The Secant Conjecture in the Real Schubert Calculus

Luis David García-Puente, Nickolas Hein, Christopher J. Hillar, Abraham Martín del Campo, James Ruffo, Frank Sottile, Zach TeitlerExp. Math 2012 · DOI

TL;DR

This paper introduces and provides strong evidence for the "Secant Conjecture," which posits that certain geometric intersections in the real Schubert calculus always yield real solutions. The authors support this conjecture with extensive computational experiments, revealing a fascinating "inner border" phenomenon related to the overlap of geometric configurations.

What Problem Does This Paper Address?

The paper delves into the realm of real algebraic geometry, specifically focusing on the number of real solutions to systems of polynomial equations that arise from geometric problems. A central theme in this area is understanding when such systems, which often have complex solutions, are guaranteed to have only real solutions. This is particularly relevant in the context of Schubert calculus, a powerful tool for enumerating geometric objects like linear subspaces with specific intersection properties.

The authors build upon existing conjectures like the Shapiro Conjecture, which deals with conditions for real polynomials and is related to osculating flags of a rational normal curve. The Secant Conjecture, the main subject here, generalizes this by considering "secant" flags, which are formed by planes that intersect the rational normal curve at multiple points. The core question is whether these secant configurations also lead to exclusively real solutions for the corresponding geometric intersection problems.

Methods

The paper's primary method for supporting the Secant Conjecture is through extensive computational experimentation. They leverage symbolic exact arithmetic and specialized software (Singular and Maple) to compute the number of real solutions for numerous instances of Schubert problems. This approach allows them to test specific cases of the conjecture and gather statistical evidence.

They also employ theoretical justifications by relating the Secant Conjecture to established results like the Shapiro Conjecture and by analyzing special cases. For instance, they prove results for configurations involving arithmetic progressions of secancy. Furthermore, they introduce a novel statistic called the "overlap number" to quantify the degree of intersection between the intervals defining the secant flags. This statistic proves crucial for organizing and understanding their experimental data.

Datasets

The paper does not introduce new datasets in the traditional machine learning sense. Instead, it presents the results of a large-scale computational experiment. The data generated from this experiment is made available online [40]. This experiment involved computing the number of real solutions for millions of instances of Schubert problems, categorized by factors like the type of flags (secant vs. osculating) and the overlap number.

Key Insights and Ideas

  • The Secant Conjecture: The central idea is that geometric configurations defined by secant flags (planes intersecting a rational normal curve at multiple points) should, like osculating flags, lead to systems with only real solutions.
  • The "Inner Border" Phenomenon: A striking observation from the computational data is the appearance of an "inner border" in tables of results. This suggests that for small overlap numbers between secant intervals, there's a minimum number of real solutions, and instances with zero or very few real solutions are absent. This hints at deeper geometric constraints.
  • Overlap Number as a Key Statistic: The introduction of the overlap number provides a way to systematically study and categorize the vast number of possible configurations of secant flags, making the experimental results more interpretable.
  • Connection to Shapiro and Monotone Conjectures: The Secant Conjecture is presented as a natural extension and generalization of existing conjectures, strengthening its plausibility.

Key Results

  • The paper provides strong computational evidence supporting the Secant Conjecture. For millions of tested instances, the conjecture held true.
  • The Generalized Secant Conjecture (which includes cases with some osculating flags) also held in every computed instance.
  • The overlap number was found to be a crucial factor in determining the number of real solutions. For small overlap numbers, a minimum number of real solutions was consistently observed, creating the "inner border" effect.
  • In specific Schubert problems, such as the one involving four secant lines, the authors were able to provide a topological explanation for why certain configurations lead to real solutions.

Limitations and Future Work

  • Computational Limits: While the experiment was massive, it's still a finite set of instances. The authors acknowledge that "computations are possible because Schubert problems are readily modeled on a computer, and for those of moderate size, we may algorithmically determine the number of real solutions with software tools." However, for very complex problems, the number of configurations is astronomical, making exhaustive study impossible.
  • The "Inner Border" Mystery: The authors express doubt about formulating a precise conjecture for the "inner border" phenomenon based on their current measure of overlap, suggesting it might be a complex geometric feature.
  • Future Work: The paper suggests exploring the "inner border" further and investigating the underlying geometric reasons for these observed lower bounds on real solutions. They also mention the existence of a Generalized Cosecant Conjecture, which is the dual statement to the Generalized Secant Conjecture, and its corresponding theoretical implications.

Summary updated

BibTeX
@article{garciapuente2012secant,
  author = {Luis David García-Puente and Nickolas Hein and Christopher J. Hillar and Abraham Martín del Campo and James Ruffo and Frank Sottile and Zach Teitler},
  title = {{The Secant Conjecture in the Real Schubert Calculus}},
  journal = {Exp. Math},
  year = {2012},
  doi = {10.1080/10586458.2012.661323},
}
Figure