Invertible linear transformation

Click For Summary
The discussion centers on proving the invertibility of a linear transformation T on R^n under the condition that the norm ||T-I|| is less than 1. It is established that if T is not invertible, then 0 would be an eigenvalue, leading to a contradiction with the spectral radius being less than 1. The participants explore the implications of the spectral radius and norms, noting that if ||T-I|| < 1, then the series sum from k=0 to infinity of (I-T)^k converges absolutely, indicating T is invertible. They also reference a theorem stating that if the spectral norm is less than 1, then higher powers of the matrix tend to 0 as n approaches infinity. The discussion highlights the relationship between the spectral radius, norms, and the convergence of series in establishing the invertibility of T.
CarmineCortez
Messages
30
Reaction score
0

Homework Statement


If T is a linear transformation on R^n with || T-I || < 1, prove that T is invertible.



The Attempt at a Solution



So a linear transformation T is invertible iff the matrix T is not singular.
and I know for any matrix A, ||A|| > spectral radius(A).

so, spectral radius(T-I) < 1.
 
Physics news on Phys.org
What would happen to T-I, if 0 was an eigenvalue of T? Is it compatible with the hypothesis?
 
if 0 was an eigenvalue of T then T would be singular..
 
CarmineCortez said:
if 0 was an eigenvalue of T then T would be singular..

Ok, so if T is not invertible then Tv=0 for some v. So v corresponds to what eigenvalue of T-I?
 
Dick said:
Ok, so if T is not invertible then Tv=0 for some v. So v corresponds to what eigenvalue of T-I?

0 = λ*v + I*v
=> -1 = λ

but I know my spectral radius is <1 so contradiction...
 
Yes. That's it. You could also say (T+I)v=(-v) means ||T+I||>=1 and not even say anything about spectral radius. Still a contradiction with ||T+I||<1.
 
I need to show: sum from k=0 to infinity of (I-T)^k converges absolutely to T^(-1)

so if ||T-I|| <1 then is ||I-T|| < 1? and all the properties I listed carry over? I'm still not too sure where to go with this.

when the spectral radius is <1, the higher powers of the matrix tend to 0, so it clearly converges...
 
For any norm \left\|v\right\|=\left\|-v\right\|. Regarding the limit, remember the form of the geometric series.
 
In fact, it's easier if you consider a matrix S, with \left\|S\right\|&lt;1 and prove that:

\sum_{n=0}^{\infty}S^n

Converges absolutely and compute the limit.
 
  • #10
JSuarez said:
In fact, it's easier if you consider a matrix S, with \left\|S\right\|&lt;1 and prove that:

\sum_{n=0}^{\infty}S^n

Converges absolutely and compute the limit.

There is a thm that says if spectral norm <1 then A^n -> 0 as n-> infinity.

and I proved above that spectral norm is <1

so I'm lost again...
 
Last edited:
  • #11
What can you say about the real series:
<br /> \sum_{n=0}^{\infty}\left\|S\right\|^n<br />
When \left\|S\right\|&lt;1? Does it converge? if yes, what's the sum? Is it related to ypur original series if S = I-T?
 

Similar threads

  • · Replies 4 ·
Replies
4
Views
3K
  • · Replies 11 ·
Replies
11
Views
2K
  • · Replies 2 ·
Replies
2
Views
1K
  • · Replies 8 ·
Replies
8
Views
1K
Replies
2
Views
1K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 3 ·
Replies
3
Views
2K
  • · Replies 1 ·
Replies
1
Views
1K
  • · Replies 18 ·
Replies
18
Views
2K
  • · Replies 1 ·
Replies
1
Views
2K