Lagrangian Dynamics#
Contents
When is \(L = K - V\)?#
I make a big deal about how the prescription of forming the Lagrangian \(L = K-V\) as the difference between kinetic and potential energies is only valid for Newtonian mechanics. How does one know if this is appropriate?
The way I think about it is to imagine writing the full Lagrangian for all of the particles in terms of their Cartesian inertial coordinates \(\vect{r}_i = (x_i, y_i, z_i)\), with all other complications inserted as constraints with Lagrange multipliers etc. If, in this case, the kinetic energy is quadratic in the velocities, then you can use the \(L = K - V\) prescription after transforming to whatever (non-singular) coordinates make the problem easiest to solve:
Note: the “mass matrix” \(\mat{M}\) does not need to be diagonal.
To see how this fails, consider special relativity. Here, it is natural to start from the Hamiltonian formulation with the relativistic energy-momentum relation (here, with no potential \(V(x)=0\))
Solving for \(p(v)\) we have the relativistic kinetic energy in terms of the time-dilation factor \(\gamma(v/c)\):
However, with a little bit of work, one can find that the Lagrangian for a relativistic particle should actually be:
which can be obtained via the Legendre transform, but is not of the form \(E(v) - V(x)\).