On Direction Derivative of Lie Groups

On Direction Derivative of Lie Groups

This article is inspired by Henry Taubes’s Differential Geometry

If $\Vert A\Vert <1$, then $(I-A{)}^{-1}=I+A+A^2+\cdots +A^n+\cdots$

proof. First, the RHS does converge if $\Vert A\Vert <1$, for
$$\Vert \sum _{n=0}^{\infty }{A}^n\Vert \le \sum _{n=0}^{\infty }\Vert A^n\Vert \le \sum _{n=0}^{\infty }\Vert A{\Vert }^n=\frac{1}{1-\Vert A\Vert }$$
Next, we show that the RHS is indeed the inverse of $I-A$, for
$$\sum _{n=0}^{\infty }{A}^n(I-A)=\sum _{n=0}^{\infty }{A}^n-\sum _{n=1}^{\infty }{A}^n=I$$
Q.E.D.

Using this, we show that the directional derivative of the map $A\mapsto A^{-1}$ as the point $A$ in the direction $V$ is $A^{-1}V{A}^{-1}$. Choose $V$ with small enough matrix norm, then
$$(A+V{)}^{-1}=(A(I+A^{-1}V){)}^{-1}=(I+A^{-1}V{)}^{-1}{A}^{-1}$$

$$=\sum _{n=0}^{\infty }(-1{)}^n{A}^{-n}{V}^n{A}^{-1}$$

Thus, the $1^{th}$ order term of this equation is $A^{-1}V{A}^{-1}$, exactly as we expected.

The next important computation is the directional derivative of $\det$. We show that ${\partial }_{ij}\det m=m_{ij}^{\star }$, where $m_{ij}^{\star }$ denotes the $(i,j)$-adjacent matrix of $m$, i.e. to delete the $i^{th}$ row and $j^{th}$ column of $m$. To me, the prove is quite elegant, for it is telling that the symmetry (instead of finiteness of vector space) of $\det$ must let the deduction stop somewhere.

Proposition: If $m\in M(n,\mathbb{R})$, then ${\partial }_{ij}\det m=m_{ij}^{\star }$.

proof.

Using the identity $\det m={\sum }_{j=1}^n{m}_{ij}\det m_{ij}^{\star }$ for any chosen $i$, we have
$$\begin{array}{lll}{\partial }_{ij}\det m& =& \det m_{ij}^{\star }+{\sum }_{k\ne j}{m}_{ik}{\partial }_{ij}\det m_{ik}^{\star }\\ & =& \det m_{ij}^{\star }+{\sum }_{k\ne j}{m}_{ik}{\partial }_{ij}({\sum }_{l\ne i}{m}_{lj}\det m_{il,kj}^{\star })\end{array}$$
where $m_{il,kj}^{\star }$ denote subbing matrix of $m$ by deleting the $i,l$ rows and $k,j$ columns. Observing the $m_{lj}$ in the above equation cannot be $m_{ij}$ since $l\ne i$, we obtain
$$\det m_{ij}^{\star }+\sum _{l\ne i,k\ne j}{m}_{ik}{m}_{lj}{\partial }_{ij}\det m_{il,kj}^{\star }$$
But ${\partial }_{ij}\det m_{il,kj}^{\star }=0$ since the $i^{th}$ row and the $j^{th}$ column have been deleted, hence the only item left is $\det m_{ij}^{\star }$, which is exactly what we expected.

Q.E.D.

From this we obtain the differential of $\det$:
$$d(\det ){|}_m=\det m\sum _{i,j}(m^{-1}{)}_{ij}d{m}_{ij}=\det m\text{ tr}(m^{-1}dm)$$