[Paper Review] On transparent boundary conditions for the high--order heat equation

Durvudkhan Suragan, Niyaz Tokmagambetov|arXiv (Cornell University)|Feb 6, 2013

Differential Equations and Numerical Methods6 references3 citations

TL;DR

This paper develops transparent boundary conditions for the high-order heat equation in bounded domains using a non-local initial boundary value problem. It proves that the solution to this artificial problem exactly matches the solution of the original Cauchy problem in the domain, leveraging a fundamental solution and integral representations to ensure uniqueness, stability, and analytical tractability.

ABSTRACT

In this paper we develop an artificial initial boundary value problem for the high-order heat equation in a bounded domain $Ω$. It is found an unique classical solution of this problem in an explicit form and shown that the solution of the artificial initial boundary value problem is equal to the solution of the infinite problem (Cauchy problem) in $Ω$.

Motivation & Objective

• To develop artificial boundary conditions that accurately mimic the behavior of solutions in unbounded domains for high-order parabolic equations.
• To ensure the resulting initial boundary value problem is uniquely solvable and stable.
• To construct boundary conditions that yield a solution identical to the infinite-domain (Cauchy) problem within the bounded computational domain.
• To provide an analytically tractable framework using integral representations and fundamental solutions for high-order heat equations.

Proposed method

• Formulates an artificial initial boundary value problem (IBVP) for the high-order heat equation using a non-local boundary condition involving time-space integrals.
• Employs the fundamental solution $ \varepsilon_{m,n}(x,t) $ of the high-order heat equation $ \left(\frac{\partial}{\partial t} - \Delta_x\right)^m \varepsilon_{m,n} = 0 $ as a Green's function.
• Derives a non-local boundary condition via integration by parts and adjoint operators, ensuring the solution matches the Cauchy problem in the domain.
• Uses the Dirichlet-to-Neumann (DtN) map concept in a non-local form to enforce transparent boundary behavior.
• Applies the method of fundamental solutions and convolution-type representations to construct explicit solutions.
• Proves uniqueness and regularity of the solution in Hölder spaces $ C^{2m+\gamma, m+\frac{\gamma}{2}} $ under appropriate smoothness assumptions on the data.

Experimental results

Research questions

• RQ1 Can transparent boundary conditions be constructed for the high-order heat equation that preserve the solution of the infinite-domain problem within a bounded domain?
• RQ2 How can non-local boundary conditions be designed to ensure both analytical solvability and numerical efficiency?
• RQ3 What role does the fundamental solution $ \varepsilon_{m,n} $ play in constructing exact solutions to the high-order heat equation with artificial boundaries?
• RQ4 Under what conditions does the solution of the artificial IBVP coincide with the solution of the original Cauchy problem?
• RQ5 Can the Dirichlet-to-Neumann (DtN) approach be generalized to high-order parabolic equations using integral boundary operators?

Key findings

• The heat potential $ u(x,t) = \int_0^t \int_Q \varepsilon_{m,n}(x-\xi,t-\tau)f(\xi,\tau)\,d\xi d\tau $ is the unique classical solution of the non-local IBVP defined by equations (2)–(4).
• The solution of the artificial IBVP with non-local boundary conditions (4) is identical to the solution of the original Cauchy problem in the domain $ \Omega $, satisfying the transparency condition.
• The fundamental solution $ \varepsilon_{m,n} $ acts as the Green's function for the non-local IBVP, ensuring exact representation of the solution.
• For the case $ m=1 $, the solution to the inhomogeneous problem (13)–(15) is given explicitly by a sum of volume and boundary integrals involving the heat kernel and the Green's function.
• The non-local boundary condition (4) ensures that the solution satisfies the required initial conditions and matches the infinite-domain solution in $ \Omega $, with the boundary terms involving time integrals of normal derivatives.
• The method guarantees stability and uniqueness under Hölder continuity assumptions on the data, with solutions in $ C^{2m+\gamma, m+\frac{\gamma}{2}} $ regularity class.