Given the Lagrangian,

$$L = \frac{1}{2}\dot{q}_i M_{ij} \dot{q}_j + f(q,\dot(q),t)$$

where $M_{ij}$ is a non-degenerate matrix and $q,\dot{q}$ are generalised coordinates & velocities and summation over repeated indices is assumed. I came across a question which asks to show only the symmetric part of $M$ enters the Euler-Lagrange equations. My attempt was to simply calculate the E.O.M as being:
$$ \frac{1}{2}\frac{d}{dt}(M_{ij}\dot{q}_j+\dot{q}_jM_{ji}) +g(q,\dot{q},t)=0 $$ where $g$ depends on the derivatives of $f$.

So I was able to show this, however I am confused as to why only the symmetric part stays. If anyone could clarify my misunderstanding that would be great.

  • 1
    $\begingroup$ Is the antisymmetric part present in the Lagrangian? $\endgroup$ – Herr_Mitesch Apr 10 '18 at 16:57
  • $\begingroup$ Technically, the coordinates could also be fermionic, so this question is then pointless. $\endgroup$ – DanielC Apr 10 '18 at 20:14

The symmetric part of $M$ is defined to be $M_S = (M + M^T)/2$.

Note that $\dot{q}_j M_{ji} = M_{ji}\dot{q}_j = M^T_{ij}\dot{q}_j$, so you can write your EOM as $$\frac{d}{dt}((M_{ij}+M^T_{ij})/2 \,\dot{q}_j +\dots = \frac{d}{dt}(M_{S,ij} \dot{q}_j)$$

More fundamentally,you can split $M$ into its symmetric and antisymmetric part, $M = M_S + M_A$, where $M_S^T = M_S$ and $M_A^T = -M_A$. Since $\dot{q}_i\dot{q}_j =\dot{q}_j\dot{q}_i$ we can see that $$\dot{q}_i M_{A,ij}\dot{q}_j = \dot{q}_j M_{A,ji}^T \dot{q}_i = -\dot{q}_i M_{A,ij} \dot{q}_j = 0$$

So $\dot{q}^T M \dot{q} = \dot{q}^T (M_S + M_A) \dot{q} = \dot{q}^T M_S \dot{q} + 0 $. Only the symmetric part of $M$ actually contributes to the Lagrangian.


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.