home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!munnari.oz.au!manuel!rsphy1.anu.edu.au!rwc124
- From: rwc124@rsphy1.anu.edu.au (Roderick Vance)
- Subject: Hermetian Matrix Lie Algebra
- Message-ID: <1992Sep14.132239.26191@newshost.anu.edu.au>
- Sender: news@newshost.anu.edu.au
- Reply-To: rwc124@rsphy1.anu.edu.au (Roderick Vance)
- Organization: Optical Sciences Centre, Australian National University
- Date: Mon, 14 Sep 92 13:22:39 GMT
- Lines: 68
-
- About a fortnight ago i posted the following question on the net:
-
- I begin with a given set H of hermetian matices, whence i form the set
- of unitary matrices:
-
- U1 = {exp(i h) : h belongs to H}
-
- I wanted to find the group generated by members of U1. A kind newsgroup
- reader offered me the following advice:
-
- "...it is (comparitively) easy to compute the group generated by all
- exponentials of a fixed set of Hermitian matrices. It is the
- exponential of the Lie algebra generated by those matrices."
-
- Now, having just begun to teach myself Lie algebra&group theory, i'd
- be most grateful if someone could comment on whether my following
- interpretation of the above advice is right...
-
- I take for example, just two hermetian matrices h1 and h2. I find
- the Lie bracket
-
- [h1,h2] = h1 h2 - h2 h1 (usual matrix product)
-
- to get a third matrix. I add this to my set, form all the Lie brackets
- possible and add them to the set and repeat until i have the smallest
- set L = {l1,l2,l3,l4,....} that both contains {h1,h2} and has the property
- that [x,y] belongs to L whenever both x and y do. Then the group
- generated by {exp(i h1),exp(i h2)} will be:
-
- {exp(i (a1 l1 + a2 l2 + ...)): l1,l2,... belong to L, a1, a2 ... are scalars}
-
- I'd also LOVE a simple reference/explanation of why this works. Presumably,
- such a proof would use the fact that commutators of L's members are
- also members of L to show that:
-
- exp(i l0) exp(i (a1 l1 + a2 l2 +...)) =
- exp(i (a1' l1 + a2' l2 + ...)) for any l0 in L
-
- and the required result would then follow by induction. I can't quite seem
- to solve it!
-
- Sorry to bother the net with non-research forefront questions but i'm sure
- someone out there knows Lie theory inside out.
-
- Many Thanks In Advance
-
- Roderick Vance
- Optical Sciences Centre
- Australian National University.
-
- By the way, the physicists amongst you may be interested to learn that
- my problem
- arises in a study of n coupled, single-moded optical waveguides. If
- field amplitudes
- [x1,x2,x3,...] drive the inputs then the ouput after distance z will be
- U(z) [x1,x2,x3,...]
- , where U(z) is an nxn square matrix fulfilling:
-
- (d/dz) U(z) = i K(z) U(z),
-
- K is an nxn square matrix of coupling co-efficicents between the guides.
- If the system
- is lossless, K is hermetian and U is therefore unitary. I am interesed
- in what input to
- output transfer functions U(z) i can realise with a restricted set of
- K.
-
-
-