last day (15 days later) » 

2:50 PM
1
A: How can I compute the differential of this map $(t,x)\mapsto \begin{pmatrix} e^t &xe^t \\ 0 & e^{-t} \end{pmatrix}$ going into a submanifold?

Sam BallasThe Lie algebra $\mathfrak{g}$ of $G$ is the tangent space at the identity of $G$ which consists of matrices of the form $$\begin{pmatrix} a & b\\ 0 & -a \end{pmatrix} $$ This space has a basis $$\begin{pmatrix} 1 & 0\\ 0 & -1 \end{pmatrix}, \begin{pmatrix} 0 & 1\\ 0 & 0 \end{pmatrix} $$ To find ...

 
Could you explain what you meant by translate the basis?
 
I just mean to multiply the basis elements on the left by the matrix corresponding to the point you want to translate to. This works because left multiplication is a diffeomorphism of $G$ whose pushforward is given by left multiplication.
 
Ah okey I see. and how do you get the partial derivatives of my map?
 
they are just the entrywise derivatives with respect to both $t$ and $x$
 
So the first matrix is w.r.t t and the second w.r.t x?
 
2:50 PM
yes, that is correct
 
But wouldn't I then have $-xe^t$?
instead of $xe^t$? @SamBallas
 
We are taking the derivative of $\begin{pmatrix}
 
Is there something missing?
@SamBallas can I explain you here in chat what I have understood so that you see where I'm stuck?
 
are you here?
 
yes
why?
 
3:02 PM
just checking
can you explain what is confusing you?
 
ah okey so I will write you here what I understood. give me a view minutes to write it properly
are you still here?
I have understood it as follows: we want to find $D(\phi)_{(t,x)}:\Bbb{R}^2\rightarrow T_{\phi(t,x)}G$. So first we need to understand $T_{\phi(t,x)}G$. I know a basis for the lie algebra of $G$ which is $T_IG$ where the basis is $\begin{pmatrix}
1 & 0\\
0 & -1
\end{pmatrix}, \begin{pmatrix}
0 & 1\\
0 & 0
\end{pmatrix}$. Now we can translate this basis to one of $T_{\phi(t,x)}G$. Which then is $\begin{pmatrix}
e^t & -xe^t\\
0 & -e^{-t}
\end{pmatrix}$, $\begin{pmatrix}
0 & e^t\\
0 & 0
\end{pmatrix}$. So we know that every element of $T_{\phi(t,x)}G$ can be written as $\lambda \begin{pmatrix}
 
yeah, let me look this over
If you take the entrywise derivatives (say with respect to x) of the map you are interested in then the resulting matrix is an element of $T_{\phi(x,t)}G$ and hence you can write it with respect to the basis that we have constructed for this space.
The $x$ derivative case is quite simple because it is just equal to our second basis element, so if we try to write it as a linear combination in this basis then the coordinates are just $(0,1)$
The other derivative is a bit more complicated, but if you write it down you find that the coordinates for the $t$ derivative in this basis is $(1,1+x)$
 
but don't I have elements of $T_{\phi(x,t)}G$ with my representation: $\lambda \begin{pmatrix}
e^t & -xe^t\\
0 & -e^{-t}
\end{pmatrix}+\mu\begin{pmatrix}
0 & e^t\\
0 & 0
\end{pmatrix}$
 
I don't understand what you mean here
 
I mean I have startet with a basis of $T_IG$ and translated it to a basis of $T_{\phi(x,t)}G$ right?
 
3:18 PM
that is correct
so now that you have a basis for $T_{\phi(x,t)}G$ we just use it to express the partial derivatives
 
ah I think I got it now. because now I want to compute the derivative of $\phi$ w.r.t $t$ and once w.r.t $x$?
and this derivatives can be written in the new basis?
 
yes, and then take those derivatives and figure out how to write them in the basis we have constructed
yes, those are the "coordinates" I mentioned above
 
but one question still remains.
can I only compute the "normal" partial derivative of $\phi$ since $G$ is isomorphic to $\Bbb{R}^2$?
because otherwise it would not make sense to speak about the "normal" partial derivatives would it?
 
I don't understand what you mean by "normal"
 
so like the one we are used to $\partial f/\partial x$ is $f$ derived w.r.t $x$
because i think partial derivatives would not make sense in any group $G$
 
3:22 PM
Well, you defined your function $\phi$ using the variables $x$ and $t$, so they are the most natural coordinates to write the Jacobian in. If you rewrote the function using different coordinates then you would be able to write the Jacobian in these coordinates
the partial derivatives come from your choice of coordinates on the domain
 
but i mean for the partial derivative I need a map $f$ which takes values in $\Bbb{R}^d$ but in our case $\phi$ takes values in $G$. SO my question is can we only take partial derivatives as we are used to since $G$ is isomorphic to $\Bbb{R}^2$?
or does it also work with an arbitrary matrix group $G$?
 
These same ideas would work for computing derivatives of functions into arbitrary matrix Lie groups. You would compute a basis for the Lie algebra (tangent space at I) and then use left multiplication to get a basis at all the other points. Then you can take partial derivatives of the matrix entries of the map and express them with respect to the relevant bases
 
aha okey. but why couldn't I immediately compute the partial derivatives of $\phi$ w.r.t $t,x$ and I'm done?
 
If you just compute the partial derivatives then you are really viewing $\phi$ as a function into $M_2(\RR)$ (2x2 real valued matrices). If you want to view it as a function into $G$ then you need to use a basis for the tangent spaces to $G$
 
Ah but If I wanted to compute it into $M_2(\RR)$ I can take the usual basis there and then write both partial derivarives in terms of this basis?
 
3:33 PM
If you take the entrywise partial derivatives then these are the coordinates with respect to the "usual" basis of M_2(RR)$
I have to go now, but I hope this helped
 
Yes thanks a lot!
Else I will write again!
 
maybe accept my answer when you get a chance:)
 
 
5 hours later…
8:08 PM
@SamBallas I did it
 
8:28 PM
thanks
 

  last day (15 days later) »