Adelic Descent: A Local-to-Global Approach

• Adelic descent is a local-to-global principle that reconstructs global structures by gluing compatible local data from adeles, completions, and localizations.
• It underpins reconstruction theorems for perfect complexes on Noetherian schemes and supports cubical descent for localizing invariants like algebraic K-theory.
• The framework extends to arithmetic and cohomology applications, offering practical insights into torsor-lifting, descent sets, and derived categorical enhancements.

Adelic descent is a local-to-global principle in which a global object is reconstructed from compatible data attached to adeles, adelic rings, or adelic local conditions. In current usage, the term covers several related but distinct frameworks: Beilinson-style co-simplicial adeles for Noetherian schemes and their descent theorems for perfect complexes; descent for localizing invariants such as Bass–Thomason $K$-theory; descent sets for adelic points under torsors and ramified covers; and a number of analogical extensions in arithmetic and cohomology. The unifying feature is that localizations and completions along places, points, or chains of specializations are assembled into a global object by a compatibility condition, totalization, homotopy limit, or global cohomology class (Groechenig, 2015, Kim, 2021, Demeio, 2021).

1. Adeles, flags, and the basic local-to-global mechanism

A classical prototype is Weil’s adelic description of bundles on a smooth projective curve. If $F$ is the function field, $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$, and

$\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$

then Weil’s theorem identifies $G$-bundles with the double quotient

$$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$

This is the archetypal adelic description: global geometric objects are recovered by gluing local data encoded in adeles (Groechenig, 2015).

For arbitrary Noetherian schemes, the ambient combinatorics is supplied by chains of points. One uses the specialization order

$x\le y \quad \text{if } x\in \overline{\{y\}},$

and the simplicial set

$$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$

Beilinson’s construction associates to these chains a co-simplicial ring of adeles $\mathbb A_X^\bullet$, while the reduced variant $A^\bullet_{\mathrm{red}(X)}$ is built from strictly increasing chains of points. These adelic diagrams are not presented as descent for a Grothendieck topology; rather, they are specific local-to-global resolutions attached to the stratification of $F$0 by chains of specializations (Kim, 2021).

The same pattern appears in the more abstract notion of adelic cohomology for a poset $F$1 with coefficient system $F$2 and localization data $F$3. The adelic cochains are defined by

$F$4

and the cohomology

$F$5

measures whether the local pieces glue globally. For a catenary Noetherian commutative ring $F$6, one of the main reconstruction statements is

$F$7

and the associated adelic cube is a homotopy pullback (Greenlees, 2019).

A persistent misconception is that adelic descent is a single theorem with a single formalism. The literature instead treats a family of related mechanisms: co-simplicial descent for perfect complexes, cubical descent for invariants, torsor-lifting conditions on adelic points, and more analogical uses in which “descent” denotes passage from local or microscopic data to a global adelic object.

2. Perfect complexes and reconstruction of Noetherian schemes

The central scheme-theoretic theorem is the adelic descent equivalence

$F$8

for every Noetherian scheme $F$9. Here $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$0 is the totalization of the co-simplicial symmetric monoidal $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$1-category obtained by applying $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$2 degreewise to $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$3. Objects on the right are cartesian co-simplicial perfect complexes, i.e. perfect adelic data satisfying the full homotopy-coherent compatibility relations (Groechenig, 2015).

The proof strategy proceeds through an adelic realization functor

$$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$4

its fully faithfulness, the construction of a right adjoint $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$5, and the verification that $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$6 is conservative and preserves perfection on adelic-perfect objects. A crucial technical input is that Beilinson’s adelic sheaves are not merely flasque but l^ache, so that for a l^ache sheaf of algebras $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$7 on a quasi-compact space one has

$$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$8

as symmetric monoidal $$\mathbb O_X=\prod_{x\in X_{cl}}\widehat{\mathcal O}_x$$9-categories (Groechenig, 2015).

The same framework yields cohomological reconstruction. For any quasi-coherent sheaf $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$0,

$\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$1

is a co-simplicial resolution by flasque sheaves, and

$\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$2

This is Beilinson’s cohomological adelic resolution (Groechenig, 2015).

A further consequence is reconstruction of the scheme itself: $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$3 Using the Tannakian formalism of Bhatt and Bhatt–Halpern-Leistner, the same reasoning extends to stacks with quasi-affine diagonal: $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$4 In group-valued cases this yields adelic cocycle descriptions such as

$\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$5

This recovers Weil’s theorem for curves as a special case (Groechenig, 2015).

An important restriction is that the theorem is formulated for perfect complexes. The analogous statement for all quasi-coherent complexes is explicitly noted to be false in general, and the good behavior over adelic product rings relies on perfectness and on Bhatt’s theorem describing perfect complexes over infinite products of rings (Groechenig, 2015).

3. Descent for localizing invariants, $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$6-theory, and categorical refinements

For a Noetherian scheme $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$7 of finite Krull dimension $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$8, Kim proves an adelic descent theorem for localizing invariants of stable $\mathbb A_X=\sideset{}{'}\prod_{x\in X_{cl}} \widehat F_x,$9-categories: $G$0 Equivalently, the cubical diagram

$G$1

is a limit diagram, and so is the augmented semi-cosimplicial diagram $G$2. Here $G$3 is any localizing invariant valued in a stable $G$4-category $G$5, and examples include Bass–Thomason nonconnective algebraic $G$6-theory, $G$7, $G$8 in some contexts, and trace-related invariants (Kim, 2021).

The mechanism is cubical rather than directly semi-cosimplicial. One filters perfect complexes by support dimension, obtains exact sequences

$G$9

and then uses that a localizing invariant sends exact sequences to fiber sequences. In the Dedekind case $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$0, with $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$1 the fraction field, $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$2 the integral adeles, and $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$3 the finite adeles, the theorem gives

$$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$4

hence a Mayer–Vietoris long exact sequence in $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$5-groups (Kim, 2021).

A solid-algebraic refinement replaces ordinary quasi-coherent sheaves by solid quasi-coherent sheaves $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$6 for schemes of finite type over $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$7. The skeletal filtration produces adelic rings $$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$8, and the descent theorem is

$$BG(X)\simeq [G(F)\setminus G(\mathbb A_X)/G(\mathbb O_X)].$$9

as an equivalence. The vertices of the fracture cube are identified with Beilinson–Parshin adelic expressions, for example

$x\le y \quad \text{if } x\in \overline{\{y\}},$0

Full faithfulness comes from the adelic decomposition of the structure sheaf, while essential surjectivity uses completeness properties of the solid tensor product (Brav et al., 2024).

A further categorical enhancement replaces compact objects by nuclear ones. For stable presentable categories of nuclear modules, one has

$x\le y \quad \text{if } x\in \overline{\{y\}},$1

and therefore, for any accessible stable localizing invariant $x\le y \quad \text{if } x\in \overline{\{y\}},$2,

$x\le y \quad \text{if } x\in \overline{\{y\}},$3

The use of nuclear objects is motivated by the fact that compact objects suffer from an Eilenberg swindle and are not suitable for localizing invariants in that setting (Konovalov, 27 Jul 2025).

4. Arithmetic descent sets for adelic points

In Diophantine geometry, adelic descent concerns subsets of $x\le y \quad \text{if } x\in \overline{\{y\}},$4 defined by lifting conditions through torsors. For a smooth geometrically integral quasi-projective variety $x\le y \quad \text{if } x\in \overline{\{y\}},$5, a torsor $x\le y \quad \text{if } x\in \overline{\{y\}},$6 under a finite étale $x\le y \quad \text{if } x\in \overline{\{y\}},$7-group scheme $x\le y \quad \text{if } x\in \overline{\{y\}},$8, and an adelic point $x\le y \quad \text{if } x\in \overline{\{y\}},$9, one says that $$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$0 survives $$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$1 if there exists $$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$2 such that

$$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$3

The finite descent obstruction is obtained by requiring survival for all such torsors in a given class (Overkamp, 2016).

Kim’s non-abelian reciprocity construction gives a descending filtration

$$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$4

using the étale fundamental group, its lower central series, local path torsors $$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$5, and global-to-local maps in non-abelian cohomology. Under Kim’s technical hypotheses $$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$6 and $$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$7, Harpaz and Schlank prove

$$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$8

and hence

$$|X|_k=\{(x_0,\dots,x_k)\in |X|^{k+1}\mid x_0\le \cdots \le x_k\}.$$9

Thus Kim’s iterative construction and the finite descent obstruction for finite étale nilpotent torsors of odd order coincide in this setting (Overkamp, 2016).

A ramified generalization starts from a finite $\mathbb A_X^\bullet$0-cover

$\mathbb A_X^\bullet$1

that is only generically a torsor. One chooses a nonempty open $\mathbb A_X^\bullet$2 over which $\mathbb A_X^\bullet$3 is an étale $\mathbb A_X^\bullet$4-torsor, defines

$\mathbb A_X^\bullet$5

and sets

$\mathbb A_X^\bullet$6

This descent set is an obstruction to the Hasse principle and to weak approximation, since

$\mathbb A_X^\bullet$7

The corresponding Brauer-Manin obstruction is encoded by a subgroup $\mathbb A_X^\bullet$8 built from the semidirect product $\mathbb A_X^\bullet$9, and one has

$A^\bullet_{\mathrm{red}(X)}$0

A key negative result is that $A^\bullet_{\mathrm{red}(X)}$1 need not be algebraic even when $A^\bullet_{\mathrm{red}(X)}$2 is commutative; the paper gives an explicit transcendental obstruction on a quotient $A^\bullet_{\mathrm{red}(X)}$3 with $A^\bullet_{\mathrm{red}(X)}$4 constant metabelian, thereby answering Harari’s question negatively. By contrast, for a $A^\bullet_{\mathrm{red}(X)}$5-cover totally ramified along a divisor, one has

$A^\bullet_{\mathrm{red}(X)}$6

in the special case cited from Colliot-Thélène and Skorobogatov (Demeio, 2021).

A different adelic descent principle appears in the adelic Mordell–Lang theorem for subvarieties $A^\bullet_{\mathrm{red}(X)}$7 of an abelian variety over a global function field $A^\bullet_{\mathrm{red}(X)}$8. Assuming that $A^\bullet_{\mathrm{red}(X)}$9 has no nonzero positive-dimensional isotrivial quotient, one has

$F$00

Assuming finiteness of $F$01, this implies density of $F$02 in the Brauer set, and the Brauer-Manin set admits a decomposition into finitely many coset Brauer sets

$F$03

Here the adelic descent statement is that every adelic point of $F$04 arising as a limit of $F$05-rational points on $F$06 is already a limit of $F$07-rational points on $F$08 (Creutz, 29 Oct 2025).

5. Cohomological and equivariant formulations

Adelic descent also appears as a method for building global cohomology theories from localizations and completions along parameter spaces. In adelic cohomology on a poset, the adelic cube gives a homotopy pullback description, and the theory recovers classical invariants such as Čech cohomology and local cohomology: $F$09 In stable equivariant homotopy theory, the $F$10-page of the relevant spectral sequence is adelic cohomology, and in the torus case it identifies additively with tom Dieck splitting (Greenlees, 2019).

For equivariant elliptic cohomology, Tomasini uses adelic descent on the parameter spaces

$F$11

to reconstruct $F$12-rationalized $F$13-equivariant elliptic cohomology, $F$14-theory, and singular cohomology from fixed-point data. In rank one, the local piece of the elliptic theory at a closed point $F$15 is

$F$16

while at the generic point

$F$17

When $F$18, the adelic model recovers Grojnowski’s equivariant elliptic cohomology: $F$19 The same adelic mechanism gives algebraic models for rationalized equivariant $F$20-theory and cohomology and comparison theorems with periodic cyclic homology, including

$F$21

These results are presented as completing a program first proposed by Roșu (Tomasini, 2023).

A recurring structural theme is that the local pieces are not independent. They are tied together by the adelic diagram, and the global object is the compatible system obtained by totalization or homotopy limit. This suggests that adelic descent should be viewed less as a single formal theorem and more as a family of reconstruction procedures adapted to different categories of local data.

6. Analogical extensions and nonstandard uses

Some works use the language of adelic descent in a precise but analogical sense rather than as a theorem about descent of sheaves or torsors. For an elliptic curve $F$22 over a number field $F$23, the adelic point group is

$F$24

and the paper on adelic point groups shows

$F$25

where $F$26 and $F$27 is the torsion closure. For almost all elliptic curves over $F$28,

$F$29

The nontrivial arithmetic variation is controlled by the Galois action on torsion,

$F$30

and the authors explicitly describe this as an adelic analogue of descent: the topology of $F$31 is read off from the division fields $F$32 and the associated torsion representations (Angelakis et al., 2017).

A more speculative extension is the Derived Adelic Cohomology Conjecture (DACC) for elliptic curves. The proposal defines local derived sheaves $F$33, glues them by a cone construction

$F$34

and forms an adelic complex

$F$35

A Postnikov filtration yields a spectral sequence whose first nonzero differential is claimed to satisfy

$F$36

and

$F$37

The paper presents this as a homotopical form of adelic descent for BSD, supported by numerical evidence, rather than as an established theorem in the classical Beilinson or Tannakian sense (Wachs, 7 Mar 2025).

The most nonstandard use in the supplied corpus is the adelic theory of stock prices. There the price is modeled as an adele-valued function

$F$38

with a conceptual chain

$F$39

The paper itself states that it does not present a rigorous adelic descent theorem in the arithmetic-geometric sense; the “descent” is a coarse-graining and completion procedure from a microscopic trader model to a $F$40-adic theory and then to an adelic synthesis (Zharkov, 2011).

Across these variants, the stable core of the subject remains the mathematical local-to-global principle: adeles organize local completions and localizations into a diagram from which one reconstructs perfect complexes, localizing invariants, descent sets of adelic points, or other global structures. The precise meaning of “descent” depends on the ambient category, but the common architecture is the same—compatibility of local data across all relevant places or strata, and recovery of the global object from that compatible system.