Take the 2-minute tour ×
Physics Stack Exchange is a question and answer site for active researchers, academics and students of physics. It's 100% free, no registration required.

Is there a proof from the first principle that for the Lagrangian $L$,

$$L = T\text{(kinetic energy)} - V\text{(potential energy)}$$

in classical mechanics? Assume that Cartesian coordinates are used. Among the combinations, $L = T - nV$, only $n=1$ works. Is there a fundamental reason for it?

On the other hand, the variational principle used in deriving the equations of motion, Euler-Lagrange equation, is general enough (can be used to to find the optimum of any parametrized integral) and does not specify the form of Lagrangian. I appreciate for anyone who gives the answer, and if possible, the primary source (who published the answer first in the literature).


Notes added on Sept 22:
- Both answers are correct as far as I can find. Both answerers were not sure about what I meant by the term I used: 'first principle'. I like to elaborate what I was thinking, not meant to be condescending or anything near to that. Please have a little understanding if the words I use are not well-thought of.
- We do science by collecting facts, forming empirical laws, building a theory which generalizes the laws, then we go back to the lab and find if the generalization part can stand up to the verification. Newton's laws are close to the end of empirical laws, meaning that they are easily verified in the lab. These laws are not limited to gravity, but are used mostly under the condition of gravity. When we generalize and express them in Lagrangian or Hamiltonian, they can be used where Newton's laws cannot, for example, on electromagnetism, or any other forces unknown to us. Lagrangian or Hamiltonian and the derived equations of motion are generalizations and more on the theory side, relatively speaking; at least those are a little more theoretical than Newton's laws. We still go to lab to verify these generalizations, but it's somewhat harder to do so, like we have to use Large Hadron Collider.
- But here is a new problem, as @Jerry Schirmer pointed out in his comment and I agreed. Lagrangian is great tool if we know its expression. If we don't, then we are at lost. Lagrangian is almost as useless as Newton's laws for a new mysterious force. It's almost as useless but not quite, because we can try and error. We have much better luck to try and error on Lagrangian than on equations of motion.
- Oh, variational principle is a 'first principle' in my mind and is used to derive Euler-Lagrange equation. But variational principle does not give a clue about the explicit expression of Lagrangian. This is the point I'm driving at. This is why I'm looking for help, say, in Physics SE. If someone knew the reason why n=1 in L=T-nV, then we could use this reasoning to find out about a mysterious force. It looks like that someone is in the future.

share|improve this question
2  
I think that Goldstein shows one similar to that one Lagrange used before Euler formulated the least action argument that we see in, say, Marion and Thorten. –  dmckee Sep 21 '13 at 22:15
3  
I think your question kind of gets this backward... saying $L = T - V$ is a nice way to give a general Lagrangian that gives you Newtonian mechanics, but it isn't the most general Lagrangian, and this form, in fact doesn't work for, say Electromagnetism. The modern view would say that the Lagrangian (density) is an object that is defined prior to a notion of energy. –  Jerry Schirmer Sep 21 '13 at 22:18
1  
@ChinYeh There is a proof that builds from Newton's Laws and D'Alembert's principle! You want something more basic than that? –  Cheeku Sep 21 '13 at 23:28
2  
@ChinYeh: sure--you define the type sof fields/particles you need, define the symmetry of the problem, and then write down the most general lagrangian that respects these two things . Then you make predictions, and refine your lagrangian. –  Jerry Schirmer Sep 22 '13 at 0:18
1  
@Qmechanic I'm amazed this question hasn't been asked before –  Larry Harson Sep 22 '13 at 18:25
show 4 more comments

4 Answers 4

We assume that OP by the term first principle in this context mean Newton's laws rather than the principle of stationary action$^1$. It is indeed possible to derive Lagrange equations from Newton's laws, cf. this Phys.SE answer.

Sketched proof: Let us consider a non-relativistic$^2$ Newtonian problem of $N$ point particles with positions ${\bf r}_1, \ldots, {\bf r}_N$, with generalized coordinates $q^1, \ldots, q^n$, and $m=3N-n$ holonomic constraints.

Let us for simplicity assume that the applied force of the system has generalized (possibly velocity-dependent) potential $U$. (This e.g. rules out friction forces proportional to the velocity.)

It is then possible to derive the following key identity

$$\tag{1} \sum_{i=1}^N \left(\dot{\bf p}_i-{\bf F}_i\right)\cdot \delta {\bf r}_i ~=~ \sum_{j=1}^n \left(\frac{d}{dt} \frac{\partial (T-U)}{\partial \dot{q}^j} -\frac{\partial (T-U)}{\partial q^j}\right) \delta q^j. $$

Here $\delta$ denotes an infinitesimal virtual displacement consistent with the constraints. Moreover, ${\bf F}_i$ is the applied force (i.e. the total force minus the constraint forces) on the $i$'th particle. The Lagrangian $L:=T-U$ is here defined as the difference$^3$ between the kinetic and the potential energy. Note that the rhs of eq. (1) precisely contains the Euler-Lagrange operator.

D'Alembert's principle says that the lhs of eq. (1) is zero. Then Lagrange equations follows from the fact that the virtual displacement $\delta q^j$ in the generalized coordinates is un-constrained and arbitrary.

D'Alembert's principle in turn follows from Newton's laws using some assumptions about the form of constrained forces. (E.g. we assume that there is no sliding friction.) See Ref. 1 and this Phys.SE post for further details.

References:

  1. H. Goldstein, Classical Mechanics, Chapter 1.

--

$^1$ One should always keep in mind that, at the classical level (meaning $\hbar=0$), the Lagrangian $L$ is far from unique, in the sense that many different Lagrangians may yield the same eqs. of motion. E.g. it is always possible to add a total time derivative to the Lagrangian, or to scale the Lagrangian with a constant. See also this Phys.SE post.

$^2$ It is possible to extend to a special relativistic version of Newtonian mechanics by (among other things) replacing the non-relativistic formula $T=\frac{1}{2}\sum_{i=1}^N m_i v^2_i $ with $T=-\sum_{i=1}^N \frac{m_{0i}c^2}{\gamma(v_i)}$ rather than the kinetic energy $\sum_{i=1}^N [\gamma(v_i)-1]m_{0i}c^2$. See also this Phys.SE post.

$^3$ OP is pondering why the Lagrangian $L$ is not of the form $T-\alpha U$ for some constant $\alpha\neq 1$? In fact, the key identity (1) may be generalized as follows

$$\tag{1'} \sum_{i=1}^N \left(\dot{\bf p}_i-\alpha{\bf F}_i\right)\cdot \delta {\bf r}_i ~=~ \sum_{j=1}^n \left(\frac{d}{dt} \frac{\partial (T-\alpha U)}{\partial \dot{q}^j} -\frac{\partial (T-\alpha U)}{\partial q^j}\right) \delta q^j. $$

So the fact that the Lagrangian $L$ is not of the form $T-\alpha U$ for $\alpha\neq 1$ is directly related to that Newton's 2nd law is not of the form $\dot{\bf p}_i=\alpha {\bf F}_i$ for $\alpha\neq 1$.

share|improve this answer
add comment

Let me assume that "first principles" means Newton's laws but in the somewhat more encompassing formulation of Hamilton's equations, which say that given a Hamiltonian function $H$, then the canonical momentum (I display a single one just for each of notation) is related to the velocities by

$$ \dot q = \frac{\partial H}{\partial p} $$

and that the dynamical equation of motion (generalizing $F = m a$) is

$$ \dot p = -\frac{\partial H}{\partial q} \,. $$

So in an infinitesimal time span $\epsilon$ the coordinates and momenta evolve as

$$ q_\epsilon = q + \frac{\partial H}{\partial p} \epsilon $$

and

$$ p_\epsilon = p - \frac{\partial H}{\partial q} \epsilon \,. $$

At the same time, the change of canonical coordinates/canonical momenta is related to the Lagrangian $L$ by ("generating functions for canonical transformations")

$$ p_\epsilon \mathbf{d}q_{\epsilon} - p \mathbf{d}q = \epsilon \mathbf{d}L \,. $$

Now we compute:

$$ \begin{aligned} p_\epsilon \, \mathbf{d} q \epsilon - p \mathbf{d} q & = \left(p - \frac{\partial H}{\partial q} \epsilon \right) \mathbf{d} \left( q + \frac{\partial H}{\partial p} \epsilon \right) - p \mathbf{d}q \\ & = \epsilon \left( p \mathbf{d}\frac{\partial H}{\partial p} - \frac{\partial H}{\partial q} \mathbf{d}q \right) \\ & = \epsilon \left( \mathbf{d}\left( p \frac{\partial H}{\partial p}\right) - \frac{\partial H}{\partial p} \mathbf{d} p - \frac{\partial H}{\partial q} \mathbf{d}q \right) \\ & = \epsilon \mathbf{d} \left( p \frac{\partial H}{\partial p} - H \right) \end{aligned} \,. $$

Hence in general the Lagrangian is

$$ L := p \frac{\partial H}{\partial p} - H \,. $$

Now if $H$ has the standard form (setting $m = 1$ for simplicity)

$$ H = H_{kin} + H_{pot} = \tfrac{1}{2}p^2 + V(q) $$

then

$$ L = H_{kin} - H_{pot} \,. $$

By the way, anyone enjoying a more general abstract perspective on what's going on here might enjoy to learn this story translated to the language of "prequantized Lagrangian correspondences", for more on this see on the nLab here,

share|improve this answer
add comment

I found a wikilink, Lagrange_multiplier, that answers my question:

"Thus, the force on a particle due to a scalar potential, $F=-\nabla V$, can be interpreted as a Lagrange multiplier determining the change in action (transfer of potential to kinetic energy) following a variation in the particle's constrained trajectory."

${\ \ }$In other words, the potential energy $V$ becomes a set of constraints for the Lagrangian $L=T-nV$ where $n$ is the Lagrange multiplier that needs to be determined. The variation
$\delta \int_{t_1}^{t_2}L(\dot q_1,...,\dot q_N,q_1, ..., q_N)dt=0 $
is turned into $2N$ equations, $N$ of which are equations of motion
$\frac{d}{dt} ( \nabla_{\dot q}T)+n \nabla{_q}V=0$
and the other $N$ equations are constraints. It turn out $n=1$.
${\ \ }$The Lagrange multiplier method makes sense because $V$ is path-independent, therefore, its variation along different paths is always zero:
$\delta \int_{t_1}^{t_2}Vdt≡0. $
When we apply variational principle to $\delta \int_{t_1}^{t_2}Ldt≡0$, only the $T$ term varies. When we add $n \int_{t_1}^{t_2}Vdt $ with arbitrary $n$, nothing changes. But if we think the $V$ term as constraints on which the particle moves, then we get the right equations of motion.

share|improve this answer
    
Comment to the answer (v1): Note that the two actions $S[q] := \int \! dt~(\frac{m}{2}\dot{q}^2-V)$ and $S[q,\lambda]:= \int \! dt~(\frac{m}{2}\dot{q}^2-\lambda V)$ describe two different physical theories. The latter model constrains the particle to move on the equipotential surface $V=0$, while the former doesn't. For instance, if $V=\frac{m}{2}(\omega q)^2$ is a harmonic potential in 1D, then the solution is $q=q_0\cos(\omega (t-t_0))$ in the former case, while $q=0$ in the latter case. –  Qmechanic Sep 24 '13 at 11:25
    
It looks like the resulting equations of motion are the same for the 2 settings. I've to scratch my head more. My point is that potential energy plays no role in variation even though it has role in trajectory. –  Chin Yeh Sep 24 '13 at 16:47
add comment

The $n$ in $L=T-nV$ can be seen as a rescaling factor of potential. $n$ does not change the physics. For example for gravity, $n$ can be absorbed in gravitational constant. See also this.

share|improve this answer
    
I think the key lies in the generality of the variational method. The Lagrangian exists as a template. You will not get an answer that does not involve Newton's Laws. I don't think it has a life independent of associated theory. Penrose has discussed these issues in 'The Road To Reality'. –  dj_mummy Sep 26 '13 at 16:14
    
Comment to the answer (v1): What do you mean $n\neq 1$ does not change physics? Changing the gravitational constant $G$ does change physics. –  Qmechanic Sep 30 '13 at 20:12
    
@Qmechanic. I probably didn't phrase it right. Or, I might not have thought it right. What I thought was this. $V$ has no role in the variation. We can add a factor $n$ to $V$, the nature of physics does not change, even though the trajectory does. The example I gave is even more risque. If we double $V$ and reduce gravitational constant by half, we still get the same equations of motion. Disclosure: I wouldn't do any calculation with $n\neq 1$. I just wanted to explain away this lump in my mind. –  Chin Yeh Oct 1 '13 at 20:17
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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