/
cohomology_sl.tex
208 lines (184 loc) · 11 KB
/
cohomology_sl.tex
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
\subsection[SU(p,q)]{$\mathrm{SU}(p,q), p+q=n \sim A_{n-1}, n \geq 2$}\label{sec:su}
\subsubsection{Root system data}
\[\alpha_i = \epsilon_i - \epsilon_{i+1}, \quad \omega_i = \epsilon_1 + \cdots + \epsilon_i \]
\begin{align*}
\roots & = \{ \epsilon_i - \epsilon_j | i\neq j, i,j=1\ldots n\}\\
\roots_c^+ & = \{ \epsilon_i-\epsilon_j | 1\leq i < j \leq p \text{ or } p+1 \leq i < j \leq n\}\\
\roots_n^+ & = \{ \epsilon_i - \epsilon_j | 1 \leq i \leq p, p+1 \leq j \leq n \}
\end{align*}
\[\beta = \epsilon_1 - \epsilon_n,\quad 2\rho = (n-1,n-3,\ldots -n+3,-n+1),\quad \zeta = (\underbrace{\frac{q}{n},\ldots,\frac{q}{n}}_{p\text{ times}},\underbrace{\frac{-p}{n},\ldots,\frac{-p}{n}}_{q\text{ times}})\]
\inserttikzfigure{diagrams/dynkin_An_p.tikz}{Marked Dynkin diagram of $\lie{su}(p,q)$ for $p+q=n$}
The reduction points of unitarizable highest weight modules are the following integral translated cones $\lambda_a + C_a$:
Let $a=(Q,R,l)$, $Q=R=\mathrm{SU}(p',q')$, $1\leq p' \leq p$, $1\leq q'\leq q$. Then
\[
C_a = \{a_{p'}\omega_{p'} + \cdots + a_p\omega_p + \cdots + a_{n-q'}\omega_{n-q'} \,|\, a_p=-a_{p'}-\cdots -a_{p-1}-a_{p+1} - \cdots - a_{n-q'} \}.
\]
\begin{gather*}
\lambda_a=\omega_{p'} + \omega_{n-q'} - (n+l+1-p'-q')\omega_p \\
\mu_a = \omega_{p'-l}+\omega_{n-q'+l}-(n+l+1-p'-q')\omega_p\\
1\leq p' \leq p,\quad 1\leq q' \leq q,\quad 1\leq l \leq \min(p',q')\\
Q(\lambda_a)=R(\lambda_a)=\mathrm{SU}(p',q')
\end{gather*}
\subsubsection{Nilpotent cohomology in detail}
Now we compute the cohomology for $\lambda = \omega_{p'} + \omega_{n-q'} - (n+l+1-p'-q')\omega_p$. We have for $k=2-(n+l+1-p'-q') = 1+p'+q'-n-l$ that
\[
(\epsilon_i, \lambda) = \begin{cases}
k, &1\leq i \leq p' \\
k-1, & p' < i \leq p\\
1, & p < i \leq n-q'\\
0, & n-q' < i \leq n.
\end{cases}
\]
The positive roots of $\lie{su}(p,q)$ are $\epsilon_i - \epsilon_j$, $i<j$ and we have
\[(\epsilon_i - \epsilon_j, \rho) = \frac{n+1-2i}{2}-\frac{n+1-2j}{2} = j-i > 0.\] Now, in order to determine the set $\Psi^+_\lambda$, we compute all possible values of $(\epsilon_i - \epsilon_j, \lambda +\rho)$ for $i<j$
\begin{center}
\begin{tabular}{C|CCCC}
& 1\leq j \leq p' & p' < j \leq p & p < j \leq n-q' & n-q' < j \leq n \\[2pt]\hline
1\leq i \leq p' & j-i & 1 + j-i & k-1 + j-i & k + j-i \\
p' < i \leq p & & j-i & k-2 + j-i & k-1 + j-i \\
p < i \leq n-q' & & & j-i & 1 + j-i \\
n-q' < i \leq n & & & & j-i \\
\end{tabular}
\end{center}
We see that only terms containing $k$ can be zero, because $j-i >0$ and $2-n \leq k \leq 0$. Thus the only singular roots are the noncompact ones. Substituting for $k$ and solving for $j$ we get
% \begin{center}
% \begin{tabular}{C|CC}
% & p < j \leq n-q' & n-q' < j \leq n \\[2pt]\hline
% 1\leq i \leq p' & p'+q'-n-l + j-i & 1+p'+q'-n-l + j-i \\
% p' < i \leq p & -1+p'+q'-n-l + j-i & p'+q'-n-l + j-i \\
% \end{tabular}
% \end{center}
% and solving for zero
\begin{center}
\begin{tabular}{C|CC}
& p < j \leq n-q' & n-q' < j \leq n \\[2pt]\hline
1\leq i \leq p' & j=i+m & j=i+m-1 \\
p' < i \leq p & j=i+m+1 & j=i+m \\
\end{tabular}
\end{center}
where \[m=1-k = n+l-p'-q'.\]
The case $j=i+m+1$ doesn't lead to any solution, since $j-i-m-1 = j-i-n-l+p'+q'-1$ is strictly negative for $ p' < i \leq p < j \leq n-q' $.
%for $i=p'+1$ we get $j=2+l+n-q'>n-q'$.
The case $j=i+m$ for $i \leq p'$ leads to
\begin{gather*}
\max\{1,q'-q+1+p'-l\}\leq i \leq p'-l\\
\max\{1+n+l-p'-q', p+1 \} \leq j \leq n-q',
\end{gather*}
which results in an empty set of singular roots if and only if $l=p'$.
The case of $j=i+m$ for $i>p'$ yields
\begin{gather*}
p'+1 \leq i \leq \min\{p, p'+q'-l \}\\
n-q'+l+1 \leq j \leq \min\{n +l-p'-q'+p,n \},
\end{gather*}
which gives an empty set of singular roots if and only if $l=q'$ or $p=p'$.
And finally the case $j=i+m-1$ gives
\begin{gather*}
p'+2-l \leq i \leq p'\\
n-q'+1 \leq j \leq n-q'-1+l
\end{gather*}
which doesn't contribute to singular roots if and only if $l=1$.
Two roots $\epsilon_i - \epsilon_j$, $\epsilon_a - \epsilon_b$ are orthogonal if and only if $\{i,j\} \cap \{a,b\}=\emptyset$. Thus the set of positive noncompact roots orthogonal to $\Psi_\lambda^+$ is
\begin{multline}\label{eq:orthogonal_to_singular}
\big\{\epsilon_i - \epsilon_j\,|\,
i \in \{ 1,\ldots, q'-q+p'-l\}\cup \{p'-l+1\}\cup \{ p'+q'-l+1,\ldots, p \} \\
j \in \{p+1,\ldots, m\} \cup \{n+l-q'\} \cup \{m+p+1, \ldots,n\} \big\},
% j \in \{p+1,\ldots, n+l-p'-q'-1\} \cup \{n+l-q'\} \cup \{n+l-p'-q'+p+1, \ldots,n\} \big\}.
\end{multline}
where a set $\{a,\ldots,b\}$ is considered empty if $a>b$.
Positive noncompact roots $\alpha = \epsilon_i - \epsilon_j$ satisfying $(\lambda + \rho, \alpha)\in \Z^+$ are given by constraints
\begin{equation}\label{eq:su_root_constraints}
\begin{split}
1 \leq i \leq p' & \Longrightarrow \max \{ i+m, p \} < j \leq n-q' \vee \max \{ i+m, n-q'+1 \} \leq j \leq n \\
p' < i \leq p & \Longrightarrow \max \{p, i+m+1 \} < j \leq n-q' \vee \max \{n-q',i+m\} < j \leq n.
\end{split}
\end{equation}
If a positive noncompact root $\epsilon_i - \epsilon_j$ with $i>p'$ is orthogonal to all singular roots and it's scalar product with $\lambda + \rho$ is positive, then necessarily
\[
i\in \{ p'+q'-l+1,\ldots, p \} \text{ and } i+m < j\leq n.
\]
But for $i=p'+q'-l+1$ we get that $i+m = n+1$ and hence we see that there are no such roots. Thus, we have to look only at the noncompact roots satisfying the first constraint of \eqref{eq:su_root_constraints}.
If a noncompact root $\epsilon_i - \epsilon_j$ with $i\leq p'$ is orthogonal to the singular roots $\Psi_\lambda^+$, then either $i=p'-l+1$ or $i \leq q'-q + p'-l$. If $i=p'-l+1$, then we get $i+m = n-q'+1$ which means that we have to look only at
\[
\max \{ i+m, n-q'+1 \} = n-q'+1 \leq j \leq n.
\]
Taking into account \eqref{eq:orthogonal_to_singular} we see that the roots with indices given by
\[
i = p'-l+1, \quad j \in \{n-q'+l\} \cup \{m+p+1, \ldots,n\}
\]
are included in $\roots_{n,\lambda}^+$. For $i\leq q'-q + p'-l$ we have $i+m \leq p$ and the constraints of \eqref{eq:su_root_constraints} reduce to
\[
p < j \leq n
\]
and thus the remaining roots of $\roots_{n,\lambda}^+$ have indices given by
\[
i \in \{ 1,\ldots, q'-q+p'-l\}, \quad
j \in \{p+1,\ldots, m\} \cup \{n+l-q'\} \cup \{m+p+1, \ldots,n\}.
\]
Because all roots have the same length $\sqrt{2}$ we finally arrive at $\roots_{n,\lambda}^+ $ being equal to the set of roots $\epsilon_i - \epsilon_j$ with indices given by the following constraints
\begin{gather}
i = p'-l+1 \quad \&\quad j \in \{n-q'+l\} \cup \{m+p+1, \ldots,n\} \\ \notag
i \in \{ 1,\ldots, q'-q+p'-l\} \quad \& \hspace{8cm} \\ \notag \hspace{4.5cm} j \in \{p+1,\ldots, m\} \cup \{n+l-q'\} \cup \{m+p+1, \ldots,n\}.\notag
\end{gather}
The Weyl group of $\lie{sl}(n)$ is generated by root reflections and is isomorphic to the symmetric group $S_n$ on the set $\{1,\ldots,n\}$. Indeed, the reflection $s_{\epsilon_i - \epsilon_j}$ acts as a transposition of the $i$th and $j$th coordinate in $\epsilon$-basis. Since there are overlaps in the ranges for $j$ for various $i$, it follows that the group $W_\lambda$ is isomorphic to the permutation subgroup of $S_n$ which permutes only a subset $M_\lambda$ of $\{1,\ldots,n\}$ and leaves all other elements fixed. This subset is
\[
M_\lambda = \{ 1,\ldots, q'-q+p'-l, p'-l+1, p+1,\ldots,m, n-q'+l,m+p+1,\ldots,n\}
\]
in the case of $q'-q+p'-l > 0$ and
\[
M_\lambda = \{p'-l+1,n-q'+l,m+p+1,\ldots,n\}
\]
if $q'-q+p'-l \leq 0.$ We remind, that a range $a,\ldots,b$ is considered empty if $a>b$. The root subsystem $\roots_\lambda$ is then given by
\[
\roots_\lambda = \{ \epsilon_i - \epsilon_j \,|\, i\neq j, i,j\in M_\lambda \}.
\]
The pair $(\roots_\lambda,\roots_{\lambda,c})$ is a pair of root systems such that the corresponding Lie algebras $(\lie{g}_\lambda,\lie{g}_{\lambda,c})$ form a Hermitian symmetric pair of type $\lie{su}(a,b)$ for some $a,b$. The formula of the theorem \ref{thm:cohomology} is basically coming (via Enright Shelton equivalences) from the classical Kostant formula for Lie algebra cohomology of the nilradical coming from $\lie{g}_\lambda$. Hence the cohomology groups can be computed in a classical way just by restriction to the indices contained in $M_\lambda$.
Let us look at some specific cases. For $p=1$ the only possible value for $l$ is $1$ and $1\leq q' \leq n-1$. The weights are $\lambda = (q'-n)\omega_1 + \omega_{n-q'}$ and the corresponding set $M_\lambda = \{ 1,n-q'+1,\ldots,n \}$.
Now let's look what happens when $p=2$. There are three possibilities
\begin{enumerate}
\item $p'=1, l=1$\\
The weight is $\lambda=\omega_1+\omega_{n-q'}-(n+1-q')\omega_2$ and the resulting set of admissible indices $M_\lambda = \{1,n-q'+1,\ldots,n\}$.
\item $p'=2, l=1$\\
The weight is $\lambda=\omega_2+\omega_{n-q'}-(n-q')\omega_2 = \omega_{n-q'} + (q'-n+1)\omega_2$ and the resulting set of admissible indices $M_\lambda = \{1,\ldots,n\}$ if $q'=q$ (in which case $\lambda = 0$) and $M_\lambda = \{2,n-q'+1,\ldots,n\}$ for $q'<q$.
\item $p'=2, l=2$\\
The weight is $\lambda=\omega_2+\omega_{n-q'}-(n+1-q')\omega_2 = \omega_{n-q'}+(q'-n)\omega_2$ and the resulting set of admissible indices $M_\lambda = \{1,n-q'+2,\ldots,n\}$.
\end{enumerate}
The most singular case occurs for $\lie{su}(k,k)$ and $p'=q'=l = k $. Then the set $M_\lambda =$ is $\{1,2k\}$. For $p'=q'=k$ and $l<k$ we get $M_\lambda = \{ 1,\ldots,k-l+1,k+l,\ldots, 2k \}$.
\subsubsection*{Examples of posets of minimal length representatives}
\begin{figure}[H]
\centering
\input{diagrams/nroots_A5_1.tikz}
\input{diagrams/bruhat_A5_1.tikz}
\caption{Poset of noncompact roots and the Bruhat graph for $\mathrm{SU}(1,5)$}
\end{figure}
\begin{figure}[H]
\centering
\input{diagrams/nroots_A5_2.tikz}
\input{diagrams/bruhat_A5_2.tikz}
\caption{Poset of noncompact roots and the Bruhat graph for $\mathrm{SU}(2,4)$}
\end{figure}
\begin{figure}[H]
\centering
\resizebox{\textwidth}{!}{
\input{diagrams/nroots_A5_3.tikz}
\input{diagrams/bruhat_A5_3.tikz}
}
\caption{Poset of noncompact roots and the Bruhat graph for $\mathrm{SU}(3,3)$}
\end{figure}
\begin{figure}[H]
\centering
\input{diagrams/nroots_A5_4.tikz}
\input{diagrams/bruhat_A5_4.tikz}
\caption{Poset of noncompact roots and the Bruhat graph for $\mathrm{SU}(4,2)$}
\end{figure}
\begin{figure}[H]
\centering
\input{diagrams/nroots_A5_5.tikz}
\input{diagrams/bruhat_A5_5.tikz}
\caption{Poset of noncompact roots and the Bruhat graph for $\mathrm{SU}(5,1)$}
\end{figure}
\begin{figure}[H]
\centering
\input{diagrams/a7_4}
\caption{The BGG graph of type $(A_7,A_3\times A_3)$}
\end{figure}
\subsubsection*{Low rank examples}