Finite-Rank Perturbations of Toeplitz Ops

• Finite-rank perturbations of Toeplitz operators are defined as operators formed by adding a finite-rank term, with classifications sensitive to the symbol class and function space.
• In the Bargmann–Fock setting, function symbols produce only the zero operator, while distributional symbols are fully characterized by finitely supported jets of delta-masses.
• In Hardy, Bergman, and block settings, finite-rank phenomena are modeled via nearly invariant subspaces, explicit factorization methods, and Mellin–Gamma conditions.

Searching arXiv for the cited papers to ground the article in current arXiv records. Finite-rank perturbations of Toeplitz operators concern operators obtained by adding a finite-rank term to a Toeplitz operator, and the subject is markedly sensitive to the underlying function space, symbol class, and operator-theoretic constraint. In the Bargmann–Fock setting, the decisive rigidity statement is that finite-rank Toeplitz operators with function symbols vanish, whereas allowing distributional symbols produces a complete classification by finitely supported jets of delta-masses (Borichev et al., 2013, Rozenblum et al., 2013). In Hardy, Bergman, harmonic Bergman, and block Toeplitz settings, finite-rank perturbations are instead described through nearly invariant subspaces, noncommutative symbol calculus, Mellin–Gamma conditions, Blaschke or Blaschke–Potapov factorizations, and finite-rank local boundary modifications (Liang et al., 2019, Le et al., 2020, Dong et al., 2014, Abhinand et al., 4 May 2026, Koekenbier et al., 15 Jun 2025).

1. Bargmann–Fock framework and Toeplitz operators with generalized symbols

In the Bargmann–Fock model, one identifies $\mathbb R^2\cong\mathbb C$ and uses the Gaussian probability measure

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$

with $dA$ the Lebesgue area measure. The Bargmann–Fock space is

$$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$

and its reproducing kernel is $K_w(z)=e^{z\bar w}$. The orthogonal projection $P:L^2(dv)\to\mathcal F^2$ acts by

$(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$

These formulas fix the canonical Toeplitz structure in the Fock space (Rozenblum et al., 2013).

For a distributional symbol $\mu\in\mathcal D'(\mathbb C)$, one introduces the weighted distribution $\tilde\mu$ by

$$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$

and for $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$0 defines the sesquilinear form

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$1

Whenever this form is bounded on $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$2, one obtains a densely defined operator

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$3

which extends by continuity to $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$4. If $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$5 is actually a function of at most Gaussian growth, this recovers the usual Toeplitz construction (Rozenblum et al., 2013).

The function-symbol theory is rigid. For $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$6, where

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$7

Borichev–Rozenblum proved that if the sesquilinear form $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$8 has finite rank on reproducing kernels, then $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$9. Equivalently, no nonzero Toeplitz operator with function symbol in $dA$0 can have finite rank (Borichev et al., 2013). This is the basic rigidity statement against which the distributional theory should be read.

2. Finite-rank theorem in the Fock space

The central classification result for finite-rank Toeplitz operators in the Fock space with distributional symbols states that if $dA$1 is such that $dA$2 is bounded on $dA$3, then $dA$4 has finite rank $dA$5 if and only if there exist points $dA$6, integers $dA$7, and constants $dA$8 such that

$dA$9

In particular, every finite-rank $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$0 is supported on finitely many points and is a finite linear combination of point-masses and their holomorphic and anti-holomorphic derivatives (Rozenblum et al., 2013).

This theorem resolves a potential misconception created by the function-symbol zero theorem. In the Fock space, finite-rank Toeplitz operators do exist, but only after enlarging the symbol class from functions to distributions. The two results are complementary rather than contradictory: function symbols in the admissible growth classes yield only the zero finite-rank operator, while distributional symbols yield precisely finitely supported jets (Borichev et al., 2013, Rozenblum et al., 2013).

The perturbative consequence is explicit. If $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$1 is a bounded Toeplitz operator with regular symbol $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$2, and $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$3 is any finite-rank operator in $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$4, then $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$5, viewed as Toeplitz with distributional symbol $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$6, must satisfy the same classification. Hence

$$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$7

where $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$8 is a finite linear combination of $$\mathcal F^2=\{f\ \text{entire}:\ \|f\|^2=\int_{\mathbb C}|f(z)|^2\,dv(z)<\infty\},$$9-terms. Conversely, adding such a $K_w(z)=e^{z\bar w}$0 produces a finite-rank upgrade of rank $K_w(z)=e^{z\bar w}$1 (Rozenblum et al., 2013). In this sense, every finite-rank perturbation is localized at finitely many points.

3. Gaussian $K_w(z)=e^{z\bar w}$2-estimates and the reduction of singularity order

The proof mechanism in the Fock-space distributional theorem is built on Gaussian-weighted solvability of the $K_w(z)=e^{z\bar w}$3-equation. The key lemma is: if $K_w(z)=e^{z\bar w}$4 and $K_w(z)=e^{z\bar w}$5 is a globally defined function satisfying

$K_w(z)=e^{z\bar w}$6

and $K_w(z)=e^{z\bar w}$7 for all $K_w(z)=e^{z\bar w}$8, then for any $K_w(z)=e^{z\bar w}$9 there is an entire function $P:L^2(dv)\to\mathcal F^2$0 with

$P:L^2(dv)\to\mathcal F^2$1

One may construct $P:L^2(dv)\to\mathcal F^2$2 by the Cauchy–Green integral

$P:L^2(dv)\to\mathcal F^2$3

and control both near-field and far-field contributions to preserve Gaussian decay (Rozenblum et al., 2013).

This estimate extends to distributions by convolution with a compactly supported mollifier. If $P:L^2(dv)\to\mathcal F^2$4 annihilates all $P:L^2(dv)\to\mathcal F^2$5, then one obtains a distributional solution $P:L^2(dv)\to\mathcal F^2$6 of $P:L^2(dv)\to\mathcal F^2$7 with one higher order of smoothness and only slightly worse Gaussian decay. The stability is formulated in the distributional topology $P:L^2(dv)\to\mathcal F^2$8 with Gaussian seminorms (Rozenblum et al., 2013).

The finite-rank theorem then follows by iteration. Starting from $P:L^2(dv)\to\mathcal F^2$9, one chooses a holomorphic polynomial $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$0 so that $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$1 still annihilates all $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$2, solves

$(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$3

multiplies again by another polynomial $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$4, and repeats. After $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$5 steps one reaches $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$6, which is an actual function in $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$7. At that point the classical finite-rank theorem for function symbols implies that the resulting function-symbol Toeplitz operator must have zero symbol. Tracing the factorization back shows that $(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$8 is supported on the common zeros of the chosen polynomials, hence on finitely many points, and admits only a finite jet at each point (Rozenblum et al., 2013).

The earlier Borichev–Rozenblum function-symbol theorem uses a different mechanism: the function

$(Pf)(z)=\int_{\mathbb C}K_z(w)\,f(w)\,dv(w).$9

has both a finite-sum entire representation and, via the integral definition, a Fourier-transform representation whose derivatives go to zero at infinity. A convex-geometric argument with Wronskians and Santalo’s inequality then forces linear dependence and finally $\mu\in\mathcal D'(\mathbb C)$0 (Borichev et al., 2013). Taken together, the two proofs show that finite-rank phenomena in the Fock space are exhausted by finitely supported distributional singularities.

4. Hardy-space kernels, near invariance, and finite defect

In the scalar Hardy space $\mu\in\mathcal D'(\mathbb C)$1, a finite-rank perturbation

$\mu\in\mathcal D'(\mathbb C)$2

has a kernel with a precise structural property: $\mu\in\mathcal D'(\mathbb C)$3 is nearly $\mu\in\mathcal D'(\mathbb C)$4-invariant with defect at most $\mu\in\mathcal D'(\mathbb C)$5, and one may take the defect space to be $\mu\in\mathcal D'(\mathbb C)$6. The proof uses the commutation $\mu\in\mathcal D'(\mathbb C)$7 and shows that if $\mu\in\mathcal D'(\mathbb C)$8 with $\mu\in\mathcal D'(\mathbb C)$9, then $\tilde\mu$0 (Liang et al., 2019).

The Chalendar–Gallardo–Partington theorem then represents such nearly $\tilde\mu$1-invariant subspaces with finite defect in terms of backward-shift-invariant subspaces of vector-valued Hardy spaces. If $\tilde\mu$2 is closed and nearly $\tilde\mu$3-invariant with defect $\tilde\mu$4, then either every $\tilde\mu$5 vanishes at $\tilde\mu$6, or every $\tilde\mu$7 admits a decomposition

$\tilde\mu$8

with $\tilde\mu$9 lying in a closed $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$0-invariant subspace $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$1 (Liang et al., 2019). For rank-one perturbations, this yields explicit model-space-type parametrizations of $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$2.

The vector-valued Hardy theory parallels the scalar case. For $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$3 and a rank-$$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$4 perturbation

$$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$5

the kernel $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$6 is nearly $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$7-invariant with defect at most $$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$8 (Chattopadhyay et al., 2020). In four symbol classes—$$\langle\tilde\mu,\phi\rangle=\langle\mu,e^{-|\cdot|^2}\phi\rangle,$$9, inner multiplier, invertible factorization $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$00, and $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$01 with $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$02 inner—the defect space can be described explicitly in terms of the perturbation data and transforms of the $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$03 (Chattopadhyay et al., 2020).

A further refinement appears for perturbations of $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$04 on vector-valued Hardy spaces when $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$05 is inner with $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$06 and $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$07. If

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$08

and $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$09 is invariant under $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$10, then $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$11 admits the representation

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$12

where $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$13 is invariant under $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$14, and the map $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$15 is unitary from $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$16 onto $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$17 (Khan et al., 8 Jul 2025). The same framework extends to almost invariant and nearly invariant subspaces with finite defect.

5. Bergman and harmonic Bergman finite-rank differences

On the Bergman space $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$18, finite-rank perturbation questions often appear not for a single Toeplitz operator but for products, commutators, and generalized semicommutators. For quasihomogeneous symbols, Le–Thilakarathna introduced a noncommutative convolution $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$19. If

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$20

then

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$21

and the main theorem states that for finite sums of quasihomogeneous functions,

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$22

has finite rank on $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$23 if and only if

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$24

with the right-hand side in $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$25 (Le et al., 2020). In the holomorphic/anti-holomorphic case, $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$26 is characterized as the unique $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$27-function on $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$28 solving a first-order PDE system together with a normalization condition (Le et al., 2020).

Dong–Zhou gave a separate complete classification of finite-rank commutators and generalized semicommutators of quasihomogeneous Toeplitz operators on the harmonic Bergman space $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$29 and on $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$30. Their criteria are expressed through Mellin–Gamma formulas and explicit arithmetic cases. On $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$31, the situation is rigid: nontrivial finite-rank commutators and generalized semicommutators are always rank $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$32, with range spanned by a single monomial (Dong et al., 2014). On $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$33, the theory is more flexible: the finite-rank cases admit canonical decompositions, explicit range descriptions, and closed-form rank formulas, and arbitrarily large finite ranks occur in cases (6)–(9) of their classification (Dong et al., 2014).

These Bergman-space results indicate a different perturbative geometry from the Fock-space theorem. In the Fock setting, finite-rank Toeplitz operators themselves are classified by delta-jets. In Bergman settings, finite-rank phenomena frequently arise as the defect between operator products and Toeplitz operators with a corrected symbol, or as commutators whose symbol data satisfy exact Mellin constraints.

6. Block Toeplitz operators, self-commutators, and finite-rank local perturbations

For scalar Toeplitz operators on $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$34, Nakazi–Takahashi established that

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$35

if and only if there exists a finite Blaschke product $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$36, and one may choose $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$37 so that $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$38 (Abhinand et al., 4 May 2026). In the block setting, if $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$39 and $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$40 contains a constant unitary matrix $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$41, then $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$42 is normal; under a mild symbol hypothesis, normality implies that $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$43 contains such a $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$44 (Abhinand et al., 4 May 2026).

A partial block-matrix analogue of the Curto–Hwang–Lee conjecture is now available. If $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$45, $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$46 is of bounded type, and $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$47 is hyponormal, then

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$48

where $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$49 and $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$50. Moreover, $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$51 may be chosen so that

$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$52

The proof proceeds through Hankel-operator formulas and Beurling–Lax theory, which converts finite-dimensional range of $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$53 into a finite Blaschke–Potapov model space (Abhinand et al., 4 May 2026).

In a matrix-asymptotic direction, local finite-rank perturbations of block Toeplitz matrices have been analyzed through a generalized Widom formula. For a block-tridiagonal Toeplitz matrix $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$54, modifying only $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$55 boundary blocks yields a perturbation $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$56 of rank at most $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$57. The continuous part of the limit spectrum depends only on the rank $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$58 of the perturbation, while the outliers depend continuously on the local perturbation data $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$59 (Koekenbier et al., 15 Jun 2025). This separates universal rank-class effects from perturbation-specific discrete spectral motion.

Related rank-$dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$60 perturbation models for nonnormal Toeplitz matrices replace the Toeplitz part by $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$61 or $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$62 and add $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$63, where $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$64 is the all-ones matrix. In these models, the perturbation creates explicit nonzero eigenvalue equations, a defective zero eigenvalue, and a pseudospectral cloud whose geometry is organized by the symbol curves of the unperturbed Toeplitz operators (Morimoto et al., 2024). In a different operator-theoretic application, finite-rank perturbations $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$65 of semi-infinite Toeplitz operators with $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$66 arise from finite-difference boundary conditions; under dissipativity of the Toeplitz symbol and a weak Kreiss–Lopatinskii condition allowing finitely many simple zeros on the unit circle, $dv(z)=\pi^{-1}e^{-|z|^2}dA(z),$67 is power bounded (Coulombel et al., 2021).

Taken together, these results show that finite-rank perturbations of Toeplitz operators are best understood as a family of classification problems rather than a single theorem. In the Fock space they are exhausted by finitely supported distributional jets; in Hardy spaces they are encoded by finite defect and near invariance; in Bergman settings they are governed by explicit symbol calculus and Mellin identities; and in block or matrix settings they are tied to Blaschke–Potapov models, boundary transfer matrices, and outlier spectral equations (Rozenblum et al., 2013, Liang et al., 2019, Le et al., 2020, Abhinand et al., 4 May 2026, Koekenbier et al., 15 Jun 2025).