Assignment 4 – Solutions
1. Note that 4x= 0 when x =0, so x =0 is a singular point. Dividing
4xy ′′+ 2y ′+ y =0 (1)
by the coefficient of y ′′ gives
y ′′+1
2xy ′+
1
4xy = 0 (2)
and by examining the denominators of (2) we see that x = 0 is a regular singular point of the differ-ential equation (1). We can therefore assume that (1) has a Frobenius series solution about x0 = 0
y(x)=∑
m=0
∞
amxm+c (3)
where a0 � 0. There is no singular point of (1) other than x = 0, therefore the distance R from x =0 to the nearest singular point is R = ∞ and so the Frobenius solution obtained will converge for0 < |x|<∞.
Substituting (3) into (1) gives
4x
∑
m=0
∞
(m + c− 1)(m + c)am
xm+c−2 + 2
∑
m=0
∞
(m + c)am
xm+c−1 +
∑
m=0
∞
am
xm+c
≡0
⇒
∑
m=0
∞
4(m + c − 1)(m + c)am
xm+c−1 +
∑
m=0
∞
2(m + c)am
xm+c−1 +
∑
m=0
∞
am
xm+c
≡ 0
and grouping like powers of x gives
∑
m=0
∞
(4(m + c− 1)(m + c)+ 2(m + c))amxm+c−1 +∑
m=0
∞
amxm+c≡ 0
⇒∑
m=0
∞
2(m + c)(2m + 2c− 1)amxm+c−1 +∑
m=0
∞
amxm+c≡ 0 (4)
The indicial equation is obtained from the coefficient of the lowest power of x. In this case, thislowest power of x is the first term of the first series of (4), so we isolate this term
2c(2c− 1)a0xc−1 +
∑
m=1
∞
2(m + c)(2m + 2c − 1)amxm+c−1 +
∑
m=0
∞
amxm+c≡ 0 (5)
and the indicial equation is
2c(2c− 1)= 0;
the indicial roots are
c =0 and c =1
2.
1
Notice that the difference of these indicial roots is
1
2− 0=
1
2
which is NOT an integer, so from Theorem 2 of Assignment 4, we will obtain two Frobenius seriessolutions
y1(x)=∑
m=0
∞
amxm+0 , y2(x) =∑
m=0
∞
amxm+
1
2 (6)
which correspond to the indicial roots c = 0 and c =1
2respectively. The series solutions (6) are
determined by their coefficients am and we shall obtain these coefficients by using recurrence rela-tions. For the series solution corresponding to c = 0 we shall need one recurrence relation and for
the solution corresponding to c =1
2we will need a different recurrence relation.
However, we can avoid some work by deriving both recurrence relations from a single general recur-rence:
let c =0 or c =1
2. For these values of c the first term of (5) vanishes
0 +∑
m=1
∞
2(m + c)(2m + 2c− 1)amxm+c−1 +∑
m=0
∞
amxm+c≡ 0.
Now reindex the second summation so that powers of x match
0 +∑
m=1
∞
2(m + c)(2m + 2c− 1)amxm+c−1 +∑
m=1
∞
am−1xm+c−1≡ 0.
Notice that both series have the same starting index ‘m = 1’ and have matching powers of x, hencewe can group them
∑
m=1
∞(
2(m + c)(2m + 2c− 1)am + am−1)xm+c−1≡ 0
and for this identity to hold, the coefficient of each power of x must vanish and we get
2(m + c)(2m + 2c− 1)am + am−1 = 0 for m =1, 2, 3�and solving for am gives the desired general recurrence relation
am =− am−1
2(m + c)(2m + 2c− 1)for m = 1, 2, 3� (7)
We now substitute c = 0 and c =1
2into (7) to get the different solutions y1(x) and y2(x) of (1).
c = 0
Substituting c =0 into (7) gives the recurrence relation
am =− am−1
2m(2m− 1)for m = 1, 2, 3� (8)
2
Substituting a few values of m into the recurrence (8) gives
a1 =(− 1)a0
(2)(1)
a2 =(− 1)a1
(4)(3)
a3 =(− 1)a2
(6)(5)
a4 =(− 1)a3
(8)(7)
and by ‘back substituting’ we have
a1 =(− 1)a0
2.1
a2 =(− 1)2a0
4.3.2.1
a3 =(− 1)3a0
6.5.4.3.2.1
a4 =(− 1)4a0
8.7.6.5.4.3.2.1The general formula for am is
am =(− 1)ma0
(2m)!for m = 0, 1, 2, 3�
Note that the above formula holds for m = 0. Substituting the coefficients am into the Frobeniussolution for c = 0 we get
y1(x)=∑
m=0
∞
amxm+0
= a0
(
∑
m=0
∞
(− 1)mxm
(2m)!
)
.
c =1
2
Substituting c =1
2into (7) gives the recurrence relation
am =− am−1
2(
m +1
2
)(
2m + 2(
1
2
)
− 1) for m = 1, 2, 3�
which simplifies to
am =− am−1
(2m + 1)(2m)for m = 1, 2, 3� (9)
Substituting a few values of m into the recurrence (9) gives
a1 =(− 1)a0
(3)(2)
a2 =(− 1)a1
(5)(4)
a3 =(− 1)a2
(7)(6)
a4 =(− 1)a3
(9)(8)
3
and by ‘back substituting’ we have
a1 =(− 1)a0
3.2
a2 =(− 1)2a0
5.4.3.2
a3 =(− 1)3a0
7.6.5.4.3.2
a4 =(− 1)4a0
9.8.7.6.5.4.3.2
The general formula for am is
am =(− 1)ma0
(2m + 1)!for m = 0, 1, 2, 3�
Note that this formula holds for m = 0. Substituting the coefficients am into the Frobenius solutionfor c =
1
2we get
y2(x)=∑
m=0
∞
amxm+
1
2
= a0
∑
m=0
∞
(− 1)mxm+
1
2
(2m + 1)!
.
We have found two linearly independent solutions
∑
m=0
∞
(− 1)mxm
(2m)!,∑
m=0
∞
(− 1)mxm+
1
2
(2m +1)!
corresponding to the indicial roots c = 0 and c =1
2respectively. Therefore the general solution of
the given differential equation is
y =A
∑
m=0
∞
(− 1)mxm
(2m)!+ B
∑
m=0
∞
(− 1)mxm+
1
2
(2m + 1)!
which, from above, is valid for 0< |x|<∞.
4
2. Note that 2x(x+ 3)= 0 when x= 0, so x= 0 is a singular point. Dividing
2x(x +3)y ′′− 3(x + 1)y ′+ 2y = 0 (10)
by the coefficient of y ′′ gives
y ′′−3(x + 1)
2x(x +3)y ′+
1
x(x+ 3)y = 0 (11)
and by examining the denominators of (11) we see that x = 0 is a regular singular point of the dif-ferential equation (10). We can therefore assume that that (10) has a Frobenius series solutionabout x0 =0
y(x) =∑
m=0
∞
amxm+c (12)
where a0 � 0. We obtain the singular points of the differential equation (10) by setting the coeffi-cient of y ′′ equal to zero
2x(x+ 3) =0 ⇒ x= 0, x =− 3
so x = − 3 is also a singular point of the differential equation (10). So the distance R from x = 0 tothe nearest singular point is R = 3
and therefore the Frobenius solution (12) will converge for 0 < |x| < 3. Substituting (12) into (10)gives
(2x2 + 6x)
∑
m=0
∞
(m + c− 1)(m + c)am
xm+c−2
− 3(x + 1)∑
m=0
∞
(m + c)am
xm+c−1 +2
∑
m=0
∞
am
xm+c
≡0
⇒
∑
m=0
∞
2(m + c− 1)(m + c)am
xm+c +
∑
m=0
∞
6(m + c− 1)(m + c)am
xm+c−1
−
∑
m=0
∞
3(m + c)am
xm+c
−∑
m=0
∞
3(m + c)am
xm+c−1 +
∑
m=0
∞
2am
xm+c
≡ 0
and grouping like powers of x gives
+∑
m=0
∞
(6(m + c− 1)(m + c)− 3(m + c))am
xm+c−1 +
∑
m=0
∞
(2(m + c − 1)(m + c)− 3(m + c) + 2)am
xm+c
≡ 0
⇒∑
m=0
∞
(6(m + c)2− 9(m + c))amxm+c−1 +∑
m=0
∞
(2(m + c)2− 5(m + c)+ 2)amxm+c≡ 0
⇒∑
m=0
∞
3(m + c)(2(m + c)− 3)amxm+c−1 +∑
m=0
∞
(2(m + c)− 1)((m + c)− 2)amxm+c +≡ 0 (13)
The indicial equation is obtained from the coefficient of the lowest power of x. In this case, thislowest power of x is the first term of the first series of (13), so we isolate this term
3c(2c − 3)a0xc−1 +
∑
m=1
∞
3(m + c)(2(m + c)− 3)amxm+c−1 +
∑
m=0
∞
(2(m + c)− 1)((m + c)− 2)amxm+c≡ 0
5
and so the indicial equation is
3c(2c− 3)= 0
and the indicial roots are
c =0 and c =3
2.
Notice that the difference of these indicial roots is
3
2− 0=
3
2
which is NOT an integer, so from Theorem 2 of Assignment 4, we will obtain two Frobenius seriessolutions
y1(x)=∑
m=0
∞
amxm+0 , y2(x)=∑
m=0
∞
amxm+
3
2 (14)
which correspond to the indicial roots c = 0 and c =3
2respectively. The series solutions (14) are
determined by their coefficients am and we shall obtain these coefficients by using recurrence rela-tions. For the series solution corresponding to c = 0 we shall need one recurrence relation and for
the solution corresponding to c =3
2we will need a different recurrence relation.
As in question 1, we can avoid some work by deriving both recurrence relations from a single gen-eral recurrence:
let c =0 or c =3
2. For these values of c the first term of (14) vanishes
0 +∑
m=1
∞
3(m + c)(2(m + c)− 3)amxm+c−1 +∑
m=0
∞
(2(m + c)− 1)((m + c)− 2)amxm+c≡0
Now reindex the second summation so that powers of x match
∑
m=1
∞
3(m + c)(2(m + c)− 3)amxm+c−1 +∑
m=1
∞
(2((m− 1)+ c))− 1)((m− 1)+ c)− 2)am−1xm+c−1
≡0
⇒∑
m=1
∞
3(m + c)(2(m + c)− 3)amxm+c−1 +∑
m=1
∞
(2(m + c)− 3)((m + c)− 3)am−1xm+c−1
≡0
Notice that both series have the same starting index ‘m = 1’ and have matching powers of x, hencewe can group them
∑
m=1
∞
(3(m + c)(2(m + c)− 3)am + (2(m + c)− 3)((m + c)− 3)am−1)xm+c−1≡ 0
and for this identity to hold, the coefficient of each power of x must vanish and we get
3(m + c)(2(m + c)− 3)am + (2(m + c)− 3)((m + c)− 3)am−1 = 0 for m = 1, 2, 3�6
and solving for am gives the desired general recurrence relation
am =− (2(m + c)− 3)((m + c)− 3)am−1
3(m + c)(2(m + c)− 3)for m = 1, 2, 3� (15)
We now substitute c = 0 and c =3
2into (15) to get the different solutions y1(x) and y2(x) of (10).
c = 0
Substituting c =0 into (15) gives the recurrence relation
am =− (2m− 3)(m− 3)am−1
3m(2m− 3)for m = 1, 2, 3� (16)
and as 2m− 3� 0 when m = 1, 2,� we may simplify (16) by cancelling the (2m− 3) factor
am =− (m− 3)am−1
3mfor m =1, 2, 3�
and substituting values m = 1, 2, 3, 4 into (16) gives
a1 =− (− 2)a0
3(1)=
2a0
3
a2 =− (− 1)a1
3(2)=
1
6
(
2a0
3
)
=a0
9
a3 =− (0)a2
3(3)= 0
a4 =− (1)a3
3(4)=
− (1)(0)
3(4)= 0
(17)
and notice that am = 0 for all m > 3. Substituting the equations of (17) into the Frobenius seriesy1(x) of (14) gives
y1(x)=∑
m=0
∞
amxm
= a0 + a1x1 + a2x
2 + a3x3 +�
= a0 +2a0
3x1 +
a0
9x2 + 0+ 0 +�
and we get that the Frobenius series solution y1(x)
y1(x)= a0
(
1 +2x
3+
x2
9
)
.
c =3
2
Substituting c =1
2into (15) gives the recurrence relation
am =−(
2(
m +3
2
)
− 3)((
m +3
2
)
− 3)
am−1
3(
m +3
2
)(
2(
m +3
2
)
− 3) for m = 1, 2, 3�
7
which simplifies to
am =− (2m− 3)am−1
3(2m +3)for m = 1, 2, 3� (18)
Substituting a few values of m into the recurrence (18) gives
a1 =(− 1)(− 1)a0
(3)(5)
a2 =(− 1)(1)a1
(3)(7)
a3 =(− 1)(3)a2
(3)(9)
and by ‘back substituting’ we have
a1 =(− 1)(− 1)a0
(3)(5)
a2 =(− 1)2(− 1.1)a0
(3)2(5.7)
a3 =(− 1)3(− 1.1.3)a0
(3)3(5.7.9)
The general formula for am is
am =(− 1)m(− 1.1� (2m− 3))a0
3m(5.7� (2m + 3))for m =1, 2, 3�
which can be simplified as follows
am =(− 1)m+1(3.5.7� (2m− 3))a0
3m−1(3.5.7� (2m− 3)(2m− 1)(2m + 1)(2m + 3))for m = 1, 2, 3�
⇒ am =(− 1)m+1a0
3m−1(2m− 1)(2m + 1)(2m +3))for m = 1, 2, 3�
Notice that this formula for am does not hold for m = 0. Substituting the coefficients am into the
Frobenius solution for c =3
2we get
y2(x)=∑
m=0
∞
amxm+
3
2
= a0x3
2 +∑
m=1
∞
amxm+
3
2
= a0
x3
2 +∑
m=1
∞
(− 1)m+1xm+
3
2
3m−1(2m− 1)(2m +1)(2m + 3))
We have found two linearly independent solutions
1 +2x
3+
x2
9, x
3
2 +∑
m=1
∞
(− 1)m+1xm+
3
2
3m−1(2m− 1)(2m +1)(2m +3))
8
corresponding to the indicial roots c = 0 and c =3
2respectively. Therefore the general solution of
the given differential equation is
y =A
(
1+2x
3+
x2
9
)
+ B
x3
2 +∑
m=1
∞
(− 1)m+1xm+
3
2
3m−1(2m− 1)(2m + 1)(2m +3))
which, from above, is valid for 0< |x|< 3.
9
3. Note that x(x− 1)= 0 when x= 0, so x= 0 is a singular point. Dividing
x(x− 1)y ′′+(3x− 1)y ′+ y = 0 (19)
by the coefficient of y ′′ gives
y ′′ +(3x− 1)
x(x− 1)y ′+
1
x(x− 1)y = 0 (20)
and by examining the denominators of (20) we see that x = 0 is a regular singular point of the dif-ferential equation (19). We can therefore assume that that (19) has a Frobenius series solutionabout x0 =0
y(x) =∑
m=0
∞
amxm+c (21)
where a0 � 0. We obtain the singular points of the differential equation (19) by setting the coeffi-cient of y ′′ equal to zero
x(x− 1)= 0 ⇒ x =0, x= 1
so x = 1 is also a singular point of the differential equation (19). So the distance R from x = 0 tothe nearest singular point is R = 1
x-axis0 1−1
R = 1
and therefore the Frobenius solution (21) will converge for 0< |x|< 1.
Substituting (21) into (19) gives
(x2− x)
∑
m=0
∞
(m + c − 1)(m + c)am
xm+c−2 + (3x− 1)
∑
m=0
∞
(m + c)am
xm+c−1 +
∑
m=0
∞
am
xm+c
≡0
⇒
∑
m=0
∞
(m + c− 1)(m + c)am
xm+c
−
∑
m=0
∞
(m + c− 1)(m + c)am
xm+c−1 +
∑
m=0
∞
3(m + c)am
xm+c
−∑
m=0
∞
(m + c)am
xm+c−1 +
∑
m=0
∞
am
xm+c
≡ 0
and grouping like powers of x gives
−
∑
m=0
∞
((m + c− 1)(m + c) + (m + c))am
xm+c−1 +
∑
m=0
∞
((m + c − 1)(m + c)+ 3(m + c) + 1)am
xm+c
≡ 0
⇒ −∑
m=0
∞
(m + c)2amxm+c−1 +∑
m=0
∞
((m + c)2 + 2(m + c)+ 1)amxm+c≡ 0
⇒ −∑
m=0
∞
(m + c)2amxm+c−1 +∑
m=0
∞
(m + c + 1)2amxm+c≡ 0 (22)
The indicial equation is obtained from the coefficient of the lowest power of x. In this case, thislowest power of x is the first term of the first series of (22), so we isolate this term
− c2a0xc−1−
∑
m=1
∞
(m + c)2amxm+c−1 +∑
m=0
∞
(m + c + 1)2amxm+c≡ 0 (23)
10
and so the indicial equation is
c2 = 0
and we only have one (repeated) indicial root
c =0.
From Theorem 2 of Assignment 4
“If c1 = c2 then there is only one Frobenius series solution about the point x = x0. The other solu-tion may be obtained by using the reduction of order method.” Therefore in this case we will obtainonly one Frobenius solution that corresponds to c = 0 and this solution will take the form
y(x)=∑
m=0
∞
amxm+0. (24)
We substitute c =0 into (23) to determine the coefficients am of (24)
0−∑
m=1
∞
m2amxm−1 +∑
m=0
∞
(m +1)2amxm≡ 0.
Reindex the second summation so that powers of x match
−∑
m=1
∞
m2amxm−1 +∑
m=1
∞
((m− 1)+ 1)2am−1xm−1≡ 0
⇒−∑
m=1
∞
m2amxm−1 +∑
m=1
∞
m2am−1xm−1≡ 0.
Notice that both series have the same starting index ‘m = 1’ and have matching powers of x, hencewe can group them
∑
m=1
∞(
−m2am + m2am−1
)
xm+c−1≡ 0
and for this identity to hold, the coefficient of each power of x must vanish and we get
−m2am + m2am−1 = 0 when m =1, 2, 3,� (25)
and as m2� 0 when m =1, 2, 3,� we can divide (25) by m2 to get
− am + am−1 = 0 when m =1, 2, 3,�⇒ am = am−1 when m = 1, 2, 3,� (26)
Substituting a few values of m into the recurrence (26) gives
a1 = a0
a2 = a1
a3 = a2
11
and by ‘back substituting’ we have
a1 = a0
a2 = a0
a3 = a0.
The general formula for am is
am = a0 for m =0, 1, 2, 3�Notice this general formula holds when m = 0. Substituting this general formula for am into theFrobenius solution (24) gives
y(x) =∑
m=0
∞
amxm
=∑
m=0
∞
a0xm
= a0
(
∑
m=0
∞
xm
)
(27)
which, from above, is valid for 0< |x|< 1. Recall that
∑
m=0
∞
xm = 1 +x + x2 +�is a geometric series with first term a = 1 and common ratio r = x. From the formula for the sum ofa geometric series we have
∑
m=0
∞
xm =1
1− x
and so we may write the Frobenius solution (27) as
y = a0
(
1
1− x
)
.
We therefore have that
y1(x)=1
1− x=(1− x)−1 (28)
is a solution of the differential equation (19). We obtain a second linearly independent solution bythe reduction of order method, that is, we assume our second solution y2(x) of the differentialequation (19) takes the form
y2(x)= (1−x)−1v(x) (29)
where we need to determine v(x). From (29) we get
y2′ = ((1− x)−1)′v + (1− x)−1v ′
y2′′= ((1− x)−1)′′v + 2((1− x)−1)′v ′+ (1−x)−1v ′′
12
and as y2 is assumed, by the reduction of order method, to be a solution of the differential equation(19) we substitute y2 into (19) to determine v(x) –
x(x− 1)y ′′+ (3x− 1)y ′+ y = 0
⇒x(x− 1)(
((1− x)−1)′′v + 2((1− x)−1)′v ′+ (1− x)−1v′′)
+ (3x− 1)(
((1− x)−1)′v + (1− x)−1v′)
+ (1−x)−1v = 0
(30)
and rearranging (30) gives
x(x− 1)((1− x)−1)′′v + (3x− 1)((1− x)−1)′v +(1− x)−1v
+ 2x(x− 1)((1− x)−1)′v ′+ (3x − 1)(1− x)−1v′+ x(x − 1)(1− x)−1
v′′= 0
⇒(x(x− 1)((1− x)−1)′′+ (3x− 1)((1− x)−1)′+ (1− x)−1)v
+ 2x(x − 1)((1− x)−1)′v ′+ (3x− 1)(1− x)−1v′+ x(x− 1)(1− x)−1
v′′= 0.
(31)
We know that (1− x)−1 is a solution of the differential equation (19), that is:
x(x− 1)((1−x)−1)′′+ (3x− 1)((1− x)−1)′+ (1− x)−1 = 0
⇒(
x(x− 1)((1− x)−1)′′+(3x− 1)((1− x)−1)′+ (1−x)−1)v = 0 (32)
and substituting (32) into (31) gives1
2x(x− 1)((1− x)−1)′v ′+ (3x− 1)(1− x)−1v ′+ x(x− 1)(1− x)−1v ′′= 0
⇒ 2x(x− 1)((1−x)−2)v ′+ (3x− 1)(1− x)−1v ′− xv ′′= 0
⇒ − 2x(1− x)−1v ′ +(3x− 1)(1− x)−1v ′− xv ′′=0
⇒ (x− 1)(1−x)−1v ′−xv ′′= 0
⇒ − v ′− xv ′′= 0
⇒ xv ′′+ v ′=0 (33)
Now use the substitution w = v ′ to convert the second order differential equation (33) to a firstorder differential equation
xw ′+ w = 0 (34)
1. Substituting equation (32) into the differential equation (31) removes the terms in v. (31) then takes the form
( )v ′+ ( )v ′′= 0.
We then reduce the order of this second order differential equation by using the substitution w = v′ to obtain the first order
differential equation
( )w + ( )w′= 0
which, in this case, can be solved by seperation of variables.
13
Write (34) as
xdw
dx+ w =0
and seperating variables gives∫
dw
w=−
∫
dx
x(35)
By integrating (35) we have
lnw =− lnAx (36)
where A is a constant. From (36) we obtain
w =B
x
and as we only seek ONE second solution we may assume B = 1, that is
w =1
x. (37)
Recall that w = v ′, so by integrating (37) we have that
v(x) = ln x
(again we only seek ONE second solution so we may assume the constant of integration is equal tozero in this case). Therefore, from (29)
y2 = (1− x)−1lnx
is a second solution to the differential equation (19). Since
y1 =(1− x)−1 and y2 = (1− x)−1lnx
are linearly independent, it follows that the general solution to the differential equation (29) is
y = A(1− x)−1 + B (1− x)−1lnx
which clearly may be written as
y = A
(
1
1−x
)
+ B
(
lnx
1−x
)
.
From above, this solution is valid for 0< |x|< 1.
14
4. Note that 4x= 0 when x =0, so x =0 is a singular point. Dividing
xy ′′− (4+ x)y ′+ 2y = 0 (38)
by the coefficient of y ′′ gives
y ′′−(4+ x)
xy ′ +
2
xy = 0 (39)
and by examining the denominators of (39) we see that x = 0 is a regular singular point of the dif-ferential equation (38). We can therefore assume that (38) has a Frobenius series solution aboutx0 = 0
y(x) =∑
m=0
∞
amxm+c (40)
where a0� 0. There is no singular point of (38) other than x = 0, therefore the distance R from x =0 to the nearest singular point is R = ∞ and so the Frobenius solution obtained will converge for0 < |x|<∞.
Substituting (40) into (38) gives
x
∑
m=0
∞
(m + c− 1)(m + c)am
xm+c−2
− (4+ x)∑
m=0
∞
(m + c)am
xm+c−1 + 2
∑
m=0
∞
am
xm+c
≡0
⇒
∑
m=0
∞
(m + c − 1)(m + c)am
xm+c−1
−
∑
m=0
∞
4(m + c)am
xm+c−1
−
∑
m=0
∞
(m + c)am
xm+c +
∑
m=0
∞
2am
xm+c
≡ 0
and grouping like powers of x gives
∑
m=0
∞
((m + c− 1)(m + c)− 4(m + c))amxm+c−1−
∑
m=0
∞
((m + c)− 2)amxm+c≡ 0
⇒∑
m=0
∞
(m + c)(m + c− 5)amxm+c−1−
∑
m=0
∞
((m + c)− 2)amxm+c≡ 0 (41)
The indicial equation is obtained from the coefficient of the lowest power of x. In this case, thislowest power of x is the first term of the first series of (41), so we isolate this term
c(c− 5)a0xc−1 +
∑
m=1
∞
(m + c)(m + c− 5)amxm+c−1
−
∑
m=0
∞
((m + c)− 2)amxm+c≡ 0 (42)
and the indicial equation is
c(c− 5)= 0;
and the indicial roots are
c = 5 and c = 0.
15
Notice that the difference of these indicial roots is
5− (0) =5
which IS an integer, so from Theorem 2 of Assignment 4
“If c1− c2 is an integer then one of the following alternatives occur
a) A Frobenius series solution about the point x = x0 can be obtained from the larger indicialroot, but a second linearly independent Frobenius series solution cannot be obtained fromthe smaller root.
b) A Frobenius series solution about the point x = x0 can be obtained from the larger indicialroot and two linearly independent Frobenius series solutions about the point x = x0 can beobtained from the smaller root”
As in questions 1, 2 & 3 we obtain the general recurrence relation. It is perhaps best to first substi-tute the smaller root into this general recurrence in case the smaller root does yield both Frobeniussolutions.
Let c =5 or c = 0. For these values of c the first term of (42) vanishes
0+∑
m=1
∞
(m + c)(m + c− 5)amxm+c−1−
∑
m=0
∞
(m + c− 2)amxm+c≡ 0
Now reindex the second summation so that powers of x match
∑
m=1
∞
(m + c)(m + c− 5)amxm+c−1−
∑
m=1
∞
(m + c− 3)am−1xm+c−1≡ 0.
Notice that both series have the same starting index ‘m = 1’ and have matching powers of x, hencewe can group them
∑
m=1
∞
((m + c)(m + c− 5)am − (m + c− 3)am−1)xm+c−1≡ 0
and for this identity to hold, the coefficient of each power of x must vanish and we get
(m + c)(m + c− 5)am − (m + c− 3)am−1 =0 for m = 1, 2, 3� (43)
However, at this point, we do not solve for am immediately as this will eventually lead to divisionby zero. We substitute the indicial roots, starting with the smaller root c = 0 in the hope that thissmaller root will yield both solutions.
c = 0
Substituting c =0 into (43) gives
m(m− 5)am − (m− 3)am−1 = 0 for m = 1, 2, 3� (44)
16
In the recurrence relation (44), notice that the coefficients m(m − 5) and (m − 3) of am and am−1
vanish at m =5 and m = 3 respectively. Substituting values m = 1, 2, 3, 4, 5 into (44) gives
m =1: − 4a1 +2a0 = 0m =2: − 6a2 + a1 = 0m =3: − 6a3 + 0= 0m =4: − 4a4− a3 = 0m =5: 0a5− 2a4 =0
(45)
From the first four equations of (45) we have
a1 =1
2a0
a2 =1
12a0
a3 = 0a4 = 0;
(46)
the fifth equation of (45) is
0a5− 2a4 =0 ⇒ 0= 0
as a4 = 0. This fifth equation imposes no condition on a5 and implies that a5 can take any value,that is, a5 is an arbitrary constant. We now show that am for m ≥ 6 can be expressed in terms of
a5. Note that the coefficient m(m − 5) of am in (44) is nonzero for m ≥ 6; it follows that solving(44) for am
am =(m− 3)am−1
m(m− 5)m = 6, 7,� (47)
does not lead to division by zero when m = 6, 7,� . Substituting a few values of m into the recur-rence (47) gives
m = 6: a6 =(3)a5
(6)(1)
m = 7: a7 =(4)a6
(7)(2)
m = 8: a8 =(5)a7
(8)(3)
and by ‘back substituting’ we have
a6 =(3)a5
(6)(1)
a7 =(3.4)a5
(6.7)(1.2)
a8 =(3.4.5)a5
(6.7.8)(1.2.3)
and so a general formula for am when m≥ 6 is
am =(3.4� (m− 3))a5
(6.7�m)(m− 5)!m = 6, 7,� (48)
17
and note this formula does not hold when m = 5. The equations (46) and (48) express the coeffi-cients am of the Frobenius solution
y(x) =∑
m=0
∞
amxm+0 (49)
corresponding to c = 0 in terms of a0 and a5 where both a5 and a0 are arbitrary constants. Substi-tuting (46) and (48) into (49) gives
y = a0 + a1x + a2x2 + a3x
3 + a4x4 + a5x
5 +∑
m=6
∞
amxm+0
⇒ y = a0 +1
2a0x +
1
12a0x
2 + 0x3 +0x4 + a5x5 +
∑
m=0
∞
(3.4� (m− 3))a5xm+c
(6.7�m)(m− 5)!
⇒ y = a0
(
1 +1
2x+
1
12x2
)
+ a5
(
x5 +∑
m=6
∞
(3.4� (m− 3))xm
(6.7�m)(m− 5)!
)
(50)
As a5 and a0 are arbitrary constants, (50) is the general solution of the given differential equation;from above this solution is valid for 0< |x|<∞.
18
5.
a) Note that x2 = 0 when x= 0, so x= 0 is a singular point. Dividing
x2y ′′+ xy ′+ (x2− ν2)y = 0 (51)
by the coefficient of y ′′ gives
y ′′+1
xy ′+
(x2− ν2)
x2y = 0 (52)
and by examining the denominators of (52) we see that x = 0 is a regular singular point thedifferential equation (51). We can therefore assume that that (51) has a Frobenius seriessolution about x0 =0
y(x) =∑
m=0
∞
amxm+c (53)
where a0� 0.
Substituting (53) into (51) gives
x2∑
m=0
∞
(m + c− 1)(m + c)am
xm+c−2 + x
∑
m=0
∞
(m + c)am
xm+c−1 + (x2
− ν2)∑
m=0
∞
am
xm+c
≡0
∑
m=0
∞
(m + c− 1)(m + c)am
xm+c +
∑
m=0
∞
(m + c)am
xm+c +
∑
m=0
∞
am
xm+c+2 −
∑
m=0
∞
ν2a
mx
m+c≡ 0
and grouping like powers of x gives
∑
m=0
∞(
(m + c− 1)(m + c)+ (m + c)− ν2)
am
xm+c +
∑
m=0
∞
amxm+c+2≡ 0
⇒∑
m=0
∞(
(m + c)2− ν2)
am
xm+c +
∑
m=0
∞
amxm+c+2≡ 0 (54)
The indicial equation is obtained from the coefficient of the lowest power of x. In this case,this lowest power of x is the first term of the first series of (54), so we isolate this term
(c2− ν2)a0x
c +∑
m=1
∞(
(m + c)2− ν2)
am
xm+c +
∑
m=0
∞
amxm+c+2≡ 0 (55)
and so the indicial equation is
(c2− ν2)= 0
and the indicial roots are
c = ν and c =− ν.
b) We desire only one Frobenius solution – the solution corresponding to c = ν
y(x)=∑
m=0
∞
amxm+ν. (56)
19
The coefficients am of (56) are obtained by substituting c = ν into (55) and obtaining arecurrence relation.
(ν2− ν
2)a0xc +
∑
m=1
∞(
(m + ν)2− ν2)
am
xm+ν +
∑
m=0
∞
amxm+ν+2≡ 0
⇒ 0 +∑
m=1
∞
m(m + 2ν)amxm+ν +∑
m=0
∞
amxm+ν+2≡ 0 (57)
and now re-index the second series of (53) so that the powers match
∑
m=1
∞
m(m +2ν)amxm+ν +∑
m=2
∞
am−2xm+ν ≡ 0 (58)
Note that we cannot immediately group the two series of (54) as the starting indices of theseries do not match. We therefore rewrite the first series of (54)
(1+ 2ν)a1x1+ν +
∑
m=2
∞
m(m +2ν)amxm+ν +∑
m=2
∞
am−2xm+ν ≡ 0
and now we may group the series to obtain
(1 + 2ν)a1x1+ν +
∑
m=2
∞
(m(m +2ν)am + am−2)xm+ν ≡ 0
and for this identity to hold, the coefficient of each power of x must vanish and we get
(1 + 2ν)a1 = 0 (59)
m(m +2ν)am + am−2 = 0 when m =2, 3,� (60)
and, since by assumption ν > 0 we have from (59) that
a1 = 0. (61)
Notice that (60) is a recurrence relation that ‘steps’ by 2. This type of recurrence seperatesinto ‘even’ a2k and ‘odd’ a2k+1, but from (61) it follows that all odd a2k+1 will be zero.Hence we consider only even a2k. First rewrite (60) as
am =− am−2
m(m + 2ν)(62)
and substituting m = 2, 4, 6 into (62) we have
a2 =− a0
2(2+ ν)=
− a0
22(1)(1 + ν)
a4 =− a2
4(4+ ν)=
− a2
22(2)(2 + ν)
a6 =− a4
6(6+ ν)=
− a4
22(3)(3 + ν)
20
and ‘back-substituting’ gives
a2 =− a0
22(1)(1 + ν)
a4 =a0
24(1.2)(1 + ν)(2+ ν)
a6 =− a0
26(1.2.3)(1+ ν)(2 + ν)(3+ ν)
and the general formula for even a is given by
a2k =(− 1)ka0
22kk!(ν +1)(ν + 2)� (ν + k)when k =1, 2, 3,� (63)
and as mentioned above
a2k+1 =0 when k =0, 1, 2, 3,� (64)
Substituting (63) and (64) into the Frobenius solution corresponding to c = ν gives
y(x)=∑
m=0
∞
amxm+ν
=∑
k=0
∞
a2kx2k+ν +
∑
k=0
∞
a2k+1x2k+1+ν
=∑
k=0
∞
a2kx2k+ν + 0
= a0
(
xν +∑
k=1
∞
(− 1)kx2k+ν
22kk!(ν +1)(ν + 2)� (ν + k)
)
.
c) When ν = n where n = 0, 1, 2,� it follows by replacing ν =n in part b) that
y(x) = a0
(
xn +∑
k=1
∞
(− 1)kx2k+n
22kk!(n + 1)(n + 2)� (n + k)
)
(65)
are solutions of Bessel’s equation of index n:
x2y ′′+ xy ′+ (x2−n2)y = 0. (66)
Letting a0 =1
2nn!
in (65) gives a particular solution of (66) which is
y(x)=1
2nn!
(
xn +∑
k=1
∞
(− 1)kx2k+n
22kk!(n + 1)(n + 2)� (n + k)
)
=xn
2nn!+
1
2nn!
∑
k=1
∞
(− 1)kx2k+n
22kk!(n +1)(n +2)� (n + k)
=xn
2nn!+∑
k=1
∞
(− 1)kx2k+n
22k2nk!n!(n +1)(n +2)� (n + k)
21
Now
n!(n +1)(n +2)� (n + k) = (n + k)!
so we have
y(x)=xn
2nn!+∑
k=1
∞
(− 1)kx2k+n
22k+nk!(n + k)!(67)
Notice that
(− 1)kx2(0)+n
22(0)+n0!(n +0)!=
xn
2nn!
that is, the general formula in the series in (67) holds when k = 0. Therefore we can rewrite(67) as
y(x)=∑
k=0
∞
(− 1)kx2k+n
22k+nk!(n + k)!(68)
and so, as required, (68) is a particular solution of Bessel’s equation of index n.
22