home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!noc.near.net!news.bbn.com!olivea!hal.com!darkstar.UCSC.EDU!ucscb.UCSC.EDU!ask
- From: ask@ucscb.UCSC.EDU (Andrew Stanford Klingler)
- Newsgroups: sci.math
- Subject: Re: Lie vs. Covariant derivatives
- Message-ID: <1e7pcsINNpoa@darkstar.UCSC.EDU>
- Date: 16 Nov 92 09:24:44 GMT
- Distribution: usa
- Organization: University of California; Santa Cruz
- Lines: 64
- NNTP-Posting-Host: ucscb.ucsc.edu
-
-
- :From: glezen@drb-mathsun1.usc.edu (Paul Glezen)
-
- : I've seen several interesting responses to the inquiry:
-
- : "What is the difference between covariant derivative and the Lie
- : derivative?"
-
- : But comments like "one requires a certain vector field to be defined at
- :a point while the other needs it to be defined on some neighborhood", while
- :correct, shed little light on why you get different answers. Perhaps a simple
- :example will show why these things are two completely different beasts. The
- :emphasis here is on simplicity and intuition; not technicalities meaningless
- :to a beginner.
-
- : Let's choose our manifold to be the 2-d plane and let P be the point (1,0).
- :Consider two vector fields V and W defined by
-
- : V = (-y)i + (x)j where i and j are the standard basis for
- : and the plane and hence for the tangent space
- : W = i to the plane.
-
- : V corresponds to the velocity field arising from a rigid counter-clockwise
- :rotation about the origin while W is the constant vector field which points to
- :the right and has length one. Allow me to denote
-
- : L(V,W) := Lie derivative of W in the direction of V
-
- : D(V,W) := Covariant derivative of W in the direction of V
-
- : Notice also the diffence between L and D when you switch V and W. Since it
- :can be shown that L(V,W) = [V,W], the Lie bracket, we expect L(V,W) = -L(W,V).
- :In our example, it is clear that the diffeomorphisms induced by W are merely
- :translations to the right. The tangent mappings for each diffeomorphism is
- :clearly the identity. If you start at (1,0) and move to the right, indeed
- :the vectors of V, the rotation field, seem to get larger in the j direction.
- :Thus when you compare V(1+dx,0), via the identity tangent map of the diffeo-
- :morphism induced by W, to V(1,0), you see that indeed the vector field V is
- :getting bigger in the j direction. Thus L(W,V) = j.
-
- : On the otherhand, it is easy to see that D(W,V) = -i at the point P if you
- :think of V as being the velocity field of a particle rotating about the origin
- :at constant speed. Then D(W,V) is just the acceleration (change in velocity)
- :which points toward the origin.
-
- D(W,V) is the covariant derivative of V in the direction of W. But
- W is a constant i pointing along the x-axis. An infinitesimal
- displacement along the x-axis causes V to change by (dx)j, so
- we get D(W,V)=j. This is merely a calculational error, but it brings
- back to mind the equation that underlay my geometric confusion.
-
- L(V,W) = [V,W] = D(V,W) - D(W,V)
-
- So in some restricted sense I can think of the Lie derivative as
- constructed out of covariant differentiation around a little loop
- (incidentally this makes antisymmetry of Lie derivative obvious).
- This is what I was mentally referring to with my previous post.
- However, the equation's only true for torsion-free connections.
-
- As a followup: what does the existence of torsion mean to those
- of us whose experience is limited to GR? (Apart from a bad choice
- of frame field!)
-
- ask@ucscb.ucsc.edu
-