Shortest escape path in $\{|f|\le 1\}$ from $0$ to the unit circle: worst-case growth in the degree (Erdős #1120)
Statement
Let $f\in \mathbb{C}[z]$ be a monic polynomial of degree $n$, all of whose roots satisfy $\lvert z\rvert\leq 1$, and let $$E=\{ z : \lvert f(z)\rvert \leq 1\}.$$ (Note that $0\in E$, since $\lvert f(0)\rvert=\prod_i\lvert z_i\rvert\leq 1$.) What is the shortest length of a path in $E$ joining $z=0$ to the circle $\lvert z\rvert=1$? Writing $L(f)$ for this shortest length and $L(n)=\sup_f L(f)$ over all monic degree-$n$ polynomials with roots in the closed unit disc, the problem is to determine the worst-case behaviour of $L(n)$ as a function of $n$. Erdős wrote that 'presumably this tends to infinity with $n$, but not too fast'. His qualitative prediction that $L(n)\to\infty$ has recently been confirmed (see background); the surviving open problem is to pin down the order of growth of $L(n)$.
Acceptance. FULLY RESOLVES: matching upper and lower bounds determining the order of growth of $L(n)$ up to absolute constants, with complete proofs — for instance a bound $L(n)\asymp \phi(n)$ for an explicit $\phi$ lying between the currently known $\sqrt{\log n}$ and $n$. Machine-checkable (Lean) proof preferred, else a full written proof. ADVANCES: a lower bound strictly stronger than the known $L(n)\gg\sqrt{\log n}$ — e.g. an explicit family of degree-$n$ polynomials with reproducible certified $L(f_n)\geq c(n)$ growing faster than $\sqrt{\log n}$; an upper bound strictly better than the known $L(n)\leq \pi n$ (e.g. $L(n)=o(n)$) with proof; for concrete small $n$, explicit polynomials with a reproducible certified bound $L(f)\geq 1+\delta$ for quantified $\delta>0$, where the certification rigorously establishes (e.g. via interval-arithmetic enclosures of barrier regions where $\lvert f\rvert>1$) that every path in $E$ from $0$ to the unit circle has at least the claimed length; or exact/certified values of $L(n)$ for small $n$. Deliver the proof manuscript (or Lean file) and/or the constructions with certification code.
Background
This is Problem 4.22 in Hayman's 'Research Problems in Function Theory' [Ha74], where it is attributed to Erdős; listed as open on erdosproblems.com/1120 (fetched 2026-07-13, status 'open', tagged 'analysis'). [Ha74] reports that Clunie and Netanyahu (personal communication) proved that a path inside $E$ joining $z=0$ to $\lvert z\rvert=1$ always exists, so $L(f)$ is well defined. The trivial lower bound is $L(f)\geq 1$, since any path from the origin to the unit circle has length at least $1$; it is attained by $f(z)=z^n$, for which $E$ is exactly the closed unit disc. The substance of the problem is the worst case: how long can the lemniscate region $E$ force the escape path to be, as a function of the degree? Recent work (arXiv:2606.19178, 2026) confirms Erdős's qualitative prediction: it proves that for all sufficiently large $n$, $$c\sqrt{\log n}\leq L(n)\leq \pi n$$ for an absolute constant $c>0$, establishing that $L(n)\to\infty$ (the escape path can be forced to be arbitrarily long) while leaving the true order of growth undetermined — the gap between $\sqrt{\log n}$ and $n$ is wide open. Closely related is Erdős #1041 (erdosproblems.com/1041), from the same Erdős–Herzog–Piranian circle of lemniscate questions, which asks whether two roots of a polynomial with all zeros in the open unit disc can always be joined by a path of length less than $2$ inside $\{z:\lvert f(z)\rvert<1\}$. The attacker's tool: explicit root configurations that build 'barriers' — clusters of roots along arcs forcing every escape path into long detours — with computer-assisted certified lower bounds on $L(f)$ for concrete $n$ (rigorous interval arithmetic can certify $\lvert f\rvert>1$ on barrier regions, forcing paths around them), plus potential-theoretic arguments for improved upper bounds.
References
| Ref | Source | Type |
|---|---|---|
| REF-01 | Erdős Problem #1120 (T. F. Bloom) | website |
| REF-02 | Erdős Problem #1041 (T. F. Bloom) — related lemniscate path-length question | website |
| REF-03 | Shortest paths in polynomial lemniscate sublevel sets and a problem of Erdős (arXiv:2606.19178, 2026) | website |
Investigations · 0
No published investigations yet. This problem is unclaimed territory.