Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

I am given a lemma and it states:

Lemma: Let A be a real n x n matrix. Then given any e > 0, there is a norm such that norm(A) <= p(A) + e Where p(A) is the spectral radius of A.

They then go on to state that based on this lemma, if p(A) < 1, then norm(A) < 1 for the correct choice of the norm.

What I'm looking for help on is understanding how they came to that conclusion based only on that lemma. For example, given e = .001, then the lemma states there is a norm such that norm(A) <= p(A) + .001

Since it is a "or equal" then this means there is the possibility that norm(A) = p(A) + .001, and in such a case then norm(A) > p(A). So it seems that based on the lemma, then p(A) < 1 does not necessarily always imply that norm(A) < 1.

I even tried to do a proof of this but still came up with the same result:

p(A) < 1 -> 
p(A) + e < 1 + e -> 
norm(A) <= p(A) +e < 1 + e
norm(A) - e <= p(A) < 1

This pretty much says the same thing, if norm(A) - e <= p(A), then p(A) < 1, but if p(A) close to 1 by a difference less than e, then norm(A) > 1 in the "equal to" case where norm(A) - e = p(A)

Now of course you could pick a smaller e, but the way the lemma reads it says that for a given e, you can find a norm s.t. A <= p(A) + e, so the "or equal" part will still get you using a counter example I gave similar to the e = .001 above

Now I have seen a different proof on wikipedia that shows if p(A) < 1 then norm(A) < 1, so I am not disputing that fact, but that proof gets into looking at entries of the Jordan Normal Form. So my point is, yes p(A) < 1 then norm(A) < 1, BUT I don't understand how you can conclude that based only on the Lemma I gave at the beginning. It's these big jumps in rational that confuse me.

share|cite|improve this question
The flaw in your reasoning is you're trying to think of $e$ as a fixed value independent of $A$, which clearly doesn't work. For any given matrix $A$, you have to choose $e$ the right way, as Jonas' answer explains. – Rahul Oct 26 '10 at 3:57
I see. I thought it was the opposite based on the wording of the lemma, that "given any e", then "there is a norm such that" i.e. there is a norm you can pick that satisfies the inequality. Is the lemma worded poorly or am I reading it wrong? – AaronLS Oct 26 '10 at 5:01
It is worded correctly. I think the problem is that you're thinking too much about what it says for some particular choices of $e$. As you said, when $e=.001$, the lemma will give you a norm(A) $\leq$ p(A) +.001. But when you said it might be equal, that is not a problem, because you can just take a smaller $e$ and apply the lemma to get another norm that does what you want. – Jonas Meyer Oct 26 '10 at 5:09
up vote 4 down vote accepted

Apply the lemma with $e=\frac{1}{2}(1-p(A))$.

Also, the fact that it is not strict inequality is insignificant. Given $e$, you could apply the lemma with $e/2$ to get strict inequality with $e$.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.