I am trying to do a simple calculation of $L_X(Y)$ and $L_Y(X)$ over the manifold $M = S^1$ where $X = \partial_{\theta} = (-\sin{\theta}, \cos{\theta})$ and $Y = \cos{\theta}\partial_\theta$ using the local coordinates $\phi(\theta) = (\cos{\theta}, \sin{\theta})$. In particular I am self-studying from Peter Peterson's Riemannian Geometry and want to really understand what the Lie derivative is of $2$ vector fields from the difference quotient definition with local flow. <span class=$Y$ has a source and a sink."> I have the following. In local coordinates, the local flow defined by $X$ is $F^t(\theta) = (\cos{(\theta + t)} , \sin{(\theta + t)})$ and is a curve on $S^1$. $$L_X(Y) := \lim\limits_{t \rightarrow 0} \frac{dF^{-t}(Y(F^t(\theta))) - Y(\theta)}{t} = (\sin^2{\theta}, - \sin{\theta}\cos{\theta}) = -\sin{\theta} \partial_{\theta}$$


$1.~~$ I am having a hard time picturing what it means for $L_X(Y) = 0$ at the rightmost and leftmost points of the circle. Relatedly, how does the vector (1,0) relate to $Y$ "flowing along $X$" at the top point of $S^1$? I get that these derivatives are really about curves in $T_pM$. In particular, the difference quotient is fixing a point $p$ on $S^1$, flowing along $X$, taking the vector given by $Y$, and pulling it back to $T_pM$, thus tracing a curve in $T_pM$ as $t$ varies.

$2.~~$ I am having trouble calculating the local flow of $Y$ to ultimately calculate $L_Y(X)$. I have tried the following: Let $G^t(\theta) = (\cos(\eta(t, \theta)), \sin(\eta(t, \theta)))$. Then $\dot{G} = (-\sin(\eta)\dot{\eta}, \cos(\eta)\dot{\eta}) = \cos(\eta) (-\sin(\eta), \cos(\eta)) = Y(G^t(\theta))$ where the dot is differentiation w.r.t. $t$. This gives $\dot{\eta} = \cos(\eta)$. I'm not sure how to proceed. I also know that $G^0(\theta) = (\cos\theta,\sin\theta)$, and $G^t(\pm \frac{\pi}{2}) = (0, \pm 1)$. A hint on how to proceed would be greatly appreciated. Separation of variables (treating $\theta$ as constant)/an ODE solver seems to give rather unpleasant solutions that I can't easily use to plug into $X$. Is thinking of these things as literal vectors as opposed to the derivation viewpoint hindering me?

  1. Yes, exactly, though since your manifold is one-dimensional these curves may be hard to visualize, since $T_pM$ is just isomorphic to $\mathbb{R}$ and the curve just moves back and forth on the line. For example at $\theta=0$, you can visualize $T_pM$ as a vertical tangent line passing though the right point of your circle. Then consider how $$f(t) = dF^{-t} \circ Y \circ F^t(0)$$ behaves as a function of $t$: at $t=0$ you get the unit vector pointing up, and for $t$ slightly negative or slightly positive the vector points up and is slightly less than unit length. It's clear then that $f$ has a critical point at $t=0$ (the "curve" $f(t)$ has a cusp where it changes from moving up to moving down) which agrees with your computation that $L_XY=0$. $$$$ Similarly at $\theta=\pi/2$ you can imagine $T_pM$ as the horizontal tangent line, and as $t$ varies form being slightly negative to slightly positive, $f(t)$ varies from a vector pointing to the left along the line to vanishing at $t=0$ to pointing right; the derivative of $f$ is now negative in agreement with your answer $L_XY = -\partial_\theta$. $$$$ Notice that since $X$ is constant, all you're really doing here is computing the covariant derivative of $Y$: $L_XY = \nabla_{X(\theta)} Y$. To really understand what's happening with the flows I suggest trying a slightly less simple example, e.g. with $X$ and $Y$ two vector fields on the sphere.

  2. I don't think you should expect any particularly nice formula for the flow. Your approach of solving the ODE $$\frac{d\theta}{dt} = \cos\theta$$ on the obvious chart seems right to me. You can get a (not particularly pleasant) closed-form solution by separation of variables, and Wolfram Alpha can't do any better either.

  • $\begingroup$ Thanks! Short of a solution, I am thinking about some properties for $L_Y(X)$. Am I correct that $L_Y(X) = 0$ at $\theta = \pm \pi/2$? What about the sign of $L_Y(X)$ at $\theta = 0, \pi$?. If my understanding is correct, I shouldn't be able to associate a sign to a lie derivative in general, but since this is a 1-manifold, I mean the sign of the coefficient in front of $\partial_\theta$. $\endgroup$ – geometry_geek Jul 19 at 11:24
  • $\begingroup$ @geometry_geek the Lie derivative of vector fields is anti-symmetric so you should get $$L_YX = -L_XY.$$ $\endgroup$ – user7530 Jul 20 at 20:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.