Let $A = [a_{ij}]$ be an $n \times n$ matrix with entries in $\mathbb{R}$. Suppose there exists an $m$ with $a_{ij} = 0$ for $i \ge m$ and $j \le m$, and $a_{i,i} \ne 0$ for $1 \le i \lt m$. Show that $A$ has no inverse.

From my understanding no diagonal value is zero, but all the non-diagonal values more than some arbitrary $m$ is zero. And, I know I must show that the determinant is zero, but I'm not sure how to do this.


Note that you do have a diagonal entry equal to 0: $a_{m,m} = 0$. And possibly $a_{k,k} = 0$ for values of $k \geq m$.

The hypothesis is saying you have a block of zeroes in the bottom left of the matrix, which includes a diagonal entry. This should allow you to take a block decomposition of your matrix, $$ A = \begin{bmatrix} A_{1,1} & A_{1,2} \\ 0 & A_{2,2} \end{bmatrix}$$ where $A_{1,1}$ is $(m-1) \times (m-1)$ and has no zeroes on the diagonal. Then $A_{2,2}$ is $(n-(m+1))\times (n-(m+1))$, the top left entry of $A_{2,2}$ is $a_{m,m}$ (which is 0). In fact, the first column of $A_{2,2}$ consists of the entries $a_{i,j}$ where $i$ runs through $m, m+1, \dots, n$ and $j = m$ (your hypothesis gives you some info on these entries).

Now: Can you show that $A_{2,2}$ is not invertible? What does this tell you about the invertibility of $A$?

  • $\begingroup$ If $A_{2,2}$'s top-left element is zero, then it's determinant should also be zero after taking the co-factor expansion. However, can't we take a non-singular matrix, and generate a zero in that very place by using row operations? $\endgroup$ – Jaigus Mar 13 at 18:43
  • $\begingroup$ @Jaigus The comment you made is not true: simply having a diagonal entry equal to $0$ does not make the determinant $0$ (consider $\begin{bmatrix} 0 & 1\\1&0\end{bmatrix}$). But consider what is happening with the rest of the entries in the first column of $A_{2,2}$. $\endgroup$ – Morgan Rodgers Mar 13 at 19:30
  • $\begingroup$ So because the first column is all zeros, that means that the first column as a vector is linearly dependent, so $A_{2,2}$ is not full rank, and therefore singular? $\endgroup$ – Jaigus Mar 13 at 19:59
  • $\begingroup$ Yes absolutely, a full column of zeros will give zero determinant with a cofactor expansion also. $\endgroup$ – Morgan Rodgers Mar 13 at 20:24
  • 1
    $\begingroup$ @Jaigus It's really the singularity of $A_{2,2}$ along with the $0$ block that will tell you the whole thing is singular. Have you seen formulas for the determinant of $\begin{bmatrix} A&B\\0&D \end{bmatrix}$? If not, consider the rank of $A$. How many pivot positions can $A$ have in the top portion of the matrix $\begin{bmatrix} A_{1,1} & A_{1,2}\end{bmatrix}$? How many in the bottom portion $\begin{bmatrix} 0 & A_{2,2}\end{bmatrix}$? $\endgroup$ – Morgan Rodgers Mar 13 at 23:03

The determinant $\det(A)$ is a sum of products $$\pm\, a_{1\,\sigma(1)}\>a_{2\,\sigma(2)}\>\cdots\> a_{n\,\sigma(n)}\ ,\tag{1}$$ where $\sigma$ runs through the $n!$ permutations of $[n]$. For such a permutation $\sigma$ we cannot have $$m+1\leq \sigma(i)\leq n\quad(m\leq i\leq n)$$ (counting elements). It follows that in each product $(1)$ there is at least one factor $a_{i\,\sigma(i)}=0$. Therefore $\det(A)=0$; hence $A$ is singular.


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.