University College London

Department of Physics and Astronomy

2B21 Mathematical Methods in Physics & Astronomy
Suggested Solutions for Problem Sheet M4 (2003–2004)

1. Rewrite the equations in matrix form,

d 2x
= Ax ,
dt2
where !
− 52 3
2 [1]
A= 3 .
2
− 52
This has eigenvalues λ given by
 
 (− 5 − λ) 3 
 2 2 =0,

3
(− 52 − λ)

 
2

which has solutions λ1 = −1 and λ2 = −4. The normal modes therefore satisfy [2]
the uncoupled equations

d2 y1
+ y1 = 0 ,
dt2
d2 y2
+ 4y2 = 0 . [1]
dt2

To relate the normal modes to the original coordinates, we must find the
rotation matrix R, i.e. the eigenvectors of A. For λ1 = −1, we require
! ! !
− 32 3
2
r11 0
3 = .
2
− 32 r21 0

By inspection, the (normalised) solution is

√ !
1/√2
r1 = . [2]
1/ 2

For the other eigenvalue of λ2 = −4, we require

! ! !
3 3
2 2
r12 0
3 3 = .
2 2
r22 0

By inspection, the (normalised) solution is

√ !
1/ √2
r2 = .
−1/ 2

1
 [2]

The rotation matrix √ √ !

1/√2 1/√2 [1]
R=
1/ 2 −1/ 2
and the two sets of coordinates are related by
1 1
x1 = √ (y1 + y2 ) : y1 = √ (x1 + x2 ) ,
2 2
1 1
x2 = √ (y1 − y2 ) : y2 = √ (x1 − x2 ) . [1]
2 2

The general solutions of the uncoupled differential equations are

y1 = α1 cos t + β1 sin t ,
y2 = α2 cos 2t + β2 sin 2t . [1]

√ √
At time t = 0, ẏ1 = ẏ2 = 0, y1 = a 2, and y2 = −a 2. At later times,
therefore,
√
y1 = a 2 cos t ,
√
y2 = −a 2 cos 2t . [2]

Rotating back to the original coordinates,

x1 = a(cos t − cos 2t) ,

x2 = a(cos t + cos 2t) . [1]

Students can actually solve this simple two-degree-of-freedom problem by much

easier methods. Adding and subtracting the two original equations gives

d2 x1 d2 x2
2 + 2 = −2x1 − 2x2 ,
dt2 dt2
d2 x1 d2 x2
2 2 − 2 2 = −8x1 − 8x2 .
dt dt
We can see directly here the uncoupled equations in x1 ± x2 and all the sub-
sequent manipulations should come out. However it does not use the matrix
diagonalisation technique asked for. The maximum mark is therefore only
10/14.

2
2. For the matrix  
1 i 3i
A =  −i 1 −3 


−3i −3 −3
the eigenvalue equation is
 
 1−λ i 3i 
 
| A − λI |=  −i 1−λ −3
 

 −3i −3 −3 − λ
 


= (1 − λ)[(1 − λ)(−3 − λ) − 9] − i[−i(−3 − λ) − 9i] + 3i[3i + 3i(1 − λ)]

= (1 − λ)(λ2 + 2λ − 12) + (λ − 6) + 3(3λ − 6) = 0 .
The characteristic equation is therefore

λ3 + λ2 − 24λ + 36 = 0 . [3]

By inspection, λ = 2 is one solution and, factorising this out,

(λ − 2)(λ2 + 3λ − 18) = (λ − 2)(λ − 3)(λ + 6) = 0 ,

and hence the eigenvalues are λ1 = 2, λ2 = 3, and λ3 = −6. [2]

(i) The sum of the eigenvalues is 2+3−6 = −1, whereas the trace is 1+1−3 =
−1, as expected. [1]
(ii) The product of the eigenvalues is 2 × 3 × −6 = −36. The determinant
   

 1 i 3i   1 i 3i 
| A |=  −i 1 −3  =  0 0 −6  = −36 ,

 −3i −3 −3   −3i −3 −3 
   

as predicted. [2]
The eigenvector equation in the case of λ = 2 is
       
1−λ i 3i u11 −1 i 3i u11 0

 −i 1 − λ −3   u21  =  −i −1 −3   u21  =  0 
     
 ,
−3i −3 −3 − λ u31 −3i −3 −5 u31 0

leading to the two independent equations

−u11 + iu21 + 3iu31 = 0 ,

−3iu11 − 3u21 − 5u31 = 0 .

This has solution  

i
1  
u1 = √  1  , [3]
2 0
where the factor in front has been inserted to ensure that the eigenvector

3
is normalised, u†1 u1 = 1. Note that this involves complex conjugation and
students might forget this point.
For λ = 3,   
−2 i 3i u12
 −i −2 −3   u22  = 0 .
  

−3i −3 −6 u32
This requires

−2u12 + iu22 + 3iu32 = 0 ,

−iu12 − u22 − 2u32 = 0 ,

from which  
i
1 
u2 = √  −1  · [3]

3 1

Finally, for λ = −6,

  
7 i 3i u13
 −i 7 −3 
  u23  = 0 .
  

−3i −3 3 u33

Hence

−iu13 + 7u23 − 3u33 = 0 ,

−3iu13 − 3u23 + 3u33 = 0 .

This has solution  

−i
1 
u3 = √  1  · [3]

6 2

Taking the scalar products,

u†2 u1 ∝ −i × i − 1 × 1 + 1 × 0 = 0 ,
u†3 u2 ∝ i × i + 1 × (−1) + 2 × 1 = 0 ,
u†1 u3 ∝ −i × (−i) + 1 × 1 + 0 × 2 = 0 . [3]

Hence the eigenvectors are orthogonal to each other.