ODE: Assignment-5

1. Solve: (i) x2 y 00 + 2xy 0 − 12y = 0 (ii)(T) x2 y 00 + 5xy 0 + 13y = 0 (iii) x2 y 00 − xy 0 + y = 0

d2 y dy
[Recall: The ODE of the form x2 + ax + by = 0, where a, b are constants, is
dx2 dx
called the Cauchy-Euler equation. Under the transformation x = et (when x > 0) for
d2 y dy
the independent variable, the above reduces to 2 + (a − 1) + by = 0, which is an
dt dt
equation with constant coefficients. ]
Solution:
(i) Using the substitution x = et , the given equation reduces to

d2 u du
+ −12u = 0 =⇒ m2 +m−12 = 0 =⇒ m = −4, 3 =⇒ u(t) = Ae−4t +Be3t = y(et ).
dt2 dt
The general solution is thus
A
y(x) = 4
+ Bx3 .
x
(ii) Using the substitution x = et , the given equation reduces to,

d2 u du
2
+ 4 + 13u = 0 =⇒ m2 + 4m + 13 = 0 =⇒ m = −2 ± 3i.
dt dt
Thus
u(t) = e−2t (A cos 3t + B sin 3t) = y(et ).
The general solution is
1

y(x) = A cos(3 ln x) + B sin(3 ln x) .
x2

(iii) Using the substitution x = et , the given equation reduces to

d2 u du
2
− 2 + u = 0 =⇒ m2 − 2m + 1 = 0 =⇒ m = 1, 1 =⇒ u(t) = et (A + Bt) = y(et )
dt dt
The general solution is thus
y(x) = ex (A + B ln x).

2. (T) (Higher order variation of parameter) Consider the n-th order linear equation
n
X
(n)
y + ai (x)y (i) = y (n) + an−1 (x)y (n−1) + · · · + a0 (x)y = r(x).
1

Assume that y1 , · · · , yn are n-independent solutions of the associated homogeneous equa-

tion. Prove that a particular integral of the given ODE is
X Ri
yp = vi yi where vi0 = .
W
Here W is the wronskian of y1 , · · · , yn and Ri is the determinant obtained by replacing
i-th column of W by [0, 0, · · · , 0, r(x)].
 Solution:
Let X
yp = vi yi − − − − − −(1).
Differentiating yp0 =
P 0
vi yi + vi yi0 . Assume
P P 0
vi yi = 0 then
X
yp0 = vi yi0 − − − − − (2).

Differentiating this yp00 = vi0 yi0 + vi yi00 . Assuming

P P 0 0
vi yi = 0 we have
X
yp00 = vi yi00 . − − − − − −(3)

Proceeding similarly, we get

X (n−1)
yp(n−1) = vi yi − − − −(n)

(n−2)
vi0 yi
P
if = 0.

X (n−1)
X (n)
yp(n) = vi0 yi + vi yi − − − − − −(n + 1).

Then n
X X (n−1)
yp(n) + ai (x)yp(i) = vi0 yi .
1

Hence yp is a solution of the given ODE if

X X X X (n−2) X (n−1)
vi0 yi = 0, vi0 yi0 = 0, vi0 yi00 = 0, · · · , vi0 yi = 0, vi0 yi = r(x).

Solution such system of linear equation is given by vi0 = R W

i
where W is the wronskian
of y1 , · · · , yn and Ri is the determinant obtained by replacing i-th column of W by
[0, 0, · · · , 0, r(x)].
1
3. (T) Consider f (x) = e− x2 for x 6= 0 and f (0) = 0. Then:
(a) Calculate f 0 , f 00 , f 000 .
2
(b) Prove derivative of xcp e−1/x consists of sum of terms of similar form. Hence deduce
2
that f (n) (x) consists of sum terms of the form xcp e−1/x for different c, p ∈ N.
(c) Prove that
c −1/x2
lim e = 0, c, p ∈ N.
x→0 xp

(d) Deduce that f (n) (0) = 0 for all n.

(e) Thus conclude that f is infinitely differentiable but f is not analytic at 0.
[Recall: A real valued function is said to be analytic at x0 if f (x) can be written as a
an (x − x0 )n on |x − x0 | < R for some R > 0. A function
P
convergent power series
is analytic on a domain Ω if it is analytic at each x0 ∈ Ω. We know that any analytic
 function is infinitely differentiable BUT there exists infinitely real differentiable functions
which are not analytic. ]
Solution:
(a)
2 −1/x2 00 4 2 6 2 8 2 36 2 24 2
f 0 (x) = 3
e , f (x) = 6 e−1/x − 4 e−1/x , f 000 (x) = 9 e−1/x − 7 e−1/x + 5 e−1/x .
x x x x x x

(b)
d c −1/x2 pc 2 2c 2
( pe ) = − p+1 e−1/x + p+3 e−1/x .
dx x x x
c −1/x2
Clearly, by induction, f (n) (x) consists of sum terms of the form xp
e for different
c, p ∈ N.
(c)
c −1/x2 p −u2 cup
lim e = lim cu e = lim 2 = 0. c, p ∈ N.
x→0 xp u→∞ u→∞ eu

(d) Combining (b) and (c) we conclude that f (n) (0) = 0 for all n.
an xn on a nbd of 0, then an = f (n) (0)/n! = 0. Hence f = 0 on a nbd of
P
(e) If f (x) =
0. This is a contradiction. So f is not analytic at 0.

4. Let y1 (x), y2 (x) are two linearly independent solutions of y 00 + p(x)y 0 + q(x)y = 0. Show
that φ(x) = αy1 (x) + βy2 (x) and ψ(x) = γy1 (x) + δy2 (x) are two linearly independent
solutions if and only if αδ 6= βγ.
Solution:
We have W (φ, ψ) = (αδ−βγ)W (y1 , y2 ). Since y1 , y2 are fundamental solutions, W (y1 , y2 ) 6=
0. If αδ 6= βγ, then W (φ, ψ) 6= 0. Conversely if W (φ, ψ) 6= 0, then αδ 6= βγ.

5. Prove that if f, g are analytic at x0 and g(x0 ) 6= 0 then f /g is analytic at x0 .

Solution:
an (x − x0 )n and g(x) =
bn (x − x0 )n with g(x0 ) = b0 6= 0.
P P
Assume f (x) =
Claim: We can find cn ∈ R such that f /g = cn (x − x0 )n i.e.
P

X X X
an (x − x0 )n = bm (x − x0 )m ck (x − x0 )k .

Equating coefficients of different xn :

a0 = b0 c0 =⇒ c0 = a0 /b0 .
a1 = b0 c1 + b1 c0 =⇒ c1 can be found using known value ofc0 .
a2 = b0 c2 + b1 c1 + b2 c0 =⇒ c2 can be found using known values of c0 , c1 .
Thus inductively we can solve for all c0k s.
6. (T)(i) Prove that zeros of an analytic function f (x), which is not identically zero, are
isolated points i.e. if x0 is a zero of f (x) then there exists  > 0 such that f (x) 6= 0 for
all 0 < |x − x0 | < .
(T)(ii) Deduce that f, g analytic on an interval I and W (f, g) = 0 on I then f, g are
linearly dependent on I.
(Compare this with the result we have proved before: if W (y1 , y2 ) = 0 and they are
solution of second order linear homogeneous equation, then y1 , y2 are linearly dependent.)
Solution: (i) Write f (x) = n≥0 an (x − x0 )n on |x − x0 | < R for some R > 0. Since a
P

power series can be differentiated term by term, we get n!an = f (n) (x0 ). Since f (x0 ) = 0,
we have a0 = 0. Since f is not zero function there exists m such that am 6= 0. Choose m
to be the least such that am 6= 0. Then f (x) = am (x − x0 )m + am+1 (x − x0 )m+1 + · · · =
(x−x0 )m [am +am+1 (x−x0 )+· · · ] = (x−x0 )m g(x) where g is analytic and g(x0 ) = am 6= 0.
By continuity of g, there exists exists  > 0 such that g(x) 6= 0 for all |x − x0 | < . Hence
f (x) 6= 0 for all 0 < |x − x0 | < .
(ii) Given that f g 0 − f 0 g = 0 on an interval I. Since zeros of f are isolated points we can
choose an interval I 0 ⊂ I such that f 6= 0 on I 0 . Then on I 0 , we have (f g 0 − f 0 g)/f 2 = 0,
implies (g/f )0 = 0, imples g = cf on I 0 . Now h = g − cf is analytic on I and h is zero
on an interval I 0 i.e. h has non isolated zero. Hence by (i), we must have h = 0 on I.

7. Is x0 is an ordinary point of the ODE? If so expand p(x), q(x) in power series about x0 .
Find a minimum value for the radius of convergence of a power series solution about x0 .
(a) (x + 1)y 00 − 3xy 0 + 2y, x0 = 1
(T)(b) (1 + x + x2 )y 00 − 3y = 0, x0 = 1.
Solution:
(a) Here p(x) = −3x/(x + 1), q(x) = 2/(x + 1). Clearly x0 = 1 is an ordinary point.
1
Now x/(x + 1) = x/(2 + x − 1) = x2 1+(x−1)/2 = 21 (x − 1 + 1) [(1 − x)/2]n valid for
P

|1 − x| < 2.
The only singular point is x = −1. Thus the minimum radius of convergence of the
solution is the distance between x0 = 1 and −1, which is 2.
(b) Here p(x) = 0, q(x) = −3/(x2 + x + 1). Clearly x0 = 1 is an ordinary point.
√
The singular points are x = (−1 ± 3i)/2. Thus the minimum radius of convergence of
√ √
the solution is the distance between x0 = 1 and (−1 ± 3i)/2, which is 3.
Now for t = x − 1
1 1 1 1X
= = = [−(t2 + 3t)/3]n
x2 + x + 1 3 + 3t + t2 3(1 + [t2 + 3t)/3]) 3
√
valid for |(t2 + 3t)/3| < 1 that is |t| < 3.
8. Locate and classify the singular points in the following:
(T)(i) x3 (x − 1)y 00 − 2(x − 1)y 0 + 3xy = 0 (ii) (3x + 1)xy 00 − xy 0 + 2y = 0
Solution:
(i) The given ODE can be written as
2 0 3
y 00 − 3
y + 2 y=0
x x (x − 1)
Hence, x = 1 regular and x = 0 irregular singular points
(ii) The given ODE can be written as
1 2
y 00 − y0 + y=0
3x + 1 x(3x + 1)
Hence, both x = 0, x = −1/3 are regular singular points

9. Consider the equation y 00 + y 0 − xy = 0.

(i) Find the power series solutions y1 (x) and y2 (x) such that y1 (0) = 1, y10 (0) = 0 and
y2 (0) = 0, y20 (0) = 1.
(ii) Find the radius of convergence for y1 (x) and y2 (x).

Solution:
(i) Substituting y = n=0 an xn into y 00 + y 0 − xy = 0, we get
P

X X X
(n + 2)(n + 1)an+2 xn + (n + 1)an+1 xn − an−1 xn = 0
n=0 n=0 n=1

Rearranging, we find
X
(2a2 + a1 ) + [(n + 2)(n + 1)an+2 + (n + 1)an+1 − an−1 ] xn = 0
n=1

Hence,
an+1 an−1
2a2 + a1 = 0, an+2 = − + , n ≥ 1.
n + 2 (n + 1)(n + 2)
Iterating we get

a2 = −a1 /2, a3 = a1 /(2 · 3) + a0 /(2 · 3), a4 = a1 /(2 · 3 · 4) − a0 /(2 · 3 · 4), · · · .

Thus,
x3 x4 x5 x2 x3 x4 4x5
   
y = a0 1 + − + + · · · + a1 x − + + − + ···
2·3 2·3·4 2·3·4·5 2 2·3 2·3·4 2·3·4·5
= a0 y1 (x) + a1 y2 (x).

Now, y1 and y2 have the desired properties.

(ii) For the given ODE, p(x) = 1 and q(x) = −x both of which have radius of convergence
R = ∞. Hence, both y1 and y2 have radius of convergence R = ∞.
10. (T) Consider the equation (1 + x2 )y 00 − 4xy 0 + 6y = 0.

(i) Find its general solution in the form y = a0 y1 (x) + a1 y2 (x), where y1 (x) and y2 (x)
are power series.
(ii) Find the radius of convergence for y1 (x) and y2 (x).

Solution:
(i) Substituting y = n=0 an xn into (1 + x2 )y 00 − 4xy 0 + 6y = 0, we get
P
X X X X
(n + 2)(n + 1)an+2 xn + n(n − 1)an xn − 4nan xn + 6an xn = 0
n=0 n=2 n=1 n=0

Rearranging we find
X
(2a2 +6a0 )+(6a3 −4a1 +6a1 )x+ [(n + 2)(n + 1)an+2 + n(n − 1)an − 4nan + 6an ] xn = 0
n=2

Hence,
a1 (n − 2)(n − 3)
a2 = −3a0 , a3 = − , an+2 = − an , n ≥ 2.
3 (n + 1)(n + 2)
Iterating we get
a1
a2 = −3a0 , a3 = − , an = 0, n ≥ 4.
3
Thus,
x3
 
2


y = a0 1 − 3x + a1 x −
3
= a0 y1 (x) + a1 y2 (x)

(ii) Both the series are polynomials and hence converges for all x. Note that here
p(x) = −4x/(1 + x2 ) and q(x) = 6/(1 + x2 ) are analytic at x = 0 and have radius
convergence R = 1. Thus the existence and uniqueness theorem for the ordinary point
guarantees existence of unique solution in |x| < 1 but actually we find the existence of
unique solution for all x.

11. Find the first three non zero terms in the power series solution of the IVP

y 00 − (sin x)y = 0, y(π) = 1, y 0 (π) = 0.

Solution: As the initial values are given at π, the expansion should be about x0 = π.
The best way to do this is to first shift x0 to 0. To do this, let t = x − π. Then
t0 = x0 − π = 0. The equation becomes

y 00 + (sin x)y = 0, y(0) = 1, y 0 (0) = 0.

(−1)n 2n+1
an xn and using sin x =
P P
Assuming y = (2n+1)!
x we get

0 = y 00 + (sin x)y = 2a2 + (6a3 + a0 )x + (12a4 + a1 )x2 + (20a5 + a2 − a0 /6) + · · · .

From initial conditions a0 = 1, a1 = 0. So a2 = 0, a3 = −1/6, a4 = 0, a5 = 1/120.

12. The function on the left side of
∞
1 X
√ = Pn (x)tn
1 − 2xt + t2
n=0

is called the generating function of the Legendre polynomial Pn . Assuming this, show
that
(T)(i) (n + 1)Pn+1 (x) − (2n + 1)xPn (x) + nPn−1 (x) = 0 (ii) nPn (x) = xPn0 (x) − Pn−1
0
(x)
0
(iii) Pn+1 (x) − xPn0 (x) = (n + 1)Pn (x) ; (iv) Pn (1) = 1, Pn (−1) = (−1)n
1 · 3 · 5 · · · · · · (2n − 1)
(v) P0 (0) = 1, P2n+1 (0) = 0, P2n (0) = (−1)n , n≥1
2n n!
Solution:
(i) Differentiating both sides w.r.t. t, we get
x−t X
2 3/2
= nPn (x)tn−1
(1 − 2xt + t ) n=1

which gives
X X
(x − t) Pn (x)tn = (1 − 2xt + t2 ) (n + 1)Pn+1 (x)tn
n=0 n=0

Equating the coefficient of tn from both sides, we get

xPn − Pn−1 = (n + 1)Pn+1 − 2xnPn + (n − 1)Pn−1 ,

which on simplification yields

(n + 1)Pn+1 (x) − (2n + 1)xPn (x) + nPn−1 (x) = 0

(ii) Differentiating both sides w.r.t. x, we get

t X
= Pn0 (x)tn
(1 − 2xt + t2 )3/2 n=0

which gives X X
(1 − 2xt + t2 ) Pn0 tn = t Pn tn
n=0 n=0
n
Equating the coefficient of t from both sides, we get

Pn0 − 2xPn−1
0 0
+ Pn−2 = Pn−1

which on replacing n by n + 1 gives

0
Pn+1 − 2xPn0 − Pn + Pn−1
0
= 0. (*)
Differentiating the relation in (i) w.r.t. x, we get
 
0
(n + 1)Pn+1 − (2n + 1) Pn + xPn0 + nPn−1
0
= 0. (**)
 0
Elimination of Pn+1 between (*) and (**) gives

0
nPn (x) = xPn0 (x) − Pn−1 (x)

(iii) Proceeding as in (ii) we arrive in relation given in (*) and (**). Eliminate p0n−1
between (*) and (**) to find

0
Pn+1 (x) − xPn0 (x) = (n + 1)Pn (x)

(iv) Substituting x = 1 into the relation we find

X 1 X
Pn (1)tn = = tn
n=0
1 − t n=0

Equating coefficients of tn , we get Pn (1) = 1.

Similarly, substituting x = −1 into the relation we find
X 1 X
Pn (−1)tn = = (−1)n tn
n=0
1 + t n=0

Equating coefficients of tn , we get Pn (−1) = (−1)n .

(v) Substitute x = 0 into the relation we get
X −1 −1 − 1 −1 − 2 · · · −1 − n + 1





X
n 1
Pn (0)t = √ =1+ 2 2 2 2
t2n
n=0
1+t 2
n=1
n!

or
X X X 1 · 3 · 5 · · · (2n − 1) 2n
P0 (0) + P2n (0)t2n + P2n+1 (0)t2n+1 = 1 + (−1)n n n!
t
n=1 n=1 n=1
2

Equating the coefficients of tn we get

1 · 3 · 5 · · · · · · (2n − 1)
P0 (0) = 1, P2n+1 (0) = 0, P2n (0) = (−1)n , n≥1
2n n!

13. Expand the following functions in terms of Legendre polynomials over [−1, 1]:
(
0 if −1 ≤ x < 0
(i) f (x) = x3 + x + 1 (T)(ii) f (x) = (first three nonzero
x if 0 ≤ x ≤ 1
terms)
Solution:
We know from Legendre Expansion Theorem that any continuous function f (x) on
[−1, 1], has Legendre series expansion as

2n + 1 1
X Z
f (x) = an Pn (x), with an = f (x)Pn (x) dx; x ∈ [−1, 1].
n=0
2 −1
 ( See N. N. Lebedev, Special Functions and Their Applications, pp. 53 − 58, Prentice-
Hall, Englewood Cliffs, N.J. , 1965.)
(i) We can use the above formula to find an . Alternately, we know that
5x3 − 3x
P0 (x), P1 (x) = x, P3 (x) = .
2
So we find
2P3 (x) + 3P1 (x)
1 = P0 (x), x = P1 (x), x3 = .
5
Hence,
2P3 (x) + 3P1 (x) 8 2
f (x) = P0 (x) + P1 (x) + = P0 (x) + P1 (x) + P3 (x)
5 5 5
R1 n R1
(Remark: Note that, if f has derivatives of all order then, −1 f (x)Pn (x)dx = (−1)
2n n! −1
f (n) (x)(x2 −
1)n dx. In particular, if f (x) is a polynomial of degree n then am = 0 for all m > n.)
(ii) Using the above formula,
1 1 5
a0 = , a 1 = , a 2 = .
4 2 16
Thus,
1 1 5
f (x) = P0 (x) + P1 (x) + P2 (x) + · · ·
4 2 16
14. For each of the following, verify that the origin is a regular singular point and find two
linearly independent solutions:
(a) 9x2 y 00 + (9x2 + 2)y = 0 (b) x2 (x2 − 1)y 00 − x(1 + x2 )y 0 + (1 + x2 )y = 0
(T) (c) xy 00 + (1 − 2x)y 0 + (x − 1)y = 0 (d) x(x − 1)y 00 + 2(2x − 1)y 0 + 2y = 0
Solution:
(a)
The given ODE can be written as
9x2 + 2
y 00 + y=0
9x2
Hence x = 0 is a regular singular point. Let y = n=0 an xn+r , a0 6= 0. This gives
P

X X X
9(n + r)(n + r − 1)an xn+r + 9an xn+r+2 + 2an xn+r = 0
n=0 n=0 n=0

which can be written as

X X X
9(n + r)(n + r − 1)an xn+r + 9an−2 xn+r + 2an xn+r = 0
n=0 n=2 n=0

This can be rearranged as (after canceling xr )

    X 
9r(r − 1) + 2 a0 + 9r(r + 1) + 2 a1 x + 9(n + r)(n + r − 1)an + 9an−2 + 2an xn = 0
n=2
This implies
    9an−2
9r(r−1)+2 a0 = 0, 9r(r+1)+2 a1 = 0 and an = − , n ≥ 2.
9(n + r)(n + r − 1) + 2

Since a0 6= 0, we have 9r(r − 1) + 2 = 0 =⇒ r = 2/3 = r1 , r = 1/3 = r2 . Here

r1 − r2 = 1/3 is not an integer and we have two independent Frobenius series solutions.
With r = r1 or r = r2 , 9r(r + 1) + 2 6= 0 =⇒ a1 = 0. This leads to a2n+1 = 0, n ≥ 0.
Also,
9an−2
an = − , n ≥ 2.
(3n + 3r − 2)(3n + 3r − 1)

With r = r1 = 2/3 we find

X 3a2n−2
y1 (x) = x2/3 a2n x2n , a0 = 1, a2n = − , n ≥ 1.
n=0
2n(6n + 1)

With r = r1 = 1/3 we find

X 3a2n−2
y2 (x) = x1/3 a2n x2n , a0 = 1, a2n = − , n ≥ 1.
n=0
2n(6n − 1)

(b)
The given ODE can be written as

1 + x2 0 1 + x2
y 00 − y + =0
x(x2 − 1) x2 (x2 − 1)

Hence x = 0 is a regular singular point. Let y = n=0 an xn+r , a0 6= 0. This gives

P

X 
n+r+2 n+r n+r n+r+2 n+r n+r+2
(n+r)(n+r−1)an (x −x )−(n+r)an (x +x )+an (x +x ) =0
n=0

which can be written as

X  X 
(n+r−2)(n+r−3)−(n+r−2)+1 an−2 xn+r − (n+r)(n+r−1)+(n+r)−1 an xn+r = 0
n=2 n=0

This can be rearranged as (after canceling xr )

    X

− r2 − 1 a0 − (r + 1)2 − 1 a1 x + (n + r − 3)2 an−2 − (n + r)2 − 1 an xn = 0
n=2

This implies

2
 
2
 (n + r − 3)2
r − 1 a0 = 0, (r + 1) − 1 a1 = 0, and an = an−2 , n ≥ 2.
(n + r)2 − 1

Since a0 6= 0, we have r2 − 1 = 0 =⇒ r = 1 = r1 , r = −1 = r2 . Here r1 − r2 = 2 is an

integer and we may or may not have two independent Frobenius series solutions.
With r = r1 , (r + 1)2 − 1 6= 0 =⇒ a1 = 0. Also,
(n − 2)2
an = an−2 , n ≥ 2 =⇒ an = 0, n ≥ 1.
n(n + 2)
Hence
y1 (x) = x, a0 = 1.
For the other solution, let y2 = y1 u(x) = xu (reduction of order technique)
u00 1 1 3
x(x2 − 1)u00 + (x2 − 3)u0 = 0 =⇒ 0
= − − =⇒ u0 = 1/x3 − 1/x
u 1+x 1−x x
which integrating again gives
1
u = − log x −
2x2
Hence y2 = x ln x + 1/(2x) (ignoring the negative sign)
(c)
The given ODE can be written as
1 − 2x 0 x − 1
y 00 +
y + =0
x x
Hence x = 0 is a regular singular point. Let y = n=0 an xn+r , a0 6= 0. This gives
P

X 
(n + r)(n + r − 1)an xn+r−1 + (n + r)an (xn+r−1 − 2xn+r ) + an (xn+r+1 − xn+r ) = 0
n=0

which can be written as

X X  X 
an−2 xn+r−1 − 2(n+r−1)+1 an−1 xn+r−1 + (n+r)(n+r−1)+(n+r) an xn+r−1 = 0
n=2 n=1 n=0

This can be rearranged as (after canceling xr−1 )

  X 
r2 a0 + (r + 1)2 a1 − (2r + 1)a0 x + (n + r)2 an − 2(n + r − 1) + 1 an−1 + an−2 xn = 0


n=2

This implies

r2 a0 = 0, (r + 1)2 a1 = (2r + 1)a0 , (n + r)2 an = 2(n + r) − 1 an−1 − an−2 , n ≥ 2

Now a0 6= 0 =⇒ r = r1 = 0, r = r2 = 0. Since the indicial equation has double roots,

the given equation has only one independent Frobenius series solution. We take r = 0
and this gives a1 = a0 . We also have
2n − 1 1
an = 2
an−1 − 2 an−2 , n ≥ 2.
n n
With a0 = 1 we get a1 = 1. This leads to a2 = 1/2!, a3 = 1/3!. We prove an = 1/n!
by induction. Clearly the induction hypothesis is true for n = 1, 2, 3. Let it be true for
n = k. For n = k + 1, we have
 
2k + 1 1 1 2k + 1 1
ak+1 = a k − a k−1 = − 1 =
(k + 1)2 (k + 1)2 (k + 1)2 (k − 1)! k (k + 1)!
Hence X xn
y1 (x) = = ex
n=0
n!
For other solution let y2 = y1 u(x) = ex u. This gives

xu00 + u0 = 0 =⇒ u0 = 1/x =⇒ u = ln x

Hence y2 (x) = ex ln x
(d)
The given ODE can be written as
2(2x − 1) 0 2
y 00 + y + y=0
x(x − 1) x(x − 1)
Hence x = 0 is a regular singular point. Let y = n=0 an xn+r , a0 6= 0. This gives
P

X 
(n + r)(n + r − 1)an (xn+r − xn+r−1 ) + (n + r)an (4xn+r − 2xn+r−1 ) + 2an xn+r = 0
n=0

which can be written as

X  X 
(n+r−1)(n+r−2)+4(n+r−1)+2 an−1 xn+r−1 − (n+r)(n+r−1)+2(n+r) an xn+r−1 = 0
n=1 n=0

This can be rearranged as (after canceling xr−1 )

X 
(r2 + r)a0 − (n + r)(n + r + 1)an − (n + r − 1)(n + r + 2) + 2 an−1 xn = 0


n=1

This implies

(r2 + r)a0 = 0, (n + r)(n + r + 1)an − (n + r − 1)(n + r + 2) + 2 an−1 = 0, n ≥ 1

Now a0 6= 0 =⇒ r = r1 = 0, r = r2 = −1. Hence r1 − r2 = 1 is an integer and hence

the ODE may or may not have two independent Frobenius series solution.
With r = r1 = 0,


n(n + 1)an = (n − 1)(n + 2) + 2 an−1 =⇒ an = an−1 =⇒ an = a0 , n ≥ 1.

Hence (with a0 = 1)
X 1
y1 (x) = xn =
n=0
1−x
For the other solution, let y2 = y1 u(x). This gives
1
xu00 + 2u0 = 0 =⇒ u0 = =⇒ u = −1/x
x2
Hence (neglecting the negative sign)
1
y2 (x) =
x(1 − x)
 We can write
1 1
y2 (x) = +
x 1−x
Since the last term is y1 (x), we can take y2 (x) = 1/x
Note: If we continue the Frobenius series method with r = r2 = −1, then from the
recurrence relation
n(n − 1)an = n(n − 1)an−1 , n ≥ 1.
For n = 1, the relation is automatically satisfied for any value of a1 . We may take
a1 = 0. This leads to an = 0 for n ≥ 1. Then we again get (taking a0 = 1)
1
y2 (x) =
x

15. Show that 2x3 y 00 + (cos 2x − 1)y 0 + 2xy = 0 has only one Frobenius series solution.
Solution:
We can write the ODE as
cos 2x − 1 0
2x2 y 00 + xy + 2y = 0
x2
Since
cos 2x − 1
lim = −2,
x→0 x2
the indicial equation is

2r(r − 1) − 2r + 2 =⇒ r2 − 2r + 1 =⇒ r = 1, 1.

Since the indicial equation has double roots, it has only one Frobenius series solution.