§5 · Agent 5 · Audit: transfer, hypotheses · 2026-10-18T04:03:11Z
Audit 5.6 · Audit of A4-8: Theorem 8.0.2 for M_θ over F_q((t)); uniqueness; (SG)
proved audit: correct ★ key step
The module transfer, the uniqueness argument and the consequences for (SG) are correct. The theorem is conditional on \((I_\Lambda)\), as stated, and also inherits \(H_{\mathrm{Sa08}}\) from Audit 4.7. Its claim to use 'no statement over \(\bar F\)' holds for the isomorphism but not for identifying the target with Sakellaridis's invariants in characteristic \(p\).
Audit 5.6.
Outcome: holds, conditionally on \((I_\Lambda)\) and \(H_{\mathrm{Sa08}}\) (Audit 5.5). Steps 1–5 of Theorem 4.8 are correct. (i) \(\Psi(\mathbf 1_{Kx_\kappa})=P_\kappa\) is a linear bijection \(C_K\to S\). It is \(\mathcal H_K\)-equivariant for every pair \((\lambda,\mu)\), each pair checked against one \(F'_e\) with \(e\ge\max(D(\lambda,\mu),D'(\lambda))\), where \(D'\) is as repaired in Objection 5.4. (ii) It is the unique \(\mathcal H_K\)-isomorphism with \(\varphi_0\mapsto1\): the conductor ideal \(\mathfrak f\subseteq R=\operatorname{im}r\) is non-zero by S4-7, which is unconditional algebra valid for every real \(q>1\). (iii) (SG) fails over \(\mathbb F_q(\!(t)\!)\), and \(C_K\) needs exactly two generators. Remarks. (a) No characteristic-\(p\) conclusion feeds back into a hypothesis (checklist item 10). (b) \(S\), \(W_X=S_3\), \(\Lambda_X^+\) and \(\delta(X)\) enter as the combinatorial data of the \(p\)-adic side. To call the result "Sakellaridis's theorem for \(X_F\)" one needs these to equal the characteristic-\(p\) invariants of \(X_{\bar F}\). That is Proposition 3.6, Proposition 3.7 (with Audit 5.10, Objection 5.11), so the claim "no statement over \(\bar F\) is used" applies to the proof of the isomorphism only.
- Bases. By Proposition 4.5, \(\{b_\kappa=\mathbf 1_{Kx_\kappa}\}_{\kappa\in\Lambda_S^+}\) is a basis of \(C_K(F)\). Over one \(F'_e\), Theorem 8.0.2 (available under \((I_\Lambda)\) by Audit 4.7(b)) makes \(\{P_\kappa\}\) a basis of \(S\), and Audit 5.5 step 5 shows it is the same set for every \(e\). So \(\Psi\) is bijective. \(P_0=1\) is stated in the proof on p. 64, and \(X(\mathcal O)=Kx_0\).
- Equivariance. For fixed \((\lambda,\mu)\): \(\Psi(T_\lambda*b_\mu)=\sum_{\nu\in N(\lambda,\mu)}c^\nu_{\lambda\mu}(F)P_\nu\) (Proposition 4.6(a)). This equals \(\sum_\nu c^\nu_{\lambda\mu}(F'_e)P_\nu\) (Proposition 4.6(b)), which is \(\Phi_{F'_e}(T'_\lambda*b'_\mu)=r(\mathcal ST'_\lambda)P_\mu\) by Theorem 8.0.2 over \(F'_e\), and \(\mathcal S T'_\lambda=\mathcal ST_\lambda\) (Proposition 4.6(c)). The \(T_\lambda\) span \(\mathcal H_K\) and the \(b_\mu\) span \(C_K\). Only finitely many \(\nu\) occur for each pair, so one \(e\) suffices per pair. The map \(r\) depends only on \(q\): \(\delta(X)^{1/2}\) is a power of \(q\).
- Uniqueness. If \(\Psi'\) is another such isomorphism, \(\Delta=\Psi'\Psi^{-1}\) is \(R\)-linear with \(\Delta(1)=1\), so \(\Delta|_R=\mathrm{id}\). S4-7(3) gives \(R=\{g:g(y_1)=g(\sigma y_1)\}\), so \(\mathfrak f=\{g:g(y_1)=g(\sigma y_1)=0\}\) is an ideal of \(S\) contained in \(R\), and it is non-zero because \(\operatorname{Spec}S\) is an irreducible surface. For \(0\ne a\in\mathfrak f\) and any \(s\): \(a\Delta(s)=\Delta(as)=as\). \(S\) is a domain (the coordinate ring of a translate of a torus, taken \(W_X\)-invariants), so \(\Delta(s)=s\).
- Consequences. \(\Psi(\mathcal H_K\varphi_0)=R\) and \(\dim S/R=1\) (S4-7), so (SG) fails. S4-8 needs only an isomorphism \(C_K\simeq S\) with \(\varphi_0\mapsto1\) and \(P_{3\omega_1^\vee}\notin R\). The latter is a statement about the explicit polynomial (lowest-weight Remark, p. 62) and holds in either label convention. Here \(3\omega_1^\vee=2e_1+e_6\in\Lambda_S^+\) (Audit 5.2).
- Convention (checklist item 8). The identity "\(f\) acts as multiplication by \(r(\mathcal Sf)\)" is transported in whatever convention Theorem 8.0.2 holds over \(F'_e\), because \(c^\nu_{\lambda\mu}\) and \(n_{\lambda\nu}\) are defined by the same formulas over both fields. The paper does not fix the conventions for \(\mathcal S\) or the action. Matching them with the left-action formulas of Proposition 4.6 can change \(r\) only by automorphisms of source and target (S4-8 step 1), which leaves (SG) and the two-generator statement unchanged.
- What remains. \((I_\Lambda)\) (Axiom 2.2.1 for \(\mathcal M_\theta\) over \(W(\mathbb F_q)\), for 45 orbits and 6 simple roots; Dead end 4.10) and \(H_{\mathrm{Sa08}}\).
Builds on
- Theorem 4.8 · Theorem: Theorem 8.0.2 for \(\mathcal M_\theta\) over \(\mathbb F_q((t))\); (SG) fails
Built on by
- Summary 5.13 · Verdict of audit A5: transfer route A4-2…A4-9 and char-p hypotheses A3-1…A3-10
- Audit 7.3 · Check: Sakellaridis's 2008 paper is stated for every local field of characteristic \(0\)
- Theorem 7.4 · Verdict: Sakellaridis's theorem over \(\mathbb F_q(\!(t)\!)\); for \(E_6/F_4\) in both characteristics, (A3a)(A3a) Rank-one models over O (Sakellaridis's Axiom 2.2.1): for every Borel orbit Y of maximal rank and every simple root α, the O-scheme YP_α/R(P_α) is one of the four standard PGL2-spaces over O, with his scheme structure on Y, and YP_α is the closure of its generic fibre minus the closure of the generic boundary. included
Sources
- Theorem 4.8 (claim audited)
- satake-two-levels/audit/S4.json S4-7 (im r, every real q>1), S4-8 (two generators)
- sakellaridis-v3.pdf Thm 8.0.2 and proof p. 64, Remark after Thm 7.2.1 p. 62
- Dead end 4.10 ((I_Λ) open)