Retarded Stieltjes-Type Kernels

• Retarded Stieltjes-type kernels are causal kernels defined on a half-axis that combine resolvent-type denominators, positivity, and analytic control.
• They appear in various forms including one-sided scalar models, operator-valued frameworks, and matrix-valued moment problems, each offering unique spectral insights.
• These kernels underpin modern applications such as nonlinear Maxwell equations, where they enforce causality via time-domain integration of retarded material laws.

Searching arXiv for the supplied papers and closely related work. Retarded Stieltjes-type kernels are one-sided or causal kernels built from Stieltjes, Cauchy–Stieltjes, or resolvent-type factors and supported on a half-axis or a time-ordered region. The terminology is not uniform across the literature. In some works it denotes semiaxis kernels such as $w(x)v(y)/(x+y)$ on $\mathbb{R}_+$; in others it arises through operator-valued Stieltjes representations on $[0,\infty)$, through matrix-valued moment problems on $[\alpha,\infty)$, or through adaptations of Cauchy–Stieltjes kernel families to causal kernels $1_{\{s\le t\}}G_\nu(g(t,s))$. A common structure is the combination of half-line support, positivity or sectoriality, and analytic control by transforms of the form $(x+y)^{-1}$, $(t-\lambda)^{-1}$, or $(z-x)^{-1}$ (Ushakova, 2011, Arlinskiĭ et al., 2021, Bryc et al., 2012).

1. Terminology and defining structures

The most direct scalar model is the Stieltjes-type operator on the semiaxis

$$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$

with kernel

$k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$

This kernel is positive, non-increasing in $\mathbb{R}_+$0, and depends on $\mathbb{R}_+$1 and $\mathbb{R}_+$2 only through the sum $\mathbb{R}_+$3. The operator is “retarded” in the sense that it acts on functions on $\mathbb{R}_+$4 and has no advanced part; all relevant kernels are of semi-axis type (Ushakova, 2011).

A second basic form is the Cauchy–Stieltjes kernel family generated by a probability measure $\mathbb{R}_+$5 with support bounded from above. It is built from

$\mathbb{R}_+$6

so the kernel is $\mathbb{R}_+$7, analytically equivalent, up to a linear change of variables, to the standard Cauchy kernel $\mathbb{R}_+$8 (Bryc et al., 2012).

A third formulation is operator-valued. For a Stieltjes family $\mathbb{R}_+$9 on $[0,\infty)$0, the canonical integral representation is

$[0,\infty)$1

with $[0,\infty)$2 nondecreasing. This realizes the kernel $[0,\infty)$3 as the basic Stieltjes factor in an operator-valued setting (Arlinskiĭ et al., 2021).

Matrix-valued moment theory places the same structure on a shifted half-line $[0,\infty)$4. There the Stieltjes class consists of holomorphic matrix functions $[0,\infty)$5 on $[0,\infty)$6 admitting

$[0,\infty)$7

with $[0,\infty)$8 and $[0,\infty)$9. The support restriction to $[\alpha,\infty)$0 is the matrix-valued version of the retarded threshold (Fritzsche et al., 2016).

2. One-sided Stieltjes operators on the semiaxis

In the semi-axis theory, the kernel

$[\alpha,\infty)$1

is the standard retarded Stieltjes-type kernel. The associated operator is studied on $[\alpha,\infty)$2, with $[\alpha,\infty)$3 and $[\alpha,\infty)$4. Its compactness theory is reduced to Hardy-type operators

$[\alpha,\infty)$5

and the paper states that $[\alpha,\infty)$6 is pointwise equivalent, up to constants, to the sum of these two one-sided Hardy operators (Ushakova, 2011).

For $[\alpha,\infty)$7, boundedness is characterized by finiteness of

$[\alpha,\infty)$8

where

$[\alpha,\infty)$9

Compactness holds iff $1_{\{s\le t\}}G_\nu(g(t,s))$0 and

$1_{\{s\le t\}}G_\nu(g(t,s))$1

For $1_{\{s\le t\}}G_\nu(g(t,s))$2 with $1_{\{s\le t\}}G_\nu(g(t,s))$3 and $1_{\{s\le t\}}G_\nu(g(t,s))$4, boundedness already implies compactness through finiteness of $1_{\{s\le t\}}G_\nu(g(t,s))$5. For $1_{\{s\le t\}}G_\nu(g(t,s))$6, compactness requires boundedness plus localized vanishing conditions encoded by $1_{\{s\le t\}}G_\nu(g(t,s))$7 and their truncated versions (Ushakova, 2011).

This semi-axis theory clarifies an important point. Retardedness here is not a Volterra restriction $1_{\{s\le t\}}G_\nu(g(t,s))$8; it is one-sidedness on $1_{\{s\le t\}}G_\nu(g(t,s))$9. The absence of negative arguments is the causal feature. The unweighted kernel $(x+y)^{-1}$0 on the whole semiaxis is therefore one-sided but not compact for $(x+y)^{-1}$1, because the controlling Hardy functionals cannot vanish at infinity (Ushakova, 2011).

3. Cauchy–Stieltjes kernel families and causal adaptation

For a non-degenerate probability measure $(x+y)^{-1}$2 whose support is bounded from above, the Cauchy–Stieltjes kernel family is parametrized by

$(x+y)^{-1}$3

With mean parametrization $(x+y)^{-1}$4, the family becomes

$(x+y)^{-1}$5

where $(x+y)^{-1}$6 is the pseudo-variance function. The associated Cauchy transform

$(x+y)^{-1}$7

is linked to $(x+y)^{-1}$8 by

$(x+y)^{-1}$9

The map $(t-\lambda)^{-1}$0 places the Stieltjes parameter to the right of the support of $(t-\lambda)^{-1}$1, where $(t-\lambda)^{-1}$2 is well defined and non-negative (Bryc et al., 2012).

The paper does not treat time variables explicitly, but it states that its structure is readily adapted to retarded kernels. A model construction is

$(t-\lambda)^{-1}$3

with $(t-\lambda)^{-1}$4. If one chooses

$(t-\lambda)^{-1}$5

then

$(t-\lambda)^{-1}$6

This enforces causality through $(t-\lambda)^{-1}$7 and keeps the spectral parameter in the analyticity region of $(t-\lambda)^{-1}$8 (Bryc et al., 2012).

A distinctive feature of Cauchy–Stieltjes families is iteration. If $(t-\lambda)^{-1}$9 is used as a new generating measure, the new pseudo-variance $(z-x)^{-1}$0 satisfies

$(z-x)^{-1}$1

so the Stieltjes parameter $(z-x)^{-1}$2 is preserved. This suggests a structured multi-step retarded kernel in which sequential interactions conserve a common resolvent parameter (Bryc et al., 2012).

The same theory also extends the domain of means beyond the natural interval. First, formula

$(z-x)^{-1}$3

continues to define a probability measure on an enlarged interval $(z-x)^{-1}$4. Second, for $(z-x)^{-1}$5, one adds an atom at

$(z-x)^{-1}$6

obtaining

$(z-x)^{-1}$7

A plausible implication is that retarded Stieltjes-type kernels can acquire discrete spectral contributions when the continuous part alone no longer defines a probability measure (Bryc et al., 2012).

4. Operator-valued and matrix-valued Stieltjes frameworks

In the operator-theoretic setting, a Stieltjes family $(z-x)^{-1}$8 is a Nevanlinna family on $(z-x)^{-1}$9 whose values on the negative real axis are selfadjoint and nonnegative. For each $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$0, the numerical range of a Stieltjes family lies in a sector, and after multiplication by a unimodular scalar $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$1, the family becomes maximal sectorial with acute semi-angle. This “up to a rotation” property is the basis for the associated closed sectorial forms (Arlinskiĭ et al., 2021).

The central form representation is

$$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$2

with $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$3 a nonnegative selfadjoint relation, $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$4 a contraction, and $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$5. The associated forms are holomorphic families of type (B) in the sense of Kato, with domain independent of $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$6 (Arlinskiĭ et al., 2021).

This representation admits a direct time-domain interpretation. Setting $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$7 with $$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$8,

$$(Sf)(x)=w(x)\int_0^\infty \frac{f(y)v(y)}{x+y}\,dy,\qquad x\in\mathbb{R}_+,$$9

so one may define

$k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$0

Then $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$1 becomes the Laplace transform of a kernel supported on $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$2. In this sense, operator-valued Stieltjes families generate causal kernels through Laplace inversion (Arlinskiĭ et al., 2021).

The matrix-valued moment problem on $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$3 supplies the corresponding algebraic classification. A sequence $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$4 is $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$5-Stieltjes non-negative definite when the Hankel matrices $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$6 and shifted Hankel matrices built from

$k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$7

are nonnegative. The exact truncated problem on $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$8 is solvable iff $k_S(x,y)=\frac{1}{x+y},\qquad x>0,\ y>0.$9 is $\mathbb{R}_+$00-Stieltjes nonnegative extendable, and the inequality problem is solvable iff $\mathbb{R}_+$01 is $\mathbb{R}_+$02-Stieltjes nonnegative definite (Fritzsche et al., 2016).

The function-theoretic counterpart is the matrix Stieltjes class on $\mathbb{R}_+$03, represented by

$\mathbb{R}_+$04

with $\mathbb{R}_+$05 and $\mathbb{R}_+$06. The $\mathbb{R}_+$07-Schur–Stieltjes algorithm acts both on moment sequences and on holomorphic matrix-valued functions, and the function-theoretic transforms

$\mathbb{R}_+$08

together with their inverse transforms, preserve the Stieltjes class under the stated range and kernel conditions (Fritzsche et al., 2016). This gives a recursive calculus for retarded kernels whose spectral support starts at $\mathbb{R}_+$09.

5. Modified transforms, continued fractions, and explicit resolvents

A different route to retarded Stieltjes-type structure begins with the modified Stieltjes transform

$\mathbb{R}_+$10

Its kernel is

$\mathbb{R}_+$11

The transform is defined on the whole real line, so it is not one-sided in itself. However, the paper states that when $\mathbb{R}_+$12 is supported on $\mathbb{R}_+$13, the gamma-mixture interpretation becomes one-sided. A natural retarded version is therefore obtained by restricting the underlying measure to $\mathbb{R}_+$14, or, as a plausible implication, by inverse Fourier transformation to a gamma-type kernel supported on $\mathbb{R}_+$15 (Klebanov et al., 2016).

Generalized indefinite strings of Stieltjes type furnish another spectral realization. Their Weyl–Titchmarsh function is a Herglotz–Nevanlinna function

$\mathbb{R}_+$16

and for discrete strings the continued fraction expansion of $\mathbb{R}_+$17 is given explicitly in terms of Hankel determinants. The paper does not introduce retarded kernels explicitly, but it describes the Green function structure from which a retarded kernel

$\mathbb{R}_+$18

can be inferred. A plausible implication is that the continued fraction coefficients determine causal Stieltjes-type Green kernels block by block (Eckhardt, 2020).

The explicit Stieltjes integral equation

$\mathbb{R}_+$19

has recently been solved in weighted space

$\mathbb{R}_+$20

by a new resolvent kernel. The classical kernels $\mathbb{R}_+$21 are of multiplicative convolution type, $\mathbb{R}_+$22. The new kernel for $\mathbb{R}_+$23 is

$\mathbb{R}_+$24

with $\mathbb{R}_+$25. It is explicit but non-convolution. The paper emphasizes that combining known convolution kernels can lead to a non-convolution kernel that is more effective (Schuur, 12 Feb 2025). Strictly speaking, this kernel is not Volterra-retarded, but the paper interprets the Stieltjes equation as multiplicative convolution on $\mathbb{R}_+$26, so the resolvent retains a one-sided Mellin-domain structure.

6. Retarded material laws and current applications

A contemporary application appears in nonlinear Maxwell equations with retarded material laws. The polarization is written as

$\mathbb{R}_+$27

The integrals over $\mathbb{R}_+$28 encode retardation directly. In the rigorous framework, the linear susceptibility is modeled by a finite signed measure $\mathbb{R}_+$29, while the nonlinear kernel is a finite measure $\mathbb{R}_+$30, reduced periodically to measures on the torus $\mathbb{R}_+$31 (Ohrem et al., 14 Mar 2025).

For time-periodic breathers, the decisive objects are the Fourier coefficients

$\mathbb{R}_+$32

where $\mathbb{R}_+$33. The assumptions require evenness in time, reality of the coefficients, the upper bound

$\mathbb{R}_+$34

and two-sided decay for the nonlinear kernel,

$\mathbb{R}_+$35

with $\mathbb{R}_+$36 in the slab case and $\mathbb{R}_+$37 in the cylindrical case. In the examples, $\mathbb{R}_+$38 and decays like $\mathbb{R}_+$39 (Ohrem et al., 14 Mar 2025).

The resulting effective operator involves

$\mathbb{R}_+$40

and its coercive quadratic form defines a variational functional of mountain-pass type. The paper does not explicitly call these kernels Stieltjes-type, but the measure-based hereditary structure, spectral positivity, and inversion of convolution operators on a one-sided time domain place them in the same family of retarded kernels governed by Stieltjes-type resolvent behavior (Ohrem et al., 14 Mar 2025).

A recurrent misconception is therefore that retarded Stieltjes-type kernels form a single classical object. The literature instead presents a family of related constructions: one-sided scalar kernels on $\mathbb{R}_+$41, Cauchy–Stieltjes probability families, operator-valued Stieltjes families on $\mathbb{R}_+$42, matrix moment kernels on $\mathbb{R}_+$43, Mellin-resolvent kernels for the Stieltjes integral equation, and measure-based retardation kernels in nonlinear electromagnetism. What unifies them is not a single formula but a shared architecture: half-line support, resolvent-type denominators, and positivity or sectoriality controlled by measures, transforms, or moment data (Ushakova, 2011, Bryc et al., 2012, Arlinskiĭ et al., 2021, Fritzsche et al., 2016, Schuur, 12 Feb 2025, Ohrem et al., 14 Mar 2025).