- The paper establishes exact boundedness and compactness criteria for operators
H_g:H^p
H^2 across all 1p, with mean Lipschitz conditions governing 1p2 and the mixed-norm space A^{2,\tilde p}_{\tilde p/2} governing 2p.
- For every 2<p<, boundedness from H^p to H^2 is equivalent to compactness and to g'A^{2,\tilde p}_{\tilde p/2}, while the H^1 endpoint is characterized by g'H^p, with explicit two-sided operator-norm estimates.
- The results show that g(p,1/p) is insufficient for boundedness when p>2 and completely resolve the H^p boundedness problem for lacunary symbols through Zygmund-type norm comparisons.
The paper "The Boundedness Problem for Generalized Hilbert Operators on Hardy Spaces" (2608.19086) by Norrbo, Peláez, and Wu advances the study of generalized Hilbert operators Hg(f)(z)=∫01f(t)g′(tz)dt on Hardy spaces Hp. The central contribution is an exact characterization of boundedness and compactness of Hg:Hp→H2 for all 1≤p≤∞, which the authors then exploit to show that the classical mean Lipschitz condition g∈Λ(p,1/p) is far from sufficient for boundedness of Hg on Hp when p>2, and to resolve the long-standing open problem completely within the class of lacunary symbols.
Background and context
For a symbol g∈H(D), the operator Hg generalizes the classical Hilbert matrix operator, recovered when Hp0. The foundational work of Galanopoulos–Girela–Peláez–Siskakis established that Hp1 is necessary for boundedness of Hp2 on Hp3 for all Hp4, and sufficient when Hp5. For Hp6, sufficiency fails: Guo and Tang (Guo et al., 30 Jul 2026) constructed symbols in Hp7 for which Hp8 is unbounded on Hp9. A complete characterization in this range has remained open. The present paper sharpens that negative result considerably and identifies a mixed-norm condition that is both necessary and sufficient in important special cases.
Boundedness from Hg:Hp→H20 to Hg:Hp→H21
The first main theorem gives a clean characterization at the endpoint domain Hg:Hp→H22: for Hg:Hp→H23, boundedness of Hg:Hp→H24 is equivalent to compactness, and both hold if and only if Hg:Hp→H25. For the codomain Hg:Hp→H26, boundedness is equivalent to Hg:Hp→H27, while compactness requires the stronger condition Hg:Hp→H28, the disc algebra. Moreover, the operator norm satisfies two-sided bounds
Hg:Hp→H29
The proof of necessity uses reproducing kernels 1≤p≤∞0, normalized in 1≤p≤∞1, together with the uniform-on-compacta convergence 1≤p≤∞2 as 1≤p≤∞3. Compactness follows from polynomial approximation of 1≤p≤∞4; the failure of automatic compactness into 1≤p≤∞5 reflects the non-density of polynomials there. A remark notes that the coefficient condition 1≤p≤∞6, which characterizes the associated Hadamard multiplier from 1≤p≤∞7 to 1≤p≤∞8, is strictly weaker than 1≤p≤∞9 — so boundedness of g∈Λ(p,1/p)0 is sufficient but not necessary for the corresponding Hadamard product to be bounded.
Boundedness from g∈Λ(p,1/p)1 to g∈Λ(p,1/p)2 for g∈Λ(p,1/p)3
In the sub-g∈Λ(p,1/p)4 range, the paper proves that g∈Λ(p,1/p)5 is bounded exactly when g∈Λ(p,1/p)6, with norm equivalence
g∈Λ(p,1/p)7
and compactness holds precisely when g∈Λ(p,1/p)8 belongs to the little-oh space g∈Λ(p,1/p)9. The sufficiency argument combines moment estimates Hg0 with dyadic block decompositions of Hg1 and the known Hg2-boundedness of the sublinear operator Hg3. Necessity is obtained by testing against explicit kernel-like functions Hg4 whose moments are uniformly large on dyadic blocks Hg5, forcing the dyadic norms Hg6 to satisfy the mean Lipschitz decay. This extends the previously known case Hg7 and shows that in this range the target-Hg8 problem is governed entirely by the same Lipschitz condition that governs boundedness on Hg9.
Boundedness from Hp0 to Hp1 for Hp2
The situation changes qualitatively above Hp3. Setting Hp4, the authors prove that for Hp5 the following are equivalent: boundedness of Hp6, its compactness, and the membership
Hp7
an analytic weighted mixed-norm space with norm comparable to Hp8. Furthermore,
Hp9
Notably, boundedness and compactness coincide here — there is no gap between them, in contrast with the p>20 case. The proof of necessity is the technical core: it employs a lemma producing test functions p>21 built from dilated symbols p>22, for which p>23 is controlled by p>24 while p>25 dominates the full symbol norm. Letting p>26 yields the lower bound. Sufficiency uses Hölder interpolation between the p>27-weighted dyadic sums of p>28 and the p>29 structure of the moments of g∈H(D)0.
Consequences: counterexamples and lacunary symbols
Two significant consequences follow. First, since one can exhibit lacunary series such as g∈H(D)1 belonging to g∈H(D)2 but with g∈H(D)3, the authors obtain, for every g∈H(D)4, a symbol g∈H(D)5 for which g∈H(D)6 is not even bounded from g∈H(D)7 into g∈H(D)8. This strengthens the earlier result of Guo and Tang, who had only shown failure of boundedness on g∈H(D)9 itself; the implication is that the necessary condition Hg0 lies strictly below any sufficient condition in the supercritical range, quantifying how far it is from characterizing boundedness.
Second, for lacunary symbols Hg1 with Hg2, the open problem is solved completely: Hg3 is bounded on Hg4 (Hg5) if and only if Hg6, with norm equivalence. The key observation is that Hg7 remains lacunary with the same step whenever Hg8 is analytic, so Zygmund's theorem comparing Hg9 norms of lacunary series reduces the Hp00 question to the already-characterized Hp01 case. The authors also observe that for lacunary Hp02 the conditions Hp03 and Hp04 coincide, whereas for general symbols the latter condition (shown sufficient but not necessary by Guo–Tang) is strictly stronger.
Limitations and open questions
The principal limitation is inherent to the subject: for general (non-lacunary) symbols and Hp05, boundedness of Hp06 on Hp07 itself remains uncharacterized. The mixed-norm condition Hp08 fully resolves the problem only for the target space Hp09 and, via lacunarity, for symbols with spectral gaps. Whether some modification of this condition, or an interpolation-theoretic framework bridging the Hp10-target result and the Lipschitz obstruction, yields a full characterization on Hp11 is left open. Additionally, the equivalence of boundedness and compactness established for Hp12 does not extend to all codomains, as the Hp13 case demonstrates, and the paper does not address Schatten-class membership or weighted Bergman analogues beyond citing prior work.
Conclusion
This paper delivers exact norm equivalences for generalized Hilbert operators mapping into Hp14 across the full range of source Hardy spaces, establishes sharp endpoint results from Hp15, and produces the strongest available counterexamples showing insufficiency of the mean Lipschitz condition for Hp16. By resolving the boundedness problem for lacunary symbols and identifying the mixed-norm space Hp17 as the correct governing condition in the supercritical regime, the work delineates precisely where current techniques succeed and isolates the remaining difficulty in the general Hp18 case.