The numerical solution of parabolic Volterra integro-differential equations on unbounded spatial domains (Q2387322)

The authors show how to use the artificial boundary method for the numerical solution of parabolic Volterra integro-partial differential equations of the type \[ {\partial u\over\partial t} + \int_0^t k(x,t-\tau)u(x,\tau)\, d\tau \, = \, \Delta u + f(x,t), \] where \(x\in {\mathbb R}\), \(t\in [0,T]\), \(u| _{t=0}=u_0(x)\) with \(u\to 0\) as \(| x| \to\infty\). The functions \(f\) and \(u_0\) are assumed to be continuous with support in \([0,d]\), and the kernel \(k\) satisfies \(k(x,t)=k_0(t)\) for \(x\notin (0,d)\), where \(k_0\) is continuous or weakly singular. They first derive fully transparent artificial boundary conditions by dividing the spatial-temporal domain into appropriate subdomains. These allow to reduce the original problem to one defined on bounded spatial-temporal domains. They then study two classes of examples for the convolution kernel which appears in the reduced problem, as well as the concrete example \(k_0(t)\equiv 1\) for the memory kernel. After explaining the connection to previous work on parabolic partial differential equations by \textit{H. Han} and \textit{Z. Huang} [ibid. 43, 889--900 (2002; Zbl 0999.65086); Comput. Math. Appl. 44, 655--666 (2002; Zbl 1030.35092)], they conclude with giving two numerical examples.