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.

Let $A$ be a $3\times 3$ matrix over $\mathbb {R}$ such that $AB =BA$ for all matrices $B$ over $\mathbb {R}$ then what can we say about such matrix $A$. or such matrix $A$ must be orthogonal matrix? Can we say anything about its eigen values? I tried by taking random examples also tried to construct such $3 \times 3$ matrices. But i am not able to get any proper conclusion and proof.

share|improve this question
$A$ must be a scalar multiple of the identity. This is a nice exercise. –  Qiaochu Yuan May 9 '12 at 7:48
can i get little hint please –  srijan May 9 '12 at 7:53
Choose $B$ to be some particularly simple matrices. –  Qiaochu Yuan May 9 '12 at 7:53
Take some trivial $B$'s like a $B$ will only one column non-zero and all the other entries being zero and you will get some constraints on the entries of $A$. –  user17762 May 9 '12 at 7:54
Do you know what $Ae_i$ and $e_i^tA$ look like for the column ($3\times1$) matrices $e_i$ having entry $1$ in the $i$th row and $0$ elsewhere? Now try making a $B$ from this somehow, say $B=(\delta_{ai}\delta_{bj})=e_ae_b^t$. –  bgins May 9 '12 at 8:22
show 5 more comments

4 Answers

up vote 3 down vote accepted

If we say that $A_i$ is the $i$th column of $A=(a_{ij})$ and $A^{(j)}$ is the $j$th row and $e_i$ is the $3\times1$ unit column vector with $1$ in the $i$th row (and consequently $0$ everywhere else), then for $B=e_ie_j^t$, which is $1$ at $(i,j)$ and $0$ elsewhere, we find that $$AB=Ae_ie_j^t=A_ie_j^t$$ and $$BA=e_ie_j^tA=e_iA^{(j)}$$ are both $3\times3$ matrices, the first with the $i$th column of $A$ in its $j$th column, the second with the $jth$ row of $A$ in its $i$th row. For these to be equal, we must have $a_{ij}=0$ whenever $i\ne j$. Furthermore, the single nonzero entry in the products above is again at $(i,j)$, where we find: $$a_{ii}=(AB)_{ij}=(BA)_{ij}=a_{jj}\,.$$ Thus $A=\lambda I$ must be a scalar (multiple of the identity) matrix.

So how does this look? Well, here goes: $$ e_1=\left[\matrix{1\\0\\0}\right],\quad e_2=\left[\matrix{0\\1\\0}\right],\quad e_3=\left[\matrix{0\\0\\1}\right], $$ $$ e_1^t=\left[\matrix{1&0&0}\right],\quad e_2^t=\left[\matrix{0&1&0}\right],\quad e_3^t=\left[\matrix{0&0&1}\right], $$ $$ e_1e_1^t=\left[\matrix{1&0&0\\0&0&0\\0&0&0}\right],\quad e_1e_2^t=\left[\matrix{0&1&0\\0&0&0\\0&0&0}\right],\quad e_1e_3^t=\left[\matrix{0&0&1\\0&0&0\\0&0&0}\right],\quad $$ $$ e_2e_1^t=\left[\matrix{0&0&0\\1&0&0\\0&0&0}\right],\quad e_2e_2^t=\left[\matrix{0&0&0\\0&1&0\\0&0&0}\right],\quad e_2e_3^t=\left[\matrix{0&0&0\\0&0&1\\0&0&0}\right],\quad $$ $$ e_3e_1^t=\left[\matrix{0&0&0\\0&0&0\\1&0&0}\right],\quad e_3e_2^t=\left[\matrix{0&0&0\\0&0&0\\0&1&0}\right],\quad e_3e_3^t=\left[\matrix{0&0&0\\0&0&0\\0&0&1}\right],\quad $$ $$ A=\left[\matrix{ a_{11}&a_{12}&a_{13}\\ a_{21}&a_{22}&a_{23}\\ a_{31}&a_{32}&a_{33} }\right],\quad A_1=\left[\matrix{a_{11}\\a_{21}\\a_{31}}\right],\quad A_2=\left[\matrix{a_{12}\\a_{22}\\a_{32}}\right],\quad A_3=\left[\matrix{a_{13}\\a_{23}\\a_{33}}\right], $$ and using a superscript for my nonstandard row notation, $$ A^{(1)}=\left[\matrix{a_{11}&a_{12}&a_{13}}\right],\quad A^{(2)}=\left[\matrix{a_{21}&a_{22}&a_{23}}\right],\quad A^{(3)}=\left[\matrix{a_{31}&a_{32}&a_{33}}\right]=(A^t)_3^t. $$ So for example with $i=2$ and $j=3$, we get $B=e_ie_j^t=e_2e_3^t$ as above so that $AB=A(e_2e_3^t)=(Ae_2)e_3^t$ since matrix multiplication is associative. Furthermore, $$ Ae_2= \left[\matrix{ a_{11}&a_{12}&a_{13}\\ a_{21}&a_{22}&a_{23}\\ a_{31}&a_{32}&a_{33} }\right] \cdot \left[\matrix{0\\1\\0}\right]= \left[\matrix{ a_{11}\cdot0 + a_{12}\cdot1 + a_{13}\cdot0\\ a_{21}\cdot0 + a_{22}\cdot1 + a_{23}\cdot0\\ a_{31}\cdot0 + a_{32}\cdot1 + a_{33}\cdot0 }\right] =\left[\matrix{a_{12}\\a_{22}\\a_{32}}\right] =A_2 $$ so that $$ AB=(Ae_2)e_3^t=A_2e_3^t =\left[\matrix{a_{12}\\a_{22}\\a_{32}}\right] \cdot\left[\matrix{0&0&1}\right] =\left[\matrix{ a_{12}\cdot0&a_{12}\cdot0&a_{12}\cdot1\\ a_{22}\cdot0&a_{22}\cdot0&a_{22}\cdot1\\ a_{32}\cdot0&a_{32}\cdot0&a_{32}\cdot1 }\right] =\left[\matrix{0&0&a_{12}\\0&0&a_{22}\\0&0&a_{32}}\right] $$ and, similarly (without all the gory details), $$ e_3^tA=\left[\matrix{0&0&1}\right]\cdot \left[\matrix{ a_{11}&a_{12}&a_{13}\\ a_{21}&a_{22}&a_{23}\\ a_{31}&a_{32}&a_{33} }\right]= \left[\matrix{a_{31}&a_{32}&a_{33}}\right] =A^{(3)} $$ so that $$ BA=(e_2e_3^t)A=e_2(e_3^tA)=e_2A^{(3)}= \left[\matrix{0\\1\\0}\right]\cdot \left[\matrix{ a_{31}&a_{32}&a_{33}} \right]= \left[\matrix{0&0&0\\a_{31}&a_{32}&a_{33}\\0&0&0}\right]. $$ Thus using $B=e_2e_3^t$, we can see from $AB=BA$ that $a_{22}=a_{33}$ and $a_{i2}=0=a_{3j}$ for $i\ne2,j\ne3$.

share|improve this answer
Note that this does not show $a_{ii}=a_{jj}$ for all $i,j$. –  Alex Becker May 9 '12 at 8:45
@AlexBecker: Thanks, I fixed it now. –  bgins May 9 '12 at 9:01
@bgins Can you explain how matrices $A$ and $B$ are being multiplied to each other. I am sorry i am not able to get you point. How this matrix $B=e_ie_j^t$ has come? –  srijan May 9 '12 at 9:41
I added a second section graphically depicting many of the the matrices, certainly enough to see what's going on. I also changed my nonstandard notation for the rows of $A$ by introducing parentheses around the superscripted row numbers. What you need to know are the definition of the $ij$th entry of a matrix, namely that it is the entry in row $i$ and column $j$, and the rule for matrix multiplication. –  bgins May 9 '12 at 20:36
add comment

Here is one way to prove this, a bit less elementary than what was suggested in the comments (but it avoids any computations). It is a fundamental fact (and easy to prove) that if $A,B$ are commuting matrices, then every eigenspace for $B$ must be $A$-stable (i.e., if $v$ is in the eigenspace of$~B$ for eigenvalue$~\lambda$, then so is $A\cdot v$). Also, a $1$-dimensional subspace $\langle v\rangle$ is $A$-stable precisely if $v$ is an eigenvector for $A$ (immediate from the definition).

On one hand if $A$ is a scalar multiple of identity, then it clearly commutes with every $B$ (by definition any linear operator commutes with scalar multiplications). On the other hand if $A$ is not a scalar multiple of identity, then there exists some vector $v$ that is not an eigenvector of $A$ (if every nonzero vector were eigenvector, one easily shows they all have the same eigenvalue $\lambda$, and $A$ would be $\lambda I$). Then choose $B$ such that $\langle v\rangle$ is an eigenspace for $B$ (easy). By the above, if $B$ would commute with $A$, then $\langle v\rangle$ would be $A$-stable, and therefore $v$ an eigenvector for $A$. But it isn't, so $A$ and $B$ do not commute.

Note that the facts that the base field is $\Bbb R$, that the dimension is$~3$ or even that it is finite are irrelevant to this question.

share|improve this answer
add comment

Here's another solution. Since $A$ is $3\times3$ and $3$ is odd, $A$ has a real eigenvalue $\lambda$ (if you're comfortable with complex eigenvalues, we don't even need the dimension to be odd.) We claim that $A=\lambda I$. To prove this, let $w = (w_1,w_2,w_3)$ be an arbitrary vector in $\mathbb R^3$; it suffices to show that $wA =\lambda w$.

Let $v$ be an eigenvector corresponding to $\lambda$. Some coordinate of $v$ must be nonzero, and $v$ can be scaled at will; without loss of generality, say the first coordinate of $v$ equals $1$. Now set $B$ to be the matrix whose columns are $w_1v$, $w_2v$, and $w_3v$. We easily compute that $AB$ is the matrix whose columns are $\lambda w_1v$, $\lambda w_2v$, and $\lambda w_3v$; in particular, the top row of $AB$ equals $(\lambda w_1,\lambda w_2,\lambda w_3) = \lambda w$. This is also the top row of $BA$, which equals the top row of $B$ times $A$, which equals $w$ times $A$. This establishes that $wA =\lambda w$ as desired.

share|improve this answer
add comment

Here is a simple exercise which I find useful few times.

Exercise Let $D$ be a diagonal matrix, with pairwise distinct diagonal entries. Then $D$ commutes only with diagonal matrices.

Proof: Let $CD=DC$. Equaling the $ij$ entries in these matrices we get:


Thus, if $i \neq j$we have $d_{ii} \neq d_{jj}$ and thus $c_{ij}=0$..

Now back to your problem. Pick $B$ a diagonal matrix with pairwise distinct diagonal entries. Then $AB=BA$ implies that $A$ is diagonal.

Next pick $B$ the matrix with all ones, that is $b_{ij}=1 \forall i,j$. $AB=BA$ and $A$ diagonal implies that all entries of $A$ are equal.

share|improve this answer
add comment

Your Answer


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.