In 2008, a few researchers (my advisors among them) published their work detailing a normal mode integrator based on Langevin dynamics. With Normal mode analysis it is possible to study the vibrational characteristics of a system. Normal modes can be found in many disciplines ranging from civil engineering to quantum mechanics. In biophysics it is possible to break the movements of a molecule into two sets of frequencies - slow, and fast - and the frequency used to define the split is arbitrary. The authors applied this concept to biomolecules and developed an integrator (function that updates positions of atoms based on physics). Next, the authors measured the performance of the integrator by varying the number of modes propagated. This idea is of particular interest to Operating Systems since we will be following a similar pattern with our final project.
One improvement I would make to the paper is to explore the idea of describing the movements of a biomolecule by the slowest frequencies. How does this affect other simulations? Will the reduction in computation time result in less accurate results? Is it possible to determine an accurate estimate of the number of modes to propagate based on the characteristics of the system in question?