CH.6 EIGENVALUES
(8) Show that
aFp=84+h md Boaz
s of AB and 3
(b) Show that the (. f) em
4%, be eigenvectors of an
ree of R* spanned
., show that Ax € S
19, Let B = S-¥AS and let x be
eigenvalue A. Show
20. Show 1 ices A and B have a
x (but not necessarily a common eigenvalue). then x W
eigenvector of any matrix ofthe forin C = a + BB.
21. Lot be an x matrix and let 2 be a nonzero eigenvalue of A, Show
is an eigenvector belonging 04, the
fF aamal oot Ata?
je matrix with eigenvalues Ay.
(04, for each i
all add up toa fixed constant 3, Show
eigenvector of A” belonging t0 is.
and y are orthogonal
25. Let A and B ben x n matrices, Show that:
(a) 1f2.is a nonzero eigenvalue of AB, then itis also an eigenvalue of
Ba.
(b) IFA = O's an cigenvalue of AB, then % = Dis also an eigenvalue
of Bi.
26. Prove that there do not exist x n matrices A and B sucl
AB ~BA=.
Hi
: See Exercises 8 and 25.)
27. Let pd) = (1) @" — ayaa! =
degree n > {and let
‘Show that if2; is avoot of pli
with eigenvector x = (> ‘
by Use pare (a) to show that if p(X) has distinct roots Ay. As
‘then p(A) isthe characteristic polynomial of C.
‘The matrix C is called the companion
ris of pl).
28, The result given in Exercise 27(b) holds even if al
of pCa) are not distinct. Prove this as follows:
fa) Let
{by Show that
det — AL) = (ayy — A)(-i)™" — det,
= pa)
WB systems OF LINEAR DIFFERENTIAL EQUATIONS
important role in the soltion of systems of linea differ:
wwe see how they are used in the solution of
‘equations with constant coefficients. We begin
by considering systems of firstorder equations of the form
Eigenvalues play a
au + ats +
= ayy + aye +292.
chs
18,
19.
20.
24.
22,
25.
26.
EIGENVALUES
(a) Show that
atp=w+h md Baan
(b) Show that the (. /) enties of AB and AB are equal and hene
that
AaB
Let xy. .0+,% be eigenvectors of an x matsix A and let $ be th
subspace of R* spanned by xy, x, ....X,-Show that Sis invariant unde
A (le, show that Ax € S whenever x € 5),
Let B = SAS and let x be an eigenvector of & belonging to an
cigenvalue A. Show that 53
an eigenvector of A belonging 0,
Show that if two x1 matrices A and B have a common eigenvecto.
x (but not necessarily a common eigenvalue), then x will also be an
eigenvector of any matrix ofthe forin C = aa + BB.
Let A be ann x matrix and let. be a nonzero eigenvalue of A, Show
that if is an eigenvector belonging to, then xis inthe column space of
‘A. Hence the eigenspace corresponding to isa subspace ofthe column
space of A,
Let (uy. tay. y} be an ot
be scalars. Define
normal basis for RY and Tet &yy...-y
As inal b dougul ++ bau
Show that 4 is a symmetric matuix with eigenvalues Ay. Aa.
that u, isan eigenvector belonging to 2, for each i
= Let be a matrix whose columns al sdd up toa fixed constant 5, Show
that dis an efgenvalue of A,
Let Ay and Ay be distinct eigenvalues of A. Let x be an eigenvector of
A belonging to ay and let y be an eigenvector of A” belonging (0 ia,
Show that x and y are orthogonal,
Let A and B ben x 1 matrices. Show that:
(a) 1f2.is a nonzero eigenvalue of AB, then itis also an eigenvalue of
Ba.
(w) 1.
of BA.
Ois an cigenvalue of AB, thea %
is also an eigenvalue
Prove that there do not exist m x m matrices A and B such that
AB BA
(Hint: See Exercises 8 and 25.)
SEC.2 SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 293
27. Let p(d) = (-Q" =,
degree n > | and let
(At-l =+—aiA ag) bea polynomial of
00
00
oa to
(a) Show that i, is a root of pA) = 0. then 2, is an eigenvalue of C
with eigenvector x = QJ1 a8, HT
by Use part (a) to show that if p(X) has distinet roots hy.
then p(A) isthe characteristic polyiomial of C.
‘The matrix C is called the companion matrix of (2)
28, The result given in Exercise 27(b) holds even if al of the eigenvalues
of pCi) are not distinct, Prove this as follows:
(ay Let
a a
00
Loa ‘
: ‘an se mathemati
det De Q)) = (Dai gah! bo bah + a)
(b1 Show that
det(C = AL) = (04-4 — 4)! — det Dy-a) = pla)
IB svsTEMs OF LINEAR DIFFERENTIAL EQUATIONS
Ejgenvalues play an important role inthe solution of systems of linear differ-
ential equations, In this section we see how they’ are used
systems of linear differential equations with constant coeffivients, We begin
by considering systems of first-order equations of the form
Y= auyy + anyy $s + ty
++ ay
Moun + ony +
1
ee294
C6 eIcenvatues
If we tet ¥ =
the system ean be written in
‘any function ofthe form
(can arbitrary constant)
satisfies this equation, A natural gener
isto uke
wherex = (6.
oF this form does
‘work we compu
Visita ay
Nos
if we choose 2 to be an eigenvalue of A and x to be an eigenvector
belonging to A, then
=Y=¥
ion to 2 is an eigenvalue of 4 snd x is
tor belonging to, on of the system Y" = AY.
her Ais eal or complex. Nove that if Y, and Vane b
to ¥' = AY, then @Y, + BYsis also a solution, since
SeAY, + BAY
= AY, +BY)
the sok
ions toann 5
'
}
|
|
I
—
SEC.2. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 4,
p.228). A problem of the form
Y=AY, Y)=Y¥,
-d an fntial value problem,
EXAMPLE 1. Solve the system
=n tds
=3n +2
am
a=)
Tienieamnarecn ener ats
‘longing to A, and x» = (1,1)! is
‘Thus any vector function of the Form
= =I. Solving (A —A}x = 0
at xy = 3)? is an ef
the system, a
In Example I suppose that we requ
Thus
Y= Gand yy =F wher
rm (822)=(0)
5) = Vand ey = 2, Hence the s
lemis given by
Yedtx, $2ehe,
(tet be
= ae — 20
Two tanks are connected as shown
200 titers of wate
Figure
% __C6 eIcenvatues
where x) = (0) is a funetion in C'fa,b] for exch i. If we let ¥ =
(idee 3)? and Y= (y), yf... 807s the system cam be writen in
the form
yaay
‘Y.and ¥ are both vector functions ofr Let us consider the simplest case first.
‘When 1 = 1, the system is simply
Clearly any function ofthe form
0)
satisfies this equation. A natural generalization ofthis solution forthe ease
n> listo take
(can arbitrary constant)
where x = (6.12... .44)7- To verify that a vector Funetion ofthis form does
‘work we compute the derivative
Vaio
ay
Now, if we choose 4 to be an eigenvalue of and x to be an
belonging to A, then
AY = May = hex = Y=
Hence ¥ isa solution tothe system. Thus iF isan eigenvalte of A and x is
vector belonging to 2, hen ex isa solution af the system Y" = AY,
‘This will be rue whether a is teal or complex. Nove that if ¥, and VY, are both
solutions to ¥" = AY, then @Y; + BYs is also a solution, since
(a) + BY.) =e; + BY,
AY, + BAY
AG@Y, + BY
{1 follows by induction that if Wy... Yy are solutions to Y’ = AY. then any
linear combination e,¥; + ++ 4-¢yY, will also be a solution,
4m general, the solutions to ann 5 1 firstorder system ofthe Form
Y=ay
Will form an sr-dimensional subspace of the veetor spnce of all continuous
‘ector-valued functions. If in addition, we require that Yr) take on a pre
scribed value Vi when? 0. the proble wvea unique solution (see [32].
APPLICATION +
SEC2_SYSTENS OF LEAR DIFERENTIALEQUATONS
P. 228). A problem of the form
Ym Ay, ¥()=¥o
is called an initia value problem,
EXAMPLE 1. Solve the system
3m + dys
3m + 2s
da Ge)
The eigenvalues of A are Ay = 6 and dy
with = dy and A = Aa, we see that xy
belonging to Ay and x» = (1, —1)" is an ei
‘Thus any vector function ofthe form
1, Solving (A —A/)x = 0
(4.3)! is am eigenvector
Verqehtsy +e.e%
a (tee! tee
= (see — eve
{sa solution 1 the system, a
In Example 1 suppose that we require thats
Ths
tee (6
vo=(3)78)-(1)
andi follows hat ey = 1 andy = 2 Hence the solution he inv
prolem ssn by
Y
Mx, Dee
=(sis20)
SIDTURES
‘Two tanks are connected as shown in Figure 6.2.1. Initially tank A contains
200 titers of water in which 60 grams of salt has been dissolved. and tank B
contains 200 liters of pure water. Liquid is pumped in and out of the wo tanks
ar rates shown in the diggram. Determine the ammount of salt in each tank atSEC.2 SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS 297
a-(4 3} (9)
d
i CH.5 EIGENVALUES
Water _Miwure where
isu Gain
‘The eigenvalues of ae iy = —
corresponding eigenvec-
When: =0.¥ = Yo
cum tex = Yo
‘and we can find c, and e: by solving,
(+ 2)(2)-(@)
FIGURE 6.21
30er 9 4306"
COMPLEX BIG": UES
rate itis being pumped out For tank the rate at which the salts ‘ at ibe arealn x ‘complex eigenvalue A = bi na let
ven by
xbe an eigenvector
and imaginary parts,
longing to 2. The vector x can be spl
wold, Ne 80 yin
sum (20 0
0 Rex) +i Rem) /Imsi
salts being pomped out is ca [Restiims | [Res |
(to :
oun) (Ay) = sin mati unse) Next) Nin
the rte of change for tank A i given by Since enties ofA ae ale, it follows that == a —bi
nO _ my of A with eigenvector
= Rex—ilmx
Oni) _ ih
NO= Soy" ~ “300 = 18. ~ T0
(0) and y(t) re mst solve protien
Y=ay, ¥Q)=¥oH.6 EIGENVALUES
Water
131L/min
Minture
(Vain
201 min
FIGURE 6.2.1
SOLUTION, Let ys(t) and y3(7) be the number of grams of salt in tanks A
and B. respectively, at time t. Initially.
-(s8))=(¢)
‘The colal amount of liquid in each tank will remain at 200.titers sine Le
amount being pumped in equals the amount being pumped out. The rate of
change in the amount of salt for each tank is equal tothe rate i is being add
‘minus the rate itis being pumped out, For tank the rate at which the salts
added is given by
ow (Yat)
« unis (B2an)
20.Limin) (ex)
‘Thus the rate of change for tank A is given by
Yo)
nit)
AO ini
10
0)
407 10
y, for lank B, the rte of change is given by
i) 20m) _ itt) _ yt)
nee 0)
To determine »(#) and ya(t) we must solve the initial valve problem
Yay, ¥@)=Y¥.
Ko
Simi
xe
COMPLEX EIG=
SEC.2. SYSTEMS OFLINEAR DIFFERENTIAL EQUATIONS 297
-(2)
ih corresponding eigenvee-
(4) = 0)
The solution must then be of the form
where
‘The eigenvalues of are iy
tors
Ya ces be:
When: = 0, ¥ = Yo. Thus
em tems
and we can find ¢; and ¢; by solving
(4 )(2)-(8)
‘The solution to this system is ¢, = ¢:
initial value problem is
TO) 3067874. 30e7128
wo=(0B)=(Bestarm) a
30. Therefore, the solution to te
“/ALUES
+ bi andlet
up imc its real
Let A bea real n xn matrix with a complex eigenvalue A
x be an eigensector belonging to 2. The vector x can be spl
‘nd imaginary pars
Rex +i Imn, Rew Ima,
Rex: téima: |_| Rex
mas
fa Rex +i tmx
Rex, tilmy) — \Rex, Im
Since entries of A are all real, it follows that k= a ~bi is also an ef
of A with eigenvector
Rex, =i lmsy
Rexy—fImss
=Rex-ilmx
Rex, —ilmay
xe both solutions to the first-order system Y"
\ay linear combination ofthese two solutions will also be a solution. Thus if
and hence ex and e298 GILG EIGENVALUES
Yi = Hex es) = Relea)
n= ex SR) = Ime")
‘Taking the real and imaginary parts of
= eM (cos bt +i sin bu)(Rex + /Imx)
we see that
Yi =e [(e0sbr) Rex ~ (Ginbt) Ima)
Vz = et ((cosb/)Imx + (sinbr) Rex)
EXAMPLE 2, Solve the system
*=(23)
The eigenvalues of A ae 2 = 24 i and X = 2~i with eigenvector
X= (LUD and X= (1,1 — 0", respec
( e*(cost + isint)
eMcost + isin \(L-+0)
= (ote a tier
Let
vmmnon (a fe)
Bie (rere)
Yay tes
on to the system.
vector fimetions Yy and Yy are real-valued solutions to Y= AY,
a
SS {
SEC 2. SYSTEMS OF LIVEAR DIFFERENTIAL EQUATIONS: Soy
the » x2 coefficient matrix A of the system ¥! = AY
\lependent eigenvector
rathave been presented,
HIGHER-ORDER SYSTEMS
Given a second-order system af the form
Yaayeay”
10 a first-order system by seting
Hewete
and
thea
Ov + Ys
and
Yeas, $az
give the 2n x 20
Gis)
the values of Y, = Vand ¥p = Y" ae specified w
valve problem wil have a unique solution
order system
= O.then the initial
EXAMPLE 3. Solve
valve problem
nt mt ye
Sy t2y + 5yj—298
CH.6 EIGENVALUES
(ext FR = Re(e')
(tc ~ oR) = Imes)
then the vector functions ¥ and Y; are real-valued solutions to ¥" = AY,
‘Taking the real and imaginary patts of
= eM (cos bt + i sin br)(Rex + Imx)
wwe see that
(eos bt) Rex — (sin bt) Im}
€" (cosbr) Tm + (sin bt) Rex}
EXAMPLE 2, Solve the system.
we
SOLUTION, Let
‘The eigenvalues of A are 2 = 24 i and T
X= (1 1-4A)7 and ¥ = (1.1 1)" respectively.
eign (, MMeost + isin)
Bs (esc avi a)
with eigenvectors
rf eF cost ie sine
e* (cost = sine) + ie% cos: sins)
Let p
are (sic)
and
ina = (¢228,,)
(e081 +sins)
Any linear combination
‘ill be solution ro the system,
SEC.2. SYSTEMS OF LINEAR DIFFERENTIAL EQUATIONS: So
~~
Wake m x0 coefcient matrix A of the system V! = AY has n incarly
independent eigenvectors, the general solution ean be obtained by the methods
thathave been presented, The ease when A hasless than linearly independem,
eigenvectors is more complicated and consequenly will not he coveted in this
book.
HIGHER-ORDER SYSTEMS
Given a second-order system of the form
Ys aveay
‘ve may translate it into a first-order system by setting
Yen) = HO)
we le:
and
thea
x
‘The equations can be combined to give the 2 x 2nfistorder system
Reo: nV y,
(°C. w()
ihe values of Yj = Vand Ya = Y's specified when = O,then the init
‘alu problem will ave unique soltion
EXAMPLE 3. Solve the initial value problem
Qk mt ye
Syyt 2ys + Si — 95
30) = y4(0)
!
i
i
\
}
|
|
|
|(CH. EIGENVALUES
nto ty
+5m- 0
o
0 0
1
A had
Corresponding to these eigenvalues are the eigenvectors
mal 5.
4 3.
“Thus the solution willbe of the form
exe” tere
‘We ean use the
have
ns t0 find c1.¢2.¢3, andy. For f = 0, we
or, equivalently,
general. if we have
YO = AV ALY tee Ag
h-order system ofthe form
where each a isan n x matrix, we can transform it into a fistonder system
by setting
Yo Yee Ye hes Yor
We wil system ofthe form
Oia o\(%
oo! o||
ve
¥
If in addition. we require
= 0. there will be exactly one solution tothe problem.
of A. xis an eigenvector bel
th coot of & and Y = 2x. then
= gMely AY
and
AY = Ax = hex SY
“Therefore, Y = ex is a solution tothe system,
(G areucanion 2: HARWONIC MOTION
In Figure 6.2.2 wo masses are
are fixed
1c masses are free to move horizontally. We will assume that
equilibrium
horizontal displacements of the masses a
respectively. We will assume that there are no retarding forces such as
“Then the only forces acting on mas:
springs 1 and 2. The force from spring 1 wi
FIGURE 6.2.2SEL. states Ur ensern vie
CH.G EIGENVALUES
SOLUTION. Set yy = yj and
In genera. if we have anmth-order system of the form.
“This gives the frstorder system
ae i YOU AEA bot AY”
we i svbere each Ay is an nx n matt, we can tansform it ino afist-one
; by seting
Ment etontn eens -
= 59) 42 +5 = 7
‘ i a \We will end up witha system ofthe Form
“The coefficient matrix for his system, x eer a
0 Ou (pad we oor 0\| %
0 oN 5 Sn ‘ ‘
Aote to)
ae y; 000-1
Aue l Ay Ay ay Ae
It inaddition. we require that ¥, take on specific values when
1 = 0. there will be exactly one solution tothe problem.
‘the system issimply ofthe form ¥""" = AY. it usually not necessary
tointroduce new variables. In thiscase one need only ealeulate the mh roots of
the eigenvalues of A. IF). isan eigenvalue of A. x isan eigenvector belonging
toca is an uth root of 2, and X = e*x. then
ye
(h-Lb=D%,
139"
‘Thus the solution willbe of the form
eymie bene bene beste”
and
‘We ean use the inital conditions to finde). €3,€3, and cy, For = 0, we
have
oxi ter bey toms hbo!
ig ted at Wei 4
es ts \i[6 4 In Figure 6.2.2 two masses are adjoined by springs and the ends A and
ens aie |e # ae fixed. The masses are free to move horizontally. We will assume that the
Bimes) a) Nc -4 ilvee springs are uniform and that initially the system isin dhe equilibrium
position. A force is exerted an the system to set the masses in motion. The
horizontal displacements ofthe masses at time r willbe denoted by y(t) and
‘cir). respectively. We will essume that there are no retarding forces such as
friction. Then the only forces acting on mass ny at time ¢ will be from the
springs | and 2. The force from spring | will be ~kxy and the fore from
“The solution to this system is
the initial value problem is
2,1, 1.0)" and hence the soluticn to
el banet tase"
Therefore,
Wa tte
“td + Se" +
det — et + 30H
mtd — Set + 3 a
FIGURE 6.2.2