Professional Documents
Culture Documents
4.6
Variable-Coefficient Equations
193
EXERCISES
4.7
16. y 5y 6y 18t 2
1
1
17. y 2y tan 2t e t
2
2
18. y 6y 9y t 3e 3t
19. Express the solution to the initial value problem
1
y y ,
y A 1 B 0 , y A 1 B 2 ,
t
using denite integrals. Using numerical integration
(Appendix C) to approximate the integrals, nd an
approximation for y(2) to two decimal places.
20. Use the method of variation of parameters to show that
y A t B c1 cos t c2 sin t
f AsB sin At sB ds
0
VARIABLE-COEFFICIENT EQUATIONS
The techniques of Sections 4.2 and 4.3 have explicitly demonstrated that solutions to a linear
homogeneous constant-coefcient differential equation,
(1)
ay by cy 0 ,
are dened and satisfy the equation over the whole interval (q, q). After all, such solutions are combinations of exponentials, sinusoids, and polynomials.
The variation of parameters formula of Section 4.6 extended this to nonhomogeneous
constant-coefcient problems,
(2)
ay by cy (t) ,
yielding solutions valid over all intervals where (t) is continuous (ensuring that the integrals in
(10) of Section 4.6 containing (t) exist and are differentiable). We could hardly hope for more;
indeed, it is debatable what meaning the differential equation (2) would have at a point where
f(t) is undened, or discontinuous.
194
Chapter 4
Therefore, when we move to the realm of equations with variable coefcients of the form
(3)
the most we can expect is that there are solutions that are valid over intervals where all four
governing functions a2(t), a1(t), a0(t), and (t)are continuous. Fortunately, this expectation is fullled except for an important technical requirementnamely, that the coefcient
function a2(t) must be nonzero over the interval.
Typically, one divides by the nonzero coefcient a2(t) and expresses the theorem for the
equation in standard form [see (4), below] as follows.
Example 1
y(t0) Y0 ,
y(t0) Y1 .
Determine the largest interval for which Theorem 5 ensures the existence and uniqueness of a
solution to the initial value problem
(5)
Solution
(t 3)
d 2y
dt
dy
2t y ln t ;
dt
y(1) 3 ,
y(1) 5 .
The data p(t), q(t), and g(t) in the standard form of the equation,
y py qy
d 2y
dt
1 dy
2t
ln t
y
g ,
(t 3) dt
(t 3)
(t 3)
are simultaneously continuous in the intervals 0 6 t 6 3 and 3 6 t 6 q . The former contains the point t0 1, where the initial conditions are specied, so Theorem 5 guarantees (5)
has a unique solution in 0 6 t 6 3.
Theorem 5, embracing existence and uniqueness for the variable-coefcient case, is difcult
to prove because we cant construct explicit solutions in the general case. So the proof is deferred
to Chapter 13. However, it is instructive to examine a special case that we can solve explicitly.
Indeed, the whole nature of the equationreduction from second-order to rst-orderchanges at points where a2(t)
is zero.
All references to Chapters 1113 refer to the expanded text Fundamentals of Differential Equations and Boundary
Value Problems, 6th ed.
Section 4.7
Variable-Coefficient Equations
195
ty trt r1 rt r ,
and substitution into the homogeneous form of (6) (that is, with g 0) yields a simple
quadratic equation for r:
ar(r 1)t r brt r ct r [ar 2 (b a)r c]t r 0 , or
(7)
ar 2 (b a)r c 0 ,
Solution
t 7 0 .
y 2(t) t 3
(for t 7 0) .
196
Chapter 4
Then we simplify as in Section 4.3 by taking the real and imaginary parts to form independent
solutions:
y1 t cos( ln t) ,
(8)
y2 t sin( ln t) .
If r is a double root of the characteristic equation (7), then independent solutions of the
CauchyEuler equation on (0, q) are given by
y1 t r ,
(9)
y2 t r ln t .
This can be veried by direct substitution into the differential equation. Alternatively, the
second, linearly independent, solution can be obtained by reduction of order, a procedure to
be discussed shortly in Theorem 8. Furthermore, Problem 23 demonstrates that the substitution t e x changes the homogeneous CauchyEuler equation into a homogeneous constantcoefcient equation, and the formats (8) and (9) then follow from our earlier deliberations.
We remark that if a homogeneous CauchyEuler equation is to be solved for t
0, then
one simply introduces the change of variable t t, where t 0. The reader should verify
via the chain rule that the identical characteristic equation (7) arises when tr A t B r is substituted in the equation. Thus the solutions take the same form as (8), (9), but with t replaced by
t; for example, if r is a double root of (7), we get A t B r and A t B r ln A t B as two linearly independent solutions on A q, 0 B .
Example 3
Find a pair of linearly independent solutions to the following CauchyEuler equations for t 0.
(a) t 2 y 5ty 5y 0
Solution
(b) t 2y ty 0
For part (a), the characteristic equation becomes r 2 4r 5 0 , with the roots r 2 i ,
and (8) produces the real solutions t2 cos(ln t) and t2 sin(ln t).
For part (b), the characteristic equation becomes simply r 2 0 with the double root r 0,
and (9) yields the solutions t 0 1 and ln t.
In Chapter 8 we will see how one can obtain power series expansions for solutions to
variable-coefcient equations when the coefcients are analytic functions. But, as we said, there
is no procedure for explicitly solving the general case. Nonetheless, thanks to the existence/
uniqueness result of Theorem 5, most of the other theorems and concepts of the preceding sections are easily extended to the variable-coefcient case, with the proviso that they apply only
over intervals in which the governing functions p(t), q(t), g(t) are continuous. Thus we have the
following analog of Lemma 1, page 162.
If y1(t) and y2(t) are any two solutions to the homogeneous differential
on an interval I where the functions p(t) and q(t) are continuous and if the Wronskian
W[y1, y2 ](t) J y1(t)y2(t) y1(t)y 2(t) `
y1(t) y2(t)
`
y1(t) y2(t)
The determinant representation of the Wronskian was introduced in Problem 34, Section 4.2.
Section 4.7
Variable-Coefficient Equations
197
a general solution on I is given by y yp yh, where yh c1y1 c2y2 is a general solution to the corresponding homogeneous equation (10) on I and yp is a particular solution
to (11) on I. In other words, the solution to the initial value problem stated in Theorem 5 must
be of this form for a suitable choice of the constants c1, c2 . This follows, just as before, from a
straightforward extension of the superposition principle for variable-coefcient equations
described in Problem 30.
If linearly independent solutions to the homogeneous equation (10) are known, then yp can
be determined for (11) by the variation of parameters method.
Variation of Parameters
Theorem 7. If y1 and y2 are two linearly independent solutions to the homogeneous
equation (10) on an interval I where p(t), q(t), and g(t) are continuous, then a particular
solution to (11) is given by yp y1y1 y2 y2, where y1 and y2 are determined up to a
constant by the pair of equations
y1Y1 y2Y2 0 ,
y1 Y1 y2 Y2 g ,
which have the solution
(12)
y1(t)
g(t) y2(t)
W[y , y ](t) dt ,
1
y2(t)
W[y , y ](t) dt .
g(t)y1(t)
1
Note the formulation (12) presumes that the differential equation has been put into
standard form [that is, divided by a2(t)].
The proofs of the constant-coefcient versions of these theorems in Sections 4.2 and 4.5
did not make use of the constant-coefcient property, so one can prove them in the general case
by literally copying those proofs but interpreting the coefcients as variables. Unfortunately,
however, there is no construction analogous to the method of undetermined coefcients for the
variable-coefcient case.
198
Chapter 4
What does all this mean? The only stumbling block for our completely solving nonhomogeneous initial value problems for equations with variable coefcients,
y p(t)y q(t)y g(t) ;
y(t0) Y0 ,
y(t0) Y1 ,
is the lack of an explicit procedure for constructing independent solutions to the associated
homogeneous equation (10). If we had y1 and y2 as described in the variation of parameters
formula, we could implement (12) to nd yp, formulate the general solution of (11) as
yp c1y1 c2y2, and (with the assurance that the Wronskian is nonzero) t the constants to
the initial conditions. But with the exception of the CauchyEuler equation and the ponderous
power series machinery of Chapter 8, we are stymied at the outset; there is no general procedure for nding y1 and y2 .
Ironically, we only need one nontrivial solution to the associated homogeneous equation,
thanks to a procedure known as reduction of order that constructs a second, linearly independent solution y2 from a known one y1. So one might well feel that the following theorem rubs
salt into the wound.
Reduction of Order
Theorem 8. Let y1(t) be a solution, not identically zero, to the homogeneous differential
equation (10) in an interval I (see page 196). Then
(13)
y2(t) y1(t)
p(t)dt
ey (t)
1
dt
This remarkable formula can be conrmed directly, but the following derivation shows
how the procedure got its name.
Proof of Theorem 8. Our strategy is similar to that used in the derivation of the variation
of parameters formula, Section 4.6. Bearing in mind that cy1 is a solution of (10) for any constant c, we replace c by a function v(t) and propose the trial solution y2(t) v(t)y1(t), spawning
the formulas
y2 vy1 vy1 ,
or, on regrouping,
(14)
The group in front of the undifferentiated v(t) is simply a copy of the left-hand member of the
original differential equation (10), so it is zero. Thus (14) reduces to
(15)
Section 4.7
Variable-Coefficient Equations
199
Indeed, (16) is separable and can be solved immediately using the procedure of
Section 2.2. Problem 50 carries out the details of this procedure to complete the derivation
of (13).
Example 4
y
1
1
y 2 y 0 ,
t
t
use the reduction of order procedure to determine a second linearly independent solution for
t 0.
Solution
Rather than implementing the formula (13), lets apply the strategy used to derive it. We set
y2 A t B v A t B y1 A t B v A t B t and substitute y2 vt v, y2 vt 2v into (17) to nd
(18)
1
1
vt 2v A vt v B 2 vt vt A 2v v B vt v 0 .
t
t
0 6 t 6 p .
As we indicated above, the tricky part is to nd a single nontrivial solution. Inspection of (19)
suggests that y sin t or y cos t, combined with a little luck with trigonometric identities,
might be solutions. In fact, trial and error shows that the cosine function works:
y1 cos t ,
y1 sin t ,
y1 cos t ,
(sin t)y1 2(cos t)y1 (sin t)y1 (sin t) (cos t) 2(cos t)(sin t) (sin t)(cos t) 0 .
Unfortunately, the sine function fails (try it).
So we use reduction of order to construct a second, independent solution. Setting
y2 A t B v A t B y1 A t) v A t B cos t and computing y2 v cos t v sin t, y2 v cos t 2v sin t
v cos t, we substitute into (19) to derive
Asin t B 3 v cos t 2v sin t v cos t 4 2 A cos t B 3 v cos t v sin t 4 A sin t B 3 v cos t 4
A sin t B A cos t B
2
sec2 t
.
tan t
Reverse osmosis is a process used to fortify the alcoholic content of wine, among other applications.
200
Chapter 4
In this section we have seen that the theory for variable-coefcient equations differs only
slightly from the constant-coefcient case (in that solution domains are restricted to intervals),
but explicit solutions can be hard to come by. In the next section, we will supplement our exposition by describing some nonrigorous procedures that sometimes can be used to predict qualitative features of the solutions.
4.7
EXERCISES
y A1B 0
dy
2t
6y 0
9. t
2
dt
dt
2
10. t 2y A t B 7ty A t B 7y A t B 0
d 2w
6 dw
4
w0
t dt
dt 2
t2
d 2z
dz
5t
4z 0
12. t 2
dt
dt 2
11.
In Problems 19 and 20, solve the given initial value problem for the CauchyEuler equation.
19. t 2y A t B 4ty A t B 4y A t B 0 ;
y A 1 B 2 ,
y A 1 B 11
20. t 2y A t B 7ty A t B 5y A t B 0 ;
y A 1 B 1 ,
y A 1 B 13
In Problems 21 and 22, devise a modication of the
method for CauchyEuler equations to nd a general
solution to the given equation.
21. A t 2 B 2y A t B 7 A t 2 B y A t B 7y A t B 0 ,
t 7 2
22. A t 1 B 2y A t B 10 A t 1 B y A t B 14y A t B 0 ,
t 7 1
23. To justify the solution formulas (8) and (9), perform
the following analysis.
(a) Show that if the substitution t ex is made in
the function y A t B and x is regarded as the new
independent variable in Y A x B J y A ex B , the chain
rule implies the following relationships:
13. 9t 2y A t B 15ty A t B y A t B 0
14. t y A t B 3ty A t B 4y A t B 0
2
dy
dY
,
dt
dx
t2
d 2y
dt
d 2Y
dx 2
dY
.
dx