I'm interested in seeing a few examples of root, parabolic, and Borel subgroups given a specific reductive group $G$. Here is what I know.

Let $G$ be a reductive algebraic group over an algebraically closed field $k$ of characteristic $p$. Fix a maximal torus $T\subset G$, and let $T$ act on Lie$(G)=\mathfrak{g}$ via the adjoint action. Then $\mathfrak{g}$ decomposes as a direct sum of eigenspaces corresponding to this action. The eigenvalues are the roots in the group of characters of the torus $T$. Let the roots of $G$ with respect to $T$ be denoted by $\Phi$. Choose a set of simple roots $\Pi\subset\Phi$.

For each root $\alpha$, we can uniquely define a root subgroup $U_{\alpha}$ of $G$ as the image of a certain one-parameter subgroup $\mathbb{G}_a\to G$. Furthermore, given a subset $J\subset\Pi$ of simple roots, we can define $P_J=\langle T,U_{\alpha}|\alpha\in J\rangle$. Finally, there is exactly one Borel subgroup which contains the root subgroups corresponding to all positive roots (which depend on the choice of $\Pi$).

Is all of this correct? If so, it doesn't mean much to me without seeing an example or two worked through. What are examples of these subgroups in the case that $G=GL_n$ and $T$ is the diagonal subgroup? If this is too much to type out here, I would appreciate a link to any online notes that work through this example thoroughly. It would be nice to see how $T$ determines the set of roots $\Phi$, what the root subgroups are for an arbitrary root, what the parabolic subgroups are for subsets of a choice of simple roots, and what the corresponding Borel subgroup is.


3 Answers 3


I am also a beginner in this field. The following is an example.

For $G=GL_n$, $T=\{\mathrm{diag}(c_1,\cdots c_n),c_i\in G_m\}$ the set of diagonal matrices in $G$. $T\simeq G_m^n$ is a torus of $G$.

  1. Consider the character group $X^*(T)$. It consists a natural basis $$ \epsilon_j:T\rightarrow G_m,\quad \epsilon_j(\mathrm{diag}(c_1,\cdots c_n))=c_j $$ Every characters of $T$ is a $\mathbb{Z}$-linear combinatinon of $\epsilon_j$.

  2. $Lie(G)=\mathfrak{gl}_n$, the set of $n\times n$ matices. It has a natrual basis $E_{ij}$, the matrix with $1$ at $(i,j)$ and $0$ otherwise.

  3. For $t=\mathrm{diag}(c_1,\cdots c_n)\in T$, consider the adjoint action on $E_{ij}$. \begin{eqnarray*} \mathrm{Ad}t.E_{ij}=\begin{cases}E_{ij},&\quad i=j\\ (c_ic_j^{-1})E_{ij}=(\epsilon_i-\epsilon_j)(t)E_{ij}&\quad i\neq j\end{cases} \end{eqnarray*} It shows that $E_{i,i}$ are in weiht zero space, and $E_{ij},i\neq j$ are non-zero weight space of weight $\alpha_{ij}=\epsilon_i-\epsilon_j$.

  4. Thus the root system of $G=GL_n$ is $$\Phi=\{\pm\alpha_{ij}=\pm(\epsilon_i-\epsilon_j),1\leq i< j\leq n\}.$$ One can choose a set of positive roots $$\Phi^+=\{\alpha_{ij}=(\epsilon_i-\epsilon_j),1\leq i< j\leq n\}.$$ and the set of simple roots $$\Delta=\{\alpha_{i,i+1}=(\epsilon_i-\epsilon_{i+1}),1\leq i< j\leq n\}.$$ It is easy to see that elements in $\Phi$ are $\mathrm{Z}$-linear combinition of $\Delta$, and elements in $\Phi^+$ are with non-negative coefficients.


We cosider $X_*(T)=\{\lambda^\vee:G_m\rightarrow T\}$, the set of $1$-parameters subgroups.

  1. It is easy to see that $X_*(T)$ is a free $\mathbb{Z}$-module with generators \begin{eqnarray} \epsilon_j^\vee:G_m\rightarrow T,\quad c\mapsto \mathrm{diag}(1,\cdots c,\cdots 1) \end{eqnarray}

  2. Consider the composition: \begin{eqnarray} \epsilon_j\circ\epsilon_i^\vee:G_m\rightarrow T\rightarrow G_m \end{eqnarray} It is easy to see that $\epsilon_j\circ\epsilon_i^\vee=0$ if $i=j$, and $\epsilon_j\circ\epsilon^\vee_j=1$ is the identity map on $G_m$. Thus we have a $\mathbb{Z}$-binlinear map \begin{eqnarray} X^*(T)\times X_*(T)\rightarrow \mathbb{Z},\quad (\lambda,\mu^\vee)\mapsto\lambda\circ\mu^\vee \end{eqnarray} and $\{\epsilon_i\}$ and $\{\epsilon_i^\vee\}$ are the dual $\mathrm{Z}$-basis.

  3. Given $\alpha_{ij}=\epsilon_i-\epsilon_j$. Consider $\alpha_{ij}^\vee=\epsilon_i-\epsilon_j$. It is an element in $X_*(T)$ and one has $$\alpha_{ij}\circ\alpha_{ij}^\vee= \epsilon_i\circ\epsilon_i^\vee+(-\epsilon_j)\circ(-\epsilon_j^\vee)=2.$$ Thus $\alpha_{ij}^\vee$ are the coroots of $\alpha_{ij}$. The set of corroots are $$ \Phi^\vee=\{\pm\alpha_{ij}^\vee=\pm(\epsilon_i-\epsilon_j),1\leq i<j\leq n\} $$


Now we have the root datum of $G=GL_n$. We also have chosen a set of positive roots $\Phi^+$ as above. For each $\alpha_{ij}\in\Phi^+$, as the above argument, the corresponding root space are $$\mathfrak{gl_n}_{\alpha_{ij}}=\mathbb{C}E_{ij}.$$ Roughly speaking, the torus $T$ and all positive root spaces corresponds to upper triangular matrix, which is just the Borel subgroup.


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.