§3 · Agent 3 · Symplectic family: fibres · 2026-10-17T02:48:26Z

Theorem 3.6 · \(\GL_{2m}/\Sp_{2m}\): the labels \((1^k0^{m-k})\), \((2^{k-m}1^{2m-k})\) and their shifts lie in \(\mathcal M\); \(\GL_4/\Sp_4\) is complete

proved audit: minor defect ★ key step

Minor defect (second audit). Fibre lemma, defect formula, maximality, shift stability and the GL_4/Sp_4 conclusion are correct: fibres recomputed for m = 2 at q = 3 and m = 3 at q = 2 by an independent lattice parametrization, including emptiness for labels with an entry 3; def(nu) = n_1(nu) recomputed for m <= 7 from the finite-type dimension count rather than from Prop. 64; the m = 3 lists of settled and open labels confirmed (g2_hnf_lattices.py, g2_defect_coverage.py). Defects that change no claim: (i) the codimensions are taken from Chen-Yi Prop. 64, whose hypothesis (LX connected) fails for GL_{2m}; the formula holds in every component by the dimension count of Lemma 3.1 (correction Correction 10.2). The scalar-shift repair proposed in Objection 7.11 would not reach the components k not divisible by m used here. (ii) 'IC[-k] is a summand of V_{omega_k} * IC_0' is off by the global shift between m_! k_Y[D] and V_{omega_k} * IC_0, which is harmless for membership in M.

Computes every fibre of the minuscule convolutions \(\Gr^{\omega_k}\mathbin{\tilde\times}L^+X\to LX\) for \(\GL_{2m}/\Sp_{2m}\) (they are Lagrangian Grassmannians), shows that the defect of a stratum equals its number of 1-blocks, and combines the maximal-defect lemma with the scalar shift. For \(\GL_4/\Sp_4\) every orbit label then lies in \(\mathcal M\), so \(\Csat=\cC_K\) given semisimplicity of Satake objects; for larger \(m\) a specific family of uncovered labels is reached.

Theorem 3.6.

Notation of Lemma 3.1, Computation 3.2, Proposition 3.4. For \(\nu\in\{0,1,2\}^m\) with \(|\nu|=k\) let \(n_i(\nu)=\#\{j:\nu_j=i\}\); \(\omega_\nu=\bigoplus_jt^{\nu_j}J_2\).

  1. Fibre lemma. For \(\lambda=\omega_k\), \(1\le k\le2m-1\),
$$F_{\omega_k}(\omega_\nu)=\{W\in\Gr(k,t^{-1}\Lambda_0/\Lambda_0):\ \bar\omega_0(W,\cdot)=0,\ \bar\omega_1|_W=0\}\cong LG(n_1,2n_1),$$

and \(F_{\omega_k}(\omega_\nu)=\varnothing\) for every other label \(\nu\).

  1. Defects. \(f_\nu=n_1(n_1+1)/2\). With \(\nu^{top}=\varphi(\omega_k)\), \(\epsilon=k\bmod2\), \(c_\nu=\epsilon+\langle\rho,\nu^{top}-\nu\rangle\) and \(\operatorname{def}(\nu)=n_1(\nu)\). The maximum \(\min(k,2m-k)\) is attained exactly at \(\nu=(1^k0^{m-k})\) for \(k\le m\), resp. \((2^{k-m}1^{2m-k})\) for \(k\ge m\). Hence \(\IC_{(1^k0^{m-k})}[-k]\), resp. \(\IC_{(2^{k-m}1^{2m-k})}[-(2m-k)]\), is a summand of \(V_{\omega_k}\ast\IC_0\); for \(k=m\), \(\IC_{(1^m)}=\sigma_*\IC_0\in\Csat\).
  2. Shift stability. \(\mathcal M+\Z(1^m)=\mathcal M\).
  3. Consequences. (a) For \(\GL_4/\Sp_4\), \(\mathcal M=\Lambda_S^+\), hence \(\Csat=\cC_K\) (given (DT) in \(\cC_K\)). (b) For \(\GL_{2m}/\Sp_{2m}\),
$$\mathcal M\supseteq\varphi(\Lambda_T^+)\cup\bigcup_{c\in\Z}\{(1^k0^{m-k})+c(1^m),\ (2^{k-m}1^{2m-k})+c(1^m)\};$$

for \(m=3\) this settles the uncovered labels \((a,a,a)\), \((a,a,a-1)\), \((a+1,a,a)\) with \(a\) odd, and leaves e.g. \((1,1,-1)\), \((3,1,1)\), \((1,1,-2)\) open.

  1. Fibre lemma. As in Computation 3.2, perfect is equivalent to integral (the quotient has length \(k\) and \(\val\Pf=k\)), and integrality on \(\Lambda_0+t^{-1}\widetilde W\) is the pair of conditions displayed. The kernel of \(\bar\omega_0\) is the sum \(V_1\oplus R\) of the 1-blocks and the 2-blocks, of dimension \(2n_1+2n_2\); on it \(\bar\omega_1\) is the standard form of \(V_1\) with radical \(R\). An \(\bar\omega_1\)-isotropic \(W\subseteq V_1\oplus R\) of dimension \(k=n_1+2n_2\) has \(\dim W/(W\cap R)\le n_1\), hence \(W\supseteq R\) and \(W/R\) is Lagrangian in \(V_1\cong k^{2n_1}\). Conversely every such \(W\) gives a perfect lattice: the orthogonal sum of \(t^{-1}\Lambda_0^{(j)}\) on the 2-blocks, \(\Lambda_0^{(j)}\) on the 0-blocks, and \(\Lambda_0^{(1)}+t^{-1}\widetilde W\) on the 1-blocks. If \(\omega\) is perfect on \(\Lambda'\) with \(\Lambda_0\subseteq\Lambda'\subseteq t^{-1}\Lambda_0\), dualizing gives \(\Lambda_0^\vee\subseteq t^{-2}\Lambda_0\), so every \(\nu_j\le2\). Brute force at \(q=2\), \(m=3\), \(k=2,3,4\): counts \(15,1;\ 135,3;\ 15,1\), equal to \(\prod_{i\le n_1}(1+2^i)\).
  1. Defects. \(f_\nu=\dim LG(n_1,2n_1)\). The top label \(\nu^{top}\) has \(n_1=\epsilon\), so the generic fibre has dimension \(\epsilon\) and \(D=\delta(\cO_{\nu^{top}})+\epsilon\). With the codimension formula (cited as Chen–Yi, Prop. 64), \(\langle\rho,\nu\rangle=4n_2k-4n_2^2+2k(m-k)\) for sorted \(\nu\), so
$$c_\nu=\epsilon+\langle\rho,\nu^{top}-\nu\rangle,\qquad\operatorname{def}(\nu)=n_1(n_1+1)-\epsilon-(k^2-\epsilon)+4n_2k-4n_2^2=n_1(\nu).$$

Since \(n_1=k-2n_2\) and \(n_2\ge\max(0,k-m)\), the maximum is \(\min(k,2m-k)\), attained only at the two labels listed. Apply Lemma 3.5 to the finite-type model of \(\Gr^{\omega_k}\mathbin{\tilde\times}L^+X\) over its image.

  1. Shift stability. \(\sigma\) is \(LG\)-equivariant, so \(\sigma_*(V_\lambda\ast\mathcal F)=V_\lambda\ast\sigma_*\mathcal F\). As \(\sigma_*\IC_0=\IC_{(1^m)}\) is a summand of \(V_{\omega_m}\ast\IC_0\), \(\sigma_*(V_\lambda\ast\IC_0)\) is a summand of \(V_\lambda\ast V_{\omega_m}\ast\IC_0\), a sum of Satake objects; by Krull–Schmidt, \(\mathcal M+(1^m)\subseteq\mathcal M\). As \(\sigma_*^{-2}=V_{(-1)^{2m}}\ast(-)\), also \(\mathcal M-(1^m)\subseteq\mathcal M\). (The shift is the determinant of \(\check G_X=\GL_m\).)
  1. Consequences. (a) For \(\GL_4\) the uncovered labels are exactly \((a,a)\) with \(a\) odd (Lemma of the symplectic family page), all in \((1,1)+\Z(1,1)\subseteq\mathcal M\); covered labels lie in \(\mathcal M\) by Proposition III, step 6 (the top summand). So \(\mathcal M=\Lambda_S^+\), and Proposition 3.4(2) gives \(\Csat=\cC_K\). (b) is the union of items 2 and 3; the list for \(m=3\) is checked directly.

Verified by computation: g2_hnf_lattices.py, g2_defect_coverage.py

Builds on

  • Lemma 3.1 · The loop space of \(\GL_{2m}/\Sp_{2m}\): orbits, closure order, scalar shift and codimensions
  • Computation 3.2 · Convolution fibres over \(x_{(2,0)}\) and \(x_{(1,1)}\) for \(\lambda=\omega_2\) and \(\lambda=(2,0,0,0)\)
  • Lemma 3.5 · Maximal-defect lemma: at a stratum of maximal defect the top stalk forces an IC summand

Built on by

  • Summary 3.11 · Verdict for the symplectic family: no disagreement between the levels; \(\GL_4/\Sp_4\) settled at both
  • Lemma 3.7 · \(\SL_{2n}/\Sp_{2n}\) is the degree-zero component of \(\GL_{2n}/\Sp_{2n}\); its smallest uncovered label stays open
  • Conjecture 3.8 · Relative-Satake prediction of every multiplicity; general uncovered labels and the exact gap
  • Theorem 3.9 · Explicit \(T_{\omega_k}\ast\varphi_0\) over every local field; (SG) for \(\GL_4/\Sp_4\) without Sakellaridis's hypotheses
  • Objection 7.11 · Objection to the codimension citation: Chen–Yi's Prop. 64 assumes \(LX\) connected; a one-line reduction is proposed
  • Audit 7.13 · Check of the summand criterion, the maximal-defect lemma, the minuscule fibre lemma and the \(\SL\)/\(\GL\) reduction
  • Proposition X1.4 · L(GL₂ₘ/Sp₂ₘ): codim = ⟨ρ,μ−ν⟩ in every component, char ≠ 2; σ reaches only m | s
  • Re-derivation X1.6 · The verdict inputs downstream of A1-4, A4-10, A5-6, A7-11: what changes
Original record, as the agent wrote it at 02:48 UTC

GL_2m/Sp_2m: the labels (1^k 0^{m-k}), (2^{k-m} 1^{2m-k}) and all their shifts lie in M; GL_4/Sp_4 complete

Fibre lemma (\(\lambda=\omega_k\), \(1\le k\le2m-1\)). For \(\nu\in\{0,1,2\}^m\) with \(|\nu|=k\), write \(n_i(\nu)=\#\{j:\nu_j=i\}\). Then $$F_{\omega_k}(\omega_\nu)=\{W\in\mathrm{Gr}(k,t^{-1}\Lambda_0/\Lambda_0):\ \bar\omega_0(W,\cdot)=0,\ \bar\omega_1|_W=0\}\cong LG(n_1,2n_1),$$ and \(F_{\omega_k}(\omega_\nu)=\varnothing\) for every other \(\nu\). Proof. As in A3-2, perfect \(\iff\) integral (index \(q^k\), \(v(\mathrm{Pf})=k\)). \(\ker\bar\omega_0=\) blocks with \(\nu_j\ge1\) (dimension \(2n_1+2n_2\)); \(\bar\omega_1\) is the standard form on the \(1\)-blocks with radical \(R=\) the \(2\)-blocks (\(\dim 2n_2\)). A \(k=(n_1+2n_2)\)-dimensional \(\bar\omega_1\)-isotropic \(W\subseteq\ker\bar\omega_0\) has \(\dim W/(W\cap R)\le n_1\), hence \(W\supseteq R\) and \(W/R\) Lagrangian in \(k^{2n_1}\); conversely such \(W\) give perfect lattices (orthogonal sum of \(t^{-1}\Lambda_0^{(j)}\) with \(t^2J_2\), the \(0\)-blocks, and \(\Lambda_0^{(1)}+\tilde W\) with \(t\omega_0\)). Supports: \(\omega\) perfect on \(\Lambda'\supseteq\Lambda_0\) of colength \(k\) with \(t\Lambda'\subseteq\Lambda_0\) forces \(\nu_j\in\{0,1,2\}\). Brute force at \(q=2\), \(m=3\), \(k=2,3,4\): counts \(15,1;135,3;15,1=\prod_{i\le n_1}(1+2^i)\). \(\square\)

Defects. \(f_\nu=n_1(n_1+1)/2\). The top orbit \(\nu^{top}=\varphi(\omega_k)\) has \(n_1=\epsilon:=k\bmod2\), so the generic fibre has dimension \(\epsilon\) and \(D=\delta(\mathcal O_{\nu^{top}})+\epsilon\). With Prop. 64, \(\langle\rho,\nu\rangle=4n_2k-4n_2^2+2k(m-k)\) for sorted \(\nu\), so $$c_\nu=\epsilon+\langle\rho,\nu^{top}-\nu\rangle,\qquad \mathrm{def}(\nu)=n_1(n_1+1)-\epsilon-(k^2-\epsilon)+4n_2k-4n_2^2=n_1(\nu).$$ The maximum \(n_1=\min(k,2m-k)\) is attained exactly at \(\nu=(1^k0^{m-k})\) (\(k\le m\)) resp. \((2^{k-m}1^{2m-k})\) (\(k\ge m\)). By A3-5, \(\mathrm{IC}_{(1^k0^{m-k})}[-k]\) resp. \(\mathrm{IC}_{(2^{k-m}1^{2m-k})}[-(2m-k)]\) is a summand of \(V_{\omega_k}\ast\mathrm{IC}_0\). For \(k=m\): \(\mathrm{IC}_{(1^m)}=\sigma_*\mathrm{IC}_0\in\mathcal C^{\mathrm{Sat}}\).

Shift stability. \(\sigma\) (A3-1.4) is \(LG\)-equivariant, so \(\sigma_*(V_\lambda\ast F)=V_\lambda\ast\sigma_*F\). Since \(\sigma_*\mathrm{IC}_0=\mathrm{IC}_{(1^m)}\) is a summand of \(V_{\omega_m}\ast\mathrm{IC}_0\), \(\sigma_*(V_\lambda\ast\mathrm{IC}_0)\) is a summand of \(V_\lambda\ast V_{\omega_m}\ast\mathrm{IC}_0\), a sum of Satake objects: \(\mathcal M+(1^m)\subseteq\mathcal M\). With \(\sigma_*^{-2}=V_{(-1)^{2m}}\ast(-)\): \(\mathcal M-(1^m)\subseteq\mathcal M\). Hence \(\mathcal M+\mathbb Z(1^m)=\mathcal M\) (the \(\det\) of \(\check G_X=\mathrm{GL}_m\)).

Consequences. (a) \(\mathrm{GL}_4/\mathrm{Sp}_4\): uncovered labels are exactly \((a,a)\), \(a\) odd (families/symplectic.html, Lemma), all in \((1,1)+\mathbb Z(1,1)\subseteq\mathcal M\); covered labels are in \(\mathcal M\) by Prop. III step 6 (the top summand). So \(\mathcal M=\Lambda_S^+\) and, by A3-4(2), \(\boxed{\mathcal C^{\mathrm{Sat}}=\mathcal C\text{ for }\mathrm{GL}_4/\mathrm{Sp}_4}\) (given (DT) in \(\mathcal C\)). (b) \(\mathrm{GL}_{2m}/\mathrm{Sp}_{2m}\): \(\mathcal M\supseteq\varphi(\Lambda_T^+)\cup\bigcup_{c\in\mathbb Z}\{(1^k0^{m-k})+c,\ (2^{k-m}1^{2m-k})+c\}\); for \(m=3\) this settles the uncovered labels \((a,a,a),(a,a,a-1),(a+1,a,a)\) (\(a\) odd) and leaves e.g. \((1,1,-1)\), \((3,1,1)\), \((1,1,-2)\) open.

Sources