5
$\begingroup$

The electromagnetic field tensor is $F_{\mu\nu}=\partial_\mu A_\nu - \partial_\nu A_\mu$. I am trying to calculate the quantity $$ \frac{\partial(F_{\alpha\beta}F^{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}. $$ This calculation arises when trying to derive the electromagnetic equations of motion (i.e. Maxwell's equations) from the Lagrangian $\mathcal{L}=C F_{\mu\nu}F^{\mu\nu}$. According to P. 14 of these online notes, this derivative is $$ \frac{\partial(F_{\alpha\beta}F^{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}=2F^{\alpha\beta}\frac{\partial F_{\alpha\beta}}{\partial(\partial_\mu A_{\nu})} .$$

This result surprises me. I can use the product rule to find

$$ \frac{\partial(F_{\alpha\beta}F^{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}=F_{\alpha\beta}\frac{\partial(F^{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}+F^{\alpha\beta}\frac{\partial(F_{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})} $$

and it is clear that if $F_{\alpha\beta}\frac{\partial(F^{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}=F^{\alpha\beta}\frac{\partial(F_{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}$ then you get the desired result. However, I can't see why this is true. In particular, I don't understand how to take the derivative $$\frac{\partial(F^{\alpha\beta})}{\partial(\partial_{\mu}A_{\nu})}=\frac{\partial(\partial^{\alpha}A^{\beta}-\partial^{\beta}A^{\alpha})}{\partial(\partial_{\mu}A_{\nu})}=\frac{\partial(\partial^{\alpha}A^{\beta})}{\partial(\partial_{\mu}A_{\nu})}-\frac{\partial(\partial^{\beta}A^{\alpha})}{\partial(\partial_{\mu}A_{\nu})} $$ where the downstairs part has lower indices and the upstairs part has upper indices.

$\endgroup$
1
  • $\begingroup$ Can't you just use the metric to lower the indices in the numerator? $\endgroup$ May 22 '18 at 20:39
2
$\begingroup$

For your first question: the components of the metric don't depend on $\partial_\mu A_\nu$, or for that matter, anything at all. So we have, e.g. $$J^\mu \partial K_\mu = J^\mu \partial (\eta_{\mu\nu} K^\nu) = J^\mu \eta_{\mu\nu} \partial K^\nu = J_\nu \partial K^\nu = J_\mu \partial K^\mu$$ where $\partial$ stands for any kind of derivative whatsoever and $J$ and $K$ are arbitrary. The proof for your case is identical.

For your second question: simply do the same thing. Note that $$\frac{\partial J^\nu}{\partial J_\mu} = \frac{\partial (\eta^{\rho\nu} J_\rho)}{\partial J_\mu} = \eta^{\rho\nu}\frac{\partial J_\rho}{\partial J_\mu} = \eta^{\rho \nu} \delta^\mu_\rho = \eta^{\mu\nu}$$ where you can adapt this reasoning to your own example. After you do this a couple times, it becomes completely second nature, and you won't have to write out the steps. Everything works out exactly how you would expect, just "lining up the indices", $$\frac{\partial J^\nu}{\partial J_\mu} = \eta^{\mu\nu}, \quad \frac{\partial J^\nu}{\partial J^\mu} = \eta^{\nu}_\mu, \quad \frac{\partial J_\nu}{\partial J_\mu} = \eta_\nu^\mu, \quad \frac{\partial J_\nu}{\partial J^\mu} = \eta_{\mu\nu}$$ where, in order to write all four results the same way, I defined $\eta^\mu_\nu = \delta^\mu_\nu$.

$\endgroup$
2
$\begingroup$

Use a metric. Write F with upper indices as F with lower indices times the appropriate metric tensor summed over indices.

$$ F^{ab} = g^{ac}g^{bd}F_{cd} $$

If you are on flat Minkowski space-time then g is a constant diagonal matrix and need not be differentiated. If you are one a curved space-time then the partial of g will be required and there are ways to handle that.

Looking at the term that is causing an issue, on Minkowski space (flat space) the factors of g can be passed out of the derivative and then acted on the tensor outside the derivative to raise indices.

$\endgroup$

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.