Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have a matrix which is of odd order and has exactly two ones in each row and column. The rest of the entries in each row/column are all zero. What will be the determinant of this matrix?

I believe the answer is $\pm 2$. My question is as to how is this derived?

Any help will be appreciated. Thanks.

share|improve this question

4 Answers 4

up vote 5 down vote accepted

As described by examples in J.M.'s answer and Paul's comment to Martin's answer, it is possible to get zero and $\pm2$. But that is not all of it. Consider matrices of the form (thanks to Martin for pointing out that I need to use an odd size) $$ A=\pmatrix{P&0&0\cr0&P&0\cr0&0&P\cr}, $$ where $P$ is the matrix from Paul's comment $$ P=\pmatrix{1&1&0\cr1&0&1\cr0&1&1\cr}. $$ Obviously $\det A=-8$, and it is clear that we can get any odd power of two in this way. I believe this covers all the possibilities: $0,\pm2^k$ for some odd natural number $k>0$.

Edit: Here's a proof that $\pm 2^{2t+1}, t\in\mathbf{N},$ and $0$ are all the possibilities. Do the following: Pick one of the 1s on row number $i_1$. Check the location of the other 1 on that row. Let the other 1 on that column be on row $i_2$. Keep doing this horizontal-vertical motion to define a set of rows $i_3,\ldots$. Sooner or later you get back to row number $i_1$. We have thus identified a subset of rows $i_1,\ldots,i_k$ for some integers $k>1$. By reordering the rows, we can bring these to the top. By reordering the columns we get a block looking exactly like the matrices in J.M.'s answer in the top-left corner. That block has determinant $2$ or $0$ as described by J.M. - according to the parity of $k$.

We may or may not have exhausted all the rows of the matrix. If we have not covered all of it, we build another block in a similar fashion. In the end we will have an odd number of blocks with an odd number of rows. The parity of the necessary row/column permutations gives us the sign.

share|improve this answer
1  
Your example does not have odd dimensions, which is one of the conditions in the question. (But this could be considered an interesting generalization of the question.) –  Martin Sleziak Dec 27 '11 at 11:35
1  
@Martin: Oops, I missed that condition. Fixing with 3 copies of $P$ to give $\det=-8$. –  Jyrki Lahtonen Dec 27 '11 at 11:38
2  
Another way of looking at the partitioning into blocks: form a colored graph with the entries $=1$ as nodes. Connect two nodes with a red arc, if they are on the same row. Connect two nodes with a blue arc, if they are on the same column. All the nodes have weight 2 in this graph. The graph is not a tree, because its number of arcs is equal to its nodes, so it has a cycle. Along that cycle the arcs alternate in colors: red-blue-red-..., so the cycle is of an even length. The rows/columns included in such a cycle form one of the blocks. Rinse. Repeat. –  Jyrki Lahtonen Dec 27 '11 at 19:58
    
David, J.M., thanks for fixing my typos and punctuation, guys! –  Jyrki Lahtonen Dec 30 '11 at 14:47

NOTE: I originally thought that the possible values are $0$ or $\pm2$ and I've tried to correct my approach after seeing Jyrki's answer and comments. I do not claim that my answer is substantially different from his, but I decided to keep it, since it still might be useful for some users.

The determinant can be $0$ or $(\pm2)^k$. We will show this for all dimensions (not only odd ones).


We can show this by induction on $n$ for any $n\times n)$-matrix.

For $n=1,2,3$: By inspection.

Inductive step. Suppose that the claim is true for smaller matrices and we work wit $(n+1)\times(n+1)$-matrix of this form.

The first row $\vec a_1$ contains two ones. There are two possibilities.

a) There is a row $\vec a_i=\vec a_1$ for $i\ne 1$. Then the determinant is zero. (Since two rows of the matrix are identical.)

b) Find rows $\vec a_j$ and $\vec a_k$ containing ones in the same columns, where $\vec a_1$ has them. First suppose that the second $1$'s in the rows $\vec a_j$ and $\vec a_k$ are in different columns.

This means that $\vec a_j+\vec a_k-\vec a_1$ is a vector containing precisely two $1$'s.

We replace one of the three rows with the vector $\vec a_j+\vec a_k-\vec a_1$. This might change the sign of the determinant (depending on the row we choose.)

Now we get (after reordering rows and columns, which only changes the sign) a block matrix. One of the blocks is a $(n-1)\times(n-1)$-matrix which fulfills all the assumptions. The whole matrix looks like this
$$\begin{vmatrix} 1 & 1 & 0 & \dots & 0\\ 1 & 0 & 1 & \dots & 0\\ 0 & 0 & * & * & * \\ 0 & 0 & * & * & * \\ 0 & 0 & * & * & * \end{vmatrix}$$ where stars denote the submatrix fulfilling inductive hypothesis.

If we use Laplace expansion with respect to the first row, one of the determinants has zero column, hence it is zero and another one is the determinant of $(2n-1)\times(2n-1)$ submatrix.

c) Now suppose that the row $\vec a_j$ and $\vec a_k$ have the second $1$'s in the same column. This means that after reordering rows and columns we get a from these three rows and corresponding columns the submatrix of the form $\begin{pmatrix} 1&1&0\\ 1&0&1 \\ 0&1&1 \end{pmatrix}$. This means we have matrix of the form $$\begin{pmatrix} 1&1&0&0&\ldots&0&\\ 1&0&1&0&\ldots&0&\\ 0&1&1&0&\ldots&0&\\ 0&0&0&*&\ldots&*&\\ 0&0&0&*&\ldots&*&\\ 0&0&0&*&\ldots&*& \end{pmatrix}$$ where the submatrix marked by stars has form from the inductive hypothesis. So the determinant is the determinat of submatrix multiplied by $\begin{vmatrix} 1&1&0\\ 1&0&1 \\ 0&1&1 \end{vmatrix}=-2$.


Example:

$\begin{vmatrix} 1 & 1 & 0 & 0 & 0\\ 1 & 0 & 1 & 0 & 0\\ 0 & 1 & 0 & 1 & 0\\ 0 & 0 & 1 & 0 & 1\\ 0 & 0 & 0 & 1 & 1 \end{vmatrix}= \begin{vmatrix} 1 & 1 & 0 & 0 & 0\\ 1 & 0 & 1 & 0 & 0\\ 0 & 0 & 1 & 1 & 0\\ 0 & 0 & 1 & 0 & 1\\ 0 & 0 & 0 & 1 & 1 \end{vmatrix} =1 \begin{vmatrix} 0 & 1 & 0 & 0\\ 0 & 1 & 1 & 0\\ 0 & 1 & 0 & 1\\ 0 & 0 & 1 & 1 \end{vmatrix} -1 \begin{vmatrix} 1 & 1 & 0 & 0\\ 0 & 1 & 1 & 0\\ 0 & 1 & 0 & 1\\ 0 & 0 & 1 & 1 \end{vmatrix} $

(In the first step we subtracted the first row from the third on and added the second row to the third one, the second step is Laplace expansion.)


Note: Other way of saying this is to say that by interchanging rows and columns you will obtain one of two situations:

$\begin{vmatrix} 1 & 1 & 0 & 0 & 0\\ 1 & 1 & 0 & 0 & 0\\ 0 & 0 & 1 & \ldots &\ldots \\\ 0 & 0 & \vdots & & \\ 0 & 0 & \vdots & & \end{vmatrix}$ or $ \begin{vmatrix} 1 & 1 & 0 & 0 & 0\\ 1 & 0 & 1 & 0 & 0\\ 0 & 1 & 0 & 1 & 0\\ 0 & 0 & 1 & 0 & 1\\ 0 & 0 & 0 & 1 & 1 \end{vmatrix} $

(NOTE: As was pointed out by Jyrki, the matrix can continue after the block of this form. It is explained nicely in his answer.)

I.e. either a submatrix in upper left corner makes determinant zero (and we do not have to care about the remaining entries) or the matrix has special form, as the one given in the above example (and determinants of such matrices can be computed by various methods).

This is essentially the same thing as I've done in the inductive proof above, but perhaps a different viewpoint might be useful.

share|improve this answer
3  
But $\left| \begin{array}{ccc} 1 & 1 & 0 \\ 1 & 0 & 1 \\ 0 & 1 & 1 \\ \end{array} \right|=-2$. Did I miss something? –  Paul Dec 27 '11 at 10:59
    
@Paul I must have made mistake somewhere. I'll try to work this out. –  Martin Sleziak Dec 27 '11 at 11:01

I assume for the purposes of this answer that your matrices take the form

$$\begin{pmatrix}1&&&&1\\1&1&&&\\&1&1&&\\&&\ddots&\ddots&\\&&&1&1\end{pmatrix}=\mathbf H+\mathbf I$$

$\mathbf H$ is a Frobenius companion matrix, with characteristic polynomial $(-1)^n(x^n-1)$. The characteristic polynomial of $\mathbf H+\mathbf I$ is $p(x)=(-1)^n((x-1)^n-1)$ Thus, $\det(\mathbf H+\mathbf I)=p(0)=(-1)^n((-1)^n-1)=1-(-1)^n$.

The matrix itself is a special case of what is sometimes referred to as a "Forsythe matrix". It can be generated in MATLAB with the command gallery('forsythe',n,1,1).


Consider also the matrices

$$\mathbf B=\begin{pmatrix}1&1&&&\\1&0&1&&\\&1&\ddots&\ddots&\\&&\ddots&0&1\\&&&1&1\end{pmatrix}$$

It can be shown (though the relationship of tridiagonal determinants and three-term difference equations) that the characteristic polynomials of these matrices are given by

$$\det(\mathbf B-\lambda\mathbf I)=(-1)^n (\lambda-2)\frac{\sin\left(n\arccos\frac{\lambda}{2}\right)}{\sin\left(\arccos\frac{\lambda}{2}\right)}=(-1)^n (\lambda-2)U_{n-1}\left(\frac{\lambda}{2}\right)$$

where $U_k(x)$ is the Chebyshev polynomial of the second kind. $\det\mathbf B$ is thus given by $2(-1)^{n+1}\sin\dfrac{n\pi}{2}$.

share|improve this answer
    
My matrices may take any form satisfying the hypothesis. H + I is just a special case. Thanks anyway. –  Shahab Dec 27 '11 at 11:27
1  
Sure. To obtain all other matrices, one just pre- and postmultiplies $\mathbf H+\mathbf I$ by permutation matrices, whose determinants take the form $(-1)^n$... –  J. M. Dec 27 '11 at 11:38
    
Can I get all such matrices this way? I sort of see that intutively but how do I prove that. What about a matrix with two identical rows? –  Shahab Dec 27 '11 at 11:50
1  
Hmm, apparently not. $\begin{vmatrix}1 & 1 & 0 & 0 & 0 \\ 1 & 1 & 0 & 0 & 0 \\ 0 & 0 & 1 & 0 & 1 \\ 0 & 0 & 1 & 1 & 0 \\0 & 0 & 0 & 1 & 1\end{vmatrix}=0$. The matrix can be treated as a block diagonal matrix, with a singular block and a Forsythe block. Hmm... –  J. M. Dec 27 '11 at 11:56
1  
@Shabab: You get all the matrices from diagonal blocks of this type by row/column permutations. My answer seeks to outline why that happens. –  Jyrki Lahtonen Dec 27 '11 at 12:15

Only matrix from $n\times n$ type have a determinant. I can't understand you very well, but if your matrix is like $\begin{vmatrix} a&b \\ c&d \end{vmatrix}$ then your determinant is $ad-cb$.

P.S. I don't know how to make it to look like matrix here.

share|improve this answer
    
I've edited your answer to show you how to use TeX in your post. You can find more information by clicking on help (question mark) in edit window; see here and here. (But anyway, I do not think that this answers the posted question.) –  Martin Sleziak Dec 27 '11 at 10:55
    
Thanks mate, I have tried to make my answer simple and on topic. I really appreciate your help. I really like your answer but you have to include in your answer how to find a triangle matrix determinant and matrix type duck foot to make your answer complete. Regards, Yoni –  speedyGonzales Dec 27 '11 at 11:03
    
the matrix has order n x n of course. By odd order I meant that n is odd. For example, the matrix may be 3 x 3 or 5 x 5 etc. In addition each row contains 1 in exactly two places and 0 in the other n-2 places. Similarly for each column. –  Shahab Dec 27 '11 at 11:22

Your Answer

 
discard

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.