0% found this document useful (0 votes)
88 views12 pages

Linear ODES

This document discusses linear systems of first-order ordinary differential equations (ODEs). It begins by proving existence and uniqueness theorems for matrix differential equations. It then applies these results to systems of first-order ODEs. Key points include: 1. A system of first-order ODEs x' = Ax + f has a unique solution for given initial conditions. 2. Solutions to the associated homogeneous system x' = Ax form a vector space. Any basis of this space is a fundamental set of solutions. 3. Solutions to the nonhomogeneous system are the sum of a particular solution and a solution to the homogeneous system. 4. Variation of parameters provides a method to solve the non

Uploaded by

Adi Subbu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
88 views12 pages

Linear ODES

This document discusses linear systems of first-order ordinary differential equations (ODEs). It begins by proving existence and uniqueness theorems for matrix differential equations. It then applies these results to systems of first-order ODEs. Key points include: 1. A system of first-order ODEs x' = Ax + f has a unique solution for given initial conditions. 2. Solutions to the associated homogeneous system x' = Ax form a vector space. Any basis of this space is a fundamental set of solutions. 3. Solutions to the nonhomogeneous system are the sum of a particular solution and a solution to the homogeneous system. 4. Variation of parameters provides a method to solve the non

Uploaded by

Adi Subbu
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 12

Linear ODE: Systems of First Order and n-th Order

S. Kumaresan
School of Math. and Stat.
University of Hyderabad
Hyderabad 500046
[email protected]

A Fundamental Existence Theorem for Matrix Equations

Theorem 1. Let J := [t0 , t1 ] R be an interval. Let A, F : J M (n, R) be continuous n n


matrix valued functions. Then the matrix DE
X 0 (t) = A(t)X(t) + F (t),

X(t0 ) = X0 ,

(1)

for a given matrix X0 has a unique solution on the entire interval J.


Proof. The given problem is equivalent to solving the matrix valued integral equation
Z t
X(t) = X0 +
(A(s)X(s) + F (s)) ds.

(2)

t0

We adopt the Picard iteration scheme:


Z
X0 (s) := X0 ,

Xm (t) := X0 +

(A(s)Xm1 (s) + F (s)) ds.


t0

Let M > 0 be such that kA(s)k M for s J. We prove by induction that


kXm (t) Xm1 (t)k

Mm
(t t0 )m ,
m!

(3)

for all t J and m N. As in the case of Picards local existence theorem, we conclude that
(Xm ) is uniformly Cauchy on J and hence they converge to an X which satisfies the integral
equation (2).
We now prove that the solution is unique. Let X1 and X2 be solutions of the DE (1).
Then their difference Y := X1 X2 is a solution of the homogeneous problem IV problem
Y 0 (t) = A(t)Y (t),

Y (t0 ) = 0.

(4)

If we set g(t) :=

Rt
t0

kY (s)k ds, we see that

Z t



0
g (t) = kY (t)k = kY (t) Y (t0 )k =
Y (s)


t0
Z t
0
Y (s) ds

t0
t

kA(s)Y (s)k ds

=
t0

kY (s)k
t0

= M g(t).
Hence we have
g 0 (t) M g(t) 0,

(t J).

But then

d
(g(t)eM t ) = [g 0 (t) M g(t)]eM t 0,
(t J),
dt
so that the function t 7 g(t)eM t is decreasing. Since g(t0 ) = 0 and is nonnegative by
definition, it follows that g = 0 on J.
A matrix DE X 0 (t) = A(t)X(t) + F (t) is said to be homogeneous if F = 0.

Theorem 2. Let the notation be as in the last theorem. Let X be a solution of (1). The
Wronskian W (t) := det X(t) of the solution satisfies the scalar DE
W 0 (t) = (Tr A(t))W (t).
This implies that if X( ) is invertible for some J, then X(t) is invertible for all t J.
Proof. We use Ex. 3 given below. Let x1 (t), . . . , xn (t) be the columns of X(t). We have
0

W (t) =

n
X

det(x1 , . . . , x0k (t), . . . , xn (t))

k=1

n
X

det(x1 , . . . , A(t)xk (t), . . . , xn (t))

k=1

= Tr(A(t)) det(x1 (t), . . . , xn (t))


= Tr(A(t))W (t).
Rt

Tr A(s) ds

Integrating this DE, we find that W (t) = ce t0


for the constant c = W (t0 ). The last
statement of the theorem follows from this observation.
Ex. 3. Let xi : J Rn be differentiable functions. Let X(t) := (x1 (t), . . . , xn (t)) where xj
is considered as the j-th column of X. Show that
n
X
d
det(X(t)) =
det(x1 (t), . . . , x0k (t), . . . , xn (t)).
dt
k=1

Hint: Recall that det is a multilinear function. The derivative can be found from the first
principles. Or, use the Laplace expansion.
2

Linear Systems: x0 = Ax + f

We use the above theorem to deduce all the results concerning the first order system
x0 (t) = A(t)x(t) + f (t),
where A : J M (n, R) and f : J Rn are continuous functions. Written componentwise,
we have
x0i (t) = ai1 (t)x1 (t) + + ain (t)xn (t) + fi (t), (1 i n).
Theorem 4. Let J := [t0 , t1 ] R be an interval. let A : J M (n, R) and f : J Rn be
continuous functions. Consider the nonhomogeneous first order system
x0 (t) = A(t)x(t) + f (t)

(5)

and the corresponding homogeneous system


x0 (t) = A(t)x(t)

(6)

(a) For any given vector x0 Rn , the system (5) has a unique solution with x(t0 ) = x0 .
(b) If x1 (t), . . . , xn (t) are n solutions of (6), then the following are equivalent.
(i) x1 ., . . . , xn are linearly dependent on J.
(ii) det(x1 (t), . . . , xn (t)) = 0 for all t J.
(iii) det(x1 (t), . . . , xn (t)) = 0 for some t J.
(c) The solution set of the homogeneous system x0 (t) = A(t)x(t) is a vector space of dimension n. Any basis of the space of solutions is known as a fundamental set.
(d) If xp is a particular solution of the nonhomogeneous system (5), then any solution is
of the form xp + x where x is a solution of the associated homogeneous system.
(e) Let (x1 , . . . , xn ) be a fundamental set of the homogeneous system. Let X(t) := (x1 (t), . . . , xn (t))
be the matrix whose i-th column is xi (t). If c(t) is any solution of the equation
X(t)c0 (t) = f (t)
then
x(t) := X(t)c(t)
is a solution of the nonhomogeneous equation.
(f) Let (x1 , . . . , xn ) be a fundamental set for the homogeneous system. Assume that uk are
real valued functions such that
u0k (t) =

det(x1 (t), . . . , xk1 (t), f (t), xk+1 (t), . . . , xn (t))


.
det(x1 (t), . . . , xn (t))

Then x(t) := u1 (t)x1 (t) + + un (t)xn (t) is a solution of the nonhomogeneous system (5).
Proof. (a) follows trivially, if we apply Theorem 1 to X0 := (x0 , . . . , x0 ) and F (t) := (f (t), . . . , f (t)).
Then X(t) is of the form X(t) = (x(t), . . . , x(t)).
(b) Let X(t) := (x1 (t), . . . , xn (t)). Then x1 , . . . , xn are solutions of x0 = Ax iff X is a
solution of X 0 = AX.
3

(c) Let xk be the unique solution of x0 = Ax with x(t0 ) = ek , the k-th basic vector of Rn .
(Existence
P of xk is assured by (a). )PLet y be any solution of the homogeneous system. Then
y(t0 ) = k ck ek . If we let x(t) := k ck xk , then x is a solution of the homogeneous system
such that x(t0 ) = y(t0 ). By the uniqueness, it follows that y = x.
(d) is trivial, if we observe that the difference of any two solutions of the nonhomogeneous
system is a solution of the homogeneous system.
(e) If x(t) = X(t)c(t), then by the product rule, we have
x0 (t) = X 0 (t)c(t) + X(t)c0 (t) = A(t)X(t)c0 (t) + f (t) = A(t)x(t) + f (t).
(f) For any fixed t J, the vector c(t) := (c1 (t), . . . , cn (t)) (considered as a column vector)
is, by Cramers rule, the solution of the linear system X(t)c0 (t) = f (t). Hence t 7 X(t)c(t)
is a solution of the nonhomogeneous equation (1) by (e). But,
X(t)c(t) = c1 (t)x1 (t) + + cn (t)xn (t),
by the definition of matrix multiplication.
Remark 5. Note that the equation X(t)c0 (t) = f (t)
of the above theorem says that
R t1in (e) 1
1
c is an anti-derivative of X(t) f (t), e.g., c(t) = t0 X(s) f (s) ds. Thus (f) of the last
theorem allows us to solve for the nonhomogeneous system given a fundamental set for the
homogeneous system. This is known as the method of variation of parameters.
If the matrix function A(t) = A is a constant matrix, we then have an explicit representation of the solution of the non-homogeneous equation.
Theorem 6. Let A be a constant matrix. Let J = [t0 , t1 ] R be an interval. Let f : J Rn
be continuous. Then the unique solution of the initial value problem
x0 (t) = Ax(t) + f (t),

x(t0 ) = x0

is given by
x(t) = e(tt0 )A x0 +

e(ts)A f (s) ds.

t0

Proof. Multiply the DE by etA on the left to obtain


etA x0 (t) etA Ax(t) = etA f (t).
The LHS of this equation is the derivative the function t 7 etA x(t). So, upon integration,
we get
Z t1
tA
e
x(t) =
esA f (s) ds + c.
t0

The constant vector is identified by taking t as t0 .


Let A be a constant matrix. We want to solve x0 = Ax with IC x(0) = x0 explicitly. This
P
(tA)k
is easily done. Let x(t) := etA x0 where etA :=
k=0 k! . Then the standard results about
the exponential of matrices tell us that the problem is solved. (See Exer. 7 below.) However,
it is usually very difficult to compute the exponential of any matrix A. Before attending to
this, we establish some useful results.
4

Ex. 7 (Exponential Map in M (n, R)). The following set of exercises introduces the exponential map in M (n, R) and its properties:
1. For X M (n, R), X := (xij ), let
kX k := max |xij |
1i,jn

be the max norm. It is equivalent to the operator norm k k on elements of M (n, R)


viewed as linear operators on Rn . We shall use the operator norm in the following.

2. We have kAB k kAk kB k for all A, B M (n, R) and Ak kAkk .
3. A sequence Ak A in the operator norm if and only if akij aij for all 1 i, j n as
k . Here we have Ak := (akij ), etc.
P
P
4. If
k=0 Ak is convergent to an element A of M (n, R).
k=0 kAk k is convergent, then
P
Xk
5. For any X M (n, R), the series
k=0 k! is convergent. We denote the sum by exp(X)
or by eX .
6. For a fixed X M (n, R) the function f (t) := etX satisfies the matrix differential
equation f 0 (t) = Xf (t), with the initial value f (0) = I. Hint: Note that the (i, j)-th
entry of f (t) is a power series in t and use (4).
7. Set g(t) := etX etX and conclude that etX is invertible for all t R and for all X
M (n, R).
8. There exists a unique solution for f 0 (t) = Af (t) with initial value f (0) = B given by
f (t) = etA B. Hint: If g is any solution, consider h(t) = g(t) etA .
9. Let A, B M (n, R). If AB = BA then we have
eA+B = eA eB = eB eA = eB+A .
Hint: Consider (t) := et(A+B) etA etB .
10. For A, X M (n, R) we have eAXA

= AeX A1 .

Definition 8. We say that a matrix X(t) with columns xi (t) is a fundamental matrix if
{xi : 1 i n} is a basis of solutions of the DE x0 = Ax. Note that by the fundamental
theorem for linear systems such a matrix exists.
Lemma 9. A matrix X(t) is a fundamental matrix for x0 = Ax iff X 0 (t) = AX(t) and
det X(t) 6= 0.
Proof. Let xj be the j-th column of X. Observe that the matrix equation X 0 = AX is
equivalent the n vector equations x0j (t) = Axj (t). By the standard uniqueness argument, the
n solutions xj (t) are linearly independent iff x1 (0), . . . , xn (0) are linearly independent. The
latter are linearly independent iff det X(0) 6= 0.
Lemma 10. The matrix etA is a fundamental solution of x0 = Ax.
5

Proof. Obvious in view of the last lemma.


Lemma 11. Let X(t) and Y (t) be two fundamental solutions of x0 = Ax. Then there exists
a constant matrix C such that Y (t) = X(t)C.
Proof. Each column yi of Y can be written as a linear combination of the columns xj of X:
yi = c1i x1 + + cni xn .
Then C := (cij ) is as required.
Theorem 12. Let X(t) be a fundamental matrix of x0 = Ax. Then
etA = X(t)X 1 (0).

(7)

In other words, any fundamental matrix X(t) is of the form X(t) = etA X(0).
Proof. Immediate consequence of the last two lemmas.
Ex. 13. Let xj be the solution of the initial value problem x0 = Ax with xj (0) = ej . Show
that etA = (x1 , . . . , xn ).
Ex. 14. Let X and Y be fundamental matrices of x0 = Ax with Y = XC for a constant
matrix C. Show that det C 6= 0.
Ex. 15. Let X(t) be a fundamental matrix of x0 = Ax and C a constant matrix with
det C 6= 0. Show that Y (t) = X(t)C is a fundamental matrix of x0 = Ax.
Ex. 16. Let X be a fundamental solution of x0 = Ax. Prove that the solution of the IV
problem x0 = Ax, x(t0 ) = x0 is x(t) = X(t)(X(t0 )1 x0 .
Ex. 17. Let X be a fundamental matrix of x0 = Ax. Show that X(t)X(t0 )1 = e(tt0 )A .
Theorem 18 (Nonhomogeneous Equation-Variation of Parameters). The solution of the IV
problem x0 = Ax + f (t), x(t0 ) = x0 is given by
x(t) = X(t)X 1 (t)x0 + X(t)

X 1 (s)f (s) ds,

(8)

t0

where X is a any fundamental matrix of the homogeneous equation x0 = Ax.


Proof. Let x1 , . . . , xn be a set of n linearly independent solutions of the homogeneous system
x0 = Ax. We seek a solution x of the IV problem for the nonhomogeneous system in the form
x(t) = u1 (t)x1 (t) + + un (t)xn (t).
This can be written as x(t) = X(t)u(t) in an obvious notation. Assuming that x(t) solves the
IV problem and plucking the expression for x(t) in the equation x0 = Ax + f , we get
X 0 (t)u(t) + X(t)u0 (t) = AX(t)u(t) + f (t).

(9)

Since X is a fundamental matrix X 0 (t) = AX(t) so that the first terms on either side of (9)
are equal. Hence, (9) reduces to
X(t)u0 (t) = f (t).
Thus, u0 = X 1 (t)f (t) so that
Z

u(t) = u(t0 ) +

X 1 (s)f (s) ds

t0

= X

(t0 )(x0 ) +

X 1 (s)f (s) ds.

t0

The result (8) follows from this.


Remark 19. Theorem 6 is a special case of the last theorem if we take X(t) = etA . Note
that the Greens kernel in this case is G(t, s) = e(ts)A .

Linear Equations of Higher Order

An n-th order linear ODE is of the form


y (n) + an1 y (n1) + + a1 y 0 + a0 y = f (t),

(10)

where the coefficient function aj and f are assumed to be continuous functions on an interval
J R. The homogeneous linear equation associated to (10) is
y (n) + an1 y (n1) + + a1 y 0 + a0 y = 0.

(11)

The crucial observation is that the study of such equations can be reduced to the study
of first order systems considered above. If y is a solution of the DE (10), then the functions
x1 := y, x2 := y 0 , x3 := y 00 , . . . , xn1 = y (n2) , xn := y (n1)
satisfy the following differential equations
x01 = x2 , x02 = x3 , . . . , x0n1 = xn , x0n = (an1 xn + an2 xn1 + + a1 x2 + a0 x1 ) + f. (12)
We introduce the matrix valued function

0
1
0
0
0
1

..
.
..
.
A(t) := .
.
.

0
0
0
a0 a1 . . .

...
...
..
.

0
0
..
.

...
1
. . . an1

With this notation, the differential equations in (12), can be recast as

0
x1 (t)
0
x1 (t)
x0 (t)
x2 (t) 0

. .
..
. = A(t) .. + ..

xn1 (t)
xn1 (t) 0
xn (t)
f (t)
x0n (t)
7

or, in an obvious notation


x0 (t) = A(t)x(t) + F (t).

(13)

Ex. 20. Every solution y of (10) is a solution of (13). Conversely, if x(t) is a solution of (13),
then y(t) := x1 (t) is a solution of (10).
The matrix A above is called the companion matrix of DE (10).
The following theorem is more or less an immediate consequence of Theorem 4.
Theorem 21. Let J := [t0 , t1 ] R be an interval. Let a0 , a1 , . . . , an1 : J R and f : J R
be continuous. Let L(y) := y (n) (t) + an1 y (n1) (t) + + a1 (t)y 0 + a0 (t)y(t). Consider the
nonhomogeneous equation L(y) = f and the homogeneous equation L(y) = 0.
(a) Given any numbers j R, for 0 j k 1, there is a unique solution y of L(y) = f
with y (j) (t0 ) = j .
(b) If y1 , . . . , yn are solutions of the homogeneous equation L(y) = 0, then their Wronskian
determinant

y1 (t)
y2 (t)
...
yn (t)
y10 (t)
y20 (t)
...
yn0 (t)

W (y1 , . . . , yn )(t) :=
..
..
..

.
.
.
(n1)

y1

(n1)

(t) y2

(n1)

(t) . . . yn

(t).

satisfies the differential equation: W 0 (t) = an1 (t)W (t).


(c) If y1 , . . . , yn are n solutions of the homogeneous equation L(y) = 0, then the following
are equivalent:
(i) y1 , . . . , yn are linearly independent on J.
(ii) W (y)(t) = 0 for all t J.
(iii) W (y)( ) = 0 for some J.
(d) The set of solutions of the homogeneous equation L(y) = 0 is an n-dimensional vector
space. Any basis of the space of solutions is called a fundamental set.
(e) If yp is a particular solution of the nonhomogeneous system, then any solution of the
nonhomogeneous system is of the form yp +y where y is a solution of the homogeneous system.
(f) Let {yj : 1 j n} be a fundamental set of the space of solutions of L(y) = 0. If
u1 , . . . , un are solutions of the matrix equation

0
y1
y2
...
yn
u1 (t)
0
y10
0

y20
...
yn0

u2 (t) 0
=
..
.. .. ..
..
.
.
. . .
(n1)

y1

(n1)

y2

(n1)

. . . yn

u0n (t)

then
y(t) := u1 (t)y1 (t) + + un (t)yn (t)
is a solution of the nonhomogeneous equation L(y) = f .

f (t)

Let D : f 7 f 0 denote the differential operator. If p(X) := X n +an1 X n1 + +a1 X +a0 ,


then we let
p(D)(y) := y (n) + an1 y (n1) + + a1 y 0 + a0 (t)y.
The polynomial p is called the characteristic polynomial of the differential operator p(D).
Theorem 22. Given a differential operator
p(D)(y)(t) := y (n) (t) + an1 y (n1) (t) + + a1 y 0 (t) + a0 y(t)
with constant coefficients. Let j be the roots of p with multiplicity mj , (1 j k). Any
solution y of the homogeneous equation p(D)y = 0 is of the form
y(t) = e1 t p1 (t) + + ek t pk (t)
where pj is an arbitrary polynomial of degree at most mj . That is, a basis of the solution is
{ej t tr : 1 j k, 0 r mj 1}.
Proof. Follows from the next three lemmas.
Lemma 23. Let Vr := C r (R) be the vector space of all r-times continuously differentiable
functions on R. Let D : Vr Vr1 be the derivation map f 7 f 0 .
(a) If f Vr , then (D I)r f = et Dr (et f ).
(b) A function f Vr lies in the kernel of (D I)r iff it is of the form
f (t) = et (b0 + b1 t + + br1 tr1 ).
Proof. (a) is proved by induction. To prove (b), observe that (DI)r f = 0 iff Dr (et f ) = 0,
by (a).
Lemma 24. Let p K[X] be a polynomial over a field K with a decomposition p = p1 pk
where pj are relatively prime. Let A : V V be a linear endomorphism of the K-vector space
V . Then we have
ker p(A) = ker p1 (A) ker pk (A).
Proof. Let qj := p/pj = p1 pj1 pj+1 pk . We first show that the sum is direct. Let
v1 + + vk = 0 where vj ker pj (A). Note that qi (A)vj = 0 whenever i 6= j. Since pi and
qi are relatively prime, there exist polynomials a and b such that ap1 + bq1 = 1. Then,
vi = a(A)pi (A)vi + b(A)bqi (A)vi
X
= b(A)qi (A)(
vj )
j6=i

b(A)qi (A)vj

j6=i

0.

j6=i

In the displayed equation, the inclusion is obvious. To prove the reverse


P inclusion, note
that qj are relatively prime. Hence there exist polynomials rj such that j rj qj = 1. Let
v ker p(A). Then vi := ri (A)qi (A)v ker pi (A) because
pi (A)vi = pi (A)qri (A)qi (A)v = ri (A)pi (A)qi (A) = ri (A)p(A)v = 0.
P
It is easily seen that v = i vi .
Lemma 25. Let p() := n + an1 n1 + + a1 + a0 = ( 1 )m1 ( glk )mk . Then a
basis solutions of the homogeneous equation p(D)y = 0 is {ej t tr : 1 j k, 0 r mj 1}.
Proof. Let pi () := ( i )mi . Then p(D) = p1 (D) pk (D). The result follows from the
last two lemmas.
We now apply our knowledge about the n-th order homogeneous equation to compute etA
of a matrix A!
Theorem 26. Let p be the characteristic polynomial of A. Let y1 , . . . , yn be a basis for the
set of solutions of the homogeneous n-th order equation p(D)y = 0. Then there exist matrices
A1 , . . . , An such that
etA = y1 (t)A1 + + yn (t)An .
(14)
Proof. We have p(t) := det(A tI). Let etA = (uij (t)). Then,
(p(D)uij (t)) = p(D)etA = p(A)etA = 0,
by Cayley-Hamilton theorem. Thus every entry uij of etA satisfies the DE p(D)uij = 0, hence
can be written as linear combination of yj s.
Remark 27. We now give an algorithm to find the exponential of a matrix. Given a matrix
A, we find its characteristic polynomial p(t). We find a basis {yj : 1 j n} of solutions
of p(D)y = 0. From the last result we know that there exist matrices Aj such that etA =
y1 A1 + + yn An . To find these matrices Aj , we differentiate the equation (14) n 1 times
with respect to t. We thus obtain n equations a follows.
etA = y1 (t)A1 + + yn (t)An
AetA = y10 (t)A1 + + yn0 (t)An
..
.
An1 etA = y (n1) (t)A1 + + y (n1) (t)An .
Evaluate them at t = 0 and solve for Aj by the standard Gaussian elimination method.




0 s
cos st sin st
tA
Example 28. Consider
. Following the above steps, we obtain e =
.
sin st
cos st
s
0


0
1
tA
Ex. 29. (i) Find e where A =
.
14 9
(ii) Transform the equation y 00 + 9y 0 + 14y = 0 with y(0) = 0 and y 0 (0) = 1 into a vector
10

DE.
(iii) Solve the IV problem in (ii). Ans: etA =

1
5

7e2t 2e7t
e2t e7t
2t
7t
14e
+ 14e
2e2t + 7e7t


and

y(t) = 15 (e2t + e7t ).


Ex. 30. Solve x00 + x = 3 with x() = 1 and x0 () = 2 using the methods of exponential
matrix and variation of parameters. Ans: 3 + 2 cos t 2 sin t.
Ex. 31. Solve x00 + 2x0 8x = et with x(0) = 1 and x0 (0) = 4 using the methods of
31 4t
exponential matrix and variation of parameters. Ans: 30
e
+ 16 e2t 15 et .
Ex. 32. Compute the exponential of the following matrices by any



 1 t
(e + e3t ) 41 (e3t et )
1 1
2
.
(i)
. Ans:
et + e3t 21 (et + e3t )
4 1


 t

1 1
e cos t et sin t
(ii)
. Ans: t
.
1 1
e sin t et cos t

2 1 0
(iii) 0 2 0.
0 0 2




1 + 2t
t
0
1
2t
.
. Ans: e
(iv)
4t 1 2t
4 4

2
Ex. 33. Compute etA by solving a third order DE where A = 1
1

method:

0 1
0 1 .
2 0

We now briefly indicate how one can solve a linear system by the eigen value-eigen vector
method. Given the system x0 = Ax with a constant matrix A. Assume that A has distinct
eigen values, say, j , 1 j n. Let vj be a nonzero eigen vector of A with eigen value
j . Then xj (t) := ej t vj is a solution. Also, {xj : 1 j n} is a fundamental set for the
equation x0 = Ax. Even if the eigen values are complex, by Ex. 35, we can find real valued
solutions. However, if A has eigen values with multiplicity, one requires a little more work.
We refer the reader to Brauns book (especially sections 3.63.8) for more details.
Remark 34. A fall-out of this approach is another method of computing etA . Let x1 (t), . . . , xn (t)
be linearly independent solutions of x0 = Ax. Then the matrix X(t) := (x1 (t), . . . , xn (t)) is a
fundamental matrix so that etA = X(t)X 1 (0). (Ex. 17!)
Ex. 35. Let x(t) = f (t) + ig(t) be a solution of x0 = Ax. Then f and gare real valued
solutions of x0 = Ax.
Ex. 36. Find all the solutions of x0 = Ax using eigen methods where

3 2 4
(i) A = 2 0 2 x.
4 2 3

1 0 0
(ii) A = 3 1 2.
2 2 1
11

Ex. 37. Solve the given initial value problem.




 
1 1
2
0
(i) x =
x, x(0) =
.
4 1
3


3 0 2
0
(ii) x0 = 1 1 0 x with x(0) = 1.
2 1 0
2
Ex. 38. Solve the IV problem

1 0 0
0
x0 = 2 1 2 x + 0 ,
3 2 1
et cos 2t
by two different methods viz.,

1
Ex. 39. Find etA if A = 0
0
is easier?

x(0) = e2 + e3

by finding the exponential of A and by eigen method.

1 1
3 2 by methods of Remarks 27 and 34. Which do you find
0 5

12

You might also like