Eigenvalues
By Hal 2001 on Sunday, December 30, 2001 -
11:08 pm:
Let A be a square matrix such that for some positive integer
m, Am = 0. Show that every eigenvalue of A is
zero.
Let A be a real 2 x 2 matrix which has non-zero non-real
eigenvalues. Show that the non-diagonal elements of A are
non-zero, but that the diagonal elements may be zero.
Thanks for your help in advance.
Hal.
By Kerwin Hui on Sunday, December 30, 2001
- 11:15 pm:
Hint: The first part - proof by
contradiction. The second part - what can you tell me about the
trace and determinant of A?
Kerwin
By Hal 2001 on Sunday, December 30, 2001 -
11:57 pm:
Hi Kerwin,
Another question,
Suppose that A is an n×n square matrix such that A-1 exists,
and let l be an eigenvalue of A. Show that l ¹ 0, and that
1/l is an eigenvalue of A-1.
A x=lx
A-1A x=A-1lx
x=lA-1 x
What do I do next?
Thanks.
Hal.
By Kerwin Hui on Monday, December 31,
2001 - 09:20 am:
Hal,
(a) Apply A x=lx m times and remember Am=0
(b) You should get the condition (a-d)2 < -4b c. Now LHS ³ 0.
(c) You have done most of the work. Now to see l ¹ 0, just
choose an arbitrary multiple of our eigenvector and get a contradiction for
l = 0. To see l-1 is an eval of A-1, divide both sides
by l.
Kerwin
P.S. They look suspiciously like example sheet questions...
By Hal 2001 on Monday, December 31, 2001 -
10:34 am:
Hi Kerwin,
Yes, they are from example sheet A&G No.6 I was going through
the questions again to make sure I understood the math. I am
thankful for your help.
I think I understand (b). Since (a-d)2 > 0. But if
a and d are equal then condition that (a-d)2 > = 0
exists - but does that contradict that eigenvalues are non-real?.
How do we justify statement 'but that the diagonal elements may
be zero'. Does that mean than both a, d could be zero or only one
of them?
a)
If you have A x=lx,
Am x=lm x
If x ¹ 0 this implies that Am=0.
Does this make sense? And xm would be meaningless, right?
(c)
I understand that the eval for A-1. But did not fully follow why
l ¹ 0.
To clarify, what is our evector here?
Thanks.
Hal.
By Kerwin Hui on Monday, December 31,
2001 - 10:47 am:
Hal,
(a) Am x=lm x.
But Am=0, so Am x=0 for all x.
Suppose l ¹ 0, we have a non-zero eigenvector x, and so RHS ¹ 0,
but LHS=0. Contradiction.
(b) What you get is the condition -4b c > 0, so b, c are non-zero.
a and d can both be zero, e.g. the matrix with b=1, c=-1, a=d=0 has
evals ±i.
(c) Suppose A has evect xwith eval l, then for any m,
we have A(mx)=lmx=0, i.e. mx is in the kernel of A.
However, the kernel must be trivial (Hit with A-1 to see why).
Kerwin
By Hal 2001 on Monday, December 31, 2001 -
11:01 am:
Thanks for your help Kerwin.
Have a nice new year.
Hal.