CHAPTER

SECOND
ORDER
LINEAR
EQUATIONS

14 INTRODUCfiON

In the preceding chapters we studied a few restricted types of differential

equations that can be solved in terms of familiar elementary functions.
The methods we developed require considerable skill in the techniques of
integration , and their many interesting applications have a tasty flavor of
practicality. Unfortunately, however, it must be admitted that this part of
the subject tends to be a miscellaneous bag of tricks, and conveys little
insight into the general nature of differential equations and their
solutions. In the present chapter we discuss an important class of
equations with a rich and far-reaching theory. We shall see that this
theory can be given a coherent and satisfying structure based on a few
simple principles.
The general second order linear differential equation is
d2y + P(x) dy + Q(x)y = R(x),
dx 2 dx
or, more simply,
y" + P(x)y ' + Q(x)y = R(x). (1)

81
82 DI FFERENTIAL EQUATIONS

As the notation indicates, it is understood that P(x), Q(x), and R(x) are
functions of x alone (or perhaps constants) . It is clear that no loss of
generality results from taking the coefficient of y" to be 1, since this can
always be accomplished by division . Equations of this kind are of great
significance in physics, especially in connection with vibrations in
mechanics and the theory of electric circuits. In addition-as we shall see
in later chapters-many profound and beautiful ideas in pure mathe­
matics have grown out of the study of these equations.
We should not be misled by the fact that first order linear equations
are easily solved by means of formulas. In general , (1) cannot be solved
explicitly in terms of known elementary functions, or even in terms of
indicated integrations. To find solutions, it is commonly necessary to
resort to infinite processes of one kind or another, usually infinite series.
Many special equations of particular importance in applications, for
instance those of Legendre and Bessel mentioned in Section 1 , have been
studied at great length ; and the theory of a single such equation has often
been found so complicated as to constitute by itself an entire department
of analysis. We shall discuss these matters in Chapters 5 and 8.
In this chapter our detailed consideration of actual methods for
solving (1) will be restricted , for the most part, to the special case in
which the coefficients P(x) and Q(x) are constants. It should also be
emphasized that most of the ideas and procedures we discuss can be
generalized at once to linear equations of higher order, with no change in
the underlying principles but only an increasing complexity in the
surrounding details . By restricting ourselves for the most part to second
order equations , we attain as much simplicity as possible without
distorting the main ideas, and yet we still have enough generality to
include all the linear equations of greatest interest in mathematics and
physics.
Since in general it is not possible to produce an explicit solution of
(1) for inspection , our first order of business is to assure ourselves that
this equation really has a solution. The following existence and unique­
ness theorem is proved in Chapter 13.

Theorem A. Let P(x) , Q(x), and R (x) be continuous functions on a closed

interval [a, b ). 1 If x0 is any point in [a, b ] , and if Yo and y0 are any numbers
whatever, then equation (1) has one and only one solution y (x) on the entire
interval such that y (xo) = Yo and y ' (xo) = Y o ·

1 If a and b are real numbers such that a < b, then the symbol (a, b ] denotes the interval
consisting of all real numbers x that satisfy the inequalities a ,; x ,; b. This interval is called
closed because it contains its endpoints. The open interval resulting from the exclusion of
the endpoints is denoted by (a, b ) and is defined by the inequalities a < x < b.
 SECOND ORDER LINEAR EQUATIONS 83

Thus, under these hypotheses, at any given point x0 in [a,b] we can

arbitrarily prescribe the values of y(x) and y '(x), and there will then exist
precisely one solution of (1) on [a,b] that assumes the prescribed values
at the given point; or, more geometrically, (1) has a unique solution on
[a,b] that passes through a specified point (x0,y0) with a specified slope
y�. In our general discussions through the remainder of this chapter, we
shall always assume (without necessarily saying so explicitly) that the
hypotheses of Theorem A are satisfied .

Example 1. Find the solution of the initial value problem

y" + y = 0, y (O) = 0 and y ' (O) = 1 .
We know that y = sin x, y = cos x, and more generally y =
c 1 sin x + c 2 cos x for any constants c 1 and c 2 , are all solutions of the
differential equation. Also, y = sin x clearly satisfies the initial conditions,
because sin 0 = 0 and cos 0 = 1 . By Theorem A, y = sin x is the only
solution of the given initial value problem , and is therefore completely
characterized as a function by this problem . In just the same way, the
function y = cos x is easily seen to be a solution, and therefore the only
solution : of the corresponding initial value problem
y" + y = 0, y (O) = 1 and y ' (O) = 0.
Since all of trigonometry can be regarded as the development of the
properties of these two functions, it follows that all of trigonometry is
contained by implication (as the acorn contains the oak tree) within the two
initial value problems stated above. We shall examine this remarkable idea
in greater detail in Chapter 4.

We emphasize again that in Theorem A the initial conditions that

determine a unique solution of equation (1) are conditions on the value
of the solution and its first derivative at a single fixed point x0 in the
interval [a,b ]. In contrast to this, the problem of finding a solution of
equation (1) that satisfies conditions of the form y(x0) = y0 and y(x 1) =
y1, where x0 and x1 are different points in the interval , is not covered by
Theorem A. Problems of this kind are called boundary value problems,
and are discussed in Chapter 7.
The term R(x) in equation (1) is isolated from the others and
written on the right because it does not contain the dependent variable y
or any of its derivatives. If R(x) is identically zero , then (1) reduces to
the homogeneous equation
y" + P(x)y ' + Q(x)y = 0. (2)
(This traditional use of the word homogeneous should not be confused
with the equally traditional but totally different use given in Section 7 . ) If
R(x) is not identically zero , then (1) is said to be nonhomogeneous.
84 DIFFERENTIAL EQUATIONS

In studying the nonhomogeneous equation (1) it is necessary to

consider along with it the homogeneous equation (2) obtained from it by
replacing R(x) by 0. Under these circumstances (1) is often called the
complete equation, and (2) the reduced equation associated with it. The
reason for this linkage between (1) and (2) is easy to understand , as
follows .
Suppose that in some way we know that y8(x,c1 ,c 2 ) is the general
solution of (2)-we expect it to contain two arbitrary constants since the
equation is of the second order-and that yp (x) is a fixed particular
solution of (1). If y(x) is any solution whatever of (1), then an easy
calculation shows that y(x) - yp (x) is a solution of (2) :
(y - yp )" + P(x)(y - yp ) ' + Q(x)(y - yp )
= [y" + P(x)y' + Q(x)y] - [y; + P(x)y� + Q(x)yp ]
= R(x) - R(x) = 0. (3)
Since y8(x,c 1 ,c 2 ) is the general solution of (2) , it follows that y(x) -
Yp (x) = y8(x,c l ,c2) or

for a suitable choice of the constants c1 and c2 • This argument proves the
following theorem.

Theorem B. If y8 is the general solution of the reduced equation (2) and yP is

any particular solution of the complete equation (1), then y8 + yP is the
general solution of (1).

We shall see in Section 19 that if y8 is known , then a formal

procedure is available for finding Yp· This shows that the central problem
in the theory of linear equations is that of solving the homogeneous
equation . Accordingly, most of our attention will be devoted to studying
the structure of y8 and investigating various ways of determining its
explicit form-none of which is effective in all cases.
The first thing we should notice about the homogeneous equation
(2) is that the function y(x) which is identically zero-that is , y(x) = 0
for all x-is always a solution. This is called the trivial solution, and is
usually of no interest . The basic structural fact about solutions of (2) is
given in the following theorem.

Theorem C. If y 1 (x) and y2 (x) are any two solutions of (2), then

C t Yt (x) + c 2 y2 (x) (4)

is also a solution for any constants c 1 and c 2 •
 SECOND ORDER LINEAR EQUATIONS 85

Proof. The statement follows immediately from the fact that

(C t Yt + C 2 Y2)" +
P (x)(c t Yt + c 2 y2) ' + Q (x)(c t Yt + c 2 y2)
= (c t )i'; + cdi) + P(x)(ct y ; + c2 y 2) + Q (x)(c t Yt + C 2 Y2)
= c t [ Y� + P(x)y ; + Q (x)y t ] + c2 [ Y� + P(x)y 2 + Q (x)y2]
·
= C t 0 + c 2 0 = 0,
• (5)
where the multipliers of c 1 and c 2 are zero because, by assumption, y 1 and
y2 are solutions of (2) .

For reasons connected with the elementary algebra of vectors , the

solution {4) is commonly called a linear combination of the solutions y1(x)
and y2 {x ) . If we use this terminology, Theorem C can be restated as
follows: any linear combination of two solutions of the homogeneous
equation (2) is also a solution.
Suppose that by some means or other we have managed to find two
solutions of equation (2) . Then this theorem provides us with another
which involves two arbitrary constants, and which therefore may be the
general solution of {2) . There is one difficulty: if either y 1 or y2 is a
constant multiple of the other, say y 1 = ky2 , then
C1 Y1 + Cz Yz = c1kYz + Cz Yz = (e l k + Cz)Yz = cyz ,
and only one essential constant is present. On this basis we have
reasonable grounds for hoping that if neither y 1 nor y2 is a constant
multiple of the other, then
CI Yt (x) + Cz Yz (x)
will be the general solution of (2) . We shall prove this in the next section .
Occasionally the special form of a linear equation enables us to find
simple particular solutions by inspection or by experimenting with power,
exponential , or trigonometric functions.

Example 2. Solve
y" + y ' = 0.
By inspection we see that y1 = 1 and y2 = e-x are solutions. It is
obvious that neither function is a constant multiple of the other, so
(assuming the theorem stated above , but not yet proved) we conclude that

is the general solution.

Example 3. Solve
x 2y" + 2xy ' - 2y = 0.
86 DI FFERENTIAL EQUATIONS

Since differentiating a power pushes down the exponent by one unit,

the form of this equation suggests that we look for possible solutions of the
type y = x". On substituting this in the differential equation and dividing
by the common factor x", we obtain the quadratic equation n (n - 1) +
2n - 2 = 0 or n 2 + n - 2 = 0. This has roots n = 1 , -2, so y1 = x and
y2 = x - 2 are solutions and

is the general solution on any interval not containing the origin.

It is worth remarking at this point that a large part of the theory of

linear equations rests on the fundamental properties stated in Theorems
B and C. An inspection of the calculations (3) and (5) will show at once
that these properties in turn depend on the linearity of differentiation,
that is, on the fact that
[ a:f(x) + pg(x)]' = a:f'(x) + f3g ' (x)
for all constants a: and p and all differentiable functions f(x) and g(x).

PROBLEMS

In the following problems, assume the fact stated above (but not yet proved) , that
if y 1 (x) and y2 (x) are two solutions of (2) and neither is a constant multiple of the
other, then c 1 y 1 (x) + c 2 y2 (x) is the general solution.
1. (a) Verify that y 1 = 1 and y2 = x 2 are solutions of the reduced equation
xy" - y ' = 0, and write down the general solution.
(b) Determine the value of a for which yP = ax 3 is a particular solution of
the complete equation xy" - y ' = 3x 2 • Use this solution and the result of
part (a) to write down the general solution of this equation. (Compare
with Example 1 in Section 1 1 .)
(c) Can you discover y . . y2 , and yP by inspection?
2. Verify that y 1 = 1 and y2 = log x are solutions of the equation xy" + y' = 0,
and write down the general solution. Can you discover y1 and y2 by
inspection?
3. (a) Show that y 1 = e - x and y2 = e 2x are solutions of the reduced equation
y" - y ' - 2y = 0. What is the general solution?
(b) Find a and b so that yP = ax + b is a particular solution of the complete
equation y" - y ' - 2y = 4x. Use this solution and the result of part (a)
to write down the general solution of this equation.
4. Use inspection or experiment to find a particular solution for each of the
following equations:
(a) x 3y" + x 2y ' + xy = 1 ; (c) y" - 2y = sin x.
(b) y" - 2y ' = 6;
5. In each of the following cases, use inspection or experiment to find particular
solutions of the reduced and complete equations and write down the general