§5 · Agent 5 · 𝔸ⁿ: lattices · 2026-10-18T18:43:27Z
Proposition 5.7 · (II$_I$) (IG) over $\mathbb F_q((t))$ puts every simple object in $\mathcal M_I$
proved ★ key step
Taking traces sends the generators $\mathrm{IC}_w\ast_I\mathrm{IC}_0$ into the span of the traces of the simple objects in $\mathcal M_I$; if (IG) makes that span all of $C_I$, linear independence of the traces of simple objects forces every simple object into $\mathcal M_I$, and (I$_I$) gives generation for sheaves over $\overline{\mathbb F}_q$. Only span containment is used, so cancellation in the traces cannot spoil the argument.
Proposition 5.7.
Sheaves over $k=\overline{\mathbb F}q$ ($p$ odd, $\overline{\mathbb Q}\ell$ coefficients), functions over $F=\mathbb F_q((t))$, $\operatorname{vol}I=1$. Assume (L), (Hk) of Proposition 5.4 and
- (L$q$) the level presentation and the actions are defined over $\mathbb F_q$, so $LX(\mathbb F_q)=X(F)$, $I(\mathbb F_q)=I_F$, and $\pi{N,M}$ on $\mathbb F_q$-points is the reduction map;
- (W) every $S\in\operatorname{Simple}(\mathcal P_I)$ is Frobenius-stable; fix a Weil structure $S_0$;
- (P) each $\mathrm{IC}_w\ast_I\mathrm{IC}_0$, with the Weil structure induced from the canonical ones, is pure;
- (Lin) the functions $\operatorname{tr}(S_0)$, $S\in\operatorname{Simple}(\mathcal P_I)$, are linearly independent in $C_I\otimes\overline{\mathbb Q}_\ell$;
- (Tr) $\operatorname{tr}(\mathrm{IC}_0)=\varphi_0$; $\operatorname{tr}(\mathcal G\ast_IA)=\operatorname{tr}(\mathcal G)\ast\operatorname{tr}(A)$ for Weil objects; $\operatorname{tr}$ is additive on triangles of Weil objects and $\operatorname{tr}(A[1])=-\operatorname{tr}(A)$.
If $\mathcal H_I\ast\varphi_0=C_I$ (IG), then $\mathcal M_I=\operatorname{Simple}(\mathcal P_I)$, hence $$\boxed{\mathcal C_I^{\mathrm{Iwa}}=\mathcal C_I\ \text{ over }\overline{\mathbb F}_q .}$$ Without (W): (IG) over $\mathbb F_q((t))$ puts every Frobenius-stable simple object in $\mathcal M_I$, and (IG) over all $\mathbb F_{q^r}((t))$ puts every simple object in $\mathcal M_I$.
Proof.
- Coefficients. $C_I$ has a $\mathbb Z$-basis of orbit indicators and the $T_w\ast\varphi_0$ are $\mathbb Z$-valued; a rational vector lies in the $\mathbb C$-span of finitely many rational vectors iff it lies in their $\mathbb Q$-span. So (IG) over $\mathbb C$ is equivalent to (IG) over $\overline{\mathbb Q}_\ell$.
- The IC basis of $\mathcal H_I$. $\mathrm{IC}w|{IwI/I}=\overline{\mathbb Q}\ell[\ell(w)]$ and $\operatorname{supp}\mathrm{IC}_w=\bigsqcup{y\le w}IyI/I$, so $\operatorname{tr}\mathrm{IC}w=(-1)^{\ell(w)}T_w+\sum{y<w}c_{y,w}T_y$ (Lang's theorem for the connected group $I$ identifies the $\mathbb F_q$-points of the cells with the double cosets). By unitriangularity, $\operatorname{span}\{\operatorname{tr}\mathrm{IC}w\}=\mathcal H_I\otimes\overline{\mathbb Q}\ell$, and by (Tr)
- Trace of a generator. By (P) and BBD 5.4.5, over $k$ $\mathrm{IC}w\ast\mathrm{IC}_0\simeq\bigoplus_iP_i[-i]$, $P_i={}^pH^i$ pure perverse; by BBD 5.3.8 each $P_i\simeq\bigoplus_SS\otimes E{S,i}$ over $k$. Frobenius maps the $S$-isotypic part to the $\mathrm{Fr}^*S$-isotypic part. For $\mathrm{Fr}^*S\simeq S$ it acts on $E_{S,i}=\operatorname{Hom}(S_0,P_i)$ by some $A_{S,i}$ and contributes $\operatorname{tr}(A_{S,i})\operatorname{tr}(S_0)$; a Frobenius orbit of length $>1$ is permuted off the diagonal and contributes $0$ at every $\mathbb F_q$-point. Hence
and $m_{S,w}\ne0$ only if $S$ is a shifted summand of $\mathrm{IC}_w\ast\mathrm{IC}_0$, i.e. $S\in\mathcal M_I$.
- Conclusion. Under (IG), steps 1–3 give $C_I\otimes\overline{\mathbb Q}\ell\subseteq\operatorname{span}\{\operatorname{tr}S_0:S\in\mathcal M_I\text{ Frobenius-stable}\}$. A Frobenius-stable simple $S'$ is pulled back from a finite level and supported on some $Y_N$, so $\operatorname{tr}(S'_0)\in C_I\otimes\overline{\mathbb Q}\ell$; by (Lin) it can lie in that span only if $S'\in\mathcal M_I$. Under (W) all simples are Frobenius-stable, so $\mathcal M_I=\operatorname{Simple}(\mathcal P_I)$; (P) gives (DT)(DT) · Decomposition: every generator IC_w ∗_I IC_0 is, equivariantly, a finite direct sum of shifted simple objects of C_I. Holds over F̄_q from a finite-type proper model of the convolution (purity and the stack decomposition theorem)., and Proposition 5.4(c) gives the box. A simple object that is not Frobenius-stable becomes stable over some $\mathbb F_{q^r}$ (finitely many simples per level), and the same argument applies over $\mathbb F_{q^r}((t))$.
- Cancellation (checklist 6). Step 4 needs only the containment of spans. Cancellation inside $m_{S,w}$ can only make $\operatorname{span}\{\operatorname{tr}(\mathrm{IC}w\ast\mathrm{IC}_0)\}$ smaller, and (IG) says it is everything. The converse direction (sheaves ⇒ functions) is the one exposed: a summand $S\in\mathcal M_I$ with $m{S,w}=0$ for all $w$ would leave $\operatorname{tr}S_0$ outside the span, and so would generation that needs retracts.
Sufficient conditions. (W) and (Lin) hold when every $H_\alpha$-orbit on $Y_\alpha$ is defined over $\mathbb F_q$ with connected stabilizer: by Lang's theorem $o(\mathbb F_q)$ is a single nonempty $H_\alpha(\mathbb F_q)$-orbit, $\operatorname{tr}\mathrm{IC}(\bar o)0=\pm\mathbf 1{o}+(\text{terms on smaller orbits})$ at a common level, and unitriangularity gives independence. (P) holds when the convolution has the proper finite-type model of Proposition 5.4 (BBD 5.1.14 and Gabber's purity of $\mathrm{IC}_w$). (Tr) is proper base change plus the Grothendieck–Lefschetz trace formula on that model, with $H^1(\mathbb F_q,I)=1$. Field. The conclusion is over $\overline{\mathbb F}_q$; reaching the Betti category over $\mathbb C$ needs a separate comparison of summand statements (spreading out, BBD §6.1), not supplied here.
Builds on
- Proposition 5.4 · (I$_I$) The Iwahori summand criterion: generation iff $\mathcal M_I$ is all
Built on by
- Corollary 5.8 · Reverse route for $\mathbb A^n$: (IG) and (II$_I$) give sheaf generation
- Summary 5.9 · Verdict: for $\mathbb A^n$ both Iwahori-level statements hold
- Counterexample 7.3 · Objection to A5-7 step 3 and A5-4 (DT over $\mathbb C$): BBD 5.4.5 splits only non-equivariantly
- Lemma 7.5 · Hypotheses (I1I1 · Finite levels: LX has a level presentation in which every object of C_I is pulled back from a finite level Y_{N,M} and each level has finitely many pieces (Iwahori-generation pages, Proposition III).), (I2I2 · Connected stabilizers: every stabilizer of I/I_{N,M} on Y_{N,M} is connected, so equivariant local systems on pieces are constant.), Frobenius-stability for $\mathbb A^n$ and the group; $I_{N,M}$ must be connected for Lang
- Audit 7.9 · Audit of A5-7: (IG) over $\mathbb F_q((t))$ $\Rightarrow\mathcal M_I=\mathrm{Simple}(\mathcal P_I)$ under (L), (Hk), (L$_q$), (W), (P), (Lin), (Tr)
- Strategy 10.1 · Registry after round one: every target has a candidate proof; three adversarial audits are launched
Original record, as the agent wrote it at 18:43 UTC
(II$_I$) (IG) over $\mathbb F_q((t))$ puts every simple object in $\mathcal M_I$
Proof.
- Coefficients. $C_I$ has a $\mathbb Z$-basis of orbit indicators and the $T_w\ast\varphi_0$ are $\mathbb Z$-valued; a rational vector lies in the $\mathbb C$-span of finitely many rational vectors iff it lies in their $\mathbb Q$-span. So (IG) over $\mathbb C$ is equivalent to (IG) over $\overline{\mathbb Q}_\ell$.
- The IC basis of $\mathcal H_I$. $\mathrm{IC}w|{IwI/I}=\overline{\mathbb Q}\ell[\ell(w)]$ and $\operatorname{supp}\mathrm{IC}_w=\bigsqcup{y\le w}IyI/I$, so $\operatorname{tr}\mathrm{IC}w=(-1)^{\ell(w)}T_w+\sum{y<w}c_{y,w}T_y$ (Lang's theorem for the connected group $I$ identifies the $\mathbb F_q$-points of the cells with the double cosets). By unitriangularity, $\operatorname{span}\{\operatorname{tr}\mathrm{IC}w\}=\mathcal H_I\otimes\overline{\mathbb Q}\ell$, and by (Tr)
- Trace of a generator. By (P) and BBD 5.4.5, over $k$ $\mathrm{IC}w\ast\mathrm{IC}_0\simeq\bigoplus_iP_i[-i]$, $P_i={}^pH^i$ pure perverse; by BBD 5.3.8 each $P_i\simeq\bigoplus_SS\otimes E{S,i}$ over $k$. Frobenius maps the $S$-isotypic part to the $\mathrm{Fr}^*S$-isotypic part. For $\mathrm{Fr}^*S\simeq S$ it acts on $E_{S,i}=\operatorname{Hom}(S_0,P_i)$ by some $A_{S,i}$ and contributes $\operatorname{tr}(A_{S,i})\operatorname{tr}(S_0)$; a Frobenius orbit of length $>1$ is permuted off the diagonal and contributes $0$ at every $\mathbb F_q$-point. Hence
and $m_{S,w}\ne0$ only if $S$ is a shifted summand of $\mathrm{IC}_w\ast\mathrm{IC}_0$, i.e. $S\in\mathcal M_I$.
- Conclusion. Under (IG), steps 1–3 give $C_I\otimes\overline{\mathbb Q}\ell\subseteq\operatorname{span}\{\operatorname{tr}S_0:S\in\mathcal M_I\text{ Frobenius-stable}\}$. A Frobenius-stable simple $S'$ is pulled back from a finite level and supported on some $Y_N$, so $\operatorname{tr}(S'_0)\in C_I\otimes\overline{\mathbb Q}\ell$; by (Lin) it can lie in that span only if $S'\in\mathcal M_I$. Under (W) all simples are Frobenius-stable, so $\mathcal M_I=\operatorname{Simple}(\mathcal P_I)$; (P) gives (DT), and Proposition 5.4(c) gives the box. A simple object that is not Frobenius-stable becomes stable over some $\mathbb F_{q^r}$ (finitely many simples per level), and the same argument applies over $\mathbb F_{q^r}((t))$.
- Cancellation (checklist 6). Step 4 needs only the containment of spans. Cancellation inside $m_{S,w}$ can only make $\operatorname{span}\{\operatorname{tr}(\mathrm{IC}w\ast\mathrm{IC}_0)\}$ smaller, and (IG) says it is everything. The converse direction (sheaves ⇒ functions) is the one exposed: a summand $S\in\mathcal M_I$ with $m{S,w}=0$ for all $w$ would leave $\operatorname{tr}S_0$ outside the span, and so would generation that needs retracts.
Sufficient conditions. (W) and (Lin) hold when every $H_\alpha$-orbit on $Y_\alpha$ is defined over $\mathbb F_q$ with connected stabilizer: by Lang's theorem $o(\mathbb F_q)$ is a single nonempty $H_\alpha(\mathbb F_q)$-orbit, $\operatorname{tr}\mathrm{IC}(\bar o)0=\pm\mathbf 1{o}+(\text{terms on smaller orbits})$ at a common level, and unitriangularity gives independence. (P) holds when the convolution has the proper finite-type model of Proposition 5.4 (BBD 5.1.14 and Gabber's purity of $\mathrm{IC}_w$). (Tr) is proper base change plus the Grothendieck–Lefschetz trace formula on that model, with $H^1(\mathbb F_q,I)=1$. Field. The conclusion is over $\overline{\mathbb F}_q$; reaching the Betti category over $\mathbb C$ needs a separate comparison of summand statements (spreading out, BBD §6.1), not supplied here.
Sources
- trace.html (satake-two-levels Proposition II, the spherical model)
- bbd-perverse.pdf, 5.1.14, 5.3.8, 5.4.5, §6.1
- laumon-1987.pdf (sheaf–function dictionary)
- generation.html (the sheaf–function dictionary)