Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

PHYS 312, Worksheet for Lecture 2

We will have a pop quiz, solving a system of two linear equations in three unknowns.
2x + y − z = 3
x+y+z =2
Lot’s of ways to solve it, for example: eliminate 1 variable (say, z, by adding the two
equations) to get 3x + 2y = 5. This can be solved by, say x = a, y = 5/2 − 3a/2. Finally,
solve for z = 2−x−y = −1/2+a/2. Let’s check: 2x+y −z = 2a+5/2−3a/2+1/2−a/2 = 3
and x + y + z = a + 5/2 − 3a/2 − 1/2 + a/2 = 2.
Another solution could be in terms of an undetermined z: then 2x + y = 3 + z and
x + y = 2 − z so x = 1 + 2z and y = 1 − 3z.
A useful way to represent these solutions is as follows:
         
x 0 1 1 2
y  =  5  + a − 3  = 1 + z −3
2 2
z − 12 1
2
0 1

To see that these two presentations are indeed the same solution, we can for example check
that when a = 1, the first solution obtains the point (x,y,z)=(1,1,0), and that the vector
multiplying a is paraller to the vector multiplying z. At this point, it will be good to talk a
little about terminology. Our set of equations, which we can write as
 
  x  
2 1 −1   3
y =
1 1 1 2
z

is inhomogeneous because the zero vector x = y = z = 0 is not a solution. The corre-


sponding homogeneous problem is
 
  x  
2 1 −1   0
y =
1 1 1 0
z

The general solution to the inhomogeneous problem can be broken down into two pieces: a
 
0
particular solution and the solution to the homogeneous problem. In our examples,  5  2
− 12
     
1 2 1
and 1 were both particular solutions, while z −3 and a − 3  were two ways to present
2
1
0 1 2
the solution to the homogenous problem. This terminology, as well as this approach (find a
particular solution + find the general solution to the homogeneous problem) will be a good
way to solve linear differential equations as well.
1 Consider the equation Av = b, where the unknown is
 
x
v= ,
y

a column 2-vector. Take A to be  


1 3
A=
−2 −6
and let 

−2
b= .
4
Fill in the missing values in the solution to the equation Av = b:
   
v= + c

where c is an arbitrary parameter.


 
−2
In the first column vector, we can put any particular solution to the system, say (by
0
inspection). In the second column vector, we need a solution to the homogeneous equation,
 
−3
such as . So,
1
   
−2 −3
v= + c
0 1
is a valid solution, and so are (for example)
   
1 −1
v= + c 1
−1 3

  
0 3
v= + c
− 32 −1
2 Consider again the equation Av = b, but now let v be a column 3-vector and let
 
6 3 0
A = 3 −2 0
0 0 3

and 

1
b= 7  .
−4
The matrix A has three eigenvalues and eigenvectors:

−1 −1
           
0 0 3 3
A 0 = 3 0
   A 3 = −3 3 
   A 1 = 7 1
  
1 1 0 0 0 0

Hint: to make full use of the eigenvectors, notice that they are orthogonal!

The matrix A has full rank (the determinant is nonzero) so it should be invertible.
The eigenvectors/matrix columns look/are linearly independent, so the matrix should
have an inverse, A−1 . Thus, we find v via v = A−1 b, but it looks like a lot of work. . ..
Since we know the eigenvectors, however, we don’t need to computer the inverse by brute
force. Let’s write our unknown vector in terms of the eigenvectors:

−1
     
0 3
v = a1 0 + a2  3  + a3 1
1 0 0

To match, let’s also assume that we can write b it terms of the eigenvectors (after all, they
should be linearly independent!):

−1
     
0 3
b = c1 0 + c2  3  + c3 1
1 0 0

Our equation Av = b now becomes

−1 −1
           
0 3 0 3
3a1 0 + (−3)a2 3 + 7a3 1 = c1 0 + c2 3 + c3 1
          
1 0 0 1 0 0

Because the eigenvectors are linearly independent, we have that a1 = c1 /3, a2 = −c2 /3 and
a3 = c3 /7. But what are the ci coefficients? These also satisfy a linear equation:

0 −1 3
    
c1 1
0 3 1 c2  =  7 
1 0 0 c3 −4

To solve for the ci s, we can invert the matrix:


−1  
0 −1 3
  
c1 1
c2  = 0 3 1  7 
c3 1 0 0 −4

We have replaced one inversion with another. . ., which might not look like progress.
We are in luck, however, because the eigenvectors of matrix A (due it it being symmetric)
are not only linearly independent, they are in fact orthogonal. This gives us a different
way to find ci (or, if you want, a simple way to find the inverse above):

−1
        
0 3 1
   
0 0 1 c1 0 + c2 3 + c3 1
       = 0 0 1  7
1 0 0 −4

which gives us c1 = −4,

−1
        
0 3 1
   
−1 3 0 c1 0 + c2 3 + c3 1
       = −1 3 0  7
1 0 0 −4

which gives us 10c2 = 20, c2 = 2 and

−1
        
0 3 1
   
3 1 0 c1 0 + c2  3  + c3 1 = 3 1 0  7 
1 0 0 −4

which gives us 10c3 = 10, c3 = 1. Substituting back, a1 = −4/3, a2 = −2/3, a3 = 1/7 and
therefore    23 
−1
   
0 3
4  2   1    2113 
v=− 0 − 3 + 1 = −7 =
3 3 7
1 0 0 − 43
(check that it works for yourself!)

3. A follow up question: re-do the previous problem with


 
6 3 0
A = 3 −2 0
0 0 0

and  
1
b = 7 .
0
You should be able to quess the eigenvectors/eigenvalues of this new A from the one above.
Comparing the form of the new A with the one from question 2, we see:

−1 −1
           
0 0 3 3
A 0 = 0 0 = 0
   A 3 = −3 3 
   A 1 = 7 1
  
1 1 0 0 0 0
Following the solution before, we write our unknown vector in terms of the eigenvectors:

−1
     
0 3
v = a1 0 + a2 3 + a3 1
    
1 0 0

To match, let’s also assume that we can write b it terms of the eigenvectors (after all, they
should be linearly independent!):

−1
     
0 3
b = c1 0 + c2 3 + c3 1
    
1 0 0

Our equation Av = b now becomes (notice the only change, in the value of the first eigen-
value)
−1 −1
           
0 3 0 3
0a1 0 + (−3)a2 3 + 7a3 1 = c1 0 + c2 3 + c3 1
          
1 0 0 1 0 0
Because the eigenvectors are linearly independent, we have that a2 = −c2 /3 and a3 = c3 /7.
Also a1 can be anything, but c1 must be zero or the equation will have no solution at all!
Now,
0 −1 3
    
c1 1
0 3 1 c2  = b 7
1 0 0 c3 0
Using orthogonality of the eigenvectiors, we have

−1
        
0 3 1
   
0 0 1 c1 0 + c2 3 + c3 1
       = 0 0 1 7

1 0 0 0

which gives us c1 = 0 (therefore, there exists solutions, see above). Also,

−1
        
0 3 1
   
−1 3 0 c1 0 + c2 3 + c3 1
       = −1 3 0 7

1 0 0 0

which gives us 10c2 = 20, c2 = 2 and

−1
        
0 3 1
   
3 1 0 c1 0 + c2  3  + c3 1 = 3 1 0 7
1 0 0 0

which gives us 10c3 = 10, c3 = 1. Substituting back, a1 is anything and we can call it, say,
c, a2 = −2/3, a3 = 1/7 and therefore
 23 
−1
       
0 3 0 21
2 1
v = c 0 −  3  + 1 = c 0 + − 13
7
 .
3 7
1 0 0 1 0

You might also like