Exponentiating the Lie Derivative: Which direction does it rotate?

293 Views Asked by At

I've been working through the book Functional Differential Geometry by Sussman and Wisdom and am having trouble with an example they give.

We can exponentiate a Lie derivative $$ e^{\mathsf{t\mathcal{L}_v}}\mathsf{y} = \mathsf{y + t \mathcal{L}_v y + \frac{t^2}{2!} \mathcal{L}^2_v y + \dots } $$ which evolves $\mathsf{y}$ along the integral curves of $\mathsf{v}$. As a concrete example they evolve the coordinate basis vector $\frac{\partial}{\partial \mathsf{y}}$ along $\mathsf{J}_z = x\frac{\partial}{\partial \mathsf{y}} - y\frac{\partial}{\partial \mathsf{x}}$ (a counter clockwise circular field or z-angular momentum generator) and give as an answer

$$\exp(a \mathcal{L}_{\mathsf{J}_z})\tfrac{\partial}{\partial \mathsf{y}} = -\sin(a)\tfrac{\partial}{\partial \mathsf{x}} + \cos(a)\tfrac{\partial}{\partial \mathsf{y}}.$$ This agrees at $a=0$ and indicates that the evolution maintains the orientation of $\mathsf{v}$ and $\mathsf{y}$ along the flow. $\tfrac{\partial}{\partial\mathsf{y}}$ just rotates along $\mathsf{J}_z$.

If I try to calculate the expansion, for the first term I get $$ a\mathcal{L}_{x\frac{\partial}{\partial \mathsf{y}} - y\frac{\partial}{\partial \mathsf{x}}}\tfrac{\partial}{\partial \mathsf{y}} = a\mathcal{L}_{- y\tfrac{\partial}{\partial \mathsf{x}}}\tfrac{\partial}{\partial \mathsf{y}} = -a[y\tfrac{\partial}{\partial \mathsf{x}}, \tfrac{\partial}{\partial \mathsf{y}}] = a\tfrac{\partial}{\partial \mathsf{x}} $$ which disagrees with the expansion of the answer by a sign, giving rotation in the opposite direction.

The intuitive notions of the Lie Derivative I have also say it should be $a\tfrac{\partial}{\partial \mathsf{x}}$. For example, beginning at $x=1, y=0$ we can travel along $\epsilon\mathsf{J}_z$ and then $\epsilon \tfrac{\partial}{\partial \mathsf{y}} $, or begin along $\epsilon \tfrac{\partial}{\partial \mathsf{y}} $ and then along $\epsilon\mathsf{J}_z$ (points slightly to the left). The difference between these paths will be $\epsilon^2 \tfrac{\partial}{\partial \mathsf{x}} $. The same goes for beginning at $x=0,y=1$ or if I do it using a pushforward. I'm tempted to say the book is mistaken, and that to get a rotation operator which coincides with the direction of $\mathsf{J}_z$ we should use $e^{-t\mathcal{L}_{\mathsf{J}_z}}$ but would appreciate some confirmation.

2

There are 2 best solutions below

2
On BEST ANSWER

You are right. I believe it is a small typo in the book. Indeed, here is a geometric interpretation of $exp(a \mathcal{L}_{J_z})\frac{\partial}{\partial y}$. You flow by $J_z$ for some time $a$, evaluate the vector field $\frac{\partial}{\partial y}$ at that new point, then take $d\varPhi_{-a}$ of that vector. As you can see, we are making use of $d\varPhi_{-a}$ (note the minus sign). Thus your calculation was right, and I believe it was a small typo in the book. I think it should be:

$\exp(a \mathcal{L}_{\mathsf{J}_z})\tfrac{\partial}{\partial \mathsf{y}} = \sin(a)\tfrac{\partial}{\partial \mathsf{x}} + \cos(a)\tfrac{\partial}{\partial \mathsf{y}}$.

One way to check would be to do the calculation explicitly (though I did not do that).

1
On

To add to @Malkoun: I would definitely say a typo. It seems the book is based upon a technical report by Sussman and Wisdom which can be found here. In that report (on page 48) they have the same calculation but the result this time is $+\sin(a)\frac{\partial}{\partial x}$, as you have calculated.

It does makes me chuckle that they preface the result of the calculation with "apparently the result is" - maybe they were unsure of the answer too!