84

2.3 Linear Equations I

An equation y 0 = f (x, y) is called first-order linear or a linear equa-
tion provided it can be rewritten in the special form

(1) y 0 + p(x)y = r(x)

for some functions p(x) and r(x). In most applications, p and r are
assumed to be continuous. The function p(x) is called the coefficient
of y. The function r(x) (r abbreviates right side) is called the non-
homogeneous term or the forcing term. Engineering texts call r(x)
the input and the solution y(x) the output.
A practical test:

An equation y 0 = f (x, y) with f continuously differentiable

is linear provided fy (x, y) is independent of y.

Form (1) is obtained by defining r(x) = f (x, 0) and p(x) = −fy (x, y).
Two examples:

Ly 0 + Ry = E The LR-circuit equation with p(x) = R/L and

r(x) = E/L. Symbols L, R and E are respectively
inductance, resistance and electromotive force.
y 0 = −h(y − y1 ) Newton’s cooling equation with p(x) = h and
r(x) = hy1 . Oven temperature y1 and meat ther-
mometer reading y(t) appear in the roast model.

Classifying Linear Equations

Algebraic complexity may make an equation y 0 = f (x, y) appear to be
non-linear, e.g., y 0 = (sin2 (xy) + cos2 (xy))y simplifies to y 0 = y.
Computer algebra systems classify an equation y 0 = f (x, y) as linear
provided the identity f (x, y) = f (x, 0) + fy (x, 0)y is valid. Equivalently,
f (x, y) = r(x) − p(x)y, where r(x) = f (x, 0) and p(x) = −fy (x, y).
Automatic simplifications in computer algebra systems make this test
practical. Hand verification can use the same method.
Elimination of an equation y 0 = f (x, y) from the class of linear equations
can be done from necessary conditions. The equality fy (x, y) = fy (x, 0)
implies two such conditions:

1. If fy (x, y) depends on y, then y 0 = f (x, y) is not linear.

2. If fyy (x, y) 6= 0, then y 0 = f (x, y) is not linear.

For instance, either condition implies y 0 = 1 + y 2 is not linear.

2.3 Linear Equations I 85

Variation of Parameters and Integrating Factors

The initial value problem

(2) y 0 + p(x)y = r(x), y(x0 ) = y0 ,

where p and r are continuous in an interval containing x = x0 , has an

explicit solution (justified on page 89)
Rx  Z x Rt 
− p(s)ds − p(s)ds
(3) y(x) = e x0 y0 + r(t)e x0 dt .
x0

Formula (3) is called variation of parameters, for historical reasons.

While (3) has some appeal, applications use the integrating factor
method below, which is developed with indefinite integrals for compu-
tational efficiency. No one memorizes (3); they remember and study the
method. See Example 11, page 87, for technical details.

Integrating Factor Identity

The technique called the integrating factor method uses the replace-
ment rule (justified on page 89)
 R 0
p(x)dx
e Y
(4) The fraction R replaces Y 0 + p(x)Y.
p(x)dx
e
R
p(x)dx
The factor e in (4) is called an integrating factor.

The Integrating Factor Method

Standard Rewrite y 0 = f (x, y) in the form y 0 + p(x)y = r(x)

Form where p, r are continuous. The method applies only in
case this is possible.
R
Find W Find a simplified
R
formula for W = e p(x)dx . The an-
tiderivative p(x)dx can be chosen conveniently.
(W y)0
Prepare for Obtain the new equation = r by replacing the
W
Quadrature 0
left side of y + p(x)y = r(x) by equivalence (4).
Method of Clear fractions to obtain (W y)0 = RrW . Apply the
Quadrature method of quadrature to get W y = r(x)W (x)dx +
C. Divide by W to isolate the explicit solution y(x).

In identity (4), functions p, Y and YR 0 are assumed continuous with p and

Y arbitrary functions. The integral p(x)dx R equals P (x)+C, where P (x)
is some anti-derivative of p(x). Because e p(x)dx = eP (x) eC , then factor
86

eC divides out of the fraction

R in (4). Applications therefore simplify
p(x)dx
the integrating factor e to eP (x) , where P (x) is any suitable
antiderivative of p(x) (effectively, we take C = 0).
Equation (4) is central to the method, because it collapses the two terms
y 0 + py into a single term (W y)0 /W ; the method of quadrature applies
to (W y)0 = rW . The literature calls the exponential factor W an inte-
grating factor and equivalence (4) a factorization of Y 0 + p(x)Y .
Simplifying an integrating factor. Factor W is simplifiedR by drop-
ping constants of integration. To illustrate, if p(x) = 1/x, then p(x)dx =
ln |x| + C. The algebra rule eA+B = eA eB implies that W = eC eln |x| =
|x|eC = (±eC )x. Let c1 = ±eC . Then W = c1 W1 where W1 = x.
The fraction (W y)0 /W reduces to (W1 y)0 /W1 , because c1 cancels. In an
application, we choose the simpler expression W1 . The illustration also
shows that the exponential in W can sometimes be eliminated.

Superposition
Formula (3) can be decomposed into two expressions, called yh and yp ,
so that the general solution is expressed as y = yh + yp . The function
yh solves the homogeneous equation y 0 + p(x)y = 0 and yp solves the
non-homogeneous equation y 0 + p(x)y = r(x). This observation is called
the superposition principle.
Equation (3) implies the homogeneous solution yh and a particular
solution yp∗ can be defined by
Rx Rx Z x Rt
− p(s)ds − p(s)ds p(s)ds
(5) yh = y0 e x0 , yp∗ =e x0 r(t)e x0 dt.
x0

Verification amounts to setting r = 0 in (3) to determine yh . The solution

yp∗ depends on the forcing term r(x), but yh does not. Initial conditions
of a problem are buried in yh . Experimentalists view the computation
of yp∗ as a single experiment in which the state yp∗ is determined by the
forcing term r(x) and zero initial data y = 0 at x = x0 .

Structure of Solutions
Formula (3), proved on page 89, directly establishes existence for the
solution to the linear initial value problem (2). The proof also determines
what other particular solutions might be used in the formula for a general
solution:
Theorem 3 (Solution Structure)
Assume p(x) and r(x) are continuous on a < x < b and a < x0 < b. Let yh
and yp∗ be defined by equation (5). Let y be a solution of y 0 + p(x)y = r(x)
on a < x < b. Then y can be decomposed as y = yh +yp∗ , where y0 = y(x0 ).
 2.3 Linear Equations I 87

In short, a linear equation has the solution structure homogeneous plus

particular. In particular, two solutions of the non-homogeneous equation
differ by some solution yh of the homogeneous equation.

Examples
11 Example (Integrating Factor Method) Solve the equation 2y 0 + 6y =
e−x by the integrating factor method.
1 −x
Solution: The solution is y = 4e + ce−3x . An answer check appears in
Example 13. The details:
y 0 + 3y = 0.5e−x Divide by 2 to get the standard form.
R
W = e3x Find the integrating factor W = e 3dx .

0
e3x y
= 0.5e−x Replace the LHS by the integrating factor quo-
e3x
tient; see page 85.

0
e3x y = 0.5e2x Clear fractions. Prepared for quadrature
e3x y = 0.5 e2x dx
R
Method of quadrature applied.
−3x 2x


y = 0.5e e /2 + c1 Evaluate the integral. Divide by W = e3x .
= 41 e−x + ce−3x Final answer, c = 0.5c1 .

12 Example (Superposition) Find a particular solution of y 0 + 2y = 3ex with

fewest terms.

Solution: The answer is y = ex . The first step solves the equation using the
integrating factor method, giving y = ex + ce−2x ; details below. A particular
solution with fewest terms, y = ex , is found by setting c = 0. The solution yp∗
of equation (5) has two terms: yp∗ = ex − e3x0 e−2x . The reason for the extra
term is the condition y = 0 at x = 0. The two particular solutions differ by the
homogeneous solution y0 e−2x where y0 = e3x0 .
Integrating factor method details:
y 0 + 2y = 3ex The standard form.
R
2dx
W = e2x Find the integrating factor W = e .

0
e2x y
= 3ex Integrating factor identity applied on the left.
e3x
e2x y = 3 e3x dx
R
Clear fractions and apply quadrature.
y = e−2x e3x + c


Evaluate the integral. Isolate y.
x −2x
= e + ce Solution found.

13 Example (Answer Check) Show the answer check details for 2y 0 + 6y =

e−x and candidate solution y = 41 e−x + ce−3x .

Solution: Details:
 88

LHS = 2y 0 + 6y Left side of the equation

2y 0 + 6y = e−x .
= 2(− 41 e−x − 3ce−3x ) + 6( 14 e−x + ce−3x ) Substitute for y.
= e−x + 0 Simplify terms.
= RHS DE verified.

14 Example (Finding yh and yp ) Find the homogeneous solution yh and a

particular solution yp for the equation 2xy 0 + y = 4x2 on x > 0.

Solution: The solution by the integrating factor method is y = 0.8x2 + cx−1/2 ;

details below. Then yh = cx−1/2 and yp = 0.8x2 give y = yh + yp .
The symbol yp stands for any particular solution. Variation of parameters gives
5/2
a different particular solution yp∗ = 0.8x2 − 0.8x0 x−1/2 . It differs from the
other particular solution 0.8x2 by a homogeneous solution Kx−1/2 .
Integrating factor method details:
y 0 + 0.5y/x = 2x Standard form. Divided by 2x.
R R
W = e0.5 dx/x The integrating factor is W = e p .
= e0.5 ln x Simplify the integration constant.
= x1/2 Used ln un = n ln u. Simplified W found.
1/2 0


x y
= 2x Integrating factor identity applied on the left.
x1/2
x1/2 y = 2 x3/2 dx
R
Clear fractions. Apply quadrature.
y = x−1/2 4x5/2 /5 + c


Evaluate the integral. Divide to isolate y.
= 4x2 + cx−1/2 Solution found.

15 Example (Classification) Classify the equation y 0 = x + ln (xey ) as linear

or non-linear.

Solution: It’s linear, with standard linear form y 0 + (−1)y = x + ln x. To

explain why, the term ln (xey ) on the right expands into ln x + ln ey , which in
turn is ln x + y, using logarithm rules. Because ey > 0, then ln(xey ) makes
sense for only x > 0. Henceforth, assume x > 0.
Computer algebra test f (x, y) = f (x, 0) + fy (x, 0)y. Expected is LHS −
RHS = 0 after simplification. This example produced ln ey − y instead of 0,
evidence that limitations may exist.

assume(x>0):
f:=(x,y)->x+ln(x*exp(y)):
LHS:=f(x,y):
RHS:=f(x,0)+subs(y=0,diff(f(x,y),y))*y:
simplify(LHS-RHS);

If the test passes, then y 0 = f (x, y) becomes y 0 = f (x, 0) + fy (x, 0)y. This
example gives y 0 = x + ln x + y, which converts to the standard linear form
y 0 + (−1)y = x + ln x.
2.3 Linear Equations I 89

Details and Proofs

Justification of Formula (3): Define
Rx Z x
− p(s)ds r(t)
Q(x) = e x0
, R(x) = dt.
x0 Q(t)
u 0 0 u
The
R x calculus rule (e ) = u e and the fundamental theorem of calculus result
( x0 G(t)dt)0 = G(x) can be used to obtain the formulas
r
Q0 = (−p)Q, R0 = .
Q
Existence. Equation (3) is y = Q(y0 + R). Existence will be established by
showing that y satisfies y 0 +py = r, y(x0 ) = y0 . The initial condition y(x0 ) = y0
follows from Q(x0 ) = 1 and R(x0 ) = 0. The steps below verify y 0 + py = r,
completing existence.
y 0 = [Q(y0 + R)]0 Equation (3), using notation Q and R.
0 0
= Q (y0 + R) + QR Sum and product rules applied.
0
= −pQ(y0 + R) + QR Used Q0 = (−p)Q.
= −pQ(y0 + R) + r Used R0 = r/Q.
= −py + r Apply y = Q(y0 + R).

Uniqueness. It remains to show that the solution given by (3) is the only
solution. Start by assuming Y is another, subtract them to obtain u = y − Y .
Then u0 + pu = 0, u(x0 ) = 0. To show y ≡ Y , it suffices to show u ≡ 0.
According to the integrating factor method, the equation u0 + pu = 0 is equiv-
alent to (W u)0 = 0 where W = eP and P(x) = x0 p(t)dt. Integrate (W u)0 = 0
Rx

from x0 to x, giving W (x)u(x) = W (x0 )u(x0 ). Since u(x0 ) = 0 and W (x) 6= 0,

it follows that u(x) = 0 for all x. This completes the proof.
Remarks on Picard’s Theorem. The Picard-Lindelöf theorem, page 61,
implies existence-uniqueness, but only on a smaller interval, and furthermore
it supplies no practical formula for the solution. Formula (3) is therefore an
improvement over the results obtainable from the general theory.

Justification of Factorization (4): It is assumed that Y (x) is a given but

otherwise arbitrary differentiable function. Equation (4) will be justified in its
fraction-free form
 0 Z
(6) eP Y = eP (Y 0 + pY ) , P(x) = p(x)dx.

 0
LHS = eP Y The left side of equation (6).
 0
= eP Y + eP Y 0 Apply the product rule (uv)0 = u0 v + uv 0 .

= eP pY + eP Y 0 Use the chain rule (eu )0 = eu u0 and P0 = p.

= eP (Y 0 + pY ) The common factor is eP .
= RHS The right hand side of equation (6).
90

Exercises 2.3
Integrating Factor Method. Apply 24. y 0 + 2y = 2
the integrating factor method, page 0
85, to solve the given linear equation. 25. 2y + y = 1
See the examples starting on page 87 26. 3y 0 + 2y = 1
for details.
27. y 0 − y = ex
1. y 0 + y = e−x
28. y 0 − y = xex
2. y 0 + y = e−2x
29. xy 0 + y = sin x (x > 0)
3. 2y 0 + y = e−x
30. xy 0 + y = cos x (x > 0)
4. 2y 0 + y = e−2x
31. y 0 + y = x − x2
5. 2y 0 + y = 1
32. y 0 + y = x + x2
0
6. 3y + 2y = 2
General Solution. Find yh and a par-
7. 2xy 0 + y = x ticular solution yp . Report the general
0 solution y = yh + yp . See Example 14,
8. 3xy + y = 3x
page 88.
9. y 0 + 2y = e2x
33. y 0 + y = 1
10. 2y 0 + y = 2ex/2 34. xy 0 + y = 2
11. y 0 + 2y = e−2x 35. y 0 + y = x
12. y 0 + 4y = e−4x 36. xy 0 + y = 2x
13. 2y 0 + y = e−x 37. y 0 − y = x + 1
14. 2y 0 + y = e−2x 38. xy 0 − y = 2x − 1
15. 4y 0 + y = 1 39. 2xy 0 + y = 2x2 (x > 0)
16. 4y 0 + 2y = 3 40. xy 0 + y = 2x2 (x > 0)

17. 2xy 0 + y = 2x Classification. Classify as linear or

non-linear. Use the test f (x, y) =
18. 3xy 0 + y = 4x
f (x, 0) + fy (x, 0)y and a computer al-
19. y 0 + 2y = e−x gebra system, when available, to check
the answer. See Example 15, page 88.
20. 2y 0 + y = 2e−x
41. y 0 = 1 + 2y 2
Superposition. Find a particular so- 42. y 0 = 1 + 2y 3
lution with fewest terms. See Example
12, page 87. 43. yy 0 = (1 + x) ln ey
2
21. 3y 0 = x 44. yy 0 = (1 + x) (ln ey )
22. 3y 0 = 2x 45. y 0 sec2 y = 1 + tan2 y

23. y 0 + y = 1 46. y 0 = cos2 (xy) + sin2 (xy)

2.3 Linear Equations I 91

47. y 0 (1 + y) = xy given in the integrating factor

method are related by yp = yp∗ +yh
48. y 0 = y(1 + y) for some solution yh of the homo-
49. xy 0 = (x + 1)y − xeln y geneous equation.

50. 2xy 0 = (2x + 1)y − xye− ln y 0

53. Let QR x= 1 + x and define y =
−1
Proofs and Details. Q e Q(x)dx. Show directly,
without appeal to theorems of this
51. Prove directly without appeal to section, that y 0 + (1 + x)y = ex .
Theorem 3 that the difference of
two solutions of y 0 + p(x)y = r(x) 0 2
is a solution of the homogeneous 54. Let Q = −1 x ln(1
R 2+ x ) and de-
equation y 0 + p(x)y = 0. fine y = Q x Q(x)dx. Show
directly, without appeal to the-
52. Prove that yp∗ givenR by equation orems of this section, that y 0 +
−1
(5) and yp = Q r(x)Q(x)dx x ln(1 + x2 )y = x2 .
92

2.4 Linear Equations II

Studied here are the subjects of variation of parameters and undeter-
mined coefficients for linear first order differential equations.

Variation of Parameters
A particular solution yp∗ (x) of the non-homogeneous equation

(1) y 0 + p(x)y = r(x)

is given by equation (5) in 2.3, reproduced below in (2). Literature calls

it the variation of parameters formula or the variation of constants
formula.
Theorem 4 (Variation of Parameters)
A particular solution of the differential equation y 0 + p(x)y = r(x) is given
by either of the formulae
Rx Z x Rt
− p(s)ds p(s)ds
(2) yp∗ (x) = e x0 r(t)e x0 dt,
x0
R Z R
(3) yp (x) = e− p(x)dx
r(x)e p(x)dx
dx.

Indefinite Integrals. The indefinite integral form (3) is used in sci-

ence and engineering applications. The answers (2) and (3) differ by a
solution of the homogeneous equation: yp∗ (x) = yp (x) + yh (x) for some
choice of the constant c in yh . Both answers (2) and (3) are solutions
of the nonhomogeneous differential equation, even though (2) generally
contains an extra term. While (2) satisfies y(x0 ) = 0, (3) may not.
Integrating Factor Formula. An integrating factor for (1) is
R
p(x)dx
Q(x) = e .

Formula (3) can be written in terms of Q(x) as

1
Z
yp (x) = r(x)Q(x)dx.
Q(x)

Compact Formula. Because xt f = xx0 f + xt0 f and

R R R R x0 Rx
x f =− x0 f,
the exponential factors in (2) can be re-written as
Z x Rt
p(s)ds
(4) yp (x) = r(t)e x dt.
x0

The reader is warned that using indefinite integrals in (4) results in the
wrong answer.
2.4 Linear Equations II 93

Terminology. The name variation of parameters comes from the idea

of varying the parameter Rc in the homogeneous solution formula yh =
cR(x), where R(x) = e− p(x)dx . Historically, c is replaced by an un-
known function y0 (x), to define a trial solution y(x) = y0 (x)R(x) of (1).
A derivation appears on page 97.

The Method of Undetermined Coefficients

The method applies to y 0 + p(x)y = r(x). It finds a particular solution
yp without the integration steps present in variation of parameters. The
requirements and limitations:

1. Coefficient p(x) of y 0 + p(x)y = r(x) is constant.

2. The function r(x) is a sum of constants times atoms.

An atom is a term having one of the forms

xm , xm eax , xm cos bx, xm sin bx, xm eax cos bx or xm eax sin bx.

The symbols a and b are real constants, with b > 0. Symbol m ≥ 0

is an integer. The terms x3 , x cos 2x, sin x, e−x , x6 e−πx are atoms.
Conversely, if r(x) = 4 sin x + 5xex , then split the sum into terms and
drop the coefficients 4 and 5 to identify atoms sin x and xex ; then r(x)
is a sum of constants times atoms.

The Method.

1. Repeatedly differentiate the atoms of r(x) until no new atoms ap-

pear. Multiply the distinct atoms so found by undetermined co-
efficients d1 , . . . , dk , then add to define a trial solution y.
2. Fixup rule: if solution e−px of y 0 + py = 0 appears in trial so-
lution y, then replace in y matching atoms e−px , xe−px , . . . by
xe−px , x2 e−px , . . . (other atoms appearing in y are unchanged).
The modified expression y is called the corrected trial solution.
3. Substitute y into the differential equation y 0 + py = r(x). Match
coefficients of atoms left and right to write out linear algebraic
equations for the undetermined coefficients d1 , . . . , dk .
4. Solve the equations. The trial solution y with evaluated coefficients
d1 , . . . , dk becomes the particular solution yp .

Undetermined Coefficients Illustrated. We will solve

y 0 + 2y = xex + 2x + 1 + 3 sin x.
94

Solution:
Test Applicability. The right side r(x) = xex + 2x + 1 + 3 sin x is a sum of
terms constructed from the atoms xex , x, 1, sin x. The left side is y 0 + p(x)y
with p(x) = 2, a constant. Therefore, the method of undetermined coefficients
applies to find yp .
Trial Solution. The atoms of r(x) are subjected to differentiation. The dis-
tinct atoms so found are 1, x, ex , xex , cos x, sin x (drop coefficients to identify
new atoms). The solution e−2x of y 0 + 2y = 0 does not appear in the list of
atoms, so the fixup rule does not apply. Then the trial solution is the expression

y = d1 (1) + d2 (x) + d3 (ex ) + d4 (xex ) + d5 (cos x) + d6 (sin x).

Equations. To substitute the trial solution y into y 0 + 2y requires a formula

for y 0 :
y 0 = d2 + d3 ex + d4 xex + d4 ex − d5 sin x + d6 cos x.
Then
r(x) = y 0 + 2y
= d2 + d3 ex + d4 xex + d4 ex − d5 sin x + d6 cos x
+ 2d1 + 2d2 x + 2d3 ex + 2d4 xex + 2d5 cos x + 2d6 sin x
= (d2 + 2d1 )(1) + 2d2 (x) + (3d3 + d4 )(ex ) + (3d4 )(xex )
+ (2d5 + d6 )(cos x) + (2d6 − d5 )(sin x)

Also, r(x) ≡ 1 + 2x + xex + 3 sin x. Coefficients of atoms on the left and right
must match. For instance, constant term d2 + 2d1 in the expansion of y 0 + 2y
matches constant term 1 in r(x). Writing out the matches gives the equations

2d1 + d2 = 1,
2d2 = 2,
3d3 + d4 = 0,
3d4 = 1,
2d5 + d6 = 0,
− d5 + 2d6 = 3.

Solve. The first four equations can be solved by back-substitution to give

d2 = 1, d1 = 0, d4 = 1/3, d3 = −1/9. The last two equations are solved
by elimination or Cramer’s rule (reviewed in Chapter 3) to give d6 = 6/5,
d5 = −3/5.
Report yp . The trial solution y with evaluated coefficients d1 , . . . , d6 becomes

1 1 3 6
yp (x) = x − ex + xex − cos x + sin x.
9 3 5 5

A Fixup Rule Illustration. Solve the equation

y 0 + 3y = 8ex + 3x2 e−3x

 2.4 Linear Equations II 95

by the method of undetermined coefficients. Verify that the general

solution y = yh + yp is given by

yh = ce−3x , yp = 2ex + x3 e−3x .

Solution: The right side r(x) = 8ex + 3x2 e−3x is constructed from atoms ex ,
x2 e−3x . Repeated differentiation of these atoms identifies the new list of atoms
ex , e−3x , xe−3x , x2 e−3x . The fixup rule applies because the solution e−3x of
y 0 + 3y = 0 appears in the list. The atoms of the form xm e−3x are multiplied
by x to give the new list of atoms ex , xe−3x , x2 e−3x , x3 e−3x . Readers should
take note that atom ex is unaffected by the fixup rule modification. Then the
corrected trial solution is

y = d1 ex + d2 xe−3x + d3 x2 e−3x + d4 x3 e−3x .

The trial solution expression y is substituted into y 0 + 3y = 2ex + x2 e−3x to

give the equation

4d1 ex + d2 e−3x + 2d3 xe−3x + 3d4 x2 e−3x = 8ex + 3x2 e−3x .

Coefficients of atoms on each side of the preceding equation are matched to give
the equations
4d1 = 8,
d2 = 0,
2d3 = 0,
3d4 = 3.
Then d1 = 2, d2 = d3 = 0, d4 = 1 and the particular solution is reported to be
yp = 2ex + x3 e−3x .

Examples
16 Example (Variation of Parameters Method) Solve the equation 2y 0 +
6y = 4xe−3x by the method of variation of parameters, verifying y = yh +yp
is given by
yh = ce−3x , yp = x2 e−3x .

Solution: Divide the equation by 2 to obtain the standard linear form

y 0 + 3y = 2xe−3x .

Solution yh . The homogeneous equation y 0 + 3y = 0 is solved by the recipe to

give yh = ce−3x .
Solution yp . Identify p(x) = 3, r(x) = 2xe−3x from the standard form. The
mechanics: let y 0 = f (x, y) ≡ 2xe−3x − 3y and define r(x) = f (x, 0), p(x) =
−fy (x, y) = 3. The variation of parameters
R formula is applied as follows. First,
p(x)dx
compute the integrating factor Q(x) = e = e3x . Then
Z
yp (x) = (1/Q(x)) r(x)Q(x)dx
= e−3x 2xe−3x e3x dx
R

= x2 e−3x .
 96

It must be explained that all integration constants were set to zero, in order
to obtain the shortest possible expression for yp . Indeed, if Q = e3x+c1 instead
of e3x , then the factors 1/Q and Q contribute constant factors 1/ec1 and ec1 ,
which multiply to one; the effectR is to set c1 = 0. On the other hand, an
integration constant c2 added to r(x)Q(x)dx adds the homogeneous solution
c2 e−3x to the expression for yp . Because we seek the shortest expression which
is a solution to the non-homogeneous differential equation, the constant c2 is
set to zero.

17 Example (Undetermined Coefficient Method) Solve the equation 2y 0 +

6y = 4xe−x + 4xe−3x + 5 sin x by the method of undetermined coefficients,
verifying y = yh + yp is given by
1 1 3
yh = ce−3x , yp = − e−x + xe−x + x2 e−3x − cos x + sin x.
2 4 4
Solution: The method applies, because the differential equation 2y 0 + 6y = 0
has constant coefficients and the right side r(x) = 4xe−x + 4xe−3x + 5 sin x is
constructed from the list of atoms xe−x , xe−3x , sin x.
List of Atoms. Differentiate the atoms xe−x , xe−3x , sin x to find the new list
of atoms e−x , xe−x , e−3x , xe−3x , cos x, sin x. The solution e−3x of 2y 0 + 6y = 0
appears in the list: the fixup rule applies. Then e−3x , xe−3x are replaced by
xe−3x , x2 e−3x to give the corrected list of atoms e−x , xe−x , xe−3x , x2 e−3x ,
cos x, sin x. Please note that only two of the six atoms were corrected.
Trial solution. The corrected trial solution is

y = d1 e−x + d2 xe−x + d3 xe−3x + d4 x2 e−3x + d5 cos x + d6 sin x.

Substitute y into 2y 0 + 6y = r(x) to give

r(x) = 2y 0 + 6y
= (4d1 + 2d2 )e−x + 4d2 xe−x + 2d3 e−3x + 4d4 xe−3x
+(2d6 + 6d5 ) cos x + (6d6 − 2d5 ) sin x.

Equations. Matching atoms on the left and right of 2y 0 + 6y = r(x), given

r(x) = 4xe−x + 4xe−3x + 5 sin x, justifies the following equations for the un-
determined coefficients; the solution is d2 = 1, d1 = −1/2, d3 = 0, d4 = 1,
d6 = 3/4, d5 = −1/4.

4d1 + 2d2 = 0,
4d2 = 4,
2d3 = 0,
4d4 = 4,
6d5 + 2d6 = 0,
− 2d5 + 6d6 = 5.

Report. The trial solution upon substitution of the values for the undetermined
coefficients becomes
1 1 3
yp = − e−x + xe−x + x2 e−3x − cos x + sin x.
2 4 4
2.4 Linear Equations II 97

Details
Historical Account of Variation of Parameters.
R
r = y 0 + py Let R(x) = e− p(x)dx . Assume y =
y0 (x)R(x) solves y 0 + py = r.
= (y0 R)0 + py0 R Substitute y = y0 (x)R(x) but suppress x.
= y00 R + y0 R0 + py0 R Apply the product rule (uv)0 = u0 v + uv 0 .
R
= y00 R − y0 pR + py0 R Let Q = e p(x)dx . Apply Q0 = −pQ.
= y00 /Q Because 1/R = Q.
0
The calculation gives
R xy0 (x) = r(x)Q(x). The method of quadrature applies to
determine y0 (x) = x0 (r(t)Q(t))dt, because y0 = 0 at x = x0 . Then y = y0 R
duplicates the formula for yp∗ given in (5), which is equivalent to (2).

Exercises 2.4
Variation of Parameters I. Report 14. y 0 + y = e−2x , x0 = 0
the shortest particular solution given
by the formulaR 15. y 0 − 2y = 1, x0 = 0
R
rQ
yp (x) = Q , Q = e p . 16. y 0 − y = 1, x0 = 0

1. y 0 = x + 1 17. 2y 0 + y = ex , x0 = 1

2. y 0 = 2x − 1 18. 2y 0 + y = e−x , x0 = 1

3. y 0 + y = e−x 19. xy 0 = x + 1, x0 = 1

4. y 0 + y = e−2x 20. xy 0 = 1 − x2 , x0 = 1

5. y 0 − 2y = 1 Atoms. Report the list of distinct

atoms of the given function f (x).
6. y 0 − y = 1
21. x + ex
0 x
7. 2y + y = e
22. 1 + 2x + 5ex
0 −x
8. 2y + y = e
23. x(1 + x + 2ex )
9. xy 0 = x + 1
24. x2 (2 + x2 ) + x2 e−x
10. xy 0 = 1 − x2 25. sin x cos x + ex sin 2x

Variation of Parameters II. Com- 26. cos2 x − sin2 x + x2 ex cos 2x

pute the particular solution given
R t by 27. (1 + 2x + 4x5 )ex e−3x ex/2
Rx
rQ p
yp∗ (x) = Q(t)
x0
, Q(t) = e x0 .
28. (1+2x+4x5 +ex sin 2x)e−3x/4 ex/2
0
11. y = x + 1, x0 = 0
x + ex
29. sin 3x + e3x cos 3x
12. y 0 = 2x − 1, x0 = 0 e−2x
x + ex sin 2x + x3
13. y 0 + y = e−x , x0 = 0 30. sin 5x
e−2x
98

Initial Trial Solution. Differentiate 51. y 0 = x3 + 5 + x2 ex (3 + 2x + sin 2x)

repeatedly f (x) and report the list of 0 2 3 x
distinct atoms which appear in f and 52. y = x + 5x + 2 + x e (2 + 3x +
all its derivatives. 5 cos 4x)

31. 12 + 5x2 + 6x7 53. y 0 − y = x3 + 2x + 5 + x4 ex (2 +

4x + 7 cos 2x)
32. x6 /x−4 + 10x4 /x−6
54. y 0 − y = x4 + 5x + 2 + x3 ex (2 +
2 x
33. x + e 3x + 5 cos 4x)
34. x3 + 5e2x 55. y 0 −2y = x3 +x2 +x3 ex (2ex +3x+
35. (1 + x + x3 )ex + cos 2x 5 sin 4x)

36. (x + ex ) sin x + (x − e−x ) cos 2x 56. y 0 − 2y = x3 e2x + x2 ex (3 + 4ex +

2 cos 2x)
37. (x + ex + sin 3x + cos 2x)e−2x
57. y 0 + y = x2 + 5x + 2 + x3 e−x (6x +
38. (x2 e−x + 4 cos 3x + 5 sin 2x)e−3x 3 sin x + 2 cos x)
39. (1 + x2 )(sin x cos x − sin 2x)e−x 58. y 0 − 2y = x5 + 5x3 + 14 + x3 ex (5 +
40. (8 − x3 )(cos2 x − sin2 x)e3x 7xe−3x )

59. 2y 0 +4y = x4 +5x5 +2x8 +x3 ex (7+

Fixup Rule. Given the homogeneous 5xex + 5 sin 11x)
solution yh and an initial trial solution
y, determine the final trial solution ac- 60. 5y 0 + y = x2 + 5x + 2ex/5 +
cording to the fixup rule. x3 ex/5 (7 + 9x + 2 sin(9x/2))
41. yh (x) = ce2x , y = d1 +d2 x+d3 e2x
Undetermined Coefficients. Com-
42. yh (x) = ce2x , y = d1 + d2 e2x + pute a particular solution yp according
d3 xe2x to the method of undetermined coeffi-
cients. Report (1) the initial trial so-
43. yh (x) = ce0x , y = d1 + d2 x + d3 x2 lution, (2) the corrected trial solution,
44. yh (x) = cex , y = d1 + d2 x + d3 x2 (3) the system of equations for the un-
determined coefficients and finally (4)
45. yh (x) = cex , y = d1 cos x + the formula for yp .
d2 sin x + d3 ex
61. y 0 + y = x + 1
46. yh (x) = ce2x , y = d1 e2x cos x +
d2 e2x sin x 62. y 0 + y = 2x − 1
0 x −x
47. yh (x) = ce2x , y = d1 e2x + 63. y − y = e + e
2x 2 2x
d2 xe + d3 x e
64. y 0 − y = xex + e−x
−2x −2x
48. yh (x) = ce , y = d1 e +
d2 xe−2x + d3 e2x + d4 xe2x 65. y 0 − 2y = 1 + x + e2x + sin x

49. yh (x) = cx2 , y = d1 + d2 x + d3 x2 66. y 0 − 2y = 1 + x + xe2x + cos x

50. yh (x) = cx3 , y = d1 + d2 x + d3 x2 67. y 0 + 2y = xe−2x + x3

0 −2x x
Undetermined Coefficients: Trial 68. y + 2y = (2 + x)e + xe
Solution. Find the form of the cor- 69. y 0 = x2 + 4 + xex (3 + cos x)
rected trial solution y but do not eval-
uate the undetermined coefficients. 70. y 0 = x2 + 5 + xex (2 + sin x)