I was computing the dual map for $k$-forms in Minkowski spacetime, and I found that any $2$-form is either self-dual or anti-self-dual if and only if it is the null form. Does this result make any sense to you? Here are the calculations.

Let the metric on Minkowski spacetime be

$$ (\eta_{\mu\nu})= \begin{pmatrix} 1&0&0&0\\ 0&-1&0&0\\ 0&0&-1&0\\ 0&0&0&-1 \end{pmatrix} $$

and let $\omega$ be the $2$-form

$$ \omega=\omega_{\mu\nu}\ dx^{\mu}\wedge dx^{\nu}=(\omega_{\mu\nu}-\omega_{\nu\mu})\ dx^{\mu}\otimes dx^{\nu} $$

The $\eta$-raised components of $\omega$ are

$$ (\omega)^{\mu\nu}=\eta^{\mu\lambda}\eta^{\nu\tau}\,(\omega_{\lambda\tau}-\omega_{\tau\lambda}) $$

We have

$$ \eta^{\mu\lambda}\eta^{\nu\tau}\omega_{\lambda\tau}=(\eta\omega\eta)^{\mu\nu} $$

Thus, if

$$ \omega_{\lambda\tau}= \begin{pmatrix} 0&\omega_{01}&\omega_{02}&\omega_{03}\\ \omega_{10}&0&\omega_{12}&\omega_{13}\\ \omega_{20}&\omega_{21}&0&\omega_{23}\\ \omega_{30}&\omega_{31}&\omega_{32}&0 \end{pmatrix} $$

$\eta\omega\eta$ has components

$$ (\eta\omega\eta)^{\mu\nu}= \begin{pmatrix} 0&-\omega_{01}&-\omega_{02}&-\omega_{03}\\ -\omega_{10}&0&\omega_{12}&\omega_{13}\\ -\omega_{20}&\omega_{21}&0&\omega_{23}\\ -\omega_{30}&\omega_{31}&\omega_{32}&0 \end{pmatrix} $$

The $\eta$-raised components of $\omega$ are thus

$$ (\omega)^{\mu\nu}= \begin{pmatrix} 0&\omega_{10}-\omega_{01}&\omega_{20}-\omega_{02}&\omega_{30}-\omega_{03}\\ \omega_{10}-\omega_{10}&0&\omega_{12}-\omega_{21}&\omega_{13}-\omega_{31}\\ \omega_{20}-\omega_{20}&\omega_{21}-\omega_{12}&0&\omega_{23}-\omega_{32}\\ \omega_{30}-\omega_{30}&\omega_{31}-\omega_{13}&\omega_{32}-\omega_{23}&0 \end{pmatrix} $$

Let $*\,\omega$ denote the Hodge-dual of $\omega$, such that (as $\sqrt{|\det(\eta)|}=1$)

$$ (*\,\omega)_{\tau\lambda}=\frac{1}{2}\ (\omega)^{\mu\nu}\,\epsilon_{\mu\nu\tau\lambda} $$

Then, for example

$$ (*\,\omega)_{01}=\frac{1}{2}\ (\omega)^{\mu\nu}\,\epsilon_{\mu\nu 01}=\frac{1}{2}\,(\omega^{23}-\omega^{32})=\omega_{23}-\omega_{32} $$ and

$$ (*\,\omega)_{23}=\frac{1}{2}\ (\omega)^{\mu\nu}\,\epsilon_{\mu\nu 23}=\frac{1}{2}\,(\omega^{01}-\omega^{10})=\omega_{10}-\omega_{01} $$ Thus

$$ \begin{cases} (*\,\omega)_{23}=(\omega)_{23}\\ \\ (*\,\omega)_{01}=(\omega)_{01} \end{cases} $$ implies

$$ \begin{cases} \omega_{01}=\omega_{10}\\ \\ \omega_{23}=\omega_{32} \end{cases} $$ and the same goes for the other indices. The result is the same provided that

$$ \begin{cases} (*\,\omega)_{23}=-(\omega)_{23}\\ \\ (*\,\omega)_{01}=-(\omega)_{01} \end{cases} $$

  • $\begingroup$ I think you mean $(\star \omega)_{23} = \omega_{10}$? $\endgroup$ – Muphrid Jun 21 '15 at 20:04
  • $\begingroup$ Where, exactly? $\endgroup$ – Giorgio Comitini Jun 21 '15 at 20:07
  • $\begingroup$ Nevermind. By "null form" you mean the zero 2-form? $\endgroup$ – Muphrid Jun 22 '15 at 2:28
  • $\begingroup$ Yes, I mean the zero 2-form $\endgroup$ – Giorgio Comitini Jun 22 '15 at 7:33

Yes, this is so. A simpler way to see this is to use clifford algebra. Suppose $\omega$ is (anti-)self-dual. Then the clifford product of $\omega$ and the unit pseudoscalar $\epsilon$ is

$$\epsilon \omega = \pm \omega$$

$\epsilon$ is invertible under the clifford product, yielding the equation

$$\omega= \pm \epsilon^{-1} \omega \implies \epsilon^{-1} \omega = \pm \omega$$

But $\epsilon^{-1} = -\epsilon$ for Minkowski spacetime, yielding

$$\epsilon^{-1} \omega = \pm \omega \implies -\epsilon \omega = \pm \omega \implies \epsilon \omega = \mp \omega$$

which is a contradiction from what we initially supposed.

In 4d Euclidean space, there is no contradiction, as $\epsilon^{-1} = \epsilon$ there.

Edit: in notation only using Hodge star, this is basically the same result as saying that, in Minkowski spacetime,

$$\star \star \omega = -\omega$$

So if $\star \omega = \pm \omega$, then $\star (\pm \omega) = + \omega$ on the one hand, but also $-\omega$ on the other.

  • $\begingroup$ Right, thank you so much. I didn't notice that in Minkowski spacetime you need to change the sign for the bidual operator (from $(-1)^{k(n-k)}$ to $(-1)^{k(n-k)+1}$). $\endgroup$ – Giorgio Comitini Jun 22 '15 at 16:46
  • $\begingroup$ Hi Muphrid. I have a question regarding the implications of the lack of anti/self-dual forms on Minkowski spacetime. Is there any chance that I could ask for your opinion in private? $\endgroup$ – Giorgio Comitini Jun 23 '15 at 9:15
  • $\begingroup$ I'm open to suggestion regarding how to go about that. $\endgroup$ – Muphrid Jun 23 '15 at 14:50
  • $\begingroup$ Well, I was thinking about doing it in a chatroom. It's better than nothing, and there's no need for extra privacy :-) $\endgroup$ – Giorgio Comitini Jun 23 '15 at 15:04
  • $\begingroup$ Fair enough, just let me know where and when. $\endgroup$ – Muphrid Jun 23 '15 at 15:12

Similarly to the second part of Muphrid answer, let $* \omega=\pm\omega$. Then applying $*$ again \begin{equation} *^2\omega=* (\pm\omega) =\omega \end{equation} on the other hand on a Lorentzian $n$-manifold $*^2=-(-1)^{p(n-p)}$ where $p$ is the degree of the form on which $*$ is acting. In this case $n=4,p=2$ hence we get $*^2\omega = - \omega$, therefore \begin{equation} \omega = -\omega \Rightarrow \omega = 0. \end{equation}


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.