n 2k 1
k=0
X
0
Pn+1
(x) =
(1 + 2n 2k)(k, n),
Pn2 (x) =
k=0
2k(k, n),
etc.
k=0
Use Sister Celines method to discover the various differential recurrence relations and the pure recurrence relation for Pn (x).
Solution 9. Let (x, n) =
(1)k ( 21 )nk (2x)n2k

.
k!(n k)!
Then
Pn (x) =
X
k=0
with our usual conventions. Also
(k, n)
xPn1 (x) =
X
(1)k ( 1 )n1k (2x)n2k
2
k=0
2 k!(n 1 2k)!
137
n 2k
(k, n)
2(n k 12
k=0
or
xPn1 (x) =
X
k=0
n 2k
(k, n).
2n 2k 1
Next
Pn2 (x)
=
=
X
(1)k ( 1 )n2k (2x)n22k
2
k=0
X
k=1
k!(n 2 2k)!
(1)k1 ( 21 )n1k (2x)n2k
,
(k 1)!(n 2k)!
so that
Pn2 (x) =
X
k
2k
(k,
n)
=
(k, n).
1
2n
2k 1
n
2
k=0
k=0
Now
xPn0 (x) =
X
(1)k ( 1 )nk (2x)n2k
2
k!(n 1 2k)!
k=0
(n 2k)(k, n),
k=0
and
0
Pn+1
(x) =
X
(1)k ( 1 )n+1k 2(2x)n2k
2
k=0
k!(n 2k)!
(2n 2k + 1)(k, n).
k=0
Finally,
0
Pn1
(x)
=
=
=
X
9 1)k ( 1 )n1k 2(2x)n22k
2
k=0
X
k=1
k!(n 2 2k)!
(1)k1 ( 21 )N k 2(2x)n2k
(k 1)!(n 2k)!
2k(k, n).
k=0
Problem 10. Apply sister Celines method to discover relations satisfied by the
Hermite polynomials of Chapter 11.
Solution 10.
Problem 11. Find the various relations of Section 114 on Laguerre polynomials
by using Sister Celines technique.
Solution 11.
Problem 12. Consider the pseudo-Laguerre polynomials (Boas and Buck [2;16])
fn (x) defined for nonintegral by
n
fn (x) =
X ()nk xk
()n
.
1 F1 (n; 1 + n; x) =
n!
k!(n k)!
k=0
Show that the polynomials fn (x) are not orthogonal with respect to any weight
function over any interval because no relation of the form
fn (x) = (An + Bn x)fn1 (x) + Cn fn2 (x)
is possible. Obtain the pure recurrence relation
nfn (x) = (x + n 1 )fn1 (x) xfn2 (x).
Solution 12. Consider fn (x) defined by
n
X ()nk xk
()n
.
fn (x) =
1 F1 (n; 1 + n; x) =
n!
k!(n k)!
k=0
Put
()nk xk
.
k!(n k)!
(k, n) =
Then
fn (x) =
(k, n)
k=0
fn1 (x) =
fn2 (x) =
X
X
()n1k xk
nk
=
(k, n)
k!(n 1 k)!
+n1k
k=0
k=0
k=0
(n k)(n k 1)
(k, n)
( + n 1 k)( + n 2 k)
and
xfn1 (x) =
X
()n1k xk+1
k=0
xfn2 (x) =
X
k=0
k!(n 1 k)!
()n2k xk+1
=
k!(n 2 k)!
X
k=0
X
k=1
X
()nk xk
=
k(k, n)
(k 1)!(n k)!
()n1k xk
=
(k 1)!(n 1 k)!
k=0
X
k=0
k(n k)
(k, n).
+ n 1 k
Now consider
Tn fn (x) An fn1 9x) Bn xfN 1 (x) Cn fn2 (x).
At once

X
Tn
1 An
k=0

nk
(n k)(n k 1)
Bn k Cn
(k, n).
+ n 1 k
( + n 1 k)( + n 2 k)
If Tn 0, then the following must be an identity in k with independent of n:
(1)
139
(+n1k)(+n2k)+An (nk)(+n2k)Bn k(+n1k)(+n2k)Cn (nk)(nk1) 0.

From the coefficient of k 3 we see that Bn = 0. Then Tn 0 is impossible because
fn1 (x) and fn2 (x) are of lower degree than fn (x).
Now consider the identity
fn (x) Dn xfn1 (x) En fn1 (x) Fn xfn2 (x) = 0.
We need to determine Dn , En , Fn to satisfy
1 Dn k En
nk
k(n k)
Fn
= 0,
+ n 1 k
+ n 1 k
or
(2)
+ n 1 k Dn k( + n 1 k) En (n k) Fn k(n k) 0.
1
1 Dn n( 1) = 0; Dn = or = 1
n
1
coeff k 2 :
Dn + Fn = 0; Fn =
n
+ n 1
k = + n 1 : ( 1)En + ( 1)Fn ( + n 1) = 0; En =
.
n
Hence, either = 1 or
k=n:
nfn (x) xfn1 (x) ( + n 1)fn1 (x) + xfn2 (x) = 0,

or
nfn (x) = [x + n 1 ]fn1 (x) xfn2 (x) = 0,
as desired.
For the polynomials in each of the following examples, use Sister Celines technique
to discover the pure recurrence relation and whatever mixed relations exist.
Problem 13. The Bessel polynomials of Section 150.
Solution 13.
Problem 14. Bendients polynomials Rn of Section 151.
Solution 14.
Problem 15. Bendients polynomials Gn of Section 151.
Solution 15.
Problem 16. Shivelys polynomials Rn of Section 152.
Solution 16.
Problem 17. Shivelys polynomials n of Section 152.
Solution 17.
In Exercises 1-8, Hn (x), Pn (x), Ln (x) denote the Hermite, Legendre, and simple
Laguerre polynomials, respectively. Derive each of the stated symbolic relations.
Problem 1.
Hn (x + y) + [H(x) + 2y]n .
X
Hn (x + y)tn
n!
n=0
= exp[2(x + y)t t2 ]
2
= exp(2yt) exp(2xt
! t )
!
n
n
X Hn (x)tn
X (2y) t
=
n!
n!
n=0
n=0
n
nk
X
X
(2y)
Hk (x) n
t .
=
)
n=0
k!(n k)!
k=0
It follows that
Hn (x + y) =
n
X
n!(2y)nk Hk (x)
k!(n k)!
k=0
or its equivalent
Hn (x + y) + {H(x) + 2y}n .
Problem 2.
n
[H(x) + H(y)]n + 2 2 Hn (2 2 (x + y)).

X
[H(x) + H(y)]n tn
n!
n=0
n
X
X
Hk (x)Hnk (y)tn
k!(n k)!
!
!
X Hn (y)tn
Hn (x)tn
=
n!
n!
n=0
n=0
2
= exp(23xt t ) exp(2yt t2 )
2
= exp[2(x
+ y)t
2t ]

x+y
= exp 2
( 2t) ( 2t)2 .
2
n=0 k=0
It follows that

n n
[H(x) + H(y)] t
+
n!
n=0
n=0
Hn
x+y
( 2t)n
,
or
n
{H(x) + H(y)} + 2
1
2n

Hn
x+y

.
Problem 3.

n
1x
1+x
Pn (x) + n! L
L
2
2

X
t(x 1)
t(x + 1)
Pn (x)tn
=
F
;
1;
F
;
1;
.
0 1
0 1
(n!)2
2
2
n=0
Now
X
Ln (y)tn
e 0 F1 (; 1; yt) =
.
n!
n=0
t
Hence
X
Pn (x)tn
(n!)2
n=0

t(x 1)
t(x + 1)
t
= e 0 F1 ; 1;
e 0 F1 ; 1;
2
2
!
!

1+x
1x
X (1)n Ln
X Ln
tn
tn
2
2
=
n!
n!
n=0
n=0

X
n
1+x
1x
k
X
(1) Lk ( 2 )Lnk 2
=
tn .
k!(n
k)!
n=0
t
k=0
Hence
Pn (x) = n!
n
X
n!(1)k L ( 1+x )Lnk ( 1x )
2
k!(n k)!
k=)
or

Pn (x) + n! L
Problem 4.

Hn
1x
2
1+x
2
n
.

1
P (x) + [H(x) P (x)]n .
2
X
Hn
n=0
1
2 P (x)
n!
tn
[2]
X
X
(1)k Pn2k (x)tn
k!(n 2k)!
!
!
X Pn (x)tn
(1)n t2n
=
n!
n!
n=0
n=0

2 2
t
(x
1)
2
= et ext 0 F1 ; 1;
! 4
!
X
X (1)n Pn (x)tn
Hn (x)tn
=
n!
n!
n=0
n=0
X
n
X
(1)k Pk (x)Hnk (x)tn
.
=
k!(n k)!
n=0
n=0 k=0
k=0
It follows that
141

Hn

1
P (x)
2
n
X
(1)k n!Pk (x)Hnk (x)
k!(n k)!
n
+ {H(x) P (x)} .
k=0
Problem 5.
[H(x) 2P (x)]2n+1 + 0,
2n
[H(x) 2P (x)]
n 2n
+ (1) 2

1
Ln (x2 1).
2 n
=
+
X
[1 + (x) 2P (x)]n tn
n!
n=0
n
X
k
X
(2) Pk (x)Hnk (x)tn
k!(n k)!
!
!
X Pn (x)(2t)n
Hn (x)t
=
n!
n!
n=0
n=0

4t2 (x2 1)
2
= exp(2xt t ) exp(2xt)0 F1 ; 1;
4
= exp(t2 )0 F1 (; 1; t2 (x2 1))
X
(1)n Ln (x2 1)t2n
=
.
n!
n=0
n=0 k=0
It follows that
[H(x) 2P (x)]2n+1 + 0
and
[H(x) 2P (x)]2n +
(1)n (2n)!
Ln (x2 1) = (1)n 22n
n!

1
Ln (x2 1).
2 n
Problem 6.
[H(x) P (2x)]2n+1 + 0,
2n
[H(x) P (2x)]
n 2n
+ (1) 2

1
1
2
Ln x
.
2 n
4
X
[H(x) P (2x)]n tn
n!
n=0
143
X
n
X
(1)k Pk (2x)Hnk (x)tn
k!(n k)!
!
!
X Pn (2x)(t)n
Hn (x)t
=
n!
n!
n=0
n=0

t2 (4x2 1)
2
= exp(2xt t ) exp(2xt)0 F1 ; 1;

4
1
= exp(t2 )0 F1 ; 1; (t2 )(x2 )
4
X
(1)n Ln (x2 41 )t2n
=
.
n!
n=0
n=0 k=0
We may conclude that

[H(x) P (2x)]2n+1 + 0
and
[H(x) P (2x)]2n +
(1)n (2n)!Ln (x2 41 )

= (1)n 22n
n!

1
1
L n x2
.
2 n
4
Problem 7.

Hn
1
H(x)
2
+ 2 2 Hn (2 2 x).
Solution 7. From
X
Hn ( 12 H(x))tn
n!
n=0
[2]
X
X
(1)k Hn2k (x)tn
k!(n 2k)!
!
!
X Hn (x)tn
(1)n t2n
=
n!
n!
n=0
n=0
= exp(t2 ) exp(2xt t2 )
2
= exp(2xt
2t)

2
x
= exp 2
2t ( 2t)
2
x
H (
X
)( 2t)n
n
2
=
n!
n=0
n=0 k=0
we may conclude that

Hn

n
x
1
H(x) + 2 2 Hn
.
2
2
Problem 8.
Hn (xy) + [H(x) + 2x(y 1)]n .
Solution 8. From
X
Hn (xy)tn
n!
n=0
= exp(2xyt t2 )
= exp(2xt t2 ) exp[2x(y
! 1)t]
!
X [2x(y 1)]tn
X
Hn (x)tn
=
n!
n!
n=0
n=0
X
n
k
X
[2x(y 1)] Hnk (x)tn
=
,
k!(n k)!
n=0
k=0
it follows that
Hn (xy) + [H9x) + 2x(y 1)]n .
Problem 9. Use Laplaces first integral for Pn (x) to derive equation (20), preceding
these exercises.
Pn (x) =
[x +
x2 1 cos ]n d.
We have defined vn (x) by

vn (x) + Ln (P (x)) =
n
X
(1)k n!Pk (x)
k=0
(k!)n (n k)!
Hence
vn (x)
Z
n
1 X (1)k n![x + x2 1 cos
d
0
(k!)2 (n k)!
k=0
Z

p
1
=
Ln x + x2 1 cos d.
0
=
Problem 10. For the vn (x) of equation (15), page 251, evaluate
Z
vm (x)Pk (x)dx,
1
and use your result to establish equation (21), preceding these exercises.
Z 1
Solution 10. For the vn (x) of Exercise 9 we wish to evaluate
vm (x)Pk (x)dx.
1
Now
Z
vm (x)Pk (x)dx =
1
Pk (x)
1
m
X
(1)s m!Ps (x)dx
s=0
(s!)2 (m s)!
Hence, if m < k, then s < k and we obtain

Z
vm (x)Pk (x)dx = 0, m < k.

1
If m k, we get
m
X
Z 1
(1)s m!
=
Pk (x)Ps (x)dx
(s!)2 (m s)! 1
s=0
Z
1
(1)k m!
=
P 2 (x)dx
(k!)2 (m k)! 1 k
(1)k 2m!
=
.
(k!)2 (m k)!(2k + 1)
vm (x)Pk (x)dx
1
145
Now consider
vN (x)vm (x)dx.
1
Let m n. Then
1
vn (x)vm (x)dx =
1
vn (x)
1
m
X
(1)k m!Pk (x)
k=0
(k!)2 (m k)!
dx.
For k > n the integrals involved in the sum are zero. We may therefore write
Z
vn (x)vm (x)dx
n
X
k=0
n
X
k=0
n
X
(1)k m!
(k!)2 (m k)!
vn (x)Pk (x)dx
1
(1)k m!
(1)k 2 n!
(k!)2 (m k)! (k!)2 (n k)!(2k + 1)
(n)k (m)k
(k!)4 (k + 21 )
k=0
n
X (n)k (m)k ( 1 )k
2
=2
.
4( 3 )
(k!)
2 k
k=0
=
Hence, since m and n are interchangeable,

1
n, m, ;
1
2
1
vn (x)vm (x)dx = 23 F4
1
3
1, 1, 1, ;
2

yx
n
(y + x)n Pn
+ n! {L(x) L(y)} .
y+x
n
X
(y + x)n Pn ( yx
y+x )t
n=0
(n!)2
0 F1 ; 1;
2
= 0 F1 (; 1; xt)0 F1 (; 1; yt)
t
= et 0 F1 (; 1; xt)e
0 F1 (; 1; y(t))!
!
X Ln (x)t4
X (1)n Ln (y)tn
=
n!
n!
n=0
n=0
X
n
k
X
(1) Lk (y)Lnk (x)tn
=
.
k!(n k)!
n=0
= 0 F1
; 1;
t(y + x)[ yx
y+x 1]
k=0
t(y + x)[ yx
y+x + 1]
2
We may conclude that

n
(y + x) Pn
yx
y+x

= n!
n
X
(1)k n!Lk (y)Lnk (x)
k!(n k)!
k=0
n
+ n! {L (x) L (y)} ,
as desired.
Problem 12. Define polynomials n (x, y) by
n (x, y) + Hn (xL(y)) .
Show that
X
n (x, y)tn
= exp(2xt t2 )0 F1 (; 1; 2xyt).
n!
n=0
Solution 12. We define n (x, y) by

n (x, y) + Hn (xL (y))
[n
2]
X
(1)k n!2n2k xn2k Ln2k (y)
=
.
k!(n 2k)!
k=0
Then
X
n (x, y)tn
n!
n=0
X
(1)k (2x)n Ln (y)tn+2k
k!n!
n,k=0
X Ln (y)(2xt)n
= exp(t2 )
n!
n=0
= exp(t2 ) exp(2xt)0 F1 (; 1; 2xgt)
= exp(2xt t2 )0 F1 (; 1; 2xyt),
as desired.
Problem 1. Let

1x
1+x
gn (x) =
.
(1 + )n (1 + )n
Use Batemans generating function, page 256, to see that
(,)
(1 + x)n Pn
gn (x)tn = 0 F1 (; 1 + ; xt)0 F1 (; 1 + ; t)
n=0
and thus show that, in the sense of Section 126, gn (x) is of -type zero with
= D( + ). Show also that gn (x) is of Sheffer A-type unity.
147
Solution 1. Let
gn (x) =
(,) 1x
( 1+x )
(1 + x)n Pn
(1 + )n (1 + )n
Then
X
n=0
gn (x)tn
(,) 1x
X
Pn
( 1+x )[t(1 + x)]n
(1 + )n (1 + )n
h
i
1x

t(1
+
x)
+
1
1+x
t(1 + x) 1 x
; 1 + ;
= 0 F1 ; 1 + ;
1
0 F1
2
1+x
2
n=0
= 0 F1 (; t + ; xt)0 F1 (; 1 + ; t).
We now see that with = D(+), gn (x) is of -type zero with A(t) = 0 F1 (; 1+
; t), H(t) = t, J(t) = t.
Since gn (x) = gn1 (x), it follows that in the Sheffer classification notation
J(x, D)
= D( + )
= D(xD + )
= (1 + )D + xD 2 .
Thus T0 (x) = 1 + , T1 (x) = x, Tk (x) 0 for k 2. So gn (x) is of Sheffer
A-type unity.
(,)
(x) because of the fact that
It might be of interest to see what we get on Pn
gn (x) is of -type zero.
(,)
(,)
2x(++n)DPn(,) (x)+[x()(++2n)]DPn1 (x) = (++n)[2nPn(,) (x)()Pn1 (x)],

which reduces to equation (2), page 159, for = = 0.
Solution 2. From equation (3), page 457 and (6) page 450 we get, with D
(,)
d
,
dx
(,)
(1)
(x1)(++n)DPn(,) (x)+(x1)(+n)DPn1 (x) = n(++n)Pn(,) (x)(+n)(++n)Pn1 (x),
(2)
(x+1)(++n)DPn(,) (x)(x+1)(+n)DPn1
(,beta)
(,)
(x) = n(++n)Pn(,) (x)+(+n)(++n)Pn1 (x).
We add (1) and (2) to get

(3)
(,)
(,)
2x(++n)DPn(,) (x)+[x()(++2n)]DPn1 (x) = (++n)[2nPn(,) (x)()Pn1 (x)].

In (3) put = = 0 to obtain
0
nxPn0 (x) nPn1
(x) = n2 Pn (x),
or
0
nPn (x) = xPn0 (x) Pn1
(x)
which is (2), page 270.

(,)
(,)
2(++n)DPn(,) (x)+[x(++2n)]DPn1 (x) = (++n)(++2n)Pn1 (x),

Solution 3. In Exercise 2 substract (1) from (2) to get
(,)
(,)
2(++n)DPn(,) (x)+[x(++2n)]DPn1 (x) = (++n)(++2n)Pn1 (x).

In the above put = = 0 and divide by 2n to get
0
Pn0 (x) xPn1
(x) = nPn1 (x),
which is (6), page 271, with a shift of index.

(,)
(,)
2n(++n+1)DPn+1 (x)+[(+)(n+2)x+n()]DPn(,) (x)+(n+1)[()x(++2n)]DPn1 (x)

(,)
= n[2(n+1)(++n)+(++n+1)(++2n+2)]Pn(,) (x)()(n+1)(++n)Pn1 (x),

Solution 4. From Exercise 3 with a shift of index we g et
(,)
2(++n+1)DPn+1 (x)+[()x(++2n+2)]DPn(,) (x) = (++n+1)(++2n+2)Pn(,) (x).

From Exercise 2 we get
(,)
2x(n + 1)( + + n)DPn(,) (x) + (n + 1)[x( ) ( + + 2n)]DPn1 (x)

(,)
= 2n(n + 1)( + + n)Pn(,) (x) ( )(n = 1)( + + n)Pn1 (x).

Let us form n times the first equation plus the second equation, we thus get
(,)
2n( + + n + 1)DPn+1 (x)

+[n( ) + x(n n 2n2 2n + 2n + 2n + 2n2 + 2 + 2 + 2n)]DPn(,) (x)
(,)
+(n + 1)[x( ) ( + + 2n)]DPn1 (x)

(,)
= n[(++n+1)(++2n+2)+2(n+1)(++n)]Pn(,) (x)()(n+1)(++n)Pn1 (x),

or
(,)
(,)
2n(++n+1)DPn+1 (x)+[n()+x(+)(n+2)]DPn(,) (x)+(n+1)[x()(++2n)]DPn1 (x)

(,)
= n[(++n+1)(++2n+2)+2(n+1)(++n)]Pn(,) (x)()(n+1)(++n)Pn1 (x),

which is the desired result.
On the above equation put = = 0 and divide by 2n(n + 1) to get
0
0
Pn+1
(x) Pn1
(x) = (2n + 1)Pn (x),
which is equation 5, page 271.
149
Problem 5. Use the method of Section 142 to show that
Pn(,) (x) =
n
(,0)
(x)
(1 + )n X (1)nk ()nk (1 + + )n+k (1 + + 2k)Pk
.
(1 + + )n
(n k)!(1 + )n+k+1
k=0
(,)
(,)
Solution 5. We wish to expand Pn

(x) in a series involving Pk
Now we have from (2), page 457, with p = 0,

(1)
1x
s
s
= (1 + )s
(x).
s
(,0)
X
(1 + + 2k)(1 + )k P
(x) (1)k s!
k
(s k)!
(1 + )s+1+k (1 + )k
k=0
and from (2), page 476,

(,)
(1 + + )n Pn
(1 + )n
(2)
(x)
n
X
(1)s (1 + + )n+s ( 1x )s
s
s=0
s!(n s)!(1 + )s
Consider the series
(,)
X
(x)tn
(1 + + )n Pn
(1 + )n
n=0
s n+s
X
(1)s (1 + + )n+2s ( 1x
2 ) t
=
s!n!(1 + )s
n,s=0
(x, t)
=
=
=
X
s
(,0)
X
(1)k s!(1 + + 2k)(1)s (1 + + )n+2s P
(x)tn+s
k
(s k)!(1 + )s+1+k s!n!
n,s=0 k=0
(,0)
(1)s (1 + + 2k)(1 + + )n+2k+2s Pk

(x)tn+s+k
s!(1 + )s+1+2k n!
n,k,s=0
X
n
s
X
(1) (1 + + )n+2k+s (1 + + 2k)Pk (x)(,0) tn+k
s!(1 + )2k+1+s (n s)!
n,k=0 s=0
Then
(x, t)
=
=
=
=
X
n,k=0
X
n,k=0
2 F1
n, 1 + + + n + 2k;
(,0)
(1 + + )n+2k (1 + + 2k)Pk
1
n!(1 + )2k+1
2 + + 2k;
(,0)
(2 + + 2k)(1 )(1 + + )n+2k (1 + + 2k)Pk
(x)tn+k
(2 + + 2k + n)(1 n)n!(1 )2k+1
(,0)
(1 + )2k+1 (1)n ()n (1 + + )n+2k (1 + + 2k)Pk

(1 + )n+2k+1 n!(1 + )2k+1
n,k=0
X
n
X
n=0 k=0
(,0)
(1)nk ()nk (1 + + )n+k (1 + + 2k)Pk

(n k)!(1 + )n+k+1
We may now conclude that
(x)tn
(x)tn+k
(x)tn+k
Pn(,) (x) =
n
(,0)
(x)
(1 + )n X (1)nk ()nk (1 + + )n+k (1 + + 2k)Pk
.
(1 + + )n
(n k)!(1 + )n+k+1
k=0
Problem 6. Use the result obtained in Section 142 to evaluate

Z
(,)
(1 x2 ) Pn(,) (x)Pk
(x)dx.
Solution 6. We know from Section 142 that
Pn(,) (x) =
n
(,)
(x)
(1 + )n X (1)nk ( )nk (1 + |alpha + )n+k (1 + 2)k (1 + 2 + 2k)Pk
.
(1 + + )n
(n k)!(1 + 2)n+k+1 (1 + )k
k=0
(,)
By using = in the orthogonality property of Pn

Z
(x) we get
1
(,)
(1 x2) Pn(,) (x)Pm
(x)dx = 0, m 6= n
and from (11), page 454,

Z
(1 x2 ) Pn(,) (x)]2 dx =
21+2 (1 + + n)(1 + + n)
.
n!(1 + 2 + 2n)(1 + 2 + n)
We may now write

Z
A(k, n)
(,)
=
1
(1 x2 ) Pn(,) (x)Pk
(1 + )n
(1 + + )n
n
X
s=0
(x)dx
(1)ns ( )ns (1 + + )n+s (1 + 2)s (1 + 2 + 2s)

(n s)!(1 + 2)n+s+1 (1 + )s
(,)
(1 x2 ) Ps(,) (x)Pk
(x)dx.
If k > n, then k > s and we get

A(k, n) = 0 for k > n.
If 0 k n, then the integrals in the sum are zero except for the one in which
s = k.
We thus arrive at
A(k, n) =
(1 + )n (1)nk ( )nk (1 + + )n+k (1 + 2)k (1 + 2 + 2k)21+2 (1 + + k)(1 + + k)

(n k)!k!(1 + + + n)(1 + 2 + n k + 1)(1 + + k)(1 + 2 + k)
or
A(k, n)
22+2 (1)nk ( )nk (1 + + n)(1 + + + n + k)(1 + 2 + k)(1 + + k)(1 + + k)

(n k)!k!(1 + + + n)(1 + 2 + n + k + 1)(1 + + k)(1 + 2 + k)
21+2 (1)nk ( )nk (1 + + k)(1 + + n)(1 + + + n + k)
=
k!(n k)!(1 + + + n)(2 + 2 + n + k
=
for 0 k n.
151
Problem 7. Use the result in Exercise 5 above to evaluate

1
(,0)
(1 x) Pn(,) (x)Pk
Z
(x)dx.
Solution 7. We wish to evaluate

1
(,0)
(1 x) Pn(,) (x)Pk
(x)dx.
We know that
= 0; k 6= s
21+ (1 + + k)(1 + k)
21+
=
=
;s = k
k!(1 + + 2k)(1 + + k)
1 + + 2k
(,0)
(1x) Pk
(x)Ps(,0) (x)dx
Now, by Exercise 5 we get

Z
(1x)
1
(,0)
Pn(,) (x)Pk
(x)dx
Z
n
(1 + )n X (1)ns ()ns (1 + + )n+s (1 + + 2s) 1
(,0)
(1x) Ps(,0) (x)Pk
(x)dx.
=
(1 + + )n s=0
(n s)!(1 + )n+s+1
1
If k > n, then k > s, so we get

Z
(,0)
(1 x) Pn(,) (x)Pk
(x)dx = 0, k > n.
If 0 k n, each integral in the sum is zero except for the one in which s = k.
Hence
Z
(,0)
(1x) Pn(,) (x)Pk
(x)dx =
(1 + )n (1)nk ()nk (1 + + )n+k (1 + + 2k)21+

.
(1 + + )n (n k)!(1 + )n+k+1 (1 + + 2k)
Problem 8. Use Theorem 84, page 269, with y = x =
2 F1
a, b;
a, b;
v 2 F1
1 + a + b c;
c;
v
to conclude that
1v
1
1
a, b, (a + b), (a + b + 1);
2
2
v = 4 F3
4v(1 v) .
a + b, c, 1 + a + b c;
Solution 8. Theorem 84 is
a, b;
2 F1
a, b;
2 F1
1x
c;
Now put x = y =

y
y
x
,
.
= F4 a, b; c, 1 c + a + b;
1y
(1 x)(1 y) (1 x)(1 y)
1 c + a + b;
x
. Then the above becomes
1x
2 F1
a, b;
a, b;
v 2 F1
= F4 (a, b; c, 1 c + a + b; v(1 v), v(1 v))
1 c + a + b;
c;
=
=
X
(a)n+k (b)n+k v n+k (1 v)n+k
k!n!(c)k (1 c + a + b)n
n,k=0
n
XX
(a)n (b)n v n (1 v)n
k!(n k)!(c)k (1 c + a + b)nk
n, c a b n;
(a)n (b)n v n (1 v)n
1
=
2 F1
n!(1 c + a + b)n
n=0
c;
X (c)(a + b + 2n)(a)n (b)n v n (1 v)n

=
(c + n)(a + b + n)n!(1 c + a + b)n
n=0
X
(a + b)2n (a)n (b)n v n (1 v)n
=
(c) (a + b)n n!(1 c + a + b)n
n=0 n
a+b a+b+1
,
;
a, b,
2
2
= 4 F3
4v(1 v) ,
n=0 k=0
c, a + b, 1 c + a + b;
as desired.
Problem 9. Use the result in Exercise 8 above, and Theorem 25, page 67, to show
that
2a, 2b, a + b;
y
y .
= 3 F2
2 F1
1
1
a+b+ ;
2a + 2b, a + b + ;
2
2
Solution 9. In Exercise 8 above replace a by (2a), b by (2b), and then put c =
1
1
1
a + b + to get (since 1 + 2a + 2b a b = a + b + )
2
2
2
1
2a, 2b;
2a, 2b, a + b, a + b + ;
v
4v(1 v)
= 4 F3
2 F1
1
1
a+b+ ;
2a + 2b, a + b + , a + b + ;
2
2
2
2a, 2b, a + b;
4v(1 v) .
= 3 F2
1
2a + 2b, a + b + ;
2
By Theorem 25, page 114, we may now write
2 F1
or
a, b;
a, b;
1
a+b+ ;
2
2a, 2b, a + b;
4v(1 v) = F
3 2
1
2a + 2b, a + b + ;
2
4v(1 v) ,
F1
2a, 2b, a + b;
y = F
3 2
1
2a + 2b, a + b + ;
2
a, b;
1
a+b+ ;
2
153
y .
Problem 1. Show that the Gegenbauer polynomial Cnv (x) and the Hermite polynomial Hn (x) are related by
n
Cnv (x)
[2]
X
2 F0 (k, v
+ n k; ; 1)
k=0
(1)k (v)nk Hn2k (x)

.
k!(n 2k)!
Solution 1. From
n
Cnv (x)
[2]
X
(1)s (v)ns (2x)n2s
s!(n 2s)!
s=0
we get
Cnv (x)tn =
n=0
X
(1)s (v)n+s (2x)n tn+2s
.
s!n!
n,s=0
We know that
[n]
2
X
(2x)n
Hn2k (x)
=
.
n!
k!(n 2k)!
k=0
Then
Cnv (x)tn
n=0
=
=
[2]
X
X
(1)s (v)n+s Hn2k (x)tn+2s
n,s=0 k=0
s!k!(n 2k)!
(1) (v)n+s+2k Hn (x)tn+2k+2s

s!k!n!
n,k,s=0
X
k
X
n,k=0 s=0
X
k
X
(1)s (v)n+2ks Hn (x)tn+2k

s!(k s)!
n!
(1 )k+s (v)n+k+s Hn (x)tn+2k

s!(k s)!
n!
n,k=0 s=0
k, v + n + k;
X
(1)k (v)n+k Hn (x)tn+2k
1
=
2 F0
k!n!
;
n,k=0
n
[2]
X
X
(1)k (v)nk Hn2k (x)tn
=
.
2 F0 (k, v + n k; ; 1)
k!(n 2k)!
n=0
=
k=0
Hence
Cnv (x)
[2]
X
2 F0 (k, v
+ n k; ; 1)
k=0
(1)k (v)nk Hn2k (x)

.
k!(n 2k)!

[n]
2
v
(v + n 2k)Cn2k
(x)
Hn (x) X
=
.
(1)k 1 F1 (k; 1 + v + n 2k; 1)
n!
k!(v)n+12k
k=0
X
X
(1)s (2x)n tn+2s
Hn (x)tn
=
.
n!
s!n!
n,s=0
n=0
Now by equation (36),

[n]
2
X
(2x)n
(v + n 2k)Cn2k (x)
=
.
n!
v k!(v)n+1k
k=0
Hence
n
X
Hn (x)tn
n!
n=0
[2]
X
v
X
(1)s (v + n 2k)Cn2k
(x)tn+2s
=
=
=
=
=
n,s=0 k=0
s!k!(v)n+1k
(1) (v + n)Cnv (x)tn+2k+2s

s!k!(v)n+1+k
n,k,s=0
X
k
X
n,k=0 s=0
X
k
X
n,k=0 s=0
(1)s (v + n)Cnv (x)tn+2k

s!(k s)!(v)n+1+ks
(1)ks (v + n)Cnv (x)tn+2k
s!(k s)!(v)n+1+s
1 F1 (k; 1
n,k=0
+ v + n; 1)
(1)k (v + n)Cnv (x)tn+2k

k!(v)n+1
Then
[n]
2
v
(1 )k (v + n 2k)Cn2k
(x)
Hn (x) X
=
.
1 F1 (k; 1 + v + n 2k; 1)
n!
k!(v)n+12k
k=0
Problem 3. Show, using the modified Bessel function of Section 65, that
e
xt

=
v
X
1
t
(v)
(v + n)Iv+n (t)Cnv (x).
2
n=0

( 2t )v+n
t2
0 F1 ; 1 + v + n;
(1 + v + n)
4
X
( 2t )v+n+2k
=
k!(1 + v + n + k)
Iv+n (t)
k=0
Now
[n]
2
v
X
(v + n 2k)Cn2k
(x)
xn
=
.
n
n!
2 k!(v)n+1k
k=0
Hence
C xt
X
xn t n
n!
n=0
n
=
=
[2]
X
v
X
(v + n 2k)Cn2k
(x)tn
2n k!(v)n+1k
n=0 k=0
(v + n)Cnv (x)tn+2k
2n+2k k!(v)n+1+k
n,k=0
XX
( t )n+2k (v)
2
(v + n)Cnv (x)
k!(v
+
n
+
1
+
k)
n=0 k=0
v X
( 2t )n+v+2k
t
(v + n)Cnv (x).
= (v)
2
k!(v
+
n
+
1
+
k)
n=0
Therefore
e
xt
v X
t
= |Gamma(v)
(v + n)Iv+n (t)Cnv (x).
2
n=0

n
Cnv (x)
[2]
X
(v 12 )k (v)nk (1 + 2n 4k)Pn2k (x)
=
.
k!( 23 )nk
k=0
Solution 4. From
n
Cnv (x)
[2]
X
(1)k (v)nk (2k)n2k
k=0
k!(n 2k)!
and
[n]
2
X
(2x)n
(2n 4x + 1)Pn2k (x)
=
n!
k!( 32 )nk
k=0
we obtain
155
Cnv (x)tn
n=0
[2]
X
X
(1)s (v)ns (2x)n2s tn
s!(n 2s)!
n=0 s=0
(1) (v)n+s (2x)n tn+2s

s!n!
n,s=0
n
=
=
=
[2]
X
X
(1)s (v)n+s (2n + 1)Pn (x)tn+2k+2s
s!k!( 32 )n+k
n,s=0 k=0
(1)s (v)(2n + 1)Pn (x)tn+2k+2s

s!k!( 32 )n+k
n,k,s=0
X
k
X
(1)s (v)(2n + 1)Pn (x)tn+2k
s!(k s)!( 32 )n+ks
n,k=0 s=0
k
X
X
(1)ks (v)n+k+s (2n + 1)Pn (x)tn+2k

s!(k s)!( 23 )n+s
n,k=0 s=0
k, v + n + k;
k
n+2k
X
1 (1) (v)n+k (2n + 1)Pn (x)t

=
.
2 F1
3
k!( 32 )n
n,k=0
+ n;
2
Therefore we get, using Example 5, page 69,
=
Cnv (x)tn
n=0
X
(1)k ( 23 )n (v 12 )k (1)k (v)n+k (2n + 1)Pn (x)tn+2k
( 32 )n+k
k!( 32 )n
n,k=0
X
(v 1 )k (v)n+k (2n + 1)Pn (x)tn+2k
2
k!( 32 )n+k
n,k=0
n
[2]
X X (v 1 )k (v)nk (2n 4k + 1)Pn2k (x)tn

2
.
k!( 23 )nk
n=0 k=0
Hence
n
Cnv (x)
[2]
X
(v 12 )k (v)nk (2n 4k + 1)Pn2k (x)
=
.
k!( 23 )nk
k=0
Problem 1. For the Bernoulli polynomial of Section 153 show that

n
x =
n
X
k=0
n!Bk (x)
.
k!(n k + 1)!
Solution 1. By definition of Bn (x) we have
X
text
Bn (x)tn
=
.
et 1 n=0
n!
157
Then
e
xt
X
1
Bn (x)tn
= (et 1)
=
t
n!
n=0
n1
X
t
n!
n=1
X
Bn (x)tn
n!
n=0
!
,
from which
X
xn tn
n!
n=0
!
!
X Bn (x)tn
tn
=
(n + 1)!
n!
n=0
n=0
X
n
n
X
Bk (x)t
=
k!(n
+ 1 k)!
n=0
k=0
Hence
n
x =
n
X
k=0
n!Bk (x)
.
k!(n + 1 k)!
Problem 2. Let Bn (x) and Bn = Bn (0) denote the Bernoulli polynomials and
numbers as t reated in Section 153. Define the differential operator A(c, D) by
A(x, D) =

X
1
B2k D2k
x
D
.
2
(2k)!
k=1
Prove that A(x, D)Bn (x) = nBn (x).

Solution 2. From
X
text
Bn (x)tn
=
t
e 1 n=0
n!
Bn (x)
is of Sheffer A-type zero. We have, in the Shefer notation,
n!
t
H(t) = t, J(t) = t, A(t) = t
.
e 1
We wish to apply Theorem 74, page 391. Now
we see that
log A(t) = log t log(et 1)

A0 (t)
1
et
1
1
= t
= 1 t
A(t)
t
e 1
t
e 1
X
tA0 (t)
t
Bn t n
=1t t
=1t
.
A(t)
e 1
n!
n=0
1
We know that B0 = 1, B1 = , B2n+1 = 0 for n 1. Hence
2
tA0 (t)
t X B2k t2k
=
.
A(t)
2
(2k)!
k=1
In the terminology of Theorem 74, since u = J(t) = t,
X
k=0
t X B2k t2k
k tk=1 =
,
2
(2k)!
k=1
vk tk+1 = tH 0 (t) = t.
k=0
1
Therefore v0 = 1, vk = 0 for k 1, and 0 = , 2k = 0 for k 1, 2k1 =
2
B2k
for k 1.
(2k)!
The identity
(k + vk )D k+1 n (x) = nn (x)
k=0
of Theorem 74 becomes
#
"

X
nBn (x)
1
B2k D 2k Bn (x)
=
.
x
D
2
(2k)!
n!
n!
k=1
Hence, with

X
1
B2k D 2k
A(x, D) = x
D
,
2
(2k)!
k=1
we may conclude that

A(x, D)Bn (x) = nBn (x).
Note that in Theorem 74, since D k+1 n (x) = 0 f or k n,
n1
X
(k + xvk )D k+1
k=0
may be replaced by
(k + xvk )D k+1 .
k=0
Problem 3. Consider the polynomials
n (c, x, y) =
(1)n ( 21 + 21 x)n
3 F2
(c)n
n,
1 1
x, 1 c n;
2 2
c,
1 1
x n;
2 2
y
.
Show that

1 F1

X
1
1
n (c, x, y)tn
1 1
x; c; yt 1 F1
= x; c; t =
,
2 2
2
2
n!
n=0
1 1 1 1
x, + x;
2 2 2 2
12 + 12 x
12 21 x
(1yt)
(t+t)
2 F1
159
X
(c)n n (c, x, y)tn
yt2
,
=
n!
(1 yt)(1 + t) n=0
c;
and that n (1, x, 1) = Fn (x), where Fn is Batemans polynomial of Section 148.
Solution 3. We define n (c, x, y) by
n (c, x, y) =
(1)n ( 1+x
2 )n
3 F2
(c)n
n,
1x
, 1 c n;
2
c,
y
.
1x
n;
2
Then
n (c, x, y) =
n
X
(1)n+k n!( 1+x )n ( 1x )(c)n ( 1+x )nk y k
2
k!(n k)!(c)n (c)k (c)nk ( 1+x

2 )n
k=0
or
n (c, x, y) =
n
X
(1)n+k n!( 1x )k ( 1+x )nk y k
2
k!(n k)!(c)k (c)nk
k=0
First consider the series
X
n (c, x, y)tn
n!
n=0
X
n
X
(1)nk ( 1x )k ( 1+x )nk y k tn
2
k!(n k)!(c)k (c)nk

n=0 k=0

1+x
1x
= 1 F1
; 0; yt 1 F1
; c; t ,
2
2
as desired. The use of c = 1, y = 1 in the above generating relation yields

1 F1
1x
; 1; t
2

1 F1
1+x
; 1; t
2

=
X
n (1, x, 1)tn
.
n!
n=0
Brafman had

1 F1

X
1x
1+x
Fn (x)tn
; 1; t 1 F1
; 1; t =
.
2
2
n!
n=0
Hence
n (1, x, 1) = Fn (x).
Next consider the series
X
n
1+x
k n
X
X
(1)n+k (c)n ( 1x
(c)n n (c, x, y)tn
2 )k ( 2 )nk y t
=
n!
k!(n k)!(c)k (c)nk
n=0
n=0 k=0
n
X (1) (c)n+k ( 1x )k ( 1+x )n y k tn+k
2
2
=
k!n!(c)k (c)n
n,k=0
X
k
n
n 1+x
X
(c + n)k ( 1x
2 )k (yt) (1) ( 2 )n t
=
k!(c)k
n!
n=0 k=0
1x
c + n,
;
n 1+x
n
X
(1) ( 2 )n t
2
=
F
2 1
yt
n!
n=0
c;
1x
n,
;
(1)n ( 1+x ) tn
X
2
1x
2 n
yt
(1 yt) 2 2 F1
=
n!
1 yt
n=0
c;
n
X
k k k
n 1+x
n
X
(1)k n!( 1x
1x
2 )k (1) y t (1) ( 2 )n t
= (1 yt) 2
k!(n k)!(c)k (1 yt)k
n!
n=0 k=0
1x
1+x
n+k
k
n+2k
X (
y t
1x
2 )k ( 2 )n+k (1)
= (1 yt) 2
k
k!n!(c)k (1 yt)
n,k=0
X
1+x
1+x
k 2k
X
( 2 + k)n (t)n (1)K ( 1x
1x
2 )k ( 2 )k y t
= (1 yt) 2
n!
k!(c)k (1 yt)k
k=0 n=0
1x
1+x
k k 2k
X(
1+x
1x
2 )k ( 2 )k (1) y t
= (1 yt) 2 (1 + t) 2
k!(c)k (1 + t)k (1 yt)k
k=0
We thus arrive at
1x 1+x
,
;
2
2
X
1+x
(c)n n (c, x, y)tn
1x
= (1yt) 2 (1+t) 2 2 F1
n!
n=0
yt2
.
(1 yt)(1 + t)
c;
Now put c = 1, y = 1 in the above. We get
1x 1+x
,
;
2
2
1+x
1x
t2
= (1 t) 2 (1 + t) 2 1 F1
1 t2
1;
1+x 1+x
,
;
1+x
1+x
2
2
2
1x
= (1 t) 2 (1 + t) 2 (1 t ) 2 2 F1
X
n=0
n (1, x, 1)tn
1;
1+x 1+x
,
;
2
2
x
= (1 t) 2 F1
1;
t2 .
t2
In Theorem 23, page 110, put a =
1+x
1
, b = to get
2
2
1+x 1
, ;
2
2
1x
F
2 = (1 t)
2
1
(t)
1;
1+x 1+x
,
;
2
2
F
1;
161
4t
.
(1 t)2
We may therefore write

1+x 1
, ;
2
2
X
n
1
n (1, x, 1)t = (1 t) 2 F1
n=0
1;
X
4t
=
Fn (x)tn ,
2
(1 t)
n=0
by (2), page 505. Hence n (1, x, 1) = Fn (x), (again).

Problem 4. Sylvester (1879) studied polynomials,

xn
1
.
n (x) =
2 F0 n, x; ;
n!
x
Show that
(1 t)x ext =
n (x)tn ,
n=0

(1 xt)c 2 F0 c, x; ;
t
1 xt
(c)n n (x)tn .
n=0
n (x) =

xn
1
F
.
n,
x;
;
2 0
n!
x
THen
n (x) =
n
X
(1)k n!xn (x)k (1)k xk
k=0
k!(n k)!n!
n
X
(x)k xnk
=
.
k!(n k)!
k=0
Hence
X
n=0
Next consider
n (x)tn
X
n
X
(c)k xnk tn
k!(n k)!
!
!
X xn t n
(x)n tn
=
n!
n!
n=0
n=0
x xt
= (1 t) e .
n=0 k=0
(c)n n (x)tn
X
n
X
(c)n (x)k xnk tn
n=0
k!(n k)!
n=0 k=0
(c)n+k (x)k xn tn+k

k!n!
n,k=0
X
X
(c + k)n (xt)n (c)k (x)k tk
=
=
n!
k=0 n=0
k=0
k!
(c)k (x)k t
.
k!(1 xt)c+k
Therefore

(c)n n (x)tn = (1 xt)c 2 F0 c, x; ;
n=0
t
1 xt

.
Problem 5. For Sylvesters polynomials of Exercise 4 find what properties you can
from the fact that n (x) is of Sheffer A-type zero.
Solution 5. For the n (x) of Exercise 4 recall that
(1 t)x ext =
n (x)tn .
n=0
Now (1 t)x = exp[x log(1 t)]. Hence
n (x)tn = exp[x{t log(1 t)}].
n=0
Therefore n (x) are of Sheffer A-type zero with A(t) = 1, H(t) = t log(1 t).
Problem 6. Show that Batemans Zn (x), the Legendre plynomial Pn (x), and the
Laguerre polynomial Ln (x) are related symbolically by
Zn (x) + Pn (2L(x) 1).
Solution 6. We know that Batemans Zn (x) has the generating relation
(1)
Zn (x)t = (1 t)
n=0

1 F1
1
4xt
; 1;
2
(1 t)2
and that the Laguerre polynomial satisfies

X

(c)n Ln (x)tn
xt
(1 t)c 1 F1 c; 1;
=
,
1t
n!
n=0
including the special case
X
( 21 )n Ln (x)v n
1
xv
; 1;
=
.
2
1v
n!
n=0

2
4t
1t
v
4t
In (2) put v =
. Then 1 v =
(1 + t)2
1+t
1 v (1 t)2
(2)
(1 v) 2 1 F1
163
and (2) becomes

1
(1 t)
(3)

(1 + t)1 F1
1
4xt
; 1;
2
(1 t)2

=
X
( 1 )k Lk (x)22k tk
2
k=0
k!(1 + t)2k
or
1
(1 t)

1 F1
1
4xt
; 1;
2
(1 t)2

=
X
( 1 )k 22k Lk (x)tk
2
k!(1 + t)1+2k
k=0
Therefore we have
Zn (x)tn
n=0
=
=
X
(1)n 22k ( 21 )k (1 + 2k)n Lk (x)tn+k
k!n!
n.k=0
n
X
(1) (n + 2k)!Lk (x)tn+k
n,k=0
n
X
X
n=0 k=0
(k!)2 n!
(1)n+k (n + k)!Lk (x) n
t .
(k!)2 (n k)!
We know from (3), page 285 that
n.n + 1;
1+x
.
2
Pn (x) = (1)n 2 F1
1;
Hence
Pn (2y 1)
n.n + 1;
= (1)n 2 F1
=
1;
n
X
(1)n+k (n + k)!y k
k=0
(k!)2 (n k)!
We therefore have, from
Zn (x) =
n
X
(1)n+k (n + k)!Lk (x)
k=0
(k!)2 (n k)!
= (1)n
n
X
(n)k (n + 1)k Lk (x)
k=0
the symbolic result

Zn (x) + Pn (2L (x) 1).
Problem 7. Show that Sister Celines polynomial

1
fn (x) = 2 F2 n, n + 1; 1, ; x
2
of equation (16), page 292, is such that
Z
ex fn (x)dx = (1)n (2n + 1).
0
k!k!

fn (x) = 2 F2

1
n, n + 1; 1, ; x .
2
Then
fn (x)tn =
n=0
X
n
X
(1)s (n + s)!xs tn
n=0 s=0
s!s!( 21 )s (n s)!
We also know from (19), page 373, that

s
X (1)s Lk (x)
xs
=
.
2
(s!)
k!(s k)!
k=0
Hence,
fn (x)tn
n=0
=
=
s
n X
X
X
(1)k+s (n + s)!Lk (x)tn
n=0 s=0 k=0
X
s
X
( 12 )s k!(s k)!(n s)!
(1)k+s (n + 2s)!Lk (x)tn+s

( 12 )s k!(s k)!n!
n,s=0 k=0
X (1)s (n + 2k + 2s)!Lk (x)tn+k+s
n,k,s=0
X
n
X
s!k!n!( 12 )s+k
(1)s (n + 2k + s)!Lk (x)tn+k

s!( 12 )k+s (n s)!k!
n,k=0 s=0
n, 1 + n + 2k;
n+k
X
1 (n + 2k)!Lk (x)t
.
=
F
1
k!( 12 )k n!
n,k=0
+ k;
2
By Example 5, page 119, we get
=
2 F1
n, 1 + n + 2k;
1
+ k;
2
n
1 = (1) (1 + 1 + 2k
( 12 + k)n
1
2
k)n
(1)n ( 23 + k)n
(1)n ( 23 )n+k ( 12 )k
=
.
( 12 + k)n
( 32 )k ( 12 )n+k
Then
fn (x)tn
n=0
X
(1)n ( 23 )n+k ( 12 )k (n + 2k)!Lk (x)tn+k
( 23 )k ( 12 )n+k k!( 12 )k n!
n,k=0
X
n
X
(1)nk ( 32 )n (n + k)!Lk (x)tn
=
.
( 23 )k ( 12 )n k!(n k)!
n=0 k=0
Therefore, since
(1)
( 32 )n
(n + 21 )
=
= 2n + 1,
( 12 )n
( 12 )
n
X
(n)k (n + 1)k Lk (x)
fn (x) = (1)n (2n + 1)
.
k!( 23 )k
k=0
165
Equation (1) will also be of use to us in later work. We now note that
Z
ex fn (x)dx = (1)n (2n + 1)
Z
n
X
(n)k (n + 1)k
k=0
k!( 32 )k
ex Lk (x)dx.
The integral in the sum is zero except for k = 0 and
ex L0 (x)dx = 1.
Hence
ex fn (x)dx = (1)n (2n + 1).
Problem 8. For Sister Celines fn (x) of Exercise 7 show that
n, n + 1, m, m + 1;
ex fn (x)fm (x)dx = (1)n+m (2n+1)(2m+1)4 F3
Z
0
n
(1) (2n + 1)(n)k (n + 1)k
x
e Lk (x)fn (x)dx =
k!( 32 )k
Z
0
3 3
1, , ;
2 2
1 ,
; 0 k n,
; k > n,
n, n + 1, k, k + 1;
ex fn (x)Zk (x)dx = (1)n+k (2n + 1)4 F3
3
1, 1, ;
2
1 .

1
Solution 8. Again we use fn (x) = 2 F2 n, n + 1; 1, ; x as in Exercise 7.
2
Recall that
fn (x) = (1)n (2n + 1)
(1)
n
X
(n)k (n + 1)k Lk (x)
k=0
k!( 32 )k
and, from Exercise 6, that

(2)
Zn (x) = (1)n
n
X
(n)k (n + 1)k Lk (x)
k=0
k!k!
At once
Z
0
ex fn (x)fm (x)dx = (1)n+m (2n+1)(2m+1)
Z
n X
m
X
(n)k (n + 1)k (m)s (m + 1)s
k=0 s=0
k!( 32 )k s!( 32 )s
Now, by the orthogonality property of Laguerre polynomials, and the fact that
Z
ex Ln2 (x)dx = 1, we get
0
ex Lk (x)Ls (x)dx.
min(n,m)
(n)k (n + 1)k (m)k (m + 1)k

k!( 32 )k k!( 32 )k
k=0
n, n + 1, m, m + 1;
1 .
= (1)n+m (2n + 1)(2m + 1)4 F3
3 3
1, , ;
2 2
ex fn (x)fm (x)dx
= (1)n+m (m + 1)(2m + 1)
Next consider
Z
ex Lk (x)fn (x)dx.
At once
ex Lk (x)fn (x)dx = 0; k > n,
from the orthogonality property of Lk (x).

If 0 k n,
ex Lk (x)fn (x)dx
= (1)n (2n + 1)
Z
n
X
(n)s (n + 1)s
s!( 32 )s
s=0
n
(1) (2n + 1)(n)k (n + 1)k
=
.
k!( 32 )k
ex Lk (x)Ls (x)dx
Finally, using (1) and (2) above we get
ex fN (x)Zk (x)dx
= (1)n+k (2n + 1)
Z
k
n X
X
(n)s (n + 1)s (k)i (k + 1)i
s!( 23 )s i!i!
s=0 i=0
min(n,k)
(n)s (n + 1)s (k)s (k + 1)s

s!( 32 )s s!s!
s=0
n, n + 1, k, k + 1;
1 .
= (1)n+k (2n + 1)4 F3
3
1, 1, ;
2
= (1)n+k (2n + 1)
Z
0
n, n + 1, k, k + 1;
ex Zn (x)Zk (x)dx = (1)n+k 4 F3
1 .
1, 1, 1, ;
Solution 9. Using equation (2) of Exercise 8 above, we get
ex Ls (x)Li (x)dx
Zn (x)Zk (x)dx
= (1)
n+k
167
Z
n X
k
X
(n)s (n + 1)s (k)i (k + 1)i
s=0 i=0
min(n,k)
s!s!i!i!
ex Ls (x)Li (x)dx
(n)s (n + 1)s (k)s (k + 1)s

s!s!s!s!
s=0
n, n + 1, k, k + 1;
1 .
= (1)n+k 4 F3
1, 1, 1;
= (1)n+k
Problem 10. Gottlieb introduced the polynomials

n (x; ) = en 2 F1 (n, x; 1; 1 e ).
Show that
X
n (x; )tn
= et 1 F1 (1 + x; 1; t(1 e )),
n!
n=0
n (x, + ) =
n
(e 1)n X n!(1 e+ )k k (x; )
,
en (1 e )n
k!(n k)!(e 1)k
k=0
est n (x; t)dt =
(s n)(s + x + 1)
,
(s + 1)(s + x n + 1)
(x n)n (x; ) = xn (x 1; ) ne n1 (x; ),

n[n (x; ) n1 (x; )] = (x + 1)[n (x + 1; ) n (x; )],
(x + n + 1)n (x; ) = xe n (x 1; ) + (n + 1)e n+1 (x; ),
and see Gottlieb for many other results on n (x; ).
Solution 10.
Problem 11. With fn denoting Sister Celines polynomials of Section 149, show
that
Z
x 2 (t x) 2 fn (; ; x)dx = Zn (t),
0
t
x 2 (t x) 2 fn (; ; x(t x))dx = Zn
0

1 2
t ,
4
and
Z
0

1
x (t x) fn ; ; x(t x) dx = Ln (t)Ln (t).
2
1
2
1
2
Solution 11. We shall use Theorem 37, page (??), to evaluate the integrals in
question.

1
Recall that fn (; ; x) = 2 F2 n, n + 1; 1, ; x .
2
1
1
1
Then use = , = , p = 2, q = 2, a1 = n, a2 = n + 1, b1 = 1, b2 = , c =
2
2
2
1, k = 1, s = 0 in Theorem 37 to get
1
n, n + 1, ;
2
1 1
=B
,
F
2 2
1
1, , 1;
2
n,
n
+
1;
1
1
( 2 )( 2 )
=
F
2 2
(1)
1, 1;
= Zn (t).
x 2 (t x) 2 fn (; ; x)dx
Next, make the same substitution except that s = 1 instead of s = 0. We thus

get
1 1
n, n + 1, , ;
2
2
= F
1 1 2
1, , , ;
2 2 2
n, n + 1;
t2
= 2 F2
4
1, 1;

1 2
= Zn
t .
4
x 2 (t x) 2 fn (; ; x(t x))dx
t2
4
Finally, consider
1
fn ; ; z
2
n, n + 1;
= 2 F3
1
In Theorem 37 we now use = , =
2
1
1, b2 = 1, b 3 = , c = 1, k = 1, s = 1, to
2

Z t
1
1
1
x 2 (t x) 2 fn ; ; x(t x) dx
2
0
1 1
1, , ;
2 2
z .
1
, p = 2, q = 2, a1 = n, a2 = n + 1, b1 =
2
get
( 21 )( 12 )
F
(1)
n, n + 1;
= Ln (t)Ln (t),
1
1, 1, ;
2
t2
4
by equation (9), page 509.

Recall Ramanujans theorem, Example 5, page 180,

1 x2
1F1 (; ; x)1 F1 (; ; x) = 2 F3 , ; , , + ;
.
2 2
2 4
Use = n, = 1, x = t to get

1
t2
,
1 F1 (n; 1; t)1 F1 (n; 1; t) = 2 F3 n, 1 + n; 1, , 1;
2
4
which is the result we used from page 509.
Problem 12. Define polynomials n (x) by
n (x) =
(c)n
2 F2 (n, c + n; 1, 1; x).
n!
Show that
n (x) =
n
X
(1)nk (c 1)nk (c)n+k (2k + 1)Zk (x)
(n k)!(n + k + 1)!
k=0
Solution 12. Let us define n (x) by

n (x) =
(c)n
2 F2 (n, c + n; 1, 1; x).
n!
Then
n (x) =
n
X
(1)k (c)n+k xk
k=0
k!(n k)!(k!)2
so
n (x)t
n=0
X
n
X
(1)s (c)n+s xs tn
n=0 s=0
(s!)3 (n s)!
(1)s (c)n+2s xs tn+s

.
(s!)3 n!
n,s=0
We know (page 199) that

xs = (s!)2
s
X
(s)k (2k + 1)Zk (x)
k=0
(s + k + 1)!
Then
n (x)t
n=0
X
s
X
(1)s+k (c)n+2s (2k + 1)Zk (x)tn+s
=
n!(s k)!(s + k + 1)!
n,s=0 k=0
X (1)s (c)n+2k+2s (2k + 1)Zk (x)tn+k+s

=
n!s!(s + 2k + 1)!
n,k,s=0
X
n
s
X
(1) (c)n+2k+s (2k + 1)Zk (x)tn+k
=
.
s!(n s)!(s + 2k + 1)!
s=0
n,k=0
We now have
169
n (x)tn
n=0
=
=
=
X
n,k=0
X
n,k=0
2 F1
n, c + n + 2k;
n+k
(c)n+2k (2k + 1)Zk (x)t

1
n!(2k + 1)!
2 + 2k;
(2 + 2k)(2 c)
(c)n+2k (2k + 1)Zk 9x)tn+k
(2 + 2k + n)(2 c n)
n!(2k + 1)!
(1)n (c 1)n (c)n+2k (2k + 1)!Zk (x)tn+k

(n + 2k + 1)!n!
n,k=0
n
nk
X X (1)
(c)n+k (c 1)nk (2k + 1)Zk (x)
(n k)!(n + k + 1)!
n=0 k=0
tn .
Hence
n (x) =
n
X
(1)nk (c 1)nk (c)n+k (2k + 1)Zk (x)
k=0
(n k)!(n + k + 1)!
Problem 13. Show that the F (t) of equation (10), page 286, can be put in the
form
v;
4zt
.
(1 z)2
F (t) = tb et (1 z)2v 1 F1
b + 1;
Solution 13. From equation (10), page 500, we get
n, 2v + n;
n
X
(2v)n z
t
F (t) = tb et
2 F2
1
n!
n=0
v + , b + 1;
2
n
X
X
(1)k n!(2v)n (2v + n)k tk z n
b t
=t e
k!(n k)!n!(v + 12 )k (b + 1)k
n=0 k=0
X (1)k (2v)n+2k tk z n+k

= tb et
k!n!(v + 12 )k (b + 1)k
n,k=0
X
X
(2v + 2k)n z n (1)k (2v)2k tk z k
= tb et
n!
k!(v + 12 )k (b + 1)k
k=0 n=0
X
(1)k tk z k 22k (v)k (v + 21 )k
= tb et
k!(v + 12 )k (b + 1)k (1 z)2v+2k
k=0
X
(1)k 22k (v)k tk z k
= tb (1 z)2v et
k!(b + 1)k (1 z)2k
k=0

4tz
b
2v t
= t (1 z) e 1 F1 v; b + 1;
.
(1 z)2
No problems in Chapter 19.
171
1
4
Problem 1. Show that 10 = 2q G3 .

1
10 = 2p 4 G
(1 q 2n )2
n=1
and that
G=
(1 q 2n ).
n=1
Hence
1
10 = 2q 4 G3 .
Problem 2. The following formulas are drawn from Section 168. Derive (9) and
(10):
42 12 (z) + 32 22 (z) = 22 32 (z),
32 12 (z) + 42 22 (z) = 22 42 (z),
22 12 (z) + 42 32 (z) = 32 42 (z),
22 22 (z) + 42 42 (z) = 32 32 (z).
(8)
(9)
(10)
(11)
Solution 2. We are given
42 12 (z) + 32 22 (z) = 22 32 (z).
Use the basic table with z||z + to obtain

2
(8)
42 22 (z) + 32 12 (z) = 22 42 (z)

or
32 12 (z) + 42 22 (z) = 22 42 (z).
1
In (9) use the basic table with z||z + to get
2
(9)
32 q 2 e2iz 42 (z) + 42 q 2 e2it 32 (z) = 22 q 2 e2iz 12 (z)

or
32 42 (z) 42 32 (z) = 22 12 (z)
from which we get
(10)
22 12 (z) + 42 32 (z) = 32 42 (z).
Problem 3. Use (8) and (10) of Exercise 2 and the equation 24 + 44 = 34 to show
that
14 (z) + 34 (z) = 24 (z) + 44 (z).
Solution 3. We may write (8) and (10) of Exercise 2 in the form

(8)
(10)
42 12 (z) 22 32 (z) = 32 22 (z),

22 12 (z) + 42 32 (z) = 32 42 (z).
We square each member and add to obtain

(44 + 24 )[14 (z) + 34 (z)] = 34 [24 (z) + 44 (z)].
But
24 + 34 = 34 ,
so we get
14 (z) + 34 (z) = 24 (z) + 44 (z).
Problem 4. The first of the following relations was derived in Section 169. Obtain
the other three by using appropriate
changes
of variable
and
the basic table, page

1
1
319. For example, change x to x + , or x to x + , etc.
2
2
22 1 (x + y)1 (x y) = 12 (x)22 (y) 22 (x)12 (y),
22 2 (x + y)2 (x y) = 22 (x)22 (y) 12 (x)12 (y),
22 3 (x + y)3 (x y) = 32 (x)22 (y) + 42 (x)12 (y),
22 4 (x + y)4 (x y) = 42 (x)22 (y) + 32 (x)12 (y).
Solution 4. We are given from Section 169 that
22 1 (x + y)1 (x y) = 12 (x)22 (y) 22 (x)12 (y).

In (A) change x to x +
and use the basic table to get
2
(A)
22 2 (x + y)2 (x y) = 22 (x)22 (y) 12 (x)12 (y).
Now in last equation above change x to x +

to obtain
2
1
22 q 2 e2ix 2 (x + y)3 (x y) = q 2 e2ix 32 (x)22 (y) + q 2 e2ix 42 (x)12 (y),

or
22 3 (x + y)3 (x y) = 32 (x)22 (y) + 42 (x)12 (y).

In last equation above change x to x +
to arrive at
2
22 4 (x + y)4 (x y) = 42 (x)22 (y) + 32 (x)12 (y).
173
Problem 5. Use the identities in Exercise 2 and the relation 24 + 44 = 34 to

transform the first relation of Exercise 4 into the first relation below. Then obtain
the remaining three relations with the aid of the basic table, page 319.
22 1 (x + y)1 (x y) = 42 (x)32 (y) 32 (x)42 (y),
22 2 (x + y)2 (x y) = 32 (x)32 (y) 42 (x)42 (y),
22 3 (x + y)3 (x y) = 22 (x)32 (y) + 12 (x)42 (y),
22 4 (x + y)4 (x y) = 12 (x)32 (y) + 22 (x)42 (y).
Solution 5. From Exercise 4 we obtain
22 1 (x + y)1 (x y) = 12 (x)22 (y) 22 (x)12 (y).
From (10) and (11) of Exercise 2 we get

12 (x) = 22 32 42 (x) 42 32 (x)

22 (y) = 22 32 32 (y) 42 42 (y) .
Hence we obtain
22 1 (x + y)1 (x y)

= 24 34 42 (x)32 (y) + 44 32 (x)42 (y) 32 44 32 (x)32 (y) + 42 (x)42 (y)
22 34 42 (y)32 (x) + 44 32 (y)42 (x) 22 42 32 (y)32 (x) + 42 (y)42 (x)
= 24 (34 44 ){42 (x)32 (y) 32 (x)42 (y)} .
Now 34 44 = 24 so that we may conclude that

22 1 (x + y)1 (x y) = 42 (x)32 (y) 32 (y) 32 (x)42 (y).
(A)
We shall now transform (A) by means of the basic table. Note that (A) is the
first of the required results
5.

inExercise
In (A) replace x by x +
to get
2
22 2 (x + y)2 (x y) = 32 (x)32 (y) 42 (x)42 (y).

In (B) replace x by x +
to obtain
2
(B)
q 2 e2ix 22 3 (x + y)3 (x y) = q 2 e2ix 22 (x)32 (y) + q 2 e2ix 12 (x)42 (y),

or
22 3 (x + y)3 (x y) = 22 (x)32 (y) + 12 (x)42 (y).

In (C) replace x by x +
to arrive at
2
(C)
(D)
22 4 (x + y)4 (x y) = 12 (x)32 (y) + 22 (x)42 (y).
Equations (A), (B), (C), (D) are those required.
Problem 6. Use equation (10) of Exercise 2 and the first relation of Exercise 5 to
obtain the first relation below; then use the basic table to get the remaining three
results.
32 1 (x + y)1 (x y) = 12 (x)32 (y) 32 (x)12 (y),
32 2 (x + y)2 (x y)22 (x)32 (y) 42 (x)12 (y),
32 3 (x + y)3 (x y) = 32 (x)32 (y) + 12 (x)12 (y),
32 4 (x + y)4 (x y) = 42 (x)32 (y) + 22 (x)12 (y).
Solution 6. From Exercise 5 we use
22 1 (x + y)1 (x y) = 42 (x)32 (y) 32 (x)42 (y).
From Exercise 2 we use equation (10) in the form
(10)
42 (x) = 32 [22 12 (x) + 42 32 (x)].
Thus we obtain
22 1 (x+y)1 (xy) = 32 [22 12 (x)32 (y)+42 32 (x)32 (y)22 32 (x)12 (y)42 32 (x)32 (x)].
Two terms drop out and we are left with
(A)
32 1 (x + y)1 (x y) = 12 (x)32 (y) 32 (x)12 (y),
which was desired.

to get
2
32 2 (x + y)2 (x y) = 22 (x)32 (y) 42 (x)12 (y).

to get
2
(B)
(C)32 3 (x + y)3 (x y) = 32 (x)32 (y) + 12 (x)12 (y),

1
from which we have removed the common factor q 2 e2ix .
to get
2
(D)
32 4 (x + y)4 (x y) = 42 (x)32 (y) + 22 (x)12 (y).
Equations (A), (B), (C), (D) are the required ones.
Problem 7. Use the first relation of Exercise 6 and the identities from Exercise 2
to obtain the frist relation below. Derive the other three relations from the first one.
32 1 (x + y)1 (x y) = 42 (x)22 (y) 22 (x)42 (y),
32 2 (x + y)2 (x y) = 32 (x)22 (y) 12 (x)42 (y),
32 3 (x + y)3 (x y) = 22 (x)22 (y) + 42 (x)42 (y),
175
32 4 (x + y)4 (x y) = 12 (x)22 (y) + 32 (x)42 (y).

Solution 7. From (9) and (11) of Exercise 2 we may write
12 (x) = 32 [22 42 (x) 42 22 (x)]
and
32 (y) = 32 [22 22 (y) + 42 42 (y)],
which we substitute into equation (A) of Exercise 5 above.
The result is
32 1 (x + y)1 (x y)
= 34 [24 42 (x)22 (y) 44 22 (x)42 (y) + 22 42 {42 (x)42 (y) 22 (x)22 (y)}]
34 [24 42 (y)22 (x) 44 22 (y)42 (x) + 22 42 {42 (y)42 (x) 22 (y)22 (x)}].
We thus obtain
32 1 (x + y)1 (x y) = 34 [(24 + 44 ){42 (x)22 (y) 22 (x)42 (x)42 (y)}].
Since 24 + 44 = 34 , we arrive at the desired result
32 1 (x + y)1 (x y) = 42 (x)22 (y) 22 (x)42 (y).

to obtain
2
(A)
32 2 (x + y)2 (x y) = 32 (x)22 (y) 12 (x)42 (y).

1
and remove the factor q 2 e2ix to get
2
(B)
32 3 (x + y)3 (x y) = 22 (x)22 (y) + 42 (x)42 (y).

to obtain the formula of the requires results:
2
(C)
(C)
32 4 (x + y)4 (x y) = 12 (x)22 (y) + 32 (x)42 (y).
Problem 8. Use equation (10) of Exercise 2 and the first relation of Exercise 6
to obtain the first relation below. Derive the others with the aid of the basic table,
page 319.
42 1 (x + y)1 (x y) = 12 (x)42 (y) 42 (x)12 (y),
42 2 (x + y)2 (x y) = 22 (x)42 (y) 32 (x)12 (y),
42 3 (x + y)3 (x y) = 32 (x)42 (y) 22 (x)12 (y),
42 4 (x + y)4 (x y) = 42 (x)42 (y) 12 (x)12 (y).
Solution 8. From (10) of Exercise 2 we g et

32 (y) = 42 [32 42 (y) 22 12 (y)]
and a similar result in x. The first equation in Exercise 6 is
32 1 (x + y)1 (x y) = 12 (x)32 (y) 32 (x)12 (y).
Then two applications of (10) yield
32 1 (x+y)1 (xy) = 42 [32 12 (x)42 (y)22 12 (x)12 (y)]42 [32 42 (x)12 (y)22 12 (x)12 (y)]
In the above, two terms drop out and 32 cancels out to yield
42 1 (x + y)1 (x y) = 12 (x)42 (y) 42 (x)12 (y).

and see the basic table to get
2
(A)
2 (x + y) (x y) = 22 (x)42 (y) 32 (x)12 (y).

4 2 2
1
and remove the common factor q 2 e2ix to obtain
2
(B)
42 3 (x + y)3 (x y) = 32 (x)42 (y) 22 (x)12 (y).

and thus arrive at
2
(C)
(D)
42 4 (x + y)4 (x y) = 42 (x)42 (y) 12 (x)12 (y).
Problem 9. The first relation below was derived in Section 169. Obtain the other
three.
42 1 (x + y)1 (x y) = 32 (x)22 (y) 22 (x)32 (y),
42 2 (x + y)2 (x y) = 42 (x)22 (y) 12 (x)32 (y),
42 3 (x + y)3 (x y) = 42 (x)32 (y) 12 (x)22 (y),
42 4 (x + y)4 (x y) = 32 (x)32 (y) 22 (x)22 (y).
Solution 9. From Section 169, equation (7), page 371, we obtain
42 1 (x + y)1 (x y) = 32 (x)22 (y) 22 (x)32 (y).

to get
2
(A)
42 2 (x + y)2 (x y) = 42 (x)22 (y) 12 (x)32 (y).

1
and remove the factor q 2 e2ix to arrive at
2
(B)
42 3 (x + y)2 (x y) = 42 (x)32 (y) 12 (x)22 (y).

to obtain
2
(C)
(D)
42 4 (x + y)4 (x y) = 32 (x)32 (y) 22 (x)22 (y).
177
Problem 10. Use the method of Section 169 together with the basic table, page
319, to derive the following results, one of which was obtain in Section 169.
3 4 1 (x + y)2 (x y) = 1 (x)2 (x)3 (y)4 (y) + 1 (y)2 (y)3 (x)4 (x),
2 4 1 (x + y)3 (x y) = 1 (x)2 (y)3 (x)4 (y) + 1 (y)2 (x)3 (y)4 (x),
2 3 3 (x + y)4 (x y) = 1 (x)2 (y)3 (y)4 (x) + 1 (y)2 (x)3 (x)4 (y),
2 3 2 (x + y)3 (x y) = 2 (x)3 (x)2 (y)3 (y) 1 (x)4 (x)1 (y)4 (y),
2 4 2 (x + y)4 (x y) = 2 (x)4 (x)2 (y)4 (x) 1 (x)3 (x)1 (y)3 (y),
3 4 3 (x + y)4 (x y) = 3 (x)4 (x)3 (y)4 (y) 1 (x)2 (x)1 (y)2 (y).
Solution 10. For brevity, since we know how to discover such formulas, let us
consider the function
1 (x)2 (x)3 (y)4 (y) + 1 (y)2 (y)3 (x)4 (y)
.
1 (x + y)2 (x y)
we shall show that 1 (x) is an elliptic function of order less than two, and then
obtain its constant value.
Let N1 (x) = the numerator of 1 (t) and D1 (x) be the denominator of 1 (x).
Then at once the basic table yields
1 (x) =
N1 (x + ) = 2 (x)1 (x)3 (y)4 (y) + 1 (y)2 (y)4 (x)3 (x) = N1 (x)

and
D1 (x + ) = 1 (x + y)[2 (x y)] = D1 (x).
Hence
1 (x + ) = 1 (x).
The periods and are also those of the zeros of the four theta functions.
Hence the only poles of 1 (x) in a cell are simple ones at x = y = 0 and x y = .
2
But x = y is also a zero of the numerator because
N1 (y) = 1 (y)2 (x)3 (y)4 (y) + 1 (y)2 (y)3 (y)4 (y) = 0.
Hence 1 (x) is an elliptic function order less than two and is constant. Now
0 + 1 (y)2 (y)3 4
= 3 4 ,
1 (y)2 (y)
since 2 (y) is an even function. Thus we obtain the first of the derived relations,
1 (0) =
(A)
3 4 1 (x + y)2 (x y) = 1 (x)2 (x)3 (y)4 (y) + 1 (y)2 (y)3 (x)4 (x).
Next let us consider

1 (x)2 (y)3 (x)4 (y) + 1 (y)2 (x)3 (y)4 (x)
.
1 (x + y)3 (x y)
Let as usual the numerator be N2 (x), the denominator D2 (x). Then
2 (x) =
N2 (x + ) = 1 (x)2 (y)3 (x)4 (y) 1 (y)2 (x)3 (x)4 (x),

N2 (x + ) = N 2(x).
Also
D2 (x + ) = 1 (x + y)3 (x y) = D2 (x + ).
Hence
2 (x + ) = 2 (x).
Next
N2 (x + ) = q 2 e4ix N2 (x)
and
D2 (x + ) = q 2 e2i(x+y) e2i(xy) D2 (x) = q 2 e4ix D2 (x).
Hence also
2 (x + ) = 2 (x).
In a cell, with periods , , D2 (x) has simple zeros at x + y = 0 and x y =

+
. But, for x = y,
2
2
N2 (y) = 1 (y)2 (y)3 (y)4 (y) + 1 (y)2 (y)3 (y)4 (y) = 0.
Therefore 2 (x) is an elliptic function of order less than two. The constant value
of 2 (x) is
0 + 1 (y)3 (y)4
= 2 4 .
1 (y)3 (y)
We may now conclude the validity of the equation
2 (0) =
(B)
2 4 1 (x + y)3 (x y) = 1 (y)2 (y)3 (x)4 (y) + 1 (y)2 (x)3 (y)4 (x).
From Section 169 we quote

(C)
2 3 1 (x + y)4 (x y) = 1 (x)2 (y)3 (y)4 (x) + 1 (y)2 (x)3 (x)4 (y).
Equations (A), (B),

and (C) are the first three of our desired results. In (C)

to get
replace x by x +
2
(D)
2 3 2 (x + y)3 (x y) = 2 (x)2 (y)3 (y)3 (x) 1 (y)1 (x)4 (x)4 (y),
179
which is the
equation, slightly rewritten, of the desired results. In (B)
fourth

replace x by x +
to find that
2
2 4 2 (x + y)4 (x y) = 2 (x)2 (y)4 (x)4 (y) 1 (y)1 (x)3 (y)3 (x),
(E)
which
is a rewritten form of a desired result. Finally, turn to (A) and replace x

by x +
to get
2
1
3 4 iq 2 ei(x+y) ei(xy) 4 (x+y)3 (xy) = iq 2 e2ix 4 (x)3 (x)3 (y)4 (y)+iq 2 e2ix 1 (y)2 (y)2 (x)1 (x),
or
3 4 4 (x + y)3 (x y) = 3 (x)4 (x)3 (y)4 (y) + 1 (x)2 (x)1 (y)2 (y).
Now change y to (y) to arrive at the desired result
(F )
3 4 3 (x + y)4 (x y) = 4 (x)4 (x)3 (y)4 (y) 1 (x)2 (x)1 (y) 2(y).
Problem 11. Use the results in Exercises 4-10 above to show that
23 2 (2x) = 24 (x) 14 (x) = 34 (x) 44 (x),
33 3 (2x) = 14 (x) + 34 (x) = 24 (x) + 44 (x),
43 4 (2x) = 44 (x) 14 (x) = 34 (x) 24 (x),
32 2 2 (2x) = 22 (x)32 (x) 12 (x)42 (x),
42 2 2 (2x) = 22 (x)42 (x) 12 (x)32 (x),
22 3 3 (2x) = 22 (x)32 (x) + 12 (x)42 (x),
42 3 3 (2x) = 32 (x)42 (x) 12 (x)22 (x),
22 4 4 (2x) = 12 (x)32 (x) + 22 (x)42 (x).
32 4 4 (2x) = 12 (x)22 (x) + 32 (x)42 (x).
Solution 11. We shall use y = x in various equations of Exercises 4-10. From
Exercise 4, second equation, we get
22 2 (2x) = 24 (x) 14 (x).
From Exercise 4 third equation, we get
22 3 3 (2x) = 22 (x)32 (x) + 12 (x)42 (x).
From Exercise 4, fourth equation, we get
22 4 4 (2x) = 22 (x)42 (x) + 12 (x)32 (x).
From Exercise 5, equations 1,2,3 in order, we obtain

23 2 (2x) = 34 (x) 44 (x),
22 3 3 (2x) = 22 (x)32 (x) 12 (x)42 (x),
22 4 4 (2x) = 12 (x)32 (x) + 22 (x)42 (x).
Next from Exercise 6, equations 2,3,4, we get
2 32 2 (2x) = 22 (x)32 (x) 12 (x)42 (x),
33 3 (2x) = 34 (x) + 14 (x),
32 4 4 (2x) = 32 (x)42 (x) + 12 (x)22 (x).
From equations 2,3,4 of Exercise 7 we obtain
2 32 2 (2x) = 22 (x)32 (x) 12 (x)42 (x),
33 3 (2x) = 24 (x) + 44 (x),
32 4 4 (2x) = 12 (x)22 (x) + 32 (x)42 (x).
From equations 2,3,4 of Exercise 8 we get
(
2 42 2 (2x) = 22 (x)42 (x) 1 x)32 (x),

3 42 3 (2x) = 32 (x)42 (x) 12 (x)22 (x),
43 4 (2x) = 44 (x) 14 (x).
From equations 2,3,4 of Exercise 9 we get
(
2 42 2 (2x) = 22 (x)4 x) 12 (x)32 (x),

3 42 3 (2x) = 32 (x)42 (x) 12 (x)22 (x),
43 4 (2x) = 34 (x) 24 (x).
We have now derived all 12 equations of Exercise 11, but we try Exercise 10 to
see whether from each of the first three equations
2 3 4 1 (2x) = 21 (x)2 (x)3 (x)4 (x),
(which we had in text), and
22 3 2 (2x) = 22 (x)32 (x) 12 (x)42 (x),
2 42 2 (2x) = 22 (x)42 (x) 12 (x)32 (x),
and
181
3 42 3 (2x) = 32 (x)42 (x) 12 (x)22 (x).

The last two appear as answers to Exercise 11, parts 5 and 7.
Problem 12. Use the method of Section 168 to show that
1
3 (z| ) = (i ) 2 exp
z2
i

3

z t
.

From the above identity, obtain corresponding ones for the other theta functions
with the aid of the basic table, page 319.
Solution 12. Consider the function
(z) =
exp
3 (z| )

.
3 ( z | 1
)
z2
|
First
the
theta functions are analytic for all finite z since Im( ) > 0 implies
1
Im
> 0.
To get the zeros of the denominator, recall that 3 (y|t) has its zeros all simple
ones located at
t
+
+ n + mt
2
2

z 1
for integral n and m. Then 3
has its zeros all simple ones located at

m
=
+ n
2
2
or at
y=
z=
+ n m,
2
2
which is equivalent to
z=

+
+ n1 + m1 ,
2
2

z 1
for integral n1 and m1 . Therefore the zeros of 3 (z| ) and those of 3

coincide. Hence the function (z0 has no singular points in the finite plane.
To show that (z) has the desired two periods, consider (z + ) and (z + ).
(z + ) =
exp
3 (z + | )

.
1
3 z+
|
(z+)2
i
Recall that (basic table) 3 (y + , t) = 3 (y|t) and that

3 (y + t|t) = eit e2iy 3 (y|t).
Since 3 (z) is an even function of z, we may now write
(z + )
3 (z| )

exp
exp
exp i
3 z | 1
3 (z| )

=
z2
i
i
exp i
exp 2iz
exp
exp exp
3 (z| )

=
z2
exp i 3 z | 1
= (z).
=
z2
i
2z
i
2iz
3 z |
Thus (z) has a period in z. Next consider (z + ) :
(z + ) =
3 (z + | )

=
(z+ )2
z+
1
exp
exp
3
i
z2
i
ei e2iz 3 (z| )

.
3 z + | 1
exp
i
We thus obtain
|psi(z + ) =
exp
3 (z| )

= (z).
3 z | 1
z2
i
Therefore ( ) also has the period in z.

We can now conclude that (z) is an elliptic function of order less than two. It
is a constant with the value
c = (z) = (0) =
3 (0| )
3 (0| )
=
.
e0 3 (0| 1 )
3 (0| 1 )
We shall evaluate c later. At present we have

(A)
3 (z| ) = c exp
z2
i

z 1
3
,

upon which we shalluse our basic table.
In (A) replace z by z +
to get
2

2
(z +
z
1

2 )
3
+
.
3 z +
= c exp
2
i
We conclude that

(B)
2 (z| ) = c exp
z2
i

4

z 1
.

Return to (A) and replace z by z +
to get
2

(z + 2 )2
1

z
3
+
3 z + = c exp

2
i
or
183

z
z
z2
1 1
exp
3 +
exp

i
4i
2
i

2
i
z
z
iz
z 1
exp
exp
= c exp
exp
exp
2
4
i
i
4i
2
z
z 1
= c exp
2
.

i
4 (z| )
= c exp
Hence we arrive at

(C)
z2
i

z 1
2
.

4 (z| ) = c exp

Now in (C) replace z by z +
. We thus obtain
2

2
(z +

1
z
2 )
4 z +
2
+
,
= c exp
2
i
or

i
i exp
4

exp(iz)1 (z| ) = c exp
z2
i

exp
z
c
exp

4i

z 1
.
(1)1

we arrive in this way at

z 1
(D)
1 (z| ) = ic1
.

We still need to find the value of c. From (D) we get

ic 0 z 1
.
(E)
= 1

We wish to put z = 0 in (E) and use the known result that 10 = 2 3 4 . From
(E) we thus get
10 (z| )
2 (0| )3 (0| )4 (0| ) =
ic
2

0|

3
0|

4
or
2 (0| )
3 (0| )
4 (0| )
ic
1
1
1 = .
4 (0| ) 3 (0| ) 2 (0| )
With the aid of (A), (B), (C), we may conclude that
ccc =
ic
,
or
c2 =
1
Hence c = (i ) 2 or c = (i ) 2 .
Now 3 (z) was defined by
1
.
i
0|

,
3 (z| ) = 1 + 2
ein
cos(2nz).
n=1
It follows that, if is pure imaginary, then

3 (0| ) = 1 + 2
en
(i )
>0
n=0
and, since 1| is also then pure imaginary,

1
3 0|
> 0.
We may therefore conclude that c = (i ) 2 .

Final result of our work is that the set of identities
2

1
z
z 1
3 (z| ) = (i ) 2 exp
3
,

i

2
1
z
z 1
4
,
2 (z| ) = (i ) 2 exp

i
2

1
z
z 1
4 (z| ) = i(i ) 2 exp
2
,

i

2
1
z
z 1
1 (z| ) = i(i ) 2 exp
1
.

i
With regard to the usefulness of these formulas, note that q = exp(i ) so that
|q| = exp[Im( )]. For Im( ) > 0, |q| < 1 and we have convergence of the series
definitions of the theta functions but that convergence is slow if Im( ) is small so
that |q| is near unity.
Since

1
Im( )
Im
=
,
| |2
the parameter ( 1 ) will be useful in computations with Im(t) small and positive.
Problem 1. Derive the preceding addition theorem (7) by the method of Section
179. You may use results from the exercises at the end of Chapter 20.
Solution 1. We seek an addition theorem for cn(u). Now
cn(u) =
4 2 (u32 )
.
2 4 (u32 )
At the end of Exercise 5, Chapter 20 we obtained

42 4 (x + 4)4 (x y) = 42 (x)42 (y) 12 (x)12 (y).
185
The next to last equation in Exercise 10, Chapter 20, is

2 4 2 (x + y)4 (x y) = 2 (x)4 (x)2 (y)4 (y) 1 (x)3 (x)1 (y)3 (y).
We divide the last equation above by the one preceding it to get
2 (x)4 (x)2 (y)4 (y) 1 (x)3 (x)1 (y)3 (y)
42 (x)42 (y) 12 (x)12 (y)
2 (x) 2 (y)
1 (x) 3 (x) 1 (y) 3 (y)
4 (x) 4 (y) 4 (x) 4 (x) 4 (y) 4 (y)
=
.
2 (x) 2 (y)
1 12 (x) 22 (y)
2 (x + y)
4 (x + y)
Now use x = u32 and y = v32 to obtain

2 2
cn(u + v) =
4 4
2
2
4 cn(u) 4 cn(v)
or
cn(u + v) =
2
3
2
3
3 sn(u) 4 dn(u) 3 sn(v) 4 dn(v)
2
2
2
sn2 (u) 22 sn2 (v)

32
3
cn(u)cn(v) sn(u)dn(u)sn(v)dn(v)
,
1 k 2 sn2 (u)sn2 (v)
as desired.
Problem 2. Derive preceding (8) by the method of Section 179, with the aid of
results from the exercises at the end of Chapter 20.
dn(u) =
4 3 (u32 )
2
,k = 2.
3 4 (u32 )
3
From the last equation of Exercise 10, Chapter 20 we get

3 4 3 (x + y)4 (x y) = 3 (x)4 (x)3 (y)4 (y) 1 (x)2 (x)1 (y)2 (y)
and from the last equation of Exercise 8 Chapter 20 we get
42 4 (x + y)4 (x y) = 42 (x)42 (y) 12 (x)12 (y).
The two equations about at once yield
3 3
dn(u + v)
4 4
3
3
4 dn(u) 4 dn(v)
2
2
2
2
3 sn(u)) 4 cn(u) 3 sn(v) 4 sn(v)
.
2
2
2
sn2 (u) 22 sn2 (v)

32
3
Hence we arrive at the desired result,

dn(u + v) =
Z
dn(u)dn(v) k 2 sn(u)cn(u)sn(v)cn(v)
.
1 k 2 sn2 (u)sn2 (v)
1
cn3 (x)dn(x)dx = sn(x) sn3 (x) + c.
3
Solution 3. We wish to evaluate

Z
We know that
cn3 (x)dn(x)dx.
d
dn(x) = cn(x)dn(x)
dx
and that
cn2 (x) = 1 sn2 (x).
Hence
Z
cn3 (x)dn(x)dx
[1 sn2 (x)]cn(x)dn(x)dx
1
= sn(x) sn2 (x) + c.
3
Problem 4. Show that if g(x) and h(x) are any two different ones of the three
functions sn(x), cn(x), dn(x), and if m is a non-negative integer, you can perform
the integration
Z
g 2m+1 (x)h(x)dx.
Z
A=
g 2m+1 (x)h(x)dx.
We know that
d
sn(x) = cn(x)dn(x);
dx
d
cn(x) = sn(x)dn(x);
dx
d
dn(x) = k 2 sn(x)cn(x).
dx
Now let g(x) and h(x) be any two different ones of sn(x), cn(x), dn(x). Let (x) be
the other of the three functions. Then
g(x)h(x)dx = c1 d[(x)],
in which c1 = 1, 1 or K 2 according to whether (x) is sn(x), cn(x), or dn(x).
Now we also know that
sn2 (x) + cn2 (x) = 1,
k 2 sn2 (x) + dn2 (x) = 1,
dn2 (x) k 2 cn2 (x) = x2 .
Therefore the square of any one of sn(x), cn(x), dn(x) is a linear function of the
square of any other one of them. Hence we may write
g 2 (x) = a1 + b1 2 (x).
We know have
Z
g 2m+1 (x)h(x)dx
[g 2 (x)]m g(x)h(x)dx
Z
= c1 [a1 + b1 2 (x)]m d(x),
which is integrable as a sum of powers.

Z
1
sn(x)dx = log[dn(x) kcn(x)] + c.
k
Z
sn(x)dx.
Z
Z
sn(x)dx =
Z
=
Z
=
187
We may write
sn(x)[dn(x) kcn(x)]
dx
dn(x) kcn(x)
sn(x)dn(x) ksn(x)cn(x)
dx
dn(x) kcn(x)
1
d[dn(x)] + k d[dn(x)]
Z dn(x) kcn(x)
1
d[dn(x)] k[cn(x)]
=
k
dn(x) kcn(x)
1
= log[dn(x) kcn(x)] + c.
k

dn(2x) kcn(2x) =
(1 k)[1 + ksn2 (x)]

.
1 ksn2 (x)
Solution 6. We wish to show that

(1 k)[1 + ksn2 (x)]
.
1 ksn2 (x)
Now from the addition formulas of Exercises 1 and 2 we obtain
dn(2x) kcn(2x) =
dn(2x) =
dn2 (x) k 2 sn2 (x)dn2 (x)

1 k 2 sn4 (x)
and
cn2 (x) sn2 (x)dn2 (x)

.
1 k 2 sn4 (x)
Let us put each of the above in terms of sn(x). We get
cn(2x) =
dn(2x) =
1 k 2 sn2 (x) k 2 sn2 (x)[1 sn2 (x)]

1 2k 2 sn2 (x) + k 2 sn4 (x)
=
.
1 k 2 sn4 (x)
1 k 2 sn4 (x)
cn(2x) =
1 2k 2 sn2 (x) + k 2 sn4 (x)

1 sn2 (x) sn2 (x)[1 x2 sn2 (x)]
=
1 k 2 sn4 (x)
1 k 2 sn4 (x)
Then
dn(2x) kcn(2x)
1 k 2 sn2 (x) k 2 sn2 (x) + k 2 sn4 (x) k + ksn2 (x) + ksn2 (k) k 3 sn4 (x)
1 k 2 sn4 (x)
1 + ksn2 (x) + ksn2 (x)[1 + ksn2 (x)] k[1 + ksn2 (x)] k 2 sn2 (x)[1 + ksn2 (x)]
=
[1 ksn2 (x)][1 + ksn2 (x)]
1 + ksn2 (x) k k 2 sn2 (x)
=
1 ksn2 (x)
1 k + k(1 k)sn2 (x)
=
.
1 ksn2 (x)
=
Hence
dn(2x) kcn(2x) =
(1 k)[1 + ksn2 (x)]

.
1 ksn2 (x)

Solutions To Rainville's Special Functions - Tom Cuchta, Leon Hall, Sylvester Pagano, Compiled 2 June 2015

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Solutions To Rainville's Special Functions - Tom Cuchta, Leon Hall, Sylvester Pagano, Compiled 2 June 2015

Caricato da

Copyright:

Formati disponibili

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

Figure 1. Photograph of Earl D. Rainville from the University of

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

Figure 2. These photographs are of Sylvester J. Pagano as a

SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

outstanding teacher and he kept building his knowledge of mathematics throughout

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

Problem 2. Show that

Solution 2. (Solution by Leon Hall)

Problem 3. Show that

Solution 3. (Solution by Leon Hall)

Problem 4. Investigate the product

Solution 4. (Solution by Leon Hall) Let Pn =

SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

So we have shown by induction that

which is a geometric series converging to (z + 1)

Solution 5. (Solution by Leon Hall) Let Pn =

Thus, because {Sn } diverges, so does {Pn }.

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

Case 1: n is even. Then

SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

Problem 10. Show that

is absolutely convergent for all finite z with

the usual convention at z = 0.

converges absolutely and uniformly for all finite z by Theorems 3 and 4.

Problem 11. Show that if c is not a negative integer,

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

Solution 11. Let

Problem 2. Show that 0 ( 21 ) = ( + 2 log 2) .

10 SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

Solution 2. By Problem 1, we know that

and so, algebra yields

et tz1 dt, for R(z) > 0, integration by parts yields

Solution 3. From (z) =

Problem 4. Show that (z) = lim nz B(z, n).

Solution 4. From page 28 (1), we know

Problem 5. Derive the following properties of the beta function:

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

Solution 5. (a) We know B(p, q) =

(note: p q and q p is this the symmetric property?)

12 SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

Problem 6. Show that for positive integral n, B(p, n + 1) =

Solution 6. For integer n and using Theorem 9 (pg. 23),

(1 + x)p1 (1 x)q1 dx.

(1 + x)p1 (1 x)q1 dx. Now let y =

1, 1 x = 2 2y = 2(1 y). Hence

= 2p+q1 B(p, q).

Problem 8. Show that for 0 k n,

Solution 8. Consider ()nk for 0 k n. Then

Note for = 1, that (n k)! =

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

Problem 9. Show that if is not an integer,

P (y)dy when P (y) = y byc

Solution 10. To evaluate

integer so that m 0. If m x < m + 1, then bxc = m and

14 SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA

Problem 13. Show that

and thus conclude that

SOLUTIONS TO RAINVILLES SPECIAL FUNCTIONS (1960)

This establishes the first set of equalities.

16 SOLUTIONS BY SYLVESTER J. PAGANO AND LEON HALL; EDITED BY TOM CUCHTA