In classical algebraic geometry, Weierstrass points are special locations on a smooth curve that "see" the curve's geometry more intensely than others. While well-understood for complex curves, their "tropical" counterparts—defined on metric graphs—present a significant challenge. A major hurdle in tropical geometry is that the set of Weierstrass points for a given divisor can sometimes be infinite (an "uncountably large" set of points), which breaks the analogy with the finite sets found in classical geometry.
This research seeks to restore order by answering two main questions:
- Under what conditions is the set of tropical Weierstrass points finite, and how many are there?
- How do these points spread out as the "degree" (complexity) of the divisor approaches infinity?
Methods: Stability and Electrical Analogies
To bypass the problem of infinite point sets, the paper introduces a "stable" version of the Weierstrass locus ($W^{st}(D)$). This set is guaranteed to be finite for any divisor and coincides with the standard locus for "generic" (typical) cases.
The author, David Harry Richman, employs a mix of graph theory and analogies from physics. The proof relies on an ABKS Decomposition, a sophisticated way to chop up the space of divisor classes into cells based on the spanning trees of the graph. This allows the author to treat the problem using the topology of branched covering maps.
Key Insights: Geometry as Electricity
One of the most striking findings is that the limiting distribution of these geometric points is entirely determined by how electricity would flow through the graph. The paper views the metric graph as an electrical circuit where edges have resistance proportional to their length. The points cluster in areas of high "effective resistance."
The distribution of points is linked to the Zhang canonical measure, which is defined using the graph Laplacian and effective resistances. Richman proves that these distribution properties are intrinsic to the metric graph itself, requiring no external scaffolding from more complex non-Archimedean spaces.
Key Results
- Theorem A (The Count): For a generic divisor of degree $n \ge g$, there are exactly $g(n - g + 1)$ Weierstrass points, where $g$ is the genus (number of loops).
- Theorem B (The Distribution): As the degree $n \to \infty$, the discrete measure of the Weierstrass points converges weakly to the Zhang canonical measure $\mu$.
- Theorem C (Quantitative Bounds): The author provides explicit bounds on the number of points found in any segment $e$ of the graph: $$n\mu(e) - 2g \le \#(W_n \cap e) \le n\mu(e) + g + 2$$ This shows the points are "evenly distributed" relative to the electrical measure $\mu$, with an error term that only depends on the graph's genus.
Limitations and Future Work
The paper highlights a curious discrepancy: a complex curve has $g(n - g + 1)^2$ Weierstrass points, while a tropical curve has only $g(n - g + 1)$. This suggests a specific $(n - g + 1)$-to-1 "collapsing" behavior when a smooth curve degenerates into a tropical one. Future work may involve tightening the error bounds in Theorem C and extending the results to graphs with "vertex weights" representing remaining genus.
Summary updated