[Paper Review] On Zeroes of Random Polynomials and Applications to Unwinding
This paper studies the asymptotic distribution of solutions to the equation $ p_n(z) = p_n(0) $, where $ p_n(z) $ is a random polynomial with roots drawn i.i.d. from a compactly supported, continuous probability measure $ \mu $ on $ \mathbb{C} $. It shows that as $ n \to \infty $, the roots of $ p_n(z) - p_n(0) = 0 $ converge in distribution to a measure $ \nu $ that reproduces $ \mu $ on the set $ A $ where the logarithmic potential exceeds its value at the origin, and concentrates on a curve set $ B $ where the potential is equal, with the density on $ B $ determined by the angular variation of the potential.
Let $μ$ be a probability measure in $\mathbb{C}$ with a continuous and compactly supported density function, let $z_1, \dots, z_n$ be independent random variables, $z_i \sim μ$, and consider the random polynomial $$ p_n(z) = \prod_{k=1}^{n}{(z - z_k)}.$$ We determine the asymptotic distribution of $\left\{z \in \mathbb{C}: p_n(z) = p_n(0) ight\}$. In particular, if $μ$ is radial around the origin, then those solutions are also distributed according to $μ$ as $n ightarrow \infty$. Generally, the distribution of the solutions will reproduce parts of $μ$ and condense another part on curves. We use these insights to study the behavior of the Blaschke unwinding series on random data.
Motivation & Objective
- To understand the limiting distribution of solutions to $ p_n(z) = p_n(0) $ for random polynomials with i.i.d. roots drawn from a compactly supported, continuous probability measure $ \mu $ on $ \mathbb{C} $.
- To characterize the asymptotic behavior of these solutions in terms of potential theory, particularly the logarithmic potential of $ \mu $.
- To extend known results on critical points of random polynomials to the case of level sets $ p_n(z) = p_n(0) $, especially when $ \mu $ is not radial.
- To provide a precise description of the measure $ \nu $ that arises as the weak limit of the root distribution of $ p_n(z) - p_n(0) $, including the structure on the level set $ B $ where the potential equals its value at the origin.
Proposed method
- Define the sets $ A = \{ z : \int \log|x-z| d\mu(x) > \int \log|x| d\mu(x) \} $ and $ B = \{ z : \text{equality holds} \} $, which partition the complex plane based on the logarithmic potential of $ \mu $.
- Use the law of large numbers for logarithmic potentials: $ \frac{1}{n} \sum_{k=1}^n \log|z - z_k| \to \int \log|z - x| d\mu(x) $ in probability as $ n \to \infty $.
- Apply the deviation principle to show that the level set $ \{ z : \sum_{k=1}^n \log|z - z_k| = \sum_{k=1}^n \log|z_k| \} $ converges uniformly to the set $ B $, away from the support of $ \mu $.
- Analyze the argument of the polynomial $ p_n(z) $ along the limiting curve $ C \subset B $, showing that the total variation of $ \arg p_n(z) $ over $ C $ is asymptotically linear in $ n $, leading to a regular distribution of solutions.
- Use the implicit function theorem to show that $ B $ is a $ C^1 $ curve locally, and derive the density on $ B $ as $ \frac{1}{2\pi} \left| \frac{\partial}{\partial t} \int \arg(\gamma(t) - z) d\mu(z) \right|^{-1} $, where $ \gamma $ is an arclength parametrization of $ C $.
- Verify the results via explicit computation for a union of two unit circles, showing that the limiting measure $ \nu $ reproduces $ \mu $ on $ A $ and concentrates on four new curves in $ B $, with numerical simulations confirming the prediction.
Experimental results
Research questions
- RQ1How does the distribution of solutions to $ p_n(z) = p_n(0) $ behave as $ n \to \infty $ for random polynomials with i.i.d. roots from a compactly supported, continuous measure $ \mu $?
- RQ2Under what conditions does the solution set reproduce the original measure $ \mu $, and when does it instead concentrate on a curve?
- RQ3What is the precise structure of the limiting measure $ \nu $ on the set $ B $, where the logarithmic potential of $ \mu $ equals its value at the origin?
- RQ4How does the angular variation of the potential along $ B $ affect the density of solutions on the limiting curve?
- RQ5Can the Blaschke unwinding series be analyzed using the asymptotic distribution of roots of $ p_n(z) = p_n(0) $?
Key findings
- For radial $ \mu $, the solutions to $ p_n(z) = p_n(0) $ converge in distribution to $ \mu $ as $ n \to \infty $, meaning the roots of $ p_n(z) - p_n(0) $ reproduce the original measure.
- For general non-radial $ \mu $, the limiting measure $ \nu $ equals $ \mu $ on the set $ A $, where the logarithmic potential exceeds its value at the origin, and has total mass $ 1 - \mu(A) $ supported on the level set $ B $.
- On the set $ B $, the density of solutions is given by $ \frac{1}{2\pi} \left| \frac{\partial}{\partial t} \int \arg(\gamma(t) - z) d\mu(z) \right|^{-1} $, where $ \gamma $ is an arclength parametrization of a connected component of $ B $, under regularity assumptions.
- Numerical simulations confirm that for $ \mu $ supported on two unit circles centered at 0 and 2, the solutions distribute according to $ \mu $ on $ A $ and concentrate on four curves in $ B $, with slight repulsion near the origin due to the root at 0.
- The origin is always a root of $ p_n(z) - p_n(0) $, and a repulsion effect creates a shrinking bubble around 0 with no roots, which vanishes as $ n \to \infty $.
- The expected logarithmic growth of $ |p_n(0)| $ is $ \frac{1}{2} \log 2 $, and the set of points where the logarithmic potential equals this value defines the limiting support of the solutions.
Better researchstarts right now
From reading papers to final review, dramatically reduce your research time.
No credit card · Free plan available
This review was created by AI and reviewed by human editors.