Background: Young symmetrizer $c_\lambda$ gives an explicit description of Schur module $S_\lambda V$, which is also the kernel of maps between exterior products (as in Fulton & Harris).

Example: Consider the Young symmetrizer $c_{(21)} = () + (12) - (13) - (132)$, constructed from:

$1 2\\3$

Then its image in $\otimes^3 V$ is spanned by:

$v_1 \otimes v_2 \otimes v_3 + v_2 \otimes v_1 \otimes v_3 - v_3 \otimes v_2 \otimes v_1 - v_3 \otimes v_1 \otimes v_2$

which also gives $S_{(2,1)}V$. And we have:

$S_{(2,1)}V = Ker(\Lambda^2V \otimes V \to \Lambda^3V )$.


Question 1: The dimension of $S_{(2,1)}V$ shall be $\frac{(n+1)n(n-1)}{3}$, where $n = dim(V)$. However, I am not sure how to directly see the expression:

$v_1 \otimes v_2 \otimes v_3 + v_2 \otimes v_1 \otimes v_3 - v_3 \otimes v_2 \otimes v_1 - v_3 \otimes v_1 \otimes v_2$

has dimension $\frac{(n+1)n(n-1)}{3}$ in $\otimes^3 V$.

That is, without using $\dim( S_{(2,1)}) = \dim(\Lambda^2V \otimes V) - \dim(\Lambda^3V)$.


Question 2: Consider the map $\Lambda^2V \otimes V \to \Lambda^3V$ again, which is defined by:

$(v_1 \wedge v_2) \otimes v_3 \mapsto v_1 \wedge v_2 \wedge v_3$

It seems there are multiple ways to write down its kernel, such as:

$(v_1 \wedge v_2) \otimes v_3 + (v_1 \wedge v_3) \otimes v_2$, or,

$(v_1 \wedge v_3) \otimes v_2 + (v_2 \wedge v_3) \otimes v_1$, etc.

And there are multiple ways to construct a map $\Lambda^2V \otimes V \to \otimes^3 V$, such as:

$(v_1 \wedge v_2) \otimes v_3 \mapsto v_1\otimes v_2 \otimes v_3 - v_2 \otimes v_1 \otimes v_3$, or,

$(v_1 \wedge v_2) \otimes v_3 \mapsto v_1\otimes v_3 \otimes v_2 - v_2 \otimes v_3 \otimes v_1$, etc.

So there are actually multiple elements in $\otimes^3 V$ corresponding to $Ker(\Lambda^2V \otimes V \to \Lambda^3V )$.

I have computed some of them, and it seems they are either:

  • Young symmetrizer $c^\prime_{(21)}$ constructed from all ways of filling the Young tableaux, such as $() + (23) - (12) - (132)$ constructed from:

$2 3\\1$

  • $g \cdot c^\prime_{(21)}$ where $g \in S_3$. Example: $(23) + (123) - (132) - (13) = (23) \cdot \Big(() + (13) - (12) - (123)\Big)$. And $() + (13) - (12) - (123)$ can be constructed from:

$1 3\\2$

So my second questions is, is the above true for other partitions $\lambda$?


Question 1: The general construction of the Schur module (see for instance Fulton's book on Young tableaux, Chapter 8; the example $S_{2,1}(V)$ appears in the introduction of Part II) yields

$S_{2,1}(V) = \mathrm{coker}(\omega: \Lambda^3(V) \to \Lambda^2(V) \otimes V),$

where $\omega$ is defined by $$\omega(a \wedge b \wedge c) = (b \wedge c) \otimes a - (a \wedge c) \otimes b + (a \wedge b) \otimes c,$$ or equivalently by the more symmetric definition $$\omega(a \wedge b \wedge c) = (a \wedge b) \otimes c + (b \wedge c) \otimes a + (c \wedge a) \otimes b.$$ Notice that $\omega$ is part of the comultiplication of the graded Hopf algebra $\Lambda^*(V)$. You can also write $S_{2,1}(V)$ as a kernel, but let me work with this cokernel description.

Notice that $\omega$ has a left inverse $s : \Lambda^2(V) \otimes V \to \Lambda^3(V)$ given by mapping $(a \wedge b) \otimes c$ to $\frac{1}{3} (a \wedge b \wedge c)$. In particular, $\omega$ is injective. Hence, $$\dim(S_{2,1}(V)) = \binom{n}{2} \cdot n - \binom{n}{3} = \frac{(n+1)n(n-1)}{3}.$$

We can also write down a basis: If $e_1,\dotsc,e_m$ is a basis of $V$, then the images of the vectors $(e_i \wedge e_j) \otimes e_k$ with $i<j$ and $i \leq k$ form a basis of $S_{2,1}(V)$. For this one checks (a) that they generate $S_{2,1}(V)$, (b) that their number is $\dim(S_{2,1}(V))$.

Here is a sketch of a completely different approach: Using the decomposition $$\Lambda^n(V \oplus W) = \bigoplus_{p=0}^{n} \Lambda^p(V) \otimes \Lambda^{n-p}(W),$$ one finds a decomposition $$S_{2,1}(V \oplus W) \cong S_{2,1}(V) \oplus S_{2,1}(W) \oplus (V^{\otimes 2} \oplus W) \oplus (V \otimes W^{\otimes 2}).$$ In particular, we have $$S_{2,1}(V \oplus \mathbb{K}) \cong S_{2,1}(V) \oplus V^{\otimes 2} \oplus V.$$ From here, it is easy to determine the dimension recursively.


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.