- The paper establishes birational comparisons between simultaneous and iterated blow-ups, transferring exceptional-divisor cycles and enabling intersection-theoretic calculations.
- The paper proves the product formula for Segre classes for arbitrary closed embeddings, removing pure-dimensionality, integrality, and equidimensionality assumptions.
- The paper derives a multidegree-to-degree formula for arbitrary closed subschemes of multiprojective spaces, extending van der Waerden’s classical result through blow-ups and Chow theory.
Overview
The paper establishes two birational comparison results for blow-up constructions and applies them to obtain intersection-theoretic formulas. The first result compares the simultaneous multigraded blow-up of a product of closed embeddings with an iterated blow-up construction, producing a birational correspondence that pushes forward fundamental cycles of exceptional divisors. The second identifies the blow-up along a product of ideals with the scheme-theoretic closure of a rational map to a multiprojective space, yielding explicit relations among Chern classes of tautological line bundles and exceptional divisors. These tools are then applied to prove the product formula for Segre classes without any pure-dimensionality hypothesis, and to derive a degree formula for arbitrary closed subschemes of multiprojective spaces, extending the classical formula of van der Waerden [vdW78]. Throughout, the author works within Fulton's formalism of Chow groups, normal cones, and refined intersection products [Ful98].
Comparison of simultaneous and iterated blow-ups
Fix closed embeddings Xi↪Yi defined by coherent ideal sheaves Ii, and let Bli→Yi×Ak1 be the blow-up along Xi embedded via the zero section, with exceptional divisor Ei. The paper considers two models: Bl, the blow-up of ∏i(Yi×Ak1) along ∏iXi, and Bl′, the blow-up of the regular embedding E1×⋯×En↪Bl1×⋯×Bln. The main structural result (Proposition 2.1) constructs morphisms Ii0 and Ii1 commuting with the blow-down maps such that:
- Ii2;
- Ii3;
- Ii4.
The proof proceeds by the universal property of blow-ups to obtain Ii5, and by analyzing generic points to establish equality of cycle classes. Since each center contains no generic points of its ambient space, both blow-ups are isomorphisms over a common dense open subscheme containing all generic points, which gives Ii6. For the exceptional divisors, the argument is more delicate: Ii7 is a Ii8-bundle over Ii9 (the center being a regular embedding), while Bli→Yi×Ak10 is described as the projectivization of a tensor product of the cones Bli→Yi×Ak11. On the dense open where all cone coordinates are nonzero, the induced morphism on Proj is explicitly given by a graded ring homomorphism sending Bli→Yi×Ak12 to Bli→Yi×Ak13, which is an isomorphism over Bli→Yi×Ak14; this yields the bijection on generic points with isomorphic local rings, hence Bli→Yi×Ak15.
This birational correspondence is the geometric mechanism underlying the first application: it transfers intersection-theoretic data from the iterated model, where Segre classes behave multiplicatively under products, to the simultaneous model.
Recall that for a closed embedding Bli→Yi×Ak16, the Segre class is defined via the projectivization of the normal cone:
Bli→Yi×Ak17
Existing proofs of the product formula generally require the schemes involved to be pure-dimensional, because standard cycle-theoretic relations in [Ful98] are invoked under that hypothesis. The paper removes this assumption entirely.
The key observation is that Bli→Yi×Ak18 for any closed embedding, proved directly from the definition using Lemma 1.7.2 of [Ful98]: the extra hyperplane summand contributes nothing since Bli→Yi×Ak19 for dimensional reasons. With this stabilization in place, the proof of multiplicativity proceeds as follows. For regular embeddings, the formula follows immediately from the splitting of the normal bundle into the direct sum of pullbacks of the factors' normal bundles. For general embeddings, one applies Proposition 2.1: the Segre class of the product embedding equals the pushforward from the exceptional divisor Xi0 of the simultaneous blow-up, which by Xi1 and the projection formula equals the pushforward from Xi2; but Xi3, and since each Xi4 is a regular embedding, the already-established regular case gives
Xi5
The conceptual point is that the exceptional divisor serves as a bridge reducing arbitrary closed embeddings to Cartier divisors on a suitable resolution, so the product formula becomes a consequence of the geometry of blow-ups rather than of cycle-theoretic identities requiring dimension hypotheses. This is a genuine strengthening: no integrality or equidimensionality of the Xi6 or Xi7 is assumed.
Graph closures versus blow-ups
The second part of the paper studies a rational map defined by global sections Xi8, not all identically zero. These determine an ideal sheaf Xi9 (the image of the evaluation map twisted by Ei0), a rational map Ei1 on Ei2, and two compactifications: the blow-up Ei3 and the graph closure Ei4, the scheme-theoretic image of Ei5. A surjection of graded algebras Ei6 produces a closed embedding of Ei7 into Ei8 with Ei9, and a natural closed embedding Bl0.
When Bl1 is integral, the embedding Bl2 is an isomorphism, shown via injectivity of Bl3 in a short exact sequence diagram. The multi-section version (Proposition 3.1) extends this: for integral Bl4 with invertible sheaves Bl5 and sections defining ideals Bl6, the blow-up along the product ideal Bl7 is identified with the graph closure of Bl8, and there exist morphisms Bl9 satisfying
∏i(Yi×Ak1)0
and consequently ∏i(Yi×Ak1)1.
The identification rests on the fiber-product decomposition ∏i(Yi×Ak1)2 together with the surjection onto ∏i(Yi×Ak1)3. Note that integrality of ∏i(Yi×Ak1)4 is used here; the paper does not claim the graph-closure identification for non-reduced or reducible bases.
Specializing to ∏i(Yi×Ak1)5 with two sets of homogeneous coordinate sections, the construction yields a blow-up ∏i(Yi×Ak1)6 embedded in ∏i(Yi×Ak1)7. Because every fiber of ∏i(Yi×Ak1)8 is isomorphic to ∏i(Yi×Ak1)9, Miracle Flatness implies this map is smooth proper of relative dimension one. Writing ∏iXi0 and ∏iXi1 for the pullbacks of the hyperplane classes, the disjointness of ∏iXi2 and ∏iXi3 forces
∏iXi4
from which powers of ∏iXi5 reduce via complete homogeneous symmetric polynomials:
∏iXi6
with ∏iXi7 for ∏iXi8 and ∏iXi9.
For a closed subscheme Bl′0, let Bl′1 and let Bl′2 be the scheme-theoretic image of the inverse image on the open locus. Lemma 1.7.1 of [Ful98] gives Bl′3 in every dimension, compatibly with cycle decompositions. The central result (Proposition 3.2) states that for every Bl′4, the codimension-Bl′5 degree of Bl′6 equals the sum of the codimension-Bl′7 bidegrees of Bl′8 — for arbitrary closed subschemes, not merely varieties. The proof pushes forward Bl′9 through E1×⋯×En↪Bl1×⋯×Bln0 using the projection formula and the quadratic relation; the term involving E1×⋯×En↪Bl1×⋯×Bln1 computes exactly the sum of bidegrees (all fibers being E1×⋯×En↪Bl1×⋯×Bln2), while the E1×⋯×En↪Bl1×⋯×Bln3 term vanishes for dimensional reasons. The paper notes the evident generalization to arbitrary multiprojective spaces: codimension-E1×⋯×En↪Bl1×⋯×Bln4 degrees equal sums of codimension-E1×⋯×En↪Bl1×⋯×Bln5 multidegrees.
This recovers van der Waerden's classical degree formula [vdW78]. When E1×⋯×En↪Bl1×⋯×Bln6 is integral of dimension E1×⋯×En↪Bl1×⋯×Bln7, E1×⋯×En↪Bl1×⋯×Bln8 is integral of dimension E1×⋯×En↪Bl1×⋯×Bln9, and the bihomogeneous equations of Ii00 define Ii01 in Ii02 — precisely van der Waerden's setting. The modern derivation differs in method and scope: it is entirely intersection-theoretic, built on Chow groups, blow-ups, and refined intersection products, hence functorial and valid for arbitrary closed subschemes. The subtlety for non-integral schemes is illustrated explicitly: for Ii03 cut out by Ii04, the bidegrees sum to Ii05 while the subscheme of Ii06 defined by the same equations has degree Ii07 — so naive equation transfer fails, and the correct statement must be formulated through the correspondence Ii08 rather than through shared defining ideals.
Limitations and open questions
Several hypotheses bear directly on the results. The graph-closure identification (Proposition 3.1) requires Ii09 to be integral; the paper does not extend the isomorphism Ii10 to non-integral bases, where the closure may acquire non-reduced structure differing from the blow-up. The multidegree formula is established for biprojective spaces with the multiprojective case asserted in a remark without full proof. The comparison of exceptional divisors relies on generic-point arguments with isomorphic local rings; extending these cycle-level equalities to finer invariants (e.g., refined Gysin pullbacks or operational classes) is not addressed. Finally, the degree formula concerns the specific correspondence Ii11; characterizing when other natural correspondences between multidegrees and degrees hold remains open.
Conclusion
The paper provides a unified blow-up framework connecting simultaneous multigraded blow-ups, iterated blow-ups, and graph closures of rational maps to multiprojective spaces. Its two principal applications remove standing hypotheses from classical results: the product formula for Segre classes now holds for arbitrary closed embeddings without pure-dimensionality assumptions, and van der Waerden's degree formula is rederived functorially and extended from integral varieties to arbitrary closed subschemes of multiprojective spaces. The methods are elementary within Fulton's intersection theory, relying on universal properties of blow-ups, generic-point comparisons, and projection formulas, making the results accessible and broadly applicable wherever Segre classes and multidegrees appear.