I am going through David Tong's QFT course, for which lecture notes and exercises are available online at http://www.damtp.cam.ac.uk/user/tong/qft.html.

In Question 1.8 we have the Lagrangian (density) $$L = -\frac{1}{4} F^{\mu \nu} F_{\mu \nu} + \frac{1}{2} m^2 C_\mu C^\mu,$$ $$F_{\mu \nu} = \partial_\mu C_\nu - \partial_\nu C_\mu,$$ which is like the standard electromagnetic field in the case $m=0$. I eventually derive the conjugate momenta $\Pi_\mu$ to $C_\mu$ and convert the Lagrangian to a Hamiltonian $$H = -\frac{1}{2} \Pi_i \Pi^i + \frac{1}{4}F^{ij}F_{ij} - \frac{1}{2} m^2 C^{\mu}C_{\mu} - \Pi_i \partial^i C_0,$$ answering the question.

However in a pdf of tutor's solutions I came across online (which I maybe shouldn't link), the tutor comments and interprets further: they rearrange the last term, $$H = -\frac{1}{2} \Pi_i \Pi^i + \frac{1}{4}F^{ij}F_{ij} - \frac{1}{2} m^2 C^{\mu}C_{\mu} - C^0(\partial_i \Pi^i) - \partial_i (\Pi^i C^0),$$ and comment

[the term] involves an irrelevant three-divergence term. Since the remainder of the Hamiltonian contains no derivatives in $C^0$, $C^0$ may be regarded as a multiplier that, in the $m=0$ theory, imposes the constraint $\nabla \cdot \Pi = m^2 C^0 = 0$, which is precisely Gauss' law.

Since we are back to examining the $m=0$ case, this is a statement about the standard electromagnetic field.

I don't understand either statement here.

How is $\partial_i (\Pi^i C^0)$ "irrelevant"? Can we just ignore this divergence, which as far as I can see has a nonzero value?

$- C^0(\partial_i \Pi^i)$ could be a (Lagrange) multiplier, how is it rearranged to include the $m^2$ term and (together) constrain to $\nabla \cdot \Pi=0$?


2 Answers 2


What you found is actually the Hamiltonian Density $\mathcal{H}$. The Hamiltonian is the spatial integral of Hamiltonian Density

\begin{equation} H=\int_\mathcal{M} d^Nx\,\mathcal{H}(x). \end{equation}

If you have an spatial divergence in your Hamiltonian Density, its contribution to the Hamiltonian can be converted into a hypersurface integral using divergence theorem

\begin{equation} \int_\mathcal{M} d^Nx\,\partial_i\Pi^i=\int_{\partial\mathcal{M}}d\sigma\,n_i \Pi^i, \end{equation}

where $\partial\mathcal{M}$ is the boundary hypersurface, and $n_i$ is the normal vector to it. As that boundary is in infinite, where $\Pi^i$ must vanish, the contribution of the $\partial_i\Pi^i$ term to the Hamiltonian Density is nule.

The Lagrange multiplier would be just $C_0$ in the $m^2=0$ case, enforcing $\partial_i \Pi^i=0$. When $m^2\neq0$ that constraint is not true.


Main points:

  • A total spacetime divergence in the Lagrangian (or Hamiltonian) does not change the EOMs, cf. e.g. this Phys.SE post.

  • If we know that the fields vanishes on the boundary, e.g. by imposing pertinent boundary conditions, we can use the divergence theorem to argue that a divergence term cannot contribute to, say the EM energy.

  • The EOM for $C_0$ reads $\nabla \cdot \Pi = m^2 C^0$.

  • In the massless limit $m=0$, this EOM becomes Gauss's law in vacuum.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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