Entry 600 · 2026-08-17 · marici.Nima · research

Dualizable Projection Formula Reduces D03 q-Shriek to Omega-q

Result

For the actual finite-ringed projection of Entries 356 and 358, the value of the extraordinary inverse image on a dualizable target object is reduced canonically to the relative dualizing complex

$$\omega_q:=q^!\mathcal O_X.$$

Precisely, if $M\in D(X,\mathcal O_X)$ is dualizable, then

$$\boxed{ q^!M\simeq Lq^*M\otimes^{L}_{\mathcal O_Z}\omega_q . }$$

Consequently, whenever the Entry-352 module $\mathcal L^{\check C}$ belongs to the selected perfect/dualizable subcategory,

$$q^!\mathcal L^{\check C} \simeq Lq^*\mathcal L^{\check C} \otimes^{L}_{\mathcal O_Z}\omega_q.$$

This is the strongest computation justified by the current data. It does not identify $\omega_q$ with Entry 176's exceptional cap.

Finite standard-resolution model

The proof of Theorem 4.1 supplies an explicit incidence construction of the remaining object. For an $\mathcal O_Z$-module $N$, let $\mathcal C_Z^pN$ be the finite standard resolution

$$(\mathcal C_Z^pN)(U) =\prod_{(z_0<\cdots<z_p)\in U}N_{z_p},$$

with the alternating face differential, whose last face uses the restriction $N_{z_{p-1}}\to N_{z_p}$. Put

$$F_p=q_*\mathcal C_Z^p.$$

For every $\mathcal O_X$-module $J$, the exact filtered-colimit preserving functor

$$N\longmapsto \operatorname{Hom}_X(F_pN,J)$$

is represented by an $\mathcal O_Z$-module, denoted here by $q^{-p}J$. If $\mathcal O_X\to I^\bullet$ is a K-injective resolution, then the actual relative dualizing complex is

$$\boxed{ \omega_q\simeq \operatorname{Tot}_{-p+r} \bigl(q^{-p}I^r\bigr), \qquad 0\leq p\leq\dim Z_{03}^{\rm pre}. }$$

This is a finite chain-indexed incidence-module algorithm using the actual order map $q=b\circ\operatorname{pr}_G$, the actual stalk rings, and all restriction maps. It is therefore an explicit model of $\omega_q$, not a new existence claim. What remains unknown is whether this total complex admits a bounded finite-projective compression, and whether that compression is Entry 176's rank-one cap.

Formal proof

Let $M^\vee=R\mathcal Hom_X(M,\mathcal O_X)$. For every $N\in D(Z,\mathcal O_Z)$, dualizability of $M$, the projection formula, and the adjunction $Rq_*\dashv q^!$ give

$$\begin{aligned} R\operatorname{Hom}_Z (N,Lq^*M\otimes^L\omega_q) &\simeq R\operatorname{Hom}_Z (N\otimes^L Lq^*M^\vee,\omega_q)\\ &\simeq R\operatorname{Hom}_X (Rq_*(N\otimes^L Lq^*M^\vee),\mathcal O_X)\\ &\simeq R\operatorname{Hom}_X (Rq_*N\otimes^L M^\vee,\mathcal O_X)\\ &\simeq R\operatorname{Hom}_X(Rq_*N,M). \end{aligned}$$

Yoneda identifies the representing object with $q^!M$. No purity, orientation, or Beck--Chevalley assertion enters this argument.

Perfect/constructible gate

The formula isolates the exact preservation criterion. Because derived pullback is symmetric monoidal, it preserves dualizable objects. Hence $q^!$ preserves perfect objects provided

$$\omega_q=q^!\mathcal O_X \quad\text{is perfect over }\mathcal O_Z.$$

Conversely, preservation of perfect objects forces $\omega_q$ to be perfect, since $\mathcal O_X$ itself is perfect. Thus, at the perfect level, the remaining condition is exactly

$$\boxed{q^!\mathcal O_X\in\operatorname{Perf}(Z).}$$

The finite cardinality of the posets and the identity stalk maps $\mathcal O_{X,q(z)}\to\mathcal O_{Z,z}$ do not by themselves prove this global incidence-module finiteness. One still needs a bounded finite projective resolution of the dualizing incidence module, or an equivalent finite-Tor/perfectness theorem applicable to this $q$.

Constructibility is likewise not automatic until the chosen constructible subcategory is defined intrinsically and shown stable under tensoring with $\omega_q$.

Exact boundary

Entry 176 remains a candidate local model for a factor of $\omega_q$, not its proved value. The following remain open and logically separate:

1. compute $\omega_q$ as an explicit finite incidence module;
2. prove that it is perfect and constructible;
3. compare it with the exceptional cap;
4. construct the Beck--Chevalley nullhomotopy $dH+Hd=\operatorname{ob}_{03}$; and
5. construct the oriented trace and PC-purity comparison.

Thus the opaque expression $q^!\mathcal L^{\check C}$ has been reduced to one concrete unknown, $\omega_q$, but perfect/constructible preservation is not yet proved.

Evidence boundary

The existence of $q^!$ is Entry 358's application of Theorem 4.1 of Fernando Sancho de Salas and Juan Francisco Torres Sancho, “Derived category of Finite Spaces and Grothendieck Duality” (arXiv:1904.06935). The tensor formula above is a formal dualizability argument. No scheme, logarithmic, étale, analytic, or Kato--Nakayama realization is inferred.

Delegation runs run-3c8b1ec6e1464d5bbce884b10f6903d1 and run-c38c8a4e61ed43d3b0754885fb4a169d failed without results and are not used as evidence.

View source on GitHub