I am learning classical field theory at the moment. Right now, I am trying to understand where the Legendre transform, or the multi-symplectic structure emerges.

To make the setting clear, say we have a fiber bundle $Y \overset{\pi}{\to} X$ over a $n$-manifold $X$. A Lagrangian density is a $n$-form on $X$ depending on the $1$-jet of a section of $Y$ : $\mathcal{L}\in\Gamma(J^1\pi,\pi^*\Lambda^n X)$.

The affine Legendre transform is a map $J^1\pi \to \Lambda^n_1 Y$ that is essentially the affine approximation to $\mathcal L$ at each point:

$$ FL : s \mapsto \mathcal{L}(s) + d_V\mathcal{L} \in \Lambda^n X \oplus \Lambda^n X\otimes V^* \simeq \Lambda^n_1 Y $$

where $V$ is the vertical bundle of $\pi$ and $d_V$ the vertical derivative, well-defined because $\pi^*\Lambda^n X$ is trivial above fibers of $\pi$.

In this approach, the Poincaré-Cartan $n$-form is defined by $\Theta_{\mathcal L} = FL^*\Theta$ where $\Theta$ is the tautological form on $\Lambda^n T^*Y$.

The key feature of the Poincaré-Cartan form is that $\mathcal L - \Theta_{\mathcal L}$ is contact, that is to say it vanishes on holonomic sections of $J^1Y$, that is $1$-jet of sections of $Y$. Hence there is the fundamental identity \begin{equation}\tag{1}\label{contact} \forall \phi\in\Gamma(\pi), j^1(\phi)^*\Theta_{\mathcal L} = j^1(\phi)^*\mathcal L = {\mathcal L}\circ j^1(\phi) \end{equation}

Now, these forms can be obtained from a variational principle [1]: for $\phi$ a section over $U_X$, the action $S(\phi)$ is defined as $\int_{U_X} \mathcal{L}\circ j^1(\phi)$. For a variation $V$ of $\phi$, the variation of $S$ can be expressed as: $$ dS_\phi\cdot V = \int_{U_X} j^1(\phi)^*(j^1(V)\lrcorner d\Theta_{\mathcal L}) + \int_{\partial U_X} j^1(\phi)^*(j^1(V)\lrcorner \Theta_{\mathcal L}) $$

Now, the article [1] mentions at p.27 (Corollary 4.2) that from this very identity we can obtain (\ref{contact}). I am missing an argument, so I am unable to derive (\ref{contact}), but I am very interested by the perspective of justifying the Legendre transform from this variational principle.

Using the more concise language of the variational bicomplex (notations and sign differ), if the Euler-Lagrange form $EL$ is defined by

$$ d_V\mathcal{L} = EL + d_H\Theta_{\mathcal{L}} $$

then it seems that $EL = d\Theta_{\mathcal{L}}$ which I can't explain.

[1] : Multisymplectic geometry, variational integrators, and nonlinear PDEs - Jerrold E. Marsden, George W. Patrick, Steve Shkoller


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.