Parameterized Integral Equation Formulation

• Parameterized integral equation formulation is an operator-based approach that retains explicit physical, geometric, and numerical parameters to convert PDE problems into dense, well-conditioned equations.
• It employs second-kind formulations, combined-field methods, and advanced discretization techniques to ensure stability, efficient solver performance, and consistent convergence across various applications.
• The method leverages practical strategies like spherical harmonics, Galerkin discretization, and fast solvers such as the Fast Multipole Method to achieve scalable and accurate numerical solutions.

Searching arXiv for recent and foundational papers on parameterized integral equation formulations. Parameterized integral equation formulation, in the literature surveyed here, denotes an integral or boundary-integral representation in which the operator, kernel, densities, or auxiliary projectors depend explicitly on physical, geometric, or numerical parameters such as a contrast function $q(x)$, sphere centres and radii, dielectric constants, wavenumbers, interface geometry, mass jumps, energy, or a characteristic length scale $\ell$ (Borges et al., 2019, Bramas et al., 2019, Helsing et al., 2019, Bal et al., 2023, Hadjesfandiari et al., 2017). Across acoustic, electrostatic, electromagnetic, transport, diffusion, and surface-flow problems, these formulations are used to convert PDE transmission, scattering, or mode problems into dense operator equations that are frequently of Fredholm second kind, or into boundary-domain systems based on parametrices, with the parameter dependence retained at the operator level rather than introduced only after discretization (Hassan et al., 2019, Imbert-Gerard et al., 2018, Fresneda-Portillo, 2020).

1. Conceptual scope and meaning of parameterization

A recurring pattern is that the governing PDE is rewritten as an operator equation whose coefficients enter through a parameterized integral operator. In inverse acoustic scattering, the contrast function $q(x)$ yields the Lippmann–Schwinger form

$u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$

and, after discretization or basis expansion, a parameterized operator equation of the form

$(I + K(q))u = u^i$

is obtained (Borges et al., 2019). Here the parameter is not a scalar tuning constant but the material inhomogeneity itself.

In the $N$-body dielectric spheres problem, the parameterization is geometric and material: the configuration is specified by the tuples $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$ together with the background dielectric constant $\kappa_0$, and the resulting boundary integral equation describes mutual polarisation among all spheres (Hassan et al., 2019). In electromagnetic transmission, the system depends explicitly and separately on the wavenumbers $k_1$ and $k_2$, and also on “uniqueness parameters” such as $\ell$0, $\ell$1, $\ell$2, and $\ell$3, which weight operator combinations to guarantee unique solvability in difficult regimes (Helsing et al., 2019).

Singular-waveguide formulations make the role of parameters especially explicit. For Klein–Gordon singular waveguides, mass dependence appears in the kernel through $\ell$4, in the jump conditions, and in the coefficients of the operators, while the geometry enters through the interface parameterization $\ell$5 (Bal et al., 2022). For the massive Dirac equation with a mass jump across an interface, the final boundary integral equation

$\ell$6

contains the mass $\ell$7, the energy $\ell$8, the interface $\ell$9, and the normal $q(x)$0 explicitly (Bal et al., 2023).

This suggests a useful distinction. In these works, “parameterized” usually does not mean reduced-order surrogates or parametric model reduction. It means that the integral equation itself is constructed so that dependence on coefficients, contrasts, geometry, or spectral parameters remains explicit in the operator.

2. Operator structures and canonical formulations

A major class of parameterized integral equations is the second-kind formulation. For dielectric spheres, the induced surface charge $q(x)$1 satisfies

$q(x)$2

and, using Calderón identities, this can be rewritten in second-kind form with the adjoint double-layer operator $q(x)$3 (Hassan et al., 2019). The second-kind character is central because it underlies the $q(x)$4-independent stability and convergence analysis developed in the same line of work (Bramas et al., 2019).

Another class is the combined-field or regularized combined-field family in computational electromagnetics. The regularized CFIE replaces the standard coupling operator by a regularizing operator $q(x)$5, and one proposed choice is

$q(x)$6

where $q(x)$7 is the scalar potential part of the EFIE operator at $q(x)$8 (Niino et al., 2024). The parameterization here is tied to the choice of regularizer and to the wavenumber.

Extended transmission formulations introduce larger operator systems to maintain solvability across broad parameter regimes. In electromagnetic transmission, the densities satisfy a Fredholm second-kind system

$q(x)$9

with operator blocks parameterized by $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$0, $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$1, and material ratios such as $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$2 (Helsing et al., 2019). A related, but structurally different, approach embeds Maxwell’s equations into a Dirac system and yields an $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$3 or $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$4 boundary integral equation,

$u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$5

with explicit dependence on the permittivity ratio $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$6, permeability ratio $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$7, and wavenumber ratio $u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$8 (Helsing et al., 2019).

Volume-integral and boundary-domain formulations extend the same principle to variable-coefficient media. For anisotropic electrostatics and acoustics, well-conditioned equations of the form

$u^s(x) = -k^2 \int_\Omega G(x,y) q(y) u(y)\,dy,$9

are derived from a new vector PDE, where $(I + K(q))u = u^i$0 or $(I + K(q))u = u^i$1 carries the medium anisotropy and inhomogeneity (Imbert-Gerard et al., 2018). For diffusion in inhomogeneous media, the formulation is built from a new family of parametrices,

$(I + K(q))u = u^i$2

leading to boundary-domain integral equations in weighted Sobolev spaces on unbounded domains (Fresneda-Portillo, 2020).

3. Discretization, basis design, and quadrature

The parameterization of the continuous operator is matched by discretizations that respect geometry and kernel structure. In the dielectric spheres literature, functions on each sphere are expanded in spherical harmonics up to degree $(I + K(q))u = u^i$3, producing a global approximation space $(I + K(q))u = u^i$4 and a reduced subspace that excludes constant components on each sphere (Bramas et al., 2019). The corresponding numerical analysis establishes $(I + K(q))u = u^i$5-independent error estimates and, for analytic data, exponential convergence in $(I + K(q))u = u^i$6 (Hassan et al., 2019).

Galerkin discretization is also central in surface and electromagnetic formulations. The regularized CFIE of (Niino et al., 2024) is designed so that the discrete system

$(I + K(q))u = u^i$7

can be assembled using only Rao–Wilton–Glisson basis functions, avoiding dual basis functions on a barycentrically refined mesh. In optical-waveguide mode calculation, the SKIE formulation places unknowns only on material interfaces and uses chunked boundary discretization with high-order polynomials, generalized Gaussian quadrature for logarithmic singularities, and adaptive Gauss quadrature for nonsingular interactions (Lai et al., 2015).

For Galerkin boundary element methods more broadly, accurate kernel integration is itself a parameterized subproblem. Double surface integrals over triangle pairs are written in the form

$(I + K(q))u = u^i$8

and the treatment depends on whether triangles are zero-touch, one-touch, two-touch, or coincident (Adelman et al., 2015). Non-singular cases are handled by spherical harmonics and multipole expansions; touching cases are treated by recursive geometric decomposition together with scaling and symmetry arguments; and the Helmholtz kernel is decomposed into a Laplace singular part plus a regular remainder (Adelman et al., 2015).

High-order surface discretization appears again in the surface Stokes setting, where the surface is partitioned into smooth patches $(I + K(q))u = u^i$9, Koornwinder polynomials are used as basis functions, and Vioreanu–Rokhlin nodes support spectral collocation (Goodwill et al., 23 Feb 2026). In singular-waveguide computations, the interface is truncated and adaptively discretized with high-order quadrature, and weak singularities are handled by special quadratures (Bal et al., 2022).

4. Fast solvers, preconditioning, and complexity control

Because integral equation discretizations are dense, solver design is inseparable from the formulation. In inverse and forward acoustic scattering with point scatterers, domain decomposition preconditioners extend restricted additive Schwarz ideas from sparse PDE systems to dense integral equations (Borges et al., 2019). The paper further combines the domain decomposition preconditioner with a low-rank correction, forming a new preconditioner for the forward problem, and then uses the forward preconditioner as a building block for a preconditioner for the Gauss–Newton Hessian in the inverse problem (Borges et al., 2019).

For the $N$0-body dielectric spheres problem, the computational strategy uses the Fast Multipole Method to accelerate the action of $N$1, giving $N$2 cost per matrix-vector product for fixed $N$3, and proves that the number of linear solver iterations required to obtain a solution is independent of $N$4 under geometric constraints (Bramas et al., 2019). The combination of $N$5-independent iteration counts and linear-scaling matrix-vector products is what produces the overall linear scaling solution strategies described in that work.

The Klein–Gordon singular-waveguide algorithm separates two nonlocal components. The surface operator $N$6 is evaluated by a sweeping algorithm based on left and right recurrences, yielding $N$7 application cost for $N$8 discretization points, while the layer-potential part is accelerated by the 2D Fast Multipole Method, and the resulting discrete system is solved by GMRES (Bal et al., 2022). A related theme appears in the time-dependent radiative transport equation, where the integral formulation enables a treecode reduction of the computational complexity from $N$9 to $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$0, with $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$1 the number of points in the physical domain (Zhao et al., 2020).

Fast direct solvers are also developed when iterative acceleration alone is not the main target. For the surface Stokes equation, proxy shell compression exploits low-rank structure in interactions between well-separated surface patches and supports recursive hierarchical compression, enabling fast direct solvers with $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$2-like complexity (Goodwill et al., 23 Feb 2026).

5. Solvability, stability, and parameter regimes

The most prominent analytical theme is that parameter dependence need not destroy well-posedness. In the dielectric spheres problem, a new a priori analysis demonstrates $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$3-independent stability of the continuous and discrete formulations and yields convergence rates that are independent of $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$4, provided radii, inter-sphere separation, and dielectric constants are uniformly bounded above and below (Hassan et al., 2019). The companion computational analysis shows that the number of linear solver iterations depends only on geometric and physical parameters, not on $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$5 (Bramas et al., 2019).

In electromagnetic transmission, the extended charge-current formulation is constructed so that the integral equation is uniquely solvable whenever the PDE problem is uniquely solvable, including the plasmonic condition with real positive exterior wavenumber and imaginary interior wavenumber, subject to $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$6 (Helsing et al., 2019). The inclusion of additional charge densities is stated to be necessary for uniqueness throughout the whole parameter regime (Helsing et al., 2019). The Dirac integral equation for dielectric and plasmonic scattering pushes this further: it is described as free from false eigenwavenumbers for a wider range of permittivities than other known formulations, applicable to magnetic materials, valid in both two and three dimensions, and not suffering from low-frequency breakdown (Helsing et al., 2019).

Waveguide formulations show a different solvability mechanism. For Klein–Gordon singular waveguides, invertibility is proved for all but finitely many parameter choices, using Fourier analysis in the flat-interface case and Fredholm arguments for curved interfaces (Bal et al., 2022). For the Dirac singular-waveguide problem, holomorphic perturbation theory is used to prove that the boundary integral equation has a unique solution for almost all choices of parameters, and the same strategy extends to the two-interface case (Bal et al., 2023).

Boundary-domain formulations in inhomogeneous media require a different functional framework. For diffusion on an unbounded domain, the mapping properties of parametrix-based potentials are analyzed on weighted Sobolev spaces, equivalence between the mixed boundary value problem and the BDIE system is proved, and uniqueness is established by the Fredholm Alternative and compactness arguments adapted to weighted Sobolev spaces (Fresneda-Portillo, 2020). This directly addresses a common misconception that parameterized kernels in unbounded or variable-coefficient settings preclude rigorous operator theory; the cited work shows the opposite under explicit coefficient and decay hypotheses.

6. Domain-specific realizations and cross-disciplinary significance

The range of applications is unusually broad. Inverse scattering formulations parameterized by the contrast $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$7 support preconditioned Gauss–Newton methods (Borges et al., 2019). Boundary integral equations parameterized by sphere radii, positions, and dielectric constants model mutual polarisation in large $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$8-body electrostatics (Hassan et al., 2019, Bramas et al., 2019). Electromagnetic transmission formulations parameterized by wavenumbers and material ratios are designed to remain uniquely solvable in plasmonic and magnetic regimes (Helsing et al., 2019, Helsing et al., 2019). Optical-waveguide mode solvers parameterize the system by the effective index $\{(\mathbf{x}_i,r_i,\kappa_i)\}_{i=1}^N$9 and place unknowns only on material interfaces, enabling the computation of bound, leaky, and complex modes for photonic crystal fibers, dielectric fibers, and waveguides with arbitrary irregular cross section (Lai et al., 2015).

Variable-coefficient and anisotropic media lead to parameterized volume or boundary-domain formulations rather than purely boundary-only systems. In electrostatics, acoustics, and electromagnetics in smoothly varying anisotropic media, anisotropy and inhomogeneity are carried by tensor fields $\kappa_0$0 and $\kappa_0$1, and the resulting equations have the robust structure $\kappa_0$2 or closely related second-kind forms (Imbert-Gerard et al., 2018). In size-dependent thermoelasticity, the characteristic length scale $\kappa_0$3 enters only the mechanical kernels, not the thermal equation, and modifies the singularity and decay structure of the fundamental solutions (Hadjesfandiari et al., 2017). In diffusion, the coefficient $\kappa_0$4 enters the very definition of the parametrix (Fresneda-Portillo, 2020). In time-dependent radiative transport, the integral formulation eliminates the angular variable and produces a space-time equation for the angular average, parameterized by attenuation, scattering, and path geometry (Zhao et al., 2020).

A plausible implication is that parameterized integral equation formulation is less a single method than a structural viewpoint. The viewpoint is to encode the essential dependence on coefficients, geometry, spectral variables, and interface physics directly into the integral operator, then choose discretization, quadrature, and solvers that preserve second-kind, contraction-plus-compact, or Fredholm properties whenever possible. The works surveyed here show that this viewpoint is compatible with dense linear algebra, high-order discretization, fast multipole and treecode acceleration, domain decomposition preconditioning, and, in several cases, parameter-uniform stability or convergence estimates (Borges et al., 2019, Imbert-Gerard et al., 2018, Goodwill et al., 23 Feb 2026).