3

Here is the question I am trying to solve:

For any Lie algebra $L$ determine the group of grouplike elements of the Hopf algebra $U(L).$

Some definitions:

1-To any Lie Algebra $L$ we assign an (associative) algebra $U(L),$ called the enveloping algebra of $L,$ and a morphism of Lie algebras $i_L: L \to L(U(L)).$ We define the enveloping algebra as follows. Let $I(L)$ be the two-sided ideal of the tensor algebra $T(L)$ generated by all elements of the form $xy - yx - [x,y]$ where $x,y$ are elements of $L.$ We define $$U(L) = T(L)/ I(L)$$

Where $U(L)$ is filtered as a quotient algebra of $T(L).$

2- Let $V$ be a vector space. Define $T^0 (V) = k, T^1(V) = V$ and $T^n(V) = V ^{\otimes n}$ (the tensor product of $n$ copies of $V$) if $n > 1.$ The canonical isomorphisms $$T^n (V) \otimes T^m(V) \cong T^{n + m}(V)$$ induce an associative product on the vector space $T(V) = \otimes_{n \geq 0} T^n (V).$ Equipped with this algebra structure, $T(V)$ is called the tensor algebra of $V.$ The product in $T(V)$ is explicitly defined by $$(x_1 \otimes \dots \otimes x_n)(x_{n + 1} \otimes \dots \otimes x_{n+m}) = x_1 \otimes \dots \otimes x_n \otimes x_{n + 1} \otimes \dots \otimes x_{n+m} \quad \quad \quad (5.1)$$

where $x_1, \dots , x_{n+1}, \dots , x_{n+m}$ are elements of $V.$ The unit for this product is the image of the unit element $1$ in $k = T^0(V).$ Let $i_V$ be the canonical embedding of $V = T^1 (V)$ into $T(V).$ By $(5.1)$ we have $$x_1 \otimes \dots \otimes x_n = i_V(x_1) \dots i_V(x_n)$$ which allows us to set $x_1 \dots x_n = x_1 \otimes \dots \otimes x_n$ whenever $x_1, \dots , x_n$ are elements of $V.$

3- A grouplike element of a coalgebra $(H, \Delta, \epsilon),$ i.e., an element $x \neq 0$ such that $$\Delta (x) = x \otimes x.$$

4- I also know that the set of grouplike elements of $H$ is a group.

Still I do not know how to answer it. Could someone show me some details please?

EDIT:

5- I also know the universal property of $U(L).$

EDIT-2:

Here is a theorem I have also but I do not know if it may help in my case or not:

enter image description here

Emptymind
  • 2,217
  • 2
    You spend a lot of time describing the algebra structure of the tensor algebra, which is not really necessary as it is very well-known, but what is actually relevant to your question is rather its coalgebra structure. Do you understand what it is? – Captain Lama Apr 13 '23 at 17:20
  • @CaptainLama you mean that $U(L)$ is a coalgebra structure? – Emptymind Apr 13 '23 at 17:47
  • @CaptainLama I am still lost in the solution that is why I defined unnecessary things :) – Emptymind Apr 13 '23 at 17:48
  • 2
    You say yourself that $U(L)$ is a Hopf algebra, so it is both an algebra and a coalgebra. But to define to group-like elements, the algebra structure is irrelevant, only the coalgebra structure matters. – Captain Lama Apr 13 '23 at 18:07
  • Yeah I see your point ..... but what exactly in the properties of the coalgebra structure of $U(L)$ will help me in calculating the grouplike elements? – Emptymind Apr 13 '23 at 18:11
  • 1
    Well first you should try to understand what this structure is (as in, what is the formula for the comultiplication). You can look at https://en.wikipedia.org/wiki/Tensor_algebra#Coalgebra to understand what the coalgebra structure looks like on the tensor algebra (it is the first one in the article, the one defining a Hopf algebra structure). – Captain Lama Apr 13 '23 at 18:14
  • Will we use the fact that it has an antipode? should I find its antipode? @CaptainLama – Emptymind Apr 13 '23 at 18:22
  • @CaptainLama what is the relation between the antipodes of the tensor algebra and the grouplike elements? – Emptymind Apr 13 '23 at 18:42
  • @CaptainLama I added a formula for comultiplication that I know – Emptymind Apr 20 '23 at 19:41
  • How do you know the set of grouplike elements forms a group? It seems you require $\epsilon(x)=1$ to be grouplike. – luxerhia Apr 23 '23 at 08:10
  • Moreover, is there any condition on the characteristic of $L$? – luxerhia Apr 23 '23 at 08:11
  • @luxerhia Nope there is not – Emptymind Apr 23 '23 at 23:54
  • @luxerhia for your comment before last ..... I have a concrete proof for this. – Emptymind Apr 24 '23 at 15:13

2 Answers2

3

Grouplike elements in a Hopf algebra are invertible, and an immediate consequence of the PBW theorem is that the only invertible elements in an enveloping algebras are the scalar multiples of the identity.

It follows that a grouplike element is of the form $x1_H$ for some scalar $x$, and from there it is easy to check that $x$ has to be $1$.

luxerhia
  • 3,657
2

Let us recall the PBW theorem. Let $\pi : T(L) \rightarrow U(L)=T(L)/I(L) $ be the canonical projection. Write $T_m = T^0(L) \oplus T^1(L) \oplus \cdots T^m(L)$ and $U_m = \pi(T_m)$. Set $U_{-1} = 0$ and $G^m = U_m/U_{m-1}$. The PBW theorem states that the graded unital associative algebra $G = \bigoplus_{m=0}^{\infty} G^m$ is isomorphic (as graded algebras) to the symmetric algebra $S(L) = \bigoplus_{m=0}^{\infty}S^m(L)$.

Let $x \in U(L)$ be a nonzero element and $x_0$ be the homogeneous component of $x$ with highest degree. In other words, $x = x_0$ mod $U_{m-1}$ where $x_0 \neq 0$ and $\deg(x_0) = m$. We will show $x^2$ has a nonzero homogeneous part of degree $2m$. According to the PBW theorem, we can consider the isomorphism of graded algebras $\psi: G= \bigoplus_{i=0}^{\infty} G^i \rightarrow S(L) = \bigoplus_{i=0}^{\infty} S^i(L)$. Here $0 \neq \overline{x} \in G^m=U_m/U_{m-1}$ is mapped to $0 \neq \psi(\overline{x}) \in S^m(L)$. Since the symmetric algebra $S(L)$ is an integral domain, $\psi(\overline{x})^2 = \psi(\overline{x^2}) \in S^{2m}(L)$ is also nonzero. Thus $\overline{x^2} \in G^{2m}$ cannot be zero.

Denote the multiplication map of $U(L)$ by $m: U(L) \otimes U(L) \rightarrow U(L)$. For every $x \in U_m$, we know $m\Delta(x)$ lies in $U_m$. This can be proved directly, but you may use the proposition in your question. However, if $x \in U_m \setminus U_{m-1}$ is a grouplike element, then $m\Delta(x)= m(x \otimes x) = x^2$ has a nonzero homogeneous part of degree $2m$. If follows that $m \geq 2m$, or $x$ is a scalar. Now $\Delta(x) = x \Delta(1)=x (1 \otimes 1)$ should be equal to $x\otimes x = x^2 (1 \otimes 1)$, whence $x = 1$.


(Added)

If $x \neq 0 $ satisfies $\Delta(x) = x \otimes x$, then $x = \mbox{id}(x) = (\epsilon \otimes \mbox{id})(\Delta x) = \epsilon(x)x$ and thus $\epsilon(x) = 1$. By the axiom of antipode $1 = xS(x) = S(x)x$, whence $x$ is invertible. From here, Mariano's answer proves the statement.

luxerhia
  • 3,657