630 CHAPTER 32. SOLUTIONS NEAR A REGULAR SINGULAR POINT

Without loss of generality, assume z is positive on (a,b). If it isn’t, multiply by −1 tomake this happen. The solution z is a linear combination of sines and cosines. It can bewritten in the form

z = Acos(s−φ)

and so z′ (a)> 0 and z′ (b)< 0.If y has no zeros on [a,b] , then again, without loss of generality, let y be positive on

[a,b].z′′y− y′′z+

(1−(e2s−ν

2))yz = 0

Thus on the open interval (a,b) ,

W (y,z)′ =((

e2s−ν2)−1

)yz > 0

where W (y,z) is the Wronskian. It follows from the mean value theorem that W (y,z)(a)<W (y,z)(b). Then ∣∣∣∣∣ y(a) 0

y′ (a) z′ (a)

∣∣∣∣∣<∣∣∣∣∣ y(b) 0

y′ (b) z′ (b)

∣∣∣∣∣positive = y(a)z′ (a)< y(b)z′ (b) = negative,

a contradiction. ■For the purposes of this book, this will suffice. The main message is that there are two

independent solutions, one bounded near 0 and the other unbounded as described above.Both oscillate about 0 and have infinitely many zeros. In many applications, the unboundedone is of no interest based on physical considerations.

32.7 Other Properties of Bessel FunctionsRecall that for m a nonnegative integer,

Jm (x) =∞

∑k=0

(−1)k

k!(k+m)!

( x2

)2k+m

and that if −m is negative,J−m (x) = (−1)m Jm (x)

This Jm (x) was the bounded solution for the Bessel equation. Note that an infinite sum ofthese functions is absolutely convergent. Indeed,

|Jm (x)| ≤∞

∑k=0

1k!m!

(∣∣∣ x2

∣∣∣)2k+m≤ 1

m!

∣∣∣ x2

∣∣∣m ∞

∑k=0

1k!

(( x2

)2)k

=1

m!

∣∣∣ x2

∣∣∣m exp(x2/4

)(32.22)

Therefore, it is permissible to sum the various series which result in what follows in anyorder desired.

Now for t ̸= 0,

ext2 =

∑l=0

1l!

( x2

)lt l , e−

x2t =

∑k=0

1k!

(−1)k( x

2

)kt−k

630 CHAPTER 32. SOLUTIONS NEAR A REGULAR SINGULAR POINTWithout loss of generality, assume z is positive on (a,b). If it isn’t, multiply by —1 tomake this happen. The solution z is a linear combination of sines and cosines. It can bewritten in the formz=Acos(s— @)and so z’ (a) > 0 and z'(b) < 0.If y has no zeros on [a,b], then again, without loss of generality, let y be positive on(a, b}.Zly—y" z+ (1 _ (Ca _ v?)) yz —0Thus on the open interval (a,b) ,W (y,z)' = ((e —v?) — 1) yz > 0where W (y,z) is the Wronskian. It follows from the mean value theorem that W (y,z) (a) <W (y,z) (b). Thenyi(b) z(b)positive = y(a)z’ (a) < y(b)z’ (b) = negative,a contradiction. MiFor the purposes of this book, this will suffice. The main message is that there are twoindependent solutions, one bounded near 0 and the other unbounded as described above.Both oscillate about 0 and have infinitely many zeros. In many applications, the unboundedone is of no interest based on physical considerations.32.7 Other Properties of Bessel FunctionsRecall that for m a nonnegative integer,and that if —m is negative,Join (X) = (-1)"Jin (2)This J; (x) was the bounded solution for the Bessel equation. Note that an infinite sum ofthese functions is absolutely convergent. Indeed,oo oo kSamal)" <malsl Ba ((3) )exp (x°/4) (32.22)Jim (x)|slat”Therefore, it is permissible to sum the various series which result in what follows in anyorder desired.Now for t 4 0,fava) ee Pacug)