Chapter 52

Power series methods

of solving ordinary
differential equations
as the differential coefficient of common functions
52.1 Introduction rises.
dy d2 y
Second order ordinary differential equations that can- (i) If y = eax , then = aeax , 2
= a 2 eax , and so
not be solved by analytical methods (as shown in on. dx dx
Chapters 50 and 51), i.e. those involving variable coeffi-
dy d2 y
cients, can often be solved in the form of an infinite series If we abbreviate as y
, as y

, … and
of powers of the variable. This chapter looks at some of dx dx 2
dn y
the methods that make this possible—by the Leibniz– as y (n) , then y
= aeax , y

= a 2eax , and the

dx n
Maclaurin and Frobinius methods, involving Bessel’s
and Legendre’s equations, Bessel and gamma func- emerging pattern gives: y(n) = an eax (1)
tions and Legendre’s polynomials. Before introducing
Leibniz’s theorem, some trends with higher differential For example, if y = 3e2x , then
coefficients are considered. To better understand this d7 y
chapter it is necessary to be able to: = y (7) = 3(27 ) e2x = 384e2x
dx 7
(i) differentiate standard functions (as explained in
Chapters 27 and 32), (ii) If y = sin ax,
 π
(ii) appreciate the binomial theorem (as explained in y
= a cos ax = a sin ax +
Chapters 7), and 2
(iii) use Maclaurins theorem (as explained in Chapter 8). y

= −a 2 sin ax = a 2 sin(ax + π)
 
2π
= a sin ax +
2
2
52.2 Higher order differential
coefficients as series y

= −a 3 cos x
 
The following is an extension of successive differentia- 3π
= a 3 sin ax + and so on.
tion (see page 296), but looking for trends, or series, 2
494 Higher Engineering Mathematics

  Since sinh ax is not periodic (see graph on page

In general, y(n) = an sin ax + nπ (2)
2 43), it is more difficult to find a general state-
ment for y (n) . However, this is achieved with the
For example, if following general series:

d5 y an
y = sin 3x, then 5 = y (5) y(n) = {[1 +(−1)n ] sinh ax
dx 2
  
5 5π π + [1 −(−1)n ] cosh ax} (5)
= 3 sin 3x + = 35 sin 3x +
2 2
For example, if
= 243 cos 3x
d5 y
(iii) If y = cos ax, y = sinh 2x, then = y (5)
dx 5
 π 25
y
= −a sin ax = a cos ax + = {[1 + (−1)5 ] sinh 2x
2 2
 
2π + [1 − (−1)5 ] cosh 2x}
y

= −a 2 cos ax = a 2 cos ax +
2 25
  = {[0] sinh 2x + [2] cosh 2x}

3π 2
y = a sin ax = a cos ax +
3 3
and so on.
2 = 32 cosh 2x

 nπ  (vi) If y = cosh ax,

In general, y(n) = an cos ax + (3)
2
y
= a sinh ax
For example, if y = 4 cos 2x, y

= a 2 cosh ax
 
d6 y 6π
(6)
then 6 = y = 4(2 ) cos 2x +
6 y

= a 3 sinh ax, and so on

dx 2
= 4(2 ) cos (2x + 3π)
6 Since cosh ax is not periodic (see graph on page
43), again it is more difficult to find a general
= 4(26 ) cos (2x + π)
statement for y (n) . However, this is achieved with
= −256 cos 2x the following general series:
(iv) If y = x a, y
= a x a−1 , y

= a(a − 1)x a−2 , an

y

= a(a − 1)(a − 2)x a−3 , y(n) = {[1 − (−1)n ] sinh ax

2
and y(n) = a(a − 1)(a − 2) . . . . . (a − n + 1) x a−n
+ [1 + (−1)n ] cosh ax} (6)
a!
or y(n) = xa−n (4) 1
(a − n)! For example, if y = cosh 3x,
9  7
where a is a positive integer. d7 y (7) 1 3
then 7 = y = (2 sinh 3x)
d4 y dx 9 2
For example, if y = 2x6 , then 4 = y (4)
dx = 243 sinh 3x
6!
= (2) x 6−4 1 1 2
(6 − 4)! (vii) If y = ln ax, y
= , y

= − 2 , y

= 3 , and so
6 × 5 × 4× 3 × 2× 1 2 x x x
on.
= (2) x
2×1 (n − 1)!
= 720x2 In general, y(n) = (−1)n−1 (7)
xn
(v) If y = sinh ax, y
= a cosh ax For example, if y = ln 5x, then

y = a sinh ax
2  
d6 y (6) = (−1)6−1 5! = − 120
= y
y

= a 3 cosh ax, and so on dx 6 x6 x6

 Power series methods of solving ordinary differential equations 495

1
Note that if y = ln x, y
= ; if in equation (7),
x 52.3 Leibniz’s theorem

0 (0)!
n = 1 then y = (−1) 1
x If y = uv (8)
1
(−1)0 = 1 and if y
= then (0)!= 1 (Check that where u and v are each functions of x, then by using the
x product rule,
(−1)0 = 1 and (0)! = 1 on a calculator).
y
= uv
+ vu
(9)
y

= uv

+ v
u
+ vu

+ u
v

Now try the following exercise

= u

v + 2u
v
+ uv

(10)

Exercise 193 Further problems on higher y = u v + vu + 2u v + 2v u + uv + v u

order differential coefficients as series = u

v + 3u

v
+ 3u
v

+ uv

(11)
Determine the following derivatives: y (4) = u (4)v + 4u (3)v (1) + 6u (2)v (2)
t
1. (a) y (4) when y = e2x (b) y (5) when y = 8e2 + 4u (1)v (3) + uv (4) (12)
1 t
[(a) 16 e2x (b) e 2 ] From equations (8) to (12) it is seen that
4
(a) the n’th derivative of u decreases by 1 moving
2. (a) y (4) when y = sin 3t from left to right,
1
(b) y (7) when y = sin 5θ
50 (b) the n’th derivative of v increases by 1 moving from
left to right,
[(a) 81 sin3t (b) −1562.5 cos5θ]
(c) the coefficients 1, 4, 6, 4, 1 are the normal binomial
3. (a) y (8) when y = cos 2x coefficients (see page 58).
2
(b) y (9) when y = 3 cos t
3 In fact, (uv)(n) may be obtained by expanding (u + v)(n)
 using the binomial theorem (see page 59), where the
29 2
(a) 256 cos2x (b) − 8 sin t ‘powers’ are interpreted as derivatives. Thus, expanding
3 3
(u + v)(n) gives:

t7 y(n) = (uv)(n) = u(n) v + nu(n−1) v (1)

4. (a) y (7) when y = 2x 9 (b) y (6) when y =
8 n(n− 1) (n−2) (2)
+ u v
[(a) (9! )x 2 (b) 630 t ] 2!
1 n(n− 1)(n −2) (n−3) (3)
5. (a) y (7) when y = sinh 2x + u v +··· (13)
4 3!
(b) y (6) when y = 2 sinh 3x
Equation (13) is a statement of Leibniz’s theorem,
[(a) 32 cosh 2x (b) 1458 sinh 3x] which can be used to differentiate a product n times.
The theorem is demonstrated in the following worked
6. (a) y (7) when y = cosh 2x problems.
1
(b) y (8) when y = cosh 3x
9
Problem 1. Determine y (n) when y = x 2 e3x .
[(a) 128 sinh 2x (b) 729 cosh 3x]

7. (a) y (4) when y = 2ln 3θ For a product y = uv, the function taken as
1
(b) y (7) when y = ln 2t (i) u is the one whose nth derivative can readily be
3  determined (from equations (1) to (7)),
6 240
(a) − (b) 7
θ4 t (ii) v is the one whose derivative reduces to zero after
a few stages of differentiation.
496 Higher Engineering Mathematics

Thus, when y = x 2 e3x , v = x 2 , since its third derivative By Leibniz’s equation, equation (13),
is zero, and u = e3x since the nth derivative is known  
from equation (1), i.e. 3n eax n(n − 1) (n)
y (n+2)(1 + x 2 ) + n y (n+1)(2x)+ y (2)+ 0
Using Leinbiz’s theorem (equation (13), 2!
+ 2{y (n+1) (x) + n y (n) (1) + 0} − 3{y (n) } = 0
(n) (n) (n−1) (1) n(n − 1) (n−2) (2)
y =u v + nu v + u v
2!
i.e. (1 + x 2 )y (n+2) + 2n x y (n+1) + n(n − 1)y (n)
n(n − 1)(n − 2) (n−3) (3)
+ u v + ···
3! + 2x y (n+1) + 2 ny (n) − 3y (n) = 0
or (1 + x 2 )y (n+2) + 2(n + 1)x y (n+1)
where in this case v = x 2 , v (1) = 2x, v (2) = 2 and
v (3) = 0 + (n 2 − n + 2n − 3)y (n) = 0
Hence, y (n) = (3n e3x )(x 2 ) + n(3n−1 e3x )(2x) i.e. (1 + x2 )y(n+2) + 2(n + 1)xy(n+1)
n(n − 1) n−2 3x
+ (3 e )(2) + (n2 + n − 3)y(n) = 0
2!
n(n − 1)(n − 2) n−3 3x
+ (3 e )(0) Problem 4. Find the 5th derivative of y = x 4 sin x.
3!
= 3n−2 e3x (32 x 2 + n(3)(2x)
If y = x 4 sin x, then using Leibniz’s equation with
+ n(n − 1) + 0) u = sin x and v = x 4 gives:
y(n) = e3x 3n−2 (9x2 + 6nx + n(n− 1))   nπ  4 
i.e.
y (n) = sin x + x
2
Problem 2. If x 2 y

+ 2x y
+ y = 0 show that:   
x y (n+2) + 2(n + 1)x y (n+1) + (n 2 + n + 1)y (n) = 0 (n − 1)π 3
+ n sin x + 4x
2
Differentiating each term of x 2 y

+ 2x y
+ y = 0   
n(n − 1) (n − 2)π
n times, using Leibniz’s theorem of equation (13), + sin x + 12x 2
2! 2
gives:
  
n(n − 1)(n − 2) (n − 3)π
  + sin x + 24x
n(n − 1) (n) 3! 2
y (n+2) x 2 + n y (n+1) (2x) + y (2) + 0
2!  
n(n − 1)(n − 2)(n − 3)
+ sin x
+ {y (n+1) (2x) + n y (n) (2) + 0} + {y (n) } = 0 4!

(n − 4)π
i.e. x 2 y (n+2) + 2n x y (n+1) + n(n − 1)y (n) + 24
  2
5π
+ 2x y (n+1) + 2n y (n) + y (n) = 0 and y (5) = x 4 sin x + + 20x 3 sin(x + 2π)
2
i.e. x 2 y (n+2) + 2(n + 1)x y (n+1)  
(5)(4) 3π
+ (12x ) sin x +
2
+ (n 2 − n + 2n + 1)y (n) = 0 2 2
or x2 y(n+2) + 2(n + 1) x y(n+1) (5)(4)(3)
+ (24x) sin (x + π)
+ (n + n + 1)y
2 (n)
=0 (3)(2)
(5)(4)(3)(2)  π
+ (24) sin x +
(4)(3)(2) 2
Problem 3. Differentiate the following
  
differential equation n times: 5π π
(1 + x 2 )y

+ 2x y
− 3y = 0. Since sin x + ≡ sin x + ≡ cos x,
2 2
 Power series methods of solving ordinary differential equations 497
 
3π
sin(x + 2π) ≡ sin x, sin x + ≡ −cos x, 52.4 Power series solution by the
2
Leibniz–Maclaurin method
and sin (x + π) ≡ −sin x,
then y (5) = x 4 cos x + 20x 3 sin x + 120x 2 (−cos x) For second order differential equations that cannot be
+ 240x(−sin x) + 120 cos x solved by algebraic methods, the Leibniz–Maclaurin
method produces a solution in the form of infinite
i.e. y(5) = (x4 − 120x2 + 120)cos x series of powers of the unknown variable. The fol-
+ (20x3 − 240x) sin x lowing simple 5-step procedure may be used in the
Leibniz–Maclaurin method:
(i) Differentiate the given equation n times, using
Now try the following exercise the Leibniz theorem of equation (13),
(ii) rearrange the result to obtain the recurrence
Exercise 194 Further problems on relation at x = 0,
Leibniz’s theorem
(iii) determine the values of the derivatives at x = 0,
Use the theorem of Leibniz in the following i.e. find ( y)0 and ( y
)0 ,
problems:
(iv) substitute in the Maclaurin expansion for
1. Obtain the n’th derivative of: x 2 y. y = f (x) (see page 69, equation (5)),
2 (n) 
x y + 2n x y (n−1) + n(n − 1)y (n−2)
(v) simplify the result where possible and apply
boundary condition (if given).
2. If y = x 3 e2x find y (n) and hence y (3) .
⎡ ⎤
y (n) = e2x 2n−3 {8x 3 + 12nx 2 The Leibniz–Maclaurin method is demonstrated, using
⎢ ⎥ the above procedure, in the following worked problems.
⎣ + n(n − 1)(6x) + n(n − 1)(n − 2)} ⎦
y (3) = e2x (8x 3 + 36x 2 + 36x + 6) Problem 5. Determine the power series solution
of the differential equation:
3. Determine the 4th derivative of: y = 2x 3 e−x . d2 y dy
2
+ x + 2y = 0 using Leibniz–Maclaurin’s
[ y (4) = 2e−x (x 3 − 12x 2 + 36x − 24)] dx dx
method, given the boundary conditions that at
dy
4. If y = x 3 cos x determine the 5th derivative. x = 0, y = 1 and = 2.
dx
[ y (5) = (60x − x 3 ) sin x +
(15x 2 − 60) cos x] Following the above procedure:

5. Find an expression for y (4) if y = e−t sin t . (i) The differential equation is rewritten as:
[ y (4) = −4 e−t sin t ] y

+ x y
+ 2y = 0 and from the Leibniz theorem
of equation (13), each term is differentiated n
6. If y = x 5 ln 2x find y (3) . times, which gives:

[ y (3) = x 2 (47 + 60 ln 2x)]

y (n+2) +{y (n+1) (x)+n y (n) (1)+0}+2 y (n) = 0
7. Given 2x 2 y

+ x y
+ 3y = 0 show that i.e. y (n+2) + x y (n+1) + (n + 2) y (n) = 0
2x 2 y (n+2) + (4n + 1)x y (n+1) + (2n 2 − n +
3)y (n) = 0. (14)

8. If y = (x 3 + 2x 2 )e2x determine an expansion (ii) At x = 0, equation (14) becomes:

for y (5).
y (n+2) + (n + 2) y (n) = 0
[ y (5) = e2x 24 (2x 3 + 19x 2 + 50x + 35)]
from which, y (n+2) = −(n +2) y (n)
498 Higher Engineering Mathematics

This equation is called a recurrence relation (v) Collecting similar terms together gives:
or recurrence formula, because each recurring 
term depends on a previous term. 2x 2 2 × 4x 4
y = ( y)0 1 − +
2! 4!
(iii) Substituting n =0, 1, 2, 3, … will produce
a set of relationships between the various 2 × 4 × 6x 6 2 × 4 × 6 × 8x 8
− +
coefficients. 6! 8!
 5
For n =0, ( y

)0 = −2( y)0 3x 3 3 × 5x 5
− · · · + ( y
)0 x − +
n =1, ( y

)0 = −3( y
)0 3! 5!

n =2, ( y (4) )0 = −4( y

)0 = −4{−2( y)0 } 3 × 5 × 7x 7
− + ···
7!
= 2 × 4( y)0

x2 x4 x6
n =3, ( y (5) )0 = −5( y

)0 = −5{−3( y
)0 } i.e. y = ( y)0 1 − + −
1 1×3 3×5
= 3 × 5( y
)0 
x8
+ − ···
n =4, ( y (6) )0 = −6( y (4) )0 = −6{2 × 4( y)0 } 3×5×7
5
= −2 × 4 × 6( y)0 x x3 x5

+ ( y )0 × − +
n =5, ( y (7) )0 = −7( y (5) )0 = −7{3×5( y
)0 } 1 1×2 2×4
6
= −3 × 5 × 7( y
)0 x7
− +···
2×4×6
n =6, ( y (8) )0 = −8( y (6) )0 =
−8{−2 × 4 × 6( y)0}= 2 × 4 × 6×8(y)0 The boundary conditions are that at x = 0, y = 1
dy
and = 2, i.e. ( y)0 = 1 and ( y
)0 = 2.
(iv) Maclaurin’s theorem from page 69 may be dx
written as: Hence, the power series solution of the differen-
d2 y dy
tial equation: 2 + x + 2y = 0 is:
x 2

x3 dx dx
y = ( y)0 + x( y
)0 + ( y )0 + ( y

)0
2! 3! 
x2 x4 x6
x 4 (4) y = 1− + −
+ ( y )0 + · · · 1 1 ×3 3 ×5
4!  
x8 x x3
+ −··· +2 −
Substituting the above values into Maclaurin’s 3 ×5 × 7 1 1×2
theorem gives: 5 7 
x x
+ − +···
2×4 2×4×6
x2
y = ( y)0 + x( y
)0 + {−2( y)0 }
2!
Problem 6. Determine the power series solution
x3 x4 of the differential equation:
+ {−3( y
)0 } + {2 × 4( y)0 }
3! 4! d2 y d y
+ + x y = 0 given the boundary conditions
dx 2 dx
x5 x6 dy
+ {3 × 5( y
)0 } + {−2 × 4 ×6( y)0 } that at x = 0, y = 0 and = 1, using
5! 6! dx
Leibniz–Maclaurin’s method.
x7
+ {−3 × 5 × 7( y
)0 }
7! Following the above procedure:
x8 (i) The differential equation is rewritten as:
+ {2 × 4 × 6 × 8( y)0 }
8! y

+ y
+ x y = 0 and from the Leibniz theorem of
 Power series methods of solving ordinary differential equations 499

equation (13), each term is differentiated n times, ( y (8) )0 = −{( y (7) )0 + 6( y (5) )0 }
which gives:
= −[25 +6(6)] = −61
(n+2) (n+1) (n) (n−1)
y +y +y (x) + n y (1) + 0 = 0 (iv) Maclaurin’s theorem states:
i.e. y (n+2) + y (n+1) + x y (n) + n y (n−1) = 0 x2 x3
y = ( y)0 + x( y
)0 + ( y

)0 + ( y

)0
(15) 2! 3!
x 4 (4)
(ii) At x = 0, equation (15) becomes: +
( y )0 + · · ·
4!
y (n+2) + y (n+1) + n y (n−1) = 0 and substituting the above values into
Maclaurin’s theorem gives:
from which, y (n+2) = −{y (n+1) + n y (n−1) } x2 x3 x4
y = 0 + x(1) + {−1} + {1} + {−3}
This is the recurrence relation and applies for 2! 3! 4!
n ≥1
x5 x6 x7
+ {6} + {−10} + {25}
(iii) Substituting n = 1, 2, 3, . . . will produce a set of 5! 6! 7!
relationships between the various coefficients.
x8
For n = 1, ( y

)0 = −{( y

)0 + ( y)0 } {−61} + · · ·
+
8!
n = 2, ( y (4) )0 = −{( y

)0 + 2( y
)0 }
(v) Simplifying, the power series solution of
n = 3, ( y (5) )0 = −{( y (4) )0 + 3( y

)0 } d2 y d y
the differential equation: + + x y = 0 is
n = 4, ( y (6) )0 = −{( y (5) )0 + 4( y

)0 } dx 2 dx
given by:
n = 5, ( y (7) )0 = −{( y (6) )0 + 5( y (4) )0 }
n = 6, ( y (8) )0 = −{( y (7) )0 + 6( y (5) )0 } x2 x3 3x4 6x5 10x6
y = x− + − + −
From the given boundary conditions, at x = 0, 2! 3! 4! 5! 6!
dy
y = 0, thus ( y)0 = 0, and at x = 0, = 1, thus 25x7 61x8

dx + − +···
( y )0 = 1 7! 8!
From the given differential equation,
y

+ y
+ x y = 0, and, at x = 0, Now try the following exercise
( y

)0 + ( y
)0 + (0)y = 0 from which,
( y

)0 = −( y
)0 = −1
Exercise 195 Further problems on power
Thus, ( y)0 = 0, ( y
)0 = 1, ( y

)0 = −1, series solutions by the Leibniz–Maclaurin

method
( y

)0 = −{( y

)0 + ( y)0 } = −(−1 +0) = 1

1. Determine the power series solution of the dif-
( y (4) )0 = −{( y

)0 + 2( y
)0 } d2 y dy
ferential equation: 2 + 2x + y = 0 using
dx dx
= −[1 + 2(1)] = −3 the Leibniz–Maclaurin method, given that at
dy
( y (5) )0 = −{( y (4) )0 + 3( y

)0 } x = 0, y = 1 and = 2.
dx
⎡  ⎤
= −[−3 +3(−1)] =6 x 2 5x 4 5 × 9x 6
⎢ y = 1 − 2! + 4! − 6! ⎥
⎢ ⎥
( y (6) )0 = −{( y (5) )0 + 4( y

)0 } ⎢   ⎥
⎢ 5 × 9 × 13x 8 3x ⎥
3
⎢ + −··· +2 x − ⎥
= −[6 + 4(1)] = −10 ⎢ 3! ⎥
⎢ 8! ⎥
⎢  ⎥
⎢ ⎥
( y (7) )0 = −{( y (6) )0 + 5( y (4) )0 } ⎣ 3 × 7x 5 3 × 7 × 11x 7 ⎦
+ − +···
5! 7!
= −[−10 +5(−3)] =25
500 Higher Engineering Mathematics

(iv) equate coefficients of corresponding powers of

2. Show that the power series solution of the dif- the variable on each side of the equation;
d2 y dy this enables index c and coefficients a1 , a2 ,
ferential equation: (x + 1) 2 + (x − 1) −
dx dx a3 , … from the trial solution, to be determined.
2y = 0, using the Leibniz–Maclaurin method,
is given by: y = 1 + x 2 + ex given the boundary This introductory treatment of the Frobenius method
dy covering the simplest cases is demonstrated, using the
conditions that at x = 0, y = = 1.
dx above procedure, in the following worked problems.
3. Find the particular solution of the differ-
d2 y dy Problem 7. Determine, using the Frobenius
ential equation: (x 2 + 1) 2 + x − 4y = 0 method, the general power series solution of the
dx dx
using the Leibniz–Maclaurin method, given d2 y d y
differential equation: 3x 2 + − y = 0.
the boundary conditions that at x = 0, y = 1 dx dx
dy
and = 1.

dx  The differential equation may be rewritten as:
x 3 x 5 x 7 3x y

+ y
− y = 0.
y = 1 + x + 2x 2 + − + +···
2 8 16 (i) Let a trial solution be of the form

4. Use the Leibniz–Maclaurin method to deter- y = x c a0 + a1 x + a2 x 2 + a3 x 3 + · · ·
mine the power series solution for the differen- 4
+ ar x r + · · · (16)
d2 y d y
tial equation: x 2 + + x y = 1 given that
dx dx where a0 = 0,
dy
at x = 0, y = 1 and = 2. i.e. y = a0 x + a1 x
c c+1
+ a2 x c+2
+ a3 x c+3
dx
⎡  ⎤ + · · · + ar x c+r + · · · (17)
x2 x4 x6
⎢ y = 1 − 22 + 22 × 42 − 22 × 42 × 62 ⎥ (ii) Differentiating equation (17) gives:
⎢  5 ⎥
⎢ ⎥
⎢ x3 x5 ⎥ y
= a0cx c−1 + a1 (c + 1)x c
⎢ + ··· +2 x − 2 + 2 ⎥
⎢ 3 3 ×5 2 ⎥
⎢  ⎥
⎢ x 7 ⎥ + a2(c + 2)x c+1 + · · ·
⎣ − 2 + · · · ⎦
3 × 52 × 72 + ar (c + r)x c+r−1 + · · ·
and y

= a0c(c − 1)x c−2 + a1 c(c + 1)x c−1

+ a2 (c + 1)(c + 2)x c + · · ·

52.5 Power series solution by the + ar (c + r − 1)(c + r)x c+r−2 + · · ·

Frobenius method (iii) Substituting y, y
and y

into each term of the

given equation 3x y

+ y
− y = 0 gives:
A differential equation of the form y

+ P y
+ Qy = 0,
where P and Q are both functions of x, such that the 3x y

= 3a0 c(c − 1)x c−1 + 3a1 c(c + 1)x c

equation can be represented by a power series, may be
+ 3a2(c + 1)(c + 2)x c+1 + · · ·
solved by the Frobenius method.
The following 4-step procedure may be used in the + 3ar (c + r − 1)(c+r)x c+r−1 +· · · (a)
Frobenius method:
(i) Assume a trial solution of the form y4 = y
= a0 cx c−1 +a1 (c + 1)x c +a2 (c + 2)x c+1
:
xc a0 + a1 x + a2 x2 + a3 x3 + · · · + ar xr + · · ·
+ · · · + ar (c + r)x c+r−1 + · · · (b)
(ii) differentiate the trial series,
−y = −a0 x c − a1 x c+1 − a2 x c+2 − a3 x c+3
(iii) substitute the results in the given differential
equation, − · · · − ar x c+r − · · · (c)
 Power series methods of solving ordinary differential equations 501
a1 a0
(iv) The sum of these three terms forms the left-hand Thus, when r = 1, a2 = =
side of the equation. Since the right-hand side is (2 × 4) (2 × 4)
zero, the coefficients of each power of x can be since a1 = a0
equated to zero. a2 a0
when r = 2, a3 = =
For example, the coefficient of x c−1 is equated (3 × 7) (2 × 4)(3 × 7)
to zero giving: 3a0 c(c − 1) + a0 c = 0 a0
or
(2 × 3)(4 × 7)
or a0 c[3c − 3 + 1] = a0 c(3c − 2) = 0 (18) a3
when r = 3, a4 =
(4 × 10)
The coefficient of x c is equated to zero giving:
a0
3a1c(c + 1) + a1 (c + 1) − a0 = 0 =
(2 × 3 × 4)(4 × 7 × 10)
i.e. a1 (3c2 + 3c + c + 1) − a0 and so on.
= a1(3c2 + 4c + 1) − a0 = 0 From equation (16), the trial solution was:

or a1 (3c + 1)(c + 1) − a0 = 0 (19)

y = x c {a0 + a1 x + a2 x 2 + a3 x 3 + · · ·+ ar x r + · · ·}
In each of series (a), (b) and (c) an x c term is
involved, after which, a general relationship can Substituting c = 0 and the above values of a1 , a2 , a3, …
be obtained for x c+r , where r ≥ 0. into the trial solution gives:
In series (a) and (b), terms in x c+r−1 are present;
replacing r by (r + 1) will give the corresponding   
a0
terms in x c+r , which occurs in all three equa- y = x a0 + a0 x +
0
x2
(2 × 4)
tions, i.e.  
a0
in series (a), 3ar+1 (c + r)(c + r + 1)x c+r + x3
(2 × 3)(4 × 7)
in series (b), ar+1 (c + r + 1)x c+r   
a0
in series (c), −ar x c+r + x4 + · · ·
(2 × 3 × 4)(4 × 7 × 10)
Equating the total coefficients of x c+r to zero 
x2 x3
gives: i.e. y = a0 1 + x + +
(2 × 4) (2 × 3) (4 × 7)
3ar+1 (c + r)(c + r + 1) + ar+1 (c + r + 1) 
x4
+ +··· (21)
− ar = 0 (2 × 3 × 4)(4 × 7 × 10)
which simplifies to:

ar+1 {(c + r + 1)(3c + 3r +1)} − ar = 0 (20) 2

(b) When c = :
Equation (18), which was formed from the coeffi- 3
 
cients of the lowest power of x, i.e. x c−1, is called 2 5
From equation (19), if c = , a1(3) − a0 = 0, i.e.
the indicial equation, from which, the value of 3 3
c is obtained. From equation (18), since a0 = 0, a0
a1 =
2 5
then c = 0 or c = 2
3 From equation (20), if c =
  3
(a) When c = 0: ar+1
2
+ r + 1 (2 + 3r + 1) − ar = 0,
3
From equation (19), if c = 0, a1 (1 × 1) − a0 = 0,  
i.e. a1 = a0 5
i.e. ar+1 r + (3r + 3) − ar
3
From equation (20), if c = 0,
= ar+1 (3r 2 + 8r + 5) − ar = 0,
ar+1 (r + 1)(3r + 1) − ar = 0,
ar ar
i.e. ar+1 = r ≥0 i.e. ar+1 = r ≥0
(r + 1)(3r + 1) (r + 1)(3r + 5)
502 Higher Engineering Mathematics
a1 a0 x3
Thus, when r = 1, a2 = = +
(2 × 8) (2 × 5 × 8) (2 × 3)(5 × 8 × 11)
a0 
since a1 = x4
5 + +···
a2 (2 × 3 × 4)(5 × 8 × 11 × 14)
when r = 2, a3 =
(3 × 11)
Problem 8. Use the Frobenius method to
a0
= determine the general power series solution of the
(2 × 3)(5 × 8 × 11) differential equation:
a3 d2 y dy
when r = 3, a4 = 2x 2 2 − x + (1 − x)y = 0.
(4 × 14) dx dx
a0
=
(2×3×4)(5×8×11×14) The differential equation may be rewritten as:
and so on. 2x 2 y

− x y
+ (1 − x)y = 0.

From equation (16), the trial solution was: (i) Let a trial solution be of the form

y = x c {a0 + a1 x + a2 x 2 + a3 x 3 + · · ·+ ar x r + · · ·} y = x c {a0 + a1 x + a2 x 2 + a3 x 3 + · · ·
+ ar x r + · · ·} (23)
2
Substituting c = and the above values of a1 , a2 ,
3 where a0 = 0,
a3 , … into the trial solution gives:
 a    i.e. y = a0 x c + a1 x c+1 + a2 x c+2 + a3 x c+3
2 0 a0
y = x 3 a0 + x+ x2 + · · · + ar x c+r + · · · (24)
5 2×5×8
  (ii) Differentiating equation (24) gives:
a0
+ x3
(2 × 3)(5 × 8 × 11) y
= a0 cx c−1 + a1 (c + 1)x c + a2 (c + 2)x c+1
  
a0 + · · · + ar (c + r)x c+r−1 + · · ·
+ x4 + · · ·
(2 × 3 × 4)(5 × 8 × 11 × 14)
and y

= a0 c(c − 1)x c−2 + a1 c(c + 1)x c−1


2 x x2 + a2(c + 1)(c + 2)x c + · · ·
i.e. y = a0 x 1 + +
3
5 (2 × 5 × 8) + ar (c + r − 1)(c + r)x c+r−2 + · · ·
x3 (iii) Substituting y, y
and y

into each term of

+
(2 × 3)(5 × 8 × 11) the given equation 2x 2 y

− x y
+ (1 − x)y = 0
 gives:
x4
+ + ··· (22)
(2 × 3 × 4)(5 × 8 × 11 × 14) 2x 2 y

= 2a0 c(c − 1)x c + 2a1 c(c + 1)x c+1

Since a0 is an arbitrary (non-zero) constant in each + 2a2 (c + 1)(c + 2)x c+2 + · · ·
solution, its value could well be different. + 2ar (c + r − 1)(c + r)x c+r + · · ·
Let a0 = A in equation (21), and a0 = B in equation (22).
(a)
Also, if the first solution is denoted by u(x) and the
second by v(x), then the general solution of the given −x y
= −a0 cx c − a1 (c + 1)x c+1
differential equation is y = u(x) + v(x). Hence,
 − a2 (c + 2)x c+2 − · · ·
x2 x3
y = A 1 +x + + − ar (c + r)x c+r − · · · (b)
(2 × 4) (2 × 3)(4 × 7)

x4 (1 − x)y = (1 − x)(a0 x c + a1 x c+1 + a2 x c+2
+ +···
(2 ×3 × 4)(4 × 7 × 10) + a3 x c+3 + · · · + ar x c+r + · · ·)

2 x x2 = a0 x c + a1 x c+1 + a2 x c+2 + a3 x c+3
+Bx3 1+ +
5 (2 × 5 ×8) + · · · + ar x c+r + · · ·
 Power series methods of solving ordinary differential equations 503

− a0 x c+1 − a1 x c+2 − a2 x c+3 when r = 4,

− a3 x c+4
− · · · − ar x c+r+1
−··· (c) a3 a3
a4 = =
4(8 + 1) 4 × 9
a0
(iv) The indicial equation, which is obtained by =
(1 × 2 × 3 × 4) × (3 × 5 × 7 × 9)
equating the coefficient of the lowest power of
x to zero, gives the value(s) of c. Equating the and so on.
total coefficients of x c (from equations (a) to (c)) From equation (23), the trial solution was:
to zero gives: 
2a0c(c − 1) − a0 c + a0 = 0 y = x c a0 + a1 x + a2 x 2 + a3 x 3 + · · ·
4
i.e. a0 [2c(c − 1) − c + 1] = 0 + ar x r + · · ·
i.e. a0 [2c2 − 2c − c + 1] = 0 Substituting c = 1 and the above values of a1 , a2 ,
i.e. a0 [2c2 − 3c + 1] = 0 a3, … into the trial solution gives:
i.e. a0 [(2c − 1)(c − 1)] = 0 
1 a0 a0
y = x a0 + x+ x2
1 (1×3) (1×2)×(3×5)
from which, c = 1 or c =
2 a0
The coefficient of the general term, i.e. x c+r , + x3
(1 × 2 × 3) × (3 × 5 × 7)
gives (from equations (a) to (c)):
a0
+ x4
2ar (c + r − 1)(c + r) − ar (c + r) (1×2×3×4)×(3×5×7×9)

+ ar − ar−1 = 0
+ ···
from which,

x x2
ar [2(c + r − 1)(c + r) − (c + r) + 1] = ar−1 i.e. y = a0 x 1 1+ +
(1×3) (1 × 2) × (3 × 5)
ar−1
and ar = (25) x3
2(c +r −1)(c +r)−(c +r) +1 +
(1 × 2 × 3) × (3 × 5 × 7)
ar−1
(a) With c = 1, ar =
2(r)(1 + r) − (1 + r ) +1 x4
+
ar−1 (1×2×3×4)×(3×5×7×9)
= 
2r + 2r − 1 − r + 1
2
ar−1 ar−1 + ··· (26)
= 2 =
2r + r r (2r + 1)
1
Thus, when r = 1, (b) With c =
2
a0 a0 ar−1
a1 = = ar =
1(2 + 1) 1 × 3 2(c + r − 1)(c + r) − (c + r) + 1
when r = 2, from equation (25)
a1 a1 ar−1
a2 = = i.e. ar =     
2(4 + 1) (2 × 5) 1 1 1
2 +r −1 +r − + r +1
a0 a0 2 2 2
= or
(1 × 3)(2 × 5) (1 × 2) × (3 × 5) ar−1
=   
when r = 3, 1 1 1
2 r− r+ − −r +1
a2 a2 2 2 2
a3 = =
3(6 + 1) 3 × 7 ar−1
=  
a0 1 1
= 2 r2 − − −r +1
(1 × 2 × 3) × (3 × 5 × 7) 4 2
504 Higher Engineering Mathematics

ar−1 ar−1 solution of the given differential equation is

= = 2
1 1 2r −r y = u(x) + v(x),
2r 2 − − − r + 1
2 2 
ar−1 x x2
= i.e. y = A x 1 + +
r(2r − 1) (1 × 3) (1 × 2) × (3 × 5)
a0 a0 x3
Thus, when r = 1, a1 = = +
1(2 − 1) 1 × 1 (1 × 2 × 3) × (3 × 5 ×7)
a1 a1
when r = 2, a2 = = x4
2(4 − 1) (2 × 3) +
a0 (1 × 2 × 3×4)×(3×5×7×9)
=  
(2 × 3) 1 x2
+··· +Bx2 1+x+
a2 a2 (2 × 3)
when r = 3, a3 = =
3(6 − 1) 3 × 5 x3
a0 +
= (2 × 3) × (3 × 5)
(2 × 3) × (3 × 5) 
a3 a3 x4
when r = 4, a4 = = + +···
4(8 − 1) 4 × 7 (2 × 3 × 4) ×(3 × 5 ×7)
a0
=
(2×3×4)×(3×5×7) Problem 9. Use the Frobenius method to
and so on. determine the general power series solution of the
d2 y
From equation (23), the trial solution was: differential equation: 2 − 2y = 0.
dx

y = x c a0 + a1 x + a2 x 2 + a3 x 3 + · · ·
4 The differential equation may be rewritten as:
+ ar x r + · · · y

− 2y = 0.
1 (i) Let a trial solution be of the form
Substituting c = and the above values of a1 , a2 ,
2 
a3 , … into the trial solution gives: y = x c a0 + a1 x + a2 x 2 + a3 x 3 + · · ·
 4
1 a0 2 a0 + ar x r + · · · (28)
y=x a0 +a0 x +
2 x + x3
(2×3) (2×3)×(3×5)
 where a0 = 0,
a0
+ x4 + · · ·
(2 × 3 × 4) × (3 × 5 × 7)
i.e. y = a0 x c + a1 x c+1 + a2 x c+2 + a3 x c+3

1 x2 + · · · + ar x c+r + · · · (29)
i.e. y = a0 x 2 1 + x +
(2 × 3)
(ii) Differentiating equation (29) gives:
x3
+
(2 × 3) × (3 × 5) y
= a0cx c−1 + a1 (c + 1)x c + a2 (c + 2)x c+1
x4 + · · · + ar (c + r)x c+r−1 + · · ·
+
(2 × 3 × 4) × (3 × 5 × 7)

and y

= a0 c(c − 1)x c−2 + a1 c(c + 1)x c−1

+ ··· (27)
+ a2(c + 1)(c + 2)x c + · · ·
Since a0 is an arbitrary (non-zero) constant in + ar (c + r − 1)(c + r)x c+r−2 + · · ·
each solution, its value could well be different.
Let a0 = A in equation (26), and a0 = B in equa- (iii) Replacing r by (r + 2) in
tion (27). Also, if the first solution is denoted by ar (c + r − 1)(c + r) x c+r−2 gives:
u(x) and the second by v(x), then the general ar+2 (c + r + 1)(c + r + 2)x c+r
 Power series methods of solving ordinary differential equations 505
 
Substituting y and y

into each term of the given 2x 2 4x 4

= a0 1 + + +···
equation y

− 2y = 0 gives: 2! 4!
5 6
2x 3 4x 5
y

− 2y = a0 c(c − 1)x c−2 + a1 c(c + 1)x c−1 + a1 x + + +···

3! 5!
+ [a2(c+1)(c + 2)−2a0 ]x c +· · ·
+ [ar+2 (c + r + 1)(c + r + 2) Since a0 and a1 are arbitrary constants
depending on boundary conditions, let a0 = P
− 2ar ] x c+r + · · · = 0 (30) and a1 = Q, then:
 
(iv) The indicial equation is obtained by equating 2x2 4x4
the coefficient of the lowest power of x to zero. y=P 1 + + +···
2! 4!
Hence, a0 c(c − 1) =0 from which, c = 0 or  3 
2x 4x5
c = 1 since a0 = 0 +Q x+ + +··· (33)
3! 5!
For the term in x c−1 , i.e. a1c(c + 1) = 0 (b) When c =1: a1 = 0, and from equation (31),
With c = 1, a1 = 0; however, when c = 0, a1 is 2a0 2a0
indeterminate, since any value of a1 combined a2 = =
(2 × 3) 3!
with the zero value of c would make the product
zero. 2ar
Since c = 1, ar+2 =
For the term in x c , (c + r + 1)(c + r + 2)
2ar
=
a2 (c + 1)(c + 2) − 2a0 = 0 from which, (r + 2)(r + 3)
2a0 from equation (32) and when r = 1,
a2 = (31)
(c + 1)(c + 2) 2a1
a3 = = 0 since a1 = 0
For the term in x c+r , (3 × 4)
when r = 2,
ar+2 (c + r + 1)(c + r + 2) − 2ar = 0
2a2 2 2a0 4a0
from which, a4 = = × =
(4 × 5) (4 × 5) 3! 5!
2ar when r = 3,
ar+2 = (32)
(c + r + 1)(c + r + 2)
2a3
a5 = =0
(a) When c = 0: a1 is indeterminate, and from (5 × 6)
equation (31)
Hence, when c = 1,
2a0 2a0  
a2 = = 2a0 2 4a0 4
(1 × 2) 2! y = x 1 a0 + x + x +···
3! 5!
2ar from equation (28)
In general, ar + 2 = and
(r + 1)(r + 2) 5 6
2a1 2a1 2a1 2x 3 4x 5
when r = 1, a3 = = = i.e. y = a0 x + + + ...
(2 × 3) (1 × 2 × 3) 3! 3! 5!
2a2 4a0
when r = 2, a4 = = Again, a0 is an arbitrary constant; let a0 = K ,
3×4 4!
 
 2x3 4x5
2a0 2 2a1 3 then y=K x+ + +···
Hence, y = x 0 a0 + a1 x + x + x 3! 5!
2! 3!
 However, this latter solution is not a separate solution,
4a0 4
+ x + ··· for it is the same form as the second series in equation
4!
(33). Hence, equation (33) with its two arbitrary con-
from equation (28) stants P and Q gives the general solution. This is always
506 Higher Engineering Mathematics

the case when the two values of c differ by an integer (i.e.

whole number). From the above three worked problems, 3. Determine the power series solution of the
the following can be deduced, and in future assumed: d2 y dy
differential equation: 3x 2 + 4 − y = 0
dx dx
(i) if two solutions of the indicial equation differ by using the Frobenius method.
a quantity not an integer, then two independent ⎡  ⎤
solutions y = u(x) + v(x) result, the general solu- x x2
tion of which is y = Au + Bv (note: Problem 7 ⎢y = A 1 + (1 × 4) + (1 × 2)(4 × 7) ⎥
⎢ ⎥
2 1 ⎢ ⎥
had c = 0 and and Problem 8 had c = 1 and ; ⎢ x3 ⎥
3 2 ⎢ + +··· ⎥
⎢ (1 × 2 × 3)(4 × 7 × 10) ⎥
in neither case did c differ by an integer) ⎢  ⎥
⎢ 2 ⎥
⎢ −3
1 x x ⎥
(ii) if two solutions of the indicial equation do differ by ⎢ + Bx 1+ + ⎥
an integer, as in Problem 9 where c = 0 and 1, and ⎢ (1 × 2) (1 × 2)(2 × 5)⎥
⎢ ⎥
⎢ x3 ⎥
if one coefficient is indeterminate, as with when ⎣ + + ··· ⎦
c = 0, then the complete solution is always given (1 × 2 × 3)(2 × 5 × 8)
by using this value of c. Using the second value
of c, i.e. c = 1 in Problem 9, always gives a series 4. Show, using the Frobenius method, that
which is one of the series in the first solution. the power series solution of the differential
d2 y
equation: − y = 0 may be expressed as
Now try the following exercise dx 2
y = P cosh x + Q sinh x, where P and Q are
constants. [Hint: check the series expansions
Exercise 196 Further problems on power for cosh x and sinh x on page 47]
series solution by the Frobenius method
1. Produce, using Frobenius’ method, a power
series solution for the differential equation: 52.6 Bessel’s equation and Bessel’s
d2 y d y functions
2x 2 + − y = 0.
dx dx
⎡ ⎤ One of the most important differential equations in
 2
⎢y = A 1 + x + x ⎥ applied mathematics is Bessel’s equation and is of the
⎢ (2 × 3) ⎥ form:
⎢ ⎥
⎢ ⎥ d2 y
⎢ 3 ⎥ x2 2 + x
dy
+ (x 2 − v 2 )y = 0
⎢ +
x
+··· ⎥
⎢ ⎥ dx dx
⎢ (2 × 3)(3 × 5) ⎥
⎢  ⎥ where v is a real constant. The equation, which has
⎢ ⎥
⎢ 1 x x2 ⎥ applications in electric fields, vibrations and heat con-
⎢ +Bx2 1+ + ⎥
⎢ (1 × 3) (1 × 2)(3 × 5) ⎥ duction, may be solved using Frobenius’ method of the
⎢ ⎥
⎢  ⎥ previous section.
⎢ x 3 ⎥
⎣ + + ··· ⎦
(1 × 2 × 3)(3 × 5 × 7) Problem 10. Determine the general power series
solution of Bessels equation.
2. Use the Frobenius method to determine the
general power series solution of the differen- d2 y dy
d2 y Bessel’s equation x 2 2
+x + (x 2 − v 2 )y = 0 may
tial equation: 2 + y = 0. dx dx
dx be rewritten as: x 2 y

+ x y
+ (x 2 − v 2 )y = 0
⎡   ⎤
x2 x4 Using the Frobenius method from page 500:
⎢ y = A 1 − 2! + 4! − · · · ⎥
⎢ ⎥ (i) Let a trial solution be of the form
⎢  ⎥
⎢ 3 5 ⎥
⎢ + B x − + − ··· ⎥
x x
y = x c {a0 + a1 x + a2 x 2 + a3 x 3 + · · ·
⎢ ⎥
⎣ 3! 5! ⎦
+ ar x r + · · ·} (34)
= P cos x + Q sin x
where a0 = 0,
 Power series methods of solving ordinary differential equations 507

i.e. y = a0 x c + a1 x c+1 + a2 x c+2 + a3 x c+3 i.e. a1[2v + 1] = 0

+ · · · + ar x c+r
+··· (35) Similarly, if c = −va1[1 − 2v] = 0
(ii) Differentiating equation (35) gives: The terms (2v + 1) and (1 − 2v) cannot both be
zero since v is a real constant, hence a1 = 0.
y
= a0cx c−1 + a1 (c + 1)x c
Since a1 = 0, then from equation (37)
+ a2 (c + 2)x c+1 + · · ·
a3 = a5 = a7 = . . . = 0
+ ar (c + r)x c+r−1 + · · ·
and
and y

= a0 c(c − 1)x c−2 + a1 c(c + 1)x c−1

a0
+ a2 (c + 1)(c + 2)x c + · · · a2 =
v 2 − (c + 2)2
+ ar (c + r − 1)(c + r)x c+r−2 + · · · a0
a4 =
(iii) Substituting y, andy
y

into each term of [v − (c + 2) ][v 2 − (c + 4)2 ]

2 2

the given equation: x 2 y

+ x y
+ (x 2 − v 2 )y = 0 a0
a6 =
gives: [v 2 − (c + 2)2 ][v 2 −(c + 4)2 ][v 2 − (c + 6)2 ]
a0 c(c − 1)x c + a1 c(c + 1)x c+1 and so on.
+ a2(c + 1)(c + 2)x c+2 + · · · When c = +v,
+ ar (c + r − 1)(c + r)x c+r + · · · + a0 cx c
a0 a0
+ a1(c + 1)x c+1 + a2 (c + 2)x c+2 + · · · a2 = =
v − (v + 2)2
2 v − v − 4v − 4
2 2

+ ar (c + r)x c+r + · · · + a0 x c+2 + a1 x c+3 −a0 −a0

= = 2
+ a2 x c+4
+ · · · + ar x c+r+2 2 c
+ · · · − a0 v x 4 + 4v 2 (v + 1)
− a1 v x 2 c+1
− · · · − ar v x 2 c+r
+··· = 0 a0
a4 =  
(36) v 2 − (v + 2)2 v 2 − (v + 4)2
(iv) The indicial equation is obtained by equating a0
=
the coefficient of the lowest power of x to zero. [−22 (v + 1)][−23(v + 2)]
a0
Hence, a0c(c − 1) + a0 c − a0 v 2 = 0 =
2 (v + 1)(v + 2)
5
from which, a0 [c2 − c + c − v 2 ] = 0 a0
=
i.e. a0 [c2 − v 2 ] = 0 24 × 2(v + 1)(v + 2)
a0
from which, c = +v or c = −v since a0 = 0 a6 =
[v −(v+2) ][v −(v+4)2 ][v 2−(v+6)2 ]
2 2 2

For the term in x c+r , a0

=
ar (c + r − 1)(c + r) + ar (c + r) + ar−2 [2 × 2(v + 1)(v + 2)][−12(v + 3)]
4

− ar v 2 = 0 −a0
=
24 × 2(v + 1)(v + 2) × 22 × 3(v + 3)
ar [(c + r − 1)(c + r) + (c + r) − v 2 ] =−ar−2
−a0
i.e. ar [(c + r)(c + r − 1 + 1) − v 2 ] =−ar−2 = and so on.
26 × 3! (v + 1)(v + 2)(v + 3)
i.e. ar [(c + r)2 − v 2 ] =−ar−2
The resulting solution for c = +v is given by:
i.e. the recurrence relation is:
ar−2 y=u=
ar = for r ≥ 2 (37) 
v2 − (c + r)2 x2 x4
A x v 1− +
For the term in x c+1 , 22 (v +1) 24 × 2! (v +1)(v +2)
a1[c(c + 1) + (c + 1) − v 2 ] = 0 
x6
− 6 +···
i.e. a1 [(c + 1)2 − v 2 ] = 0 2 × 3! (v +1)(v + 2)(v + 3)
but if c = v a1 [(v + 1)2 − v 2 ] = 0 (38)
508 Higher Engineering Mathematics

which is valid provided v is not a negative integer upper case Greek letter gamma, and the gamma function
and where A is an arbitrary constant. (x) is defined by the integral
! ∞
When c = −v,
(x) = t x−1 e−t dt (40)
a0 a0 0
a2 = 2 = 2
v − (−v + 2)2 v − (v 2 − 4v + 4) and is convergent for x > 0
−a0 −a0 ! ∞
= = 2
4 − 4v 2 (v − 1) From equation (40), (x + 1) = t x e−t dt
0
a0
a4 = 2 and by using integration by parts (see page 420):
[2 (v − 1)][v 2 − (−v + 4)2 ]
  
a0  x  e−t ∞
= 2 (x + 1) = t
[2 (v − 1)][23 (v − 2)] −1 0
a0 ! ∞  −t 
= 4 e
2 × 2(v − 1)(v − 2) − x t x−1 dx
0 −1
a0 ! ∞
Similarly, a6 = 6
2 × 3! (v−1)(v−2)(v−3) = (0 − 0) + x e−t t x−1 dt
0
Hence,
= x(x) from equation (40)
y =w=
 This is an important recurrence relation for gamma
x2 x4 functions.
B x −v 1 + 2 + 4
2 (v−1) 2 ×2! (v−1)(v−2) Thus, since (x + 1) = x(x)

x6
+ 6 +··· then similarly, (x + 2) = (x + 1)(x + 1)
2 × 3! (v − 1)(v − 2)(v − 3)
= (x + 1)x(x) (41)
which is valid provided v is not a positive integer
and (x + 3) = (x + 2)(x + 2)
and where B is an arbitrary constant.
The complete solution of Bessel’s equation: = (x + 2)(x + 1)x(x),
and so on.
d2 y dy  2 
x2 +x + x − v 2 y = 0 is:
dx 2 dx These relationships involving gamma functions are used
y= u +w = with Bessel functions.

x2 x4
A xv 1 − 2 + 4
2 (v + 1) 2 × 2!(v + 1)(v + 2) Bessel functions

x6 The power series solution of the Bessel equation may be
− 6 +···
2 × 3!(v + 1)(v + 2)(v + 3) written in terms of gamma functions as shown in worked
 problem 11 below.
−v x2
+Bx 1+ 2
2 (v − 1) Problem 11. Show that the power series solution
x4 of the Bessel equation of worked problem 10 may
+ 4 be written in terms of the Bessel functions Jv (x)
2 × 2!(v − 1)(v − 2)
 and J−v (x) as:
x6
+ +· · · (39)
2 × 3!(v−1)(v−2)(v−3)
6 AJv (x) + BJ −v (x)
 x v  1 x2
= − 2
2 (v + 1) 2 (1! )(v + 2)
The gamma function 
x4
The solution of the Bessel equation of Problem 10 may + 4 −···
2 (2! )(v + 4)
be expressed in terms of gamma functions.  is the
 Power series methods of solving ordinary differential equations 509

 x −v  1 x2
This is called the Bessel function of the first order kind,
+ − 2 of order v, and is denoted by Jv (x),
(1 − v) 2 (1! )(2 − v)
 x v 
2
 1 x2
x4 i.e. Jv (x) = − 2
+ 4 −··· 2 (v + 1) 2 (1!)(v + 2)
2 (2! )(3 − v)

x4
+ 4 −···
From Problem 10 above, when c = +v, 2 (2!)(v + 3)
−a0
a2 = 2 provided v is not a negative integer.
2 (v + 1)
1 For the second solution, when c = −v, replacing v
If we let a0 =
2v (v + 1) by −v in equation (42) above gives:
then (−1)k
−1 −1 a2k =
a2 = = 22k−v (k! ) (k − v + 1)
22 (v + 1) 2v (v + 1) 2v+2 (v + 1)(v + 1) (−1)0
−1 from which, when k = 0, a0 =
= v+2 from equation (41) 2−v (0! )(1 − v)
2 (v + 2) =
1
since 0! = 1 (see page 495)
2−v (1 − v)
a2
Similarly, a4 = from equation (37) (−1)1
v − (c + 4)2
2
when k = 1, a2 =
22−v (1! )(1 − v + 1)
a2 a2
= = −1
(v − c − 4)(v + c + 4) −4(2v + 4) =
since c = v 22−v (1! )(2 − v)
−a2 −1 −1
= 3 = (−1)2
2 (v + 2) 23 (v + 2) 2v+2 (v + 2) when k = 2, a4 =
24−v (2! )(2 − v + 1)
1 1
= =
2v+4 (2! )(v + 3) 24−v (2! )(3 − v)
since (v + 2)(v + 2) = (v + 3)
(−1)3
−1 when k = 3, a6 =
and a6 = v+6 and so on. 26−v (3! )(3 − v + 1)
2 (3! )(v + 4)
1
The recurrence relation is: = and so on.
26−v (3! )(4 − v)
(−1)r/2 
ar = r    1 x2
r Hence, y = Bx −v −
2v+r !  v + +1 2−v (1 − v) 22−v (1! )(2 − v)
2 2

And if we let r = 2k, then x4
+ 4−v −···
2 (2! )(3 − v)
(−1)k 
a2k = (42)  x −v 1 x2
2v+2k (k!)(v + k + 1) i.e. J−v (x)= − 2
for k = 1, 2, 3, . . . 2 (1 −v) 2 (1!)(2 − v)

Hence, it is possible to write the new form for equation x4
+ 4 −· · ·
(38) as: 2 (2!)(3 −v)

v 1 x2 provided v is not a positive integer.
y = Ax −
2v (v + 1) 2v+2 (1! )(v + 2)
 Jv (x) and J−v (x) are two independent solutions of the
x4 Bessel equation; the complete solution is:
+ v+4 −···
2 (2! )(v + 3) y = AJ v (x) + B J −v (x) where A and B are constants
510 Higher Engineering Mathematics

i.e. y = AJ v (x)+ BJ −v (x) From this series two commonly used function are
 x v  1 x2 derived,
=A − 2
2 (v + 1) 2 (1!)(v + 2) 1 1  x 2 1  x 4
 i.e. J0(x) = − +
(0! ) (1! )2 2 (2! )2 2
x4
+ − ···
24 (2!)(v + 4) 1  x 6
− +···
 x −v  1 x2
(3! )2 2
+B − 2
2 (1 −v) 2 (1!)(2 − v) x2 x4 x6
= 1− + − +···
 22 (1!)2 24 (2!) 2
26 (3!)2
x4
+ 4 − ···
2 (2!)(3 −v)   x 2
x 1 1
and J1(x) = −
2 (1! ) (1! )(2! ) 2
 x v ;
∞ (−1)k x 2k 
In general terms: Jv (x) = 1  x 4
2 k=0 22k (k! )(v+k+1) + −···
(2! )(3! ) 2
 x −v ;∞ (−1)k x 2k
and J−v (x) =
k=0 2 (k! )(k − v + 1) x3 x5
2 2k x
= − 3 + 5
2 2 (1!)(2!) 2 (2!)(3!)
x7
− +···
Another Bessel function 27 (3!)(4!)

It may be shown that another series for Jn(x) is given by: Tables of Bessel functions are available for a range of
values of n and x, and in these, J0 (x) and J1(x) are most
 x n  1 1  x 2
Jn (x) = − commonly used.
2 n! (n + 1)! 2
 x 4  Graphs of J0 (x), which looks similar to a cosine, and
1 J1 (x), which looks similar to a sine, are shown in
+ − ···
(2! )(n + 2)! 2 Figure 52.1.

1
y ⫽ J0(x)

0.5

y ⫽ J1(x)

0 2 4 6 8 10 12 14 x

⫺0.5

Figure 52.1
 Power series methods of solving ordinary differential equations 511

i.e. y = a0 x c + a1 x c+1 + a2 x c+2 + a3 x c+3

Now try the following exercise + · · · + ar x c+r + · · · (44)

Exercise 197 Further problems on Bessel’s (ii) Differentiating equation (44) gives:
equation and Bessel’s functions
y
= a0 cx c−1 + a1 (c + 1)x c
1. Determine the power series solution of Bes-
d2 y dy + a2 (c + 2)x c+1 + · · ·
sel’s equation: x 2 2 + x + (x 2 −v 2 )y = 0
dx dx
when v = 2,up to and including the term in x 6. + ar (c + r)x c+r−1 + · · ·
 2 4
x x
y = Ax 2 1 − + − ··· and y

= a0 c(c − 1)x c−2 + a1 c(c + 1)x c−1

12 384
+ a2 (c + 1)(c + 2)x c + · · ·
2. Find the power series solution  of
 the + ar (c + r − 1)(c + r)x c+r−2 + · · ·
Bessel function: x 2 y

+ x y
+ x 2 − v 2 y = 0
in terms of the Bessel function J3(x) when
(iii) Substituting y, y
and y

into each term of the

v = 3. Give the answer up to and including the
given
 equation:

term in x 7 .
⎡  x 3  1 ⎤ 1 − x 2 y

− 2x y
+ k(k + 1)y = 0 gives:
x2
⎢ y = AJ3 (x) = 2 −
4 22 5 ⎥
⎢  ⎥ a0 c(c − 1)x c−2 + a1 c(c + 1)x c−1
⎣ x4 ⎦
+ 5 −···
2 6 + a2 (c + 1)(c + 2)x c + · · ·

3. Evaluate the Bessel functions J0 (x) and J1 (x) + ar (c + r − 1)(c + r)x c+r−2 + · · ·
when x = 1, correct to 3 decimal places.
[J0(x) = 0.765, J1(x) = 0.440] − a0 c(c − 1)x c − a1 c(c + 1)x c+1

− a2 (c + 1)(c + 2)x c+2 − · · ·

52.7 Legendre’s equation and − ar (c + r − 1)(c + r)x c+r − · · · − 2a0 cx c

Legendre polynomials − 2a1 (c + 1)x c+1 − 2a2 (c + 2)x c+2 − · · ·

Another important differential equation in physics − 2ar (c + r)x c+r − · · · + k 2 a0 x c

and engineering applications is Legendre’s equation
d2 y dy + k 2 a1 x c+1 + k 2 a2 x c+2 + · · · + k 2 ar x c+r
of the form: (1 − x 2 ) 2 − 2x + k(k + 1)y = 0 or
dx dx + · · · + ka0 x c + ka1 x c+1 + · · ·

(1 − x )y − 2x y + k(k + 1)y = 0 where k is a real

2

constant. + kar x c+r + · · · = 0 (45)

Problem 12. Determine the general power series (iv) The indicial equation is obtained by equating the
solution of Legendre’s equation. coefficient of the lowest power of x (i.e. x c−2 ) to
zero. Hence, a0c(c − 1) = 0 from which, c = 0 or
To solve Legendre’s equation c = 1 since a0 = 0.
(1 − x 2 )y

− 2x y
+ k(k + 1)y = 0 using the Frobenius
For the term in x c−1 , i.e. a1 c(c + 1) = 0 With
method:
c = 1, a1 = 0; however, when c = 0, a1 is inde-
(i) Let a trial solution be of the form terminate, since any value of a1 combined with

the zero value of c would make the product zero.
y = x c a0 + a1 x + a2 x 2 + a3 x 3
4 For the term in x c+r ,
+ · · · + ar x r + · · · (43) ar+2 (c + r + 1)(c + r + 2) −ar (c + r − 1)
where a0 = 0, (c + r) − 2ar (c + r) + k 2 ar + kar = 0
512 Higher Engineering Mathematics

from which, a0k(k + 1)(k − 2)(k + 3) 4

+ x
 4!
ar (c+r−1)(c+r)+2(c+r)−k 2 −k
ar+2 = a1 (k − 1)(k − 3)(k + 2)(k + 4) 5
(c+r+1)(c+r +2) + x
5!

ar [(c + r)(c + r + 1) − k(k + 1)]
= + ···
(c + r + 1)(c + r + 2) (46)

When c = 0, k(k + 1) 2
i.e. y = a0 1 − x
2!
ar [r(r + 1) − k(k + 1)] 
ar+2 = k(k +1)(k − 2)(k + 3) 4
(r + 1)(r + 2) + x −···
4!
For r = 0,

a0 [−k(k + 1)] (k − 1)(k + 2) 3
a2 = + a1 x − x
3!
(1)(2)

For r = 1, (k − 1)(k − 3)(k + 2)(k + 4) 5
+ x − · · · (47)
5!
a1[(1)(2) − k(k + 1)]
a3 = From page 506, it was stated that if two solutions of
(2)(3)
the indicial equation differ by an integer, as in this case,
−a1 [k 2 + k − 2] −a1 (k − 1)(k + 2) where c = 0 and 1, and if one coefficient is indetermi-
= =
3! 3! nate, as with when c = 0, then the complete solution is
For r = 2, always given by using this value of c. Using the second
 value of c, i.e. c = 1 in this problem, will give a series
a2 [(2)(3) − k(k + 1)] −a2 k 2 + k − 6 which is one of the series in the first solution. (This may
a4 = =
(3)(4) (3)(4) be checked for c = 1 and where a1 = 0; the result will be
the first part of equation (47) above).
−a2 (k + 3)(k − 2)
=
(3)(4)
−(k + 3)(k − 2) a0 [−k(k + 1)] Legendre’s polynomials
= .
(3)(4) (1)(2) (A polynomial is an expression of the form:
a0 k(k + 1)(k + 3)(k − 2) f (x) = a + bx + cx 2 + d x 3 + · · ·). When k in equation
= (47) above is an integer, say, n, one of the solution series
4!
terminates after a finite number of terms. For example,
For r = 3,
if k = 2, then the first series terminates after the term in
a3[(3)(4) − k(k + 1)] −a3 [k 2 + k − 12] x 2 . The resulting polynomial in x, denoted by Pn (x), is
a5 = = called a Legendre polynomial. Constants a0 and a1 are
(4)(5) (4)(5)
chosen so that y = 1 when x = 1. This is demonstrated
−a3 (k + 4)(k − 3) in the following worked problems.
=
(4)(5)
Problem 13. Determine the Legendre polynomial
−(k + 4)(k − 3) −a1 (k − 1)(k + 2)
= . P2 (x).
(4)(5) (2)(3)
a1(k − 1)(k − 3)(k + 2)(k + 4) Since in P2 (x), n =k = 2, then from the first part of
= and so on. equation (47), i.e. the even powers of x:
5!
 
Substituting values into equation (43) gives: 2(3) 2
 y = a0 1 − x + 0 = a0 {1 − 3x 2 }
a0 k(k + 1) 2 2!
y = x a0 + a1 x −
0
x
2! a0 is chosen to make y = 1 when x = 1
a1 (k − 1)(k + 2) 3 1
− x i.e. 1 = a0 {1 −3(1)2 } = −2a0 , from which, a0 = −
3! 2
 Power series methods of solving ordinary differential equations 513
 
1  1 d2 x 4 − 2x 2 + 1 d(4x 3 − 4x)
Hence, P2 (x)= − 1 − 3x 2 = (3x2 − 1) and = = 12x 2 − 4
2 2 dx 2 dx
Problem 14. Determine the Legendre poly-  
1 d2 x 4 −2x 2 +1 1 
nomial P3 (x). Hence, P2 (x) = 3 2
= 12x 2 − 4
2 dx 8
Since in P3 (x), n =k = 3, then from the second part of 1 2 
i.e. P2 (x) = 3x − 1 the same as in Problem 13.
equation (47), i.e. the odd powers of x: 2

(k − 1)(k + 2) 3 Problem 16. Determine the Legendre polynomial
y = a1 x − x P3 (x) using Rodrigue’s formula.
3!
  n
(k − 1)(k − 3)(k + 2)(k + 4) 5 1 dn x 2 − 1
+ x − ··· In Rodrigue’s formula, Pn (x) = n and
5! 2 n! dx n
  when n = 3,
(2)(5) 3 (2)(0)(5)(7) 5
i.e. y = a1 x − x + x  3
3! 5! 1 d3 x 2 − 1
  P3 (x) = 3
5 3 2 3! dx 3
= a1 x − x + 0   
3
1 d3 x 2 − 1 x 4 − 2x 2 + 1
= 3
a1 is chosen to make y = 1 when x = 1. 2 (6) dx 3
   
5 2 3  
i.e. 1 = a1 1 − = a1 − from which, a1 = − 1 d3 x 6 − 3x 4 + 3x 2 − 1
3 3 2 =
  (8)(6) dx 3
3 5 1  
Hence, P3 (x) =− x− x 3 or P3 (x) = (5x3− 3x)
2 3 2 d x 6 −3x 4 +3x 2 −1
= 6x 5 − 12x 3 + 6x
dx
 
d 6x 5 −12x 3 +6x
Rodrigue’s formula = 30x 4 − 36x 2 + 6
dx
An alternative method of determining Legendre poly-  
nomials is by using Rodrigue’s formula, which states: d 30x 4 − 36x 2 + 6
and = 120x 3 − 72x
 n dx
1 dn x2 − 1  
Pn (x)= n
dxn
(48) 1 d3 x 6 − 3x 4 + 3x 2 − 1
2 n! Hence, P3 (x) =
(8)(6) dx 3
This is demonstrated in the following worked problems.
1   1 
= 120x 3 − 72x = 20x 3 − 12x
Problem 15. Determine the Legendre polynomial (8)(6) 8
P2 (x) using Rodrigue’s formula. 1 
i.e. P3 (x)= 5x3 − 3x the same as in Problem 14.
 n 2
1 dn x 2 − 1
In Rodrigue’s formula, Pn (x) = n
2 n! dx n
and when n =2, Now try the following exercise
1 d 2 (x 2 − 1)2
P2 (x) =
22 2! dx 2 Exercise 198 Legendre’s equation and
Legendre polynomials
1 d2 (x 4 − 2x 2 + 1)
= 1. Determine the power series solution of
23 dx 2
the Legendre equation:
d 4  
(x − 2x 2 + 1) 1 − x 2 y

− 2x y
+ k(k + 1)y = 0 when
dx
(a) k = 0 (b) k = 2, up to and including the
= 4x 3 − 4x
514 Higher Engineering Mathematics

term in x 5 .  ⎤ 2. Find the following Legendre polynomials:

⎡
x3 x5 (a) P1 (x) (b) P4 (x) (c) P5 (x).
⎢(a) y = a0 + a1 x + 3 + 5 + · · · ⎥ ⎡ ⎤
⎢ ⎥ 1 
⎢ : 4 ⎥
⎢ (a) x (b) 8 35x − 30x + 3 ⎥
4 2
⎢(b) y = a0 1 − 3x 2 ⎥
⎢   ⎥ ⎣
⎣ 2 1 ⎦ 1  ⎦
+ a1 x − x 3 − x 5 (c) 63x 5 − 70x 3 + 15x
3 5 8