§6 · Agent 6 · Audit: integral models · 2026-10-18T03:58:48Z

Audit 6.3 · Audit of the Borel-orbit classification in every characteristic \(\ne2\) (claim A2-3)

proved audit: correct

Second audit: correct. Re-derived Proposition 2.3: the eigenspace count i_c - i_n = 32 - 2a with at most 12 imaginary roots when a <= 3 forces a = 4 = rank H; Audit 6.3's derivation of a common maximal torus of B_1 and theta(B_1) from the Bruhat decomposition (g' = bnb', bT_1b^-1 in B_1 and B_2) is the standard argument and a valid repair of a citation that named the decomposition instead of its corollary. The bookkeeping g = b n' h, the action of theta on W, the separation of orbits by Bruhat cells through tilde-tau, and the dimension/rank count all check. Conrad Thm 1.4.12 is stated over an algebraically closed field (checked in the local text) and is used over k-bar; Steinberg 7.5, 8.1 and CGP A.8.10 / Edixhoven 3.4 are not local, but the recalled statements are the standard ones and their hypotheses (an involution is semisimple in characteristic not 2, E6 simply connected) hold.

Every step of Proposition 2.3 is correct over an algebraically closed field of characteristic \(\ne2\), \(p=3\) included, and no characteristic-0 theorem is used; one sentence (two Borel subgroups share a maximal torus) needs a line of proof, supplied here. Outcome: holds with that repair.

Audit 6.3.

Let \(k\) be algebraically closed with \(\operatorname{char}k\ne2\). Granting that \(H_k=G_k^\theta\) is smooth (CGP, Prop. A.8.10(2), or Edixhoven, Prop. 3.4) and connected (Steinberg, Mem. AMS 80, Thm. 8.1), and Steinberg's Thm. 7.5 (a semisimple automorphism of a connected linear algebraic group fixes a Borel subgroup and a maximal torus of it), items 1–4 of Proposition 2.3 hold. Repair: the step "\(B''\) and \(\theta B''\) share a maximal torus (Bruhat, Conrad Thm. 1.4.12)" is a consequence of the Bruhat decomposition, not its statement, and is proved in step 2 below. Outcome: holds with this repair.

  1. Torus conjugacy (item 1). For a \(\theta\)-stable maximal torus \(T'\), \(\theta\) acts on \(\mathfrak t'\) with eigenspaces of dimensions \(a=\dim T'_+\) and \(6-a\) (this uses \(\frac12\) and smoothness of \(T'^\theta\)). Each pair \(\{\gamma,\theta\gamma\}\) of real or complex roots contributes one line to \(\mathfrak g^{\theta}\) and one to \(\mathfrak g^{-\theta}\); an imaginary root contributes to exactly one of them. With \(\dim\mathfrak g^{\pm\theta}=52,26\) (Audit 6.2 step 7):
$$52=a+\tfrac{r+c}2+i_c,\qquad 26=(6-a)+\tfrac{r+c}2+i_n,\qquad i_c-i_n=32-2a .$$

The imaginary roots are \(\Phi(T')\) intersected with the \(+1\)-eigenspace of \(\theta\) on \(X^*(T')_{\mathbb Q}\), a simply laced root subsystem of rank \(\le a\); for \(a\le3\) it has at most 12 roots (\(A_3\); \(A_2{+}A_1\) has 8, \(A_1^3\) has 6). Then \(i_c\le12<26\le i_c-i_n\), absurd. So \(a=4\). Since \(Z_H(T^\theta)=T\cap H=T^\theta\simeq\mathbb G_m^4\) (Audit 6.2 steps 4–5), \(\operatorname{rank}H=4\), so \(T'_+\) is a maximal torus of the connected group \(H\), \(hT'_+h^{-1}=T^\theta\) for some \(h\in H(k)\), and \(hT'h^{-1}\subseteq Z_G(T^\theta)=T\). Verified; nothing depends on \(p\) beyond \(p\ne2\).

  1. Every orbit meets \(N_G(T)x_0\) (item 2), with the repair. Let \(B_1=g^{-1}Bg\), \(B_2=\theta(B_1)\). Choose a maximal torus \(T_1\subseteq B_1\) and \(g'\) with \(B_2=g'B_1g'^{-1}\). By the Bruhat decomposition (Conrad, Thm. 1.4.12, over the field \(k\)) \(g'=bnb'\) with \(b,b'\in B_1(k)\), \(n\in N(T_1)(k)\); then \(bT_1b^{-1}=bnT_1n^{-1}b^{-1}\subseteq bnB_1n^{-1}b^{-1}=B_2\) and \(bT_1b^{-1}\subseteq B_1\). So \((B_1\cap B_2)^\circ\) is a \(\theta\)-stable connected solvable group containing a maximal torus of \(G\); Steinberg's Thm. 7.5 (an involution is semisimple when \(\operatorname{char}k\ne2\)) gives a \(\theta\)-stable maximal torus \(T'\subseteq B_1\). By step 1, \(T'=h^{-1}Th\); \(hB_1h^{-1}\supseteq T\), so \(hB_1h^{-1}=n'^{-1}Bn'\) with \(n'\in N_G(T)(k)\), whence \(gh^{-1}n'^{-1}\in N_G(B)=B\) and \(gx_0\in Bn'x_0\).
  2. Distinct orbits (item 2). \(\theta\) acts on \(W=N_G(T)/T\) by \(w\mapsto w_Lww_L\) (\(\theta|_T=-w_L\) and \(-1\) is central in \(\mathrm{GL}(X^*(T))\)); hence \(\dot w\theta(\dot w)^{-1}\sigma_L\sigma_0\) represents \(ww_Lw^{-1}w_L\cdot w_Lw_0=xw_0\). \(C(B)=B^-=\sigma_0B\sigma_0^{-1}\) gives \(\theta(B)=\sigma_L\sigma_0B(\sigma_L\sigma_0)^{-1}\), so \(\tilde\tau(bz)=b\,\tilde\tau(z)\,\beta(b)^{-1}\) with \(\beta(b)=(\sigma_L\sigma_0)^{-1}\theta(b)\sigma_L\sigma_0\in B\), and \(\tilde\tau(By_x)\subseteq B\,\dot{(xw_0)}\,B\). Bruhat cells are disjoint, so distinct \(x\) give distinct orbits. Conversely lifts of a coset \(vC_W(w_L)\) give points of one \(T\)-orbit, because \(N_H(T)\to C_W(w_L)\) is onto (Audit 6.2 step 6). Verified.
  3. Dimension and rank (item 3). \(B_x=(B\cap\theta_xB)^{\theta_x}=T^{\theta_x}\ltimes(U_{\Phi^+\cap\theta_x\Phi^+})^{\theta_x}\) (fixed points of a semidirect product with both factors stable), Lie algebra of dimension \(4+12+c_x\); \(\ell(x)=12+2c_x\) (Audit 6.1); the \(B\)-eigencharacters of \(k(B/B_x)\) are the characters of \(T\) trivial on \(T^{\theta_x}\), a group of rank \(6-4=2\). Verified.
  4. Item 4 and the closing remark (Knop's action is \(x\mapsto s_\alpha xs_\alpha\) in both types, i.e. left multiplication on \(W/C_W(w_L)\)) are correct; the remark is not used.
  5. Not verified locally. Steinberg's Thms. 7.5 and 8.1, CGP A.8.10 and Edixhoven 3.4 are not among the local sources; the statements recalled above are the ones used, and they hold over algebraically closed fields of every characteristic (8.1 needs \(G\) simply connected, true here).

Outcome: holds with the repair of step 2.

Builds on

  • Proposition 2.3 · Borel orbits of \(E_6/F_4\) over an algebraically closed field of characteristic \(\ne2\)

Built on by

Sources