Dual Root Data
The dual root datum swaps characters with cocharacters and roots with coroots. It is the combinatorial source of Langlands duality — the operation that turns the root datum of \(G\) into the root datum of the Langlands dual group \(\hat G\).
Definition
For a root datum \(\Psi = (X, \Phi, X^\vee, \Phi^\vee)\) with the canonical pairing \(\langle -, - \rangle : X \times X^\vee \to \mathbb{Z}\), the dual root datum is \[ \Psi^\vee \;:=\; (X^\vee, \Phi^\vee, X, \Phi) \] with the pairing \(\langle -, - \rangle^\vee : X^\vee \times X \to \mathbb{Z}\), \(\langle x^\vee, x \rangle^\vee := \langle x, x^\vee \rangle\) (i.e.\ the swapped pairing).
The map \(\Psi \mapsto \Psi^\vee\) is a contravariant involution: \((\Psi^\vee)^\vee = \Psi\).
Properties
- The bijection \(\Phi \leftrightarrow \Phi^\vee\) of the original root datum becomes part of the new root datum: roots of \(\Psi^\vee\) are coroots of \(\Psi\) and vice versa.
- The pairing condition \(\langle \alpha, \alpha^\vee \rangle = 2\) is preserved (becomes \(\langle \alpha^\vee, \alpha \rangle = 2\) in \(\Psi^\vee\)).
- The Weyl group of \(\Psi^\vee\) is canonically isomorphic to that of \(\Psi\) (same set of reflections, dual realisation).
Effects on Dynkin type
For irreducible root systems, duality acts on Dynkin types as follows:
| Type \(G\) | Type \(\hat G\) (= dual) |
|---|---|
| \(A_n\) | \(A_n\) (self-dual) |
| \(B_n\) | \(C_n\) |
| \(C_n\) | \(B_n\) |
| \(D_n\) | \(D_n\) (self-dual) |
| \(E_6\) | \(E_6\) (self-dual) |
| \(E_7\) | \(E_7\) (self-dual) |
| \(E_8\) | \(E_8\) (self-dual) |
| \(F_4\) | \(F_4\) (long ↔ short swap) |
| \(G_2\) | \(G_2\) (long ↔ short swap) |
Long and short roots swap their roles in non-simply-laced types.
Lattice gap inversion
The lattice gap \(\mathbb{Z}\Phi \subseteq X \subseteq P\) (weight lattice) and its dual \(Q^\vee = \mathbb{Z}\Phi^\vee \subseteq X^\vee \subseteq P^\vee\) swap under duality:
- Simply connected \(G\) (\(X^\vee = Q^\vee\)) becomes adjoint \(\hat G\) (\(X = \mathbb{Z}\Phi\));
- Adjoint \(G\) (\(X = \mathbb{Z}\Phi\)) becomes simply connected \(\hat G\).
So:
- \((\mathrm{SL}_n)^\vee = \mathrm{PGL}_n\) (sc ↔ ad);
- \((\mathrm{SO}_{2n+1})^\vee = \mathrm{Sp}_{2n}\) (sc-of-B_n ↔ sc-of-C_n);
- \(\mathrm{GL}_n^\vee = \mathrm{GL}_n\) (self-dual, but with the determinant character swapped to the determinant cocharacter).
Canonical inclusions
The lattice inclusions \(Q \subseteq X \subseteq P\) dualise to inclusions \(Q^\vee \subseteq X^\vee \subseteq P^\vee\), giving: \[ \pi_1(\Psi) \;=\; X^\vee / Q^\vee \;\stackrel{\text{dual}}{\longleftrightarrow}\; X / Q \;=\; Z(\Psi^\vee). \]
This is the fundamental Kottwitz duality: \[ \pi_1(G) \;=\; X^*(Z(\hat G)). \]
Examples
\(\mathrm{GL}_n\) (self-dual)
\(X = X^\vee = \mathbb{Z}^n\); roots \(\Phi = \{\varepsilon_i - \varepsilon_j\} \subseteq X\) and coroots \(\Phi^\vee = \{\varepsilon_i^\vee - \varepsilon_j^\vee\} \subseteq X^\vee\).
Dualising swaps the two copies of \(\mathbb{Z}^n\). Under the identification, this is just the identity — \(\mathrm{GL}_n\) is self-dual.
\(\mathrm{SL}_n\) ↔ \(\mathrm{PGL}_n\)
\(\mathrm{SL}_n\): \(X = \mathbb{Z}^n/\mathbb{Z}(1, \ldots, 1)\), \(X^\vee = \{(a_i) : \sum a_i = 0\}\). Dualising swaps these, giving \(\mathrm{PGL}_n\)'s root datum.
Type \(C_2 = \mathrm{Sp}_4\)
Roots in \(X = \mathbb{Z}^2\): \(\Phi = \{\pm \varepsilon_1 \pm \varepsilon_2, \pm 2 \varepsilon_1, \pm 2 \varepsilon_2\}\). Coroots in \(X^\vee\): \(\Phi^\vee = \{\pm \varepsilon_1^\vee \pm \varepsilon_2^\vee, \pm \varepsilon_1^\vee, \pm \varepsilon_2^\vee\}\). Dualising gives \(B_2 = \mathrm{Spin}_5\) (simply connected of \(B_2\)).
Action on automorphism groups
Root-datum duality preserves the automorphism group: \[ \mathrm{Aut}(\Psi) \;=\; \mathrm{Aut}(\Psi^\vee). \]
For an outer form determined by a Galois representation \(\Gamma \to \mathrm{Out}(G)\), the dual outer form has the same representation \(\Gamma \to \mathrm{Out}(\hat G) = \mathrm{Out}(G)\).
Functoriality
Duality is contravariant on morphisms: for \(f : \Psi \to \Psi'\), the dual morphism \(f^\vee : (\Psi')^\vee \to \Psi^\vee\) is induced by the dual lattice maps.
This compatibility is what makes Langlands duality functorial: a homomorphism of L-groups corresponds to a "co-functoriality" in the forward direction.
Role in Langlands
The dual root datum is the algebraic underpinning of:
- The dual group \(\hat G\): the unique split reductive group with root datum \(\Psi^\vee\).
- Langlands duality: the principle that automorphic representations of \(G\) correspond to L-parameters into \({}^L G\), with the dual side built from \(\Psi^\vee\).
- Symplectic-orthogonal correspondence: the duality \(\mathrm{Sp}_{2n} \leftrightarrow \mathrm{SO}_{2n+1}\), which makes Arthur's classification of automorphic representations symmetric between symplectic and odd-orthogonal groups, comes from \(\Psi^\vee\) for \(B_n \leftrightarrow C_n\).
References
- Borel, Automorphic L-functions (Corvallis 1979), §2.
- Springer, Linear Algebraic Groups (1998), §7.
- Kottwitz, Stable trace formula: cuspidal tempered terms (Duke 1984), §1.