Suppose $q = \{q_1,\cdots, q_i\}$ is a coordinate system for Lagrangian $L(q,\dot{q},t)$. In this text by David Morin, on page 16 in chapter 6, it states that a symmetry is a transformation of the Lagrangian coordinates, $q'_i = q_i + \varepsilon\cdot K_i(q)$, $\varepsilon$ small and where $K_i(q)$ is a function on the set $q$, that does not lead to a first-order change in the Lagrangian.

It later stated that not having a first-order change in the Lagrangian means that $$\frac{dL}{d \varepsilon}=\sum_{i}\left(\frac{\partial L}{\partial q_i}\frac{\partial q_i}{\partial \varepsilon} + \frac{\partial L}{\partial \dot q_i}\frac{\partial \dot q_i}{\partial \varepsilon} \right) = 0.\tag{6.59}$$

Is an equivalent condition for symmetry given by requiring the Taylor expansion of $L(q',\dot{q}')$ about $L(q,\dot{q})$ to lack first order terms in $\varepsilon$? This comes from thinking that $L(q',\dot{q}')$ can be written as

$$L(q',\dot{q}') = L(q,\dot{q}) + \frac{dL}{d\varepsilon}\varepsilon + O(\varepsilon^2),$$

which would show that the two are equivalent, but I feel like I'm glossing over some details.


1 Answer 1


I don't think there is the statement that $K_i(q)$ has to be linear in $q$. This is certainly in a lot of examples. The simplest is a translation symmetry $q'=q+\epsilon$.

Going to your question though, this is certainly equivalent to asking that the Taylor expansion of $L$ around $q$ has no first order in $\varepsilon$. However, your Taylor expansion of the Lagrangian is wrong. It is simply $$L(q')=L(q)+ \epsilon\frac{\text{d}L}{\text{d}\epsilon}+\mathcal O(\epsilon^2)$$.

  • $\begingroup$ Thank you for your answer. I think I was confused by the Taylor series expansion of L(q'). My intuition was to expand $L$ like a multivariable function. However, in this case, we are only varying a single variable - $\epsilon$. I suppose that is why the given Taylor expansion is only in terms of $\epsilon$. $\endgroup$ Aug 7, 2020 at 23:05
  • $\begingroup$ Yes, we are indeed doing an expansion over a single variable. Mind you though that we can use the chain rule on $dL/d\varepsilon$ to recover the expression at the beginning of the question $\endgroup$ Aug 8, 2020 at 3:29

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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