I'm self-studying Mechanics and I have a little problem:

We can see that in Landau's book or in Wikipedia that when we inject the lagrangian in Euler Lagrange equation the term $\frac{\partial v²}{\partial q}$ vanish. So we get $\frac{\partial L}{\partial q}= - \frac{\partial U}{\partial q}$

here more details :

We want to proof that Euler-lagrange equation imply newton law :

Euler lagrange equation state that $\frac{d}{dt}\frac{\partial L}{\partial \dot{q}} = \frac{\partial L}{\partial q}$

And $ L= T-V(q)= \frac{1}{2}mv² - V(q) $

But if we inject L in Euler-Lagrange equation we will get

$\frac{d}{dt}\frac{\partial L}{\partial \dot{q}} = m\frac{dv}{dt} - \frac{d}{dt}\frac{\partial V}{\partial \dot{q}}$

And $\frac{\partial L}{\partial q} = \frac{1}{2}m\frac{\partial v²}{\partial q} + F$

In landau's book the terms $\frac{\partial V}{\partial \dot{q}}$ and $ \frac{1}{2}m\frac{\partial v²}{\partial q}$ Vanish without any explanation

Why these terms vanish?


The first term vanishes because you have assumed that the potential is independent of $\dot{q}=v$ (as you have said, $V(q)$). The second term vanishes because speed and possition are independent.

  • $\begingroup$ But suppose for exemple that we deal with a simple pendulum. If the solve de the DE, we get : $ \theta = acos(wt+\varphi ) $ , so we can find a function f such that : $t = f(\theta )$ and $\frac{d\theta }{dt}= - awsin(wt+\varphi )$ Thus $\dot{\theta}= - awsin(wf(\theta)+\varphi )$ and thus we will have $\frac{\partial L }{\partial \theta} = \frac{1}{2}ml\frac{\partial \dot{\theta²} }{\partial f}\frac{\partial f}{\partial \theta} -mglsin(\theta)$ . So why the term $\frac{1}{2}ml\frac{\partial \dot{\theta²} }{\partial f}\frac{\partial f}{\partial \theta}$ vanish ? $\endgroup$ – Med Saâd Alami Nov 15 '14 at 21:22
  • $\begingroup$ @MedSaâdAlami Newton's equations are specifically written as $m\ddot{x}=-U'(x)$, where $x$ is a coordinate in your Euclidean system. If you want to prove Newton's equations, then, you'd better choose to write $L$ in terms of Euclidean coordinates first. (The pendulum adds an extra degree of complication, therefore, because there is now a force of constraint when you move back to Euclidean coordinates!) $\endgroup$ – user12029 Nov 15 '14 at 21:58
  • $\begingroup$ @MedSaâdAlami, so certainly that term doesn't vanish. But, if it did vanish, what would that get you? (Also, $\theta=a \cos(\omega t)$ does not solve the simple pendulum problem, it solves the simple harmonic oscillator problem.) For reference I think the proof of Newton's laws from the principle of least action is given in L&L as equation 5.3. It's so simple because the first step in this would be writing $L$ in the form 5.1. $\endgroup$ – user12029 Nov 15 '14 at 22:04

The Lagrangian is defined in the most simple case as a function of $q$, $\dot{q}$ and $t$: $L(q,\dot{q},t)$. This notation implies that $q$, $\dot{q}$ and $t$ are by definition independent variables. This is how you have to interpret the partial derivatives to $q$ and $\dot{q}$, it doesn't make sense to write: $q = f(\dot{q})$, because both are considered independent variables in the lagrangian. Its just a function of three variables, so $\frac{\partial L(q,\dot{q},t)}{\partial \dot{q}}$ is mathematically just the same as $\frac{\partial L(x,y,t)}{\partial y}$ where we replaced $q$ by $x$ and $\dot{q}$ by $y$: it is the derivative to the second variable of the function.

Ofcourse when you solve the problem and find a $q(t)$ you can find a relation for this solution between $q$ and $\dot{q}$: $q = f(\dot{q})$, but this doesn't alter the fact that you should consider the variables of the $L$ above as independent, this solution has nothing to do with the independent variables of the lagrangian.

So when using the Lagrange formula: $$\frac{d}{dt}\frac{\partial L(q,\dot{q},t)}{\partial \dot{q}} = \frac{\partial L(q,\dot{q},t)}{\partial q}$$ the idea is that for the partial derivatives you just look at $q$ and $\dot{q}$ as independent variables, neglecting possible solutions you may have in mind. Calculating these derivatives will result in new functions of $q$, $\dot{q}$ and $t$, f.i.: $F(q,\dot{q},t) =\frac{\partial L(q,\dot{q},t)}{\partial \dot{q}}$. Then you have to calculate a total derivative to $t$ of $F$: to do this you have to substitute a possible solution $q(t)$, i.e. $F \circ q = F(q(t),\dot{q} (t),t) = G(t)$, the result is a function of only $t$, only when calculating this total derivative you can use dependency between $q$ and $\dot{q}$ because you have to insert a possible solution.


Why these terms vanish?


$$L = \frac{mv^2}{2} - V(q) = \frac{m(\dot q)^2}{2} - V(q)$$


$$\frac{\partial L}{\partial \dot q} = m \dot q$$

Why? Because $V(q)$ is a function of $q$ thus the partial derivative of $V$ with respect to $\dot q$ is zero.


$$\frac{\partial L}{\partial q} = -\frac{\partial V}{\partial q}$$

Why? Because $\frac{m(\dot q)^2}{2}$ is a function of $\dot q$ thus the partial derivative of $\frac{m(\dot q)^2}{2}$ with respect to $q$ is zero.

  • 1
    $\begingroup$ Ok i understood that... but the question is not as simple as you think : Why $q$ and $\dot{q}$ have to be independant ? $\endgroup$ – Med Saâd Alami Nov 15 '14 at 23:38
  • $\begingroup$ @MedSaâdAlami, they're not independent but, as I wrote and emphasized, these are partial derivatives. $\endgroup$ – Alfred Centauri Nov 15 '14 at 23:46
  • $\begingroup$ So if they're not independant we can write for exemple $q=f(\dot{q})$ and we will have $\frac{\partial V(q)}{\partial \dot{q}}= \frac{\partial V(f(\dot{q}))}{\partial q} = \frac{\partial V(f(\dot{q}))}{\partial f} \frac{\partial f}{\partial \dot{q}}$... So ? how to conclude ? (i think that their is somthing about calculus that i don't understand) $\endgroup$ – Med Saâd Alami Nov 16 '14 at 0:01

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.