---
title: Periodic Hankel Operators
url: https://www.emergentmind.com/topics/periodic-hankel-operators
type: topic
---

# Periodic Hankel Operators

Searching arXiv for recent and foundational papers on periodic Hankel operators across half-line, circle/Hardy-space, and spectral-density settings.
{"query":"periodic Hankel operators Floquet Bloch band spectrum site:arxiv.org"}
{"query":"arXiv 2509.25010 Ergodic Hankel operators periodic Hankel operators"}
Periodic Hankel operators are Hankel operators endowed with an additional periodic structure, but the phrase is used in several closely related senses. In the half-line theory, the periodicity is multiplicative: the kernel has the form
\[
(Hf)(t)=\int_0^\infty h(t+s)f(s)\,ds,\qquad h(t)=\frac{P(\log t)}{t},
\]
with \(P\) periodic in \(\log t\), so \(H\) commutes with a discrete dilation. In the circle and Hardy-space literature, the same phrase refers to Hankel operators on \(\mathbb T\), where periodicity is built into the underlying domain and the operator is generated by Fourier data or Hardy symbols. A broader harmonic-analytic generalization places these constructions on compact connected Abelian groups with ordered duals [2509.25010], [1402.1716], [2208.06215], [1611.06564].

## 1. Terminology and operator models

The common structural feature is the Hankel dependence on a sum of variables or indices. On \(L^2(\mathbb R_+)\), the kernel depends on \(t+s\); on \(\ell^2(\mathbb Z_+)\), the matrix entries depend on \(j+k\); on Hardy space \(H^2(\mathbb T)\), the operator is realized by projecting a product onto anti-analytic modes. In the half-line framework developed for ergodic and periodic Hankel operators, bounded self-adjoint Hankel operators are realized as
\[
(Hf)(t)=\int_0^\infty h(t+s)f(s)\,ds,
\]
with boundedness ensured, for example, by \(|h(t)|\le C_h/t\), and self-adjointness equivalent to \(h\) being real-valued [2509.25010].

A distinct but standard periodic model is the circle Hardy-space operator
\[
H_u(h)=\Pi(u\overline h),\qquad h\in L^2_+(\mathbb T),
\]
which is antilinear, or its linear companion
\[
H_\psi f=P_-(\psi f),
\]
acting from \(H^2\) to \(L^2\ominus H^2\). In matrix form, these correspond to classical Hankel matrices constant on anti-diagonals [1402.1716], [2205.09105]. On \(\mathbb T\), the symbol-generated Hankel matrix is
\[
\Gamma(\widehat\omega)=\{\widehat\omega(j+k)\}_{j,k\ge 0},
\]
unitarily equivalent to the Hardy-space operator
\[
H(\omega)f=P_+\omega J P_+f,
\]
with self-adjointness characterized by
\[
\omega(v)=\overline{\omega(\overline v)}.
\]
This circle model is explicitly treated as a periodic Hankel setting in the spectral-density theory for piecewise continuous symbols [1903.11572].

A plausible implication is that “periodic Hankel operator” is not a single universal definition but a family of constructions unified by Hankel symmetry together with periodicity of the ambient variable, the logarithmic variable, or the generating symbol.

## 2. Dilation-periodic operators on the half-line

The modern half-line theory identifies periodicity with discrete dilation covariance. If
\[
h(t)=\frac{P(\log t)}{t},\qquad P(\xi+\tau)=P(\xi),
\]
then the corresponding Hankel operator commutes with
\[
(D_\tau f)(t)=e^{\tau/2}f(e^\tau t),
\]
that is,
\[
HD_\tau=D_\tau H.
\]
After the unitary logarithmic change of variables
\[
(Ef)(\xi)=e^{\xi/2}f(e^\xi),
\]
the transformed operator \(K=EHE^*\) acts on \(L^2(\mathbb R)\) and commutes with translation by \(\tau\):
\[
KU_\tau=U_\tau K.
\]
This makes Floquet–Bloch theory available in direct analogy with periodic Schrödinger operators [2509.25010].

The smooth periodic class is defined by the Fourier coefficients
\[
\widetilde P_m=\frac1\tau\int_0^\tau e^{-im\frac{2\pi}{\tau}\xi}P(\xi)\,d\xi,
\]
with the summability condition
\[
\sum_{m=-\infty}^\infty |\widetilde P_m|\,(1+|m|)^{1/2}<\infty.
\]
Under this assumption, the fiber operators in the Floquet decomposition are trace class [2509.25010].

The fiber Hilbert space is \(\ell^2(\mathbb Z)\), and for a bounded self-adjoint operator commuting with \(U_\tau\) one has
\[
UAU^*=\int_{(-\pi/\tau,\pi/\tau)}^\oplus A(k)\,dk,
\]
with
\[
[(Uf)(k)]_n=\widehat f\!\left(k+\frac{2\pi}{\tau}n\right).
\]
Applied to a periodic Hankel operator, this yields
\[
UKU^*=\int_{(-\pi/\tau,\pi/\tau)}^\oplus K(k)\,dk.
\]
A concrete fiber formula is available:
\[
[K(k)]_{n,m} = B\!\left(\frac12-i\frac{2\pi}{\tau}n-ik,\ \frac12+i\frac{2\pi}{\tau}m+ik\right)\widetilde P_{n-m},
\]
or equivalently in the factorized Gamma-form recorded in the paper. This formula is central because it converts the dilation-periodic Hankel operator into a compact-operator family parametrized by quasi-momentum \(k\) [2509.25010].

An earlier treatment of essentially the same class establishes the same Floquet–Bloch framework with period \(T\), writing \(h(t)=p(\log t)/t\) and fiber matrices through the Beta function. There the key observation is that logarithmic variables convert discrete dilations into translations, so Mellin analysis plays the role normally played by Fourier analysis in periodic operator theory [2307.09242].

## 3. Band spectra, flat bands, and the integrated density of states

For smooth periodic Hankel operators on the half-line, the nonzero spectrum is described by analytic band functions. There exists a finite or countable list of nonvanishing real-analytic functions \(E(k)\), called band functions, representing all nonzero fiber eigenvalues with multiplicity encoded by repetition. The list splits into non-flat and flat bands. Non-flat bands have multiplicity one and satisfy
\[
E'(k)\neq 0,\qquad 0<k<\pi/\tau,
\]
while flat bands are constant. Distinct non-flat bands have disjoint interiors, and no non-flat band intersects the negative of another [2509.25010].

The integrated density of states for smooth periodic Hankel operators admits a complete decomposition. If \(\nu\) denotes the IDS measure, then
\[
\nu=\nu^{\mathrm{pp}}+\nu^{\mathrm{ac}},\qquad \nu^{\mathrm{sc}}=0.
\]
The absolutely continuous part has the form
\[
\nu^{\mathrm{ac}}=\frac1\tau\sum_n \nu_n,
\]
where each \(\nu_n\) is a purely absolutely continuous probability measure supported on a closed bounded interval \(\sigma_n\) separated away from \(0\). The pure point part has the form
\[
\nu^{\mathrm{pp}}=\frac1\tau\sum_n(\delta_{\lambda_n}+\delta_{-\lambda_n}),
\]
with \(\lambda_n\to 0\) if the family is infinite; the points \(\pm\lambda_n\) are eigenvalues of infinite multiplicity and are interpreted as flat spectral bands. If the operator is positive, then \(\nu^{\mathrm{pp}}=0\), so the IDS is purely absolutely continuous [2509.25010].

The support of the IDS equals the deterministic spectrum:
\[
\operatorname{supp}\nu=\sigma(K_\omega).
\]
Thus periodic half-line Hankel operators have a deterministic band spectrum consisting of absolutely continuous bands, possibly together with flat bands at isolated nonzero eigenvalues, and with \(0\) as the only possible accumulation point if infinitely many bands occur [2509.25010].

The band picture was already developed in detail in the earlier band-spectrum analysis. There the absolutely continuous spectrum has multiplicity \(2\), the singular continuous spectrum is absent, and nonzero eigenvalues have infinite multiplicity. A particularly distinctive feature is that flat bands may coexist with non-flat bands, unlike the standard one-dimensional periodic Schrödinger picture. The paper’s Mathieu–Hankel operator provides an explicit example: there exists \(A_*\in(0,1)\) such that \(\mathbf H(A_*)\) has at least one flat band and at least one non-flat band [2307.09242].

The periodic IDS also satisfies a gap-labelling statement: in a spectral gap,
\[
\int_\lambda^\infty \nu(dx)
\]
is constant and equals \(N/\tau\), where \(N\) is the number of bands above \(\lambda\), counting both flat and non-flat bands [2509.25010].

## 4. Periodic Hankel operators on the circle

On the circle, periodicity is intrinsic. The Hardy-space model takes
\[
L^2_+(\mathbb T)=\{u\in L^2(\mathbb T): \hat u(n)=0\ \text{for }n<0\},
\]
and defines the compact Hankel operator
\[
H_u(h)=\Pi(u\overline h),\qquad u\in VMO_+(\mathbb T),
\]
with compactness equivalent to \(u\in VMO_+(\mathbb T)\). The shifted operator
\[
K_u=S^*H_u=H_uS=H_{S^*u}
\]
is essential because the inverse spectral problem is controlled by the pair \((H_u,K_u)\), not by \(H_u\) alone. The relation
\[
K_u^2=H_u^2-(\cdot\mid u)u
\]
drives interlacing and multiplicity phenomena [1402.1716].

For compact periodic Hankel operators on the circle, singular values with arbitrary multiplicities are classified by a nonlinear Fourier transform. The complete spectral data consist of the interlaced singular values of \(H_u\) and \(K_u\), together with one finite Blaschke product \(\Psi_r\) for each singular value \(s_r\). The map
\[
\Phi:u\mapsto ((s_r),(\Psi_r))
\]
is bijective, and in finite rank its inverse is explicit. A singular value of multiplicity \(m\) contributes a Blaschke product of degree \(m-1\) [1402.1716].

The same circle setting also supports asymptotic spectral analysis for noncompact symbol-defined periodic Hankel matrices. For \(\omega\in PC(\mathbb T)\), the Hankel matrix
\[
\Gamma(\widehat\omega)=\{\widehat\omega(j+k)\}_{j,k\ge 0}
\]
has truncations whose logarithmic spectral density depends only on the jump half-heights
\[
\varkappa_z(\omega)=\frac{\omega(z+)-\omega(z-)}{2}.
\]
In the non-self-adjoint case,
\[
\operatorname{LogDens}_{\tau}(t;\Gamma(\widehat{\omega}))
=
\sum_{z\in\Omega}\mathsf c\!\left(t|\varkappa_z(\omega)|^{-1}\right),
\]
and in the self-adjoint case the positive and negative densities split according to conjugate jump pairs and the special points \(z=\pm1\). The result is universal with respect to the truncation scheme under the paper’s assumptions: square truncation, Abel–Poisson regularization, and other admissible Schur–Hadamard multipliers give the same asymptotic logarithmic density [1903.11572].

A more geometric analysis of circle Hankel operators studies Schmidt subspaces
\[
E_s^+(H_\psi)=\ker(H_\psi^*H_\psi-s^2I)
\]
for the linear Hardy-space Hankel operator \(H_\psi f=P_-(\psi f)\). For the top singular value \(s=\|H_\psi\|\), the Schmidt subspace is exactly the kernel of a Toeplitz operator. More generally,
\[
E_s^+(H_\psi)\subset \theta\cdot \ker T_{\theta \nu_s}\subset E_s^+(H_{s\nu_s}),
\]
and if \(E_s^+(H_\psi)\not\perp 1\), then it is nearly \(S^*\)-invariant [2205.09105]. This does not define periodicity, but it gives structural information for periodic circle symbols as a subclass.

## 5. Group-theoretic and symbol-theoretic extensions

The circle theory extends to compact connected Abelian groups \(G\) with linearly ordered dual \(X\). The Hardy spaces are defined by Fourier support in a positive cone \(X_+\):
\[
H^2(G)=\{f\in L^2(G): \widehat f(\chi)=0 \ \forall \chi\in X_-\},
\]
and the classical Hankel operator is
\[
H_\varphi f=P_-(\varphi f).
\]
When \(G=\mathbb T\), this recovers the usual periodic Hardy-space Hankel operator. In this framework, boundedness is characterized by \(L^\infty\)-symbol extension or, equivalently, by \(P_-\varphi\in BMO(G)\); compactness is characterized by \(P_-\varphi\in K_1(G)\) when the dual has a least positive element; and bounded Hankel operators are not left Fredholm [1611.06564].

A different generalization introduces \((\mu;\nu)\)-Hankel operators on Hardy spaces over compact Abelian groups. They are defined by
\[
(A^{(\mu;\nu)}_a \chi,\bar\xi)=\mu(\chi)\nu(\xi)a(\chi\xi),
\qquad \chi\in X_+,\ \xi\in X_+\setminus\{x_0\}.
\]
For \(G=\mathbb T\), this becomes
\[
(A z^k, z^{-j})=\mu_k\nu_j a_{k+j},
\]
which is exactly the weighted Hankel anti-diagonal pattern. If \(|\mu(\chi)|=1\) for all \(\chi\in X_+\), every bounded \(\mu\)-Hankel operator is of the form
\[
A=H_\varphi U
\]
for a unitary translation \(U\) and a classical Hankel operator \(H_\varphi\) [2208.06215].

For symbol-defined Hankel matrices on \(\ell^p\), the Fredholm and index theory is most naturally formulated inside the Toeplitz-plus-Hankel algebra \(TH(PC_p)\). There, the local symbol of
\[
T(a)+H(b)
\]
is a \(2\times 2\) matrix away from \(t=\pm1\) and a scalar at \(t=\pm1\), with the Hankel contribution entering through the jump differences \(b(t+)-b(t-)\). The main message for structured or periodic symbol-generated Hankel matrices is that Fredholmness is governed explicitly by local jump data of the symbol, and the index is a winding number of the resulting normalized symbol. The same paper also notes that a pure Hankel operator \(H(b)\) alone can only be Fredholm in very exceptional circumstances within this algebraic setting, because the Toeplitz part \(a\) must be invertible [1112.3140].

## 6. Explicit examples, characteristic phenomena, and scope

The periodic half-line theory has several explicit examples that isolate genuinely Hankel features. If
\[
\Sigma=\sum_{n\in\mathbb Z}\delta_{\tau n},
\]
then the fibers have rank one and there is a single nonzero band
\[
E_0(k)=\frac1\tau\sum_{n=-\infty}^\infty \pi\,\sech\!\left(\pi\left(\frac{2\pi}{\tau}n+k\right)\right),
\]
so the IDS is absolutely continuous on \([E_{\min},E_{\max}]\); because the extrema are nondegenerate, the density has square-root singularities near the endpoints [2509.25010].

A signed periodic measure gives the opposite extreme. For
\[
\Sigma=\sum_{n=-\infty}^\infty\left(\delta_{\tau n}-\delta_{\tau/2+\tau n}\right),
\]
the fibers are rank two and their eigenvalues are \(\pm E_*(k)\), but an elliptic-function identity makes \(E_*(k)\) independent of \(k\). Hence both bands are flat, the spectrum consists of
\[
-E_*,\ 0,\ E_*,
\]
and the IDS is pure point, supported at \(\pm E_*\). This is the explicit example showing that, without positivity, pure point IDS can occur [2509.25010].

The earlier band-spectrum paper gives further explicit models. For the Carleman operator, all bands are positive and non-flat, filling \([0,\pi]\) with multiplicity \(2\). For the Mathieu–Hankel operator with
\[
p(\xi)=A+\cos(\omega\xi),
\]
one has all bands flat when \(A=0\), infinitely many positive and negative bands for \(A\in(0,1)\), positivity of all fibers for \(A\ge 1\), and existence of a parameter \(A_*\in(0,1)\) for which flat and non-flat bands coexist [2307.09242].

Taken together, these developments show that periodic Hankel operators form a spectral class parallel to periodic Schrödinger operators but not reducible to them. The Floquet–Bloch picture, spectral bands, and IDS are familiar periodic-operator features; the coexistence of flat and dispersive bands, the possibility of infinite-multiplicity eigenvalues at nonzero energies, and the natural role of logarithmic variables and Mellin analysis are specifically Hankel phenomena [2509.25010], [2307.09242].

Source: https://www.emergentmind.com/topics/periodic-hankel-operators