Let $V$ be a vector space and $W_1 \subsetneq W_2 \subsetneq ... \subsetneq W_\ell \subsetneq V $ a flag of subspaces.

The associated Schubert variety is defined as : $ \Omega ( W_{ \bullet } ) = \{ \ W \in Gr_k ( V ) \ | \ \mathrm{dim} ( W \bigcap W_i ) \geq i \ , \ i = 1 , \dots , \ell \ \} $.

My question is to know what is the idea behind the fact to define this kind of varieties ?

Why did we choose to define a Shubert variety with this condition that : $ \ \mathrm{dim} ( W \bigcap W_i ) \geq i \ , \ i = 1 , \dots , \ell \ $

Thanks in advance for your help.


I get $G=\mathrm{GL}(n,\mathbb{C})$, let $B$ be a Borel subgroup of $G$, let $P$ be a parabolic subgroup of $G$ and let $T$ be a maximal torus in $G$; we know that:

  • up to conjugation: $T$ is the group of diagonal matrices, $B$ is the group of upper triangular matrices; in particular: $B$ is $T\ltimes U$, the semidirect product of $T$ with the group $U$ of unipotent upper triangular matrices;

  • $G_{\displaystyle/P}$ is the variety $\mathcal{F}$ of flags of type $(1\leq m_1<\dots<m_r\leq n)$ of $\mathbb{V}$ (a complex vector space of dimension $n$); in particular $G_{\displaystyle/B}$ is the variety $\mathcal{B}$ of complete flags of $\mathbb{V}$;

  • without loss of generality, $T$ is contained in $P$;

  • the Weyl group $W$ of $G$ is defined as $N_G(T)_{\displaystyle/C_G(T)}=N_G(T)_{\displaystyle/T}$ (the normalizer of $T$ in $G$ quotiented by the centralizer of $T$ in $G$), because $T$ is a maximal torus then $T=C_G(T)$ (ever);

and in particular, the Weyl group of $\mathrm{GL}(n,\mathbb{C})$ is the $n$-th symmetryc group $S_n$.

One can consider the action: \begin{gather} \alpha:(M,F_{\bullet})\in G\times\mathcal{F}\to M\cdot F_{\bullet}\in\mathcal{F},\\ F_{\bullet}\equiv\{\underline{0}\}\subsetneqq\left\langle e_1,\dots, e_{m_1}\right\rangle\subsetneqq\left\langle e_1,\dots, e_{m_2}\right\rangle\subsetneqq\dots\subsetneqq\mathbb{V},\\ M\cdot F_{\bullet}\equiv\{\underline{0}\}\subsetneqq\left\langle M\cdot e_1,\dots,M\cdot e_{m_1}\right\rangle\subsetneqq\left\langle M\cdot e_1,\dots,M\cdot e_{m_2}\right\rangle\subsetneqq\dots\subsetneqq\mathbb{V} \end{gather} where $\{e_1,\dots,e_n\}$ is a basis of $\mathbb{V}$.

One can prove that: \begin{equation} \mathrm{Fix}^{\alpha}_T(\mathcal{B})=\{E_w\equiv wB\in\mathcal{B}\mid w\in W\}, \end{equation} where $E_{1_W}$ is the standard complete flag \begin{equation} \{\underline{0}\}\subsetneqq\langle e_1\rangle\subsetneqq\langle e_1,e_2\rangle\subsetneqq\dots\subsetneqq\mathbb{V}; \end{equation} for example, see theorem 10.2.7 [BL].

Considered the set: \begin{gather} \mathrm{Orb}^{\alpha}_B(E_w)=\{b\cdot E_w\in\mathcal{F}\mid b\in B\}=\{u\cdot E_w\in\mathcal{F}\mid u\in U\}=\mathrm{Orb}^{\alpha}_U(E_w), \end{gather} it is called Schubert cell $C_w$ of $\mathcal{F}$; one defines Schubert variety $X_w$ the Zariski closure of $C_w$ in $\mathcal{F}$ (see definitions 1.1.2 and 1.2.2 from [B]).


  1. Any Schubert cell and variety is parametrized by an element of $W$!
  2. The Grassmannians are particular flag varieties!
  3. Any complex Lie group is embeddable in some $\mathrm{GL}(N,\mathbb{C})$, therefore $\alpha$ makes sense for any reductive, complex Lie group; and one can define again the flag varieties, the Grassmannians and the Schubert cells and varieties.

Choice $1\leq k<n$, let $E_{w_k}$ be the $k$-dimensional space in the flag $E_w$; by the previous reasoning, $E_{w_k}$ is a $T$-fixed point in $G(k,n)$ (the Grassmannian of the $k$-plane of $\mathbb{V}$); in particular, there exists $w\in W$ such that \begin{equation} E_{w_k}=\left\langle e_{w(1)},\dots,e_{w(k)}\right\rangle \end{equation} which corresponds to $\left[e_{w(1)}\wedge\dots\wedge e_{w(k)}\right]=\left[e_{\underline w}\right]\in G(k,n)\subseteq\mathbb{P}\left(\bigwedge^k\mathbb{V}\right)$.

Let \begin{equation} A_{w_k}=\left(a_j^i\right)\in\mathbb{C}_n^k\mid e_{w(i)}=\sum_{j=1}^na_i^je_j\equiv a_i^je_j, \end{equation} that is \begin{equation} a_j^i=\begin{cases} 1\iff j=w(i)\\ 0\iff \text{otherwise} \end{cases}; \end{equation} because $C_w$ in $G(k,n)$ is the $U$-orbit of $E_{w_k}$, then $C_w$ is in bijection with set \begin{equation} \left\{uA_{w_k}\in\mathbb{C}_n^k\mid u\in U\right\}=\left\{M=\left(m_i^j\right)\in\mathbb{C}_n^k\mid\begin{cases} m_{w(i)}^j=\delta_i^j\\ m_i^j=0\iff i>w(j) \end{cases}\right\}, \end{equation} that is $C_w$ is the set \begin{equation} \left\{\left[m_i^1e_{w(1)}\wedge\dots\wedge m_i^ke_{w(k)}\right]\in G(k,n)\mid\begin{cases} m_{w(i)}^j=\delta_i^j\\ m_i^j=0\iff i>w(j) \end{cases}\right\}; \end{equation} in particular, $C_w$ is an affine agebraic variety in $G(k,n)$.

For these and other details, see lemmata 1.4.3 and 1.4.4 from [L].

Defined \begin{gather} I_{k,n}=\left\{\underline{i}=(i_1,\dots,i_k)\in\{1,\dots,n\}^k\mid i_1<\dots<i_k\right\}\\ \underline{i},\underline{j}\in I_{k,n},\,\underline{i}\leq\underline{j}\iff\forall h\in\{1,\dots,k\},i_h\leq j_h; \end{gather} by the previous reasoning and definition \begin{equation} X_w=\left\{[v_1\wedge\dots\wedge v_k]\in G(k,n)\mid\forall\underline{i}\not\leq\underline{w}\in I_{k,n},\,p_{\underline{i}}([v_1\wedge\dots\wedge v_k])=0\right\} \end{equation} where the support of $\underline{w}$ is $\{w(1),\dots,w(k)\}$; in particular, $X_w$ is a projetive algebraic variety in $G(k,n)$.

As usual, $p_{\underline{i}}$ is the $\underline{i}$-th Plücker projection, that is the minor $\det\left(a_h^{i_h}\right)_{h\in\{1,\dots,k\}}=d_{\underline i}$, where $v_i=a_i^je_j$ and $\underline{i}=(i_1,\dots,i_k)$.

Now, how can we interpret $X_w$?

First of all, let $\underline{i}\in I_{k,n}$, we can define \begin{gather} w\in S_n\mid\forall h^{\prime}\in\{1,\dots,k\},\,w(h^{\prime})=i_{h^{\prime}},\\ \forall h^{\prime\prime}\in\{1,\dots,n-k\},\,w(h^{\prime\prime})\,\text{is the}\,h^{\prime\prime}\text{-th element of the ordered set}\,\{1,\dots,n\}\setminus\{i_1,\dots,i_k\}, \end{gather} in this way there exists a bijection between $S_n$ and $I_{k,n}$. Considering the flag \begin{equation} F_{\underline{i}}\equiv\{\underline0\}<\left\langle e_1,\dots,e_{i_1}\right\rangle=F_{i_1}<\left\langle e_1,\dots,e_{i_2}\right\rangle=F_{i_2}<\dots<\left\langle e_1,\dots,e_{i_k}\right\rangle=F_{i_k}<\mathbb{V}; \end{equation} let $[W]\in X=X_{\underline i}$, by construction: \begin{equation} [W]=[v_1\wedge\dots\wedge v_k]=\left[\sum_{\underline{j}\leq\underline{i}}d_{\underline j}e_{\underline j}\right] \end{equation} in particular: \begin{equation} \forall h\in\{1,\dots,k\},\,\dim\left(W\cap F_{i_h}\right)\geq h \end{equation} and vice versa; for other details, see lemma 1.4.5 from [L].

For other interpretation of the Schubert varieties, one can see the chapter 1, sections 1 and 2 from [B].


[B] Brion M. - Lectures on the geometry of flag varieties; available on arxiv.org

[BL] Brown J., Lakshimibai V. - Flag Varieties, Northeastern University

[L] Littelmann P. - Schubert varieties; available on personal web page


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.