Antonino Polimeno

At present, many software and hardware solutions to perform long molecular dynamics simulations and microsecond-long trajectories are available. Even such long trajectories still show problems of convergence with respect to observables evolving with time scales of several tens of nanoseconds and the problem of conformational sampling, which is decisive to get correct statistics, have not been uniquely solved even if a number of methodologies has been proposed, such as umbrella sampling, metadynamics, adaptive biasing force and replica exchange techniques.

We explore alternatives routes based on stochastic coarse-grained approaches, adopting the following basic hypotheses: i) a purely torsional internal dynamics, coupled to global reorientation ii) a time scale separation among subgroups of torsional angles (rigid vs. flexible domains), iii) an Fokker-Planck model with explicit roto-conformational coupling.

Formal and computational solutions based on time-separation methods are defined and tested and included in a developing open access new software package (SALEM: Stochastic Advanced Liouville Equation Model).

Nature-inspired heuristic algorithms are fundamental tools for machine learning and cognitive computing. Chemical processes optimisation is a novel approach of chemical reactions, in which biology or physics-inspired algorithms can be employed to devise optimized pathways, geometries and mechanisms.

For instance, in order to simulate the solvent motion and each involved chemical substance in microfluidic reactor, one can perform numerical simulations with continuum models. Numerical optimization methods, e.g. ANNs or genetic algorithms, can be employed to modify microchannels paths and ameliorate the outflow of a desired product generated in a specific reaction. One can explore explore diverse kinetic schemes in order to test the whole methodology and to characterize the rate of convergence of optmization algorithm.

At a more fundamental level, quantum methods (QM) can be applied, coupled to heuristic approaches, to chemical reactions involving bond breaking and re-forming. The combination of standard QM software and advanced and ad hoc developed search algorithms can be tailored to explore structural and dynamic properties of sets of related molecular substrates, devising optimized molecular libraries with desired properties.