🔢 mathematics

Suslin's cancellation conjecture on smooth real affine varieties with few real points

This paper investigates Suslin's cancellation conjecture on smooth real affine varieties, specifically focusing on cases where the real locus is either empty or possesses a small cohomological dimension.

Original authors: Sourjya Banerjee, Jean Fasel, Samuel Lerbet

Published 2026-08-19

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Sourjya Banerjee, Jean Fasel, Samuel Lerbet

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

In the vast landscape of modern mathematics, there is a branch dedicated to understanding shapes that are defined not by curves and surfaces in space, but by systems of equations. These are called algebraic varieties, and when they are "smooth," they behave like well-behaved surfaces without sharp corners or tears. Within these shapes, mathematicians study objects called vector bundles, which can be thought of as a collection of flat, flexible sheets attached to every point of the shape. A central question in this field is whether these sheets can be simplified. If you have a bundle that looks like it is made of a complicated stack of sheets, can you prove that it is actually just a simple stack plus some extra, unnecessary sheets that can be peeled away? This is known as the cancellation problem. For decades, mathematicians have known that if the underlying shape exists over a field of complex numbers, these bundles are almost always simple enough to be cancelled out. However, the situation becomes much more delicate when the shape exists over the real numbers, the kind of numbers we use to measure the physical world. The real world introduces a topological complexity: the shape might have holes, or it might be made of several disconnected pieces floating in space. These features can sometimes prevent a bundle from being simplified, even when the equations suggest it should be possible.

A team of researchers, Sourjya Banerjee, Jean Fasel, and Samuel Lerbet, has now solved a significant piece of this puzzle for a specific class of real-world shapes. They focused on smooth shapes that have very few points in the real world, or perhaps none at all. Imagine a shape that, while defined by real equations, has no actual real points where it exists, or perhaps exists only in a way that its real points do not form any closed, compact loops or islands. The authors proved that for these specific shapes, the cancellation problem has a positive answer. If you have a vector bundle of a certain size on such a shape, and you add a simple, free sheet to it, you can always remove that extra sheet to get back to the original bundle. This result confirms a long-standing conjecture by the mathematician Arvind Suslin for these particular cases, showing that the topological obstructions that usually complicate real-world shapes simply vanish when the real points of the shape are sparse or non-existent.

The path to this discovery required the team to navigate a sophisticated framework known as motivic homotopy theory. This is a way of studying shapes by treating them as if they were made of rubber that can be stretched and deformed in a specific, algebraic way. Instead of looking at the shape directly, the researchers looked at the "holes" and "twists" in the space of all possible bundles on that shape. They used a method called obstruction theory, which is like checking a map for roadblocks. To see if a bundle can be simplified, one must check if there are any hidden barriers preventing the simplification. These barriers are measured by cohomology groups, which are algebraic tools that count the number of holes or disconnected parts in a shape. The researchers found that for their specific shapes, the relevant cohomology groups were zero. In plain terms, the "roadblocks" did not exist. Because the real points of the shape were too few to form the necessary closed loops or compact islands, the algebraic machinery that usually detects these topological problems had nothing to detect.

The team's work builds on previous findings that showed cancellation works for shapes with no real points at all, and for shapes where the real points are orientable and have no compact connected components. They extended this to a broader category by proving that if the shape's real points do not generate a specific type of topological complexity in dimension equal to the shape's dimension, then cancellation holds for bundles of that rank. They also showed that if the real points lack complexity in the dimension just below that, cancellation holds for bundles of one rank lower. This distinction is crucial because it maps the exact boundary where the simplification of bundles is possible. The authors demonstrated that the only things that can stop a bundle from being cancelled are the compact, closed pieces of the shape's real existence. If those pieces are absent, the algebraic rules that govern the complex world apply just as strictly to the real world.

This result is not just a theoretical victory; it clarifies the relationship between the algebraic equations defining a shape and the topological reality of that shape. The researchers showed that when the real locus is "cohomologically small," the behavior of vector bundles is governed entirely by the same invariants that work in the complex world, specifically the Chern classes, which are algebraic numbers associated with the shape. They proved that no additional topological data is needed to decide if a bundle can be simplified. The paper also explicitly rules out the idea that this result could be improved further for general cases. The authors noted that for other types of shapes, specifically those with more complex real structures, cancellation can fail. Their work defines the precise limit of where the "nice" behavior of complex algebraic geometry extends into the messy reality of real algebraic geometry. By isolating the condition of having few real points, they provided a clear, definitive answer to a question that had remained open for a specific but important class of mathematical objects.

Technical Summary: Suslin's Cancellation Conjecture on Smooth Real Affine Varieties with Few Real Points

Problem Statement
The paper addresses the cancellation problem for projective modules over smooth real affine algebras. Specifically, it investigates under which conditions a projective module $P$ of rank $r$ is cancellative, meaning that $P \oplus A^{\oplus m} \cong Q \oplus A^{\oplus m}$ implies $P \cong Q$ for any projective module $Q$ . While Suslin proved that every rank $d$ projective module over a smooth affine algebra of dimension $d$ over an algebraically closed field is cancellative, the situation over the real numbers $\mathbb{R}$ is more delicate. Previous work established that cancellation holds for rank $d$ modules when the real locus $X(\mathbb{R})$ is empty, or for rank $d$ modules when $X(\mathbb{R})$ has no compact connected components. The central question of this paper is whether these results extend to all rank $d$ (and rank $d-1$ ) projective modules when the real locus is "cohomologically small," specifically when certain cohomology groups of $X(\mathbb{R})$ vanish.

Methodology
The authors employ motivic obstruction theory within the framework of $\mathbb{A}^1$ -homotopy theory. The core strategy involves reducing the cancellation problem to the injectivity of stabilization maps between sets of $\mathbb{A}^1$ -homotopy classes of maps into classifying spaces.

Motivic Homotopy Setup: The set of isomorphism classes of rank $n$ vector bundles on a smooth affine scheme $X$ , denoted $V_n(X)$ , is identified with the set of pointed $\mathbb{A}^1$ -homotopy classes $[X_+, BGL_n]_{\mathbb{A}^1}$ . Cancellation corresponds to the injectivity of the stabilization map $s_{n,\infty}: [X_+, BGL_n]_{\mathbb{A}^1} \to [X_+, BGL]_{\mathbb{A}^1}$ .
Obstruction Theory: The authors utilize the Moore–Postnikov tower for the map $\pi_n: BGL_n \to BGL$ . The obstructions to lifting maps (and thus to injectivity of the stabilization map) lie in Nisnevich cohomology groups of the form $H^i_{\text{Nis}}(X, \pi^{\mathbb{A}^1}_j(F)(\xi))$ , where $F$ is the homotopy fiber of the map between classifying spaces and $\xi$ is a twist determined by the bundle.
Key Technical Reduction: The problem is reduced to proving the vanishing of specific cohomology groups. For rank $d$ cancellation, the critical group is $H^d_{\text{Nis}}(X, \pi^{\mathbb{A}^1}_d(\mathbb{A}^d \setminus \{0\})(L))$ . For rank $d-1$ , the analysis involves $H^d_{\text{Nis}}(X, \pi^{\mathbb{A}^1}_d(F_{d-1})(L))$ and $H^{d-1}_{\text{Nis}}(X, \pi^{\mathbb{A}^1}_{d-1}(F_{d-1})(L))$ .
Cohomological Computations: The authors compute these homotopy sheaves using exact sequences involving Milnor–Witt K-theory ( $K^{MW}_*$ ), Milnor K-theory ( $K^M_*$ ), and Grothendieck–Witt groups ($GW$). They leverage the Rost–Schmid complex to compute Nisnevich cohomology and relate these groups to the topology of the real locus $X(\mathbb{R})$ via Jacobson's theorem and results from [CS96].

Key Hypotheses
The results rely on the vanishing of specific cohomology groups of the real locus $X(\mathbb{R})$ (endowed with the Euclidean topology):

Hypothesis (*): $H^d(X(\mathbb{R}), \mathbb{Z}/2) = 0$ . This is equivalent to $X(\mathbb{R})$ having no compact connected components.
Hypothesis ():** Hypothesis (*) holds, and for every line bundle $L$ on $X$ , $H^{d-1}(X(\mathbb{R}), \mathbb{Z}(L)) = 0$ , where $\mathbb{Z}(L)$ is the twist of the constant sheaf by the torsor associated with $L$ .
Hypothesis ('):** $H^d(X(\mathbb{R}), \mathbb{Z}/2) = 0$ and $H^{d-1}(X(\mathbb{R}), \mathbb{Z}/2) = 0$ .

Main Results

Theorem 1 (Main Cancellation Theorem): Let $A$ be a smooth $\mathbb{R}$ -algebra of dimension $d \geq 3$ with $X = \text{Spec}(A)$ .
- If $H^d(X(\mathbb{R}), \mathbb{Z}/2) = 0$ (Hypothesis *), then every rank $d$ projective $A$ -module is cancellative.
- If $H^d(X(\mathbb{R}), \mathbb{Z}/2) = 0$ and $H^{d-1}(X(\mathbb{R}), \mathbb{Z}(L)) = 0$ for all line bundles $L$ (Hypothesis **), then every rank $d-1$ projective $A$ -module is cancellative.
Theorem 3 (Technical Vanishing): Under Hypothesis (*), $H^d_{\text{Nis}}(X, \pi^{\mathbb{A}^1}_d(\mathbb{A}^d \setminus \{0\})(L)) = 0$ for every line bundle $L$ . This is the crucial technical step enabling the rank $d$ cancellation result.
Corollary 4.6: Cancellation holds in corank 0 (rank $d$ ) under Hypothesis (*). This generalizes previous results by [DTZ18] and [Ban25].
Theorem 5.5: Under Hypothesis (**), the weak Euler class group $E^0_{d-1}(X)$ , the Euler class group $E_{d-1}(X)$ , the Chow–Witt group $\widetilde{CH}_{d-1}(X)$ , and the Chow group $CH_{d-1}(X)$ are all isomorphic. This establishes a strong classification of vector bundles of corank 1 by their Euler classes in the presence of these cohomological vanishing conditions.

Significance and Claims
The paper claims to extend the known scope of Suslin's cancellation conjecture over real fields. While previous results required the real locus to be empty or specifically non-compact in a way that implied the vanishing of $H^d(X(\mathbb{R}), \mathbb{Z}/2)$ , this work demonstrates that the vanishing of this specific cohomology group is sufficient to ensure the cancellation of all rank $d$ projective modules, not just stably free ones.

Furthermore, the paper establishes that for rank $d-1$ modules, the cancellation property holds if the cohomological dimension is sufficiently small (Hypothesis **), linking the algebraic property of cancellation directly to the topological properties of the real locus. The authors note that their results cannot be improved in general, citing Mohan Kumar's counterexamples for complex algebras of dimension $p+2$ where rank $p$ modules are not cancellative.

The work unifies previous disparate results (e.g., those for empty real loci vs. orientable non-compact loci) under a single cohomological framework, showing that the "delicate" nature of splitting and cancellation over $\mathbb{R}$ is controlled by the vanishing of specific cohomology groups of the real points, much like the situation over algebraically closed fields where Chern classes suffice.

Technical Summary: Suslin's Cancellation Conjecture on Smooth Real Affine Varieties with Few Real Points

More like this