6
$\begingroup$

I'm trying to prove the $|\cdot|_{p}$ norm will become the maximum norm when $p \to \infty$.

Let $\mathbb K$ denote $\mathbb R$ or $\mathbb C$, and $x= (x_1, \ldots, x_m) \in \mathbb K^m$. Then $$\lim_{p \to \infty} \left ( \sum_{i=1}^m |x_i|^p \right )^{1/p} = \max _{1 \leq i\leq m} |x_{i}|$$

Could you please verify whether my attempt is fine or contains logical gaps/errors? Any suggestion is greatly appreciated!


My attempt:

It suffices to prove the statement in case $x \in \mathbb {(R^+)}^{m}$, where it becomes $$\lim_{p \to \infty} \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} = \max _{1 \leq i\leq m} x_{i}$$

Let $l:= \max _{1 \leq i\leq m} x_{i}$. We have $$l = (l^p)^{1/p} \le \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} \le (ml^p)^{1/p} = m^{1/p}l$$

Then $$l = \lim_{p \to \infty} l \le \lim_{p \to \infty} \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} \le \lim_{p \to \infty} (m^{1/p}l) = l$$

and thus by squeeze theorem $$\lim_{p \to \infty} \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} = l$$

This completes the proof.

$\endgroup$
  • $\begingroup$ Hi @AlexR., Please elaborate on why we should subscript $l_m$. I think I well defined $l$ by $l:= \max _{1 \leq i\leq m} x_{i}$. $\endgroup$ – LAD Aug 7 '19 at 4:56
  • 2
    $\begingroup$ @LeAnhDung I think Alex's comment is bad and that part of your proof is fine. I do, however, think your proof is flawed in that it works with limits before establishing the limit exists. You should be working with liminf's and limsup's $\endgroup$ – mathworker21 Aug 7 '19 at 5:02
  • 1
    $\begingroup$ @LeAnhDung I mean, it comes down to how you phrase it. Once you have $l \le (\sum x_i^p)^{1/p} \le m^{1/p}l$, you can say: since $\lim_{p \to \infty} l = l$ and $\lim_{p \to \infty} m^{1/p}l = l$, the squeeze theorem implies $\lim_{p \to \infty} (\sum x_i^p)^{1/p} = l$. That's what the squeeze theorem says. My objection in my previous comment was that you can't immediately go from$ l \le (\sum x_i^p)^{1/p} \le m^{1/p}l$ to $\lim_{p \to \infty} l \le \lim_{p \to \infty} (\sum x_i^p)^{1/p} \le \lim_{p \to \infty} m^{1/p}l$, since you don't know if the limits exist (yet). $\endgroup$ – mathworker21 Aug 7 '19 at 9:37
  • 1
    $\begingroup$ Yes it's correct. Good job! However, that is not what I suggested. I clearly suggested to use liminf's and limsup's. You should go from $l \le (\sum x_i^p)^{1/p} \le m^{1/p}l$ to $l = \liminf_p l \le \liminf_p (\sum x_i^p)^{1/p} \le \limsup_p (\sum x_i^p)^{1/p} \le \limsup_p m^{1/p}l = l$, so $\liminf_p (\sum x_i^p)^{1/p} = \limsup_p (\sum x_i^p)^{1/p}$ and thus the limit exists (and is then equal to $l$). $\endgroup$ – mathworker21 Aug 8 '19 at 13:41
  • 1
    $\begingroup$ @LeAnhDung no problem. upvoted your answer :) $\endgroup$ – mathworker21 Aug 8 '19 at 14:53
3
$\begingroup$

On the basis of @mathworker21's suggestion, I added a proof that $\lim_{p \to \infty} \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p}$ exists here. It would be great if someone helps me verify my attempt.


My attempt:

Lemma: $$\left( \sum_{i=1}^m (x_{i})^{p} \right)^{q} \ge \left( \sum_{i=1}^m (x_{i})^{q} \right)^{p}, \quad (x_1, \ldots,x_m) \in {(\mathbb R^+)}^m, \quad p,q \in \mathbb R^+, \quad p \le q$$

$$\text{Let } l:= \max _{1 \leq i\leq m} x_{i}.\ \text{Then } \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} \ge (l^p)^{1/p} = l.\ \text{As such, the sequence}$$

$$\left \langle \left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} \right \rangle_{p \in \mathbb N}$$ is bounded from below. Next we prove that this sequence is decreasing by showing $$\left ( \sum_{i=1}^m (x_i)^p \right )^{1/p} \ge \left ( \sum_{i=1}^m (x_i)^{p+1} \right )^{1/(p+1)}$$ or equivalently $$\left ( \sum_{i=1}^m (x_i)^p \right )^{p+1} \ge \left ( \sum_{i=1}^m (x_i)^{p+1} \right )^{p}$$, which is true by our Lemma. This completes the proof.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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