OctoInformBeta

OpenAI preprint, September 2026 · §4. Filtered local systems on surfaces

Lemma 4.4 [Lifting a filtered exact sequence]: Coordinate gradings specify retained lift entries and the additional entries selected by refinement

Lowering maps determine the filtered stabilizer.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Split coordinate filtrations on modules UU and QQ impose vanishing conditions on the entries of a matrix from QQ to UU. The library describes the corresponding stabilizer and the coordinate blocks used for lifting and refinement.

Let RR be a commutative ring, let α\alpha and β\beta be coordinate sets, and let a:α→ℕa\colon\alpha\to\mathbb N and b:β→ℕb\colon\beta\to\mathbb N be gradings. Put U=RαU=R^\alpha and Q=RβQ=R^\beta, with grades numbered from zero. For each k∈ℕk\in\mathbb N, the coordinate prefixes Uk⊆UU_k\subseteq U and Qk⊆QQ_k\subseteq Q are the submodules of vectors supported in grades strictly less than kk, defined by (1). Uk={u∈U:ui=0 whenever k≤a(i)},Qk={q∈Q:qj=0 whenever k≤b(j)}.\begin{aligned} U_k&=\{u\in U: u_i=0\text{ whenever }k\leq a(i)\},\\ Q_k&=\{q\in Q: q_j=0\text{ whenever }k\leq b(j)\}. \end{aligned}(1)✦ For finite β\beta, a matrix M∈Mat⁡α×β(R)M\in\operatorname{Mat}_{\alpha\times\beta}(R) acts by (Mq)i=∑j∈βMijqj(Mq)_i=\sum_{j\in\beta}M_{ij}q_j. Let f:ℕ→ℕf\colon\mathbb N\to\mathbb N be a function on the grades.

The Claim

Theorem 1. For the gradings aa and bb, the filtered matrix stabilizer S(a,b)S(a,b) is the submodule of Mat⁡α×β(R)\operatorname{Mat}_{\alpha\times\beta}(R) defined by (2). S(a,b)={M:Mij=0 whenever b(j)≤a(i)}.S(a,b)=\{M:M_{ij}=0\text{ whenever }b(j)\leq a(i)\}.(2)✦ Suppose that β\beta is finite, and let M∈Mat⁡α×β(R)M\in\operatorname{Mat}_{\alpha\times\beta}(R). Its stabilizer membership is characterized by (3). M∈S(a,b)⇔M(Qk+1)⊆Ukfor every k∈ℕ.M\in S(a,b) \quad\Longleftrightarrow\quad M(Q_{k+1})\subseteq U_k\quad\text{for every }k\in\mathbb N.(3)✦ For arbitrary coordinate sets α\alpha and β\beta, the set Eret(a,b)E_{\mathrm{ret}}(a,b) of retained lift entries is the subset of α×β\alpha\times\beta defined by (4). Eret(a,b)={(i,j)∈α×β:b(j)≤a(i)}.E_{\mathrm{ret}}(a,b) =\{(i,j)\in\alpha\times\beta:b(j)\leq a(i)\}.(4)✦ The set Eref(a,b;f)E_{\mathrm{ref}}(a,b;f) of refinement fiber entries is the subset of α×β\alpha\times\beta defined by (5). Eref(a,b;f)={(i,j)∈α×β:a(i)<b(j),f(a(i))=f(b(j))}.E_{\mathrm{ref}}(a,b;f) =\{(i,j)\in\alpha\times\beta:a(i)<b(j),\ f(a(i))=f(b(j))\}.(5)✦

In the paper, the result reads: for vector spaces U,QU,Q with flags (Ui),(Qi)(U_i),(Q_i) and exact sequences 0→Ui/Ui−1→Wi→Qi/Qi−1→00\to U_i/U_{i-1}\to W_i\to Q_i/Q_{i-1}\to0 for 1≤i≤m1\leq i\leq m, the flags (Vi)(V_i) on V=U⊕QV=U\oplus Q inducing those flags with compatible identifications Vi/Vi−1≃WiV_i/V_{i-1}\simeq W_i form a nonempty homogeneous space under H=Hom⁡(Q,U)H=\operatorname{Hom}(Q,U), with stabilizer K={h:h(Qi)⊆Ui−1 for every i}K=\{h:h(Q_i)\subseteq U_{i-1}\text{ for every }i\}; a further flag or direct-sum decomposition of each WiW_i may be transported to the lifted quotient, after which one forgets the coarse identification and retains the refined flags and fine quotient identifications; for compatible data, writing KcoarseK_{\mathrm{coarse}} and KintervalK_{\mathrm{interval}} for the stabilizers of the coarse and interval flags, the restriction H/Kcoarse→H/KintervalH/K_{\mathrm{coarse}}\to H/K_{\mathrm{interval}} is surjective with affine fiber Kinterval/KcoarseK_{\mathrm{interval}}/K_{\mathrm{coarse}}, both stabilizers being given by the same formula for their respective flags; these assertions hold over RR when all named graded modules are free, free splittings give free integral parameters for choices and fibers, and in fixed-rank algebraic families the maps are locally affine bundles.

What Is Checked

The constructions are over any commutative ring RR, with arbitrary coordinate sets and gradings and any function ff; the lowering characterization (3) assumes that the column coordinate set β\beta is finite.