John Stillwell - Naive Lie Theory.pdf - Index of
John Stillwell - Naive Lie Theory.pdf - Index of
John Stillwell - Naive Lie Theory.pdf - Index of
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
7.6 The Campbell–Baker–Hausdorff theorem 153<br />
that<br />
Z = log(e X e Y )=<br />
(X +Y + XY + X 2<br />
2! + Y 2<br />
···)<br />
2! +<br />
− 1 (X +Y + XY + X 2<br />
2<br />
2! + Y 2<br />
···) 2<br />
2! +<br />
+ 1 (X +Y + XY + X 2<br />
3<br />
2! + Y 2<br />
···) 3<br />
2! +<br />
−···<br />
= X +Y + 1 2 XY − 1 YX+ higher-order terms<br />
2<br />
= X +Y + 1 [X,Y ]+higher-order terms.<br />
2<br />
The hard part <strong>of</strong> the Campbell-Baker-Hausdorff theorem is to prove that<br />
all the higher-order terms are composed from X and Y by <strong>Lie</strong> brackets.<br />
Campbell attempted to do this in 1897. His work was amended by<br />
Baker in 1905, with further corrections by Hausdorff producing a complete<br />
pro<strong>of</strong> in 1906. However, these first pro<strong>of</strong>s were very long, and many<br />
attempts have since been made to derive the theorem with greater economy<br />
and insight. Modern textbook pro<strong>of</strong>s are typically only a few pages long,<br />
but they draw on differentiation, integration, and specialized machinery<br />
from <strong>Lie</strong> theory.<br />
The most economical pro<strong>of</strong> I know is one by Eichler [1968]. It is only<br />
two pages long and purely algebraic, showing by induction on n that all<br />
terms <strong>of</strong> order n > 1 are linear combinations <strong>of</strong> <strong>Lie</strong> brackets. The algebra<br />
is very simple, but ingenious (as you would expect, since the theorem is<br />
surely not trivial). In my opinion, this is also an insightful pro<strong>of</strong>, showing<br />
as it does that the theorem depends only on simple algebraic facts. I present<br />
Eichler’s pro<strong>of</strong>, with some added explanation, in the next section.<br />
Exercises<br />
7.6.1 Show that the cubic term in log(e X e Y ) is<br />
1<br />
12 (X 2 Y + XY 2 +YX 2 +Y 2 X − 2XYX − 2YXY).<br />
7.6.2 Show that the cubic polynomial in Exercise 7.6.1 is a linear combination <strong>of</strong><br />
[X,[X,Y]] and [Y,[Y,X]].