§5 · Agent 5 · Audit: transfer, hypotheses · 2026-10-18T04:04:01Z

Audit 5.7 · Audit of A4-3 (τ closed immersion) and A4-9 (general symmetric X)

proved audit: correct

Second audit: correct. Re-derived Audit 4.3 item 1: T_qQ = g^{-dtheta_q}, so twisted orbits are open and closed; dtau = 2 id; a fibrewise flat etale monomorphism is an open immersion; closedness uses (H-Cartan) only over unramified p-adic fields (a characteristic-0 input in characteristic 0) plus Lang for the finitely many special-fibre orbits. The Chen-Yi supplement matches chen-yi-v4 (k = F_p-bar; the proof of Prop. 17(i) cites GN10 and Nad04). For Proposition 4.9 the rationality of x_kappa, the splitting and G(Lambda)-conjugacy of T_H, and the (T3) membership test are correct. One point Audit 5.7 did not raise: Proposition 4.9 step 5 asserts that P_kappa is the same for all F'_e because Theta 'with its signs' is, but the signs depend on the rationality of colours over F'_e; they are e-independent when (A1)-(A5)Sakellaridis's standing assumptions for Theorem 8.0.2: (A1) characteristic-0 field and split group; (A2) integral model; (A3a) rank-one models over O, (A3b) Cartan decomposition, (A3c) Iwahori orbits (his Axioms 2.2.1, 2.4.1, 2.4.2); (A4) spherical roots and colours; (A5) homogeneous, affine, one open Borel orbit. hold over O'_e for all large e (or after restricting to a cofinal set of e), a one-line justification that changes no conclusion. Correct.

That \(\tau\) is a closed immersion over \(W(\mathbb F_q)\) is correctly proved, using (H-Cartan) only at unramified \(p\)-adic fields. The remark on Chen–Yi is confirmed from their text. Proposition 4.9 is a correct conditional theorem whose hypotheses (T4)–(T5) live at \(p\)-adic places.

Audit 5.7.

Outcome: Audit 4.3 holds (item 2 rests on the cited existence of a closed immersion \(G_\Lambda\hookrightarrow\mathrm{GL}(W)\)). Proposition 4.9 holds as a conditional statement under (T1)–(T5)Hypotheses of the characteristic-p transfer for symmetric X over W(F_q): (T1) H split with connected fibres; (T2) X(E) = G(E)x0; (T3) τ detects integral points; (T4) the Cartan decomposition over the ramified fields F'_e; (T5) Sakellaridis's theorem over F'_e = Q_q(p^{1/e}), e ≫ 0.. Supplement to the Chen–Yi remark: Prop. 17(i) of chen-yi-v4 reads "Part (i) is proved in [GN10, Theorem 8.2.9] and [Nad04, Theorem 4.2.]", which are characteristic-\(0\) sources. Moreover, Chen–Yi work over \(k(\!(t)\!)\) with \(k=\bar{\mathbb F}_p\) algebraically closed, so even a characteristic-\(p\) proof of Prop. 17(i) would concern \(\bar{\mathbb F}_p(\!(t)\!)\), not \(\mathbb F_q(\!(t)\!)\). Either way the pages' input D0D0 · Orbit geometry over F̄_q, p ≠ 2 (Chen–Yi, Prop. 17, Thm 18): orbits indexed by Λ_S^+, closures, the support bound and the stabilizers. needs replacing; Proposition 4.5 or Theorem 3.8 replaces it.

Audit 4.3, item 1.

  1. Fibres. Over \(k=\bar k\) with \(\operatorname{char}k\ne2\), the twisted stabilizer of \(q\in Q\) is \(G^{\theta_q}\) with \(\theta_q=\operatorname{Int}(q)\theta\), an involution since \(\theta(q)=q^{-1}\). Write tangent vectors at \(q\) as \(Yq\). From \(\iota(\exp(\epsilon Y)q)=\exp(-\epsilon\,d\theta_q Y)q\) one gets \(T_qQ=\mathfrak g^{-d\theta_q}\). Hence \(\dim Gq=\dim T_qQ\ge\dim_qQ\), so every twisted orbit is open, and therefore also closed. At \(x_0\), \(d\tau=2\cdot\mathrm{id}\), so \(\tau_k\) is an étale monomorphism onto \(G\cdot1\).
  2. Over \(\Lambda\). Flat on fibres gives flat (EGA IV\(_3\) 11.3.10). A monomorphism of finite type is unramified, so \(\tau\) is an étale monomorphism, hence an open immersion (EGA IV\(_4\) 17.9.1).
  3. Closedness. A point of \(\overline{\tau(X)}\smallsetminus\tau(X)\) lies in the special fibre and in a geometric orbit \(\Omega\ne G\cdot1\). \(\Omega\) has \(\mathbb F_{q^r}\)-points for large \(r\) (Lang, applied to the homogeneous space of the connected \(G\)), which lift to \(Q(\Lambda_r)\) because \(Q\) is smooth. Case \(q'_\eta\in\tau(X)\): (H-Cartan) over the unramified field \(W(\mathbb F_{q^r})[1/p]\) (S4-13) forces \(\kappa=0\), so \(y'\in\tau(X)\), a contradiction. Case \(q'_\eta\notin\tau(X)\): the orbit map \(g\mapsto g\cdot q'\) is smooth with open image missing \(\tau(X)\), again a contradiction. This uses a characteristic-\(0\) theorem only in characteristic \(0\). Closed immersions base-change, so \(X(\mathcal O)=\{y:\tau(y)\in G(\mathcal O)\}\) over \(\mathbb F_q[\![t]\!]\).
  4. Item 2. \(A\mapsto(A,A^{-1})\) and equivariance are fine. The closed immersion \(G_\Lambda\hookrightarrow\mathrm{GL}(W)\) is cited (SGA 3, not opened). The main line does not use items 1–2.
  5. Chen–Yi. Checked in chen-yi-v4 (text copy chenyi.txt, §3.1, Proposition 17 and proof).

Proposition 4.9.

  1. \(x_\kappa=\kappa(\sqrt\varpi)x_0\) is \(E\)-rational: for the non-trivial \(\sigma\in\mathrm{Gal}(E(\sqrt\varpi)/E)\), separable since \(p\ne2\), we have \(\sigma(\kappa(\sqrt\varpi))=\kappa(-1)\kappa(\sqrt\varpi)\), and \(\kappa(-1)\in S[2]\subseteq H\) because \(\theta(s)=s^{-1}=s\) there. Also \(\tau(x_\kappa)=\kappa(\varpi)\).
  2. Step 1: \(T_H\) lies in a maximal torus of the reductive \(Z_G(T_H)\) that lifts a split torus of the special fibre, hence is split over the henselian \(\Lambda\). Split maximal tori of \(G_\Lambda\) are \(G(\Lambda)\)-conjugate: the transporter is smooth with an \(\mathbb F_q\)-point, so Hensel applies. Then \(\eta(\varpi)=gt^{\eta'}g^{-1}\) with fixed \(g\in G(\Lambda)\), and Lemma 4.2(e) still applies.
  3. Step 3: by (T3), "\(x_\kappa\in Kt^\lambda X(\mathcal O)\)" is equivalent to \(\exists k\in K\) with \(t^{-\lambda}k^{-1}t^\kappa\theta(k)t^{\theta\lambda}\in K\). This is of type Lemma 4.2(e).
  4. Scope. (T1) fails whenever \(G^\theta\) is disconnected (e.g. \(\mathrm{O}_n\subseteq\mathrm{GL}_n\)). (T4) fails when \(|J|>1\) or \(A_X(\mathcal O)\ne\operatorname{im}T(\mathcal O)\) (cf. the \(\mathrm{O}_2\backslash\mathrm{GL}_2\) remark on p. 11). Proposition 4.9 states these as hypotheses, so this is not an error. In this route the characteristic-\(p\) theorem never meets Sakellaridis's axioms over \(\mathbb F_q[\![t]\!]\). That answers (Q1) in a different form from "(A1)–(A5)Sakellaridis's standing assumptions for Theorem 8.0.2: (A1) characteristic-0 field and split group; (A2) integral model; (A3a) rank-one models over O, (A3b) Cartan decomposition, (A3c) Iwahori orbits (his Axioms 2.2.1, 2.4.1, 2.4.2); (A4) spherical roots and colours; (A5) homogeneous, affine, one open Borel orbit. over \(\mathbb F_q[\![t]\!]\) imply the theorem".

Builds on

  • Audit 4.3 · Check: \(\tau\) is a closed immersion over \(W(\mathbb F_q)\), giving (H-int)
  • Proposition 4.9 · Proposition: what the characteristic-\(p\) theorem needs for symmetric \(X\)

Built on by

Sources