A sufficient criterion
How should “a combinatorial τ implies Iwahori generation” be formulated?
1 · A finite-level criterion for functions
· The data to check Definition
Let \(\Grp\) be a split connected reductive group over a non-Archimedean local field \(\Loc\), and \(\Iw\) an Iwahori subgroup. Let \(X(\Loc)\) be a locally compact left \(\Grp(\Loc)\)-space, \(\Module=C_c^\infty(X(\Loc))^{\Iw}\), and \(\basic\in\Module\) a chosen basic function. Let \(\Gamma\subseteq N_{\Grp(\Loc)}(\Iw)/\Iw\) be a subgroup. Suppose
$$\Module=\varinjlim_{\nu}\LevelModule{\nu},\qquad \dim\LevelModule{\nu}<\infty.$$Each \(\LevelModule{\nu}\) has a basis \(e_{\nu,1},\ldots,e_{\nu,d_{\nu}}\) of indicators of finite-level strata. The ordered labels include every terminal stratum.
· Translate coverage plus triangularity Proved here
For each level \(\nu\), assume there are \(\gamma_{\nu,1},\ldots,\gamma_{\nu,d_{\nu}}\in\Gamma\) whose translates lie in \(\LevelModule{\nu}\) and satisfy
$$T_{\gamma_{\nu,j}}*\basic=\sum_{i=1}^{d_{\nu}}A_{\nu,ij}e_{\nu,i}.$$ $$A_{\nu,ij}=0\quad(i<j),\qquad A_{\nu,jj}\ne0.$$Then
$$\boxed{\mathbb C[\Gamma]*\basic=\Module,\qquad \Hecke*\basic=\Module.}$$Proof
- Normalizer representatives make the Hecke operator an ordinary translation: $$T_{\gamma}*\basic(x)=\basic(\gamma^{-1}x).$$
- The finite matrix is invertible: $$\det A_{\nu}=\prod_{j=1}^{d_{\nu}}A_{\nu,jj}\ne0.$$
- Every level, hence every function, lies in the generated module: $$\LevelModule{\nu}=\operatorname{span}_j\{T_{\gamma_{\nu,j}}*\basic\}\subseteq\mathbb C[\Gamma]*\basic.$$
Proved here This is an elementary sufficient criterion, not a classification from the literature. The affine-space calculation satisfies it with \(\Gamma=\langle\shift\rangle\) and the nested-support matrix.
2 · The extra data for sheaves
· Translation reaches every simple Proved here
Let \(\Category\) be a triangulated category with an action of the geometric Iwahori–Hecke category and a chosen object \(\BasicSheaf\). Assume:
- Every object comes through an exact fully faithful functor from a bounded finite-level category whose perverse heart has finite length; the level functors preserve the stated objects and triangles.
- For every simple object \(\Simple\) at every level, there are \(\gamma\in\Gamma\) and \(d\in\mathbb Z\) such that
Here \(\delta_\gamma\) is the point kernel on the normalizer coset, and \(\gamma_!:\Category\to\Category\) is translation. Then
$$\boxed{\Category=\operatorname{Thick}\{\delta_\gamma*_I\BasicSheaf:\gamma\in\Gamma\}.}$$Proof
- Translate coverage places every simple in the proposed thick subcategory: $$\Simple[-d]\cong\delta_\gamma*_I\BasicSheaf.$$
- A Jordan–Hölder filtration builds each finite-length perverse object from these simples by extensions, hence triangles.
- Bounded perverse truncation triangles build each finite-level complex from its finitely many perverse cohomology objects: $$ {}^p\tau_{\le j-1}K\longrightarrow{}^p\tau_{\le j}K\longrightarrow{}^pH^j(K)[-j]\xrightarrow{+1}.$$
- Every object belongs to some level, so the finite-level conclusion gives the asserted equality.
Orbit supports alone do not verify this hypothesis: disconnected stabilizers may supply additional equivariant local systems. The affine-space proof checks the connected stabilizers and all simple objects explicitly in Proposition 4a.
3 · Why existence alone is insufficient
· A smooth affine symmetric counterexample Proved here
Put \(\BaseGroup=\mathrm{GL}_2\), take its upper-triangular Iwahori \(\BaseIwahori\), and let \(\MaxCompact=\mathrm{GL}_2(\Ring)\). The smooth affine variety \(X=\BaseGroup\) is spherical and symmetric for \(\Grp=\BaseGroup\times\BaseGroup\), acting by \((g_1,g_2)x=g_1xg_2^{-1}\); its stabilizer at the identity is the diagonal subgroup. Set
$$\GroupHecke=C_c(\BaseIwahori\backslash\BaseGroup(\Loc)/\BaseIwahori).$$For the ordinary finite simple reflection \(s\) of GL₂, the basic function on this integral model is
$$\basic=\mathbf1_{\MaxCompact}=T_e+T_s.$$Both factors have the length-zero shifts of the model, but the generated module is the proper two-sided ideal
$$\Hecke*\basic=\GroupHecke(T_e+T_s)\GroupHecke.$$ $$\boxed{\GroupHecke(T_e+T_s)\GroupHecke\subsetneq\GroupHecke.}$$ $$\GroupHecke=C_c^\infty(X(\Loc))^{\BaseIwahori\times\BaseIwahori}.$$Proof
- The finite Bruhat decomposition of GL₂ over the residue field has two cells, so
- The Iwahori–Hecke presentation gives a unital algebra character \(\signchar:\GroupHecke\to\mathbb C\): send every affine simple generator to \(-1\), and every length-zero generator to \(1\).
The quadratic relations vanish at −1, the braid relations hold for identical scalars, and conjugation by Ω preserves these values. [Sht, Corollary 2.33]
- Left and right Hecke operators produce exactly the displayed two-sided ideal: inversion in the second factor changes the right-convolution kernel and permutes the full Hecke algebra. This ideal is killed by the character, while the unit is not:
Therefore \(T_e=\mathbf1_{\BaseIwahori}\) cannot lie in the generated module.
Question For a specified class of smooth affine spherical G-varieties, can the orbit-closure order, local-system data and normalizer action be used to recognize the translate-coverage hypothesis? Is there a broader criterion allowing simple-reflection operators when Ω-translates do not suffice?
The criterion above is sufficient. A characterization of that class is the research question; it is not asserted here.