Partial Integral Equation (PIE) Framework

• Partial Integral Equation Framework is an operator-based approach that reformulates PDEs into unconstrained integro-differential equations using PI operators.
• It enables systematic stability analysis and controller synthesis via operator-valued LMIs and convex optimization techniques.
• The framework absorbs boundary and integral conditions within its operator structure, supporting mesh-independent numerical methods and scalable simulations.

A Partial Integral Equation (PIE) framework is an operator-theoretic methodology for the representation, analysis, and numerical solution of infinite-dimensional systems, notably partial differential equations (PDEs) with spatial integral terms and complex boundary conditions. The central idea is to transform a PDE (or system) into a PIE: an unconstrained first-order-in-time integro-differential equation characterized by operator-valued coefficients acting on $L_2$-space states. The PIE framework enables direct application of convex optimization, robust control, and simulation tools familiar from finite-dimensional systems, while eliminating ad hoc treatment of boundary and continuity constraints typically required in conventional PDE methods.

1. Mathematical Foundations and Operator Structure

The PIE framework is built around the algebra of partial-integral (PI) operators. For a state $x$ in $L_2^n[a, b]$, a standard 3-PI operator $\mathcal{P}_{\{R_0, R_1, R_2\}}$ is defined as: $$(\mathcal{P}_{\{R_0, R_1, R_2\}} x)(s) = R_0(s) x(s) + \int_a^s R_1(s, \theta) x(\theta) d\theta + \int_s^b R_2(s, \theta) x(\theta) d\theta.$$ PI operators are closed under addition, composition, and adjoint, and form a $C^*$-algebra suitable for representing time-evolution, in-domain integral operators, and boundary terms. For systems with inputs, higher-order derivatives, or ODE coupling, the framework generalizes naturally to 4-PI operator structures, allowing block-matrix representations that couple $L_2$ and finite-dimensional states (Shivakumar et al., 2022).

A general PIE system has the form: $$\mathcal{T} \dot{x}_f(t) = \mathcal{A} x_f(t), \quad x_f(0) = x_f^0 \in L_2^n[a, b],$$ where $\mathcal{T}, \mathcal{A}$ are PI operators constructed explicitly from PDE data (Shivakumar et al., 2022, Shivakumar et al., 2022).

2. Conversion of PDEs to the PIE Framework

Given a linear PDE with any combination of spatial derivatives and spatial integrals (including nonlocal and boundary integral terms), the PIE framework employs a substitution of the state with its highest, well-defined spatial derivatives: $$z(t, s) := \mathrm{col}\bigl( \partial_s^0 x_0(t, s),\, \partial_s^1 x_1(t, s),\, \partial_s^2 x_2(t, s) \bigr).$$ Through repeated application of the Fundamental Theorem of Calculus and, for higher-order systems, a generalization of Cauchy's rule for repeated integration, one constructs a PI operator $x$0 such that $x$1, with $x$2 living in $x$3 and free of boundary/continuity constraints (Shivakumar et al., 2022, Shivakumar et al., 2022). All PDE system parameters, including boundary conditions (Dirichlet, Neumann, Robin, integral, or nonlocal), are incorporated into $x$4 and the right-hand-side operator $x$5 via explicit kernel mappings.

For the most general class of linear spatially distributed systems—including arbitrary-order derivatives, boundary-valued or distributed inputs, and ODE/PDE interconnections—the construction yields a PIE of the form $x$6 with all terms expressed as block-PI operators, ensuring a one-to-one correspondence between constrained PDE solutions and unconstrained PIE orbits (Shivakumar et al., 2022, Peet, 2018).

3. PIE Formulation for Stability Analysis and Control Synthesis

A key application of the PIE framework is the systematic analysis and design (stability, robust performance, controller synthesis) of infinite-dimensional systems via direct operator-valued convex programming. For a given PIE, exponential stability is equivalent to the existence of a positive-definite PI operator $x$7 such that the following operator-valued Linear Matrix Inequalities (LMIs) are feasible: $x$8

$x$9

for some $L_2^n[a, b]$0. These inequalities encode Lyapunov-type estimates directly in operator form; via polynomial parameterization of the kernels, they reduce to tractable finite-dimensional LMIs solvable by standard SDP solvers (Shivakumar et al., 2022, Shivakumar et al., 2022).

In robust and optimal control, extensions of the PIE framework enable the use of Integral Quadratic Constraints (IQCs) and $L_2^n[a, b]$1-analysis. Structured uncertainties, delays, and dynamic nonlinearities are incorporated in the PIE state-space, so that robust stability/performance synthesis reduces to verifying feasibility of Linear Partial Integral Inequalities (LPIs) in operator variables, with decision variables $L_2^n[a, b]$2, multipliers $L_2^n[a, b]$3, and controller $L_2^n[a, b]$4 entering affinely. Implementations in PIETOOLS provide a practical computational pipeline (Lenssen et al., 18 Nov 2025).

4. Treatment of Boundary and Integral Conditions

In traditional PDE theory, boundary and continuity constraints are a major source of analysis and discretization complexity. In the PIE representation, boundary conditions—including those involving spatial integrals or nonlocal constraints—are absorbed into the operator $L_2^n[a, b]$5. This operator embeds all boundary data directly in its kernel structure, so that the PIE state $L_2^n[a, b]$6 is unconstrained in $L_2^n[a, b]$7, thus removing the need for special treatment at the discretization or solution stage (Shivakumar et al., 2022, Peet, 2018).

This transformation is valid as long as certain well-posedness/admissibility criteria are satisfied, typically checked by invertibility of an associated boundary-transfer matrix assembled from the PDE's boundary condition parameters and the structure of the derivative-to-original variable mapping.

5. Numerical Methods and Software Implementation

The PIE framework underpins a suite of numerical methods that are mesh-independent and inherently compatible with convex optimization. Discretization schemes include Galerkin approximations using Chebyshev polynomials, where the sparse structure of PI operators leads to tractable, well-conditioned linear algebra problems (Peet et al., 2020). For time integration, explicit or implicit schemes can be applied in direct analogy to ODEs.

Algorithmic pipelines—most notably PIETOOLS—automate the end-to-end conversion: symbolic or GUI-definition of the PDE (spatial orders, boundary conditions, differential/integral terms), translation to PIE operator parameter structures, formulation and solving of operator-valued LPIs/LMIs, and simulation via block-PI state evolution. Engineering examples in heat, wave, beam, reaction-diffusion, and population-dynamics PDEs, as well as delay equations, demonstrate the method’s scalability and accuracy (Peet, 2018, Shivakumar et al., 2022, Shivakumar et al., 2022, Lenssen et al., 18 Nov 2025). Stability and performance margins computed via PIE-LMI methods are numerically tight, with large-dimensional PDE systems, high-order boundary conditions, and spatially-distributed integral operators handled seamlessly.

6. Extensions and High-Dimensional Applications

Recent work has extended the PIE framework to nonlinear, stochastic, and integral-partial differential equations (integral-PDEs/PIDEs). For example, parameter grouping, Taylor expansion approximations of nonlocal integral operators, and explicit symbolic solution construction have been developed for solving high-dimensional PIDEs with machine-level precision, as in the FEX-PG method (Hardwick et al., 2024).

In stochastic control, stochastic representations for coupled HJB-Isaacs systems with integral-partial operators have linked backward stochastic differential equations with nonlocal jumps to the viscosity solutions of corresponding PIE/PIDE systems, providing a rigorous foundation for analysis of stochastic control and game-theoretic models (Luo et al., 2023).

Boundary-integral PIE formulations also underpin high-order discretizations and fast solvers (e.g., Nyström, pre-corrected FFT) for mixed-boundary-value and surface-wave problems, such as linearized water waves in partially covered domains, confirming Fredholm second-kind structure and mesh-independent conditioning (Askham et al., 26 Oct 2025).

7. Significance and Outlook

The Partial Integral Equation framework provides a unified and scalable platform for expressing, analyzing, simulating, and controlling infinite-dimensional systems. By absorbing all continuity and boundary constraints into algebraic operator structure, the PIE methodology generalizes ODE-style analysis and synthesis methods—Lyapunov, LMI/SDP, IQC-based robust control, Galerkin discretization—to broad PDE, delay, and integral-differential settings without spatial discretization artifacts.

The explicit operator algebra, symbolic-to-numeric workflow, and convex programmability of the PIE framework position it as a central tool for high-fidelity modeling, robust and optimal control, and computational design of physical and engineering systems with spatially distributed or nonlocal structure (Shivakumar et al., 2022, Shivakumar et al., 2022, Lenssen et al., 18 Nov 2025). Emerging research targets extensions to nonlinear/nonlocal and data-driven PDEs, adaptive and learning-based operator synthesis, and high-performance parallel acceleration, consolidating the PIE framework as foundational in modern infinite-dimensional systems theory.