You are on page 1of 83

Matrix algebra

Column

 a11 a12 ........ a1 n  Row


 
 a 21 a 22 ........ a2n 
 . . ........ . 
A = (a jk ) =  
 . . ........ . 
 . . ........ . 
a am 2 ........ a mn 
 m1

Row Vector:
A = (a11 a12 ………a1n)

Column vector:

 b11 
 
 b21 
B= 
M
 
b 
 n1 
Square matrix:

 c11 c12 .......... ..... c1n 


 
 c21 c22 .......... ...... . 
 . 
c= 
 . . . . 
 . 
c . .......... ..... cnn 
 n1

“TRACE” = ∑ cjk
j =k

Symmetric matrix:

1 2 3 a23
 
A = 2 5 4
3 4 6 

a32
Skew-symmetric:

 0 2 3
 
A =  − 2 0 4
 − 3 − 4 0
 
Diagonal matrix
 a11 
 
 a22 O 
A= . 
   1 0 0
 O .   
 ann  Ex :  0 2 0 

 0 0 3
Unit matrix “I”  
1 
 
 1 O 
 1 
I= 
 . 
 O . 
 
 1 

Zero matrix

0 0 . . 0
 
0 0 .
0=. . .
 
. . .
0 . . . 0 

Operations
Equality For two matrices of the same
order: A = B if and only if
ajk = bjk

Ex: 1 2 a 3− a 
A =   B =  
3 4  3 2 + 2a 

Multiplication by a number

A = (a jk )
λ − number
 λa11 λa12 .... λa1n 
 
 λa21 λa22 .... λa2 n 
λA = 
. . .... . 
 
 λa . .... λamn 
 m1

 2 3  4 6
   
Ex : 2 ⋅  4 1  =  8 2 
 − 2 0  − 4 0
   
Addition “ A+B ”

 a11 + b11 . . a1n + b1n 


 
 .
A + B = (ajk + bjk ) ⇔

. 
 
a +b . . a +b 
 m1 m1 mn mn 

i) A+B = B+C
ii) A+(B+C) = (A+B)+C
1 2  1 1  2 3
Ex :   +   =  
 3 4   −1 0   2 4 
3 + ( −1) = 2

Subtraction “A - B”

A − B = (a jk − b jk )
Multiplication of Matrices “AB”
 a11 . . a1n   b11 . . b1 p 
   
 . .   . . 
A= B=
. .  . . 
   
a . . amn  b . . amn 
 m1  n1
{ {
n p

 b11 . . . . bn1   b11 . . . . b1 p 


   
 * + L L + *   . . . . . . 
a . . . . a1n   . . . . . . 
AB =  11    =
 . . . . . .   . . . . . . 
 . . . . . .   . . . . . . 
  
a . . . . a mn  b . . . . bnp 
 m1  n1

c11 c1p
} }
 a11b11 + a12 ⋅ b21 + ....+ a1nbn1, .....a11b1p + ....+ a1nbnp 
 
= cm1. cmp.
} } 
 a b + a b + ....+ a b , .....a b + ....+ a b 
 m1 11 m2 21 mn up mn 1p mn np 
AB=C C=(Cjk)

Ex:
 3 5 
 2 1 4  
B =  2 − 1
A =  
 4 2 
 − 3 0 2 

 2 ⋅ 3 + 1 ⋅ 2 + 4 ⋅ 4, 2 ⋅ 5 + 1 ⋅ (−1) + 4 ⋅ 2 
AB =   =
 (−3) ⋅ 3 + 0 ⋅ 2 + 2 ⋅ 4, (−3) ⋅ 5 + 0 ⋅ (−1) + 2 ⋅ 2 

 24 17 
=  
 − 1 − 11

A ⋅ B = C
( m × n )( n × p ) = ( m × p )
Properties of multiplication


! AB BA

A(BC) = (AB)
A(B + C) = AB + AC Be careful
with the
(B + C)A = BA + CA sides

A⋅I = I ⋅A = A
A⋅0 = 0
Transpose of a Matrix " AT "
A= (ajk)

AT= (akj)
 2 1 4
A =  
Ex: −3 0 2 
2 − 3
 
A = 1
T
0 
4 2 

A + ≡ AT
Properties (A + B) T = A T + B T

(AB)T = B T A T
(A T ) = A  2 − 4
 
− 4 3 
AT = A
AT= -A
 1 1 2
 
 −1 3 4
− 2 −4 2 

Multiplication of vectors " X ⋅ R"
 x1   v1 
   
X =  x2  R =  v2 
x  v 
 3  3

X ⋅ R = X ⋅ R ⋅ cos Θ = x1v1 + x2 v2 + x3v3


 v1 
 
X ⋅ R = ( x1 x3 )  v 2  = x1v1 + x 2 v 2 + x 3v 3
T
x2
v 
 3 

Ex:  1 i 1 − i 
A =   A =  
 − i 1 i 1 
b)
If A = AT
A = − AT
Ex:
 5 − 2i  T  5 2i  T  5 − 2i 
A =   A =   A =  
 2i 3   − 2i 3   2i 3 
Determinants

a 11 a 12 .... a 1n
Def (A) ≡ ∆ = a 21

.
.
.
.
.
. . .
a n . .... a nn

How to find det(A)

( )
Minor a11
...
a1n
A= …ajk….
.
an1 .. ann

Minor of ajk= a11 a1n


an1 ann

Ex: 2 −1 1 3
  Minor of a32= 2 1 3
− 3 2 5 0
A= −3 5 0
1 0 −2 2
 
 4 −2 3 1  4 3 1

Cofactor

Ajk ≡ Cofactor of ajk = minor of ajk⋅ (-1)j+k

Ex: 2 1 3
Cofactor of a32= (-1)3+2 − 3 5 0
4 3 1
2 1 3
Ajk= − −3 5 0
4 3 1

n
Det(A)= ∑ ajkAjk
k =1

Ex:
. .
. .
a j1 . . . . ………… . . a j8 . .
. .
. .

det( A ) = a j 1 ⋅ A j1 + ... + a j 8 ⋅ A j 8 + ...


Ex: 2×2
a 11 a 12
= a 11 ⋅ a 22 − a 12 ⋅ a 21
a 21 a 22

3× 3
a11 a12 a13 a 11 ⋅ [a 22 ⋅ a 33 − a 23 ⋅ a 32 ] +
a 21 a 22 a 23 =
( − a 12 ) ⋅ [a 21 ⋅ a 33 − a 23 ⋅ a 31 ] +
a 31 a 32 a 33
+ a 13 ⋅ [a 21 ⋅ a 32 − a 22 ⋅ a 31 ]

Ex: Calculate the determinant of matrix A:


3 1 6
 
A = 0 − 2 3
0 0 4 

det(A) = 3 ⋅ [( − 2 ) ⋅ 4 − 0 ⋅ 3 ] +
( − 1 ) ⋅ [0 ⋅ 4 − 0 ⋅ 3 ] +
6 ⋅ [0 ⋅ 0 − 2 ⋅ 0 ] = − 24

alternatively (using 3rd row):


det( A ) = 4 ⋅ [3 ⋅ ( − 2 ) − 0 ⋅ 1 ] = − 24
Theorems (15-7)

If we multiply the elements of any row (or column)


by a given number and add to corresponding
elements of any other row ( or column), then the
value of the determinant remains the same.
 x1 x2 x3   x1 + ky1 x2 + ky2 x3 + ky3 
   
A =  y1 y2 y 3  B =  y1 y2 y3 
z  z z3 
 1 z2 z 3   1 z2

det( A ) = det( B )
Th: 15-5
If any two rows ( or columns) are the same or
proportional, the determinant is zero.
 x1 x2 x3 
 
A =  kx1 kx2 kx3  multiply by −
1
and add to
z k
 1 z2 z3 

 k k k 
(
 1 x − x1 ) ( x2 − x2 ) ( x3 − x3)  0 0 0
 k k k   
B =  kx1 kx2 kx3  =  kx1 kx2 kx3 
 z1 z z  z z3 
 2 3   1 z2
 
det(A)=det(B) using theorem 15-7
0 0 0
. . .
det(B) = 0 = All zeros in a row
. . .
. . .
 x1 y1 z1 
 
det(A)=0 A =  x2 y2 z2 
x y3 z 3 
 2
x y z
Th. 15-10: vectors
If we can find such λ1, λ2, λ3
so that not all of them = 0 0
 
And: λ1 x + λ2 y + λ3 z = 0 ≡  0  [* ]
0
 
Then det (A) = 0

• x, y , z
• A
otherwise

• x, y, z →
•A
Inverse matrix: " A −1"

A⋅B = I
−1
Then B is called inverse and denoted A

A ⋅ A −1 = I

Theorem (15-11) If A-nonsingular (i.e. det(A) ≠ 0


Then there is a unique inverse matrix A-1 which can
be expressed (found) in the following way:
( A jk )T
A −1 =
det( A)
−1 −1 −1
1. ( AB) = B A 2. (A−1 )−1 = A

3. Real numbers AT=A-1 : A is


4. Complex numbers A-T=A-1: A is
Example

 3 1 6
  Calculated earlier: det (C) = -24
C =  0 − 2 3
 0 0 4
 
−8 0 0 
 
Find C-1. Matrix of cofactors: (C jk ) =  − 4 12 0 
 15 − 9 − 6 
 
 − 8 − 4 15 
 
(C jk ) =  0 12 − 9 
 0 0 − 6 

−8 −4 15 
−1
(C jk )
T
1  
C = = − ⋅ 0 12 − 9
det( C ) 24 
 0 0 − 6 
Verification: C C-1=I
3 1 6 −8 −4 15 
   1  
0 −2 3 ⋅  0 12 − 9 =
0  24 
 0 4   0 0 − 6 

−8⋅3+ 1⋅0+ 6⋅0 −4⋅3+ 12⋅1+ 0⋅6 15⋅3− 9⋅1− 6⋅6


1
− −8⋅0+ (−2)⋅0 3⋅0 −4⋅0+ 12⋅(−2) + 0⋅0 15⋅0− −9⋅(−2)− 6⋅3
24
−8⋅0+ 0⋅0+ 0⋅ 4 −4⋅0+ 12⋅0+ 0⋅4 15⋅0− 9⋅0− 6⋅4

 − 24 0 0   1 0 0
1    
=−  0 − 24 0  =  0 1 0
24 
 0 0 − 24   0 0 1 
System of linear Equations

3V
I3  1 ⋅ I1 − 1 ⋅ I 3 = 1
R1 R2 
 3 ⋅ I 2 − 2 ⋅ I3 = 2
R3
I1 I2 − 1 ⋅ I − 2 ⋅ I + 3 ⋅ I = 3
 1 2 3

1V 2V

1 0 − 1  I1   1 
    
0 3 − 2  I 2  =  2  AX=R
 − 1 − 2 3  I   3 
  3   

A X R

How to find X :
AX=R
A-1AX=A-1R
IX=A-1R
X=A-1R
5 / 2 1 3 / 2
  −1
−1
A = 1 1 1  X=A R
3 / 2 1 3 / 2
 
 5 / 2 1 3 / 2  1  1⋅ 5 / 2 + 2 ⋅1 + 3 ⋅ 3 / 2   9 
      
X =  1 1 1  2  =  1⋅1 + 2 ⋅1 + 3 ⋅1  =  6 
 3 / 2 1 3 / 2  3  1⋅ 3 / 2 + 2 ⋅1 + 3 ⋅ 3 / 2   8 
      
9 I1=9A
 
X = 6 I2=6A
8 I3=8A
 

Possible solutions:

Det ≠ 0, R ≠ 0
 0
 
 0
Det ≠ 0, R = 0 X = 
M
 
 0
 

Det = 0, R = 0

Det = 0, R ≠ 0

**see page p348, Cramer’s rule.


Eigenvalue problem

A – square matrix of order 3.

 x1 
 
X =  y1 
z 
 1

AX = X '

?
AX = λ X  Eigenvalue problem

eigenvectors
To find
eigenvalues
Eigenvalues

AX = λ X
AX − λ X = 0 using distributive low
( A − λI ) X = 0

There are two cases for solutions:


0
 
0
1 Det ≠ 0, R = 0 X = 
M
 
0
 

2 Det = 0, R = 0

Just what we want!


Secular Determinant

∆=0 det( A − λI ) = 0

(a11 − λ ) a12 K a1n


a21 (a22 − λ ) M
=0
M O M
an1 L L (ann − λ )

p (λ ) = 0

Ex:
 2 2  2 2  x1   x1 
A =      = λ  
 −1 5   − 1 5  x2   x2 

  2 2   1 0   x1 
   − λ     = 0
 −1 5   0 1  x 
     2 

 (2 − λ ) 2  x1 
   = 0
 −1 (5 − λ )  x2 
Secular determinant:
(2 − λ ) 2
=0
−1 (5 − λ )

(2 − λ )(5 − λ ) − 2(−1) = 10 − 5λ − 2λ + λ2 + 2 =
λ2 − 7λ + 12 = 0

λ2 − 7λ + 12 = (λ − 3)(λ − 4) = 0

λ1,2 = 3, 4

Conclusion:
Eigenvectors

2−3 2  x1  − 1 ⋅ x1 + 2 ⋅ x2 = 0
   = 0 
 − 1 5 − 3  x2  − 1 ⋅ x1 + 2 ⋅ x2 = 0

x1 = 2x2

Let x2=1, then x1=2


 2
X1 =  
1

 6
Y = 3X =   eigenvector of A
 3 (corresponding to λ=3)
− 2 x1 + 2 x2 = 0  1
 Let x2=1, then x1=1 X 2 =  
 − x1 + x2 = 0  1

Eigenvalues λ1=3, λ2=4


Final result:  2  1
Eigenvectors X1 =   X 2 =  
1  1

Unit eigenvectors

X
U= X = x12 + x22 L xn2
X

 2
Ex: X1 =   X1 = 12 + 2 2 = 5
1

1  2
U1 =  
5 1
Properties of eigenvectors

 b11 b12 L b1n 


 
 M M L M 
B=
M M L M 
 
b L bnn 
 n1 bn 2

 λ1 

then −1


B AB = 
λ2

0 the diagonal
 form of A



0 O
λ


n

B-1AB  the transform of A by B


Canonical form

Theorem 15-13

Eigenvectors are orthogonal


e.g.
 y1 
X Y = ( x1 x2 )  =
+
X⊥Y⇒
 y2 
= x1 y1 + x2 y2 = 0
Basis vectors

1 0 0


     
U1 =  0  U2 =  1 U3 =  0 
0 0 1
     

U3
U1
U2
1
2
3

1 1 0 0


       
0 0 1 1
U1: A '   = λ1   U2: A '   = λ2  
M M M M
       
0 0 0 0
       
U1: 1 ⋅ a11' + 0 ⋅ a12' + 0 ⋅ a13' + L + 0 ⋅ a1' n = λ1
1 ⋅ a11' + 0 ⋅ a12' + 0 ⋅ a13' + L + 0 ⋅ a1' n = 0
1 ⋅ a11' + 0 ⋅ a12' + 0 ⋅ a13' + L + 0 ⋅ a1' n = 0
LLLLLLLLLLLLLLL

U2:
0 ⋅ a11' + 1 ⋅ a12' + 0 ⋅ a13' + L + 0 ⋅ a1' n = 0
0 ⋅ a11' + 1 ⋅ a12' + 0 ⋅ a13' + L + 0 ⋅ a1' n = λ2
0 ⋅ a11' + 1 ⋅ a12' + 0 ⋅ a13' + L + 0 ⋅ a1' n = 0
LLLLLLLLLLLLLLL

 λ1 


A =
'
λ2

 0
O 



0 λ


n
 λ1 0 L 0
 
 0 λ2 L 0
M M  U = λU
O M
 
0 0 L λn 

1 0 0


    
0 1 0
where U =  ,  , L ,  
M M M
    
0 0 1
    
U1, U2, … , Un

and λ = λ1,λ2, …, λn

AX = λ X
Ex:
Θ
U2  cos(θ ) 
X1 =   X1 = 1
X2
X1  sin(θ ) 
Θ
U1  − sin(θ )  X2 = 1
X 2 =  
 cos(θ ) 

U2
Θ
X2
X1
Θ
U1

1
X1 =   ≡ U1
0
0
X 2 =   ≡ U 2
1
Answer: B −1AB = A '

Answer: BT X = U

**) eigenvectors should be of unit length


Ex:

U1:  b11 b12 L b1n  x1   1 


    
 b21 b22 L b2 n  x2   0 
 M    = 
M O M M M
    
b
 n1 L L bnn  xn   0 

This is only possible if :

(b11 b12 L b1n ) 


(b21 b22 L b2 n ) orthogonal
 unit eigenvectors
M M O M 
(bn1 L L bnn ) 

c
T
B
( A ' − λI ) U = 0

(B −1AB − λI )BT X = 0
B −1AB ⋅ BT X − λIBT X = 0

∴ BT = B −1 ⇒ B −1AB ⋅ B −1X − λIB −1 X = 0


I
B −1A X − λIB −1 X = 0

B −1 ( A X − λI X) = 0
=0

transformations are valid


Summary
If A – Hermitian
Eigenvalue problem AX = λ X
can be transformed to a canonical form in the
basis of unit eigenvectors:

A' U = λ U

 λ1
where
 0 
 λ2  −1
A =
'
 = B AB
O
 0 
 λn 

and
U = BT X

where B is an orthogonal matrix of unit


eigenvectors (arranged as columns)
Eigenvalue problem
(practical example)

M M M
K K

X1(t) X2(t) X3(t)

Periodic solution: xk (t ) = xk eiωt

k = 1,2,3
After substitution:

− Mωx1 = − Kx1 + Kx2


− Mωx2 = Kx1 − 2 Kx2 − Kx3
− Mωx3 = Kx2 − Kx3
In matrix form:
 −1 1 0
K 
ω 2 X = AX where A= −  1 −2 1 
M 
 0 1 − 1
or equivalently:

( A − ω 2I) X = 0

Eigenvalue problem !
 x1 
 
The solutions ( i.e. vector X =  x2  )
x 
 3

Let us first redefine A


K
assuming that ≡1 λ ≡ ω2
M
 1 −1 0  Expected eigenmodes
 
A =  − 1 2 − 1 K
ω= λ
 0 −1 1  M
 
Secular determinant

R1 (1 − λ ) −1 0
−1 (2 − λ ) −1 = 0
R3 0 −1 (1 − λ )

C1 C3
(1 − λ ) 0 − (1 − λ ) R1-R3
−1 (2 − λ ) −1 = 0
0 −1 (1 − λ )

C1+C3
(1 − λ ) 0 0
−1 (2 − λ ) −2 = 0
0 −1 (1 − λ )
Characteristic equation:
(1 − λ )[(2 − λ )(1 − λ ) − 2] = 0
(1 − λ )[(λ2 − 3λ ] = λ (1 − λ )(λ − 3) = 0
Eigenvalues: λ1,2,3 = 0, 1, 3
K 3K
Resonance frequencies: ω1,2,3 = 0, ,
M M
Eigenvectors

λ1 = 0:  x1 − x2 = 0 Let x2 = 1

− x1 + 2 x2 − x3 = 0 From 1st: x1 = x2 = 1
 x2 + x3 = 0 From 3rd: x3 = x2 = 1

1 1
  X1 1  
X1 = 1 Unit vector: U1 = = 1
1 X1 3 
  1

λ2 = 1:  − x2 = 0 x2 = 0

− x1 + x2 − x3 = 0 x1= - x3 , Let x1= 1
 − x2 = 0 x3= -1

1 1
  1  
X2 =  0  Unit vector: U2 = 0
 − 1 2 
   − 1
λ3 = 0:  − 2 x1 − x2 = 0 Let x2 = 2

− x1 − x2 − x3 = 0 x1 = - 1
 − x − 2x = 0 x3 = - 1
 2 3

 − 1  − 1
  1  
X3 =  2  Unit vector: U3 = 2
 − 1 6 
   − 1

Modes of oscillation
ω = 0, No oscillation, uniform movement
1
 
X1 = 1
1
 

K
ω2 =
M
1
 
X2 =  0 
 − 1
 
1 0 -1
 − 1
  3K
X3 =  2  ω3 =
 − 1 M
 

-1 2 -1

The eigenvectors represent the states of the


system corresponding to different modes defined
by the eigenvalues

Transformation to the eigenvectro basis

'
X = BT X
Transformation requires that:

B - orthogonal

1 1 − 1
 
B = 1 0 2 
1 − 1 − 1
 
 −1 1 0
 
A = 1 −2 1 
0 1 − 1
 

 X1 ⋅ X 2 = (1 ⋅ (−1) + 0 ⋅1 + 1 ⋅1) = 0

 X1 ⋅ X 3 = (1 ⋅ (−1) + 2 ⋅1 + 1 ⋅ (−1)) = 0
X ⋅ X = (1 ⋅ (−1) + 0 ⋅ 2 + (−1) ⋅ (−1)) = 0
 2 3

Orthogonal ⇔
 λ1 0
  '
 λ2  X = λ X
'
Where
0 λ3 

 x '1   0   0 
K 3K   '  
λ = 0, , and X ' 1, 2,3 =  0 ,  x 2 ,  0 
M M  0   0   x' 
     3
Accordingly:
 − Mω 2 x '1 = 0 + 0 + 0

− Mω x 2 = 0 − Kx 2 + 0
2 ' '

 − Mω 2 x '3 = 0 + 0 − 3Kx '3


 x '1 = x1 + x2 + x3
 '
'
X = BT X ⇔  x 2 = x1 − x3
 x '3 = − x + 2 x − x
 1 2 3
Summary
(coupled linear differential equation)
Given: A physical system of order n, which can be
described by eigenvalue problem:

AX = λ X
In which we are looking for:

eigenvalues modes
eigenvectors states

Can be diagonalised by transforming to the


basis of the eigenvectors

This can only be done if A is symmetric

In the eigenvector basis matrix A has diagonal


form where the main diagonal consists of
eigenvalues
Degeneracy

2 1 1
 
A = 1 2 1
 1 1 2
 

det( A − λI ) = 0 = −(λ − 1) 2 (λ − 4) = 0
λ1, 2,3 = 1,1,4
λ=1 is used twice!

degenerate
Eigenvectors:

λ1 = 1:  x1 + x2 + x3 = 0 Let x2 = -2

 x1 + x2 + x3 = 0 x1 = 2-x3
x + x + x = 0 x1 = x3 = 1
 1 2 3

 1 
 
X1 =  − 2 
 1 
 
 − 2
 
X2 =  1 
 1 
 

1
 
λ=4: X3 = 1
1
 

However: X1 X 2 = −2 ⋅1 + (−2) ⋅1 + 1⋅1 = −3 ≠ 0

X1 ⋅ X3 = X 2 ⋅ X 3 = 0
To make the second vector ortogonal use
Gram-Schmidt orthonormalisation process.
'
X 2 = X 2 − ( X 2 ⋅ U1 )U1 =

 1   − 2
1    
U1 =  − 2 X2 =  1 
6   1 
 1   

 − 2  1  −3 
   1   1  2 
=  1 − (−2 ⋅1 − 2 ⋅1 + 1)  − 2  = 0 
 1   6   6  3 
   1   2 

 − 1
1  
U2 = 0
2 
1
Check that it satisfies the
eigenvalue problem:

 2 1 1  − 1  − 1 − 2 + 1 = −1
     
 1 2 1  0  =  0  − 1 + 1 = 0 OK
 1 1 2  1   1  − 1 + 2 = 1 
    
Differential equation as
an eigenvalue problem

So far: AX = λ X

Xn λn

Xn ⋅ Xm = 0 m≠n
Xn ⋅ Xm ≠ 0 m=n

can be also normalised so that:


2
Xn ⋅ Xn = 1 Xn = 1
The General Eigenvalue Problem

AΦ = λ Φ where

Φ = ∑ cn X n
n

Φ = c1 X1 + c2 X 2 + c3 X3 L

Ac1 X1 + Ac2 X 2 + Ac3 X3 + L = λc1 X1 + λc2 X 2 + L

 
A ∑ cn X n  = ∑ λn cn X n
 n  n

YES, we can
To start with, introduce Kroenicker delta:

0 m = n
X n ⋅ X m = δ n ,m =
1 m = n

Φ = ∑ cn X n
n

 
Φ ⋅ X p =  ∑ cn X n  X p = ∑ c n X n X p =
 n  n

Φ ⋅ X p = ∑ cn X n X p = ∑ cnδ n , p = c p
n n
1 2πnx
f n ( x) = sin( )
2L L
1 2πnx
g n ( x) = cos( )
2L L

2πn 2πm
L L
1 0 m = n
∫− L f n ( x) f m ( x) = 2L −∫Lsin( L ) sin( L )dx = 1 m = n
= δn,m

Also:

F ( x ) = ∑ cn f n ( x ) -- “sin” Fourier series


n
For cp coefficient:
L L

∫ F ( x) f p ( x)dx = ∑ cn ∫ f n ( x) f p ( x)dx =
−L n −L

= ∑ cnδ n , p = c p
n

Conclusion:

Restrictions:
Eigenvalue problem in differential equations

Suppose we have a differential operator L(x):

A∂ 2 ∂
L( x) = 2 + B + Cu ( x)
∂x ∂x

ψ ( x ) = ∑ cn f n ( x )
n

L( x)ψ ( x) = λψ ( x)
The Method

Is to try for a solution for ψ(x) in the form:

ψ ( x ) = ∑ cn f n ( x )
n

The task -- to find cn

∑ L ( x )c
n
n f n ( x ) = ∑ λ cn f n ( x )
n

The trick – use orthonormal properties of fn(x) for


solving the equation.

∫ f m* ( x) f n ( x)dx = δ n ,m
∑ n∫ m
c
n
f *
L ( x ) f n ( x ) dx = ∑ n ∫ m ( x) f n ( x)dx
λ c
n
f *

Lmn δm,n

∑c L
n
n mn = ∑ λcnδ m,n
n

Repeat for All m

m=1 L11c1 + L12 c2 + L = 1⋅ λc1 + 0 ⋅ λc2 + 0 ⋅ λc3 + L

m=2 L21c1 + L22c2 + L = 0 + 1⋅ λc2 + 0 + L


m=3 L31c1 + L32c2 + L = 0 + 0 + 1⋅ λc3 + L
---------------------------------

c
Lψ = λψ
 L11 L L1n   c1 
   
L= M O M  ψ = M 
L L L  c 
 n1 nn   n

λ 0
 
λ = O 
0 λ 
 

Result:

L( x)ψ ( x) = λψ ( x)

Lψ = λψ
Advantages

Disadvantages

1 2πnx N = 1,2,3,…infinity
Ex.: sin( )
2π L
Example:

A particle of mass M moves in the 1D box 0 < x < L.


Consider the set of functins {exp(ikx)} as a basis for the
wavefunction of the particle, i.e. as a basis for the
solution ψ(x) to the Schrödinger equation

h2 d 2
− 2
ψ ( x) = Eψ ( x)
2M dx

• What values must k take for the boundary condition


ψ(0) = ψ(L) to be satisfied.
• Form an orthonormal basis set {fn(x)}
• Write the Schrödinger equation for the particle as a
matrix eigenvalue problem and find the possible
values of E
• Next assume that a potential

2πx 2π 2 h 2
V ( x) = V0 sin( ) V0 = 1.24
L ML2

is switched on inside the box.

i. Construct the matrix elements associated with


V(x) with respect to the basis set {fn(x)}
ii. Assume that {f1,f2} forms a complete set of the
new Hamiltonian:

h2 d 2
− 2
ψ ( x) + V ( x)
2M dx

Find the two eigenvalues and the corresponding


eigenvectors.
Solution

ψ ( x) = a exp(ikx)

ψ (0) = ψ ( L) exp(ikL) = 1
kL = 2πn
2πn
k = kn =
L
n = 0,1,2, L ∞

{ f n ( x) = An exp(ikn x)}

L L


−L
f n ( x) f m ( x) = An* Am ∫ e −ikn x eikm x dx =
0
L
= An* Am ∫ e −i ( kn − km ) x dx =
0
L
e 
−i ( k n − k m ) x
= A Am 
*
n  dx =
 i (k n − k m )  0

= An* Am ⋅ 0 = 0
L L


0
f n* ( x) f n ( x) = An* An ∫ dx = An* An L = 1
0

∫0
f n* ( x) f m ( x)dx = δ n ,m


Ψ ( x ) = ∑ cn f n ( x )
n =1

h2 d 2
− 2
ψ ( x) = Eψ ( x)
2M dx
L(x)

L ( x ) ∑ n c n f n ( x ) = E ∑ n cn f n ( x )
L L

∫ f m* L( x)∑n cn f n ( x)dx = E ∫ f m* ( x)∑n cn f n ( x)dx


0 0

L L

∑c ∫fn n
*
m L( x) f n( x)dx = E ∑n cn ∫ f m* ( x) f n ( x)dx
0 0

∑cL n n mn = E ∑n cnδ m,n

∑cL n n mn = Ecm

 L11 L12 L  c1   c1 
    
 L21 O  c2   c2 
 M  M  = E  M 
    
    
    
L
Lmn = ∫ f m* L( x) f n( x)dx
0

− h 2 d 2 1 ikn x h 2 2
L( x) f n( x) = 2
e = k n f n ( x)
2M dx L 2M

L
h2 2 * h2 2
Lmn = k n ∫ f m f n( x)dx = kn δ m,n
2M 0 2M

 h2 2 
 
0
k1
 2M 
 h2 2 
Lmn = k2 
2M
 
h2 2




0 2M
k3 

O
 h 2 k12 

 2M
 2
h k 2 0  c1 
 
 c2 
 c1 
 
 c2 
 M  = E  M 
2
 2M
    
0    
O
    
 
 

h2 2 n = 0,1,2 … ∞
En = kn
2M
2πn
kn =
L

 2π 2 h 2  2
En =  2 
n
 ML 
ii) 2πx
V ( x) = V0 sin( )
L
L
Vmn = ∫ f m*V ( x) f n( x)dx =
0

2πx −ikn x
L
1
= V0 ∫ e −ik m x sin( )e dx =
L0 L
V0 1 i ( km − kn ) x i 2Lπ x − i 2π
L
x
= ∫
L 2i 0
e (e −e L
)dx =


L 2π
V0 i ( m − n +1) x i ( m − n −1) x
= ∫
2iL 0
e L
− e L
dx =

V0
= [ Lδ m − n , −1 − Lδ m − n ,1 ]dx =
2iL

V0
Vnn = ⋅0 = 0
2iL

V0 iV
Vn,n +1 = [ L, δ1, −1 − Lδ1,1 ] = 0
2iL 2
V0 iV0
Vn,n −1 = L=−
2iL 2
 iV0
− 2 n = m +1
 iV
Vnm = 0 n = m −1
 2
 0 otherwise


 iV0 
 0 0 L
 2 
 − iV0 0
iV0
0
V = 2 2 
 iV0 
 0 − 0 O
 2 
 M 0 O O

 h2 d 2 
 − 2
+ V ( x ) ψ ( x ) = E ψ
'
( x)
 2M dx 

( L + V )ψ ( x) = E ψ
'
( x)
2
ψ ( x ) = ∑ cn f n ( x )
n =1
(L + V )c = E ' c

 2π 2 h 2 
Lc = E c E n = αn 2 α ≡  2 

 ML 

 L11 L12  c1   α 0  c1 
i.e.    =   
 L21 L22  c2   0 4α  c2 

 iV0 
 0 
V = 2 
 − iV0 0 
 2 
  iV0  
 α 0   0 
2  − E 'I  c1  = 0
  +  c 
 0 4α   − iV0 0   2 

  2  
2
V
(α − E ' )(4α − E ' ) − 0 = 0 V0 = 1.24α
4

( E ' − 0.9α )( E ' − 4.1α ) = 0

Eigenvalues: E '1, 2 = 0.9α ,4.1α

 iV0
 0.1αc1 + 2 c2 = 0
E`=0.9α:
 iV
− 0 c1 + 3.1c2 = 0
 2
i  1 
c2 = c1 ⇒ c1 =  i 
31  
 31 

E`=4.1α:
i  i 
c1 = c2 ⇒ c2 =  31 
31  1 
 
2
1 ik1x 1
ψ ( x) = ∑ cn f n ( x) = c1 e + c2 eik2 x
n =1 L L

1  ik1x i ik2 x 
ψ ( x) =
'
1 e + e 
L  31 
1  i ik1x ik2 x 
ψ ( x) =
'
2  e +e 
L  31 

L L

∫ 1 ( x)ψ 1 ( x) = 1
ψ ∫ 2 ( x)ψ 2 ( x) = 1
ψ
* *

0 0
1 1 31
const = = =
c12 + c22  i 
2 30
1 +
2

 31 

31  ik1x i ik2 x 
ψ 1 ( x) =  e + e 
30 L  31 
31  i ik1x ik2 x 
ψ 2 ( x) =  e +e 
30 L  31 
Perturbation Theory

Ax = λ x
L(r )Ψn (r ) = En Ψn (r )

Many problems are not easily solved,


but are closed to one which can be…

d2
Example: L( r ) = ∇ 2
(i.e. 2 )
dx

1 ik r
ψ k (r ) = e
V
(A + a )x = λ x

Example:

M M M
K K

Described by: Ax = ω 2 x
 K K 
 − 0 
with
 M M 
K 2K K
A = − − 
 M M M
 K K 
 0 − 
 M M 
M M M
K+S K

Bx = ω 2 x

 K +S K +S 
 − 0 
 M M 
K +s 2( K + S ) K
B = − − 
 M M M
 K K 
 0 − 
 M M 

B = A+a

 S S 
 − 0
 M M 
 S 2S
a= − 0
 M M 
 0 0 0
 
 
Ax = ω 2 x

Bx = ω 2 x

using Perturbation Theory

Which says that the solution to Bx = ω 2 x


is a perturbation of the solution to A x = ω 2 x
L(r )ψ n (r ) = Enψ n (r )

∫ n (r )ψ n (r )d r = δ n,m
ψ *

Need to solve:

(L(r ) + U (r ))Ψ (r ) = EΨ (r )
For Ψ(r) and E given U(r).

Assume that U is small and Ψ(r) is close to an


eigenvector ψn(r), and E is close to an eigenvalue En.

i.e.
Ψ = ψ n + δψ

 E = En + δE
Expand Ψ(r) in terms of the basis:

Ψ (r ) = ∑ cmψ m (r ) ⇒
m

Ψ (r ) = ψ n (r ) + ∑ cmψ m (r ) =
m≠n

= ψ n + ∑ ' cmψ m ( ∑ ' means m ≠ n)


m m


m
'
cmψ m

Substitute for Ψ(r) in:

(L(r ) + U (r ))Ψ (r ) = EΨ (r )
(L + U )(ψ n + ∑ 'cmψ m ) = E (ψ n + ∑ 'cmψ m ) ⇒
m m

(L + U − E )(ψ n + ∑ 'cmψ m ) = 0
m
(Lψ n + Uψ n − Eψ n ) + ∑ 'cm ( Lψ m − Eψ m + Uψ m ) = 0
m
Enψ n Emψ m

(En − E )ψ n + Uψ n + ∑ 'cm (( Em − E )ψ m + Uψ m ) = 0
m

∫ mψ n = δ m,n
ψ *

(En − E )∫ψ n*ψ n + ∫ψ n*Uψ n +


+ ∑ ' cm (( Em − E ) ∫ψ n*ψ m + ∫ψ n*Uψ m ) = 0
m

E = En + ∫ψ n*Uψ n + ∑ ' cm ∫ψ n*Uψ m


m
0 + ∫ψ *pUψ n + ∑ ' cm (( Em − E ) ∫ψ *pψ m + ∫ψ *pUψ m ) = 0
m
δp,m

∫ p n p p
ψ *
Uψ + c ( E − E ) + ∑ m ∫ pUψ m = 0
'
c ψ
m
*

∫ pUψ n
ψ *
1
cp = −
Ep − E
− ∑
Ep − E m
'
cm ∫ψ *pUψ m

E ≈ En ∫ nUψ m << En cm << 1


ψ *

cm ∫ψ n*Uψ m
E = En + ∫ψ n*Uψ n

∫ pUψ n
ψ *

cp =
E − Ep

'∫ m
ψ *
Uψ n
Ψ =ψ n + ∑ ψm
m E − Em

'∫ m
ψ *
Uψ n
Ψ =ψ n + ∑ ψm
m E n − Em
E = En + ∫ψ n*Uψ n +

' ∫ m n ∫ nUψ m
ψ *
Uψ ψ *

+∑
m E − Em

∫ mUψ n
ψ *

cm = − −
Em − E

' ∫ m p∫
1 ψ *
Uψ ψ pUψ m
*

− ∑
Em − E p E − Ep
The almost harmonic oscillator

h2 d 2
H =− 2
+ V ( x)
2M dx
where
1
V ( x) = mω 2 x 2 + ε1 x 3 + ε 2 x 4
2

small

( L( x) + U ( x))ψ ( x) = Eψ ( x)

Accordingly:
h2 d 2 1
L( x) = − 2
+ mω 2 2
x
2m dx 2
Exact harmonic oscillator
U ( x) = ε 1 x 3 + ε 2 x 4

E = En + En(1)
first order correction

En(1) = ∫ψ n*U ( x)ψ n dx


Eigenfunctions corresponding to En
1
In our case: En = (n + )hω
2
n=0 1
E0 = hω
2
1
mωx 2
 mω  4 −
ψ0 =   e 2h

 hπ 
+∞
E0(1) = ∫ψ 0*U ( x)ψ 0 dx =
−∞
1
+∞ − mωx 2
 mω  2
= ∫  e h
(ε 1 x 3 +ε 2 x 4 )dx =
− ∞
hπ 
1
+ ∞ − mωx 2
 mω  2
= ε2 
 hπ 
∫e
−∞
h
x 4 dx =

3ε 2 h 2
=
4m 2ω 2

1 3ε 2 h 2
E = hω +
2 4m 2ω 2

You might also like