🔢 mathematics

A Hopf index for isotropic sections of orthogonal bundles

This paper establishes eight analogues of the Hopf index for isotropic sections of orthogonal bundles equipped with a quadratic form, demonstrating their applications to cosection-localized virtual cycles and DT $^4$ virtual cycles.

Original authors: Martijn Kool, Jeongseok Oh, Jørgen Vold Rennemo, Richard P Thomas

Published 2026-06-23

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Martijn Kool, Jeongseok Oh, Jørgen Vold Rennemo, Richard P Thomas

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

Imagine you are a mathematician trying to count the "zeros" of a complex equation. In the standard world of complex numbers, if you have a function that hits zero at a specific point, you can usually assign it a "multiplicity" (a weight) that tells you how many times it effectively touches zero. If the function is simple, it touches once. If it's a bit wobbly, it might touch twice or three times. This is a bit like counting how many times a spinning top wobbles before it stops; the wobble tells you the "strength" of the stop.

This paper, titled "A Hopf Index for Isotropic Sections of Orthogonal Bundles," by Martijn Kool, Jeongseok Oh, Jørgen Vold Rennemo, and Richard Thomas, tackles a much stranger version of this counting problem.

Here is the story in simple terms:

1. The Strange Rule: "The Zero Sum"

In the standard world, if you have a zero, it usually counts as a positive number (1, 2, 3...). But in this specific mathematical universe (involving "orthogonal bundles" and "isotropic sections"), there is a new rule: Zeros can cancel each other out.

The authors introduce a new way of counting called the "Orthogonal Hopf Index." Think of it like a balance scale.

Some zeros are "positive" (they add weight).
Some zeros are "negative" (they subtract weight).
The final answer is the net weight (Positive minus Negative).

The shocking discovery in this paper is that you can have a situation where the "length" of the zero (how many times it looks like it's touching zero) is 3, but the final "Orthogonal Hopf Index" is 0.

The Analogy: Imagine you have three people standing on a scale. Two of them are wearing heavy boots (positive weight), and one is wearing a jetpack that pulls them up (negative weight). If the math works out just right, the scale reads zero, even though three people are standing there. The paper explains why this happens and gives you eight different ways to calculate this "net weight."

2. The Eight Ways to Count

The paper lists eight different methods to calculate this index, similar to how you might measure the height of a building using a tape measure, a laser, or by counting the stairs.

The "Winding" Method: Imagine wrapping a string around a pole. How many times does it twist? This twist tells you the index.
The "Deformation" Method: Imagine the zero is a messy knot. You gently pull the string to untie it into simple, separate knots. You count the positive knots and subtract the negative ones.
The "Homogeneous" Method: If the equation looks the same no matter how much you zoom in or out, you can calculate the index just by looking at the shape of the equation.
The "K-Theory" Method: This is like looking at the equation through a special pair of glasses that turns the problem into a puzzle of algebraic blocks. You count the blocks in one pile and subtract the blocks in another.

The paper proves that all eight methods give the exact same answer, even though they look completely different on the surface.

3. The "Running Example" (The 3 vs. 0 Mystery)

The authors use a specific example to show off their theory. They create a mathematical object (a section) that has a zero at the origin of a 2D plane.

Visually/Geometrically: It looks like it has a "length" of 3 (it's a triple zero).
The New Index: When they apply their "Orthogonal Hopf Index" formula, the answer is 0.

Why? Because the "positive" parts of the zero cancel out the "negative" parts perfectly. The paper shows that if you try to "deform" (smooth out) this zero into simpler pieces, you get two positive zeros and two negative zeros (or a similar mix that sums to zero), proving that the "3" you saw initially was an illusion caused by the complexity of the shape.

4. Why This Matters (The "Virtual" World)

The paper mentions that this isn't just a game with numbers; it applies to Virtual Cycles.

The Concept: In advanced geometry (specifically counting shapes on "Calabi-Yau 4-folds," which are complex shapes used in string theory), mathematicians often deal with "virtual" spaces. These are spaces that should exist based on the equations, but might be messy or broken.
The Application: Usually, when you count points in these virtual spaces, you expect a positive number. But this paper shows that in the "orthogonal" world, the count can be zero or even negative.
The Takeaway: It's like counting the number of solutions to an equation. In the real world, $x^2 = -1$ has no solutions. In the complex world, it has two. In this new "orthogonal" world, you might have a solution that looks like it's there, but its "weight" is zero because of how it twists and turns.

5. The Special Case of "n=2"

When the dimension is small (specifically 2), the authors can refine their counting even further. Instead of just getting a single number (like 0 or 5), they can get a pair of numbers (like 3 and 3).

The first number counts the "positive" twists.
The second number counts the "negative" twists.
The final index is just the difference between them.

This allows them to see why the total is zero: it's not that nothing is happening; it's that two equal and opposite things are happening at the same time.

Summary

This paper is a guidebook for a new kind of counting. It tells mathematicians: "Don't just count how many zeros you see. Look at how they twist and turn. Some zeros cancel each other out. We have eight different tools to measure this cancellation, and they all agree. Sometimes, a zero that looks like it has a weight of 3 actually has a weight of 0."

It's a bit like realizing that a crowd of people standing in a circle, some holding hands clockwise and some counter-clockwise, might actually result in a net rotation of zero, even though everyone is standing there. The paper gives you the math to prove it.

Technical Summary: A Hopf Index for Isotropic Sections of Orthogonal Bundles

Problem Statement
The paper addresses the calculation of multiplicities for zeros of isotropic sections of vector bundles equipped with a nondegenerate quadratic form. In classical topology, the Hopf index theorem equates the multiplicity of a zero of a section of an oriented real vector bundle with a winding number (an element of $\pi_{2n-1}(S^{2n-1}) \cong \mathbb{Z}$ ). For holomorphic sections, this multiplicity coincides with the length of the zero locus.

However, for isotropic sections $s$ of an orthogonal bundle $(E, q)$ (where $q(s,s)=0$ ), the situation is more complex. While a "square root Euler class" $\sqrt{e}(E, s)$ exists (defined via cosection localisation in [KL] and [OT1]), it does not always equal the length of the zero locus $Z(s)$ , nor does it always behave like a standard deformation-invariant count of points. The authors highlight a specific counterexample (Equation 0.1) where $Z(s)$ has length 3, yet the assigned multiplicity is 0. The central problem is to provide explicit, computable formulae for this multiplicity $\sqrt{e}(E, s)$ and to understand its topological and algebraic properties, particularly why it can vanish or differ from the scheme-theoretic length.

Methodology
The authors develop a suite of eight analogues to the classical Hopf index theorem, tailored specifically for isotropic sections of $SO(2n, \mathbb{C})$ bundles. The methodology involves:

Topological Analysis: Utilizing homotopy groups of the space of isotropic vectors. For $n=2$ , the space of isotropic lines is isomorphic to $\mathbb{P}^1 \times \mathbb{P}^1$ , allowing the index to be refined from a single integer to a pair $(d_1, d_2)$ .
Homological and Cohomological Techniques: Employing the Hurewicz map, Leray spectral sequences, and intersection theory on quadrics to relate homotopy classes to homology classes.
Algebraic Geometry and K-Theory: Using the Koszul complex and Clifford algebras. Specifically, the authors utilize the 2-periodic complex $M_\bullet$ formed by Clifford multiplication by $s$ on the spinor module $M = M_+ \oplus M_-$ .
Deformation and Localisation: Analyzing deformations of sections to simple zeros and applying torus localisation techniques (equivariant cohomology) to compute invariants.
Cosection Localisation: Comparing their new formulae against the original definition of $\sqrt{e}(E, s)$ via the Kiem-Li cosection localisation construction [KL], demonstrating that the new formulae are often computationally superior.

Key Contributions and Results

The paper establishes that the orthogonal Hopf index $\sqrt{e}(E, s)$ can be computed via eight distinct formulae (labeled OH1–OH8), which are proven to be equivalent:

Topological Formulae (OH1, OH2): The index is the degree of the projection of the section to the sphere of the real part of the bundle, or the homotopy class of the section projected to the quadric of isotropic lines $Q \subset \mathbb{P}(E)$ . For $n=2$ , this refines to a pair of integers $(d_1, d_2) \in \mathbb{Z} \oplus \mathbb{Z}$ such that $\sqrt{e}(E, s) = d_1 - d_2$ .
Holomorphic and Deformation Formulae (OH3, OH4): While the index equals $\pm \text{length}(Z(s))$ if the section factors through a maximal isotropic subbundle, this fails generally. The authors provide a formula involving a deformation to simple zeros, summing signs determined by the maximal isotropic subspace spanned by the differential at each zero.
Homogeneous and Normal Cone Formulae (OH5, OH6): For homogeneous sections, the index is expressed in terms of the degrees of the map to the quadric components. For general sections, it is computed via the projective normal cone $P(C_{Z(s)}/C^n)$ .
K-theoretic Formula (OH7): The index is the difference in lengths of the cohomology sheaves of the 2-periodic Clifford complex: $\sqrt{e}(E, s) = \text{length } H^+(M_\bullet) - \text{length } H^-(M_\bullet)$ .
Torus Localisation (OH8): An equivariant formula expressing the index as a ratio of equivariant Euler classes.

Refinement for $n=2$
A significant portion of the paper is dedicated to the case $n=2$ . Here, the index refines to a pair $(d_1, d_2)$ . The authors prove that $d_1$ and $d_2$ correspond to the lengths of the cohomology sheaves $H^+(M_\bullet)$ and $H^-(M_\bullet)$ respectively. Crucially, while the difference $d_1 - d_2$ is the deformation-invariant virtual count, the individual terms $d_1$ and $d_2$ are not necessarily deformation invariant in the general case, though they are for $n=2$ under holomorphic isotropic deformations.

Running Example
The paper repeatedly uses the section $s = (x^2, y^2, xy, -xy)$ over $\mathbb{C}^2$ with quadratic form $XY+ZW$.

The zero locus $Z(s)$ has length 3.
Using the K-theoretic approach, the authors calculate $H^+(M_\bullet)$ and $H^-(M_\bullet)$ to have lengths 2 and 2 respectively.
The resulting index is $2 - 2 = 0$ .
This explicitly demonstrates that $\sqrt{e}(E, s)$ can vanish even when the zero locus is non-empty and has positive length, explaining the failure of naive deformation invariance.

Significance and Applications
The paper claims significance in two primary areas of algebraic geometry and mathematical physics:

Cosection Localised Virtual Cycles: The results provide a practical toolkit for computing the degrees of cosection localised virtual cycles, which are used to define enumerative invariants when the moduli space has a cosection vanishing on a locus of expected dimension. The paper shows that these invariants can be zero even for non-empty moduli spaces.
DT4 Virtual Cycles: The work applies to the definition of Donaldson-Thomas invariants for Calabi-Yau 4-folds (DT4 theory). In this context, the moduli space of stable sheaves carries a $(-2)$ -shifted symplectic structure. The paper notes that for general DT4 moduli spaces, the virtual count can be zero or negative, behaving more like a real count (where $x^2-a=0$ can have 0 or 2 solutions) than a complex count. However, in the special case where the moduli space is a $(-2)$ -shifted cotangent bundle (e.g., local Calabi-Yau 4-folds), the theory recovers the standard Behrend-Fantechi virtual cycle, behaving like a complex count.

The authors conclude that the "moral" of the work is that counting in the presence of a cosection (as in DT4 theory) is fundamentally different from standard complex enumerative geometry, often yielding zero or negative contributions, and that the orthogonal Hopf index provides the correct framework to understand and compute these phenomena.