MA1201- MATHEMATICS-II

Linear Differential Equations

Dr. Pratibhamoy Das

Department of Mathematics
Indian Institute of Technology Patna
India 801106

January 20, 2025

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 1 / 28

Contents

Ordinary Linear Differential Equations of nth order

Solutions of homogeneous and nonhomogeneous equations (Method of variation

of parameters)

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 2 / 28

Introduction

Recall: C n (I, R) of all real functions in I that have continuous derivatives up to

order n. C(I, R): the Vector Space (VS) of all real continuous functions on I .
The differential operator D : C 1 (I, R) → C(I, R) defined by D(f ) = f ′ is a Linear
Transform (LT). The same is true for Dn : C n (I, R) → C(I, R).
Similarly L : C n (I , R) → C(I , R), where L = a0 D (n) + a1 D (n−1 ) + ... + an I,
ai ∈ C(I, R), and I the identity LT. (We often write Lf for L(f ).)
For any choices of ai (x) and b ∈ C(I, R),the differential equation Ly = b, i.e.,

a0 y (n) + a1 y (n−1) + ... + an y = b, (1)

is called a linear differential equation. The order of (1) is n, if a0 ̸= 0.

The equation (1) with b = 0, (i.e., the equation Ly = 0) is said to be
homogeneous. If b ̸= 0, then Ly = b is non-homogeneous. In that case, Ly = 0 is
called the corresponding homogeneous equation of Ly = b.
A solution of Ly = b is a function ϕ ∈ C n (I, R) such that Lϕ = b.
In particular,the general solution (solution set) of Ly = 0 is KER(L), a
subspace of C n (I, R).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 3 / 28

Solving Linear Equations

Consider a first-order linear differential equation:

y ′ + P (x)y = Q(x), where P, Q ∈ C(R, R). (2)

The general
R solution isR {cϕ + ψR : cR∈ R} is an arbitrary constant and
ϕ = e− P (x)dx , ψ = e− P (x)dx (e P (x)dx Q(x))dx.
Note: ψ is a particular solution of (2), and {cϕ : c ∈ R} is the general solution of
y ′ + P y = 0.
Consider a linear equation Ly = b. Let u be a particular integral. Then the
general solution is u + W, where W = KER(L), the general solution of Ly = 0.
A basis for W is called a set of fundamental solutions of Ly = 0.
Ex. Consider the equation y ′′ + 4y = 0. Then, cos 2x and sin 2x are two of its
solutions, i.e., they are in KER(D2 + 4I). Let us allow the solutions to be in
C n (I, C), a VS on C. Then e2ix and e−2ix are also solutions.
LS{cos 2x, sin 2x}= LS{e2ix , e−2ix } as subspaces of C n (I, C).
Ex. Consider x2 y ′′ − 2xy ′ + 2y = x3 on I = (0, ∞). For any c1 , c2 ∈ R (or in the
field C) y(x) = c1 x2 + c2 x + 12 x3 is a solution. This gives the general solution iff
KER(x2 D2 − 2xD + 2I) = LS{x, x2 }.

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 4 / 28

Linearly Independent Solutions

EXISTENCE AND UNIQUENESS THEOREM (EUT): Suppose

a0 (x) ̸= 0 ∀x ∈ I, and ai (x), b(x) ∈ C(I), x0 ∈ I. Then, for any choice of
α0 , ..., αn−1 ∈ R, the initial value problem (IVP)

Ly ≡ a0 y (n) + a1 y (n−1) + ... + an y = b,

y(x0 ) = α0 , y ′ (x0 ) = α1 , · · · , y (n−1) (x0 ) = αn−1 ,

has one and only one solution. [Proof is very Simple & Exercise.]
THEOREM: DIM (KER(L)) = n.
Proof: Consider T : KER(L) → Rn : T y := [y(x0 ), y ′ (x0 ), ..., y (n−1) (x0 )]t .
Then T is linear. Appeal to EUT. For each α ∈ Rn , there is y ∈ KER(L) such
that T y = α. Therefore, T is onto. The solution u is unique, So, T is one one.
Thus, KER(L) and Rn are isomorphic, and

DIM (KER(L)) = DIM (Rn ) = n.

Pn
Corollary: Suppose Ly = 0 have order n. Then, its general solution is i=1 ci yi ,
where {yi : 1 ≤ i ≤ n} is any L.I set of solutions.
One can choose yi = T −1 ei .

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 5 / 28

Wronskian Matrix

Consider Ly = (D2 + 4)y = 0. We have {cos 2x, sin 2x} ∈ KER(L).

KER(L) = LS({cos 2x, sin 2x}) iff {cos 2x, sin 2x} is linearly independent in
C 2 (R, R).
 
α1
Suppose {cos 2x, sin 2x} is linearly dependent. Then, ∃ 0 ̸= α = ∈ R2 such
α2
that ∀x ∈ R, α1 cos 2x + α2 sin 2x = 0.
Differentiating, we get −2α1sin  2x + 2α2 cos 2x = 0.
a
Thus, α is a solution of A(x) = 0,
b
 
cos 2x sin 2x
where A(x) =
−2 sin 2x 2 cos 2x
This cannot be true, since DET (A(x)) = 2, ∀x ∈ R. Therefore, {cos 2x, sin 2x}
is linearly independent.
Suppose {u1 , ..., un } ⊆ C n (I, R). The Wronskian matrix of u1 , ..., un at x ∈ I is
defined to be  
u1 (x) u2 (x) ... un (x)
′
 u1 (x) ′ ′
u2 (x) ... un (x) 
[W (u1 , ..., un )(x)] :=  .
 ... ... ... ... 
u1 n−1 (x) u2 n−1 (x) ... un n−1 (x)

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 6 / 28

Wronskian and Linear Dependence

The Wronskian of u1 , ..., un ∈ C n (I, R) is the function (on I) defined by

DET [W (u1 , ..., un )].
Suppose u1 , ..., un ∈ C n (I, R), and W (u1 , ..., un )(x0 ) ̸= 0 for some x0 ∈ I. Then
{u1 , ..., un } is linearly independent.
THEOREM. Consider Ly = 0, where L = a0 Dn + ... + an I and a0 (x) ̸= 0 for all
x ∈ I. Let x0 be any point in I. A set of SOLUTIONS of the homogeneous linear
nth order ODE {u1 , ..., un } on I is linearly independent iff W (u1 , ..., un )(x0 ) ̸= 0.
Proof. (⇐) Follows from above.
(⇒) Suppose W (u1 , ..., un )(x0 ) = 0. Then, theP system [W (u1 , ..., un )(x0 )]X = 0
has a nontrivial solution [α1 , ..., αn ]t . Let v = ni=1 αi ui . We have Lv = 0 and

v(x0 ) = v ′ (x0 ) = ... = v (n−1) (x0 ) = 0.

By EUT, v ≡ 0, that is, ui ’s are linearly dependent.

Ex. Take distinct reals (or complex numbers) αi ∈ R, 1 ⩽ i ⩽ k. Then,
{eαi x |1 ⩽ i ⩽ k} is linearly independent in C n (R, R)(or in C n (R, C)).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 7 / 28

Linear ODE with Constant Coefficients

Consider Ly = b, where L = a0 Dn + .... + an−1 D + an I. If ai are constants, then

Ly = b is said to be linear with constant coefficients. In this case, L = p(D),
where p(λ) = a0 λn + ..... + an−1 λ + an ∈ R[λ].
p(λ) = 0 is called the auxiliary equation of Ly = 0.
Let λ ∈ R (or in C). Then, Deλx = λeλx , i.e., eλx ∈ KER(D − λI). So,
KER(D − λI) = LS{eλx }.
For any polynomial q(x) ∈ R[x], we have (D − λI)(q(x)eλx ) = eλx D(q(x)).
Suppose deg(q) ⩽ k − 1. Then (D − λI)k (q(x)eλx ) = eλx Dk (q(x)) = 0, i.e.,
q(x)eλx ∈ KER((D − λI)k ). So, KER((D − λI)k ) = LS{eλx , xeλx , ..., xk−1 eλx }.
For λ ̸= 0, KER(D2 + λ2 I) = LS{cos λx, sin λx}
Suppose α ± iβ are the roots of a2 λ2 + a1 λ + a0 = 0. then
KER(a2 D2 + a1 D + a0 I) = LS{eαx cos βx, eαx sin βx}.
Let L = p(D), where p(λ) = ((λ − α)2 + β 2 )k , k > 1, a, b ̸= 0.
Find KER(L).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 8 / 28

Solutions: Linear ODE with constant coefficients

Theorem: Suppose p, q, r ∈ R[λ] with positive degrees,

L p = qr and
GCD(q, r) = 1. Then KER (p(D)) = KER (q(D)) KER (r(D)).

Proof. Let u ∈ KER(q(D)), v ∈ KER(r(D)). Then,

p(D)(u + v) = r(D)q(D)u + q(D)r(D)v = 0,

that is, u + v ∈ KER(p(D)).

Next, since GCD(q, r) = 1, ∃s, t ∈ R[λ] such that sq + tr = 1. Thus,
u ∈ KER(q(D)) ∩ KER(r(D)) ⇒ u = (s(D)q(D) + t(D)r(D))u = 0.

Corollary: Let p, q, r be as above. Let S and T be sets of fundamental solutions

of q(D)y = 0 and r(D)y = 0, respectively. Then S ∪ T is a set of fundamental
solutions of p(D)y = 0.

Ex. Consider the equation Ly = p(D)y = 0, where

p(λ) = (λ4 − 4λ3 + 6λ2 − 8λ + 8I) = (λ − 2)2 (λ2 + 9) = 0.

The solutions of the equation are KER(L) = LS({e2x , xe2x } ∪ {cos 3x, sin 3x}).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 9 / 28

Summing up: Homogeneous Linear ODE with constant coefficients

To solve p(D)y = 0 (For p ∈ R[λ], first solve the auxiliary equation p(λ) = 0).
For each simple real root α, the corresponding fundamental solution is eαx .
If α is a real root with multiplicity k, then the fundamental solutions
corresponding to α are eαx , xeαx , ..., xk−1 eαx .
For each pair of simple complex roots α ± iβ, the corresponding fundamental
solutions are eαx cos βx, eαx sin βx.
If α ± iβ is a pair of complex roots each with multiplicity k, then the
corresponding fundamental solutions are

eαx cos βx, xeαx cos βx, ..., xk−1 eαx cos βx,

eαx sin βx, xeαx sin βx, ..., xk−1 eαx sin βx.
The span of all these fundamental solutions is the general solution.
Solve p(D)y = 0 on I = R, where
p(λ) = (λ − 3)(λ + 2)(λ − 3)3 (λ2 + 6)(λ2 − 4λ + 13)2 .

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 10 / 28

Particular solution for Non-homogeneous Linear ODE

Let Ly = b be of orderPn. Let KER(L) = LS{u1 , ..., un }. Then Ly = 0 has

general solution uc = n i=1 ci ui , ci ∈ R (or C). Here uc is called the
Complementary Function of Ly = b.
Let L = p(D), p ∈ R[λ]. Then uc is known if the roots of p are known. On the
contrary, finding a particular integral/solution up of Ly = 0 is hard, in general.
Luckily, there is a simple way to find up with the help of uc when b is one of the
following:
a polynomial in x,
an exponential function eαx ,
a trigonometric function sin(αx) or cos(αx),
a finite sum/difference of products (including constant multiples) of the above
functions.

In that case, b will have an antiderivative of one of the above types. It is easy to
find a q(λ) ∈ R[λ] such that q(D)b = 0.
Ex. Consider b(x) = x2 cos 3x + e−2x . Produce a q such that q(D)b = 0.
Easy. Take q(λ) = (λ2 + 9)3 (λ + 2).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 11 / 28

Method of Undetermined Coefficients

Consider the equation Ly ≡ p(D)y = b. Suppose q ∈ R[λ] be such that

q(D)b = 0. Then, KER(p(D)) ⪯ KER(q(D)p(D)).
Let B be a basis for KER(p(D)). Extend B to a basis B1 of KER(q(D)p(D))
Let u be any solution of Ly = b. (Existence ensured by EUT).
Now, p(D)u = b ⇒ (q(D)p(D))u = 0 ⇒ u ∈ KER(q(D)p(D))
X
⇒u= ci ui for some ci ∈ R
ui ∈B1
X X
= c i ui + cj uj = u1 + u2 , say.
ui ∈B1 /B ui ∈B
P
Now, p(D)u1 = p(D)(u − u2 ) = b, i.e., u1 = ui ∈B1 /B ci ui is a particular
solution of Ly = b.
P
So, to get a particular solution u1 , put u1 = ui ∈B1 /B ci ui , and determine ci
comparing the coefficients of both sides in p(D)u1 = b.

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 12 / 28

Be aware of your Dummy Friends who are preparing from Class Notes

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 13 / 28

But Remember

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 14 / 28

Example: General Solution = Complementary Function + Particular
Integral

Solve the equation (D2 + 9)y(x) = cos 3x.

Auxiliary equation: p(λ) = λ2 + 9 = 0 with roots λ = ±3i.
Complementary function: uc (x) = c1 cos 3x + c2 sin 3x, c1 , c2 ∈ R.
Now we find the Particular Integral up . Note that y = uc + up .
We should find a differential operator q(D) such that
q(D)(R.H.S.) = q(D) cos 3x = 0. Take q(λ) = λ2 + 9 = p(λ). The basis for
KER(p(D)q(D)) is {cos 3x, sin 3x, xcos 3x, xsin 3x}. Put
up = αxcos 3x + βxsin 3x. Solve for α, β so that p(D)up = cos 3x.

Dup = (α + 3βx) cos 3x + (β − 3αx) sin 3x.

D2 up = (6β − 9αx) cos 3x + (−6α − 9βx) sin 3x.
(D2 + 9)up = cos 3x
= (6β − 9αx) cos 3x + (−6α − 9βx) sin 3x + 9x(α cos 3x + β sin 3x)
= −6α sin 3x + 6β cos 3x

⇒ α = 0, β = 16 .
Thus, up = 61 x sin 3x. The general solution of the given non-homogeneous ODE
is y = up + uc = 16 x sin 3x + c1 cos 3x + c2 sin 3x, c1 , c2 ∈ R.
Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 15 / 28
Disadvantages of Undetermined Coefficients

Consider Lu = f.
It is only useful for constant-coefficient linear ODEs involved in the operator Lu.
The RHS of Lu = f must be functions of xm , enx , sin(px), cos(qx), or their
combinations (additions/subtractions and multiplications).
Question: Will it also allow divisions?

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 16 / 28

Operator Method
(Finding A Solution by Integration- Particular Integral)

We write D−1 y = y and call D−1 as the inverse of differential operator.

R

Consider (D − α)y = b(x), which R −αxis linear and first order ODE. A particular
integral is given by up = eαx e b(x)dx. We write
(D − α)−1 b(x) = up = eαx e−αx b(x)dx.
R

Let L = p(D) = a0 n
Q
i=1 (D − αi ). A particular solution for Ly = b(x) is

−1
L−1 b(x) = a0 n b(x) = a−1
Q Qn −1


i=1 (D − αi ) 0 i=1 (D − αi ) b(x).

Let b(x) = emx . For α ̸= m,

(D − α)−1 emx = eαx e(m−α)x dx = eαx m−α
1
e(m−α)x = m−α
1
emx .
R

If m = α, then (D − α)−1 emx = eαx e−αx emx dx = xemx .

R

Thus we have the following:

If p(m) ̸= 0, then p(D)−1 emx = 1

p(m)
emx .

If p(λ) = (λ − m)k q(λ) with q(m) ̸= 0, then

1
p(D)−1 emx = xk emx .
k! q(m)
Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 17 / 28
Operator Method continued

Let b = xm . We have
Z
(D − α)−1 xm = eαx xm e−αx dx
Z
1 1
= eαx (− xm e−αx + (Dxm )e−αx dx)
α α
xm Dxm D2 xm Dm xm
=− −( 2 )−( 3
) − ... − ( m+1 )
α α α α
1 D D2 Dn m
= − (1 + + 2 + ... + n + ...)x
α α α α

Q. Find (D2 + D + 1)−1 x3 . (Hint: (1 + (D + D2 ))−1 x3 , and think about series

expansion of (1 + x)−1 ).

To use the above procedure for finding p(D)−1 xm , express 1

p(λ)
as sum of partial
fraction.

Q. Find algorithms (as given in above) for finding p(D)−1 b, where (a) cos mx or
sin mx, (b) xk emx . Note that m may or may not be a root of p(λ).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 18 / 28

Homogeneous Equations with Variable coefficients: Reduction of Order

Consider x2 y ′′ − 3xy ′ + 4y = 0, x ∈ (0, ∞). Note that u1 (x) = x2 is a solution.

Suppose that the other fundamental solution is u2 . We put w = uu21 , i.e.,
u2 = u1 w. Then

u2 ′ = u1 ′ w + u1 w ′ ,
u2 ′′ = u1 ′′ w + 2u1 ′ w + u1 w′′ .

Therefore we have,

0 = x2 u2 ′′ − 3xu2 ′ + 4u2
= (x2 u1 − 3xu1 ′ + 4u1 )w + (2x2 u1 ′ − 3xu1 )w′ + (x2 u1 )w′′
= x3 w′ + x4 w′′ .

⇒ (w′ )′ + x1 w′ = 0 (a linear equation of order 1 in w′ )

R 1
⇒ w′ = e− x dx = x1 ⇒ w = ln(x) ⇒ u2 = u1 = x2 ln(x) for x ∈ (0, ∞).
Consider any homogeneous equation Ly = 0 with variable coefficients. In
general, finding solutions to any problem is difficult, as factoring L is not easy.
Suppose that one solution is known. Using reduction of order, the problem can
be reduced to solving another linear equation of lower order.

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 19 / 28

Reduction of Order

Consider Ly(x) = 0, where L = an Dn + an−1 Dn−1 ... + a1 D + a0 I, where

ai ∈ C(I, R), an ̸= 0. Suppose u1 is a known (non-zero) solution. Put u2 = wu1 .
Then
Du2 = wDu1 + (Dw)u1 ,
D2 u2 = wD2 u1 + 2(Dw)(Du1 ) + (D2 w)u1 ,
.
.
! ! !
n n n
Dn u2 = wDn u1 + (Dw)(Dn−1 u1 ) + (D2 w)Dn−2 u1 ... + (Dn w)u1 .
1 2 n
Substituting these values, we get
Lu2 = wLu1 + b1 (Dw) + b2 (D2 w) + bn (Dn w), where

+ ...n−1
n
b1 = a1 u1 + 2a2 (Du1 ) + ... + 1 an (D u1 ), and in general
bk = kk ak u1 + k+1 n
an (Dn−k u1 ).




k
a k+1 (Du 1 ) + ... + k
Since Lu1 = 0. Thus, Lu2 = 0 iff Dw is a solution of
L1 y = (b1 I + b2 D + ... + bn Dn−1 )y = 0. (3)
R
Suppose v is a nonzero solution (3). Then u2 (x) = u1 (x) v(x)dx is a non-zero
solution of Ly = 0. Clearly, u1 , u2 are linearly independent.
Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 20 / 28
Particular Integral by Variation of Parameters

Consider a second-order linear (possibly, non-constant coefficient) ODE

Ly ≡ [D2 + aD + bI]y = g, a, b, g ∈ C(I, R). (4)

Let Ly = 0 have two fundamental solutions u1 , u2 . So, y = c1 u1 + c2 u2 is its

general solution. We seek for a particular solution yp of (4) of the form
yp = w1 u1 + w2 u2 , where w1 , w2 are some functions in C(I, R). Then we should
have Lyp = g.
Now, Dyp = 2i=1 [(Dwi )ui + wi (Dui )] = 2i=1 wi (Dui ), assuming that
P P
(Dw1 )u1 + (Dw2 )u2 = 0.
Then D2 yp = 2i=1 [(Dwi )(Dui ) + wi (D2 ui )].
P

Therefore, Lyp = 2i=1 {[(Dwi )(Dui ) + wi (D2 ui )] + awi (Dui ) + bwi ui }

P

i.e., g = Lyp = (Dw1 )(Du1 ) + (Dw2 )(Du2 ), as u1 , u2 satisfy the homogenious

problem. Thus, w1 , w2 are to be chosen so that

u1 (Dw1 ) + u2 (Dw2 ) = 0,
(Du1 )(Dw1 ) + (Du2 )(Dw2 ) = g.

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 21 / 28

Particular solution by Variation of Parameters: Example

Since W := W (u1 , u2 )(x) ̸= 0 ∀x ∈ I, the above system has unique solution for
Dw1 and Dw2 with Dw1 = −u2 g/W and Dw2 = u1 g/W . Integrating these
functions, we get w1 and w2 .
What is the advantage of Method of ‘Variation of Parameter’ method over
‘Undetermined Coefficients’ ? Is it applicable for g = tan(x) or ln(x)?

Ex. Consider y ′′ − 4y ′ + 4y = (x + 1)e2x . Then u1 = e2x , u2 = xe2x are

fundamental solutions. Here W := W (u1 , u2 ) = e4x .
−xe2x (x+1)e2x e2x (x+1)e2x
Therefore, Dw1 = e4x
= −(x2 + x) and Dw2 = e4x
= x + 1.
3
x2 x2
We get w1 = − x3 − 2
and w2 = 2
+ x.
3 2 2 3
x2
Hence yp = (− x3 − x
2
)e2x + ( x2 + x)xe2x = ( x6 + 2
)e2x .

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 22 / 28

Example of Variation of Parameter

Solve the ODE using the method of variation of parameters, y ′′ − 2y ′ = ex sin x.

Answers: Given differential equation is

y ′′ − 2y ′ = ex sin x. (5)

The complementary function of (5) is C1 + C2 e2x where C1 and C2 are arbitrary

constants. Let u = 1 and v = e2x , g = ex sin x. The Wronskian of u and v is
2e2x (̸= 0).

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 23 / 28

Example Continued

The particular integral of (5) is uf (x) + vg(x), where

Z 2x x
e e sin x
f (x) = − dx
2e2x
x
1 e
= (sin x − cos x)
2 12 + 1 2
1
= − ex (sin x − cos x)
4
and
ex sin x
Z
g(x) = dx
2e2x
1 e−x
= (sin x − cos x)
2 (−1)2 + 12
e−x
= (sin x + cos x)
4
Hence the required solution of (5) becomes
1 x
y = C1 + C2 e2x − e sin x
2
Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 24 / 28
Book References

Book Source: "A Modern Introduction to Differential Equations" 3rd edition, by

Henry J Ricardo, Academic Press Inc., 2020

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 25 / 28

Variation of Parameters for nth order Linear ODEs

The above method is known as variation of parameters and can be applied for any
order.
Suppose Ly = g be of order n ⩾ 1. Let {u1 , ..., un } be a set of fundamental
solutions. Put yp = u1 w1 + ... + un wn . If wi are chosen so that

u1 (Dw1 ) + ... + un (Dwn ) = 0, (1)

(Du1 )(Dw1 ) + ... + (Dun )(Dwn ) = 0, (2)
........................
(Dn−2 u1 )(Dw1 ) + ... + (Dn−2 un )(Dwn ) = 0. (n − 1)
Then, the requirement Lyp = g implies
(Dn−1 u1 )(Dw1 ) + ... + (Dn−1 un )(Dwn ) = g. (n)

Since W (u1 , ..., un ) ̸= 0 ∀ x ∈ I, the system (1) − (n) has a unique solution vi for
Dwi , 1 ⩽ i ⩽ n. Choosing anti-derivatives vi for wi , a particular solution of
Ly = g is obtained.

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 26 / 28

Cauchy-Euler Equation

A linear ODE of the following form

Ly ≡ an xn Dn y + ... + a1 xDy + a0 y = g(x), (6)

where a0 , · · · , an ∈ R, g ∈ C((0, ∞), R), is called a Cauchy-Euler equation.

Idea: To convert this ODE operator Ly to a constant coefficient problem.

Put x = et , x ∈ (0, ∞). Denote dt d
by Dt . Equation (6) transforms to
bn Dt y + ... + b1 Dt y + b0 y = g(et ), b0 , ..., bn ∈ R.
n

[Hint: x = et , Dy = Dt y/Dt x = 1
D y, D2 y
x t
= 1
x2
Dt2 y − 1
x2
Dt y, ...]

Ex. Solve x2 y ′′ − 2xy ′ + 2y = x3 .

For x = et , Dy = 1
x2
Dt y, D2 y = 1
x2
Dt2 y − 1
x2
Dt y. The equation transforms to

(Dt2 − 3Dt + 2)y = e3t .

and its general solution is y(t) = c1 e2t + c2 et + 12 e3t , c1 , c2 ∈ R. So, the general
solution of the given equation on (0, ∞) is
1 3
y(x) = c1 x2 + c2 x + x , c1 , c2 ∈ R
2
Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 27 / 28
Are you still taking screenshots of Class Notes?

Dr. Pratibhamoy Das (IITP) MA1201- MATHEMATICS-II January 20, 2025 28 / 28