§6 · Agent 6 · Audit: integral models · 2026-10-18T03:57:25Z

Audit 6.1 · Independent recomputation of the \(45\times 6\) table of \(E_6/F_4\) (claim A2-6)

proved audit: correct

Second audit: correct. Recomputed every item of Computation 2.6 with my own enumeration (Weyl elements as integer matrices on simple-root coordinates; Bruhat order by the subword property on the W-orbit of 2rho, not A6's lifting property): |C| = 45, w0 in C, theta(alpha1) = -(alpha1+alpha2+2alpha3+2alpha4+alpha5), Phi_x = w(Phi_L) of size 24, no real roots, no gamma+theta_x(gamma) in Phi or 0, l(x) = 12+2c_x, dimension multiplicities 1,2,3,4,5,5,5,5,5,4,3,2,1, 90 G / 90 open U / 90 small U with the partner rule, item 5 on all 90 open type-U pairs, and choice-independent closure sets equal to Bruhat up-sets (821 pairs, transitive). Both reduced words of w_L quoted in the run were verified, and A6's remark that Computation 2.6 item 3 is elementary is right. A6's scripts, run afterwards from a copy, print identical values; nothing material was missed.

An enumeration written from scratch, before reading A2's scripts, reproduces every item of Computation 2.6: 45 orbits, dimensions 14 to 26, 90 pairs of each kind, the type-U root sets, and closure sets equal to Bruhat up-sets (821 relations), independent of all choices. Outcome: holds.

Audit 6.1.

Let \(\Phi\) be the root system of \(E_6\) (Bourbaki labels), \(w_L\) the longest element of \(W(D_4)\) on nodes \(2,3,4,5\), \(\theta=-w_L\), \(\mathcal C=\{ww_Lw^{-1}\}\), and for \(x\in\mathcal C\): \(\theta_x=-x\), \(\Phi_x=\{\gamma:\theta_x\gamma=\gamma\}\), \(c_x\) the number of pairs \(\{\gamma,\theta_x\gamma\}\subseteq\Phi^+\) with \(\gamma\ne\theta_x\gamma\). Then:

  1. \(|\mathcal C|=45\), \(w_0\in\mathcal C\), \(\theta\alpha_1=-(\alpha_1+\alpha_2+2\alpha_3+2\alpha_4+\alpha_5)\); the roots fixed by \(\theta\) are the 24 roots of the black \(D_4\).
  2. For every \(x\): \(\Phi_x=w(\Phi_L)\) (24 roots); no root with \(x\gamma=\gamma\) (no real roots); \(\gamma+\theta_x\gamma\notin\Phi\cup\{0\}\) for all \(\gamma\); \(\ell(x)=12+2c_x\).
  3. \(\dim Y_x=26-c_x\) takes the values \(14,\dots,26\) with multiplicities \(1,2,3,4,5,5,5,5,5,4,3,2,1\).
  4. Of the 270 pairs \((x,\alpha)\): 90 with \(\alpha\in\Phi_x\) (G), 90 with \(x\alpha\in\Phi^+\) (U, \(Y_x\) open), 90 with \(x\alpha\in\Phi^-\smallsetminus\Phi_x\) (U, \(Y_x\) small); \(s_\alpha xs_\alpha=x\) iff type G, otherwise the partner has dimension \(\mp1\) and the opposite kind. Open orbit \(x=w_L\): U for \(\alpha_1,\alpha_6\), G for \(\alpha_2,\dots,\alpha_5\).
  5. For each of the 90 open type-U pairs, with \(\delta=-\theta_x\alpha\): \(\delta\in\Phi^+\smallsetminus\{\alpha\}\), \(\Psi:=\Phi(P_\alpha)\cap\theta_x\Phi(P_\alpha)=(\Phi^+\cap\theta_x\Phi^+)\sqcup\{-\alpha,\delta\}\), \(\Psi\cap\{\pm\alpha\}=\{-\alpha\}\), \(-\alpha-\theta_x\alpha\notin\Phi\cup\{0\}\), and \(4+\#(\Psi\cap\Phi_x)+\tfrac12\#(\Psi\smallsetminus\Phi_x)=17+c_x\).
  6. The recursion \(\Sigma_{w_0}=\{w_0\}\), \(\Sigma_x=\Sigma_{x'}\cup\{s_\alpha vs_\alpha: v\in\Sigma_{x'},\alpha\notin\Phi_v\}\) gives the same set for every admissible \(\alpha\) at every step, and \(\Sigma_x=\{z\in\mathcal C: x\le z\}\) (Bruhat order); there are 821 pairs \((x,z)\) with \(z\in\Sigma_x\).

What was computed.

  1. The roots of \(E_6\) were taken as the 72 vectors of norm 2 in the root lattice, with Gram matrix the Cartan matrix (chain \(1\!-\!3\!-\!4\!-\!5\!-\!6\), node 2 on node 4); every one has coefficients of one sign, and 36 are positive.
  2. \(W\) acts by permutations of the 72 roots, generated by the six simple reflections. \(w_L\) was found as the element of maximal length (12) among the 192 elements generated by \(s_2,\dots,s_5\); it negates \(\alpha_2,\dots,\alpha_5\) and has reduced word \(s_2s_3s_4s_2s_3s_4s_5s_4s_2s_3s_4s_5\).
  3. \(\mathcal C\) was obtained by closing \(\{w_L\}\) under conjugation by simple reflections, recording for each \(x\) an element \(w\) with \(x=ww_Lw^{-1}\) (checked).
  4. For each \(x\) the quantities in items 2, 4, 5 were computed directly from the permutation \(-x\).
  5. The Bruhat order on \(W\) was computed by the lifting property: if \(sv<v\) then \(u\le v\) iff \(su\le sv\) (when \(su<u\)) or \(u\le sv\) (when \(su>u\)). The recursion of item 6 was run by increasing dimension, computing \(\Sigma_x\) for every simple \(\alpha\) with \(x\alpha>0\).

Results. All six items hold exactly as stated in Computation 2.6; no check failed. In item 6 the closure sets are transitive (\(\Sigma_v\subseteq\Sigma_x\) for \(v\in\Sigma_x\)), contain \(x\), and coincide with Bruhat up-sets in \(\mathcal C\).

Why it suffices. These are statements about the finite group \(W(E_6)\) and its root system, hence valid over every base; they are exactly the finite inputs used in Proposition 2.3, Proposition 2.5, Lemma 2.7 and Theorem 2.8. Note that item 3 of Computation 2.6 (every \(x\ne w_0\) has a simple \(\alpha\) with \(x\alpha>0\), and then automatically \(\alpha\notin\Phi_x\)) is elementary: \(x\alpha<0\) for all simple \(\alpha\) forces \(x=w_0\), and \(\alpha\in\Phi_x\) means \(x\alpha=-\alpha\).

Comparison. Only after this computation were A2's scripts e6_orbits.py and e6_checks.py run (from a copy); they print the same values (their labels U\(-\), U\(+\) are our open and small type U).

Outcome: holds.

Builds on

  • Computation 2.6 · The \(45\times6\) pairs: 90 of type \(G\), 180 of type \(U\)

Built on by

Sources