The universal enveloping algebra $U(\mathfrak{g})$ of a Lie algebra $\mathfrak{g}$ over $\mathbb{C}$ is defined to be $$ \dfrac{\mathbb{C}\oplus\mathfrak{g}\oplus ( \mathfrak{g}\otimes \mathfrak{g})\oplus (\mathfrak{g}\otimes \mathfrak{g}\otimes \mathfrak{g})\oplus\ldots}{\langle a\otimes b-b\otimes a -[a,b]\rangle}, $$which means it is an associative tensor algebra generated by the vector space $\mathfrak{g}$ mod out all elements of the form $ a\otimes b-b\otimes a -[a,b]$, where $a,b\in \mathfrak{g}$.

Suppose $\mathfrak{g}=\mathbb{C}[e,p,q]/\langle [p,q]=e,[p,e]=0,[q,e]=0 \rangle$. Then we have a (ring) isomorphism
$$ U(\mathfrak{g})/\langle e-1\rangle \cong \mathbb{C}[x, \partial/\partial x] $$ where we can define the map to be $U(\mathfrak{g})\stackrel{\phi}{\rightarrow}\mathbb{C}[x,\partial/\partial x]$ by sending $$ p \mapsto \partial/\partial x \mbox{ and } q \mapsto x. $$ One can check that since $$ \phi([p,q])=[\partial/\partial x,x]= 1-x\dfrac{\partial}{\partial x}, $$ we take $e=x\dfrac{\partial}{\partial x}$, which gives us the isomorphism $U(\mathfrak{g})/\langle 1-e\rangle \cong \mathbb{C}[x,\partial/\partial x]$.

Now what should $\mathfrak{g}$ be so that $U(\mathfrak{g})/I\cong \mathbb{C}[x_1,x_2,\ldots, x_n,\partial /\partial x_1,\partial /\partial x_2, \ldots, \partial /\partial x_n]$?

I could be wrong but I'm guessing that if we take $\mathfrak{g}$ to be the algebra $\mathbb{C}[p_i,q_i,e_i]/I$ where $I$ is generated by brackets of the form $$ [p_i,q_j]= \left\{ \begin{aligned} e_i &\mbox{ if } i=j\\ -q_j p_i &\mbox{ if } i\not= j \\ \end{aligned} \right. $$

$$ [p_i,e_j]= \left\{ \begin{aligned} 0 &\mbox{ if } i=j\\ - p_i &\mbox{ if } i\not= j \\ \end{aligned} \right. $$

$$ [q_i,e_j]= \left\{ \begin{aligned} 0 &\mbox{ if } i=j\\ q_i &\mbox{ if } i\not= j, \\ \end{aligned} \right. $$ then $U(\mathfrak{g})/\langle e_i-1\rangle\cong \mathbb{C}[x_1,x_2,\ldots, x_n,\partial /\partial x_1,\partial /\partial x_2, \ldots, \partial /\partial x_n]$ where we take the map to be $$ p+\langle e_i-1\rangle \mapsto \partial/\partial x_i $$ and $$ q+\langle e_i-1\rangle \mapsto x_i. $$

Also, since the order of multiplication matters, i.e.,
$$ (\overbrace{\partial/\partial x_1 + \partial/\partial x_2}^{\deg -1?})(\overbrace{x_1}^{\deg 1}) = 1+0=\overbrace{1}^{\deg 0} $$ while $$ x_1(\partial/\partial x_1 + \partial/\partial x_2) = x_1 \partial /\partial x_1 + x_1\partial/\partial x_2, $$ how should one think about multiplication in the skew polynomial algebra?

Finally, I just noticed that $x_i\dfrac{\partial}{\partial x_i}$ form mutually orthogonal idempotent elements in $\mathbb{C}[x_i,\partial/\partial x_i]$. That is,
$$ \left(\sum_i x_i \dfrac{\partial}{\partial x_i}\right)^2 = \left(\sum_i x_i \dfrac{\partial}{\partial x_i}\right).$$ I am guessing that these are the only non-scalar idempotent elements in the algebra.

Do these idempotent elements have a geometric interpretation?

  • $\begingroup$ Please have a look here and think of changing your title accordingly. $\endgroup$ – draks ... Jul 21 '12 at 7:48
  • $\begingroup$ @draks Of course, I will edit the title shortly. $\endgroup$ – math-visitor Jul 21 '12 at 7:49

$\mathfrak{g}$ should be a Heisenberg Lie algebra. A nice coordinate-invariant way to describe these is as $V \oplus \mathbb{C}e$ where $V$ is a symplectic vector space and the symplectic form gives the Lie bracket, which takes values in $\mathbb{C}e$ (which is central). This is precisely the Lie algebra generated by the operators $x_1, x_2, ... x_n, \frac{\partial}{\partial x_1}, ... \frac{\partial}{\partial x_n}$ acting on $\mathbb{C}[x_1, ... x_n]$.

I do not believe that the elements you wrote down are idempotent.

  • $\begingroup$ Thanks Qiaochu! I thought the above elements are idempotents since $\left( x_i\frac{\partial}{\partial x_i}\right)\left( x_i\frac{\partial}{\partial x_i} \right) =x_i\frac{\partial}{\partial x_i}$. $\endgroup$ – math-visitor Jul 21 '12 at 4:18
  • 3
    $\begingroup$ @math-visitor, that equaity you wrote in that comment is actually false. There are no idempotents in the Weyl algebra except zero and one; in fact, it is a domain! $\endgroup$ – Mariano Suárez-Álvarez Jul 21 '12 at 4:42
  • $\begingroup$ @MarianoSuárez-Alvarez Thank you for the comment. I am reading the links that Qiaochu posted, and I just got to the link about Weyl algebras. =) $\endgroup$ – math-visitor Jul 21 '12 at 4:44
  • 1
    $\begingroup$ @math-visitor: no. $x_i$ is the operator describing multiplication by $x_i$, so in fact $\frac{\partial}{\partial x_i} x_i = x_i \frac{\partial}{\partial x_i} + 1$ by the product rule. $\endgroup$ – Qiaochu Yuan Jul 21 '12 at 7:56
  • $\begingroup$ @QiaochuYuan Thanks very much for the comment! $\endgroup$ – math-visitor Jul 21 '12 at 7:58

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.