Matrix derivative: a useful identity
**Lemma 1** Suppose $A(t)$ and $M(t)$ are square matrices satisfying
$$\f{d}{dt}A(t) = M(t)A(t).$$
Then
$$\f{d}{dt}(\text{ln(det(}A(t))) = \text{Trace}(M(t)).$$
//Proof// We write $A=(a_{ij})$ and $M=(m_{ij})$, and let $A_{ij}$ be the [[http://en.wikipedia.org/wiki/Minor_(linear_algebra)|co-factor]] of $a_{ij}$. Then standard linear algebra tells that
$$(3)\hspace{30pt}\sum_{j}a_{ij}A_{kj} = \delta_{ik}\text{det}(A).$$
By direct computation, we have
$$\f{d (\text{det}(A)) }{dt} = \sum_{i,j}\f{d a_{ij}}{dt} A_{ij}.$$
We then calculate
\begin{equation}
\begin{split}
\f{d}{dt}(\text{ln(det(}A(t)))&=\text{det}(A)^{-1}\sum_{i,j} \f{d a_{ij}}{dt} A_{ij}~\\
&=\text{det}(A)^{-1}\sum_{i,j,l}m_{il}a_{lj}A_{ij}~\\
&=\text{det}(A)^{-1}\sum_{i,l}m_{il} \delta_{il} \text{det}(A)~\\
&=\text{Trace}(M),
\end{split}
\end{equation}
where we used $(3)$ in the last step. $\square$