13
$\begingroup$

My question is simply the title:

What is the maximum possible value of determinant of a matrix whose entries either 0 or 1 ?

$\endgroup$
11
  • $\begingroup$ Since the determinant is the sum of $\,n!\,$ products of elements of the matrix chosen in a particular way, I'd say the maximum value is $\,n!\,$ ...but I've no example to offer. $\endgroup$
    – DonAntonio
    Jun 20, 2013 at 13:09
  • 3
    $\begingroup$ $n!$ is impossible. $\endgroup$
    – Mher
    Jun 20, 2013 at 13:13
  • $\begingroup$ There can't be any example for that $\endgroup$
    – Mher
    Jun 20, 2013 at 13:13
  • 1
    $\begingroup$ An upper bound that is smaller than $n!$ is $n^{n/2}$. Google Hadamard's inequality. $\endgroup$
    – KCd
    Jun 20, 2013 at 13:14
  • $\begingroup$ I can offer an example for $n-1.$ Taking $a_{ii}=0$ and $a_{ij}=1$ if $i\neq j.$ $\endgroup$
    – Mher
    Jun 20, 2013 at 13:16

2 Answers 2

7
$\begingroup$

Quoting my question in another thread:

In fact, I don't even know how large the determinant of a 0-1 matrix can be. The Hadamard's bound for the absolute determinant of an $n\times n$ 0-1 matrix is $\frac{(n+1)^{(n+1)/2}}{2^n}$ (online ref. 1 and ref. 2), and the bound is sharp if and only if there exists a Hadamard matrix of order $n+1$. Yet, to my knowledge, there is no known sharp upper bound for the absolute determinant of a general $n\times n$ 0-1 matrix.

$\endgroup$
2
  • $\begingroup$ Hm, can someone explain (or direct me to someplace) to understand why this answer is a community wiki? I've not dealt with community wikis before, so just curious. $\endgroup$
    – Calvin Lin
    Jun 20, 2013 at 14:44
  • $\begingroup$ @CalvinLin See an explanation here. In the present case I turned my answer into a community wiki one on purpose, because I didn't think simply quoting my own question verbatim deserves any reputation points. $\endgroup$
    – user1551
    Jun 20, 2013 at 15:13
5
$\begingroup$

A few examples of $\{0,1\}$-matrices (with the largest determinants $-$ according to OEIS-A003432):

$n=2$: $\quad\det\left( \begin{array}{cc} \bf{1} & 0 \\ \bf{1} & \bf{1} \\ \end{array} \right) =1;$

$n=3$: $\quad\det\left( \begin{array}{ccc} \bf{1} & 0 & \bf{1} \\ \bf{1} & \bf{1} & 0 \\ 0 & \bf{1} & \bf{1} \\ \end{array} \right) =2;$

$n=4$: $\quad\det\left( \begin{array}{cccc} \bf{1} & 0 & \bf{1} & 0 \\ \bf{1} & \bf{1} & 0 & \bf{1} \\ 0 & \bf{1} & \bf{1} & 0 \\ 0 & 0& \bf{1} & \bf{1} \\ \end{array} \right) =3;$

$n=5$: $\quad\det\left( \begin{array}{ccccc} \bf{1} & 0 & \bf{1} & 0 & 0\\ \bf{1} & \bf{1} & 0 & \bf{1} & 0 \\ 0 & \bf{1} & \bf{1} & 0 &\bf{1}\\ 0 & 0 & \bf{1} & \bf{1} & 0 \\ \bf{1} & 0 & 0 & \bf{1} & \bf{1} \\ \end{array} \right) =5;$

$n=6$: $\quad\det\left( \begin{array}{ccccc} \bf{1} & 0 & \bf{1} & 0 & 0 & 0\\ \bf{1} & \bf{1} & 0 & \bf{1} & 0 & 0 \\ 0 & \bf{1} & \bf{1} & 0 &\bf{1} & 0\\ 0 & 0 & \bf{1} & \bf{1} & 0 & \bf{1}\\ \bf{1} & 0 & 0 & \bf{1} & \bf{1} & 0\\ \bf{1} & \bf{1} & 0 & 0 & \bf{1} & \bf{1} \\ \end{array} \right) =9;$

$n=7$: $\quad\det\left( \begin{array}{ccccc} \bf1 & 0 & \bf1 & 0 & 0 & \bf1 & \bf1 \\ \bf1 & \bf1 & 0 & \bf1 & 0 & 0 & \bf1 \\ \bf1 & \bf1 & \bf1 & 0 & \bf1 & 0 & 0 \\ 0 & \bf1 & \bf1 & \bf1 & 0 & \bf1 & 0 \\ 0 & 0 & \bf1 & \bf1 & \bf1 & 0 & \bf1 \\ \bf1 & 0 & 0 & \bf1 & \bf1 & \bf1 & 0 \\ 0 & \bf1 & 0 & 0 & \bf1 & \bf1 & \bf1 \\ \end{array} \right) =32;$

$n=8$: $\quad\det\left( \begin{array} \bf1 & 0 & \bf1 & 0 & 0 & \bf1 & \bf1 & 0 \\ \bf1 & \bf1 & 0 & \bf1 & 0 & 0 & \bf1 & \bf1 \\ \bf1 & \bf1 & \bf1 & 0 & \bf1 & 0 & 0 & \bf1 \\ 0 & \bf1 & \bf1 & \bf1 & 0 & \bf1 & 0 & 0 \\ 0 & 0 & \bf1 & \bf1 & \bf1 & 0 & \bf1 & 0 \\ \bf1 & 0 & 0 & \bf1 & \bf1 & \bf1 & 0 & \bf1 \\ 0 & \bf1 & 0 & 0 & \bf1 & \bf1 & \bf1 & 0 \\ 0 & 0 & \bf1 & 0 & 0 & \bf1 & \bf1 & \bf1 \\ \end{array} \right) =56;$

$\endgroup$
7
  • 1
    $\begingroup$ Very nice! I was also looking for patterns of symmetry but didn't search too long... (I stopped after having brute-forced $n=4$ with $2^{16}$ variants of the matrix ... ;-)) $\endgroup$ Jun 21, 2013 at 9:47
  • $\begingroup$ Yes, we can find first 8 matrices among striped matrices ($\searrow \searrow \searrow$). What about others $n$ ??? $\endgroup$
    – Oleg567
    Jun 21, 2013 at 9:51
  • $\begingroup$ Impressive! Were you looking only among the striped ones? For $n=9$ you might need $17$ stripes, so you did not reach the largest value from the list? $\endgroup$
    – orangeskid
    Jul 6, 2022 at 18:25
  • 1
    $\begingroup$ @orangeskid: I looked among the striped shapes just to reduce the search area. The striped matrices provide maximum value for determinant only for small $n$ ($n\le 8$): $$n=9: \qquad det_{max}(striped) =125 \textrm{ vs } det_{max}=144; $$ $$n=10: \qquad det_{max}(striped)=312 \textrm{ vs } det_{max} = 320;$$ $$n=11: \qquad det_{max}(striped)=1215 \textrm{ vs } det_{max}=1458;$$ $\endgroup$
    – Oleg567
    Jul 7, 2022 at 13:40
  • 1
    $\begingroup$ @Oleg567: So interesting! Btw, just found this. $\endgroup$
    – orangeskid
    Jul 8, 2022 at 10:54

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .