Вы находитесь на странице: 1из 19

L14.

htm Page 1 of 19

ORDINARY DIFFERENTIAL EQUATIONS


Lecture 14

Linear Nonhomogeneous Second-Order ODEs - Variation of


Parameters & Green's Kernel

(Revised 16 April 2009 @17:00)

Professor Stephen H Saperstone


Department of Mathematical Sciences
George Mason University
Fairfax, VA 22030

email: sap@gmu.edu

Copyright © 2009 by Stephen H Saperstone


All rights reserved

14.1 VARIATION OF PARAMETERS (VP)


Recall the linear nonhomogeneous ODE with constant coefficients from Lecture 13

We assume that is a continuous function on some interval The method of variation of parameters extends the scope of
the UC method to include any continuous forcing function on some interval

Theorem 14.1 [Variation of Parameters - VP]


Assume
1. is a FSS for
2. is the Wronskian of and
3. is a continuous function on some interval

Then a particular solution to

is given by

where and are defined by their derivatives

for all

PROOF: Click here .

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 2 of 19

The implementation of the method of variation of parameters requires the antiderivitives of the expressions for and in
Eqn. (14.3). Although the VP method allows any continuous function the antidifferentiation may not be possible as we see
later in Example 14.10. The following examples illustrate the use of the VP method.

Example 14.2
Use the VP method to compute a particular solution to the ODE

Solution: First compute a FSS for the corresponding homogeneous ODE

The characteristic equation is Factor as

to get the repeated real roots

From Table 11.1 we get the FSS

Label these as

(The labeling is important as we must maintain consistency in identifying the FSS functions in calculating a particular solution by the VP
method.) Next compute the Wronskian from Table 11.2:

Then with we have

Calculate the antiderivatives to get

(Note that is continuous on so that we may drop the absolute value bars.) Although constants of integration and
are indicated, we can take them to be zero as will be explained shortly. Thus we have a particular solution

End of Example 14.2

Remark 14.3

Why the constants of integration and may be taken to be zero.


If we did not take and to be zero in Example 14.2, the particular solution would have been

But the expression is a solution of the homogeneous ODE, Eqn (14.5). In formulating a general solution
to the nonhomogeneous ODE, Eqn (14.4), the expression would be redundant - it has the exact same
form as a general solution to the homogeneous ODE. Since a general solution to the Eqn (14.4) has the form

then the expression in Eqn (14.7) can be absorbed into thereby allowing us to choose

End of Remark 14.3

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 3 of 19

Example 14.4:
Use the VP method to compute a particular solution to the ODE

Solution: First compute a FSS for the corresponding homogeneous ODE

The characteristic equation is

to get the complex conjugate roots (with real part zero)

From Table 11.1 we get the FSS

In view of the defining relations, Eqn. (14.3) for and set

Next compute the Wronskian from Table 11.2 (noting that ), or from the definition:

Then with we have

Calculate the antiderivatives to get

(Use an integral table if you must to evaluate Note the following:

is continuous on .
is positive on so that we may drop the absolute value bars in the computation of
Take

Thus we have a particular solution

End of Example 14.4

Remark 14.5 [The Power of VP]


Why the UC method fails in Examples 14.2 & 14.4

Revisit Example 14.1: Calculate:

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 4 of 19

So

is an infinite set of LI functions on It follows that we can never apply the UC method to this

Revisit Example 14.4: Calculate:

So

is an infinite set of LI functions on It follows that we can never apply the UC method to this
End of Remark 14.5

Sometimes we use the VP method when is a UC function. Compare the solution of the particular solution in the next
example with that of Example 13.11 in which the UC method was employed.

Example 14.6
Use the VP method to compute a particular solution to the ODE

Solution: First we get the FSS for the corresponding homogeneous ODE from Example 13.11.:

(The labeling is important as we must maintain consistency in identifying the FSS functions in calculating a particular solution by the VP
method.) Next compute the Wronskian from Table 11.2

Then with we have

Calculate the antiderivatives to get

Thus we have a particular solution

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 5 of 19

the same as calculated in Example 13.11.


End of Example 14.6

Example 14.7
Use the VP method to compute a particular solution to the ODE

Solution: First determine a FSS for the homogeneous ODE

The characteristic equation is Factor as

to get the repeated real roots

A FSS is given by the pair

(The labeling is important as we must maintain consistency in identifying the FSS functions in calculating a particular solution by the VP
method.) Next compute the Wronskian from Table 11.2

Then with we have

Calculate the antiderivatives to get

Thus we have a particular solution

End of Example 14.7

The calculation of the solution to an IVP proceeds as follows: determine the general solution to the
nonhomogeneous ODE and use the initial data to calculate the coefficients in

Example 14.8
Solve the IVP

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 6 of 19

Solution: We calculated a FSS in Example 14.4. Consequently, the complementary solution is

A particular solution was also calculated in Example 14.4:

Thus the general solution to the given ODE is

To determine and first calculate :

From we have

and from we have

Hence, and So we get the solution to the IVP

End of Example 14.8

Example 14.9
Solve the IVP

Solution: The FSS for the corresponding homogeneous ODE and the Wronskian were calculated in Example 14.4:

Then with we have

Calculate the antiderivatives to get

Thus we have a particular solution

which upon simplification yields

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 7 of 19

Thus the general solution to the given ODE is

To determine and first calculate :

From we have

and from we have

Hence, and So we get the solution to the IVP

End of Example 14.9

The VP method may present us with expressions for and which we can't integrate. In this event, the particular solution can
only be expressed as an integral.

Example 14.10
Use the VP method to compute a particular solution to the ODE

Solution: The FSS for the corresponding homogeneous ODE and the Wronskian were calculated in Example 13.4:

Then with we have

The antiderivatives cannot be evaluated, so we must leave them as

Thus we have the particular solution

It is not a good idea to leave the particular solution in this form. We saw in the Integral Formula Solution for the solution to a linear first-order
ODE in Lecture 5 how to represent a solution in integral form. Replace the variable of integration with a dummy variable, say and integrate
over the interval where and are any points the interval of continuity of Since is undefined at we can take or .
Typically, would be the initial time from a set of initial conditions. For the sake of having an actual value in this example, we choose
and
Thus we can express and as

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 8 of 19

Then

Further simplification of is possible. Bring the and terms into the integrals - we can do this because is a fixed quantity. Thus

End of Example 14.10

Example 14.11
Solve the IVP

and plot the solution on the interval

Solution: We get the FSS comes from Example 14.10. Consequently, the complementary solution is

A particular solution was calculated in Example 14.10:

Thus the general solution to the given ODE is

To determine and first calculate :

Leibniz's Rule is needed for the indicated derivative of integral as the variable appears in both the upper limit of integration and the integrand.
We get

From we have

and from we have

Hence, and So we get the solution to the IVP

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 9 of 19

Figure 14.1: Solution to

The plot in Figure 14.1 was created in Maple; click here for an active Maple worksheet (you must have Maple installed on the computer in
which you are viewing this page) or here for an html (non excecutable rendition of the Maple worksheet) image of the worksheet.
End of Example 14.11

14.2 GREEN'S KERNEL


We saw in Example 14.10 how the solution was in the form of an integral. Let's revisit the construction of a particular solution
of by the VP method. Suppose that

is a continuous function on an interval


is a FSS for
is the Wronskian of and

Consider the formulas for and defined by

for all Choose any point in and integrate Eqn. (14.3) over the interval (We did a similar integration in
Lecture 6 where we developed the Integral Formula Solution to a first-order linear ODE of the form )
Typically, is specified in the initial conditions. Choose as one of the points of since it is at the point we want to
calculate the values of and . Then using the dummy variable of integration we get

The particular solution can now be expressed as

Since the variable of integration is the variable is actually just a constant - indeed, it is just the upper limit of integration.
Thus and do not depend on so and can be shifted to within the integrals as follows.

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 10 of 19

Set

so that we can write


Green's Kernel Solution for a Particular Solution:

Theorem 14.12
Suppose is a continuous function on some interval and let be any point in Then a particular solution to the linear
nonhomogeneous ODE

has the form

for all

Definition 14.13
The function is called Green's kernel (or Green's function) for the FSS Observe that Eqn. (14.10)
can be written as a quotient of determinants:

The green's kernel for a given linear homogeneous ODE is independent of the FSS as the next theorem stipulates.

Theorem 14.14 [ is Independent of FSS]

Suppose and are FSS's for Then the corresponding Green's kernels
and are identical.

PROOF: Click here .

Example 14.15
Calculate the Green's kernel for the homogeneous ODE

Solution: First compute a FSS for the corresponding homogeneous ODE. We leave it to the reader to check that and
forms a FSS. Then from Eqn. (14.12) we have

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 11 of 19

End of Example 14.15

The Green's kernel has the following representation as stated in the next theorem. This representation is remarkable as
doesn't depend on specific values for and just depends on and through their difference, The proof is important
enough to feature.

Theorem 14.16 (Green's Kernel Forms)

Suppose is a FSS for Depending on the roots of the characteristic polynomial, the Green's
kernel is given by one of the following formulas:

PROOF: Consider the three cases from Table 11.2 and calculate for each case using Eqn. (14.12).

Case 1: The roots are

Case 2: The roots are Apply L'Hopital's rule to the result of Case 1, treating as the independent variable:

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 12 of 19

Alternatively, we could have computed directly from the FSS as was done in Case 1.

Case 3: The roots are Because the formula for in Case 1 does not depend on and
being real numbers, we can substitute and in the result for Case 1 to get

Q.E.D.

The FSS and Green's kernel for each case is summarized in Table 14.1 below.

TABLE 14.1

The value of using the Green's kernel to determine a particular solution is evident when integrals that define the coefficients
and cannot be computed in terms of elementary functions.

Example 14.17
Use Green's kernel to determine a particular solution for ODE

Solution: We computed the Green's kernel for in Example 14.8:

Observe that is only defined and continuous on Choose (as we need positive value for Then use Eqn.
(14.11) to obtain

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 13 of 19

Since the last integral cannot be evaluated any further, we must stop at this point. Thus a particular solution is

End of Example 14.17

The Green's kernel is also useful to calculate a particular solution when the forcing function is defined piecewise. Observe
that the integral form of the particular solution

resembles the integral formula solution for a first-order linear IVP, Eqn. (6.6). Examples 6.7, 6.8, and 6.9 of Lecture 6. The
next example illustrates how to integrate Eqn. (14.11) for a piecewise continuous function whose graph of is given in
Figure 14.2.

Figure 14.2: Piecewise continuous forcing function

Example 14.18
The idealized single-story building frame depicted in Figure 14.3(a) is modeled by the equivalent undamped mass-spring system of Figure 14.3
(b). You may think of the frame as consisting of a slab of mass supported by two elastic columns of negligible mass (compared to the slab).
The spring constant depends only on the geometric and material properties of the columns. The frame is subjected to a linearly decreasing
blast loading that originates at and lasts for seconds (See Figure 14.3(c).) Determine the response of the frame at any time from the
blast input. We assume that the building is initially at rest.

Figure 14.3

Solution: The ODE for the motion of the structure is

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 14 of 19

where the expression for is obtained from the geometry of Figure 14.3(c). Set The normalized ODE is then

The characteristic equation for the corresponding homogeneous ODE, is so the characteristic roots are It
follows from Table 14.1 that the Green's kernel is

Thus a particular solution is given by

Because of the piecewise nature of in order to evaluate the integral in Eqn. (14.15) we must consider two cases: and We
can ignore because the structure is assumed at rest prior to

Solution for
According to Eqn. (14.15), we have

Upon integrating by parts (or using Maple), we obtain

Upon simplification we get

Solution for
Since when the upper limit of integration in Eqn. (14.15) can be replaced by Thus the solution can be computed from Eqn.
(14.16) by setting in all terms within the square brackets. Consequently,

Upon simplification we get

In summary,

To solve the IVP we first combine the complementary solution with the particular solution to get the general solution to Eqn. (14.14):

Compute since we calculate and from the values of and at we need only need the first piece of that is

Then

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 15 of 19

So

Letting in Eqn. (14.19) yields

so that

Letting in Eqn. (14.20) yields

so that

as well. It follows that the solution to the IVP is

The solution is plotted in Figure 14.4 below where and

Figure 14.4: Solution to

where

and

End of Example 14.18

As the plot in Figure 14.3 illustrates, the oscillation of the building continues indefinitely after the blast loading ends at
A more realistic model incorporates some damping. In particular, introduce the damping in Eqn. (14.14) to get the
new model

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 16 of 19

Rather than solve this ODE analytically we use Maple to plot its solution. The result is displayed in Figure 14.5.

Figure 14.5: Solution to

where

and

A more elegant but just as complicated a solution to piecewise forced ODEs is presented in Lecture 16 using convolution. A
totally different approach is presented in Lecture 20 where we use Laplace transforms to solve ODEs with piecewise
continuous forcing functions.

The result in the following example was obtained in Example 13.8. Note the the solution below is much (much) more
complicated than the (UC) method used in Example 13.8. We include these laborous calculations to demonstrate, that although
elegant, the Green's kernel solution is not practical when the UC method will do.

Example 14.19
Use the Green's kernel to determine a particular solution for the ODE

where

Solution: The Green's kernel for can be read off from Table 14.2:

Without loss of generality we can set since is defined and continuous at all Then use Eqn. (14.11) to obtain

Hence

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 17 of 19

where

The evaluation of and requires the use of some trigonometric identities, namely,

Set and Then and .so we get

Then

Another (more familiar, I hope) trigonometric identity gives us:

so that

Likewise

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 18 of 19

Yet another trigonometric identity yields:

so that

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009
L14.htm Page 19 of 19

Then

End of Example 14.19

file://F:\OnLineODE\Lectures\Lec14\L14.htm 4/16/2009

Вам также может понравиться