i-PI 2.0: A Universal Force Engine for Advanced Molecular Simulations

TL;DR

i-PI 2.0 introduces a universal, modular force engine that decouples force evaluations from dynamics and can interface with any driver code for both classical and quantum simulations. Its refactored core (System, Motion, Ensemble, ForceField) enables easy integration of advanced sampling techniques and replica-exchange schemes, drastically reducing implementation overhead. The paper highlights features including replica-exchange MD, multiple time stepping, SL-RPC, ring-polymer instantons, a PLUMED metadynamics interface, and perturbed path integrals, with demonstrations on systems like a benzene-graphene interface and Zundel cation, and a post-processing framework for energy corrections. This open-source framework broadens access to cutting-edge atomistic methods for ab initio and empirical potentials, accelerating development and deployment of sophisticated simulations.

Abstract

Progress in the atomic-scale modelling of matter over the past decade has been tremendous. This progress has been brought about by improvements in methods for evaluating interatomic forces that work by either solving the electronic structure problem explicitly, or by computing accurate approximations of the solution and by the development of techniques that use the Born-Oppenheimer (BO) forces to move the atoms on the BO potential energy surface. As a consequence of these developments it is now possible to identify stable or metastable states, to sample configurations consistent with the appropriate thermodynamic ensemble, and to estimate the kinetics of reactions and phase transitions. All too often, however, progress is slowed down by the bottleneck associated with implementing new optimization algorithms and/or sampling techniques into the many existing electronic-structure and empirical-potential codes. To address this problem, we are thus releasing a new version of the i-PI software. This piece of software is an easily extensible framework for implementing advanced atomistic simulation techniques using interatomic potentials and forces calculated by an external driver code. While the original version of the code was developed with a focus on path integral molecular dynamics techniques, this second release of i-PI not only includes several new advanced path integral methods, but also offers other classes of algorithms. In other words, i-PI is moving towards becoming a universal force engine that is both modular and tightly coupled to the driver codes that evaluate the potential energy surface and its derivatives.

Figures (8)

• Figure 1: A schematic representation of the structure of i-PI. Green boxes identify programs, red and blue boxes identify classes, and lines represent information exchange between classes, all in a loose sense. This structure is reflected in the XML input format of i-PI. The System (or a collection of systems) is evolved in steps, which are compounded by collective moves that can exchange information between systems ( SMotion) and by an Output management. The i-PI core is complemented by external drivers ("force codes") that compute energy and forces -- the interface with which is managed by a ForceField class -- and by post-processing tools that can compute observables from the output of i-PI.
• Figure 2: The panels show the temperature dependence for the density (left) and the coefficient of thermal expansion (right) calculated for 64 molecules of q-TIP4P/f water using path integral (red) and classical MD (blue). While panels (a) and (b) were calculated at 1 bar, (c) and (d) were calculated at 100 bar from $NpT$ replica exchange simulations. Panels (e) and (f) were obtained at 1 bar without any replica exchange moves. The dotted lines on the upper left and upper middle panel are fourth degree polynomials fitted to the averages obtained from the simulations. All averages are calculated from 500 ps simulations.
• Figure 3: A commented snipped of an i-PI input file containing the essential components of a MTS simulation. The right portion of the figure schematically represents the steps in an integrator that correspond to such input.
• Figure 4: (a) Comparison of the vibrational frequencies of the full system (blue bars) and of the benzene molecule at the geometry it adopts on top of graphene (red crosses). (b) Convergence of the quantum kinetic energy on the adsorbate for the SL-RPC scheme with respect to the contracted number of beads $P'$ (see Eq. \ref{['eq:v-slc']}).
• Figure 5: CH$_4$+H instanton geometries at 250 K (left) and 300 K (right) for three situations. Top: $\tau=0$; Middle: All imaginary time slices (instanton geometry); Bottom: $\tau=\beta \hbar/2$.