O.

Linear Differential Operators

1. Linear differential equations. The general linear ODE of order n is

(1) y (n) + p1 (x)y (n1) + . . . + pn (x)y = q(x).

If q(x) 6= 0, the equation is inhomogeneous. We then call

(2) y (n) + p1 (x)y (n1) + . . . + pn (x)y = 0.

the associated homogeneous equation or the reduced equation.

The theory of the n-th order linear ODE runs parallel to that of the second order equation.
In particular, the general solution to the associated homogeneous equation (2) is called the
complementary function or solution, and it has the form

(3) yc = c 1 y1 + . . . + c n yn , ci constants,

where the yi are n solutions to (2) which are linearly independent, meaning that none of
them can be expressed as a linear combination of the others, i.e., by a relation of the form
(the left side could also be any of the other yi ):
yn = a1 y1 + . . . + an1 yn1 , ai constants.
Once the associated homogeneous equation (2) has been solved by finding n independent
solutions, the solution to the original ODE (1) can be expressed as

(4) y = yp + yc ,

where yp is a particular solution to (1), and yc is as in (3).

2. Linear differential operators with constant coefficients

From now on we will consider only the case where (1) has constant coefficients. This type
of ODE can be written as

(5) y (n) + a1 y (n1) + . . . + an y = q(x) ;

using the differentiation operator D, we can write (5) in the form

(6) (Dn + a1 Dn1 + . . . + an ) y = q(x)

or more simply,
p(D) y = q(x) ,
where
(7) p(D) = Dn + a1 Dn1 + . . . + an . y

We call p(D) a polynomial differential operator with constant coefficients. We p(D)

think of the formal polynomial p(D) as operating on a function y(x), converting
it into another function; it is like a black box, in which the function y(x) goes in,
and p(D)y (i.e., the left side of (5)) comes out. p(D)y
1
2 18.03 NOTES

Our main goal in this section of the Notes is to develop methods for finding particular
solutions to the ODE (5) when q(x) has a special form: an exponential, sine or cosine, xk , or
a product of these. (The function q(x) can also be a sum of such special functions.) These
are the most important functions for the standard applications.
The reason for introducing the polynomial operator p(D) is that this allows us to use
polynomial algebra to help find the particular solutions. The rest of this chapter of the
Notes will illustrate this. Throughout, we let

(7) p(D) = Dn + a1 Dn1 + . . . + an , ai constants.

3. Operator rules.
Our work with these differential operators will be based on several rules they satisfy.
In stating these rules, we will always assume that the functions involved are sufficiently
differentiable, so that the operators can be applied to them.
Sum rule. If p(D) and q(D) are polynomial operators, then for any (sufficiently differ-
entiable) function u,

(8) [p(D) + q(D)]u = p(D)u + q(D)u .

Linearity rule. If u1 and u2 are functions, and ci constants,

(9) p(D) (c1 u1 + c2 u2 ) = c1 p(D) u1 + c2 p(D) u2 .

The linearity rule is a familiar property of the operator a Dk ; it extends to sums of these
operators, using the sum rule above, thus it is true for operators which are polynomials in
D. (It is still true if the coefficients ai in (7) are not constant, but functions of x.)
Multiplication rule. If p(D) = g(D) h(D), as polynomials in D, then
u


(10) p(D) u = g(D) h(D) u . h(D)

The picture illustrates the meaning of the right side of (10). The property h(D)u
is true when h(D) is the simple operator a Dk , essentially because
Dm (a Dk u) = a Dm+k u; g(D)

it extends to general polynomial operators h(D) by linearity. Note that a must be

a constant; its false otherwise. p(D)u
An important corollary of the multiplication property is that polynomial operators with
constant coefficients commute; i.e., for every function u(x),



(11) g(D) h(D) u = h(D) g(D) u .

For as polynomials, g(D)h(D) = h(D)g(D) = p(D), say; therefore by the multiplication

rule, both sides of (11) are equal to p(D) u, and therefore equal to each other.
The remaining two rules are of a different type, and more concrete: they tell us how
polynomial operators behave when applied to exponential functions and products involving
exponential functions.
Substitution rule.
 O. LINEAR DIFFERENTIAL OPERATORS 3

(12) p(D)eax = p(a)eax

Proof. We have, by repeated differentiation,

Deax = aeax , D2 eax = a2 eax , . . . , Dk eax = ak eax ;

therefore,
(Dn + c1 Dn1 + . . . + cn ) eax = (an + c1 an1 + . . . + cn ) eax ,
which is the substitution rule (12). 
The exponential-shift rule This handles expressions such as xk eax and xk sin ax.
(13) p(D) eax u = eax p(D + a) u .
Proof. We prove it in successive stages. First, it is true when p(D) = D, since by the
product rule for differentiation,
(14) Deax u(x) = eax Du(x) + aeax u(x) = eax (D + a)u(x) .
To show the rule is true for Dk , we apply (14) to D repeatedly:
D2 eax u = D(Deax u) = D(eax (D + a)u) by (14);
ax


= e (D + a) (D + a)u , by (14);
= eax (D + a)2 u , by (10).
In the same way,
D3 eax u = D(D2 eax u) = D(eax (D + a)2 u) by the above;


= eax (D + a) (D + a)2 u , by (14);
ax 3
= e (D + a) u , by (10),
and so on. This shows that (13) is true for an operator of the form Dk . To show it is true
for a general operator
p(D) = Dn + a1 Dn1 + . . . + an ,
we write (13) for each Dk (eax u), multiply both sides by the coefficient ak , and add up
the resulting equations for the different values of k. 

Remark on complex numbers. By Notes C. (20), the formula

(*) D (c eax ) = c a eax
remains true even when c and a are complex numbers; therefore the rules and arguments
above remain valid even when the exponents and coefficients are complex. We illustrate.

Example 1. Find D3 ex sin x .

Solution using the exponential-shift rule. Using (13) and the binomial theorem,
D3 ex sin x = ex (D 1)3 sin x = ex (D3 3D2 + 3D 1) sin x
= ex (2 cos x + 2 sin x),
since D sin x = sin x, and D3 sin x = cos x.
2

Solution using the substitution rule. Write ex sin x = Im e(1+i)x . We have

D3 e(1+i)x = (1 + i)3 e(1+i)x , by (12) and (*);
x
= (2 + 2i) e (cos x + i sin x),
by the binomial theorem and Eulers formula. To get the answer we take the imaginary
part: ex (2 cos x + 2 sin x).
4 18.03 NOTES

4. Finding particular solutions to inhomogeneous equations.

We begin by using the previous operator rules to find particular solutions to inhomoge-
neous polynomial ODEs with constant coefficients, where the right hand side is a real or
complex exponential; this includes also the case where it is a sine or cosine function.

Exponential-input Theorem. Let p(D) be a polynomial operator with onstant coefficients,

and p(s) its s-th derivative. Then
(15) p(D)y = eax , where a is real or complex
has the particular solution
eax
(16) yp = , if p(a) 6= 0;
p(a)

xs eax
(17) yp = , if a is an s-fold zero 1 of p.
p(s) (a)

Note that (16) is just the special case of (17) when s = 0. Before proving the theorem,
we give two examples; the first illustrates again the usefulness of complex exponentials.

Example 2. Find a particular solution to (D2 D + 1) y = e2x cos x .

Solution. We write e2x cos x = Re (e(2+i)x ) , so the corresponding complex equation is

(D2 D + 1) ye = e(2+i)x ,

and our desired yp will then be Re(e

yp ). Using (16), we calculate
p(2 + i) = (2 + i)2 (2 + i) + 1 = 2 + 3i , from which
1
yep = e(2+i)x , by (16);
2 + 3i
2 3i 2x
= e (cos x + i sin x) ; thus
13
2 2x 3 2x
Re(eyp ) = e cos x + e sin x , our desired particular solution.
13 13

Example 3. Find a particular solution to y + 4y + 4y = e2t .

Solution. Here p(D) = D2 + 4D + 4 = (D + 2)2 , which has 2 as a double root; using
(17), we have p (2) = 2, so that
t2 e2t
yp = .
2

Proof of the Exponential-input Theorem.

That (16) is a particular solution to (15) follows immediately by using the linearity rule
(9) and the substitution rule (12):
eax 1 p(a)eax
p(D)yp = p(D) = p(D)eax = = eax .
p(a) p(a) p(a)
1 John Lewis communicated this useful formula.
 O. LINEAR DIFFERENTIAL OPERATORS 5

For the more general case (17), we begin by noting that to say the polynomial p(D) has the
number a as an s-fold zero is the same as saying p(D) has a factorization
(18) p(D) = q(D)(D a)s , q(a) 6= 0.
We will first prove that (18) implies
(19) p(s) (a) = q(a) s! .
To prove this, let k be the degree of q(D), and write it in powers of (D a):
(20) q(D) = q(a) + c1 (D a) + . . . + ck (D a)k ; then
s s+1
p(D) = q(a)(D a) + c1 (D a) + . . . + ck (D a)s+k ;
p(s) (D) = q(a) s! + positive powers of D a;
substituting a for D on both sides proves (19). 
Using (19), we can now prove (17) easily using the exponential-shift rule (13). We have
eax xs eax
p(D) (s)
= (s) p(D + a)xs , by linearity and (13);
p (a) p (a)
eax
= (s) q(D + a) Ds xs , by (18);
p (a)
eax
= q(D + a) s!, by (19);
q(a)s!
eax
= q(a) s! = eax ,
q(a)s!
where the last line follows from (20), since s! is a constant:

q(D + a)s! = (q(a) + c1 D + . . . + ck Dk ) s! = q(a)s! .

Polynomial Input: The Method of Undetermined Coefficients.

Let r(x) be a polynomial of degree k; we assume the ODE p(D)y = q(x) has as input
(21) q(x) = r(x), p(0) 6= 0; or more generally, q(x) = eax r(x), p(a) 6= 0.
(Here a can be complex; when a = 0 in (21), we get the pure polynomial case on the left.)
The method is to assume a particular solution of the form yp = eax h(x), where h(x) is
a polynomial of degree k with unknown (undetermined) coefficients, and then to find the
coefficients by substituting yp into the ODE. Its important to do the work systematically;
follow the format given in the following example, and in the solutions to the exercises.
Example 5. Find a particular solution yp to y + 3y + 4y = 4x2 2x.
Solution. Our trial solution is yp = Ax2 + Bx + C; we format the work as follows. The
lines show the successive derivatives; multiply each line by the factor given in the ODE, and
add the equations, collecting like powers of x as you go. The fourth line shows the result;
the sum on the left takes into account that yp is supposed to be a particular solution to the
given ODE.
4 yp = Ax2 + Bx + C
3 yp = 2Ax + B
yp = 2A
2 2
4x 2x = (4A)x + (4B + 6A)x + (4C + 3B + 2A).
6 18.03 NOTES

Equating like powers of x in the last line gives the three equations

4A = 4, 4B + 6A = 2, 4C + 3B + 2A = 0;

solving them in order gives A = 1, B = 2, C = 1, so that yp = x2 2x + 1.

Example 6. Find a particular solution yp to y + y 4y = ex (1 8x2 ).

Solution. Here the trial solution is yp = ex up , where up = Ax2 + Bx + C.
The polynomial operator in the ODE is p(D) = D2 + D 4; note that p(1) 6= 0,
so our choice of trial solution is justified. Substituting yp into the ODE and using the
exponential-shift rule enables us to get rid of the ex factor:

p(D)yp = p(D)ex up = ex p(D 1)up = ex (1 8x2 ),

so that after canceling the ex on both sides, we get the ODE satisfied by up :

(22) p(D 1)up = 1 8x2 ; or (D2 D 4)up = 1 8x2 ,

since p(D 1) = (D 1)2 + (D 1) 4 = D2 D 4.

From this point on, finding up as a polynomial solution to the ODE on the right of (22)
is done just as in Example 5 using the method of undetermined coefficients; the answer is

up = 2x2 x + 1, so that yp = ex (2x2 x + 1).

In the previous examples, p(a) 6= 0; if p(a) = 0, then the trial solution must be altered
by multiplying each term in it by a suitable power of x. The book gives the details; briefly,
the terms in the trial solution should all be multiplied by the smallest power xr for which
none of the resulting products occur in the complementary solution yc , i.e., are solutions of
the associated homogeneous ODE. Your book gives examples; we wont take this up here.

5. Higher order homogeneous linear ODEs with constant coefficients.

As before, we write the equation in operator form:

(23) (Dn + a1 Dn1 + . . . + an ) y = 0,

and define its characteristic equation or auxiliary equation to be

(24) p(r) = rn + a1 rn1 + . . . + an = 0.

We investigate to see if erx is a solution to (23), for some real or complex r. According
to the substitution rule (12),
p(D) erx = 0 p(r) erx = 0 p(r) = 0 .
Therefore

(25) erx is a solution to (7) r is a root of its characteristic equation (16).

Thus, to the real root ri of (16) corresponds the solution eri x .

 O. LINEAR DIFFERENTIAL OPERATORS 7

Since the coefficients of p(r) = 0 are real, its complex roots occur in pairs which are
conjugate complex numbers. Just as for the second-order equation, to the pair of complex
conjugate roots a ib correspond the complex solution (we use the root a + ib)
e(a+ib)x = eax (cos bx + i sin bx),
whose real and imaginary parts

(26) eax cos bx and eax sin bx

are solutions to the ODE (23).

If there are n distinct roots to the characteristic equation p(r) = 0, (there cannot be more
since it is an equation of degree n), we get according to the above analysis n real solutions
y1 , y2 , . . . , yn to the ODE (23), and they can be shown to be linearly independent. Thus
the the complete solution yh to the ODE can be written down immediately, in the form:
y = c 1 y1 + c 2 y2 + . . . + c n yn .
Suppose now a real root r1 of the characteristic equation (24) is a k-fold root, i.e., the
characteristic polynomial p(r) can be factored as

(27) p(r) = (r r1 )k g(r), where g(r1 ) 6= 0 .

We shall prove in the theorem below that corresponding to this k-fold root there are k
linearly independent solutions to the ODE (23), namely:

(28) er1 x , xer1 x , x2 er1 x , . . . , xk1 er1 x .

(Note the analogy with the second order case that you have studied already.)

Theorem. If a is a k-fold root of the characteristic equation p(r) = 0 , then the k

functions in (28) are solutions to the differential equation p(D) y = 0 .
Proof. According to our hypothesis about the characteristic equation, p(r) has (ra)k
as a factor; denoting by g(x) the other factor, we can write
p(r) = g(r)(r a)k ,
which implies that

(29) p(D) = g(D)(D a)k .

Therefore, for i = 0, 1, . . . , k 1, we have

p(D)xi eax = g(D)(D a)k xi eax

= g(D) (D a)k xi eax , by the multiplication rule,


= g(D) eax Dk xi , by the exponential-shift rule,


= g(D) eax 0 , since Dk xi = 0 if k > i;
= 0,

which shows that all the functions of (20) solve the equation. 

If r1 is real, the solutions (28) give k linearly independent real solutions to the ODE (23).
8 18.03 NOTES

In the same way, if a+ib and aib are k-fold conjugate complex roots of the characteristic
equation, then (28) gives k complex solutions, the real and imaginary parts of which give
2k linearly independent solutions to (23):

eax cos bx, eax sin bx, xeax cos bx, xeax sin bx, . . . , xk1 eax cos bx, xk1 eax sin bx .

Example 6. Write the general solution to (D + 1)(D 2)2 (D2 + 2D + 2)2 y = 0.

Solution. The characteristic equation is
p(r) = (r + 1)(r 2)2 (r2 + 2r + 2)2 = 0 .
By the quadratic formula, the roots of r2 + 2r + 2 = 0 are r = 1 i, so we get
y = c1 ex + c2 e2x + c3 xe2x + ex (c4 cos x + c5 sin x + c6 x cos x + c7 x sin x)
as the general solution to the differential equation. 

As you can see, if the linear homogeneous ODE has constant coefficients, then the work
of solving p(D)y = 0 is reduced to finding the roots of the characteristic equation. This is
just a problem in algebra, but a far from trivial one. There are formulas for the roots if
the degree n 4, but of them only the quadratic formula ( n = 2 ) is practical. Beyond
that are various methods for special equations and general techniques for approximating the
roots. Calculation of roots is mostly done by computer algebra programs nowadays.
This being said, you should still be able to do the sort of root-finding described in Notes
C, as illustrated by the next example.

Example 7. Solve: a) y (4) + 8y + 16y = 0 b) y (4) 8y + 16y = 0

Solution. The factorizations of the respective characteristic equations are
(r2 + 4)2 = 0 and (r2 4)2 = (r 2)2 (r + 2)2 = 0 .
Thus the first equation has the double complex root 2i, whereas the second has the
double real roots 2 and 2. This leads to the respective general solutions

y = (c1 + c2 x) cos 2x + (c3 + c4 x) sin 2x and y = (c1 + c2 x)e2x + (c3 + c4 x)e2x .

6. Justification of the method of undetermined coefficients.

As a last example of the use of these operator methods, we use operators to show where
the method of undetermined coefficients comes from. This is the method which assumes the
trial particular solution will be a linear combination of certain functions, and finds what
the correct coefficients are. It only works when the inhomogeneous term in the ODE (23)
(i.e., the term on the right-hand side) is a sum of terms having a special form: each must
be the product of an exponential, sin or cos, and a power of x (some of these factors can be
missing).
Question: Whats so special about these functions?
Answer: They are the sort of functions which appear as solutions to some linear homo-
geneous ODE with constant coefficients.
 O. LINEAR DIFFERENTIAL OPERATORS 9

With this general principle in mind, it will be easiest to understand why the method of
undetermined coefficients works by looking at a typical example.

Example 8. Show that

(30) (D 1)(D 2)y = sin 2x.

has a particular solution of the form
yp = c1 cos 2x + c2 sin 2x .
Solution. Since sin 2x, the right-hand side of (30), is a solution to (D2 + 4)y = 0, i.e.,
(D2 + 4) sin 2x = 0.
we operate on both sides of (30) by the operator D2 + 4 ; using the multiplication rule for
operators with constant coefficients, we get (using yp instead of y)

(31) (D2 + 4)(D 1)(D 2)yp = 0.

This means that yp is one of the solutions to the homogeneous equation (31). But we
know its general solution: yp must be a function of the form

(32) yp = c1 cos 2x + c2 sin 2x + c3 ex + c4 e2x .

Now, in (32) we can drop the last two terms, since they contribute nothing: they are
part of the complementary solution to (30), i.e., the solution to the associated homogeneous
equation. Therefore they need not be included in the particular solution. Put another way,
when the operator (D 1)(D 2) is applied to (32), the last two terms give zero, and
therefore dont help in finding a particular solution to (30).
Our conclusion therefore is that there is a particular solution to (30) of the form
yp = c1 cos 2x + c2 sin 2x .

Here is another example, where one of the inhomogeneous terms is a solution to the
associated homogeneous equation, i.e., is part of the complementary function.

Example 9. Find the form of a particular solution to

(33) (D 1)2 yp = ex .

Solution. Since the right-hand side is a solution to (D 1)y = 0, we just apply the
operator D 1 to both sides of (33), getting
(D 1)3 yp = 0.
Thus yp must be of the form
yp = ex (c1 + c2 x + c3 x2 ).
But the first two terms can be dropped, since they are already part of the complementary
solution to (33); we conclude there must be a particular solution of the form
y p = c 3 x 2 ex .

Exercises: Section 2F
M.I.T. 18.03 Ordinary Differential Equations
18.03 Notes and Exercises
Arthur
c Mattuck and M.I.T. 1988, 1992, 1996, 2003, 2007, 2011