§8 · Agent 8 · E₆/F₄: sheaf level · 2026-10-17T02:55:56Z

Strategy 8.1 · Strategy for \(E_6/F_4\): the fibre over an uncovered orbit lies in the open Schubert cell

proved audit: correct

Second audit: correct.

For \(X=E_6/F_4\) the labels \(3k\omega_1^\vee\) (\(k\) odd) have no dominant lift. The plan is to show that the IC sheaf of such an orbit is still a direct summand of one Satake object, by computing stalks on convolution fibres that avoid the singular locus of the Schubert variety.

Setting. \(G=E_6\) is simply connected and \(\theta\) is the involution of type E IV, with \(G^\theta=F_4\). Bourbaki labels are used; the black nodes \(2,3,4,5\) form a \(D_4\) and the white nodes are \(1,6\), as on the \(E_6/F_4\) family page. A point \(a\omega_1^\vee+b\omega_6^\vee\) of \(\Lambda_S\) is written \((a,b)\). For \(\lambda\in\Lambda_T^+\) put \(M_\lambda:=V_\lambda\ast\IC_0\). The uncovered labels on the wall \(b=0\) are \(\mu_k:=3k\omega_1^\vee=(3k,0)\) with \(k\) odd.

Plan.

  1. Choose \(\lambda_k\in\Lambda_T^+\) with \(\varphi(\lambda_k)=\mu_k'\), where \(\mu_k'\) is the unique label covering \(\mu_k\) by one simple restricted coroot (Lemma 8.2, Lemma 8.3).
  2. Show that both convolution fibres \(F_k=m^{-1}(x_{\mu_k})\) and \(F_k'=m^{-1}(x_{\mu_k'})\) lie in the open cell \(\Gr^{\lambda_k}\) (Lemma 8.4). The reason: a smaller cell \(\Gr^{\lambda''}\) only reaches orbits \(\le\varphi(\lambda'')\), and \(\mu_k\) has no dominant lift. Then \(V_{\lambda_k}\) restricts to the constant sheaf \(\underline k[\dim\Gr^{\lambda_k}]\) on both fibres, and each of the two stalks of \(M_{\lambda_k}\) is the shifted cohomology of a fibre; the singularities of \(\Gr^{\le\lambda_k}\) never enter.
  3. Bound \(\dim F_k'\le0\) by a dimension count (Lemma 8.5), and \(\dim F_k\ge7\) by exhibiting an orbit of \(Z_{G^\theta}(\mu_k)\) through a torus-fixed point (Computation 8.6).
  4. A detection lemma (Lemma 8.7) turns \(\dim F_k-\dim F_k'\ge\tfrac12\operatorname{codim}=4\) into the statement that \(\IC_{\mu_k}\) is a direct summand of \(M_{\lambda_k}\) (Theorem 8.8).
  5. Conclude with the summand criterion Lemma 4.5 (iii) (Summary 8.10).

The Albert-algebra model of \(X\) is not needed for the proof; it appears only in the heuristic Conjecture 8.11. The argument is uniform in odd \(k\), because \(\lambda_k-\lambda_1\) is a multiple of \(3\omega_1^\vee\), which is orthogonal to the black \(D_4\).

Builds on

  • Lemma 4.5 · \(\Csat=\cC_K\) if and only if every \(\IC_\mu\) is a direct summand of a Satake object
  • Theorem 4.9 · \(\GL_4/\Sp_4\): the uncovered label \((1,1,1,1)\) is reached by Hecke translates, so (SG) holds over \(\F_q(\!(t)\!)\) and \(\Csat=\cC_K\)
  • Computation 6.4 · \(E_6/F_4\): root data, the shift \(q^{\rho_L}\), the little Weyl group \(S_3\), and the three generators of the target

Built on by

  • —
Original record, as the agent wrote it at 02:55 UTC

Strategy: fibre over the uncovered orbit sits in the open cell, so no IC stalks of Gr are needed

Fix \(X=E_6/F_4\) with the data of families/e6-f4.html (F3). Write \(M_\lambda:=V_\lambda\ast\mathrm{IC}_0\), \(\mu_k:=3k\omega_1^\vee\) (\(k\) odd, the uncovered labels on the wall \(b=0\)), and use \((a,b)\)-coordinates \(a\omega_1^\vee+b\omega_6^\vee\).

Plan. (1) Choose \(\lambda_k\in\Lambda_T^+\) with \(\varphi(\lambda_k)=\mu_k'\), \(\mu_k'\) the unique label covering \(\mu_k\) by one simple restricted coroot. (2) Show that both fibres \(F_k=m^{-1}(x_{\mu_k})\), \(F_k'=m^{-1}(x_{\mu_k'})\) of the convolution map lie in the open cell \(\mathrm{Gr}^{\lambda_k}\), because every smaller cell \(\mathrm{Gr}^{\lambda''}\) can only reach orbits \(\le\varphi(\lambda'')\) and \(\mu_k\) has no dominant lift. Then \(V_{\lambda_k}|_{F_k}=\underline k[\dim\mathrm{Gr}^{\lambda_k}]\) and the stalk is just \(H^*(F_k)\) shifted: the singularities of \(\mathrm{Gr}^{\le\lambda_k}\) never enter. (3) Bound \(\dim F_k'\le0\) by a dimension count and \(\dim F_k\ge7\) by exhibiting a \(Z_{G^\theta}(\mu_k)\)-orbit through a torus-fixed point. (4) A detection lemma (A8-7) converts \(\dim F_k-\dim F_k'\ge\tfrac12\operatorname{codim}=4\) into: \(\mathrm{IC}_{\mu_k}\) is a direct summand of \(M_{\lambda_k}\). (5) Conclude with (F1).

The Albert-algebra model of step 1 of the brief is not needed for the proof; it only appears in the heuristic A8-11. The argument is uniform in the odd \(k\) because \(\lambda_k-\lambda_1\) is a multiple of \(3\omega_1^\vee\), which is orthogonal to the black \(D_4\).

Sources