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 the following details about the matrices $A,B$:

$$A_{n\times n},B_{n\times n}$$ $$A^2 = B^2 = 0$$ $$AB=BA$$

I need to find $x \in \mathbb{N}$ so $(A+B)^x=0$

What implications can I make from the given details?

  1. $A,B$ doesn't have to equal $0$ so $A^2 = B^2 = 0$
  2. Can I imply from the 2nd and 3rd given details that $A=B$?
  3. Any other leads?
share|improve this question
2  
Write out $(A+B)^x$ for $x=2$ and $x=3$ and simplify as far as you can. You should notice something nice. You can use the binomial theorem for these powers before simplification (because $AB=BA$). Then try $x=4$ using your result from $x=2$. –  bgins Apr 6 '12 at 11:42
    
Concering 2.: $B$ might be a scaled version of $A=\begin{pmatrix}0& 1\\0 &0\end{pmatrix}$, so no. –  draks ... Apr 6 '12 at 11:49

3 Answers 3

up vote 1 down vote accepted

For all $x \ge 3$, The equation $(A+B)^x=0$ is true because for $x =3$, it is true.

share|improve this answer

$$\mbox{ let } x > 2\mbox{ and } x\in \mathbb{N}$$ $$\textbf{(A+B)}^2 = \textbf{A}^2+\textbf{B}^2+2\textbf{AB} = \textbf{2AB}$$ $$\textbf{(A+B)}^x = \textbf{(A+B)}^2\textbf{(A+B)}^{x-2}$$ $$\Rightarrow \textbf{(A+B)}^x =2\textbf{AB}\textbf{(A+B)}^{x-2}$$ $$\Rightarrow \textbf{(A+B)}^x =2\textbf{AB}\textbf{(A+B)}\textbf{(A+B)}^{x-3}$$ $$\Rightarrow \textbf{(A+B)}^x =2(\textbf{ABA} + \textbf{ABB})\textbf{(A+B)}^{x-3}$$ $$\Rightarrow \textbf{(A+B)}^x =2(\textbf{AAB} + \textbf{ABB})\textbf{(A+B)}^{x-3}\quad(\because \textbf{BA = AB})$$ $$\Rightarrow \textbf{(A+B)}^x = \textbf{0} (\because\hspace{2pt}\textbf{A}^2 = \textbf{B}^2 = 0)$$ $$\text{If x = 2, I can not show }\textbf{(A+B)}^2 = 0$$ $$\text{ can anyone to prove that or give a counter example}$$

share|improve this answer
    
I have added a counterexample showing that $(A+B)^2=0$ is not true in general. The counterexample uses $4\times4$ matrices. I have also tried to show that a counterexample cannot be found for smaller matrices. –  Martin Sleziak Jan 30 at 12:24

In another answer it was already shown that $(A+B)^3=0$ for any matrices fulfilling the given conditions. Let us give an example showing that in general the given conditions do not imply $(A+B)^2=0$.$\newcommand{\Ker}{\operatorname{Ker}}\newcommand{\inv}[1]{{#1}^{-1}}$


Consider the following $4\times 4$ matrices: $$A= \begin{pmatrix} 0 & 0 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \end{pmatrix} \qquad B= \begin{pmatrix} 0 & 1 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \end{pmatrix} $$ We have $$AB=BA= \begin{pmatrix} 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \end{pmatrix} $$ and $A^2=B^2=0$.

Then we get $$(A+B)^2= \begin{pmatrix} 0 & 1 & 1 & 0 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 1 \\ 0 & 0 & 0 & 0 \end{pmatrix}^2= \begin{pmatrix} 0 & 0 & 0 & 2 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 \end{pmatrix} $$ and $(A+B)^3=0$. (You may notice that $(A+B)^2=2AB=2BA$ which was derived in bsdshell's answer.)

This counterexample works for matrices over any field which does not have characteristic 2. If $\operatorname{char} F=2$ then we have $(A+B)^2=2AB=2BA=0$.


Now let me show how I found this counterexample. This argument also shows (if I have not missed something), that we cannot find a counterexample for $n<4$.

I will use row vectors. (If you are used to column vectors, you should modify some parts of the proof.) This means that to a $n\times n$ matrix $A$ over a field $F$ corresponds a linear map $f_A \colon F^n \to F^n$ given by $f_A(\vec x)=\vec xA$. The kernel of this map is $\Ker{f_A}=\{\vec x\in F^n; \vec xA=\vec 0\}$. The set $\Ker f_A$ is a subspace of $F^n$ and we can write the whole space as $F^n=S\oplus \Ker f_A$.

The condition $A^2=0$ means that $f_A\circ f_A=0$, i.e., $f_A(f_A(\vec x))=\vec 0$ for any vector $\vec x\in F^n$.

Let $\vec x_1,\dots,\vec x_l$ be any basis of the subspace $S$. Then vectors $f_A(\vec x_1),\dots,f_A(\vec x_l)$ belong to $\Ker f_A$. Moreover, they are linearly independent. (Indeed, $c_1f_A(\vec x_1)+\dots+c_lf_A(\vec x_l)=\vec0$ implies $f_A(c_1\vec x_1+\dots+c_l\vec x_l)=\vec0$ and thus we get $c_1\vec x_1+\dots+c_l\vec x_l\in\Ker f_A\cap S=\{\vec 0\}$. This implies $c_1=\dots=c_l=0$.) This implies that $\dim\Ker f_A\ge l$ and we can choose a basis for $\Ker f_A$ of the form $\vec x_{l+1}=f_A(\vec x_1),\dots,\vec x_{2l}=f_A(\vec x_l),\vec x_{2l+1},\dots,\vec x_n$.

Notice that the matrix of the map $f_A$ with respect to this basis is $$A_1= \begin{pmatrix} 0 & I & 0\\ 0 & 0 & 0\\ 0 & 0 & 0 \end{pmatrix}$$ where the 4 blocks in the upper left corner are all $l\times l$ and the rightmost bottom block is $(n-2l)\times(n-2l)$.

Since the matrix $A_1$ is similar to the matrix $A$, we have $A_1=PA\inv P$ for some invertible matrix $P$. If we choose $B_1=PB\inv P$, then we have $A_1B_1=PAB\inv P=PBA\inv P=B_1A_1$ and $A_1^2=PA^2\inv P=0=PB^2\inv P=B_1^2$. We also have $A+B=\inv P(A_1+B_1)P$ and $(A+B)^2=\inv P(A_1+B_1)^2P$. So the question whether we can find the matrices fulfilling given conditions such that $(A+B)^2\ne0$ is equivalent to the question whether we can find such matrices where $A$ already has a special form $A=A_1= \begin{pmatrix} 0 & I & 0\\ 0 & 0 & 0\\ 0 & 0 & 0 \end{pmatrix}$; i.e., we can work with $A_1$, $B_1$ instead of $A$, $B$.

Now we can divide the matrix $B_1$ into blocks of the same sizes. $$B_1= \begin{pmatrix} B_{11} & B_{12} & B_{13} \\ B_{21} & B_{22} & B_{23} \\ B_{31} & B_{32} & B_{33} \end{pmatrix} $$

Then we get $$A_1B_1= \begin{pmatrix} B_{21} & B_{22} & B_{23} \\ 0 & 0 & 0\\ 0 & 0 & 0 \end{pmatrix}= \begin{pmatrix} 0 & B_{11} & 0\\ 0 & B_{21} & 0\\ 0 & B_{31} & 0 \end{pmatrix} =B_1A_1$$ which implies that
$$B_{11}=B_{22}\qquad\text{ and }\qquad B_{21}=B_{23}=B_{31}=0,$$
so we have $$B_1= \begin{pmatrix} B_{11} & B_{12} & B_{13} \\ 0 & B_{11} & 0 \\ 0 & B_{32} & B_{33} \end{pmatrix}. $$

We also know that $B_1^2=0$ which implies $B_{11}^2=0$.

If we are working with $2\times2$ or $3\times3$ matrices, then $B_{11}$ must be $1\times1$ and $B_{11}^2=0$ already implies $B_{11}=0$. This, in turn implies $(A_1+B_1)^2=A_1^2+2B_1A_1+B_1^2=2B_1A_1= 2\begin{pmatrix} 0 & B_{11} & 0\\ 0 & B_{21} & 0\\ 0 & B_{31} & 0 \end{pmatrix} =0$. So no counterexample for $n\le3$ can be found.

The counterexample for $n=4$ was constructed simply by taking $B_{22}=B_{11}$ equal to some non-zero $2\times2$ matrix with $B_{11}^2=0$; and all other blocks were chosen to be zero. (In this case $l=2$ and $n=2l=4$.)


NOTE: Originally I had here more complicated proof which only worked for $2\times2$ matrices and used Jordan normal form.

share|improve this answer
    
What I have written above is solution only for $2\times2$ matrices. I guess that similar approach might work for $n\times n$ matrices, but it seems that the proof is becoming somewhat messy. –  Martin Sleziak Jan 29 at 13:31
    
Nice try, Looking for a simple proof. –  bsdshell Jan 30 at 1:58
    
Note: The above comments refer to my first attempt. (I have rewritten my post completely since then.) –  Martin Sleziak Jan 30 at 12:41

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.