Exploring Exponential Runge-Kutta Methods: A Survey

In this survey, we provide an in-depth investigation of exponential Runge-Kutta methods for the numerical integration of initial-value problems. These methods offer a valuable synthesis between classical Runge-Kutta methods, introduced more than a century ago, and exponential integrators, which date back to the 1960s. This manuscript presents both a historical analysis of the development of these methods up to the present day and several examples aimed at making the topic accessible to a broad audience.

💡 Research Summary

This survey provides a comprehensive overview of exponential Runge‑Kutta (ExpRK) methods for solving semilinear initial‑value problems of the form u′(t)=Au(t)+g(t,u(t)), u(0)=u₀. The authors begin by recalling the historical development of classical Runge‑Kutta (RK) schemes, introduced by Runge in 1895 and refined by Kutta in 1901, and they point out the well‑known limitation of explicit RK methods when applied to stiff problems. Stiffness is discussed from several perspectives: eigenvalues of the Jacobian with large negative real parts, multi‑scale dynamics, and, most pragmatically, the inability of explicit schemes to use reasonable step sizes without sacrificing stability.

The core idea of ExpRK is then presented: the linear part A is treated exactly via the matrix exponential e^{hA}, while the nonlinear term g is incorporated through a Runge‑Kutta‑type stage construction that uses the so‑called ϕ‑functions. These functions are defined by
\