Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. Join them; it only takes a minute:

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

Let $M$ be a real $n\times n$ matrix and let $\|M\|_1 = \sum_{ij} |M_{ij}|$ ("the entry-wise 1-norm").

My question is: How well can we bound $\|M\|_1 $ in terms of $|\det M|$ (from below)?

Here is an obvious bound: Let $m = \operatorname{max}_{ij} |M_{ij}|$.

Then it is obvious that $$|\det M| \leq \sum_{\sigma \in S_n} m^n = n! m^n \quad or \quad \|M\|_1 \geq m \geq (|\det M|/n!)^{\frac1n}.$$ Can this bound be improved?

(In fact, the above result is strong enough for my purposes, but this question came out of curiosity. I have played around for a while but I couldn't improve the bound, nor achieve it.)

share|cite|improve this question
Watch it ; you've forgotten to sum over all entries of your matrix when computing your lower bound ; you actually get $ m \ge (|\det M|/n!)^{1/n} $, so that $\| M \| _1 \ge n^2 m \ge n^2 (| \det M | / n! )^{1/n}$. – Patrick Da Silva May 23 '11 at 20:47
@Patrick Da Silva: You think so? $m$ is the maximum, not the minimum of the entries, in fact $n^2 m$ should be larger than $\| M\|_1$. – Myself May 23 '11 at 20:52
Got tired here... haha good. – Patrick Da Silva May 23 '11 at 21:08
up vote 2 down vote accepted

If we denote by $v_1,\ldots,v_n$ the column of $M$, we have $|\det M|\leq \prod_{j=1}^n\lVert v_j\rVert_2$. But $\lVert v_j\rVert_2^2 =\left(\sum_{k=1}^n|m_{kj}|\right)^2-2\sum_{j<k}|m_{jk}|\leq \left(\sum_{k=1}^n|m_{kj}|\right)^2$ hence $\lVert v_j\rVert_2\leq \sum_{k=1}^n|m_{jk}|\leq \lVert M\rVert_1$. We finally get that $|\det M|^{\frac 1n}\leq \lVert M\rVert_1$ which is a better bound.

share|cite|improve this answer
That is a better bound! Is the first inequality well-known or obvious or something? The rest I agree with. (By the way, note that $\|v_j \| \leq \|M\|_1$ is obvious, in fact $\|M\|_1 = \sum_j \|v_j\|$.) – Myself May 23 '11 at 20:59
I should have remembered that, thanks. Apparently $\| v_i\| = \| v_i\|_2$ here so my last comment should be read as $\|M\|_1 = \sum_j \|v_j\|_1 \geq \|v_j\|_2$. – Myself May 23 '11 at 21:21
$|\det M| \le \prod_{j=1}^n \|v_j\|_2$ because $|\det M|$ is the $n$-dimensional volume of a parallelepiped whose edges correspond to the vectors $v_j$. Now by the AM-GM inequality we get $|\det M|^{1/n} \le \frac{1}{n} \sum_j \|v_j\|_2$. Finally, $\|v_j\|_2 \le \|v_j\|_1$, so the result is $\|M\|_1 \ge n |\det M\|^{1/n}$. And this is best possible, as seen by taking $M=I$. – Robert Israel May 23 '11 at 23:42
@Robert Israel: It appears I missed your comment wich gives an optimal bound. If you could post it as an answer, I'd be glad to accept it. – Myself May 31 '11 at 18:38

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.