3
$\begingroup$

I've recently met with the Temperley-Lieb algebra in my work. I'm in no way a specialist, and it's seems like a pretty simple question, but nevertheless. I'm interested in the subalgebra generated by two neighbour generator $ U_i, \, U_{i+1}$. This subalgebra is 5 dimensional with a basis: $$ \mathbb{1}, \, U_i, \, U_{i+1}, \, U_i U_{i+1}, \, U_{i+1} U_i. $$ It seems very simple, and it'll be very helpful to me if the subalgebra is isomorphic to a well-known matrix algebra, for example. I've looked throught references, but mostly they talk about representation-theoretical questions, and simple algebraic ones like mine are not discussed.

Therefore, my question is the following. Is the Temperley-Lieb subalgebra generated by two neighbor generators isomorphic to some well-known algebra? Or is there any way I can check it myself?

$\endgroup$
2
  • 1
    $\begingroup$ In the linked Wikipedia page, there's a parameter $\delta$. What is it here? It seems the subalgebra is $TL_3(\delta)$ isn't it. It"s described nowhere? To be complete you might have considered writing down the missing relations: $U_i^2=\delta U_i$, $U_iU_{i+1}U_i=U_i$, $U_{i+1}U_iU_{i+1}=U_{i+1}$ (and $U_iU_j=U_jU_i$ for $|i-j|\ge 2$, not relevant here). $\endgroup$
    – YCor
    Mar 30, 2021 at 13:01
  • $\begingroup$ I quite quickly computed that the underlying Lie algebra is solvable iff $\delta^4=1$ (this should be double checked). The information that the underlying Lie algebra or not is solvable is relevant since it gives significant restrictions on the Wedderburn-Malcev decomposition. $\endgroup$
    – YCor
    Mar 30, 2021 at 16:54

1 Answer 1

6
$\begingroup$

As YCor mentions in the comments, this is (isomorphic to) the algebra $TL_3$, so I will call its generators $U_1,U_2$. I will henceforth write $R := TL_3$. I do not know what your ground ring is or what your parameter $\delta$ is. If $1-\delta^2$ is invertible, then $R$ contains a special idempotent $p$ named after Jones and Wenzl. The Jones–Wenzl idempotent is very special: $U_1 p = p U_1 = U_2 p = p U_2 = 0$. In particular it is central. It follows that your ring splits as $pR \times (1-p)R$. A formula is $$ p = 1 - \frac{\delta}{\delta^2 - 1} ( U_1 + U_2) + \frac{1}{\delta^2-1} (U_1 U_2 + U_2 U_1). $$

Since $U_ip = 0$, $pR$ is definitely 1-dimensional, and so $R' := (1-p)R$ is a noncommutative 4-dimensional ring. I claim it is a matrix ring. Note that $R'$ is spanned for example by the elements $1-p$, $U_1$, $U_2$, and $U_1U_2$. Consider the left $R'$-ideal $R'U_1 \subset R'$. It is 2-dimensional: the basis vectors map to $U_1, \delta U_1, U_2 U_1$, and $U_1$ respectively, and so we can take the basis $\{U_1, U_2 U_1\}$ of this ideal. In this basis, the action of $R'$ on the ideal is $$ 1-p \mapsto \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}, \quad U_1 \mapsto \begin{pmatrix} \delta & 1 \\ 0 & 0 \end{pmatrix}, \quad U_2 \mapsto \begin{pmatrix} 0 & 0 \\ 1 & \delta \end{pmatrix}, \quad U_1 U_2 \mapsto \begin{pmatrix} 1 & \delta \\ 0 & 0 \end{pmatrix}. $$ Since I insisted that $1-\delta^2$ should be invertible, these four matrices are a basis for the $2\times 2$ matrix ring, proving the claim.

$\endgroup$
2
  • $\begingroup$ I should emphasize that the case $\delta = \pm 1$ is also very interesting (and YCor's calculation in the comments makes me worry that I have a sign error somewhere). Note that $U_i \mapsto -U_i$ switches the sign of $\delta$, and so we might as well assume $\delta = 1$. Then the element that would be called "$(1-\delta^2)p$", i.e. $x := (U_1 + U_2) - (U_1 U_2 + U_2 U_1)$, solves $xU_1 = U_1 x = xU_2 = U_2 x = 0$. In particular, it is a central element solving $x^2 = 0$. $\endgroup$ Mar 31, 2021 at 13:52
  • $\begingroup$ Write "$k$" for whatever ground ring you are working over (perhaps $k = \mathbb{C}$). Then $R = TL_3(\delta=1)$ is an algebra over $k[x]/(x^2)$. The quotient $R' = R / (x)$ is 4-dimensional. Again we can look at the 2-dimensional ideal $R'U_1 \subset R'$, but now the action map $R' \to \mathrm{Mat}(2)$ is not injective. Rather, its kernel is spanned by $y := U_1 - U_1U_2 = U_2U_1 - U_2$. In other words, I am saying that $R'$ contains a 1-dimensional 2-sided ideal spanned by $y$. In particular, it is not a matrix ring. So not only $R$ but also $R'$ is not a sum of matrix rings. $\endgroup$ Mar 31, 2021 at 14:08

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.