§5 · Agent 5 · Audit: split candidate · 2026-10-17T23:24:13Z
Audit 5.9 · Audit of A3-9, A3-2(c): given D1–D2, \(\mathcal C_k^{\mathrm{Sat}}=\mathcal C_k\)
proved
Given the \(\ell\)-adic formalism for \(L^+G\backslash LX\), the deduction is correct: \(V_\lambda\ast\mathrm{IC}_0\) is perverse with multiplicity matrix \(c\), every label is a summand label, and \((\mathrm{SG})_n\) is surjectivity of the restriction of characters, which fails by fusion.
Audit 5.9.
Outcome: holds, as a conditional statement. Assume for \(X\) the inputs of Proposition I and IIb:
- H2 (equivariant recollement);
- PH and D1 (a Frobenius-stable perverse t-structure with simple objects \(\mathrm{IC}_\nu\), Verdier duality, proper base change and Grothendieck–Lefschetz for \(m_\lambda\));
- D2 (decomposition theorem for \(m_\lambda\)).
Then over \(k=\overline{\mathbb F}_q\):
- (a) \(V_\lambda\ast\mathrm{IC}_0\simeq\bigoplus_\nu\mathrm{IC}_\nu^{\oplus c_{\nu\lambda}}\);
- (b) \(\mathcal M=\Lambda_X^+\), hence \(\mathcal C_k^{\mathrm{Sat}}=\mathcal C_k\);
- (c) \(m^{(n)}_{\nu\lambda}=c_{\nu\lambda}q^{n(\langle2\rho,\lambda\rangle-\delta(\nu))/2}\);
- (d) \(\operatorname{tr}_n\mathrm{IC}_{(0,0,2)}\notin\mathcal H_{K_n}\ast\varphi_0\otimes\overline{\mathbb Q}_\ell\) for all \(n,q\).
The other inputs used (H1, H3, D3) are proved in A3-6 (Audit 5.6).
Re-derivation.
- Perversity. The support condition: on \(\mathrm{Gr}^{\lambda'}\tilde\times L^+X\) the source complex sits in degrees \(\le-\langle2\rho,\lambda'\rangle\). Hence the stalk at \(x_\nu\) vanishes above degree \(\max_{\lambda'}(2\dim F^{\lambda'}_\nu-\langle2\rho,\lambda'\rangle)\le-\delta(\nu)\) (Audit 5.8). The cosupport condition follows from Verdier self-duality of \(V_\lambda\tilde\boxtimes\mathrm{IC}_0\), since \(m_\lambda\) is proper; this needs Verdier duality in the formalism (part of D1).
- Multiplicities. Constant coefficients come from connected stabilizers (H3). In degree \(-\delta(\nu)\) at \(x_\nu\):
- \(\mathrm{IC}_{\nu'}\) for \(\nu'>\nu\) contributes nothing, by the IC support condition;
- the strata \(\lambda'<\lambda\) contribute nothing, because \(\mathrm{IC}_{\mathrm{Gr}_{\le\lambda}}|_{\mathrm{Gr}^{\lambda'}}\) lives in degrees \(\le-\langle2\rho,\lambda'\rangle-2\) and \(\dim F^{\lambda'}_\nu\le f'\).
So \(E_\nu=H^{2f}_c(F^\lambda_\nu)\), of dimension \(c_{\nu\lambda}\).
- Generation. Every label is relevant for some \(\lambda\) (Audit 5.8, step 6). So every \(\mathrm{IC}_\nu\) is a direct summand, and Proposition I gives \(\mathcal C_k^{\mathrm{Sat}}=\mathcal C_k\).
- Traces.
- Top components are defined over \(\mathbb F_q\) and geometrically irreducible (Lang–Weil, Audit 5.8 step 5), so \(A_{\nu,0}(\lambda)=q^f\cdot\mathrm{id}\) and \(m^{(n)}=D_n\,c\,D_n'\) with invertible diagonal \(D_n,D_n'\).
- IIb(iii) needs the IC traces to form a basis of \(C_{K_n}\otimes\overline{\mathbb Q}_\ell\). This holds because orbit closures are finite unions of orbits (Audit 5.6) and one \(K_n\)-orbit of rational points lies in each \(L^+G\)-orbit.
- IIa (\(\operatorname{tr}(V_\lambda\ast\mathrm{IC}_0)=\operatorname{tr}V_\lambda\ast\varphi_0\)) needs D1 and Lang's theorem for \(L^+G\).
- Fusion. Columns of \(c\) are restrictions of \(\mathrm{SO}_5^2\)-characters. The class function \(\psi(\chi)=\chi(a,b,a)-\chi(b,a,b)\), at generic \((a,b)\) (that is, \(a\ne\pm b^{\pm1}\), Audit 5.2), vanishes on every column. On \(V_{(0,2,0)}\) it equals \(b^2+b^{-2}-a^2-a^{-2}\ne0\). Numerically, the restricted characters agree at \((a,b,a)\) and \((b,a,b)\) to \(10^{-14}\) for all \(a_i\le3\).
- A3-2(c)(iii) and (b).
- (b): every irreducible of a closed reductive \(\check G_X\subseteq\check G\) is a summand of some restricted irreducible, and characters separate semisimple classes of the connected \(\check G_X\).
- (c)(iii): the hypotheses are exactly the conclusions of (a)–(c) above.
Status. This is a correct conditional statement. Its unconditional consequence, the failure of \((\mathrm{SG})_n\), is already proved at the function level (Audit 5.5). The only open content is the existence of the formalism; see Objection 5.10.
Builds on
- Corollary 3.9 · Given D1–D2 for \(X\): \(\mathcal C_k^{\mathrm{Sat}}=\mathcal C_k\), \(m^{(n)}=D_ncD_n'\)
- Proposition 3.2 · Thick closure versus span: twisted traces, restriction of class functions, fusion
- Audit 5.6 · Audit of A3-6: orbits \((d_1,d_2,e)\), connected stabilizers, \(\delta=3d_1+3d_2+e\)
- Audit 5.8 · Audit of A3-7, A3-8: position formula, semismallness, dual branching
Built on by
- Audit 5.11 · Audit of A3-11, A3-2: no n-dependence; retracts reach the trace, Hecke does not
- Summary 5.12 · Verdict: (SG) fails, proved; \(\mathcal C_k^{\mathrm{Sat}}=\mathcal C_k\) modulo (O1)–(O3)
Sources
- trace-pages/trace-b.html (IIb (i)–(iii))
- trace-pages/trace-a.html (IIa)
- props/summand.html (Proposition I, H1–H3, PH, DT)
- sun-decomposition.pdf, Thm 1.2
- decataldo-migliorini-v2.pdf