For the Hamilton's principle: $$\delta s =\int_{t_1}^{t_2}L(\mathbf {q+\delta q},\mathbf {\dot q+\delta \dot q},t) dt-\int_{t_1}^{t_2}L(\mathbf {q},\mathbf {\dot q},t) dt=0.\\$$

In the textbooks, using the Taylor's series:

$$L(\mathbf {q+\delta q},\mathbf {\dot q+\delta \dot q},t) =L(\mathbf {q},\mathbf {\dot q},t)+\frac{\partial L(\mathbf {q},\mathbf {\dot q},t)}{\partial \mathbf q} \bar \delta \mathbf q(t) +\frac{\partial L(\mathbf {q},\mathbf {\dot q},t)}{\partial \mathbf {\dot q}} \bar \delta \mathbf {\dot q(t)}+\frac{\partial}{\partial \mathbf q} (\frac{\partial L(\mathbf {q},\mathbf {\dot q},t)}{\partial \mathbf {\dot q}}\bar \delta \dot q)\bar \delta q + ...$$

But why is omitting the rest of the Taylor's series? I think it will propagate a lot of error in the derivation...

  • $\begingroup$ The variations $\delta q$ must multiply a coefficient $\epsilon$ that you will eventually take the incremental limit on. Higher order derivatives will still present the multiplicative factor after the incremental ratio and will therefore vanish anyway. $\endgroup$ – gented Aug 28 '17 at 12:23

The Euler Lagrange equation is a differential equation resulting from the search for the extremum of a functional: this extremum is given by the first variation only.

This is similar to the condition for finding a point where a function $f$ is extremum: the condition $df/dx=0$ is on the first derivative only.

In both cases, one does not seek to approximate $L$ or $f$ by a series, but rather extract from the first term in the series either a differential equation to be satisfied by a function or an algebraic equation to be satisfied by a point.


In addition, the order of the resulting Euler-Lagrange equation can be greater than 1 even if one considers only the first variation. For instance, in 1d, with a Lagrangian of the type $L=L(q,\dot{q},\ddot{q},t)$, the resulting EL equation is of second order: $$ \frac{\partial L}{\partial q}-\frac{d}{dt}\frac{\partial L}{\partial \dot{q}}+ \frac{d^2}{dt^2}\frac{\partial L}{\partial \ddot{q}}=0. $$

Finally, the first variation (like the first derivative) only guarantees an extremum. To verify if this is a local max or min, one ought to use the second variation (like the second derivative) but it is often technically complicated and unnecessary: one can simply verify if any other construction gives a greater or smaller action or shorter or longer path.

  • $\begingroup$ This is not the reason. The reason is that the variations $\delta q$ are multiplied by a coefficient that is set limit to zero afterwards. All terms in higher order derivatives still present the coefficient after the limit procedure and therefore vanish. $\endgroup$ – gented Aug 28 '17 at 12:21
  • $\begingroup$ @GennaroTedesco What is multiplied by a coefficient, in the case shown above I didn't need to declarate a coefficient what is generally declated as $\epsilon$ $\endgroup$ – Joan J. Cáceres Aug 28 '17 at 12:33
  • 1
    $\begingroup$ @GennaroTedesco This is equivalent, since the remaining coefficient is just proportional to the first variation (or the definition of the first derivative). $\endgroup$ – ZeroTheHero Aug 28 '17 at 12:34
  • $\begingroup$ @JoanJ.Cáceres That is exactly the point. You want to define $\delta L$, which by definition is the limit of the incremental ratio $L(q+ \epsilon) - L(q)$ over $\epsilon$. Your notation is just shorthand notation for the whole thing. $\endgroup$ – gented Aug 28 '17 at 12:35

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.