§12 · Agent X2 · SL₂/SO₂ traces · 2026-10-17T20:22:44Z

Lemma X2.1 · \(\sigma_n\) is translation by \(h\in L^+\mathrm{SL}_2(\overline{\mathbb F}_q)\) with \(h^{-1}\mathrm{Fr}^n(h)=-I\)

proved correction

Correction. Corrects Theorem 4.11 · \(\SL_2/\SO_2\): the sign-twisted IC sheaves are not in the Satake span, so \(\Csat\ne\cC_K\) over \(\overline{\F}_q\).

In both \(\mathbb F_q\)-forms of \(\mathrm{SL}_2/\mathrm{SO}_2\), the automorphism \(\sigma_n\) is defined over \(\mathbb F_{q^n}\) by an element of \(\mathrm{PGL}_2(\mathcal O_n)\) and is geometrically the action of \(h=\operatorname{diag}(\sqrt u,\sqrt u^{-1})\in L^+\mathrm{SL}_2(\overline{\mathbb F}_q)\). Hence the trace of any equivariant Weil complex at \(\sigma_n(y)\) equals the trace at \(y\) twisted by the action of the central element \(-I\).

Lemma X2.1.

Let \(q\) be odd, \(k=\overline{\mathbb F}_q\), \(d\in\{1,-1\}\), and \(X_d=\{A\in\mathrm{Sym}_2:\det A=d\}\) with \(G=\mathrm{SL}_2\) acting by \(g\cdot A=gAg^{\mathsf T}\). Then \(X_1\) is the \(\mathrm{SO}_2\)-form (\(\theta(g)=(g^{\mathsf T})^{-1}\), base point \(I\), stabilizer \(\mathrm{SO}(x^2+y^2)\)), and \(X_{-1}\) is the split form \(\mathrm{SL}_2/T\) (base point \(H=\begin{pmatrix}0&1\\1&0\end{pmatrix}\), stabilizer the diagonal torus \(T\)). For \(n\ge1\) put \(\mathcal O_n=\mathbb F_{q^n}[\![t]\!]\), \(F_n=\mathbb F_{q^n}(\!(t)\!)\), \(K_n=\mathrm{SL}_2(\mathcal O_n)\). Fix a non-square \(u\in\mathcal O_n^\times\), put \(s=\operatorname{diag}(u,1)\), and $$\sigma_n(A)=u^{-1}\,sAs^{\mathsf T}.$$ (a) \(\sigma_n\) is an automorphism of \(LX_d\) defined over \(\mathbb F_{q^n}\), with \(\sigma_n(g\cdot A)=(sgs^{-1})\cdot\sigma_n(A)\). It preserves \(X_d(\mathcal O_n)\) and commutes with every Hecke operator, so \(\mathcal H_{K_n}\varphi_0\subseteq (C_{K_n})^{\sigma_n}\). On \(X_{-1}=\mathrm{SL}_2/T\) it is the map \(gT\mapsto sgs^{-1}T\) of Proposition J(2).

(b) Over \(k\), \(\sigma_n(A)=h\cdot A\) with \(h=\operatorname{diag}(\sqrt u,\sqrt u^{-1})\in T(k[\![t]\!])\subseteq L^+\mathrm{SL}_2(k)\), and \(h^{-1}\mathrm{Fr}^n(h)=-I\).

(c) For every \(\mu\ne0\), \(\sigma_n\) exchanges the two \(K_n\)-orbits of \(\mathcal O_\mu(\mathbb F_{q^n})\); it preserves \(\mathcal O_0(\mathbb F_{q^n})=K_n\cdot x_0\).

(d) Trace twist. Let \((M,\Phi)\) be an \(L^+G\)-equivariant Weil complex on \(LX_d\) with \(\mathbb F_{q^n}\)-structure \(\Phi^{(n)}\). For \(y\in LX_d(k)\) let \(\varepsilon_y\) be the action of \(-I\in\operatorname{Stab}_{L^+G}(y)\) on the stalks \(\mathcal H^i(M)_y\). Then \(\varepsilon_y\) commutes with \(\Phi^{(n)}_y\), and for every \(y\in LX_d(\mathbb F_{q^n})\) $$\operatorname{tr}_n(M)(\sigma_ny)=\sum_i(-1)^i\operatorname{tr}\big(\Phi^{(n)}_y\circ\varepsilon_y\ \big|\ \mathcal H^i(M)_y\big).$$

  1. (a), the map. \(\det(u^{-1}sAs^{\mathsf T})=u^{-2}u^2\det A\), and the formula has coefficients in \(\mathcal O_n\) with inverse of the same shape, so \(\sigma_n\) is an \(\mathbb F_{q^n}\)-automorphism of \(LX_d\) preserving \(L^+X_d\). One checks \(u^{-1}s(gAg^{\mathsf T})s^{\mathsf T}=(sgs^{-1})(u^{-1}sAs^{\mathsf T})(sgs^{-1})^{\mathsf T}\). In the split form, \(s^{-1}Hs^{-\mathsf T}=u^{-1}H\), so \(\sigma_n(gHg^{\mathsf T})=(sgs^{-1})H(sgs^{-1})^{\mathsf T}\): this is \(gT\mapsto sgs^{-1}T\). In the \(\mathrm{SO}_2\)-form, conjugation by \(s\) does not preserve \(\mathrm{SO}(x^2+y^2)\), and \(\sigma_n\) is the right replacement; it is the action of \([s]\in\mathrm{PGL}_2(\mathcal O_n)\) for \(g\cdot A=gAg^{\mathsf T}/\det g\).
  1. (a), Hecke operators. \(\operatorname{Ad}(s)\) preserves \(K_n\), the Haar measure with \(\operatorname{vol}K_n=1\), and each \(K_nt^{m\alpha^\vee}K_n=\{g\in\mathrm{SL}_2(F_n):\min\operatorname{val}(g_{ij})=-m\}\), because \(s^{\pm1}\) are integral. For \(K_n\)-biinvariant \(f\), substituting \(g=sg's^{-1}\) in \((f\ast\varphi)(\sigma_nA)=\sum_{g\in G/K_n}f(g)\varphi(g^{-1}\sigma_nA)\) and using step 1 gives \((f\ast\varphi)\circ\sigma_n=f\ast(\varphi\circ\sigma_n)\). Since \(\sigma_n(X_d(\mathcal O_n))=X_d(\mathcal O_n)\), \(\varphi_0\circ\sigma_n=\varphi_0\).
  1. (b). As \(p\) is odd, \(u\) has a square root in \(k[\![t]\!]\) (the constant term in \(\mathbb F_{q^{2n}}\), the rest by the binomial series). Then \(u^{-1}sAs^{\mathsf T}=hAh^{\mathsf T}\) entrywise: \((A_{11}u,\,A_{12},\,A_{22}u^{-1})\). Since \((\mathrm{Fr}^n\sqrt u)^2=u\) and \(\sqrt u\notin\mathcal O_n\), \(\mathrm{Fr}^n(\sqrt u)=-\sqrt u\), so \(\mathrm{Fr}^n(h)=-h\).
  1. (c). For \(y=gx_\mu\in\mathcal O_\mu(\mathbb F_{q^n})\), \(g\in L^+G(k)\), the \(K_n\)-orbit of \(y\) is the class of \(c(y)=\mathrm{Fr}^n(g)^{-1}g\) in \(H^1(\mathbb F_{q^n},\pi_0\operatorname{Stab}(x_\mu))\) (Lemma 4.2, Lang's theorem, D3D3 · Lang's theorem for the arc group L⁺G and its stabilizers over F_q.). For \(\mu\ne0\), \(\pi_0\operatorname{Stab}(x_\mu)=\{\pm I\}\), generated by the central \(-I\) (Computation 2.1 for \(X_1\); for \(X_{-1}\) through the isomorphism over \(k\), which carries \(-I\) to \(-I\)); Frobenius acts trivially on a group of order two, so \(H^1=\mathbb Z/2\). Now \(\sigma_n(y)=(hg)x_\mu\) and \(c(\sigma_ny)=\mathrm{Fr}^n(g)^{-1}(-h)^{-1}hg=(-I)\,c(y)\): the class flips. For \(\mu=0\) the stabilizer is connected and \(\mathcal O_0(\mathbb F_{q^n})\) is one \(K_n\)-orbit.
  1. (d), conventions. Work with the cohomology sheaves \(\mathcal F=\mathcal H^i(M)\), which are equivariant constructible Weil sheaves. Equivariance gives stalk isomorphisms \(\alpha_{g,y}:\mathcal F_{gy}\to\mathcal F_y\) for \(g\in L^+G(k)\), with the cocycle identity \(\alpha_{gg',y}=\alpha_{g',y}\circ\alpha_{g,g'y}\). The Weil structure \(\Phi_y:\mathcal F_{\mathrm{Fr}(y)}\to\mathcal F_y\) is a morphism of equivariant sheaves, i.e. \(\Phi_y\circ\alpha_{\mathrm{Fr}(g),\mathrm{Fr}(y)}=\alpha_{g,y}\circ\Phi_{gy}\). The same holds for the \(n\)-fold composite \(\Phi^{(n)}\) with \(\mathrm{Fr}^n\). Put \(\varepsilon_y=\alpha_{-I,y}\).
  1. (d), computation. Let \(y\) be \(\mathbb F_{q^n}\)-rational and \(z=hy=\sigma_n(y)\); then \(\mathrm{Fr}^n(h)=h\cdot(-I)\) and \(\mathrm{Fr}^n(z)=z\). The compatibility with \(g=h\) gives \(\Phi^{(n)}_y\circ\alpha_{h(-I),y}=\alpha_{h,y}\circ\Phi^{(n)}_z\), and the cocycle identity gives \(\alpha_{h(-I),y}=\alpha_{-I,y}\circ\alpha_{h,(-I)y}=\varepsilon_y\circ\alpha_{h,y}\). Hence
$$\Phi^{(n)}_z=\alpha_{h,y}^{-1}\circ\big(\Phi^{(n)}_y\circ\varepsilon_y\big)\circ\alpha_{h,y},$$

so the traces on \(\mathcal F_z\) and \(\mathcal F_y\) agree after inserting \(\varepsilon_y\). Taking the alternating sum over \(i\) gives the formula. With \(g=-I\), which is rational, the compatibility gives \(\Phi_y\circ\varepsilon_{\mathrm{Fr}(y)}=\varepsilon_y\circ\Phi_y\), so \(\varepsilon\) commutes with \(\Phi^{(n)}\). \(\square\)

Status. Proved, given the standing formalism D1D1 · Traces and t-structure: proper base change and the Grothendieck–Lefschetz trace formula on convolution fibres, with a Frobenius-compatible perverse t-structure whose simple objects are the IC sheaves. of Setup 4.1: equivariant Weil complexes whose stalks carry the equivariance isomorphisms. This is the same input as the Lang-torsor trace computations of Computation 2.8(b), Proposition 1.7 and Correction 10.8, step 5. Answer to starting point (1): \(\sigma_n\) is induced \(\mathbb F_{q^n}\)-rationally by \([s]\in\mathrm{PGL}_2(\mathcal O_n)\), a group normalizing \(K_n\) and permuting the stabilizers, and geometrically by \(h\in L^+\mathrm{SL}_2(k)\). If \(u\) is a constant non-square, then \(h\in T(\mathbb F_{q^{2n}})\) is even a constant loop. Only the cocycle \(h^{-1}\mathrm{Fr}^n(h)=-I\) is used.

Builds on

  • Theorem 4.11 · \(\SL_2/\SO_2\): the sign-twisted IC sheaves are not in the Satake span, so \(\Csat\ne\cC_K\) over \(\overline{\F}_q\)
  • Lemma 4.2 · \(K\)-orbits on \(X(\F_q(\!(t)\!))\) via Lang's theorem: \(\Lambda_S^+\) exactly when \(Z_{G^\theta}(S)\) is connected
  • Computation 2.1 · Stabilizers for \(\SL_2/\SO_2\): \(\mu_2\) times a pro-unipotent group
  • Proposition 1.2 · \(\SL_2/T\) versus \(\GL_2/T\): one variety, two Hecke algebras
  • Correction 10.8 · Correction: the square-class automorphism multiplies \(\tr_n\IC_{\chi,\nu}\) by \(\chi(-1)\)

Built on by

  • Lemma X2.2 · Weil inertia lemma: \(\operatorname{tr}_n\mathrm{IC}_{\chi,\nu}\circ\sigma_n=\chi(-1)\,\operatorname{tr}_n\mathrm{IC}_{\chi,\nu}\)
  • Theorem (repair of A4-11) X2.4 · Repair of A4-11: \(\mathrm{IC}_{\mathrm{sgn},\mu}\notin\mathcal C_K^{\mathrm{Sat}}\) in both \(\mathbb F_q\)-forms, \(q\) odd
  • Proposition X2.6 · (SG) fails for \(\mathrm{SL}_2/\mathrm{SO}(x^2+y^2)\) over every \(F\), \(q\) odd
  • Re-check V2 · Re-check by V2
  • Summary 9.6 · Verdict after the repairs: every input held or repaired; \(E_6/F_4\) still the one divergence

Sources