If I make the transformation

$$ x_i \rightarrow x_i' = x_i + \delta x_i,$$

I find that the Lagrangian $L = L(x_i,\dot{x_i})$ transforms as

$$ L \rightarrow L' = L + \frac{\partial L}{\partial x_i} \delta x_i + \frac{\partial L}{\partial \dot{x_i}} \delta \dot{x}_i \\ = L + \frac{\partial L}{\partial x_i} \delta x_i + \frac{\partial L}{\partial \dot{x_i}} \frac{d}{dt}\delta x_i \\ =L + \frac{\partial L}{\partial x_i} \delta x_i + \frac{d}{dt}\bigg(\frac{\partial L}{\partial \dot{x_i}} \delta x_i\bigg) - \bigg(\frac{d}{dt}\frac{\partial L}{\partial \dot{x_i}}\bigg) \delta x_i \\ = L + \frac{d}{dt}\bigg(\frac{\partial L}{\partial \dot{x_i}} \delta x_i \bigg).$$

For a symmetry, we demand $L'=L$, which means

$$ \frac{d}{dt}\bigg(\frac{\partial L}{\partial \dot{x_i}} \delta x_i \bigg) = 0. \quad (\mathrm{1})$$

This is just Noether's theorem. However, I have also read in Goldstein, and other Classical mechanics texts, that if we transform the Lagrangian in such a way that it changes as

$$ L \rightarrow L' = L + \frac{d}{dt} F(x) \quad (\mathrm{2}), $$

$L'$ still extremises the action for the same path. So, surely for symmetry I do not have to demand the condition as given in $(\mathrm{1})$ because it is a total time derivative like $(\mathrm{2})$. Is Noether's theorem overly restrictive about Lagrangian transformations? Can I derive conserved quantities even if I allow the Lagrangian to change by a total time derivative?


3 Answers 3


I think the problem with your statement is the definition of symmetry.If you consider the family of trasformations $$x'_i \rightarrow x_i + \epsilon K_i(\vec{x},\dot{\vec{x}}) $$ Then you can demand for symmetry that: $$\dfrac{dL'}{d\epsilon}\Bigr|_{\epsilon=0}=0$$ So that a symmetry of the Langrangian is a transformation that leaves it unchanged at first order with respect to $\epsilon$.You can derive the conserved quantity as follows from Taylors theorem: $$L'(x_i',\dot{x_i'})=L'(x_i+\epsilon K,\dot{x_i}+\epsilon \dot{K})=L(x_i,\dot{x_i}) +\dfrac{\partial L}{\partial x_i}\epsilon K_i +\dfrac{\partial L}{\partial \dot{x_i}}\epsilon \dot{K_i}+\mathcal{O}(\epsilon^2)$$ From there by applying the condition of the symmetry and using the Euler-Lagrange equations you get your desired result: $$\dfrac{\partial L}{\partial x_i}K_i +\dfrac{\partial L}{\partial \dot{x_i}} \dot{K_i} = \dfrac{d}{dt}\left( \dfrac{\partial L}{\partial \dot{x_i}}\right) K_i + \dfrac{\partial L}{\partial \dot{x_i}} \dot{K_i} = \dfrac{d}{dt}\left( \dfrac{\partial L}{\partial \dot{x_i}} K_i\right)=0$$


Combining your calculation of $L'-L$ with th claim in Eq. (2), the conserved quantity is $\dfrac{\partial L}{\partial \dot{x}_i}\delta x_i-F\left( x\right)$. Similarly, $F$ must in general be subtracted from the conserved charge obtained in SpecialM's answer.


A symmetry of a Lagrangian is defined as a transformation that leaves the Lagrangian unchanged up to a total time derivative. Thus, there is no point of adding condition (1) since adding a total time derivative to the Lagrangian gives the same equations of motion.

Therefore, the Noether's theorem still holds even if the Lagrangian changes by a total time derivative.


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.