I'm trying to follow the answer to this post, that cites the identity $$ \epsilon_{i_{1} \ldots i_{n}} A_{j_{1}}^{i_{1}} \cdots A_{j_{n}}^{i_{n}}=\operatorname{det} A \epsilon_{j_{1} \ldots j_{n}} $$ as a proof of the invariance of the Levi-Civita tensor. How would I prove this identity? It seems like just the definition of the determinant, but I'm not sure about the second Levi-Civita symbol on the right hand side. I don't know if the definition of determinant is trivial, or if I'm missing a critical step?

Once I have this identity, then the invariance follows obviously because rotations are in $SU(2)$, but I'm stuck on this identity, any help would be appreciated.

  • $\begingroup$ Personally I would say this follows by definition; if you don’t like that, what other definition of the determinant would you like an answer to start from? $\endgroup$
    – knzhou
    Feb 6, 2020 at 17:18
  • $\begingroup$ this definition only gives it terms of 1 Levi-Civita, so I would say the left hand side of the equation is just equal to the determinant, without the levi-civita on the RHS, and I'm confused why that's not true. $\endgroup$ Feb 6, 2020 at 17:22
  • $\begingroup$ Why not just multiply it by a rotation matrix and show that is is the same? $\endgroup$
    – Jake
    Feb 6, 2020 at 17:25
  • $\begingroup$ @JakeRose I'm more interested in the identity than what it's used for tbh. $\endgroup$ Feb 6, 2020 at 17:26

1 Answer 1


Consider the following expression : $$\epsilon_{i_1i_2...i_n}A^{i_1}_{1}A^{i_2}_{2}...A^{i_n}_{n}$$

Now THIS is exactly the definition of the determinant of $A$. Why ? Well the determinant is defined as $$detA=\sum_{\sigma\in S_n} \epsilon(\sigma)A^{\sigma(1)}_{1}A^{\sigma(2)}_{2}...A^{\sigma(n)}_{n}$$

Which, if you look hard enough, is exactly the same as the above expression, you can see the sum over all permutations is handled by the contraction with the Levi-Civita tensor, and the sign $\epsilon(\sigma)$ comes from the sign of the Levi-Civita.

But now we have all we need. Indeed, consider the tensor defined by $T_{j_1j_2...j_n}=\epsilon_{i_1i_2...i_n}A^{i_1}_{j_1}A^{i_2}_{j_2}...A^{i_n}_{j_n}$. You can easily show that it is completely anti-symmetric in its indices, try it !

But totally anti-symmetric tensors of rank n are necessarily proportional to the Levi-Civita tensor !(So here, $T_{j_1j_2...j_n}=C\epsilon_{j_1j_2...j_n}$) Indeed, once you know one of the elements, all the other elements of the tensor are determined using the antisymmetry. That's great, because we know : $T_{12..n}=detA=C\epsilon_{12...n}=C$.

Plugging back in yields $$T_{j_1j_2...j_n}=detA\epsilon_{j_1j_2...j_n}$$ And this is exactly as we wished to show.

  • 1
    $\begingroup$ +1, or said differently this is the definition of the determinant plus the fact: permuting the columns of a matrix multiplies the determinant by the sign of this permutation which is the epsilon on the RHS of the identity $\endgroup$ Feb 6, 2020 at 18:20

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.