Erdős Problem #1153 — Local lower bounds for Lebesgue functions

About 65 years old · traced to

For distinct x1,…,xn∈[−1,1]x_1,\ldots,x_n\in[-1,1], let lk(x)=∏i≠kx−xixk−xil_k(x)=\prod_{i\ne k}\frac{x-x_i}{x_k-x_i} and λ(x)=∑k=1n∣lk(x)∣\lambda(x)=\sum_{k=1}^n|l_k(x)|. Is it true that, for every fixed −1≤a<b≤1-1\leq a<b\leq1,

max⁡x∈[a,b]λ(x)>(2π−o(1))log⁡n?\max_{x\in[a,b]}\lambda(x)>\left(\frac2\pi-o(1)\right)\log n?
References

Additional references

P. Erdős and P. Turán, An extremal problem in the theory of interpolation, Acta Mathematica Academiae Scientiarum Hungaricae 12 (1961), 221–234.

Progress summary

Refreshed
Claimed solved

A 2026 paper settles the question affirmatively: even on any fixed subinterval, the interpolation error must grow logarithmically at the sharp leading rate.

Erdős and Turán asked whether arbitrary interpolation nodes in [−1,1][-1,1] force the Lebesgue function to reach the sharp asymptotic size on every fixed subinterval. The answer is affirmative.

Known results

  • Bernstein, 1931: the full interval [−1,1][-1,1] has a logarithmic lower bound.
  • Erdős: globally, max⁡x∈[−1,1]λ(x)≥2πlog⁡n−O(1)\max_{x\in[-1,1]}\lambda(x)\geq\frac{2}{\pi}\log n-O(1); Chebyshev nodes show sharpness up to O(1)O(1).
  • Erdős and Szabados, 1978: for every fixed [a,b]⊆[−1,1][a,b]\subseteq[-1,1], the restricted maximum is ≫log⁡n\gg\log n, without the sharp constant.

March 2026 resolution

Tao's result proves that for every fixed nontrivial interval I⊂[−1,1]I\subset[-1,1] and arbitrary distinct nodes, sup⁡x∈Iλ(x)≥2πlog⁡n−o(log⁡n)\sup_{x\in I}\lambda(x)\geq\frac{2}{\pi}\log n-o(\log n). This directly proves Problem #1153 and also establishes an integral lower bound over II.

Current status (as of March 2026): The problem is resolved affirmatively by a published arXiv theorem, with the sharp leading constant 2π\frac{2}{\pi}; no substantive part remains open.

Sources

Solutions 0

No solutions have been posted yet.