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.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Why is it that if $A(t), B(t)$ are two $n\times n$ complex matrices and $${d\over dt}A=AB-BA$$ then the trace of the matrix $A^n$ where $n\in \mathbb Z$ is a constant for all $t$?

share|cite|improve this question
up vote 1 down vote accepted
Note that Trace(FE)=Trace(EF) in general.

$n>0$ : Trace$(A^n)' = n [$Trace$ (A'(t) A^{n-1})] = n[ $Trace$ ((AB - BA)A^{n-1})] = 0$

$ n=0$ : $A^0 = I$ So we are done

$n <0$ : Check $(A^{-1})' = A^{-1} B - BA^{-1}$ So this case is reduced to the first case.

share|cite|improve this answer
Note in general that $(A^n)'\neq n A^{n-1}$. It is still true for the trace, but it isn't obvious. – Thomas Andrews Nov 9 '12 at 14:53
@ThomasAndrews : Would you mind explaining that a little bit? Thanks! – Ronald Crev Nov 9 '12 at 14:54
See Mercy's answer for the correct formula for $\frac d{dt} A^n$. The reason $\text{trace}((A^n)')=n\text{trace}(A'A^{n-1})$ is because $\text{trace}(XY)=\text{trace}(YX)$ so the terms in Mercy's formula all have equal trace. – Thomas Andrews Nov 9 '12 at 14:58

For every positive integer $m$ we have $$ \frac{d}{dt}A^m=A^{m-1}\dot{A}+A^{m-2}\dot{A}A+\ldots+A\dot{A}A^{m-2}+\dot{A}A^{m-1}, $$ where $$ \dot{A}=\frac{d}{dt}A. $$ Since $$ \text{trace}: M_n(\mathbb{R}) \to \mathbb{R},\ X \mapsto \text{trace}(X) $$ is linear and satisfies $$ \text{trace}(XY)=\text{trace}(YX) \quad \forall X,Y \in M_n(\mathbb{R}), $$ it follows that \begin{eqnarray} \frac{d}{dt}\text{trace}(A^m)&=&\text{trace}(\frac{d}{dt}A^m)=m\text{trace}(A^{m-1}\dot{A})\\ &=&m\text{trace}[A^{m-1}(AB-BA)]=m[\text{trace}(A^{m-1}AB)-\text{trace}(A^{m-1}BA)]\\ &=&m[\text{trace}(A^mB)-\text{trace}(BA^m)]=0. \end{eqnarray} Hence $$ \text{trace}(A^m(t))=\text{trace}(A^m(0)) \quad \forall t. $$ Notice that $A^m$ is not necessarily defined for $m<0$.

share|cite|improve this answer
You've used that $\text{tr}(A^{m-1}\dot{A})=\text{tr}(A^{m-1})\text{tr}(\dot{A})$, which appears not to be true. And you dont need that step, since $A^{m-1}\dot{A}=A^{m-1}(AB-BA)$ gives you what you want anyway. – Thomas Andrews Nov 9 '12 at 14:48
I see now, thanks! – Mercy King Nov 9 '12 at 14:53

For $n=1$. $$ \dot{ {\rm tr} \, A}={\rm tr}\, \dot{A} ={\rm tr}\, [A, B] =0 \ . $$

For $n=2$ $$ \dot{ {\rm tr} \, A^2}={\rm tr}\, \dot{A^2} ={\rm tr}\, (A \dot{A}+\dot{A} A) = $$ $$tr(A(AB-BA)+(AB-BA)A)=tr(A^2B-BA^2)=tr[A^2, B]=0 $$

More generally by an easy induction $$ \frac{d}{ dt} {\rm tr} \, A^n = {\rm tr} \, [A^n, B]=0 \ . $$

share|cite|improve this answer

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.