Lucas
Visscher
*a,
Peter
Bolhuis
*b and
F. Matthias
Bickelhaupt
*a
aAmsterdam Center for Multiscale Modeling, VU University Amsterdam, De Boelelaan 1083, 1081 HV Amsterdam, The Netherlands. E-mail: L.Visscher@vu.nl; F.M.Bickelhaupt@vu.nl
bAmsterdam Center for Multiscale Modeling, University of Amsterdam, PO Box 94157, 1090 GD Amsterdam, The Netherlands. E-mail: P.G.Bolhuis@uva.nl
With the theoretical models in the above-mentioned fields reaching maturity, complex simulation workflows are emerging that link quantum mechanical (QM), molecular mechanics (MM), coarse-grained (CG), and continuum descriptions. The focus is thereby changed from the improvement of individual components of a workflow, calculations at a single length and/or time scale, to the improvement of the complete model and the transfer of information between the levels. Feeding the larger length scale simulations with ab initio parameters from lower length scales requires a thorough matching of the physics in the two models and efficient implementation of the entire workflow.
As multiscale modelling developments are primarily discussed in the literature of the parent fields, cross-fertilization between the different fields is still limited. This themed issue, collecting ideas on multiscale modelling across the broad field of physical chemistry and chemical physics, therefore aims to enhance the interdisciplinary exchange of ideas.
The contributions address coupling between methods at a wide range of length and time scales. For the smallest length scales it is of interest to consider quantum mechanical effects on the motion of nuclei, ideally in an adaptive scheme that allows for switching between quantum and classical mechanical descriptions in a dynamical fashion (DOI: 10.1039/c0cp02865g). More established are techniques that connect an explicit quantum mechanical (QM) description of the electrons to an implicit molecular mechanics (MM) model (DOI: 10.1039/c0cp02957b). Although both these models employ the Coulomb operator to describe electrostatic interactions, it is challenging to accurately handle the interaction between the delocalized electron density in the QM system and the localized charges in the MM part. Parameterization of such charges and inclusion of polarization in the MM description therefore remains an active field of research (DOI: 10.1039/c0cp02850a and DOI: 10.1039/c1cp20646j). Genetic algorithms can improve optimization of force field parameterization (DOI: 10.1039/c0cp02889d). The combination of QM/MM with a boundary potential for bulk solvent enhances efficiency (DOI: 10.1039/c0cp02828b). An alternative to the QM/MM partitioning is to employ a frozen density ansatz and thereby include non-electrostatic components in solute–solvent interactions (DOI: 10.1039/c0cp02874f). This model may also be employed in subsystem density functional approaches in which localized excitations form the basis for an excitonic treatment of extended systems (DOI: 10.1039/c0cp02808h). This is an example of reducing the information obtained from the calculation of an individual subsystem, a key concept in multiscale modelling.
This concept is also used at larger length scales where coarse graining (CG) techniques systematically reduce the number of degrees of freedom of a complex system. Besides increasing the efficiency of the simulation and being able to study long-timescale processes, meaningful simple models are also important to understand complex systems as stressed by Kamerlin and Warshel in their perspective (DOI: 10.1039/c0cp02823a). An example of coarse graining is to match conditional reversible work (DOI: 10.1039/c0cp02888f). Several papers combine the MM with the CG level (DOI: 10.1039/c0cp02842h), or even the continuum level (DOI: 10.1039/c0cp02936j), to investigate behaviour of protein fibers and polymers (DOI: 10.1039/c0cp02868a). Moreover, in a spirit similar to the QMMM approach CG and MM potentials can be coupled in a hybrid scheme (DOI: 10.1039/c0cp02981e). Coupling a protein model with a multi-particle collision model allows exploring the effect of hydrodynamics (DOI: 10.1039/c1cp00003a). An important problem in multiscale modelling is fine-graining or back-mapping to enable information flow from a higher to a lower level description, for instance, when studying the properties of membrane remodelling proteins (DOI: 10.1039/c0cp02978e). While most CG models have been developed to reproduce statics, a relatively new topic in multiscale modelling is the effort to preserve correct dynamics in the CG models. Kremer et al. provide a perspective of the methods that link time scales along the different levels in multiscale models (DOI: 10.1039/c1cp20247b). Another contribution introduces a new method to coarse grain the dynamics based on a Focker-Planck description (DOI: 10.1039/c0cp02826f). Similarity measures can facilitate coarse graining of dynamics in proteins (DOI: 10.1039/c0cp02675a). Finally, considering the time dimension only, rare event techniques can substantially reduce the computational effort, by focussing on essential dynamical bottlenecks of the process (DOI: 10.1039/c0cp02852e).
The papers in this issue are a showcase of current developments, and clearly demonstrate the wave of innovations that is sweeping through the multiscale modelling field. While the topics are rather diverse, the aim is always the same: to accurately and efficiently predict dynamical and structural properties of complex processes and systems, and thus provide a better insight thereof.
The future is hard to predict. Yet, it is almost certain that in the coming decades new multiscale modelling methods, comprising quantum chemistry, classical MD, static and dynamic coarse graining, multi-level and hybrid simulations in all their forms will continue to be developed, allowing researchers in the field of physical chemistry and chemical physics to access ever more complex systems on larger length and time scales, without loosing essential microscopic features.
This journal is © the Owner Societies 2011 |