8.

Linear Equations with Constant Coefficients

The general second order linear equation with constant coefficients is

ay!! + by! + cy = f (t) (8.1)

where a , b , and c are constants. We will assume throughout this section that a ! 0 .

The function f (t) on the right-hand side of equation (8.1) depends only on t and not on

y because all of the allowed y-dependence is already expressed on the left-hand side of

the equation. Corresponding to any equation of the form (8.1) we define the linear

differential operator

L = aD 2 + bD + c (8.2)

Then equation (8.1) can also be written as

Ly = f (t) (8.3)

We also define the characteristic polynomial corresponding to (8.1) as

ar 2 + br + c = 0 (8.4)

The roots of the characteristic polynomial will be important when we solve (8.1).

THEOREM 1. The operator L = aD 2 + bD + c is linear.

Proof. To prove linearity we must show that for any constants c1, c2 and for any

functions y1, y2 , that L[c1y1 + c2 y2 ] = c1Ly1 + c2 Ly2 . But

L[c1y1 + c2 y2 ] = (aD 2 + bD + c)[c1y1 + c2 y2 ]

= aD 2 [c1y1 + c2 y2 ] + bD[c1y1 + c2 y2 ] + c[c1y1 + c2 y2 ]
= ac1y1!! + ac2 y2!! + bc1y1! + bc2 y2! + cc1y1 + cc2 y2 (8.5)
= c1[ay1!! + by1! + cy1 ] + c2 [ay2!! + by2! + cy2 ]
= c1Ly1 + c2 Ly2

hence the operator is linear. 

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.1
Because equation (8.1) involves a second-order derivative the solution will in general

include two constants of integration rather than the single arbitrary constant that we had

when we were solving first order equations. Determining these constants will require two

constraints upon the solution; we will limit ourselves to the study of initial conditions,

where both the solution y(t) and its derivative y!(t) are specified at the same point t 0 .

The second order linear initial value problem with constant coefficients is then

Ly = f (t),!!y(t 0 ) = y0 ,! y!(t1 ) = y1 (8.6)

Equation (8.6) is sometimes called the Cauchy initial value problem. We refer to the

solution of equation (8.1), which involves two arbitrary constants as the general solution

of the differential equation, to distinguish it from the solution of the initial value problem

(8.6), which has no arbitrary constants.

Instead of constraining the solution and its derivative at the same point t 0 , it is also

possible to specify that value of the solution at two points as in the Dirichlet Problem or

boundary value problem with Dirichlet conditions,

Ly = f (t),!y(t 0 ) = y0 ,!y(t1 ) = y1 , (8.7)

or to specify the value of the derivative of the solution at two points as in the Neumann

Problem, or boundary value problem with Neuman boundary conditions,

Ly = f (t),!! y!(t 0 ) = y0 ,! y!(t1 ) = y1 (8.8)

to constrain the solution and its derivative at two different points, as in the initial-

boundary value problem,

Ly = f (t),!!a11y(t 0 ) + a12 y!(t 0 ) = y0 ,!a21y(t1 ) + a22 y!(t1 ) = y1 . (8.9)

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.2
where a11, a12 , a21, a22 are constants. We will only study the initial value problem

(equation (8.6)) and will not consider any of the boundary value problems, which are

considerably more complicated.

The homogeneous equation corresponding to equation (8.1) is

ay!! + by! + cy = 0 (8.10)

or in terms of the operator notation, Ly = 0 . We use the notation yH (t) to denote the

general solution of (8.10), and yP (t) to denote any particular solution of (8.1) that is not

a solution of (8.10). We will see that there may be many different solutions to the

homogeneous equation (8.10). If every solution of the homogeneous equation can be

written in the form

y = c1 y1 (t) + c2 y2 (t) (8.11)

where y1 (t) and y2 (t) are each by themselves solutions of (8.10), then we call (8.11) the

general solution of the homogeneous equation, and call the pair of functions

{y1 (t),!y2 (t)} and fundamental set of solutions to the homogeneous equation. Then the

general solution of (8.1) is

y(t) = yH (t) + yP (t) (8.12)

We will see that the functions y1 (t) and y2 (t) must not only both be solutions of (8.10),

i.e., Ly1 = Ly2 = 0 , but they must both be linearly independent (in the sense that it is

impossible to find any pair of constants c1 and c2 , both non-zero, such that

c1 y1 (t) + c2 y2 (t) = 0 for all t) for the pair of functions {y1 (t),!y2 (t)} to be considered a

fundamental set, and for (8.12) to be the general solution of Ly = f (t) .

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.3
THEOREM 2. Principle of Superposition. If y1,H (t) and y2,H (t) are both solutions

of Ly = 0 , then any linear combination

yH (t) = c1y1H (t) + c 1y2 H (t) (8.13)

is also a solution of Ly = 0 .

Proof. Since L is linear,

LyH = L[c1y1,H + c2 y2,H ] = c1Ly1,H + c2 Ly2,H = 0 .  (8.14)

Example 1. Two solutions of the homogeneous linear differential equation

y!! " 7 y! + 12y = 0 (8.15)

are y1 (t) = e 4t and y2 (t) = e3t . To set this we let L = D 2 ! 7D + 12 and calculate that

Ly1 (t) = Le 4t = 16e 4t ! 7(4e 4t ) + 12e 4t = 0 (8.16)

Ly2 (t) = Le 3t = 9e 3t ! 7(3e 3t ) + 12e 3t = 0 (8.17)

Hence

L[c1y1 (t) + c2 y2 (t)] = L[c1e4t + c2 e36 ]

!!!!!!!!!!= (16c1e46 + 9c2 e3t ) ! 7(4c1e4t + 3c2 e3t ) + (12c1e4t + 3c2 e3t ) (8.18)
!!!!!!!!!!= 0

THEOREM 3. Subtraction Principle. If y1,P (t) and y2,P (t) are two different particular

solutions of Ly = f (t) then yH (t) = y1,P (t) ! y2,P (t) is a solution of the homogeneous

equation Ly = 0 .

Proof. Since Ly1,P = f (t) and Ly2,P = f (t) , LyH = L(y1,P ! y2,P ) = Ly1,P ! Ly2,P = 0 . 

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.4
THEOREM 4. Existence and Uniqueness. The second order linear initial value

problem

a(t) y!! + b(t) y! + c(t)y = f (t), y(t 0 ) = y0 , y!(t 0 ) = y1 (8.19)

has a unique solution, except possible where a(t) = 0

COROLLARY 1. The second order linear initial value problem with constant

coefficients

Ly = f (t),!y(t 0 ) = y0 ,!y(t1 ) = y1 (8.20)

has a unique solution.

Proof. This follows immediately from theorem 4 with a(t) = a,!b(t) = b,!c(t) = c . 

Proof of theorem 4. For a(t) ! 0 we can write the differential equation as

y!! + p(t) y! + q(t)y = f (t) (8.21)

where p(t) = b(t) / a(t) , q(t) = c(t) / a(t) , and r(t) = f (t) / a(t) . Define

u(t) = y, v(t) = y! (8.22)

Then

v! = y!! = " p(t) y! " q(t)y " r(t) = "q(t)u " p(t)v " r(t) (8.23)

and therefore

d ! u$ ! 0 1 $ ! u$ ! 0 $
= + (8.24)
dt #" v &% #" 'q(t) ' p(t)&% #" v &% #" 'r(t)&%

If we define the vectors y, y 0 , and w, and the matrix A by

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.5
 ! u(t)$ ! y0 $ ! 0 $ ! 0 1 $
y(t) = # & ,!!y 0 = # & , !w(t) = # & , A(t) = # (8.25)
" v(t)% " y1 % " 'r(t)% " 'q(t) ' p(t)&%

Then the initial value problem is equivalent to the first order system

dy
y! = = f(t, y), !y(t 0 ) = y 0 (8.26)
dt

Where f(t, y) = A(t)y + w(t) . We have already proven the existence (e.g., the Peano

existence theorem in section 7) and uniqueness (section 2, theorem 1) of the solution to

(8.26).

COROLLARY 2. Ly = 0 has a solution.

Proof. Pick any numbers ! , " . Then by Theorem 4 the initial value problem Ly = 0 ,

y(t 0 ) = ! , y"(t 0 ) = # has a unique solution. Therefore the differential equation Ly = 0

must have a solution. 

If c = 0 in L then the differential equation Ly = f (t) simplifies to

ay!! + by! = f (t) (8.27)

In this case it is possible to solve the equation by making the change of variables z = y! ,

which reduces (8.27) to a first order linear equation in z.

Example 2. Find the general solution of the homogeneous linear equation y!! + 6 y! = 0 .

Making the substitution z = y! yields z! + 6z = 0 ; separating variables ( dz / z = !6 ) and

integrating gives y! = z = Ce"6t . Separating variables and integrating a second time gives

1
y = " Ce!6t dt = ! Ce!6t + C2 = C1e!6t + C2 (8.28)
6

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.6
where C1 = !C / 6 . 

Example 3. Find the general solution of y!! + 6 y! = t .

Making the substitution z = y! we obtain the first order linear equation

z! + 6z = t (8.29)

An integrating factor is µ = e6t , so that

d
! ze6t # = e6t ( z% + 6t ) = e6t t (8.30)
dt " $

and therefore

z = e!6t ( " te 6t
) #1
dt + C1 = e!6t % te6t !
$6
1 6t
36
& 1
e + C1 ( = t !
' 6
1
36
+ C1e!6t (8.31)

Substituting back z = y! we get a separable first order differential equation,

1 1
y! = t" + C1e"6t (8.32)
6 36

Integrating gives

1 2 1 1
y= t ! t ! C1e!6t + C2 (8.33)
12 36 6

It is customary to combine the (!1 / 6)C1 into a single unknown constant, which we again

name C1 ,

1 2 1
y= t ! t + C1e!6t + C2 .  (8.34)
12 36

Example 4. Separate the solution of y!! + 6 y! = t found in the previous example into its

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.7
homogeneous and particular parts.

Comparing equations (8.34) and (8.28) we see that y(t) = yP (t) + yH (t) , where

yH (t) = C1e!6t + C2 (8.35)

and

1 2 1
yP (t) = t ! t (8.36)
12 36

Two interesting observations can be made from this. First, both arbitrary constants are

part of the homogeneous solution, and not the particular solution. Second, the

homogeneous solution has the form

yH = c1er1t + c2 er2 t (8.37)

where r1 and r2 are the roots of the characteristic equation r 2 + 6r = 0 . We will see that

both of these properties hold more generally for linear equations with constant

coefficients. 

When c ! 0 the simple substitution z = y! is not sufficient to simplify the differential

equation to a first order equation. However, the following factorization property of the

linear operator L suggests an alternate substitution that is not very much more difficult.

THEOREM 5. The linear differential operator L can be factored as

L = aD 2 + bD + c = a(D ! r1 )(D ! r2 ) (8.38)

where r1 and r2 are the roots of the characteristic polynomial ar 2 + br + c = 0 ,

!b ± b 2 ! 4ac
r1,2 = (8.39)
2a

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.8
Proof. The roots of the characteristic polynomial satisfy the following identities,

!b + b 2 ! 4ac !b ! b 2 ! 4ac b
r1 + r2 = + =! (8.40)
2a 2a a

and

" !b + b 2 ! 4ac % " !b ! b 2 ! 4ac % b 2 ! (b 2 ! 4ac) c

r1r2 = $ '$ '= = (8.41)
$# 2a '& $# 2a '& 4a 2 a

hence b = !a(r1 + r2 ) and c = ar1r2 . Thus

Ly = ay!! + by! + cy
= ay!! " a(r1 + r2 ) y! + ar1r2 y
= a[ y!! " (r1 + r2 ) y! + r1r2 y] (8.42)
= a(D " r1 )( y! " r2 y)
= a(D " r1 )(D " r2 )y

Hence L = a(D ! r1 )(D ! r2 ) which is the desired factorization.

Theorem 5 provides us with the following simple algorithm for solving any second order

linear differential equation or initial value problem with constant coefficients.

ALGORITHM 1. To solve {ay!! + by! + cy = f (t),!y(t 0 ) = y0 ,! y!(t 0 ) = y1 }

(1) Find the roots of the characteristic polynomial ar 2 + br + c = 0 .

(2) Factor the differential equation as a(D ! r1 )(D ! r2 )y = f (t) .

(3) Substitute z = (D ! r2 )y = y" ! r2 y to get a(D ! r1 )z = f (t) .

1
(4) Solve the resulting differential equation z! " r1z = f (t) for z(t) .
a

(5) Substitute the initial condition z(t 0 ) = y!(t 0 ) " r2 y(t 0 ) to determine the arbitrary

constant in z(t) .

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.9
 (6) Solve the first order differential equation y! " r2 y = z(t) where z(t) is the solution

you found in steps (4) and (5).

(7) Solve for the remaining arbitrary constant in y(t) using the initial condition

y(t 0 ) = y0

Example 5. Use the factorization of theorem 5 to solve the initial value problem

y!! " 10 y! + 21y = 3sin t,!!y(0) = 1,!! y!(0) = 0 . (8.43)

The characteristic polynomial is r 2 ! 10r + 21 = (r ! 3)(r ! 7) . Hence the differential

equation can be factored as

(D ! 3)(D ! 7)y = 3sin t (8.44)

To solve equation (8.44) we can make the substitution z = (D ! 7)y = y" ! 7y . Then

(D ! 3)z = z" ! 3z = 3sin t (8.45)

This is a first order linear equation with an integrating factor of µ = e!3t .

d
" ze!3t $ = 3e!3t sin t (8.46)
dt # %

Integrating,

e!3t [!3sin t ! cost]

ze!3t = 3" e!3t sin tdt = 3 +C (8.47)
10

The boundary conditions tell us that z(0) = y!(0) " 7y(0) = "7 , so C = !6.7 .

Hence

3
y! " 7y = z = " (3sin t + cost) " 6.7e 3t (8.48)
10

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.10
Substituting back for z = y! " 7y , and therefore

3
y! " 7y = " (3sin t + cost) + C1e3t (8.49)
10

The integrating factor for this equation is µ = e!7t ; multiplying through and integrating,

" 9 3 %
ye!7t = ( $ ! e!7t sin t ! e!7t cost ! 6.7e!4t 'dt
# 10 10 &
(8.50)
9e!7t (!7 sin t ! cost) 3e!7t (!7 cost + sin t) 6.7 !4 x
=! ! + e +C
10(50) 10(50) 4

Simplifying,

60 sin t + 30 cost
y= + 1.675e3t + Ce7t (8.51)
500

From the first initial condition y(0) = 1 and thus C = !0.735 so that

60 sin t + 30 cost
y= + 1.675e 3t ! 0.735e 7t .  (8.52)
500

The following properties of the linear operator will prove useful in developing a more

precise formulation of algorithm 1. Our goal will be to find a general closed form

solution to the second order linear differential equation.

THEOREM 6. Properties of the Linear Differential Operator. L = aD 2 + bD + c . Let

z !! and P(z) = az 2 + bz + c . Then for any function y,

Ly = P(D)y (8.53)

Le zt = P(z)e zt (8.54)

Lye zt = e zt P(D + z)y (8.55)

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.11
Proof. For (8.53),

P(D)y = (aD 2 + bD + c)y = Ly (8.56)

To derive (8.54), we calculate

L(ezt ) = a(ezt )!! + b(ezt )! + c(ezt )

= az 2 ezt + bzezt + cezt
(8.57)
= (az 2 + bz + c)ezt
= P(z)ezt

To derive (8.55) we apply the differential operator to e zt and expand all of the

derivatives:

Lyezt = (aD 2 + bD + c)(yezt )

= a(yezt )!! + b(yezt )! + c(yezt )
(8.58)
= a(yzezt + y!ezt )! + b(yzezt + y!ezt ) + cyezt
= a(yz 2 ezt + 2zy!ezt + y!!ezt ) + b(yzezt + y!ezt ) + cyezt

Factoring out the common exponential and substituting back the D operator,

Lye zt = e zt "# a(yz 2 + 2zy! + y!! ) + b(yz + y! ) + cy $%

(8.59)
= e zt "# a(D 2 + 2zD + z 2 )y + b(D + z)y + cy $%

By the factorization theorem,

Lye zt = e zt !" a(D + z)2 + b(D + z) + c #$ y = e zt P(D + z)y  (8.60)

To solve

ay!! + by! + cy = f (t) (8.61)

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.12
using algorithm 1, we let r1 and r2 be the roots of the characteristic

polynomial ar 2 + br + c = 0 , and then substitute z = (D ! r2 )y = y" ! r2 y to obtain

1
z! " r1z = f (t) (8.62)
a

This is a first order linear differential equation for z; an integrating factor is µ = e!r1t , so

the general solution is

# 1 &
z(t) = er1t %C1 + " e!r1s f (s)ds ( (8.63)
$ a t '

Replacing z(t) with y! " r2 y

y! " r2 y = F(t) (8.64)

where

# 1 &
F(t) = er1t %C1 + " e!r1s f (s)ds ( (8.65)
$ a t '

An integrating factor for (8.64) is µ = e!r2 t and hence

y = er2 t #C2 + " e!r2 u F(u)du % (8.66)

$ t &

The integral in (8.66) is

!r2 u #
(r !r2 )u 1 !r1s &
"t e F(u)du = "t e 1 %$C1 + a "u e f (s)ds (' du
(8.67)
1
= C1 " e(r1 !r2 )u du + " e(r1 !r2 )u " e!r1s f (s)dsdu
t a t u

Hence

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.13
 # 1 &
y = er2 t %C2 + C1 " e(r1 !r2 )u du + " e(r1 !r2 )u " e!r1s f (s)dsdu (
$ t a t u '
er2 t
= er2 t C2 + C1er2 t " e(r1 !r2 )u du + (r !r2 )u
"t e 1 "u e
!r1s
f (s)dsdu (8.68)
t a
= y H + yP

We obtain the homogeneous solution when f (t) = 0 , hence

er2 t (r !r2 )u !r1s

yP =
a "t e 1 "u e f (s)dsdu (8.69)

yH = er2 t C2 + C1er2 t " e(r1 !r2 )u du (8.70)

t

If r1 ! r2 then

C1 r2 t r1 !r2 t
yH = er2 t C2 + e e = c1er1t + c2 er2 t (8.71)
r1 ! r2

where c1 = C1 / (r1 ! r2 ) and c2 = C2 . If r1 = r2 = r then

yH = ert C2 + C1ert ! du = (c1 + c2t)ert (8.72)

t

where c1 = C2 and c2 = C1 .

THEOREM 7. Let r1 and r2 be the roots of ar 2 + br + c = 0 . Then the general solution

of

ay!! + by! + cy = 0 (8.73)

is

"$C1er1t + C2 er2 t , if r1 ! r2
y=# (8.74)
%$(C1t + C2 )e , if r1 = r2 = r
rt

where C1 and C2 are completely arbitrary constants.

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.14
Proof. We have already proven that (8.74) is a solution of ay!! + by! + cy = 0 . To

complete the proof we need to show that every solution has the form (8.74).

Suppose that r1 ! r2 and that u(t) is a solution of (8.73) that is not a linear combination

of the two functions er1t and er2 t . Then u(t) is a solution of some initial value problem

Ly = 0, y(t 0 ) = ! , y"(t 0 ) = # (8.75)

where we define the constants ! = u(t 0 ), " = u #(t 0 ) . We ask the following question: can

we find any linear combination of

y = C1er1t + C2 er2 t (8.76)

that is also a solution of (8.76)? If so then

y(t 0 ) = C1er1t0 + C2 er2 t0 = ! (8.77)

y!(t 0 ) = C1r1er1t0 + C2 r2 er2 t0 = " (8.78)

Multiplying (8.77) by r1 and r2 , respectively,

C1r1er1t0 + C2 r1er2 t0 = ! r1 (8.79)

C1r2 er1t0 + C2 r2 er2 t0 = ! r2 (8.80)

Subtracting equations (8.79) and (8.80) from (8.78),

C2 er2 t0 (r2 ! r1 ) = " ! ar1 (8.81)

C1 (r1 ! r2 )er1t0 = " ! # r2 (8.82)

Hence

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.15
 ! " # r2 " r1t0 ! " # r1 " r2 t 0
C1 = e ,!!C2 = e (8.83)
r1 " r2 r2 " r1

and therefore y = C1er1t + C2 er2 t is a solution of { Ly = 0, y(t0 ) = ! , y"(t0 ) = # } , with C1

and C2 given by (8.83). But we already said that u(t) is also a solution of the same

initial value problem. By the uniqueness theorem they must both be identical. But this

means that our assumption that u(t) was not a linear combination of er1t and er2 t must be

false. Therefore er1t and er2 t must be a fundamental set of solutions when r1 ! r2 . A

similar argument holds when r1 = r2 = r ; the calculation is left as an exercise. 

COROLLARY 1. If the roots of ar 2 + br + c = 0 are a complex conjugate pair

r1 = µ + i! , r2 = µ ! i" , where µ, ! "! (this will occur when b 2 < 4ac ), then

yH = C1er1t + C2 er2 t = e µt [ c1 cos ! t + c2 sin ! t ] (8.84)

where c1 = C1 + C2 and c2 = i(C1 ! C2 ) .

Proof.

yH = C1er1t + C2 er2 t = C1e( µ +i! )t + C2 e( µ "i! )t

= e µt #$C1ei! t + C2 e"i! t %& (8.85)
= e µt [ C1 (cos ! t + i sin ! t) + C2 (cos ! t " i sin ! t)]
= e µt [ (C1 + C2 )cos ! t + i(C1 " C2 )sin ! t ] 

We therefore revise algorithm 1 as follows.

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.16
COMPLEX NUMBERS
A complex number z is an ordered pair of real numbers (u, v) that we write as z = u + iv
where the symbol i has the special property that i 2 = !1 . We denote the field of all
complex numbers by the symbol ! .
If z = u + iv we call u the real part of z, v the imaginary part of z, and write u = Re z ,
v = Im z . We call the number z* = u ! iv the complex conjugate of z.

The absolute value z = u 2 + v 2 is the distance of the point (u, v) from the origin.

The phase ph z = tan !1 (v / u) is the angle between the vector from the origin to (u, v) .
If z = u + iv then in polar coordinates r = z and ! = ph z , and

z = z * z = (u ! iv)(u + iv) = u 2 + v 2
2

DeMoivre’s formula: cos n! + i sin n! = (cos! + i sin ! )n .

Euler’s formula: eiz = cos z + i sin z ; ei! = "1 .
Corollary: If z = u + iv then z = rei! where r = z and ! = ph z .
Complex Roots. If r = u + iv u, v !! is the root of a polynomial with real coefficients,
then so is r* = u ! iv .
The Fundamental Theorem of Algebra: Every polynomial equation of degree n with
complex coefficients has n roots in the complex numbers.
Corollary. Every complex number has n distinct nth roots.
Example: To find the 3 cube roots of 27, we write
27 = 27 + (i)(0) = 27(cos 2n! + i sin 2n! ) = 27ei 2n! , n = 0,1, 2, 3,...
271/ 3 = (27ei 2n! )1/ 3 = (27)1/ 3 (ei 2n! )1/ 3 = 3ei 2n! / 3 = 3[cos(2n! / 3) + i sin(2n! / 3)]
"cos 0 + i sin 0 "1
$ $
= 3 #cos(2! / 3) + i sin(2! / 3) = 3 #(&1 / 2) + i( 3 / 2)
$cos(4! / 3) + i sin(4! / 3) $
% %(&1 / 2) + i(& 3 / 2)
3 3 3 3 3 3
= 3,!& +i ,& & i
2 2 2 2
Historical note. The invention of complex numbers is generally attributed to Cardan (c. 1540). Gauss first
introduced the term “complex number” in 1837. The term “conjugate” was introduced by Cauchy in 1821.
Descarte proposed in 1637 that one could imagine that there were n roots to a polynomial. Albert Girard
(1629) proposed that an nth order polynomial has n roots but that they may exist in a field larger than the
complex numbers. The first published proof of the fundamental theorem of algebra was by D’Alembert in
1746, but his proof was based on an earlier theorem that itself used the theorem. At about the same time
Euler proved it for real coefficients up to 6th. Between 1799 (in his doctoral dissertation) and 1816 Gauss
published three different proofs for the case with real coefficients, and in 1849 proved the general case with
complex coefficients.

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.17
ALGORITHM 2. To solve the initial value problem

ay!! + by! + cy = f (t),!y(t 0 ) = y0 ,! y!(t 0 ) = y1 (8.86)

(1) Find the roots r1 , r2 of the characteristic equation ar 2 + br + c = 0 .

(2) The general solution of the homogeneous equation is

$C1er1t + C2 er2 t ,!!if !r1 ! r2 ,!r1 , r2 "!

&
yH = %(C1 + C2t)ert ,!!!if !r1 = r2 = r "! (8.87)
&eµt (C cos # t + C sin # t),!!r = µ ± i#
' 1 2 1,2

(3) Find y P by operator factorization

a. Factor the differential equation as a(D ! r1 )(D ! r2 )yP = f (t)

b. Substitute z = (D ! r2 )yP = yP" ! r2 yP

c. Solve the resulting first order equation for z.

d. Set the two formulas for z found in previous two steps equal to each other.

e. Solve the resulting formula for y P

(4) As an alternative to step (3),

er2 t (r !r2 )u !r1s

yP =
a "t e 1 "u e f (s)dsdu (8.88)

(5) The general solution is y = y P + y H .

(6) Use the initial conditions to determine the values of the constants C1 and C2 .

Example 6. Find the general solution of y !! + 4 y ! + 3y = 0

The characteristic polynomial is 0 = r2 + 4r + 3 = (r + 3)(r + 1) which has distinct roots

r = !1,!3 . Hence the general solution is y = C1e!3t + C2 e!t .

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.18
Example 7. Solve the initial value problem y !! " 6y ! + 9y = 0, y(0) = 4, y !(0) = 17 .

The characteristic polynomial is 0 = r2 ! 6r + 9 = (r ! 3)2 . Since there is a repeated root

r = 3 , the general solution of the homogeneous equation is

y = Ae3t + Bte3t (8.89)

By the first initial condition, we have 4 = y(0) = A . Substituting A=4 into (8.89) and

differentiating,

y! = (12 + B + 3Bt)e3t (8.90)

From the second initial condition, 17 = y !(0) = 12 + B + 3B(0) = 12 + B , so that B = 5

and the solution of the initial value problem is

y = (4 + 5t)e3t .  (8.91)

Example 8. Find the general solution of y!! + 6 y! + 25y = 0 .

The characteristic polynomial is r 2 + 6r + 25 = 0 ; its roots are

!6 ± !64 !6 ± 8i
r= = = !3 ± 4i (8.92)
2 2

Hence the solution is

y = e!3t [C1 cos 4t + C2 sin 4t] .  (8.93)

Example 9. Solve y!! " 9y = e2t ,!y(0) = 1,! y!(0) = 2

The characteristic equation is r 2 ! 9 = (r ! 3)(r + 3) = 0 , so the roots are r1 = 3, r2 = !3

and the solution of the homogeneous equation is

yH = C1e3t + C2 e!3t (8.94)

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.19
From equation (8.88) with a = 1 and f (t) = e2t , a particular solution is

yP = e!3t " e6u " e!3s e2s dsdu = e!3t " e6u " e!s dsdu
t u t u
(8.95)
#1 & 1
= !e!3t " e6u e!u du = !e!3t % e5t ( = ! e2t
t $5 ' 5

Hence the general solution is

1 2t
y = C1e 3t + C2 e!3t ! e (8.96)
5

The initial conditions give

1 2
1 = C1 + C2 ! ,!!2 = 3C1 ! 3C2 ! (8.97)
5 5

The solution of (8.97) is C1 = 1,!C2 = 1 / 5 , so that

1 !3t 1 2t
y = e 3t + e ! e (8.98)
5 5

is the solution of the initial value problem. 

Because equation (8.88) provides a closed form expression, we can always derive a

particular solution from it; in fact, some forms of f (t) occur so frequently that the

corresponding solutions have been tabulated. For example, a particular solution of

ay!! + by! + by = Ceµt is

$ C "t
& a( µ ! r )( µ ! r ) e , if ! µ # r1 !# r2
& 1 2

& C
yP = % teµt , !!!!!!!!!!if !( µ = r1 !or! µ = r2 )!and!r1 # r2 (8.99)
a(r
& 1 2 ! r )
& K 2 µt
& t e , !!!!!!!!!!!!!if ! µ = r1 = r2
' 2a

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.20
where r1 and r2 are the roots of ar 2 + br + c = 0 . The fact that these solutions are

tabulated justifies the following technique.

ALGORITHM 3. Method of Undetermined Coefficients. To find a particular solution to

ay!! + by! + cy = f (t)

(1) Make an "educated guess" for the form of yP (t) .

(2) Plug the "guess" into the differential equation.

(3) Solve for any unknown constants.

The success of this method hinges on being able to make a good "guess" for the form of

yP . For example, (8.99) suggests that for ay!! + by! + by = Ke" t would be Ct n e! t where

n = 0,1, 2 is the number of times that ! is a root.

Example 10. Find a particular solution for y!! " 5 y! + 6y = 4e3t .

Since r = 2, 3 are roots of the characteristic equation this means that 3 is a single root. So

we try y = Kte 3t . Differentiating twice,

y! = Ke3t + 3Kte 3t
(8.100)
y!! = 6Ke3t + 9Kte 3t

Substituting this into the differential equation

( ) (
4e3t = 6Ke3t + 9Kte3t ! 5 Ke3t + 3Kte3t + 6Kte3t ) (8.101)
= Ke 3t
( 6 + 9t ! 5 ! 15t + 6t ) = Ke 3t

Hence K = 4 , yP = 4te3t , and y = C1e2t + C2 e3t + 4te3t . 

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.21
THEOREM 9. Let r1 and r2 are the roots of ar 2 + br + c = 0 . Then

(1) A particular solution of ay!! + by! + cy = Ct n eµt is

yP (t) = (a0 + a1t + ! + ant n )t k eµt (8.102)

where k is the multiplicity of µ as a root of ar 2 + br + c = 0 and the a0 , a1 ,..., an can

be determined by the method of undetermined coefficients.

(2) A particular solution of ay!! + by! + cy = t n [ Asin " t + B cos " t ] is

yP (t) = t k "#(a0 + a1t + ! + ant n )sin ! t + (b0 + b1t + ! + bnt n )cos ! t $% (8.103)

where

$"0,!±i! !are!not !roots!of !ar + br + c = 0

2

k=# (8.104)
$%1,!±i! !are!roots!of !ar + br + c = 0
2

and the a0 , a1 ,..., an and b0 ,b1 ,...,bn can be determined by the method of undetermined

coefficients.

(3) A particular solution of ay!! + by! + cy = t n eµt [ Asin " t + B cos " t ] is

yP (t) = t k eµt "#(a0 + a1t + ! + ant n )sin ! t + (b0 + b1t + ! + bnt n )cos ! t $% (8.105)

where

"$0,! µ ± i! !are!not !roots!of !ar 2 + br + c = 0

k=# (8.106)
$%1,! µ ± i! !are!roots!of !ar + br + c = 0
2

and the a0 , a1 ,..., an and b0 ,b1 ,...,bn can be determined by the method of undetermined

coefficients.

(4) A particular solution of ay!! + by! + cy = " i =1 fi (t) is

n

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.22
 yP = !
n
y (8.107)
i=1 Pi

where each y Pi is a particular solution ay!! + by! + cy = fi (t) .

Example 11. Find the general solution to y !! + 2 y! + 5y = 3sin2t

The characteristic polynomial is r 2 + 2r + 5 = 0 ; its roots are r = !1 ± 2i , and the

homogeneous solution is

yH = e!t [ C1 sin 2t + C2 cos 2t ] (8.108)

Therefore we try

yP = C cos 2t + D sin 2t (8.109)

Differentiating twice

y! = "2C sin 2t + 2D cos 2t

(8.110)
y!! = "4C cos 2t " 4D sin 2t

Plugging back into the differential equation,

3sin 2t = ( !4C cos 2t ! 4D sin 2t ) + 2 ( !2C sin 2t + 2D cos 2t )

+ 5 ( C cos 2t + D sin 2t )
(8.111)
= ( !4C + 4D + 5C ) cos 2t + ( !4D ! 4C + 5D ) sin 2t
= ( C + 4D ) cos 2t + ( D ! 4C ) sin 2t

or

0 = (C + 4D) cos2t + (D ! 4C ! 3)sin2t (8.112)

This must hold for all t; setting t = 0 and t = ! / 4 , respectively, give

C + 4D = 0,!0 = D ! 4C + 3

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.23
Solving these simultaneously for C and D give C = !12 / 17,!D = 3 / 17 . Hence

12 3
yP = ! cos2t + sin 2t (8.113)
17 17

and the general solution is

12 3
y = yH + yP = e!t [ C1 cos 2t + C2 sin 2t ] ! cos 2t + sin 2t  (8.114)
17 17

Example 11. Find the general solution to y !! " 9y = e 2t + 5e3t

The characteristic equation is r 2 ! 9 = 0 and the roots are r = ±3 so

y = C1e3t + C2 e!3t + yP1 + y P 2 (8.115)

where LyP1 = e2t and LyP 2 = 5e 3t . Since µ = 2 is not a root of the characteristic equation

we try

yP1 = Ce2t ,! yP1

! = 2Ce2t ,! yP!!2 = 4Ce2t (8.116)

in LyP1 = e2t , giving

e2t = 4Ce2t ! 9Ce2t = !5Ce2t (8.117)

so that C = !1 / 5 and hence yP1 = !(1 / 5)e2t . Since µ = 3 is a root of multiplicity 1 we

try

yP 2 = cte3t
yP! 2 = c(e3t + 3te3t ) = ce3t (1 + 3t)! (8.118)
yP!!2 = c(6e3t + 9te3t ) = 3ce3t (2 + 3t)

in LyP 2 = 5e 3t , giving

5e3t = 3ce3t (2 + 3t) ! 9cte3t = 6ce3t (8.119)

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.24
 1 2t
!! " 9y P2 = e 2t , while from
rom example 9, yP2 = ! e is a particular solution to yP2
5

5 3t
example 10, yP3 = !! " 9yP3 = 5e 3t .
te is a particular solution to yP3 A quick
6

calculation verifies that u = y P2 + y P3 is a particular solution of our current differential

equation:

u!! " 9u = y!P2 ! " 9( yP2 + y P3 ) = y!P2

! + y !P3 ! " 9y P2 + y !P3
! " 9y P3
(8.120)
= e 2t + 5e3t

Finally, from example 9, yH = Ae3t + Be !3t is a solution to the homogeneous equation.

Thus the general solution is

1 5
y = y H + y P2 + yP3 = Ae 3t + Be !3t ! e2t + te3t  (8.121)
5 6

Harmonic Oscillations. If a spring is extended from its resting length by an amount y

then a restoring force, opposite in direction from but proportional to the displacement will

attempt to pull the spring back; the subsequent motion of an object of mass m attached to

the end of the spring is described by Newton’s laws of motion:

my!! = "ky (8.122)

where k is a positive constant that is determined by the mechanical properties of the

spring. Equation (8.122) is known as Hooke’s Law. Rearranging,

y!! + " 2 y = 0 (8.123)

where ! = k / m . Equation (8.123) is called the simple harmonic oscillator equation.

The characteristic equation is r 2 + ! 2 = 0 , and since the roots are purely imaginary

( r = ±i! ) the motion is oscillatory:

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.25
 y = C1 cos ! t + C2 sin ! t (8.124)

If we define the amplitude A of the oscillations by A 2 = C12 + C22 then

y = Asin(! t + " ) (8.125)

where ! = tan "1 (C1 / C2 ) is known as the phase of the oscillations. It can be easily shown

that sin ! = C1 / A and cos ! = C2 / A .

In fact, equation (8.122) is not such a good model because it predicts the system will

oscillate indefinitely, and not slowly damp out to zero. A good approximation to the

damping is a force that acts linearly against the motion: the faster the mass moves, the

stronger the damping force. Its direction is negative, since it acts against the velocity y!

of the mass. Thus we modify equation (8.122) to

my!! = "Cy!! " ky (8.126)

where C ! 0 . Defining b = C / m and ! = k / m as before, the equation for the

damped harmonic oscillator becomes

y!! + by! + " 2 y = 0 (8.127)

The roots of the characteristic equation are now

!b ± b 2 ! 4" 2
r= (8.128)
2

The resulting system is said to be underdamped when b < 2! , critically damped when
b = 2! , and overdamped when b > 2! . In the underdamped system ( b < 2! ) the roots
are a complex conjugate pair with negative real part, r = µ ± i! where

µ = b / 2 > 0,!!! = " 1 # (b / 2" )2 (8.129)

and hence the resulting oscillations are described by decaying oscillations

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.26
 y = Ae! µt sin(" t + # ) (8.130)

The critically damped system has a single positive real repeated root r = µ = b / 2 , so

that

y = (C1 + C2t)e! µt (8.131)

The critically damped systems decays directly to zero without crossing the y-axis. The

overdamped systems has two negative real roots ! µ ± " (where ! < µ ) and hence

(
y = e! µt C1e " t + C2 e! " t ) (8.132)

The system damps to zero without oscillations, but may cross the y-axis once. The first

term in parenthesis in (8.132) does not give an exponential increase because ! < µ .

Figure 1. Harmonic oscillations with ! = 1 and initial conditions y(0) = 1 , y!(0) = 0 .

Critical damping: heavy line, b = 2 ; overdamped: long dashes, b = 2.1 ; underdamped:
short dashes, b = 0.25 .

Finally, it is possible to add a motor to the spring that produces a force f (t) . The

resulting is system is called the forced harmonic oscillator:

my!! = "Cy!! " ky + f (t) (8.133)

If we define F(t) = f (t) / m and ! and b as before, this becomes

y!! + by! + " 2 y = F(t) (8.134)

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.27
Thus any second order linear differential equation with positive constant coefficients

describes a harmonic oscillator system. The particular solution is

yP = er2 t " e(r1 ! r2 )u " e! r1 s F(s)ds (8.135)

t u

For example, suppose the system driven by a force function

F(t) = F0 sin ! t (8.136)

Then

yP = F0 er2 t # e(r1 ! r2 )u # e! r1 s sin " s!dsdu

t u

F0
= er2 t # e! r2 u (" cos " u + r1 sin " u)du (8.137)
" + r12
2 t

F0
= 2 2 %
$ !" (r1 + r2 )cos " t + (" 2 ! r1r2 )sin " t &'
(" + r1 )(" + r2 )
2 2

As with the unforced case, we can define the amplitude and phase angle by

Asin ! = "# (r1 + r2 ) and A cos! = " 2 # r1r2 then

F0 Asin(! t + " )
yP = (8.138)
(! 2 + r12 )(! 2 + r22 )

where

A 2 = [!" (r1 + r2 )]2 + [" 2 ! r1r2 ]2 = " 2b 2 + (" 2 ! # 2 )2 (8.139)

because r1 + r2 = !b and r1r2 = ! 2 (see equation (8.128)). Furthermore, the denominator

of (8.138) is the same as A 2 :

(! 2 + r12 )(! 2 + r22 ) = ! 4 + (r12 + r22 )! 2 + (r1r2 )2

= ! 4 + [(r12 + r22 )2 " 2r1r2 ]! 2 + (r1r2 )2
= ! 4 + (b 2 " 2# 2 )! 2 + # 4 (8.140)
= ! 2b 2 + ! 4 " 2# 2! 2 + # 4
= ! 2b 2 + (! 4 " # 2 )2

and therefore

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.28
 F0 sin(! t + " )
yP = (8.141)
! 2b 2 + (! 2 # $ 2 )2

Forcing the oscillator pumps energy into the system; it has a maximum at ! = " , which

is unbounded (infinite) in the absence of damping. This phenomenon – that the

magnitude of the oscillations is maximized when the system is driven at its natural

frequency – is known as resonance. If there is any damping at all the homogeneous

solutions decay to zero and all that remains is the particular solution – so the resulting

system will eventually be strongly dominated by (8.141), oscillating in synch with the

driver. If there is no damping ( b = 0 ) then

F0 sin(# t + $ )
y = C sin(! t + " ) + (8.142)
#2 %!2

where C is the natural magnitude of the system, determined by its initial conditions.

Figure 2. Magnitude of oscillations given by (8.141) for various values of damping

coefficient b = 1, 0.3, 0.1, 0.03, !0.01 (bottom to top) with ! = 1 , as a function of forcing
frequency ! .

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.29
Exercises.
Find a fundamental set of solutions to the following homogeneous equations.
(1) y!! " y! " 2y = 0
(2) y!! + 6 y! + 9y = 0
(3) 5 y!! " 20 y! + 30y = 0
(4) y!! + 5 y! = 0
(5) y!! + 8 y! + 16y = 0
(6) y!! = "36y
Find general solutions to the following differential equations
(7) y!! + 2 y! + y = 2e"t
et
(8) y!! " 2 y! + y =
1+ t2
(9) y!! " 5 y! + 6y = sin 3t

(10) y!! " 6 y! + 8y = cos 4t + e"2t

(11) y!! " 6 y! + 9 = t + e3t
(12) y!! = t 2
(13) y!! " 3y! + 17 = t 2
Solve the initial value problems
(14) y!! + y! " 2y = 0, y(0) = 1, y!(0) = 1
(15) 2 y!! + y! " 4y = 0, y(0) = 0, y!(0) = 1
(16) y!! + 4y = 0 , y(0) = 0, y!(0) = 1
(17) y!! + 4 y! + 3y = 0, y(0) = 2, y!(0) = "1
(18) 4 y!! " y = 0, y("2) = 1, y!("2) = "1
(19) y!! + 4 y! + 5y = 0, y(0) = 1, y!(0) = 0
(20) 2 y!! + 4 y! " 6y = 0, y(0) = 4, y!(0) = 0

(21) y!! + 3y! " 10y = e"5t , y(0) = 1, y!(0) = 2

(22) y!! " 4y = 0, y(0) = 1, y!(0) = 0

(23) y!! + 4y = t 2 + 3et , y(0) = 0, y!(0) = 2

(24) y!! " 2 y! " 3y = 3te2t , y(0) = 1, y!(0) = 0

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.30
(25) For the differential equation t 2 y!! + ty! + y = 0
(a) Find a solution by using the substitution z = ln t ,
(b) Find a by the method of undetermined coefficients with y = t a .
(c) Prove that the solutions are equivalent
(26) Rewrite the simple harmonic oscillator equation y!! + " 2 y = 0 as a system of two
first order equations.
(27) Letting dy / dx = y! / x ! find a closed form expression for the trajectories, in the xy
plane, of the system you found in the previous exercise.

©2005 BE Shapiro [Math 351 Spring 2005 Revised 5 May 2005] Page 8.31