Again I have issues with notations. The hodge star operator is defined as :

(m is the dimension of the manifold)

$$\star: \Omega^{r}(M) \rightarrow \Omega^{m-r}(M)$$

$$\star(dx^{\mu_{1}} \wedge dx^{\mu_{2}} \wedge ...\wedge dx^{\mu_{r}}) = \frac{\sqrt{|g|}}{(m-r)!}\epsilon^{\mu_{1}\mu_{2}...\mu_{r}}_{\nu_{r+1}...\nu_{m}}dx^{\nu_{r+1}}\wedge...\wedge dx^{v_m}$$


$$\epsilon^{\mu_{1}\mu_{2}...\mu_{m}}= g^{\mu_{1}\nu_{1}}g^{\mu_{2}\nu_{2}}...g^{\mu_{m}\nu_{m}}\epsilon_{\nu_{1}\nu_{2}...\nu_{m}}=g^{-1}\epsilon_{\mu_{1}\mu_{2}...\mu_{m}}$$

With an r-form

$$\omega = \frac{1}{r!}\omega_{\mu_{1}\mu_{2}...\mu_{r}}dx^{\mu_{1}} \wedge dx^{\mu_{2}} \wedge...\wedge dx^{\mu_{r}} \in \Omega^{r}(M)$$


$$\star\omega = \frac{\sqrt{|g|}}{r!(m-r)} \omega_{\mu_{1}\mu_{2}...\mu_{r}}\epsilon^{\mu_{1}\mu_{2}...\mu_{r}}_{\nu_{r+1}...\nu_{m}}dx^{\nu_{r+1}}\wedge...\wedge dx^{\nu_m}$$

Now I would like to derive these results

(orthogonal metric and doesn't matter if forms or vectors)

$$\star(e_{2} \wedge e_{3})=e_{1}$$

$$\star(e_{1} \wedge e_{3})=-e_{2}$$

$$\star(e_{1} \wedge e_{2})=e_{3}$$

Another example, let me calculate $r=2$, $m=3$

$$\star(dx \wedge dy)=\sqrt{|g|}\epsilon^{xy}_{\nu_{2}\nu_{3}}dx^{\nu_{3}} \wedge dx^{\nu_3}$$

I have no clue what's going on, is $\nu$ different from $\mu$ ? How does this machinery work?

I know the formula is long and annoying, but can someone give a clear example of how this works?

  • 1
    $\begingroup$ Hint: If $m = 3$, then your last formula has too many $\nu$ on the right hand side. Also, this looks like a pure math question to me. $\endgroup$ Jul 21 '14 at 0:56
  • 1
    $\begingroup$ @ACuriousMind Pure math questions that arise in a physical context have recently become on-topic. Arguably, no such context has been explicitly given in this question, but I think it's clear what the physical motivation for it is. $\endgroup$
    – Wouter
    Jul 21 '14 at 3:18
  • $\begingroup$ The dual of a $2$-form is a $1$-form in this case. The Hodge dual associates $k$ forms with $n-k$ forms uniquely. $\endgroup$ Jul 21 '14 at 3:55
  • $\begingroup$ Your expression for $*(dx\wedge dy)$ at the end isn't correct: since $r+1=2+1=3$ and $m=3$, the $\nu_2$ shouldn't be there. Similarly, the RHS should only be a one-form. $\endgroup$ Jul 21 '14 at 3:58
  • 1
    $\begingroup$ The first example is explicitely explained here. $\endgroup$
    – Dilaton
    Jul 22 '14 at 12:39

You have the formulae $$\omega = \frac{1}{r!}\omega_{\mu_{1}\mu_{2}...\mu_{r}}dx^{\mu_{1}} \wedge dx^{\mu_{2}} \wedge...\wedge dx^{\mu_{r}} \in \Omega^{r}(M)$$ $$\star\omega = \frac{\sqrt{|g|}}{r!(m-r)} \omega_{\mu_{1}\mu_{2}...\mu_{r}}\epsilon^{\mu_{1}\mu_{2}...\mu_{r}}_{\nu_{r+1}...\nu_{m}}dx^{\nu_{r+1}}\wedge...\wedge dx^{\nu_m}\in \Omega^{m-r}(M)$$ in general. Let's consider your $(r,m)=(2,3)$ case. Then these become $$\omega = \frac{1}{2!}\omega_{\mu_{1}\mu_{2}}dx^{\mu_{1}} \wedge dx^{\mu_{2}}\in\Omega^{2}(M)\implies \star\omega = \frac{\sqrt{|g|}}{2} \omega_{\mu_{1}\mu_{2}}\epsilon^{\mu_{1}\mu_{2}}_{\nu_{3}}dx^{\nu_{3}}\in \Omega^{1}(M)$$ Note that only $\nu_3$ shows up once since $r+1=m=3$. So $\star\omega$ is indeed a 1-form. Concretely, take the metric to be orthonormal i.e. $\sqrt{|g|}=1$, and suppose $\omega_{12}=-\omega_{21}=1$ and $\omega_{ij}=0$ otherwise. Then from the antisymmetry of the wedge product and the $\epsilon$ symbol, we have $\omega=dx_1\wedge dx_2$ and $$\star \omega =\frac{1}{2}\left(\epsilon^{12}_\nu dx^\nu-\epsilon^{21}_\nu dx^\nu\right)=\epsilon^{12}_\nu dx^\nu=\epsilon^{12\nu}dx_\nu=dx_3$$ as expected. Note that in the last two steps we have implicitly used $g_{\mu\nu}$ to raise/lower the index $\nu$ and $\epsilon^{12k}=\delta_{3k}.$

  • $\begingroup$ Hey thank you, that helped ! I was confused because of this $v_{3}$ index, but it has no meaning other then summation I guess. $\endgroup$
    – user117640
    Jul 25 '14 at 20:29
  • $\begingroup$ That's the game that's always played, yeah. Each repeated index is just a dummy. (That's why, you'll notice, I called it as $\nu$ in the last line---after all, it doesn't matter what it's name is as long as I'm consistent with it.) $\endgroup$ Jul 25 '14 at 20:30

It may sometimes be easier to compute Hodge duals using an orthogonal basis and clifford algebra.

Clifford algebra allows you to write a lot of complicated tensor expressions as simple products of vectors, using the "geometric product". If $a, b$ are orthogonal vectors with respect to some metric, and $u, v, w$ are arbitrary vectors, then under the geometric product,

$$aa \equiv g(a,a) = |a|^2, \quad ab = -ba, \quad (uv)w = u(vw)$$

So this captures the inner product, encoded in the metric; the need for antisymmetric tensors in $ab = - ba$, and associativity allows us to drop unneeded parentheses and talk about long products of vectors as entities in their own right.

Now then, suppose you have orthogonal coordinate one-forms $e^t, e^x, e^y, e^z$. Then there is a well-defined volume-form $i$:

$$i \equiv e^t e^x e^y e^z$$

Again, note that I said orthogonal. What happens if you take a two-form like $e^x e^y$ and multiply it by $i$?

$$i e^x e^y = e^t e^x e^y e^z e^x e^y =e^t (e^x e^y)(e^x e^y) e^z = - e^t e^z g^{xx} g^{yy} $$

where $g^{xx}$ and $g^{yy}$ arose from the inner products of forms. The sign might be "wrong" for a given convention, but this is doing exactly the same sort of thing the Hodge dual does.

I'll illustrate that, in the case of 3d with a Cartesian metric, you get the same results this way as you would using the index definition. In 3d, $i = e_1 e_2 e_3$, so we get

$$\star e_1 = i e_1 = e_1 e_2 e_3 e_1 = -e_1 e_2 e_1 e_3 = + e_1 e_1 e_2 e_3 = e_2 e_3$$

The other equalities follow similarly.

  • $\begingroup$ Hi, I'm sorry, but I know nothing about Clifford Algebra. But thx for the afford. $\endgroup$
    – user117640
    Jul 22 '14 at 20:38

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.