1
$\begingroup$

The symmetric group has two irreducible representations of degree $3$: One is the standard representation and the other is the product of the standard representation with the (degree $1$) sign representation. I'm interested in the latter one. So $S_4$ acts on a $3$-dimensional vector space, say $V \cong \mathbb{Q}^3$ generated by $x_1, x_2, x_3$, by the product of the standard and sign irreps.

Question: Is there a simple description of this action only involving the basis vectors $x_1, x_2, x_3$?

My knowledge of representation theory is still very limited. But from what I know, $S_4$ acts on say $\mathbb{Q}^4$ just by permutation of the basis vectors $v_1, v_2, v_3, v_4$. Then $v_1 + v_2 + v_3 + v_4$ is invariant under the $S_4$-action, hence $S_4$ acts on the $3$-dimensional subspace given by the orthogonal complement of $v_1 + v_2 + v_3 + v_4$. But I really don't understand how to describe this action, let alone how to describe it in the general case above where one does not start with a $4$-dimensional space and consider a $3$-dimensional subspace.

PS: This question arose when I was looking at this mathSE question.

Thanks in advance for any help!

$\endgroup$
  • $\begingroup$ Do you know that this representation is isomorphic to some isomorphisms $\mathfrak{S}_4 \rightarrow Isom^+(C_3)$ where $C_3$ is the cube in $\mathbb{R}^3$ and $Isom^+(C_3)$ is its group of linear isometries with positive determinant? $\endgroup$ – nombre Nov 26 '16 at 0:07
  • $\begingroup$ @nombre No I haven't seen this statement before. I guess this might be helpful cause then we're already in $\mathbb{R}^3$ (or $\mathbb{Q}^3$). Would you be able to explain this a little further? $\endgroup$ – Tom Bombadil Nov 26 '16 at 9:31
  • $\begingroup$ I posted a detailed explaination as an answer. $\endgroup$ – nombre Nov 26 '16 at 10:29
1
$\begingroup$

Define $C_3:=\{\varepsilon_1.e_1 + \varepsilon_2.e_2 + \varepsilon_3.e_3 \ | \ \varepsilon_1,\varepsilon_2,\varepsilon_3 \in \{-1;1\}\}$ where $(e_1,e_2,e_3)$ is the canonical basis of $\mathbb{R}^3$. So $C_3$ is a cube of length $2$ centered on $(0,0,0)$.

Let $Isom^+(C_3)$ denote the set of positive (determinant) linear isometries of the euclidean space $\mathbb{R}^3$ which fix $C_3$. Let $d_1,d_2,d_3,d_4$ respectively denote $\{(1,1,1);(-1,-1,-1)\},\{(1,-1,1);(-1,1,-1)\},\{(-1,-1,1);(1,1,-1)\},\{(-1,1,-1);(1,-1,1)\}$. $d_1,d_2,d_3,d_4$ are the great diagonals of $C_3$.

Since those diagonals realize the greatest distance between points of $C_3$, they are preserved by elements of $Isom^+(C_3)$. This gives us a morphism $\varphi:Isom^+(C_3) \rightarrow \mathfrak{S}_4$ caracterised by $\forall f \in Isom^+(C_3), \forall 1 \leq i \leq 4, d_{\varphi(f)(i)} = f(d_i)$.


Claim: $\varphi$ is an isomorphism.

Indeed, for $f \in Ker(\varphi)$, $f$ fixes the great diagonals, so for each diagonal $d_i$, $f$ fixes or interchanges its elements, which means $d_i \subset E_1(f)$ or $d_i \subset E_{-1}(f)$. Now any $3$-uple of points in three different diagonals form a basis of $\mathbb{R}$, so $E_1(f)$ and $E_{-1}(f)$, whose sum is direct, cannot both contain two diagonals, and one of them has to contain $3$, but then it is of dimension $3$. $E_{-1}(f)$ cannot be of dimension $3$, otherwise $f$ would be $-id_{\mathbb{R}^3}$, which is not positive. So $\dim(E_1(f)) = 3$, and $f = id_{\mathbb{R}^3}$.

Now to prove surjectivity, one only has to find antecedents for a system of generators of $\mathfrak{S}_4$, for instance $(1 \ 2)$ and $(1 \ 2 \ 3 \ 4)$. See that $\varphi(r_{\pi}) = (1 \ 2)$ and $\varphi(r_{\frac{\pi}{2}}) = (1 \ 2 \ 3 \ 4)$ where $r_{\pi}$ is the rotation with angle $\pi$ and axis $\mathbb{R}.(e_1+e_3)$, and $r_{\frac{\pi}{2}}$ is the rotation with angle $\frac{\pi}{2}$ and axis $\mathbb{R}.e_3$.


So $\varphi$ is an isomorphism, and $\varphi^{-1}$ is a representation of $\mathfrak{S}_4$.

To compute the values of the associated character $\chi$, one can use the fact that the trace of a rotation in $\mathbb{R}^3$ with angle $\theta$ is $1 + 2\cos(\theta)$. This yields $\chi(1 \ 2) = -1$ and $\chi(1 \ 2 \ 3 \ 4) = 1$.

$\varphi^{-1}(1 \ 2 \ 3)$ has order $3$ in dimension $3$ so its trace $\chi(1 \ 2 \ 3)$ is $1 + j + j^2 = 0$.

$\varphi^{-1}((1 \ 2)(3 \ 4))$ has order $2$ and positive determinant so its trace $\chi((1 \ 2)(3 \ 4))$ is $1 + (-1) + (-1) = -1$.

You can check that this is the same character as the one you were looking for.


This type of proof is quite classic when dealing with representions of $\mathfrak{S}_n$. If you want a degree $p$ representation of $\mathfrak{S}_n$, try to realize it with isometries of a subset $X$ of the euclidean $\mathbb{R}^p$. To this effect, try finding $n$ things that are permutated by those isometries, and you get a morphism $Isom(X) \rightarrow \mathfrak{S}_n$. If that morphism is surjective (so if $(1 \ 2)$ and $(1 \ 2 \ ... \ n)$ are values), then $\mathfrak{S}_n$ is a quotient of $Isom(X)$ and with luck, this quotient is a subgroup of $Isom(X)$. Now $Isom(X)$ is a subgroup of $Isom(\mathbb{R}^p)$, and this is nice because elements of $Isom(X)$ are actually maps on the whole space, so they can act on many things, giving you other representations of $\mathfrak{S}_n$. For instance, $Isom^+(C_3)$ permutes $e_1,e_2,e_3$, and this gives another degree $3$ representation of $\mathfrak{S}_4$ which happens to reduce to the direct sum of the trivial representation on $\mathbb{R}.(e_1+e_2+e_3)$, and the degree $2$ subrepresentation on $Ker((e_1+e_2+e_3)^*)$, which is the degree $2$ irreductible representation of $\mathfrak{S}_4$.

Of course, you can also find representations of $\mathfrak{S}_n$ without geometric tools, and you might enjoy more equivalent algebraic tools.

$\endgroup$
2
$\begingroup$

If $V$ is spanned by the basis $\{v_1, v_2, v_3, v_4\}$, and $v = v_1 + v_2 + v_3 + v_4$, then (as you have observed) the action of $S_4$ on $V$ fixes $v$. Therefore, there is a natural action of $S_4$ on the three-dimensional quotient $\overline{V} = V/\langle v \rangle$. This might be easier than thinking of the action as being on the orthogonal complement.

Then $\overline{V} \otimes W$, where $W$ is a one dimensional space with sign action, gives the other three-dimensional representation of $S_4$. If you haven't come across the tensor product yet, it is the way to construct a representation whose character is the product of the characters of two given representations.

You can describe these actions directly in terms of a basis $\{x_1,x_2,x_3\}$, by choosing a basis for $\overline{V}$ (or $\overline{V} \otimes W$), and then calculating the actions of $S_4$. But it will be a bit artificial because it forces you to break the inherent symmetry of $v_1, v_2, v_3, v_4$.

$\endgroup$
  • $\begingroup$ No need to break the symmetry. Just compute the action of $S_4$ on $V \otimes W$ - the resulting space has a similar one-dimensional subrepresentation, which you can then quotient out, or take its orthogonal complement. $\endgroup$ – Dustan Levenstein Nov 24 '16 at 8:31
  • $\begingroup$ @Ted Thank you for your answer. I guess what I'm asking for is really what you call "a bit artificial". From your description it seems like it really depends on the basis right? Are there any statements I can make that are independent of the basis of $\bar{V} \otimes W$? (Im a bit unsure myself of what I'm looking for here... Again I mention this question, which is what I really want to solve.) $\endgroup$ – Tom Bombadil Nov 24 '16 at 9:05

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.