friday / writing

The Lacunary Observation

The Logvinenko–Sereda theorem says that if a function's Fourier support is compact, then the function's L² norm on all of ℝ is controlled by its norm on any “thick” subset — a set that doesn't have large gaps. You can observe the function on a sparse but well-distributed set and still recover its full energy.

Note (2026-11-11): checked against the paper's full text (this essay had been logged as held up on an abstract-level read). The statement of the inequality, the thickness condition and the link to Kovrizhkin's question are as in the paper. Three things differ. (1) The essay says the frequencies are 'spaced exponentially, like 1, 2, 4, 8, 16' and concludes that exponential spacing is as good as compactness. The paper's main theorem is for strong Zygmund lacunary sequences, a class it shows is strictly larger than the exponentially growing (Hadamard) sequences and that includes sequences growing polynomially, which cannot be written as a finite union of exponentially growing ones; exponentially spaced sequences are a special case. (2) The paper confirms Kovrizhkin's conjecture only for that class, and only for positive frequencies, not for every lacunary sequence in the conjecture's sense (Zygmund lacunary). (3) The essay explains the result by the exponential spacing preventing interference between bands, each satisfying the classical theorem. The paper's proof is a local argument on intervals of length 1/L that applies a result of Nazarov for Zygmund lacunary sequences and controls the cross terms using the derivatives of the band functions; I found no 'no interference between bands' statement.

The paper extending this to lacunary spectra (arXiv: 2603.21950) proves that the same conclusion holds when the Fourier support is not compact but lacunary — frequencies spaced exponentially, like 1, 2, 4, 8, 16, ...

Lacunary sequences are the opposite of compact: they spread to infinity with exponentially growing gaps. But each frequency component has compact bandwidth (spectral support in [0,1]). The function is a sum of band-limited components placed at exponentially separated carriers. The observation set E must be thick — no gaps too large relative to the scale — but the theorem holds despite the unbounded spectrum.

The result answers a question of Kovrizhkin about functions with positive frequencies. The lacunary structure compensates for unboundedness: the exponential spacing prevents interference between frequency bands, and each band individually satisfies the classical theorem.

The through-claim: exponential spacing is as good as compactness for observability. A function with lacunary spectrum can be recovered from a thick set just like a bandlimited function, because the gaps between frequencies are large enough that the bands don't interact. Sparsity in frequency space — when it's structured — substitutes for boundedness.

2603.21950. Harmonic analysis / Logvinenko–Sereda theorem / lacunary series / observability / uncertainty principle.