Continuous closure of sheaves (Q715698)

In the paper under review, a purely algebraic construction of the continuous closure of any torsion-free coherent sheaf is given. It is shown that in characteristic zero taking continuous closure commutes with flat morphisms with seminormal fibers. In particular, the continuous closure of a coherent ideal sheaf is again a coherent ideal sheaf. The continuous closure also commutes with field extensions. Let \(I=(f_1,\dots, f_r)\subset k[z_1, \dots, z_n]\) be an ideal in a polynomial ring over a topological field \(k\). Then a polynomial \(g\) belongs to the continuous closure \(I^C\) of \(I\) if and only if there are continuous functions \(\phi_1,\dots, \phi_r\) such that \(g=\sum_i\phi_i g_i\). The set \(I^C\) is an ideal containing \(I\). This natural definition can be reformulated in the following way. Let \(X=\mathbb A^n=\mathrm{Spec}(k[z_1, \dots, z_n])\). Then the polynomials \(f_1, \dots, f_r\) define a morphism of locally free sheaves \(f: \mathcal O_X^r\to \mathcal O_X\), \((s_1, \dots, s_2)\mapsto \sum_i a_is_i\). It induces the homomorphism \(C^0(X(\mathbb C), \mathcal O_X^r)\to C^0(X(\mathbb C), \mathcal O_X)\) on the level of continuous sections as well. Notice that the image of \(f\) is the sheaf \(\mathcal I\) corresponding to the ideal \(I\subset k[z_1, \dots, z_n]\). In this setting we see that the continuous closure \(I^C\) of \(I\) is equal to the intersection \[ H^0(X, \mathcal O_X)\cap \mathrm{im}[C^0(X(\mathbb C), \mathcal O_X^r)\to C^0(X(\mathbb C), \mathcal O_X)]. \] Let \(S\) be a scheme. A descent problem over \(S\) is a compound object \(\mathbf{D}=(p: Y\to X, f: p^* E\to F)\), where p is a proper morphism of reduced \(S\)-schemes of finite type, \(E\) is a locally free sheaf on \(X\), \(F\) is a locally free sheaf on \(Y\), and \(f\) is a morphism of sheaves. One can define the continuous closure of the descent problem \(\mathbf{D}=(p: Y\to X, f: p^* E\to F)\) as \[ H^0(X, F)\cap \mathrm{im}[C^0(X(\mathbb C), E)\to C^0(X(\mathbb C), F)], \] which is a natural generalization of the initial definition. Let \(\mathbf{D}=(p: Y\to X, f: p^* E\to F)\) and \(\mathbf{D}_i=(p: Y\to X, f_i: p^*E\to F_i)\) be descent problems over \(S\). Then one can produce new descent problems using the following constructions. 1. For a proper morphism \(r: Y_1\to Y\) put \(r^*\mathbf{D}:= (p\circ r: Y_1\to X, r^*f: (p\circ r)^* E\to r^*F)\). 2. Put \(\bigoplus_{i=1}^m \mathbf{D}_i :=(p: Y\to X, \bigoplus_{i=1}^m f_i: p^* E\to \bigoplus_{i=1}^m F_i)\). 3. Assume that \(f\) factors as \(p^*E \mathop{\to}\limits^qF'\mathop{\hookrightarrow}\limits^j F\) such that \(F'\) is a locally free sheaf and \(\mathrm{rank}_y j=\mathrm{rank}_y F'\) for every \(y\) from some open dense subscheme \(Y_0\subset Y\). Define \(\mathbf{D}':=(p: Y\to X, f':=q: p^*E\to F')\). A scion of \(\mathbf{D}\) is a descent problem \(\mathbf{D}_s\) such that there exists a sequence of descent problems \(\mathbf{D}=\mathbf{D}_0\), \(\mathbf{D}_1,\dots, \mathbf{D}_s\), where \(\mathbf{D}_{i+1}\) is obtained from \(\mathbf{D}_0,\dots, \mathbf{D}_i\) using the three types of constructions described above. Each scion \(\mathbf{D}_s\) comes with a natural structure map \(Y_s\to Y\). The class of all scions of \(\mathbf{D}\) is denoted by \(\mathrm{Sci}(\mathbf{D})\). For a scheme \(X\) the author considers its seminormalization \(\pi: X^{\text{sn}}\to X\), which can be thought of as a partial normalization such that every morphism \(g: Y\to X\) induces a morphism of seminormalizations \(Y^{\text{sn}} \to X^{\text{sn}}\). For a coherent sheaf \(F\) on \(X\), let \(F^{\text{sn}}\) denote its pull-back to \(X^{\text{sn}}\). Let \(\mathbf{D}=(p: Y\to X, f: p^* E\to F)\) be a descent problem, with scions \[ \mathrm{Sci}(\mathbf{D})=\{ (p_i: Y_i\to X, f_i: p_i^* E\to F_i) \mid i\in I\}. \] An algebraic section of \(F\) over \(\mathrm{Sci}(\mathbf{D})\) is a collection of sections \(\Phi:=\{\phi_i\in H^0(Y_i^{\text{sn}}, F_i^{\text{sn}}) \mid i\in I\}\) commuting with 1) pull-backs for operations of type~1; 2) direct sums for operations of type~2; 3) push-forwards for operations of type~3. All such algebraic sections form an \(\mathcal O_S\)-module \(H^0(\mathrm{Sci}(\mathbf{D}), F)\). One calls \(\phi_i\) the restriction of \(\Phi\) to \(Y_i\). From the definition it follows that every section of \(F\) over \(\mathrm{Sci}(\mathbf{D})\) is uniquely defined by its restriction to \(Y\), and hence \(H^0(\mathrm{Sci}(\mathbf{D}), F)\) is an \(\mathcal O_S\)-submodule of \(H^0( Y^{\text{sn}}, F^{\text{sn}})\). Let now \(X\) be a reduced affine scheme over a field of characteristic zero and let \(J\) be a torsion-free coherent sheaf on \(X\). Realize \(J\) as the image of a map of locally free sheaves \(f: E\to F\). For \(Y=X\) and \(p=\mathrm{id}_X\), consider the descent problem \(\mathbf{D}_J=( \mathrm{id}_X : X \to X, f: E\to F)\). Define \(J^C:=H^0(\mathrm{Sci}(\mathbf{D}_J), F)\). The essence of the main result of the paper, Theorem~17, could be expressed as follows. \(J^C\) does not depend on the choice of \(f\). In the case when \(k\) a subfield of \(\mathbb C\) there is the equality \[ J^C=\mathrm{im}[C^0(X, E)\to C^0(X, F)]\cap H^0(X^{\text{sn}}, F^{\text{sn}}), \] which coincides with the natural definition of continuous closure in the case when \(X=X^{\text{sn}}\). Taking algebraic sections over scions commutes with base field extensions and with flat finite-type seminormal base changes. In particular for a descent problem \(\mathbf{D}=(p: Y\to X, f: p^* E\to F)\) the correspondence \[ U\mapsto H^0(\mathrm{Sci}(\mathbf{D})|_U, F|_U) \] defines a coherent sheaf on \(X\). For \(\mathbf{D}_J\) as above, one obtains a sheafification of the notion of continuous closure. The paper under review consists of an introduction and five sections. A short overview of the paper is given in the introduction. A formalism of descent problems and scions is introduced in the first section. Classes of continuous functions with nice properties are described in the second section. So called finitely determined descent problems, a special type of descent problems behaving well with respect to the functions from the second section, are introduced in the third section. The main theorem and its consequences are considered in the fourth section. Here the main result of the paper, Theorem~17, is stated. The meaning of this rather technically looking result is explained by means of important corollaries. The proof of Theorem~17 is presented in the last section of the paper.