Skip to main content
Post Reopened by Alex Degtyarev, Stefan Kohl, Joonas Ilmavirta, Johannes Hahn, Joseph O'Rourke
I reformulated my question because it was banned (and I do not understand why).
Source Link

An (open?) problem about a sequence of nested principal sub-matrices and their determinantdeterminants

I prefer to start with an example. Consider theProblem: Let $A$ be a $n \times n$ integer matrix $$ A = \left[ \begin{array}{ccc} 1 & 1 & 0 \\ 1 & 1 & 1 \\ 1 & 0 & 1 \end{array} \right] $$ It is invertible, since its determinant is $1$$\det(A) = \pm 1$. Now consider theUnder which conditions there exist a nested sequence of the upper-left square submatrices:principal submatrices of size $n$ such that they all have determinant $\pm 1$ ?

$$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,2 \right] = \left[ \begin{array}{cc} 1 & 1\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A $$ The second matrix is We call such a sequence a singular$\pm 1$ Principal Minor Sequence (its determinant is zero$\pm 1$PMS). However, we can find other sequence of nested "central" sub-matrices suchNote that all of them are invertible: $$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,3 \right] = \left[ \begin{array}{cc} 1 & 0\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A $$ I wonder when such a sequence exists. I need some notation to enunciate the problem.

Let $A$ beis described by a permutation of $n \times n$ matrix$n$ elements. Given $S \subset \left[ 1,n \right]$For example, we denote $A\left[ S \right]$consider the matrix obtained from $A$ by only taking the rows and columns whose index are in $$ A = \left[ \begin{array}{ccc} 1 & 1 & 0 \\ 1 & 1 & 1 \\ 1 & 0 & 1 \end{array} \right]. $$ The sequence $S$. Formally,$(1,2,3)$ $$ A\left[ S \right] = R_S \, A \, R^t_S, $$$$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,2 \right] = \left[ \begin{array}{cc} 1 & 1\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A. $$ whereis not a $S = \lbrace s_1 < \ldots < s_m \rbrace$ and$\pm 1$PMS since $R_s$ is the$\det(A \left[ 1,2 \right])=0$. The sequence $m \times n$ matrix with$(1,3,2)$ is a $1$ at$\pm 1$PMS. On the positionsother hand, the matrix $$ A = \left[ \begin{array}{ccc} 1 & 1 & 1 \\ 0 & 0 & 1 \\ 1 & 0 & 1 \end{array} \right] $$ does not have any $(i, s_i)$ and zero elsewhere$\pm 1$PMS.

Question: Let Does anybody know if this problem has already been studied? A related problem is to estimate the number of $A$ be$\pm 1$PMS of a regular $n \times n$ matrix. Under which conditions there exist a nested sequence $$ S_1 \subset S_2 \subset \cdots \subset S_n = \left[ 1,n \right], \quad |S_i| = i$$ such that each $A\left[ S_i \right]$ is regular? How to compute such a sequence? How many sequences are there?

Note: It is trivial that the matrix must have a non-zero entry in its diagonal. I am particularly interested in the case wherematrix has no $\pm 1$PMS if all theits entries in the diagonal are non-zerozero.

Does anybody know if this problem has already been studied?

An (open?) problem about a sequence of nested sub-matrices and their determinant

I prefer to start with an example. Consider the matrix $$ A = \left[ \begin{array}{ccc} 1 & 1 & 0 \\ 1 & 1 & 1 \\ 1 & 0 & 1 \end{array} \right] $$ It is invertible, since its determinant is $1$. Now consider the sequence of the upper-left square submatrices:

$$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,2 \right] = \left[ \begin{array}{cc} 1 & 1\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A $$ The second matrix is singular (its determinant is zero). However, we can find other sequence of nested "central" sub-matrices such that all of them are invertible: $$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,3 \right] = \left[ \begin{array}{cc} 1 & 0\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A $$ I wonder when such a sequence exists. I need some notation to enunciate the problem.

Let $A$ be a $n \times n$ matrix. Given $S \subset \left[ 1,n \right]$, we denote $A\left[ S \right]$ the matrix obtained from $A$ by only taking the rows and columns whose index are in $S$. Formally, $$ A\left[ S \right] = R_S \, A \, R^t_S, $$ where $S = \lbrace s_1 < \ldots < s_m \rbrace$ and $R_s$ is the $m \times n$ matrix with $1$ at the positions $(i, s_i)$ and zero elsewhere.

Question: Let $A$ be a regular $n \times n$ matrix. Under which conditions there exist a nested sequence $$ S_1 \subset S_2 \subset \cdots \subset S_n = \left[ 1,n \right], \quad |S_i| = i$$ such that each $A\left[ S_i \right]$ is regular? How to compute such a sequence? How many sequences are there?

It is trivial that the matrix must have a non-zero entry in its diagonal. I am particularly interested in the case where all the entries in the diagonal are non-zero.

Does anybody know if this problem has already been studied?

An (open?) problem about a sequence of nested principal sub-matrices and their determinants

Problem: Let $A$ be a $n \times n$ integer matrix, $\det(A) = \pm 1$. Under which conditions there exist a nested sequence of principal submatrices of size $n$ such that they all have determinant $\pm 1$ ?

We call such a sequence a $\pm 1$ Principal Minor Sequence ($\pm 1$PMS). Note that such a sequence is described by a permutation of $n$ elements. For example, consider the matrix $$ A = \left[ \begin{array}{ccc} 1 & 1 & 0 \\ 1 & 1 & 1 \\ 1 & 0 & 1 \end{array} \right]. $$ The sequence $(1,2,3)$ $$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,2 \right] = \left[ \begin{array}{cc} 1 & 1\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A. $$ is not a $\pm 1$PMS since $\det(A \left[ 1,2 \right])=0$. The sequence $(1,3,2)$ is a $\pm 1$PMS. On the other hand, the matrix $$ A = \left[ \begin{array}{ccc} 1 & 1 & 1 \\ 0 & 0 & 1 \\ 1 & 0 & 1 \end{array} \right] $$ does not have any $\pm 1$PMS.

Does anybody know if this problem has already been studied? A related problem is to estimate the number of $\pm 1$PMS of a matrix.

Note: It is trivial that a matrix has no $\pm 1$PMS if all its entries in the diagonal are zero.

Post Closed as "Not suitable for this site" by Will Jagy, coudy, Dima Pasechnik, Ricardo Andrade, Joonas Ilmavirta
edited tags
Link
darij grinberg
  • 33.8k
  • 4
  • 118
  • 253
Source Link

An (open?) problem about a sequence of nested sub-matrices and their determinant

I prefer to start with an example. Consider the matrix $$ A = \left[ \begin{array}{ccc} 1 & 1 & 0 \\ 1 & 1 & 1 \\ 1 & 0 & 1 \end{array} \right] $$ It is invertible, since its determinant is $1$. Now consider the sequence of the upper-left square submatrices:

$$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,2 \right] = \left[ \begin{array}{cc} 1 & 1\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A $$ The second matrix is singular (its determinant is zero). However, we can find other sequence of nested "central" sub-matrices such that all of them are invertible: $$ A \left[ 1 \right] = \left[ \begin{array}{c} 1 \end{array} \right], \quad A \left[ 1,3 \right] = \left[ \begin{array}{cc} 1 & 0\\ 1 & 1 \end{array} \right], \quad A \left[ 1,2,3 \right] = A $$ I wonder when such a sequence exists. I need some notation to enunciate the problem.

Let $A$ be a $n \times n$ matrix. Given $S \subset \left[ 1,n \right]$, we denote $A\left[ S \right]$ the matrix obtained from $A$ by only taking the rows and columns whose index are in $S$. Formally, $$ A\left[ S \right] = R_S \, A \, R^t_S, $$ where $S = \lbrace s_1 < \ldots < s_m \rbrace$ and $R_s$ is the $m \times n$ matrix with $1$ at the positions $(i, s_i)$ and zero elsewhere.

Question: Let $A$ be a regular $n \times n$ matrix. Under which conditions there exist a nested sequence $$ S_1 \subset S_2 \subset \cdots \subset S_n = \left[ 1,n \right], \quad |S_i| = i$$ such that each $A\left[ S_i \right]$ is regular? How to compute such a sequence? How many sequences are there?

It is trivial that the matrix must have a non-zero entry in its diagonal. I am particularly interested in the case where all the entries in the diagonal are non-zero.

Does anybody know if this problem has already been studied?