The Pardon Algebra for zero-cycles

Kai Behrend · 2026

Read The Pardon Algebra for zero-cycles by Kai Behrend (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-05534.txt.

Introduction

Behrend builds a Pardon-style homology algebra for zero-cycles in higher-dimensional varieties, mirroring Pardon's recent gadget for one-cycles in threefolds. The construction reframes point-counting enumerative problems, including the degree-zero MNOP conjecture on Hilbert schemes of points in projective three-space. It offers a new algebraic viewpoint on counting points.

Catalog record

CREATORKai Behrend
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE113981
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2604.05534

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

The Pardon Algebra for zero-cycles Kai Behrend Abstract Recently, John Pardon proved the MNOP conjecture by introducing a new mathematical gadget, which we call the Pardon homology algebra of 1-cycles in 3-folds. We work out an analogous construction for 0 0 -cycles in d d -folds. This gives a new point of view on enumerative problems involving point-counting, such as, for example, the degree zero MNOP conjecture. Introduction This is an exploration of John Pardon’s theory [ 9 ] , in a much simpler case: instead of considering 1-cycles, we consider 0-cycles. Pardon used his theory to prove the MNOP Gromov-Witten-Donaldson-Thomas correspondence. We arrive at a new point of view on point enumeration problems. Summary of main result. Fix d . A family of d -folds parametrized by S is a separated (not necessarily proper) smooth morphism of schemes X\to S of relative dimension d . We denote by \overline{Z}_{n}X=X^{n}_{S}/\!/\Sigma_{n}\to S the relative n -th symmetric power of X\to S , and let \overline{Z}X=\coprod_{n}\overline{Z}_{n}X . The Pardon homology algebra is by definition H_{\ast}=\mathop{\lim\limits_{\textstyle\longrightarrow}}\limits_{X/S}H_{\ast}(\overline{Z}X,{\mathbb{Q}})\,. Following Pardon, we construct a product and a coproduct on H_{\ast} , and prove that H_{\ast} is a graded commutative and cocommutative Hopf algebra over {\mathbb{Q}} . We define a ‘tautological’ Hopf subalgebra, which contains all classes one might conceivably be interested in, and we prove that H_{\ast}^{\mbox{\tiny taut}}={\mathbb{Q}}[q_{n\lambda}]\,, where n runs over all positive integers, and \lambda over all integer partitions with at most d parts, such that \lambda_{i}\geq n-1 for all parts \lambda_{i} of \lambda . Here, q_{n\lambda} is a class in H_{\ast} , indexed in the colimit by the family of d -folds ( e_{i} is the i -th standard basis vector of {\mathbb{Q}}^{n} ) X=\mathop{\rm tot}\nolimits\Big({\mathscr{O}}(e_{1})\oplus\ldots\oplus{\mathscr{O}}(e_{d})\Big)\longrightarrow{\mathbb{P}}^{\lambda_{1}}\times\ldots\times{\mathbb{P}}^{\lambda_{d}}=S\,. The n -fold zero section of this family defines a section \Delta of \overline{Z}_{n}X\to S , and q_{n,\lambda}=\Delta_{\ast}[{\mathbb{P}}^{\lambda_{1}}\times\ldots\times{\mathbb{P}}^{\lambda_{d}}] . The interesting part of H_{\ast}^{\mbox{\tiny taut}} , in the sense that it contains all [\overline{Z}_{n}X] for compact X\to\ast , is the part where the homological degree equals 2nd . This forms a subalgebra (not a Hopf subalgebra), and is equal to {\mathbb{Q}}[q_{n,\lambda+n-1}]\,, where n runs over all positive integers, and \lambda over all partitions of d . (By \lambda+n-1 we denote the partition obtained by adding n-1 to each of the d parts of \lambda ). This reduces the computation of the values of any multiplicative point counting theory e on all proper algebraic d -folds to evaluating e on the classes q_{n,\lambda+n-1} . (The direct analogy to Pardon’s work for 1-cycles on 3-folds would be the subalgebra of homological degree 0 , which is simply {\mathbb{Q}}[q_{n,0}] , where n runs over the positive integers. So this algebra is freely generated by the ‘equivariant multiple points’.) Outline. If X is an algebraic d -fold, the space of 0 -cycles of degree n is the symmetric product which we denote by \overline{Z}_{n}X . In order to avoid dealing with singular varieties, we prefer working with the quotient stack Z_{n}X=X^{n}/\Sigma_{n} , of which \overline{Z}_{n}X is the coarse moduli space. In terms of (co-)homology with {\mathbb{Q}} -coefficients, this makes essentially no difference. The moduli problem solved by Z_{n}X is that of morphisms D\to S\times X , where D\to S is finite étale of degree n . Thus, we can treat the theory of 0 -cycles also as a theory of maps. When passing from a fixed target variety X to a family X\to S , we consider diagrams D\to X\to S , where X\to S is a family of d -folds and D\to S is finite étale. We denote the relative space of maps of degree n by Z_{n}X=X^{n}_{S}/\Sigma_{n} . For the theory to be compatible with passing from Z_{n}X to its coarse moduli space \overline{Z}_{n}X , we need to make sure Z_{n}X has finite stabilizers. This requires X\to S to be separated. On the other hand, dropping the separatedness assumption on our target families X\to S gives rise to a perfectly consistent theory, which we develop in the first part of this paper. It is simpler than the separated theory, deformation to the normal bundle is allowed, and multiple points can be separated into sums of simple points. The tautological Pardon algebra gets smaller, and is now equal to {\mathbb{Q}}[q_{1\lambda}] . In the second part, we briefly introduce the dual of the Pardon homology algebra, the Pardon cohomology algebra. It contains the enumerative theories against which classes in H_{\ast} can be evaluated to get enumerative invariants. The third part contains the main results announced above. As an illustration, we finish with a proof of the degree zero MNOP conjecture. (Our proof apparently works for proper Deligne-Mumford stacks, and it does not use relative Donaldson-Thomas theory.) Many problems involving point counts in d -folds can be formulated (and solved) using these results. We expect that the calculations this gives rise to are essentially equivalent to well-known approaches to these kinds of problems, as developed for example by Ellingsrud-Göttsche-Lehn [ 3 ] . All geometric arguments are purely algebraic. This distinguishes this work from Pardon’s for 1-cycles, which is essentially analytic. Therefore, what we describe here should work for an algebraic (co)-homology theory such as Chow-Witt (co)homology theory, which is homotopy invariant, and should give values in the Witt-ring. This will be explored elsewhere. Conventions. All schemes and stacks are over {\mathbb{C}} , all homology and cohomology coefficients are {\mathbb{Q}} . Deligne-Mumford stacks have separated diagonal, if the inertia stack is finite over the stack, the Deligne-Mumford stack is said to have finite stabilizers. 1 The non-separated Pardon Hopf algebra 1.1 Homology and Cohomology We work with Deligne-Mumford stacks over the complex numbers. They have representable, finite type and unramified diagonal, but we do not assume separatedness, or finiteness of stabilizers. Singular homology and cohomology with coefficients in {\mathbb{Q}} are defined via double complexes associated to groupoid presentations as in [ 1 ] . In addition, we assume existence of Gysin wrong way maps f^{!}:H_{i}(Y)\longrightarrow H_{i+2d}(X)\,, and f_{!}:H^{i}(X)\longrightarrow H^{i-2d}(Y)\,, for any (not necessarily representable) proper lci morphism f:X\to Y of relative dimension d , as constructed in [ 6 ] . These Gysin maps are functorial, and commute with homology pushforward and cohomology pullback. Moreover, they are adjoint: \langle e,f^{!}\alpha\rangle=\langle f_{!}e,\alpha\rangle\,, for a cohomology class e and a homology class \alpha . Finally, if f:X\to Y is finite étale, then f_{\ast}f^{!}:H_{\ast}(Y)\to H_{\ast}(Y) and f_{!}f^{\ast}:H^{\ast}(Y)\to H^{\ast}(Y) are multiplication by the degree of f . For example, every smooth and proper Deligne-Mumford stack X of dimension d has a fundamental class [X]\in H_{2d}(X) , defined by [X]=a^{!}[\ast] , where [\ast]\in H_{0}(\ast) is the canonical element, and a:X\to\ast the structure morphism. Dually, there is an integration map \int_{[X]}:H^{2d}(X)\to{\mathbb{Q}}=H^{0}(\ast) , defined by a_{!} . For example, if the finite group G acts on the algebraic space X , then [X/G]=a^{!}[\ast]=\frac{1}{|G|}\pi_{\ast}\pi^{!}a^{!}[\ast]=\frac{1}{|G|}\pi_{\ast}[X]\,, where \pi:X\to X/G is the quotient map, and a:X/G\to\ast the structure map. In the case where X=\ast , we can identify {\mathbb{Q}}=H_{0}(\ast)=H_{0}(BG) , and via this identification, we have [BG]=1/|G| . Künneth theorem We will need to use the Künneth theorem. There are natural inverse isomorphisms {\lx@inpgf@ignorespaces H_{\ast}(X)\otimes H_{\ast}(Y)} {\lx@inpgf@ignorespaces H_{\ast}(X\times Y)} \scriptstyle{\lx@inpgf@ignorespaces G} \scriptstyle{\lx@inpgf@ignorespaces F} 1.2 Families and cycle spaces Definition 1.1 We fix an integer d\geq 0 , and define a family of d

Read the complete text