Statement

Let g\mathfrak g be a complex with h\mathfrak h and root-space decomposition

g=hαΦgα.\mathfrak g=\mathfrak h\oplus\bigoplus_{\alpha\in\Phi}\mathfrak g_\alpha.

Suppose kg\mathfrak k\subseteq\mathfrak g is , so that it can be written

k=tαΨgα,t=kh,\mathfrak k=\mathfrak t\oplus\bigoplus_{\alpha\in\Psi}\mathfrak g_\alpha, \qquad \mathfrak t=\mathfrak k\cap\mathfrak h,

with Ψ=Ψ\Psi=-\Psi and with t\mathfrak t containing the coroots of Ψ\Psi. Then its is

Cg(k)=t0βΩgβ,C_{\mathfrak g}(\mathfrak k) =\mathfrak t_0\oplus\bigoplus_{\beta\in\Omega}\mathfrak g_\beta,

where

t0={Hh:α(H)=0 for all αΨ}\mathfrak t_0 =\{H\in\mathfrak h:\alpha(H)=0\text{ for all }\alpha\in\Psi\}

and Ω\Omega consists of the roots βΦ\beta\in\Phi satisfying

βt=0and[gβ,gα]=0 for every αΨ.\beta|_{\mathfrak t}=0 \quad\text{and}\quad [\mathfrak g_\beta,\mathfrak g_\alpha]=0 \text{ for every }\alpha\in\Psi.
Simply laced simplification

Assume Φ\Phi is . Because t\mathfrak t contains the coroots of Ψ\Psi, the condition βt=0\beta|_{\mathfrak t}=0 makes β\beta orthogonal to every root in Ψ\Psi. In a simply laced system this already implies β±αΦ\beta\pm\alpha\notin\Phi for all αΨ\alpha\in\Psi. Hence

Ω={βΦ:βt=0}.\Omega=\{\beta\in\Phi:\beta|_{\mathfrak t}=0\}.

Using the to identify h\mathfrak h and h\mathfrak h^*, this says: retain the Cartan directions orthogonal to the root span of k\mathfrak k, and retain exactly those ambient whose roots are orthogonal to the entire toral part t\mathfrak t.

Why the toral part matters

If k\mathfrak k has a nontrivial center, then t\mathfrak t can be strictly larger than the span of its coroots. A root can be orthogonal to every root of k\mathfrak k while failing to vanish on these extra central directions. Such a root space does not centralize k\mathfrak k. Thus replacing βt=0\beta|_{\mathfrak t}=0 merely by orthogonality to Ψ\Psi is not valid without an additional hypothesis.

General root-system caution

The bracket condition in the definition of Ω\Omega is the general criterion. In a crystallographic , once β\beta vanishes on t\mathfrak t and t\mathfrak t contains the coroots of Ψ\Psi, orthogonality to every αΨ\alpha\in\Psi forces β±αΦ\beta\pm\alpha\notin\Phi by the root-string property, so the simply-laced simplification extends to this setting. Orthogonality to Ψ\Psi alone remains insufficient when t\mathfrak t has extra central directions.

References
  1. Eugene B. Dynkin, “Semisimple subalgebras of semisimple Lie algebras,” Matematicheskii Sbornik 30(72), no. 2 (1952), 349–462; English translation, AMS Translations, Series 2, vol. 6 (1957), 111–244. Journal record.
  2. John C. Baez, “Three Generations in E7E_7,” 2026, §§3, 6–7. arXiv record.